Final Report: Hazardous Air Pollutant Mixtures: Measuring and Modeling Complex Exposure

EPA Grant Number: R827928
Title: Hazardous Air Pollutant Mixtures: Measuring and Modeling Complex Exposure
Investigators: Adgate, John L. , Church, Timothy , Pratt, Greg , Ramachandran, Gurumurthy , Sexton, Ken , Zhang, Junfeng
Institution: University of Minnesota , Minnesota Pollution Control Agency , University of Medicine and Dentistry of New Jersey
EPA Project Officer: Chung, Serena
Project Period: December 20, 1999 through December 19, 2002 (Extended to June 19, 2004)
Project Amount: $510,012
RFA: Urban Air Toxics (1999) RFA Text |  Recipients Lists
Research Category: Air Quality and Air Toxics , Air

Objective:

The objectives of this research project were to:  (1) explore the statistical relations between ambient dispersion modeling results and outdoor volatile organic compound (VOC) measurements; (2) examine the associations between outdoor, indoor, and personal measurements of trace elements in the fine particulate matter (PM particles less than or equal to 2.5 μm in diameter) fraction; and (3) explore the relationships between VOCs and trace elements in the outdoor, indoor, and personal measurements.

This study initially was performed in three communities (Battle Creek, East St. Paul, and Phillips) in the Minneapolis-St. Paul metropolitan area (see Figure 1).  Peer-reviewed publications describing the study design (Adgate, et al., 2002) and the methods used to obtain outdoor, indoor, and personal PM2.5 (Adgate,et al., 2002; Adgate,et al., 2003; Ramachandran,et al., 2000) and VOC measurements (Pratt, Bock,et al., 2004; Sexton,et al., 2004) can be consulted for specific methods and details.

Summary/Accomplishments (Outputs/Outcomes):

At the time of this report, most of our findings have been published in peer-reviewed literature or are undergoing peer review.  Our complete findings on the relationship between VOCs and trace elements are still under development, however, and the results reported here reflect our preliminary conclusions. 

The major findings from this grant are summarized in the following five sections, which:  (1) compare the results of an ambient dispersion model using a refined emissions inventory with VOC measurement results in three urban neighborhoods; (2) explore the comparability of passive diffusion-based organic vapor monitors (OVMs) with the U.S. Federal Reference Method for measuring VOCs; (3) examine the variability in gravimetric 24-hour average and short-term (15-minute average) PM2.5 concentrations in outdoor and indoor air; (4) estimate mean differences in trace element levels between monitoring locations (i.e., outdoor-indoor, outdoor-personal, and indoor-personal) over time, using a mixed-effects statistical model that adjusts for the effects of monitor location, community, and season while accounting for within-subject and within-monitoring period correlation; and (5) report preliminary conclusions on the associations between trace elements in PM2.5 and VOCs in outdoor, indoor, and personal measurements.

Figure 1.   Study Area:  Minneapolis and St. Paul (Darker Gray); Areas of the Three Labeled Communities Where Monitoring Was Done (Black)

Ambient Dispersion Model Versus Monitoring Results

Air concentrations of nine VOCs were measured over 48-hour periods at 23 locations in three communities in the Minneapolis-St. Paul metropolitan area (Pratt and Bock, et al., 2004).  Concentrations at the same times and locations were modeled using a standard regulatory air dispersion model, Industrial Source Complex Short Term-3 (ISCST3), to evaluate model performance by comparing predictions with measurements using linear regression and estimates of bias.  The modeling, done with mobile and area source emissions resolved to the census tract level and characterized as model area sources, represents an improvement over large-scale air toxics modeling analyses done to date.  Despite the resolved spatial scale, the model did not fully capture the spatial resolution in concentrations in an area with a sharp gradient in emissions.  In a census tract with a major highway at one end of the tract (i.e., uneven distribution of emissions within the tract), model predictions at the opposite end of the tract overestimated measured concentrations.  This shortcoming was seen for pollutants emitted mainly by mobile sources (benzene, ethylbenzene, toluene, and xylenes).  We suggest that major highways would be better characterized as line sources.  The model also failed to fully capture the temporal variability in concentrations, which was expected because the emissions inventory comprised annual average values.  Based on our evaluation metrics, model performance was best for pollutants emitted mainly from mobile sources and poorest for pollutants emitted mainly from area sources. 

Important sources of error appeared to be the source characterization (especially location) and emissions quantification.  We expect that enhancements in the emissions inventory would give the greatest improvement in results.  As anticipated for a Gaussian plume model, performance was dramatically better when compared to measurements that were not matched in space or time.  Despite the limitations of our analysis, we found that the regulatory air dispersion model generally was able to predict space- and time-matched 48-hour average ambient concentrations of VOC species within a factor of two on average, results that meet regulatory model acceptance criteria.

