IJnited States Office of Air Quality EPA-450/3-82-022
Environmental-Protection Planning and Standards September 1982
Agency Research Triangle Park NC 27711
Air ,
Revisions to
Continuous Emission
Monitoring Systems-
Performance
Specifications 2 and 3-
Summary of Comments
and Responses
-------
EPA-450/3-82-022
Revisions to
Continuous Emission Monitoring Systems
Performance Specifications 2 and 3
(Proposed January 26, 1981, 46 FR 8352)
Summary of Comments and Responses
Emission Measurement Branch
Emission Standards and Engineering Division
U.S ENVIRONMENTAL PROTECTION AGENCY
Office of Air, Noise, and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
September 1982
-------
This report has been reviewed by the Emission Standards and Engineering Division of the Office of Air
Quality Planning and Standards, EPA, and approved for publication. Mention of trade names or
commercial products is not intended to constitute endorsement or recommendation for use. Copies of
this report are available through the Library Services Office (MD-35), U.S. Environmental Protection
Agency, Research Triangle Park, N.C. 27711; or, for a fee, from the National Technical Information
Services, 5285 Port Royal Road, Springfield, Virginia 22161.
-------
TABLE OF CONTENTS
Page
Chapter 1. INTRODUCTION 1
Chapter 2. SUMMARY OF CHANGES SINCE PROPOSAL 3
Chapter 3. SUMMARY OF COMMENTS AND RESPONSES 6
Table 1. List of Acronyms Used in Summary of
Comments and Responses 2
Table 2. List of Commenters 73
m
-------
Chapter 1
INTRODUCTION
On January 26, 1981, the U.S. Environmental Protection Agency (EPA)
published in the Federal Register (46 FR 8352) revised "Performance
Specification 2 - Specifications and Test Procedures for S02 and
NO Continuous Emission Monitoring Systems in Stationary Sources,"
J\
and "Performance Specification 3 - Specifications and Test Procedures
for Og and ^ Continuous Emission Monitoring Systems in Stationary
Sources." These revised specifications were proposed under the
authority of Sections 111, 114, and 301(a) of the Clean Air Act, as
amended.
Public comments were solicited at the time of proposal. To
provide interested persons the opportunity for oral presentation of
data, views, or arguments concerning the proposed specifications,
a public hearing was scheduled for February 19, 1981, beginning at
9:00 a.m. The hearing, however, was not held because no one requested
to speak. The public comment period was from January 26, 1981, to
March 27, 1981.
Twelve comment letters on the proposed revisions to the
performance specifications (PSrs) were received from industry, Federal
agencies, State air pollution control agencies, trade associations, and
State air pollution control agencies, trade associations, and equipment
manufacturers. The comments that were submitted, along with EPA's
responses, are summarized in this document. The summary of comments
and responses serves as a basis for the revisions that have been made
to the test methods between proposal and promulgation.
-------
TABLE 1. LIST OF ACRONYMS USED
CD - Calibration Drift
CEMS- Continuous Emission Monitoring System
NSPS- New Source Performance Standards
PS - Performance Specification
PST - Performance Specification Test
QA - Quality Assurance
RA - Relative Accuracy
RM - Reference Method
SIP - State Implementation Plan
-------
CHAPTER 2
SUMMARY OF CHANGES SINCE PROPOSAL
1. Section 60.13(c)(2)(ii). The revision proposed for this
subparagraph is no longer under consideration.
2. Section 60.13(d). The limit of the mandatory adjustment
of the zero and span drifts was changed from the applicable
specification to twice the applicable specification. The requirement
to quantify and record the amount of excess zero and span drift was
changed to require the quantifying and recording only when specified.
3. Section 1.1. (Note: All sections refer to PS 2, unless otherwise
specified.) A second paragraph was added to explain that performance
specification (PS) is not designed to evaluate the installed continuous
emission monitoring system (CEMS) over an extended period of time nor does
it identify specific calibration techniques and other auxiliary procedures
to assess the CEMS performance. This paragraph also reminds the source
owner or operator of his/her responsibility to properly calibrate, maintain,
and operate the CEMS. This paragraph was also added to PS 3.
4. Section 3. Parts of Section 3.1, 3.1.1, and 3.2 have been
amended to address locations at inlets to control devices.
5. Section 3.1. The last sentence in the first paragraph was
revised to read, "If the cause of failure to meet the RA test is determined
to be the measurement location and a satisfactory correction technique
cannot be established, the Administrator may require the CEMS to be
relocated." The RA stands for relative accuracy.
-------
6. Section 3.2. For stacks greater than 2.4 meters, the use
of the 0.4, 1.2, and 2.0 meter points was limited to those cases
where pollutant stratification is not expected and prohibited after
wet scrubbers or at points where two streams with different pollutant
concentrations are combined.
For those cases where the pollutant concentration changes are
due solely to diluent leakages (e.g., air heater leakages) and pollutants
and diluents are simultaneously measured at the same location, the
option of using a half-diameter in lieu of two equivalent diameters is
allowed.
7. Section 4.1. The limits of the recorder high level was
changed from 90 to 100 percent of the span value to a value between
1.5 times the pollutant concentration corresponding to the emission
standard level and the span value. The option of using a value
between 50 and 100 percent of the high-level value was allowed
provided the data recorder full-scale requirements are met. The
zero and high-level points for the determination of calibration drifts
were changed from "0 to 50" to "0 to 20" and from "80 to 100" to
"50 to 100" of the high-level value.
Inlets to scrubbers were also addressed.
8. For SOg emission standards between 130 and 86 ng/J (0.30
and 0.20 Ib/million Btu), the RA specification was changed to 15 percent;
and below 86 ng/J (0.2 Ib/million Btu), the RA specification was changed
to 20 percent of the emission standard.
-------
9. Sections 5.2 and 5.3. The minimum acceptable operational
condition was changed from 50 percent "capacity" to 50 percent
"of normal load."
10. Section 5.3. This section was changed to allow the RA
test to be conducted during the calibration drift (CD) test period.
11. Section 6. The third paragraph concerning the use of
incremental addition procedure was deleted. The first paragraph
was revised to allow other CD procedures as long as the CD can be
determined.
12. Section 7.1.2. The sentence, "A test run for grab samples
must be made up of at least three separate measurements," was added.
13. Section 7.2.2. For those cases where the pollutant
concentration is varying with time over the run, the arithmetic
average of the CEMS value recorded at the time of each grab sample
is being allowed.
14. Section 7.3. The second sentence was deleted, and a
revised sentence was placed under Section 7.1.2.
-------
CHAPTER 3
SUMMARY OF PUBLIC COMMENTS AND RESPONSES
Commenter VI-D-1
1.1 Comment: Section 4.1. (Note: All sections refer to PS 2,
unless otherwise noted.) Setting the span at the expected range of
operation would be more appropriate.
Response: The span level must be chosen so that the majority
of the measurements will be accomplished at about midscale and
still allow the calculation of a monthly emission average or control
device percent reduction average. To obtain the latter, it is best to
select a high span level; however, the CEMS would become less accurate
at the low levels. If the span level is too low, then data would be
lost when the emissions exceed the recorder full scale. To allow some
compromises, Section 4.1 has been revised to allow spans to be set at
50 to 100 percent of recorder full scale as long as a high-level value
is read between 90 and 100 percent of the data-recorder full scale. The
high-level value is selected between 1.5 times the emission standard
level and the span (which is generally 1.5 to 2.5 times the expected
emissions).
1.2 Comment: Section 3.2. In large diameter stacks there would be
no significant difference in results obtained from three sample points
compared to a single-point sample. A more representative sample could be
drawn from a single point, or points nearer the centroid of the stack.
Response: The use of three test points was based on
"Magnitude of S02, NO, C02 and 02 Stratification in Power Plant Ducts"
-------
(EPA-600/2-75-053, September 1975) and "Sampling Location for Gaseous
Pollutant Monitoring in Coal-Fired Power Plants" [Source Evaluation
Society Newsletter 4(2), May 1979]. A single point located 1 meter or
more from the stack wall could be -11 percent off from the average
emission rate, and a single point at the centroid could be +9 percent
off. The use of three points allows measurements to within +3 percent
of the average.
1.3 Comment: The proposed revisions to Appendix B are in the
best interest of all involved in continuous monitoring. I am certain
the revisions will result in a marked improvement of monitoring programs.
Response: None required.
Commenter VI-D-2
2.1 Comment: The proposed revisions will have profound effects
on State implementation plans (SIP) across the country and, in particular,
on SIP that include provisions for use of CEMS data for demonstration
of compliance with emission regulations. Yet it is difficult to fully
evaluate the impact of the revisions since EPA has given no opportunity
to review either the CEMS guidelines or Appendix E. We feel very strongly
that no revisions should be made until those documents are proposed and
reviewed.
Although, for the most part, it appears that the proposed revisions
might be appropriate for CEMS's installed to indicate proper operation and
maintenance of pollution control equipment, EPA must realize that SIP
-------
refer to the Federal regulations when specifying requirements for
compliance monitoring. With no mandatory design or installation
specifications and precious few mandatory performance specifications,
review of proposals to install a CEMS for compliance monitoring could
not be effective. It is doubtful that expenditures of the magnitude
involved in installing a CEMS could be justified without prior indication
of the ability of the CEMS to provide accurate data. Sources and
control agencies would be at the mercy of manufacturers' claims.
Sources could be tempted to purchase inexpensive, poorly engineered
systems, which would have a very low probability of passing the
certification tests. If the system is installed and fails
certification, it is likely that the source would claim economic
hardship rather than replace or relocate the system. Thus, if the
design, installation and most of the performance specifications are to
published as guidelines only, EPA should undertake a program to certify
equivalent systems (analysis and sample interface) as is done for ambient
monitors. If EPA is unwilling to provide this assurance of probable
ability of systems to provide accurate data, all specifications should
remain mandatory for CEMS's used for compliance monitoring.
Response: The primary objectives of the CEMS PS's are evaluation of
data accuracy and representativeness. To accomplish these objectives, EPA
feels that the three specifications, which include the limits of
inaccuracy and drift, listed under Section 4 of PS 2 and Section 2 of
8
-------
PS 3 are the essential requirements; i.e., the OEMS must be accurate,
must not drift, and must provide a means for determining drift.
The proposed PS's are not designed to evaluate the installed CEMS
performance over an extended period of time nor are they designed to
evaluate adequacy of quality assurance (QA) procedures. For that matter,
neither are the present PS's. The performance specification test (PST)
procedure determines the acceptability of the RA along with the monitoring
location and the acceptability of the drift for 7 consecutive days. At
this time, the only means for evaluating CEMS's for extended periods of
time is a provision in Section 60.13(c), which allows the Administrator
under Section 114 to require the owner or operator of the affected
facility to conduct additional CEMS performance evaluations.
At the present time, QA procedures are being formulated to determine
whether a CEMS produces acceptable data (i.e., with the desired RA) on a
day-to-day basis so that the data can be used for continual compliance
purposes. As was mentioned in the preamble, EPA plans to publish such
procedures at a later date to determine precision and accuracy.
The CEMS guidelines were available in draft form in a separate
document and have been available upon request since the time of proposal
of the PS's for review. As was mentioned in the preamble, the CEMS
guidelines are to assist vendors, purchasers, and operators of CEMS's with
supplementary equipment and performance specifications, and are essentially
the specifications proposed on October 10, 1979. The CEMS guidelines are
not mandatory and are intended to aid in the selection of a CEMS.
-------
The EPA agrees that it is doubtful that the owner or operator
of an affected facility would install a CEMS without prior indication
of the ability of the CEMS to provide accurate data. Most sources
ask for references to determine the monitor's past performance and
require some form of guarantee that the CEMS would pass the PST.
The Agency feels that this is a better approach than instituting a
costly certification program or mandating the present PS's.
2.2 Comment: Section 60.13(d). The wording requiring
quantification of excess drift is unnecessary since Section 60.7(d)
requires maintenance of the calibration data in the 2-year file. If
the intent was to require reporting of excess drift, that has not
been accomplished.
Response: Although Section 60.7(d) requires maintenance
of the calibration data in the 2-year file, these data may not
provide the means to quantify excess drift. For example, not all
CEMS's have strip charts. Some use computer printouts and may give
only the adjusted output. The intent was not to require reporting
of excess drift on a regular basis, but that the data should be
available upon request.
2.3 Comment: Appendix B. The only performance specifications
that could be eliminated without jeopardizing the evaluation of a CEMS
are the 2-hour zero and span drifts. These specifications do not
10
-------
relate to normal, daily system operation. All other design, installation,
and performance specifications should be retained as mandatory requirements,
Response: We agree that the 2-hour zero and span drifts
are not usually checked during the normal, daily system operation.
We also feel that response time and calibration error fall in this
same category and these instrument checks should not be made
mandatory. The CD test is also a form of calibration
error determination and would be redundant if the calibration error
test is retained.
2.4 Comment: One general requirement which we feel should be
added for a CEMS used to determine compliance with emission regulations
is a periodic three-point calibration check. The present requirement
for a daily two-point (zero and 80-100 percent of full scale) check
would not necessarily identify problems with the calibration system
(gas or cell) or the linearity of analyzer response. The monitor could
be adjusted to read the expected upscale value regardless of whether
the system waS, in fact, operating properly. Addition of the extra
upscale point would provide a check of both the linearity of analyzer
response and the integrity of the calibration gas or cell.
Response: The use of a periodic three-point calibration
check is to evaluate the CEMS on a long-term basis. Such checks
properly belong in the QA procedures.
11
-------
2.5 Comment: (a) We agree that the sample acquisition point
for RA RM testing need not be adjacent to the monitor sample
acquisition point. In fact, the RM point should be located so as
to be acceptable for new source performance standards (NSPS) compliance
testing.
Comment: (b) We agree with the relaxation of the RA
specification when actual source emissions at normal process
operating capacity are less than 50 percent of the standard. However,
we would caution against extending this concept to the point where
monitoring results would be used only as a pass/fail indication with
respect to emission regulations. The purpose of these requirements
should not be confused with alternative types of performance regulations,
such as sulfur in fuel, which try to achieve a general nonquantitative
level of emissions.
Response: None required.
Cotnmenter VI-D-3
3.1 Comment: Section 7.2.2. When NOV concentrations are not
A
linear with time, the RM average cannot be expected to agree with the
.integrated CEMS record over the set of three grab samples. It is
our suggestion that the three grab samples be taken simultaneously
(within a 3-minute period) if possible. If for some reason it is
desirable to spread out the grab samples beyond the 3-minute period
12
-------
(up to 21 minutes), we suggest that the average RM values be compared
to the average CEMS values recorded at the time of each grab sample.
Response: Section 7.2.2 has been revised to allow the use
of the arithmetic average of the CEMS values recorded at the time of
each grab sample.
3.2 Comment: Section 7.3. This section appears inconsistent
with Section 7.1.2, which requires a 21-minute maximum for each set
of grab samples.
Response: This section has been revised to make it consistent
with Section 7.1.2. The second sentence should have been placed under
Section 7.1.2 with the word "set" changed to "run." The intent was
that three Method 7 grab samples constitute a test "run." The word
"set" refers to all necessary RM tests; e.g., Method 7 and Method 3
would constitute a set.
Commenter VI-D-4
4.1 Comment: A review of these specifications shows them to be
significantly improved over the existing specifications. The following
changes have been made:
a. Calibration Error and Zero Drift requirements have been
combined with the CD specifications. This would reduce the amount of
effort required for completing the test.
13
-------
b. The CEMS Response Time requirement has been eliminated.
c. Reference Method Measurement Location and Traverse Points
are clearly specified.
d. The 1-week conditioning period prior to starting the
PS test has been eliminated.
Response: None required.
4.2 Comment: Section 4.1. One should be allowed to use a
span gas concentration of 50 to 100 percent span. This would provide
more flexibility to the operator without any significant loss in
accuracy.
Response: See Response 1.1.
4.3 Comment: Section 5.3. There is a typing error. The second
sentence should read: "To meet the specifications, the RA must be
equal to or less than 20 percent of the mean value of the RM test data
in terms of the units of the emission standard or 10 percent of the
applicable standard, whichever is greater."
Response: This section has been corrected.
4.4 Comment: Section 6. It is not clear if the reference gases
during these tests would be introduced into the analyzer or through the
entire CEMS. Since gas cells or optical filters are allowed, it would
14
-------
indicate that the intent here is to consider the drift of the analyzer
alone.
Response: The intent is to determine the CD according to
the written procedure of the source owner or operator. The
acceptability of the procedure will be determined under the QA
procedure.
Commenter VI-D-5
5.1 Comment: These regulations will apply to all sources
subject to PS 2 and 3, including some sources subject to Appendix P
of 40 CFR Part 51. We are concerned that these proposed revisions
may result in additional expenditures to the systems installed at
our facilities, without resulting in any improvement in the accuracy,
reliability or enforcement capability of those instruments. We have
now completed installing and modifying our opacity, nitric oxide,
and diluent monitors, an effort which began in 1976, and has cost
approximately $9 million. It is our estimate that these revised
specifications could require the expenditure of an additional
$300,000 for testing, and increase maintenance costs from about
$20,000/year/monitor to $30,000/year/monitor. Other users could face
more severe relocation or replacement costs. In other words, we
strongly disagree with EPA's statement that
". . . existing CEMS that met the specifications of
the current PS 2 and 3 also meet the requirements
of these revised specifications, and therefore, do not
require retesting."
15
-------
We believe that these specifications could require replacement or
testing, depending on responsible agency's interpretation, of many of
the emission monitors installed since October of 1975. We, therefore,
still urge EPA to exclude existing monitoring installations from the
proposed revisions with the following language:
"Existing monitoring installations shall be excluded
from these revisions unless the responsible agency
demonstrates by testing the inadequacy of the installation
location, opacity calibration, or RA."
As we urged in our December 7, 1979, submittal application of
these changes to existing systems should be excluded. While the intent
is reasonable, it could be misinterpreted by many local agencies.
Specifically, RM location requirements, system accuracy requirements
and other key areas have changed. It is clear that a system which
meets the existing PS might be properly located but still require
retesting due to RM requirement changes. It might also fail a
+20 percent system error.
Response: The intent of the cited statement was to let
agencies know that they should not require retesting of all existing
monitors under the revised PSrs. Existing GEMS's cannot be exempted
from being retested because it would be contrary to 60.13(c); however,
the PSvs have been amended to prevent CEMS's from being relocated to the
case where a satisfactory correction technique cannot be established.
16
-------
5.2 Comment: We appreciate your consideration of our December 7,
1979, comments on this subject and note that many significant
improvements have been made. We still have two major concerns with this
recently revised proposal:
a. Section 3.2. The RM measurement,location requirements should
exclude the "two equivalent diameter requirement" when: (1) pollutant
concentration changes are due solely to diluent leakage, and (2)
diluents and pollutants are simultaneously measured at the same location.
This allows practical siting of RM locations in large ducts, and will
not affect the accuracy of the method.
b. Section 4.2. In over 300 RA tests conducted throughout our
system of 29 GEMS's, we have often found calibration errors in excess
of +2.5 percent of span, but RA results have been well within +20
percent of the RM. While we use 2.5 percent for testing, it is overly
restrictive for normal operation. In fact, it is often counterproductive,
requiring much more adjustment and maintenance time with no resulting
improvement in RA, but with significantly reduced reliability.
In order to provide a practical and workable rule, you should
change this requirement to +_5 percent before readjustment is required in
normal operation, or change paragraph 60.13(d) to refer to 24-hour
zero and span drifts of twice the amount referenced in this paragraph.
Response: Section 3.2 and Paragraph 60.13(d) have been
revised to incorporate the suggestions.
17
-------
5.3 Comment: The EPA states that these PS's, although related
to QA, should be evaluated on the basis of their adequacy in evaluating
the GEMS's after their initial installation. One initial test is
completely inadequate to demonstrate any suitability beyond the test
period. Our experience indicates most problems occur after this initial
test, even with good maintenance. Hence, the proposed PST cannot
adequately evaluate a CEMS without a continuing long-term (6 months or
more) test. If EPA wishes to identify such long-term testing as QA,
then it should be integral to the PST for most currently available
monitors.
In any case, we again suggest that the best course of action for
EPA is to continue to review these PS's while developing a QA program
and gathering statistically significant data on a wide range of
analyzers. Then, revised PS's and a workable QA package could be
promulgated simultaneously. Such a course of action would preclude
equipment modifications and retesting at this time which may prove to
be unnecessary or counterproductive.
Response: We agree that one initial test is completely
inadequate to demonstrate any suitability beyond the test period.
The initial PST is to demonstrate the CEMS adequacy at the time
of the initial installation or soon thereafter. The QA procedure is
to evaluate CEMSrs on a long-term basis; therefore, we do not feel
that the QA procedure should be an integral part of the PST. However,
users have the option of requiring an evaluation over a longer period
of time.
18
-------
The EPA will continue to review the PS's while developing a QA
procedure; however, the determination of acceptable QA programs will be
left to the user.
5.4 Comment: Section 2.1.4. Change title to "Data Aquisition
System." The recorder is usually only a small part of a cost effective
data system.
Response: The definition includes data aquisition; therefore,
a change need not be made.
5.5 Comment: Section 3.1. The first sentence allows correction
of CEMS data ". . . so as to be representative of the total emissions
of the affected facility." We agree with this statement. However,
Section 3.1 also states, "If the cause of failure to meet the RA test
is determined to be the measurement location, the CEMS may be required
to be relocated." We believe that this sentence should be deleted and
that data correction should be allowed in all cases where RA results
can be applied to generate a correction factor. This correction factor
would then be checked and modified during periodic QA testing at the
facility.
Response: The cited statement was intended to take care of
those cases where a practical solution for correcting the data cannot
be reached. The statement has been revised to read, "If the cause of
19
-------
failure to meet the RA test is determined to be the measurement location
and a satisfactory correction technique cannot be established, the
Administrator may require the CEMS to be relocated."
5.6 Comment: Section 3.2. We recommend inserting "at least"
in front of "three traverse points" in the minimum requirements. In
large ducts, 6, 9, or more may be needed. Allowance of added points
as needed should be explicitly stated.
Response: More than three points are allowed. Note the
phrase, "minimum requirements."
5.7 Comment: Section 3.2. The point location should also be
revised. Since "measurement line" is not defined, we assume it extends
from stack wall to stack wall. For stacks with large diameters (e.g.,
5 meters), there is no guarantee that the samples will be
representative of facility emissions. We recommend deleting this sentence
and retaining the 16.7, 50, and 83.3 percent of measurement line unless
another configuration can be shown to be representative.
Response: The measurement line is defined earlier in the
paragraph. Note that the second criterion is an option, not a mandatory
requirement. It is true that for large (also for small) stacks, there
is no guarantee that the samples taken with the second criterion will be
representative of the emissions and that the first criterion of 16.7,
50.0, and 83.3 points offers a better chance, if there is stratification.
The second criterion was based on practicality for those cases where
2Q
-------
stratification is not expected. The text has been revised to make
this clear.
5.8 Comment: Section 4.1. This paragraph is confusing as
written. We assume that it suggests setting both data system
equipment and the CEMS analyzer spans with a 90-100 percent of
i
full-scale value. While this is normal for highly linear data system
equipment, it may be ill-advised for the analyzers. In that case,
the span level should be permitted at other than 90-100 percent
without requiring Administrator approval. A span level near the
regulation limit (about 50 percent of full-scale) is suggested. Some
light absorption [e.g., nondispersive infrared analysis (NDIR)]
detectors have nonlinear outputs that are not as effective near full
scale as the response curve flattens out.
Response: See Response 1..1.
5.9 Comment: Section 4.2. While it appears that a source
may adjust its CEMS as needed during the performance testing to
meet the +2.5 percent criteria (Re: Paragraph 6) this is not
immediately clear. We suggest adding the phrase "allow daily
(more often if specified) adjustment, if necessary."
Response: The CEMS is not to be adjusted unless it is
part of the normal operating procedure. Section 6 has been revised
to reflect this.
21
-------
5.10 Comment: Section 4.3. By including potential stratification
effects within the RM test as well as pollutant and diluent errors
within the +20 percent you have effectively made a +30-50 percent
requirement much more severe. While we successfully use such a standard
on our sources, it is not clear that this is economically reasonable
or technologically feasible. In any case, we suggest that this be
demonstrated - on a variety of sources - before forcing utilities into
an overly expensive and time consuming program which may never work as
intended.
Response: Many CEMS have met the +20 percent criterion for
combined pollutant-diluent systems. If a CEMS is not capable of meeting
this requirement during the initial PST, it would certainly be a poor
candidate for a long-term reliable operation. See also Response 8.3.
5.11 Comment: Section 5.2. Load or capacity should be allowed
to vary over the full range that it does in normal operation, including
nearly full loads, but not limited to 50 percent. If loads of less than
50 percent are typical of unit operation, it is important to test at
50 percent load. Large errors may often be introduced by negative duct
pressures, different temperatures, etc., at low loads. Further, the
50 percent restriction places an unreasonable burden on existing units.
22
-------
Response: This section has been revised to 50 percent
of normal operation.
5.12 Comment: Section 5.3. It is not essential to hold
the RA test period until after completion of the CD test. This
should be up to the source, based on experience with the monitors
under test and scheduling needs.
Response: The section has been revised to allow RA
testing during the CD test period.
5.13 Comment: Section 6.0. The incremental addition
procedure should not be tied to 80-95 percent of span, but rather
to an appropriate incremental cell which will provide the most
meaningful demonstration of the analyzer's operation. The value
will depend on stack concentrations and the shape of the Beer's
law curve for the given analyzer.
Response: The third paragraph has been deleted. Since
this case is mentioned in Section 4.1, it is not necessary to
make mention of it again.
5.14 Comment: Section 7.1. In our 3 years of RM testing, we
have found that it is relatively simple to pass existing PST's at one
load or boiler condition - but that in our boilers, where loads and
23.
-------
fuels vary widely, the RM tests need to be done over the full range
of operating conditions. We suggest that the tests be made over a
representative set of conditions and a long period of time, e.g.,
several months.
Response: This is one option that we have decided not
to mandate because of cost considerations. To write the PS's so
as to cover all source conditions would require the most stringent
approach, which can be costly and may be unnecessary for some or
the majority of the sources. However, the sources, understanding
their peculiar problems, may choose to use the more stringent
procedures to ensure the specified RA requirement.
5.15 Comment: Section 7.1. If the source operating
conditions vary significantly with time, the calculated emission rate
may have no relationship to the actual emission rate unless pollutant
and diluent measurements are taken simultaneously. We recommend
requiring simultaneous measurements unless the source operator can
show that time/pollutant variations are insignificant.
x
Response: See Response 3.1.
Commenter VI-D-6
6.1 Comment: The proposed rules are certainly consistent in that
the only method of detennining the acceptability of a given instrument
24
-------
is with respect to the RM's. Unfortunately, during the last 5 years
we have not seen any advancement in the accuracy, repeatability,
reliability or cost associated with the RM's. In the same period,
we have seen substantial improvements in our gas measuring instruments
to the point where, in many cases, they are a much better criteria for
establishing emissions than are the RMfs.
At the current time and under the proposed specifications, there
is no criteria which either qualify or motivate the stack sampler
to do an adequate job. We recognize that there have been interlab
comparison studies which have attempted to evaluate RM accuracy, but
how valid is the data so obtained when the intent is known, as
opposed to running a stack sampling exercise in the field with
sometimes poorly trained personnel and under trying, demanding and
adverse conditions? Since EPA has established the RMrs, it seems
that they have been very wary of quantifying inherent errors or
attempting to provide quality checks to ensure that inadequate RM
data are not used in any of the required test evaluations. In particular,
RM 7 has been nearly universally proclaimed by stack samplers and
users alike as a very imprecise method. Further, the American Society
for Testing and Materials (ASTM) describes several biases and
interferences for this method which are never mentioned in the EPA
descriptions. At the current time, we believe that RM 7
may exhibit a substantial negative interference due to chlorides
which are present in the outlet of wet scrubber type flue gas
25
-------
desulfurization systems. Reference Method 6 is certainly better than
Method 7, but still is not ideal and their consistency is, in no way,
comparable to National Bureau of Standards (NBS) reference materials.
At a minimum, under the above conditions, the RM should be verified
using a traceable NBS gas, and only that RM which meets the blind
gas sample check should be acceptable for verifying an instrument.
All manual stack sampling reference data should be qualified during
each test using blind gas samples. If the errors are believed to be
consistent during a given run, the RM data should be corrected according
to the error observed in the blind sample. Further acceptance criteria
should be established in terms of the uniformity of individual data
points established in a given run.
Response: The commenter's primary contention is that NBS
reference gases are better checks for accuracy than the RM's on the
basis that RMfs are "fickle." It is true that the RM's can give
erroneous results when the procedure is not properly executed. This is
also true of a CEMS. The Agency is presently in the process of
including QA procedures into Methods 2 through 8 to ensure high
probability of obtaining valid data.
The Agency considered for Methods 6 and 7 a periodical check against
a known gas cylinder concentration. At this time, however, we
feel that qualifying testers at each test in the use
26
-------
of the RM's against blind gas samples would be costly. We feel that
the option to use such a procedure should be left to the vendor and
the user.
We see no reason why the RM should be verified time-after-time
against traceable NBS gases. The methods have been established as
RM's and have been proven to provide accurate data, when properly
executed. It would seem prudent to inquire into the tester's
experience and secure testers with an established reputation of
providing accurate and precise results to run a stack sampling exercise
in the field under trying, demanding, and adverse conditions. If the
RM's are properly conducted and the quality control procedures in
the methods are carefully followed, the results will be accurate
measures of emissions.
We agree that Method 7 is less precise than Method 6. The
imprecision of these methods has been considered when the +20 percent
criterion was established.
Method 7 does not mention the chloride interference mentioned in
the ASTM method since high concentrations of chlorides were not expected
in nitric acid and power plants. About a 1:1 ratio of chloride to NO
A
is required before there is any appreciable reduction in concentration. '
6.2 Comment: Because of the ability to calibrate these instruments
dynamically, i.e., at stack temperature and pressure, using either direct
or traceable NBS reference gases, instruments can be made accurate, and
measurements very repeatable. Using the criteria you have proposed, one
would be forced to throw away the only consistent and accurate standard which
we have and attempt to design instruments to meet a fickle RM — this is
27
-------
certainly not in the best interests of regulator, vendor, or user. The
reason I say that you have essentially outlawed traceable gases in
your proposal is that in the event that an instrument fails the RA
(RM correlation) test, the users' most promising approach
is to reset the calibration so that it is supposedly equal to the
past RM data and run the test again. This does not achieve the desired
results of consistent, accurate calibration.
For those instruments whose calibration can be validly checked
using direct or traceable NBS gases, the testing with respect to existing
RMTs is a redundant and an unnecessary, costly exercise. It may be
reasonable to use RM's sometimes to evaluate the potential stratification
effects on the CEMS, but the NBS gases should be an acceptable measurement
qualifying media. Certainly not all instruments can be validly checked
with NBS gases under dynamic stack conditions, but those which can,
should be permitted to be operated with certification criteria which are
relevant to the capabilities of such instruments. We strongly recommend
•that the EPA should investigate the capabilities of such instrumentation
and determine if the desired accuracy and consistency can be obtained
with less costly testing procedures.
Response: We would tend to agree that the capability
to dynamically calibrate instruments is a plus in instrument
design. However, calibration gases are not identical in composition
as the stack effluent being monitored. In addition, in several
28
-------
cases, when differences were found between the OEMS and the RMrs,
the cause was traced to an incorrect moisture correction, moisture
interference, or a plugged filter, i.e., dynamic calibration, by
itself, is not sufficient to ensure accurate results.
The use of direct or NBS reference gases is not prohibited.
The PS merely allows greater flexibility of meeting the RA
requirement. For example, there are users of instruments who
obtain accurate results by using gas cylinders that are analyzed
by the RM's. The use of such cylinders is less costly than NBS
reference gases.
Presetting the calibration is not improper if improper calibration
is determined to be the cause for the difference between the CEMS and
the RMrs; however, resetting the calibration based on erroneous RM
results would not be prudent.
We agree that for those instruments whose calibration can be
validly checked using direct or traceable NBS gases in lieu of
the RM's would be a redundant, unnecessary, and costly exercise.
However, how does one determine what "constitutes" a "valid" check?
As mentioned earlier, several cases showed that the use of NBS
calibration gases by itself is insufficient to check the accuracy
of a CEMS. All CEMS operators should establish and demonstrate that
their procedure is adequate to "validly" check the CEMS. After the
demonstration,, one can apply for an alternative procedure under
Section 60.13(1) of 40 CFR Part 60.
29
-------
6.3 Comment: In several places in your proposed values, you
have reiterated the RA criteria in terms of +20 percent of the
mean reference value or +10 percent of the applicable standard,
whichever is greater. This specification is to include the pollutant
and diluent monitor, as well as the ppm to Ib/MBtu converter and
recorder. We certainly agree that we should continue to push for more
accuracy as long as it is not unduly costly, but this specification is
overly demanding, considering that the RMrs are far from absolute
and that there certainly is less intrinsic value associated well
under basic requirements.
If the proposed criteria are promulgated, it will certainly
encourage any informed operator to operate as near the emission limit
as possible during RA testing to ensure an error window of 20 percent.
At emission levels of one-half the emission standard, the +10 percent
of standard becomes effective and remains the effective criteria as
emission levels approach zero. This penalizes the conscientious operator
and substantially increases the cost of certification.
Response: The relative error criterion of +20 percent of
the mean RM value is not a proposed revision, but is the promulgated
requirement. The +JO percent of the applicable standard is the proposed
revision. This criterion was included to alleviate—not penalize—the
"conscientious operator." Let us assume that the standard is 500 ppm
30
-------
and the source is operating at an emission level of 100 ppm. If the
+.20 percent criterion is used, the CEMS must operate within +20 ppm.
However, with the +10 percent of standard criterion, the CEMS must
operate within +_50 ppm or +50 percent of the RM mean value.
6.4 Comment: It is important to note that an increasing number of
scrubber applications provide emission levels of 15-50 ppm S02 to meet
the 70 or 90 percent removal criteria. Considering the fact that,
typically, these instruments are scaled for span values of 300 to 750,
these readings are a very small percentage of full scale (2 percent to
16 percent), the error allowance in terms of ppm, not percent of reading,
is very important. As a result, I would propose that the two-tiered
accuracy criteria be eliminated. We would suggest that for a scrubber
inlet monitor, the recommended +20 percent of the emission limit be
applied to the actual inlet concentration or RM value. For an cutlet
monitor, the recommended +20 percent of the emission limit be applied
to the product of the inlet concentration multiplied by the required
removal efficiency.
I firmly believe that your proposed criteria, as compared to the one
recommended above, are overly demanding and will only cost the user more
in terms of repeat testing and will not provide any improvement in
overall instrument accuracy. We favor treatment of the system as a
whole, but allowances must be sufficient to provide adequate margins
for the RM's, each instrument (diluent and pollutant), and associated
hardware. Your proposed +_10 percent of the emission standard
31
-------
does not leave sufficient margins under any, except the most ideal,
conditions.
Response: We agree that high span values would cause higher
relative inaccuracies at the low concentrations. However, the use of
a standard +20 percent of the emission standard would penalize
instruments used on the scrubber inlet, especially if the emission
standard is low in comparison to the inlet concentration. In addition,
instead of encouraging any improvement in overall instrument accuracy,
the suggested criterion, which increases the allowable inaccuracy, would
seem to encourage greater inaccuracy. To improve the instrument accuracy
at low levels, it would be better to simply allow lower span gases
to be used. To avoid retesting, the CEMS should be reliable and
without any biases; and competent testers should be used.
For the second recommendation, the correct emission limit should
be the product of the inlet concentration multiplied by the allowed
removal inefficiency—not required removal efficiency.
6.5 Comment: Another factor needs to be clarified in this general
RA area. A scrubber operator may be required to meet a 70 or
90 percent S02 removal efficiency as well as a gross 0.6 to 1.2 Ib/MBtu
ceiling. In terms of your proposed +10 percent of the emission standard
or our recommended +20 percent of the emission standard, specifically
what is the emission standard for an inlet and outlet S0? scrubber
monitor?
32
-------
Response: The criterion is +20 percent of the RM mean value
or +JO percent of the emission standard, whichever is greater.
Therefore, at the inlet of the scrubber, the +10 percent of the emission
standard criterion would not be expected to be used, but the +20
percent of the RM mean value would be used. For the outlet, the emission
standards are defined by the regulations. See Subpart Da, Section 60.42a.
For example, for sources under 60.43a(a), the emission standard is
1.20 Ib/million Btu, or 10 percent of the potential combustion
concentration, or 0.60 Ib/million Btu, or 30 percent of the potential
combustion concentration.
6.6 Comment: In terms of determining scrubber efficiency, it is
extremely important that the inlet and outlet monitors be referenced to
a common stable reference. Again, the only consistent standard of which
we are aware is an NBS gas. Without a common reference, the inlet and
outlet monitors could be mi sealibrated by up to +20 percent of
reading, which could result in up to a 15 percent error in scrubber
efficiency, indicating 55 percent removal when the actual removal is
70 percent.
Response: We agree that the inlet and outlet monitors should
be referenced to a common stable reference. Of course, we feel that
at present the RM is that reference. In calculating the maximum
error that could result from mi sealibration, the commenter erroneously
assumed that the monitor can be miscalibrated by +20 percent. Such
33
-------
a monitor would not pass the RA test, unless the RM gave very precise
numbers, an unlikely event. A more reasonable maximum bias is about
+10 percent. Instead of seeing a possible maximum error of +15
percent at the 70 percent level, one would see about +7 percent at
the 70 percent reduction level and +2 percent at the 90 percent
reduction level.
6.7 Comment: In paragraph 60.13, you have required operators
of CEMS to check the zero and span drift at least once daily -- and to
readjust the zero and span whenever the 24-hour zero and span drift
exceeds the limits of the applicable performance specifications.
Since there are no criteria for what constitutes a check of zero and span
and since operators do not wish to readjust anything any more often than
absolutely necessary, the aggressive instrument manufacturer or user would
design this check to include as little of the overall measurement system as
possible. In fact, by using a relay to switch the drive signal input to the
final current driver, one could ensure that the zero and span checks
would be stable and that the instrument would virtually never
need readjustment, at least as-so far as "zero/span" drift values
are concerned. Obviously, this situation is not consistent
with the basic intent of the specification and must be corrected. We
would suggest that the following statement be added in 60.13(d): "The
zero and span drift check must, at a minimum, include all the active
components used or referenced in the normal stack gas measurement cycle,
except that predictable effects, such as stack temperature compensation,
34
-------
which would otherwise cause the zero and span values to vary, can be
checked independently. The zero and span drift check values shall
normally remain invariant from day-to-day if the instrument does not
exhibit any drift." This statement would ensure that zero and span
drift values provide a valid indication of the stability of the
measurement system.
Response: The basic approach of the PS's is to state what
is expected of the CEMS and the manner in which the specifications
are to be evaluated. This was done to allow greater flexibility in
the design and selection of the instruments and to allow for innovative
technology.
The first criterion, the RA, is checked with the applicable RM's.
The second criterion, the calibration drift at the zero and span levels,
is left up to the manufacturers and the users. If the manufacturer
designs an instrument to include, as you indicate, as little of the
overall measurement system as possible and if the user accepts his
design, the user is responsible for the quality of the data. The
Administrator, under Section 60.13(c), can require the owner or operator
of the affected facility to conduct additional CEMS evaluations. For
those sources that must continually comply with the emission standards,
the QA procedure will most likely involve periodical' RA checks. These
RA checks will ascertain the adequacy of the drift check procedures.
If the manufacturer does not properly design his instrument to operate
reliably and the user accepts such instruments, the user in the long
run can be subjected to costly repairs and tests.
35
-------
We agree that instruments should be designed so that the zero
and span calibration checks include all the active components; however,
we cannot say that this requirement is absolutely necessary for all
designs. The guideline will be revised to include these points.
6.8 Comment: Section 6. The effects of automatic calibration
and correction do not appear to be anticipated in this proposal. In
one of our typical automated CEMSrs, the zero and span check values from
all instruments are individaully recognized and the appropriate zero
and span calibration correction can be applied to all succeeding data
until the next calibration cycle. In this case, no readjustment of
zero and span should be necessary until the correction exceeds the
manufacturer's recommendation. This means that no zero or span drift
would be seen in the final reported emission values.
Your proposal to eliminate the potential advantages of automatic
calibration correction eliminates one of the more cost-effective tools
with which the instrument designer has to work — digital electronics.
We recommend that automatic correction be allowed, but with
qualification. When automatic correction is allowed, the automatic
correction should be monitored and the drift in correction should not
exceed three times the normal drift allowance, provided that the
corrected output is less than one-third the normal drift allowance.
Response: We are not disallowing auto correction. All we are
asking is how much is the correction.
36
-------
6.9 Comment: I would like to point out that qualification of
the zero and span drift integrity has been included and is existing for
opacity monitors -- but not gas analyzers. In 60.13(d), it states that,
"Such procedures shall provide a system check of the analyzer internal
optical surfaces and all electronic circuitry, including the lamp and
photodetector assembly." Unfortunately, the intent of this statement
is admirable, but 1t is openly violated by at least one model of
transmissometer which has been "certified" in some applications and
tacitly approved by the EPA. This fact is primarily important from the
fact that EPA must establish a reasonable floor for instrument
performance and apply it uniformly or the market will become distorted
by less expensive imitations which can get by and will be purchased on
a strictly cost-selection criteria. The eventual outcome of such actions
is an unhappy user and data of less real integrity than that specified.
Response: The specification in 60.13(d) for opacity monitors
is given because there is no easily applied RM for transmissometers.
Manufacturers of different instruments can request a determination if
certain procedures meet the intent of the regulations in principle.
Approval of such procedures, however, does not "certify" the instrument
nor does it mean that the instrument will continuously operate with or
without loss in integrity.
6.10 Comment: The argument with respect to whether the
span calibration check should be done near mid-scale, 50 percent
37
-------
or full scale, 90 percent, is somewhat mundane. From a technical
standpoint, the closer the check is to full-scale, the better
will be the check of the real span drift component as opposed to
zero drift. From the practical viewpoint, the span check material
is relatively difficult (costly) to manufacture to a specific value.
Since the desired effect is to ensure accuracy in the vicinity of
the emission standard, we would suggest that the span check be
performed at a value which is a minimum of 1.5 times the emission
1imit.
Response: See Response 1.1.
6.11 Comment: In PS 2, paragraph 6, you have specifically
allowed a single-point incremental addition type check of
instrument drift. It is very easy to prove that it takes two
points, at a minimum, to define a linear calibration function.
Normally, .these two criteria are defined as gain and offset, which
correspond to EPA's typical span and zero parameters, respectively.
By using only a one-point definition of a calibration function,
you have made it impossible to predict the performance of the
instrument at any other point on the calibration "curve." We
strongly recommend that a single-point drift check only be allowed
if the user can show that either the zero/offset or gain/span
characteristic is stable (non-variable) and predictable for that
individual instrument. It should be noted that all gas analyzers
38
-------
can be designed to incorporate a zero and span check. In some designs,
it is more expensive than others, but this should be a required
characteristic if the EPA really desires a consistent specification
which will ensure a reasonable degree of accuracy. We would recommend
that the EPA consider the experience of Pacific Gas and Electric in
San Francisco in their attempt to obtain reasonable performance from
their across-stack gas analyzers. It is our understanding that the
rotating zero pipe has been necessary to accurately define the
performance of such systems. A paper was presented on the subject at
the 1979 APCA Show by Bob Webb of Pacific Gas and Electric.
Response: We agree that a single-point drift check should be
allowed only if the user or manufacturer can show that either the
zero/offset or gain/span characteristic is stable and predictable for
that individual instrument. In the case in point, the vendor did exactly
that. It is our understanding that the one-point check was sufficient to
determine calibration drift.
6.12 Comment: It is not clear what the allowable zero and span
drift are without adjustment, considering that the PS 2
criteria involve statistical means and confidence
intervals. These data are not readily available on a day-to-day basis.
39
-------
Response: Section 4.2 specifies the allowable zero and span
drifts. The specification criteria do not involve statistical means
and confidence intervals.
5.13 Comment: It appears from this and previous proposals that
the EPA has assumed all devices are equally inadequate, not capable of
being easily checked with a high degree of integrity or predictability,
and have settled on a rigorous correlation with RM's as the only
method of determining compliance with basic accuracy requirements. This
means that all measurement techniques, even those including the highest
degree of built-in verification means, must also comply with the most
costly means of verification (RM testing) which are really only relevant
to the instruments with the least capability. In the interests of
our economy and the often-suggested high cost of environmental monitoring,
we sincerely believe that it is necessary for EPA to consider lower cost
means for certification of instruments with greater capabilities for
verification. From another viewpoint, it may be much more desirable
for a user to purchase more instrumentation verification capability on
the front end to reduce future certification, maintenance and QA costs.
Additionally, it is certainly much more inexpensive to purchase and use
traceable calibration gases than to become involved in repeat testing
according to RMKs 6 or 7.
At the current time and under the proposed rules, there is no incen-
tive to evaluate this trade off, since all devices are assumed to be
40
-------
marginal and certification costs are the same high value for all devices.
For example, all across-the-stack devices could be equipped with rotating
zero pipe facilities to provide zero and span checks with a high degree
of integrity. Without this calibration means, one is justified in
mandating rigorous RM testing in order to ensure accurate results. Another
example involves the use of NBS traceable zero and span check reference
materials. With nontraceable materials, one is left with a high degree
of suspicion regarding the meaning and integrity of such calibration checks.
With traceable references, the calibration checks can provide much higher
real calibration accuracy. A third example involves the potential validity
of a dynamic calibration check, using NBS traceable gases at stack
temperature and pressure. Without this means, a high degree of RM testing
may be necessary to ensure valid results. With this capability, the use
of traceable gases can provide a higher degree of integrity than will ever
be achieved with wet-chemistry, operator-dependent, RMrs. We would welcome
the opportunity to discuss the above options with the EPA in more detail.
We would like to make it clear that we believe that we can ensure a
higher degree of accuracy and substantially lower certification costs
(50 percent) if the EPA will consider certification means which are more
proportionate to the inherent built-in or installed capabilities of the
instrument.
Response: The Agency has not assumed that all devices
are equally inadequate. If this was true, then EPA would not have
required the use of CEMS for monitoring purposes or for continual
41
-------
compliance purposes. The Agency has concluded that the best means of
checking the instrument for accuracy is the RM tests. This, however,
do.es not preclude alternative and cheaper means to be used for checking
RA. Those measurement techniques with "the highest degree of built-in
verification means" would be more likely to be approved than those with
the lesser capabilities. And EPA will consider lower cost means for
certifying instruments if it is adequately demonstrated that such
procedures will provide the proper evaluation. The QA procedures
can be used for these demonstrations.
The EPA encourages users to consider those GEMS's that can easily verify
the accuracy, which would reduce future certification and QA costs, in
addition to its demonstrated reliability and cost of maintenance.
We agree that the use of traceable calibration gases is less
expensive than to become involved in repeat testing using RM's 6 and 7.
However, it should first be established that the traceable calibration
gases are sufficient to insure the required accuracy. See Response 6.2.
As was mentioned in the preamble to Subpart Da, EPA will be proposing
QA procedures to require the determination of precision and accuracy of
the CEMS. These procedures should provide the incentive to obtain reliable
instruments.
6.14 Comment: We believe your proposal is weak in one other area,
which is necessary to ensure accuracy measurement at emission levels
other than the one encountered during RA testing. This is the area
42
-------
of linearity, or nonlinearity, as the case may be. We would
recommend that the linearity or nonlinearity should be documented
by the equipment manufacturer at the factory. We would recommend that
if the manufacturer represents an instrument as having a linear
calibration curve, that a test similar to your existing RA test be
performed, except that that data should be evaluated in terms of its
deviation from a straight line. For such linear instruments, the
maximum deviation should not exceed +3 percent of span. For nonlinear
calibration curves, the vendor should demonstrate that the published
nonlinear curve is accurate to within +3 percent of span for that
individual instrument. The NBS traceable materials should be required
for this test. The calibration may exclude temperature compensation
if such can be shown not to affect the linearity or nonlinearity of
the system. The manufacturer should thus provide, in essence, a
certificate of calibration. Further, the manufacturers' maintenance
procedures should be directed toward verifying or maintaining the
given calibration function.
Response: The Agency agrees that linearity or nonlinearity
should be established and periodically verified. However, EPA is
leaving the option of whether to have such procedures or certifications
to the user.
6.15 Comment: We definitely believe that these proposed regulations
and revisions should not be applied retroactively. It may be reasonable
43
-------
to make them effective for instrumentation purchased after the date of
the proposal, but certainly not broadly retroactive. The basic
reason for this concern is the revised RM certification testing
which could require changing the installation site in some circumstances,
and the tighter accuracy specifications which are inherent with the
proposed consideration of pollutant and diluent monitor, ppm to
Ib/MBtu conversion, and recorder, as a system which must fall within the
same error band as previously allowed for just the pollutant monitor.
There are substantial changes which become even more significant when
considered in light of the potential errors and costs associated with
the RM testing. It may be reasonable to mandate updating to this
specification within 5 years after date of the initial certification.
Response: See Responses 5.1 and 8.3.
6.16 Comment: Section 2.6. The validity of the "established
reference value" is totally undefined, as discussed previously. Such
values can be estabished to provide minimum measurable or visible drift,
or to provide a valid check of the operation of the entire measurement
system.
Response: This is the reason that RA tests are conducted with
RM's. See also Response 6.7.
44
-------
6.17 Comment: Section 2.1.4 The last sentence needs to be
expanded as follows: "The data recorder may include automatic data
reduction and correction capabilities."
Response: Automatic data reduction does not preclude
correction capabilities. The suggested addition need not be
made.
6.18 Comment: Section 4.1. The single-point calibration
drift determining technique is wholly unsatisfactory, as previously
discussed, unless the operator can show that the offset or gain of
the system is nonvariable. There is a noticeable inconsistency
here in that you have tacitly allowed single-point calibration checks
without qualification and are, at the same time, specifying overly
tight limits on the range of allowable zero and span limits. If
checking is not possible at zero and span, the lower value should be
between 0 and 33 percent of full scale, and the upper between 66 and
100 percent.
Response: See Responses 6.11 and 1.1.
6.19 Comment: Section 5.2. There should be no restriction
for plant operating capacity as proposed. It may be reasonable to
require that the plant operate at more than 50 percent capacity for
at least 3 days out of 7, but certainly not all the time.
Response: See Response 5.11.
45
-------
6.20 Comment: Section 5.3. (a) The CD and RA tests should
be allowed to be conducted simultaneously to reduce costs, (b) One
test failure does not generally invalidate the results of the other.
(c) The last sentence could be somewhat restrictive depending on
the meaning of the word, "demonstrated." We prefer that the last
half of the sentence be modified to state, "... unless the zero
and span calibration check values are shown to be unaffected." If
one part of the system fails, the RA testing should be allowed to be
completed for the remaining available data, as long as no more than
one major system component has failed, (d) We would propose that
for diluent/pollutant measuring systems, the error criteria be
tightened up by Mfi. or 0.707, compared to the combined system tests.
Thus, an individual diluent monitor could be certified at 0.707 times
20 or 14.14 percent error allowance, as could the pollutant monitor
alone.
Response: (a) See Response 5.12. (b) It is true that
one test failure does not generally invalidate the results of the other.
This depends on the system, (c) As mentioned earlier (see
Response 6.2), the suggested modification is not sufficient to
ascertain whether adjustments made to one channel does not affect the
others. This also includes failure of major components. The degree
of restrictiveness of the word "demonstrated" depends on the particular
CEMS. (d) See Response 8.3.
46
-------
6.21 Comment: Section 7.1. It is very important that if the
plant operating conditions, such as load, fuel and excess air, are
not reasonably constant (+10-20 percent) during the RM tests, and
that the diluent, pollutant and moisture are determined as near
simultaneous as possible.
Response: See Response 3.1.
6.22 Comment: Section 7.3. We support the note in this
paragraph.
Response: None required.
6.23 Comment: Section 3.2 of PS 3. We question the accuracy
of Method 3 in its basic form. In particular, this is an area where
a dynamic calibration check, using NBS traceable gases, should
be allowed in lieu of Method 3, if the diluent analyzer is tested
alone per the tighter-suggested RA error allowance (14.14 percent).
Response: See Responses 6.1, 6.2, and 8.3.
Commenter V1-0-8
8.1 Comment: These comments on the proposed 'revisions are
submitted on behalf of the Utility Air Regulatory Group (UARG). The UARG
is composed of the Edison Electric Institute, the National Rural Electric
Cooperative Association, and 85 individual electric utilities. A list
of UARG's members is enclosed as Enclosure 1.
The UARG filed comments on the October 10, 1979, proposed revisions
in December, 1979. Our review of the reproposed PS's indicates that the
Agency has responded to several of the earlier UARG comments.
47
-------
We would like to suggest certain additional changes which appear
necessary as a result of the changes which the Agency has made since the
earlier proposal. The UARG comments are discussed in some detail in
Enclosure 2. I would like to take the opportunity to briefly summarize
in this letter some of the more important comments.
Continuous monitoring technology is a developing science.
The problems with monitor accuracy and reliability have been well
documented by both UARG and the Agency. Given the state-of-the-art
of monitoring technology, it is critical that any PS's be flexible
enough to allow sources to respond to unanticipated problems and
special circumstances. We, therefore, applaud the increased
flexibility which the Agency has introduced in the reproposed PS's.
At the same time, however, we encourage the Agency to issue at an
early date the monitoring guidelines referred to in the January 26, 1981,
Federal Register publication, in order to provide vendors, purchasers,
and operators of continuous monitors with supplementary equipment and
performance specifications. While we agree that the state-of-the-art is
not adequate to incorporate more specific guidelines in regulatory language,
more detailed guidelines would be helpful as a starting point for continued
discussions among the Agency, monitor users, and monitor vendors.
While we agree with the general approach taken in the reproposed
revisions to Appendix B, we have a variety of concerns with particular
aspects of this reproposal. These concerns are listed briefly in the
following comments.
48
-------
Response: The CEMS guideline document was available on
January 26, 1981. However, based on the comments received during the
comment period of this rulemaking, the document will be revised and
republished as soon as possible.
8.2 Comment: The reproposed PSTs address only the location of
outlet monitor systems. For sources with pollution control technology,
an inlet as well as an outlet monitoring system may be required. The
Agency, therefore, must revise these reproposed PS's in order to
account for questions concerning location and testing of inlet monitor
systems, and propose these specifications for public comment.
Response: The text has been revised to address inlets to
control devices. However, we do not feel it is necessary to propose
these specifications for public comments as the same principles apply
as to obtaining representative samples. In addition, the revisions
should not pose any problems.
8.3 Comment: The problems with continuous monitor accuracy are
well documented. In spite of these well-known problems, the reproposed
specifications require that both pollutant and diluent monitors in
combination meet the +20 percent accuracy requirement, even though the
existing PS's only require the pollutant monitor to meet these accuracy
problems. In addition, the reproposed specifications eliminate the
10 percent location error allowed in the October 1979 proposal. We
do not understand the basis for this departure from the existing PS's
49
-------
and earlier proposal. In light of the well-known problems with monitor
accuracy, the Agency should allow +10 percent error for monitor location,
as well as an error for diluent monitors.
Response: In the October 6, 1975, PS's, only the individual
pollutant monitors were required to meet the specified RA. No accuracy
requirement was placed on the diluent monitors. This was an error on
EPA's part. At that time, EPA felt that the zero and span checks were
sufficient to ensure that the concentrations obtained by the monitor
would have yielded results within +0.5 percent 02 or C02. However,
experience has shown that this was not the case.
The combined accuracy requirement is more restrictive. However,
as mentioned in the preamble, this requirement is not unattainable.
The Agency has reviewed a number of RA tests where the combined systems
were checked. Of the 24 tests that were reviewed, 13 improved its RA
and 11 got worse. But only two would have failed as a result of the
combined specification and five would have passed while failing the
pollutant monitor RA. The point is that CEMS's are capable of meeting the
combined specification.
Some monitors which presently meet the existing RA specifications
may not be capable of complying with the proposed accuracy specification.
However, this could be a result of monitor deterioration or lack of
maintenance. If a CEMS cannot be corrected outside of full system
replacement, the system is probably so inaccurate that it serves no
50
-------
useful purpose and would be advisable to have it replaced; the remedy
is not to increase the relative error specification.
The +10 percent criterion used in the definition of stratification
was not an allowance for error. The entire context of Section 4 of the
October 5, 1975, specification (40 FR 46250) is that the sampling location
should be representative of the emissions. Section 4.1 states, "For
sampling locations where effluent gases are ... nonstratified, a point
or path of average emissions may be monitored." Section 3.2 of the
October 10, 1979, proposal (44 FR 58602) states, "The tester may ...
conduct the stratification check ... to select the point, points, or
path of average gas concentration."
3.4 Comment: The Federal Register proposal states that "some
commenters erroneously assumed that QA procedures were an integral part
of the specifications." The Agency, therefore, concludes that it is
unnecessary to issue QA procedures before finalization of the PS's. The
UARG remains convinced that until we understand the procedures which are
required to maintain monitor accuracy over time, application of detailed
PS's at the time of monitor installation provides no guarantee of continued
monitor accuracy. We understand that there is little data at this
point on which the Agency could base the development of QA procedures.
The Agency, however, should recognize the seriousness of this issue in
any PS's it issues.
Response: The PS's are not designed to evaluate the CEMS over
a long period of time. Neither are they designed to establish the adequacy
of QA procedures that are to insure valid short-term and long-term
51
-------
monitoring data. The PS's merely specify what is expected of the
instrument, e.g. in PS 2, the calibration curve should remain stable
and the measurements should be accurate within +20 percent when
compared to the RM's. The PST provides the check necessary to
determine whether the CEMS is within the requirements at the time
of the initial installation. The QA procedures are intended to
demonstrate that the CEMS is properly maintained. Because EPA
recognizes that neither the PST nor the PS's guarantee continued monitor
accuracy, EPA is developing the QA procedures to evaluate the CEMS
over longer periods of'time. If the user desires to include an
evaluation of the CEMS over a long period of time along with the initial
PST or to institute other procedures, he may choose to do so. He may
also choose to include procedures to insure that the selected sample
point(s) remain representative, or unbiased, with respect to total
emissions.
8.5 Comment: The UARG is concerned with how the location procedures
specified in the reproposed specifications will affect new units. While
the application of these procedures to existing units is possible since
testing can be done, a new unit will need adequate time to test monitor
locations after the unit starts up. Industry experience indicates that
monitor installation and the testing required to establish a proper '
location can take 1 year or more after the initial startup and debugging.
The PS's should deal with this concern, and other regulations may require
modification as a result of these proposed procedures.
52
-------
Response: Taking a year or more to establish a suitable
location is unusual. However, if delays in meeting the present
requirements are justifiable, Section 60.13(1) may be used to establish
alternative requirements.
8.6 Comment: The reproposal does not specify that the drift test
procedures, and the results of the drift test must be approved prior to
the RA test. If the Agency reviews apparently successful calibration
drift tests after the completion of RA tests; and, if upon such review,
the Agency determines that the procedural requirements and specifications
of the drift test was not achieved, sources may be required to repeat
both the drift and accuracy tests. To help avoid this type of situation,
we recommend that sources be given the discretion of having the
procedures and results of the drift tests evaluated and approved by
the applicable agency before initiation of RA testing. The timing for
such approval should be decided by the source, since they will be taking
the liability for having to reconduct all tests if it is later found
that the drift test was failed.
Response: The PS's do not prohibit such discretion.
8.7 Comment: The CD test is based on a single measurement of
a calibration gas or gas cell each day. Since the imprecision of any
one measurement could cause a monitor to fail the drift specification,
it is recommended that calibration medium(s) be analyzed three
times each day by the subject monitor and that the results of the
three analyses be averaged and compared to the reference value of the
53
-------
calibration medium to determine CD. This repetition of analysis should
minimize the possibility of imprecision causing drift test failure.
Response: The imprecision is considered in the CD specification;
therefore, averages need not be used.
8.8 Comment: The drift specification, 2.5 percent of the span
value is an absolute specification, i.e., 25 ppm drift is 2.5 percent of
1000 ppm span and 12.5 ppm is 2.5 percent of 500 ppm, etc. The absolute
specification may be adequate for detecting a parallel shift of the
calibration curves away from the initial, or accepted, calibration curve.
However, if the slope of the calibration curve changes or "drifts,"
the absolute drift specification may be inadequate to quantify the
actual drift at the calibration or span range. For example, when the
slope of the calibration curve changes by 2.5 percent, the CD will show
a 25 ppm change at the 1000 ppm level, 22.5 ppm at the 900 ppm level,
20 ppm at 800 ppm, and 17.5 ppm at 700 ppm. As seen here, determination
of compliance with an absolute drift specification can vary when
differing calibration values are used for the drift test. We suggest
that in order to more accurately assess CD, the drift specification
should be expressed relative to the value of the calibration gas or gas
cell.
Response: The CD value does not vary with the calibration
values. The "span value" is fixed and specified in an applicable subpart
of the regulations.
54
-------
8.9 Comment: The calibration error specification is important
in evaluating monitor performance at sources that exhibit variable
pollutant concentrations. Although the RA test is useful in
evaluating monitor performance at any one concentration level, the
calibration error test is the only means of evaluating the calibration
response of monitors over the full measurement range. Therefore, the
calibration error specification should be included in the reproposed
specifications.
Response: The primary criterion of the PS is the RA test
using RM's. When purchasing a CEMS, the user may choose to include a
calibration error specification.
Commenter VI-0-9
9.1 Comment: Section 60.13(b). I concur with the Agency that it
is not necessary to'write a regulation that would require a pretest to
a test.
Response: None required.
9.2 Comment: Section 60.13(c)(2)(ii). I feel that the "+20
percent with a confidence level of 95 percent" should be referenced
to an applicable standard, since it is unlikely that any monitoring
equipment would be able to maintain this error at the minimum
detectablility of the monitoring equipment.
Response: The PS includes a +10 percent of emission standard
criterion to account for errors at the minimum detectability of the CEMS,
55
-------
9.3 Comment: Section 60.13(d). The requirement for daily zero
and span drift is unnecessary since the regulation already requires the
operator to follow the manufacturer's recommendation and also minimum
standards are specified in Appendix B.
