深圳对流尺度集合预报系统对台风降水预报的检验评估

目前数值模式对台风降水预报的准确率仍有待提高。为了评估深圳对流尺度集合预报系统对台风降水预报能力，选取了2015—2018年共14个影响广东台风个例，利用广东省2 300多个自动气象观测站的24小时累计降水观测资料，检验该系统的集合预报方法(含集合平均方法和概率匹配平均方法)和控制预报方法的24小时降水预报结果。(1)系统对台风24小时降水预报具有较好参考价值，三种方法的暴雨等级预报TS评分均达到0.39以上。(2)集合预报方法总体上优于控制预报方法，可改善珠江口两侧暴雨中心降水预报。其中集合平均方法总体预报效果最好，其降水预报均方根误差为38.1 mm，比控制预报方法减少18.8%，对暴雨等级预报TS评分为0.469比控制，预报方法提升20.1%，但是对特大暴雨等级预报能力不足；而概率匹配平均方法改善了小雨和特大暴雨的预报能力。(3)系统对较强台风的降水预报能力优于弱台风。在较强台风情形下，系统对粤东暴雨中心降水预报明显偏小且控制预报方法偏差最大，其他地方降水预报偏大为主；在弱台风情形下，系统对降水预报存在明显系统性偏大，但对粤西暴雨中心降水预报明显偏小且控制预报偏差最大。 

VERIFICATION AND EVALUATION OF TYPHOON PRECIPITATION FORECAST BY SHENZHEN STORM-SCALE ENSEMBLE FORECAST SYSTEM

At present, the accuracy of typhoon precipitation prediction by numerical models still needs to be improved. To evaluate the ability of typhoon precipitation forecast by the high resolution (4km) storm-scale ensemble forecast system in Shenzhen, 14 typhoons that affected Guangdong Province from 2015 to 2018 were selected. Based on the 24-hour accumulative precipitation observational data from more than 2300 automatic observation stations in Guangdong, the 24-hour precipitation forecast results of the system by using ensemble forecast method, including ensemble average method and probability matching average method, and the control forecast method were verified. The main conclusions are as follows: (1) The system can provide a good reference for typhoon 24-hour precipitation forecast, and the TS scores of rainstorm grade forecast by using the three methods are all above 0.39. (2) The ensemble forecast method is better than the control forecast method as the former can improve the rainfall forecast of the rainstorm center on both sides of the Pearl River estuary. Besides, the ensemble average method has the best overall prediction. Its root mean square error of precipitation prediction is 38.1mm, which is 18.8% less than that of the control method; the TS score of rainstorm grade prediction is 0.469, which is 20.1% higher than that of the control method. Its ability to predict extraordinary rainstorm grade is insufficient and the probability matching mean method improves its ability to predict light rain and extraordinary rainstorm. (3) The precipitation forecast ability of the system for strong typhoons is better than that for weak typhoons. In the cases of strong typhoons, there is an obvious negative deviation between the precipitation forecast for and observations in the rainstorm center in the east of Guangdong, and the deviation of the control method is the largest; in other places, the deviation is mainly positive. In the cases of weak typhoons, the precipitation forecast by the system has an obvious systematic positive deviation, but there is an obvious negative deviation for the rainstorm center in the west of Guangdong, and the deviation of the control method is the largest.