Comparison of VOC Measurement Methods

Concurrent field measurements of 10 VOCs were made using passive diffusion-based OVMs and the U.S. Federal Reference Method, which uses active monitoring with stainless steel canisters to store samples for subsequent laboratory analysis (Pratt, et al., in revision, 2004).  Measurements were obtained repeatedly throughout a range of weather conditions over the course of three seasons in Battle Creek, East St. Paul, and Phillips.  Ambient concentrations of most VOCs measured by both methods were low compared to other large metropolitan areas.  For some VOCs, a considerable fraction of measurements was below the detection limit of one or both methods.  The observed differences between the two methods were similar across measurement sites, seasons, or any meteorological variable.  A Bayesian analysis with uniform priors on the differences was applied, with accommodation of sometimes heavy censoring (nondetection) in either device.  The resulting estimates of bias and standard deviation of the OVM relative to the stainless steel canisters were computed by tertile of the canister-measured concentration.  In general, OVM and stainless steel canister measurements were in best agreement for benzene and other aromatic compounds with hydrocarbon additions (ethyl benzene, toluene, and xylenes).  The two methods were not in such good agreement for styrene and halogenated compounds (carbon tetrachloride, p-dichlorobenzene, methylene chloride, and trichloroethylene).  OVMs slightly overestimated benzene concentrations and carbon tetrachloride at low concentrations, but in all other cases where significant differences were found, OVMs underestimated relative to canisters.  Our study indicates that the two methods are in agreement for some compounds, but not all.  We provide data on and interpretation of the relative performance of the two VOC measurement methods, which facilitates intercomparisons among studies.

Comparison of Continuous Versus PM2.5 Measurements

Little is known about the within-day variability of indoor and outdoor PM levels.  Variability in measurements of daily (24-hour) average PM2.5 concentrations and short-term (15-minute average) PM2.5 concentrations were explored in outdoor and indoor microenvironments in Battle Creek, East St. Paul, and Phillips (Ramachandran, et al., 2003).  Daily average PM2.5 concentrations were measured using gravimetry, whereas measurements of 15-minute average PM2.5 mass concentrations were made using a light scattering photometer in which readings were normalized using the gravimetric measurements.  Outdoor measurements were made at a central monitoring site in each of the three communities and indoor measurements were made in 9-10 residences (with nonsmoking occupants) in each community.  Outdoor PM2.5 concentrations across the Minneapolis-St. Paul metropolitan area appear to be spatially homogeneous on a 24-hour time scale as well as on a 15-minute time scale.  Short-term average outdoor PM2.5 concentrations can vary by as much as an order of magnitude within a day.  The frequency distribution of outdoor 15-minute averages can be described by a trimodal lognormal distribution; the three modes have geometric means of 1.1 µg/m3 (geometric standard deviation [GSD] = 2.1), 6.7 µg/m3 (GSD = 1.6), and 20.8 µg/m3 (GSD = 1.3).  There is much greater variability in the within-day 15-minute indoor concentrations than outdoor concentrations (as much as ~ 40-fold).  This most likely is a result of the influence of indoor sources and activities that cause high short-term peaks in concentrations.  The indoor 15-minute averages have a bimodal lognormal frequency distribution; the two modes have geometric means of 8.3 µg/m3 (GSD = 1.66) and 35.9 µg/m3 (GSD = 1.8), respectively.  The correlation between the matched outdoor and indoor 15-minute average PM2.5 concentrations showed a strong seasonal effect, with higher values observed in spring and summer (R2adj = 0.49 ± 0.33) and lower values in fall (R2adj = 0.13 ± 0.13).

Differences in Trace Element Levels by Measurement Location

Twenty-four hour average concentrations of 22 trace metals (Ag, Al, Ca, Cd, Co, Cr, Cs, Cu, Fe, K, La, Mg, Mn, Na, Ni, Pb, Sb, Sc, Ti, Tl, V, and Zn) and S were measured (using high-resolution inductively coupled plasma-mass spectrometry) concurrently in outdoor, indoor, and personal sample pairs for 32 healthy, nonsmoking adults (Adgate, et al., in preparation).  After blanks subtraction using field blank samples, a relatively high percent of samples greater than 0 were observed for more prevalent elements such as S, Ca, and K.  Intermediate-to-high percent detectable levels of heavy metals, such as Ni and Pb, were observed, and several elements (e.g., Ag, La, Sc, and V) were observed infrequently.  Outdoor PM2.5 concentrations were relatively low compared to many other urban areas, but many trace element levels (µg/m3) do not follow the general personal greater than indoor greater than outdoor pattern observed for PM2.5 total mass measurements (µg/m3).

