SEARCH · Search NASA

Results for “Spatio-temporal dynamics”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2

All sources NASA NTRS DOE OSTI

DOE OSTI2024

Analyzing inference workloads for spatiotemporal modeling

Ensuring power grid resiliency, forecasting climate conditions, and optimization of transportation infrastructure are some of the many application areas where data is collected in both space and time. Spatiotemporal modeling is about modeling those patterns for forecasting future trends and carrying out critical decision-making by leveraging machine learning/deep learning. Once trained offline, field deployment of trained models for near real-time inference could be challenging because performance can vary significantly depending on the environment, available compute resources and tolerance to ambiguity in results. Users deploying spatiotemporal models for solving complex problems can benefit from analytical studies considering a plethora of system adaptations to understand the associated performance-quality trade-offs. To facilitate the co-design of next-generation hardware architectures for field deployment of trained models, it is critical to characterize the workloads of these deep learning (DL) applications during inference and assess their computational patterns at different levels of the execution stack. In this paper, we develop several variants of deep learning applications that use spatiotemporal data from dynamical systems. We study the associated computational patterns for inference workloads at different levels, considering relevant models (Long short-term Memory, Convolutional Neural Network and Spatio-Temporal Graph Convolution Network), DL frameworks (Tensorflow and PyTorch), precision (FP16, FP32, AMP, INT16 and INT8), inference runtime (ONNX and AI Template), post-training quantization (TensorRT) and platforms (Nvidia DGX A100 and Sambanova SN10 RDU). Overall, our findings indicate that although there is potential in mixed-precision models and post-training quantization for spatiotemporal modeling, extracting efficiency from contemporary GPU systems might be challenging. Instead, co-designing custom accelerators by leveraging optimized High Level Synthesis frameworks (such as SODA High-Level Synthesizer for customized FPGA/ASIC targets) can make workload-specific adjustments to enhance the efficiency.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Journey over Destination: Dynamic Sensor Placement Enhances Generalization

Reconstructing complex, high-dimensional global fields from limited data points is a challenge across various scientific and industrial domains. This is particularly important for recovering spatio-temporal fields using sensor data from, for example, laboratory-based scientific experiments, weather forecasting, or drone surveys. Given the prohibitive costs of specialized sensors and the inaccessibility ofcertain regions of the domain, achieving full field coverage is typically not feasible. Therefore, the development of machine learning algorithms trained to reconstruct fields given a limited dataset is of critical importance. In this study, we introduce a generalapproach that employs moving sensors to enhance data exploitation during the training of an attention based neural network, thereby improving field reconstruction. The training of sensor locations is accomplished using an end-to-end workflow, ensuringdifferentiability in the interpolation of field values associated to the sensors, and is simple to implement using differentiable programming. Additionally, we have incorporated a correction mechanism to prevent sensors from entering invalid regions within the domain. We evaluated our method using two distinct datasets; the results show that our approach enhances learning, as evidenced by improved test scores.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

ICED: An Integrated CGRA Framework Enabling DFVS-Aware Acceleration

oarse-grained reconfigurable arrays (CGRAs) are a promising solution to enable energy-efficient acceleration of applications from different domains. By leveraging reconfiguration at the functional level, they can adapt to significantly different computational patterns. Existing CGRA mapping approaches extract instruction-level parallelism, exploit loop-pipelining opportunities, guarantee the data dependency, and target high throughput of a given loop. However, the recurrence data-dependency in the DFG and the mismatch between required and available computing/communication resources complicate the mapping, and might lead to significant unbalances in the utilization of the CGRA's tiles. This results in wasted power for tiles with low utilization. Applying dynamic voltage and frequency scaling (DVFS) can potentially solve this challenge and improve energy efficiency by adjusting voltage and frequency of different tiles independently. CGRAs have also been successful in accelerating data-dependent streaming applications. However, in these applications, the execution time of each kernel in the pipeline might dynamically vary depending on the characteristics of the input. This also leads to under-utilization of resources for the dynamically changing kernels that do not limit the application throughput. DVFS can also improve energy efficiency for these applications by dynamically changing the voltage and frequency levels of tiles that host non performance-constraining kernels. This paper proposes ICEDTEA -- an integrated DVFS-aware framework to map applications on CGRAs that support power islands. ICEDTEA proposes a CGRA architecture supporting DVFS islands at varying granularity (from a single tile to a group of tiles) and the related DVFS-aware compilation and mapping toolchain. ICEDTEA is the first work that introduces DVFS support for spatio-temporal CGRAs at power-island levels. The experimental evaluation shows that ICEDTEA improves average utilization by 2.3$\times$ and energy-efficiency by 1.32$\times$ over a conventional CGRA. With streaming applications, ICEDTEA improves energy efficiency by 1.12$\times$ over a state-of-the-art CGRA that introduces partial dynamic reconfiguration to adapt to variations in kernels' throughput.

