Efficacy of recent state implementation plans for 8-hour ozone

Abstract

The development of state implementation plans (SIPs) for attainment of criteria pollutant standards is an integral component of air quality management in the United States. However, the content and efficacy of SIPs have rarely been examined systematically. Here, 20 SIPs developed in response to the 1997 8-hr ozone standard are reviewed as case studies of attainment efforts at the state level. Comparison of observed and model predicted ozone concentrations shows the U.S. Environmental Protection Agency (EPA) recommended modeled attainment test to be a somewhat conservative predictor of attainment. Among 12 SIPs for regions that sought attainment by 2009, the test correctly predicted attainment and nonattainment in four and five regions, respectively; in the other three regions, attainment was observed despite predictions of nonattainment. However, weight-of-evidence determinations and deviations from the recommended modeled attainment test methodology led five of these SIPs to predict attainment that was not in fact observed by 2009; three of those regions achieved attainment in 2010. Ozone and NO₂ concentrations declined across much of the United States during the period covered by the SIPs, with rates of improvement strongly correlated with the initial pollution levels and hence greatest in nonattainment regions. However, at monitors with mid-range levels of ozone initially, rates of reduction were largely independent of the initial attainment status of the region. This is consistent with the fact that apart from California, the majority of ozone precursor reductions documented by SIPs resulted from federal measures rather than from state or local controls specific to the nonattainment regions.

Implications

At a time of sharp debate in the United States about strengthening ambient ozone standards, this study highlights strengths and weaknesses of recent state implementation plans. Substantial improvements in air quality have been observed, bringing the majority of regions into attainment of the 1997 ozone standards. The extent to which state-level measures contributed to those improvements is unclear. The standard modeled attainment tests provided somewhat conservative predictions of future attainment, but alternate approaches predicted attainment that was not in fact observed. Greater caution may be needed in weight-of-evidence analyses to avoid overly optimistic predictions of future attainment.

Introduction

States shoulder critical responsibilities for air quality management under the United States Clean Air Act (CAA). Although the Act tasks the U.S. Environmental Protection Agency (EPA) with setting health-based National Ambient Air Quality Standards (NAAQS) for criteria pollutants, it assigns to states the responsibility to develop state implementation plans (SIPs) for attaining those standards.

Revised ambient standards for ozone, issued by EPA in 1997 but not cleared by the courts until 2002 (Bachmann, 2007), forced many states to develop new attainment plans in recent years. The 1997 standards established an 8-hr ozone limit of 0.08 parts per million (ppmv), or 84 parts per billion (ppbv) after rounding. The associated implementation rule required SIPs for moderate and above nonattainment areas to include photochemical modeling to demonstrate whether attainment would in fact be achieved (U.S. Environmental Protection Agency, 2005). Many of those plans projected attainment of ozone standards by 2009, and thus can now be evaluated by ambient observations.

Ozone NAAQS issued in 2008 (U.S. Environmental Protection Agency, 2008) strengthened the 8-hr standard to 75 ppbv, but implementation and associated SIP development have yet to occur as more stringent limits were considered (U.S. Environmental Protection Agency, 2010) and then abandoned (Obama, 2011). Nevertheless, reconsideration of the ozone NAAQS scheduled for 2013 or implementation the 2008 standard could prompt a new wave of SIP development efforts. Now is thus an especially opportune time to assess the efficacy of the SIP attainment demonstration process.

Past studies have highlighted important strengths and weaknesses of the SIP process. In terms of performance, ambient concentrations of criteria pollutants have declined dramatically nationwide (U.S. Environmental Protection Agency, 2007a), even as many regions have failed to achieve SIP-promised attainment dates (Bachmann, 2007; Fine and Owen, 2005). SIPs commit states to implement enforceable measures for air quality improvement, but EPA has long recognized the need for interstate regional and national policies to address air pollution across state boundaries (Bachmann, 2007). Some have argued that the growing profusion of federal rules such as the NO_x SIP Call, Clean Air Interstate Rule, and Federal Motor Vehicle Emissions Control Program have marginalized the importance of SIP-prompted state and local measures, which tend to be costlier and more difficult to enforce (Anderson, 2006).

