Great Sumatra-Andaman Earthquake of December 2004: New Insights That Will Change the Next 40 Years and the Plate Tectonic Paradigm

 

by Arthur E. Berman

 

Nearly one year ago, the Great Sumatra-AndamanEarthquake produced the most destructive tsunami in history, with 283,000 people dead or missing in the Indian Ocean region. Modern geophysical data recorded during the earthquake and tsunami revealed unprecedented complexity in the mechanics of plate boundary rupture. No inversion model successfully accounts for the slip or rupture pattern produced by this event. New understanding of plate boundary behavior and mechanics that arises from the Great Sumatra-Andaman Earthquake may ultimately modify the existing plate tectonic paradigm.  

The earthquake initiated just before 8:00 a.m. local time at the epicenter on the morning of December 26, 2004, approximately 255 km (158 mi) south-southwest of Banda Aceh off the western coast of northern Sumatra (Figure 1). There was no warning for the earthquake or the ensuing tsunamithat rapidly reached the shores of the Indian Ocean, devastating coastal areas of Indonesia, Thailand,Myanmar, India and Sri Lanka.

 

 

The Great Sumatra-Andaman Earthquake was exceptional in every way. Its 1300-km (807-mi) rupture length is the longest of any known earthquake (Hanson, 2005). The rupture zone was as much as 240 km (149 mi) wide (Lay and others,2005). The rupture lasted over an hour, making it the longest known rupture period of any earth- quake (Hanson, 2005). It deformed an entire hemisphere, moving global positioning system (GPS) stations in southern India 4 m (13 ft) with peak-to-peak ground motion over 9 cm (3.5 in) in Sri Lanka (Bilham, 2005). No point on Earth was undisturbed, with peak ground motion greater than 1 cm (0.4 in) everywhere (Park and others,2005).

 

The Great Sumatra-Andaman Earthquake was the second largest earthquake in instrumental history releasing 4.3 × 1018 Joules (Bilham, 2005). This is approximately equivalent to a 100-gigaton nuclear explosion or the total energy used in the United States in 6 months. The tsunamiproduced by the earthquake displaced 30 km3 (7.2 mi3) of sea water (Bilham,2005). The 9.3 moment magnitude (Mw)of the earthquake was equal to the sum of all moment magnitudes of earthquakes during the decade that preceded it (Lay and others, 2005). The earthquake andtsunami killed more people than any othernatural disaster in history.

 

The Great Sumatra-Andaman Earthquake was also exceptional as the first very large earthquake to be recorded and measured by a spectrum of digital technologies that were not available during other large earthquakes of the 20^th century. Previous great earthquakes—the 1960 Chile Earthquake(Mw = 9.5) and the 1964 Alaska Earthquake (Mw = 9.1)—saturated existing analog measurement and recording equipment. The Sumatra-Andaman event was recorded by a global network of broadband,high dynamic range, digital seismometers. It was the first major earthquake to be monitored by the GPS. In addition, it was thefirst application of digitally recorded, long-period, free-oscillation modal geophysics.

 

A moment magnitude of 9.0 was initially calculated for the Great Sumatra-Andaman Earthquake. In March 2005, researchers at Northwestern University determined that the true magnitude was 9.3—approximately 3 times the energy released by a 9.0 event—after taking into account the full length and slip of the rupture,particularly the “slow slip” of the rupture’s northern portion (Fellman, 2005).

 

 

The December 2004 earthquake trajectory may be divided into three segments, along with a fourth segment created by the contiguous March 2005 Nias Earthquake (Figure 2). Rupture initiated at a depth of about 30 km (18.7 mi) within the Sumatra segment. Rupture speed was slow and slip was small for the first 50 seconds. Rupture then expanded over the approximate 420 km (261 mi) of the Sumatra segment at an average speed of 2.7 km/s(1.7 mi/s) and rapid slip of between 5 and 20 m (16.4-65.6 ft) (Lay and others, 2005).

