Professional Documents
Culture Documents
Category: ICTs,ClimateChangeandDisasterManagement
Initiative Overview
Fiercemonsoonrainsin2010causedtheworstfloodinginPakistanin80years(AonBenfield2010). ApproximatelyonefifthofPakistan'stotalareawasunderwaterattheheightoffloods.Thefloods displacedandaffected20millionpeople,mostlyviadestructionofproperty,livelihoodand infrastructure.Closeto2,000peoplediedasaresult.Althoughthespecificlinktoclimatechangeis unclear,itisgenerallyagreedthatclimatechangeisalreadyincreasingtheriskoffloodingandthat thisriskwillgrowinfuture(IPCC2007).Inparticular,Pakistanstandsamongthegroupof developingcountrieswhichareextremelyvulnerabletotheadverseimpactsofclimatechange. (PlanningCommission2010:13).Inrecentyears,climatechangeinPakistanisevidencedbyan increaseintemperatureinsummerandtimingchangesofthemonsoonseason(Dell'Amore2010). Thesechangesareaffectingagriculturalpatterns,buttheyalsothreatenagreaterincidenceand intensityofnaturaldisasters,includingfloods. Duringsuchdisasters,informationisatapremium:thereisanurgentneedtoknowwhichareasare affectedhowtheyareaffectedwhatthepriorityproblemsareandsoforth.Thistypicallycoincides withthedisruptionoftraditionallinesofcommunication.Theadventofmobilephoneshasprovideda newdigitaldevelopmentinfrastructure,whichmaybeofsignificantvaluetodisasterresponse. Pakistanhasseenparticularlystronggrowthinmobiles,withroughly110msubscriptionsinmid2011 wellinexcessoftheadultpopulation(PTA2011). ThiscasestudyfocusesonPakreport,anICTinitiativebetweencrisismappingorganisations, engineers,reliefagenciesandcrowdsourcingcompaniesthatbeganinJuly2010asaresponsetothe floods. WewillreviewemergencycommunicationusingICTtoolsandvolunteersusingcrowdsourcing platformstoperformverification,categorisation,translation,andmappingoftheinformationinreal time.ThisinitiativedemonstratesICTs'impactonclimatechangeinthedomainsofbothadaptation (disastermanagement)andmonitoring(Heeks2009). Thiscasestudywillextendpreviousresearch oncrowdsourcedworkflowsforcrisisreliefinKenyaandHaiti.
Application Description
1 Pakreport is a customisation of Ushahidi software in Pakistan which employs two forms of crowdsourcing. First, the use of a distributed group of people to provide data reports from the ground.Second,theuseofa(verydifferent)distributedgrouptotranslate,categoriseandgeolocate 2 theincomingmessages this beingundertaken viaaCrowdFlower microtask . Oncetheinformation was processed, it wasinput and displayed on the Pakreport.org platformmost visibly via an online map:seeFigure1.
1 2
CASE STUDY
Category: ICTs,ClimateChangeandDisasterManagement
Figure1:ThePakreport.orgOperationalFlowandStakeholders ThePakreportplatform th wasdeployedon8 August2010,andcould acceptdatainputin manydifferentforms:via radiochannels,viasocial mediasitessuchas twitter,anddirectly onlineviathewebsite itself.However,the primarysourceofinitial informationwasasetof villagelevelassessments directfromtherelief agencieswhichdetailed thesituationanddamage ontheground.These reportswerecategorised andmappedvia Pakreportstaff. Theteamalsowanteda simplewayforordinary peoplecaughtupinthe disastertoreporttheir situation.SMSwasseen asthemosteffective meansforthis,andthe teamatPakreportsetup ashortcode3441 thatwasavailableon fourofthefivemobile companiesinPakistan. Itwasasharedshort code,sotheFLtag neededtobeappended beforethemessage. Detailsoftheshortcodewiththemessagewhatyouseeaboutfloodswasspreadviathemass mediainparticularviaapartnershipwiththeBBCWorldServiceandviareliefagencyworkers. Thisledtoasubstantialincreaseinthevolumeofdatabeingreceived,withtheSMSchanneltaking overastheprimaryinformationsource.Thiscreatedaneedtocategorisetheincomingmessages dependingonwhatwasbeingreported,andthen(iflocationaldetailswereavailablewiththeSMS) geolocatethemessagesothatitcouldbemapped. MessagesmightalsoneedtobetranslatedfromPashtoorUrdutoEnglish.Itwasdecidedto crowdsourcethesemicrotasksviapartnerCrowdFlower.Volunteersfromaroundtheworld participatedincompletionofthemicrotasks(seeFigure2).Becauseaccuracywascriticaltothese efforts,CrowdFloweraddedalevelofredundancyforimprovedqualitycontrol(ascomparedto previousdisasterreliefworkflows),meaningthatmultiplevolunteersevaluatedeachSMSmessage. Thisresultedinthecollectionofmultiplepointsonamap,whichtheCrowdFlowerplatformthenused tocalculatethecentroidofthepoints.Thisincreasedtheaccuracyofthefinalresultsbeingplaced ontotheonlinemap.