In terms of process, concerns have been raised that the highly technical, model-dependent approach to attainment demonstrations discourages public participation and delays abatement efforts (Fine and Owen, 2005). Each step of the air quality management process—from EPA revising NAAQS limits, clearing legal challenges, and designating nonattainment areas to states developing SIPs and implementing abatement plans—may involve years of efforts, perhaps with the NAAQS revised again before attainment is achieved (Bachmann, 2007). For example, most SIPs for the 1997 ozone standard were not submitted until almost a decade later, following delays from litigation of the standard and then issuance of nonattainment designations and implementation rules. The repeated reconsideration of the 2008 ozone standard has again put SIP development on hold. A seminal study by the National Academy of Sciences (NAS) National Research Council (NRC) recommended transforming the SIP process to emphasize periodic review and refinement of integrated multipollutant air quality management plans (National Research Council, 2004). Integrated multipollutant approaches could enhance the net benefits of air quality planning at the state level (Chestnut et al., 2006; Cohan et al., 2007; Scheffe et al., 2007).

Only a handful of published studies have systematically reviewed SIP content. Roth et al. (2005) audited the modeling approaches of ozone SIPs and other regulatory modeling efforts in North America from the early 1990s, and found that the majority of cases lacked sufficient performance evaluation and corroborating analyses. Fine and Owen (2005) reviewed the attainment planning process in the San Joaquin Valley and highlighted specific shortcomings regarding modeling and public participation. The NRC review of air quality management in the United States devoted a chapter to assessing the SIP process, including quantification of the relative importance of state and federal measures in four SIPs targeting the previous 1-hr ozone standard (National Research Council, 2004). All of these studies highlighted the dilemma that EPA-approved SIPs do not always lead to actual attainment at ambient monitors.

Here, SIPs recently developed under EPA's 1997 8-hr ozone standard (U.S. Environmental Protection Agency, 1997) are considered as case studies to critically review the state-level attainment process. As indicators of SIP efficacy, we assess the following:

•	The extent to which the abatement plans adopted new state and local controls, rather than merely tallying the impacts of existing federal measures.
•	The extent to which deviations from EPA-recommended methodology (EPA, 2007b), weight-of-evidence (WOE) arguments, or reclassifications were needed to demonstrate attainment.
•	Differences in modeled and observed ozone attainment.
•	Trends in observations of ozone and its precursor NO2, to assess whether nonattainment designation spurred more rapid improvements in air quality.

The prevalence of cost and health quantification in SIPs is also assessed.

Methods

Selection of SIPs

Screening was applied to identify SIP attainment demonstrations suitable for analysis (Table 1). Because areas under the least polluted nonattainment designation, marginal, are given only 3 years to reach attainment (U.S. Environmental Protection Agency, 2004), they did not have the full 3 years that all higher designations were given to develop a SIP, and thus were not considered. However, analysis does include the Atlanta nonattainment area, because it submitted a full SIP after being redesignated from marginal to moderate (Georgia Department of Natural Resources, 2009). Some regions originally designated as moderate were excluded from analysis because they achieved the standard early (Table 1), obviating the need for a full SIP revision. In cases where a nonattainment area is the responsibility of more than one state, the SIP from the state with the greatest share of the population of an area was considered. Twenty of the remaining 22 SIPs were considered for qualitative analysis, as we were unable to obtain those from Illinois despite repeated inquiries.

