GREAT EARTHQUAKES AND THE SUBDUCTION OF ASEISMIC RIDGES, VOLCANIC CHAINS AND FRACTURE ZONES

The unexpected 11 March 2011 Tohoku-Oki earthquake in Japan has reinvigorated a debate about the need to better understand earthquake supercycles on timescales of up to 1000 years and more, which may defy prediction using traditional methods. Global digital data sets represent a promising source of information for a multi-dimensional earthquake hazard analysis. We combine the NGDC global Significant Earthquakes database with a global strain rate map, gridded ages of the ocean floor, and a recently produced digital data set for oceanic fracture zones, major aseismic ridges and volcanic chains to investigate the association of earthquakes as a function of magnitude with age of the downgoing slab and convergence rates. Emphasis is placed on subduction zone segments that intersect with linear seafloor highs. We use a so-called Top-N recommendation method, a technology originally developed to search, sort, classify, and filter very large and often statistically skewed data sets on the internet, to analyse the association of subduction earthquakes sorted by magnitude with key parameters. The Top-N analysis is used to progressively assess how strongly particular “tectonic niche” locations (e.g. locations along subduction zones intersected with aseismic ridges or volcanic chains) are associated with sets of earthquakes in sorted order in a given magnitude range. As the total number N of sorted earthquakes is increased, by progressively including smaller-magnitude events, the so-called recall is computed, defined as the number of Top-N earthquakes associated with particular target areas divided by N. The resultant statistical measure represents an intuitive description of the effectiveness of a given set of parameters in accounting for the location of significant earthquakes on record. We find that different classes of subducting linear seafloor highs have vastly different effects in terms of their association with megathrust earthquakes, with fracture zones playing a much more significant role than aseismic ridges and volcanic chains. In addition we find that the age of the downgoing plate and convergence rates only weakly correlate with the magnitude of recorded earthquakes at intersection points between linear seafloor highs and subduction zones.