Response: The requirement for daily zero and span drift is
for long-term basis. However, Appendix B does not address the long-term
issue. Although the regulation requires the operator to follow the
manufacturer's recommendation, this paragraph sets the minimum
requirement.
9.4 Comment: Section 60.13(d). I do not believe the EPA has any
basis for determining the optical surface of opacity monitors must be
cleaned once the zero compensation exceeds 4 percent opacity. This
requirement does not allow the manufacturer any flexibility. The optical
surface should be cleaned as per the manufacturer's recommendation.
Response: The 4 percent is thought to be the minimum percent
opacity needed to assure proper quality of data. This specification guides
the manufacturer's recommendation as to how often it should be cleaned.
Commenter VI-D-10
10.1 Comment: The proposed revisions to Section 60.13 of the
General Provisions make this section consistent with the proposed revisions
to Appendix B. We have no comment with respect to these proposed revisions.
Response: None required.
56
-------
10.2 Comment: We support the proposed revision which would
allow for the complete evaluation of S0? and NO GEMS's after they have
^ /\
been installed at operating facilities. The proposed PS tests should,
therefore, be less costly to perform relative to current PS requirements.
Response: None required.
Commenter VI-0-12
12.1 Comment: We are a member company in the UARG and wholeheartedly
support the comments on this subject submitted by them. Therefore, our
comments will be in addition to those filed by UARG.
Response: None required.
12.2 Comment: Although this latest proposal is drastically
different from previous EPA proposals and has the potential to shift
the bulk of the burden to prove the accuracy and reliability of
CEMS from the vendor to the user, we are generally in favor of this
approach. But we do have reservations.
Response: None required.
12.3 Comment: Prudent specifications of conditions and requirements
in user purchase agreements may effectively swing this burden of proof
back to the vendor. However, it remains to be seen if the vendors will
attempt to meet instrument specifications included as part of a purchase
agreement if those specifications are not required by regulations. In
fact, our experience to date with conditional purchase agreements has
been less than promising. The vendors were reluctant to agree to this
procedure even when we had the previous proposal (October 10, 1979) and
an EPA sponsored research project to back us up. This reinforces the
57
-------
position that the EPA should issue the guidelines covering monitor
specifications and test procedures that were discussed in the preamble
in a timely manner. At least then, the vendors and users will have an
official definitive document on which to base these conditional agreements.
Response: See Response 8.1.
12.4 Comment: Even though we are in general agreement with the.
overall content of the proposal, we would like to point out a few
specific points that need further clarification before the proposal is
promulgated. These comments follow.
a. The proposal is apparently not intended for sources subject
to Subpart Da requirements or those sources required to have S0« scrubbers,
since it precludes locating an S02 scrubber inlet CEMS and determining its
RA. If the EPA intends for this proposal to be applicable to sources
subject to scrubber installations, then more effort should be extended
in meeting that objective. Unless another proposal is forthcoming that
will cover these sources, we recommend that the EPA rewrite this proposal
to incorporate inlet CEMS testing.
b. The determination of the CEMS RA contains a number of possible
errors that must be combined into one final accuracy number. The sources
of these possible errors are the inaccuracies associated with the CEMS
location, the inaccuracies associated with the pollutant monitor, the
inaccuracies associated with the diluent monitor, and the inaccuracies
associated with the RM.
58
-------
By allowing more leeway in choosing a CEMS location, the EPA has
increased the possibility of it being more accessible than previous
proposals would dictate. No effort is to be made to determine if the
flow is stratified or not, and no provision is provided for this
purpose. Therefore, the +10 percent location error allowed by the
last proposal (October 10, 1979) must be incorporated in the CEMS RA
determination. If stratification does exist, this error could be
significant.
Also, the pollutant monitor and the diluent combined monitor are
now considered the CEMS and are subjected to the RA determination as
such. The previous proposal (October 10, 1979) allowed 20 percent
for the pollutant monitor alone; therefore, the EPA must have already
recognized some limits in pollutant monitor accuracies.
Finally, the RM, itself, may cause other errors of differing
values depending on technique, familiarity, and other nondefinitive
problems. Though considered accurate in itself and since it is a
RM, there can still be a number of possible errors.
We doubt that the CEMS can be depended upon to meet the accuracy
requirement of +20 percent when it must incorporate them all into one
accuracy number. We also question whether the CEMS should be subjected
to this sum total of possible errors in determining their accuracy. We
recommend, at the least, requiring a separate accuracy determination
for the pollutant monitor and one for the diluent monitor.
59
-------
Response: See Responses 0.1 and 8.2. In addition, the
PS does not prohibit stratification checks. This option is left to
the user. The October 1979 proposal included the diluent monitor
as part of the CEMS (see Section 5.1).
Commenter VI-D-13
13.1 Comment: Section 5.2. We strongly oppose the "more than
50 percent capacity" provision because it would make it impossible
for certain units to comply. Such units are those that were overdesigned
by more than 100 percent for future considerations. We suggest that
calibration tests be conducted while the affected facility is operating
at 50 percent of its normal load.
Response: See Response 5.11.
13.2 Comment: Section 4.1. We anticipate difficulties in
complying with this provision because span gas suppliers often provide
span gas which is only within +10 percent of what we have requested.
For this reason, we suggest a span level of 80 to 100 percent.
Response: See Response 1.1.
13.3 Comment: Proposed regulations delete the current requirements
to offset the zero setting by at least 10 percent of the span during '
calibration tests (Sections 6.2.1 and 6.2.2 of the existing PS 2). We
support the deletion for the following reasons. First, some monitors
can be offset by only up to 5 percent. Second, offset of the zero setting
invalidates span settings and causes a nonlinear relationship between
the two settings.
60
-------
Response: None required.
13.4 Comment: Section 60.13(d). The proposed quantification
and recording requirements for zero and span drift create unnecessary
paperwork. Such an investigation can be conducted as the need arises
on an informal basis.
Response: The recording is not a proposal but is presently
required under Section 60.7(d). The section has been revised to
quantify the amount of excess zero and span drift, when specified.
13.5 Comment: In general, we are in agreement with the proposed
concept to simplify and improve procedures for determining pollutant
concentrations and for evaluating continuous monitoring systems.
Response: None required.
61
-------
COMMENTS ON OTHER VIEWS
The preamble of the January 26, 1981, proposed rules, under
"Request for Comments on Other Views," stated: "A number of
suggestions were received which were not incorporated in these
revisions. Because they represent differing views, EPA requests
comments on them to determine what course of action should be taken
in the final rulemaking." The following are the comments that were
received and EPA's responses.
1. Mandatory Conditioning Periods
a. VI-D-2. Although we believe that a conditioning period is
beneficial for identification of operational problems prior to
committing funds and personnel to a test program, we see no need to
include a conditioning period as a mandatory requirement.
b. VI-D-5. While we agree with the suggestion of a conditioning
period with the length determined by the anticipated use of the CEMS,
we do not feel it should be mandatory. Conditioning periods would
tend to reduce the need for variances if the CEMS goes "unreliable"
immediately after an RA test is conducted 1 day after startup, but
are not necessary for those who have a great deal of experience in
a particular application.
c. VI-D-6. Eliminating the 7-day conditioning period leaves
the burn-in requirements to the operator and equipment manufacturer
to define. This allows more flexibility, a possible reduction in cost,
and should not have a significant effect on the integrity of the tests.
62
-------
d. VI-D-10. Provisions of Section 60.8 require that NSPS
testing be conducted within 60 days after the date when full
production is achieved. Under these circumstances, a requirement
for CEMS's to be installed and operational 60-90 days prior to NSPS
compliance testing could actually conflict with activities
associated with bringing the unit up to full capability. Therefore,
we urge that CEMS installation and operation 60-90 days prior to
NSPS compliance testing not be required.
e. VI-D-11. Due to the high cost of certifying a CEMS, a
source will want to insure that the system is functioning properly
prior to arranging or initiating a compliance certification test.
We recommend that EPA suggest that source adopt a 30-day conditioning
period before initiating a performance test. By incorporating a
longer preliminary period, there should be a greater chance of a
CEMS passing the performance test. The longer conditioning period
would also allow plant maintenance personnel an opportunity to
become familiar with the system under actual field conditions. We
believe that these benefits clearly outweigh having a fixed
conditioning period established by regulation and recommend the
mandatory 7-day conditioning period be excluded.
f. VI-D-13. While we do not see a necessity for a separate
conditioning requirement for operation and compliance purposes, we
strongly recommend that the mandatory conditioning period requirement
be retained and perhaps even made longer. Retention and expansion
63
-------
of the mandatory conditioning period is important and needed because
of persistent plugging problems. A conditioning period requirement
will provide additional impetus for vendors to supply reliable
systems and will help insure that approved systems will continue to
operate satisfactorily.
Response: Five of the six commenters felt that the conditioning
periods were beneficial but should not be made mandatory. Therefore,
the requirements and verification of operational status have been
left to be determined between the manufacturer and the source owner
or operator.
2. Mandatory 3-Day RA
a. VI-D-11. We disagree with the view that RA tests should be
spread over 3 days instead of 1. Considering that most stack sampling
consultants base their fees on time spent in the field plus expenses,
we believe that the requirement to extend the sampling time is
unreasonable. We assume that by extending the sampling over 3 days,
the objective would be to obtain data that would be more representative
of actual source emissions. The objective of the RA tests is to insure
that the pollutant concentration, measured by an appropriate RM, is
statistically similar to pollutant measurements obtained from the CEMS.
We see no benefit in extending the sampling period. However, if the
objective is to vary the pollutant emission concentration, other means
should be employed.
64
-------
Response: The objective of the RA tests as mentioned by the
commenter is correct. Therefore, the specification was not changed
to mandate a 3-day RA. The present specifications allow the RA
tests to be conducted over a minimum of 4.5 hours. However, no
maximum is set. Therefore, the 3-day option is open to the source
owner or operator, should he desire to extend the test period.
3. Concurrent NSPS and PS Tests
a. VI-D-2. We agree that NSPS compliance tests and CEMS
performance tests should be conducted simultaneously in a manner
so as to be acceptable as NSPS compliance tests.
b. VI-D-5. Source operators should be given the option of
conducting CEMS PS tests and NSPS tests concurrently. However, we
disagree with the suggestion to mandate concurrent testing.
c. VI-D-6. The option to combine tests should be left open
to the operator, but the tests are sometimes performed under different
test conditions.
d. VI-D-13. Use of results from a concurrent test would save
resources. For this reason, we support integration of the concurrent
test provision in the final regulation.
