Abstract:
Objective The aim was to accurately map the three-dimensional (3D) spatial distribution of soil pollutants and overcome the smoothing effects of traditional spatial interpolation and achieve precise characterization of the 3D spatial distribution of soil pollutants under sparse borehole data conditions.
Method A Soil-Landscape Inference-Mixed Inverse Distance Weighting (SoLIM-IDW) 3D hybrid model was constructed using multi-source auxiliary information and soil borehole data. This model was employed to predict the 3D distribution of chloroform concentrations in the soil within a pesticide workshop region. The performance was compared with traditional spatial interpolation methods such as IDW and OK. The impact of multi-source auxiliary data on the spatial prediction of soil chloroform was evaluated using a geographic detector.
Result The results indicated that chloroform contamination hotspots were mainly distributed in the surface to deep layers of soil in the central pesticide workshop, and the upper layer of the southeast wastewater treatment area. The SoLIM-IDW method exhibited higher spatial prediction accuracy, with R2 ranging from 0.37 to 0.39, and RMSE ranging from 83.15 to 84.48 mg kg–1. The estimated volume of soil exceeding the standard through spatial prediction ranged from 7024 to 7980 m3. Multi-source auxiliary information interacted, and the integration of the multi-source auxiliary data yielded higher accuracy compared to relying on a single auxiliary data source.
Conclusion This study demonstrated that making full use of easily accessible and low-cost multi-source pollution auxiliary information under the conditions of limited cost and sparse boreholes contributes to improve the accuracy of 3D spatial prediction and risk assessment of site pollutants.
下载: