Use of Storm-Scale Ensemble Data Assimilation for Initializing the Deterministic HRRR and Use of HRRRE Storm-Scale Ensemble Forecasts to Provide Probabilistic Aviation Hazard Guidance

NOAA’s Global Systems Division (GSD) has developed a storm-scale ensemble-base Data Assimilation System to initialize the deterministic High Resolution Rapid Refresh (HRRR) version 4, which is planned for NCEP operational implementation in Spring 2020. This analysis system, known as the HRRR Data Assimilation System (HRRRDAS), provides storm-scale ensemble background error covariance information to the hybrid 3DensVar used to initialize the deterministic HRRR. Ensemble mean fields from the HRRRDAS are also used to provide initial conditions for the deterministic HRRR pre-forecast hour. In addition to this major data assimilation upgrade, the HRRRv4 (and associated RAPv5) feature numerous enhancements to the model physics and numeric aspects. Extensive real-time and retrospective test results indicate improvement for nearly all aspects of the HRRRv4 relative to the HRRRv3 and will be presented at the conference.

The HRRRDAS, which is comprised of a 36-member EnKF-based hourly update cycle, also provides initial conditions for a 9-member storm-scale ensemble forecast system. This forecast system, known as the HRRR Ensemble (HRRRE) provides forecast uncertainty information about convection and other aviation hazards. This information can then be used to create statistically reliable probability guidance, as well as a number of other hazard uncertainty products.

The HRRRDAS initialized HRRRE has been continually refined over the past three years and has shown significant improvement in most measures (reduced mean bias and RMS, improved spread-skill relationship, etc.). A recent area of focused research has been on optimizing the use of stochastically perturbed parameterization, which has yielded the desired result of increased ensemble spread without increased ensemble mean RMS errors. Recent in house and testbed assessments suggests it has achieved a skill and spread similar to that of the HREF.

This ARW-based ensemble system continues to be used as a test bed for improving single-core ensemble data assimilation and ensemble forecast methods as a prelude to planned operational implementation within the FV3-based Rapid Refresh Forecast System (RRFS) that will replace the RAP and HRRR around 2022. In anticipation of this switchover to the FV3 following the HRRRv4 (and concurrent RAPv5) implementation in spring 2020, we have been actively working on improving stand-alone regional (SAR) storm-scale model-forecast with the FV3. This work has including optimizing the Gnomonic grid domain configuration, porting the RAP/HRRR physics components to the FV3 (using the Common Community Physics Package – CCPP), and work to initialize the SAR FV3 from RAP and HRRR grids. All of this work will be summarized in the presentation.