This item is only available as the following downloads:

Wind Speed Perception and Risk ( PDF )

PAGE 1

WindSpeedPerceptionandRisk DuzgunAgdas 1 ,GregoryD.Webster 2 ,ForrestJ.Masters 1 1 EngineeringSchoolofSustainableInfrastructure&Environment,CollegeofEngineering,UniversityofFlorida,Gainesville,Florida,UnitedSta tesofAmerica, 2 DepartmentofPsychology,CollegeofLiberalArtsandSciences,UniversityofFlorida,Gainesville,Florida,UnitedStatesofAmerica Abstract Background: Howaccuratelydopeopleperceiveextremewindspeedsandhowdoesthatperceptionaffecttheperceived risk?Priorresearchonhumanwindinteractionhasfocusedoncomfortlevelsinurbansettingsorknock-downthresholds. Nosystematicexperimentalresearchhasattemptedtoassesspeople'sabilitytoestimateextremewindspeedsand perceptionsoftheirassociatedrisks. Method: Weexposed76peopleto10,20,30,40,50,and60mph(4.5,8.9,13.4,17.9,22.3,and26.8m/s)windsin randomizedordersandaskedthemtoestimatewindspeedandthecorrespondingrisktheyfelt. Results: Multilevelmodelingshowedthatpeoplewereaccurateatlowerwindspeedsbutoverestimatedwindspeedsat higherlevels.Windspeedperceptionsmediatedthedirectrelationshipbetweenactualwindspeedsandperceptionsofrisk (i.e.,thegreatertheperceivedwindspeed,thegreatertheperceivedrisk).Thenumberoftropicalcyclonespeoplehad experiencedmoderatedthestrengthoftheactualperceivedwindspeedrelationship;consequently,mediationwas strongerforpeoplewhohadexperiencedfewerstorms. Conclusion: Thesefindingsprovideaclearerunderstandingofwindandriskperception,whichcanaiddevelopmentof publicpolicysolutionstowardcommunicatingtheseverityandrisksassociatedwithnaturaldisasters. Citation: AgdasD,WebsterGD,MastersFJ(2012)WindSpeedPerceptionandRisk.PLoSONE7(11):e49944.doi:10.1371/journal.pone.0049944 Editor: LuisM.Martinez,CSIC-UnivMiguelHernandez,Spain Received March28,2012; Accepted October19,2012; Published November30,2012 Copyright: 2012Agdasetal.Thisisanopen-accessarticledistributedunderthetermsoftheCreativeCommonsAttributionLicense,whichpermits unrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalauthorandsourcearecredited. Funding: TheauthorsthanktheFederalAllianceforSafeHomes(FLASH;http://www.flash.org)forsupportingtheresearch.Publicationofthisarticlewasfu nded inpartbytheUniversityofFloridaOpen-AccessPublishingFund.Thefundershadnoroleinstudydesign,datacollectionandanalysis,decisiontopu blish,or preparationofthemanuscript. CompetingInterests: Theauthorshavedeclaredthatnocompetinginterestsexist. *E-mail:gdwebs@gmail.com Introduction Windistheprimaryagentfortwoofthemostdestructive naturalhazardsonearthhurricanesandtornadoes.Storm preparation,evacuation,andhazardmitigationdependsignificantlyonriskperception[1,2],andeffectivepolicymakingand implementationnecessitatesunderstandinghumanperceptionof hazardsandassociatedrisks[3].Priorexperimentalresearchon windhumaninteractionhasfocusedonpedestriancomfort'in urbanareas[46]byestablishingwindspeedthresholdsthatmake dailytaskschallenging,uncomfortable,orcausepeopletofeel unsafe[7].Findingswerelargelybasedontwo-choicesemantic responses(e.g.,gentle-violent,calm-gale,pleasant-annoying[4,8]) orcharacterizationsofphysicalresponses(e.g.,lossofbalance, shiftsinfootsteptrajectories[9,10]).Surprisingly,however, empiricalresearchislackingon(a)people'saccuracyinperceiving windspeedwhiletheyareexperiencingitand(b)people's perceptionofpersonalriskinresponsetowind.Understanding people'sperceptualaccuracyofextremewindspeedsisimportant becausestormsoftencausemassivepowerandcommunication disruptionsthatleavepeoplewithoutofficialweatherwarningsor reports.Theeffectivenessofweatherwarningsinconveyingthe actualrisksassociatedwithextremewindeventsmaybe suboptimalevenifinformationregardingwindspeedseverityis madeavailable[3,11].Ourgoalinthepresentexperimentwasto addresstheseshortcomingsbyexposingpeopletovariouswind speedstogainabetterunderstandingofwindandrisk perceptionfactorsthatcouldbekeyindevelopingbetterpolicy andwarningsystemsforextremewind-relatedevents.Thismight includesupplementingextremewind-relatedwarningsbyframing theminfamiliarcontexts(e.g.,awindofthisspeedorgreateris enoughtoknockoveraperson''). Thispresentexperimentexaminedhumanperceptionof extremewindsandassociatedrisks.Becausepriorresearchin establishedperceptualdomains(e.g.,vision,audition,justnoticeabledifferencesinweightperception)hasshownhumanperceptiontobeanonlineartransformofphysicalstimuli(e.g.,Weber Fechnerlaw),wepredictedthatpeoplewouldoverestimatewind speedsathighervelocities,andthatriskperceptionswouldfollowa similaracceleratingtrajectory.Wealsoexpectedthatwindspeed perceptionwouldmediatetherelationshipbetweenactualwind velocityandperceptionsofrisk;overestimatesofwindspeedwould relatetomoreperceivedrisk.Onanexploratorybasis,wealso examinedtheextenttowhichindividualdifferencesinpriorstorm experiencemoderatedtheserelationships. Methods EthicsStatement EthicalstandardsoutlinedbytheAmericanPsychological Associationwerefollowedintheconductofthisresearch,which wasapprovedbytheUniversityofFloridaInstitutionalReview PLOSONE|www.plosone.org1November2012|Volume7|Issue11|e49944

