I continue to support the expansion of nuclear power generation in the US, but found this comment (from "Nikster") interesting:

There have been so few nuclear accidents, and each one has the potential for so much damage, that death statistics are meaningless. You don't have a large enough sample. The next accident might kill Millions - what then? Is that "whoops!"-time, like in Fukushima?

Let me put this in a framework of reference - I am not a statistician but I do understand enough about it to know you usually simply assume a normalized distribution - the bell curve. And that you need a sample of > 100 at the very least to make any kind of prediction. That means in order to predict the death rate from the largest possible nuclear accidents over time, you need around 100 such accidents. So far we've had two, Chernobyl and Fukushima. Many things went wrong in these accidents - but many things also went right. Both were very bad (one still ongoing), and both could have been much worse.

One thing I am sure about: I've never heard of a solar roof panel that required a 30km evacuation radius. US scientists recommend 80km - that's 50 miles, and several million people. Fukushima could have been worse by orders of magnitude.

The interesting mesure is not how many people have died, but how many could die if everything goes wrong. By that metric nuclear power plants are an unacceptable risk. Murphy's law.