Table 1. State implementation plans for the 1997 8-hr ozone standard

Nonattainment Area	CurrentClassification ^a	ObservedAttainmentby 2009? ^b	SIP Consideredfor Analysis	Quantifies$/ton byMeasure? ^c	Quantifies TonsReduced byMeasure? ^d	QuantifiesHealthBenefits? ^e
Atlanta, GA	Moderate		Yes
Baltimore, MD	Moderate		Yes		x	x
Boston-Lawrence-Worcester (E. MA), MA	Moderate	x	Yes		x
Boston-Manchester-Portsmouth (SE), NH	Moderate	x	No—early  attainment
Charlotte-Gastonia-Rock Hill, NC-SC	Moderate		Yes
Chicago-Gary-Lake County, IL-IN	Moderate	x	No—not found
Cleveland-Akron-Lorain, OH	Moderate	x	Yes	x
Dallas-Fort Worth, TX	Moderate		Yes		x
Fredericksburg, VA	Moderate	x	No—early  attainment
Greater Connecticut, CT	Moderate	x	Qualitative
Houston-Galveston-Brazoria, TX	Severe 15	x	Qualitative
Jefferson Co, NY	Moderate	x	No—early  attainment
Los Angeles South Coast Air Basin, CA	Extreme		Qualitative	x	x	x
Los Angeles-San Bernardino Cos (West Mojave), CA	Moderate ^f		Qualitative	x	x
Milwaukee-Racine, WI	Moderate	x	No—early  attainment
New York-N. New Jersey-Long Island, NY-NJ-CT	Moderate	x	Yes	x
Philadelphia-Wilmington-Atlantic Ci, PA-NJ-MD-DE	Moderate		Yes		x
Poughkeepsie, NY	Moderate	x	Yes	x
Providence (All RI), RI	Moderate	x	Yes
Riverside Co (Coachella Valley), CA	Severe 15		Qualitative	x	x	x
Sacramento Metro, CA	Severe 15		Qualitative	x	x
San Joaquin Valley, CA	Extreme		Qualitative	x	x	x
Sheboygan, WI	Moderate	x	No—early  attainment
Springfield (Western MA), MA	Moderate	x	Yes		x
St Louis, MO-IL	Moderate	x	No—not found
Ventura Co, CA	Serious		Qualitative	x	x
Washington, DC-MD-VA	Moderate	x	Yes		x	x
Notes: ^aAs of September 2, 2010.
^bOzone DVs ≤84 ppb based on 2007–2009 observations.
^cCost-effectiveness estimated for majority of control measures.
^dTons of reduction estimated for majority of control measures.
^eProvides any quantitative analysis of health benefits resulting from SIP.
^fReclassified as Severe 15.

Quantitative analysis of SIP predictions was performed only for the SIPs from moderate designation regions, which predicted future design values (DVs) for each monitor in the area for 2009, the most recent year for which monitoring data corresponding with an attainment deadline are available. These monitors also had to be identified by their unique numerical ID including 4-digit Air Quality System (AQS) site code within the attainment demonstration so that they could be matched to monitoring data; this restriction excluded the Greater Connecticut SIP, leaving 12 SIPs for comparison of predicted and observed DVs.

Observational data

For ozone trends analysis, observational data for each monitor were obtained from U.S. EPA via http://www.epa.gov/airtrends/values.html. Ozone DVs represent the 3-year average of the fourth highest 8-hr ozone observation at each monitor each year. Data for the NO₂ trends analysis come from monthly average NO₂ column densities observed by the Ozone Monitoring Instrument (OMI) satellite over the continental United States as retrieved by the Dutch OMI NO₂ product (DOMINO) algorithm (Boersma et al., 2007)(data obtained from http://www.temis.nl).

Results

State and federal measures in SIPs

Does the SIP process spur new state and local control measures, or is it primarily a bookkeeping of federal measures that had already been enacted? NRC (2004) categorized state and federal measures in four illustrative SIPs for the 1-hr ozone standard. They found that federal measures constituted the majority of volatile organic carbon (VOC) controls in three of the four SIPs, but that local, state, and multistate measures constituted the majority of nitrogen oxides (NO_x ≡ NO and NO₂) controls in each SIP.

Updating the NRC analysis for the newer 8-hr ozone SIPs is impeded by the limited quantification of control measures within the SIP documents (Table 1). Only 12 of the 20 SIPs considered for qualitative analysis clearly indicated the tons of reduction each control measure was expected to produce, and 6 of these 12 SIPs were from California. Thus, we focus our analysis on one California SIP (Los Angeles) and the six other SIPs (Dallas-Fort Worth, Baltimore, Washington, DC, Philadelphia, and Eastern and Western Massachusetts) that clearly quantify emission reductions by control measure.

Quantified measures that were employed in photochemical modeling for these seven SIPs were categorized by the originating level of government (i.e., federal or state/local) most directly responsible for that measure. The distinction between state and federal measures can be ambiguous in some cases. For analysis purposes, any measure that was enacted and implemented primarily by a state is considered to be a state measure, even if based on an EPA model rule. Measures of primarily federal origin such as the Clean Air Interstate Rule are categorized as federal, even though they require some state implementation actions.

