Coupled Hydrogeophysical Inversion for Characterizing Heterogeneous Permeability Field at a Groundwater-River Water Interaction Zone

Abstract:

The hydrological and biogeochemical processes at the groundwater and river water interface are largely controlled by the exchange dynamics between the two water bodies. Accurate characterization of the heterogeneous permeability field at such interface is critical for modeling the bulk flow as well as the biogeochemical processes that are coupled with the flow. Taking advantage of the distinct conductivities in groundwater and rive water, time lapse electrical resistivity tomography (ERT) can provide rich spatial and temporal data for characterizing the permeability field, by imaging the change in subsurface electric conductivity driven by river water intrusion and retreat. We installed a large-scale (300 m by 300 m) 3-dimensional ERT array to monitor river water intrusion and retreat through time at a major river corridor, and the 4-dimensional electrical geophysical data is assimilated to invert for the underlying permeability field using ensemble-based algorithms (e.g., ensemble Kalman filter and ensemble smoother). We developed a new high-performance hydrogeophysical code by coupling an ERT imaging code E4D (Johnson et al., 2010) with a site-scale flow and transport code, PFLOTRAN (Hammond et al., 2012). The coupled code provides the key modeling capability of multi-physics processes, parallel efficiency, and multi-realization simulation capability for hydrogeophysical inversion. We assimilated both well-based point measurements of water table and specific conductance and spatially continuous ERT images in a sequential Bayesian way. Our study demonstrates the effectiveness of joint hydrogeophysical inversion for large-scale characterization of subsurface properties in the groundwater and river water interaction zone. Our investigation of spatial versus temporal data assimilation strategies have inspired systematic data worth analyses to identify the most valuable data sets for hydrogeophysical inversion. The high performance computing is performed on the Hopper supercomputer at NERSC. The main scientific achievements of this study are to integrate multiscale and multi-type data with increasingly complex multi-physics model through Bayesian data assimilation techniques, and to handle the intensive computational demands by using high performance computing resources.