AWE: Aviation Weather Data VisualizationEnvironment

Lilly Spirkovska�

NASAAmesResearch Center, MS269-3,Moffett Field, CA,94035-1000,USA

SureshK. Lodha

ComputerScience, University of California, SantaCruz,CA95064,USA

Abstract

Weatheris one of the major causesof aviation accidents.Generalaviation (GA) flightsaccountfor 92% of all the aviation accidents.In spite of all the official and unofficialsourcesof weathervisualizationtoolsavailableto pilots,thereis anurgentneedfor visual-izing severalweatherrelateddatatailoredfor generalaviationpilots.Oursystem,AviationWeatherDataVisualizationEnvironment(AWE), presentsgraphicaldisplaysof meteoro-logical observations,terminalareaforecasts,andwindsaloft forecastsontoa cartographicgrid specificto thepilot’sareaof interest.Decisionsregardingthegraphicaldisplayandde-signaremadebasedoncarefulconsiderationof userneeds.Integral visualdisplayof theseelementsof weatherreportsis designedfor theuseof GA pilots asa weatherbriefingandrouteselectiontool. AWE provideslinking of theweatherinformationto theflight’s pathandschedule.Thepilot caninteractwith thesystemto obtainaviation-specificweatherfortheentireareaor for his specificrouteto explorewhat-if scenariosandmake ”go/no-go”decisions.Thesystem,asevaluatedby somepilots at NASA AmesResearchCenter, wasfoundto beuseful.

Key words: weathervisualization,generalaviation,routeselection,userevaluation.

1 Introduction

Weatheris oneof the major causesof aviation accidents.Accordingto a NASAplanninggroup,it is estimatedthat approximately30 percentof commercialair-�

Correspondingauthor. Faxnumber:650-604-4036.Email addresses:spirkov@wolfie.arc.nasa.gov (Lilly Spirkovska),lodha@cse.ucsc.edu

(SureshK. Lodha).

Preprintsubmittedto Elsevier Preprint 27October2000

craft accidentshave weatherasa contributing factor. Although,thepercentageofaccidents(0.4 accidentper100,000departures),hasremainedflat in thepastfiveyearsaccordingto the NationalTransportationSafetyBoard,the total numberofaccidentswill increasebeyondpublicexpectationsdueto theprojectedincreaseinthe numberof flights in a few years.Therefore,in 1997,a nationalgoal wasde-finedin theUnitedStatesto reducethefatalaviationaccidentrateby 80percentbytheyear2007.TheUSFederalAviationAdministrationhaslaunchedanaggressiveaviation weatherresearchprogramandis pouringmillions of dollarsinto differentpartsof theaviationweatherresearch[25].

Significantadvanceshavebeenmadein thelastdecadein bothweatherforecastingandweathervisualizationfor avarietyof audiencesincludingscientists,forecastersandthegeneralpublic [11]. ProfessionalTV productionsystemsfor weatherpre-sentationsto thegeneralpublichavebeenin existencefor morethantenyearsandareconstantlybeingupgraded.An exampleis the TriVis systemoperatingsince1993[33]. An interactive 3D weathervisualizationsystemVISUAL for scientistshasbeeninstalledin the GermanMeteorologicalOffice (DWD) in collaborationwith FraunhoferIGD sincemid-1990s.PersonalizedWeather-on-Demandproductsarealsobeingofferedthroughtheinternetsince1998.Augmentedreality weathervisualizationsystemsarebeingdeveloped.Many of thesesystemsalso incorpo-ratenumericalweatherprediction(NWP)forecastsandobservations.However, theusersarerequestingtailoredvisualizationtools for their specificneeds.In partic-ular, theexisting anddevelopingweatherforecastingandvisualizationtechnologyneedsto beharnessedappropriatelyfor thebenefitof thepilots.

Broadly speaking,all aviation activities canbe classifiedinto commercialairlineoperations,generalaviations(GA) andmilitary operations.It is importantto under-standthedifferencesbetweentheneedsof weathervisualizationandrouteselectionfor commercialandgeneralaviation pilots. It is interestingto notethatonly 4% ofaircraft areassociatedwith commercialairline operations.Indeed,the restof theaviation activities, referredto asgeneralaviation, accountfor 96%of all aircraft.Thegeneralaviation aircraft rangefrom single-seat,single-engine,pistonaircraftto businessjets thatcanfly ashigh asair carriersbut typically carry lessthan20passengers.More importantly, GA pilots cover the full spectrumof flying experi-ence,from studentpilots with 20 hoursof experienceto accomplishedpilots withtensof thousandsof hours.In contrast,commercialpilots have substantialflyingexperience,fly powerful equipmentsuchastheBoeing747,andhaveanetwork ofsupportpeopleon the groundat the Airline OperationsCenterandthe FAA’s airtraffic controlcenters.In addition,GA pilotsoftenfly at loweraltitudes,fly slower,carrylessfuel onboard,andthuscovershorterdistancesin asingleflight. A typicalflight coversabout400miles in 4 hours.Becauseof the lower altitudeandslowerspeed,they spendmoretime in adverseweatherconditions.In contrast,air carrieraircraftareableto fly above muchof theweatherfor a largeportionof theflight.Commercialair carriersaccountfor 85%of all thepassengerscarried,67%of thetotal milesflown, but only 40%of thetotal hoursflown dueto thehigh speed.In-

2

deed,althoughGA accountsfor only 60%of thetotal hoursflown, it accountsforover92%of thetotal accidents.Thefatalaccidentratefor air carriersis 0.15acci-dentsper100,000hoursflown, whereasit is nearlyanorderof magnitudegreaterfor GA at 1.4 accidentsper 100,000hoursflown. Of theseaccidents,more than15%canbeattributedto weather[5].

Thefocusof thiswork is to provideweathergraphicsusefulto thegeneralaviationpilots for routeselectionandweatherbriefing.Themostimportantofficial sourceof aviation weatherreportsto the GA pilots in the United Statesis Direct UserAccessTerminals(DUATs).In addition,weatherbriefingscanbeobtainedvia tele-phone,aircraft radio,or infrequently, in personfrom Flight ServiceStation(FSS)specialists(employeesof FederalAviation Administration(FAA)) [22,23,44],orvia computerfrom theDUATs (Direct UserAccessTerminal)system[8]. Most ofthesebriefingsaretextual or verbalandareobtainedprior to flight. Face-to-facebriefingswith anFSSspecialisthave theadvantageof accessto graphicaldisplaysof the data.This advantageis outweighedby the limited availability of FSSfa-cilities. Unfortunately, DUATs doesnot provide visualizationof threeof themostimportantelementsof aweatherbriefing:airport-specificcurrentweatherobserva-tions (meteorologicalobservations,or METARs), terminalareaforecasts(TAFs),andwindsaloft forecasts.

Perhapsthemostimportantunofficial sourceof weatherinformationto pilots is theNationalWeatherService(NWS)websites[35]. Althoughthesewebsitesprovidea greatervarietyof weathergraphics,the informationprovidedto thepilots asso-ciatedwith airportsandterminalareasis difficult to use.Thepilot hasto pick anairportin orderto displaytheweather-specificinformation,whichis thendisplayedtextually without filtering asshown in Figure1, which is difficult to grasp.More-over, thisweatherinformationis hardto relateto theflight’sscheduleandpath.

Fig. 1. TheTAF squarescanbeselectedwith themouseto obtainthe full text associatedwith aNWSTAF display.

In a recentarticle[25], Perrystatesthat”Unfortunately, thetypeof weatherinfor-mationavailableto a commercialpilot is scanty;a sheetof weatherdataprintedout beforetakeoff, may be outdatedandof minimal use.” The inadequacy of theweatherinformationprovidedto thepilots throughtheDUATs andtheNWSoftenmake themturn to severalotherunofficial sourcesof weatherinformationsuchas

3

thetelevisionnewsweatherreports,theWeatherChannel[6],or avarietyof weatherwebsites[35,19,45,34,46].However, theseunofficial sourcesoftenprovideonly ageneralview of whattheweatherwill belike.

In thiswork, wepresentAWE (AviationWeatherDataVisualizationEnvironment)thatfocusesonextractingaviation-specificweatherinformationfrom textualdocu-ments,visualizingthis information,linking it to theflight’spathandschedule,andproviding asimpleuserinterfaceto thepilot to controlthedisplay. To thispurpose,we focusspecificallyon thoseweatherproductsthat have not beenvisualizedorreadilyaccessiblethroughDUATs,NWSor many othersourcesmentionedbefore.Threemainexamplesof theseproductsare:airport-specificcurrentweatherobser-vations(meteorologicalobservations,or METARs),terminalareaforecasts(TAFs),andwindsaloft forecasts.AWE provideslinking of this informationto theflight’spathandschedule,andthusfacilitatesansweringof questionstailoredto thepilot’sneedswhich aredifficult to addressusingcurrentweatherproducts.For example,AWE canbeusedto answerthefollowing questionquickly: Will thecloudsbelowenoughto requireflight underinstrumentflight rules,or canI fly undervisualflightrules?.

It is possibleto extendAWE by addingadditionalweather-relatedinformationsuchasthe informationon turbulence,wake vortices,icing, lightning andprecipitationinformation,stormcells,ceiling, etc.,or by addingnumerouscontinuousweatherinformationsbasedonnumericalweatherpredictionsystems.Laterin Section4,wereporton somepreliminarypilots’ feedbackon thesequestions.However, clearlyadditionalresearchis neededto decidehow muchadditionalweatherinformation(andwhich ones)canbe presentedin a displayso that the visualizationremainsintuitive anduncluttered.Sucha studyis beyond the scopeof this work. Rather,AWE is one steptowardsthe importantgoal of ”providing weatherinformationrelative to the pilot’s flight path,presentit to the pilot in the cockpit in an easy-to-interpretgraphicalformat,andgive him decision-makingaidsto help him usethatinformation....” outlinedby Stough,themanagerof NASA’s aviation weatherinformationsystems(AWIN) projectat theLangley ResearchCenter[25].

