D3.1 Preliminary report on the skill assessment and comparison of state‐of-the-art methods for forecasts and projections of extremes

ThisprojecthasreceivedfundingfromtheEuropeanUnion’sHorizon2020researchand innovationprogrammeundergrantagreementNo101037293               DeliverableD3.1 Preliminaryreportontheskillassessmentandcomparisonof state‐of‐the‐artmethodsforforecastsandprojectionsof extremes  September2023   ThisprojecthasreceivedfundingfromtheEuropeanUnion’sHorizon2020researchand innovationprogrammeundergrantagreementNo101037293     InnovatingClimateservicesthroughIntegratingScientificandlocalKnowledge            DeliverableTitle:DL3.1Preliminaryreportontheskillassessmentandcomparisonofstate‐of‐ the‐artmethodsforforecastsandprojectionsofextremes Author(s):LluísPesquer(CREAF),IliasPechlivanidis(SMHI),YihengDu(SMHI),Rebecca Emerton(ECMWF),PaoloMazzoli(GECOsistema),GyörgyiBela(IDEAS),Ilyas Masih(IHE),SumiranRastogi(IHE),MarcvandenHomberg(RC510),Micha Werner(IHE). DateOctober2022 Suggestedcitation:PesquerL.,PechlivanidisI.,DuY.,etal.(2022)Preliminaryreportontheskill assessmentandcomparisonofstate‐of‐the‐artmethodsforforecastsand projections Availability:☒PU:Thisreportispublic[Pleaseselect] ☐CO:Confidential,onlyformembersoftheconsortium(includingthe CommissionServices)   DocumentRevisions: AuthorRevisionDate LluísPesquer,IliasPechlivanidisFirstdraftJune2022 LluísPesquer,IliasPechlivanidis,YihengDuetal.SeconddraftSeptember2022 MichaWerner,IlyasMasihReviewOctober2022 LluísPesquer,IliasPechlivanidis FinalversionOctober2022 LluísPesquer,YihengDu,PaoloMazzoli,Györgyi Bela VersionafterPOrevisionSeptember2023  D3.1‐Methodsforforecastsandprojections 1 ExecutiveSummary ClimateServices(CS)haveacrucialroleinempoweringcitizens,stakeholdersanddecision‐makersintaking climate‐smartdecisionsthatareresilienttoclimatechangeandcompatiblewithachievingclimateneutrality. Theresultssupportedbyascientificevidencebasecontributetowardsasustainableeconomy,lifestyle, environmentalprotectionandresourceuse.CSaimtotransformclimate‐relateddataandinformationinto customisedproducts,amongothersprojections,forecasts,information,trends,economicanalysisetc.,in ordertofurthersupportadaptation,mitigationanddisasterriskmanagement.Toachievethisadvanced scientificknowledge,monitoringandmodellingofclimatechangeandtheimpactsofclimateextremesare needed.AkeybarrierthatimpedesthecurrentgenerationofCSachievingthefullopportunityoftheirvalue‐ propositionrelatestothefailuretoincorporatethesocialandbehaviouralfactorsandthelocalknowledge andcustomsoftheirusers.Additionalchallengesarein:(i)theunderstandingofthemulti‐temporalandmulti‐ scalardimensionofclimate‐relatedimpactsandactions;(ii)thetranslationofCS‐provideddataintoactionable information;(iii)theconsiderationofreinforcingorbalancingfeedbackloopsassociatedtousers’decisions basedonCS;(iv)thelackoftransdisciplinaryapproachesacrossthefullCSvaluechain;and(v)needtodeliver tailor‐madeandrobustservicesatthescalerelevanttousers. TheI‐CISKprojectaimstoseizetheseuntakenopportunitiesbydevelopingnext‐generationCSthatfollowa socialandbehaviourallyinformedapproachforco‐producingCSthatmeettheclimateinformationneedsof citizens,decisionmakersandstakeholdersatthespatialandtemporalscalerelevanttothem.Intheseven geographicallydiverselivinglabs(LL),eachwithdifferentrelevantsectors,I‐CISKshowcasesitshuman‐centred co‐design,co‐creation,co‐implementation,andco‐evaluationapproachacrosskeysectorsvulnerableto climatechangeinEuropeandbeyond. Thisdocumentfocusesmainlyontheskillassessmentofstate‐of‐the‐artpredictionsofhydro‐meteorological variablesrelevantfortheI‐CISKLLs,andconsequentlysetsthebenchmarkforquantifyingtheaddedvalue fromotherscientificmethodsexploredwithinI‐CISK.TheskillassessmentisveryimportantfortheCS producedduringtheproject;theusefulnessoftheseCSandtrustfulnessoftheirusershighlydependsonthe reliabilityofthepredictionsonwhichtheseCSarebased. Thedocumentpresentsaninitialsetofdifferentmethodsbeingusedtogeneratehydro‐meteorological forecasts,predictionsandprojections,andfurtherlistsdifferentmethodologiesusedtoassesstheskilland robustnessofpredictions.Drivenbyinherentlimitationsinmodel‐basedseasonalmeteorologicalpredictions, biasesintherawseasonalprecipitationandtemperaturepredictionsoverEuropearehighlighted,andthe significantreductionofthesebiasesachievedafterpost‐processing(bias‐adjustment).Moreover,the documentpresentsananalysisofseasonalhydro‐meteorologicalpredictionskillincludingstreamflow extremes(floodsanddroughts)atthescaleoftheLLsintheproject.Finally,abriefreviewofthestate‐of‐the‐ artoftheintegrationoflocalandscientificknowledgeisprovided.Thisreviewisdevelopedfromtherich literatureinthefield,andexplorestheintegrationofknowledgesfromtheperspectiveoftheirintegrationin climateservices.Differentdimensionsoflocalknowledgethatarerelevantinthiscontextareexplored,anda typologyforlevelsofintegrationisintroduced.  Keywords ClimateServices;Userneeds;Seasonalpredictions;Biasadjustment;Extremeevents;Localdata;Local knowledge  D3.1‐Methodsforforecastsandprojections 2 AboutI‐CISK I‐CISK’sambitionistoinnovatehowclimateinformationisused,interpretedandactedonthroughanext‐ generationofClimateServicesthatfollowahumancentred,socialandbehaviourallyinformedapproach; integratingtheknowledge,needsandperceptionsofcitizens,decisionmakersandstakeholderswithclimate informationatspatialandtemporalscalerelevanttothem. ClimateServices(CS)arecrucialtoempoweringcitizens,stakeholdersanddecision‐makersintakingclimate‐ smartdecisionsthatareinformedbyasolidscientificevidencebase,thatcontributetowardsasustainable Europeaneconomy,lifestyle,environmentalprotectionandresourceuse,andthatareresilienttoclimate changeandcompatiblewithachievingclimateneutrality.Europeanandinternationalcollaborativeresearch efforts,includingCopernicusandGEOSShaveestablishedasolidscientificfoundationforaneffectiveCSvalue chain,includingadvancedscientificknowledge,monitoringandmodellingofclimatechangeandtheimpacts ofclimateextremes.However,severalbarrierschallengethecurrentgenerationofCSinachievingthefull opportunityoftheirvalue‐proposition.Thesechallengesincludethefailuretoincorporatethesocialand behaviouralfactorsandthelocalknowledgeandcustomsofclimateservicesusers.Additionally,the effectivenessofclimateservicesischallengedby;thestillpoorlydevelopedunderstandingofthemulti‐ temporalandmulti‐scalardimensionofclimate‐relatedimpactsandactions;thetranslationofCS‐provided dataintoactionableinformation;considerationofreinforcingorbalancingfeedbackloopsassociatedtousers’ decisions;andthelackoftrans‐disciplinaryapproachesacrossthefullCSvaluechain. I‐CISKaimstoseizetheseuntakenopportunitiesthroughahuman‐centredframeworkforco‐productionof nextgenerationCSthatspansthefullCSvaluechaintakingthedownstreampartofthevaluechainasastarting point.TheI‐CISKframeworkrealisesthefullpotentialofinformationprovidedthroughCSbyempowering actorstotaketheimpactsofextremeclimaticeventsandclimatechangeintoaccountintheirdecisions. Disclaimer Useofanyknowledge,informationordatacontainedinthisdocumentshallbeattheuser'ssolerisk.Neither theI‐CISKconsortiumnoranyofitsmembers,theirofficers,employeesoragentsshallbeliableorresponsible, innegligenceorotherwise,foranyloss,damageorexpensewhateversustainedbyanypersonasaresultof theuse,inanymannerorform,ofanyknowledge,informationordatacontainedinthisdocument,ordueto anyinaccuracy,omissionorerrorthereincontained. TheEuropeanCommissionshallnotinanywaybeliableorresponsiblefortheuseofanysuchknowledge, informationordata,oroftheconsequencesthereof. ThisdocumentdoesnotrepresenttheopinionoftheEuropeanUnion,andtheEuropeanUnionisnot responsibleforanyusethatmightbemadeofit. D3.1‐Methodsforforecastsandprojections 3 TableofContents  1Introduction.................................................................................................................................................1 1.1Purposeofthisdocument....................................................................................................................1 1.2Structureofthisdocument..................................................................................................................2 2State‐of‐the‐artmethodsandsystemsforforecasts,predictionsandprojections....................................3 2.1Descriptionofthefuturetimescales...................................................................................................3 2.2Exploringmedium‐rangeweatherforecasting....................................................................................4 2.3Exploringsub‐seasonalforecasting......................................................................................................5 2.4Exploringseasonalpredictions.............................................................................................................6 2.5Exploringdecadalpredictions..............................................................................................................7 2.6Exploringcentennialprojections..........................................................................................................8 3State‐of‐the‐artinintegratinglocalknowledgeandlocaldata..................................................................9 3.1Introduction.........................................................................................................................................9 3.2Whatconstituteslocalknowledgeanddata........................................................................................9 3.3Typesoflocalknowledgeandlocaldata...........................................................................................11 3.4Typologyofintegrationoflocalandscientificknowledgeinclimateservices..................................13 3.5Challengesanddirectionslocalandscientificknowledgeinclimateservices..................................15 4Advancinglarge‐scaleclimateservicesthroughintegrationoflocaldata................................................17 4.1IntroductiontoClimateServices........................................................................................................17 4.2DescriptionofEuropeanandglobalClimateServices.......................................................................17 4.2.1TheCopernicusservices................................................................................................................17 4.2.2GlobalEarthObservationSystemofSystems...............................................................................19 4.2.3TheWMOservices........................................................................................................................20 4.3Benchmarkingpredictions.................................................................................................................21 4.3.1Evaluationofpredictions..............................................................................................................21 4.3.2Benchmarksandreferencesystems.............................................................................................21 4.4UserrequirementsfromtheI‐CISKLLs..............................................................................................22 5Impactofpost‐processingonseasonalclimatepredictionerror..............................................................24 5.1Experimentalobjectives.....................................................................................................................24 5.2Dataavailability..................................................................................................................................24 5.3Post‐processingmethodology............................................................................................................24 5.4Assessingtheimpactofpost‐processingonpredictiveerror............................................................25 6Assessmentoffit‐for‐purposemethodologiesatselectedLivingLabs.....................................................28 