10B.3 Evaluation of RRFS-like Simulations Using GEFS and EnKF IC Perturbations Compared with the HREF: Toward Effective IC Perturbation Generation, Including Blending, in the RRFS

Wednesday, 31 January 2024: 11:15 AM
323 (The Baltimore Convention Center)
Jeff Beck, NOAA/GSL, Boulder, CO; and C. Schwartz, X. Wang, A. T. Johnson, M. A. Harrold, W. Mayfield, D. Dowell, C. Zhou, J. K. Wolff, V. Vargas Jr., and D. E. Lippi

With the first implementation of the Rapid Refresh Forecast System (RRFS), NOAA will soon be replacing a number of legacy, operational deterministic and ensemble regional modeling systems. Among other requirements, the RRFS will need to provide sufficient spread to ensure forecast quality equal to or better than the High-Resolution Ensemble Forecast (HREF) system. To meet this goal, efforts are underway within NOAA and partner institutions to implement and test methods to account for model physics and initial condition (IC) uncertainty. One such effort, based on a three-year WPO project, is looking at different IC perturbation options to test in RRFS prototype simulations. This work is also evaluating the use of blending global-scale GFS information into regional ICs to assess feasibility for use within the RRFS. One potential advantage of blending is a mitigation of the complicated partial cycling paradigm, which has been used in operational CAMs, and requires extensive I/O and complex workflows.

This presentation will outline findings from RRFS-like simulations using GEFS and EnKF IC perturbations. Results will be compared to HREF, including evaluations of spread/skill, reliability, and Brier score, to assess forecast performance against the operational standard for convection-allowing ensembles. Plans to test blended ICs for the RRFS will also be described. Following testing and assessments of these methods, a recommendation will be provided to RRFS developers for potential inclusion in a subsequent version of the RRFS.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner