United States Office of Research and EPA/600/R-01/065
Environmental Protection Development August 2001
Agency Washington, D.C. 20460
<&EPA Environmental Technology
Verification Report
Explosives Detection Technology
SRI Instruments
Model 861OC, Gas Chromatograph/
Thermionic lonization Detection
Environmental Security
Technology Certification
Program
oml
Oak Ridge National Laboratory
ETVET
-------
THE ENVIRONMENTAL TECHNOLOGY VERIFICATION PROGRAM
ll*. \ifntf
V
Environmental Security
Technology Certification
Program
Oak **<>& National Lab«rato->
Joint Verification Statement
TECHNOLOGY TYPE:
APPLICATION:
TECHNOLOGY NAME:
COMPANY:
ADDRESS:
WEB SITE:
EMAIL:
GAS CHROMATOGRAPHY
MEASUREMENT OF EXPLOSIVES IN CONTAMINATED
SOIL
Model 8610C Gas Chromatograph/Thermionic lonization
Detection
SRI Instruments
20720 Earl Street
Torrance, CA 90503
www.srigc.com
hagoldsmith@earthlink.net
PHONE: (310)214-5092
FAX: (310) 214-5097
The U.S. Environmental Protection Agency (EPA) has created the Environmental Technology Verifica-
tion Program (ETV) to facilitate the deployment of innovative or improved environmental technologies
through performance verification and dissemination of information. The goal of the ETV Program is to
further environmental protection by substantially accelerating the acceptance and use of improved and
cost-effective technologies. ETV seeks to achieve this goal by providing high quality, peer-reviewed data
on technology performance to those involved in the design, distribution, financing, permitting, purchase,
and use of environmental technologies.
ETV works in partnership with recognized standards and testing organizations, stakeholder groups
consisting of regulators, buyers, and vendor organizations, with the full participation of individual
technology developers. The program evaluates the performance of innovative technologies by developing
test plans that are responsive to the needs of stakeholders, conducting field or laboratory tests (as
appropriate), collecting and analyzing data, and preparing peer-reviewed reports. All evaluations are
conducted in accordance with rigorous quality assurance protocols to ensure that data of known and
adequate quality are generated and that the results are defensible.
EPA-VS-SCM-48
The accompanying notice is an integral part of this verification statement.
August 2001
-------
The Department of Defense (DoD) has a similar verification program known as the Environmental Security
Technology Certification Program (ESTCP). The purpose of ESTCP is to demonstrate and validate the most
promising innovative technologies that target DoD's most urgent environmental needs and are projected to
pay back the investment within 5 years through cost savings and improved efficiencies. ESTCP
demonstrations are typically conducted under operational field conditions at DoD facilities. The
demonstrations are intended to generate supporting cost and performance data for acceptance or validation of
he technology. The goal is to transition mature environmental science and technology projects through the
demonstration/ validation phase, enabling promising technologies to receive regulatory and end user
acceptance in order to be field tested and commercialized more rapidly.
The Oak Ridge National Laboratory (ORNL) is one of the verification organizations operating under the Site
haracterization and Monitoring Technologies (SCMT) program. SCMT, which is administered by EPA's
National Exposure Research Laboratory, is one of 12 technology areas under ETV. In this verification test,
ORNL evaluated the performance of explosives detection technologies. This verification statement provides
i summary of the test results for SRI Instruments' Model 8610C gas chromatograph with thermionic
onization detection (GC/TID). This verification was conducted jointly with DoD's ESTCP.
VERIFICATION TEST DESCRIPTION
This verification test was designed to evaluate technologies that detect and measure explosives in soil. The
;est was conducted at ORNL in Oak Ridge, Tennessee, from August 21 through 30, 2000. Spiked samples of
cnown concentration were used to assess the accuracy of the technology. Environmentally contaminated soil
samples, collected from DoD sites in California, Louisiana, Iowa, and Tennessee and ranging in
concentration from 0 to approximately 90,000 mg/kg, were used to assess several performance
characteristics. The primary constituents in the samples were 2,4,6-trinitrotoluene (TNT); isomeric
dinitrotoluene (DNT), including both 2,4-dinitrotoluene and 2,6-dinitrotoluene; hexahydro-l,3,5-trinitro-
1,3,5-triazine (RDX); and octahydro-l,3,5,7-tetranitro-l,3,5,7-tetrazocine (HMX). The results of the soil
analyses conducted under field conditions by the GC/TID were compared with results from reference
aboratory analyses of homogenous replicate samples analyzed using EPA SW-846 Method 8330. Details of
he verification, including a data summary and discussion of results, may be found in the report entitled
Environmental Technology Verification Report: Explosives Detection TechnologySRI Instruments,
3C/JJD, EPA/600/R-01/065.
TECHNOLOGY DESCRIPTION
The SRI Model 86IOC gas chromatograph (GC) is a transportable instrument that can provide on-site
analysis of soils for explosives. Coupling this transportable gas chromatograph with a thermionic ionization
detector (TID) allows for the determination of explosives in soil matrices following simple sample
preparation procedures. Samples are extracted in acetone, diluted, and injected directly onto the GC column
within a heated injection port. The high temperature of the injection port instantaneously vaporizes the
solvent extract and explosives, allowing them to travel as a vapor through the GC column in the presence of
he nitrogen carrier gas. The stationary phase of the GC column and the programmable oven temperature
separate the components present in sample extracts based on their relative affinities and vapor pressures.
Upon elution from the column's end, compounds containing nitro groups are ionized on the surface of the
hermionic bead, and the increased conductivity of atmosphere within the heated detector is measured with a
collector electrode. In this verification test, the instrument was verified for its ability to detect and quantify
2,4-dinitrotoluene (2,4-DNT), RDX, and TNT. Analytical run times were typically less than 7 min and
reporting limits were typically 0.5 mg/kg.
EPA-VS-SCM-48 The accompanying notice is an integral part of this verification statement. August 2001
-------
VERIFICATION OF PERFORMANCE
rhe following performance characteristics of SRI's GC/TID were observed.
yredsion: The mean relative standard deviations (RSDs) for 2,4-DNT, RDX, and TNT were 15%, 14% and
23%, respectively, indicating that the determinations of all analytes were precise.
Accuracy: Accuracy was assessed using the performance evaluation (PE) soil samples, which were spiked to
lominal TNT and RDX concentrations of 0, 10, 50, 100, 250, and 500 mg/kg each by an independent
aboratory. The mean percent recoveries for RDX and TNT were 91% and 97%, respectively, indicating that
he analyses were unbiased.
False positive/false negative results: Of the 20 blank soils, SRI reported TNT in five samples (25% false
Dositives). No false positives were reported for 2,4-DNT and RDX. False positive and false negative results
;re also estimated by comparing the GC/TID result to the reference laboratory result for the environmental
md spiked samples (e.g., whether SRI reported a result as a nondetect that the reference laboratory reported
is a detection, and vice versa). For these soils, 3% of the 2,4-DNT results and 7% of the TNT results were
"eported as false positives relative to the reference laboratory results, but none of the RDX results were
"eported as false positives. Similarly, a small percentage of the results were reported as nondetects by SRI
i.e., false negatives) when the laboratory reported a detection (2% for RDX , 4% for TNT, none for 2,4-
DNT).
Completeness: The GC/TID generated results for all 108 soil samples for a completeness of 100%.
Comparability: A one-to-one sample comparison of the GC/TID results and the reference laboratory results
/as performed for all samples (spiked and environmental) that were reported as detects. The correlation
coefficient (r) for the comparison of the entire soil data set for TNT (excluding one suspect measurement for
he reference laboratory) was 0.95 (slope (m) = 1.32). When comparability was assessed for specific
concentration ranges, the r value did not change dramatically for TNT, ranging from 0.89 to 0.93 depending
the concentrations selected. RDX correlation coefficient with the reference laboratory for all soil results
LS slightly lower than TNT (r = 0.85, m = 0.91). The GC/TID's results for RDX correlated better with the
"eference laboratory for concentrations <500 mg/kg (r = 0.96, m = 0.83) than for samples where
concentrations were >500 mg/kg (r = 0.49, m = 0.56). For the limited number of data points where both the
"eference laboratory and SRI reported results for 2,4-DNT (n = 14), the correlation was 0.44 (m = 0.33).
Sample Throughput: Throughput was approximately three samples per hour. This rate was accomplished by
wo operators and included sample preparation and analysis.
?ase of Use: No particular level of educational training is required for the operator, but knowledge of
chromatographic techniques and experience in field instrument deployments would be advantageous.
EPA-VS-SCM-48 The accompanying notice is an integral part of this verification statement. August 2001
-------
Overall Evaluation: The overall performance of the GC/TID for the analysis of 2,4-DNT, RDX, and TNT
was characterized as precise and unbiased. As with any technology selection, the user must determine if this
technology is appropriate for the application and the project's 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
Jeffrey Marqusee, Ph.D.
Director
Environmental Security Technology Certification Program
Department of Defense
W. Frank Harris, Ph.D.
Associate Laboratory Director
Biological and Environmental Sciences
Oak Ridge National Laboratory
NOTICE: EPA and ESTCP verifications are based on evaluations of technology performance under specific,
predetermined criteria and appropriate quality assurance procedures. EPA, ESTCP, and ORNL make no expressed or
implied warranties as to the performance of the technology and do not certify that a technology will always operate
as verified. The end user is solely responsible for complying with any and all applicable federal, state, and local
requirements. Mention of commercial product names does not imply endorsement or recommendation.
EPA-VS-SCM-48
The accompanying notice is an integral part of this verification statement.
August 2001
-------
EPA/600/R-01/065
August 2001
Environmental Technology
Verification Report
Explosives Detection Technology
SRI Instruments
Gas Chromatograph/Thermionic
lonization Detection
By
Amy B. Dindal
Charles K. Bayne, Ph.D.
Roger A. Jenkins, Ph.D.
Oak Ridge National Laboratory
Oak Ridge Tennessee 37831-6120
Eric N. Koglin
U.S. Environmental Protection Agency
Environmental Sciences Division
National Exposure Research Laboratory
Las Vegas, Nevada 89193-3478
This verification was conducted in cooperation with the
U.S. Department of Defense
Environmental Security Technology Certification Program
-------
Notice
The U.S. Environmental Protection Agency (EPA), through its Office of Research and Development (ORD),
and the U.S. Department of Defense's Environmental Security Technology Certification Program (ESTCP)
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.
