[Federal Register Volume 62, Number 55 (Friday, March 21, 1997)]
[Proposed Rules]
[Pages 13776-13785]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 97-7215]



[[Page 13775]]

_______________________________________________________________________

Part V





Environmental Protection Agency





_______________________________________________________________________



40 CFR Part 60, et al.



Hazardous Waste Combustors; Continuous Emissions Monitoring Systems; 
Notice of Data Available and Request for Comments; Proposed Rule

  Federal Register / Vol. 62, No. 55 / Friday, March 21, 1997 / 
Proposed Rules  

[[Page 13776]]



ENVIRONMENTAL PROTECTION AGENCY

40 CFR Parts 60, 63, 260, 261, 264, 265, 266, 270 and 271

[FRL-5711-5]


Hazardous Waste Combustors; Continuous Emissions Monitoring 
Systems; Proposed Rule--Notice of Data Availability and Request for 
Comments

AGENCY: Environmental Protection Agency (EPA).

ACTION: Notice of data availability and request for comments.

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

SUMMARY: This announcement is a notice of availability and invitation 
for comment on the following reports pertaining to the proposed 
requirement for continuous emissions monitoring systems for hazardous 
waste combustors (61 FR 17358 (April 19, 1996)): Site-specific Quality 
Assurance Test Plan: Method 301 Validation of a Proposed Method 101B 
for Mercury Speciation, with Appendices, dated September 27, 1996; 
Site-specific Quality Assurance Test Plan: Total Mercury CEMS 
Demonstration, Volumes 1 and 2, dated October 11, 1996; Site-specific 
Quality Assurance Test Plan: Particulate Matter CEMS Demonstration, 
Volume 1, dated August 7, 1996; and Status Report IV: Particulate 
Matter CEMS Demonstration, Volumes 1, 2, and 3, dated February 12, 
1997.
    Readers should note that only comments about new information 
discussed in this notice will be considered. Issues related to the 
April 19, 1996, proposed rule and subsequent notices that are not 
directly affected by the documents or data referenced in this Notice of 
Data Availability are not open for further comment.

DATES: Written comments on these documents and this Notice must be 
submitted by April 21, 1997.

ADDRESSES: Commenters must send an original and two copies of their 
comments referencing Docket Number F-97-CS3A-FFFFF to: RCRA Docket 
Information Center, Office of Solid Waste (5305G), U.S. Environmental 
Protection Agency Headquarters (EPA, HQ), 401 M Street, SW., 
Washington, DC 20460. Comments may also be submitted electronically 
through the Internet to: [email protected]. Comments in 
electronic format should also be identified by the docket number F-97-
CS3A-FFFFF. All electronic comments must be submitted as an ASCII file 
avoiding the use of special characters and any form of encryption. 
Commenters should not submit electronically any confidential business 
information (CBI). An original and two copies of the CBI must be 
submitted under separate cover to: RCRA CBI Document Control Officer, 
OSW (5305W), 401 M Street, SW., Washington, DC 20460. For other 
information regarding submitting comments electronically, viewing the 
comments received, and supporting information, please refer to the 
proposed rule (61 FR 17358 (April 19, 1996)). The RCRA Information 
Center is located at Crystal Gateway One, 1235 Jefferson Davis Highway, 
First Floor, Arlington, Virginia and is open for public inspection and 
copying of supporting information for RCRA rules from 9:00 a.m. to 4:00 
p.m. Monday through Friday, except for Federal holidays. The public 
must make an appointment to view docket materials by calling (703) 603-
9230. The public may copy a maximum of 100 pages from any regulatory 
document at no cost. Additional copies cost $0.15 per page.

FOR FURTHER INFORMATION CONTACT: For general information, call the RCRA 
Hotline at 1-800-424-9346 or TDD 1-800-553-7672 (hearing impaired) 
including directions on how to access electronically some of the 
documents and data referred to in this notice electronically. Callers 
within the Washington Metropolitan Area must dial 703-412-9810 or TDD 
703-412-3323 (hearing impaired). The RCRA Hotline is open Monday-
Friday, 9:00 a.m. to 6:00 p.m., Eastern Time.
    Documents referred to in this notice are available from two 
electronic sources: the CLU-IN and EMTIC bulletin boards. The CLU-IN 
bulletin board is accessible by modem at phone number 301-589-8366 or 
by Telnet at clu-in.epa.gov. The EMTIC bulletin board is accessible by 
modem at phone number 919-541-5742 or over the Internet at http://
ttnwww.rtpnc.epa.gov/. The reader should note that figures, diagrams, 
and appendices may not be available in these electronic documents.
    For other information on this notice, contact H. Scott Rauenzahn 
(5302W), Office of Solid Waste, 401 M Street, SW., Washington, DC 
20460, phone (703) 308-8477, e-mail: [email protected].

SUPPLEMENTARY INFORMATION: On April 19, 1996, EPA proposed revised 
standards (herein referred to as ``the proposed rule'') for hazardous 
waste combustors (HWCs, i.e., incinerators and cement and lightweight 
aggregate kilns that burn hazardous waste). See 61 FR 17358.

I. Introduction and Background

    In the proposed rule, EPA proposed requiring that continuous 
emissions monitoring systems (CEMS) for particulate matter (PM) and 
total mercury (Hg) be used for compliance with the proposed PM and 
mercury emission standards. To require CEMS for compliance the Agency, 
among other things, determines that the CEMS are commercially available 
and meet certain performance specifications. To make these 
determinations, the Agency routinely tests CEMS available in the 
marketplace. EPA published a notice inviting vendors of PM and Hg CEMS 
to participate in a CEMS demonstration test program. (See 61 FR 7232, 
February 27, 1996). Ten vendors responded to the Agency's invitation. 
They donated nine devices for the test program: six PM CEMS and three 
Hg CEMS.1
---------------------------------------------------------------------------

    \1\ One Hg CEMS vendor was unable to participate.
---------------------------------------------------------------------------

    Today the Agency is providing notice and opportunity to comment on 
the following documents resulting from its CEMS demonstration test 
program: (1) Site-specific Quality Assurance Test Plan: Method 301 
Validation of a Proposed Method 101B for Mercury Speciation, with 
Appendices, dated September 27, 1996; (2) Site-specific Quality 
Assurance Test Plan: Total Mercury CEMS Demonstration, Volumes 1 and 2, 
dated October 11, 1996; (3) Site-specific Quality Assurance Test Plan: 
Particulate Matter CEMS Demonstration, Volume 1, dated August 7, 1996; 
and (4) Status Report IV: Particulate Matter CEMS Demonstration, 
Volumes 1 and 2, dated February 12, 1997. The purpose of this notice of 
data availability (NODA) is to obtain comment on the Agency's approach, 
as described in these documents, prior to the end of the demonstration 
tests. Comments received will, to the extent possible, be incorporated 
into the demonstration test programs. EPA plans to follow this NODA 
with a second notice after the testing program has completed or is near 
completion. That final notice will contain what EPA believes will be 
final draft performance specifications for these CEMS.
    The reader should note that one of these documents, the PM CEMS 
demonstration test status report, is a draft report which is evolving 
over time. The report will be added to and modified substantially as 
the program progresses. Therefore, conclusions and discussions in this 
report do not necessarily represent EPA's final views. They are 
included in this NODA so the reader can fully evaluate the Agency's

[[Page 13777]]

approach and comment on this approach prior to the end of the testing 
program.
    The remainder of this notice describes the demonstration test 
programs for PM and Hg CEMS. It serves as an overview for the reader 
and brings to the reader's attention certain areas where EPA requires 
input.

