NOAA National Water Center

Abstract We build three long short‐term memory (LSTM) daily streamflow prediction models (deep learning networks) for 531 basins across the contiguous United States (CONUS), and compare their performance: (1) a LSTM post‐processor trained on the United States National Water Model (NWM) outputs (LSTM_PP), (2) a LSTM post‐processor trained on the NWM outputs and atmospheric forcings (LSTM_PPA), and (3) a LSTM model trained only on atmospheric forcing (LSTM_A). We trained the LSTMs for the period 2004–2014 and evaluated on 1994–2002, and compared several performance metrics to the NWM reanalysis. Overall performance of the three LSTMs is similar, with median NSE scores of 0.73 (LSTM_PP), 0.75 (LSTM_PPA), and 0.74 (LSTM_A), and all three LSTMs outperform the NWM validation scores of 0.62. Additionally, LSTM_A outperforms LSTM_PP and LSTM_PPA in ungauged basins. While LSTM as a post‐processor improves NWM predictions substantially, we achieved comparable performance with the LSTM trained without the NWM outputs (LSTM_A). Finally, we performed a sensitivity analysis to diagnose the land surface component of the NWM as the source of mass bias error and the channel router as a source of simulation timing error. This indicates that the NWM channel routing scheme should be considered a priority for NWM improvement.

Abstract Numerical simulations of three of the most severe historical tropical cyclones to affect the Delaware River Basin (DRB) are used to evaluate a new numerical approach that is a candidate model for the inland‐coastal compound flood forecast. This study includes simulating interactions of tides/surges, freshwater streamflows, winds, and atmospheric pressure for the DRB. One‐way coupling between the hydrologic (National Water Model [NWM]) and the ocean/wave (ADvanced CIRCulation model/WAVEWATCH III [ADCIRC/WW3]) models for the Delaware river‐estuarine system is developed. The links between the coastal processes and the NWM are provided by two different hydraulic and hydrodynamic models: (i) a well‐calibrated public‐domain 1D hydraulic solver model (Hydrologic Engineering Center's River Analysis System [HEC‐RAS]) and (ii) 1D/2D open‐sourced hydrodynamic model (D‐Flow Flexible Mesh [D‐Flow FM]). First, the modeling system is tested to confirm model verification and stability when the system is forced with only tidal forcing. Then, the relative performance of each modeling approach (NWM/ D‐Flow FM /ADCIRC/WW3 and NWM/ HEC‐RAS /ADCIRC/WW3) is evaluated using observational data from Hurricanes Isabel (2003), Irene (2011), and Sandy (2012). Furthermore, the sensitivity of water level prediction to the streamflows, different wind products, and bed roughness are examined. Results show that the D‐Flow FM is generally accurate for water levels: the water levels near the peak of the storms have a skill ranging from 0.79 to 0.91 with a negligible phase error. Simulations show that water level predictions depend on an accurate representation of the wind conditions and bottom roughness. The work shows that hydrodynamic predictions, especially upstream, are highly dependent on the streamflow discharges.

Abstract The NOAA National Water Model (NWM) became operational in August 2016, producing the first ever real-time, distributed, continuous set of hydrologic forecasts over the continental United States (CONUS). This project uses integrated hydrometeorological assessment methods to investigate the utility of the NWM to predict catastrophic flooding associated with an extreme rainfall event that occurred in Ellicott City, Maryland, on 27–28 May 2018. Short-range forecasts (0–18-h lead time) from the NWM version 1.2 are explored, focusing on the quantitative precipitation forecast (QPF) forcing from the High-Resolution Rapid Refresh (HRRR) model and the corresponding NWM streamflow forecast. A comprehensive assessment of multiscale hydrometeorological processes are considered using a combination of object-based, grid-based, and hydrologic point-based verification. Results highlight the benefits and risks of using a distributed hydrologic modeling tool such as the NWM to connect operational CONUS-scale atmospheric forcings to local impact predictions. For the Ellicott City event, reasonably skillful QPF in several HRRR model forecast cycles produced NWM streamflow forecasts in the small Ellicott City basin that were suggestive of flash flood potential. In larger surrounding basins, the NWM streamflow response was more complex, and errors were found to be governed by both hydrologic process representation, as well as forcing errors. The integrated, hydrometeorological multiscale analysis method demonstrated here guides both research and ongoing model development efforts, along with providing user education and engagement to ultimately engender improved flash flood prediction.

