Air Weather ServiceTechnical Report 240

Forecaster’s Guide toTropical MeteorologySecond Edition, 1991

by Maj. Gary D. Atkinson, USAFrevised by C.S. Ramage

* DOES NOT CONTAIN GRAPHICS *

1 April 1971 / Revised August 1991 (Second Draft)

PDF version published by Weather Graphicsweathergraphics.comNo copyright claimedNo restriction on distribution

PREFACE TO SECOND EDITION

As correctly forecast in the Preface to the first edition, Forecasters’ Guide to Tropical Meteorol-ogy was used not only to train Air Weather Service meteorologists, but also as a text foruniversity courses. It was even translated into Chinese. Technological advances in the lasttwenty years, especially in satellite-based observations, have not seriously affected viewsexpressed in the first edition, but have allowed them to be expanded and refined. Bothclimatology and synoptic meteorology have benefited, while many hitherto obscure partsof the tropics have been exposed to the global eye.

In this edition, satellite data and wide-ranging research have allowed the tropics to betreated more evenly and extensively than before. Sections dealing with Africa, the Americasand south and southwest Asia have been expanded. Satellite pictures illustrate tropicalsystems and processes.

Besides these changes, the following topics have been either added to (bold face) or re-moved from (italics) the first edition.

Chapter 1. Terminology; heat lows.

Chapter 2. Stress-differential along a coast.

Chapter 3. Constant-level balloons; surface reports from buoys; Doppler radar; Profiler.

Chapter 4. Low-level jet streams.

Chapter 5. Upwelling; tradewinds; low-latitude westerlies.

Chapter 6. Fog; character of significant rain in the tropics; interannual variation of rainfalland El Niño; seasonal distribution of rainfall and the monsoons; pentad distribution ofrainfall.

Chapter 7. A new chapter on diurnal variation, combining the scattered discussions of thefirst edition, and expanding and adding to them. New sections on vertical mixing, stratocu-mulus and fog, and squall lines.

Chapter 8 (formerly Chapter 7). Equatorial waves; monsoon depressions; squall lines;surges; troughs in the upper-tropospheric westerlies; South Pacific convergence zone;superposition of tropical and extra-tropical disturbances; near-equatorial convergencezones

Chapter 10 (formerly Chapter 9). Duststorms.

Chapter 11 (formerly Chapter 10). Isogon analysis; kinematic frontal analysis.

Chapter 12 (formerly Chapter 11). Stability indexes (reduced); preparing to forecast in anew area; 30-60 day cycles; Atlantic tropical storms (seasonal forecasts); preparing to fore-cast in a new area.

FIGURES: 137 have been retained and 171 added.

REFERENCES: 177 have been retained and 367 added.

The range of topics and locations cannot be simply dealt with sequentially; to avoid repeti-tion, extensive cross-referencing has been used. Detailed subject, geographical and authorindexes have been added, as well as a fold-out locator map.

ACKNOWLEDGMENTS. Many present and former members of Air Weather Service con-tributed to the preparation of this edition, including the sixteen who carefully and con-structively reviewed the first and second drafts, and those who provided weather satellitepictures, bibliographies and publications. I particularly appreciated Garry Atkinson’scomments. Louis Oda drafted and revised the figures, Thomas Schroeder, was responsiblefor administration. Others who helped included Mike Gentry, William Gray, Barry Hinton,Richard Orville, Patrick Sham, Norman Wagner, J.F. White, Scott Woodruff, and X. Ziang.My wife Alice proofread.

C.S. Ramage1991

TTTTTABLE OF CONTENTSABLE OF CONTENTSABLE OF CONTENTSABLE OF CONTENTSABLE OF CONTENTS

CHAPTER 1 - INTRODUCTION

CHAPTER 2 - PRIMARY PHYSICAL CONTROLS OF THE TROPICAL CIRCULATION

CHAPTER 3 - THE OBSERVATIONAL BASIS

CHAPTER 4 - PRESSURE AND WINDS

CHAPTER 5 - TEMPERATURE AND WATER VAPOR

CHAPTER 6 - CLOUDINESS AND RAINFALL

CHAPTER 7 - DIURNAL VARIATIONS

CHAPTER 8 - TROPICAL SYNOPTIC MODELS

CHAPTER 9 - TROPICAL CYCLONES

CHAPTER 10 - SEVERE WEATHER IN THE TROPICS

CHAPTER 11 - TROPICAL ANALYSIS

CHAPTER 12 - TROPICAL FORECASTING

A. GeneralA. GeneralA. GeneralA. GeneralA. General

The urge to understand tropical meteorol-ogy has been driven by a variety of forces:the devastation caused by tropical cyclones;demands by military operations, e.g. WorldWar II, the Marshall Islands weapons tests,and the conflict in southeast Asia; and thebelief of influential mid-latitude meteorolo-gists that the tropics may hold the key tosuccessful forecasts of global weather pat-terns and climate change. Efforts by WorldWeather Watch (WWW) and the GlobalAtmospheric Research Program (GARP),including the GARP Atlantic Tropical Ex-periment (GATE) of 1974 and the Winterand Summer Monsoon Experiments(WMONEX and SMONEX) of December1978 and June-July 1979, gave many meteo-rologists their first experience of the tropics.Research papers proliferated, concentratingon the small time and space scales describedby the special observing systems (aerialreconnaissance, rawinsondes, and radars)set up for the experiments.

But where does all this leave the field fore-caster? A case study dependent on data notoperationally available, and supported bybrief, doubtful statistics, doesn’t help muchin the day-to-day grind of tropical analysisand forecasting.

Predicting tropical cyclones taps almost allthe resources available to aid the tropicalforecaster. Riehl (1979) devotes a chapter tonumerical hurricane prediction and nothingto other tropical forecasting. On an averageday in the tropics less than one tenth of onepercent of the area is affected by a tropicalcyclone. Only about 80 cyclones develop

each year, but since 1950 more than 5000operational and research reconnaissanceflights have been made into cyclone centers.By contrast, one Arabian Sea mid-tropo-spheric cyclone, one monsoon depression,and two upper-tropospheric cold lows havebeen reconnoitered. Fifteen single-aircraftresearch traverses have been made acrossthe near-equatorial convergence zone, butnone into the mysterious weak circulationsthat can cause disastrous floods. For mostgenerating regions, about one hundredyears of daily tropical cyclone positionsprovide the basis of detailed, authoritativeclimatologies. Apart from monsoon depres-sions, not even a year of tracks has beenplotted for other tropical vortexes.

This disproportion, though slightly reducedby squall-line studies in GATE, Venezuelaand the Amazon, and a 1987 field study ofearly-summer rain systems off southeastChina (Taiwan Area Mesoscale Experiment—TAMEX), means that the meteorologisttrying to forecast tropical cyclones is servedby a variety of subjective, numerical andstatistical techniques, by special servicesfrom global prediction centers, and by 15-minute satellite updates. For the other 99.9percent of the time, forecasters are prettymuch on their own. This manual tries toredress the balance. It devotes one chapterto tropical cyclones, while emphasizingother weather systems, backgroundclimatologies, and forecasting hints andpitfalls.

Observational information on the tropicsincreased significantly through the 1970s,though progress has been patchy overAfrica and the Americas. Weather radarsoperate at many tropical stations, and there

Chapter 1Chapter 1Chapter 1Chapter 1Chapter 1INTRODUCTIONINTRODUCTIONINTRODUCTIONINTRODUCTIONINTRODUCTION

is great potential for more aircraft weatherreports. Compared to twenty years ago,human observations are now fewer; auto-mated observations, especially those fromsatellites, are vastly more, and are not yetfully exploited.

There is an urgent need to synthesize fromthese diverse observational systems acoherent basis for the analysis and predic-tion of tropical circulations and associatedweather. Our ability to digest and use thisinformation in tropical numerical modelshas not kept pace with numerical weatherprediction in middle latitudes. This is par-tially due to the greater complexity ofweather systems and interactions betweenvarious scales of motion in the tropics. Thus,although world-wide predictions are beingmade, at least in the near future, tropicalweather forecasting will continue to dependheavily on subjective judgment aided byselected numerical products such as objec-tive wind analyses, global long-wave predic-tions and computer-processed satellite data.

West Pacific typhoons and Atlantic hurri-canes are the same phenomenon — theyneed not be separately described. Thunder-storms also possess the same characteristicsworld-wide. What is not so well-known isthat some other phenomena occur through-out the tropics. These include surges, coldlows, linear convergence lines and shearlines, squall lines, low-level jets, heat lows,upwelling and dust storms. Each can beexplained once, then all forecasters need toknow is its distribution and frequency.Although this manual discusses varioustheoretical hypotheses, it uses little math-ematical formulation; it assumes that thereader is familiar with the pressure/windand the thermal wind relationships, diver-gence, convergence and vertical motion,vorticity and lapse rate, and generally un-derstands mid-latitude synoptic models.The comprehensive list of references servesto facilitate further study on selected aspects

of tropical meteorology in much greaterdetail than is possible here.

The forecaster is viewed as an appliedresearcher who can evaluate hypotheses andcome up with new ideas. He should realizethat in the tropics weather mechanisms areoften not understood. “Authoritative” opin-ions may differ, for example, on the preva-lence of easterly waves, on the cause of thequasi-biennial oscillation, on predicting ElNiño, and on generation, movement anddissipation of surges. Even when there isgeneral agreement, such as on tropicalcyclone dissipation, diurnal variations alonga mountainous coast, and the role of surfacecooling in causing fog, apparent exceptionsmay occur. Chapters 2 through 6, besidesincluding a chapter (3) on data, focus on theannual cycle and monthly averages, whilemaking some attempt to interpret them interms of weather. Chapter 7 covers diurnaleffects. The remaining chapters concernweather analysis and forecasting. Hereastronomical control by the annual anddiurnal cycles is muted by complex feed-backs among atmosphere, ocean, and ter-rain.

B. TB. TB. TB. TB. Terminologyerminologyerminologyerminologyerminology.....

Definitions listed in the Glossary of Meteo-rology (Huschke, 1958) have been used. Thissection expands some of these definitions,adds new terms introduced to the tropicssince 1958, and discusses terms that havebeen made misleading by a variety of defini-tions. Terms used only once in the text (e.g.hodograph) are defined where they appear.

!!!!! BUFFER ZONE (Figure 1-1). This wasdefined by Conover and Sadler (1960). Abuffer zone straddles the equator and sepa-rates two oppositely-directed wind streams,the easterly trades and the monsoon wester-lies. The sense of rotation in the

buffer zone is clockwise in the NorthernHemisphere (NH) summer andanticlockwise in the NH winter. The zonetends to persist and may contain circulationcells. These are generally stationary, andmay last 3 to 4 days. If it moves more than afew degrees from the equator it takes on thecharacter of either a trough, if the curvatureis cyclonic, or a ridge, if the curvature isanticyclonic. The term has not been widelyaccepted, and has been criticized by Leroux(1983) “because there is generally no inter-ruption of the stream”.

!!!!! COLD LOW (upper troposphere). Incomparison to its environment, this low ismost intense and coldest in the upper tropo-sphere. Thus it is weaker at lower levels andcan be hard to find below 500 mb. Apartfrom a few cumulonimbus that may formnear the center, it is generally subsident andcloud free, except for rising motion andcloud on its eastern periphery. In summerover the tropical Atlantic and PacificOceans, and the South Indian Ocean, a lineof cold lows may comprise the TROPICALUPPER TROPOSPHERIC TROUGH (TUTT).See Figures 8-20 to 8-24.

!!!!! ELECTROMAGNETIC RADIATION.All the instruments on meteorologicalsatellites and all electrooptical devices aretriggered by electromagnetic waves emittedfrom, or reflected by objects or gases in theatmosphere. The waves are usually specifiedin terms of wavelength (e.g. 10 cm radar) orfrequency (e.g. 10µm) or more descriptively— visible, infrared. Figure 1-2 illustrates therange of electromagnetic waves of interest tothe meteorologist. The following units areused:

Frequency: 1 Hertz = 1 HZ = 1 cycle sec-1

1 Gigahertz = 1 GHz = 109HZ

Wavelength: 1 micrometer = 1µm = 10-6m

(Frequency in cycles per second) = (speed oflight = 2.997 x 1010 cm s-1) divided by thewavelength in cm.

For example, a 10 cm radar operates at afrequency of 2.977 GHz.

!!!!! HEAT LOW/HEAT TROUGH. Surfaceheat lows, occasionally strung along atrough, as over the deserts of North Africaand southwest Asia in summer, and oversouthern South America, southern Africaand along the Great Rift of Africa through-out the year (Figure 4-1), form above sur-faces made relatively hot by insolationthrough clear skies. Heating of the overlyingair causes its pressure to fall to a relativeminimum and a depression to form. Airconverging into the heat low rises, but astrong subsidence inversion, reflectingconvergence in the upper troposphere,allows only shallow cloud to form; this isquickly evaporated through mixing withvery dry air from above the inversion. Axesof heat troughs have been termed the heatequator. In the family of tropical vortexes(Figure 6-32), heat lows/troughs are linkedto surface anticyclones and upper tropo-spheric (cold) lows. In them, mid-tropo-spheric air sinks and weather is fine. Overthe deserts, as winter follows summer, andsurface air cools and becomes denser, sur-face pressure increases and anticyclonesreplace heat lows. On the eastern and west-ern edges of heat troughs, moister air andsome cloud reduce insolation, and so thesurface pressure minimum, especially overthe oceans, is less marked. However, thetroughs are still heat troughs, with thelowest pressure coinciding with the highestsurface temperature. That in turn, is ensuredby a relative cloud minimum and lightvariable winds (oceanic doldrums) along thetrough axis. The term MONSOONTROUGH has been incorrectly applied toall persistent near-equatorial troughs withwesterlies on the equatorward side andeasterlies on the poleward side of the trough

(Figure 1-1). Although this usage is too well-established to be readily changed, the readershould remember that in this report, “mon-soon” troughs may be year-long, (as overSouth America and southern Africa), andnot just summer features (as over NorthAfrica). Figure 1-3 presents a schematicmodel of the rainfall, pressure, and low-level circulation associated with the easternNorth Pacific “monsoon” trough. The me-ridional section, based on satellite observa-tions, reveals that maximum cloudiness andrainfall do not occur along the trough linebut in the westerlies south of the trough (seeFigure 6-16). A slight secondary rainfallmaximum is evident just north of thetrough. Along the trough line clouds arescattered or even absent. This confirms thatthe trough is really a heat trough, wherelight, variable winds and less cloud allowthe sun to maintain a relatively warm seasurface and relatively low air density. Simi-lar distributions have been observed overcentral India (Figure 6-15), in the southIndian Ocean (Ramage, 1971, 1974a) andacross West Malaysia (Lim, 1979).

!!!!! INTERTROPICAL CONVERGENCEZONE (ITCZ). Although this term is stillwidely used, its meaning has been hope-lessly corrupted. Besides “ITCZ”, the exten-sive low latitude surface pressure troughhas been called “equatorial trough”, “inter-tropical front”, “heat trough”, and “mon-soon trough”. In turn, as shown in Figure 1-4, ITCZ has been variously defined as“trough axis”, “axis of maximum cloudi-ness”, “axis of surface wind convergence”,and a fourth line that coincides with none ofthe others (Sadler, 1975a). Finally, the se-mantic impossibility of two ITCZs along thesame meridian has also been perpetrated.This manual replaces ITCZ by more specificterms — “heat trough”, “monsoon trough”(already somewhat contaminated), and“near-equatorial convergence zone”(NECZ).

!!!!! MARITIME CONTINENT. This regionincludes Indonesia, Malaysia, Mindanao,the western Caroline Islands, Papua NewGuinea, and the Solomon Islands. Alongwith equatorial Africa and equatorial SouthAmerica it is the source of most tropicalthunderstorms (see Figures 10-1,2,3). Thelarge, high islands generate continental-likeconvection, while the warm surroundingocean is an inexhaustible moisture source.The area is outlined on the locator chart.

!!!!! MID-TROPOSPHERIC (SUBTROPI-CAL) CYCLONE. Both systems originate inbaroclinic environments and are probablysimilar. They are most intense in the middletroposphere, where horizontal temperaturegradients are least. They weaken both up-ward and downward from this level, sincerelative to the environment, they arewarmer in the upper and cooler in the lowertroposphere. Although a vertical eye doesnot develop, these cyclones may be mis-taken for tropical cyclones in a weathersatellite picture. Extensive nimbostratuswith embedded cumulonimbus gives some-times heavy rain. They are most commonover the subtropical North Pacific and NorthAtlantic, and in summer over the northernArabian Sea. See Figures 8-7 through 8-18.

!!!!! MONSOON TROUGH. See HEAT LOW/HEAT TROUGH.

!!!!! NEAR-EQUATORIAL CONVERGENCEZONE. Throughout the year, surface pres-sure gradients are directed from east to westin near-equatorial parts of the central Pacificand the eastern Atlantic/South America.Coupled with small equator-ward directedmeridional pressure gradients, this ensuresthat the northeast tradewinds of the NHconverge with the southeast tradewinds ofthe SH in a persistent zone of unsettledweather — the near-equatorial convergencezone. The NECZ shifts between about 5°Nin April and 10°N in September.

Between February and May, a secondaryNECZ sometimes forms near 5°S over theeastern Pacific. See Figures 8-1, 8-55 and 8-56.

!!!!! QUASI-BIENNIAL OSCILLATION(QBO). Winds in the tropical stratosphereundergo a downward-propagating oscilla-tion between easterlies and westerlies, on anaverage every 26 months (quasi-biennial).The oscillation maximum occurs at 75,000 to85,000 feet (23-26 km) above the equator anddecreases downward and poleward. Thecause is still debated. See Figure 4-12.

!!!!! RAINS/SHOWERS. These terms areitalicized in the manual, where they aregiven the following meanings. Rains fallfrom deep nimbostratus with embeddedcumulonimbus, when wind shear in thevertical and surface convergence are bothlarge. Rains vary in intensity, but since skiesstay mainly overcast, heaviest falls are mostlikely during the night and around dawn.Showers fall from scattered cumulus orcumulonimbus, when wind shear in thevertical and lower tropospheric convergenceboth are small. Diurnally-varying localwinds, which readily develop during ashowers regime, in turn determine thediurnal variation of the showers.

!!!!! SURGE. Within persistent surface windregimes, such as the tradewinds and themonsoons, flowing toward the heat equator,greater speeds may suddenly appear (surge)and last for a few days. Downwind of thespeed maximum of the surge, convergenceworsens weather, and the associated verticalcirculation accelerates. Above this, in theupper troposphere, flow in the oppositedirection from the surface winds is diver-gent and exports heat to colder latitudes.Surges have been linked to the developmentof tropical cyclones and monsoon depres-sions, squall lines, equatorial convection,and onset of El Niño. Their causes seem tobe varied, and not well-observed or under-stood. See Figure 8-28.

!!!!! RESULTANT WIND. For climatologicalpurposes, scalar quantities can be simplyaveraged; their standard deviations indicatevariability. On the other hand, wind must beaveraged vectorially, by first resolving eachwind observation into components fromnorth and east, summing over the climato-logical period, obtaining the averages, andfinally reconverting the average north andeast components into a single vector — theresultant wind. The steadiness or persis-tence of the wind is defined as the ratio ofthe magnitude of the resultant wind to theaverage speed of the wind without regard todirection. It is usually expressed as a per-centage. A resultant of winds all with thesame direction would have a steadiness of100 percent.

!!!!! TRADEWIND TROUGH (Figure 1-1). Inthe first edition of this report, this wasdefined as the pressure trough along whichthe tradewinds of each hemisphere con-verge. Research aircraft traverses haveshown that the trough is very broad andweak, and that the convergence, which ismuch narrower and well-defined, may notcoincide with the lowest pressure. Thisedition replaces “tradewind trough” by themore specific “near-equatorial convergencezone” (NECZ).

!!!!! TROPICAL UPPER TROPOSPHERICTROUGH (TUTT). See COLD LOW.

!!!!! UNITS OF MEASUREMENT. The WorldMeteorological Organization (WMO) hasestablished a set of units, based on thecentimeter-gram-second system, in which toexpress meteorological measurements.Ideally, this manual should conform, butregrettably in the United States, surface airtemperature is still measured and reportedin degrees Fahrenheit (°F) instead of degreesCelsius (°C), and rainfall in inches instead ofmillimeters (mm) or centimeters (cm). Also,in aviation operations height is in feet in-stead of meters (m) or kilometers (km),distance is in nautical miles instead of kilo-

meters (km) and speed is in knots instead ofmeters per second (m s-1). There has beensome progress. In the United States, upperair temperatures and dew points are ex-pressed in °C, potential temperatures in °Kand mixing ratios in gm Kg-1. Whenever USunits appear in the text, they are followedby the approximate metric equivalents set inbrackets, except in the case of speed, wherethe transfer between knots and m s-1 can bereadily approximated by applying a factorof one-half. When convenient, figures in-clude both sets of units.

!!!!! WALKER CIRCULATION. Bjerknes(1969) identified a weak equatorial zonalcirculation in which surface flow from eastwas lifted and gave rain over Indonesia,returned from west in the upper tropo-sphere and sank over the central Pacific.Others have postulated similar circulationsassociated with the equatorial African andSouth American thunderstorms. TheWalker Circulation is much weaker than theHadley circulation and hard to track fromday to day. See Figure 2-7.

Chapter 2Chapter 2Chapter 2Chapter 2Chapter 2PRIMARY PHYSICAL CONTROLSOF THE TROPICAL CIRCULATION

A. General.A. General.A. General.A. General.A. General.The tropical circulation and resultingweather patterns are determined by a num-ber of complex and related physical con-trols. The more observational data acquired,the more the complexity and variability ofthe tropical atmosphere become evident.These characteristics are vividly displayedby the global coverage of meteorologicalsatellite data. In 1952, Palmer predicted thisincrease in apparent complexity with in-creasing data. Referring to new discoveriesin tropical meteorology, often made possiblein the past by observational requirements ofwars or atomic weapons tests, he stated,“The new regions thus opened up for explo-ration only vaguely resemble those precon-ceived by the theoreticians. It is not onlythat the griffins and basilisks described bythe philosophers are absent, it seems thatthe country is occupied by creatures ofwhich they have never dreamt.” Since 1952,much progress and many exciting discover-ies have been made in tropical meteorology,such as the biennial oscillation in the tropi-cal stratosphere, the detailed structure andinfluence of oceanic upper-tropospherictroughs revealed by jet-aircraft and satelliteobservations, the preparation of detailedclimatologies of the tropical wind field,cloudiness and rainfall, and studies of thetropical general circulation and energybalance, including the role of the monsoonsand air-sea interaction during El Niño.

This chapter reviews the major physicalfactors controlling tropical climate. Agrossly simplified description of the earth’sheat-energy budget is presented to help inunderstanding the major driving forces of

atmospheric circulations. This is followedby views of the tropical general circulation.Finally, factors controlling the tropicalcirculation and weather patterns are dis-cussed. Subsequent chapters cover them inmore detail. Lack of observations or physi-cal understanding may prevent other influ-ences from being specified.

B. The EarthB. The EarthB. The EarthB. The EarthB. The Earth’s Heat’s Heat’s Heat’s Heat’s Heat-Energy Balance.-Energy Balance.-Energy Balance.-Energy Balance.-Energy Balance.According to Sellers (1965), net radiation atthe earth’s surface, (R) is given by:

R = (Q + q) (1 -a) - I

where (Q + q) is the sum of the direct anddiffuse solar radiation incident on theearth’s surface (“insolation”),

a is the surface albedo (= 1 for a perfectlyreflecting surface; = 0 for a perfectly absorb-ing surface), and

I is the effective outgoing radiation from thesurface.

Over an average year, the terms balanceglobally. The earth absorbs solar radiation ata rate of about 164 watts m-2. In turn, itradiates at 69 watts m-2, leaving a net sur-plus of 95 watts m-2, i.e. 95 = 164 - 69.

The atmosphere also participates in theexchange. The atmospheric net radiation(Ra) is given by:

Ra = (Qa + qa) (1 - aa) - Ia

where (Qa + qa) is the total radiation incidenton the atmosphere, aa is the atmosphericalbedo, and Ia is the effective outgoingradiation from the atmosphere.

For the earth/atmosphere system to remainin long-term radiation balance, R = -Ra, andthe components of the atmospheric equa-tion are:

-95 = (60 - 155) watts m-2

Thus, there is a net balance between thegain at the earth’s surface and the loss of theatmosphere. Although the annual globalaverage radiation balance is zero, it willgenerally not balance either seasonally orannually in a given latitude zone. A sche-matic view of the radiation balance is givenin Figure 2-1. The atmosphere is uniformlya radiation sink at all latitudes, while theearth’s surface, except near the poles, is aheat source. The sum of the two, shown bythe solid line in Figure 2-1, determines theradiation balance of the earth-atmospheresystem. To keep the earth from warmingand the atmosphere from cooling, theremust be energy transfer from the surface tothe atmosphere. This vertical exchangeoccurs mainly by evaporation of water fromthe surface and condensation in the atmo-sphere and by conduction of sensible heatfrom the surface and turbulent diffusion(convection) into the atmosphere. In Figure2-1, since the horizontal axis scale is propor-tional to the earth surface area in eachlatitude band, the area of surplus radiationequals the area of deficit radiation.

The net radiation is positive equatorward of40° latitude and negative poleward of 40°.In order to keep the poles from gettingcolder and the tropics warmer, energy mustbe transported from the tropics to higherlatitudes. This horizontal heat exchange iscarried out by (1) the poleward transfer ofsensible heat in the atmosphere, (2) latentheat of condensation subsequently releasedand carried poleward in the atmosphereand (3) oceanic circulations. Over the tropi-cal oceans, the latent heat transfer is anorder of magnitude larger than the sensibleheat transfer. And so Malkus (1962) con-cluded that the atmosphere is fueled mainly

from below, with 80 percent (global aver-age) of its heat energy initially latent in theform of water vapor. Considerably morethan half of this latent heat energy is sup-plied to the atmosphere by the tropicaloceans between 30°N and 30°S.

Figure 2-2 depicts the annual water balancein the earth-atmosphere system. Polewardof about 40° and within about 10° of theequator, precipitation exceeds evaporation,while the reverse is true in the subtropicallatitudes (approximately 10° to 40°). On along-term basis, regions with excess evapo-ration must export water vapor whileregions with excess precipitation mustimport water vapor. Overall, the oceans losemore water by evaporation than they gainby precipitation; the deficit is made up byrunoff from the land areas where precipita-tion exceeds evaporation. As shown inFigure 2-2, evaporation is greatest in thesubtropics, mostly occurring in relativelydry tradewinds. Riehl and Malkus (1957)showed that regions of oceanic tradewindsprovide a significant proportion of theenergy in the form of latent and sensibleheat needed to drive the atmospheric circu-lation. This energy, primarily in the form ofwater vapor, is initially transportedequatorward by the relatively steadytradewinds near the earth’s surface. In theequatorial zone, water vapor is lifted andcondensed in large, convective cloud sys-tems producing sensible heat and potentialenergy which are then moved to higherlatitudes by the upper tropospheric circula-tion. In another study of the heat balance inthe equatorial zone, Riehl and Malkus(1958) estimated that only a very smallfraction of a 10° latitude belt near the equa-tor needs to be occupied by giant cumulon-imbus clouds or “hot towers” to maintainthe heat budget of the equatorial zone andprovide for much of its poleward energytransport. They suggested the followingdescending hierarchy of the fractional area

occupied by the various scales of phenom-ena in the equatorial zone:

Area of near-equatorial trough zone= A

Area occupied by synoptic disturbances= 0.1A

Area occupied by active rain= 0.01A

Area occupied by undilute towers= 0.001A

Thus, they estimated that at any one timeabout 30 synoptic disturbances containing atotal of about 1500 to 5000 giant cumulon-imbus is sufficient for heat balance require-ments. As mentioned previously, a latitudi-nal heat-energy balance is achieved byatmospheric and oceanic circulations.Assuming no net heat flux across the poles,Figure 2-3 shows the net annual transportof various components of the heat-energybudget across each latitudinal circle. Northof about 5°N the sensible heat-energy(sensible heat plus potential energy) trans-port by atmospheric and oceanic currents isnorthward, and south of 5°N it is south-ward. While atmospheric sensible heat-energy transport shows double maxima ineach hemisphere, the oceanic sensible heat-energy transport shows a single maximumin the subtropics of each hemisphere. Thetransport of latent heat energy by atmo-spheric circulations is more complex.Poleward of 20-25° in each hemisphere,there is net transport of water vapor

poleward, while equatorward of theselatitudes water vapor is transported towardthe near-equatorial convergence zone(NECZ), located on average near 5°N. Thetotal poleward heat-energy transport re-quired to balance the net radiation deficit inhigher latitudes combines the three compo-nents discussed above, as shown at thebottom of Figure 2-3; it is greatest near40°N and 40°S. Overall, there is a net en-ergy transport from the tropics by atmo-spheric and oceanic circulations, and a netimport to higher latitudes.

Vonder Haar and Oort (1973) used satelliteradiation measurements to estimate totalpoleward heat transport. They differed littlefrom Sellers (1965), whose estimates areshown in Figure 2-3. But because the tropi-cal oceans absorb more radiant energy(lower albedo) than previously thought,ocean currents transport a larger proportionof sensible heat across the subtropics (Table2-1). Over the globe, about 40 percent of thetotal required meridional heat transportmay be accounted for by ocean currents.Hastenrath (1985) discusses heat and waterbudgets in detail.

Besides meridional exchanges of heat,regional exchanges also occur. Alexanderand Schubert (1990) applied the method ofVonder Haar and Oort to data from FGGE,and found that in winter the mid-latitudeoceans export heat to the continents. Theflow is reversed in summer. Not surpris-

———————————————————————————————Table 2-1. Poleward transport of heat energy by ocean currents according to Sellers (1965)(S) and Vonder Haar and Oort (1973) (VO) and percentages of the total (air plus ocean)fluxes (units of 1013 W).

North SouthLat. 90 80 70 60 50 40 30 20 10 0 10(S) 0 0 12 34 75 104 152 157 107 -21 -132% 0 0 7 11 17 20 31 47 66 70 60(VO) 0 -7 5 5 113 206 244 340 187 26 -13% 0 2 4 2 28 41 48 74 66 61 6

ingly, the exchanges are much larger in theNorthern than in the Southern Hemisphere.

C. TC. TC. TC. TC. Tropical General Circulation.ropical General Circulation.ropical General Circulation.ropical General Circulation.ropical General Circulation.The expression “general circulation of theatmosphere” refers to a statistical descrip-tion of the mean large-scale atmosphericmotions over the whole earth. These statis-tics are derived from many daily flowpatterns and include not only mean condi-tions but also the variability of the flowresulting from seasonal changes and fromthe effects of transient cyclones and anticy-clones (Huschke, 1959). Many generalcirculation studies have been made, espe-cially since the worldwide expansion ofupper-air networks during and after WorldWar II. Lorenz (1967) made an excellentsurvey. A group of Massachusetts Instituteof Technology (MIT) investigators (Newellet al., 1972) concentrated on the generalcirculation of the tropics. They processedupper-air data for about 300 stations be-tween 45°N and 30°S for the period July1957 to December 1964. The troposphericresults were obtained from objective analy-ses at various pressure levels of the long-term station means and the stratosphericresults from latitude-band means givingequal weight to data from odd and evennumbered years to eliminate the effects ofthe quasi-biennial stratospheric wind oscil-lation (see Chapter 4). Mean meridionalcross-sections were computed from a linearcombination of station values for 10° lati-tude bands using a weighting scheme toreduce the bias caused by more observa-tions over the continents. These cross-sections present extremely simplified viewsof the tropical general circulation becauseof very large E-W departures from themeans. For example, during July along10°N there are regions of very strong east-erly and westerly resultant winds near thesurface; but the resultant wind for all me-ridians is very light. Figure 2-4 shows the

mean vertical motion at 500 mb for eachseason (Kyle, 1970). In general, motion isupward near the equator and downward inthe subtropics. But mean vertical motionvaries greatly along latitude circles. Overthe continents, annual variation is large andassociated with the monsoons. Otherwise,centers of upward motion lie over thecontinents and centers of downward mo-tion over the oceans and the deserts. Re-member too, that time-averaging greatlyreduces the intensity of vertical motion.Largest values in Figure 2-4 are less than 20x 10-4 mb s-1, or 30 mm s-1, two orders ofmagnitude less than vertical motions insidecumulonimbus. Where climatological dataare sparse, these patterns may have to berevised; for example, the satellite cloudclimatologies of Figures 6-1 to 6-4 showcloud along the North Pacific NECZ, whereFigure 2-4 depicts sinking motion.

Figure 2-5 shows the mean zonal windcross-sections for each season from the MITstudies. The major features are the regionsof westerly flow above the near-surfacelayer in the subtropics reaching maxima inthe subtropical jet streams near 200 mb, andthe deep easterly flow in low latitudes.

At the surface, the subtropical ridges,separating easterlies from westerlies, shiftabout 5° poleward between winter andsummer. On average, the ridges are locatedat about 30°N and 30°S, slightlyequatorward of the boundary between heatsurplus and heat deficit (Figure 2-1). Near200 mb the ridges are 30° apart in June-August and only 15° apart in December-February.

By determining the mean meridional mo-tion at various levels, averaging themaround latitude circles, and integratingwith respect to pressure between 1000 mband 100 mb (assuming zero vertical motionat the upper and lower boundaries), themean vertical motion field can be deter-mined from the continuity equation. Figure

2-6 depicts the mean meridional circula-tions and resultant vertical motion fieldsbetween 40°N and 40°S in the form of mass-flow streamlines. The spacing of thestreamlines is inversely proportional to thestrength of the flow; total transport in eachcell is given by the labeled values in thecenter. Mass continuity demands that overthe globe the streamlines must be closed.Cellular patterns are evident, with risingmotion near the equator and sinking mo-tion in the subtropics. Hadley (1735) postu-lated these cellular circulations to accountfor the tradewinds. The mean meridionalcirculation is strongest in the winter hemi-sphere, when upward motions of theHadley cells extend across the equator intothe opposite hemisphere. The Hadley cellsin large part maintain the strong subtropi-cal jet streams of the winter hemispherethrough poleward momentum transport inthe upper troposphere; and they are re-sponsible for equatorward latent-heattransport in the lower troposphere, asdescribed previously. Hadley cells are“direct” circulations — warmer air risingand cooler air sinking. Their primary en-ergy source is the latent heat of condensa-tion released in the rising branches of thecells. Short-period fluctuations, whichwhen averaged comprise a Hadley cell, arediscussed in Chapter 8, Section D-4.

Mean meridional temperature cross-sec-tions produced by the MIT investigators,show nearly uniform temperature in thetropical middle troposphere; between 15°Nand 15°S the annual range in mean tem-perature is only about 1°C. The tropo-spheric temperature maximum (heat equa-tor), though not pronounced, lies near 5°Sduring the southern summer and near 20-25°N during the northern summer. Thisasymmetry results from the stronger sur-face heating over the Northern Hemisphere,with its larger continental areas, and thecooling effect of Antarctica.

DDDDD. P. P. P. P. Primary Physical Frimary Physical Frimary Physical Frimary Physical Frimary Physical Factors.actors.actors.actors.actors.

1. General. The preceding sections dis-cussed the role of the major energy conver-sion processes (i.e., radiation, precipitation,evaporation, etc.) in the tropical large-scalecirculation. This section covers the moreimportant physical factors controlling ormodifying tropical circulation and weatherpatterns.

2. Distribution of Land and Water. Thedistribution of land and water affects allscales of motion in the tropics. The largeland masses of Africa, Asia and Australiaexperience a much larger annual cycle ofradiational heating and cooling than do theneighboring oceans, where specific heat ishigher. Consequently, pronounced seasonal(monsoonal) wind regimes dominate muchof the Eastern Hemisphere tropics. Lesspronounced annual wind changes arerelated to the Western Hemisphere landmasses. The shifts in these seasonal windpatterns stem from shifts in the mean posi-tions of the low-level monsoon and heattroughs over and near the continents. Inturn the frequencies of tropical cyclones,which develop in the troughs, are affected.Over tropical oceans, away from continen-tal influences, the NECZ, where easterlytradewinds from both hemispheres meet,has only a small annual movement.

Although latent heat of evaporation ismainly added to the air over the ocean, therelease of latent heat of condensation isgreatest over tropical land. Ramage (1968)has shown that because of frequent thun-derstorms, the near-equatorial regions ofSouth America, Africa, and Indonesiagenerate much more heat for export tohigher latitudes than do the low-latitudeoceans. The longitudinal distribution ofthese concentrated heat sources produceslarge meridional temperature gradients and

results in the strong subtropical jet streamsshown in Figure 2-5. The heat sources alsocause weak vertical circulations along theequator. Bjerknes (1969) identified a cell inwhich air rose over Indonesia and sankover the western Pacific. He dubbed it“Walker Circulation”. Flohn (1971) ex-tended the concept around the globe, iden-tifying upward branches of the circulationcells over the continents and Indonesia andsinking branches over the oceans. But in theequatorial Pacific, where the Walker Circu-lation is presumably strongest, data barelysupport its existence.

At the surface, easterlies extend from100°W to 150°E, while upper troposphericwesterlies are not found west of 180°. Notsurprisingly, others have disagreed withFlohn; Newell’s (1979) more complicatedview (Figure 2-7) fits the upper tropo-spheric and surface wind systems, butleaves a gap between 170°E and 130°E,where large interannual variations, associ-ated with El Niño, occur. The Walker Circu-lation may be a useful concept and will berevisited in Chapter 6, but forecastersshould beware of putting it on a par withHadley cells.

The Tibetan plateau is a high-level radia-tional heat source throughout the year. Insummer, with condensation along theHimalayas powerfully aiding the effects ofradiation, this area plays a significant rolein developing and maintaining the hightropospheric anticyclone over Asia and theresulting easterly jet stream south of Asia.In turn the vertical motion patterns associ-ated with this jet contribute to the extremearidity over North Africa and Arabia andthe heavy rainfall over India and SoutheastAsia (see Chapter 6, Section C-5b).

On much smaller scales, the thermal gradi-ents associated with topography producelocal circulations such as land and seabreezes over islands and along coasts, landand lake (or river) breezes around inland

water bodies, and mountain and valleybreezes (see Chapter 7). Local topographyand distance from moisture sources largelydetermine rainfall in the tropics. Some ofthe heaviest mean annual or mean monthlyrains fall at stations on the windward sidesof mountains exposed to a prevailing flowoff a large body of water. The top of Mt.Waialeale, Kauai, Hawaii (22°N, 159°W),exposed to the tradewinds at 5075 feet (1547m) elevation, experiences an average of 449inches (11415 mm) a year. Conversely,stations to leeward of large mountains aredry, reflecting the rain shadow effect associ-ated with subsiding downslope winds.

a. Stress-Differential Along a Coast. Accord-ing to Bryson and Kuhn (1961), air flowingparallel to a coast is frictionally slowedmore by the land than by the sea. Thus,surface winds over land cross isobars to-ward lower pressure at a greater angle thanover the sea. Hence, if pressure is lowerover the land (as in all upwelling regions),winds diverge and air subsides at the coast.This is partly why vigorous upwellingborders coastal deserts (see Figure 5-4). Ifpressure is higher over the land, surfacewinds converge at the coast, a conditionoften accompanying continental anticy-clones. Figure 2-8 (Fett and Bohan, 1986)shows the processes.

3. Sea-Surface Temperature. Sea-Surfacetemperature (SST) significantly influencesatmospheric circulation and weather sys-tems. Because heat and moisture are ex-changed at the air-sea interface, sea surfacetemperature largely determines the airtemperature and moisture distribution inthe surface layer over the tropical oceans.Over warm oceans latent heat is released bylifting of the surface air in the convectiveclouds that may precede development ofwarm-cored tropical cyclones; some inten-sify into hurricanes. If hurricanes move intohigher latitudes, the lower sea-surface

temperature may deprive them of the latentheat source necessary to maintain theintense circulation. Hurricanes also weakenon moving over land. This weakeningresults more from a cutoff of their latentand sensible oceanic heat source than fromincreased surface friction (see Chapter 9,Section F).

Intense tropical cyclones significantlymodify the distribution of sea-surfacetemperature. Leipper (1967) studied theeffects of Hurricane Hilda (October 1964)in the Gulf of Mexico. Over an area 100 to300 km wide near the hurricane path, sea-surface temperature decreased more than5°C. The storm circulation had increasedmixing and upwelling of cooler subsurfacewater. The sea surface stayed cooler for atleast 3 weeks. More recent measurementshave confirmed this effect. Evaporation andconduction beneath early winter cold, drypolar outbreaks can also significantly lowersea surface temperature.

“El Niño”, occurring every few years in theequatorial Pacific, exemplifies the effect ofsea-surface temperature on weather.Coastal Peru is normally arid because coldwater upwelling near the coast reinforcesthe effect of strong atmospheric subsidence.During El Niño, however, when a warmcurrent prevails, copious rain and floodsmay occur (Caviedes,1975) (see Chapter 6Section C-4).

4. Interactions with Mid-Latitude Flow.Mid-latitude circulations often affect thetropics. These interactions are vividlyillustrated by some weather satellite pic-tures in which continuous cloud bandsextend far into the tropics (see Figure 8-57).Subtropical cyclones develop from upper-level lows which became cut off from themid-latitude westerlies. In turn, mid-lati-tudes are affected by recurving tropicalcyclones.

The above discussion shows that the tropicscannot be simply defined. The area hasbeen specified variously as: lying betweenthe tropics of Cancer and Capricorn; en-closed by the 20°C mean annual isotherm,within which average temperature exceeds20°C in all 12 months; where there is a heatsurplus in the earth-atmosphere system.These definitions encompass similar areas.Some regions experience a marked annualvariation. On the south coast of China,Hong Kong (22°N, 114°E), with a July meantemperature of 83.1°F (28.4°C), indubitablylies in the tropics in summer. However, itsJanuary mean temperature of 59.7°F(15.4°C) is scarcely tropical. Suffice to say,tropical meteorologists should not confinetheir attention to circumscribed latitudes.They must comprehend the dynamics ofmid-latitude circulations and extend theiranalyses poleward enough to identify andinterpret mid-latitude effects on the tropicalatmosphere.

5. Small-Scale Controls. Tropical circula-tions are also affected on small scales,primarily those associated with cumulusconvection. The role of near-equatorialcumulonimbus has already been stressed.Gray (1970) emphasized their importancefor vertical momentum transport and theproduction of kinetic energy for export tohigher latitudes.

Cumulus convection can affect a tropicalsquall line in two ways (Zipser, 1977).Highly unsaturated downdrafts are pro-duced by entrainment of cool dry air intomiddle levels of the cloud line. The leadingedge of the downdraft air acts like a coldfront, lifting warm moist air and enhancingconvection. The downdraft air in the rear ofthe squall line may eventually becomeorganized over the whole extent of thesystem, and effectively suppress convectionthere for several hours (see Chapter 8). Thedirection of movement, environmental

wind field, and interacting synoptic-scaleinfluences of these squall lines vary indifferent parts of the world.

6. Summary. The above examples illustrateinteracting atmospheric controls operatingat a variety of scales. Although there someexceptions, such as heat lows, size andduration of systems affecting the tropics areroughly correlated, as Figure 2-9 suggests.At one end of the range is the quasi-biennialoscillation (QBO) that covers the equatorialregion and follows a 26-month cycle, and atthe other end, intense, brief rain plumes intropical cumulus, that can be detected onlyby very fast-response instruments (Fullertonand Wilson, 1975). Complicated interactionsbetween scales must be considered in tropi-cal weather analysis and forecasting. Atpresent, most of the influences can only beevaluated subjectively. Despite someprogress, it will probably be many yearsbefore sub-synoptic scale motions, whichare greatly influenced by convection andtopography, can be usefully predicted bynumerical methods.

Chapter 3Chapter 3Chapter 3Chapter 3Chapter 3THE OBSERVTHE OBSERVTHE OBSERVTHE OBSERVTHE OBSERVAAAAATIONAL BASISTIONAL BASISTIONAL BASISTIONAL BASISTIONAL BASIS

A. Present Data Sources.

Tropical meteorologists have a wide varietyof data sources available for analysis, fore-casting and research. This chapter broadlyreviews these sources and their locations.Distribution is a problem. Some areas, suchas South Asia, Australia, the western Pacific,and parts of Africa are reasonably wellcovered with surface and upper-air stations.However, over vast ocean areas and parts ofSouth America, data are inadequate.

The weather reporting system is organizedinto regions by the World MeteorologicalOrganization (1986); the boundaries of thesix WMO regions are shown in Figure 3-1.Most of the stations in Regions I (Africa), III(South America), and V (Southwest Pacific),and some of the stations in Region II (Asia),and IV (North and Central America), arelocated in the tropics. Therefore statistics ontotal number of desired stations and hoursof observations in each region given in thefollowing sections will have to be consid-ered with this limitation in mind.

1. Surface Reports from Land Stations. Thecurrent surface synoptic network in thetropics is illustrated by the stations on the

latest readily available Department of De-fense (DOD) Weather Plotting Charts (e.g.,DOD-WPC’s 2-15-11 and 2-15-13 give a 1:15million Mercator projection for the region60°N to 60°S). The hours of surface observa-tions vary; however, most of the stationsreport on at least three of the four majorsynoptic hours (00, 06, 12, and 18 UTC) andmost report the intermediate three-hourlyobservations. In many tropical areas thereare enough surface synoptic reports foranalysis, at least at the six-hourly synoptictimes. However, compared to higher lati-tudes, the smaller weather systems andfewer observations in the tropics limit diag-nostic and prognostic information. Table 3-1compares the number of synoptic observingstations requested by WMO to the numberachieving the full observing program in1988 (WMO, 1989). There has been littlechange over the past few years; predomi-nantly tropical regions have been the worstin achieving the WMO goal of more observ-ing stations. Besides the first-order surfacesynoptic stations, which transmit real-timedata, many tropical countries operate lesser-order synoptic and climatological stations.These observe various meteorological ele-ments (especially rainfall) and periodicallysend the record to central offices of their

————————————————————————————————Table 3-1. Implementation of WMO synoptic observing program, 1988 (from WorldMeteorological Organization, 1989).

Region Numbers Requested Fully % AchievedRA I—Africa 704 385 54RA II—Asia 1171 1091 93RA III—South America 348 158 45RA IV—North and Central America 584 421 72RA V—Southwest Pacific 356 152 43RA VI—Europe 842 812 96

national meteorological services. Climato-logical stations are operated by agriculturalor industrial interests. These non-synopticdata can be used in local-area synopticstudies, in preparing climatological mapsand in other research. In some tropicalcountries, military weather services andairlines make observations and pass themover local military and aviation circuits.These data sometimes get into the Auto-mated Weather Network (AWN).

2. Surface Reports from Ships. In the beltbetween 30°N and 30°S, about 75 percent ofthe surface is covered by water. Therefore,ship reports are indispensable to tropicalanalysis and forecasting. The average dailynumber of ship reports collected by theNational Meteorological Center (NMC)during June 1988 for four-degree grid areasbetween 30°N and 30°S, shown in Figure 3-2, reveals vast empty spaces.

In June 1988, of the 110,117 radioed shipobservations, 30,071 or 27 percent weremade between 30°N and 30°S. Of them, 24percent were made at 00 UTC, 25 percent at06 UTC and at 12 UTC, and 22 percent at 18UTC. Circuit overcrowding and deadlinesprevent at least 40 percent of ship observa-tions from ever being transmitted on theWorld Meteorological Organization’s GlobalTelecommunications System (GTS) circuits,which interface with the AWN. About 15percent of all the radioed reports containerrors, most commonly in position and sea-level pressure (poor calibration). In criticalweather situations, the analyst, by followingindividual ships from chart to chart, canoften detect and correct these errors.

Some tropical forecasting units may berequired to analyze sea-surface temperature(SST), particularly when longer-range fore-casts are made. Because SST changes slowly,several-day compilations from ships andsatellites are used at NMC and at the Fleet

Numerical Oceanography Center (FNOC) toprepare SST analyses that are widely dis-seminated.

3. Surface Reports from Buoys. In the Gulfof Mexico, east of Florida, and aroundHawaii, moored weather buoys (Figure 3-2)transmit a wide range of meteorological andoceanographic data. In June 1988, WMO(1989) reported that worldwide, 200 driftingbuoys transmitted about 2000 weatherreports over GTS every day. Figure 3-3shows the distribution of drifting buoysover the tropical Pacific during August 1989(Climate Analysis Center, 1989). An encour-aging number are in the southeast Pacific.Measurements made on the much smallerbuoys are more representative of openocean conditions than measurements madeon ships.

Seven land-based automatic weather sta-tions are now providing important datafrom remote tropical islands (Figure 3-2).The federal departments of Navy, Com-merce and Interior plan to install 20 morestations in the western Pacific. The dataresemble those from moored buoys.

4. Upper Air Data. In the tropics, lack ofsufficient rawinsonde/pibal upper-air obser-vations is often a serious problem. Over theoceans there may not be a single verticalsounding in millions of square kilometers.Over the world, radiosonde observations aremade almost exclusively at 00 and 12 UTC,whereas radiowind and pilot balloon obser-vations are made at 6-hour intervals at 00,06, 12, and 18 UTC. There has been littlechange in recent years (WMO, 1989). Theperformance pattern of Table 3-2 resemblesthat of Table 3-1, and is reflected in Figure 3-4.

Aircraft make valuable observations(AIREPs). Commercial airlines makeAIREPs as required, and not routinely.Figure 3-5 shows the distribution of interna-tional commercial airline flights over thetropics. Flights by national airlines greatlyoutnumber international flights over largecountries such as China, Brazil and Austra-lia and between Hawaii and the US main-land. In most of the oceanic tropics, fewerthan five flights are made each day across a10° rectangle. Figure 3-6 shows the numberof AIREPs actually received for each 10-degree grid area by the Air Force GlobalWeather Central (AFGWC) over a period offour hours. Of the 320 AIREPs, 34 percentwere made from just two 10° rectangles —20°N to 30°N, and 140°W to 160°W, demon-strating how requirements other thanweather appear to control the pattern ofreporting. Twenty-three percent of thereports were made between 20°N and 20°S;in this belt, 78 percent of the 10° rectangleslacked AIREPs. AIREPS are concentrated inthe upper troposphere between 300 and 200mb, where jet aircraft fly. Fortunately, theselevels are critical for tropical weather diag-nosis and prediction.

In 1988 WMO reported that six Aircraft toSatellite Data Relay (ASDAR) systems wereoperating on wide-bodied jet aircraft. WithASDAR a great many automated wind andtemperature measurements are transmittedvia satellite to GTS. Besides this, operationalsystems now being deployed will obtain

turbulence and maximum wind data, andvertical soundings during take-off andlanding.

Air Weather Service operates 12 hurricanereconnaissance aircraft equipped with thenew Improved Weather ReconnaissanceSystem (IWRS). The automated systemresembles that used on NOAA researchaircraft. Every second, it collects position,wind, pressure altitude, absolute altitude,D-value, temperature and humidity datafrom sensors on the aircraft. Temperature,pressure, humidity and wind beneath theaircraft are sampled by Omega dropsondewind finders.

5. Meteorological Satellite Data. Satelliteshave truly revolutionized tropical meteorol-ogy. In mid-latitudes, satellite pictures led tono great surprises since the typical clouddistributions associated with fronts andextra-tropical cyclones had already beendetermined from fairly dense surface net-works. Apart from a few modifications,satellite data have reinforced the originalpolar-front concept. But in the tropics,which has always been data sparse, satelliteshave enhanced our knowledge of develop-ment, movement, and typical distributionsof the major weather systems. These dataoften help the forecaster to relate clouddistributions to circulation patterns, infer-ring the latter from the former.

Table 3-2. Implementation of WMO radiosonde observing program, 1987 (from WorldMeteorological Organization,1989).Region Number Percentage Requested ImplementedRA I—Africa 99 49.5RA II—Asia 325 91.7RA III—South America 58 56.1RA IV—North and Central America 153 94.7RA V—Southwest Pacific 97 59.3RA VI—Europe 145 91.7————————————————————————————————

The first Television Infrared ObservationalSatellite (TIROS 1) was launched on 1 April1960. This was followed by nine more re-search satellites in the TIROS series, the lastbeing launched in July 1965. Based on theknowledge gained from the TIROS series, aNational Operational Meteorological Satel-lite System (NOMSS) was developed andbecame operational in February 1966, withthe launches of ESSA I and II. Besides theTIROS and ESSA series, several of the NIM-BUS and Applications Technology Satellite

(ATS) series have been used for research anddevelopment with some real-time opera-tional applications.

More recently, The National Oceanic andAtmospheric Administration’s (NOAA)Polar-orbiting Operational EnvironmentalSatellite (POES), the Defense MeteorologicalSatellite Program (DMSP), and the Geosta-tionary Operational Environmental Satellite(GOES) systems have provided comprehen-sive coverage. Each NOAA and DMSP

———————————————————————————————Table 3-3. Operational weather satellites (From National Aeronautics and SpaceAdministration, 1988, and Yu, 1990)).

Program Agency Objectives

POES (TIROS-N): Polar-orbiting NOAA Meteorological observationsOperational measurements of sea ice,Environmental Satellite and snow cover, assessment of

condition of vegetation

GOES: Geostationary Oper- NOAA Operational weather data,ational Environmental cloud cover, temperatureSatellite — normally at profiles, real-time storm75°W and 135°W monitoring (rapid scan),

severe-storm warning, seasurface temperature

DMSP: Defense Meteorology DOD Operational weather obserSatellite Program — vations for Department ofpolar orbiter Defense, tropical cyclone reconnaissance,

microwave observations

METEOSAT: Geostationary ESA Operational weather data,Meteorological Satellite Europe cloud cover, water-vaporat 0°E imagery

GMS; Geostationary Meteor- NASDA Operational weather data,ological Satellite — at Japan cloud cover, temperature140°E profiles, real-time storm

monitoring (rapid scan),hydrological observations

METEOR: Meteorological USSR Meteorological observationsSatellite — polar sea-surface temperature, seaorbiter ice, snow cover, vegetation condition

INSAT: Geostationary Indian India Meteorology, domestic comm-National Satellite — at unications, television pro-74°E gram distribution

FY-1: Polar orbiter China Meteorological observations, seasurface temperature, sea ice,snow cover, vegetation condition

Gaps may occur in a program when a satellite fails and a replacement is delayed.

satellite, which views each part of the earthat least twice daily, contains a suite of sen-sors, including conventional satellite imag-ery and vertical sounding sensors. GOES,anchored at a point over the equator, pro-vides pictures every 30 minutes, and moreoften on request. Foreign governments alsooperate weather satellites.

Table 3-3 lists satellite capabilities.

Satellite soundings fill large data gaps. Theyhave led to improved global analyses, andbenefited numerical weather prediction.Geostationary satellites are especially valu-able to tropical meteorologists. Besides thetwo operated by NOAA at 75°W and 135°W,a Japanese satellite is located at 140°E. Indiais responsible for 74°E and the Organizationfor the Exploitation of Meteorological Satel-lites maintains one at 0°E. The geostationarysystems indirectly provide excellent low-level and high-level wind data by allowingindividual clouds to be tracked (Figures 3-4and 3-6), again filling many data voids.

Tropical meteorologists use satellite picturesto fix the cloud system centers of tropicalcyclones, and from successive imagery, todetermine center movement. In the westernNorth Pacific, satellites have replacedweather reconnaissance aircraft in this role.Cyclone intensity can now be estimatedfrom the appearance of the clouds (Dvorak,1975). The temperature of the air column inthe eye of a tropical cyclone is inverselyrelated to the central surface pressure, anddirectly related to the maximum surfacewind. The temperature difference at 250 mbbetween the eye and the surrounding re-gion, determined by microwave (54.96 GHz)measurements from NOAA polar-orbitingsatellites were statistically related to recon-naissance-observed central pressures andwinds. Standard errors of the estimatesamounted to 8 mb and 13 KN for the Atlan-tic (103 cases) (Velden, 1989) and 13 mb and17 KN for the western Pacific (82 cases)(Velden et al., 1991).

The IR channel (10.5 to 12.5 µm) that mea-sures cloud top temperatures, and its vari-ous “enhancements”, is especially useful.Tropical meteorologists also use satelliteimagery to determine the movement ofshear lines, troughs in the TUTT, and thesubtropical ridge, as well as the extent andintensity of weather associated with each.Vapor images (in the 6.7 µm band) of cloudyareas resemble IR images, but they can alsovertically integrate moisture content wherethere is no cloud, and so delineate subsid-ence and rising motion better than cloudpictures.

The special sensor microwave/imager (SSM/I) on DMSP senses surface wind speed overthe ocean where no rain is falling (Hollidayand Waters, 1989), and enables areas ofdestructive winds around tropical cyclonesto be mapped. Every two weeks a group atthe National Meteorological Center (NMC)combines measurements from ships withsatellite-sensed infrared radiances from thesea to produce excellent global sea-surfacetemperature charts. Satellite-sensed infraredradiances from cloud tops can be translatedinto heights. From these, if convective ori-gins are assumed, rainfall can be estimated.As subsequent chapters show, enoughsatellite data have been accumulated togreatly contribute to tropical climatology.

6. Weather Radar Data. Weather radarprovides the most important short-rangelocal Nowcasting, forecasting and weatherwarning aid. It gives forecasters a three-dimensional view of rain areas out to 150km or more from the observation site. AsTable 3-4 shows, radars are scanty in Africa,South America and the southwest Pacific,with little or no improvement in recentyears.

Table 3-4. Ground weather radar stations in 1988 (from World Meteorological Organization,1989)

Region No. of Wavelengths stations 3 cm 5 cm 10 cm other

RA I—Africa 54 11 20 16 7RA II—Asia 193 107 54 32RA III—South America 17 5 3 9RA IV—North and Central America 154 5 81 68RA V—Southwest Pacific 56 2 12 42RA VI—Europe 170 118 39 13————————————————————————————————

———————————————————————————————-Table 3-5. Proposed satellite missions (from National Aeronautics and SpaceAdministration, 1988). Unless specified, United States agencies are responsible.

Program Agency/ Objectives Status

TOPEX/POSEIDON: Ocean NASA-CNES Ocean surface topographyTopography Experiment (France) Launch 1991

POES - follow-on miss- NOAA/ Advanced capabilities forions (NOAA K,L,M) Planned weather observations

GOES - follow-on miss- NOAA/ Advanced capabilities forions (e.g., GOES-Next) Planned weather observations

DMSP - follow-on DOD Advanced capabilities for missionsweather observations

MOS-2 Marine Observat- NASDA/Japan Passive and active micronSatellite-2 Launch about wave sensing 1990

EOS: Earth Observing NASA/NOAA Long-term global earthPOES Launch 1994 observations

Rainfall mission NASA/ Tropical precipitation Launch 1994 measurements

NOAA operations NOAA/ Weather observations and Planned atmospheric composition; observations of ocean and ice surfaces; land surface imaging; Earth radiation budget; data collection and location of remote measurement devices; detection and location of emergency beacons; monitoring of space environment

7. Lightning Detection. In the tropics,lightning detection systems are operating inAustralia, Hong Kong, Indonesia, SriLanka, and the southern United States(World Meteorological Organization, 1989).Information from these systems is notavailable on GTS.

B. PB. PB. PB. PB. Probable Future Data Sources.robable Future Data Sources.robable Future Data Sources.robable Future Data Sources.robable Future Data Sources.

1. Meteorological Satellites. By the end ofthe 20th century, new sensors will measuretropical winds at several levels includingthose in and around tropical cyclones,providing the data density needed by thehurricane/typhoon prognostic models andgiving general support to numericalweather prediction. Other new sensorsmaking accurate, high resolution measure-ments in the vertical could greatly enhanceanalyses used to initialize numerical fore-cast models. Rainfall measurements fromspace will greatly help tropical meteorolo-gists. Lightning has been detected at nightfrom polar-orbiting satellites (see Figure 10-2). By the mid-1990s, a new sensor flown onGOES will operate continuously and detectlightning at a 10 km resolution (Christian etal., 1989). Table 3-5 lists proposed newprograms. Should they succeed, their prod-ucts will be made available to operationalmeteorology.

2. Radar.

a. Doppler Radar. In the near future, mostcountries in the tropics will probably main-tain the current 3-,5-,or 10-cm conventionalweather radars. However, a revolutionarystate-of-the-art weather surveillance radarsystem (WSR-88D) was developed underthe Next Generation Weather Radar(NEXRAD) program. NEXRAD is a jointAir Weather Service (AWS), National

Weather Service (NWS), Federal AviationAdministration (FAA) project using Dop-pler techniques. Not only can a 10-cmDoppler radar detect and locate raindrops,but if a pulsed signal is combined withcircular scanning at a fixed elevation, theDoppler shift produced in the reflectedsignal by horizontal movement of thereflecting particles can be determined andmotion toward or away from the radarcalculated. This has been known for a longtime (e.g., Probert-Jones, 1960). Now anetwork of Doppler radars is being in-stalled in the United States and at someoverseas tropical forecast offices and mili-tary installations. This radar will increasethe user’s ability to detect severe weatherwhile reducing the false alarm rate. WSR-88D will find and alert the operator to mid-cloud vortexes (e.g., tornadoes, and possi-bly “microbursts”), gust fronts, tilted stormaxes, wind shears, and allow monitoring oftropical cyclone circulations. The operatorand severe weather forecasters at majorcentralized weather facilities will also beable to remotely monitor other terminals onthe NEXRAD network. Table 3-6 showswhere WSR-88Ds are being installed in thetropics (beyond the continental UnitedStates). A Doppler radar has been operatingat Taipei (25°N, 122°E) since 1987; a net-work of Doppler radars is planned forThailand.

b. Profiler. The Doppler principle, appliedto very-high or ultra-high frequency up-ward pointing radar, which can detectdiscontinuities in clear air, can be used todetermine vertical motion in the radarbeam. If two beams are pointed slightly offvertical, the horizontal wind can be esti-mated through the troposphere above thesite. The calculation is based on the twoslant range radial (Doppler) velocity mea-surements for each level (assuming zerovertical velocity) (Gauge et al., 1990). Windprofiles can be measured every 1-2 minutes.NOAA is now installing 30 profilers in the

central United States, and maintains re-search profilers at Pohnpei (7°N, 158°E),Christmas Island (2°N, 157°W), Piura (5°S,81°W), Darwin (12°S, 131°E) and Biak (1°S,136°E). Data from the Christmas Islandprofiler are incorporated in routine analysesat NMC and ECMWF. The other stationsand two stations planned for Saipan (15°N,146°E) and Darwin (12°S, 131°E) will alsoprovide data on GTS.

3. Summary. Data available to tropicalmeteorologists will increase in the yearsahead. Satellite communication techniqueswill be used to collect and relay the multi-tude of environmental observations tocentralized forecast offices, and then todisseminate the products to users. TheGlobal Telecommunications System (GTS),already overloaded, should be greatlyexpanded to cope with present as well asfuture needs. More detailed satellite data,and new ways of applying, combining, andenhancing multi-spectral images will un-doubtedly aid our understanding of tropicalmeteorology and perhaps enable us todevelop better conceptual and numericalmodels for the tropics.

————————————————————————————————Table 3-6. Tropical stations beyond the continental United States at which 10-cm Dopplerradars (WSR-88D) are being installed.

Station Lat. Long. Station Lat. Long.

Kadena AB 26°N 128°E Kamuela 20°N 156°WSouth Kauai 21°N 158°W Clark AB 15°N 121°EMolokai 21°N 157°W Anderson AB 14°N 145°E

Chapter 4Chapter 4Chapter 4Chapter 4Chapter 4PRESSURE AND WINDSPRESSURE AND WINDSPRESSURE AND WINDSPRESSURE AND WINDSPRESSURE AND WINDS

A. General.

In this chapter, the major characteristics oftropical pressure and wind fields are illus-trated by mean maps, cross-sections, latitu-dinal profiles, typical station climatology,and circulation models. The wind field isemphasized. In the tropics, important per-sistent features are better defined by theresultant winds than by pressure and pres-sure-height fields, which tend to have smallgradients. The tropical upper-tropospherictroughs (TUTT) over the Atlantic and Pa-cific, rarely detectable in the height fields,often show clearly in the resultant windfields.

In much of the tropics, the two main sea-sons, wet and dry, are separated by rela-tively brief transition seasons. Thus mid-season months (January, April, July, Octo-ber) describe the annual cycle pretty well.However, local forecasting often demands amore detailed background — of monthly oreven 5-day (Pentad) climatologies. A firmgrasp of upper tropospheric circulationpatterns is essential.

Standard deviations (square root of themean of the squares of the differences be-tween individual values and the meanvalue) of the surface pressure and otherfields provide useful information to theforecaster. Although observational errorscontribute, the standard deviation roughlyindicates synoptic variability.

B. PB. PB. PB. PB. Pressure.ressure.ressure.ressure.ressure.

1. Mean Sea-Level Pressure. Figure 4-1gives the mean sea-level pressure and the

standard deviation of pressure between40°N and 40°S for the mid-season months.Throughout the year, low pressure generallyprevails near the equator, and high pressureover the oceanic subtropics. The SouthernHemisphere (SH) subtropical high-pressureridge migrates over a smaller range oflatitude than its Northern Hemisphere (NH)counterpart.

In the oceanic ridges, mean pressure ishighest in July, reaching 1025 mb; dailyvalues often exceed 1030 mb. Because ofextra-tropical cyclones, standard deviationsare much larger in winter than in summer.Although weather is variable equatorwardof the ridge, standard deviations of onlyabout 2 mb prevail, not much more than theexpected observational error. Most weatherdisturbances cause pressure fluctuationsindistinguishable from noise. No wondertropical meteorologists prefer to analyzeweather in terms of the wind field.

In the winter hemisphere, the warm oceanicsubtropical ridge and the cold continentalridge form a continuous belt of high pres-sure around the hemisphere. During sum-mer, as a result of differential heating be-tween land and ocean, heat lows developover the continents adjacent to the IndianOcean. The consequent reversal of pressuregradient between winter and summer givesrise to the monsoons (see Chapter 6 SectionC-5).

Zonal averages smooth the large departuresoccurring at various longitudes. The broad-scale meridional variations remain. Figure 4-2 shows the zonally averaged sea-levelpressure for January and July computedfrom Figure 4-1. Arrows show the north-

south movement of the subtropical ridgesand the low-latitude trough between Janu-ary and July. The average trough positionvaries from 5°S to 6°N. The mean ridge inthe NH ranges over seven degrees, from30°N to 37°N, compared to five degrees forthe SH ridge, from 33°S to 28°S. Because thetrough moves more than the ridges, thetrade winds are much broader in winterthan in summer. Pressure in the zonal meansubtropical ridge rises from 1015 mb insummer to slightly over 1020 mb in winter,reflecting the summer heat lows and coldhighs of the continents.

2. Large-Scale Pressure Changes. Lasting afew days, simultaneous widespread pres-sure variations of unknown origin wereobserved over India (Eliot, 1895) and theCaribbean (Frolow, 1942). Figure 4-3 (fromBrier and Simpson, 1969) shows simulta-neous pressure anomalies for three Pacific-area stations for a 15-day period. The diur-nal and semi-diurnal variations have beenremoved by using departures from themonthly means of hourly values. In thetropics, shear lines may be over 2000 NM(3700 km) long, while vortexes are rarelymore than 1000 NM (1900 km) across. Thus,although the stations in Figure 4-3 are toofar apart for the same synoptic system toaffect all three at once, their broad-scalepressure changes are well correlated. In acase studied by Palmer and Ohmstede(1956), pressure over the entire tropicalPacific fell almost uniformly by 2.5 mb in 48hours. Sadler et al., (1968) reported an evenmore massive event over the western Pa-cific and East Asia (Figure 4-4).

More recently, Lander (1991) studied asequence of widespread pressure fluctua-tions over the Pacific during May 1986. AsFigure 4-5 shows, a north-south oscillationcovered most of the basin and impliedcorresponding fluctuations in surfacewinds. According to Guard (1985) such

changes accompany wind surges, in theforward parts of which, weather deterio-rates (see Chapter 8, Section D-4).

C. Wind.

1. General. The levels most analyzed in thetropics are the surface/gradient-level and200 mb, primarily because they have themost observations. This is fortunate sincethe lower and upper troposphere controlthe vertical motion fields that produce mosttropical weather. Regions of upward mo-tion shown in Figure 2-4 generally coincidewith troughs (convergence) at the gradientlevel (Figures 4-6 to 4-9) and with anticy-clones (divergence) at 200 mb (Figures 4-10and 4-11). Mid-tropospheric analyses delin-eate mid-tropospheric cyclones and help inpredicting tropical cyclone movement, butunfortunately, data come only from rawinand rawinsonde stations (see Figure 3-4).This section presents the resultant windclimatology for the lower and upper levelsfor the mid-season months. Local winds arediscussed in Chapter 7.

2. Resultant Gradient-Level Winds. Chartsfor mid-season months shown in Figures 4-6 to 4-9 were extracted and reduced fromAir Weather Service Technical Report 215(1970) which contains monthly charts on a1:20 million Mercator map projection. Thegradient level is defined as the lowest levelat which predominantly friction-free flowoccurs. Over most of the tropics, this isabout 3000 feet (1 km). For stations below1000 feet elevation, winds at 3000 feet wereused when available. For stations with dataonly at standard pressure levels, the 850 mbwinds were used and adjusted by compar-ing with vertical wind profiles at neighbor-ing stations. For stations above 1000 feetresultant winds at 5000 feet (1.5 km) weregenerally used. Over the oceans, resultant

surface winds from ship reports formed thebasis of subjective estimates of the resultantgradient-level winds.

The wind field is represented on the chartsby kinematic analyses (Palmer et al., 1955).Full lines depict streamlines (arrow headsshow flow direction), dashed lines theresultant isotachs (drawn for 5 knot inter-vals), and shaded areas resultant speeds <5and >15 knots. To retain patterns, stream-lines were drawn through large mountainbarriers, such as the Andes in SouthAmerica; but remember, gradient-level flowhas little or no meaning in these regions.

Resultant winds give no direct informationon variability. Steadiness serves the samepurpose with winds as standard deviationdoes with scalar quantities such as pressureand temperature. The long-term resultantwind speed is quite highly correlated withsteadiness and roughly indicates variability(see Figure 11-15).

The major features shown by the mean sea-level pressure fields (subtropical high-pressure ridges and low-latitude troughs)are equally evident in the gradient-levelwind fields. This coincidence arises fromthe fact that for long-term means, gradient-level winds parallel isobars to within about3° from the equator. Along the equator,where the Coriolis force is zero, a balance isstruck between the mean pressure gradientand mean friction. The mean resultant windthen blows perpendicular to the isobars,from high to low pressure. This is also truefrom day to day, if the direction of thepressure gradient remains unchanged. Overthe western Indian Ocean in January (July),winds blow from north (south) across theequator. Easterlies prevail in an east-westpressure gradient over the central andwestern parts of the equatorial Pacific andAtlantic, and over northern South America(Figure 4-1). During El Niño, the pressuregradient over the equatorial Pacific oftenreverses to west-east and westerlies then

replace easterlies (see Chapter 6, Section C-4). The central North and South Atlanticand the central North and South Pacific areleast influenced by the large continents.Over these regions, tradewinds prevail. Thenear-equatorial convergence zone (NECZ),where the northeast and southeast tradesmerge, lies in a broad, very weak, trough. Inthe summer hemisphere and over theoceans, the gradient-level wind patternsclosely resemble those of the surface winds.But over the continents in winter, largethermal gradients produce significantdifferences. For example, in January, meansea-level pressure shows an intense anticy-clone over East Asia north of 40°N (Figure4-1) while the gradient wind field of Figure4-6 shows a small anticyclone near 30°N.

January (Figure 4-6):

Northern Hemisphere

(1) The tropics are dominated by the oce-anic subtropical ridges which merge withthe continental anticyclones.

(2) The northeasterlies south of this ridgesystem are strongest near 10°N.

(3) The tradewinds converge near 5°N.Winds blow across the equator from Africato 150°E with a near-equatorial trough overIndonesia.

(4) Anticyclones prevail over the winter(NH) continents and heat lows over thesummer (SH) continents. The strongestnortherlies along the equator occur overand near Africa, reflecting the large pres-sure gradient between a winter high in thenorth and a summer heat low to the south.

Southern Hemisphere

(5) A trough extends from a low near 13°S,170°E, westward across northern Australia,where it becomes a monsoon/heat trough.West of Australia the trough lies near 10°S.

Near 50°E it trends poleward to link withthe monsoon/heat trough over southernAfrica. Most tropical cyclones develop inthe oceanic parts of this 15,000 km-longtrough.

(6) The trough extending southeastwardfrom the low near 13°S, 170°E, is associatedwith tropical and extratropical synopticsystems and relative maxima of cloud andrain along what is known as the SouthPacific Convergence Zone. It persiststhroughout the year and has no NH coun-terpart.

(7) The broad subtropical anticyclone overthe South Atlantic also persists throughoutthe year, as does the trough east of theAndes.

April (Figure 4-7):

Northern Hemisphere

(1) The oceanic subtropical ridges, the near-equatorial trough and the near-equatorialconvergence zones lie close to their Januarypositions.

(2) Northeasterlies south of the ridge havediminished, especially over the South ChinaSea, where the winter monsoon is almostover.

(3) Heat lows develop over China, southernIndia and western North Africa, replacingthe ridges of January.

(4) The low over northern Africa stemsfrom extratropical cyclogenesis east of theAtlas Mountains (Gleeson, 1954).

(5) Anticyclonic circulations persist over theArabian Sea and the Bay of Bengal, but byMay, give way to the southwest monsoon.

Southern Hemisphere

(6) The monsoon trough has moved northand extends from 50°E to 150°E. OverIndonesia and the Indian Ocean the near-

equatorial troughs of both hemispheres arenearly equidistant from the equator. Thisstructure, typical of spring and autumn,favors “twin” tropical cyclone formation —one in each hemisphere.

(7) The subtropical ridge over the oceansremains poleward of 30°S.

(8) The heat lows over central SouthAmerica, southern Africa and northernAustralia are replaced by tradewinds;anticyclones start to move over southernparts of the continents.

July (Figure 4-8):

Northern Hemisphere

(1) The oceanic anticyclones have movedwell north and expanded. South of theanticyclones, tradewinds are strongestbetween 15°N and 20°N.

(2) The heat/monsoon trough is well-estab-lished over southern Arabia and NorthAfrica, and over south and southeast Asia,with speeds exceeding 18 m s-1 in the Ara-bian Sea and 13 m s-1 in the Bay of Bengal.The cyclone east of the Philippines reflects aconcentration of tropical cyclone tracks.Over the Philippines, southwesterlies areweak because the monsoon trough thereranges over about 10° of latitude.

(3) A heat low has developed over Mexico.

(4) The cool-season NECZ over the north-east Pacific has been replaced by a troughextending westward from northern SouthAmerica.

Southern Hemisphere

(5) The weak monsoon trough is now con-fined to the Indian Ocean.

(6) Strong tradewinds dominate the tropics.

(7) The SPCZ now reflects passage of extrat-ropical systems.

October (Figure 4-9):

Northern Hemisphere

(1) Except for a minor heat low over theSahara, cyclonic circulations are gone fromthe continents.

(2) The southern Asia heat/monsoon troughnow lies across the central Arabian Sea andBay of Bengal. It is linked to the westernPacific near-equatorial trough, which hasremained quasi-stationary. Tropical cy-clones may form in this east-west trough,now extending to Africa.

(3) Monsoon westerlies have abated overNorth Africa and been replaced by a troughbetween 5 and 10°N.

(4) In the eastern North Pacific, the near-equatorial trough has shifted slightly south.Tropical cyclones may form in it on both thePacific and Caribbean sides of CentralAmerica.

Southern Hemisphere

(5) The broad band of transequatorialwinds east of Africa and north of Australiahave abated. Otherwise the tradewindshave not changed much.

(6) A heat low has formed over southernAfrica in advance of the rainy season, andtroughs have appeared over central SouthAmerica and northern Australia, anticipat-ing the heat lows of summer.

3. Resultant 200 mb Winds. These fields,shown in Figures 4-10 and 4-1l, will bediscussed briefly. They are based on AIREPas well as rawin and rawinsonde averages.Sadler (1975b) gives more detail. In thelongitudes where only a single ridge isfound in the tropics, it is termed the “sub-tropical” ridge. When two ridges occur, thepoleward is termed “subtropical”, and theequatorward, “subequatorial”. The inter-

vening trough is known as the tropicalupper tropospheric trough (TUTT) (seeFigure 8-21).

January (Figure 4-9, top):

Northern Hemisphere

(1) The mean subtropical jet stream (STJ)reaches 143 knots just south of Japan, afterbeing established in November.

(2) The subtropical ridge near 10°N extendsfrom 0° eastward to 170°E.

Southern Hemisphere

(3) The STJ generally lies south of 35°S.

(4) The subtropical ridge, lying between10°S and 20°S, extends from 0° eastward to180°.

(5) In the SH, TUTT’s extend to the equator,where westerlies prevail. The TUTT axesare oriented southeast-northwest. In thePacific the axis extends from 30°S, 105°W to175°W and the equator; in the Atlantic,from 30°S, 15°W across South America to75°W and the equator. The troughs areseparated by an anticyclone over westernSouth America.

April (Figure 4-9, bottom):

Northern Hemisphere

(1) The STJ shifts north of 35°N and weak-ens.

(2) The subtropical ridge moves slightlysouth to near 8°N.

(3) A narrow band of equatorial easterliesextends from 170°E westward across theIndian Ocean and Africa into the Atlantic.

Southern Hemisphere

(4) The SH STJ has moved north to about25°S to 30°S. Maximum resultant speedsexceed 70 knots east of Australia.

(5) The subtropical ridge has moved northto between 10 and 15°S.

(6). TUTTs are much weaker.

July (Figure 4-10, top):

Northern Hemisphere

The circulation changes dramatically fromApril to July.

(1) The STJ disappears and the polar jetstream moves north of 35°N.

(2) An anticyclone is established near 32°Nover North America.

(3) Over Asia the subtropical ridge movesfrom near 10°N in April to 30°N in July. Tothe south a tropical easterly jet stream isestablished over the north Indian Oceanwith mean winds exceeding 50 knots oversouthern India.

(4) Over the Indian Ocean, the strong south-erly monsoon flow across the equator at lowlevels (Figure 4-8) is matched by equallystrong flow in the opposite direction at 200mb.

(5) TUTTs oriented southwest-northeasthave appeared over the North Atlantic andNorth Pacific. The mean mutes day-to-dayshifts of the trough and the transient effectsof embedded cold-core cyclones.

(6) A subequatorial ridge extends from130°E eastward to 50°W. Anticyclones in theeastern and western Pacific parts of theridge overlie tropical cyclone activity.

Southern Hemisphere

(7) The STJ near 28°S has moved onlyslightly south since April but has strength-ened to 105 knots over Australia.

(8) A subtropical ridge, lying between 5°Sand 15°S, extends westward from 180° to20°W.

(9) Over the South Pacific the TUTT is nolonger apparent.

October (Figure 4-10, bottom):

Northern Hemisphere

(1) The NH ridge over Asia and Africa hasmoved south to between 20°N and 10°N.The North American anticyclone has shiftedsouth to near 15°N; except where equatorialwesterlies intercede, the subtropical ridge iscontinuous around the NH.

(2) The tropical easterly jet has abated.

(3) Equatorial westerlies have reappearedover the eastern Pacific and Atlantic; how-ever, small resultant speeds indicate largedaily variability.

(4) The TUTTs have weakened and thesubequatorial ridge has gone.

Southern Hemisphere

(5) The STJ has moved only slightly fromJuly, having weakened to about 75 knotsover Australia.

(6) The ridge lies between 5°S and 10°S withanticyclones over the rain forests of SouthAmerica and Africa.

Studies have been made of the mean windfield at various levels for specific regions ofthe tropics. An upper-air atlas prepared byRamage and Raman (1972) in conjunctionwith the International Indian Ocean Expedi-tion (IIOE) presents resultant wind analysesfor standard pressure levels from 850 mb to100 mb and representative meridional cross-sections for the area 45°N to 50°S and 20°Eto 155°E. Chin and Lai (1974) producedcharts for 850, 500 and 200 mb for the area46°N to 14°S and 66°E to 147°E.

Global cross-sections can provide only thebroadest view, and are particularly uninfor-mative in the monsoon regions. For ex-

ample, the strongest mean surface windsblow from southwest across the Arabian Seain summer (Figure 4-8). But Figure 2-5shows the global average to be easterly inthose latitudes.

4. Tropical Jet Streams.

a. Subtropical Jet Streams. As shown inFigures 4-10 and 4-11, subtropical jetstreams (STJ) are climatological features ofthe tropical general circulation. Over theNH in winter, the STJ possesses a three-wave pattern with ridges and maximumspeeds over eastern North America, theMiddle East and south of Japan. The meanlatitude ranges from 20°N to 35°N. Steadi-ness exceeds 90 percent and speeds mayoccasionally exceed 200 knots. Over the SHthe winter STJ exceeds 100 knots only overand to the east of Australia, where it liesnear 27°S.

b. Tropical Easterly Jet. This summer featureis much broader than the STJ. Its axis, at 150mb, extends from 5°N to 15°N and extendsacross the Indian Ocean and Africa. Averagespeeds exceed 70 knots; between 1956 and1962, 151 knots was once measured at 59,000feet (18 km) over Bombay (Flohn,1964). Thejet will be discussed again in Chapter 6.

c. Low-Level Jet Streams. Two types of low-level jet stream (LLJ) occur in the tropics.

Seasonal or year-round LLJs blow wherepersistent, large surface temperature gradi-ents enhance surface pressure gradients,and where vertical motion above about 6500feet (2 km) is inhibited by subsidence. Fa-vored locations include desert littorals,especially where coastal waters are up-welling, and over equatorial upwelling(Figure 5-4). LLJs also border heat troughsor sharply defined zones of heavy rain. Inall of these cases, the thermal wind opposesthe wind in the surface layer. As Grossmanand Friehe (1986) and others have pointed

out: (1) surface friction over the sea and anocturnal surface inversion over land ensurethat wind increases with height above thesurface; (2) the horizontal temperaturegradient, by creating a thermal wind oppo-site to the surface wind, causes the windabove the surface friction layer to decreaserapidly with height. Thus these two pro-cesses define and sharpen the LLJ.

The largest and most intense LLJ prevailsover the western Indian Ocean during theNH summer, extending from east of Mada-gascar across eastern Somalia to PeninsularIndia. At its core height of 5000 feet (1.5 km),speeds may exceed 60 knots (Figure 4-8).Over eastern Somalia, the jet is enhanced bythe temperature gradient across upwellingcoastal waters and the coastal desert. Also, itis confined and accelerated by a subsidenceinversion and by mountains to the westparalleling the flow. LLJs have been ob-served along the west coasts of SouthAmerica south of the equator, and Namibia.

At somewhat higher levels, between 5000and 15000 feet (1.5 to 4.5 km), northernSouth America experiences LLJs along about15°N during summer, while over westAfrica and south of the Mei-Yu front oversouth China (see Chapter 6, Section C-5d)they occur in spring and early summer.These jets seldom exceed 25 knots and seemto result from temperature gradients causedpartly by release of latent heat of condensa-tion. All seasonal or annual LLJs appear toaccelerate in response to surges from higherlatitudes.

Nocturnal LLJs develop over land whenfresh synoptic winds coincide with clearskies. Radiational cooling, by creating asurface inversion and reducing dissipationby friction, allows the winds between about1000 and 3000 feet (300 and 900 m) abovethe surface to increase; a bordering moun-tain range and an overlying inversion nar-row the jet layer and further increase thespeed. The jet is strongest near dawn. Along

the coastal plain of northeastern SaudiArabia between 24°N and 29°N, a nocturnaljet in the prevailing summernorthwesterlies may exceed 50 knots be-tween 800 and 1500 feet (250 and 450 m)(Vojtesak et al., 1991). At Daly Waters (16°S,133°E) in Australia, a nocturnal LLJ occurson 19 percent of all winter nights. It isgenerally a hundred or two NM wide andabout 500 NM (1000 km) long (Brook, 1985).Over land, daytime surface heating, leadingto vertical mixing, weakens and may dissi-pate all LLJs. This process in discussed inChapter 7, Section D-5b.

5. Tropical Stratospheric Winds. A quasi-biennial oscillation (QBO) of winds in thetropical stratosphere was discovered byMcCreary (1959). This phenomenon hasbeen intensively investigated by Reed et al.(1961) and many others. The oscillationbetween easterly and westerly wind re-gimes occurs on an average every 26months, but the period is quite irregular,varying between 21 and 30 months (hence“quasi-biennial”). The maximum of theoscillation occurs at 75,000-85,000 feet (23-26 km) above the equator and decreasesdownward and poleward. The easterly orwesterly wind regime, which first appearsabove 98,000 feet (30 km), sinks at about3000 feet (1 km) a month without losingamplitude in the first 23,000 feet (7 km). Butbelow 75,000 feet (23 km), it attenuatessharply and is barely noticeable at 100 mb,the average level of the tropical tropopause.

Figure 4-12 illustrates the QBO in the equa-torial stratosphere from 1953 through 1984(Naujokat, 1986). The annual cycle of strato-spheric wind change is small at the equator,where the QBO is most prominent. In thetroposphere and at more poleward stations,the larger annual cycle renders the ob-served wind pattern more complex. Forexample, the 21 km mean monthly zonalwinds at Darwin (12°S, 131°E) from 1958 to

1968 reflect both the annual and quasi-biennial cycles (Figure 4-13) (Hopwood,1968). The average amplitudes of the twocycles are shown in Figure 4-14. Along theequator, the QBO dominates; poleward ofabout 20°, the annual oscillation is muchlarger. Although the QBO has been well-described, its exact cause is still debated(Wallace, 1973).

The climatology of tropical stratosphericwinds cannot be satisfactorily representedin standard format such as resultant windson constant pressure surfaces or by verticalcross-sections. Instead, the mean winds fora given month and year can be anticipatedby extrapolating into the future the recentQBO cycle for the station and levels ofinterest. At most, the QBO affects the tropo-spheric circulations only marginally(Trenberth, 1980), and weather not at all.

Chapter 5Chapter 5Chapter 5Chapter 5Chapter 5TEMPERATEMPERATEMPERATEMPERATEMPERATURE AND WTURE AND WTURE AND WTURE AND WTURE AND WAAAAATER VTER VTER VTER VTER VAPORAPORAPORAPORAPOR

shows, performances of all electroopticaldevices can be seriously affected by atmo-spheric water and dust. Forecasters shouldmemorize this table, and use it to assess theimpact of tropical weather on electro-optics.

This chapter examines the distributions andvariations of temperature and water vaporin the global tropics.

B. Surface Air TB. Surface Air TB. Surface Air TB. Surface Air TB. Surface Air Temperature andemperature andemperature andemperature andemperature andWWWWWater Vater Vater Vater Vater Vaporaporaporaporapor.....

1. Means. The mean surface air tempera-ture and standard deviation for January,April, July, and October are shown in Fig-ure 5-1. Because of their small scale, thesemaps can show only the largest tempera-ture features and should not be used toobtain values for specific locations.Climatologies of temperature, humidityand rainfall and of other elements at indi-vidual stations are readily available in

A. General.

Most of the tropical atmosphere is charac-terized by small horizontal temperaturegradients. Near the surface, differentialheating of land and water, or cold oceanwaters upwelling near coasts, may causelarge temperature gradients; however, theeffects of these disappear or are greatlyreduced several hundred feet (a few hun-dred meters) above the surface. Even thestrongest cold fronts, which may penetratedeep into the tropics, can scarcely be foundin the surface temperature field over theoceans below 20° latitude. Whereas tem-peratures vary little, water vapor variesperceptibly throughout the troposphere.This results from annual changes in thelarge-scale tropical circulation and fromday-to-day changes caused by synopticweather systems.

From here onward, water in its atmosphericvapor and liquid forms (weather) domi-nates almost every topic. As Table 5-1

————————————————————————————————Table 5-1. Weather effects on electrooptical devices (Seagraves, 1989)

Visible and Thermal GuidanceWeather Type Near-IR Imagers Lasers Devices

Imagers

Clouds Severe Severe Severe SevereFog Severe Moderate Severe Moderate/severeRain/snow light Moderate Moderate Moderate Moderate heavy Severe Severe Severe SevereSmoke/dust/sand heavy density Severe Moderate Severe Moderate/severe moderate density Moderate Moderate Moderate ModerateHigh humidity Low Moderate Low Low/moderate

many publications, such as AWS ClimaticBriefs, Crutcher and Davis (1969), numerousUSAF Environmental Technical Applica-tions Center (USAFETAC) documents andthe eleven volumes of the US Navy MarineClimatic Atlas of the World.

Air over the tropical oceans has a meantemperature of about 82°F (28°C) through-out the year. By contrast, over the subtropi-cal continents in summer, means exceed90°F (32°C). Over the open tropical oceans,surface air temperature is high and standarddeviation small. This is especially so overthe western oceans, where upwelling isinsignificant. Large temperature gradientsnear the east coasts of North America andAsia arise from nearby warm ocean currents— the Gulf Stream and the Kuroshio Cur-rent flowing northeastward from Taiwanalong Okinawa and the coast of Japan as faras 35°N. Small shifts in wind direction overthese currents may lead to considerableheating or cooling of the air by the sea.Large values of the standard deviationresult. Persistent oceanic anticyclones causeequatorward winds to blow along the westcoasts of the continents (Figure 4-1) wherecold sub-surface waters upwell and the air iscooled. In general, mean air temperaturediffers insignificantly from mean sea surfacetemperature, being noticeably cooler onlyoff coasts that experience frequent polaroutbreaks. The heating (cooling) of subtropi-cal continents closely follows the poleward(equatorward) movement of the sun. Re-gions are warmest when the sun is mostnearly overhead and coldest when it isfarthest away. On the other hand, because ofthe sea’s large specific heat, SST changes lagthe sun’s movement by one to two months.Over tropical lands, the mean temperaturedistribution is influenced by land/sea con-trasts along coasts and lake shores, elevationdifferences, cloudiness and other localeffects. To illustrate typical surface tempera-tures and annual variations over tropicalcontinents, mean monthly temperatures are

shown for representative stations in Figure5-2. The stations lie along approximatelymeridional profiles across Africa, East Asia/Australia, and the Americas. The regions arealike in that the annual range is large in thesubtropics and small within 10° of theequator, amounting to 42°F (23°C) at Shang-hai and 2°F (1°C) at Singapore. In higherlatitudes, the annual variation approximatesa sine curve, with the extremes generallyoccurring in January and July. The curvesare more complex in lower latitudes, withdouble maxima and minima at some sta-tions. This results from the sun passingdirectly overhead twice a year and fromcorresponding changes in humidity andcloudiness. In low latitudes, highest tem-peratures are experienced during the sunnymonths just before the rainy season starts.

2. Upwelling/Downwelling. Wind blowingacross open water produces a stress on thesurface (see Sadler et al., 1987) that causes acurrent to flow in a direction turned 90°anticyclonically from the wind direction.That is to say, in the NH (SH) there is asurface stress to the right (left) of the wind.Therefore, when the wind blows parallel toa coast, surface water is forced either awayfrom or toward the coast, depending on thesense of the wind. Outward flow causeswater to upwell along the coast (Figure 5-3).If the upwelling extends below the surfacemixed layer and into the thermocline, coolerwater will be brought to the surface. Thus, anorth wind along the west coast of NorthAmerica, or a south wind along the eastcoast of Somalia, will produce cold-waterupwelling. The wind stress associated withupwelling also causes water to accumulateoff the coast and increase the ocean pressurethere. In analogy to the atmosphere, a geo-strophic ocean current develops, roughlyparallel to the wind. Thus in almost allcoastal upwelling the geostrophic currentenhances surface cooling by advecting water

from higher latitudes. The exception is offSomalia in summer. There, the southwestmonsoon upwells cold water, but the geo-strophic current brings warmer water fromthe south. Even so, the net result is signifi-cant surface cooling. The large thermalgradient that develops across the coastcauses a low-level jet stream to develop; thesurface wind increases and so further en-hances the upwelling. The net result of thispositive feedback is the strongest, steadiestsurface winds in the tropics (Figure 4-8).

If the wind flow were reversed, the resultingcyclonically turning current would producewarm water downwelling. This occurs alongthe coast of China in winter, during thenortheast monsoon. However, the stronggeostrophic current generated by the windbrings cold surface water from the northand causes a net cooling at the surface.

The Coriolis force changes sign across theequator. Thus easterly winds, blowing alongthe equator, are divergent (Gordon andTaylor, 1970); wind stress forces surfacewater away from the equator and upwellingresults. Westerlies cause downwelling.

Figure 5-4 shows upwelling regions andcoastal deserts. Off western Australia, up-welling fails to lower the SST, probablybecause warm sub-surface water driftssouthward around Northwest Cape (22°S,114°E). The winds driving coastal upwellingundergo frictionally-induced divergencealong the coast (Chapter 2, Section D-2a).Over these same regions, large-scale subsid-ence, associated with subtropical anticy-clones or with the Asian summer monsoonin the case of the western Indian Ocean,further contributes to relatively dry weather(Lydolph, 1973) and helps maintain coastaldeserts. Upwelling persists through the yearoff the west coasts of the Americas andsouthern Africa, and also off the north coastof South America. Since the water in theseregions is normally colder than the land, theannual reversal of the temperature gradient,

needed for monsoons to develop, cannotoccur.

Along the west coast of South Americaduring El Niño (see Chapter 6, Section C-4),cold water in the thermocline is too deep tobe reached by upwelling, and SST becomesanomalously high.

3. Elevation Effects. Over tropical conti-nents, away from the coasts, the meanmonthly surface temperatures show a regu-lar decrease with elevation of about 1.2°C1000 feet-1 (4.5°C km-1). With elevation, dewpoint decreases less than temperature. Thisresults in higher mean relative humiditiesand more fog at higher stations. Whencombined with rugged terrain this can poseserious problems for aircraft operations.Near the coast, the patterns of surface tem-perature and dew point related to elevationmay be influenced by land and sea breezes,marine inversions, frictionally-induceddivergence and other local effects.

C. Upper-C. Upper-C. Upper-C. Upper-C. Upper-Air TAir TAir TAir TAir Temperature and Wemperature and Wemperature and Wemperature and Wemperature and WaterateraterateraterVVVVVaporaporaporaporapor.....

1. Means. The mean structure of the upper-air temperature and humidity fields isillustrated by constant-pressure-level chartsand mean vertical soundings. Figure 5-5shows the mean 700 mb temperature andrelative humidity for winter and summer. Inthe NH winter, 700 mb temperature exceeds8°C over most of the tropics below 20°.Except over East Asia, the mean horizontaltemperature gradients are small. The 700mb 8°C isotherm corresponds to a freezinglevel of about 14,000 feet (4270 m). In theNH subtropical high-pressure ridges, 700mb relative humidities are below 30 percent.Mean relative humidities above 70 percentare confined to bad weather areas near theequator.

In the NH summer, the isotherm patternshifts northward with the sun, and warmcenters (exceeding 12°C at 700 mb) appearabove pronounced surface heating oversouthern Asia, North America and NorthAfrica. 700 mb relative humidities are high-est in regions of deep cloud, and lowest inthe subtropical ridges.

Changes in temperature and relative humid-ity between winter and summer are bestillustrated by difference maps (Figure 5-6).Large temperature changes occur over theNH mid-latitude continents, especially EastAsia. Equatorward of 20° the temperaturechanges are less than 4°C and often less than2°C. Since patterns of mean monthly relativehumidity resemble those of mean monthlycloudiness and rainfall over much of thetropics, changes in all three patterns occur inphase. In general, 700 mb relative humidi-ties decrease in the SH tropics and increasein the NH tropics from January to July.

2. Mean Precipitable Water. Precipitablewater is defined as the depth of liquid waterthat would be obtained if all the water vaporabove a unit area of the earth’s surface werecondensed. Thus it indicates total moisturecontent within the column of air having across sectional unit area. Tuller (1968) pre-pared mean monthly and annual maps ofthe world distribution of precipitable waterbased on the years 1964 through 1966; hismaps for January and July are shown inFigure 5-7. In January, mean values in thetropics range from less than 0.50 inches (13mm) over North Africa to more than 1.75inches (45 mm) over much of the low-lati-tude tropics, especially in the SH. Largegradients of precipitable water and eleva-tion coincide, as along the western Andes,the coast of eastern Africa and the southernHimalayas. Also note the large meridionalgradient over western Africa.

