United States Office of Research and EPA/600/R-98/114
Environmental Protection Development August 1998
Agency Washington, D.C. 20460
4>EPA Environmental Technology
Verification Report
Portable Gas Chromatograph/
Surface Acoustic Wave Detector
Electronic Sensor Technology
4100 Vapor Detector
^H^M^HHHHIti JtaHHiH^Hi^MlP jtjf^Mh J^^^^^lr ^UJ^^^^^MMM^ i^mm^m^^^
• v I *^y I? I ^
-------
EPA/600/R-98/114
August 1998
Environmental Technology
Verification Report
Portable Gas Chromatograph/
Surface Acoustic Wave Detector
Electronic Sensor Technology
4100 Vapor Detector
By
Amy B. Dindal
Charles K. Bayne, Ph.D.
Roger A. Jenkins, Ph.D.
Oak Ridge National Laboratory
Oak Ridge Tennessee 37831-6120
Stephen Billets, Ph.D.
Eric N. Koglin
U.S. Environmental Protection Agency
Environmental Sciences Division
National Exposure Research Laboratory
Las Vegas, Nevada 89193-3478
This demonstration was conducted in cooperation with
U.S. Department of Energy
David Bottrell, Project Officer
Cloverleaf Building, 19901 Germantown Road
Germantown, Maryland 20874
Superfund Innovative Technology
Evaluation Program
-------
Notice
The U.S. Environmental Protection Agency (EPA), through its Office of Research and Development
(ORD), and the U.S. Department of Energy's Environmental Management (EM) Program, funded and
managed, through Interagency Agreement No. DW89937854 with Oak Ridge National Laboratory, the
verification effort described herein. This report has been peer and administratively reviewed 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 of a specific product.
-------
l
•*
ltDt
Uffizj
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, D.C. 20460
LT1/
ENVIRONMENTAL TECHNOLOGY VERIFICATION PROGRAM
VERIFICATION STATEMENT
TECHNOLOGY TYPE: POLYCHLORINATED BIPHENYL (PCB) FIELD ANALYTICAL
TECHNIQUES
APPLICATION: MEASUREMENT OF PCBs IN SOILS AND SOLVENT EXTRACTS
TECHNOLOGY NAME: 4100 VAPOR DETECTOR
COMPANY: ELECTRONIC SENSOR TECHNOLOGY
ADDRESS: 1077 BUSINESS CENTER CIRCLE
NEWBURY PARK, CA 91320
PHONE: (805) 480-1994
The U.S. Environmental Protection Agency (EPA) has created a program to facilitate the deployment of innovative
technologies through performance verification and information dissemination. The goal of the Environmental Technology
Verification (ETV) Program is to further environmental protection by substantially accelerating the acceptance and use
of improved and more cost effective technologies. The ETV Program is intended to assist and inform those involved in
the design, distribution, permitting, and purchase of environmental technologies. This document summarizes the results
of a demonstration of the Electronic Sensor Technology (EST) 4100 Vapor Detector.
PROGRAM OPERATION
The EPA, in partnership with recognized testing organizations, objectively and systematically evaluates the performance
of innovative technologies. Together, with the full participation of the technology developer, they develop plans, conduct
tests, collect and analyze data, and report findings. The evaluations are conducted according to a rigorous demonstration
plan and established protocols for quality assurance. EPA's National Exposure Research Laboratory, which conducts
demonstrations of field characterization and monitoring technologies, with the support of the U.S. Department of
Energy's (DOE) Environmental Management (EM) program, selected Oak Ridge National Laboratory as the testing
organization for the performance verification of polychlorinated biphenyl (PCB) field analytical techniques.
DEMONSTRATION DESCRIPTION
In July 1997, the performance of six PCB field analytical techniques was determined under field conditions. Each
technology was independently evaluated by comparing field analysis results to those obtained using approved reference
methods. Performance evaluation (PE) samples also were used to assess independently the accuracy and comparability
of each technology.
The demonstration was designed to detect and measure PCBs in soil and solvent extracts. The demonstration was conducted
at the Oak Ridge National Laboratory (ORNL) in Oak Ridge, Tennessee from July 22 through July 29. The study was
conducted under two environmental conditions. The first site was outdoors, with naturally fluctuating temperature and
relative humidity conditions. The second site was inside a controlled environmental chamber, with generally cooler
temperatures and lower relative humidities. Multiple soil types, collected from sites in Ohio, Kentucky, and Tennessee, were
analyzed in this study. Solutions of PCBs were also analyzed to simulate extracted surface wipe samples. The results of the
soil and extract analyses conducted under field conditions by the technology were compared with results from analyses of
EPA-VS-SCM-14 The accompanying notice is an integral part of this verification statement August 1998
iii
-------
homogeneous replicate samples conducted by conventional EPA SW-846 methodology in an approved reference laboratory.
Details of the demonstration, including a data summary and discussion of results, may be found in the report entitled
Environmental Technology Verification Report: Portable Gas Chromatograph/Surface Acoustic Wave Detector,
Electronic Sensor Technology 4100 Vapor Detector, EPA/600/R-98/114.
TECHNOLOGY DESCRIPTION
A handheld, portable (35 Ibs.) chromatography system equipped with a non-specific Surface Acoustic Wave (SAW)
detector is used to speciate and quantify PCBs. The SAW detector is an integrating mass detector (micro-balance) with
the ability to quantify chromatographic peaks, with peak widths measured in milliseconds. Measurement speed makes
the instrument well suited to rapid screening of soil samples. Early separation of those soil samples below the regulatory
level from those which require laboratory validation by GC/MS reduces the cost associated with site characterization and
monitoring.
A sampling pump and loop trap are used to sample and inject analyte into a GC capillary column. Speciation is based
upon retention time measurements using a temperature programmed DB-5 column. Quantification is based upon the
frequency shift produced by the PCB congeners as they exit the GC column. By focusing the effluent onto a specific area
on the surface of a temperature controlled piezoelectric crystal, high sensitivity is achieved with a 10 second analysis
time. The 4100 Vapor Detector is able to screen selectively and quantify PCB levels of Aroclors in soil and flyash.
VERIFICATION OF PERFORMANCE
The following performance characteristics of the 4100 Vapor Detector were observed:
Detection limits: EPA defines the method detection limit (MDL) as the minimum concentration of a substance that can
be measured and reported with 99% confidence that the analyte concentration is greater than zero. Because there was
a significant "site effect" inherent to the PE samples, separate MDLs were calculated for both the outdoor and chamber
conditions. The MDL was calculated to be 26 ppm under outdoor conditions and 62 ppm under chamber conditions.
Throughput: Throughput was 5 to 6 samples/hour under outdoor conditions and 10 samples/hour under chamber
conditions. This rate included sample preparation and analysis.
Ease of Use: Two operators were used for the demonstration due to the number of samples and working conditions, but
the technology can be run by a single operator. Operators generally require several hours of training and should have a
basic knowledge of gas chromatographic techniques. These methods should be used by, or under the supervision of,
analysts experienced in the use of sampling techniques and gas chromatography.
Completeness: The 4100 generated results for all 232 PCB samples for a completeness of 100%.
Blank results: EST did not specify a method detection limit prior to the demonstration, therefore, any PCB concentration
that was detected was considered real. PCBs were detected in all of the soil blanks, resulting in 100% false positive
results. PCBs were also detected in 3 of 8 of the extract blanks or 38% false positive results. The 4100 reported 5% false
negative results for soils and no false negative results for extracts.
Precision: The overall precision, based on average relative standard deviations (RSDs), was 87% for soil samples and
65% for extract samples. The 4100 was imprecise compared to the reference laboratory's precision (21% for soils and
14% for extracts).
Accuracy: Accuracy was assessed using PE soil and extract samples. The study was conducted under two experimental
conditions to detect and control for "site effects." The data showed that the 4100 exhibited a significant site effect, and
the results were generally biased high. The overall accuracy, based on average percent recoveries, was 177% (outdoor
EPA-VS-SCM-14 The accompanying notice is an integral part of this verification statement August 1998
iv
-------
site) and 631% (chamber) for PE soil samples. For the extract samples, the results indicated a high bias (267% recovery)
on the lower concentration samples and a low bias on the higher concentration samples (54% recovery).
Comparability: The demonstration showed that the 4100 generated data that exhibited low correlation to the reference
laboratory data. The coefficient of determination (R2) which is a measure of the degree of correlation between the
reference laboratory and the 4100 data was 0.177 when all soil samples (0 to 700 ppm) were considered. For the
concentration range from 0 to 125 ppm, the R2 value was 0.115. Most of the percent difference values were greater than
100%, when the 4100 results were compared directly with the reference laboratory results. The comparability of the
extract samples also exhibited low correlation.
Regulatory Decision-making: One objective of this demonstration was to assess the technology's ability to perform at
regulatory decision-making levels for PCBs, specifically 50 ppm for soils and 100 (jg/100cm2 for surface wipes. For PE
and environmental soil samples in the range of 40 to 60 ppm, the precision was low (72% RSD) and the accuracy was
variably biased with both high and low recoveries (an average recovery of 132%). For extract samples representing
surface wipe sample concentrations of 100 (jg/100cm2 and 1000 (jg/100cm2 (assuming a 100 cm2 wipe sample),
measurements were also imprecise (65% RSD) and indicated a high bias (161% recovery).
Data quality levels: The overall performance of the EST 4100 Vapor Detector was characterized as biased and
imprecise. EST is working to improve the performance of the methodology.
The results of the demonstration show that certain cautions should be considered when using this technology for PCB
analysis due to its bias and imprecision. This technology should be employed in well-defined applications for PCB
analysis, and only in conjunction with a stringent quality assurance plan. As with any technology selection, the user must
determine if the technology is appropriate for the application and the project data quality objectives. For more information
on this and other verified technologies, visit the ETV web site at http://www.epa.gov/etv.
Gary J. Foley, Ph.D.
Director
National Exposure Research Laboratory
Office of Research and Development
NOTICE: EPA verifications are based on an evaluation of technology performance under specific, predetermined criteria and the
appropriate quality assurance procedures. EPA makes no expressed or implied warranties as to the performance of the technology and
does not certify that a technology will always, under circumstances other than those tested, operate at the levels verified. The end user
is solely responsible for complying with any and all applicable Federal, State, and Local requirements.
EPA-VS-SCM-14 The accompanying notice is an integral part of this verification statement August 1998
V
-------
-------
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the nation's
natural resources. The National Exposure Research Laboratory (NERL) is EPA's center for the
investigation of technical and management approaches for identifying and quantifying risks to human health
and the environment. NERL's research goals are to (1) develop and evaluate technologies for the
characterization and monitoring of air, soil, and water; (2) support regulatory and policy decisions; and (3)
provide the science support needed to ensure effective implementation of environmental regulations and
strategies.
EPA created the Environmental Technology Verification (ETV) Program to facilitate the deployment of
innovative technologies through performance verification and information dissemination. The goal of the
ETV Program is to further environmental protection by substantially accelerating the acceptance and use of
improved and cost-effective technologies. The ETV Program is intended to assist and inform those involved
in the design, distribution, permitting, and purchase of environmental technologies. This program is
administered by NERL's Environmental Sciences Division in Las Vegas, Nevada.
The U.S. Department of Energy's (DOE) Environmental Management (EM) program has entered into
active partnership with EPA, providing cooperative technical management and funding support. DOE EM
realizes that its goals for rapid and cost effective cleanup hinges on the deployment of innovative
environmental characterization and monitoring technologies. To this end, DOE EM shares the goals and
objectives of the ETV.
Candidate technologies for these programs originate from the private sector and must be commercially
ready. Through the ETV Program, developers are given the opportunity to conduct rigorous demonstrations
of their technologies under realistic field conditions. By completing the evaluation and distributing the
results, EPA establishes a baseline for acceptance and use of these technologies.
Gary J. Foley, Ph.D.
Director
National Exposure Research Laboratory
Office of Research and Development
vn
-------
Vlll
-------
Abstract
In July 1997, the U.S. Environmental Protection Agency (EPA) conducted a demonstration of
polychlorinated biphenyl (PCB) field analytical techniques. The purpose of this demonstration was to
evaluate field analytical technologies capable of detecting and quantifying PCBs in soils and solvent
extracts. The fundamental objectives of this demonstration were (1) to obtain technology performance
information using environmental and quality control samples, (2) to determine how comparable the
developer field analytical results were with conventional reference laboratory results, and (3) to report on
the logistical operation of the technology. The demonstration design was subjected to extensive review and
comment by EPA's National Exposure Research Laboratory (NERL) Environmental Sciences Division in
Las Vegas, Nevada; Oak Ridge National Laboratory (ORNL); EPA Regional Offices; the U.S. Department
of Energy (DOE); and the technology developers.
The demonstration study was conducted at ORNL under two sets of environmental conditions. The first site
was outdoors, with naturally variable temperature and relative humidity conditions typical of eastern
Tennessee in the summer. A second site was located inside a controlled environmental chamber having
lower, and relatively stable, temperature and relative humidity conditions. The test samples analyzed during
this demonstration were performance evaluation soil, environmental soil, and extract samples. Actual
environmental soil samples, collected from sites in Ohio, Kentucky, and Tennessee, were analyzed, and
ranged in concentration from 0.1 to 700 parts per million (ppm). Extract samples were used to simulate
surface wipe samples, and were evaluated at concentrations ranging from 0 to 100 (jg/mL. The reference
laboratory method used to evaluate the comparability of data was EPA SW-846 Method 8081.
The field analytical technologies tested in this demonstration were the L2000 PCB/Chloride Analyzer
(Dexsil Corporation), the PCB Immunoassay Kit (Hach Company), the 4100 Vapor Detector (Electronic
Sensor Technology), and three immunoassay kits: D TECH, EnviroGard, and RaPID Assay System
(Strategic Diagnostics Inc.). The purpose of an Environmental Technology Verification Report (ETVR) is
to document the demonstration activities, present demonstration data, and verify the performance of the
technology. This ETVR presents information regarding the performance of Electronic Sensor Technology's
4100 Vapor Detector. Separate ETVRs have been published for the other technologies demonstrated.
The 4100 Vapor Detector is a handheld, portable (35-lb) chromatography system equipped with a
nonspecific surface acoustic wave (SAW) detector designed to speciate and quantify PCBs. The SAW
detector is an integrating mass detector (microbalance) with the ability to quantify chromatographic peaks,
with peak widths measured in milliseconds. A sampling pump and loop trap are used to sample and inject
analyte into the gas chromatography capillary column. Speciation is based upon retention time
measurements using a temperature-programmed DB-5 column. Quantification is based upon the frequency
shift produced by the PCB congeners as they exit the column. By focusing the effluent onto a specific area
ix
-------
on the surface of a temperature-controlled piezoelectric crystal, high sensitivity is achieved with a 10-s
analysis time. Because of the short analysis time, Aroclor speciation is limited to low, medium, and high
classifications, based on the percentage of chlorine within each Aroclor.
The 4100's quantitative results were based on initial calibrations. The method detection limit (MDL) is
often defined as the minimum concentration of a substance that can be measured and reported with 99%
confidence that the analyte concentration is greater than zero. The field-based MDLs were site-specific and
were calculated to be 26 ppm under outdoor conditions and 62 ppm under chamber conditions. Electronic
Sensor Technology did not specify an MDL prior to the demonstration. The study was conducted under
two experimental conditions to detect and control "site effects" (i.e., differences in performance due to
environmental conditions). In general, the 4100's results for soils were biased high and exhibited a
significant site effect (177% recovery outdoors and 631% recovery in the chamber). For the extract
samples, the results indicated a high bias (267% recovery) on the lower-concentration samples and a low
bias on the higher-concentration samples (54% recovery). The overall precision, based on relative standard
deviation (RSD), for soil samples was 87% compared with the reference laboratory's 21%. The precision
for the extract samples was also low at 65% RSD. Comparability, based on coefficients of determination
(R2), was 0.177 for all soil samples (0 to 700 ppm), where an R2 of 1.0 denotes perfect correlation. Most of
the percent difference values were greater than 100% when the 4100 results were compared directly with
the reference laboratory results. The 4100 also exhibited low correlation for the extract samples.
During the demonstration the 4100 was found to be light, easily transportable, and rugged. The system is
shock mounted into a field-portable fiberglass shipping case that can be checked as airplane baggage. The
4100 was simple to operate in the field, requiring less than 1 h for two operators to set up initially and
prepare for sample analysis. Operators generally require several hours of training and should have a basic
knowledge of gas chromatographic techniques. These methods should be used by, or under the supervision
of, analysts experienced in the use of sampling techniques and gas chromatography. Once operational, the
4100 had a throughput of 5 to 10 samples per hour during the demonstration. Measurement speed makes
the instrument well suited to rapid screening of soil samples. Early separation of those soil samples below
the regulatory level from those that require laboratory validation by gas chromatography/mass
spectroscopy reduces the cost associated with site characterization and monitoring. The overall
performance of the Electronic Sensor Technology 4100 Vapor Detector was characterized as biased,
imprecise, and having significant "site effects." EST is working to improve the performance of the
methodology for PCB analysis.
-------
Table of Contents
Notice ii
Verification Statement iii
Foreword vii
Abstract ix
List of Figures xv
List of Tables xvii
List of Abbreviations and Acronyms xix
Acknowledgments xxiii
Section 1 Introduction 1
Technology Verification Process 2
Needs Identification and Technology Selection 2
Demonstration Planning and Implementation 3
Report Preparation 3
Information Distribution 3
Demonstration Purpose 4
Section 2 Technology Description 5
Objective 5
System Overview 5
Sample Preparation and Analysis Procedures 6
Direct Thermal Extraction/Analysis 6
Liquid Extraction and Injection/Analysis 6
Instrument Analysis Checklist 7
Aroclor Pattern Recognition/Quantification 8
Calculations 8
Section 3 Site Description and Demonstration Design 9
Objective 9
Demonstration Site and Description 9
xi
-------
Site Name and Location 9
Site History 9
Site Characteristics 10
Experimental Design 10
Environmental Conditions during Demonstration 13
Sample Descriptions 13
Performance Evaluation Materials 13
Environmental Soil Samples 14
Extract Samples 14
Sampling Plan 14
Sample Collection 14
Sample Preparation, Labeling, and Distribution 14
Predemonstration Study 16
Predemonstration Sample Preparation 16
Predemonstration Results 17
Deviations from the Demonstration Plan 17
Section 4 Reference Laboratory Analytical Results and Evaluation 19
Objective and Approach 19
Reference Laboratory Selection 19
Reference Laboratory Method 20
Calibration 20
Sample Quantification 20
Sample Receipt, Handling, and Holding Times 21
Quality Control Results 21
Objective 21
Continuing Calibration Verification Standard Results 21
Instrument and Method Blank Results 22
Surrogate Spike Results 22
Laboratory Control Sample Results 22
Matrix Spike Results 23
Conclusions of the Quality Control Results 23
Data Review and Validation 23
Objective 23
Corrected Results 24
Suspect Results 24
Data Assessment 25
Objective 25
Precision 25
Performance Evaluation Samples 25
Environmental Soil Samples 26
Extract Samples 28
Accuracy 28
Performance Evaluation Soil Samples 28
xii
-------
Extract Samples 29
Representativeness 30
Completeness 30
Comparability 31
Summary of Observations 31
Section 5 Technology Performance and Evaluation 33
Objective and Approach 33
Data Assessment 33
Precision 33
Performance Evaluation Samples 33
Environmental Soil Samples 34
Extract Samples 35
Precision Summary 36
Accuracy 36
Performance Evaluation Soil Samples 37
Extract Samples 37
Accuracy Summary 38
False Positive/False Negative Results 39
Representativeness 39
Completeness 40
Comparability 40
Summary of PARCC Observations 42
Regulatory Decision-Making Applicability 44
Additional Performance Factors 45
Detection Limits 45
Sample Throughput 45
Cost Assessment 45
4100 Vapor Detector Costs 46
Reference Laboratory Costs 48
Cost Assessment Summary 48
General Observations 48
Performance Summary 49
Section 6 Technology Update and Representative Applications 51
Objective 51
Technology Update 51
Representative Applications 51
Data Quality Objective Example 52
Section 7 References 53
Appendix A Description of Environmental Soil Samples 55
xin
-------
Appendix B Characterization of Environmental Soil Samples 59
Appendix C Temperature and Relative Humidity Conditions 63
Appendix D EST's PCB Technology Demonstration Sample Data 69
Appendix E Data Quality Objective Example 79
Disclaimer 81
Background and Problem Statement 81
DQO Goals 81
Use of Technology Performance Information to Implement the Decision Rule 82
4100 Vapor Detector Accuracy 83
Determining the Number of Samples 84
Determining the Action Level 85
Alternative FP Parameter 87
xiv
-------
List of Figures
3-1. Schematic map of ORNL, indicating the demonstration area 11
5-1. The 4100 Vapor Detector's results versus the certified PCB concentration for PE soil samples. ... 37
5-2. Paired PCB measurements for 4100 and reference measurements 41
5-3. Range of percent difference values for the comparison of the 4100 Vapor Detector soil sample results
with the reference laboratory results 41
5-4. Paired PCB extract measurements for the 4100 Vapor Detector and reference
laboratory 43
5-5. Range of percent difference values for the comparison of EST extract sample results with the reference
laboratory results 43
C-l. Summary of temperature conditions for outdoor site 66
C-2. Summary of relative humidity conditions for the outdoor site 66
C-3. Summary of temperature conditions for chamber site 67
C-4. Summary of relative humidity conditions for chamber site 67
E-l. A line fitted to the average concentration of outdoor PE samples 83
E-2. Decision performance curve for PCB drum example 87
xv
-------
-------
List of Tables
3-1. Summary of experimental design by sample type 12
3-2. Summary of the 4100's predemonstration results 17
4-1. Suspect measurements within the reference laboratory data 24
4-2. Precision of the reference laboratory for PE soil samples 26
4-3. Precision of the reference laboratory for environmental soil samples 27
4-4. Precision of the reference laboratory for extract samples 28
4-5. Accuracy of the reference laboratory for PE soil samples 29
4-6. Accuracy of the reference laboratory for extract samples 30
4-7. Summary of the reference laboratory performance 31
5-1. Precision of the 4100 Vapor Detector for PE soil samples 34
5-2. Precision of the 4100 Vapor Detector for environmental soil samples 35
5-3. Precision of the 4100 Vapor Detector for extract samples 36
5-4. Overall precision of the 4100 Vapor Detector for all sample types 36
5-5. Accuracy of the 4100 Vapor Detector for PE soils samples 38
5-6. Accuracy of the 4100 Vapor Detector for extract samples 39
5-7. Overall accuracy of the 4100 Vapor Detector for all sample types 39
5-8. Comparison of the reference laboratory's suspect data to the 4100 Vapor Detector data 42
5-9. Summary of PARCC observations for the 4100 Vapor Detector 44
5-10. Performance of the 4100 Vapor Detector for soil samples between 40 and 60 ppm 45
5-11. Estimated analytical costs for PCB soil samples 47
5-12. Performance summary for the 4100 Vapor Detector 50
A-l. Summary of soil sample descriptions 57
B-l. Summary of environmental soil characterization 61
C-l. Average temperature and relative humidity conditions during testing periods 65
D-l. EST's 4100 Vapor Detector PCB technology demonstration soil sample data 71
D-2. EST's 4100 Vapor Detector technology demonstration extract sample data 77
D-3. Corrected reference laboratory data 78
xvn
-------
-------
List of Abbreviations and Acronyms
AL action level
ANOVA analysis of variance
ASTM American Society for Testing and Materials
BHC benzenehexachloride
C concentration at which the false positive error rate is specified
CASD Chemical and Analytical Sciences Division
CCV continuing calibration verification standard
CSCT Consortium for Site Characterization Technology
DCB decachlorobiphenyl
DOE U. S. Department of Energy
DQO data quality objective
EM Environmental Management
EPA U. S. Environmental Protection Agency
ERA Environmental Resource Associates
EST Electronic Sensor Technology
ETTP East Tennessee Technology Park
ETV Environmental Technology Verification Program
ETVR Environmental Technology Verification Report
EvTEC Environmental Technology Evaluation Center
fn false negative result
FN false negative decision error rate
fp false positive result
FP false positive decision error rate
GC gas chromatography
HEPA high-efficiency particulate air
ID identifier
INEL Idaho National Engineering Laboratory
LCS laboratory control sample
xix
-------
LMER Lockheed Martin Energy Research
LMES Lockheed Martin Energy Systems
LV Las Vegas
MDL method detection limit
MS matrix spike
MSB matrix spike duplicate
n number of samples
NERL National Exposure Research Laboratory
NRC Nuclear Regulatory Commission
OPDDT open path direct desorption tube
ORD EPA's Office of Research and Development
ORNL Oak Ridge National Laboratory
ORO Oak Ridge Operations
OTD open tubular desorption
PARCC precision, accuracy, representativeness, completeness, comparability
PCB polychlorinated biphenyl
PE performance evaluation
ppb parts per billion
ppm parts per million; equivalent units: mg/kg for soils and (jg/mL for extracts
Pr probability
QA quality assurance
QC quality control
R2 coefficient of determination
RDL reporting detection limit
RH relative humidity
RFD request for disposal
RPD relative percent difference
RSD percent relative standard deviation
RT regulatory threshold
S2 variance for the measurement
SARA Superfund Amendments and Reauthorization Act of 1986
SAW surface acoustic wave
SD standard deviation
xx
-------
SDI Strategic Diagnostics Inc.
