Operationalizing Continental-Domain Hydrologic Models: What can we learn?

Understanding the interaction between terrestrial microclimate, hydrology and \necology is a key to determining the effect of land-use and climate change on \nhydrological systems. Traditionally, the hydrological impacts of climate change have \nbeen based on driving hydrological models with the output of regional climate \nmodels. These climate models often operate at spatial and temporal scales that are \nmuch larger than the scales required to analyse the effects on the hydrological \nsystem. This is in part because of computational limitations and in part because of \nthe physics of the regional models do not justify much higher resolution. \nFurthermore there is an inherent contradiction in this approach since these climate \nmodels include their own hydrological model component. Similarly in analysing the \nhydrological effects of land-use change the feedback to the meteorological system is \noften neglected.\n\nTo address these issues a dynamically coupled hydrological and meteorological model \nsystem has been developed for evaluating interactions at hydrological (catchment) \nscales. Such a coupled system provides a unique framework for investigation of land \nsurface - atmosphere interactions at hydrological scales. A comprehensive \nhydrological modelling system describing the terrestrial component of the \nhydrological cycle MIKE SHE has been modified to allow coupling to a local scale \nmeteorological model. The coupling exploits the European Open Modelling Interface \nand Environment (OpenMI) Open MI to link the two models systems. As simulations can \nbe run with and without coupling to the meteorological model, it is possible to \nevaluate the impact of feedbacks between the two systems on hydrological \npredictions. The uncoupled system has been validated against remote sensing and \neddy correlation measurements at the field and landscape scale describing the \nhydrological and energy fluxes on the land-surface. The coupled system is then \nvalidated against field data describing both the atmosphere and hydrological system. \nFinally, a sensitivity analysis is carried out to examine the sensitivity of \nhydrological predictions to atmospheric feedbacks to identify where feedback will \nsignificantly affect the simulated impact.

The National Water Model (NWM), a configuration of the Weather Research and Forecasting Hydrological model, operates as the United States’ hydrological model. The NWM predicts streamflow at more than 2.7 million river reaches; and is a subject of growing attention in the hydrological modeling community. Large‐scale computationally distributed models such as the NWM, often require technical knowledge of, and access to, cluster‐based computing environments for model compilation and simulation. User‐friendly tools capable of setting up and running such models to adjust and explore their parameter space generally do not exist. Here we present the Dockerized Job Scheduler (DJS) a Python library that takes a service approach to modeling. The library is capable of (1) generating varied parameter sets and (2) orchestrating concurrent NWM simulations via Docker. DJS is designed to automate the deployment of varied parameter simulations and lower the model usage entrance barrier. In this paper, we use a case study to demonstrate its installation and usage.

A key element in hydraulic and hydrologic modeling is the specification of representative channel geometry. For continental‐scale modeling, the large amount of high‐resolution data required, as well as the considerable computational effort needed to incorporate such data, has led to simplifying assumptions such as rectangular or trapezoidal channels for river reaches. The National Water Model (NWM) uses a trapezoidal channel representation for 2.7 million river reaches to forecast water discharge for the entire continental United States. This has created uncertainties in when to initiate hydraulic predictions. The aim of this study is to: (1) evaluate the impact of simplified (NWM trapezoidal) channel geometry representation on hydrological model predictions and (2) suggest an improved representation of channel geometry while maintaining parsimony in model input and runtime. The Hydrologic Engineering Center’s River Analysis System model was used to simulate hydraulic dynamics in three study sites under varying streamflow conditions (including flooding) with four different geometry representations at each site: NWM (trapezoidal), surveyed, and two proposed generalized geometries. Statistical analyses show that more realistic channel geometry improves simulated Muskingum‐Cunge routing parameters and stage/discharge predictions, indicating potential for geometric improvements to enhance hydrological models like the NWM.

