National Academies Press: OpenBook
« Previous: 2 Overview of Ambient-Ozone Standards Development and Benefits Assessment
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 48
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 49
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 50
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 51
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 52
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 53
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 54
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 55
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 56
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 57
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 58
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 59
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 60
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 61
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 62
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 63
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 64
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 65
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 66
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 67
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 68
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 69
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 70
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 71
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 72
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 73
Suggested Citation:"3 Ambient Ozone and Related Pollutants." National Research Council. 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution. Washington, DC: The National Academies Press. doi: 10.17226/12198.
×
Page 74

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

3 Ambient Ozone and Related Pollutants Ozone is a naturally occurring compound that is found in the troposphere (the layer of the atmosphere next to Earth’s surface) and, at greater concentra- tions, in the stratosphere (the layer about 10-50 km above the surface). In the stratosphere, ozone absorbs and thus reduces the amount of potentially harmful ultraviolet (UV) radiation. Ozone in the troposphere also absorbs UV radiation and plays a role in the degradation of toxic compounds. More important, how- ever, tropospheric ozone is responsible for various deleterious effects on humans and other organisms that are exacerbated by increased ambient ozone concentra- tions. Preindustrial tropospheric background ozone concentrations were about 10 ppb, but they have increased to about 30-40 ppb (e.g., Volz and Kley 1988; NRC 1991; Finnlayson-Pitts and Pitts 2000) because of emissions and atmos- pheric chemistry, as discussed below. In and around urban areas in the United States, ambient ozone concentrations rise much further, to about 160 ppb in Houston and 180 ppb in Los Angeles. Although such concentrations may seem extreme, they are substantially reduced from the concentrations of more than 400 ppb experienced in the 1970s (e.g., NRC 1991). OZONE FORMATION AND TRANSPORT We must understand the formation and transport of tropospheric ozone if we are to be able to interpret health-benefits assessments of ozone and conduct new ones, inasmuch as the interpretation of epidemiologic analyses must con- sider ozone’s spatial and temporal patterns and their relationships with human exposure. Processes that affect ozone dynamics also affect other pollutants, such as components of particulate matter, and lead to what can be strong correlations between pollutant concentrations. This effect hinders our ability to determine the health effects of single pollutants that are present in a mixture. That is critical because a study may falsely implicate a pollutant that is not responsible for most of the health effects found, and strategies designed to reduce that pollutant may 48

Ambient Ozone and Related Pollutants 49 be ineffective in reducing the pollutants that are responsible for the health ef- fects. An important question is whether that is the case in the approach to ozone and airborne particulate matter (PM), of which the latter seems to have greater health effects (see Chapter 2). Most ozone in the troposphere is not directly emitted to the atmosphere (although there are minor sources of such ozone, including some indoor air cleaners) (e.g., NRC 1991). Rather, it is formed from a complex series of photo- chemical reactions of the primary precursors: nitrogen oxides (NOx where [NOx] indicates the sum of [NO] and [NO2]), volatile organic compounds (VOCs), and to a smaller extent other pollutants, such as carbon monoxide (CO) (Figure 3-1) (e.g., NRC 1991). The specific reactions that form most of the tropospheric ozone are the photolysis of NO2 followed by the combination of the released oxygen atom with the abundant oxygen molecules (O2): NO2 + sunlight NO + O [3-1] O + O2 O 3 [3-2] Volatile Compounds Organic Particles Semivolatile Organic Compounds OH Compounds Secondary Organic Semivolatile and RO 2 Particles Gaseous O3 Compounds OH NO 3 O3 Inorganic N 2O 5 NO NO 2 Nitrates O3 HNO 3 hν h Ozone NH 4 NO 3 NH 3 (NH 4 )2 SO 4 OH H 2O SO 2 H 2SO 4 Chemical Deposition O3 , H 2 O 2, O 2 FIGURE 3-1 Source and chemical links between ozone and PM formation. Major pre- cursors are shown in boxes with thick sides. Secondary particle components are shown in boxes with thin solid sides. Mobile sources (cars, trucks, and off-road vehicles) and plants are major sources of VOCs, and mobile sources and electricity-generating units are dominant sources of NOx, but myriad smaller sources also contribute. Trace species, such as OH, are crucial to the formation of ozone, sulfate, nitrate, and organic-carbon particu- late matter. Ozone also leads to the oxidation of SO2 and NO2. Biologic activity and fer- tilizer use dominate ammonia (NH3) emissions. Source: Modified from NARSTO 2004. Reprinted with permission; copyright 2004, Cambridge University Press.

50 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits VOCs enter the picture by their reactions with the hydroxyl radical (OH) and other oxidants, which lead to the formation of NO2 from NO, shown in a highly simplified form as VOC + OH RO2 [3-3] RO2 + NO NO2 + OPs [3-4] In Reactions 3-3 and 3-4, RO2 is an oxygenated radical (there are many such species in the atmosphere), and OPs are oxygenated organic products, including aldehydes, ketones, acids, and condensable species that lead to secondary or- ganic aerosol (SOA) formation. Although a small fraction of ground-level ozone is transported from the stratosphere, the reactions shown above (summarizing hundreds of reactions and species) are responsible for most ambient ozone, par- ticularly on high-ozone days in urban areas (Fiore et al. 2003). Thus, ozone air- quality management strategies concentrate on reducing emissions of the ozone precursors. However, the nonlinearity of the relationship between ozone and its precursors (Figure 3-2) complicates decisions about which precursors should be controlled and to what degree. Because reactions that form ozone are driven by sunlight, ambient ozone concentrations exhibit both diurnal variation (they are typically highest during the afternoon, as seen in Figure 3-3a) and marked seasonal variation (they are highest in summer, as seen in Figure 3-3b). Ambient concentrations are highest during hot, sunny summer episodes characterized by low ventilation (a result of low winds and low vertical mixing). Diurnal variability is enhanced, particularly in urban areas, by reactions between ozone and fresh emissions of NO (such as that from automobiles): NO + O3 NO2 + O2 [3-5] (An additional, more complex set of reactions leads to local reductions in ozone in response to increased NOx emissions.) The impact of this Reaction 3-5 is par- ticularly marked after the sun sets and during the day in areas with high NOx (Figure 3-2); it forces ozone concentrations down, even below background con- centrations. NO-associated decreases tend to occur relatively close to the NO sources. Indeed, increased NOx emissions can lead to local ozone reductions (for example, near major roads and near the stack in power-plant plumes), whereas ozone formation downwind is enhanced (Figure 3-4) (see also Ryerson et al. 2001). The relationship between VOC emissions and ozone concentrations is somewhat simpler, with higher emissions typically leading to higher ambient ozone concentrations, particularly in urban areas. The long effective atmospheric lifetime of ozone, measured in weeks (IPCC 2001), leads to substantial regional and intercontinental transport, and increased global NOx emissions result in steadily increasing background ozone concentrations (e.g., Finnlayson-Pitts and Pitts 2000). That is in stark contrast

Ambient Ozone and Related Pollutants 51 FIGURE 3-2 Ozone isopleth diagrams showing the nonlinear response of ozone to emis- sions of VOCs and NOx and how they can vary from a city center to a downwind location. Source: Adapted from Dodge 1977. 0.14 Maximum Ozone (ppm) 0.16 0.12 Ozone (ppm) Maximum Ozone 0.1 0.12 0.08 0.06 0.08 0.04 0.04 0.02 0 0 0 5 10 15 20 25 30 0 60 120 180 240 300 360 Hour Day of Year (a) (b) FIGURE 3-3 Ozone concentrations in Atlanta in 2006: (a) diurnal variation of ozone on July 22; (b) daily maximum. with the initial view that ozone was an urban air pollutant that affected primarily the larger cities, such as Los Angeles (NRC 1991), and that it could be substan- tially mitigated by reducing VOCs. As pollution transport from high-NOx- emitting areas, such as the Ohio River Valley, to downwind areas and cities be- came recognized as an important contributor to regional ozone, control efforts were redirected to reduce NOx. Now industrialization of developing nations and recognition of the stability of ozone and its increased concentrations over the oceans have led to a global view of the problem. Global background concentra- tions are increasing, they are higher over the continents, and they are yet higher