PAGE 2

Board.Allparticipantsgavetheirsignedconsentpriortopartaking intheexperiment. ParticipantsandProcedure Seventy-sixcollege-agestudents(18women,58men)aged18to 40years( M =23.47, SD =4.68)participatedinthestudy. Participantswerefirstgivensurveysontheirpriorexperiences withandbeliefsaboutextremeweatherphenomenaand associateddecision-making.Next,participantsdonnedprotective gear(goggles,waders,andhoodedraincoats)andaharnessthat attachedtoahandrailsystemlocated8ftdownwindofthejet, whichtheywereallowedtohold.Participantswerethenexposed to10,20,30,40,50,and60mph(4.5,8.9,13.4,17.9,22.3,and 26.8m/s)windspeedsfor20-sintervalsinpredetermined randomizedorders(seeVideoS1).Inbetweeneachwindexposure event(whichlasted < 10s),participantscommunicatedtheir estimateofthewindspeedandtheirestimateofpersonalinjury riskonascaleof0( noperceivedrisk )to10( dangerous )toanobserver standingoutsidethewindfield.Thus,thetotalexperiment durationwas < 3minforeachparticipant.Thetestingconditions (windspeedintensities,totalexposuretime,gear)wereidenticalfor allparticipants;onlytheorderofwindspeedswasrandomizedto controlforpossibleordereffects.Participantsweregivenno informationonthewindspeedintensitiestopreventpossiblebias; however,theywereinformedofthewindspeedsafterthe experiment. WindApparatus Eight54-in(1.37-m)diametervaneaxialfansforcedairthrough a10-ft 6 10-ft(3.05-m 6 3.05-m)squarejettogeneratethewind fieldinthetestchamber(Figure1).Hydraulicpowertothefans wasindividuallycontrolledtoregulatetheangularvelocityofthe fanstoreachadesiredflow.AnRMYoungWindMonitorlocated inthetestchambermeasuredwindspeed,whichwasreadbythe equipmentoperator. DataAnalysis Becauserepeatedestimateswerenestedwithinparticipants,we analyzedthedatawithmultilevelmodeling(MLM[12,13])using HLM[14]andMplus[15].Usingmaximumlikelihoodestimation,MLMcanmodelwithin-andbetween-personeffects simultaneously.Within-person(orbetween-trial)varianceinwind orriskperceptionwasmodeledatlevel1,andbetween-person variancewasmodeledatlevel2asafunctionofbetween-person means(intercepts)and,insomemodels,individualdifferencesin numberoftropicalcyclonesexperienced(i.e.,thetropicalstorms withsustainedwinds $ 39mphor63km/h,hereafterreferredto inshorthandasstorms'').Forexample,inoneanalysiswe modeledwindspeedperceptionasafunctionofactualwindspeed (level1)andnumberofstormsexperienced(level2).Thelevel-1 modelwas: WindSpeedPerception ti ~ p 0 i z p 1 i ActualWindSpeed !" i z p 2 i ActualWindSpeed 2 !" i z e ti 1 wherePerception ti representsthewindspeedestimateforSpeed t byPerson i .Eachperson'sPerceptionscoresaremodeledas functionsoftheirmeanorintercept( p 0 i )andthelinear( p 1 i )and quadratic( p 2 i )effectsofactualwindspeed.Theerrorterm, e ti capturesthelevel-1residualvarianceforeachperson. InMLM,thelevel-1interceptsandslopesforeachpersonare modeledatlevel2asafunctionofindividualdifferencesinthe numberoftropicalcyclonesexperienced(grand-mean-centeredat 5.05storms): p 0 i ~ b 00 z b 01 Storms !" z r 0 i p 1 i ~ b 10 z b 11 Storms !" z r 1 i p 2 i ~ b 20 z b 21 Storms !" z r 2 i 2 Here, p 0 i againrepresentsthemean(intercept)foreachperson. The b 00 coefficientrepresentsthegrandmeanthebetweenpersonaverageofeachperson'saverageintensityscoreforthe averagenumberofstormsexperienced.Thecoefficients b 10 and b 20 representthebetween-personaverageofthewithin-person linearandquadraticeffects(respectively)ofactualwindspeedon windspeedperceptions.Thecoefficients b 01 b 11 ,and b 21 representtheextenttowhichthewithin-personinterceptsand linearandquadraticeffects(respectively)aremoderatedby individualdifferencesinthenumberoftropicalcyclonespeople haveexperienced.Theerrorterms r 0 i r 1 i ,and r 2 i capturethelevel2residualvariancefortheirrespectiveeffects. Inthemultilevelmoderatedmediationmodelsbelow,thisMLM frameworkisexpandedtoincludemediationatlevel1witha continuouslevel-2moderator(numberoftropicalcyclones experienced).Wefollowedproceduresoutlinedinpriorwork [1618]. Results DescriptiveStatistics Table1listsdescriptivestatisticsforperceivedwindspeedand riskbyactualintensity. WindSpeedPerceptionasaFunctionofActualWind Speed Boththelinear( b 10 =1.311, SE =0.054, t 75 =24.23,partial correlation[ r p ]=.94)andquadratic( b 20 =0.0061, SE =0.0016, t 75 =3.80, r p =.40)effectsofactualwindspeedsonpeople's perceptionsofwindspeedsweresignificantlypositiveforthe averageperson( p s .05;Figures2and3).Forthismodel,76%and 24%ofthevariancewasatthebetween-andwithin-personlevels, respectively.Theaveragepersonwasreasonablyaccurateand perceptionwasfairlylinearatslowerwindspeeds,butthe perceivedwindspeedsdepartedfrombothaccuracyandlinearity athigherwindspeeds. Simpleeffectstests[19]showedthattheaverageperceptiondid notdiffersignificantlyfromactualwindspeedsat10and20mph (4.5and8.9m/s);however,beginningat30mph(13.4m/s),the averagepersonprogressivelyoverestimatedtheactualwindspeeds (Table2,left;Figures2and3).Wealsotestedtheextenttowhich theaverageperceptionsfitordepartedfromaone-to-oneaccuracy slopeacrossthesixwindspeeds.Thesimpleslopebetween perceivedandactualwindspeedswascomputedforeachoneof thesixwindspeedlevels(i.e.,1060mph;4.526.8m/s).Thisis equivalenttoaskingwhetherthelinestangenttothecurveateach oneofthesixspeedsissignificantlydifferentthantheone-to-one line(Figure4).At10mph(4.5m/s),thesimpleslopewasnot significantlydifferentfromaone-to-onerelationship;however, startingat20mph(8.9m/s),thesimpleslopesweresignificantly morepositivethantheone-to-onerelationship,suggestingthat peoplebecamelessaccurateaboutthewindfunction(departed fromlinearity)aswindspeedsincreased(Table2,right),whichis tobeexpectedasthewindforcesexertedonthehumanbodyare proportionaltothewindspeedsquared. WindSpeedPerceptionandRisk PLOSONE|www.plosone.org2November2012|Volume7|Issue11|e49944

