ABSTRACT. TBA

ABSTRACT. This work aims at developing a fully implicit scheme for numerical simulation of two-phase reactive flows in porous media. Such processes are governed by a highly nonlinear coupled system of PDEs modeling a compositional flow to differential algebraic equations or ODEs modeling chemistry. We have integrated specific developments in the DuMuX framework. In a HPC context, three-dimensional numerical results related to geological storage of CO2 will be presented with a comparison between the fully implicit scheme and a sequential one.

ABSTRACT. Injection-induced seismicity has emerged as a central scientific and societal issue in the development of subsurface energy technologies such as enhanced geothermal energy extraction, unconventional hydrocarbon production, wastewater injection, geologic carbon sequestration, or underground gas storage. Although our knowledge of induced seismicity has increased thanks to field observations and computational models, some key aspects remain poorly understood. One of the outstanding questions is how the hydraulic properties of faults impact the nucleation and rupture of injection-induced earthquakes.

The hydraulic properties of faults can range from sealing to complete conductive, both in terms of flow along and across the fault. However, the effects of longitudinal and transverse fault permeability on the development of injection-induced earthquakes remain unexplored. Here, we explore these effects by means of sophisticated computational modeling. We develop a numerical framework for fully-coupled poromechanics in the presence of faults. Faults are modeled as contact frictional surfaces, whose frictional rheology is governed by the Dieterich-Ruina rate-and-state law. This framework allows us to reproduce and explore the interplay between two basic mechanisms responsible for injection-induced seismicity: (1) the increase of pore pressure that reduces the effective normal compressive stress on the fault, and (2) poroelastic effects, by which the dilation of the rock rises the total compressive stress on the fault, and develop additional shear stresses along the fault whose effect may be stabilizing or destabilizing depending on the well location and fault orientation.

In our computational model, we simulate earthquake triggering by fluid injection in a 2-dimensional computational model. We elucidate that the nucleation and rupture phases are driven by the pressure and shear stress profiles along the fault, which in turn are controlled by the fault hydraulic properties. Our model allows us to rationalize how longitudinal and transverse fault permeability impact the mechanisms that control the evolution of fault strength and shear stress, the stress path to instability. We find that fault permeability exerts a heretofore unrecognized fundamental control on the scaling of the nucleation length, the mode of rupture and, as a result, the magnitude of the subsequent seismic event.

ABSTRACT. CO2 sequestration in the subsurface is poised to grow, but uncertainties still exist regarding the effects of coupled flows and geomechanics on the integrity of CO2 storage. One important unknown is thermal effects. When injected CO2 is cooler than the target reservoir, thermal stresses can develop in the reservoir and in the caprock, threatening storage integrity. This work extends a multiphase simulator with thermal effects to better understand thermal stresses in the caprock by enabling high-resolution, parallelized thermoporomechanical simulation.

ABSTRACT. Coupling of geomechanics and flow in a poroelastic porous media has several subsurface applications. The geomechanical effects account for the influence of deformations in the porous media caused due to the pore pressure whereas the changes in the pore structure due to mechanical stresses affect the flow field. The fact that the fractures in the porous media have strong influence on the flow profiles and the deformations are particularly significant in such reservoirs motivates studying the coupled geomechanics and flow problems in fractured reservoirs. This work reports computational, algorithmic, and theoretical developments in the models describing the coupling of flow and mechanics in fractured media in subsurface.

ABSTRACT. Simulation of multiphase poromechanics of subsurface processes requires handling domains with geometrical complexities. Flexibility of polygonal cells helps the generation of meshes conforming to changes in physical properties.

We introduce a coupled formulation for multiphase flow and geomechanics based on a mixed hybrid (mimetic) Finite Volume and Virtual Elements approach, capable of handling general polygonal meshes. Fixed-stress based solution strategies using either sequential-implicit or fully-implicit approaches are discussed. The method robustness and accuracy are demonstrated on challenging numerical experiments.

