SVM方法在降水预报中的应用及改进

Application and Improvement of SVM Method in Precipitation Forecast

以T213数值模式输出产品为基础,结合常规观测的降水资料,利用SVM方法,进行了大量多因子的随机交叉验证,从而选出最优参数,建立了全国72个站点的降水预报模型,并用独立的样本对预报模型进行了检验。再通过计算正、负样本的贴近度来分析预报因子,实现了预报因子的筛选和降水预报模型的改进;检验结果表明:改进后的降水模型的预报结果优于改进前的。实时业务试运行的结果也显示SVM模型的降水预报效果好于T213模式直接输出的降水预报。

Based on T213 NWP （Numerical Weather Prediction）model outputs and precipitation observations, cross-validation is performed with random samples to find the samples with best predictors and optimal parameters. The forecast models of precipitation are established at 72 meteorological stations in China by the SVM （Support Vector Machine） statistical method. The models are verified with independent samples. The predictors are selected and the precipitation forecast models are improved by pressing close degree. Forecast experiments show that the improved models are better. The precipitation forecasted by SVM models is superior to the precipitation of T213 DMO （direct model output） in real - time experiments.