The remarkable predictability of inter-annual variability of Atlantic hurricanes during the past decade

[1] A newly developed global model, the Geophysical Fluid Dynamics Laboratory (GFDL) High-Resolution Atmospheric Model (HiRAM) which is designed for both weather predictions and climate-change simulations, is used to predict the tropical cyclone activity at 25-km resolution. Assuming the persistence of the sea surface temperature anomaly during the forecast period, we show that the inter-annual variability of seasonal prediction for hurricane counts in the North Atlantic basin is highly predictable during the past decade (2000–2010). A remarkable correlation of 0.96 between the observed and model predicted hurricane counts is achieved. The root-mean-square error of the predicted hurricane number is less than 1 per year after correcting the model's negative bias. The predictive skill of the model in the tropics is further supported by the successful prediction of a Madden-Julian Oscillation event initialized 7-day in advance of its onset.