Comment:SMC quoted from three Federal circuit court opinions that, in SMC's view, demonstrate that EPA's proposal to disapprove “the portion of Michigan's SIP related to BART requirements for [SMC-Charlevoix],” and “to substitute EPA's own limits in their place, is impermissible under the Clean Air Act.” Specifically, SMC asserted that the U.S. Court of Appeals for the D.C. Circuit inEME Homer City Generation, LPv.EPA,No. 11-1302 (D.C. Cir. August 21, 2012) and the U.S. Court of Appeals for the Fifth Circuit inLuminant Generation Co.v.EPA,675 F.3d 917 (5th Cir. 2012) andTexasv.EPA,No. 10-60614 (5th Cir. August 13, 2012) held that if a state plan meets the standards required by the Clean Air Act, EPA cannot force the states to adopt specific control measures.
Response:These decisions address rulemakings that are unrelated to regional haze and circumstances that do not invoke the same relationship between state and federal action. Moreover, these courts acknowledge that EPA has a valid role in assessing whether a state submittal is compliant with the Clean Air Act. EPA proposed to find that Michigan's submittal was not compliant with the Clean Air Act, insofar (in part) as Michigan failed to require BART for SMC-Charlevoix. SMC appears to be arguing that EPA may not disapprove a submittal that meets Clean Air Act requirements to force the State to adopt an alternative measure that EPA prefers, but EPA is not taking such an action here. Nor is EPA using the SIP process to force Michigan to adopt any particular control measure. Instead, EPA is simply fulfilling its responsibility to evaluate the State's submittal and, in the absence of a state submittal meeting applicable requirements, promulgating federal limits to meet these requirements.
Comment:SMC noted EPA's finding that Michigan's SIP “includes a reasonable set of measures that provide its appropriate share of reductions toward achieving reasonable progress goals.” (See 77 FR 46919.) SMC concluded that, because the emissions limits proposed by Michigan allow the State to meet the reasonable progress goals for improving visibility, “EPA cannot * * * require emissions limits for SMC which would go beyond allowing the State to meet those progress goals.” SMC stated that the BART requirements are included within the set of emission limits that EPA may require only as “necessary to make reasonable progress.”
Response:Clean Air Act section 169A(b)(2) provides that the measures that are necessary to provide for reasonable progress necessarily include measures representing BART. The fact that EPA codified BART requirements separately from the requirements for reasonable progress (in 40 CFR 51.308(e) versus 40 CFR 51.308(d)) supports an interpretation that BART requirements must be satisfied irrespective of whether reasonable progress goals are being met.
Another possible reading of section 169A(b)(2) is that a plan that lacks BART measures by definition fails to include all the measures that this section mandates be part of the plan for achieving reasonable progress. That is, under this interpretation, BART is necessarily a reasonable measure, and a plan, such as Michigan's, that fails to require BART cannot be considered to provide for reasonable progress.
In response to this comment, EPA is clarifying that, insofar as Michigan's plan fails to require BART on at least two facilities, Michigan's plan fails to include all reasonable measures. To that extent, Michigan's plan may be considered to fail to provide for reasonable progress, but EPA believes that the plan, in combination with the FIP (in conjunction with BART limits for Tilden Mining, being addressed separately), meets reasonable progress requirements.
Comment:SMC cited six factors listed in the definition of BART at 40 CFR 51.301 that are to be taken into consideration in determining BART. With respect to the first factor, the technology available, SMC believes that “EPA did not properly evaluate the capabilities of technology available for NOXcontrol at Charlevoix.” SMC provided a review of the history of the SMC-Charlevoix kiln system design, including conversion in the late 1970s to a preheater/precalciner design and installation of an indirect firing system.
Response:EPA has considered the design of the SMC-Charlevoix kiln system in evaluating BART for this facility, as discussed more fully below.
Comment:SMC maintained that “the normal variability of NOXformation in cement kilns justifies the 6.5 pounds per ton NOXemission limit contained in Michigan's SIP.” SMC provided a graph of emissions data for 2006 to 2008, and states that the “average of [these] data is 4.56 pounds per ton, but there is a significant standard deviation of 0.64 pounds per ton, leading to a 99.7 [percent] confidence number of 6.47 pounds per ton.”
Response:EPA recognizes the variability in NOXformation at SMC-Charlevoix. EPA addressed this variability in its proposal in part by proposing a limit in the form of a 30-day average. Further discussion of the appropriate limit in the context of this variability is provided below.
The statistic SMC cites as being the 99.7th percentile (the value three standard deviations above the mean) is in fact an even higher percentile, specifically the 99.87th percentile. Although EPA is basing its limits on the 95th percentile baseline emissions, this error is worth noting because EPA is avoiding the same error in estimating the 95th percentile baseline emissions. This error presumably reflects confusion between two statistical values, one being the percent of values within three standard deviations both above and below the mean, and the other being the percent of values between zero and a value that is three standard deviations above the mean. The latter statistic is the appropriate statistic in finding percentiles, since a given percentile is the value that exceeds that percentage of the entire distribution, including values down to zero, not just the portion of the distribution down to another value for example three standard deviations below the mean. In a normal distribution, 49.87 percent of values are between the mean and three standard deviations above the mean, and the same 49.87 percent of values are between three standard deviations below the mean and the mean, for a total of 99.74 percent of values within three standard deviations of the mean. In contrast, in determining percentile values, one must sum the 49.87 percent of values that are below three standard deviations above the mean but above the mean together with the full 50 percent of values that are below the mean. Thus, the value three standard deviations above the mean in a normal distribution is the 99.87th percentile value, not the 99.74th percentile value. For similar reasons, EPA is estimating 95th percentile baseline emissions as the value 1.645 standard deviations above the mean, rather than the value 1.96 standard deviations above the mean that SMC's approach would suggest.
Comment:SMC commented that it “has put in place more modern technology than its competitors, such as Lafarge's Alpena plant.” Elsewhere, SMC cited other plants with higher emission limits which, it claims have “not been upgraded to the same degree as the Charlevoix plant,” and noted that “SMC already outperforms those [limits] with the improvements it already has put in place.”
Response:With the consideration of source-specific factors, as required in determining BART at each facility, dissimilarities among facilities can yield dissimilarities in control requirements. Lafarge's Alpena facility has long wet kilns, a different design with inherently more NOXemissions than SMC-Charlevoix's preheater/pre-calciner kiln. In fact, BART at Lafarge requires similarly effective SNCR there as at SMC-Charlevoix, and BART at Lafarge requires sulfur emission control that is not required at SMC-Charlevoix.
Comment:SMC asserted that “EPA will expect compliance with its emission limit every day, not just `on average' over several years. Therefore, EPA also was incorrect when it derived its proposed NOXemission limit of 2.3 [lb per ton] for the Charlevoix plant by applying a presumed 50 percent reduction against the plant's 4.56 [lb per ton] average, which was achieved over several years. *** An `average' value means that half of the actual performance is greater than that average. Therefore, any proposed reduction should not be applied to an average performance over several years, but instead must take into consideration the normal standard deviation from that average. This is the same rationale that was recently used by EPA in its agreement with Holcim's Montana Plant. Consequently, in this instance, if there was to be any reduction, it must be applied against the 6.5 [lb per ton] value which represents the 99.7 percent confidence value of SMC's actual performance.”
Response:SMC is noting the variability in emissions at SMC-Charlevoix, observing that a several year period will include many occasions with baseline emissions that are above average, and commenting that any emission limit should be based on those elevated baseline emission conditions. EPA addressed this concern in its proposed rulemaking. EPA proposed a limit that would require an average control of approximately 50 percent. In addition to defining the limit as a 30-day rolling average, EPA's notice of proposed rulemaking describes an examination of the variability of emissions at SMC-Charlevoix and the feasibility of achieving the proposed limit even during periods with greater emissions formation. The proposed rulemaking states, “According to 2006 to 2008 data from the facility, [the proposed limit] would require slightly under 60 percent control from St. Marys Cement's 95th percentile 30-day average emission rate, which the evidence from tests at St. Marys Cement's facility in Dixon, Illinois (SMC-Dixon) indicates is readily achievable, particularly since a limit of 2.30 lb per ton of clinker would only occasionally require this level of control.” 77 FR 46924. Conversely, at the 5th percentile of the 30-day average emission rates, or 3.5 lb per ton, the proposed limit would require only about 35 percent control. In this sense, EPA proposed a limit that would sometimes require about 60 percent control, sometimes require only about 35 percent control, and on average require slightly less than 50 percent control.