-------
Table of Contents
List of Figures v
List of Tables vii
Acknowledgments ix
Abbreviations and Acronyms xi
1. INTRODUCTION 1
2. TECHNOLOGY DESCRIPTION 3
Technology Overview 3
Sample Preparation 3
Analytical Procedure 3
Instrument Calibration and Quantification of Sample Results 4
3. VERIFICATION TEST DESIGN 5
Objective 5
Testing Location and Conditions 5
Soil Sample Descriptions 5
Sources of Samples 5
Iowa Army Ammunition Plant 5
Louisiana Army Ammunition Plant 5
Milan Army Ammunition Plant 5
Volunteer Army Ammunition Plant 5
Fort Ord Military Base 5
Performance Evaluation Samples 6
Soil Sample Preparation 6
Sample Randomization 6
Summary of Experimental Design 7
Description of Performance Factors 7
Precision 7
Accuracy 7
False Positive/Negative Results 7
Completeness 8
Comparability 8
Sample Throughput 8
Ease of Use 8
Cost 8
Miscellaneous Factors 8
4. REFERENCE LABORATORY ANALYSES 10
Reference Laboratory Selection 10
Reference Laboratory Method 10
Reference Laboratory Performance 10
5. TECHNOLOGY EVALUATION 12
Objective and Approach 12
Precision 12
Accuracy 12
False Positive/False Negative Results 12
iii
-------
Completeness 13
Comparability 13
Sample Throughput 15
Ease of Use 15
Cost Assessment 15
GC/TID Costs 16
Labor 16
Equipment 17
Reference Laboratory Costs 17
Sample Shipment 17
Labor and Equipment 18
Cost Assessment Summary 18
Miscellaneous Factors 18
Summary of Performance 18
6. REPRESENTATIVE APPLICATIONS 20
7. REFERENCES 21
Appendix A GC/TID Sample Results Compared with Reference Laboratory Results 22
Appendix B Data Quality Objective (DQO) Example 25
IV
-------
List of Figures
1 SRI Model 8610C gas chromatograph 3
2 Comparison of reference laboratory results with GC/TID results for all
RDX soil concentrations 14
3 Comparison of reference laboratory results with GC/TID results for SRI TNT
soil concentrations <500 mg/kg 15
4 Range of percent difference values for 2,4-DNT, RDX, and TNT 16
-------
-------
List of Tables
1 Summary of Soil Samples 7
2 Summary of the Reference Laboratory Performance for Soil Samples 11
3 Summary of the Reference Laboratory Performance on Blank Samples 11
4 Summary of GC/TID Precision 12
5 Summary of GC/TID Accuracy 12
6 Number of GC/TID Results within Acceptance Ranges for Spiked Soils 13
7 Summary of GC/TID False Positive Performance on Blank Samples 13
8 Summary of the GC/TID Detect/Nondetect Performance Relative to the
Reference Laboratory Results 13
9 GC/TID Correlation with Reference Data for Various Vendor
Soil Concentration Ranges 14
10 Estimated Analytical Costs for Explosives-Contaminated Samples 17
11 Performance Summary for the GC/TID 19
vn
-------
-------
Acknowledgments
The authors wish to acknowledge the support of all those who helped plan and conduct the verification test,
analyze the data, and prepare this report. In particular, we recognize Dr. Thomas Jenkins (U.S. Army, Cold
Regions Research and Engineering Laboratory) and Dr. Michael Maskarinec (Oak Ridge National
Laboratory) who served as the technical experts for this project. We thank the people who helped us to obtain
the samples from the various sites, including Dr. Jenkins, Danny Harrelson (Waterways Experiment Station),
Kira Lynch (U.S. Army Corp of Engineers, Seattle District), Larry Stewart (Milan Army Ammunition Plant),
Dick Twitchell and Bob Elmore (Volunteer Army Ammunition Plant). For external peer review, we thank
Dr. C. L. Grant (Professor Emeritus, University of New Hampshire) and Harry Craig (US EPA, Region 10)
for EPA peer review. The authors also acknowledge the participation of Hugh Goldsmith of SRI Instruments
and Alan Hewitt and Tom Ranney of U.S. Army Cold Regions Research and Engineering Laboratory, who
performed the analyses during verification testing.
For more information on the Explosives Detection Technology Verification contact
Eric N. Koglin Roger A. Jenkins
Project Technical Leader Program Manager
Environmental Protection Agency Oak Ridge National Laboratory
Environmental Sciences Division Chemical and Analytical Sciences Division
National Exposure Research Laboratory P.O. Box 2008
P. O. Box 93478 Building 4500S, MS-6120
Las Vegas, Nevada 89193-3478 Oak Ridge, TN 37831- 6120
(702)798-2432 (865)574-4871
koglin.eric@epa.gov jenkinsra@ornl.gov
For more information on SRI Instruments' Model 8610C GC contact
Hugh Goldsmith
President
SRI Instruments
20720 Earl Street
Torrance, CA 90503
(310)214-5092
hagoldsmith@earthlink.net
www.srigc.com.
-------
-------
Abbreviations and Acronyms
%D percent difference
2-Am-DNT 2-amino-4,6-dinitrotoluene
4-Am-DNT 4-amino-2,6-dinitrotoluene
2,4-DNT 2,4-dinitrotoluene
2,6-DNT 2,6-dinitrotoluene
DNT isomeric dinitrotoluene (includes both 2,4-DNT and 2,6-DNT)
DoD U.S. Department of Defense
EPA U.S. Environmental Protection Agency
ERA Environmental Resource Associates
ESTCP Environmental Security Technology Certification Program (DoD)
ETV Environmental Technology Verification Program
FA false-acceptance error rate
fn false negative result
fp false positive result
FR false-rejection error rate
GC gas chromatograph
GC/TID gas chromatograph with thermionic ionization detection
HMX octahydro-l,3,5,7-tetranitro-l,3,5,7-tetrazine
HPLC high-performance liquid chromatograph
LAAAP Louisiana Army Ammunition Plant
MLAAP Milan Army Ammunition Plant
NERL National Exposure Research Laboratory (EPA)
NO2 nitro
OB/OD open burning and open detonation
ORNL Oak Ridge National Laboratory
PE performance evaluation
QA quality assurance
QC quality control
RDX hexahydro-1,3,5 -trinitro-1,3,5 -triazine
RSD relative standard deviation
SAI Specialized Assays, Inc.
SCMT Site Characterization and Monitoring Technologies Center
SD standard deviation
TBAOH 5 millimolar (mM) tetrabutyl-ammonium hydroxide
TID thermionic ionization detector
TNB 1,3,5-trinitrobenzene
TNT 2,4,6-trinitrotoluene
XI
-------
-------
Section 1 Introduction
The U.S. Environmental Protection Agency (EPA)
created the Environmental Technology Verification
Program (ETV) to facilitate the deployment of
innovative or improved environmental technologies
through performance verification and dissemination
of information. The goal of the ETV Program is to
further environmental protection by substantially
accelerating the acceptance and use of improved and
cost-effective technologies. ETV seeks to achieve
this goal by providing high-quality, peer-reviewed
data on technology performance to those involved in
the design, distribution, financing, permitting,
purchase, and use of environmental technologies.
ETV works in partnership with recognized standards
and testing organizations and stakeholder groups
consisting of regulators, buyers, and vendor
organizations, with the full participation of
individual technology developers. The program
evaluates the performance of innovative
technologies by developing verification test plans
that are responsive to the needs of stakeholders,
conducting field or laboratory tests (as appropriate),
collecting and analyzing data, and preparing peer-
reviewed reports. All evaluations are conducted in
accordance with rigorous quality assurance (QA)
protocols to ensure that data of known and adequate
quality are generated and that the results are
defensible.
ETV is a voluntary program that seeks to provide
objective performance information to all of the
participants in the environmental marketplace and to
assist them in making informed technology
decisions. ETV does not rank technologies or
compare their performance, label or list technologies
as acceptable or unacceptable, seek to determine
"best available technology," or approve or
disapprove technologies. The program does not
evaluate technologies at the bench or pilot scale and
does not conduct or support research. Rather, it
conducts and reports on testing designed to describe
the performance of technologies under a range of
environmental conditions and matrices.
The program now operates 12 centers covering a
broad range of environmental areas. ETV began
with a 5-year pilot phase (1995-2000) to test a wide
range of partner and procedural alternatives in
various technology areas, as well as the true market
demand for and response to such a program. In these
centers, EPA utilizes the expertise of partner
"verification organizations" to design efficient
processes for conducting performance tests of
innovative technologies. These expert partners are
both public and private organizations, including
federal laboratories, states, industry consortia, and
private sector entities. Verification organizations
oversee and report verification activities based on
testing and QA protocols developed with input from
all major stakeholder/customer groups associated
with the technology area. The verification described
in this report was administered by the Site
Characterization and Monitoring Technologies
(SCMT) Center, with Oak Ridge National
Laboratory (ORNL) serving as the verification
organization. (To learn more about ETV, visit
ETV's Web site at www.epa.gov/etv.) The SCMT
Center is administered by EPA's National Exposure
Research Laboratory (NERL), Environmental
Sciences Division, in Las Vegas, Nevada.
The Department of Defense (DoD) has a similar
verification program known as the Environmental
Security Technology Certification Program
(ESTCP). The purpose of ESTCP is to demonstrate
and validate the most promising innovative
technologies that target DoD's most urgent
environmental needs and are projected to pay back
the investment within 5 years through cost savings
and improved efficiencies. ESTCP responds to
(1) concern over the slow pace and cost of
remediation of environmentally contaminated sites
on military installations, (2) congressional direction
to conduct demonstrations specifically focused on
new technologies, (3) Executive Order 12856, which
requires federal agencies to place high priority on
obtaining funding and resources needed for the
development of innovative pollution prevention
programs and technologies for installations and in
acquisitions, and (4) the need to improve defense
readiness by reducing the drain on the Department's
operation and maintenance dollars caused by real
world commitments such as environmental
restoration and waste management. ESTCP
demonstrations are typically conducted under
operational field conditions at DoD facilities. The
demonstrations are intended to generate supporting
-------
cost and performance data for acceptance or
validation of the technology. The goal is to
transition mature environmental science and
technology projects through the demonstration/
validation phase, enabling promising technologies to
receive regulatory and end user acceptance in order
to be field tested and commercialized more rapidly.
(To learn more about ESTCP, visit ESTCP's web
site at www.estcp.org.)
EPA's ETV program and DoD's ESTCP program
established a memorandum of agreement in 1999 to
work cooperatively on the verification of
technologies that are used to improve environmental
cleanup and protection at both DOD and non-DOD
sites. The verification of field analytical
technologies for explosives detection described in
this report was conducted jointly by ETV's SCMT
Center and ESTCP. The verification was conducted
at ORNL in Oak Ridge, Tennessee, from August 21
through 30, 2000. The performances of two field
analytical techniques for explosives were
determined under field conditions. Each technology
was independently evaluated by comparing field
analysis results with those obtained using an
approved reference method, EPA SW-846 Method
8330. The verification was designed to evaluate the
field technology's ability to detect and measure
explosives in soil. The primary constituents in the
samples were 2,4,6-trinitrotoluene (TNT); isomeric
dinitrotoluene (DNT), including both 2,4-
dinitrotoluene (2,4-DNT) and 2,6-dinitrotoluene
(2,6-DNT); hexahydro-l,3,5-trinitro-l,3,5-triazine
(RDX); and octahydro-l,3,5,7-tetranitro-l,3,5,7-
tetrazocine (HMX). Naturally contaminated
environmental soil samples, ranging
in concentration from 0 to about 90,000 mg/kg, were
collected from DoD sites in California, Louisiana,
Iowa, and Tennessee, and were used to assess
several performance characteristics. This report
discusses the performance of SRI Instruments'
Model 86IOC gas chromatograph equipped with a
thermionic ionization detector (TID).
