the operational meteorology of convective weather volume i: operational mesoanalysis

1

NOAA TECHNICAL MEMORANDUM NWS NSSFC-5NTIS Accession No. PB83-162321

THE OPERATIONAL METEOROLOGYOF CONVECTIVE WEATHER

VOLUME I: OPERATIONAL MESOANALYSIS

Charles A. Doswell IIINational Severe Storms Forecast CenterKansas City, Missouri

November 1982

UNITED STATES DEPARTMENT OF COMMERCENational Oceanic and Atmospheric AdministrationNational Weather Service

Electronic reprint produced byTim VasquezWeather Graphics Technologies / www.weathergraphics.com

March 2000

2

I appreciate that Tim Vasquez is willing to make this available, since Icontinue to get requests for this material. There is a limited supply ofhard copy versions still available; I have the world�s supply of themand they can be had without charge by sending me a mailing address,as long as the supply holds up. I also have copies of the secondvolume in the two-volume series: Storm-Scale Analysis.

This first volume was written in 1982 and still contains material I amnot ashamed of, but a lot has happened since then to make the contentsomewhat obsolete. For example, Q-vector diagnostics, PotentialVorticity thinking, and the ideas associated with Moist SymmetricInstability and Slantwise Convection have all been developed after thistech. memo. was finished. If I had it to do over, I think I could revisethe discussion of synoptic-scale vertical motion in a more useful waythan is currently given. In general, my ideas have not remained staticsince 1982, and I�ve had to mull over occasional impulses towardrevising both of the tech. memos. since both of them are on the longslide to increasing obscelesence. However, I�ve chosen not to do so.Rather, it�s my intention eventually to co-author a textbook (or two) thatwill incorporate a lot of what I�ve learned since 1982. Note that in thesecond volume, I implied that there would be a third volume with lotsof examples of the methodology in action. Time and higher prioritieshave made it clear that this third volume simply will never get done,although I have some of the major components of it assembled (more orless), so my intentions were good. Hopefully, all of this revision andinclusion of new material will be made manifest in the textbook I haveplanned to write.

At least some of this tech. memo. should remain relevant well into thenext century, so I hope that Tim�s making it available will be useful tothe readers. In the meantime, please visit my Websites at:

http://www.nssl.noaa.gov/~doswell/

http://webserv.chatsystems.com/~doswell/

where you can find a number of essays and other Web pages devotedto some of the issues covered in this old work of mine.

Chuck DoswellNorman, Oklahoma, Spring 2000

PREFACEMarch 2000

3

PREFACE TO

THE OPERATIONAL METEOROLOGY OF CONVECTIVE WEATHERVOLUME I: OPERATIONAL MESOANALYSIS

Primary causes are unknown to us; but are subject to simple and constant laws, which may bediscovered by observation, the study of them being the object of natural philosophy.

� Joseph Fourier, Theory of Heat

There is no other species on Earth that does science. It is, so far, entirely a human invention ... It has tworules. First: there are no sacred truths; all assumptions must be critically examined; arguments fromauthority are worthless. Second: whatever is inconsistent with the facts must be discarded or revised ...The obvious is sometimes false; the unexpected is sometimes true.

L.F. Richardson was a British meteorologist interested in war. He wished to understand its causes. Thereare intellectual parallels between war and weather. Both are complex. Both exhibit regularities,implying that they are not implacable forces but natural systems that can be understood and controlled.To understand the global weather you must first collect a great body of meteorological data; you mustdiscover how weather actually behaves.

� Carl Sagan, Cosmos

There is a growing accumulation of evidence to indicate that man has no direct contact with experienceper se but that there is an intervening set of patterns which channel his senses and his thoughts, causinghim to react one way when someone else with different underlying patterns will react as his experiencedictates.

It is time, however, that we began to realize that much of what passes for science today may have beenscientific yesterday but can no longer qualify because it does not make any additional meaningfulstatements about anything. It blindly adheres to procedures as a church adheres to its ritual.

� E.T. Hall, The Silent Language

We can never have enough of nature. We must be refreshed by the sight of inexhaustible vigor, vast andtitanic features, the sea-coast with its wrecks, the wilderness with its living and its decaying trees, thethundercloud, and the rain which lasts three weeks and produces freshets. We need to witness our ownlimits transgressed, and some life pasturing freely where we never wander.

� Henry David Thoreau, Walden

4

These notes have been developed in an effort,however imperfect, to acquaint meteorologists inan operational environment with the basicconcepts of convective weather systems. It is asad fact of life that many of today�s operationalmeteorologists have never been given a physicalinterpretation of the dynamics which areunderstood to govern the atmosphere and, inparticular, convection. It is not my intent to becompletely exhaustive, although the length of thetext leads me to fear that it may be exhausting!

There are numerous threads which can beused to sew up the package I am trying to deliver.In trying to unravel them, I have at times assumedthe reader knows things with which he/she maynot, in fact, be familiar. Conversely, I have attimes assumed the reader�s ignorance of somebasic ideas which I have felt important enough toexplain in detail and, in the process, may havebored more advanced readers. I hope that bothforms of exasperation never reach the breakingpoint.

In any work of this sort, it is easy to find thematerial one wrote a few months beforesomewhat less than satisfactory in light of newfindings, recent publications, or just plain furtherthought. One has to stop the process of revisionssomewhere, but I suspect we are at the start of anexciting new era in applied meteorology and hereI am trying to summarize the proverbial �state ofthe art�. Since I cannot hope to be completely up-to-date by the time this reaches the hands of thereaders, I have tried to give enough material tobring the interested reader to the point of a self-sustaining, self-education process. If the reader iscontent to absorb only what is in these notes, myeffort will not have succeeded.

While this preface is being written, Volumes IIand III are still embryonic. The reader will notethat there are many references to other parts of.the text within the body of these notes. Theseinternal references follow an outline-type ofstructure of the form I.III.A.3.b..., where theleading, underscored Roman numeral refers to thevolume number. This is omitted when thereference is within the given volume. The secondRoman numeral refers to the chapter in thevolume, the capital letter to the sub-heading, andso forth. Since the second and third volumes are

not yet finished, I can only promise that they willbe completed as rapidly as possible. Becausethese self-references generally concernamplifications or additional discussions of thereferenced topics, it should not he terriblydetrimental for them to be as yet unavailable. Ifthe material were essential, it would have beenincluded at that point in the notes.

The reader should also note that all footnotesin a given chapter will be collected at the end ofthat chapter. This is not the most convenientapproach, but it happens to solve a nasty problemin trying to fit these notes into a readable text. Myapologies for any inconvenience.

As in any large work, numerous contributorshave made these notes possible. The Chief of theTechniques Development Unit of NSSFC, Dr.Joseph T. Schaefer, has perhaps been mostvaluable as an encourager (and occasionalpushes to complete this work are appreciated), asounding-board for many of the topics containedherein, an editor, and a respected colleague. Dr.Robert A. Maddox of NOAA�s EnvironmentalResearch Laboratories, Office of WeatherResearch and Modification, has provided manyideas, inspiration, and the encouragement only a�kindred spirit� can provide. The Deputy Directorof the National Weather Service Training Center,Mr. Larry Burns, gave me the initial support toundertake this effort and confirmed my perceptionof the need for it in the first place. Numerousindividuals have encouraged me by their interest,including Alan R. Moller (NWSFO, Fort Worth,Texas), Larry Wilson, Steve Weiss, Jim Henderson,and Mike Streib (all at NSSFC), as well as theusual host of those �too numerous to mention.�Valuable reviews were provided by Profs. WalterJ. Saucier, David A. Barber (North Carolina StateUniv.), and Richard J. Reed (Univ. of Washington).Naturally, any errors and misinterpretations aremy sole responsibility. Finally, Beverly Lamberthas suffered through the numerous revisions anddrafts and done an outstanding job with themanuscript preparation.

Charles A. Doswell III

Kansas City, Missouri

November, 1982

5

ContentsI. Introduction ............................................................................................................................... 6A. Preliminary Remarks .................................................................................................................. 6B. Scaling Concepts ....................................................................................................................... 6II. Upper-Air Data Analysis .......................................................................................................... 11A. General Remarks ..................................................................................................................... 11B. Upper Air Chart Analysis .......................................................................................................... 111. Vertical Motion ........................................................................................................................................................... 122. Production of Unstable Thermodynamic Stratification ...................................................................................................... 173. Some Kinematic Considerations .................................................................................................................................... 21

C. Sounding Analysis and Interpretation ........................................................................................ 221. General Remarks............................................................................................................. ............................................ 222. Sounding Thermodynamics ........................................................................................................................................... 233. Sounding Kinematics ................................................................................................................................................... 26

D. The Composite Chart ............................................................................................................... 30III. Surface Data Analysis ........................................................................................................... 33A. General Remarks ..................................................................................................................... 33B. Surface Discontinuities ............................................................................................................ 351. Cold Fronts ................................................................................................................. ................................................ 372. Warm Fronts............................................................................................................................................................... 393. Stationary Fronts........................................................................................................... .............................................. 404. Occluded Fronts .......................................................................................................................................................... 415. Drylines ..................................................................................................................................................................... 426. Land/Sea Breeze Fronts ...................................................................................................... ......................................... 457. Thunderstorm Outflow Boundaries ................................................................................................................................. 46

C. Boundaries Not Involving Air Masses ....................................................................................... 531. Wind Shift Lines.......................................................................................................................................................... 532. Pressure Troughs ......................................................................................................................................................... 54D. Pressure Change Analysis ........................................................................................................ 551. Applications to Synoptic Analysis .................................................................................................................................. 552. The Isallobaric Acceleration .......................................................................................................................................... 553. Mesoscale Isallobaric Analysis ..................................................................................................................................... 57

E. Thermal Analysis ...................................................................................................................... 60F. Terrain Effects.......................................................................................................................... 611. Mountain/Valley Circulations ........................................................................................................................................ 612. Upslope Flow ................................................................................................................ .............................................. 623. The Low-Level Jet ........................................................................................................................................................ 624. Mesoscale Eddies ....................................................................................................................................................... 675. Miscellaneous Examples .............................................................................................................................................. 68G. Flash Floods and Severe Weather ............................................................................................ 71IV. Objective Analysis Tools ........................................................................................................ 74A. Moisture Convergence ............................................................................................................. 74B. Surface Geostrophic Winds ...................................................................................................... 76C. Filtering by Objective Interpolation ............................................................................................ 78D. Upper-Level Divergence ........................................................................................................... 78E. Kinematic Analyses and Trajectories ......................................................................................... 79V. Interpretation of Numerical Guidance ...................................................................................... 84A. General Remarks ..................................................................................................................... 84B. Short and Long Term Error History ........................................................................................... 85C. Initialization and Adjustment .................................................................................................... 85D. Statistical Convective Weather Guidance .................................................................................. 86VI. Concluding Remarks on Mesoanalysis.................................................................................... 88REFERENCES .............................................................................................................................. 90

6

I. Introduction

process of mesoanalysis, but the reader is urged topursue these topics further by consulting thebibliographic references. Some generaldiscussions of mesoanalysis are contained inFujita et al. (1956), Magor (1959), Tepper (1959),and Fujita (1963 . Pieces of the materialconcerning practical mesoanalysis are containedin the references, but to the author's knowledge,these have not been collected in one place. Thepresentation in these notes is essentiallyqualitative and non-mathematical, since arigorous discussion is not necessary to thepracticing mesoanalyst. Many ideas are presentedwithout proof, but it is hoped that the referencematerial will be consulted when doubts arise.

B. Scaling Concepts

Under the general heading of �OperationalMesoanalysis� in these notes, a substantial varietyof phenomena and concepts is presented. It isworthwhile to discuss this in terms ofmeteorological scales at the outset. It should beemphasized that the notion of scaling is absolutelyessential to understanding current and futuremeteorological thinking. We shall attempt toreview current concepts on scales, from that ofthe extratropical cyclone (ETC) down to thosephenomena at the observation limits of the presentnetwork of routine surface reports. One exampleof a proposed ordering of meteorologicalphenomena by scales is shown in Fig. 1.1.

The large-scale limit to our discussion can begiven by some arbitrary order of magnitudeestimates for scaling lengths (say, horizontallengths of 103 km, vertical depths of 10 km, andtime scales of 105 s [about 1 day]. Note that thesethree values, suitably manipulated (as in Haltinerand Williams, 1980), can yield approximatevalues for most of the terms in the equationsgoverning large-scale flows. The suitability of the

A. Preliminary Remarks

Operational mesoanalysis is most oftenconsidered in the context of convective storms.Mesosystems significant to operational forecastingdo not only encompass deep convection, as thepatterns of heavy snowfall sometimes suggest, forexample. Also, it is not clear that the process ofmesoanalysis for convective storms can betransferred totally for application to, say, winterstorms, although good analysis techniques arerequired in both areas. In any case, these noteswill not address mesoanalysis associated withnon-convective weather.

In order best to accomplish operationalmesoanalysis, one should have a thoroughunderstanding of synoptic-scale meteorology.Further, one should be familiar with convectivestorms and their dynamics. This is easy to say, butdifficult to satisfy. No one person has a completeunderstanding of either one of these areas,especially the latter. Much remains to be learnedabout convective storm dynamics. Regrettably,there seems to have been a trend away fromsynoptic meteorology, both in the universities andwithin the operational arena as well (Doswell etal., 1981). Increasing dependence on numericalmodels has led to an overall decline in the skillsof the synoptic meteorologist (Snellman, 1977).Additional evidence for this decline can be seenin the frequent reference here to texts and journalarticles published in the 1950s. If more recentreferences were available, they would have beenused, but the lack of interest in relating dynamicto synoptic meteorology (and vice versa) over thelast two decades has led to the paucity of morerecent references.

Realistically, these notes cannot provide aworking knowledge of both synoptic meteorologyand the dynamics of convective storms. Materialin these areas will be covered, as it relates to the

7

manipulation hinges, in large measure, onknowing the answers we want before we begin.In other words, a formal scale analysis isessentially a way of justifying makingmathematical assumptions to describetheoretically a problem for which we alreadyhave observed the answer! In the process, we cangain insights which may have not been previouslyobvious, and considerable physical understandingcan be gained. Perhaps the most successfulapplication of scale analysis is in the problem ofour large-scale limit, the extratropical cyclone.

However, such a formal approach may not be theeasiest to understand from an operationalviewpoint and it suffers from a major deficiency:

namely, on our lower scale limit, we do not haveas clear a picture of the desired answer to beobtained. Instead, we consider a more physically-motivated way of establishing the scale ofphenomena which draws heavily from thediscussions by Emanuel (1980). By doing so, it ishoped that the reader can relate the discussion toobserved daily weather events and will thereforebe encouraged to pursue the topic as moreformally developed in the references (Holton,1979; Palmen and Newton, 1969; Haltiner andWilliams, 1980).

It is convenient that our upper scale limit is theextratropical cyclone, since that weather systemis probably the best understood. Without going

Fig. 1.1. Scale definitions and different meteorological phenomena with characteristic temporal andhorizontal spatial scales (after Orlanski, 1975).

8

into details, the essential physical mechanismdriving the extratropical cyclone is known asbaroclinic instability. The phenomenon itself (theETC) was first described qualitatively by the so-called �Bergen School� (Bjerknes, 1919; Bjerknesand Solberg, 1921, 1922) via the �Polar FrontModel.� This is summarized in Fig. 1.2, whichshows the basic structure and evolution of anextratropical cyclone. A variety of explanationswere put forward to explain the underlyingprocess during the ensuing decades, but the lackof adequate upper-air data prevented anysatisfactory explanation for nearly 30 years. Then,the insights of Rossby (1940) and Charney (1947)provided the long-sought answer in quantitativeterms which have come to be known asbaroclinic instability.

This instability theory can be fairly easilysummarized without mathematics. We begin withthe fact that the north-south variation in solarheating results in a north-south temperaturegradient. With the observation that this gradient isnot uniformly distributed, but is concentrated inmid-latitudes, forming the so-called polar front,physical reasoning can be used to show that overthe front the westerly winds must increase withheight (with the increase being proportional to thestrength of the temperature gradient).1 Thisincreasing westerly wind with height, or verticalshear, intensifies as the unequal heating

continues. The extratropical cyclone forms as theprimary process by which this strong gradient isalleviated. In essence, the unequal heating storesup potential energy and when enough is stored upto trigger baroclinic instability, the developingcyclone draws on this reservoir of potentialenergy to drive the circulation (thus producingkinetic energy). When the reservoir drops belowsome critical level, the system then begins todecay and the circulation slowly winds down.Along the way, the storm has moved warm airupward and northward, while cold air hastravelled downward and southward. Therefore,the flow has acted to relieve the strongtemperature gradients which initiated the system.

Fig. 1.2. Life cycle of extratropical cyclone (after J. Bjerknes, from Godske et al., 1957). In middlefigures, thin lines are sea-level isobars. Top and bottom figures show schematic clouds, frontal surfacesand tropopause along lines n, a little north and south of ETC center. The times from stages a to c andfrom c to e are roughly one day in each case.

Fig. 1.3. The balance of forces for geostrophicequilibrium (after Holton, 1979). The pressuregradient force is denoted by P and the Coriolisforce by C

o, while the resultant geostrophic wind is

Vg.

9

One of the observationally verifiable notionswhich has allowed treatment of baroclinicinstability from a theoretical viewpoint is the basicvalidity of geostrophic balance on the scale of theextratropical cyclone. That is, the observed windsare pretty close to geostrophic, except perhapsnear the surface. This observation has beenincorporated in the analysis of extratropicalstorms under the general heading ofquasigeostrophic theory (see e.g., Holton, 1979,Chap. 6 and also II.B.1). Basically, the geostrophicwind is parallel to the isobars (or the contours, inpressure coordinates), with low pressure (heights)on its left, and with speed proportional to themagnitude of the pressure (or height) gradient (Fig.1.3).

However, one might easily be led to ask somepotentially embarrassing questions about this stateof balance. For example, if the geostrophic windis so good at approximating the true wind, how dopressure systems deepen (or fill)? If the windspeed happens to be non-geostrophic (i.e.,ageostrophic) for some reason, how do the windsand/or pressures re-adjust to geostrophy? Sincethe geostrophic wind is not divergent,2 can wesay then that vertical motion is unimportant forbaroclinic instability?

We shall not explore the answers to all thesequestions in these notes, but once again refer thereader to the references. However, the subject ofhow the winds and pressure field come to adjustthemselves to a state of near-geostrophic balancehappens to be relevant to the issue of scale. Themanner in which the adjustment occurs dependson the scale of the pressure system (Rossby, 1938).Specifically, on the small scale, the pressure fieldchanges to fit the winds while on the largescale,the winds adjust the pressure field. But howsmall is �small� and how large is �large�? It turnsout that we can define a length scale called theRossby radius of deformation, lambda, which isrelated to the problem of geostrophic adjustment.Physically, the adjustment is accomplished bygravity waves which travel at relatively fastspeeds. If we take this gravity wave speed anddivide it by the Coriolis parameter3 (the reciprocalof the Coriolis parameter defines a time scaleappropriate to geostrophic balance), we obtainthe Rossby radius of deformation. This can beinterpreted as the influence radius of the gravity

waves which accomplish the adjustment. Forlength scales much less than lambda, the gravitywaves have time to reach any point in the systemand they act to adjust the pressure field. Forlength scales much greater than lambda, gravitywaves can not penetrate the entire system and thewinds have time to adjust to the pressure.

Just how large is lambda ? It happens that lambdais about 1500 km, which is a length of the sameorder as that of the ETC. Since these disturbancesare neither much larger nor much smaller thanlambda , we can conclude that �synoptic-scale�systems adjust both their wind and their pressurefields to maintain a state of near-geostrophicbalance. Such a conclusion should be readilyapparent to those who deal with the weatheroperationally. The Rossby radius of deformationalso provides a useful clue to the behavior of the�short wave� troughs in the atmosphere, and thesmaller-scale features in the jet stream. Sincethese smaller features have lengths perhaps assmall as 300 km, one expects their pressure fieldsto react to non-geostrophic winds rather thanvice-versa. Again, operational experiencesupports this conclusion.

We have established our large-scale limit as theRossby radius of deformation. In doing so, wehave made a somewhat less arbitrary choice thanis often made, since it is based on well-acceptedtheory and observational experience. That is,baroclinic instability (which is widely accepted asthe dominant physical mechanism in extratropicalcyclones) requires both wind and pressureperturbations to operate, limiting the scales ofthese weather systems to near the Rossby radiusof deformation.

Can we motivate a definition similarly for whatwe call �mesoscale� � i.e., our lower limit ofconsideration in this section? The main issue indeveloping a physical-dynamical definition ofmesoscale is whether or not there exists adominant, scale-dependent instability whichforces mesoscale systems. Emanuel (1980) hassuggested the so-called �symmetric� instability forthis purpose, but he also leaves open thepossibility that other processes may exist and bephysically significant. His basic definition ofmesoscale is that on such a scale, both Coriolisaccelerations and ageostrophic advection areimportant.4 This approach seems entirely

10

reasonable, and symmetric instabilities do,indeed, operate on such length scales (about 100km). Further, this scale definition turns out to lie atabout the resolution limit of operational surfacedata. Hence, this is probably the best choice forour lower scale limit, even if the dominantphysical process is not as clearly established ason the larger scale.

It does seem clear that on scales below 100 km,the Coriolis acceleration becomes moredynamically irrelevant, while on scales muchlarger than 100 km, the ageostrophic contributionto advection becomes decreasingly significant.Quantitatively, this is accounted for by the RossbyNumber (Ro) � the ratio of the actual to theCoriolis acceleration. Thus, Ro is small for lengthscales of 1000 km or more, and large for scalesbelow 100 km. Around 100 km, Ro~1, which saysthat the Coriolis and actual accelerations areabout the same.

Theory suggests that for these intermediate scales,a wide variety of instabilities are possible and theactually occurring combination of parametersmay determine which process is most unstable ina given situation. This variety of theoreticalinstabilities is plausible when we realize that amuch greater range of phenomena is seen to existon the mesoscale than on larger scales. The ETCis by far the dominant form of weather systemoperating at scales near the Rossby radius (at mid-latitudes), whereas we shall see that a lot offundamentally different phenomena occur in themesoscale range.

Further, it is not clear on this scale what sort ofdominant force balances exist, if any, analogousto geostrophic balance on the large scale.Hopefully, future research will provide someinsight into mesoscale instabilities and allow aclearer picture to emerge of what �mesoscale�really implies about the dynamics of systems. Atthis time, it seems plausible to suggest that frictionand latent heat are likely to have larger roles thanthey play in baroclinic instability. Since these twofactors have proven difficult to treat in theoreticalmodels, considerable time may elapse before wecan treat mesoscale processes on the same levelas we now deal with the ETC.

Finally, the density and frequency of upper airdata may well prove to be the barrier to our

mesoscale understanding that they once were onthe large scale. It is difficult for meteorologists toattempt an explanation of phenomena they havenot routinely observed, since the mathematics ofatmospheric flow allow a bewildering variety ofsolutions. Only by careful comparison withobservations can plausible theories be selectedfrom the vast array of candidates. Since�mesoscale� observations are still not routinelyavailable, only limited conclusions can beredrawn from the limited mesoscale data.

CHAPTER I FOOTNOTES

1 I.B: This physical reasoning is based on theconcept of the thermal wind (see e.g., Holton,1979, p. 68ff and also II.8.2), which is in turn anapplication of the geostrophic wind law, validonly for large-scale flow.

2 I.B: This is not exactly true, as wc shall see inIV.B.

3 I.B: The Coriolis Parameter (often denoted by�f�) can be thought of as the vorticity of the earthabout the local vertical. Thus, at the north pole,where the local vertical is also the earth�s rotationaxis, f is simply the earth�s vorticity (twice itsrotation rate, or 1.4584 x 10-4 s-1). Since the localvertical increases its departure from the earth�srotation axis as one moves away from the poles,the Coriolis parameter decreases with latitude,and vanishes at the Equator. The rate of decreasein f is slow at high latitudes (f is 1.0313 x 10-4 s-1 at45 deg N), but increases rapidly, reaching itsmaximum at the equator itself. Coriolis parameterchanges signs upon crossing into the SouthernHemisphere so, for example, the SouthernHemisphere geostrophic wind blows with lowpressure on its right.

4 I.B: In the case of large-scale motions justdescribed, the advection of atmosphericproperties is dominated by the geostrophiccontribution. In fact, this is a cornerstone ofquasigeostrophic theory.

11

II. Upper-Air Data Analysissatellite imagery to the synoptic scale analysisproblem.

A 4-dimensional understanding is possible, evenwith limited time constraints, using centrallyanalyzed charts at the standard upper levels. Asdetailed by Maddox (1979b), these upper air andsurface maps can and should be enhanced toemphasize features of importance to convectivestorm forecasting. At SELS, analysis of upper levelcharts is done by hand, as well. Althoughsubjective analysis has numerous drawbacks froma theoretical and aesthetic viewpoint, it is anexcellent way of accomplishing severalworthwhile goals. These include: (1) all of thedata are subjected to examination, thuspinpointing erroneous observations, convection-contaminated soundings, and so forth; (2) theprocess of �redrawing lines� forces an awarenessof the significant upper-air features; and (3) ananalysis of upper-air maps can be accomplishedwhich is oriented toward mesoanalysis � i.e., theheavy smoothing necessary for large-scalemodelling purposes can be avoided. Many textsexist to help guide the process of synoptic-scaleanalysis (e.g., Saucier, 1955; Petterssen, 1956a;Godske et al., 1957).

The forecast day generally begins with themorning (1200 GMT) soundings. The data at thattime are relatively free of convectivecontamination. This is somewhat less true in thelate spring and summer, when convection maycontinue through the night and on into the nextday (note the discussion by by Maddox, 1980b).Nevertheless, the morning analysis should allowthe forecaster to develop a relatively clear pictureof the synoptic-scale setting for the afternoon�sand evening�s developments.

B. Upper Air Chart Analysis

If the analyst has the option of contouring theconstant pressure level charts, rather than simplyenhancing the facsimile (or AFOS) products, the

A. General Remarks

It must be pointed out immediately that thenetwork of upper air observations is entirelyinadequate for any true mesoanalysis. Withroutine soundings over the U.S. only availableevery 12 h, at an average separation of about 400km, no analysis can be considered mesoscale.

Nevertheless, this is where mesoanalysis shouldbegin. It cannot be overemphasized that aforecast should start with a 4-dimensional mentalpicture of the atmosphere. Thus, some of theanalyst�s most important efforts should be directedtoward developing this 4-dimensionalunderstanding. With the development andapplication of sophisticated remote sensing tools(specifically, radar and satellite imagery), newunderstanding of many aspects of convection hasbeen obtained rapidly. It should be completelyobvious that analysis should not be done withoutexamination of all the available data. The processof analysis is, in no small part, heavily dependenton the skill of the analyst at integrating a varietyof data into a unified picture (i.e., a synthesis).Although these notes by themselves cannotprovide the reader with all the necessaryknowledge to interpret remote sensing data, someelements will be presented in those areas wheresuch data can be crucial in the analysis process.

Remote sensing data can have a real impact onthe upper-air analyses, in two related ways. First,the position and strength of upper air systems canbe refined, based on the cloud and precipitationpatterns. Second, and more importantly,information from the data-void areas (e.g., overthe oceans) may have a real impact, eitherdirectly (e.g., a feature in the Gulf of Mexicowhich can move onshore later in the forecastperiod) or indirectly (e.g., a misanalyzed shortwave trough which results in a faulty numericalprognosis [Hales, 1979a]). See Anderson et al.(1974) or Weldon (1979) for applications of

12

basic process is relatively straightforward. At 850and 700 mb, the Severe Local Storms ForecastUnit of NSSFC (SELS) analyzes for height (30 mcontour interval), temperature (2 deg C isotherminterval) and dewpoint temperature (2 deg Cisodrosotherms, starting with 8 deg C at 850 mband 0 deg C at 700 mb). At 500 mb, heights (60 mcontours) temperatures (2 deg C isotherms) and12-h height changes (30 m isallohypses) areanalyzed. At 250 mb, isotachs (20 kt interval) andaxes of maximum wind are depicted. Examples ofSELS-type analyses shall be shown in III.V.

Within some limits, the development of this basicset of charts follows standard analysis practice(see Saucier, 1955, ch. 4). As described by Miller(1972), the analyst should avoid drawing closedisopleths whenever possible, even at theoccasional expense of creating long, narrow�ribbons�. There is good evidence that theatmosphere really does tend to create suchfeatures and the basic idea is to emphasize thesource regions.

An important departure from synoptic scalepractice is a heavy emphasis on 12-h changes inthe observations. The SELS routines which plot theupper-air data provide a 12-h change for allplotted variables, including the winds. Rather thanemphasizing chart-to-chart continuity, the severeweather analyst needs to recognize thesignificance of the chart-to-chart changes. Ofcourse, some effort should be made to developtime continuity, but the upper air data bythemselves are too sparse in space and time toprovide a clear picture of the often subtle featureswhich move through the synoptic-scale patterns.Short wave troughs, wind maxima, vorticity�lobes� and small-scale temperature anomaliesare frequently too small to be analyzed in detailunless 12-h height changes, backing/veeringpatterns of the wind, and thermodynamic changesare examined.

There are two complicating factors in evaluatingthe change fields: normal diurnal variations (e.g.,Harris, 1959) and the contamination of therawinsonde observations by convection. Theanalyst should know and recognize the expecteddiurnal changes (e.g., high 700 mb temperaturesat 0000 GMT over the mountains; roughly 20 m12-h height rises at 1200 GMT or falls at 0000GMT in mid-latitudes at 500 mb). While diurnal

effects are at least conceptually easy to accountfor, convection can produce large changes thatare less easy to adjust. Studies by Ninomiya(1971a,b), Maddox (1979a, 1980a), and othershave shown that large thunderstorm complexes(up to 500,000 km2, often lasting for 8 hr or more)can have a dramatic influence on even synoptic-scale rawinsonde networks. Since the effects ofconvection cannot be isolated, the correction ofconvectively contaminated data is basically notpossible. Radar and satellite data should beexamined routinely during analysis so that theanalyst can exercise caution in interpreting thedata within convective regions.

Most, if not all, of the effort spent by a severeweather analyst forecaster in examining upper-airdata is directed toward finding where upwardvertical motion will occur in regions of moist,unstably stratified air (Beebe and Bates, 1955).This being the case, the real job of analysis shouldbe directed toward this end, not merely drawinglines on the charts.

1. Vertical Motion

By way of introduction, one might ask thephysical reason for a meteorologist�spreoccupation with vertical motion. Theproduction of �weather� requires condensationand the most common way the atmosphereproduces condensation is adiabatic cooling byexpansion. This results from lowering thepressure. Since in horizontal motion parcels tendto travel parallel to isobars, no important changein pressure results. Local pressure falls do, ofcourse, occur but their magnitude is so small incomparison to what is required to saturate parcelsthat their effect is not significant (but pressure fallsare important in other ways, of course; see III.D).Since pressure surfaces are so closely packed inthe vertical through the troposphere, a smallvertical displacement can result in a large changein pressure. Naturally, this is reflected in thenormal state of large-scale hydrostatic balance,where the relatively large vertical pressuregradient force is, compensated for by gravitationalacceleration. What small vertical accelerationsoccur are quite negligible in comparison, but stillare our primary source of weather. For

13

meteorologists the physical significance ofvertical motion is ultimately the reduction ofpressure following a parcel which results incondensation.

In examining upper-air data to locate �features�, abasic problem is the diagnosis of regions ofupward vertical motion. Upward motion on thescale of the upper air data is in the range of a fewcm s-1 . This illustrates the essentially horizontalnature of large-scale flow, since the verticalcomponent can be less than a tenth of onepercent of the horizontal wind. However, sincethis upward motion is sustained for long periods, itcan have dramatic effects. If one maintains a 5cm s-1 upward motion for 24 hr, the net vertical liftis more than 4 km! Further, if the parcel started ata pressure of 1000 mb, that amount of lift reducesthe pressure to about 600 mb. Since the surface ofthe earth and the tropopause act effectively asbounding surfaces for vertical motions, a region ofupward vertical motion must have convergenceat its base and divergence at its summit. This is aconsequence of the law of mass continuity. Thus,divergence undergoes a change in sign withheight, leading to the concept of the so-calledlevel of nondivergence. Actually, this �level� israrely at the same height from place to place andtime to time. Rather, it is typically a slopingsurface (Fig. 2.1), as described by Charney (1947).Therefore, the axis of strongest vertical motionsmay be somewhat tilted away from the vertical.The interest in divergence is, therefore, an

extension of the need to assess large-scalevertical motion.

A basic effort in analysis is to infer upper leveldivergence from such features as short-wavetroughs, jet maxima, vorticity advection, and soforth (see McNulty, 1978; or Kloth and Davies-Jones, 1980 for discussions on these topics).Owing to several difficulties, we often must relyon such subtle approaches to diagnosedivergence. One basic problem is that we haveavailable only 12-hourly samples: in the morningwhen mesosystems may not be well developed,and again in the evening when convection isusually already underway. Organized regions ofupper-level divergence are hard to follow as aresult. Another, frequently mentioned problem isexemplified if we consider a 5 cm s-1 upwardmotion at a height of 5 km. This implies that theaverage low-level convergence in the layer fromthe surface to 5 km is 10-5 s-1. This, in turn,suggests horizontal wind differences in the rangeof 1 m s-1 over a distance of 100 km. Smallchanges in the data (say 10% of the observedwind speed) can result in a large change (in therange of 100%) in the calculated divergence and,hence, the vertical velocity.

