Study On The Techniques For Assimilation Of Doppler Radar Data

With an ultimate goal of operational application of Doppler radar data in numerical weather prediction (NWP), a primary Doppler radar data analysis and assimilation system using three-dimensional variational (3DVAR) and Ensemble Kalman Filter (EnKF) techiques has been constructed. It consists of four sub-systems: (1) Doppler radar data preprocessing system; (2) Doppler radar data retrieval system of three-dimensional wind fields; (3) Three-dimensional variational assimilation system for Doppler radar data; (4) Ensemble Kalman Filter assimilation system for Doppler radar data. Above sub-systems have been tested with the dataset collected by the field experiment of China Heavy Rain Experiment and Study (CHeRES) in 2002 and 2003.In the data preprocessing system, traditional steps (such as isolated point and singularity removal, spectrum width thresholding) are used first, and then background is used for velocity dealiasing and removing AP (anomalous propagation) Clutter and other noise. Three interpolation schemes are further tested with the simulated radar data created by using high-order polynomials to fit the radar observed wind. The first scheme, the radar data are first interpolated to given horizontal grids on every cone while remaining on the constant elevation angle levels, and then the data are further interpolated to the Cartesian grids with a linear inter interpolation in the vertical direction (CVI);the second scheme is the three-dimension Barnes interpolation technique (3D-Barnes); the third scheme is based on the variational analysis (VAR).The results show that the 3D-Barnes scheme obviously excels the othersIn the Doppler radar data retrieval system, a two-step variational method was introduced to retrieve three-dimensional wind field from Doppler radar observations. In this method, first, a smoother wind field is retrieved in a low-order spectral space as the background for the next-step retrieving. Then a detailed structure of wind field is obtained at grid points. This method is extended for multi-Doppler retrieval. And it is introduced to analyze the stream structure of two heavy rain events. The result shows that the method can well retrieve the stream structure of the heavy rain.In the three-dimensional variational assimilation system for Doppler radar data, first, Doppler radar radial wind observations are assimilated directly with modified 3VAR system based on WRF-Var (1.3version, 2003), the microphysical variables, and potential temperature and water vapor special humility of the model air are adjusted by using Doppler radar reflectivity observations. Doppler radar data from Hefei radar are assimilated into WRF for a heavy rainfall case that occurred on 5 July 2003. The results show that the analysis wind contains more mesoscale information; and the adjusted variables are correspond well with the reflectivity observations in distribution and intensity. Assimilation of Doppler radar data can mitigates the spin-up problem with precipitation forecast, resulting a significant improvement in precipitation forecast in the first 2 hours. Second, the horizontal winds retrieved from Doppler Radar data by using a two-step variational method and Doppler reflectivity are assimilated with WRF 3D-Var. Doppler radar radial velocity observations are also assimilated directly by the method afforded by WRF 3D-Var. Results from the case studies show positive impact of the Doppler Radar data assimilation upon the precipitation forecast; assimilation of retrieved winds gives better forecast than the method that radial velocity observations are assimilated directly; assimilation of both retrieved winds and reflectivity obtains the highest threat scores. In the end, WRF-Var is further developed with Physical Initialization (PI) to assimilate Doppler radar radial velocity and reflectivity observations. In this updated 3D-Var, specific humidity, cloud water content and vertical velocity are first derived from reflectivity with PI, and then the model fields of specific humidity and cloud water content are replaced by the modified ones, finally, the estimated vertical velocity is added to the cost-function of the existing WRF-Var as a new observation type, and radial velocity observations are assimilated directly by the method afforded by WRF-Var. It is tested with a rainfall event that occurred in the middle area of the Yangtze River on 19 June 2002 and a heavy precipitation event that occurred on 5 July 2003. Results show that a significant improvement is noticed. It reduces the spin-up time significantly, and it is an effective way to improve the short-range prediction of precipitation.In the Ensemble Kalman Filter assimilation system for Doppler radar data, the observations whether they occur at the assimilation time or at some earlier or late time are all assimilated with Ensemble Square Root Kalman Filter (EnSRF) based on WRF model. Result shows that the scheme produces better analysis than the scheme that only assimilation of the observation at the assimilation time, especially in the first several assimilation cycles. The comparison of two methods of assimilating Doppler radar wind data with EnSRF is also presented. One is to assimilate simulated radar radial velocity observations are assimilated directly. The other is to assimilate the horizontal winds retrieved from Doppler Radar Data. Result shows that the second method that assimilated retrieved winds with small error gives better analysis than the first method does.