ABSTRACT. Watersheds are recognized as the Earth’s key functional unit for assessing and managing water resources. Despite significant improvements in numerical simulation capabilities, however, predictive understanding of watershed functions is often hindered by the heterogeneous and multiscale fabric of watersheds. Recent advances in remote sensing, geophysics and sensor technologies – as well as machine learning (ML) – have provided an unprecedented opportunity to understand and characterize bedrock-to-canopy heterogeneity of watersheds, and parameterize ecohydrological-biogeochemical models.

Under the Watershed Function SFA (watershed.lbl.gov) project, we are developing novel watershed-characterization methodology to quantify complex watershed systems across scales, by integrateing multi-scale multi-type datasets (point measurements, surface geophysics, airborne data and satellite images). We have developed methods to create high-resolution maps of (1) plant functional types and plant traits based on airborne LiDAR and hyperspectral data fusion critical for quantifying evapotranspiration, (2) soil thicknesses and texture based on a hybrid data/model-driven approach using LiDAR and soil transport models, (3) 3D subsurface structure (up to the depth of 200 meters) and effective bedrock porosity based on airborne geophysics, and (4) ecosystem sensitivity to early snowmelt and droughts based on multi-year satellite images. In addition, we have developed a ML-enabled zonation approach to quantify the co-varied bedrock-to-canopy properties and to identify the zones that have distinct distributions of these properties for tractable parameterization and experimental design. In parallel, we have developed an ML-based sensor network optimization methodology coupled with watershed ecohydrological models to inform the optimal number and placement of snow and soil moisture sensors. Through explicitly bridging information derived from “on the ground" observations, remote sensing data, and numerical models, we aim to quantify fundamental scientific linkages among interacting processes in the watershed and to improve the predictability of watershed functions.

ABSTRACT. In-situ remediation technologies based on electrokinetically enhanced bioremediation (EK-Bio) have the potential to remove organic contaminants from low-permeability porous media. For chlorinated solvents, which are among the most widespread groundwater contaminants, the technique relies on the delivery of electron donors and, when necessary, also specialized microorganisms to induce and/or enhance microbially mediated reductive dehalogenation. One of the major challenges of such in-situ applications consists in the ability to take into account the complex interplay between physical and biogeochemical reactions to predict the evolution of the system over time. The development of comprehensive numerical models is thus essential to interpret, design and optimize experimental applications under well-controlled laboratory conditions, as well as at pilot and full field scales. In this work we propose a process-based modeling framework for EK-Bio based on NP-Phreeqc-EK [1], a simulator coupling COMSOL Multiphysics and PhreeqcRM. The model accounts for: (i) multidimensional electrokinetic transport in saturated porous media including electromigration of charged species and electroosmosis, (ii) Coulombic interactions with the Nernst-Planck-Poisson equations, (iii) kinetics of contaminant degradation, (iv) dynamics of microbial populations including growth and decay of indigenous and delivered degraders, (v) mass transfer limitations, and (vi) geochemical reactions. The modeling study includes scenario simulations based on an EK-Bio pilot study in a contaminated clayey aquitard at the Skuldelev site (Denmark). The results show that electromigration plays a key role for the delivery of lactate, the charged electron donor used to stimulate the biodegradation activity. The transport of dissolved chlorinated compounds and the delivery of specialized microorganisms (i.e., the KB-1 reductive dehalogenating consortium) was controlled by the induced electroosmotic flow. In the considered geometry, including 9 injection and/or electrode wells, the effective delivery of lactate could be achieved upon application of a sufficient electric potential (110 V) at the electrodes for at least 100 days to ensure degradation of the chlorinated ethenes by the indigenous bacterial population. However, complete reductive dechlorination to the non-toxic product ethane could only be achieved in the zones where the delivered specialized microorganisms were distributed. The simulation outcomes show a great increase of the efficiency in substrate delivery and contaminant degradation when applying an electric field and relying on electrokinetic transport mechanisms compared to cases only based on natural transport processes. To evaluate the results of different conservative and reactive transport scenarios, we used quantitative metrics such as the relative area of substrate distribution and the relative mass of contaminant degraded. In one of the modeling scenarios we also considered the presence of the parent compound, tetrachloroethene (PCE), in the form of free product, as non-aqueous phase liquid (NAPL). When the NAPL product was present it significantly affected the degradation efficiency and the remediation time, as a result of the larger mass to degrade and of the kinetic limitations of the interphase mass-transfer. This study illustrates the potential of electrokinetically enhanced bioremediation and the capability of a newly developed modeling approach to describe the different physical, electrostatic, chemical and biological processes controlling the overall remediation efficiency in field scale applications.