California presents a unique case here because, due to its long history with ozone pollution (Bachmann, 2007), it alone among states has the ability to set its own mobile emission standards so long as they are more stringent than federal standards. This ability allows Los Angeles to rely on state and locally originated measures, replacing any relevant federal measure with a more stringent state measure. However, the extent to which emission reductions from the California measures exceed comparable federal measures was not assessed.

Of the six remaining areas, the federal government played the smallest role in Washington, DC, where it provided 57% and 43% of the NO_x and VOC reductions, respectively (Table 2). State measures there included the implementation of Ozone Transport Commission (OTC) model rules for VOC emissions and the Maryland Healthy Air Act (MHAA), which required reductions in emissions of NO_x, SO₂, and mercury from coal-fired electric generating units. In each of the other five areas, federal measures accounted for at least 70% of the reductions of NO_x and VOCs (Table 2).

Table 2. Percent of emission reductions attributed to state/local or federal measures in each SIP

	NOx		VOC
	State/Local	Federal	State/Local	Federal
Los Angeles South Coast	100	0	100	0
Dallas-Fort Worth	16.1	83.9	N/A ^a	N/A ^a
Baltimore	28.4	71.6	24.9	75.1
Washington, DC	57.1	42.9	42.9	57.1
Philadelphia	16.6	83.4	0	100
Boston (Eastern MA)	0	100	29.4	70.6
Springfield (Western MA)	0	100	27.2	72.8
Note: ^aDallas-Fort Worth SIP provides insufficient information to categorize VOC controls.

Deviations from EPA-recommended modeled attainment test

Roth et al. (2005) extensively audited the overall modeling methodologies of SIPs for the earlier 1-hr ozone standard. Here, we instead focus on the extent to which SIP modeling approaches deviated from the standard EPA-recommended modeled attainment test (EPA, 2007b) in order to demonstrate attainment by the targeted date. Deviations from the recommended methodology do not invalidate a SIP, since the EPA guidance document is not legally binding upon state modeling practices. However, significant deviations from EPA-recommended methodology might raise questions about the robustness of an attainment demonstration.

The EPA guidance document (EPA, 2007b) establishes the modeled attainment test as the primary predictor of whether a SIP strategy is sufficient. In that test, a photochemical model such as the Community Multi-scale Air Quality (CMAQ) model (Byun and Schere, 2006) is applied with future year and base year emissions inventories to compute the ratios, termed “relative response factors” (RRFs), between future and base year ozone concentrations near each monitor i:

(1)

The guidance document recommends that each RRF _i be computed based on the maximum concentration simulated within an array of grid cells surrounding i, and averaged over days on which Ozone_base,i exceeds a threshold level (U.S. Environmental Protection Agency, 2007b).

Each RRF is then multiplied by the observed base design value (DVB) at that monitor to predict whether the future design value (DVF) will be below the NAAQS limit:

(2)

Attainment requires that DVF _i < NAAQS for all i. The guidance document provides multiple methods of calculating DVBs, but recommends that they be calculated by averaging three successive 3-year averages of yearly fourth highest observed ozone concentrations. Most of the 8-hr ozone SIPs considered here computed DVB based on 2000–2002, 2001–2003, and 2002–2004 averages, leaving a weighted average around the year 2002, which is considered the base year.

The guidance document also encourages states to conduct supplemental analyses, including additional modeling, emission and concentration trend analyses, and observation-based analyses to complement the primary modeled attainment test. If the primary test showed a region to be very close to the NAAQS threshold (i.e., max(DVF _i ) = 82–87 ppb), the document recommends that the state conduct a weight-of-evidence (WOE) determination considering both the primary modeling and supplemental analyses of air quality and emissions trends and observational models to reach an aggregate conclusion as to whether the NAAQS will be achieved.

