Modeling the Effects of Changes in New Source ... - ACS Publications

Dec 15, 2007 - Nevertheless, dispute continues about the test to be used for determining whether a physical change at an EGU would increase emissions...
0 downloads 11 Views 123KB Size
Download PDF
Policy Analysis Modeling the Effects of Changes in New Source Review on National SO2 and NOx Emissions from Electricity-Generating Units D A V I D A . E V A N S , †,4 B E N J A M I N F . H O B B S , * ,‡ C R A I G OREN,§ AND KAREN L. PALMER† Resources for the Future, Washington, DC, Department of Geography and Environmental Engineering, Johns Hopkins University, Baltimore, Maryland, and Rutgers (The State University of New Jersey) School of Law-Camden, Camden, New Jersey

Received January 2, 2007. Revised manuscript received October 29, 2007. Accepted October 30, 2007.

The Clean Air Act establishes New Source Review (NSR) programs that apply to construction or modification of major stationary sources. In 2002 and 2003, EPA revised its rules to narrow NSR’s coverage of renovations. Congress mandated a National Research Council study of the revisions’ impacts. In that study, we used an electricity-sector model to explore possible effects of the equipment replacement provision (ERP), the principal NSR change directed at power plants. We find that, assuming implementation of the Clean Air Interstate Rule (CAIR), tight enforcement of the prerevision NSR rules would likely lead to no or limited decreases in national emissions compared to policies such as ERP. However, emissions might shift forward in time because the previous NSR rules would depress allowance prices, discouraging banking and encouraging allowance use. Only under the most aggressive prerevision NSR enforcement scenario, in which essentially all coal capacity is compelled to retrofit controls by 2020, do NOx emissions fall below ERP levels. Even then, total 2007–2020 SO2 emissions are unaffected. Further decreases in national emissions could be accomplished more cheaply by tighter emissions caps than through NSR because caps provide incentives for efficient operating strategies, such as fuel switching, as well as retrofits.

Introduction The Clean Air Act includes a pair of programs that are together known as New Source Review (NSR). They require that the operator of a large new or modified stationary source use advanced pollution-control technology, and show that the construction or modification would neither interfere with attainment of national ambient air quality standards nor violate a system of increments that restricts increases in ambient pollutant concentration in clean air areas. * Corresponding author phone: (410) 516-4681; fax: (410) 5168996; e-mail: [email protected]. † Resources for the Future. ‡ Johns Hopkins University. § Rutgers. 4 Current address: National Center for Environmental Economics, U.S.EPA, Washington, DC. 10.1021/es070003c CCC: $40.75

Published on Web 12/15/2007

 2008 American Chemical Society

There has been controversy about which renovations at existing facilities constitute modifications. A “modification” is defined in the Act as a physical or operational change that increases air pollution. EPA in the late 1980s announced that a renovation constitutes a “physical change” (rather than “routine maintenance”) if it meets a multifactor test that includes the nature, extent, purpose, frequency, and cost of the work (1). The agency also maintained that an alteration increases air pollution if it would raise annual emissions, even if the source’s maximum hourly emissions would not increase, i.e., the source operates more often. In the late 1990s, EPA filed lawsuits against some electricity-generating units (EGUs) that had undertaken renovations, alleging that these changes triggered NSR (2). The defendants responded that these renovations were “routine maintenance” and thus exempt from NSR, and did not increase emissions because the EGUs’ maximum hourly emissions did not rise. Some of these suits are still pending; others have resulted in settlements that include emission restrictions at the system and/or EGU level (1). In 2002 and 2003, EPA altered its rules to narrow NSR’s coverage of renovations. The most important of these changes for EGUs was the 2003 Equipment Replacement Provision (ERP). Under ERP, an equipment replacement costing less than 20% of the source’s current replacement value (cost) would generally be defined as “routine maintenance” and hence exempted from NSR. EPA’s rule changes-like EPA’s enforcement initiativearoused controversy. Congress required EPA to contract for a National Research Council (NRC) study of the effects of the rule changes on emissions, pollution control installations, and production efficiency (3). Three of us were members of the NRC committee, while the other acted as an expert consultant. During the committee’s study, ERP was invalidated by the DC Circuit Court of Appeals (4). (The Supreme Court has declined to review this decision (4)). Nevertheless, dispute continues about the test to be used for determining whether a physical change at an EGU would increase emissions. In U.S. v. Duke Energy (5), the Fourth U.S. Circuit Court of Appeals held that an emissions increase occurs only if a physical change increases maximum hourly emissions. The Supreme Court reversed this decision, ruling that EPA has no obligation to use the hourly emissions approach (6). EPA reads the decision as allowing the agency to voluntarily adopt the hourly emissions approach, and has proposed to do so (7, 8). This paper summarizes the methods and conclusions of one component of the NRC report. We use a detailed model of the U.S. electricity sector to simulate the likely effects of ERP on the temporal pattern of national emissions of sulfur dioxide (SO2) and nitrogen oxides (NOx) from EGUs. The model is run assuming that either ERP or the “prerevision” NSR rules (i.e., the NSR rules in place before ERP was proposed) are in effect. As we explain below, our model simulations of ERP are equivalent to simulating the hourly emissions approach to enforcing NSR, and thus, despite the Supreme Court’s invalidation of ERP, remain relevant to the NSR debate. Comparisons of predicted emission patterns, fuel use, pollution control retrofits, and electricity industry costs from the model simulations form the bulk of this study. We also explore how different cap-and-trade programs affect these comparisons and interact with NSR. Furthermore, we calculate the cost of an alternative cap-based policy that VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

347

accomplishes the same reductions as the most stringent of the prerevision scenarios considered. We also provide additional findings not in the committee’s report on how regional emissions patterns may change as a result of changing NSR. We focus on coal-fired EGUs for two reasons. First, these EGUs account for approximately 70% and 20% of 2004 national emissions of SO2 and NOx, respectively (9). Second, the fractions of large coal-fired capacity lacking flue-gas desulfurization (FGD) to control SO2 and selective catalytic reduction (SCR) to reduce NOx emissions are, respectively, 62% and 63% (10). If a coal-fired EGU becomes subject to NSR, it must install these controls. To our knowledge, this study is one of only two studies that models the national effects of different methods of determining whether a renovation is subject to NSR. (The other study is discussed below.) While NSR largely emphasizes controlling local air pollution, a focus on its effects on national emissions is appropriate because NSR is also intended to reduce interstate pollutant transport (11). Further, some sectors of the public and press believe that the NSR changes would affect national emissions (12). We emphasize that we do not render any overall judgment of NSR’s worth. Ours is but one of the lenses through which the program can be viewed. NSR has many purposes in addition to decreasing national emissions, and our study does not evaluate NSR’s success at achieving them.

Methods and Assumptions This analysis uses Version 2.1.9 of the Integrated Planning Model (IPM) (13). IPM is a detailed model of the electricity sector in the continental United States. It uses linear programming to find the least-cost pattern of EGU operation, investment, and retirement while meeting peak demands, regional reserve requirements, and regulatory restrictions on air pollutants. EPA has relied extensively on IPM in its analyses of regulations affecting the electricity sector. Consequently, stakeholders are familiar with the model and our findings can be easily compared to other analyses that use IPM. Furthermore, EPA has subjected the model to extensive review and to validation tests of its short-term outputs. EPA studies indicate that IPM can reasonably approximate generating sector operations on a regional scale (3). We remind the reader that “all models are wrong, but some are useful.” Although models necessarily simplify reality, they can provide useful insights about a system’s response to regulatory changes. IPM makes several structural assumptions to simplify the representation of the electricity sector. The model divides the sector into 26 market regions. It represents variation in electric loads by ten demand periods per year (five each in winter and summer). The load levels are based on historic demand with growth projections based on public and private sources. In this study, the model is run over an 18-year forecast horizon with four simulation years (2007, 2010, 2015, and 2020). Each simulation year is intended to represent conditions for that year alone. Results for intermediate years are obtained by interpolation. Compliance with any environmental regulations is assumed in the intervening years. The supply side of the model starts with the entire population of grid-connected generators in the United States. Computational limitations require that existing generators with similar characteristics in each region be aggregated into model plants. The parameters characterizing model plants include capacity, fixed and variable operating costs, and heat rate (fuel-to-generation ratio). The model responds to growth in electricity demand and changes in generation technology and fuel costs by selecting the least-cost pattern of new generating capacity from a variety of fossil fuel and renewable 348

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 2, 2008

technologies over the model’s forecast horizon. Capacity is retired if the going forward costs (fixed annual and variable operating costs) exceed the revenues attributable to that capacity. Natural gas and coal supplies are price-responsive. IPM dispatches model plants in each demand period to minimize cost given regulatory restrictions, plant limits, and load requirements. The electricity produced in a region can be supplied to that or neighboring regions, recognizing interregional transmission limits. The least-cost approach is equivalent to simulating a perfectly competitive market in which all market participants have perfect information and demand is perfectly inelastic. This general approach to modeling the power sector is widely used to analyze the effects of regulatory changes (14). Because IPM aggregates capacity into model plants, the model cannot provide results for generation or emissions at a local level. This aggregation also disregards intraregional transmission constraints and institutional barriers, such as those between vertically integrated utilities. Omitting such constraints generally underestimates production costs. However, even with this aggregation, the model still captures many details. The number of decision variables in IPM is on the order of one million or more. These variables include fuel use, operating levels, and adoption of pollution controls for each of the 2,053 model plants. Further details are provided in the IPM documentation (13). The model is simplified further by assuming that EGUs have perfect foresight of changes in electricity prices, fuel and other costs, and environmental policies. Thus, the model does not reflect the impact of risk-aversion or imperfect information on EGU decisions. Indeed, to our knowledge, no detailed large-scale model exists that includes investor perceptions of, and reactions to, uncertainty. Sensitivity analysis can assess the effect of a model user’s uncertainty concerning his or her assumptions, but budget constraints limited the number of scenarios that we could analyze.

