[Federal Register Volume 69, Number 7 (Monday, January 12, 2004)]
[Rules and Regulations]
[Pages 1786-1821]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 04-5]



[[Page 1785]]

-----------------------------------------------------------------------

Part II





Environmental Protection Agency





-----------------------------------------------------------------------



40 CFR Part 60



Amendments to Standards of Performance for New Stationary Sources; 
Monitoring Requirements; Final Rule

  Federal Register / Vol. 69 , No. 7 / Monday, January 12, 2004 / Rules 
and Regulations  

[[Page 1786]]


-----------------------------------------------------------------------

ENVIRONMENTAL PROTECTION AGENCY

40 CFR Part 60

[OAR-2003-0009, FRL-7604-9]


Amendments to Standards of Performance for New Stationary 
Sources; Monitoring Requirements

AGENCY: Environmental Protection Agency (EPA).

ACTION: Final rule.

-----------------------------------------------------------------------

SUMMARY: This action promulgates Performance Specification 11 (PS-11): 
Specifications and Test Procedures for Particulate Matter Continuous 
Emission Monitoring Systems at Stationary Sources, and Procedure 2: 
Quality Assurance (QA) Requirements for Particulate Matter Continuous 
Emission Monitoring Systems at Stationary Sources. The PS-11 and QA 
Procedure 2 will apply to sources that are required under an applicable 
regulation to use particulate matter continuous emission monitoring 
systems (PM CEMS) to monitor PM continuously. The PS-11 and Procedure 2 
will help to ensure that PM CEMS are installed and operated properly 
and produce good quality monitoring data on an ongoing basis.

EFFECTIVE DATE: January 12, 2004.

ADDRESSES: Docket Nos. OAR-2003-0009 and A-2001-10 contain supporting 
information used in developing the final rule. The docket is located at 
the Air and Radiation Docket and Information Center in the EPA Docket 
Center, (EPA/DC), EPA West, Room B102, 1301 Constitution Avenue, NW., 
Washington, DC 20460, telephone (202) 566-1744.

FOR FURTHER INFORMATION CONTACT: Mr. Daniel G. Bivins, Emission 
Measurement Center (D205-02), Emissions, Monitoring, and Analysis 
Division, U. S. Environmental Protection Agency, Research Triangle 
Park, North Carolina 27711, telephone number (919) 541-5244, electronic 
mail address [email protected].

SUPPLEMENTARY INFORMATION:
    Regulated Entities. The final rule applies to any facility that is 
required to install and operate a PM CEMS under any provision of title 
40 of the Code of Federal Regulations (CFR). If you have any questions 
regarding the applicability of this action to a particular entity, 
consult the person listed in the preceding FOR FURTHER INFORMATION 
CONTACT section.
    Docket. The EPA has established an official public docket for this 
action including both Docket ID No. OAR-2003-0009 and Docket ID No. A-
2001-10. The official public docket consists of the documents 
specifically referenced in this action, any public comments received, 
and other information related to this action. All items may not be 
listed under both docket numbers, so interested parties should inspect 
both docket numbers to ensure that they have received all materials 
relevant to the final rule. Although a part of the official public 
docket, the public docket does not include Confidential Business 
Information or other information whose disclosure is restricted by 
statute. The official public docket is available for public viewing at 
the EPA Docket Center (Air Docket), EPA West, Room B-102, 1301 
Constitution Avenue, NW., Washington, DC. The EPA Docket Center Public 
Reading Room is open from 8:30 a.m. to 4:30 p.m., Monday through 
Friday, excluding legal holidays. The telephone number for the Reading 
Room is (202) 566-1744, and the telephone number for the Air Docket is 
(202) 566-1742.
    Electronic Access. Electronic versions of the documents filed under 
Docket No. OAR-2003-0009 are available through EPA's electronic public 
docket and comment system, EPA Dockets. You may use EPA Dockets at 
http://www.epa.gov/edocket/ to submit or view public comments, access 
the index of the contents of the official public docket, and access 
those documents in the public docket that are available electronically. 
Once in the system, select ``search'' and key in the appropriate docket 
identification number.
    The EPA's policy is that copyrighted material will not be placed in 
EPA's electronic public docket but will be available only in printed, 
paper form in the official public docket. Although not all docket 
materials may be available electronically, you may still access any of 
the publicly available docket materials through the docket facility 
identified in this document.
    Worldwide Web (WWW). In addition to being available in the docket, 
an electronic copy of today's document also will be available on the 
WWW. Following the Administrator's signature, a copy of this action 
will be posted at http://www.epa.gov/ttn/oarpg on EPA's Technology 
Transfer Network (TTN) policy and guidance page for newly proposed or 
promulgated rules. The TTN provides information and technology exchange 
in various areas of air pollution control. If more information 
regarding the TTN is needed, call the TTN HELP line at (919) 541-5384.
    Judicial Review. Under section 307(b)(1) of the Clean Air Act 
(CAA), judicial review of the final rule is available only by filing a 
petition for review in the U.S. Court of Appeals for the District of 
Columbia Circuit by March 12, 2004. Under section 307(d)(7)(B) of the 
CAA, only an objection to the final rule that was raised with 
reasonable specificity during the period for public comment can be 
raised during judicial review. Moreover, under section 307(b)(2) of the 
CAA, the requirements established by the final rule may not be 
challenged separately in any civil or criminal proceedings brought by 
EPA to enforce these requirements.
    Outline. The information presented in this preamble is organized as 
follows:

I. Introduction
II. Summary of Major Changes Since Proposal
    A. Changes to PS-11
    B. Changes to Quality Assurance (QA) Procedure 2
III. Summary of Responses to Major Comments
    A. General
    B. Performance and Applicability of PM CEMS
    C. Instrument Selection
    D. Isokinetic Sampling
    E. Condensible PM
    F. Instrument Location
    G. Shakedown and Correlation Test Planning Period (CTPP)
    H. Correlation Testing
    I. Response Range
    J. Reference Method Testing
    K. Statistical Methods
    L. Statistical Criteria
    M. Routine Performance Checks
    N. Auditing Requirements
    O. Extrapolation of Correlation
    P. Requirements for Other Types of Monitors
IV. Summary of Impacts
    A. What are the impacts of PS-11 and QA Procedure 2?
    V. Statutory and Executive Order Reviews
    A. Executive Order 12866: Regulatory Planning and Review
    B. Paperwork Reduction Act
    C. Regulatory Flexibility Act
    D. Unfunded Mandates Reform Act
    E. Executive Order 13132: Federalism
    F. Executive Order 13175: Consultation and Coordination with 
Indian Tribal Governments
    G. Executive Order 13045: Protection of Children from 
Environmental Health Risks and Safety Risks
    H. Executive Order 13211: Actions Concerning Regulations that 
Significantly Affect Energy Supply, Distribution, or Use
    I. National Technology Transfer and Advancement Act
    J. Congressional Review Act

I. Introduction

    The PS-11, Specifications and Test Procedures for Particulate 
Matter Continuous Emission Monitoring Systems at Stationary Sources, 
and

[[Page 1787]]

Procedure 2, Quality Assurance Requirements for Particulate Matter 
Continuous Emission Monitoring Systems at Stationary Sources, were 
first published in the Federal Register on April 19, 1996 (61 FR 17358) 
as part of the proposed Hazardous Waste Combustion MACT standard. The 
PS-11 and Procedure 2 were published again on December 30, 1997 (62 FR 
67788) for public comment on revisions made to these procedures. Since 
then, we have continued to learn about the capabilities and performance 
of PM CEMS through performing and witnessing field evaluations and 
through discussions with our European counterparts.
    Additional experience with the procedures of PS-11 and Procedure 2 
led us to propose these further revisions, which were published on 
December 12, 2001 (66 FR 64176). Today's final rule builds upon that 
proposal and reflects the changes we have made to PS-11 and Procedure 2 
in response to the additional comments we received on the December 2001 
proposal.

II. Summary of Major Changes Since Proposal

A. Changes to PS-11

1. Instrument Selection
    Several changes were made to the requirements of PS-11 regarding 
the selection of instruments. Sections 4.2 and 6.1(1) of the proposed 
PS-11 required owners and operators of affected sources using 
extractive PM CEMS to heat the extracted samples of the exhaust gas to 
the same temperature specified by the reference method. In the final 
PS-11, we are changing this requirement to a recommendation. In Section 
4.3, we also changed from a requirement to a recommendation that owners 
and operators use a measurement technology that is free from 
interferences. In that same section, we deleted the phrase regarding 
duct flue gas conditions.
    We are no longer requiring in Section 6.1(3) that extractive PM 
CEMS used on sources with varying volumetric flow rates maintain 
isokinetic sampling. We still recommend isokinetic sampling in such 
installations. Furthermore, we changed Section 6.1(3) to allow owners 
and operators of extractive PM CEMS in applications with varying flow 
rates to use data from similar facilities to demonstrate that 
isokinetic sampling is unnecessary. In the proposed PS-11, data from 
similar facilities could not be used; only site-specific data could be 
used for such demonstrations.
    Several changes were made to Section 8.1 of PS-11 regarding 
instrument selection. In the proposed PS-11, Section 8.1 stated that 
owners or operators must select a PM CEMS that is most appropriate for 
the source, considering the source operating conditions. We have 
revised the rule to state that owners or operators should select an 
appropriate PM CEMS for the source. This change also is reflected in 
Sections 2.4(1) and 6.0 of the final rule. We changed from a 
requirement to a recommendation in Section 8.1(1)(ii) that extractive 
PM CEMS sample at the reference method filter temperature. We also 
changed from a requirement to a recommendation in Section 8.1(5) that 
owners or operators consult with instrument vendors to obtain basic 
recommendations on instrument capabilities and installation.
2. Instrument Location
    With respect to stratification, Section 2.4(2) of the proposed PS-
11 recommended performing a PM profile test if PM stratification was 
likely to be a problem. In addition, owners or operators would have 
been required to relocate the PM CEMS or eliminate stratification if 
the stratification varies by more than 10 percent. In the final PS-11, 
we have eliminated the reference to profile testing and the requirement 
for either relocating the CEMS or resolving the stratification issue. 
We also have deleted the requirement from Section 8.2(2) that owners or 
operators relocate the CEMS if failure to meet the correlation criteria 
is due to a location problem that cannot be corrected.
3. Pretest Preparations
    In Section 8.4 of the proposed PS-11, owners and operators of PM 
CEMS would have been required to conduct a shakedown period and a 
correlation test planning period (CTPP) prior to correlation testing. 
Although we continue to recommend that you conduct shakedowns and 
CTPPs, the final PS-11 does not require them. Instead of a formal 
shakedown period, the final rule recommends that owners and operators 
familiarize themselves with the operation of the CEMS prior to 
correlation testing. The elimination of shakedown periods also is 
reflected in Section 2.4(5) of the final rule, and the requirement 
regarding interruption of shakedown periods, specified in Section 
8.4(1)(ii) of the proposed rule, has been deleted.
    Section 8.4(1)(i) of the proposed PS-11 required owners or 
operators to conduct daily drift checks during the shakedown period. In 
the final rule, daily drift checks are recommended rather than required 
during the pretest preparation period when owners and operators 
familiarize themselves with the operation of the CEMS.
    With the elimination of CTPPs as a required pretest activity, we 
have deleted certain requirements that applied specifically to the 
CTPP. For example, we deleted the requirement to produce permanent 
records of 15-minute average PM CEMS responses that would have been 
required in Section 8.4(2) of the proposed PS-11, as well as the 
requirements in Sections 8.4(2)(ii) and (iii) of the proposed rule that 
data recorders record PM CEMS responses during the full range of 
routine process operating conditions and that owners or operators 
establish the relationship between operating conditions and PM CEMS 
response. We also have deleted the requirement in Section 8.4(3) of the 
proposed PS-11 that owners or operators set the response range of the 
PM CEMS so that the highest observed response is within 50 to 60 
percent of the maximum output. Instead, the final PS-11 requires owners 
and operators to set the response range to whatever range is 
appropriate to ensure that the instrument will record the full range of 
responses likely during the correlation test. We also have revised 
Section 2.2(2) of the final rule to reflect this change.
    The proposed PS-11 required owners or operators to perform a 7-day 
drift check at the end of the CTPP. Although we have eliminated the 
requirement for CTPPs, the final PS-11 still requires owners or 
operators to successfully complete a 7-day drift test prior to 
correlation testing. We have also revised Section 8.5(1), which 
explains the purpose of the 7-day drift test.
4. Correlation Testing
    Sections 2.2(2), 2.4(7), and 6.3 of the proposed PS-11 required 
correlation testing over the range of emissions established during the 
CTPP. Because PS-11 no longer requires CTPPs, we revised these sections 
to require correlation testing over the full range of normal process 
and control device operating conditions. We also deleted the 
requirement in Section 8.6 to conduct correlation testing while the 
source is operating as it did during the CTPP.
    Sections 2.4(7) and 8.6(1)(i) of the proposed PS-11 would have 
required paired sampling trains during all correlation tests. Although 
we highly recommend paired sampling trains, PS-11 no longer requires 
correlation tests to be performed using paired trains. We also have 
deleted from Sections 2.4(7) and 8.6(1)(ii) the requirement that data 
pairs meet certain criteria for precision and bias, because those 
criteria would

[[Page 1788]]

apply specifically to paired data, and we are no longer requiring 
paired trains. We plan to address data precision and bias in guidance 
materials at a later date.
    Sections 8.2(4) and 8.4(4) of the proposed PS-11 suggested using a 
bypass as a means of increasing PM emissions during correlation 
testing. In the final PS-11, we have eliminated any reference to 
bypassing control devices for this purpose. However, we have included 
PM spiking as an option for increasing PM emissions during correlation 
tests. We also have revised Section 8.6(5) to clarify how owners or 
operators should obtain zero point data during correlation tests.
    Finally, we have changed the requirements in Section 8.6(3) 
regarding the selection of test runs for developing the correlation. In 
the proposed PS-11, owners or operators could reject the results of 
test runs only if the basis for rejecting the data was specified in the 
reference method, PS-11, QA Procedure 2, or in the facility's QA plan. 
In the final PS-11, up to five test runs can be rejected without an 
explanation for the rejection, provided that the results of at least 15 
valid test runs are used to develop the correlation. If more than five 
test runs are rejected, the basis for rejecting those additional runs 
(i.e., those in addition to the first five rejected runs) must be 
reported.
5. Extrapolation of Correlation
    Section 8.8(1) of the proposed PS-11 addressed the limits for 
extrapolating the correlation equation before additional correlation 
testing would be required. The maximum allowable extrapolation under 
the proposed rule would have been 125 percent of the highest PM CEMS 
response used to develop the correlation curve. If that 125 percent 
limit was exceeded for three consecutive hours, three additional 
correlation tests runs would have been required. We have changed the 
time period that triggers this additional correlation testing. In the 
final PS-11, additional correlation testing is required only after the 
125 percent value has been exceeded for 24 consecutive hours, or a 
period of cumulative hours that exceeds 5 percent of the total valid 
operating hours for the previous 30 days, whichever occurs first. In 
addition, we have clarified in Section 8.8(1) of the final PS-11 that 
additional testing is required only when the 125 percent limit is 
exceeded while the source and control device are operating under normal 
conditions. In any case, Section 8.8(3) of the final PS-11 requires 
owners and operators to report the reason why the 125 extrapolation 
limit was exceeded.
    We have revised PS-11 to include a special provision for low 
emitting-sources that emit no more than 50 percent of the emission 
limit. For such cases, Section 8.8(4) of the final PS-11 allows 
extrapolation up to the response value that corresponds to 50 percent 
of the emission limit or 125 percent of the highest PM CEMS response 
used to develop the correlation curve, whichever is greater. Finally, 
in the event additional correlation testing is required, we have 
revised Section 8.8(2)(i) of the final PS-11 to extend the deadline for 
completing the testing and developing a new correlation equation from 
30 to 60 days.
6. Statistical Methods and Criteria
    In Section 12.3 of the final PS-11, we have clarified that, if 
paired testing is performed, paired reference method data should not be 
averaged, but should be treated individually in developing the 
correlation. In such cases, at least 15 sets of reference method and PM 
CEMS response data are still required, although for each PM CEMS 
response there will be two reference method data points, one for each 
of the two paired sampling trains.
    We also have reorganized and made several other changes to Section 
12.3. In the proposed PS-11, three types of correlation models were 
addressed: linear, polynomial, and logarithmic. The final rule 
specifies procedures for evaluating five types of correlation models; 
in addition to the linear, polynomial, and logarithmic models, we have 
added procedures for evaluating exponential and power correlation 
models. We also have made changes regarding the calculations needed for 
evaluating correlation equations. In the proposed PS-11, equations were 
presented for calculating confidence and tolerance intervals. For 
example, Equation 11-11 of the proposed rule defined the confidence 
interval in terms of the quantity [ycirc] +/- CI, where [ycirc] is the 
predicted PM concentration, and CI is the confidence interval half 
range. However, the confidence interval performance criterion was 
presented in terms of the confidence interval half range as a 
percentage of the emission limit and not in terms of the confidence 
interval itself. Consequently, we have eliminated the requirement to 
calculate confidence intervals. For the same reason, we eliminated the 
requirement to calculate tolerance intervals. In the final rule, owners 
or operators of affected PM CEMS must calculate the confidence interval 
half range and tolerance interval half range, but are not required to 
calculate the confidence and tolerance intervals.
    We also have changed the PM CEMS response values at which the 
confidence and tolerance interval half ranges are calculated. In the 
proposed PS-11, owners or operators would have been required to 
calculate the confidence and tolerance interval half ranges at the 
median PM CEMS response (x) values. The preamble to the proposed rule 
mistakenly indicated that the confidence and tolerance interval half 
ranges are smallest at the median x value. However, that statement is 
correct only for exponential and power correlations. In the final PS-
11, the x value for calculating confidence and tolerance interval half 
ranges depends on the type of correlation. For linear correlations, the 
confidence and tolerance interval half ranges must be calculated at the 
mean x value. The confidence and tolerance interval half ranges for 
polynomial correlations must be calculated at the x value that 
corresponds to the minimum value of the variable delta ([Delta]), which 
is defined by Equation 11-25 of the final PS-11. For logarithmic 
correlations, the confidence and tolerance interval percentages must be 
calculated at the mean of the log-transformed x values. For exponential 
and power correlations, the confidence and tolerance interval 
percentages must be calculated at the median x and log-transformed x 
values, respectively. These x values represent the points at which the 
confidence and tolerance intervals are smallest or narrowest. We also 
have reflected these changes in Section 2.3 of the final PS-11, which 
specifies general correlation data handling requirements, and in 
Section 13.2, which specifies the performance criteria for confidence 
and tolerance intervals. In addition, we have added a new section 12.4 
to the final PS-11 to specify procedures for selecting the best 
correlation model.
    We deleted the example correlation calculations presented in 
Section 18.0 of the proposed PS-11. We will provide example 
calculations for all five correlation models in the next revision to 
Current Knowledge of Particulate Matter (PM) Continuous Emission 
Monitoring, EPA-454/R-00-039 (PM CEMS Knowledge Document), which will 
be revised periodically to incorporate additional guidance, example 
calculations, and other information that will help in understanding and 
complying with PS-11 and QA Procedure 2.
    Finally, we have included in Section 13.2 a provision for low-
emitting sources to meet a lower correlation coefficient. In the final 
rule, a low-emitting source must meet a minimum correlation coefficient 
of 0.75 rather

[[Page 1789]]

than the 0.85 value required for sources that are not low-emitting.
7. Other Changes
    Section 2.4(4) of the proposed PS-11 addressed recordkeeping 
requirements for PM CEMS maintenance and performance data. We have 
deleted this section in the final PS-11 because recordkeeping 
requirements are already addressed, in detail, in the general 
provisions to parts 60, 61, and 63, and in most, if not all, applicable 
rules.

B. Changes to Quality Assurance (QA) Procedure 2

1. Precision and Bias
    Sections 10.1(3) and (4) of the proposed QA Procedure 2 specified 
precision and bias requirements for paired reference method sampling 
trains. Because the final PS-11 does not require paired sampling 
trains, we have removed the precision and bias criteria from QA 
Procedure 2. For the same reason, we also have deleted Section 12.0(5), 
which addressed relative standard deviation, the parameter for 
assessing paired data precision.
2. Quality Control (QC) Program
    Section 9 of QA Procedure 2 addresses QC measures. We have added 
Section 9.0(8) to the final rule to require owners and operators to 
include in their QC programs written procedures for checking extractive 
duct systems for material accumulation when extractive PM CEMS are 
used.
3. System Checks and Audits
    We made several changes to Section 10.3 of QA Procedure 2 regarding 
periodic audits. To ensure consistency in the organization of the 
section, we renumbered some of the paragraphs. We changed the required 
frequency of relative response audits (RRAs) from once every four 
quarters to the frequency specified in the applicable rule. In 
addition, we clarified that an RRA can be substituted for an absolute 
accuracy audit (ACA) during any quarter. Likewise, we clarified that a 
response correlation audit (RCA) can be substituted for an ACA or an 
RRA to satisfy the required auditing frequency. In Section 10.3(2)(iii) 
of the final QA Procedure 2, we deleted the requirement that owners and 
operators obtain audit samples from instrument manufacturers or 
vendors.
    We made two changes to the acceptance criteria for RCAs. In Section 
10.3(5)(ii) of the proposed QA Procedure 2, we required all 12 of the 
PM CEMS responses to fall within the range of PM CEMS responses used to 
develop the initial correlation. In the final QA Procedure 2, we 
relaxed this requirement somewhat. We still require all 12 PM CEMS 
responses to be no greater than the highest response used to develop 
the correlation curve. However, in Section 10.4(5) of the final rule, 
we allow three of the PM CEMS responses to fall below the range of 
responses used to develop the initial correlation curve. We made a 
similar change to the acceptance criterion for RRAs. In Section 10.4(6) 
of the final rule, the three PM CEMS responses for the RRA must be no 
greater than the highest PM CEMS response used to develop the initial 
correlation, but one of the three points may fall below that range of 
responses used to develop the initial correlation.
    Finally, we changed Equation 2-4 of Section 12.0(4), which is used 
to determine sample volume audit accuracy. In the proposed QA Procedure 
2, we changed the denominator of Equation 2-4 from the sample gas 
volume measured by the independent calibrated reference device to the 
full scale value.

