Four-dimensional Ensemble Variational Assimilation Of All-sky Satellite Data

A four-dimensional ensemble variational assimilation method based on Proper Orthogonal Decomposition(referred to as POD-4DEnVar)is an advanced hybrid assimilation method.It combines the ensemble Kalman filter assimilation method with four-dimensional variation assimilation method by POD decomposition.The method not only includes the flow-dependent background error covariance,but also does not need to solve the tangent linear adjoint model of forecast model and observation operator.Therefore,this method has great application potential in satellite data assimilation.In this thesis,the POD-4DEnVar method is applied to the satellite data assimilation,and the three/four-dimensional ensemble variational assimilation system of satellite data is constructed with Community Radiative Transfer Model(CRTM)as the observation operator.Satellite observations such as microwave temperature,humidity and imaging radiances are assimilated into the Weather Research and Forecasting(WRF)model by the assimilation system.Single observation experiments,observation system simulation experiments(OSSEs)and real observations assimilation experiments are designed to investigate the assimilation ability of the system for a heavy rainfall case over the Yangtze River and a binary typhoon case in the western Pacific.The main work and conclusions are as follows:First,the POD-3/4DEnVar assimilation method is successfully applied to the clear-sky satellite data assimilation.Three/four-dimensional ensemble variational assimilation system(PEVS-3D/PEVS-4D)are constructed to assimilate clear-sky satellite observations.The experiments show that the system can effectively assimilate the clear-sky satellite observations of multiple times in the time window and improve the prediction ability of typhoon track,intensity and precipitation.Then,the POD-3/4DEnVar assimilation method is successfully applied to the all-sky satellite data assimilation.Three/four-dimensional ensemble variational assimilation system(PEVS-3/4D-ALL)are constructed to assimilate all-sky satellite observations.Three key elements of the PEVS-3/4D-ALL are:(1)including four hydrometeor mixing ratios for cloud-water,cloud-ice,rain,and snow in the control variables;(2)turning on the scattering module of CRTM;and(3)combining the historical sampling method with physical ensemble forecasts to produce ensemble samples to estimate the flow-dependent background error covariance of the control variables,including those for hydrometeors.The experiments demonstrate that the system can effectively assimilate satellite observations in cloud and precipitation areas near typhoon center,improve the initial field of hydrometeors,and further enhance the prediction ability of the typhoon track and precipitation.Finally,the sensitivity of the POD-3/4DEnVar method to the percentage of truncated eigenvalues,the number and the construction method of ensemble members,and the horizontal localization scales(which are key parameters for POD-3/4DEnVar method)are tested.And the parameters of reference value for the application of operational forecasting are obtained.The experiments demonstrate that the percentage of truncated eigenvalues and the horizontal localization scales have great influence on assimilation and forecast results.The results of assimilation and forecast are the best when the percentage of truncated eigenvalues is 100% and the horizontal localization scales is 2500 km.