Tan, Cheng↗

DOE OSTI2025

Resolving turbulent magnetohydrodynamics: a hybrid operator-diffusion framework

We present a hybrid machine learning framework that combines physics-informed neural operators (PINOs) with score-based generative diffusion models to simulate the full spatio-temporal evolution of two-dimensional, incompressible, resistive magnetohydrodynamic turbulence across a broad range of Reynolds numbers (Re). The framework leverages the equation-constrained generalization capabilities of PINOs to predict coherent, low-frequency dynamics, while a conditional diffusion model stochastically corrects high-frequency residuals, enabling accurate modeling of fully developed turbulence. Trained on a comprehensive ensemble of high-fidelity simulations with Re ϵ {100, 250, 500, 750, 1000, 3000, 10000}, the approach achieves state-of-the-art accuracy in regimes previously inaccessible to deterministic surrogates. At Re = 1000 and 3000, the model faithfully reconstructs the full spectral energy distributions of both velocity and magnetic fields late into the simulation, capturing non-Gaussian statistics, intermittent structures, and cross-field correlations with high fidelity. At extreme turbulence levels (Re = 10 000), it remains the first surrogate capable of recovering the high-wavenumber evolution of the magnetic field, preserving large-scale morphology and enabling statistically meaningful predictions.

Diffusion-Integrated Neural Operators↗

DOE OSTI2026

Learning turbulent flows with generative models for super resolution and sparse flow reconstruction

Neural operators are promising surrogates for dynamical systems but when trained with standard L 2 losses they tend to oversmooth fine-scale turbulent structures. Here, we show that combining operator learning with generative modeling overcomes this limitation. We consider three practical turbulent-flow challenges where conventional neural operators fail: spatio-temporal super-resolution, forecasting, and sparse flow reconstruction. For Schlieren jet super-resolution, an adversarially trained neural operator (adv-NO) reduces the energy-spectrum error by 15 × while preserving sharp gradients at neural operator-like inference cost. For 3D homogeneous isotropic turbulence, adv-NO trained on only 160 timesteps from a single trajectory forecasts accurately for five eddy-turnover times and offers 114 × wall-clock speed-up at inference than the baseline diffusion-based forecasters, enabling near-real-time rollouts. For reconstructing cylinder wake flows from highly sparse Particle Tracking Velocimetry-like inputs, a conditional generative model infers full 3D velocity and pressure fields with correct phase alignment and statistics. These advances enable accurate reconstruction and forecasting at low compute cost, bringing near-real-time analysis and control within reach in experimental and computational fluid mechanics.

Fluid dynamics↗

DOE OSTI2025

Resolving the dynamic correlated disorder in KTa 1- x Nb x O 3

Understanding the complex temporal and spatial correlations of ions in disordered perovskite oxides is critical to rationalize their functional properties. Here, in this study, we provide new insights into the longstanding controversy regarding the off-centering of transition metal ions in the archetypal ferroelectric alloy KTa 1-x Nb x O 3 (KTN). By mapping the full energy (E) and wavevector (Q) dependence of the dynamical structure factor S(Q, E) using neutron scattering, and rationalizing our observations with atomistic simulations leveraging machine learning, we fully resolve the static vs dynamic nature of diffuse scattering sheets, as well as their composition (x) and temperature dependence. Our first-principles simulations, extended with machine-learning molecular dynamics, reproduce both inelastic neutron spectra and diffuse features, and establish how dynamically-correlated transition metal off-centerings couple to phonons, unifying local and collective viewpoints. This study sheds new light into an exemplary ferroelectric systems and shows the importance of mapping the full S(Q, E) to reveal critical spatio-temporal correlations of atomic disorder from which functional properties emerge.