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).

ABSTRACT. For complex processes, e.g. in evolving porous media, the upscaled models obtained by periodic homogenization may be micro-macro in the sense, that the coupling of the macroscopic equations and the equations at the reference cell is both ways. Micro-macro models allow more general interaction of processes, where the microscopic processes “compute” the constitutive laws, otherwise assumed. The pure local incorporation of microscale information allows for a new way of adaptive models deviating from macroscale models and codes only there where necessary.

ABSTRACT. Micro-macro models the sense, that the coupling of the macroscopic equations and the equations at the reference cell is both ways allow more general interaction of processes, where the microscopic processes “compute” the constitutive laws, which need no longer be assumed. We will discuss various examples and in particular numerical approaches to keep the numerical complexity in the range of pure macroscopic models.

ABSTRACT. Microbially induced calcite precipitation (MICP) is a new and sustainable leakage prevention technology which utilizes biochemical processes to create barriers by calcite cementation. We describe a field-scale model of MICP technology including transport of limiting components in the injected solutions (bacteria, substrate, and urea), bacterial attachment, biomass detachment, death of bacteria, formation of biofilm, usage of urea, and production of calcite, modifying the rock porosity and hence the effective permeability. We study diverse injection strategies to prevent leakage of CO₂ in two set of reservoir properties.

ABSTRACT. Reactive transport simulations are in general time-consuming due to costly geochemical equilibrium and/or kinetics calculations. These may account for over 99% of all computing costs when complex chemical systems are considered, because those computations are needed one or more times in each cell of high-resolution meshes, at every time step of the simulation. To reduce their computing cost by orders of magnitude, we present an on-demand learning strategy that enables geochemical calculations to be rapidly and accurately predicted using previously learned geochemical states. We use sensitivity derivatives combined with first-order Taylor estimations to achieve these fast computations. These derivatives enable a complete bypass of expensive operations such as evaluation of thermodynamic properties (e.g., species activities, fugacities, equations of state), solution of matrix equations in each Newton iteration, time integration of ordinary differential equations, and more. We present reactive transport simulations, considering realistic chemical systems and strong non-ideal thermodynamic behavior, in which geochemical calculations were speed up by a factor of 100 to 200 using this on-demand learning algorithm.

ABSTRACT. The filamentous algae Cladophora is one of the most prevalent algae worldwide. It interferes with water supplies, and negatively affects dissolved oxygen and pH in lakes and rivers. Yet, the complexity of its response to temperature, nutrients availability and hydrodynamics challenges our modeling capabilities. We developed a fully coupled computational model able to capture the temporally varying spatial distribution of Cladophora in the Clark Fork river. The model predictions match with remote sensing measurements of the spatial coverage of Cladophora.

ABSTRACT. The focus of this study is to refine the process-based understanding of riparian hollows as small-scale nitrogen-retention mechanisms. We used a multicomponent flow and reactive transport model to simulate the effects of hydrobiogeochemical perturbations on riparian hollow nitrogen cycling dynamics, using data from the East River watershed to constrain the model. Model results suggest groundwater upwelling and rainfall events act as significant hydrologic controls on the nitrogen retention capacity of riparian hollows. Future climatic perturbations related to these controls will likely increase the sink capacity of riparian hollows, enhancing their function as inhibitors to upland nitrogen reaching the stream.

