Use of environmental TLD data at a nuclear power station to estimate detection limits for radiation exposure due to station operation.
The environmental radiation monitoring network maintained by the Georgia Department of Natural Resources at the Hatch Nuclear Plant, Baxley, GA, was used to determine the variability in measured background exposures in order to estimate the detection limits for radiation due to station operation. Data were obtained from LiF thermoluminescent dosimeters (TLDs) exposed in sets of four at 46 locations for 16 consecutive quarters from 1982 to 1985. The average exposure for 695 values was 11.8 microC kg-1 y-1 (mR/quarter). At individual locations, the averages of quarterly exposures ranged from 9.3 to 14.3 microC kg-1 y-1 and had standard deviations from 0.8 to 2.2 microC kg-1 y-1 with a typical standard deviation of 1.5 microC kg-1 y-1. The standard deviation due to reading the TLDs and subtracting the transit plus storage exposure, estimated as 0.4 microC kg-1 y-1, contributed only a small fraction of the uncertainty of the mean background; the greater contribution is from temporal fluctuations. There is a spatial pattern of higher background along the Altamaha River but no distinct seasonal pattern. The observation that exposure values at most locations rose and fell in unison suggested a calculational model that adjusts the average exposure level at a location by an increment determined each quarter at control locations. This model reduced the standard deviation for a predicted quarterly background exposure to 0.9 microC kg-1 y-1, with a range of 0.5 to 1.3 microC kg-1 y-1. If the standard deviation of the net value is multiplied by an error probability constant of 3.3 (type I and II errors not greater than 5%), the detection limits range from 2 to 5 microC kg-1 y-1 at the individual locations, with an average of 3 microC kg-1 y-1. This approach or two similar ones described earlier are recommended for improving the detection limit where needed. The standard deviation, and hence the detection limit, can be further reduced by selecting a subset of indicator and control locations with more closely matched background exposure fluctuations. This was demonstrated with five indicator and five control locations, but further study is needed to determine selection guides.