Quality Control Review of Mercury and PCDD/PCDF Emissions Data From Cement Kilns Used for MACT Evaluations

David G. Gossman

Gossman Consulting, Inc.

INTRODUCTION

In order to perform Maximum Achievable Control Technology (MACT) evaluations, the United States Environmental Protection Agency (USEPA) collects data from a wide variety of sources. A review of mercury (Hg) and Polychlorinated Dibenzo-Dioxin/Polychlorinated Dibenzo-Furan (PCDD/PCDF) emission data from cement kilns gathered for MACT evaluations reveals data quality problems associated with test methodology, lab Quality Assurance/Quality Control (QA/QC) data, calculations and facility operating conditions. These factors are used to develop data quality grading systems for evaluating mercury and PCDD/PCDF emission data from cement kilns. Results indicate that MACT evaluations can be significantly impacted by poor or inappropriately utilized data.

MERCURY

The objective of the data review was to analyze the quality of mercury emission data used in the EPA's Locating and Estimating Air Emissions From Sources of Mercury and Mercury Compounds⁽¹⁾ document to assess mercury emissions from cement kilns. These data were summarized in Table C-2 of the Locating and Estimating Air Emissions From Sources of Mercury and Mercury Compounds document. In order to appropriately assess the data quality used in Table C-2, a copy of each stack test report referenced in Table C-2 was obtained. A data quality rating system was developed in order to evaluate those reports. This data quality grading system is contained in Table 1 and is based on the familiar A, B, C, D, F ranking system. The QA/QC review of the reports used to create Table C-2, as well as comments relative to the calculation found in Table C-2, are found in Table 2. Recalculated emissions factors for tests from Table C-2 are then found in Table 3. Tests that received a rating of D or F are not used in this reassessment.

It also should be noted that some of this emission data was obtained during EPA required certification tests for cement kilns burning hazardous waste fuels. Because these tests require "worst case" operation during testing, the data may reflect somewhat higher emission rates than normally might be expected. Nevertheless, because of the volatility of mercury, this is expected to be a minor factor.

Of fifteen test results used by EPA to calculate an industry-wide emission factor, five were judged to have been inappropriately used. The average emission factor for this corrected set of data dropped from 0.171 E-3 lb/ton of clinker to 0.091 E-3 lb/ton of clinker.

PCDD/PCDFs

A QA/QC review of PCDD/PCDF data⁽²⁾ for nine cement kilns in California was performed initially. This review revealed an issue that required further refinement of the grading system prior to moving forward with additional reviews. In order to provide perspective, a brief review of what had been done is provided. Options and rationale for changes are then evaluated.

A data quality rating system (Table 4) had been developed and used to evaluate nine test reports which had sufficient data, including lab data and blanks. Of these nine, five received an "F" and four received an "A" based on the initial criteria. To perform this review, each run's data were entered into a spreadsheet and TEQs were calculated in accordance with USEPA criteria. One clear difference between EPA and California guidance is that for EPA, non-detects (ND values) are treated as zero. California uses half the detection limit, therefore TEQ calculations from the original reports could not be relied upon..

All five of the "F" graded tests were the result of significant blank contamination. The dilemma is that in some cases the detected quantity of PCDD/PCDFs is very low. Therefore, some of those "F" graded tests represent very low worst case (because of blank contamination) results. In fact, because of the tendency to find some small blank contamination in every analysis, removing these data from the overall data set would likely bias the dataset high. Two examples may be found in Tables 5 through 12. Both have blank contamination, one at a high level the other at a very low level.

Given this situation, the following alternatives were considered:

1. Eliminate all "F" graded data. (This would bias the dataset higher than actual.)

2. Create a new grade of "C" for blank contaminated data that is below some defensible threshold. The thresholds could be:

a) The industry average for all "A" data

b) The 0.2 ng/dscm @ 7% O₂ limit which has been proposed by EPA

c) Some other defensible limit or criteria

Alternative 2 with threshold choice a) was chosen as most defensible since false positive values from contamination should not be used to raise the average, but could be considered in lowering the average for the industry.

The next issue affecting data quality and calculations is the appropriate use of estimated maximum possible concentration (EMPC) values⁽³⁾. What then is the correct way of handling a reported EMPC value for a TEQ calculation? One interpretation is to zero the EMPC or, in essence, not include it for calculating the TEQ. Presumably, this practice would occur because the EMPC value falls below the normal QA/QC quantitative reporting limits. This interpretation is consistent with a recent (January, 1995) draft of SW-846 Method 8280A⁽⁴⁾.