III. Summary of Responses to Major Comments

A. General

    Comment: One commenter stated that EPA's fundamental approach for 
PM CEMS is too complex and costly. The commenter noted that the 
requirements for PM CEMS place too much emphasis on reporting emissions 
in units directly comparable to the emission standard. According to the 
commenter, this approach results in a ``research-and-development 
effort.'' He noted that EPA's objective should be to establish a 
process whereby the owner or operator develops an understanding of how 
PM CEMS operate and the relationship among instrument response, process 
and control device operating parameters, and emissions. At that point, 
the owner/operator can use that information to reduce PM emissions. As 
proposed, PS-11 and QA Procedure 2 require such an understanding (by 
means of the shakedown and correlation test planning period) as a 
precursor to establishing a stringent statistical correlation between 
PM CEMS response and emissions. The commenter believes that the 
approach should be to use PM CEMS as a relative indicator of emissions 
rather than to attempt to achieve a precise correlation between PM 
emissions and PM CEMS response over the entire range of source 
operations.
    Response: The purpose of PM CEMS is to quantify PM emissions as 
accurately and precisely as possible to ensure compliance with the 
applicable PM emission limits. To meet this objective, we must 
incorporate into PS-11 and QA Procedure 2 procedures for ensuring that 
PM CEMS are installed, operated, and maintained properly. Although this 
necessitates complexity, we have taken steps to minimize the complexity 
of PS-11. In the final PS-11, we have simplified or eliminated several 
of the requirements specified in the proposed rule regarding instrument 
selection and location, correlation test preparation, and correlation 
test procedures. We also have reorganized and simplified the 
statistical procedures for developing the correlation equation, as well 
as incorporating additional flexibility into the types of correlation 
models that can be developed. We have published guidance on the 
selection and use of PM CEMS in the PM CEMS Knowledge Document, which 
may be revised periodically to incorporate additional guidance, example 
calculations, and other information that will help in understanding and 
complying with PS-11 and QA Procedure 2.
    With respect to cost, we believe that the cost of installing and 
operating a PM CEMS is relative to the application, and some 
applications will be more costly than others. However, we account for 
the costs of any required monitoring systems, such as PM CEMS, when we 
evaluate the compliance costs for a specific rulemaking that requires 
those monitoring systems.
    Finally, we would like to point out that PS-11 and QA Procedure 2 
do not specify the compliance scenario. Although this rulemaking is 
intended to apply to the monitoring of PM emission limits for 
compliance purposes, we recognize the advantages of using PM CEMS as an 
indicator of compliance for sources subject to 40 CFR 64 (Compliance 
Assurance Monitoring Rule) and other applications. Neither PS-11 nor QA 
Procedure 2 prohibit the use of PM CEMS as indicators of control device 
operation or emission levels. Furthermore, an owner or operator would 
not necessarily have to comply with PS-11 or QA Procedure 2 in a case 
where a PM CEMS is used as an indicator of control device performance 
or emissions.
    Comment: One commenter stated that the requirements of PS-11 and QA 
Procedure 2 focus primarily on establishing enforcement opportunities 
by holding owners and operators responsible for factors that are beyond 
their control. To support this contention, the commenter referenced 
Section 8.1 of PS-11, which requires owners/operators to select a PM 
CEMS

[[Page 1790]]

``* * * that is most appropriate for your source.'' The commenter 
believes that, for a specific source, the most appropriate instrument 
may not be known until after one or more instruments have been selected 
and placed into operation. The commenter also cited Section 2.3 of PS-
11, which addresses situations in which multiple correlations may be 
required. The commenter noted that, in both of these examples, the 
enforcement action would not depend on whether the control device is 
operating properly or emissions are exceeded. Instead, the enforcement 
action focuses on the type of instrument selected and the variability 
of emissions (which would require multiple correlations).
    Response: We agree that some enforcement actions associated with 
PS-11 may not necessarily depend on control device operation or 
emission levels. However, in this respect, PS-11 is similar to other 
performance specifications, such as PS-1, which specify the 
requirements that monitoring systems must meet. Individually, some of 
those requirements may not be directly related to the operation of a 
control device or emission levels, but, as a whole, the requirements 
help to ensure the proper operation of the monitoring system and the 
quality of the data generated by the monitoring system.
    With respect to the requirement of the proposed Section 8.1 of PS-
11 cited by the commenter, we have revised that section to state that 
owners and operators ``* * * should select a PM CEMS that is 
appropriate. * * * '' We believe this revised language allows for more 
flexibility in instrument selection. Although there may still be some 
trial and error involved in selecting an instrument, there are several 
PM CEMS technologies available, and some instruments clearly are more 
appropriate than others for certain applications.
    The requirement of Section 2.3 of the proposed rule regarding 
multiple correlations is meant to address sources with different 
operating modes that result from variations in operating parameters 
such as process load, charge rates, or feed materials. In such cases, 
there may be significant differences in PM emissions characteristics 
for the different source operating modes to the extent that a single 
correlation cannot satisfy all of the criteria specified in PS-11. We 
also would like to point out that PS-11 allows for, but does not 
require, multiple correlations. In the event that multiple correlations 
are needed, Section 2.3 simply requires that sufficient data be 
collected. By allowing multiple correlations under such a scenario, PS-
11 provides the owner or operator flexibility in complying with the 
rule. Therefore, we disagree with the comment that Section 2.3 simply 
focuses on establishing enforcement opportunities.
    Comment: One commenter observed that several requirements in PS-11 
and QA Procedure 2 require adherence to manufacturer's recommendations. 
He stated that those recommendations may conflict with regulatory 
requirements or good engineering practice. He believes that following 
manufacturer's recommendations cannot be a requirement unless EPA 
reviews and approves those recommendations. He noted that, regardless 
of how well EPA may understand the procedures currently recommended by 
existing manufacturers, new manufacturers can enter the market at any 
time, and they are not subject to regulation by EPA.
    Response: We agree with the commenter and have eliminated those 
specific requirements that owners and operators follow the 
recommendations of the instrument manufacturer or vendor. We believe 
that it is prudent to consider those recommendations, but owners or 
operators of affected sources must determine what is most appropriate 
for their specific installation.

B. Performance and Applicability of PM CEMS

    Comment: Four commenters commented that EPA has not demonstrated 
that PM CEMS can meet PS-11 and QA Procedure 2 on a consistent basis. 
They noted that sources, such as cement kilns, with low to moderate 
condensible PM will have particular difficulty complying with the rule. 
In addition, they commented that the basis for EPA's conclusion on the 
suitability of PM CEMS is largely from demonstrations and tests 
performed on hazardous waste combustors, which are characterized by wet 
control systems and exhaust temperatures below the temperature range 
within which most condensible matter nucleates. Consequently, those 
tests are not representative of cement kilns or other sources for which 
condensible PM is a significant concern. They also noted that 
condensible PM emissions for the cement industry are dependent on raw 
materials and are highly variable, making it less likely that 
correlation relationships will remain stable for cement kilns. The 
commenters suggested that EPA continue specifying opacity monitors as 
the technology for demonstrating compliance with PM emission limits.
    Response: Based on the results of the field studies, PS-11 and QA 
Procedure 2 have been modified to account for performance issues 
discovered during the field studies. For example, regarding the issue 
of condensible PM, the proposed rule eliminated the requirement for 
correlation testing using only EPA Method 5I. Instead, PS-11 now 
specifies that the correlation test be conducted using the same 
reference method required by the applicable rule, thereby minimizing 
the effects condensible PM could have on PM concentrations when one 
method is used to demonstrate compliance and a different method is use 
to develop the PM CEMS correlation. To further address concerns with 
characterizing exhaust streams that contain condensible PM, we also 
have included in PS-11 the recommendation that the PM CEMS be 
maintained at the reference method filter temperature. We made this 
recommendation because PM CEMS that measure samples at conditions that 
are different than the sampling conditions specified in the reference 
method may not correlate well with reference method data. Maintaining 
the measurement conditions of the PM CEMS at the reference method 
filter temperature eliminates one of the factors that can adversely 
impact the correlation between PM CEMS responses and reference method 
measurements.
    Although we did rely on field demonstrations on hazardous waste 
combustors to develop the requirements of PS-11, we believe that the PM 
CEMS field demonstrations completed to date encompass a range of 
operating conditions and emission characteristics that extend beyond 
those typical of the hazardous waste combustion industry. We also have 
provided guidance on the selection and applicability of PM CEMS. We do 
not rule out the possibility that PM CEMS may not be appropriate for 
certain source operating conditions or emission characteristics. 
However, the purpose of PS-11 and QA Procedure 2 is not to define the 
applicability of PM CEMS, but to establish basic requirements that will 
help to ensure that PM CEMS produce high-quality data on a consistent 
basis. The applicability of PM CEMS to specific sources and source 
categories must be established under the applicable rule, and it may be 
necessary to incorporate industry-specific criteria in rules that 
require the use of PM CEMS for compliance monitoring.
    Regarding the use of opacity monitors for demonstrating compliance 
with PM emission limits, we believe that opacity monitors are reliable 
indicators of

[[Page 1791]]

compliance with opacity limits and we will continue to require 
continuous opacity monitoring systems for certain rules that establish 
opacity limits. However, for rules that establish PM emission limits, 
we believe that PM CEMS are the appropriate technology for compliance 
monitoring.
    Comment: One commenter remarked that using PM CEMS has not been 
demonstrated to be a technically sound compliance method and suggested 
additional field testing be performed before PM CEMS are required in a 
rulemaking. Another commenter stated that PM CEMS should not be used as 
a compliance tool until there is a better understanding of their 
operation and limitations. A third commenter stated that EPA's 
evaluations do not support EPA's conclusions regarding the reliability 
of PM CEMS. The commenter noted that the performance of PM CEMS is 
mixed, at best, and instrument operation and calibration is a difficult 
and time-consuming task. The same commenter stated that PM CEMS are not 
appropriate compliance monitors because, unlike other CEMS, PM CEMS do 
not provide a direct measurement of the target pollutant (i.e., PM). 
The commenter also remarked that the fact that PM CEMS require a 
shakedown period is further indication that PM CEMS are not acceptable 
for compliance demonstrations. The commenter noted that shakedowns and 
CTPPs are not required for other types of CEMS, such as continuous 
nitrogen oxide (NOX) and sulfur dioxide (SO2) 
monitors.
    Response: We acknowledge that problems have been encountered in our 
field studies of PM CEMS. However, we have used the results of those 
field studies to modify PS-11 and QA Procedure 2 to account for the 
performance issues observed during the studies. For example, we have 
made changes that apply to sources characterized by condensible PM and 
incorporated procedures for developing other types of correlation 
models not previously addressed in PS-11. We agree with the comment 
that developing the correlation can be complex and time-consuming. With 
regard to the acceptability of PM CEMS for compliance determinations, 
the purpose of PS-11 is not to specify how compliance with an 
applicable emission limit is to be determined; the purpose of PS-11 is 
to specify procedures for obtaining the best correlation for using a PM 
CEMS to characterize PM emissions, and to ensure that PM CEMS are 
installed and operated properly. The applicability of PM CEMS for 
determining compliance with an emission limit, as well as the 
procedures for determining compliance using PM CEMS, must be specified 
by the applicable rule.
    We disagree with the commenter that the proposed requirements for a 
shakedown and CTPP are an indication that PM CEMS are unreliable or 
inappropriate as a compliance monitor. We proposed requiring a 
shakedown and CTTP because PM CEMS are a relatively new technology for 
many industries, and many operators are unfamiliar with their 
operation. In such cases, a shakedown and CTPP allows time for the 
operator and other personnel to become familiar with the operation of 
the instrument and to facilitate the correlation test. Although we 
still recommend that facilities conduct a shakedown and/or CTPP, we 
have eliminated these periods as requirements in PS-11.
    Comment: Two commenters stated that PM CEMS technology is not ready 
for use by hazardous waste combustors to demonstrate compliance with PM 
emission standards. One of the commenters stated that PM CEMS installed 
on hazardous waste combustors will result in additional automatic waste 
feed cutoffs that are unrelated to the stability of the combustion 
process. The other commenter pointed out the difficulties with the PM 
CEMS that were tested at the EPA-sponsored field study in Battleboro, 
North Carolina; he believes that PM CEMS used to monitor emissions from 
commercial incinerators would have even more difficulty because of the 
greater variability in feedstocks when compared to the coal-fired 
boiler that was tested at Battleboro.
    Response: We disagree with the commenters that PM CEMS technology 
are unsuitable for use as compliance monitors for the hazardous waste 
combustor industry. The DuPont Field Study demonstrated the effective 
use of several PM CEMS instruments on a hazardous waste combustor. A 
more recent study at the Department of Energy facility in Oak Ridge, 
Tennessee, provides another successful demonstration of a PM CEMS on a 
hazardous waste incinerator.
    We acknowledge that there were some difficulties with the PM CEMS 
that were tested during the Battleboro Field Study. However, those 
difficulties were primarily the result of the sampling location rather 
than variations in emission characteristics or the reliability of the 
PM CEMS instruments tested.
    Comment: One commenter commented that PM CEMS should not be 
required for facilities with low PM levels. He noted that the objective 
of protecting human health and the environment can be better achieved 
by controlling key operating parameters; installing and maintaining a 
PM CEMS on a well-designed and well-operated incinerator would be 
costly and difficult without actually reducing emissions. The commenter 
suggested allowing facilities to test at worst-case conditions and not 
requiring PM CEMS if the source operates consistently at some fraction 
of the emission standard (e.g., 40 percent).
    Response: The purpose of PS-11 and QA Procedure 2 is not to define 
the applicability of PM CEMS, but to establish basic requirements that 
will help to ensure that PM CEMS produce high-quality data on a 
consistent basis. The applicability of PM CEMS to specific sources and 
source categories must be established under the applicable rule. 
Therefore, we do not believe it is appropriate to specify in PS-11 the 
types of sources to which PS-11 should apply. However, we agree with 
the commenter that some provisions should be included in PS-11 for low-
emitting sources because less accuracy and precision are needed in such 
applications. To this end, we have incorporated into the final rule a 
provision for allowing a greater extrapolation of the correlation curve 
and a lower correlation coefficient for sources that emit no more than 
50 percent of the emission limit.
    Comment: One commenter concluded that PM CEMS are not suitable for 
determining compliance, but instead should be used as an indicator of 
compliance. To support this conclusion, he pointed to the results of 
Battleboro Field Study. He noted that, after having met the criteria 
for the initial correlation, all three instruments that were tested 
failed to meet the RCA criteria specified in QA Procedure 2. When a 
second RCA was performed, all three instruments again failed to meet 
the QA Procedure 2 criteria. The commenter also stated that the 
Battleboro results demonstrated that different PM CEMS calibrated at 
the same time using the same reference method gave different results. 
The responses for the two light-scattering instruments tracked each 
other well and gave similar results. However, when the results for the 
beta gauge instrument were compared to the light-scattering instrument 
results, more scatter was seen, indicating differences in how the two 
types of instruments respond to varying particle size and/or sampling 
location. One instrument could show a source to be in compliance, while 
another PM CEMS sampling the same exhaust stream could show the same

[[Page 1792]]

source to be out of compliance. Consequently, the commenter suggested 
that PM CEMS be used as an indicator of compliance rather than as a 
compliance monitor. He believes that correlation tests should not be 
required when a source operates below 40 percent of the emission limit 
and below the emission limit minus 10 mg/dscm. Instead, correlation 
tests should be optional, provided emission levels remain below these 
two levels (i.e., no more than 40 percent of the emission limit and at 
least 10 mg/dscm below the emission limit). If testing is performed, 
three runs should be adequate. Furthermore, a straight linear 
relationship should be used to estimate emissions. The relationship 
would be defined by the line from zero to the average of the three test 
runs. Additional correlation test runs should be required only if 
sustained emission levels exceed the level of the emission limit minus 
10 mg/dscm. If additional tests are performed, three runs should be 
adequate.
    Response: We believe that the problems encountered in the 
Battleboro Field Study regarding the failure of the instruments to meet 
the RCA criteria were due to the sampling location and the resulting 
stratification of the exhaust stream. Other field studies have 
demonstrated that PM CEMS can meet the RCA criteria when the sampling 
location is not a problem. We believe that the differences in the 
responses of the light-scattering and beta gauge instruments can be 
expected, given that light-scattering and beta gauge instruments 
operate on different physical principles. For a specific application, 
the correlation equation developed for each instrument takes into 
account these differences.
    Regarding the use of PM CEMS data as indicators, PS-11 and QA 
Procedure 2 do not prohibit the use of PM CEMS as indicators of control 
device performance or emission levels to satisfy the requirements of 
part 64. In such applications, the owner or operator of an affected 
source can propose the approach for selecting the appropriate indicator 
range that would trigger corrective action and reporting.
    Finally, although we do not agree with the commenter's specific 
suggestions regarding low-emitting sources, we have incorporated into 
the final rule provisions for low-emitting sources. Specifically, the 
final PS-11 allows for a lower correlation coefficient criterion and a 
larger allowable extrapolation range for PM CEMS responses for sources 
that emit relatively low levels of PM.

C. Instrument Selection

    Comment: Four commenters stated that Sections 4.2 and 6.1(1) of PS-
11 require that PM CEMS installed downstream of a wet air pollution 
control device be equipped with heated sample extraction lines. 
However, the commenters noted that EPA has not demonstrated that 
instruments so equipped can meet the requirements of PS-11 and QA 
Procedure 2.
    Response: Although we continue to believe that heated sample 
extraction lines are recommended in such installations, we have decided 
to eliminate this requirement from PS-11. We have no reason to believe 
that heated sample lines would prevent PM CEMS from meeting the 
requirements of PS-11 and QA Procedure 2. However, we also recognize 
that owners and operators are ultimately responsible for compliance and 
should have flexibility in determining an appropriate instrument and 
configuration for their specific application.
    Comment: One commenter pointed out that Section 8.1(1) requires 
selection of a PM CEMS that is appropriate for the PM characteristics 
and flue gas conditions at the source, but does not specify how owners 
or operators of the source are to determine which monitor is acceptable 
for their site-specific conditions. The commenter indicated that there 
are no EPA-approved tests for determining if PM characteristics are 
variable. The commenter also knew of no PM CEMS vendors who would 
acknowledge that their instrument was appropriate for variable PM 
characteristics or who would guarantee the performance of their 
instrument in applications with variable PM characteristics. In 
reference to this same requirement, four other commenters stated EPA 
has not demonstrated that there are appropriate PM CEMS for sources 
with routine variations in particle size distribution. As a result, 
industry must conduct instrument-oriented research to find the 
appropriate monitor. One commenter also remarked that there might not 
be an instrument available that ``responds appropriately'' to the flue 
gas conditions for a specific source.
    Response: In response to this concern, we have decided to change 
the wording of this section of PS-11 from a requirement to a 
recommendation that owners and operators select a PM CEMS that is 
appropriate for the source and emission characteristics. As mentioned 
previously, guidance on instrument selection can be found in the PM 
CEMS Knowledge Document. We believe that document can be a valuable 
tool in selecting an appropriate PM CEMS technology for a specific type 
of source. As we become aware of additional information that will help 
in selecting the appropriate PM CEMS technology, we plan to update the 
guidance accordingly.

D. Isokinetic Sampling

    Comment: Four commenters stated that, by requiring extractive 
instruments to sample isokinetically, PS-11 would preclude the use of 
several instruments that sample superisokinetically. Designing an 
instrument to sample superisokinetically enables the instrument to 
handle larger changes in flow rate without having to adjust 
continuously to maintain isokinetic sampling. The commenters pointed 
out that the error due to superisokinetic sampling is accounted for 
during instrument calibration. One of the commenters explained that, 
when a sample is extracted subisokinetically, the sampling system 
collects additional large particles, resulting in a response that is 
biased high. However, when sampling is superisokinetic, the response is 
biased low because a portion of the larger particles bypass the probe. 
When sampling at 150 percent isokinetic, as do the instruments 
manufactured by the commenter's company, the error that results from a 
10 percent change in volumetric flow rate amounts to 4 percent. 
Furthermore, if the particle size distribution in the gas stream is 
relatively constant, the correlation equation accounts for this error. 
Another commenter pointed out that the error due to superisokinetic 
sampling is smaller for gas streams that have smaller sized particles, 
as is characteristic of most current emission control technologies. The 
commenter also noted that field studies on hazardous waste combustors 
have demonstrated that extractive PM CEMS that sample isokinetically 
continuously try to compensate for flow rate fluctuations and have 
trouble reaching steady state. Finally, six commenters supported the 
requirement for isokinetic sampling specified in PS-11. One of the 
commenters pointed out that the effect of nonisokinetic sampling was 
evident at a field study conducted by the Electric Power Research 
Institute; after the sampling system was adjusted to sample 
isokinetically, the performance of the instrument changed 
significantly. He noted that the argument for allowing nonisokinetic 
sampling is based on the assumption that particle size and size 
distribution remain constant, but he believes that the particle size 
distribution does not remain constant,

[[Page 1793]]

regardless of the air pollution control device used.
    Response: We agree with the commenters that, provided that PM size 
is relatively small and particle size distribution does not change 
significantly, the correlation would account for any significant errors 
that might result from sampling above isokinetic conditions. However, 
we continue to believe that isokinetic sampling is necessary when those 
particle size conditions are not met. Consequently, we have decided to 
modify the requirements for isokinetic sampling. In the proposed PS-11, 
Section 6.1(3) allowed a waiver of the requirement for isokinetic 
sampling if the owner or operator provided site-specific data that show 
that isokinetic sampling is unnecessary. We have revised this provision 
to allow the use of data from other similar installations to 
demonstrate that isokinetic sampling is not warranted. In the event 
that data from a similar installation are not available, the owner or 
operator would have to provide site-specific data that demonstrate why 
it would not be necessary to sample isokinetically. We plan to address 
this issue more comprehensively in the PM CEMS Knowledge Document.
    Comment: Two commenters agreed with the provision in Section 6.1(3) 
of PS-11 that waives the isokinetic sampling requirement for extractive 
PM CEMS if the owner or operator provides site-specific data that show 
that isokinetic sampling is not necessary. However, four commenters 
commented that this provision in PS-11 was too vague. Two commenters 
suggested that isokinetic sampling should not be a requirement if the 
resulting error is less than a specified amount (e.g., less than 10 or 
20 percent). Another commenter stated that PS-11 should allow for an 
owner or operator to conduct a particle size distribution test, and, if 
the data indicate that the particle sizes are within certain limits, 
isokinetic sampling should not be required. Another commenter stated 
that isokinetic sampling should not be required for instruments with 
proven sampling systems. One commenter indicated that subisokinetic 
sampling should be allowed without having to demonstrate that there is 
no significant bias in the response. Four commenters suggested that the 
provision for allowing site-specific approval of nonisokinetic 
extractive instruments be revised to allow consideration for particle 
size distribution. If the owner or operator could demonstrate that 90 
percent of the PM mass is less than 10 micrometers in aerodynamic 
diameter, nonisokinetic sampling would be allowed.
    Response: As stated in our previous response to the issue of 
isokinetic sampling, we have modified PS-11 to allow owners or 
operators to use data from a similar installation to demonstrate that 
isokinetic sampling is unnecessary. We appreciate the commenters' 
suggestions for how this demonstration of acceptability can be 
accomplished (e.g., by showing the resulting error is less than some 
specified amount, or by using particle size distribution data). 
However, we want to avoid being overly prescriptive in what owners and 
operators can do to satisfy this requirement. Therefore, we have 
decided against providing specifics on this demonstration of 
acceptability for instruments that do not sample isokinetically. 
However, we plan to provide additional information on this issue in the 
PM CEMS Knowledge Document.

E. Condensible PM

    Comment: One commenter supported the requirement, specified in 
Section 8.1(i) of PS-11, that extractive PM CEMS must sample at the 
reference method temperature. The commenter stated that sampling at the 
reference method temperature eliminates the possibility of creating or 
destroying PM and eliminates the introduction of bias into the 
correlation procedure and PM CEMS measurements. However, six commenters 
stated that this requirement will preclude the use of all extractive 
light-scattering instruments. They pointed out that these instruments 
typically sample at 160[deg]C (320[deg]F) to ensure that acid compounds 
are in the gaseous phase. When the sampling temperature is 120[deg]C 
(248[deg]F), as required by EPA Method 5, sulfuric acid can be present 
as a mist. According to the reference method, this mist is collected on 
the reference method sample filter, which is dried prior to weighing. 
Light-scattering instruments detect this acid mist as PM, resulting in 
a response that is biased high when compared to the reference method. 
One of the commenters suggested allowing the owner, operator, or 
equipment supplier to set the sampling temperature. Another commenter 
stated that the correlation will account for interferences, such as 
those due to the presence of condensible PM or entrained water. Another 
commenter suggested that, instead of mandating that the sampling 
temperature be the same as the reference method temperature, PS-11 
should note the temperature difference as a potential source of error 
that must be addressed if there is too much scatter in the PM CEMS 
response data.
    Response: After reviewing the comments we received on condensible 
PM, we have decided to eliminate the requirement that extractive PM 
CEMS sample at the reference method filter temperature. Sampling at 
temperatures other than the reference method filter temperature is 
acceptable provided that all of the correlation criteria are satisfied. 
We continue to recommend sampling at the reference method filter 
temperature because sampling at other temperatures may affect the 
ability to develop a correlation that satisfies all of the criteria 
specified in PS-11.

F. Instrument Location

    Comment: Several commenters submitted comments on Sections 2.4(2) 
and 8.2 of PS-11, which concern PM CEMS installation location. One 
commenter expressed support for these requirements. The commenter 
specifically supported the requirement for a PM profile test to 
evaluate PM stratification and suggested that the profile test be 
incorporated into the shakedown period. However, he indicated that the 
profile should not include the first and last traverse points, which 
are closest to the duct walls, because other factors influence the flow 
rate at those locations, and the probe for the PM CEMS will likely be 
located near the center of the duct. Another commenter found the 
requirements of Section 2.4(2) to be too prescriptive. The commenter 
suggested that we remove from PS-11 the requirements for selecting the 
location of the instrument based on a stratification test. The 
commenter believes that instrument location should be addressed in 
guidance and not in the rule itself. Two commenters pointed out that PM 
stratification and PM profile tests are not defined in PS-11, and they 
were unaware of any standard tests for stratification. One of the 
commenters also stated that EPA Method 5 may not have the accuracy to 
meet the 10 percent stratification limit. The same commenter cited an 
example of a PM CEMS installation that achieved a successful 
correlation without satisfying the stratification requirement; the 
situation could occur where a source would be forced to relocate the PM 
CEMS because it failed the stratification test, even though the data 
indicated acceptable correlation. Another commenter stated that the 10 
percent stratification limit is too stringent; the commenter suggested 
increasing the limit to 20 percent. One commenter questioned how EPA 
could enforce requirements to relocate a PM CEMS

[[Page 1794]]

based on an optional test performed according to unspecified 
procedures. Four commenters commented that elimination of 
stratification may not be feasible for some sources.
    Response: Based on our observations made during the Battleboro and 
Wisconsin Electric Power Company Pleasant Prairie Field Studies, we 
have concluded that stratification can have a significant adverse 
effect on the correlation of a PM CEMS. We also agree that additional 
clarification is needed regarding the issue of stratification and that 
the proper place for that information is in guidance. Consequently, we 
have decided to eliminate the requirement in Section 2.4(2) of PS-11 
that the PM CEMS be relocated or the stratification condition 
eliminated, if stratification varies by more than 10 percent. We plan 
to address this issue more comprehensively in the PM CEMS Knowledge 
Document, including a definition of stratification, procedures for 
evaluating stratification (e.g., profile testing), and steps that can 
be taken when stratification is likely to be a problem.

