Updating exposure models of indoor air pollution due to vapor intrusion: Bayesian calibration of the Johnson-Ettinger model.

The migration of chlorinated volatile organic compounds from groundwater to indoor air--known as vapor intrusion--is an important exposure pathway at sites with contaminated groundwater. High-quality screening methods to prioritize homes for monitoring and remediation are needed, because measuring indoor air quality in privately owned buildings is often logistically and financially infeasible. We demonstrate an approach for improving the accuracy of the Johnson-Ettinger model (JEM), which the Environmental Protection Agency (EPA) recommends as a screening tool in assessing vapor intrusion risks. We use Bayesian statistical techniques to update key Johnson-Ettinger input parameters, and we compare the performance of the prior and updated models in predicting indoor air concentrations measured in 20 homes. Overall, the updated model reduces the root mean squared error in the predicted concentration by 66%, in comparison to the prior model. Further, in 18 of the 20 homes, the mean measured concentration is within the 90% confidence interval of the concentration predicted by the updated model. The resulting calibrated model accounts for model uncertainty and variability and decreases the false negatives rate; hence, it may offer an improved screening approach, compared to the current EPA deterministic approach.