Global probabilistic inversion within the latent space learned by a Generative Adversarial Network (GAN) has been recently demonstrated. Compared to inversion on the original model space, using the latent space of a trained GAN can offer the following benefits: (1) the generated model proposals are geostatistically consistent with the prescribed prior training image (TI), and (2) the parameter space is reduced by orders of magnitude compared to the original model space. Nevertheless, exploring the learned latent space by state-of-the-art Markov chain Monte Carlo (MCMC) methods may still require a large computational effort. As an alternative, parameters in this latent space could possibly be optimized with much less computationally expensive gradient-based methods. This study shows that due to the typically highly nonlinear relationship between the latent space and the associated output space of a GAN, gradient-based deterministic inversion may fail even when considering a linear forward physical model. We tested two deterministic inversion approaches: a quasi-Newton gradient descent using the Adam algorithm and a Gauss-Newton (GN) method that makes use of the Jacobian matrix calculated by finite-differencing. For a channelized binary TI and a synthetic linear crosshole ground penetrating radar (GPR) tomography problem involving 576 measurements with low noise, we observe that when allowing for a total of 10,000 iterations only 13% of the gradient descent trials locate a solution that has the required data misfit. The tested GN inversion was unable to recover a solution with the appropriate data misfit. Our results suggest that deterministic inversion performance strongly depends on the inversion approach, starting model, true reference model, number of iterations and noise realization. In contrast, computationally-expensive probabilistic global optimization based on differential evolution always finds an appropriate solution.

Abstract. Surface electrical resistivity tomography (ERT) is a widely used tool to study seawater intrusion (SWI). It is noninvasive and offers a high spatial coverage at a low cost, but its imaging capabilities are strongly affected by decreasing resolution with depth. We conjecture that the use of CHERT (cross-hole ERT) can partly overcome these resolution limitations since the electrodes are placed at depth, which implies that the model resolution does not decrease at the depths of interest. The objective of this study is to test the CHERT for imaging the SWI and monitoring its dynamics at the Argentona site, a well-instrumented field site of a coastal alluvial aquifer located 40 km NE of Barcelona. To do so, we installed permanent electrodes around boreholes attached to the PVC pipes to perform time-lapse monitoring of the SWI on a transect perpendicular to the coastline. After 2 years of monitoring, we observe variability of SWI at different timescales: (1) natural seasonal variations and aquifer salinization that we attribute to long-term drought and (2) short-term fluctuations due to sea storms or flooding in the nearby stream during heavy rain events. The spatial imaging of bulk electrical conductivity allows us to explain non-monotonic salinity profiles in open boreholes (step-wise profiles really reflect the presence of freshwater at depth). By comparing CHERT results with traditional in situ measurements such as electrical conductivity of water samples and bulk electrical conductivity from induction logs, we conclude that CHERT is a reliable and cost-effective imaging tool for monitoring SWI dynamics.

River restoration projects have been launched over the last two decades to improve the ecological status and water quality of regulated rivers. As most restored rivers are not monitored at all, it is difficult to predict consequences of restoration projects or analyze why restorations fail or are successful. It is thus necessary to implement efficient field assessment strategies, for example by employing sensor networks that continuously measure physical parameters at high spatial and temporal resolution. This paper focuses on the design and implementation of an instrumentation strategy for monitoring changes in bank filtration, hydrological connectivity, groundwater travel time and quality due to river restoration. We specifically designed and instrumented a network of monitoring wells at the Thur River (NE Switzerland), which is partly restored and mainly channelized since more than 100 years. Our results show that bank filtration – especially in a restored section with alternating riverbed morphology – is variable in time and space. Consequently, our monitoring network sensing physical and sampling chemical water quality parameters was adapted in response to that variability. Although not available at our test site, we consider long-term measurements – ideally initialized before and continued after restoration – as a fundamental step, towards predicting consequences of river restoration for groundwater quality. As a result, process-based models could be adapted and evaluated using these types of high-resolution data sets

The vadose zone is the main host of surface and subsurface water exchange and has important implications for ecosystems functioning, climate sciences, geotechnical engineering, and water availability issues. Geophysics provides a means for investigating the subsurface in a non-invasive way and at larger spatial scales than conventional hydrological sensors. Time-lapse hydrogeophysical applications are especially useful for monitoring flow and water content dynamics. Largely dominated by electrical and electromagnetic methods, such applications increasingly rely on seismic methods as a complementary approach to describe the structure and behavior of the vadose zone. To further explore the applicability of active seismics to retrieve quantitative information about dynamic processes in near-surface time-lapse settings, we designed a controlled water infiltration experiment at the Ploemeur Hydrological Observatory (France) during which successive periods of infiltration were followed by surface-based seismic and electrical resistivity acquisitions. Water content was monitored throughout the experiment by means of sensors at different depths to relate the derived seismic and electrical properties to water saturation changes. We observe comparable trends in the electrical and seismic responses during the experiment, highlighting the utility of the seismic method to monitor hydrological processes and unsaturated flow. Moreover, petrophysical relationships seem promising in providing quantitative results.

No abstract