SITE Superfund Innovative Technology Evaluation
SMO sample management office
SOP standard operating procedure
SSM synthetic soil matrix
TCMX tetrachloro-m-xylene
TSCA Toxic Substance Control Act
Z]_p the (1 - p)th percentile for the standard normal distribution
% D percent difference
xxi
-------
-------
Acknowledgments
The authors wish to acknowledge the support of all those who helped plan and conduct the demonstration,
analyze the data, and prepare this report. In particular we recognize the technical expertise of Mitchell
Erickson (Environmental Measurements Laboratory), Viorica Lopez-Avila (Midwest Research Institute),
and Robert F. O'Brien (Pacific Northwest National Laboratory) who were peer reviewers of this report; for
internal peer review, Stacy Barshick (ORNL); for technical support during the demonstration, Todd Skeen
and Ralph Ilgner (ORNL); for site safety and health support, Kim Thomas, Marilyn Hanner, and Fred
Smith (ORNL); for administrative support, Betty Maestas and Linda Plemmons (ORNL); for sample
collection support, Wade Hollinger, Charlotte Schaefer, and Arlin Yeager (LMES), and Mike Rudacille
and W. T. Wright (EET Corporation); for preliminary soil characterization support, Frank Gardner, John
Zutman, and Bob Schlosser (ORNL, Grand Junction, CO); for sample management support, Angie
McGee, Suzanne Johnson, and Mary Lane Moore (LMES); for providing performance evaluations
samples, Michael Wilson (EPA's Office of Solid Waste and Emergency Response's Analytical Operations
and Data Quality Center); and for technical guidance and project management of the demonstration, David
Garden, Marty Atkins, and Regina Chung (DOE's Oak Ridge Operations Office), David Bottrell (DOE,
Headquarters), Deana Crumbling (EPA's Technology Innovation Office), and Dr. Stephen Billets, Gary
Robertson, and Eric Koglin (EPA's National Exposure Research Laboratory, Las Vegas, Nevada). The
authors also acknowledge the participation of Electronic Sensor Technology, in particular, Ed Staples and
David McGuire, who performed the analyses during the demonstration.
For more information on the PCB Field Analytical Technology Demonstration, contact:
Eric N. Koglin
Project Technical Leader
Environmental Protection Agency
National Exposure Research Laboratory
P. O. Box 93478
Las Vegas, Nevada 89193-3478
(702) 798-2432
For more information on EST's 4100 Vapor Detector, contact:
Ed Staples
Electronic Sensor Technology
1077 Business Center Circle
NewburyPark, CA91320
(805)480-1994
xxiii
-------
Section 1
Introduction
The performance evaluation of innovative and alternative environmental technologies is an integral part of
the U.S. Environmental Protection Agency's (EPA's) mission. Early efforts focused on evaluating
technologies that supported the implementation of the Clean Air and Clean Water Acts. In 1987, the
Agency began to evaluate the cost and performance of remediation and monitoring technologies under the
Superfund Innovative Technology Evaluation (SITE) program. This was in response to the mandate in the
Superfund Amendments and Reauthorization Act (SARA) of 1986. In 1990, the U.S. Technology Policy
was announced. This policy placed a renewed emphasis on "making the best use of technology in achieving
the national goals of improved quality of life for all Americans, continued economic growth, and national
security." In the spirit of the Technology Policy, the Agency began to direct a portion of its resources
toward the promotion, recognition, acceptance, and use of U.S.-developed innovative environmental
technologies both domestically and abroad.
The Environmental Technology Verification (ETV) Program was created by the Agency to facilitate the
deployment of innovative technologies through performance verification and information dissemination.
The goal of the ETV Program is to further environmental protection by substantially accelerating the
acceptance and use of improved and cost-effective technologies. The ETV Program is intended to assist and
inform those involved in the design, distribution, permitting, and purchase of environmental technologies.
The ETV Program capitalizes upon and applies the lessons that were learned in the implementation of the
SITE Program to the verification of twelve categories of environmental technology: Drinking Water
Systems, Pollution Prevention/Waste Treatment, Pollution Prevention/ Innovative Coatings and Coatings
Equipment, Indoor Air Products, Air Pollution Control, Advanced Monitoring Systems, EvTEC (an
independent, private-sector approach), Wet Weather Flow Technologies, Pollution Prevention/Metal
Finishing, Source Water Protection Technologies, Site Characterization and Monitoring Technology [also
referred to as the Consortium for Site Characterization Technology (CSCT)], and Climate Change
Technologies. The performance verification contained in this report was based on the data collected during
a demonstration of polychlorinated biphenyl (PCB) field analytical technologies. The demonstration was
administered by CSCT.
For each pilot, EPA utilizes the expertise of partner "verification organizations" to design efficient
procedures for conducting performance tests of environmental technologies. To date, EPA has partnered
with federal laboratories and state, university, and private sector entities. Verification organizations oversee
and report verification activities based on testing and quality assurance protocols developed with input
from all major stakeholder/customer groups associated with the technology area.
In July 1997, CSCT, in cooperation with the U.S. Department of Energy's (DOE's) Environmental
Management (EM) Program, conducted a demonstration to verify the performance of six field analytical
1
-------
technologies for PCBs: the L2000 PCB/Chloride Analyzer (Dexsil Corporation), the PCB Immunoassay
Kit (Hach Company), the 4100 Vapor Detector [Electronic Sensor Technology (EST)], and three
immunoassay kits from Strategic Diagnostics Inc.: D TECH, EnviroGard, and RaPID Assay System. This
environmental technology verification report (ETVR) presents the results of the demonstration study for
one PCB field analytical technology, EST's 4100 Vapor Detector. Separate ETVRs have been published
for the other five technologies.
Technology Verification Process
The technology verification process is intended to serve as a template for conducting technology
demonstrations that will generate high-quality data that EPA can use to verify technology performance.
Four key steps are inherent in the process:
• Needs identification and technology selection
• Demonstration planning and implementation
• Report preparation
• Information distribution
Needs Identification and Technology Selection
The first aspect of the technology verification process is to determine technology needs of EPA and the
regulated community. EPA, DOE, the U.S. Department of Defense, industry, and state agencies are asked
to identify technology needs and interest in a technology. Once a technology need is established, a search is
conducted to identify suitable technologies that will address this need. The technology search and
identification process consists of reviewing responses to Commerce Business Daily announcements,
searches of industry and trade publications, attendance at related conferences, and leads from technology
developers. Characterization and monitoring technologies are evaluated against the following criteria:
• meets user needs;
• may be used in the field or in a mobile laboratory;
• is applicable to a variety of environmentally impacted sites;
• has high potential for resolving problems for which current methods are unsatisfactory;
• is cost competitive with current methods;
• performs better than current methods in areas such as data quality, sample preparation, or
analytical turnaround time;
-------
• uses techniques that are easier and safer than current methods; and
• is a commercially available, field-ready technology.
Demonstration Planning and Implementation
After a technology has been selected, EPA, the verification organization, and the developer agree to the
responsibilities for conducting the demonstration and evaluating the technology. The following tasks are
undertaken at this time:
• identifying demonstration sites that will provide the appropriate physical or chemical
environment, including contaminated media;
• identifying and defining the roles of demonstration participants, observers, and reviewers;
• determining logistical and support requirements (for example, field equipment, power and
water sources, mobile laboratory, communications network);
• arranging analytical and sampling support; and
• preparing and implementing a demonstration plan that addresses the experimental design,
sampling design, quality assurance/quality control (QA/QC), health and safety
considerations, scheduling of field and laboratory operations, data analysis procedures,
and reporting requirements.
Report Preparation
Innovative technologies are evaluated independently and, when possible, against conventional technologies.
The field technologies are operated by the developers in the presence of independent technology observers.
The technology observers are provided by EPA or a third-party group. Demonstration data are used to
evaluate the capabilities, limitations, and field applications of each technology. Following the
demonstration, all raw and reduced data used to evaluate each technology are compiled into a technology
evaluation report, which is mandated by EPA as a record of the demonstration. A data summary and
detailed evaluation of each technology are published in an ETVR.
Information Distribution
The goal of the information distribution strategy is to ensure that ETVRs are readily available to interested
parties through traditional data distribution pathways, such as printed documents. Documents are also
available on the World Wide Web through the ETV Web site (http://www.epa.gov/etv) and through a Web
site supported by the EPA Office of Solid Waste and Emergency Response's Technology Innovation Office
(http://CL U-in. com).
-------
Demonstration Purpose
The purpose of this demonstration was to obtain performance information for PCB field analytical
technologies, to compare the results with conventional fixed-laboratory results, and to provide supplemental
information (e.g., cost, sample throughput, and training requirements) regarding the operation of the
technology. The demonstration was conducted under two climatic conditions. One set of activities was
conducted outdoors, with naturally fluctuating temperatures and relative humidity conditions. A second set
was conducted in a controlled environmental facility, with lower, relatively stable temperatures and relative
humidities. Multiple soil types, collected from sites in Ohio, Kentucky, and Tennessee, were used in this
study. PCB soil concentrations ranged from approximately 0.1 to 700 parts per million (ppm). Developers
also analyzed 24 solutions of known PCB concentration that were used to simulate extracted wipe samples.
The extract samples ranged in concentration from 0 to 100 (jg/mL.
-------
Section 2
Technology Description
Objective
The objective of this section is to describe the technology being demonstrated, including the operating
principles underlying the technology and the overall approach to its use. The information provided here is
excerpted from that provided by the developer. Performance characteristics described in this section are
specified by the developer, which may or may not be substantiated by the data presented in Section 5.
System Overview
The EST 4100 Vapor Detector is a handheld, portable (35-lb) gas chromatography (GC) system equipped
with a nonspecific surface acoustic wave (SAW) detector. The unit can be used to speciate and quantify
PCBs. A sampling pump and loop trap are used to sample and inject analyte into a short gas
chromatography (GC) capillary column. The analyte mixture travels through the temperature-programmed
column and is separated into its components according to conventional chromatographic principles. The
effluent is focused onto a specific area on the surface of a temperature-controlled piezoelectric crystal that
acts as the SAW detector. Speciation of the analytes is based upon retention-time measurements using a
temperature-programmed J&W Durabond DB-5® capillary column (22.5 in. long, 0.25-mm internal
diameter, 0.25-/^m phase thickness). Quantification is based upon the frequency shift due to mass loading
of the SAW detector produced by the analytes as they exit the GC column. The system has the ability to
quantify chromatographic peaks at the picogram level, with peak widths measured in milliseconds.
Measurement speed and accuracy make the instrument well suited for rapid screening of soil samples.
Early segregation through rapid field screening of those soil samples below the regulatory level from those
that require laboratory analysis reduces the cost associated with site characterization and monitoring.
Because the GC column is short (22.5 in.), other environmental components may co-elute with the PCB
target analytes and be sensed by the nonspecific SAW detector. Any such compounds detected may be
misidentified and quantified as a PCB. If the quantification level is above a given criterion threshold, the
developer recommends that the soil sample be laboratory tested and the SAW/GC screening measurement
validated. Impurities from contaminants within the instrument or inlet train desorption tubing may interfere
with the analysis. Contamination by carryover can also occur whenever high-concentration and low-
concentration samples are analyzed sequentially. To minimize these types of interferences, use of the
system as a screening tool mandates that acceptably low instrument blank values be obtained before and
after all positive measurements.
-------
Sample Preparation and Analysis Procedures
Two procedures for extracting PCBs from soil matrices can be used with the 4100 Vapor Detector. Both
procedures have been tested on the following Aroclors: 1016, 1221, 1232, 1242, 1248, 1254, 1260, and
1262. The first procedure uses a direct thermal desorption of the PCBs in the soil matrix packed in a heated
tube (referred to as the open-path direct desorption tube [OPDDT]). This method is best suited for
relatively clean soil samples with PCB levels below 250 ppb. The second procedure involves the liquid
extraction of soil using a mixture of hexane, water, and methanol. A small amount of the hexane layer is
subsequently injected into the 4100 Vapor Detector inlet. The second procedure is best suited to testing soil
with contamination levels of 250 ppb or higher because of the sample dilution inherent in the method.
Direct Thermal Extraction/Analysis
The 4100 Vapor Detector inlet sample port is glass-lined stainless steel for sampling vapors directly into
the instrument. Total extraction from soil is performed using an open heated glass tube fitted with a glass-
to-Luer adapter attached directly to the inlet of the instrument. After loading tube with approximately 250
mg of soil, the Luer adapter is attached to one end of the sample tube. The sample tube is attached to the
Luer inlet fitting of the 4100. The heater jacket, preheated to 200°C, is slid over the sample tube, and
thermal desorption of the soil is immediately initiated, with desorption duration set to 30 s. The desorbed
PCB vapors are swept onto the head of the GC column, separated, and quantified by the SAW detector.
The thermal desorption/analysis is repeated for 30-s periods at 1-min intervals until analyte concentration
readings are less then 10% of initial sample values. The concentration mass, in nanograms, for each sample
measurement, N{, as well as the total of all sample measurements, 7VT are recorded. The sample tube packed
with soil is weighed. The weight of the empty tube is subtracted, and the result is designated as WSOIL in
grams. If the soil contents of the tube are not to be measured immediately, the ends of the glass tube are
sealed with slip-on septa covers.
Calibration is performed using a syringe to inject calibration standard solutions directly into the OPDDT.
Note that QA measurements require GC validation using only standards certified by an independent
laboratory. All spiking solutions, prior to their use in soil recovery analyses or calibration by direct
injection, must first be validated by GC measurement.
Blank samples must be run before and after each analytical run to monitor for background levels or
carryover. The analysis of blank samples is continued until the PCB levels are below preset minimums.
Each sample tube is weighed and prescreened (desorbed and analyzed) before being loaded with soil. The
instrument should be used with the SAW/GC method and instrument settings for which the calibration was
performed. Use of any other method requires the generation of a new calibration curve. The operator must
save all chromatograms, including blanks and calibration checks performed with liquid standards.
Liquid Extraction and Injection/Analysis
This method is well suited to the analysis of soils with high concentrations of PCBs. First the PCBs are
extracted from the soil using a mixture of hexane (1.0 mL), methanol (1.5 mL), and water (0.4 mL).
A weighed amount of soil (0.3-0.5 g) is shaken until the soil is well dispersed. The slurry is then allowed to
stand until the hexane layer is clearly separated and floats on top of methanol-water layer with soil
sediment resting on the bottom of the vial. Approximately 0.25 mL of the hexane is removed. The extract is
-------
filtered through a disposable pipette packed with glass wool and is transferred into a clean vial, which is
then capped with a septum.
Desorption of the PCBs from the extract solution is performed using an open-path thermal desorption tube
packed with glass wool. The tube is fitted with a glass-to-Luer adapter that attaches directly to the inlet of
the instrument.
With the heater jacket removed and the extraction tube at room temperature, an aliquot of the hexane
extract (0.2 to 10 (iL) is placed into the tube. Analysis cycles with the 4100 are initiated to remove the
volatile components from the hexane solvent. This is continued until liquid can no longer be seen in the
glass tube. Then, the heater jacket, preheated to 200°C, is slid over the sample tube, and thermal
desorption of the PCB-containing residue is immediately initiated, with desorption duration set to 30 s. The
desorbed PCB vapors are swept onto the head of the GC column, separated, and quantified by the SAW
detector. The thermal desorption/analysis is repeated for 30-s periods at 1-min intervals until analyte
concentration readings are less then 10% of initial sample values. The PCB mass, in nanograms, for each
sample measurement, N{, and the total of all sample measurements, 7VT are recorded. Calibration standards
are injected directly into the open-path desorption tube. It should be noted that for all of the samples
analyzed for the verification study, the liquid extraction method was used.
Instrument Analysis Checklist
The following items must be checked prior sample analysis:
• If the instrument has been previously calibrated in the laboratory, perform a single mid-
level calibration check for each analyte. If the value of the check is within 30% of the
laboratory value, then the response factor is confirmed. If the value is greater than 30%,
then the instrument must be recalibrated.
Check instrument status. Measure the instrument sample flow using the mass flow meter.
Record the sample flow and enter the value in the Peak File software dialog screen under
sample flow in cc/min units.
Run an instrument blank. Verify that the background is below 10 ppb for any compound in
the peak file. The blank should be a solvent injected into an empty desorption tube.
• Create a calibration standard solution. Fill a 4-mL vial with an appropriate amount of
standard solution and an appropriate amount of solute so that a concentration (measured in
nanograms per microliter) is achieved that is mid-level to the desired measurement range.
Seal the vial with a new septum lid.
To determine the instrument response factor, SF (measured in hertz per picogram), inject a
liquid with a known standard into the desorption tube. The instrument reading, Fm, in
measurement units of frequency (hertz) and the total amount of analyte injected, Ma
(measured in picograms) defines the response factor:
-------
SF = (2-1)
M
• Confirm the retention-time windows for each component to be analyzed. Make three
injections of the component and calculate the standard deviation of the retention time of
each component. For each analyte, the average retention time and response factor are
calculated and saved in the peak recognition file.
Aroclor Pattern Recognition/Quantification
PCB Aroclor mixtures typically contain 40 or more congeners. The system software provides the operator
with the capability to use either the sum of peaks over a retention time range or the sum of selected peaks
as the basis for calibration. A single average response factor for the sum of the peaks within the mixture is
used to calculate the concentration of the Aroclor mixture.
Commercial Aroclor mixtures of PCB isomers are commonly found at environmental sites; their
composition and vapor signature can readily be recognized by a trained operator. Once the peak
identification files for the Aroclor mixtures have been created, an unknown sample can be identified as
containing PCBs and the PCBs can be quantified. Data logging to Excel spreadsheets using different peak-
recognition file patterns for the raw data provides documentation and archival of all 4100 Vapor Detector
measurements.
