Joint and sequential inversion of geophysical and hydrogeological data to characterize seawater intrusion models

The integrated water resource management problems require studying efficiently
seawater intrusion at local and regional scale and identifying in real time the
seawater/freshwater interface dynamic. Hydrogeological modeling is widely used to
predict seawater intrusion if additional natural or man‐made factors are modified.
These models are currently calibrated using measured heads and salt mass fractions in
boreholes, which generally result in sparse data coverage. Within this scope, non to
minimally invasive geophysical techniques like electrical resistivity tomography (ERT) or
time‐domain electromagnetic method are becoming increasingly popular, given their
relatively greater spatial horizontal resolution compared to borehole observations. We
present a comparison between both a sequential and joint approach to calibrate
seawater intrusion models using ERT. The former consists of constraining
hydrogeological parameters using ERT derived parameters and relies on sequential
inversions of the geophysical and hydrogeological data using a given petrophysical
relationship. The second approach in based on simultaneous inversion of petrophysical
and hydrogeological data using electrical resistance data as data. It is performed by
coupling an inversion code with two hydrogeological and geophysical modeling codes
through a petrophysical conversion. This investigation was performed on a densitydependent
flow and transport numerical (three/two‐dimensional) simulation study
from complex and realistic heterogeneous models. In the sequential approach, the
simulations showed that only the shallow salt concentration of the seawater/freshwater
transition zone could be recovered for different time‐lapse, due to poorly resolved
regions in depth. The capability of image appraisal indicators (cumulative sensitivity and
resolution) has been analyzed to emphasize the discrepancy between the targeted and
imaged parameter values. On the other hand, the preliminary coupled inversion avoids
the regularization bias introduced by ERT and addresses the non‐stationarity of the
petrophysical relationship.