ABSTRACT. Poroviscoelasticity involves primary and secondary consolidation of fluid-filled porous media, i.e., the superposition of instantaneous and delayed responses to loadings. In this talk, we cast the corresponding model as a gradient flow depicting the dissipative character of the system. By exploiting the inherent minimization structure at semidiscrete level, we develop and analyze novel robust block-partitioned solvers extending the popular fixed-stress and undrained splits (for poroelasticity). The systematic approach demonstrated here is applicable to a large class of coupled non-linear models.

ABSTRACT. To assess the reactive surface is a major problem in macroscale complex reactive transport e.g. with evolving microstructure and/or in multiphase flow. We will show the benefits of micro-macro models combining macroscale and microscale ‘on demand’ for this and other questions and indicate numerical approaches which make micro-macro models feasible despite their apparent numerical complexity, ‘doubling’ the spatial dimension.

ABSTRACT. We present a mechanistic modelling approach that couples complex biological, chemical, and physical processes on the pore scale. It combines a PDE model for reactive transport of nutrients and bacteria in liquid and gas phase with a discrete cellular automaton for the development of a biofilm. Simulations on CT scans of microaggregates show that heterogeneity in the spatial distribution of organic matter and heterogeneity in the microaggregate structure influences the bacterial biodiversity.

ABSTRACT. Concentration hotspots are associated with highly localized reactivity. Hotspots are small areas of the domain characterized by high concentration values, and whose dynamics controls the system behavior at the macroscale. Here we propose a rigorous definition of “hotspots”: we investigate their spatial heterogeneity as well as their temporal ephemerality in terms of relevant dimensionless numbers in a planar fracture. Finally, we construct a phase diagram in the Peclet-Strouhal space to predict hotspot dynamics and we validate it through numerical simulations.

ABSTRACT. We use a reactive transport modeling approach to evaluate the interactions between microorganisms mediated by signaling molecules (quorum sensing, QS). Using a nondimensionalized advection-diffusion-reaction equation, QS behavior of single and multiple microbial aggregates was investigated across a range of Péclet and Damköhler numbers. The results highlight the importance of the distance between microbial aggregates and their relative location to the flow direction. Also, QS activation triggering the production of more signaling molecules facilitates the cell-to-cell communication over long distances.

ABSTRACT. Mineral reaction rates exhibit spatial scaling effects in porous media. Darcy-scale models often cannot adequately capture the effects of strong concentration gradients, heterogeneous distributions of reactive surface area and flow paths on mineral reaction rates. However, pore-scale models are computationally expensive, and the geometric information in the subsurface is hard to obtain. Therefore, we perform pore-network modeling which balances computational effort and modeling complexity.

We implement a pore-scale reactive transport solver in FEniCS[1] to study how the concentration gradients of dissolved minerals affect overall reaction rates. In Figure 1, we observe the variation of the concentration gradient of the dissolving ion along the flow path, leading to a process dependence of the average dissolution rate, which is difficult to quantify at the Darcy scale. In contrast, pore-network models are capable of capturing the pore-scale heterogeneities in the concentration field. This work investigates how spatial heterogeneities of mineral concentration and reactive surface area affect overall reaction rates at the Darcy scale. We aim to provide a generalized upscaled rate law of mineral dissolution for the Darcy-scale model, which can be further implemented to the coupled reactive transport solver using DuMux[2] and Reaktoro[4][3].

ABSTRACT. In order to analyze the dynamics that control bimolecular reactive transport, we study the irreversible chemical reaction A + B → C. We use a reactive random walk particle tracking (RWPT) pore scale model capable of simulating bimolecular reactive transport. We focus on the experimental scenario reported by Jiménez-Martı́nez et al. (2015) to determine the impact of flow heterogeneities on the irreversible reaction. These authors studied the conservative displacement B by A in a 2-dimensional porous medium characterized by a random grain packing. We determine the corresponding experimental reactivity based on the concentration data for the invading fluid. We quantitatively account for the total mass of product C (m_C (t)) and the mixing zone between the reactants from the laboratory experiment. Then, we simulate transport and reaction in the image of the experiment’s geometry to validate the reactive RWPT model through predicting the experimental results. Our model accurately predicts the evolution of m_C (t) in the experiment as it captures the degree of incomplete mixing present at the pore scale. Our results are in contrast to the advection-dispersion-reaction equation model that overpredicts m_C (t) in the medium because assumes complete pore scale mixing.

ABSTRACT. We report on a systematic investigation of the transport properties of laboratory rock samples using pulse-tracer tests in both sandstone and carbonates. One-dimensional formulations of the Advection Dispersion Equation, the Multi-Rate Mass Transfer model, and the Continuous Time Random Walk model were evaluated against the measured effluent profiles. Care was taken to identify those parameters that are solely rock dependent to reduce the complexity of the fitting exercise, while providing for a more robust interpretation and comparison of the model results.

ABSTRACT. We introduce a new, fully dynamic pore network modeling platform that can better account for the complex dynamics of pore-scale displacement and transport mechanisms and predict the corresponding macroscopic flow and transport properties. A highly scalable parallel implementation of this platform is designed to upscale the multiphase flow and transport processes to core-sized porous media. A comprehensive study of solute transport behaviors at different Peclet numbers are conducted and the simulation results are rigorously validated against their experimental counterparts.

ABSTRACT. Central Valley aquifers are recharged by precipitation that falls in the Sierra Nevada. These hydrologic links may evolve as climate change shifts Sierra precipitation from snow to rain, so we need to better understand the current links to predict and manage this evolution. We use three state-of-the-science approaches to better understand the current hydrologic links: integrated hydrologic modeling with ParFlow-CLM, EcoSLIM Lagrangian particle tracking constrained by isotope measurements, and spatial pattern constraints on land surface temperature, soil moisture, and evapotranspiration.

ABSTRACT. In recent years, record-breaking durations of droughts followed by high-precipitation events have been newly termed “water availability whip-lash” in places like the western US, where the potential for a “new-normal” has unknown consequences for water management. With the use of high performance computing, this work explicitly simulates the water-land energy budget with an integrated hydrologic model in high spatio-temporal resolutions. Hydrologic metrics are compared for recently historic dry and wet years, in addition to an end of century simulation.

ABSTRACT. Starting from flow equations in a 3D root architecture, we derived a general root water uptake equation with three characteristics root system properties: root system conductance, uptake distribution for uniform soil water potentials, and a compensatory uptake matrix. This equation can be scaled up straightforwardly and implemented in 1D soil water balance models and is a generalization of other models like the parallel root and big root models. We illustrate how the model can describe uptake by mixed vegetation.

ABSTRACT. Multiscale simulation of advection-diffusion problems has been the subject of many studies in recent years. Although methods like MsFEM and MsFVM have worked well for elliptic and some parabolic problems, there are challenges in extension to hyperbolic problems in general. We developed a theoretical framework, “the global-local analysis of PDE” to rectify this. Based on the framework, we equip the well-established Generalized FEM(GFEM) with transient local problems. Some highlights of our resulting framework include: 1) resolves multiple space and time scales; 2) computationally efficient; 3) recovers fine scale field on the coarse global mesh

ABSTRACT. Gravity fingering is a powerful hydrodynamic instability that sets in during infiltration of water in dry soil, leading to the emergence of vertical channels of preferential flow that conduct virtually all the infiltration water. This phenomenon has been characterized in detail in controlled laboratory experiments, and has been documented in the field. Mathematical and computational modeling of gravity fingering has proved challenging, but recent progress has allowed us to simulate this process quantitatively. Building on earlier work, we model unsaturated flow following a thermodynamic approach, which leads to a partial differential equation with a nonlinear, singular fourth-order term—a formulation that is endowed with an entropy function.

