United States Office of Research and EPA/600/R-98/109
Environmental Protection Development August 1998
Agency Washington, D.C. 20460
vvEPA Environmental Technology
Verification Report
Electrochemical Technique/Ion
Specific Electrode
Dexsil Corporation
L2000 PCB/Chloride Analyzer
4^^^^^^^^^^ d^^^^^^^^^^^^^t
-------
EPA/600/R-98/109
August 1998
Environmental Technology
Verification Report
Electrochemical Technique/Ion
Specific Electrode
Dexsil Corporation
L2000 PCB/Chloride Analyzer
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
cloverleafBuildin& 19901 GermantownRoad
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.
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
.•a ^^^ *•. Office of Research and Development
Washington, D.C. 20460
ENVIRONMENTAL TECHNOLOGY VERIFICATION PROGRAM
VERIFICATION STATEMENT
jV
TECHNOLOGY TYPE: POLYCHLORINATED BIPHENYL (PCB) FIELD ANALYTICAL
TECHNIQUES
APPLICATION: MEASUREMENT OF PCBs IN SOILS AND SOLVENT EXTRACTS
TECHNOLOGY NAME: L2000 PCB/CHLORIDE ANALYZER
COMPANY: DEXSIL CORPORATION
ADDRESS: ONE HAMDEN PARK DRIVE
HAMDEN, CT 06517
PHONE: (203) 288-3509
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 Dexsil L2000 PCB/Chloride Analyzer.
PROGRAM OPERATION
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's) 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 were also 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, 1997.
The study was conducted under two environmental conditions. The first site was outdoors, with naturally fluctuating
temperatures 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 homogeneous replicate samples conducted by conventional EPA SW-846 methodology in
EPA-VS-SCM-12 The accompanying notice is an integral part of this verification statement August 1998
iii
-------
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: Electrochemical Technique/Ion Specific
Electrode, Dexsil Corporation, L2000 PCB/Chloride Analyzer, EPA/600/R-98/109.
TECHNOLOGY DESCRIPTION
The L2000 PCB/Chloride Analyzer (dimensions: 8" x 8" x 4.5") is a field-portable instrument, weighing approximately
3.5 Ib, designed to quantify PCB concentration in soils, dielectric fluids, and surface wipes. Sample preparation consists
of extraction and dehalogenation of the PCB. A 10-g sample of soil is weighed into a polyethylene test tube. The soil is
extracted with a nonchlorinated solvent from a premeasured ampule. (Note that a newly developed hydrocarbon solvent
system was used for the demonstration analyses.) The soil is allowed to settle, and the supernatant is decanted onto a
Florisil column. The solution is passed through the column, where all of the water and inorganic chloride is removed. Five
milliliters of the solution are collected in a polyethylene reaction tube. Two glass ampules contained in the reaction tube
are broken, introducing metallic sodium to the extract solution. The sodium strips the covalently bound chlorine atoms
off the PCB molecule. The mixture is then shaken for 10 s and allowed to react for a total of 1 min. An aqueous
extraction solution is added to the reaction tube to adjust the pH, destroy the excess sodium, and extract and isolate the
newly formed chloride ions in an aqueous buffered solution. The aqueous layer is decanted, filtered, and collected in an
analysis vial. A chloride-ion-specific electrode is put into this aqueous solution to measure the millivolt potential of the
chloride solution. The potential is then converted to a PCB concentration in terms of parts per million (ppm).
VERIFICATION OF PERFORMANCE
The following performance characteristics of the L2000 PCB/Chloride Analyzer 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. The MDL was
calculated to be 7.1 ppm based on the performance evaluation sample analyses. By use of a line fitted to a plot of the
L2000-measured PCB concentrations versus the certified PE values, bias in the L2000 data can be corrected. After
compensation for bias, the resulting L2000 MDL agrees with Dexsil's specified MDL of 2 ppm.
Throughput: Throughput was 5 samples/hour under the outdoor conditions and 10 samples/hour under the chamber
conditions. This rate included sample preparation and analysis.
Ease of Use: Two operators analyzed samples during the demonstration, but the technology can be run by a single
trained operator. Minimal training (<1 hour) is required to operate the L2000, provided the user has a fundamental
understanding of basic chemical techniques.
Completeness: The L2000 generated results for all 232 PCB samples for a completeness of 100%.
Blank results: PCBs were detected above the L2000's MDL for four of the eight blank samples. Therefore, the
percentage of false positive results was 50%. These results were obtained for both soil and extract samples. The L2000
reported no false negative results.
Precision: The overall precision, based on average relative standard deviations (RSDs), was 23% for soil samples and
14% for extract samples. The L2000's precision was comparable to that of the reference laboratory (21% RSD for soils
and 14% RSD for extracts). At higher concentrations (>125 ppm), the L2000 was more precise than the reference
laboratory (4% versus 19% RSD).
Accuracy: Accuracy was assessed using PE soil and extract samples. The data showed that the L2000 exhibited a
significantly high bias. The overall accuracy, based on average percent recoveries, was 208% for PE soil samples and
EPA-VS-SCM-12 The accompanying notice is an integral part of this verification statement August 1998
iv
-------
149% for extract samples. Evaluation of the data generated at each site indicated that there were no significant differences
between the two data sets based on environmental conditions.
Comparability: This demonstration showed that the L2000 generated data that exhibited a linear 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 L2000 data, was 0.854 when all soil samples (0 to 700 ppm) were considered. For the
concentration range from 0 to 125 ppm, the R2 value was 0.781. Most of the percent difference values were greater than
100% when the L2000 results were compared directly with the reference laboratory results.
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 high (12% RSD), but the measured
concentrations were biased high (147% recovery). For extract samples representing surface wipe sample concentrations
of 100 (jg/100cm2 and 1000 (jg/100cm2 (assuming a 1000 cm2 wipe sample), measurements were precise (14% RSD),
but indicated a high bias (149% recovery), especially for the lower concentrations.
Data quality levels: Because the PCB data generated in this demonstration strongly correlated with the reference
laboratory results, it may be possible for Dexsil's L2000 PCB/Chloride Analyzer to be used quantitatively, but the high
bias must be considered. The overall performance was characterized as consistently biased but precise.
The results of the demonstration show that the Dexsil L2000 PCB/Chloride Analyzer can provide useful, cost-effective
data for environmental problem-solving and decision-making. Undoubtedly, it will be employed in a variety of
applications, ranging from serving as a complement to data generated in a fixed analytical laboratory to generating data
that will stand alone in the decision-making process. As with any technology selection, the user must determine if this
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-12 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's) Environmental Management (EM) program has entered into active
partnership with EPA, by 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
-------
-------
Abstract
In July 1997, the U.S. Environmental Protection Agency (EPA) conducted a demonstration of poly chlorinated
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 Dexsil's L2000 PCB/Chloride Analyzer. Separate
ETVRs have been published for the other technologies demonstrated.
The L2000 PCB/Chloride Analyzer is a field-portable instrument, weighing approximately 3.5 Ib, designed to
quantify PCB concentration in soils, dielectric fluids, and surface wipes. The L2000 utilizes a chloride-specific
electrode to determine the amount of chlorine in a sample after the sample has been digested to convert the
bound chlorine into ionic chloride. The L2000 detects the total chloride content of the sample and then
electronically converts total chloride content to PCB concentration in units of parts per million (ppm). The
L2000 provides no information on Aroclor identification.
The L2000's quantitative results were based on site-specific calibrations that were temperature-dependent.
Recalibration was required approximately every 15 min, or whenever there was an internal temperature change.
The method detection limit (MDL) is often defined as the minimum concentration of a substance that can be
ix
-------
measured and reported with 99% confidence that the analyte concentration is greater than zero. A field-based
MDL was calculated from a linear line fit to all of the PE data. The calculated field-based MDL (2.0 ppm) was
comparable to Dexsil's specified detection limit (2 ppm) when the instrument bias was mathematically
corrected (MDL without bias correction was 7.1 ppm). In general, the L2000's results were biased high (208%
recovery for soils and 149% for extracts). The overall precision, based on the relative standard deviation, for
soil samples (23%) was comparable to that for the reference laboratory (21%) . The precision for extract
samples was also comparable to that of the reference laboratory (both 14%). Comparability, based on
coefficients of determination (R2), was 0.854 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 L2000 results were
compared directly with the reference laboratory results.
The demonstration found that the L2000 was simple to operate in the field, requiring less than 2 h for initial
set-up and preparation for sample analysis. Once operational, the sample throughput of the L2000 during the
demonstration was 5 to 10 samples/h. Two operators analyzed samples during the demonstration, but the
technology can be run by a single trained person. Minimal training (<1 h) is required to operate the L2000,
provided the user has a fundamental understanding of basic chemical techniques. Because the PCB data
generated in this demonstration strongly correlated (R2 = 0.95) with the reference laboratory results, it may be
possible for Dexsil's L2000 PCB/Chloride Analyzer to be used quantitatively, but the high bias must be
considered. No "site effects" were observed in the data generated by the L2000 based on the change in
environmental conditions. The overall performance was characterized as consistently biased but precise.
