Computational Techniques for Stochastic Source-to-Dose Modeling of Human Exposures

Estimation of human exposures to co-occurring chemicals is a critical component in understanding the impact of environmental pollutants on human health. Mechanistic models are valuable tools for providing such estimates because they can consistently describe the physical and chemical processes occurring across different spatial and temporal scales, as well as account for the inherent uncertainty and variability. Comprehensive exposure analysis involves linking various aspects in the source-to-dose continuum, including sources of pollutant release, atmospheric transport and chemical transformation processes, human activity patterns, and physiological attributes of the target individual/population. Therefore, several computational issues arise in the course of mechanistic modeling of individual and population exposures. These are presented in the context of the stochastic source-to-dose modeling framework of MENTOR (Modeling ENvironment for TOtal Risk studies).

MENTOR follows a "One Atmosphere" approach to characterize cumulative exposures to co-occurring air pollutants by taking into account their physical and chemical interactions, and calculates exposure and dose profiles, while providing the ability to focus on time scales and subpopulations of interest. The system estimates individual and population exposures and doses by combining information on: demographic characteristics of the population under study, outdoor concentration distributions, indoor/outdoor air exchange rates, indoor sources, time-activity diaries, and biologically based dosimetry. It constructs a sequence of exposure events for a large number of "virtual individuals" by sampling from relevant distributions, and calculates exposures and doses associated with each activity for each virtual individual.

The presentation focuses on computational modules for generating statistically representative virtual individuals, activity events, and for performing exposure and dose calculations. Other relevant computational modules for real time application of MENTOR for emergency event analysis, as well as for efficient characterization of sensitivity and uncertainty, are also discussed.