On the seasonal predictability of the 2020 North Atlantic tropical cyclone season

Emma Lilly Levin, Mu-Ting Chien , Elizabeth Barnes, Haozhe He, Gabriel Vecchi, Wenchang Yang 

The 2020 Atlantic tropical cyclone (TC) season was unprecedented, producing a record storm count that exceeded the seasonal forecasted ranges, despite their anticipation of high activity. Here we assess the likelihood of the extreme 2020 season given the observed sea surface temperature (SST) forcing by examining a hierarchy of statistical, dynamical, and deep learning (DL) modeling frameworks, all constrained by observed SSTs. Although physics-based models anticipated a moderately active season, the observed outcome lay beyond their ensemble ranges. To investigate this result, we first test if the 2020 large-scale environment was conducive to a hyperactive season. Although the 2020 conditions were relatively favorable for TCs, they were not indicative of potentially record-breaking TC activity. We also generate 1,000 samples of plausible 2020 Atlantic TC season outcomes given the observed SSTs by leveraging a DL model. From this large ensemble, the observed season emerges as a 0.5\% event, which is highly unlikely, but not implausible. These results suggest that SST forcing provided a limited seasonal signal, and that internal atmospheric variability could have driven the observed hyperactivity on top of the moderately favorable large-scale conditions. The inability of physics-based model ensembles to encompass the observed outcome does not indicate a model failure, but reflects both limited predictability from SST forcing and the inability of relatively small ensembles (N=O(10)) to sample and estimate extreme tail risk like the 2020 season, seasons which are extremely unlikely any given year but have a substantial probability of occurring some year across decades.