CASE STUDY
Category: ICTs,ClimateChangeandDisasterManagement
Figure2:CrowdFlowerDisasterMessageAnalysisMicrotaskForms
Becausemapsofanddetailsaboutcertainlocations(especiallyruralvillagesandsmalltownsin remoteareasofPunjabandSindhprovinces)werenotwelldevelopedorreadilyavailable,the Pakreportteamcreateddetailedonlinedocumentation(http://groups.google.com/group/PakReport volunteers/web/mappinglinks)tosupplementthesegaps.Theteamalsoprovidedtrainingthrough Skypetomanage40volunteerswhoassistedwiththispartoftheprocess. Themapwithdetails,labelsandannotationsfromthemessagesandotherincomingdataflows wasmadeavailabletoreliefagenciesinPakistan,providingrealtimereportingforthem.Itwas integratedwithvariousICTimplementationsnationallyintheUNandNGOsector(e.g. http://www.pakresponse.info)andtheNationalDisasterManagementAuthority(NDMA)and provinciallyintheProvincialDisasterManagementAuthorities(PDMAs).Themapenhancedthe responseeffortsbyimprovingprioritisationandcoordinationofthedisasterresponse.
CASE STUDY
Category: ICTs,ClimateChangeandDisasterManagement
Stakeholders
AsillustratedinFigure1,thisprojectbroughttogetheraverywiderangeofstakeholders.Themain stakeholderswerethelocalcommunitiesandfloodaffectedpeoplewhousedmobilephonesto communicatetheiremergencysituation.Reliefagenciesandthegeneralpublicwhosought informationaboutthecrisissituationandwhocouldrespondtohelprequestsweretheend stakeholderswhoreceivedtheinformation.Betweentheseweremanyothers: Mediaentities(e.g.BBCWorldServiceradio,Internews)whohelpedspreadthewordabout Pakreport ThecorePakreportteam Variousvolunteergroups(CrisisCommons,CrisisMappers,HumanityRoad)whoparticipatedvia CrowdFlowerindataanalysisorwhogavedirectassistancetoPakreport Individualvolunteers(fromFletcherUniversity)whohadworkedonsimilarexercisesinthepast Technologypartners(Pakistanicellphoneproviders,andCrowdFlower)
Collectionof1500realtimereportsfromthepeopleongroundthroughSMS. Translation,categorisationandmappingofreportsinnearrealtime.Crowdvolunteerscompleted over2500labelsorcategorisationsofreports. DetailedknowledgeandmappingresourcesorganisedbythePakreportteam(toviewthesemap resources,pleasevisitwww.pakreport.orgorrefertoFigure1.) AbaseforfuturecrowdsourcingandmobileimplementationsinPakistan.Itwouldbeeasyto repurposeorduplicatethistypeofproject.Thegrowthandsocializationofshortcodeuseand shortcodedisasterreportingisanimportantcomponentofthefutureuptakeofsimilar implementations. Awarenessofanddialogueabouttheeffectsofclimatechange.Thecollaborativereportingabout thefloodinganditsaftermathcontributedtoenvironmentalawarenessaswellasaclearchannel forreportingandmonitoringtheenvironmentalchangesthroughoutPakistan.Pakreportalso representsanimportantandinnovativetechnologicalfoundationforongoing,nationallevel reporting,monitoringand/orearlywarningefforts.
CASE STUDY
Category: ICTs,ClimateChangeandDisasterManagement
Figure3:ShareofMessagesbyCategory(Source:Pakreport)
CASE STUDY
Category: ICTs,ClimateChangeandDisasterManagement
Figure4:WorkerErrorRatesvs.NumberofTasksCompleted(byWorkerGroup) (Source:Pakreport)
Thefirstlineshown,perworkererror,issimplytheaveragedegreesdifferenceperworkerafterthey rd havecompleted1task,2tasks,3tasks,etc.Itshowsimprovementfromthe3 tothe10thtask, indicatingthevolunteersbecamebetteratthetaskoncetheybecamemorefamiliarwithit,butnot immediatelyso.Thesecondline,filteredworkers,excludesworkerswhoaveragedmorethan1 degreeoferror.Predictably,afterafewtasksfiltrationisaconsistentlymoreaccuratestrategy.The finalgraph,centroid,showsthemethodactuallyusedinthedeployment:thecentroidofthelocations identifiedbydifferentworkers(itistheaverageoveralltasks,asthenumberoftasksisnotas meaningful).Itclearlyshowsthatthiswasmoreaccuratethantakinganysingleworkerslocations, evenifthatworkerwassubstantiallyexperienced. Toourbestknowledge,nohumanitarianorganizationhaspreviouslyattemptedtoincreasethe accuracyofencodingbygivingrealtimeworktomultipleworkers(crowdsourcedvolunteersor otherwise)sowehopethatouranalysesofdifferentfiltering/aggregationtechniquescanpositively influencethedecisionsofhumanitarianorganizationsthatareconsideringpossibleinformation processingstrategies.WedidnotfilterworkersduringPakReport.Rather,wesimplytookthesuper setofallcategoriestheyselectedandtheweightedaverageacrosslocations.Weexploredthe potentialforfilteringworkersinposthocanalysisofthedata,simplybecausefilteringhigherror workersisstandardpracticeincommercialmicrotaskingplatforms.
