Predicting distribution characteristics of PM2.5 concentrations across China combining land use regression model and spatiotemporally weighted stacking machine learning algorithms
文献类型：期刊论文
| 作者 | Wang, Wentao8; Zhang, Ping1,2,3,8; Han, Yuanyuan8; Zhang, Fengqian8; Huang, Yue4; Liu, Jinbao2; Pan, Ying5; Liu, Wenping6; Chen, Jian7; Li, Mingyao() |
| 刊名 | JOURNAL OF ENVIRONMENTAL CHEMICAL ENGINEERING
 |
| 出版日期 | 2026-04-01 |
| 卷号 | 14期号:2页码:121780 |
| 关键词 | PM 2.5 pollution Land use regression model Machine learning ensemble method Space-time weighting information Spatiotemporal modelling |
| ISSN号 | 2213-2929 |
| DOI | 10.1016/j.jece.2026.121780 |
| 产权排序 | 6 |
| 文献子类 | Article |
| 英文摘要 | Accurate estimation of PM2.5 concentrations is essential for understanding the health risks associated with exposure to PM2.5 and formulating strategies to address air pollution. However, most studies have overlooked the spatiotemporal heterogeneity of PM2.5 pollution, and single machine learning approaches often suffer from limited generalization capability and insufficient prediction accuracy. This study proposed a spatiotemporally weighted ensemble machine learning framework that integrated land use regression (LUR), machine learning, ensemble methods, and space-time weighting information. By incorporating predictive variables such as ground-based measurements, aerosol optical depth (AOD), meteorological parameters, and socioeconomic factors, it successfully generated distribution maps of PM2.5 concentrations across mainland China in 2019. The final selected hybrid model demonstrated exceptional overall predictive performance, achieving a 10-fold cross-validated (CV) R2 of 0.903, with root mean square error (RMSE) and mean absolute error (MAE) values of 12.13 mu g/m3 and 7.78 mu g/m3, respectively. Site-based and temporal CV-R2 reached 0.91 and 0.87, correspondingly, representing a 61% improvement in explanatory power compared to conventional LUR modelling (R2 = 0.56). The robustness and generalizability of the proposed model were further confirmed through external validation using independent data from 2020 to 2023. Global Moran's I spatial autocorrelation analysis revealed significant high-high and low-low clustering patterns of PM2.5 pollution, with severely affected regions concentrated in the Beijing-Tianjin-Hebei urban agglomeration, Shandong and Henan provinces of the North China Plain, Sichuan Basin, and Xinjiang, particularly during winter periods. The model we developed provides crucial support for the sustainable management of urban air quality. |
| URL标识 | 查看原文 |
| WOS关键词 | GROUND-LEVEL PM2.5 ; POLLUTION ; EMISSIONS |
| WOS研究方向 | Engineering |
| 语种 | 英语 |
| WOS记录号 | WOS:001694318500001 |
| 出版者 | ELSEVIER SCI LTD |
| 源URL | [http://ir.igsnrr.ac.cn/handle/311030/220880]  |
| 专题 | 资源利用与环境修复重点实验室_外文论文 |
| 通讯作者 | Zhang, Ping |
| 作者单位 | 1.Chinese Acad Sci, Res Ctr Ecoenvironm Sci, State Key Lab Urban & Reg Ecol, Beijing 100085, Peoples R China; 2.Minist Nat Resources, Key Lab Degraded & Unused Land Consolidat Engn, Xian 710075, Peoples R China; 3.Xian Univ Architecture & Technol, State Key Lab Green Bldg Western China, Xian 710055, Peoples R China; 4.Xian Inst Prospecting & Mapping, Xian 710054, Peoples R China; 5.Chinese Acad Sci, Inst Geog Sci & Nat Resources Res, Beijing 100101, Peoples R China; 6.Huazhong Agr Univ, Coll Hort & Forestry Sci, Wuhan 430070, Peoples R China; 7.Zhejiang A&F Univ, State Key Lab Subtrop Silviculture, Linan 311300, Peoples R China 8.Xian Polytech Univ, Sch Environm & Chem Engn, Xian 710048, Peoples R China; |
| 推荐引用方式 GB/T 7714 | Wang, Wentao,Zhang, Ping,Han, Yuanyuan,et al. Predicting distribution characteristics of PM2.5 concentrations across China combining land use regression model and spatiotemporally weighted stacking machine learning algorithms[J]. JOURNAL OF ENVIRONMENTAL CHEMICAL ENGINEERING,2026,14(2):121780. |
| APA | Wang, Wentao.,Zhang, Ping.,Han, Yuanyuan.,Zhang, Fengqian.,Huang, Yue.,...&Fan, Jinghao.(2026).Predicting distribution characteristics of PM2.5 concentrations across China combining land use regression model and spatiotemporally weighted stacking machine learning algorithms.JOURNAL OF ENVIRONMENTAL CHEMICAL ENGINEERING,14(2),121780. |
| MLA | Wang, Wentao,et al."Predicting distribution characteristics of PM2.5 concentrations across China combining land use regression model and spatiotemporally weighted stacking machine learning algorithms".JOURNAL OF ENVIRONMENTAL CHEMICAL ENGINEERING 14.2(2026):121780. |
入库方式： OAI收割
来源：地理科学与资源研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。