Regionalization of Hydraulic Conductivity in Porous Media Aquifers Using Genetic Algorithm

Abstract:

The development of groundwater flow and contaminant transport models require effective hydrogeological characterization of the aquifer system. Extensive field investigations are thus adopted, albeit at sparse locations since the available resources are typically limited. Spatial interpolation (e.g., kriging, inverse distance weighting) is normally then employed to provide approximate estimates of the hydrogeologic properties between measurement locations. However, the use of these techniques typically results in smooth hydrogeological maps that do not capture the heterogeneity. The present study, alternatively, employed the genetic algorithm (GA) to approximate the spatial distribution of the hydraulic conductivity in the Paskapoo aquifer based on sparse observations. The field observations were fit to a statistical distribution, which was used to constrain the GA individuals during initialization and reproduction. As the GA solution depends on the initial guess, particularly when limited resources (e.g., computational time, storage) are available, multiple initial populations were generated and the corresponding GA solutions were obtained. The ensemble average hydraulic conductivity field was subsequently estimated and compared to actual observations, as well as the results from kriging and inverse distance weighting interpolations, at selected locations. The use of GA effectively replicated the actual hydraulic conductivity measurements at the selected locations with a correlation coefficient of approximately 0.95. The results of this study support the potential of using GA, over other spatial interpolation techniques, to capture heterogeneity in estimates of the spatial distribution of hydraulic conductivity in porous media aquifers. This is crucial for the design of effective aquifer planning and management strategies.