6.1Experimentalobjectives.....................................................................................................................28 D3.1‐Methodsforforecastsandprojections 4 6.2Assessingtheseasonalhydro‐meteorologicalpredictionskill..........................................................29 6.2.1Hydrologicalmodelling.................................................................................................................29 6.2.2Evaluationframework...................................................................................................................29 6.2.3Assessmentofseasonalhydro‐meteorologicalpredictability......................................................30 6.2.4Assessmentofseasonalpredictabilityofstreamflowextremes..................................................32 6.3DroughtmodelsintheAndalucía(Spain)LivingLab..........................................................................34 6.3.1Backgroundandmethodology......................................................................................................34 6.3.2Precipitation..................................................................................................................................35 6.3.3Temperature.................................................................................................................................37 6.3.4Droughtindices.............................................................................................................................37 6.4UpperSecchiaRiver(Italy)LivingLab................................................................................................38 6.4.1Methodology.................................................................................................................................39 6.4.2Results...........................................................................................................................................40 6.5BudapestLivingLab............................................................................................................................42 6.5.1Background...................................................................................................................................42 6.5.2CalculationofhighresolutionLSTforurbanheatislandmapping...............................................43 6.5.3SatelliteLST...................................................................................................................................44 6.5.4Low‐altitudethermalinfraredremote(TIR)sensingdata(UASs)................................................44 6.5.5CNNnetworkforproducinghighresolutionLST..........................................................................44 6.5.6CitizenMeasurement:UrbanHeatIslandandThermalComfortAssessment.............................45 7Conclusionsandfuturework.....................................................................................................................47 7.1Conclusions........................................................................................................................................47 7.2MovingforwardwiththeLivingLabs.................................................................................................48 References.........................................................................................................................................................49     D3.1‐Methodsforforecastsandprojections 5 ListofFigures Figure1Timescalesrangesandtherelevanceofinitialvaluesandforcing.....................................................................4 Figure2TheseasonaloutlookavailableintheEFASserviceprovidedbyCEMS.............................................................18 Figure3TheGlobalDroughtMonitortoolofGDIS.........................................................................................................20 Figure4Biasesinrawandbias‐adjustedprecipitationpredictionsforthewinterandsummermonthsandforlead month0.ThepredictionsarebasedontheECMWFSEAS5predictionsystem.................................................................26 Figure5Biasesinrawandbias‐adjustedprecipitationandtemperaturepredictionsforthewinterandsummermonths andforleadmonth0.ThepredictionsarebasedontheCMCC‐SPS3.5predictionsystem...............................................27 Figure6LocationsoftheI‐CISKLivingLabs.....................................................................................................................28 Figure7Seasonalpredictionskill(intermsofCRPSS)fordifferenthydro‐meteorologicalvariablesdownscaledtothe scaleoftheLivingLabs.ThehydrologicalmodelsaredrivenbytheECMWFSEAS5predictions......................................31 Figure8Seasonalpredictionskill(intermsofCRPSS)fordifferenthydro‐meteorologicalvariablesdownscaledtothe scaleoftheLivingLabs.ThehydrologicalmodelsaredrivenbytheCMCC‐SPS3.5predictions........................................32 Figure9Seasonalpredictionskill(intermsofBSS)forstreamflowhigh(BSS90)andlow(BSS10)extremesdownscaled tothescaleoftheLivingLabs.............................................................................................................................................33 Figure10Ontheright,gaugeprecipitationstations(whitepointlocations)usedforthelocalmodelsintheAndalucia LivingLab(yellowboundaries),ontheleft,mapsituationofthisregion(bluepolygon)..................................................34 Figure11Flowchartforthegenerationofthefinerspatialvariabilitypatternmethod...................................................35 Figure12TheyellowlineofsyntheticNDVIshowsamorecoherentresponseofvegetationtotheannualprecipitation pattern(lineblue)thantheoriginalNDVI(ingreen).Thismodified(synthetic)productisneededinmeteorologicalstations locatedinurbanareas........................................................................................................................................................36 Figure13SPIatcoarse(0.25deg.)spatialresolutionbyEDO(upperfigure).Onbottom,SPIatfinerresolution(250m)by I‐CISK(lowerfigure)............................................................................................................................................................38 Figure14LocationofthestudyareainRER‐Italy,southofthePoRiver,intheupperprovincesofReggioEmiliaand Modena(administrativeboundariesanddotsrepresentingriverstagemonitoringstationsfromRegionalEnv.Agency networks)andtheweirattheuppercatchmentclosureofSecchiaRiver.........................................................................38 Figure15Exampleoflocaldatacollectedwiththehelpoftheusers:dischargestation(upperleft)andmeteorological stations(upperright)..........................................................................................................................................................39 Figure16Exampleofcollectedforecastandstatusmaps(multibandraster)fromlefttoright,CopernicusCDSdailyriver dischargeforecast@10kmspatialresolution,localARPAEdailyprecipitationmaps@5kmresolutionandforecasted COSMO2precipitationmaps@2kmresolution(alsoprovidedbyARPAE),withoverlayofrivercatchmentupstreamof theCastellaranoweirandtherivernetwork......................................................................................................................40 Figure17SketchofapossibleserviceGUI.A:Inputfeatures(e.g.rainfallandtemperature,groundstationsoraveraged, recordeddischarge,meteorologicalforecast);B:dischargeforecastforleadtimesofinterest;C:Timeseriesofvariables ofinterest;D:Lumpederrormetricsfortheselectedperiod;E:Configurationoptions;F:errorgraphsandgraphic indicatorsofforecastperformances(seealsonextfigure)................................................................................................41 Figure18Exampleofperformancesgraphs(ontheleftcorrelationamongobservedversuspredictedresiduals,onthe righterrorfrequencydistribution).....................................................................................................................................41 Figure19MapoftheElizabethdistrict..............................................................................................................................42 Figure20Schematicoftherelativehorizontalscalesandverticallayerstypicalofurbanareas:(a)Meso‐scaledome,(b) Meso‐scaleplume,(c)Local‐scale,and(d)Micro‐scale......................................................................................................43 Figure21CNNnetworkforhighresolutionLST.................................................................................................................45 Figure22Impressionoflocalresidenttakingameasurementwithadevicethatwasdevelopedtounderstandurbanheat islandsandheatwaves.Citizenscancontributetowell‐suitedadaptationmethodsandbetterpolicymeasures.The projectwillempowerresidentstomeasuretheirownthermalexposureandunderstandthethermalconditionsintheir livingenvironment.Coloursintheimageindicateheat,withreddercoloursindicatinghighersurfacetemperature.....45 Figure23ThermalimageofcityblocksrecordedinBudapeston7/1/22(leftimage).CitiZcansensorboxuserinterface (citizcan.com)(rightimage)................................................................................................................................................46   D3.1‐Methodsforforecastsandprojections 6 ListofTables Table1Dimensionsoflocalknowledgerelevanttoperceptionofclimaterelevantinformation.................................12 Table2Typologyofapproachestointegratinglocalandscientificknowledgeanddatainclimateservices................14 Table3Examplesoflocalknowledgeintegrationapproachesthathavebeendiscussedinliterature.........................14 Table4Propertiesoftheseasonalclimatepredictionsystems.....................................................................................24 Table5Comparedcontributionsofvegetationindicesintheregressionmodellingofprecipitation.‘out’meansthat theproductisrejectedbythemodelbecauseitscontributionisnotsignificant.Greencolouristhebestoneforamonth, lightbrownmeansthatthereisnotaclearresult,tiedresults..........................................................................................37 Table6Collectedlocaldataattheendoffirststageofanalysis....................................................................................41   D3.1‐Methodsforforecastsandprojections 1 1 Introduction 1.1 Purposeofthisdocument IntheDescriptionofWorkoftheI‐CISKproject,itisstatedthateffortswillbetargetedtowardsinnovation andenhancementofexistingclimateservices(CS)anddownstreamimpact‐basedproducts,andconsequently onthesupportofdecisionsandpoliciesinmultiplesectorsaccountingfortheirlocaltrade‐offs.InWork Package(WP)3,oneoftheaimsistoaddressthelocalneedsandsectoralgapsofexistingCSandtherefore variousstate‐of‐the‐artmethodswillbeusedtogetherwithtools/methodstointegratelocalstate‐of‐the‐art observationsandlocalknowledge.Bothcontinental/globalandlocal‐scaleprocess‐basedimpactmodels,e.g. forthewaterandagriculturesectors,willbeusedtoassesssub‐seasonal,seasonalandcentennialchanges andimpactsattheLivingLab(LL)scale.Therefore,acontinuousdialoguewithvariousWPs,e.g.WP1,WP2, WP3andWP4,hasbeenestablishedtoensureacontinuousexchangeandfeedbackofinformationrequired totranslatedatasetsintotailoredinformationandindicatorsforlocaluse. TheobjectivesofWP3areto:  Toadvancelocalimpactpredictionsandprojectionsofclimatechangeandfutureextremesthrough developingmodellingchainsthatefficientlyintegrateexistingCSwhilealsocombininglocaldataand knowledgeforlocaltailoring.  Toexploredifferentscientificstate‐of‐the‐artmethodstobridgedataandservicesthatarecurrently separatedontemporalandspatialscales(fromforecaststoprojections)andincreasethetrustinlocal predictions.  Toevaluatetheusefulnessoftheintegratedimpactpredictionsandassessmentsforlocaloperations anddecision‐makingfrombothascientificandauserperspective.  Tounlockthebenefitsoftransformationofdatatoinformationforandwithintheclimate‐sensitiveLL regionsandsectorsbyimprovingtheconfidenceinformationofindicatorswhileenhancingtheir usability.  