II. The PM CEMS Demonstration Tests

A. Background

    EPA previously tested PM CEMS at two other sites, the Rollins 
incinerator in Bridgeport, NJ, and the LaFarge cement kiln in Fredonia, 
KS. Both were short-term tests to determine whether further testing is 
warranted.
    The purpose of the Rollins Bridgeport tests was to qualitatively 
determine whether vendor claims that PM CEMS can be used for compliance 
with a PM standard was feasible and to gain insights on the scope and 
nature of future tests. Three devices were tested at Rollins: a Sick 
RM200 light-scattering CEMS; a BHA CPM1000 time dependant optical 
transmission CEMS; and an Emissions SA Beta 5M -gage 
CEMS.2 Due to the limited nature of this investigation, though, 
there were certain deficiencies in these tests which make quantitative 
comparisons of this data to other data difficult. For instance a 
calibration of the instruments cannot be performed because manual 
method data was obtained over only two particulate emission loadings, 
the measured range of emissions was less than one-third of the proposed 
HWC PM standard, and only eight valid manual method measurements were 
made. However, we did determine that optically based PM CEMS, such as 
the light-scattering and time dependant optical transmission 
instruments, had a step function increase in their output when 
entrained water droplets were encountered in the gas stream.
---------------------------------------------------------------------------

    \2\ The Beta 5M CEMS is participating in this demonstration as 
well.
---------------------------------------------------------------------------

    Additional tests were conducted in May 1995 at the LaFarge cement 
kiln in Fredonia, KS. The purpose of these tests was to conduct a full 
calibration of the instruments in accordance with the ISO 
(International Standards Organization) specification, to better 
determine whether these CEMS could be used for compliance with a PM 
standard and whether the ISO performance specification could be used as 
a basis of a proposed PM CEMS performance specification, and to gain 
insights on future testing. Two devices were tested at LaFarge: the 
Sick monitor used in the Rollins tests and an ESC P5A.3 Both are 
light-scattering devices. Both devices were installed in April 1995 and 
were operated continuously on the cement kiln through July 1995. At 
these tests, EPA successfully calibrated these devices in May 1995 
using nine valid pairs of M5 runs at three PM loadings. Additional PM 
measurements were made approximately one and two months after the 
initial calibration. EPA gained the following insights during this test 
program:

    \3\ The ESC P5A CEMS is participating in this demonstration as 
well.
---------------------------------------------------------------------------

--The ISO specification can be used as a basis for any performance 
specification EPA develops and that the instruments could be calibrated 
to particulate emissions obtained from manual method data.
--Response of the instruments to changing PM concentrations was 
generally better at this cement kiln than at the previous Rollins test.
--Statistics resulting from these calibrations barely passed the ISO 
specification. Other countries (such as Germany) suggest that 15 
measurements be made instead of 9 to improve calibration statistics. 
Therefore, more than 9 measurements may be necessary.
--The current Method 5(M5) had limitations in measuring low-level 
particulate emissions due, in large part, to the difficulty of the 
extraction, filter recovery, and weighing steps. This limitation likely 
lowered the calibration statistics determined from the data obtained 
during these tests.
--PM CEMS could be used for compliance with a PM standard, but longer 
term demonstration testing is necessary to ascertain the device's long-
term durability.

    The calibration results are summarized in Table 1, below.

                      Table 1.--May 1995 Calibration Results From the LaFarge PM CEMS Tests                     
----------------------------------------------------------------------------------------------------------------
                                                                                    Confidence       Tolerance  
                                                                    Correlation      interval        interval   
                              CEMS                                  coefficient      (CI0.95)        (TI0.95)   
                                                                        (r)          (percent)       (percent)  
----------------------------------------------------------------------------------------------------------------
ISO Performance Specification...................................            0.90              25              35
Sick RM200......................................................            0.92              17              29
ESC P5A.........................................................            0.90              20              32
----------------------------------------------------------------------------------------------------------------

B. Site Selection

    For the PM CEMS demonstration tests, EPA selected the DuPont 
Experimental Station hazardous waste incinerator in Wilmington, DE. The 
DuPont incinerator receives a variety of wastes from many DuPont 
facilities in northern Delaware. As such the waste input to the 
incinerator is like that of many commercial incinerators.
    The DuPont incinerator has a Nichols Monohearth as its primary 
combustion chamber. Waste is fed to this combustion chamber using a ram 
feeder for solid waste, a cylindrical chute for batched waste material, 
and a Trane Thermal liquid/gas waste burner. The primary chamber 
exhausts to a secondary chamber (afterburner) where waste is fed using 
a Trane Thermal burner. The flue gas then travels through a spray 
dryer, then through a cyclone separator, where dissolved and suspended 
solids are removed. The cyclone system discharges to a reverse jet gas 
cooler/condenser which reduces the gas temperature to the dew point. 
The flue gas then travels through a variable throat venturi scrubber 
which removes additional particulate and some acid gasses. The venturi 
scrubber exhausts into an absorber neutralized with soda ash scrubbing 
solution to absorb acid gasses. The absorber also subcools the flue gas 
before traveling through a chevron-type mist eliminator. After passing 
through the mist eliminator, the gas travels through a set of electro-
dynamic venturis (EDVs) which are used to remove fine particulate along 
with metals that condense onto the fine particulate as a result of the 
gas subcooling. The gas then travels through a set of centrifugal 
droplet separators and an induction fan, is reheated to eliminate any 
visible plume, and is finally discharged to the atmosphere through the 
stack. A full description of the incinerator as well as

[[Page 13778]]

a diagram of the system is contained in section 2.2 of the PM Test 
Plan.
    EPA chose to perform the PM CEMS tests at an incinerator because, 
under a normal range of operating conditions, incinerators present a 
worse case exhaust stream to challenge multiple PM CEMS technologies in 
a long-term test program. For the purpose of demonstrating the 
capabilities and limitations of PM CEMS, a worse case exhaust stream 
would consist of high moisture (i.e., greater than 20%), average PM 
levels below the proposed emission limit, and PM with a wide variation 
in physical properties (such as composition, particle size 
distribution, shape, color). Incinerators fulfill this worst-case need 
in three main ways. First, commercial incinerators and some on-site 
incinerators, including the DuPont facility, burn a wide variety of 
waste as their primary feedstream. The wide variety of the primary 
feedstock 4 has a higher potential to produce highly variable 
particulate, which is a worst case test for PM CEMS. This is not the 
case for cement or light-weight aggregate kilns (CKs or LWAKs, 
respectively.) These sources primarily feed particulate rich process 
ingredients (limestone and fly ash for CKs and slate, shale, and clay 
for LWAKs). As a result, PM in the flue gases from both CKs and LWAKs 
are likely to be overwhelmed by the process dust and be more uniform 
than those from an incinerator. Second, many incinerators are equipped 
with wet air pollution control system (APCS) technologies which are 
able to meet the proposed PM emission limit and produce high moisture. 
Finally, these APCS technologies produce a narrow PM size distribution 
(i.e., primarily less than 1 micron). This narrow size distribution is 
typical of emission levels from wet APCS technologies that are expected 
to be installed on HWCs to meet the upcoming MACT standards.
---------------------------------------------------------------------------

    \4\ Feeds which affect PM emissions include metals, other 
solids, and chlorinated solvents.
---------------------------------------------------------------------------

    The DuPont incinerator was chosen because:

--PM emissions were expected to range from 0.005 to 0.075 gr/dscf (that 
is, 17 to 250% of the proposed HWC PM standard), depending on how the 
facility operated;
--The facility accepts ``small'' batches of many waste streams and has 
limited capacity to burn many waste streams simultaneously, thereby 
assuring more dramatic changes in particulate concentrations and 
physical characteristics in shorter time intervals, relative to a 
larger commercial facility;
--The facility has no ESP or fabric filter for PM control;
--The facility was willing to participate in the test program and allow 
necessary modifications to be made;
--The facility was willing and able to vary operating conditions as 
required to perform the PM CEMS calibrations; and
--Physical access, both for sampling in the stack and for equipment and 
personnel on the adjacent platform, was available to locate six PM CEMS 
and a test crew.

