Effective methods for prediction and visualization of contaminated soil volumes in 3D with GIS

Abstract: Geographical Information Systems (GISs) have shown to be of great help in the work at contaminated sites. Today there is an increase in the development of 3D modeling in many fields. However, to combine the advantage working with spatial data in a GIS and 3D modeling for subsurface soil data is limited. In this study, the possibilities of interpolating volumes of soil contamination in 3D were investigated. The study is focused on the ability of integrating 3D modeling with the GIS-environment in projects for contaminated land. Three different interpolation techniques were evaluated; Kriging-, Inverse Distance Weighting (IDW)-, and Nearest Neighbor interpolation. The data used in the study originates from a contamination project at a former gasworks site in Norrköping, Sweden. The data was sampled by the consultancy company Sweco. Three major contaminants of different characteristics were evaluated for potential of volume interpolation in 3D (lead, benzene, and polycyclic aromatic hydrocarbons (PAHs)). The study also aimed at determining if the interpolation method with greatest potential differs in relation to contaminant type. Prior the 3D interpolation possible GIS software and other methods for 3D interpolation were identified. Geostatistical analyses were performed where the optimized parameters for the interpolations were determined. In the geostatistical analyses a spatial dependence at short distances was found for all contaminants in the vertical direction (1-2 m) but not in the horizontal plane. The lack of spatial dependence in the horizontal plane indicates an effect of the coarse sampling density (about 10 m compared to 0.5 m for the vertical direction). The distribution patterns of the three contaminants are expected. Lead and PAH are both distributed differently depending on the soil material. Benzene seems to be distributed equally in all material and was interpolated with the same parameters in all soil types. All contaminants show greatest potential for volume interpolation with Kriging and secondly IDW. The least accurate method was Nearest Neighbor. The optimized parameters for interpolation are similar for lead and PAH but differed for benzene. That reflects their difference in grade of mobility. Benzene shows the most accurate 3D volume interpolation while PAH the least. It is suggested that an effective volume interpolation of pollutants in 3D must combine regular GISs with software handling 3D volumes, or a development of the GISsoftware is necessary.