Photochemical Assessment Monitoring
Stations (PAMS)
Performance Evaluation Program
Final Report
Contract No. 68-D3-0095
Delivery Order 11
Prepared for:
Neil J. Berg
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
Prepared by:
Eastern Research Group, Inc.
P. O. Box 2010
Morrisville, North Carolina 27560
June 1997
-------
DISCLAIMER
The U.S. Environmental Protection Agency through its Office of Air Quality Planning and
Standards funded and managed the research described here under Contract No. 68-D3-0095 to
Eastern Research Group, Inc. It has been subjected to the Agency's peer and administrative
review and has been approved for publication as an EPA document. Mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
11
-------
ABSTRACT
Humidified performance evaluation samples containing volatile organic compounds were
prepared for PAMS in 6-liter SUMMA®-polished stainless steel canisters. Before preparation,
each canister was cleaned and certified for total nonmethane organic compound (NMOC) content
to less than 10 ppbC, and analyzed by gas chromatograph/mass spectrometer-flame ionization
detector (GC/MS-FID) to verify cleanliness on a compound-specific basis. Performance
evaluation samples were prepared from high pressure multiple-component gas cylinders obtained
from a specialty gas manufacturer at a pressure of approximately 30 psig with a relative humidity
of about 70%. The canisters were prepared at a concentration appropriate for direct analysis by
the analytical system without the use of any slip-stream or pressure reduction devices. Each
canister was analyzed by Eastern Research Group (ERG) prior to shipment to the PAMS
participants, and the results of the ERG analyses were pooled to provide a range of two standard
deviations around the ERG analytical result for each canister. Participants analyzed the canisters
in replicate and reported both sets of analytical results. The canisters analyzed by the participants
were analyzed on a variety of instruments, with a variety of analytical conditions. Samples were
submitted to 37 PAMS, with 98 individual sets of data received. Each PAMS site was requested
to perform and report two analyses (37 x 2=74 datasets). However, several PAMS sites elected
to perform additional replicate determinations on additional instruments (12 x 2 = 24 additional
datasets). The total of individual datasets was thus 74 + 24 = 98 datasets. Within 48 hours,
PAMS were notified of the comparison of their analytical results to the ERG analytical results for
the same canister. Approximately 30% of the returned canisters were re-analyzed by ERG on
their return to the laboratory. Canisters selected for re-analysis generally had a major discrepancy
between the ERG analysis and the external analysis. Four canisters were prepared and retained in
the ERG laboratory to be analyzed again at the end of the study to provide information on the
stability of the test compounds. The range of analytical results for the PAMS participants was
wider than the range of the ERG analyses, but this result is to be expected since the ERG
canisters were prepared in a short period of time and analyzed on one instrument over a short
period of time. The results of the analysis can certainly be used by the PAMS to point out areas
of the analytical procedure that can be improved.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
111
-------
TABLE OF CONTENTS
Page
EXECUTIVE SUMMARY 1-1
1.0 INTRODUCTION 1-4
1.1 PAMS Quality Assurance Work Group 1-6
1.2 Performance Evaluation Canister Preparation Procedures 1-7
2.0 CONCLUSIONS AND RECOMMENDATIONS 2-1
3.0 LABORATORY PROCEDURES 3-1
3.1 Cleaning of SUMMA®-Passivated Stainless Steel Canisters 3-1
3.2 Verification of Cleanliness of the SUMMA®-Polished Canisters 3-1
3.3 Verification of Canister Cleanliness on a Compound-Specific Basis (GC/MS-FID
Analysis) 3-2
3.4 Preparation of Performance Evaluation Samples 3-3
3.5 GC/MS-FID Analysis of Performance Evaluation Samples by ERG 3-4
3.6 Analysis by PAMS 3-7
3.7 Preparation of a Database of PAMS Results 3-7
3.8 Repeated Analysis of Performance Evaluation Samples 3-7
3.9 Quality Control Information 3-8
3.9.1 Precision 3-10
3.9.2 Bias 3-11
3.9.3 Completeness 3-12
4.0 RESULTS AND DISCUSSION 4-1
4.1 Analysis Results 4-1
4.2 Stability Study 4-42
4.3 Outlier Analysis 4-48
4.4 Additional Compounds Identified 4-49
4.5 ERG Re-Analysis of Performance Evaluation Canisters 4-52
4.6 Additional Statistical Calculations Relative to the Mean of the PAMS
Analyses Excluding Outliers 4-57
5.0 REFERENCES 5-1
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
-------
TABLE OF CONTENTS (Continued)
APPENDICES
A
Analytical Results for Each Analysis of Each Canister
B
Analytical Results Pooled for All PAMS
C
Equipment and Analytical Conditions Used by PAMS
D
Reanalysis Data
E
ERG Theoretical and Analytical Data
F
Percent Difference Between Analyses Arranged According to Laboratory
G
Percent Difference Between Replicate Analyses Arranged According to
Compound
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
V
-------
LIST OF TABLES
Page
1-1 Ozone Precursors 1-5
1-2 Components of Certified Gas Cylinders 1-8
3-1 Analytical Conditions for GC/MS-FID Analysis of Ozone Precursor Compounds .... 3-3
3-2 Summary of PAMS Performance Evaluation Quality Control Procedures 3-9
4-1 Original Batch of Forty Canisters: Theoretical Amounts 4-2
4-2 Original Batch of Forty Canisters: ERG Analyzed Values 4-4
4-3 Example Reported Dataset for a PAMS Site 4-7
4-4 Statistical Evaluation of Results Obtained from all PAMS 4-8
4-5 Ranges for ERG Analyses and PAMS 4-34
4-6 Stability Study Performed on Canisters Retained at the ERG Laboratory
Canister #2061 4-43
4-7 Stability Study Performed on Canisters Retained at the ERG Laboratory
Canister #2078 4-44
4-8 Stability Study Performed on Canisters Retained at the ERG Laboratory
Canister #2208 4-45
4-9 Stability Study Performed on Canisters Retained at the ERG Laboratory
Canister #GP00036 4-46
4-10 Additional Compounds Reported and Frequency of Reporting 4-50
4-11 Additional Compounds Identified in ERG Stability Study Canisters 4-51
4-12 Percent Difference Between First and Second ERG Analysis 4-53
4-13 Compounds Most Frequently Not Reported by PAMS Sites 4-78
4-14 Reproducibility of the PAMS Measurements by Compound 4-80
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD VI
-------
LIST OF TABLES (Continued)
Page
4-15 Statistical Calculations by Compound Compared to the Mean of the
Reported Values from the PAMS Sites, Excluding Zeros and Outliers 4-81
4-16 Summary of Statistical Calculations by Laboratory 4-83
4-17 Summary of Statistical Calculations by Laboratory for Absolute Bias 4-86
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
vii
-------
LIST OF FIGURES
Page
3-1 Chain of Custody Form Accompanying Each Canister Shipped 3-5
3-2 Letter of Instruction Accompanying Each Canister Shipped 3-6
4-1 Graphical Comparison of Analytical Results for Ethylene Between ERG and Participant4-9
4-2 Graphical Comparison of Analytical Results for Ethane Between ERG
and Participants 4-10
4-3 Graphical Comparison of Analytical Results for Propane Between ERG
and Participants 4-11
4-4 Graphical Comparison of Analytical Results for 1-Butene Between ERG and Participa4td 2
4-5 Graphical Comparison of Analytical Results for //-Butane Between ERG and Participants! 3
4-6 Graphical Comparison of Analytical Results for //-Hexane Between ERG and ParticipafetM
4-7 Graphical Comparison of Analytical Results for Benzene Between ERG and Participants! 5
4-8 Graphical Comparison of Analytical Results for Cyclohexane Between ERG and
Participants 4-16
4-9 Graphical Comparison of Analytical Results for Toluene Between ERG
and Participants 4-17
4-10 Graphical Comparison of Analytical Results for //-Octane Between ERG and Participants! 8
4-11 Graphical Comparison of Analytical Results for Ethylbenzene Between ERG and
Participants 4-19
4-12 Graphical Comparison of Analytical Results for m-/p-Xylene Between ERG and
Participants 4-20
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
viii
-------
LIST OF FIGURES (Continued)
Page
4-13 Graphical Comparison of Analytical Results for o-Xylene Between ERG and Participants! 1
4-14 Graphical Comparison of Analytical Results for //-Propylbenzene Between ERG and
Participants 4-22
4-15 Graphical Comparison of Analytical Results for 1,3,5-Trimethylbenzene Between
ERG and Participants 4-23
4-16 Graphical Comparison of Analytical Results for 1,2,4-Trimethylbenzene Between
ERG and Participants 4-24
4-17 Graphical Comparison of Analytical Results for n-Decane Between ERG
and Participants 4-25
4-18 Graphical Comparison of Analytical Results for 1,2,3-Trimethylbenzene Between
ERG and Participants 4-26
4-19 Graphical Comparison of Analytical Results for Total NMOC Between ERG and
Participants 4-27
4-20 Graphical Comparison of Analytical Results for Propylene for ERG
and Participants 4-28
4-21 Histogram of Laboratory Results for Propylene 4-29
4-22 Graphical Comparison of Analytical Results for Propylene Between ERG and Participants
(Outliers Excluded) 4-30
4-23 Example of the Bias of the Participants Compared to ERG Results for the
Same Canister 4-32
4-24 Example of the Absolute Value of the Bias for the Participants Compared to ERG Results
for the Same Canister 4-33
4-25 ERG Analysis Range and Participant Analysis Range, Including All Data Points (Ethylene
to Toluene) 4-35
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
IX
-------
LIST OF FIGURES (Continued)
Page
4-26 ERG Analysis Range and Participant Analysis Range, Excluding Outlier
Points for Propylene (Ethylene to Toluene) 4-36
4-27 ERG Analysis Range and Participant Analysis Range
(«-Octane to Trimethylbenzene) 4-37
4-28 Confidence Intervals for ERG and Participant Data Sets, Including Propylene
Outliers (Ethylene to Toluene) 4-39
4-29 Confidence Intervals for ERG and Participant Data Sets, Excluding Propylene
Outliers (Ethylene to Toluene) 4-40
4-30 Confidence Intervals for ERG and Participant Data Sets
(w-Octane to Trimethylbenzene) 4-41
4-31 Stability Study for Canisters Retained at the ERG Laboratory 4-47
4-32 Comparison of ERG Analysis 1 to ERG Analysis 2
1.2.3-Trimethylbenzene 4-58
4-33 Comparison of ERG Analysis 1 to ERG Analysis 2
1.2.4-Trimethylbenzene 4-59
4-34 Comparison of ERG Analysis 1 to ERG Analysis 2
1.3.5-Trimethylbenzene 4-60
4-35 Comparison of ERG Analysis 1 to ERG Analysis 2
1-Butene 4-61
4-36 Comparison of ERG Analysis 1 to ERG Analysis 2
Benzene 4-62
4-37 Comparison of ERG Analysis 1 to ERG Analysis 2
Cyclohexane 4-63
4-38 Comparison of ERG Analysis 1 to ERG Analysis 2
Ethane 4-64
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD X
-------
LIST OF FIGURES (Continued)
Page
4-39 Comparison of ERG Analysis 1 to ERG Analysis 2
Ethylbenzene 4-65
4-40 Comparison of ERG Analysis 1 to ERG Analysis 2
Ethylene 4-66
4-41 Comparison of ERG Analysis 1 to ERG Analysis 2
Propane 4-67
4-42 Comparison of ERG Analysis 1 to ERG Analysis 2
Propylene 4-68
4-43 Comparison of ERG Analysis 1 to ERG Analysis 2
Toluene 4-69
4-44 Comparison of ERG Analysis 1 to ERG Analysis 2
Total NMOC 4-70
4-45 Comparison of ERG Analysis 1 to ERG Analysis 2
m/p-Xylene 4-71
4-46 Comparison of ERG Analysis 1 to ERG Analysis 2
«-Butane 4-72
4-47 Comparison of ERG Analysis 1 to ERG Analysis 2
«-Decane 4-73
4-48 Comparison of ERG Analysis 1 to ERG Analysis 2
«-Hexane 4-74
4-49 Comparison of ERG Analysis 1 to ERG Analysis 2
«-Octane 4-75
4-50 Comparison of ERG Analysis 1 to ERG Analysis 2
«-Propylbenzene 4-76
4-51 Comparison of ERG Analysis 1 to ERG Analysis 2
o-Xylene 4-77
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
XI
-------
LIST OF FIGURES (Continued)
Page
4-52 Average Percent Bias (Considering Bias Positive or Negative) per
Organization for All Compounds 4-85
4-53 Average Percent Bias Per PAMS Site for All Compounds 4-88
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
Xll
-------
EXECUTIVE SUMMARY
The Photochemical Assessment Monitoring Stations (PAMS) Quality Assurance (QA)
Work Group requested Eastern Research Group (ERG) to work in conjunction with the States to
conduct a performance evaluation program using audit canisters in order to assist the PAMS in
assessing the quality of their analytical systems, and to assist the QA Work Group to evaluate the
overall precision and accuracy of the National PAMS program. The two primary requirements
for the proposed performance evaluation program were:
• That all ERG laboratory procedures be thoroughly documented so that canister
concentrations of analytes provided to the participating States would be accurately
known; and
• That feedback to the States reporting analytical results be rapid enough to allow
them to repeat the analysis should they desire to resolve issues relating to analysis.
Eastern Research Group accordingly designed and performed a program consisting of the
following components:
• Selecting and cleaning of canisters to be used for the program;
• Blanking canisters using Compendium Method TO-12;1
• Blanking canisters using gas chromatography/mass spectrometry with flame
ionization detection (GC/MS-FID) for the analytes to be spiked;
• Preparing performance evaluation standards;
• Analyzing each performance evaluation standard by GC/MS-FID after preparation;
• Retaining at least three performance evaluation standard canisters in the ERG
laboratories for a stability check;
• Checking final canister pressure before shipment;
• Shipping performance evaluation canisters with Instruction Sheet and Chain of
Custody to the participants;
• Creating a database (Microsoft® Access) to track all information;
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xiii
-------
• Receiving duplicate results from participants;
• Responding with analytical results within 48 hours; and
• Compiling results and preparing a project final report.
Analytical results were received from the majority of the participants within two weeks of
canister shipment. Several PAMS submitted analytical results from more than one analytical
system when PAMS analyses were performed on more than one analytical system. Although
participants were allowed to keep the performance evaluation canisters for a period of two weeks
beyond the reporting of analytical results in order to repeat the analysis, if necessary, only one
participant reported a repeated analysis.
Most of the reported results for the spiked compounds were within ±2 standard deviations
of the ERG analytical results. There was only one set of points, duplicate points for propylene
submitted by one participant, more than an order of magnitude out of the analytical range (± 2
standard deviations from the mean). These data points were rejected as outliers. The target
compounds with the widest range of results were the very volatile compounds (C2 hydrocarbons,
negative bias of 10-15%) and the least volatile compounds (C3-alkylbenzenes, positive bias
ranging up to 10%). Decane showed the lowest bias, <1%. The group of compounds showing
the lowest bias were the compounds eluting in close chromatographic proximity to the standard,
propane. Compounds that were not reported (i.e., missed) were not treated as zeros in statistical
calculations to avoid skewing data to low values.