Many of the methods used in the WOE determinations, although uniquely tailored to local conditions, are common across nonattainment areas (Table 3). Almost all of the WOE determinations focused on evidence that would support rather than contradict predictions of attainment. For example, most cited persistent downward trends in NO_x, VOCs, and ozone, and many detail unquantifiable voluntary emission reduction measures that were not included in the modeling (Table 3). Some WOE determinations also presented alternate methods for calculating DVBs (e.g., different base years) and RRFs (e.g., different treatment of the array of grid cells or the concentration threshold) that would reduce the predicted max(DVF _i ) in eq 2(2) (Table 3). Some states cited dynamic evaluation studies such as Gilliland et al. (2008) to raise the possibility that photochemical models may underestimate the responsiveness of ozone to reductions in NO_x emissions. SIPs for the Philadelphia (see p. 72 of SIP) and Baltimore (p. 146) nonattainment areas used such studies to justify increasing ozone responsiveness by 50% for recalculations of DVFs in the WOE determinations, whereas the SIP for Washington, DC, raised similar points qualitatively (pp. 10–28). Other states, presented with nonattaining DVFs from the standard modeled attainment test, chose to seek voluntary reclassification to a higher category instead of performing a WOE determination (Table 3). In seeking reclassification, states were required to show that they would reach attainment by their new deadline. Two regions (Houston and New York) that sought reclassifications actually achieved DVs below 85 ppbv in 2009 (Table 1).

Table 3. Rationale used to support weight-of-evidence determinations of attainment

Nonattainment Area		UnquantifiedVoluntaryMeasures ^a	AlternateMethodfor RRFCalculation ^b	AlternateMethodfor DVBCalculation ^c	Criticismof AirQualityModel ^d	Requests aReclassification ^e
Atlanta, GA		x	x	x
Baltimore, MD		x			x
Boston-Lawrence-Worcester (E. MA), MA
Charlotte-Gastonia-Rock Hill, NC-SC		x		x
Cleveland-Akron-Lorain, OH			x	x
Dallas-Fort Worth, TX					x
Greater Connecticut, CT		x
Houston-Galveston-Brazoria, TX		x			x	x ^f
Los Angeles South Coast Air Basin, CA						x
Los Angeles-San Bernardino Cos (W Mojave), CA		x				x
New York-N. New Jersey-Long Island, NY-NJ-CT		x				x
Philadelphia-Wilmington-Atlantic City, PA-NJ-MD-DE			x	x	x
Poughkeepsie, NY				x
Providence (All RI), RI
Riverside Co, (Coachella Valley), CA						x
Sacramento Metro, CA						x
San Joaquin Valley, CA						x
Springfield (Western MA), MA
Ventura Co, CA						x
Washington, DC-MD-VA		x	x	x	x
Notes: ^aCommits to optional measures not considered in calculation of DVF.
^bCriticizes EPA's recommended method for calculating RRFs, and recalculates using an alternate approach.
^cCriticizes EPA's recommended method for calculating DVB, and recalculates using an alternate approach.
^dArgues that the photochemical model underpredicts ozone responsiveness to NOx emission reductions.
^eAttainment is demonstrated for a date reflecting a requested reclassification to a higher category, rather than the attainment date in effect at the time of SIP development.
^fSought and granted reclassification to a higher category prior to writing the SIP.

Modeled and observed ozone attainment

Moderate nonattainment areas modeled 2009 as the “future” year for attainment demonstrations in the 8-hr ozone SIPs. Now that 2009 has passed, it is possible to compare the modeled 2009 “future” design values with the actual ozone design values that were observed in these areas.

Figure 1 compares modeled 2009 DVFs reported in 12 SIPs based on standard EPA-recommended modeling methodology against the 2009 observed ozone design values. Note that observed design values are 3-year averages of fourth highest annual 8-hr measurements, so the “2009” DV reflects ozone measurements from 2007 to 2009. The vertical line divides monitors that measured attainment from those that did not, whereas the horizontal line divides those that predicted attainment from those that did not. Only monitors that were identified by an AQS number in the SIP and that recorded ozone measurements for 2007–2009 in the AQS database were considered for this analysis. Results for three moderate regions that completed SIPs with attainment modeling are not plotted because we did not obtain their SIPs (St. Louis and Chicago) or they lacked matching monitor IDs (Greater Connecticut); all three went on to observe ozone DVs below 85 ppbv at all monitors by 2009.

Figure 1. Ozone design values predicted in SIPs by EPA-recommended modeled attainment test (y-axis) and observed at monitors in 2007–2009 (x-axis).

