An Integrated Framework for Multipollutant Air Quality Management and Its Application in Georgia
- First Online:
- Cite this article as:
- Cohan, D.S., Boylan, J.W., Marmur, A. et al. Environmental Management (2007) 40: 545. doi:10.1007/s00267-006-0228-4
- 307 Views
Air protection agencies in the United States increasingly confront non-attainment of air quality standards for multiple pollutants sharing interrelated emission origins. Traditional approaches to attainment planning face important limitations that are magnified in the multipollutant context. Recognizing those limitations, the Georgia Environmental Protection Division has adopted an integrated framework to address ozone, fine particulate matter, and regional haze in the state. Rather than applying atmospheric modeling merely as a final check of an overall strategy, photochemical sensitivity analysis is conducted upfront to compare the effectiveness of controlling various precursor emission species and source regions. Emerging software enables the modeling of health benefits and associated economic valuations resulting from air pollution control. Photochemical sensitivity and health benefits analyses, applied together with traditional cost and feasibility assessments, provide a more comprehensive characterization of the implications of various control options. The fuller characterization both informs the selection of control options and facilitates the communication of impacts to affected stakeholders and the public. Although the integrated framework represents a clear improvement over previous attainment-planning efforts, key remaining shortcomings are also discussed.
KeywordsAir pollution controlCost–benefit analysisOzoneFine particulate matterState implementation plansAttainment
The Clean Air Act requires the United States Environmental Protection Agency (U.S. EPA) to establish National Ambient Air Quality Standards (NAAQS) for the protection of public health and the environment. U.S. EPA currently administers NAAQS limits for six criteria pollutants: lead, nitrogen dioxide (NO2), sulfur dioxide (SO2), carbon monoxide (CO), particulate matter (PM), and ozone. With the promulgation of more stringent standards for ozone and fine particulate matter (PM2.5, denoting particles with aerodynamic diameter less than 2.5 microns) (U.S. EPA 1997a, 1997b), these two pollutants now comprise the bulk of NAAQS violations. Both pollutants have been linked to significant health impacts in humans, including respiratory problems and premature mortality (e.g., Brunekreef and Holgate 2002, Bell and others 2005). Many metropolitan regions, especially in the eastern United States and California, now confront nonattainment for both the ozone and PM2.5 standards (U.S. EPA 2004c, 2005a). In addition, the Regional Haze Rule (U.S. EPA 1999) requires the remedying of visibility impairment at designated pristine sites known as Class I areas.
Although the NRC report focused on air quality management in the United States, a lack of integrated assessment across multiple priorities afflicts environmental planning in countries worldwide (Sexton 1999). The NRC recommended a shift toward a multi-pollutant paradigm that seeks cost-effective and simultaneous reduction of multiple pollutants to improve overall air quality (NRC 2004).Air pollutants occur in complex mixtures, and yet SIPs are constrained to address only individual criteria pollutants. As a result, the entire, relatively cumbersome SIP process must be undertaken for a pollutant such as ozone and then again for PM in a separate process and on a different timetable, despite the fact that the exposures are simultaneous, the sources are often the same, and the two pollutants share many common chemical precursors. (p. 130).
The multi-pollutant paradigm heightens the importance for policymakers to consider not only costs but also health improvements and other benefits of pollution abatement in developing attainment plans. As demonstrated by Chestnut and others (2006), consideration of benefits may yield a multi-pollutant attainment strategy with greater net benefits than one optimized for cost alone, due to the overlapping impacts of control measures and the different benefits associated with controlling each pollutant. In practice, however, benefits have typically remained poorly quantified, both due to lack of resources and because states are mandated to demonstrate regulatory attainment, not to deliver a particular level of benefits to society.
When only a single pollutant or precursor emission species is targeted, the above process may prove adequate. In particular, if the impact of each emitted ton is roughly uniform, then cost per ton and other practical considerations will be sufficient bases for prioritizing control measures. Omitting consideration of health and other benefits may hinder the communication of impacts to stakeholders and the public, but likely would not skew option prioritization in this case because benefits would scale proportionally with emissions reductions. However, if per-ton impacts vary markedly with the time or location of emission origin, modeling only the aggregate strategy will fail to distinguish the relative effectiveness of each measure.
The inadequacies of the traditional approach become more pronounced in the case of multi-precursor and multi-pollutant attainment strategies. Modeling only the aggregate impact of an overall strategy cannot determine which component measures yield the greatest impact. If the initial strategy is modeled to be insufficient for attainment, additional measures cannot be effectively prioritized without knowledge of the relative sensitivity of ambient concentrations to each precursor. Furthermore, the isolated treatment of each attainment demonstration and lack of assessment of benefits in the traditional approach precludes cross-pollutant prioritization that could enhance overall cost-effectiveness (Chestnut and others 2006).
