Abstract
Abstract
The Sunda subduction margin produced the great 2004 Mw 9.1–9.3 Sumatra–Andaman earthquake, whose trans-oceanic tsunami killed over 200,000 people. No self-consistent framework yet explains why co-seismic slip during this earthquake varied so dramatically along strike, or why neighbouring fault systems responded so differently to the resulting stress changes. Here, by jointly inverting GPS velocities and earthquake slip vectors in a regional kinematic block model, we show that the rupture crossed a previously unrecognised plate boundary within the forearc, so that a single earthquake spanned two microplates loaded at different long-term slip rates and directions. This boundary, near ~6–7°N (which we term the Nicobar Neck), separates the Andaman and Sumatran segments of the rigid forearc microplate and coincides with a pronounced negative residual gravity anomaly, indicating a mechanically weaker zone. We further find that the offshore continuation of the Sumatran Fault (the Andaman–Nicobar Fault) accommodates ~39 mm/yr of trench-parallel slip (twice the onshore rate) yet has no record of large earthquakes, implying creep and/or distributed slip.