摘要
本文采用城镇精细化气象要素预报检验方法,对多种数值预报及释用产品的2014年度日极端气温、24小时降水、暴雨过程暴雨和8种分类暴雨过程暴雨等预报进行站点预报质量检验。结果表明:释用T639数值预报的T639-MOS日极端气温预报具有较高的参考价值,5天逐日的最高气温和3天逐日的最低气温≤2℃预报准确率均为最高;中尺度数值模式的晴雨和一般性降水预报优于全球模式,但暴雨预报评分不如全球模式和多模式集成降水预报;暴雨预报Ts评分都不高,且随预报时效降低,24小时预报最高为17.3%,而120小时预报最高仅3.4%;暴雨过程暴雨预报平均TS评分很低,均未达10%;分类暴雨过程以台风登陆类暴雨预报TS评分最高,但该类暴雨过程样本数太少,检验结果代表性差;其他分类暴雨预报效果差,对局地对流降水明显的午后热雷雨类和副高边缘类暴雨全部漏报。
By comparing the grid-to-point forecasts of daily extreme 2-meter temperature and daily precipitation with the observations, the performance of the numerical weather prediction (NWP) models and o the ensemble prediction systems (EPSs) in 2014 was evaluated. Several objective forecasts such as the T639-MOS forecasts and the consensus forecasts of multimodels were verified as well. It was shown that short-range forecasts of the T639-MOS daily extreme temperature were reliable, which were the best among the temperature forecasts evaluated. With respect to the torrential rain forecasts, the global models and the consensus forecasting were no better than the meso-scale models. The process-average threshold scores were all less than 0.1. When the torrential rains in 2014 were classified into eight categories according to their major mechanism, the models and the consensus forecasting were found to be good at predicting the torrential rains induced by a landing typhoon, while all of the models failed to forecast the torrential rains triggered by local thunderstorm.
关键词
预报检验
数值预报
释用产品
分类暴雨
forecast verification, numerical prediction, interpretation product, classified rainstorm