Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “streamflow prediction”

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.

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

At least 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Enhancing Streamflow Reanalysis Across the Conterminous US Leveraging Multiple Gridded Precipitation Data Sets

Streamflow observations, essential for various water resource applications, are often unavailable at critical locations in need. Although different models have been proposed to enhance streamflow predictability at ungauged locations, the challenge extends beyond model fidelity. Differences in meteorologic forcing data sets, precipitation in particular, can significantly affect the accuracy of hydrologic predictions. This challenge intensifies across regions characterized by diverse hydro-climatological and geographical conditions, such as in the conterminous US (CONUS) where a single precipitation product struggles to consistently replicate observed hydrographs, particularly peak flow dynamics. To enhance streamflow predictions, we utilize a VIC-RAPID hydrologic modeling framework driven by multiple commonly used meteorological forcing data sets, such as Daymet, PRISM, ST4, AORC, and their hybrids and create multiple sets of 40-year (1980–2019) hourly, daily, and monthly streamflow reanalysis, Dayflow Version 2, for 2.7 million river reaches across the CONUS. Most forcings lead to skillful streamflow performance, except for ST4 in the mountainous west, where severe radar blockage adversely affects the accuracy. The evaluation using over 6,000 hourly stream gauges shows that hourly AORC and ST4 lead to improved annual peak flow performance over Daymet—driven streamflow (Dayflow V1), particularly in smaller basins, highlighting the value of high temporal resolution forcings in hydrologic predictions. Compared with other benchmark data sets like National Water Model V3.0, AORC-driven VIC-RAPID exhibits improved regional streamflow performance, with comparable peak flow representation. We envision that multi-forcing streamflow reanalysis data can inform regions in need of forcing data enhancement, diagnose hydrologic model performance, and benefit diverse water resource applications.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2010

The Contribution of Soil Moisture Information to Forecast Skill: Two Studies

This talk briefly describes two recent studies on the impact of soil moisture information on hydrological and meteorological prediction. While the studies utilize soil moisture derived from the integration of large-scale land surface models with observations-based meteorological data, the results directly illustrate the potential usefulness of satellite-derived soil moisture information (e.g., from SMOS and SMAP) for applications in prediction. The first study, the GEWEX- and ClIVAR-sponsored GLACE-2 project, quantifies the contribution of realistic soil moisture initialization to skill in subseasonal forecasts of precipitation and air temperature (out to two months). The multi-model study shows that soil moisture information does indeed contribute skill to the forecasts, particularly for air temperature, and particularly when the initial local soil moisture anomaly is large. Furthermore, the skill contributions tend to be larger where the soil moisture initialization is more accurate, as measured by the density of the observational network contributing to the initialization. The second study focuses on streamflow prediction. The relative contributions of snow and soil moisture initialization to skill in streamflow prediction at seasonal lead, in the absence of knowledge of meteorological anomalies during the forecast period, were quantified with several land surface models using uniquely designed numerical experiments and naturalized streamflow data covering mUltiple decades over the western United States. In several basins, accurate soil moisture initialization is found to contribute significant levels of predictive skill. Depending on the date of forecast issue, the contributions can be significant out to leads of six months. Both studies suggest that improvements in soil moisture initialization would lead to increases in predictive skill. The relevance of SMOS and SMAP satellite-based soil moisture information to prediction are discussed in the context of these studies.

Koster, Randal↗
DOE OSTI2023

Closing in on Hydrologic Predictive Accuracy: Combining the Strengths of High-Fidelity and Physics-Agnostic Models

Applications of process-based models (PBM) for predictions are confounded by multiple uncertainties and computational burdens, resulting in appreciable errors. A novel modeling framework combining a high-fidelity PBM with surrogate and machine learning (ML) models is developed to tackle these challenges and applied for streamflow prediction. A surrogate model permits high computational efficiency of a PBM solution at a minimum loss of its accuracy. A novel probabilistic ML model partitions the PBM-surrogate prediction errors into reducible and irreducible types, quantifying their distributions that arise due to both explicitly perceived uncertainties (such as parametric) or those that are entirely hidden to the modeler (not included or unexpected). Using this approach, we demonstrate a substantial improvement of streamflow predictive accuracy for a case study urbanized watershed. Such a framework provides an efficient solution combining the strengths of high-fidelity and physics-agnostic models for a wide range of prediction problems in geosciences.

