降维方法在广西前汛期降水预测中的应用

对广西88个站前汛期(4～6月)降水距平百分率作自然正交展开(EOF分解),选取累积方差贡献接近70%的前5个主成分为预报量.从前期大气环流场和海温场中分别查找预报因子,进一步对初选预报因子作降维处理,分别选取其中与5个预报分量相关程度高的主成分建立5个预报方程.经独立样本预报试验证明,该方法对广西前汛期降水趋势具有较好的预测能力.

The Application of EOF Method to the Precipitation Forecast During the First Raining Season in Guangxi

By way of making empirical orthogonal function (EOF) with the precipitation percent departures during the first raining season (April to June) at 88 stations in Guangxi, the first 5 principal components with total ratio contribution round 70% are used as predictands. Predictors chosen from former atmospheric circulation and sea surface temperature fields are contracted as synthesized ones by using EOF decomposition. The synthesized predictors which are closely related to the predictands mentioned above are selected to build 5 prediction equations respectively. To estimate the effect of the statistical prediction model, individual sample tests are made. The result shows that the modal performs well on the precipitation trend forecasting during the first raining season in Guangxi.