We've heard a lot lately on the imminent earthquake about to hit Indonesia. The threatened zone is on the southern end of the rupture for the 2004 giant earthquake.
We all know about the big earthquake. It ruptured from Sumatra, practically up to Thailand! On the other hand a mere 8.7 hardly ruptured much at all, and did not cause a major tsunami. Such is the power of a log scale!
In many parts of the world, it is observed that earthquakes can march along a major fault. This happens in Turkey along the North Anatolian Fault, for example. Now, a study of beached corals shows that this also happens for this fault system. As well, it is confirmed by historical observation of the last sequence.
Thus, we have the whole Sunda Trench open to us for a major earthquake.
This, of course, leads us to attempt to mechanically model the whole system. If the upper section moved, it must lead to increased stress on the lower section. If you use standard finite elements, then it shows something like this.
Note the very rapid relative motion of the converging plates! This leads to a full cycle every 200 years or so. Upon reading all of these articles, I became more interested in the mechanics of the subduction zone. It turns out to be vastly complicated, with many interesting variations.
We all have a general idea that the oceanic plates are 'conveyor belts' forming at ridges, and sinking at subduction zones.
What is the driving mechanism for this machine? Most people now believe that it is 'slab pull' which comes when the tired old slab becomes denser after significant cooling, and sinks back into the asthenosphere. This pulls the plate. But all these physics people aren't engineers which know You Can't Pull a Piece of Crap (the extension of You Can't Push on a Rope). In other words, for a pull, you need tensile strength, and large scale rock doesn't have this!
But, I'm not fighting these artsies, and we'll go with slab pull for now. At the subduction boundary, the descending plate can do all sorts of wonderful things. The plate is all wet and juicy from the ocean sediments. In order to generate volcanoes, it has to descend to 100 km, where the water, and other gases get cooked off. In some zones, the plate is too bouyant, doesn't get to 100 km, and does not produce volcanoes. Some zones, the plate is tired, and descends rapidly. They can plunge really deep, and are most likely the source of hotspots.
Still, the mechanism for the very large earthquakes is the same for all varieties. The descending plate 'sticks' for a while, and then releases. One side goes down, the other goes up. As far as buildings go, these are fairly slow earthquakes, and do not do much damage, but if they are long enough, they can generate huge tsunamis.
Look at the maps of the rupture areas, and visualize a very thin film of water, much like water in a cookie sheet. Only the rupture zones that are much longer than their width can generate a 'sloshing disturbance', which is a tsunami. The shorter sections can not propagate a slosh, but can produce a large local tsunami, if there are underwater landslides.
In summary, we really have to go with the history of any given subduction zone. In this case, I believe we will really see some action in the next 10 years. Although Padang is only a few metres above sea level, they are really getting prepared for a tsunami, and have many tall buildings which could act as refuges.