Responses to Comments Document:

MACT Standard for Primary Magnesium Refining

Metals Group
Emission Standards Division
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency

August 18, 2003


-------
1.	Introduction

Only one short comment and one substantive comment were received on the proposed
rule. The short comment suggested we cut the proposed standard by 50 percent. However, the
commenter did not provide any information or rationale to support such a request.

Consequently, we could not consider it further and made no changes to the rule as a result of the
comment. Substantive comments (Docket Item OAR-2002-0043-0002 dated February 20, 2003)
were received from US Magnesium LLC and are addressed in the following sections.

2.	Comments and Responses

2.1 Dioxin/Furan Limit Not Needed

Comment: The commenter stated that a dioxin/furan emission limit is not appropriate
for the primary magnesium industry because EPA has applied these limits primarily to facilities
that burn wastes. Other industries, such as petroleum refineries and iron and steel foundries, are
known to emit dioxin/furan; however, EPA did not propose limits for them. The commenter also
stated that the dioxin/furan limit cannot be justified on the basis of health risk because the
facility is in a remote location, and the nearest resident is 25 miles away. The commenter
recommended that EPA use PM as a surrogate for dioxin/furan emissions from the melt reactor
because: (1) EPA established MACT for dioxin/furan as the PM control devices on the melt
reactor, (2) PM is used as a surrogate for other pollutants in this rule and has been used as a
surrogate for dioxin/furan in other rules, (3) the dioxin/furan emissions are mainly in particulate
form, (4) the dioxin/furan limit will obtain no additional reduction beyond that obtained using
PM as a surrogate, and (5) the dioxin/furan limit will add significantly to the cost of stack testing
with no apparent gain.

Response: We set a dioxin/furan limit because it is a HAP of concern with respect to
toxicity, we have adequate test data (two tests composed of three runs each) to characterize
emission control performance, and dioxin/furan formation and control is not always correlated to
PM formation and control. First, the formation of dioxin/furans in combustion devices with an
available source of chlorine is well documented, and it is not a concern only for facilities that
burn waste. The test data from this industry confirms the formation and emissions of
dioxin/furans from this emissions source. In this case, it is more appropriate to set a limit for

1


-------
dioxin/furan rather than to use PM as a surrogate. Second, we do not agree that the control
device for PM will adequately control the emissions of dioxin/furans. There are factors other
than the PM control device which may affect the formation and control of dioxin/furan, such as
the composition and concentrations of precursors, temperature, and process conditions. Dioxins
are formed in acid gases leaving the combustion device, and the means of control is not
necessarily the particulate control system but quenching of gases to control the temperature in
the device (to assure that temperature does not fall in the range which optimizes dioxin/furan
formation).

The MACT control system for dioxin/furans is the entire scrubber train - the packed
tower scrubbers (for HC1 control) and the venturi scrubber (for PM control) - and not just the PM
control device. That is, the control of dioxin/furans includes the rapid cooling of the exhaust gas
that occurs in the packed tower absorbers, which limits the dioxin/furan formation. Therefore,
we believe a dioxin/furan limit is necessary to ensure that process and control device operations
do not change in the future in a manner that might increase the formation and release of
dioxin/furan, even if the overall PM control level remains the same.

The dioxin/furan limit is not based on a determination that health risks exist; it is based
on technology and the floor level of control that has been achieved. We do not believe that stack
testing every 2.5 years is costly or unreasonable to provide assurance that the dioxin/furan limit
is being achieved. Moreover, the commenter did not provide any information as to how this
stack testing will add significantly to the costs of compliance with the NESHAP.
2.2 Dioxin/Furan Emission Limit

Comment: The commenter disagreed with the approach used to set the emission limit
for dioxin/furan and claimed it does not provide a reasonable margin of safety to ensure
continuous compliance. The commenter suggests a level of 50 ng TEQ/dscm is statistically
valid. However, the commenter recommended that a minimum safety factor of three be applied
to the average of results from the two stack tests (21.5 ng TEQ/dscm) to develop a limit of 65 ng
TEQ/dscm rather than a limit of 36 ng TEQ/dscm as proposed. The commenter believes this is
reasonable because of the high variability in the test results and because of the inherent
inaccuracies in the dioxin/furan sampling and analysis, especially at these extremely low levels
of detection.

2


-------
Response: We chose 36 ng TEQ/dscm because it was the highest result from any of the
six runs. This approach accounts for inherent variability, and an additional margin of safety is
provided by determining compliance from the average of three runs. It appears that the
commenter considered the six test runs in developing a standard deviation of ± 9.5 ng TEQ/dscm
and estimating a 99th percentile for single test runs. The variability of the average of three runs
is more appropriate than the variability of a single test run because compliance is determined
from the average of three test runs rather than for each single test run.

To illustrate the impact of using the average of three runs, we performed a Monte Carlo
simulation of 5,000 runs based on a normal distribution developed from the test results for six
runs. (Details are given in Appendix A.) From the simulation, the 99th percentile for individual
runs was 44 ng TEQ/dscm compared to a 99th percentile of 32 ng TEQ/dscm for the average of
three runs. Consequently, since the emission limit is enforced based on three-run averages, the
proposed limit of 36 ng TEQ/dscm is close to the 99th percentile of performance. We believe
that the limit as proposed is achievable, and the simulation indicates it accounts for variability.