Thereis oneadditionalveryimportantquestion:theavailability of in-flight weatherinformationto the pilots. Themostimportantissuehereis the communicationofthe information to the pilots and the advancementin the datalink technologies.In this areaagain,currently the commercialpilot getsupdatesfrom the groundstaff througha text printervia 2400baudmodem,or hearsanecdotalreportsfromotherpilots in thearea.Currently, AWE is operationalon thegroundasa briefingand routing tool for pilots prior to flight. Becauseof the minimal datatransferrequirements,AWE canbeeasilyincorporatedasanin-flight decision-makingtool.

Therestof thepaperis organizedasfollows.Section2 describesthebackground,previousandrelatedwork. Section3 presentsAviationWeatherDataVisualizationEnvironment(AWE) includinggraphicaldesign,displayandflight pathplanning

4

issues.Section4 describesusers’feedbackandexperiences.Finally, Section5 con-cludeswith asummaryanddirectionsfor futureresearch.

2 Background and Previous Work

We begin by describingin detail themostimportantsourceof official weatherin-formationavailableto thepilots– DirectUserAccessTerminals(DUATs).Wethendescribeanimportantunofficial sourceof weatherinformationavailableto pilots–AviationDigital DataService(ADDS) by theNationalWeatherService(NWS).

2.1 DUATs

Till early 1990s,Flight ServiceStationSpecialists(FSS)were the only officialsourceof informationto thepilots.With the introductionof theDUATs serviceinmid 90s,pilotsweregiventheoptionof receiving moreautomatedofficial briefingsvia a modemdial-in or throughan internetbrowser. In the late1990s,DUATs in-troducedgraphicaldisplaysto helppilotsvisualizethe”big picture”usingweathergraphics.

DUATs,Direct UserAccessTerminalsystem,is offeredby privatecompaniesun-dercontractto theFAA. It is availableto all pilots, from studentpilot onward.Weusedthe DynCorp(previously known asGTE) DUAT system.It obtainsits datafrom the FAA, which in turn obtainssomeof its datafrom the NationalWeatherService.A DUATs areabriefing provides the following information in a textualformat:

� Area Forecastincluding positionof fronts, pressuresystems,wind conditions,cloudlayers,weather(suchasrain),andvisibility conditions,� SevereWeatherWarnings,� SIGMETs (SignificantMeteorologicalConditions)and Convective SIGMETssuchasthunderstorms,� AIRMETs (Airman’sMeteorologicalInformation)for turbulence,mountainob-scuration,widespreadlow visibility conditions,andicingconditionsandfreezinglevels,� SurfaceObservationsor METARs(MeteorologicalObservations)of currentcon-ditions,includingceilings,visibility, wind,barometricpressure,temperatureanddew pointsfor certainairports,� Pilot Reports,� RadarSummariesthat textually provide informationaboutechos,echomove-ment,andechointensity,

5

� TAFs(TerminalAreaForecasts)includingceiling,visibility, andwind forecastsfor certainairports,� WindsAloft Forecastsfor relevantsitesataltitudesof 3000feetto 39000feet,atvariousincrements,and� NOTAMs (Noticesto Airmen), which provide information on suchthings asairportclosures,unlightedobstructions,outof serviceequipmentsuchasrunwaylights,etc.

Someof theaboveinformationis alsopresentedto thepilotsvisually. In particular,DUATS provides weatherchartsincluding surfaceanalysisand surfaceforecastchartsthatshow thecurrentandforecastlocationof high/low pressuresystemsandwarm/coldfronts,areasof precipitation,andinfraredsatellitecharts.An examplechartis shown in Figure2. In addition,thesechartsalsodisplayvisibility conditionscrucial to pilots. Therearethreetypesof visibility conditionsthatpilots typicallyuse:

� Instrumentflight rules(IFR).Thisisdefinedasvisibility � 3 milesand/orceiling� 1000feet.� Marginalvisualflight rules(MVFR). This is definedas(3 miles � visibility ���5 miles)and/or(1000feet � ceiling ��� 3000feet).� Visual flight rules(VFR). This is definedasvisibility � 5 milesandceiling �3000feet.

Thesevisibility conditionsaredisplayedascolor-codedregionsasshown in Fig-ure3. Thechartsareveryeffective in providing a broad(nation-wide)overview oftheweather, but they do not provide informationaboutspecificlocations,suchasairportsalongthepilot’s route.In particular, DUATsdoesnotprovidevisualinfor-mationon METARs andTAFs. Moreover, this weatherinformationis not relatedto theflight’spathor schedule.

Fig.2.An exampleof graphicsavailablethroughDUATs.Thissurfaceanalysischartshowslow pressuresystems(”L”), coldfronts(theblueline acrosssouthernCalifornia)andwarmfronts(theredline from Colorado,throughKansasandOklahoma),areasof precipitation(suchasthe rain in NorthernCaliforniaandthunderstormsin easternKansasandeasternOklahoma).

6

Fig.3.Color-codedvisibility conditionsin DUATs.Poorvisibility regionsareshown in redandmarginal visibility regionsareshown in blue.

2.2 NWS

We now describeweatherinformationandgraphicsavailableto pilots throughtheWorld Wide Web site by the Aviation WeatherCenterof the National WeatherService(NWS)[35] initially releasedin 1997.UnliketheDUATssite,theNWSsiteis experimentalandunofficial. However, it providesa greatervarietyof graphics,and allows pilots to zoom in to get more specificinformation aboutan areaofinterest.

In additionto thestandardchartsprovidedby DUATs, NWS doesattemptto pro-vide moreinformationvisually on METARs andTAFs.Figure4 showsa METARdisplayprovided by the NWS site.The color codingreflectsthe visibility condi-tions.Theamountthedisk is filled reflectsthecloudcoverageamount,with emptydisksrepresentingclearconditions,partiallyfilled disksrepresentingfew, scattered,andbrokenclouds,andcompletelyfilled for overcastconditions.Additional infor-mationsuchaswind speedanddirection,temperature,dew point,andairportiden-tifier canbe associatedwith the disksby selectingthe properoptionsthroughaninteractive menu.Windsaloft displayis shown asbarbs(Figure5) discussedlaterin Section3.2.

TheTAF displayshown in Figure1 is lesspowerful. Thestandarddisplayshowsa black squareover airportsthat have TAF reports.No graphicaldisplayof theforecastis available.Rather, the pilot can get a textual display of an individualTAF by selectingtheappropriatesquare.As shown in Figure1, thetextualdisplayobscuresthesurroundingregion.

In additionto DUATs andNWS, therearemany othernon-aviation weatherwebsites[6,19,45,20]thatcouldalsoproveusefulasasupplementto official pre-flightbriefings.For example,weathergraphicsavailableto the generalpublic, suchasthoseshown on television newscastsor theWeatherChannel,do a reasonablejobof displayinglots of weatherinformationthatpilots find useful.Overall, all these

7

Fig. 4. A METAR displayavailablethroughtheNationalWeatherServicewebsite.Eachcircle representsaMETAR report.Thecolor-codingprovidessomeinformationoncurrentvisibility conditions,suchasvisualflight rulesin effect (greencircles)or instrumentflightrulesin effect (redcircle).Cloudcoverageis depictedby progressively filling in thedisksfor greatercloudcoverage.

Fig. 5. Windsaloft displayusingbarbswithin ADDS systemby NWS.

sourcesput togetherprovide adequateinformationon low/high pressuresystems,cold/warmfronts,cloudsandcloudmovement(throughsequencesof satelliteim-ages),andareasof precipitation(throughsequencesof radarimages).Someof thesesourcesalsoprovidegraphicsfor forecastsfor wide-spreadareasof thunderstorms,highwind speeds,or fog.Thus,visualizationof mostof thetextualDUATsbriefingis alreadyavailable.

In this work, we focuson visualizingwhat hasbeenneglectedso far – the latestweatherreportsfor selectedairports(METARs),andtheforecastreportsfor theseairports(TAFs). To thesedisplays,we also add a winds aloft display. AlthoughNWS displayswindsaloft reports,we provide additionalfunctionality to help thepilot planhis flight. More importantly, the integratedvisualdisplayof thesethreeelementsallows a pilot to morequickly recognizeandunderstandthecurrentandforecastweatherin an areaof interestalong the flight’s path.AWE presentsan

8

intuitive,graphicalpresentationof theseelementsandprovidesanenvironmenttodeterminealternateroutesincludingaltitudeor alternatedestinations,andexplorehis/herrouting optionswith quick ”what-if” scenarios.We expect that the easyinteractionimprovesthe pilot’s understandingof currentandforecastconditionsduringthepre-flightbriefingin orderto makeaninformeddecisionaboutwhetherto takeoff or delaytheflight andto planasaferoute.