ABSTRACT. Biogeochemical cycles impact most ecosystem functions by controlling the distribution of nutrients in an environmental compartment. The Critical Zone accounts for a large proportion of the global carbon and nitrogen budget. The involved subsurface compartments are known to exhibit spatial heterogeneity and temporal dynamics. In this study a numerical modeling approach is used to study the impact of spatio-temporal heterogeneities on nutrient cycles in the subsurface. Results indicate that spatial heterogeneity has an impact on nutrient removal from the domain.

ABSTRACT. A macroscopic root water uptake model is considered, described by Richards' equation with a state-dependent sink term. After semi-discretizing the time derivative, we integrate in the space the second-order ODEs system. Endowed with Dirichlet boundary conditions, this problem takes the form of a boundary value problem, that can be classically solved by a shooting method. The problem is faced with Gardner's hydraulic functions and Kirchhoff transformation of water content.

ABSTRACT. We study the impact of sub-core scale heterogeneity on meter-scale flow by conducting high resolution numerical simulations using data obtained in coreflooding experiments. Saturation distributions at steady state show that the experimental data is mostly governed by capillary end effects. Furthermore, gravity effects are observed only on larger scale flows. Two types of capillary entry pressure trapping mechanisms are identified and different statistical parameters of permeability (variance and correlation) are shown to impact these mechanisms.

ABSTRACT. We will introduce our in-house developed “MF-LBM-v2” code, which employs the continuum-surface-force based color-gradient lattice Boltzmann multiphase model and geometrical wetting model and is highly optimized for modern manycore processors/coprocessors. Thus, the code enables us to perform practical pore-scale simulations in real rock geometries at realistic capillary number with consideration of inertial effects as our previous work demonstrates the importance of inertial effects in scCO2-brine displacing process. Validation and demonstration of the simulation capability of the code will be given.

ABSTRACT. The present work investigates mechanisms of permeability impairment as a result of low-salinity fluid injection into brine-saturated porous media containing dispersible clays. We present a new computational fluid dynamics model at pore-scale to simulate detachment, migration and straining of fine particles in porous media. The model uses an Eulerian-Lagrangian approach to simulate fine particles under the effects of attractive (e.g., van der Waals forces), repulsive (e.g., electric double layer forces) and hydrodynamic forces. The pore-scale simulations will be presented to reveal the mechanisms of pore clogging and permeability decline in porous media.

ABSTRACT. This study presents a theoretical investigation of the processes controlling adsorption, capillary condensation and imbibition in nanoporous media. The theoretical model, which is based on the square-gradient classical density functional approach, explicitly includes the relevant interaction forces among fluids and solids in macroscopic porous media. Application of the model to a relative humidity-controlled water uptake experiment in a nanoporous medium is presented to demonstrate the effects of fluid-pore wall interactions on adsorption and condensation-induced imbibition.

ABSTRACT. The purpose of this work is to develop new Neural Network architectures for flux regression problems in Discrete Fracture Networks (DFNs) taking advantage of the graphs representing the DFNs. We build a “Graph Informed NN” (GI-NN), with layers characterized by the adjacency matrix, introducing a novel typology of Graph Convolutional Neural Network. After an introduction to NN applications for flow simulation problems in DFNs, the creation of GI-NNs is described and the flux regression performances are analyzed.

