Reliance on models
Uncertainties in dose estimates
In official publications the radiation doses to individuals near nuclear power stations are invariably very low. These values are estimates and are not based on measurements. How these estimates are derived is not well understood by scientists, and not at all by members of the public.
The methodology is very complicated as it based on at least four kinds of computer models in sequence:
• Models for the generation of fission and activation products in reactor cores. The emission data published by utilities are derived from these models.
• Environmental transport models for radionuclides, including weather models.
• Human metabolism models to estimate radionuclide uptake, retention and excretion.
• Dose models which estimate radiation doses from internally retained radionuclides.
Each model has its inherent limitations so the result of each model has an uncertainty range. The uncertainties of each model have to be treated together to gain an idea of the overall uncertainty in the final dose estimate. Further uncertainties are introduced by 'unconservative' radiation weighting factors, dose rate reduction factors, and tissue weighting factors in the official models. The cumulative uncertainty in dose estimates could be very large.
Uncertainties in risk estimates
Risk models are used to estimate the likely level of cancers. The risk models have their inherent imperfections and uncertainties as well as the dose estimate models. The current official risk models are mainly based on studies of the Japanese survivors of the nuclear bombs in 1945. How reliable are the official risk models? Uncertainties are introduced by a number of factors, such as:
• The Japanese bomb survivor study was started five years after the bomb blasts, so the deaths in the first five years were not counted.
• The risks estimated from a sudden puls of gamma rays and high-energy neutrons are not applicable to environmental releases which result in chronic, slow, internal exposures to often low-range beta radiation, from biologically reactive radionuclides, such as tritium and carbon-14, but also from low-gamma alpha-emitting radionuclides.
• Application to adults only.
• Application of age and gender-averaged risks.
• Arbitrarely halving the risks to take account of cell studies suggesting lower risks from low doses and low dose rates.
Troublesome detection of radionuclides
An impediment for sound health risks assessments is the fact that a number of dangerous radionuclides are hard to detect with commonly used radiation counters. As a result of the difficult detectability, severe radioactive contamination with these radionuclides may escape notice during prolonged periods. Not every spill or release contains 'marker' nuclides which are easily detectable, such as cesium-137.
Examples of 'unnoticed' releases are the routine releases of nuclear power plants under nominal conditions [more i19]. For that reason it would be advisable to check on regular occasions food and drinking water on the presence of those troublesome radionuclides, even if no direct threat seems apparent. Risk estimates based on models likely will not come up to the mark.
Inherently limited significance of models
Any model in physics, economics or other field, inevitably has two kinds of limitations: inherent limitations and the specific limitations resulting from the choice of input data: constants, variables and other data.
A model is a simplified description of the reality, the practice, and is based on a number of axioms and assumptions. Models are widely used in science to describe specified phenomena in nature and to build a theory wich enables scientists to predict the occurrence of such phenomena under conditions different from the investigated ones. As a result of the simplification of the reality a model is only valid within specific system boundaries and has a limited application range. The wider the system boundaries of a model, the more complicated its structure. As a well-know scientist put it:
'If empirical observation is incompatible with a model, the model must be trashed or amended, even if it is conceptual beautiful or mathemathecally convenient.'
Two examples of scientific models used in chemistry may illustrate this statement. The simple model of atoms and molecules formulated by Dalton in the 19th century is able to describe some basic chemical phenomena. To explain why water has the formula H2O and not H3O and to predict chemical compounds not yet found, one needs the greatly more complicated atom model of Bohr. However, not all chemical phenomena can be explained by the Bohr model.
Specific limitations: the choice of input data
The results of an investigation by means of a model are determined by the input data, such as physical constants, variables and properties of the entities of the model.
How reliable are the axioms the model is based on and the input data? Are they experimentally verified and are they widely accepted by the scientific community? How large are the uncertainty ranges of the numerical input data and how do these uncertainties propagate into the results? How sure can we be that the investigatorŐs choices of the input data of his model were not outdated, or biased, wittingly or unwittingly?
From a scientific/mathematic viewpoint the radiological models seem rigid: once formulated, always and everywhere valid. Conspiciously the radiological models turn out to be flexible under economic pressure, as is proved by the recent relaxation of authorized radioactivity standards for drinking water in the USA and the relaxation of exposure standards in Japan after the Fukushima disaster.
Why not start from empirical evidence?
Which assumptions form the basis of the currently used radiological models? Which phenomena are included in the models and which are not? What was the original purpose of the models? To estimate the acute radiological risks for military personel in wartime, or to estimate the health risks for the public posed by chronic exposure to a number of radionuclides from civilian nuclear power?
More than ever the time has come to base health risk estimates on published and verifiable empirical facts, not on computer models originating from the closed nuclear industrial complex and based on outdated secret data. Epidemiologic studies in Germany anf France proved that the existing exposure and health risk models are unable to explain the empirical observations of that study, so the models must be revised.