Calculations
Windows 95, SAW/GC system software (Version 4.0), and Excel and are required to operate the system,
log data, and provide measurement documentation. Three calibration options are provided with the system
software. The operator may select individual compound peaks and calibrate based upon the measured
signal in hertz and the standard input in nanograms. Alternatively, the operator may use either the total area
of all peaks over a specified range of retention times or the sum of a set of "tagged" peaks specified in a
calibration file to determine a response factor in terms of a standard input. Soil contamination is expressed
in either ppm (mg/kg), ppb (ng/kg), or ppt (ng/kg). To calculate soil contamination, the following
calculation is performed:
- N
"SOIL "SOIL (2-2)
For liquid extractions the result from Eq. 2-2 must be multiplied by the dilution ratio (the total amount of
hexane solution divided by the amount of solution extract injected).
-------
Section 3
Site Description and Demonstration Design
Objective
This section describes the demonstration site, the experimental design for the verification test, and the
sampling plan (sample types analyzed and the collection and preparation strategies). Included in this section
are the results from the predemonstration study and a description of the deviations made from the original
demonstration design.
Demonstration Site Description
Site Name and Location
The demonstration of PCB field analytical technologies was conducted at Oak Ridge National Laboratory
(ORNL) in Oak Ridge, Tennessee. PCB-contaminated soils from three DOE sites (Oak Ridge; Paducah,
Kentucky; and Piketon, Ohio) were used in this demonstration. The soil samples used in this study were
brought to the demonstration testing location for evaluation of the field analytical technologies.
Site History
Oak Ridge is located in the Tennessee River Valley, 25 miles northwest of Knoxville. Three DOE facilities
are located in Oak Ridge: ORNL, the Oak Ridge Y-12 Plant, and East Tennessee Technology Park
(ETTP). Chemical processing and warhead component production have occurred at the Y-12 Plant, and
ETTP is a former gaseous diffusion uranium enrichment plant. At both facilities, industrial processing
associated with nuclear weapons production has resulted in the production of millions of kilograms of
PCB-contaminated soils. Two other DOE facilities—the Paducah plant in Paducah, Kentucky, and the
Portsmouth plant in Piketon, Ohio—are also gaseous diffusion facilities with a history of PCB
contamination. During the remediation of the PCB-contaminated areas at the three DOE sites, soils were
excavated from the ground where the PCB contamination occurred, packaged in containers ranging in size
from 55-gal to 110-gal drums, and stored as PCB waste. Samples from these repositories—referred to as
"Oak Ridge," "Portsmouth," and "Paducah" samples in this report—were used in this demonstration.
In Oak Ridge, excavation activities occurred between 1991 and 1995. The Oak Ridge samples were
comprised of PCB-contaminated soils from both Y-12 and ETTP. Five different sources of PCB
contamination resulted in soil excavations from various dikes, drainage ditches, and catch basins. Some of
the soils are EPA-listed hazardous waste due to the presence of other contaminants (e.g., diesel fuels).
A population of over 5000 drums containing PCB-contaminated soils was generated from 1986 to 1987
during the remediation of the East Drainage Ditch at the Portsmouth Gaseous Diffusion Plant. The ditch
was reported to have three primary sources of potential contamination: (1) treated effluent from a
radioactive liquid treatment facility, (2) runoff from a biodegradation plot where waste oil and sludge were
disposed of, and (3) storm sewer discharges. In addition, waste oil was reportedly used for weed control in
-------
the ditch. Aside from PCB contamination, no other major hazardous contaminants were detected in these
soils. Therefore, no EPA hazardous waste codes are assigned to this waste.
Twenty-nine drums of PCB-contaminated soils from the Paducah plant were generated as part of a spill
cleanup activity at an organic waste storage area (C-746-R). The waste is considered a listed hazardous
waste for spent solvents (EPA hazardous waste code F001) because it is known to contain
trichloroethylene. Other volatile organic compounds, such as xylene, dichlorobenzene, and cresol, were also
detected in the preliminary analyses of some of the Paducah samples.
Site Characteristics
PCB-contaminated environmental soil samples from Oak Ridge, Portsmouth, and Paducah were collected
from waste containers at storage repositories at ETTP and Paducah. Many of the soils contained interfering
compounds such as oils, fuels, and other chlorinated compounds (e.g., trichloroethylene). Specific sample
descriptions of the environmental soils used in this demonstration are given in Appendix A. In addition,
each sample was characterized in terms of its soil composition, pH, and total organic carbon content. Those
results are summarized in Appendix B.
Field demonstration activities occurred at two sites at ORNL: a natural outdoor environment (the outdoor
site) and inside a controlled environmental atmosphere chamber (the chamber site). Figure 3-1 shows a
schematic map of a section of ORNL indicating the demonstration area where the outdoor field activities
occurred. Generally, the average summer temperature in eastern Tennessee is 75.6°F, with July and August
temperatures averaging 79.1 °F and 76.8°F, respectively. Average temperatures during the testing periods
ranged from 79 to 85 °F, as shown in Appendix C. Studies were also conducted inside a controlled
environmental atmosphere chamber, hereafter referred to as the "chamber," located in Building 5507 at
ORNL. Demonstration studies inside the chamber were used to evaluate performance under environmental
conditions that were markedly different from the ambient outdoor conditions at the time of the test. Average
temperatures in the chamber during the testing periods ranged from 55 to 70°F. The controlled
experimental atmosphere facility consists of a room-size walk-in chamber 10 ft wide and 12 ft long with air
processing equipment to control temperature and humidity. The chamber is equipped with an environmental
control system, including reverse osmosis water purification that supplies the chamber humidity control
system. High efficiency particulate air (HEPA) and activated charcoal filters are installed for recirculation
and building exhaust filtration.
Experimental Design
The analytical challenge with PCB analysis is to quantify a complex mixture that may or may not resemble
the original commercial product (i.e., Aroclor) due to environmental aging, and to report the result as a
single number [1]. The primary objective of the verification test was to compare the performance of the
field technology to laboratory-based measurements. Often, verification tests involve a direct one-to-one
comparison of results from field-acquired samples. However, because sample heterogeneity can preclude
replicate field or laboratory comparison, accuracy and precision data must often be derived from the
analysis of QC and performance evaluation (PE) samples. In this study, replicates of all three sample types
(QC, PE, and environmental soil) were analyzed. The ability to use environmental soils in the verification
10
-------
Building
**»
Figure 3-1. Schematic map of ORNL, indicating the demonstration area.
test was made possible because the samples, collected from drums containing PCB-contaminated soils,
could be thoroughly homogenized and characterized prior to the demonstration. This facet of the design,
allowing additional precision data to be obtained on actual field-acquired samples, provided an added
performance factor in the verification test.
Another objective of this demonstration was to evaluate the field technology's capability to support
regulatory compliance decisions. For field methods to be used in these decisions, the technology must be
capable of informing the user, with known precision and accuracy, that soil concentrations are greater than
or less than 50 ppm, and that wipe samples are greater than or less than 100 (jg/100 cm2 [2]. The samples
selected for analysis in the demonstration study were chosen with this objective in mind.
The experimental design is summarized in Table 3-1. This design was approved by all participants prior to
the start of the demonstration study. In total, the developers analyzed 208 soil samples, 104 each at both
locations (outdoors and chamber). The 104 soil samples comprised 68 environmental samples (17 unique
environmental samples prepared in quadruplicate) ranging in PCB concentration from 0.1 to 700 ppm and
36 PE soils (9 unique PE samples in quadruplicate) ranging in PCB concentration from 0 to 50 ppm. To
determine the impact of different environmental conditions on the technology's performance, each batch of
104 samples contained five sets of quadruplicate soil samples from DOE's Paducah site. These were
analyzed under both sets of environmental conditions (i.e., outdoor and chamber conditions). For the
developers participating in the extract sample portion (i.e., simulated wipe samples) of the demonstration,
12 extracts, ranging in concentration from 0 to 100 (ig/mL, were analyzed in each
11
-------
location (chamber and outdoors). All samples were analyzed without prior knowledge of sample type or
concentration and were analyzed in a randomized order that was unique for each developer.
Table 3-1. Summary of experimental design by sample type
Concentration
Range
Sample ID a
Outdoor Site
Chamber Site
Total #
Samples
Analyzed
PE Materials
0
2.0 ppm
2.0 ppm
5.0 ppm
10.9 ppm
20.0 ppm
49.8 ppm
50.0 ppm
50.0 ppm
126
118
124
120
122
119
125
121
123
226
218
224
220
222
219
225
221
223
8
8
8
8
8
8
8
8
8
Environmental Soils
0.1-2.0 ppm
2. 1-20.0 ppm
20. 1-50.0 ppm
50. 1-700.0 ppm
0
10 |ig/mL
100 |ig/mL
Grand Total
101, 107, 108, 109, 113, 114
102,103,104,115
111,116
105,106,110,112,117
201,202,206
203,207,212,213
204,208,209,214,215
205,210,211,216,217
36
32
28
40
Extracts
129b/132c
127/130
128/131
116
229/232
227/230
228/231
116
8
8
8
232 d
a Each sample ID was analyzed in quadruplicate.
b Extract prepared in iso-octane for Dexsil and the reference laboratory.
c Extract prepared in methanol for Electronic Sensor Technology, Strategic Diagnostics Inc., and the
reference laboratory.
d All samples were analyzed in random order.
12
-------
Environmental Conditions during Demonstration
As mentioned above, field activities were conducted both outdoors under natural environmental conditions
and indoors in a controlled environmental atmosphere chamber to evaluate the effect of environmental
conditions on technology performance. The weather outside was relatively uncomfortable during the July
demonstration, with highs approaching 100°F and 90% relative humidity (RH). Daily average
temperatures were around 85 °F with 70% RH. While outside, the developers set up canopies to provide
shade and protection from frequent late afternoon thundershowers.
In the indoor chamber tests, conditions were initially set to 55 °F and 25% RH. An independent check of the
conditions inside the chamber revealed that the temperature was closer to 68 °F with a 38% RH on the first
day of testing. A maintenance crew was called in to address the inconsistencies between the set and actual
conditions. By the middle of the third day of testing, the chamber was operating properly at 55 °F and 50%
RH.
Appendix C contains a summary of the environmental conditions (temperature and relative humidity)
during the demonstration. The EST team worked outdoors July 22, 23, and 24, 1997, and in the chamber
on July 24 and 25, 1997.
Sample Descriptions
PCBs (C12H10.XC1X) are a class of compounds that are chlorine-substituted linked benzene rings. There are
209 possible PCB compounds (also known as congeners). PCBs were commercially produced as complex
mixtures beginning in 1929 for use in transformers, capacitors, paints, pesticides, and inks [1]. Monsanto
Corporation marketed products that were mixtures of 20 to 60 PCB congeners under the trade name
Aroclor. Aroclor mixtures are identified by a number (e.g., Aroclor 1260) that represents the mixture's
chlorine composition as a percentage (e.g., 60%).
Performance Evaluation Materials
Samples of Tennessee reference soil [3] served as the blanks. Preprepared certified PE samples were
obtained from Environmental Resource Associates (ERA) of Arvada, Colorado, and the Analytical
Operations and Data Quality Center of EPA's Office of Solid Waste and Emergency Response. The soils
purchased from ERA had been prepared using ERA's semi volatile blank soil matrix. This matrix was a
topsoil that had been dried, sieved, and homogenized. Particle size was approximately 60 mesh. The soil
was approximately 40% clay. The samples acquired from EPA's Analytical Operations and Data Quality
Center had been prepared using contaminated soils from various sites around the country in the following
manner: The original soils had been homogenized and diluted with a synthetic soil matrix (SSM). The SSM
had a known matrix of 6% gravel, 31% sand, and 43% silt/clay; the remaining 20% was topsoil. The
dilution of the original soils was performed by mixing known amounts of contaminated soil with the SSM
in a blender for no less than 12 h. The samples were also spiked with target pesticides (a, P, A, and 8-
BHC, methoxychlor, and endrin ketone) to introduce some compounds that were likely to be present in an
actual environmental soil. The hydrocarbon background from the original sample and the spiked pesticides
produced a challenging matrix. The PE soils required no additional preparation by ORNL and were split
for the developer and reference laboratory analyses as received.
13
-------
Environmental Soil Samples
As noted in the site description above, PCB-contaminated environmental soil samples from Oak Ridge,
Portsmouth, and Paducah were used in this demonstration. The soils were contaminated with PCBs as the
result of spills and industrial processing activities at the various DOE facilities. Originally, the
contaminated soils were excavated from dikes, drainage ditches, catch basins, and organic waste storage
areas. The excavated soils were then packaged into waste containers and stored at the repositories in ETTP
and Paducah in anticipation of disposal by incineration. The environmental soil samples used in this study
were collected from these waste containers. Many of the soils contained interfering compounds such as oils,
fuels, and other chlorinated compounds, while some contained multiple Aroclors. For more information on
sampling locations and sample characteristics (soil composition, pH, and total organic carbon content),
refer to Appendices A and B, respectively.
Extract Samples
Traditionally, the amount of PCBs on a contaminated surface is determined by wiping the surface with a
cotton pad saturated with hexane. The pad is then taken to the laboratory, extracted with additional hexane,
and analyzed by gas chromatography. Unlike soil samples, which can be more readily
homogenized and divided, equivalent wipe samples (i.e., contaminated surfaces or post- wipe pads) were
not easily obtainable. Therefore, interference-free solutions of PCBs were analyzed to simulate an extracted
surface wipe pad. Extract sample analyses provided evaluation data that relied primarily on the
technology's performance rather than on elements critical to the entire method (i.e., sample collection and
preparation). Because different developers required the extract samples prepared in different solvents (e.g.,
methanol and iso-octane), the reference laboratory analyzed sets of extracts in both solvents. EST analyzed
extracts prepared in methanol. A total of 12 extracts were analyzed per site; these consisted of four
replicates each of a blank and two concentration levels (10 and 100 (jg/mL).
Sampling Plan
Sample Collection
Environmental soil samples were collected from April 17 through May 7, 1997. Portsmouth and Oak Ridge
Reservation soils were collected from either storage boxes or 55-gal drums stored at ETTP. Briefly, the
following procedure was used to collect the soil samples. Approximately 30 Ib of soil were collected from
the top of the drum or B-25 box using a scoop and placed in a plastic bag. The soil was sifted to remove
rocks and other large debris, then poured into a plastic-lined 5-gal container. All samples were subjected to
radiological screening and were determined to be nonradioactive. Similarly, soil samples were collected
from 55-gal drums stored at Paducah and shipped to ORNL in lined 5-gal containers.
Sample Preparation, Labeling, and Distribution
Aliquots of several of the environmental soils were analyzed and determined to be heterogeneous in PCB
concentration. Because this is unsatisfactory for accurately comparing the performance of the field
technology with the laboratory-based method, the environmental soils had to be homogenized prior to
sample distribution. Each Portsmouth and Oak Ridge environmental soil sample was homogenized by first
placing approximately 1500 g of soil in a glass Pyrex dish. The dish was then placed in a large oven set at
35 °C, with the exhaust and blower fans turned on to circulate the air. After drying overnight, the soil was
14
-------
pulverized using a conventional blender and sieved using a 9-mesh screen (2 mm particle size). Last, the
soil was thoroughly mixed using a spatula. A comparison of dried and undried soils showed that a minimal
amount of PCBs (< 20%) was lost due to sample drying, making this procedure suitable for use in the
preparation of the soil samples. The Paducah samples, because of their sandy characteristics, only required
the sieving and mixing preparation steps. Extract sample preparation involved making solutions of PCBs in
methanol and iso-octane at two concentration levels (10 and 100 (jg/mL). Multiple aliquots of each sample
were analyzed using the analytical procedure described below to confirm the homogeneity of the samples
with respect to PCB concentration.
To provide the developers with soils contaminated at higher concentrations of PCBs, some of the
environmental soils (those labeled with an "S" in Appendix B) were spiked with additional PCBs. Spiked
soils samples were prepared after the soil was first dried in a 35 °C oven overnight. The dry soil was
ground using a conventional blender and sieved through a 9-mesh screen (2 mm particle size).
Approximately 1500 g of the sieved soil were spiked with a diethyl ether solution of PCBs at the desired
concentration. The fortified soil was agitated using a mechanical shaker and then allowed to air-dry in a
laboratory hood overnight. A minimum of four aliquots were analyzed using the analytical procedure
described below to confirm the homogeneity of the soil with regard to the PCB concentration.
The environmental soils were characterized at ORNL prior to the demonstration study. The procedure used
to confirm the homogeneity of the soil samples entailed the extraction of 3 to 5 g of soil in a mixture of
solvents (1 mL water, 4 mL methanol, and 5 mL hexane). After the soil/solvent mixture was agitated by a
mechanical shaker, the hexane layer was removed and an aliquot was diluted for analysis. The hexane
extract was analyzed on a Hewlett Packard 6890 gas chromatograph equipped with an electron capture
detector and autosampler. The method used was a slightly modified version of EPA's SW-846 dual-column
Method 8081 [4].
After analysis confirming homogeneity, the samples were split into jars for distribution. Each 4-oz sample
jar contained approximately 20 g of soil. Four replicate splits of each soil sample were prepared for each
developer. The samples were randomized in two fashions. First, the order in which the filled jars were
distributed was randomized, such that the same developer did not always receive the first jar filled for a
given sample set. Second, the order of analysis was randomized so that each developer analyzed the same
set of samples, but in a different order. The extract samples were split into 10-mL aliquots and placed into
2-oz jars. The extracts were stored in the refrigerator (at <4°C) until released to the developers. Each
sample jar had three labels: (1) developer order number; (2) sample identifier number; and (3) a PCB
warning label. The developer order number corresponded to the order in which the developer was required
to analyze the samples (e.g., EST 1001 through EST 1116). The sample identifier number was in the
format of "xxxyzz," where "xxx" was the three-digit sample ID (e.g., 101) listed in Table 3-1, "y" was the
replicate (e.g., 1 to 4), and "zz" was the aliquot order of each replicate (e.g., 01 to 11). For example,
sample identifier 101101 corresponded to sample ID "101" (an Oak Ridge soil from RFD 40022, drum
02), "1" corresponded to the first replicate from that sample, and "01" corresponded to the first jar filled in
that series.
Once the samples were prepared, they were stored at a central sample distribution center. During the
demonstration study, developers were sent to the distribution center to pick up their samples. Samples were
15
-------
distributed sequentially in batches of 12 to ensure that samples were analyzed in the order specified.
Completion of chain-of-custody forms and scanning of bar code labels documented sample transfer
activities. Some of the developers received information regarding the samples prior to analysis. This was
provided to simulate the type of information that would be available during actual field testing. EST did not
request any sample information. The developers returned the unused portions of the samples with the
analytical results to the distribution center when testing was completed. The sample bar codes were scanned
upon return to document sample throughput time.
Three complete sets of extra samples, called archive samples, were available for distribution in case the
integrity of a sample was compromised. Very few (<5) archive samples were utilized over the course of the
demonstration.
Predemonstration Study
Ideally, environmental soil samples are sent to the developers prior to the demonstration study to allow
them the opportunity to analyze representative samples in advance of the verification test. This gives
developers the opportunity to refine and calibrate their technologies and revise their operating procedures
on the basis of the predemonstration study results. The predemonstration study results can also be used as
an indication that the selected technologies are of the appropriate level of maturity to participate in the
demonstration study.
According to ORNL regulations, however, one of two conditions must exist in order to ship environmental
soils that were once classified as mixed hazardous waste. First, the recipient—in this case, the developer's
facilities—must have proper Nuclear Regulatory Commission (NRC) licensing to receive and analyze
radiological materials. Second, the soils must be certified as entirely free of radioactivity, beyond the no-
rad certification issued from radiological screening tests based on ORNL standards. Because none of the
developers had proper NRC licensing and proving that the soils were entirely free of radioactivity was
prohibitive, spiked samples of Tennessee reference soil were used for the predemonstration study. The
developers had an opportunity to evaluate the Tennessee reference soils spiked with PCBs at concentrations
similar to what would be used in the demonstration study. The developers also analyzed two performance
evaluation samples and one solvent extract. The reference laboratory analyzed the same set of samples,
which included two extracts samples, prepared in the two solvents (methanol and iso-octane) requested by
the developers.
Predemonstration Sample Preparation
Two soil samples were prepared by ORNL using Tennessee reference soil [3]. The soil was a Captina silt
loam from Roane County, Tennessee, that was slightly acidic (pH ~5) and low in organic carbons (~ 1.5%).
The soil composition was 7.7% sand, 29.8% clay, and 62.5% silt. To prepare a spiked sample, the soil was
first ground either using a mortar and pestle or a conventional blender. The soil was then sieved through a
16-mesh screen (1 mm particle size). Approximately 500 g of the sieved soil was spiked with a diethyl ether
solution of PCBs at the desired concentration. The soil was agitated using a mechanical shaker, then
allowed to air-dry overnight in a laboratory hood. A minimum of five aliquots were analyzed by gas
chromatography using electron capture detection. The PCB concentration of the spiked samples was
determined to be homogeneous. The remaining two soil samples used in the predemonstration study were
performance evaluation materials acquired from ERA and EPA (see the section "Performance Evaluation
16
-------
Materials" above). In addition, a solvent extract was prepared by ORNL to simulate an extracted surface
wipe sample. The extracts were prepared in two different solvents (iso-octane and methanol) to
accommodate developer requests.
Predemonstration Results
The predemonstration samples were sent to the developers and the reference laboratory on June 2, 1997.
Predemonstration results were received by June 26, 1997. Table 3-2 summarizes the 4100's results for the
predemonstration samples. Acceptable results for three of the five performance evaluation samples
indicated that the 4100 was ready for field evaluation.