CASE STUDY
Category: ICTs,ClimateChangeandDisasterManagement
PakreportalsomadeuseofUshahidiawelltriedandwelltrustedtechnologyplatformratherthan seekingtodeveloptechnologyfromscratch(somethingwhich,giventhedisastertimescales,wouldin anycasehavebeenimpossible). Relatedtothis,theprojectutilisedestablishedorganisationalexpertiseinanumberofways. BecauseUshahidiwasthefoundation,thenmembersfromUshahidiHaitiandChileimplementationsof crisisreportingandmappingplatformsalongwithSiliconValleyengineersjoinedthePakreportteam inthefirstweekafterdeployment.Theteamsexpertisewassubstantial,anditincludedlessons learnedfirsthandfromtheHaitiandChileimplementations.Similarly,partneringwithCrowdFlower allowedthedataanalysismicrotaskstoberapidlyandscalablyrolledout,ratherthannecessitateda neworganisationalinfrastructuretobecreated.Andlinkingupwithradiobroadcastersmadeuseof theirexpertiseandtheirexistingtechnologiestopubliciseuseoftheSMSmessagesystem. Finally,altruismcanbeseentohaveplayedarole.Ofcourse,someofthoseprovidingtheSMS inputsweremotivatedoutoftheirpersonalneedforrelieffromthisclimatelinkeddisaster.But othersweresimplyreportingwhattheysawforthebenefitofothers.Muchofthetechnical developmentandallofthedataanalysisworkwasundertakenbyvolunteersprovidingtheirtimein theserviceofthoseaffected.Andindividualsalsotookactiononthebasisofthereportstheysaw mapped,inordertoorganisetheirowndirectreliefactivities.
Constraints/Challenges
Theprojecthadaninformationalandtechnologicalnotdisasterresponsefocusatleastinthe initialstages.Thatis,themainworkthathadtobedonewasinsettingupthesoftware,web platform,datainputandanalysisandmappingprocessesinordertoproducethemapbased information.Butthiswasseparatefromthedisasterresponseeffortsothatcoordinationwiththe disasterreliefagencieswasatfirstquitelimited.Inlargepartthiswasduetothenatureof applicationdeploymentduringthedisasterratherthanpriortothedisasterwhentherecouldhave beentimetomakereliefagenciesawareoftheapplication,andtohelpensuretheinformation producedwasbeingusedtoguidefielddecisionsandactions. Amajorchallengeofanycrowdreportingprojectisverifyingtheaccuracyandauthenticityofthe datacomingfromtheground.Ingeneralthehopeisthatthevolumeofdatawillbesuchthatgood datadrivesoutbad,butthismaynotalwaysbethecase,especiallyascrowdsourcingofclimate relateddataincreases.Thereisnoauthoritativeansweryet(andthedangersareprobablygreater duringpoliticalcrisesratherthannaturaldisasters)butsomesuggestionsincludeweightingbasedon pastdatainputs,useofadditionaldatasuchaslocationalorphotographic,directquestioningofthe source,andthepossibilitiesforusinglanguageanalysissoftware(Meier2011). UsingmobilesandSMSenablesanyclimatechangerelatedprojecttoreachaverylargeuser population.However,therearestilldigitaldivideproblems.Forexamplebyrestrictinginputsto threelanguagesandbyrequiringatextbasedmessagetobesent,Pakreportrecognisesthatitdid excludesomemembersofthepopulation,suchasilliteratemobileowners.Useofthewebasthe mainreportingtoolsimilarlymeantthatonlyacertainsubsetofthepopulationwasabletoutilisethe results. Althoughrecognisingthegreatvalueofthecomplementaritiesbroughtbyhavingsomany stakeholders,thisalsobroughtthechallengesofnegotiatingagreementsandreaching consensus.Withtelecommunicationproviders,governmentdepartments,internationalagencies,the nationalmediaandothersallinvolved,thisaddedgreatlytothetransactioncostsofthePakreport initiativesomethingparticularlychallenginggiventheeverpressingtimescaleofdisasterrelief.This requiredagreatdealoftimeandeffortnotjustwhenformalagreementswereneededbutalsoin seekingtocreatelessformalconsensusforexamplearoundtheratherradicalnotionofcollecting andthendisseminatingpotentiallyunverifiedcitizenreports.