42 ENGINEERING↗

DOE OSTI2024

PSInet: a new global water potential network

Abstract Given the pressing challenges posed by climate change, it is crucial to develop a deeper understanding of the impacts of escalating drought and heat stress on terrestrial ecosystems and the vital services they offer. Soil and plant water potential play a pivotal role in governing the dynamics of water within ecosystems and exert direct control over plant function and mortality risk during periods of ecological stress. However, existing observations of water potential suffer from significant limitations, including their sporadic and discontinuous nature, inconsistent representation of relevant spatio-temporal scales and numerous methodological challenges. These limitations hinder the comprehensive and synthetic research needed to enhance our conceptual understanding and predictive models of plant function and survival under limited moisture availability. In this article, we present PSInet (PSI—for the Greek letter Ψ used to denote water potential), a novel collaborative network of researchers and data, designed to bridge the current critical information gap in water potential data. The primary objectives of PSInet are as follows. (i) Establishing the first openly accessible global database for time series of plant and soil water potential measurements, while providing important linkages with other relevant observation networks. (ii) Fostering an inclusive and diverse collaborative environment for all scientists studying water potential in various stages of their careers. (iii) Standardizing methodologies, processing and interpretation of water potential data through the engagement of a global community of scientists, facilitated by the dissemination of standardized protocols, best practices and early career training opportunities. (iv) Facilitating the use of the PSInet database for synthesizing knowledge and addressing prominent gaps in our understanding of plants’ physiological responses to various environmental stressors. The PSInet initiative is integral to meeting the fundamental research challenge of discerning which plant species will thrive and which will be vulnerable in a world undergoing rapid warming and increasing aridification.

Forestry↗

DOE OSTI2025

pyTCR: A tropical cyclone rainfall model for python

pyTCR is a climatology software package developed in the Python programming language. It integrates the capabilities of several legacy physical models and increases computational efficiency to allow rapid estimation of tropical cyclone (TC) rainfall consistent with the large-scale environment. Specifically, pyTCR implements a horizontally distributed and vertically integrated model [Zhu et al., 2013] for simulating rainfall driven by TCs. Along storm tracks, rainfall is estimated by computing the cross-boundary-layer, upward water vapor transport caused by different mechanisms including frictional convergence, vortex stretching, large-scale baroclinic effect (i.e., wind shear), topographic forcing, and radiative cooling [Lu et al., 2018]. The package provides essential functionalities for modeling and interpreting spatio-temporal TC rainfall data. pyTCR requires a limited number of model input parameters, making it a convenient and useful tool for analyzing rainfall mechanisms driven by TCs. To sample rare (most intense) rainfall events that are often of great societal interest, pyTCR adapts and leverages outputs from a statistical-dynamical TC downscaling model [Lin et al., 2023] capable of rapidly generating a large number of synthetic TCs given a certain climate. As a result, pyTCR significantly reduces computational effort and improves the efficiency in capturing extreme TC rainfall events at the tail of the distributions from limited datasets. Furthermore, the TC downscaling model is forced entirely by large-scale environmental conditions from reanalysis data or coupled General Circulation Models (GCMs), simplifying the projection of TC-induced rainfall and wind speed under future climate using pyTCR. Finally, pyTCR can be coupled with hydrological and wind models to assess risks associated with independent and compound events (e.g., storm surges and freshwater flooding).

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Trends and Drivers of Terrestrial Sources and Sinks of Carbon Dioxide: An Overview of the TRENDY Project

The terrestrial biosphere plays a major role in the global carbon cycle, and there is a recognized need for regularly updated estimates of land-atmosphere exchange at regional and global scales. An international ensemble of Dynamic Global Vegetation Models (DGVMs), known as the “Trends and drivers of the regional scale terrestrial sources and sinks of carbon dioxide” (TRENDY) project, quantifies land biophysical exchange processes and biogeochemistry cycles in support of the annual Global Carbon Budget assessments and the REgional Carbon Cycle Assessment and Processes, phase 2 project. DGVMs use a common protocol and set of driving data sets. A set of factorial simulations allows attribution of spatio-temporal changes in land surface processes to three primary global change drivers: changes in atmospheric CO 2 , climate change and variability, and Land Use and Land Cover Changes (LULCC). Here, we describe the TRENDY project, benchmark DGVM performance using remote-sensing and other observational data, and present results for the contemporary period. Simulation results show a large global carbon sink in natural vegetation over 2012–2021, attributed to the CO 2 fertilization effect (3.8 ± 0.8 PgC/yr) and climate (–0.58 ± 0.54 PgC/yr). Forests and semi-arid ecosystems contribute approximately equally to the mean and trend in the natural land sink, and semi-arid ecosystems continue to dominate interannual variability. The natural sink is offset by net emissions from LULCC (–1.6 ± 0.5 PgC/yr), with a net land sink of 1.7 ± 0.6 PgC/yr. Despite the largest gross fluxes being in the tropics, the largest net land-atmosphere exchange is simulated in the extratropical regions.

