An Evaluation of Surface Wind Structure Forecasts from the fvGFS and Operational Dynamical Models

The GFDL finite-volume Global Forecast System (fvGFS) is a model comprised of the FV3 dynamical core coupled with physics packages from the NWS Global Forecast System (GFS). In this paper, we examine methods of evaluating forecasts of tropical cyclone surface wind structure from the fvGFS and from operational dynamical models, including the GFS and the NWS Hurricane Weather Research and Forecast (HWRF) models. Methods will include the use of standard verification procedures to verify wind radii as well as procedures that directly verify the wind field forecasts from each model. Verifications of surface wind structure are important for evaluating model performance in predicting cyclone size as well as asymmetries in the surface wind field. The models will be evaluated by first verifying forecasts of the radii of 34-, 50- and 64-kt winds for the 2017 Atlantic and eastern Pacific seasons. An additional evaluation will involve the verification of model-predicted surface wind values that have been azimuthally averaged at specified radii surrounding 45-degree azimuths in each storm quadrant. This method provides a continuous wind profile for each model in each quadrant. This is in contrast to the standard method of verifying wind radii, which often results in inhomogeneous verification datasets due to variability across the models in the generation of winds at the different wind thresholds. Another method involves the application of techniques described by Powell and Reinhold (2007) to evaluate the near-surface integrated kinetic energy (IKE) in each model. Various IKE quantities will be evaluated for each of the models. A third method involves evaluating the distribution of surface winds in each storm quadrant. Comparisons between profiles of forecast and observed wind distributions will be shown, and validation statistics will be presented, including probability of detection and false alarm rates. In the talk, we will describe the methods used and provide evaluation of these methods for 2017 Atlantic and eastern Pacific storms using forecasts from the fvGFS, GFS and HWRF models.