Abstract:The hierarchical ensemble filter(HEF) and sampling error correction(SEC) localization methods can minimize sampling error without giving definition of physical distance.To examine the advantages of the two methods and the possibility of applying them to storm-scale assimilation,experiments involving assimilating radar data are conducted using the ensemble Kalman filter(EnKF).Compared with the Gaspari-Cohn(GC) experiment,the influence of the localization methods on the assimilation effect is investigated.Results show that the analysis reflectivity coverage of all the experiments is smaller than the true reflectivity.The analysis reflectivity of the HEF experiment is bigger than that of the GC experiment in both the horizontal and vertical directions.The analysis error of most model variables decreases with time and becomes lower after analysis.This indicates that radar data assimilation can help to improve the quality of the forecast field.The RMSE of the HEF experiment is the smallest and the analysis error of the SEC experiment is smaller than the GC experiment.Compared with the GC experiment,the analysis error of the U,V and W of the HEF experiment decreases more sharply than the microphysical variables,including QR(cloud-water mixing ratio),QC(rainwater mixing ratio),QI(ice mixing ratio),QS(snow mixing ratio) and QG(graupel mixing ratio).For U,V and W,the analysis error decreases by 25% and,for microphysical variables,it decreases by 17%.The spread of the HEF experiment is largest and the spread of the SEC experiment is larger than that of the GC experiment.Compared with the GC experiment,the spread of the U,V and W of the HEF experiment increases by 45%,while that of the microphysical variables increases by 42%.In the convective region,the temperature is colder than the environment,which is called the cold pool.This is caused by the evaporation of the rainwater in the convective system.The strength and coverage of the cold pool of the GC experiment are stronger than the true field,while the HEF and SEC experiments are weaker and their areas are smaller.From 60 km to 120 km in the south-north direction,and from 0 km to 12 km in the vertical direction,the areas of vertical wind and Graupel mixing ratio are bigger,while their values are larger.So,they are closer to the wind and Graupel mixing ratio of the true field,respectively.Through simulation of the analysis fields,it is found that the northern branch of the convective system of the HEF experiment is stronger than that of the SEC and GC experiments,especially at 80 km in the south-north direction.The true field and HEF forecast result can reach about 50 dBz,which corresponds well with the assimilation results.The southern branch of the convective system of the SEC experiment is stronger than that of the HEF and GC experiments.The SEC experiment can simulate the new convective cell at(40 km,60 km).The cold pool of the HEF experiment is coldest,reaching as low as 299 K.Both the HEF and SEC experiments can simulate the center of the graupel mixing ratio.These results prove that the HEF and SEC localization methods can improve the performance of the EnKF based on GC localization method.The SEC localization method is inferior to the HEF method,but it can reduce the computational expense of the HEF method and its effect is better than the GC method.So,it could be a good choice when the NWP model is complicated.