United States
Environmental Protection
Agency
Office of
Research and Development
Washington, DC 20460
EPA/600/R-04/028
May 2004
&EPA
Innovative Technology
Verification Report
Field Measurement Technology for
Mercury in Soil and Sediment
MTI Inc.'sPDVGOOO
Anodic Stripping Voltammetry
-------
EPA/600/R-04/028
May 2004
Innovative Technology
Verification Report
MTI Inc.'sPDVGOOO
Anodic Stripping Voltammetry
Prepared by
Science Applications International Corporation
Idaho Falls, ID
Contract No. 68-C-00-179
Dr. Stephen Billets
Characterization and Monitoring Branch
Environmental Sciences Division
Las Vegas, Nevada 89193-3478
National Exposure Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
-------
Notice
The U.S. Environmental Protection Agency through its Office of Research and Development funded
and managed the research described here under contract to Science Applications International
Corporation. It has been subjected to the Agency's peer and administrative review and has been
approved for publication as an EPA document. Mention of trade names or commercial products does
not constitute endorsement or recommendation for use.
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
MEASUREMENT AND MONITORING TECHNOLOGY PROGRAM
VERIFICATION STATEMENT
TECHNOLOGY TYPE: Field Measurement Device
APPLICATION: Measurement for Mercury
TECHNOLOGY NAME: MTI Inc.'s Portable Digital Voltammeter (PDV) 6000
COMPANY: Monitoring Technologies International Pty. Ltd.
ADDRESS: 1/7 Collingwood Street
Osborne Park, Perth
West Australia 6017
WEB SITE: www.monitoring-technologies.com.
U.S. CONTACT/TELEPHONE: Felecia Owen, MTI Inc.
1609 Ebb Drive
Wilmington, NC 28409
(910)392-5714
VERIFICATION PROGRAM DESCRIPTION
The U.S. Environmental Protection Agency (EPA) created the Superfund Innovative Technology Evaluation (SITE) and
Measurement and Monitoring Technology (MMT) Programs to facilitate deployment of innovative technologies through
performance verification and information dissemination. The goal of these programs is to further environmental
protection by substantially accelerating the acceptance and use of improved and cost-effective technologies. These
programs assist and inform those involved in design, distribution, permitting, and purchase of environmental
technologies. This document summarizes results of a demonstration of the Portable Digital Voltammeter (PDV) 6000
developed by Monitoring Technologies International Pty. Ltd. (MTI).
PROGRAM OPERATION
Under the SITE and MMT Programs, with the full participation of the technology developers, the EPA evaluates and
documents the performance of innovative technologies by developing demonstration plans, conducting field tests,
collecting and analyzing demonstration data, and preparing reports. The technologies are evaluated under rigorous
quality assurance (QA) protocols to produce well-documented data of known quality. The EPA National Exposure
Research Laboratory, which demonstrates field sampling, monitoring, and measurement technologies, selected Science
Applications International Corporation as the verification organization to assist in field testing five field measurement
devices for mercury in soil and sediment. This demonstration was funded by the SITE Program.
DEMONSTRATION DESCRIPTION
In May 2003, the EPA conducted a field demonstration of the PDV 6000 and four other field measurement devices for
mercury in soil and sediment. Due to inaccurate results on standards (later determined to be an oilcontaminanton the
disposable beakers that was coating the electrodes), MTI decided to discontinue the field measurements. The
demonstration of their PDV 6 000 instrument was rescheduled and was con ducted in Las Vegas, NV, in June 2003. This
verification statement focuses on the PDV 6000; a similar statement has been prepared for each of the other four
devices. The performance of the PDV 6000 was compared to that of an off-site laboratory using the reference method,
-------
"Test Methods for Evaluating Solid Waste" (SW-846) Method 7471 B (modified). To verify a wide range of performance
attributes, the demonstration had both primary and secondary objectives. The primary objectives were:
(1) Determining the instrument sensitivity with respect to the Method Detection Limit (MDL) and Practical
Quantitation Limit (PQL);
(2) Determining the analytical accuracy associated with the field measurement technologies;
(3) Evaluating the precision of the field measurement technologies;
(4) Measuring the amount of time required for mobilization and setup, initial calibration, daily calibration, sample
analysis, and demobilization; and
(5) Estimating the costs associated with mercury measurements for the following four categories: capital, labor,
supplies, and investigation-derived waste (IDW).
Secondary objectives for the demonstration included:
(1) Documenting the ease of use, as well as skills and training required to properly operate the device;
(2) Documenting potential health and safety concerns associated with operating the device;
(3) Documenting the portability of the device;
(4) Evaluating the device durability based on its materials of construction and engineering design; and
(5) Documenting the availability of the device and associated spare parts.
The PDV 6000 analyzed 52 field soil samples, 33 field sediment samples, 35 spiked field samples, and 77 performance
evaluation (PE) standard reference material (SRM) samples in the demonstration. The field samples were collected
in four areas contaminated with mercury, the spiked samples were from these same locations, and the PE samples
were obtained from a commercial provider.
Collectively, the field and PE samples provided the different matrix types and the different concentrations of mercury
needed to perform a comprehensive evaluation of the PDV 6000. A complete description of the demonstration and a
summary of the results are available in the Innovative Technology Verification Report: "Field Measurement Technology
for Mercury in Soil and Sediment — MTI Inc.'s PDV 6000 Anodic Stripping Voltammetry" (EPA/600/R-04/028).
TECHNOLOGY DESCRIPTION
The principle of analysis used by the MTI PDV 6000 is Anodic Stripping Voltammetry (ASV). ASVis a simple procedure
in which a reducing potential is applied to a "working electrode." When the potential of this working electrode exceeds
the ionization potential of the particular metal ion solution in solution, it is reduced to the metal. The metal plates onto
the working electrode surface as follows:
Mn++ ne'^ M
where: Mn+ = analyte metal ion in solution
ne" = number of electrons
M = metal plated onto the electrode
The longer the potential is applied, the more metal is reduced and plated onto the surface of the electrode. This
"deposition" or "accumulation" step concentrates the metal. After sufficient metal has been plated onto the working
electrode, the metal is stripped (oxidized) off the electrode by increasing the potential to that electrode at a constant rate.
For a given electrolyte solution and electrode, each metal has a specific potential in which the following oxidation
reaction will occur:
M -> Mn++ ne'
The electrons released by this process form a current that is measured and plotted as a function of the applied potential
to give a "voltammogram." The current at the oxidation or stripping potential for the analyte metal is seen as a peak.
To calculate the sample concentration, the peak height or area is measured and compared to that of a known standard
or solution under the same conditions.
During the demonstration, each sample was digested per MTI's standard operating procedure (SOP). Sample
preparation consisted of weighing out (in order) 2 grams of sample material, placing that material in a 70-mL digestion
IV
-------
bottle and then pipetting into that bottle 4.0 ml_ of HNO3, 4.0 ml_ of H2O2, 12 ml_ of deionized (Dl) water, and 20 ml_ of
electrolyte solution. The prepared samples were then analyzed with the PDV 6000.
ACTION LIMITS
Action limits and concentrations of interest vary and are project specific. There are, however, action limits which can
be considered as potential reference points. The EPA Region IX Preliminary Remedial Goals (PRGs) for mercury are
23 mg/kg in residential soil and 310 mg/kg in industrial soil.
VERIFICATION OF PERFORMANCE
To ensure data usability, data quality indicators for accuracy, precision, representativeness, completeness,
comparability, and sensitivity were assessed for the reference method based on project-specific QA objectives. Key
demonstration findings are summarized below for the primary objectives.
Sensitivity. The two primary sensitivity evaluations performed for this demonstration were the MDL and PQL. Both
will vary dependent upon whether the matrix is a soil, waste, or aqueous solution. Only soils/sediments were tested
during this demonstration, and therefore, MDL calculations and PQL determinations for this evaluation are limited to
those matrices. By definition, values measured below the PQL should not be considered accurate or precise and those
below the MDL are not distinguishable from background noise.
Method Detection Limit - The evaluation of an MDL requires seven different measurements of a low concentration
standard or sample following the procedures established in the 40 Code of Federal Regulations (CFR) Part 1 36. The
MDL is between 1 .67 and 3.67 mg/kg. The equivalent MDL for the referee laboratory is 0.0026 mg/kg. Examples from
analyzed samples, however, suggested that the MTI MDL may be closerto 0.81 1 mg/kg or lower. Values detected at
these lowerlevels would likely be highly inaccurate and should only be considered as a "positive hit" withoutany implied
accuracy or precision.
Practical Quantitation Lim it - The low standard calculations suggest that a PQL for the MTI field instrument is 4-8 mg/kg.
The percent difference (%D) for the average MTI result for a sample concentration of 4.75 mg/kg, tested as part of the
demonstration, is 46%. The referee laboratory PQL confirmed during the demonstration is 0.005 mg/kg with a %D
Accuracy: The results from the PDV 6000 were compared to the 95% prediction interval for the SRM materials and
to the referee laboratory results (Method 7471 B). MTI data were within the SRM 95% prediction intervals 53% of the
time. The comparison between the MTI field data and the referee laboratory results suggestthatthe two data sets are
not different but the similarityfor individual samples is often the result of high variability associated with the MTI reported
values. The number of MTI average values greater than 50% different from the referee laboratory results or SRM
reference values was only 6 for 21 different sample lots with only 2 of those 6 greater than 1 00% different. These 6 (of
the 21 ) sample lots had results greater than laboratory results, indicating a positive bias. MTI results therefore appear
to provide a rough estimate of mercury concentration for fie Id determination and may be affected by interferences not
identified by this demonstration. It should be concluded, however, that the MTI PDV 6000 did not compare well to
laboratory Method 7471 B in terms of obtaining accurate analyses of mercury in soil.
Precision: The precision of the MTI field instrument is not as good as the referee laboratory precision. The overall
average RSD for MTI is 35.1% which is above the 25% RSD objective set for the laboratory. The overall laboratory
average RSD is 22.3%.
Measurement Time: From the time of sample receipt, MTI required 38 hours to prepare a draft data package
containing mercury results for 1 97 samples. Two persons performed all setup, calibration checks, sample preparation
and analysis, and equipment demobilization. Each individual analysis, on average, took 7.5 minutes (from the time the
sample was digested until results were displayed), butthe total time per analysis averaged approximately 1 1 .6 minutes
when all sample preparation, sample analysis, and data package preparation were included in the calculation.
-------
Measurement Costs: The cost per analysis, based on measurement of 197 samples, when incurring a minimum
1-month rental fee for the PDV 6000, was determined to be $43.74 per sample. Excluding the instrument rental cost,
the cost for analyzing the 197 samples was determined to be $32.57 per sample. Based on the 3-day field
demonstration, the total cost for equipment rental and necessary supplies was estimated at $8,600. The cost breakout
by category is: capital costs, 25.5%; supplies, 32.5%; support equipment, 3.2%; labor, 17.4%; and IDW, 21.4%.
Key demonstration findings are summarized below for the secondary objectives.
Ease of Use: Based upon observations made during the demonstration, the PDV 6000 is relatively easy to operate,
requiring one fie Id technician with a basic knowledge of chemistry acquired on the job or in a university, basic computer
skills, and training on the PDV 6000. A 1-day training course on instrument operation is offered at additional cost; this
training should be considered for most potential users having no previous laboratory experience.
Potential Health and Safety Concerns: No significant health and safety concerns were noted during the
demonstration; however, during the demonstration, fumes following the addition of acidic solution, during digestion of
certain samples, was observed. The digestion procedure involves the use of potentially reactive oraggressive reagents
and ultimately there is a potential for vigorous reactions to occur.
Portability. The main components of the PDV 6000 consist of a handheld control unitand a cell assembly. The control
unit, cell assembly, and other accessories are easily portable due to their size, weight, and method by which they are
packed and transported. The handheld control unit, cell assembly, and other pertinent accessories are transported in
a hard pelican style case (approximately 36 cm by 25 cm by 15 cm). The portable control unit weighs approximately
700 grams and measures 10 cm by 18 cm by 4 cm. The unit was easy to setup, and it can be taken anywhere by foot.
It operates on 110 or 220 volt AC current and also on a rechargeable NiMH battery power supply.
Durability: The PDV 6000 was well designed and constructed of durable plastic. According to MTI, a conformal
coating on the instrument's electronics allows for unlimited humidity range; however, the laptop used to run the
voltammetric analysis system (VAS) software should be kept dry. The cell assembly is also constructed of durable
plastic, and the cell assembly stand is constructed of metal.
Availability of the Device: Per MTI, a stock of instrumentation will be available in the U.S. through the new MTI U.S.
operations. Instruments will be available for purchase, rent, or lease and can be delivered within one week of order
placement. Spare parts and consumable supplies can be added to the original PDV 6000 order. Supplies not typically
provided by MTI (pipetters and a scale) are readily available from laboratory supply firms.
PERFORMANCE SUMMARY
In summary, during the demonstration, the PDV 6000 exhibited the following desirable characteristics of afield mercury
measurement device: (1) field acceptable accuracy, (2) good precision, (3) good sensitivity compared to the PRGs, (4)
high sample throughput, (5) measurement costs comparable to laboratory analytical costs, (6) exceptional portability,
and (7) relative ease of use. During the demonstration the PDV 6000 was found to have the following limitations: (1)
sample digestion requiring nitric acid and hydrogen peroxide and (2) generation of a secondary waste stream from
sample digestion.
NOTICE: EPA verifications are based on an evaluation of technology performance under specific, predetermined criteria and appropriate
quality assurance procedures. The EPA makes no expressed or implied warranties as to the performance of the technology and does 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.
VI
-------
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the nation's natural resources.
Under the mandate of national environmental laws, the Agency strives to formulate and implement actions leading to a
compatible balance between human activities and the ability of natural systems to support and nurture life. To meet this
mandate, the EPA' s Office of Research and Development provides data and scientific support that can be used to solve
environmental problems, build the scientific knowledge base needed to manage ecological resources wisely, understand
how pollutants affect public health, and prevent or reduce environmental risks.
The National Exposure Research Laboratory is the Agency's center for investigation of technical and management
approaches for identifying and quantifying risks to human health and the environment. Goals of the Laboratory's research
program are to (1) develop and evaluate methods and technologies for characterizing and monitoring air, soil, and water;
(2) support regulatory and policy decisions; and (3) provide the scientific support needed to ensure effective
implementation of environmental regulations and strategies.
The EPA's Superfund Innovative Technology Evaluation (SITE) Program evaluates technologies designed for
characterization and remediation of contaminated Superfund and Resource Conservation and Recovery Act (RCRA) sites.
The SITE Program was created to provide reliable cost and performance data in order to speed acceptance and use of
innovative remediation, characterization, and monitoring technologies by the regulatory and user community.
Effective monitoring and measurement technologies are needed to assess the degree of contamination at a site, provide
data that can be used to determine the risk to public health or the environment, and monitor the success or failure of a
remediation process. One component of the EPA SITE Program, the Monitoring and Measurement Technology (MMT)
Program, demonstrates and evaluates innovative technologies to meet these needs.
Candidate technologies can originate within the federal government or the private sector. Through the SITE Program,
developers are given the opportunity to conduct a rigorous demonstration of their technologies under actual field
conditions. By completing the demonstration and distributing the results, the Agency establishes a baseline for acceptance
and use of these technologies. The MMT Program is managed by the Office of Research and Development's
Environmental Sciences Division in Las Vegas, NV.
Gary Foley, Ph. D.
Director
National Exposure Research Laboratory
Office of Research and Development
VII
-------
Abstract
Mil's PDV 6000 was demonstrated under the U.S. Environmental Protection Agency Superfund Innovative Technology
Evaluation Program in June 2003 in Las Vegas, NV. The purpose of the demonstration was to collect reliable performance
and cost data for the PDV 6000. Four other field measurement devices for mercury in soil and sediment were evaluated
in May 2003 at the Oak Ridge National Laboratory in Oak Ridge, TN. The key objectives of the demonstration were:
1) determine sensitivity of each instrument with respect to a vendor-generated method detection limit (MDL) and practical
quantitation limit (POL); 2) determine analytical accuracy associated with vendor field measurements; 3) evaluate the
precision of vendor field measurements using field samples and standard reference materials (SRMs); 4) measure time
required to perform mercury measurements; and 5) estimate costs associated with mercury measurements for capital,
labor, supplies, and investigation-derived wastes.
The demonstration involved analysis of SRMs, field samples collected from four sites, and spiked field samples for
mercury. The performance results fora given field measurement device were compared to those of an off-site laboratory
using reference method, "Test Methods for Evaluating Solid Waste" (SW-846) Method 7471 B.
The sensitivity, accuracy, and precision measurements were successfully completed. Results of these measurement
evaluations suggest that the MTI field instrument does not perform as well as the laboratory analytical method but does
provide a rough estimate of mercury concentrations in soils and sediments often suitable for field analysis. During the
demonstration, MTI required 38 hours for analysis of 197 samples. The cost peranalysis, based on measurement of 197
samples, when incurring a minimum 1-month rental fee for the PDV 6000, was determined to be $43.74 per sample.
Excluding the instrument rental cost, the cost for analyzing the 197 samples was determined to be $32.57 per sample.
Based on the 3-day field demonstration, the total cost for equipment rental and necessary supplies was estimated at
$8,600.
The PDV 6000 exhibited good ease of use and durability, as well as no major health and safety concerns. However, there
is the potential for gas-producing reactions to occur during the digestion procedure used to prepare the samples. MTI sells
kits, containing extraction reagents and disposable supplies, for the analyses of samples. When conducting a large
number of analyses, purchase of bulk reagents and disposable supplies should be considered to reduce costs.
VIM
-------
Contents
Notice ii
Verification Statement iii
Foreword vii
Abstract viii
Contents ix
Tables xii
Figures xiii
Abbreviations, Acronyms, and Symbols xiv
Acknowledgments xvi
Chapter Page
1 Introduction 1
1.1 Description ofthe SITE Program 1
1.2 Scope of the Demonstration 2
1.2.1 Phase I 2
1.2.2 Phase II 2
1.3 Mercury Chemistry and Analysis 3
1.3.1 Mercury Chemistry 3
1.3.2 Mercury Analysis 4
2 Technology Description 6
2.1 Description of Anodic Stripping Voltammetry 6
2.2 Description ofthe MTI PDV 6000 7
2.3 Developer Contact Information 8
3 Field Sample Collection Locations and the Demonstration Site 9
3.1 Carson River 10
3.1.1 Site Description 10
3.1.2 Sample Collection 10
3.2 Y-12 National Security Complex 11
3.2.1 Site Description 11
3.2.2 Sample Collection 11
3.3 Confidential Manufacturing Site 11
3.3.1 Site Description 11
IX
-------
Contents (Continued)
Chapter
3.3.2 Sample Collection 12
3.4 Puget Sound 12
3.4.1 Site Description 12
3.4.2 Sample Collection 12
3.5 Demonstration Site 13
3.6 SAIC GeoMechanics Laboratory 14
Demonstration Approach 15
4.1 Demonstration Objectives 15
4.2 Demonstration Design 16
4.2.1 Approach for Addressing Primary Objectives 16
4.2.2 Approach for Addressing Secondary Objectives 20
4.3 Sample Preparation and Management 21
4.3.1 Sample Preparation 21
4.3.2 Sample Management 24
4.4 Reference Method Confirmatory Process 25
4.4.1 Reference Method Selection 25
4.4.2 Referee Laboratory Selection 25
4.4.3 Summary of Analytical Methods 27
4.5 Deviations from the Demonstration Plan 28
Assessment of Laboratory Quality Control Measurements 30
5.1 Laboratory QA Summary 30
5.2 Data Quality Indicators for Mercury Analysis 30
5.3 Conclusions and Data Quality Limitations 31
5.4 Audit Findings 33
Performance of the PDV 6000 34
6.1 Primary Objectives 35
6.1.1 Sensitivity 35
6.1.2 Accuracy 37
6.1.3 Precision 44
6.1.4 Time Required for Mercury Measurement 48
6.1.5 Cost 50
6.2 Secondary Objectives 50
6.2.1 Ease of Use 50
6.2.2 Health and Safety Concerns 53
6.2.3 Portability of the Device 53
6.2.4 Instrument Durability 54
6.2.5 Availability of Vendor Instruments and Supplies 54
Economic Analysis 56
7.1 Issues and Assumptions 56
7.1.1 Capital Equipment Cost 56
7.1.2 Cost of Supplies 57
-------
Contents (Continued)
Chapter Page
7.1.3 Support Equipment Cost 57
7.1.4 Labor Cost 57
7.1.5 Investigation-Derived Waste Disposal Cost 58
7.1.6 Costs Not Included 59
7.2 PDV 6000 Costs 59
7.2.1 Capital Equipment 59
7.2.2 Cost of Supplies 60
7.2.3 Support Equipment 61
7.2.4 Labor Cost 62
7.2.5 Investigation-Derived Waste Disposal Cost 62
7.2.6 Summary of PDV 6000 Costs 62
7.3 Typical Reference Method Costs 64
8 Summary of Demonstration Results 65
8.1 Primary Objectives 65
8.2 Secondary Objectives 66
9 Bibliography 69
Appendix A - MTI Comments 70
Appendix B - Statistical Analysis 75
XI
-------
Tables
Table Page
1-1 Physical and Chemical Properties of Mercury 4
1-2 Methods for Mercury Analysis in Solids or Aqueous Soil Extracts 5
2-1 Power Supply Options for the PDV 6000 8
3-1 Summary of Site Characteristics 10
4-1 Demonstration Objectives 15
4-2 Summary of Secondary Objective Observations Recorded During the Demonstration 20
4-3 Field Samples Collected from the Four Sites 22
4-4 Analytical Methods for Non-Critical Parameters 28
5-1 MS/MSD Summary 31
5-2 LCS Summary 31
5-3 Precision Sum mary 32
5-4 Low Check Standards 32
6-1 Distribution of Samples Prepared for MTI and the Referee Laboratory 34
6-2 MTI SRM Comparison 38
6-3 ALSI SRM Comparison 38
6-4 Accuracy Evaluation by Hypothesis Testing 39
6-5 Number of Sample Lots Within Each %D Range 42
6-6 Concentration of Non-Target Analytes 42
6-7 Evaluation of Precision 45
6-8 Time Measurements for MTI 49
7-1 Capital Cost Summary for the PDV 6000 60
7-2 Supplies Cost Summary, Using Preparation Kits 61
7-3 Supplies Cost Summary, Preparing Reagents 61
7-4 Labor Costs 62
7-5 IDW Costs 62
7-6 Summary of Purchase Costs for the PDV 6000 63
7-7 PDV 6000 Costs by Category 64
8-1 Distribution of Samples Prepared for MTI and the Referee Laboratory 66
8-2 Summary of PDV 6000 Results for the Primary Objectives 67
8-3 Summary of PDV 6000 Results for the Secondary Objectives 68
A-1 Calibration Standard and Analysis Information 71
B-1 Unified Hypothesis Test Summary Information 77
XII
-------
Figures
2-1 PDV 6000 cell assembly 7
2-2 Photograph of the PDV 6000 during the field demonstration 7
3-1 Tent and field conditions during the demonstration at Oak Ridge, TN 13
3-2 Demonstration site and Building 5507 13
4-1 Test sample preparation at the SAIC GeoMechanics Laboratory 23
6-1 Data plot for low concentration sample results 43
6-2 Data plot for medium concentration sample results 43
6-3 Data plot for high concentration sample results 44
6-4 PDV 6000 sample run log with open data windows 51
6-5 PDV 6000 sample graph screen 52
6-6 PDV 6000 mercury graph for sample analyzed by method of standard additions 52
7-1 Capital equipment costs 60
XIII
-------
Abbreviations, Acronyms, and Symbols
% Percent
%D Percent difference
°C Degrees Celsius
|jg/kg Microgram per kilogram
g/L Gram per liter
AC Alternating current
AAS Atomic absorption spectroscopy
ALSI Analytical Laboratory Services, Inc.
ASV Anodic Stripping Voltammetry
bgs Below ground surface
cm Centimeter
CFR Code of Federal Regulations
Cl Confidence Interval
COC Chain of Custody
DC Direct current
Dl Deionized (water)
DOE Department of Energy
EPA United States Environmental Protection Agency
g Gram
H&S Health and Safety
Hg Mercury
HgCI2 Mercury (II) chloride
IDL Instrument detection limit
IDW Investigation derived waste
ITVR Innovative Technology Verification Report
kg Kilogram
L Liter
LCS Laboratory control sample
LEFPC Lower East Fork Poplar Creek
m Meter
MDL Method detection limit
mg Milligram
mg/kg Milligram per kilogram
mL Milliliter
mm Millimeter
MS/MSD Matrix spike/matrix spike duplicate
MMT Monitoring and Measurement Technology
MTI Monitoring Technologies International
NERL National Exposure Research Laboratory
XIV
-------
Abbreviations, Acronyms, and Symbols (Continued)
NiMH Nickel metal halide
nm Nanometer
ORD Office of Research and Development
ORR Oak Ridge Reservation
ORNL Oak Ridge National Laboratory
OSWER Office of Solid Waste and Emergency Response
PDV Portable Digital Voltammeter
PPE Personal protective equipment
ppb Parts per billion
ppm Parts per million
ppt Parts per trillion
POL Practical quantitation limit
QA Quality assurance
QAPP Quality Assurance Project Plan
QC Quality control
RPD Relative percent difference
RSD Relative standard deviation
SAIC Science Applications International Corporation
SITE Superfund Innovative Technology Evaluation
SOP Standard operating procedure
SRM Standard reference material
SW-846 Test Methods for Evaluating Solid Waste; Physical/Chemical Methods
TOC Total organic carbon
TOM Task Order Manager
UL Underwriters Laboratory
UEFPC Upper East Fork of Poplar Creek
Y-12 Y-12 Oak Ridge Security Complex, Oak Ridge, TN
VAS Voltammetric analysis system
V Volt
xv
-------
Acknowledgments
The U.S. Environmental Protection Agency (EPA) Superfund Innovative Technology Evaluation wishes to acknowledge
the support of the following individuals in performing the demonstration and preparing this document: Elizabeth Phillips
of the U.S. Department of EnergyOak Ridge National Laboratory (ORNL); Stephen Chi Ids, Thomas Early, Roger Jenkins,
and Monty Ross of the UT-Battelle ORNL; Dale Rector of the Tennessee Department of Environment and Conservation
(TDEC) Department of Energy Oversight; Felecia Owen and Colin Green of MTI, Inc.; Leroy Lewis of the Idaho National
Engineering and Environmental Laboratory, retired; Ishwar Murarka of the EPA Science Advisory Board, member; Danny
Reible of Louisiana State University; Mike Bolen, Joseph Evans, Julia Gartseff, Sara Hartwell, Cathleen Hubbard, Kevin
Jago, Andrew Matuson, Allen Motley, John Nicklas, Maurice Owens, Nancy Patti, Fernando Padilla, Mark Pruitt, James
Rawe, Herb Skovronek, and Joseph Tillman of Science Applications International Corporation (SAIC); Scott Jacobs and
Ann Vega of the EPA National Risk Management Research Laboratory's Land Remediation and Pollution Control Division;
and Brian Schumacher of the EPA National Exposure Research Laboratory.
This document was QA reviewed by George Brilis of the EPA National Exposure Research Laboratory.
XVI
-------
Chapter 1
Introduction
The U.S. Environmental Protection Agency (EPA) under
the Office of Research and Development (ORD), National
Exposure Research Laboratory (NERL), conducted a
demonstration to evaluate the performance of innovative
field measurement devices for their ability to measure
mercury concentrations in soils and sediments. This
Innovative Technology Verification Report (ITVR) presents
demonstration performance results and associated costs
of MTI's Portable Digital Voltammeter (PDV) 6000 anodic
stripping voltammetry instrument. The vendor-prepared
comments regarding the demonstration are presented in
Appendix A.
The demonstration was conducted as part of the EPA
Superfund Innovative Technology Evaluation (SITE)
Monitoring and Measurement Technology (MMT) Program.
Mercury contaminated soils and sediments, collected from
four sites within the continental U.S., comprised the
majority of samples analyzed during the evaluation. Some
soil and sediment samples were spiked with mercury (II)
chloride (HgCI2) to provide concentrations not occurring in
the field samples. Certified standard reference material
(SRM) samples were also used to provide samples with
certified mercury concentrations and to increase the matrix
variety.
