necessary to exclude particular census tracts or cancer types from the epi-demiologic study in cases where there are substantial industrial releases. This will need to be handled on a facility-by-facility basis.
The uncertainties in dose estimates for an epidemiologic study are likely to be substantial. These uncertainties arise from uncertainties in source terms (i.e., reported effluent releases; see Chapter 2) and, usually to a greater extent, uncertainties in atmospheric transport and liquid dispersion models that relate these source terms to environmental concentrations, and also uncertainties in pathway models that relate environmental concentrations to dose. Uncertainties in dose estimates have the potential to mask the “true” dose-response relationship in an epidemiologic study. Consequently, understanding and characterizing these uncertainties is important.
The magnitude of dose estimate uncertainties is also likely to vary over time. Effluent release data for early years of facilities operations are of lower quality than more recent data (see Chapter 2). As a consequence, dose estimates based on earlier data are likely to be more uncertain than doses calculated for releases for more recent years. Moreover, because effluent releases in earlier years were much higher as a result of higher airborne effluent releases (see Chapter 2), uncertainties in airborne effluent releases are likely to be relatively more important than uncertainties in liquid effluent releases. The airborne effluent release uncertainties are a function of how representative the weekly grab samples22 were with respect to the actual releases of specific nuclides, as well as to uncertainties in stack airflow rates, especially if they varied with time. There is much less uncertainty associated with the measured activities of the grab samples themselves. Furthermore, the use of an average quarterly value for batch releases rather than the actual values for each batch adds to the reported uncertainties and resultant dose estimates, particularly for PWRs.
Uncertainties in diffusion and dispersion models that relate source terms (effluent releases) to environmental concentrations as well as exposure pathway models relating environmental concentrations to doses can be high. Atmospheric dispersion estimates can also be very uncertain, particularly when releases are episodic, when there are terrain irregularities, and for locations that are distant from the facility fence line (Table 3.14). On sites with flat terrain, Gaussian plume models have been shown to provide reasonable estimates of air concentrations when integrated over a sufficient time interval, although estimates for a shorter integration times can be very uncertain. Uncertainties increase for sites with complex terrain (e.g., sites with hills or valleys). Also, local meteorology at any particular time (wind speed, direction, and atmospheric stability) can vary significantly from annual averages and result in significant errors if the latter are used to estimate doses for batch effluent releases into the atmosphere.