A hierarchical mixed-effects statistical model was used to estimate the mutually adjusted effects of monitor location, community, and season on mean differences between monitoring locations (i.e., outdoor-indoor, outdoor-personal, and indoor-personal) across low, medium, and high tertiles of personal exposure while accounting for within-subject and within-monitoring period correlation.  For outdoor-indoor differences, the data indicate that across communities, estimated outdoor concentrations of some elements were higher in outdoor air than indoor air (S and Ca), but others were higher in indoor air (K and Mn).  Phillips had a significantly higher contribution of Ni from outdoor air across all seasons and exposure tertiles.  The modeled differences indicate that indoors, most other element levels tended to be higher in East St. Paul and Phillips than in Battle Creek, especially in the low and medium tertiles of exposure.  The outdoor-personal patterns are similar to outdoor-indoor:  across communities estimated outdoor concentrations of some metals were higher in outdoor air than indoor air (S and Ca), and the intercept for S in the highest tertile is highly significant and in the microgram range.  As with outdoor-indoor, some elements (K and Mn) tended to be higher in indoor air in East St. Paul and Phillips in the highest tertile, and Phillips still has a significantly higher contribution of Ni from outdoor air.  The modeled differences in East St. Paul and Phillips indicate that personal levels of most elements tended to be higher in these two communities than in Battle Creek.  The results for the estimated indoor-personal mean differences support these conclusions but add little additional insight into the observed statistical associations.  The overall results indicate that ambient trace element measurements at central monitoring sites can both overestimate and underestimate actual exposure to these elements for urban residents. 

Associations Between Trace Elements in PM2.5 and VOCs 

Initial attempts to use a mixed-model approach to understand the associations between these two classes of air pollutants have been unsuccessful.  This analysis is complicated because the measurements have different averaging times (temporally aligned, VOC measurements are 48-hour averages, whereas PM2.5 measurements are 24-hour averages) and substantially different underlying distributions that resist use of a common transformation (e.g., log) to comply with the underlying assumptions of this statistical approach.  Whereas measured VOC levels show a consistent personal greater than indoor greater than outdoor pattern for most compounds, trace element levels do not all fit into this pattern, and thus we are working on using receptor model techniques to fully understand the complex patterns in our data.

Conclusions:

Overall, our main conclusions are the following:  

  • Compared to the Federal Reference Method, passive diffusion monitors can be used to reasonably estimate outdoor exposure to benzene and other aromatic compounds but did less well for styrene and halogenated compounds.  The data and statistical methods we developed demonstrate the relative performance of the two VOC measurement methods, which will facilitate intercomparisons among studies.
  • The ambient dispersion modeling, done with a standard regulatory model and mobile and area source emissions resolved to the census tract level, represents an improvement over large-scale air toxics modeling analyses done to date.  Despite the improved spatial resolution of our emissions inventory, the model did not fully capture the spatial resolution in concentrations for areas with sharp emissions gradients.  Model performance was best for pollutants emitted mainly from mobile sources and poorest for pollutants emitted mainly from area sources.  Important sources of error appeared to be the source characterization (especially location) and emissions quantification.
  • Outdoor PM2.5 concentrations across the Minneapolis-St. Paul metropolitan area appear to be spatially homogeneous on 24-hour and 15-minute time scales.  Given that there is as much as a 40-fold greater variability in the within-day indoor concentrations than outdoor concentrations, further exploration of peak exposures and the influence of season and indoor sources and activities that cause high short-term peaks in PM2.5 concentrations should be an important future research goal.  
  • Whereas PM2.5 concentrations follow the general pattern outdoor less than indoor less than personal, the 22 trace elements measured in this study do not follow this pattern in all cases.  Estimated mean differences between outdoor, indoor, and personal measurements using mixed-effects statistical models indicate that across communities estimated outdoor concentrations of some elements were higher in outdoor air than compared to indoor and personal (e.g., S and Ca), but others were higher in indoor air (e.g., K and Mn).  Modeled differences indicate that indoor levels of most other element levels were higher in East St. Paul and Phillips compared to Battle Creek, especially in the low and medium tertiles of exposure.  The overall results suggest that ambient trace element measurements at central monitoring sites can both overestimate and underestimate estimate actual exposure to these elements for urban residents.
  • Improved receptor modeling approaches are needed to better understand the associations between VOCs and trace element exposure in outdoor, indoor, and personal air.

