Hydraulic Head Data Research Articles

Surface Water and Groundwater Recharge Modeling in Sunsari District using Integrated SWAT-MODFLOW Model

This study employs an integrated groundwater-surface water (GW-SW) model, using SWAT-MODFLOW, to assess groundwater recharge and withdrawals at a regional scale in eastern Nepal. The model components were calibrated and validated using monthly river flow and hydraulic head data for the 1979-2017 period. Our results demonstrate that the average annual groundwater discharge is 2.43 m3/s. Seasonally, the average groundwater discharge is 2.73 m3/s during Summer and is 2.13 m3/s during winter. We observe that the maximum (6.9 m3/s) and minimum (0.036 m3/s) rates occurred in September-October and March-April, respectively. We also find that the average runoff fraction (i.e., runoff/ precipitation) varied from 0.34 to 0.72, and the ET component varied from 0.24 to 0.547 of the total precipitation within the SWAT model. The yearly fluctuation in TWS varied from ± 9.9% to ± 58% of the total precipitation. The Aquifer thickness varies from 16m to 57 m, and the specific yield in the Sunsari district varied from 0.12 lps to 890 lps. In the southern part, we see the water level declines from 2-4 m from July to August, where the bottom water slowly declines in September. The recharge of 1380 l/sec is observed in the Sapta Koshi area in the Bhabar zone. There is substantial underflow from hills simultaneously, resulting in the total recharge in the Sunsari district of 1.98 MCM/yr. At the same time, the total discharge, including surface, sub-surface, and spring water discharge, is nearly 29.2 MCM/yr. This study shows that the Sunsari district’s water resources have many groundwater and surface water, properly managed for future use.

Characterization of the non-Gaussian hydraulic conductivity field via deep learning-based inversion of hydraulic-head and self-potential data

Accurate characterization of the spatial heterogeneity of hydraulic properties such as hydraulic conductivity (K) is essential for understanding groundwater flow and contaminant transport processes. Deep learning-based ensemble smoother methods have proven to be effective in estimating non-Gaussian K fields by assimilating hydrodynamic measurements. However, traditional hydrogeological investigations (e.g., hydraulic tomography) often suffer from the limited number of invasive borehole data (e.g., hydraulic head), making the non-Gaussian K inversion problem strongly underdetermined. As a non-invasive and high-sampling-density geophysical approach, the self-potential (SP) method can measure fluctuations of the electrical field caused by subsurface fluid movement. Thus, the recorded SP signals are electro-kinetic responses of the groundwater flow and can provide complementary information for delineating non-Gaussian heterogeneity. In this paper, we performed a joint hydrogeophysical inversion of hydraulic-head and SP data to characterize the non-Gaussian K field within a deep learning-based inversion framework. We conducted three synthetic cases in a 3-D non-Gaussian aquifer with 16,000 unknown K, to assess the ability of the proposed joint inversion framework for K imaging by assimilating different types of data. The results show that using limited hydraulic-head data alone may capture the large-scale features of the non-Gaussian reference field but might lose fine features. By integrating the hydraulic-head and SP data, the non-Gaussian K field can be reconstructed with an improved accuracy and reduced estimation uncertainty. The low-cost SP method opens up new opportunities for non-Gaussian K characterization and fluid-dynamic monitoring.

Mapping discrete fracture networks using inversion of hydraulic tomography data with convolutional neural network: SegNet-Fracture

In this paper, we propose a new method to map the fracture network structure in a heterogeneous aquifer from inversion hydraulic head data measured during pumping tests in hydraulic tomography mode. This inversion tool is based on the new concept of convolutional neural networks, which provides a direct approximation to the inverse function linking fracture geometry to hydraulic data. In order to handle the highly nonlinear inverse function more effectively, an advanced neural network is developed from SegNet architecture with encoder-decoder structure, which excels in image processing to translate the water level image associated with the pumping tests at the input into a fracture map at the output. The network is trained with a synthetic dataset where the fracture structure and matrix heterogeneity are randomly generated, and the hydraulic head are obtained by solving the groundwater flow equation. The trained network accurately maps different complexity levels of fractures embedded in a matrix with heterogeneous transmissivity.As a data-driven approach, the accuracy of the mapping depends on the quantity, quality, and relevance of the synthetic dataset used in the training phase. While generating data to train the network requires effort, the trained network performs each inversion instantly. The inversion result appears to be stable even in the presence of data noise, reliably interprets the hydraulic data if they share comparable fracture and matrix properties as specified in the training models.

