High Resolution Rapid Refresh

High Resolution Rapid Refresh is a convection-allowing, hourly-updating regional numerical weather prediction system designed for short-term severe weather and aviation forecasting. The system provides high-frequency, high-resolution mesoscale analyses and forecasts used operationally by agencies such as the National Weather Service, Federal Aviation Administration, Department of Defense, and private meteorological firms. Developed within a lineage of models that includes the Global Forecast System, Rapid Refresh, and regional models run by institutions like NOAA ESRL and University of Oklahoma, it emphasizes assimilation of dense observational networks and rapid update cycles.

Overview

The model originated from efforts at the National Oceanic and Atmospheric Administration and the National Centers for Environmental Prediction to produce a high-resolution, hourly-update analysis and forecast framework to support extreme-weather warning operations and transportation decision-making. Influential programs and partners include Office of Oceanic and Atmospheric Research, Storm Prediction Center, Aviation Weather Center, and research collaborations with National Severe Storms Laboratory, Cooperative Institute for Research in Environmental Sciences, and universities such as University of Washington, Penn State University, and Massachusetts Institute of Technology. Funding and operational mandates involve coordination with Federal Aviation Administration, Department of Energy, and regional emergency managers. The system fits within the broader US modeling family alongside Global Forecast System, North American Mesoscale Model, and research systems like Weather Research and Forecasting Model.

Model Design and Resolution

The model employs nonhydrostatic dynamics, advanced microphysics schemes, and high-resolution terrain and land-surface parameterizations to resolve convection and boundary-layer processes. Its horizontal grid spacing is typically 3 km over the contiguous United States, with vertical levels numbering between 50 and 64 to capture tropospheric structure, similar in ambition to experimental configurations developed at University Corporation for Atmospheric Research and National Center for Atmospheric Research. Physics options derive from community-model developments influenced by groups at Colorado State University, Oregon State University, and Naval Research Laboratory. The HRRR uses lateral boundary conditions from global analyses such as the Global Forecast System and leverages nested-domain strategies akin to regional implementations at Met Office and ECMWF research teams.

Data Assimilation and Observations

A key strength is ingestion of dense, heterogeneous observations on an hourly cycle, including data from the NOAA GOES satellite series, NEXRAD radar composite networks, surface mesonets like Atmospheric Radiation Measurement sites, and aircraft observations processed by Aircraft Meteorological Data Relay. Assimilation techniques draw from 3D-Var and hybrid ensemble-variational methods developed in collaboration with National Centers for Environmental Prediction, ECMWF, and research groups at University of Oklahoma and Penn State. Radar reflectivity and radial velocity assimilation from the NEXRAD network, wind-profiler data, GPS radio occultation, and surface observations from METAR stations enhance short-term prediction of convective initiation, influenced by assimilation research at Cooperative Institute for Mesoscale Meteorological Studies and National Severe Storms Laboratory.

Forecast Products and Applications

Operational outputs include hourly analyses, short-range forecasts to 18 hours (commonly 9–15 hours in operational delivery), probabilistic convection guidance, and derived fields for icing, turbulence, visibility, and fire-weather indices. End users include National Weather Service forecast offices, Federal Aviation Administration traffic flow managers, airline operations centers such as Delta Air Lines and United Airlines, emergency management agencies in states like Texas, Oklahoma, and Florida, and energy sector planners at companies working with the Department of Energy. The HRRR supports applied research into hail forecasting at institutions like University of Arizona and wind-energy forecasting used by companies and labs associated with National Renewable Energy Laboratory and Sandia National Laboratories.

Performance, Verification, and Limitations

Verification studies by NOAA, university groups, and independent researchers compare HRRR against benchmark systems like the North American Mesoscale Model and ensemble suites from ECMWF and Canadian Meteorological Centre. The model generally shows improved timing and placement of convective storms and better depiction of boundary-layer phenomena relative to coarser models, but challenges remain in convective-scale predictability explored by scientists at University of Oklahoma, Colorado State University, and University of Illinois Urbana-Champaign. Limitations include sensitivity to observation gaps, spin-up problems for convection, bias in microphysics-driven precipitation amounts, and computational constraints that have driven research into GPU acceleration at institutions like Argonne National Laboratory and Lawrence Berkeley National Laboratory. Ongoing development involves hybrid ensemble data assimilation, machine-learning post-processing researched at MIT, Stanford University, and University of Washington, and coupling with ocean and wildfire models to improve end-to-end operational utility.

Category:Numerical weather prediction