 

From 230 to 350 seconds, rupture progressed along the 325-km (202-mi) Nicobar segment of the earthquake with an average 5 m (16.4 ft) of rapid slip and an average rupture speed of 1.1 km/s (0.7 mi/s). In the Nicobar segment, an additional 5 m of slow slip proceeded up to 3500 seconds after rupture initiation (Lay and others, 2005).

 

The Andaman segment of the December rupture was characterized by less than 2 m (6.6 ft) of slip from 350 to 600 seconds after rupture initiation. An additional 5 m of slow slip occurred from 600 seconds to more than 3500 seconds. Rupture speeds in the northern segment were only about 0.3 km/s (0.2 mi/s) (Lay and others, 2005). The most remarkable aspect of the Great Sumatra-Andaman Earthquake was the slow slip that followed the initial, rather characteristic rupture and unzipping of a plate boundary. Slow slip tripled the earthquake’s energy release and accounts for the revision of its moment magnitude from 9.0 to 9.3. Slip along the northern segments of the rupture zone occurred too slowly to generate tsunami waves. Had slip been as rapid on the Nicobar and Andaman segments of the rupture as on the Sumatra portion, the resulting tsunami would have been many times more devastating than what actually occurred.

 

The March 28, 2005, Nias Earthquake may be thought of as a large, late aftershock. It had a moment magnitude of 8.7 and an average 8 m (26.3 ft) slip along a 300-km (186.5-mi) segment. It was probably produced by plate boundary failure because of stress changes that resulted from the December 26 rupture. No significant tsunami was produced by this rupture.

 

Slow slip is poorly understood and its causes are largely empirical at present. Slow slip occurred where plate convergence became increasingly oblique. It also coincided with a change in age of the subducting lithospheric plate from 60 Ma to 90 Ma between the Sumatra and Andaman rupture segments. Age and accompanying textural differences may have resulted in changes in mechanical coupling along the fault plane: subduction of younger lithosphere may have provided a broader contact area (Lay and others, 2005). Subduction of  older lithosphere to the north may have contributed to displacement transfer across zones of previous back-arc spreading and associated dispersion of rupture energy (Figure 2).

 

Slow slip excited several of Earth’s fundamental resonances called free vibrational oscillations. Free oscillations were first reported after Fourier analysis of digitized analog seismic records from the 1960 Chilean earthquake. When seismic waves of different frequencies are diffracted back toward Earth’s surface, resonances are formed when these waves interfere constructively (Figure 3). Records from more than 400 stations of the Federation of Digital Seismic Networks were used to observe free oscillations during the Great Sumatra-Andaman Earthquake. Vibrational modes provide information about the composition of the Earth as well as the size and duration of the earthquake.

 

“Just like thumping a watermelon to hear if it is ripe, after a big earthquake thumps our planet, we measure the natural tones from seismograms to detect properties of Earth’s deep mantle and core,” said Jeffrey J. Park of the Department of Geology and Geophysics at Yale University. “The Sumatra-Andaman earth- quake produced the best documentation of  Earth’s free oscillations  ever  recorded”(“Thumping the Earth like a Watermelon,” 2005).

 

 

Moment magnitude or finite rupture modeling was thegeophysical standard prior tothe Great Sumatra-Andaman Earthquake. These models were based on a single, stablefault offset. Long-period seismic surface waves in the 100- to 300-second range wereused to produce an inversion that seemed to satisfy the relativelysimple yet robust modelsprovided by the plate tectonicmodel. Finite rupture was thepreferred model run in late December and early January following the Great Sumatra-Andaman Earthquake. Harvard Centroid Moment Tensor(CMT) models that yielded a 9.0-moment magnitude were typical of the generation of modeling that included the two great earth- quakes in Chile and Alaska during the second half of the 20th century. For the Sumatra earthquake, assumptions included a uniform slip of 5 m along a 1300-km fault whose width varied between 100 and 240 km and a uniform rigidity factor. This method did not account for well-documented geodetic observa- tions in the Andaman and Nicobar islands regions or the dispersal patterns and magnitude of the tsunami that followed the earthquake.

