Wrong Model Was Used: "Linear No Threshold"
It is too bad the study by a Stanford professor and grad student used the "Linear No Threshold" (LNT) model to calculate their conclusions. While it is true that the US Nuclear Regulatory Agency still uses the LNT model for a variety of environmental and risk factor calculations, even the experts at NRC are well aware that the LNT model is inaccurate and predicts far higher cancer rates than have actually been observed empirically. The LNT is based on the simple and intuitive ideas that (1) any DNA damage can result in cancer, and (2) the amount of DNA damage is linearly proportional to the amount of radiation, therefore (3) the amount of cancer will be proportional to the amount of radiation. This seemed intuitively obvious in the 1960s, however all the data analyzed since then has shown that the model is false. As just one of the many examples, hundreds of thousands of people living in certain parts of the Colorado Plateau are exposed to roughly twice as much radiation as the average for North America-- because they happen to live in areas where the topsoil and sand contain more thorium and uranium than elsewhere in the world-- yet they have NO statistically greater incidence of cancer than the rest of North America. The correct radiation cancer risk model has been figured out in the past few decades, and there is not yet enough data on HUMAN exposures (thank goodness) to completely, unequivocally demonstrate the exactness of the correct model for humans. However animal studies have proven this more complicated model to be the right one. The correct model takes into account the natural self-repair rate of DNA in animal (or human) cells, and the fact that cancer-causing DNA damage, as observed in the lab, nearly always requires two or more "double-strand breaks" of the DNA by radiation to trigger a cancer. Since the DNA repairs itself naturally, and two or more radiation hits are required to activate cancer, it is clear that the cancer rate should scale as the *SQUARE* of the radiation dosage received within a narrow time-window --the time window being the natural repair rate for DNA in any particular animal tissue in question.
Put that more accurate radiation cancer model into your global-radiation-dispersal model and your geographical population distribution model, and you will find that the radiation from Fukushima will result in somewhere between zero and five deaths-- all of them workers at the power plant itself.
When you consider that risk calculations based on the Linear No Threshold model have significantly slowed the development of nuclear fission worldwide, thus aggravating and accelerating the rate of global climate change-- it is quite possible that the Linear No-Threshold model is the most expensive statistical mistake in human history.
