基于LS-SVM的改进统计降尺度方法

A Statistical Downscaling Method Based on Least Squares Support Vector Machines

摘要: 统计降尺度方法作为一种计算量小、使用灵活的降尺度模型，被越来越多应用到气候变化研究当中。本文以湘江流域为例，开发了一种基于LS-SVM回归的改进统计降尺度算法，并与经典统计降尺度模型SDSM (Statistical Downscaling Model)进行比较。结果表明，在湘江流域，无论是降水模拟和温度模拟，基于LS-SVM回归算法的改进统计降尺度方法都能达到SDSM的效果，而温度的模拟，LS-SVM回归降尺度方法模拟结果更好。为了使得这种方法能更适合气候变化对水资源的影响研究，还需要在更多的区域进行应用证明。

Abstract:  The statistical downscaling method has been more and more utilized in the climate change study for its simplicity and flexibility. A statistical downscaling method based on LS-SVM (least squares support vector machines) was developed and compared with SDSM (Statistical Downscaling Model) to test its ability in downscaling precipitation and temperature in Xiangjiang Basin. The results showed that the method based on LS-SVM has the similar performance with the SDSM method in simulating precipitation, while it was superior to SDSM in simulating temperature. The proposed method still needs to be applied to more regions to make it more suitable for studying the impact on water resources under climate change.