9 March 2026 in Water Resources Research
Groundwater model predictions are often inaccurate due to uncertainties in model structure, heterogeneous parameters, and initial conditions, leading to error accumulation during simulations. Traditional data assimilation (DA) faces severe computational challenges in high‐dimensional systems due to the costly inversion of large covariance matrices. In addition, the inaccurate estimation of background and observation error statistics introduces further biases. To address these challenges, we develop and evaluate an integrated framework that couples a computationally efficient deep learning surrogate model for rapid prediction with Latent Data Assimilation (LDA) to accurately correct simulations. The framework employs dimensionality reduction, specifically Proper Orthogonal Decomposition (POD), to project the high‐dimensional physical state into a low‐dimensional latent space. Data assimilation is then performed in this reduced space using the Ensemble Kalman Filter (EnKF). Results demonstrate that POD provides a robust representation of simulated concentration fields and interpolated observations for dimensionality reduction. The EnKF operating in the latent space effectively reduces prediction errors. Key advantages of the LDA framework include: enabling sparse observations to effectively inform global state updates through the low‐dimensional latent variables, achieving higher accuracy comparable to traditional physical‐space DA while requiring significantly fewer observations, and inherently filtering high‐frequency noise from the initial condition during the dimensionality reduction process. Collectively, these features establish LDA as a powerful and computationally efficient methodology for enhancing predictive accuracy and managing uncertainty in complex and high‐dimensional groundwater systems.
16 January 2026 in Water Resources Research
Groundwater flow and solute transport models, governed by partial differential equations (PDEs), are computationally intensive, particularly in large‐scale. Traditional numerical models are prohibitively expensive, and existing surrogate models often fail under out‐of‐distribution (OOD) conditions, such as unseen initial conditions, boundary configurations or altered source terms. To address these challenges, we propose a novel framework based on Operator Inference (OpInf), a physics‐informed surrogate modeling approach. OpInf preserves the structure of governing equations, ensuring physical consistency and interpretability, while significantly improving computational efficiency and generalization capabilities. By leveraging Proper Orthogonal Decomposition (POD) for dimensionality reduction and inferring reduced operators directly from simulation data, OpInf enables robust prediction of system behavior. We evaluate the proposed method through two case studies: the two‐dimensional and three‐dimensional solute transport problem under different point‐source concentration fluctuation release conditions with heterogeneous hydraulic conductivity. The inversion framework is further appraised by integrating OpInf with Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN‐GP) parameterization methods and the Ensemble Smoother (ES) data assimilation. Results demonstrate that OpInf relatively outperforms other surrogate models, particularly under OOD conditions and the inversion efficiency can be increased by over 99%. We establish OpInf as a transformative tool for dynamic surrogate groundwater modeling, offering robust generalization, reduced computational costs, and strong potential for real‐world applications.
13 October 2026 in EGUsphere
Geochemical reactions can alter groundwater flow by changing solution density and porous-medium properties. Representing these feedbacks requires consistent exchange of chemical states and physical properties between reaction, flow, and transport calculations. We present MF6PQC v1.0, a Python framework coupling groundwater flow and transport in MODFLOW 6 with geochemical reactions in PhreeqcRM. MF6PQC updates solution density and converts mineral-volume changes into changes in porosity, hydraulic conductivity, and diffusion coefficients. It supports three schemes for coupling transport and reactions while retaining existing flow and geochemical model-building workflows. Comparisons with two established reactive transport models showed agreement in component transport, mineral reactions, and the evolution of porous-medium properties. An idealized evaporating salt-lake model reproduced density-driven circulation and evaporite zonation. In a heterogeneous evaporite-leaching aquifer, dissolution concentrated flow along preferential pathways, and density feedback strengthened this redistribution. MF6PQC provides configurable coupling schemes and property relationships for examining these interactions in groundwater reactive transport.