-------
Table of Contents
Notice ii
Verification Statement iii
Foreword vii
Abstract ix
List of Figures xv
List of Tables xvii
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
General Technology Description 5
Soil Sample Preparation 5
Instrument Calibration 5
Sample Analysis 6
Surface Wipe Sampling and Analysis 6
Section 3 Site Description and Demonstration Design 7
Objective 7
Demonstration Site Description 7
Site Name and Location 7
Site History 7
Site Characteristics 8
xi
-------
Experimental Design 8
Environmental Conditions during Demonstration 11
Sample Descriptions 11
Performance Evaluation Materials 11
Environmental Soil Samples 12
Extract Samples 12
Sampling Plan 12
Sample Collection 12
Sample Preparation, Labeling, and Distribution 12
Predemonstration Study 14
Predemonstration Sample Preparation 14
Predemonstration Results 15
Deviations from the Demonstration Plan 15
Section 4 Reference Laboratory Analytical Results and Evaluation 17
Objective and Approach 17
Reference Laboratory Selection 17
Reference Laboratory Method 18
Calibration 18
Sample Quantification 18
Sample Receipt, Handling, and Holding Times 19
Quality Control Results 19
Objective 19
Continuing Calibration Verification Standard Results 19
Instrument and Method Blank Results 20
Surrogate Spike Results 20
Laboratory Control Sample Results 20
Matrix Spike Results 21
Conclusions of the Quality Control Results 21
Data Review and Validation 21
Objective 21
Corrected Results 22
Suspect Results 22
Data Assessment 23
Objective 23
Precision 23
Performance Evaluation Samples 23
Environmental Soil Samples 24
Extract Samples 26
Accuracy 26
Performance Evaluation Soil Samples 27
Extract Samples 28
Representativeness 28
Xll
-------
Completeness 28
Comparability 29
Summary of Observations 29
Section 5 Technology Performance and Evaluation 31
Objective and Approach 31
Data Assessment 31
Precision 31
Performance Evaluation Samples 31
Environmental Soil Samples 32
Extract Samples 33
Precision Summary 34
Accuracy 34
Performance Evaluation Soil Samples 35
Extract Samples 36
Accuracy Summary 37
False Positive/False Negative Results 37
Representativeness 38
Completeness 38
Comparability 38
Summary of PARCC Observations 40
Regulatory Decision-Making Applicability 42
Additional Performance Factors 42
Detection Limits 42
Sample Throughput 43
Cost Assessment 43
L2000 PCB/Chloride Analyzer Costs 44
Reference Laboratory Costs 45
Cost Assessment Summary 46
General Observations 46
Performance Summary 47
Section 6 Technology Update and Representative Applications 49
Objective 49
Technology Update 49
Representative Applications 49
Data Quality Objective Example 50
Section 7 References 51
Appendix A Description of Environmental Soil Samples 53
Appendix B Characterization of Environmental Soil Samples 57
xin
-------
Appendix C Temperature and Relative Humidity Conditions 61
Appendix D PCB Technology Demonstration Sample Data 67
Appendix E Data Quality Objective Example 77
Disclaimer 79
Background and Problem Statement 79
DQO Goals 79
Use of Technology Performance Information to Implement the Decision Rule 80
L2000 PCB/Chloride Analyzer Accuracy 80
Determining the Number of Samples 82
Determining the Action Level 83
Alternative FP Parameter 84
xiv
-------
List of Figures
3-1 Schematic map of ORNL, indicating the demonstration area 9
5-1 Paired PCB measurements for the L2000 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 39
5-2 Range of percent difference values for the comparison of the L2000 soil sample results with the reference
laboratory results 40
C-l Summary of temperature conditions for outdoor site 63
C-2 Summary of relative humidity conditions for the outdoor site 64
C-3 Summary of temperature conditions for chamber site 64
C-4 Summary of relative humidity conditions for chamber site 65
E-l A linear model for predicting L2000 PCB concentrations on certified PE concentrations with
95% confidence intervals 81
E-2 L2000 standard deviations versus certified PE concentration values 82
E-3 Decision performance curve for PCB drum example 84
xv
-------
-------
List of Tables
3-1 Summary of experimental design by sample type 10
3-2 Summary of the L2000's predemonstration results 15
4-1 Suspect measurements within the reference laboratory data 22
4-2 Precision of the reference laboratory for PE soil samples 24
4-3 Precision of the reference laboratory for environmental soil samples 25
4-4 Precision of the reference laboratory for extract samples 26
4-5 Accuracy of the reference laboratory for PE soil samples 27
4-6 Accuracy of the reference laboratory for extract samples 28
4-7 Summary of the reference laboratory performance 30
5-1 Precision of the L2000 PCB/Chloride Analyzer for PE soil samples 32
5-2 Precision of the L2000 PCB/Chloride Analyzer for environmental soil samples 33
5-3 Precision of the L2000 PCB/Chloride Analyzer for extract samples 34
5-4 Overall precision of the L2000 PCB/Chloride Analyzer for all sample types 35
5-5 Accuracy of the L2000 PCB/Chloride Analyzer for PE soil samples 36
5-6 Accuracy of the L2000 PCB/Chloride Analyzer for extract samples 37
5-7 Overall accuracy of the L2000 PCB/Chloride Analyzer for all sample types 37
5-8 Comparison of the reference laboratory's suspect data to the L2000 PCB/Chloride
Analyzer data 41
5-9 Summary of PARCC observations for the L2000 PCB/Chloride Analyzer 41
5-10 Performance of the L2000 PCB/Chloride Analyzer for soil samples between 40 and 60 ppm .... 42
5-11 Estimated analytical costs for PCB soil samples 44
5-12 Performance summary for the L2000 PCB/Chloride Analyzer 48
A-l Summary of soil sample descriptions 55
B-l Summary of environmental soil characterization 59
C-l Average temperature and relative humidity conditions during testing periods 63
D-l L2000 PCB technology demonstration soil sample data 69
D-2 L2000 technology demonstration extract sample data 75
D-3 Corrected reference laboratory data 76
xvn
-------
-------
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 (ORNL)
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 (DOE)
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
LMER Lockheed Martin Energy Research
xix
-------
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 (EPA)
NRC Nuclear Regulatory Commission
ORD Office of Research and Development (EPA)
ORNL Oak Ridge National Laboratory
ORO Oak Ridge Operations (DOE)
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 relative standard deviation (percent)
RT regulatory threshold
S2 variance for the measurement
SARA Superfund Amendments and Reauthorization Act of 1986
SD standard deviation
SDI Strategic Diagnostics Inc.
SITE Superfund Innovative Technology Evaluation
SMO sample management office
SOP standard operating procedure
xx
-------
SSM synthetic soil matrix
TCMX tetrachloro-m-xylene
TSCA Toxic Substance Control Act
Zj_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, we thank 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, Colo.); 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 Stephen Billets, Gary Robertson, and Eric Koglin (EPA's National
Exposure Research Laboratory, Las Vegas, Nevada). The authors also acknowledge the participation of Dexsil
Corporation, in particular, Wendy Schutt-Young and Chris Tellone, who performed the analyses during the
demonstration.
For more information on the PCB Field Analytical Technology Demonstration, contact
Eric Koglin
Project Technical Leader
Environmental Protection Agency
Environmental Sciences Division
National Exposure Research Laboratory
P.O. Box 93478
Las Vegas, Nevada 89193-3478
(702) 798-2432
For more information on Dexsil's L2000 PCB/Chloride Analyzer, contact
Ted Lynn, Ph.D.
Director of Research
Dexsil Corporation
One Hamden Park Drive
Hamden, Connecticut 06517
(203)288-3509
xxni
-------
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
technologies for PCBs: the L2000 PCB/Chloride Analyzer (Dexsil Corporation), the PCB Immunoassay Kit
(Hach Company), the 4100 Vapor Detector (Electronic Sensor Technology), and three immunoassay kits from
1
-------
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, Dexsil's L2000 PCB/Chloride Analyzer. 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://CLU-
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.
General Technology Description
The L2000 PCB/Chloride Analyzer (dimensions: 8" x 8" x 4.5") is a field-portable instrument, weighing
approximately 3.5 Ib, designed to quantify PCB concentration in soils, dielectric fluids, and surface wipes. The
L2000 currently requires 120V AC power, but the next version of the instrument will be battery-operated. The
instrument can quantify PCBs in soil over a range of 2 ppm to 2000 ppm and has the ability to extend the range
to over 2000 ppm with the reduction of the sample size. For wipe samples, PCBs can be quantified over a range
of 2 to 2000 (jg/100 cm2. The percent error rate is specified as 5% chlorine. The total time for analysis of soil
is 10 min; for dielectric fluid, 5 min; and for surface wipes, 12 min.
Soil Sample Preparation
Sample preparation consists of extraction and dehalogenation of the PCB. A 10-g sample of soil is weighed
into a polyethylene test tube. The soil is extracted with a nonchlorinated solvent from a premeasured ampule.
(A newly developed hydrocarbon solvent system was used for the demonstration analyses.) The soil is allowed
to settle, and the supernatant is decanted onto a Florisil cartridge. The soil extract is passed through the
cartridge, where all of the water and inorganic chloride is removed. Five milliliters of the eluent is collected in
a polyethylene reaction tube. Two glass ampules contained in the reaction tube are broken, introducing metallic
sodium to the extract solution. The sodium strips the covalently bonded chlorine atoms off the PCB molecule.
The mixture is then shaken for 10s and allowed to react for a total of 1 min. An aqueous extraction solution
is added to the reaction tube to adjust the pH, destroy the excess sodium, and extract and isolate the newly
formed chloride ions in an aqueous buffered solution. The aqueous layer is decanted, filtered, and collected in
an analysis vial. The ion-specific electrode is put into this aqueous solution to measure the millivolt potential.
The potential is then converted to PCB concentration in terms of parts per million.
Instrument Calibration
A one-point calibration is analyzed prior to sample analysis. The analyst simply selects calibration mode and
inserts the electrode into a 50 ppm chloride solution supplied with the reagents. A start button is pushed and
a "wait light" illuminates for approximately 30 s. When a "read light" illuminates, the analyst calibrates the
instrument by turning the calibration knob until the display reads 50 ppm. The instrument is now calibrated.
Additional calibration is required when the recalibrate light illuminates. This occurs approximately every 20
min, or after the completion of 15 to 20 samples.