In July, the area with mean precipitablewater over 1.75 inches (45 mm) is largelyconfined to the NH. Values exceed 2.5inches (60 mm) near Bangladesh. Lowestvalues are found over Australia and south-ern Africa. The monthly maps of meanprecipitable water show a meridional shiftthat accompanies the sun over the conti-nents, and lags it by about two months overthe oceans. The annual range is small overthe equatorial oceans, high mountains, andhigher latitudes. The greatest annual range,more than 1.50 inches (38 mm) occurs overthe Bay of Bengal.

Stephens (1990) established a statistical linkbetween mean SST and mean precipitablewater over the ocean (Figure 5-8). Since therelationship is almost linear for SST > 20°C,the patterns of ocean surface air tempera-ture in Figure 5-1 and of precipitable waterin Figure 5-7 are alike. Knowing the SST, aforecaster can use Figure 5-8 to estimateprecipitable water.

3. Mean Vertical Soundings. The “Stan-dard Atmosphere”, adopted by the Interna-tional Commission for Air Navigation in1924 (List, 1963), formed the basis for “nor-mal” atmospheres (ESSA et al., 1967) for15°N and 30°N. These are representative ofthe tropics and subtropics, and were derivedby computing the zonal averages of pres-sure, temperature and humidity at standardpressure-levels. The mean vertical profilesof temperature and dew point for the sur-face to 100 mb are shown in Figure

5-9. For 30°N the months of January andJuly are given. At 15°N, where annual varia-tion is small, mean annual values suffice.The data are listed in Table 5-2.

In the 30°N normal atmosphere for January(Figure 5-9A), a fairly

stable lapse rate extends from the surface to800 mb, giving way to a lapse rate near that

of USSA above 800 mb. The mean tropo-pause lies near 200 mb. The dew pointdecreases a little with height below 900 mband more above. The 30°N normal atmo-sphere in July (Figure 5-9B) is much warmerthan in January. From the surface to 250 mbit is 14 to 17°C warmer than the USSA. Thetemperature lapse rate is large below 900mb, then slightly less than the USSA be-tween 900 and 500mb, and slightly moreabove 500 mb. The mean tropopause heightis near 130 mb. The July dew point lapserate is large below 900 mb and less above.

The 15°N normal atmosphere (Figure 5-9C)shows a temperature inversion between 800and 750 mb that is typical of the tradewinds.The inversion (or marked stable layer)separates the moist, well-mixed lower layerfrom drier air aloft. So, dew point decreasesmarkedly with height through the inversionor stable layer. Because the height, thickness,and intensity of the inversion often changequickly, it would normally be lost inmonthly and zonal averaging at constantheights. This is avoided by averaging dailyvalues of the inversion height, thickness and

Table 5-2. Normal atmospheres for 30°N in January and July and 15°N Annual.

30°N January 30°N July

P height T Td RH height T Td RH kfeet km kfeet km1000 0.57 0.17 13.5 12 80 0.39 0.12 27 23 78850 5.1 1.5 9.5 2 60 5.1 1.5 18 10 60700 10.3 3.1 1 -10 45 10.4 3.2 9 1 60500 18.9 5.7 -17 -29 30 19.3 5.9 -6 -17 40400 24.1 7.3 -27 -39 30 24.9 7.6 -18 -28 40300 31.5 9.6 -40 31.8 9.7 -33 -44 30250 34.4 10.5 -48 36.1 11.0 -42200 39.4 12.0 -57 40.7 12.4 -52150 45.9 14.0 -65 46.6 14.2 -62100 53.5 16.3 -69 54.5 16.6 -6970 61.3 18.7 -69 62.0 18.9 -6450 67.3 20.5 -64 68.9 21.0 -5930 76.4 23.3 -57 79.7 24.3 -5220 86.0 26.2 -51 88.6 27.0 -4710 100.1 30.5 -43 101.0 30.8 -39

15°N Annual

1000 0.39 0.12 26 23 75850 5.1 1.5 17 13 75700 10.4 3.2 9 -5 35500 19.3 5.9 -9 -21 35400 24.9 7.6 -19 -33 30300 31.7 9.7 -33250 35.9 10.9 -43200 40.6 12.4 -53150 46.5 14.2 -64100 52.7 16.1 -8070 61.3 18.7 -7150 67.8 20.7 -6330 78.1 23.8 -5420 87.5 26.7 -4810 101.7 31.0 -37

Geopotential Height in thousands of feet and km. P = Pressure (mb) T = Temperature (°C) Td= Dew point (°C) RH = Relative Humidity (percent)————————————————

intensity, and then synthesizing the typicalprofile shown in Figure 5-9C. Gutnick (1958)studied the climatology of the tradewindinversion in the Caribbean, including theseasonal, spatial and diurnal variation of theinversion frequency, height, strength, thick-ness, etc. He found that above the inversion,the mean temperature lapse rate wasslightly greater than in the normal atmo-sphere. The mean tropopause height at 15°Nis 100 mb, typical of the tropics equatorwardof 20°.

Bangkok, Thailand (14°N, 100°E) andPapeete, Tahiti (18°S, 150°W) are representa-tive of the continental and oceanic tropics.Figure 5-10 shows their mean vertical tem-perature and dew point soundings for dryand wet season months. Above 850 mb atBangkok, July is only 1 to 2°C warmer thanJanuary; however, the dew points at alllevels are much higher. In July, the averagedew point spread from the surface to 300mb is only 4 to 6°C. At Papeete, the meanmonthly temperatures are within 1°C of theannual mean, so only that sounding isshown. As in Bangkok, dew points are muchlower at Papeete in the dry season (August)than in the rainy season (February).

Upper-air dew points vary much more fromday-to-day than do temperatures. Overmost of the tropical troposphereequatorward of 20° the standard deviationof temperature in any month is generallyless than 2°C. Much of this stems fromoccasional cold advection from middle-latitude systems in winter and from cycloniccells in the TUTT in summer, and fromdifferential radiation effects under cloudyand clear skies. Dew point standard devia-tions are two to three times larger. Thisresults from the upward and downward airmotions associated with meso- and synop-tic-scale changes acting on the relativelylarge vertical gradients of dew points.

4. Equivalent Potential Temperature. Theequivalent potential temperature ( e) hasbeen used by various investigators to de-scribe the structure of the tropical tropo-sphere. This element, which combines theeffects of temperature and moisture varia-tions, is conservative with respect to dry-and pseudo-adiabatic processes and isproportional to the total static energy in theatmosphere. The equivalent-potential tem-perature is defined as the potential tempera-ture that an air parcel would have afterundergoing the following processes (rarelyrealized in the real atmosphere): dry adia-batic expansion until saturation, pseudo-adiabatic expansion until all moisture isprecipitated out, and dry-adiabatic com-pression to 1000 mb. The atmosphere isdefined as potentially unstable if e decreaseswith height, and potentially stable if eincreases with height. Typical lapse rates ofe show that the lower part of the tropicaltroposphere is almost always potentiallyunstable.

Garstang et al. (1967) studied soundingsmade aboard a research vessel in the tropi-cal Atlantic. The soundings were groupedinto disturbed (cloudy with rain) and undis-turbed (fair weather) as determined fromhourly weather observations and cloudphotographs. The mean temperature pro-files differed little; however, the dew pointlapse rates showed large differences that inturn resulted in large differences in e lapserates (Figure 5-11). The mean e lapse rate forall days is typical of much of the tropics; inthe lower troposphere a large lapse ratereaches a e minimum between 600 and 700mb; e then increases with height. Duringdisturbed (undisturbed) days mid-tropo-spheric e becomes larger (smaller) due tothe increased (decreased) moisture aloft.Thus, the lower troposphere is more poten-tially unstable (larger decrease of e withheight) on undisturbed days. Garstang et al.(1967), in a similar study for Barbados(13°N, 59°W), where the soundings were

divided into five synoptic groups depend-ing on the percentage of stations reportingrainfall during a 24-h interval, also foundlarge differences in mid-tropospheric evalues between greatly disturbed days(more than 85 percent of stations reportingrain) and greatly suppressed days (less than16 percent of stations reporting rain).

Harris and Ho (1969) studied the virtualequivalent-potential temperature ( ev) forstations in southeast Asia as a measure ofatmospheric structure during periods ofincreased or suppressed convection. Com-pared to e, ev more accurately assigns tem-perature to each pressure level. e and evhave essentially the same vertical distribu-tions. The degree of convection was deter-mined by a Radar Index, which representedthe percentage of the area (within a 50 NM(100 km) radius of the station) covered byradar echoes. Daily values of this indexdefined four days with increased and fourdays with suppressed convection at Saigon(11°N, 107°E) during July 1966. When the evlapse rates were compared with the monthlymean ev lapse rates (Figure 5-11), the resultsconfirmed those of Garstang et al.

The above studies show that e varies charac-teristically with height and time in associa-tion with synoptic disturbances. Paradoxi-cally, the tropical troposphere is most poten-tially unstable when convection is sup-pressed and becomes more stable as convec-tion increases. The associated distinctlydifferent rainfall regimes will be discussedin Chapter 6. The cause of the mid-tropo-spheric increase in e is the upward transportof sensible and latent heat. Thus, changes ine lapse rates (determined mainly by mois-ture variations in the mid-troposphere) andstability are often the result rather than thecause of the increased convection associatedwith tropical disturbances. This partiallyexplains why conventional stability indexesdeveloped for mid-latitudes are of limited

use in forecasting convection over much ofthe tropics.

Soundings made even in disturbed condi-tions show e increasing above 600 mb. How,then, can heat energy be exported fromsurface to high troposphere to fuel theHadley cell? The “hot tower” hypothesis ofRiehl and Malkus (1958) (see Chapter 2,Section B) provides the answer. Within thecentral region of a huge cumulonimbusundiluted by entrainment (Figure 6-10), eremains nearly constant with height andheat is transported upward. This very rea-sonable hypothesis is still unconfirmed,because no accurate sounding has yet beenmade in the core of a giant tropical cumu-lonimbus. Gray (1968) determined theaverage vertical distribution of e over thetropical oceans for about 15 years of the twowarmest summer and the two coldest win-ter months. Figure 5-12 shows the distribu-tions for the regions where tropical cyclonesform. Gray defined the term “potentialbuoyancy” of cumulus in the lower tropo-sphere as the difference in e between thesurface and 500 mb. The average potentialbuoyancy is large over all tropical oceans,decreases with increasing latitude, becomesnegative in higher latitudes, and is consider-ably less in winter than in summer. As aresult cumulonimbus are less intense andrarer in winter. The difference in e betweenthe top and bottom of an inversion is note-worthy. Where there is a solid stratus over-cast, e increases with height. A change to adecrease with height from the bottom to thetop of the inversion indicates instability andprecedes stratus breakup and eventualdevelopment of scattered cumulus.

5. Tradewinds. The tradewind inversion hasbeen extensively investigated since it wasfirst mapped over the Atlantic Ocean by theMeteor expedition (von Ficker, 1936). Riehl(1979) summarized his and others’ work.The inversion marks the lower boundary ofair subsiding in the oceanic subtropical

anticyclones. Figures 4-6 to 4-9, 5-1, and 6-1to 6-4 reveal that surface air in the anticy-clones first flows toward the western coastsof Africa and the Americas, where it iscooled and stabilized by the upwellingwater. The inversion may exceed 5°C. If theinversion stays above the surface, cooling bythe sea may reduce the surface air tempera-ture to the dew point and cause sea fog toform. At times, the inversion may reach thesurface, and clouds are absent or scatteredin the dry, subsiding air. Farther down-stream, winds first parallel the coast, andthen head out to sea. During this time, SSTincreases along the trajectory, and the airreceives considerable heat and moisturefrom below. The resulting convection raisesthe inversion to between 1500 feet (500m)and 3000 feet (1km). Since this is above thecondensation level, extensive stratus orstratocumulus develop at the inversion base.An average aerological sounding for thesouthern California coast in summer (Figure5-13) (Neiburger et al., 1961) shows not onlyan intense inversion, but e increasing from312°K at the base to 323°K at the top, indi-cating strong potential stability. No ex-change takes place through the inversionand the cloud top is uniformly flat. Wylie etal. (1989) found that in July 1987, stratocu-mulus increased 500 to 1300 NM (1000 to2500 km) west and southwest of Californiawhen the air was being heated and moist-ened from below. Deep precipitating cloudsare absent (Garcia, 1985).

Farther into the North Pacific, when the airreaches about 20°N, the inversion has beenraised above 5000 feet (1500 m), and heatingby the sea surface has diminished. A typicalsounding for Lihue (22°N, 159°W) (Figure 5-13) shows e decreasing from 319°K to 315°Kthrough the inversion. This potential insta-bility aids evaporation, as clouds penetratethe inversion, and as dry air subsidesthrough the inversion. Tradewind cumulus,typical of this regime, is scattered; tops areat various heights and often wispy. The

inversion height changes little as thetradewinds approach the equator (Kloeseland Albrecht, 1989).

Riehl has pointed out that the inversion isnot an impermeable boundary. Convectionfrom below and subsidence from abovecombine to mix moist rising air with drysinking air. Thus, in summer, surface rela-tive humidity is about 75 percent over thecentral Pacific compared to 85 percent atHong Kong (22°N, 114°E), in prevailingsouthwesterlies, where subsidence is weakor absent. Consequently, cloud bases thereare about 500 feet (150 m) lower than in thetradewinds.

Over the southeast tropical Pacific andAtlantic as the tradewinds approach thecooled upwelled water along the equator,cooling and divergence stabilize the air,causing shears in the vertical to reach 16knots in the lowest 300 feet (100 m) (Wallaceet al., 1989), while lowering the inversionand reducing clouds. The southeasttradewinds then cross the equator andconverge with the northeast tradewindsbetween 5 and 10°N, where a persistentcloud band — the near-equatorial conver-gence zone (NECZ) — is found. The charac-ter of the NECZ will be discussed in Chap-ters 6 and 8.

A satellite picture of the eastern Pacific(Figure 5-14) shows typical tradewindclouds. Figure 5-13 shows that over coastalupwelling the stratus cloud top has aboutthe same temperature as the sea surface, andcan seldom be distinguished on satellite IRimages (see Figure 6-7). Since the tradewindinversion limits vertical motion, it alsoaffects horizontal motion when thetradewinds impinge on mountains that areas high or higher than the inversion. Satel-lite images of cloud swirls that developleeward of Guadeloupe Island (29°N,118°W) have often been reproduced (see forexample, Dvorak and Smigielski, 1990). Anisland obstacle slows the flow and causes

convergence some distance upwind, accel-eration around the edges of the obstacle andswirls to leeward. A Space Shuttle photo-graph, looking eastward across the island ofHawaii (Figure 5-15), shows lines of cumu-lus paralleling the windward coast andcyclonic and anticyclonic vortexes to lee-ward. This pattern accounts for very lighttradewinds along the windward coast,strong winds around the northern andsouthern ends of the island, and a well-developed sea breeze along the leewardcoast (Figures 7-6 and 7-16). The CentralAmerican mountain chains also block thetradewinds, except where they are funneledthrough three isthmuses (Figures 8-35 and8-36). In the strait between Taiwan andChina, the northeast monsoon is similarlyfunneled.

In winter (as mentioned previously),thesubtropical oceanic anticyclones combinewith the continental anticyclones to form acontinuous high pressure ridge. Thus thenortheast winter monsoon across the SouthChina Sea and the southeast winter mon-soon across the Timor Sea are, in all impor-tant aspects, tradewind regimes, with sub-sidence aloft and an inversion between 5000and 10,000 feet (1.5 and 3 km).

6. Low-Latitude Westerlies. Equatorwardof heat troughs, surface winds with a west-erly component prevail. Most notable is thesouthwest monsoon of the NH summer,extending from West Africa to east of thePhilippines and between 5°N and 25°N overthe Indian Ocean. Away from the heattrough, surface westerlies generally con-verge, and underlie upper-troposphericdivergence. Most of the troposphere is moist(see Figure 5-10A). Where orographic liftingoccurs, as over the western Ghats and theKhasi Hills of India, and the Arakan Moun-tains of Burma, summer rainfall exceeds 200to 300 inches (5000 to 7500 mm).

Year-long heat lows cause persistent surfacewesterlies in equatorial West Africa andwestern Colombia. On windward mountainslopes, annual rainfall may reach 500 inches(12,500 mm) (Walters et al., 1985).

Chapter 6Chapter 6Chapter 6Chapter 6Chapter 6CLCLCLCLCLOUDINESS AND RAINFOUDINESS AND RAINFOUDINESS AND RAINFOUDINESS AND RAINFOUDINESS AND RAINFALLALLALLALLALL

A. General.Tropical cloudiness and rainfall have beenintensively investigated in recent years byuniversity groups at Texas A & M, FloridaState, Hawaii, Colorado State, and others. Inthis chapter, the results of these investiga-tions are summarized. The distributions andvariations of cloudiness and rainfall arepresented for various scales from the large-scale global patterns to the convective-scalepatterns associated with individual rain-storms.

B. Cloudiness.

1. Means. Various investigators have usedmeteorological satellite data to derive themean cloudiness over the tropics. Sadler etal.(1984) used operational nephanalysisprepared by the National EnvironmentalSatellite Service (NESS) to determine meanmonthly cloudiness between 30°N and 30°S(Figures 6-1 to 6-4). The following introduc-tion to the broad-scale features of tropicalcloudiness should be supplemented by acareful study of all the monthly charts in theoriginal publication, and a comparison ofthese charts with the mean circulationfeatures, especially those at the gradientlevel.

Over most of the tropics, cloudiness andrainfall are directly related. Figures 6-1 to 6-4 show that regions of average deep convec-tion are also cloudy. As mentioned in Chap-ter 5, this is not true over the eastern parts ofthe oceanic subtropical anticyclones. There,where the tradewinds first blow along thecoast and then head out to sea, they areheated and moistened from below, and

extensive stratus and stratocumulus formbeneath the tradewind inversion. Thecloudiness maximum extending southeastfrom New Guinea persists throughout theyear. The largest gradients are found east ofTibet, south of the North African desertsand near upwelled cold water.

Clouds are orographically enhanced towindward, and depleted to leeward of thePhilippines, Madagascar and CentralAmerica (see Figure 7-41).

2. Annual Variation. The largest annualvariation occurs over south Asia between70°E and 110°E, where desert-like condi-tions in winter give way to the wet summermonsoon. The oceanic cloudiness minimaassociated with the subtropical highs, andthe near-equatorial maxima, movemeridionally following the sun, with a lag ofabout three months. Except where thetradewinds converge, maximum cloudinesscoincides with surface westerly winds onthe equatorward side of a surface pressuretrough. Clouds are relatively less along thetrough, which is anchored by a surfacetemperature maximum. Over the continents,this summer heat trough is replaced inwinter by an anticyclone that is also almostcloud-free — hence the deserts (see Figure6-29).

Apart from a brief “Indian summer” inAugust and September, south China re-mains cloudy (see Figure 6-29).

In the eastern Pacific, an east-west cloudi-ness maximum is often observed along 5°Sbetween February and April. The effect mayextend to

South America. Convergence within thesoutheast tradewinds causes it. Very fewship observations and no research flightshave inhibited study.

Climatologically, the cloudiness of tropicalcyclones is counteracted by subsidence-generated clear skies on their peripheries.Hence, average cloudiness is not affected.

3. Inter-annual Variations. Sadler (1969)found that for any calendar month thegeneral locations of cloudiness maxima andminima vary little from year to year; how-ever, within these regions, cloudiness maychange significantly from year to year. In ayear when some maximum regions haveabove-normal values there is a tendency forcompensation by below-normal values innearby minimum regions, suggestingchanges in the intensity of the vertical mo-tion cells linking maxima and minima. Thisoccurs during El Niño (see Section C-4).There is little inter-annual change in thetotal average cloudiness for the wholetropical belt.

4. Zonal Averages. Averages derived fromSadler et al. (1984) are shown in Figure 6-5.Between 5°N and 5°S cloudiness changeslittle throughout the year, with higher val-ues to the north along the oceanic NECZ. Arelative minimum along the equator reflectsincreased stability due to cool upwelling inthe Atlantic and eastern Pacific. In the NHsummer, the south Asian monsoon movesthe cloudiness maximum north and intensi-fies it. Compared to the NH, annual SHvariation and meridional gradients ofcloudiness are much less and total cloudi-ness more. In this ocean hemisphere, moremoisture is available, the ocean buffers theannual cycle, and stratocumulus persists inthe eastern parts of the tradewind regions.

5. Large-scale Organization of Deep Tropi-cal Cloud Systems. A special Study Groupon Tropical Disturbances established as partof the Joint Organizing Committee of GARP(1969) made a census of cloud systems overthe tropics. Daily satellite pictures for 1967were used to determine the typical large-scale organization of deep clouds in variousparts of the tropics. Three major types wereidentified and labelled cloud clusters, mon-soon clusters, and “popcorn” cumulonim-bus.

A cloud cluster covers an area of 200 to 1200km on a side, and consists of numerouscumulonimbus (CB) cells whose tops areseen as bright patches from which cirrusstreamers emanate. Clusters are usuallyarranged in bands several thousand kmlong. Monsoon clusters are larger than cloudclusters, extending from 500 to 1000 kmnorth - south and 500 to 2000 km east west.They are distributed over the land andadjoining seas of southern Asia during thesummer (southwest) monsoon, and in mostcases are not aligned in bands. Bright elon-gated areas of CB activity occur within themonsoon clusters.

Distinguishing between cloud clusters andmonsoon clusters is not easy, and may notbe important. The term “cluster” containslittle information on the nature or cause ofcloud aggregation, and reflects meteorolo-gists’ tendency to substitute labels for expla-nations. The satellite image for 19 July 1985(Figure 6-6) shows clusters of various sizesand orientations. Over the whole area, amoderate southwest (summer) monsoonprevailed. The cloud system oriented south-west-northeast across southeast China andthe Ryukyus might comprise cloud clusters,while that covering the Philippines could betermed a monsoon cluster. Apart from avery weak 500 mb low over the Ryukyus,the clusters were not readily related tosynoptic features. This was also true of agroup of Atlantic clusters between 8°N and

14°N, studied during GATE (Martin et al.,1984). “Popcorn” cumulonimbus comprisesa few CB cells, covering an area of about10,000 km2. The cells undergo pronounceddiurnal variation; in their maturity duringthe afternoon, a collection of these clustersresemble (in satellite pictures) randomlydistributed kernels of popcorn. This distincttype occurs mainly over tropical continents,where there is sufficient low-level moistureand weak wind shear in the vertical. Thesehave been labeled showers (see Chapter 1,Section B). Over northern South America, inthe relatively undisturbed conditions shownin Figure 6-7, popcorn cumulonimbusdeveloped between 11 and 16 LT. In thetropical Pacific, most deep clouds occur asclusters of numerous cells. They make upthe near equatorial convergence zones thatare clearly revealed in maps of monthlyaverage cloudiness (Figures 6-1 to 6-4). Themean NH cloud band is prominent through-out the year (Figure 6-5), being nearest theequator over the central Pacific and farthestfrom it over the western and eastern Pacific.The SH near-equatorial mean cloud band isprominent only from October throughMarch. Within each band the space betweenclusters is about two to three times the sizeof a cluster. Clusters in the eastern andwestern North Pacific often show a vorticalstructure. Throughout the year in the SouthPacific, a prominent band of cloud clusters(the South Pacific Convergence Zone) ex-tends from near 10°S, 160°E to 30°S, 120°W.In this band, the space between cloud clus-ters is less than in the equatorial regions.(See Chapter 8 for a more detailed discus-sion of these bands).

Martin and Karst (1969), in a detailed statis-tical study of deep cloud systems over thetropical Pacific (25°N - 25°S, 160°E - 100°W)used satellite photographs from March 1967to February 1968. They identified four typesof cloud systems: oval, line, wave, andspiral-vortex. The cloud systems werecategorized on the basis of satellite images;

accompanying circulations were almostnever identified. Only the spiral vortexescould with some confidence be associatedwith tropical cyclones. The oval type oc-curred over 500 times a year, and the othertypes between 100 and 200. From near zeroat the equator frequencies of all types in-creased to maxima at 5-10° latitude. Ovals,waves, and vortexes were two-thirds morecommon in the NH than in the SH. Thevortex and oval types extended over about150,000 NM2 (500,000 km2), while the waveand line types were about 50 percent larger.The vortex type lasted longest (about 6days) and the oval type was the briefest(about 2 days). Near the equator, the cloudsystems generally moved westward, whilepoleward of 20° they moved eastward withsome poleward component. Especially inthe NH, there were significant annual andareal differences in the frequency, distribu-tion, and movement of the cloud systems.They were larger and commoner in thewestern part of the area studied.

Martin and Suomi (1972) studied cloudclusters over the tropical Atlantic for thesummers of 1969 and 1970. The averagecloud cluster lasted longer in the Atlanticthan in the Pacific, possibly because theAtlantic was studied for two summers, andthe Pacific for the whole year. Clusters werelarger in the east and central Atlantic than inthe west, perhaps because many clustersoriginated well within Africa as squall lines,before moving westward across the ocean.

Hayden (1970) measured cloud-clusterdimensions and spacing in the tropicalNorth Pacific. He used a computer programto objectively analyze satellite-derivedbrightness fields for July and October 1967and January and April 1968. He concen-trated on the 500 NM (1000 km)-wide zoneof maximum brightness between 110°W and130°E where the predominant cloud-clusterwidth ranged from about 150 NM (280 km)in winter to about 250 NM (460 km) in

summer. Clusters were most often either 360to 480 NM (670 to 890 km) apart or 600 to720 NM (1100 to 1330 km) apart, with littleannual variation. For comparison, Haydenapplied the technique to 30°N to 60°N in thePacific for October 1967. The mid-latitudeclusters were usually about twice as large asthe tropical clusters. Assuming a relation-ship between cloud-cluster width and thesize of associated circulation features,Hayden concluded that a grid size suitablefor tropical analysis should be about halfthat used in mid-latitude analysis. Figure 6-8 shows the frequency distributions ofcloud-cluster widths and separation foundby Hayden. The relative dominance (ex-pressed as a percentage) is computed byweighting the frequency of a cluster widthby its area and dividing by the total area ofall cloud clusters. According to Hayden, the600 to 720 NM separation distance corre-sponds to the principal synoptic scale of thetropics; whereas, the shorter 360 to 480 NMseparation corresponds to a faster-movingsecondary scale that is enhanced by interac-tion with the larger scale.

In a different approach, Williams and Gray(1973) classified the stages of cloud clusterdevelopment. They analyzed the character-istics of the wind, thermal, and moisturefields surrounding satellite-defined mesos-cale cloud clusters in the tradewinds of thewestern North Pacific (0°N - 30°N, 125°E160°W) for October 1966 to October 1968.They identified five categories of cloudclusters (pre-storm, developing, conserva-tive, non-conservative or developing-dying,and dying) 3 to 6° latitude wide. Clear areas,at least 10° latitude wide, was added as asixth category. Altogether, 1257 individualclusters and 553 clear areas were identified.For each category, they composited rawin-sonde data relative to the category center for16 levels from the surface to 30 mb, andfrom them, they derived fields of wind,vorticity, divergence, kinetic energy, tem-perature, moisture, stability, and isobaric

heights. The results support physical rea-soning. For example, the composited windfields showed significant low-level cyclonicshear across the cloud clusters, with pre-storm clusters having the largest shear; andanticyclonic shear across the clear areas.Since little curvature was evident in thecomposited low-level wind field, the rela-tive vorticity was produced primarily by thehorizontal wind shear. This suggests ab-sence of easterly wave-like disturbances (seeChapter 8-B-1). In the clusters, net conver-gence in the lower troposphere (below 400mb) was overlain by net divergence in theupper troposphere (resulting in upwardmotion); the reverse was true for the clearareas. Over the cloud-cluster centers, aver-age vertical wind shear (900 to 200 mb) wasrelatively small (< 10 knots). Because theauthors did not distinguish between clustersassociated with lower-tropospheric circula-tion systems and those associated withupper-tropospheric systems, the compositeenvironmental fields surrounding eachcategory of cluster could not be distin-guished. Twenty-seven percent of all theclusters observed on any one day disap-peared within 24 hours.

The organization of tropical Atlantic cloudsystems resembles those of the Pacific;however, a SH near-equatorial band is muchless evident. From near the equator in theNH winter, the NH band shifts to near 10°Nin summer. Then, circulations associatedwith clusters in the North Atlantic occasion-ally develop into hurricanes. In the SouthAtlantic, a curved band of cloud clusters,similar to the one in the Pacific, extendssoutheastward from near 20°S, 40°W intomid-latitudes.

Over the Amazon Basin, cloud clusters andmonsoon clusters predominate in the Janu-ary-May wet season (Greco et al., 1990),while in the July-October dry season, pop-corn cumulonimbus are common. Thepopcorn are rather randomly distributed

over the plains and mountains, and varydiurnally; cumulus develop in the forenoon,become thunderstorms during the after-noon, and dissipate in the evening. Overtropical Africa, clouds are organized intolarge clusters similar to those of the Atlanticand Pacific, and usually accompany squalllines (Figure 8-28). Generally the clustersfirst appear over the East African mountainsand move westward at about 250 NM (500km) day-1. During the NH summer, theytravel between 5°N and 10°N, and maydevelop into vortexes before emerging intothe North Atlantic. Although the vortexesgenerally weaken as they move across theAtlantic, some redevelop into hurricanes inthe western North Atlantic. During the SHsummer, cloud clusters are found overAfrica between 5°S and 10°S.

Over the Indian Ocean and southeast Asia,cloud systems both resemble and differfrom those in other areas. In the SouthIndian Ocean they are organized into largeclusters similar to those in the Pacific. Theyare most active in the SH summer, movingwestward along 10°S to 12°S. The accompa-nying circulations cannot be identifiedunless the clusters become vortical. Clustersalso occur during the SH winter near 5°S.

During the NH winter, over the maritimecontinent intense cloud clusters develop,generally in response to surges of the wintermonsoon. The clusters undergo strongdiurnal variation and are characterized byextensive stratiform cloud, the tops of whichappear as cirrus shields in satellite pictures.Rain comes about equally from embeddeddeep convective cells and from the strati-form cloud (Williams and Houze, 1987). Asimilar distribution has been observedelsewhere in the tropics, and in middlelatitudes (Houze, 1989).

During the NH spring (April, May) cloudclusters are generally confined between 5°Nand 10°N. Onset of the southwest monsoonis accompanied by a dramatic change. In

late May and early June, cloud clusters lienorth and south of and occasionally straddlethe equator. By mid-June one or two largecloud systems extend west to east to covermost of the area between 70° and 90°E and10° and 20°N. These large cloud systems,called “monsoon clusters”, occasionallyreach as far east as 110°E. Smaller monsoonclusters sometimes occur over the ArabianSea (50°E to 70°E, 5°S to 15°N). Many of thecumulonimbus tops embedded in theseclusters appear in satellite pictures.

Based on the results of the GARP StudyGroup, a highly idealized picture of typicalcloud organization in the tropics emerges.Figure 6-9 shows the large-scale deep tropi-cal cloud systems on a typical summer dayin each NH and SH region. Many mesoscalecloud systems would occur within each ofthe larger systems. Smaller, independentcloud systems, which may appear in dailysatellite pictures, have been omitted. In viewof the great variety of daily patterns and theprevailing westward motion of low-latitudecloud systems, the complexity and variabil-ity of the tropical atmosphere is not surpris-ing.

Tropical forecasters have long known thelarge, organized cloud systems of tropicalcyclones and monsoon depressions. Theytended to think that other rain-producingclouds were rather scattered. Weather satel-lites have now shown that rain from cloudsin clusters contributes more than 80 percentof the total amount, that scattered convec-tion is relatively unimportant, and that fairweather usually covers a large area.

6. Tropical Cloud Types. In this section,tropical cloud types, their characteristics,and relative frequency are discussed. Muchof the information is condensed fromPalmer et al. (1955,1956) with additionalinformation from recent studies. The clouddefinitions are taken from the WMO Inter-

national Cloud Atlas (1956). Pictures ofindividual cloud types are readily availablein the Atlas and so are not presented here.

The international cloud classification isapplicable worldwide. However, the relativefrequency of various cloud species doesdiffer between the tropics and higher lati-tudes and some sub-varieties found in highlatitudes may not occur in the tropics(Malkus and Riehl, 1964). While all cloudspecies are found over both land and sea inthe tropics, oceanic and orographic cloudsmust be clearly differentiated. Far fromland diurnal variation is relatively small.Over continents and large islands, diurnalvariation is pronounced and must be care-fully considered in forecasting (See Chapter7). Operational satellite images are notdetailed enough to allow individual cloudelements to be readily identified.

a. Cumulus. Cumulus predominate in thetropics, and as discussed earlier, transportmuch of the energy required to drive theatmosphere. The following discussion of thephysics of cumulus is based on Simpsonand Dennis (1972). Atmospheric movementis driven by the sun, but indirectly — heat isabsorbed at the earth’s surface from which itis transmitted upward, mainly as watervapor. When cumulus develop, water vaporcondenses into liquid cloud droplets, turn-ing latent heat of evaporation into sensibleheat of condensation. The process is madeirreversible if rain falls out of the cloud; thenthe sensible heat can be used to drive airmotion.

Figure 6-10 illustrates a hierarchy of tropicalcumulus according to size. The bubble-likesmall cumulus or the plume or jet-like largercumulus are driven upward by condensa-tion and the resultant buoyancy of the cloudwith respect to the environment. For growthto be maintained, a small supersaturation isneeded to sustain condensation in the faceof vapor removal by the growing droplets.

Measurements showed that the temperatureand cloud liquid water in a cloud are lessthan what would evolve were the cloud tobe composed only of isolated parcels of airthat had risen from the earth’s surface. Asshown in Figure 6-10, unsaturated air isdrawn (entrained) into the cloud from thesurroundings. Compared to small clouds,the buoyant central core of large clouds isless affected by entrainment, and the cloudslast longer. An added complication is intro-duced by shear (Figure 6-11). Then, airentrains on the upshear side and detrains onthe downshear side of the cloud. The largerclouds of Figure 6-10 are more slowlyeroded by entrainment than the smallerclouds, since their central cores are betterprotected from invasion by dry air.

Condensation alone cannot produce rain-drops (0.5 mm or more in diameter), butrain does fall. Two processes may ensurethis. The more important is growth bydroplet collision, causing coalescence. Theother occurs in the parts of a cloud colderthan about -5°C. Because saturation vaporpressure is lower over ice than over water,the ice particles in the cloud grow fasterthan water droplets and sometimes growlarge enough to fall.

When drops grow large enough to fall froma cloud, do not evaporate in the air beneaththe cloud, and so reach the surface, rainoccurs. Within the cloud, the falling dropsscavenge a significant fraction of all clouddroplets. The relative humidity in the cloudis thus lowered, most of the remainingdroplets evaporate, and the cloud “rainsout”; by the time the rain stops, only a fewwisps may remain of the cloud from whichthe rain originated.

Although the mechanism is not clear, onlypresence of both ice and water in a cloudcan generate the large electric gradients thatcause lightning. So in the tropics, wherecloud must extend above 16,000 feet (4 km)for ice to form, thunderstorms are confined

to only very deep convective clouds thatgenerally extend above 30,000 feet (9 km).

When rain falls from a cloud it may, throughfriction and evaporation, induce a colddowndraft of air that reaches the surface.There, cooling due to cloud shade might befurther enhanced and the chance of newconvection reduced. Or the cold air, spread-ing outward and converging with surround-ing air (the gust front) may trigger upwardmotion and new convection.

Cumulus are detached clouds, generallydense and with sharp outlines developingvertically in the form of rising mounds,domes, or towers, of which the bulgingupper part often resembles a cauliflower.The sunlit parts of cumulus are brilliantlywhite; the shaded base is relatively dark andnearly horizontal. Four species of cumulusare defined in order of increasing verticaldevelopment: fractus or irregular shreds ofcloud having a ragged appearance, humilis,of slight vertical extent, mediocris of moder-ate vertical extent, and congestis of greatvertical extent. The most striking variationamong cumulus is the extent of lean orshear of the cloud (see above). On this basis,Palmer et al., (1955) divided cumulus intodoldrum cumulus which have little or nolean or shear and trade cumulus which havesignificant shear.

The doldrum class is typical of the lightwind and small wind-shear regions in lowerlatitudes. The base is slightly wider than thetop, with dimensions usually 3000 feet (900m) and 2000 feet (600 m) respectively. Thebase is usually 1500 feet (450 m) high, lower-ing to 1300 feet (400 m) or below in showers.The tops vary between 6000 and 12,000 feet(1800 and 3600 m). Winds are light and thereis little shear below 15,000 feet (4500 m). Asthe name implies, trade cumulus are mostcommon in the tradewinds between 10° and30° latitude. Wind shear through the cloudlayer causes the cloud axis to lean (seeFigure 6-11). In the tradewinds the wind is

usually strongest near cloud base and de-creases with height. This is in response to atemperature gradient directed to the right(left) of the tradewind flow in the NH (SH).The resulting thermal wind diminishes thetradewind. This effect, combined withfrictionally reduced speed at the surface,produces the typical tradewind profile, firstexplained by Sheppard and Omar (1952)and shown in Figure 6-12. The daily shearoften exceeds the mean shear. The base oftrade cumulus is usually near 2000 feet (600m). Tops vary with atmospheric structureand synoptic situation. The mean is near7400 feet (2.3 km) (Figure 6-13), but is com-monly much lower.

b. Cumulonimbus (CB) is a heavy, densevery tall cloud in the form of a mountain orhuge tower, usually accompanied by show-ers, thunder and lightning. The upper por-tion, which is usually smooth, fibrous, orstriated cirrus and almost flattened, oftenspreads out in the form of an anvil or vastplume. Satellite pictures sometimes showplumes from individual Cbs extending forhundreds of kilometers. Over tropicaloceans, Cbs are almost always restricted tosynoptic or meso-scale disturbances. Theyare much commoner over land, with somestations reporting thunderstorms on morethan half the days of the year (see Figures10-1, 2, and 3). Over the oceans, CB topsusually range between 30,000 and 40,000feet (9 and 12 km). But in the walls andfeeder bands of intense tropical cyclonesthey can reach 50,000 to 60,000 feet (15 to 18km). Over land the tops are somewhathigher than over the oceans. A radar studyof thunderstorms over India during thesouthwest monsoon by Ghosh (1967)showed mean heights at maximum develop-ment to be 43,000 feet (13 km), with anextreme of 56,000 feet (17 km). Studies overFlorida and Africa gave similar results. Cbstypically are from 1 to 10 NM (1.5 to 15 km)wide. Despite its size, an individual CB hasa short life; from development to dissipation

it seldom lasts more than two hours. Torna-does and waterspouts occasionally accom-pany CB in the tropics, but appear to be lesssevere than in mid-latitudes.

c. Stratus and Stratocumulus prevail overthe eastern tropical oceans dominated byinversions and where the air is beingwarmed and moistened from beneath.Stratus is a grayish uniform cloud with ahorizontal base. It usually occurs in exten-sive layers, but sometimes may be in raggedpatches (fractostratus) often caused whencooling and condensation from rain fallingfrom above saturates the air. Stratocumulusis a grey or whitish cloud occurring inpatches, sheets, or layers. Its rounded orroll-shaped elements may or may not bemerged and are usually arranged in orderlygroups, lines, or undulations. Stratocumulusis more common over the oceans withmoderate low-level winds and associatedmixing. Stratus is more often observedalong coasts at night and in the early morn-ing with light winds. Stratus is also commonalong coasts where warm moist air is ad-vected over cool coastal waters. It may reachthe surface and become fog.

d. Fog. Fog prevents most visual and nearinfrared electrooptical devices from operat-ing. Conditions favoring fog are the same inthe tropics as in middle latitudes.

Radiation Fog. Over flat open ground,nocturnal cooling in the tropics rarely low-ers surface air temperature to the dew point,and so fog does not result. However, underclear skies, air that has a local moisturesource and that is trapped in river valleysmay often be cooled enough at night for fogto form. The Congo and Amazon riverbasins are the most susceptible regions inthe tropics, experiencing 60 to 120 days offog a year, although a different mechanismtends to keep very wide rivers clear of fog(see Chapter 7, Section D-5).

Advection Fog. When warm, moist airmoves over water that is colder than thedew point, condensation usually occurs andfog forms. This occasionally happens north-west of the Gulf Stream and the KuroshioCurrent. The fog may be deep and persistent— Hong Kong (22°N, 114°E) once experi-enced a week of sea fog. Known locally as“Crachin”, it develops along the southChina coast in spring as the winter monsoonweakens and surface winds turn fromnortheast to southeast, advecting warmmoist air over still cold coastal waters(Ramage, 1954).

This mechanism leads one to expect fogover cold coastal upwelling. Fog usuallydevelops if sea surface temperature dropssufficiently along the air trajectory, and ifdry subsiding air does not reach the surface.It has been observed along the Ivory Coastin summer, and the Angola coast in winter,and along the northern part of the south-eastern Arabian coast in summer. Evenwithin upwelling regions, low stratus ratherthan fog may form, if surface air is locallywarmed along its track. For example, atGuayaquil (2°S, 80°W), visibilities less than0.5 NM (1 km) have almost never beenobserved (Edson and Condray, 1989) be-cause SST is about 6 to 8°F (3 to 4°C) higherthere than at the up-wind entrance to theGulf of Guayquil. At Lima (12°S, 77°W) onthe coast of Peru, ceilings below 300 feet(100 m) are rare. In fact, since surface windsgenerally blow up the SST gradient of theupwelled water, and parallel to the coast,fog is likely only when the wind directionshifts to onshore, and the air is then cooledfrom below.

Research vessel measurements along theeast coast of Somalia in summer 1964 neverrecorded visibility less than 5 NM (10 km),and only scattered low clouds, although airmoving over the cooling surface had, only aday before, a dew point 16°F (9°C) higherthan the temperature of the upwelled water.

Aerological soundings revealed that drysubsiding air had reached the surface and soprevented cooling from causing saturation(Figure 6-14). This is typical, for summer foghas never been reported off Somalia (Meteo-rological Office, 1949) (Figure 6-15). Tworeasons can be found. On a small scale, thedifference in surface stress between landand sea along upwelling coasts alwaysfavors divergence and so causes the air tosink (Chapter 2, Section D-2-a). On a largerscale, sinking is also produced upwind of aclimatological speed maximum in the sur-face southwesterlies (Figure 4-8). Whencombined with absence of convection as SSTdecreases downstream, these two effectsbring dry subsiding air down to the surface.Similar effects occur off other upwellingcoasts (see Figure 5-8).

To the northeast, along the coasts of Arabiaand Pakistan, climatological convergence inthe summer southwesterlies (Figure 4-7) liftsthe subsidence inversion above the surface,and morning stratus and drizzle are com-mon. Fog can develop where the air iscooled downstream (Figures 6-15 and 8-16).In fog-prone areas subtle differences insurface temperature gradients, low-leveldivergence, inversion height, vertical mix-ing, insolation and local winds may deter-mine whether cloud lies on the surface (fog)or a little above the surface (stratus). Fore-casting is important because surface-basedvisual and near infrared electroopticaldevices fail in fog, but can function under alow stratus deck. Surface observations candistinguish between fog and stratus, butsatellite pictures cannot.

Satellite IR images generally cannot distin-guish between stratus top and the oceansurface, and of course give no hint of ceil-ings. Forecasters can find the stratus invisible images. Using centrally-estimatedand disseminated SST charts, they need thento predict surface wind direction and esti-mate down-trajectory surface temperature

change before considering the chance of fog.A sounding can determine the base of thesubsidence inversion. Remember that evensmall topographical features, less than 150feet (50 m) high, may prevent sea fogspreading inland.

e. Altocumulus and Altostratus. In thetropics, these clouds are typically associatedwith disturbed weather. They may persist asresidues of “rained out” cumulonimbus.Altocumulus resembles stratocumulus instructure and color, but has smaller ele-ments and appears at higher altitudes (bydefinition, from 6,500 to 25,000 feet (2 to 7.5km)). Altostratus, in the form of grey orbluish sheets or layers of striated, fibrous, oruniform appearance, may be formed fromspreading of cumulus or cumulonimbus, ormay occur independently. Extensive alto-type clouds usually accompany tropicalcyclones. Independently-formed alto-typeclouds are usually found east of uppertropospheric cyclones or troughs. Nimbos-tratus, a thick grey cloud rendered diffuseby fairly continuous rain, may develop fromaltostratus under these conditions.

f. Cirrus and Cirrostratus. These consist ofice particles and may be formed from CBactivity (remnant anvils) or independently.Cirrus is a fibrous cloud composed of de-tailed elements in the form of white fila-ments, patches, or narrow bands. Cirrostra-tus is a denser cloud and more extensivethan cirrus, appearing as a whitish veilwhich may totally cover the sky and oftenproduce halo phenomena. Thin cirrus iscommon in the tropics. Independentlyformed cirrostratus is generally found withupper-level cyclones and on the south sideof subtropical jet streams. Cirrocumulusappears as thin white patches of cloudwithout shadows, composed of very smallelements in the form of grains or rippleswhich may be merged or separate, and bemore or less regularly arranged. It is rare inthe tropics.

Cirriform clouds are usually much higher inthe tropics than in mid-latitudes. Usingaircraft reconnaissance flights over Indiaduring the southwest monsoon, Deshpande(1965) found that the height of cirriformcloud bases averaged 38,000 feet (11.5 km),and 73 percent ranged between 35,000 and45,000 feet (10.5 and 13.7 km). Tops aver-aged 41,000 feet (12.5 km), with 50 percentbetween 40,000 and 45,000 feet (12.2 and13.7 km). Both the bases and tops were 4000to 5000 feet (1.2 to 1.5 km) higher in thesouthwest monsoon than in other seasons.These findings generally agree with double-theodolite observations made at Manila andDjakarta during the International CloudYear 1896-97 (Stone, 1957). They showedcirrus bases to range between 36,000 and41,000 feet (11 and 12.5 km), and to beslightly higher in summer than in winter.

Sadler and Lim (1979) compared cirrusmotion vectors calculated from sequentialgeostationary satellite pictures to windsmeasured by radiosondes. For the tropicaleastern hemisphere, differences were least at200 mb and 150 mb. This suggests that thecirrus usually lay between those levels(41,000 to 47,000 feet (12.4 to 14.2 km), inagreement with the earlier findings. In viewof this, and because of innumerable reportsfrom jet aircraft, those meteorologists whostill assign the mid-latitude altitude of25,000 feet to tropical cirrus should desist.

7. Relation of Summer Cloudiness to otherElements over the Tropical Atlantic. War-ren et al. (1989) have analyzed the frequencyand distribution of cloud types over theoceans, using ship data for the period 1952-1981. Extracts from their charts for June -August over the tropical Atlantic are repro-duced in Figures 6-16h to 6-16m. To aidinterpretation, August long-term means ofsea-level pressure (Figure 6-16a), resultantsurface wind (Figure 6-16b) and its diver-gence (Figure 6-16c), sea surface tempera-

ture (Figure 6-16d), outgoing longwaveradiation (OLR) (Figure 6-16e), frequency ofhighly-reflective clouds (Figure 6-16f), andcloudiness (Figure 6-16g) are reproduced.The circulation changes little between Juneand August, so comparisons are possible.

Cumulus is the commonest cloud, beingobserved more than half the time in theheart of the tradewinds. The clouds arescattered, with tops limited by thetradewind inversion, and the flow is diver-gent. Outgoing longwave radiation exceeds270 W m-2, signifying tops below 15,000 feet(4.5 km).

Stratus and Stratocumulus predominate inthe band of fresh tradewinds after theyleave the cool upwelled water along theNorth African coast and are heated frombelow. Subsidence limits the tops to about3000 feet (1 km) altitude.

Cumulonimbus, Altostratus and Altocumu-lus, and Nimbostratus are most common,and total cloud exceeds 5 oktas, above aband of convergent westerlies south of themonsoon trough. Here, OLR is <240 W m-2,and highly-reflective cloud commonest.Farther west, where the tradewinds con-verge, the effect is less pronounced.

Cirrus, Cirrostratus, and Cirrocumulus. Inthe bad weather region, lower clouds oftenobscure high clouds, and so these clouds arenot reported. Hence they appear to have alower frequency. Upper tropospheric east-northeast to northeast winds (Figure 4-11)would carry high cloud generated in the badweather toward the southwest, where lowerclouds are scattered and high clouds can beobserved more often.

Along the equator, surface winds divergemost and cloudiness is least. Cumuluspredominate and highly-reflective cloudsare absent. The southeast tradewinds nearthe equator are cooled from below andconvection dies down.

Vortexes moving westward from Africa aregenerally concentrated between 10°N and15°N (Figure 9-19), north of the bad weatherclouds that tend to be confined to the low-level westerlies south of the vortex centers.The cloud charts agree well with studiesbased on satellite pictures. For example,Duvel (1989) analyzed three summers, usingthe Meteosat geosynchronous satellite.

Forecasters should note that the surfacetrough and asymptotes of confluence (Fig-ure 6-16b) lie in a broad, weakly-definedarea of low pressure (Figure 6-16a), which isroughly located where the sea surface iswarmest (Figure 6-16d). Although flow isconvergent along the asymptotes (Figure 6-16c), maximum convergence, and the worstweather, coincide where winds decreasedownstream.

Figure 6-16 samples the range of tropicalclimatological charts now available to theforecaster. Growing interest in climate andclimate-change is largely responsible for thedevelopment of these charts, but synopticmeteorology is benefiting. Although thisdiscussion has been restricted to a relativelysmall area and a single season, its findingscan be applied to other tropical oceans thatexperience upwelling, tradewinds,tradewind convergence, monsoon troughs,and low-level westerlies.

C. Rainfall.C. Rainfall.C. Rainfall.C. Rainfall.C. Rainfall. In the absence of shadowing by orogra-phy, most significant rain in the tropics fallsin either of two circumstances. The firstcircumstance is from deep nimbostratuswith embedded cumulonimbus. Wind shearin the vertical and lower tropospheric con-vergence are both large, and although rainintensity may fluctuate considerably, skiesremain predominantly overcast. Lapse ratesare small (see Figure 5-9). The second cir-cumstance is from scattered towering cumu-lus or cumulonimbus, when vertical wind

shear and lower tropospheric convergenceboth are small. Lapse rates are relativelylarge. The term rains is assigned to theformer, and showers to the latter.

1. The Character of Significant Rain in theTropics.

a. Thunderstorm Frequency and Rain-fall. Heavy rain falls from thunderstorms.The average thunderstorm lasts about thesame time everywhere — seldom over twohours. Consequently, where thunderstormsare the overwhelmingly predominant rainproducers, the average number of thunder-storm days should be directly related to theaverage rainfall. Figure 6-17, derived fromPortig (1963), shows large areal variation inthe relation. It ranges from 16 over westernAfrica and southeastern Tibet to less than 2over China, southwestern India, and theocean. At the ends of the range, Portigidentified two types of rain regimes — theWest African, in which frequency of thun-derstorm days and rainfall increase tomaxima in midsummer; and the westernIndian in which frequency of thunderstormdays decreases as the rainfall increases.Indian meteorologists have long known thatperiods of maximum thunderstorm fre-quency precede and follow the summerrains (Ram, 1929). The rains, particularlyalong the west coast of India, seldom resultfrom thunderstorms.

Between Bombay and northern India, wherefour rawinsonde stations are located, thethunderstorm days/rainfall ratio spans bothof Portig’s categories (Figure 6-17). Thegradient can be studied by comparingmonthly averages of the ratio, rainfall, andthe lapse rate and wind shear between 850and 300 mb (Figure 6-14). Jodhpur and NewDelhi, with most thunderstorm days in July,are in Portig’s West African category: atBombay and Ahmedabad, typical of thewestern Indian category, thunderstorm daysare fewest in August. However, at all fourstations, the thunderstorm days/rainfall

ratios are least in midsummer when rainfallis greatest. From May to August at Bombay,where the rains are heavy, the ratio de-creases from 6.5 to 0.06, whereas at Jodhpur,on the edge of the Great Indian Desert, thedecrease is an order of magnitude less.Thus, in both categories, the trend towardmid-summer is the same. The averageamount of rain falling from a thunderstormis unlikely to increase spectacularly as theseason advances. Perhaps then, a differentmechanism begins to produce rain, becom-ing quite important in the north and over-whelmingly predominant in the south.

As summer advances, the monsoon circula-tion develops and intensifies over the twosouthern stations. Strong and generallyconvergent westerlies, underlying strongand generally divergent easterlies, lead tolarge-scale rise through the middle tropo-sphere, increased humidities, decreasedlapse rates to near moist adiabatic, and greatcloud sheets with embedded cumulonim-bus, from which considerable rain falls. Thereduced insolational heating of the surface,near-stable lapse rate, and large shear ofwind in the vertical all inhibit thunderstormformation and a rains regime prevails.

Over the two northern stations, the intensi-fying monsoon deepens the moist layer andso decreases lapse rate and hinders thunder-storm development. However, Jodhpur andNew Delhi, which lie near the heat troughand upper-tropospheric ridge axes, experi-ence decreasing shear of wind in the verti-cal, a trend that favors thunderstorms.During summer the upper troposphere isrelatively dry, the lapse rate is conditionallyunstable, shear is small, and insolation heatsthe surface — all favoring thunderstorms,but not extensive rain. A prevailing showersregime only rarely gives way to rains. Me-ridional profiles over western Africa re-semble those of India.

In the summer monsoon trough of centralIndia (Figure 6-19), rainfall is less than on

either side of the trough. Thunderstorms,which would be expected to be more fre-quent over the higher ground on either sideof the trough, are more frequent along thetrough. Thus, during the wet season, exten-sive fair weather is also thunderstormweather. Beneath scattered convective cells,showers may be heavy but average rainfallper unit area is only a fraction of that causedby the rains. Gentle subsidence favors fineweather since the accompanying inversionor isothermal layer limits convection. How-ever, an inversion can also enhance convec-tion (Fulks, 1951). Air beneath the inversionis moist and is heated during the day; airabove the inversion may radiationally cool.And so potential instability throughout thetroposphere increases. Then, if some ther-mal or orographic agent ruptures the inver-sion, thunderstorms may suddenly develop.

b. Continuous Thunderstorms. Rarely, whenrains are falling from thick nimbostratus, ahuge, stationary, persistent thunderstormdevelops within the cloud mass. It is appar-ently initiated and sustained by upper-tropospheric divergence (Sourbeer andGentry, 1961; Daniel and Subramaniam,1966; Chen, 1969) above geographicallyanchored surface convergence. Surface windis light and variable, with no gusts orsqualls. Radiation and heat divergence inthe upper troposphere must so rapidlydissipate the heat released by condensationin the rising air that the lapse rates neces-sary for a thunderstorm are maintained forhours. These systems have never beensuccessfully probed by radiosondes. Morethan 10 inches (250 mm) of rain may fall in aday, and floods result. Hawaii, India, Malay-sia, south Florida, Puerto Rico, easternIndochina, the Philippines, and south Chinahave all experienced continuous thunder-storms.

On 2 and 3 December 1978, 20 inches (512mm) of rain fell on Singapore (1°N, 104°E)during a continuous thunderstorm, the

heaviest fall since records started in 1869and amounting to 21 percent of the annualaverage. Seven people died and 3000 wereevacuated. The weather satellite showed anintensely reflecting blob of cloud over theisland (Figure 6-20). Objective kinematicanalysis of the 950 and 200 mb winds at 00UTC 2 December, are shown in Figure 11-23.The torrential rain started about 3 hoursafter. At 200 mb over Singapore, the NH andSH Hadley cell circulations were diverging.At the surface, flow converged into a weaklow.

Over south China, continuous thunder-storms associated with Mei-Yu fronts (Sec-tion 5d) gave 24 hour rainfalls of 33.5 inches(850 mm) on 29 May 1973 at Yangjiang(22°N, 111°E) and 34.8 inches (884 mm) on31 May 1977 at Haifeng (23°N, 115°E) (Taoand Chen, 1987) (see Figure 6-34).

Continuous thunderstorms are likeliest inspring or autumn, and often develop be-neath upper-tropospheric poleward diver-gent flow typical of winter adjoiningequatorward divergent flow typical ofsummer. Surface conditions give few clues -convergence into a weak low is generally allthat can be seen -a very common occurrenceto accompany such a rare event. Over southChina they are most common in May orJune during the Mei-Yu season. Over west-ern Malaysia, development tends to shiftsouth from 22 November to 11 December. Inthe cases studied, the area enclosed by the10 inch (250 mm) isohyet ranged from 436 to4360 NM2 (1500 to 15,000 km2). In somerespects, continuous thunderstorms re-semble the mid-latitude mesoscale convec-tive complexes described by Maddox (1983),although they are generally much smaller.

Giant persistent thunderstorms do not fitthe thunderstorm model (Byers andBraham, 1949). Continuous thunderstormscan be categorized as neither rains norshowers. They are so rare, that the best aforecaster can do is to issue a warning if an

active thunderstorm appears embedded inthick nimbostratus.

c. Significant Tradewind Rainfall. Over theopen ocean, where tradewinds and thewinter monsoon prevail, monthly averagerainfall seldom exceeds 2 inches (50 mm).However, on mountain ranges exposed tothe tradewinds, monthly totals of more than30 inches (750 mm) are not uncommon. Thiswas explained in terms of orographic uplift.Doubts arose (Ramage, 1978b) when analy-sis of tradewind rainfall at the top of Mt.Waialeale on Kauai (5,240 feet (1598 m),22°N, 159°W; annual average 433 inches(11,000 mm)) showed that strong tradewinddays with rainfall < .02 inches (5 mm) wereas common as those with rainfall >2 inches(50 mm). Similar tradewind inversions werepresent in both categories. Only light moun-tain rain falls in an environment of scatteredfair weather cumulus, while persistent,generally moderate rain falls when themountain lies beneath an area or line ofclouds extending to the upwind side. Then,cloud droplets grow along the air trajectory,unhindered by dry air entrainment. Whenthis cloudy air is lifted along the mountainface, prolonged light or moderate rain fallsat the mountain top.

On 14 and 15 October 1976, 7.32 inches (186mm) of uninterrupted rain fell at Waialeale(Figure 6-21). On the east coast, 10 NM (20km) away, 0.79 inches (20 mm) fell.Tradewinds were fresh and the inversionheight was 8000 feet (2.5 km). The cumuluscloud band that caused the rain was alignedparallel to the wind. It may have been ashear line (see Chapter 8, Section D-3), andpresumably marked slight local conver-gence. Mountains exposed to thetradewinds or to the winter monsoon in thePacific, the South Indian Ocean, the Philip-pines, Indochina, southeastern Indonesia,the Caribbean, and Central America, sharethis pattern. Maximum rain occurs along the

mountain slope about 2000 to 5000 feet (600to 1500 m) below the inversion base.

The NH and SH tradewinds converge overthe central and eastern Pacific and the cen-tral and western Atlantic (see Figure 6-16),and occasionally during spring and autumnover the western Indian Ocean. Cloudpersists along the convergence zone, wherelower-tropospheric flow parallels the zoneaxis. Thus cloud particles grow steadily byaccretion, and lifting raises the inversionand increases cloud depth, though not to thelevel of equatorial continental clouds (Fig-ure 6-22, Stowe et al., 1989). Rain can belight or moderate and longlasting. SeeChapter 8 for more discussion.

2. Mean. Rainfall varies greatly over shortdistances; even where dense rain gaugenetworks exist, constructing accurate isohy-etal charts is demanding and time-consum-ing (Meisner, 1979). Over the open ocean,rainfall measurements at coral islands areprobably representative, but they are tooscanty to allow confident interpolation. Wecan turn to the continuous field measure-ments made from meteorological satellitesto obtain a fair general picture of tropicalrainfall. Since deep clouds rain more thanshallow clouds, outgoing longwave radia-tion (OLR), which is inversely related to theheight of the emitting surface, is also in-versely related to rainfall. Figure 6-22 showsthe annual mean OLR for about 10 years.Where there are good rain gauge networks,OLR less than 220 W m-2 has been found tocorrespond to annual rainfalls greater than100 inches (2500 mm) and OLR greater than260 W m-2 to rainfalls less than 20 inches(500 mm).

Rain is heaviest close to the equator overAfrica, Indonesia and South America. Thetradewind regions and subtropical Africa,Australia, and southwest Asia are dry. Indiais not particularly wet since the annual total

combines a very wet summer with a desert-like winter. Although the easterntradewinds are as cloudy as the continentalequatorial regions (Figures 6-1 to 6-4), theinversion stops clouds from growing deepenough to give significant rain. The near-equatorial convergence zones (NECZ) overthe central and eastern North Pacific and theNorth Atlantic are sharply defined in theOLR, but radiate 20 W m-2 less than thecontinental equatorial maxima (see Chapter8, Section D-8).

East of Hawaii OLR is less than what isrecorded in other tradewind regions. Sheetsof higher cloud, unconnected to convectionor rain are rather common here in winter(Morrissey, 1986), and account for the differ-ence (see Chapter 8, Section D-5).

The elevated cold surfaces of Tibet and theAndes may contaminate the data, suggest-ing higher cloud tops and more rain thanmight actually fall there.

Figure 6-23 compares the mean annualzonal OLR and cloudiness. The OLR maxi-mum and cloudiness minimum coincidenear 20°N, where the cloud-free deserts aremore extensive than the stratocumulusdecks of the eastern ocean tradewinds.Although the coldest clouds straddle theequator (Figure 6-22), the Pacific and Atlan-tic NECZs and equatorial fine weather to thesouth shift the global minimum of OLR andmaximum of cloudiness to 6°N and second-ary turning points to 1°S. Near 20°S, OLRmaximum and cloudiness minimum coin-cide. The much larger cloudiness minimumreflects dominance of eastern ocean stra-tocumulus and near absence of deserts withclear skies.

Summing up, cloudiness can give a roughidea of rainfall, except over the easterntropical oceans, where a persistent inversionallows extensive, but shallow stratocumulusto develop over a heating ocean.

3. Variability. Rainfall variability is in-versely proportional to rainfall (see forexample, Figure 6-37). Thus, rainfall vari-ability tends to be directly proportional toOLR. Consequently, droughts are uncom-mon in wet (low OLR) regions, but in thedeserts, almost all the rain falls in brief, rare,intense storms, and causes severe erosion.

4. Interannual Variation and El Niño. Inthe tropics, most significant variations havebeen associated with the phenomenonknown as El Niño (Enfield, 1989). At irregu-lar intervals, ranging from two to twelveyears, the equatorial Pacific and watersalong the coasts of Ecuador and Peru be-come anomalously warm (Table 6-1). Usu-ally, this ocean change first appears in theeast in the SH summer, and extends west-ward through the year, to weaken anddisappear at the end of the next SH summer.

Complex interactions between air and seacause the sea surface temperature to change.The sequence appears to start along theequator in the western Pacific when a surgeof anomalous westerly surface winds, oftenaccompanied by tropical cyclones to northand south, generates an equatorial oceanicKelvin wave that travels eastward. It istrapped along the equator because theearth’s rotation confines the water motionsto the near-equatorial zone. A Kelvin wavetakes about two to three months to reach thecoast of South America. It is accompaniedby a deepening of the thermocline andeastward flow of near-surface water. Succes-

sive bursts of equatorial westerlies generatea sequence of Kelvin waves that maintainand enhance El Niño by causing the SST torise. As El Niño progresses, cool upwelledwater disappears off South America, andsubsequently in the central equatorial Pa-cific, where westerlies replace easterlies andshallow scattered clouds give way to deepconvective systems. The maritime continentbecomes relatively dry. In the most intenseNiño on record, in 1982-83, active convec-tion extended all the way to South America.

El Niño is one manifestation of a basin-wideoscillation between normally high surfacepressure in the southeast Pacific and nor-mally low surface pressure over Indonesia(Figure 6-24). When southeast Pacific minusIndonesia pressure is abnormally small, thatis, when the east-west pressure gradient issmall and the tradewinds relatively weak, ElNiño occurs. This “Southern Oscillation” isfurther discussed in Chapter 12-C-3.

Since the Niño of 1972-73, extensive researchhas sought to explain why the phenomenonstarts, why it persists, and how to forecast it.Increasingly sophisticated air-sea interactingnumerical models are now being used. Theyhave generally failed, as have statisticallybased forecasts, probably because the his-torical record is too short to encompass thewide variability among El Niño events.Despite this, better observations have usu-ally made it possible to recognize the earlysigns of El Niño (though false alarms are notuncommon). For example, in mid-December1990, a surge of surface westerlies along the

————————————————————————————————Table 6-1. Starting years of El Niño and anti-Niño events (1877-1988) (from Kiladis andDiaz, 1989).

El Niño (warm) events1877 1880 1884 1888 1891 1896 1899 1902 1904 1911 1913 1918 1923 1925 19301932 1939 1951 1953 1957 1963 1965 1969 1972 1976 1982 1986

Anti-Niño (cold) events1886 1889 1892 1898 1903 1906 1908 1916 1920 1924 1928 1931 1938 1942 19491954 1964 1970 1973 1975 1988

equator in the western Pacific was followedby tropical cyclones developing to north andsouth (Figure 9-15). The Kelvin wave movedacross the Pacific and resulted in anoma-lously high SST off Peru in March 1991,possibly presaging El Niño. However, thewesterly surge proved to be isolated, andSST in the eastern Pacific returned to normala few weeks later.

Statistics have linked rainfall anomalies invarious parts of the tropics to the stages ofEl Niño and to “cool” periods between Niñoevents (Table 6-1, Figure 6-25). Hastenrath(1990a) used the distribution of highlyreflective clouds (Garcia, 1985) to determinerelationships of tropical rainfall anomaliesto the Southern Oscillation. His findingsagree with those of Ropelewski and Halpert(1987,1989) shown in Figure 6-25.Hastenrath concluded that the equatorialconvection centers over South America,Africa and occasionally the maritime conti-nent tend to vary in unison. Thus, long-range forecasting is sometimes possible.Over northern Peru and Ecuador, wheredata are too scanty to permit the statisticalanalyses depicted in Figure 6-25, El Niñomay massively affect rainfall. The coastalplains are normally deserts. At the start of

El Niño, during the SH summer, they can befive to seven times as wet as usual (Cobb,1967,1968); in the super Niño of 1982-83,coastal rainfall was more than 40 timesnormal (Goldberg et al., 1988). As Table 6-2shows, summer over India tends to be drierduring El Niño, though this is not alwaystrue. India experienced a severe drought in1979 when El Niño was absent.

The Hawaiian Islands are also affected.Seventy-five percent of the cool seasonmonths at the end of El Niño are drier thanusual; conversely, during anti-Niño, 68percent are relatively wet. In the centralPacific, Christmas Island (2°N, 157°W) andCanton Island (2°S, 172°W), normally dry,receive an order of magnitude more rainduring El Niño.

The typhoon season that is usually dyingout by mid-November is prolonged towardthe end of El Niño and development occursfarther east than usual, probably becausewarmer water there encourages persistentsurface troughs. In 1982-83, phenomenalactivity in the SH (Figure 6-26) gave FrenchPolynesia as many tropical cyclones as inthe previous century!

El Niño is not duplicated over the equatorialAtlantic. Even so, the rainfall regimes ofcoastal Angola and coastal Peru are alike.Both coasts are bordered by cold upwellingwater and are very dry (Figure 5-4). Marchand April (the Angolan “wet” season) aredry when coastal upwelling and equatorialeasterlies are strong, and wet when coastalupwelling and equatorial easterlies areweak (Hirst and Hastenrath, 1983). As withPeru, the interannual variation of Angolancoastal rainfall is large.

5. Seasonal Distribution - the Monsoons.

a. Definition and Extent of the Monsoons.The monsoons blow in response to theannual change in the difference in air pres-sure over land and sea, which in turn iscaused by the difference in temperature

———————————————————————————————Table 6-2. Relationship of moderate or strong Niño to same-year Indian summer monsoonrainfall, 1875-1980 (Ramage, 1983)

Number of Seasons El Niño Non-NiñoRainfall above normal 5.5 50.5Rainfall below normal 15.5 34.5

between land and sea. The annual change isgreatest where continents border oceans.When the sun moves north of the equator inthe NH summer, Asia, because of its rela-tively low heat capacity, and the presence ofthe Himalayas and Tibet, is rapidly warmed.On the other hand, the northern IndianOcean and the western Pacific store the sun’sheat within their deep surface layers. Conse-quently, the land gives off heat more readilythan the sea, and the air over the land be-comes warmer and the air pressure lowerthan over the neighboring ocean.

Thus during summer, air flows from theIndian Ocean toward lower pressures oversouthern Asia, ascending as it is heated overthe land, until it reaches a level at which thepressure gradient is reversed, whereupon itflows on a return trajectory from land to sea.There it descends, and is once more takenup by the landward-directed pressuregradient (Figure 6-27, top). As long as theland is significantly warmer than the sea,this great circulation persists.

In winter the reverse occurs. The low heatcapacity of Asia relative to the northernIndian and western Pacific Oceans insuresthat over the land air is colder than over thesea. Then, the winter monsoon prevails. Atlow levels air flows out from the continentover the sea where it rises and returns in themiddle and upper troposphere to the land,sinks to the surface, and resumes the cycle(Figure 6-27, bottom).

The African monsoons differ from the Asianmonsoons. During the NH summer thedeserts of North Africa heat rapidly and, asover Asia, cause pressure to fall. South ofthe equator, during the SH winter, coolingcauses a pressure rise.

Thus a pressure gradient is established fromsouth to north across Africa, setting up amassive flow of air, also from south to northacross the equator. Because the Coriolisforce changes sign at the equator, air flow is

more directly from high to low pressurethere than at higher latitudes. Africa influ-ences the circulation as far as 800 km to theeast, merging with the influence of Asia tothe north. In the NH summer, wind circu-lates in a huge gyre. It blows from southeastaround the northern edge of the SouthIndian Ocean anticyclone toward the coastof Africa, then swings to south across theequator, and then to southwest to parallelthe African, Arabian and Asian coasts, andfinally sweeps across India, Burma and theIndochina-Thailand peninsula as the south-west or summer monsoon (see Figure 4-8).Six months later, a complete reversal takesplace. Northern Africa is cold and southernAfrica is warm and so the winds blow fromnorth across the equator in the westernIndian Ocean and over East Africa.

Australia experiences less intense, butpersistent monsoons. Monsoon tendencieshave been identified in many other regions,for example, Mexico, southwestern UnitedStates, the Caspian Sea and even parts ofEurope. Since an annual shift in circulationdirection is the fundamental property of themonsoons, monsoon regions encompassJanuary and July circulations in which:

1. the prevailing wind direction shifts by atleast 120° between January and July;

2. the average frequency of prevailing winddirections in January and July exceeds 40percent;

3. the mean resultant winds in at least one ofthe months exceed 6 knots;

4. fewer than one cyclone-anticyclone alter-nation occurs every two years in eithermonth in a 250 x 250 NM (500 x 500 km)square.

These criteria are met in a contiguous regionextending from western Africa to Indonesiawith southward protrusions to Madagascarand northern Australia (Figure 6-28). Aspointed out in Chapter 5, Section B-2, up-

welling along the north and west coasts ofSouth America, and along the west coast ofsouthern Africa, prevents monsoons fromdeveloping there.

Weather is not a monsoon criterion. Nosimple connection exists between the direc-tion of the surface pressure gradient andrainfall. However, most rain falls in summeror autumn, except for a near-equatorialband possessing a double maximum (Figure6-28). The summer maximum is associatedwith heat trough intensification, althoughthe heat troughs themselves are dry. Thedouble maximum is probably associatedwith temporary intensification of near-equatorial troughs in the transition seasons.

b. Role of the Himalayan-Tibetan Massif inthe Monsoons. The chain of mountainsextending from Turkey to west China, byprotecting the land to the south from coldpolar outbreaks, produces a sharp disconti-nuity in surface monsoon characteristicsalong about 100°E. Thus in summer thesouth Asian monsoon is stronger (moreintense heat trough) than the east Asian,whereas in winter the east Asian monsoon isthe stronger (more intense polar anticy-clone).

Figure 6-29 reveals a large cloudiness gradi-ent between north-west India and westernChina, a distribution that prevails through-out the year and is reflected in the rainfall.This weather discontinuity derives from themountain-plateau mass of the Himalayasand Tibet which at times may also influenceSH monsoons.

In the broadest sense, the NH monsoonscomprise three parts: east of Tibet, wherespring and summer are wet and significantwinter precipitation falls; west of Tibet,where the climate is desert-like; and southof Tibet, where summers are wet and win-ters are desert-like. These distributions stemfrom the combined mechanical-thermaleffect of Tibet. During summer, central and

southeastern Tibet is a strong radiationalheat source; with the aid of condensationalong the Himalayas it anchors an upper-tropospheric subtropical ridge across south-ern Tibet (Figure 6-30). The heating in-creases the N-S temperature gradient of thesummer monsoon and so causes a speedmaximum in the upper-troposphericeasterlies (the easterly jet) south of India.Thus, upstream to the east of about 70°E,rising motion beneath divergent easterliesfavors clouds and rain. Downstream to thewest of 70°E subsidence beneath convergenteasterlies keeps skies clear. Karachi (25°N,67°E) and Bombay (19°N, 73°E) (see Figure8-1) typify the two regimes. Both are coastalstations upwind of mountain chains. Bothare exposed to the fresh southwesterlies ofthe summer monsoon, which are slightlymoister at Karachi. June - September rainfallaverages 6.00 inches (152 mm) at Karachi,and 56.72 inches (1441 mm) at Bombay.

Central and southeast Tibet are almost snowfree during autumn, winter and spring, andso persist as a high-level radiational heatsource. Pressure surfaces are raised there,diminishing the S-N temperature gradientand the strength of the subtropical jet to thesouth. Immediately east of the plateau,subsiding air is warmed by compressionalheating. Over central China it flows along-side very cold air which has swung aroundthe northern edge of the massif (Figure 6-31). Here the very large temperature gradi-ent produces an exceptionally strong jetstream. With speeds increasing downstreameast of 100°E, upper-tropospheric diver-gence favors large-scale upward motion andincreased cloudiness. West of 100°E, gener-ally convergent upper-tropospheric wester-lies over the Middle East and North Africafavor sinking and less cloud.

Year-long subsidence persists over the greatdeserts of southwest Asia and North Africa,extending to the surface in winter anticy-

clones and to the lower troposphere insummer heat lows.

c. Rain During the Monsoons. The mon-soons, as modified by Himalaya-Tibet, setthe stage for weather. Over and near thecontinents, the cold dry air of the wintermonsoon usually predisposes the weather tobe fine, while the warm moist air of thesummer monsoon favors unsettled weather.Although during the monsoons movingsurface disturbances are rare and surfacewinds are notably steady, rainfall on a scalefrom days to weeks is surprisingly variable.Synoptic changes and intensification anddecay of quasi-stationary bad-weathersystems are responsible.

The circulation components of the mon-soons represented in Figure 6-32 are dis-cussed in detail in Chapter 1, Section B (heatlows), Chapter 8, Section C (monsoon de-pressions, West African cyclones, mid-tropospheric cyclones and upper cyclones)and Chapter 9 (tropical cyclones). They canbe separated into three groups in the mon-soon region; predominantly summer sys-tems are shown in bold face, and predomi-nantly winter systems in italics:

1. circulations in which downward motionand fine weather predominate — polar andsubtropical anticyclones, and heat lows;

2. circulations in which upward motion andwet weather predominate — tropical cy-clones, monsoon depressions, West Africancyclones, and mid-tropospheric cyclones(Rao, 1976);

3. other systems with marked weathergradients — troughs in the upper-tropo-spheric westerlies, the Polar Front, near-equatorial troughs, quasi-stationary non-circulating disturbances, surface trans-equatorial flow, and squall lines.

Figures 6-30 and 6-31 show the regionaldistribution of some of these synoptic com-

ponents. They will be discussed in moredetail in Chapters 8 and 9.

Tropical cyclones, monsoon depressions andsquall lines interrupt the surface monsooncirculation as they develop and move.However, other synoptic systems may movevery little and do not significantly changesurface wind directions. Thus, quite variableweather may occur in what is superficially asteady monsoon. As described above inSection B-1, most rain falls in the form ofrains or showers.

d. March of the Seasons. During spring andautumn, the transition seasons betweenwinter and summer monsoons, complexchanges affect the large-scale circulationsand the mixtures of synoptic systems. Thechanges differ within the monsoon area andfrom one year to the next— Indian droughtsare well-known. The spring transition lastsabout twice as long as the autumn transi-tion, because in spring the normal equator-pole temperature gradient is reversed, whileautumn marks return to the normal gradi-ent. For example, over West Africa the heattrough takes five months to move from 9°Nto 22°N, and only three months to return to9°N. Since the monsoons arise from differ-ing heat capacities of land and ocean, in atransition season circulations change first inthe surface layers. After several weeks thechange has spread through the troposphereand the transition season is over.

In the Indian Ocean-south Asia region, aneast-west trough exists throughout the yearin the tropics of each hemisphere. In the NHsummer, the heat trough lies above 25°Nand there may be a very weak secondarytrough close to the equator (Figure 6-33). Assummer gives way to winter, the primarytrough weakens and dissipates and thesecondary trough becomes predominant.The sequence reverses from winter to sum-mer. The transition season “jumps” arecaused by rapid cooling in autumn andheating in spring of the deserts of northwest

India and Pakistan. This accounts for theusual absence of a double rainfall maximumbetween 10° and 25°N. South of the equator,despite considerable day-to-day fluctua-tions, a single trough follows the annualmarch of the sun.

Over Africa, a weak heat trough is some-times found between 5-10° latitude in eachhemisphere during the transition seasons.As the year advances, the trough in thewinter hemisphere disappears and thetrough

in the summer hemisphere intensifies andfollows the sun poleward. The sequencethen reverses through the following transi-tion season.

The Indian Ocean-south Asian regime isweakly duplicated in the northern Austra-lian region.

In late spring and early summer East Asiaexperiences an unusual transition season.North of the monsoon trough, and sepa-rated from it by a weak ridge, a very activefrontal trough (the “Mei-Yu front”) domi-nates south China and Taiwan. It was stud-ied during the Taiwan Mesoscale Experi-ment (TAMEX) from 1 May to 29 June 1987(Kuo and Chen, 1990). As the front movesslowly northward between May and July, itmay stop and even shift a little southward attimes. The front separates relatively cool airin the north from moist, conditionally un-stable air in the south. Along the front, weakmesoscale depressions form and driftslowly eastward. When one is overlain byvigorous upper tropospheric divergence,very heavy rain, and occasionally a continu-ous thunderstorm, may develop within thedepression. In the 850 to 700 mb layer, alow-level jet reaching 25 knots sometimesprecedes and sometimes accompanies heavyrain to the north. In Figure 6-34 a Mei-Yufront extends from northern Indochina tosouthern Taiwan. Hong Kong (22°N, 114°E),lying under the northern part of an ex-

tremely cold cloud mass, experienced 8inches (200 mm) of rain in seven hours.Absence of shear in the deep clouds sug-gests that they extended from a surfacetrough to an upper tropospheric ridge.

As the Mei-Yu front moves northward, it isfollowed by a brief period of better weatherunder a weak ridge (Figure 6-38). That inturn gives way to the monsoon trough andthe typhoon season proper in July.

The satellite helps forecasting. A very cold,non-shearing mass of cloud, many timeslarger than a single thunderstorm, possiblyanchored and enhanced by orography, cangive prolonged heavy rain. The Mei-Yufront is not duplicated elsewhere in thetropics. Probably this is because only overChina does persistent cloudiness hinder aheat trough from forming, and absence ofeast-west mountain ranges allows cool, mid-latitude air to reach the tropics even in earlysummer. Over the ocean, troughs near theequator spawn cloud clusters; poleward,tropical cyclones may develop in thetroughs. Over land, west of 70°E, fineweather prevails in the heat troughs, but ontheir equatorward sides hybrid cyclones (seeChapter 8, Section C-3) may develop.

e. Summary. During summer and winter,although surface disturbances are rare,middle-or upper-tropospheric disturbancesare not. In the transition seasons differencesbetween the monsoon area and surroundingareas are least. Thus, latitude for latitude,the range and complexity of annual varia-tions are greater in the monsoon area thanbeyond.

The general character of the monsoons andtheir inter-regional variations reflect thejuxtaposition of continents and oceans.However, without the great mechanical andthermal distortions produced by theHimalayas and the Tibetan Plateau, the vastNH deserts would be moister, and central

China would be much drier and no colder inwinter than India.

Within the monsoon area, annual variationsare seldom in phase. The climatologicalcycles merely determine necessary condi-tions for certain weather regimes; synopticchanges then control where and when rainwill fall and how heavily, and whetherwinds will be destructive. The great verticalcirculations comprising the monsoons(Figure 6-27) often undergo wide-rangingnearly simultaneous accelerations or decel-erations, apparently triggered by priorchanges in the cold parts of the circulations(see Chapter 8, Section D-4). Monsoonweather, on the scale of individual clouds,seems to be determined by changes occur-ring successively on the macro-and synopticscales. Rains set in — not when cumulonim-bus gradually merge, but when a synopticdisturbance develops, perhaps in responseto change in a major vertical circulation.Showers too are part of the synoptic cycle.Individually intense, but collectively lesswet, they succeed or precede rains as gen-eral upward motion diminishes.

6. Monthly Distribution. The tropics isdivided into four regions as shown in Fig-ures 6-35 and 6-36: the Americas, Africa,Asia and Australia, and the Pacific. Thegraphs show mean monthly rainfalls. Theaverage rainfall in mm is given beneath thestation location. Bar graphs are not shownfor stations with an average rainfall of 5inches (125 mm) or less. Most of the stationshave 30 or more years of data.

Some general deductions can be made fromthese graphs. Most tropical stations havewet and dry seasons or significantly morerain in some months than in others. Near theequator some stations have a double maxi-mum, while more poleward stations gener-ally experience a single maximum during orjust after the high-sun period. The rainy

season, as shown by the monthly means,starts abruptly in Bombay, India and SanSalvador, El Salvador (Figure 6-35). Therainy season begins more gradually at otherstations, such as Addis Ababa (9°N, 39°E),Ethiopia (Figure 6-36). Similar variationsoccur at the end of the rainy season. At somestations, rain may vary considerably withinthe rainy season. For example, Kingston(18°N, 77°W), Jamaica (Figure 6-35) has arelative maximum in May and June, amarked decrease in July followed by anincrease to the yearly maximum duringOctober. Several station pairs illustrate theeffects of orography, elevation, and expo-sure to the prevailing wind. For example,Honolulu (21°N, 158°W), on the lee side ofthe island of Oahu, receives an average of 22inches (559 mm) compared to Hilo, (20°N,155°W) on the windward side of the islandof Hawaii, where 137 inches (3480 mm) falls(Figure 6-36). On the island of St. Helena(16°S, 6°W), in the South Atlantic (Figure 6-36), Jamestown on the coast at 40 feet (12 m)elevation receives 5 inches (125 mm) a yearcompared to 32 inches (813 mm) at HuttsGate at 2062 feet (855 m) elevation near thecenter of the island. For further discussions,the reader is referred to climatology text-books such as Trewartha (1981), and specialtreatises on the climates of particular areas.

7. Monthly Variability. The coefficient ofvariation, Cv, is defined as the ratio of thestandard deviation of the monthly amountsS to the mean monthly amounts X, times 100percent

Cv = (S/X)100

In Figure 6-37 the coefficient of variation isplotted against mean monthly rainfall atIndian stations during the rainy season(Mooley and Crutcher, 1968). The solid lineis fitted by eye to the points; however, theindividual station months vary considerablyabout this line. The dashed line is from a

similar study by Coligardo (1967) for sta-tions in the Philippines. For large meanmonthly rainfalls, both curves approach alimiting Cv value of about 40 percent. Vari-ability increases rapidly as amounts dropbelow 8 inches (200 mm). Below 2 to 3inches (50 to 75 mm), standard deviations ofthe rainfalls exceed the means, i.e., Cv >100percent.

In the Sahel of North Africa, south of thesummer heat trough, Cv is very large. AsOsman and Hastenrath (1969) pointed out, itwas unusually wet in July 1958 when theheat trough lay to the north along about19.5°N, and unusually dry in July 1966when the trough stayed south of 16°N.Nicholson (1981) confirmed this finding forthe Sahel as a whole. Various statisticaldistributions have been used to describeannual and monthly rainfall variability. Thenormal (Gaussian) distribution approxi-mates the frequency distributions at stationswith large annual or monthly rainfalls. Fordrier stations, distributions which areskewed to the right, such as log-normal,square-root normal, or gamma distributions,are more appropriate. Although the gammadistribution is more complex than the othertypes, it generally fits rainfall series betterunder a wide range of conditions (Mooleyand Crutcher, 1968). Further details on thesestatistical distributions and their applica-tions are available in various textbooks(Brooks and Carruthers, 1953).

8. Pentad Distribution and Beginning andEnding of Rainy Seasons. Some tropicalstations compile Pentad (5-day) rainfallaverages. These often reveal climatic fea-tures hidden in the monthly averages. Oversouth China (Figure 6-38), the Mei-Yu rainsease off and are followed by a relatively dryspell from 1-15 July, generally as a weakhigh-pressure ridge moves over from thesouth. The second rainy period coincideswith more frequent tropical cyclones. This

sequence cannot be found in the monthlyaverages for June and July. Japanese stationsto north and east follow the pattern withabout a two-week delay (Ramage, 1952b).

Figure 6-39 shows long-term Pentad rainfallfor Indian stations (Ananthakrishnan andPathan, 1970). Missing from the bar graphsof monthly means (Figure 6-35) is the sec-ondary minimum near the 46th Pentad(mid-August) at all stations north of 16°N.More frequent “breaks” in the monsoonrains between 10 and 20 August(Ramamurthi, 1969) may be the cause.

Were Pentad rainfall averages more gener-ally available, climatological regimes couldbe much better defined.

Where the tropics experience distinct dryand wet seasons, as in the monsoons, thereis great interest in knowing when the rainyseason will start and finish. Staff weatherofficers supporting tropical operations areoften urged by field commanders to forecastrainy season beginning and ending dates.Estimates often depend on gross climato-logical data (such as mean monthly rain-falls) without a firm understanding of whatis being forecast. The rainy season is noteasy to define. The remainder of this sectionreviews definitions and delineations of therainy season in various parts of the tropics.

Pentad rainfalls have helped define the startand end of the rainy season over India (Rao,1976), although the task is challenging.Experienced Indian forecasters settled onthe following objective criteria for declaringthe onset of the monsoon over Kerala (in thesouthwest corner of India)(Ananthakrishnan et al., 1967).

1. Beginning from 10 May, if at least five outof the seven stations report 24-hr rainfall 1mm or more for two consecutive days, theforecaster should declare on the second daythat the monsoon has advanced over India.

2. Thereafter, the daily rainfall distributionshould be watched and if it is found thatthree or more stations out of seven report norainfall for three consecutive days, theforecaster should indicate on the third daythat the monsoon has receded from Kerala.The recession of the monsoon will thus bepreceded by weak monsoon conditions atleast for a day or two. (There is nothingwrong in saying that the monsoon hasreceded in the early stages of its onset if webear in mind the pulsatory character of themonsoon. As a matter of fact, such an-nouncements have been made in the dailyweather bulletins in the past.)

3. One important point has to be borne inmind in the practical application of rule 2.This rule can be applied only if the monsoonhas not advanced into Konkan (the provincenorth of Kerala) and is still confined tosouth of 13°N. If the monsoon has advancednorth of this latitude it is illogical to recedeit from Kerala on the rainfall criteria givenunder 1. In that case we can only say thatthe monsoon is weak over Kerala.

4. After stating that the monsoon has re-ceded on the basis of criteria 2. and 3., theforecaster should continue to keep a watchon the rainfall of the seven stations andwhen criterion 1. is again satisfied he shoulddeclare that “the monsoon has revived overKerala” or alternatively “a fresh advance ofthe monsoon has taken place over Kerala”.

5. Rules 2. and 3. can again be applied ifrequired.

6. The date of permanent onset of the mon-soon for the purpose of records may betaken as that date after which it does notbecome necessary to recede the monsoonover Kerala.

As mentioned in the earlier discussion of themonsoons, rain associated with large-scalecirculation changes falls at the whim ofsynoptic events; at any location wet weatheris interrupted by fair weather intervals.

Table 6-3 shows the normal dates the sum-mer monsoon rains start in various Indiansubdivisions, as well as the earliest andlatest dates. The apparent progression fromsouth to north is seldom realized in a singleseason. The rains set in during synopticpulses and may occasionally start earlier inthe north. Figure 6-40 traces the onset datesat Bombay (19°N, 73°E) between 1879 and1975. The term “monsoon” and the rainyseason must be clearly differentiated. “Mon-soon” applies to prevailing surface seasonalwinds, i.e., the northeast (winter) monsoonand the southwest (summer) monsoon. Laypeople, including those who use weatherforecasts, equate summer monsoon withincreased rain. However, the start or finishof the rainy season may differ significantlyfrom the changes in prevailing surfacewinds. Along the west coast of India, west-erlies generally set in at least a month beforethe rainy season starts.

Griffiths (1964) tried to define the beginningand ending of the rainy season at San Salva-dor (14°N, 89°W) in two ways.

In a climatological approach, he summedthe daily rainfall over 40 years by pentads.In a “rainy” Pentad, more than 1 inch (25mm) of rain falls. Figure 6-41 shows thenumber of years on which each

Pentad was rainy and a subjective estimateof any Pentad being rainy (dotted lines). Forexample, after about 20 May and beforeabout 15 October, there is over a 50 percentchance of any Pentad being rainy, with theprobability much less before and after thesedates respectively. Such a straight climato-logical approach gives no direct measure ofthe year-to-year variation. Therefore,Griffiths also used the “practical approach”.He computed the soil-water budget fromdaily rainfall and an assumed average dailyloss of .02 inches (5 mm) due to runoff andevaporation. Figure 6-42 shows the periodseach year at San Salvador when the soil-water budget exceeded 1 inch (25 mm). This

representation may be useful to agricultureand to the military for trafficability esti-mates. In some years, short spells of rainprecede the main rainy season while inother years the rains continue once theystart. Using the soil-water budget definitionand ignoring breaks of less than five daysonce the main rains set in, Griffiths calcu-lated the average onset date at San Salvadoras 25 May, with a range from 11 April to 26June and the average ending date as 2 Janu-ary, with a range from 21 November to 12January. Note that the average beginningdate agrees with that found by the climato-logical approach but the ending date ismuch later (Figure 6-42). This lag resultsfrom the slow drying of the soil after therains have ended.

Granzow and Henry (1972) first defined a“rainy” Pentad as having more than 25 mmof rain. They then delineated the wet seasonover Central America as being encompassedby the first and last rainy Pentad. Althoughsome stations experienced a relative dryspell in mid-summer, the method workedwell (Figure 6-43 and 6-44). The patterns arecomplex and are affected by the prevailinglow-level easterlies. The rains start first onthe Caribbean coasts of Panama and CostaRica, followed by Nicaragua, and progressacross the isthmus. Uplift on the El Salvadormountains causes an early start there. Then,progress is northward across Honduras. Therains end first in the mountains of El Salva-dor, Honduras and Nicaragua and much

sooner on the Pacific coast than on theCaribbean coast.

Horel et al. (1989) used Pentad averages ofoutgoing longwave radiation for 1974-1988to define the rainy seasons of CentralAmerica and the Amazon basin. For each,they calculated the fractional low radiance(FLR), defined as the ratio of the number of2°X2° latitude/longitude grid areas whichhave OLR less than 200 W m-2 divided bythe number of grid areas in the enveloping10°X10° box (usually 25). High FLR com-bines frequent and/or intense convectionover a 5-day period. As Figure 6-45 shows,deep convection is confined to the summerof each hemisphere, with the equator morelike the Amazon. The transition seasons arebrief, especially during October. There islittle evidence of deep convection passingback and forth across the equator during thetransitions (Figure 6-46).

Horel et al. then defined the durations of thewet season in each hemisphere for thetropical Americas and Africa (Figure 6-47).Their wet season started with the first of fivepentads during which the maximum FLR isin the opposite hemisphere. The definitionfails to take account of heavy convection inthe same pentad in both hemispheres, andso does not allow any time overlap. OverCentral America, the mean wet seasonduration agrees with the dates determinedby Griffiths (1964) and Granzow and Henry(1972). The Amazon wet season and theCentral American dry season start as the 200

Table 6-3. Advance of Indian summer monsoon.

Area Normal Earliest LatestCoastal Karnataka (13-15°N) 4 June 19 May 1962 14 June 1958North Konkan (17-20°N) 8 June 29 May 1956 25 June 1959West Bengal (22-25°N) 7 June 27 May 1962 23 JuneVidarbha and most of 12 June First week Last weekof Madhya Pradesh (18-25°N) of June JuneBihar (22-27°N) 12 June 6 June 1 JulyE. Uttar Pradesh (25-28°N) 15 June 5 June 3 JulyW. Uttar Pradesh (25-30°N) 25 June 10 June 9 July———————————————————————————————-

mb Bolivian anticyclone develops, and ceaseas it weakens (see Figures 4-10 and 4-11).Over Africa the southern wet season islonger than the northern, where the drynessmight be partly due to downstream subsid-ence induced by Himalaya/Tibet (see Sec-tion 6-b).

For the area around Saigon (11°N, 107°E),Fukuda (1968) defined the end of the rainyseason as the first day of the first 14-dayperiod when the average daily rainfall atseven stations in the Saigon area was lessthan .08 inches (2 mm) (i.e., the average totalof the daily rainfalls over a 2-week periodwas less than 1.10 inches (28 mm)). Thecriterion was selected because after 1 Octo-ber, two weeks of little rain is unlikely to befollowed by significant falls. Figure 6-48shows the three-day average of the dailyrainfalls. The arrows indicate the dates therainy seasons ended. End dates ranged from23 October (1962) to 24 December (1966)with an average date of 15 November. Insome years the rains ended abruptly, whilein others they tapered off. Interannualvariability was large between 1966 and 1968.

Tropical meteorologists trying to make long-range forecasts of rainy season beginning orending should express them in probabilityterms based on long-period data. The opera-tional user must clearly understand how therainy season is defined. Despite user pres-sure, dogmatic forecasts such as “the rainyseason will start on 15 May” must beavoided.

9. Daily Distribution. As in mid-latitudes,synoptic-scale disturbances are primarilyresponsible for creating the environmentthat favors most tropical rainfall. Within thedisturbances, rain from convective andcloud-scale processes lasts from tenths ofhours to hours (Garstang, 1966) and ac-counts for most of the accumulation (Henry,1974). Rain falls mostly within a small

fraction of the total periods with rain. Onthe island of Barbados (13°N, 59°W) half ofthe total rain falls in 10 percent of the timeintervals with rain, whether 24-hour, 1-hour,or 0.1 hour periods are used as the basicinterval (LaSeur, 1964b). Only on mountainsexposed to the tradewinds is this distribu-tion unlikely (see Figure 6-21). Garstang(1966) expanded LaSeur’s single-stationamount-frequency analysis to Florida where10 percent of the rainy days accounted formore than 70 percent of the areal rainfalls.Similarly, at stations in southeast Asia halfthe rain fell in about 10 percent of the hours(Riehl, 1967). About 80 percent of the areaencompassed by a synoptic disturbancelacked rain at any instant. In another study,Riehl (1949) showed that over central Oahu,most rain fell during synoptic disturbances.On average, ten rainstorms a year with amedian rainfall exceeding 0.75 inches (19mm) produced over two-thirds of the totalannual rainfall. Riehl also showed that asthe rainstorm intensity increases, orographybecomes less important. Riehl et al. (1973)came to a similar conclusion for Venezuela.

