We know the basic events of 26 December 2004: a giant earthquake beneath the Indian Ocean generated a devastating tsunami. But geoscientists are still learning about processes initiated during the earthquake.
Wherever tectonic plates meet, potential trouble exists for those who live nearby. The largest earthquakes threaten people who inhabit regions where one plate plunges beneath another. Such subduction boundaries have been gaining increasing attention as a result of new observations. Episodic tremor and slip have now been observed in several places, but are seen most clearly along the northern Cascadia subduction boundary1,2 off western North America. And there was the event that literally shook the world — the huge and deadly Sumatra–Andaman megathrust earthquake of 26 December 2004, which generated the large tsunami that left more than 250,000 people dead or missing3.
While geoscientists were monitoring several subduction zones thought more likely to produce a megathrust earthquake, the 26 December event occurred in a region of relatively sparse modern instrumentation, in a place where few, if any, would have forecast such a large earthquake. Fortuitously, however, several research groups had recently completed Global Positioning System (GPS) surveys in northern Sumatra and along the Andaman and Nicobar islands. The results of these and subsequent surveys are documented by Subarya et al. on page 46 of this issue4. They present a new view of the slip associated with the Sumatra–Andaman earthquake sequence (Fig. 1) that is similar to earlier seismic and GPS results5,6. But the new work is based on near-field GPS measurements and observations of geodetic uplift that are sensitive to both the rapid fault-slip of earthquakes and the slower slip that often follows (afterslip). The sparse continuous GPS observations allow our first, albeit approximate, separation of the rapid seismic and slower subsequent slip in a giant megathrust earthquake.
Subarya et al. integrate observations from several sources — a near-field continuous GPS station on northern Sumatra, a set of more distant regional continuous GPS stations, several reoccupied GPS sites, and measured patterns of coral uplift. They show that most of the recent plate motion occurred along shallow parts of the plate boundary (the upper 100–150 km). The authors construct models that include a large area with more than 5 metres of slip, peaking at more than 20–30 metres of slip off northern Sumatra (see Fig. 1 on page 47), the source region for the devastating Aceh tsunami. In the models, slip extends into the northernmost of the Andaman Islands, as surface deformation was observed as far north as Preparis Island, Myanmar. The analysis also shows that the giant earthquake (magnitude 9.1–9.2) was followed by weeks of slower deformation that accompanied thousands of aftershocks (Fig. 1). The results show that we have much to learn about how deformation is accommodated across these zones of dramatic plate convergence.
One of the most interesting aspects of the new geodetic models is the large amount of post-seismic slip, estimated at approximately 30% of that in the initial mainshock and corresponding to slip magnitudes comparable to those produced by an earthquake of magnitude 8.7. The substantial afterslip is similar to that seen with the 1960 magnitude-9.5 Chile earthquake sequence7,8. But unlike the afterslip in 1960, which is thought to have occurred at greater depth than the seismic rupture, the 2004–05 afterslip seems to have been located in the shallower regions of the plate boundary, adjacent to the seismic slip that occurred abruptly on the morning of 26 December.
The emerging picture is that of a frictionally heterogeneous plate boundary that contains large regions prone to slip during earthquakes and regions that slip more slowly, although both participate in megathrust earthquake sequences. Earthquake scientists have known for some time that mapping the regions prone to sudden slip and understanding how they evolve is essential for estimating the hazards posed by a particular fault. Understanding the correlations between the regions of large seismic slip and variations in upper-plate morphology and strength9,10,11 may help to illuminate the frictional condition of the plate interface at depth. As work to integrate the available data continues, and new data on the Sumatra– Andaman plate-boundary structure are collected, perhaps we can test some of these ideas.
Many issues require deeper investigation. On 28 March 2005, a large rupture (Fig. 1) continued the sequence begun in December southeast along the Sunda arc. But why did this rupture occur 90 days after the mainshock? Subarya et al.4 identify a particular fault in the Indian Ocean floor that possibly inhibited continuation of the initial rupture in that direction. But what process changed the situation during those 90 days — afterslip-induced fault loading, migrating fluids, an expanding deformation front, or something more complicated? And what caused the rupture to terminate in the north? Did it simply run out of strain energy as a consequence of plate-motion geometry? Or was something more complicated involved — perhaps related to the subduction of a large load of sediments from the Bay of Bengal, or an undiscovered feature of plate interactions in the region? What tectonic processes are operating in the region between this subduction zone and the Himalayan plate boundary that looms so close to billions of people?
Finally, will the sequence continue farther south? The occurrence of large earthquakes in the eighteenth and nineteenth centuries shows that large ruptures can hit this region, which includes southern Sumatra and Java (and so coastal populations of millions of people). If activity does migrate southwards, we will be better prepared with more adequate continuous GPS and seismic measurements, a result of international efforts. And at least one network is openly sharing data as they are recorded12, which is necessary for timely and innovative integrations of the data. Perhaps these data will help in understanding the nature and implications of episodic tremor and slip before the next large subduction-zone earthquake occurs.