The Great Sumatra-Andaman Islands Earthquake and Tsunami of 26 December 2004 V Posters
Presiding: M H Ritzwoller, Center for Imaging the Earth's Interior, Department of Physics, University of Colorado; s ni, University of Science and Technology of China
U53A-01 1330h
Source mechanism of the Aceh Sumatra 2004 earthquake from very long period Earth free oscillations
The giant Aceh 2004 earthquake strongly excited the free oscillations of the earth, including some usually elusive very long period modes which are conspicous in the first calculated spectra. Several source models, published in the first days after the event, obtained from teleseismic body waves, show a rupture length of 400-500 km with a source duration of 200 sec approximately. On the other hand, the spatial distribution of aftershocks fills a much longer region, going further than 1000 km north of the epicenter, and some short period teleseismic observations suggest a source duration as long as 8 minutes. Both observations suggest a giant rupture, difficult to model with the standard teleseismic bodywave analysis du to overlapping of different phases. In order to gain some insight into the details of the rupture process, we model the very long period seismic and gravitational world wide records corresponding to this event. From the amplitude of the splitted singlets of the gravest free oscillations of the earth excited by the Aceh earthquake, we study its focal mechanism, seismic moment, depth, extent and history of the source.
U53A-02 1330h
Normal-Mode Excitation by Sumatran Earthquake and Short-Timescale THERMO-MECHANICS AND RHEOLOGY OF THE EARTH'S LITHOSPHERE
Solutions of the free-oscillation amplitudes,excited by the recent Sumatran wallop, by Okal and Stein ( 2005 ) have revealed a linearly growing trend in the semi-log plot between amplitude and period from 300 seconds to around an hour. This tantalizing plot (http://www.earth.northwestern.edu/people/seth/research/sumatra.html ) is very much reminiscent of the Rayleigh-Jeans portion of the Planck function in radiation physics, which was called the ultra-violet catastrophe. This distinct signature at long periods shows that some other physics must intervene to neutralize this singular tendency at a longer timescale. Thus in earthquake thermo-mechanics the size of an earthquake or moment is analogous to temperature in statistical physics. In this vein we have studied the thermal-mechanical shear interaction within the framework of a two-dimensional time-dependent model wherein a realistic visco-elastic-plastic rheology is implemented, and the governing equations include the momentum equation without inertia, the rheological and energy equations. We have retained all mechanical heating terms and heating terms involving volumetric expansion in the energy eq uation. In our simulations wherein we have modeled a bending situation, we encou nter two basically different bifurcation phenomena at the brittle-ductile transition-zone in the lithosphere, which can be attributed to two different families of eigenmodes of the system. One in which the shear zone nucleates on thermal perturbations in the ductile field, and the second which is fully associated with elasto-plastic (brittle, pressure-dependent) displacements. A quartz slab has all two modes operating simultaneously at three different depth levels. The bottom of the crust is controlled by the elasto-visco-plastic mode while the top is controlled by the elasto-plastic mode. The exchange of the two modes appears to communicate on a sub-horizontal layer in a flip-flop fashion, which may yield a fractal-like signature in time. The timescales of instabilities are found to decrease strongly with a decrease in the activation energy of the thermally activated processes. For olivine the timescales are around 10**5 years, whereas for quartz it goes down to around one month to a few years. Therefore, we have demonstrated that with a properly formulated thermal-mechanical and rheological model, we can generate timescales now very close to those of earthquakes and of the same order as slow earthquakes, which are the hallmarks of very large earthquakes, like Sumatra. These long-period seismological results together with future data acquisition from GPS data near subduction slabs, such as the Cascades, would help to link the two regimes of timescales and enable us to develop a uniformly valid rheological law for the lithosphere, which provides a link between the two regimes of elastic behavior and nonlinear dissipative dynamics.
U53A-03 1330h
Observations of Sub-Millihertz Normal Modes Excited by The Great Sumatra-Andaman Islands Earthquake
On social side, the Great Sumatra-Andaman Islands Earthquake of 26 December 2004 devastated south Asian countries as well as other circum-Indian Ocean countries with tsunami and became one of lethal disaster in human history. On scientific side, this earthquake released the largest seismic moment in the last half century and the first such (Mw 9 class) event in the era of modern digital recordings. Low frequency seismic signal excited by this event would test and improve our knowledge on large scale Earth structure and source process of large interplate earthquake. Our targets are low frequency normal modes of the Earth, and we focus our attention on the recordings at Japanese observatories which are located 40-60 degrees away from the epicentral region. In addition to broadband seismic recordings, we use various geodetic recordings such as strainmeters and extensometers. Signal-to-noise ratio of these recordings are remarkably high, which signifies the large excitation at the source. We employ conventional spectral analysis methods such as Fourier and multi-taper methods as well as "kyosin" (resonance) method [Morii, J Geod Soc Jpn, 2001], which employs the principle of a lock-in amplifier (or a phase sensitive detector; PSD) to detect isolated spectral peaks. The latter method has an advantage in case that frequencies of target spectral peaks are known beforehand and is a noise-resistant robust detector. We are able to detect fundamental spheroidal modes up to 0S36 as distinct peak with this method. Splitting of 0S2, the lowest frequency spheroidal mode, is clearly observed with the strainmeter recordings as well as broadband seismometer recordings, for which we use as long as approximately 25 days time series. Amplitudes of 5 splitted peaks of 0S2 appear to show temporal and spatial variation. Such spatial variation is understood as a function of distance between the centroid of moment release and observatories, as the mode amplitude diminishes near its node. Temporal variation is mainly due to large aftershocks, and could bias estimates of amplitude and width (Q) of mode peaks if ignored.