Many more studies of tropical daily rainfallgave similar results. They can be describedby composite tropical rainfall frequency-versus-amount curves. To construct thecurve, first order the days with rain from thesmallest to the largest amounts and thenstart the plot at zero with the smallestamount and finish at 100 cumulative percentfrequency with the largest amount. Onesuch curve has been derived by Martin(1964) for 30 years of daily rainfall at fivestations in North and Central America(Figure 6-49). The abscissa gives the cumula-tive percent amount and the ordinate thecumulative percent frequency of rainfall.The dashed lines paralleling the curve showdepartures of two standard deviations fromthis mean. The dashed diagonal line showsthe distribution if it always rained the sameamount for each time period considered.Once again the curve demonstrates that half

the days with rain account for only about 10percent of the total, while half the total isaccounted for by 10 to 15 percent of therainy days. The curve also fits 20 additionalstations in the region and closely resemblesfrequency-amount curves derived byOlascoaga (1950) for Argentina and by Riehl(1950) for San Juan (18°N, 66°W). Its appar-ently wide applicability led Martin to call itthe “universal daily rainfall distributioncurve”. The data are listed in Table 6-4. Thecurve can also be applied to other timeintervals, such as hours.

The universal curve can be used toestimate the frequency distribution of dailyrainfalls at a tropical station where only themean monthly rainfall and mean monthlynumbers of days with rain are known. Forsimplicity, assume a station during a givenmonth has an average of 20 days with rainand a mean total of 10 inches (250 mm).Over ten years, 200 days with rain, totalling100 inches (2500 mm), could be expectedduring the given month. Table 6-5 shows theestimated frequency distribution of dailyrainfalls at this station during a ten-yearperiod based on the universal rainfall distri-bution curve values given in Table 6-4. Thepercent frequencies corresponding to each10 percent rainfall are determined by sub-tracting subsequent numbers in the percentfrequency column in Table 6-4. In this hypo-

thetical case, 60 percent of the total rain fallsduring days reporting 1 inch (25 mm) ormore.

Cobb (1968), on applying Martin’s universalcurve to daily rainfalls for southeast Asianstations, found significant differences in theannual mass-distribution curves betweenone group of stations influenced primarilyby the southwest monsoon and anothergroup influenced by the northeast monsoon.He concluded that describing rainfall distri-butions at a variety of tropical stationsmight require a family of curves. Similarconclusions were reached by Wexler (1967)in a study of rainfall in Thailand. Martin’suniversal curve estimates the distribution ofdaily rainfalls fairly well. Better estimatesdemand individual daily rainfalls, or fre-quency distributions of daily rainfalls suchas those given in the USAFETAC “SurfaceObservation Climatic Summaries”.

10. Variation with Elevation. Orography isundoubtedly a major factor in determiningrainfall distribution. In a comprehensivediscussion, Barry (1981) pointed out thatmany local or regional complications pre-vent generalizing. Figure 6-50 from Lauer(1975) illustrates this. In West Africa, thesouth side of Mt. Cameroon (4°N, 9°E),exposed to the summer monsoon, has a

———————————————————————————————Table 6-4. Data for the universal rainfall distribution curve (from Martin, 1964).

Cumulative Cumulative Standard% Amount % Frequency Deviation (%)0 0 010 50.6 5.120 66.1 3.930 75.4 3.240 82.1 2.750 87.2 2.160 91.2 1.770 94.3 1.380 96.7 0.990 98.7 0.4100 100.0 0———————————————————————————————

rainfall maximum at its base, whereas on itsnortheast side, where tradewinds dominate,the maximum lies at 5000 feet (1500 m).Convective heating over inland plateausmay further raise the maximum above 6500feet (2000 m). In general, the maximumzones tend to rise as annual rainfalls dimin-ish. That Lauer has by no means included allprofiles is confirmed by two from the Ha-waiian Islands (Giambelluca et al, 1986)added to Figure 6-50. The summit of Mt.Waialeale (22°N. 159°W) lies about 1600 feet(500 m) below the tradewind inversion, andso the inversion does not seriously impedevertical motion. However, on the island ofHawaii, with peaks reaching 13,000 feet(4000 m), the inversion hinders rising airand the tradewinds flow mostly around theobstacle. Thus the rainfall maximum at 2000feet (600 m) is lower than on Waialeale.

Cobb (1966) lists five factors that influenceterrain effect:

a. Shape, size and roughness of the moun-tains

b. Direction and distance from the moisturesource

c. Intervening terrain between moisturesource and the orographic barrier

d. Wind velocity

e. Thickness and stability of the moist layer.

To these should be added -

f. Height and intensity of any subsidenceinversion.

These factors must be considered by meteo-rologists forecasting for mountainous areasin the tropics and by climatologists prepar-ing isohyetal maps.

11. Mesoscale Distribution. As shown inthe discussion of daily rainfalls, a few dayscontribute most of the total, and as Riehl(1979) pointed out, most rain comes frommesoscale disturbances measuring about13.5-54 NM (25-100 km) across. They de-velop within favorable macroscale systemsand are suppressed when the macroscale isunfavorable (Henry, 1974). The size of me-soscale disturbances can be inferred byrelating daily rainfalls between station pairsat various distances apart. Henry (1974)studied mesoscale rainfall systems in Cen-tral and South America using the contin-gency index (CI) computed from dailyrainfalls as follows:

Table 6-5. Examples of applying the universal rainfall distribution curve to determinedistribution of daily rainfall. A given month has an average of 20 days with rain and amean rainfall of 10 inches (250 mm). The frequency of various daily rainfalls during a 10-year period are estimated below.Rainfall Number of days Required Average Rainfall to Produce this Amount per Rainy Dayin. mm in. mm10 250 50.6 X 200 = 101.2 (101)* .10 2.510 250 15.5 X 200 = 31.0 (31) .32 8.110 250 9.3 X 200 = 18.6 (19) .52 13.210 250 6.7 X 200 = 13.4 (13) .77 19.610 250 5.1 X 200 = 10.2 (10) 1.00 25.410 250 4.0 X 200 = 8.0 (8) 1.25 31.710 250 3.1 X 200 = 6.2 (6) 1.66 42.210 250 2.4 X 200 = 4.8 (5) 2.00 50.810 250 2.0 X 200 = 4.0 (4) 2.50 63.510 250 1.3 X 200 = 2.6 (3) 3.33 84.6100 in. 2500 mm 200 days* Rounded to nearest whole day———————————————————————————————-

The 1 inch (25 mm) rainfall isopleth delin-eates the extent of mesoscale rainfall sys-tems. The actual frequency of occurrence ineach of the nine blocks is given by the a’s,the column and row totals are given by thex’s and y’s respectively, and the total numberof days by T. A similar table can be con-structed showing the number of occurrencesexpected in each block by chance, indicatedby b’s.

x1y1 x2y2 x3y3

b1 = ——— , b2 = —— , ......... b9 = ——- T T T

The contingency index is then computed asfollows:

(a1 + a5 + a9) - (b1 + b5 + b9) CI = ——————————————————————————————— T - (b1 + b5 + b9)

The numerical values of CI range from 1.0 to-1.0; 1.0 indicates a perfect relationshipbetween amounts at the stations (i.e., allvalues in blocks a1, a5, or a9,), a CI of zeroindicates no relationship between the sta-tions, and CI < 0 indicates an inverse rela-tionship. In other words, the CI value indi-cates the tendency for no, light or heavy rainto occur on the same days at two stations. CIcalculations for hundreds of station pairs inCentral and South America led to a typicalprofile of CI versus station separation (Fig-ure 6-51). The CI falls off with distance

to a minimum at 19 NM (35 km), and thenincreases to a maximum at about 35 NM (65km), which suggests that mesoscale rain

systems are about 13 NM (25 km) wide andthat their centers are about 30 NM (55 km)apart. Rainbird (1968) obtained a similarresult for two summers over the SeSansubbasin of the Mekong River. Henry (1974)found that during the dry season storms aresmaller, but have larger CI than in the wetseason. The rainfall per storm is about thesame. Schroeder (1978) concluded that overthe Hawaiian Islands orography so domi-nated CI that no indisputable mesoscalestructure could be detected. It is hard tounderstand why orography over central andSouth America did not also mask the mesos-cale.

From hourly rainfall data for stations inThailand and Cambodia during the south-west monsoon, Henry (1974) determined thetypical rainfall associated with a mesoscaleconvective disturbance. In his analysis rainshowers of less than 0.4 inches (10 mm)were omitted; this eliminated 50 percent ofthe rainy days and about 15 percent of thetotal rainfall. Henry constructed a compositerainstorm for the months May throughOctober for 13 stations in Thailand andCambodia (Figure 6-52). Of course, indi-vidual rainstorms may depart significantlyfrom this composite.

Any hour may have many showers.Garstang (1959) used ten years of record toderive the frequency distributions of showeramounts and durations at Piarco (11°N,61°W) during the wet and dry seasons(Figure 6-53). Most showers total less than0.16 inches (0.4 mm) in the dry season andless than 0.4 inches (1 mm) in the wet sea-son. Showers usually last 12 minutes or lessand rarely more than 24 minutes.

Rainfall at Station A 0 <25 mm >25 mm TotalRainfall 0 a1 a2 a3 y1

at <25 mm a4 a5 a6 y2

Station >25 mm a7 a8 a9 y3 BTotal x1 x2 x3 T———————————————————————————————-

Winner (1968) applied empirical relation-ships in estimating from the mean annualrainfall how often various hourly rainfallsare equalled or exceeded at tropical stations.He used hourly rainfall data for 32 stationsin the Panama Canal Zone (Figure 6-54).When applied to 13 stations in Thailand andCambodia, the estimated and observedfrequencies agreed well. The graphs give themean number of clock-hours per year thatvarious hourly rainfalls (shown in inchesnext to each line) are equalled or exceeded.Clock-hour rainfall rate is the total amountof rain measured during a one-hour period.(The term clock-hour is used because obser-vations are usually made on the hour, that isat 07, 08 h etc.). If during this time it rainedfor only five minutes, for a total of 0.05inches, the clock-hour rainfall rate would be0.05 inches h-1. This is true, even thoughinstantaneous rates may have ranged fromzero to some value greatly exceeding theclock-hour rate. Thus, a station with a meanannual rainfall of 100 inches (2500 mm)could expect approximately 350 clock-hoursper year with an hourly rate of 0.05 inches(1.3 mm) or more, 49 hours of 0.5 inches(12.7 mm) or more and two hours of 2inches (50 mm) or more. Winner’s regres-sion equations for various hourly rainfalls

are given in Table 6-6 along with the stan-dard error of the estimate and the correla-tion coefficients between the estimated andobserved frequencies. The best relationshipswere for rates from 0.02 to 0.5 inches (0.5 to12.7 mm) h-1. Forecasters could use the tableto estimate the chances that flooding or rain-reduced visibility would interfere with orstop airport operations.

12. Extremes. The tropics experience ex-tremely heavy rain much more often thanhigher latitudes. Figure 6-55 shows theworld’s greatest observed point rainfalls forvarious time periods. Cherrapunji (25°N,92°E; 4300 feet (1313 m) altitude) holds all ofthe records for periods of four days and for30 days and above, while three mountain-ous stations on Reunion Island (21°S, 56°E)hold the record for nine hours to 15 days(except for four days). The envelope curvefor these extremes is given by R = 16.6D0.475

where R is the rainfall in inches; and R =42.16D0.475 where R is the rainfall in centime-ters. D is the duration in hours. This issimilar to a formula developed earlier byFletcher (1950) who showed that extremerainfalls (inches) were proportional to the

———————————————————————————————Table 6-6. Empirical relationships between the mean annual rainfall given by X and themean number of hours per year given by Y that various clock-hour rainfalls are equalled orexceeded. The standard errors of estimate (in hours) and the correlation coefficientbetween the estimated and observed frequencies are also shown (after Winner, 1968).

Rainfall Rate (h-1) Regression Equation Standard Correlationinches mm Error (h) Coefficient0.01 0.3 Y = -219.8 + 0.45X 219.4 0.790.02 0.5 Y = -42.1 + 0.24X 65.0 0.920.05 1.3 Y = -6.8 + 0.14X 32.1 0.940.10 2.5 Y = 4.2 + 0.09X 16.8 0.960.20 5.1 Y = 5.3 + 0.05X 9.3 0.960.25 6.3 Y = 5.7 + 0.04X 7.0 0.960.50 12.7 Y = 6.9 + 0.016X 6.0 0.870.75 19.1 Y = 2.9 + 0.008X 4.6 0.781.00 25.4 Y = 2.5 + 0.005X 3.6 0.641.50 38.1 Y = -0.6 + 0.002X 2.2 0.532.00 50.8 Y = -1.1 + 0.0011X 1.3 0.442,50 63.5 Y = -0.6 + 0.0004X 0.7 0.36

square root of duration (hours) according tothe formula R = 14.3D0.5. The correspondingequation for centimeters is R = 36.32D0.5.

Many of the tropical daily rainfall extremeshave been associated with intense tropicalcyclones, e.g., the world record 24-hour fallof 73.6 inches (186.9 cm) at Cilaos, Reunion;49.1 and 47.0 inches (124.7 and 113.4 cm) attwo stations in the mountains of Taiwan(Paulus, 1965) and 42.2 inches (107.2 cm) atKadena Air Base (26°N, 128°E) (Jordan andShiroma, 1959). This occurred as TyphoonEmma passed on 8 September 1956 and issomewhat surprising since Kadena lies nearsea-level on the fairly low-lying island ofOkinawa (compared to Reunion and Tai-wan). Jordan and Shiroma pointed out thatfor a long time the station lay under the wallcloud and a feeder band of the typhoon. Onthe other hand, a coastal station on northernKauai, in Hawaii, recorded more than 38inches (96.5 cm) in 24 hours from a winterstorm. All records for 5 to 15 days werebroken at Commerson, Reunion (Chaggar,1984). From 14 to 28 January 1980 tropicalcyclone Hyacinthe moved slowly clockwisearound the island, the track looping twice.The 15-day record of 252 inches (643.3 cm)almost equalled the 30-day record held byCherrapunji.

Extremely heavy rains can also fall in rela-tively innocent-looking situations. Forexample, over 50 cm were recorded overparts of southeast Florida in less than 24hours on 21 January 1957. Sourbeer andGentry (1961) found little synoptic-scalejustification for the downpour, thoughupper tropospheric divergence overlaylower-tropospheric convergence. The con-tinuous thunderstorms described in SectionC-1b are other examples.

Daily amounts are the most commonlyavailable rainfall data. As a result, moststudies of extreme rainfall have concen-trated on daily amounts. The Gumbeldouble-exponential distribution is the statis-

tical model most often applied to extremevalues (Gringorten, 1960; Gumbel, 1958). Itfits well the annual extremes of wind speed,rainfall, temperature, pressure, and othermeteorological elements. It can sometimesbe applied to extremes within a month orseason. To determine the Gumbel rainfalldistribution for a station, find for each yearof the record the rainfall for the wettest day,then order these from lowest to highest. Thecumulative probability will proceed bysteps = 1/n, where n = the number years ofrecord. Tests have shown that for a varietyof stations, the points fall on a straight line,if plotted on extreme probability paper witha double exponential grid. Probable returnperiods, longer than the period of record,can then be estimated by extending the line.Remember that this estimate is based onlyon what has gone on before. It gives a prob-ability forecast.

The Gumbel distributions of extreme dailyrainfalls at some tropical stations are pre-sented in Figure 6-56. Figure 6-56A showsthe distribution for Nha Trang (12°N, 109°E)for 49 years of record. The solid line is thedouble-exponential distribution fitted to theextremes. The abscissa is the annual extremedaily rainfall, the left-hand ordinate is thecumulative probability (F), and the right-hand ordinate represents the cumulativeprobability expressed as a return period (R),based on the relationship R = 1/(1-F). Theplot shows that on average, once every twoyears Nha Trang can expect daily rainfall toreach about 5 inches (125 mm), while duringa 100-year period slightly over 14 inches(355 mm) could be expected. Figures 56Band C show similar graphs for other stationsin Vietnam and West Malaysia (Lockwood,1967). East coast stations are wettest becausefrom September through December theymay be hit by typhoons or experience con-tinuous thunderstorms. In Central andSouth America (Figure 6-56D) (Griffiths,1967), Caracas (10°N, 67°W), in a high pro-tected valley inland from the north coast,

differs from La Asuncion (11°N, 64°W) onan offshore island. The wetter San Salvador(14°N, 89°W), in contrast to the Venezuelanstations, is affected by developing Pacifictropical cyclones.

For other tropical locations, Fletcher (1949)has analyzed extreme rainfalls over Venezu-ela, and the U.S. Weather Bureau has stud-ied extreme rainfalls in Panama (1943), theHawaiian Islands (1962), Puerto Rico andthe Virgin Islands (1961). These studies weremade primarily to support hydrologicalengineering but are also valuable in meteo-rological planning.

13. Rainfall Associated with TropicalCyclones. Extremely heavy rain falls be-neath the eyewalls and the rainbands ofintense tropical cyclones (see Section 12 andFigure 6-55). Although measurements arescanty over the open ocean, Gray (1978)reported averages of rain gauge readings inthe vicinity of tropical cyclones over Florida(Miller, 1958) and the west Pacific (Frank,1976), while Riehl (1979) gave measure-ments from Hurricane Donna (1960) and hiscomputations based on moisture conver-gence toward the center. Figure 6-57 showsthat the four disparate procedures giverather similar results between 50 and 250NM (100 and 500 km) from the cyclonecenter. Rainfall is an order of magnitudegreater at the inner radius than at the outerradius. Between 50 NM (100 km) of thecenter and the eyewall, hurricane windsprohibit accurate measurement. Riehl’scalculations suggest a further order ofmagnitude increase in rainfall over thisdistance.

Goodyear (1968) investigated the rainfall of46 tropical storms and hurricanes crossingthe low-lying United States Gulf Coast from1940 through 1965. Figure 6-58 shows therainfall with respect to the cyclone centerpositions for the 46 storms averaged from 24

hours before to 24 hours after land fall. Themaximum of 6 inches (150 mm) occursslightly inland at 20 to 40 NM (40 to 80 km)to the right of the storm track, where thewind is blowing onshore. This fits the dataof Figure 6-57 quite well. In the successive12-hour sub-periods of the 48-hour period(not shown), the center of maximum rainfallmoves from offshore in the first period to 43to 62 NM (80 to 115 km) inland in the fourthperiod. The maximum 6-hour point rainfallaveraged 4.2 inches (107 mm) and rangedfrom 0.95 to 11.0 inches (24 to 280 mm). For12 hours the average was 5.9 inches (150mm) and the range 1.47 to 15.55 inches (37to 395 mm). Where rugged terrain lifts thestrong onshore winds, the average maxi-mum rainfall can be much greater.

Figures 6-57 and 6-58 can be used in antici-pating areal rainfall distribution from atropical cyclone. When a meteorologist isasked to make a point forecast, however, thedata from Kadena Air Base (26°N, 128°E)should be kept in mind (Detachment 8, 20thWeather Squadron, 1988). Between 1945 and1988, 31 typhoons gave maximum sustainedwinds of 50 knots or more at Kadena. Fromthree typhoons whose eyes passed over thestation, rain amounted to 3.89 inches (98mm) from one, and over 10.46 inches (266mm) from each of the others. On everybearing of the eyes from Kadena and an eyeapproach distance of zero to 140 NM (260km), at least one total exceeded 10 inches(254 mm), though many others did not. Onecenter that passed 7 NM (13 km) to thenorthwest gave only 2.63 inches (67 mm).When Kadena lay under a rainband forseveral hours, totals were large; they weresmall when the station lay mostly betweenrainbands.

Chapter 7Chapter 7Chapter 7Chapter 7Chapter 7DIURNAL VDIURNAL VDIURNAL VDIURNAL VDIURNAL VARIAARIAARIAARIAARIATIONSTIONSTIONSTIONSTIONS

A. General.

In the tropics, except for tropical cyclones,surface pressure gradients and wind andtemperature gradients are small. In fact, thestandard deviations of these elements (Fig-ures 4-1 and 5-1) are about the same as thevariations imposed by the sun’s diurnalcycle. Thus forecasters must always under-stand and take account of diurnal variations,because often they may overwhelm changesaccompanying synoptic or mesoscaleweather systems. For example, duringsummer between 10 and 11 LT, stationsalong the east coast of West Malaysia neverrecord rain, so powerful is the diurnal cycle.

B. Temperature.

The diurnal variation in temperature, fol-lowing the daily march of the sun, is gener-ally largest at the surface, and by affectingair density, determines to a great degree thenature of the diurnal variations of pressure,wind, humidity, cloudiness and rainfall(Buchan, 1883). These in turn, by modifyinginsolation and outgoing longwave radiation,help determine the amplitude of the tem-perature cycle.

The mean diurnal temperature variation atany station depends on a number of factors:atmospheric humidity, cloudiness, windspeed, topography, etc. Measurements madeon a buoy-mounted mast, 26 feet (7.8 m)above the surface in the North Atlantictradewinds (Prümm, 1974), showed anaverage daily range of 0.47°F (0.26°C),almost the same as the diurnal variation of

the sea surface temperature (Figure 7-1).Temperature was measured in two instru-ment shelters on Willis Island (16°S, 150°E;size 1300 X 500 feet (400 X 150 m); elevation30 feet (9 m) during tradewinds (Neal, 1973)(Figure 7-2). At the upwind end of the island(beach site), the diurnal variation amountedto 1.3°F (0.7°C) and at the middle of theisland to 5.4°F (3°C) (Figure 7-1). Thus, anisland about the size of a supertanker cansignificantly heat the air, while shallowwater and occasionally exposed reefs cantriple the diurnal range observed over theopen ocean.

The range is 10°F (5°C) or less at smallislands or along coasts with a prevailingonshore flow; 10 to 20°F (5 to 10°C) atcoastal stations affected by land breezes andat interior stations during the rainy season,and greater than 20°F (10°C) at interiorstations during the dry season. Figure 7-3illustrates the mean diurnal temperatureand dew point changes at a tropical conti-nental station (Luang Prabang, 20°N, 102°E)during dry (March) and wet (August)months and on a coral island (Eniwetak,11°N, 162°E). The annual curves at Eniwetakare typical for any month. In general, thediurnal curves show a minimum near sun-rise and a maximum in the afternoon; how-ever, they depart significantly from a sinecurve. Hence, the average of the mean dailymaximum and minimum temperatures willslightly overestimate the mean temperaturedetermined by averaging all the meanhourly values. The diurnal range of 25°F(14°C) at Luang Prabang in March is morethan double the range of 12°F (6.7°C)

in August when rainfall is an order of mag-nitude more. In contrast, the mean diurnal

range at Eniwetak is only 5°F (2.8°C), butstill ten times as much as over the openocean. For these stations, the dew pointvaries in phase with the temperature al-though the mean diurnal range is much less.At Luang Prabang dew point increasesrapidly from 08 to 10 LT, then stays rela-tively constant until after sunset, when itbegins to decrease. This “square-wave”effect may result from the sun’s heat dissi-pating a nocturnal inversion. Over humidcontinents and on the windward sides ofmountainous islands, dew point is highestin the afternoon and lowest in the earlymorning, similar to the curves shown inFigure 7-3. But over dry regions and on thelee sides of mountainous islands, this diur-nal cycle is reversed (Palmer et al., 1955).

C. Pressure.

Mean hourly pressures roughly follow asystematic sequence, characterized by twominima near 04 and 16 LT and two maximanear 10 and 22 LT. Figure 7-4 shows thatamplitudes are greatest over the equatorand are about half as large at 30° latitude.Tidal forces generated by the sun, and thealternate heating and cooling of the air,account for most of the diurnal variation.Thus the magnitude of the mean diurnalpressure range is closely related to the meandiurnal temperature range, which in turndepends on the mean water vapor content,cloudiness, and surface moisture, that con-trol solar heating and cooling by terrestrialradiation. Figure 7-5 illustrates the effect ofcloudiness at Hong Kong (22°N, 114°E). Theclear day maximum is about 0.5 mb higherand the minimum about 0.3 mb lower thanon cloudy days.

Pressure tendencies can be used in extrapo-lating motion of pressure systems. In thetropics, the usually small pressure gradientscan be recognized locally only if the diurnal

variation is first removed from the stationdata. A further adjustment can be made forcloudiness.

D. Winds.

1. Open Ocean. The research vessel “Chal-lenger” cruised the world’s oceans from1873 to 1876, spending most of the time inthe tropics. Over the open ocean, windsunderwent a very small diurnal variationwith the maxima about midday and mid-night and a range of less than 1 knot(Buchan, 1883). Subsequent measurementsover the tropical oceans and upwind of verysmall coral islands (see Figure 7-2 for anexample) produced similar results. Magni-tude and timing of the variation stronglysuggest that the diurnal variation of pres-sure is responsible for the diurnal variationof wind.

2. Land-Sea Breeze. In the tropics, lower-tropospheric diurnal wind changes are mostpronounced near coast lines and over theland. Above this layer, wind changes arerelatively small. Land-sea breezes are themost familiar local circulations. Accordingto Defant (1951), along tropical coasts andthe shores of large lakes a diurnal cycle ofonshore and offshore winds occurs in re-sponse to large temperature variations overland and very small temperature variationsover water. As the land warms with respectto the sea, air over the land becomes lessdense than air over the sea and a surfacepressure gradient directed from sea to landis established. Since pressure in the land-anchored warm air column falls off withheight more slowly than in the cool aircolumn over the sea, at about 3000 feet (1km) above the surface the pressure gradientis reversed and is directed from land to sea.Thus, air flows from sea to land in thesurface layer and from land to sea above.

This sea breeze circulation sets in a fewhours after sunrise, continues during day-light and dies down after sunset. Later, asradiation cools the land with respect to thesea, the surface temperature and pressuregradients are reversed. Consequently, aseaward-directed land breeze begins toblow in the surface layer with a return flowfrom sea to land above. The land breezecirculation persists until after sunrise. Al-though the scale is much larger, the verticalcirculations associated with the monsoons(Figure 6-27) resemble the land-sea breezecirculations with the summer monsooncorresponding to the sea breeze and thewinter monsoon to the land breeze. Thesimilarity can be extended. Year-long up-welling of cold water along a coast, bymaintaining higher pressure over the seathan over the land, not only prevents mon-soons, but inhibits the land breeze andenhances the sea breeze. The effect is less-ened by nocturnal downslope winds gener-ated by mountains near the coast. Land-seabreezes are weak or absent along swampycoasts such as those of the Guyanas of SouthAmerica (Edson and Condray, 1989). Espe-cially when the prevailing flow is onshore,temperature differences across the shoreline are small. The sea breeze may extendinland up to 25 to 50 NM (50 to 100 km), butthe seaward range of the land breeze ismuch smaller. In the vertical, the sea breezereaches altitudes of 4300 to 4600 feet (1300 to1400 m) in tropical coastal areas, and isstrongest a few hundred meters above thesurface. In contrast, the nocturnal landbreeze is usually rather shallow, being onlya few hundred meters deep. The sea breezetypically reaches speeds of a few knots.

a. Observations. Many studies of the land-sea breezes have been based on surfacewind and cloud observations. At stationssome distance from the equator, the seabreeze or land breeze usually persists longenough for the earth’s rotation (Coriolisforce) to cause the wind direction to turn

clockwise in the NH and anticlockwise inthe SH. The lower the latitude (the smallerthe Coriolis force), the farther inland the seabreeze is likely to penetrate. As the seabreeze “front” moves inland, it is accompa-nied by a distinct wind shift, a pronounceddrop in temperature, and increase in relativehumidity. Convergence at the leading edgemay cause convection, and conversely atnight, convergence may accompany theleading edge of the land breeze. Sometimesstratus or fog is carried in from the sea.Some studies have included observations ofthe vertical structure of the circulations.

Leopold (1949) studied interaction of land-sea breezes with the prevailing easterlytradewinds and resulting cloud systems inthe Hawaiian Islands. He listed three factorsinfluencing the interaction — (1) the size ofthe land area, (2) the heights and shapes ofthe mountain ranges and their relation tothe height of the subsidence inversion, and(3) the aspect of the area over which theland-sea breeze develops (i.e., windward orleeward exposure to the tradewinds). Tothese should be added (4) the shape of thecoastline, (5) the strength of the tradewinds,and (6) the vertical profile of tradewindspeed. Leopold developed models of fourmajor types of interaction, named after thegeographic feature of the islands where eachfeature is most common. The original papergives further details.

Diurnal variations of surface winds for Hilo(20°N, 155°W) on the windward andWaikoloa Beach (20°N, 156°W) on the lee-ward coast of the island of Hawaii areshown in Figure 7-6. The hodograph foreach station in Figure 7-6 represents thedistribution of hourly wind vectors on apolar diagram. For example, a wind fromsouth (east) is plotted on the radius labeled360° (270°) with distance from the originproportional to the speed. The curve con-nects the ends of all the hourly wind vec-tors. The island of Hawaii penetrates the

tradewind inversion, and diverts flowaround its corners. This swirl, aided bymore leeward sunshine, produces a strongersea breeze at Waikoloa (against thetradewind) than at Hilo (with thetradewind), while the Hilo land breeze isstronger than its sea breeze (see also Figure7-16). Garrett (1980) used mountain slopemeasurements to develop a vertical-planemodel of the Hilo land-sea breeze (Figure 7-7). It shows the return flow branches.

An early study of the vertical structure ofland-sea breezes was made by vanBemmelen (1922), who used pilot balloonobservations for Djakarta (6°S, 107°E) at thehead of a bay. Figure 7-8 shows speed of theonshore or offshore winds at various levelsthroughout the day. The surface sea breezeblows from about 10 to 19 LT, becomingstrongest in mid-afternoon. Due to surfacefrictional effects, the sea-breeze appearsearlier and is much stronger several hun-dred meters above the surface than at thesurface. It deepens to about 4300 feet (1300m) near 21 LT after it has ended at the sur-face. Above the sea breeze, a pronouncedreturn flow from land to sea is centered near6500 feet (2 km). Near the surface, the night-time land breeze and its overlying returnflow from the sea are much weaker than thesea breeze system.

Hsu (1970) described a University of Texasfield study of the Texas Gulf Coast land-seabreeze, based on a mesoscale station net-work near Galveston (29°N, 95°W), duringthe summers of 1965 to 1967. Hsu, usingonly those days with very weak or zerosynoptic flow, synthesized a model of theland-sea breeze system which is shown insimplified form in Figure 7-9. The lowerportion of the onshore flow is the sea breezeand that of the offshore flow the landbreeze. The levels of maximum wind in eachcurrent and their approximate heights aredepicted by arrows. The elliptical shapesenclose lower and upper parts of the circula-

tion systems. The approximate level ofconvective condensation (900 mb) is shownby the horizontal dashed lines. At 09 LT, theair is still cooler over the land than over thesea and the land breeze is still blowing. By12 LT, the land has become warmer than thewater and the circulation has reversed. Atthis time a line of small cumulus may markthe sea breeze front. At 15 LT, the sea breezeis fully developed and showers may beobserved at the convergence zone markingthe sea breeze front 15 to 20 NM (30 to 40km) inland. Low-level divergence near thecoast causes subsidence there. At 18 and 21LT, the sea breeze is gradually weakening.By 00 LT, there is no wind at the surface, thesea breeze is barely evident a little above thesurface and a temperature inversion occa-sionally accompanied by fog appears overthe land. With the land now cooler than thewater a land breeze becomes well developedby 03 LT and becomes strongest near 06 LT.A weak land breeze convergence with a lineof cumulus develops offshore near sunrise.The land breeze continues until mid-morn-ing, when the sea breeze resumes. In thismodel, the land breeze in the near-surfacelayer is as strong as the sea breeze. Butbecause of day-night differences in stabilityand friction, the daytime sea breeze at thesurface is considerably stronger than thenight-time land breeze. These differencesalso account for the vertical circulationbeing deeper for the sea breeze, with thereturn flow above the convective condensa-tion level.

Figure 7-10 shows a well-developed seabreeze along a relatively straight coast,bordering a flat coastal plain. As with theTexas Gulf Coast model of Figure 7-9,clouds reveal that the sea breeze circulationextends over about 50 NM (100 km).

In general, the larger the land mass and thehigher the mountains, the stronger will bethe sea and land breezes. Sea breezes areenhanced and land breezes are retarded

along windward coasts; the opposite is truealong leeward coasts (see Section 5). Acrossa bay (headland or isthmus), sea breezesdiverge (converge), while land breezesconverge (diverge). b. Numerical Models.Estoque (1962) derived numerical models ofthe sea breeze for different synoptic situa-tions. The models are applied to a 6500 feet(2 km) thick layer across a straight coast andare integrated to obtain time and spacevariations of winds and temperatures. Thedriving force for the circulation is a sinefunction temperature wave applied to theearth’s surface initially at 08 LT. Figure 7-11shows the results of the integration at 11and 17 LT for three initial geostrophicwinds: zero, offshore at 10 knots and on-shore at 10 knots. The horizontal componentof the wind vectors is proportional to windspeed, as shown by the scale in Figure 7-11;however the vertical motion scale is greatlyexaggerated. With zero initial wind the seabreeze forms near the shore and movesabout 15 NM (30 km) inland by 17 LT. Withan offshore initial wind the sea breeze isabsent at 11 LT, but by 17 LT has formed andpenetrated about 5 NM (10 km) inland. Withan onshore initial wind no pronouncedvertical motion develops up to 25 NM (50km) inland since the onshore advectioninhibits the temperature from rising there.Note the resultant vertical motions are mostpronounced in the zero and offshore pre-vailing wind cases. By afternoon, this typi-cally results in a line of cumulus inland andclearing over the shoreline. The modelunderlines the importance of synoptic flowin determining the character of the seabreeze.

McPherson (1970) expanded Estoque’smodel to three dimensions and applied it tothe Texas Gulf Coast sea breeze. He simu-lated Galveston Bay (29°N, 95°W) by using asquare indentation of an otherwise straightcoast. In the numerical results, the seabreeze convergence and the associatedvertical motion fields are distorted. The

greatest upward motion first developsnorthwest of the bay and later shifts tonortheast of the bay. The distortion causedby the bay becomes damped as the conver-gence zone moves inland. The preference forconvergence and convection northwest andnortheast of Galveston Bay has been verifiedby observations (Hsu, 1970).

Pielke (1974) used an eight-level three-dimensional primitive equation model todescribe the life-cycle of the south Floridasea breeze during late spring and summer.Though the roughness difference betweenland and sea was included, Pielke foundthat the difference in heating between landand water was chiefly responsible for thesea breeze convergence patterns. One mustremember that most land surfaces are muchrougher than south Florida. In the model thesea breeze is confined below 11,800 feet (3.6km). With the most common synoptic flowfrom southeast, convergence first developsalong the east coast and moves slowlyinland (Figure 7-12). Convergence along thewest coast sets in about two hours later, andremains almost stationary. Confirming themodel, Miami often has early morningshowers. Skies clear, as cumulonimbusdevelop to the west; by late afternoon,heavy thunderstorms dominate the westernhorizon. With southwest synoptic winds,the outcome is reversed, with the west coastbecoming clear during the day and deepconvection in the east. Once more, observa-tions agree.

No model treated the complete land-seabreeze cycle, until Mak and Walsh (1976).They found that greater stability in the near-surface layer at night accounted for the landbreeze being weaker than the sea breeze.Numerical models have realistically simu-lated sea breezes between the ocean and flat,low land. Complexities introduced by rug-ged coastlines and mountains have beenaddressed with a two-dimensional model(Mahrer and Pielke, 1977). Observations

support their conclusions that the combinedland-sea breeze and mountain circulationsare stronger than either acting alone, and amountainous coast aids inland penetrationof a sea breeze.

3. Lake Winds. Large tropical lakes experi-ence pronounced land-lake breeze circula-tions. Over Lake Victoria (2°S, 32°E), con-verging land breezes cause the air to rise atnight, while diverging lake breezes causesubsidence and clear skies in the afternoon(Figure 7-13). The much smaller LakeOkeechobee (27°N, 81°W) in Florida exerts asimilar effect, reproduced in Pielke’s model.

Over Lake Victoria (Figure 7-14) (Leroux,1983), diurnal variation is modified by theenvironmental wind. The prevailingsoutheasterlies are stronger in April than inJuly. In the early morning they sweep acrossthe lake in April, but converge with a landbreeze in July. In the afternoon, lake breezesprevail everywhere. Similar patterns havebeen reported over other lakes. Figure 7-15shows that even the small bays and tributar-ies of Lake Volta (7°N, 1°W) are cloud-freeduring the afternoon, suggesting a veryweak environmental wind.

The larger the lake, and the weaker theenvironmental wind, the greater the ampli-tude of the land-lake breeze cycle. Evenextensive irrigated surfaces may producethe same effect as lakes (Hammer, 1970).

4. Mountain-Valley Winds. These have beendescribed by Defant (1951) and others. Theyare important local circulation features inthe tropics where insolation is great andsynoptic-scale wind regimes are weak. Thetemperature difference between heatedmountain slopes and air at the same altitudeover nearby valleys causes air to rise (ana-batic wind) along the slopes in daytime. Atnight, radiational cooling of the mountainsurface generates a reverse circulation anddownslope flow (katabatic wind). These

circulations are well developed on sun-facing slopes and weak or absent on shadyslopes. The upslope wind layer is generally300 to 600 feet (100 to 200 m) thick withcharacteristic speeds of 4 to 6 knots. Thenocturnal downslope wind is shallower andweaker. Figure 7-7 shows the slope windson the island of Hawaii. A more detailedstudy was made during the HawaiianRainband Project (HARP) in 1990. Figure 7-16 shows measurements at automatic sta-tions near dawn and in the afternoon duringa day on which a well-developed inversionlay near 6500 feet (2 km). Near dawn, landbreezes prevailed, except near the inversion,where calms were common. The strongestbreezes were in the direction of thetradewinds and around the northern andsouthern ends of the island. In the after-noon, the island blocked and diverted thetradewinds (see Figure 5-15), which swirledand accelerated around the ends, forming acyclonic circulation in the north and ananticyclonic circulation in the south, enhanc-ing the sea breeze near the center of the westcoast. Of the two trans-island saddles, thenorthern, 3500 feet (1060 m) below theinversion, funneled the tradewinds, whilethe central saddle, at the inversion level,obstructed the tradewinds. Winds were lightand variable near and above the inversion.Hastenrath (1985) gives an example for theMount Kenya (Eq., 37°E) area (Figure 7-17).During the day, divergence betweenupslope flows causes the valleys to be clear,and convergence results in clouds over themountains. At night, downslope flowsconverge in the valleys, where cloud islikely, and diverge from the mountain tops,which are usually clear. In an afternoonweather satellite image of South America(Figure 6-7B), lines of cumulonimbus alongthe western and eastern Andes are sepa-rated by relatively clear skies over the Alti-plano. No separation can be seen where theAltiplano is narrow. Consequently, Quito(Eq., 78°W; 10,226 feet (3120 m)) experiences

a sharp afternoon cloudiness and rainfallmaximum. A somewhat different effect(Lyons, 1979) is produced by Mt. Haleakalaon the island of Maui (21°N, 156°W). Sincethe peak penetrates the tradewind inversionby 3000 feet (1 km), maximum afternooncloud rings the summit (Figure 7-18). Atnight, convergence between the land breezeand the tradewind makes the windwardcoast cloudiest.

Land-sea breezes and upslope-downslopewind regimes reinforce one another. Forexample, in the Red Sea area (Flohn, 1965)steep escarpments rise abruptly close to thecoast, and generally fair skies allow maxi-mum surface heating. Figure 7-19 shows theresultant surface wind directions for twodaily periods (06 to 08 LT and 12 to 15 LT)during July and August. During summer,the resultant gradient-level wind blowsfrom northwest over the Red Sea and fromwest-southwest over the Gulf of Aden. Inthe morning, downslope-land breezes pre-vail causing convergence over the water; byafternoon, upslope-sea breezes cause diver-gence over the water and local convergenceover land. These circulations significantlyaffect rainfall and vegetation patterns. Thewest coast of Guatemala and the Mediterra-nean coast of Asia Minor have similar verystrong combined wind systems. As themorning advances and mountain wind andsea breeze develop together, and blow in thesame direction, a sea breeze “front” and itsassociated convergence effects can seldombe detected.

5. Vertical Mixing.

a. General. From the surface to the gradientlevel, friction decreases and wind increaseswith height. In the absence of local winds,the vertical gradient of wind speed is con-trolled by vertical mixing, which in turn iscontrolled by the lapse rate. Over the sea,lapse rate undergoes little diurnal variation,

but over land, surface cooling at nightincreases stability and reduces verticalmixing, while daytime surface heatingdecreases stability and increases verticalmixing. Thus, with no land-sea or moun-tain-valley winds, surface winds are weak-est at night and strongest during the day, asover an extensive plain. At Honolulu (21°N,158°W), during light tradewinds, a classicalland-sea breeze cycle prevails (Figure 7-19A). During typical tradewinds, no seabreeze appears; in fact, the tradewind fresh-ens during the afternoon when island heat-ing, by increasing vertical mixing, bringsstronger tradewinds down to the surface(see Figure 6-12 for typical tradewind pro-files), and so reverses the sea breeze effect(Figure 7-19B).

Observers have often reported sea breezesextending more than 50 NM (100 km) inlandby midday — impossibly far. On most ofthese occasions, synoptic wind and seabreeze blow in the same direction. Inland,vertical mixing after sunrise freshens thesurface wind; its direction suggests seabreeze, but its origin is local.

b. Low-Level Jets. Beneath strong LLJs overland, the effects of vertical mixing oftendominate other local winds. Typically, atBaledogle (3°N, 45°E) beneath the Somali jet,surface winds average less than 10 knotsaround dawn, and may gust up to 25 knotsin the afternoon (Walters,1988). Conversely,the jet is strongest in early morning andweakest in early afternoon.

Between 23 June and 3 July 1977, three-hourly pilot balloon observations weremade at Obbia (5°N, 48°E; 400 feet (120 m)elevation) on the east coast of Somalia, andat Burao (9°N, 45°E; 3420 feet (1043 m)elevation) on the plateau in northwesternSomalia (Ardanuy, 1979). Figure 7-21 showsthe Somali jet strongest at night near 2450feet (750 m) (much lower than the regionalaverage of about 5000 feet (1.5 km)) aboveground, reaching 49 knots at Obbia and 39

knots at Burao. Turbulent mixing during theday at Burao reduced the jet and increasedthe surface wind. By late afternoon, speedswere nearly constant with height. The se-quence differed at Obbia, on the coast,where daytime turbulence was less. The jetmaximum moved down to 800 feet (250 m)during the morning, and then rose. At thesame time, overland surface temperatureincreased 32°F (18°C) and caused the ther-mal wind to increase. Despite the very largeland-ocean temperature difference duringthe day, Obbia experienced no sea breezeand stayed under the control of the LLJ.

During relatively clear weather over theAmazon Basin, large rivers (> 8 NM (15 km)wide) affect the local circulation (Garstanget al., 1990). The river is 14 to 16°F (8 to 9°C)warmer than the forest canopy at night and4 to 6°F (2 to 3°C) cooler by day. At night, asurface inversion develops over the coolingforest. Reduced frictional drag leads to a jetdeveloping at 1000 to 2300 feet (300 to 700m) above the surface. Its speed reaches 30 to40 knots. Over the warm river, convectiveheating prevents a jet, winds are weakerthan above the forest, and so convergencedevelops inland from the upwind bank. Byday, since vertical mixing is greater over theheated forest, flow accelerates over the river,leading to convergence inland from thedownwind Bank (Figure 7-22). These pro-cesses cause relatively little rain along thebanks of large rivers; satellites have revealedless cloud over the rivers. The Congo Riverof central Africa should produce similareffects.

Even more complex interactions occur alongcoasts.

Onshore synoptic wind. In his study of theTexas land-sea breeze (described in Section2), Hsu (1970) selected 9 days that lackedany significant synoptic wind. For the 22remaining days of the periods studied, freshonshore synoptic flow prevailed, and thestrongest sea breeze occurred at midnight

(Eigsti, 1978). On these nights, a LLJ waspresent, with maximum speeds near dawnat about 3000 feet (1 km) altitude. The LLJ,in the same direction as the sea breeze,seems to have been responsible for increas-ing the weakening sea breeze after mid-afternoon. Eigsti suggested that at the coastonshore flow would hinder nocturnal cool-ing enough to prevent a near-surface inver-sion from developing there. Thus, the sur-face would be linked to the nocturnal jet(Figure 7-23). The jet, becoming super-geostrophic a short distance inland, and socausing pressure to fall downstream, wouldincrease the onshore surface wind (the seabreeze) at the coast. At Bombay (19°N, 73°E)in May, westerlies prevail. Their associatedLLJ enhances the sea breeze, which in-creases through the evening.

Offshore synoptic wind. When the synopticwind blows away from the coast (Figure 7-24), surface wind speed along the coast is inphase with LLJ speed, and now the landbreeze is strongly enhanced, reaching amaximum near dawn, and the sea breeze isweaker than the land breeze. The northcoast of Somalia provides an example, withthe Somali jet enhancing the land breeze andopposing the sea breeze (Vojtesak et al.,1990).

Synoptic wind parallel to coast. In thisevent, the land-sea breeze is not affectedmuch at the shoreline, but the LLJ maycause complications a few miles inland,where the diurnal heating/cooling cycle issignificant. Detailed observations areneeded. In summer, LLJs parallel the shoresof the Persian Gulf.

The LLJ poses some problems for forecast-ers. Beneath it, airports may experiencegusty winds during the day, while at night,aircraft on approach may let down througha layer with large wind shear. The rapidlydiminishing wind causes a sudden andsometimes hazardous loss of lift.

So far discussion has concentrated on diur-nal wind changes at the surface and in thelower troposphere driven by the diurnalthermal cycle. Wallace and Hartranft (1969)used data from the surface to 100,000 feet(30 km) to extend our knowledge of diurnalwind variations in the free atmosphere.They computed annual mean 12-h winddifferences between 00 and 12 UTC (or 03and 15 UTC) at 27 pressure levels for 105radiosonde stations throughout the world.In polar latitudes, a simple pattern is ob-served with flow across the poles from thedaytime to the nighttime hemisphere. Thisagrees well with atmospheric tidal theory. Inlow- and middle-latitudes, however, thewind differences are strongly related totopography, with land-sea contrasts andterrain slope appearing to be in control.Above 900 mb at tropical stations, the aver-age 12-h wind differences were less than 4knots. This about equals the accuracy of theobservations and is smaller than hour-to-hour changes or differences over a few tensof kilometers. Therefore, it is doubtful ifthese slight diurnal wind changes in the freeatmosphere could be detected or applied inroutine tropical analysis and forecasting.

E. Cloudiness.

The previous section made passing refer-ence to clouds associated with diurnal windvariations, and Figures 7-7, 7-10, and 7-12 to7-18 should be kept in mind. In this section,attention is focussed on non-precipitatingclouds. Diurnal variation of deep convectiveclouds will be discussed in the followingsection on rainfall, since the former impliesthe latter.

1. Open Ocean Tradewind Cloud. Section 2reported that over the open ocean, surfacewinds vary semi-diurnally by about 1 knot.This is in response to the semi-diurnal

pressure wave. Brier and Simpson (1969)claimed that the resulting divergence of thesurface wind could account for diurnalvariations of cloudiness and rainfall at WakeIsland (19°N, 167°E). The small divergence(5 X 10-8 s-1) corresponds to a vertical motionof only about 0.02 cm h-1, very unlikely tocause variations in cloud or rain.

Brill and Albrecht (1982) used data from theAtlantic Tradewind Experiment (ATEX) todetermine diurnal variations of the sub-cloud mixed layer and the base of thetradewind inversion. This was highest atdawn and lowest in mid-afternoon with arange of about 260 feet (80 m), somewhatless than estimated by Holle (1968) for thetropical Atlantic (13°N, 55°W) in Augustand September 1963, and in line with whatKraus (1963) found. The difference betweencloud base and inversion base (a measure ofthe depth of tradewind cloud) was greatest(2800 feet (860 m)) at 04 LT. Diurnal varia-tion of the divergence of the surface windsmeasured in the ATEX array equalled 1 X 10-

6 s-1 with a maximum at 07 and a minimumat 19 LT. Brill and Albrecht then inserted thedata into a time-dependent one-dimensionalmodel of the tradewind boundary layer,while specifying long-wave radiative cool-ing. The model simulated the diurnal varia-tion of tradewind cloudiness rather well.Varying the external parameters confirmedwhat Refsdal (1930) had suggested — thatthe diurnal variation of radiational heatingand cooling at cloud top contributes most tothe diurnal variation in cloud depth andcloudiness and that divergence due to thesemi-diurnal pressure wave has a negligibleeffect. During GATE and ATEX, cloud depthand cloudiness were generally greatest near08 LT and least near 16 LT, with a range of14 percent (Figure 7-25). The values givenhere average rather diverse data. Theygenerally apply to shallow tradewind cloud.Preponderant evidence supports a morningmaximum and an afternoon minimum andcontrol by radiation.

2. Open Ocean Stratocumulus. These exten-sive sheets in the eastern Pacific and Atlantictradewinds (see Figures 6-1 to 6-4) arecapped at about 3000 feet (1 km) by a stronginversion. Minnis and Harrison (1984)analyzed GOES data for November 1978and found very uniform diurnal variation ofcloudiness for all the stratocumulus regions.Figure 7-26 is typical, with a dawn maxi-mum and an afternoon minimum. It re-sembles the tradewind cloudiness variation(Figure 7-25) but the range (43 percent) ismuch greater. The range of 2.8°K in thetemperature of the cloud top is equivalent toa range of 1000 feet (300 m) in the height ofthe inversion base. Turton and Nichols(1987) have reproduced the same diurnalvariation of cloudiness in a numerical modelthat emphasizes radiation. Similar resultswere obtained by Short and Wallace (1980),who found cloud tops to be relatively cold(high) in the morning and warm (low) in theevening in both stratocumulus andtradewind cloud regimes.

3. Open Ocean and Coastal Fog. Whenwarm moist air moves over cold water,generally near the coast, and the subsidenceinversion is some distance above the sur-face, air temperature may be lowered to thedew point (see Chapter 6, Section C-8). Aswith ocean stratocumulus, nocturnal radia-tion from fog top makes it deepest anddensest near dawn. Daytime heating of thetop reverses the process. Fog may extendinto coastal bays and river mouths and overlow-lying shore lines. It is rather commonover coastal south China in spring (seeChapter 6, Section B-6c). Fog dissipatesquickly over land as the sun warms thesurface and the air temperature rises abovethe dew point. At the same time, the fogthins over coastal waters, though the surfacetemperature is unchanged. Besides daytimeheating of the fog top, the vertical circula-tion of the sea breeze may contribute. Sink-

ing motion over the coastal waters (seeFigure 7-11) brings drier air to the surface,and causes the dew point to fall below thesea surface temperature, thus dissipatingthe fog (Ramage, 1971).

4. Overland Stratocumulus and Fog. Alongflat exposed coasts, stratocumulus is fairlycommon and follows the oceanic pattern;radiative cooling from the top of a moistlayer favors nocturnal development, whilewarming from above and convective mixingdissipate the cloud during the morning. AtLima (12°S, 77°W), drizzle (garua) fallsbetween 23 and 09 LT, from stratus thatdeepens during the night. This accounts foralmost all the rainfall, averaging 0.39 inches(10 mm) a year (Prohaska, 1973). Similarweather prevails along the west coast ofSouth America from 8°S to 25°S.

Over the Amazon basin, during the SHwinter, surface air within a cold surge(Chapter 8, Section D-4a) is heated andmoistened from below and is capped by a3000 feet (1 km) high inversion. As in thetradewinds over the eastern oceans, exten-sive nocturnal stratus develops, sometimesgiving drizzle. The cloud generally breaksup during the morning (Walters et al., 1989).In river valleys, cooling under clear skies,coupled with surface moisture and nearstagnant conditions, may lead to fog atnight. As with the stratocumulus, dissipa-tion follows sunrise. Radiation fog maydevelop over flat coastal areas borderingdeserts. Strong sea breezes bring moist airinland during the day. At night, surface aircools markedly under clear skies, and sincethe flat terrain hinders land breezes, fogmay form in the previously moistened air.The fog burns off quickly after sunrise. Thecoasts of Pakistan, the Persian Gulf, north-ern Angola, and eastern Somalia all experi-ence such fogs. The Guyanas are also af-fected during the dry season (September-October).

Duvel (1989), analyzing METEOSAT pic-tures, found extensive altostratus and al-tocumulus decks at about 16,000 feet (5 km)altitude over tropical Africa and the Atlan-tic, and suggested that they may be notuncommon elsewhere in the tropics. As withstratocumulus, coverage is greatest arounddawn.

5. Popcorn Cumulus. Figure 6-7 showspopcorn cumulus over the Amazon region.They are a regular feature, developing in theforenoon, sometimes becoming thunder-storms in the afternoon, and dissipating atnight. Small terrain gradients, leading toweak local wind regimes, account for therandom distribution of cloud elements. Thediurnal cycle is directly controlled by radia-tion; upward motion and condensationresult from daytime heating of the ground;radiational cooling at night favors subsid-ence, and that evaporates cloud.

F. Rain.

1. Open Ocean. Amounts of cloud thatrarely produce any rain respond to the dailycycle of radiation, being greatest in earlymorning and least in the afternoon.

Ruprecht and Gray (1976) analyzed 13 yearsof cloud clusters over the tropical westPacific and found that over twice as muchrain fell on small islands from morning (07to 12 LT) as from evening (19 to 24 LT)clusters. When rain was part of an orga-nized weather system, it was heaviest, anddiurnal variation most pronounced. Fu et al.(1990) used satellite IR images over thetropical Pacific to confirm and refine Grayand Jacobson’s findings. Deep convectivecloudiness was greatest around 07 LT andleast around 19 LT. Gray and Jacobson(1977) explained the observations in termsof radiational differences between the

cloudy disturbed area and its surroundings.Figure 7-27 shows that the clear surround-ings cool more by night than by day. Theonly large diurnal variation in the distur-bance is at the cloud top, where daytimeheating contrasts with large nighttimecooling. Lower tropospheric inflow andupper tropospheric outflow maintain up-ward motion in the disturbance and pres-sure surfaces slope accordingly. In Figure 7-28, the day-night radiation differencesenhance nocturnal low-level inflow andupper-level outflow and so intensify thedisturbance. The atmospheric response lagsthe radiational forcing by 1 - 2 hours.

Fingerhut (1978), using an axisymmetricprimitive equation model, confirmed thatdiurnal variations in radiative forcing of thedisturbance and its surroundings couldaccount for the observed diurnal variationin rainfall. The same processes were uncov-ered in typhoons (Muramatsu, 1983; Zehr,1987), Atlantic tropical cyclones (Steranka etal., 1984), and tropical cyclones in the Aus-tralian region (Lajoie and Butterworth,1984). Over tropical cyclones, the area cov-ered by very cold cirrus (< -65°C) is greatestin the early morning. The cirrus is thenadvected outward from the cyclone centerand is warmed and thinned by sinking.Thus the total area covered by cirrus reachesa maximum in late afternoon (Browner etal., 1977). Figure 7-29 from Zehr (1987)shows diurnal variations in the areas of“cold” and “warm” cirrus. Infrared imagesmade through the 11.2 µm window do notdelineate the two diurnal variations, butenhanced IR images allow the area of “cold”cirrus to be determined; after the diurnalvariation has been removed, increases(decreases) in this area precede tropicalcyclone intensification (weakening).

a. Squall Lines. During the Line IslandsExperiment (LIE) and GATE, when attentionwas concentrated on the tradewinds, squalllines were common rain producers, and

formed the basis of the model described inChapter 8, Section D-2 (Zipser, 1977). In theGATE array of research vessels centerednear 10°N, 23°E, the air was least stable ataround 03 LT (Albright et al., 1981). Whensignificant convection, generally in the formof squall lines, occurred, it was usuallyunder way about then. Since the squall linestook several hours to develop fully, maxi-mum rainfall was measured on the researchvessels at about 14 LT (McGarry and Reed,1978). The maximum surface convergencethat occurred at this time is consonant withsquall line intensity. Disturbances resem-bling squall lines may also have been re-sponsible for diurnal variations reported byother investigators, for example, Reed andJaffe, 1981; Augustine, 1984; Albright et al.,1985; Duvel, 1989; Shin et al., 1990. Over theAtlantic and Pacific near equatorial conver-gence zones, cold cloud tops were morecommon between noon and mid-afternoon,while the South Pacific convergence zone(SPCZ) experienced dawn and mid-after-noon maxima, as did the Pacific tradewinds.

Summing up, the morning rainfall maxi-mum associated with western Pacific cloudclusters and the early morning instability inthe tradewinds both originate from noctur-nal radiational cooling of cloud tops. In thewestern Pacific cloud clusters, which remainunder radiational control, rainfall dimin-ishes in the afternoon. In the tradewinds,middle tropospheric dryness favors distur-bances resembling squall lines. These formin the early morning, but in the severalhours they take to intensify, they are notunder radiational control; consequently therainfall maximum is pushed into the after-noon.

Majuro Atoll (17°N, 171°E) is generallydominated by the tradewinds, and experi-ences little diurnal variation in thunder-storms (Figure 7-30). These are likely only asweather disturbances pass, and so diurnalvariation is masked.

2. Land.

a. General. So far in this chapter, except foratmospheric tides and near-shore fog, diur-nal variations of pressure, temperature,winds, cloudiness and rain have been di-rectly forced by the diurnal variation of thesun’s radiation. Now as we move from smallislands to large islands to continents andfrom offshore to interior, the effect of theradiation cycle on the local winds — land-sea breezes, mountain-valley winds —determines the distribution of divergenceand vertical motion, and that in turn cancomplicate, counteract, and sometimesoverwhelm direct radiational forcing ofcloud development and rain. So, under-standing diurnal variation of rainfall de-mands knowing the local wind regimes.They may be powerful enough to blank outsynoptic or mesoscale influences, and oftendominate weather forecasting.

b. Synoptic-Scale Weather Disturbances.Under cloudy skies, surface temperaturedoes not vary much, and so diurnal varia-tion of local winds is damped out. Then, theradiation cycle, following the Gray-Jacobson(1977) model described earlier, favors anearly morning rainfall maximum and anafternoon minimum. This happens when thesummer monsoon is active over India(Prasad, 1970), and is also true during themonths of greatest rainfall over southeastAsia (Figure 7-31), except in the extremesouth of Vietnam where less cloudiness andconverging sea breezes cause an afternoonmaximum. In the wet monsoon regions,when disturbances weaken and cloudinessdiminishes, radiational heating of theground could cause a second maximum ofrainfall in the afternoon (none of the curvesof Figure 7-31 goes to zero during the after-noon). This transitional regime is typical ofsummer over east China and southwestJapan (Ramage, 1952a). During a break inthe monsoon rains, a showers regime gives

scattered thunderstorms and most rain fallsin the afternoon and evening (Prasad, 1970).

Over the flat Korat Plateau of eastern Thai-land, Ing (1971) used weather satellite pic-tures to stratify summer days into weakmonsoon (scattered clouds) and strongmonsoon (generally overcast). The diurnalrainfall variations were very different (Fig-ure 7-32). An afternoon maximum for theweak monsoon arose from convection in-duced by surface heating, while a nocturnalmaximum for the strong monsoon wascaused by destabilizing cooling of the cloudtop, enhancing convection.

c. Squall Lines. Over the Sahel and northernSouth America, squall lines account for mostof the rain (see Chapter 8, Section D-2). Theytend to develop during mid-afternoon,when convective heating is greatest; off thenortheast coast of South America nocturnalconvergence between land breeze andtradewind may also be responsible. Formany hours, and sometimes for more than aday after development, intensity and move-ment of a squall line are not controlled bythe diurnal cycle. Consequently, Leroux(1983) failed to find any pattern in the diur-nal variation of rainfall over West Africa; thesame is probably true for northern SouthAmerica.

d. Tradewind Regimes. Diurnal variations inrainfall frequency on the island of Hawaii(Figure 7-33) (Schroeder et al., 1978) cover awide range. Along the northeast coast,exposed to the tradewinds, rainfall is mostcommon at night, when a land breeze con-verges with the trades (see Figures 7-6 and7-7), and is rarest around noon, when thecoast lies in the sinking branch of the seabreeze circulation. On the sheltered westcoast, the night is dry as the land breezesweeps out to sea (Figure 7-6) while the seabreeze produces an afternoon maximum.Along the southeast coast of the island,parallel to the tradewinds, land breeze/tradewind convergence always lies offshore;

a sharp rainfall maximum stems from seabreeze/mountain wind lifting. At all inlandhigher-level stations, the moist mountain-sea breeze produces an afternoon maxi-mum, and the dry, sinking downslope flowa nocturnal minimum (note the dew pointdifferences between day and night in Figure7-7).

The averages of Figure 7-33 reflect re-sponses to a strong diurnal heating-coolingcycle. When an upper cloud layer covers theisland of Hawaii (moist-dry-moist sand-wich, Chapter 8 Section D-5a), nocturnalrain increases along the east coast. Rainfalling from the upper layer is not the cause,but rather reduced surface cooling. Thisdiminishes the land breeze and so brings theconvergence between it and the tradewindinshore from its normal offshore position(Chen and Schroeder, 1986) and enhancesthe coastal rain.

Kousky (1980) reported similar patterns fornortheast Brazil (Figure 7-34). This areausually lies south of the NECZ and of thesquall lines, and is relatively dry. The coastsexposed to easterly onshore flow experiencea nocturnal maximum from convergencewith the land breeze. 50 to 150 NM (100 to300 km) inland, rain is greatest during theday as the sea breeze advances inland andconvective heating develops. Farther inland,mountain-valley winds dominate (see Fig-ure 7-17) and since most of the stations arein valleys, they experience a nocturnalmaximum.

e. Dry Regimes. The diurnal variation ofmonsoon rainfall at 17 stations in the Sudanwas analyzed by Pedgley (1969). Only one ofthe stations was coastal; nevertheless, awide variety of types were found. An after-noon or early evening maximum dominatesonly far from the Ethiopian highlands. Atinland stations near the highlands, rain ismore evenly distributed through the day;however, at most places a weak afternoonmaximum resulting from radiational surface

heating can still be detected. A weak second-ary maximum in the early morning is alsoevident at many stations. The Sudan istopographically rather uniform.

f. Lake Regimes. The local winds over andaround Lake Victoria in Africa (see SectionD-3) and Lake Okeechobee in Florida (seeFigure 7-12) determine the diurnal variationof rainfall. Similar effects have been ob-served over Lake Nicaragua (12°N, 85°W)and Lake Maracaibo (10°N, 72°W) and canbe expected over other tropical lakes. Since alake is warmer than its surroundings atnight and cooler during the day, convergentland breezes cause nocturnal showers overthe lake. By day, diverging lake breezes areaccompanied by a rainfall minimum.

If the lakes are surrounded by mountains,very complex diurnal patterns are typical intheir neighborhoods. On the northern shoresof Lake Victoria (Figure 7-14), Entebbe (Eq.,32°E) experiences an early morning rainfallmaximum in April, when it is part of thecentral lake regime, and an afternoon maxi-mum in July after the lake breeze sets in. AtKisumu (Eq., 35°E), northeast of the lake onan inlet perpendicular to the environmentalflow, the land breeze always lies offshore.The afternoon rainfall maximum accompa-nies the lake breeze (Lumb, 1970).

g. Mountain Slopes. On the equatorialwestern Andes, maximum rainfall occursfirst (14 to 19 LT) above the condensationlevel, and later (19 to 24 LT) below thecondensation level (Walters et al., 1989).Similar separation has been observed overthe Ethiopian highlands (Vojtesak et al.,1990) and on the island of Hawaii. Thissomewhat surprising sequence may beexplained in terms of the local winds. Aftersunrise, the upper slopes heat first, andgenerate a stronger anabatic wind than thelower slopes. Rapid condensation producescloud close to or on the ground, and raindoes not evaporate before reaching thesurface. The cloud and rain, by cooling the

air and interrupting insolation, weaken andmay even reverse the upslope wind. Airstops rising and rain eases off during theafternoon. Katabatic flow from this levelnow converges with the anabatic wind(possibly reinforced by a sea breeze compo-nent) generated over the still sunny lowerslopes. Thus, the center of rising motion isshifted downward and the rainfall maxi-mum is displaced to the evening. The lowestslopes of the eastern Andes experience aneven later maximum, extending to dawn(Johnson, 1976)

h. Moist Near-Equatorial Regimes. Overwest Malaysia extensive weather distur-bances are rare and a weak Coriolis forceresults in less complicated land and seabreezes. In summer (Ramage, 1964), diurnalvariation of the wind leads to a cycle ofconvergence and divergence over the penin-sula (Figure 7-35). There is no simple rela-tionship to the diurnal variation of rainfall.This can be explained in terms of interac-tions between local winds, the prevailingsynoptic southwesterly wind and orogra-phy. West Malaysia is a roughly ellipticalpeninsula about the size and shape ofFlorida. The western and eastern ranges riseto 6500 feet (2 km) and are oriented alongthe peninsula. Mean August rainfall reachesabout 10 inches (250 mm) along the seawardslopes of both ranges. Figure 7-36 shows thedistribution of the diurnal variation ofrainfall.

Southwest Coast Malacca Strait Regime. Atnight, land breezes from west Malaysia andSumatra converge and cause upward mo-tion and rainfall in the Malacca Strait (the“Sumatra”). Since the prevailing winds arefrom southwest the Sumatra lies near theMalay coast causing a nighttime and earlymorning rainfall maximum there. A similareffect is observed at Bayan Lepas (15°N,100°E) on the east coast of the island ofPenang where downslope winds from theisland converge with land breezes from the

mainland to produce an early morningmaximum.

West Coast Regime. There is little interac-tion between the Malaysian and Sumatranland-sea breezes to either intensify orweaken the low-level synoptic-scale conver-gence. Consequently, diurnal rainfall varia-tion is slight.

Inland-Mountain Regime. During the dayconvective heating, reinforced near the coastby the sea breeze convergence, results in anafternoon rainfall maximum.

Mountain-Valley Regime. No hourly rainfallmeasurements were made in this area;however, downslope winds converging inthe valley at night should result in a noctur-nal maximum there.

East Coast Regime. In the east, a mountainslopes regime prevails (Section g). Anabaticwinds interacting with the southwest synop-tic wind, cause an early afternoon maximuminland. As the upslope flow dies downbeneath the rain and cloud, it is replaced bythe synoptic wind. Convergence between itand the sea breeze moves eastward to give alate afternoon rainfall maximum along thecoast.

Over the peninsula, the local circulations areso powerful that as much rain falls to “lee-ward” (NE) as to “windward” (SW).

In winter under prevailing northeasterlywinds, diurnal variation in the interiorremains unchanged, but the patterns arereversed along the coasts with maxima inthe early morning along the east coast andin the afternoon along the west coast(Nieuwolt, 1968). Similarly, the Pacific coastof central America, leeward of year-roundnortheast tradewinds, experiences a sharpearly evening maximum (Condray andEdson, 1989).

In western Colombia, prevailing winds blowonshore throughout the year. Along thecoast, most rain falls between midnight and

04 LT, and in the afternoon and eveningfarther inland (West, 1957).

In northern Taiwan, summer flash floodsoccur mostly at night, as downslope flowconverges with the prevailingsouthwesterlies (Kuo and Chen, 1990). AtClark Air Base (15°N, 121°E) on the centralLuzon plain, thunderstorms are about 30times as likely between 16 and 17 LT asbetween 07 and 08 LT (Figure 7-30). Theseexamples illustrate the role of the synopticwind. By enhancing the land breeze, anoffshore synoptic wind keeps coastal skiesclear at night, while during the afternoonand evening it spreads convective clouds,generated over the mountains, to the coast.An onshore synoptic wind converges alongor near the coast with the land breeze andresults in a nocturnal rainfall maximum,while during the afternoon an enhanced sea/mountain breeze keeps convection awayfrom the coast. Satellite views of centralAmerica for July 1985 (Figure 7-37) showthat at 19 LT, in the tradewind region northof Panama, mountain convection has spreadto the west coast of Yucatan and to thesouthwest coasts of Guatemala and ElSalvador, but is confined inland over west-ern Colombia, where surface westerliesprevail. Twelve hours later, at 07 LT, Theleeward tradewind coasts are clear andshowers are well offshore, but nocturnalconvection over the Mosquito Gulf hasspread inshore, as has a massive cloudsystem south of the Gulf of Panama.

The diurnal variation of rainfall is probablymore marked over and around the island ofSulawesi in Indonesia than any where else.Sulawesi comprises four long mountainouspeninsulas and three enclosed bays (Figure7-38). From geostationary satellite picturestaken during 15 days of the wet season,cumulonimbus centers were plotted for 05and 17 LT (Figure 7-39). In the early morn-ing, offshore thunderstorms are foundwhere land breezes converge. There are very

few over land where the land breezes di-verge. In the afternoon, by contrast, seabreezes prevail; they converge over land tocause thunderstorms and diverge over thebays, where skies are almost clear. In Figure7-40 the per unit area ratio of the number ofsea to land thunderstorms reaches a maxi-mum of 9.5 at 06 LT; numbers are the samenear local noon; and at 17 LT the ratio ofland to sea thunderstorm numbers (8.5) isgreatest. Figure 6-22, showing the meanannual outgoing longwave radiation valuesover Sulawesi are about 10-15 W m-2 lowerthan over the neighboring water, suggeststhat daytime convection is more intense anddeeper (colder cloud tops) than at night.Although Figures 7-39 and 7-40 were basedon December data, such intense local effectsshould prevail throughout the year. Thesatellite images for 20 July 1985 (Figure 7-41)confirm this. In the early morning easternBorneo and most of Sulawesi were clear ofconvection, which was especially intenseover the Macassar Strait and the Gulf ofTomini. Twelve hours later, heavy convec-tion dominated eastern North Borneo andthe northern peninsula of Sulawesi, whilethe seas were generally clear. Murakami(1983) analyzed GMS pictures for December1978 and January 1979 and found theSulawesi picture repeated (though lessstrongly) throughout Indonesia and thesurrounding seas, while Keegan et al. (1987)found similar patterns over eastern Indone-sia, the Timor Sea, and northern Australiaduring the summer monsoon.

G. Summary.

At sea, diurnal variation of surface tempera-ture is insignificant, but over land in thetropics, a forecaster should first predictsynoptic scale cloudiness. If scattered cloudsare expected, then a strong surface heating-cooling cycle will affect the local winds andvertical mixing, and through them controlwhere and when rising and sinking motions

and convection will develop. The sun willnot heat a damp or flooded surface as fast asa dry surface. Thus, recent rain can post-pone development of afternoon convection.By initially shielding the surface, earlymorning stratocumulus or fog can have thesame effect, while even cirrus can reduce theamplitude of the convective cycle. If theforecaster expects cloudy skies, the surfaceheating-cooling cycle and local wind devel-opments will be muted, and diurnal varia-tion of rainfall, now largely controlled byradiation at cloud top, will favor a nocturnalmaximum.

In regions where the mid-troposphere isrelatively dry, squall lines usually developwhen the diurnal cycle favors convection.However, squall lines move, intensify anddissipate largely independent of the diurnalcycle, and so tend to blur or shift the normaldiurnal variation of rainfall. Since diurnalrainfall patterns in tropical lands stem frommany factors, they cannot be simply catego-rized. Therefore tropical meteorologistsshould obtain climatological data for thestations of interest. For most first-ordertropical stations this is available inUSAFETAC standard summaries (see, forexample, Edson and Condray, 1989) or fromvarious national meteorological services.Remember, almost all first order stations arelocated on coasts or in valleys, and that ashort distance away, diurnal variations maybe quite different. Once the climatology ofdiurnal rainfall variation is known, physicalexplanations of the causal mechanisms canbe sought.

Lacking rainfall data, the forecaster can usegeostationary satellite pictures to develop auseful climatology. In the deep tropics, arelatively short period can provide usefulinformation, as was true for Sulawesi. Veryhigh resolution IR images from polar-orbit-ing satellites can pick up nocturnal fog orstratus, and the first signs of early morningconvection.

Chapter 8Chapter 8Chapter 8Chapter 8Chapter 8TROPICAL SYNOPTIC MODELSTROPICAL SYNOPTIC MODELSTROPICAL SYNOPTIC MODELSTROPICAL SYNOPTIC MODELSTROPICAL SYNOPTIC MODELS

A. General.

A synoptic model attempts to specify theidealized or generalized space distributionof meteorological elements, such as clouds,rainfall, wind, temperature, or pressure, in adistinct type of atmospheric system(Huschke, 1959). It also attempts to con-dense the general results of extensive em-pirical investigations so as to describe theessential kinematic, dynamic, and thermo-dynamic aspects of the particular type ofatmospheric system. The most importantmid-latitude synoptic model is the extratro-pical cyclone and associated air masses andfrontal systems, developed by the Norwe-gian school of meteorology during and justafter World War II. With minor variations, ithas stood the test of time as the basic synop-tic model for mid-latitude weather systems,and was little changed by cloud imageryfrom meteorological satellites. By the early19th century, tropical cyclones were knownto rotate about a center of low pressure andto travel east to west before recurving para-bolically eastward and poleward. Duringthe first half of the 20th century, tropicalmeteorologists tried to apply mid-latitudemodels. Out of this came various conceptsincluding tropical and intertropical fronts,and three fronts pinwheeling around atropical storm. None of these satisfied themid-latitude test of air mass discontinuity,although cold fronts may penetrate below15° latitude during winter. Before the adventof satellite meteorology, tropical meteorolo-gists thought that moving systems causedchanges in tropical weather. The key to agood forecast was to know what was up-stream. After the satellite arrived, weather

forecasting did not improve significantly.Changes within a weather system may affecta particular location much more thanchanges produced there by movement of theweather system. Extrapolating intensitychanges is harder than extrapolating move-ment and this has hindered the develop-ment of useful tropical models. Also, modelsadequate for one region may not be suitableelsewhere; the reader should keep an openmind and allow operational experience todetermine which concepts are most usefulto analysis and forecasting. As LaSeur(1964a) said in a survey of tropical synopticmodels, “In the literature one encounterssuch diverse opinions and interpretations,often of essentially the same data, plusconsiderable speculation and unwarrantedgeneralization or extrapolation of conceptsand results of limited validity, that theresulting impression is one of disquietingconfusion”. In surveying the origin of tropi-cal cyclones, Gray (1968) made a similarobservation: “There has been too muchqualitative and incomplete reasoning con-cerning the physical processes of develop-ment. General conclusions have been drawnfrom atypical case studies. Theories ofdevelopment have been advanced withoutsupporting data or plausible physical sub-stantiation. Numerical model experimentshave been made where initial assumptionsare not realistic”. Despite this confusion, it isevident that a variety of tropical synopticsystems are associated with areas of dis-turbed weather. One can have more confi-dence in a model that has been found towork in more than one region.

This chapter surveys various synoptic mod-els. The organization follows LaSeur’s

(1964a) division of synoptic models intothree broad classes : waves, vortexes, andlinear disturbances. The models will also beclassified according to weather, divergenceand vertical motion (Figure 6-32).

B. WB. WB. WB. WB. Waves.aves.aves.aves.aves.

1. Waves in the Easterlies. DuringWorld War II, Riehl (1948) postulated theeasterly wave theory. Frank (1969) proposeda variation called an “inverted V”. Neitherof these helped the tropical forecaster. Pastweather analyses from many tropical sta-tions show regular successions of easterlywaves across an entire ocean; however, evencareful post-analysis cannot maintain ad-equate time-continuity for individual waves.Any report of showers in the tradewindsused to justify introducing a wave axis nearthe report, just as “cold fronts” were calledon during World War II. So for many years,easterly waves conveniently “explained”most tropical weather. Evidence is nowlacking for the notion that tropical cyclonesform from easterly waves. Among weather-producing systems in the tropics, the “clas-sical” easterly wave, if it exists, is much lesscommon and more regionally restrictedthan once believed. For example, Simpson(1969) thought that easterly waves are con-fined to the North Atlantic, whereasAspliden et al.(1965-67) could find no evi-dence for them there in satellite and conven-tional data. Although Reed and Recker(1971) claimed to find waves in the westernPacific when the region was influenced by atrain of typhoons, careful analysts have beenunable to identify easterly waves anywhereoutside of the North Atlantic. Even there,associated weather lay sometimes east andsometimes west of the wave axis, thusfurther limiting its usefulness as a synopticmodel. For an up-to-date, rather conserva-tive opinion, the reader is referred to Riehl(1979).

It is now recognized that many waves in thelow-level easterlies stem from upper-levelcyclones (cold lows); therefore, the NationalHurricane Center at Miami uses the moregeneral term “tropical waves” (Simpson etal., 1968) in lieu of easterly waves. A tropicalwave is defined as “a trough, or cycloniccurve maximum, in the tradewindeasterlies. The wave may reach maximumamplitude in the low or middle troposphereor may be the reflection from the uppertroposphere of a cold low (Figure 8-22) orequatorward extension of a mid-latitudetrough”. In addition to these mechanisms,scanty observations have allowed thepoleward extensions of low-level, low-latitude vortexes to be mistaken for easterlywaves. Since there have always been moredata in the subtropical tradewinds than inlower latitudes (especially in the Atlantic), ithas often been easier for meteorologists totrack the waves than the low-level vortexeswith which they may have been associated.Similarly, that upper cold lows caused low-level tropical waves, could not be asserteduntil wind data from jet aircraft and satellitecloud pictures became available.

C. VC. VC. VC. VC. Vorteorteorteorteortexxxxxes.es.es.es.es.

1. General. Weather satellite data showthat cyclonic cloud and circulation patterns(hereafter termed “vortexes”) are muchcommoner in the tropics than was previ-ously thought. Some weak vortexes developinto tropical storms (maximum wind 34 to63 knots) or even hurricanes (maximumwind greater than 63 knots), while otherseventually dissipate without reaching thisintensity. Almost all vortexes can be de-tected in weather satellite pictures. How-ever, pictures alone do not reveal the tropo-spheric level of maximum cyclonic vorticity.To determine this, tropical meteorologistsneed to use all available conventional dataand have a thorough knowledge of the

regional and seasonal climatology of varioustypes of vortexes.

As Figure 6-32 shows, tropical vortexes canbe schematically arranged in terms of verti-cal motion. In fine weather vortexes (anticy-clones, heat lows, and upper troposphericcyclones (cold lows)), middle troposphericair subsides, while in bad weather vortexes(tropical cyclones, monsoon depressions,West African cyclones and mid-troposphericcyclones), middle tropospheric air rises.Thus, in their associated weather, heat lowsand anticyclones are more alike than heatlows and monsoon depressions.

Heat lows are described in Chapter 1, Sec-tion B, and tropical cyclones are treated inChapter 9. Circulations in the other vortexesare strongest and pressure gradients largestsome distance above the surface — 3000 to5000 feet (1 to 1.5 km) for monsoon depres-sions, 6500 to 10,000 feet (2 to 3 km) for westAfrican cyclones, 13,000 to 16,000 feet (4 to 5km) for mid-tropospheric cyclones, and33,000 to 39,000 feet (10 to 12 km) for coldlows. Following Sanders (1984), the firstthree (bad weather) types are classified as“hybrid” cyclones. All form near orequatorward of monsoon/heat troughs; theirenvironments are (for the tropics) stronglybaroclinic, with marked easterly shear. Mostrain falls equatorward of the center. Thesecyclones differ significantly from tropicalcyclones, which are most intense at thesurface, are embedded in a barotropic envi-ronment (small horizontal wind shear in thevertical) and have a much more uniformdistribution of rain around the center. Dou-

glas (1987) suggested that hybrid cyclonesmay resemble pre-hurricane depressions,well before an eye has formed. They havealso been observed in summer over theSouth China Sea and in the north Australianregion (McBride, 1987); Davidson andHolland, 1987).

2. Monsoon Depressions. During the Indiansummer monsoon the monsoon troughnormally lies over land (see Figure 6-19). Itoccasionally extends southeastward over thenorthern Bay of Bengal where monsoondepressions may develop in it after lowertropospheric southwesterlies freshen andweather deteriorates over peninsular India(Ramanathan and Ramakrishnan, 1932). Inan extensive discussion, Rao (1976) pointedout that monsoon depressions cause most ofIndia’s summer rain. They move betweenwest and northwest to central India beforeweakening or filling. Most rain falls in thesouthwest quadrant and is heavy at times.Table 8-1 gives average frequency of mon-soon depression development. For theseason, the range is from 3 to 13. Theyusually move west-northwest at 2 to 4 knotseast of 85°E and about twice as fast to thewest. All storms weaken to depressionintensity as they cross the coast. The trackenvelope is very narrow (Figure 8-1). Mon-soon depressions usually last between threeand four days.

Half the time monsoon depressions form ina southeastward extension of the monsoontrough over the northern Bay of Bengal. In

————————————————————————————————Table 8-1. Average monthly and seasonal frequency of Bay of Bengal monsoon depressions forthe period 1891-1970 (Rao, 1976). Depressions have winds up to 33 knots, storms from 33 to62 knots.

June July August September SeasonDepressions 0.9 1.3 1.7 1.8 5.7Storms 0.4 0.5 0.3 0.4 1.6

Total 1.3 1.8 2.0 2.2 7.3

15 percent of cases, development starts froma cyclonic circulation between 850 and 500mb. The remainder originate from weaklows that have moved westward acrossBurma. Many of these may be remnants ofSouth China Sea typhoons. According toRao, techniques for forecasting developmenthave not been established. Fortunately, sincedevelopment is restricted to the north Bay ofBengal, forecasters can concentrate onlooking for early signs there.

During summer MONEX, a monsoon de-pression that lasted from 5 to 8 July 1979was intensively observed and attempts weremade to model it. Sanders (1984) called it a“hybrid” cyclone that depended onbaroclinic processes (strong shear in thevertical between surface westerlies andupper tropospheric easterlies) for its devel-opment and maintenance. Thus, it lay be-tween tropical cyclones that “avoid” shearand mid-latitudes cyclones that dependentirely on shear. Douglas (1987) suggesteda schematic of the vertical circulation of theSMONEX depression (Figure 8-2). Upwardmotion is concentrated just south of thesurface low center, most of which is clear ofcloud. Douglas ascribed the asymmetricalcloud field to the large shear of wind in thevertical.

Figure 8-3 illustrates formation of a mon-soon depression. On 14 July 1966, the mon-soon trough lay near its normal positionover land. Two days later it extended outover the Bay of Bengal. The following day(17 July) a depression of 990 mb formed inthe trough and moved inland. The satelliteimage of Figure 8-4 shows a monsoon de-pression forming over the northwestern Bayof Bengal. Almost all the cloud lies south-west of the center and may have accompa-nied a surge that moved northeastwardthree days earlier (Figure 8-16). The mon-soon depression moved west-northwest-ward. It weakened and became indistin-

guishable from the monsoon trough on the16th.

During the transition months (April-May,October-December) the monsoon trough isfarther south. Lacking the strongbaroclinicity of summer, vortices that de-velop in it differ from monsoon depressions,and often reach tropical storm or hurricanestrength before striking land.

3. West African Cyclones. Cyclones mayoriginate over West Africa, but not in thesurface heat trough, which is controlled bysubsidence. Cyclones develop in a trough6500 to 10,000 feet (2 to 3 km) above thesurface, about 500 NM (1000 km) south ofthe surface trough. As with the mid-tropo-spheric cyclones discussed later, the heattrough may exert an indirect effect on cy-clone development. The cyclones, movingwest, generally do not have an associatedsurface-wind vortex until they near thecoast. Carlson (1969a) tracked a number ofdisturbances across Africa during Augustand September 1967. The 700 mb levelvortexes appeared to originate over Africabetween 10 and 15°E and moved westwardat 12 to 20 knots. Figure 8-5 shows the10,000 and 2000 feet (3 km and 600 m)streamline analyses and sea-level pressureanalysis for a day in August 1967. Thevortexes become strongest over the westcoast of Africa.

Perhaps enhanced instability of the mid-tropospheric jet accentuates growth here.Convection is concentrated within andahead of the vortex trough and by releasinglatent heat of condensation, promotesgrowth over West Africa.

In a later paper, Carlson (1969b) expandedhis synoptic analyses to the period 5 July to18 October 1968. He used these and theearlier analyses in synthesizing a model(Figure 8-6) of a typical disturbance near thecoast during the warm season. Mean cloudi-

ness is generally less than 10 percent near20°N, but over 30 percent near 5°N, and isslightly more along the 10,000 feet (3 km)wave axis through the vortex. In individualdisturbances, convection is most intensebetween 11 and 15°N; farther south cloudsare mainly dense and persistent stratocumu-lus. As the disturbances approach the coast,a vortex (separate from the string of vor-texes often found in the surface heat troughnear 20°N) often extends down from 10,000to 2000 feet (3 km to 600 m). The jet instabil-ity produced by the sub-Saharan barocliniczone and the accentuated convection areboth greater over Africa than over the ocean.

Leroux (1983) in his extensive treatise on theclimate of Africa, failed to find any evidencefor the vortexes described by Carlson, ex-cept off the west coast; the case studies hereproduced bear him out. He classified thewestward moving weather systems as squalllines; as these pass, surface easterlies replacemonsoonal westerlies. In any case, as thevortexes start to cross the Atlantic, theyweaken. Reed et al. (1977) composited dataas far west as 30°W from 20 August to 23September 1974 (Phase III of GATE). Theirvortex model differed little from Carlson’sand confirmed an earlier finding by Burpee(1975).

4. Mid-Tropospheric Cyclones. In certaintropical regions and seasons, cyclones reachtheir greatest intensity in the middle tropo-sphere. Two types have been studied: thecool season subtropical cyclones over theeastern North Pacific and North Atlantic,and the Arabian Sea cyclones, found nearthe west coast of India during the summermonsoon.

a. Subtropical Cyclones. Simpson (1952), inthe first detailed study of subtropical cy-clones, showed them to be a major circula-tion feature of the subtropics during the coolseason. These cyclones, known as “Kona

storms” in the Hawaiian Islands, originatewhen a closed upper low is cut off from themain stream of the upper-level westerlies.Simpson identified two sources for 76 Konastorms that were observed over the NorthPacific from November through March for aperiod of 20 years. Two-thirds developedfrom occluded cyclones trapped in lowlatitudes after being blocked by a warmhigh. The other third resulted frombaroclinic development of mid- and upper-tropospheric cut-off lows which graduallyextended their circulations to the surface.Figure 8-7 shows the surface pressure and700 mb contour charts for a subtropicalcyclone that formed from an occludedfrontal wave. In the North Pacific, subtropi-cal cyclones are usually confined between15°N and 35°N and 174°E and 140°W, andmove erratically (Figure 8-8). Simpson alsofound evidence of subtropical cyclones inthe North Atlantic (15°N to 35°N and 30°Wto 60°W) most often from Novemberthrough January (16 during a 10-year pe-riod). Simpson developed a compositemodel of the Pacific storms. Maximum windand rainfall occur in a quadrant 150 to 400NM (300 to 800 km) east from the center,depending on the size of the storm. Exceptfor the few subtropical cyclones that de-velop tropical cyclone

characteristics, small pressure gradients andlight winds prevail in the central region(within 80 NM (160 km) of the center). Intheir final stages, subtropical cyclonesgenerally move to where frontogenesis isfavored and generate a tropical wave. Occa-sionally, however, latent heat of condensa-tion, released near the center, causes them toacquire the warm-core properties of tropicalcyclones before they recurve and are ab-sorbed into the polar westerlies.

Ramage (1962), using conventional data andearly TIROS satellite pictures, also studiedthe wind and weather patterns of subtropi-cal cyclones. Figure 8-9 composites the

surface wind distribution around a sub-tropical cyclone during 4-5 April 1960. Thestrongest winds, largest pressure gradientsand greatest convergence occurred in themid-troposphere between 400 and 600 mb;however, upper-air data were too few for acomposite to be constructed. Figure 8-10presents a schematic radial cross section ofthe clouds, weather, and vertical motion inthis storm. The tropopause and the low-level subsidence inversion are also shown.Horizontal convergence is greatest near 600mb, resulting in general upward motionabove this level and sinking beneath, exceptfor the near-surface layer. The cyclone eyeappears to be rather large with scatteredclouds and little weather. Significantweather is confined to an annulus from 85 to270 NM (160 to 500 km) from the center.This model differs from Simpson’s, whichlocates most rain east of the center. This mayhave resulted from Simpson not distinguish-ing between true subtropical cyclones whichare completely cut off from the polar wester-lies, and large-amplitude troughs in thepolar westerlies. In the latter, bad weather isgenerally concentrated east of the troughaxis. Subtropical cyclones may sometimeslast several weeks. Because of their energy-producing character and the insignificanteffect of surface friction, subtropical cy-clones do not decay in situ but are eventu-ally absorbed by large-amplitude troughs inthe polar westerlies.

Figure 8-11 shows a GOES image of a NorthPacific subtropical cyclone that extended tothe surface. On the previous day it had beencut off in the southern end of a sharp troughin the upper tropospheric westerlies. Threedays later, it moved rapidly northeastwardahead of a new trough approaching fromthe west. b. Arabian Sea Cyclones. Thesesystems develop between 700 and 500 mb inthe monsoon trough lying across the north-east Arabian Sea and are the major produc-ers of rain along the west coast of India.Figure 8-12 shows the mean July position of

the monsoon trough in the surface layer andat 500 mb as determined from the resultant-wind field. Note the large slope of thetrough over western India and Pakistancompared to eastern India where monsoondepressions predominate. Above 400 mb,steady easterlies prevail. Figure 8-13 is ameridional cross section of July resultantzonal winds along western India. A sub-tropical ridge line overlies the surface mon-soon trough near 30°N. The trough slopesfrom there to the 500 mb level near 20°N.Baroclinicity is greatest south of 20°N wherethe tropical easterly jet overlies strong low-level westerlies.

Miller and Keshavamurthy (1968) studiedan Arabian Sea mid-tropospheric cyclonethat was probed by research aircraft be-tween 26 June and 10 July 1963. Compositeanalyses of wind, pressure, temperature,moisture, and weather were prepared usingthe daily cyclone center at 500 mb as theorigin of a movable coordinate system.Figure 8-14 shows the composite kinematicanalyses for the near-surface and the 600 mblevels. A weak trough near the coast is theonly evidence of a surface disturbance,while at 600 mb a cyclone is well developed.The composite temperature fields show thecyclone to be colder than the environment at700 mb and warmer at 500 mb. This re-sembles the thermal structure of subtropicalcyclones. Figure 8-15 shows a composite ofthe vertical-motion and cloud distributionassociated with the cyclone. The cumulon-imbus symbols show that convection wasbest developed slightly west and well southof the cyclone center.

Ramage (1966), applying Petterssen’s (1956)views on large-scale thermodynamic influ-ences, concluded that export of excesscyclonic vorticity generated in the heat low(monsoon trough) to the north could de-velop and maintain the mid-troposphericcyclone. Subsequently, numerical diagnosticmodels were applied to Miller and

Keshavamurthy’s data (see for example Carr(1977) and Brode and Mak (1978)). AlthoughRamage’s hypothesis could not be con-firmed, the models suggested that instabilityin baroclinic flow is a prerequisite for mid-tropospheric cyclone development.

Forecasters should be aware that thesecyclones often develop after a monsoondepression has moved inland to the east andstarted to fill. The heat low may also haveintensified and increased export of cyclonicvorticity to the south. Lack of aircraft data inthe middle troposphere often handicapsanalysis.

If the circulation of a hybrid cyclone extendsto the surface over the ocean, the surfacewind distribution can mislead a forecasterwho has a tropical cyclone in mind. AsFigure 8-9 shows, a large central region oflight winds may be surrounded by an annu-lus of stronger winds. Applying a tropicalcyclone model to ship observations or SSM/Iestimates in the strong wind zone, a fore-caster may overestimate wind speed nearerthe cyclone center. Although in Figure 8-15the satellite image does not show an eye, thecloud can sometimes look like a tropicalcyclone. Reconnaissance aircraft are rarelyavailable to settle the question, and no rulesof thumb have been developed. Such hybridcyclones have been observed in summerover the South China Sea. There and else-where, they have been mistaken for tropicalcyclones, especially when the only means ofdetection and identification are satelliteimages. Forecasters should err on the safeside, but be prepared for a few surprises.The monsoon and West African depressions,presumably influenced by the stationarymonsoon/heat trough, move rather steadilywestward. Elsewhere, hybrid cyclones moveslowly and irregularly. The best predictionfor them might be “no movement”.

5. Upper-Tropospheric Cyclones. SincePalmer (1951b) first reported tropical upper-tropospheric cyclones (cold lows), they havepuzzled meteorologists. We know that theyare most intense between 200 and 300 mband that their tangential winds may exceed50 knots. Since the lows are colder than theenvironment they weaken downward.According to Frank (1970) 60 percent do notreach 700 mb, and most of these (“drylows”) have little direct effect on surfaceweather. On 8 July 1979 (Figure 8-17) a coldlow northwest of Hawaii was enclosed by acirrus annulus. The cyclonic circulationneither reached nor affected the lowertroposphere, because undisturbed bands oflow cloud in surface westerlies extendedacross the center. Thin cirrus, on the periph-ery of most dry cold lows, allows them to bereadily recognized in satellite images. Occa-sionally, deep, short-lived cumulonimbusdevelop in the center of the low. Once con-vection is triggered near the surface, weakvertical wind shear and potential instabilityfavor development. However, dry air en-training into a cloud shortens its life. Thecloud systems of cold lows that reach thesurface (“wet lows”) are composited inFigure 8-18. While the central core is rela-tively cloud-free, a cloudy ring (heavydashed line) surrounds the center at a radiusof about 135 NM (250 km). The eastern partof the ring is twice as cloudy as the westernpart. In a Caribbean cold low studied byCarlson (1967), rain was confined south andeast of the center, in agreement with Frank’scomposite.

In wet lows air rises east of the center(where cloud and rain are concentrated) andsinks west of the center, similar to the distri-bution accompanying a trough in the upper-tropospheric westerlies (Carlson, 1967).Mostly, layers of middle and high cloudspredominate and these tend to inhibit con-vection, which is usually confined to thesoutheast quadrant. The dry subsiding air inthe cold low center is sharply delineated in

weather satellite moisture channel pictures.Figure 8-19 shows a cold low that hadmoved slowly westward across Hawaii.Deep convection in the eastern semicirclegave rainfalls of up to 8 inches (200 mm)over eastern Hawaii and eastern Maui. Aftercalculating kinetic energy production forboth dry and wet lows, Frank concludedthat both were “direct” systems, in whichcold interior air sank, and warm peripheralair rose. Cold lows, usually unconnected tohigher latitude energy sources, can persistfor several days. Especially for dry lows,where condensation heating is absent, thesinking, warming cold air must beradiationally cooled more than the environ-ment if the horizontal temperature andpressure gradients are to be maintained. Thedrier sinking core should cool more, butRiehl (1979) doubted if that is enough toensure that a cold low persists.

The tropical upper-tropospheric trough(TUTT) is a well-known summer featureover all oceans except the North IndianOcean. The TUTT incorporates a series ofcold lows, appears in response to a heatingcontinent, and is most intense just after mid-summer. It is usually oriented WSW-ENE inthe NH and WNW-ESE in the SH. As ashort-wave trough in the upper tropo-spheric westerlies passes poleward of theTUTT, a jet-like surge of cold air west of thetrough sweeps westward and equatorwardaround the TUTT, where the increase incyclonic vorticity causes a cold low to de-velop and may result in an apparent split inthe TUTT. On the eastward side of the low, asmooth, sharp-edged sheet of cirrostratusdevelops, while deep convection, withfrequent thunderstorms, occurs beneath andeast of the cirrostratus. In the NH the TUTTstarts developing in late April and is apersistent feature by June. It is most intensein August (Figure 8-20) and becomes rare bymid-November. Sadler (1976a) has mademany detailed studies of cold lows and theTUTT over the Pacific. From them, he devel-

oped a model showing interaction with low-level flow. Divergence 100 to 250 NM (200 to500 km) east of the upper center induces asurface wave-like disturbance that may bemistaken for an easterly wave. On rareoccasions, the disturbance becomes a ty-phoon. Figure 8-21 is a three-stage sche-matic of the model.