-------
Sample Analysis
The analyst chooses between four different PCB settings—1242, 1260, Askarel A (60% Aroclor 1260/40%
trichlorobenzene), and total chloride—depending on the percentage of chlorine in the PCBs expected in the
samples as indicated by site information or history. If the Aroclor is not known or if there is a mixture of
Aroclors, the 1242 setting should be employed for the most conservative results. Total chloride setting is used
to quantify "odd" Aroclors (1221, 1248, etc.) or other chlorinated organics. To analyze the sample, the analyst
places the electrode into the aqueous extract solution and pushes the start button. After approximately 30 s,
the PCB concentration of the sample (in ppm) is displayed on the L2000 when the "read light" illuminates.
Note that total chlorine results must be divided by the percent chlorine of the analyte and multiplied by 100 to
calculate the PCB concentration of the sample.
Surface Wipe Sampling and Analysis
For this demonstration, simulated extract samples were provided for analysis. A sample collection and
preparation kit for surface wipes is available from Dexsil for use with the L2000 PCB/Chloride Analyzer. The
kit contains the following items for surface wipe sampling:
• chromatographic-grade hexane in 2-mL sealed glass ampules,
• disposable PCB-rated gloves,
• disposable forceps,
goggles,
• gauze pads, and
• reagents and vials.
To collect a wipe sample, a 1000-cm2 area is wiped with a gauze pad saturated with 2 mL of chromatographic-
grade hexane. (A 100-cm2 area must be wiped to obtain a low-level PCB concentration in (jg/100cm2.) After
waiting approximately 30 seconds for the hexane to evaporate, the analyst extracts the gauze pad with 10 mL
of isooctane. (Note that this is where the analysis of the demonstration extract samples began.) Five milliliters
of the extract is introduced into the sodium reaction tube. Once the extract is in the reaction tube, the procedure
is exactly the same as for the soil analysis. The PCB concentration is reported in terms of (jg/100cm2. The
concentration range of the instrument is 2 to 2000 (jg/100 cm2.
-------
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 descriptions
of the environmental soil samples 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
-------
Building
Figure 3-1. Schematic map of ORNL, indicating the demonstration area.
environmental soil) were analyzed. The ability to use environmental soils in the verification 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 (jg/mL, were analyzed in each
-------
Table 3-1. Summary of experimental design by sample type
Concentration
Range
Sample ID "
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
Extracts
129Vl32c
127/130
128/131
116
201,202,206
203,207,212,213
204,208,209,214,215
205,210,211,216,217
229/232
227/230
228/231
116
36
32
28
40
8
8
8
232"
" 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.
10
-------
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.
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 Dexsil team worked outdoors July 22, 23, 24, 25, and 26, 1997, and in the chamber
on July 26, 28, and 29, 1997.
Sample Descriptions
PCBs (C^HjQ.xClx) 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 semivolatile 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 6-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.
11
-------
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. Dexsil analyzed extracts prepared
in iso-octane. 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 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
12
-------
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 pg/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., Dexsil 1001 through Dexsil 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
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. Dexsil received
information pertaining to which Aroclors were in the samples. This was provided at the request of Dexsil to
simulate the type of information that would be available during actual field testing. The developers returned
13
-------
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 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.
14
-------
Predemonstration results were received by June 26, 1997. Table 3-2 summarizes the L2000's results for the
predemonstration samples. Results indicated that Dexsil's L2000 PCB/Chloride Analyzer was ready for field
evaluation. The same analyst, instrument, and procedure were used in the field demonstration study.
Table 3-2. Summary of the L2000's predemonstration results
Sample Description
2 ppm of Aroclor 1260
100ppm(total)of
Aroclors 1254 and 1260
1 1 ppm of Aroclor 1260
50 ppm of Aroclor 1254
5 ppm of Aroclor 1242
Matrix
Soil
Soil
Soil
Soil
Extract
Source
ORNL
ORNL
EPA
ERA
ORNL
L2000"
Result
(ppm)
2.4
82.1
8.8
45.3
6.0
Duplicate result
(ppm)
2.8
88.4
7.7
b
6.4
Reference Laboratory
Result
(ppm)
2.2
78.0
11.0
37.0
4.7
Duplicate result
(ppm)
2.3
89.0
9.5
b
4.9
" No sample information was provided to Dexsil. All samples quantified as Aroclor 1260.
6 Replicate was not analyzed because of lack of adequate sample for second analyses.
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 Dexsil team made two modifications to the procedure described in the
technology demonstration plan [5]. On the first day of testing, the Dexsil field team realized the need for an
additional pre-filter, prior to the Florisil column, to remove particulate matter. The Dexsil team attributed the
filtering problems to fine particulate matter in the sample matrix that was somewhat different from the
predemonstration study sample matrix. The second deviation from the demonstration plan was that Dexsil used
iso-octane as the solvent for the extract samples and not hexane, as was listed.
15
-------
16
-------
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.
17
-------
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
((ig/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").
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
18
-------
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 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.
19
-------
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 ,nnn,
percent recovery = x 100% (4-1)
actual amount
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.
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.
20
-------
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:
100% (4_2)
nnr, I MS recovery - MSD recovery
KrL) —
average recovery
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.
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
21
-------
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
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
22
-------
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:
„„„ Standard Deviation ,nnn,
RSD = 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
23
-------
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
3i
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%).
samples, was 21% for the worst case (including the suspect result) and 18% for the best case (excluding the
suspect result).
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.
24
-------
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).
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 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.
25
-------
Extract Samples
The extract samples, which were used to simulate surface wipe samples, were the simplest of all the
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 (jg/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 lO-pg/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
1 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.
26
-------
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 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:
c Results excluding the suspect value (results including the suspect value:
average = 67.8 ppm and recovery = 136%).
average = 52.8 ppm and recovery = 106%).
Extract Samples
27
-------
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 lO-pg/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; 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 (ig/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)
n
in 4
Recovery
104
64
' 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.
28
-------
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 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.
29
-------
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.
30
-------
Section 5
Technology Performance and Evaluation
Objective and Approach
This section presents the evaluation of data generated by Dexsil's L2000 PCB/Chloride Analyzer. 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 Dexsil's L2000
PCB/Chloride Analyzer through a statistical analysis of the data. PARCC parameters were used to evaluate
the L2000'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 measurements made
at each site indicated that there were no significant differences between the two data sets. All statistical tests
were performed at the 5% significance level. Because environmental conditions did not appear to affect the
results significantly, data from both sites were also combined for each parameter (precision and accuracy) to
determine overall performance. Appendix D contains the raw data that were used to assess the performance
of the L2000 PCB/Chloride Analyzer.
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 Equation 4-3).
For comparative information on 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 L2000 PCB/Chloride Analyzer for the analysis of PE samples.
RSDs ranged from 7 to 35% under the outdoor conditions, and from 5 to 46% under inside chamber conditions.
In Table 5-1, the data generated under both environmental conditions are combined to
31
-------
Table 5-1. Precision of the L2000 PCB/Chloride Analyzer for PE soil samples
Outdoor Site
Sample
ID
126 a
118
124
120
122
119
125
121
123
Average
Concentratio
n (ppm)
2.1
6.5
6.2
12.2
18.1
47.8
72.3
66.2
78.7
SD
(ppm)
1.0
1.0
1.9
4.3
4.5
12.9
5.2
5.9
6.1
RSD
(%)
48
15
31
35
25
27
7
9
8
Chamber Site
Sample
ID
226 a
218
224
220
222
219
225
221
223
Average
Concentratio
n (ppm)
1.8
5.0
8.3
7.8
15.1
41.6
73.3
73.4
75.4
SD
(ppm)
1.3
2.3
3.6
1.8
6.0
6.2
3.9
5.6
7.8
RSD
(%)
72
46
43
23
40
15
5
8
10
Combined Sites
Average
Concentratio
n (ppm)
1.9
5.8
7.2
10.0
16.6
44.7
72.9
69.8
77.0
SD
(ppm)
1.1
1.8
2.9
3.9
5.2
9.9
4.3
6.6
6.7
RSD
(%)
58
31
40
39
31
22
6
9
9
a The L2000 detected PCBs in the blanks. The method detection limit (specified by Dexsil) was 2 ppm. The blank data
were not included in the calculation of the overall average RSD.
provide an overall assessment of precision. The performance for the combined site data indicated RSDs ranging
from 6 to 40%.
Environmental Soil Samples
The precision of the L2000 PCB/Chloride Analyzer for the analysis of environmental soil samples is reported
in Table 5-2. RSDs ranged from 3 to 54% under the outdoor conditions, and from 2 to 51% under inside
chamber conditions. For concentrations above 15 ppm, all RSDs were below 35%. Because the majority 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 26%, in contrast to that of the high concentration range, which was
4%.
The Paducah soils (indicated by bold sample IDs in Table 5-2) were analyzed at both sites to provide an
assessment of the L2000'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. For the replicate Paducah soil sample sets, represented in bold at the bottom of the table, the 100
series were samples analyzed under the outdoor conditions and the 200 series were samples
32
-------
Table 5-2. Precision of the L2000 PCB/Chloride Analyzer for environmental soil samples
Outdoor Site
Sample
ID
108
109
102
107
103
101
104
111
105
106
110
112
113"
114
115
116
117
Average
Concentration
(ppm)
5.6
5.9
6.3
6.9
10.3
10.6
13
49.4
72.9
430.6
432.7
473.3
6.7
7.7
38.2
100.9
513.9
SD
(ppm)
2
2.5
3.2
2.4
3.8
5.7
3.1
11.9
10.9
26.1
108.1
134.5
1.4
3.1
4.5
3.3
35.3
RSD
(%)
35
42
50
34
36
54
24
24
15
6
25
28
21
40
12
3
7
Chamber Site
Sample
ID
206
207
208
209
210
211
212
213
214
215
216
217
201
202
203
204
205
Average
Concentration
(ppm)
9.5
31.7
64.2
80.5
105.0
437.1
14.1
15.4
50.6
52.6
86.7
445.7
8.7
5.5
31.3
108.7
477.1
SD
(ppm)
3.8
0.8
7.0
in
5.0
7.9
6.1
3.5
5.5
7.0
9.7
10.6
4.4
2.2
5.9
8.6
25.2
RSD
(%)
39
3
11
35
5
2
43
23
11
13
11
2
51
40
19
8
5
Combined
Sites
RSD
(%)
n/aa
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
42
41
17
7
7
' Combined site results were not applicable because these environmental samples were not replicate pairs.
