The Great Sumatra-Andaman Islands Earthquake and Tsunami of 26 December 2004 III
Presiding: R S Gross, Jet Propulsion Laboratory, California Institute of Technology; Y Bock, Scripps Institution of Oceanography
U51A-01 08:30h
Neotectonic Setting of the 2004.12.26 Sumatra Earthquake
The epicentre of the Mw 9.0 Indonesia subduction earthquake of 26 December 2004 was located 40 km NW of the island of Simeulue. Both uplift and subsidence are reported from this island and other parts of the area affected by the earthquake and they are widely explained as transient effects controlled by location relative to the megathrust rupture. Vertical movements were noted on Simeulue and several other outer-arc islands after previous earthquakes including the Mw 7.7 event of 1935. 14C dates on Holocene palaeoshore samples collected in 1987 on Simeulue and the neighbouring island of Nias have been recalculated using updated calibration and marine reservoir values. They support the earlier conclusion that over the last 6000 years Simeulue and Nias have undergone northeastward tilting in increments of ~1 m within deformation phases lasting ~1500 yr, the last of which began about AD 1695 and is still in progress, and separated by two periods of quiescence of similar duration. The geometry of uplift indicates coseismic displacement on imbricate faults within the sediment prism of which Simeulue and other outerarc islands are emerged portions supplemented by landward rotation of the prism. The imbricate model is complementary to the concept of a megathrust: faults within the weak sediments of the accretionary prism cannot account for great earthquakes, but by distributing strain along as well as across strike they influence the pattern of surface deformation resulting from major subduction events. They also release stored energy as relatively shallow aftershocks and thus contribute to the diffuse seismicity of the subduction zone. The errors inherent in teleseismic location limit its value for mapping faults within the prism, but some indication of their spacing is given by swath bathymetry obtained by the Royal Navy's HMS Scott soon after the earthquake, and reported at www.pmel.noaa.gov/tsunami/indo20041226/hms_scott.htm, which shows fold-thrust ridges trending NNW some of them 70 km long and of which the youngest are 5-15 km apart, ~500 m high and separated by sediment-filled basins. There are instructive parallels with the Iranian Makran which suggest that deformation may be serial in development. Holocene and bathymetric data may thus be of value in hazard mitigation as indicators of eventual coastal configuration once coseismic deformation has faded and as guides to possible shifts in the locus of shallow seismicity.
U51A-02 INVITED 08:45h
Southern terminus of the 2004 Aceh megathrust rupture deduced from emergence of coral microatolls
Coral microatolls on the fringing reefs of northwestern Simeuleu Island emerged as much as 150 cm during the 2004 earthquake. The pattern of uplift constrains the southern terminus of the 2004 rupture to be under the northwestern part of the island. It also constrains the downdip limit of coseismic rupture to be between Simeuleu and the mainland coast of Aceh. Our reconnaissance of the Simeuleu's reefs in mid-January 2005 revealed that the northwestern third of the island tilted northeastward, toward the mainland coast of Aceh. Reefs nearest the mainland emerged about 25 cm, while reefs farthest from the mainland emerged about 150 cm. The magnitude of slip on the megathrust needed to produce these uplift values would be about 10 m if the megathrust dips 15°, or about 20 m if the megathrust dips 8° beneath the island. The magnitude of coseismic emergence diminished southeastward to zero about midway down the island's long axis. The pattern and magnitude of emergence and tilt are similar to what occurred on the Mentawai islands farther south during the giant earthquake of 1833. The corals also reveal that uplift of up to 15 cm occurred just a couple years ago, near the terminus of the 2004 rupture. This probably reflects slip on the megathrust during the M 7.5 earthquake of 2002.