G. Shakedown and Correlation Test Planning Period (CTPP)

    Comment: One commenter voiced support for preliminary testing, 
which is recommended in Section 8.4(4) of PS-11, and suggested that 
such testing remain a recommendation and not a requirement. Another 
commenter agreed that preliminary reference method testing should be a 
recommendation, but pointed out that the specific language in PS-11 is 
too vague. Three commenters suggested that preliminary testing be 
incorporated into guidance and not be a requirement of PS-11. Although 
PS-11 does not require preliminary reference method testing, one 
commenter believes that Section II (A)(16) of the preamble to the 
December 2001 proposal implies that preliminary testing is required.
    Response: In the proposed PS-11, preliminary testing is a 
recommendation and not a requirement. We continue to believe that 
preliminary testing is advisable as a means of ensuring that the 
objectives of correlation testing are achieved. We agree that 
additional guidance on preliminary testing would be useful, and we plan 
to incorporate such guidance in later revisions of the PM CEMS 
Knowledge Document.
    Comment: Sections 8.2(4) and 8.4(4) of PS-11 suggested the use of 
bypasses as a means of achieving higher PM emissions during the CTPP; 
however, four commenters noted that the use of a bypass is prohibited 
in some jurisdictions.
    Response: We agree with the commenters that the use of a bypass may 
not be appropriate or allowed for certain installations. Therefore, we 
have revised Sections 8.2(4) and 8.4(4) to eliminate the suggestion 
that sources bypass air pollution control devices as a means of 
achieving higher emission levels during correlation testing. It was not 
our intent to require or suggest any actions that would be in violation 
of existing emission standards and other applicable requirements.
    Comment: One commenter agreed with the concept of a shakedown 
period but stated that it should not be a requirement of PS-11 because, 
as owners and operators gain experience with PM CEMS, shakedown periods 
will no longer be necessary.
    Response: We agree that operating PM CEMS for a shakedown period 
should be a recommendation and not a requirement, and we have revised 
PS-11 accordingly. We believe that shakedown periods are advisable and 
continue to recommend them, particularly for facilities with little or 
no experience in operating and maintaining PM CEMS. Owners and 
operators can benefit greatly by using a shakedown period, but 
experienced users may not feel the need to do so. In such cases, we 
believe a shakedown period may not be necessary.
    Comment: Three commenters stated that the CTPP should be a 
recommendation rather than a requirement. One of the commenters 
believes that CTPPs will no longer be necessary once owners and 
operators gain experience with PM CEMS. Another commenter supported the 
requirement for the CTPP and agrees that the time frame for the CTPP 
should not be specified. The commenter noted that each installation is 
different and requires an initial period of instrument operating time 
to characterize potential emissions. The CTPP allows the operator time 
to become familiar with instrument operation.
    Response: As is the case for the shakedown period, we urge owners 
and operators of PM CEMS to implement a CTPP to help ensure that the 
correlation tests are performed in a manner that allows development of 
a correlation over the full range of source operating conditions. 
However, we also recognize that those with experience with PM CEMS and 
familiar with their operation under various source operating conditions 
may not need to implement a CTPP. For this reason, we have decided to 
delete from PS-11 the requirement for a CTPP. We continue to believe 
that owners and operators will benefit from a CTPP and recommend that 
all owners and operators of PM CEMS give serious consideration to 
conducting a CTPP before correlation testing.
    Comment: Eight commenters objected to the requirement in Section 
8.4(2) of PS-11 that PM CEMS data recorded during the CTPP be kept as a 
permanent record. Some of these commenters pointed out that keeping the 
data as a permanent record is unnecessary because the data cannot be 
used for compliance purposes. One of the commenters indicated that this 
requirement is contrary to EPA's initiatives on reduced paperwork and 
burden. Another of the commenters believes that PS-11 should only 
require keeping the PM CEMS response range recorded during the CTPP as 
a permanent record. Six of the commenters believe that PS-11 should 
explicitly state that CTPP data cannot be used for compliance purposes. 
As proposed, they believe the recordkeeping requirements specified in 
PS-11 for the CTPP make owners and operators vulnerable to enforcement 
action. Three of the commenters questioned the need to record the CTPP 
data in 15-minute averages. One commenter stated that this requirement 
could create circumstances in which it would be difficult to recreate 
the same conditions at a later date if the data only were in 15-minute 
averages. The commenter also noted that problems could arise for 
extractive instruments with different cycle times. In the case of a 
beta gauge instrument with a 15-minute cycle time, a 15-minute 
``average'' would consist of a single measurement. He suggested that 
facilities be allowed to keep the data in the form that best suits 
their needs. One commenter supported the requirement for 15-minute data 
averages during the CTPP. The commenter believes that calculating 15-
minute averages of PM CEMS data is no more difficult than determining 
15-minute averages for gas or flow monitors. These monitoring systems 
can average the data over whatever period is required.
    Response: Because PS-11 no longer requires a CTPP, requirements 
concerning CTPP data recordkeeping also have been deleted from PS-11. 
As a result, we believe that the comments concerning the requirements 
for making a permanent record of CTPP data and recording data as 15-
minute averages are no longer relevant. This change does not 
necessarily preclude the use of CTPP data for compliance purposes if a 
facility decides to conduct a CTPP. We do not expect this issue to be a 
problem

[[Page 1795]]

because CTPP data would be generated prior to the initial compliance 
determination and before the quality of the data has been determined. 
However, the purpose of PS-11 is to specify performance criteria and 
not to define what is and what is not credible evidence. Therefore, we 
disagree that PS-11 should state that CTPP data cannot be used for 
compliance purposes.
    Comment: One commenter suggested that PS-11 allow PM spiking as a 
means of increasing the response during the CTPP. He noted that spiking 
can provide a controlled increase to instrument response without 
disrupting the process. Spiking also allows owners and operators to 
correlate PM CEMS at concentrations that approximate the emission 
limit. He pointed out that the methods suggested in Section 8.6(4)(i) 
of PS-11 for increasing PM emissions led to difficulties during EPA-
sponsored demonstration tests, and there are no such problems when PM 
spiking is used.
    Response: We concur with the commenter that PM spiking can be an 
acceptable option for increasing PM concentrations. Although we are no 
longer requiring a CTPP, owners or operators of PM CEMS will still have 
the option of conducting a CTPP. For such cases, we have indicated in 
Section 8.6(4) of PS-11 that PM spiking can be used to simulate 
increased PM concentrations during the CTPP. In addition, we have 
revised PS-11 to indicate that PM spiking is an acceptable manner for 
varying PM concentrations during correlation testing.

H. Correlation Testing

    Comment: Five commenters expressed support for the increased 
flexibility in the proposed three levels of PM emissions during the 
correlation test specified in Section 8.6(4)(iii) and (5) of PS-11. 
However, four of the commenters believe this section of the proposed 
PS-11 implies that there is greater control over PM emissions than 
there actually is for some sources. Two commenters pointed out that, 
with light-scattering instruments, the response can change with changes 
in the waste feed, making it difficult to reproduce the same response 
during correlation testing. The commenters suggested rewording Section 
8.6(5) of PS-11 to allow performing correlation testing at whatever 
range of PM concentrations the PM CEMS recorded during the CTPP.
    Response: Because we are no longer requiring a CTPP, this comment 
is largely moot. However, we have revised Section 8.6(5) of PS-11 to 
state that, in the event that the three distinct levels of PM 
concentrations cannot be achieved, owners or operators of affected PM 
CEMS must perform correlation testing over the maximum range of PM 
concentrations that is practical for that specific installation. We 
believe that this change addresses the commenters' concerns on this 
issue.
    Comment: One commenter suggested that PS-11 allow for PM spiking as 
a means of increasing the response during the correlation testing. He 
noted that spiking can provide a controlled increase to instrument 
response without disrupting the process.
    Response: We concur with the commenter that PM spiking can be an 
acceptable option for increasing PM concentrations during the 
correlation test, and we have revised Section 8.6(4)(i) of PS-11 to 
reflect that change.

I. Response Range

    Comment: Five commenters objected to the requirement of Section 
8.4(3) of PS-11, which requires owners and operators to set the 
instrument response range ``* * * such that its output is within 50 to 
60 percent of its maximum output (e.g., 12 to 13.6 mA on a 4 to 20 mA 
output) when your source is operating at the conditions that were 
previously observed to produce the highest PM CEMS output.'' The 
commenters pointed out that the resolution capabilities of current 
technology make this requirement unnecessary. In addition, allowing the 
instrument to operate below this 50 to 60 percent range at some 
installations allows more room for spikes and provides better 
measurement of low PM concentrations. The commenters believe that 
setting the response range at 50 to 60 percent of its maximum output 
should be a recommendation rather than a requirement in PS-11. One of 
the commenters pointed out that there are no such requirements for 
other types of CEMS. Another of the commenters suggested using 
preliminary testing and extrapolation to set the maximum instrument 
response at 1.1 to 1.2 times the emission limit to ensure that the 
emission limit lies within the response range of the instrument.
    Response: After considering the comments on this issue, we have 
decided to eliminate the requirement to set the response range at a 
specified percentage of the maximum PM CEMS output. Instead, PS-11 now 
requires that owners or operators set the response range at whatever 
level is necessary to ensure that the instrument measures the full 
range of responses that correspond to the range of source operating 
conditions that owners or operators will implement during correlation 
testing.

J. Reference Method Testing

    Comment: Ten commenters supported the change to allow facilities to 
use test methods other than EPA Method 5I for the correlation test. 
However, four commenters believe that sources subject to 40 CFR 63, 
subparts LLL and EEE, should be able to use EPA Method 17 for 
correlation testing instead of EPA Method 5, as required by subparts 
LLL and EEE. The commenters pointed out that EPA Method 5, which is the 
reference method specified in subparts LLL and EEE, creates a 
disincentive for light-scattering instruments because of the problems 
associated with measuring condensible PM. The same commenters also 
stated that using EPA Method 17 reduces QA problems associated with 
onsite sample analysis. One commenter suggested that EPA Method 5I be 
recommended for low emission levels.
    Response: We maintain that it is essential that correlation testing 
be performed using the same reference method that is required by the 
applicable regulation, as specified in Section 8.6(1) of PS-11, to help 
ensure that the correlation is based on emission concentration 
measurements that are consistent with the emission standard units and 
sampling method. However, we have eliminated the requirement that 
extractive PM CEMS sample at the reference method filter temperature. 
In doing so, we believe we have addressed the concern raised by the 
commenters about using EPA Method 5. Section 12.4(4) of the final PS-11 
also allows owners or operators of affected PM CEMS to petition us for 
alternative regression models or other solutions in the event that 
correlation test results cannot be modeled to satisfy the performance 
criteria for correlation coefficient, tolerance interval half range, or 
confidence interval half range specified in Section 13.2 of PS-11.
    We agree with the commenter that Method 5I may be a more 
appropriate test method for sources with low emission levels. Although 
PS-11 does not require the use of Method 5I, the method is available to 
any owner or operator that chooses to use it.
    Comment: Ten commenters agreed with the requirement for paired 
reference method trains. However, two of the commenters believe that 
other techniques to improve correlation testing also should be allowed, 
subject to approval by the Administrator. One of the commenters 
suggested that PS-11 allow an approach similar to that used in Europe 
for light-scattering

[[Page 1796]]

instruments, whereby reference method test runs are shorter in duration 
with higher flow rates. The commenter explained that this approach 
generates more data points in a shorter time frame, resulting in less 
scatter and improved correlations.
    Response: We believe that it is essential that correlation testing 
be performed using the same reference method that is required by the 
applicable regulation, as specified in Section 8.6(1) of PS-11, to help 
ensure that the correlation is based on emission concentration 
measurements that are consistent with the emission standard units and 
sampling method. However, in the event that an acceptable correlation 
cannot be achieved using the reference method specified in the 
applicable regulation, Section 12.4(4) of PS-11 allows owners or 
operators of affected PM CEMS to petition us to allow alternative 
regression models or other solutions. We also recognize that paired 
reference method sampling trains may not be necessary for obtaining 
representative PM data for certain sources. Consequently, we have 
revised PS-11 to indicate that paired sampling trains are highly 
recommended, but not required.
    We disagree with the implication that collecting more data points 
necessarily results in less scatter in the data and improved 
correlations. If the data are not collected in a manner that is 
consistent with the reference method measurements, the additional data 
may result in a correlation that is less representative of actual 
emissions. Therefore, we do not concur with the suggestion to allow 
correlation tests to be conducted with shorter test runs at higher flow 
rates.
    Comment: One commenter stated that the criteria for rejecting data 
based on the calculation of the RSD may be too restrictive. Another 
commenter expressed concern that applying the RSD criteria to paired 
data might result in valid data being rejected. If the source of error 
cannot be identified, either the data should be retained or the 
analysis should be performed both with and without the suspect data. He 
pointed out that, in the event that valid data are rejected, the 
correlation equation cannot properly characterize emissions. He also 
requested an explanation for the basis for the RSD criteria so that the 
criteria could be applied to test data for other pollutants.
    Response: We agree that data should not be rejected solely on the 
basis of a statistical criterion. The sources of error should be 
investigated in all cases. Outlying or extreme data points may be the 
result of transcription errors, data-coding errors, measurement system 
problems, and so forth. In the absence of such errors, outlying data 
may simply be an indication that the variability in the data is larger 
than expected, and we recommend keeping the data. Based on these and 
other comments on the proposed rule, we have decided to revise the 
requirements of PS-11 with respect to reference method precision. In 
the final PS-11, owners and operators would still be required to 
complete a minimum of 15 valid test runs, but can discard the results 
of up to five test runs. It is not necessary to provide an explanation 
for why the five discarded runs are rejected. We continue to believe 
that the RSD, as defined in the proposed rule, should be considered 
when deciding which test runs are to be included in the final data set. 
If the RSD for any data pair is excessive, we recommend that the data 
be investigated to determine the reason for the lack of precision. We 
are no longer requiring that the data be screened based on the RSD. 
However, we plan to provide additional information on calculating the 
RSD in guidance.
    Comment: Four commenters stated that paired data should be used as 
two discrete data points and not averaged into a single value per test 
run.
    Response: We agree that, when determining the regression relation, 
the individual data points should be used rather than the average of 
the data pairs, and we have revised Section 12.3 of PS-11 to state that 
paired data, when collected, should not be averaged. Although one 
obtains the same regression coefficients (e.g., slope and intercept) 
using either approach, a few results are different: (1) The degrees of 
freedom will increase when using all of the data points as discrete 
values; (2) the standard error of the slope and of the intercept will 
be different, which in turn will affect the width of the confidence 
intervals for the predicted mean PM concentration (y value) for a given 
response (x value); and (3) the square of the correlation coefficient 
(r\2\ value), a measure of how well the line fits the data, will 
change. Combined, these results could have an effect on the statistical 
significance of the regression relation in either direction. Using the 
average of the data points will reduce the scatter of the data, 
potentially increasing the r\2\ value, but will decrease the degrees of 
freedom and therefore increase the standard error of the intercept and 
slope estimates. On the other hand, using all the data points could 
yield more precise estimates of the slope and intercept at the cost of 
a smaller r\2\ value.
    Comment: Five commenters supported the criteria to determine 
whether the reference method data are biased. Another commenter 
believes that the slope criteria for identifying biased data may be too 
restrictive. The same commenter suggests using other statistical 
parameters, such as the t-test for evaluating the bias.
    Response: As is the case for paired data precision, we have decided 
that reference method data bias can be addressed more appropriately in 
guidance due to the complexity of the procedures for evaluating data 
bias and the need for multiple examples. Consequently, we have 
eliminated from Sections 8.6(1) and 7 of PS-11 and from Sections 2.1(3) 
and 10.1 of QA Procedure 2 the requirement for checking data for bias.
    With respect to the comment about using other statistical 
parameters to check for bias, we have concluded that a more appropriate 
statistic for determining bias is the 95 percent confidence interval 
for the slope. The confidence interval is a more widely accepted 
statistic for comparing the slopes of two regression lines. We plan to 
provide in the next revision to the PM CEMS Knowledge Document example 
calculations for checking the reference method data slope for bias.
    Comment: One commenter pointed out that the criteria for 
determining data bias consider only the slope of the correlation line. 
However, both the slope and the intercept must be considered when 
determining if the data are biased.
    Response: We agree with the commenter that the intercept must also 
be considered in the determination of data bias. The slope, or 
correlation coefficient, if different from 1, may exhibit a systematic 
difference between the two paired sampling trains. However, a 
statistically significant intercept (i.e., different from 0) would 
indicate an offset, or bias, that will not affect the slope. Although 
we have eliminated the requirements for checking reference method data 
for bias, we plan to include in guidance materials a procedure for 
checking the intercept for bias, using the 95 percent confidence 
interval for the intercept of the line. If the interval contains zero, 
it can be said with 95 percent confidence that the intercept is not 
statistically different from zero. An intercept significantly different 
from 0 would be an indication of a systematic offset between the two 
paired sampling trains, in addition to the systematic difference as 
defined by the slope of the regression line. We intend to provide 
example calculations for checking the reference

[[Page 1797]]

method data slope for bias in the next revision to the PM CEMS 
Knowledge Document.

K. Statistical Methods

    Comment: One commenter stated that the term confidence interval 
applies to the bounds within which one would predict the correlation 
line to fall. For this reason, the entire line should be considered and 
not simply the value of the confidence interval at a single point, as 
specified in Equations 11-10 and 11-33 of PS-11. The commenter believes 
the multiplier +/- (2F2, n-2, 0.05) should be used instead 
of the multiplier +/- t0.05 in the confidence interval 
equations. For 15 pairs of data, using the +/- 
(2F2, n-2, 0.05) multiplier results in a difference of 29 
percent at the 0.05 significance level. The commenter further noted 
that it is unclear whether PM CEMS would satisfy the acceptability 
criteria of PS-11 when the correct equation is used.
    Response: We agree that the definition of confidence interval in 
Section 3.4 of the PS-11 is inconsistent with Equations 11-10 and 11-33 
of the proposed PS-11. These equations represent confidence intervals 
for the predicted true mean concentration (y value) for any given PM 
CEMS response (x value). The commenter is discussing simultaneous 
confidence curves for the whole regression over its entire range. In 
this case, the commenter would be correct to replace the t-statistic 
with the F-statistic. Requiring the entire line to fall within these 
confidence bands would be a more stringent requirement than what is 
required by Equations 11-10 and 11-33 for a given value of x. In the 
final PS-11, we have replaced the definition of confidence interval 
with that of confidence interval half range, which is the parameter on 
which the correlation performance criterion is based. We believe the 
new definition is consistent with the equations presented in the final 
PS-11 for calculating this parameters. We also believe the definition 
clarifies the issue raised by the commenter.
    Comment: One commenter commented that the statistical methodology 
specified in PS-11 should also address residuals. He pointed out that, 
for the example data sets presented in Section 18 of PS-11, the pattern 
of data violates the fundamental assumption of homogeneity of the 
linear model. This violation becomes apparent when considering the 
residuals. He also noted that there is no such violation for the log-
log correlation model. Therefore, the example problem should have 
concluded that the log-log correlation model is better than the linear 
model.
    Response: We agree that residuals, which are the difference between 
the observed and predicted concentrations (y values), should be plotted 
in all regression analyses. However, we believe that residuals are best 
addressed in guidance materials rather than in PS-11. Therefore, we 
have decided against incorporating in the final PS-11 requirements for 
examining residuals. However, we intend to provide example problems and 
additional information on how to examine residuals in the PM CEMS 
Knowledge Document when it is next revised.
    Comment: One commenter opposes the elimination of the provision 
that allowed for alternative ``nonlinear'' correlation equations. In 
view of the wide range of waste types processed by hazardous waste 
combustors and the variations in how PM CEMS respond to varying 
particle characteristics, it is important to allow alternative 
calibration equations that are nonlinear. In such cases, the owner or 
operator could provide the additional correlation test data to support 
such a nonlinear correlation equation.
    Response: Section 12.4(4) of the final PS-11 allows for owners or 
operators of affected PM CEMS to petition us for alternative regression 
models or other solutions in the event that correlation test results 
cannot be modeled to satisfy the performance criteria for correlation 
coefficient, tolerance interval half range, or confidence interval half 
range specified in Section 13.2 of PS-11. We also have addressed 
additional correlation models (i.e., exponential and power 
correlations) in the final rule. We believe these provisions satisfy 
the commenter's concern by allowing for the consideration of nonlinear 
models that may be more appropriate for a specific installation.
    Comment: One commenter suggested that PS-11 should require linear 
regressions only and eliminate the criterion for a minimum correlation 
coefficient. He noted that sources with a narrow range of emissions 
will have particular difficulty in satisfying the correlation criteria. 
In such cases, the correlation could become invalid if the response 
range extrapolation limit (i.e., 125 percent of the highest response) 
is exceeded, even though the source could still be in compliance with 
the emission limit. The commenter suggested an alternative approach of 
allowing a single point correlation with the correlation line passing 
through zero, or a least-squares regression line if a range of data is 
available. The slope of the line could be adjusted to account for 
variability or uncertainty in the test method or source operation.
    Response: We disagree with the commenter that linear regressions 
are universally adequate. A straight-line regression does not always 
provide the best fit to the data, and we disagree that, in cases where 
the data exhibit a polynomial relationship, an acceptable correlation 
can be achieved by adjusting the slope of the regression line. In such 
cases, a second-order polynomial or a log transformation must be 
investigated. If the fit from such models is only marginally better 
than a linear model, then the linear model would be adequate, provided 
the residuals do not exhibit patterns.

L. Statistical Criteria

    Comment: Five commenters disagreed with specifying limits on the 
correlation coefficient, confidence interval, and tolerance interval. 
The commenters generally preferred the approach used in Europe, which 
is to determine an allowable variability or uncertainty that is then 
added to the emission limit. Sources are in compliance if their PM CEMS 
indicates that emissions are within the sum of the emission limit plus 
the allowable variability. The commenters noted that, as proposed, PS-
11 and QA Procedure 2 will be a disincentive for using PM CEMS because 
of the complexity of the statistical procedures required.
    Response: We agree with the commenter that PM CEMS compliance 
limits must account for the variability and uncertainty in the data, 
and we believe that the requirements for the correlation coefficient, 
confidence interval half range, and tolerance interval half range 
specified in the final PS-11 account for the variability and 
uncertainty in the data. The primary difference between the approach 
described by the commenters and the approach established in PS-11 is 
that the commenters' approach assumes that the uncertainty in PM CEMS 
response is one-sided, that PM CEMS invariably overestimate actual PM 
concentrations. Within the level of uncertainty, a high PM CEMS 
response that would otherwise indicate an exceedance of the emission 
limit is considered acceptable, once this uncertainty is subtracted 
from the instrument response. In our approach, we assume that there is 
uncertainty in both directions; PM CEMS responses can overestimate or 
underestimate actual PM concentrations. Just as a high PM CEMS response 
can be an overestimate of PM concentrations, our approach also accounts 
for situations in which the PM CEMS response indicates emissions are 
below the limit when an exceedance actually has occurred. Consequently, 
we

[[Page 1798]]

believe our approach is more appropriate for compliance monitoring. On 
the other hand, the requirements in PS-11 do not disallow the approach 
described by the commenters, provided that the applicable rule allows 
for such an approach.
    Comment: Nine commenters commented specifically on the reduction of 
the correlation coefficient from 0.90 to 0.85. Many of these commenters 
believe that relaxing the correlation coefficient criterion allows PM 
CEMS to be less accurate and is an admission that PM CEMS are 
inappropriate for compliance. One of the commenters stated that the 
correlation coefficient of 0.85 is evidence that the response of PM 
CEMS is highly variable and unreliable. Five of the commenters stated 
that the revised correlation criteria increase imprecision. One of the 
commenters concluded that the revised criteria ensure that defective 
technology will not be rejected by PS-11. The same commenter also 
believes that the tolerance interval criterion allows for too much 
uncertainty. Several of these commenters suggested that PS-11 should 
require PM CEMS to meet the International Standards Organization (ISO) 
correlation coefficient limit of 0.95. Two of the commenters stated 
that reducing the correlation coefficient forces a facility to operate 
even further below the emission limit to account for the increased 
uncertainty in the instrument. One commenter pointed out that the 
proposed rulemaking does not address the uncertainty inherent in 
requiring a lower correlation coefficient. One other commenter 
requested decreasing the correlation coefficient to 0.7, as is the 
practice in Germany.
    Response: We agree with the commenters that the reduction in the 
required minimum correlation coefficient value allows for more 
variability in the data, and that was our intent in changing this 
requirement. However, we do not agree that this change in the 
correlation coefficient criterion is an indication that PM CEMS are 
unreliable. We also point out that variability in correlation data can 
be accounted for in the applicable rule. If appropriate for specific 
types of sources, a higher minimum correlation coefficient can be 
specified.