An overall strategy can thereby be developed with fuller consideration of its implications for costs, benefits, and attainment across regions and pollutants. As can be seen by comparing Figures 3 and 4, the selection of measures for the overall strategy can be influenced by far more factors than simply cost and feasibility. Although the integrated approach provides objective metrics for evaluating each option, it must be recognized that other considerations such as practical realities and equity across stakeholders can and should influence the actual selection of measures. Finally, longer-term photochemical modeling must be conducted to demonstrate attainment from the overall selected strategy before it can be incorporated into SIPs and implemented, to account for interactions across measures and other details that cannot be adequately captured by the sensitivity modeling of short-term episodes.
The process of control strategy development for ozone and PM2.5 SIPs in Georgia continues to be ongoing as of March 2007. Resources devoted to the modeling and analysis described here include five staff members who have worked to examine the feasibility and cost of potential controls, and four Ph.D.-trained modelers who have conducted the photochemical sensitivity analysis and health assessments. Control options are being evaluated by managers at the Air Protection Branch and by Georgia EPD’s Director’s Office, with input from an ongoing stakeholder process. Ultimately, the SIPs must be approved by the Board of the Georgia Department of Natural Resources and by the U.S. EPA.
The following sections detail each component of the integrated framework as applied to control strategy development in Georgia.
Cost and Feasibility Analyses
The integrated process begins with a comprehensive effort to identify options for controlling emissions of each precursor. Control options are identified for each major component of the emissions inventory: major industrial facilities known as “point sources”; “mobile sources,” which include both on-road and non-road vehicles and equipment; and “area sources,” which include a variety of emitters such as prescribed fires, meat cooking, and residential fuel combustion, as well as businesses whose emissions are too small to be tracked as point sources.
For point sources, Georgia EPD applied AirControlNET (E.H. Pechan 2005) to identify potential retrofit technologies and their associated emissions reductions and costs. The software links emissions inventories with a matrix of control options compiled from existing literature. AirControlNET may omit certain options or fail to account for facility-specific conditions that influence feasibility and cost-effectiveness, but it serves as a helpful initial scoping tool of plausible options. Georgia EPD then conducts an extensive process, including input from affected industries, to analyze the emissions reductions and cost-effectiveness that could feasibly be achieved by controls at each source.
Mobile and area sources are considered on a sector-by-sector basis to identify control options. Because comprehensive assessment software is not available for these sectors, potential costs and emissions reductions are estimated by reviewing available literature or empirical evidence regarding the performance of previous projects. With peer-reviewed literature lacking for many mobile and area source control options, much of this review relies on a variety of gray literature sources (e.g., E.H. Pechan 2002; California Air Resources Board 2004, Sierra Research 2005). Interagency and stakeholder forums such as the Georgia Diesel Working Group, the Southeast Diesel Collaborative, and the Idling Reduction Stakeholder Group have informed the assessment of certain options.
Ideally, the above steps would yield a comprehensive menu of potential control options, their costs, and their impacts on emissions. However, as of March 2007, that process is still ongoing and it has proven difficult to determine the actual feasibility and costs of some potential measures.
Atmospheric Sensitivity Analyses
In order to quantify the response of ozone, PM2.5, and regional haze to emissions reductions from various sources, Georgia EPD is using the MM5/SMOKE/CMAQ modeling system to perform episodic emission sensitivity analysis (Boylan and others 2006). This modeling system consists of the NCAR/PSU Mesoscale Modeling System (MM5) meteorological model (Grell and others 1994), the Sparse Matrix Operator Kernel Emissions (SMOKE) emissions model (Houyoux and others 2000), and the Community Multiscale Air Quality (CMAQ) photochemical model (Byun and Schere 2006). This grid-based modeling system simulates hourly concentrations of ozone, PM, and other air pollutants throughout the modeling domain. Georgia EPD is applying these models to simulate ambient pollutant concentrations for 2002 (base year), 2009 (target year for ozone and PM2.5), and 2018 (target year for regional haze) on a modeling domain that covers the entire states of Georgia, Alabama, Mississippi, South Carolina, Tennessee; most of North Carolina and Florida; and small pieces of the surrounding states. This represents a subdomain of modeling developed by the Visibility Improvement State and Tribal Association of the Southeast (VISTAS) regional planning organization (Morris and others 2005).