Todevelopuser‐drivenvisualisationtoolsthatassurerobustandseamlesstransferofproduced informationfromCS,andcommunicatepredictions,explicitlyincludinguncertainty,forguided decision‐making.  Toproviderecommendationsforproductadaptations,extensionsandCSimprovements,anddeliver fitforpurposetools,methodsandproductsforuser‐tailoredreal‐timeoperationalservices Toachievepartoftheobjectiveslistedabove,thisdocumentpresentsthecurrentlyongoingworkandreports onthepreliminaryprogressinWP3,whileitaddressesaseriesofspecificobjectivesthatinclude:  Definingthedifferentfuturetimehorizonsandexemplifyingstate‐of‐the‐artmethodsusedto generatehydro‐meteorologicalforecasts,predictionsandprojections.  Listingthetraditionalmethodologiesusedtoassesstheskillandrobustnessofpredictions,including commonbenchmarkmethods.  QuantifyingthebiasesinseasonalmeteorologicalpredictionsovertheEuropeandomainand exploringthespatialbiasesafterpost‐processing(bias‐adjustment).  Benchmarkingtheseasonalhydro‐meteorologicalpredictiveskill,includingextremes,overthespatial scaleofthelivinglabs.  Reviewingdifferentapproachestotheintegrationoflocaldataandknowledgeatthescaleoftheliving labtoaddressthelocaluserneeds.  D3.1‐Methodsforforecastsandprojections 8  DecadalClimatePredictionSystem(DePreSys):Itisthedecadalpredictionsystemdevelopedandrun bytheMetOffice,HadleyCentre(UK).Theinitialisationisachievedbyrelaxingtofull‐depthanalyses ofoceantemperatureandsalinity,atmosphereanalysesofwinds,temperatureandsurfacepressure. Thisinitialisationimprovestheforecastskillofglobally‐averagedsurfacetemperatureinpasttest cases.  2.6 Exploringcentennialprojections Climateprojectionsaredependentonscenariosoffutureanthropogenicandnaturalforcing.Themain challengeofmid/longtermclimatepredictionscomparedtoweatherforecastsisthatthepredictionofsocio‐ economicdevelopmentisevenmoredifficultthanthepredictionoftheevolutionofthephysicalsystem. TheCoupledModelIntercomparisonProject(CMIP)isthemostrelevantworldwideprojectfortheclimate projectionsmodelling,notexclusivelytomid/longtermones,butmainly,italsocoversothertimescales(i.e. decadal).Since1995,CMIPhascoordinatedclimatemodelexperimentsinvolvingmultipleinternational modellingteams,currentlytheCMIP6istheactivephase. ThecurrentCMIPjoins21researchinstitutionsthateachdeveloptheirownmodels.Aselectionofthemare:  CESM(byNationalCenterforAtmosphericResearch,USA):Itisafully‐coupled,globalclimatemodel thatprovidescomputersimulationsonatmosphere(CAM),landCLM,ocean(POP,MOM6),ice(CSIM) oftheEarth'spast,present,andfutureclimatestates.(Danabasogluetal.2020).  EC‐Earth(byEurope‐wideconsortium,includingSMHI):ItintegratessomeEuropeanmodelsfrom differentinstitutions,includingECMWFmodels:IFSistheIntegratedForecastSystem,NEMOforthe ocean,LIMforsea‐ice,TM5fortheatmosphericchemistryandtransportandLPJ‐Guessforvegetation.  IPSL‐CM(byInstitutPierre‐SimonLaplace,France):ItcomprisesoftheLMDzmodelforthe atmosphere,INCAandREPROBUSforatmosphericcomposition,NEMOfortheocean,oceandynamics (NEMO‐OCE),sea‐ice(NEMO‐LIM)andoceanbiogeochemistry(NEMO‐PISCES),andtheORCHIDEE modelforterrestrialsurfaces.  HadGEM2:(byMetOffice,UK):TheHadGEM2familyincludesacoupledatmosphere‐ocean configuration,withorwithoutaverticalextensionintheatmospheretoincludeawell‐resolved stratosphere,andanEarth‐Systemconfigurationwhichincludesdynamicvegetation,oceanbiology andatmosphericchemistry.  Mk3L(byCSIRO,Australia):Itisacoupledgeneralcirculationmodelthatincorporatesaspectral atmosphericgeneralcirculationmodel,az‐coordinateoceangeneralcirculationmodel,adynamic‐ thermodynamicseaicemodelandalandsurfaceschemewithstaticvegetation.  MPI‐ESM(byMaxPlanckInstituteforMeteorology,Germany):Itcouplestheatmosphere(ECHAM6), ocean(MPIOM,JSBACH)andlandsurface(JSBACH)throughtheexchangeofenergy,momentum, waterandcarbondioxide.   D3.1‐Methodsforforecastsandprojections 9 3 State‐of‐the‐artinintegratinglocalknowledgeandlocaldata 3.1 Introduction Climateserviceshaveawell‐recognisedpotentialofempoweringdecisionmakersintakingclimatesmart decisions(GoddardandGoddard,2017);includingstakeholdersfromawiderangeofsectors,publicagencies andpolicybodies,andcitizens.Thispotentialis,however,inmanycasesnotfullyrealised,andtheuptakeof climateservicesmaybehamperedbyanumberofbarriers;includingthelackofunderstandingofuser’sneeds, layperceptions,localknowledge,andcapacitylevels(JacobsandStreet,2020),differencesbetweenthe spatialandtemporalscalesatwhichinformationisprovidedandthescalesrelevanttousers(Howardetal., 2020),difficultyofaccess,aswellaslackofsustainabilityofclimateservices(Vincentetal.,2020).Research shows,however,thattheusersclimateservicesmayserve,oftenhavewelldevelopedknowledgeofthe climatesystemsaroundthembasedontheirobservationandexperience(Tadesseetal.,2015;Plotzetal., 2017;Orloveetal.,2010),andthatrecognisingandintegratingtheseknowledgesthroughco‐creationof climateservicescanhelpclosetheusabilitygap(Plotzetal.,2017;Vincentetal.,2018;Cashetal.,2003), despitechallengestotheseknowledgesasaresultofdemographic,climaticandenvironmentalchanges(Plotz etal.,2017). Inthissectionweprovideabriefreviewofthecurrentstateoftheartintheintegrationoflocalknowledgein climateservices.Wedonotaimtoprovideafullreviewofthemultipledimensionsoflocalknowledge,rather theaimistoreviewofthecurrentstateoftheartfromtheperspectiveofhowlocalandscientificknowledge areintegratedinclimateservices.Theprocessofhowlocalknowledgeisincludedinestablishingclimate servicesinaco‐creationprocessdependsontheorganisationsinvolved(Cashetal.,2003),butthisprocessis beyondthescopeofthissection.Amorecompleteanalysisoftheintegrationoflocalknowledgeinthecontext oftheco‐creationframework,includingtheco‐identificationoflocalknowledgewillbedevelopedinWP2 (Task2.2)andrelateddeliverables. Wefirstbrieflyexplorewhatweconsideraslocalknowledge,bothwithinthescopeofthisreviewbutalsoto establishareferenceforofthedimensionsoflocalknowledgewithinthecontextofintegrationoflocaland scientificknowledgeintheI‐CISKproject.Wethenreviewhowlocalknowledgeisusedinclimateservices,and introduceabasictypologyofhowlocalknowledgeandscientificknowledgeareconsideredand/orintegrated withinclimateservices.Finally,weprovideareflectiononthechallengesanddirectionsoflocalandscientific knowledgeintegrationinclimateservices,andabriefoutlookonhowthesechallengeswillbeaddressedin theI‐CISKproject.  3.2 Whatconstituteslocalknowledgeanddata Todefinelocalknowledgeisitselfachallenge.Whatconstituteslocalknowledgeandwhotheownerofthat knowledgeis,andinparticulartheaspectsoflocalknowledgethatarerelevantwithinthecontextofaclimate service,ishighlydependentonthe(local)contextofthedecisionprocesstheclimateserviceintendstoserve, aswellasthestakeholdersanddecisionmakersthatareinvolved.TheUnitedNationsFoodandAgricultural Organisation(FAO)offersabroaddefinitionoflocalknowledge;describingitas“acollectionoffactsrelated totheentiresystemofconcepts,beliefs,andperceptionsthatpeopleholdabouttheworldaroundthem.This includesthewaypeopleobserveandmeasuretheirsurroundings,solveproblems,andvalidatenew information.Itincludestheprocesseswherebyknowledgeisgenerated,stored,appliedandtransmittedto others”(BuildingonGender,Agrobiodiversityandlocalknowledge).Withinthisframeworkitisimportantto notethatlocalknowledgeisnotonlyheldbytribalorindigenouscommunities,butalsobyothercommunities, includingthoseinruralandurbanenvironments,settledandnomadiccommunities,originalinhabitantsand migrants(ibid). D3.1‐Methodsforforecastsandprojections 10 Hermansetal.(2022)pointoutthatwhatdiscriminatesbetweenscientificandlocalknowledgeisthatwhile scientificknowledgeisdevelopedthroughaformalandagreedmethodology,localknowledgereflectsthe accumulatedknowledgeofthepeopleoftheenvironmentalcontextinwhichtheyliveandtowhichtheyare closelyrelated.Thislocalknowledgemaythenbeintrinsicallyusedinmakingpredictionsrelatedtohow environmentalconditions,includingtheweather,maydevelop,ortheconsequenceschangingandextreme theseenvironmentconditions(suchasdroughts,heatwaves,etc.)mayhave.Inthecontextofdisasterrisk reduction(DRR)andearlywarningsystems(EWS),therolelocalknowledgehasisdescribedastheexperience oflocalsurroundings,identificationandmonitoringofindicators,todetect,copeoradapttodisastersaswell ascommunicatedisasterrisk(Dekens,2007).Hadlosetal.(2022)characteriselocalandindigenousknowledge extractedfromfield‐basedstudieswithinDRRandEWStosixoverarchingforms;namely,earlywarning systems,riskknowledgeandperception,structuralmeasures,livelihood‐basedadaptation,socialcohesion andbeliefs.Muchoftheliteratureonlocalknowledgeconsiderstheknowledgeofthosecloselyrelatedtothe environment,suchasforexamplefarmerslivinginruralenvironments.Thereisnotmuchliteratureonhow localknowledgecanbeunderstoodinurbanareas.Inurbanareas,thereisamuchlargermobilityofpeople aswellaschangeintheenvironment(forexampleduetorapidurbanisation),whichwillinfluencehowpeople canbuildup(local)knowledgeoftherisksduetonaturalhazardssurroundingthem.Also,Hermansetal. (2022)foundthatmoststudiesintheirliteraturereviewonlocalknowledgeforearlywarningsystemsare geographicallyconcentratedintheGlobalSouth Scientificknowledgeincontrastisknowledgethatisestablishedthroughaformalandagreedmethodology, suchaslocalmeteorologicalorhydrologicalobservationsandindicatorsderivedfromthoseobservationsis thenreferredtoaslocaldata.Arguably,thisdistinctionbetweenscientificandlocalknowledgesbeingbased onhowformaltheprocessofknowledgegenerationis,issomewhatgrey.Someresearchersconsiderlocal indicatorsestablishedthroughtheuseofformallylocallycollectedobservationaldataaslocalknowledge (Reyes‐Garcíaetal.,2016),whilecitizenscienceisaprocessthatgeneratesandconsolidatesbothscientific andlocalknowledge(Wehnetal.,2021;Tengöetal.,2021). Localknowledgeisreferredtoinliteratureusingvariousnames,eithertohelpdistinguishwhotheownerof theknowledgeisand/ortoclarifytowhattheknowledgemaypertain.Theseincludeindigenousknowledge (Iloka,2016),traditionalknowledgeandlocaltraditionalknowledge(Plotzetal.,2017;Chisadzaetal.,2015) ortraditionalecologicalknowledge(Berkesetal.,2000)amongothers.Plotzetal.(2017)aswellasother authorsunderlinethatmostliteratureonlocalknowledgerelatedtoclimateservicesconsiderstraditionaland orindigenousknowledgediscussedwithinthecontextofdevelopmentresearch,whereformaldatasuchas fromhydro‐meteorologicalobservationsarescarce.However,werecognisethatlocalknowledgethatis relevantinthediversecontextsanddifferentsectorsintheLivingLabsoftheI‐CISKprojectmaybeknowledge thathasdevelopedovermanygenerations,orthathasevolvedinashortertime‐span,andthatbelongsboth toindigenousandnon‐indigenouspeople(ibid). Withintheclimateservicesthatareco‐createdintheI‐CISKlivinglabs,wealsoconsidertheroleoflocaldata collectedthroughaformaldatacollectionprocess.Thismayincludedatasuchaslocalmeteorologicaland/or hydrologicalobservationscollectedbyformalinstitutionswithinthegeographicalcontextofeachlivinglab, suchasthenationalhydrometeorologicalservices(NHMS),orotherpublicand/orprivatebodies.Also,local datacanbecollectedthroughmoreinformalprocesses,suchascitizenscienceorvolunteeredgeographic information.Thescientificknowledgetheselocaldataprovidemaybeusedtocomplementtheknowledge thatisobtainedfromlargerscaledatasetssuchasobtainedfromglobaland/orregionalclimatepredictions andprojections. Anotherwaytolookatlocaldatainrelationtolocalknowledgeisasfollows.Inordertobetterunderstand whatmakesuplocalknowledge,wecanpositionlocalknowledgeinthecommonlyusedDataInformation KnowledgeWisdom(DIKW)pyramid.Mulderetal.(2016)flippedthisconventionalview,wheredataarethe D3.1‐Methodsforforecastsandprojections 11 rawbuildingblocksofknowledge,arguinginsteadthatdataisgeneratedfromdifferentsourcesofknowledge. Wecandissectlocalknowledgeintoinformationanddataforeachcategoryordimensionoflocalknowledge. Forexample,localknowledgeontheecological(orfloraandfauna)dimensioncanbenarroweddownto informationaboutanimalbehaviour(aspecificfishspeciesasasignofupcomingfloods)(ŠakiTrogrlicetal., 2019).Insomecases,thisusuallyqualitativetypeofinformationcanbeconvertedintoapreferably quantifiableindicator,suchasquantifyingtheincreaseinnumberoffishes.Weemphasisethatbythisprocess of“datafying”thelocalknowledge,onenolongercapturesallthecontextualknowledgeandlivedexperience, andisreducingortransforminglocalknowledgetoscientificknowledgeandtodifferentsmallbuildingblocks.  