The commenter mentioned process variability and uncertainty associated with sampling
and analysis as reasons for a higher limit. However, the variability in the process, sampling, and
analysis are inherently included in the runs we used to derive the limit, and using the highest run
accommodates this variability. In addition, there is no need to artificially increase the limit by
multiplying the average of the test results by three because the statistical simulation shows that
the proposed limit is reasonable. With testing performed every 2.5 years and a limit at about the
99th percentile, the limit would be exceeded no more than once every 250 years if the process
and control device are operated as they were during the two performance tests.

While we were evaluating the data discussed by the commenter, we discovered an error
in the 1998 test report. The test contractor inadvertently switched the TEF for two congeners.
The net effect is that the overall average for six runs is 18 ng TEQ/dscm instead of 21.5 ng
TEQ/dscm. This correction had no effect on the highest run and did not change the limit that
was originally proposed.

2.3 Use Updated Toxicity Equivalence Factors (TEF)

3


-------
Comment: The commenter believes that the World Health Organization's 1998 TEF
scheme should be used to assign toxic equivalency, and this scheme should be stated in the final
rule.

Response: Based on our dioxin reassessment report, we agree with the commenter and
have incorporated the updated TEF scheme in the final rule. The effect on the test results was
small, and the highest run remained at 36 ng TEQ/dscm. Consequently, the level of the standard
was not changed. Details are provided in Appendix B.

2.4	Health Effects of Manganese

Comment: The commenter stated that manganese emissions from the facility are very
low. Workers at the plant do not show any evidence of the health effects described in the
preamble for chronic exposure to high levels of manganese by inhalation.

Response: We did not imply in the preamble that the workers at this plant exhibited the
characteristics of exposure to manganese. The preamble contained a generic description of the
health effects of several HAP, including manganese, and was provided only as background to
show why these pollutants are hazardous air pollutants.

2.5	Modern Electrolytic Cells

Comment: The commenter believes the rule should include descriptions and
requirements for modern electrolytic cells and the proper handling and collection of the cell off-
gases. The modernization of electrolytic cells and chlorine capture equipment has reduced
chlorine emissions by 75 percent.

Response: We did not include a detailed description of the modern cell technology, and
we agree that we should point out that the old cell technology (known as the IG Farben cells) has
been replaced. The new cell technology is a closed system and has resulted in reductions in
chlorine emissions. The old cell technology allowed chlorine gas to escape from the anode
section of the cell and infiltrate into the cathode section, where it was difficult to capture and
control. The improvement in emission control is evidenced by the reduced chlorine emissions as
reported for the Toxics Release Inventory (TRI).

Our proposed rule is based on the most current permit requirements at the time of
proposal (permit dated October 11, 2001). This permit reflects the operating conditions after the

4


-------
old cell technology was replaced, and the permit specifies the old cell technology must be
replaced by October 1, 2001. Consequently, the rule and operating permit reflect current
operations after the replacement of the old IG Farben cells.

5


-------
APPENDIX A. RESULTS OF THE MONTE CARLO SIMULATION OF TEQ RUNS
Issue

EPA proposed a dioxin limit of 36 ng/m3. Commenter OAR-2002-0043-0002 (US
Magnesium LLC) stated that the 99th percentile is 50 ng/m3 and requests a standard of 65 ng/m3
because of variability and inherent inaccuracy associated with sampling and analysis.

Background

The test results are given in Table A-l. EPA chose the highest single run (36 ng/m3) as
the limit. The limit is based on the average of 3 runs, which further allows for variability.

Table A-l. Dioxin Test Results1

Run (year)

ng/m3

Run 1 (2000)

24.2

Run 2 (2000)

36.0

Run 3 (2000)

10.5

Run 1 (1998)

13.6

Run 2 (1998)

7.4

Run 3 (1998)

17.4





Average

18.2

Standard deviation

10.5

Maximum

36

Monte Carlo Simulation

According to the Shapiro-Wilke test, the individual runs are normally distributed.

A Monte Carlo simulation was performed for the normal distribution and results for
5,000 runs were simulated. Three-run averages were calculated. Summary statistics are
given in Table A-2.

The 99th percentile for single runs is 44 ng/m3. Averaging over 3 runs reduces the
variability and results in a 99th percentile of 32 ng/m3.

1 From the test reports listed as Docket Items II-A-4 (1998) and II-A-7 (2000) in Docket
No. A-2002-0027.

6


-------
Conclusion

The proposed limit adequately accounts for variability, especially considering the limit is
enforced based on 3-run averages.

Table A-2. Results of the Monte Carlo Simulation (5,000 Runs)



Individual runs
(ng/m3)

3-run averages
(ng/m3)

Average

18.2

18.2

Standard deviation

10.5

6.3

99th percentile

44

32

Maximum

53

42

7


-------
APPENDIX B. UPDATED TOXICITY EQUIVALENCE FACTORS

Tables B-l and B-2 present the calculated toxicity equivalence (TEQ) based on the
toxicity equivalence factors (TEF) recommended by the World Health Organization (WHO) in
1998. The most significant change in terms of these test results was that the TEF for 1,2,3,7,8-
PeCDD increased from 0.5 to 1. Other TEFs that changed include a reduction by a factor of 10
for OCDD and OCDF.