PAGE 3

RiskasaFunctionofActualWindSpeed Boththelinear( b 10 =0.1336, SE =0.0045, t 75 =30.01, r p =.96) andquadratic( b 20 =0.00043, SE =0.00020, t 75 =2.14, r p =.24) effectsofwindspeedonpeople'sperceptionsofriskwereboth significantlypositivefortheaverageperson( p s .05;Figure5).A slopeof0.1336inModel1indicatesthat,forevery10-mph(4.5m/s)increaseinwindspeed,theaverageparticipant'sperception ofriskincrease1.336unitsona010scale.Forthismodel,67% and33%ofthevariancewasatthebetween-andwithin-person levels,respectively.Theaverageperson'sriskfunctionforactual windspeedswascurvilinearandconcaveup(accelerating; Figure5).Thistrendwassupportedviaaseriesofsimpleeffects testsateachwindspeed;forexample,thesimpleslopesat10and 60mph(4.5and26.8m/s)were0.112( SE =0.011, t 75 =9.80, r p =.75)and0.155( SE =0.011, t 75 =14.66, r p =.86),respectively ( p s .05). RiskasaFunctionofWindPerception Perceptionsofwindandriskwerelinearlyrelated( b 10 =0.1031, SE =0.0035, t 75 =29.46, p .05, r p =.96);nosignificantquadratic effectwaspresent( r p = 2 .06;Figure6).Aslopeof0.1031indicates that,forevery10-mph(4.5-m/s)increaseinperceivedwindspeed, theaverageparticipant'sperceptionofriskincreasedabout1.031 unitsona010scale.Forthismodel,65%and35%ofthe Figure1.Designandphotographsofthewindsimulator. Theupperleftpanelshowsthewindsimulator'sdesign;theotherthreepanels showphotographsofitfromdifferentangles. doi:10.1371/journal.pone.0049944.g001 Table1. Descriptivestatisticsforwindandriskperceptionsbyactualwindspeed. PerceivedWindSpeed(mph)PerceivedRiskona1to10Scale ActualWindSpeedmph(m/s)Range Mdn Mean SD Skew. Exc. Kurt.Range Mdn Mean SD Skew.Exc.Kurt. r 10(4.5)16010.010.28.23.4218.05051.00.810.871.846.12.15 20(8.9)44020.020.69.30.51 2 0.28052.01.681.120.43 2 0.17.45* 30(13.4)107530.033.713.60.580.21073.03.211.54 2 0.03 2 0.23.48* 40(17.9)109045.045.217.50.43 2 0.31194.04.461.690.661.23.72* 50(22.3)1511557.560.419.40.550.303106.05.991.790.26 2 0.51.45* 60(26.8)3013075.075.825.40.22 2 0.502108.07.341.89 2 0.57 2 0.18.65* Note. Nesting not takenintoaccountforthistable;dataaveragedacrosspersonsratherthanexaminingdatawithinpersons.Skew.=Skewness.Exc.Kurt.=Excess Kurtosis. r =correlationbetweenwindperceptionsandriskperceptions. N s=76participants,454observations(2datapointsmissingduetoproceduralerror). p .05. doi:10.1371/journal.pone.0049944.t001 WindSpeedPerceptionandRisk PLOSONE|www.plosone.org3November2012|Volume7|Issue11|e49944

PAGE 4

variancewasatthebetween-andthewithin-personlevels, respectively. NumberofStormsExperiencedModeratestheActual PerceivedWindRelationship Wetestedtheextenttowhichindividualdifferencesinthe numberoftropicalcyclonespeoplehadexperienced( n =75; range:010-or-more; Mdn =5.0, M =5.0, SD =3.0)moderatedthe within-personactualperceivedwindspeedrelationships.The purposewastodeterminewhetherthenumberoftropicalstorms peoplehadexperiencedrelatestotheaverageactualperceived windspeedrelationship.Numberofstormsexperiencedsignificantlymoderatedthelinear( b 11 = 2 0.037, SE =0.015, t 73 = 2 2.48, p .05, r p = 2 .28)butnotthequadratic ( b 21 = 2 0.00078, SE =0.00049, t 73 = 2 1.57, p =.12, r p = 2 .18) effectofactualwindspeedonwindperceptions(Figure7).(Itis unlikelythatfourparticipantsexperienced10ormoretropical cyclonesbasedontheirageandhistoricdata.Thereported exposureinaccuraciesmightrelatetomisperceptionsaboutthe environmentalconditionsthatconstitutetropicalcyclones.Nevertheless,whenwere-ranthemodelwithoutthesefourparticipants,numberofstormsexperiencedstillmoderatedthelinear effectofactualwindspeedsonwindspeedperception, b 11 = 2 0.034, SE =0.016, t 69 = 2 2.01, p .05, r p = 2 .23). Wedecomposedthismodelbyconductingsimpleeffecttestsat theminimum(0)andmaximum(10)reportedvaluesfornumberof tropicalcyclonesexperienced.Forpeoplewhoexperiencedno storms,boththelinear( b 10 =1.50, SE =0.10, t 73 =14.70, r p =.86) andquadratic( b 20 =0.0101, SE =0.0029, t 73 =3.52, r p =.38) effectsofactualwindspeedsonperceivedwindspeedwere significant( p s .05);moreover,people'saveragelinearslopeswere significantlydifferentfromaone-to-onerelationship( b 10 =0.50, SE =0.10, t 73 =4.93, p .05, d =1.15;Figure7,thicklight-gray curve).Incontrast,forpeoplewhoexperiencedtenormore storms,therelationshipwasstrictlylinear( b 10 =1.130, SE =0.081, t 73 =14.00, p .05, r p =.85)thequadraticeffect( b 20 =0.0023, SE =0.0030, t 73 =0.78, p =.44, r p =.09)wasnon-significant; moreover,people'saveragelinearslopesdidnotdiffersignificantly fromaone-to-onerelationship( b 10 =0.130, SE =0.081, t 73 =1.61, p =.11, d =0.38;Figure7,thickblackcurve).Onanexploratory basis,wealsotestedthesimplemoderationeffectofnumberof stormsexperiencedonwindperceptionat60mph(26.8m/s).At 60mph(26.8m/s),numberofstormsexperiencedmarginally (i.e., p .06)moderatedpeople'sperceptionsofwindspeed ( b 01 = 2 1.65, SE =0.85, t 73 = 2 1.95, p =.055, r p = 2 .22; Figure7,rightmostendsofcurves). Figure2.Multilevelmodelingresultsforperceivedwindspeedasafunctionofactualwindspeed. Thingraylinesrepresentindividual predictedcurvesfor76participants.Thickblacklinerepresentstheaveragecurve.Thinblacklinerepresentsaone-to-onerelationship. doi:10.1371/journal.pone.0049944.g002 Figure3.Perceivedwindspeedasafunctionofactualwind speed. Pointestimatesand95%confidenceintervalsareshownforthe averageslopeforeachwindspeedtested. doi:10.1371/journal.pone.0049944.g003 WindSpeedPerceptionandRisk PLOSONE|www.plosone.org4November2012|Volume7|Issue11|e49944

