Hydraulic properties of a model dike from coupled Bayesian and multi-criteria hydrogeophysical inversion

Abstract

Coupled hydrogeophysical inversion aims to improve the use of geophysical data for hydrological model parameterization. Several numerical studies have illustrated the feasibility and advantages of a coupled approach. However, there is still a lack of studies that apply the coupled inversion approach to actual field data. In this paper, we test the feasibility of coupled hydrogeophysical inversion for determining the hydraulic properties of a model dike using measurements of electrical resistance tomography (ERT). Our analysis uses a two-dimensional (2D) finite element hydrological model (HYDRUS-2D) coupled to a 2.5D finite element electrical resistivity code (CRMOD), and includes explicit recognition of parameter uncertainty by using a Bayesian and multiple criteria framework with the DREAM and AMALGAM population based search algorithms. To benchmark our inversion results, soil hydraulic properties determined from ERT data are compared with those separately obtained from detailed in situ soil water content measurements using Time Domain Reflectometry (TDR). Our most important results are as follows. (1) TDR and ERT data theoretically contain sufficient information to resolve most of the soil hydraulic properties, (2) the DREAM-derived posterior distributions of the hydraulic parameters are quite similar when estimated separately using TDR and ERT measurements for model calibration, (3) among all parameters, the saturated hydraulic conductivity of the dike material is best constrained, (4) the saturation exponent of the petrophysical model is well defined, and matches independently measured values, (5) measured ERT data sufficiently constrain model predictions of water table dynamics within the model dike. This finding demonstrates an innate ability of ERT data to provide accurate hydrogeophysical parameterizations for flooding events, which is of particular relevance to dike management, and (6) the AMALGAM-derived Pareto front demonstrates trade-off in the fitting of ERT and TDR measurements. Altogether, we conclude that coupled hydrogeophysical inversion using a Bayesian approach is especially powerful for hydrological model calibration. The posterior probability density functions of the model parameters and model output predictions contain important information to determine if geophysical measurements provide constraints on hydrological predictions.