WANG Yang，LIU Jia，YU Fuliang，LI Chuanzhe，TIAN Jiyang，QIU Qingtai
State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin，China，Beijing 100038，China
Abstract: Numerical simulation of rainfall is an important method to extend the hydrological forecast lead time. However, since the initial field and boundary field conditions provided by the driving data could not perfectly agree with the actual state of the atmosphere, the simulation results would contain a certain error. The reduction of errors in numerical rainfall simulation is a key problem to be sotted for the improvement of hydrological forecasting accuracy. With the Weather Research & Forecasting （WRF） model and the three-dimensional variational data assimilation technique, in this study the spatial distribution of the simulated rainfall on the basis of data assimilation was thoroughly examined, and the simulation results before and after assimilation were compared by using two evaluation indices, i.e., CSI and RMSE. The results demonstrate that: by using the method of data assimilation, we can capture the cumulative rainfall distribution that WRF model fails to simulate, thus improving the consistency of the spatial distribution between the simulated rainfall and the gauge observed rainfall; the simulated rainfall accumulation after assimilation presents better CSI and RMSE indexes than that before assimilation; the simulated data after assimilation can more accurately describe the spatial distribution of rainfall. The enhancement of spatial accuracy of the simulated rainfall would help provide more reliable rainfall information to hydrological models through the coupled atmospheric-hydrological modeling system, so that better flow forecasting results can be obtained.
Keywords: Numerical rainfall simulation; WRF; Data assimilation; Spatial Distribution
(Journal of China Institute of Water Resources and Hydropower Research, Vol. 16, No. 3, 2018)