Based on the statute, as interpreted in the Brick MACT decision, we believe a source's performance resulting from the presence or absence of HAP in raw materials must be accounted for in establishing floors; i.e., a low emitter due to low HAP proprietary raw materials can still be a best performer. In addition, the fact that a specific level of performance is not being intentionally achieved by the source is not a legal basis for excluding the source's performance from consideration. Sierra Club v. EPA, 479 F.3d at 63134; National Lime Ass'n, 233 F. 3d at 640.

The Brick MACT decision also reiterated that EPA may account for variability in setting floors. However, the Court found that EPA erred in assessing variability because it relied on data from the worst performers to estimate best performers' variability, and held that ``EPA may not use emission levels of the worst performers to estimate variability of the best performers without a demonstrated relationship between the two.'' 479 F. 3d at 882.

After considering the implications of this decision, EPA granted the petition for reconsideration of all the existing source standards in the 2006 rulemaking.

A second Court opinion of relevance to the Portland cement NESHAP amended here is Sierra Club v. EPA, 551 F. 3d 1019 (DC Cir. 2008). In that case, the court vacated the regulations contained in the General Provisions which exempt major sources from CAA section 112(d) standards during periods of startup, shutdown and malfunction (SSM). The regulations (in 40 CFR 63.6(f)(1) and 63.6(h)(1)) provided that sources need not comply with the relevant CAA section 112(d) standard during SSM events and instead must ``minimize emissions * * * to the greatest extent which is consistent with safety and good air pollution control practices.'' The current Portland Cement NESHAP references the now vacated rules in the General Provisions. As a result of the court's decision, we are removing the references to the vacated provisions and addressing SSM in this rulemaking. Discussion of this issue may be found in Section IV.A.
III. Background Information on the NSPS 40 CFR Part 60, Subpart F

NSPS implement CAA section 111(b) and are issued for categories of sources which cause, or contribute significantly to, air pollution which may reasonably be anticipated to endanger public health or welfare. The primary purpose of the NSPS is to attain and maintain ambient air quality by ensuring that the best demonstrated emission control technologies are installed as the industrial infrastructure is modernized. Since 1970, the NSPS have been successful in achieving longterm emissions reductions in numerous industries by assuring cost effective controls are installed on new, reconstructed, or modified sources.

Section 111 of the CAA requires that NSPS reflect the application of the best system of emission reductions which, taking into consideration the cost of achieving such emission reductions, any non air quality health and environmental impact and energy requirements, the Administrator determines has been adequately
[[Page 54974]]
demonstrated. This level of control is commonly referred to as best demonstrated technology (BDT). EPA promulgated Standards of Performance for Portland Cement Plants (40 CFR, part 61 subpart F) in 1971 ((36 FR 24876, December 23, 1971).

Section 111(b)(1)(B) of the CAA requires EPA to periodically review and revise the standards of performance, as necessary, to reflect improvements in methods for reducing emissions. We have conducted three reviews of the standards (39 FR 20793, June 14, 1974; 39 FR 39874, November 12, 1974; and 53 FR 50354, December 14, 1988).

We proposed the current review of the Portland Cement Plant NSPS on June 16, 2008. We received a total of 46 comments from the Portland cement industry, environmental groups, State environmental agencies and others during the comment period. This final rule reflects our consideration of all the comments we received. Detailed responses to the comments not included in this preamble are contained in the Summary of Public Comments and Responses document which is included in the docket for this rulemaking.

IV. Summary of EPA's Final Action on the Amendments

In this section we discuss the final amendments to 40 CFR part 63 subpart LLL and part 60 subpart F, the changes since proposal, and the rationale for the changes. Responses to specific comments may be found in the response to comment section of this document or in the response to comment documents contained in the dockets for this rulemaking.

