The 15 Costliest Nuclear Disasters and the Nuclear Risks of the Future

CC BY 2.0. Eamonn Butler

The names Chernobyl and Fukushima connote nuclear disaster. But do you remember Three Mile Island? Have you ever heard of Beloyarsk, Jaslovske, or Pickering? These names appear among the 15 most expensive nuclear disasters.

  1. Chernobyl, Ukraine (1986): $259 billion
  2. Fukushima, Japan (2011): $166 billion
  3. Tsuruga, Japan (1995): $15.5 billion
  4. Three Mile Island, Pennsylvania, USA (1979): $11 billion
  5. Beloyarsk, USSR (1977): $3.5 billion
  6. Sellafield, UK (1969): $2.5 billion
  7. Athens, Alabama, USA (1985): $2.1 billion
  8. Jaslovske Bohunice, Czechoslovakia (1977): $2 billion
  9. Sellafield, UK (1968): $1.9 billion
  10. Sellafield, UK (1971): $1.3 billion
  11. Plymouth, Massachusetts, USA (1986): $1.2 billion
  12. Chapelcross, UK (1967): $1.1 billion
  13. Chernobyl, Ukraine (1982): $1.1 billion
  14. Pickering, Canada (1983): $1 billion
  15. Sellafield, UK (1973): $1 billion

A new study of 216 nuclear energy accidents and incidents crunches twice as much data as the previously best review, predicting that

"The next nuclear accident may be much sooner or more severe than the public realizes."

The study points to two significant issues in the current assessment of nuclear safety. First, the International Atomic Energy Agency (IAEA) serves the dual masters of overseeing the industry and promoting nuclear energy. Second, the primary tool used to assess the risk of nuclear incidents suffers from blind spots.
The conflict of interest in the first issue is clear. The second issue may not be transparent to the layperson until they understand more fully how industry conducts the probabilistic safety assessments (PSAs) which are the source of the standard predictions of the risk of nuclear accidents. A PSA involves identifying every single possible thing that could go wrong, and assigning a probability that reflects the risk it will go wrong. Nuclear plants are then built with layers of interlocking safety mechanisms, that should reduce the probability to near zero that all of the failures necessary to result in a significant event could ever happen all at the same time.

It is a comprehensive and thorough method to help safety engineers reduce risks to levels that are acceptable relative to the benefits of the technology. It has certainly helped safety engineering make great strides in the effort towards 'zero accident' goals. However, the scientifically calculated risk probabilities from a PSA are only as good as the engineers' abilities to identify every single thing that could go wrong.

Every time some new thing goes wrong that wasn't thought of before, it is quickly integrated into the PSA and the assessment re-calculated and safety measures reinforced to again return the risks to the 'safe' levels. And industry keeps close track of everything that goes wrong, even when no accident occurs due to the layers of safety engineered in, which helps to fine-tune PSAs without the need for actual disasters. But every so often, a Chernobyl or Fukushima proves that our limitations outrun our technology for controlling the risks.

The new study, by researchers at the University of Sussex (England) and ETH Zurich (Switzerland), takes a different approach by submitting the data on events that have disrupted the nuclear industry to a statistical analysis. The report tracks the evolution of nuclear safety engineering that with the benefit of 20:20 hindsight in the wake of each nuclear disaster. It finds that nuclear accidents have substantially decreased in frequency, especially due to success of safety engineering in suppressing the "moderate-to-large" incidents.

But even with these optimistic trends, the report predicts that it is more likely than not that disasters at the extreme end of the IAEA scale will occur once or twice per century. Accidents on the scale of Three Mile Island have over a 50% probability of occurring every 10-20 years.

This may not spell the end of the nuclear industry though. One co-author of the study, Professor Didier Sornette, emphasizes that: "While our studies seem damning of the nuclear industry, other considerations and potential for improvement may actually make nuclear energy attractive in the future."

The team has made an open-source nuclear events database available online for further study of the issue. The papers are published in Energy Research and Social Science: Reassessing the safety of nuclear power and in the journal of Risk Analysis: Of Disasters and Dragon Kings: A Statistical Analysis of Nuclear Power Incidents and Accidents

For another perspective published on TreeHugger about nuclear power, see: The debate over nuclear power: An engineer looks at the issues