Simultaneous Estimation of a Contaminant Source and Hydraulic Conductivity Field by Combining an Iterative Ensemble Smoother and Sequential Gaussian Simulation

Joint estimation of groundwater contaminant source characteristics and hydraulic conductivity is of great significance for contaminant transport models in heterogeneous subsurface media. As for accurate characterization of hydraulic conductivities, both geostatistical modeling and groundwater inverse modeling are alternative approaches. In this study, an iterative ensemble smoother and sequential gaussian simulation (SGSIM) in geostatistics modeling were combined to realize the simultaneous inversion of contaminant sources and hydraulic conductivities, by using directly measured hydraulic conductivities and indirect hydraulic head and concentration data. To alleviate the high computational cost caused by repetitive evaluations of complex, high-dimensional groundwater models, SGSIM with the pilot points method was used. Considering the characteristics of the proposed method, four scenarios with ten cases were set up in terms of ensemble number and iteration number that affect the performance of the iterative ensemble smoother, the number of pilot points, and the observation data, respectively. The results for the synthetic example indicate that the ensemble size of 2000 and the pilot point number of 80 is an ideal combination of parameters, and the proposed method can successfully recover contaminant source information simultaneously with hydraulic conductivity.

Distinguishing Mountain Front and Mountain Block Recharge in an Intermontane Basin of the Himalayan Region.

Intermontane basin aquifers worldwide, particularly in the Himalayan region, are recharged largely by the adjoining mountains. Recharge in these basins can occur either by water infiltrating from streams near mountain fronts (MFs) as mountain front recharge (MFR) or by sub-surface mountain block infiltration as mountain block recharge (MBR). MFR and MBR recharge are challenging to distinguish and are least quantified, considering the lack of extensive understanding of the hydrological processes in the mountains. This study used oxygen and hydrogen isotopes (δ18 O and δ2 H), electrical conductivity (EC) data, hydraulic head, and groundwater level data to differentiate MFR and MBR. Groundwater level data provide information about the groundwater-surface water interactions and groundwater flow directions, whereas isotopes and EC data are used to distinguish and quantify different recharge sources. The present methodology is tested in an intermontane basin of the Himalayan region. The results suggest that karst springs (KS) and deep groundwater (DGW) recharge are dominated by snowmelt (47% ± 10% and 46% ± 9%) as MBR from adjacent mountains, insignificantly affected by evaporation. The hydraulic head data and isotopes indicate Quaternary shallow groundwater (SGW) aquifer system recharge as MFR of local meteoric water with significant evaporation. The results indicate several flow paths in the aquifer system, a local flow for KS, intermediate flow for SGW, and regional flow for DGW. The findings will significantly impact water resource management in the area and provide vital baseline knowledge for sustainable groundwater management in other Himalayan intermontane basins.

Using Freshwater Heads to Analyze Flow Directions in Saline Aquifers of the Pingtung Plain, Taiwan

The hydraulic head is the most important parameter for the study of groundwater. However, a head measured from observation wells containing groundwater of variable density should be corrected to a reference density (e.g., a freshwater head). Some previous case studies have used unknown density hydraulic heads for calibrating flow models. Errors arising from the use of observed hydraulic head data of unknown density are, therefore, likely one of the most overlooked issues in flow simulations of seawater intrusion. Here, we present a case study that uses the freshwater head, instead of the observed hydraulic head, to analyze the flow paths of saline groundwater in the coastal region of the Pingtung Plain, Taiwan. Out of a total of 134 observation wells within the Pingtung Plain, 19 wells have been determined to be saline, with Electric Conductivity (EC) values higher than 1500 μS/cm during 2012. The misuse of observed hydraulic heads causes misinterpretation of the flow direction of saline groundwater. For such saline aquifers, the determination of a freshwater head requires density information obtained from an observation well. Instead of the purging and sampling method, we recommend EC logging using a month interval. Our research indicates that EC values within an observation well within saline aquifers vary not only vertically but also by season.