' Bold sample IDs were matching Paducah sample pairs (i.e., 113/201, 114/202, 115/203, 116/204, 117/205).
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 this analysis showed that there were no
significant differences in the data generated at each site. This indicated that different environmental conditions
had no effect on the L2000's performance for this sample set, as illustrated in the following data. Under the
outdoor conditions, the RSDs for the Paducah samples ranged from 3 to 40%, and under chamber conditions,
from 5 to 51%. RSDs for the combined site data (eight replicates per paired Paducah sample IDs) ranged from
7 to 41%.
Extract Samples
Table 5-3 summarizes the L2000 PCB/Chloride Analyzer results for the extract samples which were used to
simulate surface wipe samples. Refer to Section 3 under the heading "Extract Samples" for further clarification
of this sample type. RSDs ranged from 3 to 11% under the outdoor conditions, and from 6 to 19% inside the
chamber. In terms of concentration levels, the average RSD at the 10-(jg/mL level was 9%, while the average
RSD at the lOO-^ig/mL level was 18%.
33
-------
Table 5-3. Precision of the L2000 PCB/Chloride Analyzer for extract samples
Outdoor Site
Sample
ID
129 a
127
128
Average
Concentration
(ug/mL )
2.9
21.5
81.4
SD
(ug/mL)
0.3
2.4
2.5
RSD
(%)
10
11
3
Chamber Site
Sample
ID
229 a
227
228
Average
Concentration
(ug/mL)
1.8
20.0
100.0
SD
(ug/mL)
0.3
1.2
19.3
RSD
(%)
18
6
19
Combined Sites
Average
Concentration
(ug/mL)
2.3
20.7
90.7
SD
(ug/mL)
0.70
1.9
16.2
RSD
(%)
30
9
18
a The L2000 detected PCBs in the blanks. The method detection limit (specified by Dexsil) was 2 ppm. 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 overall precision of the L2000
PCB/Chloride Analyzer for the PE samples was a mean RSD of 22% and a median RSD of 19%; the 95th
percentile of all individual RSDs was 45%. The environmental soil sample RSD results were a mean of 23%,
a median of'22%, and a 95th percentile of 50%. The overall precision for all extract samples was a mean RSD
of 14%. The 95th percentile and median data are not presented for extract samples because of the limited
number of data points. Additionally, the precision of the solvent blanks from the extract data was comparable
to the precision of the soil blanks from the environmental soil sample data (SD of 0.7 vs 1.1 ppm, respectively).
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 Equation 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 "Accuracy."
34
-------
Table 5-4. Overall precision of the L2000 PCB/Chloride Analyzer for all sample types
Statistic
Mean
Median
95th percentile
% RSD, PE Samples
Outdoor
20
20
34
Chamber
24
19
46
Combine
d
22
19
45
% RSD, Environmental Soil
Samples
Outdoor
27
25
51
Chamber
19
11
45
Combined a
23
22
50
% RSD, Extract Samples
Outdoor
7
n/ab
n/a
Chamber
13
n/a
n/a
Combine
d
14
n/a
n/a
a Combined data were generated only for the Paducah soil samples.
b Median and 95th percentile statistics were not applicable to extract samples.
Performance Evaluation Soil Samples
The performance of the L2000 PCB/Chloride Analyzer 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. All
average concentrations determined by the L2000 were outside of the acceptance ranges and all were biased
high. This was also reflected in the average percent recoveries, which were all greater than 100%. Average
percent recoveries ranged from 132 to 325% under the outdoor conditions. Under chamber conditions, average
percent recoveries ranged from 139 to 415%. When the data for the two sites were combined, the average
percent recovery ranged from 132 to 360%. Additionally, some of the PE samples (sample IDs 118/218,
119/219, 120/220, 121/221, 122/222, 123/223) were spiked with pesticides (up to 15 ppm). The results for
the pesticide-containing PE samples did not appear to be biased significantly higher than the nonpesticide PE
samples (sample IDs 124/224 and 125/225). Pesticides are potential interferences for the L2000 because they
contain organic chlorine. The results for the pesticide-containing PE samples did not appear to be biased
significantly higher than the nonpesticide PE samples (sample IDs 124/224 and 125/225). It was speculated
that the aliphatic pesticides were probably removed during sample preparation, and the aromatic pesticides
were a minor contribution to the total chlorine concentration.
While the results were biased high, the L2000 data did correlate with the certified PE values. Because a
mathematical relationship between the certified values and the biased results can be defined, it is possible to
mathematically correct for the bias evident in this study. Although the correlation between the certified values
and the L2000's results in this study was most accurately described by a quadratic equation (which draws a
curved line through the data points), a simpler linear equation (which draws a straight line through the data
points) can be used with minimal loss of predictive capability. A detailed description of how the L2000 data
could be converted to compensate for the bias is presented in Appendix E.
Several factors may have contributed to the high bias of the L2000 results. First, PCBs were detected at
concentrations that ranged from 2 to 4 ppm in four blank soil samples, and the results were not blank-corrected.
Second, a multiplier of 1.25 was built into the circuitry of the L2000 analyzer to account for the extraction
efficiencies typical of complex environmental soil samples. This multiplier ensured that a conservative estimate
was reported. Third, as discussed in Section 3, Dexsil was provided information
35
-------
Table 5-5. Accuracy of the L2000 PCB/Chloride Analyzer 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)
2.1
6.5
6.2
12.2
18.1
47.8
72.6
66.2
78.7
Average
Recovery
(%)
n/a
325
310
244
166
239
145
132
157
Chamber Site
Sample
ID
226
218
224
220
222
219
225
221
223
Average
Cone
(ppm)
1.8
5.0
8.3
7.8
15.1
41.6
73.3
73.4
75.4
Average
Recovery
(%)
n/a
250
415
156
139
208
147
147
151
Combined Sites
Average
Cone
(ppm)
1.9
5.8
7.2
10.0
16.6
44.7
72.9
69.8
77.0
Average
Recovery
(%)
n/a
290
360
200
152
224
146
140
154
a The L2000 detected PCBs in the blanks. The method detection limit (specified by Dexsil) was 2 ppm. Average recovery
calculations were not applicable to blank samples.
prior to the demonstration pertaining to which Aroclors were thought to be present in the samples. Dexsil used
this sample information to convert the total chloride concentration detected by the L2000 to a PCB
concentration. In some cases, the Aroclor suspected to be present did not match the Aroclor reported by the
reference laboratory. It is difficult to quantify the impact that this discrepancy had on the L2000's bias, but
it is thought to be minor.
Extract Samples
Percent recovery results for the extract samples are summarized in Table 5-6 for the L2000 PCB/ Chloride
Analyzer. Under the outdoor conditions, the average percent recoveries for extract samples ranged from 81 to
215% and under indoor conditions from 100 to 200%. In terms of concentration levels (i.e., for the combined
site data), the average recovery at the 10-(jg/mL level was 207%, compared to
36
-------
Table 5-6. Accuracy of the L2000 PCB/Chloride Analyzer for extract samples
Slnilrp
Concentration
(ug/mL)
Oa
10
100
Outdoor Site
Sample
ID
129
127
128
Average
Cone
(ug/mL)
2.9
21.5
81.4
Average
Recovery
(%)
n/a
215
81
Chamber Site
Sample
ID
229
227
228
Average
Cone
(ug/mL)
1.8
20.0
100.0
Average
Recovery
(%)
n/a
200
100
Combined Sites
Average
Cone
(ug/mL)
2.3
20.7
90.7
Average
Recovery
(%)
n/a
207
91
a The L2000 detected PCBs in the blanks. The method detection limit (specified by Dexsil) was 2 ppm. Average recovery
calculations were not applicable to blank samples.
91% at the 100-(jg/mL level. Blank concentrations of 2.0 to 3.2 (jg/mL may have contributed more
significantly to the bias in the 10 (jg/mL concentration results.
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 L2000 PCB/Chloride Analyzer can be characterized as biased high, with a mean
percent recovery of 208%, a median of 168%, and a 95th percentile of 420%. The overall accuracy for all
extract samples was a mean percent recovery of 149%. The 95th percentile and median data were not presented
for extract samples because of the limited number of data points.
Table 5-7. Overall accuracy of the L2000 PCB/Chloride Analyzer for all sample types
Statistic
Mean
Median
95th percentile
% Recovery, PE Samples
Outdoor Chamber Combined
215
181
355
201
157
420
208
168
420
% Recovery, Extract Samples
Outdoor Chamber Combined
148
n/aa
n/a
150
n/a
n/a
149
n/a
n/a
Median and 95th percentile statistics were not applicable to extract samples because of the limited number of data points.
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 (fh) result [10] is one in which the technology indicates that there are no PCBs
present in the sample, when PCBs actually are present. Both fp and fh results are influenced by the method
detection limit of the technology. Of the eight blank soil samples analyzed, four were reported as having
detectable levels of PCBs (i.e., fp = 50%). Of the 192 non-blank soil samples analyzed, none were reported
37
-------
as non-detects. Therefore, the percentage of fh results was 0%. The results were the same for the extract
samples (fp = 50% and fh =
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 L2000 PCB/Chloride Analyzer was capable of analyzing diverse sample types (PE, simulated
surface wipe extract, and actual field environmental soil samples) under multiple environmental conditions.
When this technology is used, quality control samples should be analyzed to assess the performance of the
L2000 under the testing conditions.
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 by the technology for all 232 samples. Therefore, the completeness
of the L2000 PCB/Chloride Analyzer was 100%.