Abstract The National Oceanic and Atmospheric Administration (NOAA) Geostationary Operational Environmental Satellite series R (GOES-R) will greatly expand the ability to observe the earth from geostationary orbit compared to the current-generation GOES, with more than 3 times as many spectral bands and a 75% reduction in footprint size. These enhanced capabilities are beneficial to rainfall rate estimation since they provide sensitivity to cloud-top properties such as phase and particle size that cannot be achieved using the limited channel selection of current GOES. The GOES-R rainfall rate algorithm, which is an infrared-based algorithm calibrated in real time against passive microwave rain rates, has been previously described in an algorithm theoretical basis document (ATBD); this paper describes modifications since the release of the ATBD, including a correction for evaporation of precipitation in dry regions and improved calibration updates. These improvements have been evaluated using a simplified version applicable to current-generation GOES to take advantage of the high-resolution ground validation data routinely available over the conterminous United States. Correcting for subcloud evaporation using relative humidity from a numerical model reduced false alarm rainfall by half and reduced the overall error by 35% for hourly accumulations validated against the National Centers for Environmental Prediction stage IV radar–gauge field; however, the number of missed events did increase slightly. Reducing the size of the calibration regions and increasing the training data requirements improved the consistency of the retrieved rates in time and space and reduced the overall error by an additional 4%.

Abstract The purpose of this study is to evaluate the components of the land surface water budget in the four land surface models (Noah, SAC‐Sacramento Soil Moisture Accounting Model, (VIC) Variable Infiltration Capacity Model, and Mosaic) applied in the newly implemented National Centers for Environmental Prediction (NCEP) operational and research versions of the North American Land Data Assimilation System version 2 (NLDAS‐2). This work focuses on monthly and annual components of the water budget over 12 National Weather Service (NWS) River Forecast Centers (RFCs). Monthly gridded FLUX Network (FLUXNET) evapotranspiration (ET) from the Max‐Planck Institute (MPI) of Germany, U.S. Geological Survey (USGS) total runoff ( Q ), changes in total water storage (d S /d t , derived as a residual by utilizing MPI ET and USGS Q in the water balance equation), and Gravity Recovery and Climate Experiment (GRACE) observed total water storage anomaly (TWSA) and change (TWSC) are used as reference data sets. Compared to these ET and Q benchmarks, Mosaic and SAC (Noah and VIC) in the operational NLDAS‐2 overestimate (underestimate) mean annual reference ET and underestimate (overestimate) mean annual reference Q . The multimodel ensemble mean (MME) is closer to the mean annual reference ET and Q . An anomaly correlation (AC) analysis shows good AC values for simulated monthly mean Q and d S /d t but significantly smaller AC values for simulated ET. Upgraded versions of the models utilized in the research side of NLDAS‐2 yield largely improved performance in the simulation of these mean annual and monthly water component diagnostics. These results demonstrate that the three intertwined efforts of improving (1) the scientific understanding of parameterization of land surface processes, (2) the spatial and temporal extent of systematic validation of land surface processes, and (3) the engineering‐oriented aspects such as parameter calibration and optimization are key to substantially improving product quality in various land data assimilation systems.

Permafrost thaw can cause an intensification of climate change through the release of carbon as greenhouse gases. While the effect of air temperature on permafrost thaw is well quantified, the effect of rainfall is highly variable and not well understood. Here, we provide a literature review of studies reporting on effects of rainfall on ground temperatures in permafrost environments and use a numerical model to explore the underlying physical mechanisms under different climatic conditions. Both the evaluated body of literature and the model simulations indicate that continental climates are likely to show a warming of the subsoil and hence increased end of season active layer thickness, while maritime climates tend to respond with a slight cooling effect. This suggests that dry regions with warm summers are prone to more rapid permafrost degradation under increased occurrences of heavy rainfall events in the future, which can potentially accelerate the permafrost carbon feedback.