3.3 Typesoflocalknowledgeandlocaldata Acknowledgementofthevaluethatlocalknowledgecanbringinimprovingthedesignanddeliveryofclimate serviceshasledtocredibleeffortstowardstappingitspotential,withagrowingbodyofliteraturethat capturesthevariousdimensionsoflocalknowledgeinthecontextofclimateservices.Tableprovidesan overviewofsomeofthesedimensionsacrossvariousstudies.Mostoftheapproacheswithinclimateservices literaturetendtofocusoncapturinglocalknowledgethroughindicatorswithinthemeteorologicalcategory (Bucherieetal.,2022;Streefkerketal.,2022).Thiscanbeattributedtoevidenceoftraditional,local knowledge‐basedforecastingmethodsaddingsignificantvalueinenhancingthespatialandtemporal resolutionofscientificforecasts(Masinde,2015)aswellasbeingusefulincommunicatingweatherandclimate informationtolocalcommunities(Tadesseetal.,2015).Literaturemakeslimitedattemptsatplacingaspecific timeframefortheoccurrenceoflocalknowledgeindicators.ŠakiTrogrlicetal.(2019),basedonresearchin Malawi,foundthatassigningatimeframeforaspecificindicatorisachallengingtask,asahighdegreeof disparityisobserved,evenwithinthesamevillage.Nevertheless,itispossibletolinktheobservationson differentlocalknowledgedimensionstoforexamplethedifferentphasesindisasterriskmanagement.For example,localknowledgeonearlywarningversuslocalknowledgeforearlyactionorresponseafterfloods havehappenedcanbemapped(ibid).Thereseemstobelessliteratureonlocalknowledgeinrelationto climatechangeadaptation,asitismorecomplicatedtocapturehowlocalknowledgechangesoverlonger timeperiodsandbecause,duetoclimatechange,possiblyalsosomeoftheextremeweathereventswillbe outsidethelivedexperience(Kelmanetal.,2012).Similarly,alsounderstandinghow“local”localknowledge is,isnotstraightforward.Theseambiguitiesanddifferencesinspatialandtemporalcoverageandresolution betweenlocalandscientificknowledgedirectlyinfluencehowtheintegrationbetweenlocalandscientific knowledgeanddatacantakeplace. Beyondclimateservices,scholarshipondisasterriskreductionandadaptationhasfocussedonother,broader dimensionsoflocalknowledge,including;riskperception,earlyactionstrategies,livelihooddiversification, roleofinstitutionsandsocialcapital,andbeliefsystemsofcommunities(ŠakiTrogrlicetal.,2019;Hadloset al.,2022).Thereisthereforeanopportunityforclimateservicestolearnfrombroaderliteratureandlookat localknowledgemoreholisticallyasawaytobetterunderstanduserneedsembeddedwithintheirlocal contexts.Thisalsomeansthatnearlyallphasesintheco‐creationframeworkofI‐CISKcanleveragelocal knowledgeandlocaldata.Forexample,intheco‐exploreclimateinformationneedsanddesiresphase,abetter understandingoflocalknowledgecanhelpinidentifyinggapsinexistingclimateservices.Intheco‐identify adaptationandDRRplanstosupportphase,thelocalknowledgeonexistinglivelihood‐basedadaptation strategiescanbeidentified.Intheco‐developclimatedataandknowledgephase,thelocalknowledgecanbe actuallyintegrated(whichisthefocusofthissection).Similarly,Tanetal.(2022)showedhowcitizenscience cancontributethroughknowledgetoallpartsofthewarningvaluechain. D3.1‐Methodsforforecastsandprojections 12 Table1Dimensionsoflocalknowledgerelevanttoperceptionofclimaterelevantinformation Dimensionsoflocal knowledge DescriptionReference(Examples) Meteorological conditions Anomaliesinambienttemperatureand/orwind directionandspeedpriortowetseasonrelatedto drought (Streefkerketal.,2022) Meteorological conditions Windspeed,temperatureandcloudformationsover LakeMalawiasprecursorstoflashfloodevents (Bucherieetal.,2022) FloraandFaunaDensityofleavesandfruitsandfloweringlevelsasan indicationofdroughtconditions (Chisadzaetal.,2015) FloraandFaunaChangestothebiophysicalsystemasanindicationof achangingclimateatthelocallevel (Reyes‐Garcíaetal., 2016) Climatechange perception Experiencedclimateknowledgetoimproveservice deliveryofclimateservices (Cliffordetal.,2020) LocalspatialknowledgeLocalknowledgeofspatialandtemporalpatternsof floods,droughtsandrainfall (Paulietal.,2021) RiskperceptionChangesintemperature,inter‐seasonalchangesin rainfallandrecurrenceofextremeevents (Singhetal.,2022) LocalimpactsAwarenessandjudgementoflocalimpactsofheavy rainfallfordisasterpreparedness (Sudmeier‐rieuxetal., 2012) Livelihood‐based adaptationStrategies Cropselection(ChenandCheng,2020) Changingplantingschedules(ŠakiTrogrlicetal., 2019) Improvingirrigationandwatermanagementsystems(LiragandEstrella, 2017)  Asdiscussedinsection1.3,wedistinguishlocalknowledgeandlocaldata.Whileinsomecaseslocaldatacan beseenasdirectlylinkedtolocalknowledge,inthiscontextwerefertolocaldataasdatathatislocally collectedthrougha(scientifically)formalprocess.Insomecases,moreinformalprocessesmayalsoresultin usefullocaldata.TherecentCitizenScienceGuidanceNotebytheWMO(2021)summarisestheinfluenceof citizens(assensors,interpreters,engagersandcollaborators)andscientists(instructing,collaborating,orco‐ creating)ondifferenttypesofcitizenscienceprojects.Whenthereisnoinfluenceatallfromthescientists involved,wecouldarguethatthisrepresentsthelocaldatathatispartoflocalknowledge.Thisisalsoinline withLeachandFairhead(2002),whoconsiderthatcitizenscienceimpliesacertainengagementwith,and usuallyamoredominantdiscursiverolefor,thescienceofexpertinstitutionsthanisthecasewithlocal knowledge.Understandingandcharacterisinglocalknowledge(andassociatedlocaldata)isusuallythrough qualitativetechniques,suchasfocusgroupdiscussionsandkeyinformantinterviews.Citizenscienceprojects typicallyusemorequantitativeandformalisedtechniques,thougharenotlimitedtothese(Hicksetal.,2019; CitizenscienceDRR,2022). D3.1‐Methodsforforecastsandprojections 13 InthecontextoftheI‐CISKproject,citizenscanrefertothoseincommunities,suchasinruralareasinLesotho, butalsotohotelownersinGreeceorolivefarmersinSpain.GivenI‐CISK’sco‐creationframework,thecitizens andscientistsintheLivingLabswilloperatemostlyinthecollaborationandco‐creatingsideofthespectrum forthetobecollectedlocaldata.Ofcourse,LivingLabactorsmayalsoalreadyhavelocaldata.Hereonecan thinkofe.g.commercialfarmsthathaverainfallrecordsoveralongperiodoftime(Landmanetal.,2020),but alsocommunitiesthathavelocalknowledgeonhydro‐meteorologicalindicators. 3.4 Typologyofintegrationoflocalandscientificknowledgeinclimateservices Theadvancinginterestinusingco‐creationprocessesfordeliveringusableclimateinformationhas consequentlyalsoledtogrowinginterestinexploringlocalknowledgeandwaystointegrateitwithscientific knowledgeinclimateservices.Severalauthorsarguetherelevanceoflocalknowledgeinthecontextof environmentaldecision‐making,disasterriskmanagementortoenhance(seasonal)climateforecasts(Jiriet al.,2016;Plotzetal.,2017;Streefkerketal.,2022).Theintegrationoflocalknowledgewithinclimateservices alsopresentstheopportunitytobettertailortheinformationtomatchtheend‐userneeds(Knivetonetal., 2014)andtoimprovecommunicationtolocalcommunities(Tadesseetal.,2015). Dryballetal.(2009)discuss,inthecontextofenvironmentalmanagement,processesofsociallearningamong actors.Theyexplainthatparticipationandinteractionamongdifferentactorscanrangefromcoercion(the willofonegroupisimposedontheother),informing,consulting,enticing,co‐creationtoco‐acting(active participation).Thisspectrumofinteractionbetweencommunitiesandexternalactorscanbeused,tosome extent,todescribethespectrumofhowtheholdersofLKandSKrelatetooneanother,reflectingpower relationsbetweentheactors.Wewillshortlygiveanexampleofthetwoextremes.Coercioncanbetheresult whenholdersofSKconsiderSKthemostvaluableknowledgesystemandwheretheyfocuson“extracting” thosepartsofLKthatcanbevalidatedscientificallyandusedtoforexamplelocalisescientificforecasts.Co‐ actingreflectsaprocessinwhichbothLKandSKco‐exist,eachhavetheirrespectivevalueandmutual learningsoccurbasedoncollaborationandnegotiation. Studieshavedefinedknowledgeintegrationinseveralways.Berggrenetal.(2011)definesitasacombination ofspecialisedknowledgetoreachanendresult,whileithasalsobeeninterpretedastheprocessof transformingindividualknowledgetoacollectiveone(OkhuysenandEisenhardt2002).Plotzetal.(2017) lookedatmethodsofintegratinglocalknowledge,comingupwiththetypologythatincludes:consensus buildingapproachesandscienceintegrationapproaches.Theformerisdescribedasaprocesswhereinthe finalproduct(inthiscaseaforecast)isanagreedoutcomeofnegotiationbetweenthelocalandthescientific knowledgeholders.Thenegotiationprocessmayfollowaverystructuredapproachoralessformalone, dependingonthecontext.Thisapproachalsoreliesonthelastmileactorsandlocalsocialnetworkstobuild acommonunderstanding(ibid).Thescienceintegrationprocess,ontheotherhand,employslocalknowledge forvalidation.Forexample,localindicatorsofweatherpatternswhichareextractedthroughsurveyswith stakeholdersareprocessedtovalidateindicatorsestablishedusingexistingscientificdatasets(ibid).Other approachesofintegrationincludeestablishingknowledgetimelinesandparticipatorydownscalingprocesses thataimtodrawonthesimilaritiesbetweenknowledgetypes,andexploringthelimitsofcurrentinformation (Knivetonetal.,2014).Wenotethatmostoftheapproachesforintegratinglocalandscientificknowledgeare whenproducingtheclimateservice,sopriortotheservicebecomingoperational,orwhenevaluatingclimate services,(Hironsetal.,2021)statethatevaluationshouldbeongoingandcombinemeteorologicalverification withdecision‐makersfeedback.However,integrationalsotakesplacewhenaclimateserviceisdelivered.For example,whenclimateserviceuserstriangulatethemselvesbetweentheirlocalknowledgeandthe knowledgeprovidedintheclimateservice.InTablewedevelopanextendedtypologyofintegration.This providesanoverviewofthewiderangeoflevelsofintegrationofLKandSKwithinclimateservicesasfound inrelevantliterature.Table3providesselectedexamplesofintegrationfromliterature,showingthatinmany oftheseseveralofthelevelsofintegrationmaybeconsidered. D3.1‐Methodsforforecastsandprojections 14 Table2Typologyofapproachestointegratinglocalandscientificknowledgeanddatainclimateservices LevelofintegrationDescriptionReference Science‐dominatedInthisapproach,theinformationprovidedthrough theclimateservicederivedfromscientificknowledge isconsideredthemostvaluable(describedas coercion).Thislevelofintegrationisoftenfoundin globalforecastingsystemsthataredevelopedusing globalscientificdatasetsandmodels. (Dryballetal.,2009) ConsensusIntheconsensusapproach,scientificknowledge(e.g. seasonalforecastsobtainedfromaclimatemodel) andlocalknowledge(e.g.seasonalforecastbasedon traditionalknowledgeofmeteorologicalsigns)are consideredequallybyscientificexpertsandtraditional knowledgeholders.Thetwoknowledgesare combinedtodevelopaconsensusforecast. (Plotzetal.,2017) ValidationLocalknowledgeisusedtoevaluateinformation providedbytheclimateservice,orscientific knowledgeisusedtoevaluatetheaccuracyoflocal knowledge‐basedforecasts.Referredtoasscience integrationinPlotzetal. (Landmanetal.,2020; Gillesetal.,2022) TriangulationScientificknowledgeprovidedthroughtheclimate serviceistriangulatedbyuserswiththeirlocal knowledgeoftheirenvironment.Thiscouldinclude comparisonof(seasonal)forecaststheclimateservice provideswithenvironmentalcuesobservedbythe user. (Shahetal.,2012)  (Gwenzietal2016)  InformingLocalknowledgeisusedtoinformhowscientific knowledgecanbeinterpreted.Examplesinclude where(Meteorological)indicatorsbasedonlocal knowledgeareusedtoinformhowscientificdatasets andmodelsareinterpreted. (Bucherieetal.,2022; Streefkerketal.,2022) ConditioningandBias Correction Notes:herelocalknowledgeandinparticularlocal dataisusedtoconditionmodeluncertaintiesand correctbiases.Thisisthroughformalmathematical approachessuchasquantilemappingorbayesian approaches.    