Except for New York, all of these moderate SIPs claimed attainment at all monitors by 2009, using alternate modeling methodology or WOE arguments as necessary when the model attainment test results plotted in Figure 1 did not predict attainment. New York sought a reclassification but then observed attaining DVs in 2009. Thus, all points to the right of the vertical line represent “false negatives” of SIPs that predicted attainment that was not in fact observed by 2009. This includes both false negatives in which the SIP and the EPA-recommended modeling methodology (Figure 1, lower right quadrant) concurred in predicting attainment and false negatives in which only the SIP predicted attainment (Figure 1, upper right quadrant). A region attains the ozone NAAQS only if none of its monitors records an ozone design value that exceeds the standard.

The standard (EPA-recommended) modeled attainment test correctly predicted attainment for 69 of the monitors and nonattainment for 6 monitors, or 79% of all monitors (Figure 1). Of the remaining monitors, 17 (18%) were “false alarms” where the monitor actually attained despite modeling that did not predict attainment, and 3 (3%) were false negatives. At the regional level, the standard methodology yielded three false alarms (Cleveland, New York, and Washington, DC) and no false negatives, whereas the actual SIPs had no false alarms and five (Atlanta, Baltimore, Charlotte, Dallas-Fort Worth, and Philadelphia) false negatives. Thus, the EPA-recommended modeled attainment test tended to provide a relatively conservative prediction of attainment, yielding more false alarms than false negatives. This is consistent with a dynamic evaluation study that found that a photochemical model underestimated ozone reductions caused by downward NO_x emission trends (Gilliland et al., 2008). Although it is possible that other hypotheses such as favorable meteorology could explain why fewer violations were observed than predicted by SIP models, so far the progress has been sustained. Updating the Figure 1 analysis to 2008–2010 data (not shown), the number of regions with a nonattaining monitor drops from five to two (Baltimore and Dallas-Fort Worth).

Observed ozone and NO₂ trends

Do nonattainment designations and subsequent SIP attainment efforts lead to accelerated improvements in air quality? To address this question, observed trends in both ozone and one of its key precursors (NO₂) were computed and classified spatially based on nonattainment status. For ozone, trends in design values from 2003 to 2009 were compared between areas that were or were not classified as nonattainment of the 84 ppbv ozone standard in the April 2004 designations (Figure 2). During this time period, the NO_x SIP Call substantially reduced point source NO_x emissions in the eastern United States (U.S. Environmental Protection Agency, 2009), and vehicle and fuel standards reduced mobile emissions nationwide (Ban-Weiss et al., 2008; Dallmann and Harley, 2010). Note that the 2003 DV reflects ozone measurements from 2001 to 2003, and the 2009 DV reflects 2007 to 2009. For this analysis, monitoring data were used from the entire AQS system over the continental United States, including regions excluded from the quantitative SIP analysis.

Figure 2. Reduction in observed ozone design values from 2003 to 2009 at monitors in regions that were or were not designated nonattainment of the 1997 ozone standard.

Data for DVs from individual monitors in 2003 and 2009 were matched by AQS number, and those monitors that did not have data for both 2003 and 2009 were removed. Reductions were then calculated and each remaining monitor was marked with 1 (nonattainment) if the monitor is in a county that has ever been designated nonattainment under the 1997 ozone standard (including early action compacts and designations under both subpart 1 and subpart 2) and 0 (attainment) if not. On average, monitors originally designated nonattainment achieved ozone reductions of 13.3 ppbv, nearly double the 7.0 ppbv achieved by attainment monitors. The ozone improvements enabled 16 of the 27 nonattainment regions in Table 1 to achieve the 84 ppbv ozone limit at all monitors by 2009; 3 more regions did so by 2010. On the other hand, comparing among monitors with initial DVs between 65 and 84 ppbv (which may have been designated attainment or nonattainment, since designation for an entire metropolitan region is determined by ozone levels at its most polluted monitor), the trends were nearly identical: 7.4 ppbv at nonattainment monitors, and 7.7 ppbv at attainment monitors.

