ECMWF模式初值场的确定

ECMWF模式初值场的确定确定性数值预报概要（一）Let’shopethatnobodywillfallasleep!AHECMWF预报技巧演变341.提高水平与垂直分辨率2.减少近似与简化来提高动力方程的精度3.采用更准确的物理过程来表达小尺度过程，如辐射、云、降水与热量、水汽与动量的湍流交换4.采用更精确的分析与四维资料同化形成模式的初值5.增加观测资料科学家们的努力引自陈静，《业务数值天气预报》，201656提纲资料同化的概念ECMWF同化系统介绍我国资料同化现状和展望78主流业务数值预报系统输入资料资料同化数值模式模式后处理产品释用产品发布3DVar/4DVar/EnKFGriddedDataTerrestrialDataObservationData动力框架(全方程组，静力/非静力)物理过程（云物理，边界层，陆面）(引自孙继松，数值预报产品应用)9数值模式新的初值OperationalconfigurationsoftheECMWFIntegratedForecastingSystem(IFS)(witherrors)Observations(witherrors)Computer(withalotofCPUs)People(withalotofgoodideas)Analysis(with-smaller–errors)12aTObservedtemperature(To):8˚CBackgroundforecasttemperature(Tb):10˚degrees.Analysis(Ta):x˚CaToTbTaTObservationsandmodelbackgroundhaveerrorsItisimportanttospecifythemaccurately背景场观测分析场背景场和观测的权重与它们的“精确度”（误差方差的倒数）成正比131415T1T10000000011()()()((()))((()))22NbbiiiiiiiiJHMHMxx-xBx-xyxRyx分析场是通过求解目标函数极小值得到的：用模式从分析时刻t0预报到观测时刻ti0iM四维变分（4DVar）16x：分析场（新的模式初值）bx：背景场（短期预报结果）y：观测（如探空，地面，辐射率，雷达径向风/反射率等）B：背景误差协方差R：观测误差协方差：观测算子（将模式变量转换为观测变量，“模拟观测”）Hbo12-h同化窗17oJObservedsatelliteradianceModelradianceHcompareMeteosatimagery–watervapourchannelModelT,u,v,q,o3卫星资料的观测算子18EUROPEANCENTREFORMEDIUM-RANGEWEATHERFORECASTSAnothermodelversusobservationexampleCLOUDSATisaNASASatellitewithacloudprofilingradaron-boardModelObservationsCloudRadarReflectivity雷达径向风的观测算子20把模式空间风矢量(u,v,w)场投影到雷达坐标平面，模拟雷达径向风000()()()()TrxxuyyvzzwvVrrr222000rxxyyzz(,,)xyz：模式格点与雷达站点000(,,)xyz的距离Tv：降水粒子的下落末速度：观测算子（模式空间到观测空间的投影）H(SunandCrook1997,1998)(WRF3DVAR)Theobservationoperatorprovidesthelinkbetweenthemodelvariablesandtheobservations(Lorenc1986;Pailleux1990).Theobservationoperatoristypicallyimplementedasasequenceofoperatorstransformingtheanalysiscontrolvariablexintotheequivalentsofeachobservedquantityy,atobservationlocations.Thissequenceofoperatorscanbemulti-variate(candependonmanyvariables)andmayinclude:“Interpolation”fromforecasttimetoobservationtime(in4D-Varthisisactuallyrunningtheforecastmodelovertheassimilationwindow)HorizontalandverticalinterpolationsVerticalintegrationIflimb-geometry,alsohorizontalintegrationIfradiances,radiativetransfercomputationAnyothertransformationtogofrommodelspacetoobservationspace.ActualimplementationofobservationoperatorsusedintheECMWF4DVariationalDataAssimilationSystem求解目标函数极小值是迭代极小化过程costfunctionJJ(xb)JminimmD1mmmmxxD222324252627282930T11()()1NbbbbbiiiiiNPxxxxB：背景场扰动←集合预报bix用集合预报成员统计背景误差协方差31集合预报统计出来的背景误差协方差表征了随时间演变的大气流型3233EDA（EnsembleofDataAssimilations）4DVARForecastX25b(tk)Y+ε25oBoundarypert.25X25a(tk)X25b(tk+1)……ε25m343536欧洲中期天气预报中心（ECMWF）的数值预报产品是我国气象业务部门的重要参考•资料同化技术发展的前瞻性•卫星资料同化-使用更多卫星仪器类型和卫星资料3DVar(1996)4DVar(1997)EDA(2011)“全天候”微波同化技术长时间窗-弱约束4DVar增加使用受云降水和气溶胶影响的红外、微波卫星观测优化资料挑选、质量控制和观测误差的定义优化先进垂直探测仪资料的使用37ConventionalobservationsusedbyECMWF’sanalysisDRIBU:MSLPressure,Wind-10mPILOT/Profilers:WindMSLPressure,10m-wind,2m-Rel.Hum.Wind,Temperature,Spec.HumidityAircraft:Wind,TemperatureSYNOP/METAR/SHIP:Radiosondeballoons(TEMP):Note:Dataassimilationonlyusealimitednumberoftheobservedvariables-especiallyoverland.38SatellitedatasourcesusedbyECMWF’sanalysisImagers:SSMI,SSMIS,AMSR-E,TMIOzoneGPSradiooccultationsSounders:NOAAAMSU-A/B,HIRS,AIRS,IASI,MHSGeostationary+MODIS:IRandAMVScatterometeroceanlow-levelwinds:ASCAT39EUROPEANCENTREFORMEDIUM-RANGEWEATHERFORECASTSWhattypesofsatellitesareusedinNWP?AdvantagesDisadvantagesGEO-RegionalcoverageNoglobalcoveragebysinglesatellite-TemporalcoverageLEO-Globalcoveragewithsinglesatellite40Metop41RoleofobservationsRMSerror(m)Time(hours)SEVIRI6.2µmEvery12hoursweassimilate~20,000,000observationstocorrectthemodel’svariables….Themodelhasmanymorevariablesthanwehaveobservations.Observationslimiterrorgrowthandmakeforecastingpossible….42FSO(ForecastSensitivitytoObservations)ECMWFSystem(June2011)0510152025SYNOPAIREPDRIBUTEMPDROPPILOTGOES-AMVMeteosat-AMVMODIS-AMVSCATHIRSAMSU-AAIRSIASIGPS-ROAMSR-ESSMISTMI-1MERISMHSAMSU-BMeteosat-RadMTSAT-RadGOES-RadO3FEC%AMSU-AGNSS-ROIASIAIRSRemark:GNSS-ROcontributes~2-3%ofobs.assimilated(引自顾建峰，气象业务现代化与信息化)43TOVSERA-Interimskillshowslittlechange1980-2000whentherewaslittlechangeintheGOS.TheATOVSerashowsagainofaround6-18hoursinpredictableskillcomparedtotheTOVSera.ERA-Interimexcludesmorerecentdatae.g.IASIandASCATsotheimpactofnewerdataishardtojudge.ThiscomparestooverallECMWFgaininskillofabout2-3hoursperyear.ATOVSATOVS+HyperspectralIRFulloperationalsystem.GOS+forecastchangeswithtimeRe-analysissystem.OnlyGOSchangeswithtime数值预报系统资料类型观测来源GRAPES-GFS2009-2015地面、船舶、浮标GTS常规观测探空、飞机报GTS常规观测AMV云导风卫星--搭载探测仪器：METEOSAT–红外GEOS–红外、水汽TERRA–红外、水汽FY-2E--红外、水汽ATOVS微波辐射率卫星--搭载探测仪器：NOAA15/16/18/19--AMSUAMETOP-A/B–AMSUAFY-3B/C--MWTSNPP--ATMS高光谱辐射率卫星--搭载探测仪器：AQUA–AIRSGNSSGPS/RO掩星折射率卫星--搭载探测仪器：COSMICMETOP-A/B--GRASGRACE卫星资料同化_GRAPES模式系统中同化的观测资料GRAPES-GFS中同化的卫星资料量占总资料量的比例：60%左右引自陈静，《业务数值天气预报》，201645GRAPES与UKMO业务模式资料使用情况对比过去5年、特别是近三年，GRAPES-GFS同化的资料量有很大提高与英国模式资料应用量的差距显著减小引自陈静，《业务数值天气预报》，201646气象现代化发展纲要（2015-2030年）一、指导思想和发展目标二、大力提升气象防灾减灾和公共气象服务水平（