Let me respond briefly to a few comments that present interesting analytical issues.
One such issue is whether, given the large, perhaps infinite, number of low-probability disaster scanarios, it is possible to defend against all of them without bankrupting a nation, or the world. It is not possible to defend against all of them; indeed, it is not even possible to imagine all of them. Ideally, they should be ranked by expected cost (probability times loss), and the expected costs compared with the costs of prevention. A disaster-protection budget could then be determined and allocated across the different disasters in such a way as to minimize the total net expected costs. The approach will not work perfectly because of uncertainty about both probability and loss with respect to many of the possible disasters (and also means and costs of prevention, in many cases). But there appears to be no better approach.
My book Catastrophe: Risk and Response lists a number of disaster scenarios, with some effort at estimating probabilities, losses, and precaution costs. However, to create a comprehensive ranking along the lines suggested above would require consideration of a number of additional disaster scenarios--including tsunamis, which I mentioned only in passing in the book. Such a ranking would be a worthwhile research project.
A related point is that poor countries may not have the resources to create tsunami warning systems or take other precautionary measures that wealthy countries could afford to do. I would rephrase the point as follows: the budget for disaster prevention will depend on the competing claims on public and private funds. The more urgent those competing claims, the rationally smaller will be the budget devoted to disaster prevention. It is really the same point that I made in my original posting in noting that value of life estimates are positively correlated with per capita income. This is not because the lives of poor people are worth less in some ethical sense than those of the rich, or even that poor people consider that their life is worth less. One must understand that the value of life estimates, while useful in cost-benefit analysis, are really just arithmetic transformations of estimates of willingness to pay to avoid a risk of death. If a person will pay $70 to avoid a .00001 risk of death, we divide the first number by the second and call the resulting figure of $7 million the "value of life." The reason for the transformation is not to make an ethical point but to facilitate comparison between the costs and benefits of precautions.
Obviously a person who can avoid starvation only by taking a risky job will demand less to assume the risk than a rich person would. That is rational behavior and if we forbid him to take the risk and force him to starve we won't be doing him any favor.
But to keep matters in perspective, although per capita incomes in the nations affected by the recent Indian Ocean tsunami are roughly 10 times less than the per capita income of the United States, the four countries principally affected--Indonesia, Thailand, India, and Sri Lanka, have an aggregate GDP of hundreds of billions of dollars. A tsunami warning system might cost only a few million dollars a year.