    A detailed description of the site selection is located in section 
1.4 of the PM Test Plan.

C. Revised Manual Method for PM

    One issue which PM CEMS vendors raised and which was noted during 
the LaFarge tests was that the current manual method for PM (Method 5, 
herein refered to as M5,) may be inadequate to make the low-level 
measurements required for PM CEMS calibrations. EPA determined that 
much of this error comes from sample recovery and analysis. Stacks with 
high acid gas, water, and/or adhesive concentrations (i.e., cement kiln 
clinker) in the flue gas make the filter stick to the filter housing. 
As a result, filter recovery is difficult. For this reason, EPA chose 
to modify M5 slightly.
    The modification employs the use of a light-weight filter assembly. 
The front-half and filter assembly are first pre-tared. The filter 
assembly then replaces the current M5 filter housing in the heated box. 
After measurement, the entire assembly is desiccated and weighed. This 
way the M5 extraction step is eliminated without making fundamental 
modifications to M5 itself. Given that this change to M5 is minor and 
only affects the extraction and analysis steps, EPA does not believe 
that a full field validation of the modification was necessary. Instead 
the Agency tested those parts of the method which changed to ensure 
that those parts of the process are as good as the current M5. EPA has 
initially determined that this modification is acceptable. Completion 
of this analysis, including a full write-up of the new method, is 
expected soon. A full description of this method will be given in the 
later CEMS NODA.
    EPA expects this modified method for particulate measurements would 
be required for use when calibrating PM CEMS.

D. The PM CEMS Demonstration Test

    The PM CEMS demonstration tests started in September 1996 and are 
expected to continue until May 1997 or later. The test program started 
with an initial calibration of the instruments and followed with 
response calibration audits (RCAs) and absolute calibration audits 
(ACAs) every four weeks. The program also involves continuously 
recording the CEMS data for the duration of the program, documenting 
daily calibration and zero checks, documenting all performed 
maintenance/adjustments, and documenting all periods in which data was 
not available.
    A second important aspect of the demonstration tests is to evaluate 
the proposed performance specification and data quality objectives 
themselves. Proposed performance specification 11 (PS 11) was drafted 
and proposed with the idea that it would be modified based on what 
these tests showed. The final promulgated specifications will be based 
on the data obtained through these tests.

E. PM CEMS Technologies Tested

    The six PM CEMS being tested represent three separate PM CEMS 
technologies: light-scattering, beta gage (-gage), and 
impaction energy devices. Each technology is described below. Full 
descriptions of each PM CEMS are found in section 2.7 of the PM CEMS 
Test Plan and in the proposals submitted by the vendors. Vendor 
proposals are found in docket item S0205. All instruments participating 
in this program are provided to the government at no charge.
    Light scattering devices work by sending a light beam across the 
flue gas and measuring the amount of light reflected back to a detector 
located at some angle (other than straight-path transmissivity) from 
the light source. These devices can be used either in-situ (i.e., in 
the stack) or extractively. These devices are not complex, relative to 
other instruments, and as such are relatively inexpensive to purchase. 
They also have few moving parts and consequently require little 
maintenance. These CEMS are, however, sensitive to PM characteristics, 
including composition, density, size distribution, and index of 
refractory. Three light scattering devices are participating in the 
program: Sigrist Photometer AG model KTNR (supplied by Lisle-Metrix), 
Durag model DR-300, and Environmental Systems Corporation (ESC) model 
P5A. All three CEMS are installed on more than a hundred stacks 
worldwide.
    -gage instruments continuously sample extracted flue gas 
PM on a filter tape. After the PM sample is collected, the tape moves 
so that the collected PM is located between a carbon-14 beta

[[Page 13779]]

radiation source and a detector. This measurement is compared to a 
measurement done on the blank filter to obtain the mass of the 
collected particulate. As such, these CEMS are continuous samplers but 
batch analyzers. These devices are quite complex and as a result cost 
more than light-scattering devices. Their complexity also means they 
require more maintenance and, as a result, experience more down-time 
than light-scattering devices. However, these devices are relatively 
independent of the PM characteristics and vendors claim a site-specific 
PM calibration is generally not required.5 Two -gage 
devices are participating in the program: Verewa model F-904-KD 
(supplied by Monitor Labs), and Emissions SA model Beta 5M (supplied by 
Environnement USA). Both CEMS are installed on more than a hundred 
facilities worldwide.
---------------------------------------------------------------------------

    \5\ All PM CEMS in this testing program will be calibrated 
against the manual method. The claim that -gage PM CEMS do 
not require a calibration will be tested as part of this test 
program.
---------------------------------------------------------------------------

    The third type of PM CEMS technology is an impaction energy device 
supplied by Jonas Consultants, Inc. This monitor operates by detecting 
shock waves caused by particles impacting a probe inserted into the 
flue gas. The device counts the number of impacts and the energy of 
each impact. This information, coupled with the knowledge of flue gas 
velocity, allows the calculation of particulate mass and thus 
concentration. However, the probe does alter the velocity profile of 
the flue gas near the probe which, in principle, affects the 
instrument's response. Thus, EPA believes a site-specific calibration 
is necessary to ensure good instrument response. This device has been 
installed at few locations, mainly for process control use in steam, 
and not for compliance with a flue gas PM standard.