In addition to a number of PAMS target compounds, four non-PAMS target compounds
were also spiked into the performance evaluation canisters. These additional compounds
provided a challenge to the compound identification procedures of several participants.
The most common compound identification errors were as follows:
• Methyl /-butyl ether was misidentified as 2,3-dimethylbutane or 2-methylpentane;
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xiv
-------
• 1-Hexene was misidentified as 2-methyl-l-pentene; and
• 1,1,1-Trichloroethane was misidentified as 2,4-dimethylpentane.
The use of non-PAMS target compounds in the standard mixture provided a useful evaluation of
the identification and data validation procedures used by the participants.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xv
-------
1.0 INTRODUCTION
Ambient air quality surveillance regulations in Title 40 Part 58 of the Code of Federal
Regulations (40 CFR Part 58)1 include several provisions for enhanced monitoring of ozone (03)
and its precursors including oxides of nitrogen (NOx), volatile organic compounds (VOCs),
including carbonyl compounds, and meteorological parameters. These revisions to the ambient air
quality surveillance regulations require the States to establish Photochemical Assessment
Monitoring Stations (PAMS). PAMS will be part of their existing State Implementation Plan
(SIP) monitoring network in ozone nonattainment areas, and monitoring data must be reported to
the U. S. Environmental Protection Agency (EPA). The additional ambient air pollutant and
meteorological data are required because the National Ambient Air Quality Standards (NAAQS)
have not been achieved for ozone. Additionally, a more comprehensive air quality database for
ozone and its precursors is needed. The authority of the EPA for proposing the enhanced
monitoring regulations stems from Title I, Section 182, of the Clean Air Act Amendments
(CAAA) of 1990.
The monitoring stations for ozone and its precursors are identified as PAMS, with
different types and frequencies of monitoring required on the basis of the population of the
Metropolitan Statistical Area (MSA) or the Consolidated Metropolitan Statistical Area (CMSA).
To assist the States in PAMS implementation, Technical Assistance Documents2'3 have been
prepared by EPA to address the siting of these monitoring stations and the sampling and analysis
of ozone precursors.
The VOCs to be measured in the PAMS network are gaseous nonmethane organic
compounds with a vapor pressure greater than 10"2 kilopascals, with a number of carbon atoms
ranging from C2 to C12. These ozone precursors are saturated, unsaturated, cyclic, and/or
aromatic hydrocarbons, as shown in Table 1-1. Some or all of these ozone precursors are
expected to be found in ambient air. This list of compounds is listed in the Technical Assistance
Document for Sampling and Analysis of Ozone Precursors.3
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xvi
-------
Table 1-1
Ozone Precursors
Ozone Precursors
ethylene
3-methylhexane
acetylene
2,2,4-trimethylpentane
ethane
//-heptane
propylene
methylcyclohexane
propane
2,3,4-trimethylpentane
isobutane
toluene
1-butene
2-methylheptane
//-butane
3-methylheptane
/ra//.s-2-butene
«-octane
c'/.s-2-butene
ethylbenzene
isopentane
/??-//;-xylene
1-pentene
styrene
//-pentane
o-xylene
isoprene
//-nonane
/ra//.s-2-pentene
isopropylbenzene
c/.s-2-pentene
//-propylbenzene
2,2-dimethylbutane
w-ethyltoluene
cyclopentane
/;-ethyltoluene
2,3-dimethylbutane
1,3,5 -trimethy lb enzene
2-methylpentane
o-ethyltoluene
3-methylpentane
1,2,4-trimethylbenzene
2-methyl-1 -pentene
«-decane
//-hexane
1,2,3-trimethylbenzene
methylcyclopentane
w-di ethylbenzene
2,4-dimethylpentane
p-$\ ethylbenzene
benzene
//-undecane
cyclohexane
2-methylhexane
Total Nonmethane Organic Cor
2,3-dimethylpentane
Note: m- and /^-Xylene are listed together because of chromatographic coelution.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xvii
-------
To identify the compounds and determine their concentrations, a sample of the ambient air
must be collected, and the organic constituents concentrated and analyzed. The sample is
typically cryogenically collected on a solid sorbent (preconcentration of organic compounds) and
desorbed directly into an automated gas chromatograph for qualitative and quantitative analysis.
Because there are a number of organic constituents of the ambient air, these constituents must be
separated before analysis. Gas chromatography, using a capillary analytical column provides the
best possible separation of the constituents prior to analysis. The quantitative analytical results
are obtained by FID with gas chromatography. GC/FID relies upon chromatographic retention
times for the identification of compounds.
The States are responsible for setting up and operating their own PAMS. A PAMS is
successfully operated by collecting scheduled samples according to the regulatory requirements,
analyzing samples according to accepted methodology, and reporting data to the appropriate
regulatory agencies. Each state has an internal program to provide quality control and quality
assurance. An external evaluation of the effectiveness of the PAMS analytical process (as
assessed by an external performance evaluation sample) provides extremely valuable information
regarding the successful operation of the PAMS. This external program was a performance
evaluation rather than an audit, conducted primarily for the benefit of the monitoring
organizations to allow the PAMS to evaluate how well they were performing their analyses. This
report presents the specifications and procedures used in preparing, certifying, distributing, and
managing a multiple-site performance evaluation program, with humidified performance
evaluation samples prepared in 6-liter SUMMA®-polished stainless steel canisters.
1.1 PAMS Quality Assurance Work Group
A PAMS Quality Assurance Work Group, formed by the U. S. EPA, included
representatives from EPA Office of Air Quality Planning and Standards (OAQPS), EPA National
Exposure Research Laboratory (NERL), EPA Region II, EPA Region VI, Texas and California.
The group was established to perform the following initial functions:
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xviii
-------
Develop Quality Assurance and Quality Control procedures to be applied
uniformly and inexpensively to raise the confidence levels of the field staff;
Explore the development of Quality Assurance/Quality Control software (including
error-checking, source profiles, and multi-variate techniques) for on-site
integration at PAMS data collection locations (i.e., "real-time" Quality Assurance
evaluations); and
Provide information/reports to the Steering Committee and other work groups on
the precision and accuracy of the PAMS data.
The Work Group tasked Eastern Research Group, Inc. (ERG) to supply performance
evaluation samples to the PAMS. Performance evaluation samples supplied by an external
laboratory are quality assurance tools that can be used to evaluate the precision and accuracy of
an analysis. ERG was required to supply timely performance sample result information to the
States in order to provide feedback that the participants could use to immediately improve their
operations. The secondary purpose of this study was to assist the QA Work Group in evaluating
the performance of the PAMS program as a whole and to help the EPA regions identify sites in
need of additional assistance. In order to operate a successful performance evaluation program, it
was essential to have accurate and precise measurement of the compounds in the performance
evaluation samples which could be supplied to all of the participants at or very near the same time.
In the event that problems were encountered, the PAMS had the opportunity to correct the
problems and re-analyze the performance evaluation standard. In this project final report, the
participants have the opportunity to evaluate their performance relative to other PAMS, although
no specific identifications of the participants are made.
1.2 Performance Evaluation Canister Preparation Procedures
Humidified performance evaluation samples were prepared in 6-liter SUMMA®-polished
stainless steel canisters and contained the compounds listed in Table 1-2. Twenty-two of the
compounds listed in Table 1-2 are ozone precursors (see Table 1-1). Two additional
Table 1-2
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xix
-------
Components of Certified Gas Cylinders
Cylinder 1
Cylinder 2
benzene
benzene
//-butane
1,3-butadiene
1-butene
cyclohexane
decane
ethylbenzene
ethane
ethylene
ethylbenzene
propylene
//-hexane
/f/V-butyl methyl ether
1-hexene
toluene
«-octane
1,1,1 -trichloroethane
propane
1,2,3 -trimethylbenzene
//-propylbenzene
1,2,4-trimethylbenzene
toluene
1,3,5 -trimethy lb enzene
w-xylene
o-xylene
/^-xylene
compounds included in the gas mixture are not ozone precursors. The performance evaluation
samples were prepared in a single batch of forty canisters. (An additional batch of ten PE
canisters was prepared when some of the original group of 40 were found to have leaked before
shipment.) Prior to preparation of the performance evaluation samples, each canister was cleaned
and certified for total nonmethane organic compound (NMOC) content to less than 10 ppbC.
Each canister was also analyzed by GC/MS-FID to verify cleanliness on a compound-specific
basis. The criterion for acceptable cleanliness of a canister was 1.0 ppbC or the detection limit for
the specific compound, whichever is greater.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xx
-------
When the cleanliness of the canisters had been verified, performance evaluation samples
were prepared from high pressure multiple-component gas cylinders obtained from a specialty gas
manufacturer. The cylinders contained the compounds shown in Table 1-2, mostly hydrocarbons
with carbon numbers from C2 through C10 (analytical accuracy of ±5%), with some non-
hydrocarbon volatile organic compounds also included. The canisters were prepared at a pressure
of approximately 35 psig with a relative humidity of about 70%. The relative humidity was
calculated based on the ideal gas law. The performance evaluation samples tested only the
analytical portion of the PAMS procedures. To eliminate potential sources of variability or error
in the analysis, the canisters were analyzed directly by the analytical system without the use of any
slip-stream or pressure reduction devices.
Prior to distribution, each canister was analyzed by GC/MS-FID to verify the final
composition and concentration. The concentration in ppbC of each component was referenced to
a NIST-certified propane standard. The pressure of the canisters was also checked prior to
shipment. Four additional canisters (baseline samples) were prepared and held by ERG for
reanalysis and additional information on stability at the end of the program.
A database of all of the participants, the site locations, the bias of the performance
evaluation results, and other relevant comments and information was created and maintained
during the program using Microsoft Access®. A Chain of Custody form was created to track
samples and collect other necessary information for the database.
Participants were requested to analyze the canister in replicate and, within two weeks, fax
both sets of results to ERG. ERG compared the analytical results to the first set of analytical
results generated by ERG and faxed a report to the PAMS within 48 hours. The participating
PAMS could then reanalyze the canister, if desired, to resolve any performance issues, and return
the canister to the ERG laboratory within one month. Once the canisters were returned to the
ERG laboratory, a subset of the canisters was selected for reanalysis based on the results from the
PAMS. The four reserved baseline canisters were also reanalyzed.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xxi
-------
To ensure the success of the performance evaluation program, close coordination among
EPA, ERG, and the site contacts was required, with timely reporting of results to the PAMS sites
and reporting of technical progress and problems to the EPA.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
1-xxii
-------
2.0 CONCLUSIONS AND RECOMMENDATIONS
On the basis of ERG laboratory experience in performing and managing this program and
the evaluation of the analytical results, the following conclusions are drawn:
• The compounds spiked into the performance evaluation canisters (Table 1-2) are
stable in the canisters over a period of eight weeks: analysis at the beginning and at
the end of the period shows minimal compound loss, both for canisters retained in
the laboratory and for canisters shipped to the PAMS sites and returned to ERG.
• With the exception of compounds not reported (which were not included in
statistical calculations), only two reported values were rejected statistically as
outliers.
• The overall PAMS site average absolute value percent bias (considering only the
magnitude of the bias, not the sign) was 11.31.
• The overall PAMS site average percent bias (considering that the bias can be
positive or negative) was -0.25.
• After shipping the canisters to PAMS sites, where two or more analyses were
performed, and returning the canisters to ERG, a second ERG analysis for sixteen
canisters agreed with the initial ERG analysis generally within 10 percent.
• Providing rapid feedback to the PAMS is a crucial component of a successful
program. With rapid feedback, the information provided by the evaluation can be
applied immediately to solve PAMS analytical problems;
• For the highest reproducibility in preparation of performance evaluation standards,
a certified cylinder (certified by the gas vendor) of gaseous standards should be
used with dynamic flow dilution;
• Strong Quality Assurance and Quality Control procedures are essential in
providing appropriate documentation for tracking the complete history of each
performance evaluation canister;
• The majority of the participants were prompt, accurate, and precise in reporting
their analytical data.
The following recommendations are made:
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
2-xxiii
-------
• To provide a good external check on laboratory procedures and build confidence
in the program both internally and externally, performance evaluation samples
should be provided on a regular schedule. Regularly-scheduled performance
evaluation samples must also provide rapid feedback of results.
• To provide a challenge to the PAMS in compound identification procedures, the
compounds to be spiked should be selected carefully. Including the complete
PAMS target list should in general be avoided and including compounds
extraneous to the PAMS target list is desirable.
• To provide a better test of the ability to perform quantitative measurements,
standards should be prepared at different concentrations to produce a range of
values. Some compounds should be spiked at concentrations close to actual
ambient measurements and some compounds spiked at higher concentrations.
Blending two certified cylinder standards at different concentrations would
produce a range of values in the final performance evaluation standard.
• To get the best performance evaluation program possible, advance planning and an
early start are necessary. Some delays are inevitable—funding is delayed, cylinder
purchase may take longer than quoted by the suppliers, instruments and equipment
need repair, etc. However, planning well ahead allows the program to compensate
for delays.
All of these factors should be considered in planning a program for preparation and
distribution of the performance evaluation standards.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
2-xxiv
-------
3.0 LABORATORY PROCEDURES
The ERG laboratory procedures for preparation of performance evaluation standards
consisted of four major parts:
• Cleaning canisters prior to preparation of audit samples;
• Verifying the cleanliness of the audit canisters by NMOC and by GC/MS-FID;
• Preparing diluted samples from high pressure multiple-component gas cylinders;
and
• Verifying the final concentration and content of the audit canisters.
Procedures for each part of the operation are described.
3.1 Cleaning of SIJMMA®-Passivated Stainless Steel Canisters
The canister cleaning procedure used for the performance evaluation standard canisters
corresponds to the procedure described in the Technical Assistance Document.3 The canisters
are purged with cleaned humidified air and then subjected to high vacuum to ensure that the
canister interior surfaces are free of contaminants. The canister must meet an NMOC cleanliness
criterion of <10 ppbC (as measured by Method TO-12) to minimize the potential for carryover of
organic pollutants from one sample to the next.
Complete and detailed records were maintained for each canister to track the cleaning
procedure and each subsequent analysis to demonstrate cleanliness.
3.2 Verification of Cleanliness of the SIJMMA®-Polished Canisters
The cleanliness of each canister was evaluated by using Compendium Method TO-12
procedures. The Method TO-12 blanking procedure requires a simple preconcentration
procedure with subsequent analysis by direct flame ionization detection (PDFID) to provide an
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxv
-------
accurate and precise measurement of total NMOC concentration, expressed as ppbC. When all
cleaned canisters had met a Method TO-12 cleanliness criterion of NMOC<10 ppbC, GC/MS-
FID was used to verify cleanliness on a compound-specific basis.
By convention, concentrations of NMOC are reported as parts per million carbon (ppmC).
For this program, the criterion for cleanliness was 10 parts per billion carbon (ppbC). All
canisters met this criterion.