ABSTRACT. Enhanced Geothermal Systems (EGS) present a significant long-term opportunity for widespread renewable power production. The EGS approach may make it possible to utilize otherwise inaccessible geothermal resources. It is estimated that within the U.S. alone the electricity production potential of EGS is in excess of 100 GWe. Hence, efforts to accurately model and predict the performance of EGS reservoirs under various reservoir conditions (e.g., formation permeability, reservoir temperature, existing fracture/fault connectivity, and the in-situ stress distribution) are vital. We present a site- and regional-scale modeling study based on field data (e.g., geochemical tracers, groundwater velocity, initial temperatures from the wells) from New Mexico. The modeling includes coupled fluid flow, energy transport, and tracer modeling by using PFLOTRAN and E4D, a massively parallel multi-physics subsurface and geoelectrical simulator. The subsurface material properties are varied to generate multiple realizations of simulation datasets. Based on the generated simulation data, we investigate the geothermal potential and the impact of associated subsurface uncertainties (e.g, in permeability, water table depth, heat flow, discharge temperature, water chemistry, flow velocity) that can influence the discovery and development of hidden geothermal resources. The datasets and simulation results generated in this study provide the foundation for ongoing supervised and unsupervised machine learning analyses. To this end, we also present a new way to analyze available geologic, geochemical, and geophysical data to reduce the risk and increase success rates associated with EGS exploration and development. We present an unsupervised machine learning method based on non-negative matrix/tensor factorization to perform exploratory data analysis on site-scale and regional-scale data from New Mexico. The goal of our work is to discover the signatures (features) characterizing hidden geothermal resources and favorable EGS sites from field and simulation datasets. Machine learning is used to clean, preprocess, and combine independent data streams (both field and simulation datasets) to analyze geothermal resources in New Mexico. Our work is a part of the ongoing initiative supported by the U.S. Department of Energy – Geothermal Technologies Office to apply machine learning for geothermal energy exploration.

ABSTRACT. This study explores the capabilities of physics-informed neural networks for shallow water problems. Obtaining reliable hydrodynamic measurements can be costly and time consuming, therefore, measurements are rarely available and numerical models are used to simulate hydrodynamic conditions in the spatial and temporal domain. Advances in computational modeling have significantly improved the accuracy of numerical methods, however at the expense of considerable computational costs. On the other hand, these methods have allowed engineers to generate vast amounts of accurate physics-based data that can be leveraged using machine learning methods like physics-informed neural networks. By adding physical constraints to the solution space, we can train neural networks on a small amount of physically accurate data and extrapolate to other domains. Once these neural networks are trained, they can be orders of magnitude faster than a numerical model.

ABSTRACT. We present two deep learning based techniques to estimate nearshore bathymetry. In the first approach, we use a conditional Generative Adversarial Network (cGAN) to generate bathymetry samples consistent with our multi-scale measurements. In the second approach, neural network residual kriging (NNRK), a combination of deep neural network and Kriging, is used to directly learn the inverse relationship between our multi-scale measurements and nearshore bathymetry. Results show that both cGAN and NNRK provide more accurate estimates than Kriging.

ABSTRACT. For geological and synthetic materials, reactive transport modeling is essential to characterize weathering capacity, toxicity, permeability, wettability, strength, and texture. Needed spatial resolution of minerals is currently limited by practical analysis. We developed a machine learning approach that uses synchrotron-based X-ray fluorescence and X-ray diffraction for 2D mineral mapping at mm scales with micron raster resolution. We demonstrate how the maps inform reactive transport modeling including mineral abundance, surface area, accessibility and grain size distribution.

ABSTRACT. Traditional communication based on electromagnetic waves have poor accuracy in complex subsurface environments and it, therefore, requires significant energy to penetrate into complex and absorbing media. Molecular Communication (MC) is a nature-inspired novel communication paradigm whereby digital information is embedded in the chemical structure of molecules. More recently, MC has been extended to consider also the role of diffusion and advection in micro- and nano-channels. These extensions allow to consider information carriers (particles, solutes, tracers) propagating through complex media. However, they do not include proper transport models suitable for realistic subsurface flows at intermediate and large scales, and they still rely on coding and modulation schemes developed traditionally for wave-based communication.

In this work, we present, for the first time, a general modulation and demodulation scheme to encode information into particles and solutes travelling through uncertain and heterogeneous porous media. We make use of a generalised multi-continuum formulation (Municchi, Icardi, 2020) to account for arbitrary heterogeneities and quantify their effects on the efficiency and bitrate of communication. The inverse problem of parametrising the porous media model ( channel estimation) will be also considered. This work throws the basis for possible applications in low-energy groundwater sensing as well as efficient experimental design for estimating pathways and connectivities in the subsurface.