Linear regression was applied to further analyze the ozone DV trends data. The data were well characterized by a linear equation (eq 3(3); Figure 2):

(3)

Here y is the reduction in ozone (in ppb) for a monitor from the 2003 DV to the 2009 DV, x is the ozone DV for 2003, and s is a dummy variable with a value of 0 (attainment) or 1 (nonattainment) as described above. The regression coefficients β₁, β₂, and β₃ were found using ordinary least-squares regression. Figure 2 shows the coefficients, the R ² value, and the P value from a t test of the null hypothesis that β₃ = 0. Although the small positive value of β₃ (0.53 ppb) is consistent with slightly faster ozone reduction in the originally nonattainment regions, the large P value (0.30) indicates that the null hypothesis that nonattainment status did not influence subsequent ozone reduction cannot be rejected. Since nonattainment status and 2003 DVs are highly correlated, causation of trends cannot be clearly defined.

The association between trends and original designation status was also assessed for the ozone precursor NO₂. Satellite observations of NO₂ column densities offer far greater spatial coverage than ground-based NO₂ monitors, which are sparse outside of polluted urban centers and may exhibit interference from other compounds (Dunlea et al., 2007; Lamsal et al., 2008). Because the OMI measurements extend back only to 2005, reductions were based on differencing the summertime (June–August) averages for 2005 and 2009. Only data from the contiguous 48 U.S. states was used. The data was well characterized by a quadratic trend line, with greater fractional declines in NO₂ in areas where NO₂ concentrations were higher. NO_x influences its own photochemical lifetime via interactions with hydroxyl radicals (Valin et al., 2011), so trends in concentrations are an imperfect proxy for trends in emissions. The data were then analyzed by eq 4(4) (Figure 3):

Figure 3. Reduction in satellite-observed summertime NO₂ column densities from 2005 to 2009 for grid cells in regions that were or were not designated nonattainment of the 1997 ozone standard.

(4)

Here y is the reduction in OMI NO₂ column density (in units 10¹³ molecules cm⁻²) from 2005 to 2009 for a given grid cell, and x is the NO₂ column density in 2005. The dummy variable s has a value of 1 if any part of the grid cell falls within a county that has ever been designated under the 1997 ozone standard (including early action compacts and designations under both subpart 1 and subpart 2) and 0 if not. The regression coefficients γ₁, γ₂, γ₃, and γ₄ were found using ordinary least-squares regression. The P value was determined by performing a t test of the null hypothesis that γ₄ = 0. Although the large amount of data leads to a low P value (<0.001), the magnitude of impact indicated by γ₄ is small, about 6.1 × 10¹³ molecules cm⁻². Thus, it appears that slightly faster NO₂ reductions were observed in nonattainment regions.

Prevalence of cost and health quantification

Although SIPs must demonstrate projected attainment of air quality standards, the Clean Air Act does not require quantification of costs or health benefits within SIPs. Some recent studies have shown that cost and benefit quantification could in theory enhance the selection of control measures for attainment plans (Chestnut et al., 2006; Cohan et al., 2007). However, the extent to which actual SIPs have quantified costs and benefits has not previously been investigated systematically.

The SIP narratives identified in Table 1 were examined to explore the extent to which quantifications of control costs and health benefits were documented. It must be cautioned that absence of such documentation in the final SIP document does not necessarily mean that the costs and health benefits of control options were not considered in the design of the attainment plan. Further research would be necessary to make empirical claims about the abundance or absence of control cost analyses used in the SIP process but outside of published SIPs. All SIPs quantified the overall reductions in NO_x and VOC emissions to develop future year inventories, but less than half quantified reductions for each individual measure (Table 1).

Costs are inherently considered in Reasonably Available Control Technology (RACT) analysis, a required component of ozone SIPs in which technological and economic feasibility helps determine which control technologies to require at RACT-applicable emission sources (U.S. EPA, 44 FR 53762, 1979). However, many possible SIP control measures, especially for mobile and nonroad sources, do not fall under RACT.

Beyond RACT, the Clean Air Act stipulates that nonattainment areas must implement “all reasonably available control measures (RACM) as expeditiously as practicable.” The only SIPs that detail cost per ton estimates for most non-RACT NO_x and VOC control measures are those from California, New York, and Ohio. In each of these cases, the cost per ton estimates were not done exclusively for one nonattainment area but instead for numerous areas by an overarching organization: the California Air Resources Board (CARB), Ozone Transport Commission (OTC), and Lake Michigan Air Directors Consortium (LADCO), respectively. In addition, Atlanta, Greater Connecticut, and Washington, DC, all note some cost considerations in less systematic ways, such as detailing the total cost or allotted funds for particular control measures.

