Formatted Title
The Shape of a Violin: Interpreting Results of the Probabilistic Vapor Intrusion Model vapintr
Background/Objectives
The Agency for Toxic Substances and Disease Registry (ATSDR) is an environmental health agency that assesses and communicates the potential for health effects from environmental exposures at hazardous waste sites. ATSDR provides advisory recommendations to protect public health and reduce exposures when a potential or known public health hazard is identified.
Vapor intrusion (VI) is subject to substantial temporal and spatial variability. Data sets for contaminated sites often include groundwater, soil gas, and indoor air concentrations sampled from multiple locations in different seasons. ATSDR developed the vapintr model as an R statistical package to use these and other data (data from soil borings, building surveys, etc.) to characterize the potential range of exposures at VI sites. The model results are not used as stand-alone estimates but as complementary results to compare to limited point-in-time and point-in-space indoor air measurements.
Approach/Activities
The vapintr R package is based on the Johnson and Ettinger model (JEM) for recalcitrant contaminants. The package conducts Monte Carlo simulations of the JEM and outputs graphical violin plots that show a range of indoor air concentrations with varying probabilities of occurrence based on the entered building conceptual site model (CSM) and subsurface media concentration inputs. The violin plots also estimate concentration quantiles (such as the 50th and 95th percentile) to represent central tendency and reasonable maximum exposures. ATSDR will share model results and discuss how CSM differences affect the shape and magnitude of the probability curve.
The vapintr Monte Carlo simulations function by iteratively calculating indoor air estimates from randomly selected model inputs. Model inputs include subsurface concentration data, soil borings, and building characteristics. Numeric parameter inputs can be defined as constants or distributions, such as triangular and program evaluation and review technique (PERT) distributions, which are both characterized by only a minimum value, central tendency estimate, and maximum value. The vapintr package will randomly select a different soil boring to characterize the subsurface in each Monte Carlo iteration. ATSDR improved on the EPA JEM spreadsheet to allow entry of an unlimited number of soil strata in each boring.
Results/Lessons Learned
In testing vapintr using representative site data, unimodal probability curves were observed for most simulated contaminants, but bimodal results were observed in some cases. Some plots were more symmetrical indicating similar tendency towards higher or lower indoor air values around the median estimate, whereas other plots yielded long narrow tails demonstrating a probability, albeit low, for more extreme low or high indoor air concentrations relative to the median. When comparing plots to measured indoor air values, this type of uncertainty analysis is valuable for investigating variable systems such as VI, i.e., measured values may fall within a high or low probability section of the curve.
This poster will present an exploration of the effect of inputs on the resultant shape and magnitude of the probability curves in order to demonstrate potential effects of different input distributions on indoor air estimates for realistic examples. Understanding the contributors to indoor air contaminant levels and the potential variability is key to public health evaluations of the potential for health effects in the ATSDR public health assessment process. The violin plots are also useful in communicating variability to communities at the vapor intrusion sites ATSDR evaluates.
Disclaimer. The findings and conclusions in this presentation have not been formally disseminated by the Agency for Toxic Substances and Disease Registry and should not be construed to represent any agency determination or policy.