基于支持向量机和FY静止卫星云参数的降水预测技术  

Precipitation Forecasting Technique Based on Support Vector Machine and FY Geosynchronous Satellite Cloud Parameters

作　　者：刘建朝[1,2] 周毓荃[2] 

机构地区：[1]南京信息工程大学大气物理学院,南京210044 [2]中国气象科学研究院,北京100081

出　　处：《气象与环境科学》2011年第2期59-63,共5页Meteorological and Environmental Sciences

基　　金：气象关键技术集成与应用项目(CMAGJ2011M71)资助

摘　　要：将支持向量机分类方法应用于降水与非降水的分类预测。选用FY静止卫星反演的云光学厚度、云顶温度、云顶高度、云有效粒子半径作为特征分量,以Micaps 1 h雨量资料作为是否降水的类别标签,建立预测降水与非降水的分类模型,针对安徽地区2008年5-11月资料进行分析。结果显示,降水类的预测准确率在40%～60%,非降水类的预测准确率在90%以上。在分类模型中多加入微波辐射计观测的液水路径特征分量后,所得预测结果对降水类的预测准确率有所提高。通过完善和改进分类模型,预测分类的准确率有望进一步提高,该方法对降水预测有很好的业务应用前景。The classification forecasting about precipitation and nonprecipitation is based on SVM.The feature components are cloud optical thickness,cloud top temperature,cloud top height and cloud effective particle radius retrieved by FY geosynchronous satellite.Taking the Micaps rainfall data of an hour to be a class label,we make a classification model about precipitation and nonprecipitation to analyze Anhui data between May and November in 2008.Results show that the accuracy rate of precipitation category is about 40%～60% and nonprecipitation is more than 90%.When the liquid water path from microwave radiometer as a kind of feature component is put in model,the forecasting accuracy rate of precipitation category will be a little raising.If you can continue to improve the model and increase the prediction model for the classification accuracy,this method will have certain business application prospect.

关 键 词：支持向量机 卫星云参数 降水分类预测 

分 类 号：P456.1[天文地球—大气科学及气象学] P412.27