Study On Doppler Radar Data Direct Assimilation

Doppler radar data with high temporal and spatial resolution could supply the resolution gap between models and observations, and many studies have been done on radar data assimilation. With the establishment of Doppler Radar Network in China, how to make a full use of the radar data in NWP to improve the rainstorm forecasting is an urgent task we are facing.The direct assimilation experiments of radial velocity and reflectivity are researched with the mesoscale numerical weather prediction model WRF V3.2, WRF-3DVAR V3.2 assimilation system and WRF-EnSRF assimilation system which is built by ourselves. They are tested with a heavy rainfall process in the mei-yu front occurred at Yangtze River during 4-5 July 2003. NCEP reanalysis data and Next-Generation Weather Radar data of Hefei are used in these experiments. The results are as follows:(1) The quality control of Radar data which includes removing ground clutter and isolated points, data filling, data smoothing, elevation angle adjustment and data thinning can remove noise and reduce the data correlation. These measures are important for the radar data assimilation.(2) When using 3DVAR and EnSRF to assimilate radar data directly, assimilating radial velocity can improve the accuracy of wind and the exact location of the strong echo. Assimilating reflectivity can change the microphysical variables and dynamic field in the model, these changes will increase the area of radar echo. Because the direct assimilation experiments of radial velocity and reflectivity have the advantage of the former experiments, the wind and radar echo will be closer to observation. Different assimilation intervals have positive effects on the assimilation results, reducing assimilation interval can improve assimilation results, but the advantage is not obvious.(3) When using EnSRF to assimilate reflectivity, it can forecast the rainfall center accurately, which does better than using 3DVAR. However, both of them forecast the range and magnitude of precipitation center largely. Assimilating radial velocity can significantly inhibit the excessive rainfall, while assimilating radial velocity and reflectivity shows better results. (4) The direct assimilation experiments using 3DVAR and EnSRF can enhance the precipitation forecast, but the 3DVAR assimilation experiments can get higher TS score and improve the results even more.(5) To 3DVAR, the influence of different background error covariance for the three-dimensional wind field mainly performs in the horizontal wind field. Cycle of assimilation does little improvement for the vertical velocity increment, but large improvement for horizontal wind. The convergence line corresponds to the rain belt well, which is more conducive to the generation of precipitation. Cycle of assimilation of radial velocity can make the radar echo location with a good agreement with the observation, and filter out some strong and isolated convective cells, which can accurately forecast the storm's rainfall centers, and improve the scope of the rain belt greatly.(6) To EnSRF, the overall assimilation effects of local disturbance are no better than that of overall disturbance. When using local disturbance, it dose obviously affects for the precipitation center of inside the disturbance range but less affects for the center of outside the disturbance range.