Predicting the spatial pollution of soil heavy metals by using the distance determination coefficient method

Highlights: • LUR simulates the spatial distribution of heavy metal concentrations well. • The optimal buffer distance determined by DCD is easier and more accurate. • The spatial distribution simulated by LUR outperforms Ordinary Kriging. Identifying the spatial distribution of heavy metal concentrations is a prerequisite for soil contamination assessment and control. In this study, soil surface samples (0–20 cm) were collected in Wuwei, China, and heavy metal concentrations were determined. The LUR (land use regression) model was used to simulate the spatial distribution of seven heavy metal concentrations in the study area, considering various factors, and the results were compared with ordinary kriging (OK) interpolation. Based on A Distance Decay REgression Selection Strategy (ADDRESS), the distance-coefficient of determination (DCD) was proposed to more easily and accurately select the optimal buffer of the relevant covariates. The simulation results showed that the adj R² values of the LUR models of the remaining heavy metals were all above 0.6, and the empirical comparison showed that LUR was better than OK. The variables that showed a more significant impact on soil heavy metal concentrations in the research area included road length, normalized difference vegetation index (NDVI), and soil sample nutrients around the sample site. In the research area, the concentration of heavy metals in the soil was greatly affected by motorway, primary roads, and secondary roads in the range of 1.2–2.1 km (r > 0.5), while building and As, Cu, Pb and Zn in the range of 3.6–4.8 km had a significant correlation (r > 0.5). This study provides scientific evidence and basic information for the prevention and control of heavy metal contamination and human health risk assessment management in arid zones.