Subduction zone segmentation and controls on earthquake rupture

The Sumatran earthquake of December 26th 2004 was the second-largest earthquake on record. The growing concentrations of population in regions prone to great earthquakes makes it a matter of urgency to study the processes that control these earthquakes. The Sumatran earthquake is the first to which modern geophysical tools can be applied, so offers a unique opportunity for such study. Most great earthquakes (magnitude 8 and larger) take place where two plates converge; such regions lie mostly under water, which makes them difficult to investigate, and which means that the hazard of tsunamis is added to the dangers of ground shaking. In the December 26th 2004 earthquake, the Australian and Eurasian plates slipped towards each other by up to 25 metres, on a fault that runs for 1200 km along the earth's surface. An obvious question is: Why was this earthquake so large? But perhaps the question should be: Why wasn't this earthquake larger? because, in March 2005 an adjacent 400 km of the plate boundary slipped in a second huge earthquake. All plate boundaries are divided into segments - pieces of fault that are distinct from one another, either separated by gaps or with different orientations. The boundaries between segments provide barriers that limit how far an earthquake can spread. A large earthquake may rupture a whole segment of plate boundary, but a great earthquake usually ruptures more than one segment at once. However, we do not know what determines whether an earthquake stays within one segment of plate boundary (and remains relatively small), or jumps across barriers between segments (to become a great earthquake). The Sumatran earthquakes give a unique framework to attack this problem. Detailed analyses of the seismic waves radiated by the Sumatran earthquakes give accurate locations for the barriers controlling the sizes of the earthquakes. The December 2004 earthquake started close to Banda Aceh and spread almost entirely northwards. Although 25 metres of slip occurred near the southern end of the rupture, almost no slip spread to the south; clearly, an important barrier here prevented the earthquake spreading to the next fault segment. In March 2005 another great earthquake occurred within this next segment, and spread southwards until it was stopped by a barrier at its southern end. We will conduct detailed geophysical surveys of the plate boundary to determine the nature of these two barriers. Large-scale (1-20km resolution) images of the plate boundary will be obtained in a combined land-sea experiment using air-gun explosions to bounce seismic waves off structures inside the plate boundary. In a longer-term experiment, seismometers left on land and on the sea-bed for several months will pick up the seismic signals from distant earthquakes. These waves, travelling upwards through the earth to the array of seismometers, can be used (in a fashion similar to CAT scanning) to form 3-dimensional images of the deeper parts of the crust and upper mantle. At the same time, new techniques will be developed to give more precise pictures of the distribution of slip in the two earthquakes, in order to link the static structure of the plate boundary to the dynamics of the earthquakes. Cores from the seabed will show when large earthquakes have occurred on these faults in the past, and whether the segmentation seen in 2004-5 was similar in the past events. In addition, we will collect evidence for the ways in which fault slip affects the seabed and generates a tsunami. The results will be significant both locally and globally. It is important to compare the barrier between the 2004 and 2005 earthquakes with the barrier at the south of the 2005 earthquake, because the plate boundary immediately to the south slipped in 1833, causing a devastating tsunami. More generally, the results will have implications for other convergent boundaries, such as those beneath Japan, and on land associated with the Himalayas.