Prediction of heavy metal spatial distribution in soils of typical industrial zones utilizing 3D convolutional neural networks

Abstract

Abstract Land resources are vital for urban development and construction. Abandoned industrial areas often contain large amounts of heavy metals from past industrial activities. Accurate knowledge of soil pollutant content and spatial distribution is crucial to avoid health risks and achieve sustainable soil use. However, due to the limitation of human, material and financial resources, it is difficult to carry out intensive detection of soil heavy metals in polluted areas. This problem can be solved by using known soil heavy metal content data to predict the heavy metals in unknown regions. This study utilizes a three-dimensional Convolutional Neural Network (3DCNN) model, combined with spatial location and soil physicochemical properties, to predict heavy metal in a typical industrial zone in Qingdao City. The results show that the $$R^2$$ of 3DCNN for predicting cadmium (Cd), lead (Pb), copper (Cu) and nickel (Ni) are 0.59, 0.59, 0.77 and 0.51, respectively. Therefore, 3DCNN can be used as an effective method for spatial prediction of soil heavy metals, which can reduce the cost of sampling and laboratory analysis. The three-dimensional spatial distribution analysis revealed that Cd and Pb were concentrated in the surface soil layer and gradually decreased with the depth, while Cu and Ni contents are mainly concentrated in the range of 3 m, exhibiting downward migration. Therefore, heavy metal enrichment has occurred in this area, and soil heavy metal treatment should be carried out before redevelopment.

Keywords

• Industrial area soil
• Soil contamination
• Heavy metals
• Content prediction
• 3DCNN
• Spatial distribution characteristics