F. Demonstration Test Report

    EPA seeks comment on the document Particulate Matter CEMS 
Demonstration: Status Report IV, provided in the above referenced 
docket. This document describes the interim results from the PM CEMS 
demonstration tests EPA is conducting. It contains an analysis of data 
obtained from the initial calibration through the Relative Calibration 
Audit (RCA) in January 1997. Specific aspects of the report are 
discussed below. Subsection 1 describes the limitations EPA has 
experienced in this test program. Subsection 2 discusses general 
testing issues. Subsection 3 describes the PM CEMS performance 
characteristics observed during the initial (and subsequent) 
calibration and the RCAs. Subsection 4 describes issues associated with 
the proposed performance specifications.
    Note that many of the issues described in this section may also 
apply to other CEMS, such as the Hg CEMS described in the next section. 
Consistency between the two programs will be maintained by handling 
similar issues in a similar manner in both programs.
    Overall, EPA believes the PM CEMS demonstration is making progress. 
EPA was able to calibrate all of the installed devices. The subsequent 
RCAs have proved those calibrations to be reliable over time. EPA also 
believes that the proposed performance specifications will need to be 
modified based on the data and experiences coming out of this program.
1. Limitations of the Test Program
    a. CEMS downtime. One limitation of the program is that, unlike 
facility personnel, EPA is not on-site all the time. As described in 
the test plans, EPA travels to the site every two weeks to, among other 
things, perform any maintenance the instruments might require. This 
causes CEMS downtime occurring during the program to be overstated 
relative to what a real facility would experience if it were using one 
of these instruments for compliance.
    In addition, CEMS purchased by a facility usually come with a 
supply of spare parts so the facility can make minor repairs without 
incurring substantial downtime. In this program however, EPA was not 
supplied with many of the spare parts it would otherwise get if it had 
purchased the instruments. Parts required for routine maintenance must 
be ordered from the supplier as needed rather than drawing them from 
the facility's store of spare parts. It takes more time to order parts 
than to draw from the store of spare parts on-site, so the CEMS are 
down longer than they would be if the CEMS were purchased by a facility 
for compliance.
    Finally, there tends to be no US-based, trained service technicians 
to conduct major repairs on many of these instrument. Technicians from 
the CEMS manufacturer's native country are often flown in to provide 
specialized service. Many of the parts must also be ordered from 
suppliers in other countries. This means that, if a major repair is 
required, service and parts must be obtained from overseas. This takes 
more time than it would if service and parts were available in this 
country, and further overstates CEMS downtime.
    This is important because one thing EPA is trying to gage in this 
program is data availability. Data availability is one minus the CEMS 
downtime, expressed as a percentage. If downtime is overstated, data 
availability will be understated. EPA anticipates remedying this 
situation by subtracting out downtime associated with these 
limitations. For instance, if a CEMS requires a minor repair and goes 
down soon after EPA leaves the facility, the CEMS will be inoperable 
for two weeks, until EPA arrives back at the facility. If the repair 
takes eight hours to perform, then EPA will count the downtime as 8 
hours, not two weeks. The same approach will be used for the Hg CEMS 
program as well.
    b. Absolute Calibration Audits. In the proposed rule, EPA proposed 
requiring facilities to conduct ``Absolute Calibration Audits'' (ACAs) 
every quarter. These tests would be conducted with NIST traceable 
standards to ensure the analytical parts of the instrument were still 
working properly. Unfortunately, only two vendors (Sigrist and Durag) 
have supplied us with these standards. EPA will conduct ACAs on the 
instruments as the standards arrive.
    At this time, NIST does not have traceable standards for these 
instruments. However, German TuV versions of these standards (called 
``linearity test kits'') exist for most of these CEMS. We believe that 
these TV standards are sufficient substitutes for the yet-to-be-
developed NIST standards to conduct the ACAs.
    If these test kits are generally not available to facilities, then 
EPA solicits comment on whether the ACA approach should be modified. 
For instance, it might be adequate to require a device to make daily 
internal zero and span drift measurements and corrections. Most devices 
are already configured to make both zero and span drift measurements 
and corrections.
    c. Inability to repeat tests. It is infeasible to repeat a test 
conducted under a set of conditions at this facility due to the wide 
variety of ``small'' batches of waste the facility processes and the 
hysteresis effect of the APCD. Like a commercial facility, this 
incinerator accepts a wide variety of wastes, both hazardous and 
industrial, from all DuPont facilities in northern Delaware. The 
incinerator often incinerates multiple wastes concurrently. Those 
wastes arrive at the incinerator in a random fashion. Batches are also 
quite small relative to what would be experienced at a commercial 
facility, meaning that transients in PM concentrations and 
characteristics are

[[Page 13780]]

more pronounced and shorter in duration.
    Further complicating this is the fact that this incinerator is a 
zero water discharge facility. This means wastewater from the wet 
scrubber is recycled to the spray dryer (upstream from the scrubber) 
and injected back into the incinerator exhaust gas. This results in a 
hysterisis effect; wastes fed to the incinerator at one time accumulate 
in the pollution control system and affect the emissions later. Both 
situations affect our ability to repeat tests and, consequently, to 
show that CEMS have the same response to the same particulate at a 
later time.
    d. Inability to test with entrained water droplets. One thing that 
attracted EPA to this facility was that it is an incinerator with a wet 
air pollution control system and a reheat system to vaporize water 
droplets that would otherwise be entrained in the stack gas. EPA 
anticipated being able to conduct tests with entrained water droplets 
in the stack by turning off the reheat system. The Rollins tests showed 
that entrained water droplets are mistaken as particulate by light-
scattering PM CEMS. EPA wished to test the light-scattering devices 
with entrained water droplets so it could quantify the effects of 
entrained water droplets on light-scattering PM CEMS.
    Such tests were planned and conducted in November 1996. But no 
entrained water droplets formed despite turning the reheat off. EPA and 
DuPont have since concluded that we are unable to predict when 
entrained water droplets will occur at the incinerator as currently 
configured.\6\ Therefore, it is unlikely that EPA will be able to 
conduct tests with entrained water droplets as part of this program.
---------------------------------------------------------------------------

    \6\ Apparently, there is a 35 deg.F temperature increase across 
the induction fan that can not be overcome.
---------------------------------------------------------------------------

    One approach EPA may take is to use the limited data EPA has from 
its earlier Rollins Bridgeport tests of PM CEMS and draw whatever 
conclusions it can from that data.\7\ However, EPA believes this data 
is insufficient to quantify the effects of entrained water droplets. 
For this reason, EPA requests data which quantifies the effects of 
entrained water droplets on the calibration of light-scattering PM 
CEMS.
---------------------------------------------------------------------------

    \7\ Based on the Rollins data, EPA qualitatively concluded that 
while entrained water droplets did induce a step function increase 
in the output from in-situ light-scattering PM CEMS, it did not 
affect the calibration so much as to cause the calibration to fail 
under this condition. The step function increase was caused by the 
in-situ light-scattering PM CEMS mistaking entrained water droplets 
for particulate.
    This leads EPA to believe that a facility which uses an in-situ 
light-scattering PM CEMS for compliance and has entrained water 
droplets in the stack gas may experience situations where the actual 
PM emissions are lower than those reported by the monitor. The risk 
for a such a facility which is in compliance with the PM standard is 
that it may experience an increased number of false non-compliances 
with the PM standard. EPA hopes to quantify the effects of entrained 
water droplets on light-scattering PM CEMS so it can quantify the 
risks associated with a false non-compliance in this situation.
    This possible false non-compliance possibility can be avoided if 
the light-scattering PM CEMS is configured such that it extracts 
flue gas from the stack, heats it to above the highest possible dew 
point temperature, and measures the heated, extracted gas. One 
light-scattering PM CEMS in the program is so configured. Others can 
be similarly configured to avoid this potential problem. The trade-
off is that extractive light-scattering PM CEMS cost more than in-
situ units.
---------------------------------------------------------------------------