52 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits over and downwind of cities. As the ozone NAAQS is tightened, the regional concentrations approach the desired concentrations, providing very little room for the bump of ozone from individual cities. Downtown Atlanta 0.100 S(2) cross Source Contributions (ppm) 0.080 S(2) VOC Ozone Concentrations & S(2) NOX 0.060 S(1) VOC S(1) NOX 0.040 Ozone Conc. 0.020 0.000 -0.020 -0.040 00 00 00 00 00 0 0 0 0 0 0 0 :0 :0 :0 :0 :0 :0 :0 0: 2: 4: 6: 8: 20 22 10 12 14 16 18 Local Time Suburban Atlanta (Cherokee Co.) 0.100 S(2) cross Source Contributions (ppm) S(2) VOC 0.080 Ozone Concentrations & S(2) NOX S(1) VOC 0.060 S(1) NOX Ozone Conc. 0.040 0.020 0.000 -0.020 -0.040 00 00 00 00 00 0 0 0 0 0 0 0 :0 :0 :0 :0 :0 :0 :0 0: 2: 4: 6: 8: 10 12 14 16 18 20 22 Local Time FIGURE 3-4 Simulated ozone concentration and sensitivity of ozone to Atlanta-area NOx and VOC emissions for downtown Atlanta and a suburban location downwind. Bars in each graph illustrate how much local emissions change ozone concentrations in down- town and suburban Atlanta. Combined red and tan bars show how emissions of NOx in the Atlanta area affect ozone. If bars are above zero (positive sensitivity), Atlanta NOx emissions increase ozone in locations shown at that time. If bars are negative, such NOx emissions lead to decreases in ozone at that location and time. Locally, increased NOx emissions lead to decrease in ozone (negative sensitivity) through much of day downtown but generally increase ozone downwind in suburban area. VOC emissions increase ozone during daytime downtown (for example, green portion of bars) but have less effect at night and downwind. Gap between ozone concentration and sum of sensitivities shows magnitude of ozone transported from outside area. Reducing local NOx emissions can decrease ozone peak but actually lead to an increase in 24-h average ozone on such a high-ozone day. On lower-ozone days, decreasing local NOx emissions can increase ozone through more of day. Source: Cohan et al. 2005. Reprinted with permission; copy- right 2005, American Chemical Society.

Ambient Ozone and Related Pollutants 53 The amount of regional transport relative to local ozone production and the amount of NOx emissions relative to VOC emissions lead to differences in the dynamics of ozone formation between cities (and in responses to control, as discussed below). Where VOC emissions are abundant (or NOx emissions are low) and ozone concentrations are more homogeneous, ozone goes down rela- tively little at night, and weekend concentrations are similar to those in the rest of the week. That is particularly true in rural areas. In a place like Los Angeles, where ozone is largely produced locally and there are relatively high local NOx emissions, ozone has a distinct diurnal trend, reaching a maximum in the after- noon and dropping toward zero in the evening and during the night. Ozone can increase on the weekend as NOx emissions decline (for example, because of decreases in traffic and construction). That is the “weekend-weekday ozone” effect (e.g., Chow 2003 and references therein). Despite the continuing rise in NOx emissions worldwide and increases in vehicle miles driven and electricity production from fossil-fuel combustion in the United States, ambient ozone concentrations in cities in the United States and other developed countries have generally decreased. The reductions have occurred as the result of decreases in VOC emissions, which tend to have a local effect on reducing ozone, and NOx controls, which lead to a more regional re- duction. In many cities, particularly in the eastern United States, ozone origi- nates primarily from distant VOC and NOx sources; it is formed as prevailing winds transport pollutants into the cities. Local VOC and NOx emissions cause more ozone to be added (e.g., NRC 1991, 2004a and references therein), and this leads to regionally high ozone concentrations with a noticeable increase in and downwind of cities (Figure 3-4; also see Box 3-1). OZONE MEASUREMENT In the United States, ozone concentrations are typically reported as a “mixing ratio,” the ratio of the number of molecules of ozone to the number of all molecules in the same volume of air. For example, the typical unit used for regulatory purposes is parts per million, although scientific studies often use parts per billion. The prior NAAQS was an average of 0.08 ppm over 8 h, which, with rounding, was effectively 84 ppb, whereas the current NAAQS is 0.075 ppm. Other nations often use units of mass per volume, such as micrograms per cubic meter.1 Ground-level ambient ozone has for many decades been one of the most well-characterized pollutants in the United States, especially in more populated areas where multiple monitors are used. For example, the Los Angeles Basin has over 30 ozone monitors. Ozone is also measured in rural locations and in na- 1 Ozone at 1 ppb is about 2 µg m-3 (within about 2% at standard temperature and pres- sure).

54 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits BOX 3-1 Policy Relevant Background EPA defines the policy relevant background (PRB) as the level that a pollutant concentration would be in the absence of anthropogenic emis- sions from the United States, Mexico, and Canada (see Chapter 2). How- ever, because of intercontinental transport of ozone and its precursor emissions, it is not feasible to directly measure the amount of ambient ozone attributable only to PRB sources. Therefore, modeling is the only viable approach to estimating a PRB. Observations in very remote places can be used to assess how well a model simulates chemistry and transport in clean regions. In the 2006 Criteria Document, EPA used results from the GEOS-Chem global air quality model (Fiore et al. 2003), with U.S., Cana- dian and Mexican anthropogenic emissions removed, to provide temporally and spatially varying PRB ozone levels (EPA 2006a). As applied, GEOS- Chem used a spatial resolution of 2o latitude by 2.5o longitude (about 200 km x 275 km). Criticisms of this approach include the model’s coarse spa- tial resolution and reliance on the assumption that Canadian and Mexican emissions would be potentially removed by U.S. policies (e.g., Brauer et al. 2007). Further, some analyses of observations at remote sites suggest that the calculated PRBs are low (Lefohn 2006; Brauer et al. 2007), noting that the model may underestimate the impact of some natural processes. Although the approach used by EPA in calculating the PRB was rea- sonable given the information available at the time, there is a need to ad- dress the criticisms levied in order to provide a better foundation for a simi- lar exercise in the future. The inclusion of Canadian and Mexican emissions removal in the process is a policy decision, and the influence of that choice is uncertain. GEOS-Chem’s coarse resolution is of some con- cern, because the effect of using a model with a coarse resolution is uncer- tain, though given the lack of local emissions substantial concentration gradients are not expected. Although comparison of simulated ozone lev- els at a remote site to those observed did not show significant bias (Fiore et al. 2003, Hudman et al. 2004), demonstrating that the simulated distribu- tions of observed ozone agree with observations at a variety of remote sites would strengthen the foundation of the modeling approach(es) cho- sen for future implementation. In regards to how to scale ozone levels be- tween current levels and those meeting various proposed NAAQS, an as- sessment of how ozone will respond to the alternate path(s) of emissions removal would be of interest (e.g., compare removing 50% of the NOx, versus 50% of the VOC versus 50% of both). Such information can be used in the uncertainty assessments conducted as part of the NAAQS re- view process. tional parks, for example, as part of the IMPROVE network (Sisler and Malm 2000). In many locations, however, ozone monitors are operated only during the ozone season, which varies from place to place. During the ozone season, moni- tors provide nearly continuous measurement, although levels are typically re-