Table3Examplesoflocalknowledgeintegrationapproachesthathavebeendiscussedinliterature TypeofintegrationDescriptionReference Modelinput(Forecast thresholdmodel); Informingand validation Meteorologicalindicatorsbasedonlocalknowledge onpredictingdryconditionsduringrainyseason (Streefkerketal.,2022) Validation(Statistical integration) Usingwind‐relatedindicatorsforforecastingrainand optimiselocalandmodernforecasts (Gbangouetal.2021) D3.1‐Methodsforforecastsandprojections 15 Participatory geographicinformation system(PGIS)or participatorymapping Understandingvulnerabilityandlocaladaptation actionsusingspatiallyexplicitmappingoflocal knowledge (Cruz‐Belloetal.,2018) Crowdsourcinglocal datatovalidatemodels orCommunity‐based observationnetworks Usecrowd‐sourcedfloodobservationsto quantitativelyassessmodelperformanceoffor examplefloodforecastingmodels (Dasguptaetal.,2022) (LeCozetal.,2016) (Alessaetal.,2016) Systemlevelintegration acrosstheearly warningvaluechain Participatoryapproachtoconnecttop‐downscientific knowledge‐basedsystemswithbottom‐up community‐basedsystemsandlocalknowledge. (Tarchianietal.,2020) Usesitespecificrecords toimproveforecastskill ofglobalorregional models Usesiterecordedfarmrainfallrecordsforthe developmentofskillfulforecastsystemsspecifictothe farm (Landmanetal.,2020) Participatory downscaling Theuseofknowledgetimelinesandparticipatory downscalingtovalidatemeteorologicalforecastsand buildtrustintheseforecastsamongfarmersinSenegal andKenya (Knivetonetal.,2014)  3.5 Challengesanddirectionslocalandscientificknowledgeinclimateservices Examplesfromrecentliteraturereviewedintheprecedingsectionsclearlyestablishtheneedforexploring localknowledgeanditsvalueinproducingclimateinformationthatismoresalienttouserneeds.However, localknowledgestillremainsunderutilisedwithindesign,deliveryandcommunicationofclimateservices, predominantlybecausethereisstillalackofaunifiedunderstandingofwhatconstituteslocalknowledge (Hadlosetal.,2022).Currently,withinclimateservicesliterature,localknowledgerelatedtometeorological indicatorsismorefrequentlydiscussedthanotherdimensions(Streefkerketal.,2022).Researchhas, however,revealedthatlocalknowledgeanditsusecanhavewidersocio‐economic,politicaland environmentaldimensions(ŠakiTrogrlicetal.,2019).Furthermore,clearerlinksneedtobedrawnbetween climateservicesandthelocalknowledgeembeddedwithinthelivelihoodpractices,copingandadaptation strategies,andsocialnetworkoflocalcommunities,sothattheinformationoradviceprovidedismoresuited toenduserneeds.Theongoingdiscussionaroundlocalknowledgealsodoesnotfullyconsiderknowledge heldatdifferentlevelsofgovernance(fromlivelihoodstolocalgovernmentbodiestohigherlevelsof government)aswellasalongtheclimateservicesvaluechainitself.Calveletal.(2020)studythisinthecontext ofdroughtwarning,exploringlocalknowledgerelatedwithcommunicationstructures(dissemination channels)anddecision‐making.Thereisalsoaneedtobetterunderstandtheroleoflocalknowledgewithin urbanenvironmentsandcommercialsectors(fore.g.,tourism).Currentrepresentationisskewedtowards agriculture,waterandnaturalresourcemanagement.Aliteraturereviewonclimateservicesforadaptation (Boonetal.,2022)foundthatthemajorityofinterventionseitherdidnotidentifyaclearsectoralfocusor mostlydiscussedtheaforementionedcategories.Thediversecontextsofthelivinglabsthathavebeen establishedintheI‐CISKproject,andtherangesectorsanddiversityoflocalknowledgeholdersineachofthe LivingLabprovidetheopportunitytoreducethisskew. D3.1‐Methodsforforecastsandprojections 16 Additionally,itisalsoimportanttolookatlocalknowledgeintermsofits‘relational’aspecti.e.,‘who’isbeing includedorexcludedintheproblemunderstandingprocess(Bouwen,2001).ŠakiTrogrlicetal.(2019)discuss theintergenerationalandgendereddifferencesinlocalknowledgeoflocalcommunitiesandtheneedtohave amoreholisticandrepresentativeviewoflocalknowledge.Thisconcernisaddressedwithintheco‐creation frameworkoftheICISKproject,andwillneedtobecarefullydocumentedwithinthecontextofeachofthe sevenLivingLabs. Whilethesearesomeofthegapsthathavebeenidentifiedwhenitcomestolocalknowledge,thereisaclear needtounpacktheseinamoresystematicmanner.Morespecifically,itisimportanttounderstandthevarious dimensionsoflocalknowledge.Thiscanshedfurtherlightonthequestionofwhatconstituteslocal knowledge,andcanhelptostartbuildingacommonunderstanding.Thiswillalsoserveasasteppingstone towardsidentifyingvariousentrypointsandpathwaysthroughwhichlocalknowledgecanhelpbuildamore salientandhuman‐centredclimateservices. Inthissection,specificpathwaysareidentifiedastohowlocalknowledgecanbeintegratedwithscientific knowledge.Beforebeingabletointegratelocalandscientificknowledge,onehastocharacteriseand understandlocalknowledge.Theoverviewofthedimensionsoflocalknowledge,aswellasthetypologyof levelsofinformation,illustratesthattheintegrationoflocalknowledgeandscientificknowledgeextendswell beyondthecombiningofquantitativedata,suchasinforexamplebiascorrectionofforecastsusinglocaldata. Breakinglocalknowledgedownintolocal(quantitative)data,hastheriskoflosingthemorecontextual qualitativeinformation,thoughthatmaybenecessaryforsomeintegrationapproaches.However,withinthe co‐creationprocessofdevelopingclimateservices,suchasinthelivinglabsinI‐CISK,theseadditional dimensionsoflocalknowledgeandtypesofintegrationshouldbeexplicitlyconsidered.  D3.1‐Methodsforforecastsandprojections 17 4 Advancinglarge‐scaleclimateservicesthroughintegrationoflocaldata 4.1 IntroductiontoClimateServices TheEuropeanCommission’sRoadmapforClimateServices(2015),setthedefinitionofclimateserviceswhich accountsforthosecovering"thetransformationofclimate‐relateddatatogetherwithotherrelevant informationintocustomisedproductssuchasprojections,forecasts,information,trends,economicanalysis, assessments(includingtechnologyassessment),counsellingonbestpracticesdevelopmentandevaluationof solutionsandanyotherservicesinrelationtoclimatethatmaybeusefulforthesocietyatlarge.Assuch,these servicesincludedata,informationandknowledgethatsupportadaptation,mitigationanddisasterrisk management(DRM)."Consequently,aclimateserviceneedstoprovidescience‐basedanduser‐specific informationrelatingtopast,presentandpotentialfutureclimate,assistingsocietyinadaptingtoclimate variabilityandchange. Typically,nationalserviceshavethemandatetoprovideforecastsandwarnings,andmanynationalandlocal organisationsproducetheirownforecastsandclimateservices.Inadditiontothese,arangeofglobaland continentalscaleforecastingsystemsandclimateservicesexisttosupporttheneedsatthelarge‐scale (continentalandglobalscale)andalsoaddressinter‐dependenciesbetweenregionsandevenoccasionally countries.Whilelocalservicesbenefitfromlocalknowledgeandexperience,andoftenbytheexistenceof high‐resolutionmodelsoverthesmalldomains,large‐scaleservicescanprovidecomplementaryinformation tosupportlocalCSandexistingcapabilities.Amongothers,large‐scaleservicesprovidedataandinformation attransboundarydomains,whilethesecanbefoundatlongleadtimesandeveninaprobabilisticapproach. Moreover,theycanprovideinformationwherenootherlocalpredictionsystemsareavailable(withouta massivescale‐upofresourcesforservicecustomization)andfororganisationsworkingatinternationalscales (suchashumanitarianorganisations)(Emertonetal.,2016). Specifically,fortheseasonaltimescale(whichisthetimescalethatthisreportismainlyfocusingon),anumber ofresearchandoperationalcentresprovidepredictionsofmeteorologicalvariablesattheglobalscaleandat atimerangefrom1to12monthsahead.Themostcommonpredictionsystemsare:ECMWFSEAS5,CMCC‐ SPS,MeteoFranceSystem7,GloSEA5fromtheUKMetOffice,NCEPCFSv2etc.AsmentionedinSection2, thesepredictionsystemsrelyoncoupledatmosphere‐ocean‐landGCMs;however,theirconfiguration(i.e. spatialresolution,ensemblemembers,initializationetc.)differsbetweenthemandhencetheforecastingskill variesasafunctionofvariableofinterest,geographicaldomain,leadtime,andaggregationperiod. Consequently,multi‐modelapproachesareexpectedtobebeneficialtounderstandbettertheuncertainty stemmingfromdifferencesinthemodelconfiguration.  4.2 DescriptionofEuropeanandglobalClimateServices Internationalorganisationshavebeencoordinatingeffortstoco‐createclimateservicesforlarge‐scale applications.Below,wesummarisetheseefforts,sincetheycanactasbenchmarkclimateservicestowhichI‐ CISKhuman‐centredCS,whichfocusonamorelocalscale,willaddvalue. 4.2.1 TheCopernicusservices CopernicusistheEuropeanUnion’sEarthObservationProgramme(www.copernicus.eu)andprovidesarange ofservicescoveringtheatmosphere,oceans,land,climatechange,securityandemergencyservices.Manyof theservicesareglobal,whileotherscoveronlytheEuropeandomain. TheCopernicusAtmosphereMonitoringService(CAMS)providesdataandinformationonatmospheric composition,forsectorssuchashealth,environmentalmonitoring,renewableenergy,meteorologyand D3.1‐Methodsforforecastsandprojections 24 5 Impactofpost‐processingonseasonalclimatepredictionerror 5.1 Experimentalobjectives Seasonalclimatepredictionslackthenecessarydownscalingandtailoring,andhenceeffortisstillongoingto improveservicepredictabilityandusability.Asmentionedinsection2,thepredictabilityofS2Spredictionsis subjecttomultiplesourcesoferroranduncertainty,whicharepresentinthevariouscomponentsofthe productionchaingoingfromclimatemodels(theirparameterization,initialization,bias‐adjustment,etc.)to theservicethatprovidesimpactindicators(impactmodelsetup,structureandparameterization).Inaddition, predictabilityischaracterisedbystrongspatialvariationandcommonlyatemporaldegradationofitsskillin longertimescales(seeFig.1). Here,theobjectiveistoexplorethebiasesthatarepresentintheseasonalclimatepredictions,andalsoapply apost‐processingmethod(alsoknownasbias‐adjustmentmethod)inordertoreducethesebiasesandresults towardsaclimatepredictionproductthatcanbeappliedforimpact(i.e.hydrology)assessmentatthelocal scale.Inaddition,hereweaimtoexploreandbetterunderstandthebiasesinspacefromdifferentseasonal predictionsystems.  5.2 Dataavailability Weassessedtheprecipitationandtemperaturepredictionswhicharedrivenbytwoclimateprediction systems;theECMWFSEAS5(Johnsonetal.,2019)andtheCMCCGlobalSeasonalEnsemblePredictionSystem version3.5(CMCC‐SPS3.5;Gualdietal.,2020).Bothsystemsgeneratetime‐seriesof6(CMCC‐SPS3.5)to7 (ECMWFSEAS5)monthsaheadwithamonthlyinitialization.Theparallelinvestigationofthetwosystems allowsdetectionofspatial‐temporalcomplementaritiesinthepredictions.Wenotethatinordertoquantify theimpactofclimatevariabilityonhydrologyatthelocalscale,bothseasonalpredictionsystemshavetobe downscaledandbias‐adjusted.Onlythencanthesepredictionsdrivethehydrologicalimpactmodel (Hundechaetal.,2016)toprovidelocalinformationofthehydrologicalconditions;seeresultsinSection5. ParticularlyforCMCC‐SPS3.5,theaccesstothepredictionswasduetoaninternalSMHI‐CMCCcollaboration. MoreinformationaboutthesystemsusedcanbefoundinTable1. Table4Propertiesoftheseasonalclimatepredictionsystems NameSourceHandledbyPost processing Ensemble members TimeframeResolution ECMWFSEAS5C3S‐CDS ECMWF‐CDSDBSmethod25 (hindcasts) 1993‐2015 (hindcasts) 0.33° CMCC‐SPS3.5CMCCSMHIDBSmethod40 (hindcasts) 1993‐20160.5°  5.3 Post‐processingmethodology Toadjustthereforecastdataforbiasesanddrifting,amodifiedversionoftheDistributionBasedScaling(DBS) methodwasconsideredandused(Yangetal.,2010).TheDBSforecastingmethodwasoriginallydevelopedto adjustclimateprojectionbiasesandhasbeenadaptedhereforseasonalpredictions.Inessence,theDBS methodimplementsaparametricquantile‐quantilemappinginwhichthetemperatureisconditionedon precipitationoccurrence.Here,notallmeteorologicalvariableswerebias‐adjustedbutinsteadonly precipitationandtemperatureusinganavailablereferencedatasetthatcoverstheentireEuropeandomain. Notethathereprecipitationandtemperaturewerebias‐adjustedseparatelyandnotfollowingamethodthat explicitlyconsiderstheinterdependencebetweenthetwovariables.However,giventhattheDBSbias‐ D3.1‐Methodsforforecastsandprojections 25 adjustmentisconductedforbothvariablestowardsobservations,itisexpectedthattheseparateadjustment stillrespectsthevariableinterdependencewhichispresentintheactualreferencedataset(HydroGFD). TheHydroGFDversion2.0(Bergetal.,2018)datasetwasusedasareferenceofbiasadjustmentforboth ECMWFSEAS5andCMCC‐SPS3.5meteorologicalforecasts(dailytemperatureanddailyprecipitation)forthe periodof1993‐2015.TheproductconsistsoftheERA5reanalysisproductcorrectedbyGPCCforprecipitation andtheCRUproductfortemperature,andhencethemonthlymeanwaterbalanceisconstrainedto observations.Theproductisavailableata0.5oresolutionandhencetheseasonalpredictionshadtobe convertedtothisresolutionpriortotheirbias‐adjustment.Usingthisreanalysisproduct,weovercomethe technicalchallengeofdailydatatypicallynotbeingavailableonalargescale(national,continental,global).  