Table 3-2. Summary of the 4100's predemonstration results
Sample Description
30 ppm of Aroclor 1242
20 ppm of Aroclor 1260
1 1 ppm of Aroclor 1260
50 ppm of Aroclor 1254
blank
4100 result
(ppm)
21.7
16.2
1.3
7.4
0
Certified
Concentration
(ppm)
30.2
21.9
11
50
0
Acceptance
Limits a
(ppm)
12-45
12-27
4-12
20-63
0
a Acceptance limits provided by supplier of performance evaluation materials.
Deviations from the Demonstration Plan
A few deviations from the demonstration plan occurred. In Appendix B of the technology demonstration
plan [5], the reference laboratory's procedure states that no more than 10 samples will be analyzed with
each analytical batch (excluding blanks, standards, QC samples, and dilutions). The analytical batch is also
stated as 10 samples in the Quality Assurance Project Plan of the demonstration plan. The reference
laboratory actually analyzed 20 samples per analytical batch. Because a 20-sample batch is recommended
in SW-846 Method 8081, this deviation was deemed acceptable.
Table 5 of the demonstration plan [5] delineates the environmental soils according to concentration. The
classification was based on a preliminary analysis of the soils at ORNL. Table 3-1 of this report arranges
the concentrations as characterized by the reference laboratory. The reference laboratory determined that
five sample sets (sample IDs 102, 105, 110, 111, and 210) were in the next highest concentration range,
differing from what was originally outlined in the demonstration plan. Also, the highest concentration
determined by the reference laboratory was 700 ppm, while the preliminary analysis at ORNL found the
highest concentration to be 500 ppm.
During the demonstration study, the EST team made several modifications to the procedure described in the
technology demonstration plan [5].
17
-------
• Reduced the sample size from ~1.5 g to ~0.5g.
If the extract appeared oily or dirty, the extract was diluted by a factor of 4 prior to
analysis.
A centrifuge step was added to the liquid extraction procedure.
• The slide heater jacket temperature was reduced from 200°C to 180°C.
The time to heat the slide heater jacket was reduced from 30 s to 10s.
These changes were applied to compensate for what the EST referred to as " high oil concentrations''
encountered in the samples.
18
-------
Section 4
Reference Laboratory Analytical Results and Evaluation
Objective and Approach
The purpose of this section is to present the evaluation of the PCB data generated by the reference
laboratory. Evaluation of the results from the analysis of PE, environmental soil, and extract samples was
based on precision, accuracy, representativeness, completeness, comparability (PARCC) parameters [6].
This section describes how the analytical data generated by the reference laboratory were used to establish
a baseline performance for PCB analysis.
Reference Laboratory Selection
The Oak Ridge Sample Management Office (SMO) has been tasked by DOE Oak Ridge Operations (DOE-
ORO) with maintaining a list of qualified laboratories to provide analytical services. The technology
demonstration plan [5] contains the SMO's standard operating procedures (SOPs) for identifying,
qualifying, and selecting analytical laboratories. Laboratories are qualified as acceptable analytical service
providers for the SMO by meeting specific requirements. These requirements include providing pertinent
documentation (such as QA and chemical hygiene plans), acceptance of the documents by the SMO, and
satisfactory performance on an on-site prequalification audit of laboratory operations. All laboratory
qualifications are approved by a laboratory selection board, composed of the SMO operations manager and
appointees from all prime contractors that conduct business with the SMO.
All of the qualified laboratories were invited to bid on the demonstration study sample analysis. The lowest-
cost bidder was LAS Laboratories, in Las Vegas, Nevada. A readiness review conducted by ORNL and the
SMO confirmed the selection of LAS as the reference laboratory. Acceptance of the reference laboratory
was finalized by satisfactory performance in the predemonstration study (see Table 3-2). The SMO
contracted LAS to provide full data packages for the demonstration study sample analyses within 30 days
of sample shipment.
The SMO conducts on-site audits of LAS annually as part of the laboratory qualification program. At the
time of selection, the most recent audit of LAS had occurred in February 1997. Results from this audit
indicated that LAS was proficient in several areas, including program management, quality management,
and training programs. No findings regarding PCB analytical procedure implementation were noted. A
second on-site audit of LAS occurred August 11-12, 1997, during the analysis of the demonstration study
samples. This surveillance focused specifically on the procedures that were currently in use for the analysis
of the demonstration samples. The audit, jointly conducted by the SMO, DOE-ORO, and EPA-Las Vegas
(LV), verified that LAS was procedurally compliant. The audit team noted that LAS had excellent
adherence to the analytical protocols and that the staff were knowledgeable of the requirements of the
method. No findings impacting data quality were noted in the audit report.
19
-------
Reference Laboratory Method
The reference laboratory's analytical method, also presented in the technology demonstration plan [5],
followed the guidelines established in EPA SW-846 Method 8081 [4]. According to LAS's SOP, PCBs
were extracted from 30-g samples of soil by sonication in hexane. Each extract was then concentrated to a
final volume that was further subjected to a sulfuric acid cleanup to remove potential interferences. The
analytes were identified and quantified using a gas chromatograph equipped with dual electron-capture
detectors. Each extract was analyzed on two different chromatographic columns with slightly different
separation characteristics (primary column: RTX-1701, 30 m x 0.53 mm ID x 0.5 (jm; confirmatory
column: RTX-5, 30 m x 0.53 mm ID x 0.5 (jm). PCBs were identified when peak patterns from a sample
extract matched the patterns of standards for both columns. PCBs were quantified based on the initial
calibration of the primary column.
Calibration
Method 8081 states that, because Aroclors 1016 and 1260 include many of the peaks represented in the
other five Aroclor mixtures, it is only necessary to analyze two multilevel standards for these Aroclors to
demonstrate the linearity of the detector response for PCBs. However, per LAS SOPs, five-point (0.1 to 4
ppm) initial calibration curves were generated for Aroclors 1016, 1248, 1254, and 1260 and the surrogate
compounds [decachlorobiphenyl (DCB) and tetrachloro-m-xylene (TCMX)]. Single mid-level standards
were analyzed for the other Aroclors (1221, 1232, and 1242) to aid in pattern recognition. All of the multi-
point calibration data, fitted to quadratic models, met the QC requirement of having a coefficient of
determination (R2) of 0.99 or better over the calibration range specified. The detection limits for soil
samples were 0.033 ppm (ng/g) for all Aroclors except Aroclor 1221, which was 0.067 ppm. For extract
samples, the detection limits were 0.010 ppm (ng/mL) for all Aroclors except Aroclor 1221, which was
0.020 ppm. Reporting detection limits were calculated based on the above detection limits, the actual
sample weight, and the dilution factor.
Sample Quantification
For sample quantification, Aroclors were identified by comparing the samples' peak patterns and retention
times with those of the respective standards. Peak height ratios, peak shapes, sample weathering, and
general similarity in detector response were also considered in the identification. Aroclor quantifications
were performed by selecting three to five representative peaks, confirming that the peaks were within the
established retention time windows, integrating the selected peaks, quantifying the peaks based on the
calibrations, and averaging the results to obtain a single concentration value for the multicomponent
Aroclor. If mixtures of Aroclors were suspected to be present, the sample was typically quantified in terms
of the most representative Aroclor pattern. If the identification of multiple Aroclors was definitive, total
PCBs in the sample were calculated by summing the concentrations of both Aroclors. Aroclor
concentrations were quantified within the concentration range of the calibration curve. If PCBs were
detected and the concentrations were outside of the calibration range, the sample was diluted and
reanalyzed until the concentration was within the calibration range. If no PCBs were detected, the result
was reported as a non-detect (i.e., "< reporting detection limit").
20
-------
Sample Receipt, Handling, and Holding Times
The reference laboratory was scheduled to analyze a total of 256 PCB samples (208 soil samples, 24 iso-
octane extract samples, and 24 methanol extract samples). Of these same samples, the developer was
scheduled to analyze a total of 232 PCB samples (208 soil samples and 24 extract samples in solvent of
choice). The samples were shipped to LAS at the start of the technology demonstration activities (July 22).
Shipment was coordinated through the SMO. Completion of chain-of-custody forms documented sample
transfer. The samples were shipped on ice in coolers to maintain <6°C temperatures during shipment.
Samples were shipped with custody seals to ensure sample integrity and to prevent tampering during
transport.
Upon receipt of the samples, the reference laboratory checked the receipt temperature and conditions of the
sample containers, assigned each sample a unique number, and logged each into its laboratory tracking
system. All samples were received at the proper temperature and in good condition. Demonstration samples
were divided into 11 analytical batches (with no more than 20 samples per batch). The samples were
analyzed in an order specified by ORNL to ensure that the analysis of sample types was randomized.
Analyses of QC samples, supplied by the reference laboratory to indicate method performance, were
performed with each analytical batch of soils.
Prior to analysis, samples were stored in refrigerators kept at 4 to 6° C to maintain analyte integrity. The
reference laboratory was required to analyze the extract samples and to extract the soil samples within 14
days of shipment from ORNL. Once the soils were extracted, the reference laboratory had an additional 40
days to analyze the soil extracts. Maximum holding times were not exceeded for any of the demonstration
samples. The final reference laboratory data package for all samples was received at ORNL in 72 days, on
October 1, 1997. The contractual obligation was 30 days.
The remainder of this section is devoted to summarizing the data generated by the reference laboratory and
to assessing the analytical performance.
Quality Control Results
Objective
The purpose of this section is to provide an assessment of the data generated by the reference laboratory's
QC procedures. The QC samples included continuing calibration verification standards (CCVs), instrument
blanks, method blanks, surrogate spikes, [laboratory control samples (LCSs)], and MS/MSD samples.
Each control type is described in more detail in the following text and in the technology demonstration plan
[5]. Because extraction of these liquid samples was not required, calibration check standards and
instrument blanks were the only control samples implemented for the extract samples. The reference
laboratory's implementation of QC procedures was consistent with SW-846 guidance.
Continuing Calibration Verification Standard Results
A CCV is a single calibration standard of known concentration, usually at the midpoint of the calibration
range. This standard is evaluated as an unknown and is quantified against the initial calibration. The
calculated concentration is then compared with the nominal concentration of the standard to determine
whether the initial calibration is still valid. CCVs were analyzed with every 10 samples or at least every 12.
The requirement for acceptance was a percentage difference of less than 15% for the CCV relative to the
21
-------
initial calibration. This QC requirement was met for all Aroclors and surrogates, except for one standard
that had a 16% difference for DCB. These results indicated that the reference laboratory maintained
instrument calibrations during the course of sample analysis.
Instrument and Method Blank Results
Instrument blanks (hexane) were analyzed prior to each CCV. The QC requirement was that instrument
blanks must contain less than the reporting detection limit for any analyte. All instrument blanks were
acceptable.
A method blank is an analyte-free soil matrix sample that is taken through the extraction process to verify
that there are no laboratory sources of contamination. One method blank was analyzed for each analytical
batch. The QC requirement was that method blanks must contain less than the reporting detection limit for
any Aroclor. No PCBs were detected in any of the eleven method blanks that were analyzed. These results
demonstrated that the reference laboratory was capable of maintaining sample integrity, and that it did not
introduce PCB contamination to the samples during preparation.
Surrogate Spike Results
A surrogate is a compound that is chemically similar to the analyte group but is not expected to be present
in the environmental sample. A surrogate is added to test the extraction and analysis methods to verify the
ability to isolate, identify, and quantify a compound similar to the analyte(s) of interest without interfering
with the determination. Two different surrogate compounds, DCB and TCMX, were used to bracket the
retention time window anticipated in the Aroclor chromatograms. All soil samples, including QC samples,
were spiked with surrogates at 0.030 ppm prior to extraction. Surrogate recoveries were deemed to be
within QC requirements if the measured concentration fell within the QC acceptance limits that were
established by past method performance. (For LAS this was 39 to 117% for DCB, and 66 to 128% for
TCMX). The results were calculated using the following equation:
measured amount iririn/
percent recovery = x 100%
actual amount (4-1)
In all undiluted samples, both of the surrogates had percentage recoveries that were inside the acceptance
limits. Surrogate recoveries in diluted samples were uninformative because the spike concentration (0.030
ppm, as specified by the method) was diluted below the instrument detection limits. The surrogate recovery
results for undiluted samples indicated that there were no unusual matrix interferences or batch-processing
errors for these samples.
Laboratory Control Sample Results
A LCS is an aliquot of a clean soil that is spiked with known quantities of target analytes. The LCS is
spiked with the same analytes and at the same concentrations as the matrix spike (MS). (MSs are described
in the next section.) If the results of the MS analyses are questionable (i.e., indicating a potential matrix
effect), the LCS results are used to verify that the laboratory can perform the analysis in a clean,
representative matrix.
22
-------
Aroclors 1016 and 1260 were spiked into the clean soil matrix at approximately 0.300 ppm, according to
the reference laboratory's SOP. The QC requirements (defined as percent recovery) for the LCS analyses
were performance-based acceptance limits that ranged from 50 to 158%. In all but one of the eleven LCSs
analyzed, both Aroclor percent recoveries fell within the acceptance limits. Satisfactory recoveries for LCS
verified that the reference laboratory performed the analyses properly in a clean matrix.
Matrix Spike Results
In contrast to a laboratory control sample (LCS), a MS sample is an actual environmental soil sample into
which target analytes are spiked at known concentrations. MS samples are used to assess the efficiency of
the extraction and analytical methods for real samples. This is accomplished by determining the amount of
spiked analyte that is quantitatively recovered from the environmental soil. A duplicate matrix spike (MSB)
sample is spiked and analyzed to provide a measure of method precision. Ideally, to evaluate the MS/MSD
results, the environmental soil is analyzed unspiked so that the background concentrations of the analyte in
the sample are considered in the recovery calculation.
For the demonstration study samples, one MS and MSB pair was analyzed with each analytical batch. The
MS samples were spiked under the same conditions and QC requirements as the LCS (50 to 158%
acceptance limits), so that MS/MSD and LCS results could be readily compared. The QC requirement for
MS and MSB samples was a relative percent difference (RPD) of less than 30% between the MS/MSD
pair. RPD is defined as:
RpD = MS recovery - MSP recovery
average recovery
(4-2)
A total of eleven MS/MSD pairs were analyzed. Because the MS/MSD spiking technique was not always
properly applied (e.g., a sample which contained 100 ppm of Aroclor 1254 was spiked ineffectively with
0.300 ppm of Aroclor 1260), many of the MS/MSD results were uninformative. For the samples that were
spiked appropriately, all MS/MSD QC criteria were met.
Conclusions of the Quality Control Results
The reference laboratory results met performance acceptance requirements for all of the samples where
proper QC procedures were implemented. Acceptable performance on QC samples indicated that the
reference laboratory was capable of performing analyses properly.
Data Review and Validation
Objective
The purpose of validating the reference laboratory data was to ensure usability for the purposes of
comparison with the demonstration technologies. The data generated by the reference laboratory were used
as a baseline to assess the performance of the technologies for PCB analysis. The reference laboratory data
were independently validated by ORNL and SMO personnel, who conducted a thorough quality check and
reviewed all sample data for technical completeness and correctness.
23
-------
Corrected Results
Approximately 8% of the results provided by the reference laboratory (20 of 256) were found to have
correctable errors. So as not to bias the assessment of the technology's performance, errors in the reference
laboratory data were corrected. These changes were made conservatively, based on the guidelines provided
in the SW-846 Method 8081 for interpreting and calculating Aroclor results. The errors (see Appendix D,
Table D-3) were categorized as transcription errors, calculation errors, and interpretation errors. The
corrections listed in Table D-3 were made in the final data set that was used for comparison with the
demonstration technologies.
Suspect Results
Normally, one would not know if a single sample result was "suspect" unless (1) the sample was a
performance evaluation sample, where the concentration is known or (2) a result was reported and flagged
as suspect for some obvious reason (e.g., no quantitative result was determined). The experimental design
implemented in this demonstration study provided an additional indication of the abnormality of data
through the inspection of the replicate results from a homogenous soil sample set (i.e., four replicates were
analyzed for each sample ID).
Data sets were considered suspect if the standard deviation (SD) of the four replicates was greater than 30
ppm and the percent relative standard deviation (RSD) was greater than 50%. Five data sets (sample IDs
106, 205, 216, 217, 225) contained measurements that were considered suspect using this criteria, and the
suspect data are summarized in Table 4-1. A number of procedural errors may have caused the suspect
measurements (e.g., spiking heterogeneity, extraction efficiencies, dilution, etc.). In the following
subsections for precision and accuracy, the data were evaluated with and without these suspect values to
represent the best and worst case scenarios.
Table 4-1. Suspect measurements within the reference laboratory data
Criteria
SD > 30 ppm
and
RSD > 50%
Qualitative Result
Sample ID
106
205
216
217
225
110
112
PCB Concentration (ppm)
Replicate Results
(ppm)
255.9,269.9,317.6
457.0,483.3,538.7
47.0, 54.3, 64.0
542.8, 549.8, 886.7
32.1,36.5,56.4
< reporting detection
limits
Suspect Result(s)
(ppm)
649.6
3,305.0
151.6
1,913.3
146.0
< 66, < 98, < 99, < 490
< 66, < 130, < 200,< 200
Data Usability
Performed data analysis with
and without this value
Used as special case for
comparison with developer
results
Samples that did not fall into the above criteria, but were also considered suspect, were non-blank samples
that could not be quantified and were reported as "< the reporting detection limit." This was the case for
24
-------
environmental soil sample IDs 110 and 112. It is believed that the reference laboratory had trouble
quantifying these soil samples because of the abundance of chemical interferences. These samples were
diluted by orders of magnitude to reduce interferences, thereby diluting the PCB concentrations to levels
that were lower than the instrument detection limits. With each dilution, the reporting detection limits
values were adjusted for sample weight and dilution, which accounts for the higher reporting detection
limits (up to 490 ppm). It is believed that these samples should have been subjected to additional pre-
analytical cleanup to remove these interferences before quantification was attempted. Sample IDs 110 and
112 were collected from the same cleanup site (see Appendix B), so it is not surprising that similar
difficulties were encountered with both sample sets. Because the results for sample IDs 110 and 112 were
not quantitative, these data were compared with the technology data only on a special case basis.
Data Assessment
Objective
The purpose of this section is to provide an evaluation of the performance of the reference laboratory
results through statistical analysis of the data. The reference laboratory analyzed 72 PE, 136 environmental
soil, and 48 extract samples. All reference laboratory analyses were performed under the same
environmental conditions. Therefore, site differentiation was not a factor in data assessment for the
reference laboratory. For comparison with the technology data, however, the reference laboratory data are
delineated into "outdoor site" and "chamber site" in the following subsections. For consistency with the
technology review, results from both sites were also combined to determine the reference laboratory's
overall performance for precision and accuracy. This performance assessment was based on the raw data
compiled in Appendix D. All statistical tests were performed at a 5% significance level.
Precision
The term "precision" describes the reproducibility of measurements under a given set of conditions. The
SD of four replicate PCB measurements was used to quantify the precision for each sample ID. SD is an
absolute measurement of precision, regardless of the PCB concentration. To express the reproducibility
relative to the average PCB concentration, RSD is used to quantify precision, according to the following
equation:
RSD = Stmdard Deviation x 100% (4.3)
Average Concentration
Performance Evaluation Samples
The PE samples were homogenous soils containing certified concentrations of PCBs. Results for these
samples represent the best estimate of precision for soil samples analyzed in the demonstration study. Table
4-2 summarizes the precision of the reference laboratory for the analysis of PE samples. One suspect
measurement (sample ID 225, 146.0 ppm) was reported for the PE soil samples. The RSDs for the
combined data ranged from 9 to 33% when the suspect measurement was excluded, and from 9 to 79%,
including the suspect measurement. The overall precision, determined by the mean RSD for all PE samples,
was 21% for the worst case (including the suspect result) and 18% for the best case (excluding the suspect
result).
25
-------
Table 4-2. Precision of the reference laboratory for PE soil samples
Outdoor Site
Sample
ID
126 a
118
124
120
122
119
125
121
123
Average
Concentration
(ppm)
0
1.6
1.7
5.0
11.1
20.1
37.9
54.6
60.1
SD
(ppm)
n/a
0.6
0.2
1.0
0.9
3.4
6.9
3.4
4.6
RSD
(%)
n/a
39
13
20
8
17
18
6
8
Chamber Site
Sample
ID
226
218
224
220
222
219
225
221
223
Average
Concentration
(ppm)
0
2.6
1.7
5.8
12.8
23.3
41.7"
44.9
55.8
SD
(ppm)
n/a
0.2
0.5
1.8
0.3
6.1
12.9"
11.3
7.7
RSD
(%)
n/a
6
29
31
3
26
31"
25
14
Combined Sites
Average
Concentration
(ppm)
0
2.1
1.7
5.4
11.9
21.7
39.5C
49.8
58.0
SD
(ppm)
n/a
0.7
0.4
1.4
1.1
4.9
9.2 c
9.3
6.3
RSD
(%)
n/a
33
21
26
9
23
23 c
19
11
a All PCB concentrations were reported as non-detects.
b Results excluding the suspect value (results including the suspect value: mean = 67.8 ppm,
c Results excluding the suspect value (results including the suspect value: mean =52.8 ppm,
SD = 53.2 ppm, and RSD =79%).
SD = 38.6 ppm, and RSD = 73%).