Given the small magnitude of synoptic-scalevertical motion and the modest changes inhorizontal wind needed to produce it, the role ofquasigeostrophic theory becomes more clear. Formost purposes, and specifically for horizontaladvection, the geostrophic flow is good enough.The divergence needed for vertical motion is notcontained in the geostrophic wind, but the theorycan be used to evaluate it. In effect, the verticalmotion is the result of a secondary flow (muchweaker) which is required to maintain a state ofnear-geostrophic (and hydrostatic) balance. Thissecondary circulation is a cornerstone ofquasigeostrophic theory (and explains why theterm is quasigeostrophic) and its validity is seen inits value for diagnosis of real weather systems.

Vorticity advection is widely accepted as anindirect means of locating large-scale upwardmotion. By vorticity advection, we mean a patternof height contours and vorticity isopleths as shownin Fig. 2.2. For this indirect method to work, avariety of assumptions is necessary. The first twoassumptions are that the actual winds are closelyapproximated by the geostrophic winds (which

Fig. 2.1. Vertical cross section (after Fleagle,1948) of horizontal divergence relative to troughand ridge lines (dotted and dash-dotted lines,respectively). Divergence contours in units of 10-6

s-1.

14

parallel the contours) and that the vorticity field isderived from the height field (i.e., is essentiallygeostrophic) and so is moving slower than thewinds. Under these restrictions, a parcel movesthrough the vorticity pattern, and finds its originalvorticity different from that of its environment.Another assumption is that the basic process bywhich the parcel changes its vorticity isdivergence (or convergence), so if a parcel ismoving into regions of lower vorticity (as in aregion of positive vorticity advection [PVA]) theremust be a tendency for divergence to bring theparcel�s vorticity down to that of ite environment.

Petterssen (1956a, p. 299ff) presents the PVAarguments as follows: at lower levels, vorticityadvection is weak since the flow is very nearlyparallel to the vorticity isopleths. Therefore, atthose levels, vorticity changes are dominated bydivergence effects. Regions of increasing vorticity

must be convergent (and vice versa) at low levels.At upper levels, vorticity advection is large butlocal changes are small in comparison. Air passesthrough the vorticity pattern since wind speedsare high, so the arguments (above) apply whichsuggest that PVA implies divergence. At middlelevels (500 mb), divergence is small and vorticityis very nearly conserved � local changes invorticity are dominated by advection. Historically,this is why 500 mb was chosen for earlynumerical forecasting models (the �Barotropic�model). Vorticity changes implied by PVA at 500mb produce convergence below and divergenceabove that level - hence, vertical motion.Panofsky (1964, p.114ff) also gives an excellentdescription of how to infer vertical motion fromvorticity concepts.

This simple physical picture is subject to manyrestrictions because so many assumptions are

Fig. 2.2. Schematic showing vorticity advection by the geostrophic wind (Vg). Solid lines are height

contours (z), dashed lines are contours of absolute vorticity (in units of 10-5 s-1). Where the height andvorticity contours intersect, they form quadrilaterals (with curved sides). The strength of the advection isproportional to the number of such quadrilatarals per unit area. Where vorticity and height contours areparallel, no advection is occurring. The hatched quadrilateral is in a region of negative vorticity advection(NVA) by V

g, since V

g is pointing from lower to higher vorticity. The stippled quadrilateral is in a region of

positive vorticity advection (PVA) by Vg.

15

involved. Although the winds are not usually toofar from geostrophic, it is often those cases oflarge ageostrophic departures which producesignificant weather [recall the geostrophic wind isessentially non-divergent!]. Also, occasionally,the vorticity pattern may move faster than thewinds, reversing the convergence/divergencepatterns associated with vorticity advection.Finally, it is not at all clear that 500 mb levelparcels conserve their vorticity, that divergence isthe only mechanism by which parcels changetheir vorticity, and that 500 mb is always near thelevel of nondivergence.

Nevertheless, in spite of all these potentialproblems, PVA patterns often prove useful. Thecareful analyst should be aware of those situationswhere PVA is less likely to tell the whole story. Anexcellent discussion of large scale vertical motioncan be found in Holton (1979, p. 136ff.). In thisdiscussion the role of PVA in producing verticalmotion is clarified. Specifically, there are two

sources for vertical motion in quasigeostrophicsystems. Rising motion is proportional to (a) therate of increase with height of PVA and (b) thestrength of warm thermal advection.1

Note that PVA must increase with height forupward vertical motion to result. This is anessential consequence of the law of masscontinuity we have described and is consistentwith the physical picture presented above. If thedivergence (related directly to PVA) does notincrease with height, then the air is not likely tobe rising, even if PVA exists at the standard 500mb level. Hales (1979b) has recently emphasizedthis important point.

The contribution of warm advection to upwardmotion is often neglected. The physicalsignificance of this effect can be described in avariety of ways. Consider the well-knownrelationship that the thickness of a layer (usuallybounded by pressure surfaces) is proportional tothe mean temperature in that layer. Thus, warmadvection is essentially related to thicknessadvection. A common situation wherein warmadvection plays a role is with a warm front. Thesoutherly flow, nearly perpendicular to thethickness contours, produces strong warm(thickness) advection, which tends to increase thethickness at a point. The vertical motion (upward)acts to cool the column by lifting and, therefore,tends to compensate for the warming. Upwardmotion induced by warm advection is oftenerroneously attributed to �overrunning�.

Most of the confusion about �overrunning� andthe effects of warm advection result from taking a2-dimensional, rather than a 3-dimensional view.Fig. 2.3 shows a cross section through a frontalzone, with potential temperature (theta) surfaces(isentropes). The actual winds are acting to pushthe theta surfaces from left to right, by advection.However, the vertical motion also acts to lift thosesurfaces, which displaces them opposite to thecontribution by advection. If the 3-dimensionalwind happens to be exactly parallel to the thetasurfaces, there is no horizontal movement, despitea horizontal wind component across the surfaces.In general, the flow is not exactly isentropic,usually giving a net horizontal displacement lessthan the normal component of the horizontalwind. This also explains why warm fronts tend tomove more slowly than cold fronts. It happens that

Fig. 2.3. Schematic illustration of how the 3-dimensional wind acts to displace isentropic (theta= constant) surfaces. The effect of the horizontalwind component (\V

H) is to push the theta surfaces

from left to right. Three different verticalcomponents are illustrated; \V1 is the typicalexample, which makes the theta surfaces rise,displacing them from right to left at any given level,such that the net displacement is less than thatindicated by the \VH contribution, but still from leftto right. In the second case, \V2 is such that the 3-dimensional flow is parallel to the theta surface,yielding no net displacement. For the third case,which is rare, the vertical component of \V

3 is so

large that the net displacement by vertical motion islarger than the \VH contribution, giving a netmovement from right to left.

16

analysis on isentropic surfaces is a good way tosee this on a 2-dimensional chart, subject to thelimitation that the actual flow may not be exactlyalong isentropes. Note that in some unusualcases, the contribution by vertical motion canexceed that by advection, so the front could�back up�, into the horizontal flow!

Recently, Trenberth (1978) and Hoskins et al.(1978) have pointed out that the PVA andthickness advection effects have a tendency tocancel each other. This can also be seen in thediscussion by Holton (1979, p. 139). Trenberth hasproposed a solution to this dilemma by using theadvection of vorticity by the thermal wind. Thosefamiliar with the pioneering work of Sutcliffe (e,g.,Sutcliffe, 1947 or Sutcliffe and Forsdyke, 1950)should recognize this approach. This requiresdoing the same thing that is currently done withPVA, but using thickness contours (to infer the

thermal wind) rather than height contours.Sangster (personal communication) has verifiedthe validity and value of this approach on a day-to-day basis. Sangster�s estimates of 850 and 700mb vertical motion are derived by using thevorticity and isotherms at each level. Theisotherms at each level ought to be fairly goodapproximations to thickness contours (for a layercontaining that level), so this is quite similar tovorticity advection using the thermal wind.

Since this revised method for locating areas ofupward motion includes both the PVA and thermaladvection terms, it has clear advantages. WithAFOS, overlaying the thickness and vorticityfields is relatively simple.

There are other ways to estimate the verticalmotion field, including the model output fields,which show forecast vertical motion directly.Since the model-generated vertical motion

Fig. 2.4. Vorticity advection by the thermal wind (VT). Thickness countours (T) are dashed lines, while solid

lines are contours of absolute vorticity (as in Fig. 2.2). Note that thickness contours and height contoursusually do not coincide, so that V

T differs from V

g.

17

patterns are not perfect, it is in the analyst/forecaster�s interest to have as many differentestimates of where upward motion is (or is goingto be) occurring as possible. This includesempirical rules as well as more objectivemethods, since no single approach appliesequally well under all conditions. Most severeconvection depends on larger-scale forcing todevelop (and/or maintain) its severity. It is worthnoting that this supportive large-scale upwardmotion may not always be obvious fromindications in mid-troposphere (about 500 mb).There is evidence (e.g., Hales, 1979b) that duringthe warm season, the �upper support� may onlybe detectable above 500 mb. Also, the forcingcan be confined to levels below 500 mb (Doswell,1977; Maddox and Doswell, 1982), as well.However, it should be recalled, that upperdivergence and lower convergence are mostfrequently related, as we have discussed. This isan essential element in the work of Uccellini andJohnson (1979), in which the coupling of upperand lower jet streaks is stressed, and which isdiscussed further in III.F.3.

2. Production of Unstable ThermodynamicStratification

Vertical motion, by itself, obviously is insufficientto develop severe thunderstorms or heavyconvective rain. In fact, large-scale verticalmotion produces large-scale regions ofcondensation. It can be argued that the major roleplayed by large-scale upward motion is toprepare the environment for convection. Onebasic property of convection is that it requires anunstable thermodynamic stratification.2 Therefore,a substantial effort in the interpretation of upper-air charts is directed also toward questions of theinstability of the �air mass�. Note that it is unusualfor severe storms to occur in a true air massregion, i.e., one with horizontally uniformproperties. Thus, it is somewhat misleading tospeak of the �unstable air mass� in which severeconvection develops. This is especially the casesince the vertical structure associated with severestorms (to be discussed later) usually revealsdifferent source regions for the air at differentlevels.

Fig. 2.5. Schematic illustration of differential advection (after Newton, 1980). Frontal symbols areconventional. Long-dashed lines are 500 mb isotherms while short-dashed lines are isotherms of low-levelparcels lifted to 500 mb (proportional to theta-w). Note that the eastward progression of the 500 mb thermaltrough and the northward progression of high theta-w air at low levels creates a condition of instability attime to### + dt### in the hatched region. That is, low-level parcels in the hatched region, when lifted to 500mb, are warmer than their environment.

18

The means by which the classic severe stormsounding develops is often the result of a processof differential advection. This process, describedby McNulty 1980, Whitney and Miller (1956), andAppleby (1954) among others, is simply the resultof vertical differences in the horizontal advectionof atmospheric properties (see Fig. 2.5). Ifdifferential advection acts to warm the lowerlayers relative to those above (or, equivalently, tocool the upper layers relative to those below), theresult is a net decrease in the stability of the aircolumn. Typical values in severe weathersoundings suggest that differential advection mayincrease the lapse rate of a sounding by as muchas 1 deg C km-1 every 3 hr (recall the dryadiabatic lapse rate is about 10 deg C km-1). Asevidence of the importance of instability changes,Newton (1980) has shown that an average parcelbuoyancy increase of 1 C over the depth of thetroposphere cap increase the cloud maximumvertical velocity by 7-12 m s-1.

Further, if differential advection results in a netmoistening of the lower layers, and/or a net dryingof the middle and upper troposphere, theconvective potential is also enhanced. In fact, anincrease in moisture content by 1 g kg-1 is aboutequivalent to increasing the temperature by 3.5deg C, if all the latent heat ran be released.McNulty (1980) has combined the influences oftemperature and moisture by considering thedifferential advection of wet-bulb potentialtemperature (theta-w)3 , since convectiveinstability is defined to exist when wet-bulbpotential temperature decreases with height.McNulty�s study was not directed beyond theshort-range correlation of differential advectionwith observed severe storms, so no clear-cutresults concerning the relationship were found.Nevertheless, over a period of days, differentialadvection must play a substantial role in creatingareas of unstable stratification. In most cases, themodification of stratification necessarily involvesthe process of differential advection. This does notimply that, once a basically unstable region hasdeveloped, differential advection is an ongoing,important process. McNulty�s conclusions supportthis view, since he suggested that during thespring, differential advection is not effective atseparating non-severe from severe storms, while itis more valuable at separating convective fromnon-convectiveregions. Because instability is

confined to relatively small regions during thespring, the additional destabilization fromconcurrent differential advection was notsignificant. In summer, the opposite conclusionwas drawn � i.e., differential advection isvaluable for delineating areas of severe weather,but not effective for locating convective regions.As NcNulty states (1980, p. 288), in summer,�Further destabilization is unnecessary forconvection and appears to contribute only tosevere convection development.�

When discussing differential advection, it isperhaps appropriate to digress briefly andexamine the concept of the thermal wind. If oneexamines the upper air charts, it is quite clear thatthe height contour pattern (and hence, thegeostrophic wind) generally varies with height atany given location. The difference between thepatterns at any two levels is simply the thicknessbetween the pressure surfaces. A relationshipknown as the Hypsometric Law can be stated asfollows: the thickness between any two pressuresurfaces is proportional to the mean virtualtemperature4 in that layer (see Table 1). Thus, thethickness contours can be regarded as isotherms,as we have already mentioned.

Since the contour patterns change with height, sothen does the geostrophic wind. By definition, thechange in the geostrophic wind with height is thethermal wind.5 Figure 2.6 shows how the thermalwind can be derived from the geostrophic windsat two pressure levels. Observe that we have twoquantities which are related to the change ofcontour patterns with height: the thickness and thethermal wind. It is logical to assume that thesequantities are related in some way to each otheras well. This is, in fact, the case. Specifically, thethermal wind blows parallel to the thicknesscontours (i.e., to the layer average isotherms),with speed proportional to the thickness gradient,and with low thickness (temperature) to its left.This is totally analogous to the geostrophic wind�srelationship to height contours (recall Fig. 2.4).

So how does all of this apply to the subject ofdifferential advection? Examine Fig. 2.6 andconsider the wind�s relationship to the thicknesscontours. If these contours are given theirinterpretation as isotherms, then it can be seenthat there is a component of the winds in the layer

19

Table 1. Factors, which when multiplied by the mean virtual temperature (Tv) in a layer for which thebounding pressures have the given ratio, yields the thickness (in m) of that layer. Thus, for example if Tv =0 deg C = 273.16 deg K in the 850-700 mb layer, then the thicknees dZ### is simply 5.687 x 273.16 =1553.5 m.

Pressure RatioPbot/Ptop Examples Factor

2.0 (1000/500, 500/250, etc) 20.3021.70 (850/500) 15.5421.6667 (500/300, 250/150, etc) 14.9621.50 (300/200, 150/100, etc) 11.8761.4286 (1000/700) 10.4471.40 (700/500) 9.8551.3333 (400/300, 200/150, etc) 8.4261.25 (500/400, 250/200, etc) 6.5361.2143 (850/700) 5.6871.20 (300/200) 5.3401.1765 (1000/850) 4.760

which is blowing across the isotherms. Thiscomponent, curiously enough, is the same ateither level! That is, the normal component (to theisotherms can be determined from the geostrophicwind at either level. The implication is that achange in geostrophic wind direction with heightis always associated with thermal advection. Asshown in Fig.2.6, when the geostrophic windbacks with height (turns counterclockwise , coldadvection is implied, while veering of thegeostrophic wind with height indicates warmadvection. Perhaps now the reason for thisdigression is clear. Thermal advection canactually be diagnosed simply by examining thechange in the contour pattern with height, even ifisotherms are not available. In fact, subject to thelimitation that the real wind may differ markedlyfrom geostrophic (especially at low levels), onecan infer temperature advection merely byknowing the profile of winds aloft, over a singlestation! See also Oliver and Oliver (1945) formore details.

Given the geostrophic thermal advectioncontribution through several layers (or at severallevels), one might be tempted to conclude thatone could diagnose the differential thermaladvection. After all, advection at any level isdominated by the geostrophic contribution.Unfortunately, this does not work. One cannotinfer destabilization when the 850 mb geostrophic

Fig. 2.6. Schematic illustration showingtemperature advection as implied by the change ingeostrophic wind with height. Geostrophic winds atthe lower and upper levels are denoted by VL andV

U, respectively, while the resultant thermal wind is

given by VT. Implied thickness contours are shownby dashed lines. The component of either V

L or V

U

normal to the thickness lines is VN. In the casewhere the geostrophic wind veers with height (turnsclockwise in the Northern Hemisphere, warmadvection is implied. Conversely, winds backingwith height imply cold advection.

20

thermal advective change is positive and thecorresponding 500 mb term is negative. In orderto see why this is so, consider how thegeostrophic advection changes with height. Sincethe geostrophic wind change is simply the thermalwind, which is parallel to the layer meantemperatures, the differential advection by thegeostrophic wind must essentially vanish.Therefore, differential advection must beaccomplished by the ageostrophic wind. Whilethe ageostrophic wind may not be the largest partof the wind itself, it has two very important roles --it supplies the significant divergent part of thewind field and it provides the means to changethe stratification via differential advection. Anexcellent discussion of how changes in stabilityoccur can be found in Panofsky�s (1964, p. 105ff)textbook.

As a final observation on differential advection, ithas often been suggested that cold air advectionat, say, 500 mb is an important contributor tosevere weather potential. Observations do notsupport this on a day-to-day basis. While casestudies certainly exist (e.g., Barnes, 1978) whichshow that cooling aloft as a result of coldadvection did play a role, it is more frequentlyfound that the environmental soundings show onlyweak thermal advection at 500 mb (either warmor cold) in the threat area. This is especially trueduring the late spring and summer (Hales, 1982).

A major contributor to the development ofinstability is the large-scale vertical motion itself.Regions undergoing large-scale lifting mustnecessarily approach an adiabatic lapse rate. Thedemonstration of this is available in any textbook(e.g., Hess, 1959, p. 102). By means of lifting,even an initially stable environment can becomefavorable for convection.

Since the classic pre-severe storm sounding(discussed in Miller, 1972) often has an inversioncapping the moist layer, the lifting process may beessential for development of storms even whenthe atmosphere is already convectively unstable.For the typical storm environmental sounding,about 6 h of synoptic-scale lift (at about 5 cm s-1)is capable of eliminating the cap (i.e., about 1 kmof net vertical lift).

Note that the negative area for the soundingassociated with the cap can be interpreted in an

interesting way. This area happens to beproportional to the square of vertical motion(Petterssen, 1956b, p. 136)! That is, for negativeareas, an upward vertical motion equal to thesquare root of twice the area is needed to cancelthat amount of negative buoyancy.

When the capping inversion is too strong to bebroken by the available sources of lift, noconvection may occur even under conditions ofextreme instability above the inversion. Thus, acoupling between �dynamics� and�thermodynamics� frequently must be present forsevere storms. The capping inversion acts toenhance severe potential by confining moisture tolow levels (Williams, 1960; Carlson and Ludlam,1968) until it can be released. Although daytimeheating from below may sometimes be sufficientto eliminate the inversion, the unmistakablerelationship between severe thunderstorms andsome source of upward motion (fronts, short-wavetroughs, etc.) suggests that in most cases, theinversion is eliminated by lifting.

It should be pointed out that layer lifting in thetraditional sense described by Hess occurs onlyfor layers of small thickness. That is, the processHess describes involves lifting the top and bottomof a layer by an equal amount. While thiscertainly has the effect described, one shouldremember that vertical velocity normally has itslargest magnitude in middle levels (say, around500 mb). Thus, layers of significant thickness willundergo stretching (or compression) whichamplifies any changes in stability. Further, it isworth emphasizing that the more stable the layeris to begin with, the greater is the change in itsstability as a result of lifting and stretching. Ineffect, the large-scale lifting process tends to drivelapse rates toward the dry adiabatic value. Alayer which is already stratified nearly dryadiabatically will not undergo much change,whereas a very stable layer is altered rapidly bythe lifting and stretching mechanisms.

Clearly, the source of vertical motion can be ondifferent scales in different situations. A case likethat of April 3-4, 1974 (Hoxit and Chappell, 1975)may be driven by large-scale lifting process.Doswell (1977) has shown that at timessubsynoptic scale lifting may provide the meansfor breaking the inversion. Beebe (1958) haspresented serial soundings where the inversion

21

clearly rises and the moist layer deepens in amesoscale area. As the scale of a translatingvertical motion source decreases, the requiredaverage upward speed must increase, since it hascorrespondingly less time to act. That is, the timescale generally decreases with size scale.Mesoscale systems can develop vertical motionsin the range of several m s-1, but their life cyclescan be completed in 6 h. Naturally, such detailcan be unavailable to the operational forecaster,but it is clear that the existing instability (say, at1200 GMT) in a region may not reflect accuratelywhat the sounding will look like at the time ofconvection. The Lifted Index6 (Galway, 1956)represents an adjustment of the sounding�sstability parameter to account for diurnal heating.The analyst/forecaster needs to provide furtheradjustments based on the upper level charts, usingthe concept of differential advection and thepossible effect of vertical motion.

3. Some Kinematic Considerations

As discussed before, many empirical rules forinterpretation of upper level winds are indirectefforts to diagnose and forecast vertical motion.

McNulty (1978) and Kloth and Davies-Jones(1980) have evaluated several of these ideas, asrelated to jet maxima. Hales (1979b) hasconsidered the use of anticyclonic (horizontal)shear in this context. It is pretty clear thatmesoscale features exist aloft, even ifconventional rawinsonde data are generallyinsufficient to reveal them. This insufficiency isrelated to the data density, to errors in the data(which tend to increase with height), to roundingwinds to 5 deg direction intervals, and to theanalyst�s bias toward recognition of winddirection changes more readily than actual vectorwind changes. The smaller scale details of thewind field can be inferred to some extent from thesatellite images, especially when animated loopsare available. The basic principle involved is thatwhere there is cloud, there is upward verticalmotion, and where there is vertical motion there issome �feature� which is forcing it (see Doswell,1982a).

Perhaps the most successful application of thisprinciple is the location of the jet stream axis(Whitney et al, 1966; Whitney, 1977). However,the often very sharp cloud edge near the jetstream axis (e.g., Fig. 2.7) may not be the result of

Fig. 2.7. Visible (a) and enhanced infrared (b, next page) satellite images showing anticyclonically curvedband of cirrus across Texas, Oklahoma, southeastern Kansas and Missouri. Such bands are associatedwrith upper level jet streams, with the jet axis from 1 deg to 5 deg poleward of the sharp cloud edge.

a b

22

the deep vertical motions associated with jetstream secondary circulations (such as thosedescribed by Cahir, 1971). Rather, the �edgemechanism� appears to be an interface betweenshallow vertical circulations, basically confined tocirrus levels (Weldon, 1975). Details of thismechanism remain unclear.

Unfortunately, conventional data do not alwaysrelate well to cloud masses observed in satelliteimagery. An interesting phenomenon whichreveals the type of problems inherent in satelliteinterpretation is the large mesoscale convectivecomplex (MCC) described by Maddox (1980b)(e.g., Fig. 2.8). As the MCC grows to maturity ithas an increasingly obvious influence on therawinsonde-sensed observations. Thedevelopment of a diverted flow around thenorthern side of an MCC creates the illusion of a�short-wave trough� or �vort max� upstream,which may have no previous or subsequenthistory. It is an �effect� rather than a �cause�,since it has a convective origin. In order todiscriminate valid mesoscale features in the largerscale fields, the satellite imagery should, ifpossible, be supplemented with corroborativeconventional data.

C. Sounding Analysis andInterpretation

1. General Remarks

Part of the early morning upper-air analysis shouldinclude an examination of plotted soundings. Thissubject has also suffered from declining interest,along with other aspects of synoptic meteorology.It seems obvious that considerable usefulinformation is available in the soundings. Anabundance of literature (Showalter, 1953;Fawbush and Miller, 1954b; Galway, 1956;House, 1958; Prosser and Foster, 1966; Miller,1972; Doswell and Lemon, 1979) exists whichstresses the detailed vertical structure, boththermodynamic and kinematic, of theenvironment in which convection develops. Anautomated sounding analysis, such as thatproduced at SELS (Doswell et al., 1982), can helpto decide which soundings to examine. However,such parameters as moisture depth, inversionstrength, and wind directional variation aredifficult to automate and can be helpful indeveloping a clear picture of the synopticsituation. Nothing can or should replace anexamination of the individual soundings. Such anexamination can also help to evaluate and correctany erroneous data that may have crept into theconstant level analyses.

Newton (1980) has presented the three types ofsoundings associated with thunderstorms (Fig.2.9). Newton�s Type A corresponds to Miller�s(1972) Type IV tornado air mass, which isgenerally characteristic of High Plains severeweather situations. Newton�s Type B is Miller�sType I tornado air mass, which is the classical�loaded gun� sounding of the Great Plains.Finally, Newton�s Type C corresponds to Miller�sType II tornado air mass, typically identified withthe Gulf Coastal regions of the southeasternUnited States. Newton does not explicitlydescribe Miller�s Type III sounding and itssimilarity to the Type II profile (except for lowertemperatures) suggests that it is a subset of theType II (or Newton�s Type C) situation.

Fig. 2.8. Enhanced infrared satellite imagerevealing Mesoscale Convective Complex (MCC)over Illinois and Indiana.

23

2. Sounding Thermodynamics

First consideration of sounding analysis is anassessment, usually via a single parameter, of thestability of the thermodynamic stratification. Thismight be the Showalter Stability Index (Showalter,1953), the Lifted Index (Galway, 1956), or theTotals Indices (Miller, 1972). These parameterskey on the amount of buoyancy available to alifted parcel at 500 mb, and have been in use for aconsiderable time. The SWEAT index developedby Miller (1972) attempts to incorporate somekinematic properties, specifically the shearbetween 850 and 500 mb. The need for and valueof these parameters are well-known and arestraightforward.

There are other factors which can be evaluatedfrom the soundings, some of which are not soeasily automated. One important example is thedepth of the moist layer. While some soundingstypify the classical �loaded gun� severe weathersounding (Fawbush and Miller, 1954b) in that theyhave a well-defined, inversion-capped moistlayer, surmounted by a substantially drier layerwith a steep lapse rate, this is not always thecase. The depth of the moisture has a large impacton the subsequent events. If the moisture is tooshallow (say, less than 50 mb deep) there may beinsufficient water vapor to support severeconvection. If the moist layer is exceptionallydeep (say, 200 mb or more), the likelihood of non-

severe heavy rainstorms is greater. Further, asdescribed in Schaefer (1974a), moist layer depthhas a dramatic influence on dryline motion (seeIII.B.5).

The occasional occurrence of a very deep layerof essentially saturated conditions to, say, above500 mb can result from convection contaminationand, hence, be unrepresentative. However, it alsocan indicate some severe potential, especially inthe southern part of the United States (Newton�stype C). In many such cases, the occurrence ofdry air aloft upstream from the threat area iscommon (Miller, 1972). This dry air typically hasarisen from subsidence (but may have otherorigins � e.g., Carlson and Ludlam, 1968) andthus is also relatively warm, so a dry intrusion isfrequently also indicative of warm advection (seeIII.E). The complete absence of dry air generallyimplies an increase in heavy rain potential and acorresponding decrease in the likelihood ofsevere thunderstorms.

However, moisture aloft in the absence of low-level moisture does not preclude severe weather,since high-based severe storms in such a situationare not uncommon, especially in the High Plainsregion of the United States (Newton�s type A). Asimple argument can show that such storms havea high potential for strong surface wind gusts.Further, when a shallow, surface-based moistlayer is found in such soundings, tornadoes canresult from High Plains thunderstorms (Doswell,

Fig. 2.9. Schematic of three distinctive soundings associated with severe convection in the U.S. (afterNewton, 1980). Type A is an “inverted V” sounding typical of the High Plains or desert severe storms. TypeB is the classical. “loaded gun” sounding characteristic of Great Plains or central U.S. severe weatheroutbreaks. Type C is common over Gulf coastal states and in summer east of the Mississippi River. Dash-dotted lines correspond to moist adiabats associated with low-levels and dashed lines to moist adiabats inlower middle levels (600-700 mb).

24

1980, Mahrt, 1977) since a few storms may beable to tap this low-level moisture by developingupdrafts with surface roots.

The reader should have realized by this time thatthe existence of dry air, generally in the mid-troposphere, is an important factor in much severeconvection. This has long been recognized(Ludlam, 1963). It appears that the enhancementof the downdraft potential, created by evaporationof cloud and precipitation into dry environmentalair, plays a key role in developing the stormstructures associated with severe weather (Lemonand Doswell, 1979),

Although it is not easily evaluated from a simpleplotted sounding, the vertical profile of wet-bulbpotential temperature (theta-w) is worth someexamination, Since theta-w incorporates bothtemperature and moisture, its vertical distributionprovides key clues about convective instability. Infact, by definition, if theta-w decreases withheight in a layer, that layer is convectivelyunstable. As noted previously, the �loaded gun�sounding is the archetypical example ofconvective instability, since its moisture andtemperature profiles combine to produce aminimum in theta-w in middle levels.

It has been argued that the difference between thetheta-w minimum at mid-levels and the theta-wmaximum at low levels (often at the surface)represents the total energy available to a severestorm (Darkow, 1968; Morgan and Beebe, 1971).This concept has been tested by Doswell andLemon (1979). They found that, for a sample ofsevere thunderstorm environmental soundingsbefore and then near severe storm occurrence, aparameter based on this difference did not seemtoo effective at delineating the region of mostsevere convection. However, they note thatduring the time from the sounding well before thestorm to the sounding closest to storm occurrence,the minimum theta-w value actually increasesslightly (about 1.5 deg C) and the height of theminimum rises (by about 80 mb). This can beinterpreted as a reflection of the action of upwardvertical motion. That is, the moist layer deepensand rises during the period before storms (Beebe,1958).

Another factor that should be evaluated fromselected soundings is the so-called negative area

in the lower part of the parcel�s ascent profile. Ifthe parcel is negatively buoyant, energy must besupplied to lift the parcel through those layers. Assuggested earlier, negative area can act toenhance severe potential by capping the releaseof energy until the optimum time (usually near thetime of maximum surface heating). The LiftedIndex can account for the contribution of surfaceheating to �cap� erosion by using a forecastmaximum surface temperature. When substantialnegative area remains after accounting for diurnalheating (if applicable -- at 0000 GMT, surfacecooling will occur), the forecaster/analyst shouldtry to determine whether there is a source ofsufficient lift (e.g., a source of low-levelconvergence or some feature supplying upwardmotion) to eliminate the cap (recall II.B.2, above).

Also valuable in operational study of soundings isthe determination of the equilibrium level for therising air parcels. The equilibrium level (EL) iswhere the rising, buoyant parcel re-crosses theenvironmental sounding curve. It is this level,rather than the tropopause, that is physicallysignificant. Anvil cloud material tends toaccumulate here, rather than at the tropopause,since it is where rising parcels are (naturally) inequilibrium with their environment. Penetrationsof the EL are indicative of strong updrafts, and theEL can be well below, near, or well above thetropopause. Naturally, depending on thecharacteristics of the tropopause, a storm whichreaches above the tropopause is usuallysignificant. However, when the EL is far belowthe tropopause, storms with tops which remainbelow the tropopause can still be severe (Burgessand Davies-Jones, 1979). Similarly, a storm whichpenetrates the tropopause may still be below theEL.

Positive area should also be evaluated. If a smallcalculator is available, the positive area can bedetermined from the hypsometric equation as thedifference in thickness between the observedheights and the heights using the parcel ascentcurve (between the LFC and the EL). As discussedearlier, this can be used to determine the parceltheory vertical motion associated with the amountof positive area. Such a vertical motion speed isgenerally an overestimate (see II.II.A), but isrepresentative of peak updraft speeds in the mostsevere storms.

25

In the past, some attempts have been made toforecast the maximum gust potential and/or themaximum hail size possible with a given sounding(Foster and Bates, 1956; Foster, 1958; Fawbushand Miller, 1954a; and Fawbush and Miller, 1953).Doswell et al., (1982) suggest that the automatedestimates previously used in SELS (Prosser andFoster, 1966) do not have much skill in predictionof observed gust speeds and hail sizes. Shown inTable 2 are the average gust speed errors fromthat study, based on 1978 severe storm reports. Ofsignificance is the fact that well over half of thereported gusts occurred when the predicted valuefrom the nearest rawinsonde was for no gusts. Byexcluding the �no gust� forecasts, it can be seenthat reported gusts under 65 knots were actuallyoverforecast while those 65 knots or greater wereconsistently underforecast.

In Table 3, the same sort of calculation is shownfor predicted maximum hailstone size.Underforecasting is the general rule, even whenexcluding the (roughly) one-fourth of reportedevents which occurred with a �no hail� forecast.Based on these statistics, it seems clear thatrelatively little skill is apparent.

Sophisticated cloud models, using soundings asinput, might be able to provide better quantitativeestimates (Chisholm, 1973), but they are notcurrently practical for operational use. Further,there are simply too many important factors in

Table 2. Errors in predicted gust speeds: 01 Apr — 30 Jun 1978, for reported gusts 50 kt or greater in thetwo categories shown. Values are the mean difference (in knots) of (observed-predicted) +### theassociated standard deviation. The number of cases is in parentheses. Values under “Preceding” refer topredictions from the sounding preceding the report by more than 6 hr; whereas “Next” refers to predictionsfrom the sounding immediately following. Values are given for “All” predictions and also excluding caseswhen the prediction is “no gusts”.

