2010 Progress Report: Development and Assessment of Environmental Indicators: Application to Mobile Source Impacts on Emissions, Air Quality and Health Outcomes

EPA Grant Number: R833626
Title: Development and Assessment of Environmental Indicators: Application to Mobile Source Impacts on Emissions, Air Quality and Health Outcomes
Investigators: Russell, Armistead G. , Darrow, Lyndsey , Klein, Mitchel , Mulholland, James , Pachon, Jorge , Sarnat, Stefanie Ebelt , Sarnat, Jeremy , Tolbert, Paige
Current Investigators: Russell, Armistead G. , Klein, Mitchel , Mulholland, James , Sarnat, Stefanie Ebelt , Sarnat, Jeremy , Tolbert, Paige
Institution: Georgia Institute of Technology , Emory University
EPA Project Officer: Nolt-Helms, Cynthia
Project Period: October 1, 2007 through September 30, 2010 (Extended to September 30, 2011)
Project Period Covered by this Report: October 1, 2009 through September 30,2010
Project Amount: $499,512
RFA: Development of Environmental Health Outcome Indicators (2006) RFA Text |  Recipients Lists
Research Category: Health Effects , Health

Objective:

  • Develop approaches to identify outcome-based indicators and apply those approaches to mobile sources in Atlanta for the period 1998-2004 using data and methods of varying detail and complexity.
  • Test a range of integrated indicators for the impact of mobile source emission changes on air quality and cardiovascular health.
  • Develop and apply approaches for assessing these indicators for their ability to represent a range of outcomes associated with mobile source emissions and policies.
  • Evaluate select indicators using an independent, new data set of emissions, air quality and cardiovascular health endpoints for 2005-2009 to assess the approaches developed for identifying, testing, assessing and refining outcome indicators.

Progress Summary:

At this point of the project, we concluded our estimation of the secondary organic carbon (SOC) by four different methods and evaluated the association of both primary and secondary fractions with health outcomes. Our results are in agreement with other studies, in which the primary fraction of the organic carbon resulted in a significant association with cardiovascular diseases (CVD). This primary fraction is related to combustion activities, such as traffic and biomass burning.

We have advanced in the development and assessment of an outcome-based, multipollutant indicator for use in air quality and epidemiologic analysis, in which pollutants are mixed based on their emissions ratios. Such an indicator will represent a source of emissions. For our analysis, we chose mobile sources because they represent a significant fraction of pollution in urban centers. The association with health was based on CVD outcomes, which are aggravated by combustion sources. We found that the proposed multipollutant indicator (here called EB-IMSI) not only represents adequately the impact of mobile sources on air quality, but also has a stronger association with CVD than any of the single species (EC, CO and NOx). A sensitivity analysis of fractions in which pollutants can be combined to form the indicator shows that the combination that constitutes the proposed indicator (a weighting based on emission ratios) is near the strongest associated with CVD. This outcome-based, multipollutant indicator, built on readily available air quality data and a straightforward calculation, provides support for the setting of multipollutant air quality standards.

During the period covered in this report, we have advanced in the following activities:
· Health associations of Primary and Secondary Organic Carbon fractions
· Design of outcome-based, multipollutant indicators for air quality and epidemiological analysis
 
Health Associations of Primary and Secondary Organic Carbon Fractions
 
We concluded our estimation of the secondary organic carbon (SOC) fraction in Atlanta for the years 1999-2007 using four different methods. The results were published in a peer-reviewed journal. Briefly, we tested four methods (EC tracer, multiple regression, CMB and PMF) and found that the multiple regression resulted in a more accurate estimation of SOC when compared with daily measurements of the water soluble organic carbon (WSOC) used as a surrogate of SOC, and SOC estimates from CMB using molecular markers. Results from the regression method showed the least uncertainty among the four methods, a greater value in summer than winter, and less day-to-day variability, as expected for a secondary pollutant.
 
We use estimates of the primary (POC) and secondary organic carbon fractions to assess the association of carbonaceous PM species with health outcomes. The POC fraction is mainly emitted in combustion activities, such as traffic and biomass burning, especially important in Atlanta. The SOC fraction is formed in the atmosphere by the condensation of volatile organic compounds (VOC), both biogenic and anthropogenic, under photochemical conditions.
 
