基于决策树的雷暴大风客观预报方法研究  

On the Objective Forecasting Method for Thunderstorms and Gales Based on the Algorithm of Classification and Regression Tree

作　　者：陈荣泉 周静 李厚伟 CHEN Rongquan;ZHOU Jing;LI Houwei(Zhaoqing Meteorological Service,Zhaoqing 526060)

机构地区：[1]肇庆市气象局,广东肇庆526060

出　　处：《广东气象》2025年第1期10-15,共6页Guangdong Meteorology

基　　金：广东省气象局科学技术研究项目(GRMC2021M31)。

摘　　要：为进一步做好雷暴大风的预报,通过对2005-2023年肇庆市雷暴大风分析,得到雷暴大风的气候特征和空间分布,并归纳出5种主要天气形势特征。以环境物理量:K、Li、CAPE、T85、t_(g)、SHR3、SHR6、DCAPE为特征因子,运用决策树方法建立雷暴大风天气预报模型,通过模型可知CAPE(阈值:2 086.05 J/kg),SHR3(阈值:8.65 m/s),抬升指数(阈值:-0.94℃)对判别雷暴大风有较好的效果,模型的准确率达到83.5%,并尝试对2019年6月4日雷暴大风过程进行回报,效果较好。For performing better forecasting of thunderstorms and gales,this paper determined the climatological characteristics and spatial distributions of the thunderstorms and gales taking place in Zhaoqing from 2005 to 2023 and summarized five main synoptic patterns associated with them.The results are shown as follows.Taking as characteristic factors the following environmental physical quantities of K,Li,CAPE,T85,t_(g),SHR3,SHR6 and DCAPE,this paper used the algorithm of classification and regression tree to set up a model for forecasting the weather of thunderstorms and gales.It is known from the model that the CAPE,with the threshold of 2086.05 J/kg,SHR3,with the threshold of 8.65 m/s,and a lifting index,with the threshold of-0.94℃,are doing a good job in distinguishing thunderstorms and gales,with the model accuracy as high as 83.5%.A retrospective forecast was carried out for such weather that happened locally on June 4,2019 and the result was satisfactory.

关 键 词：天气学 决策树 预报方法 环境物理量 雷暴大风 

分 类 号：P44[天文地球—大气科学及气象学]