CASE STUDY
Category: ICTs,ClimateChangeandDisasterManagement
Recommendations/Lessons Learned
Thefollowinglessonswerelearnedfromthiscasestudy: 1)CrowdsourcingcanbehighlypowerfulandeffectiveforICTenabledclimatechange applications.Asseen,crowdsourcingcanbeusedinatleasttwodifferentwaysforgathering disaster/climatechangedatafromaverybroadsetofusersandlocationsandforthenanalysing thatdatasothatitcanbeeffectivelydisplayedandutilised.Althoughthisparticularapplication relatedtoaclimatelinkeddisaster,itisentirelyfeasibletousethesamemodelformonitoring climatechangee.g.byaskingpopulationstoprovideinformationonlocaldrought,rainfall, temperature,waterflow,etc.Thiswouldofitselfalsohelptoraiseawarenessaboutclimate changesomethingthatcouldbeenhancedwithafeedbackloopbywhichthoseprovidingdata wouldthemselvesalsobesentshortreportsandclimatealertsviaSMS,web,etc.Locationsat highriskofnaturaldisasterand/orclimatechangeeffectsshouldconsiderproactiveestablishment ofthistypeofsimplereportingworkflow. 2)MobileplusInternetplusserversequalsasystemwithreachandpower.Ontheirown, mobileandInternetandservertechnologieshavegreatvalue.However,thetechnologicalkeyto Pakreporthasbeenitscombinationofthethree.Mobileprovidedthereachdowntothe"bottomof thepyramid"populationswhoareonthefrontlineofdisastersandotherclimatechangerelated vulnerabilities.TheInternetprovidedthereachandpowertohelpcoordinatevolunteersacrossthe world,anddisseminateresultstoreliefagencies.Asthefoundation,serversprovidedthepowerto collect,analyse,storeanddisplaytheprocessedinformation.OtherICTbaseddisasterandclimate changeapplicationscanthereforeidentifyhowtocombinethereachandpowerofthese technologiesintoanoverallsystem. 3)Thefull"informationchain"mustbeinplace:theprovisionofinformationonclimatelinked disastersiscriticaltoeffectivedisasterresponseandbroadermanagement.Butthelatterareonly possibleifthereisafull"informationchain"(seeFigure5)thatisifthereisamechanismbywhich thatinformationisturnedintodecisionsaboutwhattodoandwhereandhowtodoitandthose decisionsarethenturnedintoactionsontheground.ICTsandclimatechangeprojectsmust thereforebedesignedaroundtheentirechain,typicallystartingthatchainbackwardsfromthe resultsthataresoughtandtheactionsnecessarytoachievethoseresults.
Figure5:TheInformationChain(Source:adaptedfromHeeks&Kanashiro2009)
CASE STUDY
Category: ICTs,ClimateChangeandDisasterManagement
References
Aon Benfield (2010) Pakistan Flood Event Recap Report, Aon Benfield, Chicago, IL. http://www.aon.com/attachments/reinsurance/201008_pakistan_flood.pdf Dell'Amore,C.(2010)Russiafires,Pakistanfloodslinked?,NationalGeographic,11Aug. http://news.nationalgeographic.com/news/2010/08/100812russianwildfirespakistanfloodsglobal warmingscienceenvironment Heeks,R.B.(2009)OverviewmodelofICTs,climatechangeanddevelopment,NotesonICTs, ClimateChangeandDevelopment,14Dechttp://niccd.wordpress.com/2009/12/14/overviewmodel ofictsclimatechangeanddevelopment/ IPCC(2007)WorkingGroupIIContributiontotheIntergovernmentalPanelonClimateChangeFourth AssessmentReport,IntergovernmentalPanelonClimateChange,Geneva. http://www.ipcc.ch/pdf/assessmentreport/ar4/wg2/ar4wg2spm.pdf Meier,P.(2011)Howtoverifysocialmediacontent,iRevolution,21June. http://irevolution.net/2011/06/21/informationforensics/ PlanningCommission(2010)TaskForceonClimateChange,PlanningCommission,Islamabad http://www.planningcommission.gov.pk/usefull%20links/Taskforces/TFCC%20Final%20Report.pdf PTA(2011)TelecomIndicators,PakistanTelecommunicationsAuthority,Islamabad. http://www.pta.gov.pk/index.php?Itemid=599