Supercritical dynamics governs extreme seismic hazard

Earthquake hazard assessment relies on the principle that seismic events follow predictable size distributions, with the largest quakes limited by fault geometry and slip rates. Nevertheless, some devastating earthquakes, like in Japan in 2011, transcend these rules, exposing weaknesses in our understanding of extreme seismic events. Here, I show that conventional models fail to account for supercritical dynamics, a physical mechanism where stress cascades dynamically trigger faults failures, enabling whole-scale ruptures far larger than expected. By analysing 20 global regions on the base of historical data and seismic catalogues, I demonstrate that such mechanism produces heavy-tailed distributions, increasing the likelihood of megaquakes by order of magnitudes. This reconciles long-standing discrepancies between geodetic measurements and paleoseismic evidence revealing the occurrence of past mega-events. My findings challenge the assumption that fault segmentation inherently limits earthquake size, revealing instead that near-critical stress states can enable ruptures across multiple fault systems. This result implies that global hazard assessment, mostly based on subcritical seismicity models, must integrate physics-based supercriticality, geodetic and paleoseismic data to reliably quantify megaquake hazard.