As a preliminary matter, EPA notes that certain portions of the existing rules are not being amended substantively but are being reprinted, sometimes with editorial changes, in today's regulatory text. As explained at proposal, EPA did so either for readers' convenience or to make certain nonsubstantive ``plain English'' changes to rule text. 74 FR at 21140. The final rule text makes these same nonsubstantive changes (which did not occasion public comment), and reprints certain existing provisions. Provisions from the existing rules which do not change substantively include the PM emission limits for kilns currently subject to the NSPS, the opacity limits for raw materials dryers, raw mills, and finish mills, and the limits for dioxin furan (D/F) for cement kilns. We reorganized the testing and monitoring requirements of both rules to make them more consistent, and modified the rule language to better conform with the June 1, 1998, Executive Memorandum on Plain Language in Government Writing. A. What are EPA's final actions on 40 CFR part 63, subpart LLL? 1. What are the final actions on emission limits under 40 CFR part 63, subpart LLL?

In this action, we are amending the emission limits for mercury, THC, and PM from new and existing kilns located at a major or area source, and for HCl from new and existing kilns located at major sources. We identify these standards below for the emission sources in a typical Portland cement production process. We have applied the limits for existing and new sources in this final rule for mercury and THC to area sources consistent with CAA section 112(c)(6). As noted above, mercury is one of the pollutants specifically singled out by Congress in CAA section 112(c)(6), and THC is a surrogate for POM and PCB, which are also section 112(c)(6) HAP. See 63 FR 14193, March 24, 1998 (determination to control all THC emissions from the source category under MACT standards). Finally, Portland cement kilns are a listed area source category for urban HAP metals pursuant to CAA section 112(c)(3), and control of these metal HAP emissions (via the standard for the PM nonmercury HAP metal surrogate) is required to ensure that area sources representing 90 percent of the area source emissions of urban metal HAP are subject to CAA section 112 control, as required by CAA section 112(c)(3). The PM standards for area sources reflect MACT, as explained below.

a. Changes to Overall Floor Setting Procedure

The MACT floor limits for each of the HAP and HAP surrogates (mercury, THC, HCl, and PM) are calculated based on the performance of the lowest emitting (considered best performing in this rulemaking) sources in each of the MACT floor pools for each HAP or HAP surrogate. We ranked all of the sources for which we had data based on their emissions and identified the lowest emitting 12 percent of the sources for which we had data, which ranged from two kilns for THC to 11 kilns for mercury for existing sources. For new source MACT, the floor was based on the best controlled source.

In assessing sources' performance, EPA may consider variability both in identifying which performers are ``best'' and in assessing their level of performance. Brick MACT, 479 F. 3d at 88182; see also Mossville Envt'l Action Now v. EPA, 370 F.3d 1232, 124142 (DC Cir 2004) (EPA must exercise its judgment, based on an evaluation of the relevant factors and available data, to determine the level of emissions control that has been achieved by the best performing sources considering these sources' variability).

Variability in cement kilns' performance has a number of causes. For many of the pollutants, notably mercury and THC, most kilns do not have addon control devices. The main source of variability for these pollutants consequently is the differing mercury and organic concentrations in the raw materials and fuels which are fed to the kiln. For particulate matter, which is wellcontrolled by baghouses, the variability is chiefly due to variations in performance of the control device for which both runtorun and testtotest variability must be accounted.\8\
\8\ Runtorun variability is essentially withintest
variability, and encompasses variability in individual runs comprising the compliance test, and includes uncertainties in correlation of monitoring parameters and emissions, and imprecision of stack test methods and laboratory analysis. 72 FR at 54877 (Sept. 27, 2007). Testtotest variability results from variability in pollution device control efficiencies over time (depending on many factors, including for fabric filters the point in the maintenance cycle in which a fabric filter is tested). Testtotest variability can be termed longterm variability. 72 FR at 54878.

In determining the MACT floor limits, we first determine the floor, which, as explained above, for existing sources is the level achieved in practice by the average of the top 12 percent of existing sources, or the level achieved in practice by the best controlled similar source for new sources. In this rule, EPA is using lowest emissions as the measure of best performance.

