SVM方法在武汉市大雾预警预报中的应用  

Application of SVM Method in Warning and Forecasting of heavy fog in Wuhan City

作　　者：李才媛[1] 韦惠红[1] 邓红[1] 

机构地区：[1]武汉中心气象台,武汉430074

出　　处：《暴雨灾害》2008年第3期264-267,共4页Torrential Rain and Disasters

基　　金：湖北省科技攻关项目(2006AA301B45)资助

摘　　要：在简单介绍支持向量机(SVM)基本原理的基础上,根据2003—2006年武汉市探空和地面气象要素资料,使用支持向量机方法(SVM)和CMSVM应用软件平台,通过对训练样本进行交叉验证和模型核参数的逐渐逼近,建立了武汉市0～24 h大雾支持向量机预报模型。经过2007年试报,该模型的TS评分为62.5%;同时,通过对2007年武汉市大雾特点和预警信号发布情况的分析发现,该模型所选取的建模时段和设计的输出产品时间均比较合理,能够较及时地为大雾预警预报提供参考依据。Based on the fundamental rules of the Support Vector Machine （SVM） and the conventional observation data of Wuhan from 2003 to 2006, the cross-checking of train samples and the gradual approaching of core parameters of model are made to establish the 0-24 hour＇s forecasting model of the heavy fog of Wuhan city with SVM method and the application software platform of Chinese Meteorological Support Vector Machine （CMSVM）. The Ts score of this model is 62.5% during its initial forecasting in 2007. The features and warning information issuing of the heavy fog in Wuhan in 2007 are analyzed to show that both the model-building date design and the time of output are proved to be reliable to provide the heavy fog warning and forecasting in good time.

关 键 词：武汉市 大雾 预警预报 支持向量机 

分 类 号：P456[天文地球—大气科学及气象学]