Two groups of health outcomes were evaluated for significance with air pollution: respiratory and cardiovascular diseases. The respiratory group (all resp) includes asthma-wheezing (asthwhez), chronic obstructive pulmonary disease (COPD), upper respiratory infections (upresp) and pneumonia (pneu). The cardiovascular group (all CVD) includes dysrythmia (dysryth), cardiac arrest (cardarrest), congestive heart failure (CHF), ischemic heart disease (ischemia), myocardial infarction (myocardial), peripheral + cerebrovascular disease (peri) and disease of circulatory system (all circ).
 
The POC fraction shows a significant association (p-value < 0.05) with the "all-card" group, especially with "peri" outcomes and in a minor scale with "all-circ" outcomes (Figure1). Both, POC and SOC show association with "COPD," which is in the group of "all-resp" outcomes, the significance being very comparable between each other and slightly larger than the significance of total OC (Figure 2).
 
 
 
 
These results suggest that POC, and therefore combustion-related activities, are more associated with CVD than respiratory outcomes; an important conclusion that has been pointed in similar studies (Metzger et al., 2004; Sarnat et al., 2008)
 
Design of Outcome-Based, Multipollutant Indicators for Air Quality and Epidemiological Analysis
 
Air pollution has been traditionally managed on a single pollutant basis, from the setting of air quality standards to the association with health outcomes. However, we breathe and interact in a multipollutant atmosphere, and therefore, a multipollutant approach would be desirable in the management of air pollution. The National Research Council (NRC), NARSTO(2008) and EPA (2008) have elaborated on the tasks needed to move to a risk-based, accountable, multipollutant air quality approach, identifying as the main limitation the lack of knowledge on the health impacts of mixtures of pollutants. In fact, multipollutant models in epidemiologic analysis have been limited to the evaluation of two or more pollutants at a time, more with the rationale to identify confounders than mixtures of pollutants representing a source and affecting health.
 
We propose a multipollutant indicator for use in air quality and epidemiologic analysis, in which pollutants are mixed based on their emissions ratios. In this way, such an indicator will truly represent a source of emission. For our analysis, we chose mobile sources because they represent a significant fraction of pollution in urban centers. The association with health was based on cardiovascular diseases (CVD), which are known to be aggravated by combustion sources (Sarnat et al., 2008).
 
The steps that we follow on the development and assessment of the multipollutant indicator are:
i. Selection of pollutants
ii. Analysis of emissions and ambient air concentrations
iii. Development and assessment of multipollutant indicators
iv. Sensitivity analysis of the mixtures
 
i. Selection of pollutants
 
Traditionally, carbon monoxide (CO) and nitrogen oxides (NOx) have been used as tracers of vehicular activity. CO is generated under fuel-rich combustion conditions, proper of spark engines and hence, is a pollutant associated to gasoline fleet. On the other hand, NOx is generated as a result of fuel-lean (air-rich) combustion conditions typical of diesel engines. Another pollutant of interest is PM2.5, generated not only in combustion activities but also mechanical processes and even secondary formation (i.e., formation in the atmosphere from other precursors and photochemical conditions). Because PM2.5 can have several sources, the carbonaceous fraction is a better indicator of combustion sources. Elemental and organic carbon (EC, OC, respectively) are species formed under diverse combustion conditions, OC being produced in early stages of the combustion and EC at later stages at higher temperatures. Both, gasoline and diesel vehicles emit EC and OC in different proportions. The gasoline fleet usually has an OC/EC ratio greater than the diesel fleet ratio, with values around 3.0-4.0 for gasoline and below 1.0 for diesel. EC has been pointed as a stronger tracer of diesel engines than OC is for gasoline fleet (Lee et al., 2007; Zheng et al., 2007).
 
Other pollutants have also been used as tracers to identify mobile sources, and specifically to split between gasoline and diesel fleet. Heavy metals found in the PM2.5, such as zinc (Zn), nickel (Ni), vanadium (V), copper (Cu) and lead (Pb), are the most common referenced in the literature (Lee et al., 2008). Zn is used as a tracer of gasoline vehicles, Pb and Cu are from brake wear and road traffic, and Ni and V from fuel-oil combustion. Organic compounds, such as hopanes and polycyclic aromatic hydrocarbons (PAH) are also used as tracers (Zheng et al., 2002). Although these compounds are very useful in the identification of the sources, their measurements are resource intensive and they are not widely available, for which we focus our study in readily available species.
 
ii. Analysis of emissions and ambient air concentrations
 
To select the pollutants that best represent mobile source activity and the emissions ratios in which they can be combined to form the indicator, we performed an analysis of emissions and ambient air concentrations for CO, NOx, PM2.5, EC and OC.
 
