决策树算法在雷达反射率因子质控中的应用  

Application of Decision Tree Algorithm in the Quality Control for Radar Reflectivity

作　　者：袁凯 李明 庞晶 李武阶 YUAN Kai;LI Ming;PANG Jing;LI Wujie(Wuhan Meteorological Observatory,Wuhan 430040,China)

机构地区：[1]武汉市气象台,湖北武汉430040

出　　处：《沙漠与绿洲气象》2023年第1期139-144,共6页Desert and Oasis Meteorology

基　　金：武汉市短时临近预警预报技术创新团队(2021年度);武汉市气象局科研课题(WH202101)。

摘　　要：基于决策树算法,利用2020年武汉多普勒天气雷达探测资料,分析决策树算法对反射率因子的质控效果,得到以下主要结论:(1)决策树算法能有效滤除湖北东部地区由于山脉和城市建筑所导致的固定地物回波和超折射地物回波。(2)对于降水回波而言,决策树算法不改变雷达近距离范围内回波的位置、形状和强度,而远距离处(≥160 km)回波则存在一定程度的过度质控,主要是由于雷达波束随距离增加而抬高以及回波垂直梯度较大导致。(3)决策树算法对辐射状和弧状杂波的质控能力较好,其滤除率达到95%以上。In this paper,the utility of the decision tree algorithm on the quality control for radar reflectivity was examined by using the Doppler radar data of Wuhan in 2020,the main conclusions are as follows:(1)the decision tree algorithm can effectively filter the fixed and hyperrefraction surface echoes caused by mountains and urban buildings in the eastern Hubei.(2)For the radar echo of precipitation,the decision tree algorithm does not change the position,shape,and intensity of the radar echo within the short distance,but there is a certain degree of over quality-controlled radar echo for the long distance(generally farther than 160 km),which mainly caused by the rising radar beam and vertical gradient of echo with distance increase.(3)The statistical results show that the decision tree algorithm has a good quality control utility for those echoes,with a filtering rate of 95% at least for radial and arcuation radar echoes.

关 键 词：决策树算法 反射率因子 质量控制 地物杂波 

分 类 号：P412.25[天文地球—大气科学及气象学]