2. General Test Issues
    a. Handling of Outliers. Two types of outliers were experienced so 
far in the program: paired data and statistical outliers. Each is 
discussed below.
    i. Paired Data Outliers. EPA is conducting its PM measurements in 
such a way that a pair of (two) trains concurrently sample the flue gas 
at the same time in the same plane of the stack. The average of these 
two concurrent trains is the PM emissions measured by the manual method 
for a given run. This methodology usually means the results from the 
two trains are similar. This conclusion is substantiated by most of the 
data obtained during the test program.
    However, there were instances when the results from the pair of 
concurrent trains differed substantially. This leads EPA to believe 
that there was a problem with one or both of the trains which comprise 
such a run. As a result, EPA developed a quality criteria requiring 
that the pair of trains which comprise a run not differ substantially. 
EPA quantitatively defined this substantial difference by looking at 
historical M5 data. Data indicate that results from paired trains such 
as these agree with a relative standard deviation (RSD) of 10%. 
Therefore, nearly all data should agree to within three times this RSD, 
or to within 30% of each other. If the results of paired trains 
disagree by more than 30%, the whole run would be thrown out.
    EPA anticipates that other situations will arise in which it will 
need to disregard data which substantially differs from the historical 
data in other ways. The Agency would have serious reservations 
regarding this practice of defining what is or is not acceptable data 
after the fact if this were a compliance determination. However, this 
is not a compliance evaluation, and EPA does not believe the same 
cautions apply. In addition, EPA is unable to develop these quality 
criteria prior to the start of the program because it does not have a 
history of PM data from this facility upon which to base such quality 
criteria. EPA believes this approach of developing quality criteria as 
the program progresses is reasonable given the unique situation here.
    ii. Statistical Outliers. Another type of outlier data experienced 
during the program is referred to as ``statistical outliers.'' 
Statistical outliers are data which are more than three (3) standard 
deviations away from the linear regression line that represents the 
calibration of the instruments. EPA does not have an opinion on how to 
handle statistical outliers and requests comment on how to proceed.
    In implementing manual calibration tests for other CEMS, EPA 
routinely allows outliers of this kind to be disregarded when 
developing the calibration curve. The Agency's logic for disregarding 
this data is that it is known that the manual method sometimes 
dramatically under-reports emissions for unknown reasons.\8\ Using this 
outlier data in the calibration of other CEMS is unwise because: (1) 
The error cannot be accounted for by known science; and (2) eliminating 
the data causes the slope of the calibration curve to be steeper (i.e., 
numerically larger), therefore it is protective of the environment to 
exclude this outlier data. In the case of other CEMS, the statistical 
difference is reasonable justification for proving that a problem 
occurred while obtaining that data point.
---------------------------------------------------------------------------

    \8\ It is believed that the cause is a manual method sample is 
not obtained due to spacial differences between the sampling 
locations of the manual method and CEMS.
---------------------------------------------------------------------------

    Upon closer examination here, however, it is not clear whether the 
fact that the data appears to be statistically different is ample 
justification to disregard the data. Statistical tests usually involve 
the testing of a sample population to see if the sample data is from 
the same population as data you are comparing it to. It involves 
establishing a null hypothesis which states that the sample is part of 
the population (H0: s = p or 
s2 = p2, that the mean or 
variance, respectively, of the sample data is equal to that of the 
population). If the statistical test infers that the null hypothesis is 
not true, then an alternate hypothesis, stating that the data is from a 
different population than the one you are comparing it to, is accepted 
(HA: s  p or 
s2  p2, that the mean or 
variance, respectively, of the sample data is different than that of 
the population). In our case here, the

[[Page 13781]]

fact that a statistical outlier is more than three standard deviations 
away from the linear regression line would likely lead one to reject 
the null hypothesis and say that the outlier is from a different 
population than what is represented by the linear regression line. 
After determining the data are different, one must determine why. This 
analysis is important because it will help determine whether the data 
should be kept in the original sample or disregarded.
    This situation differs from the other CEMS case because PM CEMS 
have known sensitivities to changes in what they are measuring, i.e., 
moisture and the particle's characteristics, such as density, shape, 
size distribution, refractory (color), etc. Unfortunately, we do not 
know the effects of these changes on the outputs of PM CEMS.9 In 
other words, EPA is uncertain whether the statistical outliers were 
caused by an error in the manual method measurement process (in which 
case the data would be thrown out) or if the error was caused by the 
CEMS overstating (or understating) the PM emissions due to changing 
particulate properties (in which case the data would be kept in the 
data set). In addition, most of the statistical outliers experienced in 
this program are ones in which the manual method result is higher than 
what the PM CEMS report. Therefore, it would likely be more appropriate 
to keep the data in the data set in this case.
---------------------------------------------------------------------------

    \9\  EPA's Office of Research and Development recently concluded 
a study of how changes in particulate properties affect the output 
of PM CEMS. The report describing the results of this study is not 
expected to be completed until September 1997.
---------------------------------------------------------------------------

    EPA is currently pursuing statistical ways of dealing with outliers 
and requests comment on how to deal with this situation. One 
alternative is to establish a stringent specification for the 
correlation coefficient (yet within the bounds of the data obtained in 
this program) and allow facilities to throw out, but report, data that 
is farthest away from the linear regression line. ``Farthest away'' 
could be defined on either a relative or absolute standard deviation 
basis. The facility would then substitute in better data if needed to 
meet the minimum number of samples or other performance specification 
requirements.
    b. Extrapolating Data. Another issue is that this facility, while 
having a PM permit limit of 0.08 gr/dscf, cannot emit that much 
particulate. We expect that similar situations exist throughout the 
industry. This is a concern because EPA proposed that facilities 
calibrate their PM CEMS to up to two (2) times the emission limit. If 
it is physically impossible for a facility to emit this much 
particulate, it obviously cannot calibrate the instruments that high.
    Therefore, the Agency seeks comment on whether the following 
approach is acceptable. EPA believes a facility should calibrate the 
CEMS up to the point where, based on historical data the facility has, 
the facility is producing the most particulate. This point will serve 
as the ``high'' calibration range for this facility's PM CEMS. In 
addition, the facility would use the available data and extrapolate the 
linear regression line beyond the high calibration range for instances 
where the emissions are higher than the historical data indicate. As 
the historical data grow for this facility, the facility may notice 
times when the PM emissions are more than what the previous historical 
data indicated. In this event, the facility would re-calibrate the CEMS 
under the previously unknown condition(s) which result in higher 
emissions than the old historical data indicated.
    A unique case exists when the highest possible emission level is 
less than the emission standard. In this case the calibration data 
point resulting in the highest PM CEMS output would be the point where 
the confidence and tolerance interval tests would be conducted.10
---------------------------------------------------------------------------

    \10\ The proposed performance specification states that the 95% 
confidence interval for the calibration curve must be no more than 
 20% of the emission limit at the emission limit. This 
is a single point test. In the case where a facility cannot 
calibrate up to or above the emission limit, the 95% confidence 
interval test for the calibration curve would be  20% of 
the emission limit at the point resulting in the highest PM CEMS 
output. The same approach would be used for the tolerance interval 
test as well.
---------------------------------------------------------------------------

    c. Correcting for temperature and dry basis. Some of the PM CEMS 
need to correct their output for stack temperature and moisture. These 
corrections have not been done for the data in this report. While these 
corrections will have a minor effect, the data will change slightly 
after the corrections are made. EPA is now correcting the data to 
account for changes in temperature and moisture. Future reports will 
report all data properly corrected for temperature and moisture.
3. PM CEMS Performance Characteristics
    One important aspect of the program is to test and verify that the 
performance of these devices meet the characteristics described in the 
proposed performance specification (PS) 11. It also serves as a test of 
the performance specification itself and proposed data quality 
objectives for CEMS described in the proposed Appendix to Subpart EEE. 
Data from this demonstration test program will be used to revise PS 11 
and the data quality objectives as necessary.
    Table 2 lists each of the monitors being tested, the proposed 
performance requirement for the devices, and the actual performance 
observed during the test program.11 The results in Table 2 do not 
include data outliers which have been excluded from the analysis, such 
as ``paired data outliers.'' The reader should note that the 
correlation coefficient, confidence interval, and tolerance interval 
tests apply only to the calibration. These values are reported, 
however, for subsequent RCAs even though they do not apply in this 
situation. Specific discussion on each performance specification is 
discussed, below.
---------------------------------------------------------------------------

    \11\ Results from the Jonas analyzer, however, is not reported. 
The Jonas analyzer reported results in terms of emission rate (g/s) 
rather than emission concentration (mg/dscm). Time is needed to 
analyze and correct how to correct these values to the proper units.
---------------------------------------------------------------------------

    EPA has been able to generate a linear regression line for the 
various PM CEMS. Performance of the devices are nearly identical when 
one compares the performance of an in-situ device to the other in-situ 
device and extractive devices to one another. In-situ units seem to 
show better performance than extractive units regardless of technology 
of the CEMS.

[[Page 13782]]



                                            Table 2.--Performance Characteristics of the PM CEMS Being Tested                                           
--------------------------------------------------------------------------------------------------------------------------------------------------------
         Performance specification             Correlation     Confidence       Tolerance                                                               
--------------------------------------------   coefficient      interval        interval        RCA test       Calibration drift      Zero drift  (ZD)  
         CEMS              Date of test            (r)          (CI0.95)        (TI0.95)        (TI0.95)              (CD)                              
--------------------------------------------------------------------------------------------------------------------------------------------------------
                                                             20  35  75%  2% of the    2% of 
                                                  >0.90       % at emission   % at emission      of data      calibration standard       the emission   
                                                                  limit           limit          within                                     limit       
                                                                                                 TI0.95                                                 
--------------------------------------------------------------------------------------------------------------------------------------------------------
ESA..................  Cal.................            0.55              26              38  ..............  Pass.................  Pass.               
                       01/97...............            0.92              21              25              75  Pass.................  Pass.               
                       All Data............            0.46              35              40  ..............                         ....................
Verewa...............  Cal.................            0.69              27              32  ..............  No data..............  Pass.               
                       12/96...............            0.86              24              20             100  No data..............  Pass.               
                       01/97...............            0.93              18              25             100  Fail.................  Pass.               
                       All data............            0.76              18              23  ..............                         ....................
Durag................  Cal.................            0.72              22              36  ..............  Pass.................  Pass.               
                       11/96...............           -0.38              52              77              75  Pass.................  Pass.               
                       12/96...............            0.91              45              73             100  Pass.................  Pass.               
                       01/97...............            0.93              20              22             100  Pass.................  Pass.               
                       All data............            0.61              20              35  ..............                         ....................
ESC..................  Cal.................            0.71              22              36  ..............  Pass.................  Pass.               
                       11/96...............            0.87              24              31              88  Pass.................  Pass.               
                       12/96...............            0.92              42              69             100  Fail.................  Pass.               
                       01/97...............            0.93              20              23             100  Pass.................  Pass.               
                       All data............            0.68              18              32  ..............                         ....................
Sigrist..............  Cal.................            0.64              25              40  ..............  Pass.................  No data.            
                       11/96...............            0.87              24              31              88  Pass.................  No data.            
                       12/96...............            0.90              47              77             100  Pass.................  No data.            
                       01/97...............            0.92              21              24             100  Pass.................  No data.            
                       All data............            0.64              19              33  ..............                                             
--------------------------------------------------------------------------------------------------------------------------------------------------------
 Note to Table 2: The initial calibration for the Durag, ESC, and Sigrist units was performed in September and October 1996. The initial calibration for
  the Verewa unit was performed in September and November. The initial calibration for the ESA unit was performed in September and December.            

    a. Correlation Coefficient (r). Proposed PS 11 states that the 
correlation coefficient be at least 0.90 (See Sec. 4.2.1). Tests to 
date indicate that EPA may be unable to produce a linear regression 
line which correlates as well as the proposed performance specification 
indicates.12 EPA believes this is caused by the fact that this 
facility is a worst-case facility for this demonstration test program. 
EPA believes the correlation coefficient specification may have to be 
lower based on the results of this testing.
---------------------------------------------------------------------------

    \12\ Time constraints have required the Agency to temporarily 
ignore the quadratic regression approach described in the 
performance specification. This analysis will be done for the final 
report and the curve which best fits the data will be presented.
---------------------------------------------------------------------------

    Particulate properties depend largely on the wastes fed and the 
accumulated particulate in the APCS. These properties vary considerably 
at this facility, just as the types of wastes fed to the unit vary. 
This variability in the particulate properties causes a varied response 
from the PM CEMS, which in turn causes the correlation coefficient to 
be lower than anticipated. This can be avoided by developing a 
calibration curve for every possible set of particulate properties. 
However as described in the next paragraph, this may not be possible at 
this facility.
    The calibration tests were done under as wide a variety of 
operating conditions as possible. The proposed performance 
specifications and data quality objectives would make a facility such 
as this incinerator to have one calibration for every given operating 
condition, not one that fits all situations as EPA did here. The Agency 
now believes the proposed approach may not be possible for this source. 
As mentioned, particulate does accumulate in the APCS causing wastes 
fed at one time to influence the type of particulate that is emitted 
later. In addition, the wastes arrive at the incinerator in a random, 
uncontrollable manner and in ``small'' batches. This makes it extremely 
difficult for a facility such as this one to determine what 
calibrations it needs and which of those calibrations to use at any 
given time. It might be best for a facility such as this one to have 
one, not many, calibrations to simplify compliance. EPA seeks comment 
on whether this approach, having one calibration curve to cover every 
circumstance rather than several for each circumstance, is acceptable.
    b. Confidence Interval (CI0.95). Proposed PS 11 states that 
CI0.95 be within  20% of the emission limit at the 
emission limit. (See Sec. 4.2.2). This test is done by taking the data 
from the initial (or subsequent) calibration, calculating the 95% 
confidence interval for the regression line, and verifying that the 
upper confidence limit at the emission standard is less than the 
emission standard plus 20% and that the lower confidence limit at the 
emission standard is more than the emission limit minus 20%. In other 
words, this is a single point test at the emission limit.13 Based 
on standards proposed for HWCs, this means the upper confidence limit 
must be less than 83 mg/dscm and the lower confidence limit more than 
55 mg/dscm calculated at the emission limit.
---------------------------------------------------------------------------

    \13\ See above for the discussion of what to do when it is not 
possible to calibrate to the emission limit.
---------------------------------------------------------------------------

    Confidence intervals calculated for these PM CEMS are higher than, 
but close to the proposed specification. This higher value for the 
confidence interval is probably the result of EPA's approach of 
generating one calibration curve at this worst-case facility.
    c. Tolerance Interval (TI0.95). Proposed PS 11 states that 
TI0.95 be within 35% of the emission limit at the 
emission limit. (See Sec. 4.2.3). This test is done by taking the data 
from the initial (or subsequent) calibration, calculating the 95% 
tolerance interval for the regression line, and verifying that the 
upper tolerance limit at the emission standard

[[Page 13783]]

is less than the emission standard plus 35% and that the lower 
confidence limit throughout the calibration range is more than the 
emission limit minus 35%. Like the correlation interval test this is 
also a single point test.14
---------------------------------------------------------------------------

    \14\ Again, see above for the discussion of what to do when it 
is not possible to calibrate to the emission limit.
---------------------------------------------------------------------------

