Testing the atmospheric dispersion model of CSA N288.1 with site-specific data

The atmospheric dispersion component of CSA Standard N288. 1, which provides guidelines for calculating derived release limits, has been tested. Long-term average concentrations of tritium in air were predicted using site-specific release rates and meteorological data and compared with measured concentrations at 43 monitoring sites at all CANDU stations in Canada. The predictions correlate well with the observations but were found to be conservative, overestimating by about 50% on average. The model overpredicted 84% of the time, with the highest prediction lying a factor of 5.5 above the corresponding observation. The model underpredicted the remaining 16% of the time, with the lowest prediction about one-half of the corresponding measurement. Possible explanations for this bias are discussed but no single reason appears capable of accounting for the discrepancy. Rather, the tendency to overprediction seems to result from the cumulative effects of a number of small conservatisms in the model. The model predictions were slightly better when site-specific meteorological data were used in the calculations in place of the default data of N288.1. Some large discrepancies between predictions and observations at specific monitoring sites suggest that it is the measurements rather than the model that are at fault. The testing has therefore provided a check on the observations as well as on the model. Recommendations on model use and data collection are made to improve the level of agreement between predictions and observations in the future.