In the debate about nuclear power, some people argue that the number of fatalities that can be attributed to the coal or gas industry is much higher than fatalities caused by nuclear power – and conclude that this argument successfully makes the case for the continued use of nuclear reactors.

I find the discussion about which of the dirty technologies (as we actually can and need to get rid of both fossil and nuclear) is largest killer quite indecent and do not want to fuel it by throwing in more abstract figures that never respect the human lives behind them. But for the purpose of getting some clarity around this argument, let’s look closer at those claims.

The perhaps most authoritative comparison of fatalities appeared in a number of Organisation for Economic Co-operation and Development (OECD) papers during last decade: “Environmental and Health Impacts of Electricity Generation” published by International Energy Agency in 2002, “Risks and Benefits of Nuclear Energy" published by Nuclear Energy Agency in 2007, “Comparing Nuclear Accident Risks with Those from Other Energy Sources”, also published by Nuclear Energy Agency 2010, and “Energy Technology Perspectives 2008“ by International Energy Agency. The arguments on fatalities from those papers have been frequently quoted elsewhere.

The first major problem with the fatalities argument that if one takes the effort to look closely, all the relevant figures and graphs in the OECD publications actually refer down to one single collective of authors, which is the Paul Scherer Institute in Switzerland. In references you will find Paul Scherer himself, or his colleagues R. Dones, U. Gantner or S. Hirschberg. Now, having such an important argument based on one single source is not exactly scientific, is it?

Let’s look further. Here one fancy version of their graphs, the one that has been reproduced in most influential International Energy Agency (IEA) paper, “Energy Technology Perspectives 2008”:

This chart plots frequency of accidents and their fatalities. Therefore, the closer to the left bottom corner the technology appears, the safer it is – and vice versa, the more towards top right corner, the more risky.

You need to note two things that are apparently wrong with this chart. First, it talks about the OECD – which allows authors to entirely exclude the worst nuclear accident - Chernobyl in 1986 - from the picture. If we were to plot Chernobyl, it would appear at the frequency of once in 8,000 GWy (accumulated nuclear generation history by now), and fatalities in the range between 9,000 (according to World Health Report (WHO) report in 2006, future victims in Russia, Byelorussia and Ukraine) and 33,000 (modest estimate of overall potential death toll based on collective dose, see for example TORCH report from 2006). The graph would then actually look like this, with the red bar showing more realistic range for nuclear power’s record, making it clearly the most hazardous technology:


The second thing that is terribly wrong with the original chart, is that while it pretends to include nuclear power, it is using double standards to plot it. Contrary to other technologies, where the risks are taken from real world experience, nuclear risks are based on PSA. This is so-called Probabilistic Safety Analysis and is nothing more than a theoretical modeling of reactor risks.

It is acknowledged as a method to identify weak points in the technology, but even the nuclear industry says it is not good tool for coming up with reliable figure of the overall chance of heavy reactor accident. We have seen this method going totally wrong in the case of the US Space Shuttle program, where the engineers estimated a chance of accident as 1 in 100,000 but in fact two of the four Shuttles in the fleet ended up meeting disaster (a great insight by Richard Feynmann can be found here).

We can also see how wrong this method went for the nuclear power sector globally: while the industry is assuring us that the chance of a heavy accident involving reactor damage is in the order of 1 in 100,000 years, in fact we have seen at least five such accidents (one reactor at Three Mile Island in 1979, one reactor at Chernobyl in 1986, and three reactors at Fukushima in 2011) during an accumulated 14,000 reactor-years of experience – this brings the empirical figure of more than 1 in 3,000 chance. Nevertheless, OECD publications use this theoretical calculation, done for a single Swiss reactor (sic!), to make generic conclusions on hazards. This is already a third major methodological mistake that makes the original IEA charts figures heavily biased and unreliable.

So the recipe for cooking up false conclusions about nuclear safety is as easy as this:

1)       Take your data from one single institute in all publications (there is a lot of referring in circles but finally it all ends up with the Paul Scherer Institute);

2)       Ensure sure you find a methodology that allows you to exclude the worst nuclear disaster in the history from your statistics (such as talking about OECD only); and

3)       Use double standards in comparing technologies (theoretical modeling for nuclear versus historical record for other technologies).

While if plotted properly, you will find out that nuclear energy probably still presents the largest risk to lives of the known energy technologies. But there is also an important qualitative difference here. If we talk about accidents in the coal or oil industries, those risking their lives are mostly workers directly involved in the operation. In that sense, it is certainly their choice by choosing the job and they also have the means to mitigate and control the risks they are taking. In the case of nuclear power, most of the risk is imposed on a population that has no information, no expertise, no training, and most importantly no choice as they are denied from having the power to decide if they want a nuclear reactor in their vicinity.

Apart from the International Energy Agency, more graphs and figures have been appearing widely, such as this one


This is coming from yet another angle, which is not only looking at fatalities from accidents but also the health impacts of routine air pollution etc. That this approach gives even much more space for voluntary speculation, it can be illustrated by how much the figures differ if they come from different sources.

Take for example the above mentioned blog and source data for those impressive boxes: they attribute 161 fatalities per one TWh (billion kilowatt hours) of electricity to coal, 36 fatalities per 1 TWh to oil, 4 fatalities per TWh for natural gas, and 0.04 fatalities per TWh to nuclear. That would indicate that nuclear is at least by three magnitudes (1,000 times) safer than fossil fuels.

Contrary to that, New Scientist comes up with figures that still suggest fossil fuels are bigger killers, but the contrast with nuclear is by far less impressive. It concludes that per 1 TWh of generated electricity, there is between 0.02 to 0.12 fatalities attributable to nuclear, 0.03 to 0.16 to gas, and 0.28 to 3.3 to coal. That is roughly a difference of just one magnitude (10-30 times) between nuclear and coal, and nuclear is just comparable to gas.

The nuclear figures seem to be in line with what the International Energy Agency reported in 2002, that is, routine nuclear power related fatalities to be in the range of 0.03 and 0.16 fatalities per TWh only due to health impacts related to uranium mining – even if we simply forget about reactor accidents including Chernobyl.

So the more proper illustration comparing health and death impacts of energy sources – that is if we skip accidents such as Chernobyl and only talk about standard operation – would rather look like this (grey and black squares representing the range):

And yes, there is no doubt that for various reasons – including the related fatalities but much beyond – we need to get rid of dirty and risky energy sources, be it coal, oil, nuclear, or even gas. The future is with renewable energy technologies which have economic, safety, security of supply, as well as sustainability advantages over fossil and nuclear energy. The fatalities ratio is no differently in clear favour of renewables.

The wind power fatalities per unit of energy was reported to be twice as that of nuclear in Germany in 2000 (IEA 2002). In this chart above, we use the figure of 0.15 per TWh copied from the Next Big Future blog. It is probably safe to assume that in past decade, due to significant upscaling of energy output per one wind turbine, as well as maturing of the technology, the fatality ratio has substantially further decreased.

Jan Beránek is Head of Greenpeace International's Nuclear Campaign.