PNNL

Abstract

In support of the development of nuclear weapons, the United States government began acquiring uranium, and on a smaller scale, thorium, beginning around 1940. Much of the processing of ore, compounds, and metals was done by private companies for the Manhattan Engineer District (MED) and later the U.S. Atomic Energy Commission (AEC). The MED Health and Safety Division evolved into the Health and Safety Laboratory (HASL) of the AEC. HASL personnel visited many of the private plants to assess exposures to radiation and radioactive materials, and developed a time-and-task approach to determining “daily weighted average” (DWA) concentrations representing exposures to airborne uranium, thorium, radon, and radon decay products. Generally the surveys and analyses were of the highest quality. Short-term exposures >105 dpm m-3 of uranium and >105 pCi L-1 of radon were observed, but DWAs were much lower. Exposure estimates in the form of DWAs are useful as inputs to dose reconstruction under the Energy Employees Occupational Illness Compensation Program as performed by the Office of Compensation Analysis and Support (OCAS) of the National Institute for Occupational Safety and Health (NIOSH). However, the software used for determining probability of causation, the Interactive Radioepidemiological Program (IREP), requires not only dose estimates, but numerical expressions of uncertainties in dose estimates. The reported DWAs had no uncertainty associated with them. In this work, we show that Monte Carlo methods can be used to assess uncertainty and variability in the DWA results. The methods were tested for 63 job titles from 5 sites that processed U, U ore, Th, or 226Ra-222Rn between 1948 and 1955. Most groups of repeated air samples could be well described by lognormal distributions. Combining samples associated with different tasks often resulted in a reduction of the geometric standard deviation of the DWA to less than those GSDs typical of individual tasks. This work provides support for the practice of assuming a GSD of 5 when information on uncertainty in DWA exposures is unavailable. Errors involving arithmetic, transposition, and transcription were discovered. In 5 cases out of 63, these blunders resulted in overestimates of DWAs by a factor of up to 2.5, and in 2 cases, underestimates of DWAs by factors of 3 and 10, respectively.