Future year emissions were projected for each source category to account for “on-the-books” control programs at the federal, state, and local level and expected economic growth. In 2009, it was estimated that Georgia would see a 34% reduction in SO2 emissions (mostly due to electric-generating unit (EGU) point sources), 23% reduction in NOx emissions (mostly due to on-road mobile and EGU point source reductions), and 14% reduction in anthropogenic volatile organic compounds (VOC) emissions (mostly from on-road and non-road mobile sources) relative to 2002. Because of computational limitations, the specific modeling episodes for sensitivity analysis consist of one summer and one winter 30-day period, selected on the basis of Classification and Regression Tree (CART) analysis for ozone and PM2.5 (Douglas and others 2005a, 2005b). CART examines the frequency of various meteorological regimes and can be used to identify episodes best suited for representing longer periods of time (e.g., entire years or summer ozone seasons).
Actual 2002 model estimates for ozone, PM, and gaseous precursors were compared against observations from various monitoring networks in order to assess model performance (Morris and others 2005, Tesche and others 2006). Once satisfactory model performance was achieved, future year emissions were input into the modeling system (with 2002 meteorological fields) to produce future-year modeled pollutant distributions. To increase confidence in the model’s ability to predict future design values, EPA guidance (U.S. EPA 2006) suggests that modeling results be used in a relative sense rather than looking at the absolute change in concentrations. Therefore, future design values are calculated by multiplying the 2002 5-year design value (used for EPA’s model attainment test) by a Relative Reduction Factor (RRF) where the RRF is the ratio of future modeled mean concentrations to base year modeled mean concentrations. For ozone, only days when the modeled base year ozone concentrations exceeded 85 ppb were included in the RRF calculation.
Costs and benefits of scenarios for reducing ozone in Atlanta
Annual cost (106 $)
Emissions reduceda (tpd)
Ozone response at monitorb (ppb)
Ozone sensitivity at monitorb (ppt/tpd)
Annual statewide benefitsc (106 $)
Annual cost per 1 ppt reduction at monitor ($)
Cost per $1 health benefits ($)
10% Atlanta NOxd
10% Atlanta VOCd
SCRs at Power Plant 1
SCRs at Power Plant 2
Costs and benefits of scenarios for reducing fine particulate matter (PM2.5) in Atlanta
Annual cost (106 $)
Emissions reduceda (tpd)
PM2.5 response at monitorb (µg/m3)
PM2.5 sensitivity at monitorb (ng/m3/tpd)
Annual statewide benefitsc (106 $)
Annual cost per 1 ng/m3 reduction at monitor ($)
Cost per $1 health benefits ($)
10% Atlanta PCd
10% Atlanta NH3d
Scrubbers at Power Plant 1
Scrubbers at Power Plant 2
“Regional” emissions sensitivities examine the impact of emissions from specific geographic regions (e.g., Atlanta 20 county non-attainment area). These regional sensitivities are computed by comparing ozone, PM2.5, and haze levels in a standard target year simulation to levels in a similar simulation in which anthropogenic emissions of nitrogen oxides (NOx), VOCs, SO2, ammonia, or primary organic and elemental carbon particles (PC) have been reduced by 10% in a given region. The individual sensitivity responses can be scaled up or down to approximate pollutant responses to other magnitudes of emission reduction, although the accuracy of extrapolations diminishes with the size of the extrapolation and the nonlinearity of the response (Hakami and others 2004, Cohan and others 2005).
Georgia point source emissions of NOx and SO2 are dominated by seven large coal-fired EGUs at which specific control technologies may be applicable. Thus, rather than considering these EGUs as part of the arbitrary 10% regional emission sensitivity simulations described above, Georgia EPD has directly modeled the potential impacts of installing selective catalytic reduction (SCR) for NOx and flue gas desulfurization (“scrubbers”) for SO2 at the facilities. The potential emissions reductions associated with these sensitivities range from 65% to 95%, depending on the current control technologies operating at each power plant. The final sensitivity results are presented as an absolute change (ppb or µg/m3) and as a relative change on a tons per day basis (ppt/tpd or ng/m3/tpd) so that the total impacts and per-ton efficiencies of EGU controls can be considered (Tables 1 and 2). Although the Clean Air Interstate Rule establishes interstate cap-and-trade markets for EGU NOx and SO2 emissions, Georgia EPD does not model any changes in out-of-state emissions that might result from additional Georgia EGU emissions reductions beyond the base case.