We then assess variability of the best performers by using a statistical formula designed to estimate a MACT floor level that is equivalent to the average of the best performing sources based on future compliance tests (or calculated inputs in the case of mercury). Specifically, the MACT floor limit is an upper prediction limit (UPL) calculated with the Student's ttest using the TINV function in Microsoft Excel^[reg]. The Student's ttest has also been used in other EPA rulemakings (e.g., NSPS for Hospital/Medical/ Infectious Waste Incinerators, NESHAP for Industrial, Commercial, and Institutional Boilers and Process Heaters) in accounting for variability. A prediction interval for a future observation is an interval that will, with a specified degree of confidence, contain the next (or some other prespecified) randomly selected observation from a population. In other words, the prediction interval estimates what the upper bound of future values will be, based upon present or past background samples taken. The UPL
[[Page 54975]]
consequently represents the value which we can expect the mean of future observations (3run average for HCl, 30day average for mercury, PM, HCl (sources not having wet scrubbers or otherwise electing CEM based compliance), and THC) to fall below within a specified level of confidence, based upon the results of an independent sample from the same population. In other words, if we were to randomly select a future test condition from any of these sources (i.e., average of 3 runs or 30day average) we can be 99 percent confident that the reported level will fall at or below the UPL value. Use of the UPL is appropriate in this rulemaking because it sets a limit any single or future source can meet based on the performance of members of the MACT pool.

This formula uses a pooled variance (in the s \2\ term) that encompasses all the datapoint to datapoint variability of the best performing sources comprising the MACT floor pool for each HAP. Where variability was calculated using the UPL statistical approach (i.e., for the Hg, HCl, and PM standards), we used the average (or sample mean) and sample standard deviation, which are two statistical measures calculated from the data distributions for mercury, HCl, and PM. The average is a central value of a data set, and the standard deviation is the common measure of the dispersion of the data set around the average. We describe in detail in the preamble sections on mercury, HCl and PM and in the memorandum ``Development of the MACT Floors for the Final NESHAP for Portland Cement'', August 6, 2010'' how these averages were developed. We note here that the methodology accounts for both shortterm and longterm variability and encompasses runtorun and testtotest variability. The formula also applies differently depending on how the underlying data set is distributed. To this end, EPA carefully evaluated the data sets for each HAP to ascertain whether the data were normally distributed, or distributed in some other manner (i.e., log normally). After applying standard and rigorous statistical tests (involving the degree of ``skewness'' of the data), we determined that the distributions for mercury and particulate matter were approximately a normal distribution, which in turn determined the final form of the UPL equation. See Floor Calculations for Final Portland Cement NESHAP, August 6, 2010; see also 75 FR at 3201920.

EPA was able to reasonably calculate variability for the THC and HCl standards without needing to use predictive statistics. Specifically, the data set for THC contains a sufficient number of observations to estimate the variability without the need of any type of statistical intervals (no UPL needed to be calculated). For HCl, although EPA applied the UPL formula in developing the HCl standard, the key issue for the HCl data set is the HCl analytic method's detection limit, which ultimately dictated the level of the standard.

At proposal we adopted a form of the UPL equation that has been used in a previous rulemaking. 69 FR 21233 April 20, 2004. Commenters stated correctly that there was an error in the equation used at proposal. As a result of these comments, EPA corrected the formula in the final rule. The UPL used in the final rule is calculated by: [GRAPHIC] [TIFF OMITTED] TR09SE10.000
Where:
x = the mean of the sample data set
n = the number of test runs
m = the number of test runs in the compliance average
s\2\ = observed variance
t = student t distribution statistic
This calculation was performed using the following Excel functions: Normal distribution: 99 percent UPL = AVERAGE(Test Runs in Top 12percent) + [STDEV(Test Runs in Top 12percent) x TINV(2 x probability, n1 degrees of freedom)*SQRT((1/n)+(1/m))], for a onetailed tvalue, probability of 0.01, and sample size of n
This is the same UPL equation that EPA used in more recent rulemakings. See 75 FR 32020 (June 4, 2010) and 75 FR 31905 (June 4, 2010). The value of ``m'' denotes the number of future observations, and it is used to calculate an estimate of the variance of the average of m future observations. For example, if 30day averages are used to determine compliance (m=30), the amount of variability in the 30day average is much lower than the variability of the daily measurements in the data base, which results in a lower UPL for the 30day average.

As an illustration of the effects that correcting the UPL had on the emission limits, we calculated the UPLs for mercury and PM using the proposal version of the UPL formula, and the version used in this final rule. The results of these calculations are presented in Table 1. Both calculated limits are about 20 percent lower when the corrected UPL formula is used.