58 GEOSCIENCES↗
NASA NTRS2023

Late-Fall Satellite-Based Soil Moisture Observations Show Clear Connections to Subsequent Spring Streamflow

Because runoff production is more efficient over wetter soils, and because soil moisture has an intrinsic memory, soil moisture information can potentially contribute to the accuracy of streamflow predictions at seasonal leads. In this work, we use surface (0-5 cm) soil moisture retrievals obtained with the National Aeronautics and Space Administration’s Soil Moisture Active Passive satellite instrument in conjunction with streamflow measurements taken within 236 intermediate-scale (2,000 – 10,000 km2) unregulated river basins in the conterminous United States to show that late-fall satellite-based surface soil moisture estimates are indeed strongly connected to subsequent springtime streamflow. We thus show that the satellite-based soil moisture retrievals, all by themselves, have the potential to produce skillful seasonal streamflow predictions several months in advance. In poorly instrumented regions, they could perform better than reanalysis soil moisture products in this regard.

Randal D Koster↗
DOE OSTI2026

MOSAIC-CONUS: A Multimodal, Multi-Temporally Paired Dataset for Earth Sciences

Earth embeddings—vector representations of geographic locations indexed in space and time—are emerging as a unifying interface for geospatial AI. However, their quality depends not only on model design, but on how multimodal Earth observation (EO) data are spatially indexed, temporally aligned, and cross-modally associated during pretraining. We introduce MOSAIC-CONUS (Multimodal Observations with Spatially Aligned Imagery, Urban Points of Interest, In-Situ Measurements and Text Captions), a large-scale EO dataset over the contiguous United States, organized around 250,000 stratified point indices that serve as stable spatial keys across seven modalities: active radar, passive optical imagery, lidar-derived elevation, land cover, functional context, hydrometeorological measurements, and textual summaries. Unlike existing EO datasets, MOSAIC-CONUS introduces four contributions not jointly addressed in prior work: 1. an open-source, large-scale multimodal EO corpus structured around point-indexed data designed to support Earth embedding learning; 2. explicit radar-optical pairing tables spanning twelve temporal alignment regimes, formalizing cross-sensor alignment as a controllable variable for analyzing how temporal mismatch across modalities influences learned embeddings quality; 3. a benchmark suite spanning cross-modal retrieval, annual nightlights regression, and basin-held-out streamflow prediction, positioning MOSAIC-CONUS as a benchmark-ready resource for multimodal AI systems; and 4. a language-based embedding layer through co-registered textual summaries, enabling Earth embeddings to function as a queryable interface for agentic AI systems. The dataset and pairing protocols are publicly released.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

The suitability of differentiable, physics-informed machine learning hydrologic models for ungauged regions and climate change impact assessment

As a genre of physics-informed machine learning, differentiable process-based hydrologic models (abbreviated as δ or delta models) with regionalized deep-network-based parameterization pipelines were recently shown to provide daily streamflow prediction performance closely approaching that of state-of-the-art long short-term memory (LSTM) deep networks. Meanwhile, δ models provide a full suite of diagnostic physical variables and guaranteed mass conservation. Here, we ran experiments to test (1) their ability to extrapolate to regions far from streamflow gauges and (2) their ability to make credible predictions of long-term (decadal-scale) change trends. We evaluated the models based on daily hydrograph metrics (Nash–Sutcliffe model efficiency coefficient, etc.) and predicted decadal streamflow trends. For prediction in ungauged basins (PUB; randomly sampled ungauged basins representing spatial interpolation), δ models either approached or surpassed the performance of LSTM in daily hydrograph metrics, depending on the meteorological forcing data used. They presented a comparable trend performance to LSTM for annual mean flow and high flow but worse trends for low flow. For prediction in ungauged regions (PUR; regional holdout test representing spatial extrapolation in a highly data-sparse scenario), δ models surpassed LSTM in daily hydrograph metrics, and their advantages in mean and high flow trends became prominent. In addition, an untrained variable, evapotranspiration, retained good seasonality even for extrapolated cases. The δ models' deep-network-based parameterization pipeline produced parameter fields that maintain remarkably stable spatial patterns even in highly data-scarce scenarios, which explains their robustness. Combined with their interpretability and ability to assimilate multi-source observations, the δ models are strong candidates for regional and global-scale hydrologic simulations and climate change impact assessment.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Leveraging Groundwater Dynamics to Improve Predictions of Summer Low-Flow Discharges