Abstract. Large-scale hydrologic models are increasingly being developed for operational use in the forecasting and planning of water resources. However, the predictive strength of such models depends on how well they resolve various functions of catchment hydrology, which are influenced by gradients in climate, topography, soils, and land use. Most assessments of hydrologic model uncertainty have been limited to traditional statistical methods. Here, we present a proof-of-concept approach that uses interpretable machine learning techniques to provide post hoc assessment of model sensitivity and process deficiency in hydrologic models. We train a random forest model to predict the Kling–Gupta efficiency (KGE) of National Water Model (NWM) and National Hydrologic Model (NHM) streamflow predictions for 4383 stream gauges in the conterminous United States. Thereafter, we explain the local and global controls that 48 catchment attributes exert on KGE prediction using interpretable Shapley values. Overall, we find that soil water content is the most impactful feature controlling successful model performance, suggesting that soil water storage is difficult for hydrologic models to resolve, particularly for arid locations. We identify nonlinear thresholds beyond which predictive performance decreases for NWM and NHM. For example, soil water content less than 210 mm, precipitation less than 900 mm yr−1, road density greater than 5 km km−2, and lake area percent greater than 10 % contributed to lower KGE values. These results suggest that improvements in how these influential processes are represented could result in the largest increases in NWM and NHM predictive performance. This study demonstrates the utility of interrogating process-based models using data-driven techniques, which has broad applicability and potential for improving the next generation of large-scale hydrologic models.

High‐resolution, coupled, process‐based hydrology models, in which subsurface, land‐surface, and energy budget processes are represented, have been applied at the basin‐scale to ask a wide range of water science questions. Recently, these models have been developed at continental scales with applications in operational flood forecasting, hydrologic prediction, and process representation. As use of large‐scale model configurations increases, it is exceedingly important to have a common method for performance evaluation and validation, particularly given challenges associated with accurately representing large domains. Here, we present phase 1 of a comparison project for continental‐scale, high‐resolution, processed‐based hydrologic models entitled the Continental Hydrologic Intercomparison Project (CHIP). The first phase of CHIP is based on past Earth System Model intercomparisons and comprised of a two‐model proof of concept comparing the ParFlow‐CONUS hydrologic model, version 1.0 and a NOAA US National Water Model configuration of WRF‐Hydro, version 1.2. The objectives of CHIP phase 1 are: (a) describe model physics and components, (b) design an experiment to ensure a fair comparison, and (b) assess simulated streamflow with observations to better understand model bias. To our knowledge, this is the first comparison of continental‐scale, high‐resolution, physics‐based models which incorporate lateral subsurface flow. This model intercomparison is an initial step toward a continued effort to unravel process, parameter, and formulation differences in current large‐scale hydrologic models and to engage the hydrology community in improving hydrology model configuration and process representation.

Improved runoff forecasting performance through error predictions using a deep-learning approach

Seasonal hydrological forecasts are of critical importance for the effective management of water resources, particularly in complex and vulnerable basins such as the Mediterranean area which have both a deficitary water regimen and a high anthropogenic pressure. Nevertheless, inaccuracies in meteorological inputs can propagate through hydrological models, amplifying uncertainties in flow predictions. It is imperative to rectify these forecasts, whether meteorological or hydrological, to enhance prediction reliability and provide robust data for informed decision-making. This study proposes an advanced framework for seasonal hydrological forecasting that integrates raw and corrected weather forecasts with distributed hydrological modeling and sophisticated post-processing techniques to enhance flow prediction accuracy in the Júcar River in Spain as a representative case study of the Mediterranean area.The catchment hydrological model was implemented using the model TETIS v9.1 wich was calibrated and validated using observed records from 1981 to 2019 at seven control points. To do this, a split sample test was conducted using the period 2009–2019 for calibration, and the rest of the time series for validation (1981–2008). The process involved refining parameter maps to ensure good or acceptable performance at all control points. Meteorological data were sourced from the W5E5 dataset, downscaled to a 0.09° resolution using ERA5-Land. This improved the spatial and temporal resolution of the hydrological model. Once we have established an acceptable hydrological model,, the seasonal forecast hindcasts were evaluated using meteorological inputs from global forecasting systems, including ECMWF-SEAS5, CMCC_SPSv35, DWD_GCFS21, and MeteoFrance System8. To address uncertainties in the meteorological forecasts and their propagation to hydrological outputs, two complementary correction strategies were implemented. First, artificial intelligence(fuzzy logic) was applied to correct meteorological inputs before integration into the hydrological model, assuming errors originate solely from meteorological data and treating the hydrological model as a “perfect” simulator. Second, a hydrological error model was developed to identify and adjust discrepancies between simulated and observed flows, addressing systematic biases and errors in the hydrological simulation.The results demonstrated that forecasts based on corrected meteorological inputs exhibited significant accuracy improvements compared to those using unprocessed inputs. The hydrological error model further enhanced prediction reliability by mitigating systematic biases. These findings underscore the effectiveness of combining meteorological forecasts with AI-driven corrections to address uncertainties, thereby improving the robustness of seasonal hydrological predictions. This study highlights the potential for integrating advanced correction techniques into seasonal hydrological forecasting frameworks, offering a replicable methodology for other basins with similar complexities. The proposed framework enhances both the reliability and applicability of forecasts, ensuring their relevance for effective decision-making in complex hydrological systems as the Mediterranean area. The improved accuracy of these forecasts provides a sound scientific support for adaptive water resource management, particularly in the face of increasing climatic variability and environmental changes.Acknowledgments: This study was funded by the Colombian Ministry of Science, Technology, and Innovation (MINCIENCIAS) through the Call for Doctorates Abroad 885-2; by the Valencian Regional Government through the WATER4CAST 2.0 (CIPROM/2023/5) research project; and Spanish Ministry of Science and Innovation through the research project TETISPREDICT (PID2022-141631OB-I00). 