Ambient Ozone and Related Pollutants 55 ported hourly. Thus, information on ozone, both spatial and temporal, is rela- tively complete in contrast with information on other pollutants. The wealth of information on ozone can be contrasted with that on airborne fine PM with an aerodynamic diameter less than 2.5 µm (PM2.5), which has been widely moni- tored for less than a decade and of which many of the measurements are for 24-h periods as opposed to hourly in the case of ozone. Chemiluminescent monitors or UV absorption monitors are typically used for routine ozone measurements. A recent Environmental Protection Agency (EPA) analysis found that when averaged over 8 h monitor precision is better than 3%, which is why the ozone design value (for an emission-control pro- gram) determined from an area’s monitoring data is within about 1.3 ppb (about 3 µg m-3) (EPA 2007a). For regulatory purposes, data are rounded to the last figure in the level of the standard; thus, the previous standard of 0.08 ppm has led to an effective standard of 0.084 ppm (EPA 2007a). In other words, a meas- ured value of 0.084 ppm would be rounded down to meet the 0.08 standard. This improved measurement accuracy has prompted EPA and its Clean Air Science Advisory Committee to recommend making the revised ozone NAAQS more precise, that is, adding an extra digit to the level of the standard. In March 2008, EPA set the level of the standard at 0.075 ppm (EPA 2008a). SOURCES OF OZONE PRECURSORS AND OZONE CONTROL There are myriad major outdoor sources of VOCs, including vegetation, solvent use, and mobile sources (Figure 3-5). Ambient sources of NOx include fuel combustion (for example, in cars, trucks, construction equipment, factories, and power plants) and to a lesser extent biogenic activity (Figure 3-5). The dis- tribution of sources and the other factors that contribute to ozone cause the high- est ozone concentrations to be in or just downwind of larger cities, particularly cities that are sunny during summer, such as Los Angeles, Houston, cities in the Northeast Corridor, and Atlanta. However, smaller cities can experience ozone concentrations above the previous NAAQS of 0.08 ppm; with the promulgation of a tighter ozone standard, even more areas are expected to be out of attainment of the NAAQS for ozone (see Figures 2-1 and 2-2 in Chapter 2). Early ozone-control programs concentrated on VOC-emission reductions because they were thought to be cost-effective, decreased concentrations of many additional air toxicants, and, unlike control of NOx, were unlikely to raise ozone concentrations. However, as the regional nature of the ozone problem became apparent, the focus shifted to NOx-emission controls, which had a greater influence on regional ozone concentrations. For example, the recent Clean Air Interstate Rule (CAIR) (EPA 2007h), which is designed to reduce ozone and PM2.5, focuses on reducing NOx and SO2 from electricity-generating units and other large sources in 28 eastern states and the District of Columbia.

56 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits VOCa NOx FIGURE 3-5 Anthropogenic sources of 2002 ozone-precursor emissions of VOCs and NOx. Biogenic sources are not included but dominate nationwide VOC emissions. Source: EPA 2003c, Figures 2-19 and 2-36. Automobile standards, which originally limited VOC and CO emissions, now include limits on NOx and PM emissions. If the imposition of CAIR, tighter automobile and truck emission standards, and other controls are effective to the extent expected by EPA, they will lower ozone substantially in the near future (Figure 3-6). Those controls would reduce NOx by about 42% and anthropo- genic VOCs by 28% by 2020 (Woo et al. 2006). Given that those steps are not likely to bring all areas into attainment, state improvement plans will need to include further controls. RELATED POLLUTANTS In assessments of the mortality risks associated with ozone exposure, it is important to consider how emission sources, meteorology, and the chemistry of ozone formation affect the formation of other pollutants of concern, including components of particulate matter (for example, the sulfate fraction of PM2.5) and other gases, such as aldehydes, and acids (Figure 3-1). The sharing of many of

FIGURE 3-6 Left, ozone and PM2.5 nonattainment areas in the eastern United States in 2006. Right, implementation of CAIR and other con- trols is expected to bring future ozone concentrations in many areas in the eastern United States into attainment of the previous 0.08-ppm ozone NAAQS. Source: EPA 2005c. 57

58 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits the sources and the intertwining of chemical and meteorologic processing lead to strong correlations between ambient ozone concentrations and concentrations of the other pollutants, such as sulfate and particulate organic carbon in summer (see Tables 3-1a, b, c and Table 3-2). The strong linkages and the fact that the other pollutants may have health effects are captured in a regulatory context in TABLE 3-1a Day-to-Day Correlation Between Air Pollutants in Boston, MA, by Season (Summer Nonshaded, Winter Shaded), 1999-2004a O3 PM2.5 SO42- BC PN NO2 CO SO2 O3 1.00 0.55b 0.59b -0.06 -0.34b 0.10 -0.03 0.01 PM2.5 -0.62b 1.00 0.74b 0.43b -0.37b 0.27b 0.36b 0.11 2- b b b b b SO4 -0.43 0.78 1.00 0.34 -0.31 0.25 0.14 0.13 b b b b b BC -0.60 0.67 0.64 1.00 0.11 0.55 0.39 0.17 PN -0.13 -0.04 0.10 0.10 1.00 0.25b 0.03 0.22b NO2 -0.56b 0.60b 0.57b 0.64b 0.33b 1.00 0.54b 0.32b CO -0.58b 0.50b 0.48b 0.72b 0.23b 0.59b 1.00 0.33b SO2 -0.42b 0.64b 0.50b 0.47b 0.32b 0.63b 0.60b 1.00 a Pearson correlation coefficients. Summer defined as May-August, winter as November- February. BC = black carbon. PN = particle number. bindicates p-value < 0.0001. TABLE 3-1b Day-to-Day Correlation Between Air Pollutants in St. Louis, MO, by Season (Nonwinter Clear, Winter Shaded), April 2001-September 2002a O3 PM2.5 SO42- NO3- BC OC NO2 CO SO2 O3 1.00 0.39b 0.37b -0.54b 0.02 0.30b -0.10c -0.21b 0.06 PM2.5 -0.29 c 1.00 0.72b 0.06 0.50b 0.60b 0.23b 0.24b -0.03 2- b b c c SO4 -- -- 1.00 -- 0.47 0.69 0.18 0.22 0.02 b b NO3- -- -- -- 1.00 0.07 -0.06 0.28 0.32 -0.01 b b b b b BC -0.39 0.61 -- -- 1.00 0.77 0.69 0.52 0.03 OC -0.20 0.68b -- -- 0.92b 1.00 0.47b 0.42b 0.17c NO2 -0.23 c 0.53b -- -- 0.70b 0.70b 1.00 0.49b 0.16c CO -0.35b 0.53b -- -- 0.86b 0.79b 0.53b 1.00 0.05 b c c b c SO2 -0.02 0.30 -- -- 0.29 0.27 0.35 0.25 1.00 a Pearson correlation coefficients. Winter defined as November-March, nonwinter as April-October. Data from St. Louis Supersite. PM2.5 measured with beta-attenuation gauge monitor. BC = black carbon. OC = organic carbon. Sulfate and nitrate data not available during winter 2001-2002. Sample sizes for pairwise comparisons generally above 70 in winter and from about 110 to 390 in nonwinter (except 53 for SO42- and OC comparisons in nonwinter). bindicates p-value < 0.0001. cindicates p-value < 0.05.