Environmental Soil Samples
The precision of the reference laboratory for the analysis of environmental soil samples is reported in Table
4-3. In this table, results including suspect measurements are presented in parentheses. Average
concentrations were reported by the reference laboratory as ranging from 0.5 to 1,196 ppm with RSDs that
ranged from 7 to 118% when the suspect results were included. Excluding the suspect results, the highest
average concentration decreased to 660 ppm, and the largest RSD decreased to 71%. Because the majority
of the samples fell below 125 ppm, precision was also assessed by partitioning the results into two ranges:
low concentrations (< 125 ppm) and high concentrations (> 125 ppm). For the low concentrations, the
average RSD was 23% excluding the suspect value and 26% including the suspect value. These average
RSDs were only slightly larger than the RSDs for the PE soils samples of comparable concentration (18%
for best case and 21% for worst case). Five soil sample sets (sample IDs: 106, 117, 205, 211 and 217)
were in the high-concentration category. The average precision for high concentrations was 56% for the
worst case and 19% for the best case. The precision estimates for the low and high concentration ranges
were comparable when the suspect values were excluded. This indicated that the reference laboratory's
precision for the environmental soils was consistent (approximately 21% RSD), and comparable to the PE
soil samples when the suspect values were excluded.
The Paducah soils (indicated as bold sample IDs in Table 4-3) were analyzed by the technologies under
both outdoor and chamber conditions to provide a measure of the effect that two different environmental
26
-------
Table 4-3. Precision of the reference laboratory for environmental soil samples
Outdoor Site
Sample
ID
101
102
103
104
105
106
107
108
109
110
111
112
113 c
114
115
116
117
Average
Concentration
(ppm)
0.5
2.0
2.3
9.4
59.4
281.0 (373.2) a
1.3
1.8
2.0
n/ab
38.7
n/a
1.1
1.3
14.8
41.3
383.9
Standard
Deviation
(ppm)
0.1
0.3
0.6
4.0
16.5
32.4(186.2)
0.3
0.1
0.4
n/a
4.3
n/a
0.6
0.3
1.8
5.9
55.2
RSD
(%)
16
16
27
43
28
12 (50)
20
8
20
n/a
11
n/a
55
20
12
14
14
Chamber Site
Sample
ID
206
207
208
209
210
211
212
213
214
215
216
217
201
202
203
204
205
Average
Concentration
(ppm)
1.9
18.8
30.5
40.2
88.6
404.5
3.2
8.1
25.2
26.7
55.1 (79.2)
659.8(973.2)
0.9
1.4
13.9
44.3
493.0(1196.0)
Standard
Deviation
(ppm)
0.9
3.5
7.9
28.5
25.6
121.8
1.6
1.6
3.7
3.2
8.5 (48.7)
196.6(647.0)
0.2
0.2
1.7
2.9
41.7(1406.4)
RSD
(%)
49
19
26
71
29
30
50
20
15
12
15 (62)
30 (66)
24
12
12
7
8(118)
a Data in parentheses include suspect values.
bn/a indicates that qualitative results only were reported for this sample.
c Bold sample IDs were matching Paducah sample pairs (i.e., 113/201, 114/202, 115/203, 116/204, 117/205).
conditions had on the technology's performance. Although this was not an issue for the reference laboratory
(because all the samples were analyzed under laboratory conditions), the reference laboratory's results were
delineated into the different site categories for comparison with the technologies. Sample IDs 113 and 201,
114 and 202, 115 and 203, 116 and 204, and 117 and 205 each represent a set of eight replicate samples
of the same Paducah soil. The RSDs for four of the five Paducah pairs (excluding the suspect value for
sample ID 205) ranged from 11 to 17%. The result from one pair (sample IDs 113 and 201) had an RSD
of 42%, but the reported average concentration was near the reporting limits.
Extract Samples
The extract samples, which were used to simulate surface wipe samples, were the simplest of all the
27
-------
demonstration samples to analyze because they required no extraction and were interference-free. Three
types of extract samples were analyzed: solvent blanks, spikes of Aroclor 1242 at 10 (ig/mL, and spikes of
Aroclor 1254 at 100 (jg/mL. Identical extract samples were prepared in two solvents (iso-octane and
methanol) to accommodate the developer's request. The reference laboratory analyzed both solvent sets. A
Student's t-test [7, 8] was used to compare the reference laboratory's average PCB concentrations for the
two different solvents and showed that no significant differences were observed at either concentration.
Therefore, the reference laboratory results for the two extract solvents were combined. Additionally, all
blank samples were quantified as non-detects by the reference laboratory.
Table 4-4 summarizes the reference laboratory results for the extract samples by site. RSDs for the four
replicates for each sample ID ranged from 3 to 24%. For the combined data set (16 replicate
measurements), the average RSD at the 10-(jg/mL level was 19%, while the average RSD at the 100-
(jg/mL level was 8%. For the entire extract data set, an estimate of overall precision was 14%. The overall
precision for the extract samples was comparable to the best-case precision for environmental soil samples
(21%) and PE soil samples (18%).
Table 4-4. Precision of the reference laboratory for extract samples
Outdoor Site
Sample
ID
129 a
132 a
127
130
128
131
Average
Cone
(ug/mL)
0
0
10.9
12.1
67.4
63.8
SD
(ug/mL)
n/a
n/a
0.4
2.9
2.3
5.0
RSD
(%)
n/a
n/a
4
24
3
8
Chamber Site
Sample
ID
229
232
227
230
228
231
Average
Cone
(ug/mL)
0
0
9.6
8.9
65.2
57.7
SD
(ug/mL)
n/a
n/a
0.8
1.4
5.1
3.1
RSD
(%)
n/a
n/a
8
16
8
5
Combined Sites
Average
Cone
(ug/mL)
0
10.4
63.5
SD
(ug/mL)
n/a
1.9
5.2
RSD
(%)
n/a
19
8
a All PCB concentrations reported as non-detects by the laboratory.
Accuracy
Accuracy represents the closeness of the reference laboratory's measured PCB concentrations to the
accepted values. Accuracy was examined by comparing the measured PCB concentrations (for PE soil and
extract samples) with the certified PE values and known spiked extract concentrations. Percent recovery
was used to quantify the accuracy of the results. The optimum percent recovery value is 100%. Percent
recovery values greater than 100% indicate results that are biased high, and values less than 100% indicate
results that are biased low.
Performance Evaluation Soil Samples
The reference laboratory's performance for the PE samples is summarized in Table 4-5. Included in this
table are the performance acceptance ranges and the certified PCB concentration values. The acceptance
ranges, based on the analytical verification data, are guidelines established by the provider of the PE
materials to gauge acceptable analytical results. As shown in Table 4-5, all of the average concentrations
28
-------
were within the acceptance ranges, with the exception of sample ID 218. The average result of sample ID
225 was outside of the acceptance range only when the suspect result was included. All of the replicate
measurements in sample ID 225 were biased slightly high. Average percent recoveries for the PE samples
(excluding suspect values) ranged from 76 to 130%. Overall accuracy was estimated as the average
recovery for all PE samples. The overall percent recovery was 105% as a worst case when the suspect
value was included. Excluding the suspect value as a best case slightly lowered the overall percent recovery
to 101%. A regression analysis [9] indicated that the reference laboratory's results overall were unbiased
estimates of the PE sample concentrations.
Table 4-5. Accuracy of the reference laboratory for PE soil samples
Certified
Concentration
(ppm)
(Acceptance
Range, ppm)
Oa
(n/a)
2.0
(0.7-2.2)
2.0
(0.9-2.5)
5.0
(2.1-6.2)
10.9
(4.0-12.8)
20.0
(11.4-32.4)
49.8
(23.0-60.8)
50.0
(19.7-63.0)
50.0
(11.9-75.9)
Outdoor Site
Sample
ID
126
118
124
120
122
119
125
121
123
Average
Cone
(ppm)
0
1.6
1.7
5.0
11.1
20.1
37.9
54.6
60.1
Recovery
(%)
n/a
79
85
99
102
100
76
109
120
Chamber Site
Sample
ID
226
218
224
220
222
219
225
221
223
Average
Cone
(ppm)
0
2.6
1.7
5.8
12.8
23.3
41.7"
44.9
55.8
Recovery
(%)
n/a
130
85
117
117
116
84"
90
112
Combined Sites
Average
Cone
(ppm)
0
2.1
1.7
5.4
11.9
21.7
39.5 c
49.8
58.0
Recovery
(%)
n/a
105
85
108
109
109
79 c
100
116
a All PCB concentrations reported as non-detects by the laboratory.
b Results excluding the suspect value (results including the suspect value: average = 67.!
'Results excluding the suspect value (results including the suspect value: average = 52.S
ppm and recovery = 136%).
ppm and Recovery = 106%).
Extract Samples
Percent recovery results for extract samples are summarized in Table 4-6 for the reference laboratory. The
average percent recoveries for extract samples ranged from 58 to 121%. In terms of concentration levels,
the average recovery at the 10-(jg/mL level (for both solvents) was 104%, compared with 64% at the 100-
(jg/mL level. The reference laboratory classified all 16 samples spiked at 10 (jg/mL as Aroclor 1016;
29
-------
however, these samples were actually spiked with Aroclor 1242. Despite this misclassification, the results
did not appear to be biased. In contrast, the samples spiked at 100 (jg/mL were correctly classified as
Aroclor 1254 but were all biased low. Although these results suggested that Aroclor classification had little
effect on the quantification of the extract samples, there was an obvious, consistent error introduced into
the analysis of the 100-(jg/mL samples to cause the low bias. For the entire extract data set, the overall
percent recovery was 84%.
Table 4-6. Accuracy of the reference laboratory for extract samples
Spike
Concentration
(ug/mL)
Oa
Oa
10
10
100
100
Outdoor Site
Sample
ID
129
132
127
130
128
131
Avg
Cone
(ug/mL)
0
0
10.9
12.1
67.4
63.8
Recovery
(%)
n/a
n/a
109
121
67
64
Chamber Site
Sample
ID
229
232
227
230
228
231
Avg
Cone
(ug/mL)
0
0
9.6
8.9
65.2
57.7
Recovery
(%)
n/a
n/a
96
89
65
58
Combined Sites
Avg
Cone
(ug/mL)
10.4
63.5
Recovery
(%)
n/a
104
64
a All PCB concentrations reported as non-detects by the laboratory.
Representativeness
Representativeness expresses the degree to which sample data accurately and precisely represent the
capability of the method. Representativeness of the method was assessed based on the data generated for
clean-QC samples (i.e., method blanks and laboratory control samples) and PE samples. Based on the data
assessment (discussed in detail in various parts of this section), it was determined that the
representativeness of the reference laboratory data was acceptable. In addition, acceptable performance on
laboratory audits substantiated that the data set was representative of the capabilities of the method. In all
cases, the performance of the reference laboratory met all requirements for both audits and QC analyses.
Completeness
Completeness is defined as the percentage of measurements that are judged to be usable (i.e., the result was
not rejected). Usable results were obtained for 248 of the 256 samples submitted for analysis by the
reference laboratory. Eight results (for sample IDs 110 and 112) were deemed incomplete and therefore not
valid because the measurements were not quantitative. To calculate completeness, the total number of
complete results were divided by the total number of samples submitted for analysis, and then multiplied by
100 to express as a percentage. The completeness of the reference laboratory was 97%, where a
completeness of 95% or better is typically considered acceptable.
30
-------
Comparability
Comparability refers to the confidence with which one data set can be compared with another. The
demonstration study was designed to have a one-to-one, sample-by-sample comparison of the PCB results
obtained by the reference laboratory and the PCB results obtained by the technology being evaluated.
Based on thorough examination of the data and acceptable results on the PE samples, it was concluded that
the reference laboratory's SOPs for extraction and analysis, and the data generated using these procedures,
were of acceptable quality for comparison with the field technology results. Additional information on
comparability was available because the experimental design incorporated randomized analysis of blind,
replicate samples. Evaluation of the replicate data implicated some of the individual data points as suspect
(see Table D-2). The reference laboratory's suspect data were compared with the technology data on a
special-case basis, and exceptions were noted.
Summary of Observations
Table 4-7 provides a summary of the performance of the reference laboratory for the analysis of all sample
types used in the technology demonstration study. As shown in Table 4-7, the precision of the PE soils was
comparable to the environmental soils. A weighted average, based on the number of samples, gave a best-
case precision of 21% and a worst-case precision of 28% for all the soil data (PE and
Table 4-7. Summary of the reference laboratory performance
Sample Matrix
Blank
Environmental soil with
interferences
Soil
Best Case
(excluding suspect data)
Soil
Worst Case
(including suspect data)
Extract
Sample Type
Soil
Extract
Sample ID 110
Sample ID 112
PE
Environmental
< 125 ppm
> 125 ppm
overall
PE
Environmental
< 125 ppm
> 125 ppm
overall
10 ppm
100 ppm
overall
Number of Samples
8
16
4
4
63
107
17
187
64
108
20
192
16
16
32
Precision
(Average % RSD)
n/aa
n/aa
18
23
19
21
21
26
56
28
19
8
14
Accuracy
(Average %Recovery)
All samples were
reported as non-detects.
All samples were
reported as non-detects.
101
n/ab
n/ab
101
105
n/ab
n/ab
105
104
64
84
a Because the results were reported as non-detects, precision assessment is not applicable.
b Accuracy assessment calculated for samples of known concentration only.
-------
environmental). The extract samples had a smaller overall RSD of 14%. Evaluation of overall accuracy
was based on samples with certified or known spiked concentrations (i.e., PE and extract samples). The
overall accuracy, based on percent recovery, for the PE samples was 105% for the worst case (which
included the suspect value) and 101% for the best case (which excluded the suspect value). These results
indicated that the reference laboratory measured values were unbiased estimates of the certified PE
concentrations (for samples that contained <50 ppm of PCBs). Accuracy for the extract samples at 10 ppm
was also unbiased, with an average percent recovery of 104%. However, the accuracy for the extract
samples at 100 ppm was biased low, with an average recovery of 64%. Overall, the average percent
recovery for all extract samples was 84%. The reference laboratory correctly reported all blank samples as
non-detects, but had difficulty with two soil sample IDs (110 and 112) that contained chemical
interferences. In general, the reference laboratory's completeness would be reduced, at the expense of an
improvement in precision and accuracy, if the suspect measurements were excluded from the data analysis.
Based on this analysis, it was concluded that the reference laboratory results were acceptable for
comparison with the developer's technology.
32
-------
Section 5
Technology Performance and Evaluation
Objective and Approach
The purpose of this section is to present the evaluation of data generated by EST's 4100 Vapor Detector.
The technology's precision and accuracy performance are presented for the data generated in the
demonstration study. In addition, an evaluation of comparability, through a one-to-one comparison with the
reference laboratory data, is presented. An evaluation of other aspects of the technology (such as detection
limits, cost, sample throughput, hazardous waste generation, and logistical operation) is also presented in
this section.
Data Assessment
The purpose of the data assessment section is to present the evaluation of the performance of EST's 4100
Vapor Detector through a statistical analysis of the data. PARCC parameters were used to evaluate the
technology's ability to measure PCBs in PE, environmental soil, and extract samples. The developer
analyzed splits of replicate samples that were also analyzed by the reference laboratory (72 PE soil
samples, 136 environmental soil samples, and 24 extract samples). See Section 4 for a more detailed
analysis of the reference laboratory's results. Replicate samples were analyzed by the developer at two
different sites (under outdoor conditions and inside an environmentally controlled chamber) to evaluate the
effect of environmental conditions on performance; see Section 3 for further details on the different sites.
Evaluation of the data sets indicated that there were no significant differences in the precision of the
measurements made at each site. There were significant differences, however, in the accuracy of the
measured concentrations determined at each site. In cases where the environmental conditions did not affect
results significantly, data from both sites were combined for each parameter (precision and accuracy) to
determine overall performance. All statistical tests were performed at the 5% significance level. Appendix
D contains the raw data that were used to assess the performance of the 4100 Vapor Detector.
Precision
Precision, as defined in Section 4, is the reproducibility of measurements under a given set of conditions.
The SD and RSD of four replicate measurements were used to quantify the technology's precision. The
average PCB concentration for a replicate set was used to calculate the RSD for each sample ID (see Eq.
4-3). For more information regarding the reference laboratory's precision, refer to the data presented in
Section 4 under the heading of "Precision."
Performance Evaluation Samples
Table 5-1 summarizes the precision of the 4100 Vapor Detector for the analysis of PE samples. Operating
under the outdoor conditions, the RSDs ranged from 51 to 162%. RSDs ranged from 37 to 116%, while
operating inside the chamber. In Table 5-1, the data generated under both environmental conditions were
also combined to provide an overall assessment of precision. The performance for the combined site data
indicated RSDs ranging from 62 to 165%.
33
-------
Table 5-1. Precision of the 4100 Vapor Detector for PE soil samples
Outdoor Site
Sample
ID
126 a
118
124
120
122
119
125
121
123
Average
Concentration
(ppm)
33.5
9.2
8.3
8.2
20.9
2.3
30.9
39.1
18.4
SD
(ppm)
42.5
10.5
4.5
11.0
33.7
3.7
19.9
20.0
12.5
RSD
(%)
127
115
54
133
161
162
64
51
68
Chamber Site
Sample
ID
226 a
218
224
220
222
219
225
221
223
Average
Concentration
(ppm)
77.8
39.8
32.1
27.0
8.1
45.6
120.1
142.9
44.3
SD
(ppm)
78.9
25.2
11.9
14.1
9.4
29.9
73.1
68.9
16.5
RSD
(%)
101
63
37
52
116
65
61
48
37
Combined Sites
Average
Concentration
(ppm)
55.6
24.5
20.2
17.6
14.5
24.0
75.5
91.0
31.1
SD
(ppm)
63.3
24.2
15.2
15.4
23.9
30.4
68.8
72.7
19.4
RSD
(%)
114
99
75
88
165
127
91
80
62
a The 4100 detected PCBs in the blanks. The blank data were not included in the calculation of the overall average RSD.
Environmental Soil Samples
The precision of the 4100 Vapor Detector for the analysis of environmental soil samples is reported in
Table 5-2. Operating under the outdoor conditions, the RSDs ranged from 33 to 187%. RSDs ranged from
25 to 151%, while operating inside the chamber. Because most of the measurements fell below 125 ppm,
precision was also assessed by partitioning the results into two ranges: low concentrations (reference
laboratory values < 125 ppm) and high concentrations (reference laboratory values > 125 ppm). See
Section 4 for delineation of sample IDs in each concentration range. For the low-concentration range, the
average RSD was 93%, in contrast to that of the high-concentration range, which was 50%.
The Paducah soils (indicated by bold sample IDs in Table 5-2) were analyzed at both sites to provide an
assessment of the 4100's performance under different environmental conditions. For these samples, the data
generated under both environmental conditions were also combined to provide an overall assessment of
precision. Sample IDs 113 and 201, 114 and 202, 115 and 203, 116 and 204, and 117 and 205 represented
replicate Paducah soil sample sets, where the "100" series were samples analyzed under the outdoor
conditions and the "200" series were samples analyzed inside the chamber. An analysis of variance
(ANOVA) test was used to compare the effect of the two environmental conditions on the average
measurements. Results from the ANOVA analysis showed that there were no significant differences in the
RSDs generated at each site, however, the average measured concentrations were different, indicating that
environmental conditions had an effect on the 4100's ability to measure PCB concentrations, but that the
precision of the measurements was similar at each site, as illustrated in Table 5-2. When the 4100 Vapor
Detector was used under the outdoor conditions, the RSDs for the Paducah samples ranged from 33 to
34
-------
116%, and from 48 to 116% operating inside the chamber. RSDs for the combined site data (8 replicates
per paired Paducah sample ID) ranged from 65 to 160%.
Table 5-2. Precision of the 4100 Vapor Detector for environmental soil samples
Outdoor Site
Sample
ID
107
104
108
101
105
109
103
110
102
111
106
112
113 a
114
115
116
117
Average
Concentration
(ppm)
3.7
5.6
7.3
14.5
27.8
35.2
41.8
53.5
55.2
73.3
151.3
693.3
5.9
16.0
8.2
4.4
119.5
SD
(ppm)
2.3
5.3
10.9
14.3
18.8
36.4
49.3
100.0
101.1
93.1
105.6
891.9
4.5
17.2
9.5
4.4
39.1
RSD
(%)
62
95
150
99
68
103
118
187
183
127
70
129
77
108
116
101
33
Chamber Site
Sample
ID
206
212
213
215
214
207
210
209
216
208
217
211
201
202
203
204
205
Average
Concentration
(ppm)
53.5
58.9
135.9
205.3
253.5
258.5
396.9
406.4
430.3
470.4
537.8
717.9
11.2
23.0
14.2
104.1
295.1
SD
(ppm)
80.5
19.9
96.0
145.6
239.4
158.5
145.3
259.9
108.8
271.4
375.7
204.1
13.0
26.1
14.5
104.5
142.1
RSD
(%)
151
34
71
71
94
61
37
64
25
58
70
28
116
113
103
100
48
Combined
Site
RSD
(%)
n/ab
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
111
107
106
160
65
a Bold sample IDs were matching Paducah sample pairs (i.e., 113/201, 114/202, 115/203, 116/204, 117/205).
b Combined site results were not applicable because these environmental samples were not replicate pairs.
Extract Samples
Table 5-3 summarizes the 4100 Vapor Detector results for the extract samples used to simulate surface
wipe samples. Refer to Section 3 under the heading of "Extract Samples" for further clarification of this
sample type. When the 4100 Vapor Detector was used under the outdoor conditions, the RSDs ranged from
9 to 60%. RSDs ranged from 47 to 63% when it was used inside the chamber. For the combined site data,
the average RSD at the 10-(jg/mL level was 71%; the average RSD at the 100-(ig/mL level was 59%.
35
-------
Table 5-3. Precision of the 4100 Vapor Detector for extract samples
Outdoor Site
Sample
ID
132 a
130
131
Average
Concentration
(ug/mL)
0.2
14.1
40.5
SD
(ug/mL)
0.4
8.5
3.7
RSD
(%)
200
60
9
Chamber Site
Sample
ID
232 a
230
231
Average
Concentration
(ug/mL)
33.3
39.3
68.3
SD
(ug/mL)
66.2
18.6
43.3
RSD
(%)
199
47
63
Combined Sites
Average
Concentration
(ug/mL)
16.7
26.7
54.4
SD
(ug/mL)
46.8
19.0
32.1
RSD
(%)
280
71
59
a The 4100 detected PCBs in the blanks. The blank data were not included in the calculation of the overall average RSD.