Thus, EPA considered the variability of baseline emissions but also considered the variability of control effectiveness in determining its proposed emission limit. Nevertheless, as discussed below, EPA is modifying its view of achievable control efficiencies and is modifying its approach for determining appropriate limits accordingly.
Comment:“Although better performing than other old plants, unique Charlevoix design features increase NOXformation compared to the most modern kiln designs.” SMC discussed the ratio of the kiln length to kiln diameter at SMC-Charlevoix, as well as the need to operate the kiln in an oxidizing atmosphere to minimize the likelihood of formation and buildup of calcium sulfate. SMC concluded that these factors raise the amount of energy needed to produce a kilogram of clinker from about 800 kilocalories to about 930 kilocalories, which raises expected NOXemissions per ton of clinker.
Response:Average NOXemissions at SMC-Charlevoix are about 4.5 lb per ton of clinker. According to the Compilation of Air Pollution Emission Factors (AP-42), average emissions for a representative cement plant of the design of SMC-Charlevoix, i.e., a preheater/precalciner kiln, is 4.2 lb per ton of clinker. Thus, SMC-Charlevoix has very typical NOXemissions for a facility of its type.
While it may be true that NOXemissions at SMC-Charlevoix are slightly higher than those at newer plants, EPA is also setting a higher limit for SMC-Charlevoix than we have set fornew cement plants. The new source performance standards for cement plants require NOXemission rates not to exceed 1.5 lb per ton of clinker. Were EPA to require similar rates for SMC-Charlevoix, but allow for the 16 percent increase in heat input noted in the comment, EPA would be imposing an emission limit of 1.74 lb per ton of clinker, rather than the 30-day average limit of 2.8 lb per ton of clinker finalized in this rule.
Comment:“EPA's conclusion that SNCR will allow a 50 percent reduction in NOXemissions from the Charlevoix plant is incorrect because the plant's design is incompatible with effective SNCR use.” SMC argued that the achievement of emission rates as low as 2.3 lb per ton requires kiln design features “(e.g., proper kiln length to diameter dimensions and increased calciner retention time)” that are not present at SMC-Charlevoix. SMC provided a figure identifying temperatures and residence times at various locations within the kiln system, and concludes that “nowhere in the kiln riser or flash calciner regions of the system does the plant reach the optimum temperature profile to support an effective SNCR reaction.” SMC also found that the “residence time at Charlevoix is not adequate for use of SNCR.” SMC provided a graph entitled “SNCR Efficiency based on Residence Time (Lab Trial).” SMC stated that at SMC-Charlevoix, “there is only a 0.11 second retention time between the reagent injection point and the time the system reaches the low end of efficiency point for the SNCR reaction.” SMC further quotes EPA and other work suggesting that “larger plants had lower efficiencies than smaller sized plants.”
SMC stated, “Actual test results demonstrate that SNCR will have only limited success in NOXcontrol at Charlevoix.” SMC presented results of trial urea injections conducted in 2005 to test the NOXreductions that an SNCR system might be expected to achieve. SMC described these tests as demonstrating that urea injection achieved less NOXreduction than expected. SMC provided results in a table that gives average NOXreduction percentages for four sets of tests, each conducted with urea injection at a different location in the kiln system and with a different urea injection rate. The table also gives urea injection rate in terms of the normalized stoichiometric ratio (NSR).2
“In one test run, [with an NSR equal to 1.07], the reduction was 36.8 percent. * * * However, that was coupled with a significant amount of ammonia slip, based on the theoretical calculations from the NOXpresent. The time frames for this trial were short, roughly several 10 minute runs to consolidate the average, and thus SMC is not confident that these reductions are sustainable.” SMC provided a photograph that it considers to document excess ammonia (ammonia slip) appearing as a visible detached plume occurring at SMC-Charlevoix.
2Normalized stoichiometric ratio expresses the ratio of the number of moles of ammonia equivalent to the pre-control number of moles of NOX. Each molecule of urea yields the equivalent of two molecules of ammonia. Thus, for example, if 0.6 moles of urea (yielding 1.2 moles of ammonia) are injected per mole of NOX, NSR = 1.2.