Abstract Despite advancements in numerical modeling and the increasing prevalence of convection-allowing guidance, flash flood forecasting remains a substantial challenge. Accurate flash flood forecasts depend not only on accurate quantitative precipitation forecasts (QPFs), but also on an understanding of the corresponding hydrologic response. To advance forecast skill, innovative guidance products that blend meteorology and hydrology are needed, as well as a comprehensive verification dataset to identify areas in need of improvement. To address these challenges, in 2013 the Hydrometeorological Testbed at the Weather Prediction Center (HMT-WPC), partnering with the National Severe Storms Laboratory (NSSL) and the Earth System Research Laboratory (ESRL), developed and hosted the inaugural Flash Flood and Intense Rainfall (FFaIR) Experiment. In its first two years, the experiment has focused on ways to combine meteorological guidance with available hydrologic information. One example of this is the creation of neighborhood flash flood guidance (FFG) exceedance probabilities, which combine QPF information from convection-allowing ensembles with flash flood guidance; these were found to provide valuable information about the flash flood threat across the contiguous United States. Additionally, WPC has begun to address the challenge of flash flood verification by developing a verification database that incorporates observations from a variety of disparate sources in an attempt to build a comprehensive picture of flash flooding across the nation. While the development of this database represents an important step forward in the verification of flash flood forecasts, many of the other challenges identified during the experiment will require a long-term community effort in order to make notable advancements.

Abstract. A meteo-hydrological modelling system has been designed for the reconstruction of long time series of rainfall and river runoff events. The modelling chain consists of the mesoscale meteorological model of the Weather Research and Forecasting (WRF), the land surface model NOAH-MP and the hydrology–hydraulics model WRF-Hydro. Two 3-month periods are reconstructed for winter 2011 and autumn 2013, containing heavy rainfall and river flooding events. Several sensitivity tests were performed along with an assessment of which tunable parameters, numerical choices and forcing data most impacted on the modelling performance.The calibration of the experiments highlighted that the infiltration and aquifer coefficients should be considered as seasonally dependent.The WRF precipitation was validated by a comparison with rain gauges in the Ofanto basin. The WRF model was demonstrated to be sensitive to the initialization time and a spin-up of about 1.5 days was needed before the start of the major rainfall events in order to improve the accuracy of the reconstruction. However, this was not sufficient and an optimal interpolation method was developed to correct the precipitation simulation. It is based on an objective analysis (OA) and a least square (LS) melding scheme, collectively named OA+LS. We demonstrated that the OA+LS method is a powerful tool to reduce the precipitation uncertainties and produce a lower error precipitation reconstruction that itself generates a better river discharge time series. The validation of the river streamflow showed promising statistical indices.The final set-up of our meteo-hydrological modelling system was able to realistically reconstruct the local rainfall and the Ofanto hydrograph.

Abstract Reservoir management is a critical component of flood management, and information on reservoir inflows is particularly essential for reservoir managers to make real‐time decisions given that flood conditions change rapidly. This study's objective is to build real‐time data‐driven services that enable managers to rapidly estimate reservoir inflows from available data and models. We have tested the services using a case study of the Texas flooding events in the Lower Colorado River Basin in November 2014 and May 2015, which involved a sudden switch from drought to flooding. We have constructed two prediction models: a statistical model for flow prediction and a hybrid statistical and physics‐based model that estimates errors in the flow predictions from a physics‐based model. The study demonstrates that the statistical flow prediction model can be automated and provides acceptably accurate short‐term forecasts. However, for longer term prediction (2 h or more), the hybrid model fits the observations more closely than the purely statistical or physics‐based prediction models alone. Both the flow and hybrid prediction models have been published as Web services through Microsoft's Azure Machine Learning (Azure ML ) service and are accessible through a browser‐based Web application, enabling ease of use by both technical and nontechnical personnel.

