Study Of Assimilating Wind,Temperature And Humidity Remote Sensing Data In Hybrid Rapid Refresh System

The accuracy of model initial values is one of the key problems that restrict the accuracy of numerical weather prediction.Compared with ensemble-variational(hybrid)data assimilation which produces flow-dependent background error covariances,three-dimensional variational(3DVAR)data assimilation used in conventional rapid refresh systems has a defect that utilizes a Isotropy,homogeneous and static background error covariances.In order to improve the analysis and adapt to the high quality requirements of the model initial values of the rapid refresh system,a Hybrid Rapid Refresh(Hybrid RR)system which could well reflect the flowdependent forecast error of real atmosphere was built using ensemble-variational data assimilation method.With data of wind profile radar detection and microwave radiometer in the Beijing metropolitan observation network assimilated,based on experiments of assimilating temperature,humidity and wind remote sensing data in Hybrid Rapid Refresh system,parameters optimization study and rapid refresh contrast experiments of different schemes were conducted.The main conclusions are as follows:1)Hybrid RR system could well reflect the flow-dependent forecast error of real atmosphere.Test result of background error covariances in Hybrid RR system showed that flow-dependent background error covariances were led into Hybrid RR system,and well reflected the weather situation.Test result of ensemble spread showed that RMAPS-EN ensemble members,Ensemble Transform Kalman Filter(ETKF)and multi-physics ensemble forecast could effectively produce high quality ensemble perturbation with reasonable amplitude and fitting motion characteristics of real atmosphere;2)Result of parameters optimization study showed that the best lengthscale factor and variance factor of temperature,relative humidity,u-wind and v-wind were respectively 0.7/1.0,1.0/1.0,0.7/1.0 and 0.7/1.0,the best localization scale and ensemble weighting factor were respectively 11.2km and 0.5.Sensitivity tests indicated that tuning variance factor could under the guidance of distribution of background,observation and truth;3)Rapid refresh contrast experiments of different schemes appeared that hybrid data assimilation could make analysises and forecasts better than 3DVAR in the rapid refresh system,and the best parameters settings could further improve analysises and forecasts of Hybrid RR system.