Precision Summary
The overall precision was characterized by three summary values for the RSD: mean (i.e., average),
median (i.e., 50th percentile value at which 50% of all individual RSD values are below and 50% are
above), and 95th percentile (i.e., the value at which 95% of all individual RSD values are below and 5%
are above). These values are summarized in Table 5-4 for each of the sample types. The 4100 Vapor
Detector's overall precision for the PE samples was a mean RSD of 81%, a median RSD of 64%, and the
95th percentile of all individual RSDs was 161%. The environmental soil sample RSD results were a mean
of 90%, a median of 95%, and a 95th percentile of 162%. The overall precision for all extract samples was
a mean RSD of 65%; the 95th percentile and median data were not presented because the number of data
points was limited .
Table 5-4. Overall precision of the 4100 Vapor Detector for all sample types
Statistic
Mean
Median
95th nercentile
PE Samples
%RSD
Outdoor Chamber Combined
101
91
162
60
57
98
81
64
161
Environmental Soil Samples
%RSD
Outdoor Chamber Combined a
107
103
184
73
70
123
90
95
162
Extract Samples
%RSD
Outdoor Chamber Combined
34
n/ab
n/a
55
n/a
n/a
65
n/a
n/a
a Combined data were only generated for the Paducah soil samples.
b Median and 95th percentile statistics were not applicable to extract samples.
Accuracy
Accuracy, as defined in Section 4, represents the closeness of the technology's measured PCB
concentrations to the accepted values. Accuracy was examined in terms of percent recovery (see Eq. 4-1),
and average percent recoveries were calculated by averaging the four replicates within a sample ID. For
comparative information on the performance of the reference laboratory, refer to Section 4 under the
heading of "Accuracy."
36
-------
Performance Evaluation Soil Samples
The 4100 Vapor Detector's performance for the PE samples is summarized in Table 5-5. Included in this
table are the performance acceptance ranges and the certified PCB concentration values. Most of the
average concentrations determined by the 4100 were outside of the acceptance ranges and most were biased
high. This was also reflected in the average percent recoveries, which ranged as high as 1,990%. Average
percent recoveries ranged from 12 to 460% while operating under the outdoor conditions. Under chamber
conditions, average percent recoveries ranged from 74 to 1,990%.
A regression analysis [9] indicated that there were significant differences between outdoor and chamber
results. These differences could be caused by changes in the analytical procedure, which are noted in
Section 3 under the heading of "Deviations from the Demonstration Plan." Because there appeared to be
significant differences in the data generated at the two sites, the data were not combined. Additionally, there
was low correlation of the 4100's measured PCB concentrations with the certified PE values. This is
illustrated in Figure 5-1.
Extract Samples
Percent recovery results for the extract samples are summarized in Table 5-6 for the 4100 Vapor Detector.
The average percent recoveries for extract samples ranged from 41 to 141% when the 4100 was operated
under the outdoor conditions and ranged from 68 to 393% inside the chamber. In terms of concentration
levels (i.e., for the combined site data), the average recovery at the 10-(jg/mL level was 267%, compared
with 54% at the 100-(jg/mL level. Of the eight blank samples analyzed, five were reported as non-detects,
two as <1 ppm, and one as 133 (jg/mL. Note that the one anomalous blank result was obtained after the
zju -
E 200 -
Q.
Q. 1
O
15 150 -
'c
Ol
o
o
0 100 j
CO
V. '
o
° 50-
•* 1
,
1
1
1
1
O 1
o o
1
.858
Ix
10 20 30 40
Certified PCB Concentration (ppm)
50
Figure 5-1. The 4100 Vapor Detector's results versus the
certified PCB concentration for PE soil samples.
37
-------
analysis of a 100-(jg/mL sample; it may have resulted from carryover if insufficient instrument blanks were
not analyzed. Refer to Section 2 under the heading of "Sample Preparation and Analysis Procedures" for
more information on the technology's requirements for the analysis of instrument blanks.
Accuracy Summary
The overall accuracy was characterized by three summary values for the percent recovery: mean, median,
and 95th percentile. These values are summarized in Table 5-7 for the PE and extract samples. For the PE
samples, the overall accuracy of the 4100 Vapor Detector can be characterized as biased with a significant
influence based on environmental conditions. The mean percent recovery of the 4100 operating under
outdoor conditions was 177% with a median of 53% a and 95th percentile of 693%. Under chamber
conditions, the overall accuracy was a mean percent recovery of 631%, a median of 257%, and a 95th
percentile of 2,150%. The overall accuracy for all extract samples was a mean percent recovery of 161%;
the 95th percentile and median data were not presented because the number of data points was limited.
Table 5-5. Accuracy of the 4100 Vapor Detector for PE soils samples
Certified
Concentration
(ppm)
(Acceptance
Range, ppm)
Oa
(n/a)
2.0
(0.7-2.2)
2.0
(0.9-2.5)
5.0
(2.1-6.2)
10.9
(4.0-12.8)
20.0
(11.4-32.4)
49.8
(23.0-60.8)
50.0
(19.7-63.0)
50.0
(11.9-75.9)
Outdoor Site
Sample
ID
126
118
124
120
122
119
125
121
123
Average
(ppm)
33.5
9.2
8.3
8.2
20.9
2.3
30.9
39.1
18.4
Recovery
(%)
n/a
460
415
164
192
12
62
78
37
Chamber Site
Sample
ID
226
218
224
220
222
219
225
221
223
Average
(ppm)
77.8
39.8
32.1
27.0
8.1
45.6
120.1
142.9
44.3
Recovery
(%)
n/a
1,990
1,605
540
74
228
241
286
89
a The 4100 Vapor Detector detected PCBs in the blanks. Average recovery calculations were not applicable to blank samples.
38
-------
Table 5-6. Accuracy of the 4100 Vapor Detector for extract samples
Spike
Concentration
(ug/mL)
Oa
10
100
Outdoor Site
Sample
ID
132
130
131
Average
Cone
(ug/mL)
0.2
14.1
40.5
Recovery
(%)
n/a
141
41
Chamber Site
Sample
ID
232
230
231
Average
Cone
(ug/mL)
33.3
39.3
68.3
Recovery
(%)
n/a
393
68
Combined Sites
Average
Cone
(ug/mL)
16.7
26.7
54.4
Recovery
(%)
n/a
267
54
1 The 4100 Vapor Detector detected PCBs in the blanks. Average recovery calculations were not applicable to blank samples.
Table 5-7. Overall accuracy of the 4100 Vapor Detector for all sample types
Statistic
Mean
Median
95th nercentile
PE Samples
%Recovery
Outdoor Chamber Combined
177
53
693
631
257
2,150
n/aa
n/a
n/a
Extract Samples
%Recovery
Outdoor Chamber Combined
91
n/ab
n/a
231
n/a
n/a
161
n/a
n/a
a Combined site results are not applicable because of significant site-specific differences.
b Median and 95th percentile statistics were not applicable to extract samples.
False Positive/False Negative Results
A false positive (fp) result [10] is one in which the technology detects PCBs in the sample when there
actually are none. A false negative (fn) result [10] is one in which the technology indicates that there are no
PCBs present in the sample, when there actually are. Both fp and fn results are influenced by the method
detection limit of the technology. Because EST did not specify a method detection limit prior to the
demonstration, any PCB concentration that was detected was considered real. Of the blank soil samples
analyzed, PCBs were reported in all eight (fp = 100%). Of the 192 non-blank soil samples analyzed, ten
were reported as non-detects (i.e., fn = 5%). For the extract samples, the percentage of fp results was 38%
(three of eight blank samples were reported as containing PCBs), with 0% fn results.
Representativeness
Representativeness expresses the degree to which the sample data accurately and precisely represent the
capability of the technology. The performance data were accepted as being representative of the technology
because the 4100 Vapor Detector was capable of analyzing diverse sample types (PE, simulated surface
wipe extract, and actual environmental soil samples) under multiple environmental conditions. When this
technology is used, independent quality control samples should be analyzed to assess the performance of
the 4100 under the testing conditions.
39
-------
Completeness
Completeness is defined as the percentage of measurements that are judged to be useable (i.e., the result
was not rejected). Useable results were obtained by the technology for all 232 samples. Therefore, the
completeness of the 4100 Vapor Detector was 100%.
Comparability
Comparability refers to the confidence with which one data set can be compared with another. A one-to-one
sample comparison was performed to assess the comparability of the PCB concentrations found in all soil
samples (PE and environmental) for the 4100 measured values versus the reference laboratory results.
Additional statistical analysis of the PCB soil concentrations for paired samples showed that the 4100
measured values were significantly different from the reference laboratory results. This is illustrated in
Figure 5-2, which is a plot of the 4100 measured PCB soil concentrations versus the corresponding
reference laboratory measured concentrations (excluding the suspect values listed in Table 4-1). Figure 5-2
(a) is a plot of all of the soil data, while (b) is a plot of the concentration region from 0 to 125 ppm, where
most of the variation can be viewed. Note that the diagonal lines drawn in Figure 5-2 represent the line of
theoretically perfect correlation (R2 =1.0) between the reference laboratory data set (plotted along the x-
axis) and the 4100 data set (plotted along the j-axis). A value above the diagonal line indicated that the
4100's measurement was higher than the reference laboratory's measurement, while those below the
diagonal line indicated a lower result. Coefficients of determination (R2) [9] were computed using a linear
model fitted to the plot of the 4100 PCB concentrations versus the reference laboratory PCB
concentrations. Excluding the reference laboratory's suspect measurements, the coefficient of determination
(R2) was 0.177 when all soil samples (0 to 700 ppm) were considered. As shown in Figure 5-2(b), most of
the soil samples were in the concentration range of 0 to 125 ppm. The R2 value for this concentration range
was 0.115.
A direct comparison between the 4100 and reference laboratory data was performed by evaluating the
percent difference (% D) between the measured concentrations, defined as:
%D= [4100] - [RefLab]
[Ref Lab] (5-1)
Figure 5-3 provides a summary of the range of percent difference values for the soil samples, as calculated
using Eq. 5-1. The graph represents the percentage of samples that fall within each range of percent
difference values; however, the graph does not reflect any grouping according to the actual concentrations
of the replicate sets. Results for sample IDs 110, 112, 126, and 226 were not included because the
reference laboratory did not report quantitative results for them. As shown in Figure 5-3, most of the
percent difference values were greater than 100%, and 40% of the samples had a negative bias (<-!%)
relative to the reference laboratory results. Approximately 10% of the soil sample results had %D values
within the range of ± 25%.
40
-------
0 100 200 300 400 500 600 700 800 900
Reference Laboratory PCB Concentration
0 25 50 75 100 125
Reference Laboratory PCB Concentration (ppm)
(a)
(b)
Figure 5-2. Paired PCB measurements for 4100 and reference measurements for (a) all soil samples and (b) soil
samples where the reference laboratory result was less than or equal to 125 ppm. Lines denote perfect correlation.
60 T
-100% to -75% to- -50% to- -25% to- 0% to 25% to
-76% 51% 26% 1% 24% 49%
50% to 75% to
74% 100%
101%
Range of percent difference values
Figure 5-3. Range of percent difference values for the comparison of the 4100 Vapor Detector soil sample
results with the reference laboratory results.
41
-------
Comparability was also assessed for the extract samples. Figure 5-4 is a plot of the 4100 measured extract
concentrations versus the reference laboratory results. The coefficient of determination (R2) was 0.187 for
a line fit to the data, indicating a low correlation between the 4100 extract values and the reference
laboratory results. The percent difference values for the extract samples were also assessed, and are shown
in Figure 5-5. The bias in the 4100 results was evenly distributed (positive and negative) compared with
that of the reference laboratory results. Approximately 19% of the extract results had %D values within the
range of ± 25%.
The soil data not included in previous comparability evaluations (because the replicate data for the
reference laboratory were considered suspect) are shown in Table 5-8. Refer to Section 4, in particular
Table 4-1, for more information on the reference laboratory's suspect measurements. The reference
laboratory's suspect data were compared with the 4100's matching results. For sample IDs 110 and 112,
the reference laboratory obtained qualitative results only, while EST reported quantitative PCB
concentrations. For the other five suspect reference laboratory measurements, quantitative results were
obtained; however, one of the four replicates was considered suspect. For those samples, the 4100
generated quantitative results that were not always consistent with the replicate means or comparable with
the reference laboratory's corresponding suspect value.
Table 5-8. Comparison of the reference laboratory's suspect data to the 4100 Vapor Detector data
Sample ID
110
112
106
205
216
217
225
Reference Laboratory
Suspect Measurement
(ppm)
-------
20 40 60 80 100 120 140
Reference Laboratory PCB Concentration (ppm)
Figure 5-4. Paired PCB extract measurements for the
4100 Vapor Detector and reference laboratory.
-75% to - -50% to - -25% to - 0% to 25% to 50% to 75% to
51% 26% 1% 24% 49% 74% 100%
Range of percent difference values
> 101%
Figure 5-5. Range of percent difference values for the comparison of the 4100 Vapor Detector
extract sample results with the reference laboratory results.
43
-------
Table 5-9. Summary of PARCC observations for the 4100 Vapor Detector
Sample
Matrix
Blank
Soil
Extract
Sample
Type
Soil
Extract
PE
Environmental
< 125 ppm b
> 125 ppm c
Sample ID 110
Sample ID 112
overall
10 ppm
100 ppm
overall
4100's Number
of Samples
8
8
64
108
20
4
4
200
8
8
16
Precision (Average % RSD)
4100
114
280
81
93
50
187
129
87
71
59
65
Reference
Laboratory
n/a
18a
23 a
19a
not quantified
not quantified
21 a
19
8
14
Accuracy (Average %Recovery)
4100
Background
contributions
were detected.
177 (outdoors)
631 (chamber)
177 (outdoors)
631 (chamber)
267
54
161
Reference
Laboratory
All reported as
non-detects.
101 a
101 a
104
64
84
a Average result excluding the suspect measurements.
b Samples where the reference laboratory values were < 125 ppm.
c Samples where the reference laboratory values were > 125 ppm.
comparison with the reference laboratory's overall RSD of 21%. For the extract samples, the overall
average RSD for the 4100 was 65%, compared with that of the reference laboratory, which was 14%.
In terms of accuracy, the 4100's PE soil measurements were generally biased high. The results also
indicated significant differences in the percent recoveries of the measurements performed outdoors (177%
recovery) and those performed inside the chamber (631% recovery). In comparison, the reference
laboratory reported unbiased PCB concentrations for these PE soil samples. Extract measurements by the
4100 were also biased high at 10 ppm (267% recovery), but were biased low at 100 ppm (54% recovery).
In contrast, the reference laboratory results were unbiased at 10 ppm (104% recovery), but were biased low
at 100 ppm (64% recovery).
The 4100 detected PCBs in all soil blanks (i.e., 100% false positive results), while the reference laboratory
correctly reported all blank samples as non-detects. For the 4100, the percentage of false negative results
was 5%. Overall, the performance of the 4100 Vapor Detector for the PCB demonstration samples was
characterized as biased and imprecise.
Regulatory Decision-Making Applicability
One of the objectives of this demonstration was to assess the technology's ability to perform at regulatory
decision-making levels for PCBs, specifically 50 ppm for soils and 100 (jg/100 cm2 for surface wipes. To
44
-------
assess this, the 4100's performance for soil samples (PE and environmental soils) ranging in concentration
from 40 to 60 ppm can be used, and the data are provided in Table 5-10. The performance of the 4100 for
this concentration range was consistent with the conclusions outlined above (which found the technology to
provide results that were biased both high and low, and that were imprecise). Additionally, most of the
percent difference values were greater than 100% when compared with the corresponding reference
laboratory result.
The 4100 Vapor Detector's performance on extract samples was provided in Tables 5-4 and 5-7. Assuming
a 10-mL extract volume, extract samples (at 10 and 100 (jg/mL) represented surface wipe sample
concentrations of 100 and 1000 (jg/100 cm2. For the simulated wipe extract samples, the 4100's precision
was 65% RSD with a high bias (267% recovery) on the lower concentration samples, and a low bias (54%
recovery) on the higher concentration samples.
Table 5-10. Performance of the 4100 Vapor Detector for soil samples between 40 and 60 ppm
Overall Performance
Mean
Median
95th percentile
Precision (% RSD)
72
64
114
Accuracy (% Recovery)
132
76
377
Comparability (% Difference)
296
60
630
Additional Performance Factors
Detection Limits
The method detection limit (MDL) is often defined as the minimum concentration of a substance that can
be measured and reported with 99% confidence that the analyte concentration is greater than zero. An
MDL is determined from repeated analyses of a sample in a given matrix containing the analyte [11]. EST
did not specify a method detection limit prior to the demonstration study. An MDL was calculated from the
data for the PE samples. Because there was a significant "site effect" (i.e., differences in performance due
to environmental conditions) inherent to the PE samples, separate MDLs were calculated for both the
outdoor and chamber conditions. The MDL calculated for the outdoor conditions was 26 ppm, while the
MDL for the chamber conditions was 62 ppm.
Sample Throughput
Sample throughput is representative of the average amount of time required to extract the PCBs, to perform
appropriate reactions, and to analyze the sample. Operating under the outdoor conditions, EST's sample
throughput rate was approximately 5 to 6 samples/h, but improved to 10 samples/h under the chamber
conditions. This increased sample throughput may be attributed to the analysis order (EST may have
gained experience by analyzing samples under the outdoor conditions first), or difficulty with the sample
matrices that were analyzed only under the outdoor conditions.
Cost Assessment
The purpose of this economic analysis is to provide an estimation of the range of costs for an analysis of
PCB-contaminated soil samples using the 4100 Vapor Detector and a conventional analytical reference
45
-------
laboratory method. The analysis was based on the results and experience gained from this demonstration,
costs provided by EST, and representative costs provided by the reference analytical laboratories that
offered to analyze the samples. To account for the variability in cost data and assumptions, the economic
analysis was presented as a list of cost elements and a range of costs for sample analysis by the 4100 and
by the reference laboratory.
Several factors affected the cost of analysis. Where possible, these factors were addressed so that decision
makers can independently complete a site-specific economic analysis to suit their needs. The following
categories were considered in the estimate:
• sample shipment costs,
• labor costs,
• equipment costs,
• waste disposal costs.
Each of these cost factors is defined and discussed in the following section; the cost factors compose the
basis for the estimated cost ranges presented in Table 5-11. Sample acquisition and preanalytical sample
preparation, which were tasks common to both methods, are costs that were not included here.
4100 Vapor Detector Costs
Because the samples were analyzed on site, no sample shipment charges were associated with the cost of
operating the 4100. Labor costs included mobilization/demobilization, travel, per diem, and on-site labor.
• Labor mobilization/demobilization: This cost element included the time for one person to
prepare for and travel to each site. The estimate ranged from 5 to 8 h, at a rate of $50/h
• Travel: This element was the cost for the analyst(s) to travel to the site. If the analyst is
located near the site, the cost of commuting to the site (estimated to be 50 miles at $0.30
per mile) would be minimal ($15). The estimated cost of an analyst traveling to the site for
this demonstration ($1,000) included the cost of airline travel and rental car fees.
Per diem: This cost element included food, lodging, and incidental expenses, and was
estimated ranging from zero (for a local site) to $150 per day per analyst.
• Rate: The cost of the on-site labor was estimated at a rate of $30 to $75/h, depending on
the required expertise level of the analyst. This cost element included the labor involved
with the entire analytical process, comprising sample preparation, sample management,
analysis, and reporting.
Equipment costs included mobilization/demobilization, purchase of equipment, training, and the reagents
and other consumable supplies necessary to complete the analysis.
46
-------
• Equipment mobilization/demobilization: This included the cost of shipping the equipment
to the test site. If the site were local, the cost would be zero. For this demonstration, the
cost of shipping equipment and supplies was estimated at $150.
• Purchase: At the time of the demonstration study, the cost of purchasing the 4100 Vapor
Detector was $24,950. The SAW detector is sold separately for $1,500.
Training: EST offers a 3-d training course on the use of the 4100 Vapor Detector at a cost
of $2,400.
• Reagents/supplies: These items are consumable and are purchased on a per-sample basis.
At the time of the demonstration, the cost of the reagents and supplies needed to prepare
and analyze PCB soil samples using the 4100 was $1 to $2 per sample.
Waste disposal costs were estimated based on the 1997 regulations for disposal of PCB-contaminated
waste. Using the 4100, EST generated approximately 27 Ib of solid PCB waste that could be incinerated
(i.e., vials containing soils and liquid solvents) and approximately 20 Ib of solid PCB waste (i.e., used and
unused soil, gloves, paper towels, and ampules). The disposal costs for solid PCB wastes by incineration at
a commercial facility was estimated at $1.50 per pound. The cost for solid PCB waste disposal at ETTP
was estimated at $18 per pound.
Table 5-11. Estimated analytical costs for PCB soil samples
4100 Vapor Detector
Electronic Sensor Technology
Sample throughput rate: 5-6 samples per hour (outdoors)
10 samples per hour (chamber)
Cost Category Cost ($)
Sample Shipment 0
Labor
Mobilization/demobilization 250 - 400
Travel 15 - 1,000 per analyst
Per diem 0-150 per day per analyst
Rate 30 - 75 per hour per analyst
Equipment
Mobilization/demobilization 0-150
4 1 00 Vapor Detector price 24,950
SAW detector price 1 ,500
3-Day Training 2,400
Reagents/supplies 1-2 per sample
Waste Disposal 70 - 850
EPA SW-846 Method 8080/8081/8082
Reference Laboratory
Typical turn-around time: 14 -
Cost Category
Sample Shipment
Labor
Overnight shipping charges
Labor
Mobilization/demobilization
Travel
Per diem
Rate
Equipment
Mobilization/demobilization
Purchase of equipment
Reagents/supplies
Waste Disposal
30 days
Cost (S)
100-200
50-150
included a
included
included
44 - 239 per sample
included
included
included
included
a "Included" indicates that the cost is included in the labor rate.