PAGE 5

WindPerceptionMediatestheWindRiskRelationship andStrengthenswithInexperience Wetestedamultilevelmoderatedmediationmodel(mediation atthelowerlevel,moderationattheupperlevel[16,17])toassess (a)ifwindperceptionsmediatedthedirectrelationshipbetween actualwindandriskperceptionsand(b)ifindividualdifferencesin experiencewithtropicalcyclonesmoderatedthestrengthofthe mediation.Wetestedonlylineareffectsbecause(a)theywere substantiallystrongerthanthequadraticeffectsand(b)the quadraticeffectofactualwindspeedonriskwasnon-significant aftercontrollingforwindperceptions.Becausealldirecteffects remainedsignificant,allresultsshowedpartial(vs.complete) mediation.AsshowninFigure8a,whenassessedatthemean numberoftropicalcyclonesexperienced,thedirectrelationship betweenactualwindandriskwassignificantlyattenuatedafter controllingforwindperception;theindirecteffectviawind perceptionwassignificant.Thedirectandindirecteffects accountedfor38%and62%ofthetotaleffect,respectively. Wenexttestedthestrengthofthemediation(viasimpleeffects tests)forpeoplewhohadexperiencednostormsor10-or-more storms(Figures8band8c,respectively).Peoplewhohad experiencednostormshadanespeciallystrongactualperceived windrelationshipandshowedasignificantmediationpattern.The directandindirecteffectsaccountedfor36%and64%ofthetotal effect,respectively.Incontrast,peoplewhohadexperienced10or morestormshadaweakerbutmoreaccurate(theiraverage slopedidnotdiffersignificantlyfromaone-to-onerelationship) actualperceivedwindrelationshipandshowedanon-significant mediationpattern,becausethe95%CIfortheindirecteffect includedzero(Figure8c).Thedirectandindirecteffects accountedfor41%and59%ofthetotaleffect,respectively. Figure4.Perceivedwindspeedasafunctionofactualwindspeed:Simpleslopes. Examplesofsimpleslopestangenttotheaveragecurve (thicksolidline)at20(dottedline)and50(dashedline)mph(8.9and22.3m/s).Slopesareshowninreferencetoaone-to-onerelationship(thinsoli d line). doi:10.1371/journal.pone.0049944.g004 Table2. Simpleeffects:Windperceptionasafunctionofactualwindspeed. ActualWindSpeedmph(m/s) Intercept(differencefromactual)Slope(differencefromone-to-one) b 00 SEt 75 d b 10 SEt 75 d 10(4.5)0.2510.8800.280.080.0060.0710.080.02 20(8.9)0.9201.0900.840.220.1280.0522.48*0.66 30(13.4)2.7981.3852.02*0.540.2510.0485.19*1.37 40(17.9)5.9101.6933.49*0.920.3720.06355.85*1.55 50(22.3)10.2342.1304.80*1.270.4930.0885.59*1.48 60(26.8)15.7702.8395.55*1.470.6140.1165.28*1.40 Note.N s=76participants,454observations(2datapointsmissingduetoproceduralerror). p .05. doi:10.1371/journal.pone.0049944.t002 WindSpeedPerceptionandRisk PLOSONE|www.plosone.org5November2012|Volume7|Issue11|e49944