Ambient Ozone and Related Pollutants 59 TABLE 3-1c Day-to-Day Correlation Between Air Pollutants in Los Angeles, CA, by Season (Nonwinter Clear, Winter Shaded), June 2002-December 2003a O3 PM2.5 SO42- NO3- EC OC NO2 SO2 O3 1.00 0.08 0.32c -0.08 -0.56b -0.32 -0.38b -0.07 PM2.5 -0.32b 1.00 0.87b 0.78b 0.33c 0.49b 0.46b -0.04 2- b b SO4 -0.08 0.79 1.00 0.68 -0.01 0.17 -0.05 0.09 b b c c c NO3- -0.27 0.97 0.82 1.00 0.33 0.44 0.31 -0.02 EC -0.80b 0.65 c 0.21 0.52c 1.00 0.87b 0.86b 0.28c OC -0.60c 0.82 b 0.40c 0.69b 0.84b 1.00 0.89b 0.31c NO2 -0.64b 0.65 b 0.41c 0.64c 0.87b 0.90b 1.00 0.21b SO2 -0.30b 0.28 c 0.25 0.27 0.51c 0.41c 0.55b 1.00 a Pearson correlation coefficients. Winter defined as November-February, nonwinter as March-October. Data from EPA Speciation Network. EC, elemental carbon. OC, organic carbon. In winter, sample sizes for pairwise comparisons with particles 27 or 28 and for comparisons between gases, about 180; in summer, samples sizes for comparisons with particles above 60 and for comparisons between gases, 370. bindicates p-value < 0.0001. c indicates p-value < 0.05. that the NAAQS is for ozone and other photochemical oxidants; ozone is used as an indicator of the presence of the wide array of photochemical oxidants pre- sent in ambient air. (Other photochemical oxidants include a wide variety of both organic and inorganic species, such as nitrogen dioxide, organic and inor- ganic acids [sulfuric, nitric, formic, etc.], organic and inorganic peroxides, other reactive oxygen species such as quinones, and many more. [see also Figure 3-1 for chemical linkages].) Similarly, epidemiologic studies of ozone-health asso- ciations tacitly include the potential exposure to co-occurring pollutants unless they control for such additional exposure. However, the complete suite of other pollutants has never been monitored sufficiently to be included in such studies. Many of the photochemical oxidants that would be expected to covary with ozone are hardly, if ever, monitored. Correlations between ozone and other pollutants, both measured and un- measured, have led to an awareness of the possible confounding of ozone-health effects analyses. As shown on Tables 3-1a-c, correlations between ambient ozone and other pollutants, such as PM, vary substantially by geographic loca- tion, season, and pollutant. For example, associations between ambient ozone and PM2.5 differ substantially in the Western and Eastern United States, between summer and winter, and for different PM components. Also, correlations among ambient ozone and other pollutants have been shown to vary by averaging pe- riod. These results suggest that the potential for confounding of ozone-health effects also varies by these factors, as discussed in Chapter 4. Strong correla-

60 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits tions between ambient ozone and its co-pollutants further suggest that it may be difficult to separate the effects of ambient ozone from its co-pollutants in epi- demiologic studies based on using ambient concentrations as an estimate of ex- posure. Of ozone’s copollutants, ambient PM2.5 has been of particular concern with regard to confounding because of its observed effects on health and the often strong, seasonally varying correlations between ambient concentrations of ozone and PM2.5 (Tables 3-1 and 3-2). PM2.5 is a mixture of many compounds that originate from a variety of primary emission sources and reactions in the atmosphere. Not only do the formation and fate of specific compounds vary temporally and spatially but health assessments suggest that the components have different health effects (e.g., Laden et al. 2000; Peel et al. 2005; Thurston et al. 2005; Sarnat et al. 2008). Strong summer correlations between ozone and PM2.5 may be attributed to similar formation processes. For example, in New York and other eastern U.S. states (Figure 3-6), PM2.5 in summer consists largely of sulfate, which originates primarily from the oxidation of emitted SO2 by the hydroxyl radical (Figure 3-1, Reaction 3-6), the same molecule that leads to oxidation of VOCs: SO2 + OH ... H2SO4 sulfate aerosol. [3-6] In winter, particulate nitrate formation from emitted NOx can be important. It, too, is oxidized by the hydroxyl radical: NO2 + OH HNO3 ... nitrate aerosol. [3-7] In addition, some organic products (OPs) formed in Reaction 3-4 can lead to SOA production and thus also contribute to PM2.5 concentrations. In contrast, ozone concentrations can be depressed in areas with high NO and NO2 (such as areas close to highways). That also means that ozone concentrations can be lower in areas that have pollutants coming from substantial NOx sources. For example, high concentrations of ultrafine particles (less than 100 nm in diame- ter) are found near freeways, which also have higher concentrations of CO and NOx (Zhu et al. 2002; Sardar et al. 2004). However, ozone and PM2.5 can corre- late positively away from the source, for example, because of similar formation routes and precursor emission sources of ozone and sulfate (Figure 3-1). The relation between ambient ozone and PM2.5 can differ substantially be- tween winter and summer, especially in areas with low winter sulfate, such as the eastern United States (Tables 3-1a,b,c). As a result of lower sulfate forma- tion, primary PM2.5 sources—such as motor vehicles, wind-blown dust, and combustion—contribute a larger fraction of ambient PM2.5. Furthermore, secon- dary nitrates increase in winter because of a decrease in competition for ammo- nia (NH3) by sulfate and because of the decrease in temperature, both of which

TABLE 3-2 Pollutant Correlations at Jefferson Street SEARCH Site, Atlanta, GA, 1999-2006a 24-h O3 1-h max O3 8-h max O3 24-h O3 1-h max O3 8-h max O3 (O3 season) (O3 season) (O3 season) PM2.5 0.31 0.52 0.47 0.60 0.74 0.73 SO4 0.56 0.65 0.63 0.55 0.61 0.61 Organic carbon PM (OC) 0.02 0.24 0.19 0.55 0.70 0.70 24-h O3 1.00 0.87 0.91 1.00 0.80 0.87 1-h max O3 0.87 1.00 0.98 0.80 1.00 0.97 8-h max O3 0.91 0.98 1.00 0.87 0.97 1.00 24-h O3 (O3 season) 1.00 0.80 0.87 1.00 0.80 0.87 24-h NO -0.31 -0.06 -0.11 -0.07 0.25 0.23 24 h average NO2 -0.13 0.12 0.07 0.16 0.45 0.42 24-h average PM10 0.31 0.53 0.49 0.55 0.74 0.73 a This SEARCH site (Hansen et al. 2006) is the only location that has such a long record of daily speciated PM2.5, PM10, NOx, and O3 for such an analysis. 61

62 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits increase the stability of ammonium nitrate. At the same time, the decrease in solar radiation leads to lower hydroxyl-radical concentrations and reduced VOC oxidation and thus to lower NO conversion to NO2. The primary NO emissions then scavenge ozone. Thus, on days when primary PM and NO are poorly dis- persed, concentrations of both increase, and the increased NO lowers ozone (of- ten to well below background concentrations). Ambient concentrations of ozone and PM2.5 may not correlate or may even correlate negatively in winter (Tables 3-1a,b,c). The seasonal variability in the ambient ozone-PM2.5 association sug- gests that the potential for confounding of ozone health effects by PM2.5 also differs by season, with confounding most likely when they correlate strongly in summer (see also Bell et al. 2007). The potential for confounding of ambient-ozone health effects by ambient PM2.5 also probably depends on particle composition, which varies with location and season (Figure 3-7). As shown above, ambient ozone concentrations tend to correlate strongly and positively with ambient sulfate but not with ambient traf- fic-related elemental carbon, where a negative correlation is often found. Since studies have suggested that PM2.5 health impacts may differ by component, these results indicate that confounding by specific PM2.5 components, including sul- fate, elemental carbon, metals, and secondary organics, should also be examined. It will be difficult to address such confounding with currently available data, however, because data on the PM2.5 component have only recently been col- lected routinely in many sites. EPA’s Speciation Trends Network (STN) for monitoring PM2.5 components started in 2000, and individual monitors came on line at different times. Most monitors measure a standard set of characteristics, including PM2.5 mass and various components (Peterson et al. 2000; Flanagan et al 2006). In the near future, new speciated measurements may provide the needed data, but, unlike ozone data, such data are now generally available only once every 3 or 6 d. To understand how short-term variation in PM2.5 compo- nents might confound ozone-mortality associations, more frequent measure- ments may be needed. One other approach to avoiding the confounding by PM2.5 is to consider the spatial distribution of the two pollutants, although they have similar spatial distributions (Figure 3-4b), which are tied to their similarities in sources, transport, and chemistry. Confounding of epidemiologic studies of ozone-mediated mortality and other health effects by PM2.5, especially by such specific components of PM2.5 as sulfate, presents formidable challenges to the interpretation of both time-series and spatial distributions. A further challenge to epidemiologic studies is that there are hundreds of gaseous and particle-bound ozone-related pollutants, including peroxides, other reactive oxygen species, and oxidized organics—most of which are not routinely monitored. The correlations between ozone and PM, as well as the lack of characterization of the variety of ambient pollutants that may also contribute to adverse health outcomes, presents a challenge to epidemiologic studies of PM effects and to studies of ozone ef- fects because of the potential for confounding by many of the pollutants in the atmospheric mixture.