 

Geophysicists began to unravel the new, very long-period seismic and geodetic data acquired in late 2004 and early 2005 to reach a more satisfactory solution. It was soon learned that little or no conventional body or surface seismic waves were generated during much of the extremely long period rupture process in the Indian Ocean. More complicated models or series of models were therefore required to approximate both geodetic and tsunami- related observations. The modeling process was, to say the least, experimental. The results of the modeling were understandably not entirely convincing. It could be argued that, without knowing the answer, modeling could not have achieved even the less-than- satisfactory solutions we now have.

 

The problems that result, of course, from using magnitude to quantify earthquakes are a direct consequence of trying to summarize a process as complex as an earthquake in a single number. In the case of the Indian Ocean event there was, for the first time in history, enough data to characterize this complexity. The possible casualty of this information extravaganza may be our revered plate tectonic paradigm.

 

The overload of data that was produced by the Great Sumatra- Andaman Earthquake reminds me of the expectations that many oil companies have had with 3D seismic. If a prospect seems complicated with only 2D coverage, many companies have hoped and assumed that 3D seismic data would make things clear. Generally the opposite occurs. A confusing situation often becomes overwhelmingly complex when all the details can be imaged using 3D.

 

Without going into the details of the modeling geophysics, the Great Sumatra-Andaman Earthquake was a compound process of seismic energy release that involved variable slip amplitudes, rupture velocities, and slip durations. Most of the slip occurred in the southern third of the rupture region, with diminishing rapid slip that became increasingly slower toward the north. The causes for these variables are poorly understood at present.

 

Seismic finite rupture models that most closely agree with empirical observations are complicated and use teleseismic (distantly observed) waveforms to determine moment magnitude (P or body wave, Rayleigh and SH, or second harmonic, surface wave magnitudes). No fewer than three separate models were required to account for the Sumatra-Andaman slip and rupture patterns. Each method involved different parameterization of seismic source as a function of time and restricted fault plane slip to specific time windows for each step of the model.

 

In some areas, modeling efforts using free oscillation geophysics provided a better match between observed and predicted responses than the seismic finite rupture models. On the other hand, oscillation mode models had to rely on seismic body and surface waves as well as GPS data to achieve a reasonable solu- tion. Part of  the weakness in this interesting approach to earthquake modeling is that free oscillation information was not produced until the slow slip phase of rupture began. In other words, free oscillation models could not begin until the late, northern phases of the Sumatra-Andaman Earthquake occurred. This solution, incidentally, results in a full 4.5 degrees northerly shift of the earthquake’s centroid (the 3D center of energy release), an adjustment of nearly 500 km (311 mi)!

 

Equally laudable and unsatisfying models were generated using geodetic GPS data from 41 continuously operating reference sta- tions in the Indian Ocean region. This approach yielded an average calculated slip of slightly greater than the 5 m used by the Harvard CMT models and an accompanying 40% increase in moment magnitude of 9.1 (though less than the 9.3 calculation that resulted from the free oscillation models). The drawbacks to the GPS approach were two-fold: first, only the horizontal component of global position was used to resolve a mostly vertical problem owing to the limits of vertical resolution using GPS methods; second, data for the model began five days after rupture initiation because of uncertainties introduced during reclamation of GPS surface sites.

 

Modern instrumentation provided a broad frequency spectrum of time-variable seismic data, free oscillation, and geodetic infor- mation not available in prior large earthquakes. Efforts to cope with the apparent complexity of the Sumatra-Andaman event have produced complicated models which try this layman’s ability and willingness to accept. The models still do not account for important observations.

 

 

 

The various models that I have summarized contain substantive flaws. The principal problem is that the seismic model does not account for 3 to 7 m (10 to 23 ft) of slip over 160 km (99 mi) of the Andaman and Nicobar fault segments. This was because slip in these regions occurred on a time scale beyond the normal seismic band and generated little seismic response. Geodetic constraints require two to three times more slip on this northern rupture portion than the seismic model provides.