Definition of Scenarios The IPM scenarios are specified on two dimensions. One consists of different versions of EPA’s policy regarding NSR’s coverage along with varying assumptions about NSR’s effects on EGU decisions. A second dimension, referred to as the regulatory setting, represents assumptions about what other air pollution regulations are in place. (For a tabular depiction of the different model runs, see the Supporting Information.) Strictness of NSR Policy. We simplify the NSR policies into two basic alternatives: enforcement of the prerevision NSR multifactor approach versus ERP. It is difficult to use IPM to directly analyze the effects of NSR because the model neither explicitly represents maintenance or life-extension decisions that could trigger ERP or the prerevision rules, nor their costs or effects on EGU performance. Therefore, we must make assumptions about how NSR is represented in IPM. We assume that ERP does not appreciably affect EGU operations and investment decisions because it is rare for a renovation to exceed 20% of the source’s replacement value, which is the NSR applicability trigger under ERP. This allows us to assume that ERP is already represented in “base case” IPM runs (15). These are the runs that EPA uses as a reference against which the effects of proposed policies are compared. Those base cases incorporate all of the adopted major regulations (except NSR) affecting the electricity sector. The ERP runs can also be interpreted as representing emissions and generation patterns that would result from the EPA’s proposed hourly emissions approach. This is because renovations at EGUs are also unlikely to increase hourly emission rates, and thus EGU renovations would similarly be unlikely to trigger NSR. We define two possible consequences of the prerevision rules for generator investment: (1) “avoid,” in which genera-

tors could avoid triggering NSR but at the cost of worsening performance, and (2) “R/R/R,” where a given quantity of coal-fired capacity either retrofits pollution controls, repowers with NSR-compliant technology, or retires. The retrofit option requires the installation of FGD and SCR controls. Several alternative R/R/R scenarios are defined by assuming that various shares of capacity become subject to NSR. Economic, policy, and legal uncertainties are too large for us to determine which scenario is most likely to be correct, so we choose instead to explore and bound the consequences of alternative assumptions. The regulatory impact analysis of ERP (16) reports results for the “avoid” scenario for the prerevision rules. This scenario anticipates that prerevision requirements would cause EGU owners to avoid NSR by deferring maintenance. The EPA assumed that the consequence of deterring maintenance is a 0.1%/yr deterioration in efficiency and capacity. Generation costs are higher, but NOx and SO2 emissions patterns are essentially the same in the “avoid” scenario as under ERP. Hence, if the prerevision rule motivates EGUs to avoid NSR, and the consequence is the assumed deterioration in efficiency and capacity, there would be only minor national NOx and SO2 emission differences between the prerevision rule and ERP. Our analysis goes beyond the EPA’s regulatory impact analysis by considering R/R/R scenarios and accounting for recently adopted cap-and-trade programs, which we describe below. But we do not compare ERP to an “avoid” scenario that accounts for the new cap-and-trade policies. This comparison is unnecessary because, if emission patterns were already similar between the “avoid” scenario and ERP, then the same would be true if the caps on total SO2 and NOx emissions were tightened. The R/R/R scenarios assume that the plants that make renovations that trigger NSR are those where the cost of retrofitting, repowering, or retiring is the lowest, net of the value of emissions allowances that are freed up. That is, the model selects the least costly set of retrofits while meeting demand, accounting for the net effects of fuel and operating costs, compliance with emissions caps constraints, and demand (total generation) requirements. This least-cost assumption was used for modeling convenience; it may not be the order in which EGUs are subject to R/R/R. Thus, our analysis may understate the cost of the prerevision NSR, while alternative assumptions might increase or decrease emissions. Different R/R/R scenarios assume varying lower bounds on the amount of EGU capacity (defined as a fraction of 2004 capacity) that must retrofit, retire, or repower in each year. The low scenario assumes that capacity equivalent to 2% of the 189 gigawatts of coal capacity without FGD as of 2004 (190 gigawatts of capacity without SCR) undergoes R/R/R in each year from 2007 through 2020. As a result, a minimum of 2% of coal-fired capacity would be either retrofitted, repowered, or retired by 2008, 4% by 2009, and so forth, reaching a lower bound of 26% in 2020 and every year thereafter. The two other scenarios assume 5% and 7.5% annual retrofit rates. The 5%/yr (middle) scenario implies that at least 65% of the capacity presently without FGD/SCR would be subject to R/R/R by 2020 while the 7.5%/yr (high) scenario reaches 97.5% by 2020. The retrofit adoption rate in the high scenario is unlikely because some fraction of uncontrolled generation is likely instead to avoid NSR by deferring maintenance. Nevertheless, we analyze the high scenario, treating it as a bounding case. It might be argued that even the rate of NSR applicability under the lower R/R/R scenarios is implausibly high. However, if EPA pays attention to alleged violators, other EGU owners may more frequently conclude that NSR is applicable to a particular maintenance project. Keohane et al. (17) show that plants responded to the possibility of being

