Development and Implementation of Global High-Resolution Systems for Seasonal Hurricane Prediction and Research

A focus of this presentation will be efforts to build an real-time prediction system within a research environment, while contributing to operations (in fact, being a part of the now operational North American Multi-Model Ensemble for seasonal prediciton - NMME). It will be argued that a key to "R20" is to replace the preposition "to" with the conjuction "and".

A family of high-resolution (50km and 25km atmospheric/land resolution) global coupled climate models, with an atmospheric component built on the GFDL Cubed-Sphere Finite Volume Dynamical Core (GFDL-FV3), provide a unified framework towards the understanding, intraseasonal-to-decadal prediction and decadal to multi-decadal projection of regional and extreme climate, including tropical cyclones. Initialized predictions of global hurricane activity show skill on regional scales, comparable to the skill on basin-wide scales, suggesting that regional seasonal TC predictions may be a feasible forecast target. The 25km version of the model shows skill at seasonal predictions of the frequency of the most intense hurricanes (Cat. 3-4-5 and Cat. 4-5). It is shown that large-scale systematic errors in the mean-state are a key constraint on the simulation and prediction of variations of regional climate and extremes, and methodologies for overcoming model biases are explored. Improvements in predictions of regional climate are due both to improved representation of local processes, and to improvements in the large-scale climate and variability from improved process representation. The systems are applied in a unified manner to real-time predictions and to broad research questions.