SMC provided a report from DeNOXTechnologies describing the urea trials. SMC quoted from this report: “Typically, NOXreduction at a NSR of 1.0 is 40-60 percent; Charlevoix demonstrated 25-30 percent.” In addition, SMC stated, “DeNOX's owner noted * * * that he had seen SNCR effectively solve NOXissues in multiple cement plants. However, he commented to SMC that he was amazed that SNCR is not as efficient in SMC's system, and he believed it must be because of Charlevoix's calciner design.”
Response:EPA believes that the tests of SNCR at SMC-Charlevoix do not demonstrate that SNCR would be ineffective in reducing emissions, and in particular do not demonstrate that SMC could not meet the emission limits established in this final action. EPA notes that the tests SMC described were performed with urea rather than with ammonia, which is both more commonly used for this application and significantly more effective.
SMC-Charlevoix's test results were the subject of “SNCR emission control,” published in the August 2006 edition of the journal International Cement Review (the Horton article).3
The article presents NOXreductions resulting from urea injection at “Plant B,” which are the results found at SMC-Charlevoix. The article also includes contrasting results from testing at “Plant A,” a plant with the same type of design as SMC-Charlevoix, demonstrating that NOXreductions of more than 50 percent could be achieved with ammonia injection at an NSR as low as 0.56 (i.e., the injection of only 0.56 moles of ammonia per mole of NOX). The article includes a graph showing that use of ammonia achieves higher NOXreductions than urea and has maximum efficiency at lower temperatures than urea. EPA views the 50 percent reduction at Plant A as more representative of the level of emission reduction that a properly designed and operated SNCR at SMC-Charlevoix could achieve. In fact, at the temperatures at SMC-Charlevoix cited by SMC, use of ammonia is expected to provide at least 40 percent more, and possibly greater than twice as much, NOXreduction as is expected from use of urea. Thus, while SMC's concerns may apply to SNCR using urea, EPA believes that SMC can address these concerns by using ammonia.
3Joe Horton, Suwannee American Cement/Votorantim Cimentos North America, Al Linero, Florida Department of Environmental Protection, and F. MacGregor Miller, Cement Etc., Inc, “SNCR Emission Control,” International Cement Review, August 2006.
EPA also believes that the DeNOXTechnologies report cited by SMC demonstrates that SMC-Charlevoix can achieve significant NOXemission reductions even using urea. Table 1 presents relevant information derived from the DeNOXTechnologies report. During these trials, urea was injected at three locations: (1) After the kiln but before the tertiary air inlet, (2) in a duct after the tertiary air but before the precalciner, and (3) after the first stage of the preheater that is after the precalciner. In Table 1, the reduction per mole of reagent (ammonia equivalent) is computed by dividing the NOXreduction percentage by the NSR.
Table 1—NOXEmission Reductions at SMC-Charlevoix From Injection of Urea
Reduction per mole reagent(percent)
Before Tertiary Air
Before Tertiary Air
After Tertiary Air
These results suggest the relationship between the quantity of reagent and the NOXreduction. Notably, as increasing amounts of urea are injected, the resulting NOXreductions increase correspondingly. Examined in terms of NOXreduction per mole of ammonia equivalent injected, while some loss of efficiency is expected, the efficiency of urea utilization even at the highest urea injection rate is similar to the efficiency of urea utilization at the lowest urea injection rate. These results also suggest that the control efficiency is similar across several urea injection locations.
EPA believes that these tests demonstrate that SNCR at SMC-Charlevoix as it is currently configured can readily achieve at least 30 to 37 percent NOXreduction. As discussed above, EPA believes that use of ammonia would provide significantly greater control than was found in the tests at SMC-Charlevoix using urea. The tests, being short tests, by definition did not test the sustainability of control, but SMC provides no evidence that these short-term results could not also be achieved over longer periods. In addition to the change in reagent, SMC has a range of options for optimizing SNCR effectiveness and addressing the potential operational issues arising from SNCR use. These include: Use of facility design modifications that either reduce NOXemissions directly or facilitate use of SNCR or both; use of reagent injection both before and after the calciner; use of lime injection; adjustment of air flows; and other changes in operating characteristics. SMC in its written comments and in discussion during meetings with EPA did not address the option of using ammonia, either to dispute the feasibility of its use or to provide evidence regarding its effectiveness at SMC-Charlevoix. Since the tests at SMC-Charlevoix used urea and are not indicative of the NOXreduction that can be achieved using ammonia, the most pertinent evidence regarding potential effectiveness of SNCR using ammonia is the results of tests at SMC-Dixon, corroborated by results of tests at “Plant A” in the Horton paper and elsewhere. This evidence indicates that the 50 percent NOXemission reduction required at other cement plants is also achievable at SMC-Charlevoix.