Ambient air quality data are obtained from the Jefferson Street station, a site operated by the SEARCH project (Edgerton et al., 2005; Hansen et al., 2003). JST is a sophisticated station with filter-based and real-time measurements available through a web portal (ARA Inc). JST is located in downtown Atlanta, particularly in Fulton County. To evaluate emissions from mobile sources in the period 1999-2007 in Fulton County, we used information from three sources: (1) the National Emission Inventory NEI (US-EPA) with emissions estimates for the years 1999, 2002, 2005 and 2008; (2) application of the EPA MOVES 2010 (US-EPA, 2010) software to estimate local on-road emission based on nation-wide vehicle miles traveled (VMT) data; and (3) estimates from the VISTAS project for 2002 (Air Resources Specialists, 2007). MOVES only provides emissions from on-road sources, while NEI and VISTAS provide emissions from all sources (non-road, point, non-point).
 
From the VISTAS project in 2002, only 20 percent of the total OC emissions are from mobiles sources (on-road and non-road), so in our analysis we are not considering this pollutant as a good indicator for mobile sources.
 
As a summary of the comparison between emissions and ambient air concentrations of single pollutant indicators (Figures 3-5), we observe that CO is a good indicator of the gasoline fleet and EC is a good indicator of diesel engines. NOx is an indicator of diesel fleet, but is largely influenced by emissions from gasoline vehicles.
 
 
 
 
Nitrogen Oxides (NOx)
 
MOVES estimates that 60 percent of the on-road emissions are from gasoline fleet and 40 percent are from diesel engines. Although diesel engines emit more NOx than spark engines given the richer air/fuel ratio, the gasoline fleet is significantly larger than the diesel fleet for Fulton County, compensating the effect of lower emissions. Other sources, contributing up to 13 percent of NOx emissions are classified as area and point sources, in particular fuel combustion in electrical generating utilities (EGU) and biomass burning. The results indicate that NOx emissions are a significant, but not dominant, indicator of mobile sources.
 
 
 
 
Elemental Carbon (EC)
 
The EPA-NEI does not report emissions of EC. From the VISTAS estimates, 92 percent of the total EC emissions in 2002 were from mobiles sources (on-road and non-road), being only 8 percent from other sources, such as biomass burning. From the on-road fraction, MOVES estimates 91 percent from diesel engines and 9 percent from gasoline fleet. These results indicate that EC emissions are dominated by mobile sources for Fulton County.
 
 
 
 
iii. Development and assessment of multipollutant indicators
 
From the numerous ways pollutants can be mixed, we chose a combination that would represent an actual emission source, and therefore, use information from emissions to estimate ratios in which pollutants can be blended. We develop this approach for mobile sources, using as a single species CO-1h, NOx-1h and EC. We call these indicators emission-based Integrated Mobile Source Indicators (EB-IMSI). The emissions information is used here in the form of ratios of pollutants that can be used to weigh the combination of the single species. Because the original pollutants have different units, we normalize the ambient air concentrations by the mean of each variable (EC, NOx, CO) during the period 1999-2007. The proposed indicator has the following expression.
 
 
The ratios between mobile sources and total emissions for CO and NOx were obtained from the NEI. The value of the (NOxmob/NOx total) ratio is 0.84 ± 0.03 and (COmob/COtot) is 0.97 ± 0.01. For EC, a further calculation was necessary, obtaining a (ECmob/ECtot) ratio of 0.91 ± 0.28. When ambient air concentrations of EC, NOx and CO are equal to their means, the weighted contribution of each factor is approximately the same (EC = 0.33, NOx = 0.31, CO = 0.36), which captures information from all these pollutants in one indicator.
 
The EB-IMSI is also normalized by the sum of the emission ratios, in such a way that the indicator has an average value of 1.0 and can be easily compared with other metrics. Similarly, we define integrated indicators for gasoline engines (EB-IMSIGV) and diesel fleet (EB-IMSIDV) using particular emission ratios from gasoline or diesel.
 
The uncertainty in EB-IMSI is estimated by propagating uncertainties from individual terms, taking into account that individual species are correlated between each other (ISO, 1993).
 
 
 
In this way, the uncertainty in EB-IMSI before being normalized is expressed as:
 
 
Where e=(ECmob/ECtot), c=(COmob/COtot), n=(NOxmob/NOxtot), and EC, CO and Nox represent the normalized concentrations (i.e. divided by the mean).
 
The temporal trend of the EB-IMSI indicator (Figure 6) is similar to the trend of the single species they are integrating (Figure 3-5). EB-IMSI shows a good correlation with EC (R2 = 0.74), CO (R2 = 0.86) and NOx (R2 = 0.81), which is expected since they are the species forming the indicator.
 