subject to an NSR lawsuit by lowering their 1999 SO2 emission rates to avoid scrutiny. Admittedly, this effect may be ephemeral. In the long run, sources may simply avoid the types of investments that attracted EPA’s attention, and NSRinduced retrofits may not increase at all. Bushnell and Wolfram (18) provide evidence that supports this: EGUs that are more likely to be subject to NSR avoided making large maintenance investments. Other Regulations. We impose two alternative regulatory settings to explore how different stringencies of cap-andtrade programs for SO2 and NOx interact with NSR to influence the spatial and temporal pattern of emissions. Cap-and-trade programs give EGUs a fixed annual amount of emissions allowances that they can trade or bank for use or sale in the future. The purpose of a cap-and-trade program is to reach emission targets at least cost over a large area, as opposed to NSR, which is primarily intended to limit individual source emissions over time. For example, the recently adopted Clean Air Interstate Rule (CAIR) is a cap-and-trade program covering emissions of SO2 and NOx from EGUs in the east and midwest (19). In the “without CAIR” simulations, the only cap-andtrade programs in place are the 1990 Title IV acid rain control program and the 1998 NOx SIP Call. The former is a national cap-and-trade program for SO2 from large coal-fired power plants. Its annual allowance allocations varied little from 1995 until 2000, when they fell as the program went into its second phase. Allowance allocations will be close to constant in the future. Nonetheless, annual emissions are expected to decline over time as allowances banked early in the program are drawn down at a declining rate. The NOx SIP Call is a cap-and-trade program affecting summer (May-September) emissions in 19 eastern and midwestern states. This program became fully effective in 2005. The “with CAIR” simulations assume implementation of CAIR along with the Clean Air Mercury Rule, which caps nationwide mercury emissions from coal-fired boilers, and the Best Available Retrofit Technology rule (or Clean Air Visibility Rule), which addresses visibility around national parks and wilderness areas (20, 21). We use “CAIR” below to refer to the combination of these programs. CAIR begins in 2010 and will be fully implemented by 2015. It imposes both an annual and seasonal NOx cap. For SO2, sources affected by CAIR must hold two (rather than one) Title IV allowances per ton of emissions from 2010 to 2014 and 2.86 Title IV allowances per ton thereafter. We conduct analyses for a pre-CAIR setting even though CAIR has been adopted because we are interested in learning how the effect of NSR varies with the stringency of SO2 and NOx caps. We hypothesize that the likelihood that NSR would pull aggregate emissions below their caps depends on the levels of the caps. We also conjecture that the incremental cost of decreasing emissions below the caps increases as the caps on SO2 and NOx are tightened, since more facilities will already have pollution controls. To our knowledge, this is the first study to simultaneously capture the incentives that each policy generates and how they interact to affect temporal and spatial emissions patterns. An earlier study by EIA looked at two levels of NSR enforcement in combination with different levels of national caps on emissions of SO2, NOx, and carbon dioxide (22). Unlike this analysis, the EIA study assumed an unrealistically aggressive schedule for both NSR enforcement and the implementation of stricter emissions caps. The EIA study did not discuss the effects of the policies on spatial emissions patterns or the cost effectiveness of NSR as a policy to reduce emissions, two aspects that we explore below. We assume that no allowances are surrendered for compliance with NSR. In fact, EGUs have surrendered allowances in most of the NSR settlements to ensure that VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

349

TABLE 1.

Retrofits and Coal Use for Each Scenario 2007

2010

2015

2020

Cumulative FGD Retrofits (gigawatts) ERP low R/R/R (2%) middle R/R/R (5%) high R/R/R (7.5%)

7.8 8.0 8.0 8.0

10.4 11.0 27.8 39.8

16.0 29.8 74.9 110.4

17.1 48.7 122.1 181.1

with CAIR

ERP low R/R/R (2%) middle R/R/R (5%) high R/R/R (7.5%)

8.0 8.0 8.0 8.0

46.4 46.4 47.0 39.6

88.3 88.1 86.6 110.3

107.9 108.1 120.3 181.0

without CAIR

ERP low R/R/R (2%) middle R/R/R (5%) high R/R/R (7.5%)

20.2 21.7 21.8 22.3

25.9 24.7 28.2 40.2

33.3 35.8 75.5 111.0

35.8 49.2 122.8 181.4

with CAIR

ERP low R/R/R (2%) middle R/R/R (5%) high R/R/R (7.5%)

17.1 17.1 17.9 18.8

41.1 41.2 42.2 43.7

70.6 70.6 74.0 110.8

72.9 72.9 120.9 181.3

without CAIR

Cumulative SCR Retrofits (gigawatts)

Appalachian and Interior Coal Consumption (million tons) without CAIR

ERP low R/R/R (2%) middle R/R/R (5%) high R/R/R (7.5%)

498 502 522 542

486 493 532 563

477 499 597 624

474 563 658 661

with CAIR

ERP low R/R/R (2%) middle R/R/R (5%) high R/R/R (7.5%)

437 437 443 454

472 472 474 479

504 503 506 565

555 556 589 684

without CAIR

ERP low R/R/R (2%) middle R/R/R (5%) high R/R/R (7.5%)

577 572 551 525

603 594 554 513

631 603 496 468

714 600 495 505

with CAIR

ERP low R/R/R (2%) middle R/R/R (5%) high R/R/R (7.5%)

628 628 625 617

589 589 587 587

568 568 570 514

568 568 536 463

Western Coal Consumption (million tons)

emissions decreases at a plant are not offset by increases at other plants. Additional surrenders might be possible under the prerevision rules, but we do not know how many may occur and thus we do not model them.

