摘要
采用多数表决集成法、评分权重集成法、多元回归集成法对承德市西来系统形成的降水进行晴雨和分级集成预报试验。结果表明:在每种单一的数值预报方法TS值均较高,总体预报质量相差不大的情况下,多数表决集中集成法可大幅度提高预报准确率;如果几种数值预报对于降水分级预报的预报准确率相差较大时,可采用评分权重集成法来提高总体预报质量;在样本足够的条件下,细化的预报结果多元回归集成法是晴雨预报和分级降水预报中比较可行的集成方法之一。
The graded precipitation and rainy/fine forecasting experiments are conducted by means of several consensus methods such as majority voting,weighted scoring,and multivariable regression on precipitation caused by weather systems from west in Chengde.The results indicate that when there is relative high individual TS s and slightly varied total forecasting performance,the majority voting method can improve the forecasting accuracy greatly;when the accuracies of various numerical forecasting methods for grade...
出处
《气象科技》
2008年第5期520-523,共4页
Meteorological Science and Technology
关键词
集成方法
降水预报
预报准确率
consensus method
precipitation forecasting
forecasting accuracy