Abstract This paper compares the annual and monthly components of the simulated energy budget from the North American Land Data Assimilation System phase 2 (NLDAS‐2) with reference products over the domains of the 12 River Forecast Centers (RFCs) of the continental United States (CONUS). The simulations are calculated from both operational and research versions of NLDAS‐2. The reference radiation components are obtained from the National Aeronautics and Space Administration Surface Radiation Budget product. The reference sensible and latent heat fluxes are obtained from a multitree ensemble method applied to gridded FLUXNET data from the Max Planck Institute, Germany. As these references are obtained from different data sources, they cannot fully close the energy budget, although the range of closure error is less than 15% for mean annual results. The analysis here demonstrates the usefulness of basin‐scale surface energy budget analysis for evaluating model skill and deficiencies. The operational (i.e., Noah, Mosaic, and VIC) and research (i.e., Noah‐I and VIC4.0.5) NLDAS‐2 land surface models exhibit similarities and differences in depicting basin‐averaged energy components. For example, the energy components of the five models have similar seasonal cycles, but with different magnitudes. Generally, Noah and VIC overestimate (underestimate) sensible (latent) heat flux over several RFCs of the eastern CONUS. In contrast, Mosaic underestimates (overestimates) sensible (latent) heat flux over almost all 12 RFCs. The research Noah‐I and VIC4.0.5 versions show moderate‐to‐large improvements (basin and model dependent) relative to their operational versions, which indicates likely pathways for future improvements in the operational NLDAS‐2 system.

Abstract Sea level rise and intense hurricane events make the East and Gulf Coasts of the United States increasingly vulnerable to flooding, which necessitates the development of computational models for accurate water level simulation in these areas to safeguard the coastal wellbeing. With this regard, a model framework for water level simulation over coastal transition zone during hurricane events is built in this study. The model takes advantage of the National Water Model’s strength in simulating rainfall–runoff process, and D‐Flow Flexible Mesh’s ability to support unstructured grid in hydrodynamic processes simulation with storm surges/tides information from the Advanced CIRCulation model. We apply the model on the Delaware Estuary to simulate extreme water level and to investigate the contribution of different physical components to it during Hurricane Isabel (2003). The model shows satisfactory performance with an average Willmott skill of 0.965. Model results suggest that storm surge is the most dominating component of extreme water level with an average contribution of 78.16%, second by the astronomical tide with 19.52%. While the contribution of rivers is mainly restricted to the upper part of the estuary upstream of Schuylkill River, local wind‐induced water level is more pronounced with values larger than 0.2 m over most part of the estuary.

Headwater wetlands provide many benefits such as water quality improvement, water storage, and providing habitat. These wetlands are characterized by water levels near the surface and respond rapidly to rainfall events. Driven by both groundwater and surface water inputs, water levels (WLs) can be above or below the ground at any given time depending on the season and climatic conditions. Therefore, WL predictions in headwater wetlands is a complex problem. In this study a hybrid modeling approach was developed for improved WL predictions in wetlands, by coupling a watershed model with artificial neural networks (ANNs). In this approach, baseflow and stormflow estimates from the watershed draining to a wetland are first estimated using an uncalibrated Soil and Water Assessment Tool (SWAT). These estimates are then combined with meteorological variables and are utilized as inputs to an ANN model for predicting daily WLs in wetlands. The hybrid model was used to successfully predict WLs in a headwater wetland in coastal Alabama, USA. The model was then used to predict the WLs at the study wetland from 1951 to 2005 to explore the possible teleconnections between the El Niño Southern Oscillation (ENSO) and WLs. Results show that both precipitation and the variations in WLs are partially affected by ENSO in the study area. A correlation analysis between seasonal precipitation and the Nino 3.4 Index suggests that winters are wetter during El Niño in Coastal Alabama. Analysis also revealed a significant negative correlation between WLs and the Nino 3.4 Index during the El Niño phase for spring. The findings of this study and the developed methodology/tools are useful to predict long-term WLs in wetlands and construct more accurate restoration plans under a variable climate.

