Robin Thibaut

The groundwater-surface water (GW-SW) exchange fluxes are driven by a complex interplay of subsurface processes and their interactions with surface hydrology, which have a significant impact on the water and contaminant exchanges. Due to the complexity of these systems, the accurate estimation of GW-SW fluxes is important for quantitative hydrological studies and should be based on relevant data and careful experimental design. Therefore, the effective design of monitoring networks that can identify relevant subsurface information are essential for the optimal protection of our water resources. In this study, we present novel deep learning (DL)-driven approaches for sequential and static Bayesian optimal experimental design (BOED) in the subsurface, with the goal of estimating the GW-SW exchange fluxes from a set of temperature measurements. We apply probabilistic Bayesian neural networks (PBNN) to conditional density estimation (CDE) within a BOED framework, and the predictive performance of the PBNN-based CDE model is evaluated by a custom objective function based on the Kullback-Leibler divergence to determine optimal temperature sensor locations utilizing the information gain provided by the measurements. This evaluation is used to determine the optimal sequential sampling strategy for estimating GW-SW exchange fluxes in the 1D case, and the results are compared to the static optimal sampling strategy for a 3D conceptual riverbed-aquifer model based on a real case study. Our results indicate that probabilistic DL is an effective method for estimating GW-SW fluxes from temperature data and designing efficient monitoring networks. Our proposed framework can be applied to other cases involving surface or subsurface monitoring and experimental design.

A document by Robin Thibaut. Click on the document to view its contents.

Temperature logs are an important tool in the geothermal industry. Temperature measurements from boreholes are used for exploration, system design, and monitoring. The number of observations, however, is not always sufficient to fully determine the temperature field or explore the entire parameter space of interest. Drilling in the best locations is still difficult and expensive. It is therefore critical to optimize the number and location of boreholes. Due to its higher spatial resolution and lower cost, four-dimensional (4D) temperature field monitoring via time-lapse Electrical Resistivity Tomography (ERT) has been investigated as a potential alternative. We use Bayesian Evidential Learning (BEL), a Monte Carlo-based training approach, to optimize the design of a 4D temperature field monitoring experiment. We demonstrate how BEL can take into account various data source combinations (temperature logs combined with geophysical data) in the experimental design (ED). To optimize the ED and determine the best data source combination, we use the Root Mean Squared Error (RMSE) of the predicted target in the low dimensional latent space where BEL is solving the prediction problem. The generated models agree well with the true models and are accurate enough to be used in optimal ED.Furthermore, the method is not limited to monitoring temperature fields and can be applied to other similar experimental problems.The method is computationally efficient and requires little training data. A training set of only 200 is sufficient for the considered optimal design problem.

Temperature logs are an important tool in the geothermal industry. Temperature measurements from boreholes are used for exploration, system design, and monitoring. The number of observations, however, is not always sufficient to fully determine the temperature field or explore the entire parameter space of interest. Drilling in the best locations is still difficult and expensive. It is therefore critical to optimize the number and location of boreholes. Due to its higher spatial resolution and lower cost, four‐dimensional (4D) temperature field monitoring via time‐lapse Electrical Resistivity Tomography has been investigated as a potential alternative. We use Bayesian Evidential Learning (BEL), a Monte Carlo‐based training approach, to optimize the design of a 4D temperature field monitoring experiment. We demonstrate how BEL can take into account various data source combinations (temperature logs combined with geophysical data) in the Bayesian optimal experimental design (BOED). To determine the optimal data source combination, we use the Root Mean Squared Error of the predicted target in the low dimensional latent space where BEL is solving the prediction problem. The parameter estimates are accurate enough to use in BOED. Furthermore, the method is not limited to monitoring temperature fields and can be applied to other similar experimental design problems. The method is computationally efficient and requires little training data. For the considered optimal design problem, a training set of only 200 samples and a test set of 50 samples is sufficient.

With the growing population and the adverse effects of climate change, the pressure on coastal aquifers is increasing, leading to a larger risk of saltwater intrusion (SI). SI is often complex and difficult to characterize from well data only. In this context, electrical resistivity tomography (ERT) can provide high-resolution qualitative information on the lateral and vertical distribution of salinity. However, the quantitative interpretation of ERT remains difficult because of the uncertainty of petrophysical relationships, the limitations of inversion, and the heterogeneity of aquifers. In this contribution, we propose a methodology for the semiquantitative interpretation of ERT when colocated well data are not available. We first use existing wells to identify freshwater zones and characterize the resistivity response of clayey deposits. Then, we approximate the formation factor from water samples collected in the vicinity of ERT data to derive a resistivity threshold to interpret the saline boundary. We applied the methodology in the shallow aquifers of the Luy River in the Binh Thuan province, Vietnam, where water resources are under pressure due to agricultural, aquacultural, and industrial production. Twenty-one ERT profiles were collected and revealed a much larger intrusion zone, compared to the previous study. Saltwater is present in lowland areas of the left bank over almost the whole thickness of the aquifer, while the right bank is constituted of sand dunes that are filled with freshwater. At a larger distance from the sea, a complex distribution between fresh and saltwater is observed. Our methodology could be applied to other heterogeneous aquifers in the absence of a dense monitoring network.

KEYWORDS: Saltwater intrusion, Geochemistry, Groundwater extraction   With an average annual rainfall of 800-1150 mm/year, the Binh Thuan province is one of the driest places in Vietnam. The quantity and quality of groundwater play a significant role in the agriculture, aquaculture development and daily life of the local communities. In 2012, the national center for water resources (Nawapi, 2012) delineated the seawater intrusion extent in Binh Thuan based on the total dissolved solids (TDS) content of water samples taken from shallow boreholes. The threshold of 3g/L and 1.5g/L were exceeded in the estuaries of the Luy, Long Song and Ca Ty rivers. In recent years, the prolonged droughts combined with the sea level rise and the over-extraction of groundwater during the dry season increased dramatically the seawater intrusion process especially in the estuaries of the province.The geochemistry of groundwater in the Luy River catchment was studied to investigate the contamination of the aquifers and identify the processes taking place. From 1991 to 2015, 98 water samples had been taken from the wells in the area in both dry and rainy seasons. 71% of the water samples were fresh while 21% and 5% were lightly saline and moderately saline respectively. In summer 2020, 110 new water samples from both shallow and deep wells were collected in the Luy river catchment in wells from 3m to 40m. The TDS values are ranging from 105 to 23080 mg/L and can be classified into 4 groups: freshwater (48%), slightly saline (40%), moderately saline (8%) and very saline (4%). The samples show that the seawater intrusion expands not only horizontally at shallow depth along the river but also deeper down the aquifer in most of the study area, what is also confirmed by geophysical data. Freshwater samples were mostly collected at a depth lower than 10m. The chemical composition of water samples were analyzed showing evidence of seawater intrusion, but also the occurrence of freshening processes within the study area. Together with the presence of saltwater at larger depths, this points towards a situation more complex than previously thought. Saltwater intrusions are likely not only related to interaction with the river estuary, but also to the presence of fossil saltwater in the aquifer, and to groundwater pumping and irrigation practices.REFERENCESNAWAPI. 2012. Hydrogeological mapping at scale 1:50000 in Ninh Thuan and Binh Thuan provinces. In Vietnamese.  *Corresponding Author. Email: Linh.PhamDieu@UGent.be