ABSTRACT. We describe the basic conceptual and computational issues of mixing and its implications for reactive transport. We start by the basic mixing calculations and their generalization to dynamic problems. We revise the Water Mixing Approach (WMA) to represent transport as mixing water instead of individual solute concentrations. This representation simplifies calculations as it decouples transport from chemical calculations. We illustrate its application to reactive transport problems. We then analyze representations of solute transport for properly reproducing mixing.

ABSTRACT. We present a family of multi-stage preconditioners for coupled thermal-compositional problems. For thermal simulations, inadequate treatment of the energy equation can cause severe convergence degradation in the iterative solvers. We use Schur-complement decompositions to extract a temperature subsystem and improve the treatment of the energy equation. We show improved performance and robustness across different thermal regimes. The number of iterations is reduced by 40-85% compared to standard CPR, and the new methods exhibit almost no sensitivity to the thermal regime.

ABSTRACT. The feedback between varying ecohydrologic conditions and biogeochemistry in wetlands remains a significant challenge in environmental modeling. We explored the effect of climate and nutrient load on wetlands using a mechanistic biogeochemical model that integrates carbon, nitrogen, and sulphur cycles (BAMS4). The model was initially validated at site specific conditions and it was then extended to a spatial scale. Our results show a relative high resilience to climate change and variability with nutrient load.

ABSTRACT. In carbon storage, in deep geologic zones the potential exists for the formation brine to leak through natural faults, pressure-induced fractures, or failed well casings affecting water quality in shallow aquifers. An eight-meter long intermediate scale test aquifer is used to generate data to validate numerical models used in developing optimal monitoring systems to detect such leakage events. This paper shows the importance in representing the leakage source configuration in matching the model simulations to the observations.

ABSTRACT. Exchange processes across a porous-medium free-flow interface occur in a wide range of environmental systems. In the course of these processes, flow dynamics in the porous domain and in the free-flow domain exhibit strong coupling, often controlled by mechanisms at the common interfaces. Therefore, understanding the underlying processes is decisive. An example of such an environmental problem is soil-water evaporation. We will presented a novel coupled model concept that includes a (Navier-) Stokes model for the free-flow region and a pore-network model for the porous domain. Both sub-models are coupled in a fully monolithic way, i.e., all balance equations are solved in a single matrix and no coupling iterations are required. The coupling conditions implicitly account for the conservation of mass and momentum fluxes at the interface between the two domains.

ABSTRACT. Dimensional reduction strategy is a key ingredient to derive reliable conceptual models to describe flow in fractured porous media. This avoids constructing a computational grid, which resolves the fracture aperture. The typical procedure identifies the aperture to length ratio as the small parameter ε with the fracture permeability scaled as an exponent of ε. We derive the reduced model as the vanishing limit of ε. We have a two-phase flow where we mix polymer into the water phase.

ABSTRACT. In this presentation a new Semi Direct Pore-scale Model (SD-PSM) is developed. In the proposed method, simulations are performed on pore networks where the pore geometry has not been simplified into idealized shapes. Instead, each element of the network represents a component of the micro-CT image of the pore space. Thermodynamically optimal fluid configurations within each voxel image are determined by minimize the Helmholtz free energy. Results of a variety of simulations are analyzed to validate the SD-PSM.

ABSTRACT. We present a novel and highly scalable Dynamic Pore Network Modeling (DPNM) platform developed for simulating two-phase flow through large-scale porous media under various wettability and flow conditions. We validate the platform by simulating flow of oil and water through different miniature core samples and comparing the simulation results against their experimental counterparts. To the best of our knowledge, this is the first time DPNM has been successfully applied to study waterflooding in core-scale porous media under oil-wet conditions.

ABSTRACT. The Anacostia River Watershed is a catchment that covers 460 square km, extends 14 km long, and hosts over 1 million people. The head waters originate in Maryland (Montgomery and Prince George County) and flow south through the nation's capital, eventually discharging into the Potomac River. This study simulates the hydrological flow conditions on the Anacostia Watershed to help improve water quality.