Table B-l. Summary of TEQ Results - US Magnesium (March 1998)



WHO

PCDD/PCDF (ng/mJ)

TEQ (ng/m3)

Congener

TEFs













(1998)

Run 1

Run 2

Run 3

Run 1

Run 2

Run 3

2,3,7,8 -TCDD

1

0.05

0.04

0.05

0.050

0.040

0.050

1,2,3,7,8-PeCDD

1

0.14

0.08

0.22

0.140

0.080

0.220

1,2,3,4,7,8-HxCDD

0.1

0

0

0

0.000

0.000

0.000

1,2,3,6,7,8-HxCDD

0.1

0.1

0.1

0.2

0.010

0.010

0.020

1,2,3,7,8,9-HxCDD

0.1

0.1

0.1

0.1

0.010

0.010

0.010

1,2,3,4,6,7,8—HpCDD

0.01

0.1

0.1

0.1

0.001

0.001

0.001

OCDD

0.0001

0.1

0

0.1

0.000

0.000

0.000

2,3,7,8-TCDF

0.1

17.2

9

18.5

1.720

0.900

1.850

1,2,3,7,8-PeCDF

0.05

21.9

13.4

30.7

1.095

0.670

1.535

2,3,4,7,8-PeCDF

0.5

7.9

4.7

9.9

3.950

2.350

4.950

1,2,3,4,7,8-HxCDF

0.1

13.1

8.8

18.5

1.310

0.880

1.850

1,2,3,6,7,8-HxCDF

0.1

7.7

4.7

11.1

0.770

0.470

1.110

2,3,4,6,7,8-HxCDF

0.1

4.4

2.6

5.1

0.440

0.260

0.510

1,2,3,7,8,9-HxCDF

0.1

0.8

0.6

1.2

0.080

0.060

0.120

1,2,3,4,6,7,8-HpCDF

0.01

8.3

4.6

11.1

0.083

0.046

0.111

1,2,3,4,7,8,9-HpCDF

0.01

1.9

1.2

2.5

0.019

0.012

0.025

OCDF

0.0001

0

4.9

8.3

0.000

0.000

0.001

Total TEQ (ng/in ')









9.7

5.8

12.4

Percent 02









11.0

10.0

11.0

02 Correction Factor









1.4

1.3

1.4

Total TEO Tne/m3 at 7% O.)







13.6

7.4

17.4

8


-------
Table B-2. Summary of TEQ Results - US Magnesium (May 2000)



WHO

PCDD/PCDF (ng/mJ)

TEQ (ng/m5)

Congener

TEFs













(1998)

Run 2

Run 3

Run 4

Run 2

Run 3

Run 4

2,3,7,8 -TCDD

1

0.04

0.06

0.03

0.040

0.060

0.030

1,2,3,7,8-PeCDD

1

0.26

0.47

0.16

0.260

0.470

0.160

1,2,3,4,7,8-HxCDD

0.1

0.18

0.29

0.05

0.018

0.029

0.005

1,2,3,6,7,8-HxCDD

0.1

0.53

0.91

0.16

0.053

0.091

0.016

1,2,3,7,8,9-HxCDD

0.1

0.44

0.74

0.13

0.044

0.074

0.013

1,2,3,4,6,7,8—HpCDD

0.01

0.97

1.86

0.27

0.010

0.019

0.003

OCDD

0.0001

0.78

1.17

0.14

0.000

0.000

0.000

2,3,7,8-TCDF

0.1

7.64

9.82

5.67

0.764

0.982

0.567

1,2,3,7,8-PeCDF

0.05

21.16

32.05

9.25

1.058

1.603

0.463

2,3,4,7,8-PeCDF

0.5

12.12

21.08

8.45

6.060

10.540

4.225

1,2,3,4,7,8-HxCDF

0.1

36.69

71.22

11.13

3.669

7.122

1.113

1,2,3,6,7,8-HxCDF

0.1

21.33

39.79

6.15

2.133

3.979

0.615

2,3,4,6,7,8-HxCDF

0.1

9.73

17.31

2.91

0.973

1.731

0.291

1,2,3,7,8,9-HxCDF

0.1

8.45

15.31

2.48

0.845

1.531

0.248

1,2,3,4,6,7,8-HpCDF

0.01

41.17

72.04

12.3

0.412

0.720

0.123

1,2,3,4,7,8,9-HpCDF

0.01

17.41

30.9

3.12

0.174

0.309

0.031

OCDF

0.0001

124.15

185.83

18.13

0.012

0.019

0.002

Total TEQ (ng/m;i)









16.5

29.3

7.9

Percent 02









11.4

9.6

10.4

02 Correction Factor









1.5

1.2

1.3

Total TEO Tne/m3 at 7% CM







24.2

36.0

10.5

9


-------