TABLE 2

Less than 65 kt 65 kt or Greater Preceding Preceding Next

All: 30.5 ± 30.0 (167) 41.3 ± 29.6 (45) 48.1 ± 30.9 (43)Excl 0�s: -5.7 ± 6.1 (62) 12.9 ± 8.8 (22) 13.2 ± 10.5 (19)

105 23 24

105/167 = 62.9% observed gusts 50 =< V < 65 kt with 0 calculated (Preceding)23/45 = 51.1% observed gusts V >= 65 kt with 0 calculated (Preceding)24/43 = 55.8% observed gusts V >= 65 kt with 0 calculated (Next)

producing hail and surface wind gusts that arepoorly understood, much less routinely observed.

An example of a plotted sounding is shown in Fig.2.10 with relevant features labelled. Any textbook(e.g., Hess, 1959) provides enough understandingto plot and analyze the typical sounding. Thereare several aspects of this sounding worth noting.First, it is interesting to observe that the moisturecuts off just below 850 mb, so that the ShowalterIndex is unrepresentative of the sounding�sconvective instability � the Showalter Index hasa value of +0.2 deg C, whereas the Lifted Index(based on a forecast surface temperature of 100 F)is -6.3 C! Further it can be seen that even at asurface temperature of 100 F, a substantialnegative area remains to be overcome before theconvective instability can be released. Also, notethe large positive area in this sounding. The sizeof the positive area may be a more relevantparameter than 500 mb buoyancy, since updraftspeed is only crudely related to the accelerationat any (arbitrary) single level. As discussed inII.III.A.3, the updraft speed is probably a goodmeasure of storm severity.

This sounding also shows an equilibrium levelsomewhat above the tropopause. Thus, a stormwhich slightly overshoots the tropopause in thisenvironment is not necessarily severe. In fact, thesounding suggests that a storm which realizes

26

most of the energy implied by the positive areashown will, in all likelihood, penetrate above 100mb!

3. Sounding Kinematics

Perhaps one of the most widely accepted ideasabout the severe thunderstorm environment is thatvertical shear is a prime ingredient without whichstorms are unlikely to develop severecharacteristics. This idea is presented quitesuccinctly by Ludlam (1963): �When there is littleor no wind shear the updraft is upright and theprecipitation falls through and impedes it.�

The common presence of substantial shear insevere storm environments actually has presenteda minor paradox. Certain theory suggests thatshear can be detrimental to convection, whileobservations indicate, in general terms, theopposite tendency. This has been reconciled bythe suggestion that while shear certainly inhibitsweak convection, a sufficiently strong updraft canovercome this tendency and can, in fact, beenhanced by it, The details of this apparentcooperation between shear and updraft havenever been completely understood although avariety of mechanisms have been proposed (e.g.,Newton and Newton, 1959; Alberty, 1969;

Table 3. Errors in predicted hailstone size: 02 Apr — 30 Jun 1978, for reported hailstones 3/4 inch orgreater, in the two categories shown. Values are the mean difference (in inches) of (observed-predicted) ±the associated standard deviation. The number of cases is in parentheses. See Table 2 for explanation of“Preceding” and “Next”. Values are given for “All” predictions and also excluding cases when theprediction is “no hail.

TABLE 3

Less than 2 inches 2 inches or Greater Preceding Preceding Next

All: 0.95 ± 0.64 (449) 2.21 ± 0.85 (70) 2.08 ± 0.85 (71)Excl 0�s: 0.76 ± 0.62 (317) 1.94 ± 0.73 (51) 1.97 ± 0.93 (54)

132 19 17

132/449 = 29.4% observed hail 3/4 <= d < 2 inches with 0.0 calculated (Preceding)19/70 = 27.1% observed hail d >= 2 inches with 0.0 calculated (Preceding)17/71 = 23.9% observed hail d >= 2 inches with 0.0 calculated (Next)

Charba and Sasaki, 1971; Rotunno and Klemp,1982). Model results (Weisman and Klemp, 1982)suggest that supercell storms arise only for arestricted range of shear, with any given amountof instability. The operational utility of theseresults remains unproven, but they are indicativeof the delicate balance required to producesevere thunderstorms. Such details may be belowthe resolution of operation data sets.

Many believe that the essential feature whichallows this interaction to benefit convection is theestablishment of a tilted updraft (Fig. 2.11). By thismeans, precipitation formed within the updraftcan fall out of the updraft, rather than having topass through it. Once again, a potential problemarises since the updraft must somehow tiltupshear, or else the precipitation falls into air thatthe storm is ingesting, which cools (and therebydecreases the buoyancy) the inflow. This isgenerally resolved by noting the common severestorm hodograph shows not only shearing, butveering with height.8 Since parcels rising in theupdraft should tend to conserve their horizontalmomentum, this is one means of developing theappropriate tilt. Ludlam (1963) also shows (hisFigure 21) how upshear updraft tilt can arise fromthe combined effects of storm motion and thefinite time required by successive updraft parcelsto rise to a given level.

27

Fig. 2.10. Sample plotted sounding (Oklahoma City, Oklahoma, 1200 GMT, 8 June 1974) on a skew T-logp thermodynamic diagram. Heavy solid line is the temperature (T) observation during balloon ascent,heavy dashed line is for dewpaint (Td). The thin solid line labelled q=39 (deg C) is the dry adiabat throughthe forecast maximum surface temperature, while the thin dashed line labelled w=16 g kg-1 is the meanmixing ratio in the lowest 100 mb. These intersect at the Convective Condensation Level (CCL).7 From theCCL, the parcel, ascends along the pseudo-adiabat labelled qw=25.8 (deg C), denoted by a dash-dot line.Pseudo-adiabats labelled qw=24 and qw=28 are the thin solid lines included for reference. The ascendingparcel is initially negatively buoyant and the negative area is depicted by hatching. As the parcel risesthrough the Level of Free Convection (LFC) it becomes positively buoyant, with the positive area stippled.After the parcel rises through the tropopause (about 250 mb) it once again crosses the environmentalascent curve, at the Equilibrium Level (EL).

28

There are some problems with this picture of howthe storm structure arises from its environment.These are also discussed in II.III.A.5.Documentation of many supercell storms suggeststhat their updrafts are essentially vertical throughgreat depths in the storm. Further, McNulty (1978)has found that �crossover� (the veering of windswith height) is not an essential feature in allsevere weather situations. Rather, he found thatsevere thunderstorms can occur in environmentswith relatively little directional shear. Crossovermay be more important in the process ofdifferential advection than as a means of locatingsevere convection directly. Finally, Doswell andLemon (1979) have found that supercell storms(see II.III.A.5.b) can be found in environmentswhich have a rather wide range of cloud-bearinglayer shear values.

Marwitz (1972) has stressed that the subcloudlayer veering of wind with height is important indifferentiating supercell environments from thoseinvolving multicellular storms. Doswell andLemon have supported this concept by suggestingthat subcloud layer average shear is bettercorrelated with their supercell sample. Thisconcept is further supported by recentthunderstorm model studies (Weisman and Klemp,1982) which reveal that hodographs having aveering subcloud wind profile, especially within

the lowest 3 km (as well as an appropriatethermal instability structure) seem to be mostsuccessful for model simulations of supercellthunderstorms. Such a hodograph is shown in Fig.2.12.

It is not yet entirely clear that severethunderstorms are dependent on a given type ofhodograph, at least in terms of their large scalesetting, as sensed by operational rawinsondes. Asdeveloped in II.II.A.3 and II.III.A.4, it seems morelikely that the relative flows are more physicallysignificant in producing a particular stormstructure. If this is the case, then the explanationof the climatological fact that severethunderstorms occur most frequently in shearedenvironments becomes more clear. It is in thesheared environment that a given convectiveelement can develop the appropriate relative flowmost easily. Shear, per se, may not be a necessarycondition.

This situation creates a forecast problem, since amajor element in knowing the relative flowstructure is the storm motion, which is notgenerally known, a priori. From a forecastviewpoint it is probably best to examinehodographs with respect to shear structure first.However, one should also be aware of features inthe environment which could result in stormmotions favorable for developing appropriate

Fig. 2.11. Schematic cross section though squall line thunderstorm of 21 May 1961 (after Newton, 1966).Note the upshear tilt of the updraft. Although cumuliform clouds are shown in the downdraft region to therear of the storm, there should not be cloud material in the downdraft region; rather, precipitation isdeecending and becoming less intense as one moves away from the updraft. Vertical scale is exaggeratedfor clarity.

29

relative flow. For example, Weaver (1979) haspointed out that a persistent stationary source oflow-level convergence can dominate the effectsof advection, yielding a storm which hasrelatively little movement. A very slow-movingstorm, or one which moves substantiallydifferently from the flow in which it is embedded,can have strong relative winds even when theflow is not highly sheared and/or with only modestwind speeds.

There are certainly features in the hodographwhich can be useful in identifying environmentswhich favor, or oppose development of severeconvection (e.g., see Darkow and McCann,1977). As Maddox et al. (1979) describe, thetypical non-severe convective rainstorm exists inan unstably stratified environment with very weakwinds through a great depth. This agrees withconventional ideas about shear. Unless storms ina weakly-sheared environment move in ways

much different than the mean flow, they cannotdevelop strong relative flows.

At times, strong shear at very high levels canoverlie regions of weak shear. This is not anenvironment conducive to severe convection,although the average shear in the cloud-bearinglayer might be fairly high. The tops of convectionmay well be literally sheared off and persistentupdrafts are unlikely. This is supportedtheoretically by Schaefer and Livingston (1982),since they find that moderate �shear of the shear�is favorable, whereas excessive changes in shearare detrimental to convection.

The analyst should certainly be prepared toexamine the hodographs in situations withrelatively weak winds. If the winds are onlymoderate in speed, directional shear can result inlarge relative windspeeds, especially since thestorms may not move very quickly under thoseconditions (e.g., Doswell, 1977). Further, in

Fig. 2.12. Mean proximity hodograph for 62 cases of severe thunderstorms under southwesterly flow aloft(after Maddox, 1976).

30

moderate upper-level winds, the low-level jetmay assume greater importance. It is notuncommon for the strongest wind in the soundingbelow, say, 300 mb to be associated with a low-level jet. This sort of structure can, and does,produce very severe storms even though thecloud-bearing layer shear is only moderate. Thisis consistent with the enhanced significance of thesubcloud layer shear (see the discussion ofnumerical model results in II.III.D), and with thephysical significance of relative winds.

Note that the analyst can evaluate numericalshear values quickly and easily. AFOS can beprogrammed to calculate shear over variouslayers, at the discretion of the analyst. This canalso be done with a programmable handcalculator. Plotting the hodograph can allow thegraphical calculation of shear, as shown in Fig.2.13. Doswell and Lemon (1979) havesummarized their findings on shear values,showing that the average value for supercellstorms is about 3 x 10-3 s-1 in the cloud-bearinglayer, and confirming the values of Marwitz(1972) for supercell storms. For subcloud shears,Doswell and Lemon find an average value of 7 x10-3 s-1 . As discussed, the analyst should be

aware of the broad range of values which can befound in supercell environments.

D. The Composite Chart

The final product of the morning analyses shouldbe the so-called Composite Chart. Guidelines forthis have been developed (e.g., Miller, 1972) andthere is some value in having a consistentlystructured composite chart. However, the relevantparameters on any given day may not be usefulon some other day. Owing to their obviouslycentral role in severe convection, moisture,vertical motion, and instability parameters shouldalways be included on the composite analyses.Choices for additional parameters should be madeby the analyst/forecaster, depending on thesituation. For example, instability at analysis timemay be rather weak but a low-level jet stream israpidly increasing moisture in the threat area. Onsome other occasion, instability may already besubstantial so the advection via the low-level jetis not important, whereas an upper level jet streakmay suggest a region of vertical motion necessaryto break a capping inversion.

Fig. 2.13. Schematic showing how the hodograph can be used to calculate the shear vector (the dashedvector) graphically between the surface and 850 mb. By translating this vector to the origin, one can seeits magnitude and direction. Then, knowing the distance between the top and bottom of the layer, one caneasily find the magnitudes of the shear.

31

Thus, the weather situation should, in part, dictatethe parameter choices. However, there exists anecessarily small set of standard features whichrepresent a reliable starting point. For example,this might include the basic features analyzed onthe surface chart (fronts, drylines, etc.), low-levelmoisture fields, static stability, and jet stream axesat high and low levels.

The prognosis should also influence parameterchoices. This can be derived largely fromnumerical guidance, since broad trends are thebasic controlling factors. However, one should notbe limited to mere reproduction of the modelforecasts. Since the analysis should be used, inpart, to assess the quality of numerical guidance,a forecast composite is somewhat dependent onthe analysis. The decision about what isnecessary in constructing the composite analysisand forecast is a feedback process betweencurrent and anticipated conditions.

The old adage about expecting the severeweather where the Composite Chart is mostillegible has some basis in fact, although only to alimited extent. The forecaster benefits most by theprocess of preparing the Composite Chart, not bylooking at the end result. The necessity fordetermining the spatial relationships amongfeatures (at the various levels in the vertical)ought to be self-evident. Preparation of theComposite Chart accomplishes this. A by-productis that the Composite Chart can be used as aquick reference throughout the day�s activitiesand for briefing the next shift. A mesoanalystworking without a Composite Chart is severelyhandicapped, and this step should be routineduring severe weather analysis.

CHAPTER II FOOTNOTES

1 II.B.1: It is commonly stated that vertical motionis related to the Laplacian of the thermaladvection. This is not true! In effect, thequasigeostrophic equation for vertical motion saysthat the Laplacian of the vertical motion isproportional to the Laplacian of thermal advection� the Laplacian operators effectively �cancel�,giving the relationship as stated by Holton (1979,p. 136ff).

2 II.B.2: There are numerous definitions by whichwe express the stability of the stratification (seee.g., Hess, 1959, p. 95ff). The two of mostimportance here are conditional and convectivestability. Conditional instability simply refers to thelapse rate of temperature. If the environmentaltemperature decreases at a rate between dry andmoist adiabatic, the stratification is said to beconditionally unstable. The �condition� inquestion is whether or not rising parcels canreach condensation. For dry ascent, such anenvironment is stable, since the dry parcel coolsmore rapidly than the environment. For moistascent, the opposite is true. Convective stability issomewhat more subtle. It accounts for the changeof moisture with height, as well as thetemperature lapse rate. In physical terms, if. thebottom of a layer being lifted reaches saturationfirst it will then cool at the moist adiabatic rate.Should the upper portion continue to rise withoutcondensation, it cools at the much larger dryadiabatic rate. Thus, the upper portion is coolingmore rapidly than the bottom. Therefore, withinthat layer the lapse rate is rapidly increasing.Such a situation results from environments wheremoisture content decreases rapidly with height.Clearly, there must be enough low-level moistureto allow moist ascent � if the air dries out tooquickly off the surface, rising parcels will mix intoo much dry air to maintain moist ascent.

3 II.B.2: These notes make extensive referencesto wet-bulb potential temperature (theta-w). Onecan easily argue that equivalent potentialtemperature (theta-e) is easier to calculate andequally useful. No attempt will be made torationalize the choice of theta-w over theta-e � itsimply depends on what one is accustomed to.They are essentially the same � there is a (non-linear) one-to-one correspondence (see Hess,1959, p. 103). However, theta-w is no longermore difficult to compute, if one has a computerand the algorithms given in Doswell et al. (1982).

4 II.B.2: The virtual temperature (Tv) accounts for

changes in density as a result of water vaporbeing present. As discussed by Saucier (1955), Tvis always greater than the actual temperature. Thedifference (Tv - T), therefore, is always positive. Ifthe moisture content is given by the mixing ratio(r), then to a good approximation, Tv = T + (r/6),where r is given in g kg-1. Since mixing ratios

32

above 500 mb are normally smaller than 1 g kg-1,the difference is negligible at those upper levels.The correction can be substantial in lower levels.

5 II.B.2: It may be noted that since the windsabove, say, 850 mb are pretty close togeostrophic, the change in the real wind oftenmay be a good approximation to the thermalwind.

6 II.B.2: Note that all uses of the term �LiftedIndex� in these notes are based on the originalversion developed by Galway (1956). Otherdefinitions exist, some of which bear littleresemblance to the original � these are still usedfor facsimile (or AFOS) analysis and forecastcharts.

7 II.C.2: The definition of CCL used here differssomewhat from that of Huschke (1959, p. 134). Asused here, the CCL is the condensation level forthe well-mixed boundary layer (the lowest 100mb) at the time associated with the forecastmaximum surface temperature. Thus, it is theintersection of the boundary layer mean mixingratio line with the dry adiabat from the forecastsurface temperature.

8 II.C.3: Recall in the discussion of the thermalwind in II.B.2 above, we found that veering meanswarm advection. Also remember that warmadvection implies upward vertical motion! Thisline of reasoning must be used with caution, sinceveering with height is much more common thanupward vertical motion. The contributions tovertical motion from differential vorticityadvection and from non-quasigeostrophic effectscannot be disregarded, and may often be thedominant factors. Also, low-level flow is lesslikely to be in geostrophic balance, soquasigeostrophic arguments do not apply.

33

III. Surface Data Analysis

�corrections� to sea level. The surrounding terraincan cause the winds to favor a particulardirection. Although every effort is made to get�uniform� instrument exposure, it is simply notpossible. Furthermore, the general trend forobservation sites at airports creates specialproblems, since some airports are in urbanenvironments while others are more rural. It is theresponsibility of the analyst to determine andcorrect for as many of these local biases aspossible. This should not be confined to one�s ownstation, but should include all reporting siteswithin the confines of the analysis region.Generally, mesoanalysis will be confined to alimited area, usually encompassing 3-5 states,called a �sectional chart�. This is done to allowthe plotting and analysis at hourly intervals. Itwould be desirable to document these biases andhave the documentation available for trainingnew local analysts, as necessary.

The problem of occasional �errors� in the data issomewhat more difficult to deal with. Somepercentage (unknown) of these are actually validobservations of the variables, but areunrepresentative in some sense. A stationexperiencing a thunderstorm will frequently, aswe shall see, have surface conditions that appearanomalous in comparison with its neighbors. It isprecisely this sort of observation we wish to retainand use to construct an operational mesoanalysis.Not all such mesoscale features are possible toidentify with such ease. Further, even when thephenomenon itself is recognized, its quantitativeinfluence on the station data is not always clear.This is especially important for objective analysis(see IV.C). When doubts about the validity of anobservation arise, it is probably wise to err on theside of retention of the datum. One additional testthat may be valuable is to examine how theanalysis looks with the datum accepted � if theresulting pattern does not fit any recognizablemesoscale structure, it is likely to be erroneous.Clearly, in this context, it is to the analyst�sadvantage to be familiar with convectivemesoscale systems. Naturally, it is also beneficial

A. General Remarks

With the exception of remote sensing (radar andsatellite imagery), the only source of data suitablefor �mesoanalysis� is at the surface. In the centralUnited States, station separation averages about125 km, with routine data collection every hour.Under certain conditions, special observations areavailable between the hourly observations. Thestation density has become increasingly diurnallydependent, with a substantial number of stationsclosing at night. Also, not every station reports athourly intervals.

With these data, it is possible to accomplish whatis referred to loosely as mesoanalysis in thesenotes. Under most definitions of mesoscale, theseobservations are still not dense enough to performa truly mesoscale analysis. The author haspreviously chosen to refer to the scale definableby the routine observations as �subsynoptic� scale(Doswell, 1976). Others have referred to this as�meso-alpha� scale analysis (Orlanski, 1975;Maddox, 1979a). Of course, the name isirrelevant, but the intent here is to emphasizeanalysis on the smallest scale allowed by the datadensity of routine surface observations. As weshall see, this process can be enhanced and theeffective scale limitations reduced somewhat byusing remotely sensed data.

Before plunging into the analysis itself, somethings need to be emphasized concerning thedata. It is an aphorism that mesoanalysts neverdiscard any observations, making every datumpart of the analysis. Like most aphorisms, thiscontains a sizable element of truth. However,there are clearly limitations to this.

Anyone who has done surface analysis comeseventually to realize that some stations havepersistent biases in their observations. Assumingthat the instruments themselves are not faulty, themain cause of persistent biases is the uniquenature of the station location. Station elevationsinfluence the pressure observations, even after the

34

Fig. 3.1. Sectional plot of surface observations at 1800 GMT, 8 June 1974. Station model for plotted datagiven in Fig. 3.2.

Fig. 3.2. Station model for plottedsurface observations. Note that thesecond digit of the wind direction isindicated — in this example, a wind fzom210 deg has a “2” shown. Wind speedsare in kts, temperature in deg F, altimetersetting in hundredths of an inch ofmercury (leading digit suppressed — inthis example, 29.57 in Hg = 957) and 3-hpressure tendencies in tenths of a millibar(down 1.5 mb over past 3 h = -15).

35

to consider independent data sources such assatellite images to help determine a datum�svalidity. A final tool to use in this process iscontinuity in time � if a feature persists throughseveral observations, it is probably valid.

In the typical situation, surface observations comein as a �sequence� (or on AFOS, in batches). Inorder to do an analysis, these need to be plottedon a map. In the interest of saving time, a givenforecast office has only to plot a sectional map, anexample of which is given in Fig. 3.1. Data areplotted according to the station model shown inFig. 3.2. Using AFOS, such a chart can bemachine plotted rather quickly. It is unlikely that areally detailed �mesoanalysis� (recall that this isactually a subsynoptic scale analysis) can bedone hourly for an area much larger than thatshown. Even national centers like NSSFC mustfocus attention on a confined region in order toaccomplish a detailed hourly analysis.

Note that the altimeter setting is plotted,1 ratherthan the so-called �sea level pressure.� This isdone for several reasons (Magor, 1958) . Theprimary reason is that there is a 30 to 40 percentincrease in the number of data points. Almostevery reporting station transmits an altimetersetting, even with the off-hourly �special�observations. This, by itself, is a substantialadvantage. Further, the altimeter setting is relatedto the station pressure in an identical manner ateach observation point.

Also note that significant remarks are plotted.These can frequently provide important clues tothe analyst, and are a valuable complement to theremote sensors like radar and satellite cameras. Ifpossible under the time constraints, these additiveremarks should he included in the station plot,perhaps abbreviated (e.g., �PRR� for �PRESRR�).

B. Surface Discontinuities

Much emphasis has been put on discontinuities inthe surface fields, and rightly so. The identificationand prognosis of boundaries is generally a keyelement in producing a forecast. In terms of whatthe atmosphere does, not much �weather� goeson in regions of uniformly distributed weatherelements. Generally speaking, what we perceive

as weather is the result of atmospheric processesacting to relieve some form of non-uniformity.Weather continues until the non-uniformity hasbeen eased to the point where the process can nolonger be sustained. This statement applies to allscales of weather phenomena, from globalcirculation patterns down to microscale activity.

While these notes place a substantial emphasis onproper analysis of surface boundaries, the analyst/forecaster is cautioned not to place so much effortinto analysis of boundaries that the basic physicalreasoning process suffers. There are many factorsto consider, and one should avoid getting �hung-up� in a complex weather situation in trying tofine-tune the surface analysis. A surface analysisis, after all, a working chart, not a work of art.

Probably the most commonly recoqnized surfacediscontinuities are air-mass boundaries. This broadclass of boundaries includes true fronts, drylines,sea and land breeze fronts, and thunderstormoutflow boundaries. These notes do not dwell onthe details of each of these, and once again thereader is urged to examine the references. Rather,we wish to point out some features to look for inhelping to locate these boundaries properly on thesectional mesoanalysis. As in the case of upper airanalysis, the end result of surface analysis isgenerally to locate and forecast where upwardmotion is found to coincide with regions ofconvective instability. The lift associated withsurface convergence tends to be concentrated inthe vicinity of these boundaries.

First of all, consider the true front. By definition, afront is a zone where atmospheric density variessubstantially. Basically, the frontal zone separatestwo distinct air masses. The width of the frontalzone can vary greatly, and only approaches atrue discontinuity in an idealized, limiting case.Nevertheless, we choose to draw �the front� as aline on the weather chart. Typical large-scalesurface frontal gradients are in the range of 10deg K per 100 km. The question might arise as towhere in the frontal zone do we place thehypothetical line called �the front�? Theconvention is to place the front on the warm-airside of the frontal zone. If we consider theisotherm analysis of Fig. 3.3, derived from thedata shown in Fig. 3.1, we would then have apossible frontal structure as indicated. It should bepointed out that if 20 meteorologists started with

36

Fig. 3.3. Isotherms (deg F)and one possible frontalanalysis for data in Fig. 3.2.

Fig. 3.4. Isotherms of theta-w(deg C) for data in Fig. 3.1.

37

the data of Fig. 3.1, we would probably have 20different analyses of the fronts and other surfaceboundaries. This not necessarily bad. A carefulexamination of the data in Fig. 3.1 suggests thatthe frontal positions are quite clear in some areasand not so clear in others. This situation is mostcommon. Real surface data simply do not fit theidealized patterns which are usually chosen forpresentation in textbooks. This is not a fault intextbooks, since their aim is generally to increasethe reader�s basic physical understanding, not toteach one how to analyze real data.

It is also possible to define air-mass boundaries(not necessarily fronts!) quite effectively bycalculating and plotting the observations of wet-bulb potential temperature (theta-w). Since airmass density is influenced by pressure,temperature and moisture content, the theta-wfield is quite well-suited for this purpose. See Fig.3.4 for a theta-w analysis associated with the datafrom Fig. 3.1. Note that virtual temperature (recallII.B.2, above) is directly proportional to density, sothat a Tv analysis gives as clear a picture of frontallocations as is possible.

A well-recognized method for locating fronts is toposition the front in the zone of strong cyclonicwind shear which usually accompanies a changeof air mass. That such a wind shift is associatedwith frontal zones is a subject well-covered intextbooks (see, e.g., Hess, 1959; p. 230ff), and is anatural consequence of the strong densitygradients involved. Furthermore, such a wind shiftis implied also by the pressure trough in whichfronts tend to occur. Since a front is locatedgenerally in a pressure trough, pressuretendencies can be used to locate the frontalboundary (see III.D.1). These relationships are alltheoretically justifiable and show a remarkableagreement with the typical real weather situation.However, the analyst should certainly be awarethat all wind shift lines and pressure troughs arenot frontal in nature. Further, not all fronts haveeasily recognized cyclonic wind shear acrossthem, and occasionally liberties must be takenwith the pressure pattern to put the front in apressure trough.

This brings to mind the often-discussed issue ofwhether or not to �kink� the isobars across a front.At times, in actual practice, it is more clearlyjustified than others. Naturally, as we have

already mentioned, a front is only a transitionzone and not a true discontinuity. Therefore, itwould seem reasonable to suggest that the �kink�is only a theoretical artifact and not a reflection ofthe true pressure distribution. Nevertheless, thisissue is actually more an aesthetic or pedanticone. In short, it really doesn�t matter except topurists (on both sides of the issue). Rather, oneshould concentrate on the proper pressureanalysis all over the chart.

1. Cold Fronts

By definition, a cold front is a frontal boundaryalong which warm air is being replaced by cold(Fig. 3.5). In the northern hemisphere then, thebasic flow behind the front typically has anortherly component, since cold air generally hasits source region to the north. When the cold airbegins its southward movement away from itssource, the contrast between the cold air and theair it replaces is often large, and the front is quiteeasy to locate. By moving cold air to the south,the atmosphere is trying to equalize thetemperatures over the earth. As the cold, usuallydry, air moves southward over warmer ground, itis gradually heated from below. Since the warmair created by contact with the surface naturallytends to rise, this heating is quickly spread togreat depths. Thus, it takes a relatively long timeto note the surface modification of the air mass,since the heating is so quickly dispersedvertically. The opposite is true when the front ismoving over colder surfaces. Situations where acold front moves over colder ground are relativelyrare in the United States, except perhaps whenmaritime polar air from the west displaces

Fig. 3.5. Schematic cross section through cold font,showing cloud and precipitation (after Byers,1959).

38

modified arctic air masses. In such situations thecold front may be hard to find at the surface, sincea surface inversion is created which tends tomask the normal features associated with a front(see Saucier, 1955, p. 297).

While the basic large-scale flows imply surfaceconvergence along the frontal zone, care must beexercised in using this generalization. Theobvious picture of the cold front acting as awedge, lifting the air ahead of it is probably not abad concept, but the presence of lift along thefrontal boundary depends critically on the

relative, normal components of the winds oneither side of the boundary (Fig. 3.6), as describedby Saucier (1955, p. 292ff). One really needs toexamine the wind field in the vicinity of the frontcarefully, as well as the speed of frontalmovement, to complete the picture. Note that thestructure of the pressure and temperature fieldsalong a cold front tend to increase the cyclonicshear across the front near the surface (Petterssenand Austin, 1942). Therefore, cold fronts areusually characterized by abrupt wind shifts.

By simple density considerations, the stronger the

Fig. 3.6. Schematic of two types of cold fronts (after Saucier, 1955). In upper figures, solid lines showcontours of the 1000 mb surface (with frontal analysis), dashed lines are contours of an upper levelisobaric surface (say, 700 mb), and dotted lines depict the thickness contours between them. In (a) the frontis moving faster than the normal wind component aloft. Therefore, relative to the front, warm air is lifted“upslope”, producing precipitation behind the frontal zone. In (b), the front is moving more slowly than thenormal component aloft. Thus, there is “downslope” movement and subsidence of the warm air over thefront, giving clear air behind the surface boundary. Any weather associated with type (b) is likely to occurahead of the front, perhaps by several hundred km.

39

density contrast, the shallower the cold air masstends to be. Therefore, the overall frontal slope isless when the cold air mass is markedly colderthan the air it displaces. As a result, the airmassboundary aloft may not be easily identifiablemuch above 850 mb, and the horizontaldisplacement of the boundary with height can belarge. When the air masses are only weaklydifferent, the frontal zone can be more nearlyvertical. Note in Fig. 3.7, that the resultingisotherms on an isobaric surface which intersectsthe boundary may have similar gradients in bothcases, or the �weak� front can actually show astronger horizontal gradient on a given isobaricsurface!

A final word on cold fronts concerns the use ofdewpoint gradients to locate the boundary. As weshall see, a dewpoint difference is not necessarilyindicative of a frontal boundary. Further, ifmaritime air is displacing continental air, thedewpoint difference may be negligible or, indeed,may be such that the cold air is also wetter. Aswith any rule of thumb, the use of dewpoints tolocate fronts must be done with caution, usingother information like winds, temperatures, andupper-air data, as well as common sense. Thisfurther suggests that one should not anticipate thepost-frontal weather based on classical models. Inmany cases, cold-frontal passage marks the onsetof cloudiness, not a clearing line.

2. Warm Fronts

As one might expect, along a warm front cold airis being replaced by warm (Fig. 3.8). In thenorthern hemisphere, this suggests that the basicflow behind the front generally has a southerlycomponent. Unlike cold fronts, when a warm airmass begins moving northward, it typicallyoriginates in a region where the air mass contrastis weak and the front may well be difficult tolocate at the surface. Since the air is often warmerthan the surface over which it moves, thepresence of a surface inversion can make theboundary difficult to find early in the day, untilsurface heating can break the inversion.

As with the cold front, there is a general picture ofthe flow pattern associated with a warm frontwhich has wide applicability. That is, it isgenerally conceived in terms of the warm airgliding up over the retreating cold air. The sameproblems exist with this concept as with thatassociated with the cold front, since the truepicture depends on the relative normal windcomponents and how they vary with height. Theconcept of �overrunning� can be valuable whencombined with a clear understanding of howvertical motion arises. However, �overrunning� isa term often overused and inappropriatelyapplied.

By finding the 850 mb warm front, it is oftenpossible to return to the surface data and findclues as to the location of the warm front thatmight have been too subtle to see at first glance.Since the thermal contrast at low levels across awarm front tends to be weaker than thatassociated with cold fronts, the picture shown inFig. 3.7 suggests that a possible aid to warmfrontal location is by first checking the isothermpattern at 850 mb. Experience tends to bear this

Fig. 3.7. Schematic showing how fronts withdifferent slopes can give the same horizontaltemperature field on a given isobaric surface.

Fig. 3.8. As in Fig. 3.5, except for a warm front(after Byers, 1959).

40

out. As Petterssen and Austin (1942) have shown,the temperature and pressure fields along a warmfront act to decrease the cyclonic shear across theboundary, so the surface wind shift is often verysubtle. A convenient rule of thumb which can behelpful in locating warm fronts is the observationthat isobars in the warm sector of an extratropicalcyclone are usually more or less straight lines.Isobars with substantial curvature are generallyon the cold side of the warm frontal boundary.

As with cold fronts, the dewpoints can be useful,provided the analyst is aware of the source regionfor the advancing air. In the central United States,most warm air masses have a trajectory over theGulf of Mexico�s warm waters and are, therefore,usually more moist than the air they displace.However, the descent of moisture fromprecipitation into the cold air mass in case of�overrunning� can act to smear out or displacethe dewpoint gradients. Of. course, some warmair masses are continental in origin or may beassociated with subsidence, implying that theymight be drier than the displaced air. As with coldfronts, the analyst/forecaster should avoiduncritical use of classical models of warm frontweather sequences.

3. Stationary Fronts

Although the obvious definition for a stationaryfront is one which has neither air mass advancing,it must be modified somewhat, since it is rare thatthe boundary is truly stationary along its entirelength. The term frequently applied is�quasistationary� which generally requires thatthe movement be slow (say 10 knots or less) orerratic. It is not uncommon for a �stationary� frontto move slowly across a state over a 24 hourperiod.