Table 1Comparison Emission Limits Calculated Using Proposal UPL Formula Versus Corrected UPL Formula for Existing Sources Proposal Proposal (uncorrected UPL (corrected UPL formula) formula) Mercury, (lb/MM tons feed) [lb/ 29.6 [48.8] 22.5 [37.1] MM tons clinker]...............
PM (lb/ton clinker)............. 0.05 0.04 b. Ramifications of EPA Statistical Approach

A number of commenters maintained that this final rule raises the (perceived) quandry voiced by Judge Williams in his concurring opinion in Brick MACT where an achieved level of performance for purposes of CAA section 112(d)(3) results in a standard which is unachievable under CAA section 112(d)(2) because it is too costly or not costeffective. Brick MACT, 479 F. 3d at 88485. EPA is of course mindful of the repeated admonitions (with accompanying vacaturs and remands) from the DC Circuit that MACT floors must reflect achieved performance, that HAP content of process inputs (raw materials and fuels) must be accounted for in ascertaining sources' performance, and that costs cannot be considered by EPA in ascertaining the level of the MACT floor. See, e.g., Brick MACT, 479 F. 3d at 88081, 88283; NRDC v. EPA, 489 F. 3d 1364, 1376 (DC Cir. 2007) (``Plywood MACT''); see also Cement Kiln Recycling Coalition v. EPA, 255 F. 3d 855, 86162 (DC Cir. 2001) (``achievability'' requirement of CAA section 112(d)(2) cannot override the requirement that floors be calculated on the basis of what best performers actually achieved). EPA is also mindful of the need to account for sources' variability (both due to control device [[Page 54976]]
performance and variability in inputs) in assessing sources' performance when developing technologybased standards. See, e.g., Mossville Environmental Action Now v. EPA, 370 F. 3d 1232, 1242 (DC Cir. 2004); National Lime I, 627 F. 2d 416,43334(DC Cir. 1980). EPA has carefully developed data for each standard, assessing both technological controls and HAP inputs in doing so. For mercury, EPA used the pooled variance from all of the best performing kilns in the MACT floor pool in order to fully assess these kilns' intraquarry and other variable mercury levels. EPA also used pooled variance to assess the variability of HCl and PM emissions for the MACT floor pool kilns. See 70 FR at 59438 (Oct. 12, 2005) (explaining when use of such pooled variances can be reasonable). EPA has also adopted 30day averaging periods for all of the standards, further allowing short term fluctuations to be averaged out over the 30day period.

The result are floors which reasonably estimate the performance over time of the best performing sources, as do the standards based on those floors. It is true that many sources will need to install controls to meet these standards, and that these controls have significant costs (although EPA estimates that the rule's costs are substantially outweighed by its benefits). See Section VI below. This is part of the expected MACT process where, by definition, the averaged performance of the very best performers sets the minimum level of the standard. The Agency believes that it has followed the statute and applicable case law in developing its floor methodology.

Industry commenters nonetheless maintained that EPA had not properly accounted for variability of the best performing sources because not even these sources can meet the standards which are predicated on their own performance without adding controls. This contention lacks a basis in the record. For mercury, all performers in the MACT floor poolnot just those with emissions below the average of the best performers meet the promulgated standard (highest 30day average in MACT pool is 41.63 lb/MM tons clinker; the standard is 55 lb/MM tons clinker (30day average). In addition, several additional kilns, which are not in the pool of best performers, meet the standards. For THC, all kilns in the pool of best performers meet the promulgated standard (highest 30day average in MACT pool is 5.68 ppmv; the standard is 24 ppmv). In addition, seven additional kilns which are not in the pool of best performers meet the standards. Indeed, nine of the 11 kilns for which EPA has CEM data are meeting the promulgated standards for THC. For PM, all six kilns in the MACT pool as well as twelve kilns overall meet the promulgated 30day standard even though the measurements in the data base are stack tests (i.e., unlike for mercury and THC, these are not averaged values).\9\ Virtually all kilns in the MACT floor pool are meeting the HCl standard, although this is largely the result of setting the standard at a level reflecting analytic method quantitation limits.
\9\ Development of The MACT Floors For The Final NESHAP For Portland Cement. August 6, 2010.