Results NSR Compliance Strategies. The IPM runs show that coalfired EGUs would nearly always respond to an assumed mandate to retrofit, repower, or retire by retrofitting emission controls. Imposition of even the high (7.5%) R/R/R scenario results in less than 2% of the uncontrolled capacity retiring or repowering. Further, the different scenarios show relatively little difference in the national share of coal-fired generation. Table 1 shows the cumulative amount of capacity with FGD and SCR for each simulation year for the R/R/R and ERP solutions. Without CAIR, the amount of capacity with FGD increases linearly with the %/yr assumption in the R/R/R scenarios. But with CAIR, only the middle and high R/R/R scenarios force more FGD and SCR retrofits relative to ERP, and then only in the later years. Otherwise, the CAIR emissions caps motivate more retrofits than required by the R/R/R constraints. Temporal Distributions. The temporal emissions pattern of each prerevision R/R/R scenario differs between SO2 and NOx. Figures 1 and 2 present the total U.S. emissions in tons for those scenarios, as well as for ERP. Again, we interpret the latter as also representing the hourly emissions approach proposed by EPA. The different levels of emissions over time 350

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 2, 2008

FIGURE 1. National SO2 and NOx emissions, R/R/R and ERP scenarios without CAIR.

FIGURE 2. National SO2 and NOx emissions, R/R/R and ERP scenarios with CAIR. are influenced by our assumption that the model optimizes over the entire forecast horizon. For reference, the figures also show historical emissions in 2003. In the absence of CAIR, all R/R/R scenarios yield some emission decreases compared to ERP (Figure 1). In the high R/R/R scenario, NOx emissions in 2020 are 46% lower than with ERP. This is because there is no cap-and-trade program restricting national NOx emissions. The temporal pattern of SO2 shows some changes for the middle R/R/R scenario, but the anticipated 2% emissions decrease in 2010 is more than matched by a predicted increase of 10% in 2007, with only negligible changes in total emissions from 2007 to 2020. This occurs because the price of allowances is depressed by imposing R/R/R, thus motivating generators to draw down their bank of allowances more quickly. Only the high R/R/R scenario causes SO2 emissions to fall significantly below emissions with ERP, and then only in 2020. By that year, nearly all coal capacity has FGD, and SO2 emissions fall 41% relative to ERP. Emissions are pulled below the Title IV cap, and the price of allowances goes to $0. Thus, without CAIR, prerevision rules could significantly affect total national emissions compared to ERP or EPA’s proposed hourly emissions approach. The results under CAIR are quite different (Figure 2). The low and middle R/R/R scenarios indicate that, except for NOx in the year 2020, the prerevision rules would not pull national emissions below ERP emissions levels. In the middle R/R/R scenario, NOx from CAIR-affected sources falls 10% below ERP emissions in 2020. However, this total decrease in NOx relative to the ERP level is considerably smaller than if CAIR were not implemented. In contrast, the high R/R/R scenario under CAIR illustrates more complex interactions of the prerevision rules with emission caps. In particular, the SO2 decreases in 2015 and 2020 in the high R/R/R scenario are matched almost tonfor-ton by increases in 2007 and 2010. The 2007 and 2010 emissions are higher because the bank of Title IV SO2 allowances is drawn down more rapidly. As the amount of required FGD increases in later years in response to the high