2.3 RelatedWork

Thereis a large body of literatureon weathervisualizationandrelatedproducts[37]. Visualizationof large scientificdatafor earthsciences[12], oceanography[31], meteorology[39], climate modeling [17], and for environmentaldecisionmaking[28,29,7,27]hasbeenaroundfor morethana decade.In additionto DU-ATs and NWS, we have also mentioneda numberof weatherrelatedweb sites[1,6,19,20,45,35,34,46]andprofessionalweatherpresentationsystems[33]. A num-berof researchprojectsareunderway to build moreaccurateweatherforecastingtools[38,18].Wood[48] describesasystemfor presentingenvironmentaldataovertheworld wide web. Haase[11] describesaninteractive 3D weathervisualizationsystemfor scientists.Treinish[40] emphasizestask-specificvisualizationandcate-gorizesdifferentweathervisualizationtoolsbasedon thetypeof visualization(2Dor 3D) andsupport(level of interactive analysisandbrowsingcapability)into dif-ferentclasses.In particularhementionedAdvancedWeatherInformationProcess-ing System(AWIPS)by NWS, Vis5D developedby theUniversityof Wisconsin,andeffortsby ForecastSystemsLaboratory(FSL).

Thestate-of-the-artin weathervisualizationfor aviation-specificneedsis alsoad-dressedby several researchers[25,8,13,22,23,44].Numerousaviation weathervi-sualizationefforts areunderway includingthoseat NOAA, FAA, NRL, RockwellScienceCenter, Honeywell International,WSICorporation,BFGoodrich,EchoFlight,andtheMIT Lincoln Lab.

In particular, Perry[25] emphasizestheneedfor developingmorepreciseweatherforecastingtools for turbulence,wake vortices,icing, and fog. We have alreadydescribedDUATS andNWS in Sections2.1and2.2 respectively. Theweathervi-sualizationsystemdescribedby Scanlon[32] perhapscomesclosestto ourwork.

Scanlonpresentsfour weathermapsto thepilots.Thesemapsarea nationalradarmosaicfor precipitationdata,an air-to-groundlighteningstrike map,a categorymapdisplayingvisibility conditionsrelevant to commercialairline flights, andaceiling/visibility mapshown in Figure6. Of thesefour maps,the lastmapis mostrelevant to our work. It presentsfour elementsof a METAR/TAF: ceiling (lowestbroken (BKN) or overcast(OVC) cloud layer),visibility, wind speedonly if it isabove 30 knots,anda notationfor theexistenceof precipitationor anotherhazard

9

known to exist at the airport. The glyph usedis shown in Figure6 andconsistsof two stacked rectanglescolor-codedto signify visibility andceiling conditionswith a gapwhich flasheswhite for winds above 30 kts, an adjacentred squareifa flight hazardexists,andthecharacter”P” if precipitationis reported.Furtherin-formation,in the form of the original METAR/TAF text, is availableby selectingthe glyph for anairport.By selectinga glyph, the pilot canhave the METARs (ahistoryof thelastfivereports)for asingleairportdisplayedor themostrecentTAFfor thatairport.TheoriginalTAF is displayedwithoutany filtering basedonantic-ipatedarrival time.Thetextual informationis displayedasa separatewindow thatobscurestheceiling/visibility mapsogeographicalcontext is temporarilycovered.Scanlondesignedhis systemfor in-flight useby thecommercialairline pilots andtestedthe systemusingsimulateddata.AWE, in contrast,is designedfor generalaviationflights.

Fig. 6. Ceiling/visibility mapby Scanlon[32].

Many otherresearchershavefocusedondifferentaspectsof aviationneedsthatarecomplementaryto theweathervisualizationfocusof AWE. PruynandGreenberg[26] visualizeddesiredheadingandaltitudeat differentpositionsin the airspacesurroundingthe airport to facilitatethe approachandlandingat airports.Thoughtheir emphasisis not on visualizingweatherdata,they do considershowing thepilot a simulatedwind sock(showing thewind directionandspeed,similar to thewind sockat the airport) to give him extra preparationtime for whenhe breaksout of the cloudsand is readyto land. We consideredutilizing a wind sock torepresentwind velocity in AWE, but decidedagainstit for reasonsdiscussedinSection3.2.3.Azumaetal. [2,3] utilize visualizationtechniquesto presentconflictresolutionscenarioswith otherflights,leaving FAA air traffic controllerswith moreopportunityto dostrategic planningratherthanjust immediatecontrol.

Finally, the designof glyphs(shape,orientation,placement,color etc.) is an im-portant,often a critical, elementin the visualizationof scientificdata[4,42,43].

10

Appropriateuseof color [16,36] andperceptualprinciples[10,41] is importanttoconstructaccurateandcognitively easyto deciphervisualdisplays[30]. In AWE,wehavegivencarefulconsiderationto eachglyphdesignandattemptedto validateour designsby involving theusers(pilots) throughout thedevelopmentprocessasdiscussedandreportedlaterin thiswork.

3 AWE

In this section,we presentAWE (Aviation WeatherDataVisualizationEnviron-ment).Theinput to theAWE prototypeis aDUATs briefingfor a specificarea,forexample,a95nauticalmile (nm)radiusfrom thePaloAlto airport,KPAO.Wemapthis briefingontoa grid specificto thepilot’s routeor his areaof interestandonlyincludeinformationrelevant to his flight. For instance,for route-specificweather,ratherthandisplayingcurrentdataavailablefor his destination,we useforecastsavailablefor his proposedtime of arrival (automaticallycomputedfrom his routeandchosendeparturetime). We make carefulandwell thoughtout decisionsus-ing soundvisualizationprinciples[4,42,43]to designgraphicaldisplaysin ordertopresentthe informationto the pilots in a cognitively easyto decipherformat.Wevalidateour resultsthroughusers’feedback.

AWE allowstheuserto specifyhisflight includingroute,desiredaltitude,trueair-speed,andproposeddeparturetime;selectwhetherhewantsto seecurrentweatheror forecastweather;andselectwhethertheareaof interestis justairportsalongtherouteor all airportsin thearea.Theuseris ableto modify any of therouteparam-etersandseethe effect on weatherhe might encounter. We discusseachof theseissuesin thefollowing sections.

Section3.1 focuseson extractingaviation-specificweatherinformationfrom tex-tual documents.Section3.2 discussesthe graphicaldesignchoiceswe madeandwhy. Section3.3describesthelinking of thevisualinformationto theflight’s pathandschedule,andproviding a simpleuserinterfaceto thepilot to control thedis-play. Finally, Section3.4briefly describestheimplementationof AWE.

3.1 Aviation-SpecificWeatherDataExtraction

As statedbefore,AWE focusesonextractingaviation-specificweatherinformationfrom textual documentsdescribingwindsaloft display, METARs (MeteorologicalConditions),andTAFs (TerminalArea Forecasts).We needto go througha pre-processingphaseto cleanthe textual elementsof thesedocuments.The originaltextual documenthasthreeseparatesectionsdealingwith METARs, TAFs, andwindsaloft information.We startby deletingtheinput formatcommentsandextra

11

notationsassociatedwith METARsandinsertingsemi-colonsinto TAFssowecansimplify our parser. For tacitly understoodinformation,we make it explicit in themodifiedfile. For instance,if no information is provided in the DUATs briefingregardingobstructionto visibility, we insert”NO”. We alsoseparatethe originaltextualdocumentinto threeseparatedocuments.An exampleof eachof thesethreetextual documents(alreadyslightly modifiedfor easeof parsing)is shown in Fig-ure7, Figure8, andFigure9 respectively. Theretrieval of relevantweatherinfor-mationfrom modifiedtextualdocumentsis straightforward.Wenow describeeachof thesethreedocumentsin detail.

Fig. 7. Samplewinds aloft file. The airport identifier is followed by groupsof threeele-ments:wind direction,wind speed,andtemperature.Thesegroupsof threeareavailablefor prespecifiedaltitudesfrom 3000feetto 39000atvariousincrements.

Fig. 8. SampleMETAR file. Eachline representsaMETAR for aspecificairport,specifiedby theairportidentifier(first element).It alsospecifiesthetimeof observation,wind direc-tion andwind speed,visibility, visibility restrictions,cloud layers(with coverageamountandaltitude),temperature,dew point,andbarometricpressure.

Winds aloft: The winds aloft reportbegins with the airport identifier (i.e. KSFO)andthengivesthedirectionthewind is comingfrom (in tensof degrees),thewindspeed,and the temperatureat eachof the prespecifiedaltitudes.It is implicit inthesefilesthatthesethreevaluesarespecifiedataltitudesfrom3000feetto39000atvariousincrements.Moreprecisely, thewindsaloft valuesareassociatedwith 3000,6000,9000,12000,18000,24000,30000,34000,and39000feet. For example,the first row in Figure7 example,the KSFO (SanFrancisco,CA) airport windsaloft report”KSFO 11 11 0 10 10 2 9 8 -4 ...” statesthat thewind at 3000feet iscomingfrom 110degreesat11knots.Thetemperatureis notspecifiedat3000feet;however, for theeaseof uniformparsing,wehave introducedazero.Similarly, thewind at 6000feet(thesecondsetof 3 values)is from 100degreesat 10 knotsandthetemperatureis 2 degreesCelsius.

Since,DUATs only provideswindsaloft for certainlocationsandfor certainalti-tudes,thereis aneedto computethewind speedanddirectionat thepilot-specified

12

Fig. 9. SampleTAF file. A TAF for anairportextendsfrom theairportidentifierto thedot(”.”). EachTAF providesthetime theforecastwascreatedfollowedby theeffective timesof eachforecastelement.Eachelementspecifiesthe wind directionandspeed,visibilityandrestrictionsto visibility, andcloudlayers(with coverageamountandaltitude).

altitudeandlocation.In AWE, we have useddistance-basedinterpolation,that isoneof themethodsusedin meteorologicalcomputations[9,21], to fill in themiss-ing values.We areawareof the datauncertaintyproblemsarisingdueto this ap-proach,but wehaveaddressedtheseconcernselsewhere[24,47,15,14].