Commenters presented virtually no quantified data that floor plants are unable to meet the standards. See National Association of Metal Finishers v. EPA, 719 F. 2d 624, 649 (3d Cir. 1983) (unquantified assertions are entitled to little if any weight). Rather, their comments (comment 2845 at Table 1, echoed by many other industry commenters) provided narrative descriptions purporting to demonstrate that floor plants would not be able to achieve the standards.\10\ In those instances where commenters provided actual data on these plants' performance, EPA took the information into account in developing the final standards. Indeed, EPA adjusted all of the standards based on actual data presented. However, EPA is not willing to act on pure supposition and conjecture regarding variability, particularly in the face of record information indicating that not only all floor plants but a number of additional plants are already meeting the promulgated standards.
\10\ For example, the commenter asserted, without providing support, that for the floor kilns the standards were ``achieved in practice, but not under foreseeable operations''; ``achieved in practice based on limited stack tests''; ``data shows that proposed standard was not achieved in practice when malfunction emission [sic] are included in compliance determination'' (although no such data were provided to EPA).

c. Mercury Limits for Kilns

i. Floor Determination. We proposed mercury emissions limits of 43 lb/million (MM) tons clinker for existing sources and 14 lb/MM tons clinker for new sources. The proposed floor was based on 30 days of data on all kiln inputs for 89 kilns. See 74 FR at 2114243. For all kilns but the five equipped with wet scrubbers, emissions were assumed to equal the total mass of mercury fed to each kiln. Scrubberequipped kilns were considered to emit all mercury minus an assumed amount representing the average performance of the wet scrubbers. For kilns that waste cement kiln dust (CKD), the mercury component of the CKD was subtracted from inputs to calculate emissions. Id. By conducting a total mass balance for mercury and then assuming that all mercury inputted is emitted (minus conservatively estimated removals for scrubber usage and dust wastage), EPA made a near worst case assumption as to kilns' mercury emission levels. The kilns were then ranked from best to worst based on the extrapolated mercury emissions, normalized to clinker production. EPA further proposed that no beyond the floor standard was appropriate for either existing or new sources. Id. at 21149.

Since proposal we received updated data on certain kilns' raw materials usage and mercury content \11\ and used that data to revise our average mercury emissions estimates from the best performing kilns at proposal.\12\ We have also revised upward the floor kilns' projected emissions based on their reasonably estimated intraquarry variability (explained further below). As a result, estimated emissions from these kilns increased, and one of the kilns in the group of sources used to set the existing source floor is no longer one of the best performing kilns. At proposal, the average mercury emissions of the top 12 percent of the kilns was 27.4 pounds per million (lb/MM) tons clinker, and the average emissions of the best performing source were 13.4 lb/MM ton clinker. After revising our mercury emissions estimates, the averages were 32 and 14 lb/MM tons clinker, respectively, as shown in Table 2. \11\ See Portland Cement Association Comments on the NESHAP Proposed Rule (Docket Number: EPAHQOAR20020051) (September 4, 2009) at pp. 3135.
\12\ Development of The MACT Floors For The Final NESHAP For Portland Cement, August 6, 2010.
Table 2Mercury MACT Floor
Mercury emissions Kiln code (lb/MM ton feed) 1589............................................... 8.48 1650............................................... 9.53 1315............................................... 15.26 1302............................................... 15.28 1248............................................... 16.63 1259............................................... 21.33 1286............................................... 22.65 1594............................................... 25.23 1435............................................... 25.51 1484............................................... 25.51 1364............................................... 25.91 MACTExisting Kilns
Average: lb/MM tons feed (lb/MM tons clinker)...... 19.21 (31.7) Total variance..................................... 272.3 [[Page 54977]]

FOR FURTHER INFORMATION CONTACT

Mr. Keith Barnett; Office of Air Quality Planning and Standards; Sector Policies and Programs Division, Metals and Minerals Group (D24302); Environmental Protection Agency; Research Triangle Park, NC 27711; telephone number: (919) 5415605; fax number: (919) 5415450; email address: barnett.keith@epa.gov.