Geodetic Observations of an Earthquake Cycle at the Sumatra Subduction Zone: The Role of Interseismic Strain Segmentation

We use geodetic observations to examine the role of interseismic strain segmentation on the earthquake cycle at the Sumatra subduction zone, site of the 26 December 2004, Mw = 9.1 Sumatra-Andaman and 28 March 2005, Mw = 8.7 Nias-Simeulue earthquakes. Our data consist of survey-mode and continuous GPS data, and vertical deformation recorded by coral microatolls. The survey-mode data span the 15-year period 1991-2006, up to one year after the Nias-Simeulue earthquake, thus allowing us to observe the end of one earthquake cycle and the beginnings of a new one. Our modeling of these data shows that the Mentawai segment of the Sumatra megathrust that ruptured in an Mw = 8.8 earthquake in 1833, between 0.5° S and 5° S, is consistent with a fully locked thrust zone, and is flanked by two regions, the Batu Islands in the NW and Enggano Island in the SE, that are consistent with a freely slipping subduction interface. The Nias-Simeulue earthquake ruptured a 300-km-long segment of the Sumatra subduction zone directly SE of and abutting the rupture zone of the Sumatra-Andaman earthquake, the same region that ruptured in 1861. Our modeling of the interseismic, coseismic and postseismic phases of this earthquake cycle confirms then that the Sumatra megathrust is highly segmented, and that this is probably a long-lasting characteristic of this subduction zone. Recent studies suggest that the segmentation is related to geometry that continues at depth, and may thus have more to do with structural anomalies than with properties of the subduction interface. Our modeling of postseismic deformation indicates that afterslip on the fault migrated away from the main rupture zone. Afterslip from the Sumatra-Andaman earthquake was concentrated updip, whereas afterslip from the Nias-Simeulue event was concentrated downdip from the main shock. In both cases postseismic deformation occurred on parts of the fault that did not slip seismically, most of it taking place during the first few weeks after the main shocks and decaying logarithmically, consistent with rate-strengthening brittle creep, back toward the long-term interseismic rate. The deformation pattern on the Mentawai islands and observations of coral heads on the Mentawai segment of the megathust that last ruptured in 1833 suggests that the occurrence of a similar event in the near future is likely. Our kinematic analysis of preseismic CGPS data for the four hours prior to the Nias- Simeulue earthquake indicates that any short-term precursory motion to the next great event is difficult to detect.