Abstract: Four-dimensional variation data assimilation (4D-VAR) is a logical and rigorous mathematical method to obtain the "best" estimate of the model initial conditions from observations and a priori knowledge of the atmospheric state. It is one of the most advanced data assimilation methods today. Automation weather station (AWS) precipitation data is assimilated by 4D-VAR in experiments. Experiment results show that, due to addition of information of AWS precipitation data, the initial field of test is enhanced in meso-scale information, and it matches the model better in thermo-dynamical mechanism. After assimilation, the simulation is improved. The precipitation during the start period in simulation is increased, and the situation of simulating precipitation matches real situation better. The "spin-up" problem of the model is weakened. Experiment results also show that temporal information of AWS precipitation data is very important for assimilation.

Key words: four-dimensional variation data assimilation, AWS precipitation assimilation, meso-scale numerical model, torrential rain simulation

摘要： 用自动站降水资料作了四维变分同化试验。试验表明，由于它的加入，增加了初始场中的中尺度信息，改进了中尺度数值模式MM5的预报，增强了模拟开始阶段的降水量, 改进了降水量的落区预报，减弱了模式开始阶段的“spin-up"现象。试验还表明，自动站降水资料的时间变化信息，在同化时也起重要作用。

关键词: 四维变分, 自动站降水量同化, 中尺度数值模式, 暴雨模拟