地理研究

地理研究 ›› 2009, Vol. 28 ›› Issue (5): 1243-1254.doi: 10.11821/yj2009050011

• 气候与全球变化 • 上一篇    下一篇

基于主成分神经网络的台风灾害经济损失评估

娄伟平1, 陈海燕2, 郑 峰3, 吴 睿1   

  1. 1. 新昌县气象局,浙江 新昌 312500;
    2. 浙江省气象台,杭州 310000;
    3. 温州市气象局,温州 325027
  • 收稿日期:2008-10-21 修回日期:2009-05-15 出版日期:2009-09-25 发布日期:2009-09-25
  • 作者简介:娄伟平(1970-),男,浙江新昌人,高级工程师。研究方向为气象灾害风险评估。 E-mail:xclwp@163.com
  • 基金资助:

    国家气象局新技术推广项目(CMATG2008M40)

Economic loss assessment of typhoon based on principal component analysis and neural network

LOU Wei-ping1, CHEN Hai-yan2, ZHENG Feng3, WU Rui1   

  1. 1. Xinchang County Meteorological Bureau, Xinchang 312500, Zhejiang, China;
    2. Zhejiang Provincial Meteorological Office, Hangzhou 310017, China;
    3. Wenzhou City Meteorological Bureau, Wenzhou 312500, Zhejiang, China
  • Received:2008-10-21 Revised:2009-05-15 Online:2009-09-25 Published:2009-09-25
  • Supported by:

    国家气象局新技术推广项目(CMATG2008M40)

PDF (PC)

赞

被引次数

47 | 17

摘要：

本研究建立了浙江省台风灾害直接经济损失评估模型。把浙江省台风灾害直接经济损失资料换算成直接经济损失指数,运用主成分分析法对表示致灾因子、孕灾环境与承灾体的评估因子进行数据处理,提取主成分作为BP神经网络模型的输入,从而建立评估模型。模型历史拟合结果和实际一致。在2007年和2008年影响浙江省的5个台风的实际评估中,强台风"Vipa"灾后评估值比实际值偏大2.16,其余4个台风灾后评估值比实况偏大0.2~0.7,反映了人们对影响大的台风防灾减灾工作的重视和防灾减灾效果。根据台风开始影响时过程风雨预报值进行预评估,过程风雨预报值较准确的台风,预评估结果和灾后评估值一致;过程风雨预报值误差较大的台风,预评估效果较差。因此,该模型可用于实际台风灾害直接经济损失评估,提高台风影响前风雨预报准确率是提高预评估准确率的关键。

关键词: 台风, 直接经济损失, 主成分分析, BP神经网络, 评估

Abstract:

The assessment model of direct economic losses from typhoon disaster in Zhejiang Province is established in this research. The data of direct economic losses in the study region are converted into direct economic losses indexes. Using principal component analysis method, the assessment factors representing disaster inducing factor, disaster-formative environment and disaster-affected body are processed, and the principal component is abstracted as the input of the BP neural network model, thus the assessment model is established. Historical fitting results are consistent with the reality. It is found in the actual assessments of five typhoons affecting Zhejiang in 2007 and 2008 that the post-disaster assessment values of typhoons are higher than the actual situations, and the severer impacts the storms have, the narrower the gap between the assessment values and the actual situation is, which reflects the impact of the disaster prevention and alleviation efforts against typhoons of great influence. According to the forecast values of wind and precipitation when the typhoon began to exert some affect, pre-assessments are conducted and the consequence shows that the pre-assessment results with relatively accurate forecast values are in accordance with the post-disaster assessment values, while the ones with less accurate forecast values are unsatisfactory. Therefore, this model can be applied in the actual assessment of direct economic loss from typhoon damage, and the accurate forecast of wind and precipitation before the typhoons have effect is crucial to the improvement of the accuracy of pre-assessments.

Key words: typhoon, direct economic losses, principal component analysis, BP neural networks, assessment