The following (dated, but relevant) refers to a reactor accident. A spent fuel pool fire could be much worse.
Calculation of Reactor Accident Consequences (CRAC-2) at U.S. Nuclear Power Plants
Sandia National Laboratory 1982 Report
“Peak” refers to the highest calculated values — it does not mean worst case scenario. This is due to uncertainties in the meteorological modeling acknowledged by Sandia. The model only considered one year’s worth of data and does not model for precipitation beyond a 30-mile radius. This is significant because the highest consequences are predicted to occur when a radioactive plume encounters rain over a densely populated area.
Peak Early Fatalities are deaths that result within the first year. The red area represents the zone for peak fatalities. This radius is the largest calculated distance from the plant at which early fatalities are expected to occur for a core melt.
Peak Early Injuries are radiation-induced injuries occurring in the first year that require hospitalization of other medical attention – such as sterility, thyroid nodules, vomiting and cataracts. The orange area represents the zone for peak early injuries. This radius is the largest calculated distance from the plant at which early injuries are expected to occur for a core melt.
Peak Cancer Deaths are predicted to occur over a lifetime. However, this is not the case with leukemia which is assumed to have occurred within the first 30 years following an accident.
Spent Fuel Accident — In the case of a spent fuel pool accident, red, orange and yellow areas would experience more than 10 times the radioactivity released in Chernobyl, and the consequences, 10 times worse!
CRAC II’s above figures are conservative because:
1. census data from 1970 was used;
2. it was assumed that the entire 10-mile EPZ would be evacuated within at most six hours after issuance of the evacuation order – the current evacuation times are longer due to traffic congestion;
3. they sampled only 100 weather sequences out of over 8 thousand (an entire year’s worth) – a method which underestimates the peak value occurring over the course of a year by about 30%;
4. they assumed aggressive medical treatment for all victims of acute radiation exposure in developing estimates of the number of early fatalities, and employed a now obsolete correlation between radiation dose and cancer risk that underestimated the risk by a factor of 4 relative to current models.