基于人工智能模型的土壤重金属污染成因解析

Heavy metal pollution in soil is a major environmental issue in China. The analysis of the causes of pollution is a key prerequisite for solving this problem. The analysis of the causes of pollution on a large-scale region is challenging. Because the quantitative assessment of the contribution of the major impact factors to heavy metal pollution, and the inference of causal and interactive mechanisms between the evolution of heavy metals in soil and influencing factors are difficult. Therefore, the Pearl River Delta, one of the rapidly developing regions across China during the past three decades, will be taken as a case study. Based on the probability graph model and the causal network model, the influences of the counterfactual and the intervention will be quantified using mathematical algorithms, and the parameters and structure of the causal network model will be inferred. An analytical model for the causes of heavy metal pollution in soil will be developed based on artificial intelligence. Additionally, the main pollution pathways of soil heavy metals in different soil parent materials and land use types will be identified on the basis of this model and hierarchical clustering analysis. The contribution of the key influencing factors of soil heavy metal evolution will be quantitatively evaluated using the PC algorithm and the IDA algorithm. Moreover, the causal and interactive mechanisms between the evolution of heavy metals in soil and the influencing factors will be uncovered, and the causes of heavy metal pollution in soil will be further elucidated. This study will provide insights into the quantitative analysis of the causes of heavy metal pollution in soil on a regional scale.

土壤重金属污染是我国亟待解决的重大环境问题。解析污染成因是解决这一难题的关键前提。针对大尺度区域污染成因分析，主要影响因子对重金属污染贡献的定量评估、土壤重金属演变与影响因子之间的因果及交互机制的解析，是研究难点。因此，本项目以近三十年来快速发展的珠三角为例，基于概率图模型和因果网络模型，算法化虚拟事实(Counterfactual)和干预逻辑(Intervention)的影响，推演因果网络模型的参数和结构，构建基于人工智能的土壤重金属污染成因解析模型；在此模型基础上，结合分层聚类分析，识别不同土壤母质与土地利用类型的土壤重金属主要污染途径；应用PC算法结合IDA算法，定量评估不同土壤母质与土地利用类型的土壤重金属演变关键影响因子的贡献，揭示土壤重金属演变与其影响因素间因果及交互机制，深入阐明土壤重金属污染成因机制。预期可为区域性土壤重金属污染成因定量解析提供基于人工智能模型的创新思路。

土壤重金属污染是我国亟待解决的重大环境问题。解析污染成因是解决这一难题的关键前提。针对大尺度区域污染成因分析，主要影响因子对重金属污染贡献的定量评估、土壤重金属演变与影响因子之间的因果及交互机制的解析，是研究难点。因此，本项目以近三十年来快速发展的珠三角为例，基于概率图模型和因果网络模型，算法化虚拟事实(Counterfactual)和干预逻辑(Intervention)的影响，推演了因果网络模型的参数和结构，构建了基于人工智能的土壤重金属污染成因解析模型；在此模型基础上，结合分层聚类分析，识别了不同土壤母质与土地利用类型的土壤重金属主要污染途径；应用PC算法结合IDA算法，定量评估了不同土壤母质与土地利用类型的土壤重金属演变关键影响因子的贡献，揭示了社会经济发展对重金属显著影响主要取决于经济的规模、结构和技术效应的强度，Hg和Cr污染减轻主要受经济增长结构和技术效应的强烈影响(65%和77%; 81%和71%)，而经济增长规模效应主要影响Cd、As和Pb污染增加(80%,72%和69%), 深入阐明了土壤重金属污染成因机制。研究结果为区域性土壤重金属污染成因定量解析提供了基于人工智能模型的创新思路。