Bayesian optimization of submarine landslide parameters for the 2020 Alaska Sand Point tsunami

This is the authors' abstract. We don't add key points for this paper.

In the Alaskan subduction zone, a magnitude 7.6 earthquake in October 2020 generated a tsunami substantially larger than predicted by the earthquake magnitude alone. Here, we hypothesize that this additional source—a submarine landslide—contributed to this discrepancy. We used Bayesian optimization to efficiently estimate landslide parameters that reproduce the observed tsunami. The ensemble of best-fitting solutions yielded a thickness of 9.5–11.5 m, width of 108–140 km, length of 18–33 km, and delay time of 262–319 s. The relatively small thickness suggests a shallow landslide occurring within weak seafloor sediments. These results support the interpretation that a submarine landslide contributed to the non-seismic component of the tsunami.