Ambient Ozone and Related Pollutants 63 FIGURE 3-7 Composition of PM2.5 in representative urban and rural locations. Urban sites are Toronto, ON, Canada; Washington, DC; Atlanta, GA; Mexico City, Mexico; Los Angeles, CA; and Fresno, CA. Averaging periods and average PM2.5 mass are indicated. All sites have at least 1 y of sampling except Mexico City, for which average was deter- mined for 14 d in 1 mo. More recent short-term measurements from December 1995 and January 1996 at Fresno and Kern Wildlife Refuge in California show lower PM2.5 mass concentrations than displayed here but similar composition. Colorado Plateau data are averages of IMPROVE sites in Bryce Canyon, Canyonlands, Grand Canyon, Petrified Forest, Mesa Verde, and Zion National Parks. Source: NARSTO 2003. Reprinted with permission; copyright 2005, Cambridge University Press. OZONE CONTROL EPA uses morbidity and mortality risk estimates to assess the benefits of its pollutant-control programs and thus the benefits of reductions in specific emissions from specific sources. That approach has a subtle but profound effect on the metric chosen to link ambient ozone concentrations with mortality risk. Various studies (e.g., Bell et al. 2005; Levy et al. 2005) have considered differ- ent metrics to characterize short-term ozone exposure, including the 24-h aver-

64 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits age, the 8-h maximum (similar to the NAAQS), and the 1-h maximum (similar to the prior NAAQS). In meta-analyses of studies of the relationships between ozone and mortality (e.g., Bell et al. 2005), ozone concentrations were linearly scaled to the same metric to allow comparisons across studies; for example, a 24-h ozone average was used by Bell et al. (2005). Scaling of the 1-h maximum, the 24-h average, and the 8-h maximum ambient ozone concentrations is reason- able given their typically strong correlations for a given location, but it is not generalizable to ozone-control strategies in that the 24-h average and 1-h and 8- h maximum ambient ozone concentrations will respond differently to specific control strategies (Table 3-3) (Liao et al. 2008). For example, although measures to reduce ground-level NOx emissions in a city, such as that from cars, may re- sult in lower daily maximum ambient ozone concentrations on sunny stagnant days (for example, see Figure 3-8) (Cohan et al. 2005) in or downwind of such cities as Atlanta, Houston, and New York (Sillman et al. 1990; NRC 1991), the measures can also result in higher 24-h averages because of the higher nighttime ozone concentrations due to reduced ozone scavenging at night (as discussed above). Furthermore, on days that are cloudy or otherwise less conducive to ozone formation, NOx-reduction measures can result in both higher peak and higher average ozone; that is, the sensitivities of ozone to NOx controls are nega- tive (Figure 3-8). Thus, strategies aimed at meeting the NAAQS (based on re- ducing the fourth-highest 8-h maximum) can lead to increases in average ozone level concentrations (Table 3-3). This issue is directly related to assessments of the characteristics of ozone exposure that lead to death and other end points of concern. As shown in Table 3-2 and noted in various studies (e.g., Bell et al. 2004), various ozone metrics for data collected in a specific location tend to be highly correlated, so it may seem unimportant to distinguish between them; given the degree of correlation, it may be difficult for an epidemiologic study to do so. However, the choice of metric can be influential in determining benefits of control programs. For exam- ple, a program that lowers NOx emissions could reduce peak ozone concentra- tions but raise average concentrations. A cost-benefit analysis based on mortal- ity and average ozone could appear to show a negative benefit although the association with peak summertime concentrations appears to be positive. As discussed later, it is unknown which is more accurate, so evaluating the benefits of urban NOx reduction is uncertain in both magnitude and sign. Complicating the process is that the association might depend on the metric of exposure, that is, whether the response is taken to have a threshold or not. In the recent NAAQS review process, EPA assumed a linear nonthreshold response (EPA 2007a). Control programs for ozone will affect other pollutant concentrations be- cause ozone precursors are also responsible for formation of other species; pol- lutants have many sources in common. The effects of these ozone-control pro- grams on other species will vary in space and time and, as discussed above for ozone, can even change direction, depending on conditions. In general, reduc-

Ambient Ozone and Related Pollutants 65 TABLE 3-3 Simulated Sensitivity of Annual Average Ozone to NOx and Sensitivity of Fourth-Highest 8-h Average Ozone to NOxa Los Atlanta New York Chicago Houston Angeles 2001 annual average -0.06 -0.11 -0.05 -0.07 -0.11 ozone 2001 fourth-highest 8-h 0.42 0.26 0.28 0.05 -0.29 average ozone a Negative sensitivity means that ozone will go up if NOx emissions are reduced. Sensi- tivities provided are local (for example, in given conditions, how ozone will respond to emission changes) and are given as parts per billion per 1% change in domainwide NOx emissions. As shown, annual average ozone goes up in each location, although fourth- highest 8-h averages typically are reduced. Source: Adapted from Liao et al. 2008. Re- printed with permission; copyright 2008, American Chemical Society. 1.0 1.0 1.0 Atlanta (2001) Atlanta (2050) Atlanta (2001) SMDAO3, NOx (ppb) 0.5 0.5 0.5 0.0 0.0 0.0 0 50 100 150 0 50 100 150 0 50 100 150 -0.5 -0.5 -0.5 -1.0 -1.0 -1.0 1.0 1.0 1.0 Chicago (2001) Chicago (2050) Chicago (2001) SMDAO3, NOx (ppb) 0.5 0.5 0.5 0.0 0.0 0.0 0 50 100 150 0 50 100 150 0 50 100 150 -0.5 -0.5 -0.5 -1.0 -1.0 -1.0 1.0 1.0 1.0 Houston (2001) Houston (2050) Houston (2001) SMDAO3, NOx (ppb) 0.5 0.5 0.5 0.0 0.0 0.0 0 50 100 150 0 50 100 150 0 50 100 150 -0.5 -0.5 -0.5 -1.0 -1.0 -1.0 1.0 1.0 1.0 Los Angeles (2001) Los Angeles (2050) Los Angeles (2001) SMDAO3, NOx (ppb) 0.5 0.5 0.5 0.0 0.0 0.0 0 50 100 150 0 50 100 150 0 50 100 150 -0.5 -0.5 -0.5 -1.0 -1.0 -1.0 1.0 1.0 1.0 New York (2001) New York (2050) New York (2001) 0.5 0.5 0.5 SMDAO3, NOx (ppb) 0.0 0.0 0.0 0 50 100 150 0 50 100 150 0 50 100 150 -0.5 -0.5 -0.5 -1.0 MDA8hr O3 (ppb) -1.0 MDA8hr O3 (ppb) -1.0 MDA8h O3 (ppb) FIGURE 3-8 Calculated daily sensitivities of maximum 8-h averaged ozone in Atlanta, Chicago, Houston, Los Angeles, and New York regionwide changes in NOx emissions and corresponding simulated concentrations. Shown are responses both at the city center and at the location of the regional maximum for 2001, and for the city-center case in 2050 in response to a 51% reduction in NOx emissions. Results shown are for change in ozone in parts per billion per 1% change in domainwide NOx emissions. Although highest 8-h ozone average may decrease with NOx controls, lower peak daily concentrations can increase. Ozone formations at the location of the regional maximums tend to be more NOx-limited. Source: Liao et al. 2008. Reprinted with permission; copyright 2008, American Chemical Society.

