Comparative evaluation of kriging as an interpolator for groundwater contaminant plumes

This study analyzed the capabilities of three interpolation methods as applied to groundwater contaminant plumes. The three methods selected for evaluation were (1) kriging, (2) inverse distance squared, and (3) regression using a fifth order polynomial. The methods were evaluated over a range of values for four variables: (1) sample density, (2) sample location tolerance, (3) data quality, and (4) network age. Sample values and truth values were determined synthetically using a two-dimensional solute transport equation. The three interpolation methods were also applied to a case study to determine if the properties indicated by the synthetic data were valid for actual sample values. It was concluded that the kriging method consistently provided the highest quality estimates within the range of variables used in this study, including the case study. The inverse distance squared method provided the second best results and the regression method consistently provided the lowest quality estimates. When variables other than estimate quality were considered, no best method could be determined, with each method demonstrating strengths and weaknesses specific to that method.