Sensitivity modeling by Georgia EPD shows ozone to be far more responsive to NOx than to VOCs, indicating that Atlanta is in a NOx-limited regime. Atlanta ozone is also responsive to the installation of SCRs at two of the major power plants, one located inside and the other larger one located outside of the 20-county Atlanta non-attainment area (Power Plant 1 and Power Plant 2, respectively). Controls at other plants showed substantially less impact on ozone.
Seasonal average PM2.5 sensitivities were calculated for the summer and winter episodes, and “annual” sensitivities were calculated by assigning different weighting factors to each episode day based on how important that day was to the annual average (Table 2). For PM2.5 controls, the largest benefits are achieved from additional controls of regional PC from Atlanta. Controls of regional SO2, NOx, and VOCs have a much smaller benefit. Atlanta PM2.5 was also responsive to the installation of scrubbers at all major power plants in Georgia, and reported here are the results for the above-mentioned Power Plants 1 and 2. In addition, atmospheric sensitivity analysis indicated that local ammonia emissions contribute strongly to wintertime PM2.5, which prompted an intensified search for control options for this often-neglected precursor. Unfortunately, no new ammonia controls were identified as being cost effective.
Overall strategies for ozone and PM2.5 are still under development in Georgia. After sensitivity analysis is complete and control measures are selected, it will still be necessary to model the impact of the overall control strategy so that attainment over the full time period can be demonstrated and nonlinear interactions among component measures (Cohan and others 2005) are not neglected. If modeling of an initial overall control strategy indicates that further control is necessary, the sensitivity results can inform the search for additional measures and thereby lessen the need for iterations between modeling and strategy development (Figure 4). Although not currently slated for use by Georgia EPD, advanced sensitivity analysis techniques such as the high-order decoupled direct method (Hakami and others 2004) or response surface modeling (Dennis and others 1999) could also assist in the search for additional controls.
One limitation of air quality modeling and sensitivity analysis is that it relies on the accuracy of the underlying meteorology and emissions inventory. For many emissions components, including primary PM, there are considerable uncertainties in the inventory. Thus, Georgia EPD has chosen to supplement its emissions-based modeling with observational analyses to better understand the sources of ambient pollutants. In particular, observation-based source apportionment analyses have been conducted using the Chemical Mass Balance model (U.S. EPA 2004a, 2004b) and positive matrix factorization (Paatero 1997) upon speciated PM2.5 measurements to estimate the contribution of various emissions categories to overall concentrations (Marmur 2006a). Taken together, inventory-based sensitivity analysis and observation-based source apportionment provide a fuller picture of the likely impacts of controls and the associated uncertainties (Marmur and others 2005, 2006).
The control of air pollution may yield significant benefits for a wide range of factors including human health, visibility, agriculture, forestry, building materials, and natural ecosystems. However, the protection of human health is the driving motivation behind the Clean Air Act. Comprehensive assessments of air pollution abatement benefits (e.g., U.S. EPA 1997c) generally find health effects to dwarf all other impacts on a monetized basis. Thus, efforts to quantify the benefits of air pollution control have largely focused on health impacts.
Georgia EPD is applying the Environmental Benefits Mapping and Analysis Program (BenMAP) (Abt Associates 2003) to assess reductions in population exposure to ozone and PM2.5 and the potential health benefits associated with these reductions. BenMAP derives these estimates of health-related benefits by utilizing concentration-response (CR) functions, which relate a change in the concentration of a pollutant to a relative change in the incidence of morbidity and mortality (e.g., Pope and others 2002). Spatial estimates of ambient pollutant concentrations (as modeled by CMAQ) are input into BenMAP, which considers population distributions and CR functions to compute the health impacts resulting from those pollutant levels. BenMAP can also associate each health outcome with a monetized per-incident value to translate morbidity and mortality into associated economic valuations.
BenMAP can be applied to compare the health impacts of various control options considered by atmospheric sensitivity analysis, or to quantify the health impacts of an overall strategy modeled by CMAQ. Benefits analysis allows environmental authorities to evaluate and prioritize various options from the standpoint of reducing overall population exposure to air pollution, on top of the requirement to show attainment at specific monitoring sites (which is also the driving factor for sensitivity analyses). Communicating the benefits of air quality control strategies to decision-makers, stakeholders, and the general public helps to translate unfamiliar quantities such as µg/m3 of PM into tangible impacts such as premature death, asthma attacks, and associated monetary valuations.