Stage 1. In the upper cell, 500 to 1000 NM(1000 to 2000 km) across, winds reach 40knots or more in all quadrants. Flow di-verges east and southeast of the center,generally accelerating downstream towardthe strong southwesterlies of the TUTT tothe east. Beneath it, the divergence induces atrough that extends up to 700 mb. This iswhere organized convection is first ob-served. Locating the clouds with respect tothe low-level trough is difficult because, atthis stage, the trough is weak. Whether theextensive convection, initiated from aloft,develops prior to the surface trough cannotbe determined from available observations.

Stage 2. Organized convection beneath theupper divergence east of the upper cellsignificantly expands. This in turn alters theupper-level flow as the resulting heatingdue to condensation builds a sharp ridgeeast-northeast of the cell, forces the cellnorthward and splits the TUTT. At the sametime, the cold low shrinks and no longerpenetrates down to the 500 mb level. In lowlevels, the wave develops into a weak de-pression with a closed vortex below 500 mb.The main convective cloud lies east of thelow-level vortex. Other clouds may bedistributed around the circulation.

Stage 3. Upper-tropospheric flow becomesmore complex as increased convectionintensifies the ridge or even builds an anti-cyclone whose outflow interacts with andfurther distorts the TUTT. In the westernend of the northern branch a small cycloneoften forms, with its own cloud system. Themodel only depicts it for stage 3, but it mayoccur in stage 2. In the lower layers, the

depression becomes a tropical storm extend-ing up through 500 mb and capped by ananticyclone. The upper outflow is not im-peded to the east where it merges with thelarger scale westerlies south of the TUTT,and so the cirrus shield is skewed towardthe east. North of about 15°N, storms thatare triggered by cold lows in the uppertrough cannot tap an upper-level outflowchannel to the south — into large-scaleeasterlies. Perhaps this is why these stormsseldom intensify into strong typhoons.

Between 27 and 31 July 1970, an upper coldlow moved westward across Midway Island(28°N, 177°W). In the time cross section(Figure 8-22) (Sadler, 1976a), easterliesprevailed below 700 mb. The upper lowcentered near 250 mb changed little inintensity. It was 4° to 5°C colder than itssurroundings. Beneath it, the weather wasfine and the air dry. Humidity rose after itspassage, as an induced trough in thetradewinds (135 NM (250 km) east of thecold low center) approached Midway, whereshowers developed. The low-level troughwas capped by an anticyclone at about 250mb. As the trough passed Midway, windsveered from east-northeast to southeast.This cross section fits stage 1 of Figure 8-21.Lacking upper-tropospheric data, a fore-caster could easily label the trough in thetradewinds an easterly wave.

Cyclones develop during the warm seasonin the TUTT over the South Pacific. TheJanuary position of this trough is shown inFigure 4-10. As over the NH tropical oceans,these cold lows occasionally trigger tropicalcyclones. Since the low-level monsoontrough in the South Pacific seldom extendseast of the dateline, it seems likely that mostof the rare tropical storm developments eastof this meridian stem from cold lows (exceptin the 1982-83 Niño—Figure 6-26).

Kousky and Gan (1981) described a SouthAtlantic summer TUTT, closely resemblingthe NH systems. Although the embedded

cold lows move erratically, those between10° and 15°S tend to move westward andsignificantly affect the distribution andintensity of convection over eastern Brazil.Lyons (1991) has identified a rather weakerTUTT over the southeast Indian Ocean,suggesting that it is a channel throughwhich mid-latitude waves can affect theweather of equatorial Africa. Jet aircraftobservations generally allow cold lows to beanalyzed and tracked. Lows that extenddownward into the middle tropospheretend to persist and to move steadily; shal-low lows are usually transient and moveerratically. After cold lows and the TUTT areidentified in the wind field, their cloudsignatures can be recognized in satelliteimages, especially the moisture channel.Since the lows are mostly free of cloud, theimages alone cannot provide unequivocalidentification. The pictures reveal the drycentral area as well as outbreaks of deepconvection east of the center — early evi-dence that the low is modifying weather andcirculation in the lower troposphere.

Hybrid storms and cold lows are puzzling.We do not understand why interactionbetween higher latitudes and the tropicssometimes results in cold lows and at othertimes in subtropical cyclones, nor do weunderstand what causes a cold low to inten-sify enough to produce bad weather on itseastward/equatorward periphery. Why doArabian Sea mid-tropospheric cyclonesrarely reach the surface, while those in theSouth China Sea often do, and in that sense,resemble subtropical cyclones? And whatcauses mid-tropospheric cyclones veryrarely to intensify into hurricanes, and in atleast one instance, for a cold low to be sotransformed (Chapter 9, Section E-5)?

All these cyclones are direct and energy-exporting, and do not readily dissipate. Incontrast to tropical cyclones, they are notlocally dependent on heat and moisture fortheir survival, and so often persist over land.

Forecasters can use satellite images to helpidentify hybrid cyclones. They lack eyes andgenerally spiral convective bands, usuallyappearing as amorphous masses of densecirrostratus. Cold lows may possess anannulus of cirrus around a clear center inwhich a few cumulonimbus may be scat-tered. More often, the cirrus is in one quad-rant or semicircle only, with a very sharpinner edge lying along a jet stream. Usually,by differentiating cloud top heights, IRimages are more informative than visualimages, while the moisture channel givesthe best picture of subsiding dry air in thecenters of cold lows.

6. Temporal Storms of Central America. Aphenomenon that may be variously causedby tropical depressions, hybrid cyclones orcold lows is the temporal of the Pacific coastof Central America and the north coast ofHonduras (Portig, 1976). Each location onthe Pacific coast experiences one or twotemporals a year. They are most likely inSeptember and October, with a secondarymaximum in June. They resemble hurri-canes in their rainfall and clouds; however,the associated low-level winds are relativelyweak. Pallman (1968) studied temporalsusing conventional and satellite data, butwas hampered by lack of upper-air dataover the eastern Pacific. His model of thelow-tropospheric contour pattern andweather distribution (Figure 8-23) indicatesextensive altostratus on the east side of thestorm with embedded cumulonimbus andcontinuous rain in the northeast quadrant.Thunderstorms are very rare. Sincetemporals often remain almost stationaryfor many days, they can be responsible forcatastrophic rain and major flooding, accen-tuated by orographic effects.

DDDDD. Linear Disturbances.. Linear Disturbances.. Linear Disturbances.. Linear Disturbances.. Linear Disturbances.

1. General. A linear disturbance is a synop-tic system in which the vorticity or diver-gence, or both, tend to be concentrated in azone whose length is much greater than itswidth (LaSeur, 1964a). Through the use ofsatellite imagery, these disturbances aremuch better understood. Before satellites,tracking them over the oceans was difficult.Conventional data are sparse and differ-ences in temperature, moisture and windare subtle. As a line passes, winds, cloudsand rain increase. Often, during World WarII, such weather was ascribed to fronts,although no air-mass discontinuity could bedetected. Later, easterly waves were blamed,but satellites showed how short-lived manyof the disturbances were. Despite the daunt-ing variety, especially between 10°N and10°S, the following line disturbances occurrather often — squall lines; remnant orreactivated cold fronts; shear lines;tradewind or monsoon “surges”; troughsuperpositions; near equatorial convergencezones.

2. Squall Lines. These are generally definedas non-frontal lines of active thunderstormsseveral to some tens of nautical miles wideand hundreds of nautical miles long whichlast much longer than the lifetimes of thecomponent cumulonimbus elements. In thetropics, they are mainly observed duringspring and autumn in monsoon regions.Based on squall lines he investigated duringthe Line Islands Experiment (LIE) in 1967and GATE in 1974, Zipser (1977) describedan archetypal squall line for the tropics(Figure 8-24). Quoting him:

“1. Relative winds are directed into the squallline from the front at all levels, with a minimumrelative flow often noted about 700-600 mb. Wellbehind the squall, the relative flow is stronglyoutward from the squall near the surface, stilloutward — but much less so — just above themixed layer, and from either direction in the

600-900 mb layer, with indications that the flowfrom the rear may increase with time.

2. New cumulonimbus cloud growth takes placeabove the squall front along the leading edge,with older towers joining the anvil mass trailingbehind. The anvil is 20,000 to 33,000 feet (6-10km) thick, and rain falls from the anvil base for50 NM (100 km) or more to the rear in a regionhaving no significant clouds below the anvilbase, which is near 13,000 to 16,000 (4-5 km).

3. At 500 feet (150 m) in altitude, the regionfrom 5 to 15 NM (10 to 30 km) behind the squallfront, where almost all active cumulonimbusclouds are located, has very cool, near-saturatedair. The remainder of the system to the rear hasair varying widely in w and in relative humidity.

On a Skew T diagram, the wet bulb potentialtemperature ( w) can be found by first plottingthe wet bulb temperature, and then followingthe saturation adiabat from that point to 1000mb, and reading the temperature at the intersec-tion. As with e and ev (Chapter 5, Section C-4)an atmospheric layer is potentially unstable(stable) if w decreases (increases) with height. 4.Soundings taken in the rear portion of the squallsystem show cool, near-saturated air of interme-diate w [21.5°C] near the surface [the mixedlayer]; a deep layer of relatively warm air withlow relative humidity and low w just above that;and near-saturated conditions again near anvilbase.

5. The stable layer which separates the twolowest layers (4,above) has its base, which marksthe top of the mixed layer, from 130 to 1600 feet(40 to 500 m), with 330 to 1300 feet (150 to 400m) rather common.

6. The winds in the surface layer and in themixed layer behind the squall line are divergentin the range of 1 to 5 X 10-4 s-1 over distances of15-55 NM (30-100 km). On the same scale, themesoscale sinking just above the mixed layer isin the 5-25 cm s-1 range at 1600 feet (500 m).

7. The dew point at the surface drops withsquall-front passage, but often drops still further

in the anvil rain area, reaching an absoluteminimum about 50-100 NM (100-200 km)behind the squall line.

8. A mesoscale high-pressure region at thesurface tends to accompany the squall line itself,followed by a mesoscale low-pressure regionabout 50-100 NM (100-200 km) behind thesquall line, but with considerable variations inindividual cases..... If the relative flow is fromthe front at 700 mb, much of the air must passbetween active cumulonimbus towers to arrivein the heavy rain area behind the squall andeventually in the anvil rain area....because thecumulonimbus towers are not continuous eitherin distance along the squall line or in time....Ifthe relative flow is from the rear, the air entersthe anvil rain area directly. There is evidencethat air in some squall lines enters from bothfront and rear, converging near the heavy-rainarea.... Evaporation of rain into this air is stillbelieved to be the most important factor allowingsome of it to sink several kilometers to within afew hundred meters of the surface”.

This model is similar to the tropical squallline model conceived independently byHouze (1977).

As mentioned in Chapter 2, Section D-5,convection is generated along the leadingedge of the cold downdraft air, while thedowndraft air to the rear of the squall linemay suppress convection. So the squall linemoves forward. For squall lines to develop,persist and propagate, they require not onlymoist surface air, but also relatively dry(low w) middle tropospheric air, which, onbeing cooled by evaporation from rain, sinksand establishes a cold front at the leadingedge of the squall line. Zipser (1969), in astudy of a cloud system that passed acrossthe Line Islands between 31 March and 1April 1967, described how unsaturatedconvective downdrafts in the rear of a squallline suppressed convection. The squall lineformed rapidly on the evening of the 31st,was most intense around midnight, andthen quickly dissipated during daylight of

the 1st. Since it developed where wind shearin the vertical was strong (westerlies replac-ing easterlies above 600 mb), Zipser con-cluded that mid-tropospheric air with lowequivalent-potential temperature was en-trained into the mesoscale system andcarried to the surface in organizeddowndrafts that thereupon inhibited furthercumulus production (Figure 8-25). In Figure8-25C the disturbance is dissipating and thedowndraft is maintained primarily by rainfalling from the extensive cloud shield. Atthis time satellite pictures revealed a thinline of cumulus moving outward from themain cirrus cloud mass. Successive positionsof this surface outflow boundary are shownin Figure 8-26. The main cloud system,which had appeared as a bright mass on themorning satellite picture, had almost com-pletely gone by 1711 LT. Sequences of satel-lite pictures have shown that where twooutflow boundaries intersect, large cumulusis likely to develop. Had the Zipser squallline been close to a coast, where orographycould be involved, intensification, ratherthan dissipation, might have ensued.

Over West Africa, where squall lines areresponsible for most of the rain, and evenfor generating summer duststorms (Leroux,1983) (Chapter 10, Section E-3), they maydevelop at the leading edges of anticyclonicsurges from mid-latitudes (see Section 4).The surges move in from northeast anddisplace the monsoon westerlies. The initiat-ing impulse may quickly dissipate. How-ever, in a numerical simulation (Rennick,1976), instability in the 700 mb easterly jetalong about 14°N favored squall line devel-opment. At the western edge of the squallline, low-level cold outdraft continuallyregenerates the convection, and since squalllines are embedded in an environmentpossessing easterly vertical shear, they maymove westward faster than the environmen-tal flow (Eldridge, 1957). Most rain falls inthe low-level westerlies, accelerated towardthe

squall line and lifted by the cold air. Asshown by Figure 8-27 (Leroux, 1983), amountain chain affects weather some dis-tance ahead of the line, generating showersto windward, and subsident drying toleeward.

Although some West African squall lineslast several days, 80 percent die out in lessthan 24 hours (Cochemé and Franquin,1967). They are most common from Junethrough October and best developed be-tween 10 and 15°N. Combining severalstudies, leads to the following rough statis-tics:

Time interval between squall line passages:2.5 to 5 days

Distance between lines: 1000 to 2000 NM(2000 to 4000 km)

Westward speed: 6 to 7° longitude day-1, buta few are much faster.

Squall lines are most likely to develop overthe southern Sudan, generally during theafternoon.

Figure 8-28 shows successive 24-hour IRimages of a West African squall line that wastypically convex toward the west. It movedwest at about 30 knots.

Over tropical South America, east of theAndes, squall lines resembling those of WestAfrica (Fernandez, 1982) are responsible formost of the rain (see Section 8b). This is alsotrue over the Australian tropics when east-erly winds prevail (Drosdowsky and Hol-land, 1987). Scattered synoptic observationsmade every six hours, or twice-daily pic-tures from polar-orbiting satellites, areseldom frequent or detailed enough to allowthe forecaster to maintain analytical conti-nuity, let alone to predict squall lines. Con-tinuous radar coverage or hourly geosta-tionary satellite pictures are essential if therapid intensity changes are to be monitoredor extrapolated. Even with these aids, fore-casts rarely succeed beyond a few hours.

Disturbances are likely to become squalllines where a moist lower troposphere isoverlain by a relatively dry middle tropo-sphere. Thus they are favored in thetradewinds, the general neighborhood ofheat troughs, and during the transitionseasons in monsoon regions. Squall lines areextremely rare in the deep, moist air of thesummer monsoon and within 5 to 10° of theequator.

The impact of squall lines on the diurnalvariation of rainfall is discussed in Chapter7, Section F, and their association withdestructive winds, in Chapter 10, Section E-2.

3. Surface Cold Fronts and Shear Lines.Cold fronts often penetrate to low latitudesover tropical continents during winter. Eventhough currents of polar air are usuallyrather shallow by the time they reach thesubtropics, surface temperature and dewpoint discontinuities and a shear line can bedetected to low latitudes because of re-peated nocturnal cooling in the clear, dry airmass behind the front (Stone et al., 1942).Over the oceans, satellite imagery andcontinuity facilitate tracking fronts. Thefront’s leading edge is usually marked by aline of active convection; a series of convec-tive lines may parallel the front on itspoleward side. Cloud tops average about10,000 to 15,000 feet (3 to 4.5 km). Associatedlow ceilings and rain may be enhanced byorography. Frontal shear lines occasionallyreaching Guam in winter cause extendedperiods of low ceilings and poor visibility.Figure 8-29 schematically relates the shearline to a cold front farther poleward. Afterbecoming stationary, the shear line coincideswith an asymptote of convergence withinthe tradewinds, and the mechanism de-scribed in the discussion of near-equatorialconvergence zones (Section 8 below) en-sures growth of cloud droplets and some-times rain. In shear lines, stratiform cloudsgenerally reach 8000 to 12,000 feet (2.4 to 3.7

km) with embedded convection. Whereshear lines intersect windward mountainslopes significant rain may fall (see Chapter6, Section C-1c). On 31 December 1987 and 1January 1988, a shear line, along which freshnortheasterlies converged with weak east-southeasterlies, lay stationary across theisland of Oahu in Hawaii. Orographic liftingby the Koolau Range, aided by a trough tothe west in the upper tropospheric wester-lies, built cloud tops to 19,000 feet (5.8 km).Although thunderstorms were absent,continuous rain amounting to up to 25inches (635 mm) in 24 hours fell over and towindward of the mountains on southeasternOahu (National Research Council, 1981).

Figure 8-30, for 00 UTC 25 December 1985,shows a satellite view of a cold front cross-ing Hawaii, where winds shifted fromsouthwest to northwest. Below 15°N thefront merged into a shear line, whose cloudtops ranged between 7000 and 12,000 feet (2and 3.5 km); tradewinds freshened as itpassed. To the east, shear lines had formedin deep southwesterlies beneath a subtropi-cal jet stream. During the normally drymonths of February and March, a shear lineoriginating from a NH cold front may reachthe coastal mountains of Venezuela and givecontinuous rain that is heaviest between 500and 1500m on the north-facing windwardslopes (Goldbrunner, 1963). This “inviernode las chicarras” strikingly resembles thetradewind rainfall periods on Mt. Waialeale(Chapter 6-C-1-c).

Shear lines are sometimes very hard toidentify, even with satellite pictures, andeventually they dissipate.

As described below, a cold front may sud-denly disappear within a strongequatorward surge.

4. Surges. For more than 50 years, tropicalmeteorologists have often observed rathersudden accelerations within major surface

wind currents such as the tradewinds andthe monsoons. Downwind of the accelera-tion maximum, convergence and increasedcyclonic shear to the left (right) of the maxi-mum axis in the NH (SH) enhanced convec-tion and sometimes favored cyclogenesis.Lack of data hindered finding any develop-ment continuity, and isolated speed in-creases were often discounted since meteo-rologists were looking for downstreammovement of acceleration centers. No localgenerating force could be identified. Thereis now no doubt that throughout the tropics,and in all seasons, wind currents locallyaccelerate and decelerate, usually withoutapparent cause, but often with significantimpact on weather. The best-understoodsurges, originating in the winter hemi-sphere, will be discussed first, followed bythe more mysterious, but no less important,surges in the summer monsoon and thetradewinds. Oppositely-directed uppertropospheric surges almost certainly accom-pany surface surges, but are often poorlyobserved. a. Winter Surges. In the winterhemisphere, meridional temperature gradi-ents are largest and branches of the HadleyCell best developed across the continents.The vertical circulation of a Hadley cellbranch may suddenly accelerate in responseto a deep trough in the upper-troposphericwesterlies (Ramage, 1971). Southeast Asiaand the South China Sea experienced suchan event in mid-January 1967 (Figure 8-31).A cold front crossed Hong Kong (22°N,114°E) on the 14th; on the same day surfacewinds increased across the South China Sea,and heavy rain began over North Borneo.Apart from a remnant, moving southwardalong the coast of Vietnam (Chang et al.,1979) the cold front had disappeared as thenortherly winds south of the front freshenedsimultaneously. Figure 8-32

shows the vertical structure of such a sys-tem. In the middle and upper troposphere,cold air also swept southward and subsidedon the eastern side of a ridge; on the south-

ern and western sides of the ridge, easterliesaccelerated over Singapore (1°N, 104°E) andsoutherlies over Saigon (11°N, 107°E).

Once established, surges do not move much.They last about 3 to 5 days. Rather weaksurges generated during Winter MONEXtravelled south at 80 knots across the north-ern and central South China Sea, and ap-peared to have the character of synoptic-scale gravity waves (Chang et al, 1983).During surges, the cross-isobaric componentof the surface wind is much greater thannormal.

On 9 December 1985, a cold front passedHong Kong. At about the same time north-easterly winds increased by about 10 to 15knots over the South China Sea. Prior to thesurge, the satellite showed scattered convec-tion near 10°N, north of a surface trough(Figure 8-33). Everywhere to the northclouds were limited by a subsidence inver-sion. Twenty-four hours later (Figure 8-34)shallow low cloud had reached about 20°N,but to the south the convection near 10°Nhad intensified. During this time, the jetstream south of Japan increased from 110 to150 knots. Subsequent surges further in-creased winds over the South China Sea,and led to a tropical depression developingnear 8°N on the 15th.

Vigorous cold outbreaks channeled bynorth-south mountain ranges also affectCentral America and the Caribbean. “Typi-cally, increase of wind speed and shift indirection are observed several hours prior tothe arrival of cold air” (Hastenrath, 1985).These “Northers” funnel through the moun-tain passes of southern Mexico, Nicaraguaand Panama. Each winter the Isthmus ofTehuantapec experiences an average of 20Northers and the Canal Zone, three(Trewartha, 1981). In the Pacific, hundredsof kilometers downstream from the passes,the resulting high winds, cool upwelledwater, and a cloud minimum are features of

the long-term climatology (Figures 8-35, 8-36) (Sadler and Lander, 1986).

Significant surface pressure rises at HongKong (22°N, 114°E) have been statisticallyrelated to heavy rain along the east coast ofWest Malaysia within a day or so (Gan,1963). Similarly, Brooks (1987) used increasein the pressure gradient between Houston(30°N, 95°W) and Merida (21°N, 90°W) inforecasting onset of a Norther in Honduraswithin 48 hours. Kousky (1979) counted 152cold fronts at 18°S on the coast of Brazil in aten-year period. They were most commonbetween March and December. Occasionally,the fronts reached 5°S and enhanced rain-fall. Surface pressures varied nearly simulta-neously between 18°S and 4°S. Parmenter(1976) used satellite pictures in tracking acold front across South America in July1975. When the front reached about 15°S,winds turned southerly and accelerated asfar north as 5°S, and by the following day,had reached 10°N. Equatorial convectionincreased as well. These observations againsuggest near-simultaneous accelerations ofthe South American Hadley Cell branch.

In winter, when a vigorous cold front startsmoving south over North Africa, the strongwinds may cause a severe dust storm(Chapter 10. Section E-3). However, onceover the desert, the front is hard to follow.Hamilton and Archbold (1945) and Johnson(1964) noted that freshening northeastwinds often start to blow dust well ahead ofwhere the front might reasonably have beenexpected. Did a branch of the Hadley Cellsuddenly accelerate? Figure 8-37 uses theleading edge of a North African dust stormto define a cold front position. Since thisimplies that the front moved at more than 30knots over the previous 24 hours (a surpris-ing rate for such low latitudes), could not asurge be the cause?

Dvorak and Smigielski (1990) used satellitepictures to identify outbreaks over thetropical Americas in March 1986 and over

North Africa in October 1978. In both cases,divergence beneath the upper troposphericlow-latitude ridge west of the outbreakoverlay intense convection. Walters et al.(1989) reported that over central America,equatorward-advancing upper tropospheric“cold pools” often accompany surfacesurges. Once again, these observationssuggest near-simultaneous accelerations of abranch of the Hadley Cell.

During winter, over southern Africa andAustralia, cold fronts may be preceded by“leader fronts” up to 300 to 400 km ahead ofthe main cold front (Taljaard, 1972). Oftenwind and temperature changes are greaterat the former. Possibly every three to fivedays, similar sequences occur during the SHwinter along and off the east coast of Africa.In a case study by Cadet and Dubois (1981),a deep mid-latitude low moved eastward tothe south of the continent, and southerliesincreased in the Mozambique Channel, wellahead of the surface cold front. Within a dayor two, the surge had crossed the equator,intensified the Somali jet and increasedconvection over Somalia. Although surgesare best observed at the surface, severalinvestigators (e.g. Chang and Lau, 1980,1982) have confirmed that upper tropo-spheric poleward flow diverges from theenhanced convection in the deep tropics.Downstream from where the poleward flowmerges with the subtropical jet, the wester-lies accelerate. Cause and effect are hard toascribe. Is the sequence started in the tropicsor in higher latitudes? That everythingseems to happen at once suggests that asurge cycle goes as follows: Initially, as asurge develops, convergence and cyclonicvorticity downwind of the speed maximumenhance upward motion and so increaserainfall. In the upper troposphere, above therain, heat accumulating from condensationis exported by the divergent flow to colderregions. Thus convective instability is main-tained. Quiescence returns when the surfacesurge weakens and heat is no longer ex-

ported aloft (no longer needed?). Convec-tion then warms the upper troposphere, andby increasing stability, inhibits furtherconvection. Once more, everything seems tohappen at once.

Over the maritime continent generally fairweather prevails when the northeast mon-soon is weak. When circulation strengthens,rising motion increases, but is often streaky.For example, in January 1967 (Figure 8-31),the mid-month surge was much wetter overNorth Borneo than the surge at the start ofthe month, whereas the reverse was trueover eastern West Malaysia, which experi-enced floods in the earlier period. Rainfallstreakiness probably also typifies the equa-torial regions of Africa and South America.Thus, when a Hadley Cell branch is weak,local forecasts of relatively fair equatorialweather can be made with some confidence.When a surge develops, figuring just wherethe heavy rain will fall is much more diffi-cult, although rain is more likely wheresurface winds curve cyclonically.

Analysts should watch carefully for theseregional accelerations and be prepared toeliminate fronts from their analyses of thestrongest cold outbreaks. They shouldexpect to retain fronts in analysis polewardof 20° latitude, and for longer times duringspring, or in the mid-ocean tradewinds, thanequatorward of the continents in the depthof winter. They should remember that theHimalayas prevent central Asian cold out-breaks from reaching the Indian Ocean.

b. Other Surges. As mentioned in Section C-2-d, a surge in the southwest monsoon overPeninsular India precedes development of amonsoon depression over the northern Bayof Bengal. Surges generated in the winterhemisphere apparently crossed the equatorand contributed to the generation of tropicalcyclones in the South Pacific (Arakawa,1940) and in the eastern North Atlantic(Morgan, 1965). Attempts to link the surgesto fronts failed. In the Australian region,

when a tropical cyclone develops in themonsoon trough, low-level westerlies northof the trough and low-level easterlies southof the trough usually have already increased(McBride and Keenan, 1982). Here, and inthe western Pacific (Douglas, 1987), tropicalcyclones in their early stages resemblemonsoon depressions.

Lander (1990) studied simultaneous devel-opment of tropical cyclones on either side ofthe equator in the western Pacific, andfound that equatorial surface westerliesfreshened prior to cyclogenesis. Chu andFrederick (1990) suggested that equatorialsurges might stem from pressure rises to thewest originated by surges out of East Asia.The case they studied lasted 11 days; 3 to 5days is more usual. In line with this, Love(1985a,b) concluded that surges originatingin the winter hemisphere, by increasingequatorial pressure, could induce surges inthe surface westerlies of the summer hemi-sphere and so favor cyclogenesis there.

Jeandidier and Rainteau (1957) emphasizedthat forecasting westerly monsoon surges isprerequisite to forecasting weather in theCongo Basin.

What else might cause the pressure gradientto increase and the summer monsoon tofreshen? Guard (1985) thinks that the pres-sure falls accompanying tropical cyclonessometimes generate upstream surges, evenin the opposite hemisphere, while the wide-spread pressure changes described in Chap-ter 4, Section B-2 could also be responsible;about two a year develop east of the Philip-pines. According to Guard, these pressurefalls might originate from intensifyingupper cyclones (Section C-5). Minor surgesmight also result from the heat/monsoontrough deepening under clear skies.

Guard has categorized summer monsoonsurges (Figure 8-38). In a weak surge, themonsoon is shallow, the upper level returnflow from northeast is weak, and a showers

regime prevails. In a strong surge, thesouthwesterlies are deep and vigorous andare overlain by strong northeasterlies. Thislarge vertical wind shear accompanies arains regime.

Surges in the tradewinds seem to comprisehigh pressure pulses originating on theeastern sides of subtropical cyclones. Thesurges generally lack fronts. They may affectpatterns of convergence and convection inthe NECZ, while on the equatorward side ofa speed maximum, cyclonic shear favorsdevelopment of a cloud line.

Over Africa in the warm part of the year,anticyclonic pulses/surges can reach thetropics between the intense heat lows cen-tered over central West Africa and south-west Asia (see Figure 4-1). This interveningzone stretches across Libya, Egypt andSudan. According to Leroux (1983) surgestrigger the haboobs of Khartoum (16°N,33°E) (Chapter 10, Section D-3-c), and far-ther south, where more moisture is avail-able, the squall lines of West Africa (SectionD-2). The time-sequence is schematicallyshown in Figure 8-39, with dust stormspredominating in the dry regions aroundand north of the heat trough, and squalllines where moist southwesterlies prevail.The schematic incorporates Leroux’s viewthat westward progression of the anticy-clonic pulse causes the squall line/duststorm also to move west. Strong environ-mental easterly shear is a much more likelycause of this. Similar surge-induced squalllines move westward across tropical SouthAmerica (Section 8b).

During the SH summer, upper troposphericpoleward outflow from enhanced tropicalconvection is accompanied by a strength-ened subtropical jet stream. Hurrell andVincent (1990) observed this over the south-west Pacific and south Indian Oceans. In aPacific case study they reported that convec-tion came from a developing tropical cy-clone in the South Pacific Convergence Zone

(SPCZ). A time sequence could not be estab-lished.

Perhaps the continuous thunderstormsdescribed in Chapter 6, Section C-b, aresometimes triggered by surges from bothnorth and south, directed toward a weaklycyclonic circulation in the surface layers,and away from an overlying upper tropo-spheric anticyclone.

c. Summary. Rapid wave-like propagationwith convergence in the surface windsflowing toward the heat equator, matchedby divergence and heat export in the oppo-sitely-directed upper tropospheric flow,causes surges suddenly to “appear” in thetropics; once there, the surface speed maxi-mum and bad weather downstream of themaximum may not move much for severaldays. This could account for stationary badweather in the low-level easterlies over EastAfrica (Johnson, 1962) and near-absence ofmoving disturbances over Indonesia (Braak,1921-29). Upstream of the speed maximum,weather is better than usual — in the wintermonsoon and the tradewinds, the inversionlowers.

In surges, heat is exchanged between thetropics and the winter hemisphere, andsometimes between the tropics and higherlatitudes in the summer hemisphere. Heatexport in the upper tropospheric part of asurge causes instability and so contributesto convection along the forward edge of thesurface surge. Cause and effect are hard todetermine; convergence in low levels, diver-gence aloft and enhanced convection appearto coincide and may arise from distantcauses; the surprising variability of tropicalweather (Lau and Lau, 1990) may reflectthis. Surges contribute to climatology. Theyare most common during winter in longi-tudes dominated by continents — Africa,the Americas, East Asia/Australia. There,low-latitude rainfall is heaviest and uppertropospheric flow to colder latitudes (some-times across the equator) is greatest and the

subtropical jets are strongest (Yang andWebster, 1990). Even in summer, uppertropospheric outflow from the Sudan into ajet maximum over Turkey fit the pattern.

Surges range widely in strength, extent,duration and weather effects. Except withintense polar outbreaks they can seldom beanticipated. Regional and annual variationsare poorly understood. But long-time evi-dence from disparate sources confirms theirimportance. Their effects on tropicalweather demand to be investigated; and aspointed out by Keenan and Brady (1988) forthe Australian summer monsoon, theseeffects may well have been overemphasized.In the meantime, tropical forecasters shouldwatch for changes in surface wind speedswithin extensive monsoon and tradewindstreams. An isolated report of stronger windassociated with a thunderstorm, needscareful evaluation before being dubbed asurge.

Forecasters must avoid repeating the errorof their predecessors with fronts in the 1930sand easterly waves in the 1950s — not allmysterious weather is surge-induced!

5. South Pacific Convergence Zone (SPCZ).The charts of average cloudiness (Figures 6-1 to 6-4) show a persistent maximum ex-tending east-southeastward from NewGuinea. It also appears as a relative mini-mum in annual average OLR (Figure 6-22),though not as well-marked south of 10°S asthe cloudiness maximum. The clouds aredeepest from October to March.

a. Winter. Hill analyzed two periods ofextensive altostratus between Australia andFiji in June of 1962 and 1963. Light rain fell.In both cases a remnant cold front wasmoving northeastward beneath divergentflow east of a trough in the upper-tropo-spheric westerlies. The divergence maxi-mum was created by passage of an isotachmaximum. In the lower troposphere, sink-

ing dried the air. Hill suggested a model forthis decoupling between upper and lowertropospheres (Figure 8-39) that may alsoaccount for the sandwich described inSection 6. Middle- and upper-troposphericupward motion responded to the jet streammaximum, while lower-tropospheric sink-ing accompanied outflow from the surfaceanticyclone. Upper troughs that slow downand often become stationary cause theclimatological cloudiness maximum.

At 03 LT 2 July 1985 (Figure 8-41), a well-developed SPCZ extended southeastwardfrom New Guinea. It lay in divergentnorthwesterlies east of an upper trough thathad recently crossed the Australian coast.The western edge of the SPCZ cloud sheetcoincided with sharply sheering uppertropospheric winds west of a strong jetstream that increased (diverged) down-stream from 55 knots over southeasternNew Guinea to 150 knots at 27°S, 160°E. Fora few days, the system was almost station-ary. At the surface in the SPCZ, southeast-erly tradewinds prevailed, reaching 25 to 30knots beneath the jet stream. A weak depres-sion remained over the Solomon Islands.Observers reported skies overcast withmiddle and high clouds, little or no convec-tion, and scattered, generally light rain. Thejet stream and its accompanying badweather moved east on the 4th, along withthe upper tropospheric trough; thetradewinds moderated.

The Hill model seems to describe spring andautumn situations east of Asia (Guard,1990). Frequently, in the region of coldfronts, dense cloud below 20,000 feet (6 km)underlies dense cirrus above 26,000 feet (8km), with little cloud between. Anticipatingthis distribution helps in planning air refuel-ing.

b. Summer. In summer, the convergencezone lies between surface northeasterlies tothe north and southeasterlies to the south,and includes a climatological low east of

New Guinea. As in winter, the SPCZ isalmost stationary. Vincent (1982, 1985)analyzed the SPCZ for the period 10-18January 1979. An upper trough lay west ofthe surface trough. As shown in Figure 8-42,upper divergence, lower convergence and aminimum in OLR coincided. Three tropicaldepressions formed and moved east orsoutheast. Such developments are commonduring summer, and depressions mayremain almost stationary in the convergencezone for many days before moving away.Some quickly intensify into tropical stormsor hurricanes. Forecasting is difficult. Un-necessary warnings may be given if earlyslight intensification is extrapolated. On theother hand, a low that has remained station-ary for days can catch a forecaster by sur-prise if it suddenly intensifies and starts tomove. Sequences of satellite pictures canhelp.

6. Troughs in the Upper-TroposphericWesterlies. In middle latitudes, rising mo-tion ahead of these eastward-movingtroughs intensifies fronts and may causewave cyclones to form. Except over Asia andNorth Africa in summer, the troughs oftenextend into the tropics and may even reachthe equator. Figure 8-43 (Ramage, 1971)shows a winter-time trough almost station-ary over Indochina after having movedeastward south of the Himalayas. A front layacross the northern part of the South ChinaSea, with fresh surface northeasterlies to thenorth. West of the upper trough, skies wereclear (Figure 8-44) as upper-level conver-gence caused strong subsidence down to930 mb (Figure 8-45). East of the uppertrough, upper-level divergence producedascent and extensive cloud. In the south,upglide along the front (SSW winds), andheat and moisture added to the cold surfaceair by coastal waters, almost saturated airbelow and above the front. Visibilities werepoor, but little rain fell. Wang et al. (1985)

ascribed springtime heavy rain over Taiwanto a similar interaction between surfacefront and upper trough.

During the dry cool season, upper tropo-spheric troughs affecting Senegal (13 to16°N), and very rarely, other parts of tropi-cal North Africa (Sak, 1962) resemble thecool season troughs of the SPCZ (Section 5).Skies east of the trough become overcastwith altostratus; the underlying anticycloniceasterlies are even drier than the SouthPacific tradewinds, and so very little rainreaches the ground. Australia (Southern etal., 1970) and southern Africa (Fox, 1969)experience similar situations. At the startand end of the cool season, when moremoisture is available, frontal thunderstormsand even hail may precede the trough(Figure 8-46).

In Zambia (8 to 18°S) the rainy season lastsfrom December through March. Kumar(1979) analyzed five rainy seasons (1972-73through 1976-77), finding that an average offive to seven upper tropospheric westerlytroughs extended into Zambia each month.Rainfall increased east of about 80 percent ofthe troughs and decreased after they hadpassed.

Over the Hawaiian Islands, mean resultant200 mb winds blow from slightly north ofwest in winter and slightly south of west insummer (Figures 4-10 and 4-11). Troughs inthe upper westerlies can occur throughoutthe year. The most intense, giving heavyrain east of the trough axis, are confined tothe cool season. On 19 April 1974, northeast-ern (windward) Oahu experienced up to 250mm of rain as it lay east of a stationarytrough in the upper-tropospheric westerlies(Schroeder, 1977). Although the tradewindscontinued to blow, the inversion disap-peared (Figure 8-47) and a “continuous”thunderstorm developed, anchored by themountains that were also orographicallylifting the tradewinds. If low-level conver-gence is already concentrated east of the

upper trough, say along a front, and if thetrough is stationary or slow-moving andsurface flow light or moderate, upwardmotion east of the trough may then act longenough on a volume of air for deep convec-tion and severe weather to develop. How-ever, rain is not always enhanced east of anupper trough. Although Han (1970) in hisstudy of 300 mb troughs concluded thatover northern Oahu daily rainfall greaterthan 0.10 inches (2.5 mm) is more likely withthe trough near to or west of the island thanin other locations, average trough-dayrainfall (0.15 inches (3.8 mm)) was not muchmore than average non-trough-day rainfall(0.10 inches (2.5 mm)). In the tradewinds,the relationship between the height of thesubsident dry layer and the direction of the200 mb wind could indicate vertical motion.For 1976 at Lihue (22°N, 159°W), the medianheight of the dry layer base was 800 mb withnorthwest winds at 200 mb and 785 mb withsouthwest winds at 200 mb. Figure 8-48shows that dry layer bases above 650 mb aremuch more likely with 200 mbsouthwesterlies but on two-thirds of occa-sions the difference between the effects ofnorthwest and southwest flows is negligible.

Often, east of an upper-tropospheric trough,an extensive sheet of altostratus is separatedby a subsiding dry layer from the tradewindinversion and the moist tradewinds beneath.This condition may last several days. Thedry layer prevents any rain falling from thealtostratus from reaching the surface. Risingmotion east of the trough is apparentlydecoupled from the lower troposphere bythe widespread subsidence associated withthe subtropical anticyclone. This and an-other possible cause of “moist-dry-moistsandwiches” are discussed in the followingparagraphs.

a. Moist-Dry-Moist Sandwiches. In theregion of Hawaii on about 20 percent ofdays, east of an upper-tropospheric trough,a moist-dry-moist sandwich is observed.

The sounding shown in Figure 8-49, madeduring a prolonged rain spell on the top ofMt. Waialeale (Figure 6-21) is typical. Mod-erate moisture convergence caused by risingmotion east of the trough causes middlecloud to develop within a normally dryzone, though why rising motion fails toextend to the surface is puzzling. What ismore puzzling, moist-dry-moist sandwichesare just as common beneath upper-tropo-spheric northwesterlies, though generallyunaccompanied by middle cloud. Duringthe Line Islands Experiment (LIE), 70 per-cent of the soundings at Fanning Island(4°N, 159°W) had sandwiches. Of 1200dropsonde observations made over theeastern tropical Pacific in 1979, 68 percenthad sandwiches (Kloesel and Albrecht,1989), suggesting that in convectively sup-pressed conditions two distinct cloud layersmay exist. The lowest OLR of any tradewindarea, recorded east of Hawaii (Figure 6-22),as well as dryness above 700 mb east of theupper trough in Figure 8-45, could also beevidence of the phenomenon.

In the absence of middle-troposphericmoisture convergence east of an uppertrough, how does moisture accumulate insubsiding air? Kloesel and Albrecht (1989)postulated the nearby existence or pre-existence of deep cumulus, which moistenlevels above the inversion by advection, andwhose downdrafts bring relatively drier airto the top of the inversion. Detailed observa-tions are needed to test this hypothesis.

Decoupling of upper and lower tropo-spheres also occurs in winter over the south-west Pacific (Section 5) and North Africa,and may be not uncommon elsewhere in thetradewinds and the winter monsoon. Fore-casters must often choose between little rainor a storm east of an upper tropospherictrough. Upper winds often fail to revealwhether upper and lower tropospheres arecoupled (Figure 8-47) or decoupled (Figure8-49). Weather satellites may not help much,

viewing only the upper cloud deck, al-though in a storm this will be denser andcolder than in the decoupled mode.

Interaction between a trough in the upperwesterlies and the near-equatorial conver-gence zone is discussed in Section 8.

7. Superposition of Tropical and Extratro-pical Disturbances. Between 15° and 25°latitude, disturbances of tropical origin aretypical of summer, whereas in winter,troughs in the upper-tropospheric westerliesmay appear. Rising motion east of thesetroughs is often insufficient to break down atradewind-type inversion and bad weatherseldom results (Sections 5 and 6). Occasion-ally, a remnant tropical vortex may be mov-ing westward at a low latitude as an upper-tropospheric trough approaches from thewest. As the two systems superpose, therelatively deep moist layer accompanyingthe vortex is further deepened east of theupper trough, and an unseasonable wetspell may result. Then, as the two systemsdraw apart, weather becomes seasonallynormal again.

In December 1962 (see Ranganathan andSoundarajan, 1965; Ramage, 1971), a weaktropical vortex moved across southernPeninsular India on the 1st and becamestationary off the west coast, just east of analmost stationary trough in the upper-tropospheric westerlies. From the 2ndthrough the 4th, both systems intensified,and widespread rain resulted (Figures 8-50and 8-51). At Bombay (19°N, 73°E) (Figure8-52), southwesterlies prevailed above 500mb from the 3rd through the 5th. In thelower troposphere, pressure fell on the 3rdand 4th as the tropical vortex intensified,and southerlies set in. Moist air had spreadthrough the troposphere by the 4th. Then asthe upper trough moved eastward, subsid-ence dried the air above 700 mb and the rainceased. The southern vortex had moved

westward and both systems quickly weak-ened. At Bombay 1.9 inches (48 mm) fell injust over 24 hours on the 4th and 5th. TheBombay December average is 0.08 inches (2mm) and four out of five Decembers thereare completely dry.

Superposition can also occur in a transitionseason. In September 1970, a late monsoondepression was moving west over India.Instead of weakening, it intensified as itcame under the influence of three earlywaves in the middle- and upper-tropo-spheric westerlies. It stalled beforerecurving eastward. Rainfalls reached 11.8inches (300 mm) (Mishra and Singh, 1977).

Frank (1969) showed that as westward-moving cloud systems in the North Atlanticmoved beneath southwest flow east of anupper-tropospheric trough, convectionrapidly increased, and decreased as theymoved west of the upper trough. Accordingto Simpson (1970b), these changes stemfrom initial stimulation and then constraintof the upper-level outflow as the lowersystem moves beneath the upper trough(Figure 8-53). Since superposition is unsea-sonable, forecasters should be on the alertfor the unseasonable disturbance — in thefirst example, the tropical vortex, in thesecond, both the monsoon depression andthe upper westerly troughs. Predicting thattwo approaching disturbances will super-pose is not too hard, but anticipating whenthey will interact and for how long can bedifficult. Frequent satellite pictures, show-ing deepening clouds, help greatly.

8. The Near-Equatorial Convergence Zone(NECZ). a. The Convergence Zone over theOceans. Over the southeast Pacific and theSouth Atlantic, the southeast tradewindscross the equator and converge with thenortheast tradewinds between 5° and 10°N,where a cloud band, the NECZ, is found. Inearlier chapters, mean charts of gradient-

level winds, cloudiness and OLR all con-firmed that the near-equatorial convergencezones of the North Atlantic and the centraland eastern North Pacific are the mostpersistent disturbed weather in the tropics.Within the zones, ships report rain morethan 30 percent of the time; elsewhere overthe tropical oceans 20 percent is rarelyexceeded (Crutcher and Davis, 1969). Notonly are there more rain days under theNECZ, but the rainfall per rain day is lessthan elsewhere. Two coral islands, Washing-ton (5°N, 160°W) and Kwajalein (9°N,168°E), have the same average rainfall (2592and 2630 mm) but ships report twice asmany rain days at the former, influenced bythe NECZ, than at the latter, where movingdisturbances usually cause the rain. TheNECZ never moves east/west and onlyslowly north/south, probably because itoverlies a warm eastward-moving oceancurrent. OLR measurements show thatNECZ clouds are not as deep as those overthe equatorial continents and Indonesia(Figure 6-22). This is confirmed by researchaircraft flying over the cloud at 16,000 feet (5km), rare satellite-detected lightning dis-charges (Figure 10-2) (Orville andHenderson, 1986), and absence of cirrus inNimbus 7 satellite pictures (Stowe et al.,1989). The cloud tops may well be overlainby remnants of the tradewind inversion.

Twelve meteorological/oceanographicresearch flights shuttled between Honoluluand Tahiti from 29 November 1977 to 5January 1978 (Ramage et al., 1981) (Figure 8-54). They revealed large intensity changes inthe central Pacific NECZ, apparently stem-ming from changes in the South Pacific andNorth Pacific tradewinds. Figure 8-55 showsthe daily changes in satellite pictures of theNECZ along 150°W; 158°W was equallyvariable. For the period encompassing theshuttle flights (38 days), intensities were thesame along 150°W and 158°W (476 NM (882km) apart) on only 15 days and persisted

from one day to the next on 15 days at150°W and on 11 days at 158°W.

In the NECZ region the aircraft flew at 820feet (250 m) altitude. The northern andsouthern edges of the zone were sharplydefined in the wind and moisture fields nomatter how intense the zone. When theNECZ was active, it comprised lines ofconvergence and raining cloud separated byrelatively dry strips in which the air wassinking. There were no thunderstorms. Theschematic cross section of a vigorous NECZ(Figure 8-56) is based on the shuttle flightsand on subsequent 20,000 feet (6 km) flightsover the NECZ along 110°W. The easterlycomponent of the wind goes into the paper.Air follows helical trajectories with east-west axes. Consequently, air parcels are keptwithin the cloud long enough for droplets togrow to raindrops and fall out. The lines ofcloud spread out just beneath the raisedtradewind inversion; the uniform surfacethus presented to weather satellites oftenhides the lines beneath.

The NECZ is a mechanical phenomenon,resembling in some ways the orographi-cally-lifted cloud lines that account for mostof the rain on Mt. Waialeale (Chapter 6Section C-1-c). The rapid intensity changesseem to arise from mutually independentvelocity changes (some possibly caused bysurges) in the NH and SH tradewinds, andfrom complex interactions with the uppertroposphere. Equatorward-moving shearlines might sometimes be responsible,although smaller-scale, more local fluctua-tions are probably more common. No evi-dence has been uncovered of westward-moving easterly waves.

A trough in the upper-tropospheric wester-lies, often preceded by a wide band ofmiddle and high clouds (see Section 5), mayinteract with the NECZ (Giambelluca, 1986).In the Pacific, as the trough approaches fromthe west, it induces the tradewinds to be-come more easterly and weaken. Conver-

gence into the NECZ then diminishes andclouds break up. The middle- and upper-tropospheric rising motion associated withthe cloud band just east of the upper troughmay either be decoupled from the NECZ orinduce deep convection and thunderstormsin it. Once the upper trough has passed, thetradewinds freshen and shift from east backto northeast, and the NECZ is reestablished.

McGurk and Ulsh (1990) averaged 35 casesof interaction between a trough in the uppertropospheric westerlies and the NECZduring the winter of 1983-84 over the east-ern North Pacific. They used vapor imagesfrom GOES to estimate moisture distribu-tion. Before the trough extended far enoughsouth for interaction to occur, NECZ mois-ture (cloudiness) diminished to the east. Atinteraction, north of the NECZ, moisturegreatly increased east of the trough anddiminished west of the trough. The responsewas similar, but much less in the NECZ. Bycomparison, when upper tropospherictroughs were absent, the NECZ was gener-ally moister and more convectively un-stable. During the period studied, troughsand NECZ were interacting 80 percent ofthe time over the eastern North Pacific.

At 00 UTC, 2 December 1985, an uppertropospheric trough was approaching Ha-waii from the west. It was preceded by avigorous STJ, clearly delineated in a satelliteimage (Figure 8-57) by a broad band ofmiddle and high clouds. A well-developedNECZ lying along about 9°N was somewhatweakened just ahead of the upper trough.Conforming to the McGurk and Ulst aver-ages, the line linking deepest cloud to theeast and least cloud to the west of the troughlay about 5° north of the NECZ.

Over southeast Asia during winter, thenortheast monsoon extends to the equator(Figure 4-6). Then, as with the NECZ, anapproaching trough in the upper tropo-spheric westerlies weakens the monsoon

and improves equatorial weather (Figures 8-43 and 8-44).

Except when a trough in the upper-tropo-spheric westerlies is involved, flyingweather in the NECZ is not severe. Turbu-lence is slight or moderate and jet aircrafteasily top the clouds. At the surface, pro-longed rain under one of the cloud bandsmight significantly lower visibility.

b. The Convergence Zone Over NorthernSouth America. Over South America, sur-face pressure decreases westward to clima-tological heat lows over western Colombiaand the Gulf of Panama, and over Bolivia. Inresponse, easterlies blow down the gradient,and the Atlantic NECZ extends as far westas the Andes (Walters et al., 1989). Intensi-fied by orography, weather is usually worseand more variable along the continentalNECZ than out to sea, and thunderstormsare common (Figure 10-2). As over theoceans, the convergence zone is farthestnorth in the NH autumn and farthest southin the NH spring, being oriented roughlyeast-west. In March/April, it enters theBrazilian coast south of the Amazonmouths, moves rapidly north in May andJune, and reaches 10°N in August/Septem-ber. Then follows a rapid return to south ofthe Amazon by December. WesternAmazonia and the eastern foothills of theAndes, within 5° of the equator, experiencelittle annual variation in rainfall. Farthersouth and farther north, rains are heaviest inthe summer. As over the oceans, weathersome distance from the NECZ is generallyfair, with the tradewind inversion limitingconvection. When the inversion rises orweakens, popcorn cumulus develops duringthe afternoon.

Extensive field programs in Venezueladuring the wet seasons of June to October of1969 and 1972 (Riehl, 1979b) and over theAmazon basin during the dry season (July -August) of 1985 (Garstang et al., 1988) andduring the wet season (April-May) of 1987

(Greco et al., 1990) included research aircraftflights and special aerological soundings.The Venezuelan programs have been exten-sively documented; papers on the Amazonprojects are still appearing. The regionsadjoin, and although the topographiesdiffer, disturbances may affect both at thesame time, producing more rain near theNECZ. What follows summarizes manyresearch findings.

In the NH summer, when the NECZ oftenlies over Venezuela, surface westerliessometimes appear along the north coast,and may be as deep as 7000 to 10000 feet (2to 3 km), suggesting development of a weakmonsoon trough to the north. Rarely, distur-bances that resemble monsoon depressionsform in the trough and cause heavy rain tofall on the eastern slopes of the Andes(Riehl, 1979b). Farther east, weak, shallowvortexes have been detected by researchaircraft. A surge from the SH may cause thewesterlies. When they are absent, Venezu-elan wet-season rainfall is less than normal,especially along east-west segments of thecoast, where the tradewinds diverge inresponse to stress-differential (see Figure 2-8).

Over South America north of 10°S and eastof the Andes, easterly winds prevail below300 mb throughout the year (Chu, 1985),while an easterly jet has been observed near10000 feet (3 km) over Venezuela. Squalllines, resembling those of West Africa,develop in this environment or even overthe Atlantic, and move westward(Fernandez, 1982). The associated cloudclusters are oriented roughly north-southand account for most of the rainfall. Duringtheir passage, they disrupt the NECZ.

Surges from either hemisphere may gener-ate squall lines. A favored source region isjust off the east coast between 7°N and 5°S,where land breeze and tradewinds convergeat night and cause vigorous convection. Bylate afternoon the squall line has moved

west and intensified over the coastal moun-tains. It continues inland at speeds rangingfrom 5 knots (weak) to 32 knots (strong).Squall lines may also develop over land. Theassociated cloud clusters undergo diurnalvariations as they move. Extensive cloudclusters tend to be most active at night;when convection is scattered, it is mostcommon during the afternoon and evening.

Annual variation in rainfall is controlled bythe meridional movement of the NECZ andby the frequency of cloud clusters. There aremore of these in the wet season than in thedry season. Rainfall may vary considerablyfrom year to year, in response to variationsin NECZ latitude and number of squalllines. Northeast Brazil is especially prone todroughts. Only when the NECZ movesanomalously south of its normal southern-most latitude is northeast Brazil assured ofadequate rain (Hastenrath, 1985).

Forecasts should incorporate westwardmovement of weather systems, allowing forlarge ranges in speed and duration. Fairtradewind weather and scattered convectioncan occur at any time of the year, but aremore likely during the dry season. Con-versely, although cloud clusters can occuryear-round, they are more common in thewet season, and give most rain near theNECZ. Forecasters may get a jump on squallline development by detecting surge genera-tion from middle latitudes. Southern Hemi-sphere winter surges ahead of vigorous coldfronts may be the easiest to anticipate.

9. Mid-Summer Dry Spell. Over southeastAsia this spell shows up in long-term aver-age pentad rainfall (Figure 6-38) and can befollowed synoptically. Figure 8-58, fromSadler et al. (1968) shows that by 15 July1967 a ridge line had moved northward oversoutheast Asia; satellite pictures revealeddecreasing convection. The ridge continuednorthward on 19 July. Cheng (1978), in adetailed study of the dry spell, used one-dayrainfall averages and concluded that the

spell usually comprises three shorter peri-ods. Of these, 92 percent were associatedwith ridges and 8 percent occurred in theregion of divergence west of a typhoon.

Over India, “breaks” in the summer mon-soon rains occur when the monsoon trough,usually lying along the Ganges Valley,moves northward to the Himalayas and isreplaced by a ridge (Rao, 1976). At times thisshift can extend as far east as the Philip-pines. Although a break is more likely in thesecond half of August (Figure 6-39), it canoccur at any time in the summer and may bea manifestation of the 30 to 60 day oscilla-tion (see Chapter 12).

On 11 July 1985, the weather satelliteshowed a distribution typical of the mid-summer dry spell (Figure 8-59). A ridge ofhigh pressure, accompanied by fineweather, extended along about 20°N fromthe central Pacific to Burma. It persistedfrom the 10th to the 18th July across southChina and the northern South China Sea.

Chapter 9TROPICAL CYCLONES

A. General.

The general term “tropical cyclone” is sub-ject to various interpretations. In this reporttropical cyclone is used in its broadest sense,and is defined as a non-frontal, synoptic-scale cyclone, developing over tropical orsubtropical waters and having a well-orga-nized circulation. This definition impliesnothing about wind speed or weather;however, the more destructive tropicalcyclones are warm-cored and have strongestwinds near the surface that decrease withheight.

Intense tropical cyclones (i.e., hurricanes ortyphoons) are the tropics’ most impressivephenomenon. The destructive potential oftheir high winds, rain, and storm tides iswell known. And so tropical meteorologistsmust be thoroughly versed in them. Mostresearch in tropical meteorology concen-trates on tropical cyclones. There are manyreferences to this subject; only relatively fewwill be mentioned here. Useful generalworks include those by Riehl (1979), theWorld Meteorological

Organization (1979), Anthes (1982) andElsberry (1988). Satellite observations oftropical cyclones and their analysis wereextensively discussed by Dvorak (1984), andDvorak and Smigielski, (1990). In this chap-ter, the essential global characteristics oftropical cyclones are summarized withsections on structure, classification, climatol-ogy, formation, dissipation, movement, andforecasting.

B. Structure of Mature TropicalCyclones.

1. General. Even though tropical cycloneforecasting is highly centralized, meteorolo-gists must be familiar with the major fea-tures of intense tropical cyclones so as toproperly interpret and apply these forecaststo terminal and local weather conditions.Detailed observations, especially thoseobtained by reconnaissance aircraft, haveenabled the structure of tropical cyclones tobe described. Satellite pictures have alsosignificantly contributed. This section usesan excellent survey paper by Miller (1967) indiscussing the characteristics of the wind,temperature, and cloud distributions intropical cyclones. Wherever they occur,tropical cyclones have similar features.

2. Winds. The low-level circulation com-prises three distinct areas:

a. In the outer portion, extending from thestorm periphery inward to the edge of thezone of maximum winds, wind increasestoward the center.

b. The annulus of maximum winds sur-rounding the eye is the most outstandingfeature of the mature tropical cyclone. It isabout 5 to 10 NM (10 to 20 km) wide andcoincides with the wall cloud, site of themost vigorous convection and the heaviestrain in the storm.

c. The eye is the innermost part of the storm.Here the wind weakens rapidly toward theeye center. As determined by the radius ofthe eye-wall cloud, the eye can be as little as3 NM (6 km) to as much as 100

NM (185 km) across.

Izawa (1964) prepared composite models ofthe circulation in Pacific typhoons, based on14 storms approaching Japan. His compositevertical cross section of the mean tangentialspeeds is shown in Figure 9-1. Between thesurface and 3000 feet (1 km), the windincreases considerably. Between 3000 and20,000 feet (1 and 6 km) there is littlechange. Izawa attributed the lower surfacewinds to friction since most of his data werecollected from coastal and large islandstations. Over the ocean, winds increase lesswith height near the surface. In general,western Pacific typhoons are larger thanNorth Atlantic hurricanes. Tropical cycloneadvisories for the North Atlantic and NorthPacific include the observed and forecastradii of 50 knots sustained winds in additionto the maximum sustained winds.

The circulation of intense tropical cyclonesextends upward to around 46,000 to 49,000feet (14 to 15 km) (close to the tropicaltropopause). Since the cyclones are warm-cored, the cyclonic circulation weakens withheight (Figure 9-1). But up to about 20,000feet (6 km), the shear of the wind in thevertical is small. The circulation may bedivided into three layers. The inflow layerextends from the surface to about 10,000 feet(3 km). It contains a pronounced componentof motion toward the storm center. Thisinflow is largely confined to the planetaryboundary layer, below 3000 feet (1 km).However, tropical cyclone circulation com-posites constructed by Gray (1978) showinflow extending up to 23,000 feet (7 km). Inthe middle layer from about 10,000 to 25,000feet (3 to 7.6 km), the flow is mostly tangen-tial, although Gray would disagree. Theoutflow layer extends from 25,000 feet (7.6km) to the top of the storm with maximumoutflow in mature storms near 39,000 feet(12 km). The characteristic horizontal circu-lation patterns in the inflow and outflowlayers are shown in Figure 9-2. Low-level

inflow is most pronounced in the easternsemicircle. Typically, winds are strongest tothe right of the direction of movement andin the eye wall. In the upper troposphere,outflow is cyclonic. This circulation is muchsmaller than near the surface and is sur-rounded by anticyclonic flow (correspond-ing to the negative values of the tangentialwinds in Figure 9-1). Strongest winds andstrongest divergence extend outward in thenorth semicircle of the storm. Such a patternis often reflected in the cirrus streamersshown in satellite pictures. A storm’s out-flow pattern depends on the environmentalwinds and on the dynamics of the stormitself (see Section E-5).

Very few measurements have been madeabove 20,000 feet (6 km) in the central re-gions of intense tropical cyclones, and soFigures 9-1 and 9-2B may not be truly repre-sentative of the upper tropospheric circula-tion. On 17 September 1990 the NASA DC-8research aircraft flew at 41,000 feet (12.5 km)across super typhoon Flo. The center waslocated near 24°N, 129°E, and was movingnorth at 6 to 8 knots. An omega sonde in theeye recorded a surface pressure of 891 mb,and maximum surface winds were esti-mated to be 140 knots. Maximum flight-level winds on four crossings of the eyewallaveraged 96 knots, a much smaller decreasewith height than suggested by Figures 9-1and 9-2. Even though pressure in the eyewas about 2 mb higher than in the eyewall,the circulation was strongly cyclonic, andflowed away from the center only beyond aradius of 160 to 270 NM (300 to 500 km).Fett (1964) was one of the first to studyhurricanes using manned spacecraft andearly satellite photographs. He identifiedtwo features found in many cyclones atsome time during their life-cycle: (1) arelatively clear annular zone of subsidencearound the rim of the storm’s high-cloudshield, and (2) an outer convective bandbeyond this annular zone. Later, Fujita et al.(1967) proposed a model (Figure 9-3) of the

outflow pattern from storms that have bothinner and outer rainbands or only innerrainbands (to simplify the analysis therainbands are shown as concentric circlesrather than the spirals to which they actu-ally conform). They deduced that verticaltransport of momentum by outer rainbandsalters the outflow wind pattern, resulting information of an upper-level shear-line and acloud-free annulus which separates theinner cirrus shield from the outer rainbands.

In summary, the three-dimensional windstructure of tropical cyclones comprises airflowing into the cyclone in the lower layers,rising primarily in the eye-wall cloud andthe other rainbands, and finally flowingoutward from the cyclone top and sinkingsome distance away. Forced sinking insidethe eye, by warming the air through adia-batic compression, contributes to the lowsurface pressure.

3. Temperatures. Tropical cyclones of atleast tropical storm intensity are warm-coredirect atmospheric circulations (i.e., warmair rises and cold air sinks) that convert heatenergy to potential energy and potentialenergy to kinetic energy. Their primaryenergy sources are latent heat of condensa-tion released in the eye wall and in the spiralrainbands, and sensible heat supplied fromthe ocean surface.

As surface winds spiral in toward the wallcloud of a tropical cyclone, speed increasesalong the trajectory. Evaporation from theocean, and the rate at which latent heat isadded to the air, also increase. Along thetrajectory, pressure decreases. The resultingadiabatic cooling is counteracted by sensibleheat added from the ocean. In the annulussurrounding the eye, the pressure gradientis very large and the winds are strong. Inthis zone, the air receives most latent andsensible heat from the ocean. This causes thewall cloud to be much warmer than the

peripheral air, and so maintains a direct,energy-producing circulation. Figure 9-4shows the temperature anomalies measuredin Hurricane Cleo on 18 August 1958(LaSeur and Hawkins, 1963). The uppertroposphere warmed most with tempera-tures 10°C or more above normal. Thelargest horizontal temperature gradientsoccur in mid-troposphere and are concen-trated in a narrow band extending acrossthe wall-cloud. The record may have beenset by super typhoon Tip on 12 October 1979between Luzon and Guam. At 700 mb, areconnaissance aircraft measured 30°C inthe eye, for an anomaly of 22.1°C. Gradientsare small within the eye, especially at lowerlevels. Outside the eye-wall, temperatures inthe lower troposphere are slightly belownormal, especially in the rear quadrant ofthe cyclone, where the surface waters havebeen cooled by evaporation and upwellingas the eye passed. The ultimate source ofboth latent heat (evaporation) and sensibleheat is the warm sea-surface (see Section E).

Riehl (1954) has compared tropical cyclonesto simple but very inefficient heat engineswith only about 3 percent of the total re-leased latent heat being converted intokinetic energy. Much of the rest of the heat isconverted into potential energy and ex-ported through the outflow layer.

4. Clouds. The major convective cloudsystems (“rainbands”) in tropical cycloneslie along spirals as shown by radar scansand satellite pictures. Upward motion isconcentrated in the rainbands and especiallyin the wall-cloud, where updrafts of 5 to 13m s-1 have been measured. Vertical transportof heat and conversion of potential to kineticenergy are concentrated in the rainbands.Figure 9-5 reproduces the Kadena (26°N,128°E) radarscope as the eye of typhoonNelson lay about 85 NM (160 km) southeastof the station, moving northeast at 12 knots.At Kadena, pressure fell to 987 mb, maxi-

mum sustained winds reached 38 knots, andrainfall totaled 8.35 inches (212 mm). Noteevidence for an inner eye, and how attenua-tion of the radar echo renders the southeast-ern wall cloud almost invisible. The cirrusshield near the cyclone center often preventsthe convective banded structure from beingseen in satellite pictures at visible wave-lengths. However, computer techniques canenhance the brighter areas caused by cumu-lonimbus penetrating the cirrus shield andthus help to locate the rainbands. Micro-wave imagers can also detect the spiralrainbands beneath a cirrus overcast. High-resolution infrared sensors give more detailon the convective activity. The more intensethe storm, the higher the convective tops. Itis not uncommon for them to exceed 49,000feet (15 km), especially in the eyewall.

Figure 9-6 shows an IR image of supertyphoon Flo and the enhancement curveused. At the time of the picture, the centralsurface pressure was 891 mb and the maxi-mum estimated surface wind, 140 knots.Emitting temperature of the uniform massof cirrostratus surrounding the eye (white)ranged between -70.2 and -75.2°C. Within it,the temperature of protruding cumulonim-bus tops (gray) ranged between -76.2 and -80.2°C. The inner surface of the eyewallranged over eight enhancement segments(1-8) from the sea surface to -75.2°C.

C. Classification and Definition ofTropical Disturbances.

Tropical cyclones form over all the tropicaloceans except for the South Atlantic and theSouth Pacific east of about 130°W. Regionaldifferences in terminology are listed in Table9-1.

In addition to the classification of Table 9-1,the United States meteorological servicesdefine a tropical depression as a weaktropical cyclone with a surface circulationincorporating one or more closed isobars,and highest sustained winds (averaged overone-minute or longer periods) of less than34 knots. Extreme surface wind gusts intropical cyclones may be 30 to 50 percenthigher than the reported sustained surfacewind. US meteorologists also use the term“tropical disturbance”, defining it as adiscrete system of apparently organizedconvection, generally 75 to 250 NM (150 to500 km) across, originating in the tropics orsubtropics, having a non-frontal migratorycharacter and having maintained its identityfor a day or more. Tropical disturbancesmay subsequently intensify into tropicalcyclones.

———————————————————————————————-Table 9-1. Areas of occurrence of intense tropical cyclones and regional terminology(obtained from World Meteorological Organization, 1979). Range of Maximum Wind Speeds (knots)Region 34-63 64-165Western North Pacific Tropical Storm TyphoonBay of Bengal and Cyclone Severe CycloneArabian SeaSouth Indian Ocean Tropical Depression Tropical CycloneSouth Pacific Tropical Depression Tropical CycloneNorth Atlantic and Tropical Storm HurricaneEastern North Pacific

In some regions, “Tropical Storm” is subdivided into Tropical Storm — 34 to 47 knots, andSevere Tropical Storm — 48 to 63 knots.

Table 9-2. Monthly and average number of tropical cyclones per year for each major basin(from Crutcher and Quayle, 1974). — BASIN AND STAGE Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec YearNORTH ATLANTICTropical Storms - - - - 0.1 0.4 0.3 1.0 1.5 1.2 0.4 - 4.2Hurricanes - - - - - 0.3 0.4 1.5 2.7 1.3 0.3 - 5.2Total - - - - 0.2 0.7 0.8 2.5 4.3 2.5 0.7 0.1 9.4EASTERN NORTH PACIFICTropical Storms - - - - - 0.9 2.3 1.6 1.2 1.0 0.3 - 7.4Hurricanes - - - - 0.3 1.0 1.7 2.4 1.8 0.9 - - 8.2Total - - - - 0.5 1.9 4.0 4.0 3.0 1.9 0.3 - 15.6

WESTERN NORTH PACIFICTropical Storms 0.2 0.3 0.3 0.2 0 4 0 5 1.2 1.8 1.5 1.0 0.8 0.6 7.5Typhoons 0.3 0.2 0.2 0.7 0.9 1.2 2.7 4.0 4.1 3.3 2.1 0.7 17.8Total 0.4 0.4 0.5 0.9 1.3 1.8 3.9 5.8 5.6 4.3 2.9 1.3 25.3

SOUTHWEST PACIFIC AND AUSTRALIAN AREATropical Storms 2.7 2.8 2.4 1.3 0.3 0.2 - - - 0.1 0.4 1.5 10.9Hurricanes 0.7 1.1 1.3 0.3 - - 0.1 0.1 - - 0.3 0.5 3.8Total 3.4 4.1 3.7 1.7 0.3 0.2 0.1 0.1 - 0.1 0.7 2.0 14.8

SOUTHWEST INDIAN OCEANTropical Storms 2.0 2.2 1.7 0.6 0.2 - - - - 0.3 0.3 0.8 7.4Hurricanes 1.3 1.1 0.8 0.4 - - - - - - - 0.5 3.8Total 3.2 3.3 2.5 1.1 0.2 - - - - 0.3 0.4 1.4 11.2

NORTH INDIAN OCEANTropical Storms 0.1 - - 0.1 0.3 0.5 0.5 0.4 0.4 0.6 0.5 0.3 3.5Cyclones(1) - - - 0.1 0.5 0.2 0.1 - 0.1 0.4 0.6 0.2 2.2Total 0.1 - 0.1 0.3 0.7 0.7 0.6 0.4 0.5 1.0 1.1 0.5 5.7

- = less than .05; (1) = Winds 50 knots or moreMonthly values cannot be combined because single storms overlapping twomonths were counted once in each month and once in the annual.———————————————————————————————————

D. Global Climatology of TropicalCyclones.

Tropical meteorologists should realize thathistorical records of tropical cyclone inten-sity are incomplete. Over oceans lackingaircraft reconnaissance, the strongest surfacewinds were seldom encountered by shipsor island stations. Even today, storm inten-sity estimates based on satellite picturesoccasionally differ from aircraft reconnais-sance observations by about 20 knots. Theselimitations must be kept in mind when thefrequencies of various classes of tropicalcyclones are compared. Before weathersatellites, many smaller tropical cycloneswere missed entirely. Especially in theeastern North Pacific and the South Indian

Ocean, tropical cyclones are observed to bemuch more numerous with the advent ofweather satellite observations. This sectionpresents the global climatology of tropicalcyclones. An excellent survey paper by Gray(1978) provided most of the data.

Average annual frequency of tropical cy-clones in each generating area is shown inFigure 9-7. An annual average of only 80tropical cyclones develops world-wide(Gray, 1985). Three years of representativetracks are shown in Figure 9-8. Nearly halfof the storms occur in the North Pacific,about one-third in the Indian Ocean andAustralian waters and only about 11 percentin the North Atlantic. In Figure 9-9, theaverage number of tropical cyclones permonth relative to both the calendar and

solar year are presented. Note that themaximum frequency in the SH occurs ear-lier in the solar year (January) than in theNH (August). Table 9-2 shows the averagemonthly frequency of tropical cyclones of atleast tropical storm intensity for each of thestorm-development areas. These data arefrom Crutcher and Quayle (1974), except forthe eastern North Pacific, for which 22 yearswith satellite observations (1967-1988) havebeen used. In this area, in earlier years, alltropical cyclones were not picked up by shipreports (Sadler, 1964). Cyclones are ascribedto the month in which they began, and sothe monthly values add up to the annualvalue. For the other areas, records compriseat least 70 years.

— The data are discussed in more detaillater in this section.

In many areas, the climatology of tropicalcyclone occurrence goes back a hundredyears or so. However, if significant differ-ences are apparent between earlier cyclonestatistics and those based on more recentdata, only the latter are used. Any long-termtrends in tropical cyclone frequency defyidentification in most areas because of long-term changes in data and methods of track-ing.

Tropical cyclones originate in surfacetroughs, characterized by westerly winds ontheir equatorward sides and easterly windson their poleward sides. Figure 9-10 showsthe average locations of the monsoontroughs and tradewind convergence zonesfor February, May, August and November.Following Sadler (1967), the areas of vortexorigin are labeled with capital letters andthe characteristics of each are briefly dis-cussed. Western North Pacific (Area A). Thisis the most active region. July throughNovember accounts for 77 percent of thecyclones, but they can occur in any month,and can be as intense in January or Februaryas at the height of the typhoon season. 70percent of the tropical storms intensify into

typhoons. Tropical depressions have asimilar distribution to tropical storms andtyphoons. Over 80 percent of tropical de-pressions further intensify into tropicalstorms or typhoons.

From September to December, and often inEl Niño years, the monsoon trough mayextend east to the date line or beyond, tocause cyclones to form in the MarshallIslands area or even to the south of Hawaii.South China Sea (Area A). About 15 percentof western North Pacific storms develop inthe South China Sea. The season extendsfrom May to December, with most activityin July through September. A minimumduring June coincides with the ridge of highpressure that is moving north to give theearly July dry spell over south China (Fig-ures 6-38 and 8-58).

Eastern North Pacific (Area B). This arearanks second only to the western NorthPacific in number of tropical cyclones. Moststorms occur from June through October,and over half in August and Septemberwhen the monsoon trough is farthest north.53 percent of the tropical storms becomehurricanes. Once formed, most cyclonesmove westward or northwestward into aregion of cooler sea surface and strongvertical wind shear, as illustrated by Figure9-11 (Sadler, 1964). Consequently the major-ity dissipate before they reach populatedareas (see Section F). Rarely, however, avortex travelling westward south of Hawaiiwill intensify into a hurricane and affect thecentral Pacific, as did Hurricane Sarah in1967. It passed Johnston Island (17°N,170°W) and moved directly over WakeIsland (19°N, 167°E), causing considerabledamage. The monsoon trough usuallyextends westward to about 120°W; however,during active periods it can reach to near thedate line.

North Atlantic (Area C). In this region about80 percent of the tropical storms and hurri-canes occur in August to October. 55 percent

of the tropical storms reach hurricane inten-sity.

Over the eastern Atlantic, Area C resemblesArea B. Both are narrow warm-water zonesbetween cooler water to the north and south(see Figure 5-1). Maximum cloudiness alsooccurs in the westerlies south of the mon-soon trough (see Figure 6-16). As in Area B,the monsoon trough merges in the west intoa tradewind convergence and the Africanvortexes usually decay as they move intothe unfavorable environment of thetradewinds. However, several vortexesintensify into storms and hurricanes in themid-Atlantic before leaving the trough,while others may persist as low-level vor-texes (not necessarily at the surface) all theway to the Caribbean before dying, or rarelyintensifying. These observations are illus-trated by the life histories of cyclonesemerging from Africa during August 1963(Aspliden et al., 1965-1967) (Figure 9-12). Ofthe 11 cyclones tracked during this month,only two developed into hurricanes. Simi-larly, during Phase III of GATE, between 20August and 23 September 1974, ten Africanvortexes moved over the Atlantic; six dissi-pated before reaching 45°W. Of the remain-ing four, two became hurricanes (Sadler andOda, 1979). During Phases I and II of GATE,between 26 June and 19 August 1974, noWest African vortexes reached 45°W. Therather better performance during Phase IIISadler and Oda ascribed to northwardmovement of the trough over the ocean, andsouthward movement of the trough overWest Africa, resulting in an east-west align-ment. This helped prolong the lives of westAfrican vortexes, and resembles patterns inother years. Vortexes that leave West Africabetween 20 August and 10 September havethe best chance of crossing the Atlantic.

Sadler (1967)developed the schematic modelassociated with low-level cyclones in theNorth Atlantic shown in Figure 9-13. Visual-ize the model as either a synoptic picture of

a chain of cyclones or as the life history ofone cyclone. In the model, the initial vortexfails to develop beyond tropical depressionintensity (A and B). The circulation remainsa vortex in the easterlies (C) to beyond mid-Atlantic before decaying further to an openwave (D). Associated cloud masses movewestward with the circulation; the dominantsystem south of the trough may fluctuategreatly from day to day. North Indian Ocean(Area D). Over both the Arabian Sea and theBay of Bengal, most tropical storms areconfined to early summer and autumn,when the surface trough lies near 10°N(Figure 6-33). Between June and September,the heat/monsoon trough, which has devel-oped discontinuously over Pakistan and theGanges Valley, ensures that no tropicalstorms can form over the Arabian Sea. Onlywhen the monsoon trough extends over theextreme north of the Bay of Bengal candevelopment occur there (see Chapter 8,Section C-2), and because the environmentis baroclinic, it is usually restricted to mon-soon depressions.

Southwest Pacific and Australian Area(Areas F and G). In this area and over thesouthwest Indian Ocean, a third or less of alltropical storms intensified into hurricanes, asignificantly lower proportion than over theNorth Pacific and the North Atlantic. Thismay be true, although poor observingnetworks could be to blame for unreliableintensity estimates.

The median recurvature latitude for tropicalcyclones is 15°, compared to 24° north of theequator. In general, tropical cyclones lyingequatorward of the upper troposphericsubtropical ridge move toward the west andpoleward. As they cross the ridge the zonalcomponent of their movement shifts to theeast. Over the western North Pacific the 200mb subtropical ridge moves 1000 NM(2000km) poleward between winter andsummer, but over the southwest Pacific/Australian area the ridge remains between

10° and 15°S throughout the year, andtropical cyclones recurve soon after devel-opment. The season is almost as prolongedas in the western North Pacific. This may beno coincidence. Early or late season cyclonesin one region could trigger late or earlyseason cyclones across the equator in theopposite region, and so extend the tropicalcyclone season in both regions. Keen (1982)identified 22 cross-equatorial pairs over thePacific in about 10 years (Figure 9-14). Thesatellite image of Figure 9-15 shows a tropi-cal cyclone pair over the west Pacific. Thenorthern cyclone became super typhoonRuss, and later passed south of Guam (fron-tispiece). The southern system did notintensify further. Surface westerlies and badweather coincide along the equator, wherecirrus plumes reveal strong upper tropo-spheric easterlies, evidence of an activeWalker Circulation (see Figure 2-7). South-west Indian Ocean (Areas E and F). Thisregion incorporates the ocean from the coastof Africa to 110°E. 73 percent of the stormsoccur from January through March and 50percent reach hurricane intensity. As overthe western and southwestern Pacific, in 11years 14 twin developments occurred northand south of the equator in the IndianOcean (nine in November-December).Figure 9-16 shows the tracks (Mukerjee andPadmanabham, 1977).

The latitude at which initial disturbances,which later became tropical storms, werefirst detected is shown for each develop-ment region in Figure 9-17A. The regionsdiffer significantly, especially in the NH. Inthe North Atlantic, proportionately morestorms develop above 20°N than in the otherregions. In the western North Pacific, stormsdevelop over a broad meridional rangebecause of relatively large seasonal andshorter period movements of the monsoontrough and because of developments initi-ated by upper-tropospheric cyclones. In theSH, the monsoon trough moves less, withmost storms developing within 250 NM (500

km) of the mean trough position near 13°S.Over the Bay of Bengal, if monsoon depres-sions are not considered, the peak develop-ment latitude shifts south to 10°N. Thecombined graphs for NH, SH, and globe areshown in Figure 9-17B. Overall, 65 percentof the disturbances that became tropicalstorms were first detected between 10° and20° latitude, 13 percent poleward of 20°, and22 percent equatorward of 10°.

The number of tropical cyclones in anymonth or year undergoes a largeinterannual variability. For August in thewestern North Pacific, the average is seven,but since 1940 the range has been from twoin 1947 and 1977 to 18 in 1950. Figure 9-18shows the frequency distribution of theannual number of tropical cyclones in sixdevelopment areas for 1958 - 1978 (Gray,1978).

The global climatology presented here canbe supplemented by regional publicationsthat incorporate much more detailedclimatologies. They include tropical cyclonetracks for each month (or, in some areas, forsemimonthly or ten-day intervals), fre-quency distributions and means of tropicalcyclone movement for various latitude-longitude rectangles, probability of tropicalcyclones affecting particular locations basedon their present position, and many otherdetailed statistics. The following annualreports give tropical cyclone tracks andintensities and movement forecast errors, aswell as illustrative analyses and satellitepictures:

“Annual Tropical Cyclone Report”, JointTyphoon Warning Center, Guam: west andSouth Pacific and Indian Ocean.

“Tropical Cyclone Summaries”, Royal Ob-servatory, Hong Kong: North Pacific west ofdate line.

“Monthly Weather Review”: North Atlanticand eastern North Pacific.

“Australian Meteorological Magazine”:southwest Pacific and Australian area.

“Technical Report - C5”, Mauritius Meteoro-logical Service: southwest Indian Ocean.

E. Formation.

Several conditions must be met beforetropical cyclones can form.

1. Adequate Source of Surface Energy. Thisgenerally limits the area to seas or oceanswith surface temperatures of 79°F (26°C) orhigher. According to Gray (1978), these arealso the regions in which the warm water isat least 200 feet (60 m) deep and wheremixing by the winds will not lower thesurface temperature too much. The 26°Cminimum requirement may not be necessar-ily valid. Tropical-type storms have formedin the Baltic Sea (Bergeron, 1954) and theMediterranean (Meyencon, 1983), and evenover the Arctic Ocean (Emmanuel andRotunno, 1989; Businger, 1991), where thesea-surface temperature was certainly lessthan 26°C. Rather, a small horizontal SSTgradient is needed — most often foundwhere the temperature is high. Then, asurface air parcel spiralling inward from thecyclone periphery toward the eye will re-place heat lost by adiabatic expansion byheat gained from the ocean, and reach theeye wall much more buoyant than the envi-ronmental air surrounding the cyclone. If asurface air parcel undergoes a 60 mb pres-sure drop as it moves along the pressuregradient toward the eye, it should cooladiabatically by about 5°C, but is observedto maintain a nearly constant temperature.The only way this can happen is for theocean to warm the air. Conversely, the lowheat capacity of the earth’s surface preventsit from heating the storm and so, by inhibit-ing buoyancy in the center of the storm,

either prevents development or causes adeveloped storm to dissipate.

2. Wind Shear. Small shear in the vertical ofthe tropospheric wind, allows the heatreleased by condensation to concentrate in avertical column, enhancing the surfacepressure fall. Figure 9-19 shows where shearis favorable. Above the tradewinds, windsmust be easterly or weak westerly. Duringsummer, surface westerlies of the Asianmonsoon may extend across the Philippinesand into the Caroline Islands, causing smallwind shear and favoring cyclone formationin Micronesia. A similar effect is producedover the western Atlantic. The large summershear over the Arabian Sea and the Bay ofBengal accounts for the absence of tropicalstorms there. The significant summer shearover the South China Sea is sometimesreduced when an upper-tropospherictrough extends westward from the NorthPacific across the Philippines and so favorstropical storm development (Sadler et al.,1968). In contrast, the low shear throughoutthe year near 10°N over the western NorthPacific allows development in any month,while during summer this region encom-passes the largest area of weak climatologi-cal shear, from 5° to 30°N. In the westernNorth Atlantic, mean wind shear less than10 knots extends to 33°N; in this regionproportionately more cyclones develop athigher latitudes than elsewhere (see Figure9-17).

3. Preexisting Disturbance. In general,active cyclone formation is confined toregions of small cyclonic shear in a summer-time surface trough (Gray, 1968). Localconcentrations of vorticity in the trough canhelp triggering mechanisms start tropicalcyclogenesis. Besides strong wind shear,absence of a surface trough over the ArabianSea also contributes to absence of summer

tropical cyclones there. A similar, but lessmarked, effect is present over the Bay ofBengal and the South China Sea. In all threeareas, tropical cyclones are more likely toform in the transition seasons, when surfacetroughs are present, and wind shear is least.Surface troughs are absent throughout theyear over the southeast Pacific east of about130°W, and over the South Atlantic.

4. Earth’s Rotation. Almost all cyclonesform poleward of 5° latitude, since a strongrotation can be generated only where theCoriolis force exceeds a certain minimumvalue. Disturbances near the equator rarelydevelop into tropical storms. However,Typhoon Kate (1970) grew to typhoonintensity at 4.5°N and moved east to westalong 5°N before recurving northward(Holliday and Thompson, 1986). Kate wasvery small compared to other typhoons withsimilar central pressures.

5. Upper-Tropospheric Outflow. All thepreconditions for tropical cyclone develop-ment so far listed prevail together over vastocean areas for weeks or even months at atime. Yet, in an average year, only 80 distur-bances intensify to or beyond tropicalstorms. In numerical models, the precondi-tions will lead to hurricanes only if themodel starts out with a surface cyclone ofabout 20 knots or more (Gray, 1978), morethan double what has usually been ob-served. Even then, no model duplicatesoccasionally-observed surface pressure fallsof more than 50 mb in 24 hours. So some-thing is missing.

The notion that the upper troposphere holdsthe key to rare and often rapidintensifications has become more generallyaccepted (see, for example, Riehl, 1979). Inmost generating regions, climatologicallydivergent upper-tropospheric easterliesoverlie and extend equatorward of the

monsoon trough (Figures 4-10 and 4-11).Poleward of the generating region, troughsin the upper-tropospheric westerlies or coldlows in TUTT may also cause divergence,far enough east and southeast of the strongperipheral winds to ensure small shear ofwind in the vertical (Figure 8-21). Thesedivergent outflows to north and south,acting on a surface disturbance, may causethe surface pressure to fall enough for atropical cyclone to form (see Sadler, 1976a;Elsberry, 1988). As a tropical cyclone devel-ops, the surface pressure fall in the eye isusually punctuated by short periods of nochange or of slight rises (see Sadler, 1976b).This suggests that the circulation respondsto a succession of intensification impulses.Tropical meteorologists, used to near-simul-taneous divergence/convergence compensa-tions in the atmospheric column leading toinsignificant surface pressure changes, havetrouble imagining an unbalanced upperdivergence sudden and intense enough tocause the observed surface pressure falls.Perhaps the divergence stems from ephem-eral small-scale jet “streaks” that escapediscovery in jet aircraft winds or in cloudmotion vectors derived from satellite im-ages. Reconnaissance and research aircrafthave only rarely probed the upper tropo-sphere near potential tropical storms. In theUS Navy typhoon field project of 1990(Elsberry, 1989), one-second measurementsmade on the NASA DC-8 flying at 190 mbover typhoon Flo may give some clues.

To summarize, there is general agreementon conditions necessary for tropical cyclonedevelopment. These are a warm, small-temperature-gradient ocean, more than 5°from the equator, a surface trough overlainby small wind shear through the tropo-sphere, and an essential condition of intensedivergence (outflow channel) in the hightroposphere. For example, if large-scaleatmospheric circulation anomalies at vari-ous levels combine to produce large windshear above a development area usually

characterized by small shear, this will pre-vent cyclone development. The conversewould favor development where cyclonesare normally rare. There may be regionaldifferences. In the North Atlantic, Africanvortexes in the monsoon trough have beenidentified as initiating disturbances. TheTUTT in the North Pacific and North Atlan-tic is a source of cyclones, especially thoseforming north of 20°N. The sequence illus-trated by Figure 8-21 is usually prolonged.Divergence east of a cold low, leading toconvection, may eventually reduce surfacepressure enough to produce a depression inthe tradewinds. If upper-level conditionsremain favorable, this circulation maybecome a tropical storm, and rarely, a ty-phoon or hurricane.

It was previously thought that easterlywaves grew into tropical cyclones. Over theAtlantic from 1960 to 1964, only six of the 43storms or hurricanes (14 percent) wereattributed to easterly waves. From 1965 to1969, with almost continuous satellite cover-age and more aircraft reports, none of thestorms were associated with easterly waves.(It should be noted that in that period theterm “easterly wave” was replaced by themore general term “tropical wave”, whichwas defined in Chapter 8). Even so, tropicalwaves were said to account for only four ofthe 28 storms occurring between 1967 and1969. Lack of conventional data over thecentral tropical Atlantic still handicapsresearch into tropical cyclone origins. Aerialreconnaissance starts only after a distur-bance has moved west of 55°W. Annualsummaries published in the MonthlyWeather Review suggest that mid- andupper-tropospheric circulation features andtheir interaction with lower levels may bemore important over the

Atlantic than elsewhere. Early in the hurri-cane season, mid-latitude influences areimportant. For example, development oftwo hurricanes and one tropical storm in

June 1968 was related to baroclinic processesin the mid-troposphere. Carlson (1969a,b)has shown that during the heart of thehurricane season (mid-July to mid-Septem-ber), most Atlantic tropical storms resultfrom intensification of low-level cyclonicvortexes that originate over Africa (Chapter8, Section C-3). They generally intensify overthe warmer waters of the western NorthAtlantic. One purpose of GATE in 1974 wasto observe the evolution of African vortexesinto tropical cyclones. This was thwarted bytoo little tropical cyclone activity. Early andlate in the Atlantic season cyclones tend toform in the southwestern Caribbean whenthe monsoon trough extends eastward fromthe eastern Pacific. The role of cold lows inthe subtropics has already been mentioned.In the western North Pacific 85 to 90 percentof tropical cyclones form in the surfacemonsoon trough. Ten to 15 percent developfarther north as the TUTT acts on the sur-face tradewinds (Figure 8-21). In the othertropical cyclone regions most developmentstake place when vortexes in the surfacemonsoon trough intensify.

On 11 July 1989, Typhoon Gordon suddenlydeveloped from convection located near thecenter of a cold low within the TUTT (JointTyphoon Warning Center, 1990). Within 30hours, the central pressure fell 70 mb. Asmentioned earlier, isolated, short-lived deepconvection often develops near the centersof cold lows, but a typhoon? Upper tropo-spheric convergence into the cold lowwould seem to preclude this possibility,making this very rare event a mystery.

F. Dissipation.

When conditions that favor development orpersistence disappear, hurricanes or ty-phoons dissipate.

1. Surface Energy Source Removed. Whenthe surface supply of sensible heat to thecore is cut off, the buoyancy of the system isdestroyed (Bergeron, 1954; Miller, 1964).Only a uniformly warm ocean surface cansupply this heat. Thus, the circulation weak-ens rapidly when it moves over land, theisothermal inflow at the surface stops, theair is now cooled by adiabatic expansionand the horizontal buoyancy gradient disap-pears. The same effect is produced whenrelatively cold or dry surface air penetratesthe core.

Forecasters should remember that cold airreaches the eye wall from west orequatorward of the center. If air swirlsaround the center before reaching the wall,the outward-directed temperature gradient(and the cyclone intensity) may temporallyincrease before the system starts to fill(Brand and Guard, 1979). After recurvature,a tropical cyclone may move too fast for thecold air to catch up; hurricane winds thenpersist into high latitudes. The worst windsever experienced in New Zealand werecaused by a tropical cyclone that movedsouth-southeast at 22 knots from the vicinityof New Caledonia in April 1968 (Hill, 1970).The stress of cyclonic winds transportsocean-surface water outward from a tropicalcyclone center (see Chapter 5, Section B-2).Ensuing upwelling of a stratified ocean(more likely above 20° latitude) may lowerthe sea-surface temperature by as much as9°F (5°C) (Leipper, 1967; confirmed by manymore recent measurements), but with a lagtoo great for the energetics of a movingcyclone to be affected. However, should thecyclone be stationary, it would weaken as itsbuoyancy decreased. For example, in De-cember 1989, Typhoon Jack remained sta-tionary for 48 hours. In the first 36 hours,maximum winds decreased from 120 to 30knots.

2. Excessive Wind Shear. If a tropical cy-clone moves beneath a strongly shearinglayer, the low-level center may becomedetached from the rest of the circulation,may be weakened by subsidence, andquickly dissipate. This sequence usuallyshows clearly in satellite images. Dissipationby shearing often occurs in the easternNorth Pacific (Figure 9-11) and in the east-ern North Atlantic, as tropical cyclonesmove west or northwest into thetradewinds.

3. Upper Tropospheric Convergence. Gen-erally, as a tropical cyclone approaches anupper tropospheric trough, it recurvespoleward beneath upper troposphericdivergence east of the trough, and remainsstrong. Occasionally, if the upper trough ismoving rapidly across the tropical cyclonetrack, it may pass over the cyclone. Uppertropospheric convergence then fills thecyclone (Figure 8-53).

G. Movement.

Various climatological publications providegeneralized mean tracks of tropical cyclonesby month or season. These tracks are drawnalong axes of maximum cyclone frequency.They are useful for some purposes; how-ever, they hide the characteristically largevariability of tropical cyclone tracks that canbe revealed by charts of the tracks of indi-vidual cyclones compiled from many yearsof record (see Figure 9-8). Recent tropicalcyclone tracks, based on aircraft reconnais-sance and satellite pictures, are more accu-rate than tracks for earlier years, especiallywhere other data are sparse.

The annual meridional movement of thesubtropical ridge lines controls the meanlatitude of recurvature of tropical cyclones.Data for the western North Pacific are

shown in Figure 9-20. January and Februarystorms generally dissipate before they canrecurve. In the western North Atlantic,mean recurvature positions determined as afunction of both latitude and longitude areshown in Figure 9-21 (Colon, 1953). Theenvironment in which a cyclone is embed-ded largely controls the track.

As a tropical cyclone recurves and movespoleward the influence of the environmentalwind field changes. South of the subtropicalridge and within the tropics, the cyclonedominates the circulation for a considerabledistance from the center and throughout thetroposphere. The warm core of the cyclonetransports large amounts of latent heat ofcondensation to the middle and uppertroposphere, inflating isobaric surfaces anddeveloping a ridge at these levels northeastof the cyclone center. If a cyclone recurvespoleward where the ridge is weak or bro-ken, its depth decreases from the top down.Then, as it moves beneath the upper wester-lies and starts travelling eastward, theclosed circulation is confined to the lowertroposphere. Thus, the influence of thecyclone on its surroundings decreases afterrecurvature and the large-scale environmentis then largely responsible for steering thestorm.

H. Forecasting.

1. General. Forecasting tropical cyclonemovement is highly specialized and can bestbe accomplished in central locations byexperienced forecasters using all availabledata. The National Hurricane Center inMiami, Florida, prepares forecasts for alltropical cyclones of depression intensity orgreater in the North Atlantic, Caribbean,Gulf of Mexico and eastern Pacific. In thewestern North Pacific, tropical cycloneforecasting is centralized at the Joint Ty-phoon Warning Center (JTWC) on Guam

(13°N, 145°E), which issues warnings as wellfor the Indian Ocean and the southwestPacific. Other weather centers also issuewarnings; in some regions the warning areasoverlap and then differing forecasts for thesame storm may confuse the user. Thetechniques used at the various centerschange from time to time in response toresearch results and to changes in methodsof observing and tracking the storms. Thesetechniques will be discussed briefly below,since field forecasters should know thecapabilities and the limitations of tropicalcyclone forecasts. The various centers com-pare and evaluate old and new techniquesin real time, but until now, no one systemhas proved to be best under all circum-stances. So, an official forecast is always asubjective decision; the techniques help, andestablish a kind of probability envelope, butthe forecaster has the last word. Resultssupport this procedure -rarely, over thecourse of a season, does any one techniquebetter the official forecast. The detailedcomparisons made at JTWC are illustratedby Table 9-3. In the table, “X-axis” refers totechniques listed horizontally, and “Y-axis”to techniques listed vertically. In each box,only forecasts made at the same times forthe same tropical cyclones are tallied. Thusthe comparisons are homogeneous.

2. Positioning. Satellite imagery has revolu-tionized cyclone reconnaissance. At leasttwo-thirds of all cyclone-center fixes arebased on satellite pictures. Ending aircraftreconnaissance in the western North Pacifichas further enhanced the value of weathersatellite observations. The geosynchronoussatellite makes it possible to continuouslymonitor cyclone positions and cloud mo-tions. Although the cyclone center positionas indicated by the satellite may differ fromthe true position, it is sufficiently accurate tosupport operations. Just what constitutesthe cyclone center — the geometrical center

of the eye as defined by the wall cloud(satellite), the minimum pressure (aircraftreconnaissance), the maximum temperatureor the circulation center is debated. Differ-ences can amount to as much as 30 NM (55km). Because of this and because of occa-sional mapping displacement by the satel-lite, there will always be some disagreementon center location, particularly in thepoorly-defined cases.

3. Techniques. Most tropical cyclone fore-casting centers combine subjective andobjective techniques to forecast movement.The subjective techniques depend on theexperience of individual forecasters andrules of thumb. Objective techniques includepersistence forecasts based on extrapolationof past positions, average tropical cyclonemovements in various areas based on clima-tology, and dynamic or statistical forecastmethods. Those based on satellite-observedcloud patterns (e.g. Lajoie, 1981) haveproved hard to apply, and are not usedmuch.

