An Approach To Assimilating Radar Data Assimilation With A Large-scale Data Constraint

Radar data assimilation can improve the prediction of severe convective weather,but some case studies showed that the location and intensity of predicted precipitation obtained by radar data assimilation prediction are still significantly different from observations.The deviation of the large-scale field in the regional model may be one of the reasons why the prediction has error in position.The large-scale bias has been widely concerned in dynamical downscaling fields,but only a few studies document the importance of large-scale bias on weather forecasting in recent years.For this,we firstly verify the large-scale bias between Weather Research and Forecasting(WRF)model simulation and its driving field through two severe weather events occurred over the Yangtze-Huaihe river basin(a squall line case on June 14,2009 and a severe convective case on July 20,2013),and then perform the bias correction using spectral nudging(SN)technique.Based on the above,a method(LSC-RDA)is developed to mitigate the effects of large-scale bias on analysis and forecast results.A series of experiments are conducted with the WRF model and its three-dimensional variational(3DVar)module to investigate the effectiveness of LSC-RDA.The main conclusions are listed as follows:(1)Obvious large-scale biases occur between the WRF simulations and its driving fields for two severe convective cases.Even for a short-term(eg,6 hours),the simulated large-scale components of temperature and wind fields has a significantly difference from NCEP FNL analysis,in both pattern and intensity.The large-scale bias increased with the increasing integration time.The large-scale bias occurred in the squall line case was stronger than that in the severe convective case occurred on July 20,2013.(2)The spectral nudging(SN)method could effectively correct large-scale bias occurred in convective-scale simulations.The correction effect gets better with the increase of nudging wave number.However,the result of the large-scale correction has a small difference when nudging wave number is greater than 2(2~4 in this paper).(3)A radar data assimilation approach with a large-scale constraint(LSC-RDA)was proposed to mitigate the effects of large-scale bias,in which the global analysis data are introduced by SN method to obtain more accurate background information for radar data assimilation and to consequently improve the forecast skill.The backgrond inforamtion includes first guess field(FG)and the background error covariance(BE)statistics.The scheme of LSC-RDA is built on the WRF model and its 3DVar system for carrying out the high-resolution radar data assimilation.(4)The LSC-RDA method performs bettern than WRF 3DVar,as shown by the experimental results on two severe cases.The LSC-RDA method has better performance when large-scale constraints were applied simultaneously to FG and BE.The forecasted composite reflectivity is consistent with observation in the pattern and location.However,the larger nudging wavenumber was not appropriate for radar data assimilation domain,maybe because the use of larger wavenumber could result in an excessive adjustment of regional-scale information and consequently suppress the absorption of meso-and small-scale information from radar observations.(5)The BE statistics with large-scale constraint plays an important role in radar data assimilation.The length scales of BE with the large-scale constraint are smaller than those without the large-scale constraint,which in some way mitigates the overestimate by the NMC method,providing a more reasonable BE for radar data assimilation.