Validation and Verification of the FV3-GFS at NCEP (327)

Geoffrey Manikin, NOAA/NWS/NCEP/EMC, College Park, MD

Logan Dawson, NOAA/NWS/NCEP/EMC and I.M. Systems Group

Tracey Dorian, NOAA/NWS/NCEP/EMC and I.M. Systems Group


Mallory Row, NOAA/NWS/NCEP/EMC and I.M. Systems Group



As an initial major step in its effort to simplify and unify the operational Production Suite, the National Centers of Environmental Prediction (NCEP) will implement the Next Generation Global Prediction System (NGGPS) in 2019, using the Finite Volume 3 (FV3) model core. In preparation for this implementation, a Beta version of the FV3-GFS model will be run during 2018 to validate the new core as a global model. By May 2018, the FV3-GFS and operational GFS will be run simultaneously by NCEP Central Operations (NCO), allowing for real-time comparisons. A set of retrospective runs covering most of 2017 will also be made.The new Verification, Post-Processing, and Product Generation (VPPG) Branch of the Environmental Modeling Center (EMC) will conduct verification and validation activities of the FV3-GFS which will be critical to the overall NGGPS development. A combination of statistical and subjective verification will be performed, engaging the entire customer community in this critical effort. While results by the time of this conference will be preliminary, this presentation will discuss the process needed to validate and verify a new global model and help conference attendees learn how they can participate.EMC's Model Evaluation Group (MEG) will organize the evaluation and use its weekly forum to communicate and share results with the field. Statistical verification will involve a series of metrics computed over different regions of CONUS and the globe, starting with basic standard global verification metrics (including 500 mb anomaly correlation coefficients, RMS and bias scores for 2-m temperature and dewpoint, and precipitation skill scores), but taking advantage of Model Evaluation Tools (MET+) verification software to greatly enhance and expand the verification effort. Statistics covering multiple parameters over multiple regions over multiple forecast hours will be consolidated into scorecards. New diagnostic tools and existing tools such as MODE verification for precipitation and upper jets will also be applied. Subjective evaluations, both real-time and on retrospective model runs, will be performed on high-impact cases to make sure that the FV3-GFS can provide at least the same insight into significant events that the operational GFS does. Standard 2-D model forecast maps are used, but tools such as forecast soundings, station time-series plumes, and cross sections are also employed. These subjective evaluations offer a nice complement to statistical evaluations, as the statistics sometimes fail to tease out critical aspects of model performance. A supplemental team of National Weather Service Science and Operations Officers (SOOs) will assist in the formal subjective evaluation.