Spatial Interpolation for the Distribution of Groundwater Level in an Area of Complex Geology Using Widely Available GIS Tools

Abstract

The present study is an attempt to implement several spatial interpolation methods for the distribution of groundwater level in a wider area with multiple aquifers having variable hydraulic characteristics. Moreover, the goal of this study is to compare the results of these methods and check their accuracy and reliability, considering mainly the physical meaning of the outcome. Finally, we try to figure out which of these methods manage to identify hydrogeological features like groundwater divides, hydraulic conductivity barriers and no flow boundaries, and to highlight the hydraulic relationship between aquifers. Exploratory Spatial Data Analysis proved to be a necessary step prior to the implementation of spatial interpolation methods, since normalization of datasets, removal of general trends and data declustering was necessary for the proper implementation of geostatistical methods and reduction of the uncertainty of the results. Inverse Distance Weight, Radial basis functions, simple Kriging and Cokriging methods were implemented. None of the implemented methods produced results that were totally unreliable or erroneous and each method added pieces of information that were useful for the deeper understanding of the hydrogeological processes in the study area. The most appropriate spatial interpolation method for generating a groundwater level distribution surface, in an area with multiple aquifers and significant heterogeneity in hydraulic properties proved to be the Ordinary Cokriging method with altitude as a second parameter, which was highly correlated to groundwater level values in the study area. Cokriging method succeeds to accurately represent both the local variations within the individual aquifers and also to highlight the hydraulic relationships between them.

Highlights

• All spatial interpolation methods produced realistic surfaces.

• Geostatistical methods produced smoother surfaces and lower hydraulic gradients.

• Applying block declustering to sample data significantly reduces prediction uncertainty.

• The most appropriate method for groundwater level distribution is Ordinary Cokriging.