66 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits tions in NOx will directly reduce nitrate formation (e.g., Russell and Cass 1986; Boylan et al. 2006), and there is evidence that they will slightly increase sulfate. For example, Liao et al. (2007b) found that the effect of a 1% reduction in NOx emissions would be a 0.2-ppb reduction in average fourth-highest 8-h maximum ozone, a 0.01-µg/m3 reduction in average annual nitrate aerosol, and a slight increase in average annual sulfate, with some areas experiencing small increases and others seeing decreases. Reductions in anthropogenic VOC emissions would reduce SOA formation, although the relationship between anthropogenic and biogenic VOC emissions and SOA formation is uncertain, and it would depend on which VOC species were being controlled (e.g., Robinson et al. 2007). Some organic compounds (such as benzene) are toxicants or (as in the case of formal- dehyde) result in formation of toxicants in the atmosphere. Effects on toxic compounds will depend on the type of control and the source. For example, re- ducing automobile VOC emissions will probably reduce benzene, and changing paint solvents will affect the concentrations of other compounds. VOC controls will lead to minor changes in sulfate and nitrate concentrations (Napelenok et al. 2007). Much less is known about how ozone-related controls will affect concen- trations of reactive oxygen species (ROS), in large part because of the general lack of specific knowledge about ROS. The chemistry of oxidant species and the presence of their precursors suggest that reductions in VOC emissions would lead to reductions in organic ROS. If the ROS is associated with metals, the ef- fect of ozone-related controls is less obvious, because the dominant precursor sources are less common and the chemistry not as well explored. Some PM-related controls (such as NOx controls, as discussed above) will affect ozone concentrations and (as in the case of control of fly ash) could re- duce metal-associated ROS. SO2 reductions will have a minor effect on ozone concentrations (Napelenok et al. 2007), and the direction of the response is un- certain (Dickerson et al. 1997; Jacobson 1998). OZONE DYNAMICS AND MONITORING: IMPLICATIONS FOR HEALTH STUDIES Time-series studies of ozone typically assess health effects in terms of ambient concentrations measured over time at a central monitoring site or at several sites in the study area. The use of ambient concentrations to estimate exposures is a source of considerable uncertainty in understanding ozone- mediated mortality with respect to both how well they indicate ozone exposures over time and space and how well ozone effects can be separated from effects of other pollutants or weather conditions. The relation of ambient concentrations to ozone exposures of a study population depends on several factors, including the spatial distribution of ambient ozone and the typical activity patterns and home- ventilation patterns of the study community. Spatial variability in ambient ozone concentrations is generally less important in that concentrations of ambient

Ambient Ozone and Related Pollutants 67 ozone and other secondary pollutants tend to be relatively homogeneous (e.g., Wade et al. 2006) over large areas. Over greater distances, ambient ozone con- centrations do vary, being sometimes higher in suburban and rural areas than in urban areas (Liu et al. 1993; Waldman et al. 1990) and higher away from roads than near roads. Transport of ozone and its precursors can also lead to a gradient in concentrations in an area. Such causes of spatial variability in ozone concentrations in a community increase exposure error in time-series studies based on ambient concentrations, because this spatial variability can vary over time. The increased error may re- sult in biased risk estimates (Wakefield and Shaddick 2006). Nevertheless, spa- tial variability in ozone concentrations is generally a minor concern in time- series epidemiologic studies because other factors, such as indoor concentrations and activity patterns, usually introduce larger errors in estimates of ozone expo- sure over time. Indoor ozone generally originates outdoors; indoor sources— such as photocopiers, laser printers, and some air cleaners—are not present in most homes. Indoor residential ozone concentrations are thus determined primarily by outdoor concentrations and by factors that affect the ability of ozone to penetrate into and persist in the home, including removal by reaction on surfaces, air exchange between indoors and outdoors, air filtration, and reactions between ozone and other indoor pollutants. Under normal conditions, the half-life of ozone indoors is only 7-10 min (Weschler 2000). For averaging periods of 24 h, the short half-life results in generally low indoor concentrations (and often below the sampling method limit of detection).2 These low concentrations reflect high rates of ozone removal by surface reactions. Indoor ozone concentrations are typically 10-50% of outdoor concentrations (Weschler 2006). Even at their lows, 24-h indoor ozone concen- trations exhibit some diurnal and seasonal variation that reflects the outdoor variation. Indoor ozone concentrations tend to be higher in summer than in win- ter, when they are generally at or below the limit of detection of some measure- ment methods in personal exposure studies (Liu et al. 1993; 1997; Avol et al. 1998; Sarnat et al. 2006). For example, in a Southern California study of 126 homes (February-December 1994), mean 24-h indoor ozone concentrations (13 ± 12 ppb) were lower than corresponding outdoor concentrations (37 ± 19 ppb) (Liu et al. 1997). Indoor:outdoor ratios were greater during the summer pollu- tion period. Higher summer ratios indicate the importance of home ventilation, especially in homes that do not have air conditioning or tight sealing. In non-air- conditioned and other well-ventilated homes, 24-h indoor concentrations can be a larger fraction of corresponding outdoor concentrations (Liu et al. 1993; Gold et al. 1996; Sarnat et al. 2006; Weschler 2006). Similarly, in shorter averaging periods, indoor residential ozone concen- trations can be a larger fraction of those outdoors when windows are opened. In 2 With the exception of periods when indoor concentrations of nitric oxide (NO) are elevated, indoor ozone concentrations are normally not below the limit of detection of UV photometric methods (Weschler and Shields 1994).

68 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits a study by Zhang and Lioy (1994), for example, indoor ozone concentrations in six New Jersey homes measured during the high-ozone afternoon hours (2:30- 7:30 pm) were found to track outdoor concentrations, with a mean ratio of 5-h indoor to outdoor ozone concentrations of 0.22 (+ 0.09) to 0.62 (+ 0.11). The indoor:outdoor ratios were highest (0.59 + 0.16) when windows were open and gas stoves were off. For homes with air conditioning and homes with closed windows and no air conditioning the indoor:outdoor ratios over the 5-h high- ozone periods remained relatively low, 0.28 (+ 0.12) and 0.26 (+ 0.12), respec- tively. Use of a gas stove lowered indoor:outdoor ratios further for all ventila- tion conditions. The low indoor ozone concentrations explain why 24-h personal ozone exposures are low: people spend on the average 87% of their time indoors (Klepeis et al. 2001). The average amount of time people spend outside during their day varies little across the United States (Klepeis et al. 2001). Therefore, daily averaged ozone concentrations to which people are exposed (averaged over the time spent indoors and outdoors) are generally low relative to outdoor concentrations averaged over the same period regardless of geographic location, with measured personal ozone concentrations generally below the limit of detec- tion of some measurement methods. Like 24-h outdoor and indoor concentra- tions, 24-h personal ozone exposures are higher in summer than in winter, al- though summertime concentrations to which individuals are exposed (averaged over 24-h) are still a relatively small fraction of corresponding 24-h outdoor values (Liu et al. 1993, 1997; Sarnat et al. 2000; Sarnat et al. 2006). For example, in a Steubenville, OH, cohort of older adults which was conducted over a seven month period, the mean 24-h personal:ambient ratio of ozone concentration, 0.24, was substantially lower than that of any of the other measured pollutants (PM2.5, EC, sulfate, NO2, and SO2). It is important to note that over periods of less than a day, such as 1 h, per- sonal ozone exposures can be relatively high; Hourly personal exposure concen- trations were shown in a scripted-exposure study to be comparable with corre- sponding outdoor concentrations when the trained technician spent the time outdoors (Chang et al. 2000). Based on these findings, it is possible that hourly and peak ambient ozone concentrations are appropriate surrogates for personal ozone exposure concentration, especially in summer and during peak afternoon high ozone hours, which account for much of the individual outdoor time (Fig- ure 3-9). Further study should examine whether this is in fact the case, espe- cially for individuals who are susceptible to ozone exposures. Hourly personal exposures in the scripted-exposure study were also strongly associated with hourly outdoor concentrations when the trained techni- cian was outdoors (Chang et al. 2000). In contrast, when the technician spent the hour indoors or in transit, associations between hourly personal exposures and ambient concentrations were not significant. Similarly, over 24 h, the vast ma- jority of which is spent indoors, many studies suggest that personal ozone expo- sure concentrations are not strongly associated with corresponding ambient con-