 
 
 
 
To assess how the integrated indicators represent mobile sources, we compare daily time series of indicators with source impacts from receptor models. The chemical mass balance CMB (Watson et al., 1984) and the Positive Matrix Factorization PMF (Norris and Vedantham, 2008) were run on exactly the same period of time and the source/factor associated with mobile sources was selected for further analysis. EB-IMSI adequately tracks mobile source daily impacts from CMB (R2 = 0.83) and PMF (R2 = 0.84) (Figure 7).
 
 
 
The comparison of uncertainties between ambient measurements, integrated indicator and source impacts from receptor models, led to conclude that ambient concentrations have the lowest uncertainties, especially CO and NOx. The integrated indicators show uncertainties larger than the ambient measurements, since uncertainties from emissions ratios are involved. However, uncertainties of EB-IMSI are less than uncertainties in CMB results, which include not only the uncertainties in air concentrations, but also uncertainties in source profiles and fitting of the mass balance method.
 
iv. Health association and sensitivity analysis
 
To evaluate the health association of our indicators, we use health outcome residuals from the epidemiologic model. As explained in our previous progress report, our epidemiologic model (Eq. 4) was executed without the pollutant variable to generate the residual between the estimated and observed emergency department (ED) count. Our hypothesis is that the residual should contain information about the association of the health outcome with the air pollutant (not controlled in the model).
 
 
 
The significance of the association between the health outcome and the pollutant is evaluated through the correlation coefficient and the p-value (< 0.05). We assessed the association with indicators from various levels: both single pollutant and multipollutant indicators. We observe an increase in the significance of the association (lower p-values) as it move towards a multipollutant approach (Figure 8).
 
 
 
 
Among the single species indicators, EC was the most significant associated with CVD, followed by NOx-8h and CO-1h, the latter one with a p-value greater than 0.05 but below 0.1 (i.e., significant at the 90% confidence level). However, mixtures of CO1h and NOx1h became significant, which shows a "synergism" between these pairs of pollutants. The multipollutant indicator EB-IMSI resulted more significantly than any of its precursors (EC, CO1h, NOx8h). The most significant association was in a mixture of EC and NOx1h, as in the case of the diesel fraction of the EB-IMSI, which confirms previous studies about the impact of diesel sources on cardiovascular health.
 
Because different mixtures of pollutants can result in different association with health, we performed a sensitivity analysis of our multipollutant indicators. First, we assess mixtures of CO and NOx in the form ‘alfa*NOx+(1-alfa)*CO’, simulating emissions from gasoline fleet. Second, we assess mixtures of EC and NOx in the form ‘alfa*NOx+(1-alfa)*EC’, simulating emissions from diesel engines (Figure 9).
 
 
 
 
We found that a mixture of CO and NOx becomes more significant as we move towards NOx because CO was not significantly associated with CVD outcomes. The fraction of gasoline vehicles of the EB-IMSI corresponds approximately to a mixture of 0.5*NOx1h+0.5*CO1h where the significance is larger than CO1h but less than NOx1h. On the other hand, a mixture of EC and NOx has a strongest association with CVD outcomes at 0.5*NOx1h+0.5*EC, that corresponds approximately to the diesel fraction of the EB-IMSI, indicating a positive association between diesel emissions and CVD.
 
In conclusion, we can state that using health outcome residuals are useful to screen for association with air pollution. A multipollutant indicator increases significance in the association with CVD outcomes. Indicators from emission outcomes identify diesel emissions strongly associated with CVD outcomes.

We have developed and assessed outcome-based, multipollutant indicators for mobile sources that use readily available data and require a simple calculation. These indicators show good correlation with mobile source activity and a stronger association with health outcomes than the individual pollutants. The EB-IMSI can be the base for the setting of air quality multipollutant standards.

Future Activities:

Our focus for the subsequent period will be conclusion of the project with the development of indicator sets, for which we account for the indicator structure, the associated uncertainty and the outcome relationship. We will be evaluating this new set of indicators using an independent set of air quality and health data from the period 2005-2009.

References:

Air Resources Specialists. Visibility Improvement State and Tribal Association of the Southeast (VISTAS) Conceptual Description Support Document. Asheville, NC: Georgia Department of Natural Resources, 2007.
 