3.3 Verification of Canister Cleanliness on a Compound-Specific Basis (GC/MS-FID
Analysis)
Since a background level of one or more of the analytes would significantly bias the results
of the performance audit, the cleaned blanked canisters were analyzed for speciated ozone
precursors by GC/MS-FID.
The GC/MS-FID analytical system incorporates a sample concentration trap, a capillary
gas chromatograph, a capillary column, a post-column splitter, and both a mass selective detector
(MSD) and a FID. The FID is used to quantify the compounds of interest; the MSD is used to
identify the compounds of interest. Moisture is removed from the analytical system using a
permeable membrane drying device. The samples are cryogenically concentrated by using a trap
consisting of chromatographic-grade stainless steel tubing packed with commercially available
60/80 mesh deactivated glass beads maintained at -185° C during sample concentration. The
concentrated VOCs are thermally desorbed to revolatilize them for transfer to the analytical
column for separation. All of the cleaned canisters met individual compound cleanliness criteria.
Analytical conditions are shown in Table 3-1.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxvi
-------
Table 3-1
Analytical Conditions for GC/MS-FID Analysis of Ozone Precursor
Compounds
Analytical System
GC
Mass Selective Detector
Sample Concentration System
Concentrator
Chromatographic Conditions
Column
Carrier gas
Carrier Flow
GC Program
MS Conditions
Scan cycle
Electron voltage
Hewlett-Packard 5890 Series II
Hewlett-Packard 5971 MSD
NuTech, 16 valve autosampler
J&W DB-1, 60 m x 0.32 mm, 1 |i film thickness
Helium
1 mL/min
-60 °C for 5 min, then 6°C/min to 150°C,
then 180°C for 8 min for a total run time of 50 min
1 sec/scan, 5 min solvent delay
70 eV, nominal
3.4 Preparation of Performance Evaluation Samples
Performance evaluation canister samples at a specific concentration (a concentration range
of 10-50 ppbC) were prepared by dilution from cylinders containing the compounds shown in
Table 1-2. The same procedure was used to prepare both the PAMS performance evaluation
samples and GC/MS calibration standards.
Standards and performance evaluation samples were blended and diluted to the desired
concentration with humidified air. A concentration of 25-50 ppbC per compound was targeted.
The performance evaluation samples were prepared at a pressure slightly above 30 psig and a
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxvii
-------
relative humidity of about 70%. The canisters were pressurized to slightly above 30 psig since
they would be opened for analysis in the ERG laboratory before being shipped at a pressure of 30
psig. The relative humidity was a calculated value based on the ideal gas law. To eliminate
sources of error, the performance evaluation canisters were prepared to be ready for analysis, with
no further dilution or preparation required. The canisters were shipped with a Chain of Custody
(Figure 3-1) and a Letter of Instruction (Figure 3-2) to ensure that the history of the canister
could be traced throughout the entire preparation and analysis process.
3.5 GC/MS-FID Analysis of Performance Evaluation Samples hv ERG
Using the analytical procedure described above, each performance evaluation canister was
analyzed once by GC/MS-FID to verify the composition and concentration prior to shipment.
The ppbC concentration of each component in the canister was referenced to a NIST-traceable
propane standard. Four additional performance evaluation standard canisters were retained in the
ERG laboratory as baseline samples and held for reanalysis and for additional stability information
at the end of the program. The interval between the two ERG analyses was approximately eight
weeks: first analysis, 8/8/96; second analysis, 10/3/96.
The pressure of the performance evaluation standard canisters was checked immediately
before shipment to ensure that participants would receive a canister at 30 psig. Five of the
original batch of 40 canisters were found to have leaked. These canisters were rejected for
shipment, and a second batch of ten performance evaluation canisters was prepared following the
same procedures as were followed for the original batch of 40 canisters. The second batch of ten
canisters was treated as a second data set.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxviii
-------
VERG
EASTERN RESEARCH GROUP. INC.
Audit Data Sheet
Lab ID #
Can Serial #
Canister Cleaned:
Date:
Blank NMOC Analysis:
Date:
Analvst:
Blank GC/MSD Analysis:
Date:
Analvst:
Filename:
Standard Prep:
Date:
Analvst:
Humiditv: Pressure:
GC/MSD Analysis:
Date:
Analvst:
Filename:
Shipment:
Sent
Date:
Operator:
Fed Ex #: Canister P (Dsia):
Received
Date:
Operator:
Fed Ex #: Canister P (Dsia):
Lab Analysis:
First Analysis:
Date:
Analvst:
Filename:
Second Analysis:
Date:
Analvst:
Filename:
Note: If more than two analyses are done on the samples, please attach a page with explanation of results.
Shipment:
Sent
Date:
Operator:
Fed Ex #: Canister P (Dsia):
Received
Date:
Operator:
Fed Ex #: Canister P (Dsia):
Second GC/MSD Analysis:
Date:
Analvst:
Filename:
White: Original Copy Canary: Lab Copy Pink: Auditing Lab Copy Goldenrod: EPA Copy
Figure 3-1. Chain of Custody
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD 3 "XXIX
-------
feERG
EASTERN RESEARCH GROUP. INC
PAMS QA Audit Sample Instructions
This canister contains a PAMS Performance Audit standard to be analyzed in your laboratory.
Instructions for preparing the documentation are below.
1) Fill in the Received date. Operator. Fed Ex # and Canister Pressure fpsiu) under "Shipment" on the Audit
2) Run each sample as received twice. (NO FURTHER PREPARATION IS REQUIRED). If the
samples are analyzed more than two times, please include these results and explanations on an attached
3) Fill in Data Results Page, complete with analytical conditions and results (in ppbC).
4) Fill in the Date. Analyst, and Filename on the Audit Data Sheet under the "Lab Analysis" for the First
and Second Analysis. If more than two analyses were done on the sample, please provide this
information on an attached page.
5) Fax the Data Results Page and a copy of the Audit Data Sheet to:
6) the turn around time for sample results is two weeks (2 weeks) from date of receipt. Fax the data back by
7) Fill in the Sent-Date. Operator. Fed Ex #. and Canister Pressure fpsiu) under "Shipment" on the Audit
Data Sheet.
sheet.
Julie Swift
Fax#: (919) 461-1579
Data Sheet.
8) Send the canister back no later than
using the shipping information supplied.
If you have any questions, please contact either:
Julie Swift
Tel. (919) 461-1245
or
Joan Bursey
Tel. (919) 461-1334
Ship all canister, chain-of-custody, and hardcopy results to:
Shipping Address
Julie Swift
ERG
900 Perimeter Park, Dock C
Morrisville, NC 27560
Figure 3-2. Letter of Instruction
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxx
-------
3.6 Analysis hv PAMS
After completion of the GC/MS-FID analysis of the performance evaluation samples by ERG,
the canisters were shipped, with custody documentation, to the PAMS. The PAMS in general
analyzed the canisters in replicate within two weeks of receipt according to their analytical
procedures, and both sets of results were faxed to ERG. The ERG laboratory compared the
results to the analytical results obtained for that canister prior to shipment and a report was faxed
to the site contact at the PAMS. The PAMS had the option of repeating the analysis of the
performance evaluation canister after data were received from ERG. However, only one PAMS
exercised the option to repeat the analyses.
3.7 Preparation of a Database of PAMS Results
A database of the participants, site locations, result bias, and other relevant comments and
information was created using MicroSoft® Access. The custody documentation and the form
used in sample tracking contributed information to the database. The information generated by
the PAMS (the replicate results for the analysis of the performance evaluation canister) was
entered into the database and verified to ensure that no transcription errors were made. The
database is available to EPA at the conclusion of the study.
3.8 Repeated Analysis of Performance Evaluation Samples
Upon completion of analysis at the PAMS the canisters were shipped back to ERG. When the
canisters were received at the ERG, some of the total number of canisters returned
(approximately 30%) were selected for re-analysis based on the PAMS analytical results. The
results from the repeated analysis were added to the database. The four baseline canisters
retained at ERG were also re-analyzed.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxxi
-------
3.9 Quality Control Information
The laboratory quality control procedures used at ERG are summarized in Table 3-2.
The minimum requirement of a laboratory quality control program is an initial demonstration
of laboratory capability and an ongoing analysis of internal audit standards to evaluate and
document data quality. Laboratory records are used to document the quality of the data
generated. Ongoing data quality checks are compared with established performance criteria to
determine if the results of analyses meet performance requirements. When analytical results of
audit samples indicate atypical performance for the analytical method, a quality control check
standard must be analyzed to confirm that the measurements were performed in an in-control
mode of operation for the analytical instrumentation.
Before processing any samples, a blank sample is analyzed to demonstrate that interferences
from the analytical system are under control. The blank samples for this program consisted of a
clean canister of clean humidified air which, under analysis, demonstrated that no analytes were
observed above the detection limit.
The tune of the GC/MS system to meet />bromofluorobenzene specifications was verified
every 12 hours. An initial multipoint calibration was performed to meet laboratory acceptance
criteria. A daily calibration check standard was analyzed on all analytical instruments to verify
that the chromatographic systems were operating properly. The chromatographic profile was
examined to ensure that adequate chromatographic properties and resolution were being
maintained by the analytical system. Consistency between the response of the analytical system to
the calibration check standard was evaluated on a day-to-day basis to ensure that instrument
sensitivity was maintained. Response factors generated from analysis of the daily calibration
check standard were compared to the mean response factors generated by the analysis of the
multipoint calibration to verify the stability of the GC/MS system. Standard signal levels were
monitored on a run-to-run basis to demonstrate the stability of the analytical system.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxxii
-------
Table 3-2
Summary of PAMS Performance Evaluation Quality Control Procedures
Quality Control
Check
Frequency
Acceptance Criteria
Corrective Action
System Blank
Daily, following
1.0 ppbC or Method
1)
Repeat analysis.
Analysis
calibration check,
Detection Limit,
2)
Check system for
prior to analysis
whichever is greater
leaks
for target compounds
Five-point calibration
Monthly
Correlation
1)
Repeat individual
for propane (five
coefficient (r)
sample analysis
concentrations) for
> 0.995
2)
Repeat linearity
both MS and FID,
RSD of response
check
bracketing the
factors < 30%
3)
Prepare new
expected sample
calibration
concentration
standards and
repeat
Calibration check
Daily (every 12
Response within
1)
Repeat check
using mid-point of
hours) on the days of
±30% difference of
2)
Repeat calibration
calibration curve
sample analysis
calibration curve
curve
slope
Canister cleaning
All canisters prior to
<10 ppbC total
1)
Reclean canister
certification
sample preparation
and repeat
analysis
On the FID and TO-12 analytical systems, a daily check standard (NIST-certified
propane) is analyzed to verify the stability of the analytical system. A multipoint calibration was
performed monthly and checked daily to verify that the calibration is still valid and meets
acceptance criteria.
Quality control procedures applied to this program include:
Adequately-trained personnel;
Supervision to ensure that the Quality Assurance Project Plan was followed;
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxxiii
-------
Documentation of experimental results in dedicated project notebooks in a manner
that will facilitate reconstruction of project activities and verification of data
accuracy;
Review of the progress of the work by senior technical reviewers and independent
Quality Assurance staff to assess the effectiveness of the internal quality control
program;
Use of traceable standards and accurately prepared test substances;
Use of equipment that is calibrated and tested according to laboratory Standard
Operating Procedures;
Documenting routine and non-routine maintenance of analytical instrumentation to
allow an assessment of reliability;
Following sample and data labeling procedures to help prevent sample and data
mix-ups; and
Testing quality control samples to assess the appropriateness of the analytical
system and sample preparation instrumentation.
3.9.1 Precision
The precision of the preparation of the performance evaluation canisters was determined
from the GC/MS-FID analysis of each prepared standard. The precision objective for the entire
batch of performance evaluation canisters was a variation of less than 10% relative standard
deviation. This objective was achieved for all of the compounds except ethane (precision, as
measured by relative standard deviation, was 21.19%). Ethane also failed to meet the 10%
relative standard deviation in the second batch of canisters (precision, as measured by relative
standard deviation, was 13.54%). This variability in the measurement resulted in a broadening of
the range for analyses of ethane.
The standard deviation (SD) of the ERG measured values for the performance evaluation
samples was determined for each analyte using the following equation:
SD =
y (x. - x )
f v i avg7
n - 1
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxxiv
-------
where:
number of determinations
mean of n determinations
ith value
RSD=C V =_SD x 100
V
avg
where:
RSD = relative standard deviation
CV = coefficient of variation
The standard deviations of the measured values for the PAMS were calculated in the same
way. The measured value for each compound in each PAMS performance evaluation canister was
expressed as ERG measured value ± 25, where 5 = SD.
3.9.2 Bias
The bias (B) of each of the replicate results submitted by the PAMS was calculated for
each analyte using the following equation:
n/„. Measured Value - True Value
%Bias = x 100
True Value
X
avg
X;
where:
Measured Value = Value measured by the PAMS
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxxv
-------
True Value = Value measured by ERG for each compound in each
canister prior to shipment of the canister.
The percent bias was determined and reported individually for each of the replicate values
reported by the PAMS.
The bias between the theoretical value for each analyte in the canister and the ERG
measured value was also calculated.
3.9.3 Completeness
The quality assurance objective for completeness was 100 percent, with no samples
invalidated because they were damaged or lost. This objective was achieved with respect to
damage or loss of samples. All canisters shipped were received by the PAMS participants in
acceptable condition for analysis. The projected data set was not complete, however, since
analytical results were not reported from one laboratory that received a canister.
cah\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
3-xxxvi
-------
4.0 RESULTS AND DISCUSSION
The performance evaluation canisters were prepared and analyzed by ERG in two batches
prior to shipment to the participants. An initial batch of 40 canisters was prepared and analyzed,
with the results for this set of data presented in Tables 4-1 (Theoretical Amounts) and 4-2
(Analyzed Amounts).
4.1 Analysis Results
Cylinder contents were not analyzed directly (without dilution) prior to preparation of the
diluted canisters because the undiluted cylinder gas concentration was too high for the analytical
system.
With isolated exceptions, ERG analyzed values for the canisters show a low bias relative
to the theoretical/calculated values, indicating that the concentrations of the compounds in the
cylinders have become lower over time. The low bias is greatest in magnitude for ethane and for
the least volatile compounds (decane and later-eluting compounds). Ethane and the other
compounds less volatile than decane also showed the largest range (analytical mean ± 2 standard
deviations) in the analysis. The wide range for ethane is very different from the narrow range and
low bias shown for ethylene, also a C2-hydrocarbon. Theoretical and analytical results for all
performance evaluation standards prepared, both the initial batch of 40 and the second batch of
ten, are shown in Appendix E.
