A recent issue of EOS, the weekly newspaper of the American Geophysical Union, has a very compact but informative figure in it (p. 227 of the 5 July 2011 issue). (Sorry, but it appears I can access the full figure only from my USGS desktop. I'll try to describe it:). It helps the average science-interested individual figure out how size of a fault-plane can tell you how big an earthquake will be. The diagram also warns us potentially how large a tsunami run-up (the wall of water you will meet at the coast shortly afterwards) might* follow one of these monsters.
It’s been known for a long time that the size of an earthquake correlates fairly well with how much surface area is torn in the formerly “stuck” rock on a fault surface.
Some brittle-vs-plastic-rock basics:
If you have a vertically-oriented fault like the San Andreas, the vertical dimension for the fault “tear” can be only about 10 kilometers - below that depth the rock is so hot and pressurized that it turns plastic and doesn’t “break”. A magnitude 7.8 event is a Very Big One for the San Andreas. Even if it rips horizontally for 200 kilometers, it can’t get enough surface area torn to be bigger than that.
An ocean-floor subduction fault, however, is a different cat. These things dips shallowly... almost flat in some places. You can therefore get a lot more “down-dip” rock breakage or “tear” with this kind of fault before you get down to the “plastic” zone.
The Tohoku earthquake off northeast Japan in March, 2011, is calculated to have been in the magnitude 9+ range. That’s 10 times more energy released than a magnitude 8 event, and close to 25 times more energy than a “piddly” San Andreas 7.8 event (like the one that destroyed San Francisco in 1906).
The EOS diagram (click here) lays this all out graphically:
A 60 km by 120 km tear, with 5 meters slip along the fault-face, will give you a magnitude 8 event - and a 10 meter Tsunami run-up. That's a wall of water over 30 feet tall.
A 200 km by 500 km rip, with 10 meters slip (what happened off the Sendai coast of Japan), will give you a magnitude 9 event - and a Tsunami run-up of up to 20 meters. This explains the monster wall of water hitting and destroying the Fukushima Daiichi nuclear plant, and destroying villages MILES inland. This Fukushima nuclear plant debacle is now looking more and more like the Chernobyl disaster that depopulated a lot of the Ukraine in 1986 (135,000 people were permanently evacuated from their homes in a little over a day).
It’s been known for a long time that the rate of subduction - how fast a continent is over-riding an oceanic floor - seems to correlate with the frequency of volcanic eruptions inside the continent’s edge. Mount St Helens has erupted twice in the last 31 years (but no other eruptions have occurred elsewhere in the Cascades since 1917's Lassen event). The Juan de Fuca plate "only" moves about 2.5 cm per year towards North America.
However, at least one - and generally several - volcanoes in Kamchatka are erupting all the time. The Kamchatka Peninsula is moving eastward over the Pacific plate at over 8 cm per year - much faster than North America is moving relative to the Juan de Fuca plate. More plate gets subducted down to the mantle, faster, and this means more partial melting takes place. Think lava-lamp with three times the heating coils.
What are subduction-related volcanoes, anyway? Examples are Mount Rainier on Seattle's skyline, Mount St Helens near Portland, and Mount Shasta in Northern California: the Cascades range. Their equivalents elsewhere: Bezymiani, Sheveluch, Alaid and a boatload of other volcanoes in Russia's far east Kamchatka Peninsula, Mt. Fuji and Mt. Unzen in Japan, Mt. Pinatubo in the Philippines, and Krakatau ("east of Java") and Merapi in Indonesia.
Does this subduction rate thing also hold for the frequency for large earthquakes?
The same diagram suggests that subduction earthquake frequency and size don’t seem to correlate with how fast the plates are moving. This is probably because of complex fault geometries, and how often so-called “silent” or “slow” earthquakes take place (they tend to redistribute accumulating fault strain).
Bottom line: the last huge subduction earthquake on the Pacific Northwest coast happened in January 1700 AD (there's another very eye-opening figure in that link). At least 7 of these magnitude 8+ events have occurred in the last 3,500 years, but that means nothing in terms of predicting the next monster.
The Next Big One could occur tomorrow or 200 years from now.
What can you do about this? If you live in Kansas, you need not worry.
If you live in Portland or Seattle, however, it would be a good idea to earthquake-reinforce your house... and buy earthquake insurance. The problem is that if a Cascadia earthquake hits, the damage could be so massive and so far-reaching that it could wipe out many North American insurance companies. Hurricane Andrew (which slammed south Florida in 1986 - the same year as Chernobyl) caused about $24 Billion in damage, and even with the modern practice of spreading risk, it stretched some insurance companies to the limit.
You CAN, however, steadily build up a year’s supply of food, and develop some sort of water storage system. Again, this is as much for your neighbors as for yourself. You are your brother's keeper.
* Depending on fault geometry, there could possibly be only a small tsunami. Knowing the geometry ahead of time makes all the difference.