METAR: The METAR reportalsobegins with the airport identifier. An exampleis shown in Figure8. We will decipherthe first row of this figure.KSQL standsfor the SanCarlos(CA) airport. ”181646Z” specifiesthe date(18) andthe timeof observation (16:46).The monthandyear is implicit in the file. All timesarespecifiedrelative to theGreenwichMeanTime, alsoknown asUniversalCoordi-natedTime (UTC) anddenotedby Z (or Zulu). ”18014KT” givesthesurfacewinddirection (180 degrees)and the speed(14 knots)at the airport. The visibility is”8SM” (statutemiles).Therearenorestrictionsto visibility. Comparethiswith theMETAR for ”KSAC”, where”FG,” or fog, is restrictingvisibility. Thecloud lay-ersarespecifiednext. SanCarloshasa few cloudsat 1200feet,scatteredcloudsat 5000feet,andbrokencloudsat 8000feet.The temperatureis 11 degreesCel-siusandthedew point is 8 degrees.Finally, thebarometricpressureis 30.22inHg(inchesof Mercury).

TAF: Interpretingthe TAF reportis similar to the METAR report.An exampleisshown in Figure9. As anexample,we will translatepartof thesecondsetof dataassociatedwith theKSJC(SanJose)airport.Thereporttime is on the18thof themonthat 17:30Zulu. The forecastis valid for the 18th from 18:00Zto the 19th(next dayis implicit) at 18:00Z.Thewind is from 0 degreesat 3 knots,visibility 6statutemilesor greater, with scatteredcloudsat 20000feet.Thesecondrow statesthatbetween22:00Zand23:00Zthewindswill become(”BECMG”) 320degreesat17knots,plus6 smvisibility, with afew cloudsat5000feetandscatteredclouds

13

at20000feet.Thefifth row in thissetstartingwith FM0900statesthatfrom 09:00Z(thefollowingday),thewindwill befrom180degreesat20knots,thevisibility willbe3 statutemileswith mist (”BR”), with brokencloudsat 2000feetandovercastcloudsat 8000 feet. Note that the FAA providesa standardinterpretationof thecloudcoveramount:FEWmeans1/8of thesky or lessis obscuredby clouds,SCTmeansbetween1/8 and3/8 of the sky hasclouds,BKN means4/8 to 7/8 of thesky, and OVC meansthe entire (8/8) sky hasclouds.The FAA also provides alist of standardcontractionsfor visibility obscurations.Our examplesarenot forcomprehensive coverage,but ratherfor generalunderstanding.Hence,only a fewoptionsarediscussed.

3.2 GraphicalDesignandDisplay

Wenow turn to thedisplayformats.Webegin by describingthebackgroundof ourgraphicaldisplay. We thendiscussthedisplayof winds,METARsandTAFs.

3.2.1 Background

Both DUATs andtheNWS weathergraphicsusea stateoutlineof thecontinentalUS asthebackgroundimage.This approachprovideslittle contextual informationusefulto pilots. AWE usesa VFR (Visual Flight Rules)aeronauticalchartfor itsbackground.One can obtain thesechartsat different resolutionsof 3-1/2 miles(terminal),7 miles(sectional)and14miles(world) perinch.Currently, AWE usesVFR sectionalaeronauticalcharts.Figures10,11,and12 show theVFR sectionalaeronauticalchartfor SanFranciscoin thebackground.

We consideredotheroptionssuchasIFR (InstrumentFlight Rules)charts,and3Ddepictionof terrainsaswell. We felt that the3D depictionof terrainswill requiremuchmoreinteractionandperceptualunderstandingof 3D displaysbeforethesedisplayscanbeadaptedfor thepilots’ use.AlthoughIFR chartsaremuchlessbusythantheVFR charts(describedbelow), they donot provideadequatedatafor low-altitudeflights that canbe crucial in avoiding weather-relatedaccidents.We alsoreportonusers’feedbackregardingourchoiceof VFR asthebackgroundimage.

TheVFR sectionalchartshows thelocationof airports(magentaor bluecirclesorshortlinesthatmimic therunway layout),airways(”highways” in thesky aslightblue straightlines), navigation aids (suchas VHF Omni Rangesalso known asVORsandmostlydepictedby acompassrose),controlledandspecialuseairspace,obstructions,naturalterrainfeatures(suchaswaterandhills, depictedusingcolorcodedaltitudes),demographicfeatures(suchas cities, depictedin yellow), andmaximumelevation in eacharea(depictedwith numberswith superscripts).Fi-nally, thehorizontalwhite linesappearingin Figures10,11,and12 indicatethatadetailedterminalareachartis alsoavailablefor theencompassingregion.Although

14

Fig. 10.AreaTAF displayusingtriangularwarningicons.Thetop, lower left, lower right,andthemiddlesubtrianglesrepresentwinds,visibility, cloudsandtemperature/dew pointspreadconditionsrespectively. Red,yellow, andwhite colors indicatealert, cautionandnormalconditionsrespectively. Grey color indicatesthattheinformationis notavailable.

thebackgroundmaylookclutteredandcomplex, thechartbackgroundtexturegivespilots a familiar environmentwith which to interactandprovidesthemwith addi-tional informationfor makingtheir ”go/no-go”decision.For example,a 2000footceilingpresentsadifferentsituationif theairportis in aflat regionversusanarrowvalley surroundedby tall mountains.Overlayingtheweatheron thechartconsoli-datestheweatherandtheterrainsurroundingtheairportallowing thepilot to makeadecisionby lookingonly atonesource.

In particular, we expectthatthis displaywill bevery usefulin routeselection.The

15

Fig. 11. RouteMETARs andwinds aloft shown alongsidea pilot-selectedroute.A com-binationof textual METAR at KMOD (Modesto,California) andsymbolicMETARs atKPAO (Palo Alto, California),KSAC (Sacramento,California),andKSNS(Salinas,Cal-ifornia) is shown. The red borderat KSAC indicatespoor visibility due to fog, and theyellow borderindicatesmarginal visibility at KMOD.

route selectionprocesstypically involvesdetermininga minimum distancepaththatavoidstheabovehazards,aswell asany hazardspresentedby adverseweatherconditions.Seeingeachof theseelementson onedisplaysimplifies the process.UsingAWE, apilot is ableto plot apaththathasaccessiblenavigationaids;avoidsspecialuse,prohibited,or restrictedairspaces;approachesmountainousterrainatthe properangleandaltitudebasedon the winds aloft; avoids flight over inhos-pitableterrain;optimizeshisemergency landingchoices;andavoidsflight throughadverseweather.

16

Fig. 12. AreaMETARs. All theMETAR reportsavailablefor thechartedarea.METARscanbedisplayedin symbolicor textual formatasshown here.

3.2.2 WindsAloft Display

Themostcommonmethodof displayingwindsfor aviationpurposes,alsoadoptedby FSS,DUATsandNWS,is theuseof wind arrowswith barbsasshown in Figure5. Eachbarbrepresentseither5 knotsor 10 knots,dependingon its length.Al-thoughthis representationis familiar to pilots, onehasto countthe barbsandtryto determinewhetherit’ s a shortbarbor a long one.Thedirectionof the wind isencodedin thetilt of thearrows.

In AWE, we consideredtwo choicesfor displayingwinds aloft data.The twochoiceswerebarbsandwind arrows.After someinitial discussionwith pilots,we

17

decidedto usewind arrows over thebarbs.We rotatethewind arrow to show thedirectionof thewind at that locationanddisplaythe wind speedalongside.Mostpilotsfind thisdisplaysimplerthanthebarbsdisplay. In thismethod,it is easiertoquickly reada number(thewind speed).Furthermore,this methodseemsto bringoutwind directionmoreclearlythanthebarbs.Laterin Section4,wereportonthischoiceby providing users’feedback.

In AWE,wehaveusedthewindsaloft displayin boththeroutespecificweatherandtheareawideweather. If thepilot requestsrouteweather, wedisplaywind informa-tion alongsideeachairportalonghis route,asillustratedin Figure11 by theblackarrows. This visual representationallows a pilot to comparehis flight pathto thepathof thewind anddeterminewhetherto expecta tailwind, headwind,or cross-wind. He/shecanthencompensatefor a crosswindeasily, or expectto go slowerandusemorefuel for a strongheadwind,or slow down to save fuel but still arriveat the scheduledtime in caseof a strongtailwind. If the pilot hasrequestedareawideweather, AWE displaysall known windsaloft forecastsfor centralCalifornia.

Becausewindsaloft arealtitudedependent,in AWE, thepilot canmodify his se-lectedaltitudeto determinewherethewindsaremostfavorable.Thishelpsthepilotto chooseanappropriateflight pathfor cruising.

3.2.3 METAR/TAF Display

Determininghow to displayaMETAR or aTAF to make it easilydecipherableyetprovideall thenecessaryinformationis challenging.TheFAA approach,usedwithchartsavailableat Flight ServiceStations,arevery informative, yet very cryptic.Theirsymbols,shown in Figure13,requirea lot of memorizationandcanbeeasilyforgottenif not usedregularly. Instead,we could chooseto follow the WeatherChannelapproachanddisplayonly a small setof symbols.Onedisadvantageofthisapproachis thatmuchof theavailableinformationis not represented.

In AWE, we presentoptionsto the pilot to view all the informationavailableorto view only part of it. In particular, the pilot has the option to ask for all theinformation,askfor partof theinformationin avisualrepresentationthatprovideshim/herwith a ”feel” for the weather, that helpsin making”go/no-go” decisionsprior to flight, or askfor avisualpresentationwhichprovideshim with anoverviewof theweather, that is helpful in makingroutingandpathplanningdecisions.Wenow discusstheseoptionsin greaterdetail.