Ambient Ozone and Related Pollutants 69 FIGURE 3-9 Activity patterns by time of day. Adapted from U.S. National Human Ac- tivity Pattern Survey. Source: Klepeis et al. 2001. Reprinted with permission; copyright 2001, Journal of Exposure Science and Environmental Epidemiology. centrations regardless of whether individuals are followed over time or at one point in time, with consistently low slopes and R2 values for regressions of per- sonal ozone on ambient concentration (Brauer et al. 1989; Liu et al. 1993, 1997; Linaker et al. 2000; Patterson and Eatough 2000; Sarnat et al. 2000; Sarnat et al. 2006). It is important to note that these low R2 values may be due in part to the low measured personal exposures, which were often below the limit of detection of the measurement methods used in those studies. Despite this, the low R2 val- ues do suggest that estimates from time-series studies may be biased toward the null hypothesis. More recent studies suggest that the associations are substantially weaker than those for PM2.5 and sulfate (Sarnat et al. 2000; Sarnat et al. 2006) and stronger in summer than in winter. The seasonal variability probably reflects increased home ventilation in the hotter summer months. Consistently with that, the slope of the regression between 24-h ambient ozone concentrations and cor- responding personal exposures in Steubenville, OH, was 100% higher for eld- erly people spending time in well-ventilated indoor environments (slope, 0.18 ± 0.03; t value, 7.34) than for other people (slope, 0.08 ± 0.04; t value, 1.89). As supported by Levy et al. (2005), those findings suggest that ventilation (and thus season) is an important modifier of ozone-mortality risk estimates, and further provide potential explanations for observed seasonal or between study variabil- ity in other studies.

70 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits Even in well-ventilated conditions, however, the slope of the ozone asso- ciation is small, and this suggests that only minor changes in ozone exposure occur in response to moderate changes in outdoor concentration. Together, low personal exposure concentrations and weak personal-ambient concentration rela- tionships suggest that 24-h ambient ozone concentrations are poor proxies for personal exposure. For shorter averaging periods, results of the one scripted- exposure study suggest that hourly or peak ambient ozone concentrations may be appropriate proxies for corresponding hourly or peak personal exposures. Additional short-term personal-ozone exposure studies are needed, given ozone’s large contribution to uncertainty in ozone-mortality risk estimates; how- ever, these studies will require the development of new measurement methods that have sufficient sensitivity to measure these likely low, short-term exposures. The relatively poor correlation of ambient ozone concentration with per- sonal exposure—combined with the strong associations among ambient ozone concentration, ambient PM2.5 concentration, and temperature—raises concerns about whether ozone mortality attributed to ozone is due instead to these or other correlated factors. Evidence from a series of 24-h multipollutant exposure panel studies performed generally over two seasons at locations—including Bal- timore, MD (Sarnat et al. 2001), Steubenville, OH (Sarnat et al. 2005), and Bos- ton, MA (Rojas-Bracho et al. 2004)—indicates that potential confounding by correlated pollutants and weather is an important source of uncertainty in ozone- mortality risk estimates. Results from longitudinal analyses have consistently shown, for example, that 24-h ambient ozone concentrations are substantially weaker proxies for corresponding exposures than PM2.5. In addition, studies have shown 24-h ambient ozone concentrations to be important proxies for per- sonal PM2.5 exposures over time. Associations between 24-h ambient ozone concentrations and personal PM2.5 exposures tended to be stronger in Boston and Steubenville than in Baltimore, possibly because of better home ventilation in the two cities. It is consistent with that idea that for all cities 24-h ambient ozone was a stronger proxy for personal PM2.5 exposures for individuals followed in summer than winter and for people spending most of their time in well- ventilated than in poorly ventilated environments. Because exposure studies of confounding have been based on 24-h values and for individuals living within the eastern United States, the generalizability of these results to exposures measured over shorter periods and to other populations is not known. OZONE-EXPOSURE MODELING One method for linking personal ozone exposure to ambient ozone con- centration, and thus linking ambient ozone to health end points, is to model ex- posure in ways that account for individual behavior and individual ozone expo- sure concentrations. Such simulation models are valuable for estimating air- pollution exposures in the absence of direct measurement. They can provide a framework for extending risk analyses across large geographic domains, even at

Ambient Ozone and Related Pollutants 71 the national level. They can also be used to help to estimate the effect of ambient pollutant reductions on personal exposure, as is done in developing the ozone criteria document. Such analysis has been used in advising the EPA administra- tor on proposing standards (e.g., EPA 2007a). Exposure-simulation models can be used in epidemiologic models to study the association between human exposure to ozone and mortality or morbidity instead of relying on a monitored ozone value and providing a potentially more accurate representation of individual ozone exposure. Such human-exposure models could be powerful tools for drawing inferences about the health risks associated with exposure to tropospheric ozone. That approach was taken in the recent EPA ozone staff paper as part of the policy assessment (EPA 2007a). Uncertainty in such models should be characterized and quantified properly be- fore they are used for ozone-mortality risk assessment for regulatory purposes. The ozone-exposure model developed by EPA to estimate human popula- tion exposure to the criteria and other toxic pollutants is called Air Pollutants Exposure (APEX). APEX is a probabilistic model designed to account for the numerous sources of variability that affect people’s exposures. APEX simulates the movement of people through time and space and estimates their exposure to a given pollutant in indoor, outdoor, and vehicular microenvironments. Daily activity patterns for individuals in a study area, one of the inputs to APEX, are estimated from detailed time-location-activity diaries that are compiled in the Consolidated Human Activity Database (CHAD) (McCurdy et al. 2000; EPA 2003d). The diaries contain information regarding age, sex, race, employment status, occupation, day of week, daily maximum hourly average temperature, each location during the day, and start time, duration, and type of activity per- formed. APEX estimates the concentration in the microenvironment associated with each event in an individual’s activity pattern and sums the event-specific exposures within each hour to obtain a continuous series of hourly exposures spanning the period of interest. Activity-specific simulated breathing rates of individuals are used in APEX to characterize intake during each period. The breathing, or ventilation, rates are derived from energy-expenditure estimates for each activity included in CHAD and are adjusted for age-specific and sex- specific physiologic characteristics associated with each simulated individual. Ozone concentrations in each microenvironment are estimated with mass- balance or transfer factors, and the user specifies prior probability distributions for the parameters to be used in the model; the prior distributions are used to model the uncertainties and variabilities in the parameters. APEX combines the estimated time series of exposure concentrations that a simulated individual experiences during the modeled period with the estimated time spent in each of the microenvironments visited according to the activity diary. The hourly average exposures of each simulated individual are time- weighted averages of the within-hour exposures. APEX then statistically sum- marizes and tabulates the 1-h, 8-h, and 24-h individual exposures, and then pro- vides a distribution of ozone exposure for the population in a given census as