The consistent negative bias for the late-eluting compounds may represent cylinder loss or
may represent a reproducible laboratory error such as an inadequately heated transfer line. Since
the bias was very reproducible, a cause was not pursued.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xxxvii
-------
Table 4-1
Original Batch of Forty Canisters: Theoretical/Calculated Amounts
Compound
Theoretical
Value
(Mean)1
Range2
Median3
90%
Confidence
Limits4
90th
Percentile5
benzene
77.42
76.82-78.02
77.38
0.08
77.69
n-butane
27.82
27.60-28.04
27.81
0.03
27.92
1-butene
27.82
27.60-28.04
27.81
0.03
27.92
cyclohexane
35.59
35.15-36.03
35.55
0.06
35.69
/?-dccanc
69.55
69.01-70.09
69.51
0.07
69.79
ethane
14.47
14.35-14.59
14.46
0.01
14.52
ethylbenzene
99.91
99.11-100.71
99.86
0.10
100.26
ethylene
11.92
11.82-12.02
11.91
0.01
11.96
/?-hcxanc
41.73
41.41-42.05
41.71
0.04
41.87
n-octane
55.64
55.20-56.08
55.61
0.06
55.83
propane
20.86
20.70-21.02
20.85
0.02
20.94
/7-propylbcnzene
62.60
62.12-63.08
62.56
0.06
62.81
propylene
17.66
17.52-17.80
17.65
0.02
17.72
toluene
90.39
89.69-91.09
90.34
0.09
90.70
1,2,3 -trimethylbenzene
52.82
52.40-53.24
52.79
0.05
53.00
1,2,4-trimethylbenzene
51.58
51.18-51.98
51.55
0.05
51.76
1,3,5 -trimethylbenzene
49.71
49.33-50.09
49.68
0.05
49.88
m-lp-nylene
87.71
87.03-88.39
87.65
0.09
88.00
o-xylene
51.32
50.92-51.72
51.29
0.05
51.49
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xxxviii
-------
Table 4-1
Continued
1 Value calculated from nominal cylinder values and dilution factors.
2 Range = (mean ERG theoretical value for benzene in 40 canisters ) ± 2 standard deviations,
for benzene, the mean of the ERG theoretical values for all 40 canisters is 77.42, with a
standard deviation of 0.30. Range therefore equals 2(0.30)-77.42 + 2(0.30) = 76.82-
78.02.
3 Median of 40 theoretical values.
4 A confidence interval is a range on either side of a sample mean used to estimate the
population mean with a specified probability. For small sample size (usually less than 100),
confidence interval is calculated as follows:
Confidence Interval = MEANs ± tf| a2n x
/ \
STD
f-
where:
Confidence Interval = the range for the estimate of the population mean for the
specified probability, 1-a (for a 95% probability level,
1-c£=0.95 and therefore a=0.05)
MEANS = the sample mean
t(W2,n-i) = the two-sided t-value for the specified probability (1-a) and
degrees of freedom (n-1)
n = the sample size
5 Percentile establishes a threshold of acceptance, i.e., the values that would be above the 90th
percentile.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xxxix
-------
Table 4-2
Original Batch of Forty Canisters: ERG Analyzed Values
Compound
Analyzed
Value
(Mean)1
Range2
Median3
90%
Confidence
Limits4
90th
Percentile5
% Bias
(Analyzed
vs.
Theoretical)
(Mean)
benzene
69.80
66.66-72.94
70.13
0.40
71.33
-9.84
«-butane
27.97
26.35-29.59
27.89
0.21
28.94
0.54
1 -butene
25.82
22.82-28.82
26.03
0.38
27.42
-7.19
cyclohexane
32.38
30.94-33.82
32.52
0.18
33.22
-9.02
«-decane
53.41
43.33-63.49
54.92
1.29
57.65
-23.21
ethane
10.10
5.82-14.38
9.58
0.55
13.29
-30.20
ethylbenzene
83.56
8.48-88.64
84.28
0.65
85.68
-16.36
ethylene
11.04
9.28-12.80
11.15
0.22
11.52
-7.38
«-hexane
9.60
37.38-41.82
39.76
0.29
40.45
-5.10
«-octane
9.33
46.93-51.73
49.69
0.31
50.39
-11.34
propane
21.55
20.49-22.61
21.53
0.14
22.22
3.31
«-propylbenzene
49.85
44.17-55.53
50.61
0.73
51.71
-20.37
propylene
16.65
15.73-17.57
16.65
0.12
17.23
-5.14
toluene
5.80
72.26-79.34
76.26
0.46
77.55
-16.14
1,2,3-trimethylbenzene
38.17
30.91-45.43
38.97
0.93
40.31
-27.73
1,2,4-trimethylbenzene
38.76
32.46-45.06
39.49
0.81
40.71
-24.85
1,3,5-trimethylbenzene
39.80
34.42-45.18
40.48
0.69
41.37
-19.94
m-lp-xylene
74.66
69.08-80.24
75.24
0.72
77.47
-14.88
o-xylene
39.36
36.52-42.20
39.78
0.36
40.28
-23.30
Total NMOC 861.42 804.20- 867.51 7.35 885.52
918.64
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xl
-------
Table 4-2
Continued
1 Value calculated from nominal cylinder values and dilution factors.
2 Range = (mean ERG theoretical value for benzene in 40 canisters ) ± 2 standard deviations,
for benzene, the mean of the ERG theoretical values for all 40 canisters is 77.42, with a
standard deviation of 0.30. Range therefore equals 2(0.30)-77.42 + 2(0.30) = 76.82-
78.02.
3 Median of 40 theoretical values.
4 A confidence interval is a range on either side of a sample mean used to estimate the
population mean with a specified probability. For small sample size (usually less than 100),
confidence interval is calculated as follows:
Confidence Interval = MEANs ± tf| r(2n x
I \
STD
f-
where:
Confidence Interval = the range for the estimate of the population mean for the
specified probability, 1-a (for a 95% probability level,
1-c£=0.95 and therefore a=0.05)
MEANS = the sample mean
t(W2,n-i) = the two-sided t-value for the specified probability (1-a) and
degrees of freedom (n-1)
n = the sample size
5 Percentile establishes a threshold of acceptance, i.e., the values that would be above the 90th
percentile.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xli
-------
After the performance evaluation canisters were analyzed by ERG, the canisters were
given a final pressure check and shipped to the PAMS participants together with a Chain of
Custody form and an Instruction sheet. When the pressure of the canisters was checked prior to
shipment, five canisters had leaked. While the leaking canisters were being repaired, the standards
were re-prepared and re-analyzed in other leak-free canisters, and the performance evaluation
samples were shipped to arrive at the participating laboratories as soon as possible after ERG
analytical verification.
The PAMS participants analyzed the performance evaluation samples according to their
own procedures and reported the results to ERG. ERG compared the analytical results to the
reference analytical results generated by ERG and faxed a report to the participants within
48 hours. Analytical results included the ERG range (mean ERG analytical value ± 2 standard
deviations), percent bias for each analysis relative to the ERG analyzed value for that canister, and
the bias between the two reported results. Some laboratories performed a second and even a
third set of replicate analyses on different analytical systems. Individual results for each analysis
are shown in Appendix A. A representative set of results is shown in Table 4-3.
Combined results for all of the PAMS are shown in Appendix B. In Appendix B, every
individual set of results submitted by each laboratory is considered a data set. There were 98 sets
of individual results. Calculated statistical parameters for the entire set of results submitted by the
PAMS are summarized in Table 4-4.
A compound-specific comparison of ERG and PAMS analytical results is shown
graphically in Figures 4-1 through 4-19. By far the highest degree of scatter in the analyses, as
illustrated by the percent coefficient of variation, is shown by propylene. The graph for propylene
(Figure 4-20) shows two points (replicate analyses on the same analytical system) that are more
than an order of magnitude higher than all of the other analyses. A histogram of all of the
laboratory results (Figure 4-21) illustrates that these two data points are statistical outliers. When
these two outliers for propylene are omitted from the data set, the correspondence between ERG
and PAMS results is far closer (Figure 4-22).
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xlii
-------
Table 4-3
Example Reported Dataset for a PAMS Site
Compound
Analysis
1
Analysis
2
ERG
Range1
%
Bias 1
% Bias
2
Precision
ethylene
7.70
7.70
9.60-13.10
-32.16
-32.16
0.00
ethane
11.00
10.70
14.42-14.64
-24.29
-26.36
0.21
propylene
14.60
14.50
16.15-17.99
-14.49
-15.08
0.07
propane
17.80
17.40
21.33-23.45
-20.49
-22.27
0.28
1-butene
27.90
27.50
22.83-28.83
8.02
6.48
0.28
n-butane
29.00
28.60
27.41-30.67
-0.13
-1.51
0.28
/?-hcxanc
37.50
37.80
37.61-42.07
-5.87
-5.11
0.21
benzene
66.90
66.60
67.67-73.99
-5.56
-5.98
0.21
cyclohexane
32.10
31.80
31.37-34.23
-2.15
-3.06
0.21
toluene
70.10
70.40
73.60-80.68
-9.12
-8.73
0.21
n-octane
46.70
47.00
47.98-52.80
-7.31
-6.72
0.21
ethylbenzene
79.80
81.30
80.56-90.72
-6.82
-5.07
1.06
m-lp-nylene
72.10
73.30
71.14-82.28
-6.01
-4.44
0.85
o-xylene
38.30
38.90
37.75.43.43
-5.64
-4.17
0.42
/7-propylbcnzene
50.50
50.80
45.89-57.27
-2.10
-1.52
0.21
1,3,5 -trimethylbenzene
38.90
39.00
36.46-47.22
-7.02
-6.78
0.07
1,2,4-trimethylbenzene
37.50
38.20
34.82-47.42
-8.81
-7.11
0.49
/?-dccanc
51.80
52.50
46.84-67.01
-9.01
-7.78
0.49
1,2,3 -trimethylbenzene
38.60
39.40
33.55-48.07
-5.42
-3.46
0.57
Total NMOC
831.20
836.20
829.93-944.39
-6.31
-5.74
3.54
1 ERG range = ERG analyzed value for this compound in this canister ± 2 standard deviations for this compound in the
ERG dataset.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xliii
-------
Table 4-4
Statistical Evaluation of Results Obtained from All PAMS1
Compound
Mean
Std. Dev.2
cv
%3
Median
Conf. Lts.4
90%
90th
Pet.5
ethylene
9.72
3.67
37.76
10.50
0.62
12.48
ethane
12.03
4.31
35.83
13.40
0.73
15.20
propylene
21.73
32.22
148.27
17.52
5.47
19.84
propane
20.83
3.21
15.41
20.65
0.54
23.62
1 -butene
25.60
7.12
27.81
27.32
1.21
31.05
«-butane
28.63
3.75
13.10
28.35
0.64
32.00
«-hexane
41.08
8.18
19.91
41.80
1.39
47.22
benzene
74.62
11.23
15.05
75.65
1.91
85.48
cyclohexane
34.22
5.14
15.02
34.06
0.87
38.93
toluene
84.27
10.64
12.63
84.60
1.81
94.60
«-octane
53.99
5.93
10.98
54.62
1.01
62.02
ethylbenzene
94.30
13.70
14.53
95.11
2.32
106.61
m-lp-xylene
80.36
17.50
21.78
84.27
2.97
95.31
o-xylene
43.99
6.32
14.37
44.55
1.07
49.80
«-propylbenzene
56.36
12.23
21.70
57.95
2.07
65.07
1,3,5-trimethylbenzene
43.31
7.06
16.30
44.17
1.20
50.26
1,2,4-trimethylbenzene
42.65
10.67
25.02
44.19
1.81
52.05
«-decane
52.66
22.59
42.90
59.10
3.83
70.73
1,2,3-trimethylbenzene
39.62
17.10
43.16
44.00
2.88
53.43
Total NMOC6
958.32
120.06
12.53
960.98
20.37
1086.55
1 Total of 98 sets of data (37 participating PAMS sites x 2 analyses per site) = 74 sets of data. Some sites
performed additional sets of duplicate determinations, 12 additional determinations x 2 datasets per determination
= 24 additional datasets. Total datasets=74+24=98.
2Standard deviation.
'Coefficient of variation
"•Confidence Limits (90%)
590th percentiles
6Total NMOC represents 94 data sets. All laboratories did not report a value for Total NMOC.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xliv
-------
Ethylene
Canister
—•— ERG Analysis —*— Participants
Figure 4-1. Graphical Comparison of Analytical Results for Ethylene Between ERG and Participants
-------
Ethane
—•— ERG Analysis —*— Participants
Figure 4-2. Graphical Comparison of Analytical Results for Ethane Between ERG and Participants
-------
Propane
Canister
ERG Analysis —*— Participants
Figure 4-3. Graphical Comparison of Analytical Results for Propane Between ERG and Participants
-------
1-Butene
Canister
—•— ERG Analysis —*— Participants
Figure 4-4. Graphical Comparison of Analytical REsults for 1-Butene Between ERG and Participants
-------
n-Butane
Canister
ERG Analysis —*— Participants
Figure 4-5. Graphical Comparison of Analytical Results for n-Butane Between ERG and Participants
-------
n-Hexane
Canister
ERG Analysis —*— Participants
Figure 4-6. Graphical Comparison of Analytical Results for «-Hexane Between ERG and Participants
-------
120
110
100
90
80
70
60
50
40
30
20
10
0
Benzene
Canister
ERG Analysis —*— Participants
Graphical Comparison of Analytical Results for Benzene Between ERG and Participants
-------
Cyclohexane
Canister
—•— ERG Analysis —*— Participants
Figure 4-8. Graphical Comparison of Analytical Results for Cyclohexane Between ERG and Participants
-------
Toluene
Canister
ERG Analysis —*— Participants
Figure 4-9. Graphical Comparison of Analytical Results for Toluene Between ERG and Participants
-------
n-Octane
Canister
ERG Analysis —*— Participants
Figure 4-10. Graphical Comparison of Analytical Results for n-Octane Between ERG and Participants
-------
Ethylbenzene
Canister
—•— ERG Analysis —*— Participants
Figure 4-11. Graphical Comparison of Analytical Results for Ethylbenzene Between ERG and Participants
-------
120
110
100
90
80
70
60
50
40
30
20
10
0
m-/p-Xylene
Canister
ERG Analysis —*— Participants
Graphical Comparison of Analytical Results for m-/p-Xylene Between ERG and Participants
-------
o-Xylene
60
55
50
45
40
^ 35
CL
O 30
CO
c 25
0
o
c
O Of)
o zu
15
10
5
0
167601 2002 2005 2010 2013 2049 2051 2054 2057 2058 2059 2075 2080 2126 2161 2166 2210 2212 2213 2966
Canister
ERG Analysis
Participants
Figure 4-13. Graphical Comparison of Analytical Results for o-Xylene Between ERG and Participants
-------
n-Propylbenzene
Canister
ERG Analysis —*— Participants
Figure 4-14. Graphical Comparison of Analytical Results for «-Propylbenzene Between ERG and Participants
-------
1,3,5-T rimethylbenzene
Canister
ERG Analysis —*— Participants
Figure 4-15. Graphical Comparison of Analytical Results for 1,3,5-Trimethylbenzene Between ERG and Participants
-------
1,2,4-T rimethy Ibenzene
Canister
ERG Analysis —*— Participants
Figure 4-16. Graphical Comparison of Analytical Results for 1,2,4-Trimethylbenzene Between ERG and Participants
-------
Decane
Canister
ERG Analysis —*— Participants
Figure 4-17. Graphical Comparison of Analytical Results for Decane Between ERG and Participants
-------
1,2,3-T rimethy Ibenzene
Canister
ERG Analysis —*— Participants
Figure 4-18. Graphical Comparison of Analytical Results for 1,2,3-Trimethylbenzene Between ERG and Participants
-------
Total NMOC
Canister
ERG Analysis —*— Participants
Figure 4-19. Graphical Comparison of Analytical Results for Total NMOC Between ERG and Participants
-------
Propylene
-in I I I I I I
- iu TT
167601 2002 2005 2010 2013 2049 2051 2054 2057
2058 2059 2075
Canister
2080 2126 2161 2166 2210 2212 2213 2966
ERG Analysis-
Participants
Figure 4-20. Graphical Comparison of Analytical Results for Propylene for ERG and Participants
-------
Histogram of Lab Results: Propylene
<
75
70
65
60
55
50
45
cT
-------
Propylene
270
167601 2002 2005 2010 2013 2049 2051 2054 2057 2058 2059 2075 2080 2126 2161 2166 2210 2212 2213 2966
Canister
ERG Analysis
Participants
Figure 4-22. Graphical Comparison of Analytical Results for Propylene Between ERG and Participants (Outliers Excluded)
-------
The bias of the PAMS compared to the ERG analytical results for the same canister is
summarized in Figure 4-23, with the outliers for propylene excluded. Four compounds (ethylene,
ethane, propane, and decane) show a negative bias; the remainder show a positive bias. If the
absolute value of the bias is plotted (Figure 4-24), ethylene and ethane have the largest bias. As a
group, the compounds eluting closest to propane, the standard, show the lowest bias if the
outliers for propylene are excluded.
The ranges for the ERG analyses of the initial set of 40 canisters and the ranges of the
analyses performed by the PAMS are shown in Table 4-5. The range was calculated as the
mean ±2 standard deviations. The range of the analyses for the PAMS is larger than the range for
the ERG analyses. The wider range for the PAMS can be attributed to the following factors:
• ERG analyses were all performed on the same instrument, under the same
conditions and with the same calibration, in very close time proximity;
• Canisters sent to the PAMS were shipped and subjected to far more handling than
the canisters in the ERG laboratory;
• PAMS used a wide range of analytical instruments and a wide range of conditions,
summarized in Appendix C.
• Analyses performed by PAMS constitute 98 sets of data; ERG analyses encompass
40 canisters.
The ranges of ERG and PAMS analyses are shown graphically in Figures 4-25 through
4-27. On each of these figures, the ERG range and the PAMS range are paired within one
compartment of the figure. Figure 4-25 includes the outlier data points for propylene;
Figure 4-26 excludes these points.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxvii
-------
Bias of PAMS
Ethylene Propane n-Hexane Toluene m-/p-Xylene 1,3,5-Trimethylbenzene1,2,3-Trimethylbenzene
Ethane 1-Butene Benzene n-Octane o-Xylene 1,2,4-Trimethylbenzene Total NMOC
Propylene n-Butane Cydohexane Ethylbenzene n-Propylbenzene Decane
Figure 4-23. Example of the Bias of the Participants Compared to ERG Results for the One Canister
-------
Bias of PAMS
20
15
w
TO
in 10
Ethylene
Ethane
Propane
1-Butene
n-Hexane
Benzene
Toluene
n-Octane
Propylene
m-/p-Xylene 1,3,5-Trimethylbenzene 1,2,3-Trimethylbenzene
o-Xylene 1,2,4-Trimethylbenzene Total NMOC
n-Butane
Cyclohexane
Ethylbenzene
n-Propylbenzene
Decane
Figure 4-24. Example of the Absolute Value of the Bias for the Participants Compared to ERG Results for the One Canister
-------
Table 4-5
Ranges for ERG Analyses and PAMS
Range
Compound
ERG Analyses
PAMS Analyses
ethylene
9.28 - 12.80
2.38 - 17.06
ethane
5.82- 14.38
3.41-20.65
propylene
15.73 - 17.57
42.71 - 86.17
propane
20.49-22.61
14.41-27.25
1-butene
22.82-28.82
11.36 - 39.84
n-butane
26.35 -29.59
21.13-36.13
/?-hcxanc
37.38 -41.82
24.72 - 57.44
benzene
66.66 - 72.94
52.16-97.08
cyclohexane
30.94 - 33.82
23.94-44.50
toluene
72.26 - 79.34
62.99- 105.55
n-octane
46.93 -51.73
42.13 - 65.85
ethylbenzene
78.48 - 88.64
66.90- 121.7
m-lp-nylene
69.08 - 80.24
45.36 - 115.36
o-xylene
36.52-42.20
31.35 - 56.63
/7-propylbcnzene
44.17 - 55.53
31.90 - 80.82
1,3,5 -trimethylbenzene
34.42-45.18
29.19-57.43
1,2,4-trimethylbenzene
32.46-45.06
21.31-63.99
/?-dccanc
43.33 - 63.49
7.48 - 97.84
1,2,3 -trimethylbenzene
30.91 -45.43
5.62-73.62
Total NMOC
804.20-918.64
718.20- 1198.44
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxx
-------
Ranges 1
110
100 --
90 --
80 --
70 --
60 +
O
-go
Q.
c40
o
R30H-
§>0
c
o
OlO
I
I
=" I
I
I
-10
-20 --
-30 --
-40 --
-50
H H
H H
H h
H h
H h
H 1 1 h
Ethylene (Anal) Ethane (All Anal) Propane (Anal) 1-Butene (All Anal) n-Hexane (Anal) Benzene (All Anal) Toluene (Anal)
Ethylene (All Anal) Propylene (Anal) Propane (All Anal) n-Butane (Anal) n-Hexane (All Anal) Cyclohexane (Anal) Toluene (All Anal)
Ethane (Anal) Propylene (All Anal) 1-Butene (Anal) n-Butane (All Anal) Benzene (Anal) Cyclohexane (All Anal)
Figure 4-25. ERG Analysis Range and Participant Analysis Range, Including All Data Points (Ethylene to Toluene)
-------
Ranges 1
110
100
90
80
O 70
.Q
Q.
Q.
I
I
CO
60
50
c
0
o
c
o
O 40
30
20
10
I
I
0
Ethylene (Anal) Ethane (All Anal) Propane (Anal) 1-Butene (All Anal) n-Hexane (Anal) Benzene (All Anal) Toluene (Anal)
Ethylene (All Anal) Propylene (Anal) Propane (All Anal) n-Butane (Anal) n-Hexane (All Anal) Cyclohexane (Anal) Toluene (All Anal)
Ethane (Anal) Propylene (All Anal) 1-Butene (Anal) n-Butane (All Anal) Benzene (Anal) Cyclohexane (All Anal)
Figure 4-26. ERG Analysis Range and Participant Analysis Range, Excluding Outlier Points for Propylene
(Ethylene to Toluene)
-------
Ranges 2
I
i
i
n-Octane (Anal) Ethylbenzene (All Anal) o-Xylene (Anal) n-Propylbenzene (All Anal)1,2,4-Trimethylbenzene (Anal) Decane (All Anal)
n-Octane (All Anal) m-/p-Xylene (Anal) o-Xylene (All Anal) 1,3,5-Trimethylbenzene (Art5ff),4-Trimethylbenzene (All Art^J);3-Trimethylbenzene (Anal)
Ethylbenzene (Anal) m-/p-Xylene (All Anal) n-Propylbenzene (Anal)1,3,5-Trimethylbenzene (All Anal) Decane (Anal) 1,2,3-Trimethylbenzene (All Anal)
Figure 4-27. ERG Analysis Range and Participants Analysis Range (n-Octane to Trimethylbenzene)
-------
A confidence interval is a range on either side of the sample mean used to estimate the
population mean with a specified probability. For small sample sizes (usually less than 100), the
confidence interval is calculated as follows:
STD
Confidence Interval = MEAN ± t., ,, ,, x 1
vn
where:
Confidence Interval =
the range for the estimate of the population mean for the
Specified probability, 1-a (for a 95% probability level, 1-a
= 0.95 and therefore a = 0.05)
MEAN.
the sample mean
t(i-«/2, n-i) = the two-sided lvalue for the specified probability (1 - a)
and degrees of freedom (n-1)
STDs = the sample standard deviation
n = the sample size
The confidence intervals for the ERG and PAMS data sets are shown graphically in
Figures 4-28 through 4-30. Figure 4-28 includes the two propylene outlier points, Figure 4-29
does not include these points. Again, the ERG ranges are far narrower than the PAMS ranges,
for the reasons described above.
The percent difference between replicate analyses arranged by laboratory is shown
graphically in Appendix F, arranged by compound in Appendix G.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxiv
-------
o §-
§2
ig c
t3 co
d [n
Confidence Intervals 1
<
100
90
80
70
O
.Q
Q.
Q.
C
o
ro
c
0
o
c
o
O
60
50
40
30
20
10
0
Ethylene (Anal) Ethane (All Anal) Propane (Anal) 1-Butene (All Anal) n-Hexane (Anal) Benzene (All Anal) Toluene (Anal)
Ethylene (All Anal) Propylene (Anal) Propane (All Anal) n-Butane (Anal) n-Hexane (All Anal) Cyclohexane (Anal) Toluene (All Anal)
Ethane (Anal) Propylene (All Anal) 1-Butene (Anal) n-Butane (All Anal) Benzene (Anal) Cyclohexane (All Anal)
Figure 4-28. Confidence Intervals for ERG and Participant Data Sets, Including Propylene Outliers (Ethylene to Toluene)
-------
Confidence Intervals 1
100
90
80
70
60
O
.Q
Q.
Q.
C
o
ro 5(3
c
0
o
c
o
O
40
30
20
10
—
I
—
I
--
= i
= i
—
— zc
— ^
= X
_ =n
__
i
i
zn
i
i
i
i
i
i
—I—
—I—
i
i
i
i
i
i
i
0
Ethylene (Anal) Ethane (All Anal) Propane (Anal) 1-Butene (All Anal) n-Hexane (Anal) Benzene (All Anal) Toluene (Anal)
Ethylene (All Anal) Propylene (Anal) Propane (All Anal) n-Butane (Anal) n-Hexane (All Anal) Cyclohexane (Anal) Toluene (All Anal)
Ethane (Anal) Propylene (All Anal) 1-Butene (Anal) n-Butane (All Anal) Benzene (Anal) Cyclohexane (All Anal)
Figure 4-29. Confidence Intervals for ERG and Participant Data Sets, Excluding Propylene Outliers (Ethylene to Toluene)
-------
Confidence Intervals 2
100
90
80
70
60
50
40
30
20
10
0
—
I—I
H
I—I
H
X
I
I 1
H
—
I
I
I
I—I
H
i
i
I I
I I
I I
I I
I I
I I
I I
I I
I I
I I
I I
I I
1
1
n-Octane (Anal) Ethylbenzene (All Anal) o-Xylene (Anal) n-Propylbenzene (All Anal)1,2,4-Trimethylbenzene (Anal) Decane (All Anal)
n-Octane (All Anal) m-/p-Xylene (Anal) o-Xylene (All Anal) 1,3,5-Trimethylbenzene (Art5ff),4-Trimethylbenzene (All Art^J);3-Trimethylbenzene (Anal)
Ethylbenzene (Anal) m-/p-Xylene (All Anal) n-Propylbenzene (Anal)1,3,5-Trimethylbenzene (All Anal) Decane (Anal) 1,2,3-Trimethylbenzene (All Anal]
Figure 4-30. Confidence Intervals for ERG and Participant Data Sets (n-Octane to Trimethylbenzene)
-------
4.2 Stability Study
Four canisters were designated at the beginning of the study as Stability Study Canisters.
These canisters were filled and analyzed as part of the original batch, then set aside (not shipped
to PAMS) to be re-analyzed approximately four weeks later (the actual time for the re-analysis
was approximately eight weeks). Results are shown in Table 4-6 through 4-9. Figure 4-31 plots
the results for the first analysis against the results for the second analysis.
If both sets of analytical results were in exact agreement (i.e., no change in canister
concentration over the time period), all of the points in Figure 4-31 would fall on the 45 ° line.