 

The most important thing learned from the Great Sumatra- Andaman Earthquake is that the complexity and uneven aspect of fault slip in great earthquakes is at least as great as some feared, only now the complexity can be partly quantified. We have confirmed what was widely suspected: moment magnitude is not necessarily the main factor in generating tsunamis. Rapid vertical ocean-bottom displacement produces tsunamis. Only the southern one-third of this earthquake generated tsunami waves. There are apparently limits to the obliquity of plate boundary convergence beyond which rapid vertical slip is unlikely, or at least less, likely to occur. Subduction of older lithosphere, especially where strike- slip displacement transfer occurs (back-arc spreading), is less likely to produce tsunamis.

 

Three different and elaborate modeling approaches were taken after the initial Harvard CMT model was announced. These methods relied on teleseismic observations, free oscillation geophysics and GPS-based geodetic observations. None was fully satisfying. None was remotely successful without borrowing somewhat from at least one or both of the other methods, yet no concerted effort was apparently made to fully integrate the three approaches.

 

The Russian novelist Anton Chekhov was also a physician. He said that when a patient has many symptoms, there surely can be no cure. The complexity introduced by modern recording and modeling technology has, I believe, revealed many symptoms. The plate tectonic paradigm may have reached a limit and, while still vital and useful, may need to be modified. Plate tectonics may follow a path similar to classic Newtonian physics: still useful at many scales of observation but less useful in more precise and finite domains of time and space. Some scales of observation and prediction may be beyond the simple plate model. For all of its beauty and appeal, the plate tectonic paradigm has not resolved Wegener’s dilemma. Plate tectonics provided undeniable proof that Earth’s crustal plates move, but it did not resolve the mechanism concisely or convincingly.

 

In spite of its inherent conservatism, the process of scientific observation and interpretation produces revolutions from time to time (Kuhn, 1962). While the striking feature of scientific inquiry is its intent to discover what is already known or presumed in advance, sufficient new information is revealed to overthrow most ruling theories or paradigms at some time in their histories. A paradigm is, after all, nothing more than an interpretation of observations that a particular community acknowledges for a time as supplying the basis and foundation for its further practice. Once accepted, alternatives to the paradigm are strenuously resisted. In time, however, new paradigms are grudgingly accepted partly as a reason to conduct still more self-fulfilling inquiry!

 

Our measurement and modeling technologies for earthquakes and tsunamis have advanced tremendously between the Alaska Earthquake  of   1964  and  the  Great  Sumatra-Andaman Earthquake of 2004. Because of the punctuated occurrence of great earthquakes, this steady advance in technology seems to have overwhelmed our understanding of  tectonics almost overnight. While prediction is still not possible, probabilistic deduction becomes increasingly practical. What is needed at this moment is clearer understanding of where rapid, early earth- quake displacements are most likely to occur. New observations of convergence obliquity and mechanical behavior of different lithospheric plate ages provide productive pathways to pursue for better tsunami risk assessment.

 

 

 

I received another letter from Jakarta this week. In it, my colleague Dr. Busztin György Nagykövet tells me that Indonesia is proceeding with its plan to deploy an ocean buoy sensor tsunami warning system similar to the DART (Deep-ocean Assessment and Reporting of Tsunamis) array that the United States uses along our Pacific coast. He included a photo of people inspecting a giant buoy being deployed in Jakarta’s port. The system will be installed by the time this article is published.

 

I have previously described the obvious flaws in DART technology in its present form (Berman, 2005).  The United States is at considerable tsunami risk along the Washington and Oregon coasts for all of the reasons that the Indian Ocean was at risk one year ago. With what has been learned as a result of data acquired from the Great Sumatra-Andaman Earthquake of December 2004, the danger to our northwestern coastline is amplified.

 

I replied to Dr. Busztin that I feared my efforts to inspire the geo- logical community to benefit the Indian Ocean region had failed. The modern Trojans are bringing the enemy’s gift inside the city walls as I write. It has been just 10 months since the inevitable Indian Ocean disaster occurred and we continue to lament our inability to predict tsunamis.

 

We live on a restless and often dangerous planet. Earth science has made impressive progress toward understanding those dangers in the brief history of our science. Though we cannot yet predict when or precisely where natural disasters related to geological processes will occur, we can certainly say something about the probability that some regions are at greater risk than others.

