基于随机森林和逻辑回归的舟山梅雨期暴雨关键指标研究  

Rainstorm Indicators During Meiyu Period Over Zhoushan Based on Random Forest and Logistic Regression

作　　者：王琼 马恩点 胡潇杰 张鑫 孙轶 卢琪 傅娜 Wang Qiong;Ma Endian;Hu Xiaojie;Zhang Xin;Sun Yi;Lu Qi;Fu Na(Zhoushan Meteorological Bureau,Zhoushan 316021,Zhejiang,China;Dinghai District Meteorological Bureau of Zhoushan City,Zhoushan 316004,Zhejiang,China)

机构地区：[1]舟山市气象局,浙江舟山316021 [2]舟山市定海区气象局,浙江舟山316004

出　　处：《科技通报》2024年第9期7-12,共6页Bulletin of Science and Technology

基　　金：舟山市公益性科技项目(2022C31074)。

摘　　要：本文分析了2010—2020年舟山梅雨期(6—7月)暴雨的时空分布特征和天气形势特征。利用ERA5再分析资料和随机森林算法分析了700 hPa风速和风向、850 hPa风速和风向、850 hPa的温度和露点温度、500 hPa温度以及强天气威胁指数等共8个物理量与舟山梅雨期暴雨序列的相关性,计算得到各物理量相关性,筛选出前6个物理量建立预报方程,通过逻辑回归方法可得舟山梅雨期暴雨预报模型,用来判断暴雨是否发生,对舟山梅雨期发生暴雨的概率可提供参考依据。This paper analyzes the spatial and temporal distribution characteristics and weather situation characteristics of rainstorm during the Meiyu period(June to July)in Zhoushan from 2010 to 2020.Based on ERA5 reanalysis data and the random forest algorithm,the correlation between eight physical quantities including 700 hPa wind speed and direction,850 hPa wind speed and direction,850 hPa temperature and dew point temperature,500 hPa temperature,and strong weather threat index.Filter out the top six physical quantities to establish a prediction equation,through the method of logistic regression,the rainstorm prediction model of Zhoushan Meiyu period can be obtained,which can be used to judge whether rainstorm occurs,and can provide a reference basis for the probability of rainstorm occurring in Zhoushan Meiyu period.

关 键 词：舟山梅雨 暴雨指标 随机森林 逻辑回归 

分 类 号：P456[天文地球—大气科学及气象学]