Numerical Simulation And Data Assimilation Of Heavy Rainfall Over Tianshan Mountain Area

Based on the Weather Research and Forecasting model (WRF), three heavy rainfall scenarios, i.e., through July17,2004to July21,2004, July15,2007to July18,2007and August11,2008to August14,2008, respectively, occurred in Xinjiang Tianshan region were simulated and three-dimensional variational data assimilation and forecasting tests were conducted for the third precipitation process. The major results are summarized as follows:(1) The WRF model shows a good ability to simulate heavy rain process in Tianshan region. The simulations confirm that Xinjiang rainstorm is mainly related to the baroclinic disturbance, which occurrs in the context of large-scale weather systems superimposed on the mesoscale convective system activity. And it is quite different with the storm rain events in eastern China, which is greatly influenced by the local convection systems in a more uniform underlying surface thermal field.(2) As for the large-scale-forcing related rainstorm, especially for the stratiform precipitation, the impact of cloud microphysical precipitation is much larger than that of cumulus convection, radiation and the planetary boundary layer parameterization. Comparatively speaking, the cumulus convection has the minimal impact in all physical processes on the physical element fields including precipitation, dynamic and thermal fields but except for the high-level water vapor conditions. The quality of the model precipitation forecast is almost entirely from the contribution of the grid scale precipitation. In the context with a resolution of9km fine grid scale, Thompson scheme seems to perform best for the heavy rainfall forecast as compared with all the other microphysical parameterization schemes.(3) Construct of a suitable background error covariance matrix is very important for numerical weather prediction in Tianshan region. Compared with the simulation effects (magnitude and time) of precipitation in the application of sounding data assimilation tests, the use of localized background error covariance to adjust the characteristic scale shows significantly better performance than that in use of globalization background error covariance, due to the fact that the former could effectively absorb more observations and retain the dynamic coordination of the initial fields in the meanwhile.(4) By comparing the assimilations of the encrypted automatic weather stations data, sounding data and brightness temperature data in xinjiang area, respectively, it shows that their analysis fields other than the background field are all closer to the actual atmospheric conditions, and different data have different impacts on the background field with the most significant improvement appeared in the assimilation of sounding data. As viewed from the skill score and statistics, data assimilation could stably improve the skill level of numerical prediction of heavy rainfall in Tianshan region, especially for the assimilation of sounding data which could improve0.115of the precipitation forecast skill score. In Addition, based on the sensitivity data assimilation experiments of the introduction of elements of the automatic weather stations, results show that the assimilation of each element (temperature, pressure, wind, moisture) into the three-dimensional variational system has a relative impact on the results of model forecast, with wind data most on physical field forecasts and pressure data on precipitation forecasts. Increase the assimilation frequency of the automatic station data could reduce the underreporting phenomenon, and lead the simulated precipitation values closer to the observations.