Comparison of coupled hydrogeophysical inversion techniques for salt tracer experiments

Evolutionary optimization algorithms are known as reliable optimization methods, but they are typically associated with a high computational effort. This is especially true for cases where a large number of model parameters are being optimized, as is the case with the application of electrical resistivity tomography (ERT) to infer the spatio-temporal state of the subsurface water or soil hydrological properties. Suitable alternatives are represented by Bayesian methods, such as the ensemble Kalman filter (EnKF). This is a Monte Carlo-based data assimilation approach that can be effectively used for combined state and parameter estimation. In this contribution, we compare a hierarchical approach for optimization using a genetic algorithm (GA), which was specifically developed to reduce the computational effort, with an EnKF-based approach. The test case for this comparison is focused on the determination of hydraulic conductivity from monitoring of salt tracer tests by ERT. We report on the retrieval performance of the two approaches for a two-dimensional synthetic experiment simulating the evolution of a saline tracer in a shallow aquifer and we discuss some of the strengths and weaknesses of the GA and EnKF optimization strategies.