 


Journal Articles on this Report : 7 Displayed | Download in RIS Format

Other project views: All 9 publications 7 publications in selected types All 7 journal articles
Type Citation Project Document Sources
Journal Article Adgate JL, Ramachandran G, Pratt GC, Waller LA, Sexton K. Spatial and temporal variability in outdoor, indoor, and personal PM2.5 exposure. Atmospheric Environment 2002;36(20):3255-3265. R827928 (2002)
R827928 (2003)
R827928 (Final)
R825241 (Final)
  • Full-text: Science Direct-Full Text HTML
    Exit
  • Abstract: Science Direct-Abstract
    Exit
  • Other: Science Direct-Full Text PDF
    Exit
  • Journal Article Adgate JL, Ramachandran G, Pratt GC, Waller LA, Sexton K. Longitudinal variability in outdoor, indoor, and personal PM2.5 exposure in healthy non-smoking adults. Atmospheric Environment 2003;37(7):993-1002. R827928 (2002)
    R827928 (2003)
    R827928 (Final)
    R825241 (Final)
  • Full-text: Science Direct-Full Text HTML
    Exit
  • Abstract: Science Direct-Abstract
    Exit
  • Other: Science Direct-Full Text PDF
    Exit
  • Journal Article Adgate JL, Mongin SJ, Pratt GC, Zhang J, Field MP, Ramachandran G, Sexton K. Relationships between personal, indoor, and outdoor exposures to trace elements in PM(2.5). Science of the Total Environment 2007;386(1-3):21-32. R827928 (Final)
    R833627 (Final)
  • Abstract from PubMed
  • Full-text: ScienceDirect-Full Text HTML
    Exit
  • Abstract: ScienceDirect-Abstract
    Exit
  • Other: ScienceDirect-Full Text PDF
    Exit
  • Journal Article Pratt GC, Wu CY, Bock D, Adgate JL, Ramachandran G, Stock TH, Morandi M, Sexton K. Comparing air dispersion model predictions with measured concentrations of VOCs in urban communities. Environmental Science & Technology 2004;38(7):1949-1959. R827928 (2002)
    R827928 (2003)
    R827928 (Final)
    R825241 (Final)
  • Abstract from PubMed
  • Full-text: ACS-Full Text HTML
    Exit
  • Abstract: ACS-Abstract
    Exit
  • Other: ACS-Full Text PDF
    Exit
  • Journal Article Pratt GC, Bock D, Stock TH, Morandi M, Adgate JL, Ramachandran G, Mongin SJ, Sexton K. A field comparison of volatile organic compound measurements using passive organic vapor monitors and stainless steel canisters. Environmental Science & Technology 2005;39(9):3261-3268. R827928 (Final)
    R825241 (Final)
  • Abstract from PubMed
  • Full-text: Academia.edu-Full Text PDF
    Exit
  • Abstract: ACS-Abstract
    Exit
  • Other: ACS-Full Text PDF
    Exit
  • Journal Article Sexton K, Adgate JL, Ramachandran G, Pratt GC, Mongin SJ, Stock TH, Morandi MT. Comparison of personal, indoor, and outdoor exposures to hazardous air pollutants in three urban communities. Environmental Science & Technology 2004;38(2):423-430. R827928 (2003)
    R827928 (Final)
    R825241 (Final)
  • Abstract from PubMed
  • Full-text: ACS-Full Text HTML
    Exit
  • Abstract: ACS-Abstract
    Exit
  • Other: ACS-Full Text PDF
    Exit
  • Journal Article Sexton K, Adgate JL, Mongin SJ, Pratt GC, Ramachandran G, Stock TH, Morandi MT. Evaluating differences between measured personal exposures to volatile organic compounds and concentrations in outdoor and indoor air. Environmental Science & Technology 2004;38(9):2593-2602. R827928 (2003)
    R827928 (Final)
    R825241 (Final)
  • Abstract from PubMed
  • Full-text: ACS-Full Text HTML
    Exit
  • Abstract: ACS-Abstract
    Exit
  • Other: ACS-Full Text PDF
    Exit
  • Supplemental Keywords:

    EPA Region 5, Midwest, particulate matter, PM, personal PM2.5, fine particles, volatile organic compound, VOC, Minneapolis-St. Paul metropolitan area, Minnesota, MN, ambient particle properties, air toxics., RFA, Scientific Discipline, Toxics, Air, Geographic Area, Waste, Ecosystem Protection/Environmental Exposure & Risk, air toxics, Environmental Chemistry, Chemistry, State, chemical mixtures, Monitoring/Modeling, indoor air, tropospheric ozone, Biology, 33/50, carbon tetrachloride, environmental monitoring, monitoring, particulate matter, cumulative exposure, Minnesota, MN, air pollutants, ambient particle properties, chemical characteristics, particulate, Toluene, stratospheric ozone, mass spectrometry, VOCs, air pollution, benzene, Chloroform, analytical chemistry, PM2.5, atmospheric monitoring, indoor air quality, Volatile Organic Compounds (VOCs), air quality, atmospheric chemistry, heavy metals

    Relevant Websites:

    http://www.pca.state.mn.us/air/airtoxics.html Exit

    Progress and Final Reports:

    Original Abstract
  • 2000 Progress Report
  • 2001 Progress Report
  • 2002 Progress Report
  • 2003 Progress Report