Textual METAR/TAF Display with Color-coded Borders: Whenthepilot asksfor all theinformationrelatedto theMETAR/TAFs,wechooseto displaytheinfor-mationmostly in a textual mannerby presentingtheessentialsin a compactway.A textualdisplayof METAR informationis shown in Figure14.Contrastthiswiththetextualdisplayof TAFsusingNWSusedcurrently(Figure1). To makethetaskof recognizingcrosswindconditionsquicker, we still representthe surfacewind

18

Fig. 13.FAA weathersymbolsandtheir associatedmeanings.Thesesymbolsareusedonchartsavailableat Flight ServiceStations,andmorerecentlyon DUATs andNWS chartsavailableon theweb.

direction(not to be confusedwith the winds aloft, which is at varying altitudes)graphicallyaswell astextually. Thedetailsof the textual displayaredescribedinthecaptionof Figure14.

Fig.14.Close-upview of atextualrepresentationof aMETAR.Thewinddirectionisshownby thearrow. The wind directionandwind speedarealsoshown by the text in theupperright corner. In this case,the wind is comingfrom 60 degreesat 3 knots.The visibilityis 5 miles with ’BR’ or mist. Cloud layersare given as a few cloudsat 1500 feet andscatteredcloudsat5000feet.Finally thetemperatureanddew pointareboth4 degreesandthebarometricpressureis 30.21inHg. Thecoloredrectangle(yellow) representsmarginalvisual flight conditions(MVFR). It would be coloredred for poor visibility conditions(IFR) or grayfor goodvisibility (VFR).

Thetextual displaysaresupplementedwith color-codedbordersto warn thepilotof possibleadversevisibility conditions.We describedthe visibility conditions,namelyIFR, MVFR and VFR in Section2.1. Many pilots are preventedeitherlegally (by not having appropriatecertification)or practically(by not beingpro-

19

ficient) from landingat airportswith IFR conditions.MVFR conditionsareonly apractical,not a legal,deterrent(many pilots feel lesssafein marginal conditions).AWE displaysairportswith IFR conditionswith a redborder, airportswith MVFRconditionswith ayellow border, andthosewith VFR conditionswithoutaborder.

Visual METAR/TAF Display for Overview: Thereare four primary elementsthataffecta pilot’s ”go/no-go”decision:wind conditions,visibility, cloudaltitude,andtemperature/ dew point spread.Thefirst threeelementsareavailableboth inMETARs and TAFs. The fourth element(temperature/dew point spread)is onlyavailablein METARs.We have alreadymentionedtheimportanceof wind condi-tions,visibility, andcloudconditionsearlier. Wenow brieflycommentonthefourthelement– thespreadbetweentemperatureanddew point, that is temperaturemi-nusdew point. The temperature/dew point spreadprovidesinformationregardingfog, andis clearly an importantpieceof information,particularly in areaswheremorningandeveningfog is commonfor examplein theSanFranciscoBay Area.If theairportis currentlyexperiencingfog, thespreadgivesinformationonwhenitmaybecomeclear(especiallyif youhavetheprevioushour’sMETAR andcanseeatrendin how thespreadis changing).Thereareconditionsunderwhichtemperaturemaybean importantpieceof informationaswell. For example,in theheatof thesummer, temperatureis important,particularlyathighaltitudeairports,sinceit hasadirecteffectonaircraftflight characteristics.Similarly, low temperatureshaveaneffectonstartingtheengineandthey cancontributeto possibleiceontheairframe.Usually, though,temperaturescanbe safelyclassifiedasnon-criticalinformationandarenot representedin theiconic formatin AWE.

Our iconsaredesignedto presenta quick overview of all four primaryelements.We first consideredsomeof the icons in usesuchascoloredregions in DUATs(Figure3), color-codeddisks,andweatherchannelrepresentation.All theserep-resentationsseemto do a goodjob of providing onepieceof informationsuchasvisibility conditions;however, they fall shortof providing lower level detailssuchascloudinformationor windsetc.Color-codeddiskswith auxiliary informationasutilizedin ADDSattempttoovercomethisdifficulty; however, thismethoddependsuponakey andalsothedensebarbsoftenoverlapwith otherinformation.

Therefore,in orderto presentthefour quantitiesin anintegratedmanner, we con-sideredsomeadditionalgraphicaloptionsfor displayingthem– rectangularrepre-sentationstackedon top of eachother, rectangularrepresentationstackedhorizon-tally, circular representationwith onequarterassignedto eachelement,andtrian-gular representationsubdividedinto four subtriangleswith eachtriangleassignedto eachelement.We chosethe rectangularrepresentationstacked on top of eachotherfor thedetailedMETAR/TAF visualdisplay(to bediscussednext), becausethis representationis ideally suitedfor cloudlayersat differentheights.Therefore,wewantedto chooseanalternativerepresentationfor distinction.Of theremainingthree,an earlierdiscussionanduserfeedbackseemedto favor the circular or thetriangularrepresentationswhich appearedto bemorewholistic andeasierto pro-

20

cess.We finally decidedon the triangularicon becausetriangulariconsareusedfor warningsandaremorefamiliar to thepilots.Also, thetemperature/dew pointspreadinformationis not availablefor TAFs andis not ascrucialastherestof theinformation.Therefore,weassignedthecentersubtriangleto this information.

An area-wideTAF displayusingtriangulariconsis shown in Figure10. The top,lower left, lower right andthe middle trianglesrepresentwind, visibility, clouds,andtemperature/dew pointspreadrespectively. Figure15 describesthecolor cod-ing usedin thetriangularicon.In theTAF displays,sincethey donotprovide tem-peratureor dew point forecasts,thecentertriangleis alwaysshown in gray. In eachof thesecases,thecolor codingservesto alert thepilot of possibleadversecondi-tions.Thethresholdswerechosento coincidewith theFAA definitionsof IFR andMVFR conditionsfor visibility andceiling. The thresholdsfor wind speedsandtemperaturedew point spreadswerechosenbasedon a typical pilot andweatherprofile. A moreflexible userinterfacewould allow eachpilot to sethis/herownthresholdsfor cautionandalertconditions.

Fig. 15.Triangulariconcolorcodingscheme.

Visual METAR/TAF Display with Details: AWE also usesvisual displaysforpresentingdetailedMETAR/TAF information.Severalexamplesareshown in Fig-ure 11. Figure 16 shows a close-upview of the symbolic representations.ThisMETAR/TAF displayencodessurfacewind speedanddirection,cloudconditionsatdifferentaltitudes,andvisibility conditions.

Winds: We consideredseveralalternativesfor displayingthis information.An im-portantdecisionwasto choosebetweenarrow glyphsandwind socksto representwind speedanddirection.Everyairporthasawind sockandpilotsareableto inter-pretthemreadily. Thedirectionis displayedby theorientationof thewind sockandthespeedis displayedby theamountthewind sockis straightenedto the90degreeorientation(with respectto thepoleholdingit up).Weattemptedto useawind sockto displaywind informationassuggestedby PruynandGreenberg[26] but foundthe3D wind sockdid notmergewell with theoverheadperspectiveof our2D view.Also,oneof theadvantagesof displayingthewind directionisdiminishedwhenus-ing thewind sock:whenusingawind arrow, thepilot caneasilycomparethewind

21

Fig. 16.Close-upview of symbolicrepresentationof METARs.Thewind is comingfromthe SouthEastat about20 knots in the left symbol,from the NorthEastat about5 knotsin the middle symbolandis calm in the right symbol.The red borderin the left symbolshows indicatesstrongwinds. The left symbolshows a broken layer of cloudsat about3000 feet, the middle symbol shows a scatteredlayer at about4000 feet, and the rightsymbol indicatesclearconditionsbelow 12000feet. Finally, the text below the symbolscorrespondsto 6 miles visibility with showers,5 miles visibility with mist, and5 milesvisibility with haze.

directionwith the orientationof the runwaysasdisplayedon the sectionalchart(AWE’s backgroundimage).Thus,crosswindlandingconditionsbecomevisuallyobvious.The3D wind sockrequiresmoreanalysisto extractsimilar information.Therefore,we decidedin favor of thearrow glyphs.Thesearrow glyphsaresimi-lar to glyphsusedin windsaloft display, but with the following modifications.Inwinds aloft display, we usethe samesizearrow andthe wind speedis presentedtextually adjacentto the arrow. We felt that this approachis justified becausethewind directionis typically muchmoreimportantthanthe wind speed.Thewindsaloft displaythusfacilitatesthequick detectionof altitudesat which thedirectionof the winds aremost favorableto the flight path.The speedof the wind is alsoimportantat theselectedaltitudeandis usedto computethetime of arrival at var-ious checkpoints.However, in recentpractice,this task is often delegatedto theon-boardGPS(globalpositioningsystem)unit, furtherreducingtheimportanceofdisplayingthewind speed.

In contrast,in METAR/TAF display, the width of the surfacewind arrow varieswith thespeedof thewind. Light winds(that is, low wind speeds)arerepresentedby thin arrows, whereasstrongwinds are representedby thick arrows. Comparethe wind symbol(20 knots)shown in the left symbolof Figure16 and(5 knots)shown in themiddlesymbolof Figure16.Also noticethewindsshown in therightsymbolof Figure16. Thewind vectorin theright symbol0 degreesat 0 knots,orcalm.Calmwinds arerepresentedwith just a dot sinceno directionis associatedwith them.