The issues raised in SMC's comments suggest that SMC may need more than three years to explore the various alternatives for reducing NOXemissions at SMC-Charlevoix. Therefore, EPA is promulgating a compliance deadline for SMC that is extended by one year from the compliance deadline that EPA proposed, requiring compliance within approximately four years from the date of this rulemaking.
In response to this comment, EPA also reevaluated the appropriate NOXlimits. While EPA proposed a limit based on 50 percent control on average, effectively requiring 60 percent control when emission rates are at the 95th percentile level, EPA is promulgating a limit that will require only 50 percent control when emission rates are at the 95th percentile level.
EPA proposed a single limit, based on a 30-day average. Reconsidering the basis for determining the level of the limit, in particular considering a limit based on the 95th percentile emission level rather than based on the mean emission level, requires reconsidering the form of the standard. Whereas the proposed limit was intended to require a reasonable degree of control at all times, a 30-day average limit derived from 95th percentile emissions would allow substantially less emission reduction on other occasions. For example, at SMC-Charlevoix, a limit requiring 50 percent reduction from 95th percentile emissions would only require about 20 percent emission reduction at the 5th percentile emission level.
BART reflects controlling emissions at all times, not just on occasions with elevated emissions. For this reason, along with a 30-day average emission limit, EPA is also promulgating a limit on 12-month average emissions. In this pair of limits, the 30-day average limit ensures that days with high baseline emissions are well controlled, and the 12-month average limit ensures that BART control is achieved on days with lower baseline emissions as well.
EPA used the most recent three years of emissions data available, from 2006 to 2008, to compute 30-day averages and 12-month averages. EPA is setting the 30-day average limit as a daily-rolling average limit, based on values recomputed every operating day to include the most recent 30 operating days, and EPA is setting the 12-month average as a block average, based on values recomputed at the end of each calendar month to include the preceding 12 calendar months. EPA used these averaging approaches to determine the distribution of 30-day and 12-month averages of NOXemissions during the 2006 to 2008 period. The 95th percentiles among these sets of values (more precisely, 1.645 standard deviations above the means, calculated assuming a normal distribution) are a 30-day average of 5.6 lb per ton of clinker and a 12-month average of 4.7 lb per ton of clinker. EPA is setting limits based on a 50 percent reduction from these values, which with rounding equal a 30-day average limit of 2.80 lb per ton of clinker and a 12-month average limit of 2.40 lb per ton of clinker.
EPA had several reasons for selecting the 95th percentile of baseline emissions as the starting point for determining the limits. First, use of the 95th percentile is an approach that EPA commonly uses in setting emission limits for similar sources in other contexts. For example, the consent decree for Lafarge Cement, which requires BART at Lafarge's Alpena facility, mandates control at the 95th percentile level. That is, this approach is responsive to SMC's concerns about EPA providing equity in its regulation of SMC and Lafarge. (Lafarge is also subject to both a 30-day average limit and a 12-month average limit.) Second, EPA considers the 95th percentile an appropriate compromise between setting the limit based on too low a percentile, which creates a higher percentage of time when the limit is more difficult to meet, and setting the limit based on too high a percentile, which too infrequently requires the company to achieve fully effective emission control. Third, EPA believes that the variability of the emission rates after control is likely to be less than the current variability. This is in part because the emission control can be operated in a manner that minimizes the difference in emission rates between the upper and the lower end of the distribution, in part because emissions control tends to be more effective when emission rates are higher, and in part because the limit will give the company incentive to use its knowledge about operating parameters that influence emission rates to minimize emissions on occasions with higher emission rates. Fourth, since emission rates above the 95th percentile by definition rarely occur, any extra effort needed to achieve the limit on such occasions would rarely be needed.