U51A-03 09:00h
Holocene turbidite and onshore paleoseismic record of great earthquakes on the Cascadia Subduction Zone: relevance for the Sumatra 2004 Great Earthquake
Marine turbidite stratigraphy, onshore paleoseismic records of tsunami sand beds and co-seismic subsidence (Atwater and Hemphill-Haley, 1997; Kelsey et al., 2002; Witter et al., 2003) and tsunami sands of Japan (Satake et al., 1996) all show evidence for great earthquakes (M ~ 9) on the Cascadia Subduction Zone. When a great earthquake shakes 1000 kilometers of the Cascadia margin, sediment failures occur in all tributary canyons and resulting turbidity currents travel down the canyon systems and deposit synchronous turbidites in abyssal seafloor channels. These turbidite records provide a deepwater paleoseismic record of great earthquakes. An onshore paleoseismic record develops from rapid coseismic subsidence resulting in buried marshes and drowned forests, and subsequent tsunami sand layer deposition. The Cascadia Basin provides the longest paleoseismic record of great earthquakes that is presently available for a subduction zone. A total of 17 synchronous turbidites have deposited along ~700 km of the Cascadia margin during the Holocene time of ~10,000 cal yr. Because the youngest paleoseismic event in all turbidite and onshore records is 300 AD, the average recurrence interval of Great Earthquakes is ~ 600 yr. At least 6 smaller events have also ruptured shorter margin segments. Linkage of the rupture length of these events comes from relative dating tools such as the "confluence test" of Adams (1990), radiocarbon ages of onshore and offshore events and physical property correlation of individual event "signatures". We use both 14C ages and analysis of hemipelagic sediment thickness between turbidites (H), where H/sedimentation rate = time between turbidite events to develop two recurrence histories. Utilizing the most reliable 14C and hemipelagic data sets from turbidites for the past ~ 5000 yr, the minimum recurrence time is ~ 300 yr and maximum time is ~ 1300 yr. There also is a recurrence pattern through the entire Holocene that consists of a long time interval followed by 2 to 5 short intervals that is apparent as well in the land records. This pattern has repeated five times in the Holocene. Both onshore paleoseismic records and turbidite synchroneity for hundreds of kilometers, suggest that the Holocene turbidite record of the Cascadia Subduction Zone is caused dominantly by triggering of great earthquakes similar in rupture length to the Sumatra 2004 earthquake. The recent Sumatra subduction zone great earthquake of 2004 and the 1700 AD Cascadia tsunami sand of 3m height preserved in Japan (Satake et al., 1996) show that ocean-basin wide tsunami deposits result from these great earthquakes, which rupture the seafloor for hundreds of kilometers. Cascadia and Sumatra share many geological and physiographic similarities that favor the deposition of turbidites from great earthquakes, and tend to filter non earthquake turbidites from the record. Thus the paleoseismic methods developed in Cascadia could be applied to the Sumatran Subduction Zone and we expect that the turbidite record would yield a similar record ~10,000 yr in length. In Sumatra, the dearth of such records led to the lack of widespread recognition of the hazard, particularly from the northern Sumatra and Andaman-Nicobar region where geodetic data suggested weak plate locking. Evidence of a tsunami similar to the 2004 event from satellite imagery suggests the previous event was in the recent past.
U51A-04 09:15h
ICESat Observations of Topographic Change in the Northern Segment of the 2004 Sumatra-Andaman Islands Earthquake Rupture Zone
The Andaman Islands are located 120 km east of the Sunda trench in the northern quarter of the 1300 km long rupture zone of the 2004 Sumatra-Andaman Islands earthquake inferred from the distribution of aftershocks. Initial field reports indicate that several meters of uplift and up to a meter of submergence occurred on the western and eastern shorelines of the Andaman Islands, respectively, associated with the earthquake (Bilham, 2005). Satellite images also document uplift of western shoreline coral reef platforms above sea level. Body-wave (Ji, 2005; Yamamaka, 2005) and tide-gauge (Ortiz, 2005) slip inversions only resolve coseismic slip in the southern one-third to one-half of the rupture zone. The amount of coseismic slip in the Andaman Islands region is poorly constrained by these inversions. The Ice, Cloud, and land Elevation Satellite (ICESat), a part of the NASA Earth Observing System, is being used to document the spatial pattern of Andaman Islands vertical displacements in order to constrain models of slip distribution in the northern part of the rupture zone. ICESat carries the Geoscience Laser Altimeter System (GLAS) that obtains elevation measurements from 80 m diameter footprints spaced 175 m apart along profiles. For surfaces of low slope, single-footprint absolute elevation and horizontal accuracies of 10 cm and 6 m (1 sigma), respectively, referenced to the ITRF 2002 TOPEX/Poseidon ellipsoid are being obtained. Laser pulse backscatter waveforms enable separation of ground topography and overlying vegetation cover. During each 33-day observing period ICESat acquires three profiles crossing the Andaman Islands. A NNE-SSW oriented track consists of 1600 laser footprints along the western side of North, Middle, and South Andaman Islands and 240 laser footprints across the center of Great Andaman Island. Two NNW-SSE tracks consist of 440 footprints across Middle Andaman Island and 25 footprints across the west side of Sentinel Island. Cloud-free profiles were acquired in the fall of 2003 and 2004. During February-March, 2005 ICESat's precise pointing capability will be used to exactly repeat these three profiles, with a cross-track accuracy of better than 100 m, providing trench-parallel and -perpendicular observations of topographic change of the Andaman Islands that will compliment geodetic field surveys. The observed elevation changes will be compared to models of coseismic deformation associated with the mainshock and large aftershocks in the Andaman Islands region.
