The Impact of Warmest Tropical SSTs on Seasonal North Atlantic Hurricane Frequencies and Future Projections

Modeling north Atlantic hurricane seasonal frequencies is an area of intense research, and predicting their response to anthropogenic global warming is of great societal interest. Prior research has found that a simple model using just north Atlantic (local) SSTs and tropical average (remote) SSTs can model north Atlantic hurricane frequencies with high accuracy. The non-local SSTs are believed to impact upper tropospheric temperatures and stability in the Atlantic. However, the main mechanism by which upper tropospheric temperatures are affected by the surface is through deep convection, which is collocated with just the warmest tropical SSTs. In this work, we show that the warmest tropical waters, not the tropical average, is a better predictor of hurricane activity. We use 1000s of years of simulated high-resolution atmospheric model data forced with a variety of uniformly and non-uniformly perturbed SST fields to train a simple model to predict seasonal Atlantic hurricane totals across a wide variety of climates. The model results show the response of north Atlantic hurricane activity is much more strongly tied to the 30\% warmest tropical SSTs and not the tropical average. We also highlight the role of the warmest waters in setting upper tropospheric temperature data in high-resolution model and ERA5 data. Finally, we discuss the utility of applying these statistical models to CMIP6 data. We obtain significant differences in late-century projections of hurricane activity when we use warmest tropical SSTs instead of tropical average. However, in many CMIP6 models the warmest SSTs are surprisingly not correlated with upper tropospheric temperatures. We explain this discrepancy, and the caveats that come along with using CMIP6 data and simple statistical models for future projections.