一种数值模式的嵌入式动力统计订正技术研究

Research on an Embedded Dynamic-statistics Correction Method of Numerical Model

目前数值天气预报效果的提高主要依靠改进模式初始场和模式输出结果统计订正两种方式。改进模式初始场虽然可在一定程度上提高模式预报准确率,但是模拟结果中的系统误差仍然存在。而模式输出结果统计订正虽然能消除一部分的系统误差,但是无法使预报变量之间相互协调。鉴于以上原因,本研究提出嵌入式模式逐步订正方法(EMPC)的思想和技术路线,以WRF模式为试验,结合FNL分析场历史资料,将模式输出订正方法嵌入到模式积分过程中,对2015年6月和7月进行回报。同FNL资料、地面观测及探空资料对比检验发现,经过EMPC订正的预报结果在短期内优于WRF模式直接预报结果。由于对初始场中物理量之间的协调性考虑不全,可能破坏了模式积分过程中的守恒性,使模式存在不稳定风险。

Currently,the improvement of accuracy of numerical weather prediction depended on modified initial fields and the statistics correction of model output. Although the accurate initial fields could promote the precision of prediction to a certain extent,the systematic errors of predicted output exist. Similarly,the statistics correction to model output can eliminate partly the uncertainty of model in terms of the systematic error,but it cannot make predicted variables to integrate each other. For the reasons above,the embedded model processing correction( EMPC) method was brought up. In this research,the temperature fields in June and July in China were simulated by WRF model and EMPC method. Combined with the corresponding period FNL data,ground observation at 943 weather stations and sounding data at 9 radiosonde stations in China,the model outputs were tested. The results showed that the model output by using EMPC method was better than the directly predicted output by WRF in a short time. However,the conservation of model might be broken due to considering poorly the coherence of physical quantities in initial field,which might make the model instability. Nevertheless,this study may provide a new thought to model revision techniques.