M. Routine Performance Checks

    Comment: Three commenters oppose specifying routine checks in PS-11 
and QA Procedure 2. They believe that the facility should decide how 
best to maintain its instruments. One of the commenters suggested that 
QA procedure 2 should require facilities to prepare a site-specific 
inspection and maintenance program that would address all of the 
components of their PM-CEMS. Although another commenter did not object 
to the routine checks specified in QA Procedure 2, he suggested that 
owners and operators be given the option of deciding which checks are 
appropriate for their installation. The same commenter objected to any 
requirements for daily checks. He noted that weekly or monthly checks 
may be adequate for certain components of the system. He believes the 
frequency of these checks should also be left up to the facility to 
determine. One commenter noted that photometric instruments generally 
require less frequent checks than do beta gauge instruments.
    Response: Although we recognize the importance of allowing 
flexibility in how facilities maintain their PM CEMS, we believe that 
it is necessary to check instrument operation on a daily basis to 
ensure that data quality is maintained. We also would like to point out 
that daily checks are required for other types of CEMS under QA 
Procedure 1. Owners and operators who believe that daily checks are not 
necessary have the option of applying for alternative monitoring under 
Sec.  63.8(f) of the General Provisions to part 63.
    Comment: Four commenters stated that Sections 4.2(1) and (2) of PS-
11 imply that there should be routine checks for particle formation in 
extractive duct systems and for material accumulation in extractive 
duct systems. However, the procedures for performing these checks are 
unclear.
    Response: We agree with the commenters that procedures for checking 
extractive duct systems are not addressed in PS-11 or QA Procedure 2. 
Consequently, we have revised Section 9.0 of QA Procedure 2, which 
addresses the requirements of quality control (QC) programs for PM 
CEMS. We have added paragraph 9.0(8) to require owners and operators of 
affected sources to include in their QC programs written procedures for 
checking extractive duct systems for material accumulation. Rather than 
specify in PS-11 or QA Procedure 2 the required procedures for checking 
extractive ducts, we believe that the owners and operators should 
determine the most appropriate methods for accomplishing this.
    Comment: One commenter stated that several PM CEMS on the market 
eliminate the need for daily zero and upscale drift checks, and QA 
Procedure 2 should make some allowance for such instruments. If the 
facility can show that the instruments remain stable over long periods 
of time, daily drift checks should not be required. He pointed out that 
FTIR instruments used for compliance are not required to perform 
automatic zero and upscale drift checks. Another commenter also stated 
that daily drift checks are not needed for certain types of 
instruments. He suggested allowing facilities to establish the 
appropriate frequency for drift checks during the shakedown period. The 
same commenter also submitted data from a demonstration project to 
support his argument.
    Response: We believe that it is necessary to check instrument 
operation on a daily basis to ensure that data quality is maintained. 
Requiring daily checks also is consistent with QA Procedure 1. Owners 
and operators who believe that daily checks are not necessary have the 
option of applying for alternative monitoring under Sec.  63.8(f) of 
the General Provisions to part 63.
    Comment: One commenter suggested that the daily sample volume drift 
check required in Section 10.2(5) for extractive PM CEMS be expressed 
as either of the following:
[GRAPHIC] [TIFF OMITTED] TR12JA04.000

or
[GRAPHIC] [TIFF OMITTED] TR12JA04.001

where

    SVD = sample volume drift.

He noted that the purpose of drift checks is to measure stability 
rather than accuracy. Therefore, the calculation method must depend on 
a value that does not change with time, rather than depending on the 
expected value. He stated that the output from a flow monitor used in 
an extractive instrument can deviate from the expected value over time. 
If different reference values are used, it is more appropriate to use 
the monitor's full scale or span value in the denominator of the 
equation.

    Response: We agree with the commenter that using the suggested 
expression will provide more consistency in the calculation of sample 
volume drift. Therefore, we have revised Equation 2-4 of the proposed 
QA Procedure 2 accordingly. The revised equation is as follows:
[GRAPHIC] [TIFF OMITTED] TR12JA04.002

where

VR = the expected response;
VM = the actual response; and
FS = the full scale value.

[[Page 1799]]

N. Auditing Requirements

    Comment: Three commenters commented that the requirement in Section 
10.3 of QA Procedure 2 for relative response audits is unnecessary. 
They believe that, if source operating conditions do not change, the 
correlation should not change. Two other commenters suggested that 
relative response audits be required only if the source is operating 
near the emission limit. Four commenters commented that there is 
insufficient information for determining the necessary frequency of 
relative response audits.
    Response: In the proposed QA Procedure 2, relative response audits 
were required every four calendar quarters. We continue to believe that 
these audits should be performed at least annually as a means of 
ensuring that correlations remain valid. Based on our field studies, we 
have concluded that changes in emission characteristics, which may not 
be apparent to the operator, may result in correlations that are no 
longer reliable. Relative response audits are a simple means of 
checking the validity of the correlation. However, we also believe that 
it is more appropriate to specify the frequency of relative response 
audits in the applicable rule than in QA Procedure 2. Therefore, we 
have revised Section 10.3 of QA Procedure 2 to indicate that relative 
response audits must be conducted at the frequency specified in the 
applicable rule. The section also has been revised to indicate a 
recommended frequency of at least once per year.
    Comment: Four commenters supported the acceptance criterion 
specified in Section 10.4(6) of QA Procedure 2 that at least two of 
three data points must fall within the tolerance interval. However, 
they pointed out that QA Procedure 2 does not specify the allowable 
time for completing a successful relative response audit in the event 
of a failed relative response audit.
    Response: The commenters are correct in that QA Procedure 2 does 
not specify a time frame for completing a relative response audit 
successfully following a failed audit. However, following a failed 
relative response audit, PM CEMS are considered to be out of control. 
Until a successful relative response audit is completed, the data 
recorded by the PM CEMS are not considered valid and cannot be counted 
toward data availability. Consequently, the data availability 
requirements specified in the applicable rule help to ensure that 
successful relative response audits are completed in a timely manner.
    Comment: Six commenters supported the increased flexibility in the 
audit point ranges for absolute correlation audits, as specified in 
Section 10.3(2) of QA Procedure 2. However, one of the commenters 
believes that absolute correlation audits should be required only if 
the source is operating near the emission limit (within 10 percent of 
the emission limit for more than 70 percent of the operating data). 
Four of the commenters concluded that there are insufficient data to 
determine the necessary frequency for absolute correlation audits.
    Response: We believe that it is necessary to characterize 
instrument drift periodically, and quarterly absolute correlation 
audits provide the mechanism for accomplishing this objective. 
Requiring quarterly absolute correlation audits is analogous to the 
requirement of quarterly gas audits for other types of CEMS. 
Consequently, we have decided against changing the requirement for 
quarterly absolute correlation audits.
    Comment: Six commenters supported the requirement for sample volume 
audits. However, four of the commenters had reservations about some of 
the specifics of the sample volume audit requirements. They believe 
sample volume audits need only be performed annually, rather than 
quarterly as specified in Section 10.3 of QA Procedure 2. The same four 
commenters believe that the 5 percent limitation specified in Section 
10.4(4) of QA Procedure 2 is too stringent. They pointed out that the 
accuracy of EPA Methods 2, 3, and 4 are not within this 5 percent 
limit. Finally, they stated that PM CEMS should not be considered out 
of control if the instrument reads higher than actual sample flow rates 
because, in such cases, the instrument would indicate that emissions 
were higher than they actually were.
    Response: Accurate sample volume measurements are critical for 
extractive PM CEMS; otherwise, emission concentrations cannot be 
properly characterized. Therefore, we believe it is appropriate to 
require sample volume audits every quarter. Regarding the acceptance 
criterion, the data we obtained during our field studies demonstrate 
that extractive instruments can meet the 5 percent limit. In the 
absence of data that indicate otherwise, we believe the 5 percent 
acceptance criterion is appropriate.
    Comment: Six commenters expressed support for the increased 
flexibility in the requirements for response correlation audits, as 
specified in Section 10.4(5) of QA Procedure 2. Two of the commenters 
believe that the procedure should be revised to require that all 12 
data points fall below the maximum of the PM CEMS output range 
established during the correlation test, rather than within that output 
range. Four of the commenters stated that requiring all 12 data points 
to fall within the output range established during correlation testing 
is unnecessarily stringent; they suggested that QA Procedure 2 allow 
for one of the data points to fall below the output range for the 
correlation test.
    Response: We agree with the commenters that PM CEMS responses that 
fall below the range of the responses used to develop the correlation 
curve are less critical than responses that fall above the correlation 
curve response range. However, we believe that the majority of PM CEMS 
responses should occur within the range of PM CEMS responses that were 
used to develop the correlation curve. Consequently, we have revised 
Section 10.4(5) of QA Procedure 2 to require that all 12 data points 
fall below the maximum PM CEMS response used to develop the correlation 
curve, and 9 of the 12 points fall within the range of PM CEMS 
responses used to develop the correlation curve. This change provides 
additional flexibility for sources with relatively low PM emissions 
concentrations.
    Comment: Four commenters stated that response correlation audits 
should be required no more frequently than once every 5 years unless 
the source fails the relative response audit.
    Response: We believe that the required frequency of response 
correlation audits should depend on source operation and emission 
characteristics. Consequently, we continue to believe that it is 
appropriate for the frequency of response correlation audits to be 
specified in the applicable regulation or operating permit, rather than 
in QA Procedure 2. Although it may be appropriate for some sources to 
perform response correlation audits once every 5 years, as the 
commenters suggested, more frequent audits may be appropriate for other 
sources. Therefore, we have decided against revising QA Procedure 2 to 
specify a required frequency for response correlation audits, as 
suggested by the commenters.

O. Extrapolation of Correlation

    Comment: Nine commenters oppose the limits on PM CEMS extrapolation 
to 3 consecutive hours in excess of 125 percent of the highest response 
used to develop the correlation curve before additional correlation 
testing is required, as specified in Section 8.8(1) of the proposed PS-
11. Four of the

[[Page 1800]]

commenters suggested that additional flexibility be allowed for sources 
that operate well below the emission limit. Although one of the 
commenters stated that he generally agreed with this requirement, he 
had reservations about some of the specific requirements. One commenter 
suggested that the basis for requiring additional correlation testing 
should be the proximity of emissions to the emission limit rather than 
the exceedance of the correlation test response range. He suggested 
that additional testing be required only for situations in which the 
source persistently operates close to the emission limit when it had 
previously operated well below the emission limit. Four commenters 
found the provisions regarding exceedances of 125 percent of the 
correlation range to be too vague and suggested revising the section to 
not require additional testing in cases where the three hourly averages 
exceeding 125 percent of the highest PM CEMS response occur only 
infrequently.
    Response: We agree with the commenters that the 125 percent limit 
on extrapolation of the correlation equation should not apply to 
sources that operate well below the emission limit. We have revised 
Section 8.8(1) of PS-11 to allow sources that operate below 50 percent 
of the emission limit to extrapolate up to 50 percent of the emission 
limit or 125 percent of the highest PM CEMS response used in developing 
the correlation, whichever results in a higher PM concentration.
    Comment: Regarding the requirement in Section 8.8(1) of PS-11 for 
additional correlation testing, two commenters indicated that, even if 
the facility begins corrective action immediately, it may take more 
than 3 hours to correct the problem. Four commenters stated that, when 
a 3-hour exceedance occurs, it is typically due to an unusual event 
that is difficult to reproduce. The same four commenters believe that 
three consecutive hourly averages do not constitute a pattern and that 
it could be difficult to re-create a high PM event for additional 
correlation testing. Two of the commenters suggested allowing the 
facility to make the determination as to whether such an event was 
routine or unusual.
    Response: We agree with the commenters that PS-11 should allow more 
time before additional correlation testing is required following a PM 
CEMS response in excess of 125 percent of the highest response used to 
develop the correlation curve. We have revised Section 8.8(1) of PS-11 
to increase the time period that triggers additional correlation 
testing from 3 consecutive hours to 24 consecutive hours or 5 percent 
of the valid operating hours for the previous 30-day period, whichever 
occurs first. We believe that 24 hours is a reasonable length of time 
for source operators to be alerted of the event, determine the cause, 
identify the corrective action needed, and complete the corrective 
action. We included the 5 percent criterion to address recurring 
problems or events that individually may not last 24 consecutive hours, 
but nonetheless represent a change in operation or emissions 
characteristics that must be accounted for by the PM CEMS correlation.
    We have also included in Section 8.8(4) of the final rule a 
requirement that the owner or operator of an affected PM CEMS report 
the reason for the higher PM responses. In that report, that owner or 
operator must specify if the higher responses resulted from normal 
operation or from an atypical event. We believe this provision 
addresses the comment about the facility making the determination of 
whether or not the higher PM CEMS responses were due to normal 
operation.
    Comment: Five commenters commented that 30 days is inadequate for 
setting up and conducting a test following an exceedance that is more 
than 125 percent of the response range for the correlation test. Two of 
the commenters believe that PS-11 should allow up to 60 days to conduct 
additional correlation tests, and one of the commenters believes up to 
120 days should be allowed for testing in such cases.
    Response: We agree with the commenters that 30 days is inadequate 
for scheduling and conducting additional correlation tests and 
developing a revised correlation. We recognize that scheduling an 
emission test and bringing the testing contractor on site can take 
several weeks; the test itself may last several days for setup, 
testing, and breakdown; analyzing samples, compiling the data, and 
performing emissions calculations typically take several days; and 
developing the revised correlation also may require several days. 
Consequently, we have revised QA Procedure 2 to allow 60 days to 
complete these activities. We believe that 60 days is a reasonable 
length of time for completing all of the activities needed to develop a 
revised correlation curve.

P. Requirements for Other Types of Monitors

    Comment: One commenter commented that PS-11 requires additional 
monitoring systems to satisfy the requirement that emissions are 
recorded in the same units as the emission standard, but does not 
address performance requirements for those supplemental monitoring 
systems. He noted that the emission limit in 40 CFR part 63, subpart 
LLL, is specified in units of pounds per ton of clinker. To report PM 
emissions in these units requires converting PM emission concentrations 
and clinker production rates to units of mass per unit time, and, to do 
so requires monitoring exhaust gas flow rates and production mass flow 
rates. However, there currently are no performance specifications or QA 
procedures for either type of monitoring system. The commenter also 
stated that measurement error and uncertainty in these supplemental 
monitoring systems will influence the error and uncertainty in the 
emission data that are reported.
    Response: We recognize the need for performance specifications and 
QA procedures that address continuous parameter monitoring systems 
(CPMS). We are currently developing these specifications and procedures 
and expect to propose them in the near future. The performance 
specifications and QA procedures for CPMS would apply to all sources 
subject to a part 63 rule that requires continuous parameter 
monitoring.

IV. Summary of Impacts

A. What Are the Impacts of PS-11 and QA Procedure 2?

    The PS-11 and QA Procedure 2 will apply only to PM CEMS that are 
required under an applicable rule. Rules, such as PS-11 and QA 
Procedure 2 that establish performance specifications and QA 
requirements, impose no costs independent from the emission standards 
that require their use, and such costs are fully reflected in the 
regulatory impact assessments for those emission standards. Likewise, 
the other impacts associated with the monitoring requirements specified 
in PS-11 and QA Procedure 2 are already addressed under the applicable 
emission standards as they are proposed and promulgated. Consequently, 
we have concluded that no separate estimate of the impacts is warranted 
for this rulemaking.

V. Statutory and Executive Order Reviews

A. Executive Order 12866: Regulatory Planning and Review

    Under Executive Order 12866 (58 FR 51735, October 4, 1993), EPA 
must determine whether the regulatory action is ``significant'' and, 
therefore, subject to review by the Office of Management and Budget 
(OMB) and the requirements of the Executive Order. The Executive

[[Page 1801]]

Order defines ``significant regulatory action'' as one that is likely 
to result in a rule that may:
    (1) Have an annual effect on the economy of $100 million or more or 
adversely affect in a material way the economy, a sector of the 
economy, productivity, competition, jobs, the environment, public 
health or safety, or State, local, or tribal governments or 
communities;
    (2) Create a serious inconsistency or otherwise interfere with an 
action taken or planned by another agency;
    (3) Materially alter the budgetary impact of entitlements, grants, 
user fees, or loan programs, or the rights and obligation of recipients 
thereof; or
    (4) Raise novel legal or policy issues arising out of legal 
mandates, the President's priorities, or the principles set forth in 
the Executive Order.
    Pursuant to the terms of Executive Order 12866, it has been 
determined that this rule is not a ``significant regulatory action'' 
because none of the listed criteria applies to this action. 
Consequently, this action was not submitted to OMB for review under 
Executive Order 12866.

B. Paperwork Reduction Act

    This final rule does not contain any information collection 
requirements subject to the Office of Management and Budget review 
under the Paperwork Reduction Act of 1980, 44 U.S.C. 3501 et seq. The 
recording, recordkeeping, and information collection requirements 
associated with PS-11 and QA Procedure 2 have already been accounted 
for under the applicable regulations that require the use of PM CEMS.

C. Regulatory Flexibility Act (RFA)

    The RFA generally requires an agency to prepare a regulatory 
flexibility analysis for any rule subject to notice and comment 
rulemaking requirements under the Administrative Procedure Act, or any 
other statute, unless the agency certifies that the rule will not have 
a significant economic impact on a substantial number of small 
entities. Small entities include small businesses, small organizations, 
and small governmental jurisdictions.
    For purposes of assessing the impacts of today's rule on small 
entities, small entity is defined as: (1) A small business as defined 
by the Small Business Administrations' regulations at 13 CFR 121.201; 
(2) a small governmental jurisdiction that is a government of a city, 
county, town, school district or special district with a population of 
less than 50,000; and (3) a small organization that is any not-for-
profit enterprise which is independently owned and operated and is not 
dominant in its field.
    After considering the economic impacts of today's final rule on 
small entities, I certify that this action will not have a significant 
economic impact on a substantial number of small entities. This final 
rule will establish performance specifications and QA requirements and 
will not impose any costs. Likewise, the other impacts associated with 
the monitoring requirements specified in PS-11 and QA Procedure 2 are 
already addressed under the applicable emission standards as they are 
proposed and promulgated. Consequently, we have concluded that no 
separate estimate of the impacts is warranted for this rulemaking.

D. Unfunded Mandates Reform Act

    Title II of the Unfunded Mandates Reform Act of 1995 (UMRA), Public 
Law No. 104-4, establishes requirements for Federal agencies to assess 
the effects of their regulatory actions on State, local, and tribal 
governments and the private sector. Under section 202 of the UMRA, EPA 
generally must prepare a written statement, including a cost-benefit 
analysis, for proposed and final rules with ``Federal mandates'' that 
may result in expenditures by State, local, and tribal governments, in 
the aggregate, or by the private sector, of $100 million or more in any 
1 year. Before promulgating an EPA rule for which a written statement 
is needed, section 205 of the UMRA generally requires EPA to identify 
and consider a reasonable number of regulatory alternatives and adopt 
the least costly, most cost-effective, or least burdensome alternative 
that achieves the objectives of the rule. The provisions of section 205 
do not apply when they are inconsistent with applicable law. Moreover, 
section 205 allows EPA to adopt an alternative other than the least 
costly, most cost-effective, or least burdensome alternative if the 
Administrator publishes with the final rule an explanation why that 
alternative was not adopted. Before EPA establishes any regulatory 
requirements that may significantly or uniquely affect small 
governments, including tribal governments, it must have developed under 
section 203 of the UMRA a small government agency plan. The plan must 
provide for notifying potentially affected small governments, enabling 
officials of affected small governments to have meaningful and timely 
input in the development of EPA's regulatory proposals with significant 
Federal intergovernmental mandates, and informing, educating, and 
advising small governments on compliance with the regulatory 
requirements.
    The EPA has determined that today's final rule does not contain a 
Federal mandate that may result in expenditures of $100 million or more 
for State, local, and tribal governments in the aggregate, or the 
private sector in any 1 year. Rules establishing performance 
specifications and quality assurance requirements impose no costs 
independent from national emission standards which require their use, 
and such costs are fully reflected in the regulatory impact assessment 
for those emission standards. We have also determined that this final 
rule does not significantly or uniquely impact small governments. 
Therefore, the requirements of the Unfunded Mandates Act do not apply 
to this action.

E. Executive Order 13132: Federalism

    Executive Order 13132, entitled ``Federalism'' (64 FR 43255, August 
10, 1999), requires EPA to develop an accountable process to ensure 
``meaningful and timely input by State and local officials in the 
development of regulatory policies that have federalism implications.'' 
``Policies that have federalism implications'' is defined in the 
Executive Order to include regulations that have ``substantial direct 
effects on the States, on the relationship between the national 
government and the States, or on the distribution of power and 
responsibilities among the various levels of government.''
    The final rule does not have federalism implications. It will not 
have substantial direct effects on the States, on the relationship 
between the national government and the States, or on the distribution 
of power and responsibilities among the various levels of government, 
as specified in Executive Order 13132. The requirements of PS-11 and QA 
Procedure 2 are addressed under the applicable emission standards that 
require the use of PM CEMS. Thus, the requirements of section 6 of the 
Executive Order do not apply to this final rule.

F. Executive Order 13175: Consultation and Coordination With Indian 
Tribal Governments

    Executive Order 13175, entitled ``Consultation and Coordination 
with Indian Tribal Governments'' (65 FR 67249, November 9, 2000), 
requires EPA to develop an accountable process to ensure ``meaningful 
and timely input by tribal officials in the development of regulatory 
policies that have tribal implications.'' The final rule does not

[[Page 1802]]

have tribal implications. It will not have substantial direct effects 
on tribal governments, on the relationship between the Federal 
government and Indian tribes, or on the distribution of power and 
responsibilities between the Federal government and Indian tribes, as 
specified in Executive Order 13175. The requirements of PS-11 and QA 
Procedure 2 are addressed under the applicable emission standards that 
require the use of PM CEMS. Thus, Executive Order 13175 does not apply 
to the final rule.

G. Executive Order 13045: Protection of Children From Environmental 
Health Risks and Safety Risks

    Executive Order 13045 (62 FR 19885, April 23, 1997) applies to any 
rule that: (1) Is determined to be ``economically significant'' as 
defined under Executive Order 12866, and (2) concerns an environmental 
health or safety risk that EPA has reason to believe may have a 
disproportionate effect on children. If the regulatory action meets 
both criteria, EPA must evaluate the environmental health or safety 
effects of the planned rule on children, and explain why the planned 
rule is preferable to other potentially effective and reasonably 
feasible alternatives that EPA considered.
    The EPA interprets Executive Order 13045 as applying only to those 
regulatory actions that are based on health or safety risks, such that 
the analysis required under section 5-501 of the Executive Order has 
the potential to influence the regulation. Today's final rule is not 
subject to Executive Order 13045 because this rule does not establish 
an environmental standard intended to mitigate health or safety risks. 
Furthermore, the final rule has been determined not to be 
``economically significant'' as defined under Executive Order 12866.

H. Executive Order 13211: Actions Concerning Regulations That 
Significantly Affect Energy Supply, Distribution, or Use

    Today's final rule is not subject to Executive Order 13211 (66 FR 
28355, May 22, 2001) because it is not a significant regulatory action 
under Executive Order 12866.