Here, we assess the impact of gravity fingering on deep drainage, that is, the infiltration water that bypasses evaporation and reaches deep groundwater bodies. We do so by coupling our model of unsaturated flow with evapotranspiration, which we model as a spatially- and temporally-variable nonlinear sink term in the equation. We apply the model of coupled infiltration and evapotranspiration to the soil and climatic conditions of the Kalahari transect. We reproduce not only the mean annual precipitation (MAP) and potential evapotranspiration (PET), but also the rainfall statistics that characterize the intensity and frequency of the precipitation events. We conduct unprecedented simulations of infiltration that are decades long (to capture the long-term rainfall statistics and system’s deep drainage behavior), with sub-minute temporal resolution (to capture the infiltration flux during storms) and with sub-centimeter spatial resolution (to resolve the gravity fingers accurately).

We compare the predictions of two models: the classic infiltration model based on Richards equation (which cannot reproduce the gravity-fingering instability) and our proposed model capable of reproducing fingered flow. We find that the wetting front instability has a dramatic impact on deep drainage fluxes. Fingered flow causes water to quickly traverse the shallow root zone, by-passing most of the soil column and effectively reducing losses due to evaporation and transpiration from shallowly-rooted plants. As a result, water may be expected to percolate below the topsoil, and be found at depth in arid and semiarid climates, where PET far exceeds MAP. Water from subsequent infiltration events tends to be diverted towards the fingering channels, which may thus persist over many rainfall cycles. Our results suggest that models of vegetation dynamics based on simplified estimates of infiltration depth may underestimate root biomass and groundwater recharge in dryland environments, and that gravity fingering moderates the response of water-stressed ecosystems to climate variability, increasing their resilience to a future scenario of higher aridity.

ABSTRACT. Inverse modeling is widely used to improve the predictive accuracy of subsurface flow and transport models by calibrating them against dynamic flow response measurements. While conceptually simple, inverse modeling in this domain is fraught by several challenges, including data limitation, geological uncertainty, nonlinearity of parameter-data relations, as well as the complex heterogeneity and high dimensionality of the unknown parameters (e.g., distributed rock flow properties). Recent advances in deep learning have inspired new efforts to tackle some of these challenges. This talk presents deep convolutional neural network (DCNN) as an effective dimensionality reduction and inverse modeling framework for subsurface flow systems. Specifically, DCNN architectures are designed to extract complex patterns from training data and use them to construct low-dimensional data-parameter manifolds in which the inverse problem is solved more effectively. Subsurface flow inverse modeling examples are presented to illustrate the performance and main properties of DCNN models compared to traditional methods.

ABSTRACT. Microtopography can play a very relevant role in rainfall partitioning into runoff and infiltration. Usually, however, do to data and computational constrains, microtopography is not explicitly accounted for. In this study, we perform simulations with a physically-based surface water solver together with time-dependant and constant infiltration rates, over a large number of surfaces with microtopography. Our results show that microtopography can have significant effects of runoff and infiltration, depending on slope and microtopographic features. Time-dependant infiltration may also magnify such effects.

ABSTRACT. The Standardized Precipitation Index (SPI) is a drought index used throughout the world which normalizes accumulated precipitation by fitting univariate probability distributions for each month or day of the year. The current approach suffers several methodological limitations, notably that time steps are fit independently, zero probability mass is estimated empirically, and non-stationarity is not explicitly considered. A novel Bayesian approach is described to address these limitations and is compared with traditional frequentist approaches using synthetic and real-world precipitation.

ABSTRACT. We present a rapid online decision support tool capable of predicting long term, high resolution spatio-temporally distributed well contamination from spatio-temporally distributed nonpoint sources. The online tool efficiently computes several hundred years of concentration breakthrough at receiving wells and streams. The tool has been validated for the Central Valley, California, with over 20,000 production wells and 150,000 domestic wells. The tool is able to efficiently evaluate historic and alternative land management scenarios and their effect on nitrate in well water.