The Great Sumatra-Andaman Islands Earthquake and Tsunami of 26 December 2004 I
Presiding: J Park, Yale University; S Bilek, New Mexico Institute of Mining and Technology
U43A-01 INVITED 13:30h
Source characteristics of the great Sumatra earthquake and its aftershocks
Rapid centroid-moment-tensor (CMT) analysis of the December 26, 2004, Sumatra earthquake in the period range 300--500~s gave a seismic scalar moment of 4×1029~dyne-cm (MW~9.0). The strike, dip, and rake of the presumed fault plane were found to be 329°, 8°, and 110°, respectively. The centroid location was determined to be 3.1°N, 94.3°E, and the centroid depth 29~km. The centroid time shift was found to be ~140~s. Surface-wave directivity indicates rupture toward the northwest, consistent with the locations of aftershocks occurring in the first month following the mainshock. We present further analysis of the mainshock source characteristics, including evaluation of the possibility that the total moment and source duration were significantly larger than those determined in our rapid CMT analysis. We also present focal-mechanism solutions for aftershocks having moment magnitudes MW~5.0 and larger. A large number of these aftershocks occurred during an unprecedented earthquake swarm east of the Nicobar Islands. The swarm began one month after the mainshock, on January 26, 2005. We detected more than 150 earthquakes of M~5.0 or larger during a period of three days, more earthquakes of this size than normally occur globally during a two-week period. The centroid locations of the earthquakes comprising the swarm are tightly clustered, lying within a radius of ~30~km. These earthquakes show strike-slip and normal faulting, with axes of maximum tension oriented northwest--southeast.
U43A-02 INVITED 13:45h
Ultra-long period seismic moment of the Sumatra earthquake: implications for the slip process and tsunami generation
We have estimated the seismic moment at ultra-long periods using split singlets of the normal modes 0s2, 0s3, and 0s4, as well as the radial modes 0s0 and 1s0. Consistent estimates of moment and Q for the split modes result from fitting both amplitude spectra and the decay of narrow-band filtered singlets (we leave Q constrained for the radials); Q estimates are consistent with previously reported values. We find that seismic moment increases steadily with period from 4E29 dyn-cm inverted by the CMT project at T = 300 s to 1.3E30 dyn-cm (Mw = 9.3) at the period of 0s2 (3231 s), making the Sumatran earthquake the second largest ever recorded instrumentally. The larger magnitude likely reflects slow slip along the entire rupture zone suggested by aftershocks. The systematic increase in moment with increasing period is consistent with this idea, and has not been previously observed for other earthquakes, raising important issues notably at what period the moment ultimately stabilizes. The variation of moment with frequency follows an f**(-1/2) law, thus indicating a source time function more complex than a simple boxcar. The larger moment is fit by a slip of 13 m on a fault 1200 km long by 200 km wide, a larger area than implied by body wave inversions that limited slip to its southern part. The new source model is supported by the successful modeling of the tsunami on the high seas, as detected by the JASON satellite, using a source including the northern segment [V. Titov, pers. comm.], and by the directivity of the tsunami towards Sri Lanka and East India. If the entire aftershock zone slipped, strain accumulated on the northern part of the rupture has also been released, leaving no immediate danger of a comparable tsunami being generated on this segment of the plate boundary. However, the danger of a local tsunami due to a large aftershock or of a large tsunami from a great earthquake on segments to the south remains.
U43A-03 INVITED 14:00h
slip distribution and rupture history of the 2004 Sumatra-Andaman islands earthquake
We model the slip-distribution and rupture history of the 2004 Sumatra-Andaman islands earthquake using broadband waveforms downloaded from the IRIS DMC. Three independent studies which cover a frequency band from 0.2 mHz to 5 Hz, have been preformed to understand this complex rupture process. In the first study, a preliminary model was obtained from teleseismic P and SH waves bandpassed from 2 to 200 sec. This model indicates that most of seismic energy within this frequency band was radiated from a 450 km long segment of the southern end of the aftershock zone. We verified this model by matching the static displacement at the nearby GPS site SAMP (Sampali, Medan, Sumatra, 260 km east of the hypocenter). We have also forward-computed long period (80 to 1000 sec) SEM (Spectral Element Method) waveforms at all GSN stations. The synthetics fit the observations remarkably well at period of 200 to 500 sec, but display some discrepancy at very long period (500 to 1000 sec). This preliminary result is limited because the seismic records used for inversion are not long enough to contain the energy coming from the northern segment. Our forward modeling of broadband waveforms at nearby station PALK (distance =15 deg.) as well as the discrepancies at very long periods (500 to 1000 s) suggests that some slip occurred as late as 600 sec after the origin time. This result is consistent with the second study which investigates the envelope function of high frequency (2 to 4 Hz) P waves. The azimuthal variation of the duration of envelope functions suggests a total rupture length of about 1200 km and rupture duration of 500 sec or longer. The third study investigates the normal modes spectra. The result also suggests significant slip over the northern segment of the fault. To better constrain the slip history of such a long duration earthquake, we are developing finite fault inverse procedures utilizing the longer seismic waveforms and normal-mode spectra. The inverted result will be verified using the 3D SEM simulations.