Summer streamflow predictions are critical for managing water resources; however, warming-induced shifts from snow to rain regimes impact low-flow predictive models. Additionally, reductions in snowpack drive earlier peak flows and lower summer flows across the western United States increasing reliance on groundwater for maintaining summer streamflow. However, it remains poorly understood how groundwater contributions vary interannually. Here, we quantify recession limb groundwater (RLGW), defined as the proportional groundwater contribution to the stream during the period between peak stream flow and low flow, to predict summer low flows across three diverse western US watersheds. We ask (a) how do snow and rain dynamics influence interannual variations of RLGW contributions and summer low flows?; (b) which watershed attributes impact the effectiveness of RLGW as a predictor of summer low flows? Linear models reveal that RLGW is a strong predictor of low flows across all sites and drastically improves low-flow prediction compared to snow metrics at a rain-dominated site. Results suggest that strength of RLGW control on summer low flows may be mediated by subsurface storage. Subsurface storage can be divided into dynamic (i.e., variability saturated) and deep (i.e., permanently saturated) components, and we hypothesize that interannual variability in dynamic storage contribution to streamflow drives RLGW variability. In systems with a higher proportion of dynamic storage, RLGW is a better predictor of summer low flow because the stream is more responsive to dynamic storage contributions compared to deep-storage-dominated systems. Overall, including RLGW improved low-flow prediction across diverse watersheds.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2010

Relationship Between Satellite-Derived Snow Cover and Snowmelt-Runoff Timing and Stream Power in the Wind River Range, Wyoming

Earlier onset of springtime weather including earlier snowmelt has been documented in the western United States over at least the last 50 years. Because the majority (>70%) of the water supply in the western U.S. comes from snowmelt, analysis of the declining spring snowpack (and shrinking glaciers) has important implications for streamflow management. The amount of water in a snowpack influences stream discharge which can also influence erosion and sediment transport by changing stream power, or the rate at which a stream can do work such as move sediment and erode the stream bed. The focus of this work is the Wind River Range (WRR) in west-central Wyoming. Ten years of Moderate-Resolution Imaging Spectroradiometer (MODIS) snow-cover, cloud- gap-filled (CGF) map products and 30 years of discharge and meteorological station a are studied. Streamflow data from six streams in the WRR drainage basins show lower annual discharge and earlier snowmelt in the decade of the 2000s than in the previous three decades, though no trend of either lower streamflow or earlier snowmelt was observed using MODIS snow-cover maps within the decade of the 2000s. Results show a statistically-significant trend at the 95% confidence level (or higher) of increasing weekly maximum air temperature (for three out of the five meteorological stations studied) in the decade of the 1970s, and also for the 40-year study period. MODIS- derived snow cover (percent of basin covered) measured on 30 April explains over 89% of the variance in discharge for maximum monthly streamflow in the decade of the 2000s using Spearman rank correlation analysis. We also investigated stream power for Bull Lake Creek Above Bull Lake from 1970 to 2009; a statistically-significant trend toward reduced stream power was found (significant at the 90% confidence level). Observed changes in streamflow and stream power may be related to increasing weekly maximum air temperature measured during the 40-year study period. The strong relationship between percent of basin covered and streamflow indicates that MODIS data is useful for predicting streamflow, leading to improved reservoir management.

Hall, Dorothy K.↗
NASA NTRS2010

Relationship Between Satellite-Derived Snow Cover and Snowmelt-Runoff Timing and Stream Power in the Wind River Range, Wyoming

