Application Of Radar Data Assimilation In Severe Convective Weather Forecasting

In this paper,three-dimensional variational(3DVar)and ensemble Kalman filter(EnKF)are used to assimilate radar data.This study focuses on the performances of 3DVar and EnKF in storm-scale data assimilation,as well as developing various methods to improve their analysis and forecast skills.Based on these studies,a hybrid EnKF-En3DVar system that aims to combine the advantages of the enhanced EnKF and 3DVar is developed.Furthermore,conventional and radar data are assimilated using two-step method to improve the forecasting of severe convective weather.The main contents and conclusions are as follows:(1)Based on the WRFDA 3DVar indirect radar data assimilation system,a neighborhood-based scheme to assimilate no-rain radar observations was developed,and its impact on precipitation forecast is evalued using a severe convective system.A number of sensitivity experiments were also run to examine how sensitive the scheme is to the neighborhood size and the proportional number of no-rain data in the neighborhood(alpha threshold).The reason of the improvement of precipitation skill was also explored.It is shown that the no-rain data assimilation method reduces the BIAS and FAR of precipitation over its counterpart without that assimilation.Although all sensitivity experiments were able to reduce precipitation bias,a larger radius and moderate alpha threshold produced superior results.It is also shown that the advantage of the scheme is in its ability to conserve total water content in cycled radar data assimilation,which cannot be achieved by assimilating only precipitation echoes.(2)To conquer the drawback of static background error of 3DVar approach,more advanced EnKF method is applied to assimilate radar data to study a squall line that occurred in southern China.The ability of radar data assimilation with the EnKF method over China has been explored.The Bayesian inflation method with the advantage of space-time adaptive theory is further introduced to assimilate radar data to reduce the sampling error.The advantages of Bayesian inflation method was first examined using a simulated supercell through comparing to the experiment using the multiplicative inflation method.Based on the EnKF method to assimilate multiple Doppler Radar data,the Bayesian inflation method is further applied to an MCS in eastern China.Results show that the ensemble members with radar data assimilation is better than that without radar data assimilation with higher ETS.Experimental results of the simulated data case indicate that the analysis fileds from Bayesian inflation method are more close to the true fileds and the spread of the ensemble is bigger while the root mean square error is smaller.Experimental results of the real data case show that:compared with the multiplicative inflation experiment,the analyzed and simulated composite reflectivity are more close to the observertion,and wind and cold pool from the Bayesian inflation experiment are more reasonable.ETS of composite reflectivity from Bayesian inflation experiment is higher than that from the multiplicative inflation experiment for various thresolds.It is found that Bayesian inflation method can give more weight to radar observations by producing large inflation parameters and provides bigger analysis increment when the root mean square innovation of background is bigger.(3)The improved 3DVar method and EnKF are further systematically compared to investigate the impact of different radar DA methods on both the analysis and the subsequent reflectivity and precipitation forecasts of a severe convective system.It is found that assimilation of radar data resulted in considerable improvement of the analysis,and the analyzed convective structure produced by EnSRF was more realistic than 3DVar.EnKF improved the quantitative reflectivity and precipitation forecast skills measured by FSS,ETS,BIAS,and FAR over 3DVar.The neighborhood probability of the reflectivity forecasts indicated that EnKF produced more skillful probabilistic forecast with higher probabilities than 3DVar in the observed reflectivity region for different thresholds.Additionally,compared with 3DVar,the root mean square error of wind,temperature,and humidity near the surface and at upper levels are also reduced by EnKF.(4)Based on the studies above,a coupled hybrid EnKF-En3DVar system is developed to assimilate conventional and radar observations by using the two-step procedure.The system was examined by comparing with the 3DVar experimental results.The results show that the assimilation of radar data in either 3DVar or hybrid experiments could considerably improve short-term precipitation forecasts up to 7 hours.The hybrid experiment more accurately reproduced observed rainfall amounts than the experiments using 3DVar,especially regarding the structure and intensity of the heavy rainfall center.The improvement is more evident when assimilating radar data.The improved precipitation and reflectivity forecast skill is attributable to improvements in the forecast background field of wind,temperature,and water vapor mixing ratio.Radar data assimilation with the hybrid method enhanced low-level cooling because of rainwater evaporation and mid-level warming associated with latent heating,in addition to stronger winds in the convective region.