The link between great earthquakes and the subduction of oceanic fracture zones

Giant subduction earthquakes are known to occur in areas not previously identified as prone to high seismic risk. This highlights the need to better identify subduction zone segments potentially dominated by relatively long (up to 1000 yr and more) recurrence times of giant earthquakes. We construct a model for the geometry of subduction coupling zones and combine it with global geophysical data sets to demonstrate that the occurrence of great (magnitude >= 8) subduction earthquakes is strongly biased towards regions associated with intersections of oceanic fracture zones and subduction zones. We use a computational recommendation technology, a type of information filtering system technique widely used in searching, sorting, classifying, and filtering very large, statistically skewed data sets on the Internet, to demonstrate a robust association and rule out a random effect. Fracture zone-subduction zone intersection regions, representing only 25% of the global subduction coupling zone, are linked with 13 of the 15 largest (magnitude M-w >= 8.6) and half of the 50 largest (magnitude M-w >= 8.4) earthquakes. In contrast, subducting volcanic ridges and chains are only biased towards smaller earthquakes (magnitude < 8). The associations captured by our statistical analysis can be conceptually related to physical differences between subducting fracture zones and volcanic chains/ridges. Fracture zones are characterised by laterally continuous, uplifted ridges that represent normal ocean crust with a high degree of structural integrity, causing strong, persistent coupling in the subduction interface. Smaller volcanic ridges and chains have a relatively fragile heterogeneous internal structure and are separated from the underlying ocean crust by a detachment interface, resulting in weak coupling and relatively small earthquakes, providing a conceptual basis for the observed dichotomy.