Multi-scaleIncrementalAnalysisUpdateSchemeandItsApplicationtoTyphoonMangkhut(2018)PredictionYanGAO1,JialiFENG1,XinXIA1,JianSUN2,YulongMA1,DongmeiCHEN1,andQilinWAN*11Guangdong-HongKong-MacaoGreaterBayAreaWeatherResearchCenterforMonitoringWarningandForecasting,Shenzhen518038,China2CMAEarthSystemModelingandPredictionCentre,ChinaMeteorologicalAdministration,Beijing100081,China(Received5December2021;revised5May2022;accepted11May2022)ABSTRACTInthetraditionalincrementalanalysisupdate(IAU)process,allanalysisincrementsaretreatedasconstantforcinginamodel’sprognosticequationsoveracertaintimewindow.Thisapproacheffectivelyreduceshigh-frequencyoscillationsintroducedbydataassimilation.However,asdifferentscalesofincrementshaveuniqueevolutionaryspeedsandlifehistoriesinanumericalmodel,thetraditionalIAUschemecannotfullymeettherequirementsofshort-termforecastingforthedampingofhigh-frequencynoiseandmayevencausesystematicdrifts.Therefore,amulti-scaleIAUschemeisproposedinthispaper.Analysisincrementsweredividedintodifferentscalepartsusingaspatialfilteringtechnique.Foreachscaleincrement,theoptimalrelaxationtimeintheIAUschemewasdeterminedbytheskilloftheforecastingresults.Finally,differentscalesofanalysisincrementswereaddedtothemodelintegrationduringtheiroptimalrelaxationtime.Themulti-scaleIAUschemecaneffectivelyreducethenoiseandfurtherimprovethebalancebetweenlarge-scaleandsmall-scaleincrementsinthemodelinitializationstage.Toevaluateitsperformance,severalnumericalexperimentswereconductedtosimulatethepathandintensityofTyphoonMangkhut(2018)andshowedthat:(1)themulti-scaleIAUschemehadanobviouseffectonnoisecontrolattheinitialstageofdataassimilation;(2)theoptimalrelaxationtimeforlarge-scaleandsmall-scaleincrementswasestimatedas6hand3h,respectively;(3)theforecastperformanceofthemulti-scaleIAUschemeinthepredictionofTyphoonMangkhut(2018)wasbetterthanthatofthetraditionalIAUscheme.Theresultsdemonstratethesuperiorityofthemulti-scaleIAUscheme.Keywords:multi-scaleincrementalanalysisupdates,optimalrelax...
