Assessing Variability in Petroleum Vapor Intrusion Jim Weaver United States Environmental Protection Agency Office of Research and Development National Risk Management Research Laboratory Ground Water and Ecosystem Restoration Division Ada, Oklahoma
Outline Vapor Intrusion and Petroleum Vapor Intrusion Characteristics of Environmental Models PVIScreen model Examples –Cumulative Frequency Curves –Screen Shots Availability and Peer Reviews 9/4/20152
Vapor Intrusion and Petroleum Vapor Intrusion 9/4/20153 Chlorinated Solvent (left) petroleum (right) are distinguished by prospects for biodegradation
Why vapor intrusion and models? Technical Challenges –ambient air contamination, internal sources/sinks, temporal changes Social –RP or homeowner reluctance to sample In some cases—redeveloping a site—no building exists for testing, so models are relied upon 9/4/20154
Vapor Intrusion and Models Series of articles in the Denver Post in 2000 –The vapor intrusion model (Johnson-Ettinger) over-predicted indoor air concentrations sometimes and under-predicted indoor air concentrations sometimes –(Model used with defaults and very few site specific values) 9/4/20155
Environmental Transport and Fate Models Environmental transport models have an inherently empirical two-part basis: –relationships between physical, chemical and biological quantities –Experimentation to determine the values of coefficients 9/4/20156
Example: Darcy’s Law Darcy flux q = -K dh/dl –Relationship from Darcy’s sand tank experiments –Empirical coefficient, the hydraulic conductivity (K), from experiment: measuring the flow (q) MaterialValue (m/d) Gravel10 2 to 10 4 Sand10 -1 to 10 3 Clay10 -8 to Sandstone10 -5 to 10 Basalt10 -6 to /4/20157
Limits to Predictability Note the work of N. Oreskes on ideal applications for models: –Classical physics experiments –Planetary motion Planets are observed to check model predictions –Weather forecasting Forecast given and received with uncertainties Oreskes, Naomi, 2003, The role of quantitative models in science, in Models in Ecosystem Science, C.D. Canham and W.K. Lauenroth, eds. Princeton University Press, /4/20158
Calibrated and Uncalibrated The planetary motion models can be calibrated, while the weather forecast can only be calibrated after the fact. –The forecasts of the uncalibrated models need uncertainty analysis. Vapor Intrusion models are being used in a partially/wholly uncalibrated fashion. 9/4/20159
PVIScreen PVIScreen includes: –A petroleum hydrocarbon lens as a source (can be omitted to create ground water source) –BioVapor equations, recoded in Java Johnson-Ettinger model if no biodegradation –Automated Monte Carlo uncertainty analysis –Flexible and customizable unit choices –User Interface –Automated Report Primary raison d’ ê tre –To make uncertainty analysis practical 9/4/201510
frequency parameter value frequency parameter value A constant Uniform distribution: min, max Data-driven empirical distribution Results: indoor air concentration frequency Multiple Model Runs Inputs Results All other inputs, Other distributions can be used: Triangular, truncated normal Log normal 9/4/2015
Example Problem Definition Reformulated Gasoline, BTEX + TPH Groups Some of the variable parameters –Air Exchange Rate 0.1 to 1.5 hr -1 –Moisture content 0.05 to 0.2 –Porosity 0.3 to 0.5 –Fraction Organic Carbon to –“Q soil ” 1 to 10 L/min –Crack width 0.5 to 5 mm 9/4/201512
Example Problem Schematic 9/4/201513
9/4/201514
9/4/201515
9/4/201516
9/4/201517
9/4/201518
9/4/201519
9/4/201520
Model Output – all parameter values saved with results Method to display parameter values which exceed risk levels is being developed 9/4/201521
Availability and Peer Review Command Line or “DOS” version –Input and output from Excel spreadsheet –Flexible input, but requires exactitude in entering required quantities User Interface – being completed –Available for review The views expressed in this presentation are those of the author and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency 9/4/201522