INTEGRATING EARTH SCIENCE AND EPIDEMIOLOGY: MODELING THE PROBABILITY OF ARSENIC IN NEW ENGLAND GROUNDWATER FOR EXPOSURE ASSESSMENT
Recent ecological evidence suggests that bladder cancer mortality in the region is correlated with private well use. An epidemiologic study of bladder cancer in northern New England is underway. The purpose is to explain the excess of bladder cancer, with a major focus on the risk of long-term consumption of arsenic in drinking water. The study includes an estimate of arsenic exposure and exposure uncertainty during adult life for each study subject based on residential history, arsenic concentrations in drinking water wells at current and past residences, and water-quality and use data from public water supplies. In some cases, there are gaps in the exposure profile because of unavailable data. To estimate arsenic concentrations in the gaps, a process-based statistical model has been developed to identify geographic areas in New England with a probability of arsenic concentrations exceeding 5 µg/L in drinking water wells.
In the model, multivariate logistic regression is used to estimate the probability of occurrence of elevated arsenic concentrations in ground water. The model is based on geologic, hydrologic, and landscape information, including specific geologic formation units, arsenic concentrations in stream sediments, areas of Pleistocene marine inundation, and proximity to intrusive granitic plutons. These and other hydrologic and landscape variables related to arsenic sources and ground-water geochemical factors and residence time are shown to increase the probability of elevated arsenic concentrations in ground water. Previous studies suggest that arsenic in bedrock ground water may be partly from past arsenical pesticide use but variables representing historic agricultural inputs do not improve the model, indicating that this source does not significantly contribute to current arsenic concentrations.