72 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits characterized by the simulated individuals. The estimated exposure is a linear function of the ambient ozone concentrations. Although there is the potential to use those exposure simulation models for future analyses, APEX has not yet been used to characterize ozone exposure in epidemiologic models for national ozone mortality analysis. Therefore, it has not been given more extensive consideration in this report. The parameter uncertainty in APEX is represented by prior distributions assigned to most of the parameters in the model. The prior distributions given to the model parameters are based on regional studies; to extend the prior distribu- tions to the national level, more exposure studies should be conducted. There also is uncertainty in the formulation of the model and in the inputs. The differ- ent sources of error and uncertainties in APEX result from variability that is not modeled or is modeled incorrectly; from erroneous or uncertain inputs; from errors in coding; from simplifications of physical, chemical, and biologic proc- esses to form the conceptual models; and from flaws in the conceptual model. In particular, the uncertainty in the estimation of ambient air quality will be propa- gated by APEX. The APEX output may also be very sensitive to the prior distri- butions used in the microenvironmental models. APEX relies heavily on the assumption that exposure to ozone is a linear function of ambient concentration, and this assumption needs more justification. CHAD may be the best available source of human-activity data for use in exposure modeling, but issues regarding how well CHAD diaries represent the simulated populations remain. CHAD contains 20,000 people that are used to represent several million over long periods, the diary data are relatively old (some data were generated in the 1980s), and there are diary structure differ- ences (real-time data collection vs data collection by recall); each of these char- acteristics may lead to errors. The human diaries used by the model might be too limited to characterize and represent the simulated populations. EPA has con- ducted some validation studies of APEX (EPA 2007k). The committee agrees that the APEX approach is a useful way to assess population exposure, but re- maining uncertainties require further study to estimate their magnitude and, to the extent possible, reduce them. SUMMARY Interpretation of studies relating measurements of ambient ozone to health end points is made difficult by the dynamics of ozone and associated pollutants, particularly how temporal and spatial variations affect individual human expo- sure. (Other questions about the existing evidence are addressed in Chapter 4.) Correlations between ozone and PM2.5 vary spatially, seasonally, and with PM2.5 components, all of which can confound the interpretation of epidemiologic analyses. Furthermore, exposure studies which typically relied on passive ozone monitors found that ambient ozone concentrations are not highly correlated with concentrations indoors, where people spend most of their lives, or with observed

Ambient Ozone and Related Pollutants 73 personal exposures. Indoor ozone concentrations exhibit some diurnal and sea- sonal variations that reflect the outdoor variations. Correlations with ambient concentrations may be artificially low given that measured 24-h personal expo- sures concentrations and indoor concentrations were often below the method limit of detection. One way to improve characterization of indoor ozone expo- sure is to use continuous ozone monitors, which have greater sensitivity than passive ozone monitors. At the population level, characterization of ozone expo- sures can be improved using exposure modeling, although additional uncertain- ties are involved in model application. Accounting for those complexities pre- sents a formidable obstacle in interpreting results of epidemiologic studies of ozone and health. Finding: Potential confounding of ozone-mortality studies by PM presents a challenge, particularly in considering seasonal and component effects. Some species are much more (or less) correlated with ozone when specific seasons are considered (for example, summer, when ozone is high) than over an annual cy- cle. Considering only PM mass would not account for effects of specific PM components. Recommendation: Both exposure and epidemiologic researchers should investigate seasonal and regional associations between ambient concentrations and exposures to ozone and PM2.5, (and its components), how they affect control of confounding, and correlations between the various pollutant concentrations. When possible, researchers should address those issues by focusing on groups of individuals who are sensitive to ozone exposures and by using data on the chemical and physical components and size distribution of PM2.5. EPA and the scientific community should increasingly include the grow- ing Speciated Trends Network (STN) database in future analyses of potential confounding of the ozone associations. EPA should work with the scientific community to ensure that the STN collects data frequently enough on the parti- cle components most relevant to the potential for confounding. Finding: Personal exposure, considering time spent outdoors and indoors on a 24-h basis, is poorly correlated with monitored ambient ozone concentra- tions, with low slopes of the regression of personal exposure on ambient ozone levels. However, findings on exposure in the afternoon, when both outdoor ac- tivity and ozone concentrations can peak, suggest that exposures to higher ozone concentrations are better captured by ambient monitoring. Control programs based on reducing peak afternoon ozone concentrations on days when they are greater than the standard can result in an increase in the 24-h concentrations on days that are conducive to ozone formation and can also increase afternoon con- centrations on days less conducive to ozone formation. Recommendation: Future studies should determine whether and how much daily peak exposures, such as 1h or 8h exposures, and longer-term average exposures, such as over 24h, are associated with ozone-related mortality so as to guide control decisions for protecting public health. Benefits assessors at EPA and elsewhere should use the results to identify the appropriate exposure metrics

74 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits to estimate how efforts to attain the ozone NAAQS will affect ozone exposure and health. The committee notes that EPA has apparently not considered the use of two or more averaging times jointly. Does adding metric C to a model that al- ready contains metric A, or metrics A and B, improve the fit of the model? Such analyses should be fairly easy to conduct and might show that 1-hour, 8-hour, or 24-hour averaging times are roughly equivalent and that any one of these can substitute for the others. Conversely, one might find that one or two of these metrics add little when the third is already in the model. Or, it is possible that some combination of two or three of these is a better measure for analysis than any one alone. Measurement error is likely to blur the actual contributions of each metric and the relationships among them, but that is not relevant when the goal is to find the best empirical fit to the data. Regulators need to consider that control strategies may affect 24-h average concentrations quite differently from how they affect shorter-term exposures and that peak short-term concentrations on lower-ozone days will respond differ- ently, and often in the opposite direction, from the same measure of ozone on a high-ozone day. Finding: Human-exposure simulation models, such as APEX, present an approach to estimating ozone exposure different from reliance solely on obser- vations. However, such models introduce their own uncertainties, and they need to be further evaluated and their uncertainties characterized. Recommendation: EPA should conduct more detailed evaluations of APEX and other models used to improve characterization of ozone exposure at the population level by taking human activity into account. Detailed evaluation will require further human-exposure studies, which will need improved instru- mentation and approaches for monitoring personal exposures. The extent to which the human diaries represent the population under study, as used by APEX, warrants further explication, and more extensive and more up-to-date diary data should be collected, in particular for children and minority groups. Those data, when tied to data on pollution concentrations and ozone-alert days, can be used to assess the type and degree of changes in behavior to avoid ozone exposure. The cost of such changes in behavior associated with increased pollution cur- rently is not addressed in RIAs prepared by EPA. Finding: Epidemiologic studies of ozone health effects are limited, in part, by the reduced availability of ozone data in winter. Recommendation: EPA and states should extend operation of ozone monitoring into winter and report the measurements. The size of the winter pro- gram should be sufficient to enhance researchers’ ability to examine (1) seasonal differences in risk, (2) how these seasonal risk differences vary spatially be- tween communities with warmer and cooler winters, and (3) ozone-mortality relationships at lower ozone concentrations. In recognition that ozone is a re- gional pollutant, winter measurements need not be collected in all the summer locations but, when they are collected, should be collected at the frequency of summer measurements.

Next: 4 Contributions of Relevant Health Studies to the Estimation of Reductions in Premature Mortality »
Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution Get This Book
×
Buy Paperback | $55.00 Buy Ebook | $44.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

In light of recent evidence on the relationship of ozone to mortality and questions about its implications for benefit analysis, the Environmental Protection Agency asked the National Research Council to establish a committee of experts to evaluate independently the contributions of recent epidemiologic studies to understanding the size of the ozone-mortality effect in the context of benefit analysis. The committee was also asked to assess methods for estimating how much a reduction in short-term exposure to ozone would reduce premature deaths, to assess methods for estimating associated increases in life expectancy, and to assess methods for estimating the monetary value of the reduced risk of premature death and increased life expectancy in the context of health-benefits analysis.

Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Air Pollution details the committee's findings and posits several recommendations to address these issues.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!