影响广西的热带气旋年频数的BP神经网络预测模型  

A BP NEURAL NETWORK PREDICTION MODEL FOR THE ANNUAL FREQUENCY OF TROPICAL CYCLONES AFFECTING GUANGXI

作　　者：何慧[1] 欧艺[1] 李艳兰[1] 

机构地区：[1]广西壮族自治区气候中心,广西南宁530022

出　　处：《热带气象学报》2009年第4期407-412,共6页Journal of Tropical Meteorology

基　　金：广西自然科学基金(桂科自0728074);中国气象局广州区域气象中心气象科技攻关重点项目(GRMC2007Z02)共同资助

摘　　要：对影响广西的热带气旋(TC)年频数与大气环流的关系进行分析表明,TC年频数与全球范围大气环流异常有密切关系,特别是春季南半球中高纬度环流异常和低纬越赤道气流异常。利用相关分析从春季全球大气环流场中选择初选预报因子,然后对初选预报因子作EOF展开构造综合预报因子,运用BP神经网络方法建立TC年频数预报模型,并对所建立的模型进行独立样本试验。结果表明,该预报模型对历史样本拟合精度高,试报效果优于传统的逐步回归模型,可在汛期预测业务中应用。The relationship between atmospheric circulation and the annual frequency of tropical cyclones （TC） affecting Guangxi is analyzed. The results show that the annual TC number is closely correlated with the anomalous global atmospheric circulation. Especially, the mid-and higher-latitude anomalous atmospheric circulation in the Southern Hemisphere and the anomalous cross-equatorial flow in spring are very important for the annual TC number. Based on the correlation analysis and EOF decomposition method, comprehensive predictors are constructed from those preliminarily selected in the springtime global atmospheric circulation field. A BP neural network prediction method is used to create an annual TC number prediction model. The model is tested with independent samples. The result shows that this prediction model is significantly better than traditional regression models and has a good prospect of application in routine flood-season forecasts.

关 键 词：TC年频数 EOF展开 综合预报因子 BP神经网络 

分 类 号：P444[天文地球—大气科学及气象学]