5.4 Assessingtheimpactofpost‐processingonpredictiveerror Herewepresentboththebiasesintherawandpost‐processedseasonalpredictionsfromthetwosystems (ECMWFSEAS5andCMCC‐SPS3.5),takingwintermonths(DJF)andsummermonths(JJA)asexamples.As expected,thebiasesintherawdatadonotfollowthesamepatternintermsofmagnitudeandspatial variability,whileaftertheDBS‐basedbias‐adjustment(BA),theyaresignificantlyreducedbothfor precipitationandtemperature.TheseasonalprecipitationpredictionsinFigures4(ECMWFSEAS5)and5 (CMCC‐SPS3.5)bothdisplaylargepositiveandnegativebiases,especiallyinregionswithcomplextopography andcoastalareas(e.g.,Spain,Franceandsouth‐easternEurope).Ingeneral,innorthernEurope,ECMWF SEAS5tendstooverpredictprecipitationinthewinter(rainandsnowfallseason)andunderestimate precipitationinthesummer.WhileinsouthernEurope,ECMWFSEAS5showsaclearoverestimationinthe summerprecipitation.InmostpartsofEurope,temperatureisunderpredictedbyECMWFSEAS5onaverage 1and2°Cinallmonthsandallleadmonths,exceptFebruary.CMCC‐SPS3.5showsasimilarbiaspatternin termsofspatialvariabilityonlywithslightlylargeramplitudes. WhiletheDBSmethodishighlyeffective,somebiasesstillremaininmeteorologicalpredictions(especiallyin precipitation,sincethebiasesfortemperaturearecloseto0),whichoriginatefromanassumptionofa theoreticaldistributionofdailydata.Henceitisimportanttonotethatinaproductionchain,suchremaining biaseswillbefurtherpropagated,potentiallyaffectingthequalityofhydrologicalpredictions.Nevertheless, theresultshereindicatethatstate‐of‐the‐artseasonalpredictionsystemsarestillsubjecttobiasesfor Europeanimpactassessments.Althoughthesesystemscanbeusedtoextractseasonalinformationof predictedanomalies,theirusabilityforactuallocalimpactassessmentsisquestionable,andhenceapost‐ processingisnecessary.Thepost‐processingappliedhereshowsthatthebiascorrecteddatasethasfewer remainingbiasesandcanbeconsideredasaninputtoimpactmodelsforlocalassessments. D3.1‐Methodsforforecastsandprojections 26  Figure4Biasesinrawandbias‐adjustedprecipitationpredictionsforthewinterandsummermonthsandforleadmonth 0.ThepredictionsarebasedontheECMWFSEAS5predictionsystem.  D3.1‐Methodsforforecastsandprojections 27   Figure5Biasesinrawandbias‐adjustedprecipitationandtemperaturepredictionsforthewinterandsummermonths andforleadmonth0.ThepredictionsarebasedontheCMCC‐SPS3.5predictionsystem. D3.1‐Methodsforforecastsandprojections 28 6 Assessmentoffit‐for‐purposemethodologiesatselectedLivingLabs 6.1 Experimentalobjectives ThemainobjectiveoftheI‐CISKprojectistodevelopnext‐generationCSthatfollowasocialandbehaviourally informedapproachforco‐producingCSthatmeettheclimateinformationneedsofcitizens,decisionmakers andstakeholdersatthespatialandtemporalscalerelevanttothem.I‐CISKshowcasesitshuman‐centredco‐ design,co‐creation,co‐implementation,andco‐evaluationapproachacrosskeysectorsvulnerabletoclimate changeinEuropeandbeyond.Thisisdoneinsevengeographicallyandsectorallydiverselivinglabs(seeFigure 6).Inthissection,wepresent:  theperformanceofseasonalhydro‐meteorologicalpredictionsconditionedtothescaleofeachLLs, includingthegeneralaccuracyandpredictabilityofextremes.Theseresultsarealsoconsideredas benchmarksduringthecontinuousscientificeffortsoftailoringthestate‐of‐the‐artclimateservices totheneedsoflocalusers.  localexamplesinthreeLLslocatedinthreedifferentclimaticregionsandhencesubjecttodifferent vulnerabilities.Theseare:theAndalucia(Spain)subjecttodroughts,theUpperSecchiaRiver(Italy) subjecttowateravailability,andtheBucharest(Hungary)focusingonurbanheat.  Figure6LocationsoftheI‐CISKLivingLabs. D3.1‐Methodsforforecastsandprojections 29 Finally,wenotethattheexamplespresentedinthissectionareatapreliminarystageandtheywillbe completed/complementedinDeliverable3.2“Skillassessmentandcomparisonofstate‐of‐the‐artmethods forforecastsandprojectionsofextremes”. 6.2 Assessingtheseasonalhydro‐meteorologicalpredictionskill 6.2.1 Hydrologicalmodelling TheHydrologicalPredictionsfortheEnvironment(HYPE)isasemi‐distributedprocess‐basedmodelcapable ofsimulatingthehydrologicalprocessesfromasinglebasintoglobalscale.Themodelhasconceptualroutines formostofthemajorlandsurfaceandsubsurfaceprocesses.Thesnowaccumulationandmeltprocessesare modelledusingthedegree‐daymethodwithlandusedependentparameters.HYPEsimulatesthewaterflow pathsinsoil,whichisdividedintothreelayerswithafluctuatinggroundwatertable.Afractionofrainfallor snowmeltinfiltratesintothetopsoil,whichislimitedbyasoiltypedependentmaximumrate.Ifthesoil moistureintheuppersoillayerexceedsathresholdformacroporeflow,partoftheremainingwaterforms macroporeflow.Potentialevaporation(PET)isestimatedusingthemodifiedJensen‐Haisemodel(Oudinet al.,2005),whilstPETisachievedonlyifeithertheactualsoilmoistureexceedsalargeportionofthesoilfield capacityorthesubbasinisdefinedasawaterbody.Forsoilmoisturebelowthislimitinnon‐waterbodyareas, theactualevaporation,computedusingthecropcoefficientmethodinAllenetal.(1998),decreaseslinearly tozeroatthewiltingpoint.Runofffromthesoilzoneiscomputedwhenthesoilmoistureexceedsfield capacityanditpercolatesfromuppertolowersoillayerswhenthesoilmoistureintheupperlayersexceeds fieldcapacity.Thegroundwaterlevelisestimatedbasedonthelevelinthesoilzonewheretheporespaceis filled. Here,weusetwosetupsoftheHYPEmodel;oneatthecontinentalscalecoveringtheentirepanEuropean region(Hundechaetal.,2016),andanotherattheworld‐widescalecoveringtheentireglobe(Arheimeretal., 2021).TheEuropeanmodelhasaspatialresolutionofabout35400sub‐basins,i.e.inaverage215km2andis referredtoasE‐HYPEv3.0.Theglobalmodelhasaspatialresolutionofmorethan130,000sub‐basins,i.e.in averageabout1000km2,andisreferredtoasWWH.Bothmodelsrunatadailytimestep.InI‐CISK,theLiving LabsthatlieintheEuropeandomainwereinvestigatedwiththeE‐HYPEmodel,whilethoseoutsidethe Europeandomain(GeorgiaandLesotho)wereinvestigatedwiththeWWHmodel. 6.2.2 Evaluationframework TheskillsofseasonalpredictionswereassessedbyContinuousRankProbabilityScore(CRPS;Hersbach,2000) andBrierScore(BS;Brier,1950)onbothhigh(90thpercentile)andlow(10thpercentile)streamflowextremes. Theskillsofthesetwoscores(CRPSSandBSSrespectively)wereachievedbyusingsimulatedclimatologyasa benchmark.TheskillsofseasonalpredictionsonfiveE‐HYPEoutputvariableswereconductedandanalysed asafunctionofleadweeksandforeachseason.Thehydro‐meteorologicalvariablesconsideredhereare: streamflow(COUT),temperature(CTMP),precipitation(CPRC),soilmoisture(SRFF)andevapotranspiration (EVAP).Moreover,theseasonalpredictionskillsonstreamflowextremesfromboththeE‐HYPEandWWH hydrologicalmodelswerefurtheranalysedusingtheBSS10(lowextreme;10thpercentileasathreshold)and BSS90(highextreme;90thpercentileasathreshold)metricsfordifferentleadweekswithinthelowandhigh streamflowperiods(definedbytheclimatologicalterciles;lowstreamflowperiodduring<33rdpercentile,and highstreamflowperiodduring>66thpercentile).Tofurtherextracttheinformationatthelocalscale,wefocus oneachLivingLabandgeneratetheskillscoresaccordingly. FortheLivingLabsinthepan‐Europeanregion(theNetherlands,Hungary,Italy,SpainandGreece),theE‐HYPE modelwithmeteorologicalforcingfrombothECMWFSEAS5andCMCC‐SPS3.5re‐forecastswereanalysed. FortheLivingLabsinGeorgiaandLesotho,theWWHmodelforcedwiththeECMWFSEAS5meteorologicalre‐ forecastswasused.Inallcasestheseasonalre‐forecastswerebias‐adjustedpriortobeintroducedinthe hydrologicalmodels. D3.1‐Methodsforforecastsandprojections 30 6.2.3 Assessmentofseasonalhydro‐meteorologicalpredictability Herewepresenttheresultsofpredictiveskillforanumberofhydro‐meteorologicalvariablesatthelivinglab scale.TheskillsofseasonalpredictionsforcedbyECMWFSEAS5wereaggregatedforeachseasonandassessed asafunctionofleadweek(seeFigure7).Ingeneral,theskilldeteriorateswithincreasingleadtime,butthe deteriorationratediffersdependingonthevariable,seasonandLivingLab.IntheSpanishLL(ES),the predictionsshowedhighpositiveskillforstreamflowespeciallyinthespringandsummermonths(MAMand JJA).Theskillstartedfromover0.8rightaftermodelinitializationandremainedabove0.5foraratherlong leadtime;20weeksinMAMand10weeksinJJA.Itisalsoworthynoticingthatinthewintermonths(DJF), thestreamflowpredictionsactuallyremainedskilfulfortheentiretimehorizon,withaCRPSSbeingover0.2 atthefurthestleadweek.Thepredictionsfortheothervariablesalsoshowedpositiveskillcomparedtothe simulatedclimatology(benchmark),onlywithlowerskillandfasterdeteriorationspeedcomparedto streamflow.SimilarpatternswerealsorevealedfortheskillintheGreekLL(GR).IntheLLsinHungary(HU), Italy(IT)andtheNetherlands(NL),thepredictionsforthefivehydro‐climaticvariableshadsimilarskilland deteriorationspeed.Ingeneral,higherskillwasachievedinthefirst4leadweeksinthesethreeLLs,ranging from0.3to0.8dependingontheseasonandvariable.IntheLLinGeorgia(GE)andLesotho(LS),the predictionsderivedfromtheWWHhydrologicalmodelforcedwiththeECMWFSEAS5seasonalpredictions wasassessedatamonthlyscale.Overall,higherskillwasfoundforthehydrologicalvariables,including streamflow,soilmoistureandevapotranspiration,whiletheskillforprecipitationandtemperaturesometimes reachednegativevalues,indicatingnoskillcomparedtoclimatology.Thiswasalsoobservedfortemperature intheLLsinGeorgiaandLesotho. D3.1‐Methodsforforecastsandprojections 31 Figure7Seasonalpredictionskill(intermsofCRPSS)fordifferenthydro‐meteorologicalvariablesdownscaledtothescale oftheLivingLabs.ThehydrologicalmodelsaredrivenbytheECMWFSEAS5predictions Theskillsofseasonalhydro‐meteorologicalpredictionsforcedwithCMCC‐SPS3.5werealsoaggregatedand assessedsimilarlytothoseforECMWFSEAS5(seeFigure8).Ingeneral,theskillhassimilarpatternsastheone forECMWFSEAS5,withadeteriorationpatternwithincreasedleadweeks.Nevertheless,onlysmall differencesbetweentheECMWFSEAS5andCMCC‐SPS3.5systemswereobservedbycomparingFigures5and D3.1‐Methodsforforecastsandprojections 32 6.Forexample,inthewintermonths(DJF)intheHungaryLL,theskillofallthehydro‐meteorologicalvariables fromCMCC‐SPS3.5wereslightlylowerthantheoneforECMWFSEAS5,especiallyforprecipitation(CPRC), temperature(CTMP)andevapotranspiration(EVAP).MeanwhileintheLLinItaly,theskillfromCMCC‐SPS3.5 showedaslowerdeteriorationspeedduringthefirst2leadweek(s)thanthatfromECMWFSEAS5.Those resultshighlightthepossibilityofimprovingtheseasonalpredictionskillfordifferentleadweeksbyoptimising (overevenaveraging/combining)thepredictionsystemsforeachLL.  Figure8Seasonalpredictionskill(intermsofCRPSS)fordifferenthydro‐meteorologicalvariablesdownscaledtothescale oftheLivingLabs.ThehydrologicalmodelsaredrivenbytheCMCC‐SPS3.5predictions. 6.2.4 Assessmentofseasonalpredictabilityofstreamflowextremes Thescientificliteraturehasrecognisedthatbothdroughtsandfloodshaveincreasedinfrequencyand magnitudeoverEurope,posingimmediatesocio‐economicthreats,whichcreatesaneedforhigh‐quality hydrologicalpredictionsonextremes.Thepredictionbeyondthemedium‐rangescaleaidsstrategicplanning forenergyproduction,agriculture,andotheractivitiesthatusuallyhappenonaseasonalscale.Hence, assessingthequalityofseasonalpredictionofstreamflowextremesisfundamentaltoI‐CISKsinceitalsosets thebenchmarkforthefuturework. D3.1‐Methodsforforecastsandprojections 33 Here,weassessthepredictionsintermsoftheirskillforthehydrologicalextremesandateachLL.TheBrier SkillScore(BSS;Brier,1950)wascalculatedforeachsub‐basinineachLL,foreachtargetweekandleadtime forbothlow(BSS10)andhigh(BSS90)streamflowextremes.Targetweeksaredefinedaslow‐streamflow/ high‐streamflowweeksforeachsub‐basinbasedonthetercilesderivedfromsimulatedclimatology.TheBSS forthetargetweeksisthenpooledandanalysedfordifferentleadweeks.Resultsofseasonalpredictionsfor streamflowextremesareshowninFigure9.Theskilloflow/highstreamflowextremesforthedifferentLLsis overallhigh(greaterthan0.6)forthemedium‐rangefuturehorizons(i.e.1–2weeksahead)forbothECMWF SEAS5andCMCC‐SPS3.5.However,asexpectedtheskilldeterioratesastheleadweeks’increase.Afaster deteriorationrateisobservedforthehighstreamflowextremescomparedwiththerateforthelow streamflowextremes.ThisisespeciallyobviousintheLLinSpain(ES)andGreece(GR),wherethepredictions oflowstreamflowextremesremainskilful(BSS>0)untilthefurthesttimehorizon.Basedonthisinvestigation, furtherinterpretationscanbemadebylinkingtheskillofstreamflowextremestothehydrologicalregimes (seealsoPechlivanidisetal.,2020).Forexample,intheLLsinItalyandtheNetherlands,whichareconsidered asriversystemswithsmallmemory(streamflowbeinghighlyresponsivetoprecipitation),afaster deteriorationofpredictionskilltakesplaceforbothhighandlowstreamflowextremes.However,intheLLsin SpainandGreece,wheretheareasarewithhighlyvariablestreamflowregimesandresponseissometimes drivenbysnowmeltingbesidesprecipitation,thepredictionshaveahigherandlongerskillforthelowthan thehighstreamflowextremes.  