U53A-04 1330h
Estimate the Sumatra Earthquake Source Based on Satellite Observations and an Ocean-Bottom-Pressure Tsunami Model
Several satellite radar altimeters and tide gauges have observed the Indian Ocean tsunami triggered by the M9.0 earthquake off the west coast of northern Sumatra, Indonesia, on December 26. These observations are unique and of tremendous value for testing tsunami prediction models. Earthquake tsunamis are difficult to predict because the exact force that strikes the ocean bottom water which generates a tsunami is difficult to measure. Even if the force is measured, conventional tsunami wave models are not designed to utilize the measured information for better predictions. Consequently, an unacceptable 75 percent false alarm rate has prevailed. Here we show that by applying the earthquake data as ocean-bottom-pressure (OBP) forcing in a non-Boussinesq three-dimensional ocean model, we can simulate the Indian Ocean tsunami better in comparison with the satellite-observed sea-surface-heights and tide gauge observed waves. The OBP force is consistent with seismic estimates and theories [Ji et al., 2002]. It is also found that the tsunami surface wave is not equivalent to the corresponding OBP, which is much larger than previously thought after being converted into water thickness. The sensitivity study, verified by the satellite and other observations, indicates that the ocean-bottom-pressure variation caused by the ground motion is the key to the successful prediction of earthquake tsunamis.
U53A-05 1330h
Tracking the Rupture of the Great Sumatra-Andaman Islands Earthquake With Teleseismic Arrays
The waveform recordings of the Great Sumatra-Andaman Islands earthquake at different broadband networks located at teleseismic distances allowed imaging its rupture propagation in space and time by using a migration approach. The results suggest, that the rupture initiated at 3.6N latitude and 96.0E longitude being in very good agreement to the epicentral location estimated by NEIC from body-wave arrival data (lat: 3.3N, lon: 96.0E). From the epicenter, the rupture first propagated unilaterally North-northwest for about 220 s until it reached the northern tip of Sumatra. Then, the rupture turned to the North following the strike of the subduction zone before it ends after a total duration of around 430 s in the North of the Andaman Islands. Two major phases of energy release can be distinguished from the teleseismic observations which correspond to the two distinctly oriented segments of the rupture. Indications for a change in the radiation pattern are obtained from the observed variations in the energy-time release patterns at the different arrays. The inferred total length of the rupture is about 1150 km and the average rupture velocity is estimated between 2.5 km/s to 2.9 km/s. Later secondary phases limit the resolution capabilities of the method. However, using high-resolution array techniques, strong secondary phases can partly be resolved and ambiguities in the results can be reduced.
U53A-06 1330h
Low Frequency Data Analysis From the 2004 Great Sumatra Earthquake
For seismologists, the great Sumatra earthquake of 12/26/04 brought home several striking facts: 1) As the largest earthquake in 40 years, and the first one since the establishment of regional and global networks of digital, high dynamic range, broadband seismic sensors. Data of unprecedented quality have been acquired for this event, in particular at frequencies corresponding to the gravest modes of oscillation of the Earth. It is a unique opportunity to verify fundamental properties of the earth's global structure, previously acquired at great effort through stacking of lesser quality data. 2) The Mw 9.0 event caught seismologists unaware: until now, the sophisticated procedures developed in the last 20 years to model earthquake ruptures have been targeted at the more frequent Mw 7.0-8.0 earthquakes. Consequently, a whole new dimension of the earthquake source problem has suddenly emerged, necessitating the consideration of longer source durations, larger source lengths, more complex source rupture models, all this using a wider bandwidth. We will present preliminary results and inferences from measurements and modeling of splitting and attenuation of low frequency normal modes from the Sumatra earthquake, as well as attempt to address the controversy concerning the mode of strain release in the northern part of the rupture, using data from very long period records.
U53A-07 1330h
Does the Great Sumatrian Earthquake (Dic. 2004) Significantly Influence the Surrounding Volcanic Systems?