-------
Section 2 Technology Description
In this section, the vendor (with minimal editorial changes by ORNL) provides a description of the
technology and the analytical procedure used during the verification testing activities.
Figure 1. SRI Model 86 IOC gas chromatograph.
Technology Overview
The SRI Model 86IOC gas chromatograph (GC) is a
transportable instrument that can provide on-site
analysis for explosives (Figure 1). Coupling this
field-portable gas chromatograph with a thermionic
ionization detector allows for the determination of
explosives in soil matrices following simple sample
preparation procedures. The instrument has a TID
that uses an electrically heated emission source
composed of alkali metals impregnated into a
ceramic bead. When compounds containing nitro
(NO2) functional groups impinge on the bead's
surface, they are selectively ionized and measured
with a collector electrode. The stationary phase of
the GC column and the programmable oven
temperature separate the components present in
sample extracts based on their relative affinities and
vapor pressures.
For instrumental analysis, sample extracts are
injected directly onto the GC column within a
heated injection port. The high temperature of the
injection port instantaneously vaporizes the solvent
extract and explosives, allowing them to travel as a
vapor through the GC column in the presence of the
nitrogen carrier gas. Upon elution from the column's
end, compounds containing nitro groups are ionized
on the surface of the thermionic bead, and the
increased conductivity of atmosphere within the
heated detector is measured with a collector
electrode. Analytical run times are typically <7 min
long, and baseline resolution often is achieved
between explosives that are frequently identified at
munition manufacturing facilities, depots, training
ranges, and military test centers.
Sample Preparation
Soil samples were prepared by extracting 20 to 40 g
of soil with a 40 mL volume of acetone. Extraction
was performed by intermittently shaking (manual)
the soil:acetone slurry for several short time
intervals (2 min) over a 30 min extraction period,
then allowing the soil to settle. A clear aliquot of the
extractant was filtering by passing it through a
Millex SR (0.5-|J,m) filter using a disposable plastic
syringe with a Luer-Lock type fitting. To screen
sample extracts for high concentrations of
nitroaromatic compounds (i.e., TNT), a 0.25 mL
volume was transferred to a clear 2 mL vial and 0.01
mL of 5 mM tetrabutyl-ammonium hydroxide
(TBAOH [Aldrich]) was added. The formation of a
dark purple or red solution gives a visual indication
that a high concentration of nitroaromatics are
present. Depending on the color formed, extraction
aliquots were diluted from 1:2000 to 1:10, or not
diluted at all, prior to analysis.
Analytical Procedure
Manual injections of 1 |_iL volumes of the acetone
extracts are made with a 10-|J,L glass syringe
equipped with an extra long needle (6.0 to 7.0 cm),
into the injection port of a field-portable SRI Model
86IOC gas chromatograph equipped with a TID
detector. The injection port is heated to 225°C and
the oven holding the 15-m MXT-1 column (i.e.,
0.53 mm; 1.5-|J,m crossbond 100% dimethyl
polysiloxane film coating) is programmed to
separate and elute the explosives of interest. The
detector voltage and temperature are set at -3.4 V
and 250°C, respectively. The nitrogen carrier gas is
supplied at a pressure of 10 psi or greater and an on-
board air compressor set at 5 psi supplies make up
gas to the detector. Operation under these conditions
-------
requires only that electrical service and a source of the field trial, and when instrumental response for
nitrogen gas be available. an explosive of interest has changed by more than
±20%. Concentrations of explosives in sample
Instrument Calibration and extracts were calculated from curves generated from
Quantification of Sample Results tne calibration standards. Instrument performance
A five-point calibration ranging from 0.4 to 40 was continuously monitored by reanalysis of
mg/L in acetone was established at the beginning of standards after every fifth sample.
-------
Section 3 Verification Test Design
Objective
The purpose of this section is to describe the
verification test design. It is a summary of the test
plan (ORNL 2000).
Testing Location and Conditions
The verification of field analytical technologies for
explosives was conducted on the grounds outside of
ORNL's Building 5507, in Oak Ridge, Tennessee.
The temperature and relative humidity were
monitored during field testing. Over the five days of
testing, the average temperature was 83°F, and
ranged from 63 to 98°F. The average relative
humidity was 58%, and ranged from 27 to 95%.
The samples used in this study were shipped to the
testing location for evaluation by the vendors.
Explosives-contaminated soils from Army
ammunition plants in Iowa, Louisiana, Tennessee,
and a former Army base in California (Fort Ord)
were used in this verification. Because samples were
obtained from multiple DoD sites, the samples
represented a reasonable cross section of the
population of explosives-contaminated matrices,
such that the versatility of the field technology could
be evaluated. More specific details about the
samples are presented in the following sections
Soil Sample Descriptions
The primary contaminants in the soil samples were
TNT, DNT, RDX, and HMX. The samples also
contained trace amounts of 2-amino-4,6-
dinitrotoluene (2-Am-DNT) and 4-amino-2,6-
dinitrotoluene (4-Am-DNT), which are degradation
products of TNT. The total concentration of
explosives ranged from 0 to approximately
90,000 mg/kg. The following sections describe the
sites from which the samples were collected.
Sources of Samples
Iowa Army Ammunition Plant
Currently an active site, the Iowa Army Ammunition
Plant was constructed to load, assemble, and pack
various conventional ammunition and fusing
systems. Current production includes 120-mm tank
rounds, warheads for missiles, and mine systems.
During the early years of operation, the installation
used surface impoundments, landfills, and sumps for
disposal of industrial wastes containing explosives.
The major contaminants in these samples are TNT,
RDX, and HMX.
Louisiana Army Ammunition Plant
The Louisiana Army Ammunition Plant (LAAAP),
near Shreveport, Louisiana, is a government-owned
facility that began production in 1942. The facility
is currently an Army Reserve plant. Production
items at LAAAP have included metal parts for
artillery shells; the plant also loads, assembles, and
packs artillery shells, mines, rockets, mortar rounds,
and demolition blocks. As a result of these activities
and the resulting soil and groundwater
contamination, EPA placed LAAAP on the National
Priorities List of contaminated sites (Superfund) in
1989. The major constituents in the samples from
this site are TNT, RDX, and HMX, with trace levels
of 1,3,5-trinitrobenzene (TNB), DNT, 2-Am-DNT,
and 4-Am-DNT.
Milan Army Ammunition Plant
Currently active, the Milan Army Ammunition Plant
(MLAAP) in Milan, Tennessee, was established in
late 1940 as part of the pre-World War II buildup.
The facility still has ten ammunition loading,
assembly, and packaging lines. Munitions-related
wastes have resulted in soil contamination. The
primary contaminants in these soils are RDX and
TNT.
Volunteer Army Ammunition Plant
The Volunteer Army Ammunition Plant, in
Chattanooga, Tennessee, was built in 1941 to
manufacture TNT and DNT. All production ceased
in 1977. Past production practices resulted in
significant soil and groundwater contamination. In
the samples from this site, concentrations of TNT
and DNT ranged from 10 to 90,000 mg/kg, with
significantly smaller concentrations of Am-DNT
isomers.
Fort Ord Military Base
Fort Ord, located near Marina, California, was
opened in 1917 as a training and staging facility for
infantry troops and was closed as a military
installation in 1993. Since then, several nonmilitary
uses have been established on the site: California
State University at Monterey Bay has opened its
doors on former Fort Ord property, the University of
California at Santa Cruz has established a new
-------
research center there, the Monterey Institute of
International Studies will take over the officer's club
and several other buildings, and the post's airfield
was turned over to the city of Marina. The Army
still occupies several buildings.
An Army study conducted in 1994 revealed that the
impact areas at the inland firing ranges of Fort Ord
were contaminated with residues of high explosives
(Jenkins, Walsh, and Thorne 1998). Fort Ord is on
the National Priorities List of contaminated sites
(Superfund), requiring the installation to be
characterized and remediated to a condition that
does not pose unacceptable risks to public health or
the environment. The contaminant present at the
highest concentration (as much as 300 mg/kg) was
HMX; much lower concentrations of RDX, TNT,
2-Am-DNT, and 4-Am-DNT are present.
Performance Evaluation Samples
Spiked soil samples were obtained from
Environmental Resource Associates (ERA, Arvada,
Colorado). The soil was prepared using ERA's
semivolatile blank soil matrix. This matrix was a
40% clay topsoil that had been dried, sieved, and
homogenized. Particle size was 60 mesh and
smaller. The samples, also referred to as
performance evaluation (PE) samples, contained
known levels of TNT and RDX. The concentrations
that were evaluated contained 10, 50, 100, 250, and
500 mg/kg of each analyte. Prior to the verification
test, ORNL analyzed the spiked samples to confirm
the concentrations were within the performance
acceptance limits established by the preparation
laboratory. The method used was a modified
Method 8330, similar to the reference laboratory
method described in Section 4. For the verification
test, four replicates were prepared at each
concentration level.
Blank soil samples were evaluated to determine the
technology's ability to identify samples with no
contamination (i.e., to ascertain the false positive
error rate). The soil was collected in Monroe
County, Tennessee, and was certified by ORNL to
be free of contamination prior to verification testing.
A reasonable number of blanks (N = 20) was chosen
to balance the uncertainty for estimating the false
positive error rate and the required number of blank
samples to be measured.
Soil Sample Preparation
A few weeks prior to the verification test, all of the
soil samples were shipped in plastic bags at ambient
temperature to ORNL. The samples were stored
frozen (<0°C) prior to preparation. To ensure that
the vendors and the reference laboratory analyzed
comparable samples, the soils were homogenized
prior to sample splitting. The process was as
follows. The sample was kneaded in the Ziplock
plastic bag to break up large clumps. Approximately
1500 g of soil was poured into a Pyrex pan, and
debris was removed. The sample was then air dried
overnight. The sample was sieved using a 10-mesh
(2-mm particle size) screen and placed in a 1-L
widemouthed jar. After thorough mixing with a
metal spatula, the sample was quartered. After
mixing each quarter, approximately 250 g from each
quarter was placed back in the 1-L widemouthed jar,
for a total sample amount of approximately 1000 g.
Analysis by the ORNL method confirmed sample
homogeneity (variability of 20% relative standard
deviation or less for replicate measurements). The
sample was then split into subsamples for analysis
during the verification test. Each 4-oz sample jar
contained approximately 20 g of soil. Four replicate
splits of each soil sample were prepared for each
participant. The design included a one-to-one
pairing of the replicates, such that the vendor and
reference lab samples could be directly matched.
Three replicate sets of samples were also prepared
for archival storage. To ensure that degradation did
not occur, the soil samples were frozen (<0°C) until
analysis (Maskarinec et al. 1991).
Sample Randomization
The samples were randomized in two stages. First,
the order in which the filled jars were distributed
was randomized so that the same vendor did not
always receive the first jar filled for a given sample
set. Second, the order of analysis was randomized so
that each participant analyzed the same set of
samples, but in a different order. Each jar was
labeled with a sample number. Replicate samples
were assigned unique (but not sequential) sample
numbers. Spiked materials and blanks were labeled
in the same manner, such that these quality control
samples were indistinguishable from other samples.
All samples were analyzed blindly by both the
developer and the reference laboratory.
-------
Summary of Experimental Design
The distribution of samples from the various sites is
described in Table 1. A total of 108 soil samples
were analyzed, with approximately 60% of the
samples being naturally contaminated environmental
soils, and the remaining 40% being spikes and
blanks. Four replicates were analyzed for each
sample type. For example, 4 replicate splits of each
of 3 Fort Ord soils were analyzed, for a total of
12 individual Fort Ord samples.
Table 1. Summary of Soil Samples
Sample source or type
Fort Ord
Iowa
LAAAP
MLAAP
Umatilla
Volunteer
Spiked
Blank
Total
No. of soil samples
12
4
16
20
0
12
24
20
108
Description of Performance Factors
In Section 5, technology performance is described in
terms of precision, accuracy, completeness, and
comparability, which are indicators of data quality
(EPA 1996). False positive and negative results,
sample throughput, and ease of use are also
described. Each of these performance characteristics
is defined in this section.
Precision
Precision is the reproducibility of measurements
under a given set of conditions. Standard deviation
(SD) and relative standard deviation (RSD) for
replicate results are used to assess precision, using
the following equation:
RSD = (SD/average concentration) x 100% (Eq. 1)
The overall RSD is characterized by three summary
values:
mean (average);
median (50th percentile value, at which 50% of
all individual RSD values are below and 50%
are above); and
range (the highest and lowest RSD values that
were reported).
The average RSD may not be the best representation
of precision, but it is reported for convenient
reference. RSDs greater than 100% should be
viewed as indicators of large variability and possibly
non-normal distributions.
Accuracy
Accuracy represents the closeness of the tech-
nology's measured concentrations to known (in this
case, spiked/PE) values. Accuracy is assessed in
terms of percent recovery, calculated by the
following equation:
% recovery = (measured concentration/
known concentration) x 100% (Eq. 2)
As with precision, the overall percent recovery is
characterized by three summary values: mean,
median, and range.
False Positive/Negative Results
A false positive (fp) result is one in which the
technology detects explosives in the sample when
there actually are none (Berger, McCarty, and Smith
1996). A false negative (fn) result is one in which
the technology indicates that no explosives are
present in the sample, when there actually are
(Berger, McCarty, and Smith 1996). The evaluation
of fp and fn results is influenced by the actual
concentration in the sample and includes an
assessment of the reporting limits of the technology.
False positive results are assessed in two ways.
First, the results are assessed relative to the blanks
(i.e., the technology reports a detected value when
the sample is a blank). Second, the results are
assessed on environmental and spiked samples
where the analyte was not detected by the reference
laboratory (i.e., the reference laboratory reports a
nondetect and the field technology reports a
detection). False negative results, also assessed for
environmental and spiked samples, indicate the
frequency that the technology reported a nondetect
(i.e., < reporting limits) and the reference laboratory
reported a detection. Note that the reference
laboratory results were validated by ORNL so that
-------
fp/fn assessment would not be influenced by faulty
laboratory data. The reporting limit is considered in
the evaluation. For example, if the reference
laboratory reported a result as 0.9 mg/kg, and the
technology's paired result was reported as below
reporting limits (<1 mg/kg), the technology's result
was considered correct and not a false negative
result.
Completeness
Completeness is defined as the percentage of
measurements that are judged to be usable (i.e., the
result is not rejected). The acceptable completeness
is 95% or greater.
Comparability
Comparability refers to how well the field
technology and reference laboratory data agree. The
difference between accuracy and comparability is
that accuracy is judged relative to a known value,
and comparability is judged relative to the results of
a standard or reference procedure, which may or
may not report the results accurately. Note that the
reference laboratory result is not assumed to be the
"correct" result. This evaluation is performed for
comparison of the field analytical technology result
with what a typical fixed analytical laboratory might
report for the same sample. A one-to-one sample
comparison of the technology results and the
reference laboratory results is performed in
Section 5.
A correlation coefficient quantifies the linear
relationship between two measurements (Draper and
Smith 1981). The correlation coefficient is denoted
by the letter r; its value ranges from -1 to +1, where
0 indicates the absence of any linear relationship.
The value r = -1 indicates a perfect negative linear
relation (one measurement decreases as the second
measurement increases); the value r = +1 indicates a
perfect positive linear relation (one measurement
increases as the second measurement increases). The
slope of the linear regression line, denoted by the
letter m, is related to r. Whereas r represents the
linear association between the vendor and reference
laboratory concentrations, m quantifies the amount
of change in the vendor's measurements relative to
the reference laboratory's measurements. A value of
+1 for the slope indicates perfect agreement. (It
should be noted that the intercept of the line must be
close to zero [i.e., not statistically different from
zero], in order for the slope value of+1 to indicate
perfect agreement.) Values greater than 1 indicate
that the vendor results are generally higher than the
reference laboratory's, while values less than 1
indicate that the vendor results are usually lower
than the reference laboratory's.
In addition, a direct comparison between the field
technology and reference laboratory data is
performed by evaluating the percent difference
(%D) between the measured concentrations, defined
as:
%D = ([field technology] - [reflab\)l(reflab)
x 100% (Eq. 3)
The range of %D values is summarized and reported
in Section 5.
Sample Throughput
Sample throughput is a measure of the number of
samples that can be processed and reported by a
technology in a given period of time. This is
reported in Section 5 as the number of samples per
hour times the number of analysts.
Ease of Use
A significant factor in purchasing an instrument or a
test kit is how easy the technology is to use. Several
factors are evaluated and reported on in Section 5:
What is the required operator skill level (e.g.,
technician or advanced degree)?
How many operators were used during the
verification test? Could the technology be run
by a single person?
How much training would be required in order
to run this technology?
How much subjective decision-making is
required?
Cost
An important factor in the consideration of whether
to purchase a technology is cost. Costs involved
with operating the technology and the standard
reference analyses are estimated in Section 5. To
account for the variability in cost data and
assumptions, the economic analysis is presented as a
list of cost elements and a range of costs for sample
analysis. Several factors affect the cost of analysis.
Where possible, these factors are addressed so that
decision makers can independently complete a site-
specific economic analysis to suit their needs.
-------
Miscellaneous Factors generated during the analyses, the ruggedness of the
Any other information that might be useful to a technology, the amount of electrical or battery
person who is considering purchasing the power necessary to operate the technology, and
technology is documented in Section 5. Examples of aspects of the technology or method that make it
information that might be useful to a prospective easy to use.
purchaser are the amount of hazardous waste
-------
Section 4 Reference Laboratory Analyses
Reference Laboratory Selection
The verification process is based on the presence of
a statistically validated data set against which the
performance of the technology may be compared.
The choice of an appropriate reference method and
reference laboratory are critical to the success of the
verification test. To assess the performance of the
explosives field analytical technologies, the data
obtained from verification test participants were
compared to data obtained using conventional
analytical methods.
The first evaluation of explosives-detection
technologies under the ETV program occurred in
1999. Specialized Assays, Inc. (SAI), now known as
TestAmerica Inc., of Nashville, Tennessee, was
selected as the reference laboratory for that study. A
sample holding time study performed by ORNL in
May 2000 indicated that the concentration of
explosives in the samples had not changed
significantly. Therefore, archived soil samples and
the reference laboratory data generated in 1999 were
used for comparison with the vendor results.
The following describes how SAI was chosen to
perform the 1999 analyses. Specialized Assays, Inc.
was selected to perform the analyses based on
ORNL's experience with laboratories capable of
performing explosives analyses using EPA SW-846
Method 8330. ORNL reviewed Specialized Assays'
record of laboratory validation performed by the
U.S. Army Corps of Engineers (Omaha, Nebraska).
EPA and ORNL decided that, based on the
credibility of the Army Corps program and ORNL's
prior experience with the laboratory, Specialized
Assays would be selected to perform the reference
analyses.
ORNL conducted an audit of Specialized Assays'
laboratory operations on May 4, 1999. This
evaluation focused specifically on the procedures
that would be used for the analysis of the
verification test samples. Results from this audit
indicated that Specialized Assays was proficient in
several areas, including quality management,
document/record control, sample control, and
information management. Specialized Assays was
found to be compliant with implementation of
Method 8330 analytical procedures. The company
provided a copy of its QA plan, which details all of
the QA and quality control (QC) procedures for all
laboratory operations (Specialized Assays 1999).
The audit team noted that Specialized Assays had
excellent procedures in place for data backup,
retrievability, and long-term storage. ORNL
conducted a second audit at Specialized Assays
while the analyses were being performed. Since the
initial qualification visit, management of this
laboratory had changed because Specialized Assays
became part of TestAmerica. The visit included
tours of the laboratory, interviews with key
personnel, and review of data packages. Overall, no
major deviations from procedures were observed,
and laboratory practices appeared to meet the QA
requirements of the technology test plan (ORNL
1999).
Reference Laboratory Method
The reference laboratory's analytical method,
presented in the technology test plan, followed the
guidelines established in EPA SW-846 Method 8330
(EPA 1994). According to Specialized Assays'
procedures, soil samples were prepared by
extracting 2-g samples of soil in acetonitrile by
sonication for approximately 16 h. An aliquot of the
extract was then combined with a calcium chloride
solution to precipitate out suspended particulates.
After the solution was filtered, the filtrate was ready
for analysis. The analytes were identified and
quantified using a high-performance liquid
chromatograph (HPLC) with a 254-nm UV detector.
The primary analytical column was a C-18 reversed-
phase column with confirmation by a secondary
cyano column. The practical quantitation limit for
soil was 0.5 mg/kg.
Reference Laboratory Performance
ORNL validated all of the reference laboratory data
according to the procedure described in the test plan
(ORNL 2000). During the validation, the following
aspects of the data were reviewed: completeness of
the data package, adherence to holding time
requirements, correctness of the data, correlation
between replicate sample results, evaluation of QC
sample results, and evaluation of spiked sample
results. Each of these categories is described in
detail in the test plan. The reference laboratory
reported valid results for all samples, so
10
-------
completeness was 100%. Preanalytical holding time
requirements (14 days to extract; 40 days to analyze)
were met. A few errors were found in a small
portion of the data (-4%). Those data were
corrected for transcription and calculation errors
that were identified during the validation. One data
point, a replicate Iowa soil sample, was identified as
suspect. The result for this sample was 0.8 mg/kg;
the results from the other three replicates averaged
27,400 mg/kg. Inclusion or exclusion of this data
point in the evaluation of comparability with the
field technology (reported in Section 5) did not
significantly change the r value, so it was included
in the analysis. The reference laboratory results for
QC samples were flagged when the results were
outside the QC acceptance limits. The reference
laboratory results were evaluated by a statistical
analysis of the data. Due to the limited results
reported for the other Method 8330 analytes, only
the results for the major constituents in the samples
(2,4-DNT, HMX, RDX, and TNT) are evaluated in
this report.