R/R/R policy, it becomes easier to comply with the CAIR constraint on national emissions, thus lowering demand for allowances and depressing allowance prices. There is therefore less incentive for sources to hold allowances for later use, and so emissions rise in earlier years. Thus, the main effect of the high R/R/R scenario is to redistribute SO2 emissions over the study period relative to ERP scenario, but not to reduce the total. However, it is possible that post-2020 emissions will be lower under the high scenario than under ERP if emission caps are not tightened further. NOx emissions in the high R/R/R scenario with CAIR do not present the intertemporal shift exhibited by SO2. Instead, NOx emissions do not increase in earlier years, and are lower by 7% in 2015 and by 34% in 2020, relative to the ERP scenario. These emission reduction benefits, though, largely or completely disappear under the more realistic low or middle scenarios. The differing temporal patterns of NOx and SO2 emissions are partly due to the greater flexibility that generators may have to alter SO2 emissions. SO2 emissions are affected not only by postcombustion controls but also by the sulfur content of the fuel that is used. A permit-writing authority, such as a state or EPA, may well allow a coal-fired generator that previously burned low-sulfur coal to switch to less expensive high-sulfur coal once FGD is installed. This diminishes the effect of the retrofit on SO2 emissions. In addition, once EGUs are forced to use more FGD in the future, it is less attractive for sources to bank SO2 allowances for future consumption. Consequently, they would decide to use those allowances now and purchase less expensive higher sulfur coal. The effect on coal use is illustrated in the bottom half of Table 1. Without CAIR, production of high sulfur coal in the Appalachian and Interior regions is about 9% greater in 2007 in the high R/R/R scenario than in the ERP scenario. In 2020 this difference is 39%. With CAIR, the increase in Appalachian and Interior coal production in the high R/R/R scenarios compared to ERP is only 2% in 2007 and 23% in 2020. In all these cases, the increase in Appalachian and Interior coal output is roughly offset by reduced western coal production. In contrast, options for reducing NOx emissions are typically limited to pollution control retrofits or curtailment of operations. There is no easy opportunity to increase NOx emissions by substituting fuels. Spatial Distributions: CAIR vs Non-CAIR. CAIR affects only eastern and midwestern states and so we might expect that R/R/R constraints would affect emissions from sources within the CAIR region differently than sources elsewhere. Non-CAIR affected EGUs comprise about 19% of U.S. coalfired capacity. We focus on the high R/R/R scenario because it is the only scenario that yields annual emissions that noticeably differ from ERP. The results indicate that most NOx reductions under the high R/R/R scenario, relative to ERP, occur at non-CAIR-affected EGUs, although after 2015, emissions from CAIR-affected units are reduced as well. Yet SO2 reductions in 2015 and 2020 under the high scenario, relative to ERP, primarily occur at CAIR-affected plants. This contrasting pattern of emissions is explained by the interplay of emissions caps with amount retrofits at CAIR vs non-CAIR plants. Once the NOx cap no longer binds and NOx allowance prices fall to zero, which occurs around 2010 in the high R/R/R scenario, EGUs would see no financial benefit from freeing up allowances by installing R/R/R mandated-controls in the CAIR region. At the same time, low-cost NOx controls will have already been installed earlier in the CAIR region when the cap was binding (i.e., before 2010). Thus, retrofits in later years would occur in non-CAIR, i.e., western, areas where low cost retrofits are still possible. The share of capacity with SCR retrofits at non-CAIR affected coal-fired EGUs is 7% in 2015 and 2020 with ERP. But under VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

351

the high R/R/R scenario, 40% of non-CAIR affected coal capacity has SCR retrofits in 2015, rising to 92% by 2020. SO2 presents a different picture. Assuming CAIR, the existing share of non-CAIR affected capacity with FGD is already high in the ERP scenario. This is partly because these EGUs, though outside the CAIR region, are still part of the national Title IV SO2 cap-and-trade program. The share of non-CAIR sources with FGD is 59%, 63%, and 85% in 2007, 2010, and 2015, respectively, in both the ERP and high R/R/R scenarios. Only in 2020, when 88% of the non-CAIR capacity has FGD under ERP and 96% has FGD under the high R/R/R scenario, is there a noticeable difference. However, this difference is slight relative to CAIR-affected coal-fired EGUs, where 64% have FGD in 2020 with ERP while 93% have FGD with the high R/R/R scenario. The benefit of installing FGD in the CAIR region when forced by the high R/R/R scenario is to allow sources to use cheaper high sulfur coal. Thus, EGUs receive some financial benefit from installing controls in the CAIR region as long as the SO2 cap remains binding and allowances have value. NSR Emissions Reduction Costs. To assess the cost of reducing emissions through the prerevision rules, we consider the cost-effectiveness in dollars per ton reduced for each R/R/R scenario relative to the ERP scenario in both the without-CAIR and with-CAIR settings. “Cost-effectiveness” is defined as the increase in the present value of the total cost of electricity production divided by the reduction in emissions. For simplicity, we weight NOx and SO2 reductions equally. Costs and emissions from 2007 through 2020 are considered; values for years between the simulation years are obtained by linear interpolation. Two cost-effectiveness measures are calculated for each R/R/R scenario: one based on discounting emission reductions and the other considering a simple sum of reductions over those years. The costs are discounted in both measures. Calculating the ”cost-effectiveness” of reductions is therefore equivalent to calculating a levelized cost per ton of emissions reductions that discounts the costs and, in essence, discounts the benefits of these reductions. We assume a 5%/yr discount rate in the following discussion. This is the average of the two rates, 3% and 7%, that EPA uses in its regulatory impact analyses. Results of the cost-effectiveness analysis for 0%, 3%, and 7% discount rates are provided in the Supporting Information. Without CAIR, the emissions reductions from the prerevision rules, relative to ERP, cost between $850 and $5,900 per ton (1999 $). With CAIR, emissions (both undiscounted or discounted) increase in the near future as fewer allowances are banked, while emissions do not appreciably fall in the distant future. Under the low R/R/R scenario, both emissions and costs are actually higher. In the middle and high R/R/R scenarios, when emissions are reduced by the prerevision rule compared to ERP, the incremental cost is between $2,900 and $53,000/ton. Both with and without CAIR, the cost per ton of reductions is highest in the scenarios where only small emission reductions occur. For example, relative to ERP with CAIR, the discounted emission reduction from the middle R/R/R scenario is 18,000 tons (+340,000 tons of SO2, -358,000 tons of NOx) and costs an additional $950 million, yielding an average reduction cost of $53,000/ton. In contrast, under the high R/R/R scenario, the discounted reduction is about 700,000 tons (+789,000 tons of SO2 and -1,480,000 tons of NOx), costing $9.2 billion or an average of $13,000/ton. We would not expect such a fluctuating pattern of costeffectiveness, with average emission reduction costs falling as emissions decline, from tightening caps on national emissions. As an emissions cap is tightened, the average cost of reductions would always rise. In contrast, the R/R/R program, as modeled here, does not target emissions, but rather the amount of capacity that is retrofitted or retired. 352