Examples of statewide valuations are reported for the sensitivity cases discussed in the previous section (Tables 1 and 2). Ozone benefits presented here are based on CR functions for exposure of up to 8 hours (Marmur 2006b). These once more indicate the efficiency of NOx controls for reducing ozone levels in Georgia (Table 1). Although the ozone benefits at the Confederate Avenue monitor of controlling Power Plants 1 and 2 are nearly identical, the statewide health benefits are almost three times higher for controlling Power Plant 2. This reflects the much larger size of Power Plant 2 and the closer proximity of Power Plant 1 to the monitor. A similar pattern is observed for PM2.5 (Table 2), where the installation of scrubbers at Power Plant 2 produces PM2.5 benefits at Fire Station 8 that are twice those of Power Plant 1; however, the statewide monetary benefits are more than three times higher due to the regional nature of sulfate. Controls of PC emissions also show a large benefit, because of the location of emissions in a highly populated area.
Considering the costs, atmospheric sensitivities, and health benefits together, the cost-effectiveness of control options can be compared with respect to regulatory attainment (cost per ambient improvement at the monitor) and health effect mitigation (cost per health benefit achieved) objectives. These metrics are computed by dividing the cost of a measure by the modeled ambient and health effect responsiveness.
As an illustrative example of cost-effectiveness comparisons, the final two columns of Tables 1 and 2 evaluate options for the two power plants discussed earlier. The more favorable cost-effectiveness of controls at Power Plant 1 than at Power Plant 2 reflects higher initial emission rates per Btu at Power Plant 1 and its closer proximity to densely populated regions and the Atlanta monitors. On a health effects basis, the results suggest more favorable cost-effectiveness for scrubbers than for SCRs, because of the greater monetized health impacts attributed to PM2.5 than ozone. This does not necessarily argue against installation of SCRs, because both ozone and PM2.5 standards must be attained and there may be other benefits beyond the health effects quantified here.
Incorporation of sensitivity analyses and cost and benefits assessments up front in the environmental management process can potentially enhance the selection of multi-pollutant control measures and the communication of the associated implications to the public. With that incorporation, each control option can be considered for its contribution both toward the attainment of air quality standards and toward the protection of human health. The integrated approach is especially valuable in the context of multi-pollutant attainment planning because it facilitates the comparison of disparate options on common metrics. As industrialized nations tighten standards for various air pollutants, the importance of multi-pollutant approaches will continue to increase.
The ongoing experience of Georgia EPD in shifting to an integrated framework for its air quality planning demonstrates the potential usefulness of this approach as well as some limitations. The emergence of faster and more affordable computers for atmospheric sensitivity analysis and improved tools for health benefits assessment has enabled sophisticated modeling of concentration-emission sensitivities and potential health impacts. Adoption of the integrated approach has already paid dividends in Georgia by (1) prompting a heightened focus on the precursor emissions most responsible for ambient pollutant concentrations, in particular NOx for ozone and SO2 and primary particles for PM2.5; (2) yielding objective benefit metrics (both in terms of averted health effects and reduced concentrations at monitors per amount of emissions reduced) that inform the strategy selection process; and (3) facilitating Georgia EPD’s mission to communicate the rationale and implications of its actions to the public and affected stakeholders. However, the identification and cost estimation of control options has lagged behind the other steps, hampering comprehensive comparisons of measures. Some control measures that may be crucial for attainment require long lead times, so in practice they are being pursued even before the full cost–benefit comparisons are available.
Beyond the particular experience of Georgia, several general shortcomings of the integrated approach should be noted as well. Atmospheric sensitivity analysis and cost and benefit assessments are only as accurate as the assumptions, data, and models that drive them. Important sources of uncertainty include the emissions inventories for photochemical modeling and the CR functions (via epidemiological studies) for estimating air pollutant health effects. Continued improvements are needed in emission inventories, control measure evaluation, and the representation of atmospheric science within air quality models. Analysis tools for other benefits such as visibility and ecosystem protection are even less mature than the health models. Given these uncertainties, the objective metrics are not the final arbiters of environmental decision-making, which should consider unquantifiable factors such as equity, practicality, and political realities.
Nevertheless, the integrated framework represents a clear improvement over previous approaches to attainment planning. As environmental authorities increasingly face multi-pollutant attainment challenges driven by interrelated precursor emissions, cost per ton is no longer a sufficient metric for prioritizing control options. The integrated framework provides a more complete understanding of the implications of each option so that both attainment and health objectives can more effectively be addressed.
Tom Shillock (Georgia EPD) created the map in Figure 2. This project was conducted and funded by the Georgia Environmental Protection Division.