    The calculated tolerance intervals for the various PM CEMS being 
tested are, like the confidence interval, higher than but close to the 
proposed specification.
    d. Relative Calibration Audit (RCA) Tests. The proposed data 
quality objectives state that, to pass an RCA, 75% of the RCA data must 
lie within the 95% tolerance interval. (See Sec. 5.2.3.1 of the 
proposed appendix to Subpart EEE.) All the CEMS passed all the RCAs. 
Therefore, the initial calibration is still valid over time despite the 
changing operating conditions at the facility.
    e. Calibration and Zero Drift (CD and ZD). Proposed PS 11 states 
that CD be within 2% of the calibration standard and that 
the ZD be within 2% of the emission limit. (See Secs. 4.3 
and 4.4, respectively). This test would be done during the ACA tests, 
which the proposed quality assurance requirements stated would be done 
on a quarterly basis.
    As discussed above, most of these CEMS internally check zero and/or 
calibration drift every day. In cases where one or more of the checks 
are not internally done, traceable standards are required to perform 
the check. Most vendors have not supplied these NIST traceable 
standards (or an acceptable substitute) for the ACAs. The ACAs for this 
test program will be done every 4 weeks for these missing parameters as 
soon as EPA obtains these standards from the vendors.
    Where data is available, most CEMS routinely pass the zero and 
calibration checks. In instances where the drift test is failed, the 
CEMS automatically adjusts the failed parameter to within 
specifications. EPA is now quantifying the ``Pass'' and ``Fail'' 
indicators shown in the table. Future reports will quantify values for 
zero and calibration drift rather than express them in the qualitative 
terms ``Pass'' and ``Fail''.
    f. Response Time--Continuous CEMS. Proposed performance 
specification 11 states that continuous-type CEMS respond to a step 
increase in such a way that the CEMS achieve 95% of the final stable 
reading within 2 minutes of the start of the step increase. (See 
Sec. 4.5.1). This requirement is to be certified by the vendors. The 
vendors participating in this program have done so. This specification 
will not be tested during this program unless EPA believes the response 
time is suspect.
    g. Response Time--Batch CEMS. Proposed performance specification 11 
states that the response (i.e., sampling) time for batch-type CEMS be 
no more than one-third of the averaging period. (See Sec. 4.5.2). The 
sampling time for these CEMS are on the order of minutes while the 
averaging period for the PM standard was proposed to be two hours, so 
this requirement has been met.15 But this specification does raise 
several issues which deserve consideration here.
---------------------------------------------------------------------------

    \15\ One of the -gage devices has two sample collection 
tapes to allow for the continuous sampling of flue gas, but the 
other does not. The truly continuous unit collects particulate on 
one tape as the second tape is being analyzed. The unit with one 
tape samples the extracted gas onto that tape and then analyzes it. 
This unit is not a continuous sampler since it is not sampling stack 
gas while measuring the accumulated particulate on the tape. The 
device with one tape could be configured with two tapes to allow for 
the continuous sampling of stack gas. It was not configured with two 
tapes for this test program because the vendor was unwilling to 
incur the cost of supplying such a device for this test program.
---------------------------------------------------------------------------

    While the response time requirement is met for the averaging period 
associated with the standard, ten minute and one hour averages were 
proposed for PM CEMS when used as an operating parameter, i.e., all 
times other than during a comprehensive performance test. The sampling 
time for these devices is less than one third of the one hour average, 
but not less than one third of the short term ten minute average. 
Further complicating this is the fact that the sampling period for 
these devices, while less than ten minutes, is more than half of ten 
minutes. The result is that no averaging could be done to assure 
compliance with a ten minute average. This raises the issue of whether 
the sampling period for PM CEMS needs to be less than one third of the 
ten minute average. EPA believes not, since this requirement is based 
on EPA's belief that a facility will want to base compliance on the 
average of at least three data points. If a facility is willing to base 
compliance on fewer than three data points, it could be allowed to do 
so. This is particularly true for a ten minute average which is likely 
to be quite high relative to the standard or the one hour average. 
Nonetheless, EPA seeks comment on how to address this for the final 
rule.
4. Issues Relative to the Draft Performance Specification 11 for PM 
CEMS
    a. Performance Specifications which Apply for the Calibration 
Curve. The proposed data quality objectives were specific that a 
tolerance interval test be used during an RCA to determine whether the 
calibration curve is still valid. However, the data quality objectives 
were silent on what tests apply to determine whether the initial 
calibration is valid. For this reason, EPA wishes to clarify this 
point.
    To test the validity of the calibration curve, one must check to 
make sure the calibration curve passes the correlation coefficient (r), 
confidence interval, and tolerance interval tests. The correlation 
coefficient is a test of the curve's overall fit. If the calculated 
correlation coefficient is greater than the one published in the 
performance specification, the calibration curve is acceptable. The 
confidence interval test is a single point test at the emission 
limit.16 This verifies the fit of the calibration at the emission 
limit, a critical point when using a CEMS for compliance. The tolerance 
interval test is similar to the confidence interval test in that it is 
a single point test.
---------------------------------------------------------------------------

    \16\ See above for the special case where a facility cannot 
calibrate the PM CEMS to the emission limit.
---------------------------------------------------------------------------

    The confidence interval and tolerance interval tests, though, may 
be redundant. Therefore, we seek comment on whether only one of these 
tests should be used. There is merit to keeping both tests, though. The 
confidence interval test, for instance, ensures that the calibration 
curve is accurate at the standard, a point where a high degree of 
accuracy is required. The tolerance interval test is unique in that it 
sets the maximum deviation the tolerance interval lines can be from the 
linear regression curve at 35% of the emission limit. Therefore, 
commenters should focus their comments on whether these tests are 
indeed redundant. Would a failure of one test conclusively mean the 
other test is also failed? Conversely, would passing one conclusively 
mean a facility would pass the other? If so, which of these tests is 
more stringent?
    b. Number of Tests for the RCA. During the course of this test 
program, EPA has learned that it might be wise to standardize the 
number of tests required for the RCA.17 Other Appendix B 
performance specifications require that 12 tests be performed for 
relative accuracy test audits (RATAs). (RATAs are the equivalent to the 
RCA here.)
---------------------------------------------------------------------------

    \17\ Section 7.3 of the proposed performance specification 11 
for PM CEMS states that the number of tests required for a response 
calibration is 15. This should not be confused with the number of 
tests for a relative calibration audit. Section 7.3 does not apply 
to RCAs.
---------------------------------------------------------------------------

    EPA believes the following approach would be acceptable for RCAs 
and requests comment on it. A facility

[[Page 13784]]

would perform up to 12 manual method measurements. Manual method tests 
may be disqualified and fewer than 12 used if they fail method QA/QC or 
the facility's internal data quality standard, but in no case may the 
number of RCA tests be lower than 9. If fewer than 9 measurements 
remain after the quality audit of the data, a new RCA test is required. 
To pass an RCA, more than 75% of the qualifying, good data must lie 
within the tolerance interval lines.

III. The Hg CEMS Demonstration Tests

A. Site Selection

    For the Hg CEMS demonstration tests, EPA selected the Holnam cement 
kiln #2 in Holly Hill, SC. This cement kiln co-fired hazardous waste 
with other fuels, including fossil fuels such as coal. As such, this 
cement kiln is like many other hazardous waste burning cement kilns.
    Holnam Holly Hill kiln #2 is 18.5 feet wide and 580 feet long with 
a design capacity of 2,100 tons of clinker per day. The main 
ingredients in the cement production are limestone, clay, alumina, and 
iron. The facility also obtains additional raw materials, such as fly 
ash, to supplement raw materials. Raw materials are ground, mixed with 
water, and fed to the cold end of the kiln at a solids content of about 
65%. The hot (discharge) end of the kiln is fired primarily by coal, 
but petroleum coke, waste carbon, shredded tires, hazardous waste, fuel 
oil, and natural gas can also be fired. Kiln #2 has a rated capacity of 
600 M-Btu/hr. Gases pass through the electrostatic precipitators (ESPs) 
specifically designed and built for this facility, through a transfer 
duct, and out the exhaust stack.
    EPA chose to perform the Hg CEMS tests at a cement kiln for many 
reasons. CKs tend to have higher levels of mercury in their flue gas, 
relative to an incinerator or an LWAK, because mercury is fed to the 
kiln in the raw material used for cement production. Other HWCs can 
better avoid mercury in their feed materials, so it is less likely that 
mercury would be present in the flue gases of those sources. Therefore, 
a useful Hg CEMS demonstration program can be conducted at a CK since 
it has mercury in the flue gas. CKs also have higher PM emissions 
relative to other sources. The PM is also likely to contain mercury. 
This is because the PM is derived in large part from the raw material 
that, in turn, can be a significant source of the mercury fed to the 
kiln. Particle bound mercury is difficult for Hg CEMS to measure, so 
this represents a worst case for these instruments. Finally, CKs tend 
to have air pollution control equipment to control PM only. Other 
pollutants are uncontrolled and may be present at high concentrations. 
Since these pollutants, such as SO2 and NOX, may interfere 
with the Hg CEMS's ability to measure mercury, this again is a worst 
case situation for Hg CEMS.
    The Holnam Holly Hill cement kiln #2 was chosen because:

--Data indicated the mercury concentration in the flue gas is 17 
g/dscm without the need to spike mercury;
--The facility was willing to host the demonstration and allow the 
necessary facility modifications;
--Physical access was available at the transfer duct and stack; and
--There was room for installing the Hg CEMS analyzers close enough to 
the sampling point to meet the monitor's maximum sample line 
requirements.

    A detailed description of this selection decision is found in 
Section 1.3 of the Hg Test Plan.

B. Speciated Hg Manual Method.

    One aspect of the program is to determine how well these Hg CEMS 
measure all species of mercury.18 Some mercury monitors measure 
just elemental mercury. Total mercury monitors, or Hg CEMS like those 
participating in this program, measure all mercury regardless of 
species. Most Hg CEMS measure total mercury by first converting all 
mercury to elemental mercury and measuring the amount of elemental 
mercury in the treated flue gas. Converting all mercury to elemental 
mercury adds much complexity to the instrument.
---------------------------------------------------------------------------

    \18\ For the purposes of this discussion, mercury species are 
defined as particle bound, ionic, and elemental mercury.
---------------------------------------------------------------------------

    At the start of this program, no method had been validated to 
measure mercury by species. Many types of speciated mercury methods are 
currently being developed, so EPA chose to validate one of those 
methods to use in this program. This speciated mercury method is 
tentatively called Method 101B.19 The report, Site-specific 
Quality Assurance Test Plan: Method 301 Validation of a Proposed Method 
101B for Mercury Speciation, describes the methodology used to validate 
the method.
---------------------------------------------------------------------------

    \19\ The reader should note that many methods are currently 
being developed to speciate mercury emissions. One of those other 
methods may be better than the method chosen here. This method was 
chosen because EPA knows how to perform this method. Other methods 
are not so well documented. Eventually a method other than the one 
used here may be adopted as the EPA method for mercury speciation.
---------------------------------------------------------------------------

    While EPA is not ready to release the final report on this 
validation, some mention of its validation status is warranted here. 
The method passed all Method 301 criteria without correction with the 
exception of ionic mercury. The Agency has not yet concluded whether 
the method passed for ionic mercury. The issue for ionic mercury is 
that the HgCl2 spiking used for the validation varied so much that 
it caused the calculated relative standard deviation to be much greater 
than Method 301's criteria of 0.50. EPA is now studying how to 
eliminate the effects of HgCl2 spiking from the data. We will 
release the final validation report after this concern has been 
addressed.
    Finally, EPA has not yet determined whether this validation at this 
cement kiln can be transferred to other sources. Mercury species, 
primarily ionic forms such as HgCl2, are very difficult to 
generate, transport, and measure. EPA plans to use this method at other 
sources. Prior to doing so, though, we will perform tests to determine 
how well the validation at the cement kiln transfers to these other 
sources. After this work is completed, EPA will be able to determine 
whether this method should work at other sources. Until this is done, 
however, EPA recommends that a facility wishing to measure mercury by 
species first conduct a full Method 301 validation of the speciated 
mercury method prior to using it.
    The reader should note that EPA has no plans to require facilities 
to use M101B. It was validated so EPA could answer questions it had 
regarding the ability of the Hg CEMS to measure all species of mercury 
simultaneously. A facility would continue to use Method 29 20 to 
measure stack mercury emissions, including any stack tests required for 
Hg CEMS.
---------------------------------------------------------------------------

    \20\ Likewise, a facility could also use Method 101A. M101A is a 
mercury-only emission measurement method. M101A is identical to M29 
except it uses mini-impingers.
---------------------------------------------------------------------------

C. The Test Plan

    Testing started in August 1996 and continued through September. In 
October we discovered that all the Hg CEMS had suffered equipment 
failures. EPA met with the Hg CEMS vendors soon after the problem was 
discovered, and vendors responded to EPA's data availability concerns 
by increasing the ruggedness of their equipment. Testing resumed in 
December 1996. The monitors have responded with less failures since the 
modifications were made.
    As was the case in the PM CEMS testing program, relative accuracy 
test

[[Page 13785]]

audit (RATA) and ACA tests are being performed every four weeks. 
Testing is expected to continue through May 1997 or later.
    An important aspect of the Hg CEMS demonstration tests is to test 
the performance specifications themselves. Revised specifications will 
be promulgated based on the data obtained here and comments received in 
response to the CEMS NODAs.
    Vendor proposals for this test program are found in docket number 
S0205.

D. Hg CEMS Demonstration Test

    Due to the sudden stop and restarting of the Hg CEMS demonstration 
test program, EPA is not prepared to release an interim report for this 
test program. EPA does request comment on the approach we are using to 
demonstrate these Hg CEMS and how to address the variability of spiking 
during the ACA test.
1. Hg CEMS Demonstration Test Approach
    The Agency's approach to demonstrating the Hg CEMS can be found in 
the document, Site-specific Quality Assurance Test Plan: Total Mercury 
CEMS Demonstration.
2. ACA Tests and Spike Variability
    As described in the section above concerning the Method 101B 
validation, similar problems have been encountered spiking known 
concentrations of elemental (Hg0) and ionic (Hg+2) mercury to 
the CEMS. NIST traceable permeation tubes are available for Hg0, 
but not for Hg+2. As a result, performing ACA tests on the Hg CEMS 
with Hg+2 is very difficult. EPA believes it may need to modify 
the proposed performance specification to take this into account.
    Therefore, EPA now believes it is prudent to have facilities 
conduct ACA (i.e., linearity) tests with Hg0 only. Facilities 
would then use this ACA to determine whether the calibration of the 
monitor is still valid and, if it fails the ACA (or if this ACA is the 
first performed), use the ACA results as the basis for a new 
calibration. Spiking with Hg+2 would be done only for the purposes 
of ensuring the Hg CEMS adequately measured Hg+2. In other words, 
the Hg+2 test would resemble the NOX converter efficiency 
test prevalent for NOX CEMS. In this case a facility would spike 
an amount of Hg+2 within some range (for instance, within 75 to 
125% of the emission standard) and ensure that the measured amount of 
Hg reported by the analyzer is within an acceptable range (for 
instance, within 20%) of the actual Hg+2 spike 
determined by the manual method. The actual ranges will be determined 
based on data obtained from these tests. EPA requests comment on 
whether this approach is appropriate.

    Dated: March 12, 1997.
Elizabeth Cotsworth,
Acting Director Office of Solid Waste.
[FR Doc. 97-7215 Filed 3-20-97; 8:45 am]
BILLING CODE 6560-50-P