Phytoplankton blooms are sporadic events in time and are isolated in space. This complex phenomenon is produced by a variety of both natural and anthropogenic causes. Early detection of this phenomenon, as well as the classification of a water body under conditions of bloom or non-bloom, remains an unresolved problem. This research proposes the use of Inherent Optical Properties (IOPs) in optically complex waters to detect the bloom or non-bloom state of the phytoplankton community. An IOP index is calculated from the absorption coefficients of the colored dissolved organic matter (CDOM), the phytoplankton ( phy ) and the detritus (d), using the wavelength (λ) 443 nm. The effectiveness of this index is tested in five bloom events in different places and with different characteristics from Mexican seas: 1. Dzilam (Caribbean Sea, Atlantic Ocean), a diatom bloom (Rhizosolenia hebetata); 2. Holbox (Caribbean Sea, Atlantic Ocean), a mixed bloom of dinoflagellates (Scrippsiella sp.) and diatoms (Chaetoceros sp.); 3. Campeche Bay in the Gulf of Mexico (Atlantic Ocean), a bloom of dinoflagellates (Karenia brevis); 4. Upper Gulf of California (UGC) (Pacific Ocean), a diatom bloom (Coscinodiscus and Pseudo-nitzschia) and 5. Todos Santos Bay, Ensenada (Pacific Ocean), a dinoflagellate bloom (Lingulodinium polyedrum). The diversity of sites show that the IOP index is a suitable method to determine the phytoplankton bloom conditions.

This study presents an innovative, automated deep learning-based technique for near real-time satellite monitoring of river ice conditions in northern watersheds of the United States and Canada. The method leverages high-resolution imagery from the VIIRS bands onboard the NOAA-20 and NPP satellites and employs the U-Net deep learning algorithm for the semantic segmentation of images under varying cloud and land surface conditions. The system autonomously generates detailed maps delineating classes such as water, land, vegetation, snow, river ice, cloud, and cloud shadow. The verification of system outputs was performed quantitatively by comparing with existing ice extent maps in the northeastern US and New Brunswick, Canada, yielding a Probability of Detection of 0.77 and a False Alarm rate of 0.12, suggesting commendable accuracy. Qualitative assessments were also conducted, corroborating the reliability of the system and underscoring its utility in monitoring hydraulic and hydrological processes across northern watersheds. The system’s proficiency in accurately capturing the phenology of river ice, particularly during onset and breakup times, testifies to its potential as a valuable tool in the realm of river ice monitoring.

While the National Weather Service and its River Forecast Centers and Weather Forecast Offices produce visuals, graphics, and outreach designed to support weather forecasts and warnings and inform decisions about natural resource management and emergency services, opportunities exist for risk communication scholarship to refine theory and promote best practices for communicating such information to the various stakeholders who need it. In September 2019, two focus groups were conducted with a sample (N = 14) of National Weather Service-Memphis’ core partners to gauge perceptions about how the Weather Forecast Office provides technical information about flood risk patterns, paying particular attention to evaluations of its Mississippi River Outlook product. Research findings demonstrated that core partners may benefit from targeting risk information depending on partners’ information needs and their technical knowledge/expertise. Similarly, the results suggested a need to offer context or interpretation for unique data points (e.g., gage tables, experimental forecasts, charts, and graphs) to successfully communicate important risk information and to clarify potential misunderstandings; this consideration was underscored by the finding that core partners tended to disseminate the Mississippi River Outlook product to others in the community (e.g., business owners; residents). These findings highlight the importance of audience testing in the development of risk communication and decision-making tools.

The availability of high-resolution digital elevation data (submeter resolution) from LiDAR has increased dramatically over the past few years. As a result, the efficient storage and transmission of those large data sets and their use for geomorphic feature extraction and hydrologic/environmental modeling are becoming a scientific challenge. This letter explores the use of multiresolution wavelet analysis for compression of LiDAR digital elevation data sets. The compression takes advantage of the fact that, in most landscapes, neighboring pixels are correlated and thus contain some redundant information. The space-frequency localization of the wavelet filters allows one to preserve detailed high-resolution features where needed while representing the rest of the landscape at lower resolution. We explore a lossy compression methodology based on biorthogonal wavelets and demonstrate that, by keeping only approximately 10% of the original information (data compression ratio ~94%), the reconstructed landscapes retain most of the information of relevance to geomorphologic applications, such as the ability to accurately extract channel networks for environmental flux routing, as well as to identify geomorphic process transition from the curvature-slope and slope-distance relationships.

Remote sensing based river extent mapping can be used for detecting inundation extents in retrospective and near-realtime applications. Synthetic aperture radar (SAR) has been proven useful for this purpose due to its high-spatial resolution, low atmospheric attenuation, and self-illumination but suffers from radiometric scattering and unwanted reflections from vegetation, terrain, and anthropogenic features. A novel four step procedure is proposed that augments a standard unsupervised classification of the SAR backscatter values with a terrain index to extract stages. The extracted stages are filtered, exploiting the dendritic nature of river networks, utilizing graph signal processing then remapping the filtered stages using the hydrologically relevant terrain model. Using a reference map derived from in-situ observations, the final inundation product significantly enhances the mapping skill when compared to the segmented SAR only method.

