Experimental Sub-seasonal Prediction of North Atlantic Hurricanes using the GFDL HiRAM

In this talk, we will present evaluations of the performance of HiRAM (GFDL’s non-hydrostatic high-resolution atmospheric model) for the simulation and prediction of the North Atlantic Hurricane activity on the sub-seasonal time scale. Based on a 25-year period of seasonal predictions, we find the HiRAM not only well captures the impact of Madden-Julian Oscillation (MJO) on the global TC genesis, but also faithfully reproduces the modulation of MJO on Hurricane activity (genesis, landfall, etc.) in the North Atlantic, including Gulf of Mexico and west Caribbean Sea region. A series of sub-seasonal (1-month duration) retrospective forecasts were then performed over the years 2000-2014 using two grid configurations of HiRAM: a) global uniform 25km-resolution grid and b) two-way nested grid with a 8km-resolution nest over North Atlantic. Overall, both grid configurations show very encouraging performance: the correlation between the predicted and observed Accumulated Cyclone Energy exceeds 0.7 over the 15-year period. The two-way interactive nested grid shows better performance than the uniform grid in the prediction of Major Hurricanes (Categories 3-5), which is because the increased resolution more realistically resolves the inner-core storm structure.