| 董天浩,任传猛,任清盛,董贝,张仁杰,李承永,潘淑芳,郭焱,纪雄辉,谢运河.基于PMF和机器学习模型的蒸水中下游流域农田土壤重金属污染状况及来源解析[J].农业环境科学学报,2026,45(1):69-81. |
| 基于PMF和机器学习模型的蒸水中下游流域农田土壤重金属污染状况及来源解析 |
| Source apportionment of heavy metal pollution in farmland of the middle and lower reaches of the Zhengshui River basin based on PMF and machine learning models |
| 投稿时间：2025-01-13 |
| DOI：10.11654/jaes.2025-0045 |
| 中文关键词: 土壤重金属 源解析 PMF模型 SOM模型 LightGBM模型 |
| 英文关键词: soil heavy metal source apportionment PMF model SOM model LightGBM model |
| 基金项目:国家重点研发计划项目（2022YFD1700103）；国家重点研发青年科学家项目（2023YFD1703000，2023YFD1703100）；国家地区联合基金（U22A20606，U21A20291）；山东省生态农业技术体系济南综合试验站（SDAIT-30-10） |
| 作者 | 单位 | E-mail | | 董天浩 | 湖南省耕地与农业环境生态研究所, 长沙 410125
济南市农业科学研究院, 济南 250100
农业农村部长江中游平原农业环境重点实验室, 长沙 410125
农田土壤重金属污染防控与修复湖南省重点实验室, 长沙 410125 | | | 任传猛 | 济南市农业科学研究院, 济南 250100 | | | 任清盛 | 济南市农业科学研究院, 济南 250100 | | | 董贝 | 济南市农业科学研究院, 济南 250100 | | | 张仁杰 | 湖南省耕地与农业环境生态研究所, 长沙 410125
农业农村部长江中游平原农业环境重点实验室, 长沙 410125
农田土壤重金属污染防控与修复湖南省重点实验室, 长沙 410125 | | | 李承永 | 济南市农业科学研究院, 济南 250100 | | | 潘淑芳 | 湖南省耕地与农业环境生态研究所, 长沙 410125
农业农村部长江中游平原农业环境重点实验室, 长沙 410125
农田土壤重金属污染防控与修复湖南省重点实验室, 长沙 410125 | | | 郭焱 | 中国农业大学土地科学与技术学院, 北京 100193 | | | 纪雄辉 | 湖南省耕地与农业环境生态研究所, 长沙 410125
农业农村部长江中游平原农业环境重点实验室, 长沙 410125
农田土壤重金属污染防控与修复湖南省重点实验室, 长沙 410125 | | | 谢运河 | 湖南省耕地与农业环境生态研究所, 长沙 410125
农业农村部长江中游平原农业环境重点实验室, 长沙 410125
农田土壤重金属污染防控与修复湖南省重点实验室, 长沙 410125 | 45069794@qq.com |
|
| 摘要点击次数: 652 |
| 全文下载次数: 432 |
| 中文摘要: |
| 为了探究湘江支流蒸水中下游流域农田土壤重金属污染风险，对蒸水中下游流域农田土壤进行了采样分析及污染源解析。结果表明：该区域土壤存在较大的Cd污染风险，部分存在As、Pb和Cu污染风险。研究区识别出4个污染源，即自然-大气沉降混合源、自然源、大气沉降源和工业源。PMF模型判定土壤As和Hg主要受自然-大气沉降混合源的影响，Cr、Ni和Cu主要受自然源影响，Pb主要受大气沉降源影响，Cd和Zn主要受工业源影响；4个污染源的贡献率依次为30.8%、27.0%、22.6%和19.6%。SOM模型污染源分类结果与PMF模型源解析结果均吻合度较高。LightGBM模型结果表明距蒸水主流距离对Cd、Pb、Ni和Zn的影响均较大，PM2.5浓度对Cd和Pb的影响较大；对As、Hg和Cr影响最大的因子均为母岩类型；与交通相关的因子对Cu和Zn的影响较大。研究表明，该研究区农田土壤有一定的重金属污染风险，且污染来源较复杂，LightGBM模型可对PMF模型结果进行一定程度的补充，受体模型结合机器学习模型能够更合理地判别各土壤重金属主要的污染来源。 |
| 英文摘要: |
| To investigate the risk of heavy metal contamination in agricultural soils in the middle and lower reaches of the Zhengshui River, a tributary of the Xiangjiang River, soil sampling and pollution source apportionment were conducted. The results indicate the following： The study area exhibits a significant risk of cadmium（Cd）contamination in the soil, with partial risks of arsenic（As）, lead（Pb）, and copper （Cu）contamination. Four pollution sources were identified in the study area：a mixed natural-atmospheric deposition source, a natural source, an atmospheric deposition source, and an industrial source. The PMF（Positive Matrix Factorization）model determined that soil As and Hg are primarily influenced by the mixed natural-atmospheric deposition source, Cr, Ni and Cu are mainly affected by the natural source, Pb is predominantly influenced by the atmospheric deposition source, and Cd and Zn are primarily associated with the industrial source. The contribution rates of these four sources are 30.8%, 27.0%, 22.6%, and 19.6%, respectively. The SOM（Self-Organizing Map） model's classification results for pollution sources showed high consistency with the PMF model's source apportionment results. The LightGBM（Light Gradient Boosting Machine）model results indicated that the distance from the mainstream of the Zhengshui River has a significant impact on Cd, Pb, Ni, and Zn, while PM2.5 concentration has a notable influence on Cd and Pb. The most influential factor for As, Hg, and Cr is the parent rock type. Traffic-related factors have a considerable impact on Cu and Zn. The study reveals that the agricultural soils in the study area face certain risks of heavy metal contamination, with complex pollution sources. The LightGBM model can complement the PMF model results to some extent, and the combination of receptor models and machine learning models can more reasonably identify the primary sources of heavy metal contamination in the soil. |
| HTML
查看全文
查看/发表评论 下载PDF阅读器 |
|
|
|