The demonstration was conducted at the Department of
Energy (DOE) Oak Ridge National Laboratory (ORNL) in
Oak Ridge, TN during the week of May 5, 2003. The
purpose of the demonstration was to obtain reliable
performance and cost data for field measurement devices
in order to 1) provide potential users with a better
understanding of the devices' performance and operating
costs underwell-defined field conditions and 2) provide the
instrument vendors with documented results that can assist
them in promoting acceptance and use of their devices.
The results obtained using the five field mercury
measurement devices were compared to the mercury
results obtained for identical sample sets (samples, spiked
samples, and SRMs) analyzed ata referee laboratory. The
referee laboratory, which was selected prior to the
demonstration, used a well-established EPA reference
method.
1.1 Description of the SITE Program
Performance verification of innovative environmental
technologies is an integral part of the regulatory and
research mission of the EPA. The SITE Program was
established by EPA's Officeof Solid Waste and Emergency
Response (OSWER) and ORD under the Superfund
Amendments and Reauthorization Act of 1986.
The overall goal of the SITE Program is to conduct
performance verification studies and to promote the
acceptance of innovative technologies that may be used to
achieve long-term protection of human health and the
environment. The program is designed to meetthree main
objectives: 1) identify and remove obstacles to the
development and commercial use of innovative
technologies; 2) demonstrate promising innovative
technologies and gather reliable performance and cost
information to support site characterization and cleanup
activities; and 3) develop procedures and policies that
encourage the use of innovative technologies at Superfund
sites, as well as at other waste sites or commercial
facilities.
The SITE Program includes the following elements:
The MMT Program evaluates innovative technologies
that sample, detect, monitor, or measure hazardous
and toxic substances in soil, water, and sediment
samples. These technologies are expected to provide
-------
better, faster, or more cost-effective methods for
producing real-time data during site characterization
and remediation studies than conventional
technologies.
The Remediation Technology Program conducts
demonstrations of innovative treatment technologies to
provide reliable performance, cost, and applicability
data for site cleanups.
The Technology Transfer Program provides and
disseminates technical information in the form of
updates, brochures, and other publications that
promote the SITE Program and participating
technologies. The Technology Transfer Program also
offers technical assistance, training, and workshops in
the support of the technologies. A significant number
of these activities are performed by EPA's Technology
Innovation Office.
The Field Analysis of Mercury in Soils and Sediments
demonstration was performed under the MMT Program.
The MMT Program provides developers of innovative
hazardous waste sampling, detection, monitoring, and
measurement devices with an opportunity to demonstrate
the performance of their devices under actual field
conditions. The main objectives of the MMT Program are
as follows:
Test and verify the performance of innovative field
sampling and analytical technologies that enhance
sampling, monitoring, and site characterization
capabilities.
Identify performance attributes of innovative
technologies that address field sampling, monitoring,
and characterization problems in a cost-effective and
efficient manner.
Prepare protocols, guidelines, methods, and other
technical publications that enhance acceptance of
these technologies for routine use.
The MMT Program is administered by the Environmental
Sciences Division of the NERL in Las Vegas, NV. The
NERL is the EPA center for investigation of technical and
management approaches for identifying and quantifying
risks to human health and the environment. The NERL
mission components include 1) developing and evaluating
methods and technologies for sampling, monitoring, and
characterizing water, air, soil, and sediment; 2) supporting
regulatory and policy decisions; and 3) providing technical
support to ensure the effective implementation of
environmental regulations and strategies.
1.2 Scope of the Demonstration
The demonstration project consisted of two separate
phases: Phase I involved obtaining information on
prospective vendors having viable mercury detection
instrumentation. Phase II consisted of field and planning
activities leading up to and including the demonstration
activities. The following subsections provide detail on both
of these project phases.
1.2.1 Phase I
Phase I was initiated by making contact with
knowledgeable sources on the subject of "mercury in soil"
detection devices. Contacts included individuals within
EPA, Science Applications International Corporation
(SAIC), and industry where measurement of mercury in soil
was known to be conducted. Industry contacts included
laboratories and private developers of mercury detection
instrumentation. In addition, the EPA Task Order Manager
(TOM) provided contacts for "industry players" who had
participated in previous MMT demonstrations. SAIC also
investigated university and other research-type contacts for
knowledgeable sources within the subject area.
These contacts led to additional knowledgeable sources on
the subject, which in turn led to various Internet searches.
The Internet searches were very successful in finding
additional companies involved with mercury detection
devices.
All in all, these research activities generated an original list
of approximately 30 com panies potentially involved in the
measurement of mercury in soils. The list included both
international and U.S. companies. Each of these
companies was contacted by phone or email to acquire
further information. The contacts resulted in 10 companies
that appeared to have viable technologies.
Due to instrument design (i.e., the instrument's ability to
measure mercury in soils and sediments), business
strategies, and stage of technology development, only 5 of
those 10 vendors participated in the field demonstration
portion of phase II.
1.2.2 Phase II
Phase II of the demonstration project involved strategic
planning, field-related activities for the demonstration, data
analysis, data interpretation, and preparation of the ITVRs.
-------
Phase II included pre-demonstration and demonstration
activities, as described in the following subsections.
1.2.2.1 Pre-Demonstration Activities
The pre-demonstration activities were completed in the fall
2002. There were six objectives forthe pre-demonstration:
Establish concentration ranges for testing vendors'
analytical equipment during the demonstration.
Collect soil and sediment field samples to be used in
the demonstration.
Evaluate sample homogenization procedures.
Determine mercury concentrations in homogenized
soils and sediments.
Selecta reference method and qualify potential referee
laboratories for the demonstration.
Provide soil and sediment samples to the vendors for
self-evaluation of their instruments, as a precursor to
the demonstration.
As an integral part of meeting these objectives, a pre-
demonstration sampling event was conducted in
September 2002 to collect field samples of soils and
sediments containing different levels of mercury. The field
samples were obtained from the following locations:
Carson River Mercury site - near Dayton, NV
Y-12 National Security Complex - Oak Ridge, TN
A confidential manufacturing facility - eastern U.S.
Puget Sound - Bellingham Bay, WA
Immediately after collecting field sample material from the
sites noted above, the general mercury concentrations in
the soils and sediments were confirmed by quick
turnaround laboratory analysis of field-collected
subsamples using method SW-7471B. The field sample
materials were then shipped to asoil preparation laboratory
forhomogenization. Additional pre-demonstration activities
are detailed in Chapter 4.
1.2.2.2 Demonstration Activities
Specific objectives for this SITE demonstration were
developed and defined in a Field Demonstration and
Quality Assurance Project Plan (QAPP) (EPA Report #
EPA/600/R-03/053). The Field Demonstration QAPP is
available through the EPA ORD web site
(http://www.epa.gov/ORD/SITE) or from the EPA Project
Manager. The demonstration objectives were subdivided
into two categories: primary and secondary. Primary
objectives are goals of the demonstration study that need
to be achieved for technology verification. The
measurements used to achieve primary objectives are
referred to as critical. These measurements typically
produce quantitative results that can be verified using
inferential and descriptive statistics.
Secondary objectives are additional goals of the
demonstration study developed for acquiring other
information of interest about the technology that is not
directly related to verifying the primary objectives. The
measurements required forachieving secondary objectives
are considered to be noncritical. Therefore, the analysis of
secondary objectives is typically more qualitative in nature
and often uses observations and sometimes descriptive
statistics.
The field portion of the demonstration involved evaluating
the capabilities of five mercury-analyzing instruments to
measure mercury concentrations in soil and sediment.
During the demonstration, each instrument vendor received
three types of samples 1) homogenized field samples
referred to as "field samples", 2) certified SRMs, and 3)
spiked field samples (spikes).
Spikes were prepared by adding known quantities of HgCI2
to field samples. Together, the field samples, SRMs, and
spikes are referred to as "demonstration samples" for the
purpose of this ITVR. All demonstration samples were
independently analyzed by a carefully selected referee
laboratory. The experimental design forthe demonstration
is detailed in Chapter 4.
1.3 Mercury Chemistry and Analysis
1.3.1 Mercury Chemistry
Elemental mercury is the only metal thatoccurs as a liquid
at ambient temperatures. Mercury naturally occurs,
primarily within the ore, cinnabar, as mercury sulfide (HgS).
Mercury easily forms amalgams with many other metals,
including gold. As a result, mercury has historically been
used to recover gold from ores.
Mercury is ionically stable; however, it is very volatile for a
metal. Table 1-1 lists selected physical and chemical
properties of elemental mercury.
-------
Table 1-1. Physical and Chemical Properties of Mercury
Properties Data
Appearance
Hardness
Abundance
Density @ 25 "C
Vapor Pressure @ 25 "C
Volatilizes @
Solidifies @
Silver-white, mobile, liquid.
Liquid
0.5% in Earth's crust
13.53g/mL
0.002 mm
356 "C
-39 "C
Source: Merck Index, 1983
Historically, mercury releases to the environment included
a number of industrial processes such as chloralkali
manufacturing, copper and zinc smelting operations, paint
application, waste oil combustion, geothermal energy
plants, municipal waste incineration, ink manufacturing,
chemical manufacturing, paper mills, leather tanning,
pharmaceutical production, and textile manufacturing. In
addition, industrial and domestic mercury-containing
products, such as thermometers, electrical switches, and
batteries, are disposed of as solid wastes in landfills (EPA,
July 1995). Mercury is also an indigenous compound at
many abandoned mining sites and is, of course, found as
a natural ore.
At mercury-contaminated sites, mercury exists in mercuric
form (Hg2+), mercurous form (Hg22+), elemental form (Hg°),
and alkylated form (e.g., methyl or ethyl mercury). Hg22+
and Hg2+ are the more stable forms under oxidizing
conditions. Under mildly reducing conditions, both
organically bound mercury and inorganic mercury may be
degraded to elemental mercury, which can then be
converted readily to methyl or ethyl mercury by biotic and
abiotic processes. Methyl and ethyl mercury are the most
toxic forms of mercury; the alkylated mercury compounds
are volatile and soluble in water.
Mercury (II) forms relatively strong complexes with Cl'and
CO32". Mercury (II) also forms complexes with inorganic
ligands such as fluoride (F~), bromide (Br~), iodide (I"),
sulfate (SO42"), sulfide (S2~), and phosphate (PO43~) and
forms strong complexes with organic ligands, such as
sulfhydryl groups, amino acids, and humic and fulvicacids.
The insoluble HgS is formed under mildly reducing
conditions.
1.3.2 Mercury Analysis
There are several laboratory-based, EPA promulgated
methods for the analysis of mercury in solid and liquid
hazardous waste matrices. In addition, there are several
performance-based methods for the determination of
various mercury species. Table 1-2 summarizes the
commonly used methods for measuring mercury in both
solid and liquid matrices, as identified through a review of
the EPA Test Method Index and SW-846. A discussion of
the choice of reference method is presented in Chapter 4.
-------
Table 1-2. Methods for Mercury Analysis in Solids or Aqueous Soil Extracts
Method
Analytical
Technology
Type(s) of
Mercury analyzed
Approximate
Concentration Range
Comments
SW-7471B CVAAS
SW-7472 ASV
SW-7473
SW-7474
TD,
amalgamation,
and AAS
AFS
inorganic mercury 10-2,000 ppb
organo-mercury
inorganic mercury 0.1-10,000 ppb
organo-mercury
inorganic mercury 0.2 - 400 ppb
organo-mercury
inorganic mercury 1 ppb - ppm
organo-mercury
Manual cold vapor technique widely
used for total mercury determinations
Newer, less widely accepted method
Allows for total decomposition analysis
Allows for total decomposition analysis;
less widely used/reference
EPA 1631 CVAFS
EPA 245.7 CVAFS
EPA 6200 FPXRF
inorganic mercury 0.5-100ppt
organo-mercury
inorganic mercury
organo-mercury
0.5 - 200 ppt
inorganic mercury >30 mg/kg
Requires "trace" analysis procedures;
written for aqueous matrices; Appendix
A of method written for sediment/soil
samples
Requires "trace" analysis procedures;
written for aqueous matrices; will
require dilutions of high-concentration
mercury samples
Considered a screening protocol
AAS = Atomic Absorption Spectrometry
AAF = Atomic Fluorescence Spectrometry
AFS = Atomic Fluorescence Spectrometry
ASV = Anodic Stripping Voltammetry
CVAAS = Cold Vapor Atomic Absorption Spectrometry
CVAFS = Cold Vapor Atomic Fluorescence Spectrometry
FPXRF = Field Portable X-ray Fluorescence
EPA = U.S. Environmental Protection Agency
mg/kg = milligram per kilogram
ppb = parts per billion
ppm = parts per million
ppt = parts per trillion
SW = solid waste
TD = thermal decomposition
-------
Chapter 2
Technology Description
This chapter provides a detailed description of 1) anodic
stripping voltammetry (ASV), which is the type of
technology on which Mil's instrument is based, and 2) a
detailed description of the actual MTI PDV 6000
instrument.
2.1 Description of Anodic Stripping
Voltammetry
The principle of analysis used by the MTI PDV 6000 is
ASV. ASV is a sensitive method that can be used for the
analysis of trace concentrations of metals in solution,
including digestion solutions from metal-contaminated soils
and sediments. The method involves initially plating metals
onto an electrode by applying a negative voltage, then
stripping the metals back into solution by applying a
positive voltage to the electrode. The ramping of the
positive voltage generates a small but measurable current
(MTI, 2002).
The ASV technique was first discovered in the 1920s by
Jaroslav Heyrovsky, who won the Nobel prize in 195 9 (MTI,
2002). The technique was originally developed with a
hanging Hg drop electrode. However, to limit the quantity
of Hg needed, thin Hg films can be pre-deposited onto an
electrode such as glassy carbon, or co-deposited with the
analyte metal ions. With such films, sensitivities in the low
part-per-billion (ppb) range can be achieved. A negative
potential is applied to the glassy carbon working electrode.
When the electrode potential exceeds the ionization
potential of the analyte metal ion in solution (Mn+), it is
reduced to the metal, which plates onto the working
electrode surface as follows:
where: Mn+ = analyte metal ion in solution
ne" = number of electrons
M = metal plated onto the electrode
The first step of the process is referred to as the
"deposition" or "accumulation" step. This step involves
concentrating the metal on the electrode. The longer the
potential is applied, the more metal is reduced and plated
onto the surface of the electrode. The deposition time is
predetermined. When sufficient metal has been plated
onto the working electrode, the metal is stripped (oxidized)
off the electrode by applying, at a constant rate, a positive
potential applied to the working electrode. For a given
electrolyte solution and electrode, each metal has a
specific potential at which the following oxidation reaction
will occur:
M
Mn++ ne'
Mn++
M
The electrons released by this process form a current. The
current is measured and may be plotted as a function of
the applied potential to give a "voltammogram." The
currentat the oxidation orstripping potentialforthe analyte
metal is seen as a peak. To calculate the sample
concentration, the peak height or area is measured and
compared to that of a known standard solution under the
same conditions. As a metal is identified by the potential
at which oxidation occurs, a number of metals may often
be determined simultaneously due to their differing
oxidation potentials.
The plating step makes it possible to detect very low
concentrations of metal in the sample, and laboratory
versions of an ASV device can measure concentrations in
the parts per trillion (ppt) range. The length of the plating
step can be varied to suit the analyte concentration of the
sample. For example, analysis of a 10-ppb solution of Pb
-------
may require a 3-5 minute accumulation step, while a
solution in the parts per million range would require less
than 1 minute.
2.2 Description of the MTI PDV 6000
The PDV 6000 Portable Analyzer is an instrument thatcan
be used for field screening or laboratory analysis of heavy
metal ions such as silver, arsenic, gold, cadmium,
chromium, copper, iron, mercury, manganese, nickel, lead,
tin, and zinc. The instrument can be used for metal ion
detection in a wide variety of matrixes, including chemicals,
materials, food, beverages, water, industrial effluent, and
Pharmaceuticals. The PDV 6000 consists of two major
components: 1) the main instrument control unit and 2) the
cell assembly. Figure 2-1 is a schematic of the top of the
cell assembly. Within the cell assembly, there are three
separate electrodes surrounding a stirrer motor. The
reference electrode, found in the cell's assembly with the
other electrodes, is a critical component of the PDV 6000.
CELL TOP
Cell Connector
to PDV 6000
Figure 2-1. PDV 6000 eel I assembly.
Applications and Specifications - The MTI PDV 6000
can be operated as a stand-alone instrument or attached
to a laptop or desktop computer to run accompanying
voltammetric analysis system (VAS) software. The VAS
software incorporates features such as standard addition
calibration, simultaneous multi-element analysis, and
storage of data. For stand-alone use, the control unit is
programmed to analyze 10 metals in the concentration
range of 10 ppb to 30 ppm. According to MTI, the
stand-alone unit should only be used for screening
purposes (i.e., determining whether a sample may be
above or below a particular threshold).
According to MTI, the VAS software allows for better
detection limits and more accurate analysis. A specific
electrolyte and electrode combination are used to optimize
results for specific metals. This is essential for detection
limits in the low mg/kg range. Where the detection range
is in the g/kg range, it is possible to analyze a larger range
of metals per scan, but the reproducibility will be around
10%, as opposed to the 3% typically seen when optimum
conditions are used. High silver concentrations can
interfere with mercury determinations.
Operation - The PDV 6000 control unit is a hand-held
portable device, weighing approximately 700 g, and has a
dimension of 10 cm by 18 cm by 4 cm. When used in the
field (Figure 2-2), the PDV 6000 should only be used to
indicate the approximate concentration range of the metal
of interest. This is true for all field analysis since many
factors can reduce the accuracy and precision of any
analysis. (When used in the laboratory, the vendor reports
the PDV 6000 can provide accurate and reproducible
data). According to MTI, it is realistic to expect the PDV
6000 to obtain data from the field that is within 20% of the
true value. For this reason, it is best to use the PDV 6000
to classify samples as "above a threshold concentration" or
"below a threshold concentration."
Figure 2-2. Photograph of the PDV 6000 during the field
demonstration.
-------
Certain field conditions may affect the accuracy and
precision of results. These include the following:
Sample homogeneity
Sampling handling errors
Pipetting errors
Unpredictable matrix effects
Sample and cell contamination
Essential Components and Supplies- Fortypical use as
a field instrument, the PDV 6000 is shipped with a carrying
case, the handheld control unit, cell assembly and cell
stand, Ag/AgCI reference electrode, platinum counter
electrode, glassy carbon working electrode, cable to link
the analytical cell to the control unit, a DB9 serial cable to
link the control unit to a computer, a reference electrode
plating accessory, a main-powered 12 volt (V) direct
current (DC) supply, a nickel metal halide (NiMH) battery
pack and battery pack charger, the PDV 6000 operation
manual, and VAS software installation disks and the VAS
User's Guide.
For solids analysis, ancillary supplies are required to
conduct extractions. These include a portable scale for
measuring the correct sample mass to extract; a set of air
displacement pipetters (50-500 uL, 100-1000 uL, and
200-1000 uL); a repeating pipetter (0.5-50 ml_); and
extraction reagents. It is recommended that kits be
purchased to supply the extraction reagents and
disposable supplies (e.g., bottles) necessary to extract and
analyze the samples. To "digest" the solids, a slightly
modified Method 3050B is used from EPA's Test Methods
for Evaluating Solid Waste; Physical/Chemical Methods
(SW-846).
Power Supply - The PDV 6000 can operate using either
a 110 V alternating current (AC) source or direct current
battery. The specific power supply options for the PDV
6000, as presented in Table 2-1, affect the number of
analyses that may be performed on a daily basis.
Table 2-1. Power Supply Options for the PDV 6000
Power Source No. of Analyses
9VPP3 10-20
Rechargeable Battery Pack 50
9V main power supply a Continuous
a Main power supply also recharges the battery pack and
powers the reference electrode plating accessory.
Instrument Calibration - The standard curve method
compares the sample response with that of one or more
known standards. Voltage readings can allow calibration
curves of between 1-10 standards to be constructed and
then compared with up to 15 samples. Generally,
calibration is based on a single pointcomparison, whereby
the voltage generated by the standard is compared to the
voltage generated by the sample. The response for a
particular analyte is proportional to its concentration in the
analytical cell; therefore, dilution by electrolyte or other
reagents must be taken into consideration. For best
results, the sample concentration in the cell should be
close to the cell concentration of the standard with which it
is being compared. Standard addition calibration involves
analyzing a sample and then "spiking" the same sample
solution with a small volume of standard before reanalyzing
that solution. The same sample can be spiked and
reanalyzed once or several times, as necessary. The
spiking enables the result to be calculated by the method
of standard additions. The results from the sample and
spiked sample runs are then plotted, and a line of
regression is fitted and used to calculate the sample
concentration.
VAS Software - The use of the supplied VAS software
allows more flexible and accurate analyses, even in the
field. This software incorporates features such as multiple
point calibration curves or standard addition calibration,
simultaneous multi-element analysis, easy parameter
optimization and manual baseline, and peak center
adjustment. It also stores data, and allows comments to
be added about the sample location and a description of
the sample. The VAS software requires a computer that
has a minimum specification of a 300 Mhz 586 processor
(Pentium or equivalent), 64 MB RAM, 10 MB of free hard
disk space, and runs Windows 98, ME, NT, 2000, or XP.
2.3 Developer Contact Information
Additionalinformation aboutthe PDV6000can be obtained
from the following source:
International Contact
Monitoring Technologies International Pty. Ltd.
1/7 Collingwood Street
Osborne Park, Perth
West Australia 6017
Internet email: support@mti.com.au
Web address: www.monitoring-technologies.com.
U.S. Contact
Felicia Owen, MTI Inc.
1609 Ebb Drive
Wilmington, NC 28409
Telephone:(910)392-5714 Fax:(910)392-4320
email: fowen@owenscientific.com
-------
Chapter 3
Field Sample Collection Locations and Demonstration Site
As previously described in Chapter 1, the demonstration in
part tested the ability of all five vendor instruments to
measure mercury concentrations in demonstration
samples. The demonstration samples consisted of field-
collected samples, spiked field samples, and SRMs. The
field-collected samples comprised the majority of
demonstration samples. This chapter describes the four
sites from which the field samples were collected, the
demonstration site, and the sample homogenization
laboratory. Spiked samples were preparedfrom these field
samples.
Screening of potential mercury-contaminated field sample
sites was conducted during Phase I of the project. Four
sites were selected for acquiring mercury-contaminated
samples thatwere diverse in appearance, consistency, and
mercury concentration. A key criterion was the source of
the contamination. These sites included:
Carson River Mercury site - near Dayton, NV
The Y-12 National Security Complex (Y-12) - Oak
Ridge, TN
A confidential manufacturing facility - eastern U.S.
Puget Sound - Bellingham Bay, WA
Site Diversity - Collectively, the four sites provided
sampling areas with both soil and sediment, having
variable physical consistencies and variable ranges of
mercury contamination. Two of the sites (Carson River
and Oak Ridge) provided both soil and sediment samples.
A third site (a manufacturing facility) provided just soil
samples and a fourth site (Puget Sound) provided only
sediment samples.
Access and Cooperation - Site representatives were
instrumental in providing site access, and in some cases,
guidance on the best areas to collect samples from
relatively high and low mercury concentrations. In addition,
representatives from the host demonstration site (ORNL)
provided a facility for conducting the demonstration.
At three of the sites, the soil and/or sediment sample was
collected, homogenized by hand in the field, and
subsampled for quick turnaround analysis. These
subsamples were sent to analytical laboratories to
determine the general range of mercury concentrations at
each of the sites. (The Puget Sound site did not require
confirmation of mercury contamination due to recently
acquired mercury analytical data from another, ongoing
research project.) The field-collected soil and sediment
samples from all four sites were then shipped to SAIC's
GeoMechanics Laboratory for a more thorough sample
homogenization (see Section 4.3.1) and subsampled for
redistribution to vendors during the pre-demonstration
vendor self-evaluations.
All five of the technology vendors performed a self-
evaluation on selected samples collected and
homogenized during this pre-demonstration phase of the
project. For the self-evaluation, the laboratory results and
SRM values were supplied to the vendor, allowing the
vendor to determine how well it performed the analysis on
the field samples. The results were used to gain a
preliminary understanding of the field samples collected
and to prepare for the demonstration.
Table 3-1 summarizes key characteristics of samples
collected at each of the four sites. Also included are the
sample matrix, sample descriptions, and sample depth
intervals. The analytical results presented in Table 3-1 are
based on referee laboratory mercury results for the
demonstration samples.
-------
Table 3-1. Summary of Site Characteristics
Site Name
Carson River
Mercury site
Y-12 National
Security Complex
Confidential
manufacturing site
Puget Sound -
Bellingham Bay
Sampling Area
Carson River
Six Mile Canyon
Old Hg Recovery Bldg.
Poplar Creek
Former plant building
Sediment layer
Underlying Native Material
Sample
Matrix
Sediment
Soil
Soil
Sediment
Soil
Sediment
Sediment
Depth
water/sediment
interface
3 - 8 cm bgs
0 - 1 m bgs
0 - 0.5 m bgs
3. 6 -9m bgs
1.5-1.8 m thick
0.3 m thick
Description
Sandy silt, with some
organic debris present
(plant stems and leaves)
Silt with sand to sandy silt
Silty-clay to sandy-gravel
Silt to coarse sandy gravel
Silt to sandy silt
Clayey-sandy silt with
various woody debris
Medium-fine silty sands
Hg Concentration
Range
10 ppb - 50 ppm
10 ppb- 1,000 ppm
0.1 - 100 ppm
0.1 - 100 ppm
5- 1,000 ppm
10 -400 ppm
0.16- 10 ppm
bgs = below ground surface.
3.1 Carson River
3.1.1 Site Description
The Carson River Mercury site begins near Carson City,
NV, and extends downstream to the Lahontan Valley and
the Carson Desert. During the Comstock mining era of the
late 1800s, mercury was imported to the area for
processing gold and silver ore. Ore mined from the
Comstock Lode was transported to mill sites, where it was
crushed and mixed with mercury to amalgamate the
precious metals. The Nevada mills were located in Virginia
City, Silver City, Gold Hill, Dayton, Six Mile Canyon, Gold
Canyon, and adjacent to the Carson River between New
Empire and Dayton. During the mining era, an estimated
7,500 tons of mercury were discharged into the Carson
River drainage, primarily in the form of
mercury-contaminated tailings (EPA Region 9, 1994).
Mercury contamination is present at Carson Riveras either
elemental mercury and/or inorganic mercury sulfides with
less than 1%, if any, methylmercury. Mercury
contamination exists in soils presentat the former gold and
silvermining mill sites; waterways adjacentto the millsites;
and sediment, fish, and wildlife over more than a 50-mile
length of the Carson River. Mercury is also present in the
sediments and adjacent flood plain of the Carson River,
and in the sediments of Lahontan Reservoir, Carson Lake,
Stillwater Wildlife Refuge, and Indian Lakes. In addition,
tailings with elevated mercury levels are still presentat, and
around, the historic mill sites, particularly in Six Mile
Canyon (EPA, 2002a).
3.1.2 Sample Collection
The Carson River Mercury site provided both soil and
sediment samples across the range of contaminant
concentrations desired for the demonstration. Sixteen
near-surface soil samples were collected between 3-8 cm
below ground surface (bgs). Two sediment samples were
collected at the water-to-sediment interface. All 18
samples were collected on September 23-24, 2002 with a
hand shovel. Samples were collected in Six Mile Canyon
and along the Carson River.
The sampling sites were selected based upon historical
data from the site. Specific sampling locations in the Six
Mile Canyon were selected based upon local terrain and
visible soil conditions (e.g., color and particle size). The
specific sites were selected to obtain soil samples with as
much variety in mercury concentration as possible. These
sites included hills, run-off pathways, and dry river bed
areas. Sampling locations along the Carson River were
selected based upon historical mine locations, local terrain,
and river flow.
When collecting the soil samples, approximately 3 cm of
surface soil was scraped to the side. The sample was
then collected with a shovel, screened through a
6.3-millimeter (mm) (0.25-inch) sieve to remove larger
material, and collected in 4-liter (L) scalable bags identified
with a permanent marker. The sediment samples were
also collected with a shovel, screened through a 6.3-mm
sieve to remove larger material, and collected in 4-L
scalable bags identified with a permanent marker. Each of
the 4-L scalable bags was placed into a second 4-L
10
-------
sealable bag, and the sample label was placed onto the
outside bag. The sediment samples were then placed into
10-L buckets, lidded, and identified with a sample label.