I. National Technology Transfer and Advancement Act

    Section 12(d) of the National Technology Transfer and Advancement 
Act (NTTAA) of 1995 (Public Law No. 104-113; 15 U.S.C. 272 note) 
directs the EPA to use voluntary consensus standards in their 
regulatory and procurement activities unless to do so would be 
inconsistent with applicable law or otherwise impractical. Voluntary 
consensus standards are technical standards (e.g., materials 
specifications, test methods, sampling procedures, business practices) 
developed or adopted by one or more voluntary consensus standards 
bodies. The NTTAA directs EPA to provide Congress, through annual 
reports to the OMB, with explanations when an agency does not use 
available and applicable voluntary consensus standards.
    During this rulemaking, we searched for voluntary consensus 
standards that might be applicable. An International Organization for 
Standardization (ISO) standard, number 10155, Stationary source 
emissions--Automated monitoring of mass concentrations of particles--
Performance characteristics, test methods and specifications, was 
applicable. The use of the ISO 10155 was found to be inadequate to 
fulfill the performance specification needs for our compliance 
monitoring. The use of ISO 10155 would be impractical because:
    (1) The number of test runs for a correlation test, 9, was 
insufficient for a comprehensive statistical evaluation of the PM CEMS 
correlation.
    (2) The PM concentration ranges required for a correlation test 
were too vague.
    (3) The measurement location for the PM CEMS and RM were vague.
    (4) The correlation coefficient limit of greater than 0.95 was too 
stringent for most of the PM CEMS correlations we evaluated.
    Also, ISO 10155 lacks quality assurance and quality control 
procedures.

J. Congressional Review Act

    The Congressional Review Act, 5 U.S.C. 801 et seq., as added by the 
Small Business Regulatory Enforcement Fairness Act of 1996, generally 
provides that before a rule may take effect, the agency promulgating 
the rule must submit a rule report, which includes a copy of the rule, 
to each House of the Congress and to the Comptroller General of the 
United States. The EPA will submit a report containing the rule and 
other required information to the U.S. Senate, the U.S. House of 
Representatives, and the Comptroller General of the United States prior 
to publication of the rule in the Federal Register. A major rule cannot 
take effect until March 12, 2004. This action is not a ``major rule'' 
as defined by 5 U.S.C. 804(2).

List of Subjects in 40 CFR Part 60

    Environmental protection, Air Pollution Control, Continuous 
emission monitoring; Performance specification; Particulate matter.

    Dated: December 23, 2003.
Michael O. Leavitt,
Administrator.

0
For the reasons stated in the preamble, title 40, chapter I, part 60 of 
the Code of Federal Regulations is amended as follows:
0
1. The authority citation for part 60 continues to read as follows:

    Authority: 42 U.S.C. 7401, et seq.

0
2. Appendix B of Part 60 is amended by adding Performance Specification 
11 to read as follows:

Appendix B of Part 60--Performance Specifications

* * * * *

PERFORMANCE SPECIFICATION 11--Specifications and Test Procedures for 
Particulate Matter Continuous Emission Monitoring Systems at Stationary 
Sources

1.0 What Are the Purpose and Applicability of Performance Specification 
11?
    The purpose of Performance Specification 11 (PS-11) is to establish 
the initial installation and performance procedures that are required 
for evaluating the acceptability of a particulate matter (PM) 
continuous emission monitoring system (CEMS); it is not to evaluate the 
ongoing performance of your PM CEMS over an extended period of time, 
nor to identify specific calibration techniques and auxiliary 
procedures to assess CEMS performance. You will find procedures for 
evaluating the ongoing performance of a PM CEMS in Procedure 2 of 
Appendix F--Quality Assurance Requirements for Particulate Matter 
Continuous Emission Monitoring Systems Used at Stationary Sources.
    1.1 Under what conditions does PS-11 apply to my PM CEMS? The PS-11 
applies to your PM CEMS if you are required by any provision of Title 
40 of the Code of Federal Regulations (CFR) to install and operate PM 
CEMS.
    1.2 When must I comply with PS-11? You must comply with PS-11 when 
directed by the applicable rule that requires you to install and 
operate a PM CEMS.
    1.3 What other monitoring must I perform? To report your PM 
emissions in units of the emission standard, you may need to monitor 
additional parameters to correct the PM concentration reported by your 
PM

[[Page 1803]]

CEMS. Your CEMS may include the components listed in paragraphs (1) 
through (3) of this section:
    (1) A diluent monitor (i.e., O2, CO2, or 
other CEMS specified in the applicable regulation), which must meet its 
own performance specifications (also found in this appendix),
    (2) Auxiliary monitoring equipment to allow measurement, 
determination, or input of the flue gas temperature, pressure, moisture 
content, and/or dry volume of stack effluent sampled, and
    (3) An automatic sampling system. The performance of your PM CEMS 
and the establishment of its correlation to manual reference method 
measurements must be determined in units of mass concentration as 
measured by your PM CEMS (e.g., milligrams per actual cubic meter (mg/
acm) or milligrams per dry standard cubic meter (mg/dscm)).
2.0 What Are the Basic Requirements of PS-11?
    The PS-11 requires you to perform initial installation and 
calibration procedures that confirm the acceptability of your CEMS when 
it is installed and placed into operation. You must develop a site-
specific correlation of your PM CEMS response against manual 
gravimetric reference method measurements (including those made using 
EPA Methods 5, 5I, or 17).
    2.1 What types of PM CEMS technologies are covered? Several 
different types of PM CEMS technologies (e.g., light scattering, Beta 
attenuation, etc.) can be designed with in-situ or extractive sample 
gas handling systems. Each PM CEMS technology and sample gas handling 
technology has certain site-specific advantages. You should select and 
install a PM CEMS that is appropriate for the flue gas conditions at 
your source.
    2.2 How is PS-11 different from other performance specifications? 
The PS-11 is based on a technique of correlating PM CEMS responses 
relative to emission concentrations determined by the reference method. 
This technique is called ``the correlation.'' This differs from CEMS 
used to measure gaseous pollutants that have available calibration 
gases of known concentration. Because the type and characteristics of 
PM vary from source to source, a single PM correlation, applicable to 
all sources, is not possible.
    2.3 How are the correlation data handled? You must carefully review 
your manual reference method data and your PM CEMS responses to include 
only valid, high-quality data. For the correlation, you must convert 
the manual reference method data to measurement conditions (e.g., wet 
or dry basis) that are consistent with your PM CEMS. Then, you must 
correlate the manual method and PM CEMS data in terms of the output as 
received from the monitor (e.g., milliamps). At the appropriate PM CEMS 
response specified in section 13.2 of this performance specification, 
you must calculate the confidence interval half range and tolerance 
interval half range as a percentage of the applicable PM concentration 
emission limit and compare the confidence interval and tolerance 
interval percentages with the performance criteria. Also, you must 
calculate the correlation coefficient and compare the correlation 
coefficient with the applicable performance criterion specified in 
section 13.2 of this performance specification.
    Situations may arise where you will need two or more correlations. 
If you need multiple correlations, you must collect sufficient data for 
each correlation, and each correlation must satisfy the performance 
criteria specified in section 13.2 of this performance specification.
    2.4 How do I design my PM CEMS correlation program? When planning 
your PM CEMS correlation effort, you must address each of the items in 
paragraphs (1) through (7) of this section to enhance the probability 
of success. You will find each of these elements further described in 
this performance specification or in the applicable reference method 
procedure.
    (1) What type of PM CEMS should I select? You should select a PM 
CEMS that is appropriate for your source with technical consideration 
for potential factors such as interferences, site-specific 
configurations, installation location, flue gas conditions, PM 
concentration range, and other PM characteristics. You can find 
guidance on which technology is best suited for specific situations in 
our report ``Current Knowledge of Particulate Matter (PM) Continuous 
Emission Monitoring'' (PM CEMS Knowledge Document, see section 16.5).
    (2) Where should I install my PM CEMS? Your PM CEMS must be 
installed in a location that is most representative of PM emissions, as 
determined by the reference method, such that the correlation between 
PM CEMS response and emissions determined by the reference method will 
meet these performance specifications. Care must be taken in selecting 
a location and measurement point to minimize problems due to flow 
disturbances, cyclonic flow, and varying PM stratification.
    (3) How should I record my CEMS data? You need to ensure that your 
PM CEMS and data logger are set up to collect and record all normal 
emission levels and excursions. You must ensure that your data logger 
and PM CEMS have been properly programmed to accept and transfer status 
signals of valid monitor operation (e.g., flags for internal 
calibration, suspect data, or maintenance periods).
    (4) What CEMS data should I review? You must review drift data 
daily to document proper operation. You must also ensure that any audit 
material is appropriate for the typical operating range of your PM 
CEMS.
    (5) How long should I operate my PM CEMS before conducting the 
initial correlation test? You should allow sufficient time for your PM 
CEMS to operate for you to become familiar with your PM CEMS.
    (i) You should observe PM CEMS response over time during normal and 
varying process conditions. This will ensure that your PM CEMS has been 
properly set up to operate at a range that is compatible with the 
concentrations and characteristics of PM emissions for your source. You 
should use this information to establish the range of operating 
conditions necessary to determine the correlations of PM CEMS data to 
manual reference method measurements over a wide operating range.
    (ii) You must determine the types of process changes that will 
influence, on a definable and repeatable basis, flue gas PM 
concentrations and the resulting PM CEMS responses. You may find this 
period useful to make adjustments to your planned approach for 
operating your PM CEMS at your source. For instance, you may change the 
measurement range or batch sampling period to something other than 
those you initially planned to use.
    (6) How do I conduct the initial correlation test? When conducting 
the initial correlation test of your PM CEMS response to PM emissions 
determined by the reference method, you must pay close attention to 
accuracy and details. Your PM CEMS must be operating properly. You must 
perform the manual reference method testing accurately, with attention 
to eliminating site-specific systemic errors. You must coordinate the 
timing of the manual reference method testing with the sampling cycle 
of your PM CEMS. You must complete a minimum of 15 manual reference 
method tests. You must perform the manual reference method testing over 
the full range of PM CEMS responses that correspond to normal operating 
conditions for your source and control device and will result in the

[[Page 1804]]

widest range of emission concentrations.
    (7) How should I perform the manual reference method testing? You 
must perform the manual reference method testing in accordance with 
specific rule requirements, coordinated closely with PM CEMS and 
process operations. It is highly recommended that you use paired trains 
for the manual reference method testing. You must perform the manual 
reference method testing over a suitable PM concentration range that 
corresponds to the full range of normal process and control device 
operating conditions. Because the manual reference method testing for 
this correlation test is not for compliance reporting purposes, you may 
conduct the reference method test runs for less than the typical 
minimum test run duration of 1 hour.
    (8) What do I do with the manual reference method data and PM CEMS 
data? You must complete each of the activities in paragraphs (8)(i) 
through (v) of this section.
    (i) Screen the manual reference method data for validity (e.g., 
isokinetics, leak checks), quality assurance, and quality control 
(e.g., outlier identification).
    (ii) Screen your PM CEMS data for validity (e.g., daily drift check 
requirements) and quality assurance (e.g., flagged data).
    (iii) Convert the manual reference method test data into 
measurement units (e.g., mg/acm) consistent with the measurement 
conditions of your PM CEMS.
    (iv) Calculate the correlation equation(s) as specified in section 
12.3.
    (v) Calculate the correlation coefficient, confidence interval half 
range, and tolerance interval half range for the complete set of PM 
CEMS and reference method correlation data for comparison with the 
correlation performance criteria specified in section 13.2.
    2.5 What other procedures must I perform? Before conducting the 
initial correlation test, you must successfully complete a 7-day drift 
test (See section 8.5).
3.0 What Special Definitions Apply to PS-11?
    3.1 ``Appropriate Measurement Range of your PM CEMS'' means a 
measurement range that is capable of recording readings over the 
complete range of your source's PM emission concentrations during 
routine operations. The appropriate range is determined during the 
pretest preparations as specified in section 8.4.
    3.2 ``Appropriate Data Range for PM CEMS Correlation'' means the 
data range that reflects the full range of your source's PM emission 
concentrations recorded by your PM CEMS during the correlation test 
planning period or other normal operations as defined in the applicable 
regulations.
    3.3 ``Batch Sampling'' means that gas is sampled on an intermittent 
basis and concentrated on a collection medium before intermittent 
analysis and follow-up reporting. Beta gauge PM CEMS are an example of 
batch sampling devices.
    3.4 ``Confidence Interval Half Range (CI)'' means the statistical 
term for one-half of the width of the 95 percent confidence interval 
around the predicated mean PM concentration (y value) calculated at the 
PM CEMS response value (x value) where the confidence interval is 
narrowest. Procedures for calculating CI are specified in section 
12.3(1)(ii) for linear correlations and in section 12.3(2)(ii) for 
polynomial correlations. The CI as a percent of the emission limit 
value (CI%) is calculated at the appropriate PM CEMS response value 
specified in Section 13.2(2).
    3.5 ``Continuous Emission Monitoring System (CEMS)'' means all of 
the equipment required for determination of PM mass concentration in 
units of the emission standard. The sample interface, pollutant 
monitor, diluent monitor, other auxiliary data monitor(s), and data 
recorder are the major subsystems of your CEMS.
    3.6 ``Correlation'' means the primary mathematical relationship for 
correlating the output from your PM CEMS to a PM concentration, as 
determined by the PM reference method. The correlation is expressed in 
the measurement units that are consistent with the measurement 
conditions (e.g., mg/dscm, mg/acm) of your PM CEMS.
    3.7 ``Correlation Coefficient (r)'' means a quantitative measure of 
the association between your PM CEMS outputs and the reference method 
measurements. Equations for calculating the r value are provided in 
section 12.3(1)(iv) for linear correlations and in section 12.3(2)(iv) 
for polynomial correlations.
    3.8 ``Cycle Time'' means the time required to complete one 
sampling, measurement, and reporting cycle. For a batch sampling PM 
CEMS, the cycle time would start when sample gas is first extracted 
from the stack/duct and end when the measurement of that batch sample 
is complete and a new result for that batch sample is produced on the 
data recorder.
    3.9 ``Data Recorder'' means the portion of your CEMS that provides 
a permanent record of the monitor output in terms of response and 
status (flags). The data recorder may also provide automatic data 
reduction and CEMS control capabilities (see section 6.6).
    3.10 ``Diluent Monitor and Other Auxiliary Data Monitor(s) (if 
applicable)'' means the portion of your CEMS that provides the diluent 
gas concentration (such as O2 or CO2, as 
specified by the applicable regulations), temperature, pressure, and/or 
moisture content, and generates an output proportional to the diluent 
gas concentration or gas property.
    3.11 ``Drift Check'' means a check of the difference between your 
PM CEMS output readings and the established reference value of a 
reference standard or procedure after a stated period of operation 
during which no unscheduled maintenance, repair, or adjustment took 
place. The procedures used to determine drift are specific to the 
operating principles of your specific PM CEMS. A drift check includes 
both a zero drift check and an upscale drift check.
    3.12 ``Exponential Correlation'' means an exponential equation used 
to define the relationship between your PM CEMS output and the 
reference method PM concentration, as indicated by Equation 11-37.
    3.13 ``Flagged Data'' means data marked by your CEMS indicating 
that the response value(s) from one or more CEMS subsystems is suspect 
or invalid or that your PM CEMS is not in source-measurement operating 
mode.
    3.14 ``Linear Correlation'' means a first-order mathematical 
relationship between your PM CEMS output and the reference method PM 
concentration that is linear in form, as indicated by Equation 11-3.
    3.15 ``Logarithmic Correlation'' means a first-order mathematical 
relationship between the natural logarithm of your PM CEMS output and 
the reference method PM concentration that is linear in form, as 
indicated by Equation 11-34.
    3.16 ``Low-Emitting Source'' means a source that operated at no 
more than 50 percent of the emission limit during the most recent 
performance test, and, based on the PM CEMS correlation, the daily 
average emissions for the source, measured in the units of the 
applicable emission limit, have not exceeded 50 percent of the emission 
limit for any day since the most recent performance test.
    3.17 ``Paired Trains'' means two reference method trains that are 
used to conduct simultaneous measurements of PM concentrations. 
Guidance on the use

[[Page 1805]]

of paired sampling trains can be found in the PM CEMS Knowledge 
Document (see section 16.5).
    3.18 ``Polynomial Correlation'' means a second-order equation used 
to define the relationship between your PM CEMS output and reference 
method PM concentration, as indicated by Equation 11-16.
    3.19 ``Power Correlation'' means an equation used to define a power 
function relationship between your PM CEMS output and the reference 
method concentration, as indicated by Equation 11-42.
    3.20 ``Reference Method'' means the method defined in the 
applicable regulations, but commonly refers to those methods 
collectively known as EPA Methods 5, 5I, and 17 (for particulate 
matter), found in Appendix A of 40 CFR 60. Only the front half and dry 
filter catch portions of the reference method can be correlated to your 
PM CEMS output.
    3.21 ``Reference Standard'' means a reference material or procedure 
that produces a known and unchanging response when presented to the 
pollutant monitor portion of your CEMS. You must use these standards to 
evaluate the overall operation of your PM CEMS, but not to develop a PM 
CEMS correlation.
    3.22 ``Response Time'' means the time interval between the start of 
a step change in the system input and the time when the pollutant 
monitor output reaches 95 percent of the final value (see sections 6.5 
and 13.3 for procedures and acceptance criteria).
    3.23 ``Sample Interface'' means the portion of your CEMS used for 
one or more of the following: sample acquisition, sample delivery, 
sample conditioning, or protection of the monitor from the effects of 
the stack effluent.
    3.24 ``Sample Volume Check'' means a check of the difference 
between your PM CEMS sample volume reading and the sample volume 
reference value.
    3.25 ``Tolerance Interval half range (TI)'' means one-half of the 
width of the tolerance interval with upper and lower limits, within 
which a specified percentage of the future data population is contained 
with a given level of confidence, as defined by the respective 
tolerance interval half range equations in section 12.3(1)(iii) for 
linear correlations and in section 12.3(2)(iii) for polynomial 
correlations. The TI as a percent of the emission limit value (TI%) is 
calculated at the appropriate PM CEMS response value specified in 
Section 13.2(3).
    3.26 ``Upscale Check Value'' means the expected response to a 
reference standard or procedure used to check the upscale response of 
your PM CEMS.
    3.27 ``Upscale Drift (UD) Check'' means a check of the difference 
between your PM CEMS output reading and the upscale check value.
    3.28 ``Zero Check Value'' means the expected response to a 
reference standard or procedure used to check the response of your PM 
CEMS to particulate-free or low-particulate concentration conditions.
    3.29 ``Zero Drift (ZD) Check'' means a check of the difference 
between your PM CEMS output reading and the zero check value.
    3.30 ``Zero Point Correlation Value'' means a value added to PM 
CEMS correlation data to represent low or near zero PM concentration 
data (see section 8.6 for rationale and procedures).
4.0 Are There Any Potential Interferences for My PM CEMS?
    Yes, condensible water droplets or condensible acid gas aerosols 
(i.e., those with condensation temperatures above those specified by 
the reference method) at the measurement location can be interferences 
for your PM CEMS if the necessary precautions are not met.
    4.1 Where are interferences likely to occur? Interferences may 
develop if your CEMS is installed downstream of a wet air pollution 
control system or any other conditions that produce flue gases, which, 
at your PM CEMS measurement point, normally or occasionally contain 
entrained water droplets or condensible salts before release to the 
atmosphere.
    4.2 How do I deal with interferences? We recommend that you use a 
PM CEMS that extracts and heats representative samples of the flue gas 
for measurement to simulate results produced by the reference method 
for conditions such as those described in section 4.1. Independent of 
your PM CEMS measurement technology and extractive technique, you 
should have a configuration simulating the reference method to ensure 
that:
    (1) No formation of new PM or deposition of PM occurs in sample 
delivery from the stack or duct; and
    (2) No condensate accumulates in the sample flow measurement 
apparatus.
    4.3 What PM CEMS measurement technologies should I use? You should 
use a PM CEMS measurement technology that is free of interferences from 
any condensible constituent in the flue gas.
5.0 What Do I Need To Know To Ensure the Safety of Persons Using PS-11?
    People using the procedures required under PS-11 may be exposed to 
hazardous materials, operations, site conditions, and equipment. This 
performance specification does not purport to address all of the safety 
issues associated with its use. It is your responsibility to establish 
appropriate safety and health practices and determine the applicable 
regulatory limitations before performing these procedures. You must 
consult your CEMS user's manual and other reference materials 
recommended by the reference method for specific precautions to be 
taken.
6.0 What Equipment and Supplies Do I Need?
    Different types of PM CEMS use different operating principles. You 
should select an appropriate PM CEMS based on your site-specific 
configurations, flue gas conditions, and PM characteristics.
    (1) Your PM CEMS must sample the stack effluent continuously or, 
for batch sampling PM CEMS, intermittently.
    (2) You must ensure that the averaging time, the number of 
measurements in an average, the minimum data availability, and the 
averaging procedure for your CEMS conform with those specified in the 
applicable emission regulation.
    (3) Your PM CEMS must include, as a minimum, the equipment 
described in sections 6.1 through 6.7.
    6.1 What equipment is needed for my PM CEMS's sample interface? 
Your PM CEMS's sample interface must be capable of delivering a 
representative sample of the flue gas to your PM CEMS. This subsystem 
may be required to heat the sample gas to avoid PM deposition or 
moisture condensation, provide dilution air, perform other gas 
conditioning to prepare the sample for analysis, or measure the sample 
volume or flow rate.
    (1) If your PM CEMS is installed downstream of a wet air pollution 
control system such that the flue gases normally or occasionally 
contain entrained water droplets, we recommend that you select a 
sampling system that includes equipment to extract and heat a 
representative sample of the flue gas for measurement so that the 
pollutant monitor portion of your CEMS measures only dry PM. Heating 
should be sufficient to raise the temperature of the extracted flue gas 
above the water condensation temperature and should be maintained at 
all times and at all points in the sample line from where the flue gas 
is extracted, including the pollutant monitor and any sample flow 
measurement devices.