Fronts become quasistationary when the flownormal to the boundary becomes negligible. Thiscan arise either when the front has a weak flowfield or when the winds become essentiallyparallel to it. In the former situation, the thermalcontrast across the front is also usually weak andthe frontal zone is quite diffuse. In such a case, itmay be said that the front has �washed out� andthe front dropped from the analysis. It isnoteworthy that the old frontal zone may still be

characterized by a pressure trough (see III.C.2)and significant weather may be found roughlycolocated with it.

Stationary fronts which retain a recognizablecontrast are quite significant, since �overrunning�may be a possible weather threat. This isessentially a situation where some mechanismcreating vertical motion moves over the cold airand has a source of warm, moist air (generally tothe south of the boundary) to sustain significantconvective weather. In these cases, depending onhow far north of the boundary the vertical motionis occurring, there may not be any easilyrecognized surface feature prior to the weatherevents. Satellite imagery is very useful forlocating and forecasting regions of upward motionwith roots above the surface. Also, numericalmodel forecasts may give an indication of someupper-level feature which is moving over thesurface cold air dome.

Most of what relates to locating warm and coldfronts can also be applied to stationary fronts.With respect to winds at the surface, it is notuncommon for substantial flow normal to a front tooccur on the warm air side of the boundarywithout significant frontal movement. This is thetypical case in �overrunning� situations.Examination of the winds aloft often reveals thatthe winds become essentially parallel to the fronta short distance off the surface, which hardly fitsthis simple picture. In such cases, some source forvertical motion over the front can usually befound (recall the discussion in II.B.1).

When a vertical motion field passes near thesurface boundary, it is common for cyclogenesisto occur, and an extratropical cyclone maydevelop along the boundary. This process is theclassical sequence of development for suchstorms and is thoroughly discussed in thereferences.

Alternatively, a similar redevelopment can takeplace, but on a smaller scale, in which a weak�frontal wave� is formed (Fig. 3.9), which nevergoes on to develop into a synoptic-scale weathersystem. Such systems are only hundreds of km indiameter, as opposed to thousands of km forextratropical cyclones. This process of weakwave formation can occur several days insuccession along a quasistationary boundary,

41

producing a situation where heavy precipitationand severe weather of various types (dependingon the season) occur repeatedly over a limitedarea. The dynamics of such weather systems arenot well understood, but careful analysis can be ofvalue to the forecaster, since these systems tendto concentrate low-level convergence (values oforder 10-4 s-1, implying upward motion of order 10cm s-1) into localized areas (Tegtmeier, 1974;Doswell, 1976).

4. Occluded Fronts

In the standard picture of the extratropicalcyclone, the cold air moves southeastward fasterthan the warm air moves northeastward. Thisleads to the cold air overtaking the warm frontand the initiation of the occlusion process. Interms of the extratropical cyclone�s life cycle,intensification of the large-scale storm occurswhen it has energy available from the air masscontrast. By moving warm air upward andnorthward, as well as cold air downward andsouthward, the storm acts to diminish theavailable energy used for intensification. By thetime occlusion begins, this air movement hasusually exhausted the cyclone�s potential forfurther development. Occlusion proceedsbecause the flow pattern established duringcyclogenesis does not just cease after its

amplification stops. Rather, mixing of thecontrasting air masses continues essentially on itsown inertia, When a cyclone is thoroughlyoccluded, it has succeeded in mixing the airmasses so well that near the center of the low itbecomes quite difficult to locate any significantair mass boundaries. However, wind shift linesand pressure troughs may continue to rotatearound the low like �spokes in a wheel� (seeKreitzberg and Brown, 1970 for more detailsabout the features in an occluded system).

Traditionally, discussions of occluded fronts go togreat pains to differentiate between �warm type�and �cold type� occlusions (see Fig. 3.10; alsoSaucier, 1955, p. 268 ff). The cold type is probablythe most common in the United States. As Saucier(1955, p. 271) points out, the upper warm front ina cold occlusion is not usually significant insofaras surface weather is concerned. Since the windshifts and pressure trough need not be coincidentwith the surface location of an occluded front, it isdebatable whether or not occluded fronts per seare really a significant aspect of surface analysis.Generally, the surface trough (in which theoccluded front is usually inserted) actually reflectsthe axis of the deepest warm air (a thermal ridge),rather than a front in the strict sense. In fact, theregion near the center of an occluded cyclonecan be so thoroughly mixed that the analyst mightbetter spend time on other aspects of surface

Fig. 3.9. Surface analysis showing example of aweak frontal wave in Texas and Oklahoma.

Fig. 3.20. Schematic exampleshowing formation of anoccluded front, with coldfront overtaking warm front(after Saucier, 1955). In theexample on the bottom, thecold air ahead of the warmfront is colder than thatbehind the cold front, givingrise to a warm-type occlusion.This is relatively rare in theU.S. In the case just above,the cold air behind the coldfront is colder than thatahead of the warm front,yielding a cold-typeocclusion.

42

analysis than the confusing problem of locating anoccluded front.

As a final word with regard to occluded systems,the analyst needs to be alert to the possibility ofintrusions of �fresh� air masses into thecirculation, and to structures conducive tofrontogenesis. Although classical pictures ofoccluded systems indicate little likelihood foradditional significant convective weather, thereare two situations that bear watching. First, if thecirculation (which may persist for days afterocclusion begins) can tap another air mass, a new�boundary� can be created which may result inan abrupt shift in the location of significantweather with respect to the circulation. Second,air mass modification can occur in the clear zonefollowing a closed system�s frontal passage. Ifsurface heating is sufficient, a rapiddestabilization can occur in this clear air (oftennear the core of the vertically stackedcirculation), resulting in convection and severeweather on a local scale. Such developments canalso he associated with secondary shortwavetroughs rotating through the large scale occludedsystem (see III.C.2). These secondary systemsoften have frontogenetic circulations and enoughvertical motion to contribute to destabilization.

5. Drylines

The dryline is a subject which has receivedrelatively little formal attention in textbooks, butforecasters in the Great Plains must frequentlydeal with it in relation to convective weather.Schaefer (1973a,1974a,b) has given the mostdetailed accounts of the structure and origins ofthe dryline as a synoptic scale feature, whileRhea (1966) has discussed the occurrence ofthunderstorms in relation to the dryline. Since thisfeature is not well-recognized outside the regionswhere it frequently occurs, it is often mistakenlyanalyzed as a �Pacific front� � i.e., the leadingedge of an air mass with a Pacific source region.

Research has emphasized the regional characterof the dryline � its occurrence at the surface isgenerally confined to the Great Plains, west of theMississippi River and south of the Dakotas. This isgenerally recognized to be the result of thesloping terrain. As moisture returns (from the Gulf

of Mexico) to the plains following the passage ofan anticyclone, the rising terrain limits itswestward penetration and shunts it northward(Schaefer, 1974b). Thus, a natural tendency existsfor the development of. a north-south boundaryseparating dry from moist air. The moisturegradients observed with mesoscale networks canbe enormous (Fujita, 1958: McGuire, 1962), withmixing ratio changes of 5 g kg-1 over a distance of1 km. The operational data do not provide thedetail to resolve such intense gradients, of course.However, time series observations at specificstations often reveal dewpoint temperature dropsof 30 deg F (about 16 deg C) in a matter ofminutes.

The dynamics of the dryline flow regime aredominated by boundary layer processes, assuggested by Schaefer�s work (1974a,b). Tosummarize briefly, the development of a well-mixed boundary layer during the morning acts todisperse moisture vertically. If the moisture isshallow, this can be seen as a �movement� of thedryline past the station, at which time the surfacedewpoint drops while the temperature rises. Suchmovement often is related very poorly to the windfield, under quiescent conditions (i.e., when thedryline is not involved in a synoptic-scalecirculation). With the re-establishment of theinversion during the evening, the dryline �backsup� and its motion then is related much moreclearly to the winds in the moist air as it returns.

It should be pointed out that the dryline is not afront, in the sense of a density discontinuity.During the morning hours, the air on the dry sideof the boundary is quite cool, since dry air(usually cloudless) enhances radiational cooling.However, by early afternoon, the dry air normallyis warmer than that on the moist side and theresulting surface virtual temperatures (or,equivalently, the densities) across the dryline areessentially equal on both sides. The morningthermal contrast is not a real frontal characteristic,but it can be quite deceptive to an unwaryanalyst.

Further evidence of the non-frontal nature of thedryline is the finding that the interface betweendry and moist air is nearly vertical off the surface,and quasihorizontal to the east (Fig. 3.11; recallthe discussion in III.B.1). As Rhea (1966) andSchaefer (1975) point out, the dryline is typically

43

located in a zone of small scale convergence.This convergence is reflected typically in anarrow line of cumulus clouds which is colocatedwith the dryline. On a somewhat larger scale, theconvergence of the routinely observed surfacewind field often actually reaches its maximumsomewhat behind the dryline, While the moistureconvergence maximum is right along theboundary (Doswell, 1976). The velocityconvergence maximum behind the dryline usuallylies within a region of strong downstream speeddecreases. Although the dryline itself may belocated in the area of rapid directional changes,the convergence is usually dominated by thespeed changes upstream. Since moistureconvergence also depends on the moisture field,the combined effect usually is to put the strongestmoisture convergence near the dryline axis. As anaside, the reader should exercise caution in�eyeballing� convergence patterns from theplotted wind data.

Recall that frontal boundaries are defined to lie onthe warm side of the gradient, by convention. Nosuch convention exists for the much less well-documented dryline. As suggested by Fig. 3.11,the moisture gradient can be so strong as to beindistinguishable from a true discontinuity on anormal surface map used for analysis. Of course,

conventional surface data are nowhere nearbeing dense enough to sense this intense gradient� so the gradient is seen to be smeared out.Schaefer�s work follows, insofar as possible, theconvention that the dryline is on or near the 45deg F (about 7 deg C) isodrosotherm. This alsocorresponds roughly to a mixing ratio of 9 g kg-1.Adjustments can be made on the basis of veeringwinds, as the dryline moves past an observationsite. It should be emphasized here that, especiallyduring the morning hours when the dryline isusually not moving rapidly, winds can be verydeceptive for dryline location. The morninginversion has de-coupled the surface winds fromthe significant free atmosphere flow, thus makinglocal effects more important.

Although most of Schaefer�s studies concern thedryline under quiescent conditions, they are ofwider applicability, subject to some modification.The forecaster should be aware of moisture depthas a clue to dryline motion � if the moisture isshallow, it is likely that the solar heating canbreak the morning inversion quickly, thussuggesting an early dryline �passage�. Deepermoisture will slow down the apparent movementof the dryline during the day.

When large synoptic-scale systems are present

Fig. 3.11. Aircraft traverse through dryline (after Staff, NSSP, 1963)showing mixing ratio isopleths (g kg-1). Note the very intensemoisture gradient, the nearly vertical boundary, and the evidence ofwave-like perturbations on the wet side of the dryline.

44

over the plains, the dryline is often a major factorin the day�s severe weather forecast problem. It iswell-known that a deep surface-based layer ofnearly dry-adiabatic lapse rates (as shown in Fig.3.12)2 is present on the dry side of the dryline(Carlson and Ludlam, 1968). This very dry,unstable air mass is not likely to produce anyclouds or weather, but it does play a role byallowing strong mid-tropospheric winds(occasionally as high as 25 m s-1 [about 50 knots]or more) to penetrate to low levels. With theelimination of the surface inversion, momentumfrom higher levels is free to be transported all theway to the surface. This high-speed flow (usuallyfrom the southwest) can generate dust storms inthe dry air, and influences the motion of thedryline, as well as the resulting moistureconvergence along the boundary. Theappearance of a strong �push� (high winds) at thesurface in the dry air is a reliable indicator thatthe dryline will become active with severethunderstorms. While thunderstorms of a non-severe character may occur on the drylinewithout such conditions, the dryline is not usuallyactive unless the strong winds appear in the dryair.

Note that the structure of the dryline can beexceedingly complex, as suggested by thereferences. Often, the convection is most activeon a second boundary, ahead of the �true� surfacedryline (as revealed by the air with dewpoints lessthan 45 deg F). This may be a reflection of anupper dryline (Fig. 3.13), with shallower moisturebetween this boundary and the true dryline(Siebers, et al., 1975). Further, the surface drylinerarely intrudes east of the Mississippi, but Miller(1972) refers to upper level �dry prods� as asignificant ingredient for severe weatherthroughout the country. Such features aresometimes reflected in the surface data as a subtlebreak in the dewpoint temperature gradients. Atother times, the surface data may not reveal the�upper dryline� at all.

Much work remains to be done in illuminating thestructure and behavior of drylines in conditionssupporting severe convection (see McCarthy andKoch, 1982), but the careful analyst needs to beaware of the dryline and its interaction with large-scale flows. It is a common feature ofextratropical cyclones in the Great Plains, withthe dryline typically intersecting the front at ornear the low center, forming what is often referred

Fig. 3.12. Samplesoundings through dryair on west side ofdryline (Midland, Texason 8 June 1974). Thinlines show temperature(solid) and dewpoint(dashed) curves at 1200GMT. Thick lines showsimilar curves at thefollowing 0000 GMTsounding time. Dash-dotlines show representativemixing ratio lines on thechart. Note shallow,surface superadiabatic“contact” layer at 0000GMT.

45

to as a �triple point� (Fig. 3.14). A commonmistake in analysis, as mentioned, is to put a coldfront in the trough containing the dryline and toignore the real thermal contrast to the northwest.This error can be avoided by examining the 850mb chart to locate the air mass discontinuity, bycontinuity checks on the location of the cold front,and by careful examination of the wind fields at ornear the surface. In the dry air, there is lesstendency for a northerly component to the flowthan in the cold air behind a true cold front.Further, the dewpoints in the cold air mass areoften relatively high.

6. Land/Sea Breeze Fronts

Unlike the dryline phenomenon, the thermallydriven circulation associated with the boundarybetween land and water is well-recognized. Alsoin contrast with the dryline, this feature is a truemesoscale feature, with its influence generallyconfined to the near-shoreline zone, say within 30km either side of the shoreline (O�Brien andPillsbury, 1974). The theory of such a flow hasbeen fairly thoroughly developed (Haurwitz,1947; Defant, 1951; Estoque, 1961; Pielke, 1973).

Generally speaking, the land/sea breeze is not afactor in most severe weather. This is a direct

result of the fact that most severe weather ofconcern to SELS occurs well inland, away fromlarge bodies of water. Further, the circulationsinvolve wind speeds of only a few m s-1 , overlimited areas, which limits the overall significanceof the flow. However, the mesoanalyst needs tobe cognizant of this phenomenon since therecertainly have been occasions where it has beena significant influence (e.g., Fujita and Caracena,1977; Lyons, 1966). Note that the Great Lakes andLake Okeechobee in Florida are large inlandbodies of water which show clear land/seabreeze circulations (see Neumann and Mahrer,1975).

In physical terms, the resistance that water has tochanges in temperature, compared to theadjacent land areas, is the basic factor leading tothe circulation. During the warm season, a watersurface typically is cooler than the land during theday and warmer at night. Allowing for a time lagin the heating and for circulation development,the surface flow reaches its peak during themorning and again in the afternoon (see e.g., Hsu,1969 or Neuman and Mahrer, 1974).

Basically, as the land surface heats during theday, it warms the air above it relative to that overwater, giving the air over land a greater tendencyto rise. This creates a horizontal flow off the water(the sea breeze) and compensating subsidenceabove the water. The cooler air flowing off the

Fig. 3.13. Schematic cross sections along lines ABand A’B’ showing different depths of moist layers indifferent locations along the dryline (after Sieberset al., 1975). Clouds are indicative of convectiveintensity.

Fig. 3.14. Schematic example of triple pointintersection of fronts and dryline in a developingfrontal wave.

46

water penetrates inland in the form of a small-scale front, with the amount of penetrationdepending partly on the strength of the flow,which in turn depends on the land-sea thermalcontrast. This basic flow is modified by frictionand Coriolis forces, resulting in the developmentof along-shore components to the wind (Dutton,1976, p. 375ff). A compensating flow is requiredaloft to complete the circulation (Fig. 3.15), withits height and strength depending on the overallstability of the air mass. If the overall flow iscapped by a persistent inversion, thecompensating horizontal current aloft will bestronger. During the night, the situation isreversed, with a period of transition separating thedaytime �sea breeze� from the nighttime �landbreeze�.

Although local non-severe thunderstorm activityoften is influenced by convergence along the seabreeze front (Neumann, 1951), the analyst is oftenmore concerned about possible interactions ofother mesoscale features with the front. By itself,the sea-breeze circulation usually is not strongenough to produce severe convection. In fact, thefront itself may be masked by larger scale featureswhich can dominate the wind and thermal effectcirculation. Often, the land/sea breeze front isoffset by prevailing flows, which displace theboundary downstream.

Owing to its mesoscale nature, the frontalboundary is often difficult to analyze fromconventional data and the best tool for locatingthe sea breeze front is usually the satellite image(e.g., see Fig. 3.16). Clearly, this is a small-scalefrontal boundary in the true sense of the term, somany of the clues for locating synoptic scalefronts can be used here. The analyst is cautionednot to expect the winds to blow directly off thewater surface, especially late in the day when theCoriolis force has had time to develop along-shorewind components.

7. Thunderstorm Outflow Boundaries

Since thunderstorms are almost totally limited toenvironments with unstable thermodynamicstratification (mechanically forced convectiondoes occur � e.g., Carbone and Serafin, 1980),one interpretation of thunderstorms is that theyexist to exchange warm, moist air at low levelswith cold, dry air at upper levels. While this is anoversimplification, observations generally supportthis basic view. Thus, the downdrafts whichaccompany thunderstorms bring down to the

Fig. 3.15. Isotachs of land/sea breeze regime as afunction of time (after van Bemmelen, 1922).

Fig. 3.16. Visible satellite image showing seabreeze-induced cloud lines along the Texas andLouisiana gulf coasts, and around the Floridapeninsula, as well (arrows).

47

surface a mass of relatively cold air (comparedwith the air rising in the updraft). The greaterdensity of this cold air forcing the downdraft alsoserves to keep it at lower levels, just as in thecase of the advancing cold air behind a cold front.Therefore, upon reaching the surface, the cold airis forced to spread laterally, forming a small scalecold front, usually called a gust front. Temperaturecontrasts across the gust front can be in the rangeof 10 deg K km-1.

In many ways, the mesoscale outflow boundaryresembles a true cold front. There have beennumerous studies of this feature (Charba, 1974;Goff, 1976; Mitchell and Hovermale, 1977) whichemphasize the details of its structure. From theviewpoint of the mesoanalyst, most of thesedetails are not adequately resolved byconventional surface data, but they can beimportant in trying to interpret the weathersituation. Figures 3.17 and 3.18, taken fromCharba (1974), show many details of the outflow�scharacter, including the fact that the initial windshift and the beginning of the pressure rise can besome distance ahead of the temperature breakand the really strong winds. Further, the results ofGoff (1976) show there is some tendency for alow-level backflow, which can create observedsurface flows which appear to be blowing throughthe gust front. In such cases, the cooltemperatures and high pressures are the mainclue to the gust front�s passage.

Obviously, the general effects associated withgust front passage at an observing site are abruptchanges in wind direction and speed, usuallyaccompanied by rapid cooling and rising surfacepressure (Fujita, 1959). A typical sequence ofevents observed at a station during gust frontpassage is shown in Fig. 3.19. Since the gust frontusually is associated with a rainy downdraft (seethe discussion of storm types and structure,II.III.A.5), precipitation begins after gust frontpassage. When the outflow has just begun, theprecipitation may be quite close to the gust front.Later in the storm�s life cycle, this first surge ofoutflow can move well out ahead of theprecipitation area. Because the outflow spreadsout in all directions although predominately in thedirection of storm motion), some areas mayexperience gust front passage and never receiveany precipitation.

Most convective situations produce more than asingle thunderstorm cell, so several surges of coldair may be produced in succession. Further,storms separated in space and time can produceseparate regions of outflow which eventuallymerge into a larger area of basically similar rain-cooled air. With time, this produces a subsynopticscale �bubble� (or �mesohigh�), characterized byhigher pressures and cooler temperatures, easilyresolved by the conventional surface network.The leading edge of this bubble is the combinedgust fronts from the storms which produced it andis often analyzed as a �squall line�. As pointed outby Fujita (1955), this feature can mask the truefronts in the area, making the analysis quitedifficult. This is true particularly when the stormsoccur near a front and have not moved far enoughaway from it to make the distinction clear.Occasionally, the outflow boundary can dominatethe true front, and the squall line effectively

Fig. 3.17. Detailed structure of thunderstormoutflow, showing spatial relationship of outflowboundary to radar echo at a specific time (afterCharba, 1974). Note the complex nature of theoutflow boundary with the windshift line andpressure jump leading the temperature break andgust surge.

48

becomes the frontal boundary. This seems tooccur most often in the case of �southward burst�-type squall lines, as described by Porter et al.(1955).

Outflow boundaries are generally convergent, sonew convection frequently develops along themin response. These boundaries can haveconvergence values of 10-3 s-1 associated withthem (implying vertical motions of 1 m s-1 at aheight of 1 km). It is important to analyze andforecast these boundaries properly. It is also ofvalue to be able, if possible, to distinguish the gustfronts from the true fronts, since new activity candevelop on the front after the first line ofconvection has moved away. Wind and pressurepatterns behind the squall line�s gust front havebeen well-documented by the pioneering work ofFujita (1955), Fujita and Brown (1958), Zipser(1977), and others. After the storm�s passage,pressure may fall below the pre-storm valuebehind the mesohigh, forming what Fujita hascalled a �wake depression�. Often the windsreturn to a southerly direction, the clouds breakup, and temperatures and dewpoints recover.Naturally, this sequence will not occur when thetrue front retains its identity and passes beforerecovery can occur. The recovery to near pre-storm conditions following gust front passage isthe primary clue which allows the analyst todistinguish a gust front from the true front.

Fig. 3.18. Schematiccross section throughthunderstorm outflow,showing relative depthsand vertical structure offeature (after Charba,1974).

Fig. 3.19. Example of time sequence of eventsduring gust front passage at a point (after Charba,1974). See Fig. 3.17 for explanation ofabbreviations.

49

Many of the details concerning behavior of thegust front and the flow behind it are connectedintimately to the storm(s) generating the outflows.The basic configuration, shown by Fig. 3.20 iswidely applicable, hut some features in a givensituation may differ as a result of the interactionsbetween storms. As described by Lemon (1976),Barnes (1978), and Lemon and Doswell (1979),the gust fronts from previous storms can have asignificant role in the organization of severe�supercell� storms. In fact, Lemon and Doswelldescribe in some detail how a single supercellcan develop two downdrafts, with the resultinggust front interaction possibly having a major rolein tornadogenesis. Such a process generally isbelow the resolution limits of surface data. As thesupercell collapses, it is a common event for thedissipating storm to become imbedded in a moreor less solid line of storms, with the gust frontsmerging, as already described. The details ofstorm type occasionally can be inferred fromradar and satellite data, and this topic will becovered elsewhere (II.III.A.1).

Available surface data can aid the analyst in hisassessment of possible future behavior of gustfront convection in much the same way as it canwith true fronts. That is, as described recently by

Maddox et al. (1980), the surface convergenceand vorticity fields associated with interactingsurface boundaries have a clear relationship tosubsequent activity. The analyst should monitorthe situation hourly to detect important clues as tohow the interaction(s) are proceeding. Althoughthe time-to-space conversion technique ofmesoanalysis,3 used by Fujita (e.g., 1955) andBarnes (1973), among others, is not advocated inthese notes, there is no question that time seriesobservations (Maddox et al., 1980; Moller, 1979)are of great value in helping relate theobservations to what is known about howconvective storms behave.

A reasonable approach for the analyst is to plotthe hourly observations (and any received�specials�) along a time line, for selected stations.This makes a convenient reference tool whenconstructing spatial charts, and should not requiretoo much time to accomplish and to update on anhourly basis. The key idea here is to compare thesequence of observations against the spatialanalysis, to check on the consistency of theanalysis with the analyst�s view of the events.

One problem which has become widespreadneeds to be mentioned here. There is a tendencyfor the terms �instability line�,�gust front�, and

Fig. 3.20. Model ofthunderstorm-inducedmesoscale weathersystem (after Fujita,1955). UPD denotesupraft, white DWDstands for downdraft.

50

�squall line� to be used interchangeably. Theappearance of thunderstorm lines on the surface(and radar) chart has received considerableattention (Newton, 1950; Fulks, 1951; House,1959). In the author�s opinion, this has resulted inan overemphasis on �squall lines�. Confusionexists between what are referred to in thisdocument as squall lines (see II.III.A.5.d) and otherpre-existing, non-frontal, non-convective linearfeatures in the analysis, Clearly, there is a markedtendency for storms to form in lines (with thespacing between storms quite variable in spaceand time) which is not entirely understood. Therelationships, if any, among these phenomena(i.e., non-convective linear analysis features, solidlines of radar ehoes, thunderstorms in lines,connected gust fronts, etc.) is not at all obvious.Considerable research remains to be done andsome of the present confusion is apparentlyrelated to a failure to distinguish betweenphenomena of different scales.

Intersecting solid lines of radar echoes are not acommon event, If one can assume a roughequivalence between instability lines and squalllines, one might suspect that the so-called lineecho wave pattern or �LEWP� (see Nolen, 1959;and II.III.A.5.c.ii for more details) configuration ofradar echoes is related to this �intersecting lines�phenomenon. However, recent studies (Fujita,1978) suggest that the LEWP and the so-called

�Bow Echo� are basically the same phenomenonand are the result of downdraft accelerations, notcirculation around a mesolow. This is not to saythat mesolows are not associated with LEWPs. Weshall return to mesolows later. However, thesubsynoptic scale aspects of the mesolowphenomenon may have been overemphasized inthe recent past, since the picture described byMagor (1959; his Fig. l) is not particularlycommon, as sensed by conventional surface data.

Remarks on surface observations often provideuseful information and one facet of these remarkshas a direct bearing on the mesoanalysis. Theleading edge of a cold outflow is often marked bya �shelf� or �roll� cloud, the latter term appearingmost frequently in additive remarks. Actual roll-type clouds look like Fig. 3.21, while the morecommon shelf cloud appears in Fig. 3.22. Thepresence of either type of outflow cloud is usuallya clear indication of a gust front [or on rareoccasions, a true cold front (Livingston, 1972)].When remarks include the observation of such acloud, the location of the gust front can be madequite accurately, at least in the vicinity of theobservation. If the radar echoes (precipitation) arequite distant from the shelf cloud observation, thenit is clear that the outflow boundary has movedwell away from the downdraft which initiated it.

In the area of thunderstorm outflow boundaryidentification and tracking, satellite images make

Fig. 3.21. Example of aroll-type outflow cloud,the detached, horizontaltube-like cloud near thebottom of the photograph.Occasionally, this tubemay appear to be rotatingslowly about a horizontalaxis, with the forwardedge rising and the rearside sinking.

51

an important contribution to mesoanalysis. Itcannot be overemphasized that concurrentsatellite images (preferably in the form ofanimated loops] need to be integrated thoroughlyinto the process of mesoanalysis, It is by thismeans that the positioning of surface boundariescan be refined and clarified. At the same time, thesignificance of cloud features can be assessed bycomparison with the surface (and upper air)observations.

The stabilizing influence of the outflow air can beseen readily in images of �air mass�-typethunderstorms (i.e., the small, short-lived, isolatedconvective cell). The air which has been cycledthrough the storm is typically cool, subsiding, andstably stratified. This suppresses clouds in thewake of the storm, forming a cloud-free area (Fig.3.23), perhaps surmounted by an anvil remnant. Itis common that such a feature is too small to beresolved by the surface network and yet it mayhave an influence on a local weather forecast.

The outflow boundaries from more significantconvection may live on after moving away fromthe storm. This typically takes the form of the�arc� cloud, which Purdom (1973) has describedin detail. The are cloud usually coincides with thegust front and results from cumulus and toweringcumulus which concentrate in the zone ofconvergence along the boundary. On someoccasions, the are may be associated with apressure jump line (Shreffler and Binkowski,1981).

There are two aspects of are clouds which needclarification during the act of mesoanalysis. Thefirst question is which, if any, are clouds are likelyto become reactivitated during their lifetimes?Since gust fronts (and their associated are clouds)occur with virtually all significant (i.e., thoseproducing downdrafts) convective storms, whileonly some persist and re-develop thunderstorms,this is an important problem. The second questionis where are new developments likely to occur,given the observation that storms do not alwaysdevelop uniformly along the outflow boundary?Purdom (1973, 1979) has made considerableprogress in helping to resolve the secondquestion. He has indicated that wherever outflowboundaries intersect other surface boundaries(including fronts, sea breeze fronts, other outflowboundaries, etc.), these are preferred locations fordevelopment. This observation is an obviousargument for the value of a careful hourlymesoanalysis of conventional data, since surfacefeatures are not always well-delineated bysatellite imagery alone. Purdom�s most recentstudies also suggest that the cloud types andamounts ahead of the are cloud can be useful in aqualitative assessment of the stability of the airmass which will be influenced by the arc. Heasserts that arc clouds propagating into clear skiesare unlikely to produce strong convection. Theseideas are not yet developed fully, sinceexceptions occur and their day-to-day value

Fig. 3.22. Example of a shelf-type outflow cloud, in thiscase showing a somewhatterraced appearance. Ifmotions can be seen, they willbe upward along the upperportion of the wedge-shapedshelf cloud. Note that theshelf is attached to cloudbase above it.

52

remains to be demonstrated. Nevertheless, theysuggest ways in which the satellite data can beintegrated with the conventional, and represent arapidly evolving line of research.

When storms mature, their anvils frequentlyobscure any low-level features like gust fronts. Itis during this phase of storm development thatradar is a natural tool for supplementation ofconventional observations. There are three typesof radar data that may be available: onsite,remote, and facsimile (or AFOS) displays. Onsiteradar can be controlled, within its operationalguidelines, to give the best possible depiction ofthe storm configurations for purposes of themesoanalysis. An identification of storm types (seeII.III.A.5,6) can be valuable, but a broad-scalepicture of the precipitation distribution is mostvaluable for mesoanalysis. That is, the analystneeds to know the location of any thunderstormlines, isolated storm cells, radar fine lines (whichoften reveal fronts and/or gust fronts), and theheaviest precipitation cores. These provide detailswhich may be important when the storm anvilscover the area of interest and, of course, at othertimes as well. �Spearhead� or �bow� echoconfigurations (Fujita, 1978) may revealaccelerating gust fronts, between conventionalobservation sites and times. Although radar data

Fig. 3.23. Visible satellite imageshowing a small, nearly circularoutflow boundary in the Gulf ofMexico (arrow). Note the ring ofenhanced cumulus surroundingthe area of nearly cloud free air.Remnants of the thunderstormanvil can be seen on thenortheast side of the ring.

are considered most important for warningdecisions and discussed in that context in III.IV.8,they cannot be ignored by the mesoanalyst.

When radar is not located at the station where theanalysis is done, the alternatives are remotedisplays of one (or more) radar(s) in the area, andthe facsimile (or AFOS) display of the nationalradar data. These displays are not as flexible, butcan still provide enough information to get arough idea of the precipitation distribution.Although storm type identification is more difficultunder these circumstances, enough informationcan often be obtained to be of help to the analyst(Wilson and Kessler, 1963).

As shown clearly by some of the earlier studies inmesoanalysis (e.g., Fujita and Brown, 1958), afterthe mesosystem has evolved into a mature stormcomplex, the area influenced by outflow can bedramatically larger (perhaps 20 to 100 times) thanthe area actually covered by precipitation echoes(see Fig. 3.24) at any given moment. However, asalso seen in their study, the area eventuallyreceiving precipitation is usually a substantialfraction of the outflow �bubble� (Fig. 3.25).

One can readily see that outflow boundaries tendto have a reasonable continuity from hour to hour.When conventional data are meshed with radar

53

and satellite information, the analysis andprognosis of the boundaries are relativelystraightforward, provided the analyst is aware ofthe basic types of structures seen.

As a final word on outflow boundaries, the readerneeds to be aware of the flash flood potential insituations involving this phenomenon (also seeIII.F and II.III.C). Maddox et al. (1979) have notedthat many of the flash floods which they studied(34%) were associated with storms developingalong an outflow boundary. Therefore, the analystneeds to be concerned with factors relating toflash flood potential. A significant factor in thefrequent occurrence of flash floods during thenighttime hours is the low-level jet phenomenon.Although the low-level jet will be discussed morefully later, the analyst should note the axis of thelow level flow and where the speed maxima arealong that axis. Any situation wherein strong low-level flow impinges on an outflow boundary hasflash flood potential, when the undisturbed low-level flow is moist and unstable.

C. Boundaries Not Involving AirMasses

1. Wind Shift Lines

The careful analysis of surface data frequentlyreveals organized windshift lines which areapparently non-frontal in character. Since, bydefinition, no clear change of air mass is involved,it is often difficult to explain or understand theirorigins. Nevertheless, they do occur and, sincethey may give rise to surface convergence, theycan be involved in severe weather events. Attimes, the wind shift may be traceable to an oldoutflow boundary from thunderstorms occurring,say, the previous day. Modification of the outflowair may have erased any apparent temperature

Fig. 3.24. Relationship between radar echoes(black areas) to overall mesosystem produced bythe storms associated with the echoes (after Fujitaand Brown, 1958).

Fig. 3.25. Time averaged precipitation producedin several periods during the life of themesosystem in Fig. 3.24 (after Fujita and Brown,1958). Reported wind gusts are plotted in theupper left.

54

and dewpoint differences. The same thing canhappen with old quasistationary fronts.

