Identification of subduction zone asperities and their seismogenic mechanisms using self-organizing map clustering

The rupture sources of strong earthquakes in subduction zones are generally asperities with spatio-temporal distribution. These asperities, which are patches featured by high stress and strong coupling in the late stage of seismogenic processes, can be identified by methods of multiple disciplines. It is important to comprehensively and uniquely identify asperities from different disciplines. Here we use the Self-Organizing Map (SOM) clustering method, integrating geodetic and seismological products, taking the 2011 M_W 9.1 Tohoku-Oki Earthquake as a case, to recognize the hazard sources and to investigate the seismogenic mechanism in subduction zones. The results show that the identified asperities coincide well with the main rupture zones. By subdividing 10 years before the mainshock into overlapping windows, we find that b value, which undergoes rapid changes after long-term stabilization, is an ideal indicator of asperity’s geometry. A comparative study with the 2014 M_W 8.1 Iquique earthquake indicates that the size of the asperity is mainly influenced by the megathrust curvature. We propose a model to explain the seismogenic behaviours of subduction zones, in which patches can transform between a stick-slip state and a conditionally stable frictional state. This study opens the possibility of identifying asperities and seismogenic mechanisms in subduction zones.