47
-------
Reference Laboratory Costs
Sample shipment costs to the reference laboratory included overnight shipping charges as well as labor
charges associated with the various organizations involved in the shipping process.
Labor: This cost element includes all of the tasks associated with the shipment of the
samples to the reference laboratory. Tasks included packing the shipping coolers,
completing the chain-of-custody documentation, and completing the shipping forms.
Because the samples contained PCBs, the coolers were inspected by qualified personnel to
ensure acceptance with the U.S. Department of Transportation's shipping regulations for
PCBs. The estimate to complete this task ranged from 2 to 4 h at $50/h.
• Overnight Shipping: The overnight express shipping service cost was estimated to be $50
for one 50-lb cooler of samples.
The labor bids from commercial analytical reference laboratories that offered to perform the PCB analysis
for this demonstration ranged from $44 per sample to $239 per sample. The bid was dependent on many
factors, including the perceived difficulty of the sample matrix, the current work-load of the laboratory, and
the competitiveness of the market. In this case, the wide variation in bids may also be related to the cost of
PCB waste disposal in a particular laboratory's state. LAS Laboratories was awarded the contract to
complete the analysis as the lowest qualified bidder ($44 per sample). This rate was a fully loaded
analytical cost, including equipment, labor, waste disposal, and report preparation.
Cost Assessment Summary
An overall cost estimate for the 4100 versus the reference laboratory was not made because of the extent of
variation in the different cost factors, as outlined in Table 5-11. The overall costs for the application of
each technology will also be based on the number of samples requiring analysis, the sample type, and the
site location and characteristics. Decision-making factors, such as turnaround time for results, must also be
weighed against the cost estimate to determine the value of the field technology versus the reference
laboratory.
General Observations
The following are general observations regarding the field operation of the 4100 Vapor Detector:
• The system was light (approximately 35 Ib), easily transportable, and rugged. The system
was shock mounted into a rugged field-portable fiberglass shipping case that could be
checked as airplane baggage. It took the EST team less than 1 h to prepare to analyze
samples on the first day of testing.
Two operators were used for the demonstration because of the number of samples and
working conditions, but the technology can be run by a single person.
• Operators generally require several hours of training and should have a basic knowledge of
gas chromatographic techniques. These methods should be used by, or under the
supervision of, analysts experienced in the use of sampling techniques and gas
48
-------
chromatography. The analysts should also be skilled in the interpretation of gas
chromatograms and in the use of chromatography as a quantitative tool.
The system requires 120 VAC or battery power. Also, the system requires high-purity
helium for the chromatographic column.
• The data acquisition system ran under a Windows 95 operating system.
• EST generated approximately 20 Ib of PCB-contaminated solid hazardous waste (i.e., used
and unused soil, gloves, paper towels, and ampules) using the 4100 Vapor Detector. In
addition, approximately 27 Ib of small sample vials (containing less than 1 g soil,
methanol, water, and/or hexane) were also generated as hazardous waste.
Performance Summary
The performance characteristics of EST's 4100 Vapor Detector presented previously in this chapter are
summarized in Table 5-12. The overall performance of the 4100 Vapor Detector was characterized as
biased and imprecise.
49
-------
Table 5-12. Performance summary for the 4100 Vapor Detector
Feature/Parameter
Blank Samples
Method Detection Limit
Precision
Accuracy
False Positive Results
False Negative Results
Comparison with Reference Laboratory Results
Regulatory Decision-Making Applicability
Sample Throughput
Power Requirements
Operator Requirements
Cost
Hazardous Waste Generation
Performance Summary
Soils: PCBs detected in all 8 blanks (0.7 to 188 ppm)
Extracts: PCBs detected in 3 of 8 blanks (0.5 to 133 pg/mL)
EST specified: none
Calculated: 26 ppm (outdoors); 62 ppm (chamber)
Average RSD
PE Soils: 81% (range: 51 to 165%)
Environmental Soils: 90% (range: 25 top 187%)
Extracts: 65% (range: 9 to 60%)
Average Percent Recovery
PE Soils: 177% recovery (outdoors, range: 12 to 460%) ); 631%
(chamber, range: 74 to 1,990%)
Extracts: 161% recovery (range: 41 to 393%)
recovery
Blank Soils: 100% (8 of 8 samples)
Blank Extracts: 38% (3 of 8 samples)
PE and Environmental Soils: 5% (10 of 192 samples)
Spiked Extracts: 0% (0 of 16 samples)
PE and Environmental Soil Samples
Percent Difference: 51% of samples were > 100% D
Coefficients of determination (R2): 0. 177 (all data)
0.115(< 125 ppm)
Extract Samples
Percent Difference: 44% of samples were > 50% D
Coefficient of determination (R2): 0.187
40 to 60 ppm PE and Environmental Soil Samples
precision: 72% average RSD (range: 37 to 127%)
accuracy: 132% average recovery (range: 0 to 482%)
comparability: 296% average difference (range: -100 to 6,922%)
100 ug/100cm2 and 1000 ug/100cm2 Extract Samples
precision: 65% average RSD (range: 59 to 71%)
accuracy : 161% average recovery (range: 54 to 267%)
comparability: 124% average difference (range: -69 to 558%)
5-6 samples/hour (outdoor)
10 samples/hour (chamber)
120V AC or battery-operated
Basic knowledge of chromatographic techniques; optional 3-day
course for $2,400
training
Equipment purchase: $24,950 (GC); $1,500 (SAW)
$1 to 2 per sample
approximately 27 Ib of vials with soils/solvents (solid)
approximately 20 Ib solid waste (gloves, pipettes, etc.)
50
-------
Section 6
Technology Update and Representative Applications
Objective
In this section, EST describes new technology developments that have occurred since the demonstration
activities. In addition, the developer has provided a list of representative applications where the 4100 Vapor
Detector has been or is currently being utilized.
Technology Update
A field-portable chromatography system equipped with an SAW detector has been used to speciate and
quantify PCB contamination in soil and flyash with a 10-s analysis time. Measurement speed and accuracy
make the instrument well suited to rapid screening of soil samples. The technology participated in a
performance demonstration study under EPA's ETV program in August 1997. In the three months
following the study, a number of improvements have been made in the GC method as well as in the
instrument hardware. These improvements are as follows:
• Pipette filtering has been eliminated as part of the liquid extraction method. The use of a
portable centrifuge to separate particulate materials improves the speed of the method and
reduces the amount of waste produced (disposable pipettes).
• A new temperature program for the GC column has been developed that doubles the
resolving power of the instrument and improves the separation and identification of
different Aroclor mixtures.
• An improved open tubular desorption (OTD) apparatus has been engineered and is now in
production. The new OTD is capable of reaching 300 °C (previous units were limited to
200 °C) and can be temperature-programmed to reduce interference from high-boiling-
point compounds (oils).
• Software provided with the instrument now contains a full manual and mpeg movies with
sound describing the PCB/dioxin measurement method. Additional mpeg files provide the
user with a graphic description of instrument maintenance procedures and a fully
illustrated manual.
Representative Applications
Full-scale production of GC/SAW instruments began in July 1997. Since then, a number of users have
reported on their performance. The following is a short list of relevant user sites as representative
applications.
51
-------
Savannah River DOE Site
Joe Rossabi, (803) 725-5220
(EPA-ETV program for VOCs in water matrices)
Hanford DOE Site
Marcus Stauffer, (509) 373-9928
(Tank farm, headspace analysis—VOC-PCB)
Lawrence Livermore National Laboratory
Paula Kato, (510)423-6241
(Volatile organic screening)
Idaho National Engineering and Environmenetal Laboratory (INEEL)
Rod Shurtliff, (208) 526-3325
(Volatile organic screening)
Japan, Nissho Engineering
Yoshinobu Inoue, 011-033-952-0261
(Dioxin monitoring— incinerators)
Data Quality Objective Example
This application of EST's 4100 Vapor Detector is based on data quality objective (DQO) methods for
project planning advocated by the American Society of Testing and Materials (ASTM) [12, 13] and EPA
[14]. ORNL derived a DQO example from the performance results in Section 5. This example, which is
presented in Appendix E, illustrates the use of the 4100's performance data from the ETV demonstration in
the DQO process to select the number of samples and to quantify the action level for the decision rule.
52
-------
Section 7
References
[1] M. D Erickson, Analytical Chemistry ofPCBs, 2nd ed. CRC Press/Lewis Publishers, Boca Raton,
Fla., 1997.
[2] "Polychlorinated Biphenyls (PCBs) Manufacturing, Processing, Distribution in Commerce, and
Use Prohibitions, Code of Federal Regulations, 40 CFR Part 761, rev. 7, December 1994.
[3] M. P. Maskarinec et al., Stability of Volatile Organics in Environmental Soil Samples: Final
Report, ORNL/TM-12128, Oak Ridge Natl. Lab., November 1992.
[4] U.S. Environmental Protection Agency, "Organochlorine Pesticides and PCBs as Aroclors by Gas
Chromatography: Capillary Column Technique," Test Methods for Evaluating Solid Waste,
Physical/Chemical Methods, U.S. EPA SW846, Final Update II, Method 8081, September 1994.
[5] Oak Ridge National Laboratory, Technology Demonstration Plan: Evaluation of Polychlorinated
Biphenyl (PCB) Field Analytical Techniques, July 1997.
[6] U.S. Environmental Protection Agency, Data Quality Objectives for Remedial Response
Activities, EPA 540/G-87/003, Washington D.C., March 1987.
[7] Lothar Sachs, Applied Statistics: A Handbook of Techniques, 2nd ed., Springer-Verlag, New
York, 1984.
[8] G. W. Snedecor and William G. Cochran, Statistical Methods, The Iowa State University Press,
Ames, Iowa, 1967.
[9] N. R. Draper and H. Smith, Applied Regression Analysis, 2nd ed., John Wiley & Sons Inc., New
York, 1981.
[10] Walter Berger, Harry McCarty, and Roy-Keith Smith, Environmental Laboratory Data
Evaluation, Genium Publishing Corporation, Schenectady, NY., 1996.
[11] "Definition and Procedure for the Determination of the Method Detection Limit," Code of Federal
Regulations. CFR 40 Part 136, Appendix B, Revision 1.11.
[12] American Society for Testing and Materials (ASTM), Standard Practice for Generation of
Environmental Data Related to Waste Management Activities Quality Assurance and Quality
Control Planning and Implementation, D5283-92, 1997.
53
-------
[13] American Society for Testing and Materials (ASTM), Standard Practice for Generation of
Environmental Data Related to Waste Management Activities Development of Data Quality
Objectives, D5792-95, 1997.
[14] U.S. Environmental Protection Agency, Guidance for Data Quality Assessment, EPA QA/G-9;
EPA/600/R-96/084, July 1996.
54
-------
Appendix A
Description of Environmental Soil Samples
55
-------
-------
Table A-l. Summary of soil sample descriptions
Location
Oak Ridge
Oak Ridge
Oak Ridge
Oak Ridge
Oak Ridge
Paducah
Portsmouth
Tennessee
Reference Soil
Request for
Disposal
(RFD)#
40022
40267
24375
43275
134555
97002
7515
n/a
Drum#
02
01
02
03
04
01
02
03
01
02
03
01
02
03
04
858
1069
1096
1898
2143
2528
3281
538
940
4096
n/a
Description
Soil from spill cleanup at the Y-12 Plant in Oak Ridge, Tennessee.
This soil is PCB-contaminated soil excavated in 1992.
Soil from the Elza Gate area, a DOE Formerly Utilized Sites Remedial
Action Program site in Oak Ridge, Tennessee. This soil is PCB-
contaminated soil that was excavated in 1992.
Catch-basin sediment from the K-71 1 area (old Powerhouse Area) at
the DOE East Tennessee Technology Park (formerly known as Oak
Ridge Gaseous Diffusion Plant) in Oak Ridge, Tennessee. This soil is
PCB-contaminated storm drain sediment that was excavated in 1991.
Soil from the K-25 Building area at the DOE East Tennessee
Technology Park (formerly known as Oak Ridge Gaseous Diffusion
Plant) in Oak Ridge, Tennessee. This soil is PCB-contaminated soil
that was excavated in 1993.
Soil from the K-707 area at the DOE East Tennessee Technology Park
(formerly known as Oak Ridge Gaseous Diffusion Plant) in Oak Ridge,
Tennessee. This soil is PCB-contaminated soil from a dike spillage that
was excavated in 1995.
Soil from the DOE Paducah Gaseous Diffusion Plant in Kentucky. This
soil is PCB-contaminated soil from a spill cleanup at the C-746-R
(Organic Waste Storage Area) that was excavated in 1989.
Soil from the DOE Portsmouth Gaseous Diffusion Plant in Ohio. This
soil is PCB-contaminated soil from a probable PCB oil spill into the
East Drainage Ditch that was excavated in 1986.
Captina silt loam from Roane County, Tennessee; used as a blank in
this study (i.e., not contaminated with PCBs)
57
-------
-------
Appendix B
Characterization of Environmental Soil Samples
59
-------
-------
Table B-l. Summary of environmental soil characterization
Location
Oak Ridge
Paducah
Portsmouth
Sample
ID
101
102
103
104
105
106
107
108
109
110
111
112
126, 226
113,201
114,202
115,203
116,204
117,205
206
207
208
209
210
216
211
217
212
213
214
215
RFD
Drum # a
40022-02
40267-03
40267-01
40267-04
40267-01 Sb
24375-03
24375-01
40267-02
24375-02
43275-01
134555-03Sb
43275-02
non-PCB soil
97002-04
97002-01
97002-03
97002-02
97002-02S b
7515-4096
7515-1898
7515-1096
7515-2143
7515-0940
7515-0538
7515-0538Sb
7515-0538Sb
7515-2528
7515-3281
7515-0858
7515-1069
Composition
% gravel
0
0.5
0.2
0.6
0.5
0.5
2.5
0.4
0.3
0
0.5
0.1
0
0
0.2
0.1
0.4
0
0.2
0.4
0
0.3
0.5
0.5
0.5
0
1.3
% sand
91.8
99.3
96.7
98.2
94.8
87.8
92.5
94.2
93.1
89.2
88.1
91.4
85.6
92.4
87.6
83.6
93.7
87.1
78.0
74.4
74.3
73.0
73.3
70.4
72.6
65.8
75.0
% silt + clay
8.2
0.2
3.1
1.2
4.7
11.7
5.0
5.4
6.6
10.8
11.4
8.5
14.4
7.6
12.2
16.3
5.8
12.9
21.8
25.2
25.7
26.7
26.3
29.1
26.8
34.2
23.7
Total Organic
Carbon
(mg/kg)
5384
13170
13503
15723
14533
19643
1196
9007
1116
14250
10422
38907
9249
1296
6097
3649
4075
3465
3721
3856
10687
7345
1328
5231
5862
6776
4875
PH
7.12
7.30
7.21
7.07
7.28
7.36
7.26
7.30
7.48
7.57
7.41
7.66
7.33
7.71
7.64
7.59
7.43
7.72
7.66
7.77
7.71
7.78
7.78
7.92
7.67
7.85
7.56
a Request for disposal drum number (see Table A-l).
b "S" indicates that the environmental soil was spiked with additional PCBs.
61
-------
-------
Appendix C
Temperature and Relative Humidity Conditions
63
-------
-------
Table C-l. Average temperature and relative humidity conditions during testing periods
Date
7/22/97
7/23/97
7/24/97
7/25/97
7/26/97
7/27/97
7/28/97
7/29/97
Outdoor Site
Average
Temperature
(°F)
85
85
85
80
85
80
79
b
Average
Relative Humidity
(%)
62
70
67
70
55
75
88
b
Chamber Site
Average
Temperature
(°F)
70 a
60 a
58
56
57
55
57
55
Average
Relative Humidity
(%)
38 a
58 a
66
54
51
49
52
50
' The chamber was not operating properly on this day. See discussion in Section 3.
' No developers were working outdoors on this day.
65
-------
100
Temperature (deg. F)
6 § §
20
—
—
_
"
—
—
—
—
D High Temp
H Low Temp
QAvg Temp
7/22/97 7/23/97 7/24/97 7/25/97 7/26/97
7/27/97
7/28/97
Figure C-l. Summary of temperature conditions for outdoor site.
120
7/22/97 7/23/97 7/24/97 7/25/97 7/26/97
7/27/97
7/28/97
Figure C-2. Summary of relative humidity conditions for the outdoor site.
66
-------
7/22/97 7/23/97 7/24/97 7/25/97 7/26/97 7/27/97 7/28/97 7/29/97
Figure C-3. Summary of temperature conditions for chamber site.
90
80
70 --
— 60
S?
2 50
™
u
30
20
10 --
7/22/97 7/23/97 7/24/97 7/25/97 7/26/97 7/27/97
Figure C-4. Summary of relative humidity conditions for chamber site.
67
7/28/97 7/29/97
-------
68
-------
Appendix D
EST's 4100 Vapor Detector
PCB Technology Demonstration Sample Data
69
-------
Legend for Appendix D Tables
Obs
Sample ID
Rep
4100 Result
Ref Lab Result
Reference Aroclor
Type
Order
Observation
Sample identification
101 to 126 for Outdoor Site soil samples
127 to 130 for Outdoor Site extract samples
201 to 226 for Chamber Site soil samples
227 to 230 for Chamber Site extract samples
Replicate of Sample ID
(1 through 4)
4100's measured PCB concentration (ppm).
LAS reference laboratory measured PCB concentration (ppm).