Figure9Seasonalpredictionskill(intermsofBSS)forstreamflowhigh(BSS90)andlow(BSS10)extremesdownscaledto thescaleoftheLivingLabs. D3.1‐Methodsforforecastsandprojections 40 Figure16Exampleofcollectedforecastandstatusmaps(multibandraster)fromlefttoright,CopernicusCDS dailyriverdischargeforecast@10kmspatialresolution,localARPAEdailyprecipitationmaps@5kmresolutionand forecastedCOSMO2precipitationmaps@2kmresolution(alsoprovidedbyARPAE),withoverlayofrivercatchment upstreamoftheCastellaranoweirandtherivernetwork. Bymeansofdevotedtoolstobedevelopedthismultibandrastershallbeaveragedoverthecatchmentarea toextractasingle“inputsignal”foreachvariable,storedinasimpledatabase(e.g..csvformat)togetherwith otherinputs(e.g.groundstationmeteorologicaldata)andtargetoutputvariables(e.g.timeseriesofrecorded discharge).Thisoperationincludesgapfillingtoaddmissingvaluesinthedatabase.Fortheenvisaged workflowdealingwithdifferentspatialresolutionisnotabigissueasthelumpedinformationoverthe catchmentisthevariabletoseek.ThisdatabaseshallbepassedtoselectedMLalgorithmstobetunedto retrievedesiredforecast,andresultsshallbestoredinthesamedatabaseasbefore. 6.4.2 Results Thefirstpartoftheworkflowhasresultedwithidentificationoflocalandupstreamknowledgeandsketchof theworkflowforthefollowingactivities.Despitebeingearlyatthedevelopmentstage,wehaveidentified targetachievementsandindicatorsofperformance,basingonuserexpectationsandpreviousexperiences frompastH2020projectsandliteraturepublications(Essenfelderetal.,2020;DeGregorioetal.,2018)tobe refinedalongthedevelopmentoftheCS. Concerningforecastskillspracticalapplicationsofdischargeforecastmainlyreliesonestablishederrormetrics indicatorsbothofforecastaccuracy(sucha%RootMeanSquareError,withexpectedtargetaround30to 50%intherangesofdischargeofmajorinterestforpracticalapplications)andpredictivepower(NashSutcliffe IndexandcorrelationcoefficientofthepredictedVsrecordedtimeserieswellabove0.5).Skillsshallbe evaluatedseparatingthedatasetin(past)valuesusedfortuningforecastalgorithms,andmostrecently(10to 20%)justforderivingerrormetrics. Furtherreflectionshavebeendonealsoonwaystodisplaytheproducedknowledge,withanearlymock‐up ofthepossibleGUIanditsmainfunctions(Figures17and18). D3.1‐Methodsforforecastsandprojections 41  Figure17SketchofapossibleserviceGUI.A:Inputfeatures(e.g.rainfallandtemperature,groundstationsor averaged,recordeddischarge,meteorologicalforecast);B:dischargeforecastforleadtimesofinterest;C:Timeseriesof variablesofinterest;D:Lumpederrormetricsfortheselectedperiod;E:Configurationoptions;F:errorgraphsand graphicindicatorsofforecastperformances(seealsonextfigure).  Figure18Exampleofperformancesgraphs(ontheleftcorrelationamongobservedversuspredictedresiduals, ontherighterrorfrequencydistribution). Thetablebelow(Table6)identifiesthelocaldatacontributionattheendofthefirststageoftheanalysis. Table6Collectedlocaldataattheendoffirststageofanalysis.  TypeofdataFormatLocationTime start Time end TimestepSourceResolution Riverdischargecsvstation 2021dailyARPAE Precipitationgribcatchment2001 hourly/dailyARPAE5km Maxtemperaturegribcatchment2001 hourly/dailyARPAE5km Mintemperaturegribcatchment2001 hourly/dailyARPAE5km Evapotranspirationgribcatchment2001 hourly/dailyARPAE5km Relativehumiditygribcatchment2001 hourly/dailyARPAE5km Windspeedgribcatchment2001 hourly/dailyARPAE5km Solarradiationgribcatchment2001 hourly/dailyARPAE5km Soilmapshpcatchment20012001 JRC1:10001 km D3.1‐Methodsforforecastsandprojections 42 Landuseclassmapshpcatchment20172017 RER1:10000 Digitalelevationmodelgeotifcatchment20152015 RER5m ShorttermforecastP/T COSMO gribcatchment2001 72hoursRER5km  6.5 BudapestLivingLab ThemaingoaloftheBudapestexperimentistoshowanotherpotentialCSinwhichthespatialresolutionof existingCSisfarfromtheuserdemands.ThisLLinanurbanenvironmentisacomplexsystemwherethelocal dataisabsolutelyneeded.Inthiscase,thecitizencontributionisveryrelevantandalsotheremotesensing productscanimprovetheskillassessmentofI‐CISKpredictions. 6.5.1 Background IntheLivingLablocatedinErzsébetváros(Elizabethdistrict),weinitiatedaparticipatoryresearchprocessto beabletounderstandtheurbanheatislandphenomenonmorefullyinthedistrictandtocapturethe perceptionsofcitizensontheheatstresstheyhavetoendureduringheatwaves.Forthis,we'veencouraged allcitizenstoparticipateintheresearchprocess.  Figure19MapoftheElizabethdistrict. UrbanresidentsoftheElizabethdistrictmaybeexposedtohigherheatloadsduringheatwavesthanthe populationofperi‐urbandistrictsingeneral,duetotheurbanheatisland(UHI)phenomenon,whichcauses highertemperaturesoverthisinner‐cityareathanoverthesurroundingruralareas.Thisproblemwillbe exacerbatedinthefuturebyglobalclimatechangeandurbanpopulationgrowth. Heatdomescanrelocateandaffectnearbylocationswithinaweekortwo.Becauseoftheweakbreezesand increasedhumiditytypicallycausedbythestationaryweatherpatternofheatdomes,theseeffectscanbe extremelyharmfultopeople.Theinabilityofthehumanbodytocoolitselfheatimpactsdomeevenharsher andmoredamaging. D3.1‐Methodsforforecastsandprojections 43 Inthecity,UHIcanbeinterpretedatthreelevels:meso,localandmicro.Thepresentstudyisprimarily concernedwithunderstandingtheclimaticprocessesatthelocalandmicroscales,andwithidentifyinghot andcoldspotsandheattraps.Tofacilitatetheadoptionofmitigationplansandtoquantifytheimpactsof urbanheatislands,itmaybeimportanttoprojectcurrentandfuturelandsurfacetemperatures(LST)and identifydistributionpatterns.  Figure20Schematicoftherelativehorizontalscalesandverticallayerstypicalofurbanareas:(a)Meso‐scale dome,(b)Meso‐scaleplume,(c)Local‐scale,and(d)Micro‐scale. Theresultsofourstudyareexpectedtohaveimplicationsforheatandclimatechangeplanningstrategiesfor theElizabethDistrict.Innovativeinterventionstomitigateurbanheatcouldbedevelopedthatarefully adaptiveandcollaborative.Mitigationstrategieswillneedtobeconsideredinlightofthesefindings. 6.5.2 CalculationofhighresolutionLSTforurbanheatislandmapping TheLandSurfaceTemperatureistheskintemperatureofground.Fromaclimateperspective,theaccurate understandingofLSThelpstoevaluatelandsurface–atmosphereexchangeprocessesandsurfaceenergy budgetsinmodels.Moreover,whencombinedwithotherpropertiessuchasalbedo,vegetationandsoil moisture,LSTprovidesavaluablemetricofthesurfacestate.LSTisdefinedbyGCOSasanessentialclimate variable(ECV). Thecalculationoflandsurfacetemperatures(LST)fromsatelliteimagesisessentialformanyfine‐scale applications(Zhouetal.2019).Itisalsousedinawidevarietyofapplicationssuchasmonitoringofglobal climatechange,studiesofdifferentkindsofsurfaces.However,theaccuracyofthecalculationisoftenlimited bydifferentenvironmentalandgeographicalconditions.WeperformacomputationalformofLSTsusingdata frommultiplesources.Weuseanumberofsourcestoobtainourdata,includinggroundmeasurements,UAV remotesensingandsatellitereadings.However,highresolutionLSThasbeenachallengingsubjectfor researchersforquitesometimebecauseseveralsources(missingpixelsetc.)cancauseuncertaintiesinthe calculation.Landsurfacetemperatureisakeyvariablewhenquantifyingtheimpactsofurbanheatislands. D3.1‐Methodsforforecastsandprojections 44 6.5.3 SatelliteLST LSTderivedfromsatellitethermalinfraredbands.Satellite‐basedsensorsareunabletorecordSTwithbotha highspatialandtemporalresolution.TheLSTdataisaffectedbytheincomingsolarradiation,whichaffects thesurfacetemperature.Thetemperaturecanalsochangeoverthecourseofadaybecauseofthewindand duetoaninversioneffect.Furthermore,theLSTvariationisgreatlyaffectedbylandcoverandtemperature inversion,twoofwhichcanaffecttheLSTdatagreatly. 6.5.4 Low‐altitudethermalinfraredremote(TIR)sensingdata(UASs) LSTmeasurementsfromsmallUASwillbeutilizedincombinationwiththermalimagingtechnologyand parallelimageprocessingalgorithmstoimproveandexpandonexistingmethodsforurbanheatislandstudies. SmallUnmannedAerialSystems(UASs)andtheminiaturizationofthermalcameratechnologyhaveenabled higherresolutionairbornethermalmapping.Morebroadly,theconceptalsoexploredthepotentialfor recalibratingdesignstrategiestopreventurbanheatislandeffecttomaximizeimpact. 6.5.5 CNNnetworkforproducinghighresolutionLST WewilldevelopaCNN(convolutionalneuralnetwork)thatwillprocessthesatelliteandUASbaselineimagery, whichwillproduceahighresolutionLSTmapofthedistrict.IfthisCNNnetworkislargeenough,itcandetect differentstructures,andtheAIdecideswhattheabsorptionmaybebasedonimagesegmentsratherthan pixels. ThisCNNnetworkcanberunasasoftwaremoduleandthuscanbeusedlater,eveninotherneighbourhoods, andwillbeabletopredicttheLSTofagivenneighbourhoodwithahighenoughdegreeofconfidence.  Pre‐processingphase:productionofthebasicsatelliteandUASLSTimages(upgradingofsatellite images,productionofLSTimagesfromthermalcameradroneimages)fortrainingandvalidation.  Trainingphase:(seeFigure20):usedronemeasurementLSTimages,satellitephotosandLSTimages toteachCNN.Dronemeasurementsareonlyavailableforafewdays;however,satelliteimagecapture isfrequent.  Processingphase:Usingnewlyacquiredsatelliteimage,highresolutionLSTcanbedetermined.Street measurementscanbeusedtocheckthevalidity.Furthermore,usingimagefusiontechniques,map visualizationscanbeproducedaccordingtotheneedsofthelocalpopulation.Imagefusionisa complexprocessinvolvingimageswithdifferentresolutionsanddifferentradiometricfeatures. However,therearedifferentalgorithmsthatcanbeusedtofusetheseimagesintoasingleproduct. Machinelearningmethodsrepresentapromisingsolutionforimagefusion.  Re‐trainingphase:ifnewUASdataareavailable,systemre‐trainingiscapabletogeneratecorrected highresolutionLST Duringthestreetsurvey(seenextsection),measurementsaretakenofheatradiatingindifferentdirections. Wemeasureatselectedlocationshowmuchoftheheatradiatedindifferentdirectionsisthesameasthe heatradiatedupwards(thermalradiationanisotropy). D3.1‐Methodsforforecastsandprojections 45  Figure21CNNnetworkforhighresolutionLST. 6.5.6 CitizenMeasurement:UrbanHeatIslandandThermalComfortAssessment Wearelaunchingacitizen‐sensingcampaignthatwillmeasureseveralcomponentsofurbanheatwavesand urbanheatislands(airtemperature,humidity,dust,andthermalheatpictures)inthedistrict.  