Dynamic techniques apply the same nu-merical prediction methods that are a fea-ture of mid-latitude forecasting. Thus cy-clone motion is automatically included inthe predicted pressure or pressure-heightfields. Since the dynamic model does notdepend on statistical relationships betweenthe current cyclone and historical cyclones,the model’s reliability should be indepen-dent of the details of the cyclone’s pastbehavior. The cyclone track is thus explicitlypredicted. Some dynamic models incorpo-rate sophisticated procedures such as nestedgrids and two-way boundary interactionswith larger-scale model output. Insufficientdata prevent the synoptic field from beinganalyzed accurately enough to meet theinitial field requirements of numericalmodels. In the North Atlantic, measure-ments by reconnaissance aircraft have beenused to fill in gaps. Perhaps new sensors

being developed for future satellites willlead to improvements in analyses.

Statistical techniques have been applied tomovement forecasting. They use multiple-discriminate-analysis methods to selectstatistical predictors from grid-point fieldsof surface pressure, upper-air pressure-heights, winds, thickness values, etc. Persis-tence, incorporating past positions of thestorm, is often included. Performance ofstatistical techniques is also sensitive toinput data and will generally be best wherethere is an adequate network of surface andupper-air observations.

Intensity forecasting relies heavily on inter-preting satellite pictures (Dvorak, 1975,1984). Enhanced infrared and digital infra-red satellite data now help determine inten-sity. The techniques all use cloud featuresand rules based on a statistical model oftropical cyclone development to determinethe current, and to estimate the 24-hr, inten-sity of a tropical cyclone. The visible andinfrared techniques differ mainly in thecloud features that are used in the analysisand how they are measured.

4. Position Forecast Accuracy. The annualreports listed earlier tabulate errors intropical cyclone position forecasts (see alsoTable 9-3). For each forecast, the error is thedistance between the forecast position andthe observed position as determined by the“best track” based on careful post-analysis.Forecasts of tropical cyclone intensity orperipheral winds are not usually evaluated.At JTWC, Guam (Annual Tropical CycloneReports), NHC, Miami (De Maria et al.,1990) and in the Australian Bureau of Me-teorology (Barnes, 1990) forecast modelscombining statistics and dynamics gave thesmallest position errors.

Figure 9-22 and Table 9-4 present and ana-lyze position errors in 24-, 48-, and 72-hourforecasts made for western North Pacific

tropical cyclones at JTWC, Guam, and forNorth Atlantic tropical cyclones made atNHC, Miami, from 1970 through 1989.Average errors over the 20-year perioddiffer little. Beyond that, significant differ-ences may stem from the mean latitude ofNorth Atlantic tropical hurricanes beingseven degrees farther north (27.7°N) thanthe mean latitude of western North Pacificcyclones (20.4°N) (Pike and Neuman, 1987).For the North Atlantic, the modest improve-ment over time is to be expected from bettermid-latitude numerical weather predictions,particularly for more than one day in ad-vance. No corresponding effect is discern-ible in the western North Pacific wheretropical cyclones stay deep within the trop-ics for most of their lives. The latitude effectis also seen in the standard deviation offorecast errors, which are more than twiceas large in the North Atlantic.

Average position errors are exactly corre-lated with forecast periods. This means thata statistical forecast that can be issued assoon as the observations are received has a15 to 25 NM (30 to 50 km) advantage over aforecast for the same valid time, but madelater by numerical models. Said anotherway, in the western North Pacific, shorten-ing the time to react to warnings by twohours would “improve” forecasts muchmore than they have been improved overthe past 20 years. Warning users should bemade aware of this. The best standard to usein evaluating forecast accuracy is persis-tence, i.e., extrapolation from past positions.

When cyclones move along climatologicaltracks, both forecast and persistence errorstend to be low. Errors increase when cy-clones move erratically, especially duringrecurvature. Not surprisingly, improvementover persistence increases with increasingforecast interval. For 12-hour forecasts,extrapolations from positions 6 to 12 hoursbefore are hard to beat. Ramage (1980)found that more than half the forecast errorvariance could be explained by the errors inpersistence forecasts. He also found thaterrors in the western Pacific and the Atlanticpersistence forecasts showed the sametrends -improvement to 1971 or 1972, thendeterioration to the levels of the mid-60s.The moderately significant correlationsbetween Guam and Miami forecast errors inTable 9-4 suggest that this relationship haspersisted into the 1980s and that the vari-ability in tropical cyclone tracks may re-spond to hemispheric causes.

Movement forecasting is more difficult insome tropical cyclone basins than in others.Pike and Neumann (1987) used the samepersistence/climatology model to make 24-hour movement forecasts for every basin forperiods of 33 years or more. The errorsmeasure the forecast difficulty (Table 9-5),and are generally related to mean stormlatitude. The southwest Pacific/Australiaregion is the exception, possibly becausethere, storms recurve at lower latitudes thanelsewhere (see Section D).

————————————————————————————————-Table 9-4. Average errors (NM) and trends in tropical cyclone positionforecasts made at JTWC, Guam (13°N, 145°E) and NHC, Miami (26°N, 80°W)for 1970 through 1989. 24-Hr 48-Hr 72-Hr Miami Guam Miami Guam Miami GuamAverage Error 113 119 239 235 374 353Standard Deviation 20 10 61 26 103 47Time Trend Correlation -0.30 -0.01 -0.25 0.08 -0.38 0.13Trend (NM yr-1) -1.03 -0.03 -2.64 0.32 -6.85 0.27

Miami/Guam Correlation 0.33 0.55 0.46

Average tropical cyclone position forecasterrors can be applied to daily forecast posi-tions for various time intervals to giveprobability density distribution. The distri-bution of position errors around the forecastpositions is closely approximated by circu-lar-normal, or in some cases elliptical-normal, distributions (US Navy WeatherResearch Facility, 1963). In the circular-normal distribution, the probabilities ofobserved storm positions lying withinvarious distances from the forecast positionscan be represented by a family of concentriccircles, whose radius kSv is related to theprobability level desired. Sv is the standardvector deviation of the observed about theforecast positions and the multiplier krelates to the various probability levels, P, asshown below.

P(%) = 30 50 70 90 99

k = .60 .83 1.10 1.52 2.15

For example, there is a 70 percent probabil-ity that the observed position of the stormwould be within a radius of 1.10Sv of theforecast position. A study of statisticalforecast-error data for Atlantic stormsshowed that the Sv values were nearly equal

to the average forecast error. Therefore, as afirst approximation, representative averageforecast errors obtained from Table 9-4 canbe used for Sv to establish probability levelsin tropical cyclone forecast positions. Forwestern North Pacific tropical cyclones, the24-hour forecast error is 119 NM (221 km)and Sv equals 119 x 0.83 = 99 NM, which isthe 50 percent probability radius. Figure 9-23 illustrates the procedure.

Despite the disproportionate effort made toobserve and understand tropical cyclones(Chapter 1), the results shown in Figure 9-22are discouraging. In a recent review, Chanand Holland (1989) were rather pessimistic.They concluded that satellites and aircraftreconnaissance have not helped improveintensity and position forecasts much.Synoptic flights to gather peripheral mea-surements, say in outflow areas, might beworth trying, since they could improve theinitial data for numerical prediction models.

Table 9-5. Average errors in 24-hour persistence/climatology movement forecasts oftropical cyclone centers for tropical cyclone basins (Pike and Neumann, 1987).

Mean ErrorBasin Storm Latitude NM kmSouthwest Pacific/Australian 20.1°S 130 241North Atlantic 27.6°N 113 210Western North Pacific 20.4°N 99 184Southwest Indian 18.4°S 87 161Eastern North Pacific 17.9°N 78 144North Indian 15.7°N 63 117————————————————————————————————-

Chapter 10Chapter 10Chapter 10Chapter 10Chapter 10SEVERE WEASEVERE WEASEVERE WEASEVERE WEASEVERE WEATHER IN THE TROPICSTHER IN THE TROPICSTHER IN THE TROPICSTHER IN THE TROPICSTHER IN THE TROPICS

A. General.A. General.A. General.A. General.A. General.

The tropics experience a wide variety ofweather, from the persistent, almost cloud-less skies of many subtropical lands duringthe dry season to the violent winds and rainof intense tropical cyclones. The term “se-vere weather” with its range of meanings isoften defined in relation to its effect onhuman activities. In this chapter, the follow-ing phenomena, often classified as severeweather, are discussed: thunderstorms,tornadoes, waterspouts, hail and relatedgusty surface winds; dust storms; extremewinds associated with tropical cyclones;turbulence; icing; and ocean waves. Heavyrains, which are also severe weather, havebeen treated in Chapter 6.

B. Thunderstorms.B. Thunderstorms.B. Thunderstorms.B. Thunderstorms.B. Thunderstorms.

1. Frequency. Thunderstorms are morecommon in the tropics than in higher lati-tudes. Figure 10-1 shows the mean annualnumber of thunderstorm days over the earth(World Meteorological Organization, 1956b).WMO defines a “thunderstorm day” as anobservational day during which thunder isheard at a station. Tropical continents re-ceive the most thunderstorms, with centersof over 140 days annually in South America,Africa, and southeast Asia. This distributionis confirmed by lightning observationsmade from DMSP satellites (Orville andHenderson, 1986) (Figure 10-2). Also, thereare probably more thunderstorms per thun-derstorm day over the continents than overthe oceans. Note how few lightning flasheswere recorded over the near-equatorial

convergence zones in the central and easternPacific and the western Atlantic (see Chap-ter 8, Section D-8). Over tropical oceans,almost all thunderstorms are associatedwith synoptic disturbances, while over land,air-mass thunderstorms due to convectiveheating and orographic lifting are common.In the equatorial belt, most thunderstormsoccur over land (Figure 10-3). 82 percent ofobserved thunderstorms are confined toSouth America, Africa and Indonesia, whilethe oceans account for only 18 percent.Figures 10-1, 2, and 3 give only the broad-scale distribution. More detail, such asmonthly frequencies, diurnal variations, etc.,can be found in climatological publications.A word of caution — of all commonly re-ported weather elements such as tempera-ture, humidity, rainfall, pressure, etc., thun-derstorm reports are often the least reliable.For example, stations making observationsonly during daylight invariably recordfewer thunderstorms than comparable first-order stations operating 24 hours a day.Although the WMO definition of a thunder-storm day is now generally accepted, thiswas not true in the past. The followingexamples illustrate the problem of obtainingreliable thunderstorm statistics.

(1) Figure 10-4 (Atkinson, 1967a) analyzesthe mean annual number of thunderstormdays over southeast Asia based on a carefulstudy of all available data. In some areas,the annual thunderstorm frequency is morethan double that shown in Figure 10-1.

(2) Figure 10-5 shows the mean annualnumber of thunderstorm days over

India from 1931 to 1960 (Alvi and Punjabi,1966). Analysis in this figure is based prima-

rily on thunderstorm statistics from first-order 24-hour observing stations; comparethese isobronts (lines connecting equalaverage numbers of days with thunder-storms) with the reported annual means formany other stations also plotted on the map.

(3) In Honolulu (21°N, 158°W), during thewinter of 1970-71, the first-order airportstation reported thunder or lightning on 32days. A sferic detector, with a range of 15NM (30 km) and operated at the Universityof Hawaii, 6 NM (11 km) from the airport,was activated on 69 days (Sherretz et al.,1971).

These examples emphasize the great careneeded to correctly interpret thunderstormstatistics. Ground-based lightning detection,although restricted to a few locations, isboth accurate and reliable; continuouslightning detection from GOES is plannedfor the mid-1990’s (see Chapter 3). Mostareas experience large year-to-year varia-tions in annual and monthly numbers ofthunderstorm days. A detailed study(Atkinson, 1967a) of thunderstorm variabil-ity over 12 years at 28 stations in Thailand,illustrates this. To increase the sample size,the individual monthly numbers of thunder-storm days were combined for station-months which had the same mean numberof thunderstorm days and standard devia-tions, and frequency distributions deter-mined for these grouped samples. Onlystation-months with a mean of three ormore thunderstorm days were considered.Figure 10-6A shows the resulting frequencydistributions for selected categories ofmonthly means, e.g., the bar graph showinga monthly mean of 3.6 days includes theindividual yearly values for station-monthswith a mean of between 3.0 and 3.9 days.Since the resulting distributions are approxi-mately normal, standard deviations werecomputed for each one-day mean incre-ment. In Figure 10-6B these standard devia-tions are plotted against the mean monthly

number of thunderstorm days and a regres-sion line fitted by eye to these points. Themeans and standard deviations increasetogether up to a mean of about 16 days,above which the standard deviation de-creases for further increases in the mean. Todetermine the percentage range in thunder-storm days that can be expected in anymonth, cumulative frequency distributionsof the yearly values for station-months withsimilar means (such as those in Figure 10-6A) were plotted. The initial family ofcurves was smoothed slightly and used tointerpolate the percentage values formonthly means of 3 to 24 thunderstormdays. This derived family of curves (Figure10-6C) can serve as a model for estimatingthe year-to-year variation in thunderstormdays for any month, given the meanmonthly value. Climatological recordsusually contain only mean monthly or meanannual numbers of thunderstorm days. Ifthe frequencies are assumed to be normallydistributed, then the probability of thunder-storm days for a month (Figure 10-6C) or ayear (Figure 10-6D) can be estimated.

The model is used as follows: assume weseek the probability of observing 15 thun-derstorm days in a month having a long-term mean of nine days. We move up thecurve representing this mean until it inter-sects the horizontal line corresponding to 15days and read the percentage of occurrencefrom the top scale (in this case, 10 percent).Thus, on average, we could expect 15 ormore thunderstorm days in this month inone year out of ten. Suppose we need toknow the 90 percent range (from the 95th tothe 5th percentile) of thunderstorm days fora station with a monthly mean of 16 thun-derstorm days. Entering the curve repre-senting this mean we read the number ofdays at the 95th and 5th percentiles (in thiscase 9 and 24 days). This tells us that in 90percent of the years this station can expectbetween 9 and 24 thunderstorm days in thismonth.

The same procedure was used (Figure 10-6D) to relate the variation in the annualnumber of thunderstorm days to the meanannual value. Because of the small samplesize, which gives much larger errors ofestimate at the extreme percentiles, thecurves are limited to the 90 percent range ofvalues. Although the models of Figure 10-6were developed for stations in Thailand,they might also be used for other tropicallands where similar year-to-year variationsin the annual and monthly numbers ofthunderstorm days occur. Diurnal variationis discussed in Chapter 7.

2. Duration. Rain falls from individualthunderstorm cells for about an hour or less;however, continuous thunderstorms lastmany hours (Chapter 6, Section C-1-b).Climatological summaries are usually basedon hourly- or three-hourly synoptic observa-tions — inadequate to determine thunder-storm durations. Local and special observa-tions are needed. Figure 10-7A shows thedurations of all thunderstorms at MactanAir Base (10°N, 124°E) in the Philippines forJune through August 1965, based on theWBAN forms for surface observations(Atkinson, 1967b). Figure 10-7B shows thefrequency distribution of thunderstormdurations at Mactan for the rainy seasonsJune-November 1965 and May-November1966 based on data similar to those of Figure10-7A. Each period of thunderstorm activity

is counted as a separate occurrence regard-less of the length of the intervening interval.The graph of Figure 10-7C combines succes-sive periods of thunderstorm activity sepa-rated by less than two hours into one occur-rence. Of course this produces more long-duration storms. At Mactan, thunderstormscommonly last from 30 minutes to one hour,while very few last more than two hours.Mactan, on a small island 10 km east ofCebu City, experiences mostly nocturnalthunderstorms (Figure 10-7A) probablygenerated from convergence between a landbreeze and the prevailing offshore wind.Representations similar to Figure 10-7 areeasy to prepare and readily acquaint newly-assigned forecasters with the thunderstormactivity patterns at any station.

The average thunderstorm duration at anystation can be roughly estimated by divid-ing the mean number of hourly observationsof thunderstorms in any period (i.e., month,season, year) by the mean number of thun-derstorm days in that period (Table 10-1).Durations range from 0.9 hours atChiangmai and Songkla to 2.1 hours atBangkok. The overall average of 1.5 hours isprobably typical of most of the tropics.

3. Heights. As in higher latitudes, theheights of tropical thunderstorm tops varyconsiderably; most range between 30,000and 50,000 feet (9 and 15 km). In general,

Table 10-1. Thunderstorm duration — The ratio of the mean annual number of hourlyobservations reporting thunderstorms to the mean annual number of thunderstorm days forstations in southeast Asia.

THAILANDChiangmai (19°N, 99°E) 0.9 Korat (15°N, 102°E) 1.5Bangkok (14°N, 101°E) 2.1 Songkla (7°N, 101°E) 0.9

VIETNAMPleiku (14°N, 108°E) 1.6 Saigon (11°N, 107°E) 1.5Soctrang (10°N, 106°E) 1.6

PHILIPPINESClark (15°N, 121°E) 1.5 Cebu (10°N, 124°E) 1.7

tops are higher over the continents thanover the oceans (see Figure 6-23), probablybecause of surface heating and orographiclifting. Except in intense tropical cyclones,where thunderstorm tops in the eye-wallcloud and feeder bands may range between50,000 and 60,000 feet (15 and 18 km), thun-derstorm tops over the tropical oceans liebetween 30,000 and 40,000 feet (9 and 12km). Deshpande (1964) analyzed cumulon-imbus tops over India during the summermonsoon (June-September) of 1957 through1962 using 322 reports made from jet aircraftflying near the tops. The observations werefairly evenly distributed over India. Thefrequency distributions were as follows (seeinset).

Most of the tops lie between 35,000 and50,000 feet (10.7 and 15.2 km) with a medianheight between 40,000 and 45,000 feet (12.2and 13.7 km). 60,000 feet (18.3 km) markedthe upper limit. There was no significantregional variation.

Five years of radar data were used to deter-mine the height of cumulonimbus tops overthe plateau and plains of central India(where local orographic effects are slight)(Thomas and Raghavendra, 1977). Duringthe pre-monsoon months of March to May,when a showers regime prevails, 12 percentof all tops exceeded 40,000 feet (12 km).During the summer monsoon months ofJune to September, the figure was only 6.2percent. This is considerably lower thanwhat Deshpande found and emphasizes thedifference between a continuous series ofmeasurements such as these and the spotobservations Deshpande used. Puniah(1973) analyzed about six years of mainlyover ocean flights east and southeast ofCalcutta (23°N, 88°E). 86 percent of the tops

ranged between 32,000 and 42,000 feet (9.8and 12.8 km), and were very rarely higher.Overall, tops were about 6500 feet (2 km)lower than the summer monsoon topsreported by Deshpande (1964).

In the tropics, cumulonimbus tops rangeover the same height as cirrus (see Chapter6, Section B-6g), while they must reach30,000 feet (9 km) before developing intothunderstorms.

C. TC. TC. TC. TC. Tornadoes and Wornadoes and Wornadoes and Wornadoes and Wornadoes and Waterspouts.aterspouts.aterspouts.aterspouts.aterspouts.

A tornado is a violently rotating column ofair, pendant from a cumulonimbus, andtouching the ground. The sense of rotationis usually cyclonic. On a local scale, it is themost destructive of all atmospheric phe-nomena with wind speeds estimated at from90 to 270 knots in its vortex, several hundredmeters in diameter. All continents havetornadoes, but they are most common in theUnited States and Australia.

Although they can occur throughout theyear and at any time of the day, they aremore frequent in spring and during themiddle and late afternoon. A tornado overwater — a waterspout — is much less vio-lent. Waterspouts develop more often in thetropics and subtropics (Huschke, 1959). A“funnel cloud” is a tornado that does nottouch the ground. Tornadoes and theiraccompanying death and destruction arewell-known to mid-latitude residents anddetailed statistics have been compiled.Tornadoes are rare in the tropics, of whichfew or no statistics are available. Perhapstheir reported frequency will increase as

————————————————————————————————————INSET. Frequency of thunderstorm height.

Height (000s feet) 25-30 30-35 35-40 40-45 45-50 50-55 >55 (km) 7.6-9 9-10.7 10.7-12.2 12.2-13.7 13.7-15.2 15.2-16.8 >16.8% Frequency 2.5 6.2 21.1 42.5 20.8 6.2 0.6

observing networks expand and communi-cations improve.

The India Meteorological Department hasrecorded 42 tornadoes between 1951 and1980 (Singh, 1981). Thirty-three occurred inMarch through May, and as Figure 10-8shows, 27 were confined to West Bengal andBangladesh. Development there was helpedby low-level moisture from the Bay of Ben-gal and the lakes and rivers of Bangladesh.The tornadoes were associated with scat-tered thunderstorms and generally formedeast of troughs in the upper-troposphericwesterlies. Usually, upper-tropospheric andmiddle- to lower-tropospheric wind maximawere present and wind shear was largebetween the lower maximum and the sur-face. Thus, preconditions resembled those ofthe Gulf Coast region of the United States.

Prior to 1960 tornado data were generallylacking for southeast Asia. Since then funnelclouds have been reported in the region,most often in the Mekong Delta of Vietnam(Sands, 1969), and waterspouts have beenobserved near Hong Kong (22°N, 114°E) andGuam (13°N, 145°E).

Tornadoes, waterspouts and funnels arefairly common during winter in Hawaii(Schroeder, 1976). The clustering of observa-tions around airports (Figure 10-9) stronglysuggests many more unreported occur-rences. Funnels are most likely just east of adeep trough in the upper-troposphericwesterlies, when the tradewinds and thetradewind inversion are absent. Heavy rainoften falls.

Gerrish (1967) prepared a detailed climatol-ogy of tornadoes, waterspouts, and funnelclouds over south Florida within 75 NM(140 km) of Miami (26°N, 80°W) for 1957through 1966. As in Hawaii, many occur-rences are missed by regular reportingstations, so Gerrish added to the reportsobservations from small airfields, parkrangers, news media, and other unofficial

observers. He found that in the 10-yearperiod at least 56 tornadoes and 218 water-spouts occurred during 47 and 139 daysrespectively. In addition, on 224 days 314funnel clouds were observed in the absenceof tornadoes and waterspouts. Figure 10-10A shows the monthly frequency distribu-tions and Figure 10-10B the diurnal varia-tions (hourly values smoothed by a three-hourly running mean). Figure 10-11 locatesthe events. Most occurred during the warmseason (May through October). Tornadoesare about equally likely through the sum-mer, while waterspouts and funnel cloudshave a sharper summer maximum. Diurnalvariations differ. Tornadoes are most com-mon in late afternoon, and funnel clouds inearly afternoon, whereas waterspouts favorthe morning with a slight secondary maxi-mum in late afternoon. In Figure 10-11 allphenomena are concentrated near the coast;but once again this distribution is probablybiased by population density.

Florida tornadoes, particularly those insouth Florida, are generally neither as se-vere nor as destructive of life and propertyas those in the central United States. Gerrish(1967) found that tornadoes are more oftenaccompanied by lightning than are water-spouts. He cautioned, however that water-spouts are potentially dangerous to lightaircraft and boats. During

the cool season (November through April),almost all tornadoes and many waterspoutsare associated with fronts or squall lines.But during summer, they usually accom-pany surface ridges or other “benign”tradewind patterns with no upper-level jetstream. This makes warm season forecastingextremely difficult. Tornadoes generallygenerate significant radar echoes; however,waterspouts occasionally occur with cloudshaving small or no radar echoes. Relatingtornado occurrences to temperature andmoisture aloft, Gerrish found that lessinstability and moisture are prerequisite to

tornadoes in south Florida than along theGulf Coast. Also, tornadoes, waterspoutsand funnel clouds are unlikely when thetradewind inversion is lower and strongerthan normal.

During the Lower Florida Keys WaterspoutProject of May to September 1969 (Golden,1973), 397 waterspouts and funnel cloudswere observed within 50 NM (90 km) of KeyWest (25°N, 82°W), many more than wererecorded by Gerrish (Figure 10-10). 95waterspouts were documented by aircraftduring the project. They generally formed inlines of growing cumulus congestus, andlasted from 1 to 22 minutes. In the moreintense, surface tangential winds exceeded140 knots (Golden, 1974). In September1974, instrumented aircraft flying above1300 feet (400 m) altitude penetrated 16lower Florida Keys waterspouts (Leversonet al., 1977), seven of them more than once.Never did maximum vertical velocitiesexceed 20 knots nor tangential velocities 60knots.

1. Hurricane-Generated Tornadoes.Novlan and Gray (1974) and Gentry (1983)developed a climatology of hurricane-spawned tornadoes over the United States.For the period 1948 through 1981 (excludingHurricane Beulah (1967) which had 141tornadoes), tornado-producing hurricanesspawned an average of ten each. The geo-graphical distribution shown in Figure 10-12emphasizes the contributions of a few cy-clones. These reports pretty well agree withothers (Smith, 1965; Pearson and Sadowski,1965; and Hill et al., 1966) in locating torna-does generally in the right front quadrant ofthe hurricane, within 150 NM (300 km) ofthe center (see Figure 10-13). The area ofgreatest tornado concentration lies outsidethe general area of hurricane-force winds,the mean radius of which is only about 35NM (70km) during tornado occurrence.After first eliminating tornadoes associated

with the extratropical stage of the tropicalcyclones, Hill et al. (1966) concluded that 94percent occurred between the azimuths of10° to 120°. Most of these tornadoes were 60to 240 NM (110 to 440 km) from the cyclonecenter. Hurricane tornadoes accompaniedthunderstorm cells in both the inner andouter rainbands (as well as in scatteredouter convective cells); the stronger cells ofthe outer rainbands of the storm are mostlikely to spawn tornadoes.

Smith (1965) found that 82 percent of hurri-cane tornadoes from 1955 to 1962 occurredbetween 03 to 04 and 15 to 16 LT. For Hurri-cane Isbell in 1964, Pearson and Sadowski(1965) found a pronounced peak between 15and 18 LT; however, little diurnal variationwas evident for the remaining 1964 cases.For Hurricane Beulah (1967), 80 percent ofthe tornadoes occurred between 03 and 15LT with a pronounced peak (32 percent)between 09 and 12 LT (Orton, 1970). Hurri-cane tornadoes tend to occur in daylight,when they are easier to observe. Gentrypointed out that all hurricanes crossing thecoast of the United States betweenBrownsville, Texas, and Long Island, NewYork have associated tornadoes. Convectionis usually most intense in the right frontquadrant of the hurricane. Tornadoes usu-ally develop in this quadrant (Figure 10-13)in air that has moved over land for less than25 NM (50 km). The frictional slowing of thesurface wind then creates a large verticalshear strongly favoring development.

Forecasters could use Novlan and Gray’s(1974) advice:

“When a hurricane moves inland and startsto decay, persons located particularly in theright front section of the storm should beobservant for surface winds of 13 - 17 knotswhile overhead the low level clouds appearto be moving with much greater velocities.These conditions signify tornado potentialin the regions where cumulus convection is

occurring”. Table 10-2 provides a forecastwork sheet.

Over Japan, between 1950 and 1971, tor-nado-producing typhoons averaged only 2.3tornadoes each (Fujita et al., 1972). Com-pared to the United States, hazards in otherregions are likely to be fewer. Over theoceans, where frictionally-induced increasein the shear of wind in the vertical is verysmall, waterspouts have never been re-ported in association with tropical cyclones.

DDDDD. Hail.. Hail.. Hail.. Hail.. Hail.

Hail is defined as precipitation in the formof balls or irregular lumps of ice, producedby convective clouds, usually cumulonim-bus (Huschke, 1959). By WMO convention,hail is 5 mm or more wide; smaller iceparticles are termed ice pellets. Large hail ismore than 2 cm across. The following shortsummary of hail in mid-latitudes, from Coleand Donaldson (1968), probably applies tothe tropics. Large hail is found in, alongside,and under well-developed thunderstormswhose tops may extend above 50,000 feet

Table 10-2. Hurricane-tornado forecast work sheet (Novlan and Gray, 1974).TORNADOES LIKELY TORNADOES UNLIKELY1. Intense hurricanes or those Weak hurricanes ortropical cyclones increasing filling tendency justin intensity just before land fall. before land fall.

2. After hitting land After hitting landhurricanes fill at the rate of hurricanes fill atgreater than 30 mb 12 h-1. less than 10 mb 12 h-1.

3. Once on shore the center Only small centralrapidly cools and becomes storm cooling.6°C colder than 100 NM (185 km)from the storm center.

4. Vertical wind shear profilesSurface to 850-mb wind surface to 5000 feet (1.5 km) ofshears are less than 40 knots or more. Surface winds40 knots. only 15-20 knots.

RECOMMENDED TORNADO WATCH AREAGeneral area: Surface pressure is between 1004-1012 mb.

Specific area: Areas of vertical wind shear greater than 40 knots from the surface to 850mb. Surface winds 15-20 knots.

Specific area: Within the “preferred sector”: 60-250 NM (110-460 km) from the center ofthe hurricane and at an azimuthal range of 0° to 120° with respect to true north.

Specific area: Along strong radar-observed rainbands, particularly the outer rainbands.

Begin tornado watch: When the center is 100-150 NM (180-280 km) off shore and the firstrainbands start to come on shore.

End tornado watch/warning: When the rainbands begin to break up and the vertical windshear (Sfc-850 mb) falls below 40 knots.

Note: Significant dry air intrusions in the right rear quadrant indicatea potential for tornado “family” outbreaks.————————————————————————————————

(15 km); however, small hail or ice pelletsmay occur in any thunderstorm and even insome large cumulus. Hail reaching theground is most common over mid-latitudemountains and adjacent plateaus, wheresome places may experience five to tenhailstorms a year. The thunderstorm cellsare about 4 to 8 NM (8 to 16 km) wide; theinstantaneous width of the hailing area atthe ground ranges from 1 to 3 NM (2 to 5km). Although hail can form only at alti-tudes above the freezing level, it may beencountered in flights from the surface tovery high levels.

The chance of hail reaching the groundincreases with the height of the radar-echotops of the associated thunderstorm. Forexample, half of the thunderstorms in NewEngland with radar-echo tops above 50,000feet (15 km) caused hail at the ground. InTexas, hailstorm echo tops occasionallyexceed 60,000 feet (18 km). Hail melts muchless before reaching the ground in theUnited States than in the tropics. Thus, it isnot surprising that hail measured at theground in the United States is twice as likelyto be larger than 1 inch (25 mm) than overIndia.

Over the tropics, hail is fairly rare comparedto higher latitudes. This is because thefreezing level is usually above 15,000 feet(4.5 km) and thunderstorm updrafts areseldom strong enough to support hailgrowth. However, it has been observed, andso meteorologists must consider the possi-bility of at least small hail in all severetropical thunderstorms.

In Hawaii, hail falls five to ten times a year.It is almost always quite small and affects asmall area. But on 30 January 1985, as a coldfront passed through, thunderstormsdropped hail over much of the northeasternand southeastern parts of the island ofHawaii. Takahashi (1987) analyzed 107hailstones with an average width of 0.71inches (18 mm). Since records were kept,

south Florida has experienced three stormsgiving large hail. On 29 March 1963, a se-vere thunderstorm produced hailstones upto 3 inches (75 mm) wide in Miami as theaxis of a cold upper low crossed the city(Neumann, 1965).

Frisby (1966) and Frisby and Salmon (1967)made the most complete survey of hail inthe tropics (between the Tropics of Cancerand Capricorn). They gathered data formany years from the literature and throughextensive correspondence with individualsand meteorological services. Since hail is sorare and localized, routine climatologicalsummaries often miss most of the events.

For Africa, India, and Australia, data weresufficient to allow isopleths of hail occur-rence to be drawn (Figure 10-14). Althoughhail is rare at African coastal stations, mostinland stations, including those near sealevel, have occasionally reported hail. Hail iscommonest over the highlands of EastAfrica where in a few places the meanfrequency exceeds five days a year. In theKenya highlands, hailstones more than oneinch (25 mm) across are not uncommon,although here and elsewhere in the tropicshailstones are generally less than 0.4 inches(10 mm) wide. In India, hail is most com-mon in the north and along the southwestcoast. In tropical Australia, hail is primarilyconfined to Queensland and is rarer thanover parts of Africa and India.

Frisby (1966) analyzed a scatter diagram ofannual hail frequency at tropical stationsversus station elevation. She found that onaverage most stations below 4000 feet (1200m) experience fewer than two hail days peryear, while at higher stations, hail falls onthree to eight days per year. There wasconsiderable scatter of points about the lineof best fit, suggesting that other factorsbesides elevation play roles in hail occur-rence. Frisby also prepared monthly hail-storm distribution graphs for stations withreliable data and then composited the

graphs for areas that appeared to be part ofsimilar hail regimes. Three main tropicalhail regimes were revealed — Zone 1 (10°Nto 23.5°N), Zone 2 (10°N to 10°S), and Zone3 (10°S to 23.5°S). Figure 10-15 indicates themonths of highest frequency in each zone.In Zones 1 and 3, hail principally falls inspring — the transition season precedingthe summer rains. This is when increasedinsolation combined with incursions of coolair aloft, associated with upper westerlytroughs, create maximum instability. In theNH (Zone 1), the months are usually Febru-ary to May and in the SH (Zone 3), Septem-ber to December. In the equatorial region(Zone 2), hail appears to have two seasonsand in some places it occurs throughout theyear.

EEEEE. Extreme Winds.. Extreme Winds.. Extreme Winds.. Extreme Winds.. Extreme Winds.

1. General. In the tropics, extreme windsare usually caused by thunderstormdownrush or by tropical cyclones. In thesubtropics strong winds may also accom-pany deep extratropical cyclones. Channel-ing by terrain may further accentuate strongwinds. A vigorous surge in the winter north-east monsoon can cause 50 knot windsthrough the Taiwan Strait and passes of theAnnam mountains in southeast Asia, whileFigure 8-36 shows how passes across Cen-tral America funnel the tradewinds. Stronglow-level winds often blow over the Horn ofAfrica during summer. These thermally-driven winds can exceed 100 knots in thelowest 6500 feet (2 km) (Findlater, 1977).

The highest sustained wind speeds shouldbe distinguished from peak gusts. Thesustained speed is the mean speed oversome short averaging period — 1 min, 5min, or 10 min —while the peak gust is thehighest value recorded by an anemograph.Peak gusts usually last less than 20 seconds.Compared to one-minute sustained winds,

the peak gusts may be 20 to 50 percenthigher (the gust factor), depending on thetopography. anemograph type, its heightand exposure, and other circumstances andshould be determined for each location.Inland, a 50 percent gust factor is typical,while along the coast or over the ocean,where friction is less, 20 percent is morelikely. For example, as typhoon Russ ap-proached and passed to the south of Guam(Frontispiece), near the center of the island,at Agana Naval Air Station (13°N, 145°E)gusts were highly correlated with sustainedsurface wind speed (r = 0.96) and were 37percent stronger. In what follows in thisSection, peak gusts are used since they areeasier to determine from anemograms thanare sustained winds. If desired, the sus-tained winds associated with peak gusts canbe estimated by applying the appropriategust factors.

2. Thunderstorm Winds. Almost all tropi-cal thunderstorms (except continuous thun-derstorms) produce downrush winds;however, speeds depend on geography,season, time of day, and the character ofindividual thunderstorm cells. The strongestwinds are seldom measured by the surfacestation network. Even so, enough data onthunderstorm winds have been recorded onanemograms to allow a fairly good descrip-tion.

In India, strong downrush winds producedby thunderstorms are called “squalls”,which are defined as sudden increases inwind of at least three intervals on the Beau-fort scale, reaching a speed of 24 knots ormore and lasting at least one minute. A“severe squall” reaches 43 knots or more.Sharma (1966) used anemograms for 1954 to1963 to study squalls at Nagpur (21°N, 79°E)in central India (Figure 10-16). Squalls wererecorded in every month except November.They were most common from Marchthrough July, and reached an average of 12

in June. Highest gusts of 74 knots occurredin both March and April. With the squalls,pressures generally rose and temperaturesfell suddenly (bottom of Figure 10-16). Thepressure changes ranged from -1.3 mb to+4.1 mb and temperature changes from+11.2°F to -27.4°F (+6.2°C to -15.2°C). Squallfrequency averaged 36 over the year andranged from 16 in 1954 to 51 in 1962. 80percent of the squalls and all of the severesqualls occurred with thunderstorms, theremainder with heavy showers. Over 80percent of the squalls occurred between 12and 21 LT. The frequency distribution ofsquall duration is shown below.D 1-5 6-15 16-30 >30P 54 45 9 2 D = Duration (min.) P = Percent Frequency

Leeward of mountain ranges, monsoonwinds are dried by descent and convectivelydestabilized, favoring a showers regime.Afternoon thunderstorms may then developand cause squalls. At Madras (13°N, 80°E)on the east coast of peninsular India, anaverage of one to two squalls occur eachyear, 74 percent during summer. The squallsare not severe, 65 to 70 percent of themranging between 27 and 41 knots. Gustsover 54 knots are extremely rare (SubbaReddy, 1985). Squally lee-side thunder-storms have also been observed over easternnorth Sumatra when westerlies prevail; ingeneral, they are rare in the dry subsidingair and may depend on moisture broughtinland by a sea breeze.

Carter and Schuknecht (1969) analyzedwind gusts occurring within 1 1/2 hours ofthunderstorms at Cape Kennedy (29°N,81°W) Florida for the period 1957 to 1967.The seasonal averages and the peak gustsare shown below (November is included inwinter).

Season Winter Spring Summer FallAverage Peak Gust (knots) 20 19 17 17Absolute Peak Gust (knots) 34 53 48 46

Thunderstorm gusts over 40 knots are rareat Cape Kennedy. The site seems to typifymany tropical and subtropical coasts wherethunderstorm gusts are generally less thanthose observed inland.

Atkinson (1966) used data for 1956 to 1966in studying extreme winds at 13 stations inThailand. Apart from tropical cyclonesaffecting stations in the south coast andpeninsular region, almost all extreme windswere caused by thunderstorms. Peak gustsranged from 40 to 73 knots. The strongestamong these generally were confined to themountains of north Thailand or adjacent tothe Gulf of Thailand. In the interior, thehighest gust in any year tends to occur inApril and May, while along the coast ofsouthern Thailand the peaks occur from Julythrough November. Extreme wind analysisof these data is presented below in Section 5.

Over West Africa, squall lines are mostintense and accompanying winds strongestat the end of a long dry season, when lapserates are most unstable. For the same rea-son, the second of two closely-spaced squalllines is less intense than the first, and squallsare more vigorous in the east than in thewest of the region. There does not appear tobe much diurnal variation. Squall lines arerare over southern Africa and may haveoccurred during the SH summer in Zambiaand Angola (Leroux, 1983). In the UnitedStates, major weather-related passenger

aircraft disasters near airfields led Fujita andCaracena (1977) to ascribe the cause tosevere downdrafts (microbursts) generatedby nearby thunderstorms. As a microburstdescends, the ground diverts it horizontally.When an aircraft is either taking off from, orin final approach to, an airfield, a burst of

wind in the direction of flight reduces theheadwind component and lift, and maycause stalling and a crash. Wilson et al.(1984), using Doppler radar measurements,proposed the model shown in Figure 10-17and recommended installation of Dopplerradars at airports to detect microbursts andwarn pilots. In the weaker thunderstorms ofthe tropics, severe microbursts are less likelythan in the United States. A thunderstormnear or over an airport is still dangerous toapproaching and departing aircraft.Caracena et al. (1990) explained microburstsand illustrated them with color photographsthat included time-lapse sequences.

3. Duststorms.

a. General. Any gust of wind raises dust inthe desert, and the conditions necessary fora duststorm are simple — a dry surface,covered to some depth with dust or finesand, and a wind strong enough to lift theseparticles into the air. Great duststormsusually develop in the cool season. Thenstrong winds in polar outbreaks may sweepacross the desert. Friction and convectionraise clouds of dust that fill the air beneaththe usual subsidence inversion, and oftenobscure the sun and reduce visibility to foglevel. Winter dust storms have been wellobserved and described in the Sahara.Figure 8-37 illustrates a duststorm-generat-ing cold front moving across North Africa.Intense mesoscale duststorms, or “ha-boobs”, summer phenomena, commonlyoccur around and to the south of Khartoum(16°N, 33°E).

b. Winter Duststorms. A vigorous cold frontmoving rapidly southeastward across NorthAfrica and southwest Asia heralds onset ofthe “harmattan” and possibly a severeduststorm. North of the subtropical ridge,increasing southwesterlies ahead of thefront first raise the dust. Scattered showersat the front may locally lay the dust, but in

general, strong gusty winds at the front andstrong northwesterlies in the cold air main-tain the duststorm. South of the subtropicalridge a surge often develops (Chapter 8,Section D-4). Freshening northeast windsthen begin to blow dust well ahead of wherethe front might reasonably have been ex-pected to be. Surface winds in the surgesoften exceed 30 knots (Kalu, 1979). Dustcrossing the Sahel and reaching the Gulf ofGuinea, may have been raised in the regionsouth of the Tibesti massif in northern Chad,and been carried in a plume toward south-west. Jalu and Dettwiller (1965) described anexceptionally severe duststorm that reached10°N on 27 March 1963 and raised thickdust over an area of 87,000 NM2 (300,000km2) in the central Sahara. Behind the rathersharply defined edge of the advancing dustcloud, surface temperatures were more than10°C lower than in the clear air ahead. Jaluand Dettwiller ascribed the difference tointerception of insolation by the thick dust.Their conclusion that the dust served toaccentuate the cold front along its leadingedge is not necessarily valid. If events hadlikely followed the sequence described inChapter 8, Section D-4, the front would havelong since disappeared as surface northeastwinds increased well ahead of the surfaceair-mass discontinuity. Then, the dust raisedby these winds would create its own accom-panying “front”. The large daytime tem-perature gradient across the leading edge ofthe dust cloud would have produced astrong low-level jet that would raise moredust. Probably then, once a threshold ofradiational opacity is crossed, local feedbackmay well intensify a duststorm. Subsidencefollowing the cold surge usually restrictsdust to the lowest three km.

About two or three times a month, coldfronts moving down the Persian Gulf arefollowed by strong north or northwestwinds — the “Shamal”. Raised dust mayreduce visibility for 24 to 36 hours. Once ortwice each winter, an intense Shamal accom-

panying a cold surge lasts three to five days(Walters and Sjoberg, 1988). Winds mayreach 50 knots and duststorms seriouslyreduce visibility. Figure 10-18 shows aShamal duststorm moving south acrossnortheastern Arabia and the Persian Gulf.Infrared imagery can clearly distinguish thedust from the warm Gulf waters, but this ismuch more difficult over the desert. InMarch, a low moving east across centralArabia is preceded by a hot, dry southerlywind — the “Aziab” — that raises dust oversoutheastern Arabia.

c. Summer Duststorms. Table 10-3 showsthat southward-moving winter duststormsaffect Khartoum, but that summerduststorms, which generally move in fromthe south, are most common. According toSutton (1925) they are convective phenom-ena; one in three is followed by a thunder-storm and two in three by rain. The stormfronts are usually between 10 and 15 NM(20 and 30 km) long and affect a locality forabout three hours. These “haboobs” aregenerally of mesoscale size, most oftenoccur between 16 and 18 LT and can seldombe traced between neighboring stations.They originate in complex

interactions between synoptic disturbances,probably surges (Figure 8- 39), and localorographic and heating gradients.

Rain accompanying the storm lays the dustand reduces the chances of a rapid sequenceof storms. Sutton thought that the frequencydecrease between June and August at

Khartoum might be partially due to sprout-ing of a dust-inhibiting ground cover. Simi-lar duststorms are common during summernear the borders of other deserts (Powelland Pedgley, 1969).

In summer the Persian Gulf lies north of theheat trough, and the Shamal prevails(Walters and Sjoberg, 1988). When it fresh-ens, it raises dust over the desert. Distur-bances similar to those responsible forhaboobs can lower visibilities to near zero,as winds exceed 30 knots.

At Agra (27°N, 78°E), 80 percent of allduststorms (“Andhi”) occur in May andJune (one day in four) prior to summermonsoon rain. The storms are also mesos-cale phenomena; 64 percent occur between15 and 20 LT and only 25 percent between14 and 11 LT. In spring they may be trig-gered by disturbances moving in from west(Sreenivasaiah and Sur, 1939).

Joseph et al., (1980) made case studies ofAndhi at Delhi (29°N, 77°E), which experi-ences on average one in April and three eachin May and June. Maximum gusts rangedfrom 26 to 54 knots and minimum visibilityfrom 360 to 2950 feet (110 to 900 m). Oftenno rain fell at Delhi; radar showed that thesquall line responsible for the dust could beas far as 15 NM (30 km) from the originatingcumulonimbus.

Over the deserts subsidence in winter limitsthe depth to which dust can penetrate togenerally below 10,000 feet (3 km). In sum-mer vigorous convection spreads dust

——————————————————————————————————————Table 10-3. Frequency of duststorms at Khartoum (16°N, 33°E) based oneight years of observations (from Sutton, 1925). Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total 5 7 11 13 21 48 37 23 25 5 0 1Direction of Approach*NE-SW 5 7 8 4 3 5 1 1 1 2 0 1SE-SW 4 12 26 16 11 15 2 0 0

*** EDITOR’S NOTE: USE WITH CAUTION; ERRORS IN THIS TABLE EXIST IN ORIGINAL MANUSCRIPT ***

* Direction of approach could not be determined for all storms.

upward to between 13,000 feet (3.9 km) and16,500 feet (5.1 km). The very fine dustabove 10,000 feet (3 km) does not dissipatebetween duststorms, but persists throughthe summer, seriously reducing visibilityand the range of electrooptical devices. Thesatellite image of 25 June 1985 (Figure 10-19)reveals widespread atmospheric dust withno obvious source region. The swirl of dustover southeastern Arabia has probably beencaused by a heat low acting for some timeon a reservoir of atmospheric dust.

High latitude meteorologists have troubledistinguishing snow from cloud in satelliteimages. Differentiating a dust cloud fromthe underlying surface may be equally hard.Infrared images don’t help much when thesurface temperature is close to the dustcloud temperature (recall a similar problemwith stratus or fog over cold upwelledwater). In winter, over the sea, which ismuch warmer than the air, dust can beeasily recognized in both IR and visibleimages. Over the relatively cold desert,however, the temperature difference, andconsequently the contrast in the IR image, ismuch less (see Figure 10-18). In summer, thesituation is reversed. Over the relatively coolsea, visible images show a dust cloud well,but the small temperature difference be-tween sea and dust cloud blurs the IR im-age. Over the hot land, IR picks up thetemperature contrast with the dust cloud,but as usual, the visible image seldomdifferentiates dust cloud and desert. Brooks(1990) suggested merging the IR image overland with the visible image over the sea.

Figure 10-20 shows IR (A) and visible (B)images of the Red Sea and its neighborhoodtaken on 25 August 1989. A duststorm overthe eastern Nubian Desert is clearly differ-entiated from clouds by IR, but only thevisible image shows the dust extending overthe Red Sea. Figure 10-20C combines (A)and (B) and clearly defines the limits of theduststorm.

d. Dust Devils. These are summer convec-tive phenomena, not associated with con-vective cloud. Hess and Spillane (1990)studied pilot reports of dust devils overAustralia and combined their findings withearlier results. Conditions favoring dustdevils were:

(1) Relatively level and fairly smooth ter-rain.

(2) Intense surface heating, with lapse ratesnear the surface much greater than dryadiabatic.

(3) Winds near the surface relatively lightand variable. Some shear in the vertical.

(4) Strong downdrafts producing horizontalwind shear and vorticity. Characteristicsranged widely. Height: up to 6500 feet (2km); diameter: less than a foot (0.3 m) tomore than 330 feet (100 m); duration: fewseconds to an hour; peak winds: up to 40knots. At one time, up to four dust devilscould exist over an area of 3.5 square nauti-cal miles (12 km2). Dust devils should besimilar over other tropical deserts.

In the Persian Gulf region (Walters andSjoberg, 1988), the strongest dust devilsoccur near the coast just inland of the edgeof the sea breeze. Here, environmentalsurface winds are light and horizontal windshear large, The sandy coastal zones ofSomalia, Yemen, and Djibouti often experi-ence dust devils (Vojtesak et al., 1990)

Forecasters should remember that majorwinter duststorms accompany strong coldoutbreaks when a branch of the Hadley cellaccelerates and a surge develops. Theduststorm edge may advance much fasterthan the following winds would indicate,and poor visibility and the LLJ endangerflying. In summer, the same weather condi-tions that produce severe thunderstorms ortornadoes may also cause duststorms, andsurges may sometimes be responsible.Besides the meteorological prerequisites, the

ground surface must be favorable — verydry, with ample dust or fine sand.

Duststorms and dust devils not only reducevisibility, but are also accompanied byturbulence that may endanger light aircraftor helicopters. Dust in the air warns ofturbulence, but no dust aloft may not meanno turbulence, only that the surface is dust-free.

4. Tropical Cyclone Winds. The strongestwinds at the earth’s surface (excludingtornado winds) are caused by intense tropi-cal cyclones. Hurricane Camille, which hitthe Gulf Coast of the United States in Au-gust 1969, had a peak recorded gust of 175knots. Storms are rarely this intense in theNorth Atlantic; however, “super typhoons”with peak winds of 130 to 200 knots arewell-known to residents of coasts and is-lands in the western North Pacific. Exceptfor the specially-designed Dines pressure-tube anemograph at the Royal Observatory,Hong Kong, anemometers often break or areblown away when wind gusts exceed 100knots. Maximum winds can be roughlyestimated from the resulting storm damage.A better method uses relationships betweenthe minimum sea-level pressure in thestorms’ eyes (obtained from surface ordropsonde measurements) and the strongestwinds (maximum sustained one-minuteaverages) in the tropical cyclones. Fromanemograms at coastal and island stationsin the western Pacific, Atkinson andHolliday (1977) selected 76 cases when itwas very likely maximum winds wereexperienced during cyclone passage. Theydetermined that

Vm = 6.7 (1010 - Pe)0.644

where Vm is the maximum sustained surfacewind in knots and Pe is the lowest surfacesea-level pressure (mb) in the storm center(Figure 10-21). The lowest pressure everrecorded -870 mb in super typhoon Tip on

12 October 1979, corresponds to a maximumsustained surface wind of 162 knots. Eyepressures are measured on reconnaissanceflights. As there now much fewer flights,this technique can seldom be used.

Coastal and island stations can suffer thefull fury of the maximum winds in tropicalcyclones; however, they quickly weaken asthey move inland — surface friction in-creases and the oceanic source of heat andbuoyancy is cut off. Topography plays alarge part in locally determining how strongthe winds will be. Taniguchi (1967) showedthat typhoons passing over northern Luzonin the Philippines seldom cause damagingwinds at Clark Air Base (15°N, 121°E) be-cause of the protection by surroundingmountains. Since 1945, the highest recordedgust at Clark was 83 knots, while gustsexceeding 50 knots were rarely observed. InNovember 1957, the center of typhoon Kit(which on the previous day had a centralpressure of 915 mb) passed within about 24NM (45 km). Surface winds at Clark neverexceeded 10 knots, although there wasextensive damage on the periphery of thebase. Similarly, stations on the west coast ofTaiwan experience much lighter winds fromtropical cyclones than stations on the eastcoast.

Strong winds damage structures in twoways — through static effects, in which theforce exerted is proportional to the square ofthe wind speed, and dynamic effects, inwhich unstable oscillations may be set up bythe wind. Old buildings with thick walls aresusceptible to static effects, and forecastersshould remember that the static force of a100 knot gust is double that of a 70 knotgust. Tall, flexible steel-framed buildingsrespond to dynamic effects. These can bereduced by structural damping, but sincethis is hard to extrapolate from models, aninstrumented eight-story steel-framedbuilding on an exposed site in Hong Kong is

being used to directly monitor the effects oftyphoon winds (Mackey, 1969).

The National Weather Service uses theSaffir/Simpson Damage Potential Scale(Table 10-4) to alert public safety officialswhen a hurricane is within 72 hours of landfall (Simpson and Riehl, 1981).

5. Extreme Wind Analysis. Extreme-valuestatistical distributions are often used toestimate the probability of various meteoro-logical extremes, e.g., wind speeds, tem-peratures, rainfall (see Chapter 6, Section C-12), etc. These distributions are usuallyapplied to series of annual extremes (i.e., thehighest value of each year over a period ofyears) to determine realistic design valuesfor structures -dams, air conditioning orheating needs, etc. Because of its simplicityand wide applicability, the Gumbel double-exponential distribution (1958) is the mostfrequently used model. (It has already beendescribed and applied in Chapter 6, SectionC-12.) Figure 10-22 shows double-exponen-tial distribution fitted to the annual peakgusts at Kadena Air Base (26°N, 128°E),Okinawa, for the period 1945 to 1988. Onceevery 100 years on average, Kadena canexpect a peak gust of 136 knots, while thetwo-year return period or median value is76 knots. In Figure 10-23 peak gust valuesfor 2-year and 100-year return periods areshown for some tropical stations, based ondata from a variety of sources. Values forother return periods can be determined byplotting the 2-year and 100-year return

period values on Figure 10-24, drawing astraight line connecting these points, andinterpolating the value for the desiredreturn period.

Stations with the greatest extremes oftenexperience tropical cyclones; however eventhere, station location and topography affectthe results. For example, the 100-year returnperiod wind at Guantanamo Bay (20°N,75°W) is only 71 knots compared to 124knots at San Juan (18°N, 66°W).Guantanamo is sheltered by nearby moun-tains and by the island of Hispaniola to theeast, whereas San Juan is exposed to the fullforce of Atlantic hurricanes. Coastal stationsin Australia and the southeastern UnitedStates have about the same expected ex-tremes, but the extremes in the westernNorth Pacific are significantly greater. Ex-tremes are much less at inland stationswhere thunderstorms are chiefly respon-sible; values range from 35 to 55 knots forthe expected 2-year extremes and from 60 to90 knots for the 100-year extremes.

Okulaja (1968) analyzed extreme winds atLagos (6°N, 3°E), Nigeria, where thunder-storms are the primary cause. Based on a 14-year record, the 2-year and 100-year ex-pected extreme gusts for Lagos are 44 and67 knots respectively.

FFFFF. T. T. T. T. Turbulence.urbulence.urbulence.urbulence.urbulence.

1. General. Turbulence affecting aircraftoperations remains a major problem in

Table 10-4. Saffir/Simpson Damage Potential Scale ranges.Scale Central Pressure Winds Surf Damage Number (mb) (knots) (feet)1 980 or more 64-82 4-5 Minimal2 965-979 83-96 6-8 Moderate3 945-964 97-113 9-12 Extensive4 920-944 114-135 13-18 Extreme5 less than 920 > 135 > 18 Catastrophic————————————————————————————————

higher latitudes. Saxton (1966) gives therelative frequency of significant turbulenceat various altitudes (Figure 10-25) for mid-latitude land areas.

Mechanical turbulence, which is a functionof surface wind speed and terrain rough-ness, dominates the lowest few kilometersof the atmosphere. Convective turbulence,generally associated with cumuliformclouds, dominates the middle troposphereand extends as high as 18 km around severethunderstorms. Clear air turbulence (CAT)dominates the upper troposphere andstratosphere; although it is infrequent be-tween 15 and 25 km, no upper limit isknown. These general definitions apply aswell to the tropics, even though turbulenceis much rarer there.

In the tropics convective turbulence overland is likely in association with thunder-storms and is more severe with more in-tense thunderstorms. It also depends on thestage of the thunderstorm, being mostsevere during the formative and maturestages. In 1964-65, the India MeteorologicalDepartment (1968) had pilots observe turbu-lence in or near thunderstorms or toweringcumulus. Of the 182 reports, 55 percent wereof light turbulence, 44 percent of moderateturbulence and 1 percent of severe turbu-lence. Tropical thunderstorms pose littlethreat to aircraft equipped with weatherradar or guided by ground-based radar,since the more intense cells can usually becircumnavigated. Thunderstorms are mosthazardous during landing and take off,when microbursts may destroy lift andcause an aircraft to stall (see discussion inSection E-2). At jet aircraft flight levels,tropical cyclones pose few problems. In thewestern Pacific, commercial airlines often flyover the eyes, and rarely experience morethan moderate turbulence.

Because of generally light winds in thetropics, low-level mechanical turbulence isless common there than in higher latitudes,

although it occurs leeward of mountainranges when surface winds reach about 30knots. Daytime insolation of land surfaces inthe tropics characteristically causes “bumpi-ness” up to 10,000 to 20,000 feet (3 to 6 km),especially when a LLJ is present.

2. Clear Air Turbulence (CAT). In associa-tion with mountain waves (caused by strongflow perpendicular to mountain ranges) orwith strong wind shear in the vertical nearjet streams, CAT is rarer in the tropics thanin middle latitudes, perhaps because favor-able synoptic situations are also rarer. Asurvey paper by Saxton (1966) that includesfrequency distribution with altitude (Figure10-25), is the basis for the following informa-tion. CAT is normally found in thin horizon-tal sheets, about 2000 feet (600 m) thick, 10to 20 NM (20 to 40 km) wide and 50 to 100NM (100 to 200 km) long, oriented along thewind flow. Its dimensions suggest that CATis related to the mesoscale structure of theatmosphere and is mainly caused by break-ing (unstable) gravity waves. In the freeatmosphere, gravity waves often occur instable layers with large wind-shear in thevertical. Occasionally, in the presence ofsufficient moisture, cloud patterns revealgravity waves 3000 feet (1000 m) or more inlength. However, these long waves are notparticularly turbulent for subsonic aircraft.They are more properly classified as“undulance” and are characterized bylaminar flow. Under certain conditions ofdensity stratification and wind shear, wavesshorter than some critical value amplifyexponentially and break down the laminarflow into turbulent eddies. Likewise, forgiven wave length and density stratification,increasing wind shear in the vertical maymake the flow unstable. In most of theempirical studies of CAT, strong verticalwind shear is most consistently correlatedwith CAT.

Aircraft observations of CAT comprise threecategories: pilot reports; those obtainedusing the aircraft as a sensor; and thoseobtained from a sensor mounted on anaircraft. These data have provided consider-able insight into the relative climatologicalfrequencies of CAT intensity over manyregions and altitudes. Many investigatorshave tried to find numerical values of me-teorological elements or combinations ofelements which correlate well enough withCAT to be used in forecasting occurrence ofCAT and how strong it will be. However,progress has been painfully slow. The rea-sons for this are, according to Saxton: First,most CAT reports are subjective, and onlypositive encounters are usually reported.Second, meteorological data available tooperational forecasters lack the detail, hori-zontally, vertically and temporally for com-puting the desired parameters in the mesos-cale where turbulence occurs. Third, evenwhere CAT and observed meteorologicalelements are significantly related, inabilityto forecast these elements restricts applica-tion of the relationships.

Clear air turbulence forecasts combineobjective and subjective techniques. Fore-casts will become more automated as nu-merical models improve. Saxton’s concern

that the CAT problem can never be ad-dressed through forecasting because of thelimitations imposed by the spacing of up-per-air stations and the long intervals be-tween soundings may be partly removed ifwind and temperature profilers becomecommon (see Chapter 3, Section B-3b).

3. Frequency of Clear Air Turbulence. TheInternational Civil Aviation Organization(ICAO) sponsored a world-wide data collec-tion program to determine the relativefrequency of CAT. During the four 5-daycollection periods, one for each season

in 1964-65, pilots were asked to make turbu-lence reports on all flights at and above20,000 feet (6 km). They used the followingscale: Very Light CAT — Perceptible; LightCAT — Slight discomfort; Moderate CAT —Difficulty in walking; Severe CAT — Looseobjects dislodged; and Extreme CAT -Aircraft violently tossed about. Turbulencein cirrus or cirrostratus, not connected toconvection, was also classified as CAT.These data, collected and analyzed by vari-ous meteorological centers, are summarizedfor the tropics in Table 10-5.

————————————————————————————————Table 10-5. Percent frequency of various categories of CAT by five-degree latitude-longitude rectangles in the tropics, for flights above 20,000 feet (6 km) (for the 1964-65ICAO collection periods).

Area Turbulence Category None L M SCaribbean 86.2 9.4 4.4 0.0Mexico and Central America 92.6 4.9 2.5 0.0South America and South Atlantic 92.1 5.8 1.9 0.0North Pacific 0° - 15°N 94.2 5.3 0.5 0.0 15° - 30°N 89.8 6.2 3.9 0.1Africa South of 15°N 94.8 3.7 1.4 0.1Australia, South Pacific 92.5 5.3 2.0 0.2Middle East and Southeast Asia 15°S - 0° 96.0 2.7 1.3 0.0 0° - 15°N 93.1 3.4 3.1 0.4 15°N - 30°N 92.3 5.6 2.1 0.0

CAT is not a serious problem in the tropics.Extreme turbulence was not reported andsevere turbulence was very rare. These casesand many of the moderate cases occurrednear cool season upper-tropospherictroughs extending from mid-latitudes intothe tropics. Mountain-induced waves incentral Africa were responsible for somemoderate turbulence.

A later study of CAT on air routes to andfrom India for 1966 through 1969 (Gupta etal., 1972) found moderate and sometimessevere CAT above 25°N over India associ-ated with the subtropical jet stream andtroughs in it. Only light or moderate CAToccurred south of 25°N. Severe CAT wasconfined to winter. As already mentioned,CAT was most often associated with largewind shear in the vertical. Severe turbulenceoccurred only with shears of greater than 4knots 1000feet-1 (13.5 knots km-1), with thewind increasing with height.

Between July and October 1963 a specially-instrumented jet aircraft was used in investi-gating CAT in the vicinity of the wintersubtropical jet stream over Australia(Project TOPCAT) (Spillane, 1968). Near thejet stream core (approximately 200 mb) allintensities of CAT were five times morecommon than for all flights over Australianear 20,000 feet (6 km) altitude.

Summarizing, severe CAT is extremely rarein the tropics, and so is almost impossible toforecast accurately. In satellite images, cloudbands transverse to the flow, generallyindicate moderate to severe turbulence.During the cool season, near the sharpupper-tropospheric troughs that extenddeep into the tropics, light and occasionallymoderate CAT may be experienced. It mayalso occur in strongly shearing layers nearthe tropical tropopause (about 52,000 feet(16 km)). For example, during the LineIslands Experiment in March and April1967, radiosonde observations between 6°Nand 1°N revealed extremely strong wind

shears exceeding 12 knots 1000feet-1 near the16 km level (Madden and Zipser, 1970) whothought it likely that they would be accom-panied by considerable turbulence.

GGGGG. Icing. Icing. Icing. Icing. Icing.....

Aircraft icing is of minor concern in thetropics. Few aircraft fly at the altitudes ofgreatest icing potential, visual or radaravoidance of cumulus congestus or cumu-lonimbus is practiced, and aircraft equip-ment reduces or eliminates certain icinghazards. The physical factors affectingaircraft icing are the same in the tropics as inhigher latitudes.

Icing is possible if the aircraft surface iscolder than 0°C and if supercooled waterdroplets are present. Even then, the amountand rate of icing depends on the relation-ships among meteorological and aerody-namic factors: temperature, liquid-watercontent of the air, droplet size, collectionefficiency, and aerodynamic heating. In thetropics, the freezing level lies at about 15,000feet (4.5 km), with little synoptic, annual ordiurnal variation.

In stratiform clouds, icing is most likelybetween about 17,000 and 22,000 feet (5 and7 km). In the tropics, middle- or upper-levelcyclones or dissipating convective systemsmay produce widespread altostratus ornimbostratus at these levels. Higher up,icing is much less likely. In cumuliformclouds, the zone of probable icing is smallerhorizontally but larger vertically than instratiform clouds. The cloud’s stage ofdevelopment controls the icing risk. Icingintensities may range from a trace in smallsupercooled altocumulus to moderate tosevere in cumulus congestus or cumulonim-bus. In cumuliform clouds, icing is usuallyclear or mixed. Aircraft icing is most likelywith temperatures between 0°C and -20°C;however, it has been reported at tempera-

tures below -40°C in the upper parts ofthunderstorms.

HHHHH. Ocean W. Ocean W. Ocean W. Ocean W. Ocean Waves.aves.aves.aves.aves.

1. Storm Surges. A storm surge is an abnor-mal rise of the sea along a shore, and resultsprimarily from the winds of a storm. It ismeasured as the difference between theactual elevation of the sea surface and theelevation expected in the absence of a storm(i.e., the predicted astronomical tide).

Along low coasts, the storm surge is usuallythe chief destructive agent of tropical cy-clones; with intense cyclones, the sea maypenetrate 10 to 20 NM (20 to 40 km) inland.Since the surge generated at the coast mustmove inland as a gravity wave, the peak riseat the end of a long channel may lag thepeak storm winds by many hours. Interac-tion with the tide may also shift the time ofseverest flooding several hours away fromthe time of greatest storm intensity. Besidesthe winds and the tides, the extent of coastalflooding depends also on the pressuredeficiency in the storm compared with thestorm’s environment, the size and travelspeed of the cyclone, the bottom topographynear the cyclone’s land fall and other factors.Most important in determining the maxi-mum storm-surge height is the maximumwind speed in the cyclone as it heads inland.Since this in turn is related to the cyclone’sminimum sea-level pressure, Conner et al.(1957) related the maximum storm-surgeheight H on the open coast to the minimumsea-level pressure Po [in mb] for 30 hurri-canes entering the United States from theGulf of Mexico prior to 1956. In his linearregression equation: H [in feet] = 0.154(1019 - Po). The resulting correlation coeffi-cient was 0.68. The equation worked wellwith ten Atlantic hurricanes that movedinland. When hurricanes skirted the coast,

the equation overestimated the maximumsurge height.

Hoover (1957) made a similar study ofAtlantic and Gulf Coast hurricanes. Hisresults are shown in Figure 10-26 for 18Atlantic coast hurricanes (r = 0.86) and 24Gulf of Mexico hurricanes (r = 0.81).Conner’s regression line for Gulf Coasthurricanes is also shown in Figure 10-26.According to Hoover, his regressions tend tounderestimate the probable maximumstorm-surge heights since only observedstorm surge data were used in their deriva-tion. Four typical profiles are shown inFigure 10-27. In each case, the cyclone trackwas almost perpendicular to the coast. Thestorm surge was highest just to the right ofthe cyclone’s center (near the location ofmaximum winds). From here, the surgeheight fell off rapidly with distance.Hoover’s regression equation can be used tocompute the maximum storm-surge heightaccompanying hurricane Camille in August1969. Just before land fall, the hurricane’scentral pressure was 905 mb, which gives apredicted peak storm-surge of 19.2 feet (5.85m). The observed peak was 26.4 feet (7.5 m).Hoover had no cyclones of Camille’s inten-sity in his sample.

Jelesnianski (1966, 1967) developed numeri-cal models to compute storm-surge heightsassociated with tropical cyclones. Besidesthe elements identified by earlier investiga-tors, he included latitude, wind stress andthe depth contours of the basin. His “stan-dard cyclone” was centered at 30°N and hada maximum wind of 87 knots. His standardbasin had a linear slope typical of the conti-nental shelf. He computed storm-surgeheight profiles for various coast-crossingangles and radii of maximum winds. Adetailed description of these models or ofthe many assumptions used in their deriva-tion is beyond the scope of this report. Sugg(1969) based a mean storm-surge profile ondata from 19 great Atlantic hurricanes

(central pressure of 950 mb or less) makingland fall in the United States south of 35°Nfrom 1900 to 1969. In Figure 10-28 are plot-ted 84 maximum storm-surge height mea-surements made at various distances fromhurricane centers as they crossed the coast.The dashed curves enclose the estimatedrange of storm-surge heights. Data pointsfalling outside these curves reflect localinfluences of bays, rivers,

inland waterways, etc. The highest storm-surge, of 26.4 feet (7.5 m), near Pass Chris-tian (30°N, 89°W), Mississippi, was causedby Hurricane Camille in August 1969. Thecomputed storm surge profile from one ofJelesnianski’s (1974) models is included inFigure 10-28 for comparison. There is prettygood agreement, although the maximumheights with respect to the storm centerdiffer somewhat.

Some extreme events defy explanation. At2215 LT on 21 October 1972, Hurricane Bebedrove a solitary wave across the coral atollof Funafuti (9°S, 179°E) to a depth of 13 feet(4 m), and created a new island along theeastern shore, 20 NM (37 km) long andrising 11.5 feet (3.5 m) above the sea. Coralrubble that had previously accumulated upto 40 feet (12 m) beneath the sea surface onthe open ocean side of the atoll had beentorn out and piled up by the wave. The totalarea of the atoll was increased by one-third(Maragos et al., 1973). Although Bebepassed across other islands in the Ellicechain before hitting Viti Levu, Fiji, the expe-rience of Funafuti was not repeated. Sinceneither orographic nor bathymetric featurescould account for the event, one is left witha coincidence -the eye of Bebe was movingin a tight loop near the island when thewave struck. It is obvious that the success ofany storm-surge model depends criticallyon accurate forecasts of storm movement,intensity and wind profile.

2. Tsunamis. The tsunami is not a “tidalwave” but is an ocean wave produced by asubmarine earthquake, landslide or volcaniceruption. These waves may have enoughenergy to cross and sometimes recross entireoceans as gravity waves. Because of theirlong wave length, tsunamis cannot be de-tected from a ship in mid-ocean, but theirenormous dimensions become obvious atcoasts. Tsunamis steepen and increase inheight as they begin to move over shallowwater, inundating low-lying areas; wherelocal submarine topography is extremelysteep, they may break and cause devasta-tion. Meteorologists may become involvedin receiving and disseminating tsunamiwarnings since the international meteoro-logical telecommunication system has beengiven this task. The following description oftsunamis and the tsunami warning system issummarized from Argulis (1970).

The most popular theory holds that verticaldisplacement of water by underwater earth-quakes generates tsunamis; however, only asmall fraction of ocean-bottom earthquakescause tsunamis. Earthquakes of magnitude6.5 or more are usually necessary before atsunami is likely. Other factors affectingtsunami generation and magnitude arewater depth, earthquake depth, and effi-ciency of energy transfer. Tsunamis propa-gate radially from their source with wavelengths up to 400 NM (750 km) and waveperiods of several minutes to an hour. Theirspeed S through water is given approxi-mately by:

S = (gh)0.5

where g is the acceleration of gravity (32 feets-2 or 9.8 m s-2) and h is the depth of waterthrough which the wave is travelling. Forexample, in water 18,000 feet (5500 m) deepthe tsunami travels at about 450 knots; in900 feet (274 m) the speed is 100 knots; andin 60 feet (18 m) it is 25 knots. This physicalrelationship enables arrival time to be pre-dicted if ocean depth and time of the tsu-

nami-generating shock are known. As thetsunami begins to cross the rising bathym-etry near a coast, friction becomes an activeforce, causing the wave to slow and itsheight to increase.

The variability of tsunami run up at differ-ent locations can be extreme. The 1960tsunami, generated by an earthquake offChile, caused a 35 feet (10.7 m) run up atHilo (20°N, 155°W) and only 9 feet (2.7 m) atPapaikou seven kilometers away. Run up iscontrolled by change in coastal slope, anglebetween the direction the wave is propagat-ing and the coastline, coastal shape, andunderwater terrain features. Also, althougha shoreline may not be in the direct path ofan advancing tsunami, bathymetry cancause a deflection. Thus, all major tsunamisshould be considered potentially dangerous,although they may not uniformly damageevery coast they strike. Potential dangerzones are less than 50 to 100 feet (15 to 30 m)above sea-level and within a mile (one totwo km) of the shore.

The National Ocean Survey, NOAA, whichis responsible for issuing tsunami warnings,operates a Seismic Sea Wave Warning Sys-tem (SSWWS) that employs a world-widenetwork of seismological stations to detectand locate earthquakes, a network of tide-gauges to record passing tsunami waves,and a communication system to speedtransmission of information and warningsto various countries.

Chapter 11Chapter 11Chapter 11Chapter 11Chapter 11TROPICAL ANALTROPICAL ANALTROPICAL ANALTROPICAL ANALTROPICAL ANALYSISYSISYSISYSISYSIS

A. General.A. General.A. General.A. General.A. General.

Godske et al. (1957) state the aim ofweather map analysis to be:

“After careful consideration of their repre-sentativeness and reliability, all availablemeteorological data must be fitted into themost probable system of ideal and modifiedthree dimensional tropospheric models”[italics added].

A skilled analyst’s reasoning can seldom bequantified, and so his knowledge is hard totransfer. Analysis, carried out as an integralpart of forecasting and operational research,is important, even though immeasurable.During analysis, the meteorologist mentallystores impressions of data that consciouslyor subconsciously he taps when preparingthe forecast. Equally significant and subtleimpressions cannot stem from contemplat-ing machine-analyzed processed data.

Post-analysis may lead to understanding ofpast errors and decrease the chance ofsimilar errors in the future. No better wayhas yet been devised than analysis-forecast-ing, closely interacting with operationalresearch, for rapidly identifying problems,developing, testing and modifying hypoth-eses and compiling useful statistics. Ideally,the analyst-forecaster and the operationalresearcher should be one.

Analysis techniques for the mid-latitudesand the tropics differ greatly, largely be-cause in the tropics simplified pressure/wind relationships such as the geostrophicwind are of questionable validity. In mid-latitudes, pressure (height) analyses com-bined with the Norwegian air mass/frontal

models can describe most synoptic-scaleweather disturbances in the lower tropo-sphere. In the tropics, except in cyclones,pressure (height) and weather are not toowell linked. In mid-latitudes, most cyclonesdevelop and intensify in response tobaroclinic instability associated with largehorizontal temperature gradients. In thetropics, mid-tropospheric cyclones developin a baroclinic environment (wind shearingin the vertical) while tropical cyclonesdevelop in a barotropic environment (windshearing in the horizontal). Modelers oftendisagree on which is more important todevelopment (Lighthill and Pearce, 1981). Inthe tropics, most synoptic-scale systems areweak and so diurnal variations, local topo-graphic effects and cumulus convection playlarger roles in determining weather than inmid-latitudes, and assume greater impor-tance in daily analysis and short-rangeforecasting. For this reason, sub-synopticanalysis is essential in the tropics.

In this chapter, tropical analysis techniquesare reviewed and evaluated. Very little ofthis material is new or original: subjectiveanalysis techniques have not changed muchsince the classic texts of Palmer et al., (1955),Riehl (1954), and Saucier (1955). Improve-ments can generally be attributed to moreconventional data (especially upper-airobservations), more and better aircraftreports, and cloud motion vectors derivedfrom sequences of geostationary satellitepictures. In addition, cloud patterns re-vealed by satellites have been identifiedwith tropical synoptic models (Chapters 8and 9), and so can greatly increase analyticalprecision in regions lacking any other data.As a result,

meteorologists have come to rely on satellitedata much more and should become expertsin their interpretation and application.

Four major topics are covered. First, datacollection and evaluation show how tointerpret the various data available to tropi-cal analysis. Second, manual analysis tech-niques, especially those using streamlinesand isotachs (kinematics) are summarized.Third, automated analysis techniques andproducts and the outlook in this field aresurveyed. Fourth, auxiliary aids to tropicalanalysis are described. The role of weather-radar data in tropical analysis and forecast-ing is covered in Chapter 12.

B. Data Collection and Evaluation.B. Data Collection and Evaluation.B. Data Collection and Evaluation.B. Data Collection and Evaluation.B. Data Collection and Evaluation.

1. General. As shown in Chapter 3, awide variety of data is available for tropicalanalysis. Nevertheless, data collection andevaluation are more difficult in the tropicsthan in higher latitudes, due to fewer sta-tions, poorer communications, and localeffects on the observations. Tropical weatherunits must have an aggressive data-collec-tion and quality-control program aimed atacquiring all observations and making themavailable to the analysts. This demandsintimate knowledge of station networksincluding the type and frequency of obser-vations for each station, the communicationschedules for weather-data transmissions,meteorological codes, and special datasources, such as aircraft reconnaissanceobservations and drifting data buoys.

Information on networks, observational andbroadcast schedules and codes is containedin Air Weather Service pamphlets andregulations, Federal Meteorological Hand-books (prepared by the US Departments ofCommerce and Defense), and in variousWMO publications (World MeteorologicalOrganization, 1986). In addition, AWS unitsmay wish to use automated data surveys

made periodically by the Air Force GlobalWeather Central (AFGWC) and other cen-tralized units to compare their local data-receipt with those of the Centrals. In thisway, deficiencies in the station’s programcan be revealed.

Paradoxically, in recent years some localweather units have received more central-ized products than they could effectivelyuse. Thus, they must be aware not only ofall data sources, but must also select onlythe products that contribute to their dailyanalysis and forecasting. Ideally, at eachstation, the station chief monitors routineconventional observations and ensures thatall available data required are collected andplotted; the station chief surveys and evalu-ates centralized products and determinestheir usefulness.

Tropical meteorologists, even more thantheir mid-latitude counterparts, must knowhow the observations are made, their accu-racy, and their representativeness in order toanalyze distributions and to integrate vari-ous types of data into one composite analy-sis (e.g., upper wind data from radar andvisually-tracked balloons, aircraft reports,and satellite cloud motion vectors).

As an integral part of synoptic analysis, dataare evaluated as the analysis proceeds. Thisis best achieved by comparing observationsto neighboring observations for the sametime, to observations above and below theanalysis level, and to earlier observations atthe same station. In other words, the analystevaluates data horizontally, vertically, andchronologically. Observations can bechecked against climatology to identifygross errors; this is especially useful forisolated stations. Synoptic analysis con-structs, from spot observations, a graphicalpicture of the entire field of a meteorologicalvariable over a given region for a givenmoment in time. Since analysis extrapolatesoutward from points of observation, synop-tic reports should ideally be error-free and

be representative on the analysis scale, thatis, data used in a synoptic-scale analysisshould be free from purely local effects.However, in the tropics, local effects oftenovershadow synoptic changes, especiallynear the surface. Ignoring radiational heat-ing and cooling, topography, local winds,and convection at or near the station atobservation time can lead to serious errorsin data interpretation and analysis. On theother hand, local effects reflected by the datamay be entirely appropriate for a mesoscaleor microscale analysis. Tropical forecastersmust be able to separate local from synopticinfluences. A useful rule of thumb holds thatlocal effects are least around local noon andmidnight — transition times between day-time and nighttime circulations. The analystfirst assumes that observations are correct,even when they apparently contradictpreconceived ideas. They are rejected onlyon the basis of careful, physical reasoning.

2. Surface Observations.

a. Pressure. Various factors cause surfacepressure reports to be unrepresentative.Mountain chains produce large wave-likedistortions in the surrounding surfacepressure patterns. For example, the moun-tains of the Philippines and Central Americacause lee-side troughs in strong tradewinds(see Figure 8-36) and the Western Ghats ofIndia act the same way on the southwestmonsoon. Much smaller topographic fea-tures also produce effects comparable to thepressure changes accompanying weatherdisturbances. On the island of Hawaii,pressure on the windward side can be asmuch as 2 mb higher and that on the lee-ward side as much as 4 mb lower thanpressure would be without the island. Aventuri effect in channels between largeislands can reduce the pressure by 1 mb ormore. Heavy thunderstorms can cause brieflocal pressure rises of about 0.5 mb.

Instrument errors can make surface pressuremeasurements unrepresentative. In manyregions, barometers are not calibrated. Thisproblem is magnified in the tropics. First,observations are scattered with stationsusually hundreds of kilometers apart, sopressure reports can seldom be checkedagainst nearby measurements. Second,pressure gradients are very small, especiallynear the equator, so errors of 1 or 2 mb canlead to significant errors in locating pressurecenters. Third, passing synoptic distur-bances may cause the pressure to vary lessthan with the semi-diurnal pressure wave.Ship pressures are often wrong. The Na-tional Meteorological Center (NMC) (USWeather Bureau, 1963) assumes a standarddeviation of the errors of 1 to 2 mb formerchant-ship pressures and 1 mb forweather ship pressures. This is large enoughto handicap tropical pressure analysis.

Over tropical continents, reduction of pres-sure to sea-level may introduce anothererror. NMC assumes that the sea-levelpressure will have a standard deviation oferror of 1.6 mb for each kilometer of stationelevation. This error, when combined withthe other errors, renders sea-level pressuresfrom high-level tropical stations virtuallyworthless for synoptic analysis. Sea-levelpressure analyses are of little use especiallyequatorward of 20°. This will be discussedfurther in Section C.

Local pressure changes are undoubtedlymore accurate than the absolute values ofpressure. In mid-latitudes, three-hourlypressure tendencies can indicate motion anddevelopment of synoptic systems. In thetropics, the large diurnal pressure variation(comparable to the small changes stemmingfrom synoptic influences) renders three-hourly changes useless. However, for astation on a small island, the average oce-anic diurnal pressure variation (Figure 7-4)can be subtracted from the observations,and so allow most of the synoptic signal to

be revealed. Because of local heating andcooling cycles, individual average diurnalpressure variation curves need to be deter-mined for each land station, but it is worthdoing. At Hong Kong (22°N, 114°E), forexample, the pressure, measured everyhour, is “adjusted” in this way. It could befurther adjusted to take account of thedifferent variations associated with clearand cloudy skies (see Figure 7-5). Of coursethe adjusted values are not included insynoptic reports. 24-hour pressure tenden-cies eliminate the diurnal variation, al-though the coarse time resolution hides theeffects of small-scale systems. Their use isdiscussed in Section E.

b. Temperature and Dew point. In additionto the normal instrumentation and exposureerrors, unrepresentative temperatures in thetropics are often caused by local convection.This is due primarily to evaporative coolingby rain of the air beneath clouds and to thedownrush of cold air from showers andthunderstorms. These downdrafts can lowerthe surface temperature well below therepresentative surface wet-bulb tempera-ture. For example, a record low temperatureof 21°C at Canton Island (2°S, 172°W) oc-curred in heavy rain (Figure 11-1). Colddowndrafts may well have caused recordcold at many other tropical stations. Besidesrapid local temperature fluctuations, con-vection and land-sea breezes cause largehorizontal temperature gradients at thesurface. Over homogeneous surfaces inhigher latitudes such gradients are oftenassociated with fronts but in the tropics theydo not separate different air masses. Overtropical continents, more particularly insummer, diurnal temperature variationsdominate and are generally much largerthan temperature changes due to synopticinfluences.

The representativeness of the surface dewpoint depends on the same local influencesthat affect temperature, although the dew

point is not so responsive (see Figure 7-3). Inhumid regions, dew point varies diurnallyin phase with the temperature, with anafternoon maximum and a morning mini-mum. In arid regions, and to the lee ofmountains, the cycle is reversed. Diurnaldew point variations may often overshadowsynoptically-induced dew point changes.Over tropical continents in winter, dewpoint changes as a front passes. Widespreadsubsidence and suppressed convection inthe tradewinds may be reflected in below-normal dew points (see Chapter 5, SectionC-5). Over some tropical continents, such asparts of Africa, dew points, in conjunctionwith low-level kinematic analyses, can beused to delineate boundaries between moistand dry air masses.

24-hour temperature changes are not wellcorrelated with synoptic influences in thetropics. Sometimes 24-hour dew pointchanges are useful, since cool-season frontspenetrating the tropics may lower the dewpoint more than the temperature, especiallyif the front is accompanied by little cloud.24-hour dew point change charts are usefulduring these periods. Overall, dew pointsare more representative and conservativethan temperatures.

c. Wind. If measured properly, surfacewinds are representative of synoptic influ-ences over the open ocean. But even tiny,flat, isolated islands slow the wind. At WillisIsland (16°N, 150°S) (Figure 7-2), windmeasured at 12 feet (3.7 m) elevation on thereef averaged the same as wind measured at33 feet (10 m) on the island. A typical verti-cal profile would require the latter to be 20percent stronger. Over tropical continentsand mountainous islands, winds do notusually reflect synoptic influences. Pro-nounced land-sea breeze regimes affectcoasts and mountain-valley winds theinterior. At inland stations on flat terrain,wind directions may reflect synoptic influ-ences around midday, when low-level

mixing is taking place and when any diur-nal cycle is in transition. Differences ininstrumentation, surrounding vegetation,and anemometer heights and exposure maycause spurious differences in wind veloci-ties between neighboring stations. As withtemperatures and dew points, surface windsare affected by nearby convection. In re-gions not affected by tropical cyclones, thestrongest winds are usually caused bydownrush from heavy showers or thunder-storms.

The representativeness of surface windreports depends on the observing procedureand the limitations of various meteorologi-cal codes. In surface synoptic reports, windspeed represents the sustained speed overone minute for US stations and over tenminutes for other countries. Maximum gustslasting for less than one minute are notreported. In aviation weather reports, thewind observation coded in METAR repre-sents the mean velocity over the ten minutespreceding the observation. “Maximum windspeed” is included in the report, if duringthis ten minutes it exceeds the mean speedby at least 10 knots. The averaging periodmay vary among stations. For example,stations with anemographs may give ten-minute averages while reports from otherstations may be based on shorter-periodaverages. These and other observationaldifferences require that surface wind obser-vations for all land stations be carefullyevaluated. Only those reports that appear to

reflect synoptic influences should be used inpressure and kinematic analyses.

Surface wind reports from ships are prob-ably the most representative and useful forsynoptic analysis. Verploegh (1967) hasestimated the standard error of ship wind-speed observations to be 0.6I (knots), whereI is the Beaufort interval reported by theobserver. Table 11-1 converts Beaufortnumber to knots at a height of 33 feet (10 m)above the surface. For example, the esti-mated standard error of a wind speed of 20knots would be 3 knots (0.6 X 5). Over thewestern Pacific between 10°N and 10°S,Morrissey et al.(1988) estimated the error tobe 5 knots. The direction error is estimatedto be +/- 10° when wind is reported in tensof degrees. Besides the standard error ofestimating winds, wind measurementsdepend on anemometer height and expo-sure. Bunting (1968) found a large range inthe distribution of anemometer heights fornon-Navy ships. The average height wasabout 66 feet (20 m) with a range from 26 to121 feet (8 to 37 m).

The one-seventh power law gives a goodestimate of the variation of wind speed withheight near the surface:

V/Vo = (Z/Zo)1/7

Where V is the wind at some level Z and Vois the wind at some reference level Zo. Table

Table 11-1. Relation between Beaufort wind-scale number and wind-speed (knots) as includedin synoptic reports (from World Meteorological Organization, 1966b).

Beaufort Mean Speed at 10 m Beaufort Mean Speed at 10 mNumber Above Surface Number Above Surface0 < 1 7 28 - 331 1 - 3 8 34 - 402 4 - 6 9 41 - 473 7 - 10 10 48 - 554 11 - 16 11 56 - 635 17 - 21 12 > 636 22 - 27————————————————————————————————-

11-2 shows the ratio of the wind speed atvarious levels to the wind speed at 66 feet(20 m). Thus, for the usual range of an-emometer heights, wind speed differencesof up to +/-10 percent from the winds mea-sured at 66 feet (20 m) could be expected.Ships do not include anemometer heights intheir synoptic reports. More and more shipsare being equipped with anemometers,rising from 20 percent in 1962 to 75 percentin 1986. On average, winds measured byanemometers exceed winds estimated fromsea-surface conditions by about 2 knots(Ramage, 1987). Synoptic reports now indi-cate whether an anemometer is being used,so the analyst can make a rough adjustment.Indirect wind indicators can be used. Forexample, wind “packs” cloud against awindward range, while a shear line parallelsthe flow.

d. Cloudiness, Rain, and Visibility. Oncemore, local effects and large diurnal varia-tions make cloud and rain observationstaken from tropical continental or largeisland stations unrepresentative of thegeneral conditions over the surroundingareas. Observations from ships or atolls aremore representative, but are usually tooscattered to allow cloudiness boundaries tobe determined. These difficulties are obvi-ous to anyone who has tried to use surfaceobservations in constructing nephanalysis inthe tropics. Satellites have now removedmost of the problems. Nevertheless, toprovide accurate local-area or point fore-casts, tropical meteorologists must be thor-oughly familiar with local effects on cloudi-ness and rain and their diurnal variations.To achieve this, detailed study of current

and past satellite pictures is essential. Satel-lite pictures are available from the NationalEnvironmental Satellite, Data and Informa-tion Service (NESDIS), in the form of polarhemispheric visual and infrared mosaics.Geostationary satellite pictures are alsoavailable. The Cooperative Institute forResearch in the Environmental Sciences(CIRES) of the University of Colorado ar-chives DMSP records. Station files of GOES,GMS, METEOSAT, ESA pictures can also beconsulted.

During daylight, reports of cloud types andamounts are limited by the experience andtraining of the observer and by the meteoro-logical codes that do not permit a completedescription of tropical clouds. At night,especially when there is no moon, cloudreports are much more uncertain. For ex-ample, cirrus is usually present over most ofthe tropics but often cannot be seen at nightand so is not reported. Baer (1956) foundthat at Tucson (32°N, 111°W), Arizona,average cirrus cloudiness on moonlessnights was about half that on nights with afull moon. This leads to biases in meancloudiness and in the difference betweenmean daytime and nighttime amounts.Observations and statistics of low cloudi-ness are generally more reliable than thoseof middle and high cloudiness and are morelikely to show true diurnal variations.

6-hour and 24-hour rainfalls should beplotted on surface charts to aid in identify-ing disturbed weather. Even though theconvective nature of tropical rain causes theamounts to cover a wide range, the knowl-edge is still useful and should supplementsuccessive satellite pictures in determining

Table 11-2. Ratio of the wind speed at various levels to the wind speed at 66 feet (20 m)elevation, based on the one-seventh power law

Height (m) 10 15 20 25 30 35 45 50 (feet) 33 49 66 82 98 115 148 164Ratio .91 .96 1.00 1.03 1.06 1.08 1.10 1.14————————————————————————————————-

the movements and intensity changes ofrain areas. It is important to differentiateshowers and steady rain. In the tropics,showers predominate, but steady rain ismore common than once believed. Wide-spread steady rain may fall from nimbostra-tus east of upper-tropospheric cyclones or inthe later stages of dying tropical cyclones(Zipser, 1977). It is very heavy beneathcontinuous thunderstorms (see Chapter 6,Section C-1b).

Movement and varying intensities of show-ers or thunderstorms can cause visibility attropical stations to fluctuate rapidly. Forexample, as a heavy shower passes, visibil-ity can go from “unlimited” to less than 400m and back to unlimited in 15 to 30 minutesor less. On the other hand, poor visibilityassociated with fog and low stratus canpersist for days when warm, moist airmoves over cool water or up a mountainslope. Fog may develop in mountain valleysin response to radiational cooling, but usu-ally dissipates during the day. The synopticsituation, the topography of the stationsreporting poor visibilities, and the sequenceof weather must all be considered in deter-mining the representativeness of visibilityreports.

The World Meteorological Organization hasyet to devise a procedure for reportingvisibility that adequately meets the needs ofaviation. In its code the minimum visibilityin the horizon circle surrounding the ob-server is reported. But for aviation weatherreports many countries use the visibilityindex, which is the greatest visibility at-tained or surpassed through half of thehorizon circle, not necessarily in contiguouselements. When a shower or a thunderstormrestricts visibility, the minimum visibilitycan differ greatly from the visibility index(sometimes called the prevailing visibility).Therefore to evaluate visibility reports, theforecaster must know which coding proce-dures are being used.

3. Upper-Air Observations. Direct observa-tions are made by sensors travelling throughor suspended in the atmosphere. Rawin-sonde, pilot-balloon and aircraft measure-ments are examples. Indirect observationsinclude estimates of winds at various levelsderived from profilers, single or successivesatellite pictures and vertical temperaturesoundings made from satellite or ground-based radiometers.

a. Rawinsonde and Pilot-Balloon. The raw-insonde is an instrument

package carried aloft by a balloon. Duringthe flight, it measures and transmits tem-perature, pressure and humidity data, whilethe ground tracking of the balloon by radio-direction-finding equipment or by radardetermines wind velocity. The balloonsgenerally reach 65,000 to 100,000 feet (20 to30 km). Careful field trials (Air WeatherService, 1965; Meteorology Working Group,1965) have determined the following root-mean-square (RMS) accuracies of the mea-surements:

Temperature 0.5 to 1.0°C

Pressure 2 mb (surface to 50 mb), 1 mb(above 50 mb)

Humidity 5 percent (temperature above0°C), 10 percent to 20 percent (temperaturebelow 0°C).

Such accuracies are unlikely to be realizedin routine measurements. Several varietiesof radiosonde are used in the tropics. Somehave been so poorly calibrated that datafrom them cannot be analytically meshedwith neighboring data. The European Cen-ter for Medium Range Weather Forecasts(ECMWF) identifies the offending stations.

The square of the error in a calculated con-stant-pressure surface height equals the sumof the squares of the height errors due toerrors in the measurements of temperature

and pressure. Table 11-3 shows the distribu-tion of height errors resulting from an RMStemperature error of 1.0°C and an RMSpressure error of 2.0 mb. The temperatureerror largely determines the height error.

The RMS temperature and pressure-heighterrors are enough to limit the usefulness oftemperature and pressure-height analyses inthe tropics. A further handicap is the ques-tionable validity of the geostrophic andthermal wind relationships in low latitudes.However, 24-hour pressure-height changesat an individual station can be useful indetermining trough and ridge passages andintensity changes.

Radiosonde relative humidity measure-ments are not very accurate. At best, theRMS error is about 5 percent at tempera-tures above 0°C. This accuracy is possibleonly if the humidity element has not alreadybeen exposed to high humidities in passingthrough clouds or rain. In disturbed weatherin the tropics, RMS errors of 10 percent fortemperatures above 0°C and 10 to 20 percentfor temperatures below 0°C may be morelikely. These errors correspond to a dewpoint error of about

2 to 3°C in the lower troposphere, and anerror of 3 to 6°C in the upper troposphere.Most humidity elements cease to functionproperly at temperatures below -40°C,which corresponds to a pressure of about

250 mb in the tropics. The humidity elementcan seldom measure the extreme dryness ofthe air subsiding above the tradewindinversion. This prevents inferring where theair came from.

The winds in radiosonde and pilot-balloonreports represent vector winds averagedthrough atmospheric layers of varyingthickness. Between the surface and 23,000feet (7 km), the average is calculated over atwo-minute interval with a one-minuteoverlap. Thus, wind data for the thirdminute is determined from the balloonpositions projected on a horizontal surfaceat the second and fourth minutes. From 7 to14 minutes, two-minute non-overlappingintervals are used, and above 46,000 feet (14km), four-minute intervals with two-minuteoverlaps are used. Since 30-gram pilotballoons rise at 600 to 650 feet (180 to 200 m)min-1 and 100-gram pilot balloons andradiosonde balloons rise at 900 to 1100 feet(280 to 340 m) min-1, wind observationsbelow 46,000 feet (14 km) represent theaverage vector wind through layers about1300 to 2300 feet (400 to 700 m) thick.

The accuracy of radiosonde wind measure-ments depends on altitude and wind speed.Direction is generally accurate to within 5°;however, since wind direction is oftenreported or plotted to the nearest 10°, adeviation of +/- 10° from the plotted direc-tion should be assumed for streamline

Table 11-3. Root-mean-square (RMS) pressure-height errors (meters) arising in radiosondeobservations with an RMS temperature error of 1.0°C and an RMS pressure error of 2.0 mb.

Pressure Due to Temperature Due to Pressure TotalLevel (mb) Error Error Error 1000 0 0 0 700 10 1 10 500 20 3 20 300 35 7 36 200 47 11 48 100 67 22 71 50 87 9* 87* Pressure-height error reverses sign above the tropopause, which is usually near the 100mb level in the tropics.————————————————————————————————-

analysis. When the speed is less than 5knots, the direction should be assumed to bevariable and more leeway given in analysis.The probable RMS errors are related toaltitude in Table 11-4.

In many parts of the tropics, the zonal-windcomponent reverses with height in thetroposphere. This results in a small anglebeing maintained between the local zenithand a line from the tracking radar to theballoon. In these cases fairly accurate mea-surements may be expected, provided thetracking system can cope with rapidchanges in elevation. For most of the tropics,the mean wind vector is less than 30 knots sothat errors of only about 2 m s-1 would beexpected. Beneath a strong subtropical jetstream, the mean wind vector through thetroposphere could exceed 30 knots. Forpractical purposes, an accuracy of +/-10° indirection and +/-10 percent in speed can beused.

Since pilot balloons must be tracked visu-ally, the height achieved by the observationsis limited by darkness (unless the balloontrails a light), or rain or clouds interveningbetween theodolite and balloon.