A commonly observed non-frontal wind shift lineis one which occurs ahead of a cold front. Thissituation can make the analysis difficult sincefronts themselves often are analyzed (mistakenly)along the wind shift line. This wind shift may bethe focus for development of the pre-frontal�squall lines�. At other times, the front itself is themain source of convective development (seeIII.B.1). Analysts should always be alert to theformation of these pre-frontal wind shifts. Whilethe mechanism for such a phenomenon is not yetcompletely understood [frequently-mentionedcandidates are the �pressure jump� proposed byTepper (1950) and described by Shreffler andBinkowski (1981), ancl Clarice et al. (1981); andthe gravity wave (Uccellini, 1975).], analystsneed to be aware of this phenomenon, and takecare to distinguish the true front from this pre-frontal wind shift line. Note that if the wind shift isnot present until after storms have begun, onemight suspect in such a case that the wind shift ismerely the gust front.

Since the nature of windshift lines tends to besomewhat obscure, it is difficult to generalizeabout them. Certainly, the analyst needs tomonitor them during an hourly analysis routine, inorder to check on their continuity, If they showsome tendency to dissipate during the diurnalcycle, the suggestion is that the windshift isrelated to topography in some fashion, whichindicates that such a windshift is probably notsignificant. If, on the other hand, the windshift islocated in a clearly defined pressure trough, eventhough no obvious temperature/dewpointdifferences are present, the indication is that thismay be an important surface boundary.

2. Pressure Troughs

Non-frontal pressure troughs are a commonphenomenon. For reasons discussed in III.D.2,they can have an important influence on winds.Perhaps the most commonly observed non-frontalpressure trough is that which develops in the leeof the Rocky Mountains, usually in the wake of alarge polar anticyclone. While there are manytextbook discussions of �Lee Side Trough�

development, (see Panofsky, 1964, p. 118, for asimple summary; Palmen and Newton, 1969, p,344ff, for an excellent detailed treatment) theseneed not concern the analyst.

Basically, trough development requires at leastmoderate westerly components across themountains, By the creation of a lee side region oflower pressure, a southerly wind regime isestablished over the Plains which acts to returnmoisture and warmer temperatures driven out byprior anticyclone passage. This process is anecessary precursor to the establishment ofdrylines (see III.B.5) and to �Lee Cyclogenesis�.The problem of lee cyclogenesis per se is beyondthe scope of this report and the reader isencouraged to examine the numerous referenceson the topic (e.g., Bleck, 1977; Tibaldi et al ,,1980; Chung et ol., 1976; Kasahara, 1966; Hage,1111; Klein, 1957; and Bolson, 1950).

Another common type of non-frontal pressuretrough appears in the polar airstream behind acold front, when the parent cyclone is well intothe process of occlusion. These troughs areapparently similar to what are sometimes knownas �polar lows� (see Reed, 1979; Harold andBrowning, 1969; Rasmussen, 1977). Several canappear in the northwesterly flow behind the lowpressure center (Fig. 3.26), and may be supportedby windshifts as well. It is common that small-scale �comma cloud� formations are associatedwith these secondary pressure troughs, and theycan be significant convective weather producers.As described earlier (see III.B.4 above), when apool of cold air aloft is situated near the center ofthe occluded system and sufficient heating at thesurface occurs, these secondary troughs canresult in severe thunderstorm activity (includingtornadoes, at times). The co-existence ofsecondary troughs accompanied by clouds(associated with upward motion) indicates thatsignificant deformation (see Doswell, 1982a,b)and frontogenesis may be occurring. Thus, aninitially non-frontal trough may develop into asecondary frontal system of significantproportions. In fact, the work of Johns (1982a,b)suggests that many of the northwest flow severeweather cases have origins in these secondarytroughs. Such developments can occur repeatedlyas the larger, overall occluded cyclone graduallydecays.

55

D. Pressure Change Analysis

1. Applications to Synoptic Analysis

It is not an exaggeration to say that an analysis ofthe changes in surface pressure is probably morevaluable than the pressure pattern itself. At SELS,automated plots of pressure changes over severaldifferent periods are available and analyzedroutinely. Synoptic analysis makes most use of the3- and 12-h changes, with which it is possible toobtain a pretty clear picture of the movement ofthe large-scale pressure systems. The relativestrengths, speed and direction of movement, andthe tendency for deepening or filling can also beevaluated. Naturally, some of the diurnal andsemidiurnal variations are included in theobserved pressure changes, but these are easilyaccounted for, at least in principle.

By having a history of the 12-h changes, at 6-hintervals, the broad pattern of synoptic-scalefeatures can be easily grasped. Further, significantlarge-scale events often are seen first by rapidchanges in the isallobaric fields, While the theoryinvolved in pressure changes is more thanadequately covered by textbooks (e.g., Saucier,1955; Hess, 1959 [p. 219ff]; Petterssen, 1956a

[Ch. 19]), some simple concepts are useful toreview.

A well-known equation, logically enough knownas the pressure tendency equation (see Panofsky,1964, p. 124ff), governs the change in pressure atany fixed point. Since pressure is simply theweight of the atmosphere per unit area above thepoint, it is obvious that pressure changes reflectthe vertically averaged effect of processesoccurring aloft. The tendency equation states thatthe change of pressure at a point is the combinedresult of: the vertically averaged horizontaldivergence, the vertically averaged horizontaladvection of mass, and the vertical motion (abovethe level of interest).

It is unfortunate that our ability to measure allthree of these influences is not adequate for adirect application of the tendency equation. Sincesurface pressure change is the result of smalldifferences among these three, poorly measuredvariables, which usually act partially tocompensate for one another, the tendencyequation is only of academic (or instructional)value.

However, the atmosphere has no difficulty solvingthe equation for itself and we see the result ofimbalances among the three effects as pressurechange patterns which can reveal features ofinterest to the analyst. For example, turning ourattention to the 3-h pressure changes, the locationof fronts may be clarified by the isallobaricpattern. The large temperature changesassociated with frontal. passage may not beclearly revealed at the surface (e.g., as inmountainous terrain) because of local ortopographic influences, but the pressuretendencies are less likely to be effected becausethey reflect changes through a deeper layer.

2. The Isallobaric Acceleration

Just as the pressure field can be used, via thetendency for winds to be in near-geostrophicbalance, to infer the basic wind flow pattern, thepattern of isallobars can be a valuable indicator ofaccelerations in the wind field. At times, thecontribution of isallobaric accelerations is themajor contribution to departures from geostrophic

Fig. 3.26. Schematic diagram of polar low andpolar trough.

56

balance (Brunt and Douglas, 1928 as referencedby Saucier, 1955 [p. 242]). Be aware, however,that this is not always the case, and othercontributions to the ageostrophic wind can beimportant. See the discussions in Saucier (1955, p.240ff or in Petterssen (1956a, Ch. 4). As Saucierstates, � .. it is improper to attribute existingageostrophic winds only to the isallobaricpattern....�

It is important to remember that the isallobaric�wind� is not a real wind (the geostrophic wind isnot a real wind either) � it is part of the totalacceleration which acts on air parcels (Hess,1959, p. 225ff). Specifically, it is that part createdby local changes in the pressure field. Like anyacceleration, it changes the velocity, but thosechanges do not take immediate affect. Rather,they operate over time to produce ageostrophicwinds. If isallobaric accelerations are the onlyeffect producing ageostrophic winds, theageostrophic component will oscillate (as shownby Hess) about a mean value which is theisallobaric wind, with a period of about 17 h (seeIII.F.3, below). It takes several hours for thisoscillation to develop and stabilize.

What all this means is that parcels movingthrough a region of pressure changes must beacted upon for a period of time sufficient toproduce a significant velocity change. This canoccur when parcels are moving (at least initially)slowly. Alternatively, the region of pressurechange can be large enough (in the direction ofparcel motion) that a fast-moving parcel remainsunder its influence for extended periods. Thisclearly implies a scale dependence of theisallobaric contribution to parcel velocity. Verysmall-scale, brief isallobaric patterns do not implyimmediate ageostrophic contributions equal to thecalculated isallobaric �wind�.

The acceleration induced by a changing pressurefield increases with the gradient of the isallobars,just as the geostrophic wind increases with thepressure gradient. For example, an isallobaricgradient of 1 mb per 3 hr over a distance of 100km (a large value) yields an isallobaric �wind� ofabout 10 m s-1. Unlike the geostrophic wind, theisallobaric wind is directed perpendicular to theisallobars (the geostrophic wind being directedparallel to the isobars, of course), and towardfalling tendencies. This is illustrated in Fig. 3.27,

showing how the isallobaric acceleration acts toturn the geostrophic wind toward pressure fallcenters.

The results of isallobaric accelerations can easilybe seen in the analysis process. One of thereasons for the strong ageostrophic flow nearlyperpendicular to the isobars behind a cold front isthe large contribution from the pressuretendencies. It is easy to visualize the effect, sinceit is quite clear intuitively that flow should tend torush into an area of falling pressure and out of aregion of rising pressure.

An important contribution of the isallobaricacceleration to severe weather situations isbacking of the winds ahead of an approachingtrough (or into a deepening stationary trough, likethe Lee Side Trough�). This backing of the flowcan be seen aloft, as well as at the surface, andmay serve to enhance low-level moisture influxand convergence. Note that this application of theisallobaric wind concept is on a relatively largescale, so that it is reasonable to expect isallobaricaccelerations to produce ageostrophic windswhich resemble the isallobaric wind.

Fig. 3.27. Illustration of isallobaric wind effect for atranslating low pressure center (adapted fromPetterssen, 1956a).

57

3. Mesoscale Isallobaric Analysis

For purposes of mesoanalysis (as we have usedthe term), the short term pressure changes are ofmaximum interest. In what follows, it shouldbecome clear that their main value for mesoscaleanalysis is to detect mesoscale pressure systems.White these pressure systems seem to beassociated frequently with severe thunderstorms,they are not completely understood. As suggestedabove, the analyst should not attempt to interpretthe isallobaric analysis in terms of the isallobaric�wind�, on this time and space scale.

Two-hourly pressure changes are routinelyanalyzed at SELS and it should be clear that inorder to make full use of these change fields, theyneed to be produced and analyzed hourly duringthose periods of greatest threat of severeconvection. Owing to its greater spatial coverage,one should use the altimeter setting (recall III.Aabove) for this purpose. An appropriate contouringinterval for isallobars is somewhat dependent onthe field�s extreme values, but a generallyacceptable pattern can be obtained with a 0.02inches of mercury (about 0.68 mb) per 2 hinterval. As mentioned several times previously,diurnal and semidiurnal effects are present.

Assuming the isallobaric analysis is done, theanalyst will likely have a complex picture to try tosort out. Some observations will be missing, somereported changes will represent errors, and somewill represent real atmospheric phenomena. Twoexamples showing 2-h altimeter setting changesare presented in Figs. 3.28 and 3.29. The exampleof Fig. 3.28 is one which provides an insight intothe large-scale changes, as well as the smallerscale. Note the broad area of rises, behind thefront in the Great Lakes. Most obvious, perhaps, isthe small-scale system in southeastern Nebraskaand northeastern Kansas. Also, there is an area ofconcentrated falls in Kansas ahead of the risesassociated with a �bubble� high in Nebraska.Considerable severe weather occurred inNebraska and Kansas in association with thismesosystem.

Lest one be led to assume that detectable 2-hpressure changes always accompany severeweather, consider Fig. 3.29. This is a goodexample of how substantial severe weather canoccur (in this case in central Oklahoma andsouthern Kansas) without any clear-cut 2-hpressure change features. Moller (1979) hasexamined 1-h altimeter changes for this case andhas been able to detect very small-scale, fastmoving pressure change couplets associated with

Fig. 3.28. Isallobaric analysis (hundredths of an inch of Hg, per 2 hr), at 1800 GMT, 31 May 1980.

58

the Oklahoma severe weather. If one uses 2-hchanges routinely, it may not be possible to detectsuch small systems.

Of the real phenomena, part of the pattern shouldreflect the large-scale pattern seen in the three-and twelve-hourly pressure change analysis.Embedded in this broad scale field occasionallywill be non-convective small scale phenomenawhich typically do not have the strength ofconvective isallobaric patterns. These may bereal atmospheric phenomena (gravity wavescome to mind as an example), but they generallydo not represent a significant factor inmesoanalysis. Although this suggests that theanalyst could safely ignore these, there existexamples (Uccellini, 1975; Eom, 1975), where ithas been argued that they have an exceedinglyimportant role in the convection. The key conceptin separating the wheat from the chaff here iscontinuity in space and time. Very small-scale,transient phenomena are indistinguishable fromnoise, even if their origins are in real atmosphericprocesses. Only those which affect severalobservation points an space and/or time are likelyto have some influence on mesoscale weather.

Further, the appearance of intermediate-scalepressure change structures can be a valuablecheck on the importance of satellite imagery. Ifamore or less shapeless cloud mass (as opposed toa �comma� cloud) is accompanied by a localizedregion of concentrated surface pressure falls,there is obviously reason to believe that it islinked to some dynamic �feature�. Should thistravelling feature be moving toward an area ofconvective instability, this corroboration of thesignificance of the cloud mass is of great value tothe forecaster/analyst. Recall the discussion inIII.D.5 about the effects of pressure falls on thewind field.

Many of the high-amplitude mesoscale pressurechanges associated with convection manifestthemselves only after convection has developed.As one might expect, an indication of thedevelopment of a substantial downdraft/outflow israpidly rising pressures. This pressure rise isinitially confined to a small area and is usuallyonly observed at one station, if at all. With time,the area covered by substantial pressure risesexpands, moving with the convection, Naturally,as this �bubble� moves, pressures in its wake may

then begin to fall rapidly. Thus, we have a rise/fallcouplet developed as a result of the outflow, withthe fall trailing the rises, just as the wake lowfollows the mesohigh. This is common even withrelatively non-severe storms.

In the severe weather situation, there usuallyexists a zone of concentrated pressure falls aheadof the mesohigh-associated rises. It is not entirelyclear that this is indicative of an undetectedmesolow. Hoxit et al. (1976) have proposed amechanism wherein the mesoscale verticalmotion fields induced by the convection (far awayfrom the convective drafts themselves) can lead tofalling pressures ahead of the storms. Whether ornot this is a valid suggestion in the majority ofsituations remains to be demonstrated, but it doesallow for the creation of a pressure fall centerwithout requiring a travelling mesolow to be itssource. Moller (1979) has observed mesoscalepressure falls, which he found to precede mostsevere thunderstorm outbreaks in the SouthernPlains. Magor (1971) has also noted the influenceof isallobaric accelerations associated with suchpressure falls, which act to advect heat andmoisture into area ahead of the advancingconvection.

Regardless of the origins of the falling pressuresahead of the area of rises associated with outflow,the mesoanalyst should be alert for the fall/risecouplet. If a wake low is trailing the bubble high,a fall/rise/fall triplet may result. Such a feature canalso evolve into a double couplet structure, shouldnew convection develop in the wake of the first.

Although the actual mesolow has received littleattention in these notes, there is no doubt thatsuch phenomena occur. The reader should becareful to distinguish between the �mesocyclone�(and accompanying low pressure area), which isdirectly related to the convection (seeII.III.A.5.b.(2)), and a pre-existing (or at leastlarger-scale) mesolow which apparently sets toenhance convection but is not directly tied to thestorm�s draft. Once we reach the scale of themesolow (which is from several tens of km to afew hundreds of km in diameter), we have movedinto an area of relatively little understanding. Ithas been suggested that at least some subsynopticcirculations have their origins in weak frontaldisturbances (Doswell, 1976). These systems areat the upper end of the size range for what loosely

59

can be grouped together under the heading ofmesolows. As we decrease the size of themesolow, correspondingly less is known abouttheir structure and origins. As Magor�s worksuggests, knowledge of mesolow existence isbased at times solely on localized pressure falls. Itis logical to suggest mesolow presence atintersecting boundaries, each of which generallylies in a trough of lower pressure. This isenhanced by the already described tendency forsevere convection to occur in the vicinity ofintersecting boundaries. However, Magor (1959)readily admits that �the meso-low [at theintersection of instability lines] would be suspectrather than directly observed.�

Certainly the literature includes some well-defined examples of mesolows, which have beendetected when they fortuitously passed through adata-rich region (Brooks, 1949; Mogil, 1975;Hales, 1980; Magor, 1958). From this one mightbe tempted to suggest their existence even whendata are not available to define them. However,such inferences are essentially unscientific. In asense, it is irrelevant whether or not one actuallyhas a mesolow present in a given situation. Themesolow is most important insofar as it canenhance the potential for severe weather. It isprobable but not yet proved that it does soprimarily by means of its ageostrophic

Fig. 3.29. As in Fig. 3.28 except at 2100 GMT, 8 June 1974.

60

accelerations of the flow. The isallobariccontribution to ageostrophic accelerations canfrequently be diagnosed from available data. Itshould be recalled that the isallobaric contributionis not necessarily the most important one.Naturally, when a mesolow can be detectedreliably, it is probably a significant feature.

Finally, the short time scales of most subsynopticpressure systems do not usually allow the windstime to react appreciably to Coriolis acceleration.Thus, one should anticipate that the wind flows inmesoscale pressure systems will not often be innear-geostrophic balance. Further, it is notuncommon for the circulation center ofsubsynoptic cyclones to be removed somedistance from the center of subsynoptic lowpressure systems. The analyst should constantlybe aware of the distinction between the actualand geostrophic flow, since departures fromgeostrophy are related to flow field accelerations(see IV.B).

E. Thermal Analysis

It has long been recognized that severethunderstorms, and especially tornadic storms, donot occur randomly within the warm, moist air ina large-scale extratropical cyclone�s warm sector.Rather, there are moisture and temperaturepatterns which have proven to be reliable asindicators and locators of severe storm potential.For example, when a thermal ridge axis is presentupstream (generally, to the west) from the axis ofmaximum moisture (dewpoint temperature), afavorable configuration is present. Convectionusually begins on the west side of the moist axis,between the thermal and moisture ridges. This isprobably the result of upward motion (owing towarm advection) encountering abundantmoisture. An example of this configuration isshown in Fig. 3.30. This is generally valid both forthe lower levels in upper-air analyses, say, at 850mb (Miller, 1972) and also at the surface (Moller,1979). Considerable attention has been focusedon the surface thermal ridge in the past (Kuhn etal., 1957; Darkow et al., 1958; Whiting, 1957), butrecently this awareness has waned somewhat(probably inappropriately).

The general picture of a thermal axis upstreamfrom the moisture axis in a severe weathersituation is certainly consistent with drylinestructures we have seen earlier (III.B.5). Since thedryline is a persistent feature in the SouthernPlains, Moller�s (1979) results for Southern Plainstornado outbreaks should not be surprising. Thatthe basic idea applies in a much greater area thanthat influenced by the surface dryline may besomewhat unexpected. Recall that considerableempirical and theoretical evidence suggests thatmost severe storms are accompanied by theintrusion of dry air aloft. If a dryline exists aloft, itseems clear that there is no essential differencebetween such situations and those involving asurface dryline. Note that the thermal, moisture,and windflow structures associated with bubblehighs can create a mesoscale version of thepattern we are discussing (recall Fig. 3.20). Thecommon occurrence of severe weather on theseold thunderstorm-created boundaries no doubtresults from the favorable thermodynamicstructures left behind by the convection (Maddox,et al., 1980), as a return to pre-storm conditionsproceeds.

Temperature and dewpoint analyses are useful forlocating air mass boundaries. The analyst shouldperform them before attempting to delineate thefeatures on the surface map. The details of thesurface temperature pattern are crucial to properlocation of thunderstorm outflow boundaries,which may not have clearly defined wind andpressure fields (i.e., the wind and pressurestructure may not resemble the classical patternsshown previously). As with larger scale analyses,the theta-w field can be very useful for thispurpose. Recall that the boundary between airmasses is, by convention, put on the warm side ofthe zone.

In any case, a careful analysis of the surfacetemperature (and dewpoint) pattern,4 keeping inmind the above concepts, is in order. As withpressure data, the short term (one- to two-hour)change patterns of temperature and dewpoint canbe valuable in locating the thermal ridge (andmoisture boundaries), as well as in making ashort-range forecast of its movement. Sincethunderstorms typically develop between thethermal ridge and the moisture axes, an accuratelocation and forecast of these features can be

61

critical. In many cases, isotherms at 5 deg Fintervals will suffice to locate the appropriateaxes and boundaries � in the warm season, itmay be valuable to do analysis at 2 deg Fintervals. At the same time, areas of thermal (andmoisture) advection can be diagnosed, noting thatadvection is an obviously important factor in localtemperature (and moisture) changes. As discussedin III.B, thermal and moisture changes (apart fromdiurnal effects) do not usually occurindependently of the pressure and windflowpatterns, so this aspect of analysis should be aroutine part of surface mesoanalysis.

F. Terrain Effects

Much of the climatology of severe weather canbe directly related to topographic features. It iscommonly asserted that the overall pattern of high

severe weather frequency in the central UnitedStates is a result of the combined effects of severallarge-scale topographic features (Carlson andLudlam, 1968). Without casting doubt on this basicidea, one should also be aware of limitations inthe climatological record (Kelly et al., 1978;Galway, 1977; Doswell, 1980; Snider, 1977).Simply because a region lacks a large number ofreported severe weather events is not sufficientevidence to infer that severe weather is rare inthat region. Similarly, the occurrence of. severethunderstorms without much reported tornadoactivity is not necessarily clear proof of theabsence of tornadoes.

The central plains region of the United States hasthe world�s highest observed concentration ofstrong-to-violent tornadic activity. The existenceof a source region for warm, moist air near anorth-south mountain chain and the lack ofsignificant east-west topographic barriers seemsideal for the creation of convectively unstableenvironments. The mountains act to wring outmuch of the moisture in the upper-level westerliesand the general subsidence of the flow to the leeof the barrier further reduces the humidity of thewesterly flow.

Dynamically, the preferred lee side location forcyclogenesis creates a strong tendency for low-level moisture to be drawn from the moisture-richair mass over the Gulf of Mexico into anydeveloping circulations. Further, the cool air tothe north has no restrictions to its southwardadvance, thereby creating a strong barocliniczone along which circulations can intensifyrapidly by drawing on the potential energy whichis the result of the thermal contrast.

It is precisely these factors which are classicallyseen as creating a favorable environment forsevere thunderstorms to develop. However, this isnot the whole story of the terrain�s influence.There can be small scale terrain-relatedphenomena which act to enhance the severeweather potential on a local scale. These are ofspecial interest to the analyst/forecaster.

1. Mountain/Valley Circulations

Physically, the mountain/valley circulationoriginates in a manner similar to the land/sea

Fig. 3.30. Surface temperature and dewpoint (solidand dotted lines, respectively; deg F) analysis at1900 GMT, 8 June 1974 (after Moller, 1979).Thermal and moisture axes are shown as dashedknee, while open circles locate surface observationsites.

62

breeze. Since the eastern slopes of mountainsface the sun more directly during the earlymorning (when the sun as at a low angle, they areheated more quickly. A rising plume of warmedair on the sun-facing side of the mountain results,with upslope flow developing to replace the risingair. During the afternoon, the sun is shadedsoonest on that eastern face, so the oppositeprocess eventually results in downslope flow.

As described by Defant (1951), this basic pictureis complicated by several factors. Figure 3.31shows the air currents in a valley adjacent to aregion of plains. The extremely complex terrainfeatures in any mountainous zone can complicatethis basic pattern, but the essence remains fairlysimple. A period of experience over perhaps atwo-year time span should familiarize the analystwith most of the persistent mountain/valleycirculations in the local area. There is bound to besome dependence on seasonal changes and, likeland/sea breezes, influences from the prevailingflow regime.

2. Upslope Flow

The trend for flow up- and downslope, as we havejust seen, is partially a diurnal effect. As shown byDirks (1969) in a numerical simulation, this basiccirculation can affect the flow over a large portionof the Rocky Mountain chain�s lee slopes. Inaddition to the upslope flow directly over theslopes, a compensating downward current wasfound by Dirks to exist from near the foot of theslopes, out over the plains for about 40 km. Fromthis point out to about 300 km, a weak upwardmotion regime could be found, wherever theplains are also slightly sloped. This shall bediscussed somewhat more fully in the nextsection. What is important for the analyst torealize is that the collective effect of the ratherabrupt transition from mountainous terrain togently sloping plains can be seen on a ratherlarge scale. Naturally, upslope flow produces anet upward vertical motion which is clearlyassociated with the frequent occurrence ofthunderstorms along the mountain ranges west ofthe plains. even in synoptically quiescentconditions. These thunderstorms can be purelylocal in nature, or may propagate into the plains

under the right synoptic conditions (George,1979).

It should be observed that the strength of thecontribution to upward motion by upslope flow isscale-dependent. Although the fine-scale detail ofthe terrain can create areas of locally largeupward motion, these details are smoothed out aswe shift our point of view upscale and the verticalmotion magnitudes are correspondingly reduced.On the large scale, a terrain-induced verticalmotion of several cm s-1 is associated withhorizontal slopes in the range of several km per1,000km ( for a horizontal wind speed of order 10m s-1). Similarly, for a mesoscale vertical motionvalue of several m s-1, the corresponding terraingradients are several km per 10 km, which isextremely fine scale topographic detail. SeeSchaefer (1973b) for some mathematical details ofthis aspect of low-level divergence in relation toterrain.

Under the appropriate circumstances, thesynoptic-scale pattern can enhance the diurnaltrend for upslope flow. Doswell (1980) has offeredthe idea that the majority of High Plains severethunderstorms occur under just such conditions.Following anticyclone passage to the east, intothe Great Lakes for example, an easterly low-level upslope flow regime can be establishedwhich augments the diurnal upslope tendencyand the influx of low-level moisture behind anormally-present quasistationary front (Fig. 3.32shows an example). Given appropriate upper-level flow and adequate instability, severethunderstorms frequently result, This surfacepattern can also be responsible for flash-floodproducing convection (with little or no severeweather) under weak flow regimes aloft (Maddoxet al., 1978).

It is important to the analyst that conditions ofupslope flow be recognized whenever the airbeing forced to rise over the terrain has adequatemoisture. Such conditions are not confined to theimmediate lee of the Rockies, as we shall see.

3. The Low-Level Jet

There are actually two somewhat differentphenomena involved in the well-recognized�low-level jet�.

63

The first of these is a nocturnal low-level windmaximum, distinct from the upper-level jet stream.This phenomenon is generally acknowledged tobe a result of the de-coupling of the surface fromthe flow above it, as an inversion is established atnight (Blackadar, 1957). This drastic reduction ofthe friction just above the atmospheric boundarylayer creates an oscillation in the wind at thatlevel about its equilibrium position (geostrophicbalance). The period of the oscillation (called ahalf-pendulum day which, at 45 deg N, is about17 h] is dependent on the local value for theCoriolis parameter, but the resulting flow speedgenerally reaches its maximum in the earlymorning hours (Fig. 3.33).

A second major phenomenon is the so-called low-level jet stream (the terminology here issomewhat confusing). This is a narrow ribbon ofhigh-speed flow, analogous to the jet stream aloft,

but restricted to low levels. An example is shownin Fig. 3.34. A variety of detailed theoreticalexplanations exist, but it is accepted generallythat an essential aspect in its creation is thesloping terrain, via a diurnal variation in thegeostrophic wind.

A basic element in establishing a classic low-leveljet stream is the large-scale synoptic setting.Following passage of an anticyclone through thecentral United States, a region of anticyclonicallycurved return flow is created, westward from theGulf of Mexico, and swinging northward into theplains. This current usually is augmented byestablishment of a lee side trough (see III.C.2above) and can evolve slowly over periods of upto several days (Djuric and Damiani, 1980). Bythis means, a favorable environment for severethunderstorms is re-established by low-level warmadvection and a return of adequate moisture.

Fig. 3.31. Schematicillustration of the normaldiurnal variations of the aircurrents in a valley,beginning at sunrise (a), andat about 3 h intervalsthrough early morning (b-h)[after Defant, 1951].

64

Fig. 3.32. Example of upslopeflow in the cold side of a surfacefront, leading to a High Plainssevere weather episode (afterDoswell, 1980).

Fig. 3.33. Diurnalvariation of windcomponents from theirdaily mean value at FortWorth, Texas; values inm s-1 (after Bonner andPaegle, 1970).

65

Since the low-level jet stream is a means ofenhancing the destabilization of the environment,it is of fundamental importance to the severethunderstorm analysis/forecast problem. Uccelliniand Johnson (1979) have presented evidencesuggesting that the upper-level jet streak (a localmaximum within the jet stream) and the low-leveljet stream are coupled. This seems to be areasonable hypothesis � one needs to considermore than boundary layer processes to developan adequate diagnosis of the low-level jet stream.

Having the appropriate synoptic environment,processes in the boundary layer (the lowest 1 to 2

km) enhance the tendency for development of thelow-level jet stream. The speed maximum may beless than 1 km above the surface and the width ofthe zone of strong winds is generally severalhundred km (see Fig. 3.35). Speeds in a well-developed low-level jet stream are in the range of20-30 m s-1.

In addition to the frictional changes, giving rise tothe nocturnal boundary layer wind maximum, thesloping terrain forces a modification of thepressure gradients. This change is not alwaysapparent in the �sea-level pressure� field since itdoes not properly account for terrain slope. Thealtimeter correction method of diagnosing the

Fig. 3.34. Time evolution of a lowlevel jet stream (after Djuric andDamiani, 1880). Winds shown areat the level of maximum wind inthe lowest 2 km, isotachs are in ms-1 while relative humidities over80% are stippled. The dashed linelocates the ridge at 850 mb.

66

geostrophic wind (Bellamy, 1945) hasbeen incorporated in the analysis ofsurface geostrophic winds by Sangster(1960). When this is done, it is foundthat some portion of the diurnal windvariation can be accounted for by thevariations in pressure gradient.Sangster�s method is examined furtherin III.F.3 and IV.B.

Numerical models (e.g., Bonner andPaegle, 1970; Chang, 1976) have beenformulated which incorporateboundary layer processes. Thesetheoretical models have generallysucceeded in reproducing the basicfeatures of the low-level jet. The meansby which these phenomena interactwith the large-scale setting to producea concentrated jet stream are notcompletely understood. Considerationsof potential vorticity (beyond the scopeof these notes) suggest that the slopingterrain forces an increasing southerlycomponent as the flow off the Gulf ofMexico encounters higher elevations.Also, when the geostrophic wind isparallel to the terrain, the frictional partof the ageostrophic wind is therebydirected upslope. This upslope flowmay interact with the diurnal variationsto create a concentrated core ofsoutherly flow (Chang, 1976; Schaeferet al., 1982).

Although its destabilizing influencesvia advection often are emphasized,the low-level jet stream also isassociated with fields of verticalmotion. While it is not as well-understood as the vertical motion fieldof the upper-level jet stream (seeMcNulty, 1978; Beebe and Bates,1955), the low-level jet stream�svertical motion field is a likelyexplanation for the nocturnalthunderstorms which frequent thecentral plains of the United States(Sangster, 1958; Pitchford and London,1962; Wallace, 1975). This can beeasily seen in the pioneering work byMeans (1944, 1952, 1954) which

Fig. 3.35. Isotachs of southerly wind component (m s-1)along the line from Amarillo, Texas (AMA) to Little Rock,Arkansas (LIT) during the day of 28 May 1961 (afterHoecker, 1963). Jet cores are indicated by “J” with othermaxima and minima by an “H” or an “L”, respectively.

67

reveals a distinct nocturnal peak in low-levelwarm advection, obviously related to the low-level jet. If one recalls that warm advection isdirectly related to upward vertical motion,5 aconnection between the low-level jet stream andnocturnal thunderstorms can be seen readily.

The space and time scales of this phenomenonare large enough that it often can be detected andmonitored. The core of maximum winds usuallycan be seen easily in the 1200 GMT 850 mbanalysis. The normal nocturnal inversion preventsits easy diagnosis in the surface wind field untilabout 1500 GMT, after which the low-level windmaxima can be monitored via surface analysisthrough the remainder of. the day. Satelliteimagery can also be useful in this regard, sincethe cumulus field can often be tracked(subjectively) and may provide clues about thelow-level winds when animated loopingcapability is available. Further, the advection ofmoisture may be seen and tracked via the infraredimagery: the so-called �dark stratus� (e.g.,Parmenter, 1976; Gurka, 1980). This phenomenonresults from the creation of stratiform clouds andfog in the regions of high low-level moisture.Since these clouds are low, their infrared (IR)emission temperature is relatively high, comparedto surrounding clear areas. The clear, dry areasradiate more effectively and become cooler(brighter, in the gray scale used for IR satelliteimages) than the regions of high moisture andclouds (see Fig. 3.36). While clouds and/or fog areusually present, the effect can be produced by themoisture content alone, perhaps somewhat lessdramatically.

4. Mesoscale Eddies

On occasion, the winds and terrain can interact tocreate mesoscale vortices. One fairly frequentobservation of these is unrelated to convection �the wake vortex phenomenon. Zimmerman(1969), Chopra and Hubert (1965), Hubert andKrueger (1962) and others have prescribed theoccurrence of long �vortex streets� which can beseen in stratiform clouds downstream from anisland. This process is well-recognized in fluidmechanics (Milne-Thomson, 1968, p. 377ff) andtakes the form of a series of counterrotating

vortices, which are shed into the wake of a flowpast an obstacle (Fig. 3.37). These mesoscaleeddies are typically observed in convectivelystable environments, and are not consideredsignificant in severe weather. Fujita and Grandoso(1968) have proposed that splitting thunderstorms(see II.III.A.5.b.(2)) may be the result of. thecreation of counterrotating vortices in the wake ofa blocking updraft. While this is an attractivehypothesis, it is not generally accepted as theprimary mechanism for storm splitting, nor is itterrain-associated.