However, most points fall slightly below the 45 ° line. The plot indicates that results for the
second analysis are slightly lower than the results for the first analysis. The calculated percent
difference between the two runs ranges from -0.09 to -3.53, with the exception of ethylene and
ethane. In two of the four canisters, 1-butene shows an increase, while ethylene and ethane show
a decline larger than most of the other compounds. In the four canisters, the decline of ethane
ranges from -15 to nearly -60% difference. Ethylene shows a slight increase (approximately 1%)
in two of the canisters, and a decline of -15 to approximately -40% difference in the other two
canisters. If the reason for the negative percent difference were a decline in the canister
concentration of ethylene and ethane, the two compounds would be expected to decline in
concentration together, in proportional amounts. This type of variation
is not observed. In fact, in two of the canisters, ethylene shows a slight increase while ethane
declines. At the same time, propylene, the next compound in chromatographic elution, shows a
nearly constant level. The erratic measurements of ethylene and ethane may thus be due to factors
associated with the analysis: possibly the sample introduction system, interaction of the
compounds with the Nafion® drier, or the presence of small amounts of water in the early portion
of the chromatogram. Analytical procedures for ethane and ethylene should be studied carefully
for possible modification.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxviii
-------
Table 4-6
Stability Study Performed on Canisters Retained at the ERG Laboratory
Canister # 2061
Compound
1st Analysis
8/8/96
2nd Analysis
10/3/96
Precision
%Difference
ethylene
10.58
7.79
1.97
-15.19
ethane
8.18
2.63
3.92
-51.34
propylene
15.53
15.47
0.04
-0.19
propane
20.13
20.08
0.04
-0.12
1-butene
24.15
28.85
3.32
8.87
n-butane
25.68
25.38
0.21
-0.59
/?-hcxanc
36.42
36.52
0.07
0.14
benzene
63.53
62.88
0.46
-0.51
cyclohexane
29.55
28.93
0.44
-1.06
toluene
69.21
68.63
0.41
-0.42
n-octane
45.36
45.07
0.21
-0.32
ethylbenzene
76.73
75.70
0.73
-0.68
m-lp-nylene
68.07
67.13
0.66
-0.70
o-xylene
37.25
35.40
1.31
-2.55
/7-propylbcnzene
46.47
46.71
0.17
0.26
1,3,5 -trimethylbenzene
35.14
34.69
0.32
-0.64
1,2,4-trimethylbenzene
33.52
34.12
0.42
0.89
/?-dccanc
52.35
51.61
0.52
-0.71
1,2,3 -trimethylbenzene
32.61
32.90
0.21
0.44
Total NMOC
791.33
790.06
0.90
-0.08
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxix
-------
Table 4-7
Stability Study Performed on Canisters Retained at the ERG Laboratory
Canister # 2078
1st Analysis 2nd Analysis
Compound
8/8/96
10/3/96
Precision
% Difference
ethylene
10.69
10.99
0.21
1.38
ethane
8.02
5.98
1.44
-14.57
propylene
15.78
15.65
0.09
-0.41
propane
20.38
20.56
0.13
0.44
1-butene
23.62
23.53
0.06
-0.19
n-butane
26.12
25.83
0.21
-0.56
/?-hcxanc
36.97
36.39
0.41
-0.79
benzene
64.73
63.96
0.54
-0.60
cyclohexane
30.06
29.65
0.29
-0.69
toluene
7-.47
69.51
0.68
-0.69
n-octane
46.03
45.95
0.06
-0.09
ethylbenzene
78.19
76.46
1.22
-1.12
m-lp-nylene
69.46
67.63
1.29
-1.33
o-xylene
36.32
35.64
0.48
-0.94
/7-propylbcnzene
47.22
46.35
0.62
-0.93
1,3,5 -trimethylbenzene
36.01
35.19
0.58
-1.15
1,2,4-trimethylbenzene
34.61
33.94
0.47
-0.98
/?-dccanc
52.85
51.23
1.15
-1.56
1,2,3 -trimethylbenzene
33.81
32.78
0.73
-1.55
Total NMOC
802.64
793.20
6.68
-0.59
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxx
-------
Table 4-8
Stability Study Performed on Canisters Retained at the ERG Laboratory
Canister # 2208
1st Analysis 2nd Analysis
Compound
8/8/96
10/3/96
Precision
% Difference
ethylene
10.62
4.86
4.07
-37.21
ethane
9.03
2.40
4.69
-58.01
propylene
15.57
15.40
0.12
-0.55
propane
19.97
20.02
0.0-4
0.13
1-butene
23.38
27.62
3.00
8.31
n-butane
25.68
25.30
0.27
-0.75
/?-hcxanc
36.03
36.43
0.28
0.55
benzene
63.70
62.36
0.95
-1.06
cyclohexane
29.56
28.90
0.47
-1.13
toluene
69.60
68.03
1.11
-1.14
n-octane
45.94
44.52
1.00
-1.57
ethylbenzene
77.32
74.47
2.02
-1.88
m-lp-nylene
68.85
66.08
1.96
-2.05
o-xylene
36.00
34.90
0.78
-1.55
/7-propylbcnzene
46.91
44.85
1.46
-2.24
1,3,5 -trimethylbenzene
36.37
34.31
1.46
-2.91
1,2,4-trimethylbenzene
35.28
33.46
1.29
-2.65
/?-dccanc
52.44
49.06
2.39
-3.33
1,2,3 -trimethylbenzene
34.91
32.53
1.68
-3.53
Total NMOC
795.73
768.22
19.45
-1.76
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxxi
-------
Table 4-9
Stability Study Performed on Canisters Retained at the ERG Laboratory
Canister # GP00036
1st Analysis 2nd Analysis
Compound
8/8/96
10/3/96
Precision
% Difference
ethylene
10.70
20.93
0.16
1.06
ethane
11.89
6.16
4.05
-31.75
propylene
15.77
15.53
0.17
-0.77
propane
20.42
20.58
0.11
0.39
1-butene
26.97
24.12
2.02
-5.58
n-butane
26.09
25.83
0.18
-0.50
/?-hcxanc
36.93
36.24
0.49
-0.94
benzene
64.11
63.47
0.45
-0.50
cyclohexane
29.83
29.47
0.25
-0.61
toluene
68.99
68.82
0.12
-0.12
n-octane
46.79
46.16
0.45
-0.68
ethylbenzene
75.15
74.87
0.20
-0.19
m-lp-nylene
65.66
65.42
0.17
-0.18
o-xylene
34.41
34.33
0.06
-0.12
/7-propylbcnzene
45.50
45.55
0.04
0.05
1,3,5 -trimethylbenzene
32.40
32.31
0.06
-0.14
1,2,4-trimethylbenzene
29.86
30.31
0.32
0.75
/?-dccanc
51.51
51.26
0.18
-0.24
1,2,3 -trimethylbenzene
28.37
28.61
0.17
0.42
Total NMOC
782.13
776.89
3.71
-0.34
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxxii
-------
Stability Study
10
20
30 40 50
Analysis 1 (ppbc)
60
70
80
2061
~ 2078
2208
GP00036
Figure 4-31. Stability Study for Canisters Retained at the ERG Laboratory
-------
4.3 Outlier Analysis
An outlier analysis was conducted for each compound using all nonzero values from all
PAMS sites for both Analysis 1 and Analysis 2. Any outlier analysis contains subjective elements,
but the philosophy used in this outlier analysis was very conservative—only the most blatantly
obvious of the candidate outliers were labeled as such.
Candidate outliers were identified by examing the data distribution for gaps between the
main body of data and extreme values on either end of the distribution. Two methods, the NSI
method and the F-spread method, were used to determine cutoff points, beyond which, data
points would be considered as candidate outliers for further scrutiny.
In the NSI method, the data distribution was classified as normal, lognormal, or other (not
normal or lognormal), using the SAS UNIVARIATE procedure. Candidate outliers were
identified according the data distribution classification, in the following sequence:
• Normal Distribution: any data point a distance of more than 3 times the standard
deviation from the mean
• Lognormal Distribution (but not Normal): any data point a distance of more than
3 times the standard deviation from the mean of the natural logarithms
• Other Distribution (not Normal or Lognormal): any data point a distance of more
than 6 times the standard deviation from the mean
In the F-spread method, the F-spread was calculated as the difference between the 75th
and 25th percentiles. Any data point a distance of more than 1.5 times the F-spread from the
mean was identified as a candidate outlier.
When the candidate outliers were identified from the NSI and F-spread methods, they
were visually identified on a plot of the data distribution. If there was a substantial gap between
the main body of data and the candidate outlier(s), if no other data points were located in this
region of the data distribution, and the value of the candidate outlier(s) was substantially different
from the other data points, the candidate outlier(s) was officially classified as an outlier(s).
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxxiv
-------
Only two points, both for replicate determinations of propylene from the same PAMS site
and canister, were officially classified as outliers.
4.4 Additional Compounds Identified
In addition to the ozone precursors spiked into the performance evaluation canisters, four
additional compounds were spiked: 1,3-butadiene, tert-butyl methyl ether, 1-hexene, and
1,1,1-trichloroethane. All of these additional compounds could be observed in the PAMS
analysis, with the exception of tert-butyl methyl ether which would have been removed if the
analytical system used in the analysis was equipped with a drier, tert-Butyl methyl ether,
1-hexane, and 1,1,1-trichloroethane were reported in only four analyses; 1,3-butadiene was
reported in only eight analyses.
The introduction of the additional compounds provided a challenge to the compound
identification procedures at a number of PAMS. Many PAMS reported additional compounds
that were not spiked into the canisters. Some laboratories carefully characterized the slightest
peak observed above the chromatographic baseline. The additional compounds reported and the
number of times they were reported are shown in Table 4-10. The most frequently identified
additional compound, 2-methyl-l-pentene, was not spiked into the canisters. One of the
additional spiked compounds, 1-hexene, was frequently misidentified as 2-methyl-l-pentene.
However, ten analyses reported 2,3-dimethylbutane (not present) and six analyses reported
2-methylpentane, both misidentifications of tert-butyl methyl ether. Five analyses reported
2,4-dimethylpentane, a misidentification of 1,1,1-trichloroethane.
More than twenty analysis reports each identified isopropylbenzene, w-ethyltoluene,
o-ethyl toluene, w-diethylbenzene, and //-undecane, none of which were components of the
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxxv
-------
Table 4-10
Additional Compounds Reported and Frequency of Reporting
Compound
Frequency
Compound
Frequency
acetylene
2
2,4-dimethylpentane
5
isobutane
15
1,1,1 -trichloroethane
4
1,3-butadiene
8
2-methylhexane
1
/ra/?.v-2-butcnc
24
3-methylhexane
5
c/.v-2-Butcnc
18
2,2,4-trimethylpentane
2
isopentane
4
n-heptane
5
1-pentene
7
methylcyclohexane
2
/?-pcntanc
2
3-methylheptane
28
isoprene
5
styrene
26
ftans-2-pentene
1
/?-nonanc
9
c/.v-2-pcntcnc
1
isopropylbenzene
43
2,2-dimethylbutane
2
/M-cthyltolucnc
22
cyclopentane
1
/7-ethyltoluene
7
2,3-dimethylbutane
10
m- /p-cthyltoluene
2
3-methylpentane
6
o-ethyltoluene
27
methyl /-butyl ether
4
m-diethylbenzene
39
3-methylpentane
4
/p-dicthylbenzcne
16
1-hexene
4
m- /p-dicthylbenzcne
2
2-methyl-1 -pentene
54
/?-undccanc
21
methylcyclopentane
8
unknown (one or more)
22
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxxvi
-------
gaseous mixture spiked into the canisters. Because of the frequency of the reports of the
occurrence of these compounds, ERG exhaustively characterized the four stability study canisters
retained in the ERG laboratory. From the examination of this limited sample (Table 4-11), ERG
concludes that several of these compounds are indeed present in the canisters at trace levels
(< 1 ppbC). It is not known whether these compounds are present as a residual from previous
samples that were not completely removed in the canister cleaning process, as an impurity in the
cylinder gases used to prepare the performance evaluation standards, or as a residual in the
analytical system. Blank canister samples (humidified zero air) prepared and analyzed in the ERG
laboratory do not show the presence of these three- and four-carbon alkylbenzenes or undecane at
concentrations at or above the method detection limit.
Table 4-11
Additional Compounds Identified in ERG Stability Study Canisters
Additional Compounds Frequency (Out of Four Canisters)
/ra/?.v-2-butcnc
2
c/s-2-butene
2
isopentane
1
2-methylpentane
2
trans-2-hexene
1
c/.v-2-hcxcnc
2
3-methylhexane
1
3-methylheptane
1
m-diethylbenzene
2
p-diethylbenzene
4
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxxvii
-------
4.5
ERG Re-Analvsis of Performance Evaluation Canisters
As an additional feature of the Performance Evaluation program, sixteen of the canisters
returned from the PAMS were re-analyzed by ERG upon their return. The results for the original
ERG analysis, the second ERG analysis, the percent difference between the two ERG analyses, as
well as the results of the PAMS analyses for each canister are shown in Appendix D. A general
criterion for selection of a particular canister for re-analysis was a significant difference between
the PAMS results and the ERG original analysis results (the repeated ERG analysis always
reinforced the first ERG analysis results). For example, the canister that showed the outlier points
for the propylene analysis was re-analyzed upon its return to ERG. The repeated analysis by ERG
showed a percent difference of approximately -4% from the original analysis. The values for
percent difference (equation below) for the ERG re-analysis are summarized for each analyte in
Table 4-12.
ERG, - ERG,
% Difference = X 100
ERGj
where:
ERG, = first ERG analysis
ERG2 = second ERG analysis
There are some isolated spikes in the values of the differences, but the results are generally
negative and approximately 10% lower than the original analysis. The widest ranges are shown
by propylene, 1-butene, //-butane, and benzene. It is interesting to note that the range of
differences for the ERG TNMOC analyses is no wider than most of the individual analytes.
The repeated analysis of these canisters was performed at the same time as the repeated
analysis of the stability study canisters retained by ERG. However, the drop in compound
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxxviii
-------
Table 4-12
Percent Difference Between First and Second ERG Analysis
Ethylene
Ethane
Propylene
Propane
1-Butene
-10.41
-10.07
-11.18
-9.99
-9.28
-10.82
-8.43
-19.07
-10.56
-9.89
-10.28
-7.36
-15.75
-10.08
-11.23
-11.44
-8.64
-9.13
-10.77
-4.98
-10.06
-7.35
-26.07
-9.78
-5.83
-10.23
-8.49
-0.28
-10.14
30.03
-10.85
-9.58
36.54
-10.49
-7.54
0.28
-0.46
1.08
-0.94
3.43
-10.86
-7.63
-44.89
-10.13
-9.73
-10.97
-8.34
-19.17
-10.56
-11.64
-9.75
-12.49
-15.50
-9.66
3.78
-9.91
-8.75
49.21
-9.57
-7.09
-1.17
1.09
14.80
-1.20
-2.75
-10.16
-8.68
-4.06
-9.70
-3.95
-10.29
-7.27
-6.23
-9.75
-7.68
-5.56
-4.39
3.90
-5.40
-5.97
Mean: -8.91
-7.31
-4.11
-8.67
-3.77
SD: 3.57
3.42
22.95
3.21
10.08
%CV: -39.95
-46.83
-558.02
-37.00
-267.40
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-lxxxix
-------
Table 4-12
Continued
w-Butane
w-Hexane
Benzene
Cyclohexane
Toluene
-29.23
-9.55
-4.14
-8.64
-10.76
0.65
-10.48
-1.82
-10.95
-11.10
13.10
-10.96
-3.61
-9.40
-10.45
10.79
-12.26
-3.62
-10.85
-11.57
-7.71
-8.63
-2.01
-9.35
-10.14
-6.72
-6.51
-1.08
-9.53
-7.68
7.41
-9.39
-3.47
-11.74
-10.69
8.68
-0.22
3.47
0.44
-0.97
16.70
-9.83
-5.03
-11.63
-10.18
-9.48
-10.27
-4.02
-11.32
-11.01
-16.80
-8.28
-3.63
-9.09
-9.49
8.02
-8.22
-3.03
-8.99
-10.25
-3.01
-2.59
3.43
0.16
-2.98
1.54
-9.45
-2.44
-9.57
-10.18
0.52
-9.55
-3.69
-9.51
-10.29
-57.37
-4.78
1.90
-6.07
-6.25
Mean: -3.93
-8.19
-2.04
-8.50
-9.00
SD: 18.53
3.19
4.57
3.71
3.06
%CV: -471.20
-38.99
-222.84
-43.58
-34.03
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xc
-------
Table 4-12
Continued
w-octane
ethylbenzene
/?7-//;-xylene
o-xylene
w-propylbenzene
-8.74
-9.92
-9.30
-9.59
-9.23
-6.28
-11.27
-8.37
-10.71
-11.75
-6.46
-12.87
-8.83
-9.95
-11.25
-5.81
-14.01
-7.81
-11.63
-14.65
-6.17
-8.35
-7.61
-9.19
-6.67
-6.40
0.72
-7.43
-8.04
-4.42
-5.96
-9.29
-8.28
-9.91
-9.06
1.55
1.64
-0.01
0.33
2.14
-6.08
-10.31
-9.78
-10.12
-9.81
-7.07
-11.07
-9.69
-10.59
-10.80
-7.34
-6.99
-8.76
-8.25
-6.82
-6.10
-8.69
-8.46
-9.46
-6.34
0.60
-2.54
-0.28
-1.26
-3.29
-6.07
-9.28
-7.68
-10.27
-8.65
-6.47
-9.61
00
00
00
i
-9.70
-8.58
-2.47
-5.40
-3.53
-4.80
-6.60
Mean: -5.23
-7.95
-7.16
-8.32
-7.86
SD: 2.79
4.49
3.09
3.43
3.91
%CV: -52.41
-56.42
-43.09
-41.25
-49.67
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xci
-------
Table 4-12
Continued
1,3,5-
trimethylbenzene
1,2,4-
trimethylbenzene
«-decane
1,2,3-
trimethylbenzene
Total
NMOC
-9.40
-10.47
-9.68
-9.93
-8.62
-10.80
-11.18
-10.64
-11.05
-9.14
-11.07
-10.55
-12.24
-10.03
-8.86
-13.83
-13.40
-12.66
-12.59
-9.76
-6.75
-8.55
-8.54
-9.03
-7.31
-1.11
-4.49
-4.41
-7.17
-6.88
-8.33
-9.69
-9.21
-9.48
-8.58
8.96
0.73
-0.03
1.19
2.90
-9.52
-10.24
-9.92
-9.87
-9.04
-10.14
-10.72
-10.25
-10.38
-9.30
-6.26
-7.99
-9.17
-7.88
-7.18
-7.23
-7.80
-6.89
-9.57
-7.08
-2.44
-2.88
-3.22
-2.10
-0.25
-8.39
-9.35
-9.61
-9.85
-7.97
-8.45
-9.61
-9.71
-9.80
-8.23
-5.60
-5.90
-5.04
-5.37
-4.59
Mean: -6.90
-8.26
-8.20
-8.30
-6.86
SD: 5.28
3.56
3.42
3.51
3.48
%CV: -76.53
-43.11
-41.70
-42.29
-50.71
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xcii
-------
concentrations for the canisters returned from the external laboratories is considerably larger
because:
• These canisters have been shipped twice and losses could have occurred in
shipping and handling.