Use of convolutional neural networks with encoder-decoder structure for predicting the inverse operator in hydraulic tomography

In this manuscript, we discuss the capabilities of a deep learning algorithm implemented with the Conventional Neural Network concept to characterize the hydraulic properties of aquifers. The algorithm called CNN-HT is designed to predict the inverse operator of hydraulic tomography using a synthetic training dataset in which the hydraulic head data associated with pumping tests are linked to hydraulic transmissivity field. This approach relies on an adaptation of the SegNet network that was initially developed to process image segmentation. The SegNet is composed of encoders and decoders networks. In the encoder, sequential operations with multiple filters, as convolution, batch normalization, max-pooling are performed to identify feature maps of the input data. In the decoder, the up-sampling, convolution, batch normalization and regression operations are used to prepare the output by recovering the loss of spatial resolution that occurred in the encoder process. In this adaptation, we used the least-square iterative formulation at the initial iteration with Jacobian matrix to resize the hydraulic head data to match the size of the output (transmissivity field). This protocol was applied to the hydraulic head data computed numerically by solving the groundwater flow equation for a given transmissivity field, generated geostatistically with Gaussian and spherical variograms. A part of this data was used for training the network and the other part to test its performance. The test step confirmed the effectiveness of this tool in reconstructing the main heterogeneities of the hydraulic properties, and its effectiveness is related to the nature and quantity of the training data. Moreover, the CNN-HT method provided inversion results of the same quality than those obtained with the Gauss-Newton algorithm using the finite difference or adjoint state method in the computation of the Jacobian matrix. However, the computational time is longer in CNN-HT but this time can be less or of the same order as that of Gauss-Newton using finite difference method.

Hamiltonian Monte Carlo for Simultaneous Interface and Spatial Field Detection (HMCSISFD) and its application to a piping zone interface detection problem

AbstractIn practice, inverse problems related to explicit interface detection rely on scarcely available information of the background continuous spatial fields. To accurately carry on inversion in such cases, a method called Hamiltonian Monte Carlo for Simultaneous Interface and Spatial Field Detection (HMCSISFD) is developed. Update of the interface parameters is done in a reversible mesh moving framework leading to a change in domain geometry which forces a recomputation of the spatial field covariance function as it is domain dependent. The eigenvalue problem and its derivatives w.r.t. the parameters are solved again on the updated domain to enable dimensionality reduction of the spatial field parameters through the use of the truncated Karhunen‐Loève expansion. This simultaneous parameter update procedure is applied to the detection of the interface and spatial field properties of a piping zone. Inversion is done considering hydraulic head and discharge rate data, with a priori known observation noise, from a carefully designed laboratory seepage flow experiment in a domain consisting a predefined piping zone. The static pipe in the experiment is representative of a pipe in equilibrium beneath an embankment/levee where the current hydraulic gradient isn't sufficient to cause further pipe progression. Markov chains obtained from the HMCSISFD method show convergence and the true interface and spatial field parameters are found to lie well within the 95% high probability density regions and the 0.05 and 0.95 quantiles, respectively, provided a proper choice for the observation noise is made.

Bayesian Inversion of Multi‐Gaussian Log‐Conductivity Fields With Uncertain Hyperparameters: An Extension of Preconditioned Crank‐Nicolson Markov Chain Monte Carlo With Parallel Tempering

AbstractIn conventional Bayesian geostatistical inversion, specific values of hyperparameters characterizing the prior distribution of random fields are required. However, these hyperparameters are typically very uncertain in practice. Thus, it is more appropriate to consider the uncertainty of hyperparameters as well. The preconditioned Crank‐Nicolson Markov chain Monte Carlo with parallel tempering (pCN‐PT) has been used to efficiently solve the conventional Bayesian inversion of high‐dimensional multi‐Gaussian random fields. In this study, we extend pCN‐PT to Bayesian inversion with uncertain hyperparameters of multi‐Gaussian fields. To utilize the dimension robustness of the preconditioned Crank‐Nicolson algorithm, we reconstruct the problem by decomposing the random field into hyperparameters and white noise. Then, we apply pCN‐PT with a Gibbs split to this “new” problem to obtain the posterior samples of hyperparameters and white noise, and further recover the posterior samples of spatially distributed model parameters. Finally, we apply the extended pCN‐PT method for estimating a finely resolved multi‐Gaussian log‐hydraulic conductivity field from direct data and from head data to show its effectiveness. Results indicate that the estimation of hyperparameters with hydraulic head data is very challenging and the posterior distributions of hyperparameters are only slightly narrower than the prior distributions. Direct measurements of hydraulic conductivity are needed to narrow more the posterior distribution of hyperparameters. To the best of our knowledge, this is a first accurate and fully linearization free solution to Bayesian multi‐Gaussian geostatistical inversion with uncertain hyperparameters.