Another example of a vortex which can beterrain-induced is that which can occur when theflow encounters a �corner�. This phenomenon isalso recognized in fluid mechanics (Prandtl andTietjens, 1934, p. 217ff). It has been observed tooccur, but it is uncertain how frequently this effecthas an influence on severe convection. Onepossible example of this might be found in a studyby Reed (1980). Once again, Reed�s case is non-convective in nature, but it involves a terrain-related small scale cyclone which was associatedwith damaging surface winds.

Recently, Johnston (1978, 1982) has documentedthe occurrence of what he calls Meso-scaleVorticity Centers (MVC). These appear during and

Fig. 3.36. Unenhanced infrared satellite image,showing the intrusion of moisture into easternKansas and western Missouri by means of “darkstratus” (arrows).

68

after the dissipation of a Mesoscale ConvectiveComplex (MCC � see Maddox, 1980b) ascyclonic circulations in mid-level clouds. Theyhave been observed to persist for long periods(from several hours, to more than one day) andoccasionally can serve to initiate new convectionunder certain conditions. Their reflection insurface data is rather subtle and they may notappear at the surface at all. Those which serve todevelop new convection (less than 50% of thosesampled) usually have a pronounced low-levelconvergence boundary and move into regions ofunstably stratified air. They seem to be induced bythe MCC rather than being the result of terraineffects.

It is certainly possible that mesoscale vorticesmight provide the initial disturbances in anenvironment where such a disturbance couldintensify by other mechanisms. Once a mesolowhas developers, from these relatively obscureorigins in a terrain-related or convectively-induced disturbance, convection may be forcedby a larger system. The connection between suchvortices and severe weather is obscure, but theanalyst needs to be aware that such featurescould influence convection.

5. Miscellaneous Examples

a. The Black Hills Region

In contrast to the exceedingly complex featuresassociated with the Rocky Mountains, the BlackHills area of southwestern South Dakota nearRapid City is a relatively isolated, compact regionof uplands. As may be seen in high-resolutionsatellite images (Fig. 3.38), the region is clearlydarker than surrounding plains, making its namequite appropriate. Initial thunderstormdevelopment commonly occurs in this area (Kuoand Orville, 1973). There are two possiblecontributing factors to the tendency forconvection develop earlier in the vicinity of theBlack Hills than in the surrounding region.

First, the rather abrupt transition from rollingplains to a region of uplands suggests that low-level flow impinging on the hills will he forced torise. This may allow earlier breakthrough of thecapping inversion than in the relatively uniformsurrounding terrain. This feature has beenexploited in the numerical modelling efforts ofOrville (e.g., Orville and Sloan, 1970), to providethe initial impulse for the modelled convection. Asmay be clear from our earlier discussion ofmountain/valley circulations, there is a naturaltendency for upslope motion over at least theeastern portion of the Black Hills during theafternoon. This could augment any large-scaleupslope flow dictated by the synoptic pattern.

A second contributor to early convection is thedarker terrain, which should result in a localtemperature anomaly. This would be like the�sea-breeze� around a small island in theafternoon, with a tendency for low-level flow toconverge into the Black Hills from all directions.This effect would be distinct from the asymmetricmountain/valley flow, and should certainly be anenhancing factor in the development ofconvection.

b. The Caprock Escarpment

Much like the Black Hills, the CaprockEscarpment of west Texas combines several

Fig. 3.37. Vortex trail revealed in wake ofGuadalupe Island (arrow) in layer of stratocumulusclouds.

69

terrain-related phenomena in creating afavorable situation for convective storms (underthe proper synoptic scale setting). This terrainfeature is an extensive plateau, rising abruptlyfrom the plains to its east, about 100 to 300 m ina horizontal distance of a few km, along aroughly north-south line. Figure 3.39 shows thisfeature. Note that these terrain slopes yield amesoscale upward motion of from l0 to 50 cm s-1

with an upslope wind component of 10 m s-1.Further, the escarpment is cut by severalcanyons, with two major canyons being theWhite River Canyon to the east of Lubbock,extending southeastward, and the Palo DuroCanyon to the south of Amarillo (associated witha major fork of the Red River), which opens tothe east-southeast (see Fig. 3.39a).

If this structure is considered, one can see thatthe escarpment rises relatively sharply from aregion of gently sloped terrain, as do the RockyMountains further north. Although the barrier iscertainly much less imposing than the Rockies,the same basic diurnal tendency for upslopeflow and low-level wind structures exists in thisregion. In addition to the diurnal upslope effectover the plains to the east, the escarpment

provides an additional mechanical lift, since itforces easterly low-level flow upward and alsoacts as an early heat source, just as the eastslopes of mountains do.

Furthermore, under conditions of southeasterly toeasterly low-level flow, moisture can bechannelled into the canyons (Fig. 3.39b) and beforced upward into relatively drier surroundingsby the rising canyon floor. This establishes thearea over canyons as a favored location for storminitiation or intensification, via forced lifting andlocally enhanced moisture. Fankhauser (1971)and Marshall (1980) suggest that the local stormclimatology reflects this preference.

Another role played by the canyon has beendescribed by Marshall (1980). The south-facingnorthern walls of the canyons (Fig. 3.39a) receivemore direct sunlight (and, hence, �heating) duringthe day than do the canyon floor and north-facingwalls. This is analogous to the east-facing slopesof the escarpment which received heating earlierin the morning. Thus, these walls should also beassociated with a rising plume of heated air. Thismay be augmented under southerly flowconditions at low levels, which would tend tofollow the terrain and be lifted along the northerncanyon walls. As Marshall (1980) notes, thecombined effects of the escarpment�s topographicfeatures is reflected in the area�s precipitationclimatology. It is noteworthy that, despite arelatively low population density, the area of theCaprock is characterized by a well-definedtornado frequency maximum (Kelly et al., 1978).

c. Urban �Terrain�

Recently, it has been recognized that large urbanareas have an effect on convection (Changnon,1978; Braham and Dungey, 1978). With largemetropolitan populations, there is enough industryand construction to have a demonstrablemeteorological impact. This is generally seen asthe combined influences of pollution and heatretention, although the details of the relativecontributions by these effects are not completelyresolved. Nevertheless, the data collected byBraham and Dungey show an unmistakableconcentration of radar �first echoes� in the area

Fig. 3.38. Visible satellite image shying isolatedthunderstorm complex developing over darkterrain of Black Hills area in southwestern SouthDakota (arrow).

70

Fig. 3.39a. Physiographicmap of west Texas showingthe character of the CaprockEscarpment. The abrupt riseruns basically N-S from theCanadian River Valley (northof Amarillo) to well south ofLubbock. The “LlanoEstacado” (Staked Plains),which forms the surface ofthe Caprock, is somewhatmore resistant to erosion thanthe Gypsum Plains to theeast. Observe the majorcanyons which trend WNW-ESE and the especially deepcut into the Caprock at thePalo Duro Canyon (see textfor discussion) S and E ofAmarillo. Note that theCaprock also has a westernescarpment in New Mexico(although it is not as abrupt),part of which can be seen inthis figure WSW of Amarillo.

Fig. 3.39b. Illustration of theeffect of the CaprockEscarpment and the Palo DuroCanyon. This unenhancedinfrared image at 1000 GMT,15 April 1982 shows the “darkstratus” associated with low-level moisture (see Fig. 3.36)backed up along the Caprockand, especially, into the PaloDuro Canyon (arrow). Comparewith Fig. 3.39a.

71

adjacent to and downwind from a largemetropolitan complex. Although Changnon�stornado data were too limited to make any clearconclusions, the hail and strong wind reports (aswell as damaging lightning strikes) were sufficientto indicate that severe thunderstorm frequency isalso enhanced by urban effects.

In severe weather climatology, urban areasprovide an increase in severe thunderstormreports merely by virtue of an increasedpopulation density. The carefully controlledexperiments which provided data for Changnon�sand Braham and Dungey�s reports do indicate thatsome of this enhanced severe storm frequency isreally meteorological, rather than a reportinganomaly.

Further, the tendency for enhanced convectiveactivity may be present well downstream from theurban complex (Changnon, 1980). The evidenceaccumulated for the �La Porte anomaly� suggeststhat a variety of climatological factors, as well asmicroscale influences can create a conditionwherein an isolated maximum in convection isestablished quite a distance downstream from theinitiating urban complex. This maximum can shiftin location and strength as a result of long-termchanges in weather patterns.

This tendency for large urban areas to influencesevere weather (and convective rainfall) isdifficult to assess on a day-to-day basis. Thenatural tendency of the analyst/forecaster toignore these effects in the face of active weathersystems is understandable. However, theevidence indicates that an effort should be madeto incorporate these findings into the forecast,when large metropolitan areas are involved.

G. Flash Floods and Severe Weather

While these notes have emphasized the analysisproblem with respect to severe weather,convective flash flooding is certainly aspotentially dangerous as any aspect ofthunderstorms. Many of the analytical toolsdeveloped here for severe thunderstorms can hedirectly applied (shifting the emphasis inparameters somewhat, of course) to the

convective flash flood problem (as done, forexample, by Hales, 1977). This is a reflection oftwo related aspects of thunderstorms - first, it isnot always obvious how to distinguish thosesituations which will produce severe weatherfrom those which are predominately heavy rainproducers. Second, the same thunderstorms whichproduce severe weather phenomena are alsocapable of copious rainfall, and vice versa.

To start with, Maddox et al. (1979) have providedan excellent summary of the features common toflash flood events. These are:

(1) Flash floods are associated with convectivestorms.

(2) Storms occur in regions with high surfacedewpoint temperatures.

(3) Relatively high moisture contents are presentthrough a deep tropospheric layer.

(4) Weak to moderate vertical shear of thehorizontal wind is present through the clouddepth.

(5) Convective storms and/or cells repeatedly formand move over the same area.

(6) A weak, mid-tropospheric, mesoscale troughhelps to trigger and focus the storms.

(7) The storm area is very near the mid-tropospheric, large-scale ridge position.

(8) Storms often occur during night-time hours.

With the possible exception of points (3), (4), and(7) these findings could apply equally well tosevere thunderstorms. Further, there is ampleevidence to indicate that severe thunderstormshave occurred, even when (3), (4), and (7) arevalid. Thus, while there is a slight shift inemphasis on certain features of the environmentwhen the analyst/forecaster is considering heavyprecipitation potential, it is not always possible tobe confident that severe weather is unlikely. Theproblem of dealing with convective weather in anoperational environment is compounded whenboth severe weather and flash floods areoccurring (see Maddox and Dietrich, 1981). Thisis also true for tropical cyclones, which arepredominantly heavy precipitation producers.While hail is not often observed with tropicalstorms, tornadoes are certainly not rare (Novlan

72

and Gray, 1974; Smith, 1965; Pearson andSadowski, 1965). Tropical cyclone-associatedtornadoes are still not very well understood, but itis now recognized that they are more commonthan formerly thought (Gentry, 1982).

Second, it should be clear that a large, long-lasting severe thunderstorm ingests largequantities of water vapor during its life cycle(estimated as high as 8000 metric tons persecond). This will be discussed in more detail inII.III.C, but it is adequate at this point to observethat many convective flash flood situations alsoinvolve severe weather, and vice versa. It is anatural consequence of the processes whichproduce severe weather (not all of which are wellunderstood, of course) that heavy precipitationwill often accompany severe phenomena.

CHAPTER III FOOTNOTES

1 III.A: It is often mistakenly believed thataltimeter setting is a station pressure value. This isnot so; at most sites. the altimeter setting is readdirectly from an aneroid barometer - it is alreadya �sea level� pressure. The reduction is via thestandard atmosphere. If one needs the stationpressure, it can be found by inverting thereduction equation (see List, 1966), using altimetersetting. The so-called �Sea Level Pressure�cannot be used for finding station pressure, sincethere are �corrections� made in the reductionprocess which are difficult to reconstruct. Notethat altimeter setting can be dangerous to use inregions of mountainous terrain.

2 III.B.5: The reader should observe carefullyseveral features in this example (Fig. 3.12),especially in regard to mixed layer models (e.g.,Keyser and Anthes, 1977). There are extensivechanges below about 650 mb from the 1200 GMTsounding to that at 0000 GMT, mostly in the �well-mixed layer�. Observe the shallow superadiabatic�contact� layer just above the surface � this is acommon feature of evening soundings in the dryair. Also, note that the afternoon well-mixed layerhas a value of theta (potential temperature) verynearly that of the average theta seen in themorning, lending credence to the notion that nochange of air mass is involved. The slight increase

in moisture within the mixed layer during the dayis not significant for our purposes here.

3 III.B.7: It is, perhaps, worthwhile to describe themajor weak points in time-to-space conversion,since it has been so widely applied in severestorms research. Naturally, the whole processlends itself much more readily to post-storm(rather than real-time) analysis. However, thereare two more rather significant objections to theprocess. The first is that the �event� which istracked to provide a conversion vector (a line,along which to plot successive observations) mustbe assumed to be steady-state. While it can beargued that severe storms are essentially steady-state, there is a growing acceptance that mostsevere storms are continuously evolving and thedetails of that evolution are crucial to theproduction of severe weather (Lemon andDoswell, 1979). A second major objection centersaround the choice of a time-to-space conversionvector. It is typical to apply the same vectoreverywhere within the analysis region. It is notobvious that this should be the case, sinceindividual storms, squall lines, fronts, etc.,typically are characterized by different motionsand to apply any given vector to all the surfaceobservations is of questionable validity. Note thatHolle and Naier (1980) have used two motionvectors to account for the motions of two separateoutflow boundaries. If a field of vectors is allowedfor, then the problem can become complicatedbeyond any hope of a plausible solution. A thirdobjection can also be made: under certaincircumstances, it can be shown that time-to-spaceconversion may distort the field erroneously � ineffect, by creating nonlinear-appearingboundaries from ones which are in fact, linear. Itcan be argued that it is safer to �leave theobservations in the place where they weremade�. They still may influence later analyses,but ought not to dictate the structure by rigorousapplication along the conversion vector(s).

4 III.E: It is easy to argue in favor of replacingtemperature/dewpoint analyses with potentialtemperature (theta) and mixing ratio versions. Theconversion is simple, but these notes will continueto use more conventional fields.

5 III.F.3: Based on scaling arguments, it ispossible to argue that the connection betweenwarm advection and upward motion (rooted as it

73

is on quasigeostrophic theory) does not exist forlow level jet streams. This is not necessarily thecase. First, if is certainly plausible to suggest thatparts of quasigeostrophic reasoning may wellapply to flows where the whole scaling argumentdoes not apply. Second, and more importantly, foradiabatic flow there is certainly isentropic upliftassociated with warm advection.

74

IV. Objective Analysis Tools

Further, the plotting process for localmesoanalysis ideally should be done by machine.The apparatus to do so is in the AFOS system, andthis capability should be expanded to include allavailable observations, as well as the capabilityto plot �change� variables (like short-termpressure changes). These programs are currentlyavailable at SELS. with the discussion of changevariables having been included above.

Finally, having determined the overall pattern,there are many derived parameters which requirefar too much computation to be accomplished byhand. One can use the data to calculatedivergence, vorticity, streamlines, andgeostrophic winds, to name a few of the host ofpotentially valuable parameters. It is simply notpossible to do this quantitatively with the eye, noris it practical to compute them laboriously byhand. Meteorological literature abounds withparameters which some one has felt can make acontribution to analysis. Several of theseapproaches have already been included in theproducts routinely developed in SELS, and theseare to be described here.

A. Moisture Convergence

Two of the primary factors in developing severeweather potential are low-level convergence anda supply of moisture. These may be combined inthe field of moisture convergence (more properly,the moisture flux convergence). This fieldcombines the influences of convergence andmoisture advection. Thus, the divergence (amathematical operation usually to beaccomplished with data interpolated to a grid) ofthe product of the wind and some measure ofmoisture (e.g., mixing ratio) is computed. This canbe done at any level in the atmosphere, but thesurface data density and frequency make it themost-often chosen level. An example of the SELS

While the overall emphasis in these notes istoward an analysis which is done by the analyst/forecaster, there are certain facets of the analysiswhich are more easily and precisely done via thecomputer. In general, machine interpolation hasbeen overemphasized as a replacement forhuman analysis. The computer can draw linesbeautifully and the pleasing appearance of theresult, coupled with the assumption that�objective is always best�, has led to a nearlyuniversal replacement of hand-drawn operationalcharts with machine-produced versions.

The key factor in evaluating this trend is thegeneral lack of a distinction between the terms�analysis� and �interpolation�. Analysis is theprocess of developing the 4-dimensionalunderstanding of atmospheric events described inI.A. To the extent that machine-drawn isoplethscan facilitate the process of analysis, which is theprovince of the human analyst/forecaster, thecomputer is a valuable tool. When computer-based interpolation is used as an excuse toeliminate analysis by humans, a damagingprecedent is established. This latter concept doesnot represent the correct �man-machine mix� andhas led directly to �meteorological cancer� asdescribed by Snellman (1977).

What, then, is the appropriate role for computerapplications in the analysis process? Perhaps themost obvious is in the realm of large-scaleanalysis. The sheer volume of charts to be drawnindicates that much of the preliminaryinterpolation should be done by machines. Theneed to establish a smoothed representation of thelarge, synoptic-scale pattern should be apparent.In order to grasp the whole setting, upon whichare superimposed the subsynoptic scale weathersystems, one must necessarily examine the dataon the large scale (e.g., all of North America, orthe whole Northern Hemisphere) and smooth outthe details (which are to be examined during theanalysis phase). This is clearly most easilyaccomplished objectively.

75

version of moisture convergence (available onAFOS) is shown in Fig. 4. l.

A wealth of literature supports the basic idea(Hudson, 1971; Ostby, 1975; Doswell, 1977;Ulanski and Garstang, 1978) that this is avaluable parameter. In general, surface moistureconvergence precedes the development ofconvection, allowing this diagnostic field to haveshort-range prognostic value. This makes physicalsense, in that once moisture convergence begins,it should take a few hours to break through thecapping inversion (normally present) and toaccumulate a supply of moist air upon which thestorms will draw.

It is worthwhile to consider some of the limitationsin applying moisture convergence computationsto the analysis. First, the strength of moistureconvergence is highly scale-dependent. On thesynoptic scale, values are generally in the rangeof 10-4 g kg-1 s-1. As pointed out by Fritsch (1975)(see also Fritsch et al., 1976), this rate of moistureconvergence can supply only about 20% of thatneeded to sustain a severe thunderstorm complex.Thus, on the subsynoptic (or meso-alpha) scale,there must be a large increase in moistureconvergence (a factor of 5 or more) to maintain a

long-lived storm complex. Doswell�s (1977)results substantiate that this increase does occur.Since the strength of moisture convergence isscale-dependent, the availability of datadetermines, to some extent, the resulting field.Clearly, in some situations, the detail necessary todiagnose the moisture convergence properly maynot be available, and the computed field may notbe representative of what the threat area isactually experiencing. The analyst should bewary of computed values in data void (or missingdata) areas.

Further, it is not always the case that conditions atthe surface level reflect a clear picture of what isoccurring. When an inversion is present, the rootsof storm updrafts (and the zone of precedingmoisture convergence) may be aloft. This is oftentrue for nocturnal thunderstorm situations. In suchcases, the surface moisture convergence may becompletely unrelated to the weather, sincesurface-based sensors do not detect the physicallysignificant events.

A problem related to the discussion in III.B.1 is thatthe surface wind observations are the primaryfactor in determination of moisture convergence.Errors and unrepresentative observations mayproduce completely fictitious centers of moisture

Fig. 4.1. Example ofsurface moistureconvergence values atgrid points produced atNSSFC and displayed onAFOS console.

76

convergence. These can usually be easilydetected by the associated �bulls-eye� patterns,but the truly representative value may be lostirretrievably. A carefully thought-out screeningprocedure for the input data is essential. Asdiscussed in III.A, this may not be entirelystraightforward, since the observations may beboth real and actually representative in somesense. Having the plotted input data available forexamination alongside the resulting moistureconvergence field is a way for the analyst toarrive at an assessment of the quality of the result.

Since moisture convergence usually includes boththe product of moisture with convergence and theadvection of moisture, this parameter may besomewhat misleading. It may be useful to have aseparate measure of convergence alone(normally, the dominant term). If one has this, it ispossible to keep track of travelling convergencecenters which may be distinct from the moistureconvergence pattern. That is, the moistureadvection may combine with the convergencecontribution to yield a relatively slow-movingmoisture convergence field � yet what one reallyhas is a moving convergence field intruding into azone of strong moisture advection. The distinctioncould be significant.

Finally, satellite imagery can be a usefulsupplementary tool in checking on the moistureconvergence field. Areas of enhanced cumuliformcloudiness frequently precede the onset of deepconvection. As these develop and evolve, theyshould be compared with the moistureconvergence field. It may be possible to locatemoisture convergence zones in this mannerwhich are not detected by conventional data.Small scale features in the moisture convergencecan often be easily seen in the satellite data �e.g., the lines of cumulus congestus along �arcclouds� and fronts, compared to large areas ofordinary cumulus clouds.

B. Surface Geostrophic Winds

The production and application of surface data-generated geostrophic wind charts has beenpioneered by Sangster (1960). An increasingacceptance of the value of this approach can beseen in its current availability via teletypewriterand its incorporation in the AFOS products. Anexample of the AFOS-available surfacegeostrophic wind field is given in Fig. 4.2.

Fig. 4.2. Example ofsurface geostrophicwind field produced atNSSFC and displayedon AFOS console.

77

Therefore, it is important to emphasize that thisevaluation of geostrophic wind is not astraightforward application of the well-knowngeostrophic wind law to the conventional �sea-level pressure� field. Rather, it uses the so-calledaltimeter correction system (Bellamy, 1945) toincorporate the earth�s surface topography. Thedetails of the derivation are not important here,but the resulting product normally differssomewhat from what one might expect by lookingat an NMC surface pressure analysis. This ispartially caused by the influence of topography,and partially by the process by which a solution isobtained (a high degree of smoothing is done).

One of the benefits of Sangster�s approach is thatthe diurnal variation in the geostrophic wind isdirectly incorporated. As described in III.F.3, thisgeostrophic wind variation is often reflected in thereal winds and plays a role in the low-level jetstream (Sangster, 1967). This diurnal windvariation,in turn, is an important factor inthunderstorms (Means, 1952; Sangster, 1958;Pitchford and London, 1962; Bonner, 1966].

Although the geostrophic wind is essentially non-divergent, this chart can be used effectively inconvective forecasting. The zone of upwardmotion associated with the low-level jet liesgenerally to the left of the jet axis. Exceptions tothis are either when the jet impinges on aboundary, or the calculated geostrophic speeddecreases rapidly at the �nose� of the jet. Thesesituations imply upward motion ahead of themaximum wind core along the axis, animplication which is generally substantiated inpractice. This, as we have seen, is generallyassociated with strong warm advection.

Another advantage of the surface geostrophicwind chart and its main advantage over either thesurface pressure map or the observed winds is itscontinuity. The changes in the field are relativelyslow to occur and they accurately reflect theoverall march of events. Observed windsfluctuate substantially and this makes it difficult tomonitor the actual time evolution and movementof the low-level jet stream.

It is noteworthy that the surface geostrophic windsdo not always relate clearly to the observedwinds. It is obvious that much of the time, theobserved winds are much slower than

geostrophic. This is a natural consequence ofsurface friction. Schaefer and Doswell (1980)have recently incorporated surface friction intothe force balance in an objective way, producingfields of the so-called antitriptic wind. This has notyet been implemented on an operational basis,but the suggestion is that by incorporating friction,a theoretical wind is obtained which is moreappropriate at the surface than the geostrophic.This effort has been motivated in part, by theproblem of inferring accelerations from thedifference between observed and geostrophicwinds.

The concept of ageostrophic acceleration isimportant to understand, and the reader shouldconsult textbooks (e.g., Saucier, 1955, p. 240ff) fora more thorough discussion. Briefly stated, whenthe wind is not geostrophic, accelerations existwhich act to turn and change the speed of theflow, in an effort to reach the balancedgeostrophic equilibrium state. These accelerationsare critical to understanding weather since, as wehave discussed, the divergence (and, henceupward motion) associated with geostrophic flowis not physically significant.1 As Schaefer andDoswell (1980) have pointed out, the frictionalcontribution to the ageostrophic wind is large atthe surface. Thus, a force balance ignoring frictionis just not adequate. In order adequately todiagnose significant accelerations, one must firstaccount for the friction � what remains in theway of �non-antitriptic� winds (analogous toageostrophic) is then more likely to be physicallyimportant.

Needless to say, since the antitriptic wind isslower than and directed to the left of thegeostrophic, should one encounter observedsurface winds in excess of geostrophic, then(assuming it is not an error) something significantcertainly is occurring. The equations governingmotion state that supergeostrophic flow isaccelerated to the right of the geostrophic wind. Amoment�s thought should reveal to the reader thatthis is toward high pressure. Similarly,subgeostrophic winds are shunted toward lowpressure (as may readily be concluded fromexamination of most surface maps).Supergeostrophic (or perhaps more appropriately,superantitriptic) winds are an indication thatimportant events are underway and need to bemonitored.

78

C. Filtering by Objective Interpolation

Although the primary surface data analysisresponsibility lies with the analyst/forecaster,there are areas where an objective interpolationof the primary fields can provide new insights.Specifically, it has been shown that carefullydesigned objective interpolation schemes(Doswell, 1977; Maddox, 1980a) can be used toseparate mesoscale features from the large-scalepattern. When using surface data, Doswell (1977)has included time series data (without goingthrough the time-to-space conversion), in order tosimulate the time continuity that subjectiveanalysts can impose. This has a variety ofbeneficial effects, including a reduction of thetendency that most objective interpolationtechniques have to place extreme valuesbetween observing sites. It also limits the hour-to-hour �jumps� in the location and strength of theextrema.

The advantage this objective interpolationscheme provides the analyst is that it helps toisolate and enhance those mesoscale features thatare truly supported by the observations. Further, itcan be used to develop the derived parameters,such as divergence and vorticity, associated withthose mesoscale phenomena. This techniqueprobably comes the closest to reproducing whatthe human can do. By providing the opportunity toexamine derived parameters at the small-scalelimits of the data, a substantial benefit is gained. Itcan also be an important analysis aid to the lessexperienced individual.

An obvious disadvantage is the introduction of anadditional processing step which requiressubstantial on-site computer capability. The speedand timeliness of this analysis aid is directlyproportional to that capability, which for the timebeing makes it impractical for universalapplication. In order to compensate for the timedelay, an objective analysis tool should providethe analyst with some parameter or insightotherwise unavailable.

D. Upper-Level Divergence

Having discussed low-level moistureconvergence analysis in IV.A, some mention ofthe required upper-level divergence is necessary.As in the case of low-level data, it is simply notpossible to diagnose upper divergence by eye. Itis possible for the subjective analyst to locatecertain indirect indicators of upper divergence.We have already considered some of these,especially positive vorticity advection. Anothercommonly used indicator of upper divergence is�difluence�, but it should be recognized thatdifluence and divergence are not equivalent! Asdetailed by McNulty 1978 , the left front and rightrear quadrants of upper jet maxima are favoredfor divergence (with the justification calling uponvorticity advection arguments). No doubt amyriad of empirical rules governing the use ofupper-level charts for convective forecasting canbe theoretically justified through some connectionwith upper-level divergence.

With all these indirect methods, of varying quality,one could legitimately ask: Why not computeupper-level divergence directly? There areseveral problems with direct calculation ofdivergence which we have already considered(III.B.1), and which are described in textbooks(e.g., Haltiner and Martin, 1957, p.314ff). Themain argument is that it is difficult to do sobecause the two terms used are relatively largeand of opposite sign, so we end up taking thedifference of two large numbers. Such anoperation can result in the creation of substantialerrors. Another not as well-recognized problem isthe quality and quantity of upper-level wind data.When the sonde has reached, say 300 mb, it maywell have traversed enough horizontal distancethat it is low on the horizon. Low elevation anglescan create substantial errors in wind computations(see, e.g., Middleton and Spilhaus, 1953, Ch. VII).In fact, in situations of greatest interest � i.e.,those with strong winds aloft�this low elevationangle problem is at its worst.

Therefore, winds above 300 mb may havesubstantial errors associated with them. Finally,each rawinsonde terminated below 300 mbreduces the overall number of observations. It isnot a rare event that at least one or two

79

rawinsondes do not reach 300 mb in a given set ofsynoptic observations.

In spite of these problems, there is good reason tobelieve that a careful analysis of divergence ispossible. The keys to producing a meaningfuldivergence calculation are a pre-analysisscreening for obviously erroneous data and theproper smoothing of the wind field to be used inthe divergence calculation (see Panofsky, 1964, p.33f) .

McNulty�s (1978) approach provides a meandivergence between 300 and 200 mb (in effect,an upper level vertical motion field).2 This verticalaveraging tends to smooth out the irregularitieswhich might be found at any given level. Hisresults (an example is shown in Fig. 4.3) suggestthat the analysis does contain meaningfulinformation about the divergence field. In arelated area, Schaefer and Doswell (1979) haveshown that a different method for calculatingdivergence and vorticity (using line integrals) isinherently superior to the conventional method

(using derivatives of the wind components). Insome unpublished examples, McNulty has foundthat the line integral method does, indeed,produce a field which better relates computeddivergence to satellite images of cloud patterns.The line integral approach to upper divergenceremains to be tested operationally, as it issomewhat more time-consuming thanconventional methods.

E. Kinematic Analyses and Trajectories

We have already mentioned vorticity anddivergence many times. These parameters aretwo of the four main properties of the wind field.The other two are stretching and shearingdeformation. It is possible to use either divergenceand vorticity or the two components ofdeformation to reconstruct the wind.3 When usingdivergence and vorticity, for example, thedivergence can be used to find the �irrotational�wind contribution and the vorticity to find the

Fig. 4.3. Example ofmean 300-200 mbdivergence contoursproduced at NSSFConmap background. Thezero line is along theboundary between “0”and “-”, with contours atintervals of +### 5 x 10-5

s-1.

80

�non-divergent� contribution. Non-divergentwinds have been used extensively in theinitialization process of numerical models and inthe study of atmospheric dynamics (e.g., Haltinerand Williams, 1980). The reader should recall thatthe geostrophic wind is an example of anessentially non-divergent wind (to a firstapproximation). The isallobaric �wind� is anexample of an irrotational field, but it should benoted that it does not represent a true wind (seeIII.D.2).

While these notes cannot provide much workingknowledge of the often-neglected topic ofkinematic analysis, some topics deserve specialattention as they relate to objective �analysis�and also to forecasting. Readers are urged topursue the whole range of kinematics in thereferences (esp. Saucier, 1955, Ch. 10; orPetterssen, 1956a, Ch. 2).

Deformation is a kinematic property to whichrelatively little attention is given. This isunfortunate since deformation is that propertywhich is characteristic of fluid flow. Note thatthere are two components: stretching andshearing. It is possible to combine these into asingle, resultant deformation (which is alwaysnon-negative) with a resultant axis of dilatation.This is analogous to combining the twocomponents of the vector wind into a (non-negative) wind speed and a result. direction.Shown in Fig. 4.4 are the two components ofdeformation. Note that each has both an axis

where the winds are �converging� and an axiswhere the winds are �diverging� (perpendicularto the former). When the two components ofdeformation are combined mathematically, theresultant deformation has an axis of �diverging�winds along the resultant dilatation axis. Theorientation of the resultant dilatation axis shifts,depending on the relative contributions of theshearing and stretching components (just as in thewind component analogy). The variation of thisaxis with the components is shown in Fig. 4.5.Note that the range of directions is only 180 deg,since the axis of dilation is not directed.4

One should be careful, in considering Fig. 4.4, tonote that these two idealized components of. thetotal deformation need not be apparent in actualflows. It is all too common to ignore deformationunless the flow has this �hyperbolic� appearance- this is simply not valid. The pictures of Fig. 4.4should be thought of as relative flows, since thetotal flow can have intense deformation withoutany such pattern apparent (see Saucier, 1955;Doswell, 1982b) .

Distribution of atmospheric properties liketemperature, moisture, etc., are influenceddramatically by deformations. This is apparent toanyone who has observed cloud motions in ananimated satellite loop. Cloud patterns arestretched and sheared by the flow, and thisprocess often results in the formation of cloudlines. Not only do thunderstorms frequently

Fig. 4.4. The two types of horizontal deformation (after Saucier, 1955). On the right is “shearing”deformation and on the left is “stretching” deformation.

81

develop in lines, but all cloud types can occur inlinear features.The banded character of cloudpatterns may have its origins in processes otherthan deformation at times (see e.g., Kuettner,1959), but deformation is a highly visible aspect ofatmospheric flow.

Since clouds typically arise from vertical motion,it can be seen that divergence patterns shouldalso be influenced by deformation. The bandedcharacteristic often is not apparent whenobjectively analyzed fields of divergence areexamined. Such by more or less circular patterns,rather than bands, of divergence andconvergence.

Why is this the case? A large part of theexplanation is the nature of objective interpolationto uniformly distributed grid points. This has beenrecognized, but is not often emphasized. Wiin-Nielson (1959) has pointed out:

... there is a tendency to deform the initialrather regular pattern into the structure of

elongated bands ... The stretching is in someregions so large that the bands disappearbetween the grid points ... We are thereforebound to get a smooth picture. This does,however, not mean that we should neglectthe deformation properties of the fields.

