1、Adaptive Estimation and Tuning of Satellite Observation Error in Assimilation Cycle with GRAPESHua ZHANG,Dehui CHEN,Xueshun SHEN,Jishan XUE,Wei HAN China Meteorological Administration(CMA)OUTLINE Introduction of GRAPES-3DVar Tuning of obervation error in data assimilation Latest development in the g
2、lobal assimilation/prediction experiment：2008 Summary1.Introduction of GRAPES-3DVarMain features of GRAPES_GASGrid analysisA+P with flexible resolution setupincrementalxa=xb+xVariable optionsanalysis/T,u,v,rhcontrol,u,rhpreconditioningcontrol space model space x=Uw,U UpUvUhRegional:Recursive filter
3、for UhGlobal:Spectral filterfor UhMinimizationLimited memory BFGS methodMass-wind constraintLinear balance equation(now)Nonlinear balance equation(on testing)ProgrammingFortran90,Modular structure,to be paralleledPreprocessingRaw ATOVSDATAQuality ControlPreprocessingConventionalDATAQuality Control10
4、D ForecastGRAPES GLOBAL 3D-VARGRAPES GLOBAL MODELINCREMENTAL SIDIGITAL FILTER INITIALIZATION6h ForecastcycleAt 00/12ZGRAPES_GFSanalysis：1.875 forecast：1?bbtxx Cost function 1111()()()()()22b TboToboJJJxx xBx xyHx RyHxootyHxBacground error:Observation error:()0bE()0oE()TbbEB()TooER lBasic hypothesis:
5、l l l()0TboE Optimality criterion(Bennet 1992;Talagrand,1999)211min111()()()()2222TpE JEETrd D dD D 2.Tuning of background and observation error in data assimilation(Wei HAN and Jishan XUE,2007)TTDddRHBH111(,)(,)(,)(,)(,)aoTooTooTbkkbakkbbkkbooTkakkbTaoTkbkkbEEEEEPdRB ddRB ddRB dRdRB dPHBHdRB dPinno
6、vation covariance:Iterative fixed-point method:Desrosies et al.,200511()()()ooTTooTabbbEEd dR RHBHd dRD D11()()()aoTTTooTTbbbbEEd dHBHRHBHd dHBH D D()ooTTbbEd dRHBH(1)(2)only Sonde RH observation assimilation in GRAPES regional 3DVAR20070601-0614Only RH obs.are assimilated to test the approach,since
7、 it is thus a univariate analysisBlue dot:initial obs.error of rhBlue dash dot:initial background error of rhNOAA16,AMSUA20070601-0614diagnosisObs erroBak.erro ITWG NWP WG list of assumed observation errors Against Radiosonde：humididy information of AMSUB has a proper response in GRAPES-3DVAR58238,N
8、anjing59948,SanyaRed :xbBlue :xa(amsub)Black:SoundeIndependent verification:RHxa(amsub)-Y(sonde)Before TuningAfter Tuning2007060900,500hPaBlack:Before Tuning;Red:After tuning10 cases statistics12300.10.20.30.40.50.611089414Threat Score(24h)operamsu(e1)12300.10.20.30.40.50.611559414Threat Score(48h)o
9、peramsu(e1)Tuning of observation error improve GRAPES(30km)QPF01/1301/1401/1501/1601/1701/1801/1901/2001/2101/2200.51Threat Score95951091101251179997127134operamsu(e1)01/1301/1401/1501/1601/1701/1801/1901/2001/2101/2200.51EQ.Threat Scoreoperamsu(e1)01/1301/1401/1501/1601/1701/1801/1901/2001/2101/220
10、0.511.5Bias Scoreoperamsu(e1)01/1301/1401/1501/1601/1701/1801/1901/2001/2101/2200.51Threat Score0951091101251179997127134operamsu(e1)01/1301/1401/1501/1601/1701/1801/1901/2001/2101/220.20.4EQ.Threat Scoreoperamsu(e1)01/1301/1401/1501/1601/1701/1801/1901/2001/2101/220.511.52Bias Scoreoperamsu(e1)3.La
11、test development in the global assimilation/prediction experiment：2008(Xueshun SHEN et al,2008)n Re-estimate the obs.error of sonde and Re-estimate the obs.error of sonde and radiancesradiancesn SEMI-Bias Correction in backgroundn Modify the QC of satellite radiancesModify the QC of satellite radian
12、cesn Introduce NOAA-15Introduce NOAA-15n Improve the surface albedoImprove the surface albedon Introduce the diagnostic cloud ref.ECMWFIntroduce the diagnostic cloud ref.ECMWFn Introduce the new OIntroduce the new O3 3 datadatan Daily SSTDaily SST ATOVS microwave (NOAA15 16 17)radiances Sondes geop/
13、humidity/wind Synops geop/humidity/wind Ships geop/humidity/wind Airep temp/wind Satob windData application of GRAPES-3DVAR 500hPa ACC against NCEP 2TObRadRadHBHRl海温资料更新海温资料更新l湿度分析（水物质相态）湿度分析（水物质相态）l采用增量插值标准初始化方案采用增量插值标准初始化方案l针对南极探空的质量控制针对南极探空的质量控制l考虑模式偏差的卫星资料偏差订正考虑模式偏差的卫星资料偏差订正l模式偏差订正参数调整模式偏差订正参数调整
14、(0.9,0.3)l背景误差背景误差(循环同化预报结果统计循环同化预报结果统计)l探空资料质量控制探空资料质量控制l卫星资料背景场误差和观测误差的估计卫星资料背景场误差和观测误差的估计l卫星资料卫星资料质量控制新方案质量控制新方案(Background Check)2ObRadRadR走过四季走过四季冬春夏秋冬春夏秋北半球，北半球，10天预报的天预报的500hPa ACC检验检验(.vs.NCEP ANA.)(20061201122007013112,62cases)南半球，南半球，10天预报的天预报的500hPa ACC检验检验(.vs.NCEP ANA.)(2006120112200701
15、3112,62cases)南半球预报检验，南半球预报检验，31cases(200612),against NCEP ANA.NOAA-15NOAA-15资料的影响试验资料的影响试验Summary It is promising for the new implementation of the tuning observation error.GRAPES is progressing,which improve its performance.Sondes are important in southern pole region.more satellite data applicatio
16、nSuggestions?Assimilation:more satellite data application,especially in SH and ocean any possible data(real-time)&experiences?ModeluWeak subtropical highuExcessive precipitation over the maritime continentuLarge cooling bias at top(10hPa)uCoupling of SISL dynamics&physicsuHybrid vertical coordinate
17、in non-hydrostatic model考虑了背景偏差的卫星资料观测偏差订正方案：“半偏订正”5678910-0.500.511.52AMSUA ChannelsBias of H(xb)-yo 90S-60S60S-30S30S-00-30N30N-60N60N-90NIt is obvious that the systematic departure:H(xb)-Yo,Is due to model bias,So we make a“Semi-Bias”correction0,1.0()()Njjiikkkbsc pxx12()()()(,)()(,)()()ToToTJHH1bbxx-xBx-xyxOFyxW，()()TWAs a regularization term in VarBC
