NUCENG said:
Actually the 2011 tsunami of 14+ m was an 800 to 1100 year event per the following:
http://coastalcare.org/2011/03/nuclear-plant-and-tsunami-risk-3000-years-of-geological-history-disregarded/
The seawall design basis event of 5.7 m turns out to be about a 100 year event.
It may be a thousand year event for the Fukushima Daiichi location, but it certainly is not one if we take whole japan into consideration.
While certain spots along the coast seem to attract giant tsunamis (let's take the village of Taro for example), it's not the case for the coast as a whole.
To summarize: Yes, it's true that a tsunami as high as in Fukushima may have been a thousand year event for THAT specific location. But giant tsunamis hit the japanese coast not every thousand years, but every few decades, as history proves.
Let's make a little made up math example:
Nuclear power plants run for, let's say, 50 years. The Fuku I Unit 1 reactor was 40 years old and should've been in service for another 10 years.
So the possibility that a thousand year event (tsunami) hits a plant during its lifetime is 1/20. But we don't have only one plant. Let's say we have ten along the coast. For every plant, the possibility to be hit by a thousand year tsunami during its lifetime is 1/20.
So overall, the possibility that ONE of these ten plants is hit by a thousand year event during its lifetime is 50%.
Conclusion: We absolutely HAVE TO secure plants against thousand year and even ten thousand year events. The more plants there are, the higher the possibility for one is to get "lucky". The german Biblis plant for example is secured against a flood two meters higher than a once in ten thousand years event.
A big german (conservative btw) newspaper made a little comparison between NPPs and the lottery:
Operating NPPs is like playing the lottery: The overall chances for "success" (a disaster) are very, very low. But the more players there are, the higher the possibility is for at least one to get "lucky".