Response: Two commenters felt that simultaneous NSPS
and PS tests should be mandated; two commenters felt that they should
not be. The proposed specifications allow simultaneous testing. If
simultaneous testing is economically advantageous, source owners or
operators would naturally do it without being regulated. Mandating
65
-------
simultaneous testing could increase costs, e.g., if the CEMS breaks
down during the PST, both tests would have to be halted. However,
leaving the option as is, the performance testing could proceed
and the PST conducted at a later date.
4. CEMS Calibration Checks
a. VI-0-2. Failure to periodically introduce calibration gases
as close to the point of sample acquisition as possible could result
in inaccurate data due to deterioration of the sample transport
system. If the calibration checks are not conducted in this manner
during performance testing, the CEMS may still meet the required
RA but be subject to an unidentified leakage in the sample transport
system which could vary with time and ambient conditions. Our
experience with sources that have installed extractive CEMSrs indicates
that the requirement to introduce the calibration gases at the probe
is not unduly burdensome and does not result in a significantly higher
cost than for introduction at the analyzer.
b. VI-D-5. We agree with the concept of checking the entire
CEMS starting at the probe for extractive systems. However, this type
of check is difficult to implement and can yield misleading results.
If the system runs at negative pressure, the calibration cylinder
pressure must be adjusted to properly balance the system pressure to
normal operating conditions. If the cylinder regulator pressure is
too high, leaks will not show up because the system will be pressurized.
For systems that have a sample pump near the probe, the need for a
66
-------
daily check is minimized because a small leak on a positive pressure
system will not affect analyzer response. If diluent measurement is
done simultaneously, leaks may be corrected for twice.
Sample losses due to adsorption or absorption will be minimal
if proper materials are used, again minimizing the need for a daily
system check.
Most calibration gases purchased in aluminum cylinders will not
change concentration significantly over periods much longer than
6 months. Only an initial calibration gas calibration check should
be required provided the source operator uses certified gases in
followup QA/RA tests.
c. VI-D-6. We concur with the opinion that zero and span
gases should be introduced at the stack in extractive systems to
verify the entire measurement system integrity.
d. VI-D-13. We agree with the EPA analysis that the RA test
would validate the CEMS regardless of the location for introducing
calibration gases into the system. We are opposed, however, to a
requirement which specifies the location at which calibration gases
are introduced. A requirement to introduce calibration gases at a
probe located at the stack wall would require either long calibration
gas supply lines or would require another person since monitors are
usually located at some distances away from the probe. This is
unnecessary.
67
-------
Response: The EPA reconsidered requiring the introduction
of calibration gas mixtures at the probe at the stack wall, but
determined that such a requirement would be inconsistent with the
basis for the reproposal, i.e., comparisons against the RMrs determines
the GEMS's acceptability. However, EPA agrees that checking the whole
system including the interface is desirable and, with proper demonstration,
could replace the comparisons against the RM's. Mention of its
desirability will be made in the guideline document.
5. Performance Specifications
a. VI-D-2. The comments on audit and data recovery are very apropos
and should serve to illustrate to EPA that QA procedures are an integral
part of these specifications. It is naive not to recognize that QA
specifications do govern CEMS design and evaluation and more naive not
to account for all operational requirements before selection and
installation of CEMS.
b. VI-D-5. For very low emission rates, the RA requirement
of 10 percent of the applicable standard is within our CEMS sensitivity.
However, an alternative to consider is to allow RA within 5 percent CEMS
full-scale for low emission rates.
c. VI-D-6. We concur with decreasing value of the accuracy of
emissions as they approach zero. We believe the external check is
extremely desirable if one is to ensure valid measurement data. The
minimum data recovery specification is certainly desirable from a
user's standpoint, however, vendors will raise the associated costs
63
-------
to cover their risks. It would help differentiate reliable from
unreliable equipment and vendors.
d. VI-D-13. We oppose a minimum data recovery specification
of at least 18 hours in at least 22 out of 30 days for two reasons.
First, it would mean that a minimum of 30 days is needed to complete
a source test. This is a long test by itself. Second, there would be
less time to comply with the "within the 60 days after achieving the
maximum production" provision.
The fixed percentage specification for determining RA, +10 percent
of the applicable standard, is inherently biased against the sources
combusting relatively clean fuels. Hence, we recommend a use of a
sliding scale which relaxes the subject specification for sources having
to meet low emission standards.
Response: These comments have to do with evaluating the
reliability of the CEMS over a long-term basis. See Responses 2.1,
5.3, and 5.14.
Concerning the sliding scale which relaxes the required accuracy
when emission standards are low, EPA agrees that the RM variability needs
to be considered. At present, the only low emission standard is for
S02 (Subpart Da). Therefore, Section 4.3 has been revised to allow
greater relative errors at these levels.
6. Separate vs. Combined Units of Measurement
a. VI-D-5. We agree that all units should be recorded, but only
the regulated quantity needs to be reported.
69
-------
b. VI-D-6. We recommend this from the standpoint of backup. If
the ppm to Ib/MBtu conversion fails, one would have a backup position with
a recording of raw data.
Response: Since the recording of separate units is viewed as
backup in nature, this option should be left to the owner or operator.
If it is advantageous, then this would become normal procedure without
being regulated.
7. The RA Terminology
a. VI-D-5. The term "relative accuracy" is defined accurately
in the definitions, so we would leave the term the same even though
the commenter is accurate in stating that an RA test determines the
relative error between the reference test and the CEMS.
b. VI-D-6. This is somewhat academic in that error and accuracy
are often used interchangeably. The +20 percent requirement is
obviously an error allowance.
Response: Both comments are correct. Since a change could
cause confusion, the use of the terminology will not be changed.
8. Outlier Rejection Technique
a. VI-0-2. In order to allow an agency to assess the true results
of RA testing, results for all tests conducted should be required to
be submitted, along with reasons for requesting deletion of any data.
The Agency, should make the final decision as to deletion of data and
should use the results for all acceptable tests in calculating the
RA of the CEMS.
70
-------
b. VI-D-5. We agree that the tester should be allowed to
reject up to three samples in determining RA tests results. While
it is probably true that the tester will throw out the three tests
that increases his RA, he will also have the discretion to throw
out results that do not appear reasonable.
c. VI-D-6. We highly recommend this. Specific stack samples
are often obviously faulty or erroneous. A stack sampler should be
able to take extra samples to allow for an occasional mistake. This
will, in effect, reduce the risks and costs of performing Method 6
or 7. Any one sample should be able to be eliminated without specific
justification. Elimination of more than one sample value should be
left to the subjective analysis of the data by the user.
d. VI-D-11. We recommend that statistical techniques be used
to reject outliers, rather than an arbitrary decision on the part
of the tester. However, test results that are outliers because of
testing difficulties or source operating conditions should be rejected
without being subjected to statistical techniques.
Response: The Agency originally decided that a data set
of a minimum of nine samples was sufficient to evaluate the CEMS.
Statistical techniques for outliers within the set of nine were
investigated, but were found to be too restrictive because of the
small sample size. The proposed technique does not allow any
rejection of data from the set of nine, but allows the replacement
of up to three samples by other samples. Since the evaluation involves
71
-------
a comparison of two methods, EPA feels that the proposed procedure is
satisfactory for rejecting outliers.
72
-------
TABLE 2. LIST OF COMMENTERS
Docket No. OAQPS-79-4
Docket item no. Commenter/Aff111atlon
VI-D-1 Joe Francis, Environmental Specialist
State of Nebraska Department of
Environmental Control
Mail Box 94877 Statehouse Station
Lincoln, NE 68509
VI-D-2 James K. Hambright, Director
Bureau of Air Quality Control
State of Pennsylvania
P.O. Box 2063
Harrisburg, PA 17120
VI-D-3 Richard H. Russell, Environmental Sampling
Supervisor
Owens-Illinois, Inc.
P.O. Box 1035
Toledo, OH 43666
VI-D-4 Russell 0. Blosser, Technical Director
NCASI
260 Madison Avenue
New York, NY 10016
VI-D-5 David J. Williamson
Pacific Gas and Electric Company
P.O. Box 7442
San Francisco, CA 94106
VI-D-6 Gerald F. McGowan, Vice President,
Engineering
Lear Siegler, Inc.
74 Iverness Drive East
Englewood, CO 80112
VI-D-8 F. William Browne!! for Utility Air
Regulatory Group
Hunton and Williams
1919 Pennsylvania Avenue, N.W.
Washington, D.C. 20036
VI-D-9 J.L. Bechthold, Manager
Regulatory Compliance and Services
Northern States Power Company
414 Nicollet Mall
Minneapolis, MN 55401
73
-------
TABLE 2. LIST OF COMMENTERS
(Continued)
Docket No. OAQPS-79-4
Docket item no. Commenter/Affiliation
VI-D-10 Mark P. Steinberg, Superintendent
Air Quality Division Environmental
Department
Wisconsin Electric Power Company
231 W. Michigan
Milwaukee, WI 53201
VI-D-11 F.W. Chapman, Jr., Manager
Environmental and Energy Conservation
515 South Flower Street
Los Angeles, CA 90051
VI-D-12 Morris L. Brehmer, Executive Manager
Environmental Services
Virginia Electric and Power Company
Richmond, VA 23261
VI-D-13 J.G. Holmes, Director
Environmental Planning
TOSCO Corporation
10100 Santa Monica Blvd.
Los Angeles, CA 90067
74
-------
TECHNICAL REPORT DATA
(flcase read Instructions on the reverse before completing)
EPA-450/3-82-022
3. RECIPIENT'S ACCESSION NO.
Revisions to Continuous Emission Monitoring Systems
Performance Specifications 2 and 3 - Summary of Comments
and Responses
5. REPORT DATE
September 1982
6. PERFORMING ORGANIZATION CODE
Emission Measurement Branch
8. PERFORMING ORGANIZATION REPORT NO.
VIE AND ADDRESS
Emission Standards and Engineering Division
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research TRiangle Park, NC 27711
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
Director, Office of Air Quality Planning and Standards
Office of Air, Noise, and Radiation
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
t4. SPONSORING AGENCY CODE
EPA/200/04
This document contains introductory material regarding the proposed revisions
(January 26, 1981, 46 FR 8352) to Performance Specifications 2Pand 3 The changes
edre "1th the "' Deceived and 9
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS fc! COSATI Field/G
13 B
Unlimited
19. SECURITY CLASS (ThisReport)
Unclassified
20. SECURITY CLASS (This page)
Unclassified
21. NO. OF PAGES
80
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE 75
------- |