PAGE 6

Discussion Webeganbyaskinghowaccuratelypeopleperceiveextreme windspeedsandhowtheirestimatesaffecttheirperceptionsof personalrisk.Thesearekeyquestionsforboththepsychologyof humanperceptionandpublicpolicyinresponsetoextremewindrelatedweatherevents.Thestudyresultsindicatethatonaverage (a)peopleoverestimatehigherwindspeeds( $ 20mphor8.9m/s) Figure5.Multilevelmodelingresultsforperceivedriskasafunctionofactualwindspeed. Thingraylinesrepresentindividualpredicted curvesfor76participants.Thickblacklinerepresentstheaveragecurve. doi:10.1371/journal.pone.0049944.g005 Figure6.Multilevelmodelingresultsforperceivedriskasafunctionofperceivedwindspeed. Thingraylinesareindividualpredicted curvesfor76participants.Thickblacklinerepresentstheaveragecurve. doi:10.1371/journal.pone.0049944.g006 WindSpeedPerceptionandRisk PLOSONE|www.plosone.org6November2012|Volume7|Issue11|e49944

PAGE 7

butarereasonablyaccurateatjudginglowerwindspeeds,(b)the simplerelationshipbetweenwindspeedandperceivedrisk becomesincreasinglypositiveathigherwindspeeds,(c)wind perceptionmediatestherelationshipbetweenactualwindspeed andrisk,and(d)thismediationpatternisstrongeramongpeople withnopriorexperiencewithtropicalcyclonesandweakenswith exposuretoeachadditionalstorm. Thenewknowledgegeneratedbythisresearchisusefulnotonly becauseitexpandsourunderstandingofhowpeopleperceive windandwind-relatedriskonapsychologicallevel,butalso becauseithaspotentiallylife-savingpublicpolicyimplicationson howinformationiscommunicatedpriortoandduringextreme weatherevents(e.g.,tornadoes,hurricanes).Althoughtheaverage personoverestimateshigherwindspeeds(e.g.,theyperceive60mph[26.8-m/s]windstobe75mph[33.5m/s]),thisrelationship ismoderatedbyindividualdifferencesinstormexposure;with eachadditiontropicalcyclonepeopleexperienced,peoplemade moreaccuratewindspeedestimatesonaverage.Thissuggeststhat exposuretorealstormsmayhelpcalibratepeople'sperceptions regardinghigherwindspeeds.Withsomeexposure,peoplemaybe abletogaugewindspeedsmoreaccurately.Futureresearchshould strivetoexaminetheprocessesbywhichindividualdifferencesin perceptualandriskjudgmentsform.Nevertheless,ourresultsalso highlightadisconnectbetweenwindperceptionandreality, perhapsbecauseofpeople'slackofexposuretohigh-velocitywind speeds.Forexample,inFloridathemosthurricane-pronestate intheU.S.manycoastalresidentshailfromoutsidethestate [20]andthushavenopriorexperiencewithlandfalling hurricanes.Thefindingsindicatedthatpeoplewhohavenot experiencedsustainedtropicalstormorhurricane-forcewindsare morepronetooverestimatingwindspeed,whichmaynegatively affecttheirdecision-makingaboutpreparationandevacuation. Forexample,amajorcivilproblemwithgovernment-issued evacuationsisthephenomenonofshadowevacuation,''inwhich peoplewhodo not needtoevacuatechosetodosoanyway, therebyunnecessarilyexacerbatingtrafficjamsalongevacuation routes,andfillinglimitedspacesinsheltersandhotelrooms[21]. Furtherresearchisrequiredtovalidatethesefindings,notonly Figure7.Perceivedwindspeedasafunctionofactualwind speedandnumberofstormsexperienced. doi:10.1371/journal.pone.0049944.g007 Figure8.Multilevelmoderatedmediationmodelresults. PanelA showsamoderatedmediationmodelshowingthat(a)perceivedwind speed(mph)partiallybutsignificantlymediatedtherelationshipsbetweenactualwindspeed(mph)andperceivedrisk(linearrelationshipsonly) and (b)numberofstormsexperienced(grand-meancentered)moderatedtherelationshipbetweenactualandperceivedwindspeed. PanelsB and C showsimpleeffectstestsofthemediationmodelatzeroandten-or-morestormsexperiences,respectively.Valuesareunstandardizedregression coefficients[95%CIs].Thedirectrelationshipbetweenactualwindspeedandperceivedriskisshowninparentheses. doi:10.1371/journal.pone.0049944.g008 WindSpeedPerceptionandRisk PLOSONE|www.plosone.org7November2012|Volume7|Issue11|e49944