58 GEOSCIENCES↗

DOE OSTI2024

Single-molecule imaging of lignocellulose deconstruction by SCATTIRSTORM microscopy

The goal of this project is to build a multimodal optical microscope to measure the binding, processive degradation, and pausing behaviors of cellulases as they interact with and degrade both synthetic and naturally occurring lignocellulosic walls. To achieve this, we will use high spatio-temporal single-molecule imaging to track cellulases, while visualizing specific molecular components of cellulose, lignin and hemicellulose, that make up their lignocellulose substrate. The microscope will combine Interferometric Scattering (iSCAT), which provides unprecedented spatiotemporal resolution; Total Internal Reflection Fluorescence (TIRF), which provides single-molecule resolution of multiple fluorophore-labeled molecules; and Stochastic Reconstruction (STORM), which allows for three-dimensional super-resolution imaging of intact plant cell walls during degradation. Initial studies will investigate cellulase dynamics on in vitro-assembled cell wall analogs, and later work will progress to using native plant cell walls.

09 BIOMASS FUELS↗

DOE OSTI2025

Parallel-in-Time Solution of Hyperbolic PDE Systems via Characteristic-Variable Block Preconditioning

We consider the parallel-in-time solution of both linear and nonlinear hyperbolic partial differential equation (PDE) systems in one spatial dimension. In the nonlinear setting, the discretized equations are solved with a preconditioned residual iteration based on a global linearization. The linear(ized) equation systems are approximately solved parallel-in-time using a block preconditioner applied in the characteristic variables of the underlying linear(ized) hyperbolic PDE. This change of variables is motivated by the observation that intervariable coupling between characteristic variables is weak, at least locally where spatio-temporal variations in the eigenvectors of the associated flux Jacobian are sufficiently small, while that between the original variables is not. For an ℓ-dimensional system of PDEs, applying the preconditioner consists of solving a sequence of ℓ scalar linear(ized)-advection-like problems, each associated with a different characteristic wave-speed in the underlying linear(ized) PDE. Furthermore, we approximately solve these linear advection problems using multigrid reduction-in-time (MGRIT); however, any other suitable parallel-in-time method could be used. Numerical examples are shown for the (linear) acoustics equations in heterogeneous media and for the (nonlinear) shallow water equations and Euler equations of gas dynamics with shocks and rarefactions. For many test problems, the solver converges in just a handful of iterations and with mesh-independent convergence rates.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Interactive multiscale modeling to bridge atomic properties and electrochemical performance in Li-CO 2 battery design

Li-CO 2 batteries are promising energy storage systems due to their high theoretical energy density and CO 2 fixation capability, relying on reversible Li 2 CO 3 /C formation during discharge/charge cycles. Here, we present a multiscale modeling framework integrating Density Functional Theory (DFT), Ab-Initio Molecular Dynamics (AIMD), classical Molecular Dynamics (MD), and Finite Element Analysis (FEA) to investigate atomic and cell-level properties. The considered Li-CO 2 battery consists of a lithium metal anode, an ionic liquid electrolyte, and a carbon cloth cathode with Sb 0.67 Bi 1.33 Te 3 catalyst. DFT and AIMD determined the electrical conductivities of Sb 0.67 Bi 1.33 Te 3 and Li 2 CO 3 using the Kubo–Greenwood formalism and studied the CO 2 reduction mechanism on the cathode catalyst. MD simulations calculated the CO 2 diffusion coefficient, Li + transference number, ionic conductivity, and Li + solvation structure. The FEA model, parameterized with atomistic simulation data, reproduced the available experimental voltage–capacity profile at 1 mA/cm 2 and revealed spatio-temporal variations in Li 2 CO 3 /C deposition, porosity, and CO 2 concentration dependence on discharge rates in the cathode. Accordingly, Li 2 CO 3 can form large and thin film deposits, leading to dispersed and local porosity changes at 0.1 mA/cm 2 and 1 mA/cm 2 , respectively. The capacity decreases exponentially from 81,570 mAh/g at 0.1 mA/cm 2 to 6200 mAh/g at 1 mA/cm 2 , due to pore clogging from excessive discharge product deposition that limits CO 2 transport to the cathode interior. Therefore, the performance of Li-CO 2 batteries can be improved by enhancing CO 2 transport, regulating Li 2 CO 3 deposition, and optimizing cathode architecture.