Even pibals that reach only 1500 to 3000 feet(500 to 1000 m) before entering cloud arevery useful to gradient-level wind analysis,especially when the surface winds aregreatly influenced by local effects.

b. Aircraft. Aircraft reports supplement theinadequate network of ground-based upper-air stations, and enhance tropical analysis.On jet aircraft, winds are measured with

Doppler radar or inertial navigation equip-ment. The estimated RMS temperatureerrors are slightly larger, while the windand height errors are about the same as forradiosondes. Aircraft reports include infor-mation on cloudiness, icing, turbulence, andother significant weather. The elevationdifference between aircraft and cloudsgenerally determines how accurately cloudbase and top heights can be estimated. Aheight difference of 10,000 feet (3000 m)means that the estimates will be good towithin 2000 feet (600 m). Cloudiness esti-mates (scattered, broken, or overcast) aregenerally reliable. Cirrus is not often re-ported. Consequently, cirrus may be presentwhen it is not reported. Aircraft cloudreports supplement satellite photographsand surface observations.

Extent and frequency of AIREPs will in-crease as more ASDAR systems are installed(Chapter 3, Section A-4).

c. Weather Reconnaissance. As pointed outin Chapter 1, weather reconnaissance flightsare almost entirely confined to tropicalcyclones and even these were seriouslycurtailed when tropical cyclone reconnais-sance in the western North Pacific wassuspended. Reconnaissance is limited toNorth Atlantic tropical cyclones. Data areprocessed on board the aircraft and auto-matically transmitted to NHC, Miami. RMSerrors for aircraft reconnaissance observa-tions are 0.5°C for temperature and 3° and 5knots for Doppler wind observations. Inheavy rain, wind speed may be underesti-mated.

Table 11-4. Root-mean-square errors (knots) in rawinsonde wind speeds according toaltitude and the magnitude of the mean-wind vector. (Meteorology Working Group, 1965).

Altitude Range Mean Wind Vector (knots)Zero to000s feet km <30 30-60 60-9010 3 2 5 1020 6 3 7 1539 12 4 14 3059 18 6 21 45

In addition to flight-level observations,reconnaissance aircraft probe the atmo-sphere beneath the aircraft with dropsondeswhich are similar to radiosondes, but arelowered to the surface by parachute. Theymeasure temperature, pressure and humid-ity and their position is tracked by satellite-based OMEGA navigation to give the winds.Performance is about the same as for radio-sondes. Once again, measurement of relativehumidity is the problem both at the aircraftand on the sondes.

d. Satellite-Derived Wind and Temperature.Geostationary satellite data are used incalculating winds. Central processing officesuse a pattern-matching technique to calcu-late low cloud and high cloud motions (seeFigure 3-7). These vectors fairly accuratelyportray the true winds at cloud level(Hubert and Whitney, 1971). At many sta-tions receiving sequences of geostationarysatellites pictures, cloud motion vectors canbe readily determined. Thus, the informa-tion normally available can be enhancedwhere other data are scanty or where sig-nificant weather is occurring. In rain-freeareas over the ocean, surface wind speedsare estimated from microwave images(SSM/I) on DMSP.

Atmospheric temperature and moistureprofiles can be calculated from radiancesmeasured from satellites. The TIROS Opera-tional Vertical Sounder (TOVS) has forseveral years provided synoptic data thathave been incorporated into numericalweather prediction models, and have some-what improved the forecasts (Menzel andChedin, 1990). But the vertical resolution ofthe profiles remains inadequate to replace orsupplement radiosonde measurements inthe tropics. Better ground-truth data fromsuperior radiosonde measurements andnew radiometers with better horizontalresolution should improve matters in the1990s.

C. Manual Analysis TC. Manual Analysis TC. Manual Analysis TC. Manual Analysis TC. Manual Analysis Techniques.echniques.echniques.echniques.echniques.

1. General. This report has already demon-strated that for both synoptic and climato-logical understanding, the tropical windfield contains more detailed informationthan the pressure or pressure-height fields.Thus what follows focuses on analyzing thewinds. Best results are likely if the analystalso sketches in isobars or height contoursand makes sure that general pressure-windrelationships, true everywhere, are satisfied.For example, winds should flow cycloni-cally around a low and anticyclonicallyaround a high; they should curve cycloni-cally through a trough and anticyclonicallyacross a ridge. Except near the equator, andin upper levels over tropical cyclones, windsshould blow across the isobars toward lowerpressure with the angle being greatest in thesurface layers. On the equator, flow shouldbe perpendicular to the isobars towardlower pressure. In other words, wind andpressure analyses should always be mutu-ally consistent. Although some automatedtropical analysis programs have been devel-oped, manual analysis is still widely used inthe tropics. The techniques have changedlittle since they were developed by Riehl,Palmer Saucier, and others during the 1940sand 1950s. The most significant change hasbeen the integration of satellite data into theanalyses. The principles of tropical analysisand the examples presented here provideonly a foundation for skill. To be an accom-plished analyst demands considerablepractice in working with sequences of realdata. To this end, the USAF conducts tropi-cal meteorology courses and training pro-grams.

2. Instruction in Kinematic Analysis.

a. General. All analysts should follow cer-tain procedures. Kinematic analysis con-structs a continuous representation of the

wind field from observations of two-dimen-sional horizontal wind vectors for eachanalysis surface. Since wind is a vector, withboth direction and speed, it can be repre-sented by a stream function field, in whichdistance between streamlines is inverselyproportional to the wind speed (Byers, 1974)and the field is non-divergent. It is evidentthat isobars are stream function lines forgradient-equilibrium flow on a horizontalsurface. Stream function analysis is some-times incorporated on maps disseminatedby weather centrals; although it gives abetter depiction than isobars in low lati-tudes, it is a rather crude representation ofthe flow. The preferred analysis methodrequires two sets of lines (streamlines andisotachs) to represent the field. Streamlinesare everywhere tangential to the winddirection, while an isotach links points withthe same speed. If the lines are sufficientlyclose, wind direction can be determinedfrom the streamlines and wind speed fromthe isotachs at any point on the chart. Fig-ures 4-6 to 4-11 depict low-level and uppertropospheric kinematic analyses for thetropics. In these relatively simple patterns,several anticyclones are embedded in thesubtropical ridge; the low-level troughcontains a series of cyclones with easterlytradewinds equatorward of the subtropicalridge. Circulation features are now defined.

Asymptotes. These are streamlines awayfrom which neighboring streamlines diverge(positive asymptotes) or toward which theyconverge (negative asymptotes). Asymp-totes may or may not represent lines of truehorizontal mass divergence or convergence,depending on the distribution of windspeed in the area. Therefore, it is better touse the terms asymptotes of diffluence (orconfluence) rather than asymptotes of diver-gence or convergence.

Waves. These perturbations in the stream-lines are analogous to the wave-like ar-rangement of troughs and ridges in isobars.

Waves that do not extend across the entirewidth of the current in which they areembedded are called damped waves. In thiscase, the streamlines on one or both sides ofthe current have smaller amplitude thanthose in which the wave is most pro-nounced.

Singular Points. These are points into whichmore than one streamline can be drawn, orabout which streamlines form a closedcurve. The wind is calm at singular points,and the speeds immediately adjacent to asingular point are always relatively light.There are three classes of singular points:cusps, vortexes, and neutral points.

Cusps. These are intermediate patterns inthe transition between a wave and a vortex.Cusps are relatively unimportant in routinesynoptic analysis since they exist onlybriefly in any one horizontal plane; there-fore data are rarely sufficient to determinetheir presence. Figure 11-2 illustrates cy-clonic and anticyclonic cusps in an east-westcurrent.

Vortexes. These are cyclonic or anticycloniccirculation centers and anticyclonic andcyclonic outdrafts and indrafts. Figure 11-3illustrates the six basic types of vortexpossible in streamline analysis plus pureoutdrafts and pure indrafts (the latter twomay be occasionally observed on or near theequator). At low levels, anticyclonicoutdrafts and cyclonic indrafts are commonand coincide with highs and lows in thepressure field. At upper levels, any of thecombinations may occur — cyclonicoutdrafts are common in the upper tropo-sphere over intense tropical cyclones. Dataare usually too sparse at upper levels todetermine the outdraft or indraft character-istics of vortexes; when in doubt favorcyclonic indrafts and anticyclonic outdrafts.Above newly-developed intense convection,particularly in the deep tropics, pureoutdrafts have been observed in the uppertroposphere. After a time, those away from

the equator change into anticyclonicoutdrafts.

Neutral Points. At these points, two asymp-totes, one of directional diffluence and oneof directional confluence, appear to inter-sect. They correspond to cols in the pressurefield, representing a “saddle” between twoareas of anticyclonic flow (high pressure)and two areas of cyclonic flow (low pres-sure).

b. Streamline Analysis. Wind direction canbe analyzed by first constructing an isogonfield, as in Figure 8-52. An isogon joinspoints in a plane with the same wind direc-tion. Even in research, isogon analysis isbeing supplanted by numerical objectiveanalysis programs and is no longer used inthe field. Streamlines are drawn parallelto the observed wind directions. The endresult is a set of streamlines everywheretangential to the wind arrows. To illustratethe procedure, Figure 11-4 shows the basicwind data and Figure 11-5 the streamlines.An isotach analysis is also included in thisfigure. In streamline analysis, interpolationsamong the winds are carried out by eye andthe streamlines are drawn directly using atrial and error approach. The accuracyachieved depends significantly on the skillof the analyst. This procedure should befollowed:

(1) Before starting the analysis, tentativelylocate and mark the dominant features ofthe chart by using the plotted data, includ-ing the pressure or pressure-height fields,continuity (previous analyses), and themonthly wind climatology of the level beinganalyzed. These include features such as thesubtropical ridge lines and monsoontroughs that generally lie east-west and areaxes of maximum streamline curvature,asymptotes of confluence, fronts, etc.

(2) Tentatively locate and mark anticyclonicand cyclonic vortexes along the ridge andtrough lines respectively by using the plot-

ted data, isobaric charts, continuity andsatellite pictures.

(3) Tentatively locate and mark neutralpoints associated with the vortexes.

(4) Mark center positions of tropical depres-sions, tropical storms or typhoons (hurri-canes) according to information from tropi-cal cyclone advisories or warnings.

(5) Having completed these steps, theanalyst should visualize approximately howthe finished analysis will appear, before thefirst streamline is drawn.

(6) First sketch streamlines around thedominant features, such as the NH and SHsubtropical ridges and associated anticy-clones and neutral points. Next turn to anylarge areas of undisturbed flow, such as thetradewinds or the southwest monsoon.Tropical cyclones, with centers fixed byreconnaissance or satellite, can then belocated. Finally the analysis can be com-pleted in equatorial regions; the analyst willdepend mainly on satellite pictures to locatesignificant features.

(7) On low-level maps, work outward fromanticyclonic centers and inward towardcyclonic centers to ensure a more rapid andelegant analysis. This guarantees that diver-gent flow around anticyclones and conver-gent flow around cyclones are readily ap-parent and accurately depicted.

(8) Do not draw a streamline through everyreport as this clutters the map and preventsproper spacing of the streamlines. Analyststend to distort streamlines to make them runthrough data points; this may cause sharpbends or departures from the requirementsof tangency.

(9) Keep streamline-free areas aroundneutral points relatively small. For example,on the surface-gradient-level analysis thisshould be encompassed by the 5 knot iso-tach and on upper-level charts by the 10knot isotach. To accomplish this, draw the

neutral-point intersections first and then theremaining streamlines associated with theneutral points while bearing in mind thereported wind speeds.

(10) Avoid distorting the analysis over toolarge an area on the basis of one report thatlacks substantiating data. It is sometimeseasy to draw a wave in the tradewinds onthe basis of one report. Maintaining continu-ity reduces this risk. Also, before introduc-ing a cyclonic vortex into a broad zonalcurrent, the analyst should consult satellitepictures for evidence of accompanyingdisturbed weather.

(11) Ensure that streamlines approach eachother at very small angles along asymp-totes; avoid “Y-shaped” mergers.

(12) Remember that streamlines usuallycross isobars. Despite their recognition thatin the tropics, especially at the surface,geostrophic and actual winds differ, someanalysts unconsciously try to draw stream-lines parallel to isobars.

(13) Beware of concluding that light wind(less than 10 knots) is evidence of a neigh-boring singularity; it may reflect a singular-ity above or below the analysis level.

(14) Never forget the requirement thatstreamlines parallel wind directions and soavoid the most common error in streamlineanalysis. However, in regions of very lightwinds, some leeway is permitted to achievea reasonable pattern.

(15) Streamlines should not be crowded,but there should be enough to permit easyinterpolation of wind direction at any point.

(16) Having finished the preliminarystreamline analysis, analyze the speed fieldfollowing the suggestions in the next sub-section.

c. Isotach Analysis. After the preliminarystreamline analysis is completed, the wind

speed field is analyzed. The analyst hasalready considered wind speeds during thestreamline analysis to help place singularpoints, ridges, troughs, etc. Isotach patternsresemble those of other scalar fields such aspressure or temperature —there are centersof maximum and minimum values andsaddles or cols. The following relationshipsbetween isotach and streamline patterns canbe applied during isotach analysis. Many ofthem are illustrated by the kinematic analy-ses of Figures 4-6 to 4-11, and 11-5.

(1) The major axes of elongated speed-maxima roughly parallel the streamlines,particularly in the major zonal currents.There is usually an elongated speed-maxi-mum near the center of each streamlinecurrent. In very broad currents, two or moreelongated speed maxima may lie side byside.

(2) Isotachs on either side of a speed maxi-mum also roughly parallel the streamlines.The isotachs are closer together here thanalong the axis of the maximum.

(3) At all singular points, the speed is zero.All other speed minima must have valuesgreater than zero.

(4) Winds are generally light where stream-lines curve sharply, and where a singularpoint exists just above or below the levelbeing analyzed.

(5) Close to neutral points, isotachs shouldapproximate ellipses. Farther from a neutralpoint, the isotach ellipse tends to resemble afour-pointed star with the points lying onthe asymptotes that cross at the neutralpoint.

Finally, make any mutual adjustments in thestreamline and isotach analyses to ensure areasonable and consistent analysis thatlogically follows the analyses made for theprevious synoptic time. The finished pat-terns should be generally smooth and el-

egant, difficult to define but easy to recog-nize.

d. Frontal Analysis. Satellites have shownthat fronts extend into the tropics during thecool season. Even though an air mass dis-continuity may no longer exist, a band ofcloud can persist along a shear line exten-sion of a cold front (see Figures 8-29 and 8-30). The tropical meteorologist should befamiliar with surface-layer wind flow asso-ciated with two types of front — A, at whichboth warm and cold air are rising; B, atwhich warm air rises and cold air sinks(Figure 11-6). Since on the synoptic scale, aircannot be exchanged across the front, thefront must move at the same speed as thecomponent of the wind perpendicular to it.It follows then, that an asymptote can onlycoincide with a stationary front. Finally,cyclonic vorticity is concentrated in thefrontal zone.

These specifications are used in constructingthe NH frontal wind models shown inFigures 11-7 and 11-8. The frontal segment isshort enough to allow the assumption thatwinds are constant parallel to the front. InFigure 11-7, if the front is oriented SW - NE,the circulation typifies that around a coldfront north of the subtropical ridge, wherenorthwesterlies replace southwesterlies asthe front passes. In the type

A front, although convergence is greatest atthe front, the asymptote of confluence isdisplaced ahead of the front and may notnecessarily coincide with a cloud line. Withthe type B front an asymptote of diffluenceappears behind the front. Successive analy-ses showing asymptote(s) approaching thefront indicate that the front is movingslower. In Figure 11-8, with the front ori-ented WSW - ENE, the circulation typifiesthat around a cold front south of the sub-tropical ridge, moving ahead of a surge inthe tradewinds or in the winter monsoon.The asymptotes appear as in Figure 11-7,but most importantly, there need be no

change in wind direction as the front passes,only an increase in speed. The tropicalanalyst should remember that a streamlinerepresents the instantaneous field of winddirection, and not the trajectory of an airparcel. Thus, all moving fronts are crossedby streamlines; as long as the wind compo-nent perpendicular to the front equals thefrontal speed, air mass separation is en-sured. Figure 11-9 shows a synoptic ex-ample of isobaric and kinematic analysesaround a frontal system. Note the patternsurrounding the wave cyclone. Frontalmodels underline how important the isotachfield is in determining where convergenceor divergence is concentrated (see alsoFigure 6-16b, c, and g). Without it, the as-ymptotes would tend to overpower inter-pretation, and even force the analyst todistort the streamline pattern so as to makea confluence asymptote coincide with themost severe weather.

At many stations, middle latitude contour-height or pressure analyses are merged withtropical kinematic analyses. For levels abovethe surface, this poses no problem, as longas the analyst remembers that streamlinespacing is not related to wind speed. How-ever, merging surface charts produces avery messy result, since the cross-isobariccomponent of the wind ensures that isobarsand streamlines cannot coincide. Preferably,kinematic analysis should be extended as farpoleward as operations require.

3. Use of Aircraft Reports. Aircraft reports(AIREPs) have become more important totropical analysis, because of more flights,better accuracy, and improved communica-tions. Sadler (1965) early demonstrated theusefulness of AIREPs when he preparedglobal tropical streamline analyses forvarious levels for 10 December 1963. Byusing AIREPs, off-time reports, and PIBALs,he more than doubled the informationavailable from RAWINs alone. Since then,

numbers at 250 mb (jet aircraft) have contin-ued to increase, but have decreased at thelower levels. In a typical 250 mb operationalchart for the global tropics (part of which isshown in Figure 11-10), AIREPs total 97,compared to 135 rawin and 146 cloud mo-tion vector reports.

Aircraft wind measurements made withDoppler radar and inertial navigation sys-tems are as accurate as radiosonde observa-tions. Nevertheless, when many AIREPs areplotted on constant-pressure-level charts,inconsistencies are evident, especially inreported wind speeds. Much of this stemsfrom height and time differences betweenthe AIREPs and the charts. On upper tropo-spheric charts the wind shear between the300 and 200 mb levels, as determined fromrawinsonde observations, can be applied toAIREPs to adjust the reported speeds to theanalysis level. (If the maximum wind levellies between 300 and 200 mb, then the shearbetween 300 mb and the level of maximumwind may more appropriately be used,depending on the analysis level.) Since winddirection does not vary much through thelayer between 300 and 200 mb, wind shearper

kilometer can be estimated by dividing thedifference between 300 and 200 mb windspeeds by three. This shear, interpolatedbetween radiosonde stations, can be appliedto AIREP data (Figure 11-11). In this ex-ample, the adjusted AIREP wind speed of 67knots, compared to the 200 mb wind speedof 80 knots at the rawinsonde station to thewest, indicates that the jet-stream core isprobably north of the AIREP location. Basedonly on the two rawinsonde reports, the jet-stream core could easily have been analyzedas being on or south of the AIREP position.Where there are few rawinsondes, climato-logical shear values may be used in adjust-ing AIREP data to the analysis level. Thewind speeds at the analysis level, interpo-

lated from AIREPs, can be plotted in brack-ets to facilitate isotach analysis.

It is not so easy to adjust AIREP data tocompensate for variations with time. Appar-ent inconsistencies can often be removed bymaking the effort. Lenhard (1967) used anumber of previous observational studies ofwind variability to derive the followingequations relating wind variability to windspeed and time or distance:

Time Variability St = (2.9 +0.1W)t0.5

Horizontal Space Variability Sd = 4(2.9 +0.1W)(1.85t)0.5

St or Sd is the RMS change in wind with timeor distance (knots), W is the resultant windspeed (knots), t is the time in hours, and d isdistance in nautical miles. Thus the changein wind over a distance of 8.6 NM (16 km) isequivalent to the change in wind at a pointduring an hour. Figure 11-12 gives thegraphical solution to these equations forperiods of up to six hours (or horizontaldistances of up to 52 NM (96 km)), and windspeeds up to 120 knots. For example, for t =4 hours and W = 60 knots, the resulting RMSwind variability is 18 knots. While it is notpossible to make objective corrections toaccount for wind variability with time, theRMS variability values can roughly helpdetermine how much the AIREP windspeeds should be smoothed (after they havebeen adjusted for height differences).

In some cases, locations of off-time AIREPscan be adjusted with respect to movingweather features such as tropical cyclones ortroughs in the westerlies. AIREPs madebefore synoptic map time are displaced inthe direction the system is moving andAIREPs made after map time are displaced

in the opposite direction. The displacementdistance for an AIREP equals the distancetravelled by the system between map timeand AIREP time. Figure 11-13 illustrates thelocation adjustment for two AIREPs near arapidly-moving mid-latitude trough. Beforeadjustment, the AIREPs seem to show aridge; after adjustment, they pinpoint thetrough position. Aircraft reconnaissancereports around tropical cyclones can besimilarly adjusted. AIREP positions need beadjusted only near sharply defined circula-tion features; therefore, on any one maponly a few adjustments are made. Furtherexamples of the use of asynoptic data inanalysis are given in AWS TR 225 (1960).

Once the AIREP wind data are adjusted forwind shear in the vertical and location(when appropriate) and the normal variabil-ity of wind with time is taken into account,more than 80 - 90 percent of AIREP data arefound to fit well with conventional andsatellite-derived wind data. Where AIREPsare plentiful, reports that are obviouslywrong usually stand out and can be re-moved. Isolated AIREPs, where other winddata are sparse, can be compared to clima-tology and satellite pictures to identify thosethat appear to be badly wrong.

4. Use of Climatology.

a. General. It is virtually impossible toproduce consistently good kinematic analy-ses without carefully considering the windclimatology, a most important analysis aid ifsynoptic data are scanty. In addition, theshort-tour policies of AWS for most tropicalareas mandate that each weather unit main-tain comprehensive climatological files toprovide newly-arrived forecasters with“instant experience”. Kinematic analyses ofresultant wind fields are indispensable (seeFigures 4-6 through 4-11). The resultantwind is the vector, averaged over a period ofyears, of all winds at a particular level at one

station. If possible, at least five years shouldbe averaged. The climatological chartsshould depict resultant winds and (if pos-sible) steadiness (but see Figure 11-15).Steadiness is the ratio of the resultant windspeed to the mean scalar wind speed, ex-pressed in percent. Although not all thestandard pressure levels are routinely ana-lyzed, the wind climatology for these levels— surface/gradient, 850 mb, 700 mb, 500mb, 300 mb, and 200 mb — should be avail-able for training and operational planning.The following sections describe how to useresultant-wind climatology in analysisparticularly where few or no synoptic dataare available.

b. Streamlines. From the resultant stream-lines one can determine the mean locationsof major circulation features, e.g., subtropi-cal ridges, low-level monsoon troughs, heatlows, near-equatorial convergence zones,upper-tropospheric troughs, etc. On theresultant charts, lines should be drawnthrough the centers, neutral points, andpoints of maximum curvature of each sys-tem. The slope in the vertical of major circu-lation systems, as shown by resultant windcharts and mean meridional cross sections,may be applied to synoptic analysis. Com-posite monthly charts superimposing thelocations of major circulation features of thevarious levels (Figure 11-14), and includingjet-stream locations, aid analysis and train-ing.

Since climatological systems are usuallypresent on synoptic charts, wind climatol-ogy will alert the analyst as to what toexpect. In most cases, the systems will bebetter defined on the synoptic charts than onthe mean charts because climatologysmooths out short period variations. Forexample, over the North Pacific and NorthAtlantic during the warm season, resultantwinds turn cyclonically through the upper-tropospheric troughs by 90° or less. On

synoptic charts, wind shifts of 180° acrossthe troughs are common.

Resultant wind charts help identify signifi-cant circulation anomalies that often accom-pany significant departures from the normalcloud and weather patterns for that month(see also Chapter 12). They also help deter-mine the boundaries of local-area charts tobe analyzed by forecasting units. The analy-sis area should encompass the mean posi-tion and normal variations of the majorfeatures that affect local-area weather. Sta-tions with a large annual variation in circu-lation patterns, especially in monsoon re-gions, may shift their analysis areas so as toadequately track tropical systems duringsummer and mid-latitude systems duringwinter.

c. Isotachs. Resultant wind isotachs give apretty good idea of wind variability. Figure11-15 relates monthly mean resultant-windspeeds and steadiness at the gradient level,based on monthly data from 23 stations welldistributed throughout the tropics. Similarrelationships are found for higher levels;however, the lines of best fit may differ.Thus, for practical purposes, isotachs of theresultant wind could stand in for steadiness.Figure 11-16 relates resultant gradient-levelwind speed and the percent of time thewind blows from +/- 45° of the resultantdirection, computed for the same stationmonths used in Figure 11-15. Wiederanders(1961) related modal wind direction (winddirection most often observed) to resultantwind direction and steadiness at six Pacificarea stations. His sample consisted of 360cases (12 months x 5 levels x 6 stations)(Table 11-5). The results imply that resultantdirections are close to modal directions

when steadiness exceeds 30 percent or whenresultant speeds (at gradient level) exceed10 knots.

These relationships between resultant-windspeed and wind variability help synopticanalysis and forecasting. Where resultantwinds are relatively light, wind directionwill fluctuate greatly from day-to-day. Thetype of circulation (trough or ridge) associ-ated with a resultant wind-speed minimumindicates the type of synoptic circulation(cyclone or anticyclone) to be expectedthere. Conversely, circulation centers wouldseldom be found in regions with strongresultant wind. Height determines what ismeant by “light” and “strong”. At the gradi-ent level, relatively light winds may be lessthan 10 knots and relatively strong winds bemore than 15 knots. At the 200 mb level, thecriteria are more likely to be 20 knots and 40knots respectively.

Charts of resultant winds can be used inchecking for errors in synoptic reports. Forexample, a single report of a west wind inan area normally having moderate to strongeasterlies should be carefully examined forplotting or transmission errors and com-pared to surrounding reports and satellitepictures before it is accepted. Occasionally,such errors cause vortexes to be wronglyanalyzed in the tradewinds and even carriedon subsequent analyses on the basis of“continuity”. In another example, a report ofa 20 knot wind in the lower tropospherewhere the resultant wind speed is less than5 knots might be a clue to storm formation,but one should look at the latest satellitepicture for supporting evidence of a vortex.

Table 11-5. Percentage frequency of modal wind directions differing by less than onecompass point from resultant-wind directions (after Wiederanders, 1961).Steadiness (percent) 0-19 20-29 30-49 50-69 70-100Percentage frequency 25 50 75 90 100————————————————————————————————-

5. Use of Satellite Data. Chapter 4 of AWSTR 212 (or ESSA NESC TR 51) discusses theuse of satellite data in tropical analysis andincorporates satellite images. Since thereport is readily available, details will not bepresented here. Other publications includeWeldon (1979), Dvorak (1984), Fett andBohan (1986), Dvorak and Smigielski (1990).Weather satellite images have illustrated theearlier chapters of this report and haveformed the basis for the cloudiness andrainfall climatologies of Chapter 6. In sum-mary, the following useful information canbe obtained by applying the techniquescovered in AWS TR 212:

Estimates of upper-tropospheric winds fromcumulonimbus plumes and cirrus shields.

Estimates of lower- and upper-troposphericwinds from cloud-motion measurementsfrom consecutive geosynchronous satellitepictures.

Estimates of maximum wind speeds intropical cyclones.

Location of major trough and ridge systemsin the mid-latitude part of the analysis area.

Location of subtropical jet stream.

Estimate of low-level wind direction fromorientation of cumulus lines.

Location of frontal zones.

Location of cyclonic vortexes in the upperand lower troposphere.

Location of low-level ridges in the subtrop-ics.

Location of large-amplitude upper troughsthat extend into the tropics.

Estimates of surface wind direction and/orspeed from sea breezes, topographic effects,sun glints, anomalous cloud lines in fog andstratus, etc. Many diagrams in Chapter 7show diurnal variation curves crossing zero

near local noon and midnight. Thus, se-quences of geostationary satellite images forthese times give the forecaster the leastdistorted view of synoptic and mesoscaleweather systems, and help determine tracksand intensities.

Weather satellite images have become essen-tial to tropical analysis, but it is worth re-membering that they have not solved all theforecasting problems, as many expectedback in 1960. Well-defined weather systems,such as tropical cyclones, troughs in theupper tropospheric westerlies, near-equato-rial convergence zones, fronts and shearlines are easy to recognize and to track fromone image to the next. This is true neitherfor squall lines, nor for that child of weathersatellites -the cloud cluster. Most of the time,cloud signatures cannot be linked to circula-tion features; models based on the imagesapply to a small fraction of the “organized”cloud systems seen by the satellite. Satellitestell us what clouds are there, but most of thetime they don’t tell us why, or help us fore-cast development, dissipation or motion.From the satellites we have learnt that atypical tropical cloud system is much morevariable than we had expected, and thatmerely extrapolating movement rarelyensures a good forecast. Perhaps exhaustivestudies of surges will lead to new modelsand to explaining some of what is nowmysterious. This will come only if satelliteimages are closely linked to the other impor-tant variables of winds, lapse rates andmoisture, and satisfy three-dimensionalcontinuity.

6. Post-Analysis Programs. LaSeur (1960)recommended an active post-analysis pro-gram. In it, after the immediate pressure ofissuing forecasts is past, the most recentcharts are reanalyzed. The analyses arerevised in the light of late data and subse-

quent synoptic developments. In addition,the analyses and satellite data are carefullycompared to ensure consistency. Post-analy-sis should compare charts at various levelsfor vertical consistency and prepare chrono-logical continuity charts showing successivelocations of major circulation features.Depending on available staffing, schedules,and other demands, post-analysis can becarried out either once or twice daily. Ag-gressive post-analysis ensures that the bestpossible continuity is established, to thebenefit of the next analysis cycle.

7. Examples of Kinematic Analyses. Toillustrate analysis principles, a set of kine-matic analyses for the western North Pacific/Asia region is shown in Figures 11-17through 11-19. Figure 11-17 shows thecomposite lower-tropospheric chart inwhich the 850 mb flow (the arrows and thefirst set of winds in parentheses directlyunder the station) is analyzed. To facilitatethe analysis, the 3000 feet (1 km) winds(shown as the second set of winds in paren-theses) and ship winds (shown in wind-barbform) are included. Scalloped lines encloseareas of significant cloudiness (multi-lay-ered broken to overcast cloud and/or areasof active convection). Although used in theanalysis, surface synoptic data have beenomitted here to avoid cluttering. Windreports considered to be not representativeor wrong are marked with “X”s. The majorsynoptic features are tropical storm Dot, justsouthwest of Japan, the monsoon troughlying generally east-west near 20°N, andbroad westerlies between about 5° and20°N. Tropical Storm Dot interrupts themonsoon trough circulation with westerliesbetween about 120° and 130°E and 5° and30°N. This situation is common during thetyphoon season when typhoons or tropicalstorms may move across the Japan/Okinawaregion. Over southeast Asia, the strongsouthwest monsoon is accompanied by

cloudiness and convection. Note the sharpbreak in cloudiness along the east coast ofVietnam caused by the downslope flow overthe Annam Mountains.

In the 500 mb analysis (Figure 11-18) thelow-latitude ridge and the monsoon troughalong 20°N are well defined. Anticyclonescover China while the cyclonic circulation ofTropical Storm Dot is clearly evident south-west of Japan. Note the generally lightwinds at 500 mb; only near the tropicalstorm do winds exceed 20 knots.

The 200 mb analysis is shown in Figure 11-19. The circulation features are simpler thanat the lower levels. Easterlies dominate mostof the region south of 25°N. The cycloniccirculation of tropical Storm Dot has givenway to an anticyclone east of the low-levelcenter. Over eastern China an intense cy-clone overlies generally anticyclonic circula-tions at 850 and 500 mb. East of the uppercyclone, Figure 11-17 shows extensive con-vection (see Figure 8-18). After developingin the upper-tropospheric trough over thewestern North Pacific, this cyclone driftedwestward over China.

This analysis set illustrates how most of thesynoptic-scale cloud features are related tocirculation features at various levels and toterrain influences. As over much of thetropics, the lower- and upper-troposphericcirculations in this area and season arelinked more closely to the cloud andweather patterns than are the 500 mb circu-lations. These associations may not be soobvious in the parts of the tropics wheredata are scanty.

8. Recommended Analysis Levels. In gen-eral, the levels best suited to relating circula-tion features to weather patterns in thetropics are a near-surface level which isrelatively free of frictional effects and a levelin the upper troposphere. There are excep-tions, particularly mid-tropospheric cy-

clones, which have already been discussedin Chapter 8. In most tropical areas, thegradient-level over land, depicting thefriction-free flow at about 3000 feet (900 m)elevation, combined with the surface levelover the sea, and the 200 or 250 mb level aresuitable choices for the basic charts to beanalyzed. Fortunately for tropical meteo-rologists, most observations are made atthese levels. The data available and theiraccuracies have been discussed in earliersections.

In view of the wide variety of wind-datasources it may be more appropriate toconsider making “layer” analyses ratherthan analyses at fixed levels. Such charts cangive the most information to analysts. Asshown in Figure 11-20, a near-surface layerdepiction could include the regular surfaceobservation (including pressure), the 3000feet (900 m) wind, indicated by an arrow,and the 850 mb wind velocity indicatednumerically. Adding the 850 mb windgreatly aids subjective interpolation of thelow-level wind speed profile when makingan isotach analysis. It also helps streamlineanalysis when the wind direction changessignificantly with height.

Growing interest in climate change and inair-sea interaction is resulting in moresurface observations over the tropicaloceans from ships, buoys (both anchoredand drifting) and small islands, includingremote weather stations (see Chapter 3).Thus, in the near-surface layer, emphasisshould shift toward observations made atthe surface, supplemented by gradient-levelwinds. Since low cloud motion vectorscalculated from successive geostationarysatellite pictures are far more numerousthan winds from all other sources over thetropical oceans, how might they bettercontribute to analysis? Sadler and Kilonsky(1985) derived climatological shears be-tween ship winds and low cloud motionvectors for the eastern tropical Pacific.

(Figure 11-21 shows shears exceeding 10knots along the equator.) They then success-fully applied the climatological shear tocloud motion vectors averaged over a singlemonth to obtain corresponding surfacewinds. Lander (1986) has found support forthe hypothesis that the shear equals thethermal wind, which in turn is determinedby the pattern of sea surface temperature.Thanks to satellites, the SST can now beaccurately estimated at weekly intervals.Research at the University of Hawaii raisesthe possibility that central offices, whichnow distribute low cloud motion vectors,could before long also derive and distributeuseful surface winds.

The DMSP microwave imager (SSM/I)senses surface wind speed over the ocean inrain-free areas (see Chapter 2, Section A-5).For the period July 1987 through June 1988,Atlas et al. (1991) first made kinematicanalyses of winds observed from ships andbuoys. They then derived wind directionsfrom the analyses, assigned them to speedsestimated by SSM/I, and combined the twodata sets. Increasing the number of observa-tions in this way should be operationallyfeasible

For the upper-tropospheric chart, rawindata at two or three pressure levels (e.g.,300, 250 and 200 mb) may be presented innumerical form with arrows depicting winddirections at the analysis level. To allowtime and height adjustments, this informa-tion must be indicated for each AIREP.Level-of-maximum-wind reports may alsobe plotted. Including temperature andheight or D-value data should be optional,depending on the season and the location ofthe reports. In the subtropics during the coolseason, pressure-height/temperature dataused in conjunction with the geostrophicand thermal-wind relationships can im-prove wind analysis. Avoid plotting infor-mation that has no value to diagnosis orforecasting. Figure 11-22 recommends

plotting models for rawins (pilot balloons)and aircraft wind reports on upper-aircharts. Offtime data should be plotted in redand late data (added to the chart after thepreliminary or operational analysis has beenmade) in green.

Although time, height and distance adjust-ments can be made to jet-aircraft winds, itmay happen that stations heavily dependenton AIREPs would prefer to analyze the 250mb rather than the 200 mb level and soreduce the height adjustments needed, sincemost jets fly closer to 250 mb. If 250 mb ischosen as the analysis level, rawins may notalways include 250 mb data. In this case, anaverage of the 300 and 200 mb winds will bean adequate estimate of the 250 mb wind,because in the tropical troposphere, thestrongest winds are usually above or near200 mb.

Besides the basic charts described above, aunit’s operational requirements must deter-mine what other levels should be analyzed.For example, forecasts for non-jet aircraftoperating between 3 and 6 km may require700 mb or 500 mb analyses, if adequatewind analyses or forecasts are not availablefrom weather centrals. Over south Asiaduring the southwest monsoon, 700 mb or500 mb analyses help identify and trackridges and mid-tropospheric cyclones.During the northeast monsoon, 500 mbanalyses do the same with mid-troposphericshort waves (vorticity centers) that maytrigger surface cyclogenesis and monsoonsurges. In summer, over northwest Africa,weather-causing cyclones are often best-defined near 700 mb. Over the plateau ofeastern and southern Africa, the 700 mblevel is relatively free of the terrain effectsthat influence lower-level winds. Althoughmany factors determine choice of analysislevels, it is well to remember (Gabites, 1963)that the forecaster should concentrate ononly two or three, and not disperse hisenergy in sketchily treating many levels.

9. Frequency and Scale of Analysis. Com-pared to higher latitudes, tropical weathersystems move slowly and usually developslowly. It often takes days for a tropicalcyclone to grow from an initial disturbance,whereas extratropical cyclones can explo-sively deepen in 12 to 24 hours. In higherlatitudes, three- or six-hourly analyses ofsurface pressure and frontal systems canbenefit short-range local-area forecasts. Inthe tropics, however, surface pressure analy-ses intermediate between the times of up-per-air soundings (00 and 12 UTC) are likelyto be of little use and often not worth theeffort. Therefore, analyses made at 12-hourintervals usually suffice for operational andforecasting purposes. It is better to makehigh-quality analyses twice a day than tomake less accurate and less detailed analy-ses four times a day. In fact, Riehl (1966)suggested that in some parts of the tropicsduring the dry or “good” weather season,once-daily analyses would satisfy localforecasting needs. Military operations prob-ably require at least twice-daily analyses.The chosen times may depart from 00 and12 UTC. Over Africa, more lower-tropo-spheric wind data are available at 06 UTCthan at 00 UTC, while in India the favoredhour is 09 UTC.

The map scale for analysis should dependon many factors -the spacing of reportingstations in the densest part of the network,the total area covered by the analysis, practi-cal chart-size limits, etc. For tropical analy-sis, a Mercator projection is best, regardlessof the scale. The scale should be the samefor all levels and for cross sections, to allowoverlaying on light tables. A 1:10 millionscale may be appropriate for regional analy-ses, a 1:20 million or 1:40 million scale forhemispheric or global analyses. It may notalways be necessary to plot all surfaceobservations. Instead, only stations thatmake upper-wind measurements and other

stations selected to provide a good coveragesuffice. Because of their representativewinds, all ship reports should be plotted.Map scales of 1:2.5 million or even largerscales may be suited to briefing or displayor for plotting radar scans or reconnaissancereports.

10. Operational Procedures. Kinematicanalysis of the wind field at various levels ismost commonly practiced in the tropics.Some tropical stations, whose area of re-sponsibility extends into mid-latitudes,combine sea-level pressure or pressure-height analyses in middle latitudes withkinematic analyses in the tropics. The twolines cannot be easily merged, especially inthe surface layer. The area covered by eachtype of analysis may depend on season andsynoptic situation. Stations whose primaryinterest is the tropics should produce high-quality kinematic analyses, while relying onfacsimile charts from mid-latitude weathercenters for pressure (pressure-height) andfrontal analyses. Synoptic mid-latitudeanalyses are usually available by the time allthe tropical data have been received andplotted. Despite projection differences,frontal positions and selected isobars (con-tours) can be quickly transferred to tropicalMercator-projection charts. Satellite imagescan help verify frontal positions, especiallytheir extensions into the tropics. Preferably,the analyst can use this information inapplying frontal wind models (Section C-2-d) and so extend his kinematic analysis intomiddle latitudes. Once credible analyses arebeing received from one or more mid-latitude centers, a tropical station maygreatly reduce the number of mid-latitudereports plotted on its charts.

The following charts should be displayedfor the analyst: monthly climatologicalcharts for all levels analyzed, and large-scaleterrain maps showing station locations. Bystudying these, the analyst can become

familiar with the local topographic effectsthat influence the representativeness ofweather reports and often control short-period weather changes. The terrain con-tours on weather plotting charts are notsufficiently detailed. All analyses should bemade on a light table with the previouscharts underlain to ensure continuity. Toooften, the analyst ignores continuity orclimatology, so ensuring poor work. Acetatesheets of proper size and marked at selectedlatitude-longitude intersections are essentialto good analysis. There should be enoughacetates for each set of isopleths -isobars/contours, streamlines, isotachs — at eachanalysis level.

a. Surface/Gradient-Level Charts. Thesecharts contain the most information anddemand more analytical effort than higher-level charts. Step-by-step:

(1) Underlay plotted chart with the last two12-hourly analyses and overlay acetate forpressure analysis.

(2) Transfer sea-level isobars and frontsfrom centrally-prepared mid-latitude analy-sis. Rely heavily on satellite pictures andcontinuity to locate fronts.

(3) Use later conventional data, ship/aircraftreports and satellite pictures in adjustingfrontal positions.

(4) From weather satellite pictures, preparea rough nephanalysis of the major cloudsystems and transpose it on to the plottedmap in a suitable color. Outline only thosecloud systems that are several hundreds ofkilometers wide and which appear to con-tain active convection. Apparent centers ofvortical cloud systems should be especiallynoted. Abbreviated remarks can furtherdescribe the cloud systems, e.g., ISOLD,SCT, BKN, or many Cbs, multi-lyrd clds, jet-stream cirrus shield.

(5) Complete the pressure analysis of thechart for the tropics

using a two mb interval in the tradewindsand a one mb interval near the equator,where gradients are small. Adjust the analy-sis to fit the plotted gradient-level winds,especially more than 5 to 10° from the equa-tor.

(6) Trace the isobaric analysis (in lightpencil) and the frontal analysis (in the ap-propriate color or symbols) from the acetateto the plotted map and place a clean acetateover the plotted chart for the streamlineanalysis. (This step is optional, as the iso-baric/frontal analysis could be left on itsacetate and traced later when the kinematicanalysis is finished.)

(7) Make a surface streamline analysis byconsidering and evaluating the plotted data,the pressure field, continuity, climatologyand the latest satellite data. Apply frontalkinematic models.

(8) Place a clean acetate over the streamlineanalysis and analyze the surface windspeeds. Isotachs should be drawn for 5 or 10knot intervals, except near tropical cyclones,where greater intervals are appropriate.Considerable smoothing may be needed toensure a reasonable-looking analysis.

(9) Plot data received since starting theanalysis and adjust the analyses as neces-sary.

(10) Trace the analyzed fields onto theplotted chart. The isobars should be drawnlightly with a lead pencil, allowing thepressure patterns to be seen but not topredominate. Fronts should be traced in theappropriate colors or by a lead pencil withthe appropriate symbols. Streamlinesshould be traced with a purple pencil oranother selected color, with enough arrow-heads to identify flow direction at any point.Isotachs should be traced as dashed lead-pencil lines, and be labeled for wind speeds.It is recommended that isotach minima beshaded light green and isotach maximashaded light yellow. Cyclonic centers should

be labeled C and anticyclonic centers, A. Theappropriate symbol locates tropical cyclonecenters. Beneath, the maximum surfacewind speed, the storm name, and the advi-sory or warning number should be entered.

(11) If the chart is to be used for briefing,the appropriate colored symbols should beentered at stations reporting rain or loweredvisibility.

b. Upper-Level Charts. Analyzing upper-aircharts is more straightforward, since what isneeded is an accurate kinematic analysisrather than the kinematic, pressure, andweather analyses performed on the surface/gradient-level charts. Step-by-step:

(1) Underlay the plotted chart with the lasttwo 12-hourly analyses and overlay anacetate sheet for streamline analysis.

(2) If available, transpose selected heightcontours from a centrally-prepared mid-latitude analysis. In the mid-latitudes andsubtropics, where westerlies generallyprevail, height contours usually parallel theflow and can be used as streamlines. Thewest-east contour that borders the zonalwesterlies on their equatorward side shouldfirst be drawn in, and then the contours topoleward should be added. Height contoursfrom the 300 mb or 200 mb analyses can beapplied to a 250 mb streamline analysis.Lack of a centrally-prepared pressure-heightanalysis requires a streamline analysis bemade of the subtropical and middle latitudewinds.

(3) Equatorward of the height contourstransposed to the acetate in the previousstep, consider and evaluate the plotted data,continuity, climatology and satellite data inmaking a streamline analysis.

(4) Place a clean acetate sheet over thestreamline analysis and analyze the wind

speeds. For an upper-tropospheric chart,isotachs should be drawn at 10 knot inter-vals up to 20 knots and beyond that at 20knot intervals.

(5) Trace analyzed fields on the plottedchart with a soft lead pencil, streamlines insolid lines with enough arrowheads todefine the flow direction at any point, isot-achs in dashed lines, labeled for speed.Isotach minima should be shaded in lightgreen and isotach maxima in light yellow.Jet streams are located by heavy lines witharrowheads drawn parallel to the stream-lines along the axes of isotach maximagreater than 60 knots. Cyclonic and anticy-clonic centers are labeled C and A respec-tively. Tropical cyclone centers should beshown in the same way as on the surface/gradient level chart.

DDDDD. Automated Analysis T. Automated Analysis T. Automated Analysis T. Automated Analysis T. Automated Analysis Techniques.echniques.echniques.echniques.echniques.

1. General. Global weather prediction andclimate prediction have begun to emphasizesurface heat-exchange in the tropics. Sincethe most important component of this is thesurface wind, numerical analysis of tropicalwinds has become more than peripherallysignificant to global analysis centers.

Weather centrals are undertaking two typesof automated kinematic analysis. Theyroughly correspond to direct subjectiveanalysis of the wind field as carried out inthe tropics and to subjective extrapolation ofpressure analysis from higher latitudes intothe tropics.

2. Objective Wind Analysis. As an example,since September 1983, the Australian Bureauof Meteorology has analyzed troposphericwinds between 40°N and 40°S and 70°E and180°E (Davidson and McAveney, 1981). Theanalysis uses an optimum interpolationscheme (Bedient and Vederman, 1964) in

which the mean square interpolation error isminimized. The analysis sequence startswith climatology as a first guess field. Thezonal and meridional components of thewind are first analyzed and then combined(Figure 11-23). The objective analyses agreedwell with subjective analyses made duringWinter MONEX. Compared to rawin mea-surements, the near-surface chart had anRMS error of 6.8 knots and the 200 mb chartan RMS error of 8.9 knots. Divergence fieldscalculated from the winds matched wellwith cloudy and clear areas in satellitepictures. Provided preliminary error-check-ing equals that of a human analyst, objectivewind analyses are probably as good assubjective analyses.

3. Numerical Weather Prediction Analysis.According to Trenberth and Olson (1988), atboth NMC and ECMWF,

“... analyses are produced using a four-dimensional data assimilation system thatuses a set of first guess fields as the base forintegrating observations to produce theanalysis. The first guess analysis at bothcenters is either a 6- or 12-hour forecast froma Numerical Weather Prediction (NWP)model using a previous analysis as its initialcondition. The first guess carries forwardinformation from the previous analyses. It isdependent upon the veracity of the NWPmodel and it can be biased. The subsequentprocedure itself is based upon a scheme thatmakes use of the statistical mean errorsexpected in both the first guess and in theobservations. The analysis is initializedusing nonlinear normal mode initialization,a procedure designed to ensure that theresulting fields are dynamically consistentwith each other while appropriately empha-sizing relatively slow-moving, meteorologi-cally significant components and dampingspurious gravity waves.” Although thecenters use similar procedures, Trenberthand Olson found that in the tropics RMS

differences in the east-west and north-southcomponents of the analyzed winds oftenexceeded 10 knots above 500 mb, and thecomputed divergent wind fields, and henceassociated vertical motions, significantlydisagreed (see also Lambert, 1989).

The WMO Working Group on NumericalExperimentation (1987) is continuing tocompare operational tropical (40°N to 40°S)analyses made at various numerical weathercenters. For May 1987, between 90°E and thedateline, the ECMWF and the AustralianBureau of Meteorology showed RMS differ-ences of about 8 knots at 850 mb and 14 to16 knots at 200 mb; these were about thesame as differences between the analysesand points at which observations weremade. Studies underway will include analy-ses from the United Kingdom Meteorologi-cal Office in the comparisons.

Errors of this size between analysis andobservations (on which the analyses werebased) mean that tropical forecast officescannot depend on centralized numericalanalyses. They must continue to make theirown careful kinematic analyses, preferringobjective wind analyses as back-up.

4. Computer-Aided Analysis. The Auto-mated Weather Distribution System(AWDS) now being installed in AWS fore-cast offices, obtains grid-point data fromGWC and analyses them on a screen. Newdata can be added, the field reanalyzed bythe operator, and a hard copy made. Zoom-ing allows the scale of the screen map to bechanged. Various fields can be overlaid,including satellite images.

EEEEE. Auxiliary Aids.. Auxiliary Aids.. Auxiliary Aids.. Auxiliary Aids.. Auxiliary Aids.

Auxiliary charts, diagrams, and other aidssupport tropical analysis and diagnosis.Their types and numbers differ among

operational weather units, which shouldperiodically check whether the ones theyuse are still needed.

1. Time Cross Sections. These have beenused in both operations and research. Theyshow the distribution of radiosonde and/orwind measurements above a station over aperiod of time. In the Johnston Island (17°N,170°W) vertical cross section shown inFigure 11-24, wind directions are indicatedby arrows (north toward the top of the page)and speeds by numerals. The six-hourlysurface synoptic observations are also plot-ted. In the tropics, disturbances usuallymove from east to west, that is, from right toleft on a chart. In the time sections, timeincreases from left to right and so priorconditions are shown to the left of subse-quent conditions. Radiosonde data may alsobe plotted on time sections. Dew points, orthe depressions of dew points below tem-peratures, including those at significantlevels, help monitor changes in moisturedistribution, especially within the near-surface moist layer.

Time sections help maintain continuity inthe plotted elements during the periodsbetween synoptic charts and in the layersbetween analysis levels. Time sections makeidentifying wrong or non-representativereports easier. For example, the circled windreport in Figure 11-24 suggests an abruptchange with height from easterly to south-erly winds and back to easterly near 5000feet (1.5 km). A transmission is probably toblame and the actual wind direction is likelyto be 70° rather than 170°. Had this windbeen plotted as 170° and so analyzed on asynoptic chart, a serious error would haveresulted. Time sections indicate the heightand thickness of transition layers betweeneasterlies and westerlies. Synoptic kinematicanalysis in such transition layers is oftenhard because winds are light and variable.

Time sections should be maintained for thehome station, for other stations for whichforecasts are issued, and for nearby up-stream stations. Since hindsight is an exactscience, time section post-analysis oftenhelps research. In operations, however, timesections are analyzed step wise as databecome available. In general, extrapolatingtrends does not help forecasting much.Possible types of analysis are now listed:

a. Wind Direction. Wind direction can beanalyzed in two ways — by drawingisogons (see Figure 8-52) or, more com-monly, by drawing pseudo-streamlinesparallel to the wind directions as plotted onthe section (see Figure 8-22). On the section,singular points are found at the intersectionsof the sloping axes of moving singularitiesin the horizontal wind field with the planeof the time section. Associating singularpoints on the horizontal and vertical chartshelps the meteorologist to better understandthe three-dimensional structures of theparent circulations.

b. Wind Speed. At some stations, cloud andweather changes are associated more withspeed changes than with direction changes.Therefore isotach analysis on time sectionsmay help reveal this weather-modifyingmechanism. As on horizontal wind charts,isotachs are drawn at 5, 10, or 20 knot inter-vals. Some smoothing is necessary. Isotachanalysis on sections helps detect wind-speedmaxima and minima, and over time deter-mine the usual levels at which the extremesoccur.

c. Moisture. Moisture can be analyzed inseveral ways: dew points, dew point spread,or departures of dew point from normal. Indew point analysis, a 4 or 5°C interval maybe sufficient; for dew point spread or depar-ture from normal, a 2°C interval should beused. The boundary of the moist layer canalso be drawn. LaSeur (1960) recommendedusing the height of the 5 g kg-1 moisturesurface during the rainy season and the 3 g

kg-1 surface during the dry season to indi-cate the depth of the moist layer.

d. Temperature/Height. Isotherms or con-tours (or their departures from normal)could be analyzed on time sections but formany tropical stations would provide littleuseful information. The 24-hour pressure-height changes are useful. When they areplotted in the middle of each 24-hour periodfor which the difference is determined, theyare controlled by the intensity and move-ment of circulation systems. For example, apressure-height rise before a trough passesindicates weakening; equal and oppositepressure-height changes ahead of and be-hind a trough indicate no movement orintensity changes; the zero isopleth shouldthen coincide with the wind shift line.

2. Space Cross Sections. These give a syn-optic picture in the vertical of the distribu-tion of various elements at a number ofstations along a particular route, spacedalong the abscissa according to distance.Analyses resemble those for time sections;time section forms can usually be modifiedfor use as space sections. In some cases,daily space cross sections may be preparedfor a route instead of a time cross section foreach station along the route. Space crosssections may also be used in route briefing;time cross sections could not.

The Automated Weather Distribution Sys-tem (AWDS) can plot, analyze and store thescalar variables depicted on time and spacecross sections. New soundings can be addedand old soundings removed, the analyseskept up-to-date, and hard copies made.

3. Thermodynamic Diagrams. These areless useful to tropical analysis than they areto frontal, air-mass, cloud, and stabilityanalyses in higher latitudes. Nevertheless,they are of some help in tracking the vertical

distribution and changes of atmosphericmoisture.

For AWS units, the standard thermody-namic diagram is the Skew T-Log P chartwhich is printed in several different ver-sions. Stability indices may not help tropicalforecasts, so they should be computed onlywhere and when local forecast studies haveconfirmed their value. Generally, Skew-Tdiagrams are more useful over tropicalcontinents than over tropical oceans. This isbecause over land the diurnal convectivecycle and day-to-day changes in low-levelmoisture are larger. Even then, weathersystems control lapse rates. During the wetseason, the air is more stable in unsettledrainy weather and less stable in fair weatherwith isolated thunderstorms (see Chapter 6,Section C-1). Thus, a Skew-T diagram con-tains very little predictive information.

Vertical dew point profiles are useful. Dailyvalues should be compared to the meandew point profile for the same calendarmonth, plotted on the same Skew-T dia-gram. Alternatively, plot the monthly meandew point profiles on acetates which can beoverlain on the daily profiles for compari-son. Mean monthly dew points for standardpressure levels are available in monthlyclimatic data summaries (NOAA/NESDIS,1948 —) or can be obtained by AWS unitsfrom USAFETAC.

4. Checkerboard Diagrams. Checkerboarddiagrams, on which hourly-weather obser-vations are plotted, have been used exten-sively by AWS units. A sample is shown inFigure 11-25. (This checkerboard has beentruncated; in standard checkerboards 24boxes in each row allow a day of hourlyobservations to be plotted.) The same formcan also be used to depict hourly or three-hourly observations for many stations on agiven day. Then the stations should beordered geographically. For example, sta-

tions along an air route should be arrangedsequentially from one end of the route to theother. This readily reveals the currentweather distribution. Checkerboards alsomake computation of 24-hour changes easy.Various weather phenomena should beemphasized on checkerboards by colorshading or symbols.

Checkerboards provide a statistical aid toanalysis and forecasting. For example,during the rainy season, the diurnal cycle ofconvection can be deduced from the hourlyobservations made over the previous weeks.The relative importance of local and synop-tic effects on winds and convection can beestimated by comparing hourly observa-tions with corresponding synoptic analyses.Typical diurnal cycles of fog and stratus canalso be established. For example, when fogoccurred at Don Muang (14°N, 101°E) dur-ing January 1968 (Figure 11-25), it generallyset in between 05 and 06 LT and broke upbetween 08 and 10 LT. Thus, when fog wasexpected during this period, it could havebeen confidently forecast to occur between06 and 09 LT.

5. Pressure-Change Charts. Twenty-four-hour pressure-change (isallobaric) chartshave been used by various tropical weatherunits for many years. The charts can beprepared by subtracting yesterday’s fromtoday’s pressure at fixed reporting stations.Where stations are few, pressures can beinterpolated from the isobaric analyses at 5°grid points and the interpolated values usedto compute the 24-hour changes. In theanalysis, negative isallobars are drawn inred, positive isallobars in blue and zeroisallobars in purple. One mb spacing isappropriate.

According to LaSeur (1960), 24-hour pres-sure tendencies have synoptic value morethan about 10° from the equator, but havelittle value near the equator. Dunn (1940)

used 24-hour isallobaric charts in studyinghurricane development over the NorthAtlantic. They have since been used else-where in the tropics to monitor potentialtropical cyclone formation. In particular,areas with 24-hour pressure falls of two orthree mb or more, were closely watched.Nowadays, isallobaric charts are less usedbecause developing cyclones can usually bedetected in satellite pictures before theycause pressure falls in the sparse stationnetwork. If the very large area changes insurface pressure discussed in Chapter 4,Section B-2 are monitored on charts of 24-hour pressure tendencies, the forecaster canthen distinguish between synoptic andregional causes for the changes, and may behelped in detecting surges.

Twenty-four-hour isallobaric patterns aremuch harder to relate to minor tropicalweather disturbances. Observational errors,local convective and radiational effects onsurface-pressure, scarce data in many areas,and widespread synchronous pressure falls(see Figure 4-4) complicate the problem.Therefore, each unit must decide whetherplotting 24-hour pressure change charts isworth the effort.

6. Wind-Shear Charts. Charts of the totaltropospheric wind shear help in forecastingtropical cyclone development. As discussedin Chapter 9 and further elaborated by Gray(1978), small wind-shear in the vertical isone of the necessary conditions for tropicalcyclone development. Since the shear canchange considerably from day-to-day, theanalyzed shear field should be frequentlyexamined to determine which of manyminor tropical disturbances shown in satel-lite pictures are most likely to develop. Inthe past, scarce upper-air data, and the timerequired to plot the information, stoppedroutine preparation of wind-shear charts.Nowadays, more data and automated objec-tive wind analysis programs, as well as an

office personal computer, make wind-shearcharts easy to prepare. For example, theNational Hurricane Center prepares inte-grated wind velocity charts for the lower-(1000 to 600 mb) and upper- (600 to 200 mb)tropospheric layers and a chart of the windshear between the two layers (Simpson,1970a). Stations making kinematic analysesof lower- and upper-troposphere on thesame map scale can superimpose them andestimate and outline areas of small verticalwind-shear, say less than 10 to 20 knots.

7. Rainfall Analyses. Separate analyses of6-hour or 24-hour rainfall may help deter-mine the presence, intensity and movementof tropical disturbances. The more observa-tions, the better the charts. As mentioned inChapter 6, Section C-11, the mesoscalenature of rainfall ensures large point-to-point differences in amounts. Therefore,only generalized analyses should be basedon rainfall reports from synoptic stations,e.g., outlining areas receiving measurablerain. The maximum amounts measured maygive a rough idea of how intense a distur-bance is. Whenever available, satellite dataand weather radar reports should be inte-grated into the rainfall analysis. For local-area rainfall analysis, radar is the best tool(see Chapter 12, Section B-6).

Rainfall can be analyzed on acetates over-lain on surface synoptic charts, or the rain-fall reports can be plotted on separate mapsand then analyzed. Stations that did notinclude rainfall in their reports should beindicated. In India, rainfall accompanyingmonsoon depressions is plotted and care-fully analyzed. Rainfall charts can contrib-ute to forecast studies when combined withother data. For example, they can helpdetermine whether disturbances move intoan area or develop and dissipate withoutmoving much.

Chapter 12Chapter 12Chapter 12Chapter 12Chapter 12TROPICAL FORECASTINGTROPICAL FORECASTINGTROPICAL FORECASTINGTROPICAL FORECASTINGTROPICAL FORECASTING

A. General.A. General.A. General.A. General.A. General.

Palmer (1951) surveyed the history ofthe various approaches to tropical analysisand forecasting. He described three meth-ods or “schools of thought” which havebeen applied to the tropics: the climatologi-cal method, the air-mass method, and theperturbation method. The climatologicalapproach was based on the conviction thatday-to-day tropical weather differs littlefrom the monthly and annual means. There-fore, the best guide to forecasting would bea detailed knowledge of the mean values ofthe elements being forecast. In the air-massapproach, the techniques of air-mass andfrontal analysis developed in higher lati-tudes were applied to the tropics. The“equatorial front” was thought to resemblethe polar front; tropical storms resultedwhen waves on it occluded. The perturba-tion method evolved from tropical meteorol-ogy research conducted during and afterWorld War II at the Institute of TropicalMeteorology in Puerto Rico and the Univer-sity of Chicago. In this concept, the basiccurrents in the tropics were considered to begenerally zonal, but subject to wave-like andsometimes unstable perturbations accompa-nied by characteristic weather and pressurepatterns that could be identified on synopticcharts.

Subsequently, tropical meteorology researchhas tended to confirm the basic concepts ofthe perturbation school. More synopticobservations and weather satellite data haveshown the tropics to be dominated by avariety of circulation and weather systems(or perturbations) and that significant day-to-day weather changes are common, espe-

cially during the rainy seasons. Our abilityto forecast these changes, however, hasmade painstakingly slow progress due toour incomplete knowledge of tropical dy-namics and still inadequate data. Riehl’sopinions on tropical weather predictionmade in a 1966 planning report to the WorldWeather Watch are still relevant: “No gener-ally valid methods to predict synopticdisturbances on the 24- to 48-hour timescale, based on analogue, physical, or statis-tical techniques, have been perfected. Thevariety of synoptic-scale disturbances indifferent parts of the tropics is great. Be-cause of severe deficiencies in networks ithas not yet been possible to obtain theircomplete description and to analyze theirmechanics and energetics. Models of certainsimple types of disturbances do exist, buttheir practical use is restricted by the limitednumber of days when the models apply.Hence, no powerful dynamic tools compa-rable to those outside the tropics are avail-able for prediction.”

Riehl went on to say that the primary toolsavailable to tropical meteorologists areclimatological and kinematic. The kinematicmethod relies heavily on extrapolation,occasionally tempered by qualitative consid-eration of interactions between tropical andmid-latitude circulations, and betweenlower- and upper- tropospheric circulationswithin the tropics. The kinematic techniquesare applied mainly to wind-circulationfeatures to predict tropical weather patterns.Their application to upper-air temperatureand height fields has not helped prediction.

Alaka assessed tropical prediction in a 1964WMO Technical Note:

“Short-range weather forecasting consistsessentially of predicting the developmentand movement of weather-producing dis-turbances in the tropics. Before the futurestate of the atmosphere can be accuratelypredicted, its present state must be known.Therefore, the problems of forecasting in thetropics stem largely from inadequate analy-sis and deficiencies in our knowledge of thestructure and properties of tropical distur-bances. Improvement in forecasting iscontingent on the removal of those deficien-cies.” Being conditionally unstable most ofthe time, the tropical troposphere is often ina delicate balance. Small changes in stability,possibly too small to measure, can result inlarge weather changes.

Prior to weather satellites, tropical weatherforecasts seldom bettered climatology orpersistence. Most practitioners believed thatthe usually unexpected weather changesresulted from movement of previouslyunobserved weather. They also believed thata sequence of satellite pictures would enablethem to identify and track disturbances.Then, by simply extrapolating the move-ment, they could greatly improve weatherforecasts. This did not happen. Althoughsatellite pictures filled observation gaps,they also showed that in the tropics, andespecially between 10°N and 10°S, distur-bances not only moved erratically, butrapidly changed intensity.

By 1990, many of the data shortages be-moaned by Alaka have been removed, buteven with the most investigated of all tropi-cal phenomena, the tropical cyclone,progress has been slow. As Figure 9-22shows, the only improvement in forecastingtropical cyclone movement over the past 20years has probably arisen from improve-ments in mid-latitude forecasts. So moredata have demonstrated the complexity oftropical weather forecasting.

The introduction of global dynamic modelshas improved the prospect of a useful pre-

diction model, but their performance in thetropics has barely bettered persistence, anduntil now (Section C) their forecasts havenever been compared to an optimal combi-nation of persistence and climatology. Inrecent years, conflicts in southeast Asia,southwest Asia, Africa, and Central Americaand field programs such as IIOE, BOMEX,ATEX, GATE, MONEX and TAMEX havestimulated research in tropical meteorology.

To develop forecasting techniques for south-east Asia, the Air Force Cambridge ResearchLaboratories (AFCRL) (now GeophysicsLaboratory) and the Naval Weather Re-search Facility (NWRF) set up concentratedapplied research programs, both in-houseand at the University of Hawaii. As a conse-quence, daily forecasts were improved,largely because all available types of datawere integrated — conventional, satellite,radar, aircraft reports, climatology — intothe research. The same approach might leadto better forecasts in other parts of thetropics.

In the past, most of the operational effort intropical meteorology was devoted to analy-sis; forecasting was little emphasized.Nowadays, automated data-handling pro-grams are freeing time for forecasting.

Forecasting suggestions made in earlierchapters will not be repeated here. Thischapter focuses more on the philosophy oftropical forecasting than on proven forecast-ing techniques. The techniques that areincluded illustrate the variety of approachespossible. Experience confirms that withoutsatellite data, accurate tropical analysis andforecasting would be almost impossible.Likewise, very short range local area andterminal forecasts depend on satellite pic-tures. We have also learned that simple-minded extrapolation from a series of satel-lite pictures (especially when intensitychanges cannot be foreseen) can makeforecasts worse than they would have beenwithout the pictures. This chapter assumes

that satellite and radar data are available atfirst-order tropical weather units. Althoughshort-range forecast techniques are empha-sized, the chapter includes sections onmedium-and long-range forecasting andnumerical weather prediction.

B. ShortB. ShortB. ShortB. ShortB. Short-Range F-Range F-Range F-Range F-Range ForecastingorecastingorecastingorecastingorecastingTTTTTechniques.echniques.echniques.echniques.echniques.

1. General. Short-range forecasts are thosemade for periods or valid times up to 36hours in advance. Within this period, fore-casts for 0 to 12 hours ahead have beentermed “Nowcasts”. This subdivision mayhave some value in higher latitudes, but islargely irrelevant in the tropics. Short-rangeforecasts based on careful analysis andextrapolation of existing circulation andweather patterns modified for diurnal andtopographical effects, show a little skill. Theshorter the forecast period and the strongerthe diurnal variation, the better the forecast.A number of short-range forecast tools arediscussed: circulation and cloud prognosticcharts, climatological aids, local-forecaststudies, stability indices and radar. The mostsuccessful tropical forecasters use a system-atic approach and intelligently integratedata from many sources into their routines.

2. Circulation and Cloud PrognosticCharts. The starting point for short-rangeweather forecasts, especially those in the 12-to 36-hour time period, should be circula-tion and cloud prognostic charts based onrecent synoptic analyses and satellite pic-tures. These charts should be on the samemap projections and scales as the analyses.

a. Circulation Prognostic Charts. In mosttropical areas, the basic chart should be forthe surface/gradient level. The forecastpositions of mid-latitude circulation andcloud features are generally available in

transmissions from higher latitude weathercentrals. The tropical parts of the chartsfeature extrapolation of systems shown inthe most recent kinematic analyses. Just asisotach fields are harder to analyze thanstreamlines, predicting them is also hardand generally can be attempted only wheredata are numerous. This means that fore-casting the movement and evolution ofweather systems associated with features inthe wind speed field (e.g., regions of speedconvergence or strong horizontal wind-shear) will be more difficult than forecastingthose weather systems associated withchanges in wind direction (e.g., troughs orcyclones). However, the attempt should bemade, because divergence and convergence(and hence vertical motion and weather) aremore commonly determined by speedgradients than by streamline patterns.

b. Cloud Prognostic Charts. These should bebased on climatology and recent nephanaly-sis prepared from satellite and othersources. Chapter 11 recommended thatnephanalysis outlining convective regionsbe entered directly on the gradient/surfaceanalyses. In this way, successive charts canbe used in extrapolating moving circulationand cloud systems. Forecasts should bemade of changes in orographically-influ-enced cloud systems as they move or as thecirculation changes. Climatology, includingdiurnal variation of convection (especiallyover land) and relationships between circu-lation and cloud patterns (e.g., cloudinessmaxima equatorward of the monsoontrough and along asymptotes of conver-gence) can be incorporated into the cloudprognostic charts. However, the forecastershould never forget that even along themost persistent tropical systems — the near-equatorial convergence zones — cloudinessoften fails to persist (Figure 8-55). Satellitestudies by Simpson et al. (1968) revealedcloud clusters moving westward across theNorth Atlantic tradewinds during the tropi-cal storm season. Regardless of the circula-

tion features associated with the clusters,direct extrapolation of the clusters them-selves may help in forecasting changes incloudiness and rainfall along their tracks.Remember though, that cloud clusters oftenappear or disappear without any detectablereason because lack of observations preventthe accompanying circulation changes frombeing monitored, and that in the statisticalstudy of cloud clusters by Williams andGray (1973), 27 percent of the clusters disap-peared from one day to the next (Chapter 6,Section B-5). In using cloud prognosticcharts there is a tendency to over-forecastthe extent of active convection. This maysometimes be desirable, to compensate forerrors in forecasting movement and devel-opment of the cloud system.

3. Climatological Aids.

a. General. A thorough knowledge of clima-tology has been stressed throughout thisreport. Although tropical meteorologistsagree, many tropical stations lack climato-logical information in the formats bestsuited to analysis and forecasting. Theimportance of displaying monthly climato-logical resultant-wind maps for all levelsanalyzed has already been emphasized. Atmany stations, excellent climatologicaldisplays help forecasting and acquaintaircrews and other users with the localclimatology. Formats are many; a sample,based on data for Saigon (11°N, 107°E) isshown in Figure 12-1.

b. Conditional Climatology Summaries.These give the probabilities of variouscategories of some meteorological elementsoccurring at a range of future times basedon the values at the time the forecast ismade (the initial time). For example, havingCategory Xi at the initial time, a summarygives the probabilities of Categories X1, X2,X3 .... being observed Y hours later. Wheremarginal weather is uncommon, conditionalclimatologies (CCs) do not provide muchmore information than climatology. Al-

though most commonly applied to ceilingand visibility, conditional climatologies canbe constructed for any categorizable meteo-rological element — rain occurrences oramounts, cloudiness, thunderstorms, etc.Where convective rainfall dominates, CCs ofceiling heights help forecasting to someextent, but CCs of visibility are of limitedvalue. Samples of CC ceiling and visibilitysummaries for a station with a low cloudand fog regime (Danang - 16°N, 108°E) forDecember, and a convective regime (Saigon- 11°N, 107°E) for June are shown in Figure12-2.

Wind stratified CCs have been widely usedby AWS units in the extratropics and havebeen prepared for many tropical stations.Where wind direction and various levels ofmarginal weather are highly correlated,wind-stratified CCs can significantly aidforecasting. For example, at Clark Air Base(15°N, 121°E) when low-level winds shift tosouthwest, weather is likely to deteriorate.The wind-stratified CCs were designed toaccount for synoptic and local effects. Inmany tropical areas, large day-to-dayweather changes are often unaccompaniedby changes in wind direction. Also, windsover tropical continents tend to be lighterand more variable than in higher latitudes,while wind directions can be changed bynearby convection. These reasons limit theusefulness of wind-stratified CCs in thetropics.

c. Diurnal Pressure and Temperature Data.Since in the tropics diurnal variations oftemperature and pressure are normallymuch larger than interdiurnal changes, theirmean hourly values for month or season canhelp forecasting. Charts, such as thoseshown in Figures 7-4 and 7-5, or tables canbe readily prepared. In many cases, meanseasonal charts can serve (e.g., wet and dryseasons). Short-range forecasts of altimetersettings can be based on these tables, usingthe following formula:

Forecast Altimeter Setting = (Current Altim-eter Setting)

+ (Departure from the Mean at ForecastHour)

+ (Departure from the Mean at the Hour theForecast is Made)

Of course, adjustments should be madewhenever well-defined pressure systemsare forecast to affect the station. Diurnaltemperature variation can be treated simi-larly (Figure 7-3). Since cloudiness and raininfluence the magnitude of the diurnalcycles of temperature and pressure, a highaccuracy would require separate diurnaltemperature tables for rainy and fair days.When the pressure tables are combined withtables of diurnal temperature variation,forecasts of density altitude can be pre-pared. Guidance is given in AWS/TR-79/005.

d. Diurnal Variation of Weather. Monthlytables or graphs showing percent frequencyby hour of various weather elements shouldbe prepared for the local and representativestations of interest (see Figures 7-31 and 7-33). These presentations may include se-lected ceiling and visibility categories,rainfall, thunderstorms, fog, wind velocityand other categories. Knowledge of thisclimatology keeps forecasts within reason-able limits and indicates most likely occur-rence times. A sample climatological displayfor Saigon (11°N, 107°E) is shown in Figure12-3. Note the double maximum in thefrequency of visibility values below certainlimits. Comparing the visibility and rainfalland fog frequency curves reveals that fogusually reduces morning visibility, whileshowers or thunderstorms reduce afternoonand evening visibility. If fog is forecast, it ismost likely between 04 and 09 LT; visibilityquickly improves after 08 LT. The ceiling-frequency curves show that ceilings below

1000 feet (300 m) are most likely duringearly morning fog, while ceilings from 1000to 3000 feet (300 to 900 m) are commonaround midday due to convective clouds.Note that ceilings between 3000 and 10,000feet (900 and 3000 m) are rare during theday but occur occasionally at night, prob-ably after showers and thunderstorms havedissipated.

e. Sources of Climatological Data. If therequired data are not available locally, theycan usually be obtained from USAFETAC.Most station climatological data are in-cluded in “Revised Summaries of SurfaceWeather Observations” (RUSSWOs) pre-pared from hourly airways observations orthe “N Summaries” prepared from 3-hourlysynoptic observations (now referred to asSurface Observation Climatic Summaries).Much of this information appears in publi-cations such as the Weather and Climate orMarine Climate sections of the NationalIntelligence Surveys or the “WorldwideAirfield Climatic Data” volumes preparedfor various areas of the world (1967 —).Climatic atlases and other climatologicalpublications from various national meteoro-logical services contain data useful to fore-casting. The climatological bibliographiesprepared by NOAA/EDS and USAFETACcan be consulted to determine other sourcesof climatological data.

In addition, data from weather satellites andradar provide detailed climatologies for thetropics. Weather satellite climatologiesinclude total cloudiness (Sadler et al., 1984),highly reflective clouds (Garcia, 1985), andoutgoing long-wave radiation (Janowiak etal., 1985). Climatologies based on weatherradar measurements determine seasonaland diurnal variations in convection over anarea, and preferred areas for echoes todevelop. These help local-area forecastingand are discussed in detail in Section 6.

4. Local-Forecast Studies.

a. General. The most successful tropicalforecast studies use all available observa-tions (i.e., synoptic surface and upper-aircharts, satellite data, weather-radar picturesor reports, WBAN surface observations,cross sections, checkerboards, thermody-namic diagrams, climatology, etc.). Thus,units should try their hardest to obtain thesedata. Each station should retain at least thelast several years of weather observations,charts, and diagrams taken or preparedlocally. More weather data, including satel-lite data for wider areas and longer periods,should be retained by centralized forecast-ing units to support forecast-study andtraining. Almost complete destruction ofunequalled tropical data collected duringWorld War II severely handicapped tropicalmeteorology.

Forecast studies, regardless of results,should be better documented. This has beena serious problem for military meteorologyunits in the tropics, because of short toursand the resulting rapid staffing changes. Inthe past, useful techniques that were notdocumented or evaluated have been dis-carded. When an unsuccessful effort toimprove forecasting for a particular areawas not recorded, it was doomed to berepeated. Eliminating these deficienciesdemands careful control of planning, prepa-ration, evaluation, and documentation of allforecast studies.

It is beyond the scope of this report to sur-vey the range of local and area forecaststudies, whether successful or not. Thiswould be more appropriate for regionalmanuals on tropical analysis and forecast-ing. Two examples must suffice. Brooks(1987) found that a significant increase in N-S pressure gradient over the western Gulf ofMexico gave about 48 hours warning of awinter outbreak of northerly winds overHonduras. Applying this relationship de-mands no experience. Palecek (19??) related

occurrence of warm season thunderstormsat Clark Air Base (15°N, 121°E) to priorwind speeds in three tropospheric layers,low-level moisture and depth, an inversionbelow 10,000 feet (3 km) and a complicatedstability index. Eleven steps led to a forecast.Since synoptic changes had also to be con-sidered, success must have depended onforecaster experience. In this section, ex-amples of forecast studies are used in illus-trating and evaluating efforts to solve tropi-cal forecasting problems. These approachesare generally objective or synoptic andsometimes a combination of the two.

b. Objective Forecast Studies. Performanceof a purely objective forecast technique isindependent of the forecasting experience orthe skill of the meteorologist using it.

Wind forecasts. One of the most widelyapplicable objective techniques was used byLavoie and Wiederanders (1960) to forecastupper-winds in the tropics by combiningpersistence and climatology in variousways, based on the equations:

Uf = (1 - Rtu)Uc + RtuUp

Vf = (1 - Rtv)Vc + RtvVp

where U and V are the zonal and meridionalwind components and the subscripts f, c,and p denote the forecast, climatological,and persistence values respectively. The lag-correlation coefficients for the zonal andmeridional components are Rtu and Rtvrespectively. Accuracy is best when the lag-correlation coefficients are determinedstatistically for each level, season, and timeperiod. It turned out that for 24-hour fore-casts of 300 mb winds over the tropicalPacific, an average of the climatologicalwind and the persistence wind was almostas accurate as the optimum combination ofthe two. An evaluation made during May1959 showed the statistical forecasts to be 11

percent better than persistence and 15 per-cent better than climatology (Figure 12-4).The authors indicated that similar resultscould probably be obtained for other levelsabove the near-surface layer. This meansthat 24-hour wind forecasts for point loca-tions can easily be prepared by taking thevector average of the current wind (forpersistence) and the climatological wind.Reed (1967) used upper winds fromEniwetak (11°N, 162°E) in developing statis-tical wind forecast techniques, which hefound applied throughout the tropicalPacific. His paper can be consulted fordetails on the theoretical basis of the methodand for the lag-correlation coefficients forvarious levels and periods.

The method can be computerized to pro-duce statistical wind forecasts or to combinepersistence and climatology to use as first-guess fields in automated wind-analysisprograms. For wind forecasts near andequatorward of the subtropical ridges, thisstatistical technique is seldom surpassed byeither numerical or subjective techniques.Forecasters find this hard to believe. In a testthey consistently forecast too much changein the winds and were regularly beaten bythe statistics (Lavoie, 1963). The rule to liveby — this is a very good method. Don’tchange a persistence/climatology forecastunless the evidence is very strong. Since themethod always predicts a regression towardthe mean, it should not be used where greatvalue is attached to predicting large depar-tures from normal. For example, in theregion of the subtropical jet stream it couldgreatly reduce the strength of the jet core,possibly causing large route-wind forecasterrors.

Combinations of persistence and climatol-ogy have been successfully used to forecasttropical cyclone movement in Miami, Guam,India and Hong Kong and in terminalweather forecasting (McCabe, 1961). Persis-tence-climatology combinations (even in

higher latitudes) would test numericalpredictions much more effectively than theseparate persistence and climatology fieldsnow used. In Section C, 24-hour numericalpredictions of upper winds in the tropics arecompared to persistence/climatology fore-casts as well as persistence and climatologyforecasts.

Short-range forecasts of maximum surfacewind gusts (excluding those produced bylocal convection) can be helped by an objec-tive technique that relates observed surfacewinds to winds in the lowest 10,000 feet (3km). It depends on increased mixing andmomentum transfer during the time ofmaximum surface heating. Figure 12-5illustrates the relationship. It allows maxi-mum wind gusts to be estimated to within+/- 5 knots about 90 percent of the time.Similar relationships can be derived forother locations at which the surface windundergoes a pronounced diurnal variation.Statistically-based nomograms have beendeveloped for US bases in the westernPacific that are often affected by tropicalcyclones. The nomograms relate expectedsustained winds, average gusts, and peakgusts to tropical cyclone intensity, andbearing and distance from the base. Thus,local effects and distortions are allowed forin the nomograms.

Weather Forecasts. Objectively-derivedpredictors based on sound physical prin-ciples have the best chance of success andfor the relationships between predictors andforecasts to remain stable. For example, ithas long been known that subsiding airahead of tropical storms can produce a largearea of unusually fair weather (Sadler et al.,1968).

The dynamic and subtly complex nature oftropical weather systems limit the value ofobjective methods to tropical weather fore-casting. Even so, since there are few proventechniques for forecasting pressure, wind,temperature or moisture patterns, objective

methods must generally rely on persistenceof observed or derived elements, whereas inhigher latitudes forecast variables are oftenincluded in objective forecasts. An objectivetechnique usually works best when theelement being forecast experiences a largediurnal variation. So, in general the methodworks better over tropical continents thanover tropical oceans. Objective methodsmay be rather crude, often able to indicateonly whether an event will or will not occur.Forecasting the onset time, duration, and/orintensity must depend on other methods —the synoptic situation, climatology, etc.Thus, carefully developed objective aidsshould always be integrated with all theother aids the tropical forecaster has at hisdisposal.

c. Synoptic Studies. These can be dividedinto two categories: case studies and com-posite models. Synoptic case studies focuson the circulation and weather patterns overa specific area for a given period. Consecu-tive synoptic analyses at various levels areusually shown along with satellite imagesand auxiliary charts such as vertical time orspace cross sections, thermodynamic dia-grams, etc. Many of the synoptic modelsdescribed in Chapter 8 were developed fromsynoptic case studies. Composite synopticmodels average together individual synop-tic analyses or observations so as to producea “representative” model.

Case studies have preponderantly dealtwith weather extremes, such as floods,tornadoes or hail — rare events. They neverwork as well as one might expect. Thereason — the investigator knew not only theinitial situation, but also the outcome. Thus,he almost unconsciously filters and analyzesthe data, ensuring that everything comes outright in the end. Since filtering may distortperceptions of the synoptic sequence, anuncertainty is introduced that the forecastercannot allow for. A case study benefits

forecasting most if it represents typicalsituations recurring often.

In case studies, all available data should beintegrated into the analyses. Some excellentexamples are given in reports published bythe Naval Weather Research Facility(NWRF, 1969; Ramage et al., 1969). Thesereports present the results of several work-shops on the weather of southeast Asiaduring the northeast and southwest mon-soons. For each day of the case studies, twoadjacent large pages show large-scale sur-face analysis, local-area surface analysis,thermodynamic diagrams for key stationsand temperature-advection charts (duringnortheast monsoon), upper-air analyses,weather satellite pictures, and pilot-reportsummaries. This representation clearly linkscirculation and weather patterns and showstopographical influences.

Each tropical forecast unit should maintaina notebook of case studies for reference andtraining. For each case study, the file shouldat least show the synoptic analyses at a fewselected levels, satellite pictures, radarpictures, if available, and a record or sum-mary of surface observations. The period ofsignificant weather studied should start oneor two days before the weather affected thearea of interest and continue for at least oneday after it ended. It would thus encompassthe complete sequence, and show any rela-tionships between weather and circulationthat may be peculiar to the area. The frontis-piece samples data collected and analyzedby JTWC immediately after the eye of Ty-phoon Russ passed 30 NM (55 km) south ofGuam on 21 December 1990.

Composite synoptic studies can be preparedin several ways, but all combine a limitednumber of similar synoptic (weather) situa-tions and identify and quantify their mostsignificant features. The days included inthe composite may be determined by thecirculation patterns or by occurrence of acertain type of weather over a particular

area. Some examples of composite synopticstudies are given in the following para-graphs.

In studying the daily variation of convectionaround selected stations in southeast Asia,Conover (1967) developed a radar index (RI)that represented the percent echo coverageas determined from radar photographs orreports within 50 NM (90 km) of a station.Figure 12-6 shows RI values for Saigon(11°N, 107°E) for June, July and August,1967. Convection underwent large diurnaland day-to-day changes. After being aver-aged for each day, the RI values were sepa-rated into groups, with each group contain-ing one-third of the cases. Figure 12-7 showskinematic analyses of the composite meanresultant gradient-level winds for group 1(above average convection) and group 3(below average convection). On active daysthe monsoon trough is near its normalposition and Saigon lies in westerly windsthat are shearing cyclonically. Conversely,on inactive days, the monsoon trough ispoorly defined and winds are more south-erly over Saigon.

Atkinson and Penland (1967) developedcomposite models of maximum wind gusts,rainfall and flying weather at Kadena AirBase (26°N, 128°E) with respect to the loca-tion and intensity of tropical storms. Forexample, Figure 12-8 shows the probabilityduring a one-hour period of ceiling/visibil-ity less than 1500 feet and/or 3 miles (460 mand/or 5 km) and less than 200 feet and/or ahalf mile (60 m and/or 1 km) occurring atKadena related to the position of a tropicalstorm with respect to Kadena. Similar mod-els can be prepared for other stations andfor other weather categories.

Both objective and synoptic studies can helptropical forecasting. But there are limita-tions. Point forecasts of rainfall have littlechance of success. Forecasts of above, nearor below normal rainfall might be at-

tempted, and these are more likely to beright for an area than for a point.

5. Use of Stability Indexes.

a. General. The discussion of temperatureand water vapor in Chapter 5 pointed outthat changes in vertical profiles of equiva-lent-potential temperature (and hencechanges in stability) are often the resultrather than the cause of increased convec-tion. This implies that conventional stabilityindexes (SIs) may not help convection fore-casting much over tropical areas. Manytropical meteorologists who have had lim-ited or no success with SI would agree.Since the largest day-to-day changes invertical temperature and moisture distribu-tion occur over the tropical continentsduring the transition months between thewet and dry seasons, SIs are likely to bemost useful then. They are probably of leastuse over the tropical oceans where changesin mid-level moisture are often the resultrather than the cause of convection, andover the tropical continents during themiddle of the rainy season where stabilitychanges relatively little from day-to-day.

b. Stability-Index Studies. Sansom (1963)reported that over East Africa stabilityindexes gave no better forecasts than persis-tence, while at Darwin (12°S, 131°E) theyfailed to aid summertime thunderstormforecasting in an atmosphere that wasalways conditionally unstable (Hyson et al.,1964).

Subramanian and Jain (1966) compared theShowalter Index (Showalter, 1953) andLifted Index (Galway, 1956) as aids to fore-casting thunderstorms around New Delhi(29°N, 77°E) during the hot months (Marchto June) of 1963 to 1966. They compared theindex values from the 06 LT soundings tothunderstorm occurrence during the follow-ing 24 hours within distances of 50 and 100miles (93 and 185 km) as determined from

radar observations, and found (Figure 12-9)a relation between thunderstorm probabili-ties and SI values. Slightly better discrimina-tion between thunderstorm and non-thun-derstorm days is achieved by the LiftedIndex, especially for the large-negativecases. The authors failed to link SI values tothe maximum heights attained by thunder-storm cells. During the months studied,New Delhi is dry (March to June rainfall =4.41 inches (112 mm)) and a showers regimeprevails. Vertical temperature and moistureprofiles resemble those of the continentalmid-latitudes for which the indexes weredeveloped, but differ markedly from thoseof the tropical rainy season (see Figure 6-18).

c. Thunderstorm Wind Gusts. Not surpris-ingly, R.C. Miller (1967) found that south-east Asian thunderstorms resemble those ofthe southeastern United States, and that thestrength of thunderstorm downrush windgusts is a function of the lapse rate. Hedeveloped two indexes to quantify therelationship:

Dry-Instability Index (T1). This measures thelow-level instability and is determined intwo ways: (1) If the sounding has an inver-sion, the moist adiabat is followed from thewarmest point on the inversion to the 600mb level. The temperature difference be-tween the intersection of the moist adiabatat 600 mb and the 600 mb free air tempera-ture is T1. (The top of the inversion must bewithin 200 mb of the surface and is notexpected to be eliminated by surface heat-ing.) (2) If no inversion appears on thesounding (or if the inversion top is morethan 200 mb above the surface), the forecastmaximum temperature is lifted moist-adiabatically to 600 mb to determine T1 asbefore.

Delta-T Index (�T). This measures instabilityin the middle and upper levels and is deter-mined by lowering moist adiabatically theintersection of the wet-bulb curve and the0°C isotherm to the surface, then finding the

difference between the temperature at thispoint and the ambient surface temperature.

Miller compared wind-gust cases for south-east Asia and the southeastern United Statesto the simultaneous values of the T1 andDelta-T indexes (Figure 12-10). The roughbest fit is shown by the heavy dashed line,while the lighter dashed lines enclose theobserved maximum and minimum values.These curves probably apply to other partsof the tropics. Each station can determinewhich curve best describes the air-massvariations that produce thunderstorms withgusty winds. If thunderstorms are precededby intrusions of drier air between 700 and500 mb, the Delta-T index will probably bestestimate the wind-gust. If thunderstormsare generated by low-level temperatureincreases (surface heating or low-level warmadvection), then the T1 index is recom-mended. But remember, a thunderstormmust first be forecast and SIs are not muchgood for that. Also, both indexes require thesurface temperature at the time of the thun-derstorm to be forecast. To realize the fullpotential of either index, moderate or heavyrain is needed.

6. Radar.

a. General. Weather radar is the best aid tovery short-range forecasts (up to threehours) in areas dominated by convection. Itis used the same way in the tropics as inmiddle latitudes, and so the Federal Meteo-rological Handbook 1, and Whiton andHamilton (1976) are applicable. This sectionsupplements the guidance provided by theHandbook. As Doppler radar comes moregenerally into use, indirect methods forobtaining wind velocities from echo move-ments, as described below, will give way todirect methods that measure motion ofindividual drops.

Forecasters should remember that rain notonly reflects the radar signal but also attenu-

ates it. Rain behind rain always appearslighter than it really is, and often (especiallyin the farther eye-wall of a tropical cyclone)may disappear altogether from the radarscope (see Figure 9-5). Three-centimeterradars were operated simultaneously atUdon (17°N, 103°E) and Ubon (15°E, 105°E)on the Korat Plateau, Thailand, during thewet seasons of 1967 and 1968 (Ing, 1971).Figure 12-11 shows echo frequencies. Eachradar’s zero isopleth passes through theother’s maximum, while the line joiningisopleths of equal frequency passes midwaybetween the stations. Longer wave-lengthradars would do better, but attenuationwould still be a problem. Weather satellitepictures could help evaluate location andintensity of attenuation.

b. Tropical Storm Tracking. Although na-tional centers such as NHC and JTWCforecast tropical cyclone intensities andtracks, detailed forecasts of winds, rain andseas are still largely a local responsibility,and are based on continuous radar informa-tion. The literature on weather radar use inthe tropics emphasizes tracking tropicalstorms (see Figure 9-5). For tropical cyclone-prone islands and coasts, the cost of weatherradars is quickly amortized by the accurateand timely public warnings they provide.

The first radar indication of an approachingtropical storm will be one of the outerrainbands, which appears like a squall line.The distance between the rainband and thestorm’s eye can usually be determined froma recent satellite picture. This outer bandmay lie 50 NM (100 km) ahead of the firstspiral band of the storm proper. In the outerportions of the storm, spiral bands are about20 NM (40 km) apart; near the storm center,they tend to merge into the eye-wall cloud.The storm’s eye appears as a nearly circularecho-free region that is usually from 10 to 50NM (20 to 100 km) wide. Once 180° or moreof the eye-wall cloud appears on the radar-scope, the storm center can be located with

an accuracy of about 5 NM (10 km) or better.A well-defined eye may not always bevisible on radar, even when the storm iswithin range, because heavy rain betweenthe eye and the radar can severely attenuatethe signal. Some intense storms show twoconcentric wall clouds.

For storms whose eye is out of range or notdistinguishable on the radar, the eye posi-tion can be estimated by using a logarithmicspiral overlay (Senn et al., 1957). Therainbands in tropical storms approximateequiangular spirals that cross the isobars atangles ranging from 10° to 20°. A 15° spiraloverlay (Figure 12-12) is usually accurateenough. After the overlay is adjusted to fit aportion of a well-defined spiral band, thecentral dot of the overlay indicates thecenter of the storm. Jordan (1963a) com-pared spiral-overlay center positions deter-mined from 59 operational weather radarreports to the best-track (post analysis)center positions of hurricanes affecting theUnited States from 1959 through 1961. Theoverlays were often used when the centerwas more than 100 NM (200 km) from theradar, and rarely used when centers werewithin 50 NM (100 km) of the radar. Allspiral-overlay errors were less than 50 NM(100 km), and 60 percent of them were lessthan 25 NM (50 km).

Even though eye positions can be deter-mined quite accurately by land-based radaronce enough of the eye appears on thescope, these observations are of limited useto short-range forecasting. This is becausefrequent aircraft reconnaissance and radarfixes have revealed irregular small-scalefluctuations in the movement of eyes, mak-ing extrapolation very difficult. Jordan(1963b) discussed these fluctuations, whichrange from periodic oscillations with aperiod of about two days down to nearlyrandom fluctuations lasting less than anhour. An excellent example, from the trackof Hurricane Carla on 10-12 September 1961,

is shown in Figure 12-13. Because of suchirregularities, Jordan concluded that obser-vations separated by less than an hourcannot be relied on to indicate storm motionover the following three to six hours. Simi-larly, observations made a few hours apartare of limited use in indicating storm mo-tion over the following 12 to 24 hours. Thus,to avoid being misled, forecasters mustrelate frequent storm-center fixes to otherinformation such as the smoothed stormtrack during the past few days, steeringcurrents, climatology, etc. Six- to twelve-hour forecasts based on extrapolation ofsmoothed track positions of the previous 12to 24 hours will usually better those ex-trapolated from fixes made over the previ-ous several hours.

In tropical cyclone rainbands, individualcells move tangentially around the eye.New cells tend to develop on the upwindend of the spiral rainbands, become morestratiform farther downstream, and eventu-ally dissipate near the downwind end of theband (Hardy et al., 1964). Figure 12-14illustrates the movements of the rainbandsand of the individual cells relative to thestorm eye. The motion of a cell tracked forabout 15 minutes, either at the end of a bandor between bands, approximates the stormwind. Since in fully-developed storms thereis little wind shear between the near-surfacelayer and 20,000 feet (6 km) elevation, echomotion is close to lower-tropospheric windspeed. Wind speed varies considerablyalong a radius; thus, observations at differ-ing distances from the center are requiredbefore the radial wind-speed profile can bedetermined. Also, as illustrated by Figure 9-2, wind speeds are asymmetrically distrib-uted around the storm. In the lower tropo-sphere, strongest winds usually occur to theright (left) of the storm’s track in the NH(SH). Because of the earth’s curvature, winddetermination from

echo movement should probably be limitedto ranges of 100 NM (200 km) or less fromthe radar site. The Doppler radars nowbeing deployed will allow direct determina-tion of the winds from the motions of indi-vidual raindrops out to 125 NM (250 km).

c. Echo Movement. The same rules forforecasting echo movement in higher lati-tudes (Federal Meteorological Handbook I)apply in the tropics. The movement of singlecells (or small clusters that may appear asone cell on the radar) is closely related to thewind velocity at 700 mb (or to the meanwind between 850 mb and 500 mb). Succes-sive positions of small cells at 15- or 20-minute intervals determine cell motion.Averaging velocities of several cells gives arepresentative value, and overcomes thehandicap that individual cells generally lastless than 30 to 60 minutes. For forecastingrainfall at a point, only cells at least 30minutes upwind can be extrapolated withany confidence.

Tsonis and Austin (1981) analyzed 27 long-lived (> 100 minutes) convective cells duringGATE. They found that applying simplepersistence of movement and area gaveforecasts as good as more complex schemes,and that the best forecasts were made whenthe cell had already existed for 30 minutesand motion could be established. The besttwo-hour forecasts erred 77 percent in arealrainfall. This represents optimal perfor-mance; forecasts of shorter-lived cells woulddo worse, and forecasts over land, worsestill.