Comparability
Comparability refers to the confidence with which one data set can be compared to 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 L2000 measured values vs the reference laboratory results. Additional
statistical analysis of the PCB soil concentrations for paired samples showed no significant difference between
the L2000 and reference laboratory results for the higher concentrations (i.e., >125 ppm). In contrast, for the
lower concentrations (i.e., <125 ppm), the L2000 reported PCB concentrations that were significantly higher
than those reported by the reference laboratory. This is illustrated in Figure 5-1, which is a plot of the L2000-
measured PCB soil concentrations vs the corresponding reference laboratory-measured concentrations
(excluding the suspect values listed in Table 4-1). Figure 5-l(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. The diagonal
lines in Figure 5-1 represent the line of theoretically perfect correlation (R2 = 1.0) between the reference
laboratory data set (plotted along the x-axis) and the L2000 data set (plotted along the j-axis). A value above
the diagonal line indicated that the L2000'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 L2000 PCB concentrations vs the reference
laboratory PCB concentrations. Excluding the reference laboratory's suspect measurements, the coefficient of
determination (R2) was 0.854 when all soil samples (0 to 700 ppm) were considered. As shown in Figure 5-
l(b), the majority of the soil samples were in the concentration range of 0 to 125 ppm. The R2 value for this
concentration range was 0.781.
A direct comparison between the L2000 and reference laboratory data was performed by evaluating the percent
difference (%D) between the measured concentrations, defined as
0/ „ [L2000] - [RefLab] inno/
% D = — v-^- — x 100% (5_n
[RefLab] ^ '
38
-------
900
0 100 200 300 400 500 600 700 800 900
Reference PCB Concentration (ppm)
(a)
(b)
25 50 75 100 125
Reference PCB Concentration ( ppm )
Figure 5-1. Paired PCB measurements for the L2000 and reference laboratory for (a) all soil samples and (b) soil
samples where the reference laboratory results were less than or equal to 125 ppm. Lines denote perfect correlation.
Figure 5-2 provides a summary of the range of percent difference values for the soil samples, as calculated
using Equation 5-1. The graph represents the percentage of samples that fall within each range of percent
difference values, but 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 these samples. As shown in Figure 5-2, the majority of the percent difference
values were greater than 100%. Fewer than 10% of the samples were biased low (%D < -1%) relative to the
reference laboratory. Approximately 14% of the soil sample results had percent difference values within the
range of ±25%.
Comparability was also assessed for the extract samples. Analysis of the reference laboratory extract results
vs the L2000 results indicated a high bias for the L2000's extract PCB concentrations. The coefficient of
determination (R2) for a line fitted to this data was 0.954. The correlation was near perfect
(i.e., R2 = 1.0) primarily because of the precision at the two concentration levels (10 and 100 ppm) that were
analyzed. The percent difference values for the extract samples were also assessed. Percent differences were
evenly distributed around the 76%D range (e.g., 8 of 16 samples had percent differences >76%, while 8 of 16
samples were between 0 and 75%D). None of the extract samples analyzed by the L2000 had a negative
percent difference relative to the reference laboratory.
39
-------
-100% to -75% to - -50% to-
-76% 51% 26%
-25% to
0%
1 % to
24%
25% to
49%
50% to
74%
75% to
100%
> 101%
Range of percent difference values
Figure 5-2. Range of percent difference values for the comparisons of the L2000 soil sample results
with the reference laboratory results.
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 L2000's matching results. For sample IDs 110 and 112, the reference laboratory
obtained qualitative results only, while Dexsil reported quantitative PCB concentrations for the four replicates
that were precise. 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 L2000 generated
results for all four replicates that were consistent and that were comparable with the reference laboratory's
replicate means that excluded the suspect value. These comparisons demonstrate the L2000's ability to analyze
successfully samples that were troublesome for the reference laboratory.
Summary of PARCC Observations
Table 5-9 provides a summary of the performance of Dexsil's L2000 PCB/Chloride Analyzer for the analysis
of all sample types used in this demonstration. The reference laboratory's performance (excluding suspect data)
is also presented in this table for comparison. In terms of precision, the overall average RSD for the L2000,
weighted for the number of samples, was 23% for the soil samples. This is comparable to the reference
laboratory's overall RSD of 21%. For the extract samples, the overall average RSD was the same (at 14%)
for both the L2000 and the reference laboratory.
In terms of accuracy, the L2000's soil measurements were generally biased high by approximately a factor of
2 for PE PCB concentrations less than the regulatory limit of 50 ppm [2]. In comparison, the reference
laboratory reported unbiased PCB concentrations for these PE soil samples. Extract
40
-------
Table 5-8. Comparison of the reference laboratory's suspect data to the L2000 PCB/Chloride Analyzer data
Sample ID
110
112
106
205
216
217
225
Reference Laboratory Method
Suspect Measurement
(ppm)
125 ppm c
Sample ID 110
Sample ID 112
Overall
10 ppm
100 ppm
Overall
L2000
Number of
Samples
8
8
64
108
20
4
4
200
8
8
16
Precision (Average % RSD)
L2000
58
30
22
26
4
25
28
23
9
18
14
Reference
Laboratory
n/a
18a
23 a
19a
Not quantified
Not quantified
21 a
19
8
14
Accuracy (Average % Recovery)
L2000
PCB levels in
blanks were
2-A ppm
208
208
207
91
149
Reference
Laboratory
All reported as
non-detects
101 a
101 a
104
64
84
"Average result excluding the suspect measurements.
Samples for which the reference laboratory values were < 125 ppm.
c Samples for which the reference laboratory values were > 125 ppm.
41
-------
measurements by the L2000 were also biased high at 10 ppm (207% recovery), but were unbiased at 100 ppm
(91% recovery). In contrast, the reference laboratory results were unbiased at 10 ppm (104% recovery), but
were biased low at 100 ppm (64% recovery) for extract samples.
The L2000 detected PCBs in four blanks (i.e., 50% fp results), while the reference laboratory correctly reported
all blank samples as non-detects. No fh results were reported by the L2000. Thirteen suspect measurements
that the reference laboratory had difficulty in analyzing were reported precisely by the L2000. Overall, the
performance of the L2000 PCB/Chloride Analyzer for the PCB demonstration samples was characterized as
biased but precise.
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/100cm2 for surface wipes. The
L2000's performance for soil samples (both PE and environmental) ranging in concentration from 40 to 60
ppm can be used to assess this ability, and the data are provided in Table 5-10. The performance of the L2000
for this concentration range showed a slightly lower RSD and percent recovery compared to the entire PE and
environmental soil sample data set. In general, the variability of the measurements was low (the mean RSD was
12%), but the bias was still high by almost a factor of 2 (147% mean percent recovery). The mean percent
difference value was 83% when the L2000 results were compared to the corresponding reference laboratory
results for the 40- to 60-ppm concentration range.
Table 5-10. Performance of the L2000 PCB/Chloride Analyzer for soil samples between 40 and 60 ppm
Statistic
Mean
Median
95th percentile
Precision (% RSD)
12
8
29
Accuracy (% Recovery)
147
148
168
Comparability (% Difference)
83
53
188
The L2000 PCB/Chloride Analyzer's performance on extract samples is shown 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 (ig/100cm2. For the simulated wipe extract samples, the L2000 was precise
(14% average RSD), but had a high bias (149% average percent recovery) overall. For the L2000, a 1000 cm2
sample is collected, and the results are reported in (jg/100 cm2 according to a conversion algorithm built into
the instrument's software. Note that the manufacturer recommends the collection of 1000 cm2 samples to
ensure that detection limits are exceeded.
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]. The reported
42
-------
MDL for the L2000 was 2 ppm. An MDL, calculated from a linear line fitted to all PE data, was 7.1 ppm. As
previously discussed above (under the heading "Accuracy"), the use of a plot of the L2000 values for the PE
samples vs the certified PE values permits compensation for the bias in the L2000 results. By this method, the
calculated MDL (7.1 ppm) can be corrected to an unbiased result of 2.0 ppm, which then agrees with the
L2000's specified MDL of 2 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. Dexsil's sample throughput rate was consistently around 5
samples/hour under the outdoor conditions but improved to nearly 10 samples/hour under the chamber
conditions. This increased sample throughput may be attributed to (1) the analysis order (Dexsil may have
gained experience by analyzing samples under the outdoor conditions first); (2) difficulty with the sample
matrices that were analyzed only under the outdoor conditions (i.e., filtering problems, as noted in Section 3
under "Deviations from the Demonstration Plan"); or (3) the reduced need for recalibration when operating in
the controlled temperature environment of the chamber.
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 L2000 PCB/Chloride Analyzer and a conventional analytical reference
laboratory method. The analysis was based on the results and experience gained from this demonstration, costs
provided by Dexsil Corporation, and representative costs provided by the reference analytical laboratories who
offered to analyze these 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 L2000 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, and
• waste disposal costs.
Each of these cost factors is defined and discussed below and serves as the basis for the estimated cost ranges
presented in Table 5-11. This analysis assumed that the individuals performing the analyses were fully trained
to operate the technology. Dexsil does not offer a specific training course on the use of the L2000, but does
provide free assistance, on an as-needed basis, through its technical service department.
43
-------
Table 5-11. Estimated analytical costs for PCB soil samples
L2000 PCB/Chloride Analyzer
Dexsil Corporation
Sample throughput rate: 5 samples per hour (outdoors)
10 samples per hour (chamber)
Cost Category Cost (S)
Sample Shipment 0
Labor
Mobilization/demobilization 250^00
Travel 1 5-1 000 per analyst
Per diem 0-150 per day per analyst
Rate 30-75 per hour per
analyst
Equipment
Mobilization/demobilization 0-150
L2000 rental fee 500 per month
L2000 purchase price 3500
Reagents/supplies 8-10 per
sample
Waste Disposal 35^170
EPA SW-846 Method 8080/8081/8082
Reference Laboratory
Typical turnaround time: 14-30 days
Cost Category Cost (S)
Sample Shipment
Labor 100-200
Overnight shipping charges 50-150
Labor
Mobilization/demobilization Included a
Travel Included
Per diem Included
Rate 44-239 per sample
Equipment
Mobilization/demobilization Included
Rental/purchase of system Included
Reagents/supplies
Included
Waste Disposal Included
a "Included" indicates that the cost is included in the labor rate.
