More and more data comes to light after the massive earthquake in Japan last week. As Callan Bentley points out in his blog, the vertical component of the displacement vector field shows a general subsidence in all the East coast of Honshu (Japan's main island). The maximum subsidence is 75 cm, while the horizontal displacement has been of up to ca. 4.4 m (eastwards).
At first sight, it could look awkward, as the affected part of Japan is in the Okhotsk Plate, which is the overriding plate, while the Pacific plate is the subsiding element. Shouldn't the overriding plate uplift, instead of subside? Let's analyse it more in detail.
A megathrust fault is the tectonic boundary between a subsiding plate and an overriding plate. Megathrust faults are the causing element of the majority of most powerful earthquakes in the Earth. Probably, the most famous megathrust is beneath Cascadia in North America, and you can find most of the information you need about that in this great site: Giant earquakes of the Pacific NW.
Coming back to Japan, which is part of the Eurasian plate, it overrides the Pacific plate over a megathrust. This is the fault that, with its slip due to elastic rebound, caused the 9.0 Mw earthquake some days ago. But of course, the slip occurs suddenly, after accumulating energy for certain time.
As the following figure shows (taken from here), the overriding plate develops a bulge -due to strain build-up- when the upper plate locks over the subducting plate. Behaving elastically, the bulge grows as shortening continues, and the overriding plate continues accumulating strain in the bulge area. At some moment, the contraction is so strong that it exceedes the hold of friction along the fault, and an earthquake is triggered. The strain build-up relaxes, and the previoulsy uplifted region subsides. The cycle starts all over again.
This is what has happened in Japan. An example may be to imagine ourselves pushing a mat, with our feet, against another mat. One of them will form a fold in the rim, until its resistance is broken, it slips over the "subsiding" mat, and the fold disappears.
In 1700, that happened in Cascadia (http://blog.lib.umn.edu/stgeorge/geog5426/2010/11/13/Jacoby%20Geology/201997.pdf), when a massive earthquake caused the subsidence of a region, where many trees drowned and just a handful of them survived... for telling us through their ring what happend with the topographic base level. (Photo from here)NASA has published two MODIS images, showing the coastline in Sendai before and after the eartquake. It is still early to estimate the new coastline, as in the images of this week many lands near the coastline are still flooded by the tsunami wave, and will probably not be submerged in the near future.