Characterizing Variability and Uncertainty in Exposure Assessments Improves Links to Environmental Decision-Making.

Environmental decisions often rely upon observational data or model estimates. For instance, evaluation of human health or ecological risks often includes information on pollutant emission rates, environmental concentrations, exposures, and exposure/dose-response data. Whether measured or modeled, each of these elements of information has certain underlying limitations. In addition to basic accuracy and precision issues, spatial and temporal representativeness of data and their applicability to different population or receptor groups of interest are important concerns. Understanding the various sources of variability and uncertainty in exposure and risk information is relevant to many types of environmental decision-making, including setting standards, determining emissions controls, and mitigating exposures to pollutants. Moreover, critical gaps in knowledge can be determined based on evaluation of uncertainties, so that priorities for collecting new data can be specified to improve confidence in future assessments. Risk managers typically consider likely impacts of alternative regulatory or risk management choices by performing risk comparisons, examining risk-risk trade-offs, or comparing risks to benchmarks. Incorporating variability and uncertainty surrounding the estimates of human exposures is key to making sound decisions and maximizing the benefits attained from such decisions. Consider, for example, the hypothetical case shown in Figure 1, where the population risk associated with a pollutant in the air (i.e., the “Baseline” case) is currently projected to exceed either of the two alternative regulatory “Brightlines” (i.e., dashed lines labeled “Brightline-1” and “Brightline-2”). Note that distribution of risks among the individuals in the population may vary greatly, depending on the inherent variability or uncertainties in the different factors influencing personal exposures. The baseline case suggests that more than one 25th of the population exceeds the Brightline-1 risk. However, there may be a number of alternative regulatory exposure mitigation or risk management choices that may result in reducing risks below the Brightline-1 risk level. Each of these may have different implications as to how exposures and health risks may be distributed within the affected population. Figure 1 Population Distribution of Exposures Associated with Hypothetical Risk Reduction Scenarios The two scenarios (A and B) show how either targeted or uniform across-the-board exposure reductions may be considered to achieve the desired risk reduction. Under Scenario A, only the mitigation options targeting the high-end exposure groups are adopted. Example situations may include populations residing near major sources or roadways and population groups that have higher rate of contact with the pollutant of concern, either due to their personal activity, mobility/commuting patterns, or housing characteristics influencing the indoor infiltration of outdoor pollutants. Thus, Scenario A may target more localized air pollution sources or selected population subgroups for exposure and risk reduction. In contrast, Scenario B will implement general or more uniform reductions in emissions or exposures across the entire population to meet the targeted risk levels below the Brightline-1. In this situation, large-scale urban or even regional improvements in air quality most likely will need to be considered. Note that under Scenario B population average (mean) risks are lower than for Scenario A. These distributional differences may have important implications for population versus individual level risk assessments, especially at the community level. Finally, neither scenario fully complies with the more stringent Brightline-2 risk limits. To meet these limits, much lower target risks (e.g., Scenario C) and much more comprehensive exposure and risk reduction programs, ranging from local to urban to regional scale, must be devised. Examples of recent probabilistic exposure assessments include a sensitivity analysis of a probabilistic exposure assessment of children’s exposure to arsenic in chromated copper arsenate pressure-treated wood that was used to prioritize the need to collect specific data regarding dislodgeable residue and potential exposure;1 a probabilistic risk assessment of polychlorinated biphenyls (PCBs) in the Hudson River that lead to greater public understanding of the findings;2 and a probabilistic assessment of human exposure to ozone that was influential in the recent rulemaking regarding a revised National Ambient Air Quality Standard.3