Abstract Operational forecast models require robust, computationally efficient, and reliable algorithms. We desire accurate forecasts within the limits of the uncertainties in channel geometry and roughness because the output from these algorithms leads to flood warnings and a variety of water management decisions. The current operational Water Model uses the Muskingum‐Cunge method, which does not account for key hydraulic conditions such as flow hysteresis and backwater effects, limiting its ability in situations with pronounced backwater effects. This situation most commonly occurs in low‐gradient rivers, near confluences and channel constrictions, coastal regions where the combined actions of tides, storm surges, and wind can cause adverse flow. These situations necessitate a more rigorous flow routing approach such as dynamic or diffusive wave approximation to simulate flow hydraulics accurately. Avoiding the dynamic wave routing due to its extreme computational cost, this work presents two diffusive wave approaches to simulate flow routing in a complex river network. This study reports a comparison of two different diffusive wave models that both use a finite difference solution solved using an implicit Crank–Nicolson (CN) scheme with second‐order accuracy in both time and space. The first model applies the CN scheme over three spatial nodes and is referred to as Crank–Nicolson over Space (CNS). The second model uses the CN scheme over three temporal nodes and is referred to as Crank–Nicolson over Time (CNT). Both models can properly account for complex cross‐section geometry and variable computational points spacing along the channel length. The models were tested in different watersheds representing a mixture of steep and flat topographies. Comparing model outputs against observations of discharges and water levels indicated that the models accurately predict the peak discharge, peak water level, and flooding duration. Both models are accurate and computationally stable over a broad range of hydraulic regimes. The CNS model is dependent on the Courant criteria, making it less computational efficient where short channel segments are present. The CNT model does not suffer from that constraint and is, thus, highly computationally efficient and could be more useful for operational forecast models.

Abstract Ice‐wedge polygon troughs play an important role in controlling the hydrology of low‐relief polygonal tundra regions. Lateral surface flow is confined to troughs only, but it is often neglected in model projections of permafrost thermal hydrology. Recent field and modeling studies have shown that, after rain events, increases in trough water levels are significantly more than the observed precipitation, highlighting the role of lateral surface flow in the polygonal tundra. Therefore, understanding how trough lateral surface flow can influence polygonal tundra thermal hydrology is important, especially under projected changes in temperatures and rainfall in the Arctic regions. Using an integrated cryohydrology model, this study presents plot‐scale end‐of‐century projections of ice‐wedge polygon water budget components and active layer thickness with and without trough lateral surface flow under the Representative Concentration Pathway 8.5 scenario. Trough lateral surface flow is incorporated through a newly developed empirical model, evaluated against field measurements. The numerical scenario that includes trough lateral surface flow simulates discharge (outflow from a polygon) and recharge (rain‐induced inflow to a polygon trough from upslope areas), while the scenario that does not include trough lateral surface flow ignores recharge. The results show considerable reduction (about 100–150%) in evapotranspiration and discharge in rainy years in the scenarios ignoring trough lateral surface flow, but less effect on soil water storage, in comparison with the scenario with trough lateral surface flow. In addition, the results demonstrate long‐term changes (~10–15 cm increase) in active layer thickness when trough lateral surface flow is modeled. This study highlights the importance of including lateral surface flow processes to better understand the long‐term thermal and hydrological changes in low‐relief polygonal tundra regions under a changing climate.