There are, nevertheless, EPA personnel who believe it to be more appropriate to use EMPC values in a calculation of TEQ⁽⁵⁾. Because of this confusion, ongoing reviews of PCDD/PCDF data will include calculations of TEQ both using and not using EMPC values. Tables 9 through 12 provide an excellent example illustrating the impact of EMPC values. As much as half the total PCDD/PCDF and 25% of the TEQ PCDD/PCDF are accounted for by EMPC values. Any comparison of PCDD/PCDF emission levels, standards or limits must clearly evaluate this factor and standardize this calculation to be technically defensible.

CONCLUSION

Mercury and PCDD/PCDF emissions from cement kilns have been examined for data quality relative to their use in establishing MACT limits. A significant percentage of reported data was found to be of suspect quality for a variety of reasons. These reasons include:

1) contaminated field or laboratory blanks,

2) inappropriate sampling methods,

3) erroneous calculations, and

4) inappropriate (or non-normal) operating conditions that could affect emissions.

The use of poor quality data can significantly impact MACT evaluations. If the data used by EPA is to be used to establish MACT standards, it is clear that industry must carefully and thoroughly scrutinize EPA's database to insure the development of technically defensible limits and testing criteria.

Table 1. Mercury data quality rating system.

Rating	Description
A	All QA/QC data available. No problems with blanks or recoveries. All values represent detected and quantified levels of emissions.
B	All values detected and quantified, but missing QA/QC blank and recovery data; or, data represents half the detected limit.
C	QA/QC recovery data out of range; or, values represent approximations. They are below quantitation limits and greater than detection limits; or, there is no indication of whether or not the data represent quantitation/detection limits.
D	Data obtained during abnormal operating conditions that could significantly effect emissions; or, totals incorrectly calculated.
F	High blank data (field or lab) relative to quantified amount; or Data obtained while artificially spiking the emission parameter. Inappropriate sampling methodology that could miss the parameter being tested.

Table 2. QA/QC review of EPA L&E document Table C-2.

Facility	Grade	Comments
Lone Star Cape Girardeau Phase 1	A	All data, calculations, and QA/QC satisfactory.
Lone Star Cape Girardeau Phase 2	F	Hg was spiked during test.
Lafarge Corp. Demopolis, AL	C	Not a COC test. Significant blank contamination from a variety of metals including Hg in portions of the train; but, primary sources of Hg do not appear significantly affected. Insufficient data in report to review/repeat calculations, i.e., no sampled gas volumes. Test report did not note problems with blank contamination.
Ash Grove Foreman, AR Kiln #3	B	Values reported as "<", yet QA/QC and Hg input rates suggest data is real emission rate. Clinker production rate is low-should be 54 tons/hr. EPA reported emission factors approximately 10% high because of clinker production error.
Ash Grove Foreman, AR Kiln #1	B	Values reported as "<", yet QA/QC and Hg input rates suggest data is real emission rate.
Ash Grove Chanute, KS Kiln #1	D	COC report shows abnormally high Hg input rate. Input during test was approximately ten times average '94 levels.
Ash Grove Chanute, KS Kiln #2	B	Values reported as "<", yet QA/QC and Hg input rates suggest data is real emission rate.
Ash Grove Louisville, NE Kiln #1	C	Values reported as "<". Number of runs "5" not "3". Clinker production rate is 58 tons/hr, not 44. EPA reported emission factors approximately 25% high because of clinker production error.
Ash Grove Louisville, NE Kiln #2	B	Values reported as "<", yet QA/QC and Hg input rates suggest data is real emissions rate. Clinker production rate 75 tons/hr, not 63 tons/hr. EPA reported emissions factor approximately 20% high because of clinker production error.
ESSROC Fredrick, MD Test No. 1	A	All data, calculations and QA/QC are satisfactory. Not a COC test.
ESSROC Fredrick, MD Test No. 2	A	Minor error in blank correction on two runs that does not significantly effect data. All other calculations and QA/QC are satisfactory. Not a COC test.
Lafarge Corp. Paulding, OH	A	All test data calculations and QA/QC are satisfactory. Not a COC test. Clinker production rate is 30 tons/hr, not 61 tons/hr. EPA reported emissions factor approximately one-half the actual value because of clinker production error.
Lone Star Oglesby Test No. 1	F	Not a COC test. Used inappropriate stack method. Analyzed for Hg in stack particulate - did not use metal/Hg impinger train.
Lone Star Oglesby Test No. 2	F	Not a COC test. Used inappropriate stack method. Analyzed for Hg in stack particulate - did not use metal/Hg impinger train.
Holnam Clarksville, MO	A	All data, calculations, QA/QC satisfactory.

Table 3. Recalculation of Hg emission factors for portland cement manufacture using EPA Table C-2 QA/QC corrected data.