Finally, theborderof thewind squareis color coded.A redbordersignifiesstrongwinds ( ��� 20 knots),an orange/yellow bordersignifiesmediumwinds ( ��� 15knots),andblackbordersignifieswindsbelow 15knots.

Clouds:Thenext elementof theMETAR/TAF symbolis a rectanglethatpresentsthecloudlayers.Therectanglerepresentsthesky from 0 to 12,000feet.We chose12,000feetbecauseautomatedweatherobservationsystems(AWOS)usethesame

22

heightthresholdfor reportingclouds.We thenpseudo-colortherectangleto showthecloudlayers[16,36].As suggestedby Bertin[4] andTufte[42],we chosea grayscaleto representthecloudamountssopilotsdonotneedto rememberacolorkey.Darkercolorsrepresentthickercoverage.Hence,whiterepresentsaclearsky. Verylight grayrepresentsa few clouds,a darker light grayrepresentsscatteredclouds,mediumgray representsbrokenclouds,andfinally, very dark (nearlyblack)grayrepresentsanovercastsky. Ceilings(definedasbrokenor overcastlayers)arethusquickly recognizedby scanningfor darker grays.

Visibility: The borderof the rectangleis color codedto instantly show whetherthe visibility conditionsare poor (IFR with red border),marginal (MVFR withorange/yellow border),or normal(VFR with black/noborder).

Thefinal elementof aMETAR/TAF symbolis thetext specifyingthevisibility andobstructionsto visibility. Thatinformationis presentedin blackandblendsin morewith the backgroundthan the restof the symbol.In this way, the informationisthereif it is needed,but it is not overwhelming.

The METAR/TAF symbolsdo not representthe temperature,dew point, or baro-metric pressurevalues.Also, the valuesfor cloud altitudesareshown only indi-rectly (by theamountof rectanglefilled). If any of thosevaluesaredesiredexactly,thesymbolcanbetransformedinto a text box containingall availableinformationasdiscussedpreviously.

3.3 UserInterfaceandPathPlanning

The Aviation WeatherDataVisualizationEnvironment,AWE, providesan envi-ronmentfor the pilot to interactwith andobtainthe informationhe/sheneedstoeffectively plan a flight. Theuserinterfacefor the systemis shown in Figure17.The pilot is ableto setthe true airspeedof his airplane,selectan altitudefor theflight, andspecifya departuretime. He canalsoselectwhetherhe wantsto dis-play weatherfor the entire areaor just along the flight path.Moreover, he canchooseto view any of the graphicaldisplaysdiscussedearliersuchasMETARswith overview, detailsor textual, TAFs, or winds aloft. Finally, he canspecifytodisplay the graphicsonly for selectedairportsor usethe closestairport in casetheinformationis unavailablefrom a selectedairport.Theseselectionsareusedtodeterminewhatspecificinformationto display, asdiscussedbelow.

3.3.1 SelectingInformationOverlays

AWE providesthepilot with a numberof optionsin selectingwhatinformationtodisplay. Theseinclude:

23

Fig. 17.Userinterfaceof AWE

� Displayfocus:displayareawideweathervs. routespecificweather;� Type of weather:display currentweatherobservations(METAR) or forecastweather(TAF), and/orwindsaloft;� Displayformats:displayanoverview visualrepresentation,adetailedvisualrep-resentation,or a textual representationof eitherMETARsor TAFs.� Reportlocality: displayonly weatherfor airportswith reportingcapabilityoralsodisplayclosestavailableweatherfor airportswithout reportingcapability.

Wenow describeeachof theseoptionsbelow.

DisplayFocus:Thepilot canchooseto displayweatheratairportseitherjustalonghis routeof flight, asshown in Figure11 or at airportsin the entireareaknownby AWE, asshown in Figure12.Theareawideweatheroptionis especiallyusefulduringtherouteselectionphase.Thepilot canview all availableweatherandthenchoosearouteof flight. Conversely, if hehasalreadychosenaroute,routespecificweathershowshim only informationrelevantto hisflight.

Typeof Weather:Thepilot canchooseto view eithercurrent(METAR) or forecast(TAF) weather, aswell aswindsaloft information.Most(probablyall) airportsthatprovide TAF forecastsalsoprovide METAR observations.Therefore,we imple-mentedtheoptionsto displayMETARs or to displayTAFs to bemutuallyexclu-sive to avoid screenclutter; that is, eitherMETARs or TAFs canbedisplayed,butnot both simultaneously. Winds aloft, on the otherhand,provide complementaryinformationto both sourcesandcanbe displayedeitheraloneor with a TAF orMETAR. METARs andTAFs provide surfacewindsassociatedwith thereportingairport.Winds aloft reportsprovide winds at variousaltitudesandareassociatedwith amuchwiderareaaroundtheairport.

24

DisplayFormats:AWE providesthepilot with threeoptionsonhow to displaytheMETAR andTAF reports.We discussedtheseoptionsin detailin Section3.2.

Reportlocality: Not all airportsAWE knowsabouthaveweatherreportingcapabil-ities. If thepilot is viewing routespecificweather, we providehim with theoptionof choosingto displaythenearestreport(METAR or TAF) for airportsthatfit thiscriteria.To accomplishthis, thepilot choosesthe”show closestavailableweather”option.

3.3.2 RouteSelection

In contrastto therouteselectionprocessof a commericalairliner, which is mainlycontrolledby theflight controllerson theground,a generalaviation pilot hascon-siderableflexibility in choosingtheflight path.As statedbefore,a typical generalaviationflight lastsabout4 hoursandcoversabout400miles.Dependinguponthepurposeof the flight – pleasureor business– the pilot alsohasthe flexibility ofchoosingthe landingairport.Weatherplaysa major role in determininga safeaswell asa fuel-efficientor a time-efficient route.

In AWE, the pilot canexplorewhat-if scenariosby choosingalternateroutesandobservingtheweatherconditionsalongthedifferentroutes.Thepilot specifieshisrouteof flight by selecting(with themouse)asequenceof airports.Theuseris ableto extendhis routeby addingan airport to the end,modify his routeby deletingairportsoff theenduntil themodificationpoint is reached(eventuallybacktrackingto thebeginningif desired),or specifyanew route.As airportsareaddedor deleted,theinterfaceisupdatedto reflectthecurrentspecifiedroute.Thebackgroundscreenwith a routeselectedby thepilot is shown asasolid line in Figure11.

Flight ScheduleandTime-dependentInformation:Pilotsneedtheforecastweatherat thetimeof arrival ateachen-routecheckpoint(airport,in AWE),nottheforecastfor thedeparturetime.To eliminatetheneedfor thepilot to specifyafull flight planwith expectedarrival times,we calculatearrival timesautomaticallybasedon thespecifiedtrueairspeedanddeparturetime. If thepilot is looking at route-specificTAFs,hewill begiventheappropriateforecastbasedon this information.

Choosingan AppropriateForecast:Determiningtheappropriateforecastis not assimplefor a computerasit is for a person.DUATs forecastsarenot specifiedatmutuallyexclusive time ranges.Rather, asillustratedin Figure9, thegeneralfore-cast(givenfirst) coversa24hourperiod.Forecastsfor morespecifictimesarethengiven. Even thesespecifictime periodscanoverlap.As in the KSCK (Stocktonairport)TAF of Figure9, weseethatthreeforecastsapplyfor 10:30.First,thegen-eralonethatspanstherange18:00on the18thof themonthto 18:00on the19thstatesthattheweatherwill be

�wind 300@6,visibility 6 statutemile (sm)or better,

sky clear� . Then,the”FM1000”, statesthatfrom 10:00,theweatherwill be�wind

calm,visibility of 1smwith mist,sky clear� . Finally, the”TEMPO1015”statesthat

25

therewill bea temporaryconditionfrom 10:00to 15:00of weather�visibility 1/4

smwith fog, andaverticalvisibility of 200feet� . AWE mustdeterminewhichoneof theseto presentto theuser. Currently, AWE usesa setof ruleswhich considerthevisibility, cloudcovers,andwind speedsto extractthemostcautiousscenario.Thus,in this case,it choosesthe temporaryfoggy conditionasthe representativeweather. Moreresearchis neededto extractandrepresenttheinformationfor over-lappingtimeperiods.

3.4 Implementation

AWE is writtenusingC++,OpenGL,andXformsandrunsonanSGIworkstation.It wasalsoeasilyportedto aSony SuperslimProNotebookcomputerrunningMan-drakeLinux. Thisplatformwill beusefulto provideAWE duringin-flight usewiththehelpof data-linktechnology. Theunderlyingcapability, or AWE’s foundation,consistsof object-orientedprogrammingclassesthat dealwith informationaboutairports,METAR, TAF, and winds aloft. It also includessupportingclassesthatknow how to dealwith latitudeandlongitudecoordinatesandcanfind distancesbetweenlat/longlocations.Four datafiles areusedby AWE. Airport identifiersaswell asthe latitudeandlongitudecoordinatesof theairportarespecifiedin a userreadabledatafile, andthe DUATs briefing is translatedinto threeseparatefiles:onecontainingcurrentMETARs, onecontainingTAF forecasts,andonecontain-ing windsaloft forecasts.

The two final foundationclassesareAwe interfaceandAwe. Awe interfacedealswith interactionswith theuser, properlyupdatingthe input forms.TheAwe classmaintainsa list of known airports,andreadsandupdateswindsaloft, METAR, andTAF reports.It alsoprovidesvarioussearchmethods,suchasfinding the closestairport to anarbitrarylatitude/longitudeposition;gettingtheMETAR/TAF for anairport with reportingcapability; gettinga representative METAR/TAF for non-reportingairports;finding theclosestandsecondclosestwindsaloft forecasts;orkeepingtrack of the user’s chosenroute.Eachof thesemethodsis usedby otherfoundationclassesto helpprovidewhattheuserwants.

