1. Source analysis and distribution prediction of soil heavy metals in a typical area of the Qinghai-Tibet Plateau
- Author
-
Xinjie Zha, Liyuan Deng, Wei Jiang, Jialu An, Hongcai Wang, and Yuan Tian
- Subjects
Qinghai-Tibet Plateau ,Soil ,Heavy metals ,Geodetector ,Random Forest ,Ecology ,QH540-549.5 - Abstract
The excessive presence of heavy metals (HMs) in soil poses a significant threat to both ecosystems and human health. Consequently, there is a compelling need for quantitative analysis of HMs concentration in soil and the prediction of potential contamination. In this study, 58 surface soil samples were systematically collected from 11 different townships in Luolong County. Using ArcGIS 10.7, the fishing net interpolation resampling was performed to obtain model data. The GeoDetector model was employed to determine the key driving factors and their interrelationships affecting soil composition. Subsequently, influential driving factors with higher explanatory power and employed a Random Forest (RF) model to generate a contamination prediction map. The results revealed that arsenic (As), cadmium (Cd) and lead (Pb) exceeded the risk screening values by 8.62%, 10.34%, and 10.34%, respectively. The GeoDetector model identified factors such as elevation, annual average precipitation, distance to the nearest river, geomorphic type in natural sources, geological type in geological sources, distance from roads, proximity to mining sites, per capita income of inhabitants, total potassium content and organic matter content in anthropogenic sources significantly influencing the spatial distribution of HMs concentration in the soil. The interactions among the primary driving factors increased their explanatory capacity. By using RF model to predict the spatial distribution of the main influencing factors of HMs, it was found that areas with a high probability of As contamination were mainly concentrated in the northern, central and southeast regions of Luolong County. Regions with Cd concentration exceeding the risk screening value were primarily concentrated in the east, northeast and a few northern areas of Luolong County, while the likelihood of Pb contamination was higher in the northern and southwestern regions of Luolong County. This study integrates spatial stratified heterogeneity with the random forest model to mitigate overfitting in the prediction of soil HM contamination, a common issue in traditional machine learning methods. This approach is essential for elucidating the environmental drivers of soil HM pollution, predicting high-risk areas in regions with complex environmental conditions and limited data, and ensuring the safety and stability of agricultural production as well as the well-being of local residents.
- Published
- 2024
- Full Text
- View/download PDF