选取1958～2000年的74项环流因子作为预测因子,用逐步回归法对因子进行初筛选,采用最优子集回归方法建立预测方程,AIC准则作为最优方程的判别准则,分别对入梅日期、出梅日期、梅雨量3个梅雨特征量进行预测,并利用2001 ～2012年的资料对方程的预测效果进行检验.结果表明,预测方程对梅雨偏多偏少(旱涝趋势)的预测比较准确,但对降水异常偏多(少)的年份数值误差较大；方程对入梅日期预测效果较好,大部分年份误差在3～4d以内；方程对出梅日期预测大部分年份误差在7d或以上.%Selecting 74 circulation factors from 1958 to 2000 as the predictive factors, after the initial screening of the factors by stepwise regression method, using the optimal subset regression method to establish the prediction equation, AIC criterion as the criterion of the optimal equation, three characteristics of Meiyu including onest and ending date of Meiyu and Meiyu precipitation were forecasted. And the prediction test of equation was detected by using data during 2001 -2002. The results showed that; (1) The prediction result of the equation for the more(less) Meiyu is comparatively accurate, but for more (less) precipitation is not accurate. (2) The prediction result of the equation for the onest date of Meiyu is more accurate, the error of most years is within 3 to 4 days. (3 ) The prediction result of the equation for the ending date of Meiyu is not very ideal, the error of most years is 7days or more.
展开▼
