基于SVM分类的淮河流域夏季降水预测模型  

Prediction model for summer precipitation in the huaihe river basin based on support vector machine

作　　者：吴有训[1] 刘勇[2] 叶金印[3] 余品忠 

机构地区：[1]安徽省宣城市气象局,安徽宣城242000 [2]安徽省气象局,合肥230061 [3]淮河流域气象中心,安徽蚌埠233040

出　　处：《中国农业大学学报》2011年第5期157-162,共6页Journal of China Agricultural University

基　　金：中国气象局气象新技术推广项目资助(CMATG2005M34)

摘　　要：采用1959--2009年逐月74项大气环流特征量序列、500hPa月平均高度场和月平均海温场，计算与预报对象淮河流域夏季降水量相关系数，选取预测因子；用主分量分析方法组合预测因子。用支持向量分类机方法分别建立山东淮河流域、河南淮河流域、江苏淮河流域、安徽淮河流域共4个区域夏季降水短期气候预测模型。对2007--2009年夏季降水量SVM分类预测，4个区域的训练集回预测正确率为85％～99％，平均训练集回预测正确率91％；预测结果误差最大不超过1级，绝对值平均为0．4级。结果表明，该模型具有较强的预测能力和推广前景，可在气候预测业务中使用。Based on Support Vector Machine （SVM）, four short-range summer precipitation prediction models were established for four areas in the Huaihe River basin, respectively. Using the monthly data of 74 circumfluent eigen values,the monthly data of sea surface temperature, the monthly data of 500 hPa height from 1959 to 2009, forecast factors were chosen. Combination of the forcast factors was done by using Principal component analysis. A categorical prediction was performed on summer precipitation data from 2007 to 2009. The results show that the accuracy of four regional training set to predict is 85%-99% and the mean accuracy is 91% ,and the lagest grade of error of summer precipitation prediction is no more than 1, and the mean absolute grade is 0.4. These results indicate that the short- range climatic prediction model based on Support Vector machine has a good performance on predictions of summer precipitation.

关 键 词：支持向量分类机(SVM) 淮河流域 夏季降水 短期气候预测 

分 类 号：P457.6[天文地球—大气科学及气象学]