Earlier onset of springtime weather including earlier snowmelt has been documented in the western United States over at least the last 50 years. Because the majority (>70%) of the water supply in the western U.S. comes from snowmelt, analysis of the declining spring snowpack (and shrinking glaciers) has important implications for streamflow management. The amount of water in a snowpack influences stream discharge which can also influence erosion and sediment transport by changing stream power, or the rate at which a stream can do work such as move sediment and erode the stream bed. The focus of this work is the Wind River Range (WRR) in west-central Wyoming. Ten years of Moderate-Resolution Imaging Spectroradiometer (MODIS) snow-cover, cloud- gap-filled (CGF) map products and 30 years of discharge and meteorological station data are studied. Streamflow data from six streams in the WRR drainage basins show lower annual discharge and earlier snowmelt in the decade of the 2000s than in the previous three decades, though no trend of either lower streamflow or earlier snowmelt was observed using MODIS snow-cover maps within the decade of the 2000s. Results show a statistically-significant trend at the 95% confidence level (or higher) of increasing weekly maximum air temperature (for three out of the five meteorological stations studied) in the decade of the 1970s, and also for the 40-year study period. MODIS-derived snow cover (percent of basin covered) measured on 30 April explains over 89% of the variance in discharge for maximum monthly streamflow in the decade of the 2000s using Spearman rank correlation analysis. We also investigated stream power for Bull Lake Creek Above Bull Lake from 1970 to 2009; a statistically-significant end toward reduced stream power was found (significant at the 90% confidence level). Observed changes in streamflow and stream power may be related to increasing weekly maximum air temperature measured during the 40-year study period. The strong relationship between percent of basin covered and streamflow indicates that MODIS data is useful for predicting streamflow, leading to improved reservoir management

Hall, Dorothy K.↗
NASA NTRSDate not supplied

Lessons learned from replicating services for flood prediction and monitoring in Asia to the assessment of hurricane impacts in Central America

In October and November 2020, two dangerous back-to-back hurricanes, Eta and Iota, made landfall in Central America. The SERVIR program - a joint effort of NASA and the U.S. Agency for International Development, and whose motto is “connecting space to village” was able to leverage two tools originally developed for use in other regions for predicting and assessing the flood impacts of the hurricanes. The GEOGLoWS Streamflow Prediction tool - originally implemented in the Hindu Kush Himalayan region - was used for predicting potential flooding ahead of landfall by Eta and Iota. In conjunction, the Hydrologic Remote Sensing Analysis for Floods (HYDRAFloods) framework - originally developed along with SERVIR-Mekong - was used for post-event flood mapping, leveraging its ability to map floods in cloud-covered areas using synthetic aperture radar (SAR) imagery from the Copernicus program. Both tools were used in support of disaster coordination efforts being led by the Central American Regional Disaster Prevention Center (CEPREDENAC), in conjunction with its sister agency, the Regional Water Resources Committee (CRRH). The support provided to regional entities - and to their stakeholder national governments - served as an example of rapid generation of Earth observation products for disaster response. Feedback on those products was also provided, especially in terms of the implications of (i) calibration of predicted river volumes, and (ii) the latency of the input Earth observation imagery and attempts to map the floods’ maximum extents. An upcoming NASA DEVELOP project will also seek to strengthen the capability of CEPREDENAC and CRRH to apply HYDRAFloods for future extreme events. The application of the tools also provides a useful case study on capacity building, in terms of how Earth observation data and models can be replicated across regions.

Capacity building↗
NASA NTRS2011

Snow Cover, Snowmelt Timing and Stream Power in the Wind River Range, Wyoming

Earlier onset of springtime weather, including earlier snowmelt, has been documented in the western United States over at least the last 50 years. Because the majority (is greater than 70%) of the water supply in the western U.S. comes from snowmelt, analysis of the declining spring snowpack (and shrinking glaciers) has important implications for the management of streamflow. The amount of water in a snowpack influences stream discharge which can also influence erosion and sediment transport by changing stream power, or the rate at which a stream can do work, such as move sediment and erode the stream bed. The focus of this work is the Wind River Range (WRR) in west-central Wyoming. Ten years of Moderate-Resolution Imaging Spectroradiometer (MODIS) snow-cover, cloud-gap-filled (CGF) map products and 30 years of discharge and meteorological station data are studied. Streamflow data from streams in WRR drainage basins show lower annual discharge and earlier snowmelt in the decade of the 2000s than in the previous three decades, though no trend of either lower streamflow or earlier snowmelt was observed within the decade of the 2000s. Results show a statistically-significant trend at the 95% confidence level (or higher) of increasing weekly maximum air temperature (for three out of the five meteorological stations studied) in the decade of the 1970s, and also for the 40-year study period as a whole. The extent of snow-cover (percent of basin covered) derived from the lowest elevation zone (2500-3000 m) of the WRR, using MODIS CGF snow-cover maps, is strongly correlated with maximum monthly discharge on 30 April, where Spearman's Rank correlation, rs,=0.89 for the decade of the 2000s. We also investigated stream power for Bull Lake Creek above Bull Lake; and found a trend (significant at the 90% confidence level) toward reduced stream power from 1970 to 2009. Observed changes in streamflow and stream power may be related to increasing weekly maximum air temperature measured during the 40-year study period, possibly contributing to a reduction in snow cover. In addition, the strong relationship between percent of basin that was snow covered, and maximum monthly streamflow indicates that MODIS snow-cover maps are useful for predicting streamflow, and can be used to improve management of water resources in the drought-prone western United States.