ABSTRACT. Understanding the process of moisture sorption and desorption in multi-material system is important for a variety of applications, such as electronic packaging, food packaging, hygrothermal building performance, and material aging and compatibility. In this work we model the sorption-desorption process in multi-material system by coupling diffusion with equilibrium Henry's absorption and kinetic Langmuir adsorption. A finite volume method with implicit Euler time-stepping scheme was applied to solve the coupled unsteady diffusion-sorption equations. A level set framework with smoothed Heaviside function was used to implicitly treat the moisture concentration and flux at the material interface. The code was validated against existing experimental data for moisture uptake and outgassing in polymeric materials. The proposed numerical method is shown to be well adapted for predicting interfacial mass transfer during sorption-desorption in multi-material system.

ABSTRACT. River corridor systems in snow-dominated, mountainous regions often express complex biogeochemistry and river water nutrient indicators as a function of hydrologic exchange variability and snowmelt conditions. Watershed ecological control points (ECPs) (e.g. hyporheic zones, riparian hollows, stream bed bedrock fractures) are important for solute and nutrient processing at small scales, yet can have major impacts on large scale watershed exports. A major motivating factor for our work is a five-year concentration-discharge (C-Q) time series which display declines in inorganic nitrogen over time as well as down the watershed network, indicating the importance of the passive versus active transient nature of these ECPs[1]. In our research, we develop a predictive understanding of the subsurface and surface controls on hyporheic biogeochemical behavior through data-model integration. We investigate a loose-coupling strategy for hyporheic systems that allows river gross primary productivity (GPP), hillslope runoff, and bedrock contributions to augment hyporheic zone function. We apply this model to the hyporheic zone along the East River, Colorado [2]. Additionally, we showcase a novel microbial capability in reactive transport codes [3],[4] whereby Monod kinetics and biological permeability functions [5] are used to allow feedbacks between flow and microbial growth—as microbes grow, they fill the pore-spaces and limit the flow that initially supported their growth. Across the hyporheic zone and floodplain, we measured surface and subsurface gases, geochemistry, isotopes, and used this data to constrain our model in the presence of transient hydrological flow conditions. A Bayesian approach was used for the river model that allows GPP, respiration, and diffusion parameters to vary with season, constrained by radiation, barometric pressure, water depth, temperature, pH, DIC, and atmospheric CO2. These river simulations were used as boundary conditions to test the dynamic nature of the hyporheic zone in response to projected future temperature and atmospheric CO2 representing carbon emission futures, and to compare future and current hyporheic zone processing. Our data coupled with the predictive power of our numerical model reveal that hyporheic zones can serve many different roles throughout the year and indicate the importance of hyporheic cycling as a critical control point on watershed scale exports. The reliance of active versus passive ECPs on the timing of meltwater infiltration, including the possibility of a longer vernal window under future climate change indicates the importance of ECPs as controls of river-based indicators of river corridor hydrobiogeochemical function.

ABSTRACT. Water infiltration into heterogeneous soil controls the distribution of soil moisture in the vadose zone, and the dynamics of groundwater recharge, providing the link between climate, biogeochemical soil processes and vegetation dynamics. The transport of heat and solutes under variable water saturation conditions remains poorly understood due to the complex flow patterns induced by soil heterogeneity and hydrodynamic instabilities. Preferential flow paths, created by fingering or soil structure, are essential to understand vadose zone processes, as they become pathways for the fast migration of solutes and pollutants. They may also help infiltration water bypass evaporation and reach deep groundwater bodies.

We use modeling and numerical simulation to characterize the hydrodynamic controls on scalar fluxes in unsaturated soil. Two-phase fluid displacements in porous and fractured media are governed by various pore-scale processes, like abrupt pore invasion cascades and interfacial reconfigurations, flow by wetting films along the pore walls, and interfacial pinch-offs and reconnections. At the Darcy scale, these flow mechanisms lead to the appearance of flowing and stagnant regions in the soil, with mass and heat exchange between them, inducing non-equilibrium transport. We use pore-scale simulation to elucidate the basic processes behind preferential and non-equilibrium flow and transport, and propose Darcy-scale continuum models that allow to incorporate these nonequilibrium pore scale processes into field-scale simulations.