Tracing the movement of large groups ofconvective echoes 15 NM (30 km) or moreacross, and consisting of many thunder-storms or showers, rather than individualcells, helps prediction more because thegroups usually maintain their identity forhours. Unlike smaller cells, echo groups donot move with the wind at any one level oreven with the mean wind. The results ofConover’s (1967) study of echo groups

wider than 50 NM (93 km) near Saigon(10°N, 107°E) from June through August,1966, are presented in Figure 12-15. Dura-tions ranged from 1 to 20 hours, with mostbetween three and four hours (Figure 12-15A). In about two-thirds of the cases, echogroups moved in the direction of the windbetween 8000 and 13,000 feet (2400 and 4000m) (Figure 12-15B). For the level of bestagreement between the direction of echomovement and wind direction aloft, theratio of echo speed to wind speed at thatlevel was computed. Figure 12-15C showsthat ratios of 0.3 to 1.0 predominate. Theisolated cases of ratios greater than 1.3probably reflect the effect of large windshear and little change in wind direction inthe vertical. Over West Africa, where this iscommon (Eldridge, 1957), new convectionforms at the leading edge of the echo groupand dissipates at the trailing edge. Overall,this causes the group to move faster than theenvironmental wind. The wide range in therelationship between echo group move-ments and winds aloft suggests ignoring thewind, and instead plotting successive posi-tions of the centroids or boundaries of thegroup on the radar screen. Once a group hasformed, its movement is relatively persis-tent; therefore successive group positions (athalf-hourly intervals) can be extrapolatedfor short-range forecasting.

As in higher latitudes, radar echoes in thetropics are often organized into lines (seeChapter 8, Section D-2) that tend to move tothe right of the upper winds at an anglebetween 0 and 90°. The line moves at ap-proximately the speed of the component ofthe 700 mb wind normal to the line. At line-speeds below 20 knots, the line may moveslightly faster than the upper wind compo-nent due to new cell growth ahead of theline (see above). Where wind-shear isstrong, as in the West African rainy season,lines may propagate westward above sur-face westerlies. As for large echo groups,extrapolation of past positions most accu-

rately predicts line movement; however,when a line forms close upwind, upperwinds need to be used in predicting itsarrival.

d. Severe Weather Forecasting. Weatherradar allows forecasters to monitor severeweather (i.e., heavy rain and strongdownrush winds) caused by convection.Radars with an Iso-echoing capability canpinpoint the most intense thunderstormcells. The thunderstorm echo top can bedetermined from a range-height-indicator(RHI) scope or by tilting the radar antennauntil the echo disappears. On any convec-tive day, the potential maximum wind gustsas indicated by the Dry Instability (T1) orDelta-T Indexes (see the previous section)will probably occur in the most intensethunderstorm cells. However, since intensecells cover only a small fraction of the areaencompassed by a radar, the strongest gustsin that area will seldom be measured bysurface observing networks.

In a mature thunderstorm, strong down-ward-flowing currents spread out frombeneath the cloud base as a pseudo-coldfront. Therefore thunderstorm-producedshifts in wind-direction may occur up toseveral miles from the edge of radar echoes.The strongest downrush winds occur inmoderate to heavy rain beneath the cloudbase. These microbursts (see Figure 10-17)can threaten airfield operations, e.g., pro-ducing strong crosswind or shiftingheadwinds into tail winds during aircraftapproach or takeoff. Therefore, forecastersmust be alert to these hazards and establishreliable procedures to quickly advise pilotsof rapid wind changes.

When a line of thunderstorms approaches astation, the radar scope may show a thinband or line ahead of the thunderstorm line.This precursor convection develops alongdensity and convergence discontinuities atthe leading edge of the cold airdownrushing from the thunderstorms.

Rakshit (1966) studied several thin lines atCalcutta (23°N, 88°E), and concluded:

(1) A single thin line occurred only with athunderstorm (squall) line and preceded itby up to 15 NM (28 km).

(2) As the thin line passed over a station,the first gust was accompanied by a pres-sure jump and temperature fall, but therewas no rain.

(3) The thin line had a fine-grained struc-ture with no convective cells evident anddisappeared from the radar scope when theantenna was tilted by two to three degrees.Since thin lines do not always precedethunderstorm lines, their absence does notsignal absence of dangerous gusty winds.

A source of mid-tropospheric dry air isneeded before evaporative cooling from raincan cause strong downdrafts and lead tosqualls or microbursts. Thunderstormsembedded in a deep moist environment(rains regime) may persist for several hours(continuous thunderstorms), but lacking adry air source, cannot generate strongdowndrafts.

e. Visibility Forecasting. Radars that arecalibrated to make accurate reflectivitymeasurements can help in making roughestimates of visibility. Wilson (1968) usedrainfall and visibility measurements col-lected from the Atlantic City meso-netduring heavy rainstorms from August 1965to 1966 to compute a correlation of 0.85between visibility in km and rainfall rate Rin mm hr-1. The regression found by theleast squares technique is V = 12.5R-0.68. Thisresembles a regression determined by Atlas(1953) from drop-size distributions andliquid water content

information — V = 11.6R-0.63.

Rainfall rate and reflectivity values Z inmm6m-3 have been related (Battan, 1959). Z =200R1.6 for steady rain and Z = 486R1.37 forthunderstorm rain are based on average

values for many storms (Federal Meteoro-logical Handbook I); however, there isconsiderable variation from these relation-ships for individual cases. For example,Cheng and Kwong (1973) calculated Z =175R1.68 for summer at Hong Kong (22°N,114°E), while Woodley and Herndon’s (1970)equation for Florida was Z = 300R1.4.

From combining Wilson’s V-R relationshipand the Z-R relationship for thunderstormor steady rain, the following formula relatesvisibility to radar reflectivity: V = 266Z-0.5.This relationship, plotted in Figure 12-16,can be used to estimate the lowest surfacevisibility associated with the maximumradar reflectivity in a thunderstorm orsteady rain. The most intense thunderstormcells with reflectivity of 105 to 106 will lowervisibilities to about 1500 feet (460 m) whilein light rain (Z less than 104) visibility isusually 2 NM (3.7 km) or more.

f. Rainfall Estimates. Weather-radar datahave been used to estimate rainfall for riverstage and flood forecasts, soil moisture andtrafficability estimates, weather modifica-tion experiments, etc. Even though (at apoint) radar estimates when compared torain gauge measurements leave a lot to bedesired, they often give a better idea ofrainfall over an area than any practicablenumber of gauges distributed over the samearea. This complex problem is discussed insome detail in the Federal MeteorologicalHandbook I, and in greater detail in manyarticles and technical reports, some of whichare referenced in the Handbook. Proceed-ings of annual AMS Weather Radar Confer-ences and reports from the Illinois StateWater Survey contain excellent papers.

The SSM/I, now aloft on DMSP, can estimaterainfall at resolutions of 13,5 NM (25 Km)and +/- 0.2 inches (5 mm).

g. Radar Climatology. According to theFederal Meteorological Handbook I, radarclimatology provides:

(1) frequency of precipitation echoes overinaccessible regions such as mountain andwater areas,

(2) estimates of seasonal variations of rain-fall intensity,

(3) frequency of echoes over specific regionsas a function of hourly, monthly, or seasonalvariations and differing synoptic situations,

(4) locations of favored regions for echodevelopment,

(5) terrain effects on orographic rainfallduring different synoptic situations. Thefollowing paragraphs discuss some tech-niques for preparing radar climatologiesand present examples of radar climatologystudies. Doppler radar, such as WSR-88D,can be programmed to record radar scanson video tape, and these can then be playedback and the climatology compiled. Withother radars, the scope can be photographedby a video camera at specified time inter-vals. Time-lapse movies made from thesetapes are excellent forecaster training aids.For example, they can depict the diurnalvariation of convection around the radarstation, or the distribution of rain around anapproaching tropical storm. Great care mustbe taken to insure that the desired rangesettings, power adjustment, antenna eleva-tions, etc., are properly recorded (along withthe observation times) on the films. Toincrease the climatological usefulness of thedata, the control settings should be stan-dardized.

Despite hundreds of radars being deployedin the tropics, only a few dozen radar clima-tology studies have been made. The reason— at least until the advent of video — is thatthe work is unexciting and time-consuming.Kulshrestha and Jain (1968) developed aradar climatology for the New Delhi (29°N,77°E) area of India (Figure 12-17). Figure 12-17A shows that echo tops are highest duringthe summer monsoon (June through Sep-tember) and lowest in winter. The diurnal

variation of the area of maximum radar-echo varies annually (Figure 12-17B). Possi-bly because of its prevailing fine weather,November is the only month that does notseem to fit the annual cycle. Figures 12-17Cand D show the frequency distribution ofindividual echo widths during the hotseason (March through May). The averageecho was about 3.9 NM (7.2 km) wide andthe average space between echoes was about9.5 NM (17.7 km).

Gerrish (1971) used the University of Miamiradar in developing a radar climatology ofsouth Florida summer rain. Figure 12-18shows some of the main results. The fre-quencies of echoes over land and water areshown in Figure 12-18B. Echoes are mostcommon over the sea at night and over landduring the day, while the tallest echoes areconfined to land. The areal/time patternsshown in Figure 12-18C provide a newmodel for summer convection in the Miamiarea. Over the ocean away from the coastthe pattern is cellular, implying largelyunorganized or random convection; how-ever, near the coast and over land, patternsare organized and trends may persist for aday. Separate maxima appear near sunrisein zones centered about 15 NM (28 km)apart and seem to move inland in unison.Wind data suggest that line 2, at least dur-ing part of its history, is associated with thesea breeze (see Figure 7-12); however, theoverall pattern with multiple lines of maxi-mum frequency is much more complex thana model of a single sea-breeze convergenceline. Figure 12-18D shows the frequencydistribution of echo heights under disturbedconditions. A pronounced mode is foundbetween 18,000 and 20,000 feet (5.5 and 6km) and relatively few echo tops above30,000 feet (9 km). There is little differencebetween land and sea.

C. Tropical Numerical WeatherPrediction.

Numerical weather prediction (NWP) inmid-and high-latitudes has slowly im-proved. At NMC, atmospheric models haveevolved from an initial, single-levelbarotropic model to the 12-layer globalspectral model. Although this model in-cludes the tropics, its skill is only marginallybetter than persistence at 24 hours. TheECMWF global spectral model also hasproblems in the tropics. Heckley (1985)reported that between 18°N and 18°S theRMS 24-hour wind forecast error amountedto 12 knots at 200 mb and 6 knots at 850 mb.These compared to 18 knots and 7 knots inpersistence forecasts. But remember, this is“persistence” of the NWP analyzed fieldsthat in turn depart as much or more fromreality (Chapter 11, Section D-3). Refiningconvective parameterization and increasingresolution between 1984 and 1985 (Tiedtkeet al., 1988) failed to improve the 24-hourwind forecasts, but did give better resultsbeyond two to three days.

Section B-4 noted the value of persistence/climatology forecasts (Figure 12-4) andsuggested comparing them to numericalpredictions. ETAC did this for 32 tropicalrawinsonde stations for March and April1991. The NMC operational NWP made 24-hour wind forecasts at 00 and 12 UTC at 850and 200 mb for each station location. Half

(persistence + climatology) forecasts as wellas persistence and climatology forecasts,were also made; all were compared to valid-time observed winds at each station (Figures12-19 and 12-20). The averages of the aver-age station errors are shown below in knots.From a sensitivity analysis————————————————————————————————-NMC 1/2(P + C) P C————————————————————————————————-850 mb 10.0 8.8 10.2 10.3200 mb 19.3 20.1 22.4 26.1————————————————————————————————-

ETAC concluded that error differences of 0.2knots at 850 mb and 0.4 knots at 200 mbwould be significant.

At 850 mb, 1/2(persistence + climatology) issuperior, while simple persistence or clima-tology forecasts are almost as good as thenumerical predictions. At 200 mb numericalpredictions are best, but even here, theimprovement of 3 knots over persistenceforecasts would hardly justify waitingseveral hours for a numerical product.

Table 12-1 supports the common assump-tion that numerical prediction skill is partlyan inverse function of latitude. At 850 mb, 1/2(P +C) is best at all latitudes, but its superi-ority to NMC diminishes with latitude. At200 mb, 1/2(P + C) is best between 0 and 10°,while NMC is 4 knots better between 20 and30°.

Table 12-1. March and April 1991 tropical wind forecasting. Average errors (knots) of 24-hour forecasts at 850 and 200 mb made by NMC, 1/2(persistence + climatology), persistence,and climatology; stratified by latitude.Latitude Number of NMC 1/2(P + C) P C Stations850 mb0 - 10° 7 10.0 7.8 9.3 8.610 - 20° 15 9.7 8.8 10.0 10.320 - 30° 10 10.4 9.6 11.1 11.5

200 mb0 - 10° 7 16.8 15.1 17.5 17.710 - 20° 13 19.3 21.0 21.3 25.320 - 30° 10 21.1 25.3 27.3 33.0

These comparisons confirm the value ofpersistence/climatology forecasts in thetropics. 1/2(P + C) may not be optimal for24-hour wind forecasts; the best proportionof persistence and climatology could de-pend on altitude, location and season, and isworth determining. The best combinationsfor wind forecasts beyond 24 hours (pre-sumably with the climatology contributionincreasing with forecast time) should also becalculated. Besides being operationallyvaluable, such statistical forecast modelsshould be used to test new numerical pre-diction techniques.

Numerical weather prediction models areusing smaller and smaller grid spacing,more and more levels in the vertical, andshorter and shorter iterative time intervals.For forecasts to benefit, the number ofobservations must also increase. Satellitesare being used (TIROS Operational VerticalSounders -TOVS) to estimate the verticaldistribution of mixing ratio. Tests haveshown that when this information is in-cluded in the input of the ECMWF numeri-cal model, rainfall forecasts in the tropicsare slightly improved (Illari, 1989). Betterresolutions attainable after 1993 with thenew NOAA satellites may help more.

We still lack operational NWP models forthe tropics. This means that in the nearfuture, improvements in tropical analysisand forecasting will depend on subjectiveand statistical techniques.

DDDDD. Medium and Long-Range. Medium and Long-Range. Medium and Long-Range. Medium and Long-Range. Medium and Long-RangeFFFFForecastingorecastingorecastingorecastingorecasting.....

1. General. For this report, medium-rangeforecasts are those made for periods from 36hours up to one week in advance, whilelong-range forecasts are for longer periods.Medium-range forecasts may specify gen-eral cloudiness and rainfall, while long-range forecasts are less specific, e.g., rainfall

above or below normal during a certainperiod, or a rainy season starting or endingearlier or later than normal. In the tropics,little progress has been made in eithermedium-range forecasting (MRF) or long-range forecasting (LRF).

2. Medium-Range Forecasting. In parts ofthe NH subtropics during the cool season,some MRF skill may be achieved whereweather can be linked to circulation featuresin mid-latitudes. For example, over south-east Asia, the occurrence and timing ofnortheast monsoon surges can be antici-pated from numerical weather predictionsof upper trough passage across Siberia.Similarly, most of the heavy rain along theeast coastal regions of Central Americaduring winter stems from fronts or shearlines; forecasts can be made several days inadvance on the basis of increasingly reliablenumerical upper-air and surface prognosesover North America.

Thirty- to Sixty-Day Cycle. In 1971, Maddenand Julian detected a 40- to 50-day oscilla-tion in the zonal winds over the tropicalPacific. During FGGE, in 1979, the oscilla-tion was well-marked and could also befound in cloud systems. Much researchresulted that included as yet unconfirmedhypotheses on the origin and strength of theoscillation. Knutson and Weickmann (1987)list the following characteristics of theoscillation. It:

(1) comprises global-scale tropical windand convection anomalies;

(2) is not strictly periodic but has a pre-ferred time scale of about 30 to 60 days;

(3) tends to propagate eastward at 6 to 12knots over the tropical Indian and westPacific Oceans;

(4) has zonal wind anomalies in the lowertroposphere that are out of phase with thosein the upper troposphere;

(5) occurs throughout the year.

The oscillation generally appears to start inthe central equatorial Indian Ocean as alarge convective mass that spawns smallereastward-moving systems. Rather unexpect-edly, the main cloud mass starts eastwardacross Indonesia and the western Pacific; itsintensity fluctuates, but at least in hindsight,it can sometimes be followed around theglobe. Figure 12-21 shows composite OLRanomalies accompanying passage of anoscillation in the equatorial region. Monthsmay pass without an oscillation. At anytime, convective clusters may move east-ward along the equator at synoptic-scaleintervals (Hayashi and Nakasawa, 1989)and so hide or distort the 30- to 60-dayoscillation.

During the NH summer, other oscillationsmove northward from the equatorial IndianOcean, taking about 20 days to reach 25°N.These can be linked to the monsoon cycleover India, where “active” and “break”conditions alternate. The ridge movingnorthward in Figure 8-58 is probably part ofthis sequence.

It is one thing to recognize passage of anoscillation after the fact, and another toanticipate it. Forecasters should wait untilmotion has been well established beforeextrapolating and predicting. Even then,subsequent variations in speed and intensityand interaction with shorter-period oscilla-tions are likely, and the forecast shouldincorporate such a range. The west-eastoscillations are strongest between the centralIndian Ocean and the date line, and thesouth-north oscillations between the centralIndian Ocean and south and southeast Asia.

Otherwise, during summer, about all thatcan be attempted for MRF is extrapolationof existing circulation and weather systemsmodified by climatology. To make accuratemedium-range forecasts in some tropical

areas, development and movement of tropi-cal cyclones must be forecast days in ad-vance, a feat well beyond the skills of today.

Meteorologists constantly seek patternsbesides the astronomically-controlled an-nual and diurnal cycles. Since FGGE, thesearch has intensified in the tropics; most ofthe seekers agree that the 30- to 60-day cycleis real. However, rigorous analysis hasquestioned its statistical basis (Maliekal,1990). Forecasters should suspect any claimfor a periodicity that is based on a relativelyshort time-series of data that have beenaveraged or filtered in some way. Remem-ber, a periodicity that is not visually obviousin a long time-series plot of the raw dataprobably does not exist.

3. Long-Range Forecasting.

a. General. Flohn (1961) concluded that welack: (1) useful, practical techniques forLRF; (2) a fundamental and physicallysound concept of the large-scale weatheranomalies for LRF; and (3) an evaluation ofthe climatological data needed for anyfundamental research directed toward LRF.He cautioned against assuring users thatLRF is possible in the tropics or that success-ful techniques could be developed within afew years. Despite some research in the pastseveral decades, Flohn’s view still holds.

Since dynamic techniques show little or noskill in making short-range forecasts in thetropics, it is easy to see why they have notbeen applied to LRF problems. Almost allattempts to make tropical LRF have usedstatistical techniques where the predictand(element to be forecast) is correlated tomany predictors (usually observed values ofa variety of weather elements) at variouslocations.

b. Indian Summer Rainfall. The most fa-mous attempts at LRF in the tropics weremade by Sir Gilbert Walker and his disciples

who tried to forecast seasonal rainfall inIndia using predictors from stationsthroughout the world. Walker (1924) com-puted several correlation

coefficients between Indian seasonal rain-falls and possible predictors. He selectedonly a few pairs that appeared to be signifi-cantly correlated. The most important pre-dictions were for the summer monsoonrainfall of Peninsula and Northwest India,and for these the following predictors wereselected:Area ElementFor the PrecedingPeninsulaSouth American Pressure Apr-MaySouth Rhodesia Rain Oct-AprDutch Harbor Temperature Dec-AprJava Rain Oct-FebZanzibar Rain MayCape of Good Hope Pressure Sep-NovNorthwest IndiaSouth American Pressure Apr-MaySouth Rhodesia Rain Oct-AprDutch Harbor Temperature Mar-AprEquatorial Pressure Jan-MayHimalayan Snow Depth May

These predictors gave a multiple correlationcoefficient of 0.76 for both areas for thedependent data sample. Rao (1965) summa-rized Walker’s work and later attempts atseasonal forecasting in India. He found thatthe correlation coefficients of most of theinitial predictors chosen varied greatly fromone decade to another. Figure 12-22 showsthe variation with time of the best of theoriginal predictors, South American pres-sure.

This instability in Walker’s initial formulaehas caused them to be periodically revisedsince 1924. Showing the opposite trend toSouth American pressure, the correlation ofMarch, April, May surface air temperatureover western central India with Indiansummer rainfall changed from slightlynegative to significantly positive around1940. Parthasarathy et al. (1990), who re-ported this, used dependent data for 1951 to1980 to show that a relatively warm lower

and middle troposphere over north andcentral India is followed by good monsoonrain. Whether such new relationships,summarized by Hastenrath (1990b), arestatistically stable and help forecasts ofseasonal rainfall remains to be seen.

Even though unsuccessful in LRF, Walker inthe process discovered various global atmo-spheric oscillations or tendencies for pres-sure anomalies of opposite phase. Bestknown, the “Southern Oscillation” is a largenegative correlation in surface pressureanomalies between Djakarta (6°S, 107°E),Indonesia and the eastern South Pacific. Theassociated El Niño and the possibilities itraises for LRF are discussed in Chapter 6,Section C-4 (see Figure 6-25). Meteorologicalcenters should receive monthly ClimateDiagnostics Bulletins from the NMC Cli-mate Analysis Center (1989). If a Bulletinindicates that El Niño or the opposing anti-Niño is well established, this informationshould be passed to the field, where thestatistical rainfall relationships shown inFigure 6-25 can be applied to making LRF.Even here, forecasters should be cautious. Inthe past, some apparently well-establishedNiños have suddenly and inexplicably quit.A vigorous Niño, although favoring synop-tic developments that will ensure its persis-tence, does not guarantee them; should theybe absent, then, as in 1975 and 1990, El Niñoand its associated weather anomalies canquickly fade.

c. Atlantic Tropical Storms. Gray (1984a andb), using data from 1950 through 1982,discovered statistical relationships betweenthe Atlantic tropical cyclone season andpreceding (by the end of May) informationon the quasi-biennial oscillation (QBO) (seeChapter 4, Section C-5), El Niño, and aver-age sea-level pressure over the Caribbean inApril-May. For the latter two linkages, hegave reasonable physical explanations. (1)During the QBO, Atlantic tropical cyclones

are more frequent when 30 mb winds arewesterly and increasing, rather than easterlyand increasing. Gray gave no explanation.(2) During El Niño, high SST over the east-ern Pacific causes more deep convectionthere. The resultant outflow aloft enhancesupper tropospheric westerlies over theCaribbean and western equatorial Atlantic.Consequently, the 200 mb anticyclonic flownecessary for tropical cyclones to develop isreduced. (3) During the hurricane season inthe Caribbean basin, below normal monthlymean sea level pressure is associated withincreased hurricane activity. Pressureanomalies tend to persist from springthrough summer. The three predictors arenot correlated with one another, and so canbe combined using multiple regressionanalysis to provide forecast equations thathave more predictive skill than forecastsmade using individual predictors.

Gray developed prediction equations for (1)number of hurricanes, (2) number of hurri-canes and tropical storms, (3) number ofhurricane days, (4) number of hurricane andtropical storm days. He made seven hurri-cane season predictions in late May 1984through 1990 and updated them in late July.Performances of the May forecasts for num-ber of hurricanes and tropical storms andnumber of days with hurricanes and tropi-cal storms are shown in Figure 12-23. TheJuly forecasts were somewhat better. For thefirst five years, results were encouraging,but 1989 changed the picture. The RMSerrors of the forecasts differ little from thoseof climatological forecasts. Reminiscent ofthe Indian experience, Gray (1990) incorpo-rated west Sahel rainfall in his predictionsfor 1990.

d. Other Forecasts. During the tropicalcyclone season in the Australian region,Darwin (12°S, 131°E) surface pressureanomalies are inversely related to the num-ber of tropical cyclone days. Since Darwinwinter pressure anomalies tend to persist,

Nichols (1985) related them to subsequenttropical cyclone days by the regressionequation:

Cyclone days = 224.5 - [11.6 (July to Sept.pressure) - 1000 mb]

For northeastern Brazil, Hastenrath (1990c)found that half of the interannual variabilityof March-September rainfall can be pre-dicted from the rainfall of the precedingOctober-January. The persistence predictorsof Nicholls and Hastenrath are linked to thetendency of Southern Oscillation or El Niñoanomalies to persist. This does not alwayshappen.

For some years, May to October rainfall inHong Kong (22°N, 114°E) was fairly success-fully forecast in the previous February. Thisresulted from the January surface pressuregradient between Irkutsk (52°N, 104°E) andTokyo (36°N, 140°E) being highly negativelycorrelated with the subsequent Hong Kongrainfall (Bell, 1976). There was no physicalexplanation. As Figure 12-22 shows, thecorrelation varied greatly since 1890. For the15 years ending in 1988 it had fallen to -0.58(Lau and Chan, 1990). Many other simpleand multiple regression techniques to fore-cast summer rainfall in Hong Kong havebeen developed and tested (Lau and Chan,1990). All give forecast errors of about 12inches (300 mm), or more than 15 percent ofthe normal summer rainfall of Hong Kong.The extremely wet summer of 1982 wasunderforecast, on average, by 30 inches (755mm). Lau and Chan agreed with Reynolds(1978), who concluded that a regressionequation with a correlation of less than 0.90is unlikely to produce useful forecasts.

One of the major problems in deriving LRFtechniques is that data are rarely, if ever,sufficient to allow a technique not only to bedeveloped but to be tested on a large sampleof independent data. For the present, tropi-cal meteorologists should rely on climatol-

ogy when they prepare long-range estimatesof future weather conditions.

In summary, consistently skillful medium-range or long-range forecasts have not yetbeen achieved in the tropics. As a substitute,carefully tailored applied climatology canusually provide sufficient information foroperational planning. Thanks to more data,especially from satellites, meteorologistscan now make much better estimates. Thor-ough knowledge of diurnal variations andlocal influences on weather can also beincorporated.

EEEEE. P. P. P. P. Preparing to Freparing to Freparing to Freparing to Freparing to Forecast in a Neworecast in a Neworecast in a Neworecast in a Neworecast in a NewArea.Area.Area.Area.Area.

Air Weather Service operates few tropicalforecast offices.

Future military emergencies in the tropicsmay suddenly demand forecasts for placesthat lack any but rudimentary publishedinformation on local and areal weather.What has been discussed in this report canbe used to improve the chances of makingsuccessful forecasts. The work should firstconcentrate on the immediate season, andthen move (ahead of the calendar) to succes-sive seasons. Proceed to:

(1) Collect large-scale topographic maps ofthe area of interest (AOI). From the CDROMDefense Mapping Agency computer disks,plot 3-D representations of the AOI terrain(see Figure 7-38). On a larger scale, and fordifferent aspects, plot views of the terrainsurrounding key locations, such as airfields,staging zones, or ports. The 3-D images cangreatly aid understanding of orographiceffects and local circulations.

(2) From the latest Climate DiagnosticsBulletin, published by the Climate AnalysisCenter, determine the stage of El Niño/Southern Oscillation (ENSO) cycle and itsprognosis. From Figure 6-25 decide if the

AOI is likely to be affected. If it is, makesure that weather satellite and other databeing analyzed come from earlier periods atthe same stage of the ENSO cycle.

(3) Collect all available climatological datafor the AOI and its environment fromUSAFETAC. Most will comprise meanmonthly and annual rainfalls. Data fromanalogous areas might be used, for example— eastern Madagascar, eastern north Luzonand northern Central America; coastalAngola and coastal Peru. Sea surface tem-perature and tropical storm tracks andfrequencies would be useful.

(4) Collect satellite-based monthlyclimatologies of cloudiness (Sadler et al.,1984), frequency of highly reflective clouds(Garcia, 1985), and outgoing longwaveradiation (Janowiak et al., 1985). Lightningclimatologies may soon be available. Fromthem, deduce mean monthly cloudiness andprecipitating clouds at key locations in theAOI. Compare these values with similarly-determined values for stations with rainfallclimatologies, and deduce the AOI rainfallclimatology. From this, estimate frequencydistributions of daily rainfalls using Tables6-4 and 6-5, the coefficient of variation ofmean monthly rainfall using Figure 6-37,and the chances of rain of different intensi-ties using Table 6-6.

(5) From at least a month of geostationaryIR images, determine the diurnal variationof deep clouds and their favored locationswithin the AOI. Construct distributioncharts (see Figure 7-39) and link to the 3-Dtopographic representations. If time presses,use only near-dawn and mid-afternoonimages to encompass the range of the varia-tion. Combine IR and visible images todetermine locations and diurnal variationsof fog, stratus, dust, land-sea breezes, moun-tain-valley winds and wind channeling.

(6) Using the same weather satellite images,select a sequence of near-midnight and near-

midday views in order to reduce or elimi-nate small-scale local distortions due todiurnal variations. This will help the meteo-rologist identify and follow synoptic andmesoscale disturbances and determine theirclimatologies.

(7) Deduce the likelihood of significant lowlevel jets from estimates of surface tempera-ture gradients.

(8) As time permits, extend the analyses tosubsequent seasons (ahead of the calendar).

In anticipation of urgent requirements,USAFETAC should maintain theclimatologies mentioned in (4) above, ontapes as well as in Atlases. It should collect,and then extend, at least three years’ of allavailable geostationary satellite images, inhigh quality hard copy, and also in videoform for quick reference. The excellentclimatological files should be often updated,and should be enhanced by unpublisheddata from remote places.

FFFFF. Summary. Summary. Summary. Summary. Summary.....This report broadly surveys our knowledgeof tropical meteorology. Lack of publishedresearch for South America and of recently-published research for Africa has unavoid-ably imposed a regional bias. The firstedition covered the advances made since thetropical manual written by Palmer et al.(1955).

The second edition has taken advantage oftropical research conducted during GATEand FGGE, and including MONEX. Greatlyimproved climatology derived from satellitedata and from growing interest in climatechange has made this basis of tropical me-teorology firmer than before. Despite afocused effort, progress in understandingand predicting tropical cyclones has beenslow; progress elsewhere in the tropics,starting from much lower levels, has beenmore encouraging. To keep abreast of devel-

opments, tropical meteorologists mustperiodically review professional journalarticles, technical reports and other litera-ture.

More progress has been made in diagnosingand understanding structure of the tropicalatmosphere than in accurately predictingtropical circulation and weather patterns. Inthe tropics, and especially between 10°Nand 10°S, following a weather system fromone synoptic chart to the next is often pos-sible only in retrospect. A surge is discov-ered after it has affected the weather. Asquall line may be distinguished withinpatternless convection only after it starts tomove, and then the chances of it persistingcan seldom be foretold. As a result, tropicalforecasting is still handicapped. Thus, thebest possible meteorological support mustbalance requirements and limitations. Tropi-cal meteorologists should concentrate ontechniques that are most likely to showforecasting skill. The following limitationsshould be clearly recognized:

(1) It is often possible to make an arealforecast of whether and where convectionwill be above or below normal; but it israrely possible to pinpoint where and whenrain will fall 30 minutes or an hour in ad-vance.

(2) The convective, cellular nature of tropi-cal rainstorms makes forecasts of pointrainfall very unreliable.

(3) Rarely, when large-scale circulationsbecome persistent (such as El Niño) somemedium- and long-range forecasts can bemore skillful than chance, but for most ofthe time, over most of the tropics, medium-and long-range forecasts are useless.

(4) Tropical forecasting is handicapped byobservation and forecast codes, which weredesigned for higher latitudes and havedeficiencies when used in the tropics. Fore-casts tend to be too stereotyped for manytropical stations or areas and there is exces-

sive use of INTER or TEMPO groups. Thisneeds major study.

Good analyses and diagnoses of the causesand precursors of present weather areprerequisite to accurate forecasts. Manytimes, however, the forecaster will be unableto account for some weather patterns.Hence, forecasting changes in these patternswill often fail. In general, weather systemsthat form and dissipate without moving arethe most troublesome. Also, disturbances inthe low-level wind speed field, associatedwith horizontal shear or convergence inzonal currents, are trickier than disturbancesassociated with shifts in wind direction,such as vortexes or troughs.

Forecasting fine weather is easier thanforecasting unsettled weather because of theformer’s larger time and space scales ofdivergence and vertical motion. The fore-casts could reflect this by assigning higherprobabilities when fine weather is expected.From another viewpoint, the rarer the me-teorological event, the harder it is to forecastand the more important it is to the forecastuser. Generally speaking, tropical forecast-ers must work harder than mid-latitudeforecasters to achieve equal skill. Althoughthe tropics lack good numerical weatherpredictions, meteorologists are makingbetter wind forecasts because of more obser-vations, better communications, and objec-tive wind analysis and forecast programs.Short-range forecasts based on these analy-ses are usually accurate enough for aircraftoperations since large-scale tropical circula-tion patterns change slowly.

Forecasting should be an integral part ofmeteorology, because it provides the acidtest of hypotheses and is a prolific source ofnew ideas. Forecasters should also be re-searchers and researchers should test theirfindings by using them in forecasting. At theNational Hurricane Center in Miami and toa less degree at the Joint Typhoon WarningCenter in Guam, forecasters can undertake

applied research in the “off-season”. Duringsummer 1990, when a field program tostudy tropical cyclone movement was basedat JTWC (Elsberry, 1989), researchers regu-larly forecasted, and continually interactedwith forecasters. In China, most forecasterswork shift for three months and then spendthe next three months on research. Researchmeteorologists instruct operational meteo-rologists and routinely work forecast shifts(Grice et al., 1986).

The Stormscale Operational and ResearchMeteorology (STORM) program, plannedfor the 1990s over the United States (STORMProject Office, 1988), includes experimentalforecast centers. In these, forecaster/re-searchers will use forecasting in applyingthe scientific method to solving meteoro-logical problems.

Each of the tropical synoptic models de-scribed in Chapter 8 was developed from afew case studies and so tends to be unrepre-sentative. In a case study, the end as well asthe beginning is known, and the range of themodel and its exceptions — the statistics —are usually unknown. The forecaster usingthe model has the time neither to report onits value, nor to modify it on the basis of hisexperience. An Experimental Forecast Cen-ter would overcome these problems andcontinually correct and refine the modelsfollowing the procedure illustrated in Figure12-24.

In most tropical forecast offices, meteorolo-gists are fully occupied on shift, with littletime to tackle the host of local problemscontributing so much to forecast errors.Great distances separate forecasters fromresearchers, who are thus deprived of evalu-ations and cut off from new ideas. Even so,on a very modest scale, turning tropicalforecast offices into Experimental ForecastCenters would do more for forecasting skillthan would scores of isolated investigatorsworking on what they thought were impor-tant problems.

ABBREVIATIONS and ACRONYMS

ASDAR Aircraft to satellite data relay

ATEX Atlantic Tradewind Experiment

ATS Applications technology satellite(NASA)

AWN Automated Weather Network

AWDS Automated weather distributionsystem (AWS)

AWS Air Weather Service

AWSTR Air Weather Service TechnicalReport, now AWS/TR

BOMEX Barbados Oceanographic andMeteorological Experiment

CAT Clear air turbulence

CC Conditional climatology

CDROM Compact disc read only memory

CI Contingency index

CIRES Cooperative Institute for Research inEnvironmental

AB Air Base

AFCRL Air Force Cambridge ResearchLaboratories (USAF),

became Air Force Geophysics Laboratory(AFGL) and

then Geophysics Laboratory (GL)

AFGWC Air Force Global Weather Central(USAF)_

AIAA American Institute of Aeronauticsand Astronautics

AIREP Aircraft report

AMEX Australian Monsoon Experiment

AMS American Meteorological Society

AMSU Advanced microwave soundingunit

ANMRC Australian Numerical Meteorol-ogy Research Center (now part

of the Bureau of Meteorology ResearchCenter —BMRC)

AOI Area of interest

Sciences

cm centimeter (= 0.39 inches)

CNES Centre National D’Etudes Spatiales

DMSP Defense Meteorological SatelliteProgram

DOD Department of Defense

D-value Difference between the “stan-dard” height of a

pressure surface and its actual height

ECMWF European Center for MediumRange Weather Forecasts

EDS Environmental Data Service (NOAA)

EMEX Equatorial Mesoscale Experiment

ENSO El Niño Southern Oscillation cycle

EOS Earth Observing Satellite (NASA)

EOSAT Earth Observation Satellite Com-pany

ERS Earth Remote Sensing

ESA European Space Agency

ESSA Environmental Sciences ServicesAdministration

(absorbed by NOAA in 1970)

EUMETSAT European meteorologicalsatellite

FAA Federal Aviation Administration

FGGE First GARP Global Experiment

FLR Fractional low radiance

FNOC Fleet Numerical OceanographyCenter

GARP Global Atmospheric Research Pro-gram

GATE GARP Atlantic Tropical Experiment,1974

GMS Geostationary meteorological satel-lite (Japan)

GOES Geostationary operational environ-mental satellite

(NASA/NOAA)

GTS Global telecommunications system

HARP Hawaiian Rainband Project

HK Hong Kong

ICAO International Civil Aviation Organi-zation

IIOE International Indian Ocean Expedi-tion

INSAT Geostationary Indian nationalsatellite

INTER Intermittent (in WMO TerminalAirfield Meteorological

Code)

IRIG Inter-Range Instrumentation Group

ITCZ Intertropical convergence zone

IWRS Improved weather reconnaissancesystem

JTWC Joint Typhoon Warning Center(USAF and US Navy)

K Degrees Kelvin = °C + 272.16

km kilometer (= 3281 feet; 0.54 nauti-cal miles)

LIE Line Islands Experiment, 1967

LLJ Low-level jet

LRF Long-range forecasting

LT Local time

m meter (= 3.28 feet)

mb millibar

METAR Meteorological aviation report

METEOR Meteorological satellite (USSR)

METEOSAT Geostationary Meteorologi-cal satellite (ESA)

MIT Massachusetts Institute of Technology

mm millimeter (= 0.039 inches)

MONEX Monsoon Experiments of FGGE

MOS Marine observations satellite (Japan)

MRF Medium-range forecasting

NASA National Aeronautics and SpaceAdministration

NASDA National Space DevelopmentAgency (Japan)

NAWAC National Weather Analysis Center

NCAR National Center for AtmosphericResearch

NECZ Near-equatorial convergence zone

NESS National Environmental SatelliteService (now part of

NESDIS)

NESDIS National Environmental Satellite,Data, and Information

Service

NEXRAD Next generation weather radar,now WSR-88D

NH Northern Hemisphere

NHC National Hurricane Center (NOAA)

NM Nautical mile (= 1.85 kilometers)

NMC National Meteorological Center(NOAA)

NOAA National Oceanic and AtmosphericAdministration

NOAA-Next Polar-orbiting meteorologi-cal satellites (1993—)

NOMSS National oceanographic and me-teorological satellite

system

NWP Numerical weather prediction

NWRF Naval Weather Research Facility,became Naval Environmental

Prediction and Research Facility (NEPRF)and now Navy

Oceanographic and Atmospheric ResearchLaboratory-WEST

(NOARL-WEST)

NWS National Weather Service

OLR Outgoing longwave radiation

OMEGA

PIBAL Pilot balloon observation

POES Polar-orbiting environmental satel-lite program

QBO Quasi-biennial oscillation

RAWIN Radio wind observation

RECCO Reconnaissance code

RH Relative humidity

RHI Range height indicator (radar)

RI Radar index

RMS Root mean square

RUSSWO Revised uniform summary ofsurface weather observations

(AWS)

s second

SATCOM Satellite telecommunications

SEA Southeast Asia

SH Southern Hemisphere

SI Stability index

SMONEX Summer Monsoon Experiment ofFGGE

SPCZ South Pacific convergence zone

SSM/I Special sensor microwave/imager

SST Sea surface temperature

SSWWS Seismic Sea Wave Warning System

STJ Subtropical jet stream

STORM Stormscale Operational and Re-search Meteorology Program

TAMEX Taiwan Area Mesoscale Experi-ment, 1987

TEJ Tropical easterly jet

TEMPO Temporary (in WMO TerminalAirfield Meteorological Code)

TIROS Television infrared observationalsatellite

TOPCAT Project to study clear air turbu-lence (CAT) near the

subtropical jet over Australia

TOPEX/ POSEIDON Ocean TopographyExperiment (US/France)

TOVS TIROS - operational vertical sounder

TUTT Tropical upper-tropospheric trough

USAF United States Air Force

USWB United States Weather Bureau

USAFETAC USAF Environmental Tech-nical Applications Center

USSA United States standard atmosphere

UTC Universal time coordinated

V Visibility

WBAN Weather Bureau, Air Force, Navy

WBFC Weather Bureau Forecast Center

WMO World Meteorological Organization

WMONEX Winter Monsoon Experiment ofFGGE

WPC Weather plotting chart

WSR-88D Weather surveillance radar -88 Doppler

WWW World Weather Watch

Z-R Reflectivity-rainfall relationship(radar)

FIGURE LEGENDS

Frontispiece. The effect of Typhoon Russ as it passed south of Guam (from JTWC, Guam).

A. DMSP visible image at 2245 UTC 19 December 1990 when the typhoon was centered 130NM (200 km) southeast of Guam, with maximum winds of 125 knots.

B. Track of Russ.

C. Barogram for Agana Naval Air Station.

D. Sustained winds (knots) at Agana.

E. Directions of destructive winds and storm surge heights (feet) over southern Guam.

F. Estimated total rainfall (inches) over Guam.

Figure 1-1. Schematic representations of monsoon and tradewind troughs, and near-equa-torial convergence zone (top), and near-equatorial buffer zones (bottom). Stippling denotesareas of unsettled weather.

Figure 1-2. Electromagnetic energy categories, frequencies and wavelengths (based on Elrikand Meade, 1987).

Figure 1-3. Model of the low-level summer circulation over the tropical eastern NorthPacific and associated meridional rainfall and pressure patterns. Dashed line marks themonsoon trough (after Sadler, 1967).

Figure 1-4. Four definitions of the intertropical convergence zone (ITCZ) over the tropicalNorth Atlantic (Sadler, 1975).

Figure 2-1. Average annual meridional distribution of the net radiation fluxes of the earth’ssurface, atmosphere, and earth-atmosphere system (from Sellers, 1965).

Figure 2-2. Average annual meridional distribution of precipitation, evaporation, andprecipitation minus evaporation (from Sellers, 1965).

Figure 2-3. Average annual meridional distribution of the components of the polewardheat-energy transport (from Sellers, 1965).

Figure 2-4. Mean vertical velocity (in units of 10-4 mb s-1 or 1.5 mm s-1) at 500 mb in thetropics. Areas of upward motion are shaded (from Kyle, 1970).

Figure 2-5. Meridional cross-sections of zonal wind speed (m s-1; westerlies positive,easterlies negative; subtropical jet streams labeled “J”) (from Newell et al. 1972).

Figure 2-6. Meridional cross-sections of the mean meridional circulation in the tropics inthe form of mass-flow streamlines. Units: 1012 g s-1. (from Newell et al., 1972).

Figure 2-7. Zonal circulation cells along the equator (from Newell, 1979).

Figure 2-8. Schematic of the effects of coastal friction on low-level wind flow (Fett andBohan, 1986).

Figure 2-9. Approximate dimension and time scales of tropical systems (after Orlanski,1975). Key: S = mainly in stratosphere; U = mainly in upper troposphere; M = mainly inmiddle troposphere; L = mainly in lower troposphere.

Figure 3-1. Boundaries of the six WMO regions (World Meteorological Organization,(1986).

Figure 3-2. Average daily number of shipboard surface weather observations received atthe National Meteorological Center in June 1988, totalled for 4-degree latitude/longituderectangles. * shows locations of moored weather buoys. � shows locations of automatedisland weather stations.

Figure 3-3. Drifting buoy observations in the tropical Pacific during August 1989. Linesindicate weekly displacements (Climate Analysis Center, 1989).

Figure 3-4. Tropical stations making radiosonde (o) and rawin/pibal (o) observations at thebeginning of 1991 (at least one observation a day on two-thirds of the days).

Figure 3-5. Maximum weekly number of international commercial airline flights crossingeach 10° rectangle between 30°N and 30°S (based on ICAO charts).

Figure 3-6. Number of aircraft weather observations in each 10° rectangle between 30°Nand 30°S, received at AFGWC for the period 00 to 14 UTC 19 December 1990.

Figure 3-7. Cloud motion vectors (full, low-level; dashed, cirrus-level) obtained from pic-ture sequences made by GOES at 0°, 135°W on 20 December, 1989 (from NESDIS).

Figure 4-1. Sea-level pressure in millibars for January, April, July and October (from NavalOceanography Command, 1981).

(a) Mean

(b) standard deviation.

Figure 4-2. Average annual meridional distribution of sea-level pressure (mb) for Januaryand July (derived from Crutcher and Davis, 1969).

Figure 4-3. Simultaneous hourly surface pressure anomalies (mb) for three tropical Pacificstations (from Brier and Simpson, 1969).

Figure 4-4. Surface pressure change (mb) between 00 UTC 21 and 00 UTC 23 July 1966(from Sadler et al., 1968).

Figure 4-5. Correlations between daily surface pressures at Tarawa (1°N, 173°E) and otherPacific stations for May 1986 (from Lander, 1991).

Figure 4-6. Resultant gradient-level winds (knots) for January.

(from Atkinson and Sadler, 1970).

Figure 4-7. As in Figure 4-6, but for April.

Figure 4-8. As in Figure 4-6, but for July.

Figure 4-9. As in Figure 4-6, but for October.

Figure 4-10. Resultant 200 mb winds (knots) for January and April (from Sadler and Wann,1984).

Figure 4-11. As in Figure 4-10, but for July and October.

Figure 4-12. Time-height section of the monthly mean zonal wind components (m s-1) atCanton Island (4°S, 174°W; Jan. 1953 to Aug. 1967), Gan (1°S, 73°E; Sept. 1967 to Dec. 1975),and Singapore (1°N, 104°E; Jan. 1976 to Apr. 1985) (Naujokat, 1986).

Figure 4-13. Mean monthly zonal wind speed at Darwin (12°S, 131°E) at 70,000 feet (21 km)(from Hopwood, 1968).

Figure 4-14. Average amplitude (knots) of the quasi-biennial and annual cycles of the meanmonthly zonal wind speed in the tropical stratosphere (from Reed, 1964).

Figure 5-1. Surface air temperature in °C for January, April, July, and October (from NavalOceanography Command, 1981).

(a) Mean (regions where mean sea-surface temperature exceeds mean air temperature bymore than 2°C are shaded).

(b) Standard deviation over the oceans.

Figure 5-2. Annual march of mean monthly temperatures at selected tropical stations alongmeridians through the continents.

Figure 5-3. Schematic showing the effects of surface wind blowing parallel to the west coastof a SH continent (e.g. South America, southern Africa).

Top. Cross section with the surface wind directed into the paper. Note that wind stress hasraised the surface water level off the coast and lowered it along the coast.

Bottom. Plan view. Arrow with a wavy shaft denotes a geostrophic surface current gener-ated by the change in surface water level.

Figure 5-4. Upwelling regions (stippled). The littoral bordering upwelling is always rela-tively dry (heavy lines) and may sometimes be a desert (Meigs, 1973).

Figure 5-5. Mean 700 mb temperature (°C) (solid lines) and relative humidity (percent)(dashed lines) for winter and summer (after Crutcher and Davis, 1969).

Figure 5-6. Annual ranges of mean 700 mb temperature (°C) and relative humidity (per-cent), NH summer (SH winter) minus NH winter (SH summer)

(derived from Crutcher and Davis, 1969).

Figure 5-7. Mean precipitable water in mm for January and July (after Tuller, 1968).

Figure 5-8. The annual mean relationship between precipitable water and SST. The shadingranges over one standard deviation above and below the averaged data points (Stephens,1990).

Figure 5-9. Normal atmospheres for the tropics: (A) 30°N in January, (B) 30°N in July, and(C) 15°N annual. Plotted on Skew - T diagrams; solid lines temperature, dashed lines dewpoint (from ESSA et al., 1967).

Figure 5-10. Mean vertical temperature and dew point soundings for (A) Bangkok (14°N,100°E) and (B) Papeete (18°S, 150°W) during dry and wet season months, plotted on Skew -T diagrams; solid lines temperature, dashed lines dew point.

Figure 5-11. Vertical structure of equivalent potential temperature ( e) for disturbed andundisturbed days in the tropical Atlantic (after Garstang et al., 1967) and at Saigon (11°N,107°E) (after Harris and Ho, 1969).

Figure 5-12. Average vertical distribution of equivalent potential temperature ( e) during thetwo warmest summer and two coldest winter months over various tropical oceans (afterGray, 1968).

Figure 5-13. Average sounding for the southern California coast in summer (Neiburger etal., 1961) (full line), and a typical summer sounding for 19 June 1976 at Lihue (22°N, 159°W)(dashed line. Relative humidities are entered.

Figure 5-14. GOES visible image of the eastern Pacific for about 15 LT 19 January 1984.

Figure 5-15. Tradewind clouds photographed from the Shuttle in June 1985. View across theisland of Hawaii toward the east (NASA).

Figure 6-1. Mean cloudiness (oktas) for January (from Sadler et al., 1984). Areas with exten-sive deep convection extending above 40,000 feet (12 km) enclosed by heavy dot-dashed

lines (from Janowiak et al., 1985: Duvel, 1989).

Figure 6-2. As in Figure 6-1 but for April.

Figure 6-3. As in Figure 6-1 but for July.

Figure 6-4. As in Figure 6-1 but for October.

Figure 6-5. Mean zonal cloudiness in oktas for each month (from Sadler et al., 1984).

Figure 6-6. GMS IR image for 09 UTC 19 July 1985 showing cloud clusters over and east ofsoutheast Asia.

Figure 6-7. GOES IR images for 11 LT (A) and 16LT (B), 25 October 1988, showing develop-ment of popcorn cumulus over northern South America.

Figure 6-8. (A) Frequency distributions (percent) of cluster widths by months for cloud-clusters in the tropical North Pacific for the 10° latitude band of maximum brightness. Forcomparison the frequency distribution of cluster widths for the mid-latitude North Pacificfor October 1967 is included. (B) Frequency distribution of cloud-cluster separation dis-tances as (bottom to top) smaller clusters are excluded. Distributions are an aggregate ofthe four months for the 10° latitude band used for (A) (adapted from Hayden, 1970).

Figure 6-9. Idealized view of the larger-scale tropical cloud systems on a typical summerday in each hemisphere.

Figure 6-10. Schematic hierarchy of tropical cumulus (from Simpson and Dennis, 1972).

Figure 6-11. How cumulus lean with the wind shear in the vertical; on the left, the cloudvelocity, UC is less than the wind velocity, UE, while the reverse is true of the cloud on theright.

Figure 6-12. Mean zonal tradewind speed in lowest 10,000 feet (3 km) for three tropicalPacific stations, Eniwetak Atoll (11°N, 162°E), Johnston Island (17°N, 170°W), and WakeIsland (19°N, 167°E) (derived from Adams, 1964).

Figure 6-13. Percent frequencies of cumulus top heights for area 2.5°N to 25°N and 155°E to170°W (after Palmer et al., 1955).

Figure 6-14. Measurements made along about 50°E by R. V. Discovery between 16 and 21August 1964.

Top. Kinematic analysis of surface winds and Discovery’s route.

Middle. Isotherms (°C, full line), isohumes (percent, dashed line), and inversion layer(stippled). Arrows denote locations of radiosonde ascents.

Bottom. Air temperature at deck level (full line), sea surface temperature (dashed line) andrelative humidity at deck level (dot-dashed line) (Ramage, 1971).

Figure 6-15. DMSP visible image for 7 August 1986 showing clear skies off Somalia, and apatch of fog or stratus along the southeast coast of Arabia.

Figure 6-16. Mean summer cloudiness and other elements over the tropical Atlantic. Thesurface trough is shown by dots and the surface convergence asymptote by dashes.

a. August sea-level pressure in mb (1000+) (from Sadler et al., 1987).

b. August resultant surface winds (speed in m s-1) (from Sadler et al., 1987)

c. August divergence (in 10-6s-1) derived from b.

d. August sea-surface temperature (in °C) (from Sadler et al., 1987).

e. August outgoing longwave radiation (in 10’s of W m-2) (from Janowiak at al., 1985).

f. August frequency of highly reflective clouds (in days month-1) (from Garcia, 1985).

g. August cloudiness (in oktas) (from Sadler et al., 1984.

h. June-August percentage frequency of cumulus (from Warren et al., 1989).

i. June-August percentage frequency of stratus and stratocumulus (from Warren et al.,1989).

j. June-August percentage frequency of nimbostratus (from Warren et al., 1989).

k. June-August percentage frequency of cumulonimbus (from Warren et al., 1989).

l. June-August percentage frequency of altostratus and altocumulus (from Warren et al.,1989).

m. June-August percentage frequency of cirrus, cirrostratus and cirrocumulus (from War-ren et al., 1989).

Figure 6-17. Ratio of the annual mean number of thunderstorm days to the mean annualrainfall in decimeters. Areas with ratio > 8 are hatched (from Portig, 1963).

Figure 6-18. Monthly means of thunderstorms, rainfall, lapse-rate, and shear of the wind inthe vertical at four stations in western India, located on Figure 6-13.

Bombay 19°N, 73°E Full lines

Ahmedabad 23°N, 73°E Dotted lines

Jodhpur 26°N, 73°E Dot-dashed lines

New Delhi 29°N, 77°E Dashed lines

A. Ratio of thunderstorm frequency to rainfall in decimeters (heavy lines), where numbersdenote thunderstorm frequencies; rainfall (light lines).

B. Difference in virtual temperature between 850 and 300 mb (heavy lines, at the moistadiabatic lapse rate the difference ranges between 33.5°C (T850 = 29°C and 43.5°C (T850 =20°C); Wind shear between 850 and 300 mb (light lines).

Figure 6-19. Comparison of July mean rainfall and mean sea-level pressure over northernIndia. Full lines are isohyets (in cm) and are not shown above 50 cm (20 inches); hatchingdenotes areas where the rainfall average exceeds 40 cm (15 inches); dashed lines are isobars(in mb). Stippling denotes areas with an average of more than ten thunderstorm days in themonth (data from the India Meteorological Department).

Figure 6-20. Satellite picture showing storm cloud over Singapore (1°N, 104°E) at 13 LT 2December 1978.

Figure 6-21. Rainfall record at Mt. Waialeale (22°N, 159°W) for 14 and 15 October 1976.Tipping bucket activated the recorder at 1 mm intervals. Total 186 mm. Local time = UTC -10 h.

Figure 6-22. Mean annual outgoing longwave radiation in W m-2 (from Janowiak et al.,1985).

Figure 6-23. Mean annual zonal cloudiness (in oktas) (from Sadler et al., 1984) and outgoinglongwave radiation (in W m-2) (from Janowiak et al.,

1985).

Figure 6-24. Top. The Southern Oscillation Index (SOI) calculated by subtracting the surfacepressure (in mb) at Darwin (12°S, 131°E) from the surface pressure at Easter Island (27°S,109°W), smoothing, and plotting the departures from the long-term mean difference.

Bottom. Puerto Chicama (8°S, 79°W). Departures of monthly sea surface temperatures fromthe long-term means (°C).

Figure 6-25. Schematic representation of principal rainfall anomalies found to accompanyEl Niño and anti-Niño years.

A. Associated with El NiÑo (Ropelewski and Halpert, 1987).

B. Associated with anti-NiÑo or a cool year (Ropelewski and Halpert, 1989). The bracketed“0” indicates the months of a NiÑo or cool year from January through December, thebracketed “+” the months of the following year.

Figure 6-26. Tracks of South Pacific tropical cyclones from tropical depression stage onwardfor December 1982 through May 1983 (Sadler and Kilonsky, 1983).

Figure 6-27. Schematic representation of the vertical circulation associated with the summerand winter monsoons.

Figure 6-28. The area satisfying monsoon criteria is enclosed by the full line. Within thisarea, deserts (annual average rainfall <10 inches (250 mm)) are stippled and the area with abimodal annual rainfall variation is hatched. Except for a small portion of eastern Peninsu-lar India the remainder of the monsoon area experiences a summer or autumn rainfallmaximum.

Figure 6-29. Mean annual cloudiness in oktas based on three years of weather satellitephotographs (from Sadler, 1969).

Figure 6-30. JULY. 1. Resultant winds at the surface (full lines, thickened where speedsexceed 14 knots). Major pressure troughs are denoted by wriggly lines. 2. Resultant windsat 200 mb (dashed and thickened where speeds exceed 40 knots). Major pressure ridges aredenoted by joined diamond lines. Labelling indicates preferred regions for synoptic sys-tems.

Figure 6-31. JANUARY. As in Figure 6-30 except that thickened dashed lines denote 200 mbwind speeds > 120 knots.

Figure 6-32. Circulation components of the monsoon arranged schematically according toweather, divergence, and vertical motion. Heavy lines denote levels of non-divergence. (seeChapters 8 and 9 for discussions of synoptic weather systems).

Figure 6-33. Annual latitudinal variation of lower-tropospheric pressure troughs over theIndian Ocean.

Figure 6-34. GMS IR image of a Mei-Yu front extending from northern Indochina to south-ern Taiwan at 18 UTC 28 May 1982.

Figure 6-35. Mean monthly rainfall for selected stations in Asia/Australia and the Americas.

Figure 6-36. Mean monthly rainfall for selected stations in the Pacific and Africa.

Figure 6-37. Mean monthly rainfall versus coefficient of variation for stations in India(derived from Mooley and Crutcher, 1968). Dashed line is from a similar study for stationsin the Philippines (after Coligardo, 1967).

Figure 6-38. Pentad means. Rainfall (full lines) for nine stations in southern China and fivestations in and south of Japan. Pressure (dashed lines) for Hong Kong and the five Japanesestations, and 50-year tropical cyclone frequencies (dotted line) for southern China.

Figure 6-39. Pentad mean rainfalls and monthly means (bars) for Indian stations(Ananthakrishnan and Pathan, 1970).

Figure 6-40. Onset dates of southwest monsoon rains over Bombay (19°N, 73°E) from 1879through 1975 (Rao, 1976).

Figure 6-41. Number of years each pentad received more than 1 inch (25 mm) of rain at SanSalvador (14°N, 89°W) based on a 40-year period of record. Dotted line shows linear best fit(after Griffiths, 1964).

Figure 6-42. Periods each year 1918-57 at San Salvador when the soil-water budget ex-ceeded 1 inch (25 mm) (after Griffiths, 1964).

Figure 6-43. Mean date of start of the rainy pentads (> 1 inch (25 mm)) over CentralAmerica (Granzow and Henry, 1972).

Figure 6-44. Mean date of end of the rainy pentads (last pentad with > 1 inch (25 mm)) overCentral America (Granzow and Henry, 1972).

Figure 6-45. Histograms showing the climatological annual progression of fractional lowradiance (FLR) based on OLR data for 1974-88 for boxes centered on a. 11°N, 85°W; b.Equator, 65°W; c. 11°S, 55°W. (Horel et al., 1989).

Figure 6-46. Time-latitude diagram of the 8-year (1980-87) climatology of pentad OLRacross the Americas. Areas with OLR < 200 W m-2 — heavy shading; 220 to 200 W m-2 —light shading. Contour interval 20 W m-2 (Horel et al., 1989).

Figure 6-47. The durations of wet seasons over the tropical Americas and Africa (see text).The summer wet season in each hemisphere is represented by a black bar. Transition sea-sons, when a wet season is not established, appear as gaps in the bar. A striated bar repre-sents the first and last months of periods without data. Hatched regions within a bar de-note “break” periods, when the maximum FLR shifts briefly into the opposite hemisphere(Horel et al., 1989).

Figure 6-48. Consecutive three-day means of the daily rainfall at seven

stations near Saigon (11°N, 107°E) for each year 1955-68 (less 1963) and 13-year average.Arrows show ending dates of rainy season by Fukuda’s criterion. Adapted and expandedfrom Fukuda (1968).

Figure 6-49. Relationship between cumulative rainfalls and cumulative rainfall frequency(see text) (after Martin, 1964).

Figure 6-50. Generalized profile of mean annual precipitation (cm) versus altitude in thetropics. The zone of maximum is shaded (from Lauer, 1975 and Giambelluca et al., 1986(dotted)).

Figure 6-51. Composite curve of Contingency Index (CI) as a function of distance betweenCentral American rainfall stations. Mesoscale clouds and circulations drawn to the distancescale are added to illustrate physical relationships (after Henry, 1974).

Figure 6-52. Typical rainfall associated with a single mesoscale convective system (fromHenry, 1974).

Figure 6-53. Frequency distribution of individual rain shower amounts (mm) and durations(minutes) at Piarco, Trinidad (11°N, 61°W) during dry and wet seasons (after Garstang,1959).

Figure 6-54. Nomogram based on mean annual rainfall to determine mean number ofclock-hours that a specified rainfall rate is equalled or exceeded (after Winner, 1968).

Figure 6-55. World’s greatest observed point rainfalls for various time periods (after Paulus,1965; Dhar, 1977; and Chaggar, 1984).

Figure 6-56. Probable extreme daily rainfall for selected tropical stations (adapted fromAtkinson, 1968; Lockwood, 1967; and Griffiths, 1967).

Figure 6-57. Rainfall at radial distances from tropical cyclone centers. * computed frommoisture convergence (Riehl, 1979). X measured in Hurricane Donna (Riehl, 1979). � mea-sured in Florida around tropical cyclones (Miller, 1958). measured on west Pacific islandsaround typhoons (Frank, 1976).

Figure 6-58. Average 48-hour rainfall (inches) for 46 Gulf Coast hurricanes (after Goodyear,1968).

Figure 7-1. Diurnal variation of surface air temperature at a buoy in the North Atlantictradewinds averaged between 4 and 13 February 1969 (Prümm, 1974), and at Willis Island(16°S, 150°E) for 59 undisturbed tradewind days in July-September 1964 (Neal, 1973).

Figure 7-2. Locations of instrument sites on Willis Island (Neal, 1973). Thermometer screensat TM (Main Site), and TB (Beach Site). Anemometers on island (AM) and on reef to wind-ward (AR).

Figure 7-3. Mean diurnal temperature and dew point curves at: 1. Luang Prabang (20°N,102°E) — temperature for August and March, and dew point

for August. 2. Eniwetak (11°N, 162°E) — temperature and dew point for the year.

Figure 7-4. Mean diurnal variation of surface pressure (mb) over the tropical oceans (fromMeteorological Office, 1968; Jenkins, 1945). -1, 0, and +1 isobars are labeled.

Figure 7-5. Mean diurnal variation of surface pressure (mb) for Hong Kong (22°N, 114°E) inJanuary. Full line — days with cloudiness greater than 95 percent; dashed lines — dayswith cloudiness less than 20 percent.

Figure 7-6. Hodographs of summer diurnal variation of surface winds at Hilo (20°N,155°W) on windward Hawaii (full line) and at Waikoloa Beach (20°N, 156°W) on leewardHawaii (dashed line) (Ramage, 1978).

Figure 7-7. Concept of east-west land- and mountain-sea breeze circulations made duringsummer over the eastern part of the island of Hawaii. a. daytime upslope-sea breeze; b.nighttime drainage-land breeze. At each surface station and on the radio sounding aboveHilo, dry bulb temperatures (°C) are plotted above dew points, and on the wind arrowsone full feather denotes 10 knots. The Mauna Loa Observatory (MLO) (20°N, 156°W) dataare boxed because the terrain slopes north-south there rather than east-west (Garrett, 1980).

Figure 7-8. Isotachs (m s-1) for the land-sea breeze at Djakarta (6°S, 107°E). Winds with a seabreeze component are shaded (after van Bemmelen, 1922).

Figure 7-9. Synthesized empirical model of the land-sea breeze system along the Texas GulfCoast. Arrow lengths are proportional to wind speeds (modified from Hsu, 1970).

Figure 7-10. Western Bight of Benin photographed by an astronaut during the afternoon.Clouds show the sea breeze extending up to 30 NM (55 km) inland. Air sinks and keepsskies clear along the coast and up to 40 NM (75 km) offshore (NASA).

Figure 7-11. Numerical model of the sea breeze at 11 and 17 LT with three prevailing syn-optic wind flows: A. calm, B. offshore wind at 10 knots and C. onshore wind at 10 knots(modified from Estoque, 1962).

Figure 7-12. Numerical model of the south Florida sea breeze. Arrows show horizontalwind velocity at the 160 feet (50 m) level, 3, 5, 8, and 10 hours after simulated sunrise. Fulllines show axes of convergence (from Pielke, 1974).

Figure 7-13. METEOSAT IR image of Lake Victoria (2°S, 32°E) at 14 LT 26 July 1990, show-ing the effects of lake winds on clouds.

Figure 7-14. Surface circulations over Lake Victoria.

A. April; 08 LT on left, 14 LT on right.

B. July; 08 LT on left, 14 LT on right. From Leroux (1983).

Figure 7-15. Lake Volta photographed by an astronaut during the

afternoon. Cumulus are uniformly distributed over flat land. All the water areas are clear ofcloud (NASA).

Figure 7-16. Surface winds on the island of Hawaii at 05 and 17 LT 2 August 1990. Thetradewind inversion remained close to 6500 feet (2 km). Land contours are at 1 km inter-vals.

Figure 7-17. Diurnal circulation and cloudiness patterns in the

Mt. Kenya area (Eq., 37°E). A. Afternoon, B. Night. Vertical

exaggeration about 1:25 (Hastenrath, 1985).

Figure 7-18. Cloud pattern associated with the effect of Mt. Haleakala (21°N, 156°W; 10,200feet (3111 m) altitude) showing anabatic (upslope) and katabatic (downslope) winds. Hindicates heating and C indicates cooling (Lyons, 1979).

Figure 7-19. Resultant surface wind streamlines over the Red Sea/Gulf of Aden in the morn-ing (left) and afternoon (right) during July-August (modified from Flohn, 1965).

Figure 7-20. Hodograph of diurnal variation of surface winds at Honolulu Airport (21°N,158°W) (Ramage, 1978a).

A. Tradewinds averaging 4 knots or less and winds from other directions.

B. All tradewinds.

Figure 7-21. Composite 3 h profiles of the wind speed at Obbia (5°N, 48°E) and Burao (9°N,45°E). The X-axis shows a staggered scale of wind speed, beginning at zero for each profile(from Ardanuy, 1979).

Figure 7-22. Schematic diagram of circulations over and near large rivers of Amazonia (a)during the night, and (b) during the day (Garstang et al., 1990).

Figure 7-23. Nocturnal ocean-land interface with onshore prevailing wind. A low-level jetcoincides with the boundary layer top and causes the sea breeze to persist into the night.

Figure 7-24. Nocturnal ocean-land interface with offshore prevailing wind. The low-level jetoften reaches the surface along the coast and just offshore, and so enhances the land breeze(from Hsu, 1979).

Figure 7-25. Average diurnal variation of the inversion height (I) and the cloud base height(B) for a period of undisturbed tradewinds during ATEX (Brill and Albrecht, 1982).

Figure 7-26. Mean November 1978 cloudiness (C) and cloud top temperature (Tc) for a 135X 135 NM2 (250 X 250 km2) region at 21.4°S, 86.3°W (Minnis and Harrison, 1984).

Figure 7-27. Estimated day and night rates of net radiational warming within a tropicaldisturbance and in the surrounding clear or mostly clear regions (Gray and Jacobson, 1977).

Figure 7-28. Idealized slopes of pressure surfaces between a disturbance and the surround-ing clear air, during the day (dashed) and at night (full) and the resultant night-versus dayinward-outward radial wind patterns (Gray and Jacobson, 1977).

Figure 7-29. Diurnal variation of satellite-observed radiating cloud surfaces associated withtropical cyclones. Areas of “cold” (<-65°C), “intermediate” (-40°C to -65°C), “warm” (-15°Cto -40°C), and total (<-15°C) (from Zehr, 1987).

Figure 7-30. Percent of the total thunderstorm activity occurring during any hour at MajuroAtoll (7°N, 171°E) and Clark Air Base (15°N, 121°E).

Figure 7-31. Diurnal variation of rainfall frequency during the month of maximum rainfallat southeast Asian stations.

Figure 7-32. Mean percent of total rainfall for 6-hourly periods at 11 stations on the KoratPlateau, Thailand (June-August 1967 and 1968).

Solid = during weak monsoon;

Stippled = during strong monsoon (from Ing, 1971).

Figure 7-33. Diurnal variation in the frequencies of rainfall at stations on the island ofHawaii. In the boxes, noon LT is marked by a vertical line and 10 percent frequency by ahorizontal line. Station identification numbers are shown. (Schroeder et al., 1978).

Figure 7-34. Positions of the rainfall maxima over northeast Brazil by observation times(Kousky, 1980).

Figure 7-35. Mean horizontal divergence over West Malaysia for 25 days in August 1957and 1958 derived from coastal pilot balloon observations. Full line — 1000 feet (300 m)elevation; dashed line — 2000 feet (600m) elevation; dot-dashed line -3-hourly runningmeans of rainfall at 11 uniformly distributed stations (Ramage, 1964).

Figure 7-36. Areal extent of diurnal rainfall regimes over West Malaysia in August (see textfor discussion) (adapted from Ramage, 1964).

Figure 7-37. GOES IR images of central America. A. 19 LT 6 July 1985. B. 07 LT 7 July 1985.

Figure 7-38. Graphical representation in 3-D of the island of Sulawesi, as viewed from theeast.

Figure 7-39. Centers of cumulonimbus activity over and around Sulawesi for 1-15 Decem-ber 1978. + marks mountain peaks above 6000 feet (1800 m). Based on infrared picturesfrom GMS. A. 05 LT, B. 17 LT.

Figure 7-40. Diurnal variation of the ratio (normalized by area) between the numbers ofcumulonimbus centers over Sulawesi and over the neighboring sea. Based on 3-hourlyinfrared pictures from GMS. As the curve passes through L/S X .465 = 1, the ratio is re-versed, in order to emphasize the symmetry of the variation.

Figure 7-41. GMS IR images of eastern Borneo and Sulawesi for 20 July 1985. A. 05 LT. B. 17LT.

Figure 8-1. Summer. Mean areal frequency of surface monsoon depressions. Isoplethslabeled in times per season (full) and in years between occurrences (dashed) (Ramage,1969).

Figure 8-2. Schematic of the vertical circulation across a monsoon depression (Douglas,1987).

Figure 8-3. Streamline analysis for 850 mb over the Bay of Bengal for 14 and 16 July 1966.The monsoon trough (dashed) migrated southward prior to a depression forming on 17July (after Sadler, 1967).

Figure 8-4. DMSP visible image of a monsoon depression developing over the northern Bayof Bengal on 12 August 1986. Note the relatively cloudless trough extending west-north-west from the depression center, and the mass of convective cloud south of the trough. Amid-tropospheric cyclone is filling over the northern Arabian Sea.

Figure 8-5. Streamlines at 10,000 and 2000 feet (3 km and 600 m) and sea-level pressureanalysis over West Africa for 12 UTC 24 August 1967 (adapted from Carlson, 1969a).

Figure 8-6. Disturbances over West Africa.

A. Schematic flow pattern at 10,000 feet (3 km) (full lines) and 2000 feet (600 m) (dashedlines).

B. Mean percentage cloudiness with respect to wave axis (after Carlson, 1969b).

Figure 8-7. Surface and 700 mb pressure-height charts for 7 January 1949 illustrating asubtropical cyclone in the North Pacific (after Simpson, 1952).

Figure 8-8. Typical tracks of North Pacific subtropical cyclones (adapted from Simpson,1952).

Figure 8-9. Composite of surface winds (knots) around a North Pacific subtropical cycloneduring 4-5 April 1960 (adapted from Ramage, 1962).

Figure 8-10. Schematic radial cross section of clouds, weather, and vertical motion in asymmetrical subtropical cyclone. Plus signs denote horizontal divergence and minus signs,convergence (adapted from Ramage, 1962).

Figure 8-11. GOES visible image on 1 January 1986. A subtropical cyclone is centered near23°N, 139°W.

Figure 8-12. Mean July positions of the monsoon trough over India at 3000 feet (900 m) and500 mb (adapted from Miller and Keshavmurthy, 1968).

Figure 8-13. Meridional cross section along western India of the July mean zonal winds(knots) (adapted from Miller and Keshavamurthy, 1968).

Figure 8-14. Composite kinematic analyses (knots) for 1-10 July 1963.

A. A near-surface layer (1600 to 3000 feet (500 to 900 m))

B. The 600 mb level, showing a well-developed mid-tropospheric cyclone over westernIndia (adapted from Miller and Keshavamurthy, 1968).

Figure 8-15. The mid-tropospheric cyclone shown in Figure 8-18.

A heavy dot locates the cyclone center at 600 mb.

A. East-west cross section of clouds and vertical motion. Arrows show relative verticalmotion computed from composite kinematic analyses.

B. Horizontal distribution of cloudiness and rainfall. Hatching delineates rainfall >1.57inches (4 cm) day-1, stippling areas of broken to overcast middle or high cloud (adaptedfrom Miller and Keshavamurthy, 1968).

Figure 8-16. DMSP visible image of a mid-tropospheric cyclone over the northern ArabianSea on 8 August 1986. The cyclone moved slowly westward and became very weak by the12th (Figure 8-4). Note orographic cloud along the western Ghats and relatively little cloudalong the monsoon trough over central India. The cloud mass south of India moved north-east during the next day. By the 12th it lay southwest of a developing monsoon depression(Figure 8-4).

Figure 8-17. GOES visible image of a “dry” upper tropospheric (cold) cyclone over thecentral North Pacific taken on 8 July 1979.

Figure 8-18. Composite mean cloudiness (tenths) derived from 13 cases of “wet” upper-tropospheric (cold) cyclones in the North Atlantic for 1961-1966 (adapted from Frank,1970).

Figure 8-19. GOES moisture channel image of a cold low centered west of Hawaii on 24May 1984. Note moist air east of the cyclone center and dry air in the rest of the circulation.

Figure 8-20. Mean locations of 200 mb ridges and troughs over the Northern Hemisphere inAugust (after Sadler, 1964).

Figure 8-21. Schematic model of a tropical cyclone initiated by an upper tropospheric low(Sadler, 1976a).

Figure 8-22. Time-altitude section and 12-hourly surface observations for Midway Island(28°N, 177°W) from 27 to 31 July 1970. Dotted lines delineate troughs. Boundaries of 50percent relative humidity are dashed lines (from Sadler,1976a).

Figure 8-23. Model of the lower-tropospheric contour and weather patterns of a temporal(adapted from Pallman, 1968).

Figure 8-24. Schematic cross section through a squall system typical of the tropics. Airflows are relative to the squall line which is moving from right to left. Circled numbers aretypical values of w in °C. See text for detailed discussion (Zipser, 1977).

Figure 8-25. Schematic streamlines of airflow relative to convective disturbances in theLine Islands area. (A) east-west, and (B)

north-south sections represent intensification and downdraft production, with many activeCB’s either in individual clusters or in organized lines. (C) is the north-south section of thedissipating phase, when the downdraft is maintained primarily by rain falling from theextensive upper-cloud shield (after Zipser, 1969).

Figure 8-26. Isochrones of the leading edge of the downdraft air associated with a distur-bance in the Line Islands area on 1 April 1967. based on satellite, surface and aircraft obser-vations. Full lines mark high-confidence locations, dashed lines are more conjectural. Satel-lite pictures suggested that the bordering dashed lines marked the limit of the downdraftarea (after Zipser, 1969).

Figure 8-27. Schematic of cloud development over a ridge west of an approaching WestAfrican squall line (from Leroux, 1983).

Figure 8-28. DMSP IR images of West Africa on 18 (top) and 19 (bottom) July 1987. Acurved squall line extended from about 11°N to the equator. At 7°N it moved westwardfrom 15°E to 3°E in 24 hours, at a speed of about 30 knots. Note how other clouds changedbetween the images.

Figure 8-29. Model of the surface kinematic pattern associated with a cold front extendinginto a tropical ocean shear line. Isotachs (dashed) labeled in knots (adapted from Palmer etal., 1955).

Figure 8-30. GOES visible image at 00 UTC 25 December 1985 showing a cold front crossingHawaii. South of 15°N the front merged with a shear line.

Figure 8-31. Daily values for January 1967 over the South China Sea.

Upper. Hong Kong sea-level pressures, dashed line and dew points, dotted line; Saigonsoutherly components of the 200 mb winds, dot-dashed line; Singapore 200 mb winds, fullline.

Middle. Surface isotachs (m s-1) averaged zonally between 110° and 118°E.

Bottom. Histogram of rainfall averaged for five coastal stations in North Borneo.

Figure 8-32. Schematic meridional cross section over the South China Sea during a surge ofthe winter (northeast) monsoon.

Top. Cloud, weather and temperature profile. Dot-dashed line lies at the base of the frontalinversion.

Bottom. Latitudinal averages of sea-surface, air and dew point temperatures (adapted fromNavy Weather Research Facility Staff, 1969).

Figure 8-33. GMS IR image for 09 UTC 9 December 1985, prior to a cold surge across theSouth China Sea.

Figure 8-34. GMS IR picture for 09 UTC 10 December 1985, following a cold surge acrossthe South China Sea.

Figure 8-35. Topography of Central America showing the three main low-lying links be-tween the Gulf of Mexico/Caribbean and the Pacific. Land above 2000 feet (600 m) ishatched and above 6000 feet (1800 m) is checkered (Sadler and Lander, 1986).

Figure 8-36. Analysis of long-term ship data for January.

a. Isotachs.

b. Kinematics.

c. Sea-surface temperature.

d. Sea-level pressure

Wide dashed lines delineate the main valleys shown in Figure 8-35 (Sadler and Lander,1986).

Figure 8-37. Synoptic sequence over North Africa for a 48-hour period centered on 23February 1943. Previous and following day frontal positions indicated by dashed lines.Shading shows area of dust-restricted visibility (adapted from Solot, 1943).

Figure 8-38. Schematic vertical profile of winds and clouds associated with weak, moderateand strong surges of the southwest (summer) monsoon. Heights in thousands of feet(Guard, 1985).

Figure 8-39. Schematic time-sequence (according to Leroux, 1983) of a surge (dashed line)progressing from Egypt to West Africa during the warm season. Streamlines depict surfaceflow. North of the heat trough (double line) the surge raises dust (stippling); near the coast,influx of moist, maritime air inhibits this effect. South of the heat trough the surge causessqualls.

Figure 8-40. Three-dimensional synoptic model of an altostratus layer overlying a surfacefront in the region of the southwest Pacific convergence zone (SPCZ) during winter. Thenorth-south line usually lies near 160°E (Hill,1964).

Figure 8-41. GMS IR image at 16 UTC 1 July 1985 of a South Pacific convergence zone in itsnormal climatological position.

Figure 8-42. 10 to 18 January in the southwest Pacific.

Top. Average outgoing longwave radiation (W m-2); area < 225 W m-2 shaded; it probablycontained deep convection.

Bottom. Average mean sea-level pressure (Vincent, 1985).

Figure 8-43. Circulation at 300 mb for 00 UTC 6 February 1968. Isotachs in m s-1. The surfacefront and the soundings of Figure 8-45 are located.

Figure 8-44. Clouds photographed by ESSA 3 at 05 UTC 6 February 1968.

Figure 8-45. Soundings, relative humidities and winds (one full barb denotes 10 knots) at 00UTC 6 February 1968.

1. West of trough in the subtropical westerlies — Chiangmai (19°N, 99°E) (dashed line).

2. East of trough in the subtropical westerlies — USS Belnap (19°N, 107°E) (full line) (NavyWeather Research Facility Staff, 1969).

Figure 8-46. Clouds over southern Africa photographed from ESSA 3, 17 April 1968. Thecloud zone was moving northeastward ahead of a sharp trough in the upper troposphericwesterlies. Rain from showers and thunderstorms occasionally exceeded 2 inches (50 mm)(Fox, 1969).

Figure 8-47. Sounding at Lihue (22°N, 159°W) at 00 UTC 19 April 1974. Temperature fullline, dew point dot-dashed line. For winds, one full barb denotes 10 knots (Schroeder,1977).

Figure 8-48. Lihue, 1976. Frequency distribution of the heights of the dry layer base as afunction of 200 mb wind direction.

Figure 8-49. Sounding at Lihue at 00 UTC 15 October 1976. Temperature full line, dew pointdashed line. For winds, one full barb denotes 10 knots.

Figure 8-50. Total rainfall (mm) over India from 1 through 6 December 1962.

Figure 8-51. Clouds photographed from TIROS VI at 09 UTC 4 December 1962 when uppertrough and lower disturbance were superposed.

Figure 8-52. Time cross section for Bombay (19°N, 73°E) from 3 to 6 December 1962.Isogons are shown full, isotachs (m s-1) dashed and upper limit of relative humidities above50 percent thick and full. Arrow denotes time of satellite photo (Figure 8-51).

Figure 8-53. Role of the Atlantic upper-tropospheric trough in the stimulation or constraintof mass circulation in a tropical cyclone. As the cyclone approaches the wind maximumsouth of the trough (anticyclonic-shear area), mass circulation in the cyclone is stimulated.As the cyclone passes across the upper trough axis and moves beneath cyclonic shear, masscirculation is constrained (after Simpson, 1970b).

Figure 8-54. Flight tracks and altitudes flown by NOAA P-3 research aircraft, November1977 to January 1978 (Ramage et al., 1981).

Figure 8-55. Time-latitude cross section of GOES images along a two-degree wide stripcentered at 150°W for the period 29 November 1977 through 5 January 1978. Images madedaily at 1015 LT. Arrows indicate days of aircraft traverses (Ramage et al., 1981).

Figure 8-56. Schematic meridional cross section of a vigorous near-equatorial convergencezone.

Figure 8-57. GOES visible image at 00 UTC 2 December 1985 of an upper tropospherictrough interacting with a near-equatorial convergence zone.

Figure 8-58. A. Locations of 700 mb ridge on alternate days from 11 to 21 July 1967. 700 mbanalyses for 11 (B), 15 (C), and 19 (D) July 1967 (adapted from Sadler et al., 1968).

Figure 8-59. GMS IR image at 16 UTC 11 July 1985, showing a mid-summer dry spell oversouth China and the South China Sea.

Figure 9-1. Vertical radial cross section of the mean tangential velocity (m s-1) in 14 Pacifictyphoons. Positive values denote cyclonic and negative values anticyclonic circulation(after Izawa, 1964).

Figure 9-2. Kinematic analyses of (A) the lower-tropospheric, and (B) the upper-tropo-spheric circulation in Hurricane Donna on 10 September 1960. Areas of speed maxima areshaded (after B.I. Miller, 1967).

Figure 9-3. Model of typhoons with inner rainband (right) and outer and inner rainbands(left).

Top. Upper tropospheric outflow winds.

Bottom. Vertical cross section.

The pressure along both the outer and inner shear lines is a little low, creating convergentoutflow winds which subside somewhat (after Fujita et al., 1967).

Figure 9-4. Vertical cross section of temperature anomalies (°C) relative to the mean tropicalatmosphere, Hurricane Cleo, 18 August 1958 (after LaSeur and Hawkins, 1963).

Figure 9-5. Radarscope of Typhoon Nelson centered 85 NM (157 km) southeast of Kadena(RODN) at 1138 UTC 6 October 1988.

Figure 9-6. Top. Enhanced GMS IR image of super typhoon Flo, centered near 26°N, 129°Eat 0538 UTC 17 September 1990. Bottom. Enhancement curve used to shade-code the image(from NOAA/NESDIS, 1983).

Figure 9-7. Average annual number (and percent of global total) of tropical cyclones oftropical storm or greater intensity in each development area for 1958 to 1977 (from Gray,1978).

Figure 9-8. The tracks of tropical cyclones for a three-year period (Gray, 1978).

Figure 9-9. Average monthly number of tropical cyclones of tropical storm or greater inten-sity relative to calendar and solar years for the northern and southern hemispheres and theglobe (from Gray, 1978).

Figure 9-10. Locations of the tradewind convergence and monsoon trough zones at thegradient level during February, May, August and November (from Atkinson and Sadler,1970).

Figure 9-11. Mean position of the 200 mb ridge in July and August and mean August 80°F(26.7°C) sea surface isotherm in the eastern North Pacific (Sadler, 1964).

Figure 9-12. Tracks of surface cyclones over the tropical Atlantic during August 1963 (afterAspliden et al., 1965-1967).

Figure 9-13. Model of low-level cyclones in the tropical North Atlantic depicting either achain of cyclones or the life history of one cyclone. Major satellite-observed cloud areas arestippled; the clear areas within the stippling represent deep, more convective cloud groups(after Sadler, 1967).

Figure 9-14. Twenty-two cross-equatorial named tropical cyclone pairs over the PacificOcean, September 1971 through January 1980. Circled numbers indicate points of originand the season of occurrence (1 =

1971-72 and 9 = 1979-80) (from Keen, 1982).

Figure 9-15. DMSP visible image of twin cyclone development at 2330 UTC 17 December1990. Observed surface winds are plotted.

Figure 9-16. Tracks of pairs of tropical cyclones over the Indian Ocean, 1964 to 1974(Mukerjee and Padmanabham, 1977).

Figure 9-17. Latitude at which initial disturbances that later became tropical storms werefirst detected in each development area. Average number of years of data are given inparentheses (adapted from Gray, 1968).

Figure 9-18. Frequency distribution of the annual number of tropical cyclones of tropicalstorm or greater intensity in various development areas.

Figure 9-19. Average zonal wind-shear (knots) between 850 mb and 200 mb over the tropicsfor January, April, July and October (after Gray, 1968).

Figure 9-20. Mean recurvature latitudes and ranges for western North Pacific tropicalcyclones, 1965-1982 (from Guard, 1983).

Figure 9-21. Mean monthly recurvature positions for tropical storms and hurricanes in theNorth Atlantic (after Colon, 1953).

Figure 9-22. Yearly averages of tropical cyclone position-forecast errors compared to besttracks, 1970 through 1989 for NHC, Miami (full lines), and JTWC, Guam (dashed lines).Heavy lines show linear trends.

Figure 9-23. How to use average root-mean-square vector errors to establish areal confi-dence limits for typhoon forecasts (see text for discussion).

Figure 10-1. Mean annual number of thunderstorm days over the globe (World Meteoro-logical Organization, 1956b).

Figure 10-2. One year of midnight lightning locations for 365 consecutive days from Sep-tember 1977 through August 1978 (Orville and Henderson, 1986).

Figure 10-3. Mean annual percentage of days with thunder by 5° longitude intervals for theequatorial belt between 10°N and 10°S (adapted from Ramage, 1968).

Figure 10-4. Mean annual number of thunderstorm days over southeast Asia (afterAtkinson, 1967a).

Figure 10-5. Mean annual number of thunderstorm days over India for 1931 to 1960. Indi-vidual station means are plotted (after Alvi and Punjabi, 1966).

Figure 10-6. The mean monthly and annual variability in the number of thunderstorm daysfor stations in Thailand.

A. Frequency distributions for selected categories of monthly means.

B. Standard deviations for selected categories of monthly means and line of best fit.

C. Cumulative frequency distributions of the monthly number of thunderstorm days ac-cording to the mean values.

D. As in C but for the annual values (adapted from Atkinson, 1967a).

Figure 10-7. Thunderstorm occurrence and duration at Mactan Air Base (10°N, 124°E).

A. Thunderstorm occurrence, June-August 1965.

B. Frequency distribution of thunderstorm durations considering all separate occurrences.

C. Frequency distribution of thunderstorm duration derived by considering successiveperiods of thunderstorm activity separated by less than two hours as one occurrence(adapted from Atkinson, 1967b).

Figure 10-8. Percentage frequency of occurrence of tornadoes over India, 1951 to 1980. Totalnumber — 42 (Singh, 1981).

Figure 10-9. Locations of reported funnels aloft (v) and waterspouts (tornadoes) () for theHawaiian Islands, 1961-1974. Thirty km radius circles are drawn around the major airports.The area around Oahu is enlarged below (Schroeder, 1976).

Figure 10-10. (A) Monthly, and (B) diurnal variation of the occurrence of funnel clouds,waterspouts, and tornadoes within 140 km of Miami, Florida, 1957 to 1966 (adapted fromGerrish, 1967).

Figure 10-11. Distribution of tornadoes, waterspouts and funnel clouds in the south Floridaarea, 1957 to 1966. Arrows show reported direction of movement (after Gerrish, 1967).

Figure 10-12. Distribution of hurricane tornadoes, 1948 to 1972 (Novlan and Gray, 1974).

Figure 10-13. Hurricane tornadoes over the United States, 1955 to 1964, located with refer-ence to the center and direction of movement of the hurricanes. Isopleths enclose percent-ages of the total number of cases (adapted from Pearson and Sadowski, 1965).

Figure 10-14. Frequency of hail occurrence at the surface in parts of the tropics (adaptedfrom Frisby and Sansom, 1967).

Figure 10-15. Hail regimes for three tropical zones. The X’s indicate months when hail at thesurface is most likely (adapted from Frisby, 1966).

Figure 10-16. Climatology of thunderstorm squalls and related phenomena at Nagpur(21°N, 79°E), India: (A) mean and maximum squall frequency, (B) monthly extreme gustspeeds with squalls, and frequency distribution of (C) maximum squall winds, (D) pressurechanges and (E) temperature changes associated with squalls (adapted from Sharma, 1966).

Figure 10-17. Vertical cross section of the evolution of the microburst

wind field. T is the time of initial divergence at the surface. Shading denotes the vectorwind speed (from Wilson et al., 1984).

Figure 10-18. DMSP IR image for 19 October 1984 of a duststorm generated by strongnorthwest winds (Shamal) over the Persian Gulf and northeast Arabia. The dust edge canbe seen south of the clouds of a subtropical jet stream. The northern edge of the duststormis hard to detect.

Figure 10-19. DMSP IR image for 25 June 1985, showing widespread dust covering Arabiaand the neighboring seas.

Figure 10-20. DMSP images for 25 August 1989, showing a duststorm extending from theNubian Desert over the Red Sea.

A. IR image. B. Visible image. C. Composite image (IR over land, visible over sea) (Brooks,1990).

Figure 10-21. Curve of best fit between maximum sustained surface wind speeds andminimum sea-level pressures for 76 western Pacific tropical cyclones (from Atkinson andHolliday, 1977).

Figure 10-22. Double-exponential distribution (full line) fitted to the annual peak gusts atKadena Air Base (26°N, 128°E), Okinawa, 1945 to 1988.

Figure 10-23. Expected extreme wind gusts (knots) for 2-year and 100-year return periodsfor selected tropical stations.

Figure 10-24. Nomogram for determining expected extreme wind-gusts for various returnperiods based on the 2-year and 100-year return-period values (based on the Gumbeldouble-exponential distribution).

Figure 10-25. Relative frequency of various types of turbulence versus altitude (afterSaxton, 1966).

Figure 10-26. Regression of maximum storm-surge height on central pressure for AtlanticCoast and Gulf Coast hurricanes (adapted from Hoover, 1957 and Conner et al., 1957).

Figure 10-27 Storm-tide height profiles for four hurricanes that entered the United StatesGulf coastline west of Tallahassee, Florida, from the south-southeast, south and south-southwest (adapted from Hoover, 1957).

Figure 10-28. Storm-surge heights on the United States coasts south of 35°N for great hurri-canes (central pressures 950 mb or less). Solid line gives the mean value and the dashedlines the limiting values. Double circle denotes two occurrences at that point (adapted fromSugg, 1969). Dash-dot line is computed storm-surge profile of a standard storm (see text)derived from Jelesnianski (1974). Note higher surges on the right-hand (on-shore winds)sides of the storms.

Figure 11-1. Hourly temperature, dew point, and rainfall for Canton Island for 12 to 13April, 1953 (after Palmer et al., 1955).

Figure 11-2. Example of cyclonic and anticyclonic cusps in an

east-to-west current (after Palmer et al., 1955).

Figure 11-3. Models of pure outdrafts, pure indrafts, and six types of vortexes possible instreamline analysis (after Palmer et al., 1955).

Figure 11-4. Wind data for kinematic analysis shown in Figure 11-5 (wind speeds in knots)(prepared by Dept. of Weather Training, Chanute AFB, Illinois).

Figure 11-5. Kinematic analysis based on wind data in Figure 11-4 (prepared by Dept ofWeather Training, Chanute AFB, Illinois).

Figure 11-6. Schematic cross section of two types of front.

A : Both cold and warm air rising.

B : Cold air sinking, warm air rising.

Figure 11-7. Schematic kinematic analysis of surface winds around a moving front (doubleline).

A : Cold and warm air rising.

B : Cold air sinking, warm air rising.

Vs = wind flow vector relative to the front; this changes by 180° through the front. Vf =frontal movement vector. V = resultant of Vs and Vf = actual wind (Ramage, 1957).

Figure 11-8. As in Figure 11-7, except that Vs does not change direction through the front.

Figure 11-9. Surface chart for 12 UTC 22 July 1956.

A : Isobaric analysis (mb).

B : Kinematic analysis (isotachs labeled in knots).

Figure 11-10. Chart for 250 mb winds (knots) and temperatures (°C) for 12 UTC 26 Novem-ber 1985). Stream function and isotachs shown. Areas with speeds greater than 70 knots,hatched.

Figure 11-11. Illustrating how AIREP wind speeds can be adjusted to analysis level usingthe computed wind-shear in the vertical from surrounding rawinsonde stations.

Figure 11-12. Root-mean-square variability (knots) of upper-air wind speeds as a functionof wind speed and time or distance (adapted from Lenhard, 1967).

Figure 11-13. Illustrating how AIREP positions can be adjusted according to the movementof synoptic-scale features.

Figure 11-14. Chart of the mean positions of major circulation features over the westernNorth Pacific and south Asia for October.

Figure 11-15. Relationship between the monthly mean resultant wind speed at the gradientlevel and the steadiness. Dashed lines enclose departures of less than +/- 10 percent from

the line of best fit (Atkinson and Sadler, 1970).

Figure 11-16. Relationship between the monthly mean resultant wind speed at the gradientlevel and the percent of time that the wind is within +/- 45° of the resultant direction.Dashed lines enclose departures of less than +/- 10 percent from the line of best fit(Atkinson and Sadler, 1970).

Figure 11-17. 850-mb kinematic analysis and nephanalysis for the western North Pacificand Asia for 00 UTC 26 July 1967. Wind reports that are probably wrong or not representa-tive on this analysis scale, are indicated by “X”s. Wind speeds in knots (from analysesmade at the University of Hawaii).

Figure 11-18. As in Figure 11-17, but for 500-mb.

Figure 11-19. As in Figure 11-17, but for 200-mb.

Figure 11-20. Recommended plotting model for composite lower-tropospheric wind charts.

Figure 11-21. January long-term mean (A) ship winds and (B) satellite cloud motion vec-tors; (C) wind shear in the vertical between the two data sets (m s-1) (Sadler and Kilonsky,1985).

Figure 11-22. Recommended plotting models for composite upper-tropospheric windcharts. Barbs can be substituted for the numerals to indicate wind speed.

Figure 11-23. Objective kinematic analyses for 00 UTC 2 December 1978. Wind speed inknots. A : at 950-mb. B : at 200-mb (Davidson and McAvaney, 1981).

Figure 11-24. Vertical time cross section of winds (in knots) above Johnston Island (17°N,170°W), 2-6 May 1954. Circled report is probably wrong (prepared by Dept. of WeatherTraining, Chanute AFB, Illinois).

Figure 11-25. Checkerboard plot of hourly surface observations at Don Muang (14°N,101°E), Thailand, 19 to 31 January 1968 (prepared by Dept. of Weather Training, ChanuteAFB, Illinois).

Figure 12-1. Monthly climatologies of weather variables at Saigon (11°N, 107°E).

Figure 12-2. Conditional climatology summaries for Saigon (11°N, 107°E) and Danang(16°N, 108°E).

Figure 12-3. Climatological display for Saigon (11°N, 107°E) in August.

Figure 12-4. Mean vector errors of the wind at 300 mb over the tropical Pacific during May1959 for climatology, persistence and forecasts based on 50 percent climatology plus 50percent persistence (after Lavoie and Wiederanders, 1960).

Figure 12-5. Relationship between maximum wind speed in the lowest 3 km at 08 LT andmaximum surface wind gust observed in the subsequent 12

hours at Mactan Air Base (10°N, 124°E), Philippines (First Weather Wing, 1968).

Figure 12-6. Hourly radar index percent coverage of radar echoes within 50 NM (90 km) ofstation) for Saigon (11°N, 107°E) for June, July and August 1967 (after Conover, 1967).

Figure 12-7. Kinematic analyses of the gradient-level resultant wind for: (A) days withabove average (left) and (B) below average (right) radar-index values for the Saigon area(shown by shading). Isotachs are labeled in knots (after Conover, 1967).