Naturally, these statements apply to grid pointmodels as well as to grid point analyses. This isnot a fault of the grid point analysis schemechosen or of the parameters of the scheme. AsBarnes� (1964) method exemplifies, it is possible tointerpolate objectively so as to reproduce theobservations to whatever degree is desired.Rather, it is a characteristic of all such schemes.

Can this defect of objective interpolation beovercome? As discussed in II.A, the humananalyst can draw (or re-draw, if using machine-prepared isopleths) long, narrow ribbons duringthe analysis. To guide this process, use may bemade of satellite images and temporal continuity(i.e., trends toward developing the bands).Another, more objective approach for negatingthis tendency (inherent in using grid points) istrajectory analysis and forecasting. As withkinematics, the details of trajectory approachesare beyond the scope of this text. References

Fig. 4.5. Illustration of how the two components ofdeformation may be combined into a cingle“resultant” deformation (after Saucier, 1955). Thestretching deformation value (a) is plotted alongthe x-axis, while the shearing component (a’) isplotted along the y-axis. In this example, a = +5and a’ = +3, to give an angle (at the upper right ofthe rectangle) with respect to the x-axis of 31 degfor the axis of dilatation, the angle is half of thisvalue, or 15.5 deg.

Fig. 4.6. Illustrating frontogenesis. The line offrontogenesis (hatched) moves toward the axis ofdilatation, while the temperature contrast increases(after Petterssen, 1956a).

82

should be consulted (e.g., Saucier, 1955; Doswell,1982a, Wiin-Nielson, 1959; Reap, 1968, 1972).

In the most simple terms, trajectories trace out thepaths of air parcels over some finite time period.One should be aware that trajectories andstreamlines are not generally the same. Unlessthe flow is completely steady-state, thetrajectories will differ from streamlines in possiblyimportant ways. The advantage to calculation anduse of trajectories (usually called Lagrangianmethods) is that trajectories can account for thedeformation in the flow directly. Parcels tend tocollect in deformation zones, allowing for moreresolution in precisely those areas wheregradients are becoming most intense.

At this point, we should digress briefly to considera topic implied by the previous sentence:frontogenesis. In essence, frontogenesis is theprocess by which gradients of atmosphericquantities are intensified. Deformation fields, asone might expect by our initial discussion on thesubject, have a crucial role in frontogenesis(Saucier, 1955; Petterssen, 1956a; Miller, 1948;Stone, 1966; Hoskins and Bretherton, 1972, etc.).The circulations involved are not forced by thefrontogenesis but, rather, accompany it. The topicis complex and still the subject of ongoingresearch. However, we should be aware of thestrong linkages among deformation, frontogenesis,vertical motion, and divergence.

Briefly, if we consider frontogenesis to mean theaccumulation of gradient for, say, potentialtemperature (as in the common definition), thenthere are several factors which need examination.The horizontal accumulation of gradient can beshown to be the product of the original gradientvalue and the combined effects of convergenceand contraction via deformation. Thus, the largerthe initial gradient, the more rapid thefrontogenesis. Convergence is clearlyaccumulative, but in order for the deformation tocontribute to frontogenesis, the axis of dilationmust be tilted at less than a 45 deg angle to theproperty lines (see Fig 4.6).

One should also note that horizontal differences invertical velocity can act to �tilt� the normallystrong vertical gradients into the horizontal,resulting in an increase of horizontal gradients. Itis a process which is frequently ignored, but this

neglect is perilous. The justification usually statesthat vertical motions are so weak, that gradientsin vertical velocity are unimportant. This issubstantially in error, since vertical gradients ofatmospheric variables are often quite strong. Thecontribution by differential vertical advection caneasily be as strong as, or stronger than, thatproduced by horizontal advective effects. Onealso should observe that, in general, a verticalcirculation in which relatively warm air rises andcold air sinks (a direct circulation) acts to destroyhorizontal gradients of potential temperature. Thiscan be seen easily in Fig. 4.7. Clearly, a verticalcirculation of the opposite sense (an indirectcirculation) increases the horizontal potentialtemperature gradient.

By increasing resolution in deformation zones, thetrajectory method offers some distinct advantagesover grid point schemes (usually called Eulerianmethods). There are, of course, somedisadvantages. For example, it is hard to construct�weather maps� from the distorted structureswhich results (see e.g., Welander, 1955). Reap(1968) uses backward trajectories to compensatefor this � i.e., he traces parcels backward in timefrom a uniform grid. However, this creates an endproduct which still �suffers� from the uniformresolution implicit in a grid. Other problems exist,but it suffices to say that the grid point approach isbest suited for numerical models which forecastfor periods beyond 12 h. Many of the problemsare of substantially lesser significance forforecasts of 12 h or less (Doswell, 1982a).

Fig. 4.7. Illustration of the effect of a directcirculation (warm air rising; cold air sinking) onhorizontal gradients of potential temperature. Thesame picture arises if the air is rising (or sinking)everywhere, but the warm air is rising relative tothe cold.

83

CHAPTER IV FOOTNOTES

1 IV.B: Observe, however, that this is not strictlytrue. When the geostrophic wind is strong anddirected northward (southward), the convergence(divergence) associated with the changingCoriolis parameter may reach significant values.

2 IV.D.: Since we can calculate the mean upper-level divergence, one might reasonably ask whynot compute mean divergence at low levels also,and find vertical motion? The answer is that thishas been tried and the results do not repay theeffort. What one finds is a diagnostic field withlittle or no prognostic value. This may be relatedto the 1200 and 0000 GMT rawinsonde times.However, it does seem that the low-levelrawinsonde divergence field is more susceptibleto weather �contamination� and has lesscontinuity from synoptic time to synoptic time.While the upper-level fields are also �noisy� inthis way, they have been found to be more usefulin a prognostic sense than the old low-level (SFCto 10000 ft MSL) vertical motion charts onceproduced at SELS.

3 IV.E: Actually, this can be done only to within aconstant vector � the �translation� property ofthe wind. Since translation is everywhereconstant within the field, it cannot be describedby derivatives. See Saucier (1955, Ch. 10),Schaefer and Doswell (1979), Doswell (1982b)and their references for further discussion.

4 IV.E: In visual terms of the wind analogy, awind vector has one �arrowhead� showing whichdirection along the line is specified; whereas, theaxis of dilation has two �arrowheads�, so neitherdirection is preferred.

84

V. Interpretation of NumericalGuidance

It is important, therefore, that theforecaster be conversant with theunderlying theories, assumptions, andmodels. In particular, it is important thathe be able to identify the �abnormalsituations� when the idealized models(be they dynamical or statistical) arelikely to be inadequate.

This text is not really the appropriate forum for adiscussion in detail of the human forecaster�s role,but the forecaster concerned with convectiveweather needs to be aware of the inherentlimitations within the models (see Doswell, et al.,1981). With time, the current limitations may besuperceded by others, as our understanding of theatmosphere (as reflected by numerical models)changes.

As discussed in Weiss and Ferguson (1982),numerical guidance in severe storm forecasting(and, perhaps, other areas as well) poses twodifferent dilemmas. The first is to understand howthe large-scale analyses and forecasts relate tothe production of severe convection. It is not astraightforward process to go from an analysis to adepiction of where severe storms will occur. Thesame statement applies even to a perfect forecast.Presumably, combinations of large-scaleparameters (perhaps involving different parametersets in different synoptic-scale settings) can bedeveloped to guide this process. Our current levelof understanding is such that even a perfect modelforecast (and/or analysis) often can lead to animperfect forecast of convective weather.

The second dilemma concerns how well themodels actually perform. This is especially true inforecasting convection, since all the parameterswhich are conceivably of interest to thunderstormforecasting are not currently available directlyfrom model output. It may be possible to constructmost, if not all, of these parameters from theinternal variables in a given model. However, it is

A. General Remarks

The task of analysis is diagnostic in character.Knowing what is going on now is an essentialbeginning to forecasting what will be going on inthe future. Prior to the advent of numericalprognosis, this first step of analysis received thebenefit of considerable attention. The act offorecasting had to proceed in a largely intuitiveway, with a heavy emphasis on �rule of thumb�(i.e., if this happens, then something else willfollow) and extrapolation (i.e., it�s been movingthis way, it should continue to do so).

With the development of increasinglysophisticated numerical models, it has beenpossible to incorporate much of what weunderstand about the weather into the models.Not only can they extrapolate, they can predictnew developments, dissipate old systems, andfilter out �noise�. The amount of detailedmeteorological theory actually brought to bear bythe models far exceeds that available to theanalyst/forecaster. So what role does the humanhave in the process of producing a forecast?Perhaps the obvious answer is that despite theirsophistication, the models still err. At times, theirerrors are far in excess of what a human forecastwould create. Also, the machines occasionallybreak down for one reason or another and ahuman is still needed to salvage the product.Petterssen (1956a) says it quite well:

While the machines provide the answersthat can be computed routinely, theforecaster will have the opportunity toconcentrate on the problems which canbe solved only by resort to scientificinsight and experience. Furthermore,since the machine-made forecasts arederived, at least in part, from idealizedmodels, there will always be anunexplained residual which invites study.

85

not clear that that model (or any other) does asgood a job with those variables as it does with,say, 500 mb heights. Thus, model guidance needsto be considered in light of which parameters canreliably be derived from model output (present orfuture). This is a cornerstone in the model outputstatistics (MOS) approach to forecasting (Glahnand Lowry, 1972). As Weiss and Ferguson (1982)suggest, we are far from a perfect knowledge ofthis, as well.

B. Short and Long Term Error History

At the current time, a multiplicity of models existsand the forecaster is sometimes faced with thedilemma of contradictory output from the differentmodels. Some of this can be clarified by keepingtrack of how the various models have beenbehaving. If a particular model is doing quite wellwith system movement, but has been treating thedevelopment/decay of systems badly over, say thepast week, then this should be considered whenevaluating its latest forecast.

Naturally, short-term model behavior can changefrom day to day, and the way any given modelhandles any given weather system can change.For example, primitive equation models (the LFMis one) have a tendency for what is called�locked-in error� (Fawcett, 1969). This is a model-specific error (and, as such, really represents along-term error) which can routinely be treated,and can (potentially) be eliminated or reduced bychanges in the model. At certain stages in theevolution of a weather system, a given model cando rather poorly, whereas later in its life cycle, thesystem is well-handled.

A basic element in the short-term evaluation ofany given model is the forecaster�s knowledgeand experience of how the atmosphere behaves.When model output is examined, the analyst/forecaster already should have in mind what isanticipated in terms of the overall trend. Shouldthe model output contradict this trend, a furtherexamination of the possible explanations for thisdifference is called for. If the analyst/forecasterexamines the model output first, then there is atendency to be biased by what is seen, and toaccept the model results less critically.

For any given model, it is possible to develop astatistical picture of the model�s behavior over along period. There may be biases and consistenterrors which can be accounted for. This sort oferror analysis should be routinely done at thelarge forecast centers where the models are run,and transmitted to the field offices on a regularbasis (e.g., Fawcett, 1969; Brown and Fawcett,1972). With this information at hand, it should bepossible to modify the model results subjectivelyto account for these consistent errors.

While this process is as yet imperfectlyaccomplished, some effort along these lines isbeing macle (Leary, 1971; Tsui and Brody, 1982).Part of the reason for the imperfect application ofthis concept is the need for lengthy compilation ofthe appropriate statistics. Further, the models arealtered at fairly frequent intervals, compromisingthe value of any accumulated statistics. Theproblems and magnitude of the task tend to inhibitits proper execution. Also, the lack of acommunications channel by which forecastoffices can be made aware of problems perceivedin the field limits the process. Field offices areonly vaguely aware of how the various modelswork, so their knowledge of model limitations iscorrespondingly vague. Further, they are often notaware of development efforts at NMC until aftertheir implementation, if at all.

C. Initialization and Adjustment

One area of great potential value for the analyst/forecaster�s contribution to model outputenhancement is in an assessment of itsinitialization. The numerical forecast begins withinitial data which have been interpolated to themodel grid, If the analyst has done a careful job ofproducing an internally consistent 4-dimensionalpicture of the atmosphere�s structure, he/she isprepared to evaluate how adequately the modelsare initialized. This presupposes that the analysthas available for examination the initial fieldsupon which the model operates.

Hales (1979a) has provided several goodexamples of how this can be accomplished whenthe model input has deficiencies in the data-voidocean areas. This sort of careful examination of

86

model input, in comparison with observations, canbe applied to data-rich areas as well. Forexample, if the analyst has concluded that asignificant feature is found in �the data byconsidering all available information, includingsatellite imagery, the model�s initialization shouldbe studied. If the model fields seem inconsistentwith the analyst�s assessment (the feature mayhave been poorly located or subjected toexcessive smoothing), then the model forecastshould be appropriately modified.

Not infrequently, this sort of initialization errormay represent the major reason for disagreementbetween the subjective forecast and the modeloutput. By changing the location and or strengthof an analyzed feature based on detailed analysisprocedures, the analyst/forecaster can providewhat is lacking in the models. Mesoscale detailsare simply too complex and too small for currentnumerical models and it is unrealistic to expectthem in the models. Current and foreseeablemodels have very limited (if any!) capability toforecast convection and the forecaster analystshould anticipate that a major effort is required toprovide mesoscale detail to model output.

Further, this process should recognize that currentmodels (and expected future revisions) are leastaccurate in the very long-term and very short-term ranges. The long-term range is not ofoverwhelming concern to the forecast ofconvection, but the short-term (12 hr or less) rangeis the area of greatest concern. Even with carefulinitialization procedures,: numerical models starttheir forecasts with a period of �adjustment�,wherein the variables within the model come intoa state of balance dictated by the model�sgoverning equations. This adjustment period cantake 6-12 hrs of forecast time, and ischaracterized by rapidly oscillating variables,while the mass and velocity fields reach a�mutual understanding�. During this period, theoutput is unreliable.

The reasons for this adjustment period need notconcern us here, but the result is that the 12-hrforecast may not be as reliable as the 24-hrprognosis. This fact, coupled with the problemsposed by coarse initial data and inadequateconvective physics, places great responsibility onthe analyst/forecaster. The model output cangenerally be relied upon (within the limits

supplied by its error history to provide a broad-scale background of change, upon which aresuperimposed the details of specific interest to theconvection forecast. It is up to the analyst/forecaster to provide those details, based onphysical understanding not currently incorporatedin the models.

D. Statistical Convective WeatherGuidance

Within the last few years, a statistical approach toconvective storms forecasting has beendeveloped and put into operation. There are twodistinct products, one for short-range use (Charba,1979a) analogous to the SELS �Watch� product,and the other for medium-range use (Reap andFoster, 1979) analogous to the SELS �ConvectiveOutlook� product. Both take the form of a map ofconvective storm probability, and both arederived by a mathematical process known asscreening regression.

Briefly, screening regression proceeds in thefollowing manner. Given a data set ofoccurrences (�predictands�) for a particular event,a parameter set is offered to the screeningregression program. The list of candidateparameters (�predictors�) may run into thehundreds. The program searches the parametersto find the one parameter which explains thegreatest amount of the variation (in space andtime) in the event�s occurrence. Then, given thefirst such parameter, what parameter incombination with the first explains the mostvariation? The process continues in this fashionuntil some chosen threshold is reached, whereadding new parameters has reached the point ofdiminishing returns. The result is an equationwhich forms a weighted sum of all the chosenpredictors such that, when values of the actualpredictors are plugged into the equation, aprobability for the event is produced.

The short-range statistical guidance makes heavyuse of observed data, from the surface networkand from radar, to provide predictors. Themedium-range product uses LFM-derived forecastpredictors. There is considerable art involved indeveloping predictors, owing to certain limitations

87

imposed by the screening regression method. Bothproducts find that a substantial amount of thenatural variation is associated with what theyhave termed an �interactive� predictor. This isessentially a modulated climatology, in which theclimatological frequency is modified according tothe value of another parameter (surface pressurefor severe convection, the K-index for generalthunderstorms).

In order to transform the probability values into ayes-no forecast, thresholds have been developed.These thresholds are still being experimentedupon, and the exact methods await furtherresearch, especially in the medium-rangeproducts.

Comparative verification of these statisticalproducts with those produced at SELS has not ledto any definitive conclusions. Generally speaking,the short-range products do not detect as manysevere occurrences as do the SELS watches, butthey can have somewhat better �lead time�(Charba, 1979a). Conversely, the medium-rangeproducts have a greater chance of including thesevere events within their thresholds, but theymay do so at the expense of falsely alerting asubstantially larger area (Weiss et al., 1980a,Weiss et al., 1980b). Recently, experimentalmedium-range statistical forecasts havedemonstrated better skill at reducing the falselyalerted area (Reap et al., 1982).

From an analyst�s viewpoint, these statisticalguidance products are best dealt with in the sameway as more conventional guidance. That is, thebest strategy is to form a conception of when andwhere severe weather is likely to develop withouthaving seen the guidance. Then, if there is adifference, the analyst/forecaster should try tounderstand the difference and make adjustments(if necessary) to the first conception, based on anexamination of the differences.

An obstacle to this procedure (which also appliesto all facets of interpreting numerical guidance) isthat it is not always clear why the guidance isperforming in the way it does. If the analyst isuncertain what went into the guidance product,there is not much available which can provideany insights, since the entire process is out of his/her hands. This is further complicated byunannounced changes to the model and to the

statistical routines. Hopefully, enough stability inthe system will eventually exist that the analystcan begin to determine those circumstances whenthe guidance is best and when the guidance ismost likely to fail. Charba�s (1979b) efforts todocument these synoptic situations when hisproduct performs most poorly are a step in theproper direction. Much more of this sort of self-evaluation is necessary.

88

VI. Concluding Remarks onMesoanalysis

situation. Severe weather episodes under unusualmeteorological circumstances, such asnorthwesterly flow aloft (Johns, 1977, 1982a,b), orwith exceptionally low dewpoints (Johns, 1982c),and in unique geographical locations like NewEngland (David, 1977) or the High Plains(Doswell, 1980) should not be seen as anomalies,but as elements of the same basic picture.

Similarly, we should be aware that severethunderstorms are not exclusively confined tospringtime situations involving strongcyclogenesis, although the majority of strongstorms occur in the period April through June(Kelly et al., 1978). When wintertime severestorms do develop (Galway and Pearson, 1979;Burgess and Davies-Jones, 1979), they can easilybe integrated into the patterns we have described.Summertime severe thunderstorms also fit(Maddox and Doswell, 1982). The seasonalvariations in weather patterns can create a severeweather threat in a variety of ways. Differentseasons are dominated by different parameters,but the basic building blocks of unstably stratified,moist, rising air are vital to severe storms and theanalyst�s job is to diagnose if, when, and wherethose basic building blocks will come together.

This concept also rules out a rigidly structuredanalysis program. Since the primary ingredientsmay be developed in a large variety of ways, it isnot productive to lay out a rigid set of rules foranalysis. Charts, parameters, and even conceptsvaluable to forecast a given situation may not besignificant in the next. Further, although the basicelements of unstably stratified, rising moist air arenecessary conditions, they are not by themselvessufficient. There are many unknown or poorlyunderstood factors (e.g., microphysicalinteractions, the exact role of vertical shear, etc.)and it is naive to expect that our current conceptsof how these factors interact to produce severeweather shall survive unchallenged for very long.

We have not spent a great deal of time givingspecific details about how weather map analysis

It is hoped that this text has conveyed one ideaabove all. That idea is that mesoanalysis must bebased upon physical understanding. Rather thantie the analysis process to a particular weatherchart (or charts) and develop some all-powerfulparameter (or set of parameters), the concept ofintegrating all available analysis tools into aphysically consistent picture is heavily stressed.

If one works on the assumption that severethunderstorms occur when unstably stratified airhaving sufficient moisture is lifted, then theanalysis (which includes the physicalinterpretation) of data is dramatically simplified.�Features� in the data are no longer mysteriouslycombined to produce severe thunderstormforecasts. Of course, this places a burden on theanalyst/forecaster � namely, he/she mustunderstand how the �features� contribute to (a)development of unstably stratified air, (b) thepresence [and the sufficiency] of moisture, and (c)the occurrence of upward vertical motion.Further, an awareness of how weather systemswork (as we currently understand them) isessential to the proper accomplishment of theanalysis. The development of conditions favorableto severe thunderstorms involves complex butbasically understandable interactions among thethree basic ingredients. That is, for example,lifting can act to destabilize the stratification, andthe advective processes which introduce moistureand instability can also create upward motion.The basic elements of synoptic analysis are alsoapplicable to the convective forecast � the focusis modified from the evolution of large-scaleweather systems by having to concentrate ondetails which are usually not important to thesynoptic scale. However, as we have learned, thestorms themselves and their mesoscale effects canhave a dramatic influence on the large-scalesystems, as well.

With a better physical understanding of how thelarge-scale pattern sets the stage, we no longerneed to depend on a different set of rules for each

89

relates directly to severe weather. Someapplications of the contents in this volume (andthe next) are explored in Vol. III. but it is notpossible to be exhaustive in any treatise of thissort. Rather, the interested reader will consult thereferences. It is the reader�s responsibility (anddistinctly to his/her advantage) to pursue furtherthe topics mentioned in these notes � to avoiddoing so is to miss the point.

While we have asserted that there is a �bigpicture� based on physical understanding, oneshould not be deceived into thinking that anyone(especially the author!) fully understands that bigpicture. However, the analyst/forecaster canapply some fairly simple dynamical ideas whichare valid on the synoptic and subsynoptic scaleand use them to improve everyday weatheranalysis. By similar reasoning, the reader shouldbe aware of the basic physical processesoccurring on the thunderstorm scale. This is thegoal of Volume II, Storm Scale Analysis.

90

REFERENCES

Meteorological conditions for the formation ofrain. Geof. Publ., 2, No. 3.

__________, and __________, 1922: Life cycle ofcyclones and the polar front theory of atmosphericcirculation. Geof. Publ., 3, No. 1.

Blackadar, A.K., 1957: Boundary layer windmaxima and their significance for the growth ofnocturnal inversions. Bull. Amer. Meteor. Soc., 5,283-290.

Bleck, R., 1977: Numerical simulation of thecyclogenesis in the Gulf of Genoa. Mon. Wea.Rev., 105, 428-445.

Bolin, B., 1950: On the influence of the earth�sorography on the general character of thewesterlies. Tellus, 2, 184195.

Bonner, W.D., 1966: Case study of thunderstormactivity in relation to the low level jet. Mon. Wea.Rev., 94, 167-178.

__________, and J. Paegle, 1970: Diurnalvariations in boundary layer winds over the south-central United States in summer. Mon. Wea. Rev.,98, 735-744.

Braham, R.R., Jr., and M.J. Dungey, 1978: A studyof urban effects on radar first echoes. J. Appl.Meteor., 17, 644-654.

Brooks, E.M., 1949: The tornado cyclone.Weatherwise, 2, 32-33.

Brown, H.E., and E.B. Fawcett, 1972: Use ofnumerical guidance at the National WeatherService National Meteorological Center. J. Appl.Meteor., 11, 1175-1182.

Brunt, D. and C.K.M. Douglas, 1928: Themodification of strophic balance for changingpressure distribution, and its effect on rainfall.Memoirs of the Royal Meteor. Soc., III, No. 22.

Burgess, D.W., and R.P. Davies-Jones, 1979:

Alberty, R.L., 1969: A proposed mechanism forcumulonimbus persistence in the presence ofstrong vertical shear. Mon. Wea. Rev., 97, 590-596.

Anderson, R.K., G.R. Farr, J.P. Ashman, V.J. Oliver,E.W. Ferguson, A.H. Smith, J.F.W. Purdom andR.W. Skidmore, 1974: Applications ofmeteorological satellite data in analysis andforecasting. ESSA Tech. Rep. NESC 51 [NTISAccession No. AD 697-033).

Appleby, J.F., 1954: Trajectory method of makingshort-range forecasts of differential temperatureadvection, instability, and moisture. Mon. Wea.Rev., 82, 320-334.

Barnes, S.L., 1964: A technique for maximizingdetails in numerical weather map analysis. J.Appl. Meteor., 3, 396-409.

__________, 1973: Mesoscale objective mapanalysis using weighted time-series observations.NOAA Tech. Memo. ERLTM-NSSL-62, 60 pp.[NTIS Accession No. COM-7310781].

__________, 1978: Oklahoma thunderstorms on29-30 April 1980. Part II: Radar-observed mergerof twin hook echoes. Mon. Wea. Rev., 106, 685-696.

Beebe, R.G., 1958: Tornado proximity soundings.Bull. Amer. Meteor. Soc., 39, 195-201.

__________, and F.C. Bates, 1955: A mechanismfor assisting in the release of convectiveinstability. Mon. Wea. Rev., 83, 1-10.

Bellamy, J.C., 1945: The use of pressure altitudeand altimeter corrections in meteorology. J.Meteor., 2, 1-79.

Bjerknes, J., 1919 On the structure of movingcyclones. Geof. Publ., 1, No. 2.

__________, and H. Solberg, 1921:

91

Unusual tornadic storms in eastern Oklahoma on5 December 1975. Non. Wea. Rev., 107, 451-457.

Byers, H.R., 1959: General Meteorology.McGraw-Hill, New York, 540 pp.

Cahir, J., 1971: Implications of circulations in thevicinity of jet streaks at subsynoptic scales. Ph.D.Dissertation, Pennsylvania State Univ., UniversityPark, PA, 170 pp.

Carbone, R., and R. Serafin, 1980: A severe wintersquall line. Part II: Kinematic structure deducedfrom triple Doppler radar observations, Preprints,19th Conf. Radar Meteorology (Miami, FL , Amer.Meteor. Soc., Boston, 9-1

Carlson, T.N., and F.H. Ludlam, 1968: Conditionsfor the occurrence of severe local storms. Tellus,20, 203-226.

Chang, L.W., 1976: A numerical study of thediurnal variation of the low-level jet. Ph.D.dissertation, Univ. of Oklahoma, Norman, OK,110 pp.

Changnon, S.A., 1978: Urban effects on severelocal storms at St. Louis. J. Appl. Meteor., 17, 578-586.

__________, 1980: More on the La Porteanomaly: A review. Bull. Amer. Meteor. Soc., 61,702-711.

Charba, J.P., 1974: Application of gravity currentmodel of analysis of squall-line gust front. Mon.Wea. Rev., 102, 140-156.

__________, 1979a: Two to six hour severe localstorm probabilities: An operational forecastingsystem. Mon. Wea. Rev., 107, 268-282.

__________, 1979b: Recent performance ofoperational two-six hour objective forecasts ofsevere local storms on outbreak days. Preprints,11th Conf. on Severe Local Storms (Kansas City,MO , Amer. Meteor. Soc., Boston, 600-607.

__________, and Y. Sasaki, 1971: Structure andmovement of the severe thunderstorms of 3 April1964 as revealed from radar and surfacemesonetwork data analysis. J. Meteor. Soc. Japan,49, 191-214.

Charney, J.G., 1947: The dynamics of long wavesin a baroclinic westerly current. J. Meteor., 4, 135-162.

Chisholm, A.J., 1973: Alberta hailstorms. Part I:Radar case studies and airflow models. Meteor.Monogr., 14, No. 36, 1-36.

Chopra, K.P. and L.F. Hubert, 1965: Mesoscaleeddies in wake of islands. J. Atmos. Sci., 22, 652-657.

Chung, Y.S., K.D. Hage, and E.R. Reinelt, 1976:On lee cyclogenesis and airflow in the CanadianRocky Mountains. Mon. Wea. Rev., 104, 879-891.

Clarke, R.H., R.K. Smith, and D.G. Reid, 1981:The morning glory of the Gulf of Carpentaria: Anatmosphere undular bore. Mon. Wea. Rev., 109,1726-1750.

Darkow, G.L., 1968: The total energyenvironment of severe storms. J. Appl. Meteor., 7,199-205.

__________, and D.W. McCann, 1977: Relativeenvironmental winds for 121 tornado-bearingstorms. Preprints, 10th Conf. Severe Local Storms(Omaha, NE , Amer. Meteor. Soc., Boston, 413-417.

__________, V.E. Suomi, and P.M. Kuhn, 1958:Surface thermal patterns as a tornado forecast aid.Bull. Amer. Meteor. Soc., 39, 532-537.

David, C.L., 1977: A study of synoptic conditionsassociated with New England tornadoes.Preprints, 10th Conf. on Severe Local Storms(Omaha, NE , Amer. Meteor. Soc., Boston, 180-185.

Defant, F., 1951: Local winds. Compendium ofMeteor,, T.F. Malone, Ed., Amer. Meteor. Soc.,Boston, 655-672.

Dirks, R.A., 1969: A theoretical investigation ofconvective patterns in the lee of the ColoradoRockies. Atmos. Sci. Paper No. 145, Colo. St.Univ., Ft. Collins, CO, 122 pp.

Djuric, D., and M.S. Damiani, Jr., 1980: On theformation of. the low level jet over Texas. Mon.Wea. Rev., 108, 1854-1865.

Doswell, C.A. III, 1976: Subsynoptic scaledynamics as revealed by use of filtered surfacedata. NOAA Tech. Memo., ERL-NSSL-79,Norman, Oklahoma, 40 pp.

__________, 1977: Obtaining meteorologicallysignificant surface divergence fields through the

92

filtering property of objective analysis. Mon, Wea.Rev., 105, 885-892.

__________, 1980: Synoptic scale environmentsassociated with High Plains severe thunderstorms.Bull. Amer. Meteor. Soc., 60, 1388-1400.

__________, 1982a: Short-range forecasts ofupper-level divergence using 250 mb parceltrajectories. Preprints,, 12th Conf. Severe LocalStorms (San Antonio, , Amer. Meteor,. Soc.,Boston, 75-78.

__________, 1982b: Diagnosis of weather eventsvia kinematic analysis. Preprints, 9th Conf. Wea.Forecasting and Analysis (Seattle, WA) , Amer.Meteor. Soc., Boston, 300-303.

__________, and L.R. Lemon, 1979: Anoperational evaluation of certain kinematic andthermodynamic parameters associated withsevere thunderstorm environments. Preprints, 11thConf. on Severe Local Storms {Kansas City, MO ,Amer. Meteor. Soc., Boston, 397-402.

__________, and R.A. Maddox, 1981: Forecastertraining - A review and analysis. Bull. Amer.Meteor. Soc., 62, 983-988.

__________, J.T. Schaefer, D.W. McCann, T.W.Schlatter. and H.B. Wobus, 1982: Thermodynamicanalysis procedures at the National Severe StormsForecast Center. Preprints, 9th Conf. Wea.Forecasting and Analysis (Seattle, WA) , Amer,Meteor. Soc., Boston, 304-309.

Dutton, J.A., 1976: The Ceaseless Wind: AnIntroduction to the Theory of Atmospheric Motion.McGraw-Hill, New York, NY, 579 pp.

Emanuel, K.A., 1980: Forced and free mesoscalemotions in the atmosphere. Proceedings of CIMMSSymposium (Y.K. Sasaki, N. Monji, and S. Bloom,Ed�s.). Norman, OK, 191-259.

Eom, J.K., 1975: Analysis of the internal gravitywave occurrence of 19 April 1970 in the Midwest.Mon. Wea. Rev., 103, 217-226.

Estoque, M.A., 1961: A theoretical investigation ofthe sea breeze. Quart. J. Roy. Meteor. Soc., 87,136-146.

Fankhauser, J.C., 1971: Thunderstorm-environment interactions determined from aircraftand radar observations. Mon. Wea. Rev., 99, 171-192.

Fawbush, E.J., and R.C. Miller, 1953: A method forforecasting hailstone size at the earth�s surface.Bull. Amer. Meteor. Soc., 34, 235-244.

__________, and __________, 1954a: A. basis forforecasting peak wind gusts in non-frontalthunderstorms. Bull. Amer. Meteor. Soc., 35, 14-19.

__________, and __________, 1954b: The typesof air masses in which North American tornadoesform. Bull. Amer. Meteor. Soc., 35, 154-165.

Fawcett, E.B., 1969: Systematic errors inoperational baroclinic prognoses at the NationalMeteorological Center. Mon. Wea. Rev., 97, 670-682.

Fleagle, R.G., 1948: Quantitative analysis offactors influencing pressure change. J. Meteor., 5,281-292.

Foster, D.S., 1958: Thunderstorm gusts comparedwith computed downdraft speeds. Mon. Wea.Rev., 86, 91-94.

__________, and P.C. Bates, 1956: A hail-sizeforecasting technique,. Bull. Amer. Meteor. Soc.,37, 135-141.

Fritsch, J.M., 1975: Cumulus dynamics: Localcompensating subsidence and its implications forcumulus parameterization. Pageoph., 113, 851-867.

__________, C.F. Chappell, and L.R. Hoxit, 1976:The use of large-scale budgets for convectiveparameterization. Mon. Wea. Rev., 104, 1408-1418.

Fujita, T.T., 1955: Results of detailed synopticstudies of squall lines. Tellus, 4, 405-436.

__________, 1958: Structure and movement of adry front. Bull. Amer. Meteor. Soc., 32, 1 9.

__________, 1959: Precipitation and cold airproduction in mesoscale thunderstorm systems. J.Meteor., 16, 454-466.

__________, 1963: Analytical mesometeorology:A review. Meteor. Monogr., 5, No. 27, 77-125.

__________, 1978: Manual of downburstidentification for Project NIMROD. Satellite andMesometeorology Research Project Res. PaperNo. 156, Univ. of Chicago, Chicago, IL, 104 pp.

93

__________, and H.A. Brown, 1958: A study ofmesosystems and their radar echoes. Bull. Amer.Meteor. Soc., 39, 538-544.

__________, and H. Grandoso, 1968: Split of athunderstorm into anticyclonic and cyclonicstorms and their motion as determined fromnumerical model experiments. J. Atmos. Sci., 25,416-439.