PAGE 8

withamorediversesample,butalsowithamorespecificmeasure ofriskthatcandistinguishbetweenprobabilityandseverity. Althoughadditionalresearchneedstobedonebeforewecan makeanystrongrecommendationsregardingpublicpolicy,the presentresearchmaysuggestthepossibilityforintroducingarisk metricorcontextualaidtocharacterizewindspeedsinstorm advisories,particularlygiventhatthedifferencebetweenperceived andactualwindspeedcanbeshowntobeinterpretedasa differenceofoneortwocategoriesontheSaffirSimpson HurricaneWindScaleafive-categoryclassificationsystemfor hurricaneintensitybasedonwindvelocity.Suchadualsystemis usedforhailadvisories,wheretheU.S.NationalWeatherService reportsbothhaildiameterinformation(infractionalinches)and thesizeofacommonobject(e.g.,dime-sized,''quarter-sized,'' softball-sized'').Perhapswindspeedscouldbeaccompaniedby relevantinformationsuchasthiswindspeedissufficienttoknock overtheaverageperson.''Nevertheless,wecautionthatthe presentresearchispreliminary,andadditionalresearchthat focusesonpublicpolicyapplicationswillneedtobeundertaken beforeanyrecommendationscanbemade.Wehopeournovel experimentalfindingsonwindperceptionwillinformnotonlythe psychologyofriskbutalsofutureresearchonthebroaderpolicy implicationsextremeweatherpreparationandresponse. SupportingInformation VideoS1 Thisvideoshowsparticipantsbeingexposed tovariouswindspeedinthewindsimulator(seeMethod sectionoftextfordetails). Inthisarticle,participantswere exposedtodrywinds;however,forbettervisualization,thisvideo showswind-drivenrain,whichwasappliedseparatelyfora companionstudy. (MOV) Acknowledgments Wethankthefollowingengineers,technicians,andgraduatestudentsfor theirhelpincarryingoutthisresearch:SevcanAgdas,Antonio Balderrama,ScottBolton,CoreyCook,KevinFrost,AlexEsposito, AmandaGesselman,JamesJesteadt,CarlosLopez,CarlosRodriguez,and JasonSmith.Preliminaryfindingsfromthisresearchwerepresentedatthe NationalHurricaneCenter,Miami,Florida(February2012),atthe NationalHurricaneConferenceinOrlando,Florida(March2012),the FloridaGovernor'sHurricaneConferenceinFt.Lauderdale,Florida(May 2012),andtheSocietyforJudgmentandDecisionMakinginMinneapolis, Minnesota(November2012). AuthorContributions Conceivedanddesignedtheexperiments:DAGDWFJM.Performedthe experiments:DA.Analyzedthedata:DAGDW.Contributedreagents/ materials/analysistools:DAGDWFJM.Wrotethepaper:DAGDWFJM. References 1.HorneyJA,MacdonaldPDM,VanWilligenM,BerkePR,KaufmanJS(2010) Individual,actual,orperceivedpropertyfloodrisk:Diditpredictevacuation fromhurricaneIsabelinNorthCarolina,2003?RiskAnalysis30:501511. 2.PeacockWG,BrodySD,HighfieldW(2005)Hurricaneriskperceptionsamong Florida'ssinglefamilyhomeowners.LandscapeandUrbanPlanning73:120 135. 3.SlovicP(1987)Perceptionofrisk.Science236:280285. 