Sample acquisition and pre-analytical sample preparation, which were tasks common to both methods, are costs
that were not included here.
L2000 PCB/Chloride Analyzer Costs
• Sample shipment costs. Because the samples were analyzed on-site, no sample shipment charges were
associated with the cost of operating the L2000.
• Labor costs. 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 hours, at a rate of $50 per
hour.
— 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 ($1000) included the cost of airline travel and rental car fees.
44
-------
— Per diem: This cost element included food, lodging, and incidental expenses. Estimates ranged
from zero (for a local site) to $150/day per analyst.
— Rate: The cost of the on-site labor was estimated at a rate of $30 to $75 per hour, 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. Equipment costs included mobilization/demobilization, rental fees or purchase of
equipment, and the reagents and other consumable supplies necessary to complete the analysis.
— Equipment mobilization/demobilization: This included the cost of shipping the equipment to
the test site. If the site is local, the cost would be zero. For this demonstration, the cost of
shipping equipment and supplies was estimated at $150.
— Rental/purchase: The fee to rent the L2000 at the time of the demonstration study was $500
per month. At the time of the demonstration, the cost of purchasing the equipment was $3500.
The purchase price included the L2000 PCB/Chloride Analyzer, chloride specific electrode,
power cube, portable electronic balance, 5-mL pipettor, carrying case, vial rack, timer,
instructions, and 20 soil test reagents.
— 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 L2000 was $8 to $10 per sample. The price may be
reduced if large quantities are purchased.
Waste disposal costs. Waste disposal costs were estimated based on the 1997 regulations for disposal
of PCB-contaminated waste. Using the L2000, Dexsil generated approximately 10 Ib of liquid PCB
waste and approximately 20 Ib of solid PCB waste. The disposal costs for the PCB waste by
incineration at a commercial facility was estimated at $0.25/lb for liquids and $ 1.50/lb for solids. For
comparison, the cost for PCB waste disposal at ETTP was estimated at $11/lb for liquids and $18/lb
for solids.
Reference Laboratory Costs
• Sample shipment 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.
— Overnight shipping: The overnight express shipping service cost was estimated to be $50 for
one 50-lb cooler of samples.
— Labor: This cost element included 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 hours at $50 per hour.
45
-------
• Labor, equipment, and waste disposal costs. The labor bids from commercial analytical reference
laboratories who 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 workload 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 that included
labor, equipment, waste disposal, and report preparation.
Cost Assessment Summary
An overall cost estimate for the L2000 vs the reference laboratory was not made due to 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 vs the reference laboratory.
General Observations
The following are general observations regarding the field operation and performance of the L2000:
The system was light, easily transportable, and rugged. It took less than 2 h for the Dexsil
team to prepare to analyze samples on the first day of testing. While working at the outdoor
site, the Dexsil team completely disassembled their work station and brought everything inside
at the close of each day. It took the Dexsil team less than 1 h each morning to prepare for
sample analysis.
Two operators were used for the technology demonstration because of the number of samples
and working conditions, but the technology can be run by a single person.
• Operators generally require less than an hour of training and should have a basic knowledge
of chemistry and lab techniques.
Reagent handling was minimized by the use of premeasured, breakable glass ampules.
The system required 120-V AC power. A battery-operated L2000 system is expected to be
commercially available in 1998.
• The L2000 has a temperature setting adjustment. The Dexsil team monitored the temperature
during the technology demonstration and adjusted the setting accordingly. This procedure
increased sample throughput time during the outdoor site activities because adjustment of the
temperature setting required recalibration of the instrument.
The system required calibration approximately every 15 min. It also was recalibrated
whenever there was an internal temperature change. Typically, a calibration standard was
analyzed approximately every 12 samples. Calibration seemed to be more frequent when the
46
-------
Dexsil team was working outdoors, presumably because the outdoor temperature fluctuated
throughout the day.
• The Dexsil team usually analyzed three reagent blanks with each batch of 12 samples.
Reagent blanks were analyzed to establish a baseline correction for the system. On average,
the correction was less than 1 ppm. The blank results also provided Dexsil with information
concerning the daily temperature fluctuations.
The Dexsil analysts used the L2000 in the "total chloride" mode. Using the sample
information provided at their request [i.e., which Aroclor(s) was in each sample; see Section
3 for more details], Dexsil calculated the most conservative (highest) PCB concentration. In
cases where no Aroclor information is known, Dexsil advises that the sample should be
quantified as Aroclor 1242.
Approximately 20 Ib of solid waste (used and unused soil, gloves, paper towels, ampules, etc.)
were generated from the analysis of the technology demonstration samples. In addition,
approximately 10 Ib of liquid hazardous waste were generated.
Performance Summary
A summary of the performance characteristics of Dexsil's L2000 PCB/Chloride Analyzer, presented previously
in this chapter, is shown in Table 5-12. Because the PCB data generated by the technology could be correlated
with the reference laboratory results, it may be possible for the L2000 to be used quantitatively, but the high
bias must be considered. The overall performance of the L2000 PCB/Chloride Analyzer was characterized as
consistently biased but precise.
47
-------
Table 5-12. Performance summary for the L2000 PCB/Chloride Analyzer
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 4 of 8 blanks at 2 to 4 ppm
Extracts: PCBs detected in 4 of 8 blanks at 2 to 3 ppm
Dexsil specified: 2 ppm
Calculated: 7.1 ppm (corrected: 2.0 ppm)
Average RSD
PE soils: 22%
Environmental soils: 23%
Extracts: 14%
Average Percent Recovery
PE soils: 208%
Extracts: 149%
Blank Soils: 50% (4 of 8 samples)
Blank Extracts: 50% (4 of 8 samples)
PE and Environmental Soils: 0% (0 of 192 samples)
Spiked Extracts: 0% (0 of 16 samples)
PE and Environmental Soil Samples
Percent Difference: 52% of samples were > 100% D
Coefficients of determination (R2): 0.854 (all data),
0.781 (<125 ppm)
Extract Samples
Percent Difference: 50% of samples were > 76% D
Coefficient of determination (R2): 0.954
40 to 60 ppm PE and Environmental Soil Samples
Precision: 12% average RSD
Accuracy: 147% average recovery
Comparability: 83% average difference
100 ug/100cm2 and 1000 ug/100cm2 Extract Samples
Precision: 14% average RSD
Accuracy : 149% average recovery
Comparability: 70% average difference
5 samples/hour (outdoors)
10 samples/hour (chamber)
120 VAC
Basic knowledge of chemical techniques; <1 hour
technology-specific training
Equipment purchase: $3500
$5 to $16 per sample (matrix-dependent)
Approximately 10 Ib liquid waste
Approximately 20 Ib solid waste
48
-------
Section 6
Technology Update and Representative Applications
Objective
In this section, Dexsil describes new technology developments that have occurred since the demonstration
activities. In addition, the developer has provided a list of representative applications where the L2000 has been
or is currently being utilized.
Technology Update
A new version of the L2000 will be on the market in 1998. The L2000 DX Analyzer comes equipped with an
LCD (2-in. by 16-in.) backlit display. The display is easy to read in all lighting situations. Pertinent information
regarding the program in use, blank subtraction values, reporting units, and concentration values are visible
on the display.
L2000 DX is preprogrammed with conversion factors for all major Aroclors and most chlorinated pesticides
and solvents. The built-in methods include corrections for extraction efficiencies, dilution factors, and blank
contributions. Programs are easily selected from a menu to perform routine analyses of common chlorinated
organic compounds. For less common analytes or for custom measurement protocols, user-defined methods
can be easily built and stored using the method development menus.
Analysis results can either be stored electronically, using the parallel port or uploading to a PC via the RS-232
serial port, or can be printed directly to the on-board 40-column thermal printer. The analyzer itself utilizes
rechargeable batteries, which allow fully mobile operation in remote locations without access to power.
In addition, Dexsil has developed a new solvent system that provides extraction efficiency for PCBs from all
soils tested that is equal to or better than the traditional extraction solvent. This new solvent system, which was
the system used in the demonstration, is presently available for purchase. Over the next year, the traditional
solvent system will be phased out and replaced by the more universal extraction solvent.
Representative Applications
The L2000 PCB/Chloride Analyzer is designed to provide quick quantitative results for PCB concentrations
in soils, in dielectric fluids, and on surfaces. The analyzer operates on the principle of total organic chlorine
detection. Co-contamination of the matrices with hydrocarbons such as transformer oil does not affect the
quantification of the PCB. Water and salt content also has no effect on the test.
The insensitivity of the L2000 analyzer to common test interferences makes the technology ideal for a wide
range of environmental applications. Numerous field applications have benefitted from the use of the L2000.
Due to our policy to protect client confidentiality, Dexsil will not name specific projects. The following four
49
-------
companies regularly use the L2000 analyzer for a variety of applications and may assist a potential user in
verifying they would benefit from the use of this technology.
1. GZA GeoEnvironmental, Inc.
N4140 Duplainville Road
Pewaukee, WI 53072
Contact: Mark Krumenacher
414-691-2662
(PCB Soil Analyses)
2. Public Electric and Gas
150 Circle Ave
Cliffton,NJ07011
Contact: Joe Fink
201-365-2901
(Transformer Oil Analyses)
3. Phoenix Soil
P.O. Box 1750
Waterbury, CT 06723
Contact: Dave Green
203-759-0053
(PCB Soil Analyses)
4. Long Island Lighting
175 East Old Country Road
Hickesville, NY 11801
Contact: Bart Polizotti
516-545-5511
(Transformer Oil Analyses)
Data Quality Objective Example
This application of Dexsil's L2000 PCB/Chloride Analyzer is based on data quality objective (DQO) methods
for project planning advocated by the American Society for Testing and Materials (ASTM) [12, 13] and EPA
[14]. A DQO example was derived by ORNL from the performance results in Section 5. This example, which
is presented in Appendix E, illustrates the use of the L2000'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.
