High-Resolution Rapid Refresh (HRRR) Model and Production Advancements for 2013 with Targeted Improvements for Reliable Convective Weather Guidance in the National Airspace System

The High-Resolution Rapid Refresh (HRRR) is a CONUS 3-km convection permitting atmospheric prediction system run hourly in real-time at the NOAA Earth System Research Laboratory. The HRRR uses a specially configured version of the Advanced Research WRF (ARW) model (including Thompson microphysics, MYJ PBL, and RUC LSM). The HRRR is run out to fifteen hours over a domain covering the entire coterminous United States using initial and boundary conditions from an hourly-cycled 13-km mesoscale model, the WRF-ARW-based Rapid Refresh (RAP). The RAP assimilates many novel and most conventional observation types including satellite observations on an hourly basis using Gridpoint Statistical Interpolation (GSI) and includes a procedure for initializing ongoing precipitation systems from observed radar reflectivity data using a digital filter, a cloud analysis system to initialize stable layer clouds, and special techniques to enhance retention of surface observation information.

In this presentation we will review the performance of 2012 HRRR forecasts with an emphasis on warm-season convection in real-time and retrospective runs. We will document the reduction in moist bias of soil moisture, dewpoints, precipitation and convective initiation, particularly in the first few forecast hours of each model cycle, and show improved development and maintenance of mesoscale convective systems. We will also present an improvement in the HRRR echo top height forecasts that was applied in July 2012.

We will also preview the development of the 2013 HRRR forecast system with a focus on four areas including (1) establishment of data assimilation (including radar observations) at the 3-km scale to further reduce convective-scale “spin-up” in the first few forecast hours, (2) enhancement in model dynamics and physics including shallow convective parameterization to improve the timing of convective initiation in weakly-forced weather regimes, (3) reduction of latency in HRRR model forecast production through an accelerated 3-km analysis and more efficient post-processing, and (4) improved reliability and availability of HRRR forecasts through redundant high performance computer systems hosted in Boulder, CO and Fairmont, WV. We will also update progress on other anticipated changes in the cloud analysis and ensemble data assimilation in an hourly update cycle that will improve year-round performance of the HRRR. Finally, we will discuss the development of time-lagged ensemble convective probabilities produced from HRRR runs.