Outdoor Temperature and Differential Temperature as Indicators for Timing of VI Sampling to Capture the RME EPA VI Workshop AEHS San Diego, March 20, 2018
What is the current status of sampling approaches? Continuous monitoring? Sequential 14-21 day passive samplers for a year? Controlled Pressure Testing? Lots of Summa canister samples? Seasonal focused sampling?
How Do We Get to an Acceptable Sampling Approach?
Required Number of Samples to Observee RME Once 160 140 120 100 Required Number of Unguided Random Samples Per Location/Zone to Observe RME Once at Various Confidence Levels This analysis is just the mathematics of probability. No assumptions about the distribution have been made, only the assumption of random independent sampling. RME defined as percentile. 4 80 5% Prob. Of Underestimating RME 10% Prob. Of Underestimating RME 60 40 20 58 45 28 31 22 23 15 11 20% Prob. Of Underestimating RME 30% Prob. Of Underestimating RME 0 0.88 0.9 0.92 0.94 0.96 0.98 1 Percentile Defined as RME = Chance of Not Seeing RME With One Sample
Required Number of Samples to Observee RME Once Required Number of Guided Samples Per Location/Zone to Observe RME with 5% Probabilty of Underestimating 70 60 50 40 30 58 28 An indicator does not need to be perfect to be very helpful. It just needs to load the dice by significantly increasing the odds of observing a sample toward the top of the VOC distribution. 20 10 0 13 8 6 Note the 0.05 guided true positive is a guide no better then chance 4 3 2 2 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Guided True Positive Rate = Chance of Seeing RME (Here defined as 95th Percentile with One Guided Sample) 5
Season and Temperature as Indicators Season cited in number of state VI guidance documents as important to consider for characterization of VI Many northern states require one or more winter indoor air sample Some specify multiple seasons for sampling with one in winter Data available from numerous sites in temperate climates show generally higher indoor CVOC concentrations in winter One study (Johnson & Gibson, 2013) in San Antonio suggests higher indoor CVOCs during the cooling season (air conditioning and closed house conditions)
ASU SDM Data Reprise
TCE in Indoor Air [ppb v ] 10 SDM Daily Average Concentration Data Set* Spring Summer Fall Winter Spring Summer Fall Winter Spring Summer Daily Average Concentrations Median (<0.01 ppbv) Average (0.078 ppbv) 50% of Exposure (25 days >0.6 ppbv) 1 25 of 723 days (3.5%) contribute 50% of total exposure over this time frame 0.1 0.01-180 -120-60 0 60 120 180 240 300 360 420 480 540 600 660 720 8 Aug 15, 2010 (08:00) Time [d] August 14, 2012
24 hr TCE (ppbv) SDM TCE vs Differential Temperature 24 h Average TCE in Indoor Air [ppbv] vs Differential T (SDM) 1.4 1.2 1.0 0.8 0.6 R² = 0.1607 0.4 0.2 0.0-0.2-10 -5 0 5 10 15 20 25 30 35 Differential T (C)
SDM AEHS 2013 Conclusions
SDM 2015 AEHS Workshop Presentation
An Indicator Approach to Sampling What if we use Season, Outdoor Temperature or Differential Temperature as an Indicator? What criteria work (at least for SDM & Indianapolis)? How does this help us with the number of samples required to identify the RME? *Note that this approach is very likely to seriously overestimate the long term average exposure concentration
TCE (ppbv) SDM Outdoor T Indicator Approach for RME 24 hr Average TCE in Indoor Air (ppbv) vs Outdoor Temperature (degrees F) (N=579) 3 2.5 Outdoor T < 10 th Percentile and TCE > 95 th Percentile 2 1.5 1 0.5 0 0.00 0 0 10 20 30 40 50 60 70 80 90 100 Outdoor Temperature (degrees F)
PCE (ug/l) Indianapolis Duplex Outdoor T Indicator Approach for RME 422 Basement PCE vs Outdoor Temperature 9 8 7 Indicator <15th%; PCE>95th% 6 5 4 3 2 1 0 0 10 20 30 40 50 60 70 Outdoor Temperature (degrees F)
422 Basement South Weekly [PCE] Temperature Differential Another Potential Indicator Time Series Analysis Results (Indianapolis) The strength of the stack effect predicted by the temperature differential between the 422 basement south and outdoors was significant at the 1% level. Increasing strength of the stack effect was associated with higher VOC concentrations, not primarily high values in and of themselves Data from EPA/600/R-15/070 October 2015 Third Indy Report 422 Basement North Temperature (C) PCE Data from Indy Test House: Jan 2011 to Feb2012 (includes locally weighted scatterplot smoothing line [blue], with a 95% confidence interval [shaded]) 15
Key Time Series Conclusions from Indianapolis The change in the differential temperature and thus stack effect strength was more important than the absolute value of the differential temperature. Indoor air concentrations of VOCs are expected to be high when the weather is getting colder, but would not necessarily be expected to be as high during a period of sustained cold weather. Not all winter sampling times are equivalent.