Regulatory practitioners use hydroclimatic data to provide context to observations typically collected through field site visits and aerial imagery analysis. In the absence of site‐specific data, regulatory practitioners must use proxy hydroclimatic data and models to assess a stream's hydroclimatology. One intent of current‐generation continental‐scale hydrologic models is to provide such hydrologic context to ungaged watersheds. In this study, the ability of two state‐of‐the‐art, operational, continental‐scale hydrologic modeling frameworks, the National Water Model and the Group on Earth Observation Global Water Sustainability (GEOGloWS) European Centre for Medium‐Range Weather Forecasts (ECMWF) Streamflow Model, to produce daily streamflow percentiles and categorical estimates of the streamflow normalcy was examined. The modeled streamflow percentiles were compared to observed daily streamflow percentiles at four United States Geological Survey stream gages. The model's performance was then compared to a baseline assessment methodology, the Antecedent Precipitation Tool. Results indicated that, when compared to baseline assessment techniques, the accuracy of the National Water Model (NWM) or GEOGloWS ECMWF Streamflow Model was greater than the accuracy of the baseline assessment methodology at four stream gage locations. The NWM performed best at three of the four gages. This work highlighted a novel application of current‐generation continental‐scale hydrologic models.

A well-configured, verified hydrological operational forecasting system is an invaluable tool for hydrological forecasting and warning services. Target users of such a service can be water managers, power generation planning, navigation, civil protection, and the public, whose priority is to obtain the best possible forecast for their area of interest. This was one of the reasons why SHMU proceeded to a more complex assessment of hydrological forecasts.The main objective of the assessment was to analyse the uncertainties that significantly affect the quality of the forecast itself. The evaluation was conducted for 47 selected water-gauging profiles. It showed that in Slovak physical-geographical conditions, the precipitation data (measured and predicted) and the configuration of the hydrological model are the most significant sources of uncertainties. Forcing data for hydrological forecasts come from the deterministic ALADIN/SHMU model with 4.5 km resolution, generated 4 times per day with hourly time-step and lead time 69 hours. The HBV model efficiency was tested on a total of 138 forecast profiles during the period 08/2016 – 12/2020. The input data used was precipitation from a combined radar product (qPrec) with 1 km resolution, which also enters the models in operation. Model performance was expressed by NSE and KGE statistics as well as visual inspection of the hydrographs. It showed very good model simulation results for most of the catchments. A weak point was the simulation and forecast of peak flows, which the model underestimated in many cases. It was therefore necessary to proceed to a more detailed analysis of the precipitation input, both measured and predicted, in relation to the predicted flows. Monthly precipitation totals and for selected catchments also daily ones were analysed and feedback was sent to the precipitation data providers for hydrological models. Monthly precipitation totals were compared with totals obtained from spatial interpolation of 568 rain gauge stations in a GIS environment. From these comparisons, systematic errors are visible as well as their temporal evolution for the specific catchments. Such analyses are not a routine part of hydrological forecasting systems.The work also includes a quantification of the uncertainty of the meteorological forecast and hydrological model separately expressed for different forecast lead times for a specific forecast profile. In the future, we would like to apply the methodology used for other profiles in order to detect possible systematic errors affecting the quality of the hydrological forecast.This work was supported by the Slovak Research and Development Agency under the Contract no. APVV-19-0340.

Reconstructed historical streamflow time series can supplement limited streamflow gauge observations. However, there are common challenges of typical modeling approaches: process‐based hydrologic models can be data/computation‐intensive, and statistics‐based models can be region/stream‐specific. Here we present a nationally scalable modeling framework integrating the simulated runoff from the Variable Infiltration Capacity (VIC) model with the Routing Application for Parallel computatIon of Discharge (RAPID) routing model leveraging high‐performance computing. We demonstrate an efficient method of assimilating streamflow at US Geological Survey (USGS) streamflow monitoring sites using a simple hierarchical approach in the VIC‐RAPID framework. The result is a reconstructed 36‐year (1980–2015) daily and monthly streamflow dataset (Dayflow) at ∼2.7 million NHDPlusV2 stream reaches in the conterminous US (CONUS). We perform a comprehensive evaluation at 7,526 USGS sites and characterize their error statistics. The results demonstrate that 49% of the USGS sites demonstrate Kling–Gupta Efficiency (KGE) > 0.5 and 58% of the sites show percentage bias within ±20% for the daily naturalized streamflow. Streamflow data assimilation across CONUS shows an overall improvement over naturalized streamflow, notably in the western semiarid‐to‐arid regions. Comparison to other national and global streamflow reanalysis datasets such as the National Water Model and Global Reach‐scale A priori Discharge Estimates for SWOT demonstrates improved KGE, reduced bias, and directions for Dayflow improvements. Investigations of error statistics with key hydrologic, hydroclimatic, and geomorphologic basin characteristics reveal region‐specific patterns which may help improve future framework applications. Overall, Dayflow may enable a better understanding of hydrologic conditions in a changing environment, especially in locations currently not represented by streamflow monitoring networks.

Hydrological modeling and the hydrological response to land-use/land-cover changes induced by human activities have gained enormous research interest over the last few decades. The study presented here analyzes the spatial and qualitative changes in the rainfall–runoff that have resulted from the land-cover changes between 1985–2014 in the Godavari River Basin using the Hydrologic Engineering Centre-Hydrologic Modeling System(HEC-HMS) model and remote sensing—GIS (geographic information system) techniques. The purpose of this paper is to analyze the dynamics of land-use/land-cover (LULC) changes for the years 1985, 1995, 2005, and 2014 for the Godavari Basin. The findings reveal an increase of 0.64% of built-up land, a decrease of 0.92% in shrubland, and an increase of 0.56% in waterbodies between 1985–2014. The LULC change detection results between the years 1985–2014 indicated a drastic change in the cropland, forest, built-up land, and water bodies among all of the other classes. The urbanization and agricultural activities are the major reasons for the increase of cropland, built-up land, and water bodies, at the expense of decreases in shrubland and forest. The study had an overall classification accuracy of 92% and an overall Kappa coefficient of 0.9. The HEC-HMS model is used to simulate the hydrology of the Godavari Basin. The analyses carried out were mainly focussed on the impact of LULC changes on the streamflow pattern. The surface runoff was simulated for the year 2014 to quantify the changes that have taken place due to changes in LULC. The observed and the simulated peak streamflow was found to be the same i.e., 56,780 m3/s on 9 September 2014. In the validation part, the linear regression method was used to correlate the observed and simulated streamflow data at the prominent gauge station of the Badrachalam outlet for the Godavari River Basin and give a correlation coefficient value of 0.83. It was found that the HEC-HMS model is compatible and works better for the rainfall–runoff modeling, as it takes into account the various parameters that are influencing the process. The hydrological modeling that was carried out using the HEC-HMS model has brought out the significant impact of LULCC on rainfall–runoff at the Pranhita sub-basinscale, indicating the model’s ability to successfully accommodate all of the environmental and landscape variables. The study indicates that deforestation at the cost of urbanization and cropland expansions leads to decreases in the overall evapotranspiration (ET) and infiltration, with an increase in runoff. The results of the study show that the integration of remote sensing, GIS, and the hydrological model (HEC-HMS) can solve hydrological problems in a river basin.

