At the same time, forecasts of Atlantic tropical cyclone intensity have improved little over the past 25 years. I will present evidence that, in contrast to track forecasts, we are currently far from fundamental predictability limits on tropical cyclone intensity, and that at lead times of a few days and less, this is mostly owing to some combination of model error and errors in the initial condition. At longer lead times, intensity errors become increasingly dominated by track errors and errors in forecasts of the large-scale environmental wind shear. Partly for this reason, I will advocate for a migration away from the traditional track-intensity pair to probabilistic prediction of winds at fixed points in space, which are more societally relevant.
The talk will finish with a discussion of seasonal prediction and centennial projection of tropical cyclone activity.