3.2 Y-12 National Security Complex
3.2.1 Site Description
The Y-12 site is located at the DOE ORNL in Oak Ridge,
TN. The Y-12 site is an active manufacturing and
developmental engineering facility that occupies
approximately 800 acres on the northeast corner of the
DOE Oak Ridge Reservation (ORR) adjacent to the city of
Oak Ridge, TN. Built in 1 943 by the U.S. Army Corps of
Engineers as part of the World War II Manhattan Project,
the original mission of the installation was development of
electromagnetic separation of uranium isotopes and
weapon components manufacturing, as partof the national
effort to produce the atomic bomb. Between 1950 and
1963, large quantities of elemental mercury were used at
Y-12 during lithium isotope separation pilot studies and
subsequent production processes in support of
thermonuclear weapons programs.
Soils at the Y-12 facility are contaminated with mercury in
many areas. One of the areas of known high levels of
mercury-contaminated soils is in the vicinity of a former
mercury use facility (the "Old Mercury Recovery Building"
- Building 8110). At this location, mercury-contaminated
material and soil were processed in a Nicols-Herschoff
roasting furnace to recover mercury. Releases of mercury
from this process, and from a building sump used to
secure the mercury-contaminated materials and the
recovered mercury, have contaminated the surrounding
soils (Rothchild, et al., 1984). Mercury contamination also
occurred in the sediments of the East Fork of Poplar Creek
(DOE, 1998). The Upper East Fork of Poplar Creek
(UEFPC) drains the entire Y-12 complex. Releases of
mercury via building drains connected to the storm sewer
system, building basement dewatering sump discharges,
and spills to soils, all contributed to contamination of
UEFPC. Recent investigations showed that bank soils
containing mercury along the UEFPC were eroding and
contributing to mercury loading. Stabilization of the bank
soils along this reach of the creek was recently completed.
3.2.2 Sample Collection
Two matrices were sampled at Y-12 in Oak Ridge, TN,
creek sediment and soil. A total of 10 sediment samples
was collected; one sediment sample was collected from
the Lower East Fork of Poplar Creek (LEFPC) and nine
sediment samples were collected from the UEFPC. A total
of six soil samples was collected from the Building 8110
area. The sampling procedures that were used are
summarized below.
Creek Sediments - Creek sediments were collected on
September 24-25, 2002 from the East Fork of Poplar
Creek. Sediment samples were collected from various
locations in a downstream to upstream sequence (i.e., the
downstream LEFPC sample was collected first and the
most upstream point of the UEFPC was sampled last).
The sediment samples from Poplar Creek were collected
using a commercially available clam-shell sonar dredge
attached to a rope. The dredge was slowly lowered to the
creek bottom surface, where it was pushed by foot into the
sediment. Several drops of the sampler (usually seven or
more) were made to collect enough material for screening.
On some occasions, a shovel was used to remove
overlying "hardpan" gravel to expose finer sediments at
depth. One creek sample consisted of creek bank
sediments, which was collected using a stainless steel
trowel.
The collected sediment was then poured onto a 6.3-mm
sieve to remove oversize sample material. Sieved samples
were then placed in 12-L sealable plastic buckets. The
sediment samples in these buckets were homogenized
with a plastic ladle and subsamples were collected in 20-
milliliter (mL) vials for quick turnaround analyses.
Soil - Soil samples were collected from pre-selected
boring locations September 25, 2002. All samples were
collected in the immediate vicinity of the Building 8110
foundation using a commercially available bucket auger.
Oversize material was hand picked from the excavated soil
because the soil was too wet to be passed through a sieve.
The soil was transferred to an aluminum pan,
homogenized by hand, and subsampled to a 20-mL vial.
The remaining soil was transferred to 4-L plastic
containers.
3.3 Confidential Manufacturing Site
3.3.1 Site Description
A confidential manufacturing site, located in the eastern
U.S., was selected for participation in this demonstration.
The site contains elemental mercury, mercury amalgams,
and mercury oxide in shallow sediments (less than 0.3 m
deep) and deeper soils (3.65 to 9 m bgs). This site
provided soil with concentrations from 5-1,000 mg/kg.
The site is the location of three former processes that
resulted in mercury contamination. The first process
11
-------
involved amalgamation of zinc with mercury. The second
process involved the manufacturing of zinc oxide. The
third process involved the reclamation of silver and gold
from mercury-bearing materials in a retort furnace.
Operations led to the dispersal of elemental mercury,
mercury compounds such as chlorides and oxides, and
zinc-mercury amalgams. Mercury values have been
measured ranging from 0.05 to over 5,000 mg/kg, with
average values of approximately 100 mg/kg.
3.3.2 Sample Collection
Eleven subsurface soil samples were collected on
September 24, 2002. All samples were collected with a
Geoprobe® unit using plastic sleeves. All samples were
collected at the location of a former facility plant. Drilling
locations were determined based on historical data
provided by the site operator. The intention was to gather
soil samples across a range of concentrations. Because
the surface soils were from relatively clean fill, the sampling
device was pushed to a depth of 3.65 m using a blank rod.
Samples were then collected at pre-selected depths
ranging from 3.65 to 9 m bgs. Individual cores were 1-m
long. The plastic sleeve for each 1-m core was marked
with a permanent marker; the depth interval and the bottom
of each core was marked. The filled plastic tubes were
transferred to a staging table where appropriate depth
intervals were selected for mixing. Selected tubes were cut
into 0.6-m intervals, which were emptied into a plastic
container for premixing soils. When feasible, soils were
initially screened to remove materials larger than 6.3-mm
in diameter. In many cases, soils were too wet and clayey
to allow screening; in these cases, the soil was broken into
pieces by hand and, by using a wooden spatula, oversize
materials were manually removed. These soils (screened
or hand sorted) were then mixed until the soil appeared
visually uniform in color and texture. The mixed soil was
then placed into a 4-L sample container for each chosen
sample interval. A subsample of the mixed soil was
transferred into a 20-mL vial, and it was sent for quick
turnaround mercury analysis. This process was repeated
for each subsequent sample interval.
3.4 Puget Sound
3.4.1 Site Description
The Puget Sound site consists of contaminated offshore
sediments. The particular area of the site used for
collecting demonstration samples is identified as the
Georgia Pacific, Inc. Log Pond. The Log Pond is located
within the Whatcom Waterway in Bellingham Bay, WA, a
well-established heavy industrial land use area with a
maritime shoreline designation. Log Pond sediments
measure approximately 1.5 to 1.8-m thick, and contain
various contaminants including mercury, phenols,
polyaromatic hydrocarbons, polychlorinated biphenyls, and
wood debris. Mercury was used as a preservative in the
logging industry. The area was capped in late 2000 and
early 2001 with an average of 7 feet of clean capping
material, as part of a Model Toxics Control Act interim
cleanup action. The total thickness ranges from
approximately 0.15 m along the site perimeter to 3 m within
the interior of the project area. The restoration project
produced 2.7 acres of shallow sub-tidal and 2.9 acres of
low intertidal habitat, all of which had previously exceeded
the Sediment Management Standards cleanup criteria
(Anchor Environmental, 2001).
Mercury concentrations have been measured ranging from
0.16 to 400 mg/kg (dry wt). The majority (98%) of the
mercury detected in near-shore ground waters and
sediments of the Log Pond is believed to be comprised of
complexed divalent (Hg2+) forms such as mercuric sulfide
(Bothner, et al., 1980 and Anchor Environmental, 2000).
3.4.2 Sample Collection
Science Applications International Corporation (SAIC) is
currently performing a SITE remedialtechnology evaluation
in the Puget Sound (SAIC, 2002). As part of ongoing work
at that site, SAIC collected additional sediment for use
during this MMT project. Sediment samples collected on
August 20-21, 2002 from the Log Pond in Puget Sound
were obtained beneath approximately 3-6 m of water, using
a vibra-coring system capable of capturing cores to 0.3 m
below the proposed dredging prism. The vibra-corer
consisted of a core barrel attached to a power head.
Aluminum core tubes, equipped with a stainless steel
"eggshell" core catcher to retain material, were inserted
into the core barrel. The vibra-core was lowered into
position on the bottom and advanced to the appropriate
sampling depth. Once sampling was completed, the
vibra-core was retrieved and the core liner removed from
the core barrel. The core sample was examined at each
end to verify that sufficient sediment was retained for the
particular sample. The condition and quantity of material
within the core was then inspected to determine
acceptability.
The following criteria were used to verify whether an
acceptable core sample was collected:
Target penetration depth (i.e., into native material) was
achieved.
12
-------
Sediment recovery of at least 65% of the penetration
depth was achieved.
Sample appeared undisturbed and intact without any
evidence of obstruction/blocking within the core tube or
catcher.
The percentsediment recovery was determined by dividing
the length of material recovered by the depth of core
penetration below the mud line. If the sample was deemed
acceptable, overlying water was siphoned from the top of
the core tube and each end of the tube capped and sealed
with duct tape. Following core collection, representative
samples were collected from each core section
representing a different vertical horizon. Sediment was
collected from the center of the core that had not been
smeared by, or in contact with, the core tube. The volumes
removed were placed in a decontaminated stainless steel
bowl or pan and mixed until homogenous in texture and
color (approximately 2 minutes).
After all sediment for a vertical horizon composite was
collected and homogenized, representative aliquots were
placed in the appropriate pre-cleaned sample containers.
Samples of both the sediment and the underlying native
materialwere collected in a similarmanner. Distinct layers
of sediment and native material were easily recognizable
within each core.
3.5 Demonstration Site
The demonstration was conducted in a natural
environment, outdoors, in Oak Ridge, TN. The area was
a grass covered hill with some parking areas, all of which
were surrounded by trees. Building 5507, in the center of
the demonstration area, provided facilities for lunch, break,
and sample storage for the project and personnel.
Most of the demonstration was performed during rainfall
events ranging from steady to torrential. Severe puddling
of rain occurred to the extent that boards needed to be
placed under chairs to prevent them from sinking into the
ground. Even when it was not raining, the relative humidity
was high, ranging from 70.6 to 98.3 percent. Between two
and four of the tent sides were used to keep rainfall from
damaging the instruments. The temperature in the
afternoons ranged from 65-70 degrees Fahrenheit, and the
wind speed was less than 10 mph. The latitude is 36°N,
the longitude 35°W, and the elevation 275 m. (Figure 3-1
is a photograph of the site during the demonstration and
Figure 3-2 is a photograph of the location.)
Figure 3-1. Tent and field conditions during the
demonstration at Oak Ridge, TN.
Figure 3-2. Demonstration site and Building 5507.
13
-------
3.6 SAIC GeoMechanics Laboratory
Sample homogenization was completed at the SAIC
GeoMechanics Laboratory in Las Vegas, NV. This facility
is an industrial-type building with separate facilities for
personnel offices and material handling. The primary
function of the laboratory is for rock mechanics studies.
The laboratory has rock mechanics equipment, including
sieves, rockcrushers, and sample splitters. The personnel
associated with this laboratory are experienced in the areas
of sample preparation and sample homogenization. In
addition to the sample homogenization equipment, the
laboratory contains several benches, tables, and open
space. Mercury air monitoring equipment was used during
the sample preparation activities for personnel safety.
14
-------
Chapter 4
Demonstration Approach
This chapter describes the demonstration approach that
was used for evaluating the field mercury measurement
technologies at ORNL in May 2003 and in Las Vegas in
June 2003. It presents the objectives, design, sample
preparation and management procedures, and the
reference method confirmatory process used for the
demonstration.
4.1 Demonstration Objectives
The primary goal of the SITE MMT Program is to develop
reliable performance and cost data on innovative,
field-ready measurement technologies. A SITE
demonstration must provide detailed and reliable
performance and cost data in order that potential
technology users have adequate information needed to
make sound judgements regarding an innovative
technology's applicability to a specific site and to be able to
compare the technology to conventional technologies.
Table 4-1 summarizes the project objectives for this
demonstration. In accordance with QAPP Requirements
for Applied Research Projects (EPA,1998), the technical
project objectives for the demonstration were categorized
as primary and secondary.
Table 4-1. Demonstration Objectives
Objective
Description
Method of Evaluation
Primary Objectives
Primary Objective # 1
Primary Objective # 2
Primary Objective # 3
Primary Objective # 4
Primary Objective # 5
Determine sensitivity of each instrument with respect to vendor-generated MDL and
PQL.
Determine potential analytical accuracy associated with vendor field measurements.
Evaluate the precision of vendorfield measurements.
Measure time required to perform five functions related to mercury measurements:
1) mobilization and setup, 2) initial calibration, 3) daily calibration, 4) sample
analysis, and 5) demobilization.
Estimate costs associated with mercury measurements for the following four
categories: 1) capital. 2) labor. 3) supplies, and 4) investigation-derived wastes.
Independent laboratory
confirmation of SRMs,
field samples, and
spiked field samples.
Documentation during
demonstration; vendor-
provided information.
Secondary Objectives
Secondary Objective # 1
Secondary Objective # 2
Secondary Objective # 3
Secondary Objective # 4
Secondary Objective # 5
Document ease of use, skills, and training required to operate the device properly.
Document potential H&S concerns associated with operating the device.
Document portability of the device.
Evaluate durability of device based on materials of construction and engineering
design.
Document the availability of the device and its spare parts.
Documentation of
observations during
demonstration; vendor-
provided information.
Post-demonstration
investigation.
15
-------
Critical data support primary objectives and non critical data
support secondary objectives. With the exception of the
cost information, primary objectives required the use of
quantitative results to draw conclusions regarding the
technology performance. Secondary objectives pertained
to information that was useful and did not necessarily
require the use of quantitative results to draw conclusions
regarding technology performance.
4.2 Demonstration Design
4.2.1
Approach for
Objectives
Addressing Primary
The purpose of this demonstration was to evaluate the
performance of the vendor's instrumentation against a
standard laboratory procedure. In addition, an overall
average relative standard deviation (RSD) was calculated
for all measurements made by the vendor and the referee
laboratory. RSD comparisons used descriptive statistics,
not inferential statistics, between the vendorand laboratory
results. Other statistical comparisons (both inferential and
descriptive) for sensitivity, precision, and accuracy were
used, depending upon actual demonstration results.
The approach for addressing each of the primary
objectives is discussed in the following subsections. A
detailed explanation of the precise statistical determination
used for evaluating primary objectives No. 1 through No. 3
is presented in Chapter 6.
4.2.1.1 Primary Objective #1: Sensitivity
Sensitivity is the ability of a method or instrument to
discriminate between small differences in analyte
concentration (EPA, 2002b). It can be discussed in terms
of an instrument detection limit (IDL), a method detection
limit (MDL), and as a practical quantitation limit (PQL).
MDL is not a measure of sensitivity in the same respect as
an IDL or PQL. It is a measure of precision at a
predetermined, usually low, concentration. The IDL
pertains to the ability of the instrument to determine with
confidence the difference between a sample that contains
the analyte of interest at a low concentration and a sample
that does not contain that analyte. The IDL is generally
considered to be the minimum true concentration of an
analyte producing a non-zero signal that can be
distinguished from the signals generated when no
concentration of the analyte is present and with an
adequate degree of certainty.
The IDL is not rigidly defined in terms of matrix, method,
laboratory, or analyst variability, and it is not usually
associated with a statistical level of confidence. IDLs are,
thus, usually lower than MDLs and rarely serve a purpose
in terms of project objectives (EPA, 2002b). The PQL
defines a specific concentration with an associated level of
accuracy. The MDL defines a lower limit at which a
method measurement can be distinguished from
background noise. The PQL is a more meaningful
estimate of sensitivity. The MDL and PQL were chosen as
the two distinct parameters for evaluating sensitivity. The
approach for addressing each of these indicator
parameters is discussed separately in the following
paragraphs.
MDL
MDL is the estimated measure of sensitivity as defined in
40 Code of Federal Regulations (CFR) Part 136. The
purpose of the MDL measurement is to estimate the
concentration at which an individualfield instrument is able
to detect a minimum concentration that is statistically
different from instrument background or noise. Guidance
for the definition of the MDL is provided in EPA G-5i (EPA,
2002b).
The determination of a MDL usually requires seven
different measurements of a low concentration standard or
sample. Following procedures established in 40 CFR Part
136 for water matrices, the demonstration MDL definition
is as follows:
where: t(n_1?099) =
MDL = Vi.o.Q9)s
99 percentile of the t-distribution
with n -1 degrees of freedom
number of measurements
standard deviation of replicate
measurements
PQL
The PQL is another important measure of sensitivity. The
PQL is defined in EPA G-5i as the lowest level an
instrument is capable of producing a result that has
significance in terms of precision and bias. (Bias is the
difference between the measured value and the true
value.) It is generally considered the lowest standard on
the instrument calibration curve. It is often 5-10 times
higher than the MDL, depending upon the analyte, the
instrument being used, and the method for analysis;
however, it should not be rigidly defined in this manner.
During the demonstration, the PQL was to be defined by
the vendor's reported calibration or based upon lower
16
-------
concentration samples or SRMs. The evaluation of
vendor-reported results for the PQL included a
determination of the percent difference (%D) between their
calculated value and true value. The true value is
considered the value defined by the referee laboratory for
field samples or spiked field samples, or, in the case of
SRMs, the certified value reported by the supplier. The
equation used for the %D calculation is:
%D =!.
' calculated
x100
'true
where: Ctri
'calculated "
true concentration as determined
by the referee laboratory or SRM
reference value
calculated test sample
concentration
The PQL and %D were reported for the vendor. The %D
for the referee laboratory, at the same concentration, was
also reported for purposes of comparison. No statistical
comparison was made between these two values; only a
descriptive comparison was made for purposes of this
evaluation. (The %D requirement forthe referee laboratory
was defined as 10% or less. The reference method PQL
was approximately 10 ug/kg.)
4.2.1.2 Primary Objective #2: Accuracy
Accuracy was calculated bycomparing the measured value
to a known or true value. For purposes of this
demonstration, three separate standards were used to
evaluate accuracy. These included: 1) SRMs, 2) field
samples collected from four separate mercury-
contaminated sites, and 3)spiked field samples. Foursites
were used for evaluation of the MTI field instrument.
Samples representing all three standard types were
prepared at the SAIC GeoMechanics Laboratory. In order
to preventcross contamination, SRMs were prepared in a
separate location. Each of these standards is discussed
separately in the following paragraphs.
SRMs
The primary standards used to determine accuracy for this
demonstration were SRMs. SRMs provided very tight
statistical comparisons, although they did not provide all
matrices of interest nor all ranges of concentrations. The
SRMs were obtained from reputable suppliers, and had
reported concentrations at associated 95% confidence
intervals (CIs) and 95% prediction intervals. Prediction
intervals were usedforcomparison because they represent
a statistically infinite number of analyses, and therefore,
would include all possible correct results 95% of the time.
All SRMs were analyzed by the referee laboratory and
selected SRMs were analyzed by the vendor, based upon
instrument capabilities and concentrations of SRMs that
could be obtained. Selected SRMs covered an appropriate
range for each vendor. Replicate SRMs were also
analyzed by the vendor and the laboratory.
The purpose for SRM analysis by the referee laboratory
was to provide a check on laboratory accuracy. During the
pre-demonstration, the referee laboratory was chosen, in
part, based upon the analysis of SRMs. This was done to
ensure a competent laboratory would be used for the
demonstration. Because of the need to provide confidence
in laboratory analysis during the demonstration, the referee
laboratory analyzed SRMs as an ongoing check for
laboratory bias.
Evaluation ofvendorand laboratory analysis of SRMs was
performed as follows. Accuracy was reported for
individual sample concentrations of replicate
measurements made at the same concentration.
Two-tailed 95% CIs were computed according to the
following equation:
l(n-1.0.975)
.3
where: t(n_..
'(n-1,0.975)
97.5th percentile of the
t-distribution with n-1 degrees of
freedom
number of measurements
standard deviation of replicate
measurements
The number of vendor-reported SRM results and referee
laboratory-reported SRM results that were within the
associated 95% prediction interval were evaluated.
Prediction intervals were computed in a similar fashion to
the Cl, except that the Student's "t" value use "n" equal to
infinity and, because prediction intervals represented "n"
approaching infinity, the square root of "n" was dropped
from the equation.
A final measure of accuracy determined from SRMs is a
frequency distribution that shows the percentage of vendor-
reported measurements thatare within a specified window
of the reference value. For example, a distribution within
a 30% window of a reported concentration, within a 50%
17
-------
window, and outside a 50% window of a reported
concentration. This distribution aspect could be reported
as average concentrations of replicate results from the
vendorfora particular concentration and matrix compared
to the same sample from the laboratory. These are
descriptive statistics and are used to better describe
comparisons, but they are not intended as inferential tests.
Field Samples
The second accuracy standard used for this demonstration
was actual field samples collected from four separate
mercury-contaminated sites. This accuracy determination
consisted of a comparison of vendor-reported results for
field samples to the referee laboratory results for the same
field samples. The field samples were used to ensure that
"real-world" samples were tested for each vendor. The
field samples consisted ofvariable mercury concentrations
within varying soil and sediment matrices. The referee
laboratory results are considered the standard for
comparison to each vendor.
Vendor sample results for a given field sample were
compared to replicates analyzed by the laboratory for the
same field sample. (A hypothesis test was used with alpha
= 0.01. The null hypothesis was that sample results were
similar. Therefore, if the null hypothesis is rejected, then
the sample sets are considered different.) Comparisons
fora specific matrix orconcentration were made in order to
provide additional information on that specific matrix or
concentration. Comparison of the vendor values to
laboratory values were similar to the comparisons noted
previously for SRMs, except that a more definitive or
inferential statistical evaluation was used. Alpha = 0.01
was used to help mitigate inter-laboratory variability.
Additionally, an aggregate analysis was used to mitigate
statistical anomalies (see Section 6.1.2).
Spiked Field Samples
The third accuracy standard for this demonstration was
spiked field samples. These spiked field samples were
analyzed by the vendors and by the referee laboratory in
replicate in order to provide additional measurement
comparisons to a known value. Spikes were prepared to
cover additional concentrations not available from SRMs or
the samples collected in the field. They were grouped with
the field sample comparison noted above.
4.2.1.3 Primary Objective #3: Precision
Precision can be defined as the degree of mutual
agreement of independent measurements generated
through repeated application of a process under specified
conditions. Precision is usually thought of as repeatability
of a specific measurement, and it is often reported as RSD.
The RSD is computed from a specified number of
replicates. The more replications of a measurement, the
more confidence is associated with a reported RSD.
Replication of a measurement may be as few as 3
separate measurements to 30 or more measurements of
the same sample, dependent upon the degree of
confidence desired in the specified result. The precision
of an analytical instrument may vary depending upon the
matrix being measured, the concentration of the analyte,
and whether the measurement is made for an SRM or a
field sample.
The experimental design for this demonstration included a
mechanism to evaluate the precision of the vendors'
technologies. Field samples from the four mercury-
contaminated field sites were evaluated by each vendor's
analytical instrument. During the demonstration,
concentrations were predetermined only as low, medium,
or high. Ranges of test samples (field samples, SRMs,
and spikes) were selected to cover the appropriate
analytical ranges of the vendor's instrumentation. It was
known prior to the demonstration that not all vendors were
capable of measuring similar concentrations (i.e., some
instruments were better at measuring low concentrations
and others were geared toward higher concentration
samples or had otherattributes such as cost or ease of use
that defined specific attributes of their technology).
Because of this fact, not all vendors analyzed the same
samples.
During the demonstration, the vendor's instrument was
tested with samples from the four different sites, having
different matrices when possible (i.e., depending upon
available concentrations) and having different
concentrations (high, medium, and low) using a variety of
samples. Sample concentrations for an individual
instrument were chosen based upon vendor attributes in
terms of expected low, medium, and high concentrations
that the particular instrument was capable of measuring.
The referee laboratory measured replicates of all samples.
The results were used for precision comparisons to the
individual vendor. The RSD for the vendor and the
laboratory was calculated individually, using the following
equation:
$
%RSD = -x100
18
-------
where: S = standard deviation of replicate results
x = mean value of replicate results
Using descriptive statistics, differences between vendor
RSD and referee laboratory RSD were determined. This
included RSD comparisons based upon concentration,
SRMs, field samples, and different sites. In addition, an
overall average RSD was calculated for all measurements
made by the vendor and the laboratory. RSD com parisons
were based upon descriptive statistical evaluations
between the vendor and the laboratory, and results were
compared accordingly.
4.2.1.4 Primary Objective #4: Time per Analysis
The amount of time required for performing the analysis
was measured and reported for five categories:
Mobilization and setup
Initial calibration
Daily calibration
Sample analyses
Demobilization
Mobilization and setup included the time needed to unpack
and prepare the instrument for operation. Initial calibration
included the time to perform the vendor recommended
on-site calibrations. Daily calibration included the time to
perform the vendor-recommended calibrations on
subsequent field days. (Note that this could have been the
same as the initial calibration, a reduced calibration, or
none.) Sample analyses included the time to prepare,
measure, and calculate the results for the demonstration
and the necessary quality control (QC) samples performed
by the vendor.
The time per analysis was determined by dividing the total
amount of time required to perform the analyses by the
number of samples analyzed (197). In the numerator,
sample analysis time included preparation, measurement,
and calculation of results for demonstration samples and
necessary QC samples performed by the vendor. In the
denominator, the total number of analyses included only
demonstration samples analyzed by the vendor, not QC
analyses nor reanalyses of samples.
Downtime that was required or that occurred between
sample analyses as a part of operation and handling was
considered a part of the sample analysis time. Downtime
occurring due to instrument breakage or unexpected
maintenance was not counted in the assessment, but it is
noted in this final report as an additional time. Any
downtime caused by instrument saturation or memory
effect was addressed, based upon its frequency and
impact on the analysis.
Unique time measurements are also addressed in this
report (e.g., if soil samples were analyzed directly, and
sedimentsamples required additional time to dry before the
analyses started, then a statement was made noting that
soil samples were analyzed in X amount of hours, and that
sediment samples required drying time before analysis).
Recorded times were rounded to the nearest 15-minute
interval. The number of vendor personnel used was noted
and factored into the time calculations. No comparison on
time per analysis is made between the vendor and the
referee laboratory.
4.2.1.5 Primary Objective #5: Cost
The following four cost categories were considered to
estimate costs associated with mercury measurements:
Capital costs
Labor costs
Supply costs
Investigation-derived waste (IDW) disposal costs
Although both vendor and laboratory costs are presented,
the calculated costs were not compared with the referee
laboratory. A summary of how each cost category was
estimated for the measurement device is provided below.
The capital cost was estimated based on published
price lists for purchasing, renting, or leasing each field
measurement device. If the device was purchased,
the capital cost estimate did not include salvage value
for the device after work was completed.
The labor cost was based on the number of people
required to analyze samples during the demonstration.
The labor rate was based on a standard hourly rate for
a technician or other appropriate operator. During the
demonstration, the skill level required was confirmed
based on vendor input regarding the operation of the
device to produce mercury concentration results and
observations made in the field. The labor costs were
based on: 1) the actual number of hours required to
complete all analyses, quality assurance (QA), and
reporting; and 2) the assumption that a technician who
worked for a portion of a day was paid for an entire
8-hour day.
The supplycosts were based on any supplies required
to analyze the field and SRM samples during the
demonstration. Supplies consisted of items not
included in the capital category, such as extraction
19
-------
solvent, glassware, pipettes, spatulas, agitators, and
similar materials. The type and quantity of all supplies
broughtto the field and used during the demonstration
were noted and documented.
Any maintenance and repair costs during the
demonstration were documented or provided by the
vendor. Equipment costs were estimated based on
this information and standard cost analysis guidelines
used in the SITE Program.
The IDW disposal costs included decontamination
fluids and equipment, mercury-contaminated soil and
sediment samples, and used sample residues.
Contaminated personal protective equipment (PPE)
normally used in the laboratory was placed into a
separate container. The disposal costs for the IDW
were included in the overall analytical costs for each
vendor.
After all of the cost categories were estimated, the cost per
analysis was calculated. This cost value was based on the
number of analyses performed. As the numberof samples
analyzed increased, the initial capital costs and certain
other costs were distributed across a greater number of
samples. Therefore, the per unit cost decreased. For this
reason, two costs were reported: 1) the initial capital costs
and 2) the operating costs per analysis. No com parison to
the referee laboratory's method cost was made; however,
a generic cost comparison was made. Additionally, when
determining laboratory costs, the associated cost for
laboratory audits and data validation should be considered.
4.2.2 Approach for Addressing Secondary
Objectives
Secondary objectives were evaluated based on
observations made during the demonstration. Because of
the number of vendors involved, technology observers
were required to make simultaneous observations of two
vendors each during the demonstration. Four procedures
were implemented to ensure that these subjective
observations made by the observers were as consistent as
possible.
