Duo Chan, Gabriel A. Vecchi, Wenchang Yang , Peter Huybers

Changes in the statistics of North Atlantic hurricanes are known to depend upon the pattern of tropical sea surface temperatures (SSTs). Dynamical and statistical models are key tools to predict future hurricane activity, with our confidence in this application rooted in the models’ ability to skillfully reproduce hurricane variations over the past 30-40 years, when satellite data allows accurate reconstruction of observed ocean temperature variations. Extending the evaluation of simulations forced with historical SSTs against hurricane activity to century scales provides a more complete assessment of predictive skill, but which is limited in part by uncertainty in historical SST estimates. Here we show that recent corrections for systematic offsets in bucket SST measurements improve model skill in reproducing North Atlantic hurricane counts and lead to consistent reproducibility since the late 19th century. Changes in hurricane frequency introduced by revising historical SST data are of similar magnitude to projected changes for 2081-2100 in response to increasing greenhouse gases, highlighting the importance of accurately assessing SST patterns for both the historical period and the future.