The use of stochastic simulations and geophysical logs to characterize spatial heterogeneity in hydrogeologic parameters

Characterization of the spatial distribution of hydrogeologic parameters in an aquifer is important to understanding the hydrodynamics of a groundwater flow system. The operational procedure presented in this paper uses core permeability and porosity data and geophysical logs to characterize hydrogeologic parameters, especially hydraulic conductivity (K). The procedure is illustrated with a geostatistical analysis of the permeability distribution along a 120 km cross section of the Milk River aquifer in Alberta, Canada. Geologic and hydrogeologic data from aquifers come in a variety of forms. In deep, regional aquifers, the most ubiquitous form usually is geophysical logs that are used to determine spatial variations in the thickness, porosity, and permeability as well as other rock properties of hydrostratigraphic units. Several methods of deriving hydraulic conductivity values from geophysical logs are evaluated with respect to the Milk River aquifer. Based on a statistical evaluation, a direct relation between porosity and permeability was selected. Once the hydrogeologic data were analyzed and evaluated, a stochastic approach using Bayesian updating with Cholesky decomposition is used to describe the spatial heterogeneity of hydraulic conductivity. This approach produces random-correlated fields of hydraulic conductivity that are conditioned at specific locations by the geophysically derived hydraulic conductivity values. The conditioned, random-correlated fields of hydraulic conductivity are a description of relatively small-scale heterogeneity in the hydraulic conductivity field that can be used in a numerical transport model as a detailed, spatial description of hydraulic conductivity.