Since quantification of health benefits is not a requirement in the SIP process, the documentation of health considerations in SIPs is even more limited. Although the majority of SIPs contain a paragraph or two detailing the general health damages due to exposure to ozone in their introductions, only four of the nonattainment areas provide more extensive information. South Coast AQMD/Riverside provides a more extensive general description of threats, Baltimore estimates the number of at-risk people in its area, and Washington, DC, estimates the at-risk population by geographic area. Only San Joaquin Valley provides detailed estimates of monetary damages to health as well as the number of people harmed; this analysis is not original to the SIP but instead drawn from separate research on the harm of ozone to residents of the Valley (Hall et al., 2006). Thus, the SIPs provide no direct evidence that any state conducted health benefits quantification to inform the selection or prioritization of control measures.

Discussion

Air quality for ozone and NO₂ improved substantially during the examined period across most of the country. Ozone DVs declined by an average of more than 13 ppbv at monitors in nonattainment regions, enabling the majority of those regions to observe attainment by 2009. This is a remarkable amount of progress in just 6 years, although most of the regions remain above the current 75 ppbv standard.

The contribution of the SIP process to these improvements is unclear. Average improvements were steepest in nonattainment regions. However, among locations with similar ozone or NO₂ levels initially, those in regions facing the impetus of nonattainment did not experience dramatically sharper trends. This is consistent with the fact that, apart from California, the majority of emission reductions documented in SIPs resulted from federal measures. The relative role of federal measures is likely to increase in the near term, as major federal policies for power plants (U.S. Environmental Protection Agency, 2011) and mobile sources (Walsh, 2009) take effect while delays in implementing new NAAQS (Obama, 2011) have put SIP development on hold. Nevertheless, the SIP process did document substantial state and local measures and may have motivated some federal measures such as the NO_x SIP Call.

Given the uncertainties in predicting future emission trends and the impossibility of knowing how future meteorology will differ from base years, SIPs following EPA-recommended methodology for modeled attainment tests performed reasonably in predicting future ozone levels. The standard attainment tests produced far more false alarm than false negative predictions of nonattainment for the single observational period examined (Figure 1). Since each region must maintain attainment at all monitors in all future periods, the low number of false negatives may be a desirable result.

However, WOE arguments and deviations from the recommended methodology allowed five SIPs to claim attainment that was not observed by 2009. WOE determinations were originally intended to explore beyond the default modeling to ensure that SIP control strategies would in fact be sufficient to achieve attainment over the long term. The consideration of supplemental approaches in SIP development is justified, given the limitations of standard attainment models (Jones et al., 2005; Sistla et al., 2004; Vizuete et al., 2010). However, the fact that WOE was used almost exclusively to argue that future ozone levels would be lower than originally modeled suggests that greater scrutiny of WOE determinations may be needed to avert false negative predictions.

Integrated approaches to air quality management could deemphasize the role of inherently uncertain SIP attainment predictions. A shift toward comprehensive multipollutant air quality management plans as recommended by NRC (2004) could enable ongoing collaborative review by states and EPA to evaluate actual performance in improving air quality and dynamically adjust strategies as warranted. A collaborative approach, if practicable, would better reflect the fact that both state and federal measures are often needed to achieve attainment. The plans could also include integrated consideration of costs and benefits across multiple pollutants, as recommended by several studies (Chestnut et al., 2006; Cohan et al., 2007; National Research Council, 2004; Scheffe et al., 2007).

Implementation of more stringent ozone standards may soon provide opportunities to incorporate best practices into the attainment process as States confront a new round of nonattainment challenges. The successes and shortcomings of recent SIPs can offer valuable lessons for informing that process.

Acknowledgments

The work of A. Pegues was funded by the Century Scholars Program at Rice University. The work of C. Douglass was funded by a Brown Undergraduate Research Fellowship and National Science Foundation grant ATM-0847386. The work of D. Cohan, A. Digar, and R. Wilson were funded by U.S. EPA through Science to Achieve Results (STAR) grant no. R833665. However, the work has not been subjected to the Agency's required peer and policy review and therefore does not necessarily reflect the views of the Agency and no official endorsement should be inferred. The authors acknowledge the free use of tropospheric NO₂ column data from the OMI sensor from http://www.temis.nl.