4.HuntJCR,PoultonEC,MumfordJC(1976)Theeffectsofwindonpeople:New criteriabasedonwindtunnelexperiments.BuildingandEnvironment11:15 28.doi:10.1016/0360-1323(76)90015-9 5.MelbourneWH(1978)Criteriaforenvironmentalwindconditions.Journalof WindEngineeringandIndustrialAerodynamics3:241249.doi:10.1016/01676105(78)90013-2 6.PenwardenAD,GriggPF,RaymentR(1978)Measurementsofwinddragon peoplestandinginawindtunnel.BuildingandEnvironment13:7584. doi:10.1016/0360-1323(78)90026-4 7.BottemaM(2000)Amethodforoptimisationofwinddiscomfortcriteria. BuildingandEnvironment35:118.doi:10.1016/S0360-1323(98)00065-1 8.JacksonPS(1978)Theevaluationofwindyenvironments.Buildingand Environment13:251260.doi:10.1016/0360-1323(78)90016-1 9.JordanSC,JohnsonT,SterlingM,BakerCJ(2008)Evaluatingandmodelling theresponseofanindividualtoasuddenchangeinwindspeed.Buildingand Environment43:15211534.doi:10.1016/j.buildenv.2007.08.004 10.MurakamiS,DeguchiK(1981)Newcriteriaforwindeffectsonpedestrians. JournalofWindEngineeringandIndustrialAerodynamics7:289309. doi:10.1016/0167-6105(81)90055-6 11.DowK,CutterS(2000)Publicordersandpersonalopinions:Household strategiesforhurricaneriskassessment.GlobalEnvironmentalChangePartB: EnvironmentalHazard2:143155.doi:10.1016/S1464-2867(01)00014-6 12.NezlekJB(2011)Multilevelmodelingforsocialandpersonalitypsychology. ThousandOaks,CA:Sage. 13.RaudenbushSW,BrykAS(2002)Hierarchicallinearmodels:Applicationsand dataanalysismethods.ThousandOaks,CA:Sage. 14.RaudenbushSW,BrykAS,CheongYF,CongdonR(2004)HLM6: Hierarchicallinearandnonlinearmodeling.Lincolnwood,IL:Scientific SoftwareInternational. 15.Muthe nLK,Muthe nBO(2010)Mplususer'sguide(6thed.).LosAngeles,CA: Muthe n&Muthe n. 16.MullerD,JuddCM,YzerbytVY(2005)Whenmoderationismediatedand mediationismoderated.JournalofPersonalityandSocialPsychology89:852 862.doi:10.1037/0022-3514.89.6.852 17.BauerDJ,PreacherKJ,GilKM(2006)Conceptualizingandtestingrandom indirecteffectsandmoderatedmediationinmultilevelmodels:Newprocedures andrecommendations.PsychologicalMethods11:142163.doi:10.1037/1082989X.11.2.142 18.PreacherKJ,ZyphurMJ,ZhangZ(2010)AgeneralmultilevelSEMframework forassessingmultilevelmediation.PsychologicalMethods15:209233. doi:10.1037/a0020141 19.AikenLS,WestSG(1991)Multipleregression:Testingandinterpreting interactions.NewburyPark,CA:Sage. 20.TobinG(1999)Sustainabilityandcommunityresilience:Theholygrailof hazardsplanning?GlobalEnvironmentalChangePartB:Environmental Hazards1:1325.doi:10.1016/S1464-2867(99)00002-9 21.SteinRM,Duen as-OsorioL,Buzcu-GuvenB,SubramanianD,KahleD(2011) Howriskperceptionsinfluenceevacuationsfromhurricanes.JamesA.BakerIII InstituteforPublicPolicyofRiceUniversity.http://www.bakerinstitute.org/ publications/POL-pub-SteinHurricaneRiskPerception-080311.pdf WindSpeedPerceptionandRisk PLOSONE|www.plosone.org8November2012|Volume7|Issue11|e49944