MODIS↗
DOE OSTI2025

Improving the prediction of daily reservoir releases over the CONUS using conditioned LSTM

Reservoirs play a vital role in regulating streamflow timing and variability for hydroelectricity, flood control, water supply, irrigation, and recreation. Despite their importance, many reservoirs lack comprehensive operational guidelines, making their management complex due to conflicting operational objectives. Hence traditional policy-based reservoir models often fail to capture real-world conditions accurately and they depend on perfect streamflow predictions, which are not always available. In contrast, data-driven models like Long Short-Term Memory (LSTM) networks offer a robust alternative. This study introduces an approach that integrates reservoir characteristics—such as main use, climate, and maximum capacity—into the LSTM model to enhance reservoir release predictions. Using data from nearly 200 reservoirs in the contiguous United States (CONUS), our conditioned LSTM model (LSTM_cond) was compared with both the vanila LSTM and a traditional policy-based approach. Furthermore, our results show that while both LSTM_cond and LSTM perfoms better than the policy-based approach, LSTM_cond consistently outperforms LSTM for hydroelectric, water supply, irrigation, and recreation reservoirs. The KGE median values for LSTM_cond for out-sample reservoirs are 0.764, 0.565, 0.821, and 0.779, respectively, for the aforementioned reservoir types, which are consistently higher that the corresponding KGE values of 0.737, 0.413, 0.775, and 0.713 of LSTM, demonstrating its advantages in improving generalizability.

CONUS↗
NASA NTRSDate not supplied

LIS-Hydro: A Comprehensive Framework for Flood Prediction, Analysis, and Management

A subset of agencies from across the U.S. Federal enterprise are collaborating on the development of a fully-integrated hydro-modeling and streamflow prediction system (i.e., LIS-Hydro). The hydrological products and services will, for the very first time, establish a routinely available authoritative source of global water intelligence information supporting war-fighters, planners, and decision makers at all echelons and services of the U.S. military, Federal government, and intelligence community. Completion of the project and sustainment of the operational capability by Air Force Weather will establish a national asset to assist federal agencies in implementing government-wide strategies around water resources (U.S. Global Water Strategy, 2017). A summary of the interagency scientific collaboration in addressing some of the key gaps and needs identified during a 2017 joint NASA, USACE/ERDC, and U.S. Air Force co-sponsored workshop will be presented.

Jerry W Wegiel↗
DOE OSTI2021

Explore Spatio‐Temporal Learning of Large Sample Hydrology Using Graph Neural Networks

Abstract Streamflow forecasting over gauged and ungauged basins play a vital role in water resources planning, especially under the changing climate. Increased availability of large sample hydrology data sets, together with recent advances in deep learning techniques, has presented new opportunities to explore temporal and spatial patterns in hydrological signatures for improving streamflow forecasting. The purpose of this study is to adapt and benchmark several state‐of‐the‐art graph neural network (GNN) architectures, including ChebNet, Graph Convolutional Network (GCN), and GraphWaveNet, for end‐to‐end graph learning. We explicitly represent river basins as nodes in a graph, learn the spatiotemporal nodal dependencies, and then use the learned relations to predict streamflow simultaneously across all nodes in the graph. The efficacy of the developed GNN models is investigated using the Catchment Attributes and MEteorology for Large‐sample Studies (CAMELS) data set under two settings, fixed graph topology (transductive learning), and variable graph topology (inductive learning), with the latter applicable to prediction in ungauged basins (PUB). Results indicate that GNNs are generally robust and computationally efficient, achieving similar or better performance than a baseline model trained using the long short‐term memory (LSTM) network. Further analyses are conducted to interpret the graph learning process at the edge and node levels and to investigate the effect of different model configurations. We conclude that graph learning constitutes a viable machine learning‐based method for aggregating spatiotemporal information from a multitude of sources for streamflow forecasting