Application of Indicator Kriging to hydraulic head data to test alternative conceptual models for spring source aquifers

Hydraulic head distributions can inform spring source-water characterisation by determining whether aquifers meet the thresholds required to sustain spring flow. Assessing hydraulic head data can be challenging in areas where data are sparsely distributed and subject to variable measurement uncertainty. Geostatistical methods can be used to estimate hydraulic head values at unmeasured locations with quantitative uncertainty estimates. While these methods have been applied extensively for hydraulic head estimation in management contexts, no studies have used these approaches for spring source-water conceptual model testing. In this study, an investigation into the hydraulic head distribution was conducted through the application of Ordinary Indicator Co-Kriging. Interpolated hydraulic head estimates were used to quantitatively assess the plausibility of source-aquifer conceptual models for the Doongmabulla Springs Complex (DSC), Queensland, Australia. The results offer insights into the likelihood of alternative source aquifers having adequate head to support flow to eight springs within the DSC. Probabilities of adequate head to support the springs ranged from 0.03 to 0.12 for the Permian Formations, compared to <0.01 to 0.49 for the Triassic Formations. Analyses indicated that the Triassic Formations are more likely to have adequate hydraulic head to support spring flow. However, significant uncertainty exists in the conceptual model assessment due to hydraulic head measurement scarcity, particularly in the vicinity of the springs. These findings have important implications given that the Permian Formations will be dewatered by the operation of the nearby Carmichael coal mine. The techniques employed here can inform conceptual model uncertainties arising from the interpretation of sparsely distributed hydraulic head datasets, a major benefit over traditional interpolation methods.

Geochemical and isotopic evidence for pumping-induced impacts to bedrock groundwater quality in the City of Guelph, Canada

Groundwater can be a key water resource in urban environments, providing a source of freshwater for community needs. The City of Guelph relies on groundwater for a population of over 130,000 people. Thus, groundwater quality is a critical component of long-term water management. This study assesses the effect of urban, industrial and agriculture activities on groundwater quality using a suite of multidisciplinary methods including geological, hydrological, geochemical, and isotopic tools. Multi-level systems (with depth-discrete ports), conventional well clusters, and municipal production wells were used to monitor hydraulic head and collect groundwater samples. Geochemical and isotopic measurements included redox parameters, major anions and cations, VOC, tritium and δ34S and δ18O in sulfate. Hydraulic head data show vertical gradient profiles characteristic of a multi-layered flow system within a shallow and deep Silurian bedrock aquifer and an intermediate aquitard varying in occurrence. Head loss disappears near production wells, showing enhanced vertical hydraulic connectivity between shallow and deep aquifers, attributed to pumping and long open intervals. Hydrochemical data show no impact of nitrate and high chloride is observed in the shallow and deep aquifer, attributed to seasonal road salt and the underlying aquitard unit, respectively. The aquitard unit also control the high sulfate in the deep aquifer which is supported by the isotope data on sulfate. Facilitated transport of shallow groundwater into the deeper aquifer is supported by the tritium data, showing the presence of recent groundwater throughout the whole depth of the two aquifers in some areas, likely due to the varying aquitard presence and enhanced vertical flow due to proximity from pumping municipal wells and active rock quarry. The results of this research suggest long term implications for groundwater resource management of sedimentary bedrock aquifers, where there is increasing groundwater demand due to population growth and potential for continual water quality degradation.

Spatio-temporal estimation of monthly groundwater levels from GPS-based land deformation

Significant subsidence is susceptible to groundwater level variations in aquifer systems. The relation between groundwater level change and global positioning system (GPS) estimated subsidence is spatially variable. Time-dependent spatial regression can be used for the estimation of groundwater level changes using GPS based deformation data. Furthermore, the model can be validated using observed hydraulic head data from available monitoring stations. This study uses GPS station data to estimate the monthly groundwater levels in the west-central Taiwan for the period: 2016–17.Time-dependent spatial regression provides a more realistic estimation of groundwater level changes in response to highly heterogeneous aquifer properties than other methods. The high correlation (r = 0.95) between observed and estimated groundwater levels shows that GPS estimated deformations represent an alternative approach for estimating seasonal groundwater changes. Due to availability of spatially broad/low cost GPS data (compared to the sparse availability groundwater monitoring stations), the use of GPS data represents a powerful solution for future monitoring of estimated seasonal groundwater level changes in areas where only few groundwater observations are available.

Integrating hydraulic tomography, electrical resistivity tomography, and partitioning interwell tracer test datasets to improve identification of pool-dominated DNAPL source zone architecture

High-resolution characterization of complex dense non-aqueous phase liquid (DNAPL) contaminated sites is crucial for developing effective remediation strategies. The DNAPL source zone is usually characterized by hydraulic/partitioning tracer tomography (HPTT). However, the HPTT method may fail to capture the highly saturated pool-dominated DNAPL source zone architecture (SZA), because partitioning tracers tend to bypass the low-permeability zones where the pool DNAPL accumulates, resulting in a low-resolution DNAPL estimation. With a limited number of measurements, the estimation errors may be significant. To overcome these difficulties, time-lapse electrical resistivity tomography (ERT) was integrated with the partitioning interwell tracer test (PITT) and hydraulic tomography (HT) to characterize the pool-dominated DNAPL SZA. Herein, we proposed an iterative joint inversion framework coupling the multiphase flow model with the ERT forward model to estimate the heterogeneous permeability distribution and DNAPL SZA. Under this framework, permeability was estimated using the hydraulic head data from HT in stage 1, and the DNAPL SZA was subsequently estimated by assimilating both the PITT and ERT observations in stage 2. The permeability estimated from stage 1 was used as prior information for stage 2, and the DNAPL saturation estimated from stage 2 was served as prior information for stage 1 in the next loop to form an iterative loop to improve the estimation of both permeability and DNAPL SZA. The iterative joint inversion framework was evaluated in two numerical experiments with different heterogeneous structures by assimilating multi-source datasets, including hydraulic head, partitioning interwell tracer concentration, and electrical resistivity. Results show that with limited measurements of HPTT method, one can roughly capture the DNAPL distribution, missing the fine structure of the DNAPL SZA. In contrast, by incorporating multi-source datasets, the heterogeneous permeability and DNAPL SZA can be reconstructed with a higher resolution. Furthermore, the inversion accuracy of the DNAPL SZA improves progressively as the iteration proceeds, which demonstrates the advantage of utilizing complementary information from permeability and DNAPL distribution through the iteration framework. Comparing with the results without loop iteration, the estimation error is reduced by 17.3% for permeability and 8.6% for DNAPL saturation in Experiment 1; by 14.7% for permeability and 11.2% for DNAPL saturation in Experiment 2 through the iterative framework. To further evaluate our framework, we preformed the prediction of the depletion process of the DNAPL source zone and plume based on the estimated DNAPL SZA. Results show that using the iterative framework, the prediction of the SZA depletion is greatly improved, i.e., the estimation error of the dissolved downstream plume from the DNAPL source zone after 3 years is reduced by 20.9% in Experiment 1, and by 43.2% in Experiment 2, respectively, through the iterative framework. This significant improvement is because the iterative method can better capture the spread of DNAPL pool.

Contaminant Spill in a Sandbox with Non-Gaussian Conductivities: Simultaneous Identification by the Restart Normal-Score Ensemble Kalman Filter

The joint identification of the parameters defining a contaminant source and the heterogeneous distribution of the hydraulic conductivities of the aquifer where the contamination took place is a difficult task. Previous studies have demonstrated the applicability of the restart normal-score ensemble Kalman filter (rNS-EnKF) in synthetic cases making use of sufficient hydraulic head and concentration data. This study shows an application of the same technique to a non-synthetic case under laboratory conditions and discusses the difficulties found on its application and the avenues taken to solve them. The method is first tested using a synthetic case that mimics the sandbox experiment to establish the minimum number of ensemble members and the best technique to prevent the filter collapsing. The synthetic case shows that among different techniques based on update damping and covariance inflation, the Bauser’s covariance inflation method works best in preventing filter collapse. Its application to the sandbox data shows that the rNS-EnKF can benefit from Bauser’s inflation to reduce the number of ensemble realizations substantially in comparison with a filter without inflation, arriving at a good joint identification of both the contaminant source and the spatial heterogeneity of the conductivities.

3D modelling of solute transport and mixing during managed aquifer recharge with an infiltration basin

Artificial basins are used to recharge groundwater by many municipalities to improve the sustainability of storm water management. Despite its increasing operational implementation, artificial recharge still raises numerous questions related to its impact on groundwater quality. In this paper, a 3D numerical model of MAR basin/aquifer system was implemented in order to simulate the fate of water and pollutants. It was used to illustrate the complex distribution in time and space of a tracer contaminant injected in the basin. The model was based on a well instrumented storm water infiltration basin located in Chassieu (Lyon area, France). The well-known Richards model was used to simulate the water flow in the saturated and unsaturated zone of the study site. The transfer of solutes in the basin/aquifer system was modelled by the advection-dispersion-equation (ADE). The model was calibrated during a rain event using hydraulic head and electric conductivity data from a set of piezometers located around the basin. The flow model was validated on a one month period of basin operation presenting several rain events. The model was then used to simulate the fate of a solute pollutant considered as a tracer during a high intensity rain event. This simplified test case illustrated the mechanism of capillary trapping in the vadose zone and the effect of sampling point location on concentration measurements. Three main results were obtained: (1) capillary trapping promoted a retention of up to 20% of the injected tracer in the vadose zone, (2) 0 to 24% of the injected solute concentration could be recovered depending on the piezometer location, (3) the averaged concentration decreased by 50% if the measuring device is lowered by 5 m under the water table. These results were strongly site and event dependant but observed trends should be considered while discussing punctual water quality measurements used to monitor MAR systems. It also allowed to suggest some guidelines for sampling point positioning.

Implementation of the Kalman Filter for a Geostatistical Bivariate Spatiotemporal Estimation of Hydraulic Conductivity in Aquifers

The estimation of the hydraulic parameters of an aquifer such as the hydraulic conductivity is somehow complicated due to its heterogeneity, on the other hand field and laboratory tests are both time consuming and costly. The use of geostatistical-based techniques for data assimilation could represent an alternative tool that allows the use of space-time aquifer behaviour to characterize hydraulic conductivity heterogeneity. In this paper, a spatiotemporal bivariate methodology was implemented combining historical hydraulic head data with hydraulic conductivity sparse data in order to obtain an estimate of the spatial distribution of the latter variable. This approach takes advantage of the correlation between the hydraulic conductivity (K) and the hydraulic head (H) behaviour through time. In order to evaluate this approach, a synthetic experiment was constructed through a transitory numerical flow-model that simulates hydraulic head values in a horizontally-heterogeneous aquifer. Geostatistical tools were used to describe the correlation between simulated spatiotemporal data of hydraulic head and the spatial distribution of the hydraulic conductivity in a group of model nodes. Subsequently, the Kalman filter was used to estimate the hydraulic conductivity values at nonsampled sites. The results showed acceptable differences between estimated and synthetic hydraulic conductivity data, with low estimate error variances (predominating the 1 m2/day2 value for K for all the cases, however, the smallest number of cells with values above 2 m2/day2 correspond to the bivariate spatiotemporal case) and the best agreement between the estimated errors and the selected model variance (SMSE values of 0.574 and 0.469) were found for the bivariate cases, which suggests that the implemented methodology could be used for reducing calibration efforts, particularly when the hydraulic parameters data are scarce.

Coupling ensemble smoother and deep learning with generative adversarial networks to deal with non-Gaussianity in flow and transport data assimilation

Dynamic data such as hydraulic head and concentration data can be integrated into the groundwater flow and contaminant transport model to improve its predictive ability for groundwater resource management and aquifer remediation. Ensemble Smoother with Multiple Data Assimilation (ES-MDA) has gained popularity for data assimilation in the field of hydrogeology, where aquifer parameters such as hydraulic conductivity are calibrated by conditioning on observed dynamic data. The ES-MDA has an optimal solution if aquifer parameters follow a multi-Gaussian distribution. However, fluvial deposits commonly exhibit a strong heterogeneity with channels (i.e., connectivity). In other words, the hydraulic conductivity does not follow the multi-Gaussian distribution. To deal with data assimilation in channelized aquifers, we propose to couple ES-MDA with deep learning. Specifically, Generative Adversarial Networks (GAN), a deep learning algorithm, are used to re-parameterize the channelized aquifer with a low-dimension latent variable. The ES-MDA is then used to update the latent variable by assimilating dynamic data into the groundwater model. Synthetic studies of groundwater flow and contaminant transport models are used to demonstrate the proposed method. The results illustrate that the coupling of GAN and ES-MDA is able to reconstruct the channel structures and reduce the uncertainty of hydraulic head and contaminant concentration predictions.

Worth of hydraulic and water chemistry observation data in terms of the reliability of surface water-groundwater exchange flux predictions under varied flow conditions

This study assesses the worth of routinely collected hydraulic data (groundwater head, stream stage and streamflow) and lesser collected water chemistry data (Radon-222, Carbon-14, electrical conductivity (EC)) in the context of making regional-scale surface water-groundwater (SW-GW) exchange flux predictions. Using integrated SW-GW flow and transport numerical models, first-order, second-moment (FOSM) analyses were employed to assess the extent of the uncertainty reduction or lack thereof in SW-GW exchange flux predictions following acquisition of hydraulic and water chemistry observation data. With a case study of the Campaspe River in the Murray-Darling Basin (Australia), we explored the apparent information content of these data during low, regular and high streamflow conditions. Also, a range of spatial and temporal prediction scales were considered: catchment-wide and reach-based spatial scales and annual and monthly temporal scales. Generally, the data worth evaluations showed significant variability across predictions that were dependent on the spatiotemporal scale of the SW-GW exchange, the magnitude and direction of the SW-GW exchange flux and the prevailing streamflow conditions. These dependencies serve to emphasise the importance of prediction specificity with respect to SW-GW exchange. Among existing data, the most worth was found in Radon-222, groundwater hydraulic head, EC, and streamflow data showing average reductions in uncertainty of 41%, 38%, 32%, and 23% respectively. Assessment of type and spatiotemporal locations of potential data showed Radon-222 to be the next most important observation type across many predictions in locations with data paucity of all data types. Hydraulic observation data types were found to inform SW-GW exchange flux best under high- and regular- streamflow conditions when the magnitude of exchange fluxes were largest, whereas the water chemistry data was of highest value for low- and regular-streamflow conditions where groundwater is discharging to the stream.

Use of Groundwater Level Fluctuations near an Operating Water Supply Well to Estimate Aquifer Transmissivity.

We developed a method to estimate aquifer transmissivity from the hydraulic-head data associated with the normal cyclic operation of a water supply well thus avoiding the need for interrupting the water supply associated with a traditional aquifer test. The method is based on an analytical solution that relates the aquifer's transmissivity to the standard deviation of the hydraulic-head fluctuations in one or more observation wells that are due to the periodic pumping of the production well. We analyzed the resulting analytical solution and demonstrated that when the observation wells are located near the pumping well, the solution has a simple, Dupuit like form. Numerical analysis demonstrates that the analytical solution can also be used for a quasi-periodic pumping of the supply well. Simulation of cyclic pumping in a statistically heterogeneous medium confirms that the method is suitable for analyzing the transmissivity of weakly or moderately heterogeneous aquifers. If only one observation well is available, and the shift in the phase of hydraulic-head oscillations between the pumping well and the observation well is not identifiable. Prior knowledge of aquifer's hydraulic diffusivity is required to obtain the value of the aquifer transmissivity.

Conditioning Multi-Gaussian Groundwater Flow Parameters to Transient Hydraulic Head and Flowrate Data With Iterative Ensemble Smoothers: A Synthetic Case Study

Over the last decade, data assimilation methods based on the ensemble Kalman filter (EnKF) have been particularly explored in various geoscience fields to solve inverse problems. Although this type of ensemble methods can handle high-dimensional systems, they assume that the errors coming from whether the observations or the numerical model are multivariate Gaussian. To handle existing nonlinearities between the observations and the variables to estimate, iterative methods have been proposed. In this paper, we investigate the feasibility of using the ensemble smoother and two iterative variants for the calibration of a synthetic 2D groundwater model inspired by a real nuclear storage problem in France. Using the same set of sparse and transient flow data, we compare the results of each method when employing them to condition an ensemble of multi-Gaussian groundwater flow parameter fields. In particular, we explore the benefit of transforming the state observations to improve the parameter identification performed by one of the two tested algorithms. Despite the favorable case of a multi-Gaussian parameter distribution considered, we show the importance of defining an ensemble size of at least 200 to obtain sufficiently accurate parameter and uncertainty estimates for the groundwater flow inverse problem considered.