S22A-01 INVITED
Persistent Behaviors of the Sunda Megathrust in Sumatra: Opportunities to Forecast Destructive Earthquakes
The 2000-km-long Sumatran section of the Sunda megathrust is a treasure trove of information about the long- term behavior of a great active fault. This is due principally to the presence of an island chain along the Sumatran outer arc, which provides a platform for both man-made and natural instruments atop the shallow, seismogenic part of the megathrust. My students, colleagues and I have been using these "instruments" to understand the megathrust's past and to forecast its future behavior. We recognize three long segments that produce great earthquakes separated by two short patches of low seismic coupling. Two of the long segments have just ruptured and the other is nearing the end of its seismic cycle. The largest seismic rupture along the Sumatran megathrust in the 20th century was a M 7.7 earthquake at the Equator in 1935 (G&R, 1954). GPS geodesy and coral paleogeodesy, historical seismograms and coral paleoseismology show that the rupture is embedded in a 120-km section of the megathrust that slips primarily aseismically. The ~2.3-m rupture of 1935 is the largest seismic event on this Batu Islands patch in at least the past 250 years, despite its high (45 mm/yr) rate of plate convergence. Immediately north of the Equator is the 350-km-long section of the megathrust that generated the Mw8.7 Nias- Simeulue earthquake of 2005. Analysis of records from several continuous GPS stations above the rupture and about a hundred vertically displaced living corals shows that large seismic slippage was confined to depths between 14 and 35 km. In contrast, post-seismic deformations recorded at the cGPS stations reflect concentrations of slip updip and downdip of the seismic rupture. Fossil corals record previous large earthquakes in 1861 and in about 1800 and 1845. This history, though still incomplete, suggests that a cluster of three overlapping large seismic ruptures ended the previous earthquake cycle there. The 1600-km Mw9.15 Aceh-Andaman rupture of 2004 began still farther north, on the north side of the poorly coupled Central Simeulue patch. Corals above the rupture in Aceh display a paleoseismic history characterized by very infrequent large uplifts. Prior uplifts occurred about 600 and 1000 years before 2004. The very long interseismic intervals implied by these dates suggest that the Aceh-Andaman section will lie dormant for several hundred more years, accumulating the strains necessary to generate another Mw9.2 earthquake and great tsunami. The 700-kilometer-long Mentawai section of the Sunda megathrust, south of the Equator, has been dormant since a great-earthquake couplet in 1797 and 1833. GPS measurements and rapid subsidence of modern corals show that this section is predominantly locked and storing strain, despite occasional very large aseismic events on the shallowest part of the megathrust. Coral paleoseismology reveals that great earthquakes or earthquake clusters have occurred about every two centuries since the mid-14th century. Thus the next great earthquake or earthquake sequence may be no more than a few decades in the future. Although this forecast is a bit fuzzy, it has led to the establishment of additional scientific experiments there and to modest preparation in coastal communities for the next great earthquake and tsunami. Nonetheless, the poor connections between scientific knowledge of hazards and their actual mitigation in our world most likely mean that hundreds of thousands of coastal Sumatrans will be killed, injured and left homeless by the next great Sumatran earthquake and tsunami.