CHPS: River forecasts for the entire United States
2 September 2021
The National Oceanic and Atmospheric Administration’s (NOAA) National Weather Service (NWS) issues forecast products for rivers across the United States. The NWS has thirteen (13) River Forecast Centers (RFCs), each of which is associated with at least one major river system. The RFCs provide river and flood forecasts for the protection of lives and property and basic hydrologic forecast information for environmental and economic wellbeing.
In 2008, NWS Office of Hydrologic Development began modernizing their river forecast system to more easily accommodate advances in hydrologic science within the public and private community. The Community Hydrologic Prediction System (CHPS) program was born, with Delft-FEWS as its technical core. Migrating to Delft-FEWS allowed the system architecture to evolve from a single procedural application to a much more modular and flexible suite of components.
Over the following years, Deltares and Deltares USA Inc. supported the RFCs in the successful migration to CHPS. A customized configuration was developed for each RFC, which focused on the river system(s) in its area of responsibility. In January 2012, all 13 RFCs went operational with their new systems. Along the way, several scientific and software development milestones were achieved (summarized below). Today, the system generates forecasts for the entire US, running thousands of models every day.
CHPS is used 24/7 by a teams of forecaster at each of the 13 RFCs to provide hydrological forecast information to the public (https://water.weather.gov/ahps/) and produce river forecasts and warnings in support of the protection of life and property and enhancement of national commerce.
Image source: https://www.weather.gov/safety/flood-products
The CHPS systems are also used in stand-alone mode to conduct research, run hindcasts, and calibrate models in the same software environment. This functionality has allowed the NWS to build a community dedicated to hydrological forecasting, in which it cooperates with academia on issues such as expanding its model base, model interoperability and data assimilation using the actual forecasting environment.
More information about CHPS can be found at https://training.weather.gov/nwstc/CHPS/roe.pdf.
CHPS is based on the principle of interactive forecasting: the hydrologist adjusts model states, parameters, or forcing time series of the hydrologic models in order to correct for differences between observations and model predictions, thus improving the forecasts. This process occurs every day in real time.
The figure below summarizes the main forecasting components and model chain for an individual river basin. CHPS automatically imports and processes hydrological and meteorological data, which are used to force a variety of snow, rainfall-runoff, unit hydrograph, routing, reservoir, river, and error correction models or routines, both in deterministic and probabilistic mode. The models include HEC-RAS for hydrodynamic calculations, HEC-ResSim for reservoir operations, and 17 other models developed by the NWS, including the Sacramento Soil Moisture Account Model (SAC-SMA) and Snow-17. Ensemble streamflow prediction is used for long term forecasts.
CHPS is a large-scale forecasting system that covers the entire US and is used daily by approximately 50 hydrologists to automatically runs thousands of hydrological models. The power of CHPS is the ability to handle the large amount of data and the constant interaction from different terminals (users). In the process of developing CHPS, many milestone innovations and software developments were achieved, which contributed significantly to the broader Delft-FEWS community, including:
Since the original implementation, the NWS RFCs have evolved their systems with new displays focused on climatological data, past forecasts and forecast skill, and additional model state time series. As a result of the collaborative development process used to implement the original system, the NWS RFC staff have been able to complete these modifications themselves.
In collaboration with Deltares, a new calibration capability was added to Delft-FEWS. The modifier functionality was extended to include model parameter modifiers, which permit the user to adjust model parameters in a manual search for the best parameter set. Configurations for displays and statistical calculations to compare volumes, peaks, and seasonal flows were also developed and now support the NWS model calibration efforts. More recently, automated parameter search during calibration has been made possible by coupling Delft-FEWS and OpenDA. These enhancements are available in all Delft-FEWS builds and potentially useful to all Delft-FEWS users.
The NWS also used Delft-FEWS to implement their Hydrologic Ensemble Forecast System (HEFS) and Multi-Model Ensemble Forecast System (MMEFS). The HEFS setup required hindcasting in Delft-FEWS to quantify uncertainty in the meteorological forcing and hydrologic flows. HEFS is run operationally from Delft-FEWS to provide bias-corrected ensemble flow forecasts, displays, and forecast exports.
The National Water Center (NWC) in Tuscaloosa, Alabama now hosts off-site backup operations for the NWS RFCs. If the duty-standby systems go down at an RFC, a Delft-FEWS system at the NWC will take over, ensuring continuity in operations. The backup system uses data synchronization functionality in Delft-FEWS. Such off-site backup was put into recent use during a hurricane event that compromised operations at one of the RFCs.