Values with "<" are samples that the reference laboratory
reported as "< reporting detection limit"
Aroclor(s) identified by the reference laboratory
"sample" indicates environmental soil; "1242", "1248", "1254",
"1260" indicates Aroclor in the PE samples;"blank" indicates
non-PCB contaminated sample
Order of sample analysis by EST
(started with 1001 through 1116, then 2001 through 2116)
70
-------
Table D-l. EST's 4100 Vapor Detector PCB technology demonstration soil sample data
Sample
Obs
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
ID
101
101
101
101
102
102
102
102
103
103
103
103
104
104
104
104
105
105
105
105
106
106
106
106
107
107
107
107
108
108
108
108
109
109
109
109
110
110
110
110
Rep
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
4100
Result
(ppm)
14.5
2.5
34.6
6.2
206.8
3.1
6.9
4
4.1
12.1
38.6
112.3
4.4
4.9
13
0.2
27.8
38.3
1.3
43.6
237.3
220
5.4
142.3
3.6
3.4
1.1
6.7
2.6
0.9
23.5
2.0
15.8
20.8
89.6
14.5
0
8
203.4
2.7
Ref Lab
Result
(ppm)
0.6
0.4
0.5
0.5
2.2
2.1
1.7
2.5
3.0
2.4
2.0
1.6
6.8
6.0
14.8
9.9
49.7
84.1
50.6
53.2
269.6
255.9
317.6
649.6
1.0
1.6
1.2
1.2
1.7
2.0
1.7
1.9
1.5
2.1
1.8
2.4
<490.0
<99.0
<66.0
<98.0
Reference
Aroclor
1254
1254
1254
1254
1254
1254
1260
1260
1254
1254
1260
1260
1260
1254
1254
1254
1260
1260
1260
1260
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
1254
Non- Detect
Non- Detect
Non- Detect
Non- Detect
Type
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Order
1078
1103
1063
1022
1013
1015
1006
1014
1055
1048
1064
1069
1037
1077
1016
1011
1035
1020
1034
1073
1074
1050
1017
1039
1041
1018
1003
1090
1026
1001
1030
1102
1080
1083
1100
1038
1008
1019
1085
1005
71
-------
Sample
Obs
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
ID
111
111
111
111
112
112
112
112
113
113
113
113
114
114
114
114
115
115
115
115
116
116
116
116
117
117
117
117
118
118
118
118
119
119
119
119
120
120
120
120
Rep
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
4100
Result
(ppm)
19.1
54.7
9.4
209.8
102.6
548.7
1995.4
126.5
0
6.3
6.1
11
35.8
3.3
25
0
0
18.5
0.2
14.1
0.5
9.4
6.7
0.8
139.3
141 . 1
60.9
136.5
8.3
24.3
1
3.1
0
7.8
0
1.4
3.2
24.6
1.2
3.9
Ref Lab
Result
(ppm)
44.5
36.0
39.3
35.1
<66.0
<200.0
<130.0
<200.0
0.7
1 . 1
0.6
1.9
1 . 1
1.2
1.3
1.7
14.9
12.4
15.0
16.9
41.4
41.2
48.5
34.0
431.6
406.3
304.7
392.8
2.1
1.9
0.7
1.6
21.2
17.2
17.4
24.4
4.5
4.0
6.3
5.0
Reference
Aroclor
1254
1254
1254
1254
Non- Detect
Non- Detect
Non- Detect
Non- Detect
1260
1260
1260
1248/1260
1260
1260
1260
1260
1248
1016
1248
1248
1248
1016
1248
1016
1016
1016
1016
1016
1248
1016
1248
1248
1016
1248
1248
1248
1254
1254
1254
1254
Type
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
1248
1248
1248
1248
1248
1248
1248
1248
1254
1254
1254
1254
Order
1065
1070
1044
1021
1051
1060
1082
1081
1054
1010
1092
1086
1097
1093
1094
1042
1012
1088
1031
1068
1028
1024
1075
1056
1098
1071
1032
1087
1049
1033
1057
1104
1067
1099
1072
1076
1036
1052
1002
1059
72
-------
Sample
Obs
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
ID
121
121
121
121
122
122
122
122
123
123
123
123
124
124
124
124
125
125
125
125
126
126
126
126
201
201
201
201
202
202
202
202
203
203
203
203
204
204
204
204
Rep
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
4100
Result
(ppm)
65.2
24.4
44.1
22.5
6.6
71.4
3.3
2.3
21.5
27.6
24.3
0
14.8
5.7
8
4.8
60.6
20.3
20.2
22.3
17.7
19.5
0.7
96.0
0
0
20.3
24.3
61.4
3.3
13.6
13.6
32.9
18.3
3.3
2.1
75
257.9
26.3
57.1
Ref Lab
Result
(ppm)
58.7
55.7
53.2
50.9
12.2
10.9
11.3
10
59.2
56.9
66.8
57.5
1.8
1.4
1.9
1.8
32.0
41.3
46.0
32.2
<0.1
<0.1
<0.2
<1.3
1.0
1.0
1.1
0.6
1.4
1.6
1.2
1.5
14.0
12.8
16.2
12.4
43.1
45.3
41
47.7
Reference
Aroclor
1254
1254
1254
1254
1260
1260
1260
1260
1260
1260
1260
1260
1254
1260
1254
1254
1254
1254
1254
1260
Non- Detect
Non- Detect
Non- Detect
Non- Detect
1016/1260
1016/1260
1016/1260
1260
1260
1260
1260
1260
1248
1248
1248
1248
1248
1248
1248
1248
Type
1254
1254
1254
1254
1260
1260
1260
1260
1260
1260
1260
1260
1254/1260
1254/1254
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
1254/1260
Blank
Blank
Blank
Blank
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Order
1096
1047
1101
1029
1058
1007
1004
1095
1066
1084
1040
1009
1062
1023
1079
1043
1091
1053
1046
1025
1045
1061
1027
1089
2021
2015
2083
2072
2037
2062
2049
2092
2095
2042
2026
2008
2089
2033
2017
2093
73
-------
Sample
Obs
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
ID
205
205
205
205
206
206
206
206
207
207
207
207
208
208
208
208
209
209
209
209
210
210
210
210
211
211
211
211
212
212
212
212
213
213
213
213
214
214
214
214
Rep
1
2
3
4
1
2
3
4
1
2.0
3.0
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
4100
Result
(ppm)
500
218.6
183.3
278.3
16.7
173.6
21.2
2.4
311.9
52.9
238
431.3
385.1
416.2
857.1
223.1
396.7
278.8
772.4
177.6
216.7
433.3
566.7
370.7
451.6
783.3
940.0
696.6
56.6
55.0
37.9
85.9
226
60.5
45.5
211.4
571.4
269
0
173.6
Ref Lab
Result
(ppm)
3305
538.7
457
483.3
2.9
1 . 1
1.1
2.5
17.8
14.3
21.6
21.6
42
27.7
24
28.4
32.7
79.3
11.0
37.9
123.2
61.5
84.1
85.5
387.8
581.4
330.0
318.7
3.8
3.9
4.3
0.8
6.9
7.3
7.8
10.5
26.0
25.6
29.1
20.2
Reference
Aroclor
1016/1260
1016
1016
1016
1260
1260
1016/1260
1260
1260
1260
1260
1254
1260
1016/1260
1254
1260
1260
1260
1260
1260
1260
1260
1260
1260
1254
1254
1254
1254
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
1260
Type
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Order
2047
2057
2029
2103
2022
2031
2079
2025
2085
2098
2005
2081
2001
2045
2086
2096
2055
2009
2078
2052
2048
2038
2082
2070
2027
2091
2061
2035
2074
2019
2063
2068
2020
2002
2023
2071
2065
2030
2016
2011
74
-------
Sample
Obs
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
ID
215
215
215
215
216
216
216
216
217
217
217
217
218
218
218
218
219
219
219
219
220
220
220
220
221
221
221
221
222
222
222
222
223
223
223
223
224
224
224
224
Rep
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
4100
Result
(ppm)
249.3
342.9
228.9
0
372.1
325
573.9
450
146.9
537.1
422.2
1045
71.1
11.3
45.7
31.0
47.9
82.6
9.7
42.2
33.7
23.1
41.9
9.2
240.9
137
110.3
83.3
2.8
21.8
1.2
6.6
52.5
32.2
28.9
63.4
33.8
39.0
40.8
14.8
Ref Lab
Result
(ppm)
25.1
24.1
26.2
31.2
151.6
47.0
54.3
64.0
886.7
549.8
542.8
1913.3
2.8
2.4
2.6
2.6
22.4
26.0
29.4
15.2
8.5
4.9
4.7
5.2
32.0
44.1
43.8
59.6
13.2
12.4
12.7
12.7
56.6
50.3
49.9
66.4
2.2
1.2
1.4
2.1
Reference
Aroclor
1260
1260
1260
1016/1260
1260
1260
1260
1260
1254
1254
1254
1016/1260
1248
1248
1248
1248
1248
1016
1248
1248
1254
1254
1254
1254
1016/1260
1016/1260
1254
1254
1260
1260
1260
1260
1260
1260
1260
1260
1254
1260
1260
1254
Type
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
1248
1248
1248
1248
1248
1248
1248
1248
1254
1254
1254
1254
1254
1254
1254
1254
1260
1260
1260
1260
1260
1260
1260
1260
1254/1260
1254/1260
1254/1260
1254/1260
Order
2060
2084
2059
2013
2028
2094
2104
2043
2058
2010
2051
2069
2088
2101
2004
2075
2090
2034
2056
2066
2012
2076
2039
2050
2087
2032
2073
2064
2003
2036
2053
2018
2080
2006
2097
2102
2007
2067
2044
2024
75
-------
Sample 4100 Ref Lab Reference
Obs ID Rep Result Result Aroclor Type Order
(ppm) (ppm)
201 225 1 25.6 56.4 1260 1254/1260 2014
202 225 2 194 36.5 1016/1260 1254/1260 2077
203 225 3 103.8 32.1 1260 1254/1260 2041
204 225 4 156.9 146.0 1254 1254/1260 2040
205 226 1 77.6 <0.1 Non-Detect Blank 2100
206 226 2 187.9 <0.8 Non-Detect Blank 2099
207 226 3 38.7 <0.1 Non-Detect Blank 2046
208 226 47 <0.1 Non-Detect Blank 2054
76
-------
Table D-2. EST's 4100 Vapor Detector technology demonstration extract sample data
DBS
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Sample
ID
130
130
130
130
131
131
131
131
132
132
132
132
230
230
230
230
231
231
231
231
232
232
232
232
Rep
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
4100
Result
(ppm)
12.6
13.2
25.6
5.1
43.8
38.8
43.3
36.0
0.0
0.8
0.0
0.0
53.0
42.0
50.0
12.3
16.8
51.0
115.0
90.5
0.0
0.0
132.5
0.5
Ref Lab
Result
(ppm)
16.0
11.0
10.3
11.0
67.0
57.0
63.0
68.0
<0.1
<0.1
<0.1
<0.1
9.8
10.0
7.6
7.9
55.0
55.0
61.0
59.0
<0.1
<0.1
<0.1
<0.1
Reference
Aroclor
1016
1016
1248
1016
1254
1254
1254
1254
Non- Detect
Non- Detect
Non- Detect
Non- Detect
1016
1016
1016
1016
1254
1254
1254
1254
Non- Detect
Non- Detect
Non- Detect
Non- Detect
Spike"
(ppm)
10
10
10
10
100
100
100
100
0
0
0
0
10
10
10
10
100
100
100
100
0
0
0
0
Type
1242
1242
1242
1242
1254
1254
1254
1254
Blank
Blank
Blank
Blank
1242
1242
1242
1242
1254
1254
1254
1254
Blank
Blank
Blank
Blank
Order
1111
1114
1113
1105
1110
1109
1116
1107
1115
1106
1108
1112
2108
2111
2112
2109
2115
2107
2113
2105
2110
2116
2106
2114
"Nominal spike concentration of the extract sample prepared by ORNL.
77
-------
Error
Transcription
Calculation
Interpretation
Sample ID
106
130
207
119
214
a
101 a
113"
119
201
219
(ppm)
<490
32,000
160
3.6
2.3
29.0
<
0.7
<
18.0
0.9
<
7.2
1.0
Corrected Result
10.3
17.8
17.4
26.0
0.5
1.2
0.7
21.2
26.0
Two of four measurements in Sample ID 101 were corrected.
78
-------
Appendix E
Data Quality Objective Example
79
-------
-------
Disclaimer
The following hypothetical example serves to demonstrate how the information provided in this report may
be used in the data quality objectives (DQO) process. This example serves to illustrate the application of
quantitative DQOs to a decision process but cannot attempt to provide a thorough education in this topic.
Please refer to other educational or technical resources for further details. In addition, since the focus of
this report is on the analytical technology, this example makes the simplifying assumption that the contents
of these drums will be homogeneous. In the real world, however, this assumption is seldom valid, and
matrix heterogeneity constitutes a source of considerable uncertainty that must be adequately evaluated if
the overall certainty of a site decision is to be quantified.
Background and Problem Statement
An industrial company discovered a land area contaminated with PCBs from an unknown source. The
contaminated soil was excavated into waste drums. Preliminary characterization determined that the PCB
concentration in a single drum was homogenous, but PCB concentrations varied greatly from drum to
drum. The company's DQO team was considering the use of EST's 4100 Vapor Detector to measure the
PCB concentration in each drum. The DQO team decided that drums will be disposed of by incineration if
the PCB concentration is greater than or equal to 50 ppm ("hot"). A concentration of 50 ppm is the Toxic
Substances Control Act (TSCA) regulatory threshold (RT) for this environmental problem. Those drums
with PCB concentrations less than 50 ppm will be put into a landfill because incineration of soil is very
expensive. With regulator agreement, the DQO team determined that a decision rule for disposal would be
based on the average concentration of PCBs in each drum.
General Decision Rule
If the average PCB concentration is less than the action level, then send the soil drum to
the landfill.
If the average PCB concentration is greater than or equal to the action level, then send the
soil drum to the incinerator.
DQO Goals
EPA's Guidance for Data Quality Assessment [14] states the following in Section 1.2: "The true condition
that occurs with the more severe decision error . . . should be defined as the null hypothesis." The team
decided that the more severe decision error would be for a drum to be erroneously sent to a landfill if the
drum's PCB concentration actually exceeded 50 ppm. Therefore, the null hypothesis is constructed to
assume that a drum's true PCB concentration is greater than 50 ppm; and as a "hot" drum, it would be sent
to an incinerator. Drums would be sent to the landfill only if the null hypothesis is rejected and it is
concluded that the "true" average PCB concentration is less than 50 ppm.
With the null hypothesis defined in this way, a false positive decision is made when it is concluded that a
drum contains less than 50 ppm PCBs (i.e., the null hypothesis is rejected), when actually the drum is "hot"
81
-------
(i.e., the null hypothesis is true). The team required that the error rate for sending a "hot" drum to the
sufficient number of samples must be taken from each drum so that the false positive decision error rate
(FP) is 0.05 (or less) if the true drum concentration is 50 ppm. This scenario represents a 5% chance of
The DQO team did not want to send an excessive number of drums to the incinerator if the average PCB
concentration was less than 50 ppm because of the expense. In this situation, a false negative decision is
the drum contains soil with less than 50 ppm PCBs (i.e., the null hypothesis is actually false). After
considering the guidelines presented in Section 1.1 of EPA's Guidance for Data Quality Assessment [14]
0.10 if the true drum concentration was 40 ppm. That is, there would be a 10% probability of sending a
drum to the incinerator (denoted as Pr[Take Drum to Incinerator]) if the true PCB concentration for a drum
Permissible FP and FN Error Rates and Critical Decision Points
FP: Pr[Take Drum to Landfill] 0.05 when true PCB concentration = 50 ppm
FN: Pr[Take Drum to Incinerator] 0.10 when true PCB concentration = 40 ppm
Use of Technology Performance Information to Implement the Decision Rule
Technology performance information is used to evaluate whether a particular analytical technology can
of the 4100 Vapor Detector, the performance of this technology (as reported in this ETV report) was used
to assess its applicability to this project. Two questions arise:
How many samples are needed from a single drum to permit a valid estimation of the true average
assumption was made that the PCB distribution throughout the soil within a single drum is
homogeneous, and thus, matrix heterogeneity will not contribute to overall variability. The only
analytical method, which is determined by precision studies.
2 for using the 4100 Vapor Detector to make decisions in
the field? After the required number of samples have been collected from a drum and analyzed, the
using the 4100 Vapor Detector, what is the value (here called "the action level for the decision
rule") with which that average is compared to decide whether the drum is "hot" or not? This
-------
method-specific or site-specific action level is derived from evaluations of the method's accuracy
using an appropriate QC regimen.
4100 Vapor Detector Accuracy
The ETV demonstration indicated that the results from EST's 4100 Vapor Detector were biased and
sensitive to environmental conditions. In addition, average PCB measurements on blank PE samples were
33.5 ppm and 77.8 ppm for outdoor and chamber conditions, respectively. These results from blank
samples were near or greater than the regulatory threshold of 50 ppm. The DQO Team decided to review
the PE data for the outdoor conditions with no blanks. These conditions were the best match for their
application. Figure E-l is a plot of a fitted line to the average concentration for the data. EST's
measurements have a weak correlation (R2 = 0.57) with certified PCB values for the performance
evaluation samples.
Figure E-l shows that the predicted results from the
4100 Vapor Detector would likely be biased low when
compared with certified PE values. For example, at
PE values of 40 and 50 ppm, the 4100 Vapor
Detector could be expected to produce an average
result of about 24 and 28 ppm, respectively. The
DQO Team knew that if they selected the 4100 Vapor
Detector for their project they would have to
compensate for this negative bias. They decided to
apply a correction factor to every result obtained from
the 4100 Vapor Detector to obtain a conservative,
unbiased result. Based on the information from the
ETV study, the Team expected that they would need
to multiply each 4100 Vapor Detector result by a
correction factor. The exact value of the correction
factor would be determined after detailed examination Certified PCB Concentration (ppm)
of a rigorous site-specific QC program. Figure E-l. A line fitted to the average concentrations of
outdoor PE soil samples (no blanks) with 95% confidence
intervals (dashed lines).
The DQO Team decided they would need to be very
careful about interpreting the PCB results generated using EST's 4100 Vapor Detector. They would use
the site-specific QC samples to assess the performance of the 4100 Vapor Detector under their site-specific
conditions. The Team would have to design an extensive QC regimen (which included PE samples, matrix
spikes, split samples sent for confirmatory laboratory analysis, and duplicates) that would verify whether
the 4100 Vapor Detector was performing as expected under their site-specific conditions. Additionally, to
address the possibility that these QC samples would reveal that the kit was performing differently from
what they expected, the Team would have to create a backup plan (which would become part of the
Sampling and Analysis Plan) that would permit them to document and account for deviations in expected
performance. The backup plan would lay out the courses of action to follow if the kit's performance did not
meet their expectations so that verifiable and defensible data could be produced to support decision-making
at the site without the need for extensive resampling at a future time.
83
-------
Determining the Number of Samples
With the critical decision points (which correspond to CFN and RT in the following equations) selected for
use with the kit, the Team could determine the number of samples needed from each drum to calculate its
"true" average PCB concentration. For a homogeneous matrix, the number of samples required depends on
the precision of the analytical method.
The ETV demonstration results indicated that the 4100 Vapor Detector's SD increased with concentration
level, so the RSD would be a more appropriate precision measurement than the standard deviation (SD).
The DQO Team used the precision value determined from the ETV demonstration (average RSD of 101%)
as a consistent RSD for outdoor PE samples. This estimate of measurement variability was used to
calculate the number of soil samples required to be measured on each drum to achieve the DQO objectives.
A formula (Eq. E-l) is provided in EPA's Guidance for Data Quality Assessment [14] (pp. 3.2-3, Box
3.2-1) that can be adapted to this example for calculating the number of samples required to meet the FP
and FP requirements. This formula uses a constant SD for the analytical method's precision, but can be
modified to use RSD by dividing the numerator and denominator by (RT)2 and multiplying by (100%)2, as
shown in E-2. The final form of the formula appears as Eq. E-3.
(SD)2 ( Z, Fp + Z, )2 2
n = + (0.5)Z,_pp (E-l)
(RT - cFN)2
1_FP ,_FN
n = + (0.5)7^ (E-2)
- CFN)IRT}2
RSD ( Z, ™ + Z, „, ) 9
1-JP 7 1~FNJ + (0.5)7^ (E-3)
(%D)2
where
n = number of samples from a drum to be measured,
RSD = RSD at the regulatory threshold [e.g., RSD2 = (101%)2 ],
RT = regulatory threshold (e.g., RT = 50 ppm),
CFN = concentration at which the FN is specified (e.g., CFN = 40 ppm),
%D = percent difference of CFN relative to RT [e.g., (%D)2 = (20%)2]
FP = false positive decision error rate (e.g., FP = 0.05),
FN = false negative decision error rate (e.g., FN= 0.10), and
Z^p = the (\-p)th percentile of the standard normal distribution (see EPA QA/G-9,
Table A-l of Appendix A). Example Z(1_FP) = Z095 = 1.645.
84
-------
Equation E-3 is then used to determine the number of samples to be analyzed from each drum («):
( 1.645 + 1.282)'
(20% )
2
Therefore, 220 samples would need to be analyzed from each drum by EST's 4100 Vapor Detector to meet
the criteria established by the DQO process. The AL for the decision rule can then be calculated based on a
sample size («) of 220 samples from each drum.
Determining the Action Level
The results from the 220 samples from each drum will be corrected for bias based on the site-specific QC
sample analyses. These 220 unbiased estimates (of the PCB concentration in each sample) will be averaged
(arithmetic mean) to generate an unbiased PCB concentration for the drum. This unbiased estimate of the
"true" PCB concentration in each drum will be compared with the AL. The AL for the decision rule is
calculated based on regulation-driven requirements (the TSCA regulatory threshold of 50 ppm), and on
controlling the FP requirement established in the DQO process. Recall that the Team set the permissible FP
at 5%.
ASTM D5283-92 [12] provides the formula for the action level based on a constant SD over the
concentration range (Eq. E-4). Since the 4100 Vapor Detector did not produce data with a constant SD,
this formula can be adapted to this example by using the relationship between SD and RSD, which is SD =
(Concentration) x RSD/100%. Thus Eq. E-4 becomes Eq. E-5, and the regulatory threshold (RT = 50 ppm)
is the concentration used in the formula.
AL = RT - Z^p x - (E-4)
{n
AT r>^r r7 RTXRSD tT^ .-.
AL = RT - Zj Fp x (E-5)
Decision Rule for 5% FP and 10% FN
If the corrected average PCB concentration of 220 random soil samples from a drum is less than 44.4
ppm, then send the drum to the landfill.
If the corrected average PCB concentration of 220 random soil samples from a drum is greater than or
equal to 44.4 ppm, then send the drum to the incinerator.
85
-------
AT r^ /i SAT\ .. .
AL = 50 ppm - (1.645) x —£.£- = 44.4 ppm
100% x 1/220
The average PCB concentration for a single drum will be calculated from the 220 samples from that drum.
The results would be corrected for bias and the average of the corrected results would be compared with an
AL = 44.4 ppm. The decision rule using EST's 4100 Vapor Detector to satisfy a 5% FP and a 10% FN
thus becomes:
A decision performance curve for this environmental problem [14] calculates the probability of sending a
drum to the incinerator for different values of true PCB concentration in a drum. Figure E-2 shows that the
decision performance curve has the value of Pr[ Take Drum to Incinerator] = 0.95 for True = 50 ppm. This
indicates that the decision rule meets the DQO Team's FP of 5%. The Pr[ Take Drum to Incinerator] =
0.05 for True = 40 ppm which is better than the FN of 10% that the DQO Team had specified. This
improved performance is due to rounding up the number of samples to the next integer in the calculation of
number of samples required.
Alternative FP Parameter
Because of random sampling and analysis error, there is always some chance that analytical results will not
accurately reflect the true nature of a decision unit (such as a drum, in this example). Often, 95% certainty
(a 5% FP) is customary and sufficient to meet stakeholder comfort. But suppose that the DQO Team
wanted to be even more cautious about limiting the possibility that a drum might be sent to a landfill when
its true value is 50 ppm. If the Team wanted to be 99% certain that a drum was correctly sent to a landfill,
the following describes how changing the FP requirement from 5% to 1% would affect the decision rule.
Using FP = 0.01, the sample size is calculated to be 333 and the action level is calculated to be 43.6 ppm.
The decision performance curve has the value of Pr[ Take Drum to Incinerator] = 0.99 for True = 50 ppm.
This indicates that the decision rule meets EPA's FP of 1%. The Pr[ Take Drum to Incinerator] = 0.01 for
True = 40 ppm is better than the FN of 10% that the DQO Team had specified. This improved
performance is due to rounding up the number of samples to the next integer in the calculation of number of
samples required. The decision rule for the lower FP would be:
86
-------
40 45 50
True PCB Concentration ( ppm )
55
Figure E-2. Decision performance curve for PCB drum example.
Decision Rule for FP = 1% and FN = 10%
If the corrected average PCB concentration of 333 random soil samples on a drum is less than
43.6 ppm, then send the drum to the landfill.
If the corrected average PCB concentration of 333 random soil samples on a drum is greater than
or equal to 43.6 ppm, then send the drum to the incinerator.
87
------- |