The accuracy and precision of the reference
laboratory results are summarized in Table 2.
Accuracy was assessed using the spiked samples,
while precision was assessed using the results from
both spiked and environmental samples. The
reference laboratory results were unbiased
(accurate), as mean percentage recovery values were
near 100%. The reference laboratory results were
precise; all but one of the mean RSDs were less than
30%. The one mean RSD that was greater than 30%
(soil, DNT, 56%) was for a limited data set of three,
and the problem was caused by one replicate of one
sample.
Table 3 presents the laboratory results for blank
samples. A false positive result is identified as any
detected result on a known blank. For the soil
samples, one false positive detection appeared to be
a preparation error because the concentration was
near 70,000 mg/kg. Overall, it was concluded that
the reference laboratory results were unbiased,
precise, and acceptable for comparison with the
field analytical technology.
Table 2. Summary of the Reference Laboratory Performance for Soil Samples
Statistic
Mean
Median
Range
Accuracy
(% recovery)
RDX
N = 20
102
99
84-141
TNT
N = 20
100
96
76-174
Precision"
(% RSD)
2,4-DNT
NR = 3*
56
25
19-123
HMX
NR=13
29
30
12-63
RDX
NR=13
25
21
4-63
TNT
NR=18
29
25
2-72
"Calculated from those samples where all four replicates were reported as a detect.
6NR represents the number of replicate sets; N represents the number of individual samples.
Table 3. Summary of the Reference Laboratory
Performance on Blank Samples
Statistic
Number of data points
Number of detects
% of fp results
Soil
DNT
20
0
0
HMX
20
0
0
RDX
20
0
0
TNT
20
2
10
11
-------
Section 5 Technology Evaluation
Objective and Approach
The purpose of this section is to present a statistical
evaluation of the GC/TID data and determine the
instrument's ability to measure explosives in
contaminated soil samples. The technology's
performance is presented for 2,4-DNT, RDX, and
TNT, including an evaluation of comparability
through a one-to-one comparison with the reference
laboratory data. Other aspects of the technology
(such as cost, sample throughput, hazardous waste
generation, and logistical operation) are also
evaluated in this section. Appendix A contains the
raw data, provided by the vendor during the
verification test, that were used to assess the
performance of the GC/TID.
Precision
Precision is the reproducibility of measurements
under a given set of conditions. Precision was
determined by examining the results of blind
analyses for four replicates of each sample. Data
were evaluated for only those samples where all four
replicates were reported as a detect. For example,
for RDX, NR = 13 (13 sets of four replicates)
represents a total of 52 individual sample analyses.
A summary of the overall precision of the GC/TID
for the soil sample results is presented in Table 4.
For the soil samples, the mean RSDs for 2,4-DNT,
RDX, and TNT were 15%, 14%, and 23%,
respectively, indicating that the analyses were
precise.
Accuracy
Accuracy represents the closeness of the GC/TID's
measured concentrations to the known content of
spiked samples. A summary of the GC/TID's overall
accuracy for the soil results is presented in Table 5.
Note that the PE samples were spiked with only
TNT and RDX, so accuracy for 2,4-DNT could not
be evaluated. For the soil samples, the recoveries for
both RDX and TNT were similar, ranging from 74
to 112% overall. The mean recovery values for RDX
= 91% and TNT = 97% suggested that the results
were unbiased. Based on the performance
acceptance ranges shown in Table 6, which are the
guidelines established by the provider of the spiked
materials to gauge acceptable analytical results, 90%
of the results (18 of 20) met the acceptance criteria
for RDX, while 100% (20 of 20) met the criteria for
TNT.
False Positive/False Negative Results
Table 7 shows the GC/TID performance for false
positive (fp) results for blank samples. Of the
20 blank soils, SRI did not report 2,4-DNT or RDX
in any samples, but reported TNT in 5 samples
(25% fp). Table 8 summarizes the GC/TID's fp and
fn results relative to the reference laboratory results.
(See Section 3 for a more detailed discussion of this
evaluation.) For the environmental and spiked soils,
3% of the 2,4-DNT results, 0% of the RDX results,
and 7% of the TNT results were reported as false
positives relative to the reference laboratory results
(i.e., the laboratory reported the analyte as a
nondetect when SRI reported it as a detect). In the
case where the laboratory reported a detection and
SRI reported a nondetect (i.e., false negatives), 0%
Table 4. Summary of the GC/TID Precision
Statistic
Mean
Median
Range
SoilRSDfl
(%)
2,4-DNT
NR = 4*
15
9.0
8.9-31
RDX
NR=13
14
10
5-44
TNT
NR=17
23
13
2-107
"Calculated from only those samples where all four
replicates were reported as a detect.
b NR represents the number of replicate sets.
Table 5. Summary of the GC/TID Accuracy
Statistic
Mean
Median
Range
Soil recovery (%)
RDX
N = 20
91
90
74-112
TNT
N = 20
97
96
87-110
12
-------
Table 6. Number of GC/TID Results within Acceptance Ranges for Spiked Soils
Spike
concentration
(mg/kg)
10
50
100
250
500
RDX
Acceptance range
(mg/kg)
8-11
38-57
76-113
190-283
379-566
No. of results
within range
3 out of 4
4 out of 4
3 out of 4
4 out of 4
4 out of 4
TNT
Acceptance range
(mg/kg)
7-13
35-63
70-126
174-315
348-630
No. of results
within range
4 out of 4
4 out of 4
4 out of 4
4 out of 4
4 out of 4
Table 7. Summary of GC/TID False Positive Performance on Blank
Samples
Statistic
Number of data points
Number of fp results
% of total results which were fp
2,4-DNT
20
0
0%
RDX
20
0
0%
TNT
20
5
25%
Table 8. Summary of the GC/TID Detect/Nondetect Performance
Relative to the Reference Laboratory Results
Statistic
Number of data points lab reported as non-detect
Number of fp results by GC/TID
% of total results which were fp
Number of data points lab reported as detection
Number of fn results by GC/TID
% of total results which were fn
2,4-DNT
74
2
3%
14
0
0%
RDX
33
0
0%
55
1
2%
TNT
14
1
7%
74
3
4%
of the 2,4-DNT, 2% of the RDX, and 4% of the
TNT results were false negatives.
Completeness
Completeness is defined as the percentage of
measurements that are judged to be usable (i.e., the
result was not rejected). Valid results were obtained
by the technology for all 108 soil samples.
Therefore, completeness was 100%.
Comparability
Comparability refers to how well the GC/TID and
reference laboratory data agreed. In this evaluation,
the laboratory results are not presumed to be the
"correct" answers. Rather, these results represent
what a typical fixed laboratory would report for
these types of samples. A one-to-one sample
comparison of the GC/TID results and the reference
laboratory results was performed for all
13
-------
environmental and spiked samples that were
reported as a detect. (Please refer to Appendix A to
review the raw data. See Section 4 for a complete
evaluation of the reference laboratory results.) In
Table 9, the comparability of the results are
presented in terms of correlation coefficients (r) and
slopes (m) of the linear regression analysis.
A limited number of comparable data points (N =
14) was available for 2,4-DNT. All of these values
were less than 50 mg/kg. The correlation coefficient
for the comparison was 0.44 (m = 0.33) for the
entire soil data set of 2,4-DNT results. Figure 2 is a
plot of the GC/TID RDX results versus those for the
reference laboratory for all results (N = 52). One
unusual SRI result at approximately 8,000 mg/kg is
highlighted by a box surrounding the point. This
data point greatly influenced the correlation
coefficient value. The GC/TID correlation
coefficient with the reference laboratory for all
Table 9. GC/TID Correlation with Reference Data for Various Vendor Soil Concentration
Ranges
Vendor
concentration
range
All values *
< 500 mg/kg d
>500 mg/kg
>10,000 mg/kg
2,4-DNT
Correlation
coefficient
(r)
0.44
0.44
n/ae
n/a
Slope
(m)
0.33
0.33
n/a
n/a
RDX
Correlation
coefficient
(r)
0.85/0.96 c
0.96
0.49/0.84
n/a
Slope
(m)
0.91/0.86
0.83
0.56/0.64
n/a
TNTa
Correlation
coefficient
(r)
0.95
0.89
0.93
0.89
Slope
(m)
1.32
0.72
1.46
1.46
"Excluding the one reference laboratory TNT unusual value.
'Excluding those values reported as "< reporting limits."
"Including/excluding the one SRI unusual value.
rfBased on SRI's reported values.
"No values above were reported at this concentration level.
1000
2000 3000 4000 5000 6000
Reference laboratory RDX result (mg/kg)
7000
Figure 2. Comparison of reference laboratory results with GC/TID results for all RDX soil
concentrations. The data point highlighted by a box is considered an unusual value
for SRI. Please refer to Table 9 for regression constants calculated with and
without this unusual value.
14
-------
samples was 0.85 (#7 = 0.91) including the unusual
value and 0.96 (m = 0.86) excluding it. The effect is
even more pronounced when the results are
evaluated exclusively for samples where
concentrations were >500 mg/kg [r = 0.49 (m =
0.56) including the unusual value, while r = 0.84
(m = 0.64) excluding it]. The GC/TID's correlation
coefficient with the reference laboratory for RDX
concentrations <500 mg/kg was r = 0.96 (m = 0.83).
The correlation coefficient for the comparison of the
entire soil data set for TNT (excluding one suspect
measurement for the reference laboratory) was
0.95 (m = 1.32). When comparability was assessed
for specific concentration ranges, the r value did not
change dramatically for TNT, ranging from 0.89 to
0.93 depending on the concentrations selected.
Figure 3 presents a plot of the GC/TID TNT results
versus those for the reference laboratory for
concentrations <500 mg/kg. As this figure indicates,
the GC/TID TNT soil measurements generally
agreed with but were slightly lower than the
reference laboratory results.
Another metric of comparability is the %D between
the reference laboratory and the GC/TID results.
The ranges of %D values for 2,4-DNT, RDX, and
TNT are presented in Figure 4. Acceptable %D
values would be between -25% and 25% or near the
middle of the x-axis of the plots. For 2,4-DNT, the
values were mostly negative, indicating that the
GC/TID result was usually less than the reference
laboratory result. For RDX, 65% of the %D values
were between -25% and 25%, supporting the
conclusions that the RDX results generally agreed
with the reference laboratory results. For TNT, most
of the %D values are near the middle of the x-axis,
with 45% of the results were between -25% and
25%. The median absolute %D values for 2,4-DNT,
RDX, and TNT were 66%, 19%, and 28%,
respectively.
