WRF-FDDA-Dudhia

WRFFour-DimensionalDataAssimilation(FDDA)JimyDudhiaFDDAMethodofnudgingmodeltowardsobservationsoranalysisMaybeusedforDynamicalinitialization(pre-forecastperiod)Creating4Dmeteorologicaldatasets(e.g.forairqualitymodel)Boundaryconditions(outerdomainnudgedtowardsanalysis)ARWonlyMethodModelisrunwithextranudgingtermsforhorizontalwinds,temperatureandwatervaporInanalysisnudging,thesetermsnudgepoint-by-pointtoa3dspace-andtime-interpolatedanalysisfieldInobs-nudging,pointsnearobservationsarenudgedbasedonmodelerroratobssiteThenudgingisarelaxationtermwithauser-definedtimescalearoundanhourormoreNudgingwillworkwithnestingandrestartsDynamicInitializationModeldomainsarenudgedtowardsanalysisinapre-forecastperiodof6-12hoursThishasbenefitofsmoothstartupatforecasttimezero-6h0hForecastperiodNudgingperiodBoundaryConditionsNudgeanouterdomaintowardsanalysisthroughforecastThishasbenefitofprovidingsmootherboundaryconditionstodomainofinterestthanif15kmdomainistheouterdomainwithinterpolated-analysisboundaryconditions45155Nudge45kmdomainonlyFDDAMethodsTwoMethodsGridoranalysisnudging(suitableforcoarseresolution)Observationorstationnudging(suitableforfine-scaleorasynopticobs)Nudgingcanbeappliedtowinds,temperature,andwatervaporNote:nudgingtermsarefakesources,soavoidFDDAuseindynamicsorbudgetstudiesAnalysisNudging(grid_fdda=1)Eachgrid-pointisnudgedtowardsavaluethatistime-interpolatedfromanalysesInWRFp*ismuAnalysisNudgingGisnudginginversetimescaleWisverticalweight(upperairandsurface)εisahorizontalweightforobsdensity(notimplementedyet)AnalysisNudging3danalysisnudgingusestheWRFinputfieldsatmultipletimesthatareputinwrffddafilebyprogramrealwhenrunwithgrid_fdda=1Withlowtime-resolutionanalyses,itisrecommendednottouse3dgrid-nudgingintheboundarylayer,especiallyfortemperatureSurface(2d)analysisnudgingnotavailableyet(maybelaterafterVersion3)Analysis-NudgingnamelistoptionsCanchooseFrequencyofnudgingcalculations(fgdtinminutes)Nudgingtimescaleforeachvariable(guv,gt,gqininverseseconds)WhichvariablesnottonudgeinthePBL(if_no_pbl_nudging_uv,etc.)Modellevelforeachvariablebelowwhichnudgingisturnedoff(if_zfac_uv,k_zfac_uv,etc.)Rampingperiodoverwhichnudgingisturnedoffgradually(if_ramping,dt_ramp_min)ObsNudging(obs_nudge_opt=1)EachgridpointisnudgedusingaweightedaverageofdifferencesfromobservationswithinaradiusofinfluenceandtimewindowObsNudgingGridpointobservationObsNudging•Risradiusofinfluence•DisdistancefromobmodifiedbyelevationdifferenceObsNudging•τisthespecifiedtimewindowfortheobs•ThisisafunctionthatrampsupanddownObsNudgingwσistheverticalweighting–usuallytheverticalinfluenceissetsmall(0.005sigma)sothatdataisonlyassimilatedonitsownsigmalevelobsinputfileisaspecialasciifilewithobssortedinchronologicalordereachrecordistheobs(u,v,T,Q)atagivenmodelpositionandtimeObs-NudgingnamelistoptionsCanchooseFrequencyofnudgingcalculations(iobs_ionf)Nudgingtimescaleforeachvariable(obs_coef_wind,etc.)Horizontalandverticalradiusofinfluence(obs_rinxy,obs_rinsig)Timewindow(obs_twindo)Rampingperiodoverwhichnudgingisturnedoffgradually(obs_idynin,obs_dtramp)FDDASummary•FDDAgridnudgingissuitableforcoarsergridsizeswhereanalysiscanbebetterthanmodel-producedfields•Obsnudgingcanbeusedtoassimilateasynopticorhigh-frequencyobservations•Gridandobsnudgingcanbecombined•FDDAhasfakesourcesandsinksandsoshouldnotbeusedonthedomainofinterestandinthetimeperiodofinterestforscientificstudiesandsimulationsFurtherplansAdd2d(surface)nudgingandintegratewith3dnudgingIntegratewithanalysesproducedbyWRF-VarEnd