First, forms were developed for each of the five secondary
objectives. These forms assisted in standardizing the
observations. Second, the observers met each day before
the evaluations began, at significant break periods, and
after each day of work to discuss and compare
observations regarding each device. Third, an additional
observerwas assigned to independently evaluate onlythe
secondary objectives in order to ensure that a consistent
approach was applied in evaluating these objectives.
Finally, the SAIC TOM circulated among the evaluation
staff during the demonstration to ensure that a consistent
approach was being followed by all personnel. Table 4-2
summarizes the aspects observed during the
demonstration for each secondary objective. The
individual approaches to each of these objectives are
detailed further in the following subsections.
Table 4-2. Summary of Secondary Objective Observations Recorded During the Demonstration
SECONDARY OBJECTIVE
General
, , . . Secondary Objective # 1
Information Ease of Use
- Vendor Name - No. of Operators
- Observer Name - Operator Names/Titles
- Instrument Type - Operator Training
- Instrument Name - Training References
- Model No. - Instrument Setup Time
- Serial No. - Instrument Calibration Time
- Sample Preparation Time
- Sample Measurement Time
Secondary Objective # 2
H&S Concerns
- Instrument Certifications
- Electrical Hazards
- Chemicals Used
- Radiological Sources
- Hg Exposure Pathways
- Hg Vapor Monitoring
- PPE Requirements
- Mechanical Hazard
- Waste Handling Issues
Secondary Objective # 3
Instrument Portability
- Instrument Weight
- Instrument Dimensions
- Power Sources
- Packaging
- Shipping & Handling
Secondary Objective # 4
Instrument Durability
- Materials of Construction
- Quality of Construction
- Max. Operating Temp.
- Max. Operating Humidity
- Downtime
- Maintenance Activities
- Repairs Conducted
H&S = Health and Safety
PPE = Personal Protective Equipment
20
-------
4.2.2.1 Secondary Objective #1: Ease of Use
The skills and training required for proper device operation
were noted; these included any degrees or specialized
training required by the operators. This information was
gathered by interviews (i.e., questioning) of the operators.
The number of operators required was also noted. This
objective was also evaluated by subjective observations
regarding the easeof equipment useand major peripherals
required to measure mercury concentrations in soils and
sediments. The operating manual was evaluated to
determine if it is easily useable and understandable.
4.2.2.2 Secondary Objective #2: Health and Safety
Concerns
Health and safety (H&S) concerns associated with device
operation were noted during the demonstration. Criteria
included hazardous materials used, the frequency and
likelihood of potential exposures, and any direct exposures
observed during the demonstration. In addition, any
potential for exposure to mercury during sample digestion
and analysis was evaluated, based upon equipment
design. Other H&S concerns, such as basic electrical and
mechanical hazards, were also noted. Equipment
certifications, such as Underwriters Laboratory (UL), were
documented.
4.2.2.3 Secondary Objective #3: Portability of the
Device
The portability of the device was evaluated by observing
transport, measuring setup and tear down time,
determining the size and weightof the unit and peripherals,
and assessing the ease with which the instrument was
repackaged for movement to another location. The use of
battery power or the need for an AC outlet was also noted.
4.2.2.4 Secondary Objective #4: Instrument Durability
The durability of each device and major peripherals was
assessed by noting the quality of materials and
construction. All device failures, routine maintenance,
repairs, and downtime were documented during the
demonstration. No specific tests were performed to
evaluate durability; rather, subjective observations were
made using a field form as guidance.
4.2.2.5 Secondary Objective #5: Availability of Vendor
Instruments and Supplies
The availability of each device was evaluated by
determining whether additional units and spare parts are
readily available from the vendor or retail stores. The
vendor's office (or a web page) and/or a retail store was
contacted to identify and determine the availability of
supplies of the tested measurement device and spare
parts. This portion of the evaluation was performed after
the field demonstration, in conjunction with the cost
estimate.
4.3 Sample Preparation and Management
4.3.1 Sample Prepara tion
4.3.1.1 Field Samples
Field samples were collected during the pre-demonstration
portion of the project, with the ultimate goal of producing a
set of consistent test soils and sediments to be distributed
among all participating vendors and the referee laboratory
for analysis during the demonstration. Samples were
collected from the following four sites:
Carson River Mercury site (near Dayton, NV)
Y-12 National Security Complex (Oak Ridge, TN)
Manufacturing facility (eastern U.S.)
Puget Sound (Bellingham, WA)
The field samples collected during the pre-demonstration
sampling events comprised a variety of matrices, ranging
from material having a high clay content to material
composed mostly of gravelly, coarse sand. The field
samples also differed with respect to moisture content;
several were collected as wet sediments. Table 4-3 shows
the number of distinct field samples that were collected
from each of the four field sites.
Prior to the start of the demonstration, the field samples
selected for analysis during the demonstration were
processed at the SAIC GeoMechanics Laboratory in Las
Vegas, NV. The specific sample homogenization
procedure used by this laboratory largely depended on the
moisture content and physical consistency of the sample.
Two specific sample homogenization procedures were
developed and tested by SAIC at the GeoMechanics
Laboratory during the pre-demonstration portion of the
project. The methods included a non-slurry sample
procedure and a slurry sample procedure.
A standard operating procedure (SOP) was developed
detailing both methods. The procedure was found to be
satisfactory, based upon the results of replicate samples
during the pre-demonstration. This SOP is included as
Appendix A of the Field Demonstration Quality Assurance
Project Plan (SAIC, August 2003, EPA/600/R-03/053).
Figure 4-1 summarizes the homogenization steps of the
SOP, beginning with sample mixing. This procedure was
21
-------
used for preparing both pre-demonstration and
demonstration samples. Prior to the mixing process (i.e.,
Step 1 in Figure 4-1), all field samples being processed
were visually inspected to ensure that oversized materials
were removed and that there were no clumps that would
hinderhomogenization. Non-slurry samples were air-dried
in accordance with the SOP so that they could be passed
multiple times through a riffle splitter. Due to the high
moisture content of many of the samples, they were not
easily air-dried and could not be passed through a riffle
splitter while wet. Samples with very high moisture
contents, termed "slurries," were not air-dried, and
bypassed the riffle splitting step. The homogenization
steps for each type of matrix are briefly summarized as
follows.
Table 4-3. Field Samples Collected from the Four Sites
No. of Samples / Matrices
Field Site Collected Areas For Collecting Sample Material
Volume Required
Carson River
Y-12
Manufacturing Site
Puget Sound
12 Soil
6 Sediment
10 Sediment
6 Soil
12 Soil
4 Sediment
Tailings Piles (Six Mile Canyon)
River Bank Sediments
Poplar Creek Sediments
Old Mercury Recovery Bldg. Soils
Subsurface Soils
High-Level Mercury (below cap)
Low-Level Mercury (native material)
4 L each for soil
12 L each for sediment
12 L each for sediment
4 L each for soil
4 L each
1 2 L each
Preparing Slurry Matrices
For slurries (i.e., wet sediments), the mixing steps were
sufficiently thorough that the sample containers could be
filled directly from the mixing vessel. There were two
separate mixing steps for the slurry-type samples. Each
slurry was initially mixed mechanically within the sample
container (i.e., bucket) in which the sample was shipped to
the SAIC GeoMechanics Laboratory. A subsample of this
premixed sample was transferred to a second mixing
vessel. A mechanical drill equipped with a paint mixing
attachment was used to mix the subsample. As shown in
Figure 4-1, slurry samples bypassed the sample riffle
splitting step. To ensure all sample bottles contained the
same material, the entire set of containers to be filled was
submerged into the slurryas a group. The filled vials were
allowed to settle for a minimum of two days, and the
standing water was removed using a Pasteur pipette. The
removal of the standing waterfrom the slurry samples was
the only change to the homogenization procedure between
the pre-demonstration and the demonstration.
Preparing "Non-Slurry" Matrices
Soils and sediments having no excess moisture were
initially mixed (Step 1) and then homogenized in the
sample riffle splitter (Step 2). Prior to these steps, the
material was air-dried and subsampled to reduce the
volume of material to a size that was easier to handle.
As shown in Figure 4-1 (Step 1), the non-slurry subsample
was manually stirred with a spoon or similar equipment
until the material was visually uniform. Immediately
following manual mixing, the subsample was mixed and
split six times for more complete homogenization (Step 2).
After the sixth and final split, the sample material was
leveled to form a flattened, elongated rectangle and cut into
transverse sections to fill the containers (Steps 3 and 4).
After homogenization, 20-mL sample vials were filled and
prepared for shipment (Step 5).
For the demonstration, the vendor analyzed 197 samples,
which included replicates of up to 7 samples per sample
lot. The majority of the samples distributed had
concentrations within the range of the vendor's tech no logy.
Some samples had expected concentrations at or below
the estimated level of detection for each of the vendor
instruments. These samples were designed to evaluate
the reported MDL and PQL and also to assess the
prevalence of false positives. Field samples distributed to
the vendor included sediments and soils collected from all
four sites and prepared by both the slurry and dry
homogenization procedures. The field samples were
segregated into broad sample sets: low, medium, and high
mercury concentrations. This gave the vendor the same
general understanding of the sample to be analyzed as
they would typically have for field application of their
instrument.
22
-------
1
Test Material Mixed Until
Visually Uniform
For Non-slurries
Mix manually
For Slurries
a) Mix mechanically the entire
sample volume
b) Subsample slurry, transfer to
mixing vessel, and mix
mechanically
Slurries transferred
directly to 20 ml vials
(vials submerged into slurry)
Non-slurries to
riffle splitter
Combined splits
are reintroduced
into splitter (6 X)
\ /
Transfer cut
sections to
20 mL vials
TEFLON SURFACE
I Elongated
octangular pile
(from 6* split)
Sample aliquots made
by transverse cuts
across sample piles
1
Samples shipped @ 4 °C to
referee lab and Oak Ridge
(Container numbers will vary)
Figure 4-1. Test sample preparation at the SAIC GeoMecharries Laboratory.
23
-------
In addition, selected field samples were spiked with
mercury (II) chloride to generate samples with additional
concentrations and test the ability of the vendor's
instrumentation to measure the additional species of
mercury. Specific information regarding the vendor's
sample distribution is included in Chapter 6.
4.3.1.2 Standard Reference Materials
Certified SRMs were analyzed by both the vendors and the
referee laboratory. These samples were homogenized
matrices which had a known concentration of mercury.
Concentrations were certified values, as provided by the
supplier, based on independent confirmation via multiple
analyses of multiple lots and/or multiple analyses by
different laboratories (i.e., round robin testing). These
analytical results were then used to determine "true"
values, as well as a statistically derived intervals (a 95%
prediction interval) that provided a range within which the
true values were expected to fall.
The SRMs selected were designed to encompass the
same contaminant ranges indicated previously: low-,
medium-, and high-level mercury concentrations. In
addition, SRMs of varying matrices were included in the
demonstration to challenge the vendor technology as well
as the referee laboratory. The referee laboratory analyzed
all SRMs. SRM samples were intermingled with site field
samples and labeled in the same manner as field samples.
4.3.1.3 Spiked Field Samples
Spiked field samples were prepared by the SAIC
GeoMechanics Laboratory using mercury (II) chloride.
Spikes were prepared using field samples from the
selected sites. Additional information was gained by
preparing spikes at concentrations not previously
obtainable. The SAIC GeoMechanics Laboratory's ability
to prepare spikes was tested prior to the demonstration
and evaluated in order to determine expected variability
and accuracyof the spiked sample. The spiking procedure
was evaluated by preparing several different spikes using
two different spiking procedures (dry and wet). Based
upon results of replicate analyses, it was determined that
the wet, or slurry, procedure was the only effective method
of obtaining a homogeneous spiked sample.
4.3.2 Sample Management
4.3.2.1 Sample Volumes, Containers,and Preservation
A subset from the pre-demonstration field samples was
selected for use in the demonstration based on the
sample's mercury concentration range and sample type
(i.e., sediment versus soil). The SAIC GeoMechanics
Laboratory prepared individual batches of field sample
material to fill sample containers for each vendor. Once all
containers from a field sample were filled, each container
was labeled and cooled to 4 °C. Because mercury
analyses were to be performed both by the vendors in the
field and by the referee laboratory, adequate sample size
was taken into account. Minimum sample size
requirements for the vendors varied from 0.1 g or less to
8-10 g. Only the referee laboratory analyzed separate
sample aliquots for parameters otherthan mercury. These
additional parameters included arsenic, barium, cadmium,
chromium, lead, selenium, silver, copper, zinc, oil and
grease, and total organic carbon (TOC). Since the mercury
method (SW-846 7471B) being used by the referee
laboratory requires 1 g for analysis, the sample size sent to
all participants was a 20-mL vial (approximately 10 g),
which ensured a sufficient volume and mass for analysis
by all vendors.
4.3.2.2 Sample Labeling
The sample labeling used for the 20-mL vials consisted of
an internal code developed by SAIC. This "blind" code was
used throughout the entire demonstration. The only
individuals who knew the key to the coding of the
homogenized samples to the specific field samples were
the SAIC TOM, the SAIC GeoMechanics Laboratory
Manager, and the SAIC QA Manager.
4.3.2.3 Sample Record Keeping, Archiving, and
Custody
Samples were shipped to the laboratory and the
demonstration site the week prior to the demonstration. A
third set of vials was archived at the SAIC GeoMechanics
Laboratory as reserve samples.
The sample shipment to Oak Ridge was retained at all
times in the custody of SAIC at their Oak Ridge office until
arrival of the demonstration field crew. Samples were
shipped under chain-of-custody (COC) and with custody
seals on both the coolers and the inner plastic bags. Once
the demonstration crew arrived, the coolers were retrieved
from the SAIC office. The custody seals on the plastic
bags inside the cooler were broken by the vendor upon
transfer.
Upon arrival at the ORNL site, the vendor set up the
instrumentation at the direction and oversight of SAIC. At
the start of sample testing, the vendor was provided with a
sample set representing field samples collected from a
particular field site, intermingled with SRM and spiked
samples. Due to variability of vendor instrument
24
-------
measurement ranges for mercury detection, not all vendors
received samples from the same field material. All
samples were stored in an ice cooler prior to demonstration
startup and were stored in an on-site sample refrigerator
during the demonstration. Each sample set was identified
and distributed as a set with respect to the site from which
it was collected. This was done because, in any field
application, the location and general type of the samples
would be known.
The vendor was responsible for analyzing all samples
provided, performing any dilutions or reanalyses as
needed, calibrating the instrument if applicable, performing
any necessary maintenance, and reporting all results. Any
samples that were not analyzed during the day were
returned to the vendor for analysis at the beginning of the
next day. Once analysis of the samples from the first
location were completed by the vendor, SAIC provided a
set of samples from the second location. Samples were
provided at the time that they were requested by the
vendor. Once again, the transfer of samples was
documented using a COC form.
This process was repeated forsamples from each location.
SAIC maintained custody of all remaining sample sets until
they were transferred to the vendor. SAIC maintained
custody of samples that already had been analyzed and
followed the waste handling procedures in Section 4.2.2 of
the Field Demonstration QAPP to dispose of these wastes.
4.4
Reference
Process
Method Confirmatory
The referee laboratory analyzed all samples that were
analyzed by the vendor technologies in the field. The
following subsections provide information on the selection
of the reference method, selection of the referee
laboratory, and details regarding the performance of the
reference method in accordance with EPA protocols.
Other parameters that were analyzed by the referee
laboratory are also discussed briefly.
4.4.1 Reference Method Selection
The selection of SW-846 Method 7471B as the reference
method was based on several factors, predicated on
information obtained from the technology vendors, as well
as the expected contaminant types and soil/sediment
mercury concentrations expected in the test matrices.
There are several laboratory-based, promulgated methods
for the analysis of total mercury. In addition, there are
several performance-based methods forthe determination
of various mercury species. Based on the vendor
technologies, it was determined that a reference method
for total mercury would be needed (Table 1-2 summarizes
the methods evaluated, as identified through a review of
the EPA Test Method Index and SW-846).
In selecting which of the potential methods would be
suitable as a reference method, consideration was given to
the following questions:
Was the method widely used and accepted? Was the
method an EPA-recommended, or similar regulatory
method? The selected reference method should be
sufficiently used so that it could be cited as an
acceptable method for monitoring and/or permit
compliance among regulatory authorities.
Did the selected reference method provide QA/QC
criteria that demonstrate acceptable performance
characteristics over time?
Was the method suitable for the species of mercury
that were expected to be encountered? The reference
method must be capable of determining, as total
mercury, all forms of the contaminant known or likely
to be present in the matrices.
Would the method achieve the necessary detection
limits to evaluate the sensitivity of each vendor
technology adequately?
Was the method suitable for the concentration range
that was expected in the test matrices?
Based on the above considerations, itwas determined that
SW-846 Method 7471B [analysis of mercury in solid
samples by cold-vapor atomic absorption spectrometry
(AAS)] would be the best reference method. SW-846
method 7474, (an atomic fluorescence spectrometry
method using Method 3052 for microwave digestion of the
solid) had also been considered a likely technical
candidate; however, because this method was not as
widely used or referenced, Method 7471B was considered
the better choice.
4.4.2 Referee Laboratory Selection
During the planning of the pre-demonstration phase of this
project, nine laboratories were sent a statement of work
(SOW) for the analysis of mercury to be performed as part
of the pre-demonstration. Seven of the nine laboratories
responded to the SOW with appropriate bids. Three of the
seven laboratories were selected as candidate laboratories
based upon technical merit, experience, and pricing.
These laboratories received and analyzed blind samples
25
-------
and SRMs during pre-demonstration activities. The referee
laboratory to be used for the demonstration was selected
from these three candidate laboratories. Final selection of
the referee laboratory was based upon: 1) the laboratory's
interest in continuing in the demonstration, 2) the
laboratory-reported SRM results, 3) the laboratory MDL for
the reference method selected, 4) the precision of the
laboratory calibration curve, 5) the laboratory's ability to
support the demonstration (scheduling conflicts, backup
instrumentation, etc.), and 6) cost.
One of the three candidate laboratories was eliminated
from selection based on a technical consideration. It was
determined that this laboratory would not be able to meet
demonstration quantitation limit requirements. (Its lower
calibration standard was approximately 50 ug/kg and the
vendor comparison requirements were well below this
value.) Two candidates thus remained, including the
eventual demonstration laboratory, Analytical Laboratory
Services, Inc. (ALSI):
Analytical Laboratory Services, Inc.
Ray Martrano, Laboratory Manager
34 Dogwood Lane
Middletown, PA 17057
(717)944-5541
In order to make a final decision on selecting a referee
laboratory, a preliminary audit was performed by the SAIC
QA Manager at the remaining two candidate laboratories.
Results of the SRM samples were compared for the two
laboratories. Each laboratory analyzed each sample (there
were two SRMs) in triplicate. Both laboratories were within
the 95% prediction interval for each SRM. In addition, the
average result from the two SRMs was compared to the
95% Cl for the SRM.
Calibration curves from each laboratory were reviewed
carefully. This included calibration curves generated from
previously performed analyses and those generated for
other laboratory clients. There were two QC requirements
regarding calibration curves; the correlation coefficient had
to be 0.995 or greater and the lowest point on the
calibration curve had to be within 10% of the predicted
value. Both laboratories were able to achieve these two
requirements for all curves reviewed and for a lower
standard of 10 ug/kg, which was the lower standard
required for the demonstration, based upon information
received from each of the vendors. In addition, an analysis
of seven standards was reviewed for MDLs. Both
laboratories were able to achieve an MDL that was below
1 ug/kg.
It should be noted that vendor sensitivity claims impacted
how low this lower quantitation standard should be. These
claims were somewhat vague, and the actual quantitation
limit each vendor could achieve was uncertain prior to the
demonstration (i.e., some vendors claimed a sensitivity as
low as 1 ug/kg, but it was uncertain at the time if this limit
was actually a PQLora detection limit). Therefore, it was
determined that, if necessary, the laboratory actually
should be able to achieve even alowerPQL than 10 ug/kg.
For both laboratories, SOPs based upon SW-846 Method
7471B were reviewed. Each SOP followed this reference
method. In addition, interferences were discussed
because there was some concern that organic
interferences may have been present in the samples
previously analyzed by the laboratories. Because these
same matrices were expected to be part of the
demonstration, there was some concern associated with
how these interferences would be eliminated. This is
discussed at the end of this subsection.
Sample throughput was somewhat important because the
selected laboratory was to receive all demonstration
samples at the same time (i.e., the samples were to be
analyzed at the same time in order to eliminate any
question of variability associated with loss of contaminant
due to holding time). This meant that the laboratory would
receive approximately 400 samples for analysis over the
period of a few days. It was also desirable for the
laboratory to produce a data report within a 21-day
turnaround time for purposes of the demonstration. Both
laboratories indicated that this was achievable.
Instrumentation was reviewed and examined at both
laboratories. Each laboratory used a Leeman mercury
analyzer for analysis. One of the two laboratories had
backup instrumentation in case of problems. Each
laboratory indicated that its Leeman mercury analyzer was
relatively new and had not been a problem in the past.
Previous SITE program experience was another factor
considered as partof these pre-audits. This is because the
SITE program generally requires a very high level of QC,
such that most laboratories are not familiar with the QC
required unless they have previously participated in the
program. A second aspect of the SITE program is that it
generally requires analysis of relatively "dirty" samples and
many laboratories are not use to analyzing such "dirty"
samples. Both laboratories have been longtime
participants in this program.
Other QC-related issues examined during the audits
included 1) analyses of otherSRM samples not previously
examined, 2) laboratory control charts, and 3) precision
26
-------
and accuracy results. Each of these issues was closely
examined. Also, because of the desire to increase the
representativeness of the samples for the demonstration,
each laboratory was asked if sample aliquot sizes could be
increased to 1 g (the method requirement noted 0.2 g).
Based upon previous results, both laboratories routinely
increased sample size to 0.5 g, and each laboratory
indicated that increasing the sample size would not be a
problem. Besides these QC issues, other less tangible QA
elements were examined. This included analyst
experience, management involvement in the
demonstration, and internal laboratory QA management.
These elements were also factored into the final decision.
Selection Summary
There were very few factors that separated the quality of
these two laboratories. Both were exemplary in performing
mercury analyses. There were, however, some minor
differences based upon this evaluation that were noted by
the auditor. These were as follows:
ALSI had backup instrumentation available. Even
though neither laboratory reported any problems with
its primary instrument (the Leeman mercury analyzer),
ALSI did have a backup instrument in case there were
problems with the primary instrument, or in the event
that the laboratory needed to perform other mercury
analyses during the demonstration time.
As noted, the low standard requirement for the
calibration curve was one of the QC requirements
specified for this demonstration in order to ensure that
a lower quantitation could be achieved. This low
standard was 10 ug/kg for both laboratories. ALSI,
however, was able to show experience in being able to
calibrate much lower than this, using a second
calibration curve. In the event that the vendor was
able to analyze at concentrations as low as 1 ug/kg
with precise and accurate determinations, ALSI was
able to perform analyses at lower concentrations as
part of the demonstration. ALSI used a second, lower
calibration curve for any analyses required below 0.05
mg/kg. Very few vendors were able to analyze
samples at concentrations at this low a level.
Management practices and analyst experience were
similar at both laboratories. ALSI had participated in a
few more SITE demonstrations than the other
laboratory, but this difference was not significant
because both laboratories had proven themselves
capable of handling the additional QC requirements for
the SITE program. In addition, both laboratories had
internal QA management procedures to provide the
confidence needed to achieve SITE requirements.
Interferences for the samples previously analyzed were
discussed and data were reviewed. ALSI performed
two separate analyses for each sample. This included
analyses with and without stannous chloride.
(Stannous chloride is the reagent used to release
mercury into the vapor phase for analysis. Sometimes
organics can cause interferences in the vapor phase.
Therefore, an analysis with no stannous chloride would
provide information on organic interferences.) The
other laboratory did not routinely perform this analysis.
Some samples were thought to contain organic
interferences, based on previous sample results. The
pre-demonstration results reviewed indicated that no
organic interferences were present. Therefore, while
this was thought to be a possible discriminator
between the two laboratories in terms of analytical
method performance, it became moot for the samples
included in this demonstration.
The factors above were considered in the final evaluation.
Because there were only minor differences in the technical
factors, cost of analysis was used as the discriminating
factor. (If there had been significant differences in
laboratory quality, cost would not have been a factor.)
ALSI was significantly lower in cost than the other
laboratory. Therefore, ALSI was chosen as the referee
laboratory for the demonstration.
4.4.3 Summary of Analytical Methods
4.4.3.1 Summary of Reference Method
The critical measurement for this study was the analysis of
mercury in soil and sediment samples. Samples analyzed
by the laboratory included field samples, spiked field
samples, and SRM samples. Detailed laboratory
proceduresforsubsampling, extraction, and analysis were
provided in the SOPs included as Appendix B of the Field
Demonstration QAPP. These are briefly summarized
below.
Samples were analyzed for mercury using Method 7471 B,
a cold-vapor atomic absorption method, based on the
absorption of radiation at the 253.7-nm wavelength by
mercury vapor. The mercury is reduced to the elemental
state and stripped/volatilized from solution in a closed
system. The mercury vapor passes through a cell
positioned in the light path of the AA spectrophotometer.
Absorbance (peak height) is measured as a function of
mercury concentration. Potassium permanganate is added
27
-------
to eliminate possible interference from sulfide. As per the
method, concentrations as high as 20 mg/kg of sulfide, as
sodium sulfide, do not interfere with the recovery of added
inorganic mercury in reagent water. Copper has also been
reported to interfere; however, the method states that
copper concentrations as high as 10 mg/kg had no effect
on recovery of mercury from spiked samples. Samples
high in chlorides require additional permanganate (as much
as 25 ml_) because, during the oxidation step, chlorides are
converted to free chlorine, which also absorbs radiation at
254 nm. Free chlorine is removed by using an excess (25
ml_) of hydroxylamine sulfate reagent. Certain volatile
organic materials that absorb at this wavelength may also
cause interference. A preliminary analysis without
reagents can determine if this type of interference is
present.
Prior to analysis, the contents of the sample container are
stirred, and the sample mixed prior to removing an aliquot
for the mercury analysis. An aliquot of soil/sediment (1 g)
is placed in the bottom of a biochemical oxygen demand
bottle, with reagent water and aqua regia added. The
mixture is heated in a water bath at 95 °C for 2 minutes.
The solution is cooled and reagent water and potassium
permanganate solution are added to the sample bottle.
The bottle contents are thoroughly mixed, and the bottle is
placed in the water bath for 30 minutes at 95 °C. After
cooling, sodium chloride-hydroxylamine sulfate is added to
reduce the excess permanganate. Stannous chloride is
then added and the bottle attached to the analyzer; the
sample is aerated and the absorbance recorded. An
analysis without stannous chloride is also included as an
interference check when organic contamination is
suspected. In the event of positive results of the non-
stannous chloride analysis, the laboratory was to report
those results to SAIC so that a determination of organic
interferences could be made.
4.4.3.2 Summary of Methods for Non-Critical
Measurements.
A selected set of non-critical parameters was also
measured during the demonstration. These parameters
were measured to provide a better insight into the chemical
constituency of the field samples, including the presence of
potential interferents. The results of the tests for potential
interferents were reviewed to determine if a trend was
apparent in the event that inaccuracy or low precision was
observed. Table 4-4 presents the analytical method
reference and method type for these non-critical
parameters.
Table 4-4. Analytical Methods for Non-Critical Parameters
Parameter Method Reference Method Type
Arsenic, barium,
cadmium,
chromium, lead,
selenium, silver,
copper, and zinc
SW-846 3050/6010 Acid digestion, ICP
Oil and Grease
TOC
Total Solids
EPA 1664
SW-846 9060
EPA 2540G
n-Hexane
extraction,
Gravimetric
analysis
Carbonaceous
analyzer
Gravimetric
4.5
Deviations
Plan
from the Demonstration
Three deviations to the demonstration plan occurred. The
first was that the demonstration was to be conducted in
Oak Ridge, TN, on May 5-8, 2003. On the morning of May
6, MTI was able to confirm, through the analyses of
standards, that the obtained results were erroneous.
While several possibilities existed as to why erroneous
results were being generated, the exact reason could not
be determined and corrected in time to complete the
analyses during the scheduled demonstration. Several
discussions ensued between the EPA TOM, SAIC
personnel on-site, and MTI personnel. These discussions
resulted in an agreement that if the cause of the erroneous
results could be identified and a new location and date
could be arranged, then MTI would be given a second
attempt at the demonstration.