Figure22Impressionoflocalresidenttakingameasurementwithadevicethatwasdevelopedtounderstand urbanheatislandsandheatwaves.Citizenscancontributetowell‐suitedadaptationmethodsandbetterpolicymeasures. Theprojectwillempowerresidentstomeasuretheirownthermalexposureandunderstandthethermalconditionsin theirlivingenvironment.Coloursintheimageindicateheat,withreddercoloursindicatinghighersurfacetemperature. Specifically,wearelookingathowheatisdistributedoncitystreetsthroughoutthewholeofElizabethDistrict. Ourmethodologyistoconductvisualfieldsurveysusingathermalimagingcamera(FLIR).Wecarryoutaseries ofstreet‐scalesurveyswithvolunteersintheElizabethdistrict.Theyareusingthermalimagingcamerasto takestreet‐leveltemperaturemeasurements.Theyarecomparingthethermalcharacteristicsofdifferent physicalelementsinthecitystreetsandtryingtofindouthowheatisgeneratedinthecity.Thistypeof researchisusefulforunderstandingthethermalenvironmentinthedistrict,whichaffectsthehealthandwell‐ beingofcitydwellers. Thermalimagingisatechniquethatallowsresidentsto"see"howhot(orcold)thingsarebymeasuringthe infraredradiationemittedbyobjects,differentphysicalelementsofurbanstreets(pavements,walls,grids, manholecovers,etc.).Theresultingdatacanrevealdetailsthatarenotvisibletothenakedeye.Thermal mappingcanhelpdesignersidentifyareasthatneedsidewalks,shadingorshadetrees. D3.1‐Methodsforforecastsandprojections 46 Figure23ThermalimageofcityblocksrecordedinBudapeston7/1/22(leftimage).CitiZcansensorboxuser interface(citizcan.com)(rightimage). D3.1‐Methodsforforecastsandprojections 47 7 Conclusionsandfuturework 7.1 Conclusions Inrecentyears,climateserviceshave(increasingly)receivedattentionbythescientific,developmentand decision‐makingcommunities,sincethedataandinformationprovidedcansupportadaptation,mitigation anddisasterriskmanagement.The(co‐)generatedproductsofsuchservicescoverdifferenttimehorizons fromhistoricaltopresentandtofuture,includingobservations,forecasts,predictionsandprojections.Despite recenteffortsforco‐creatingclimateservices,climateservicesdevelopmentprocedureshavenotputusersat thecentre,whichhaslimitedthepotentialforintegratinglocaldataandknowledgethataddvaluetolocal decision‐makingandactions.TheI‐CISKprojectisputtingeffortontheco‐creationofhuman‐centredclimate services,andthescientificworkconductedintheprojectaimstoexplorefit‐for‐purposemethodologies, tailoredtoaddresslocalneeds.Thisdocumentissettingthescenetotheavailablestate‐of‐the‐artClimate ServicesandpresentsthoseCSthataddresstheneedsofthewater‐andclimate‐relatedsectors.Thissummary ofthestate‐of‐the‐artClimateServicesfortheEuropeanandglobaldomainsiskeyfortheI‐CISKprojectthat aimstoaddvaluetotheproductsoftheseservicesfrombothascientificanduserperspective. Theanalysispresentedinthisreportbenchmarkstheseasonalpredictiveskillforeachofthesevenlivinglabs andexplorestheirlocaldataavailability.Wenotethatthisreportisapreliminary,withmorecompleteresults andinsightsplannedtobepresentedinDeliverable3.2.Thescientificworkpresentedinthisreportis conductedattwospatialscales:theEuropean‐widescaleandthelocalscaleofthelivinglabs.TheEuropean analysisisaimingtoacknowledgethebiasesintherawpredictionsfromtwodifferentseasonalclimatemodels (ECMWFSEAS5andCMCC‐SPS3.5)andtofurtherhighlighttheneedforpost‐processing(bias‐adjustment)in ordertoreducethebiasesandgenerateaproductthatcanbeusedforlocalimpactassessments. Weconcludethatthesebiasesarenotsimilarintermsofmagnitudewhileasexpectedthespatialvariability ofbiasesdiffersdependingontheseasonalclimatemodelused.However,evenwhenabias‐adjustmentpost‐ processingmethodisapplied,remainingbiasesstillexist,withtheirmagnitudedependingonthevariableof interest.Inparticular,weconcludethatremainingbiasesaremoreapparentforprecipitationthanfor temperature,whereremainingbiasesalmostreachzero. Despitetheremainingbiasesinthemeteorologicalforcing,analysisofseasonalhydro‐meteorological predictabilityatthelivinglabscaleshowsskillforthefirstleadtimes(upto2monthsahead).Asexpected, differentvariablesshowdifferentlevelofskill,i.e.precipitationislessskilfulthansoilmoistureorstreamflow, andthisconclusionholdsforpredictionswithbothseasonalclimatemodels.Moreover,dependingonthe hydro‐climaticpropertiesofthelivinglab,thehydro‐meteorologicalvariablesshoweddifferentprediction skill.Additionaltothepredictiveskillacrossthefulldistributionofflows,theanalysisalsofocusesonthehigh andlowstreamflowextremes.Resultsshowthatingenerallowstreamflowextremes(droughts)havehigher predictabilitythanhighstreamflowextremes(floods).Thisconclusioniswelllinkedtopreviousfindings indicatingthattherivermemoryisakeyfactorcontrollingtheseasonalhydrologicalpredictability. ThreecasestudiesaredevelopedfromaselectionofrepresentativeclimatevulnerabilitiesinthreeI‐CISKLL. Theselectedvulnerabilities:drought,wateravailabilityandheaturbanislandsareidentifiedasthemain demandsofthestakeholdersintheseLL,butthevariables,timescales,spatialresolutionandotherproperties aregoingtofittotheuserdemandsintheI‐CISKnextstages.Thethreepresentedstudiesshowtherelevant roleoflocaldataandknowledgeforunderstandingthelocalspatialpatternsofthevariabilityofthevariables involved(monthlyprecipitation,landsurfacetemperature,riverdischarge,etc…).Theknowledgeofthese D3.1‐Methodsforforecastsandprojections 48 patternsisgoingtobeusedfortheimprovementthecorrespondingforecasts,predictionsandprojectionsin thenextstepsoftheLLmodellingstudiesandthenext‐generationofCS.  7.2 MovingforwardwiththeLivingLabs Priortodefiningthestepsforfuturework,wecommunicatesomeoftheconceptualpillarsthataredriving thescientificsteps.  ThecurrentlyongoingI‐CISKresearchisconductedinclosecollaborationwiththeLLstakeholders withintheco‐creationprocess.  WestronglyarguethatI‐CISKclimateservicesneedtheintegrationoflocaldataandknowledgeto improvingtheboththerelevanceandthequalityoftheforecasts,predictionsandprojectionsatthe requestedtimehorizons(futureperiodsandaggregationwindows)andspatialresolution.  Theco‐createdI‐CISKclimateservicesareexpectedtocommunicateaccuratelyandeffectivelythe uncertainty(e.g.throughmapsand/orgraphs)inthelocalimpactindicatorsinordertobetterinform thedecision‐makers. Theplannedfutureworkwillbedevelopedintwomaindirections:  ToextendthepresentedpreliminarymethodsandmodelstoallI‐CISKLivingLabs:Inthispreliminary report,weshowresultsfromtwoseasonalpredictionsystemsatthepan‐Europeandomainandthree localstudiesspecificallyaddressedtotheircorrespondingLL:Andalucía(Spain),UpperSecchiaRiver (Italy)andBudapest(Hungary).Deliverable3.2willincludeexamplesforthesevenlivinglabs,andalso someexamplesofamergedanalysisbetweensomeofthem,forinstancethosefromthesamesector andclimatevulnerability(seeFigure4).  Tofitthemodelstothelivinglabrequirementsinordertoachievethemaximumusabilityinthe involvedsectors.InthecurrentstageoftheI‐CISKproject,wecollectedlivinglabuserneedsbasedon existingknowledge.Theongoingco‐designactivitiesineachofthelivinglabswillallowustorefinein moredetailtheserequirementsandtocontinueremainingmodellingeffortsatthelocalconditions forgeneratingI‐CISKuser‐tailoredclimateservices. Finally,methodologicallywewillcontinuetheeffortstonarrowthescalingandpredictabilitygap.Todoso, wewilltestvarioustechniquestoincreaseaccuracyandreliabilityatlocalconditionsanddecreasebiasand uncertaintyinclimateprojections.Thiswillbedoneby:  Post‐processing(includingdownscalingandbias‐adjustment)ensemblemeteorologicalpredictionsto theresolutionofimpactmodellingusinglocaldata.  Implementingdynamicsub‐samplingmethodsbasedonteleconnectionindicesinordertoimprove theseasonalhydrologicalpredictability.  Assessingthebenefitofamulti‐modelensembleapproachandaveragingmethods,particularlyatthe seasonaltimehorizons,giventhespatiotemporalcomplementaritythatwasobservedbetween ECMWFSEAS5andCMCC‐SPS3.5. D3.1‐Methodsforforecastsandprojections 49 References Alessa,L.,Kliskey,A.,Gamble,J.,Fidel,M.,Beaujean,G.,andGosz,J.:TheroleofIndigenousscienceand localknowledgeinintegratedobservingsystems :movingtowardadaptivecapacityindicesandearly warningsystems,Sustain.Sci.,91–102,https://doi.org/10.1007/s11625‐015‐0295‐7,2016. Arheimer,B.,Pimentel,R.,Isberg,K.,Crochemore,L.,Andersson,J.C.M.,Hasan,A.,Pineda,L.(2020)Global catchmentmodellingusingWorld‐WideHYPE(WWH),opendata,andstepwiseparameterestimation, Hydrol.EarthSyst.Sci.,24,535–559,https://doi.org/10.5194/hess‐24‐535‐2020 BazileR.,BoucherM.A.,PerreaultL.,LeconteR.:VerificationofECMWFSystem4forseasonalhydrological forecastinginanorthernclimate.HydrologyandEarthSystemSciences,21(11),5747–5762. https://doi.org/10.5194/hess‐21‐5747‐2017,2017 BerggrenC.,BengtssonL.,BergekA.,HobdayM.,Söderlund,J.(Eds.),KnowledgeIntegrationandInnovation: CriticalChallengesFacingTechnology‐BasedFirms,OxfordUniversityPress,Oxford(2011),pp.3‐28 Berkes,F.,Colding,J.,andFolke,C.:REDISCOVERYOFTRADITIONALECOLOGICALKNOWLEDGEASADAPTIVE MANAGEMENT,Ecol.Appl.,10,1251–1262,https://doi.org/https://doi.org/10.1890/1051‐ 0761(2000)010[1251:ROTEKA]2.0.CO;2,2000. BergP.,DonnellyC.,GustafssonD.:Near‐real‐timeadjustedreanalysisforcingdataforhydrology.Hydrology andEarthSystemSciences(22),989‐1000.https://doi.org/10.5194/hess‐22‐989‐2018,2018. Boon,E.,Judith,S.,Biesbroek,R.,Goosen,H.,andLudwig,F.:Successfulclimateservicesforadaptation : Whatweknow,don’tknowandneedtoknow,Clim.Serv.,27,100314, https://doi.org/10.1016/j.cliser.2022.100314,2022. Bouwen,R.:Developingrelationalpracticesforknowledgeintensiveorganizationalcontexts,CareerDev. Int.,6,361–369,https://doi.org/10.1108/EUM0000000006057,2001. Bucherie,A.,Werner,M.,VanDenHomberg,M.,andTembo,S.:Flashfloodwarningsincontext:Combining localknowledgeandlarge‐scalehydro‐meteorologicalpatterns,Nat.HazardsEarthSyst.Sci.,22,461–480, https://doi.org/10.5194/nhess‐22‐461‐2022,2022. BoerG.J.,SmithD.M.,CassouC.,Doblas‐ReyesF.,DanabasogluG.,KirtmanB.,KushnirY.,KimotoM.,Meehl G.A.,MsadekR.,MuellerW.A.,TaylorK.E.,ZwiersF.,RixenM.,Ruprich‐RobertY.,EadeR.(2016)The DecadalClimatePredictionProject(DCPP)contributiontoCMIP6,Geosci.ModelDev.,9,3751–3777 https://doi.org/10.5194/gmd‐9‐3751‐2016,2016 Calvel,A.,Werner,M.,Homberg,M.VanDen,Flamini,A.C.,Streefkerk,I.,Mittal,N.,Whitfield,S.,Vanya,C. L.,Rosa,E.,andArcher,M.:CommunicationStructuresandDecisionMakingCuesandCriteriatoSupport EffectiveDroughtWarninginCentralMalawi,2,https://doi.org/10.3389/fclim.2020.578327,2020. Cash,D.W.,Clark,W.C.,Alcock,F.,Dickson,N.M.,Eckley,N.,Guston,D.H.,Jäger,J.,andMitchell,R.B.: Knowledgesystemsforsustainabledevelopment,Proc.Natl.Acad.Sci.U.S.A.,100,8086–8091, https://doi.org/10.1073/pnas.1231332100,2003. Chen,T.andCheng,H.:InternationalJournalofDisasterRiskReductionApplyingtraditionalknowledgeto resilienceincoastalruralvillages,Int.J.DisasterRiskReduct.,47,101564, https://doi.org/10.1016/j.ijdrr.2020.101564,2020. D3.1‐Methodsforforecastsandprojections 56  D3.1‐Methodsforforecastsandprojections 1 Appendix1Glossary  AcronymDefinition ACAnomalyCorrelation ANNArtificialNeuralNetwork BABias‐Adjustment CIIClimateImpactIndicator CRPSContinuousRankedProbabilityScore CSClimateService CMIPCoupledModelIntercomparisonProject C3SCopernicusClimateChangeService DCPPDecadalClimatePredictionProject DRMDisasterRiskManagement DSTDecisionSupportTool DLDeepLearning ECVEssentialClimateVariable EDOEuropeanDroughtObservatory EFIExtremeForecastIndex ENSOElNiñoSouthernOscillation ESPEnsembleStreamflowPrediction GCMGlobalCirculationModel GEOSSGlobalEarthObservationSystemofSystems IQRInterquantileRange LLLivingLab LSTLandSurfaceTemperature MAEMeanAbsoluteError MJOMaddenJulianOscillation MLMachineLearning NDVINormalisedDifferenceVegetationIndex NDWINormalisedDifferenceWaterIndex NWPNumericalWeatherPrediction PETPotentialEvapotranspiration RPSRankedProbabilityScore RMSRootMeanSquare SCFSeasonalClimateForecast SEASSeasonalEnsemblePredictionSystem SPEI SPI S2S UHI WMO StandardisedPrecipitationEvapotranspirationIndex StandardisedPrecipitationIndex Sub‐seasonaltoSeasonal UrbanHeatIsland WorldMeteorologicalOrganization WPWorkPackage WWHWorld‐WideHYPE     ThisprojecthasreceivedfundingfromtheEuropeanUnion’sHorizon2020researchand innovationprogrammeundergrantagreementNo101037293                   Colophon : ThisreporthasbeenpreparedbytheH2020ResearchProject“InnovatingClimateservicesthroughIntegrating ScientificandlocalKnowledge(I‐CISK)”.ThisresearchprojectisapartoftheEuropeanUnion’sHorizon2020 FrameworkProgrammecall,“Buildingalow‐carbon,climateresilientfuture:Researchandinnovationin supportoftheEuropeanGreenDeal(H2020‐LC‐GD‐2020)”,andhasbeendevelopedinresponsetothecall topic“Developingend‐userproductsandservicesforallstakeholdersandcitizenssupporting climateadaptationandmitigation(LC‐GD‐9‐2‐2020)”.ThisprojecthasreceivedfundingfromtheEuropean Union’sHorizon2020researchandinnovationprogrammeundergrantagreementNo101037293. Thisfour‐yearprojectstartedNovember1st2021andiscoordinatedbyIHEDelftInstituteforWaterEducation. Foradditionalinformation,pleasecontact:MichaWerner(m.werner@un‐ihe.org)orvisittheprojectwebsite atwww.icisk.eu 