The great Sumatrian earthquake (Dic. 2004) is the third earthquake ever recorded in term of energy released, and it occurred in a strongly active volcanic region. The strong stress perturbations due to this event give the opportunity to constrain our knowledge about the effects of those perturbations on the surrounding volcanic systems. Here, we propose a forward test to compare the spatio-temporal distribution of the eruptions which follow the earthquake, to the coseismic (static) and the postseismic (time dependent) stress field due to the earthquake. With this aim, we apply a standard procedure which account for the preceding distribution of eruptions in the area perturbed by the stress field, and which fixes the rules of the comparison with future data.
U53A-08 1330h
Constraints on Source Rupture From the Directivity of P Waves Excited by the 2004 Sumatra-Andaman Megathrust
The rupture of the 2004 Sumatra-Andaman megathrust progressed from southeast toward the northwest, starting offshore central Sumatra and ending near the northern extreme of the Andaman Island chain. From the duration of large amplitude P wave energy highpassed at 1 Hz, the primary rupture appears to have lasted 400-500 s, long enough for the P wave to overlap with the PP and PPP body phases in observed seismic records. Source directivity causes the apparent rupture duration to vary from station to station, shorter for source azimuths aligned with rupture propagation (i.e. to the NW) and longer for source azimuths opposed to the rupture propagation. Using body wave data from stations of the Global Seismographic Network (GSN), Geoscope, GEOFON, Mednet and other broadband seismic stations, we examine the details of source directivity using body waves. We apply multiple-taper spectrum analysis to estimate the coherence of body wave signals recorded at different stations, on the assumption that correlated waveform features represent source effects, not site effects. Doppler shifts associated with rupture directivity should be manifest as coherence peaks at mismatched frequencies. We will examine whether rupture directivity exhibits detectable variation in different intervals of the source rupture.
U53A-09 1330h
Constraints on Source Rupture From the Initial Phase and Amplitude of Free Oscillations Excited by the 2004 Sumatra-Andaman Megathrust
The initial phase of a long-period seismic free oscillation, referenced to the onset of an earthquake, can be used to estimate the time centroid of the source rupture process if the rupture duration is short compared to one half-cycle of the oscillation. For the 2004 Sumatra-Andaman megathrust, we estimate the initial phase of several well-excited free oscillations nSl from stations of the Global Seismographic Network, Geoscope, GEOFON and other broadband seismic stations. We apply multiple-taper spectrum analysis to determine amplitude and initial phase of phase-coherent decaying sinusoids via maximizing an F-variance ratio test [Park, Lindberg and Thomson, 1987], using spherical harmonic stacking to isolate individual vibrational singlets. Preliminary results suggest that moment release did not follow a simple symmetric function about a centroid time. The initial phase of the "breathing mode" 0S0 (T=1227 s) corresponds to an apparent centroid time of nearly 400 s. If the source time function is a simple boxcar with 800 s duration, the amplitude of 0S0 should be deficient, relative to the longer-period 0S2 (m=0 singlet, T=3236 s), by roughly 50%. This amplitude deficiency is consistent with the relative amplitudes of the two modes, as predicted for a Mw=9.3 event. However, a simple boxcar time function for the Sumatra rupture does not match the intial phase measurements of either 0S2 (m=0 singlet) or the overtone 1S0 (T=613 s). Therefore the rupture process is more complicated than a simple boxcar or a similar symmetric function about a 400-s centroid time. An asymmetric source rupture time function is not surprising, because high-frequency body wave radiation lasted roughly 500 s, and a simple Sumatra CMT source with 400-s time shift creates surface waves that are very late relative to the data. If the Sumatra fault zone switched to "silent slippage" after 500 sec, the remainder of its moment release may have occurred over a much longer time. Park, J., C. R. Lindberg and D. J. Thomson, Multiple taper spectral analysis of terrestrial free oscillations: Part 1, Geophys. J. Roy. Astron. Soc., v91, 755-794, 1987.
U53A-10 1330h
Rupture Process of the December 26, 2004 Mw = 9.0 Sumatra Earthquake
We use primarily SH-wave data from 20 seismic stations at distances between ~40° and 85° to investigate the rupture history of this earthquake. The aftershocks define a 1200 km long zone along the Indo-Eurasian plate boundary. We use the aftershocks to identify the fault plane and the initial rupture area. We invert seismograms using the method of Das and Kostrov (JGR, 1990; PEPI, 1994) which uses constraints such as positivity of the slip rate at each time step, constraint on total moment, etc. to stabilize the solution. The mechanism of the Harvard CMT solution for this earthquake is used as the starting solution, but we search for alternate solutions with different strike, dip and rake as part of our inversion. We use only pure SH- and P-waves, that is, we terminate the seismograms when unmodelled phases such as ScS or PcP arrive, so we are able to study only about the first three minutes of the rupture process. The faulting process was found to be very heterogeneous. Details of faulting process, including moment distribution history, peak slip, peak slip rates, and average stress drops on different portion of the fault will be presented.