In 2015, the National Oceanic and Atmospheric Administration (NOAA) established the Office of Water Prediction (OWP) in the National Weather Service (NWS). The mission of the OWP is to “collaboratively research, develop and deliver timely and consistent, state-of-the-science national hydrologic analyses, forecast information, data, guidance, and decision-support services to inform essential emergency management and water resources decisions across all time scales” (NOAA 2021). The OWP team works to create a consistent and unified NWS hydrologic program with a goal of supporting a “water-ready nation” capable of addressing the nation’s water challenges. To support the mission of the OWP, the NWS also established the National Water Center (NWC), which first opened its doors in May 2015. The NWC, located on the campus of the University of Alabama in Tuscaloosa, was created to deliver a new generation of water information and services to the nation that will strengthen the nation’s water forecast capabilities for floods and droughts and improve preparedness for water-related disasters. In 2015, the NWC, in partnership with the Consortium of Universities for the Advancement of Hydrologic Science, Inc. (CUAHSI) created the Summer Innovators Program Summer Institute (SI) to engage the academic community in research to advance the mission of the National Water Center (CUAHSI 2018, 2021). The National Water Model (NWM) is the current basis for continental-scale water predictions by the NWS OWP. The NWM became operational on August 16, 2016 and delivers 18-h streamflow forecasts for about 2.7 million stream reaches in the continental United States (U.S.), as well as 10- and 30-day ensemble forecasts. The NWM is based on the National Center for Atmospheric Research’s WRF-Hydro framework (NCAR 2021). The SI is a seven-week residential program at the NWC, during which graduate students, postdoctoral scientists, faculty advisors, NWC staff, and other water professionals collaboratively conduct projects related to the NWM and water prediction. The fourth annual SI was held during June 12–July 28, 2018 and involved 23 graduate students from 18 unique universities. Cumulatively, 146 students from 77 unique universities participated in the SI during 2015–2019. In addition to the students, 64 individuals have participated in the five SIs as instructors and/or mentors, representing universities; federal, state, and local agencies; the private sector; and nonprofit organizations. A Featured Collection reporting papers from the 2019 SI is currently in review. The 2020 SI was canceled due to the pandemic and the 2021 SI is being planned as a virtual event. SI participants self-form teams of 2–5 participants, and each team conducts research on a project related to one of the predetermined SI themes that seek to advance some aspect of continental-scale water prediction. Themes for the 2018 SI were (1) Groundwater — Surface Water Interactions to help better understand the performance of the NWM in regions in which groundwater exchange is a dominant process at lower flows; (2) Hyper-Resolution Modeling to understand how models operating at large grid scales might perform for local-scale issues such as flash flooding; and (3) Computational Aspects of the NWM to potentially improve channel routing processes and incorporate nonstandard types of data. All 2018 student projects are summarized in CUAHSI Technical Report 15 (Aristizabal et al. 2018). Projects during the 2018 SI utilized NWM version 1.2 and WRF-Hydro version 5.0. This featured collection of five papers, authored by both students and faculty advisors, is representative of work performed at the 2018 SI. Previous SI Featured Collections in JAWRA are summarized by Bales (2019), Cohen et al. (2018), and Nelson (2017). Addressing the nation’s water challenges includes predicting streamflow during both floods and droughts. Although much of the recent focus has been on flood and flood inundation forecasting, NWM performance under low flow conditions has been an area of ongoing research at the SI. This research is motivated by the fact that the NWM represents the exchange of streamflow with groundwater as a one-way exchange of water from the aquifer to the stream. The aquifer is represented as a two-meter soil column, which potentially limits the amount of water that can be contributed to streamflow. Loss of the water from a stream to the underlying aquifer, which can occur when the unsaturated zone falls below the streambed level or when induced by nearby groundwater pumping, is not explicitly included in the NWM algorithms. The effects of these limitations, which are in place to allow continental-scale forecasts to be produced in a computationally efficient manner, certainly have an impact on forecasts. During the 2017 SI, Hansen et al. (2019) evaluated NWM performance during low flows and drought in the Colorado River Basin (CRB). Their results suggest that the NWM generally underestimates low flows in the CRB, although performance was better in the Upper CRB and at sites with higher precipitation, snow, and baseflows. They conclude that model performance is better in regions where the flows are driven primarily by snowmelt. Jachens et al. (2021) expanded on the work of Hansen et al. (2019) by evaluating NWM predictions in the Northern High Plains Aquifer. This system is well-studied (e.g., Peterson et al. 2016; McGuire 2017) and is regionally important for both agricultural irrigation and human water supply. Moreover, streams