4 Users’ Feedback and Experiences

Thefirst authorof this paperis a generalaviation pilot herselfwith a commericallicensewith instrumentratingandover 800hoursof flight experience.During thedesignof AWE, feedbackwas taken on many issuesdiscussedbelow at severalstagesfrom differentpilots to ensurethat the systemremainspilot-friendly andusable.

26

AWE, asdiscussedin this work, wasalsoevaluatedby six generalaviation pilots(fiveof themwork atNASA AmesResearchCenter, California).Fourof thepilotshave a commerciallicense,two have a private license,andonealsohasa flightinstructorlicense.Fourof thepilotsareinstrumentratedandarehencequalifiedtofly underpoorvisibility conditions(IFR). Theflight instructoris qualifiedto teachunderpoor visibility conditionsaswell. The approximatetotal flight time of thepilots rangedfrom 120 hoursto 2000hourswith an averageof 675 hours.Eachpilot reportedthatnearlyhalf hishourswasoutsidehis localarea.

Eachevaluationsessionconsistedof aformaldemonstrationof AWE followedby apracticesessionandaquestion/answerperiod.Eachof thepilotswasthenrequestedto fill out a questionnaire.Finally, morefeedbackwasobtainedthroughanopen-endeddiscussionof currentanddesiredfeatures.

The first sectionof the questionnaireaimedto gatherinformationabouteachpi-lot’sflying backgroundaswell ashis familiarity with DUATsandADDS. Wehavealreadydescribedthe flying backgroundabove. The familiarity of the pilots withtheDUATs systemvariedfrom 1 (very familiar) to 4 (not familiar) on a scaleof 1to 5 (not familiar at all) with anaveragescoreof 2.33.Most pilots statedthat theywerenot familiarwith theADDS systematall with anaverageof 4.67.In anopen-endedquestionasto which weathervisualizationsystemsthey aremostfamiliarwith, fiveof thepilotsmentionedthatthey useDUATsfor theirpre-flightbriefings.A few alsousedothermethods,suchastalking with anFSSspecialistandgettingunofficial informationvia televisionor theWorld WideWeb.

Therestof thequestionssoughttheiropinionsonvariousaspectsof AWE’sdesignaswell ashow it comparedto otherweathervisualizations.We comparedAWEto Scanlon’s work, ADDS, graphicsavailable throughDUATs, and the WeatherChannelgraphics.Wehavedescribedall of thesesystemsin Section2. Ratherthanaskingabouteachsystemby name,we presentedthe pilots with a representativedisplayof thesesystemsandanexplanationwhenneeded.As amemoryaidandtoallow for side-by-sidecomparisons,wepresentedarepresentativedisplayfor AWEdisplaysaswell.

We begin by presentingthe resultsof the survey in the order that the questionswereasked in the questionnaire.Overall, the evaluationof the individual piecesof information in AWE wereratedasbetterthanotheroptionsin all cases.Thefirst setof questionsweredesignedto evaluatethebackground,windsaloft display,METAR display, andTAF display. Amongthefour choices– VFR sectionalchart(usedby AWE), IFR chart,genericUS/state/point-of-interestmap(suchasusedbythe GPSunits) and3D depictionof terrain/airports/points-of-interests– the VFRsectionalchartasusedby AWE wasratedas”very desirable”(1.33)on a scaleof1 to 5 (not at all desirable).The secondchoicewasan IFR chartwith a meanof2.0. This is not surprisingbecausethe majority of the pilots areinstrumentratedandoftenfly IFR. Theremainingtwo choices– agenericUS/state/point-of-interest

27

mapanda 3D depictionof terrain/airports/points-of-interest– earned3.17and3.0andappearto beneutral.In anopen-endedquestion,somepilots expresseddesireto selectively decluttersomeof thebackgroundmaterial.

AWE’swindsaloft display(blackarrow) rated1.67onascaleof 1 (verydesirable)to 5 (not at all desirable).All otheroptions– arrow with color-codedwind speed,FAA wind speedwith barbs,and3D Windsock– ratedworsewith aratingof 2.33,3.0and4.17respectively. Someof thesuggestedchangesin displayingwindsaloftdisplaywasto includethinnerarrowsandlength-encodedwind speeds.

For theMETAR display, wepresentedthepilotswith ninechoicesincludingAWE’sdesignandthedesignsusedby othersystemsasmentionedabove. Thesechoicesincludedcolor-codeddisks,color-codeddisks with textual information,weatherchanneltypegraphics,color regionspresentingonly IFR/MVFR/VFRetc.AWE’srectangularcloud/windsymbolswith color-codedborderswerepreferred(ratingof1.67)over theotherrepresentations,followedby AWE’s textualdisplay(1.83)andAWE’striangularicons(2.0).Thenext bestlikedrepresentationswereasimpletex-tualoverlayasyoumouseover theairport(2.67),theDUATsrepresentationshow-ing regionsof IFR/MVFR/VFR(2.83),andtherepresentationusedby ADDS (3.0).A desirablefeaturenot currentlyincludedin AWE is thepresentationof trendsinweatherwith theability to extractthemorerecentMETARs.

Not asmany systemscurrentlygraphicallyencodeforecasts,hencethepilotswerepresentedwith fewer choicesfor theTAF display. Of thefive choices,AWE’s de-tailed displaywaspreferred(rating of 1.67)alongwith AWE’s overview display(alsoa ratingof 1.67but a worseminimumrating).Thenext bestoptionwascon-sideredthe ADDS presentation;that is, a textual overlay asyou mouseover theairport(2.5).

We alsoaskedabouttheuserinterface.The”Use closest”functionalitywasfoundto beuseful(ratingof 1.5)but almosteveryonecommentedthat thereshouldbeavisual referencethat lets the pilot know which airport’s reportis beingused.Theautomaticselectionof the TAF sub-forecastbasedon arrival time wasalsocon-sidereduseful(ratingof 1.5).Becausearrival timesarenot precise,theautomaticselectionof a TAF sub-elementwould beconsideredevenmoreusefulif thepilothadtheoptionto look at theotherforecastsfor thattime frameor aroundthattimeframe.

The secondsetof questionsevaluatedthe overall utility of differentcapabilitiesof AWE in comparisonto the other systemsas judgedby the pilots. AWE waswell ratedin all categoriesand was found to be betteror much betterthan theotheroptionspresented.For weatherbriefings,AWE earnedan averagerating of1.5on a scaleof 1 (”much better”) to 5 (”much worse”) for overall comparisontotheir currentsystem.The pilots alsofoundAWE useful(rating of 1.67) for routeselection.For possiblein-flight briefings,it was rateda 1.33 on the samescale.

28

In an open-endedquestionto determinefor what purposeswould AWE be mostuseful,thepilots listedinitial planningespeciallyin unfamiliarareas,helpingwiththe go/no-godecision,andenroutedecisionmakingsuchascontinuingto fly theoriginal plan, replanning,or choosingan alternatedestination.All pilots thoughtit would be usefulboth in their local areaaswell asin an unfamiliar areawith acoupleof themstressingits greaterusefulnessin unfamiliar territory.

Thediscussionfollowing the questionnairefocusedon high level advantagesandshortcomingsof AWE. Overall, thepilots werepleasedwith AWE, specificallyitsdisplaysallowing for the quick interpretationof the data,and its helpfulnessinchoosingroutesof flight aswell asalternates.Severalsuggestionswerementionedto addto theexistingfunctionality. Themostimportantonesincludedthedisplayof(or at leastaccessto) weathertrendsencodedby METARsandTAFs,anoptiontodisplaywindsaloft for all (DUATsprespecified)altitudessimultaneously, selectivede-clutterof the backgroundanddisplayedsymbols,anda methodof interactingwith AWE in thecockpitthatdid notrely onakeyboardandmouse.A touchscreenandvoiceinteractionwerebothmentionedaspossiblealternatives.

5 Conclusions and Future Work

We have presentedan Aviation WeatherDataVisualizationEnvironment(AWE)for GeneralAviation (GA) pilots. Thesystemwasdesignedkeepingthe needsofthepilots in mind. The focusis to displaywindsaloft, METARs andTAFs infor-mationagainsta usefulbackgroundin an integratedmannerto assistthepilots inmakingusefuldecisions.Thesystemcanbeusedfor pre-flightweatherbriefings,routeselection,andto make a ”go/no-go” decisionprior to theflight. Thesystemwasevaluatedby pilotsandfoundto bevery usefulin comparisonto severalothersystemsincludingDUATswhichis themostcommonmethodfor pre-flightweatherbriefings.

An importantstepforward would be to make this weathervisualizationenviron-mentavailableduringflights.TheFederalAviationAdministration(FAA), NationalAeronauticsandSpaceAdministration(NASA), andaviationindustryarecurrentlydevelopingdatalink technologywhichpromisestobringmuchneededdatadirectlyto thein-flight pilot [13]. Thedriving goalfor datalink is theFreeFlight programthatwill allow commercialandgeneralaviation pilots to determinetheir routeofflight with minimal coordinationwith FAA air traffic controllers.Besidestrafficdata[3],datalink canbeusedto transferweatherdatato thepilot or sendweatherdatafrom on-boardsensorsto acentrallocationfor dissemination(aspilot reports)to otherflights in thegeneralarea.As mentionedearlier, AWE hasbeenportedtoa Sony SuperslimProNotebookcomputerrunningMandrake Linux. We areableto accesstheaircraft’s currentpositionby connectingthe laptopto a GPS(globalpositionsystem)unit. AWE canusethis real-timepositiondatato automatically

29

scroll theareadisplayedon thechart.We expectthattheweatherupdatesin AWEwill not requiremuchhead-down time (Head-down time is time spentfocusingoninstrumentsor datainsidethecockpitratherthanscanningoutsidefor traffic). Wehopeto provide AWE for in-flight weatherbriefingsandroutemodificationplansfor GA pilots. An importantcomponentfor interactionwould have to be a touchscreenor voiceinteractioncapability.

