基于CatBoost模型和SHAP解释方法的土壤重金属影响因素与程度定量分析

Quantitative analysis of the influence factors and extent of soil heavy metals based on CatBoost model and SHAP interpretation method

掌握土壤重金属空间特征及定量解析污染源对于精准治污具有重要支撑作用,但定量分析影响因子的贡献和影响程度面临巨大挑战.本文选取空间位置、地形地貌、土壤特征三大类16个影响因素,采用CatBoost-SHAP耦合模型定量评估了土壤重金属空间分布的影响因子和范围边界.结果表明,CatBoost在预测土壤Cd、Pb和Cr含量上具有较高精度,R^(2)分别为0.76、0.71和0.81.空间位置对土壤Cd、Pb分布的影响分别可达71.28%和73.54%,其中,与冶炼厂的距离(DI_(Smp))是土壤Cd空间分布的关键影响因素,影响范围在7.5 km内;与二级公路的距离(DR_(Sec))是土壤Pb空间分布的关键影响因素,影响范围可达8 km;Fe_(2)O_(3)是土壤Cr空间分布的关键影响因素,土壤中氧化铁含量达到4.5%后开始正向影响土壤Cr含量.本研究定量解析了土壤重金属影响因素,可为土壤重金属污染防治提供科学决策依据.

Understanding the spatial characteristics and pollution sources for heavy metals in soil is helpful to the precise control of heavy metal pollution,but there is still a great challenge that the relative contribution of influencing factors to heavy metal pollution would be quantitative analysed.In this study,the coupled CatBoost-SHAP model was used to quantitatively assess the influencing factors and range boundaries for heavy metals spatial distribution,in which the three major categories(16 influencing factors)consisting of spatial location,topography,and geomorphology were selected.The results indicated that CatBoost had high accuracy for predicting the concentrations of Cd,Pb and Cr,since their R^(2) were 0.76,0.71 and 0.81,respectively.The influence extent of spatial location on the distribution of Cd and Pb could reach 71.28%and 73.54%,respectively,where the distance from the smelter(DI_(Smp))was the key factor,affecting the spatial distribution of Cd in soil with an influence range of 7.5 km.The distance from the secondary road(DR_(Sec))was the key factor in the affecting range of 8 km for the spatial distribution of Pb in soil.While for Cr,the content of Fe_(2)O_(3) was the key factor affecting the spatial distribution,in which the concentration of Cr increased after the content of Fe_(2)O_(3) reaching 4.5%.The above quantitative information on the factors affecting heavy metals spatial distribution in soil,which could provide the scientific strategies for the prevention and control of heavy metal pollution.