U43A-04 14:15h
Rupture Process of the Great Sumatra-Andaman Earthquake
We are using Rayleigh waves with periods between 100 and 500 seconds to map seismic moment release during the Mw 9.3 Sumatra-Andaman earthquake along the northern Indonesian subduction zone. To isolate source effects from the observed waveforms we deconvolved normal-mode synthetic seismograms computed using PREM and the Harvard CMT mechanism located at the USGS-NEIC epicenter. We obtain about 200 useable R1 source time functions with a good azimuthal distribution, admittedly more densely sampled to the north. Our preliminary results suggest that the major normal seismic moment release began about one minute after the initial slip (identified by the USGS-NEIC origin time) and then continued for about 400-450 additional seconds. Using observed source time functions for stations located perpendicular to the rupture direction and assuming a rupture velocity of 2.5 km/s, we estimate that the average slip along a width of about 150 km is on the order of 10 to 20 meters along much of 1250-km long rupture zone. Still, although the total moment associated with this slip equals that of the Harvard CMTsolution, it underestimates longer-period moment estimates by about a factor of three. We used an inverse Radon Transform (IRT) to extract rupture constraints from the azimuthal variations of the available source time functions. The IRT approach limits us to constraining first-order moment-release variations along the rupture strike but allows complete freedom to the temporal release of moment (i.e. no prescribed rupture velocity). Initial IRT results suggest relatively normal seismic moment release from near the hypocenter stretching at least 800 km northward. The initial minute of rupture appears to be slow (but within our resolution this could reflect been a hesitation in rupture). The long-period duration of 800 seconds (J. Park) and the substantially larger moment (S. Stein and E. Okal) suggest that at least 2/3 of the moment was released very slowly and predominantly influenced waves with periods greater than about 500 seconds. The location of this missing moment is an intriguing first-order problem associated with this very large earthquake.
U43A-05 14:30h
Very Long Period Normal Mode Analysis on 2004 Sumatra Giant Earthquake
2004 Sumatra giant earthquake is the largest event since 1964 Alaska earthquake and Harvard CMT gives a moment magnitude of 9.0 .Such a shallow thrust event produced huge tsunami across the Indian ocean and cause tremendous damages. However, seismic stations around globe honestly record the ground motion and provide an unprecedented dataset to retrieve details of source processes of such a giant event, while entire seismic frequency band (0.3 mHz ~ 4 Hz) can be used. Although aftershock zone extends about 1200 km north to the Andaman Island, direct mapping from the length of the aftershock zone to the rupture length is not straightforward. Teleseismic finite fault inversion (Ji et al) first gives slip distribution over a 400 km fault surface and promises a large slip about 100 km north of the CMT location. 3-D SEM synthetics based upon this finite fault model produce a very good fit to the 200-500 secs surface waves directivity (Hjorleifsdottir et al). However, high frequency P-wave envelope (Ni et al), very long period (500-1000 secs) surface waves (Hiroo et al) and the 0S2 normal mode splitting pattern suggest the rupture may actually extend further to the north and produce significant long period energies. We focus on the normal mode analysis of very long period at 0.2-3 mHz. Very long period records from GSN, Geoscope, TriNet and Berkerley netwok are collected and analyzed. Large numbers of records show clear splitting pattern of the gravest mode 0S2 and other modes due to rotation, ellipticity and 3-D heterogeneities. We compute coupled-mode synthetics to fully take into account the effect of rotation, ellipticity and 3-D Earth structure. Synthetics from a point CMT source and a finite fault model (Ji et al) are compared with the data. Preliminary result indicates long period moment (< 1 m Hz )deficit that is not explained by the 400 km finite fault model.
U43A-06 INVITED 14:45h
The Boxing Day Tsunami: Could the Disaster have been Anticipated?
The occurrence of the 26 December, 2004 Sumatra-Andaman earthquake and the accompanying "Boxing Day" Tsunami, which killed over 280,00, has been described as one of the most lethal natural disasters in human history. Many lives could have been saved had a tsunami warning system, similar to that which exists for the Pacific Ocean, been in operation for the Indian Ocean. The former exists because great subduction zone earthquakes have generated destructive, Pacific-wide tsunami in the Pacific Ocean with some frequency. Prior to 26 December, 2004, all of the world's earthquakes with magnitude > 9 were widely thought to have occurred in the Pacific Ocean, where they caused destructive tsunami. Could the occurrence of similar earthquakes and tsunami in the Indian Ocean been predicted prior to the 2004 Box Day Tragedy? This presentation will argue that the answer is "Yes". Almost without exception (the exception being the 1952 Kamchatka earthquake) the massive subduction zone earthquakes and tsunami of the Pacific Ocean have been associated with the subduction of relatively young ocean lithosphere (< 60 Ma), and the theory for why this should be so seems well established. Although the eastern part of the Sunda Arc off Java does not meet this criterion, the western part of the Sunda Arc offshore Sumatra does. Although there appears to be no reference to the great earthquakes off Sumatra which occurred in 1833 and 1861 in widely-used earthquake catalogs, these events have been reported in the literature and were the subject of recent research. In particular, research by Zachariasen et al. (1999 and 2000) had inferred that the magnitude of the 1833 event may have been as high as 9.2. Calculations for the tsunami that might have been associated with this event had shown, prior to 26 Dec, that it would affect the entire Indian Ocean basin, although due to the earthquake's location 1000 km southeast of the Boxing day event, the effects in the Bay of Bengal would not have been as severe. Thus, it seems to this author that the Boxing Day event could and should have been anticipated. This presentation will further consider why it was not, and what steps can be taken to anticipate and mitigate the effects of future events that may occur in the Indian Ocean and elsewhere.