Real-time probabilistic tsunami forecasting in Cascadia from sparse offshore pressure observations

Abstract

Near-field tsunami early warning in the Cascadia Subduction Zone is limited by sparse offshore observations. We investigate whether a hypothetical network of 175 ocean-bottom pressure sensors can support real-time Bayesian inference of the full spatiotemporal seafloor velocity field and probabilistic tsunami forecasting for a margin-wide and a partial fully-coupled Cascadia earthquake dynamic rupture-tsunami scenario. The simulated oceanic acoustic, Rayleigh, and tsunami wavefields are similar during the first two minutes after nucleation but diverge thereafter, enabling rapid earthquake scenario discrimination. Using an acoustic-gravity inversion with assimilation of pressure data, tsunami wave height forecasts are obtained in less than a second. We leverage a Bayesian inversion-based framework that splits the computations into an offline precomputation phase performed with large-scale computing facilities, and an online phase that computes forecasts and can be executed on a laptop. Forecast errors remain low at 22.1% for the margin-wide and 19.6% for the partial rupture.