• The canisters have been opened at least four times for analyses (at least twice by
the PAMS and twice by ERG).
Even under these conditions of shipping and handling for these canisters, the concentration
of most of the analytes declined 10% or less.
The comparison of ERG results for the first and second analyses of the four canisters
retained at the ERG laboratory and the sixteen canisters shipped to PAMS sites and returned is
shown graphically by compound in Figures 4-32 through 4-51. If ERG Analysis 1 has the same
result as ERG Analysis 2, the point falls on the line drawn at a 45-degree angle in the plot. A
point above the line indicates that Analysis 2 had a higher result than Analysis 1, while points
below the line show that Analysis 2 had a lower result than Analysis 1.
4.6 Additional Statistical Calculations Relative to the Mean of the PAMS Analyses
Excluding Outliers
The statistical calculations discussed in the previous sections have compared analytical
results obtained by the PAMS sites to the analytical results obtained by ERG in the ERG
laboratories. Additional statistical calculations were performed comparing the analytical results
obtained by the PAMS sites to the mean of the results obtained by the PAMS sites, excluding
outliers. As discussed in Section 4.3, only two data points were removed from the dataset as
outliers according to the statistical evaluation. Values for compounds not reported were not
included in the statistical calculations: i.e., if ten determinations did not report
1,2,3-trimethylbenzene, the number of data points used in the statistical calculations was 88 rather
than 98. The compounds most frequently not reported by the PAMS sites are shown in Table 4-
13.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-xciii
-------
Comparison of Analyses 1 Versus 2
1,2,3-Trimethylbenzene
Figure 4-32. Comparison of ERG Analysis 1 to ERG Analysis 2:
1,2,3-Trimethylbenzene
-------
Comparison of Analyses 1 Versus 2
1,2,4-Trimethylbenzene
Figure 4-33. Comparison of ERG Analysis 1 to ERG Analysis 2:
1,2,4-Trimethylbenzene
-------
Comparison of Analyses 1 Versus 2
1,3,5-Trimethylbenzene
Figure 4-34. Comparison of ERG Analysis 1 to ERG Analysis 2:
1,3,5-Trimethylbenzene
-------
Comparison of Analyses 1 Versus 2
1-Butene
Figure 4-35. Comparison of ERG Analysis 1 to ERG Analysis 2:
1-Butene
-------
Comparison of Analyses 1 Versus 2
Benzene
72
70
•
•
?••••
68
66
64
62
62
64
66 68
Analysis 1
70
72
Figure 4-36. Comparison of ERG Analysis 1 to ERG Analysis 2:
Benzene
-------
Comparison of Analyses 1 Versus 2
Cyclohexane
33
32
31
30
29
•
eJ
29 30 31 32 33
Analysis 1
Figure 4-37. Comparison of ERG Analysis 1 to ERG Analysis 2:
Cyclohexane
-------
Comparison of Analyses 1 Versus 2
Ethane
Figure 4-38. Comparison of ERG Analysis 1 to ERG Analysis 2:
Ethane
-------
Comparison of Analyses 1 Versus 2
Ethylbenzene
Figure 4-39. Comparison of ERG Analysis 1 to ERG Analysis 2:
Ethylbenzene
-------
tfl
55
>
cd
>
O
&
H
o
CO
o
o
4^
o
CM
CO
'(/)
CO
c
<
Comparison of Analyses 1 Versus 2
Ethylene
22
20
18
16
V,/ A .
tr\ 14
12
10
8
8
10 12 14
Analysis 1
16
18
20
22
Figure 4-40. Comparison of ERG Analysis 1 to ERG Analysis 2:
Ethylene
-------
Comparison of Analyses 1 Versus 2
Propane
Figure 4-41.
Comparison of ERG Analysis 1 to ERG Analysis 2:
Propane
-------
Comparison of Analyses 1 Versus 2
Propylene
Figure 4-42. Comparison of ERG Analysis 1 to ERG Analysis 2:
Propylene
-------
Comparison of Analyses 1 Versus 2
Toluene
Figure 4-43. Comparison of ERG Analysis 1 to ERG Analysis 2:
Toluene
-------
Comparison of Analyses 1 Versus 2
Total NMOC
760 780 800 820 840 860 880 900
Analysis 1
Figure 4-44. Comparison of ERG Analysis 1 to ERG Analysis 2:
Total NMOC
-------
Comparison of Analyses 1 Versus 2
m/p-Xylene
Figure 4-45. Comparison of ERG Analysis 1 to ERG Analysis 2:
m/p-Xy\tnt
-------
Comparison of Analyses 1 Versus 2
n-Butane
Figure 4-46. Comparison of ERG Analysis 1 to ERG Analysis 2:
tt-Butane
-------
Comparison of Analyses 1 Versus 2
n-Decane
35 40 45 50 55 60
Analysis 1
Figure 4-47. Comparison of ERG Analysis 1 to ERG Analysis 2:
tt-Decane
-------
Comparison of Analyses 1 Versus 2
n-Hexane
Figure 4-48. Comparison of ERG Analysis 1 to ERG Analysis 2:
tt-Hexane
-------
Comparison of Analyses 1 Versus 2
n-Octane
51
50
49
48
47
46
45
44
•
1
- •«
44
45
46
47 48
Analysis 1
49
50
51
Figure 4-49. Comparison of ERG Analysis 1 to ERG Analysis 2:
ft-Octane
-------
Comparison of Analyses 1 Versus 2
n-Propylbenzene
45 46 47 48 49 50 51 52 53
Analysis 1
Figure 4-50. Comparison of ERG Analysis 1 to ERG Analysis 2:
tt-Propylbenzene
-------
Comparison of Analyses 1 Versus 2
o-Xylene
Figure 4-51. Comparison of ERG Analysis 1 to ERG Analysis 2:
o-Xylene
-------
Table 4-13
Compounds Most Frequently Not Reported by PAMS Sites
Compound
Maximum Possible
Data Points
Number of
Reported Values
1,2,3 -trimethylbenzene
98
88
1,2,4-trimethylbenzene
98
96
1,3,5 -trimethylbenzene
98
98
1-butene
98
97
benzene
99
99
cyclohexane
98
98
ethane
98
91
ethylbenzene
98
98
ethylene
98
90
propane
99
99
propylene
98
95
toluene
99
99
m-lp-nylene
98
98
n-butane
99
99
/?-dccanc
98
91
/?-hcxanc
98
98
n-octane
99
99
/7-propylbcnzene
98
96
o-xylene
98
98
Total NMOC
98
95
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-cxiv
-------
The compound reported least frequently (i.e., missed in the analysis most frequently;
reported in 88 out of 98 datasets) was 1,2,3-trimethylbenzene, a late-eluting compound. Next
lowest in order of frequency of reporting was ethylene (reported in 90 out of 98 datasets),
followed by ethane and //-decane (both reported in 91 out of 98 datasets). The compounds most
likely not to be reported in the analysis occurred at the early and late extremes of
chromatographic elution.
The reproducibility of the measurement by compound for all of the PAMS sites is shown
in Table 4-14. The compound name, the number of values in the database, and the standard
deviation of the absolute bias is presented. Compounds with the highest values for standard
deviation of the absolute bias showed the widest range in the reported measurements.
Compounds with the highest standard deviation (>15) were: 1,2,3-trimethylbenzene,
1-butene, and //-decane. Two of these compounds, 1,2,3-trimethylbenzene and //-decane, were
also compounds that were not reported several times in the 98 datasets. A reason why these
particular compounds should be both more difficult to find and more difficult to quantify
accurately is not obvious.
A summary of the statistical calculations by compound compared to the mean of the
reported values from the PAMS sites (excluding zeroes and outliers) is shown in Table 4-15. A
summary of the statistical calculations for laboratory bias (both positive and negative bias),
arranged according to laboratory, is shown in Table 4-16, with a graphic presentation of the
overall average percent bias shown in Figure 4-52. If the magnitude (i.e., absolute value) of the
bias is considered, summary statistics by laboratory are shown in Table 4-17 and are presented
graphically in Figure 4-53. The overall mean bias (considering bias as a signed value) is -0.25,
very nearly zero, whereas the overall mean absolute bias (considering only the magnitude of the
bias) is 11.31. Thus, on the whole, an individual PAMS site was within approximately 10% of the
mean value of all the PAMS data points.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-cxv
-------
Table 4-14
Reproducibility of the PAMS Measurements by Compound
Compound
Number of
Reported Values
Standard Deviation of
Absolute Bias
1,2,3 -trimethylbenzene
88
16.62
1,2,4-trimethylbenzene
96
11.47
1,3,5 -trimethylbenzene
98
8.18
1-butene
97
18.63
benzene
99
11.67
cyclohexane
98
7.21
ethane
91
13.73
ethylbenzene
98
6.63
ethylene
90
13.91
propane
99
11.00
propylene
95
9.62
toluene
99
9.13
m-lp-nylene
98
14.41
n-butane
99
9.61
/?-dccanc
91
20.60
/?-hcxanc
98
13.65
n-octane
99
7.20
/7-propylbcnzene
96
8.41
o-xylene
98
6.95
Total NMOC
95
8.32
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-cxvi
-------
Table 4-15
Statistical Calculations by Compound Compared to the Mean of the Reported Values from the PAMS Sites,
Excluding Zeros and Outliers
COMPOUND
N1
MEANCONC2
STDCONC3
CVCONC4
LO90CONC5
UP90CONC'
1,2,3 -trimethylbenzene
88
44.66
10.13
22.68
42.86
46.45
1,2,4-trimethylbenzene
96
43.99
7.56
17.18
42.71
45.27
1,3,5 -trimethylbenzene
98
43.77
5.51
12.59
42.84
44.69
1-butane
97
26.18
6.12
23.38
25.15
27.21
benzene
99
74.66
11.19
14.98
72.79
76.53
cyclohexane
98
34.59
3.77
10.91
33.96
35.22
ethane
91
13.11
2.45
18.67
12.68
13.53
ethylbenzene
98
95.32
9.75
10.23
93.68
96.95
ethylene
90
10.69
2.04
19.11
10.33
11.05
propane
99
20.83
3.19
15.31
20.30
21.37
propylene
95
17.51
2.26
12.93
17.12
17.89
toluene
99
84.30
10.59
12.57
82.54
86.07
Total NMOC
95
958.77
119.50
12.46
938.41
979.14
m /p-xylcnc
98
81.22
15.46
19.04
78.62
83.81
/7-butane
99
28.66
3.74
13.06
28.04
29.29
/?-dccanc
91
57.45
16.82
29.28
54.52
60.38
/?-hcxanc
98
41.54
7.04
16.95
40.36
42.72
/?-octanc
99
54.02
5.90
10.92
53.03
55.00
/7-propylbenzene
96
58.15
7.02
12.08
56.96
59.34
o-xylene
98
44.45
4.45
10.01
43.70
45.20
ALL
1,922
91.46
201.11
219.88
83.92
99.01
1 Sample size (number) of datasets reporting this compound. 5 Lower 90% confidence bound. 9 The mean of the signed bias values is zero.
2 Arithmetic average. 6 Upper 90% confidence bound. 10 The mean of the magnitude of the bias values
3 Standard deviation. 7 Lower 95% confidence bound. is a non-zero value.
4 Coefficient of variation: (STD/MEAN) x 100. 8 Upper 95% confidence bound.
-------
Table 4-15
Continued
COMPOUND
L095C0NC7
UP95CONC8
MEANB29
STDBIAS3
MEANABS2'10
STDABS
1,2,3 -trimethylvenzene
42.51
46.80
0.00
22.68
15.35
16.62
1,2,4-trimethylbenzene
42.46
45.52
0.00
17.18
12.72
11.47
1,3,5 -trimethylbenzene
42.66
44.87
0.00
12.59
9.53
8.18
1-butane
24.95
27.41
0.00
23.38
14.05
18.63
benzene
72.43
76.89
0.00
14.98
9.35
11.67
cyclohexane
33.83
35.35
0.00
10.91
8.15
7.21
ethane
12.60
13.62
0.00
18.67
12.58
13.73
ethylbenzene
93.36
97.27
0.00
10.23
7.76
6.63
ethylene
10.26
11.12
0.00
19.11
13.03
13.91
propane
20.20
21.47
0.00
15.31
10.60
11.00
proplene
17.05
17.97
0.00
12.93
8.60
9.62
toluene
82.19
86.42
0.00
12.57
8.59
9.13
Total NMOC
934.43
983.12
0.00
12.46
9.23
8.32
m/p-xylene
78.12
84.32
0.00
19.04
12.39
14.41
/7-butane
27.91
29.41
0.00
13.06
8.80
9.61
/?-dccanc
53.94
60.95
0.00
29.28
20.68
20.60
/?-hcxanc
40.13
42.95
0.00
16.95
9.99
13.65
/?-octanc
52.84
55.19
0.00
10.92
8.17
7.20
/7-propylbenzene
56.72
59.57
0.00
12.08
8.62
8.41
o-xylene
43.56
45.34
0.00
10.01
7.17
6.95
ALL
82.47
100.46
0.00
16.28
10.69
12.28
1 Sample size (number) of datasets reporting this compound. 5 Lower 90% confidence bound. 9 The mean of the signed bias values is zero.