SDM Differential T Indicator (>90 th %) Approach for RME Differential T>90 th percentile, TCE>95 th percentile 34% True Positives 66% False Positives
Indianapolis Duplex Differential T Indicator Approach for RME
Required Number of Samples to Observee RME Once How Many Samples Are Needed? Required Number of Guided Samples Per Location/Zone to Observe RME with 5% Probabilty of Underestimating 70 60 58 50 40 30 28 20 10 0 13 8 6 4 3 2 2 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Guided True Positive Rate = Chance of Seeing RME (Here defined as 95th Percentile with One Guided Sample)
SDM Diagnostics Table for Temperature Indicators Parameter # (TP) # (FN) # (FP) # (TN) Value (%) Guided Samples to Measure RME Temperature Outdoor (N=579) Sensitivity 21 8 37 513 72.4% Specificity 98.5% Positive Predictive Value 36.2% 7 Negative Predictive Value 93.3% Temperature Differential (N=373) Sensitivity 13 6 25 329 68.4% Specificity 92.9% Positive Predictive Value 34.2% 7 Negative Predictive Value 98.2% Definitions for Exposure Screening Sensitivity (TP/(TP+FN) Proportion of all RME conditions predicted by positive elevated Indicator results Specificity (TN/(TN+FP) Proportion of all <RME conditions predicted by non-elevated Indicator Positive Predictive Value (TP/(TP+FP) Probability that RME is present when the Indicator is positive Negative Predictive Value (TN/(TN+FN) Probability that <RME is present when the Indicator is negative True Positive (TP) = Correctly identified False Positive (FP) = Incorrectly identified True Negative (TN) = Correctly rejected False Negative (FN) = Incorrectly rejected
How can we reduce the number of samples further?
Previous Sampling Strategy Simulations Holton and others (2015) AEHS presented some sampling simulations suggesting that traditional seasonal sampling approaches were inadequate. No indicator guidance was considered in these simulations (see slide 11 for strategies evaluated). What happens if we use simple indicator guided sampling and some statistics?
Sampling Strategy Simulation (SDM) Sun Devil Manor (N=593 24 hr TCE points, 5000 simulations of random seasonal sampling) Probability of 1 or more indoor air samples exceeding Target Concentration (Percentile of data set) Seasons Sampled 4 seasons 2 Winter 3 Winter 4 Winter 5 Winter 6 Winter 7 Winter 8 Winter Total Samples 4 2 3 4 5 6 7 8 1 or more sample > 90th% 34% 47% 62% 73% 81% 86% 90% 93% 1 or more sample > 95th% 19% 26% 36% 45% 52% 59% 65% 70% 95 UCL of mean = % of distribution >92 th % >97 th % >95 th % >95 th % >95 th % >94.5 th % 94.5 th % for Student T distribution
Implications Indicators can provide guide for WHEN to sample Indicators can limit the NUMBER of samples Statistics for limited number of indicator guided samples can provide a protective estimate of the RME (especially when data are highly variable) Applied successfully at CDOT MTL & Redfield Sites For dissolved GW plumes (No vadose NAPL)
Potential Implementation, Examples and Exclusions
Practical Aspects Advanced Predictability Temperature readily available forecasts, probably sufficient short term accuracy for purpose Graphic reprinted from: http://blog.extension.uga.edu/climate/ 2015/07/when-weather-apps-go-bad/
Applicability
R. Brewer et al.,2014, Groundwater Monitoring & Remediation 34, no. 4: 79 92 Vapor Intrusion Risk Zones based on Climate (Brewer, 2014)
Long-term Stewardship Application
Redfield Post-Mitigation 1,1-DCE Seasonality
Redfield Post-Mitigation Application Example Evaluation of >200 mitigated homes at Redfield Site with 10-15 years seasonal post-mitigation data (Kurtz, 2013 AEHS) 2 winter post-mitigation samples sufficient to capture >90% of homes requiring system upgrades over 10-15 years 3 winter post-mitigation samples sufficient to capture >95% of homes requiring system upgrades
Preferential Pathways Fair amount of data suggesting winter worst for CVI from dissolved phase plumes in northern temperate climates at numerous sites, with presence of preferential pathways poorly known. 2 intensively studied residences (SDM and Indi) both have preferential pathways, but temperature (& differential T & seasonal changes) still appear to be good indicators for near worst case CVI. Stack effect is likely to be important driving force for VI, regardless of presence or absence of a preferential pathway in temperate climates with dissolved phase plumes.
When doesn t this work? Shallow NAPL Source Barnes & McRae 2017 Atmospheric Environment 150 p15-23
When doesn t this work? Pressure Driven Flow Landfill gas driven VI Gas generation that occurs at municipal solid-waste (MSW) landfills can result in pressure-driven flow into overlying or nearby buildings. In some cases, methane (& VOCs) in soil gas can be induced to move by pressure gradients resulting from barometric pressure changes or infiltrating water (Eklund, AWMA 2010). Atypical pipe flow (pressure driven)?
When doesn t this work? Climate? Possibly reverse situation in humid to tropical climates with air conditioning (closed house conditions) and higher groundwater temperatures (Johnston and Gibson, 2013). Consistent with findings that air conditioned homes have lower than average air exchange rates (Yamamoto et. al., 2009) & air conditioned homes tend to have higher radon levels (Radford, 1985). BUT, Venable and others (2015) found that winter sampling was a very significant predictor of TCE indoor concentration in a dataset of U.S. DOD industrial buildings nationwide predominantly southern or coastal sites
Data Needs to Extend Application of ITS Tropical climates Mediterranean climates Pressure induced flow (landfills) Atypical preferential pathways Broader call for sites/data at end of workshop