Objective The aims were to develop a precise spatial interpolation method for heavy metal arsenic (As) in soil, further to accurately identify As pollution pattern in the study area, and to provide reliable data support for the precise assessment of the soil As pollution level in the Daye Lake Basin and the analysis of the pollution sources.

Method  Total 354 soil sampling points and soil heavy metal As in the Daye Lake Basin was selected as the representative sampling points and element, respectively. Four interpolation methods, including Ordinary Kriging (OK), Inverse Distance Weighting (IDW), Local Polynomial Interpolation (LPI), and Radial Basis Function (RBF), as well as their respective different parameters, were analyzed and compared in terms of their effects on the spatial interpolation, spatial distribution, and pollution risk assessment of soil As. Additionally, the cross-validation method was utilized to verify the interpolation accuracy of each approaches, and three different pollution assessment methods, such as the single-factor pollution index, the potential ecological risk index, and the Index of Geo accumulation, were adopted to evaluate the As pollution in the soil of the Daye Lake Basin.

Result The content of As in the study area ranged from 3.83 to 1174.27 mg kg⁻¹, and the coefficient of variation was 2.39, indicating a high spatial discretization and a significant influence by human activities. The high-value areas of As were concentrated in the central and northern regions. The single-factor pollution index and the potential ecological risk index showed that 49.44% and 89.55% of the sampling points in the basin were slightly polluted, respectively. The Index of Geo accumulation revealed that the sampling points with no accumulation and no-moderate accumulation accounted for 83.05%. Overall, the soil As pollution in the basin was mild, and point source pollution or small-scale non-point source pollution might exist in high-pollution areas. There were significant differences in the spatial interpolation results of As among the four methods. The deterministic interpolation method IDW outperformed the other methods in terms of accuracy indicators, with a Mean Absolute Error (MAE) of 0.3084, a Mean Relative Error (MRE) of 0.1638%, and a Root Mean Square Error (RMSE) of 3.9112.

Conclusion The spatial distribution of soil As in the Daye Lake Basin varied greatly, and mild pollution existed in the basin. Different parameters of the same interpolation method could lead to inconsistent interpolation results. The interpolation accuracy of IDW and LPI methods improved with the increase of weight. Among different functions of RBF, the Completely Regularized Spline (CRS) had the best accuracy. The OK method was suitable for the exponential model in the semi-variogram fitting function. Different weights of interpolation methods could also cause changes in interpolation results, requiring multiple fittings of different parameters. Based on the comprehensive analysis of each interpolation result, the recommendation order of the four interpolation methods was IDW > RBF > OK > LPI.