RESEARCH ARTICLES exposes their highest corallites to the atmo- Deformation and Slip Along the sphere during lowest tides. This subaerial ex- posure kills the uppermost corallites in the Sunda Megathrust in the Great 2005 colony, thus restricting future upward growth. Hence the coral heads provide an opportunity to measure the difference between the highest Nias-Simeulue Earthquake level of survival (HLS) formed just before and 1 1 1 2 1 that formed just after a large uplift event (4, 6) Richard W. Briggs, * Kerry Sieh, Aron J. Meltzner, Danny Natawidjaja, John Galetzka, and even to extract interseismic histories of Bambang Suwargadi,2 Ya-ju Hsu,1 Mark Simons,1 Nugroho Hananto,2 Imam Suprihanto,3 2 1 4 4 vertical deformation (7, 8). Dudi Prayudi, Jean-Philippe Avouac, Linette Prawirodirdjo, Yehuda Bock When coseismic uplift occurs, those portions of the microatoll colony raised above lowest Seismic rupture produced spectacular tectonic deformation above a 400-kilometer strip of the tides die. But if lower parts of the coral head are Sunda megathrust, offshore northern Sumatra, in March 2005. Measurements from coral still below lowest tides, its uppermost living microatolls and Global Positioning System stations reveal trench-parallel belts of uplift up to tissues demarcate a new, post-earthquake HLS 3 meters high on the outer-arc islands above the rupture and a 1-meter-deep subsidence trough (4) (Figs. 2A and 3A). Most of our uplift mea- farther from the trench. Surface deformation reflects more than 11 meters of fault slip under the surements are derived from the difference be- islands and a pronounced lessening of slip trenchward. A saddle in megathrust slip separates the tween pre- and post-earthquake HLS, often on northwestern edge of the 2005 rupture from the great 2004 Sumatra-Andaman rupture. The the same coral head. Where corals rose during southeastern edge abuts a predominantly aseismic section of the megathrust near the equator. both the 2004 and 2005 earthquakes, we can differentiate between the two uplifts (Figs. 2A giant megathrust earthquake is the surements allows us to construct one of the and 3B). rare expression of the most dramatic most detailed and accurate maps of deforma- At locations where uplift was greater than the A moment of a subduction zone_s life tion obtained for a subduction megathrust height of the coral heads and at sites that cycle—the culmination of centuries of strain ac- earthquake. experienced subsidence (Fig. 2, B and C), we cumulation across a convergent plate boundary. Methods. Most of our measurements come record the elevation difference between the Robust seismic signals around the globe allow from massive corals of the genus Porites.Be- coral’s pre-earthquake HLS and average water estimation of the gross nature of the event, but cause these are sensitive natural recorders of level at the site at the time of our coral mea- the details of rupture are usually obscure due to lowest tide levels (3–5), they are ideal natural surement. We then use a numerical tidal model a lack of geodetic measurements directly above instruments for measuring emergence or sub- to obtain an estimate of the lowest annual low or nearby. mergence relative to a tidal datum. Massive tide expected at each survey site relative to the E 0 ^ The great moment magnitude (Mw) 9.2 Porites coral heads grow radially upward and water level at the time of measurement. Our Sumatra-Andaman earthquake of December outward until they reach an elevation that tidal calculations are based on harmonic tidal 2004 (Fig. 1) was unusual for a subduction megathrust event in that geodetic measurements Fig. 1. Regional map of of coseismic motions were available from is- the 28 March 2005 rup- lands directly above the rupture (1). These near- ture and previous large field data enabled a detailed investigation of the ruptures of the Sunda source of the earthquake. Even so, near-field megathrust. The 2005 Global Positioning System (GPS) geodetic mea- rupture occurred in a surements were sparse. Furthermore, they were 400-km gap between collected in survey mode, with long periods great ruptures in 2004 between measurements that led to ambiguities in and 1797. Islands above separating pre-, co-, and postseismic motions. the rupture allowed de- 0 The great (Mw 8.7) Nias-Simeulue earth- tailed measurement of quake, 3 months later and immediately to the coseismic deformation south (Fig. 1), presents a substantially better with corals and GPS. opportunity to constrain rupture processes. Sev- An, Andaman islands; eral continuously recording GPS (CGPS) sta- Nb, Nicobar islands; Ac, tions had just been established directly above Aceh province; Ni, Nias or immediately adjacent to the rupture (2). island; Sm, Simeulue is- Moreover, the presence of a tropical archipel- land; Bt, Batu islands; ago above the rupture enabled the use of coral Mt, Mentawai islands; Sfz, Sumatran fault zone; microatolls to measure coseismic uplift and NER, Ninety East ridge; submergence. The resulting rich set of mea- WFR, Wharton fossil ridge; IFZ, Investigator 1Tectonics Observatory, Division of Geological and Plane- fracture zone. Previous tary Sciences, California Institute of Technology, Pasadena, earthquake locations and CA 91125, USA. 2Research Center for Geotechnology, Indo- 3 magnitudes are from nesian Institute of Sciences, Bandung, Indonesia. Jl. Mahoni, (1, 8, 35, 36). Indian Blok E, Gang I, No. 13, Rt. 002/015, Kota Jakarta 14270, Indonesia. 