基于多模式的巢湖降水集成预报检验及应用
Preliminary Study of Precipitation Ensemble Prediction Based on Multimodel in Chaohu Lake
中文摘要：
利用2014年2月—2015年9月中尺度模式预报（INCA、WRF)、全球模式（ECMWF、JMA）数据，结合巢湖湖泊周边区域站降水实况，应用自适应最小二乘集成法（简称统计集成）和加权平均集成法（简称加权集成）开展多模式集成预报试验，并对统计集成和加权集成效果进行检验与分析。研究表明：1）2种集成预报比单一模式预报的误差明显降低；2）在6～24 h时效内统计集成误差比加权集成误差大，在48～72 h时效内低于加权集成；3）在小雨级别，统计集成TS评分最高，中雨以上级别，统计集成TS评分低于加权集成，加权集成正确率始终高于统计集成。
中文关键词：
多模式，加权集成，统计集成，降水，巢湖
KeyWords：
multi-model, weighted average, adaptive least square, rainfall, Chaohu Lake
Abstract：
Based on the data from INCA, WRF(mesoscale model), ECMWF, JMA (Global model) and observed data around Chaohu Lake, multimodel test equations are established to access the following two methods: the Adaptive least square statistical ensemble, method and the weighted average method. Results show: 1) The error of forecasts from the 2 integrated forecasts is significantly small than that from the single model. 2) The error resulted by the Adaptive least square method is larger than that by the weighted average method in 6-24 h term, but is small in 48-72 h term. 3) For the light rain grade the, TS score of the Adaptive least square method was the highest. For above moderate rain grads, the TS score of the weighted average method is higher than other one. In addition, the correct rate of the weighted average method is always higher than that of the Adaptive least square method.