During the following weeks, MTI traced the source of the
problem to disposable beakers that were used in the
analyses. The beakers had an oil film in them that
gradually coated the electrodes, preventing accurate
readings. In order to correct the problem, MTI disposed of
the beakers in stock, and added to the operating procedure
a sodium hydroxide rinse between samples that prevented
the buildup of residues on the electrodes.
While in Oak Ridge, MTI only received samples from the
Oak Ridge sampling site. None of the results from this first
demonstration are presented ordiscussed in this ITVR. All
samples were collected from MTI and returned to the SAIC
GeoMechanics Laboratoryforstorage purposes. Samples
28
-------
that were opened by MTI were replaced with unopened
sample bottles held in reserve by SAIC, when available.
The second demonstration for MTI was held in Las Vegas,
NV, June 14-17, 2003.
A second and third deviation occurred in the distribution of
the samples (second deviation) and an unforeseen and
unrelated emergency (third deviation). The intent in the
demonstration plan was to give the samples from one
sampling site to the vendors and, when that sample set
was completed, the next sample set would be given to the
vendor. This process was repeated until all four sample
sets were completed by the vendor. MTI's two personnel
on-site divided the workload into two parts. The first
individual extracted the samples and the second individual
analyzed the samples with the PDV 6000. When the
extraction of the first set of samples was completed, the
second set of samples was given to MTI for extraction.
The extraction was faster than the analysis; therefore, the
first individual completed all of the extractions in two days.
The second individual completed the analyses in three
days. This was complicated by an unrelated emergency,
which prevented any work from being performed on the
third day at the site; therefore, no sample preparation or
analyses were performed on June 16. All samples were
prepared June 14-15. Sample analyses were completed
June 14, 15, and 17. MTI personnel are unsure as to
whether the one day delay would have any effect on
results; however, it is common practice to analyze low
concentration mercury samples as soon as possible upon
preparation.
29
-------
Chapter 5
Assessment of Laboratory Quality Control Measurements
5.1 Laboratory QA Summary
QA may be defined as a system of activities, the purpose
of which is to provide assurance that defined standards of
quality are met with a stated level of confidence. A QA
program is a means of integrating the quality planning,
quality assessment, QC, and quality improvement efforts
to meet user requirements. The objective of the QA
program is to reduce measurement errors to agreed-upon
limits, and to produce results of acceptable and known
quality. The QAPP specified the necessary guidelines to
ensure that the measurement system for laboratory
analysis was in control, and provided detailed information
on the analytical approach to ensure that data of high
quality could be obtained to achieve project objectives.
The laboratory analyses were critical to project success, as
the laboratory results were used as a standard for
comparison to the field method results. The field methods
are of unknown quality, and therefore, for comparison
purposes the laboratory analysis needed to be a known
quantity. The following sections provide information on the
use of data quality indicators, and a detailed summary of
the QC analyses associated with project objectives.
5.2 Data Quality Indicators for Mercury
Analysis
To assess the quality of the data generated by the referee
laboratory, two important data quality indicators of primary
concern are precision and accuracy. Precision can be
defined as the degree of mutual agreement of independent
measurements generated through repeated application of
the process under specified conditions. Accuracy is the
degree of agreement of a measured value with the true or
expected value. Both accuracy and precision were
measured by the analysis of matrix spike/matrix spike
duplicates (MS/MSDs). The precision of the spiked
duplicates is evaluated by expressing, as a percentage, the
difference between results of the sample and sample
duplicate results. The relative percent difference (RPD) is
calculated as:
(Maximum Value - Minimum Value)
(Maximum Value -(-Minimum Value)/2
To determine and evaluate accuracy, known quantities of
the target analytes were spiked into selected field samples.
All spikes were post-digestion spikes because of the high
sample concentrations encountered during the
demonstration. Pre-digestion spikes, on high-
concentration samples would either have been diluted or
would have required additional studies to determine the
effect of spiking more analyte and subsequent recovery
values. To determine matrix spike recovery, and hence
measure accuracy, the following equation was applied:
%R=
C
x100
where,
Css = Analyte concentration in spiked
sample
Cus = Analyte concentration in unspiked
sample
Csa = Analyte concentration added to
sample
Laboratory control samples (LCSs) were used as an
additional measure of accuracy in the event of significant
30
-------
matrix interference. To determine the percent recovery of
LCS analyses, the equation below was used:
„,,_. Measured Concentration .__
%R = xi 00
Theoretical Concentration
While several precautions were taken to generate data of
known quality through control of the measurement system,
the data must also be representative of true conditions and
comparable to separate sample aliquots.
Representativeness refers to the degree with which
analytical results accurately and precisely reflect actual
conditions present at the locations chosen for sample
collection. Representativeness was evaluated as part of
the pre-demonstration and combined with the precision
measurement in relation to sample aliquots. Sample
aliquoting by the SAIC GeoMechanics Laboratory tested
the ability of the procedure to produce homogeneous,
representative, and comparable samples. All samples
were carefully homogenized in order to ensure
comparability between the laboratory and the vendor.
Therefore, the RSD measurement objective of 25% or less
for replicate sample lotanalysis was intended to assess not
only precision but representativeness and comparability.
Sensitivity was another critical factor assessed for the
laboratory method of analysis. This was measured as a
practical quantitation limit and was determined by the low
standard on the calibration curve. Two separate calibration
curves were run by the laboratory when necessary. The
higher calibration curve was used for the majority of the
samples and had a lower calibration limit of 25 ug/kg. The
lower calibration curve was used when samples were
below this lowercalibration standard. The lowercalibration
curve had a lower limit standard of 5 ug/kg. The lower limit
standard of the calibration curve was run with each sample
batch as a check standard and was required to be within
10% of the true value (QAPP QC requirement). This
additional check on analytical sensitivity was performed to
ensure that this lower limit standard was truly
representative of the instrument and method practical
quantitation limit.
5.3 Conclusions and Data Quality
Limitations
Critical sample data and associated QC analyses were
reviewed to determine whether the data collected were of
adequate quality to provide proper evaluation of the
project's technical objectives. The results of this review
are summarized below.
Accuracy objectives for mercury analysis by Method 7471B
were assessed by the evaluation of 23 spiked duplicate
pairs, analyzed in accordance with standard procedures in
the same manner as the samples. Recovery values for the
critical compounds were well within objectives specified in
the QAPP, except for two spiked samples summarized in
Table 5-1. The results of these samples, however, were
only slightly outside specified limits, and given the number
of total samples (46 or 23 pairs), this is an insignificant
number of results that did not fa II within specifications. The
MS/MSD results therefore, are supportive of the overall
accuracy objectives.
Table 5-1. MS/MSD Summary
Parameter Value
QC Limits
Recovery Range
Number of Duplicate Pairs
Average Percent Recovery
No. of Spikes Outside QC
Specifications
80%- 120%
85.2%- 126%
23
108%
An additional measure of accuracywas LCSs. These were
analyzed with every sample batch (1 in 20 samples) and
results are presented in Table 5-2. All results were within
specifications, thereby supporting the conclusion that QC
assessment met project accuracy objectives.
Table 5-2. LCS Summary
Parameter
QC Limits
Recovery Range
Number of LCSs
Average Percent Recovery
No. of LCSs Outside QC
Specifications
Value
90%- 110%
90% -100%
24
95.5%
0
Precision was assessed through the analysis of 23
duplicate spike pairs for mercury. Precision specifications
were established prior to the demonstration as a RPD less
31
-------
than 20%. All but two sample pairs were within
specifications, as noted in Table 5-3. The results of these
samples, however, were only slightly outside specified
limits, and given the number of total samples (23 pairs),
this is an insignificant number of results that did not fall
within specifications. Therefore, laboratory analyses met
precision specifications.
Table 5-3. Precision Summary
Parameter Value
QC Limits
MS/MS D RPD Range
Number of Duplicate Pairs
Average MS/MSD RPD
No. of Pairs Outside QC
Specifications
RPD<
0.0%
23
5.7%
2
20%
to 25%
Sensitivity results were within specified project objectives.
The sensitivity objective was evaluated as the PQL, as
assessed by the low standard on the calibration curve. For
the majority of samples, a calibration curve of 25-500 ug/kg
was used. This is because the majority of samples fell
within this calibration range (samples often required
dilution). There were, however, some samples below this
range and a second curve was used. The calibration range
for this lower curve was 5-50 ug/kg. In order to ensure that
the lower concentration on the calibration curve was a true
PQL, the laboratory ran a low check standard (lowest
concentration on the calibration curve) with every batch of
samples. This standard was required to be within 10% of
the specified value. The results of this low check standard
are summarized in Table 5-4.
Table 5-4. Low Check Standards
Parameter Value
QC Limits
Recovery Range
Number of Check Standards
Analyzed
Average Recovery
Recovery 90% - 110%
88.6%-111%
23
96%
There were a few occasions where this standard did not
meet specifications. The results of these samples,
however, were only slightly outside specified limits, and
given the number of total samples (23), this is an
insignificant number of results that did not fall within
specifications. In addition, the laboratory reanalyzed the
standard when specifications were not achieved, and the
second determination always fell within the required limits.
Therefore laboratory objectives for sensitivity were
achieved according to QAPP specifications.
As noted previously, comparabilityand representativeness
were assessed through the analysis of replicate samples.
Results of these replicates are presented in the discussion
on primary project objectives for precision. These results
show that data were within project and QA objectives.
Completeness objectives were achieved for the project. All
samples were analyzed and data were provided for 100%
of the samples received by the laboratory. No sample
bottles were lost or broken.
Other measures of data quality included method blanks,
calibration checks, evaluation of linearity of the calibration
curve, holding time specifications, and an independent
standard verification included with each sample batch.
These results were reviewed for every sample batch run by
ALSI, and were within specifications. In addition, 10% of
the reported results were checked against the raw data.
Raw data were reviewed to ensure that sample results
were within the calibration range of the instrument, as
defined by the calibration curve. A 6-point calibration curve
was generated at the start of each sample batch of 20. A
few data points were found to be incorrectly reported.
Recalculations were performed for these data, and any
additional data points that were suspected outliers were
checked to ensure correct results were reported. Veryfew
calculation or dilution errors were found. All errors were
corrected so that the appropriate data were reported.
Another measure of compliance were the non-stannous
chloride runs performed by the laboratory for every sample
analyzed. This was done to check for organic interference.
There were no samples that were found to have any
organic interference by this method. Therefore, these
results met expected QC specifications and data were not
qualified in any fashion.
Total solids data were also reviewed to ensure that
calculations were performed appropriatelyand dry weights
reported when required. All of these QC checks met
32
-------
QAPP specifications. In summary, all data quality
indicators and QC specifications were reviewed and found
to be well within project specifications. Therefore, the data
are considered suitable for purposes of this evaluation.
5.4 Audit Findings
The SAIC SITE QA Manager conducted audits of both field
activities and of the subcontracted laboratory as part of the
QA measures for this project. The results of these
technical system reviews are discussed below.
The field audit resulted in no findings or non-
conformances. The audit performed at the subcontract
laboratory was con ducted during the time of project sample
analysis. One non-conformance was identified and
corrective action was initiated. It was discovered that the
laboratory PQL was not meeting specifications due to a
reporting error. The analyst was generating the calibration
curves as specified above; however, the lower limit on the
calibration curve was not being reported. This was
immediately rectified and no other findings or non-
conformances were identified.
33
-------
Chapter 6
Performance of the PDV 6000
MTI analyzed samples on May 5-6, 2003 in Oak Ridge, TN.
As discussed in Section 4.5, on the morning of May 6, MTI
determined through the analyses of standards that the
results they were obtaining were erroneous. A second
demonstration was performed in Las Vegas from June 14-
17, 2003, and it proved successful in accomplishing the
specified objectives. The observations conducted as part
of this second attempt at the demonstration were reviewed,
and the primary and secondary objectives were completed.
The results of the primary and secondary objectives,
identified in Chapter 1, are discussed in Sections 6.1 and
6.2, respectively.
Due to an unrelated emergency during the second
demonstration, analysis activities were not able to be
performed on June 16. For this reason, samples that were
analyzed on June 17 had been extracted on June 15, and
stored in plastic bottles. This may have resulted in a
reduced mercury concentration in the extract, especially for
low concentration samples. The extracts that remained
for analysis on June 17 were from low concentration
samples. There was, however, no additional study
performed to determine if time elapsed between extraction
and analysis influenced mercury concentrations, and it is
believed that this short time period was not significant in
the overall evaluation of results.
The distribution of the samples prepared for MTI and the
referee laboratory is presented in Table 6-1. From the four
sites, MTI received samples at 36 different concentrations
for a total of 197 samples. These 197 samples consisted
of 22 concentrations in replicates of 7,13 concentrations in
replicates of 3, and 1 concentration in a replicate of 4.
Although all these samples were analyzed by MTI, a few
samples were not used as part of the evaluation. Some
sample results were judged invalid by MTI field personnel
and some were found to be below MTI's detection limit, as
explained in more detail in the following sections.
Table 6-1. Distribution of Samples Prepared for MTI and the Referee Laboratory
Site
Concentration Range
Soil
Sample Type
Sediment Spiked Soil
SRM
Carson River
(Subtotal = 48)
Puget Sound
(Subtotal = 51)
Oak Ridge
(Subtotal = 54)
Manufacturing
(Subtotal = 44)
Subtotal
(Total = 197)
Low(1-500ppb)
Mid (0.5-50 ppm)
Hiqh(50->1,000ppm)
Low (1 ppb - 10 ppm)
Hiqh (10-500 ppm)
Low (0.1 -10 ppm)
Hiqh (10-800 ppm)
General (5-1,000 ppm)
3
0
0
13
0
0
13
23
52
10
0
0
0
10
3
10
0
33
7
0
7
7
7
0
0
7
35
0
7
14
3
11
14
14
14
77
34
-------
6.1 Primary Objectives
6.1.1 Sensitivity
Sensitivity objectives are explained in Chapter 4. The two
primary sensitivity evaluations performed for this
demonstration were the MDL and PQL. Determinations of
these two measurements are explained in the paragraphs
below, along with a comparison to the referee laboratory.
These determinations setthe standard forthe evaluation of
accuracy and precision for the MTI field instrument. Any
sample analyzed by MTI and subsequently reported as
below their level of detection was not used as part of any
additional evaluations. This was done because of the
expectation that values below the lower limit of instrument
sensitivity would not reflect the true instrument accuracy
and precision.
The sensitivity measurements of MDL and PQL are both
dependent upon the matrix and method. Hence, the MDL
and PQL will vary, depending upon whetherthe matrix is a
soil, waste, or water. Only soils and sediments were tested
during this demonstration and, therefore, MDL calculations
for this evaluation reflect soil and sediment matrices. PQL
determinations are not independent calculations, but are
dependent upon results provided by the vendor for the
samples they tested.
Comparison of the MDL and PQL to laboratory sensitivity
required that a standard evaluation be performed for all
instruments tested during this demonstration. PQL, as
previously noted, is defined in EPA G-5i as the lowest level
of method and instrument performance with a specified
accuracyand precision. This is often defined by the lowest
point on the calibration curve. Our approach was to let the
vendor provide the lower limit of quantitation as determined
by their particular standard operating procedure, and then
test this limit by comparing results of samples analyzed at
this low concentration to the referee laboratory results, or
comparing the results to a standard reference material, if
available. Comparison of these data are, therefore,
presented for the lowest concentration sample results, as
provided by the vendor. If the vendor provided "non-detect"
results, then no formal evaluation of that sample was
presented. In addition, the sample(s) was not used in the
evaluation of precision and accuracy.
Method Detection Limit - The standard procedure for
determining MDLs is to analyze a low standard or
reference material seven times, calculate the standard
deviation, and multiply the standard deviation by the "t"
value forseven measurements at the 99th percentile (alpha
= 0.01). (This value is 3.143, as determined from a
standard statistics table.) This procedure for determination
of an MDL is defined in 40 CFR Part 136, and while
determinations for MDLs may be defined differently for
other instruments, this method was previously noted in the
demonstration QAPP and is intended to provide a
comparison to other MDL evaluations. The purpose is to
provide a lower level of detection with a statistical
confidence at which the instrument will detect the presence
of a substance above its noise level. There is no
associated accuracy or precision determined or implied.
Several blind standards and field samples were provided to
MTI at their estimated lower limit of sensitivity. The MTI
lower limit of sensitivity was previously estimated at 0.100
mg/kg. Because there a re several different SRM sand fie Id
samples at concentrations close to the MDL, evaluation of
the MDL was performed using more than a single
concentration. Samples chosen forcalculation were based
upon: 1) concentration and how close it was to the
estimated MDL, 2) number of analyses performed for the
same sample (e.g., more than 4), and 3) if non-detects
were reported by MTI for a sample used to calculate the
MDL. Then the next highest concentration sample was
selected based upon the premise that a non-detect result
reported for one of several samples indicates the selected
sample is on the "edge" of the instruments detection
capability.
Seven replicates were analyzed by MTI of a sample for
which the laboratory reported an average concentration of
0.734 mg/kg. (Sample lot 57 from the Puget Sound site.)
The average concentration reported by MTI forth is sample
was 1.58 mg/kg, and the standard deviation was 1.17
mg/kg. An SRM with a reference value of 0.62 mg/kg
(sample lot 38) was analyzed seven times by MTI, with a
reported average concentration of 2.71 mg/kg and a
standard deviation of 0.632 mg/kg. Calculations of the
respective MDLs based upon each of these standards are
3.67 and 1.99 mg/kg.
As a further check of the MDL, both MTI and the referee
laboratory analyzed sample lot 56 from the Carson River
samples. The referee laboratory reported an average
concentration of 0.231 mg/kg, and MTI reported "non-
detect" for 6 of 7 replicates analyzed. This confirms that
the MTI field instrument sensitivity is above this sample
concentration. The referee laboratory reported a
concentration of 0.811 mg/kg for sample lot 11 (from the
Puget Sound samples), while MTI reported an average
concentration of 3.15 mg/kg for three replicate analyses,
with a standard deviation of 1.96 mg/kg. Therefore, it
appears that the method detection limit for this instrument
35
-------
is between 0.231 and 0.811 mg/kg. Also, based on results
from sample lots 57 and 38 (noted above in the MDL
calculation), there is additional evidence that sample
concentrations in this range can be detected but not
accurately quantitated by the MTI field instrument. The
referee laboratory reported an average value of 0.06 mg/kg
for sample lot 2 from the Puget Sound site, while the MTI
average value for seven separate results was 1.61 mg/kg,
with a standard deviation of 0.534 mg/kg. While this is 2
orders of magnitude above the value reported by the
referee laboratory, it does indicate that, for this sample,
MTI was able to detect a much lower concentration than
indicated by the MDL calculations noted above. Therefore,
MTI instrumentation is sometimes able to detect lower
concentrations of mercury, but this appears to be
inconsistent based on the results for sample lot 56 (referee
laboratory measured value of 0.231 mg/kg) that resulted
primarily in reported data as "non-detects." This apparent
difference in the MDL is probably attributable to the fact
that, when analyzing samples on the PDV 6000, the
operator selects a concentration, based upon the expected
sample set concentration range, and prepares and
analyzes a standard at the selected concentration.
The MDL is between 1.67 and 3.67 mg/kg. This
calculation, however, did not prove accurate for
instrument/method sensitivity. As noted by the results
above, the MTI field instrument was able to detect
concentrations well below this calculated value. It should
be noted, however, that in one sample MTI detected
mercury at 0.06 mg/kg (as determined by the referee
laboratory result), but in another sample it was unable to
detect mercury at 0.231 mg/kg. The reason for this
discrepancy is unknown. It should be concluded, however,
that values below 0.811 mg/kg may or may not be detected
by the MTI field instrument and that these values, if
detected, would likely be highly inaccurate and should only
be considered as a "positive hit" and do not represent a
value that is close to the true concentration.
Practical Quantitation Limit - This value is usually
calculated by determining a low standard on the instrument
calibration curve, and it is estimated as the lowest standard
at which the instrument will accurately and precisely
determine a given concentration within specified QC limits.
The PQL is often around 5-10 times the MDL. This PQL
estimation, however, is method- and matrix- dependent. In
order to determine the PQL, several low standards were
provided to MTI and subsequent %Ds were calculated.
The lower limit of sensitivity previously provided by the
vendor (0.10 mg/kg) appears to be an inaccurate estimate
of instrument sensitivity, but, as noted above, sometimes
the MTI field instrument may be able to detect values close
to this lower concentration. If one considers the MDL
around 0.8 mg/kg, then the PQL could be between 4 and
8 mg/kg. This, however, is only an estimate. The
relationship between sensitivity and precision is such that
the lower the concentration, the higher the variation in
reported sample results. The PQL should have a precision
and accuracy that match the instrument capabilities within
a certain operating range of analysis.
Values in the range between 4 and 8 mg/kg were chosen
for estimating the PQL and associated %D between the
MTI reported average and the reference value if it is an
SRM or the average value reported by the referee
laboratory. Also compared are the 95% CIs for additional
descriptive information. In addition, values below the
estimated value of 4 mg/kg are included to determine if the
instrument capabilities can provide an even lower PQL.
Sample lot 14 (Oak Ridge site) has an average value of
4.75 mg/kg reported by the referee laboratory, with a
standard deviation of 1.31 mg/kg. The 95% Cl for this
sample is 3.38 to 6.12 mg/kg. The MTI average value is
6.95 mg/kg, which is just outside the range of the 95% Cl.
The %D between this value and that obtained from the
referee laboratory is 46.3%.
Sample lot 57 (Puget Sound) has an average value of
0.734 mg/kg reported by the referee laboratory, with a
standard deviation of 0.119 mg/kg. The 95% Cl for this
sample is 0.624 to 0.844 mg/kg. The MTI average value
is 1.58 mg/kg, which is outside the range of the 95% Cl.
The %D between this value and that of the referee
laboratory is 115%.
These results suggest that the instrument PQL is 4-8
mg/kg. Given the information associated with the MDL
determination, the PQL is probably above the MDL range
noted previously as between 1.67 and 3.67 mg/kg.
Sensitivity Summary
The low standard calculations suggest that a PQL for the
MTI field instrument is 4-8 mg/kg. The referee laboratory
PQL confirmed during the demonstration is 0.005 mg/kg.
The %D for the average MTI result for the sample
concentration of 4.75 mg/kg is 46%. The range for the
calculated MDL is between 1.67 and 3.67 mg/kg, based on
the results of seven replicate analyses for low standards.
The equivalent MDL for the referee laboratory is 0.0026
mg/kg. The MDL determination, however, is only a
statistical calculation that has been used in the past by
EPA and is currently not considered a "true" MDL by
36
-------
SW-846 methodology. SW-846 is suggesting that
performance-based methods be used and that PQLs be
determined using low standard calculations.
More definitive information for the PQL is unavailable
because additional standards in the calculated range of the
PQL are not available. Vendor estimations of the PQL
were actually much lower than determined during the
demonstration; therefore, several lower standards below
Mil's PQL were provided for analysis (see accuracy tables
below) and subsequently, additional standards orsamples
in the actual PQL range were not tested.
6.1.2 Accuracy
Accuracy is the instrument measurement compared to a
standard or true value. For this demonstration, three
separate standards were used for determining accuracy.
The primary standard is SRMs. The SRMs are traceable
to national systems. These were obtained from reputable
suppliers with reported concentration and an associated
95% Cl and 95% prediction interval. The Cl from the
reference material is used as a measure of comparison
with the Cl calculated from replicate analyses for the same
sample analyzed by the laboratory or vendor. Results are
considered comparable if CIs of the SRM overlap with the
CIs computed from the replicate analyses by the vendor.
While this is not a definitive measure of comparison, it
provides some assurance that the two values are
equivalent.
Prediction intervals are intended as a measure of
comparison for a single laboratory or vendor result with the
SRM. When computing a prediction interval, the equation
assumes an infinite number of analyses, and it is used to
compare individual sample results. A 95% prediction
interval would, therefore, predict the correct result from a
single analysis 95% of the time for an infinite number of
samples, if the result is comparable to that of the SRM. It
should be noted that the corollary to this statement is that
5% of the time a result will be outside the prediction interval
if determined for an infinite number of samples. If several
samples are analyzed, the percentage of results within the
prediction interval will be slightly above or below 95%. The
more samples analyzed, the more likely the percentage of
correct results will be close to 95% if the result for the
method being tested is comparable to the SRM.
All SRMs were analyzed in replicates of three or seven by
both the vendor and by the referee laboratory. In some
instances, analyses performed by the vendor were
determined to be invalid measurements and were,
therefore, not included with the reported results. There
were 9 different SRMs analyzed by both the vendorand the
laboratory, for a total of 53 data points by the vendor and
63 data points by the laboratory. One SRM (sample lot 44)
was not included because 6 of 7 analyses performed by the
vendor were judged to be invalid, and this would not
provide a sufficient number of analyses for comparison.
The second accuracy determination used a comparison of
vendor results of field samples and SRMs to the referee
laboratory results for these same sam pies. Field samples
were used to ensure that "real-world" samples were tested
for each vendor. The referee laboratory result is
considered as the standard for comparison to the vendor
result. This comparison is in the form of a hypothesis test
with alpha = 0.01. (Detailed equations along with additional
information about this statistical comparison is included in
Appendix B.)
It should be noted that a laboratory bias is evident. This
bias was determined by comparing average laboratory
values to SRM reference values and is discussed below.
The laboratory bias is low in comparison to the reference
value. A bias correction was not made when comparing
individual samples (replicate analyses) between the
laboratory and vendor; however, setting alpha = 0.01 helps
mitigate for this possible bias by widening the range of
acceptable results between the two data sets.
An aggregate analysis, or unified hypothesis test, was also
performed for all 32 sample lots. (A detailed discussion of
this statistical comparison is included in Appendix B.) This
analysis provides additional statistical evidence in relation
to the accuracy evaluation. A bias term is included in this
calculation in order to account for the laboratory data bias
previously noted.
The third measure of accuracy is obtained by the analysis
of spiked field samples. These were analyzed by the
vendor and the laboratory in replicate in order to provide
additional measurement comparisons and are treated in
the same manner as the other field samples. Spikes were
prepared to cover additional concentrations not available
from SRMs or field samples. There is no comparison to
the spiked concentration, only a comparison between the
vendor and the laboratory reported value.
The purpose for SRM analyses by the referee laboratory is
to provide a check on laboratory accuracy. During the
pre-demonstration, the referee laboratory was chosen, in
part, based upon the analysis of SRMs. This was done in
order to ensure that a competent laboratory would be used
for the demonstration. The pre-demonstration laboratory
qualification showed that the laboratory was within
37
-------
prediction intervals for all SRMs analyzed. Because of the
need to provide confidence in laboratory analysis during the
demonstration, the referee laboratory also analyzed SRMs
as an ongoing check of laboratory bias. As noted in Table
6-3, not all laboratory results were within the prediction
interval. This is discussed in more detail below. All
laboratory QC checks, however, were found to be within
compliance (see Chapter 5).
Evaluation of vendor and laboratory analysis of SRMs is
performed in the following manner. Accuracy was
determined by comparing the 95% Cl of the sample
analyzed by the vendor and laboratory to the 95% Cl for
the SRM. (95% CIs around the true value are provided by
the SRM supplier.) This is provided in Tables 6-2 and 6-3,
with notations when the CIs overlap, suggesting
comparable results. In addition, the number of SRM
results for the vendor's analytical instrumentation and the
referee laboratory that are within the associated 95%
prediction interval are reported. This is a more definitive
evaluation of laboratory and vendor accuracy. The
percentage of total results within the prediction interval for
the vendor and laboratory are reported in Tables 6-2 and
6-3, respectively.
Table 6-2. MTI SRM Comparison
Sample
No.
Lot SRM Value/
95% Cl
MTI Avg./95%CI
Cl Overlap
(yes/no)
No. of 95% Prediction
Samples Interval
Analyzed
a
b
37
35
48
50
38
53
54
49
52
0.158/0.132-0.184
0.017/0.010-0.024
77.8/71.5-84.0
203 / 183-223"
0.62/0.61 -0.63b
910/821 -999b
1120/1009- 1231 b
99.8/81.9- 118
608 / 490- 726 b
Total Samples
% of samples w/in
prediction interval
2.12/0.767-3.47
1.11/0-2.87
62.6/20.9-104
188 / 121 -255
2.71 / 2.05 -3.37
323 / 253 - 393
733/511 -955
123/90.8- 155
390/270-510
Prediction interval is estimated based upon n=30. A 95% Cl
Cl is estimated based upon n=30.
