集合动力因子在内蒙古暴雨预报中的适用性分析

Applicability Analysis of the Aggregate Dynamic Parameters in Rainstorm Forecast in Inner Mongolia

针对2017年7月6—7日内蒙古东部地区的冷涡降水过程,利用NCEP的GFS预报资料为初始值得到的集合动力因子预报系统结果,从中优选出10个动力因子以及对其进行4种方法的集合,以此对强降水过程进行诊断分析和预报研究。结果表明:动力因子预报降水在空间分布上与6h观测降水基本一致,代表动力因子对降水落区具有较好的指示意义,中位数和等权平均集合预报降水的强度比观测降水偏弱,而多元线性回归和加权平均集合动力因子对强降水中心的预报相对较好。预报技巧评分计算结果表明,集合动力因子对降水预报均有一定的技巧,而多元线性回归集合结果与观测降水最为吻合,空间相关系数达到0.76,ETS评分为0.49,加权平均次之,但多元线性回归集合方法预报的降水发生频率会较实际降水偏高。

According to the cold eddy precipitation process in the eastern Inner Mongolia from July 6 to 7, 2017, the GFS forecast data of NCEP as initial value was used to gain the result of the set dynamic factor prediction system. From this, 10 dynamic factors are selected and a set of 4 methods were used to diagnose and predict the precipitation process. The main conclusions are as follows. The dynamic factor forecast precipitation is basically consistent with the 6h observation precipitation in spatial distribution, which means that the dynamic factor has a good indication to the precipitation area. The median and equal weighted average ensembles predict that the intensity of precipitation is weaker than that of observed precipitation, while the multiple linear regression and weighted average aggregate dynamic factors are relatively good for heavy precipitation centers. The calculation results of forecasting skill scores show that the aggregate dynamic factor has certain skills for precipitation forecasting, and the multivariate linear regression set results are most consistent with the observed precipitation. The spatial correlation coeffcient is 0.76, the ETS score is 0.49, the weighted average is second. However, the frequency of precipitation predicted by the multiple linear regression set method will be higher than the actual precipitation.