The userevaluationstudy pointedout someareasof improvement.Noteworthyamongstthemarethe availability of weathertrendsasencodedby METARs andTAFs, to display all winds aloft simultaneously, and to selectively declutterthebackgroundor displayedsymbols.In addition,weareinvestigatingvisualizationoftwo additionalelementsof DUATsbriefingsthatarecurrentlybeingdisplayedonlytextually – pilot reportsandnoticesto airmen.PIREPsareusefulin confirmingtheforecastor pointingoutareaswhereit wasnotaccurate.They areespeciallyhelpfulfor determiningtheextentof icing conditionsandthetop of thecloudlayers.No-ticesto airmen(NOTAMs) provideavarietyof informationpotentiallyhelpful to apilot. Someexamplesareunlightedobstructionsneartheairport,acrobaticpracticeareas,active parachutejumping sites,temporarilyprohibitedareas,andareasofpossiblyreducedGPSreception.Webelieve thatthecombinationof thesefeatureswill makeAWE a veryusefultool for localandnon-localpilotsalike.

6 Acknowledgement

Theauthorsgratefullyacknowledgethesuggestionsfrom fellow pilotsDavid Iver-son, Cedric Walker, and Butler Hine. This researchwas partially supportedbyLLNL AgreementNo. B347879underDOE ContractNo. W-7405-ENG-48andtheMultidisciplinaryResearchInitiative(MURI) grantby DOD.

References

[1] FederalAviationAdministration.WordWideWeb. www.faa.gov/aua/awr.

[2] R. Azuma,H. Neely III, M. daily, andR. Geiss. Visualizationtools for free flightair-traffic management.IEEE ComputerGraphicsand Applications, 20(5):32–36,September-October2000.

[3] R. Azuma, H. Neely, M. Daily, and M. Correa. Visualizationof conflicts andresolutionsin a ’free flight’ scenario. Proc. of IEEE Visualization, pages433–436,October1999.

[4] JacquesBertin. Semiology of Graphics:Diagrams,Networks,Maps. University ofWisconsinPress,Madison,WI, 1983.

30

[5] AOPA 1998AviationFactCard.1998.

[6] TheWeatherChannel.World WideWeb. www.weather.com.

[7] R. Denzer. Graphicsfor environmentaldecisionmaking. IEEE ComputerGraphicsandApplications, pages58–64,March1993.

[8] DynCorpDUATS. World WideWeb. www1.duats.com.

[9] DennisGlaeser, SanfordGum,andBruceWalters.AnInvitation to Fly: Basicsfor thePrivatePilot. Wadsworth PublishingCo.,1985.

[10] G. GrinsteinandH. Levkovitz. PerceptualIssuesin Visualization. Springer, Berlin,1995.

[11] H. Haase,M. Bock, E. Hergenrother, C. Knopfle, H.-J. Koppert, F. Schroder,A. Trembilski, and J. Weidenhausen.Meteorologymeetscomputergraphics– alook at a wide rangeof weathervisualizationsfor diverseaudiences.Computers andGraphics, 24:391–397,2000.

[12] W. HibbardandD. Santek. Visualizinglarge datasetsin the earthsciences.IEEEComputer, 1989.

[13] ThomasA. Horne.Beamingup theweather. AOPA Pilot, pages97–102,March2000.

[14] S.K. Lodha,A. T. Pang,R. E. Sheehan,andC. M. Wittenbrink.UFLOW: Visualizinguncertaintyin fluid flow. In Proceedingsof IEEE Visualization’96, pages249–254,October1996.

[15] S. K. Lodha, C. M. Wilson, and R. E. Sheehan. LISTEN: soundinguncertaintyvisualization. In Proceedingsof IEEE Visualization’96, SanFrancisco,California,pages189–196.IEEE,October1996.

[16] L.W. MacDonald. Using color effectively in computergraphics. IEEE ComputerGraphicsandApplications, pages20–35,July/Aug.1999.

[17] N. Max, R. Crawfis, and D. Williams. Visualizationfor climate modeling. IEEEComputerGraphicsandApplications, pages34–40,July1993.

[18] P. T. McCaslin,P. A. McDonald,andE. J. Szoke. 3d visualizationdevelopmentatnoaaforecastsystemslaboratory. ComputerGraphics, pages41–44,February2000.

[19] MIT. World WideWeb. cirrus.mit.edu/gempak/forecastmaps.html.

[20] NASA. World WideWeb. http://awin.larc.nasa.gov.

[21] W. Nuss and D. W. Titley. Use of multiquadric interpolationfor meteorologicalobjective analysis.MonthlyWeatherReview, pages1–21,August1994.

[22] U.S. Departmentof Transportation.Aviation Weather:AC 00-6A. US GovernmentPrintingOffice,Washington,D.C.,1975.

[23] U.S. Departmentof Transportation. Aviation WeatherServices:AC 00-45C. USGovernmentPrintingOffice,Washington,D.C.,1985.

31

[24] A. Pang,C.M. Wittenbrink,andS.K. Lodha.Approachesto uncertaintyvisualization.TheVisualComputer, 13:370–390,November1997.

[25] T. S. Perry. Trackingweather’s flight path. IEEE Spectrum, 37(9):38–45,September2000.

[26] P.W. PruynandD.P. Greenberg. Exploring3d computergraphicsin cockpitavionics.IEEEComputerGraphicsandApplications, pages28–35,May 1993.

[27] T. M. Rhyne.Scientificvisualizationin thenext millenium. IEEEComputerGraphicsandApplications, 20(1):20–21,January-February2000.

[28] T. M. Rhyne,M. Bolstad,P. Rheingans,L. Petterson,andW. Shackelford. Visualizingenvironmentaldataat the epa. IEEE ComputerGraphicsand Applications, pages34–38,March1993.

[29] P. K. RobertsonandD. J.Abel. Graphicsandenvironmentaldecisionmaking. IEEEComputerGraphicsandApplications, pages25–33,March1993.

[30] B. Rogowitz andL. Treinish.How notto lie with visualization.Computers in Physics,10(3):268–274,May-June1996.

[31] L. Rosenblum. Visualizing oceanographicdata. IEEE ComputerGraphics andApplications, pages14–19,May 1989.

[32] CharlesH. Scanlon. Cockpit graphical weather information shown to enhanceefficiency, safety, and situationl awareness. In Proceedingsof the 39th AnnualCorporate Aviation SafetySeminar(CASS),Flight SafetyFoundation, pages83–94,April 1994.

[33] F. Schroder and M. Lux.Trivis: Professionaltelevision weatherpresentation.World Wide Web,October1997.http://www.igd.fhg.de/www/igd-a4/projects/docs/trivis/trivis e.html.

[34] Aviation Digital Data Service. World Wide Web.adds.awckc.noaa.gov/projects/adds/index.html.

[35] NOAA National Weather Service. World Wide Web.www.awc-kc.noaa.gov/awc/aviation weathercenter.html.

[36] Shubin,Falck,andJohansen.Exploringcolor in interfacedesign.ACM Interactions,pages36–48,July/Aug.1996.

[37] P. Stough,M. Rudisill, P. Schaffner, andK. Martzaklis.Aviationweatherinformationsystemsresearchanddevelopment.SAEPaperNo. 1999-01-1579,April 1999.

[38] P. StoughandC. Scanlon. Numerousresearchprojectssupportefforts to overcomeweather-relatedhazards.ICAO Journal, pages20–29,January-February1999.

[39] L. A. Treinish. Visualizationof scatteredmeteorologicaldata. IEEE ComputerGraphicsandApplications, 15(4):20–26,July1995.

[40] L. A. Treinish. Task-specificvisualizationdesign. IEEE ComputerGraphicsandApplications, 19(5):72–77,September-October1999.

32

[41] L. A. Treinish. Multi-resolutionvisualizationtechnqiuesfor nestedweathermodels.In Proceedingsof theIEEEVisualizationConference, SanFrancisco,CA, pages513–516.IEEEComputerSocietyPress,October2000.

[42] Edward R. Tufte. TheVisual Display of QuantitativeInformation. GraphicsPress,Cheshire,CT, 1983.

[43] EdwardR. Tufte. EnvisioningInformation. GraphicsPress,Cheshire,CT, 1990.

[44] S. Uckun, C. Ruokangas,P. Donohue,and S. Tuvi. Aware: Technologiesforinterpretingandpresentingaviationweatherinformation. In Proceedingsof theIEEEAerospaceConference, Snowmass,CO, March1999.

[45] AviationWeather. World WideWeb. aviationweather.com.

[46] K. Whitehouse. Weatherwithout the weatherman. IEEE ComputerGraphicsandApplications, pages12–15,March1996.

[47] C. M. Wittenbrink,A. T. Pang,andS.K. Lodha.Glyphsfor visualizinguncertaintyinvectorfields. IEEETransactionsonVisualizationandComputerGraphics, September1996.266–279.

[48] J. Wood, K. Brodlie, and H. Wright. Visualizationover the world wide web andits applicationto environmentaldata. In Proceedingsof the IEEE VisualizationConference, SanFrancisco,CA, pages81–86.IEEEComputerSocietyPress,October1996.

33