ABSTRACT. Recent advances in sensor and telecommunication technologies have enabled the establishment of real-time in situ monitoring networks for a variety of hydrological applications, including snow, soil moisture, groundwater table, and surface/groundwater quality. Their datasets have contributed significantly to improving and calibrating hydrological models. However, there is still a challenge of designing such a network; particularly to determine how many sensors are needed and where they need to be placed. In this study, we have developed a machine learning-based methodology to optimize sensor locations for hydrological monitoring. In particular, we take advantage of hydrological simulations to create the spatiotemporal maps of target variables to be measured in the network. The methodology consists of three steps in order to determine sensor locations in a systematic and flexible manner: (1) including priority locations, such as existing or regulatory compliance points, (2) diversifying locations that cover key environmental controls known to influence target variables (such as topography, plant functional types, geology), and (3) capturing the spatiotemporal heterogeneity of target variables. For the second step, we use unsupervised learning methods to cluster multiple environmental variables and to identify zonation for representing their co-variability and their value range. For the third step, we first use a principal component analysis method to compress the temporal dimension, and then use a Gaussian process model to capture and estimate the spatial heterogeneity of principal components across the domain. We demonstrate our approach in two case studies. At the East River watershed in Colorado, we design a soil moisture monitoring network at the watershed-scale based on ecohydrological simulations, using an integrated hydrological model PARFLOW-CLM (including ET, surface runoff, vadose zone flow, and groundwater flow). At the Savannah River Site F-Area in South Carolina, we design a groundwater monitoring network for water table and contaminant concentrations (uranium, tritium), based on a reactive transport model Amanzi. The acquired data can, in turn, improve the models further, iteratively enhancing and co-designing the monitoring network.

ABSTRACT. This study uses a bayesian network to identify the connectivity and dependency among key variables within a mechanistic ecosystem model. To demonstrate the method, controls on variabliity in non-growing season carbon and nitrogen cycling in an Alaskan watershed are explored.

ABSTRACT. We performed 3D simulations and uncertainty propagation of seismic waves generated by underground chemical explosions in granite as part of the Source Physics Experiments. The goal is to understand the near-field shear motion generation under conditions of uncertainty. We probabilistically show that significant shear motions can be generated by sliding on the joints caused by spherical wave propagation. Polarity of the shear motion may change during unloading when the stress state may favor joint sliding on a different joint set.

ABSTRACT. The non-linear effects of the asteroid impact on oceans are simulated using the hydrocode GEODYN to create the impact source for the shallow water wave propagation code, SWWP. The GEODYN-SWWP coupling is based on the structured adaptive mesh refinement infrastructure; SAMRAI. We also present the coupling scheme between hydrodynamic source using GEODYN and global climate circulation code GEOSCCM. We illustrate the scheme on the PDC 2017- 2019 asteroid impact scenarios. And their impact on the US, Europe and Asia shorelines.

ABSTRACT. A transient numerical study of conjugate flow, heat and mass transfer by natural convection of gases – air, carbon dioxide, noble gases – within an underground cavity partially filled with molten rock is presented.

ABSTRACT. We present an open-sourced solver to simulate two-phase flow in hybrid systems containing both solid-free regions and deformable porous matrices. This micro-continuum model is rooted in elementary physics and volume averaging principles, where a unique set of partial differential equations is used to represent flow in both regions and scales. The crux of the proposed model is that it tends asymptotically towards the Navier-Stokes volume-of-fluid approach in solid-free regions and towards multiphase Biot theory in porous regions. In this talk we go over the underlying physics of the model and showcase its flexibilty trough various applications related to fracturing in porous media and wave dissipation in poroelastic coastal barriers.