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 2, 2008

With such a policy, there is no assurance that the EGUs that are cheapest to retrofit (consistent with our assumptions concerning the targeting of NSR enforcement) are necessarily those that achieve the cheapest emissions reductions, and potentially lower cost fuel switching or redispatch options are not rewarded. Comparison of Cost-Effectiveness of NSR and Caps. The costs per ton of reduction via the prerevision rules are large compared with the costs of achieving emission reductions using a cap alone. One indication of this is that, under ERP, the cost of CAIR-induced reductions relative to the regulatory setting without CAIR is $470/ton (undiscounted emissions) to $730/ton (discounted). Thus, starting from a regulatory setting with ERP and without CAIR, the incremental cost per ton of adopting CAIR is up to an order of magnitude smaller than the incremental cost per ton of reducing emissions by employing an R/R/R approach to NSR. (The cost per ton of R/R/R in this setting is $850 to $5,900 per ton, as reported above). It is even more revealing to examine the annual national NOx and SO2 emissions results achieved in the high R/R/R scenario and calculate the cost of instead using tightened caps to obtain the same level of national emissions in each simulation year, allowing emissions trading. We perform this analysis assuming CAIR is adopted and find that the cost of the incremental reductions achieved by this hypothetical cap-based policy to achieve the same reductions as the R/R/R scenario is $960/ton (undiscounted) and $2,600/ton (discounted). Those costs are one-third and one-fifth, respectively, of the cost of achieving the same emission decreases through the prerevision rule, assuming the high R/R/R scenario ($3,100 and $13,000, respectively). This shows that the prerevision rule is not a cost-effective way to achieve national emission reductions. Furthermore, relative to the high R/R/R scenario, this hypothetical tightened cap policy yields fewer FGD and SCR retrofits (30% less in 2020); more consumption of western low sulfur coal (14% more in 2020) and natural gas; and more Selective Non-Catalytic Reduction (which reduces NOx by about a third as much as SCR, but costs less). The prerevision NSR approach provides sources less flexibility in choosing controls than emissions trading. Sources are thus deprived of the opportunity to arrive at least-cost solutions (23), increasing the expense of achieving the same national emissions reductions. However, as with all of our analyses, we cannot say how the different approaches affect local emissions, much less whether the distribution of emissions over space is more or less harmful than the distribution under the prerevision high R/R/R scenario. We can understand why caps are more cost-effective by comparing the high R/R/R scenarios with and without CAIR, thus providing an indication of the effectiveness of economic incentives to take advantage of all abatement strategies. Those two solutions have similar amounts of FGD retrofits in every year (Table 1). But SO2 emissions with CAIR are nearly 30% less than without CAIR in 2007 and 2010, and 46% less in 2015 (compare Figures 1 and 2). The story is similar for NOx emissions. There are two reasons why these solutions have similar levels of control but different emissions. First, tighter NOx and SO2 restrictions under CAIR motivate retrofits of controls at locations where they are more cost-effective in reducing emissions, while R/R/R tends to lead to the installation of controls in order of lowest to highest cost per megawatt of capacity regardless of their cost-effectiveness in emissions reduced. Second, CAIR’s tightened emissions caps and resulting higher allowance prices motivate fuel-switching and emission-dispatch strategies that decrease emissions at uncontrolled EGUs. By contrast, the hardware orientation of

NSR provides no incentive to adopt operating strategies for reducing emissions. Sensitivity Analyses. If less pollution-intensive generating technologies turn out to be cheaper than is assumed in IPM, then the consequence of the R/R/R scenarios might be more retirements of coal EGUs and lower emissions than reported above. This would imply that adopting ERP would have more of an impact than under our base assumptions. To address this possibility, we undertook two sensitivity analyses. One assumed a 20% reduction in the investment cost for new renewable and coal gasification technologies compared to the original assumption. The other assumed lower natural gas prices: 15% lower in 2010, and 25% lower in 2020 relative to the original assumptions. The results of these analyses, reported in ref (3), show no significant change in national emissions and coal EGU retirements over the time period of study.