Sun, Alexander Y.↗
DOE OSTI2026

Hydrology in the Age of Artificial Intelligence: From Fragmentation to Coherent Terrestrial Hydrosphere Science

The rapid rise of machine learning (ML) in hydrology has prompted debate about the discipline's scientific relevance. While ML often outperforms traditional models in streamflow prediction, we argue that this reflects a deeper limitation: persistent fragmentation of hydrological science itself. Narrow focus on isolated components has hindered the development of coherent, scale‐relevant understanding of the integrated terrestrial hydrosphere. This is illustrated, for example, by widely divergent estimates of groundwater–streamflow interactions and of water balance‐implied ongoing storage changes. We argue that hydrology's future lies not in choosing between ML and physics, but in integrating data‐driven and process‐based approaches to advance consistent, realistic, and societally relevant understanding of the terrestrial hydrosphere and its multifaceted roles in the Earth System.

Painter, Scott L. [Oak Ridge National Laboratory (↗
DOE OSTI2020

Contaminant Transport Modeling for Technology Evaluation and Long-Term Monitoring in the Tims Branch Testbed, SC - 20343

Studies conducted at U.S. DOE sites have shown the presence of heavy metals, radionuclides, and volatile organic compounds in surface water, groundwater, and soil as a result of nuclear activity in the Cold War era. Since the 1990's, innovative cleanup methods have been implemented in the Tims Branch watershed at Savannah River Site (SRS) to limit the contaminant flux to the stream that have reduced the contaminant concentrations to acceptable regulatory levels in the dissolved phase. A tin-based treatment which effectively eliminated all local anthropogenic mercury inputs to this ecosystem resulted in a known step function addition of inert tin oxide particles which now serve as a potential tracer for sedimentation and particle transport processes in the stream. The long-term effectiveness of this and other remediation techniques and the potential for remobilization of adsorbed contaminant in sediment during extreme hydrologic conditions however remains unclear. It is therefore important to understand not only the fate and transport of dissolved contaminants, but also the movement of sediment and the relevant interactions with dissolved contaminant. To narrow this knowledge gap, a study is being conducted using the Tims Branch watershed as a stream-scale ecosystem test-bed to identify the primary transport processes of major contaminants of concern (such as mercury, nickel and uranium) with an emphasis on interactions with sediment transport. This involves the development of a fully distributed hydrologic watershed model of the Tims Branch watershed to predict streamflow under extreme weather conditions, as well as the development of a comprehensive contaminant transport model that can properly account for coupled contaminant and sediment transport. Review of relevant research reports and peer-reviewed journals revealed that aside from advection-dispersion transport of dissolved contaminants, adsorption and desorption with suspended solids and bed sediment also play an important role in the transport of those contaminants of concern. To develop the fully distributed hydrologic watershed model, the MIKE SHE 2-dimensional (2D) land surface/3D groundwater model that simulates surface/subsurface hydrologic processes (such as overland flow, evapotranspiration, and infiltration) was coupled with a 1D streamflow model that accounts for stream water hydraulics (such as hydraulic structures, cross-sections, and network). To model contaminant transport, the MIKE 11 streamflow component was coupled with the MIKE 11 AD module that simulates solute transport through advection and dispersion, and MIKE ECO Lab module that accounts for both sediment transport and interactions with dissolved contaminant. At this stage, the development and optimization of the fully distributed hydrologic model has been completed, achieving satisfactory statistical results between observed and predicted discharge as indicated by a root mean square error (RMSE) of 0.039 cms and a Nash-Sutcliffe efficiency coefficient (NSE) of 0.764. The ongoing development of the contaminant transport model has also yielded realistic results from preliminary tests. Results from this study are a key to evaluating the effectiveness of tin (II)-based mercury treatment of wetlands at the SRS site, and are also relevant to evaluating the potential of using this type of novel remediation technology in other mercury-contaminated stream systems. Knowledge acquired from this research will also support interpretation of historical data on the trends of contaminant concentration distribution in Tims Branch, particularly considering the effect of extreme hydrological events on the stream flow and pollutant transport. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2022

