Comparison of common spatial interpolation methods for analyzing pollutant spatial distributions at contaminated sites.

Accurate prediction of the spatial distribution of pollutants in soils based on applicable interpolation methods is often the basis for soil remediation in contaminated sites. However, the applicable interpolation method has not been determined for contaminated sites due to the complex spatial distribution characteristics and stronger local spatial variability of pollutants. In this research, the prediction accuracies of three interpolation methods (including the different values of their parameters) for the spatial distribution of benzo[b]fluoranthene (BbF) in four soil layers were compared. These included inverse distance weighting (IDW), radial basis function (RBF), ordinary kriging (OK). The results indicated: (1) IDW1 is applicable for the first layer, RBF-IMQ is applicable to the second, third, and fourth layers. (2) For IDW, the prediction error is bigger with high weight where high values and low values intersect, while the prediction error is smaller where high (or low) values aggregated distribution. (3) For RBF, if the pollutant concentration trend at the predicted location is consistent with the known points in its neighborhood, the prediction accuracy is higher. (4) IDW is suitable for fitting more drastic curved surfaces, while RBF is more effective for relatively gentle curved surfaces and OK is reasonable for curved surfaces without local outliers. (5) The interpolation uncertainty is positively associated with the contaminant concentration and local spatial variability. Therefore, we suggest the selection of the applicable interpolation model must be based on the principle of the model and the spatial distribution characteristics of the pollutants.