Sample Throughput
Sample throughput is representative of the estimated
amount of time required to prepare and analyze the
sample and perform the data analysis. Operating in
the field, the two-person SRI team accomplished a
sample throughput rate of approximately three
samples per hour for the 108 soil analysis. The
instrument run time for each analysis was
approximately 7 min.
Ease of Use
Two operators were used for the test because of the
number of samples and working conditions, but the
technology can be operated by a single person. SRI
Instruments provides a free one-half day of training
at their Torrance, California facility with an
instrument purchase. Users unfamiliar with gas
chromatography may need one or two days of
additional training to operate the instrument. No
particular level of educational training is required
for the operator.
Cost Assessment
The purpose of this economic analysis is to estimate
800 -,
700 -
600 -
r=0.89
m = 0.72
intercept = 8 mg/kg
100 200 300 400 500 600
Reference laboratory TNT result (mg/kg)
700
800
Figure 3. Comparison of reference laboratory results with GC/TID results for SRI TNT soil
concentrations <500 mg/kg.
15
-------
25 -,
c.
cs
I
z
Range of percent difference values
Figure 4. Range of percent difference values for 2,4-DNT, RDX, and TNT.
the range of costs for analysis of explosives-
contaminated soil samples using the GC/TID and a
conventional analytical reference laboratory method.
The analysis was based on the results and
experience gained from this verification test, costs
are provided by SRI, and representative costs
provided by the reference analytical laboratories that
offered to analyze these samples. To account for the
variability in cost data and assumptions, the
economic analysis is presented as a list of cost
elements and a range of costs for sample analysis by
the GC/TID instrument and by the reference
laboratory.
Several factors affected the cost of analysis. Where
possible, these factors were addressed so that
decision makers can complete a site-specific
economic analysis to suit their needs. The following
categories are considered in the estimate:
sample shipment costs,
labor costs, and
equipment costs.
Each of these cost factors is defined and discussed
and serves as the basis for the estimated cost ranges
presented in Table 10. This analysis assumed that
the individuals performing the analyses are fully
trained to operate the technology. Costs for sample
acquisition and preanalytical sample preparation,
which are tasks common to both methods, were not
included in this assessment.
GC/TID Costs
The costs associated with using the GC/TID
instrument included labor, equipment, and waste
disposal costs. No sample shipment charges were
associated with the cost of operating the instrument
because the samples were analyzed on-site.
Labor
Labor costs included mobilization/demobilization,
travel, per diem expenses and on-site labor.
Mobilization/demobilization. This cost element
included the time for one person to prepare for
and travel to each site. This estimate ranged
from 0 (if the analyst was on-site) to 5 h, at a
rateof$50/h.
Travel. This element was the cost for the
analyst(s) to travel to the site. If the analyst is
located at the site, the cost of commuting to the
site would be zero. The estimated cost of an
analyst traveling to the site for this verification
test ($1000) included the cost of airline travel
and rental car fees.
Per diem expenses. This cost element included
food, lodging, and incidental expenses. The
estimate ranged from zero (for a local site) to
16
-------
Table 10. Estimated Analytical Costs for Explosives-Contaminated Samples
Analysis method:
Analyst/manufacturer:
GC/TID
SRI Instruments
Sample throughput: 3 samples/h
Analysis method:
Analyst/manufacturer:
EPA SW-486 Method 8330
Reference laboratory
Typical turnaround: 21 working days
Cost category
Cost ($)
Cost category
Cost ($)
Sample shipment
0
Sample shipment
Labor
Overnight shipping
100-200
50-150
Labor
Mobilization/demobilization 0-250
Travel 0-1,000 per analyst
Per diem expenses 0-150/day per analyst
Rate 30-75/h per analyst
Equipment
Mobilization/demobilization 0-150
Instrument purchase price 8,995
Reagents/supplies variable
Labor
Mobilization/demobilization Included "
Travel
Per diem expenses
Rate
Included
Included
150-188 per sample
Equipment
Included
""Included" indicates that the cost is included in the labor rate.
$150/day for each analyst.
Rate. The cost of the on-site labor was estimated
at a rate of $30-75/h, depending on the required
expertise level of the analyst. This cost element
included the labor involved with the entire
analytical process, comprising sample
preparation, sample management, analysis, and
reporting.
Equipment
Equipment costs included mobilization/
demobilization, rental fees or purchase of
equipment, and the reagents and other consumable
supplies necessary to complete the analysis.
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
verification test, the cost of shipping equipment
and supplies was estimated at $150.
Instrument purchase. At the time of the
verification test, the SRI Model 86IOC gas
chromatograph equipped with TID detector,
heated injector, built-in air compressor,
PeakSimple serial data system and 15-m MXT-1
capillary column has a list price of $8,995. It
comes with a reusable plastic shipping container
that allows the GC to ship via overnight express
services, and even meets the size and weight
limits for airline baggage. As with any
instrument, the cost on a per-sample basis would
decrease as the number of analyses performed
increases. The instrument can also be rented
from SRI for $67.46 per day (12-day minimum)
or $1799 per month. With the purchase of an
instrument, SRI offers one-half day training at
their Torrance, California facility on the
functional aspects of the instrument. If the
operator requires additional training in basic gas
chromatography, SRI has a list of
recommendations of qualified people in their
manual.
Reagents/supplies. These items are consumable
and are purchased on a per sample basis.
Reference Laboratory Costs
Sample Shipment
Sample shipment costs to the reference laboratory
included the overnight shipping charges, as well as
labor charges associated with the various
organizations involved in the shipping process.
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,
17
-------
completing the chain-of-custody
documentation, and completing the shipping
forms. The estimate to complete this task
ranged from 2 to 4 h at $50/h.
Overnight shipping. The overnight express
shipping service cost was estimated to be $50
for one 50-lb cooler of samples.
Labor and Equipment
The labor bids from commercial analytical reference
laboratories that offered to perform the reference
analysis for this verification test ranged from $150
to $188 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.
This rate was a fully loaded analytical cost that
included equipment, labor, waste disposal, and
report preparation.
Cost Assessment Summary
An overall cost estimate for use of the GC/TID
instrument versus use of the reference laboratory
was not made because of the extent of variation in
the different cost factors, as outlined in Table 10.
The overall costs for the application of any
technology would 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's
providing immediate answers versus the reference
laboratory's provision of reporting data within 30
days of receipt of samples.
Miscellaneous Factors
The following are general observations regarding
the field operation and performance of the GC/TID
instrument:
The system, which weighs approximately 75 Ib,
was transportable by one person. The instrument
comes with a plastic carrying case that can be
used to ship the instrument or be checked as
baggage on an airplane.
The instrument appeared to be rugged, as the
analysts were able to run the instrument during a
late afternoon storm that had strong winds.
The SRI team completely disassembled their
work station at the close of each day. It took the
team less than an hour each morning to prepare
for sample analyses.
The instrument required 110 V of electrical
power for operation.
The SRI team employed a colorimetric method
to screen samples for high concentration of TNT
and related nitroaromatic compounds. This
undoubtedly prevented unnecessary overloading
of the instrument and potential downtime.
Sample preparation, including extraction,
colorimetric screening, and dilutions was
completed for all 108 samples in 18 h of labor
by one analyst (approximately 10 min per
sample).
Other SW-846 Method 8330 analytes (e.g.,
trinitrobenzene, tetryl, HMX) could potentially
be determined by this method, but this was not
verified in this study.
Hazardous waste generated during the test
included the following, which was classified as
RCRA waste: 0.2 L of vials with acetone and
trace explosives; 0.5 L of syringe filters with
spent acetone and trace explosives; 0.3 L of
acetone used for rinsing; and 4.3 L of acetone
and soil mixtures.
Summary of Performance
A summary of performance is presented in Table 11.
Precision, defined as the mean RSD, was 15%, 14%,
and 23% for 2,4-DNT, RDX, and TNT soil sample
results, respectively. Accuracy, defined as the mean
percent recovery relative to the spiked
concentration, was 91% and 97% for RDX and TNT
soil sample results, respectively, indicating that the
soil results were unbiased. Of the 20 blank soil
samples, SRI reported TNT in five samples (25%
false positives); no false positives were reported for
2,4-DNT or RDX . Additionally, false positive and
false negative results were determined by comparing
the GC/TID result to the reference laboratory result
for the environmental and spiked samples. None of
the RDX results were reported as false positives, but
3% of the 2,4-DNT and 7% of the TNT results were,
relative to the reference laboratory results. A low
percentage of results were also found to be false
negatives (0% for 2,4-DNT, 2% for RDX, and 4%
for TNT). The SRI results were comparable to the
reference laboratory results, with r values of 0.85
and 0.95 for RDX and TNT, respectively.
18
-------
The verification test found that the GC/TID
instrument was relatively simple for a trained
analyst to operate in the field, requiring less than an
hour for initial setup. The sample throughput of the
GC/TID was three samples per hour. Two operators
analyzed samples during the verification test, but the
technology can be run by a single trained operator.
The overall performance of the GC/TID for the
analysis of 2,4-DNT, RDX, and TNT was
characterized as unbiased and precise for soil
analyses.
Table 11. Performance Summary for the GC/TID
Feature/Parameter
Precision
Accuracy
False positive results on blank
samples
False positive results relative to
reference laboratory results
False negative results relative to
reference laboratory results
Comparison with reference
laboratory results (all data,
excluding suspect values)
Completeness
Weight
Sample throughput (2 operators)
Power requirements
Training requirements
Cost
Performance summary
Mean RSD
2,4-DNT:
RDX:
TNT:
Mean recovery
RDX:
TNT:
2,4-DNT:
RDX:
TNT:
2,4-DNT:
RDX:
TNT:
2,4-DNT:
RDX:
TNT:
15%
14%
23%
91%
97%
none
none
25%
3%
none
7%
none
2%
4%
Correlation Absolute median
coefficient Slope percent difference
2,4-DNT: 0.44 0.33 66%
RDX: 0.96 0.86 19%
TNT: 0.95 1.32 28%
Range of
percent difference
-83% to 2 16%
-51% to 275%
-85% to 475%
100% of 108 soil samples
75 Ib
3 samples per hour
110V
One-half day instrument-specific training
Instrument: $8,995
19
-------
Section 6 Representative Applications
In this section, the vendor (with minimal editorial changes by ORNL) provides a list of representative
applications in which its technology has been used.
The SRI Model 86IOC GC/TID has been used on-
site for the analysis of soil samples contaminated
with explosives that have been obtained from
locations used for the open burning and open
detonation (OB/OD) of obsolete munitions and from
a land mine test facility. The GC/TID analysis of
soil samples taken from the OB/OD sites established
the presence of nitroaromatic, nitramine, and nitrate
ester explosive compounds. Reanalysis of the soil
sample extracts from the OB/OD sites by EPA SW-
846 Method 8095 (EPA 1999) confirmed the on-site
GC/TID concentrations established for TNB, TNT,
and RDX, and qualitatively confirmed the presence
of nitroglycerin (NG) and pentaerythritol tetranitrate
(PETN); these two compounds were not present in
the field calibration standards.