in the region are known to be both losing (streamflow recharges groundwater) and gaining (groundwater discharges to the surface), with some reaches varying by season (Jachens et al. 2021, figures 4 and 5). Jachens et al. (2021) used a unique multi-approach method for classifying stream reaches as losing or gaining. NWM results were then compared with observations for 42 floods and 116 low-flow periods. In general, flood peaks were overpredicted in magnitude and low-flow predictions were systematically greater than observations; observed zero-flow days were not replicated by the NWM, due in part to flow-routing constraints. Karki et al. (2021) also focused on the effects of NWM simplifications to groundwater — surface water exchange, focusing on the baseflow simulation performance in the Northern High Plains Aquifer. A seven-year NWM retrospective (2003–2009) with no streamflow assimilation was used in the evaluation. In addition, two alternative formulations for the WRF-Hydro groundwater algorithm were tested. In general, baseflow was more closely reproduced by the NWM in basins characterized by clay soils with reduced infiltration compared to basins with sandy soils and higher infiltration rates. Because of the absence of a lag between infiltration and discharge of excess groundwater to the stream, water that entered the groundwater reservoir generally was immediately discharged to the stream, resulting in an overestimation of stormflow and an underestimation of baseflow. Combined, the work by Hansen et al. (2019), Jachens et al. (2021), and Karki et al. (2021) identified causes for specific deficiencies in NWM predictions and offered potential alternative approaches to existing groundwater algorithms. Moreover, this work clearly demonstrates the contributions of the SI to the water prediction enterprise in the U.S. To preserve computational speed, the NWM uses the relatively simplistic Muskingum-Cunge approach with an idealized channel cross section for routing flows through the stream network (NCAR 2021). 2018 SI participants explored two aspects of this assumption along with potential alternative approaches. The NWM v1.2 uses a trapezoidal channel cross section for the entire model domain, with channel depth and top width varying spatially. The Muskingum–Cunge flow routing approach depends on stage to estimate flow routing parameters, so an error in stage resulting from the cross-sectional channel representation affects forecast discharge and timing of forecast flood peaks. In order to evaluate the effects of that assumption and to explore alternatives, the one-dimensional flow routing model, HEC-RAS (USACE 2016), which allows ready substitution of channel geometries, was used at three sites under varying flows. Four different cross-sectional configurations were explored at each site: NWM (trapezoidal), surveyed, and two simplified alternatives. The admittedly limited study results suggest that simple geometric cross sections that could be obtained by volunteers through nonsurvey grade approaches can reduce NWM errors in stage prediction, especially peak flow predictions. Meselhe et al. (2021) explored another approach for improving flow routing in the NWM. For continental scale applications, the ambient channel conditions (bed slope, roughness, flow impedance features, etc. presence) are strongly variable in time and space. To accommodate such variability, they hypothesized that a heterogeneous routing strategy could be deployed without degrading computational stability and without greatly increasing computational time. The utilization of computationally demanding approaches, such as the dynamic wave approach, should be limited to conditions such as low-gradient environments or when backwater effects are prominent. Otherwise, efficient and robust methods, such as the currently used Muskingum–Cunge method, are appropriate. Meselhe et al. (2021) conducted a set of numerical experiments and provided guidance on when flow conditions could trigger a transition from a simplified approach to dynamic wave routing within the NWM. The final paper in this collection by Wu et al. (2021) is somewhat unique from papers in previous featured collections but could supplement the work of Brackins et al. (2021) as relates to channel geometry data collected by volunteers. Recognizing that (1) data collected by agencies are limited (2) these data could be supplemented by data collected by volunteers (Kampf et al. 2018), and (3) data collected by volunteers could have, or be perceived to have, less reliability than agency-collected data, Wu et al. explored approaches for determining the uncertainty in these types of data. They demonstrated that a decision tree approach methodology with specific rules could be used to identify outliers in volunteer data. The approach was evaluated using concurrently collected pressure transducer stage data and stage data obtained by volunteers using visual observations of the stage. This featured collection is the fourth such collection of papers published in JAWRA from research conducted during the SI. In addition to this collection, Nelson (2017) summarized work on the Flood Interoperability Experiment from the first SI; Cohen et al. (2018) provided an introduction to the second JAWRA featured collection describing work from the second SI on the NWM, and Bales (2019) introduced the third collection. Research represented in these past collections and current collection is helping advance the science of water prediction in the U.S.