Figure 12-8. Percent probability of: (A) ceiling/visibility less than 1500 feet (460 m) and/or 3miles (5 km), and (b) less than 200 feet (60 m) and/or a half-mile (800 m) during a one-hourperiod at Kadena Air Base (26°N, 128°E), Okinawa, in relation to tropical cyclone-centerlocation (adapted from Atkinson and Penland, 1967).

Figure 12-9. Relation between Showalter-Index and Lifted-Index values at 06 LT and prob-ability of thunderstorms during the following 24 hours at New Delhi (29°N, 77°E), andwithin radii of 50 and 100 NM (93 and 185 km) (adapted from Subramaniam and Jain,1966).

Figure 12-10. Reported maximum thunderstorm-produced surface wind gusts (knots) atstations in the southeastern United States and southeast Asia according to observed valuesof (A) the Delta-T Index, and (B) the T1 Index (adapted from R.C. Miller, 1967).

Figure 12-11. Radar-echo frequencies measured simultaneously by 3-cm radars located atUdon (17°N, 103°E) and Ubon (15°N, 105°E). for June-August, 1967 and 1968. The doubleline is equidistant from each station; the heavy line joins points of equal echo-frequency(adapted from Ing, 1971).

Figure 12-12. A 15-degree logarithmic spiral overlay for estimating the center positions oftropical cyclones from the orientation of the storms’ spiral rainbands (after Senn et al.,1957).

Figure 12-13. Hourly center positions of Hurricane Carla, 10-12 September 1961, based onradar fixes from Galveston (29°N, 95°W), Texas WSR-57 radar (adapted from Jordan,1963b).

Figure 12-14. Kinematic model of a hurricane spiral rainband. A letter identifies an echoand the subscript indicates units of time. Relative to the hurricane eye, new echoes developradially outward (not necessarily at the same azimuth), but once formed the echoes movein circular paths and eventually dissipate at the downward end of the rainband (afterHardy et al., 1964).

Figure 12-15. Statistics of large radar-echo groups (greater than 50 NM (93 km) wide)) nearSaigon (11°N, 107°E) during the southwest monsoon: (A) echo-group duration, (B) level ofbest agreement between echo-group motion and wind direction, and (C) ratio of echo-group speed to wind-speed at level of best agreement (adapted from Conover, 1967).

Figure 12-16. Surface visibility (V) in showers related to reflectivity (Z) measured by radar.Dashed lines define limits of V-R relationships based on various observational studies(adapted from Wilson, 1968; Cataneo, 1969, and ESSA, 1967).

Figure 12-17. Radar climatology for New Delhi (29°N, 77°E): (A) monthly distribution ofecho tops, (B) time of occurrence of maximum areal coverage by month, (C) frequencydistributions of echo sizes, and (D) inter-spacing of echoes during the hot season (March-May) (adapted from Kulshrestha and Jain, 1968).

Figure 12-18. Radar climatology of the southeast Florida area; (A) areas and grid used tocompile the climatology — A = land, B = water. (B) diurnal variation of echo frequency, (C)percent frequency variation by time of day and distance from the coast, and (D) relativefrequencies of heights of echo tops for 12 to 18 LT (adapted from Gerrish (1971).

Figure 12-19. Wind forecasts for March and April 1991 at 32 tropical stations. Comparisonaf NMC numerical prediction of 850 mb winds 24 hours in advance to objective — 1/2

(persistence + climatology); persistence; and climatology predictions. Mean vector errorsare given in knots.

Figure 12-20. As in Figure 12-19, but for 200 mb.

Figure 12-21. Composite 30-60 day anomalies of outgoing long wave radiation between theequator and 10°N (isopleth interval 5 W m-1) for May to October (left side) and Novemberto April (right side). Ordinate extends over approximately a 60-day interval (from Knutsonand Weickmann, 1987).

Figure 12-22. Fifteen-year running mean correlation coefficients between: (1) average ofApril-May surface pressure at three South American stations near 30°S and the subsequentJune-September rainfall over Peninsular India (thick line); (2) the difference between theJanuary mean sea-level pressures at Irkutsk (52°N, 104°E) and Tokyo (36°N, 140°E) and thesubsequent May-October rainfall at Hong Kong (22°N, 114°E). Values are plotted at themid-years of the 15-year periods.

Figure 12-23. Predictions of North Atlantic hurricanes and tropical storms made each Mayfor the following season by Gray (1990) (full line), compared to the outcomes (dashed lines)and to climatology (dot-dashed lines). Upper curves show number of days with tropicalcyclones, lower curves show number of tropical cyclones.

Figure 12-24. Iterative procedure for evaluating and modifying case studies through fore-cast performance.

TTTTTABLESABLESABLESABLESABLES

2-1. Poleward transport of heat energy by ocean currents according to Sellers (1965) (S) and VonderHaar and Oort (1973) (VO). Units of 1013 W.

3-1. Implementation of WMO synoptic observing program, 1988 (from World Meteorological Organi-zation, 1989).

3-2. Implementation of WMO radiosonde observing program, 1987 (from World Meteorological Orga-nization, 1989).

3-3. Operational weather satellites (from National Aeronautics and Space Administration, 1988; Yu,1990).

3-4. Ground weather radar stations in 1988 (from World Meteorological Organization, 1989).

3-5. Proposed satellite missions (from National Aeronautics and Space Administration, 1988).

3-6. Tropical stations, beyond the continental United States, at which 10 cm Doppler radars (WSR-88D) are being installed.

5-1. Weather effects on electrooptical devices (Seagraves, 1989).

5-2. Normal Atmospheres for 30°N in January and July and 15°N Annual.

6-1. Starting years of El Niño and anti-Niño events (1877-1988) (from Kiladis and Diaz, 1989).

6-2. Relationship of moderate or strong Niño to same year Indian summer monsoon rainfall, 1875-1980 (Ramage, 1983).

6-3. Advance of Indian summer monsoon.

6-4. Data for the universal rainfall distribution curve (from Martin, 1964).

6-5. Examples of applying the universal rainfall distribution curve to determine distribution of dailyrainfall. A given month has an average of 20 days with rain and a mean rainfall of 10 inches (250 mm).The frequency of various daily rainfalls during a 10-year period are estimated below.

6-6. Empirical relationships between the mean annual rainfall given by X and the mean number ofhours per year given by Y that various clock-hour rainfalls are equalled or exceeded. The standarderrors of estimate (in hours) and the correlation coefficient between the estimated and observed fre-quencies are also shown (after Winner, 1968).

8-1. Average monthly and seasonal frequency of Bay of Bengal monsoon depressions for the period1891-1970 (Rao, 1976). Depressions have winds

up to 33 knots, storms from 33 to 62 knots.

9-1. Areas of occurrence of intense tropical cyclones and regional terminology (obtained from WorldMeteorological Organization, 1979).

9-2. Monthly and average number of tropical cyclones per year for each major basin (from Crutcherand Quayle, 1974).

9-3. 1989 error statistics for selected objective techniques in the western North Pacific — 24-hourmean forecast error (nautical miles) (Joint Typhoon Warning Center, 1990).

9-4. Average errors (knots) and trends in tropical cyclone position forecasts made at JTWC, Guam(13°N 145°E) and NHC, Miami (26°N, 80°W) (for 1970 through 1989).

9-5. Average errors in 24-hour persistence/climatology movement forecasts of tropical cyclone centersfor tropical cyclone basins (Pike and Neumann, 1987).

10-1. Thunderstorm duration — the ratio of the mean annual number of hourly observations reportingthunderstorms to the mean annual number of thunderstorm days for stations in southeast Asia.

10-2. Hurricane-tornado forecast worksheet (Novlan and Gray, 1974).

10-3. Frequency of duststorms at Khartoum (17°N, 33°E) based on eight years of observations (fromSutton, 1925).

10-4. Saffir/Simpson Damage Potential Scale ranges.

10-5. Percent frequency of various categories of CAT by five-degree latitude-longitude rectangles inthe tropics for flights above 20,000 feet (6 km) (for the 1964-65 ICAO collection periods).

11-1. Relation between Beaufort wind-scale number and wind speed (knots) as included in synopticreports (from World Meteorological Organization, 1966b).

11-2. Ratio between wind speed at various levels to the wind speed at 66 feet (20 m) elevation, basedon the one-seventh power law.

11-3. Root-mean-square (RMS) pressure-height errors (meters) arising in radiosonde observations withan RMS temperature error of 1.0°C and an RMS pressure error of 2.0 mb.

11-4. Root-mean-square errors (knots) in rawinsonde wind speeds according to altitude and the magni-tude of the mean-wind vector (Meteorology Working Group, 1965).

11-5. Percentage frequency of modal wind directions differing by less than one compass point frommean resultant-wind directions (after Wiederanders, 1961).

12-1. March and April 1991 tropical wind forecasting. Average errors (knots) of 24-hour forecasts at850 and 200 mb made by NMC, 1/2(persistence + climatology), persistence, and climatology; strati-fied by latitude.

REFERENCES

Adams, C.W., 1964: Monthly upper-air means for selected stations of the central Pacific 1956-1959.Hawaii Inst. Geophys., HIG 64-24.

Air Weather Service, 1955: Accuracies of radiosonde data. AWS TR-105-133, 12 pp.

Air Weather Service, 1960: Use of asynoptic data in analysis and forecasting. AWS TR-225, 27 pp.

Air Weather Service, 1969: Application of meteorological satellite data in analysis and forecasting.AWS TR-212, 233 pp.

Alaka, M.A., 1964: Problems of tropical meteorology. WMO Tech. Note 62, 36 pp.

Albright, M.D., E.E. Recker and R.J. Reed, 1985: The diurnal variation of deep convection in thecentral tropical Pacific during January-February 1979. Mon. Wea. Rev., 113, 1163-1680.

Alexander, M.A., and S.D. Schubert, 1990: Regional earth-atmosphere energy balance estimates basedon assimilations with a GCM. J. Climate, 3, 15-31.

Alvi, S.M.A., and K.G. Punjabi, 1966: Diurnal and seasonal variations of squalls in India. Indian J.Meteor. Geophys., 17, 207-216.

Ananthakrishnan, R., V.R. Acharya and A.R. Ramakrishna, 1967: On the criteria for declaring theonset of the southwest monsoon over Kerala. India Meteor. Dept. Forecasting Manual Part IV; 18:Monsoons of India. 20 pp.

Ananthakrishnan, R., and J.M. Pathan, 1970: North-south oscillations of the equatorial trough andseasonal variations of rainfall in the tropics. Proc. Symp. Tropical Meteor. Honolulu, Amer. Meteor.Soc., FV-1-6.

Anthes, R.A., 1982: Tropical cyclones, their evolution, structure and effects. Meteor. Monogr., 19, 208pp.

Arakawa, H., 1940: The formation of hurricanes in the South Pacific and the outbreaks of cold air fromthe North Polar regions. J. Meteor. Soc. Japan, Ser.2, 1, 1-6.

Ardanuy, P., 1979: On the observed diurnal oscillation of the Somali jet. Mon. Wea. Rev., 107, 1694-1700.Argulis, R.P., 1970: Tsunami. ESSA Tech. Memo, WBTM-WR-48, 25 pp.

Aspliden, C.I., G.A. Dean and H. Landers, 1965-1967: Satellite Study, Tropical North Atlantic, 1963.Dept. Meteor. Florida State Univ.

Atkinson, G.D., 1966: A preliminary estimate of extreme wind speeds in Thailand. 1st Weather WingTech. Study, 3, 26 pp.Atkinson, G.D., 1967a: Thunderstorms in southeast Asia. 1st Weather Wing Tech. Study, 11, 55 pp.

Atkinson, G.D., 1967b: Climatology of significant weather at Mactan AB, Philippines. 1st WeatherWing Tech. Study, 13, 15 pp.

Atkinson, G.D., 1968: Extreme daily rainfall amounts in the Republic of Vietnam. 1st Weather WingTech. Study, 16.

Atkinson, G.D., and C.R. Holliday, 1977: Tropical cyclone minimum sea level pressure/maximumsustained wind relationship for the western North Pacific. Mon. Wea. Rev., 105, 421-427.

Atkinson, G.D., and H.E. Penland, 1967: Typhoon weather models — Kadena AB, Okinawa. 1stWeather Wing Tech. Study, 12, 19 pp.

Atkinson, G.D., and J.C. Sadler, 1970: Mean cloudiness and gradient-level wind charts over the trop-ics. Vol. I (text), Vol. II (charts). AWS Tech. Rept., 215.

Atlas, D., 1953: Optical extinction by rainfall. J. Meteor., 10, 486-488.

Atlas, R., S.C. Bloom, R.N. Hoffman, J.V. Ardizzone and G. Brin, 1991: Space-based surface windvectors to aid understanding of air-sea interactions. Eos, 72, 18, 201-208.

Augustine, J.A., 1984: The diurnal variation of large-scale inferred rainfall over the tropical PacificOcean during August 1979. Mon. Wea. Rev., 112, 1745-1751.

Baer, F., 1956: The influence of illumination on cirrus reports. Bull. Amer. Meteor. Soc., 37, 366-367.

Barnes, G.M., 1990: 18th Conference on hurricanes and tropical meteorology. Bull. Amer. Meteor.Soc., 71, 558-570.

Barry, R.G., 1981: Mountain Weather and Climate. Methuen, London, 313 pp.

Battan, L.J., 1959: Radar Meteorology. U. Chicago Press, 161 pp.

Bedient H.A., and J. Vederman, 1964: Computer analysis and forecasting in the tropics. Mon. Wea.Rev., 92, 565-567.

Bell, G.J., 1976: Seasonal forecasts of Hong Kong summer rainfall. Weather, 31, 208-212.

Bergeron, T., 1954: Review of tropical hurricanes. Quart. J. Roy. Meteor. Soc., 80, 131-164.

Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev., 97, 163-172.

Braak, C., 1921-29: Het climaat van Nederlandsch Indie. Magnet. Meteor. Observ., Batavia, Verhand.,8.

Brand, S., and C.P. Guard, 1979: An observational study of extratropical storms evolved from tropical

cyclones in the western North Pacific. J. Meteor. Soc. Japan, 57, 479-482.

Brier, G.W., and J. Simpson, 1969: Tropical cloudiness and rainfall related to pressure and tidal varia-tions. Quart. J. Roy. Meteor. Soc., 95, 120-147.

Brill, K., and B. Albrecht, 1982: Diurnal variation of the trade-wind boundary layer. Mon. Wea. Rev.,110, 601-613.

Brode, R.W., and M.K. Mak, 1978: On the mechanism of the monsoonal mid-tropospheric cycloneformation. J. Atmos. Sci., 35, 1473-1484.

Brook, R.R., 1985: Koorin nocturnal low-level jet. Bound. Layer Meteor., 32, 133-154.

Brooks. B., 1987: Forecasting the “Atemporalado” in Honduras. AWS/FM-87/001, 9pp.

Brooks, B., 1990: Personal communication.

Brooks, C.E.P., and N. Carruthers, 1953: Handbook of Statistical Methods in Meteorology. Her Majesty’sStationery Office, London, 412 pp.

Browner, S.P., W.L. Woodley and C.G. Griffith, 1977: Diurnal oscillation of the area of cloudinessassociated with tropical storms. Mon. Wea. Rev., 105, 856-864.

Bryson, R.A., and P.M. Kuhn, 1961: Stress-differential induced divergence with application to littoralprecipitation. Erdkunde, 15, 287-294.

Buchan, A., 1883: Meteorology. Encyl. Britannica, 9th Ed., Vol. 16, 114-159.

Bunting, D.C., 1968: Ship anemometer heights above mean water level. Naval Oceangr. Off. InformalRept. 68-74, 18 pp.

Burpee, R.W., 1975: Some features of synoptic-scale waves based on compositing analysis of GATEdata. Mon. Wea. Rev., 103, 921-925.

Businger, S., 1991: Arctic hurricanes. Amer. Scientist, 79, 18-33.

Byers, H.R., 1974: General Meteorology, Fourth Edition. McGraw-Hill, New York, 461 pp.

Byers, H.R., and R.R. Braham, 1949: The Thunderstorm. U.S. Govt. Printing Off., Washington, D.C.,287 pp.

Cadet, D., and M. Desbois, 1981: A case study of a fluctuation of the Somali jet during the Indiansummer monsoon. Mon. Wea. Rev., 109,182-187.

Caracena, F., R.L. Holle and C.A. Doswell III, 1990: Microbursts: A Handbook for Visual Identifica-tion, Second Edition. Dept. of Commerce, 35 pp.

Carlson, T.N., 1967: Structure of a steady-state cold low. Mon. Wea. Rev., 95, 763-777.

Carlson, T.N., 1969a: Synoptic histories of three African disturbances that developed into Atlantichurricanes. Mon. Wea. Rev., 97, 256-276.

Carlson, T.N., 1969b: Some remarks on African disturbances and their progress over the tropical At-lantic. Mon. Wea. Rev., 97, 716-726.

Carr, F.H., 1977: Mid-tropospheric cyclones of the summer monsoon. Pure Appl. Geophys., 115, 1383-1412.

Carter, E.A., and L.A. Schuknecht, 1969: Peak wind statistics associated with thunderstorms at CapeKennedy, Florida. NASA Contr. Rept. No. 61304, 21 pp.

Cataneo, R., 1969: A method for estimating rainfall rate radar reflectivity relationships. J. Appl. Me-teor., 8, 815-819.

Caviedes, C.N., 1975: El Niño 1972: its climatic, ecological, human and economic implications. Geogr.Rev., 65, 493-509.

Chaggar, T.S., 1984: Réunion sets new rainfall records. Weather, 39, 12-14.

Chan, J.C.L., and G.J. Holland, 1989: Observing and forecasting tropical cyclones — where next?Bull. Amer. Meteor. Soc., 70, 1560-1563.

Chang, C.-P., J.E. Erickson and K.M. Lau, 1979: Northeasterly cold surges and near-equatorial distur-bances over the winter MONEX area during December 1974; Part I: synoptic aspects. Mon. Wea.Rev., 107, 812-829.

Chang, C.-P., and K.M. Lau, 1980: Northeasterly cold surges and near-equatorial disturbances over thewinter MONEX area during December 1974. Part II: planetary-scale aspects. Mon. Wea. Rev., 108,298-312.

Chang, C.-P., and K.M. Lau, 1982: Short-term planetary scale interactions over the tropics andmidlatitudes during northern winter. Part I: contrasts between active and inactive periods. Mon. Wea.Rev., 110, 933-946.

Chang, C.-P., J.E. Millard and G.T.J. Chan, 1983: Gravitational character of cold surges during winterMONEX. Mon. Wea. Rev., 111, 293-307.

Chen, T.Y., 1969: The severe rainstorms in Hong Kong during June 1966. Roy. Obs. Hong Kong,Suppl. Meteor. Results, 79 pp.

Chen, Y.-L., and T.A. Schroeder, 1986: The relationship among localwinds, synoptic-scale flow and precipitation at Hilo during HAMEC. Int. Conf. Monsoon and Mesos-cale Meteor. Taipei, Taiwan, 5 pp.

Cheng, P.T.T., and W.P. Kwang, 1973: Radar climatology of Hong Kong for the years 1967-1969. Roy.Obs. Hong Kong Tech. Note, 34, 26 pp.

Cheng, S.S.M., 1978: Dry spells during mid-summer over south China. Roy. Obs. Hong Kong Tech.Note, 44, 35 pp.

Chin, P.C., and M.H. Lai, 1974: Monthly mean upper winds and temperatures over southeast Asia andthe western North Pacific. Roy. Obs. Hong Kong Tech. Mem., 12, 115 pp.

Christian, H.J., R.J. Blakeslee and S.J. Goodman, 1989: The detection of lightning from geostation-ary orbit. J. Geophys. Res., 94, 13329-13337.

Chu, P.-S., and J. Frederick, 1990: Westerly wind bursts and surface heat fluxes in the equatorial west-ern Pacific in May 1982. J. Meteor. Soc. Japan, 68, 523-537.

Climate Analysis Center, 1989: Climate Diagnostics Bulletin, August 1989. NOAA/NWS/NMC, 57pp.

Cobb, L.G., 1966: Maximum precipitation values on the lower slopes of mountainous terrain. Re-search on Tropical Rainfall Patterns and Associated Mesoscale Systems. Dept. Meteor., Texas A&MUniv.

Cobb, L.G., 1967 and 1968: The El Niño phenomenon and rainfall in Peru and Ecuador. Research onTropical Rainfall Patterns and Associated Mesoscale Systems, Repts. 2 and 3. Dept. Meteor., TexasA&M Univ.

Cobb, L.G., 1968: The annual and daily distribution of rainfall in southeast Asia. Research on TropicalRainfall Patterns and Associated Mesoscale Systems. Dept. Meteor., Texas A&M Univ., 53-75.

Cochemé, J., and P. Franquin, 1967: An agroclimatology survey of a semi-arid area in Africa south ofthe Sahara. WMO Tech. Note, 86.

Cole, A.E., and R.J. Donaldson, 1968: Precipitation and clouds: a revision of Chapter 5, Handbook ofGeophysics and Space Environments. A.F. Surv. Geophys., 212, 1-108.

Coligardo, M., 1967: An investigation of rainfall variability and distribution in Luzon and a mesoscalestudy of rainfall of the province of Laguna and adjacent areas, Philippines. Research on TropicalRainfall Patterns and Associated Mesoscale Systems. Dept. Meteor., Texas A&M Univ.

Colòn, J.A., 1953: A study of hurricane tracks for forecasting purposes. Mon. Wea. Rev., 81, 53-66.

Condray, P.M., and R.T. Edson, 1989: The Caribbean basin. An electrooptical climatology for the 8-12micron band. Vol.1. Central America. USAFETAC/TN-89/004, 159 pp.Conner, W.C., R.H. Craft and D.L. Harris, 1957: Empirical methods for forecasting the maximumstorm tide due to hurricanes and other tropical storms. Mon. Wea. Rev., 85, 113-116.

Conover, J.H., 1967: Studies of clouds and weather over southeast Asia. AFCRL Meteor. Branch [notpublished].

Conover, J.H., and J.C. Sadler, 1960: Cloud patterns as seen from altitudes of 250 to 350 miles -preliminary results. Bull. Amer. Meteor. Soc., 41, 291-297.

Crutcher, H.L., and O.M. Davis, 1969: U.S. Navy Marine Climatic Atlas of the World, Vol. VIII, theWorld. Naval Weather Service Command, NAVAIR 50-1C-54.

Crutcher, H.L., and R.G. Quayle, 1974: Mariners’ Worldwide Climatic Guide to Tropical Storms atSea. NOAA/EDS National Climatic Center, 114 pp and 312 charts.

Daniel, C.E.J., and A.H. Subramaniam, 1966: Exceptionally heavy rain in south Kerala on 17/18 Octo-ber, 1964. Indian J. Meteor. Geophys., 17, 253-256.

Davidson, N.E., and B.J. McAvaney, 1981: The ANMRC tropical analysis scheme. Aust. Meteor.Mag., 29, 155-168.

Davidson, N.E., and G.J. Holland, 1987: A diagnostic analysis of two intense monsoon depressionsover Australia. Mon. Wea. Rev., 115, 380-392.Defant, F., 1951: Local winds. Comp. Meteor., 655-672.

De Maria, M., M.B. Lawrence and J.T. Kroll, 1990: An error analysis of Atlantic triopical cyclonetrack guidance models. Wea. Forecast., 5, 47-61.

Deshpande, D.V., 1964: Heights of Cb clouds over India during the southwest monsoon season. IndianJ. Meteor. Geophys., 15, 47-54.

Deshpande, D.V., 1965: Cirriform clouds over India — heights and temperatures. Indian J. Meteor.Geophys., 16, 635-644.

Detachment 8, 20th Weather Squadron, 1988: Typhoon Information Package. USAFETAC, 14 pp.

Dhar, O.N., 1977: Cherrapunji breaks the world rainfall record for 4-day duration and a list of heavyrainfall sataions in India. Vayu Mandal, July-Dec. 1977, 44-45.

Dhar, O.N., and P.R. Mhaisker, 1973: Areal and point distribution of rainfall associated with depres-sions/storms on the day of crossing the east coast of India. Indian J. Meteor, Geophys., 24, 271-278.

Douglas, M.W., 1987: The structure and dynamics of monsoon depressions. Dept. Meteor., FloridaState Univ., Rept. 86-15, 273 pp.Drosdowsky, W., and G.J. Holland, 1987: North Australian cloud lines. Mon. Wea. Rev., 115, 2645-2659.

Dunn, G.E., 1940: Cyclogenesis in the tropical Atlantic. Bull. Amer. Meteor. Soc., 21, 215-229.

Duvel, J.P., 1989: Convection over tropical Africa and the Atlantic Ocean during northern summer.

Part I: interannual and diurnal variations. Mon. Wea. Rev., 117, 2782-2799.

Dvorak, V.F., 1975: Tropical cyclone intensity analysis and forecasting from satellite imagery. Mon.Wea. Rev., 103, 420-430.

Dvorak, V.F., 1984: Tropical cyclone intensity analysis using satellite data. NOAA Tech. Rept. NESDIS11, 47 pp.

Dvorak, V.F., and F. Smigielski, 1990: A Workbook on Tropical Clouds and Cloud Systems Observedin Satellite Imagery. NOAA/NESDIS.

Edson, R.T., and P.M. Condray, 1989: The Caribbean basin. An electrooptical climatology for the 8-12 micron band. Vol. III. Northern South America. USAFETAC/TN-89/006, 250 pp.

Eigsti, S.L., 1978: The coastal diurnal wind cycle at Port Aransas, Texas. Atmos. Sci. Group, Univ.Texas Rept. 48, 79 pp.

Eldridge, R.H., 1957: A synoptic study of West African disturbances. Q. J. Roy. Meteor. Soc., 83, 303-314.

Eliot, J., 1895: A preliminary discussion of certain oscillatory changes of pressure of long period and ofshort period in India. Mem. India Meteor. Dept., 16, 185-307.

Elrick, J.R., and A.C. Meade, 1987: Weather sensitivities of electrooptical weather systems. AWS/TN-87/003, 54 pp.

Elsberry, R.L. (ed.), 1988: A Global View of Tropical Cyclones. Naval Postgraduate School, 185 pp.

Elsberry, R.L., 1989: ONR tropical cyclone motion research initiative; field experiment planningworkshop. Naval Postgraduate School, 6B-89-002, 79pp.

Emmanuel. K.A., and R. Rotunno, 1989: Polar lows as Arctic hurricanes. Tellus, 41A, 1-17.

Enfield, D.B., 1989: El Niño, past and present. Rev. Geophys., 27, 159-187.

ESSA, NASA and USAF, 1967: U.S. Standard Atmosphere Supplement, 1966. General Post Office,Washington, D.C., 289 pp.

Estoque, M.A., 1962: The sea breeze as a function of the prevailing synoptic situation. J. Atmos. Sci.,19, 244-250.Fernandez, W., 1982: Environmental conditions and structure of the West African and eastern tropicalAtlantic squall lines. Arch. Meteor. Geophys. Bioclimatol., Ser. A, 31, 71-89.

Fett, R.W., 1964: Aspects of hurricane structure: new model considerations suggested by TIROS andProject Mercury observations. Mon. Wea. Rev., 92, 43-60.

Fett, R.W., and W.A. Bohan, 1986: Navy Tactical Applications Guide, Volume 6, Part 1 — Tropics,

Weather Analysis and Forecast Applications, Meteorological Satellite Systems. Bohan, Park Ridge,Illinois.

Findlater, J., 1977: Observational aspects of the low-level cross-equatorial jet stream. Pure Appl.Geophys., 115, 1251-1262.

Fingerhut, W.A., 1978: A numerical model of a diurnally varying tropical cloud cluster disturbance.Mon. Wea. Rev., 106, 255-264.

First Weather Wing, 1968: Objective method to forecast gusty surface winds at Mactan AB, Philip-pines. Tech. Study, 18, 9 pp.

Fletcher, R.D., 1949: A hydrometeorological analysis of Venezuelan rainfall. Bull. Amer. Meteor.Soc., 30, 1-9.

Fletcher, R.D., 1950: A relationship between maximum observed point and areal rainfall values. Trans.Amer. Geophys. Union, 31, 344-348.

Flohn, H., 1961: Some remarks on long-range forecasting in tropical and subtropical regions. Proc.Symp. Tropical Meteorology in Africa. Munitalp Foundation, Nairobi, 221-225.

Flohn, H., 1964: Investigations on the tropical easterly jet. Bonner Meteor. Abhand., 4, 69 pp.

Flohn, H., 1965: Studies on the meteorology of tropical Africa. Bonner Meteor. Abhand., 5, 57 pp.

Flohn, H., 1971: Tropical circulation patterns. Bonner Meteor. Abhand., 15, 1-55.

Fox, T., 1969: An example of a medium level westerly wave over South and Central Africa. Lusaka,Meteor. Notes, Ser. A, 2.

Frank, N.L., 1969: The inverted V cloud pattern — an easterly wave? Mon. Wea. Rev., 97, 130-140.

Frank, N.L., 1970: On the energetics of cold lows. Proc. Symp. Tropical Meteorology. Honolulu,Amer. Meteor. Soc., EIV-1-6.

Frank, W.M., 1976: The structure and energetics of the tropical cyclone. Dept. Atmos. Sci., ColoradoState Univ., Paper 258.

Frisby, E.M., 1966: Hail incidence in the tropics. US Army Electr. Comm. Tech. Rept., 2768, 7 pp.Frisby, E.M., and H.W. Sansom, 1967: Hail incidence in the tropics. J. Appl. Meteor., 6, 339-354.

Frolow, S., 1942: On synchronous variations of pressure in tropical regions. Bull. Amer. Meteor. Soc.,23, 239-254.

Fu, R., A.D. Del Genio and W.B. Rossow, 1990: Behavior of deep convective clouds in the tropicalPacific deduced from ISCCP radiances. J. Climate, 3, 1129-1152.

Fujita, T.T., and F. Caracena, 1977: An analysis of three weather-related aircraft accidents. Bull. Amer.Meteor. Soc., 58, 1164-1181.

Fujita, T., T. Izawa, K. Watanabe and I. Imai, 1967: A model of typhoon accompanied by inner andouter rainbands. J. Appl. Meteor., 6, 3-19.

Fujita, T.T., K. Watanabe, K. Tsuchiya and M. Shimada, 1972: Typhoon-associated tornadoes in Japanand new evidence of suction vortices in a tornado near Tokyo. J. Meteor. Soc. Japan, 50, 431-453.

Fukuda, H.T., 1968: The end of the rainy season in the III Corps Tactical Zone of the Republic ofVietnam. 1st Wea. Wing Sci. Serv. Spec. Tech. Rept.

Fulks, J.R., 1951: The instability line. Comp. Meteor., 647-652.

Fullerton, C.M., and S.K. Wilson, 1975: An analysis of four showers with rainfall rates > 250 mm/hr.Dept. Meteor. Univ. Hawaii, UHMET 75-04, 45 pp.

Gabites, J.F., 1963: Low latitude analysis. Proc. Interregional Seminar on Tropical Cyclones. JapanMeteor. Agency Tech. Rept. 21, 239-244.

Gage, K.S., B.B. Balsley, W.L. Ecklund, R.F. Woodman and S.K. Avery, 1990: Wind-profiling dopplerradars for tropical atmospheric research. Eos, 71, 1851-1854.

Galway, J.G., 1956: The lifted index as a predictor of latent instability. Bull. Amer. Meteor. Soc., 37,528-529.

Gan, T.L., 1963: A study of some heavy rainfalls on the east coast of Malaya during the northeastmonsoon season. Mem. Malayan Meteor. Serv., 6, 15 pp.

Garcia, O., 1985: Atlas of Highly Reflective Clouds for the Global Tropics: 1971-1983. NOAA/ERL,Boulder, Colorado, 365 pp.

Garrett, A.J., 1980: Orographic cloud over the eastern slopes of Mauna Loa volcano, Hawaii, related toinsolation and wind. Mon. Wea. Rev., 108, 931-941.

Garstang, M., 1959: A study of the rainfall distribution of Trinidad, West Indies. Woods Hole Oceanogr.Inst., 60-33.

Garstang, M., 1966: Atmospheric scales of motion and rainfall distribution. Proc. 1966 Army Conf. onTropical Meteorology. Inst, Mar. Sci., Univ. Miami, 24-35.

Garstang, M., N.E. LaSeur and C. Aspliden, 1967: Equivalent potential temperature as a measure ofthe structure of the tropical atmosphere. Dept. Meteor., Florida State Univ. Rept., 67-10.

Garstang, M., and 11 other authors, 1988: Trace gas exchanges and convective transports over theAmazonian rain forest. J. Geophys. Res., 93, 1528-1550.

Garstang, M., and 11 other authors, 1990: The Amazon boundary-layer experiment (ABLE 2B): ameteorological perspective. Bull. Amer. Meteor. Soc., 71, 19-32.

Gentry, R.C., 1983: Genesis of tornadoes associated with hurricanes. Mon. Wea. Rev., 111, 1793-1805.

Gerrish, H.P., 1967: Mesoscale Studies of Instability Patterns and Winds in the Tropics. Inst. Mar. Sci.,Univ. Miami, 72 pp.

Gerrish, H.P., 1971: A model of summer convection in south Florida. J. Appl. Meteor., 10, 949-957.

Ghosh, B.P., 1967: A radar study on thunderstorms and convective clouds around New Delhi duringthe southwest monsoon season. Indian J. Meteor. Geophys., 18, 391-396.

Giambelluca, T.W., 1986: The structure of cirrus surges in the tropical central and eastern Pacificduring January and February 1979 [unpublished manuscript].

Giambelluca, T.W., M.A. Nullet and T.A. Schroeder, 1986: Rainfall Atlas of Hawaii. State of Hawaii,Dept. Land, Natural Resources, 267 pp.

Gleeson, T.A., 1954: Cyclogenesis in the Mediterranean region. Arch. Meteor. Geophys, Bioclimatol.,Ser.A, 6, 154-171.

Godske, C.L., T. Bergeron, J. Bjerknes and R.C. Bundgaard, 1957: Dynamic Meteorology and WeatherForecasting. Amer. Meteor. Soc., 800 pp.

Goldberg, R.A., M. Tisnado and R.A. Scofield, 1987: Characteristics of extreme rainfall events innorthwestern Peru during the 1982-83 period. J. Geophys. Res., 92, 14225-14241.

Goldbrunner, A.W., 1963: Las Causas Meteorologicas de la Lluvias de Extraordinaria Magnitud enVenezuela, 2nd ed. Minist. Defensa, Venezuela, 230 pp.

Golden, J.H., 1973: Some statistical aspects of waterspout formation. Weatherwise, 26, 108-117.

Golden, J.H., 1974: Life cycle of Florida Keys’ waterspouts, I. Scaleinteraction implications for the waterspout life cycle, II. J. Appl. Meteor., 13, 676-709.

Goodyear, H.V., 1968: Frequency and areal distributions of tropical storm rainfall in the United Statescoastal region on the Gulf of Mexico. ESSA Tech. Rept. WB-7.

Gordon, A.H., and R.C. Taylor, 1970: Numerical steady-state friction layer trajectories over the oce-anic tropics as related to weather. Int. Indian Ocean Exped. Meteor. Monogr., 7, 112 pp.

Gramzow, R.H., and W.K. Henry, 1972: The rainy pentads of Central America. J. Appl. Meteor., 11,637-642.

Gray, W.M., 1968: Global view of the origin of tropical disturbances and storms. Mon. Wea. Rev., 96,669-700.

Gray, W.M., 1970: Fundamental role of cumulus convection for kinetic energy transformation in thetropics and general circulation. Proc. Symp. on Tropical Meteorology, Honolulu, Amer. Meteor. Soc.,DV, 1-8.

Gray, W.M., 1978: Hurricanes: their formation, structure and likely role in the tropical circulation.Meteorology over the Tropical Oceans. Roy. Meteor. Soc., 155-218.

Gray, W.M., 1984a: Atlantic seasonal hurricane frequency. Part I: El Niño and 30 mb quasi-biennialoscillation influences. Mon. Wea. Rev., 112, 1649-1668.

Gray, W.M., 1984b: Atlantic seasonal hurricane frequency. Part II: forecasting its variability. Mon.Wea. Rev., 112, 1669-1683.

Gray, W.M., 1985: Tropical cyclone global climatology. WMO Tech. Doc. WMO/TD, 72, Vol.1, 3-19.

Gray, W.M., 1990: Summary of 1990 Atlantic Tropical Cyclone Activity and Seasonal Forecast Verifi-cation. Dept. Atmos. Sci., Colorado State Univ., 29 pp.

Gray, W.M., and R.W. Jacobson, 1977: Diurnal variation of deep cumulus convection. Mon. Wea.Rev., 105, 1171-1188.

Greco, S., and 8 other authors, 1990: Rainfall and surface kinematic conditions over central Amazoniaduring ABLE 2B. J. Geophys. Res., 95, 17,001-17,014.

Grice, G.K., J.D. Belville, C.F. Chappell, K.G. Crawford and R.A. Scofield, 1986: Summary of the firstSino-American academic exchange conference on heavy rain in Wuhan, China. Bull. Amer. Meteor.Soc., 67, 851-855.

Griffiths, J.F., 1964: Statistical rainfall analysis. Research on Tropical Rainfall Patterns and Associ-ated Mesoscale Systems. Dept. Meteor. Texas A&M Univ.Griffiths, J.F., 1967: The analysis of extreme values. Extreme value analysis in Venezuela. Researchon Tropical Rainfall Patterns and Associated Mesoscale Systems. Dept. Meteor. Texas A&M Univ.

Gringorten, I.I., 1960: Extreme-value statistics in meteorology — a method of application. AF Surv.Geophys., 125.

Grossman, R.L., and C.A. Friehe, 1986: Vertical structure of the southwest monsoon low-level jet overthe central and eastern Arabian Sea. J. Atmos. Sci., 43, 3266-3272.

Guard, C.P., 1983: A study of western North Pacific tropical storms and typhoons that intensify afterrecurvature. First Wea. Wing Tech. Note, 83/002, 28 pp.

Guard, C.P., 1985: An observational assessment of the southwest monsoon east of East Asia. Proc. Int.Conf. Climatology, Appl. Meteor., Manila, 11 pp.

Guard, C.P., 1990: Personal communication.

Gumbel, E.J., 1958: Statistics of Extremes. Columbia Univ. Press, New York.

Gupta, G.R., M.M. Nayasthani and R.K. Verma, 1972: Clear air turbulence (CAT) in India andneighbourhood. Indian J. Meteor. Geophys., 23, 379-384.

Gutnick, M., 1958: Climatology of the tradewind inversion in the Caribbean. Bull. Amer. Meteor.Soc., 39, 410-420.

Hadley, G., 1735: Concerning the cause of the general trade-winds. Phil. Trans., 29, 58-62.

Hamilton, R.A., and J.W. Archbold, 1945: Meteorology of Nigeria and adjacent territory. Quart. J.Roy. Meteor. Soc., 71, 231-264.

Hammer, R.M., 1970: Cloud development and distribution around Khartoum. Weather, 25, 411-415.

Han, S.-C., 1970: Rainfall over Oahu and location of 300 mb trough during tradewind regimes. HawaiiInst. Geophys, Rept. 70-28, 14 pp.

Hardy, K.R., D. Atlas and K.A. Browning, 1964: The structure of hurricane spiral bands. Proc. WorldConf. on Radar Meteorology. Boulder, 342-345.

Harris, B.E., and F.P. Ho, 1969: Structure of the troposphere over southeast Asia during the summermonsoon month of July. Hawaii Inst. Geophys. Rept. 69-4, 39 pp.

Hastenrath, S., 1985: Climate and Circulation of the Tropics. D. Reidel, Dordrecht, 455 pp.

Hastenrath, S., 1990a: The relationship of highly reflective clouds to tropical climate anomalies. J.Climate, 3, 353-365.Hastenrath, S., 1990b: Tropical climate prediction : a progress report, 1985-1990. Bull. Amer. Meteor.Soc., 71, 819-825.

Hastenrath, S., 1990c: Prediction of northeast Brazil rainfall anomalies. J. Climate, 3, 893-904.

Hayashi, Y.-K., and T. Nakasawa, 1989: Evidence of the existence and eastward motion of superclus-ters at the equator. Mon. Wea. Rev., 117, 236-243.

Hayden, C.M., 1970: An objective analysis of cloud cluster dimensions and spacing in the tropicalNorth Pacific. Mon. Wea. Rev., 98, 534-540.

Heckley, W.A., 1985: Systematic errors of the ECMWF operational forecasting model in tropical re-gions. Quart. J. Roy. Meteor. Soc., 111, 709-738.

Henry, W.K., 1974: The tropical rainstorm. Mon. Wea. Rev., 102, 717-725.

Hess, G.D., and K.T. Spillane, 1990: Characteristics of dust devils in Australia. J. Appl. Meteor., 29,498-507.

Hill, E.L., W. Malkin and W.A. Schulz Jr., 1966: Tornadoes associated with cyclones of tropical origin— practical features. J. Appl. Meteor., 5, 745-763.

Hill, H.W., 1964: The weather of lower latitudes of the southwest Pacific associated with passages ofdisturbances in the middle latitude westerlies. Proc. Symp. on Tropical Meteorology. New ZealandMeteor. Serv., 352-365.

Hill, H.W., 1970: The precipitation in New Zealand associated with the cyclone of early April 1968.New Zealand J. Sci., 13, 641-662.

Hirst, A.C., and S. Hansterath, 1983: Atmosphere-ocean mechanisms of climate anomalies in the Angola-tropical Atlantic sector. J. Phys. Ocean., 13, 1146-1157.

Holle, R.L., 1968: Some aspects of tropical oceanic cloud populations. J. Appl. Meteor., 7, 173-183.

Holliday, C.R., and A.H. Thompson, 1986: An unusual near-equatorial typhoon. Mon. Wea. Rev., 114,2674-2677.

Holliday, C.R., and K. Waters, 1989: SSM/I observation of tropical cyclone gale force vicinity winds.Preprints, 4th Conf. Satellite Meteor. Oceanogr., Amer. Meteor. Soc., 267-270.

Hoover, R.A., 1957: Empirical relationships of the central pressures in hurricanes to the maximumsurge and storm tide. Mon. Wea. Rev., 85, 167-174.

Hopwood, J.M., 1968: A note on the quasi-biennial oscillation at Darwin. Aust. Meteor. Mag., 16, 114-117.Horel, J.D., A.N. Hahmann and J.E. Geisler, 1989: An investigation of the annual cycle of convectiveactivity over the tropical Americas. J. Climate, 2, 1388-1403.

Houze, R.A. Jr., 1977: Structure and dynamics of a tropical squall-line system observed during GATE.Mon. Wea. Rev., 105, 1540-1567.

Houze, R.A. Jr., 1989: Observed structure of mesoscale convective systems and implications for large-scale heating. Quart. J. Roy. Meteor. Soc., 115, 425-461.

Hsu, S., 1970: Coastal air-circulation system observations and empirical model. Mon. Wea. Rev., 98,487-509.

Hsu, S.A., 1979: Mesoscale nocturnal jet-like winds within the planetary boundary layer over a flatopen coast. Bound. Layer Meteor., 17, 485-494.

Hubert, L.F., and L.F. Whitney Jr., 1971: Wind estimates from geostationary satellite pictures. Mon.Wea. Rev., 99, 665-672.

Hurrell, J.W., and D.G. Vincent, 1990: Relationships between tropical heating and subtropical westerlymaxima in the Southern Hemisphere during SOP-1, FGGE. J. Climate, 3, 751-768.

Huschke, R.E. (ed.), 1959: Glossary of Meteorology. Amer. Meteor. Soc., 638 pp.

Hyson, P., R.M. Leigh and R.C. Southern, 1964: Observational and forecasting aspects of the convec-tion cycle at Darwin. Proc. Symp. on Tropical Meteorology. New Zealand Meteor. Serv., 306-315.

Illari, L., 1989: The quality of satellite precipitable water content data and their impact on analysedmoisture fields. Tellus, 41A, 319-337.

India Meteorological Department, 1968: ICAO Improved Short-Term High-Level Turbulence Report-ing Programme, Final Report of Analysis. Meteor. Off., Poona.

Ing, G., 1971: Summer circulation and weather over the Korat Plateau, Thailand. Dept. Meteor., Univ.Hawaii, UHMET 71-04, 41 pp.

Izawa, T., 1964: On the mean wind structure of typhoon. Meteor. Res. Inst. Japan Meteor. Agency,Typhoon Res. Lab. Tech. Note 2.

Jalu, R., and J. Dettwiller, 1965: Advection froide vers les basses latitudes (Mars 1963). La Météorologie,7, 113-128.

Janowiak, J.E., A.F. Krueger and P.A. Arkin, 1985: Atlas of outgoing longwave radiation derived fromNOAA satellite data. NOAA Atlas 6, 44 pp.

Jeandidier, G., and P. Rainteau, 1957: Prévision du temps sur le bassin du Congo. Météor. Nat. Mongr.,France, 9.

Jelesnianski, C.P., 1966: Numerical computations of storm surges without bottom stress. Mon. Wea.Rev., 94, 379-394.

Jelesnianski, C.P., 1967: Numerical computations of storm surge with bottom stress. Mon. Wea. Rev.,95, 740-756.

Jelesnianski, C.P., 1974: Special program to list amplitudes of surges from hurricanes: II General trackand variant storm conditions. Tech. Mem.II, NWS 7DL-46.

Jenkins, G.R., 1945: Diurnal variation of the meteorological elements. Handb. Meteor., McGraw-Hill,New York, 746-753.

Johnson, A.H., 1976: The climate of Peru, Bolivia and Ecuador. In: Climates of Central and SouthAmerica. Elsevier, Amsterdam, 147-218.

Johnson, D.H., 1962: Rain in East Africa. Quart. J. Roy. Meteor. Soc., 88, 1-19.Johnson, D.H., 1964: Forecasting weather in West Africa. WMO Tech. Note 64, 83-94.

Joint GARP Organizing Committee, 1969: Appendix F, Report of the Study Group on Tropical Distur-bances, Second Session.

Joint Typhoon Warning Center, 1990: 1989 Annual Tropical Cyclone Report, 254 pp.

Jordan, C.L., 1963a: The accuracy of center positions of hurricanes as determined by the spiral overlaytechnique. Proc. 10th Weather Radar Conf. Amer. Meteor. Soc., 202-207.

Jordan, C.L., 1963b: Tracking of tropical cyclones. Naval Wea. Res. Facil., NWRF 12-0763-075, 39pp.

Jordan, C.L., and M. Shiroma, 1959: A record rainfall at Okinawa. Bull. Amer. Meteor. Soc., 40, 609-612.

Joseph, P.V., D.K. Raipal and S.N. Deka, 1980: “Andhi”. the convective duststorm of northwest India.Mausam, 31, 431-442.

Kalu, A.E., 1979: The African dust plume: its characteristics and propagation across West Africa inwinter. In: Saharan Dust (C. Morales, ed.), 95-118. Wiley, New York.

Keen, R.A., 1982: The role of tropical cyclone pairs in theSouthern Oscillation. Mon. Wea. Rev., 110, 1405-1416.

Keenan, T.D., and L.R. Brody, 1988: Synoptic-scale modulation of convection during the Australiansummer monsoon. Mon. Wea. Rev., 116, 71-85.

Keenan, T.D., J. McBride, G. Holland, N. Davidson and B. Gunn, 1989: Diurnal variations during theAustralian Monsoon Experiment (AMEX) Phase II. Mon. Wea. Rev., 117, 2535-2552.Kiladis, G.N., and H.F. Diaz, 1989: Global climatic anomalies associated with extremes in the South-ern Oscillation. J. Climate, 2, 1069-1090.

Kloesel, K.A., and B.A. Albrecht, 1989: Low-level inversions over the tropical Pacific — thermody-namic structure of the boundary layer and the above-inversion moisture structure. Mon. Wea. Rev.,117, 87-101.

Knutson, T.R., and K.M. Weickmann, 1987: 30-60 day atmospheric oscillations: composite life cyclesof convection and circulation anomalies. Mon. Wea. Rev., 115, 1407-1436.

Kousky, V.E., 1979: Frontal influences on northeast Brazil. Mon. Wea. Rev., 107, 1140-1153.

Kousky, V.E., 1980: Diurnal variation of rainfall in northeast Brazil. Mon. Wea. Rev., 108, 488-498.

Kousky, V.E., and M.A. Gan, 1981: Upper tropospheric cyclonic vortices in the tropical South Atlan-tic. Tellus, 33, 538-550.

Kraus, E.B., 1963: The diurnal precipitation change over the sea. J. Atmos. Sci., 20, 551-556.

Kulshrestha, S.M., and P.S. Jain, 1968: Aspects of radar climatology of the region around New Delhi innorth India. Proc. 13th Radar Meteorology Conf. Amer. Meteor. Soc., 294-297.

Kumar, S., 1979: Interaction of upper westerly waves with intetropical convergence zone and theireffect on the weather over Zambia during the rainy season. Mausam, 30, 423-438.

Kuo, Y.-H., and G.T.-J. Chen, 1990: The Taiwan Area Mesoscale Experiment (TAMEX): An overview.Bull. Amer. Meteor. Soc., 71, 488-503.

Kyle, A.C., 1970: Longitudinal variation of large-scale vertical motion in the tropics. Mass. Inst. Tech.MS thesis.

Lajoie, F.A., 1981: Cloud top equivalent black-body temperatures and tropical cylone forecasting inthe Australian region. Aust. Meteor. Mag., 29, 73-88.

Lajoie, F.A., and I.J. Butterworth, 1984: Oscillation of high-level cirrus and heavy precipitation aroundAustralian region tropical cyclones. Mon. Wea. Rev., 112, 535-544.

Lambert, S.J., 1989: A comparison of divergent winds from the National Meteorological Center andthe European Center for Medium Range Weather Forecasts global analyses for 1980-1986. Mon. Wea.Rev., 117, 995-1005.

Lander, M., 1986: On the contribution of the sea surface temperature pattern to the regional differencesin the marine boundary layer shear. Proc. COADS Workshop, Boulder, Colorado 176-178.

Lander, M., 1990: Evolution of the cloud pattern during the formation oftropical cyclone twins symmetrical with respect the equator. Mon. Wea. Rev., 118, 1194-1202.

Lander, M., 1991: Personal communication.

LaSeur, N.E., 1960: Synoptic analysis in the tropics. The general problem and methods of tropicalsynoptic analysis. Proc. Symp. on Tropical Meteorology in Africa. Munitalp Foundation,Nairobi, 7-13 and 24-34.

LaSeur, N.E., 1964a: Synoptic models in the tropics. Proc. Symp. on Tropical Meteorology. NewZealand Meteor. Serv., 319-328.

LaSeur, N.E., 1964b: Some preliminary results of the Barbados field program, a statistical measure ofthe importance of convection in tropical rainfall. Proc. Army Conf. on Tropical Meteorology. Inst.Mar. Sci., Univ. Miami, 20-25.

LaSeur, N.E., and H.F. Hawkins, 1963: An analysis of Hurricane Cleo (1958) based on data fromresearch reconnaissance aircraft. Mon. Wea. Rev., 91, 694-709.

Lau, K.-H., and N.-G. Lau, 1990: Observed structure and propagation characteristics of tropical sum-mertime synoptic scale disturbances. Mon. Wea. Rev., 118, 1888-1913.

Lau, R., and M.Y. Chan, 1990: The Royal Observatory long range forecast methods. In: East Asian andWestern Pacific Meteorology and Climate (P. Sham and C.P Chang, eds.), World Scientific, Singapore,485-493.

Lauer, W., 1975: Klimatische Grundzuge der Höhenstufung tropischer Gebirge. Tagunsbericht undWissenschlaftliche Abhandlungen. F.Steiner, Innsbruck, 76-90.

Lavoie, R.L., 1963: Research on the meteorology of the tropical Pacific and its applications. HawaiiInst. Geophys. Rept. 37, 42 pp.

Lavoie, R.L., and C.J. Wiederanders, 1960: Objective wind forecasting over the tropical Pacific. Ha-waii Inst. Geophys. Sci. Rept. 1, 38 pp.

Leipper, D.F., 1967: Observed ocean conditions and Hurricane Hilda, 1964. J. Atmos. Sci., 24, 182-196.

Lenhardt, R.W., 1967: Subsynoptic atmospheric variability up to 300 mb. AFCRL Interim Notes Atmos.Properties, 78, 22 pp.

Leopold, L.B., 1949: The interaction of tradewind and sea breeze, Hawaii. J. Meteor., 6, 312-320.

Leroux, M., 1983: Le Climat de l’Afrique Tropical. Editions Champion, Paris, 633 pp.

Lettau, H., 1976: Dynamic and energetic factors which cause and limit aridity along South America’sPacific coast. In: Climates of Central and

South America, Elsevier, Amsterdam, 188-192.

Leverson, V.H., P.C. Sinclair and J.H. Golden, 1977: Waterspout wind, temperature and pressure struc-ture deduced from aircraft measurements. Mon. Wea. Rev., 105, 725-733.

Lighthill, J., and R.P. Pearce (eds.), 1981: Monsoon Dynamics. Cambridge Univ. Press, 735 pp.

Lim, J.T., 1979: Characteristics of the winter monsoon over the Malaysian region. Dept. Meteor. Univ.Hawaii [PhD thesis].

List, R.J., 1963: Smithsonian Meteorological Tables (6th edition). Smithsonian Institution, Washing-ton, D.C., 527 pp.

Lockwood, J.G., 1967: Probable maximum 24-hour precipitation over Malaya by statistical methods.Meteor. Mag., 96, 11-19.

Lorenz, E.N., 1967: The nature and theory of the general circulation of the atmosphere. WMO 218, TP115.

Love, G., 1985a: Cross-equatorial influence of winter hemisphere subtropical cold surges. Mon. Wea.Rev., 113, 1487-1498.

Love, G., 1985b: Cross-equatorial interactions during tropical cyclogenesis. Mon. Wea. Rev., 113,1499-1509.

Lumb, F.E., 1970: Topographic influences on thunderstorm activity near Lake Victoria. Weather, 25,404-410.

Lydolph, P.E., 1973: On the causes of aridity along a selected group of coasts. In: Coastal Deserts,Their Natural and Human Environments (D.H.K. Amirau and A.W. Wilson, eds.), Univ. Arizona Press,57-72.

Lyons, S.W., 1979: Summer weather over Haleakala, Maui. Dept. Meteor. Univ. Hawaii, UHMET 79-09, 56 pp.

Lyons, S.W., 1991: Origins of convective variability over equatorial southern Africa during australsummer. J. Climate, 4, 23-39.

Mackey, S., 1969: How typhoons affect tall buildings — HK study. Far East Builder, January, 37-40.

Madden, R.A., and P.R. Julian, 1971: Detection of a 40-50 day oscillation in the zonal wind in thetropical Pacific. J. Atmos. Sci., 28, 702-708.

Madden, R.A., and E.J. Zipser, 1970: Multi-layered structure of the wind over the equatorial Pacificduring the Line Islands Experiment. J. Atmos. Sci., 27, 336-342.

Maddox, R.A., 1983: Large-scale meteorological conditions associated with midlatitude, mesoscaleconvective complexes. Mon. Wea. Rev., 111, 1475-1493.

Mahrer, Y., and R.A. Pielke, 1977: The effects of topography on sea and land breezes in a two-dimen-sional model. Mon. Wea. Rev., 105, 1151-1162.

Mak, M.K., and J.E. Walsh, 1976: On the relative intensities of sea and land breezes. J. Atmos. Sci.,33, 242-251.

Maliekal, J.A., 1989: An analysis of temporal and spatial characteristics of tropical weather. Dept.Meteor. Univ. Hawaii (PhD thesis).

Malkus, J.S., 1962: Large scale interactions. The Sea, Vol. I. Wiley, New York, 88-294.

Malkus, J.S., and H. Riehl, 1964: Cloud structure and distributions over the tropical Pacific. Tellus, 16,275-287.

Maragos, J.E., G.B.K. Baines and P.J. Beveridge, 1973: Tropical cyclone Bebe creates a new landformation on Funafuti Atoll. Science, 181, 1161-1164.

Martin, D., and O. Karst., 1969: A census of cloud systems over the tropical Pacific. Studies in Atmos.Energetics Based on Aerospace Probings, Ann. Rept. —1968. Space Sci. Engin. Cent., Univ. Wiscon-sin, 37-50.

Martin, D.W., B. Auvine and D. Suchman, 1984: Synoptic forcing and control of deep convection onDay 261 of GATE. Mon. Wea. Rev., 112, 1936-1959.

Martin D.W., and V.E. Suomi, 1972: A satellite study of cloud clusters over the tropical North AtlanticOcean. Bull. Amer. Meteor. Soc., 53, 135-156.

Martin, L.A., 1964: An investigation of the rainfall distribution for stations in North and Central America.Research on Tropical Rainfall Patterns and Associated Mesoscale Systems. Dept. Meteor. Texas A&MUniv.

McBride, J.L., 1987: The Australian summer monsoon. Monsoon Meteorology. Clarendon Press, Oxford,60-92.

McBride, J.L., and T.D. Keenan, 1982: Climatology of tropical cyclone genesis in the Australian re-gion. J. Climatol., 2, 13-33.

McCabe, J.T., 1961: (1) A systematic approach to the use of climatology and persistence in terminalweather forecasting. (2) An application of the climatology of typhoon movement to operational fore-casting and estimation of typhoon threat to a particular installation. US Asian Wea. Symp. First Wea.Wing.

McCreary, F.E., 1959: Stratospheric winds over the tropical Pacific Ocean (abstract). Bull. Amer.Meteor. Soc., 40, 370.

McGarry, M.M., and R.J. Reed, 1978: Diurnal variations in convective activity and precipitation dur-ing Phases II and III of GATE. Mon. Wea.Rev., 106, 101-113.

McGurk, J.P., and D.J. Ulst, 1990: Evolution of tropical plumes in VAS water vapor imagery. Mon.Wea. Rev., 118, 1758-1766.

McPherson, R.D., 1970: A numerical study of the effect of a coastal irregularity on the sea breeze. J.Appl. Meteor., 9, 767-777.

Meigs, P., 1973: World distribution of coastal deserts. In: Coastal Deserts, Their Natural and HumanEnvironment (D.H.K. Amirau and A.W. Wilson, eds.), Univ. Arizona Press, 3-12.

Meisner, B.N., 1979: Ridge-regression — time extrapolation applied to Hawaiian rainfall normals. J.Appl. Meteor., 18, 904-912.

Menzel, W.P., and A. Chedin, 1990: Summary of the fifth international TIROS Operational VerticalSounder (TOVS) study conference. Bull. Amer. Meteor. Soc., 71, 691-693.

Meteorological Office, 1968: Monthly Meteorological Charts of the Eastern Pacific Ocean. 2nd Edi-tion. H.M. Stationery Office, London, 122 pp.

Meteorology Working Group, 1965: Meteorological equipment data accuracies. IRIG Document 110-64, 11 pp.

Meyencon, R., 1983: Cyclone in the Mediterranean. Mar. Wea. Log, 27, 141-143.

Miller, B.I., 1958: Rainfall rates in Florida hurricanes. Mon. Wea. Rev., 86, 258-264.

Miller, B.I., 1964: A study of the filling of Hurricane Donna (1960) over land. Mon. Wea. Rev., 92,389-406.

Miller, B.I., 1967: Characteristics of hurricanes. Science, 157, 1389-1399.

Miller, F.R., and R.N. Keshavamurthy, 1968: Structure of an Arabian Sea summer monsoon system.Int. Indian Ocean Exped. Meteor. Monogr., 1, 94 pp. East-West Center Press, Honolulu.

Miller, R.C., 1967: Forecasting thunderstorm-produced surface wind gusts in southeast Asia. Milit.Wea. Warning Center Kansas City, 14 pp [unpub. rept.]

Minnis, P., and E.F. Harrison, 1984: Diurnal variability of regional cloud and clear sky parametersderived from GOES data. Part II November 1978 cloud distributions. J. Climatol. Appl. Meteor., 23,1012-1031.

Mishra, D.K., and M.S. Singh, 1977: A study of the monsoon depression intensifying into cyclonicstorm over land. Indian J. Meteor. Hydrol. Geophys., 28, 321-327.

Mooley, D.A., and H.L. Crutcher, 1968: An application of the gammadistribution function to Indian rainfall. ESSA Tech. Rept. ESD 5.

Morgan, M.R., 1965: Outbreaks of Antarctic air in relation to the hurricane season 1962-63. Meteor.Branch Dept. Transportation, Canada, Tec. 571, 24 pp.

Morrrissey, M.L., 1986: A statistical analysis of the relationships among rainfall, outgoing longwaveradiation and the moisture budget during January-March 1979. Mon. Wea. Rev., 114, 931-942.

Morrissey, M.L., M.A. Lander and J.A. Maliekal, 1988: A preliminary evaluation of ship data in theequatorial western Pacific. J. Atmos. Ocean. Technol., 5, 251-258.

Mukherjee, A.K., and K.P. Padmanabham, 1977: Simultaneous occurrence of tropical cyclones oneither side of the equator in the Indian Ocean area. Indian J. Meteor. Hydrol. Geophys., 28, 211-222.

Murakami, M., 1983: Analysis of the deep convective activity over the western Pacific and southeastAsia. J. Meteor. Soc. Japan, 61, 60-76.

Muramatsu, T., 1983: Diurnal variations of satellite-measured TBB areal distribution and eye-diameterof mature typhoons. J. Meteor. Soc. Japan, 61, 77-90.

National Aeronautics and Space Administration, 1988: Earth System Science, a Closer View. NASA,Washingotn, D.C., 208 pp.

National Research Council, 1991: The New Year’s Eve flood on Oahu, Hawaii. Decmber 31, 1987 -

January 1, 1988. Natural Disasters Studies, 1, 72 pp. National Academy Press, Washington, D.C.

Naujokat, B., 1986: An update of the observed QBO of the stratospheric winds over the tropics. J.Atmos. Sci., 43, 1873-1877.

Naval Oceanography Command, 1981: U.S. Navy Marine Climatic Atlas of the World. Vol.IX, World-Wide Means and Standard Deviations. NAVAIR 50-1C-65. 169 pp.

Navy Weather Research Facility Staff, 1969: The Diagnosis and Prediction of SE Asia Northeast Mon-soon Weather. Norfolk, Virginia.

Neal, A.B., 1973: The meteorology of the Australian trades. Part 2: the perturbing effect of WillisIsland on the oceanic trade wind flow. Aust. Bur. Meteor. Tech. Rept., 5(2), 17 pp.

Neiburger, M., D.S. Johnson and C.-W. Chen, 1961: Studies of the structure of the atmosphere over theeastern Pacific Ocean in summer. I. The inversion over the eastern North Pacific Ocean. Univ. Calif.Pub. Meteor., 1, No. 1, 94 pp.

Neumann, C.J., 1965: Mesoanalysis of a severe south Florida hailstorm. J. Appl. Meteor., 4, 161-171.

Newell, R.E., 1979: Climate and ocean. Amer. Sci., 67, 405-416.

Newell, R.E., J.W. Kidson, D.C. Vincent and G.J. Boer, 1972: The General Circulation of the TropicalAtmosphere and Interaction with Extratropical Latitudes, Vol. 1. MIT Press, Cambridge, 258 pp.

Nicholls, N., 1985: Predictability of interannual variations of Australian seasonal tropical cyclone ac-tivity. Mon. Wea. Rev., 113, 1144-1145.

Nicholson, S.E., 1981: Rainfal and atmospheric circulation during drought periods and wetter years inWest Africa. Mon. Wea. Rev., 109, 2191-2208.

Nieuwolt, S., 1968: Diurnal variation of rainfall in Malaya. Ann. Assoc. Amer. Geogr., 58, 320-326.

NOAA/NESDIS, 1948 —: Monthly Climatic Data for the World.

NOAA/NESDIS, 1983: The GOES User’s Guide.

Novlan, D.J., and W.M. Gray, 1974: Hurricane-spawned tornadoes.Mon. Wea. Rev., 102, 476-488.

Okulaja, F.O., 1968: The frequency distribution of Lagos/Ikeja wind gusts. J. Appl. Meteor., 7, 379-383.

Olascoaga, M.J., 1950: Some aspects of Argentine rainfall. Tellus, 2, 312-318.

Orlanski, I., 1975: A rational subdivision of scales for atmospheric processes. Bull. Amer. Meteor.Soc., 56, 527-530.

Orton, R., 1970: Tornadoes associated with Hurricane Beulah onSeptember 19-23, 1967. Mon. Wea. Rev., 98, 541-547.

Orville, R.E., and R.W. Henderson, 1986: Global distribution of midnight lightning. Mon. Wea. Rev.,114, 2640-2653.

Osman, O.E., and S. Hastenrath, 1969: On the synoptic climatology of summer rainfall over centralSudan. Arch. Meteor. Geophys. Bioclimatol., Ser. B, 17, 297-324.

Palecek, D.A., 19??: Forecasting thunderstorms at Clark AB, Philippines, March through October.[unpub. manuscript]

Pallman, A.J., 1968: The Synoptics, Dynamics and Energetics of the Temporal Using Satellite Radia-tion Data. Dept. Geophys. St. Louis Univ., 98 pp.

Palmer, C.E., 1951a: Tropical meteorology. Comp. Meteor., 859-880.

Palmer, C.E., 1951b: On high-level cyclones originating in the tropics. Trans. Amer. Geophys. Un., 33,683-696.

Palmer, C.E., 1952: Tropical meteorology. Quart. J. Roy. Meteor. Soc., 78, 126-164.

Palmer, C.E., J.R. Nicholson and R.M. Shimaura, 1956: An Indirect Aerology of the Tropical Pacific.Inst. Geophys. Univ. Calif., 131 pp.

Palmer, C.E., and W.D. Ohmstede, 1956: The simultaneous oscillation of barometers along and nearthe equator. Tellus, 8, 495-507.

Palmer, C.E., C.W. Wise, L.J. Stempson and G.H. Duncan, 1955: The practical aspect of tropical me-teorology. AWS Manual 105-48, Vol. 1, 195 pp.

Parmenter, F.C., 1976: A Southern Hemisphere cold front passage at the equator. Bull. Amer. Meteor.Soc., 57, 1435-1440.

Parthasarathy, B., K. Rupah Kumar and N.A. Sontakke, 1990: Surface and upper air temperatures overIndia in relation to monsoon rainfall. Theor. Appl. Climatol., 42, 93-110.

Paulus, J.L.H., 1965: Indian Ocean and Taiwan rainfalls set new records. Mon. Wea. Rev., 93, 331-335.

Pearson, A.D., and A.F. Sadowski, 1965: Hurricane-induced tornadoes and their distribution. Mon.Wea. Rev., 93, 461-464.

Pedgley, D.E., 1969: Diurnal variation of the incidence of monsoon rainfall over the Sudan, Parts I andII. Meteor. Mag., 98, 97-107; 129-134.

Petterssen, S., 1956: Weather Analysis and Forecasting. Second Edition. Vol.I Motion and MotionSystems. McGraw-Hill, New York, 428 pp.

Pielke, R., 1974: A three-dimensional numerical model of the sea breezes over south Florida. Mon.Wea. Rev., 102, 115-139.

Pike, A.C., and C.J. Neumann, 1987: The variation of track forecast difficulty among tropical cyclonebasins. Wea. Forecasting, 2, 237-241.

Portig, W.H., 1963: Thunderstorm frequency and amount of precipitation in the tropics. Arch. Meteor.Geophys. Bioklimatol., Ser. B., 13, 21-35.

Portig, W.H., 1976: The climate of Central America. Climates of Central and South America. Elsevier,Amsterdam, 405-478.

Powell, J., and D.E. Pedgley, 1969: A year’s weather at Termit, Republic of Niger. Weather, 24, 247-254.

Prasad, B., 1970: Diurnal variation of rainfall in India. Indian J. Meteor. Geophys., 21, 443-450.

Probert-Jones, J.R., 1960: Meteorological uses of pulse Doppler Radar. Nature, 186, 271-273.

Prohaska, F.J., 1973: New evidence on the climatic controls along the Peruvian coast. In: CoastalDeserts Their Natural and Human Environments (D.H.K. Amirau and A.W. Wilson, eds.), Univ. Ari-zona Press, 91-107.

Prümm, D., 1974: Zeitliche Variationen meteorologischen Grössen in der wassernahen Luftschicht desatlantischen Nordost-Passats. Ber. Inst. Radiometeor. Mar. Meteor. Univ. Hamburg, 23, 103 pp.

Puniah, K.B., 1973: Heights of tops of Cb clouds over southeast Asia. Indian J. Meteor. Geophys., 24,386-387.

Rainbird, A.F., 1968: Weather Disturbances over Tropical Continents and Their Effects on GroundConditions. Dept. Atmos. Sci. Colorado State Univ.

Rakshit, D.K., 1966: A radar study of thin band or precursor line in association with thunderstorm line.Proc. 12th Conf. Radar Meteor. Amer. Meteor. Soc., 334-338.

Ram, N., 1929: Frequency of thunderstorms in India. India Meteor. Dept. Sci. Notes, 1, 49-55.

Ramage, C.S., 1952a: Diurnal variation of summer rainfall over east China, Korea and Japan. J. Me-teor., 9, 83-86.

Ramage, C.S., 1952b: Variation of rainfall over south Chinathrough the wet season. Bull. Amer. Meteor. Soc., 33, 308-311.

Ramage, C.S., 1954: Non-frontal crachin and the cool season clouds of the China Seas. Bull. Amer.

Meteor. Soc., 35, 404-411.

Ramage, C.S., 1957: Surface Weather Chart Analysis in Low Latitudes. Dept. Meteor. Univ. Hawaii,23 pp.

Ramage, C.S., 1962: The subtropical cyclone. J. Geophys. Res., 67, 1401-1411.

Ramage, C.S., 1964: Diurnal variation of summer rainfall of Malaya. J. Trop. Geog., 19, 62-68.

Ramage, C.S., 1966: The summer atmospheric circulation over the Arabian Sea. J. Atmos. Sci., 23,144-150.

Ramage, C.S., 1968: Role of a tropical “maritime continent” in the atmospheric circulation. Mon.Wea. Rev., 96, 365-370.

Ramage, C.S., 1969: Summer drought over western India. Yearbook Assn. Pacific Coast Geographers,30, 41-54.

Ramage, C.S., 1971: Monsoon Meteorology. Academic Press, New York, 296 pp.

Ramage, C.S., 1974: Structure of an oceanic near-equatorial trough deduced from research aircrafttraverses. Mon. Wea. Rev., 102, 754-759.Ramage, C.S., 1978a: Effect of the Hawaiian Islands on the tradewinds. Prepr. Conf. Clim. and En-ergy: Clim. Aspects and Indust. Operations. Amer. Meteor. Soc., 62-67.

Ramage, C.S., 1978b: Orographic, mesoscale and synoptic influences related to Mt. Waialeale rainfall.Bull. Amer. Meteor. Soc., 11, 1544 [abstr.].

Ramage, C.S., 1980: Forecasting tropical cyclone movement — retrospect and prospect. Sel. Pap.13th Tech. Conf. Hurricanes and Trop. Meteor. Amer. Meteor. Soc., 65-67.

Ramage, C.S., 1983: Teleconnections and the seige of time. J. Climatol., 3, 223-231.

Ramage, C.S., 1987: Secular change in reported surface wind speeds over the ocean. J. Appl. Meteor.,26, 525-528.

Ramage, C.S., L.R. Brody, R.F. Adler and S. Brand, 1969: A diagnosis of the summer monsoon ofsoutheast Asia. NWRF Tech. Pap.,10-69, 50 pp.

Ramage, C.S., S.J.S. Khalsa and B.N. Meisner, 1981: The central Pacific near-equatorial convergencezone. J. Geophys. Res., 86, 6580-6598.

Ramage, C.S., and C.R.V. Raman, 1972: Meteorological Atlas of the International Indian Ocean Expe-dition, Vol. 2 Upper Air. Nat. Sci. Found. (Charts).

Ramamurthi, K., 1969: Some aspects of the “break”in the southwest monsoon in July and August.India Meteor. Dept. FMU Rept. 4.

Ramanathan, K.R., and K.P. Ramakrishnan, 1932: The Indian southwest monsoon and the structure ofdepressions associated with it. Mem. India Meteor. Dept., 26, 13-36.

Ranganathan, C., and K. Soundarajan, 1965: A study of a typical case of interaction of an easterly wavewith a westerly trough during the postmonsoon period. Indian J. Meteor. Geophys., 16, 607-616.

Rao, K.N., 1965: Seasonal forecasting — India. WMO-IUGG Symp. Res. Devel. Aspects of Long-Range Forecasting. WMO, 17-30.

Rao, Y.P., 1976: Southwest monsoon. India Meteor. Dept., Meteor. Monogr. Synop. Meteor., 1/76, 367pp.

Reed, J.W., 1967: Some notes on forecasting winds aloft by statistical methods. J. Appl. Meteor., 6,360-372.

Reed, R.J., 1964: A climatology of wind and temperatures in the tropical stratosphere between 100 mband 10 mb. NWRF 26-0564-092.

Reed, R.J., W.J. Campbell, L.A. Rasmusson and D.G. Rogers, 1961: Evidence of a downward-propa-gating annual wind reversal in the equatorial stratosphere. J. Geophys. Res., 66, 813-818.Reed, R.J., and K.D. Jaffe, 1981: Diurnal variation of summer convection over West Africa and thetropical eastern Atlantic during 1974 and 1978. Mon. Wea. Rev., 109, 2527-2534.

Reed, R.J., D.C. Norquist and E.E. Recker, 1977: The structure and properties of African wave distur-bances as observed during Phase III of GATE. Mon. Wea. Rev., 105, 317-333.

Reed, R.J., and E.E. Recker, 1971: Structure and properties of synoptic-scale wave disturbances in theequatorial western Pacific. J. Atmos. Sci., 28, 1117-1133.

Refsdal, A., 1930: Der Feuchtlabile Niederschlag. Geophys. Publik. 5, No. 12.

Rennick, M.A., 1976: The generation of African waves. J. Atmos. Sci., 33, 1955-1969.

Reynolds, G., 1978: Two statistical heresies. Weather, 33, 74-76.Riehl, H., 1948: Waves in the easterlies. Misc. Rept. 17, Dept. Meteor. Univ. Chicago.

Riehl, H., 1949: Some aspects of Hawaiian rainfall. Bull. Amer. Meteor. Soc., 30, 176-187.

Riehl, H., 1950: On the role of the tropics in the general circulation of the atmosphere. Tellus, 2, 1-17.

Riehl, H., 1954: Tropical Meteorology. McGraw-Hill, New York, 392 pp.

Riehl, H., 1966: Upper air observations in the tropics. World Wea. Watch Planning Rept., 1, WMO, 16pp.

Riehl, H., 1967: Southeast Asia Monsoon Study. Dept. Atmos. Sci. Colorado State Univ., 33 pp.

Riehl, H., 1979a: Climate and Weather in the Tropics. Academic Press, New York, 611 pp.

Riehl, H., 1979b: Occurrence and structure of the equatorial trough zone in Venezuela. Quart. J. Roy.Meteor. Soc., 105, 217-229.

Riehl, H., L. Cruz, M. Mata and C. Muster, 1973: Precipitation characteristics of the Venezuela rainyseason. Quart. J. Roy. Meteor. Soc., 99, 746-757.

Riehl, H., and J.S. Malkus, 1957: On the heat balance and maintenance of circulation in the trades.Quart. J. Roy. Meteor. Soc., 83, 21-29.

Riehl, H., and J.S. Malkus, 1958: On the heat balance in the equatorial trough zone. Geophysica, 6,503-537.

Rogers, E.B., J. Stout, J. Steranka and S. Chang, 1990: Tropical cyclone-upper atmospheric interactionas inferred from satellite totalozone observations. J. Appl. Meteor., 29, 934-954.

Ropelewski, C.F., and M.S. Halpert, 1987: Global and regional scale precipitation patterns associatedwith the El Niño/Southern Oscillation. Mon. Wea. Rev., 115, 1606-1626.

Ropelewski, C.F., and M.S. Halpert, 1989: Precipitation patterns associated with the high index phaseof the Southern Oscillation. J. Climate, 2, 268-284.

Ruprecht, E., and W.M. Gray, 1976: Analysis of satellite-observed cloud clusters, Papers I and II.Tellus, 28, 391-425.

Sadler, J.C., 1964: Tropical cyclones of the eastern North Pacific as revealed by TIROS observations.J. Appl. Meteor., 3, 347-366.

Sadler, J.C., 1965: The feasibility of global tropical analyses. Bull. Amer. Meteor. Soc., 46, 118-130.

Sadler, J.C., 1967: On the origin of tropical vortices. Proc. Working Panel Dyn. Meteor., NWRF 12-1167-132, 39-76.

Sadler, J.C., 1969: Average cloudiness in the tropics from satellite observations. Int. Indian OceanExped. Meteor. Monogr. 2, East-West Center Press, Honolulu, 23 pp and plates.

Sadler, J.C., 1975a: The monsoon circulation and cloudiness over the GATE area. Mon. Wea. Rev.,103, 369-387.

Sadler, J.C., 1975b: The upper tropospheric circulation over the global tropics. Dept. Meteor. Univ.Hawaii, UHMET 75-05, 35 pp.

Sadler, J.C., 1976a: Tropical cyclone initiation by the tropical upper tropospheric trough. Dept. Me-teor. Univ. Hawaii, UHMET 75-02, 103 pp.

Sadler, J.C., 1976b: Typhoon intensity changes in the Pacific Storm Fury area. Dept. Meteor. Univ.Hawaii, UHMET 76-02, 13 pp and figures.

Sadler, J.C., W.R. Brett, B.E. Harris and F.P. Ho, 1968: Forecasting minimum cloudiness over the RedRiver during the summer monsoon. Hawaii Inst. Geophys. Rept. 68-16, 104 pp.

Sadler, J.C., and B.J. Kilonsky, 1983: Meteorological events in the central Pacific during 1983 associ-ated with the 1982-83 El Niño. Trop. Ocean Atmos. Newsl., 21, 3-5.

Sadler, J.C., and B.J. Kilonsky, 1985: Deriving surface winds from satellite observations of low-levelcloud motions. J. Climate Appl. Meteor., 24, 758-769.

Sadler, J.C., B. Kilonsky, L. Oda and A. Hori, 1984: Mean Cloudiness over the Global Tropics fromSatellite Observations. NAVENVPREDRSCHFAC Contractor Rept. CR 84-09, 73 pp.

Sadler, J.C., and M. Lander, 1986: Regional scale circulations produced by the orography of CentralAmerica. Proc. COADS Workshop, Boulder, Colorado, 169-171.

Sadler, J.C., M.A. Lander, A.M. Hori and L.K. Oda, 1987: Tropical marine climatic atlas, Vol I: IndianOcean and Atlantic Ocean. Dept. Meteor. Univ. Hawaii, UHMET 87-01 (charts).

Sadler, J.C., and J.T. Lim, 1979: Monitoring monsoon outflow from geosynchronous data. In: Mon-soon Dynamics, Cambridge Univ. Press, London, 80-96.

Sadler, J.C., and L.K. Oda, 1979: The synoptic (A) scale circulations during the second phase of GATE17 July - 19 August 1974. Dept. Meteor. Univ. Hawaii, UHMET 78-02, 41 pp.

Sadler, J.C., and T.C. Wann, 1984: Mean upper tropospheric flow over the global tropics. AWS Tech.Rept. 83/002, Vol. II (charts).

Sanders, F., 1984: Quasi-geostrophic diagnosis of the monsoon depression of 5-8 July 1979. J. Atmos.Sci., 41, 538-552.

Sands, W., 1969: A note on severe thunderstorms in SEA. 1st Wea. Wing Tech. Topics, 1-4.

Sansom, H.W., 1963: A moisture/stability index as an aid to local forecasting in East Africa. EastAfrican Meteor. Dept. Tech. Memo, 9, 8 pp.

Saucier, W.J., 1955: Principles of Meteorological Analysis. Univ. Chicago Press, 438 pp.

Saxton, D.W., 1966: The nature and causes of clear-air turbulence. AIAA Paper 66-966, 5 pp.

Schroeder, T.A., 1976: Hawaiian waterspout-tornado of 26 November 1975. Weatherwise, 29, 172-177.

Schroeder, T.A., 1977: Meteorological analysis of an Oahu flood. Mon. Wea. Rev.,105, 458-468.

Schroeder, T.A., 1978: Mesoscale structure of Hawaiian rainstorms. Dept. Meteor. Univ. Hawaii,UHMET 78-03, 69 pp.

Schroeder, T.A., B.J. Kilonsky and W. Kreisel, 1978: Der Tagesgang des Niederschlages auf den Ha-waii-Inseln. Erdkunde, 32, 89-101.

Seagraves, M.A., 1989: Weather and environmental effects on electro-optical devices. U.S. ArmyLaboratory Command, Atmos. Sci. Lab., TR-0254, 40 pp.

Seck, A., 1962: The Heug or dry rainy season in Senegal. Ann. Geogr., 385, 225-246.

Sellers, W.D., 1965: Physical Climatology. Univ. Chicago Press, 272 pp.

Senn, H.V., H.W. Hiser and R.C. Bourret, 1957: Studies of hurricane spiral bands as observed on radar.Nat. Hurricane Res. Proj. Rept., 12, 13 pp.

Sharma, K.K., 1966: Squalls at Nagpur. Indian J. Meteor. Geophys., 17, 77-82.

Sheets, R.C., 1968: The structure of Hurricane Dora (1964). ESSA Res. Lab. Tech. Mem. NHRL-83,64 pp.

Shepard, P.A., and M.H. Omar, 1952: The wind stress over the ocean from observations in the trades.Quart. J. Roy. Meteor. Soc., 78, 583-589.

Sherretz, L.A., J.G. Pitko and R.A. Dorr, 1971: Hawaiian weather forecast studies. Dept. Meteor. Univ.Hawaii, UHMET 71-03, 94 pp.

Shin, K.-P., G.R. North, Y.-S. Ahn and P.A. Arkin, 1990: Time scales and variability of area-averagedtropical oceanic rainfall. Mon. Wea. Rev., 118, 1507-1516.

Short, D.A., and J.M. Wallace, 1980: Satellite-inferred morning-to-evening cloudiness changes. Mon.Wea. Rev., 108, 1160-1169.

Showalter, A.K., 1953: A stability index for thunderstorm forecasting. Bull. Amer. Meteor. Soc., 34,250-252.

Simpson, J.S., and A.S. Dennis, 1972: Cumulus clouds and their modification. NOAA Tech. Mem.ERL OD-14, 148 pp.

Simpson, R.H., 1952: Evolution of the Kona storm, a subtropical cyclone. J. Meteor., 9, 24-35.

Simpson, R.H., 1969: Synoptic analysis models for the tropics. AWS Tech. Rept. 217, 189-200.

Simpson, R.H., 1970a: Mean layer analysis for tracking and prediction of disturbances in the tropics.Proc. Symp. on Tropical Meteor., Honolulu, Amer. Meteor. Soc., G IV, 1-7.

Simpson, R.H., 1970b: A reassessmment of the hurricane prediction problem. ESSA Tech. MemoWBTM SR-50, 16 pp.

Simpson, R.H., N. Frank, D. Shideler and H.M. Johnson, 1968: Atlantic tropical disturbances, 1967.Mon. Wea. Rev., 96, 251-259.

Simpson, R.H., and H. Riehl, 1981: The Hurricane and its Impact. Louisiana State Univ. Press, BatonRouge, 398 pp.

Singh, R., 1981: On the occurrence of tornadoes and their distribution in India. Mausam, 32, 307-314.

Smith, J.S., 1965: The hurricane-tornado. Mon. Wea. Rev., 93, 453-459.Solot, S.B., 1943: The meteorology of central Africa. AWS Tech. Rept., 105-50.

Sourbeer, R.H., and R.C. Gentry, 1961: Rainstorm in southern Florida, January 21, 1957. Mon. Wea.Rev., 89, 9-16.

Southern, R.L., W.R. Kininmonth and N.R. Prescod, 1970: Derivation of convective forecasting mod-els for northern Australia from a climatology of lightning discharges. Proc. Conf. Summer Monsoon,Southeast Asia, NWRF, 239-254.

Spillane, K.T., 1968: The winter jet stream of Australia and its turbulence. Aust. Meteor. Mag., 16, 64-71.

Sreenivasaiah, B.N., and K.N. Sur, 1939: A study of the duststorms of Agra. Mem. India Meteor.Dept., 27, 1-30.

Stephens, G.L., 1990: On the relationship between water vapor over the oceans and the sea surfacetemperature. J. Climate, 3, 634-645.

Steranka, J., E.B. Rogers and R.C. Gentry, 1984: The diurnal variation of Atlantic tropical cyclonecloud distribution inferred from geostationary satellite infrared measurements. Mon. Wea. Rev., 112,2338-2344.

Stone, R.G., 1957: A compendium on cirrus and cirrus forecasting. AWS Tech. Rept., 105-130.

Stone, R.G., G.E. Emmons, J.E. Miller and R.N. Culman, 1942: Caribbean Weather Characteristics andMethods of Analysis. Air Corps Tactical School, Maxwell Field, Alabama, 165 pp and maps.

STORM Project Office, 1988: Storm Watch. NCAR, Boulder, 9 pp.

Stowe, L.L., Y.Y.M. Yeh, T.F. Eck, C.G. Wellemeyer and H.L. Kyle, 1989: Nimbus-7 global cloudclimatology. Part II: First year results. J. Climate, 2, 671-709.

Subba Reddy, E.V., 1985: A study of squalls at Madras Airport. Mausam, 36, 91-96.

Subramaniam, D.V., and P.S. Jain, 1966: Stability index and area forecasting of thunderstorms. Proc.

12th Conf. on Radar Meteor. Amer. Meteor. Soc., 156-159.

Sugg, A.L., 1969: A mean storm-surge profile. ESSA Tech. Memo WBTM-SR-49, 6 pp.

Sutton, L.J., 1925: Haboobs. Quart. J. Roy. Meteor. Soc., 51, 25-30.

Takahashi, T., 1987: Hawaiian hailstones — 30 January 1985. Bull. Amer. Meteor. Soc., 68, 1530-1534.

Taljaard, J.J., 1972: Synoptic meteorology of the Southern Hemisphere. Meteorology of the SouthernHemisphere. Amer. Meteor. Soc., 139-213.Taniguchi, G., 1967: The effects of typhoons on weather at Clark Air Base. 1st Wea. Wing Tech. Study,10, 64 pp.

Tao, S.-Y., and L.-X. Chen, 1987: A review of recent research on the east Asian summer monsoon inChina. Monsoon Meteorology. Clarendon Press, Oxford, 60-92.

Thomas, S.I.T., and V.K. Raghavendra, 1977: Heights of cumulonimbus cloud tops over the DeccanPlateau and adjoining plains of Andhra Pradesh and east Maharashtra — a radar study. Indian J. Me-teor. Hydrol. Geophys., 28, 479-482.

Tiedtke, M., W.A. Heckley and J. Slingo, 1988: Tropical forecasting at ECMWF: the influence ofphysical parameterization on the mean structure of forecasts and analyses. Quart. J. Roy. Meteor. Soc.,114, 639-664.

Trenberth, K.E., 1980: Atmospheric quasi-biennial oscillations. Mon. Wea. Rev., 108, 1370-1377.

Trenberth, K.E., and J.G. Olsen, 1988: An evaluation and intercomparison of global analyses from theNational Meteorological Center and the European Center for Medium Range Weather Forecasts. Bull.Amer. Soc., 69, 1047-1057.

Trewartha, G.T., 1981: The Earth’s Problem Climates, 2nd Edition. Univ. Wisconsin Press, 371 pp.

Tsonis, A.A., and G.L. Austin, 1981: An evaluation of extrapolation techniques for the short-term pre-diction of rainfall amounts. Atmos. Ocean, 19, 54-65.

Tuller, S.E., 1968: World distribution of mean monthly and annual precipitable water. Mon. Wea. Rev.,96, 785-797.

Turton, J., and S. Nicholls, 1987: A study of the diurnal variation of stratocumulus using a multiplemixed layer model. Quart. J. Roy. Meteor. Soc., 113, 969-1010.

USAFETAC, 1967 —: World-wide Airfield Climatic Data.

U.S. Navy Weather Research Facility, 1963: An examination of the distribution of hurricane forecasterrors using probability ellipses. NWRF 12-1063-81.

U.S. Weather Bureau, 1961: Generalized estimates of probable maximum precipitation and rainfallfrequency data for Puerto Rico and Virgin Islands. USWB Tech. Paper 42.

U.S. Weather Bureau, 1962: Rainfall-frequency atlas of the Hawaiian Islands. USWB Tech. Paper 43.

U.S. Weather Bureau, 1963: Synoptic meteorology as practiced by the National Meteorological Cen-ter. NAWAC Manual.

U.S. Weather Bureau Hydrometeorological Section and U.S. Army Corps of Engineers, EngineeringDepartment, 1943: Maximum possible precipitation over the Panama Canal Basin. Hydrometeor. Rept.,4.

van Bemmelen, W., 1922: Land- und Seebrise in Batavia. Beitr. z. Phys., freien Atmos., 10, 169-177.

Velden, C.S., 1989: Observational analyses of North Atlantic tropical cyclones from NOAA polar-orbiting satellite microwave data. J. Appl. Meteor., 28, 59-70.

Velden, C.S., B.M. Goodman and R.T. Merrill, 1991: Western Pacific tropical cyclone intensity esti-mation from NOAA polar-orbiting satellite microwave data. Mon. Wea. Rev., 119, 159-168.

Verploegh, G., 1967: Observation and analysis of the surface wind over the ocean. Mededel. Verhand.,89, 67 pp.

Vincent, D.G., 1982: Circulation features over the South Pacific during 10-18 January 1979. Mon.Wea. Rev., 110, 981-993.

Vincent, D.G., 1985: Cyclone development in the South Pacific convergence zone during FGGE, 10-17 January 1979. Quart. J. Roy. Meteor. Soc., 111, 155-172.

Vojtesak, M.J., K.P. Martin and G. Myles, 1990: SWANEA (Southwest Asia and Northeast Africa). Aclimatological study. Vol.1 — The Horn of Africa. USAFETAC/TN-90/004, 242 pp.

Vojtesak, M.J., K.P. Martin, G. Myles and M.T. Gilford, 1991: SWANEA (Southwest Asia and North-east Africa). A climatological study. Vol. III — The Middle East peninsula. USAFETAC/TN-91/002,246 pp.

Vonder Haar, T., and A.H. Oort, 1973: New estimate of annual poleward energy transport by NorthernHemisphere oceans. J. Phys. Oceanogr., 2, 169-172.

von Ficker, H., 1936: Die Passatinversion. Veröff. Meteor. Inst. Berl., (1) 4, 1-33.

Walker, G.T., 1924: Correlations in seasonal variations of weather. X. Applications to seasonal fore-casting. Mem. India Meteor. Dept., 24, 333-345.

Wallace, J.M., and F.R. Hartranft, 1969: Diurnal wind variations, surface to 30 kilometers. Mon. Wea.Rev., 97, 446-455.

Wallace, J.M., T.P. Mitchell and C. Deser, 1989: The influence of sea-surface temperature on surface

wind in the eastern equatorial Pacific: seasonal and interannual variability. J. Climate, 2, 1492-1499.

Walters, K.R., Sr., 1988: A descriptive climatology for Baledogle, Somalia. USAFETAC/TN-88/001, 23 pp.

Walters, K.R., Sr., A.G. Korik and M.J. Vojtesak, 1989: The Caribbean basin — a climatological study.USAFETAC/TN-89/003, 342 pp.

Walters, K.R., Sr., and W.F. Sjoberg, 1988: The Persian Gulf region — a climatological study.USAFETAC/TN-88/002, 71 pp.

Wang, S.-T., H. Cheng, C.-H. Hsu and T.-K. Chiou, 1985: Environmental conditions for heavy precipi-tation during May-June over Taiwan Area. Conf. Wea. Anal. Forecasting, Taipei. Meteor. Soc. Taipei,55-72.

Warren, S.G., C.J. Hahn, J. London, R.M. Chervin and R.L. Jenne, 1989: Global distribution of totalcloud cover and cloud type amounts over the ocean, NCAR Tech. Note TN-317-STR.

Weldon, R., 1979: Cloud patterns and the upper air wind field. AWS/TR-79/003, 101 pp.

West, R.C., 1957: The Pacific Lowlands of Colombia, a Negroid Area of the American Tropics. Loui-siana State Univ. Press, Baton Rouge.

Wexler, R.L., 1967: Annual versus daily rainfall: Southeast Asia. AWS Tech. Rept., 196, 226-242.

Whiton, R.C., and R.E. Hamilton, 1976: Radarscope interpretation: severe thunderstorms and torna-does. AWS-TR-76-266, 29 pp.

Wiederanders, C.M., 1961: Analysis of monthly mean resultant winds for standard pressure levels overthe Pacific. Hawaii Inst. Geophys. Rept., 13, 83 pp.

Williams, K.T., and W.M. Gray, 1973: Statistical analysis of satellite-observed tradewind cloud clus-ters in the western North Pacific. Tellus, 25, 313-336.

Williams, M., and R.A. Houze,Jr., 1987: Satellite-observed characteristics of winter monsoon cloudclusters. Mon. Wea. Rev., 115, 505-519.

Wilson, J.W., 1968: Use of radar in short-period terminal weather forecasting. Proc. 13th Radar Me-teor. Conf., Amer. Meteor. Soc., 400-407.

Wilson, J.W., R.D. Roberts, C. Kessinger and J. McCarthy, 1984: Microburst wind structure and evalu-ation of doppler radar for airport wind shear detection. J. Climate Appl. Meteor., 23, 898-915.

Winner, D.C., 1968: Climatological estimates of clock-hour rainfall rates. AWS Tech. Rept., 202.

Woodley, W.L., and A. Herndon, 1970: A raingauge evaluation of the Miami reflectivity-rainfall raterelationship. J. Appl. Meteor., 9, 258-263.

Working Group on Numerical Experimentation, 1987: Report of the third session. WMO/TD 223, 34pp and appendices.

World Meteorological Organization, 1956a: International Cloud Atlas (abridged). WMO, 62 pp., 73plates.

World Meteorological Organization, 1956b: World distribution of thunderstorm days. WMO No. 21TP 6 Part 2.

World Meteorological Organization, 1966: International meteorological tables. WMO No. 186 TP 94.

World Meteorological Organization, 1979: Operational techniques for forecasting tropical cycloneintensity and movement. WMO 528.

World Meteorological Organization, 1986: Weather reporting; observing stations. WMOP/OMM 9.

World Meteorological Organization, 1989: World Weather Watch: fourteenth status report on imple-mentation. WMO 714.

Wylie, D., B.B. Hinton and K. Kloesel, 1989: The relationship of marine stratus clouds to wind andtemperature advections. Mon. Wea. Rev., 117, 2620-2625.

Xu, J.M., 1990: Chinese polar orbiting meteorological satellite. In: East Asia and Western PacificMeteorology and Climate (P. Sham and C.P. Chang, eds.), World Scientific, Singapore, 131-138.

Yang, S., and P.J. Webster, 1990: The effect of summer tropical heating on the location and intensity ofthe extratropical westerly jet streams. J. Geophys. Res., 95, 18,705-18,721.

Zehr, R.M., 1987: The diurnal variation of deep convective clouds and cirrus with tropical cyclones.Preprints, 17th Conf. on Hurricanes and Tropical Meteorology, April 7-10, Miami, Florida. Amer.Meteor. Soc., 276-279.

Zehr, R.M., 1989: Improving objective satellite estimates of tropical cyclone intensity. Preprints, 4thConf. Satellite Meteorology and Oceanography, May 16-19, 1989, San Diego, CA.Amer. Meteor. Soc., J25-J28.

Zipser, E.J., 1969: The role of organized unsaturated convective downdrafts in the structure and rapiddecay of an equatorial disturbance. J. Appl. Meteor., 8, 799-814.

Zipser, E.J., 1977: Mesoscale and convective-scale downdrafts as distinct components of squall-linestructure. Mon. Wea. Rev., 105, 1568-1589.