Battery performance↗

DOE OSTI2025

A new biogeochemical modelling framework (FLaMe-v1.0) for lake methane emissions on the regional scale: development and application to the European domain

This study presents a new physical-biogeochemical modelling framework for simulating lake methane (CH 4 ) emissions at regional scales. The new model, FLaMe-v1.0 (Fluxes of Lake Methane), rests on an innovative, computationally efficient lake clustering approach that enables the simulation of CH 4 emissions across a large number of lakes. Building on the Canadian Small Lake Model (CSLM) that simulates the lake physics, we develop a suite of biogeochemical modules to simulate transient dynamics of organic Carbon (C), Oxygen (O 2 ), and CH 4 . We first test the performance of FLaMe-v1.0 by analyzing physical and biogeochemical processes in two theoretical lakes with characteristics that can be considered representative for many lakes (an oligotrophic, deep lake driven by cold climate versus a eutrophic, shallow lake driven by warm climate). Next, we evaluate the model by comparing simulated and observed timeseries of CH 4 emissions in four well-surveyed lakes. We then apply FLaMe-v1.0 at the European scale to evaluate simulated diffusive and ebullitive lake CH 4 fluxes against in-situ measurements in both boreal and central European regions. Finally, we provide a first assessment of the spatio-temporal variability in CH 4 emissions from European lakes with a surface area comprised between 0.1–1000 km 2 (n= 108 407, total area = 1.33 × 105 km 2 ), indicating a total emission of 0.97 ± 0.23 Tg CH 4 yr −1 , with the uncertainty constrained by combining FLaMe-v1.0 and machine learning techniques. Moreover, 30 % and 70 % of these CH 4 emissions are through diffusive and ebullitive pathways, respectively. Annually averaged CH 4 emission rates per unit lake area during 2010–2016 have a South-to-North decreasing gradient, resulting in a mean over the European domain as 7.39 g CH 4 m −2 yr −1 . Our simulations reveal a strong seasonality (with ice-blocking effects accounted for) in European lake CH 4 emissions, with nearly ten times higher emissions during late summer than during winter. This pronounced seasonal variation highlights the importance of accounting for the sub-annual variability in CH 4 emissions to accurately constrain regional CH 4 budgets. In the future, FLaMe-v1.0 could be embedded into Earth System Models to investigate the feedback between climate warming and global lake CH 4 emissions.

Maisonnier, Manon [Free Univ. of Brussels (Belgium↗

DOE OSTI2025

Hybrid learning techniques for scientific data reduction with performance guarantees

The research initiatives supported by the U.S. Department of Energy (DOE) Grant DE-SC0022265 are fundamentally aimed at pioneering advanced machine learning (ML) techniques for scientific data compression within high-performance computing (HPC) environments. This comprehensive body of work addresses the critical challenge posed by the exponential growth of data generated by scientific simulations in domains such as fusion energy, climate modeling, and computational fluid dynamics (CFD). A core objective is to develop compression algorithms that achieve substantial data reduction—often by orders of magnitude—while rigorously ensuring the fidelity of both the primary data (PD) and scientifically crucial derived quantities of interest (QoI). The methodologies deployed under this grant integrate sophisticated deep learning architectures, prominently featuring autoencoders, advanced generative models like conditional diffusion, and hybrid learning techniques. Key innovations include the development of Guaranteed Autoencoders (GAE) and the Guaranteed Conditional Diffusion with Tensor Correction (GCDTC) framework, which provide explicit, instance-level error bounds on reconstructed data. Furthermore, specialized strategies such as nonlinear constraint satisfaction are employed to preserve the integrity of QoI, a vital requirement for the trustworthiness of downstream scientific analyses. This research also focuses on the design and implementation of scalable, GPU-accelerated software pipelines that seamlessly integrate into existing HPC workflows, ensuring both computational efficiency and practical applicability. The CAESAR framework, for example, unifies foundation and generative models to create an adaptive and efficient compression solution for spatio-temporal scientific data. Collectively, these efforts represent a significant advancement in mitigating the scientific data deluge, enabling more effective data management, accelerated scientific discovery, and optimized utilization of HPC resources.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Final report- UFL - RAPIDS2: A SciDAC Institute for Computer Science, Data, and Artificial Intelligence

