多雷达反演参量联合的短时强降水识别方法研究

Research of Very Short-Term Heavy Precipitation Identification Method Combinated Multi-Radar Retired Parameters

借鉴美国ANC(NCAR Auto-Nowcast System)基于模糊逻辑原理的临近预报技术思路,采用SWAN(Severe Weather Automatic Nowcast System)系统中雷达基本产品及其反演产品,研发了多参量联合的短时强降水落区识别算法,重点研究因子选取、误差分析及参数设置。基于历史资料分析结果筛选出1h定量降水预报QPF(Quantity Precipitation Forecast)、雨强RI(Rain Intensity)、垂直累积液态水含量VIL(Vertically Integrated Liquid Water)、回波顶高ET(Echo Top)预报量作为识别因子,并设计出相应的隶属函数。结果表明,多因子联合识别短时强降水算法的CSI(Critical Success Index)指数为0.18,性能高于QPF判别法及仅使用单因子识别的模糊逻辑方法。强降水指数HRI(Heavy RainIndex)阈值设置测试表明,选择合适的HRI阈值有助于发挥算法性能,统计结果表明取0.5～0.6比较合理。同时,较高的空报率是影响算法性能的主要因素,包括地物杂波、因子场定常外推假设、不合适的Z-R关系、局地性强降水以及多种因素的共同作用。

Based on the Severe Weather Automatic Nowcast System radar products,a very short-range heavy precipitation identification algorithm which inherits nowcasting method in NCAR Auto-Nowcast System and multi-radar parameters and fuzzy logic method is proposed.The research has been done on predictor chosen,error analysis and parameter threshold set.Identification factors such as quantity precipitation forecast,rain intensity,vertically integrated liquid water,echo top are chosen according to history data statistic analysis and the corresponding member functions are designed.Evaluation shows that the critical success index of very short-term heavy precipitation identification algorithm combined with multi predictors is 0.18,and its performance is higher than that of QPF determination method and fuzzy logic method using the single predictor.Test of heavy rain index threshold shows that the appropriate threshold is helpful to improve algorithm performance and it is assigned as 0.5~0.6 in this research.The main cause affecting the algorithm performance is high false alarm rate which is caused by clutter echo,const-extrapolation prediction assumption,inappropriate Z-R relationship and not uniformity on heavy precipitation location.In general,more than one factor cause together false alarm due to rainfall complexity.