To detect explosive residues around buried land
mines the GC/TID was optimized for the analysis of
2,4-DNT, TNT, 2-Am-DNT at concentrations
between 0.005 and 0.1 mg/kg, and for 4-Am-DNT at
concentrations between 0.05 and 0.5 mg/kg. Most of
the land mines at this test facility contained TNT as
the main charge. This on-site capability made it
possible to establish whether these four explosives
were present in surface soil samples taken above the
land mines. For a couple of mines where mg/kg
levels of explosive residues were detected, an
extensive, iterative sampling protocol was
performed to delineate the surface boundaries of the
explosive-related chemical signature. The reanalysis
of soil sample extracts from this mine-field site by
Method 8095 confirmed the on-site GC/TID results
for these explosive compounds.
Additional applications using the GC/TID
instrument have been (1) to analyze explosive
vapors collected on solid phase microextraction
fibers after thermal desorption in the heated inlet,
and (2) to characterize explosive residues on range
scrap following wipe sampling or solvent
immersion.
20
-------
Section 7 References
ASTM (American Society for Testing and Materials). 1997a. Standard Practice for Generation of
Environmental Data Related to Waste Management Activities: Quality Assurance and Quality Control
Planning and Implementation, D5283-92.
ASTM (American Society for Testing and Materials). 1997b. Standard Practice for Generation of
Environmental Data Related to Waste Management Activities: Development of Data Quality Objectives,
D5792-95.
Berger, W., H. McCarty, and R-K. Smith. 1996. Environmental Laboratory Data Evaluation. Genium
Publishing Corp., Schenectady, N.Y.
Draper, N. R., and H. Smith. 1981. Applied Regression Analysis. 2nd ed. John Wiley & Sons, N.Y.
EPA (U.S. Environmental Protection Agency). 1994. "Method 8330: Nitroaromatics and Nitramines by High
Performance Liquid Chromatography (HPLC)." In Test Methods for Evaluating Solid Waste: Physical/
Chemical Methods, Update II. SW-846. U.S. Environmental Protection Agency, Washington, B.C.,
September.
EPA (U.S. Environmental Protection Agency). 1996. Guidance for Data Quality Assessment, EPA QA/G-9;
EPA/600/R-96/084, EPA, Washington, D.C., July.
EPA (U.S. Environmental Protection Agency). 1999. "Method 8095: Nitroaromatics and Nitramines by GC-
ECD." In Test Methods for Evaluating Solid Waste: Physical/Chemical Methods, Update II. SW-846. U.S.
Environmental Protection Agency, Washington, B.C..
Jenkins, T. F., M. E. Walsh, and P. G. Thorne. 1998. Site Characterization for Explosives Contamination at a
Military Firing Range Impact Area. Special Report 98-9. U.S. Army Cold Regions Research and Engineering
Laboratory, Hanover, N.H. Available at http://www.crrel.usace.army.mil/.
Maskarinec, M. P., C. K. Bayne, L. H. Johnson, S. K. Holladay, R. A. Jenkins, and B. A. Tomkins. 1991.
Stability of Explosives in Environmental Water and Soil Samples. ORNL/TM-11770. Oak Ridge National
Laboratory, Oak Ridge, Tenn., January.
ORNL (Oak Ridge National Laboratory). 1999. Technology Demonstration Plan: Evaluation of Explosives
Field Analytical Techniques. Oak Ridge National Laboratory, Oak Ridge, Tenn., August.
ORNL (Oak Ridge National Laboratory). 2000. Technology Verification Test Plan: Evaluation of Explosives
Field Analytical Techniques. Oak Ridge National Laboratory, Oak Ridge, Tenn., August.
Specialized Assays, Inc. 1999. Comprehensive Quality Assurance Plan. SAL-QC-Rec 5.0. January 6.
21
-------
Appendix A
GC/TID Sample Results Compared with
Reference Laboratory Results
Sample ,
., Sample
site or
, no.
type
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Blank
Fort Ord
Fort Ord
Fort Ord
Fort Ord
Fort Ord
Fort Ord
Fort Ord
Fort Ord
Fort Ord
Fort Ord
Fort Ord
Fort Ord
Iowa
Iowa
Iowa
Iowa
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
1
1
1
1
2
2
2
2
3
3
3
3
1
1
1
1
Sampl
replicai
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
2,4-DNT
e (mg/kg)
SRI
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
11.0
<500
18.0
10.0
RefLab
O.5
0.5
O.5
0.5
0.5
<51
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
<51
0.5
<532
<50.5
RDX
(mg/kg)
SRI
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
<50
<500
<50
<50
RefLab
O.5
0.5
O.5
0.5
0.5
<51
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.6
O.5
0.5
0.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
<51
0.5
<532
<50.5
TNT
(mg/kg)
SRI
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
1.1
0.5
0.5
O.5
0.7
0.5
O.5
0.5
6.3
0.5
0.5
O.5
0.5
O.5
0.6
0.5
3.0
0.5
0.5
O.5
0.5
O.5
21000.0
31000.0
23000.0
22000.0
RefLab
O.5
0.5
O.5
0.5
0.5
70900.0
0.5
O.5
O.5
0.5
O.5
0.5
0.9
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
0.8
0.8
O.5
0.8
2.1
0.8
0.8
0.5
O.5
0.5
O.5
20400.0
0.8
33400.0
28300.0
SRI
Analysis
Order "
1079
1076
1062
1078
1070
1108
1038
1054
1043
1052
1008
1102
1024
1018
1101
1022
1088
1046
1006
1053
1050
1073
1092
1013
1034
1031
1098
1067
1026
1084
1066
1030
1077
1003
1021
1023
22
-------
Sample . .
., Sample Sample
site or .. ,
no. rephcat
type
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Milan
Spike/PE
Spike/PE
Spike/PE
Spike/PE
1
1
1
1
2
2
2
2
3
3
3
o
J
4
4
4
4
1
1
1
1
2
2
2
2
3
3
3
o
J
4
4
4
4
5
5
5
5
1
1
1
1
1
2
3
4
1
2
o
J
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
2,4-DNT
; (mg/kg)
SRI
<0.5
<50.0
<50.0
<5.0
<5
<5
<5
<5
<50
<50
5.5
<5
14.0
8.8
8.5
7.2
<0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
<50
<50
<50
<50
<5
<50
<50
<50
0.7
0.6
0.7
0.6
O.5
0.5
O.5
0.5
RefLab
O.5
0.5
O.5
<25.0
0.5
O.5
0.5
O.5
O.5
<50
<50
0.5
80.0
11.4
11.9
9.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
<50
<200
0.5
<50
O.5
0.5
O.5
2.1
2.7
1.7
1.6
0.5
O.5
0.5
O.5
RDX
(mg/kg)
SRI
2500.0
2400.0
2300.0
2200.0
1400.0
1100.0
1300.0
1200.0
4800.0
3500.0
3400.0
4000.0
6.1
7.1
6.9
4.6
110.0
130.0
110.0
150.0
22.0
20.0
26.0
18.0
7900.0
3400.0
4100.0
3800.0
2500.0
2500.0
2100.0
2700.0
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
RefLab
3460.0
3520.0
2140.0
1900.0
1180.0
1450.0
1170.0
320.0
4300.0
3550.0
4650.0
5850.0
12.0
10.7
10.8
7.7
149.0
118.0
72.2
308.0
34.8
16.4
28.0
22.9
2350.0
1950.0
4080.0
3880.0
2740.0
2640.0
2600.0
3070.0
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
TNT
(mg/kg)
SRI
150.0
120.0
120.0
99.0
66.0
76.0
61.0
63.0
81.0
89.0
45.0
150.0
80.0
84.0
82.0
73.0
o o
j.j
9.3
0.9
1.1
0.5
O.5
0.5
O.5
260.0
82.0
300.0
110.0
110.0
68.0
45.0
80.0
7.7
7.3
7.1
6.8
87.0
87.0
92.0
92.0
RefLab
109.0
120.0
111.0
125.0
50.0
51.0
51.0
10.6
205.0
170.0
300.0
400.0
89.0
78.0
81.5
67.5
2.7
1.1
1.4
1.7
0.5
O.5
0.5
O.5
190.0
270.0
320.0
273.0
220.0
260.0
80.0
162.0
11.5
10.2
11.3
10.6
81.8
104.0
90.0
124.0
SRI
Analysis
Order "
1107
1090
1100
1025
1010
1027
1029
1012
1082
1041
1055
1037
1081
1007
1056
1087
1097
1019
1083
1039
1014
1074
1064
1072
1069
1065
1016
1033
1086
1028
1036
1005
1048
1047
1060
1059
1103
1044
1095
1094
23
-------
Sample . .
., Sample Sampl
site or ,.
no. rephcai
type
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Spike/PE
Volunteer
Volunteer
Volunteer
Volunteer
Volunteer
Volunteer
Volunteer
Volunteer
Volunteer
Volunteer
Volunteer
Volunteer
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
1
1
1
1
2
2
2
2
3
3
o
3
3
1
2
3
4
1
2
o
5
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
2,4-DNT
e (nig/kg)
SRI
<0.5
<0.5
<0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
<5
<5
0.5
<5
<5
<5
O.5
<500
59.0
60.0
67.0
30.0
29.0
35.0
34.0
0.7
0.7
0.6
0.6
RefLab
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
0.5
O.5
0.5
0.5
O.5
0.5
O.5
O.5
<25
O.5
0.5
<50
<25
19.0
<250
<53.2
<538
<5.4
45.2
2.0
3.0
2.2
2.2
RDX
(mg/kg)
SRI
84.0
76.0
88.0
74.0
49.0
56.0
47.0
46.0
7.6
9.7
8.9
8.7
440.0
490.0
490.0
450.0
210.0
220.0
230.0
270.0
<500
<50
<50
<50
<50
<50
<50
<50
0.5
O.5
0.5
O.5
RefLab
111.0
90.5
98.0
127.0
49.5
45.0
63.5
51.0
9.1
8.4
8.6
9.1
460.0
455.0
705.0
445.0
260.0
255.0
335.0
250.0
<50
<25
<5
<250
<53.2
<538
6.5
<5.4
0.5
O.5
0.5
O.5
TNT
(mg/kg)
SRI
O.5
0.5
2.4
0.5
11.0
10.0
9.9
10.0
45.0
50.0
45.0
48.0
260.0
240.0
260.0
260.0
480.0
480.0
480.0
500.0
190000.0
110000.0
110000.0
94000.0
9000.0
6200.0
8300.0
9400.0
9.9
8.0
7.7
7.7
RefLab
O.5
0.5
O.5
0.5
8.4
7.6
10.0
8.5
47.5
48.5
48.5
47.0
230.0
205.0
435.0
205.0
535.0
505.0
675.0
510.0
108000.0
75500.0
117000.0
61000.0
11300.0
12600.0
26200.0
8920.0
12.0
10.3
13.8
10.4
SRI
Analysis
Order "
1105
1057
1020
1063
1049
1001
1058
1061
1104
1096
1071
1106
1068
1004
1075
1045
1099
1042
1093
1017
1002
1091
1089
1015
1085
1051
1011
1009
1035
1032
1040
1080
"Indicates order of analysis by SRI; for example, 1001 was analyzed first, then 1002, etc.