A 95% prediction interval
no
yes
yes
yes
no
no
no
ves
ves
was provided
was provided
6
3
3
7
6
7
7
7
7
53
by the SRM
by the SRM
0
0
45.6
97.4
0.54
437
582
31.3
292
- 0.357
-0.0358s
-110
-308
-0.70
- 1380
- 1701
- 168
-924
MTI No. w/in
Prediction
Interval
0
1
2
7
0
1
5
7
5
28
53%
supplier but no prediction interval was given.
supplier but no
Cl was given.
Table 6-3. ALSI SRM Comparison
Sample
No.
a
b
37
35
48
50
38
53
54
49
52
Lot SRM Value/
95% Cl
0.158/0.132-0.184
0.017/0.010-0.024
77.8/71.5-84.0
203/183-223b
0.62/0.61 -0.636
910/821 -999b
1120/ 1009- 1231 b
99.8/81.9- 118
608 / 490 - 726 b
Total Samples
% of samples w/in
prediction interval
ALSI Avg./ 95% Cl
0.139/0.093-0.185
Cl No. of 95% Prediction
Overlap Samples Interval
(ves/no) Analyzed
yes
0.00874/0.00782-0.00966 no
82.9/56.7- 109
167 / 140- 194
0.533 / 0.502 - 0.564
484 / 325 - 643
71 1 / 552 - 870
84.2/74.5-93.9
424/338-510
Prediction interval is estimated based upon n=30. A 95% Cl
Cl is estimated based upon n=30.
A 95% prediction interval
yes
yes
no
no
no
ves
ves
was provided
was provided
7
7
7
7
7
7
7
7
7
63
by the SRM
by the SRM
0-0.357
0 - 0.0358 "
45.6- 110
97.4 - 308
0.54 - 0.70
437 - 1380
582 - 1701
31.3- 168
292 - 924
supplier but no prediction
ALSI No. w/in
Prediction
Interval
7
7
5
7
7
4
5
7
7
56
89%
interval was given.
supplier but no Cl was given.
38
-------
The single most importantnumberfrom these tables is the
percentage of samples within the 95% prediction interval.
As noted for the MTI data, this percentage is only 53% with
n = 53. As seen from the tabulated data, average results
for MTI fall evenly both above and below the reference
value. (Four results are above and five results are below.)
This would suggest that there is no particular high or low
bias.
SRM values of 0.017, 0.158, and 0.62 mg/kg show a
percent difference between the MTI result and the
reference value greater than 300%, indicating that MTI
results in this range are not accurate. The sensitivity
evaluation suggested that the MTI PQL may be
somewhere between 4 and 8 mg/kg. If these results are
not included in the accuracy evaluation, 27 of 38, or 71%,
of the MTI results are within the prediction interval.
(Considerably greater than the 53% noted for all results.)
Also, 25 of these 38 samples, or 66%, are within 50% of
the SRM reference value, and 16 of these 38 samples or
Table 6-4. Accuracy Evaluation by Hypothesis Testing
42% are within 30% of the SRM value. Therefore, there is
a correlation with accuracy and concentration, at least for
the very low concentrations.
The percentage of samples within the 95% prediction
interval for the laboratory data is 89% with n = 63. For 8 of
the 9 different SRMs, ALSI average results are below the
reference value. This would suggest that the ALSI data are
biased low. Because of this bias, the percentage of
samples outside the prediction interval is slightly below the
anticipated number of results, given that the number of
samples analyzed (63) is relatively high. Nonetheless, the
referee laboratory data should be considered accurate and
not significantly different from the SRM value. Because
there is no bias correction term in the individual hypothesis
tests (Table 6-4), alpha is set at 0.01 to help mitigate for
laboratory bias. This in effect widens the scope of vendor
data that would fall within an acceptable range of the
referee laboratory.
Sample Lot No./ Site
14/ Oak Ridge
MTI
ALSI
21/ Oak Ridge
MTI
ALSI
221 Oak Ridge
MTI
ALSI
241 Oak Ridge
MTI
ALSI
261 Oak Ridge
MTI
ALSI
31/Oak Ridge
MTI
ALSI
37/Oak Ridge
MTI
ALSI
67/Oak Ridge
MTI
ALSI
02/Puget Sound
MTI
ALSI
05/Puget Sound
MTI
ALSI
11/Puget Sound
MTI
ALSI
Avg. Cone.
mg/kg
6.95
4.75
38.3
11.2
65.9
81.6
212
221
147
77.0
1211
947
2.12
0.14
1203
835
1.61
0.06
1.87
0.21
3.15
0.81
RSD or CV
37.5%
27.5%
118%
23.8%
31.1%
9.44%
22.7%
44.8%
56.4%
13.2%
12.9%
13.2%
60.8%
36.4%
5.1%
14.8%
33.2%
23.6%
9.8%
33.3%
62.3%
32.7%
Number of
Measurements
3
7
3
3
3
3
7
7
7
7
3
3
6
7
6
7
7
4
3
2
3
7
Significantly Different at
Alpha = 0.01
no
no
no
no
no
no
no
no
yes
yes
no
Relative Percent
Difference (MTI to
ALSI)
38.7%
109 %
-21.2%
-4.2%
62.8%
24.5%
175%
36.1%
174%
160%
118%
39
-------
Table 6-4. Continued
Sample Lot No./ Site
25/Puget Sound
MTI
ALSI
27/Puget Sound
MTI
ALSI
35/Puget Sound
MTI
ALSI
48/Puget Sound
MTI
ALSI
50/Puget Sound
MTI
ALSI
57/Puget Sound
MTI
ALSI
62/Puget Sound
MTI
ALSI
04/Carson River
MTI
ALSI
38/Carson River
MTI
ALSI
53/Carson River
MTI
ALSI
54/Carson River
MTI
ALSI
63/Carson River
MTI
ALSI
1 //Manufacturing Site
MTI
ALSI
20/Manufacturing Site
MTI
ALSI
32/Manufacturing Site
MTI
ALSI
33/Manufacturing Site
MTI
ALSI
49/Manufacturing Site
MTI
ALSI
52/Manufacturing Site
MTI
ALSI
66/Manufacturing Site
MTI
ALSI
Avg. Cone.
mg/kg
56.6
16.6
76.7
45.7
1.61
0.009
62.6
82.9
188
167
1.58
0.73
27.6
14.6
0.29
0.11
2.71
0.53
323
484
733
711
230
169
13.4
10.5
53.2
63.9
876
592
1048
1204
123
84.2
390
424
949
892
RSD or CV
70.9%
12.3%
25.5%
22.2%
1 1 .4%
6.3%
26.8%
34.2%
38.2%
17.7%
74.1%
16.2%
21.1%
28.3%
56.0%
9.1%
23.5%
6.2%
23.5%
35.5%
32.7%
21.0%
34.8%
6.6%
16.9%
14.6%
20.9%
25.3%
40.2%
12.7%
36.6%
13.3%
28.2%
12.5%
33.1%
21.9%
26.6%
11.2%
Number of
Measurements
3
3
7
7
3
7
3
7
7
7
6
7
7
7
3
7
6
7
7
7
7
7
7
7
2
7
7
7
7
7
6
7
7
7
7
7
6
7
Significantly Different at
Alpha = 0.01
no
yes
yes
no
no
no
yes
no
yes
no
no
no
no
no
no
no
no
no
no
Relative Percent
Difference (MTI to
ALSI)
109%
50.6%
198%
-27.9%
73.3%
73.6%
61 .6%
90%
134%
-39.9%
3.0%
30.6%
24.3%
-18.3%
38.7%
-13.8%
37.5%
-8.4%
6.2%
CV = Coefficient of variance
40
-------
Hypothesis Testing
Sample results from field and spiked field samples for the
vendor com pared to sim ilar tests by the referee laboratory
are used as another accuracy check. Spiked samples
were used to cover concentrations not found in the field
samples, and they are considered the same as the field
samples for purposes of comparison. Because of the
limited data available for determining the accuracy of the
spiked value, these were not considered the same as
reference standards. Therefore, these samples were
evaluated in the same fashion as field samples, but they
were not compared to individual spiked concentrations.
Using a hypothesis test with alpha = 0.01, vendor results
for all samples were compared to laboratory results to
determine if sample populations are the same or
significantly different. This was performed for each sample
lotseparately. Because this comparison doesnotseparate
precision from bias, if MTI's or ALSI's computed standard
deviation was large due to a highly variable result
(indication of poor precision), the two CIs could overlap.
Therefore, the fact that there was no significant difference
between the two results could be due to high sample
variability. Accordingly, associated RSDs have also been
reported in Table 6-4, along with results of the hypothesis
testing for each sample lot.
Of the 31 sample lots (including results below 8 mg/kg), 6
results are significantly different, based upon the
hypothesis test noted above. Of these six results, four are
at concentrations of 0.5 mg/kg, or less. This is due to the
inability of the MTI field instrument to accurately measure
results in this low range, as was noted previously.
Therefore, only two results within the range of the MTI
analysis capability are significantly different than the
laboratory results. This suggests that there is no difference
between these two data sets.
The most striking feature of the information in Table 6-4 is
the high variability in reported RSDs for the MTI data. This
high variability is a major reason that the comparison
between vendor and referee laboratory data are not found
to be different. Discounting the results that are too low to
accurately quantitate, there appears to be very little
difference between MTI and referee laboratory data. It is
not, however, a result of the MTI instrument accurately
determining sample concentrations. As was noted in the
discussion of SRMs, the MTI field analysis was successful
at measuring accurate concentrations of mercury only
about 50% of the time. The increased similarity noted
between MTI and the laboratory is a result of the inability of
the hypothesis test to note a difference between the two
results when sample variability is high.
Most of the relative percent differences are positive, which
indicates that the MTI result is generally higher than the
laboratory result. This is indicative of the previously noted
low bias associated with the laboratory data. There are
some MTI results that are less than the laboratory result;
therefore, no overall MTI high or low bias is apparent. It
appears that MTI data are subject to more random
variability.
In determining the number of results significantly above or
below the value reported by the referee laboratory, only 6
of 21 MTI average results were found to have relative
percent differences greater than 50% for sample
concentrations above the estimated PQLof 8 mg/kg. Only
2 of 21 MTI average results have relative percent
differences greater than 100% for this same group of
samples (see Table 6-5). Interferences may be a problem
but, because of the random variability associated with the
data, no interferences are specifically apparent from the
data collected. Table 6-6 shows the results of additional
data collected for these same samples.
In addition to the statistical summary presented above,
data plots (Figures 6-1, 6-2, and 6-3) are included in order
to present a visual interpretation of the accuracy. Three
separate plots have been includedforthe MTI data. These
three plots are divided based upon sample concentration
in order to provide a more detailed presentation.
Concentrations of samples analyzed by MTI ranged
approximately from 0.01 to over 1,200 mg/kg. The
previous statistical summary eliminated some of these data
based upon whether concentrations were interpreted to be
in the analytical range of the MTI field instrument. This
graphical presentation presents all data points. It shows
MTI data compared to ALSI data plotted against
concentration. Sample groups are shown by connecting
lines. Breaks between groups indicate a different set of
samples at a different concentration. Sample groups were
arranged from lowest to highest concentration.
As can be seen by this presentation, samples analyzed by
MTI below about 50 mg/kg did not match well with the ALSI
results with some exceptions. For higher concentrations,
sample results were much closer to ALSI with some
deviations present. This is only a visual interpretation and
does not provide statistical significance, ltdoes, however,
provide a visual interpretation that supports the previous
statistical results for accuracy, as presented above.
41
-------
Table 6-5. Number of Sample Lots Within Each %D Range
<30% >30%, <50%
>50%, <100%
>100%
Total
Positive %D
Negative %D
Total
4
6
10
4
1
5
4
0
4
2
0
2
14
7
21
Only those sample lots with the average result greater than the PQL are tabulated.
Table 6-6. Concentration (in mg/kg) of Non-Target Analytes
Lot # Site
1 Carson River
2 Puget Sound
4 Carson River
5 Puget Sound
6 Carson River
11 Puget Sound
13 Manufacturing Site
14 Oak Ridge
17 Manufacturing Site
20 Manufacturing Site
21 Oak Ridge
22 Oak Ridge
24 Oak Ridge
25 Puget Sound
26 Oak Ridge
27 Puget Sound
31 Oak Ridge
32 Manufacturing Site
33 Manufacturing Site
35 SRM Can met SO-3
37 SRMCRM-016
38 SRM NWRI TH-2
44 SRM CRM 021
46 SRM CRM 032
48 SRM CRM 023
49 SRM CRM 025
50 SRM RTC spec.
52 SRM RTC spec.
53 SRM RTC spec.
54 SRM RTC spec.
56 Spiked Lot 1
57 Spiked PS- X1.X4
62 Spiked Lot 5
63 Spiked Lot 23
66 Spiked MS-SO-08
67 Spiked Lot 26
TOC O&G
870 190
3500 290
2400 200
3500 210
7200 200
3800 130
3200 100
7800 180
2400 90
2000 <50
7800 320
6600 190
6600 250
46000 1200
88000 340
37000 1100
5000 80
4700 120
<470 120
NR NR
NR NR
NR NR
NR NR
NR NR
NR NR
NR NR
NR NR
NR NR
NR NR
NR NR
870 190
3500 290
3500 210
5700 100
NA NA
88000 340
Aa
<0.5
<0.5
<0.5
<0.5
<0.5
<0.5
<0.5
0.32
<0.5
<0.5
1.9
1.7
<0.5
<0.5
9.1
<0.5
0.59
<0.5
<0.5
NR
0.7
5.8
6.5
81
NR
130
NR
NR
NR
NR
<0.5
<0.5
<0.5
37
NA
9.1
As
9
3
8
3
4
4
2
2
<2
<2
4
5
5
2
10
3
4
2
<2
NR
7.8
8.7
25
370
380
340
NR
NR
NR
NR
9
3
3
11
NA
10
Ba
210
23
240
28
32
20
110
41
180
150
150
120
89
46
140
33
120
160
340
300
79
570
590
120
76
1800
NR
NR
NR
NR
210
23
28
280
NA
145
Cd
<0.5
<0.5
<0.5
<0.5
<0.5
<0.5
<0.5
0.4
<0.5
<0.5
2.8
<0.5
<0.5
0.7
1.9
0.7
<0.5
<0.5
<0.5
NR
0.47
5.2
1.2
130
0.92
370
NR
NR
NR
NR
<0.5
<0.5
<0.5
0.9
NA
1.9
Cr
19
16
17
18
16
18
42
16
48
35
22
44
6.3
35
47
39
41
190
9.7
26
14
120
11
15
31
440
NR
NR
NR
NR
19
16
18
25
NA
47
Cu
13
10
32
11
9
8
51
9
20
52
40
36
7
33
73
29
32
47
31
17
16
120
4800
590
8.9
7.8
NR
NR
NR
NR
13
10
11
170
NA
73
Pb
3
1
12
3
1
1
7
11
15
5
23
23
10
31
82
31
16
6
8
14
14
190
6500
4600
210
1450
NR
NR
NR
NR
3
1
3
140
NA
82
Se
<2
<2
<2
<2
<2
<2
<2
<2
<2
2
<2
<2
<2
<2
<2
<2
<2
<2
<2
NR
1
0.83
NR
170
120
520
NR
NR
NR
NR
<2
<2
<2
<2
NA
<2
Sn
<5
<5
<5
<5
<5
<5
<5
<4
<5
<5
<4
<5
<5
6
5
5
<5
<5
<5
NR
NR
NR
300
1300
NR
NR
NR
NR
NR
NR
<5
<5
<5
<5
NA
5
Zn
60
24
66
28
24
24
61
74
120
68
340
160
31
98
250
110
96
78
110
52
70
900
550
2600
94
52
NR
NR
NR
NR
60
24
28
170
NA
250
Ha
0.19
0.04
0.10
0.16
0.23
0.63
5.5
78
10
83
14
88
220
35
100
120
870
650
1300
0.02
0.16
0.62
4.7
21
78
100
200
600
900
1100
0.20
0.61
23
270
980
740
CRM = Canadian Reference Material
RTC = Resource Technology
NA = Not Analyzed
Corporation
NR = Not Reported by Standard Supplier
42
-------
15
r*
Figure 6-1. Data plot for low concentration sample results.
Figure 6-2. Data plot for medium concentration sample results.
-------
iooo
1300
Figure 6-3. Data plot for high concentration sample results.
Unified Hypothesis Test
SAIC performed a unified hypothesis test analysis to
assess the comparability of analytical results provided by
MTI and those provided by ALSI. (See appendix B for a
detailed description of this test.) MTI and ALSI both
supplied multiple assays on replicates derived from a total
of 31 different sample lots, consisting of fie Id materials and
reference materials. The MTI and ALSI data from these
assays formed the basis of this assessment.
Results from this analysis suggest that the two data sets
are not the same. The null hypothesis tested was that, on
average, MTI and ALSI produce the same results within a
given sample lot. The null hypothesis is rejected in part
because MTI results tended to exceed those from ALSI for
the same sample lot. Even when a bias term is used to
correctthis discrepancy, the null hypothesis is still rejected.
Additional information about this statistical evaluation is
included in Appendix B.
Accuracy Summary
In summary, MTI data were only within SRM 95%
prediction intervals about 50% of the time, which suggests
significant non-equivalence. ALSI data, however,
compared favorably to SRM values and were within the
95% prediction interval 89% of the time, indicating
statistical parity found to be biased low.
The comparison between the MTI field data and the ALSI
results suggest that the two data sets are not different, but
this similarity between individual samples is often the re suit
of high variability associated with the MTI reported values.
When a unified hypothesis test is performed, the results
suggest that the two data sets are in fact not the same.
MTI data was found to be both above and below referee
laboratory concentrations; however, the sample lot
distribution shown in Table 6-5 implies a positive bias when
compared to the low bias previously noted for the
laboratory reported results. The number of MTI average
values greater than 50% different from the referee
laboratory results or SRM reference values, however, was
only 6 of 21 different sample lots and those greater than
100% different were only 2 of 21 different sample lots. MTI
results therefore appear to provide a rough estimate of
accuracy for field determination and may be affected by
interferences not identified by this demonstration.
6.1.3 Precision
Precision is usually thought of as repeatability of a specific
measurement, and it is often reported as RSD. The RSD
is computed from a specified number of replicates. The
44
-------
more replications of a measurement, the higher confidence
associated with a reported RSD. Replication of a
measurement may be as few as 3 separate measurements
to 30 or more measurements of the same sample,
depending upon the degree of confidence desired in the
specified result. Most samples were analyzed seven times
by both MTI and the referee laboratory. In some cases,
samples may have been analyzed as few as three times
and some MTI results were judged invalid and were not
used. This was often the situation when it was believed
that the chosen sample, or SRM, was likely to be below the
vendorquantitation limit. The precision goal forthe referee
laboratory, based upon pre-demonstration results, is an
RSD of 25% or less. A descriptive evaluation for the
differences between MTI RSDs and the referee laboratory
RSDs was determined. In Table 6-7, the RSD for each
separate sample lot is shown for MTI compared to the
referee laboratory. The average RSD was then computed
for all measurements made by MTI, and this value was
compared to the average RSD for the laboratory.
In addition, the precision of an analytical instrument may
vary depending upon the matrix being measured, the
concentration of the analyte, and whether the
measurement is made for an SRM or a field sample. To
evaluate precision for clearly different matrices, an overall
average RSD for the SRMs is calculated and compared to
the average RSD for the field samples. This comparison
is also included in Table 6-7 and shown for both MTI and
the referee laboratory.
The purpose of this evaluation is to determine the field
instrument's capability to precisely measure analyte
concentrations under real-life conditions. Instrument
repeatability was measured using samples from each of
four different sites. Within each site, there may be two
separate matrices, soil and sediment. Not all sites have
both soil and sediment matrices, nor are there necessarily
high, medium, and low concentrations for each sample
Table 6-7. Evaluation of Precision
site. Therefore, spiked samples were included to cover
additional ranges.
Table 6-7 shows results from Oak Ridge, Puget Sound,
Carson River, and the manufacturing site. It was thought
that because these four different field sites represented
different matrices, measures of precision may vary from
site to site. The average RSD for each site is shown in
Table 6-7 and compared between MTI and the referee
laboratory. SRM RSDs are not included in this comparison
because SRMs, while grouped with different sites for
purposes of ensuring that the samples remained blind
during the demonstration, were not actually samples from
that site, and were, therefore, compared separately.
The RSDs of various concentrations are compared by
noting the RSD of the individual sample lots. The ranges
of test samples (field, SRMs, and spikes) were selected to
cover the appropriate analytical ranges of MTI's
instrumentation. Average referee laboratory values for
sample concentrations are included in the table, along with
SRM values, when appropriate. These are discussed in
detail in Section 6.1.2 describing the accuracy evaluation
and are included here for purposes of precision
comparison. Sample concentrations were separated into
approximate ranges: low, medium, and high, as noted in
Table 6-7 and Table 6-1. Samples reported by MTI as
below their approximated MDL were not included in Table
6-7. Also not included in this table are samples where a
high number of results were reported as invalid by MTI.
Other than the low concentrations previously noted, there
appears to be no correlation between concentration (low,
medium, or high) and RSD; therefore, no other formal
evaluations of this comparison were performed.
The referee laboratory analyzed replicates of all samples
analyzed by MTI. This was used for purposes of precision
comparison to MTI. RSD for the vendor and the laboratory
were calculated individually and shown in Table 6-7.
Sample
Lot No. MTI and Lab
Avg. Cone, or Reference
SRM value
RSD
Number of
Samples
w/in 25% RSD Goal?
OAK RIDGE
Lot no. 14
MTI
ALSI
Lot no. 21
MTI
ALSI
Lot no. 22
MTI
ALSI
4.75 (medium)
1 1.2 (medium)
81. 6 (high)
37.5%
27.5%
118%
23.8%
31.1%
9.4%
3
7
3
3
3
3
no
yes
no
yes
no
yes
45
-------
Table 6-7. Continued
Sample Lot No. MTI and Lab
Lot no. 24
MTI
ALSI
Lot no. 26
MTI
ALSI
Lot no. 31
MTI
ALSI
Lot no. 37
MTI
ALSI
Lot no. 67
MTI
ALSI
Oak Ridge Avg. RSD
MTI
ALSI
Avg. Cone, or Reference
SRM value
221 (high)
77.0 (high)
947 (high)
0.11 (low)
835 (high)
RSD
22.7%
44.8%
56.4%
13.2%
12.9%
13.2%
60.8%
36.4%
5.0%
14.8%
40.6%
20.9%
Number of
Samples
7
7
7
7
3
3
6
7
6
7
w/in 25% RSD Goal?
yes
no
no
yes
yes
yes
no
no
yes
yes
no
yes
PUGET SOUND
Lot no. 05
MTI
ALSI
Lot no. 1 1
MTI
ALSI
Lot no. 25
MTI
ALSI
Lot no. 27
MTI
ALSI
Lot no. 35
MTI
ALSI
Lot no. 48
MTI
ALSI
Lot no. 50
MTI
ALSI
Lot no. 57
MTI
ALSI
Lot no. 62
MTI
ALSI
Puget Sound/ Avg. RSD
MTI
ALSI
0.21 (low)
0.81 (low)
16.6 (medium)
45.7 (medium)
0.11 (low)
0.017(lo
77.78 (high)
200 (high)
0.73 (low)
14.6 (medium)
9.8%
33.3%
62.3%
32.7%
70.9%
12.3%
25.5%
22.2%
1 1 .4%
6.3%
26.8%
34.2%
38.2%
17.7%
74.1%
16.2%
21.1%
28.5%
44.0%
24.1%
3
2
3
7
3
3
7
7
3
7
3
7
7
7
6
7
7
7
yes
no
no
no
no
yes
no
yes
yes
yes
no
no
no
yes
no
yes
yes
no
no
yes
CARSON RIVER
Lot no. 04
MTI
ALSI
Lot no. 38
MTI
ALSI
0.11 (low)
0.62 (high)
56.0%
9.1%
23.3%
6.2%
3
7
6
7
no
yes
yes
ves
46
-------
Table 6-7. Continued
Sample Lot No. MTI and Lab
Lot no. 53
MTI
ALSI
Lot no. 54
MTI
ALSI
Lot no. 63
MTI
ALSI
Carson River/ Avg. RSD
MTI
ALSI
Avg. Cone, or Reference RSD
SRM value
900 (high)
1100 (high)
169 (high)
23.5%
35.5%
32.6%
21.0%
34.8%
6.6%
49.3%
16.3%
Number of
Samples
7
7
7
7
7
7
w/in 25% RSD Goal?
yes
no
no
yes
no
yes
no
yes
MANUFACTURING SITE
Lot no. 17
MTI
ALSI
Lot no. 20
MTI
ALSI
Lot no. 32
MTI
ALSI
Lot no. 33
MTI
ALSI
Lot no. 49
MTI
ALSI
Lot no. 52
MTI
ALSI
Lot no. 66
MTI
ALSI
Manufacturing Site/ Avg. RSD
MTI
ALSI
10.5 (medium)
63.9 (high)
592 (high)
1204 (high)
99.8 (high)
600 (high)
892 (high)
16.9%
14.6%
20.9%
25.4%
40.2%
12.7%
36.6%
13.3%
28.2%
12.5%
33.1%
21.9%
26.6%
11.2%
28.2%
15.4%
2
7
7
7
7
7
6
7
7
7
7
7
6
7
yes
yes
yes
yes
no
yes
no
yes
no
yes
no
yes
no
yes
no
yes
SUMMARY STATISTICS
Overall Avg. RSD
MTI
ALSI
Field Samples/ Avg. RSD
MTI
ALSI
SRMs/Avg. RSD
MTI
ALSI
35.1%
22.3%
39.4%
19.4%
30.9%
25.1%
no
yes
no
yes
no
yes
47
-------
As noted from Table 6-7, MTI precision is often more
variable than the referee laboratory. The single most
important measure of precision provided in Table 6-7,
overall average RSD, is 22.3% for the referee laboratory,
compared to the MTI average RSD of 35.1%. Only the
laboratory RSD is within the predicted 25% RSD objective
for precision expected from both analytical and sampling
variance.
In addition, field sample precision compared to SRM
precision shows that there may be some difference
between these two sample lots (field sample RSD is 19.4%
for ALSIand 39.4% for MTI; SRM RSD is 25.1% for ALSI
and 30.9% for MTI), but that this difference is likely not
significant. For purposes of this analysis, spiked samples
are considered the same as field samples because these
were similar field matrices, and the resulting variance was
expected to be equal to field samples. The replicate
sample RSDs confirm the pre-demonstration results,
showing thatsample homogenization procedures mettheir
originally stated objectives.
There appears to be no significant site variation between
Oak Ridge, Puget Sound, and the Carson River sites.
(See Table 6-7 showing average RSDs for each of these
sample lots. These average RSDs are computed using
only the results of the field samples and not the SRMs.)
The manufacturing site had a lower average RSD for both
the vendor and the laboratory, but this difference was not
significant in results from different vendors and, therefore,
may not be significant.
Precision Summary
The precision ofthe MTI field instrument is notas good as
the measured laboratory precision. The overall RSD for
MTI is 35.1%, which is above the 25% RSD objective set
for the laboratory. The overall laboratory RSD is 22.3%.
Because MTI precision shows a wider variance in sample
results, MTI data did not prove to be significantly different
from the ALSI data when performing a hypothesis test
procedure for data comparison.
6.1.4 Time Required
Measurement
for Mercury
During the demonstration, the time required for mercury
measurement activities was measured. The following
specific activities were timed: instrument setup, sample
preparation,sample analysis, and instrumentdisassembly.
Two MTI technical representatives performed these
operations during the demonstration; one individual
performed sample preparation (i.e., acid digestion of the
soil and sediment samples) and the other individual
conducted instrument setup/disassembly and all sample
analyses with the PDV 6000.
Setup and disassembly times were measured once.
Sample preparation and analytical time were measured
separately each day. Combined, sample preparation and
sample analysis comprised the operational time.
Recording of sample preparation time began when the first
aliquot of sample material was weighed out on a portable
scale and continued until the last digestion bottle was
completed at the end of the day. Recording of analytical
time began with preparation and initial analysis of the
mercury standard and continued until the last sample
analysis was completed at the end of the day.
To acquire the total operational time for all three days, the
sum of the total sample preparation time and total sample
analysis time was divided by the total number of analyses
performed over the 3-day demonstration. For this
calculation, analyses of blanks and calibration standards
and reanalyses of samples were not included in the total
number of samples. Anydowntime (i.e., lunch breaks)was
noted and subtracted from the total daily operational time.