Facility	Location	EPA L&E Average 10-3 lb/ton Clinker	Grade	Corrected Average 10-3 lb/ton clinker
Lone Star	Cape Girardeau, MO	0.02	A	0.02
Lone Star	Cape Girardeau, MO	0.43	F	---------
Lafarge Corp	Demopolis, AL	0.16	C	0.16
Ash Grove	Foreman, AK	0.035	B	0.032
Ash Grove	Foreman, AK	0.07	B	0.07
Ash Grove	Chanute, KS	0.97	D	---------
Ash Grove	Chanute, KS	0.15	B	0.15
Ash Grove	Louisville, NE	0.095	C	0.072
Ash Grove	Louisville, NE	0.03	B	0.025
ESSROC	Fredrick, MD	0.22	A	0.22
ESSROC	Fredrick, MD	0.22	A	0.22
Lafarge Corp	Paulding, OH	0.032	A	0.059
Lone Star	Oglesby, IL	0.0045	F	---------
Lone Star	Oglesby, IL	0.028	F	---------
Holnam	Clarksville, MO	0.097	A	0.097
Averages		0.171		0.091*

* Second ESSROC test on same kiln not included in average.

Table 4. PCDD/PCDF data quality rating system.

Rating	Description
A	All QA/QC data available. No problems with blanks or recoveries. Totals and TEQs represent detected and quantified levels of emissions. High resolution GC/MS used.
D	Data obtained during abnormal operating conditions that could significantly effect emissions.
F	High blank data (field or lab) relative to quantified amount. Inappropriate sampling or laboratory methodology that could miss the tested for parameter.
NR	Not Rated - missing QA/QC data for adequate evaluation.

1. Midwest Research Institute, Locating and Estimating Air Emissions From Sources of Mercury and Mercury Compounds, U.S. Environmental Protection Agency, Research Triangle Park, 1993, Appendix C.

2. (a) Report of Air Pollution Source Testing of a Cement Plant Rotary Kiln Fired on Rubber Tires and Coal At Mitsubichi Cement Company in Lucerne Valley, California, Prepared by Engineering-Science, Inc. 18 Mason Irvine, CA 92718, 1990.

(b) Report of Air Pollution Source Testing of a Cement Plant Rotary Kiln Fired on Rubber Tires and Coal At Mitsubichi Cement Company in Lucerne Valley, California, Prepared by Engineering-Science, Inc., 18 Mason Irvine, CA 92718, 1993.

(c) Report of California AB2588 Air Pollution Source Testing on the Rotary Kiln Baghouse Stack at California Portland Cement Company in Mojave, California, Prepared by Engineering-Science, Inc., 6060 North Irwindale Ave. Suite J Irwindale, CA 91706, 1992.

(d) Report of Air Pollution Source Testing for California AB2588 at RMC Lone Star Cement Co. in Davenport, California, Prepared by Engineering-Science, Inc., 75 North Fair Oaks Ave. Pasadena, CA 91109, 1990.

(e) Emissions Associated With the Combustion Pulp and Paper Mill Praimary Sludge on a Cement Kiln, Prepared by Radian Corporation, 10395 Old Placerville Road Sacramento, CA 95827, 1989.

(f) Report of AB2588 Air Pollution Source Testing at Colton Cement Corporation, Prepared by Engineering-Science, Inc., 75 North Fair Oaks Ave. Pasadena, CA 91109, 1990.

(g) Air Toxics "Hot Spots" Information and Assessment Act (AB2588) - Air Toxics Inventory Report, Prepared by Sierra Research, Inc., 1521 I Street Sacramento, CA 95814, 1990.

(h) Toxic Air Emissions Assessment White Cement Kiln, Unit No. 1 Riverside Cement Company Crestmore Facility Volume I - Test Results, Prepared by Energy and Environmental Research Corporation, 18 Mason Irving, CA 92718, 1990.

(i) Report of Air Pollution Source Testing for California AB2588 at Calaveras Portland Cement Company in Monolith, California, Prepared by Engineering-Science, Inc., 6060 North Irwindale Ave. Suite J Irwindale, CA 91706, 1992.

3. C. Cape, "A Caution in the Use of Analytical Data in Calculating TEQ Values for Dioxin Reporting," GCI Tech Notes, 3(12):3 (1995).

4. Office of Solid Waste, Test Methods for Evaluating Solid Waste, , Method SW-846 8280A "The Analysis of Polychlorinated Dibenzo-P-Dioxins and Polychlorinated Dibenzofurans," U.S. Environmental Protection Agency, Research Triangle Park, Proposed 1995.

5. J. Woodford, Gossman Consulting, Inc. Springboro, OH, personal communication with G. McCallister, USEPA, 1995.