24
-------
Appendix B
Data Quality Objective (DQO) Example
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 it cannot attempt to provide a thorough education in this topic
(EPA 1996, ASTM 1997a, ASTM 1997b). Please refer to other educational or technical resources for further
details. Additionally, because the focus of this report is on the analytical technology, this example makes
simplifying assumptions (such as the sample is homogeneous and the reference laboratory results represent
the true concentration) in the example that may not be valid in the real world.
Background and Problem Statement
An Army Ammunition Plant that produced TNT was recently decommissioned. Past practices had resulted in
contamination of four areas around the plant. Soils at each site were mixtures of clay, silt, and organic matter
with initial concentrations of about 1500 mg/kg of TNT. Forty cubic yards of TNT-contaminated soil were
loaded into a bioreactor. After three months of processing, the soil mixture was dewatered and put into
drums. The simplifying assumption was made that the soil in each drum was homogeneous based on process
knowledge. In agreement with regulators, the treatment goal established for the site was to reduce the soil
concentration to <15 mg/kg of TNT. Soil with <15 mg/kg of TNT would be returned to the four areas around
the plant. Those drums containing soil with TNT concentrations > 15 mg/kg would be stored for additional
processing (i.e., incineration).
The company's DQO team considered using SRI Instruments' Model 8610C gas chromatograph (GC) to
measure the TNT concentration in each drum, based on the data generated in the ETV verification study. The
SRI Model 86 IOC GC is an on-site gas chromatograph equipped with a thermionic ionization detector that
allows for the determination of explosives in a soil matrix. The plan was to randomly select soil samples
from each drum and determine the TNT concentration with the SRI Model 86IOC GC. The DQO team
decided that drums will be disposed by incineration if the TNT concentration is > 15 mg/kg ("hot"). Those
drums with TNT concentrations <15 mg/kg will be put into a landfill because incineration of TNT-
contaminated 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 TNT in each drum.
General Decision Rule
If the average TNT concentration is less than the action level, then the TNT drums
are sent to the landfill.
If the average TNT concentration is greater than or equal to the action level, then the
TNT drum is sent to the incinerator.
DQO Goals
The DQO team's primary goal was to calculate how many samples would need to be analyzed by the SRI
Model 86IOC GC in order to confidently make a decision about remediating the processed soil, given the
uncertainties of the technology's results. Because the team decided that inadvertently sending soil to a
landfill that exceeded 15 mg/kg of TNT was the worst possible mistake, the number of samples measured is
primarily related to this false-rejection decision error rate. A secondary decision error would be to
25
-------
unnecessarily incinerate soil with TNT concentrations <15 mg/kg. This decision would be a false-acceptance
decision error. Consideration of both the false-rejection decision error and the false-acceptance decision error
was used to determine the final sampling plan.
EPA required that a sufficient number of samples be measured from each drum so that the false-rejection
error rate (FR) for the decision rule was 0.05 or less if the true soil concentration in a drum was 15 mg/kg or
greater. This DQO goal represents a 5% chance of returning a drum with soil containing 15 mg/kg or more of
TNT to a landfill.
The DQO team did not want to incinerate an excessive number of drums if a drum's TNT soil concentration
was <5 mg/kg because of the expense. Therefore, the DQO team recommended that the false-acceptance
error rate (FA) for the decision rule be 0.10 if the true drum concentration was less than 5 mg/kg. That is,
there would be a 10% chance of incinerating a drum if the true TNT concentration for a drum was <5 mg/kg.
Permissible FR and FA Error Rates and Critical Decision Points
FR: Pr[take drum to landfill] < 0.05 when true TNT concentration =15 mg/kg
FA: Pr[take drum to incinerator] < 0.10 when true TNT concentration = 5 mg/kg
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 SRI Model 86 IOC GC, 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 estimate of the true average
concentration of TNT in the drum to the specified certainty? Recall that the simplifying assumption was
made that the TNT distribution throughout the soil within a single drum is homogeneous; thus, matrix
heterogeneity will not contribute to overall variability. The only variability, then, to be considered in this
example is the variability in the SRI Model 86IOC GC's analytical method, which is determined by precision
studies.
2. What is the appropriate action level (AL) for using the SRI Model 8610C GC 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" TNT concentration of the drum. When using the SRI Model
86IOC GC, 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.
Determining the Number of Samples
With the critical decision points selected, the DQO team could then determine the number of samples needed
from each drum to calculate the drum's "true" average TNT concentration. For a homogeneous matrix, the
number of samples required depends on the precision of the analytical method.
Figure B-l shows that the standard deviations for the SRI Model 86IOC GC versus the average TNT
concentration for the reference laboratory. Although the fitted line increases with average TNT
26
-------
concentration, the linear model is not significantly
different than a constant standard deviation over the
0 to 50 mg/kg concentration range. Therefore, the
precision of the SRI Model 86 IOC GC can be
represented by a pooled standard deviation of
3.1 mg/kg within the concentration range of 0 to
60 mg/kg (see Figure B-l). Note that the pooled
standard deviation is calculated by first calculating
the average variance then taking the square root.
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 as determined in the DQO Goals. The
following formula is provided in EPA's Guidance
for Data Quality Assessment (EPA 1996) that can
be adapted to this example for calculating the
number of samples required to meet the FR and FA
requirements:
10
20
30
50
60
Average TNT (rrgfcg) - Reference Laboratory
Figure B-l. Linear model for the standard deviation for
SRI Model 86IOC GC versus average TNT
(mg/kg) result for the reference laboratory
with 95% confidence intervals (dashed
lines).
_
"
\-FR
l-FA
(Eq B-l)
where
n
S2
RT
CFA
FR
FA
number of samples from a drum to be measured,
variance for the measurement [e.g., S2 = (3.1)2 ],
regulatory threshold (e.g., RT= 15 mg/kg),
concentration at which the FA is specified (e.g., CFA = 5 mg/kg),
false-rejection decision error rate (e.g., FR = 0.05),
false-acceptance decision error rate (e.g., FA = 0.10),
(l-p)thpercentile of the standard normal distribution (see EPA 1996, Table A-l of Appendix A).
Example: Z(l-pR) = Z095 = 1.645 and Z(l_FA) = Z090 = 1.282.
__ (3-D2 d.645 + 1.282)3
(15-5)2
(0.5)(1.645)2 =
2 =
Therefore, three soil samples from each drum would be analyzed by SRI Model 86IOC GC to meet the
criteria established by the DQO process. To be conservative, the number of samples was rounded up to the
next integer. The TNT results from the three samples are averaged (by taking the arithmetic mean) to produce
an SRI Model 86IOC GC value for a drum's TNT concentration.
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 can be calculated. The action level is the decision criterion (or "cut-off value)
that will be compared to the unbiased average TNT concentration determined for each drum. The AL for the
27
-------
decision rule is calculated based on controlling the FR established in the DQO process. Recall that the DQO
team set the permissible FR error rate at 5%.
The formula (EPA 1996) to compute the action level is
AL = RT - ;
_S_
&
(Eq. B-2)
AL = \5ppm - (1.645) x M. = 12.1 mg/kg
To summarize, three random samples from each
drum are analyzed.. The three results are averaged to
produce the average TNT concentration for the
drum, which is then compared to the action level
(i.e., AL = 12.1 mg/kg) for the decision rule.
Therefore, the decision rule using the SRI Model
8610C GC to satisfy a 5% FR and a 10% FA is as
follows:
0.0
0 5 10 15 20
True TNT Concentration (mg/kg)
Figure B-2. Decision performance curve to TNT drum
example.
Decision Rule for FR = 5% and FA = 10%
If the average TNT concentration of three random soil samples on a drum is < 12.1 mg/kg, then send the
drum to the landfill.
If the average TNT concentration of three random soil samples on a drum is > 12.1 mg/kg, then send the
drum to the incinerator.
The decision performance curve (for more information, see EPA 1996, pp. 34-36) calculates the probability
of sending a drum to the incinerator for different values of true TNT soil concentration in a drum. Figure B-2
shows that the decision performance curve has the value of Pr[take drum to incinerator] = 0.95 for true = 15
mg/kg. This indicates that the decision rule meets the DQO team's FR of 5%. The actual Pr[take drum to
incinerator] = 0.00004 for a true TNT concentration = 5 mg/kg, which is better than the FA 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.
Alternative FR Parameter
Because of random sampling and analysis error, some chance always exists 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% FR) 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 15 mg/kg. If the DQO team wanted to be 99% certain that a drum was correctly sent to a landfill, the
28
-------
following describes how changing the FR requirement from 5% to 1% would affect the decision rule. Using
FR = 0.01, the sample size is calculated to be four and the action level is calculated as 11.4 mg/kg. The
decision performance curve has the value of Pr[take drum to incinerator] = 0.99 for true =15 mg/kg. This
indicates that the decision rule meets the DQO team's FR of 1%. The Pr[take drum to incinerator] = 0.00002
for true = 5 mg/kg is better than the FA of 10% that the DQO team had specified. This improved performance
is due to rounding up the number of samples to the next integer in the calculation of number of samples
required. The decision rule for the lower FR would be as follows:
Decision Rule for FR = 1% and FA = 10%
If the average TNT concentration of four random soil samples on a drum is <11.4 mg/kg, then send the
drum to the landfill.
If the average TNT concentration of four random soil samples on a drum is > 11.4 mg/kg, then send the
drum to the incinerator.
Comparison to Sending the Samples Off-Site for Analysis
The DQO team wanted to compare the sampling plan using the SRI Model 86IOC GC field measurements
with a sampling plan using an off-site reference laboratory. For the off-site reference laboratory, the DQO
team assumed a precision of SD = 0.9 mg/kg based on the ETV reference laboratory values. They also
specified that the FA percentage would be at a TNT concentration of 10 mg/kg because the reference
laboratory measurements are more precise. This specification means that there is only a 10% chance of
sending a drum to the incinerator if the true TNT concentration in the drum is 10 mg/kg. A formula provided
in EPA's Guidance for Data Quality Assessment (EPA 1996) shows that the number would be n = 2 for the
5% FR percentage and 10% FA percentage. The decision rule would be to take two randomly selected
samples and send them to the reference laboratory for analysis. If the average TNT concentration is less than
an action level of 14.0 mg/kg, restore the soil to the plant site; otherwise, send the soil drum to storage. In
this example, the off-site laboratory appears to require fewer samples to be analyzed (two for the off-site
laboratory versus three for the SRI Model 86IOC GC). Other factors (such as a cost-benefit analysis, ability
to detect other analytes) would need to be considered before choosing a method for this application.
29
------- |