Verification of Operational Global Ensemble Forecast System in the Short to Medium Range

Following its strategic plan of developing unified modeling capabilities, the Global Ensemble Forecast System version 12 (GEFSv12) was implemented into operations on 23 September 2020 at NOAA National Centers for Environmental Prediction (NCEP). GEFSv12 employs the Finite Volume Cubed-Sphere Dynamical Core (FV3), same as that used by the Global Forecast System version 15, with the stochastically perturbed physics tendencies (SPPT) scheme and stochastic kinetic energy backscatter (SKEB) scheme to generate 30 members of ensemble forecasts, along with a control member. GEFSv12 is run on a horizontal resolution of about 25 km and provides forecasts out to 35 days.

To monitor operational GEFSv12 performance in real time, NCEP’s Environmental Modeling Center (EMC) developed a forecast verification system using the METplus software package. This new EMC Verification System (EVS) was implemented into operations on 26 March 2024, providing a wide range of statistical metrics and graphics for the entire NCEP operational modeling suite at this EMC website: https://www.emc.ncep.noaa.gov/users/verification/. The goals of the EVS are to have a set of community-vetted metrics for all NCEP development efforts. Specific requirements covering validation data sources, regions, and temporal attributes were recommended by the 2021 UFS Evaluation Metrics Workshop participants.

In this presentation, we provide a skill assessment of GEFSv12 ensemble forecasts in the short to medium range from retrospective analysis using operational forecasts and real-time verification from the EVS. We will briefly describe the system framework and capabilities of the global ensemble component of the EVS, which verifies operational GEFSv12 ensemble forecasts for up to 16 days. Performance metrics of selected variables, such as precipitation, temperature, and sea surface temperature, will be highlighted. More in-depth analysis will be given to precipitation to understand GEFSv12 prediction skill (both deterministic and probabilistic) and the sources of predictability. The spatial and temporal distributions of precipitation forecast skill will be examined. Additionally, GEFSv12 performance on selected high-impact events (e.g., extreme rainfall events) will be showcased to illustrate the strength and weakness of GEFSv12. Knowledge gained from this study can help identify potential model improvements for the upcoming GEFS version 13.