Investigating Global Correlations Between Earthquake-generated Tsunamis and Subduction Zone Characteristics

Tsunamigenic earthquakes pose a large hazard in subduction zones, but it is currently unclear in which - if any - tectonic setting they preferentially occur. Here, we compile the global Subduction Nature & Interconnected Tsunamigenic earthquake Characteristics (SNITCH) database with parameters on subduction geometry, tectonics, and megathrust seismicity, and 329 tsunami events caused by earthquakes that were recorded in subduction zones between 1962 and 2018. We aim to identify potential correlations between tsunamigenic earthquake behaviour and subduction zone characteristics. We first use a bivariate regression analysis to find first-order correlations between the normalised number of earthquake-generated tsunami events Nt and the megathrust seismicity and tectonic parameters characterising a subduction zone. We confirm self-evident correlations between Nt and the number and magnitude of earthquakes and subduction kinematics. We then apply a multivariate Fisher analysis to see which combination of tectonic parameters best distinguishes subduction zone segments in which relatively many and few tsunami events caused by earthquakes have occurred. Despite the scarcity of the tsunami data, we consistently find that the type of margin (i.e., erosional or accretionary), the trench-normal component of the subduction and convergence velocity, the amount of trench sediments, and the roughness of the incoming plate are related to the number of earthquake-generated tsunami events. Our results therefore suggest that tsunamigenic earthquakes may occur more frequently in tectonic settings where plates subduct relatively fast beneath a sediment-starved, erosional margin with a complex, shallow subduction interface, characterised by multiple faults and fractures.