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 ›› 2023, Vol. 39 ›› Issue (5): 106-112.doi: 10.3969/j.issn.1673-503X.2023.05.013

• 快报 • 上一篇    

基于BP神经网络的地基微波辐射计雷暴预报试验研究

蒋维东(),张荣智,陈博,张天旗,黄海玲,周林   

  1. 中国民用航空华东地区空中交通管理局,上海 200335
  • 收稿日期:2022-03-10 出版日期:2023-10-28 发布日期:2023-11-28
  • 作者简介:蒋维东, 男, 1989年生, 工程师, 主要从事航空气象及短临预报预警研究, E-mail: jiangdd_1024@163.com
  • 基金资助:
    上海市2020年度“科技创新行动计划”社会发展科技攻关项目“针对极端天气事件的长三角航空安全运行应对支持平台的开发与应用”(20dz1200703);华东空管局创新实验室

Experimental study on thunderstorm forecast of ground-based microwave radiometer based on BP neural network

Weidong JIANG(),Rongzhi ZHANG,Bo CHEN,Tianqi ZHANG,Hailing HUANG,Lin ZHOU   

  1. The East China Regional Air Traffic Management Bureau under the Civil Aviation Administration of China (CACC), Shanghai 200335, China
  • Received:2022-03-10 Online:2023-10-28 Published:2023-11-28

摘要:

利用上海浦东国际机场地面气象观测站2018—2019年MWP967KV型地基微波辐射计资料、民航气象地面观测数据及宝山站常规探空资料, 分析了微波辐射计探测数据的可靠性。在此基础上, 计算与雷暴发生相关的大气参数, 选取合适的参数作为预报因子, 建立了机场雷暴预报的BP(Back Propagation)人工神经网络模型, 并对模型的预报效果进行了评估。结果表明: 由MWP967KV型地基微波辐射计得到的温度、相对湿度和水汽密度与探空相应数据的平均绝对偏差分别为1.94 ℃、16.05%、0.82 g·m-3, 均方根误差分别为1.41 ℃、20.14%、1.90 g·m-3, 相关系数分别为0.99、0.66、0.85。建立的BPNN模型能够较好地预报雷暴的发生, 2、3、6 h预报命中率分别为93.27%、93.33%和89.47%, 漏报率分别为6.73%、6.67%和10.53%, 空报率分别为4.90%、4.78%和2.86%, 临界成功指数分别为89.99%、80.33%和81.18%。该研究在一定程度上实现了雷暴天气的智能预报, 能够应用于机场及单站雷暴天气的预报预警。

关键词: 地基微波辐射计, BP神经网络, 雷暴预报, 预警

Abstract:

Using the MWP967KV ground-based microwave radiometer data at Shanghai Pudong International Airport Ground Meteorological Observation Station from 2018 to 2019, the civil aviation surface observation data, and the conventional sounding data of Baoshan Station, we analyzed the reliability of the microwave radiometer detection data.On this basis, we calculated the atmospheric parameters related to the occurrence of thunderstorms, selected the appropriate parameters as the forecasting factors, established the BP (Back Propagation) artificial neural network model for airport thunderstorm forecast, and evaluated the forecasting effect of the model.The results show that the mean absolute deviations of temperature, relative humidity, and water vapor density obtained by the MWP967KV ground-based microwave radiometer with the corresponding sounding data are 1.94 ℃, 16.05%, and 0.82 g·m-3, respectively, the root mean square errors are 1.41 ℃, 20.14%, 1.90 g·m-3, respectively, and the correlation coefficients are 0.99, 0.66, and 0.85, respectively.The established BPNN model can predict the occurrence of thunderstorms accurately.The forecast accuracy rates of 2 h, 3 h, and 6 h reach 93.27%, 93.33%, and 89.47%, respectively, and the missing rates are 6.73%, 6.67% and 10.53%, respectively.The reporting rates reach 4.90%, 4.78%, and 2.86%, and the critical success indices reach 89.99%, 80.33% and 81.18%, respectively.Therefore, this study realizes the intelligent forecast of thunderstorms to some extent, and the model can be applied to the forecasting and early warning of thunderstorm weather at airports and single stations.

Key words: Ground-based microwave radiometer, Back-propagation neural network, Thunderstorm forecast, Warning

中图分类号: