统计学降尺度方法是国内外研究全球气候模型尺度降解的热点问题。研究和探讨了基于光滑支持向量机的统计学降尺度方法；建立大尺度气候观测资料和实测降水之间的统计关系；模拟和预测汉江流域降水变化，并同传统的多元线性回归分析方法相比较。结果表明，基于光滑支持向量机的统计学降尺度方法的模拟精度不仅高于多元线性回归分析方法，而且明显优于CGCM2气候模型的输出降水结果。

The statistical downscaling method is a hot topic of the downscaling to the GCMs(global climate models). To establish the statistical relationship between the larger scale climate predicted data and observed precipitations in Hanjiang River basin, a statistical downscaling method based on smooth support vector machine(SVM) was discussed and studied. The results showed the SSVM is superior to the traditional multi linear regression method(MLR) and the precision of forecasted precipitations by using SSVM is much higher than the directly output precipitation of the CGCM2. The SSVM is suitable for conducting climate impact studies.