云南民族大学学报（自然科学版）2026，Vol.35Issue(1)：57-65,9.DOI:10.3969/j.issn.1672-8513.2026.01.008
基于机器学习的德宏州PM2.5与O3质量浓度预测及潜在源分析
Prediction and potential source analysis of PM2.5 and O3 concentrations in Dehong prefecture based on machine learning
摘要
Abstract
In spring,air pollution in Dehong Prefecture is highly susceptible to cross-border transport resulting from agricultural practices such as slash-and-burn farming in South Asia and the Indochina Peninsula.Based on long-term time-series data from 2015 to 2022,this study integrates univariate statistical analysis,a convolutional neural network-long short-term memory(CNN-LSTM)model,and the potential source contribution function(PSCF)to develop a comprehensive method for predicting particulate matter 2.5(PM2.5)and ozone(O3)concentrations and identifying potential pollution sources.The results indicate that:the CNN-LSTM model demonstrates good predictive performance,with R2 values of 0.854 for PM2.5 and 0.821 for O3 concentration predictions;PM2.5 concentrations show negative correlations with dew point,humidity,precipitation,and temperature,while O3 concentrations are negatively correlated with the first three factors but exhibit a significant positive correlation with temperature;The overlap rates between potential pollution sources identified using observed values and those derived from predicted values are 92.135％for PM2.5 and 90.157％for O3.This method effectively predicts pollutant concentrations and traces cross-border sources,providing a valuable tool for precise air pollution control and scientific source identification in border regions.关键词
德宏州/PM2.5/O3/机器学习/潜在源贡献因子分析Key words
Dehong prefecture/PM2.5/O3/machine learning/potential source contribution factor analysis分类
资源环境引用本文复制引用
李懿琨,陈丹,胡天皓,潘自斌,马林转,陆飞翔,贾丽娟..基于机器学习的德宏州PM2.5与O3质量浓度预测及潜在源分析[J].云南民族大学学报（自然科学版）,2026,35(1):57-65,9.基金项目
国家自然科学基金(22476171) （22476171）
云南省科技厅科技计划(202403AC100027-1). （202403AC100027-1）
