主办单位：中国气象局沈阳大气环境研究所
国际刊号：ISSN 1673-503X
国内刊号：CN 21-1531/P
气象与环境学报

气象与环境学报 ›› 2016, Vol. 32 ›› Issue (5): 147-153.doi: 10.3969/j.issn.1673-503X.2016.05.020

• 论文 • 上一篇    下一篇

基于Logistic回归蚌埠地区致灾雷电预报研究

周丽雅1, 王凯2, 刘倪3   

  1. 1. 蚌埠市气象局, 安徽 蚌埠 233000;
    2. 安徽省防雷中心, 安徽 合肥 230061;
    3. 安徽省气象台, 安徽 合肥 230061
  • 收稿日期:2015-08-31 修回日期:2015-12-31 出版日期:2016-10-31 发布日期:2016-10-31
  • 作者简介:周丽雅,女,1981年生,工程师,主要从事雷电防护研究,E-mail:823827447@qq.com。
  • 基金资助:
    蚌埠市气象局项目“蚌埠地区闪电定位资料的分析处理及应用”（BQKY201506）和安徽省气象局项目“安徽省防雷业务综合应用平台”（KM201410）共同资助。

Forecasting study of catastrophic lightning based on Logistic Regression in Bengbu region

ZHOU Li-ya1, WANG Kai2, LIU Ni3   

  1. 1. Bengbu Meteorological Service, Bengbu 233000, China;
    2. Anhui Lightning Protection Center, Hefei 230061, China;
    3. Anhui Meteorological Observatory, Hefei 230061, China
  • Received:2015-08-31 Revised:2015-12-31 Online:2016-10-31 Published:2016-10-31

PDF (PC)

452

被引次数

4

摘要： 利用2007-2013年安徽省闪电定位资料和蚌埠地区雷电灾害资料，选取安徽省蚌埠地区36个致灾雷电样本作为建模样本，采用全球预报系统GFS（Global Forecasting System）模式的强对流参数作为预报因子，利用Logistic回归法建立了蚌埠地区4种类型致灾雷电的预报模型。结果表明：CAPE、LFTX和PW共3个参数可作为蚌埠地区致灾雷电预报的指标；建立的4种类型致灾雷电模型的预报效果较好，其中08-14时雷电预报模型的总体预报准确率最高，可达83.00%。当雷电预报模型输出概率偏高且超过0.7时，4种致灾雷电预报模型的预报技巧评分最高，且预报准确率最高，失误率和虚警率较低，预报效果较好。夏季致灾雷电预报模型的整体预报效果较好，平均预报准确率达82.20%，特别是在雷电发生的密集地区，预报出了雷电的高概率分布。

关键词: 致灾雷电, Logistic回归, 强对流参数, 预报

Abstract: Based on data from the lightning positioning system in Anhui province and lightning disaster in Bengbu region from 2007 to 2013,four kinds of catastrophic lightning forecasting models were established using the Logistic regression method.36 catastrophic lightning cases were selected as modeling samples,and the strong convective parameters from the GFS (Global Forecast System) model were used as forecasting factors.The results show that three parameters,i.e.,CAPE (Convective Available Potential Energy),LFTX (surface LiFTed indeX) and PW (Precipitable Water),can be used as the indexes for catastrophic lightning forecasting in the Bengbu region.These catastrophic lightning forecasting models perform well in forecasting,in which the overall forecasting accuracy of the 08:00-14:00 model is the best,reaching 83%.When the output probability of the forecasting models is higher than 0.7,these models receive the highest skill score and accuracy,as well as lower error and false alarm probability.In general,the summer forecasting model has the best forecasting effect with an average forecasting accuracy of 82.2%.At the dense zones where the lightning occurs,this model can forecast the high probability distribution of lightning.

Key words: Catastrophic lightning, Logistic regression, Strong convective parameter, Forecasting

中图分类号: