ABSTRACT. Recently we have introduced a dynamic formulation of the PDE-based optimal transport problem with L1 cost, the so-called Monge-Kantorovich equations (MK) proposed by Evans&Gangbo (1999). In our formulation, called Dynamic MK (DMK), the transport density µ and the Kantorovich potential u are solutions of a nonlinearly coupled system consisting of a possibly degenerate elliptic equation in space for u and an autonomous ODE in time for µ. The equilibrium solution of this model converges as t → ∞ towards the solution (µ∗,u∗) of the M-K equations. The dynamics is controlled by a Lyapunov-candidate functional that decreases along the µ(t)-trajectories and reaches its minimum at equilibrium, when µ = µ∗. This functional is the sum of the Dirichlet Energy and the total mass.

We use physical interpretations of this Lyapunov functional to develop models of landscape formation and plant root evolution. Indeed, the Lyapunov-candidate functional can be interpreted as the sum of the cost of transporting mass and the cost of building the transportation infrastructure. In the case of rivers, the functional is the sum of the cost of transporting the sediments plus the cost of erosion, which is proportional to the length of the landscape forming incision. In the case of plant roots, the first term of the Lypaunov functional represents the cost of transporting water and nutrients through the root network, while the second term is proportional to the cost of biomass allocation. Preliminary numerical experiments will be shown to highlight the capabilities of the proposed framework.

ABSTRACT. A conceptual model is presented for flowback of hydrofracturing fluid in an elastic fracture of plane or radial geometry with rigid walls. The fluid follows the power-law or Ellis constitutive equation. The aperture and pressure fields are derived under the lubrication approximation for different boundary conditions and the two rheologies. Late-time scalings for the aperture confirm and extend earlier literature findings. Model results are verified experimentally with an ad hoc device obtaining a good match between theory and experiments.

ABSTRACT. Homogenisation can be combined with a generalised formulation of the multiple rate mass transfer method to obtain closure parameters that include the effect of the Peclet number without fitting experimental data. In this work, we present a numerical approach for calculating these parameters solving the required cell problems. Finally, we show some applications of the upscaled model in heterogeneous fractured media. Our code was implemented using the opensource finite volume library OpenFOAM and it is fully available online.

ABSTRACT. We present a method to infer the characteristics of a discrete fracture network (DFN) by hydraulic or tracer tomography. The experimental set up delivers information on depth-dependent propagation of pressure signals or tracer breakthroughs. These are reproduced to construct the DFN based on statistical information by a reversible-jump Markov chain Monte Carlo method. In our presentation, the probabilistic inversion of a synthetic two-dimensional example is introduced. Based on this, we show how to extend the method to three-dimensional inversion.

ABSTRACT. Session: Environmental flows of complex fluids

Non-Newtonian fluids are commonly used in subsurface applications. However modeling of such fluids has been heretofore intractable, requiring experimentation or microscale simulation of each fluid flowing in each media to be predictive. Here we develop the theory to relate macroscale parameters to microscale quantities which had previously only been posed at the macroscale. We use this theory to derive an a priori relationship between hydraulic resistance and known fluid and system properties.

ABSTRACT. Viscoelastic fluids are non-Newtonian fluids that exhibit the properties of both viscous fluids and elastic solids. In subsurface engineering problems, such as the recovery of hydrocarbons, viscoelastic polymer solutions are commonly used in enhanced oil recovery (EOR). The primary purpose of polymer flooding is to improve mobility control (higher viscosity of displacing phase), but many studies have also shown reduction in residual, capillary-trapped oil [1, 2]. However, due to the complexity in rheology along with interfacial dynamics, it is difficult to fully predict the mechanisms of multiphase viscoelastic flow. In this work, we aim at discovering the pore-scale insights of viscoelastic displacing fluids in recovering trapped non-wetting fluids. A series of micromodel experiments comparing the non-wetting droplet displacement by different solutions in a single pore-throat (contraction-expansion) microchip are performed. The displacement of oil (tetradecane) by a Newtonian fluid (50wt% diluted glycerin) shows that the size of the droplet decreases with increasing flow rates and it completely goes through the pore throat when a critical flow rate is reached. The correlation between the trapped droplet size (L) and the Capillary Number (Ca) follows the liner model L~Ca-1/3 given by Ke et al. [3]. However, when the droplet is displaced by a viscoelastic solution (partially hydrolyzed polyacrylamide 3630s), it never goes through and starts to oscillate periodically at high flow rates. To understand the mechanisms of viscoelastic effects on oscillation and recovery of trapped non-wetting fluids, we further present pore-scale modelling of these processes using a lattice Boltzmann method [4]. The oscillation behaviors are in qualitative agreement with the micromodel experiments. The disorder of streamlines in viscoelastic fluids is reported in the presence of another phase and a vortex downstream of the droplet is found to prevent the droplet from entering the throat, which explains the observation of oscillations. Additional simulations in a "dead-end" tube geometry show that the viscoelastic oscillation helps release of trapped non-wetting droplets. A linear relationship between the average mean vorticity and the square root of the Deborah number before the release of droplets.

References [1] Wang, D. and J. Cheng, et al. (2000). "Viscous-elastic polymer can increase microscale displacement efficiency in cores", SPE 63227. [2] Clarke, A. and A. M. Howe, et al. (2016). "How viscoelastic-polymer flooding enhances displacement efficiency." SPE Journal 21 (03): 675-687. [3] Xu, K. and P. Zhu, et al. (2015). "Microfluidic investigation of nanoparticles’ role in mobilizing trapped oil droplets in porous media." Langmuir 31 (51): 13673-13679. [4] Xie, C. and W. Lei, et al. (2018). "Lattice Boltzmann model for three-phase viscoelastic fluid flow." Physical Review E 97 (2): 023312.

ABSTRACT. We simulate non-Newtonian flow in rough fractures. Random aperture fields with isotropic spatial correlations are generated using an FFT-based algorithm. We first solve the flow with a finite-difference code based on the lubrication approximation for power law and Carreau rheologies (2D approach). We then solve a modified Navier-Stokes equation incorporating both rheologies, in the 3D fracture space (3D approach). The 2D approach’s efficiency to predict the flow distribution in the fracture plane depends on the fracture closure and correlation length.

ABSTRACT. Non-intrusive reduced-order models of high-fidelity, large-scale, parametrized dynamical systems employ different regression-based approaches to establish a mapping from parameter values to projection coefficients in a lower-dimensional reduced-basis space. This work focuses on using reduced-basis spaces generated by POD and evaluates several regression techniques including classical kernel-based methods like radial basis function interpolation, Bayesian non-parametric approaches like Gaussian process regression, and neural network-based machine-learning algorithms. Their comparative accuracy, reliability, and efficiency in multi-query, fast-replay applications involving 2D shallow water equations are studied.

ABSTRACT. A Monte Carlo study is conducted to examine the joint influence of aperture variability and rheological parameters on non-Newtonian flow in a single fracture. The aperture field is generated stochastically and the flow problem is solved under the lubrication approximation. Three fluid rheologies (Newtonian, power-law and Carreau) are considered to ascertain the impact of the rheological law on the effective flow parameters. Extensive simulations over large statistics of fractures allow determining the uncertainty associated with the prediction.

ABSTRACT. Basin scale Integrated Surface and Subsurface Hydrological Models (ISSHM) are of particular interest when low flow processes are of primary concern to guide water policy decision in a climate change context. A Basin Scale ISSHM (HydroGeoShere) has been developed. To reduce computation time, groundwater recharge is computed externally and assigned as the forcing function to the integrated model. The model has been applied to a large basin in the province of Quebec, Canada. It was calibrated with multiple observation types while the Iterative Ensemble Smoother approach was used for the final calibration step at low computational cost.

ABSTRACT. Riverine bathymetry plays an essential role in safe maritime transportation, prediction of land sinking, and flood risk management. Nevertheless, direct measurement of riverbed profile has been proven costly, making alternative approaches such as estimation of bathymetry through surface flow velocity measurement a favorable substitution. Here, we learn the functional relationship between riverine bathymetry and surface flow velocities using autoencoder, as a data-driven algorithm, in the presence of varying boundary conditions and compare our results with both model-based and data-driven approaches.

ABSTRACT. We developed a probabilistic framework to spatially assess the potential for sustainable groundwater development in the arid interior of Australia, based on an explicit definition of sustainability criteria. We iteratively updated the conceptualisation as new information becomes available; the initial desktop-based conceptualisation is updated based on an AEM data and subsequently updated after a drilling campaign. Stochastic grids of hydraulic properties, salinity and water table are generated using data fusion algorithms.

ABSTRACT. The identification of dense non-aqueous phase liquid (DNAPL) distribution in the subsurface is important for optimizing remediation strategies. However, the DNAPL distribution is highly sensitive to subsurface heterogeneity and is controlled by multiphase physics. Typical multiphase models are computationally too expensive to be applied for inverse modeling. Therefore, we first propose a sequential inversion strategy to estimate the hydraulic conductivity (K) and DNAPL saturation (S_N) fields separately, which can implicitly consider the interdependences between K and S_N by using indirect conditioning data. Nevertheless, when the observation data is limited and less informative, ignoring the underlying multiphase physics may result in poor estimation. To improve DNAPL imaging, we further propose a machine learning-based method to capture the underlying multiphase physics between K-S_N. In conclusion, we suggest that: (1) With enough and informative measurements, one can use the sequential strategy to produce accurate DNAPL estimation. (2) With limited data, one can use the machine learning-based method to capture the multiphase physics in a computationally efficient way.

ABSTRACT. Geological storage of CO₂ in deep saline reservoirs is being studied widely to reduce carbon emission from power plants. The physics of CO₂-brine flow at pore-scale is a key part of prediction of amount and fate of residual trapped CO₂. The description of this flow system in the form of pore-level flow models through pore-bodies and pore-throats of an extracted pore-network (PN) from micro-CT images of real rock is a practical approach to obtain important characteristic curves during a drainage-imbibition cycle. However, this description can be improved by more specific and accurate relations for CO₂-brine flow that can come from direct numerical simulation methods. This study presented a new set of pore-level flow models during pore-body filling and snap-off events of imbibition process in PN modeling of CO₂-brine flow. Lattice-Boltzmann simulations were carried out on several designed PN configurations and the threshold local capillary pressure was evaluated to develop modified equations of threshold capillary pressure as a function of shape factor. The modified equations of local capillary were incorporated in a quasi-static PN solver. This modified model resulted a new pattern of invasion during imbibition process due to a different order of competing pore-level events compare to the original model. We applied the modified model on extracted PN of a Berea sandstone sample to obtain relative permeabilities and saturation of residual trapped CO₂ after a drainage-imbibition cycle. The statistics of pore-level imbibition events changed by replacing the original model with the modified model. The occurrence of snap-off in pore-throats was reduced by about 10% which means more frontal displacement pattern across the sample. As a result, our modified model was in closer agreement than the original model based on the comparison of the residual trapped CO₂ with experimental data.

ABSTRACT. Wettability alteration may be a time-dependent process controlled by exposure to a reactive fluid, leading to dynamic changes in constitutive functions over time. New dynamic models for capillary pressure and relative permeability have been developed based on pore-scale simulation of displacement processes where the contact angle changes dynamically. The form of the dynamic models is simple and the parameters can be easily predicted from a laboratory data measuring contact angle change over time. A convergent and robust iterative linearization scheme with a stabilization term was developed for simulation of multiphase flow with a capillary pressure function that is non-local in time. The impact of dynamic wettability alteration on migration and trapping will be studied for realistic CO2 storage scenarios.

ABSTRACT. Traditional macroscale models of two-fluid flow through porous media suffer from a number of shortcomings. Much effort has gone into the parameterization, numerical solution, and application of these models despite their recognized shortcomings. Over the past fifteen years, thermodynamically constrained averaging theory (TCAT) has been developed as a way to address the limitations associated with traditional macroscale formulations. Here, we begin the process of developing numerical techniques for recently formulated TCAT models of two-fluid flow.

ABSTRACT. In this work, we simulate gravity effects, namely gravity override and convective mixing, during miscible displacement of CO2 and oil. The flow behavior due to the competition between viscous and gravity effects is complex, and can only be accurately simulated with a very fine grid. We demonstrate that convection occurs rapidly, and has a strong effect on breakthrough of CO2 at the outlet. The implications for the fine-scale flow behaviour at the field scale have been studied.

ABSTRACT. This work presents a procedure based on the method of Krause et al. (2013) to estimate sub-core permeability distribution and characteristic relative permeability from coreflooding experimental data. The new procedure employs a simplified model for characteristic curve estimation developed in Rabinovich and Darkwah (2019), requiring only solutions of steady state single-phase flow equations. Therefore, the computational time is significantly reduced. The method is also extended to combine data from a number of experiments with different fractional flow of injection.

ABSTRACT. There are many biogeochemical factors simultaneously influencing the fate of chemical compounds in soil; abiotic factors like grain/pore size distribution, saturation state and pore connectivity and biotic factors such as the diversity of microbial community in terms of having aerobic/anaerobic as well as their activities and metabolic pathways. Primary aim of this study is to establish a novel reactive transport modelling approach covering the full range of spatio-temporal heterogeneity at pore scale to assess underlying biogeochemical phenomena in porous media.

ABSTRACT.  Injection of fluid in faulted subsurface formations cause flow, transport and deformation processes that are strongly affected by the pre-existing faults. Simultaneously, faults and fractures are also affected by the processes induced by the injection, causing them to slip, dilate and possibly mechanically open and propagate. The coupled process-structure interaction prevents the identification of simple causality and, in identifying governing mechanisms and forecasting dynamics, mathematical models and simulation technology are valuable tools to complement field and experimental data. Despite limitations in predictive capability, modelling can reveal dynamics that are crucial to understand to improve engineering operations related to subsurface fluid injection and avoid unacceptable environmental impact. In this talk, we consider conceptual, mathematical and numerical modeling of injection-induced reactivation of faults. Based on numerical examples, we study thermo-hydro-mechanical processes interacting with deforming faults and fractures. The  examples illustrate the extreme coupling in the dynamics, showing that it is not possible to study different mechanisms on their own without accounting for their mutual interaction. Finally, we  show recent results on how numerical modeling complement field observations based on a case study of hydraulic stimulation of a well in the Reykjanes Geothermal field.

ABSTRACT. The analysis of the combined impact of climate change and contamination on groundwater is a critical issue at the global scale given the primary role of subsurface water resources to satisfy water demand for the different uses and as a support for surface water bodies and groundwater dependent ecosystems. We suggest an approach based on the Probabilistic Collocation Method which allows accelerating risk assessment to predict groundwater status under future conditions and account for the sustainability of alternative management strategies.

ABSTRACT. Physically-based models, global sensitivity analysis and meta-modeling strategies are merged to quantify fluxes to groundwater from a network of irrigation channels under uncertainty of key parameters and boundary conditions. As the local increase in the phreatic water level allows reducing irrigation needs, our results have been translated into a GIS-based decision support system for irrigation management in the study area. We derived a general formulation able to handle the contribution to groundwater provided by each channel of the irrigation network.

ABSTRACT. Modern reactive transport models (RTMs) contain a wealth of process knowledge about contaminant transport. However, most model simulations are fixed in time and not designed to handle incoming streams of new data; a calibrated RTM, for example, has to be completely re-calibrated from the beginning if new observations are not in agreement with model predictions. With the increasing availability of large geochemical datasets – including high-frequency sensor measurements and long-term monitoring data – we propose a workflow for incorporating both RTM process knowledge and real-time data streams into prediction-focused recurrent neural networks (RNNs). We explore modern variants of gating-based RNNs, including time-aware long short-term memory networks (T-LSTM; Baytas et al., 2017) and continuous-time gated recurrent units (GRU-ODE-Bayes; De Brouwer et al., 2019), then demonstrate their performance on a series of synthetic and real datasets.

ABSTRACT. Evapotranspiration (ET) – the combination of evaporation and transpiration – is a critical component in the water cycle, accounting a significant portion of the water budget. ET has been measured with various approaches, including weighing lysimeters based on the soil water balances, remote sensing techniques based on surface energy balance, and eddy-covariance flux towers based on atmospheric dynamics models. Each approach involves uncertainty, since they rely on indirect measurements and certain assumptions. Although remote sensing techniques can provide spatially extensive data, the uncertainty is known to be quite high. Recent advances in sensor and telecommunication technologies have enabled us to monitor key properties such as soil moisture at depth-discrete locations in real-time, which provides a great potential for estimating actual ET. The vertical profile of soil moisture, when properly calibrated, can capture the water balance within soil, and therefore, provide the estimate of ET. The ET is equivalent to the rate of water loss, particularly when the downward water flux can be neglected or quantified. In this study, we aim to develop a hybrid machine learning-based inverse modeling framework for estimating ET in real-time based on in situ soil moisture sensors and vadose-zone flow modeling. The inverse modeling framework couples an ecohydrological model – the Richards’ equation and plant root uptake model – with the ensemble recursive Bayesian filter. We demonstrate our approach using the soil moisture data at the East River watershed in Colorado. We also evaluate the effects of various model parameters and assumptions, such as root density distribution and soil hydraulic parameters.

ABSTRACT. Porous media in nature exhibit complex and irregular geometry, and understanding of the underlying heterogeneity is key to the accurate description of groundwater flow and transport processes. Specifically, the appropriate representation of hydraulic conductivity at different scales is the first step for constructing a numerical groundwater model for use in field applications. In this presentation, we revisit the hydraulic conductivity upscaling approach of Kitanidis [1990], which is valid under gradually varying flow assumption, and develop an efficient open-source computational tool to provide field practitioners upscaled hydraulic conductivity fields, in a tensor form that accounts for anisotropy, of any arbitrary size from fine resolution ones. We test our tool with high dimensional 3D fine scale model upscaling examples and compare fine-scale head fluctuations with coarse scale counterparts. Lastly, the upscaling equation used in this presentation is directly included in a neural network to test data-driven upscaling tool development.

ABSTRACT. We simulate the propagation of reducing conditions from fine-grained, organic carbon enriched lenses into a more pervious surrounding aquifer, based on a series of column experiments, using the reactive transport code CrunchFlow. We show that such reduced lenses exert an outsized influence on groundwater chemistry by exporting large amounts or organic carbon that stimulate microbial activity. The precipitation of reduced minerals inside the aquifer (iron sulfurs) both buffers the aqueous concentrations and impedes the reversibility of hydrologic transitions.

ABSTRACT. The soil thickness plays an important role in the hydrology of watersheds such as surface and base flow runoff and water storage. It is also the zone where roots uptake water and nitrogen, organic matter decomposes, and microbes active. However, the mechanisms that control the soil thickness are still not clear. Here we combine modeling and statistic approaches to reveal the high heterogeneity of soil thickness with a fine resolution (i.e. lidar dem) at a hillslope and watershed scale in a mountainous area.

ABSTRACT. We present a framework for hydrogeological simulations in fault zones, based on discontinuity network analysis - to assist in setting up numerical simulations - as well as a numerical simulator treating fault zones as lower dimensional features in a conforming mesh and accounting for their longitudinal and transversal permeabilities. We investigate the impact of conceptualizations on the regional flow field and show where the simplification of fault zones as either barriers or conduits are appropriate.

ABSTRACT. We study the upscaling of bacteria dispersion in porous media due to hydrodynamic transport and their own motion. Based on the analysis of experimental particle tracking data of motile and non-motile bacteria, we derive a stochastic transport model that accounts for both bacteria motility and hydrodynamic flow fluctuations. The model predicts accurately the propagators of motile and non-motile bacteria, their center of mass velocity and dispersion.

ABSTRACT. Electrokinetic techniques are used in different fields of science and engineering to enhance the transport of charged species and non-charged compounds in low-permeability porous media. The modeling of electrokinetic transport processes is a challenging task as it requires the description of the interplay between physical, electrostatic and biogeochemical processes occurring upon application of an electric field. In this work we developed a modeling tool, NP-Phreeqc-EK [1], consisting of a coupling between COMSOL Multiphysics and PhreeqcRM [2]. The former is used to solve fluid flow and Nernst-Planck-Poisson-based solute transport in saturated porous media, the latter simulates biogeochemical reactions. The multidimensional multiphysics model uses an operator splitting approach and is implemented in a MATLAB LiveLink interface. The code integrates fluid flow, solute transport including electromigration and electroosmosis, Coulombic interactions between transported species, and a wide range of kinetic and equilibrium reactions. NP-Phreeqc-EK has been benchmarked in different dimensions (1D, 2D and 3D) with analytical solutions, numerical simulations with other software, and data from previously published electrokinetic experiments. This study highlights the flexibility of the proposed approach in simulating electrokinetic reactive transport processes in saturated porous media and emphasizes the importance of charge interactions during electrokinetic transport. The developed modeling tool is suitable to simulate electrokinetic processes at laboratory and field scales and can be applied in a number of studies focusing both on fundamental aspects of electrokinetic transport and on practical electrokinetic remediation applications in homogeneous and heterogeneous media.

References [1] R. Sprocati, M. Masi, M. Muniruzzaman and M. Rolle. Modeling electrokinetic transport and biogeochemical reactions in porous media: a multidimensional Nernst-Planck-Poisson approach with PHREEQC coupling. Advances in Water Resources. 127, 134–147, (2019). [2] D.L Parkhurst, L. Wissmeier. PhreeqcRM: A reaction module for transport simulators based on the geochemical model PHREEQC. Advances in Water Resources. 83, 176–189, (2015).

ABSTRACT. The spatial distribution of physical and chemical heterogeneities greatly impacts solute transport in groundwater and is critical in many subsurface applications. Previous studies have focused on imaging physical heterogeneity and the spatial distribution of hydraulic conductivity. However, the applications of such approaches to image geochemical heterogeneity based on water quality data and reactive transport modeling are still rare [1]. In this work we perform flow-through experiments in multidimensional setups and we combine them with forward and inverse reactive transport modeling to explore the capability of imaging pyrite inclusions in the subsurface. We studied the oxidative dissolution of pyrite in multidimensional setups, including 1-D columns and a 2-D flow-through chamber. Pyrite dissolution was recently studied in batch setups where the consumption of oxygen, the change in pH and the release of dissolved iron and sulfur were measured to constrain the dissolution kinetics of the reactive minerals [2]. In this work, we embedded pyrite inclusions in a sandy porous matrix. The reactive mineral and the inert sandy matrix had the same grain size in order to obtain physically homogeneous but chemically heterogeneous media. Spatially distributed optode measurements of dissolved oxygen allowed tracking the propagation of an oxic groundwater front in the initially anoxic 1-D columns and 2-D flow-through chamber. The non-invasive optode sensor measurements of O2 were performed at high-resolution (i.e., spacing of 2.5 mm) along the axis of the 1-D column and at different cross sections in the 2-D flow-through setup. These data, as well as the analysis of dissolved species originating from pyrite oxidation and measured at the outlet of the flow-through systems, were compared with the outcomes of a reactive transport model describing the main physical and geochemical processes in the setups. Finally, inverse modeling based on the Principal Component Geostatistical Approach (PCGA) [3] was applied to image the spatial distribution of the pyrite inclusions. The results show the capability of the proposed methodology to accurately identify both the spatial locations and the concentration of the reactive mineral zones embedded in the 1-D and 2-D systems and open interesting perspectives on the potential of common groundwater quality parameters to enable the estimation of reactive processes and geochemical heterogeneities in groundwater flow systems.

References [1] S. Fakhreddine, J. Lee, P.K. Kitanidis, S. Fendorf and M. Rolle. Imaging geochemical heterogeneities using inverse reactive transport modeling: An example relevant for characterizing arsenic mobilization and distribution. Advances in Water Resources, 88, 186-197, (2016). [2] M. Battistel, M. Muniruzzaman, F. Onses, J. Lee and M. Rolle. Reactive fronts in chemically heterogeneous porous media: Experimental and modeling investigation of pyrite oxidation. Applied Geochemistry, 100, 77-89 (2019). [3] J. Lee and P.K. Kitanidis. Large-scale hydraulic tomography and joint inversion of head and tracer data using the Principal Component Geostatistical Approach(PCGA). Water Resources Research, 50, 5410-5427, (2014).

ABSTRACT. We illustrate the capability of a mechanistic discrete-continuum multi-scale model for transport, biofilm development and solid restructuring in soil in three examples. First, we study structure formation of microaggregates which are the fundamental building blocks of soils and are important for soil structure, properties, and functions. Second, we quantify the effect of biomass development and root exudate on the macroscopic soil hydraulic properties. Third, we examine the way spatial distribution of organic matter and microaggregate structure can influence microbial diversity.

ABSTRACT. Fractures are of great importance for most geoengineering applications as they greatly influence both the hydrodynamic properties of natural formations and their geomechanical response. Thus, understanding their impact on both the flow and the mechanics is crucial to operate efficiently and safely any engineering system involving the subsurface, such as geothermal fields, hydrocarbon reservoirs and CO2 storage sites.

In this work, we employ a finite-volume discretization for the multiphase flow equations coupled with a finite-element scheme for the mechanical problem. Fractures are considered as lower dimensional entities, embedded in the rock matrix. As such, it is not required to generate a conforming mesh that honors the complex fracture geometry. The effect of fractures on the flow problem is accounted for using the embedded discrete fracture model (EDFM), which enriches the finite-volume formulation to include mass transfer between the fractures and the rock matrix. The contribution of the fractures to the mechanical behavior is, instead, captured by employing embedded finite-element method (EFEM), which relies on a local enrichment of the finite-element space to represent the displacement jump at each fracture surface. Both the use of piece-wise constant and piece-wise linear enrichments are investigated. One key advantage of is the element-based nature of the enrichment, which reduces the geometric complexity of the implementation and leads to linear systems with advantageous properties that can be exploited when devising an effective preconditioner.

The convergence and accuracy of the proposed method is studied through synthetic numerical examples. The method is also applied to more realistic scenarios, involving heterogeneous reservoirs with complex fracture distributions, to demonstrate that it is a viable option for field-scale applications.

ABSTRACT. Hydrogeophysical methods, such as electrical resistivity tomography (ERT), are increasingly used to study subsurface soil moisture dynamics, yet its applications beyond the soil compartment are limited. To examine how hydrogeophysical methods can be jointly utilized with other measurements to improve model-based ET estimation under soil water stress conditions, we conducted a pilot-scale field study at an experimental maize plot. The goal is to develop streamlined procedures from (1) acquiring and inverting field ERT monitoring data, (2) translating ERT into hydrogeological parameters, (3) spatially computing soil water stress coefficients that can be directly used in ET model, to (4) evaluating the performance of ERT-based ET computation. Our results show that soil moisture was better characterized by ERT than point sensors concerning the irrigation schedule. Moreover, the spatial extent and resolution of ERT information proved it of great use to compute soil constraints at various vertical depths of interests. The importance of ET simulation with soil constraint was also shown by comparing the transpiration under two water irrigation regimes. The integration of the hydrogeophysical datasets can be jointly used with other types of data to further understand ET and SPAC hydraulic dynamics.

ABSTRACT. The purpose of this research is to compare results for a levee at Port Arthur, Texas from two transient finite element seepage programs where the first program considers only hydraulic loading, and the second program does coupled seepage/soil deformation computations. Each program is parallelized for high performance computing to consider 6000 realizations of the material properties. This Monte Carlo type analysis is used to compute the probability of unsatisfactory performance for under seepage, through seepage, uplift, and slope stability.

ABSTRACT. Estimation of the permeability tensor is crucial to many natural and industrial porous media applications. Homogenization provides a rigorous approach to calculate effective parameters from pore-scale images of porous media by solving a boundary value problem on a representative unit cell. However, it requires the enforcement of global constraints that are hard to satisfy for arbitrarily complex geometries. We have developed a novel computational algorithm to rigorously calculate the permeability tensor. We show that the proposed algorithm has high accuracy.

ABSTRACT. We developed a multiscale algorithm, in which traditional upscaling is substituted by deep learning. The algorithm is constructed to model reactive transport in a multiscale fracture network, composed by main fractures and micro fractures. The new approach is first validated against analytical solutions and fully resolved simulations. Using the algorithm, we can model micro fracture clogging during mineral reactions, which cannot be accurately captured by classical upscaling methods.

ABSTRACT. We present simulations from the novel coupled model CrunchREWT, which merges the principal reactive features of the reactive transport model CrunchTope with the spatially heterogeneous transport framework of the hydrobiogeochemical model REWT (Root Exudation in Watershed-scale Transport). CrunchREWT's unique framework facilitates high-resolution, process-based simulations of complex interactions according to stoichiometrically-balanced reaction networks, including the influence of roots on biogeochemical cycling via root exudation and the role of deep roots on weathering and mineral or nutrient availability.

ABSTRACT. Evaluation of the dissolved oxygen ingress into sedimentary basins during glaciation events has been studied in the past. Due to computational limitations, these simulations were simplified to 2D. In this study, the advanced MIN3P-HPC code is used to perform large-scale and long-term 3D simulations of dissolved oxygen ingress. The meltwater recharge over the major glaciation periods and the hydro-mechanical coupling effects are considered. The results demonstrate the versatility and enhanced performance for large-scale and long-term reactive transport simulations.

ABSTRACT. In this work, an open-source multi GPU hydrodynamics code called TRITON is presented. It includes a novel bed slope upper and lower limits that are derived mathematically and allow to constrain the source strength instead of reducing the time step size. This fact ensures the positivity and accuracy of the solution at the same time as the efficiency is not compromised. It is proved to scale up to 512 GPUs using Summit supercomputer using a 68M grid cells test case.

ABSTRACT. Field observational data at an intensive study site (East River, Colorado) indicate that some aqueous geochemical components, such as sulfate and calcium, exhibit distinct characteristics under snowmelt and baseflow conditions. Subsurface geologic structure and mineral composition may have a strong influence on concentration-discharge (C-Q) and geochemical responses under different water infiltration and groundwater level scenarios. Numerical modeling is highly desirable to understand the spatial and temporal variability of hydrological and geochemical processes. The Advanced Terrestrial Simulator (ATS) is an integrated hydrology model based on solving various forms of Richards equation coupled to a surface flow equation, developed by multiple US DOE national laboratories under an open-source license. Besides hydrology, ATS is also coupled with reactive transport models, for example, PFLOTRAN and CrunchFlow, via the Alquimia interface. Inspired by the C-Q relationship conceptual model, the ATS code is applied to simulate the integrated hydrology, biogeochemistry and reactive transport processes between surface and subsurface at representative sites. One is the East River PLM hillslope intensive study site, and the other one is Copper Creek, one of the largest catchments in the watershed of East River. The meteorological forcing combines the high-resolution PRISM precipitation reanalysis data, and the simulated evapotranspiration and snowpack from another model (ParFlow-CLM). The hydrology and geochemistry simulation results are compared with the observations at monitoring wells and rivers at the outlet of the catchment. Two key minerals and geochemical processes, including calcium dissolution and pyrite oxidation, are examined with river discharge under the scenarios of precipitation, snowmelt and groundwater level fluctuations. The simulations are performed on NERSC supercomputers owing to the large computational requirements in these fine-resolution models. This study aims to understand the interactions of hydrology and major subsurface geochemistry processes, provide insights into the variability of water quality in a watershed scale under a changing climate environment.