Biggest statistical analysis of historical nuclear accidents
Prof Dr Didier Sornette and Dr Spencer Wheatley, D-MTEC and a researcher at the University of Sussex, England, have carried out the biggest-ever statistical analysis of historical nuclear accidents.
FRS Principal Investigator Prof Dr Didier Sornette, Dr Spencer Wheatley, from the department of Management, Technology and Economics (D-MTEC) at ETH Zurich, and Professor Benjamin Sovacool of the Sussex Energy Group at the University of Sussex, England, have carried out the biggest-ever statistical analysis of historical nuclear accidents.
In the article 'A Rethink of Nuclear Risk Assessment' the authors suggest that nuclear power is a currently underappreciated extreme risk and that major changes will be needed to prevent future disasters.
The authors analysed more than 200 nuclear accidents, and – estimating and controlling for effects of industry responses to previous disasters – provide a grim assessment of the risk of nuclear power: The next disaster on the scale of Chernobyl or Fukushima may happen much sooner than the public realizes.
Their worrying conclusion is that, while nuclear accidents have substantially decreased in frequency, this has been accomplished by the suppression of moderate-to-large events. They estimate that Fukushima- and Chernobyl-scale disasters are still more likely than not once or twice per century, and that accidents on the scale of the 1979 meltdown at Three Mile Island in the USA (a damage cost of about 10 Billion USD) are more likely than not to occur every 10-20 years.
The studies, published in two papers in the summer issues of the journals Energy Research & Social Science and Risk Analysis, put fresh pressure on the nuclear industry to be more transparent with data on incidents. The articles have also been picked up by popular media, e.g. very recently by the German Spiegel Online(in German).
The team also calls for a fundamental rethink of how accidents are rated, arguing that the current method (the discrete seven-point INES scale) is highly imprecise, poorly defined, and often inconsistent.
In their new analysis, the research team provides a cost in US dollars for each incident, taking into account factors such as destruction of property, the cost of emergency response, environmental remediation, evacuation, fines, and insurance claims. And for each death, they added a cost of $6 million, which is the figure used by the US government to calculate the value of a human life.
That new analysis showed that the Fukushima accident in 2011 and the Chernobyl accident in 1986 cost a combined $425 billion - five times the sum of all the other events put together. However, these two extremes are rated 7 - the maximum severity level - on the INES scale. Fukushima alone would need a score of between 10 and 11 to represent the true magnitude of consequences.
Further, the authors emphasise that such frequency-severity statistical analysis of holistic consequences should be used as a complementary tool to the industry standard Probabilistic Safety Assessment, especially when aggregate consequences are of interest.
The open-source database of all 216 analysed events is available at https://innovwiki.ethz.ch/index.php/Nuclear_events_database, containing dates, locations, cost in US dollars, and official magnitude ratings. This is the largest public database of nuclear accidents ever compiled.