气象2025，Vol.51Issue(4)：460-472,13.DOI:10.7519/j.issn.1000-0526.2024.111401
一种基于机器学习的自动气象观测站风向异常识别方法
A Method for Identifying Abnormal Wind Direction at Automatic Weather Stations Based on Machine Learning
摘要
Abstract
To address the issue of high-concealed abnormal wind directions in automatic weather station(AWS)data,this study establishes an abnormal wind direction identification method based on the density-based spatial clustering of applications with noise(DBSCAN)clustering algorithm.Historical wind direc-tion data from 16 weather events affecting Guangzhou between 2016 and 2022,including cold waves,cold air masses,and typhoons,as well as observed wind direction data from AWSs during the impact of Ty-phoon Saola(No.2309),are used to detect abnormal wind directions.The analysis results reveal that the proportion of AWSs with suspicious wind directions in historical cases ranges from 0.46％to 5.56％,while the proportion of AWSs with erroneous wind directions varies from 0.25％to 2.05％.During the case of Typhoon Saola,the method identifies 13 AWSs with significantly deviating wind directions from the dominant surface wind direction,which is primarily due to wind direction sensor malfunctions and envi-ronmental impacts on AWS observations.Compared to that by the traditional method,the accuracy of wind direction error identification has improved by 20.32 percentage point.The new method provides a novel approach for the quality control of historical wind direction data from AWSs and offers an effective reference for the operational monitoring and on-site verification of AWS equipment.关键词
机器学习/DBSCAN/自动气象观测站/风向异常识别/数据质量控制Key words
machine learning/DBSCAN/automatic weather station/identification of abnormal wind direc-tion/data quality control分类
天文与地球科学引用本文复制引用
张志坚,张静,伍光胜..一种基于机器学习的自动气象观测站风向异常识别方法[J].气象,2025,51(4):460-472,13.基金项目
广州市科技重点研发计划(2023B04J0704、2023B04J0667)共同资助 （2023B04J0704、2023B04J0667）
