逐段线性化判别方法及其在短期气侯预测中的应用

THE PHASE BY PHASE LINEAR DISCRIMINANT METHOD AND ITSAPPLICATION IN SHORT TERM CLIMATE FORECAST

针对经典判别方法中存在的不足,将半截多项式引入到判别函数中,提出了逐段线性化判别方法及其一整套求解方案。此后,将此新方法引入到短期气候预测中,对云南昆明和蒙自两站的5月雨量的初步试用结果为:在y*＝y 的条件下昆明和蒙自的历史预报准确率分别为 85．7%和 82．9%,外推预报准确率均为 6/7;在 y*＝y+40%y条件下的历史预报准确率分别为 77．1%和 71．4%,外推预报准确率分别为7/7和 5/7;在 y*＝y－40%y条件下的历史预报准确率分别为80．0%和74．3%,外推预报准确率分别为5/7和7/7。表明逐段线性化判别方法有良好的历史预报准确率和外推预报准确率。

To overcome the deficiency of classic dicriminant method, the half polynomial was introduced in the discriminant,and the phase by phase linear discriminant method was developed. To test the new method, the May rainfall of Kunming and Mengzi was forecasted with the new method. The fitting accuracy of Kunming andMengzi are 85. 7 % and 82. 9 % for y = y , but 77. 1 % and 71.4% for y = y + 40 %y , and 80. 0 % and 74. 3 % but for y* = y - 40 %y respectively. The forecasting accuracyof Kunming and Mengzi are 6/7 and 6/7 for y* = y , 7/7 and 7/5 but for y* = y + 40 %y , and 5/7 and 7/7 but for y* = y - 40 %y respectively. The testingresults indicated that the phase by phase discriminant method had excellent fitting and forecasting ability.