WRF-Hydro and the National Water Model


Over the past three years, a team of RAL scientists and engineers have developed and transitioned the community WRF-Hydro modeling system into the National Water Model (NWM), the first operational, high-resolution, physics-based hydrologic prediction model ever implemented across the continental U.S.  Developed in close collaboration with the National Weather Service (NWS) Office of Water Prediction, the USGS, and a number of university partners, the RAL team delivered the first version of the model in only one year (two years ahead of schedule) on a modest budget. The model is now being used, to solve a complex problem—tracking water flow across the country to aid local communities and emergency managers in responding to water-related threats.  It is also providing consistent, reliable, high-resolution data to decision makers across the country, helping them address increasingly complex societal issues related to flooding, drought, water availability and water quality. In addition to providing significant benefit to the operational community, the NWM is proving to be a valuable tool in the conduct of research and as a mechanism for moving research to operations. As an open source, open platform model, the NWM will continue to evolve to meet the nation’s complex needs for water resource planning and management information while also providing a seamless pathway for academic researchers to innovate new improvements in water prediction.


The RAL team delivered version 1.0 of the National Water Model ahead of schedule and on budget in August 2016.   This first iteration of the model met NOAA’s Initial Operating Capability requirements, providing high-resolution (1 km) characterizations of evapotranspiration, snowpack, and surface energy and radiation fluxes as well as "hyper-resolution" (250 m) depictions of soil moisture, shallow groundwater and surface inundation.  Most importantly, the model was also able to provide continuous streamflow analyses and forecasts from 0-30 days for 2.67 million channel reaches across the U.S. It is designed to run in four configurations: analysis and assimilation mode, assimilating real-time streamflow data; hourly short-range (0-18 hrs) deterministic forecasts; daily medium-range 10-day deterministic forecasts; and long-range 30-day ensemble forecasts. Now in operations, the model generates a host of streamflow, flood inundation, soil moisture, snowpack and evapotranspiration forecast products.

Subsequent versions delivered in 2017 have provided important new capabilities and services and marked improvements in forecast skill. Versions 1.1 and 1.2 of the model have improved its hydrological analysis and prediction skill, created an “on-demand” event-based modeling capabilities, expanded the model’s operational prediction domain beyond the CONUS (to Hawaii and Puerto Rico), and added a new mirror development capability for code development, optimization, debugging, training and re-forecast development.  In parallel, efforts to further develop the basic infrastructure for the research WRF-Hydro modeling system are continuing with the addition of new Earth system process models (e.g. coastal dynamics, water quality, deeper groundwater, etc.).  And finally, a strong education and outreach component has been launched to engage the broader NWC, NOAA, federal and academic communities in both contributing to the development of the model and using it in their own research.

Improvements in Scientific/Technical Capabilities: 

The RAL development team has essentially constructed a new NWM WRF-Hydro “Ecosystem” that provides an operational modeling capability to the National Weather Service and an important testbed for continuous model development by the research community.  Throughout this project, the team has made substantial improvements to the core WRF-Hydro model physics modules (i.e., NoahMP land surface model, terrain, channel and reservoir routing modules, and streamflow data assimilation) and added a number of new features to expand the model’s ability to provide enhanced and forecasts at the spatial and temporal scales required by NOAA and other operational prediction entities. Transforming a research model into an operational model for use by the National Weather Service required the development of new algorithms, systems to ingest and assimilate new operational data streams, advances in code parallelization to significantly improve computational efficiency and forecast timeliness, and substantial re-engineering and hardening of code to enable the model run on NCEP’s supercomputers.  It also required the development of a new research to operations environment that would allow the research and operational communities to work together to continuously improve a model the national model.   Several scientific and technical advances are highlighted here:

  • Core-model physics were improved to better characterize changes in snow volume and meltout timing, channel flow processes, as well as soil infiltration rates and lateral transport to improve depiction of soil moisture.
  • An improved geospatial and hydrologic fabric was created to allow the NWM to represent in unprecedented spatial detail (for an operational, national model) hydrologic information on a flexible framework that is now widely used across agencies and disciplines (e.g., hydrology, water quality, emergency response).
  • A new meteorological forcing engine was created to downscale, bias-correct, and blend atmospheric forcings from 5 different NWP models and the national operational radar system in real-time and for retrospective simulation and re-analysis.
  • An automated time-varying bias-correction methodology was implemented as a new innovation to the system’s operational streamflow data assimilation.
  • A new capability to rapidly implement “hyper-resolution” model domains during active hydrologic events like Hurricane Harvey was developed and demonstrated.
  • A new community model calibration toolkit was developed to efficiently calibrate model parameters within the distributed, process-based NWM over hundreds to thousands of basins.
  • An new open-source community model evaluation and benchmarking system called “Rwrfhydro” was developed and greatly expanded, providing a common set of tools to evaluate multiple hydrologic and meteorological model variables against observations.
  • A capability to conduct extended period model retrospective simulations, re-analyses and selected model hindcasts for analysis and product development by NWC staff was created.
  • A new, dynamic, web-mapping service visualization tool called the “HydroInspector” was developed to give team members and collaborators access to NWM outputs and verification metrics in real-time.

The NWM provides an important bridge between the numerical weather prediction community and the terrestrial hydrology community. From a research perspective, the NWM backbone WRF-Hydro model is natively a coupling system for hydrology and atmospheric models and serves as an advanced modeling tool for weather and climate studies.  Recent studies by Kenandi et al. (2017, Theor. Appl. Climatol.) and Arnault et al. (2016, JHM) suggest WRF-Hydro is a useful tool in quantifying the regional atmospheric-terrestrial water partitioning and precipitation recycling and a range of space and time scales. Verri et al. (2017, NHESS) demonstrated how WRF-Hydro can be coupled to an estuary model for integrated coastal predication as well. For the forecasting community, the NWM configuration of WRF-Hydro provides a high-resolution,, national-scale, real-time, on-the-ground realization of the current state of operational NWM forced hydrologic response in the U.S. This ground-level translation of NWP forecasts opens new channels of research opportunities through evaluation against observations not typically used by the atmospheric community (streamflow, high-resolution soil moisture and flood inundation area). Given the research to operations environment in which the NWM is being developed, improvements made to WRF-Hydro are readily available to the research community on the model’s annual upgrade cycle, which accelerates new advances from the broader academic community to the nation’s weather and water operational forecast models.  The initial implementation of the NWM opens the door to a future where high-resolution hydrologic forecasts are more tightly coupled to regional and continental scale operational NWP models

In September 2017 the Department of Commerce recognized the development and implementation of the National Water Model with a Gold Medal for Scientific/Engineering Achievement; this is the highest award given by the agency. The award noted that “the NWM is the first high-resolution CONUS-wide water forecast model executed in the operational NOAA supercomputing environment.  It provides first-of-its-kind water resource guidance to NWS River Forecast Centers and other end users…” The NOAA/NCAR development team was specifically praised for demonstrating “exceptional dedication, hard work, expertise and creativity.”  The WRF-Hydro modeling system, which was developed at NCAR, has been successfully transitioned into the nation’s new state-of-the-art operational water prediction system through this exemplary NOAA/NCAR partnership.