Hydrogeological spatial modelling: A comparison between frequentist and Bayesian statistics

Abstract

Abstract Traditional and modern spatial prediction techniques are applied in the analysis of water quality, evidencing a new approach that allows the modeling of a hydrogeological system in the central area of Boyacá, Colombia. The objective is to determine the quality status of groundwater for human consumption. In the process, spatial predictions were made based on frequentist methods (kriging, cokriging) and Bayesian methods (R-INLA stochastic partial differential equation) as an alternative to Markov chain Monte Carlo methods that require a large computational cost. From the application of these methods, a comparison is made by statistical tests that determine the goodness of adjustment of the predictions. Our interest here is in the implementation of future more robust, economic and scientific solutions, particularly for hydrogeological data, and the proper management of water resources. Finally, the vulnerability of aquifers is analyzed with the DRASTIC method, which takes into account the surrounding media of the aquifers through the variables depth (D), net recharge (R), type of aquifer (A), soil (S), topography (T), impact of the vadose zone (I) and hydraulic conductivity (C).