Acknowledgments Nathan Wilson, a Presidential Management Fellow at RfF in 2005, provided research assistance. The IPM runs were undertaken by B. Venkatesh, ICF Consulting, under the direction of Meg Victor, USEPA. We also thank the NRC study director Ray Wassel, other members of the NRC NSR committee, and former committee member Brian Mannix for many valuable suggestions. We also thank four referees for their thoughtful comments. The opinions expressed here are those of the authors, and do not necessarily express the views of the USEPA.

Supporting Information Available Costs per ton of the NSR scenarios assuming discount rates other than 5% value used above; display of the scenarios in tabular form; information on the distribution of emissions and retrofits between CAIR and non-CAIR regions. This information is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) WI Electric Power v. Reilly, 893 F.2d 901, 910 (7th Cir. 1990). (2) U.S. Department of Justice. New Source Review: an analysis of the consistency of enforcement actions with the Clean Air Act and implementing regulation; www.usdoj.gov/olp/nsrreport.pdf; January 2002. (3) National Research Council. New Source Review for Stationary Sources of Air Pollution; The National Academies Press: Washington, DC, 2006. (4) New York v. EPA, 443 F.3d 880 (D.C. Cir. 2006), cert. denied, 127 U.S. 2127(2007). (5) U.S. v. Duke Energy Corp, 411 F.3d 539 (4th Cir. 2005). (6) Environmental Defense v. Duke Energy Corp., 127 S.Ct. 1423 (2007).

(7) U.S. EPA. Prevention of Significant Deterioration, Nonattainment New Source Review, and New Source Performance Standards: Emissions Test for Electric Generating Units. Fed. Regist. 2005, 70, 61081–61103. (8) U.S. EPA. Supplemental Notice of Proposed Rulemaking for Prevention of Significant Deterioration and Nonattainment New Source Review: Emission Increases for Electric Generating Units. Fed. Regist. 2007, 72, 26202. (9) U.S. EPA. Acid Rain Program 2004 Progress Report; www.epa.gov/ airmarkets/cmprpt/arp04/2004report.pdf; October 2005. (10) U.S. EPA. National Electric Energy System (NEEDS) Database for IPM 2004; www.epa.gov/AIRMARKET/epa-ipm/ needs_2004.xls. (11) Clean Air Act. §160(4) 42 U.S.C. 7470(4). (12) Barcott, B. Changing all the rules. New York Times, April 4, 2004. (13) U.S. EPA. Documentation Summary for EPA Base Case 2004 (V.2.1.9), using the Integrated Planning Model; http://www.epa.gov/airmarkets/progsregs/epa-ipm/past-modeling.html; October 2004. (14) Energy Modeling Forum. Prices and Emissions in a Restructured Electricity Market, Vol I: Final Report of EMF Working Group 17; Stanford University: Stanford, CA, 2001. (15) U.S. EPA. Results of the Integrated Planning Model (IPM) for the Final Amendments to the Regional Haze Regulations and Guidelines for BART Determinations; www.epa.gov/oar/ visibility/actions.html#bart1; July 2005. (16) U.S. EPA. Regulatory Impact Analysis for the Specification of Categories of Activities As Routine Maintenance, Repair and Replacement for the New Source Review Program; http:// yosemite.epa.gov/ee/epa/riafile.nsf/vwAN/A2002-26.pdf/$File/ A2002-26.pdf; August 2003. (17) Keohane, N. O.; Mansur, E. T.; Voynov, A. Averting Enforcement: Strategic Response to the Threat of Environmental Regulation; University of California Energy Institute, CSEM WP-160; www.ucei.berkeley.edu/PDF/csemwp160.pdf; October 2006. (18) Bushnell, J. B.; Wolfram, C. The Economic Effects of Vintage Differentiated Regulations: the Case of New Source Review; University of California Energy Institute, CSEM WP-157; www.ucei.berkeley.edu/PDF/csemwp157.pdf; July 2006. (19) U.S EPA. Rule to Reduce-Interstate Transport of Fine Particulate Matter and Ozone (Clean Air Interstate Rule); Revisions to Acid Rain Program; Revisions to the NOx SIP Call. Fed. Regist. 2005, 70, 25162–25405. (20) U.S EPA. Standards of Performance for New and Existing Stationary Sources: Electric Utility Steam Generating Units. Fed. Regist. 2005, 70, 28606–28700. (21) U.S EPA. Regional Haze Regulations and Guidelines for Best Available Retrofit Technology (BART) Determinations; Final Rule. Fed. Regist. 2005, 70, 39104–39172. (22) U.S. Energy Information Agency. Analysis of Strategies for Reducing Multiple Emissions from Power Plants: Sulfur Dioxide, Nitrogen Oxides, and Carbon Dioxide; www.eia.doe.gov/oiaf/ servicerpt/powerplants/; December, 2000. (23) National Research Council. Air Quality Management in the United States; The National Academies Press: Washington, DC, 2004.

ES070003C

VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

353