__________, and F. Caracena, 1977: An analysisof three weather-related aircraft accidents. Bull.Amer. Meteor. Soc., 58, 1164-1181.

__________, H. Newstein and M. Tepper, 1956:Mesoanalysis, An important scale in the analysisof weather data. Res. Paper No. 39, Washington,D,C., USWB, 83 pp.

Fulks, J.R., 1951: The instability line. Compendiumof Meteorology, T.F. Malone, Ed., Amer. Meteor.Soc., Boston, 647-652.

Galway, J.G., 1956: The lifted index as a predictorof latent instability. Bull. Amer. Meteor. Soc., 37,528-529.

__________, 1977: Some climatological aspectsof tornado outbreaks. Mon. Wea. Rev., 105, 477-484.

__________, and A.D. Pearson, 1979: Wintertornado outbreaks. Preprints, 11th Conf. on SevereLocal Storms Kansas City, MO , Amer. Meteor.Soc., Boston, 1-6.

Gentry, R.C., 1982: Tornadoes associated withhurricanes, Preprints. 12th Conf. on Severe LocalStorms San Antonio, TX , Amer. Meteor. Soc.,Boston, 367-370.

George, R.L., 1979: Evolution of mesoscaleconvective systems over mountainous terrain.Atmos. Sci. Paper No. 318, Colo. St. Univ., Ft.Collins, CO, 160 pp.

Glahn, H.R., and D.A. Lowry, 1972: The use ofmodel output statistics in objective weatherforecasting. J. Appl. Meteor., 11, 1203-1211.

Godske, C.L., T. Bergeron, J. Bjerknes, and R.C.Bundgaard, 1957: Dynamic Meteorology andWeather Forecasting. Amer. Meteor. Soc., Boston,and Carnegie Institution, Washington, D.C., 800pp.

Goff, R.C., 1976: Vertical structure of

thunderstorm outflows. Mon. Wea. Rev., 104,1429-1440.

Gurka, J.J., 1980: Observations of advection-radiation fog formation from enhanced IR satelliteimagery. Preprints, 8th Conf. Weather Forecastingand Analysis Denver, CO , Amer. Meteor. Soc.,Boston, 108-114.

Hage, K.D., 1961: On summer cyclogenesis in thelee of the Rocky Mountains. Bull. Amer. Meteor.Soc., 42, 20-33.

Hales, J.E., 1977. The Kansas City flash flood of 12September 1977. Bull. Amer. Meteor. Soc., 59,706-710.

__________, 1979a: A subjective assessment ofmodel initial conditions using satellite imagery.Bull. Amer. Meteor. Soc., 60, 206-211.

__________, 1979b: On the relationship of 250mb positive vorticity advection and horizontaldivergence to tornado and severe thunderstormdevelopment. Preprints, 11th Conf. on SevereLocal Storms Kansas City, MO , Amer. Meteor.Soc., Boston, 28-31.

__________, 1980: The Windsor Locks tornado.Preprints, 8th Conf. Wea. Forecasting andAnalysis Denver, CO , Amer. Meteor. Soc.,Boston, 69-73.

__________, 1982: Relationship of selectedsynoptic scale parameters to significant tornadooccurrences in 1980. Preprints, 12th Conf. onSevere Local Storms (San Antonio, TX), Amer.Meteor. Soc., Boston, 139-142.

Haltiner, G.J., and F.L. Martin, 1957: Dynamicaland Physical Meteorology. New York, McGraw-Hill, pp 202-204,

__________, and R.T. Williams, 1980: NumericalPrediction and Dynamic Meteorology, 2nd Ed.,John Wiley & Sons, New York, NY, 477 pp.

Harris, M.F., 1959: Diurnal and semidiurnalvariations of wind, pressure, and temperature inthe troposphere at Washington, D.C. J. Geophy.Res. 64, 983-995.

Harrold, T.W., and K.A. Browning, 1969: Thepolar low as a baroclinic disturbance. Quart. J.Roy. Meteor. Soc., 95, 710-723,

Haurwitz, B., 1947: Comments on the sea-breeze

94

circulation. J. Meteor., 4, 1-8.

Hess, S.L., 1959: Introduction to TheoreticalMeteorology. Holt, Rinehart and Winston, 250-253.

Hoerker, W.H., Jr., 1963: Three southerly lowlevel jet streams delineated by the WeatherBureau special pibal network of 1961. Mon. Wea.Rev., 91, 573-582.

Holle, R.L., and M.W. Maier, 1980: Tornadoformation from downdraft interaction in the FACEmesonetwork. Mon. Wea. Rev., 108, 1010-1028.

Holton, J.R., 1979: An Introduction to DynamicMeteorology, 2nd M. Academic Press, New York,NY, 391 pp.

Hoskins, B.J.and F. P. Bretherton, 1972:Atmospheric frontogenesis models: Mathematicalformulation and solution. J. Atmos. Sci., 29, 11-37.

__________, I. Draghici and H.C. Davies, 1978:A new look at the w-equation. Quart. J. Roy.Meteor. Soc., 104, 31-38.

House, D.C., 1958: Airmass modification andupper level divergence. Bull. Amer.Meteor. Soc.,39, 137-143.

__________, 1959: The mechanics of instability-line formation. J. Meteor., 16, 108-120.

Hoxit, L.R., and C.F. Chappell, 1975: Tornadooutbreak of April 3-4, 1974: Synoptic analysis.NOAA Tech. Rept., ERL 338-APCL 37,Environmental Research Laboratories, Boulder,CO, 48 pp.

__________, and J.M. Fritsch, 1976: Formation ofmesolows or pressure troughs in advance ofcumulonimbus clouds. Mon. Wea. Rev., 104,1419-1428.

Hsu, S.A., 1969: Mesoscale structure of the Texascoast sea breeze. Report No. 16, NSF Grant GA-367Z, Univ. of Texas at Austin, College ofEngineering, Atmospheric Science Group, 237 pp.

Hubert, L.F. and A.F. Krueger, 1962: Satellitepictures of mesoscale eddies. Mon. Wea. Rev., 90,457-463.

Hudson, H.R., 1971: On the relationship betweenhorizontal moisture convergence and convectivecloud formation. J. Appl. Meteor., 10, 755-762.

Huschke, R.E., 1959: Glossary of Meteorology.Amer. Meteor. Soc., Boston, 639 pp.

Johns, R.H., 1977: A climatology of northwest flowsevere weather outbreaks. Preprints, 10th Conf.on Severe Local Storms (Omaha, NE , Amer.Meteor. Soc., Boston, 174-179.

__________, 1982a: A synoptic climatology ofnorthwest flow severe weather outbreaks. Part I:Nature and significance. Mon. Wea. Rev., 110 (Inpress).

__________, 1982b: A synoptic climatology ofnorthwest flow severe weather outbreaks. Part II:Meteorological parameters and synoptic patterns.Mon. Wea. Rev., 110 (In press).

__________, 1982c: Severe weather occurring inareas of low surface dew points. Preprints, 12thConf. Severe Local Storms (San Antonio, TX),Amer. Meteor. Soc., Boston, 143-146.

Johnston, E.C., 1978: Mesoscale low-levelvorticity centers associated with convection asviewed by satellite. Nat. Wea. Dig., 3, 12-16.

__________, 1982: Mesoscale vorticity centersinduced by mesoscale convective complexes.Preprints, 9th Conf. on Weather Forecasting andAnalysis Seattle, WA , Amer. Meteor. Soc., Boston,196-200.

Kasahara, A., 1966: The dynamical influence oforography on the large-scale motion of theatmosphere. J. Atmos. Sci., 23, 259-271.

Kelly, D.L., J.T. Schaefer, R.P. McNulty, C.A.Doswell III and R.F. Abbey, Jr., 1978: Anaugmented tornado climatology. Mon. Wea. Rev.,106, 1172-1183.

Keyser, D.. and R.A. Anthes, 1977: Theapplicability of a mixed layer model of theplanetary boundary layer to real data forecasting.Mon. Wea. Rev., 105, 1351-1371.

Klein, W.H., 1957: Principle tracks and meanfrequencies of cyclones and anticyclones in theNorthern Hemisphere. Wea. Bur. Res. Paper, No.40, 60 pp.

Kloth, C.M., and R.P. Davies-Jones, 1980: Therelationship of the 300 mb jet stream to tornadooccurrence. NOAA Tech. Memo. ERL NSSL-88,Nat�l. Severe Storms Lab., Norman, OK, 62 pp.

95

Kreitzberg, C.W., and H.A. Brown, 1970:Mesoscale weather systems within an occlusion.J. Appl. Meteor., 9, 417-432.

Kuettner, J., 1959: The band structure of theatmosphere. Tellus, 11, 267-294.

Kuhn, P.M., G.L. Darkow, V.E. Suomi, 1958: Amesoscale investigation of pre-tornado thermalenvironments. Bull. Amer. Meteor. Soc., 39, 224-227.

Kuo, J.-T. and H.D. Orville, 1973: A radarclimatology of summertime convective clouds inthe Black Hills. J. Appl. Meteor., 12, 359-368.

Leary, C., 1971: Systematic errors in operationalNational Meteorological Center primitiveequation surface prognosis. J. Atmos. Sci., 31,982-995.

Lemon, L.R., 1976: The flanking line, a severethunderstorm intensification source. J. Atmos. Sci.,33, 686-694.

__________, and C.A. Doswell III, 1979: Severethunderstorm evolution and mesocyclonestructure as related to tornadogenesis. Mon. Wea.Rev., 107, 1184-1197.

List, R.J., 1966. Smithsonian MeteorologicalTables, 6th Ed., Smithsonian Institution,Washington D.C., 527 pp.

Livingston, R.L., 1972: An unusual arcus cloud.Mon. Wea. Rev., 100, 817-818,

Ludlam, F.H., 1963: Severe local storms. Areview. Meteor. Monogr., 5, Boston, Amer. Meteor.Soc., 1-30.

Lyons, W.A., 1966: Some effects of Lake Michiganupon squall lines and summertime convection.Proceedings, 9th Conf. Great Lakes Res., Univ. ofMichigan, Great Lakes Res. Div., Pub. 14, 259-273.

Maddox, R.A., 1976: An evaluation of tornadoproximity wind and stability data. Mon. Wea.Rev., 104, 133-142.

__________, l979a: The evolution of middle andupper tropospheric features during a period ofintense convective storms. Preprints, 11th Conf. onSevere Local Storms (Kansas City, MO), Amer.Meteor. Soc., Boston, 41-48.

__________, 1979b: A methodology forforecasting heavy convective precipitation andflash flooding. Nat. Wea. Dig., 4, 30-42.

__________, 1980a: An objective technique forseparating macroscale and mesoscale features inmeteorological data. Mon. Wea. Rev., 108, 1108-1121.

__________, 1980b: Mesoscale convectivecomplexes. Bull. Amer. Meteor. Soc., 61, 1374-1387.

__________, and W. Dietrich, 1981: Synopticconditions associated with the simultaneousoccurrence of significant severe thunderstormsand flash floods. Preprints, 4th Conf.Hydrometeorology (Reno, NV), Amer. Meteor.Soc., Boston, 181-187.

__________, and C.A. Doswell III, 1982: Anexamination of jet stream configurations, 500 mbvorticity advection and low level thermaladvection patterns during extended periods ofintense convection. Mon. Wea. Rev., 184-197.

__________, C.F. Chappell, and L.R. Hoxit, 1979:Synoptic and mesoscale aspects of flash floodevents. Bull. Amer. Meteor. Soc., 60, 115-123.

__________, L.R. Hoxit, and C.F. Chappell, 1980:A study of tornadic thunderstorm interactions withthermal boundaries. Mon. Wea. Rev., 108, 322-336.

__________, and F. Caracena, 1978: Comparisonof meteorological aspects of the Big Thompsonand Rapid City flash floods. Mon. Wea. Rev., 106,375-389.

Magor, B.W., 1958: A meso-low associated with asevere storm. Mon. Wea. Rev., 86, 81-90.

__________, 1959: Meso-analysis: Someoperational analysis techniques utilized in tornadoforecasting. Bull. Amer. Meteor. Soc., 40, 499-511.

__________, 1971: Statistics of selected surfaceconditions found within the hour precedingtornado occurrence, having identified a meso low.Proceedings, 7th Conf. on Severe Local StormsKansas City, MO , Amer. Meteor. Soc., Boston, 17-22.

Mahrt, L., 1977: Influence of low-levelenvironment on severity of High Plains moist

96

convection. Mon. Wea. Rev., 105, 1315-1329.

Marshall, T.P., 1980: Topographic influences onAmarillo radar echo climatology. Master�s Thesis,Texas Tech. Univ., Lubbock, TX, 65 pp.

Marwitz, J.D., 1972: The structure and motion ofsevere hailstorms. Part I: supercell storms. J. Appl.Meteor., 11, 166-179.

McCarthy, J. and S.E. Koch, 1982: The evolution ofan Oklahoma dryline. Part I: A. meso- andsubsynoptic-scale analysis. J. Atmos, Sci., 39, 225-236.

Means, L.L., 1944: The nocturnal maximumoccurrence of thunderstorms in the MidwesternStates. Univ. Chicago, Dept. Meteor., Misc. Reps.,No. 16, 38 pp.

__________, 1952: On thunderstorm forecastingin the Central United States. Mon. Wea. Rev., 80,165-189.

__________, 1954: A study of the mean southerlywind-maximum in low levels associated with aperiod of summer precipitation in the middle west.Bull. Amer. Meteor. Soc., 35, 166-170.

McGuire, E.L., 1962: The vertical structure ofthree drylines as revealed by aircraft traverses.Rept. No. 7, Nat�l Severe Storms Lab., 10 pp.

McNulty, R.P., 1978: On upper tropospherickinematics and severe weather occurrence. Mon.Wea. Rev., 106, 662-672.

__________, 1980: Differential advection of wet-bulb potential temperature and convectivedevelopment. Preprints, 8th Conf. WeatherForecasting and Analysis Denver, CO , Amer.Meteor. Soc., Boston, 286-291.

Middleton, W.E.K., and A.F. Spilhaus, 1953:Meteorological Instruments. Univ. of TorontoPress, Toronto, 286 pp.

Miller, J.E., 1948: On the concept of frontogenesis.J. Meteor., 5, 169-171.

Miller, R.C., 1972: Notes on Analysis and Severe-Storm Forecasting Procedures of the Air ForceGlobal Weather Central. Air Weather ServiceTech. Rep. 200 (Rev.), Air Weather Service, ScottAir Force Base, IL, 190 pp.

Milne-Thomson, 1968: Theoretical

Hydrodynamics, 5th Ed., MacMillan, New York,743 pp.

Mitchell, K.E., and J.B. Hovermale, 1977: Anumerical investigation of the severethunderstorm gust front. Mon. Wea. Rev., 105,657-675.

Mogil, H.M., 1975: Pressure change-its use inforecasting the Atlanta tornado of March 24, 1975.Preprints, 9th Conf. Severe Local Storms Norman,OK , Amer. Meteor. Soc., Boston, 25-32.

Moller, A.R., 1979: The climatology and synopticmeteorology of southern plains� tornadooutbreaks. Master�s Thesis, Univ. of Oklahoma,Norman, OK, 70 pp.

Morgan, G.M., Jr., and R.G. Beebe, 1971: Analysisof the spacetime behavior of the field ofequivalent potential temperature during a severeweather situation. Preprints, 7th Conf. on SevereLocal Storms Kansas City, MO , Amer. Meteor.Soc., Boston, 54-59.

Neumann, J., 1951: Land breezes and nocturnalthunderstorms. J. Meteor., 8, 60-67.

__________, and Y. Mahrer, 1974: A theoreticalstudy of the sea and land breezes of circularislands. J. Atmos. Sci., 31, 2027-2039.

__________, and __________, 1975: Atheoretical study of the sea and land breezes ofcircular lakes. Mon. Wea. Rev., 103; 474-485,

Newton, C.W., 1950: Structure and mechanism ofthe pre-frontal squall line. J. Meteor., 7, 210-222.

__________, 1966: Circulations in large shearedcumulonimbus. Tellus, 18, 699-713.

__________, 1980: Overview on convectivestorm systems. Proceedings of CIMMS Symposium(Y.K. Sasaki, N. Monji, and S. Bloom, Eds.),Norman, OK, 3-107.

__________, and H.R. Newton, 1959: Dynamicalinteractions between large convective clouds andenvironment with vertical shear. J. Meteor., 15,483-496.

Ninomiya, K., 1971a: Dynamical analysis ofoutflow from tornado producing thunderstorms asrevealed by ATS III pictures. J. Appl. Meteor., 10,275-294.

97

__________, 1971b: Mesoscale modification ofsynoptic situations from thunderstormdevelopment as revealed by ATS III andaerological data, J. Appl. Meteor., 10, 1103-1121.

Nolen, R.H., 1959: A radar pattern associatedwith tornadoes. Bull. Amer. Meteor. Soc,, 40, 277-279.

Novlan, D.J., and W.M. Gray, 1974: Hurricane-spawned tornadoes. Mon. Wea. Rev., 102, 476-488.

Obrien, J.J. and R.D. Pillsbury, 1974: Rotary windspectra in a sea breeze regime. J. Appl. Meteor.,13, 820-825.

Oliver, V.J., and M.B. Oliver, 1945: Weatheranalysis from single station data. Handbook ofMeteorlogy (F.A. Barry, E. Bollay, and N.R. Beers,Ed�s), McGraw-Hill, New York, 858-879.

Orlanski, I., 1975: A rational subdivision of scalesfor atmospheric processes. Bull. Amer. Meteor.Soc., 56, 527-530.

Orville, H.D., and L.J. Sloan, 1970: A numericalsimulation of the life history of a rainstorm. J.Atmos. Sci., 27, 1148-1159.

Ostby, F.P., 1975: An application of severe stormforecast techniques to the outbreak of June 8,1974. Preprints, 9th Conf. Severe Local StormsNorman, OK , Amer. Meteor. Soc., Boston, 7-12.

Palmen, E., and C.W. Newton, 1969: AtmosphericCirculation Systems. Academic Press, 603 pp.

Panofsky, H., 1964: Introduction to DynamicMeteorology. Penn. State Univ. Press, UniversityPark, 243 pp.

Parmenter, F.C,, 1976: Low-level moistureintrusion from infrared imagery. Mon. Wea. Rev.,104, 100-184.

Pearson, A.D., and A.F. Sadowski, 1965:Hurricane-induced tornadoes and theirdistribution. Mon. Wea. Rev., 93, 461-464.

Petterssen, S., 1956a: Weather Analysis andForecasting., Vol. I (Motion and Motion Systems ,2nd Ed., McGraw-Hill, New York, 428 pp.

__________, 1956b: Weather Analysis andForecasting. Vol. II (Weather and WeatherSystems , 2nd Ed., McGraw-Hill, New York, 266

pp.

__________, and J.M. Austin, 1942: Fronts andfrontogenesis in relation to vorticity. Pap. in Phys,Ocean. and Meteor., VII, No. 2, Cambridge andWoods Hole, 37 pp.

Pielke, R.A., 1973: A three-dimensional model ofthe sea breeze. NOAA Tech. Memo. ERL WMP0-2, Weather Modification Program Office, Boulder,CO, 136pp.

Pitchford, K.L., and J. London, 1962: The low-leveljet as related to nocturnal thunderstorms overmidwest United States. J. Appl. Meteor., 1, 43-47.

Porter, J.M., L.L. Means, J.E. Hovde and W.B.Chappell, 1955: A synoptic study on the formationof squall lines in the North Central United States,Bull. Amer. Meteor, Soc., 36, 390-396.

Prandtl, L., and O.G. Tietjens, 1934: Fundamentalsof Hydro- and Aeromechanics. DoverPublications (1957), New York, 270 pp.

Prosser, N.E., and D.S. Foster, 1966: Upper airsounding analysis by use of an electroniccomputer. J. Appl. Meteor., 5, 296-300.

Purdom, J.F.W., 1973: Satellite imagery and themesoscale forecast problem. Preprints, 8th Conf.Severe Local Storms (Denver, CO , Amer. Meteor.Soc., Boston, 244-251.

__________, 1979: The development andevolution of deep convection. Preprints, 11thConf. on Severe Local Storms Kansas City, MO ,Amer. Meteor. Soc., Boston, 143-150.

Rasmussen, E., 1977: The polar low as a CISK-phenomenon. Rep. No. 6, Inst. Teoret. Meteor.,Kobenhavns Universitet, Copenhagen, 77 pp.

Reap, R.M., 1968: Prediction of temperature anddew point by three dimensional trajectories. Wea.Bur. Tech. Memo., WBTM TDL-15, 20 pp.

__________, 1972: An operational three-dimensional trajectory model. J. Appl. Meteor., 11,1193-1202.0

__________, and D.S. Foster, 1979: Automated12-36 hour probability forecasts of thunderstormsand severe local storms. J. Appl. Meteor., 18,1304-1315.

__________, and S.J. Weiss, 1982: Development

98

and evaluation of an automated convectiveoutlook (AC) chart. Preprints, 12th Conf. on SevereLocal Storms San Antonio, TX , Amer. Meteor.Soc., Boston, 110-114.

Reed, R.J., 1979: Cyclogenesis in polar airstreams. Mon. Wea. Rev., 107, 38-52.

__________, 1980: Destructive winds caused byan orographically induced mesoscale cyclone.Bull. Amer. Meteor. Soc., 61, 1346-1355.

Rhea, J.O., 1966: A study of thunderstormformation along dry lines. J. Appl. Meteor., 5, 59-63.

Rossby, C.G., 1938: On the mutual adjustment ofpressure and velocity distributions in certainsimple current systems. J. Marine Res. 1, 239-263.

__________, 1940: Planetary flow patterns in theatmosphere. Quart. J. Roy. Meteor. Soc , 66,Suppl., Toronto, 68-97.

Rotunno, R. and J.B. Klemp, 1982: The influence ofthe shear-induced pressure gradient onthunderstorm motion. Mon. Wea. Rev., 110, 136-151.

Sangster, W.E., 1958: An investigation of nighttimethunderstorms in the United States. Tech. Rep. No.5, Contract No. AF 19 (604)-2179, Univ. ofChicago, Dept. of Meteor., 37 pp.

__________, 1960: A method of representing thehorizontal pressure force without reduction ofstation pressures to sea level. J. Meteor., 17,

__________, 1967: Diurnal surface geostrophicwind variations over the Great Plains.Proceedings, 5th Conf. on Severe Local Storms(St. Louis, MO), Amer. Meteor. Soc., Boston, 146-154.

Saucier, W.J., 1955: Principles of MeteorologicalAnalysis. The University of Chicago Press, 438 pp.

Schaefer, J.T., 1973a: The motion and morphologyof the dryline. NOAA Tech. Memo. ERL NSSL-66,[NTIS Accession No. COM-74-10043] Norman,OK, 81 pp.

__________, 1973b: On the computation of thesurface divergence field. J. Appl. Meteor., 12, 546-547.

__________, 1974a: A simulative model of dry

line motion. J. Atmos. Sci., 31, 956-964.

__________, 1974b. The life cycle of the dryline.J. Appl. Meteor., 13, 444-449.

__________, 1975: Nonlinear biconstituentdiffusions: A possible trigger of convection. J.Atmos. Sci., 32, 2278-2284.

__________, and C.A. Doswell III, 1979: On theinterpolation of a vector field. Mon. Wea. Rev.,107, 458-476.

__________, and __________, 1980: The theoryand practical application of antitriptic balance.Mon. Wea. Rev., 108, 746-756.

__________, and R.L. Livingston, 1982: A thermo-hydrodynamic indicator of severe convectivepotential, Preprints, 12th Conf. Severe LocalStorms (San Antonio, TX , Amer. Meteor. Soc.,Boston, 71-74.

__________, L.R. Hoxit and C.F. Chappell, 1982:Thunderstorms and their mesoscale environment.Thunderstorms � A Social and TechnologicalDocumentary E. Kessler, Ed. (in press).

Shreffler, J.E., and F.S. Binkowski, 1981:Observations of pressure jump lines in theMidwest, 10-12 August 1976. Mon. Wea. Rev.,109, 1713-1725.

Showalter, A.K.. 1953: A stability index forthunderstorm forecasting. Bull. Amer. Meteor.Soc., 6, 250-252.

Siebers, J.O., F. Hidalgo, S.A. Tegtmeier and M.Young, 1975: Guide for using GOES/SMS imageryin severe weather forecasting. Unnumberedtechnical guide from USAFETAC, Andrews AFB,Md., 56 pp.

Smith, J.S., 1965: The hurricane-tornado. Mon.Wea. Rev., 93, 453-459.

Snellman, L.W., 1977: Operational forecastingusing automated guidance. Bull. Amer. Meteor.Soc., 58, 1036-1044.

Snider, C.R., 1977: A look at Michigan tornadostatistics. Mon. Wea. Rev., 105, 1341-1342.

Staff, NSSP, 1963: Environmental andthunderstorm structure as shown by NationalSevere Storm Project observations in Spring 1960-61. Mon. Wea. Rev., 91, 271-292.

99

Stone, P.H., 1966: Frontogenesis by horizontalwind deformation fields, J. Atmos. Sci., 23, 455-465.

Sutcliffe, R.C., 1947: A contribution to the theoryof development. Quart. J. Roy. Meteor. Soc., 73,370-383.

__________, and A.G. Forsdyke, 1950: Thetheory and use of upper air thickness patterns inforecasting. Quart. J. Roy. Meteor. Soc., 76, 189-217.

Tegtmeier, S.A., 1974: The role of the surface,subsynoptic low pressure system in severeweather forecasting. Master�s thesis, Univ. ofOklahoma, Norman, OK, 66 pp.

Tepper, M., 1950: A proposed mechanism of squalllines: The pressure jump line. J. Meteor., 7, 21-29.

__________, 1959: Mesometeorology - the linkbetween macroscale atmospheric motions andlocal weather. Bull. Amer. Meteor. Soc., 40, 56-72.

Tibaldi, S., A. Buzzi, and P. Malguzzi, 1980:Orographically induced cyclogenesis: Analysis ofnumerical experiments. Mon. Wea. Rev., 108,1302-1314.

Trenberth, K.E., 1978: On the interpretation of thediagnostic quasi-geostrophic omega equation.Mon. Wea. Rev., 106, 131-137.

Tsui, T.L. and L.R. Brody, 1982: Objective stormtracking system. Preprints, 9th Conf. on WeatherForecasting and Analysis (Seattle, WA), Amer.Meteor. Soc., Boston, 289-295.

Uccellini, L.W., 1975: A case study of apparentgravity wave initiation of severe convectivestorms. Mon. Wea. Rev., 103, 497-513.

__________, and D.R, Johnson, 1979: Thecoupling of upper and lower tropospheric jetstreaks and implications for the development ofsevere convective storms, Mon. Wea. Rev., 107,682-703.

Ulanski, S.L., and M. Garstang, 1978: The role ofsurface divergence and vorticity in the life cycleof convective rainfall. Part I: Observations andanalysis. J. Atmos. Sci., 35, 1047-1062.

van Bemmelen, W., 1922: Land und Seebrise inBatavia. Beitr. Phys. Frei. Atmos., 10, 169-177.

Wallace, J.M., 1975: Diurnal variations inprecipitation and thunderstorm frequency over theconterminous United States. Mon. Wea. Rev., 103,406-419.

Weaver, J.F., 1979: Storm motion as related toboundary-layer convergence. Mon. Wea. Rev.,107, 612-619.

Weisman, M.L., and J.B. Klemp, 1982: Thedependence of numerically simulated convectivestorms on vertical wind shear and buoyancy.Mon. Wea. Rev., 110, 504-520,

Weiss, S.J. and E.W. Ferguson, 1982: Anexperiment in medium range quantitative severelocal storm forecasting: Preliminary results.Preprints, 12th Conf. Severe Local Storms (SanAntonio, TX), Amer. Meteor. Soc., Boston, 116-119.

__________, C.A. Doswell III, and F.P. Ostby,1980a: Comments on �Automated 12-36 hourprobability forecasts of thunderstorms and severelocal storms�. J. Appl. Meteor., 19, 1328-1333.

__________, D.L. Kelly, and J.T. Schaefer, 1980b:New objective verification techniques at theNational Severe Storms Forecast Center.Preprints, 8th Conf. on Weather Forecasting andAnalysis (Denver, CO), Amer. Meteor. Soc.,Boston, 412-419.

Welander, P., 1955: Studies on the generaldevelopment of motion in a two-dimensional idealfluid. Tellus, 7, 141-156.

Weldon, B.R., 1975: Part I, Basic cloud systems.Course Notes, Satellite Training, ApplicationsGroup, National Environmental Satellite Service,Camp Springs, MD, 11 pp.

__________, 1979: Part IV, Cloud patterns and theupper air wind field. Course Notes, SatelliteTraining, Applications Group, NationalEnvironmental Satellite Service, Camp Springs,MD, 80 pp.

Whiting, R.M., 1957: The surface chart as asynoptic aid to tornado forecasting. Bull, Amer.Meteor. Soc., 38, 353-356.

Whitney, L F., Jr., 1977: Relationship of thesubtropical jet stream to severe local storms. Mon.Wea. Rev., 105, 398-412.

__________, and J.E. Miller, 1956: Destabilization

100

by differential advection in the tornado situationof. 8 June 1953. Bull. Amer. Meteor. Soc., 37, 224-229.

__________, A. Timchalk and T.I. Gray, Jr., 1966:On locating jet streams from TIROS photographs.Mon. Wea. Rev., 94, 127-138.

Wiin-Nielsen, A., 1959: On the application oftrajectory methods in numerical forecasting.Tellus, 11, 180-196.

Williams, D.T., 1960: The role of a subsidencelayer. In �The Tornadoes at Dallas, Tex., April 2,1957�, Wea. Bur. Res. Paper No. 41, 143-158.

Wilson, J.W., and E. Kessler, III, 1963: Use of radarsummary maps for weather analysis andforecasting. J. Appl. Meteor., 2, 1-11.

Zimmerman, L.I., 1969: Atmospheric wakephenomenon near the Canary Islands. J. Appl.Meteor., 8, 896-907.

Zipser, E.J., 1977: Mesoscale and convective-scale downdrafts as distinct components of squall-line structure. Mon. Wea. Rev., 105, 1568-1589.

101

NOAA SCIENTIFIC AND TECHNICAL PUBLICATIONS

The National Oceanic and Atmospheric Administration was established as part of the Department of Commerceon October 3, 1970. The mission responsibilities of NOAA are to assess the socioeconomic impact of natural andtechnological changes in the environment and to monitor and predict the state of the solid Earth, the oceans andtheir living resources, the atmosphere, and the space environment of the Earth.The major components of NOAA regularly produce various types of scientific and technical information in thefollowing kinds of publications:

PROFESSIONAL PAPERS — Important definitive research results, major techniques, and special investigations.CONTRACT AND GRANT REPORTS — Reports prepared by contractors or grantees under NOAAsponsorship.ATLAS — Presentation of analyzed data generally in the form of maps showing distribution of rainfall, chemicaland physical conditions of oceans and atmosphere, distribution of fishes and marine mammals, ionosphericconditions, etc.TECHNICAL SERVICE PUBLICATIONS — Reports containing data, observations, instructions, etc. A partiallisting includes data serials; prediction and outlook periodicals; technical manuals, training papers, planning reports,and information serials; and miscellaneous technical publications.TECHNICAL REPORTS — Journal quality with extensive details, mathematical developments, or data listings.TECHNICAL MEMORANDUMS — Reports of preliminary, partial, or negative research or technology results,interim instructions, and the like.Information on availability of NOAA publications can be obtained from:ENVIRONMENTAL SCIENCE INFORMATION CENTER (OA/D812)ENVIRONMENTAL DATA AND INFORMATION SERVICENATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATIONU.S. DEPARTMENT OF COMMERCERockville, MD 20852

102

NOAA TECHNICAL MEMORANDA

National Weather Service

National Severe Storms Forecast Center

The National Severe Storms Forecast Center (NSSFC) has the responsibility for the issuance of severethunderstorm and tornado watches for the contiguous 48 states. Watches are issued for those areas wherethunderstorms are forecast to produce one or more of the following: (1) hailstones of 3/4-inch diameter or greater,(2) surface wind gusts of 50 knots or greater, or (3) tornadoes.

NOAA Technical Memoranda in the NWS, NSSFC subseries are produced under the technical guidance of theNSSFC, Techniques Development Unit. They facilitate rapid dissemination of material of general interest in thefield of severe storm meteorology. These papers may be preliminary in nature, and may be formally publishedelsewhere at a later date.

These papers are available from the National Technical Information Service (NTIS), U.S. Department ofCommerce, Sills Building, 5285 Port Royal Road, Springfield, Virginia 22161. Price varies, $3.50 for microfiche.

Previous issues in this series:

No. 1. New Severe Thunderstorm Radar Identification Techniques and Warning Criteria: A Preliminary Report.Leslie R. Lemon, July 1977, 60 p. (PB 273049).

No. 2. A Subjective Assessment of Model Initial Conditions Using Satellite Imagery. John E. Hales, Jr., November1981, 19 p. (PB 291593).

No. 3. Severe Thunderstorm Radar Identification Techniques and Warning Criteria. Leslie R. Lemon, April 1980,60 p. (PB 231439).

No. 4. The Enhanced-V, A Satellite Observable Severe Storm Signature. Donald W. McCann, March 1981, 31 p.(PB 230336).

the operational meteorology of convective weather volume i: operational mesoanalysis

Documents

volume series

thesecond volume

useful tothe readers

worlds supply of themand

useful waythan

scale vertical motion

stormscale analysis

centerkansas city