This technical note describes recent efforts to integrate machine learning (ML) models, specifically long short‐term memory (LSTM) networks and differentiable parameter learning conceptual hydrological models (δ conceptual models), into the next‐generation water resources modeling framework (Nextgen) to enhance future versions of the U.S. National Water Model (NWM). We address three specific methodology gaps of this new modeling framework: (1) assess model performance across many ungauged catchments, (2) diagnostic‐based model selection, and (3) regionalization based on catchment attributes. We demonstrate that an LSTM trained on CAMELS catchments can make large‐scale predictions with Nextgen across the New England region and match the average flow duration curve observed by stream gauges for streamflow with low exceedance probability (high flows), but diverges from the mean in high exceedance probability (low flows). We demonstrate improvements in peak flow predictions when using δ conceptual model, but results also suggest that performance increases may come at a cost of accurately representing hydrologic states within the conceptual model. We propose a novel approach using ML to predict the most performant mosaic modeling approach and demonstrate improved distributions of efficiency scores throughout the large sample of basins. Our findings advocate for the future development of ML capabilities within Nextgen for advancing operational hydrological modeling.

The resolution of digital elevation model (DEM) used for watershed delineation and data extracted from the delineated watershed relevant to hydrography plays an important role in the calibration of hydrological models. The effect of DEM resolution on the drainage length, watershed area and average slope are studied in many earlier works. DEM resolution can have impact on hydrologic modeling parameters, water quality simulation results. The impacts of DEM resolution on results of the watershed response and water quality modeling are studied using HSPF (Hydrologic Simulation Program Fortran) model applied to a watershed in Kentucky in this study. Three major streams in this watershed are impaired due to pathogen pollution according to the current state of Kentucky water quality standard set for recreation. The effects of DEM resolution on water quality modeling and calibration for fecal coliform bacteria are studied using a HSPF model. Results from this study indicate that the DEM resolution affects hydrography parameters of the delineated watershed, influence the calibration process, and the non-point and point source loadings to the streams in the watershed.

The large-scale application of Water Management Models (WMMs) is limited by fragmented, non-standardized water management data, and a lack of integration with hydrologic models. To address this gap, we developed automated, reproducible workflows that extend the United States’ reference hydrofabric to explicitly include water management infrastructure such as reservoirs and diversion points while populating a relational database containing operational and administrative records. Piloted across the entire Upper Colorado River Basin (UCRB), this work integrates 1,318 diversions and 37 major reservoirs using persistent identifiers to ensure interoperability with existing national hydrologic models including the National Hydrologic Model and the National Water Model. The resulting contributions are, first, a basin-scale Extended Hydrofabric data product for the UCRB, directly usable for constructing WMMs and coupling with hydrologic models, and second, designs and workflows that enable generation of Extended Hydrofabric for other managed basins and provide a foundation for future conterminous United States-scale applications. • Scalable designs to extend the reference hydrofabric to include reservoirs, diversions, and water rights data. • Automated, reproducible Python workflows for water management data integration. • Model-ready data product for the Upper Colorado River Basin with 1,318 diversions and 37 reservoirs.

The integration of remote sensing observations into hydrological modeling frameworks presents a significant opportunity for improving spatial and temporal predictive capabilities across continental domains. This research introduces a novel hybrid distributed hydrological model that addresses key challenges in computational efficiency, by using a GPU-enabled computational infrastructure, and in predictive accuracy by assimilating multi-source remote sensing datasets, specifically satellite-based soil moisture and evapotranspiration, at a high spatial resolution (1km×1km) and temporal scale (hourly). The model addresses critical challenges in regional hydrological forecasting by leveraging advanced data assimilation techniques and machine learning methodologies.The proposed hybrid modeling framework synthesizes physically-based distributed hydrologic modeling principles with data-driven machine learning approaches, facilitating a more comprehensive representation of land surface hydrological processes. A key innovation is the GPU-enabled cell-to-cell routing algorithm, which enables fast and efficient computational processing of complex hydrological connectivity and water movement across large spatial domains. By integrating remote sensing observations, the methodology enables enhanced initial condition specification and improved parameter estimation, particularly in regions characterized by sparse ground-based measurement networks.Preliminary analytical results demonstrate significant improvements in model performance, particularly in capturing spatial and temporal variability of hydrological states and fluxes. The approach substantively advances current methodological capabilities in hydrological forecasting, offering a promising framework for developping enriched tensorial numerical solvers, addressing complex hydroclimatic prediction challenges in data-limited environments.