Thursday, March 17, 2011

The Latest Disaster in Energy Production

-- Posted by Neil H. Buchanan

Last Spring, in the early days of the Deepwater Horizon oil disaster in the Gulf of Mexico, I wrote a FindLaw column (remember those?) and a related Dorf on Law post, arguing that the spill had provided new evidence about the choices between oil, coal, and nuclear power. Specifically, I argued that we tend to underestimate the costs of low-probability/high-cost events, precisely because they are low probability (and thus are rarely or never actually experienced). Even now, we really have no idea how much damage has been done by the spill in the Gulf, nor do we know whether some of the damage is still getting worse. Most people were surprised to learn that there actually had been at least one other oil spill of similar size, but even two or three such disasters do not provide us with enough evidence to know with any precision the true consequences of a major catastrophe.

Today, of course, the unfolding energy-related disaster relates to nuclear power. After last week's earthquake in Japan, there has been a serious leak of radiation from some damaged nuclear power plants. At this hour, the risk of an actual meltdown remains unrealized, but it has not been ruled out. At best, we are facing years of costly clean-up, with attendant uncertainty about how much damage is being inflicted on human and animal life locally and downwind from the crippled reactors.

In my FindLaw column last June, I wrote the following: "We now have direct evidence, in the form of the Gulf disaster, of how much more costly oil production is than we used to believe.” An angry reader responded via email: "How can a single data point be used to support the conclusion that oil is more costly? The fact that a one in a million chance happens does not mean that the original odds are invalid." Do we have to simply respond to this latest disaster by saying, "We knew that earthquakes happen, so there's no reason to change anything we're doing"? Germany's decision to shut down 7 of its nukes, while the E.U. plans tests of all 143 nuclear power plants in its 27 countries, demonstrates that policymakers are not taking that view. But should they?

Note that my argument was not a statement about probabilities, but about costs. The Gulf spill provided new evidence about just how bad a low-probability event could be, with more oil being released than anyone had anticipated, over a longer period of time, and with such pathetically inadequate attempts to mitigate damage until the well was (at long last) capped. Even if the only new evidence was that the costs of such a rare spill are higher than had been believed, therefore, the calculus of low-probability/high-cost events would have changed. Estimates of the expected risk-adjusted cost of deepwater drilling had been too low, and responsible scientists would change their calculations appropriately.

Even on the probability side of the computation, however, the story changes after we observe such an event. Note that my reader's objection was based on "[t]he fact" that the disaster in question was "a one in a million chance." If we knew that the event really was a manifestation of a one-in-a-million random process (or any known statistical probability), then of course we would not need to adjust our expectations in response to observed outcomes. Not only would we not be surprised if we threw snake-eyes (a one-in-36 chance with fair dice), but we would not be surprised if we threw snake-eyes fifty times in a row. That is no more nor less likely than any other sequence of fifty specific results of throwing fair dice.

Would any sensible person, however, not reconsider whether the dice are fair, after throwing snake eyes fifty consecutive times? Even when we have fairly strong reasons to believe that a process is random, low-probability events are good reason to revisit, and possibly update, our beliefs about the facts.

The issue, moreover, is not that we are surprised that there was an earthquake in Japan. The low-probability event was the degree of damage that the reactors sustained due to the earthquake. Precisely because Japan is prone to earthquakes, that country has been at the forefront of earthquake preparedness. As an article in The New York Times put it this past Sunday, this disaster "showed the limits of what even the best preparation can do." The article quoted one seismologist: "I'm still in shock." Whereas the BP spill last year showed that all of the oil companies (with at least the passive consent of the Bush and Obama administrations) had failed to prepare for such a disaster, the recent events in Japan showed that the consequences of a large quake (and resulting tsunami) were simply not conceivable, even to the very conscientious planners in that country.

The response in Europe (and, to a lesser degree, here), therefore, is not an irrational overreaction to some one-off event. We had become accustomed to viewing the probability of major damage to nuclear plants as near-zero, because we take (or believe that we are taking) extraordinary precautions. It is now completely rational to stop and ask just how much we really know about the effectiveness of those precautions. If our confidence in them has gone down -- as it should -- then the expected probability of an earthquake-related disaster should have gone up (even if the expected probability of earthquakes themselves is unchanged).

This adjustment in our estimates of the probability of low-probability events must be paired with adjustments in our beliefs about the costs. What would be the real cost to society of a Chernobyl-plus event in, say, California? Is the cost of ten million deaths simply the cost of one death times ten million, or is there something much worse about so many deaths? (Or is it less bad, because it is all so numbing?) I strongly suspect that, when U.S. politicians talk blithely about the safety of U.S. nuclear power, they view the low-probability estimates as an excuse not even to think about the costs that such a disaster would entail.

The initial response to the disaster in Japan is, therefore, entirely appropriate. Far from being an overreaction to known risks (with admittedly sad consequences), efforts to reconsider the costs and risks of nuclear power are entirely responsible. The only legitimate concern is that such reconsideration will be brief and superficial. I will have more to say about that in a future post.

2 comments:

Jamison Colburn said...

Neil,

Thanks so much for your attention to this point.

NRC absolutely should be learning from this data point. "Natural experiments" of the kind show us what we may not have *noticed* before: a coincidence of risks that can combine to form unprecedented failures in complex technological systems. Strictly speaking, the failures at Fukushima stem from a loss of operating power -- not a defective core vessel design as many are now emphasizing. On the drawing board, the immediate cool down of a core that is stabilized (98.5% of its heat) looks conclusive. But it turns out that 1.5% of core heat compounded over hours, days, and perhaps even weeks, can be too much.

So shouldn't someone from the press ask NRC whether it has collected information from each of the US installations on the following:

(1) exactly how robust their (largely generator-based) back-up power systems are; and

(2) whether the redundancies in their back-up power systems correlate in any way with the identifiable risks (seismic or otherwise) they face in their particular sites.

The fact that such an analysis was prepared at the time of licensing is no guarantee of its sufficiency today (in some cases, decades later). Correlations of this kind should assure us a lot more re: the relative vulnerabilities of installations in the US to the (in Charles Perrow's terms) "common mode failures" that arise in their operations than should the presence or absence of any particular operating system(s) or mechanism(s) at an installation. Note, too, that NRC's generic seismic risk estimates were updated in November 2010 -- and many of the estimates changed substantially.

NRC could have this information from each licensee overnight under 10 C.F.R. Part 50 -- which speaks of "probabilistic risk assessment" in clear terms. Bayesian updating in the real world!!

And supposing NRC has obtained information like this, the question remains: why hasn't some synopsis been released to the public?

All the best,
Jamie

chengnuo said...

They build up characters and their gold stash, which will also ensure you the Cheap WOW Gold Eu
. What’s more, we usually launch big promotion, to provide the Buy WOW Gold EU to you, you can use less money to get more Tera Gold by using the special discount code. And the more you order, the more bonus you will get, rather competitive.