Expanding The HIGH-RESOLUTION Rapid Refresh (HRRR) From Deterministic To Ensemble Data Assimilation, Forecasts and Post-processing

Abstract

The 13-km Rapid Refresh (RAP) and 3-km convective-allowing High-Resolution Rapid Refresh (HRRR) are hourly updating weather forecast models that use a specially configured version of the Advanced Research WRF (ARW) model and assimilate many novel and most conventional observation types on an hourly basis using Gridpoint Statistical Interpolation (GSI). Included in this assimilation is a procedure for initializing ongoing precipitation systems from observed radar reflectivity data, a cloud analysis to initialize stable layer clouds from METAR and satellite observations, and special techniques to enhance retention of surface observation information. The RAP is cycled hourly with forecasts to twenty one hours covering much of North America and the HRRR is run hourly out to eighteen forecast hours over a domain covering the entire conterminous United States using boundary conditions from the hourly-cycled RAP. Experimental RAP and HRRR model development throughout 2014 and early 2015 culminated in a set of data assimilation and model enhancements that were incorporated into the first simultaneous upgrade of both the operational RAP and HRRR (to versions three and two respectively) in mid-2016. Development of the RAP and HRRR data assimilation and model physics has continued throughout 2016 and in this presentation we will highlight changes to the experimental RAP and HRRR that will form the foundation of the next operational RAP and HRRR upgrades scheduled by early 2018. Key among the data assimilation changes include 3-km hourly-cycling of atmospheric conditions to improve retention and evolution of forecasts across the first few hours along with the assimilation. Additional observational data sets are also assimilated including radar radial velocities, atmospheric motion vectors over land, and additional aircraft data along with refinements to the cloud analysis and assimilation of surface observations. In addition to the development of the deterministic HRRR, two efforts to extend the hourly updating forecast capability into ensemble prediction are underway. The first of these efforts leverages the HRRR forecasts in a cost-effective time-lagged ensemble (HRRR-TLE) to estimate hourly-updating likelihood probabilities of various weather hazards associated with heavy precipitation and severe thunderstorms over the CONUS out to 24 hours. We will highlight the post-processing techniques used to produce these HRRR-TLE products including quantile-mapping bias correction and temporal-spatial filtering to produce statistically reliable forecast probabilities. The second effort involves the development of a more expensive HRRR 3-km 40-member data assimilation and forecast ensemble (HRRRE) for a limited sub-CONUS domain. Ensemble spread is produced through initial condition perturbations and hourly-cycling of assimilated conventional observations using a GSI-based Ensemble Kalman Filter system. We will show examples of HRRR-TLE and HRRRE forecasts through both retrospective and real-time case-study examples and highlight the potential for future improvements in HRRRE spread-to-forecast-skill relationships with assimilation of radar reflectivity data, incorporation of stochastic physics and inclusion of HRRR-TLE post-processing with the ultimate goal of producing improved deterministic and ensemble HRRR forecasts.