华南前汛期降水预测模型及其预测试验

PREDICTION MODEL FOR ANNUALLY FIRST RAINY SEASON PRECIPITATION IN SOUTH CHINA AND PREDICTION TESTS

将中国华南区域分为东、西2个区,对每个区(8个站)的前汛期(4—6月)平均降水量作自然正交展开(EOF),选取各区累积方差贡献超过75%的前4个主分量作为预报分量。再利用偏最小二乘回归方法结合均生函数方法,提出一种同时考虑预报量自身显著变化周期和前期物理量因子对预报量未来变化影响的预报模型,分别建立东、西区前汛期平均降水量的偏最小二乘回归预报方程。试验结果表明,新的预报模型的预报效果比单纯采用前期物理量因子的逐步回归模型更好,并且其预报能力的提高具有合理的分析依据。

Mean precipitation during the annually first rainy season（April-June） of South China（divided into an east and west region with eight stations in each） was studied using the Empirical Orthogonal Function（EOF） method,and the first four principal components in each region,accounting for a total of more than 75% of the variability,were used as the predictands.A prediction model,which involves the significant period of a predictand and physical factors influencing the future changes of the predictand,was developed based on the partial least-squares regression method（PLS） in combination with mean generation function,establishing PLS equations for the east and west regions.Results showed that the new model is superior in prediction accuracy over the stepwise regression method,which uses physical factors only,and the improvement has reasonable analysis basis.