A Comparative Study Of Multiple Data Assimilation Schemes For High-density Automatic Weather Stations In East China And Coastal Areas

GSI-3DVAR(Gridpoint-Statistical Interpolation–3-Dimensional Variational data assimilation)is widely used in numerical weather prediction operations and research.Based on a 3km-grid WRF(Weather Research and Forecasting)model and high-density surface automatic weather station(AWS)observations in East China,the effect of GSI-3DVAR and FDDA assimilation methods on the temperature,wind direction,wind speed and relative humidity of surface AWS data in the areas with different terrain over the eastern China was studied.The results are summarized as follows:(1)When using GSI-3DVAR to assimilate surface AWS observation,the value of R_HZSCLis critical,and the accuracy of the surface analysis can be effectively improved by selecting appropriate R_HZSCL.The most appropriate R_HZSCLcan reduce the RMSE of background field surface temperature and vector wind difference(VWD)by more than 35%.However,if R_HZSCLis too large,an excessive expansion on the high and low temperature centers occurs and the analysis error is increased.For the surface wind analyses,the default R_HZSCLprovided in GSI-3DVAR produced overly smooth wind and little mesoscale structures.Furthermore,a too small R_HZSCLwould cause false strong local winds.For the sea surface and western areas with sparse observation distribution,a larger R_HZSCLis required to better propagate the observation information.By carrying out observation data-thinning experiments,it is found that the optimal R_HZSCLfor surface weather analysis varies in respondence to station densities of surface temperature and wind observations.(2)The hourly GSI-3DVAR updates improved the analysis accuracy of the model surface temperature,wind and water vapor.The surface temperature and water vapor analysis errors over the flat terrain area were less than over the complex.Compared with the smooth wind of the control experiment,the surface wind field after hourly 3DVAR updates assimilation obtained some mesoscale structures in boundary layer and improved the high wind speed in control tests.(3)GSI-3DVAR assimilation with hourly update cycles and FDDA assimilation can improve the analytical accuracy of the WRF model.Overall,GSI-3DVAR cyclic assimilation was slightly better at analyzing surface temperature and wind fields in the flat area than the FDDA test,and vice versa in complex terrain area.The diurnal evolution of the surface temperature of the assimilation experiments is consistent with observations,which have less error than control result.GSI-3DVAR cyclic assimilation performs better at night,but less stable than the FDDA;GSI-3DVAR cyclic assimilation and FDDA have better simulation effects on the surface wind in the flat and complex terrain areas,respectively.GSI-3DVAR cyclic assimilation causes higher wind speed and large errors in complex terrain region.Over the flat area,both methods performed well during the night and morning.The surface wind speed analyzed by the GSI-3DVAR cyclic assimilation at night was too strong,and the improvement of FDDA was also weakened.Finally,the analysis of the surface water vapor of FDDA was better than GSI-3DVAR cyclic assimilation.(4)With assimilating of the surface-based AWS observation data in coastal areas,both GSI-3DVAR cyclic assimilation and FDDA can have a certain impact on offshore weather filed.The impact range is concentrated in the sea area within about 50 km.FDDA reduced the offshore wind speed and increased the vapor value over the East Sea.GSI-3DVAR cycle assimilation tends to produce a center of wind speed analysis increment at the near-shore ocean due to an existence of a large difference in observation density,which may cause issue on the wind field simulation with WRF.