 

For those who live near convergent plate boundaries in the Indian Ocean and Pacific Northwest of North America, earth- quakes and tsunamis will occur. We should do everything possible to use science and technology to provide early warning of these events. DART pressure sensing buoys may be part of the long-term solution but will not provide sufficient warning for coastal areas close to earthquake epicenters.

 

 

Modern digital seismic methods offer the greatest potential for early tsunami warning despite the lack of direct correlation between earthquake magnitude and tsunami occurrence. The Great Sumatra-Andaman Earthquake of December 2004 has given us many new insights into this relationship. We must apply this knowledge in the study of plate boundary convergence geometries and age of subducting lithosphere in order to better understand causes of the rapid slip that produces tsunamis. In the meantime, we should promote full and appropriate utiliza- tion of seismic technology, coupled with public awareness training, to minimize the effects of tsunamis.

 

 

 

 

 

Ammon, C. J., C. Ji, H-K Thio, D. Robinson, S. Ni, V. Hjorleifsdottir, H. Kanamori, T. Lay, S. Das, D. Helmberger, G. Ichinose, J. Polet and D. Wald, 2005, Rupture process of the 2004 Sumatra-Andaman Earthquake: Science, v. 308, no. 5725, p. 1133-1138.

 

Berman, A. E., 2005a, The Northern Sumatra Earthquake of 2004: Forty Years of Ignoring Plate Tectonics: Houston Geological Society Bulletin, v. 47, no. 6, p. 9-19.

 

Berman, A. E., 2005b, Letters from Jakarta:  Indian Ocean Nations Select a Tsunami Warning System: Houston Geological Society Bulletin, v. 47, no. 8, p. 9-19.

 

Berman, A. E., 2005c, The Debate Over Subsidence in Louisiana and Texas:  Houston Geological Society Bulletin, v. 48, no. 1, p. 47-54. Bilham, R., 2005, A flying start, then a slow slip:  Science, v. 308, no. 5725, p. 1126-1127.

 

Cacciani, A. and R. DeSalvo, 2005, Strange things you can do with under- ground GW interferometers:  California Institute of Technology Laser Interferometer   Gravitational    Wave   Observatory   Document LIGO-G050050-00-R, http://www.ligo.caltech.edu/docs/G/G050050-00/G050050-00.pdf.

 

Fellman, M., 2005, Sumatra earthquake three times larger than originally thought: Eureka Alert, AAAS, The Science Society,http://www.eurekalert. org/pub_releases/2005-02/nu-set020705.php .

 

Hanson, B., 2005, Learning from natural disasters:  Science, v. 308, no.5725, p. 1125.

 

Kuhn, T. S., 1962, The Structure of Scientific Revolutions: University of Chicago Press.

 

Lay, T., H. Kanamori, C.J. Ammon, M. Nettles, S.N. Ward, R.C. Aster, S. L. Beck, Sl L. Bilek, M.Rl Brudzinski, R. Butler, H.R. DeShon, G. Eckstrom, K. Satake and S. Sipkin, 2005, The Great Sumatra-Andaman Earthquake of 26 December 2004:  Science, v. 308, no. 5725, p. 1127-1133.

 

Park, J., T-R. A. Song, J. Tromp, E. Okal, S. Stein, G. Roult, E. Clevede, G. Laske, H. Kanamori, P. Davis, J. Berger, C. Braitenberg, M. Van Camp, X. Lei, H. Sun, H. Xu and S. Rosat, 2005, Earth’s free oscillations excited by the 26 December 2004 Sumatra-Andaman Earthquake: Science, v. 308, no. 5725, p. 1139-1144.

 

The Indian Ocean Disaster: Tsunami Physics and Early Warning Dilemmas, 2005:   EOS, Transactions, American Geophysical Union, vol. 86, no. 7 (February 15, 2005).

 

Thumping the Earth like a Watermelon, 2005: Astrobiology Magazine, http://www.astrobio.net/news/modules.php?op=modload&name=News &file=article&sid=1565&mode=thread&order=0&thold=0.

 

USGS Preliminary Earthquake Report, 2004: USGS National Earthquake Information Center, http://neic.usgs.gov/neis/bulletin/mag7.html.