U51A-05 09:30h
Real Time GPS Magnitudes for the Largest Earthquakes, and Application to Tsunami Warning
Seismology has two well-known limitations in the rapid determination of magnitudes and source models for the largest earthquakes. The first is that all magnitude scales used in seismology saturate, with the exception of moment magnitude determined from very long period data. This causes initial magnitudes to be underestimated. For example, in the case of the December 2004 Sumatra-Andaman earthquake, the initial magnitude estimate was 8.0. About 2 hours after the earthquake the magnitude was revised upward to 8.5 based on additional longer-period data, and the Harvard CMT solution with Mw 9.0 was released about 5 hours after the earthquake. The second limitation is that for ruptures that last longer than several minutes, finite source modeling becomes difficult because secondary phases from the initial part of the rupture arrive at the same time as direct phases from the later part of the rupture. Both of these limitations can be overcome given time, effort, and later-arriving data, but for near real time applications, magnitude determination of the largest earthquakes remains problematic. Real time GPS data from near-regional distances can be used to determine magnitude for the largest events within minutes of the earthquake, with no upper limit on magnitude. In fact, the larger the earthquake, the better GPS will perform because the signal to noise ratio for displacements increases. This makes real time GPS recordings highly complementary to seismic methods. Displacement records from large earthquakes show that the static displacements are largely complete after the passage of the direct shear waves. After this point, a reliable estimate of the static displacement can be made as long as the dynamic displacements from the surface waves are avoided. In the Sumatra-Andaman earthquake, a continuous GPS site in Medan, Sumatra, about 400 km from the epicenter, was displaced by ~14 cm horizontally. An epoch-by-epoch GPS solution shows that this displacement could be measured with a precision of ~2 cm using data no more than 15-20 minutes after the start of the rupture. Production of a GPS displacement record in near real time is feasible, and given data from a few sites within near-regional distances (within a few to several hundred km of the rupture), seismic moment can be estimated reliably using a simple and reliable inversion scheme. I use real data from the 2002 Denali earthquake (Mw 7.9) and synthetic data from the 2004 Sumatra-Andaman earthquake to illustrate the feasibility of the system, and evaluate how many GPS sites would be required to provide accurate magnitude and finite source model determination.
U51A-06 09:45h
GPS Measurements of Coseismic and Postseismic Deformation for the Mw 9.0 2004 Sumatra Earhquake: Constraints on Seismographic Inversions of the Earthquake Source and Tsunami Models
The active tectonics of western Indonesia is dominated by convergence of the Australia plate with Sumatra and Java. Along Sumatra the direction of convergence is highly oblique to the trench strike, and is partitioned into nearly arc-perpendicular thrusting at the trench and arc-parallel, right-lateral slip at the Sumatran fault (SF). Both faults are the source of strong and frequent earthquake activity. Two great (M > 8-9) subduction earthquakes occurred in 1833 and 1861, and several large (M > 7) strike-slip events are centered on or near the SF. We present coseismic and postseismic deformations for the devastating Mw 9.0 Sumatra earthquake of 26 December 2004, which ruptured to the NW of the 1861 rupture zone. We are currently performing resurveys with GPS of a network of tens of monuments established and surveyed repeatedly in Northern Sumatra, Banda Aceh, and the Mentawai Islands in the period 1989-2001, to study the mechanics of strain partitioning at this oblique subduction zone. The monuments have expected coseismic displacements ranging up to a few meters, and include stations within 100 km of the earthquake's epicenter, and within the zone of greatest destruction in Banda Aceh. A resurvey of these points is the best way to recover the permanent deformation of the Aceh and North Sumatra regions associated with the earthquake, and to monitor the subsequent postseismic deformation with good spatial resolution. We also present results from a continuous GPS network in North Sumatra and the Mentawai Islands, operating for nearly two years. Together, the resurveyed monuments and the continuous GPS stations will provide important constraints on seismographic inversions of the earthquake source and tsunami models, which up until now have had very little geodetic input. The repeat measurements will provide critical information on segmentation of the great earthquake rupture zones and their boundaries. Furthermore, we will be able to determine the wavelength of deformation that will be important for quantifying and understanding the (expected large) time dependent deformation. We also postulate the impact that a high-rate real-time GPS network would have had on mitigating the loss of human life resulting from this event, and present plans for such a network in Sumatra.