4Cecil H. and Ida M. Green Institute of Geophysics and Australian plate mo- and Planetary Physics, Scripps Institution of Oceanography, tions relative to Sunda University of California San Diego, La Jolla, CA 92093, USA. are from (37) and faults *To whom correspondence should be addressed. E-mail: are generalized from [email protected] (24, 38). www.sciencemag.org SCIENCE VOL 311 31 MARCH 2006 1897 RESEARCH ARTICLES constituents extracted from a regional satellite- fit to estimate three-dimensional coseismic rections for these relatively small postseismic based model for Indonesia (9), using the soft- displacements (13). motions in our depiction of coseismic defor- ware package NLOADF (10, 11). The uplift or Results. The northern half of Simeulue is- mation. Instead, we leave close examination of subsidence value is the difference between the land, from about 2.5- to 2.9-N, rose during the pre- and post-earthquake deformation to a sub- pre-earthquake HLS (old extreme low tide ele- December 2004 event (Fig. 4A; table S2). Tilts sequent manuscript (14). vation) and the model value of post-earthquake were toward the northeast and southeast, with The vertical deformation pattern of March lowest tide. Where we can directly compare maximum uplift of 1.45 m on the island’s 2005 comprises principally two arc-parallel post-earthquake HLS and post-earthquake tide northwestern tip. The uplift of Simeulue con- ridges of uplift on the forearc islands of Nias elevations, we find that a band of living coral strains tightly the southeastern limit of mega- and Simeulue and a broad subsidence trough about 4 cm high can survive above the eleva- thrust rupture during the 2004 earthquake to between these islands and the mainland coast tion of extreme low tide. Thus, we apply this about 2.5-N(1, 11). The southeastern third of (Fig. 4B). This pattern—uplift nearer the defor- correction to all measurements that use the tidal Simeulue subsided in December 2004, but we mation front and subsidence nearer the arc (fig. model (12). had little time to document that subsidence in S2)—is like that observed after a few other At a few survey sites, coral records are the field before the subsequent March event. megathrust ruptures, principally in Alaska, unavailable. There we estimate uplift using We have only a few field measurements from Chile, and Japan (15–18). The asymmetry of geomorphic or cultural features. These measure- January 2005 and observations from satellite the pattern, with maximum uplift (2.9 m) ments often have relatively large uncertainties, imagery (11), as well as the subsidence pre- greater than maximum subsidence (1.15 m), is but they are still useful in that they offer dicted from an elastic dislocation model of the also similar to the patterns in these previous unambiguous evidence of the direction of land- 2004 rupture (1). Rather than apply these cases. The ridge crests are sharper than the level change. We also augment our field mea- largely model-derived corrections to our mea- trough. The contours of uplift and subsidence surements in a few locations with limits on uplift surements of March 2005 uplift on southeastern are predominantly arc-parallel, but have a pro- and subsidence derived from satellite imagery Simeulue, we contour net uplift values for sites nounced misalignment near the Banyak Islands, (table S1). Finally, we also use coseismic dis- that subsided in the 2004 earthquake (Fig. 4B). between Nias and Simeulue. placements recorded by CGPS stations of the Moreover, measurements by CGPS show that Modeling. We combine our coral observa- Sumatran GPS Array (SuGAr) (2) and a station the postseismic elevation changes by the time tions with coseismic three-component displace- (SAMP) operated by the Indonesia National of our field survey, 1.5 to 3.2 months after the ments from 16 CGPS stations of the SuGAr Coordinating Agency for Surveys and Mapping earthquake, were only rarely more than a few network to constrain an elastic dislocation model (BAKOSURTANAL). The CGPS data were percent of total coseismic motion (fig. S1). of coseismic slip on the megathrust (Fig. 5). To analyzed (13) in 24-hour segments (0- to 24- Therefore, we have chosen not to include cor- construct the model, we assume a 10- dipping hour GMT) with data from 10 additional con- tinuous GPS sites on Java, Cocos Islands, Diego Garcia, Singapore, India, Australia, and Guam.