2 Arithmetic average. 6 Upper 90% confidence bound. 10 The mean of the magnitude of the bias values
3 Standard deviation. 7 Lower 95% confidence bound. is a non-zero value.
4 Coefficient of variation: (STD/MEAN) x 100. 8 Upper 95% confidence bound.
cah/G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
4-cxviii
-------
Table 4-16
Summary of Statistical Calculations by Laboratory
LAB
N1
MEANBIAS2
STD3
CV4
NORMSTAT5
PROBNORM6
MIN7
Ol8
MEDIAN9
0310
MAX11
1
40
-12.04
6.72
-55.84
0.89
0.00
-27.97
-15.42
-12.89
-9.79
6.57
2
80
-10.60
19.52
-184.25
0.61
0.00
-98.26
-8.39
-5.03
-1.73
21.84
3
40
-5.45
7.49
-137.39
0.81
0.00
-14.47
-9.35
-7.67
-4.00
19.21
4
80
0.63
6.79
1083.91
0.95
0.01
-21.70
-3.00
1.18
5.38
13.15
5
40
4.32
5.57
128.97
0.97
0.54
-10.20
0.79
4.46
6.75
16.98
6
80
16.68
28.20
169.09
0.93
0.00
-68.03
4.49
14.91
32.70
88.10
7
40
2.41
4.88
202.49
0.79
0.00
-20.50
0.56
2.41
5.69
8.70
8
60
2.30
8.70
377.79
0.96
0.10
-14.40
-3.29
1.13
6.53
19.78
9
80
-3.60
11.86
-329.63
0.95
0.01
-46.68
-7.93
-3.07
1.90
33.43
10
36
-2.15
20.40
-947.22
0.96
0.28
-47.06
-10.83
-3.61
9.76
45.65
11
78
-22.02
14.88
-67.60
0.69
0.00
-98.47
-26.22
-21.67
-14.71
7.81
12
60
9.66
11.78
122.02
0.93
0.00
-10.72
-2.05
10.73
19.02
30.57
13
40
-7.32
26.95
-367.98
0.80
0.00
-95.13
-10.57
-1.44
9.12
24.52
14
57
4.35
10.61
244.16
0.95
0.05
-22.54
-0.04
2.56
9.95
25.90
15
57
4.34
7.76
178.71
0.93
0.01
-12.62
-0.65
1.82
8.13
23.62
16
120
2.41
6.66
275.95
0.93
0.00
-23.39
-0.87
3.48
6.35
14.14
17
38
-6.86
7.77
-113.19
0.73
0.00
-40.79
-8.39
-6.03
-3.23
3.75
19
115
1.78
6.23
349.69
0.98
0.28
-15.81
-2.33
1.10
5.49
23.07
1 N = number of data points submitted by the PAMS site. PAMS sites were requested to perform and report replicate determinations (i.e., 40 data points). Some PAMS sites performed multiple sets of replicate analyses on different instruments.
2 MEANBIAS = arithmetic mean of bias values for all compounds reported by the particular PAMS sites.
3 STD = sample standard deviation.
4 CV = coefficient of variation = (STD/MEAN) x 100.
5 NORMSTAT = normality statistic (from SAS, for sample sizes<2000). This is the Shapiro-Wilk "w" statistic which tests the null hypothesis that the data sample comes from a normal data distribution.
6 PROBNORM = probability level for NORMSTAT (from SAS, for samples sizes <2000, probability of a smaller "w" statistic). For the data distribution to be considered NORMAL, the probability level must be 0.05 or larger for a 95% confidence level (accept
the null hypothesis).
7 MEN = minimum value from sample.
8 Q1 = 25th percentile from sample.
9 MEDIAN = 50th percentile from sample.
10 Q3 = 75th percentile from sample.
11 MAX = maximum value from sample.
-------
Table 4-16
Continued
LAB
N1
MEANBIAS2
STD3
CV4
NORMSTAT5
PROBNORM6
MIN7
Ol8
MEDIAN9
0310
MAX11
20
36
-0.55
22.83
-4167.66
0.83
0.00
-86.57
-3.61
1.61
14.19
27.97
22
40
5.55
7.96
143.38
0.95
0.11
-7.20
-0.35
3.99
10.70
22.20
23
40
-13.72
10.69
-77.86
0.93
0.01
-43.47
-22.84
-13.43
-5.29
11.76
24
80
2.45
10.51
429.31
0.89
0.00
-36.13
-1.97
4.86
8.02
21.51
25
36
21.40
10.22
47.75
0.98
0.80
0.02
16.25
21.45
27.85
41.92
26
40
3.19
8.74
274.00
0.95
0.08
-9.91
-3.46
1.44
8.67
26.24
27
40
10.23
14.97
146.34
0.77
0.00
-44.03
6.15
11.83
17.33
37.03
28
40
10.23
6.27
61.26
0.98
0.86
-4.07
5.81
10.00
14.64
21.98
29
40
3.19
6.64
208.42
0.96
0.23
-7.76
-1.26
3.12
8.07
16.27
30
38
-2.06
15.42
-750.21
0.96
0.22
-44.23
-12.05
-4.25
4.68
35.23
31
40
-10.62
4.59
-43.16
0.95
0.08
-22.12
-13.03
-10.23
OO
o
OO
1
-2.70
32
52
1.80
25.96
1442.15
0.75
0.00
-91.82
-1.67
8.14
15.30
40.51
33
80
0.10
11.96
12027.13
0.84
0.00
-15.83
-9.89
1.26
5.86
66.79
34
80
2.82
10.71
380.13
0.73
0.00
-50.92
-0.51
3.74
8.19
17.97
35
23
-8.69
37.65
-433.22
0.93
0.13
-98.47
-17.91
-1.83
11.21
66.43
36
40
-6.71
5.46
-81.37
0.95
0.09
-15.40
-11.24
-6.70
-3.03
4.66
37
36
-6.62
20.92
-334.19
0.76
0.00
-44.61
-31.31
6.33
9.36
15.25
ATT,
35
-0.25
8.64
-3422.71
0.98
0.72
-22.02
-6.71
1.78
4.32
21.40
1 N = number of data points submitted by the PAMS site. PAMS sites were requested to perform and report replicate determinations (i.e., 40 data points). Some PAMS sites performed multiple sets of replicate analyses on different instruments.
2 MEANBIAS = arithmetic mean of bias values for all compounds reported by the particular PAMS sites.
3 STD = sample standard deviation.
4 CV = coefficient of variation = (STD/MEAN) x 100.
5 NORMSTAT = normality statistic (from SAS, for sample sizes<2000). This is the Shapiro-Wilk "w" statistic which tests the null hypothesis that the data sample comes from a normal data distribution.
6 PROBNORM = probability level for NORMSTAT (from SAS, for samples sizes <2000, probability of a smaller "w" statistic). For the data distribution to be considered NORMAL, the probability level must be 0.05 or larger for a 95% confidence level (accept
the null hypothesis).
7 MEN = minimum value from sample.
8 Q1 = 25th percentile from sample.
9 MEDIAN = 50th percentile from sample.
10 Q3 = 75th percentile from sample.
11 MAX = maximum value from sample.
-------
Average Percent Bias per Organization
24
-24 *
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38
Organization Number
Figure 4-52. Average Percent Bias (Considering Bias Positive or Negative)
per Organization for all Compounds
-------
Table 4-17
Summary of Statistical Calculations by Laboratory for Absolute Bias
LAB
N1
MEANBIAS2
STD3
CV4
NORMSTAT5
PROBNORM6
MIN7
Ql8
MEDIAN9
Q310
MAX11
1
40
12.68
5.38
42.45
0.93
0.02
0.21
9.79
12.89
15.42
27.97
2
80
11.34
19.09
168.37
0.56
0.00
0.04
2.10
5.20
8.66
98.26
3
40
8.06
4.46
55.31
0.96
0.18
0.13
4.74
8.02
9.84
19.21
4
80
5.31
4.23
79.58
0.88
0.00
0.10
2.53
4.64
7.02
21.70
5
40
5.54
4.33
78.17
0.90
0.00
0.02
2.07
4.59
8.08
16.98
6
80
24.93
21.16
84.86
0.88
0.00
0.03
8.59
17.44
34.91
88.10
7
40
3.99
3.66
91.55
0.79
0.00
0.10
1.40
3.23
5.95
20.50
8
60
7.02
5.57
79.41
0.90
0.00
0.02
2.12
5.55
11.26
19.78
9
80
8.54
8.94
104.74
0.78
0.00
0.02
2.84
4.95
11.40
46.68
10
36
15.25
13.49
88.45
0.80
0.00
0.09
5.97
10.02
20.44
47.06
11
78
22.30
14.45
64.78
0.65
0.00
3.23
14.71
21.67
26.22
98.47
12
60
12.85
8.11
63.15
0.93
0.00
0.79
5.51
11.08
19.02
30.57
13
40
17.16
21.90
127.66
0.69
0.00
0.11
4.96
9.92
18.32
95.13
14
57
8.28
7.88
95.27
0.85
0.00
0.04
1.75
4.63
13.67
25.90
15
57
6.15
6.39
103.97
0.83
0.00
0.04
1.07
4.03
9.43
23.62
16
12
5.54
4.39
79.17
0.89
0.00
0.01
2.36
4.37
7.84
23.39
17
38
7.23
7.42
102.52
0.66
0.00
0.23
3.39
6.03
8.39
40.79
19
11
4.77
4.37
91.70
0.84
0.00
0.04
1.48
3.84
6.62
23.07
analyses on
0.05 or larger
1 N = number of data points submitted by the PAMS site. PAMS sites were requested to perform and report replicate determinations (i.e., 40 data points). Some PAMS sites performed multiple sets of replicate
different instruments.
2 MEANBIAS = arithmetic mean of bias values for all compounds reported by the particular PAMS sites.
3 STD = sample standard deviation.
4 CV = coefficient of variation = (STD/MEAN) x 100.
5 NORMSTAT = normality statistic (from SAS, for sample sizes<2000). This is the Shapiro-Wilk "w" statistic which tests the null hypothesis that the data sample comes from a normal data distribution.
6 PROBNORM = probability level for NORMSTAT (from SAS, for samples sizes <2000, probability of a smaller "w" statistic). For the data distribution to be considered NORMAL, the probability level must be
for a 95% confidence level (accept the null hypothesis).
7 MEN = minimum value from sample.
8 Q1 = 25th percentile from sample.
9 MEDIAN = 50th percentile from sample.
10 Q3 = 75th percentile from sample.
11 MAX = maximum value from sample.
-------
Table 4-17
Continued
LAB
N1
MEANBIAS2
STD3
CV4
NORMSTAT5
PROBNORM6
MIN7
Ql8
MEDIAN9
Q310
MAX11
20
36
14.59
17.39
119.16
0.76
0.00
0.05
2.90
7.60
20.07
86.57
22
40
7.29
6.36
87.31
0.88
0.00
0.05
2.16
5.21
10.70
22.20
23
40
14.31
9.86
68.92
0.89
0.00
1.29
5.51
14.98
22.84
43.47
24
80
8.46
6.63
78.35
0.86
0.00
0.03
3.91
6.72
11.55
36.13
25
36
21.40
10.22
47.75
0.98
0.80
0.02
16.25
21.45
27.85
41.92
26
40
6.95
6.11
87.81
0.88
0.00
0.09
2.20
6.43
9.73
26.24
27
40
14.58
10.64
72.98
0.85
0.00
1.04
6.61
12.33
17.79
44.03
28
40
10.45
5.89
56.39
0.96
0.32
0.25
5.81
10.00
14.64
21.98
29
40
5.86
4.40
75.08
0.92
0.01
0.05
2.18
5.31
8.07
16.27
30
38
11.56
10.24
88.59
0.89
0.00
0.19
4.60
8.04
17.02
44.23
31
40
10.62
4.59
43.16
0.95
0.08
2.70
8.08
10.23
13.03
22.12
32
52
16.74
19.78
118.14
0.67
0.00
0.43
6.00
11.95
18.64
91.82
33
80
8.77
8.08
92.11
0.64
0.00
0.60
4.45
7.51
11.29
66.79
34
80
7.07
8.49
120.11
0.65
0.00
0.21
1.78
4.93
9.30
50.92
35
23
25.80
28.30
109.69
0.81
0.00
0.27
3.30
12.98
50.67
98.47
36
40
7.39
4.48
60.66
0.95
0.08
0.09
4.37
6.70
11.24
15.40
37
36
16.99
13.45
79.14
0.83
0.00
1.38
6.41
11.20
31.31
44.61
ALL
35
11.31
5.86
51.80
0.89
0.00
3.99
7.02
8.77
14.59
25.80
1 N = number of data points submitted by the PAMS site. PAMS sites were requested to perform and report replicate determinations (i.e., 40 data points). Some PAMS sites performed multiple sets of replicate analyses on
different instruments.
2 MEANBIAS = arithmetic mean of bias values for all compounds reported by the particular PAMS sites.
3 STD = sample standard deviation.
4 CV = coefficient of variation = (STD/MEAN) x 100.
5 NORMSTAT = normality statistic (from SAS, for sample sizes<2000). This is the Shapiro-Wilk "w" statistic which tests the null hypothesis that the data sample comes from a normal data distribution.
6 PROBNORM = probability level for NORMSTAT (from SAS, for samples sizes <2000, probability of a smaller "w" statistic). For the data distribution to be considered NORMAL, the probability level must be 0.05 or
larger for a 95% confidence level (accept the null hypothesis).
7 MEN = minimum value from sample.
8 Q1 = 25th percentile from sample.
9 MEDIAN = 50th percentile from sample.
10 Q3 = 75th percentile from sample.
11 MAX = maximum value from sample.
-------
Average Percent Bias per Organization
26
24
22
CO
20
CD
in
18
-t—>
c
0
16
o
8
<
6
4
2
0
0
8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38
Organization Number
Figure 4-53. Average Percent Bias per PAMS Site for all Compounds
-------
5.0
REFERENCES
1. Compendium Method TO-12. The Determination of Non-Methane Organic Compounds
(NMOC) in Ambient Air Using Cryogenic Preconcentration and Direct Flame Ionization
Detection (PDFID). U. S. Environmental Protection Agency, Compendium of Methods
for the Determination of Toxic Organic Compounds in Ambient Air, EPA-600/4-89/017,
Research Triangle Park, NC, June 1988.
2. U. S. Environmental Protection Agency. Code of Federal Regulations. Draft Title 40,
Part 58. Enhanced Ozone Monitoring Regulations. Washington, D. C. Office of the
Federal Register. August 23, 1991.
3. U. S. Environmental Protection Agency. Photochemical Assessment Monitoring Stations
Implementation Manual: Network Design and Siting for PAMS. Office of Air Quality
Planning and Standards, Research Triangle Park, NC. EPA-454/B-93-051. March, 1994.
4. U. S. Environmental Protection Agency. Technical Assistance Document for Sampling
and Analysis of Ozone Precursors. Office of Air Quality Planning and Standards,
Research Triangle Park, NC. EPA-600/8-91-215. October 20, 1994.
can\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
5-cxxv
-------
Appendices
The appendices of this report contain the raw data which has been summarized in tables
included in the body of the report. These appendices are very large and are not available
in Adobe format.
can\G:\USER\SHARE\PAMS\REPORT\ONEBOOK.WPD
5-cxxvi
------- |