50
-------
Section 7
References
[1] Erickson, M. D. 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, pt. 761, rev. 7, December 1994.
[3] Maskarinec, M.P., et al. Stability of Volatile Organics in Environmental Soil Samples, ORNL/TM-
12128, Oak Ridge National Laboratory, Oak Ridge, Tenn., November 1992.
[4] U.S. Environmental Protection Agency. "Method 8081: Organochlorine Pesticides and PCBs as
Aroclors by Gas Chromatography: Capillary Column Technique," in Test Methods for Evaluating
Solid Waste: Physical/Chemical Methods (SW-846), 3d ed., Final Update II, Office of Solid Waste
and Emergency Response, Washington, D.C., September 1994.
[5] Oak Ridge National Laboratory. Technology Demonstration Plan: Evaluation of Polychlorinated
Biphenyl (PCB) Field Analytical Techniques, Chemical and Analytical Sciences Division, Oak Ridge
National Laboratory, Oak Ridge, Tenn., July 1997.
[6] U.S. Environmental Protection Agency. Data Quality Objectives for Remedial Response Activities,
EPA 540/G-87/003, EPA, Washington, D.C., March 1987.
[7] Sachs, Lothar. Applied Statistics: A Handbook of Techniques, 2nd ed., Springer-Verlag, New York,
1984.
[8] Snedecor, G. W., and William G. Cochran. Statistical Methods, Iowa State University Press, Ames,
Iowa, 1967.
[9] Draper, N. R., and H. Smith. Applied Regression Analysis, 2nd ed., John Wiley & Sons, New York,
1981.
[10] Berger, Walter, Harry McCarty, and Roy-Keith Smith. Environmental Laboratory Data Evaluation,
Genium Publishing Corp., Schenectady, NY., 1996.
[11] "Definition and Procedure for the Determination of the Method Detection Limit," Code of Federal
Regulations, 40 CFR, pt. 136, appendix B, rev. 1.11.
51
-------
[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.
[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, D5 792-95, 1997.
[14] U.S. Environmental Protection Agency. Guidance for Data Quality Assessment, EPA QA/G-9;
EPA/600/R-96/084, EPA, Washington, D.C., July 1996.
52
-------
Appendix A
Description of Environmental Soil Samples
53
-------
-------
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)
55
-------
-------
Appendix B
Characterization of Environmental Soil Samples
57
-------
-------
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.
59
-------
-------
Appendix C
Temperature and Relative Humidity Conditions
61
-------
-------
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
a The chamber was not operating properly on this day. See discussion in Section 3.
No developers were working outdoors on this day.
120
100 -
60
40 -
20
o 4
[ft
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.
63
-------
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.
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.
64
-------
90
80
70
T 60
>*
:§ so
i
g 4
™
"^ 30
20
10
m
fin
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-4. Summary of relative humidity conditions for chamber site.
65
-------
66
-------
Appendix D
Dexsil's L2000 PCB/Chloride Analyzer
PCB Technology Demonstration Sample Data
67
-------
Legend for Appendix D Tables
Table Heading
Obs
Sample ID
Rep
L2000 Result
Ref Lab Result
Reference Aroclor
Type
Order
Definition
Observation
Sample identification
101 to 126 = outdoor site soil samples
127 to 130 = outdoor site extract samples
201 to 226 = chamber site soil samples
227 to 230 = chamber site extract samples
Replicate of sample ID (1 through 4)
L2000'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 = environmental soil
1242, 1248, 1254, 1260 = Aroclor in PE samples
Blank = non-PCB-contaminated sample
Order of sample analysis (started with 1001-1 1 16, then
2001-2116)
68
-------
Table D-l. L2000 PCB technology demonstration soil sample data
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
Sample
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
L2000
Result
(ppm)
6.1
18.5
11.2
6.7
5.1
11.0
3.9
5.3
14.5
7.9
6.5
12.4
11.6
17.5
12.4
10.6
60.7
66.8
83.3
80.9
411.7
461.8
406.6
442.2
6.1
10.4
5.1
6.1
7.5
3.1
6.7
4.9
4.5
4.5
5.1
9.6
582.5
327.4
394.0
427.0
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
1012
1077
1098
1071
1068
1042
1063
1086
1078
1070
1058
1100
1029
1079
1088
1002
1003
1019
1033
1031
1041
1056
1104
1044
1089
1075
1016
1072
1023
1062
1092
1085
1067
1013
1051
1035
1007
1006
1005
1022
69
-------
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
Sample
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
L2000
Result
(ppm)
38.9
46.0
46.2
66.6
443.5
346.5
663.5
439.8
5.7
6.2
8.7
6.0
6.7
4.1
11.4
8.4
42.7
40.8
32.7
36.4
105.8
99.3
99.9
98.7
478.7
562.4
501.4
513.1
6.4
5.5
6.4
7.8
37.4
37.6
51.8
64.2
10.0
7.3
16.3
15.3
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
1052
1054
1026
1048
1032
1011
1034
1073
1037
1015
1059
1004
1021
1055
1099
1069
1043
1045
1097
1093
1074
1080
1083
1101
1082
1024
1090
1076
1050
1057
1018
1030
1010
1066
1046
1047
1091
1017
1036
1081
70
-------
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
Sample
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
L2000
Result
(ppm)
62.9
75.0
64.3
62.5
24.2
18.2
16.5
13.4
74.3
87.1
79.4
73.9
9.0
5.1
5.5
5.1
73.9
69.5
67.6
79.2
3.1
2.1
2.3
0.7
6.9
3.7
10.1
14.0
5.1
8.8
4.1
4.1
24.1
31.4
38.5
31.2
109.7
120.2
104.5
100.3
Ref Lab
Result
(ppm)
58.7
55.7
53.2
50.9
12.2
10.9
11.3
10.0
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
41.0
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
1254/1260
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
1049
1096
1053
1020
1084
1103
1094
1064
1087
1039
1028
1025
1095
1001
1008
1014
1060
1027
1009
1040
1102
1065
1038
1061
2008
2023
2050
2057
2063
2079
2039
2046
2041
2084
2058
2065
2053
2091
2018
2029
71
-------
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
Sample
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
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
L2000
Result
(ppm)
493.6
439.5
487.0
488.1
7.9
9.0
14.9
6.3
32.0
32.6
30.6
31.6
58.7
74.4
61.7
61.9
66.2
69.1
64.6
122.0
108.7
110.0
100.6
100.8
429.8
442.4
431.0
445.3
22.6
14.3
10.4
9.0
13.6
18.5
11.4
18.1
46.2
58.5
49.9
47.9
Ref Lab
Result
(ppm)
3305.0
538.7
457.0
483.3
2.9
1.1
1 . 1
2.5
17.8
14.3
21.6
21.6
42.0
27.7
24.0
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
2033
2076
2102
2104
2070
2083
2077
2035
2019
2080
2073
2074
2081
2025
2016
2012
2048
2062
2092
2032
2005
2055
2097
2071
2098
2082
2061
2068
2056
2072
2004
2020
2047
2049
2026
2015
2001
2011
2043
2028
72
-------
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
Sample
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
L2000
Result
(ppm)
49.1
48.5
49.5
63.1
86.4
77.6
82.7
100.2
431.9
457.1
449.6
444.1
4.6
3.0
4.1
8.4
33.9
46.1
47.0
39.2
7.7
6.7
10.4
6.3
67.2
79.6
70.5
76.4
12.1
21.3
8.3
18.7
84.6
75.2
65.6
76.1
8.4
3.7
8.4
12.6
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
2009
2103
2040
2060
2067
2086
2027
2006
2087
2059
2013
2085
2042
2021
2044
2094
2038
2051
2017
2078
2034
2088
2100
2045
2064
2095
2096
2054
2022
2093
2037
2010
2090
2089
2002
2036
2101
2024
2069
2052
73
-------
Sample L2000 Ref Lab Reference
Obs ID Rep Result Result Aroclor Type Order
(ppm) (ppm)
201 225 1 73.5 56.4 1260 1254/1260 2003
202 225 2 68.0 36.5 1016/1260 1254/1260 2075
203 225 3 77.4 32.1 1260 1254/1260 2007
204 225 4 74.4 146.0 1254 1254/1260 2066
205 226 1 0.7 <0.1 Non-Detect Blank 2030
206 226 2 1.0 <0.8 Non-Detect Blank 2014
207 226 3 3.6 <0.1 Non-Detect Blank 2099
208 226 4 1.9 <0.1 Non-Detect Blank 2031
74
-------
Table D-2. L2000 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
127
127
127
127
128
128
128
128
129
129
129
129
227
227
227
227
228
228
228
228
229
229
229
229
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
L2000
Result
(ppm)
21.0
18.3
22.9
23.6
79.6
78.9
82.6
84.3
3.2
3.0
2.7
2.6
18.8
20.5
21.4
19.3
127.8
97.0
83.9
91.3
1.8
1.9
2.0
1.3
Ref Lab
Result
(ppm)
10.9
11.4
10.9
10.4
66.9
70.7
66.5
65.6
<0.1
<0.1
<0.1
<0.1
9.2
10.7
8.9
9.7
72.8
62.0
63.8
62.2
<0.1
<0.1
<0.1
<0.1
Reference
Aroclor
1016
1016
1016
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
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
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
Order
1105
1112
1114
1115
1111
1113
1116
1107
1106
1109
1108
1110
2105
2108
2111
2112
2113
2115
2109
2114
2106
2116
2107
2110
"Nominal spike concentration of the extract sample prepared by ORNL.