Scalable deep learning for watershed model calibration

Watershed models such as the Soil and Water Assessment Tool (SWAT) consist of high-dimensional physical and empirical parameters. These parameters often need to be estimated/calibrated through inverse modeling to produce reliable predictions on hydrological fluxes and states. Existing parameter estimation methods can be time consuming, inefficient, and computationally expensive for high-dimensional problems. In this paper, we present an accurate and robust method to calibrate the SWAT model (i.e., 20 parameters) using scalable deep learning (DL). We developed inverse models based on convolutional neural networks (CNN) to assimilate observed streamflow data and estimate the SWAT model parameters. Scalable hyperparameter tuning is performed using high-performance computing resources to identify the top 50 optimal neural network architectures. We used ensemble SWAT simulations to train, validate, and test the CNN models. We estimated the parameters of the SWAT model using observed streamflow data and assessed the impact of measurement errors on SWAT model calibration. We tested and validated the proposed scalable DL methodology on the American River Watershed, located in the Pacific Northwest-based Yakima River basin. Our results show that the CNN-based calibration is better than two popular parameter estimation methods (i.e., the generalized likelihood uncertainty estimation [GLUE] and the dynamically dimensioned search [DDS], which is a global optimization algorithm). For the set of parameters that are sensitive to the observations, our proposed method yields narrower ranges than the GLUE method but broader ranges than values produced using the DDS method within the sampling range even under high relative observational errors. The SWAT model calibration performance using the CNNs, GLUE, and DDS methods are compared using R 2 and a set of efficiency metrics, including Nash-Sutcliffe, logarithmic Nash-Sutcliffe, Kling-Gupta, modified Kling-Gupta, and non-parametric Kling-Gupta scores, computed on the observed and simulated watershed responses. The best CNN-based calibrated set has scores of 0.71, 0.75, 0.85, 0.85, 0.86, and 0.91. The best DDS-based calibrated set has scores of 0.62, 0.69, 0.8, 0.77, 0.79, and 0.82. The best GLUE-based calibrated set has scores of 0.56, 0.58, 0.71, 0.7, 0.71, and 0.8. The scores above show that the CNN-based calibration leads to more accurate low and high streamflow predictions than the GLUE and DDS sets. Our research demonstrates that the proposed method has high potential to improve our current practice in calibrating large-scale integrated hydrologic models.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

High‐Resolution National‐Scale Water Modeling Is Enhanced by Multiscale Differentiable Physics‐Informed Machine Learning

Abstract The National Water Model (NWM) is a key tool for flood forecasting, planning, and water management. Key challenges facing the NWM include calibration and parameter regionalization when confronted with big data. We present two novel versions of high‐resolution (∼37 km 2 ) differentiable models (a type of hybrid model): one with implicit, unit‐hydrograph‐style routing and another with explicit Muskingum‐Cunge routing in the river network. The former predicts streamflow at basin outlets whereas the latter presents a discretized product that seamlessly covers rivers in the conterminous United States (CONUS). Both versions use neural networks to provide a multiscale parameterization and process‐based equations to provide a structural backbone, which were trained simultaneously (“end‐to‐end”) on 2,807 basins across the CONUS and evaluated on 4,997 basins. Both versions show great potential to elevate future NWM performance for extensively calibrated as well as ungauged sites: the median daily Nash‐Sutcliffe efficiency of all 4,997 basins is improved to around 0.68 from 0.48 of NWM3.0. As they resolve spatial heterogeneity, both versions greatly improved simulations in the western CONUS and also in the Prairie Pothole Region, a long‐standing modeling challenge. The Muskingum‐Cunge version further improved performance for basins >10,000 km 2 . Overall, our results show how neural‐network‐based parameterizations can improve NWM performance for providing operational flood predictions while maintaining interpretability and multivariate outputs. The modeling system supports the Basic Model Interface (BMI), which allows seamless integration with the next‐generation NWM. We also provide a CONUS‐scale hydrologic data set for further evaluation and use.

Song, Yalan [Civil and Environmental Engineering T↗