Setup time for the PDV 6000 consisted of 1) transporting
the PDV 6000 carrying case, preparation kits, laptop
computer, and other ancillary equipment from a rental
vehicle to the measurement site; 2) unpacking and
spreading the items out on a on a level working surface; 3)
connecting the cell assembly to the PDV 6000 control unit
and connecting otherperipheral devices, such as the serial
cable used to link the PDV 6000 to the laptop computer;
and 4) securing all power connections.
It should be noted that the VAS software was already
installed on MTI's laptop computer. If this were not the
case for potential users, then the software installation
would add to the setup time.
The main components of the PDV 6000 were contained
within a hard shell, pelican-style carrying case. These
components included the main PDV 6000 control unit, the
cell assembly with stirrer, the cell stand, a Ag/AgCI
reference electrode, a platinum counter electrode, a glassy
carbon working electrode, a connecting cable to link the
analytical cell to the control unit, a DB9 male/female serial
cable link for connecting the PDV 6000 to a computer, a
reference electrode plating accessory, a main powered 12
VDCsupply(plug pack), a NiMH rechargeable battery pack
and battery pack charger, the PDV6000 operation manual,
VAS software installation disks, and a printed VAS User's
Guide and warranty card.
48
-------
Of these components, the reference electrode plating
accessory, the NiMH battery pack, and the charger were
not used during the demonstration. (The pre-plating
procedure is not required for mercury analysis and the
instrument was powered by line electricity.)
In addition to the main instrument components, there was
ancillary equipment that was unpacked as part of setup.
Most of this equipment related directly to sample digestion
procedures. These included the following items:
Portable scale
Air displacement pipettes (5-50 uL, 100-1000 uL, and
200-1000 uL)
Repeating pipetter (0.5 - 50 ml_)
Sample processing kits
Incidentals (i.e., paper wipes, felt labeling pens, etc.)
These items, along with other supplies, were transported
in a cardboard box. Setup of all of the necessary
equipment and supplies took approximately 15 minutes.
It should be noted that, although MTI utilized two people
during the demonstration, one person could perform both
the sample preparations and sample analyses. According
to MTI, the number of individuals required for conducting
full analyses should be based on the number of sam pies to
be analyzed. For instance, if there are 15 samples or less
per day to be analyzed, MTI would advise that one person
would be sufficient; if there were to be more than 15
samples analyzed daily, the user may decide to use two
people. In the case of the demonstration, where over 50
samples were analyzed daily, two people were deemed
necessary by MTI to efficiently process the samples.
Individual sample analysis times were not measured
during the demonstration. Analysis time was estimated by
recording start and stop times each day and accounting for
any instrument downtime due to operator breaks or device
failure and maintenance activities. Therefore, the total time
foranalyses included preparing a mercury standard and an
electrolyte standard solution each day, analyzing blanks
and the calibration standards, and conducting reanalyses
on certain samples; however, the total numberof analyses
performed includes only demonstration samples (i.e.,
samples, spikes, and SRMs). The sample total does not
include blanks, calibration standards, or reanalyses. Table
6-8 presents the time measurements recorded for each of
the three days of PDV 6000 operation.
Table 6-8. Time Measurements for MTI (minutes)3.
Day 1 Day 2 Day 3
Activity
Instrument Setup
Sample Preparation
Sample Analysis
Operation Time b
Disassembly
a Times are rounded
b Operation Time =
Analysis Time.
15
370
370
740
-
to nearest
Sample
-
430 0
560 550
990 550
10
15 minutes.
Preparation Time
3-Day
Totals
15
800
1,480
2,280
10
+ Sample
Disassembly of the instrumentand ancillary equipmentwas
measured from the completion time of the last sample
analyses until the instrument components were
disassembled and placed back into the original shipping
container. Disassembly of the PDV 6000 involved turning
off the power, disconnecting the electrical power source,
the serial cable link between the control unit and laptop
computer, and removal of the cable link connecting the
analytical cell to the control unit. This complete process
took about 15 minutes (earlier stated 15-minute rounding).
Packaging for shipping is not included in the time
measurement; however, the PDV 6000 carrying case
appears durable enough to ship as is or in a box, without
the need for reinforced corners and buffer spaces. The
pipettes, scale, and any unused supplies would also need
to be packaged for shipment. It is estimated that this
complete process would take approximately 1 hour,
including the time to ship any returnable items to suppliers.
Analysis Time Summary
In total, MTI analyzed 197 samples during the
demonstration. Sample preparation took a total of 800
minutes (roughly 4 minutes per sample) and sample
analysis took 1,480 minutes (7.5 minutes per analysis).
Using the total operational time reported in Table 6-8
(2,280 minutes), 11.6 minutes is the estimated time per
complete sample analysis for the demonstration. As
previously noted, the number of analyses does not include
blanks, standards, and reanalyzed samples.
It is realized that actual times will vary from site to site
depending on project goals. For example, if the results
can be reported as "greater than" or "less than" a specific
target concentration, then the time per sample analysis
could be significantly reduced. On the other hand, if high
49
-------
concentration samples require one or more dilutions, or
heterogeneous samples require duplicate analysis, the
sample time per analysis may increase. If project goals
require all samples to be quantified, the number of
reanalyses and blanks required could be higher, thus also
increasing the time per analysis. During the
demonstration, no "greater than values" were reported;
however, some "less than values" were reported.
6.1.5 Cost
Background information, assumptions used in the cost
analysis, demonstration results, and a cost estimate, are
provided in Chapter 7.
6.2 Secondary Objectives
This section discusses the performance results for the
PDV 6000 in terms of the secondary objectives described
in Section 4.1. These secondary objectives were
addressed based on observations of the PDV 6000 made
during the demonstration and information provided by MTI
during and after the demonstration.
6.2.1 Ease of Use
Documents the ease of use, as well as the skills and
training required to properly operate the device.
Based on observations made during the
demonstration and review of the PDV 6000
operation manual, the instrument appears to be
easy to operate as a stand-alone unit or in
conjunction with the VAS software. Training on
the unit and software would be recommended for
first-time users. A laboratory or field technician
with a basic knowledge of chemistry and basic
computer knowledge could operate the
equipment after a 1-day training course.
The vendor provides a SOP that summarizes the sample
preparation and sample analysis procedures for the PDV
6000. Also provided with the unit is the PDV 6000
Operations Manual, version 2.2 (47 pages in length), and
the VAS User's Guide, version 2.1 (50 pages in length).
This SOP was evaluated during the demonstration and the
step-by-step procedures were easy to understand.
The analysis portion of the MTI mercury measurement
process may require analytical experience to gain
efficiency in the field. Familiarity with calibrating analytical
instruments, running blanks before processing samples,
running calibration standards, and overall familiarity with
sample peaks, would be an advantage for a prospective
user. The VAS User's Guide does provide some examples
of sample peaks in the chapter titled "Analyzing Data."
There is also some discussion of peak measurement with in
the PDV 6000 Operations Manual in Chapter 3
(Introduction to Voltammetry).
In addition to providing the written instructions, MTI
provides a 1 -day training course on the PDV 6000 for up to
eight individuals (at the purchaser's cost). Currently, there
is no specific training course for the VAS software. The
combination of instrument training, the field SOP, the PDV
6000 Operations Manual, and the VAS User's Guide
should provide a user with adequate direction on basic use
of the PDV 6000. Once in the field, MTI does supply some
ongoing support. MTI has a phone number for domestic
support (910-617-8367) and an Internet email support
address with support provided from Australia. Chapters of
the PDV6000 Operations Manual does include information
on troubleshooting. Neither the training course nor email
support was evaluated during the demonstration.
MTI chose to operate the PDV 6000 with two MTI
representatives during the demonstration. One individual
who conducted the sample preparation held B.S. degrees
in biology and chemistry. The other individual conducting
the analysis held a degree in chemistry/pharmacology.
The two MTI representatives were able to perform sample
preparation and analysis on a somewhat continuous basis,
although sample preparation was more routine and, thus,
a quicker operation. The sample digestion procedure kept
far enough ahead of the sample analysis that all 197
samples were prepared for analysis more than a day
ahead of completing all the analyses. The separation of
sample extraction and analysis is not recommended by
MTI; however, it was performed by MTI as a means of
completing the 197 samples over the 3-day time frame.
Sample preparation took approximately four minutes per
sample. Sam pie preparation consisted of weighing out a
specific mass of sample material, placing the material into
a 70-mL digestion bottle, and then pipetting the following
into the bottle.
• 4mLofHNO3
4 ml_ of H2O2 (1 ml_ at a time)
12 ml_ of Dl water
20 ml_ of electrolyte solution
This procedure was easy to understand, and could be
performed by a trained technician. The two main
50
-------
peripheral items used during the sample preparation
procedure are a portable analytical scale and one or more
pipettes.
MTI typically would not supply the analytical balance and
pipettes, although the company indicates that these items
could be purchased as part of an accessory kit; however,
MTI does supply disposable pipette tips within the sample
preparation kits. The reader should note that brands and
models of the balance and pipettes, otherthan those used
for the demonstration, may be adequate for use.
Sample analysis on average took approximately 10-12
minutes per sample. Because sample analysis was a
separate procedure and involved running blanks,
calibration standards, and some reanalysis, the procedure
lagged behind sample preparation. If a single individual
were to perform both preparation and analysis, it would be
advantageous to prepare several samples in advance of
conducting analyses.
The sample analysis process is more involved than the
sample preparation procedure and should be performed by
a technician trained in using the PDV 6000 with the VAS
software. As sam pies are analyzed, VAS software screens
do allow the analyst to track the stage at which the
analytical process is progressing, which would be
especially beneficial to the novice user. Figure 6-4 is an
example of the PDV 6000 run log screen with open data
windows. Figure 6-5 shows the sample graph screen for
a mercury analysis and Figure 6-6 has a sample graph and
data for a sample analyzed by the method of standard
additions.
E0 Sample
0 Sample
0 Sample
^ 15377
^j Sample
£| Standard Addit
g Results
13447
£3 Blank
^j Standard
^j Sample
L|£j Results
Q 10148
^ Standard
^] Sample
3 Results
Q 11373
] Blank
j Sample
j Standard Addit
LH Results
19444
j^l Sample
17445
[72! Sample
|5?j Standard Addit
17148
0 Blank
0 Standard
22:43:57
.22:46:28
22:48:16
22:54:27
Initiate Run
17Jun2003
17Jun2003
17Jun2003
17Jun2003
Sample-file: SOppbstd
. Comment; 'I
Start
Cancel
Using:: flun-Configuration... |
Run Type—
^ Cample f*1 Standard Addition
r Standard f Blank
Help
- Concentrations -
Standards-Set >
f Hg - 20 pprn
.Save;., Standards .-libtaij.,.
Acid-
Edit...
Delete
--Volumes—
SafflpleVolurne [mLJ.1: |.1
lotal Standard Volume (ml);: [05
Electrolyte Volume (ml): f20~
Delect Peaks
Detection Limit |%): [75
hing.m: |75
DK
Cancel
Help
.Blank Subtraction
Artificial Blank Subtraction
Generate For Entire F3e l~ Retain Hand Modified Peaks
Figure 6-4. PDV 6000 sample run log with open data windows.
51
-------
Jj.!sDl&i<
JHelp
[¥.9J
WW" IEPA H° ""' LV 2-11334. Sample
260mV 280* 300mV
]1x1 Zoom- [Anode |1338mV |98.8 .j4
Put Helit,.prs&n I Instrument: PDV3000
Figure 6-5. PDV 6000 sample graph screen.
lEOHt I RiFI
plvst: [iBLisy ^etutJ W«i
jJil§|B|g|ileJ
TOW 150 M/ 200 mV 250W/ 300 mV 350*' 4001* 450* 500 iW 550 nnV 'BflOmV
Figure 6-6. PDV 6000 mercury graph for sample analyzed by
method of standard additions.
52
-------
6.2.2 Health and Safety Concerns
Documents potential health and safety concerns
associated with operating the device.
The main health and safety concern observed
during the demonstration was potentialexposure
to fumes resulting from vigorous reactions
during sample preparation. This occurrence
necessitates the use of appropriate PPE when
conducting acid digestions.
Health and safety concerns, including chemical hazards,
radiation sources, electrical shock, explosion, and
mechanical hazards were evaluated.
Chemicals were used in the preparation or processing of
samples. These included the following:
Nitric acid (HNO3) (concentrated)
Hydrogen peroxide (H2O2) (1 00%)
An electrolyte solution containing sodium acetate,
acetic acid, and trace metals
Potassium chloride (KCI)
Mercury standard
Of these chemicals, the two posing the greatest exposure
risk are HNO3 (a strong oxidizer) and H2O2 (a strong
oxidant). MTI does include Material Safety Data Sheets
(MSDS) for HNO3, H2O2, the electrolyte solution, and the
mercury standard.
The main potential health and safety concern regarding
MTI's process is, therefore, potential exposure to spilled
chemicals or fumes caused by reactions of these
chemicals with one another or with unknown sample
constituents. Ampules of these chemicals were handled
with gloves, and the operator wore safety glasses with side
shields at all times. Such standard laboratory precautions
mitigate the potential for dermal exposure. In addition, the
MTI representative conducting the sample preparation
wore a long-sleeved laboratory coat.
During the demonstration, a few samples reacted
vigorously, releasing pungent reddish-yellow fumes from
the digestion bottles, following the addition of H2O2 to the
solution. In some instances, an effervescing solution
frothed over the top of the digestion bottle and onto the
table.
It should also be noted that the demonstration samples
may pose a mercury inhalation hazard. During the original
demonstration in Oak Ridge, TN, in May 2003, vapor
measurements were recorded for the same samples that
were analyzed by MTI in Las Vegas in June. The
measurements were collected with a Jerome 431-x gold
film mercury vapor analyzer, manufactured by Arizona
Instruments Corporation. The instrument has a rangefrom
0.000 to 0.999 mg/m3. In all cases, readings were 0.000
mg/m in the breathing zone of the operator, indicating no
inhalation hazard.
In evaluating electrical shock potential, two factors were
evaluated: 1) obvious areas where electrical wires are
exposed and 2) safety certifications. No exposed wires
were noted during the demonstration. All connections
between equipment were made using standard electrical
lines, interface cables, and 8-pin cords. The power cord
was grounded with a ground fault interrupter.
No obvious explosion hazards were noted. The PDV 6000
has a CE electrical certification, which is the European
equivalent of the Underwriters Laboratory (UL) electrical
certification in the United States. The unit also has a
N4266 electrical certification (Australia's equivalent to UL),
but it does not have UL certification.
No serious mechanical hazards were noted during the
demonstration. All equipment edges were smooth,
minimizing any chance of cuts or scrapes. If an electrical
cord is used to provide power to the PDV 6000, it needs to
be installed in a manner designed to prevent trip hazard
(e.g., taped down).
6.2.3 Portability of the Device
Documents the portability of the device.
The PDV 6000 is a handheld device that is
transported in a carrying case along with its main
components. Due to its packaging and compact
size and weight, the unit can be taken anywhere
that is accessible by foot and can be operated on
a small-level surface.
The PDV 6000 display unit measured approximately 20 cm
(L) by 10 cm (W) by 4 cm (H) when lying flat. The weight
of the control unit was estimated at less than 1 kg. Also
included as standard with the PDV 6000 were the cell
assembly, stand for the cell assembly with integral stirrer,
a set of electrodes, connecting cables, a NiMH batteryand
53
-------
battery charger, and the accompanying VAS software
installation disk.
All of these accessories are lightweightand easily portable,
and fit into a pelican-style case measuring approximately
35 cm (L) by 25 cm (W) by 15 cm high. As a result, the
PDV6000 is practical for field applications, including use in
remote areas accessible only by foot.
The sample digestion required a flat, stable surface,
which would also be suitable for a laptop computer. A
table was supplied during the demonstration for preparing
and staging samples. The majority of materials and
supplies required forconducting sample preparation come
in a kit that is the approximate size of a shoe box. Each kit
comes with enough supplies to prepare 10 samples.
Additionally, a small portable scale and pipettes are
needed to properly prepare samples for analysis. These
additional items, along with the kits, could be transported
in a heavy duty bag or cardboard box.
The PDV 6000 is equipped with a NiMH battery for
operation in the absence of electrical power. The battery
powers the unit for about 4 hours and would allow for
analyzing 10-20 samples. Line operation of the instrument
requires a standard electrical source of 110 volts.
For the demonstration, the vendor was supplied with a
folding table, two chairs, and a tent to provide shelter from
inclement weather. In addition, two 1-gallon containers
were provided for waste soil and for decontamination
residual extract waste. A 2-gallon zip-lock bag was
furnished for disposal of used gloves, wipes, and other
wastes which were contaminated during the demonstration.
Finally, a large trash bag was supplied for disposal of non-
contaminated wastes.
6.2.4 Instrument Durability
Evaluates the durability of the device based on its
materials of construction and engineering design.
The PDV 6000 is well designed and constructed
for durability. A conformal coating on the
device's electronics allows for use in humid
climates.
However, the outer shell of the instrument display unit
appears to be well designed and constructed, indicating
that the device should be durable under most field
conditions. No evaluation could be made regarding the
long-term durability of the cell assembly, or display circuitry.
Visual inspection did not indicate that any problems were
likely.
MTI indicated that a conformal coating on the device
electronics allows for operation in an unlimited humidity
range (i.e., other than the unit being submersed in water,
no damage should result due to moisture). The PDV 6000
was observed in both wet and dry conditions (the
instrument was operated at both the Oak Ridge, TN, and
Las Vegas, NV, demonstration locations.
6.2.5 Availability of Vendor Instruments and
Supplies
Documents the availability of the device and spare
parts.
The MTI PDV 6000 units are stocked and
available for purchase, rent, or lease. They are
available within one week of order placement
through the new MTI US operations. Spare
parts and consumable supplies can be added
to the original PDV 6000 order. Supplies not
typically provided by MTI (pipetters and a scale)
are readily available from laboratory supply
firms.
The outside of the PDV 6000 is constructed of sturdy
plastic. The main display unit was secured with four
screws. No environmental (e.g., corrosion) or mechanical
(e.g., shear stress or impact) tests were performed.
EPA representatives contacted MTI regarding the
availability of the PDV 6000 and supplies. MTI indicated
thatapproximately 90% of its currentbusiness is purchase,
but that units are stocked and available for purchase, rent,
or lease. According to MTI, such systems are available
within one week of order placement through the new MTI
U.S. operations.
The instrument comes standard with the following items:
PDV 6000 handheld control unit
Analytical cell assembly
Analytical cell stand
Ag/AgCI reference electrode
Platinum counter electrode
Glassy carbon working electrode
Cable to link analytical cell to control unit
54
-------
DB9 serial cable to link PDV 6000 to a computer
Reference electrode plating accessory
Main powered 12 VDC supply (plug pack)
NiMH rechargeable battery pack
NiMH battery pack charger
PDV 6000 operation manual
VAS software installation disks
Printed VAS User's Guide
Warranty and contact sheet
Consumable supplies that are required for sample
preparation are provided by MTI as kits and can be ordered
with the PDV 6000 equipment. The kits contain the
following items:
Digestion bottles (10)
Electrolyte medium (one 500-mL bottle)
Electrolyte ingredients
Hydrogen peroxide (two 30-mL bottles)
Deionized water (one 250-mL bottle)
Mercury standard (one 30-mL bottle)
Waste container (one 500-mL bottle with kitty litter)
Analysis cups (15)
PVC gloves (2 pair)
5-ml soil grabber samplers (10)
10-100 uL pipette tips (14)
100-1,000 uL pipette tips (15)
5,000 uL pipette tips (14)
Part I - Electrode polishing kit - dry cloth
Part II - Electrode polish mixture
Common laboratory equipment also needed for sample
preparation include air displacement pipettes, a repeating
pipette, and a portable balance. This equipment can be
purchased from MTI as part of an accessory kit but also is
readily available from laboratory supply companies.
55
-------
Chapter 7
Economic Analysis
The purpose of the economic analysis was to estimate the
total cost of mercury measurement at a hypothetical site.
The cost per analysis was estimated; however, because
the cost per analysis would decrease as the number of
samples increased, the total capital cost was also
estimated and reported. Because unit analytical costs are
dependent upon the total number of analyses, no attempt
was made to compare the cost of field analysis using the
PDV 6000 to the analytical costs associated with the
referee laboratory. "Typical" unit cost results, gathered
from analytical laboratories, were reported to provide a
context in which to review the PDV 6000 costs. No attempt
was made to make a direct comparison between these
costs for different methods because of differences in
sample throughput, overhead factors, total equipment
utilization factors, and other issues that make a head-to-
head comparison impractical.
This chapter describes the issues and assumptions
involved in the economic analysis, presents the costs
associated with field use of the PDV 6000, and presents a
cost summary for a "typical" laboratory performing sample
analyses using the reference method.
7.1 Issues and Assumptions
Several factors can affect mercury measurement costs.
Wherever possible in this chapter, these factors are
identified in such a way that decision-makers can
independently complete a project-specific economic
analysis. MTI offers purchase, rental, and lease options for
potential PDV 6000 users. Rentals have a 1-month
minimum and MTI has a variation of the rental option
which allows a 50% credit for rental fees toward 50% of the
instrument purchase price. Lease agreements are for
three months, in which the 3-month lease equals the
purchase price. Because site and user requirements vary
significantly, each of these three options is discussed to
provide each user with the information to make a case-by-
case decision.
A more detailed cost analysis was performed on the
instrument purchase option because MTI has indicated that
instrument purchase comprises approximately 90% of
company orders. The results of that cost analysis are
provided in Section 7.2.
7.1.1 Capital Equipment Cost
The PDV 6000 (the analytical instrument) is marketed as
a standard package that includes all necessary equipment
needed to operate the PDV 6000. The instrument and its
accessories are regarded as capital equipment since they
represent a relatively significant capital expenditure that
typically is depreciated. Other equipment is required for
sample preparation; however, these are typically non-
depreciable items of lower cost and are better categorized
as supplies (see Section 7.1.2).
It should be noted upfront that the PDV 6000 can be used
as a stand-alone unit (without the VAS software). For
stand-alone use, the controller unit is programmed to
analyze 10 metals in the concentration range of 1 0 ppb to
30 ppm. According to MTI, the stand-alone unit is more
appropriately used for field screening, and it does not
provide the accuracy of laboratory analysis. MTI has
indicated that the stand-alone unit should only be used for
determining whether a sample may be above or below a
particular threshold.
During the demonstration, MTI utilized the VAS software in
conjunction with the PDV 6000. Because this software
incorporates features such as standard addition calibration,
simultaneous multi-element analysis, and storage of data,
56
-------
it allows for more flexible and accurate analysis. The cost
of the software adds less than 3% to the total cost of the
PDV 6000 unit (i.e., the PDV 6000 costs $7,900 as a
stand-alone unit and $8,100 with the VAS software). For
this reason, an assumption was made that the user will
utilize the VAS software with the PDV 6000 unit, whether
purchased or rented. Costing the unit with the software
reflects the instrumentation configuration used during the
demonstration.
The VAS software requires a computer that has a
minimum specification of 300 Mhz586 processor(Pentium
or equivalent), 64 MB RAM, 10MB of free hard disk space,
and runs on Windows 98, ME, NT, 2000, orXP. The cost
of a computer is not included. For this economic analysis,
an assumption was made thatthe userwill supply a laptop
computer that meets those specifications.
The instrument cost quoted by MTI does not include any
packaging orfreightcosts to ship the instrument to the user
location. For rental and lease agreements, the first month
rental or lease cost is required up front. A user manual is
provided with the unit at no cost. An 8-hour training
session is available for up to eight individuals, at an
additional fee of $950.
7.1.2 Cost of Supplies
The cost of supplies was estimated based on the supplies
required to analyze demonstration samples and on
discussions with MTI. For purposes of this cost estimate,
the supplies required to analyze solid samples were
factored into the cost estimate. Supplies utilized by MTI
during the demonstration consisted of both consumable
and non-consumable items. The consumable supplies
were contained within "sample processing kits" and
consisted of the following items:
Digestion bottles
Electrolyte medium
Electrolyte ingredients
Nitric acid ampules
Hydrogen peroxide
Deionized water
Mercury standard
Waste electrolyte jar
- Analysis cups
- PVC gloves
- 5-mL soil samplers
- 10-100 uL pipetter tips
- 100-1000 uL pipetter tips
- 5000 uL pipetter tips
- Electrode polishing kit
- Electrode polish mixture
These consumable items contained in the kits are readily
available from MTI. Because the user cannot return
unused portions of the kit contents, no salvage value is
included in the cost of supplies. Personal protective
equipment (PPE) supplies are assumed to be part of the
overall site investigation or remediation costs; therefore, no
PPE costs are included as supplies.
Non-consumable supplies utilized by MTI consisted of
reusable items that were not included in the kits. These
items were necessary to conduct the sample digestion
procedure. These items included a portable laboratory
scale required to measure out the correct mass of sample
to digest; a set ofairdisplacement pipetters (5-50 uL, 100-
1,000 uL, and 200-1,000 uL); and a repeating pipetter (0.5
-50 ml_). It should be noted that this type of equipment
may or may not be already owned by a potential PDV 6000
user; however, for this economic analysis, an assumption
was made that the user does not possess these items.
7.1.3 Support Equipment Cost
During the demonstration, the PDV 6000, control unit, cell
assembly, and laptop computer that ran the VAS software
were operated using AC power. The costs associated with
providing the power supply and electrical energy were not
included in the economic analysis; the demonstration site
provided AC power at no cost. None of the items
mentioned above can operate on DC power; however, the
PDV 6000 can run on a NiMH rechargeable battery. The
laptop computer can also run on a rechargeable battery.
Because of the large number of samples expected to be
analyzed during the demonstration, EPA provided support
equipment, including tables and chairs for the two field
technician's comfort. In addition, EPA provided a tent to
ensure that there were no delays in the project due to
inclement weather. These costs may not be incurred in all
cases; however, such equipment is frequently needed in
field situations, so these costs were included in the overall
cost analysis.
7.1.4 Labor Cost
The labor cost was estimated based on the time required
for PDV 6000 and sample preparation equipment setup,
sample preparation, summary data preparation, and
instrument packaging at the end of the day. Setup time
covered the time required to unpack the instrument, set up
all components, and ready the device for operation.
Sample preparation is somewhat labor intensive, albeit a
routine operation. Sample preparation involved the
following steps:
1. 2.0 grams of sample is weighed out and placed into a
digestion bottle,
2. 4.0 mL of HNO3 is pipetted into the bottle,
57
-------
3. 4.0 mL of H2O2 is pipetted into the bottle, 1 ml_ at a
time.
4. 12 mL of deionized (Dl) water is added to quench the
sample, and
5. 20 mL of electrolyte medium is added to the bottle
(thus, a total 40 mL solution is created).
Due to the large number of samples to be prepared and
analyzed, MTI had decided to utilize two individuals, one to
conduct sample preparation and the other to conduct
sample analysis. Sample preparation was the faster
component of the entire procedure. As a result, sample
digestion of all 197 demonstration samples was completed
near the end of the second day of operation. Sample
analysis was the time required to analyze all samples and
submit a data summary. The data summary was strictly a
tabulation of results in whatever form the vendor chose to
provide. In this case, the MTI analyst electronically
transferred the results of all 197 samples from the
computer database at the end of the demonstration.
The time required to perform all tasks was rounded to the
nearest 5 minutes; however, for the economic analysis,
times were rounded to the nearest hour and it was
assumed that a field technician who had worked for a
fraction of a day would be paid for an entire 8-hour day.
Based on this assumption, a daily rate for a field technician
was used in the analysis.
During the demonstration, EPA representatives evaluated
the skill level required for 1) conducting the field sample
preparation and 2) to analyze the samples with the PDV
6000 utilizing the VAS software and report the results for
mercury samples. Based on these field observations, a
field technician with basic chemistry skills acquired on the
job or in a university setting was considered qualified to
conduct the sample preparation procedure. If this
individual also possessed basic computer skills and
completed a 1-day training course specific to the PDV
6000, and for the VAS software, that individual was also
considered qualified to operate the instrument.
Analysis with the PDV 6000 can be performed with either
one or two operators, depending upon the number of
samples expected to be analyzed on a daily basis. For
analyzing 15 or fewer samples per day, MTI advises that
1 person would be sufficient. For analyzing more than 15
samples per day MTI advises using 2 people. One also
has to consider that in remote areas, where the instrument
could be used, a minimum of two field staff are typically
required as a health and safety precaution. (In the case of
the demonstration, where over 50 samples were analyzed
daily, 2 people were deemed necessary by MTI to
efficiently process the samples.)