[[Page 1806]]

    (2) You must consider the measured conditions of the sample gas 
stream to ensure that manual reference method test data are converted 
to units of PM concentration that are appropriate for the correlation 
calculations. Additionally, you must identify what, if any, additional 
auxiliary data from other monitoring and handling systems are necessary 
to convert your PM CEMS response into the units of the PM standard.
    (3) If your PM CEMS is an extractive type and your source's flue 
gas volumetric flow rate varies by more than 10 percent from nominal, 
your PM CEMS should maintain an isokinetic sampling rate (within 10 
percent of true isokinetic). If your extractive-type PM CEMS does not 
maintain an isokinetic sampling rate, you must use actual site-specific 
data or data from a similar installation to prove to us, the State, 
and/or local enforcement agency that isokinetic sampling is not 
necessary.
    6.2 What type of equipment is needed for my PM CEMS? Your PM CEMS 
must be capable of providing an electronic output that can be 
correlated to the PM concentration.
    (1) Your PM CEMS must be able to perform zero and upscale drift 
checks. You may perform these checks manually, but performing these 
checks automatically is preferred.
    (2) We recommend that you select a PM CEMS that is capable of 
performing automatic diagnostic checks and sending instrument status 
signals (flags) to the data recorder.
    (3) If your PM CEMS is an extractive type that measures the sample 
volume and uses the measured sample volume as part of calculating the 
output value, your PM CEMS must be able to perform a check of the 
sample volume to verify the accuracy of the sample volume measuring 
equipment. The sample volume check must be conducted daily and at the 
normal sampling rate of your PM CEMS.
    6.3 What is the appropriate measurement range for my PM CEMS? 
Initially, your PM CEMS must be set up to measure over the expected 
range of your source's PM emission concentrations during routine 
operations. You may change the measurement range to a more appropriate 
range prior to correlation testing.
    6.4 What if my PM CEMS does automatic range switching? Your PM CEMS 
may be equipped to perform automatic range switching so that it is 
operating in a range most sensitive to the detected concentrations. If 
your PM CEMS does automatic range switching, you must configure the 
data recorder to handle the recording of data values in multiple ranges 
during range-switching intervals.
    6.5 What averaging time and sample intervals should be used? Your 
CEMS must sample the stack effluent such that the averaging time, the 
number of measurements in an average, the minimum sampling time, and 
the averaging procedure for reporting and determining compliance 
conform with those specified in the applicable regulation. Your PM CEMS 
must be designed to meet the specified response time and cycle time 
established in this performance specification (see section 13.3).
    6.6 What type of equipment is needed for my data recorder? Your 
CEMS data recorder must be able to accept and record electronic signals 
from all the monitors associated with your PM CEMS.
    (1) Your data recorder must record the signals from your PM CEMS 
that can be correlated to PM mass concentrations. If your PM CEMS uses 
multiple ranges, your data recorder must identify what range the 
measurement was made in and provide range-adjusted results.
    (2) Your data recorder must accept and record monitor status 
signals (flagged data).
    (3) Your data recorder must accept signals from auxiliary data 
monitors, as appropriate.
    6.7 What other equipment and supplies might I need? You may need 
other supporting equipment as defined by the applicable reference 
method(s) (see section 7) or as specified by your CEMS manufacturer.
7.0 What Reagents and Standards Do I Need?
    You will need reference standards or procedures to perform the zero 
drift check, the upscale drift check, and the sample volume check.
    7.1 What is the reference standard value for the zero drift check? 
You must use a zero check value that is no greater than 20 percent of 
the PM CEMS's response range. You must obtain documentation on the zero 
check value from your PM CEMS manufacturer.
    7.2 What is the reference standard value for the upscale drift 
check? You must use an upscale check value that produces a response 
between 50 and 100 percent of the PM CEMS's response range. For a PM 
CEMS that produces output over a range of 4 mA to 20 mA, the upscale 
check value must produce a response in the range of 12 mA to 20 mA. You 
must obtain documentation on the upscale check value from your PM CEMS 
manufacturer.
    7.3 What is the reference standard value for the sample volume 
check? You must use a reference standard value or procedure that 
produces a sample volume value equivalent to the normal sampling rate. 
You must obtain documentation on the sample volume value from your PM 
CEMS manufacturer.
8.0 What Performance Specification Test Procedure Do I Follow?
    You must complete each of the activities in sections 8.1 through 
8.8 for your performance specification test.
    8.1 How should I select and set up my equipment? You should select 
a PM CEMS that is appropriate for your source, giving consideration to 
potential factors such as flue gas conditions, interferences, site-
specific configuration, installation location, PM concentration range, 
and other PM characteristics. Your PM CEMS must meet the equipment 
specifications in sections 6.1 and 6.2.
    (1) You should select a PM CEMS that is appropriate for the flue 
gas conditions at your source. If your source's flue gas contains 
entrained water droplets, we recommend that your PM CEMS include a 
sample delivery and conditioning system that is capable of extracting 
and heating a representative sample.
    (i) Your PM CEMS must maintain the sample at a temperature 
sufficient to prevent moisture condensation in the sample line before 
analysis of PM.
    (ii) If condensible PM is an issue, we recommend that you operate 
your PM CEMS to maintain the sample gas temperature at the same 
temperature as the reference method filter.
    (iii) Your PM CEMS must avoid condensation in the sample flow rate 
measurement lines.
    (2) Some PM CEMS do not have a wide measurement range capability. 
Therefore, you must select a PM CEMS that is capable of measuring the 
full range of PM concentrations expected from your source from normal 
levels through the emission limit concentration.
    (3) Some PM CEMS are sensitive to particle size changes, water 
droplets in the gas stream, particle charge, stack gas velocity 
changes, or other factors. Therefore, you should select a PM CEMS 
appropriate for the emission characteristics of your source.
    (4) We recommend that you consult your PM CEMS vendor to obtain 
basic recommendations on the instrument capabilities and setup 
configuration. You are ultimately responsible for setup and operation 
of your PM CEMS.
    8.2 Where do I install my PM CEMS? You must install your PM CEMS

[[Page 1807]]

at an accessible location downstream of all pollution control 
equipment. You must perform your PM CEMS concentration measurements 
from a location considered representative or be able to provide data 
that can be corrected to be representative of the total PM emissions as 
determined by the manual reference method.
    (1) You must select a measurement location that minimizes problems 
due to flow disturbances, cyclonic flow, and varying PM stratification 
(refer to Method 1 for guidance).
    (2) If you plan to achieve higher emissions for correlation test 
purposes by adjusting the performance of the air pollution control 
device (per section 8.6(4)(i)), you must locate your PM CEMS and 
reference method sampling points well downstream of the control device 
(e.g., downstream of the induced draft fan), in order to minimize PM 
stratification that may be created in these cases.
    8.3 How do I select the reference method measurement location and 
traverse points? You must follow EPA Method 1 for identifying manual 
reference method traverse points. Ideally, you should perform your 
manual reference method measurements at locations that satisfy the 
measurement site selection criteria specified in EPA Method 1 of at 
least eight duct diameters downstream and at least two duct diameters 
upstream of any flow disturbance. Where necessary, you may conduct 
testing at a location that is two diameters downstream and 0.5 
diameters upstream of flow disturbances. If your location does not meet 
the minimum downstream and upstream requirements, you must obtain 
approval from us to test at your location.
    8.4 What are my pretest preparation steps? You must install your 
CEMS and prepare the reference method test site according to the 
specifications in sections 8.2 and 8.3.
    (1) After completing the initial field installation, we recommend 
that you operate your PM CEMS according to the manufacturer's 
instructions to familiarize yourself with its operation before you 
begin correlation testing.
    (i) During this initial period of operation, we recommend that you 
conduct daily checks (zero and upscale drift and sample volume, as 
appropriate), and, when any check exceeds the daily specification (see 
section 13.1), make adjustments and perform any necessary maintenance 
to ensure reliable operation.
    (2) When you are confident that your PM CEMS is operating properly, 
we recommend that you operate your CEMS over a correlation test 
planning period of sufficient duration to identify the full range of 
operating conditions and PM emissions to be used in your PM CEMS 
correlation test.
    (i) During the correlation test planning period, you should operate 
the process and air pollution control equipment over the normal range 
of operating conditions, except when you attempt to produce higher 
emissions.
    (ii) Your data recorder should record PM CEMS response during the 
full range of routine process operating conditions.
    (iii) You should try to establish the relationships between 
operating conditions and PM CEMS response, especially those conditions 
that produce the highest PM CEMS response over 15-minute averaging 
periods, and the lowest PM CEMS response as well. The objective is to 
be able to reproduce the conditions for purposes of the actual 
correlation testing discussed in section 8.6.
    (3) You must set the response range of your PM CEMS such that the 
instrument measures the full range of responses that correspond to the 
range of source operating conditions that you will implement during 
correlation testing.
    (4) We recommend that you perform preliminary reference method 
testing after the correlation test planning period. During this 
preliminary testing, you should measure the PM emission concentration 
corresponding to the highest PM CEMS response observed during the full 
range of normal operation, when perturbing the control equipment, or as 
the result of PM spiking.
    (5) Before performing correlation testing, you must perform a 7-day 
zero and upscale drift test (see section 8.5).
    (6) You must not change the response range of the monitor once the 
response range has been set and the drift test successfully completed.
    8.5 How do I perform the 7-day drift test? You must check the zero 
(or low-level value between 0 and 20 percent of the response range of 
the instrument) and upscale (between 50 and 100 percent of the 
instrument's response range) drift. You must perform this check at 
least once daily over 7 consecutive days. Your PM CEMS must quantify 
and record the zero and upscale measurements and the time of the 
measurements. If you make automatic or manual adjustments to your PM 
CEMS zero and upscale settings, you must conduct the drift test 
immediately before these adjustments, or conduct it in such a way that 
you can determine the amount of drift. You will find the calculation 
procedures for drift in section 12.1 and the acceptance criteria for 
allowable drift in section 13.1.
    (1) What is the purpose of 7-day drift tests? The purpose of the 7-
day drift test is to demonstrate that your system is capable of 
operating in a stable manner and maintaining its calibration for at 
least a 7-day period.
    (2) How do I conduct the 7-day drift test? To conduct the 7-day 
drift test, you must determine the magnitude of the drift once each 
day, at 24-hour intervals, for 7 consecutive days while your source is 
operating normally.
    (i) You must conduct the 7-day drift test at the two points 
specified in section 8.5. You may perform the 7-day drift tests 
automatically or manually by introducing to your PM CEMS suitable 
reference standards (these need not be certified) or by using other 
appropriate procedures.
    (ii) You must record your PM CEMS zero and upscale response and 
evaluate them against the zero check value and upscale check value.
    (3) When must I conduct the 7-day drift test? You must complete a 
valid 7-day drift test before attempting the correlation test.
    8.6 How do I conduct my PM CEMS correlation test? You must conduct 
the correlation test according to the procedure given in paragraphs (1) 
through (5) of this section. If you need multiple correlations, you 
must conduct sufficient testing and collect at least 15 pairs of 
reference method and PM CEMS data for calculating each separate 
correlation.
    (1) You must use the reference method for PM (usually EPA Methods 
5, 5I, or 17) that is prescribed by the applicable regulations. You may 
need to perform other reference methods or performance specifications 
(e.g., Method 3 for oxygen, Method 4 for moisture, etc.) depending on 
the units in which your PM CEMS reports PM concentration.
    (i) We recommend that you use paired reference method trains when 
collecting manual PM data to identify and screen the reference method 
data for imprecision and bias. Procedures for checking reference method 
data for bias and precision can be found in the PM CEMS Knowledge 
Document (see section 16.5).
    (ii) You may use test runs that are shorter than 60 minutes in 
duration (e.g., 20 or 30 minutes). You may perform your PM CEMS 
correlation tests during new source performance standards performance 
tests or other compliance tests subject to the Clean Air Act or other 
statutes, such as the Resource Conservation and Recovery Act. In these 
cases, your reference

[[Page 1808]]

method results obtained during the PM CEMS correlation test may be used 
to determine compliance so long as your source and the test conditions 
and procedures (e.g., reference method sample run durations) are 
consistent with the applicable regulations and the reference method.
    (iii) You must convert the reference method results to units 
consistent with the conditions of your PM CEMS measurements. For 
example, if your PM CEMS measures and reports PM emissions in the units 
of mass per actual volume of stack gas, you must convert your reference 
method results to those units (e.g., mg/acm). If your PM CEMS extracts 
and heats the sample gas to eliminate water droplets, then measures and 
reports PM emissions under those actual conditions, you must convert 
your reference method results to those same conditions (e.g., mg/acm at 
160[deg]C).
    (2) During each test run, you must coordinate process operations, 
reference method sampling, and PM CEMS operations. For example, you 
must ensure that the process is operating at the targeted conditions, 
both reference method trains are sampling simultaneously (if paired 
sampling trains are being used), and your PM CEMS and data logger are 
operating properly.
    (i) You must coordinate the start and stop times of each run 
between the reference method sampling and PM CEMS operation. For a 
batch sampling PM CEMS, you must start the reference method at the same 
time as your PM CEMS sampling.
    (ii) You must note the times for port changes (and other periods 
when the reference method sampling may be suspended) on the data sheets 
so that you can adjust your PM CEMS data accordingly, if necessary.
    (iii) You must properly align the time periods for your PM CEMS and 
your reference method measurements to account for your PM CEMS response 
time.
    (3) You must conduct a minimum of 15 valid runs each consisting of 
simultaneous PM CEMS and reference method measurement sets.
    (i) You may conduct more than 15 sets of CEMS and reference method 
measurements. If you choose this option, you may reject certain test 
results so long as the total number of valid test results you use to 
determine the correlation is greater than or equal to 15.
    (ii) You must report all data, including the rejected data.
    (iii) You may reject the results of up to five test runs without 
explanation.
    (iv) If you reject the results of more than five test runs, the 
basis for rejecting the results of the additional test runs must be 
explicitly stated in the reference method, this performance 
specification, Procedure 2 of appendix F, or your quality assurance 
plan.
    (4) Simultaneous PM CEMS and reference method measurements must be 
performed in a manner to ensure that the range of data that will be 
used to establish the correlation for your PM CEMS is maximized. You 
must first attempt to maximize your correlation range by following the 
procedures described in paragraphs (4)(i) through (iv) of this section. 
If you cannot obtain the three levels as described in paragraphs (i) 
through (iv), then you must use the procedure described in section 
8.6(5).
    (i) You must attempt to obtain the three different levels of PM 
mass concentration by varying process operating conditions, varying PM 
control device conditions, or by means of PM spiking.
    (ii) The three PM concentration levels you use in the correlation 
tests must be distributed over the complete operating range experienced 
by your source.
    (iii) At least 20 percent of the minimum 15 measured data points 
you use should be contained in each of the following levels:
    [sbull] Level 1: From no PM (zero concentration) emissions to 50 
percent of the maximum PM concentration;
    [sbull] Level 2: 25 to 75 percent of the maximum PM concentration; 
and
    [sbull] Level 3: 50 to 100 percent of the maximum PM concentration.
    (iv) Although the above levels overlap, you may only apply 
individual run data to one level.
    (5) If you cannot obtain three distinct levels of PM concentration 
as described, you must perform correlation testing over the maximum 
range of PM concentrations that is practical for your PM CEMS. To 
ensure that the range of data used to establish the correlation for 
your PM CEMS is maximized, you must follow one or more of the steps in 
paragraphs (5)(i) through (iv) of this section.
    (i) Zero point data for in-situ instruments should be obtained, to 
the extent possible, by removing the instrument from the stack and 
monitoring ambient air on a test bench.
    (ii) Zero point data for extractive instruments should be obtained 
by removing the extractive probe from the stack and drawing in clean 
ambient air.
    (iii) Zero point data also can be obtained by performing manual 
reference method measurements when the flue gas is free of PM emissions 
or contains very low PM concentrations (e.g., when your process is not 
operating, but the fans are operating or your source is combusting only 
natural gas).
    (iv) If none of the steps in paragraphs (5)(i) through (iii) of 
this section are possible, you must estimate the monitor response when 
no PM is in the flue gas (e.g., 4 mA = 0 mg/acm).
    8.7 What do I do with the initial correlation test data for my PM 
CEMS? You must calculate and report the results of the correlation 
testing, including the correlation coefficient, confidence interval, 
and tolerance interval for the PM CEMS response and reference method 
correlation data that are use to establish the correlation, as 
specified in section 12. You must include all data sheets, 
calculations, charts (records of PM CEMS responses), process data 
records including PM control equipment operating parameters, and 
reference media certifications necessary to confirm that your PM CEMS 
met the requirements of this performance specification. In addition, 
you must:
    (1) Determine the integrated (arithmetic average) PM CEMS output 
over each reference method test period;
    (2) Adjust your PM CEMS outputs and reference method test data to 
the same clock time (considering response time of your PM CEMS);
    (3) Confirm that the reference method results are consistent with 
your PM CEMS response in terms of, where applicable, moisture, 
temperature, pressure, and diluent concentrations; and
    (4) Determine whether any of the reference method test results do 
not meet the test method criteria.
    8.8 What is the limitation on the range of my PM CEMS correlation? 
Although the data you collect during the correlation testing should be 
representative of the full range of normal operating conditions at your 
source, you must conduct additional correlation testing if either of 
the conditions specified in paragraphs (1) and (2) of this section 
occurs.
    (1) If your source is a low-emitting source, as defined in section 
3.16 of this specification, you must conduct additional correlation 
testing if either of the events specified in paragraphs (1)(i) or (ii) 
of this section occurs while your source is operating under normal 
conditions.
    (i) Your source generates 24 consecutive hourly average PM CEMS 
responses that are greater than 125 percent of the highest PM CEMS 
response (e.g., mA reading) used for the correlation curve or are 
greater than the

[[Page 1809]]

PM CEMS response that corresponds to 50 percent of the emission limit, 
whichever is greater, or
    (ii) The cumulative hourly average PM CEMS responses generated by 
your source are greater than 125 percent of the highest PM CEMS 
response used for the correlation curve or are greater than the PM CEMS 
response that corresponds to 50 percent of the emission limit, 
whichever is greater, for more than 5 percent of your PM CEMS operating 
hours for the previous 30-day period.
    (2) If your source is not a low-emitting source, as defined in 
section 3.16 of this specification, you must conduct additional 
correlation testing if either of the events specified in paragraph (i) 
or (ii) of this section occurs while your source is operating under 
normal conditions.
    (i) Your source generates 24 consecutive hourly average PM CEMS 
responses that are greater than 125 percent of the highest PM CEMS 
response (e.g., mA reading) used for the correlation curve, or
    (ii) The cumulative hourly average PM CEMS responses generated by 
your source are greater than 125 percent of the highest PM CEMS 
response used for the correlation curve for more than 5 percent of your 
PM CEMS operating hours for the previous 30-day period.
    (3) If additional correlation testing is required, you must conduct 
at least three additional test runs under the conditions that caused 
the higher PM CEMS response.
    (i) You must complete the additional testing and use the resulting 
new data along with the previous data to calculate a revised 
correlation equation within 60 days after the occurrence of the event 
that requires additional testing, as specified in paragraphs 8.8(1) and 
(2).
    (4) If your source generates consecutive PM CEMS hourly responses 
that are greater than 125 percent of the highest PM CEMS response used 
to develop the correlation curve for 24 hours or for a cumulative 
period that amounts to more than 5 percent of the PM CEMS operating 
hours for the previous 30-day period, you must report the reason for 
the higher PM CEMS responses.
9.0 What Quality Control Measures Are Required?
    Quality control measures for PM CEMS are specified in 40 CFR 60, 
Appendix F, Procedure 2.
10.0 What Calibration and Standardization Procedures Must I Perform? 
[Reserved]
11.0 What Analytical Procedures Apply to This Procedure?
    Specific analytical procedures are outlined in the applicable 
reference method(s).
12.0 What Calculations and Data Analyses Are Needed?
    You must determine the primary relationship for correlating the 
output from your PM CEMS to a PM concentration, typically in units of 
mg/acm or mg/dscm of flue gas, using the calculations and data analysis 
process in sections 12.2 and 12.3. You develop the correlation by 
performing an appropriate regression analysis between your PM CEMS 
response and your reference method data.
    12.1 How do I calculate upscale drift and zero drift? You must 
determine the difference in your PM CEMS output readings from the 
established reference values (zero and upscale check values) after a 
stated period of operation during which you performed no unscheduled 
maintenance, repair, or adjustment.
    (1) Calculate the upscale drift (UD) using Equation 11-1:
    [GRAPHIC] [TIFF OMITTED] TR12JA04.003
    
Where:

UD = The upscale (high-level) drift of your PM CEMS in percent,
RCEM = The measured PM CEMS response to the upscale 
reference standard, and
RU = The preestablished numerical value of the upscale 
reference standard.

    (2) Calculate the zero drift (ZD) using Equation 11-2:
    [GRAPHIC] [TIFF OMITTED] TR12JA04.004
    
Where:

ZD = The zero (low-level) drift of your PM CEMS in percent,
RCEM = The measured PM CEMS response to the zero reference 
standard,
RL = The preestablished numerical value of the zero 
reference standard, and
RU = The preestablished numerical value of the upscale 
reference standard.

    (3) Summarize the results on a data sheet similar to that shown in 
Table 2 (see section 17).
    12.2 How do I perform the regression analysis? You must couple each 
reference method PM concentration measurement, y, in the appropriate 
units, with an average PM CEMS response, x, over corresponding time 
periods. You must complete your PM CEMS correlation calculations using 
data deemed acceptable by quality control procedures identified in 40 
CFR 60, Appendix F, Procedure 2.
    (1) You must evaluate all flagged or suspect data produced during 
measurement periods and determine whether they should be excluded from 
your PM CEMS's average.
    (2) You must assure that the reference method and PM CEMS results 
are on a consistent moisture, temperature, and diluent basis. You must 
convert the reference method PM concentration measurements (dry 
standard conditions) to the units of your PM CEMS measurement 
conditions. The conditions of your PM CEMS measurement are monitor-
specific. You must obtain from your PM CEMS vendor or instrument 
manufacturer the conditions and units of measurement for your PM CEMS.
    (i) If your sample gas contains entrained water droplets and your 
PM CEMS is an extractive system that measures at actual conditions 
(i.e., wet basis), you must use the measured moisture content 
determined from the impinger analysis when converting your reference 
method PM data to PM CEMS conditions; do not use the moisture content 
calculated from a psychrometric chart based on saturated conditions.
    12.3 How do I determine my PM CEMS correlation? To predict PM 
concentrations from PM CEMS responses, you must use the calculation 
method of least squares presented in paragraphs (1) through (5) of this 
section. When performing the calculations, each reference method PM 
concentration measurement must be treated as a discrete data point; if 
using paired sampling trains, do not average reference method data 
pairs for any test run.
    This performance specification describes procedures for evaluating 
five types of correlation models: linear, polynomial, logarithmic, 
exponential, and power. Procedures for selecting the most appropriate 
correlation model are presented in section 12.4 of this specification.
    (1) How do I evaluate a linear correlation for my correlation test 
data? To evaluate a linear correlation, follow the procedures described 
in paragraphs (1)(i) through (iv) of this section.
    (i) Calculate the linear correlation equation, which gives the 
predicted PM concentration ([ycirc]) as a function of the

[[Page 1810]]

PM CEMS response (x), as indicated by Equation 11-3:
[GRAPHIC] [TIFF OMITTED] TR12JA04.005

Where:

[ycirc] = the predicted PM concentration,
b0 = the intercept for the correlation curve, as calculated 
using Equation 11-4,
b1 = the slope of the correlation curve, as calculated using 
Equation 11-6, and
x = the PM CEMS response value.

    Calculate the y intercept (b0) of the correlation curve 
using Equation 11-4:
[GRAPHIC] [TIFF OMITTED] TR12JA04.006

Where:

x = the mean value of the PM CEMS response data, as calculated using 
Equation 11-5, and
y = the mean value of the PM concentration data, as calculated using 
Equation 11-5:
[GRAPHIC] [TIFF OMITTED] TR12JA04.007

Where:

xi = the PM CEMS response value for run i,
yi = the PM concentration value for run i, and
n = the number of data points.

    Calculate the slope (b1) of the correlation curve using 
Equation 11-6:
[GRAPHIC] [TIFF OMITTED] TR12JA04.008

Where:

Sxx, Sxy = as calculated using Equation 11-7:
[GRAPHIC] [TIFF OMITTED] TR12JA04.009

    (ii) Calculate the half range of the 95 percent confidence interval 
(CI) for the predicted PM concentration (y) at the mean value of x, 
using Equation 11-8:
[GRAPHIC] [TIFF OMITTED] TR12JA04.010

Where:

CI = the half range for the 95 percent confidence interval for the mean 
x value,
tdf,1-a/2 = the value for the t statistic provided in Table 
1 for df = n-2, and
SL = the scatter or deviation of [ycirc] values about the 
correlation curve, which is determined using Equation 11-9:
[GRAPHIC] [TIFF OMITTED] TR12JA04.011

    Calculate the confidence interval half range at the mean x value as 
a percentage of the emission limit (CI%) using Equation 11-10:
[GRAPHIC] [TIFF OMITTED] TR12JA04.012

Where:

CI = the confidence interval half range at the mean x value, and
EL = PM emission limit, as described in section 13.2.

    (iii) Calculate the half range of the tolerance interval at the 
mean x value (TI) using Equation 11-11:
[GRAPHIC] [TIFF OMITTED] TR12JA04.013

Where:

TI = the tolerance interval half range at the mean x value,
kt = as calculated using Equation 11-12, and
SL = as calculated using Equation 11-9:
[GRAPHIC] [TIFF OMITTED] TR12JA04.014

Where:

n[min] = the number of test runs (n),
un[min] = the tolerance factor for 75 percent provided in 
Table 1, and
vdf = the value from Table 1 for df = n-2.
    Calculate the tolerance interval half range at the mean x value as 
a percentage of the emission limit (TI%) using Equation 11-13:
[GRAPHIC] [TIFF OMITTED] TR12JA04.015

Where:

TI = the tolerance interval half range at the mean value of x, and
EL = PM emission limit, as described in section 13.2.

    (iv) Calculate the linear correlation coefficient (r) using 
Equation 11-14:
[GRAPHIC] [TIFF OMITTED] TR12JA04.016

Where:

SL = as calculated using Equation 11-9, and
Sy = as calculated using Equation 11-15:
[GRAPHIC] [TIFF OMITTED] TR12JA04.017

    (2) How do I evaluate a polynomial correlation for my correlation 
test data? To evaluate a polynomial correlation, follow the procedures 
described in paragraphs (2)(i) through (iv) of this section.
    (i) Calculate the polynomial correlation equation, which is 
indicated by Equation 11-16, using Equations 11-17 through 11-22:
[GRAPHIC] [TIFF OMITTED] TR12JA04.018

Where:

[ycirc] = the PM CEMS concentration predicted by the polynomial 
correlation equation, and
b0, b1, b2 = the coefficients 
determined from the solution to the matrix equation Ab=B where:
[GRAPHIC] [TIFF OMITTED] TR12JA04.019


[[Page 1811]]


[GRAPHIC] [TIFF OMITTED] TR12JA04.020

Where:

xi = the PM CEMS response for run i,
yi = the reference method PM concentration for run i, and
n = the number of test runs.