SMC cites the limit for a Holcim plant in Montana as precedent for basing a limit on an upper point on the distribution, and yet SMC recommends basing the limit for SMC-Charlevoix on a more extreme statistic than was used for Holcim in Montana. EPA set the NOXlimit for Holcim by assuming a 58 percent reduction from the 99th percentile of baseline emissions. In that case, EPA had limited information on emissions of the facility; in particular, EPA did not have information on 95th percentile emissions. SMC does not explain why it seeks the use of a more extreme statistic (supposedly the 99.7th percent, but in fact the 99.87thpercentile), but the availability of more information allows EPA to use a more appropriate statistic (the 95th percentile) for SMC-Charlevoix.
Comment:SMC stated that “ammonia slip is a likely result of use of SNCR at Charlevoix.” SMC quoted from EPA and the Portland Cement Association that use of SNCR under suboptimal conditions can result in unwanted ammonia emissions.
Response:SMC does not demonstrate that proper use of SNCR at SMC-Charlevoix would cause ammonia slip at problematic levels. The photo of a detached plume at SMC-Charlevoix provided by SMC in its comments does not demonstrate that ammonia concentrations in the plume were high, and SMC does not provide information about operating conditions at the time of the picture to be able to judge this and other potential explanations of a detached plume at the facility. A theoretical comparison of urea input to NOXlevels does not establish the presence or absence of ammonia slip, because such an approach fails to consider other factors reducing ammonia levels such as oxidation. In addition, for reasons discussed in the Horton paper cited above, describing the relative merits of using ammonia rather than urea, evidence that ammonia slip occurred during injection of urea does not necessarily indicate that ammonia slip would occur with a properly designed and operated SNCR using ammonia. While SMC would have to design an SNCR system carefully to avoid causing excess ammonia emissions, many other cement plants have successfully implemented SNCR without ammonia slip problems, and SMC has provided no evidence that this would be a challenge that cannot be solved at SMC-Charlevoix. As discussed above, EPA anticipates that SMC will conduct a variety of trials to assess the most effective NOXcontrol program, and EPA anticipates that one of the parameters to be addressed in these trials is to avoid emitting excess ammonia.
Comment:SMC stated that the “size of Charlevoix reduces its ability to control NOXusing SNCR.” SMC quoted an EPA report regarding NOXcontrol at coal-fired electric utility boilers stating that “whereas smaller boilers may be able to achieve >60 percent NOXreduction, larger boilers may be capable of achieving reductions of only ∼30 percent.” SMC comments that a study of cement kilns also noted a “correlation between plant size and reduction efficiency.” SMC provided a graph labeled “SNCR Test Results based on Capacity.” SMC concludes that SMC-Charlevoix “should not be expected to have NOXreduction efficiencies of the smaller plants.”
Response:SMC does not clarify its size in relation to the other facilities addressed in these studies. Since SMC-Charlevoix has lower heat input than many electric utility boilers, this comment would seem to suggest that SMC should be able to achieve the higher rather than the lower end of the range of utility boiler control efficiencies. The graph addressing cement plants that SMC provided is illegible, and so it is indeterminable from this graph how the size of SMC-Charlevoix compares to the size of other cement plants tested.
However, EPA also examined the size of SMC-Charlevoix relative to the size of cement plants that have been subject to best available control technology determinations for new sources or major modifications in the last 6 years. These facilities have capacities quite similar to the capacity of SMC-Charlevoix. As seen in the EPA's RACT/BACT/LAER Clearinghouse, these facilities were typically issued permits that allowed 1.95 lb of NOXemissions per ton of clinker. Thus, even if smaller facilities are capable of even better NOXcontrol, this evidence makes clear that the size of SMC-Charlevoix should not prevent SMC from achieving the level of control that EPA proposed to require.
Comment:SMC submitted several comments regarding the second factor to be considered in determining BART, namely the costs of compliance. The first of these comments reflected concerns about material buildup exacerbated by injection of urea and the costs that SMC would face in addressing that problem. SMC commented “Both SMC and EPA recognize that there are potential solutions [to this problem.] * * * The most effective solution is an extensive modification to the flash calciner including geometry changes to the process ductwork.” SMC estimated that a new in-line calciner would cost $18,000,000. SMC also discussed a second option in which SMC uses its existing kiln system configuration. In conjunction with criticism of EPA's cost estimates, SMC provided its own cost estimates for these two options.
Response:EPA agrees that SMC has multiple opti