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Multiscale drivers of extreme southern California flooding: ENSO, MJO, North Pacific jet, and atmospheric rivers

Extreme rainfall and flooding, driven by a powerful atmospheric river (AR) and a persistent Madden-Julian Oscillation (MJO), hit Southern California in February 2024 during the 2023–2024 El Niño, affecting over 10 million people. ARs are key contributors to extreme rainfall and flooding along the U.S. West Coast. Although the AR-MJO link has been documented, its spatio-temporal variability remains a major forecasting and risk-management challenge. Combining precipitation, stream gauge and demographic data, we quantify the physical drivers and population exposure to this extreme event. Leveraging a Lagrangian MJO precipitation tracking algorithm, we unravel the multiscale interactions responsible for the AR’s development. El Niño favored a large, long-lived MJO that interacted with the North Pacific Jet (NPJ) over more than three weeks. The MJO convective outflow modulated the NPJ by inducing negative potential vorticity advection along the tropopause. The ensuing NPJ extension and acceleration induced explosive cyclogenesis, whose AR-driven moisture transport resulted in extreme rainfall.

Atmospheric dynamics↗

DOE OSTI2025

Graph-Based Prediction of Spatio-Temporal Vaccine Hesitancy From Insurance Claims Data

Growing vaccine hesitancy is contributing to the decline in immunization rates for highly contagious, vaccine-preventable childhood diseases. Therefore, there has been a significant interest in understanding how hesitancy is spreading at higher spatio-temporal resolutions, enabling more targeted interventions. Motivated by this, we study the problem of prediction of vaccine hesitancy at the ZIP Code level, referred to as the VaxHesitancy problem. A significant challenge for this problem is the lack of high-resolution data that indicates hesitancy. Here, we develop a hybrid VaxHesSTL framework that combines a Graph Neural Network (GNN) and a Recurrent Neural Network (RNN) to address the VaxHesitancy problem. The GNN uses a ZIP Code-level network to capture spatial signals from neighboring areas, while the RNN models the temporal dynamics present in the data. We train and evaluate VaxHesSTL using a large dataset, namely the All-Payer Claims Databases (APCD), for Virginia, consisting of insurance claims from over five million individuals for six years. We find that an aggregated contact network or graph, developed from a detailed activity-based population network, plays an important role in the performance of VaxHesSTL, compared to graph models based solely on spatial proximity. Experiments demonstrate that VaxHesSTL outperforms a range of state-of-the-art baselines, which rely solely on historical time series data without accounting for spatial relationships. Since hesitancy data at higher spatial resolution is often unavailable or hard to get, we incorporate an active learning approach with our VaxHesSTL framework to optimize the training set without compromising the prediction performance. We find that hesitancy data for only 18% of ZIP Codes selected by active learning allows us to forecast hesitancy for all the ZIP Codes in the Virginia.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2026

Bayesian Physics Informed Spatio-Temporal Network for Streamflow Data Imputation

Reliable reconstruction of incomplete streamflow records is critical for improving hydrological forecasting, flood preparedness, and water resource management. However, large observational gaps and uncertainties in governing physical parameters limit the accuracy of traditional statistical and machinelearning imputation frameworks. To address these challenges, we develop a Bayesian Physics-Informed Spatio-Temporal Network (BPI-STNet) that jointly captures spatial and temporal dependencies while enforcing hydrologic consistency through embedded physical constraints. The framework integrates a GraphSAGE-LSTM architecture to model spatial connectivity across gauges and temporal flow dynamics, coupled with a Bayesian update mechanism to estimate uncertain parameters in a simplified water-balance framework. Unlike conventional physics-informed networks that rely on sampling-based posterior estimation, BPI-STNet derives an analytic solution to the inverse problem, allowing closed-form Bayesian updates of uncertain parameters Λ={α,β,k} using Gaussian priors and likelihoods. Applied to daily observations from the Susquehanna River Basin (1980-2022), BPI-STNet achieves substantial improvements over a purely data-driven RGNN baseline, which reduced RMSE by 23 % and MAE by 9 %, and achieving an average NSE values up to 0.96. The results demonstrate that coupling Bayesian inference with physics-informed learning yields physically consistent, uncertainty-aware reconstructions that preserve the temporal persistence and statistical distribution of observed flows. The proposed framework establishes a generalizable paradigm for data-sparse hydrologic systems where both data fidelity and physical interpretability are essential.

Krishnan Kutty Ambika, Anukesh [ORNL] (ORCID:00000↗