Why The Nuclear Industry Cannot Predict The Past

The title of the post is not that original1 and you will have to bare with me to understand it.

I read a yet another news report about a nuclear power (Pilgrim in the USA) plant unexpectedly shutting down2. However, everything worked as it should, there was no radioactive releases and there was no need for concern – or is there?

First of all we have to look into how the probability of an accident that we are going to be concerned about is calculated. While the nuclear industry often tells us that things are perfectly safe and certain accidents ‘cannot happen’ this is not true. What they mean is that the risk of certain accidents is extremely unlikely3.

Let us consider a hypothetical accident. First of all event A happens (e.g a leak at the spent fuel pond). To progress any further event B must happen (e.g. the water level detector fails to work) and then event C must happen (e.g. radiation alarms are ignored by the workers) which leads to event D (e.g a fuel pond fire with significant offsite release of radiation).


The probability of event A happening is PA, event B is PB etc. Say that you want the risk of a significant offsite release to be 10-6 per year (sometimes called ‘one in a million’ year event but this is misleading4). You can then work back – if you do not meet the required probability then you could reduce PB by adding more water level detectors in case one of them fails etc.

In nearly all the scenarios including the final event where no release happens (probability of 1.00-PD) there is no offsite release and ‘nothing to be worried about’.

Predicting the Past

Even if we never get to event D how many times have we got to A, B or C? Let us take for example the leak at Sizewell A in 20075. If we had asked Sizewell A in 2006 what the probability of this event what would they have said? The general public would probably be told that it could ‘never happen’. If we pressed them we might be told something like it would happen once in 1000 years (they might not since I have pressed them and they have refused to say). Yet it happened – probability of 1.00!

Of course we  could have just been unlucky and the 1 in 1000 year event happened after 40 years. If we had the various probabilities then we could work out what the probability of it happening by chance is and we could use this to do a significance test (see Probability and Significance) on their risk model. Would we have confidence in their risk model – I doubt it.

What they usually say is ‘we have fixed the problem and it won’t happen again’ – what they really mean is that they have set it probability back to a ‘1 in 1000 year event’.

Back to Pilgrim

We are told that this ‘one off’ event at Pilgrim is nothing to worry about. However, it is not a ‘one off’ event and is the seventh outage due to a problem in two years6. If type A events are happening so often there is plenty to be worried about.

Let us take another example. In March 2010 Sizewell B was shut down due to ‘excess moisture in the containment building’7. A subsequent report by the Nuclear Installations  Inspectorate (NII) disclosed that their was a leak of 4.5L (about 1 gallon) per minute due to a failure around a heater in the pressuriser8.

This report also disclosed that such failures have happened ten times in ten year. Was this what was expected when the plant was built?

When these ‘minor’ incidents happen it does not fill me with confidence that ‘we were lucky this time’. If the nuclear industry expects these incidents to happen once every year or once every ten year then why do they not say so. However, most of these incidents are unexpected i.e. they do not match the probabilities they expect. If their analysis cannot explain what has happened in the past what confidence can we have that they can avoid a disaster in the future?

Experts in both the nuclear and finance industries assured us that new technology had all but eliminated the risk of catastrophe. Events proved them wrong: not only did the risks exist, but their consequences were so enormous that they easily erased all the supposed benefits of the systems that industry leaders promoted.

Joseph Stiglitz Senior  former Vice President and Chief Economist of the World Bank9

1 The title is based on Mike Godfrey’s “Why Economists Cannot Predict the Past.” (https://sites.google.com/site/michaeldgodfrey/economics/SET_slides.pdf?attredirects=0)

2 Valve leak at Pilgrim nuclear plant forces shutdown; no safety threat, officials say, The Boston Globe, January 21 2013 (http://bostonglobe.com/metro/2013/01/21/valve-leak-pilgrim-nuclear-plant-forces-shutdown-safety-threat-officials-say/kQa4DvsFplgiC9GyFcFGoN/story.html)

3 Even the possibility of a nuclear power plant creating a massive nuclear explosion is not impossible. Since nuclear fission is a random event then thereis a finite possibility that enough atoms will undergo spontaneous fission at the same time to produce a ‘nuclear bomb’. I have not attempted the calculation but I expect that the probability is so low that it is unlikely to happen within many times the age of the universe.

4 You have to be careful if probabilities. A 1 in 100 year event does not mean that an event will definitely happen in 100 year or every one hundred year. It means that the probability in any one year is 1/100. It might not happen or it might happen once or twice or three times. To work out the probability you will have to work out the probability that it happens the in the first year, then not the first year but the second year etc. It easier to work out the probability that it does not happen. There is a .99 (99 percent) probability that it does not happen in any one year. In one hundred years the probability that it does not happen is .99100 = 0.37. Therefore the probability that it happens once in 100 years is 1.00 – 0.37 = 0.63 (63%).

5 Sizewell nuclear disaster averted by dirty laundry, says official report, Guardian 2009

6 Service returns to Pilgrim as questions linger, Boston Globe, January 24th 2013 (http://www.bostonglobe.com/business/2013/01/24/pilgrim-nuclear-plant-back-online-questions-linger-plymouth/k25udeba1UpwJlvQrPUa4M/story.html)

7 Sizewell B outage will extend into summer, Nuclear Engineering International, 11 June 2010 (http://www.neimagazine.com/story.asp?sectionCode=132&storyCode=2056609)

8 Justification for Return to Service from forced outage 43 (FO43). NII, 2010 (http://www.hse.gov.uk/nuclear/foi/sizewell-b-fo43-par.pdf)

9 Gambling with the Planet, Joseph E Stiglitz, Project Syndicate 6 April 2011 (http://www.project-syndicate.org/commentary/gambling-with-the-planet)


Leave a Reply

Your email address will not be published. Required fields are marked *

Captcha: * Time limit is exhausted. Please reload CAPTCHA.