    Calculate the polynomial correlation curve coefficients 
(b0, b1, and b2) using Equations 11-19 
to 11-21, respectively:
[GRAPHIC] [TIFF OMITTED] TR12JA04.021

[GRAPHIC] [TIFF OMITTED] TR12JA04.022

Where:
[GRAPHIC] [TIFF OMITTED] TR12JA04.023

    (ii) Calculate the confidence interval half range (CI) by first 
calculating the C coefficients (C0 to C5) using 
Equations 11-23 and 11-24:
Where:
[GRAPHIC] [TIFF OMITTED] TR12JA04.024

Where:
[GRAPHIC] [TIFF OMITTED] TR12JA04.025

    Calculate [Delta] using Equation 11-25 for each x value:
    [GRAPHIC] [TIFF OMITTED] TR12JA04.026
    
    Determine the x value that corresponds to the minimum value of 
[Delta] ([Delta]min). Determine the scatter or deviation of 
[ycirc] values about the polynomial correlation curve (SP) 
using Equation 11-26:
[GRAPHIC] [TIFF OMITTED] TR12JA04.027

    Calculate the half range of the 95 percent confidence interval (CI) 
at the x value that corresponds to [Delta]min using Equation 
11-27:
[GRAPHIC] [TIFF OMITTED] TR12JA04.028

Where:

df = n -3, and
tdf = as listed in Table 1 (see section 17).

    Calculate the confidence interval half range at the x value for 
[Delta]min as a percentage of the emission limit (CI%) using 
Equation 11-28:
[GRAPHIC] [TIFF OMITTED] TR12JA04.029

Where:

CI = the confidence interval half range at the x value that corresponds 
to [Delta]min, and
EL = PM emission limit, as described in section 13.2.

    (iii) Calculate the tolerance interval half range (TI) at the x 
value for [Delta]min, as indicated in Equation 11-29 for the 
polynomial correlation, using Equations 11-30 and 11-31:
[GRAPHIC] [TIFF OMITTED] TR12JA04.030

Where:

[[Page 1812]]

[GRAPHIC] [TIFF OMITTED] TR12JA04.031

un' = the value indicated in Table 1, and
vdf = the value indicated in Table 1 for df = n-3.

    If the calculated value for n is less than 2, then n = 2.
    Calculate the tolerance interval half range at the x value for 
[Delta]min as a percentage of the emission limit (TI%) using 
Equation 11-32:
[GRAPHIC] [TIFF OMITTED] TR12JA04.032

Where:

TI = the tolerance interval half range at the x value that corresponds 
to [Delta]min, and
EL = PM emission limit, as described in section 13.2.

    (iv) Calculate the polynomial correlation coefficient (r) using 
Equation 11-33:
[GRAPHIC] [TIFF OMITTED] TR12JA04.033

Where:

SP = as calculated using Equation 11-26, and
Sy = as calculated using Equation 11-15.

    (3) How do I evaluate a logarithmic correlation for my correlation 
test data? To evaluate a logarithmic correlation, which has the form 
indicated by Equation 11-34, follow the procedures described in 
paragraphs (3)(i) through (iii) of this section.
[GRAPHIC] [TIFF OMITTED] TR12JA04.034

    (i) Perform a logarithmic transformation of each PM CEMS response 
value (x values) using Equation 11-35:
[GRAPHIC] [TIFF OMITTED] TR12JA04.035

Where:

xi' = is the transformed value of xi, and
Ln(xi) = the natural logarithm of the PM CEMS response for 
run i.

    (ii) Using the values for xi' in place of the values for 
xi, perform the same procedures used to develop the linear 
correlation equation described in paragraph (1)(i) of this section. The 
resulting equation has the form indicated by Equation 11-36:
[GRAPHIC] [TIFF OMITTED] TR12JA04.036


Where:

x' = the natural logarithm of the PM CEMS response, and the variables 
[ycirc], b0, and b1 are as defined in paragraph 
(1)(i) of this section.

    (iii) Using the values for xi' in place of the values 
for xi, calculate the confidence interval half range at the 
mean x' value as a percentage of the emission limit (CI%), the 
tolerance interval half range at the mean x' value as a percentage of 
the emission limit (TI%), and the correlation coefficient (r) using the 
procedures described in paragraphs (1)(ii) through (iv) of this 
section.
    (4) How do I evaluate an exponential correlation for my correlation 
test data? To evaluate an exponential correlation, which has the form 
indicated by Equation 11-37, follow the procedures described in 
paragraphs (4)(i) through (v) of this section:
[GRAPHIC] [TIFF OMITTED] TR12JA04.037

    (i) Perform a logarithmic transformation of each PM concentration 
measurement (y values) using Equation 11-38:
[GRAPHIC] [TIFF OMITTED] TR12JA04.038

Where:

yi' = is the transformed value of yi, and
Ln(yi) = the natural logarithm of the PM concentration 
measurement for run i.

    (ii) Using the values for yi in place of the values for 
yi' perform the same procedures used to develop the linear 
correlation equation described in paragraph (1)(i) of this section. The 
resulting equation will have the form indicated by Equation 11-39.
[GRAPHIC] [TIFF OMITTED] TR12JA04.039

Where:

[ycirc]i' = the natural logarithm of the predicted PM 
concentration values, and the variables b0, b1, 
and x are as defined in paragraph (1)(i) of this section.

    (iii) Using the values for yi' in place of the values 
for yi, calculate the confidence interval half range (CI), 
as described in paragraph (1)(ii) of this section. However, for the 
exponential correlation, you must calculate the value for CI at the 
median x value, instead of the mean x value for linear correlations. 
Calculate the confidence interval half range at the median x value as a 
percentage of the emission limit (CI%) using Equation 11-40:
[GRAPHIC] [TIFF OMITTED] TR12JA04.040

Where:

CI = the confidence interval half range at the median x value, and
Ln(EL) = the natural logarithm of the PM emission limit, as described 
in section 13.2.

    (iv) Using the values for yi' in place of the values for 
yi, calculate the tolerance interval half range (TI), as 
described in paragraph (1)(iii) of this section. For the exponential 
correlation, the value for TI also must be calculated at the median x 
value. Calculate the tolerance interval half range at the median x 
value as a percentage of the emission limit (TI%) using Equation 11-41:
[GRAPHIC] [TIFF OMITTED] TR12JA04.041

Where:

TI = the tolerance interval half range at the median x value, and
Ln(EL) = the natural logarithm of the PM emission limit, as described 
in section 13.2.

    (v) Using the values for yi' in place of the values for 
yi, calculate the correlation coefficient (r) using the 
procedure described in paragraph (1)(iv) of this section.
    (5) How do I evaluate a power correlation for my correlation test 
data? To evaluate a power correlation, which has the form indicated by 
Equation 11-42, follow the procedures described in paragraphs (5)(i) 
through (v) of this section.
[GRAPHIC] [TIFF OMITTED] TR12JA04.042

    (i) Perform logarithmic transformations of each PM CEMS response (x 
values) and each PM concentration measurement (y values) using 
Equations 11-35 and 11-38, respectively.
    (ii) Using the values for xi' in place of the values for 
xi, and the values for yi' in place of the values 
for yi, perform the same procedures used to develop the 
linear correlation equation described in paragraph (1)(i) of this 
section. The resulting equation will have the form indicated by 
Equation 11-43:
[GRAPHIC] [TIFF OMITTED] TR12JA04.043

Where:

[ycirc]' = the natural logarithm of the predicted PM concentration 
values, and
x' = the natural logarithm of the PM CEMS response values, and the 
variables b0 and b1 are as defined in paragraph 
(1)(i) of this section.


    (iii) Using the values for yi' in place of the values 
for yi, calculate the confidence interval half range (CI), 
as

[[Page 1813]]

described in paragraph (1)(ii) of this section. You must calculate the 
value for CI at the median x' value, instead of the mean x value for 
linear correlations. Calculate the confidence interval half range at 
the median x' value as a percentage of the emission limit (CI%) using 
Equation 11-40.
    (iv) Using the values foryi, in place of the values for 
yi, calculate the tolerance interval half range (TI), as 
described in paragraph (1)(iii) of this section. The value for TI also 
must be calculated at the median x' value. Calculate the tolerance 
interval half range at the median x' value as a percentage of the 
emission limit (CI%) using Equation 11-41.
    (v) Using the values for yi' in place of the values for 
yi, calculate the correlation coefficient (r) using the 
procedure described in paragraph (1)(iv) of this section.
    12.4 Which correlation model should I use? Follow the procedures 
described in paragraphs (1) through (4) of this section to determine 
which correlation model you should use.
    (1) For each correlation model that you develop using the 
procedures described in section 12.3 of this specification, compare the 
confidence interval half range percentage, tolerance interval half 
range percentage, and correlation coefficient to the performance 
criteria specified in section 13.2 of this specification. You can use 
the linear, logarithmic, exponential, or power correlation model if the 
model satisfies all of the performance criteria specified in section 
13.2 of this specification. However, to use the polynomial model you 
first must check that the polynomial correlation curve satisfies the 
criteria for minimum and maximum values specified in paragraph (3) of 
this section.
    (2) If you develop more than one correlation curve that satisfy the 
performance criteria specified in section 13.2 of this specification, 
you should use the correlation curve with the greatest correlation 
coefficient. If the polynomial model has the greatest correlation 
coefficient, you first must check that the polynomial correlation curve 
satisfies the criteria for minimum and maximum values specified in 
paragraph (3) of this section.
    (3) You can use the polynomial model that you develop using the 
procedures described in section 12.3(2) if the model satisfies the 
performance criteria specified in section 13.2 of this specification, 
and the minimum or maximum value of the polynomial correlation curve 
does not occur within the expanded data range. The minimum or maximum 
value of the polynomial correlation curve is the point where the slope 
of the curve equals zero. To determine if the minimum or maximum value 
occurs within the expanded data range, follow the procedure described 
in paragraphs (3)(i) through (iv) of this section.
    (i) Determine if your polynomial correlation curve has a minimum or 
maximum point by comparing the polynomial coefficient b2 to 
zero. If b2 is less than zero, the curve has a maximum 
value. If b2 is greater than zero, the curve has a minimum 
value. (Note: If b2 equals zero, the correlation curve is 
linear.)
    (ii) Calculate the minimum value using Equation 11-44.
    [GRAPHIC] [TIFF OMITTED] TR12JA04.044
    
    (iii) If your polynomial correlation curve has a minimum point, you 
must compare the minimum value to the minimum PM CEMS response used to 
develop the correlation curve. If the correlation curve minimum value 
is less than or equal to the minimum PM CEMS response value, you can 
use the polynomial correlation curve, provided the correlation curve 
also satisfies all of the performance criteria specified in section 
13.2 of this specification. If the correlation curve minimum value is 
greater than the minimum PM CEMS response value, you cannot use the 
polynomial correlation curve to predict PM concentrations.
    (iv) If your polynomial correlation curve has a maximum, the 
maximum value must be greater than the allowable extrapolation limit. 
If your source is not a low-emitting source, as defined in section 3.16 
of this specification, the allowable extrapolation limit is 125 percent 
of the highest PM CEMS response used to develop the correlation curve. 
If your source is a low-emitting source, the allowable extrapolation 
limit is 125 percent of the highest PM CEMS response used to develop 
the correlation curve or the PM CEMS response that corresponds to 50 
percent of the emission limit, whichever is greater. If the polynomial 
correlation curve maximum value is greater than the extrapolation 
limit, and the correlation curve satisfies all of the performance 
criteria specified in section 13.2 of this specification, you can use 
the polynomial correlation curve to predict PM concentrations. If the 
correlation curve maximum value is less than the extrapolation limit, 
you cannot use the polynomial correlation curve to predict PM 
concentrations.
    (4) You may petition the Administrator for alternative solutions or 
sampling recommendations if the correlation models described in section 
12.3 of this specification do not satisfy the performance criteria 
specified in section 13.2 of this specification.
13.0 What Are the Performance Criteria for My PM CEMS?
    You must evaluate your PM CEMS based on the 7-day drift check, the 
accuracy of the correlation, and the sampling periods and cycle/
response time.
    13.1 What is the 7-day drift check performance specification? Your 
daily PM CEMS internal drift checks must demonstrate that the average 
daily drift of your PM CEMS does not deviate from the value of the 
reference light, optical filter, Beta attenuation signal, or other 
technology-suitable reference standard by more than 2 percent of the 
upscale value. If your CEMS includes diluent and/or auxiliary monitors 
(for temperature, pressure, and/or moisture) that are employed as a 
necessary part of this performance specification, you must determine 
the calibration drift separately for each ancillary monitor in terms of 
its respective output (see the appropriate performance specification 
for the diluent CEMS specification). None of the calibration drifts may 
exceed their individual specification.
    13.2 What performance criteria must my PM CEMS correlation satisfy? 
Your PM CEMS correlation must meet each of the minimum specifications 
in paragraphs (1), (2), and (3) of this section. Before confidence and 
tolerance interval half range percentage calculations are made, you 
must convert the emission limit to the appropriate units of your PM 
CEMS measurement conditions using the average of emissions gas property 
values (e.g., diluent concentration, temperature, pressure, and 
moisture) measured during the correlation test.

[[Page 1814]]

    (1) The correlation coefficient must satisfy the criterion 
specified in paragraph (1)(i) or (ii), whichever applies.
    (i) If your source is not a low-emitting source, as defined in 
section 3.16 of this specification, the correlation coefficient (r) 
must be greater than or equal to 0.85.
    (ii) If your source is a low-emitting source, as defined in section 
3.16 of this specification, the correlation coefficient (r) must be 
greater than or equal to 0.75.
    (2) The confidence interval half range must satisfy the applicable 
criterion specified in paragraph (2)(i), (ii), or (iii) of this 
section, based on the type of correlation model.
    (i) For linear or logarithmic correlations, the 95 percent 
confidence interval half range at the mean PM CEMS response value from 
the correlation test must be within 10 percent of the PM emission limit 
value specified in theapplicable regulation, as calculated using 
Equation 11-10.
    (ii) For polynomial correlations, the 95 percent confidence 
interval half range at the PM CEMS response value from the correlation 
test that corresponds to the minimum value for [Delta] must be within 
10 percent of the PM emission limit value specified in the applicable 
regulation, as calculated using Equation 11-28.
    (iii) For exponential or power correlations, the 95 percent 
confidence interval half range at the median PM CEMS response value 
from the correlation test must be within 10 percent of the natural 
logarithm of the PM emission limit value specified in the applicable 
regulation, as calculated using Equation 11-40.
    (3) The tolerance interval half range must satisfy the applicable 
criterion specified in paragraph (3)(i), (ii), or (iii) of this 
section, based on the type of correlation model.
    (i) For linear or logarithmic correlations, the tolerance interval 
half range at the mean PM CEMS response value from the correlation test 
must have 95 percent confidence that 75 percent of all possible values 
are within 25 percent of the PM emission limit value specified in the 
applicable regulation, as calculated using Equation 11-13.
    (ii) For polynomial correlations, the tolerance interval half range 
at the PM CEMS response value from the correlation test that 
corresponds to the minimum value for [Delta] must have 95 percent 
confidence that 75 percent of all possible values are within 25 percent 
of the PM emission limit value specified in the applicable regulation, 
as calculated using Equation 11-32.
    (iii) For exponential or power correlations, the tolerance interval 
half range at the median PM CEMS response value from the correlation 
test must have 95 percent confidence that 75 percent of all possible 
values are within 25 percent of the natural logarithm of the PM 
emission limit value specified in the applicable regulation, as 
calculated using Equation 11-41.
    13.3 What are the sampling periods and cycle/response time? You 
must document and maintain the response time and any changes in the 
response time following installation.
    (1) If you have a batch sampling PM CEMS, you must evaluate the 
limits presented in paragraphs (1)(i) and (ii) of this section.
    (i) The response time of your PM CEMS, which is equivalent to the 
cycle time, must be no longer than 15 minutes. In addition, the delay 
between the end of the sampling time and reporting of the sample 
analysis must be no greater than 3 minutes. You must document any 
changes in the response time following installation.
    (ii) The sampling time of your PM CEMS must be no less than 30 
percent of the cycle time. If you have a batch sampling PM CEMS, 
sampling must be continuous except during pauses when the collected 
pollutant on the capture media is being analyzed and the next capture 
medium starts collecting a new sample.
    13.4 What PM compliance monitoring must I do? You must report your 
CEMS measurements in the units of the standard expressed in the 
regulations (e.g., mg/dscm @ 7 percent oxygen, pounds per million Btu 
(lb/mmBtu), etc.). You may need to install auxiliary data monitoring 
equipment to convert the units reported by your PM CEMS into units of 
the PM emission standard.
14.0 Pollution Prevention [Reserved]
15.0 Waste Management [Reserved]
16.0 Which References Are Relevant to This Performance Specification?
    16.1 Technical Guidance Document: Compliance Assurance Monitoring. 
U.S. Environmental Protection Agency Office of Air Quality Planning and 
Standards Emission Measurement Center. August 1998.
    16.2 40 CFR 60, Appendix B, ``Performance Specification 2--
Specifications and Test Procedures for SO2, and 
NOX, Continuous Emission Monitoring Systems in Stationary 
Sources.''
    16.3 40 CFR 60, Appendix B, ``Performance Specification 1--
Specification and Test Procedures for Opacity Continuous Emission 
Monitoring Systems in Stationary Sources.''
    16.4 40 CFR 60, Appendix A, ``Method 1--Sample and Velocity 
Traverses for Stationary Sources.''
    16.5 ``Current Knowledge of Particulate Matter (PM) Continuous 
Emission Monitoring.'' EPA-454/R-00-039. U.S. Environmental Protection 
Agency, Research Triangle Park, NC. September 2000.
    16.6 40 CFR 266, Appendix IX, Section 2, ``Performance 
Specifications for Continuous Emission Monitoring Systems.''
    16.7 ISO 10155, ``Stationary Source Emissions--Automated Monitoring 
of Mass Concentrations of Particles: Performance Characteristics, Test 
Procedures, and Specifications.'' American National Standards 
Institute, New York City. 1995.
17.0 What Reference Tables and Validation Data Are Relevant to PS-11?
    Use the information in Table 1 for determining the confidence and 
tolerance interval half ranges. Use Table 2 to record your 7-day drift 
test data.

               Table 1.--Factors for Calculation of Confidence and Tolerance Interval Half Ranges
----------------------------------------------------------------------------------------------------------------
                            df or n'                                    tdf             vdf          un' (75)
----------------------------------------------------------------------------------------------------------------
2...............................................................           4.303           4.415           1.433
3...............................................................           3.182           2.920           1.340
4...............................................................           2.776           2.372           1.295
5...............................................................           2.571           2.089           1.266
6...............................................................           2.447           1.915           1.247
7...............................................................           2.365           1.797           1.233
8...............................................................           2.306           1.711           1.223
9...............................................................           2.262           1.645           1.214
10..............................................................           2.228           1.593           1.208

[[Page 1815]]

 
11..............................................................           2.201           1.551           1.203
12..............................................................           2.179           1.515           1.199
13..............................................................           2.160           1.485           1.195
14..............................................................           2.145           1.460           1.192
15..............................................................           2.131           1.437           1.189
16..............................................................           2.120           1.418           1.187
17..............................................................           2.110           1.400           1.185
18..............................................................           2.101           1.385           1.183
19..............................................................           2.093           1.370           1.181
20..............................................................           2.086           1.358           1.179
21..............................................................           2.080           1.346           1.178
22..............................................................           2.074           1.335           1.177
23..............................................................           2.069           1.326           1.175
24..............................................................           2.064           1.317           1.174
25..............................................................           2.060           1.308           1.173
26..............................................................           2.056           1.301           1.172
27..............................................................           2.052           1.294           1.172
28..............................................................           2.048           1.287           1.171
29..............................................................           2.045           1.281           1.171
30..............................................................           2.042           1.274           1.170
31..............................................................           2.040           1.269           1.169
32..............................................................           2.037           1.264           1.169
33..............................................................           2.035           1.258           1.168
34..............................................................           2.032           1.253           1.168
35..............................................................           2.030           1.248           1.167
36..............................................................           2.028           1.244           1.167
37..............................................................           2.026           1.240           1.166
38..............................................................           2.025           1.236           1.166
39..............................................................           2.023           1.232           1.165
40..............................................................           2.021           1.228           1.165
41..............................................................           2.020           1.225           1.165
42..............................................................           2.018           1.222           1.164
43..............................................................           2.017           1.219           1.164
44..............................................................           2.015           1.216           1.163
45..............................................................           2.014           1.213           1.163
46..............................................................           2.013           1.210           1.163
47..............................................................           2.012           1.207           1.163
48..............................................................           2.011           1.205           1.162
49..............................................................           2.010           1.202           1.162
50..............................................................           2.009           1.199           1.162
51..............................................................           2.008           1.197           1.162
52..............................................................           2.007           1.194           1.162
53..............................................................           2.006           1.191           1.161
54..............................................................           2.005           1.189           1.161
55..............................................................           2.005           1.186           1.161
56..............................................................           2.004           1.183           1.161
57..............................................................           2.003           1.181           1.161
58..............................................................           2.002           1.178           1.160
59..............................................................           2.001           1.176           1.160
60..............................................................           2.000           1.173           1.160
61..............................................................           2.000           1.170           1.160
62..............................................................           1.999           1.168           1.160
63..............................................................           1.999           1.165           1.159
----------------------------------------------------------------------------------------------------------------


                                         Table 2.--7-Day Drift Test Data
----------------------------------------------------------------------------------------------------------------
                                                            PM CEMS
  Zero drift day    Date  and  time     Zero check         response         Difference      Zero drift  ((RCEMS-
                               value  (RL)         (RCEMS)         (RCEMS-RL)         RL) /RU) x 100
----------------------------------------------------------------------------------------------------------------
1
----------------------------------------------------------------------------------------------------------------
2
----------------------------------------------------------------------------------------------------------------
3
----------------------------------------------------------------------------------------------------------------
4
----------------------------------------------------------------------------------------------------------------
5
----------------------------------------------------------------------------------------------------------------

[[Page 1816]]

 
6
----------------------------------------------------------------------------------------------------------------
7
----------------------------------------------------------------------------------------------------------------


 
                                                            PM CEMS                            Upscale drift
Upscale drift day   Date  and  time   Upscale  check       response         Difference       ((RCEMS-RU)/RU) x
                               value  (RU)         (RCEMS)         (RCEMS-RU)              100%
----------------------------------------------------------------------------------------------------------------
1
----------------------------------------------------------------------------------------------------------------
2
----------------------------------------------------------------------------------------------------------------
3
----------------------------------------------------------------------------------------------------------------
4
----------------------------------------------------------------------------------------------------------------
5
----------------------------------------------------------------------------------------------------------------
6
----------------------------------------------------------------------------------------------------------------
7
----------------------------------------------------------------------------------------------------------------


0
3. Appendix F, part 60 is amended by adding Procedure 2 to read as 
follows:

Appendix F--Quality Assurance Procedures

* * * * *

PROCEDURE 2--Quality Assurance Requirements for Particulate Matter 
Continuous Emission Monitoring Systems at Stationary Sources

