Real-Time, Continental Scale Streamflow Nudging with WRF-Hydro: A First Benchmark

The community WRF-Hydro modeling system is currently being implemented for national streamflow analysis and prediction in collaboration with the U.S. National Weather Service, the NOAA National Water Center and multiple university partners. The analysis configuration of the WRF-Hydro system will cycle hourly over CONUS to produce streamflow analyses and initialize short-term and medium-range forecasts. Streamflow (volumetric discharge) is one of the primary variables of interest, and the first (and currently only) variable treated by data assimilation. The analysis and assimilation model configuration consists of a 1km-resolution land surface model (NoahMP), 250m-resolution surface and subsurface routing models, a baseflow parameterization, and a Muskingum-Cunge, reach-based channel model on the NHD+ channel network. USGS streamflow observations at over 7,000 gaging stations will be collected in real time and ingested into the channel model using a nudging data assimilation technique (where errors at the observations are gradually reduced with model advances in time). The goal of this assimilation is to reduce discharge bias and improve channel initial conditions for short term discharge forecasting.

Nuding a model state on a stream channel network topology presents several unique challenges. Given the heterogeneity of physical processes governing streamflow, we solve the nudging parameters for individual observation locations (instead of a global solution). Details of the implementation and tuning will be presented. Performance results for both the analysis and forecast phases of the model cycle are presented and a first WRF-Hydro DA benchmark is established.