Edgerton ES, Hartsell BE, Saylor RD, Jansen JJ, Hansen DA, Hidy GM., The Southeastern Aerosol Research and Characterization Study:  part II. Filter-based measurements of fine and coarse particulate matter mass and composition. Journal of the Air & Waste Management Association 2005;55(10):1527-1542.
 
Hansen DA, Edgerton ES, Hartsell BE, Jansen JJ, Kandasamy N, Hidy GM, Blanchard CL. The Southeastern Aerosol Research and Characterization Study:  part 1-overview. Journal of the Air & Waste Management Association 2003;53(12):1460-1471.
 
ISO. Guide to the expression of uncertainty in measurements. Geneva, Switzerland:  International Organization for Standardization ISO, 1993.
 
Lee S, Liu W, Wang YH, Russell AG, Edgerton ES. Source apportionment of PM2.5:  comparing PMF and CMB results for four ambient monitorinig sites in the southeastern United States. Atmospheric Environment 2008;42(18):4126-4137.
 
Lee S, Russell AG, Baumann K. Source apportionment of fine particulate matter in the southeastern United States. Journal of the Air & Waste Management Association 2007;57(9):1123-1135.
 
Metzger KB, Tolbert PE, Klein M, Peel JL, Flanders WD, Todd K, Mulholland JA, Ryan PB, Frumkin H. Ambient air pollution and cardiovascular emergency department visits. Epidemiology 2004;15(1):46-56.
 
NARSTO. Technical challenges of risk- and results-based multipollutant air quality management: a summary for policy makers. NARSTO, 2008,
 
Norris G, Vedantham R. EPA Positive Maxtrix Factorization (PMF) 3.0. Research Triangle Park, NC:  U.S. Environmental Protection Agency, 2008, EPA 600/R-08/108.
 
Sarnat JA, Marmur A, Klein M, Kim E, Russell AG, Sarnat SE, Mulholland JA, Hopke PK, Tolbert PE. Fine particle sources and cardiorespiratory morbidity:   an application of chemical mass balance and factor analytical source-apportionment methods. Environmental Health Perspectives 2008;116(4):459-466.
 
US-EPA. Emissions Inventories - NEI.
 
US-EPA. The multi-pollutant report: technical concepts & examples. Environmental Protection Agency, 2008.
 
US-EPA. Motor Vehicle Emissions Simulator MOVES 2010, Userguide. Research Triangle Park, NC:  U.S. Environmental Protection Agency, 2010, EPA-420-B-09-41.
 
Watson JG, Cooper JA, Huntzicker JJ. The effective variance weighting for least-squares calculations applied to the Mass Balance Receptor Model. Atmospheric Environment 1984;18(7):1347-1355.
 
Zheng M, Cass GR, Ke L, Wang F, Schauer JJ, Edgerton ES, Russell AG. Source apportionment of daily fine particulate matter at Jefferson street, Atlanta, GA, during summer and winter. Journal of the Air & Waste Management Association 2007;57(2):228-242.
 
Zheng M, Cass GR, Schauer JJ, Edgerton ES. Source apportionment of PM2.5 in the southeastern United States using solvent-extractable organic compounds as tracers. Environmental Science & Technology 2002;36(11):2361.


Journal Articles on this Report : 2 Displayed | Download in RIS Format

Other project views: All 17 publications 12 publications in selected types All 12 journal articles
Type Citation Project Document Sources
Journal Article Pachon JE, Balachandran S, Hu Y, Weber RJ, Mulholland JA, Russell AG. Comparison of SOC estimates and uncertainties from aerosol chemical composition and gas phase data in Atlanta. Atmospheric Environment 2010;44 (32):3907-3914. R833626 (2010)
R833626 (Final)
R833866 (Final)
  • Full-text: Science Direct-Full Text HTML
    Exit
  • Abstract: Science Direct-Abstract
    Exit
  • Other: Science Direct-Full Text PDF
    Exit
  • Journal Article Sarnat JA, Moise T, Shpund J, Liu Y, Pachon JE, Qasrawi R, Abdeen Z, Brenner, S, Nassar K, Saleh R, Schauer JJ. Assessing the spatial and temporal variability of fine particulate matter components in Israeli, Jordanian, and Palestinian cities. Atmospheric Environment 2010;44(20):2383-2392. R833626 (2010)
    R833626 (Final)
  • Full-text: ScienceDirect-Full Text HTML
    Exit
  • Supplemental Keywords:

    Indicators, IMSI, Secondary Organic Carbon (SOC), mobile sources, health residuals

    Progress and Final Reports:

    Original Abstract
  • 2008 Progress Report
  • 2009 Progress Report
  • Final Report