1.0 What Are the Purpose and Applicability of Procedure 2?
    The purpose of Procedure 2 is to establish the minimum requirements 
for evaluating the effectiveness of quality control (QC) and quality 
assurance (QA) procedures and the quality of data produced by your 
particulate matter (PM) continuous emission monitoring system (CEMS). 
Procedure 2 applies to PM CEMS used for continuously determining 
compliance with emission standards or operating permit limits as 
specified in an applicable regulation or permit. Other QC procedures 
may apply to diluent (e.g., O2) monitors and other auxiliary 
monitoring equipment included with your CEMS to facilitate PM 
measurement or determination of PM concentration in units specified in 
an applicable regulation.
    1.1 What measurement parameter does Procedure 2 address? Procedure 
2 covers the instrumental measurement of PM as defined by your source's 
applicable reference method (no Chemical Abstract Service number 
assigned).
    1.2 For what types of devices must I comply with Procedure 2? You 
must comply with Procedure 2 for the total equipment that:
    (1) We require you to install and operate on a continuous basis 
under the applicable regulation, and
    (2) You use to monitor the PM mass concentration associated with 
the operation of a process or emission control device.
    1.3 What are the data quality objectives (DQOs) of Procedure 2? The 
overall DQO of Procedure 2 is the generation of valid, representative 
data that can be transferred into useful information for determining PM 
CEMS concentrations averaged over a prescribed interval. Procedure 2 is 
also closely associated with Performance Specification 11 (PS-11).
    (1) Procedure 2 specifies the minimum requirements for controlling 
and assessing the quality of PM CEMS data submitted to us or the 
delegated permitting authority.
    (2) You must meet these minimum requirements if you are responsible 
for one or more PM CEMS used for compliance monitoring. We encourage 
you to develop and implement a more extensive QA program or to continue 
such programs where they already exist.
    1.4 What is the intent of the QA/QC procedures specified in 
Procedure 2? Procedure 2 is intended to establish the minimum QA/QC 
requirements for PM CEMS and is presented in general terms to allow you 
to develop a program that is most effective for your circumstances. You 
may adopt QA/QC procedures that go beyond these minimum requirements to 
ensure compliance with applicable regulations.
    1.5 When must I comply with Procedure 2? You must comply with the 
basic requirements of Procedure 2 immediately following successful 
completion of the initial correlation test of PS-11.
2.0 What Are the Basic Requirements of Procedure 2?
    Procedure 2 requires you to perform periodic evaluations of PM CEMS 
performance and to develop and implement QA/QC programs to ensure that 
PM CEMS data quality is maintained.
    2.1 What are the basic functions of Procedure 2?
    (1) Assessment of the quality of your PM CEMS data by estimating 
measurement accuracy;
    (2) Control and improvement of the quality of your PM CEMS data by 
implementing QC requirements and corrective actions until the data 
quality is acceptable; and
    (3) Specification of requirements for daily instrument zero and 
upscale drift checks and daily sample volume checks, as well as routine 
response correlation audits, absolute correlation audits, sample volume 
audits, and relative response audits.
3.0 What Special Definitions Apply to Procedure 2?
    The definitions in Procedure 2 include those provided in PS-11 of 
Appendix B, with the following additions:
    3.1 ``Absolute Correlation Audit (ACA)'' means an evaluation of 
your PM CEMS response to a series of reference

[[Page 1817]]

standards covering the full measurement range of the instrument (e.g., 
4 mA to 20 mA).
    3.2 ``Correlation Range'' means the range of PM CEMS responses used 
in the complete set of correlation test data.
    3.3 ``PM CEMS Correlation'' means the site-specific relationship 
(i.e., a regression equation) between the output from your PM CEMS 
(e.g., mA) and the particulate concentration, as determined by the 
reference method. The PM CEMS correlation is expressed in the same 
units as the PM concentration measured by your PM CEMS (e.g., mg/acm). 
You must derive this relation from PM CEMS response data and manual 
reference method data that were gathered simultaneously. These data 
must be representative of the full range of source and control device 
operating conditions that you expect to occur. You must develop the 
correlation by performing the steps presented in sections 12.2 and 12.3 
of PS-11.
    3.4 ``Reference Method Sampling Location'' means the location in 
your source's exhaust duct from which you collect manual reference 
method data for developing your PM CEMS correlation and for performing 
relative response audits (RRAs) and response correlation audits (RCAs).
    3.5 ``Response Correlation Audit (RCA)'' means the series of tests 
specified in section 10.3(8) of this procedure that you conduct to 
ensure the continued validity of your PM CEMS correlation.
    3.6 ``Relative Response Audit (RRA)'' means the brief series of 
tests specified in section 10.3(6) of this procedure that you conduct 
between consecutive RCAs to ensure the continued validity of your PM 
CEMS correlation.
    3.7 ``Sample Volume Audit (SVA)'' means an evaluation of your PM 
CEMS measurement of sample volume if your PM CEMS determines PM 
concentration based on a measure of PM mass in an extracted sample 
volume and an independent determination of sample volume.
4.0 Interferences. [Reserved]
5.0 What Do I Need To Know To Ensure the Safety of Persons Using 
Procedure 2?
    People using Procedure 2 may be exposed to hazardous materials, 
operations, and equipment. Procedure 2 does not purport to address all 
of the safety issues associated with its use. It is your responsibility 
to establish appropriate safety and health practices and determine the 
applicable regulatory limitations before performing this procedure. You 
must consult your CEMS user's manual for specific precautions to be 
taken with regard to your PM CEMS procedures.
6.0 What Equipment and Supplies Do I Need?
    [Reserved]
7.0 What Reagents and Standards Do I Need?
    You will need reference standards or procedures to perform the zero 
drift check, the upscale drift check, and the sample volume check.
    7.1 What is the reference standard value for the zero drift check? 
You must use a zero check value that is no greater than 20 percent of 
the PM CEMS's response range. You must obtain documentation on the zero 
check value from your PM CEMS manufacturer.
    7.2 What is the reference standard value for the upscale drift 
check? You must use an upscale check value that produces a response 
between 50 and 100 percent of the PM CEMS's response range. For a PM 
CEMS that produces output over a range of 4 mA to 20 mA, the upscale 
check value must produce a response in the range of 12 mA to 20 mA. You 
must obtain documentation on the upscale check value from your PM CEMS 
manufacturer.
    7.3 What is the reference standard value for the sample volume 
check? You must use a reference standard value or procedure that 
produces a sample volume value equivalent to the normal sampling rate. 
You must obtain documentation on the sample volume value from your PM 
CEMS manufacturer.
8.0 What Sample Collection, Preservation, Storage, and Transport Are 
Relevant to This Procedure?
    [Reserved]
9.0 What Quality Control Measures Are Required by This Procedure for My 
PM CEMS?
    You must develop and implement a QC program for your PM CEMS. Your 
QC program must, at a minimum, include written procedures that 
describe, in detail, complete step-by-step procedures and operations 
for the activities in paragraphs (1) through (8) of this section.
    (1) Procedures for performing drift checks, including both zero 
drift and upscale drift and the sample volume check (see sections 
10.2(1), (2), and (5)).
    (2) Methods for adjustment of PM CEMS based on the results of drift 
checks, sample volume checks (if applicable), and the periodic audits 
specified in this procedure.
    (3) Preventative maintenance of PM CEMS (including spare parts 
inventory and sampling probe integrity).
    (4) Data recording, calculations, and reporting.
    (5) RCA and RRA procedures, including sampling and analysis 
methods, sampling strategy, and structuring test conditions over the 
prescribed range of PM concentrations.
    (6) Procedures for performing ACAs and SVAs and methods for 
adjusting your PM CEMS response based on ACA and SVA results.
    (7) Program of corrective action for malfunctioning PM CEMS, 
including flagged data periods.
    (8) For extractive PM CEMS, procedures for checking extractive 
system ducts for material accumulation.
    9.1 What QA/QC documentation must I have? You are required to keep 
the written QA/QC procedures on record and available for inspection by 
us, the State, and/or local enforcement agency for the life of your 
CEMS or until you are no longer subject to the requirements of this 
procedure.
    9.2 How do I know if I have acceptable QC procedures for my PM 
CEMS? Your QC procedures are inadequate or your PM CEMS is incapable of 
providing quality data if you fail two consecutive QC audits (i.e., 
out-of-control conditions resulting from the annual audits, quarterly 
audits, or daily checks). Therefore, if you fail the same two 
consecutive audits, you must revise your QC procedures or modify or 
replace your PM CEMS to correct the deficiencies causing the excessive 
inaccuracies (see section 10.4 for limits for excessive audit 
inaccuracy).
10.0 What Calibration/Correlation and Standardization Procedures Must I 
Perform for My PM CEMS?
    You must generate a site-specific correlation for each of your PM 
CEMS installation(s) relating response from your PM CEMS to results 
from simultaneous PM reference method testing. The PS-11 defines 
procedures for developing the correlation and defines a series of 
statistical parameters for assessing acceptability of the correlation. 
However, a critical component of your PM CEMS correlation process is 
ensuring the accuracy and precision of reference method data. The 
activities listed in sections 10.1 through 10.10 assure the quality of 
the correlation.
    10.1 When should I use paired trains for reference method testing? 
Although not required, we recommend that you should use paired-train 
reference method testing to generate data used to develop your PM CEMS 
correlation and

[[Page 1818]]

for RCA testing. Guidance on the use of paired sampling trains can be 
found in the PM CEMS Knowledge Document (see section 16.5).
    10.2 What routine system checks must I perform on my PM CEMS? You 
must perform routine checks to ensure proper operation of system 
electronics and optics, light and radiation sources and detectors, and 
electric or electro-mechanical systems. Necessary components of the 
routine system checks will depend on design details of your PM CEMS. As 
a minimum, you must verify the system operating parameters listed in 
paragraphs (1) through (5) of this section on a daily basis. Some PM 
CEMS may perform one or more of these functions automatically or as an 
integral portion of unit operations; for other PM CEMS, you must 
initiate or perform one or more of these functions manually.
    (1) You must check the zero drift to ensure stability of your PM 
CEMS response to the zero check value. You must determine system output 
on the most sensitive measurement range when the PM CEMS is challenged 
with a zero reference standard or procedure. You must, at a minimum, 
adjust your PM CEMS whenever the daily zero drift exceeds 4 percent.
    (2) You must check the upscale drift to ensure stability of your PM 
CEMS response to the upscale check value. You must determine system 
output when the PM CEMS is challenged with a reference standard or 
procedure corresponding to the upscale check value. You must, at a 
minimum, adjust your PM CEMS whenever the daily upscale drift check 
exceeds 4 percent.
    (3) For light-scattering and extinction-type PM CEMS, you must 
check the system optics to ensure that system response has not been 
altered by the condition of optical components, such as fogging of lens 
and performance of light monitoring devices.
    (4) You must record data from your automatic drift-adjusting PM 
CEMS before any adjustment is made. If your PM CEMS automatically 
adjusts its response to the corrected calibration values (e.g., 
microprocessor control), you must program your PM CEMS to record the 
unadjusted concentration measured in the drift check before resetting 
the calibration. Alternately, you may program your PM CEMS to record 
the amount of adjustment.
    (5) For extractive PM CEMS that measure the sample volume and use 
the measured sample volume as part of calculating the output value, you 
must check the sample volume on a daily basis to verify the accuracy of 
the sample volume measuring equipment. This sample volume check must be 
done at the normal sampling rate of your PM CEMS. You must adjust your 
PM CEMS sample volume measurement whenever the daily sample volume 
check error exceeds 10 percent.
    10.3 What are the auditing requirements for my PM CEMS? You must 
subject your PM CEMS to an ACA and an SVA, as applicable, at least once 
each calender quarter. Successive quarterly audits must occur no closer 
than 2 months apart. You must conduct an RCA and an RRA at the 
frequencies specified in the applicable regulation or facility 
operating permit. An RRA or RCA conducted during any calendar quarter 
can take the place of the ACA required for that calendar quarter. An 
RCA conducted during the period in which an RRA is required can take 
the place of the RRA for that period.
    (1) When must I perform an ACA? You must perform an ACA each 
quarter unless you conduct an RRA or RCA during that same quarter.
    (2) How do I perform an ACA? You perform an ACA according to the 
procedure specified in paragraphs (2)(i) through (v) of this section.
    (i) You must challenge your PM CEMS with an audit standard or an 
equivalent audit reference to reproduce the PM CEMS's measurement at 
three points within the following ranges:

------------------------------------------------------------------------
            Audit point                          Audit range
------------------------------------------------------------------------
1.................................  0 to 20 percent of measurement range
2.................................  40 to 60 percent of measurement
                                     range
3.................................  70 to 100 percent of measurement
                                     range
------------------------------------------------------------------------

    (ii) You must then challenge your PM CEMS three times at each audit 
point and use the average of the three responses in determining 
accuracy at each audit point. Use a separate audit standard for audit 
points 1, 2, and 3. Challenge the PM CEMS at each audit point for a 
sufficient period of time to ensure that your PM CEMS response has 
stabilized.
    (iii) Operate your PM CEMS in the mode, manner, and range specified 
by the manufacturer.
    (iv) Store, maintain, and use audit standards as recommended by the 
manufacturer.
    (v) Use the difference between the actual known value of the audit 
standard and the response of your PM CEMS to assess the accuracy of 
your PM CEMS.
    (3) When must I perform an SVA? You must perform an audit of the 
measured sample volume (e.g., the sampling flow rate for a known time) 
once per quarter for applicable PM CEMS with an extractive sampling 
system. Also, you must perform and pass an SVA prior to initiation of 
any of the reference method data collection runs for an RCA or RRA.
    (4) How do I perform an SVA? You perform an SVA according to the 
procedure specified in paragraphs (4)(i) through (iii) of this section.
    (i) You perform an SVA by independently measuring the volume of 
sample gas extracted from the stack or duct over each batch cycle or 
time period with a calibrated device. You may make this measurement 
either at the inlet or outlet of your PM CEMS, so long as it measures 
the sample gas volume without including any dilution or recycle air. 
Compare the measured volume with the volume reported by your PM CEMS 
for the same cycle or time period to calculate sample volume accuracy.
    (ii) You must make measurements during three sampling cycles for 
batch extractive monitors (e.g., Beta-gauge) or during three periods of 
at least 20 minutes for continuous extractive PM CEMS.
    (iii) You may need to condense, collect, and measure moisture from 
the sample gas prior to the calibrated measurement device (e.g., dry 
gas meter) and correct the results for moisture content. In any case, 
the volumes measured by the calibrated device and your PM CEMS must be 
on a consistent temperature, pressure, and moisture basis.
    (5) How often must I perform an RRA? You must perform an RRA at the 
frequency specified in the applicable regulation or facility operating 
permit. You may conduct an RCA instead of an RRA during the period when 
the RRA is required.
    (6) How do I perform an RRA? You must perform the RRA according to 
the procedure specified in paragraphs (6)(i) and (ii) of this section.
    (i) You perform an RRA by collecting three simultaneous reference 
method

[[Page 1819]]

PM concentration measurements and PM CEMS measurements at the as-found 
source operating conditions and PM concentration.
    (ii) We recommend that you use paired trains for reference method 
sampling. Guidance on the use of paired sampling trains can be found in 
the PM CEMS Knowledge Document (see section 16.5 of PS-11).
    (7) How often must I perform an RCA? You must perform an RCA at the 
frequency specified in the applicable regulation or facility operating 
permit.
    (8) How do I perform an RCA? You must perform the RCA according to 
the procedures for the PM CEMS correlation test described in PS-11, 
section 8.6, except that the minimum number of runs required is 12 in 
the RCA instead of 15 as specified in PS-11.
    (9) What other alternative audits can I use? You can use other 
alternative audit procedures as approved by us, the State, or local 
agency for the quarters when you would conduct ACAs.
    10.4 What are my limits for excessive audit inaccuracy? Unless 
specified otherwise in the applicable subpart, the criteria for 
excessive audit inaccuracy are listed in paragraphs (1) through (6) of 
this section.
    (1) What are the criteria for excessive zero or upscale drift? Your 
PM CEMS is out of control if the zero drift check or upscale drift 
check either exceeds 4 percent for five consecutive daily periods or 
exceeds 8 percent for any one day.
    (2) What are the criteria for excessive sample volume measurement 
error? Your PM CEMS is out of control if sample volume check error 
exceeds 10 percent for five consecutive daily periods or exceeds 20 
percent for any one day.
    (3) What are the criteria for excessive ACA error? Your PM CEMS is 
out of control if the results of any ACA exceed +/- 10 percent of the 
average audit value or 7.5 percent of the applicable standard, 
whichever is greater.
    (4) What is the criterion for excessive SVA error? Your PM CEMS is 
out of control if results exceed +/- 5 percent of the average sample 
volume audit value.
    (5) What are the criteria for passing an RCA? To pass an RCA, you 
must meet the criteria specified in paragraphs (5)(i) through (iii) of 
this section. If your PM CEMS fails to meet these RCA criteria, it is 
out of control.
    (i) For all 12 data points, the PM CEMS response value can be no 
greater than the greatest PM CEMS response value used to develop your 
correlation curve.
    (ii) For 9 of the 12 data points, the PM CEMS response value must 
lie within the PM CEMS output range used to develop your correlation 
curve.
    (iii) At least 75 percent of a minimum number of 12 sets of PM CEMS 
and reference method measurements must fall within a specified area on 
a graph of the correlation regression line. The specified area on the 
graph of the correlation regression line is defined by two lines 
parallel to the correlation regression line, offset at a distance of +/
- 25 percent of the numerical emission limit value from the correlation 
regression line.
    (6) What are the criteria to pass an RRA? To pass an RRA, you must 
meet the criteria specified in paragraphs (6)(i) and (ii) of this 
section. If your PM CEMS fails to meet these RRA criteria, it is out of 
control.
    (i) For all three data points, the PM CEMS response value can be no 
greater than the greatest PM CEMS response value used to develop your 
correlation curve.
    (ii) For two of the three data points, the PM CEMS response value 
must lie within the PM CEMS output range used to develop your 
correlation curve.
    (iii) At least two of the three sets of PM CEMS and reference 
method measurements must fall within the same specified area on a graph 
of the correlation regression line as required for the RCA and 
described in paragraph (5)(iii) of this section.
    10.5 What do I do if my PM CEMS is out of control? If your PM CEMS 
is out of control, you must take the actions listed in paragraphs (1) 
and (2) of this section.
    (1) You must take necessary corrective action to eliminate the 
problem and perform tests, as appropriate, to ensure that the 
corrective action was successful.
    (i) Following corrective action, you must repeat the previously 
failed audit to confirm that your PM CEMS is operating within the 
specifications.
    (ii) If your PM CEMS failed an RRA, you must take corrective action 
until your PM CEMS passes the RRA criteria. If the RRA criteria cannot 
be achieved, you must perform an RCA.
    (iii) If your PM CEMS failed an RCA, you must follow procedures 
specified in section 10.6 of this procedure.
    (2) You must report both the audit showing your PM CEMS to be out 
of control and the results of the audit following corrective action 
showing your PM CEMS to be operating within specifications.
    10.6 What do I do if my PM CEMS fails an RCA? After an RCA failure, 
you must take all applicable actions listed in paragraphs (1) through 
(3) of this section.
    (1) Combine RCA data with data from the active PM CEMS correlation 
and perform the mathematical evaluations defined in PS-11 for 
development of a PM CEMS correlation, including examination of 
alternate correlation models (i.e., linear, polynomial, logarithmic, 
exponential, and power). If the expanded data base and revised 
correlation meet PS-11 statistical criteria, use the revised 
correlation.
    (2) If the criteria specified in paragraph (1) of this section are 
not achieved, you must develop a new PM CEMS correlation based on 
revised data. The revised data set must consist of the test results 
from only the RCA. The new data must meet all requirements of PS-11 to 
develop a revised PM CEMS correlation, except that the minimum number 
of sets of PM CEMS and reference method measurements is 12 instead of 
the minimum of 15 sets required by PS-11. Your PM CEMS is considered to 
be back in controlled status when the revised correlation meets all of 
the performance criteria specified in section 13.2 of PS-11.
    (3) If the actions in paragraphs (1) and (2) of this section do not 
result in an acceptable correlation, you must evaluate the cause(s) and 
comply with the actions listed in paragraphs (3)(i) through (iv) of 
this section within 90 days after the completion of the failed RCA.
    (i) Completely inspect your PM CEMS for mechanical or operational 
problems. If you find a mechanical or operational problem, repair your 
PM CEMS and repeat the RCA.
    (ii) You may need to relocate your PM CEMS to a more appropriate 
measurement location. If you relocate your PM CEMS, you must perform a 
new correlation test according to the procedures specified in PS-11.
    (iii) The characteristics of the PM or gas in your source's flue 
gas stream may have changed such that your PM CEMS measurement 
technology is no longer appropriate. If this is the case, you must 
install a PM CEMS with measurement technology that is appropriate for 
your source's flue gas characteristics. You must perform a new 
correlation test according to the procedures specified in PS-11.
    (iv) If the corrective actions in paragraphs (3)(i) through (iii) 
of this section were not successful, you must petition us, the State, 
or local agency for approval of alternative criteria or an alternative 
for continuous PM monitoring.
    10.7 When does the out-of-control period begin and end? The out-of-
control period begins immediately after the last test run or check of 
an

[[Page 1820]]

unsuccessful RCA, RRA, ACA, SVA, drift check, or sample volume check. 
The out-of-control period ends immediately after the last test run or 
check of the subsequent successful audit or drift check.
    10.8 Can I use the data recorded by my PM CEMS during out-of-
control periods? During any period when your PM CEMS is out of control, 
you may not use your PM CEMS data to calculate emission compliance or 
to meet minimum data availability requirements described in the 
applicable regulation.
    10.9 What are the QA/QC reporting requirements for my PM CEMS? You 
must report the accuracy results for your PM CEMS, specified in section 
10.4 of this procedure, at the interval specified in the applicable 
regulation. Report the drift and accuracy information as a Data 
Assessment Report (DAR), and include one copy of this DAR for each 
quarterly audit with the report of emissions required under the 
applicable regulation. An example DAR is provided in Procedure 1, 
Appendix F of this part.
    10.10 What minimum information must I include in my DAR? As a 
minimum, you must include the information listed in paragraphs (1) 
through (5) of this section in the DAR:
    (1) Your name and address.
    (2) Identification and location of monitors in your CEMS.
    (3) Manufacturer and model number of each monitor in your CEMS.
    (4) Assessment of PM CEMS data accuracy/acceptability, and date of 
assessment, as determined by an RCA, RRA, ACA, or SVA described in 
section 10, including the acceptability determination for the RCA or 
RRA, the accuracy for the ACA or SVA, the reference method results, the 
audit standards, your PM CEMS responses, and the calculation results as 
defined in section 12. If the accuracy audit results show your PM CEMS 
to be out of control, you must report both the audit results showing 
your PM CEMS to be out of control and the results of the audit 
following corrective action showing your PM CEMS to be operating within 
specifications.
    (5) Summary of all corrective actions you took when you determined 
your PM CEMS to be out of control, as described in section 10.5, or 
after failing on RCA, as described in section 10.6.
    10.7 Where and how long must I retain the QA data that this 
procedure requires me to record for my PM CEMS? You must keep the 
records required by this procedure for your PM CEMS onsite and 
available for inspection by us, the State, and/or local enforcement 
agency for a period of 5 years.
11.0 What Analytical Procedures Apply to This Procedure?
    Sample collection and analysis are concurrent for this procedure. 
You must refer to the appropriate reference method for the specific 
analytical procedures.
12.0 What Calculations and Data Analysis Must I Perform for my PM CEMS?
    (1) How do I determine RCA and RRA acceptability? You must plot 
each of your PM CEMS and reference method data sets from an RCA or RRA 
on a graph based on your PM CEMS correlation line to determine if the 
criteria in paragraphs 10.4(5) or (6), respectively, are met.
    (2) How do I calculate ACA accuracy? You must use Equation 2-1 to 
calculate ACA accuracy for each of the three audit points:
[GRAPHIC] [TIFF OMITTED] TR12JA04.045

Where:

ACA Accuracy = The ACA accuracy at each audit point, in percent,
RCEM = Your PM CEMS response to the reference standard, and
RV = The reference standard value.

    (3) How do I calculate daily upscale and zero drift? You must 
calculate the upscale drift using to Equation 2-2 and the zero drift 
according to Equation 2-3:
[GRAPHIC] [TIFF OMITTED] TR12JA04.046

Where:

UD = The upscale drift of your PM CEMS, in percent,
RCEM = Your PM CEMS response to the upscale check value, and
RU = The upscale check value.
[GRAPHIC] [TIFF OMITTED] TR12JA04.047

Where:

ZD = The zero (low-level) drift of your PM CEMS, in percent,
RCEM = Your PM CEMS response of the zero check value,
RL = The zero check value, and
RU = The upscale check value.

    (4) How do I calculate SVA accuracy? You must use Equation 2-4 to 
calculate the accuracy, in percent, for each of the three SVA tests or 
the daily sample volume check:
[GRAPHIC] [TIFF OMITTED] TR12JA04.048


[[Page 1821]]


Where:

VM = Sample gas volume determined/reported by your PM CEMS 
(e.g., dscm),
VR = Sample gas volume measured by the independent 
calibrated reference device (e.g., dscm) for the SVA or the reference 
value for the daily sample volume check, and
FS = Full-scale value.


    Note: Before calculating SVA accuracy, you must correct the 
sample gas volumes measured by your PM CEMS and the independent 
calibrated reference device to the same basis of temperature, 
pressure, and moisture content. You must document all data and 
calculations.

13.0 Method Performance. [Reserved]
14.0 Pollution Prevention. [Reserved]
15.0 Waste Management. [Reserved]
16.0 Which References are Relevant to This Method? [Reserved]
17.0 What Tables, Diagrams, Flowcharts, and Validation Data Are 
Relevant to This Method? [Reserved]
[FR Doc. 04-5 Filed 1-9-04; 8:45 am]
BILLING CODE 6560-50-P