75
-------
Table D-3. Corrected reference laboratory data
Error
Transcription
Calculation
Interpretation
Sample ID
106
130
205
207
210
118
119
209
214
219
ior
ior
107
109
113b
113b
119
127
201
219
Reported Result
(ppm)
<490
5.6
32,000
180
160
3.6
4.3
2.3
43.0
29.0
<0.7
<0.7
<1.3
18.0
<0.9
<1.0
18.0
7.2
< 1.0
21.0
Corrected Result
(ppm)
255.9
10.3
3,305.0
17.8
123.2
2.1
17.4
37.9
26.0
22.4
0.5
0.6
1.2
1.5
0.6
0.7
21.2
10.9
0.6
26.0
a Two of four measurements in sample ID 101 were corrected.
Two of four measurements in sample ID 113 were corrected.
76
-------
Appendix E
Data Quality Objective Example
77
-------
-------
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 which 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 Dexsil's L2000 PCB/Chloride Analyzer 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 average PCB concentration < than action level, then send the soil drum to the landfill.
If average PCB concentration action level, then send the soil drum to the incinerator.
DQO Goals
EPA's Guidance for Data Quality Assessment [14] states 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" (i.e., the
null hypothesis is true). The team required that the error rate for sending a "hot" drum to the landfill (i.e., the
false positive error rate for the decision) could not be more than 5%. Therefore, a sufficient number of samples
79
-------
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 sending a drum containing 50 ppm or more
of PCBs to the landfill.
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 made
when it is concluded that a drum is "hot" (i.e., the null hypothesis is not rejected), when in actuality, 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] for developing limits
on decision errors, the team selected the false negative decision error rate (FN) to be 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 was 40 ppm.
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 produce
data of sufficient quality to support the site decision. Because the DQO team is considering the use of the
L2000 PCB/Chloride Analyzer, the performance of this technology (as reported in this ETV report) was used
to assess its applicability to this project. Two questions arise:
1. How many samples are needed from a single drum to permit a valid estimation of the true average
concentration of PCBs in the drum to the specified certainty? Recall that the simplifying 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 variability, then, to be
considered in this example is the variability in the L2000's analytical method, which is determined
by precision studies.
2 What is the appropriate action level (AL) for using the Dexsil L2000 to make decisions in the field? After
the required number of samples have been collected from a drum and analyzed, the results are averaged
together to get an estimate of the "true" PCB concentration of the drum. When using the L2000, what is
the value (here called "the action level for the decision rule") to which that average is compared to decide
if 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 quality control regimen.
L2000 PCB/Chloride Analyzer Accuracy
The ETV demonstration results indicated that the PCB concentrations determined by the L2000 were biased
high when compared with reference concentration values. However, the L2000 data had a strong linear
correlation (R2 = 0.95) with the certified values for the performance evaluation samples. This correlation is
80
-------
represented by a line fitted to the data that predicts
the expected PCB concentration from the certified
PE value. Figure E-l shows this linear relationship
with the L2000's average PCB concentrations
plotted against the certified PCB values for the PE
samples, which were in the concentration range of 0
to 50 ppm. The arrows on the plot in Figure E-l
demonstrate a method to quickly estimate a corrected
PCB concentration from a L2000 PCB
measurement. The equation for the PCB prediction
line is
L2000result = 4.33 + 1.41
x (Certified PE value)
(E-l)
40
Certified PCB Concentration (ppm)
Figure E-l. A linear model for predicting L2000 PCB
concentrations from certified PCB concentrations with
95% confidence intervals (dashed lines).
Both Figure E-l and the equation of the PCB
prediction line (E-l) can be used to correct an L2000
result for bias. As shown in Figure E-l, an L2000
result (i.e., the j-axis) can be projected over to the
prediction line. Then, a perpendicular line can be
dropped to the certified PCB concentration axis (i.e.,
the x-axis). Two examples are illustrated by the
arrows drawn on Figure E-l. The first arrow
projects an L2000 result of 61 ppm to the certified PCB concentration of 40 ppm, and the second arrow
projects an L2000 result of 75 ppm to the certified PCB concentration of 50 ppm.
Compensation for bias may also be performed mathematically, by solving the linear PCB prediction line (Eq.
E-l) for the certified PE value (a "true" unbiased result). The resulting equation is Equation E-2:
Corrected result = (L2000 result -4.33)^1.41 (E-2)
For example, if the L2000 Result = 61 ppm the corrected result would be:
Corrected result = ( 61.0 - 4.33 ) + 1.41 = 40.2 ppm.
The DQO team discussed the two options available to them to deal with the bias in the L2000 results. One
option was to use the L2000 results as they were and accept the positive bias inherent in the L2000 results as
a conservative margin of safety. The other option was to compensate for the positive bias. If the DQO team
decided to correct for the bias, compensation could be performed by the graphical method discussed above,
where a result of 61 ppm obtained by the L2000 corresponds to a corrected result estimated to be 40 ppm.
Based on their site-specific circumstances, the team decided to compensate for the bias. They would design an
appropriate quality control regimen, which became part of the Sampling and Analysis Plan, to use with the
L2000 during their site work. This quality control program would enable them to adequately document the site-
specific performance of the L2000, and to prepare site-specific performance graphs (which would resemble
81
-------
Figure E-l), from which they would produce verifiable and defensible site-specific unbiased data. Since
unbiased data would be produced, the critical decision points of 40 ppm and 50 ppm (which correspond to CFN
and RT in Eq. E-3 below) were selected for use with the L2000 in calculating the sample size and action level
to meet the project's DQO goals.
Determining the Number of Samples
With the critical decision points selected, the team
could then determine the number of samples needed
from each drum to calculate the drum's "true" average
PCB concentration. For a homogeneous matrix, the
number of samples required depends on the precision
of the analytical method.
The L2000's results from the ETV demonstration
established that the standard deviation for PE samples
was approximately 6.2 ppm within the concentration
range of 20 to 50 ppm (see Figure E-2). This estimate
of analytical variability (precision) is used to calculate
the number of soil samples required to be analyzed
from each drum to achieve the DQOs for the project.
A formula provided in EPA's Guidance for Data
Quality Assessment [14] (pp. 3.2-3, Box 3.2-1) can be
20
40
50
Certified PCB Concentration (ppm)
adapted to this example for calculating the number of Figure E-2. L2000 standard deviations versus certified
samples required to meet the FN and the FP PCB concentrations.
requirements:
n =
(SD)
(RT - CFN)2
(E-3)
where
n
(SD)2
RT
LFN
FP
FN
number of samples from a drum to be measured,
variance for the measurement [e.g., (SD)2 = (6.2)2 ],
regulatory threshold (e.g., RT = 50 ppm),
concentration at which the FN is specified (e.g., CFN = 40 ppm),
false positive decision error rate (e.g., FP = 0.05),
false negative decision error rate (e.g., FN = 0.10), and
the (1 - p)th percentile of the standard normal distribution (see [14], Table A-l of
Appendix A). Example: Z(1_FP) = Z095 = 1.645.
Incorporating the appropriate values for the L2000 into Equation E-3 gives
82
-------
_ (1.645 + 1.282)2 + ^
(50-40)2
Therefore, five samples from each drum would be analyzed by Dexsil's L2000 PCB/Chloride Analyzer to meet
the criteria established by the DQO process. Note that, to be conservative, the sample size was rounded up to
the next integer. These five samples are averaged (by taking the arithmetic mean) to produce an L2000 value
for a drum's PCB concentration. As discussed earlier, this L2000 value can then be converted to a corrected
average drum concentration by using a graph such as Figure E-l or an equation for the PCB prediction line
(such as Eq. E-2).
Determining the Action Level
Now that the number of samples that need to be analyzed from each drum to meet the DQO goals has been
determined, the action level (AL) can be calculated. The action level is the decision criterion (or cutoff value)
that will be compared to the unbiased average PCB concentration determined for each drum. The AL for the
decision rule is calculated on the basis of regulation-driven requirements (the TSCA regulatory threshold of
50 ppm), and on controlling the FP established in the DQO process. Recall that the team set the permissible
FP error rate at 5%.
The formula [14] to compute the AL is
SD
AL = RT - Z^Fp x (R_4)
Computing the AL in this instance we find the following:
AL = 5Qppm - (1.645) * — = 45 A ppm
To summarize, five random samples from each drum are analyzed, and the biased results are corrected. The
five corrected results are averaged to produce the average PCB concentration for the drum, which is then
compared to the AL for the decision rule (45.4 ppm). Therefore, the decision rule using Dexsil's L2000
PCB/Chloride Analyzer to satisfy a 5% FP and a 10% FN (after correcting the L2000 results for bias) is as
follows:
83
-------
Decision Rule for 5% FP and 10% FN
If the corrected average PCB concentration of 5 random soil samples on a drum < 45.4 ppm,
then send the drum to the landfill.
If the corrected average PCB concentration of 5 random soil samples on a drum 45.4 ppm,
then send the drum to the incinerator.
The decision performance curve (see EPA
QC/G9, pp. 34-36) calculates the probability
of sending a drum to the incinerator for
different values of true PCB concentration in a
drum. Figure E-3 shows that the decision
performance curve has the value of PrjTake
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] = 0.03 for True = 40
ppm, which is better (at 3%) than the FN of
10% that the DQO team had originally
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.
E
3
Q
0)
.Q
O
40 45 50
True PCB Concentration ( ppm )
Figure E-3. Decision performance curve for PCB drum
example.
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 8 and the action level is calculated to be 44.9 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 the DQO team'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
84
-------
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 as shown below.
Decision Rule for FP = 1% and FN = 10%
If the corrected average PCB concentration of 8 random soil samples on a drum < 44.9 ppm,
then send the drum to the landfill.
If the corrected average PCB concentration of 8 random soil samples on a drum 44.9 ppm,
then send the drum to the incinerator.
85
------- |