五种短期气候预测统计模型的研制

A Study on Five-Statistical Models Of Short_Range Climatic Prediction

根据“九五”贵州短期气候预测研究课题总体安排 ,研制了一套短期气候预测模型 ,共建立了 5种统计模型 :逐步回归预报模型、多因子综合相关相似预报模型、50 0hPa环流相似预报模型、北太平洋海温相似预报模型、贵州分区RT回归预报模型。各预报模型的建模样本年限为 1957～ 1996年 ,独立试报期为 1997年及其以后年份。每一个预报模型均可在每月作出其后 6个月的逐月雨量和平均气温预报、其后 3个季度的季雨量及气温预报和气候灾害指数预报。经 1997～ 2 0 0 0年试报表明 。

In this paper 5_short_range climatic prediction models are established according to“95”scientific research item of short_range climatic prediction in Guizhou.These models consist of stepwise regression method,multitudinous factor correlation and similarity method,similarity method of atmospheric circulation in 500hPa,similarity method of SST in the northern pacific,multivariate regression method of divisional region RT factor in Guizhou. The models are established for historical sample 1957～1996. are tested for an independent sample of 1997～2000.Everly model monthly predict later precipitation and average temperature from 2 to 6 months,later precipitation and average temperature and climatic disaster index from 1to 3 seasons.The result shows that forecast accuracy(whether tends or skill)of the models can improve obviously compared with previously issued operational forecast of precipitation and average temperature monthly.