The use of two individuals may increase costs if travel and
per diem were required; however, these cost are not
considered, as explained in Section 7.1.6. For this
economic analysis, an assumption was made that one
technician will be able to conduct both sample preparation
and sample analysis, with the understanding that there are
ongoing activities at the site, allowing the technician to be
in constant communication with other individuals.
An hourly rate of $15 was used for the field technician. A
multiplication factor of 2.5 was applied to labor costs to
account for overhead costs. Based on this hourly rate and
multiplication factor, and an 8-hour day, a daily rate of
$300 was used for the economic analysis. Monthly labor
rates are based on the assumption of an average of 21
work days per month. This assumes 365 days per year,
and non work days totaling 113 days per year (104
weekend days and 9 holidays; vacation days are
discounted assuming vacations will be scheduled around
short-term work or staff will be rotated during long
projects). Therefore, 252 total annual work days are
assumed.
7.1.5 Investigation-Derived Waste Disposal
Cost
MTI was instructed to segregate its waste into three
categories during the demonstration: 1) general trash;
2) lightly contaminated PPE and wipes; and 3) excess
contaminated soil (both analyzed and unanalyzed) and
other highly contaminated wastes such as the digestion
liquid. General trash was not included as investigation-
derived waste (IDW) and is not discussed in this document.
A separate container was provided for each waste
category.
Lightly contaminated wastes consisted primarily of used
surgical gloves and wipes. The surgical gloves were
discarded for one of three reasons: 1) they posed a risk of
cross contamination (noticeably soiled), 2) they posed a
potential health and safety risk (holes or tears), or 3) the
operator needed to leave the analysis area to perform
other tasks (e.g., using a cell phone, etc.). The rate of
waste PPE generation was in excess of what would be
expected in a typical application of this instrument since
EPA evaluators occasionally contributed used gloves to
this waste accumulation point.
The specific wastes that were generated by MTI's
measurement process consisted of the following:
58
-------
Empty HNO3 ampules
Waste PPE (nitrile gloves) and wipes
Spent 70-mL digestion bottles containing a mixture of
residual contaminated soil, HNO3, and H2O2
Empty analysis cups
Spent electrolyte solution, containing a
concentration of mercury and other metals
small
The empty HNO3 ampules were considered general trash,
the PPE and wipes were considered lightly contaminated
material, and the remaining items were considered
hazardous wastes for purposes of this cost analysis.
7.1.6 Costs Not Included
Items for which costs were not included in the economic
analysis are discussed in the following subsections, along
with the rationale for exclusion of each.
Oversight of Sample Analysis Activities. A typical user
of the PDV 6000 would not be required to payfor customer
oversight of sample analysis. EPA representatives
observed and documented all activities associated with
sample analysis during the demonstration. Costs for this
oversight were not included in the economic analysis
because they were project specific. For the same reason,
costs for EPA oversight of the referee laboratory were also
not included in the analysis.
Travel and Per Diem for Field Technician. Field
technicians may be available locally. Because the
availability of field technicians is primarily a function of the
location of the project site, travel and per diem costs for
field technicians were not included in the economic
analysis.
Sample Collection and Management. Costs for sample
collection and management activities, including sample
homogenization and labeling, are site specific and,
therefore, were not included in the economic analysis.
Furthermore, these activities were not dependent upon the
selected reference method or field analytical tool.
Likewise, sample shipping, COC activities, preservation of
samples, and distribution of samples were specific
requirements of this project that applied to all vendor
technologies and may vary from site to site. None of these
costs were included in the economic analysis.
Items Costing Less than $10. The costs of inexpensive
items, such as paper towels, were not included in the
economic analysis.
Documentation Supplies. The costs for digital cameras
used to document field activities were not included in
project costs. These were considered project-specific
costs that would not be needed in all cases. In addition,
these items can be used for multiple projects. Similarly,
the cost of supplies (logbooks, copies, etc.), used to
document field activities, was not included in the analysis
because they are project specific.
Health and Safety Equipment. Costs for rental of the
mercury vapor analyzer and the purchase of PPE were
considered site specific and, therefore, not included as
costs in the economic analysis. Safety glasses and
disposable gloves were required for sample handlers and
would likely be required in most cases. However, these
costs are not specific to any one vendor or technology. As
a result, these costs were not included in the economic
analysis.
Mobilization and Demobilization. Costs for mobilization
and demobilization were considered site specific, and not
factored into the economic analysis. Mobilization and
demobilization costs actually impact laboratory analysis
more than field analysis. When a field economic analysis
is performed, it may be possible to perform a single
mobilization and demobilization. During cleanup or
remediation activities, several mobilizations,
demobilizations, and associated downtime costs may be
necessary when an off-site laboratory is used because of
the wait for analytical results.
7.2 PDV 6000 Costs
This section presents information on the individual costs of
capital equipment, supplies, support equipment, labor, and
IDW disposal for the PDV 6000.
7.2.1 Capital Equipment
The PDV 6000 sells for $7,900 as a stand-alone unit, and
for $8,100 with the VAS software. Whether purchased or
rented, the unit comes with the following components and
items:
• Handheld control unit
• Analytical cell assembly
• Analytical cell stand
• Reference electrode
• Counter electrode
• Working electrode
• Cable to link unit/cell
• Cable to link unit/laptop
- Plating accessory
- 12 VDC supply
- NiMH battery
- NiMH battery charger
- Operation manual
- VAS installation discs
- VAS User's Guide
- Warranty/contact sheet
59
-------
During the demonstration, the PDV 6000 was operated for
approximately 3 days and was used to analyze 197
samples. Table 7-1 summarizes the PDVGOOOcapitalcost
for four procurement options: purchase, rental, lease, and
rental with an option to purchase. These scenarios cover
only capital cost, not the cost of optional or user-supplied
equipment, supplies, support equipment, labor, and IDW
disposal.
Table 7-1. Capital Cost Summary for the PDV 6000
Item
Quantity Unit Cost ($) Total Cost for Selected Project Duration
1-Month 3-Month 6-Month 12-Month 24-Month
Purchase PDV6000/VAS 1
Monthly Rental of PDV 6000 a 1
Lease PDV 6000/VAS 1
Rental with purchase option b 1
$8,100
$2,200
$2,700
$2,200
$8,100
$2,200
$2,700
$2,200
$8,100
$6,600
$8,100
$6,600
$8,100
$13,200
$8,100
$12,850°
$8,100
$26,400
$8,100
$12,850°
$8,100
$52,800
$8,100
$12,850°
The standard rental period is generally a minimum of 1-month; however, the PDV 6000 can be rented for a 1-week period for $600 (with
pipettes) or $800 (without pipettes).
b 50% of monthly rental fee is credited to 50% of the purchase price.
c After 4 months of renting the unit, $8,800 of rentals fees are incurred. Up to $4,050 can be credited toward purchase.
Therefore, a balance of $4,050 would be owed to purchase the instrument and the total cost would be $12,850.
Figure 7-1 shows the relative costs for the basic capital
equipment for the purchase, rental, and rental/purchase
option (PO) scenarios. The lease arrangement is not
shown since it is very similar to direct purchase for the time
scales presented. These costs reflect the basic PDV 6000
unit and accessories (with VAS software included). No
options (e.g., laboratory scale or multi-volume dispensing
pipetters) and no supply or shipping costs are included.
Purchase
Rental
Rental/PO
Months
Figure 7-1. Capital equipment costs.
As would be expected, this chart shows that purchase is
the most cost-effective option (in terms of capital costs) for
long-term use, and the rental with a purchase option may
be desirable for product trial. Rental is shown to be a
logical choice only for short-term projects (i.e., three
months or less).
MTI's policy of applying a portion of rental costs toward
instrument purchase would be a consideration in instances
where the project duration is unknown, or when the user's
acceptability and reliance on the instrument increase with
usage. As noted in Table 7-1, after4 months of renting the
PDV 6000, $4,050 of the $8,800 expended for rental fees
can be credited toward instrument purchase. Therefore,
a balance of $4,050 would be owed to purchase the
instrument, and the total cost of the rent-to-own option
would be $12,850.
7.2.2 Cost of Supplies
During the demonstration, there were two types of supplies
used by MTI. The first type was consumable items that
were contained within "sample preparation kits." One kit
contained enough supplies for 10 sample measurements,
which correlated to approximately $10 per analysis. Based
on the assumed 10-hourworkday and 21-day work month
(estimated in Section 7.1.4), and MTI's sample
measurement rate of roughly 50 samples per day (as
calculated in Section 6.1.4), a total of 5 "sample
preparation kits" would be required per day. MTI prices
60
-------
these kits at $99 each and offers a 15% discount on orders
of 20 kits or more. Therefore, using a discounted price of
$84 per kit cost, a 21-day (i.e., 1 month) supply would total
approximately $8,800.
In addition to the consumable items contained within the
sample preparation kits, there was reusable sample
preparation equipment used during the demonstration.
These items included a portable laboratory scale used to
measure out the proper mass of sample and a set of
variable-sized pipetters used to dispense accurate and
precise volumes of reagents and standards.
The scale used by MTI was an Aculab portable scale. A
compact electronic scale manufactured by Ohaus is priced
at about $110 (www.Fishersci.com, 2003). This
comparable scale would be sufficient for the sample mass
measurements in the field.
MTI was equipped with four pipetters during the
demonstration. Three of the instruments were air
displacement pipetters with a 50-1000 uL dispensing
capacity, and one of the instruments was a repeating
pipetter with a 0.5-50 ml_ dispensing capacity. The
disposable pipetter tips used for the demonstration were
provided in the kits which prevented cross contamination
between samples.
The microliter range air displacement pipetters are priced
at $230 each (www.Fishersci.com, 2003); therefore, a set
of three air displacement pipetters would cost $690. For
dispensing set amounts of larger liquid volume, MTI used
a 0.5-50 ml_ HandyStep repeating pipetter. This type of
pipetter prices at $320 (www.Fishersci.com, 2003). The
lifetime of these pipetters could vary significantly, but it is
assumed that the pipetters would last several years. The
total annual costs for these supplies (i.e., kits, scale, and
pipetters) would total roughly $10,000.
Table 7-2 summarizes the costs for the supplies, assuming
that the same number of samples are run per day during
each 3-month period. As indicated, for continuous daily
measurements the sample preparation kits dominate the
cost of supplies. For 1 month of continuous sampling, the
kits comprise about 89% of total supply costs. This
percentage escalates with duration until the kits comprise
nearly the entire supply cost percentage. Table 7-3
indicates the costs associated with analysis when the user
prepares their reagents and purchases disposable supplies
as opposed to purchasing preparation kits. The 1-month
column is indicated as not applicable (NA) because it is not
advantageous to prepare reagents for short periods of
time. In addition to the chemical and supply costs, $1,000
per month amount is added to the cost to account for
preparation time and reagent analysis. Based on this
observation, the instrument supplies (i.e., scale and
pipetters)should be considered negligible costs, especially
after the first few months of daily measurements.
Table 7-2. Supplies Cost Summary, Using Preparation Kits
Item
Prep. Kits
Scale
ML Pipetters
mL Pipetter
Total Cost
1
$8,800
$110
$690
$320
$9,900
a Cost values are
b MTI can arrang<
3
$26,000
$110
$690
$320
$27,000
Months a'b
6
$53,000
$110
$690
$320
$54,000
12
$110,000
$110
$690
$320
$110,000
rounded to two significant digits.
3 to make bulk kits available with
24
$210,000
$110
$690
$320
$210,000
electrolyte
and water in 25 L reusable drums, the standard as the small
20-mL vials, the acid and peroxide in reusable 5 L containers
with accurate dispenser and plasticware in a big bag (no
packaging) for $4,000 for 1,000 samples, subject to a
minimum of order of $12,000. The 24-month cost for bulk
supplies would be $96,000.
Table 7-3. Supplies Cost Summary, Preparing Reagents
Item Monthsa
Reagents b
Scale
uL Pipetters
mL Pipetter
Total Cost
1
NA
NA
NA
NA
NA
3
$8,900
$110
$690
$320
$10,000
6
$16,000
$110
$690
$320
$17,000
12
$31,000
$110
$690
$320
$32,000
24
$62,000
$110
$690
$320
$63,000
a Cost values are rounded to two significant digits.
b The "reagents" cost includes disposable supplies and
preparation labor.
7.2.3 Support Equipment
MTI was provided with a 10x10 foottentfor protection from
inclement weather during the demonstration. It was also
provided with one table and two chairs for use during
sample preparation and analytical activities. The rental
cost for the tent (including detachable sides, ropes, poles,
and pegs) was $270 per week. The rental cost for the
table and two chairs for one week totaled $6. Total support
equipment costs were $276 per week for rental.
61
-------
For longer projects, purchase of support equipmentshou Id
be considered. Two folding chairs would cost
approximately $40. A 10x10 foot tent would cost between
$260 and $1,000, depending on the construction materials
and the need for sidewalls and other accessories (e.g.,
sand stakes, counterweights, storage bag, etc.). A cost of
$800 was used for this cost analysis. A folding table would
cost between $80 and $250, depending on the supplier.
For purposes of this cost analysis, $160 was used. Total
purchase costs for support equipment are estimated at
$1,000.
7.2.4 Labor Cost
MTI utilized two people to analyze 197 samples; a sample
preparation person spent almost two days performing
sample digestions and an instrument analyst spent three
days analyzing and reporting data results. Combined, the
total operational time was 2,280 minutes. Including
instrument setup and disassembly, the total labor time
expended during the demonstration was roughly 3,000
minutes. This time correlates to 50 hours, or five 10-hour
days for one individual. Based on a labor rate of $300 per
day, total labor cost for using the PDV 6000 during the
demonstration was $1,500.
Labor costs for the hypothetical site assume qualified
technicians are available locally (i.e., no lodging or per
diem costs are apply). Table 7-4 summarizes labor costs
for various operational periods, assuming 21 workdays per
month (on average) and 252 work days per year. The
costs presented do not separate supervision and quality
control because these would be associated with use of any
analytical instrument and are a portion of the overhead
multiplier built into the labor rate.
Table 7-4. Labor Costs
Item
1 3
Months
6
12
24
Technician $6,300 $18,900 $37,800 $75,600 $151,200
Supervisor NA NA NA NA NA
Quality Control NA NA NA NA NA
Total $6,300 $18,900 $37,800 $75,600 $151,200
NA = Not applicable.
7.2.5 Investigation-Derived Waste Disposal
Cost
MTI generated PPE waste, digestate solution with excess
soil waste, and waste electrolyte solution. The PPE waste
was charged to the overall project due to project
constraints. The minimum waste volume is a 5-gallon
container. Mobilization and container drop-off fees were
$1,040; a 5-gallon soil waste drum was $400, and a
5-gallon liquid waste drum was $400. (These costs were
based on a listed waste stream with hazardous waste
number U151.)
The total demonstration IDW disposal cost was $1,840.
These costs may vary significantly from site to site,
depending on whetherthe waste is classified as hazardous
or nonhazardous and whether excess sample material is
generated that requires disposal. Table 7-5 presents IDW
costs for various operational periods, assuming that waste
generation rates were similar to those encountered during
the demonstration.
It should be noted that when a large amount of digestion
samples are conducted for using the PDV 6000, the waste
digestion liquid would be poured out of each digestion
bottle and into a liquid waste container (as opposed to just
discarding the 70 ml_ digestion bottle containing the liquid,
as was done during the demonstration). The empty
digestion bottles could then be thrown away as non-
hazardous waste. Since there is 40 ml_ of digestion liquid
generated per sample analyzed, there would be
approximately 42 liters, or 11 gallons, of digestion liquid
generated per month (40 ml_ x 50 samples x 21 days).
Table 7-5. IDW Costs
Item
1 3
Months
6 12
24
Drop Fee $1,040 $3,120 $6,240 $12,480 $24,960
Disposal" $1,400 $2,000 $3,400 $6,000 $10,000
Total
$2,440 $5,120 $9,640 $18,480 $34,960
Disposal costs are estimated at $1,400 for a 20-gallon
container and $2,000 for 55-gallon drum. The most
economical combination of containers is used for disposal cost
estimates.
7.2.6 Summary of PD V 6000 Costs
The total cost for performing mercury analysis is
summarized in Table 7-6. This table reflects costs for
project durations of from 1 to 24 months. The purchase
option was used for estimating the equipment cost. For the
first month of use, the total cost is primarily the instrument
purchase and labor. As time progresses, supplies and
labor become dominant cost categories, the sample
62
-------
preparation kits being the dominant supplies cost. When
operating the PDV 6000 for periods in excess of three
months, it may be more economical to prepare the
Table 7-6. Summary of Purchase Costs for the PDV 6000
Item
Quantity Unit Unit Cost
($)
reagents and purchase the disposable items, as opposed
to purchasing the sample extraction kits.
Total Cost for Selected Project Duration
1-Month 3-Month
6-Month
12-Month 24-Month
Capital Equipment3
Purchase
1
NA
$8,100
$8,100
$8,100
$8,100
$8,100
$8,100
Supplies b
Reagents and Disposable
Micro Pipetters (set of 3)
Repeating Pipetter
Portable Scale
Total Supply Cost
Support Equipment a c
Table (optional) - weekly
Chairs (optional) - weekly
Tent (for inclement
Total Support Equip. Cost
Labor
Field Technician
IDW
Drop Fee
Disposal
Total IDW Costs
Total Cost d
a Costs are rounded to a
b The number of sample
1
1
1
1
-
1
2
1
1
NA
NA
each
set
each
each
each
each
each
hour
week
$84
$690
$320
$110
$5
$1
$270
$38
$1,040
$400
maximum of three significant digits.
preparation kits (or reagents) provides
$8,800
$690
$320
$110
$9,900
$20
$10
$800
$6,300
$1,040
$1,400
$2,440
$28,000
sufficient
$8,900
$690
$320
$110
$10,000
$60
$25
$800
$18,900
$3,120
$2,000
$5,120
$43,000
supplies to analyze
$16,000
$690
$320
$110
$17,000
$120
$40
$800
$960
$37,800
$6,240
$3,400
$9,640
$74,000
50 samples
$31,000
$690
$320
$110
$32,000
$160
$40
$800
$1,000
$75,600
$12,480
$6,000
$18,480
$136,000
$62,000
$690
$320
$100
$63,000
$160
$40
$800
$1,000
$151,000
$24,960
$10,000
$34,960
$260,000
per day. Significant cost savings
are achieved if reagents are purchased in bulk. The 1-month cost uses sample preparation kits, all other months are based upon preparing
reagents and purchasing disposable supplies (e.g., beakers, pipette tips).
c Rental costs were used through the 3-month period for chairs and the 6-month period for the table. Purchase costs were used for longer
periods. Purchase costs for the tent were used for all periods.
d Totals are rounded to two significant digits.
From these data, it is apparent that the cost to purchase
the PDV 6000 instrument and accompanying software
becomes relatively insignificant after 6 months of continual
use. This may imply that the instrument purchase option
is the most cost-effective in many instances; however, the
decision on which procurement option to utilize should be
made on a case-by-case basis.
An alternative to the vendor-supplied kits may be desirable
due to the relatively high cost of the kits when used over an
extended period of time. MTI will supply the reagent
concentrations to anyone interested is preparing their own
extraction reagents.
Table 7-7 summarizes costs for the actual demonstration.
Note that the 1-month rental cost of the PDV 6000 was
used for capital costs.
63
-------
The cost per analysis based on 197 samples when renting
the PDV6000 is $43.74 per sample. Thecostperanalysis
for the 197 samples, excluding instrument cost, is $32.57
per sample.
Table 7-7. PDV 6000 Costs by Category
Category
Instrument
Supplies b
Support
Equipment
Labor
IDW Disposal
Total
Category Cost
$2,200
$2,800
$276
$1,500
$1,840
$8,616
Percentage of
Total Costs a
25.5%
32.5%
3.2%
17.4%
21.4%
100.0%
The percentages are rounded to one decimal place; the total
percentage is 100%.
Includes 20 sample preparation kits, a portable scale, 3-uL
capacity pipetters, and a milliliter capacity pipetter.
7.3 Typical Reference Method Costs
This section presents costs associated with the reference
method used to analyze the demonstration samples for
mercury. Costs for other project analyses are not covered.
The referee laboratory utilized SW-846 Method 7471 B for
all soil and sediment samples. The referee laboratory
performed 421 analysis over a 21-day time period.
A typical mercury analysis cost, along with percent
moisture for dry-weight calculation, is approximately $35.
This cost covers sample management and preparation,
analysis, quality assurance, and preparation of a data
package. The total cost for 197 samples at $35 would be
approximately $6,895. This is based on a standard
turnaround time of 21 calendar days. The sample
turnaround time from the laboratory can be reduced to 14,
7, or even fewer calendar days, with a cost multiplier
between 125% to 300%, depending upon project needs
and laboratory availability. This results in an approximate
cost range from $6,895 to $20,685. The laboratory cost
does not include sample packaging, shipping, or downtime
caused to the project while awaiting sample results.
64
-------
Chapter 8
Summary of Demonstration Results
As discussed previously in this ITVR, the MTI PDV 6000
was evaluated by having the vendor analyze 197 soil and
sediment samples. These 197 samples included high-,
medium-, and low-concentration field samples from four
sites, SRMs, and spiked field samples. Table 8-1 provides
a breakdown of the numbers of these samples for each
sample type and concentration range or source.
Collectively, these samples provided different matrices,
concentrations, and types of mercury needed to perform a
comprehensive evaluation of the PDV 6000.
8.1 Primary Objectives
The primary objectives of the demonstration were centered
on evaluation of the field instrument and performance in
relation to sensitivity, accuracy, precision, time for analysis,
and cost. Each of these objectives was discussed in detail
in previous chapters and is summarized in the following
paragraphs. The overall demonstration results suggestthat
the experimental design was successfulforevaluation of the
MTI PDV 6000. Quantitative results were reviewed. The
results from this instrument were found not to be
comparable to standard analyses performed by the
laboratory in terms of precision and accuracy, and the
collected data provide evidence to support this statement.
The two primary sensitivity evaluations performed for this
demonstration were the MDL and PQL. Following
procedures established in 40 CFR Part 136, the MDL is
between 1.67 and 3.67 mg/kg. The equivalent MDL for the
referee laboratory is 0.0026 mg/kg. Examples from
analyzed samples, however, suggestthatthe MTI MDL may
be closer to 0.811 mg/kg or lower. Values detected at these
lowerlevels; however, would likely be highly inaccurate and
should only be considered as a "positive hit" without any
implied accuracy or precision. The calculated MDL is only
intended as a statical estimation and not a true test of
instrument sensitivity.
The low standard calculations suggest that a PQL for the
MTI field instrument is 4-8 mg/kg. The %D for the average
MTI result for a sample concentration of 4.75 mg/kg is
46%. The referee laboratory PQL confirmed during the
demonstration is 0.005 mg/kg with a %D of 10% or less,
based upon a lower calibration standard. Both the MDL
and PQL were determined for soils and sediments.
Accuracy was evaluated by comparison to SRMs and
comparison to the referee laboratory analysis for field
samples. This included spiked fie Id samples for evaluation
of additional concentrations not otherwise available. In
summary, MTI data were within the SRM 95% prediction
intervals about 50% of the time. ALSI data compared to
SRM values were within the 95% prediction interval 89% of
the time. The comparison between the MTI field data and
the ALSI results suggests that the two data sets are not
different, but the similarity for individual samples is often
the result of high variability associated with the MTI
reported values.
In determining the number of results significantly above or
below the value reported by the referee laboratory, the
number of MTI average values greater than 50% different
from the referee laboratory results or SRM reference
values was only 6 for 21 different sample lots and those
greater than 100% different were only 2 for 21 different
sample lots. MTI results, therefore, appear to provide a
rough estimate of accuracy forfield determination, and may
be affected by interferences not identified by this
demonstration. It should be concluded, however, thatthe
MTI PDV 6000 did not compare well to laboratory Method
7471B in terms of obtaining accurate analyses of mercury
in soil.
65
-------
Precision was determined by analysis of replicate samples.
The precision of the MTI field instrument did not compare
well to the measured laboratory precision. The overall RSD
for MTI is 35.1% which is above the 25% RSD objective set
for the laboratory. The overall laboratory RSD is 22.3%.
Time measurements were based on the length of time the
operator spent performing all phases of the analyses,
including setup, calibration, and sample analyses (including
all reanalyses). MTI analyzed 197 samples in 2,280
minutes over three days, which averaged to 11.6 minutes
per sample result. Based on this, an operator could be
expected to analyze 41 samples (8 hours x 60 minutes •*•
11.6 minutes/sample) in an 8-hour day.
Cost of the MTI sample analyses included capital, supplies,
labor, support equipment, and waste disposal. The cost per
sample was calculated both with and without the cost of the
instrument included. This was performed because the first
sample requires that the instrument is either purchased or
rented, and as the sample number increases, the cost per
sample would decrease. A comparison of the field MTI
cost to off-site laboratory cost was not made. To compare
the field and laboratory costs correctly, it would be
necessary to include the expense incurred to the project
due to waiting for analysis results to return from the
laboratory (potentially several mobilizations and
demobilizations, stand-by fees, and other aspects
associated with field activities).
Table 8-2 summarizes the results of the primary
objectives.
8.2 Secondary Objectives
Table 8-3 summarizes the results of the secondary
objectives.
Table 8-1. Distribution of Samples Prepared for MTI and the Referee Laboratory
Site
Concentration Range
Soil
Sediment
Sample Type
Spiked Soil
SRM
Carson River
(Subtotal = 48)
Puget Sound
(Subtotal = 51)
Oak Ridge
(Subtotal = 54)
Manufacturing
(Subtotal = 44)
Subtotal
Low(1-500ppb)
Mid (0.5-50 ppm)
High (50->1, 000 ppm)
Low (1 ppb - 10 ppm)
High (10-500 ppm)
Low (0.1 -10 ppm)
High (10-800 ppm)
General (5-1,000 ppm)
3
0
0
13
0
0
13
23
52
10
0
0
0
10
3
10
0
33
7
0
7
7
7
0
0
7
35
0
7
14
3
11
14
14
14
77
66
-------
Table 8-2. Summary of PDV 6000 Results for the Primary Objectives
Demonstration Performance Results
Objective Evaluation Basis PDV 6000
Instrument
Sensitivity
MDL. Method from 40 CFR Part 1 36 Between 1 .67 and 3.67
mg/kg
Reference Method
0.0026 mg/kg
Accuracy
Precision
PQL. Low concentration SRMs and
samples.
Comparison to SRMs, field, and spiked
samples covering the entire range of the
instrument calibration.
Determined by analysis of replicate samples
at several concentrations.
Approximately 4-8 mg/kg 0.005 mg/kg
Mil's PDV 6000 is not within expected accuracy for
laboratory analyses. Mil's field instrument was only
within the SRM 95% prediction intervals about 50% of the
time. The comparison between the MTI field data and the
ALSI results suggests that the two data sets are not
different but similarity for individual samples is often the
result of high variability associated with the MTI reported
values. The MTI PDV does, however, appear to provide
a rough estimate of mercury concentrations in soil and
sediment.
Overall RSD for the MTI PDV 6000 was computed to be
35.1% compared to the referee laboratory RSD of 22.3%.
This is a combined measure of precision which includes
sampling and aliquoting variations.
Time per Analysis
Cost
Timed daily operations for 3 days and
divided the total time by the total number of
analyses.
Costs were provided by MTI and
independent suppliers of support equipment
and supplies. Labor costs were estimated
based on a salary survey. IDW costs were
estimated from the actual costs encountered
at the Oak Ridge demonstration.
Two MTI representatives performed all setup, calibration
checks, sample preparation, sample analysis, and
equipment disassembly. Individual analyses took 7.5
minutes each, but the total time for preparation and
analysis averaged approximately 11.6 minutes per
sample.
The cost per analysis, based on measurement of 197
samples, when incurring a minimum 1-month rental fee
for the PDV 6000, is $43.74 per sample. Excluding the
instrument rental cost, the cost for analyzing the 197
samples is $32.57 per sample. The total cost for
equipment rental and necessary supplies during the
demonstration is estimated at $8,600. The cost breakout
for the demonstration by category is: capital costs,
25.5%; supplies, 32.5%; support equipment, 3.2%; labor,
17.4%; and IDW disposal, 21.4%.
67
-------
Table 8-3. Summary of PDV 6000 Results for the Secondary Objectives
Demonstration
Objective
Evaluation Basis
Performance Results
Ease of Use
Field observations during the demonstration.
Health and Safety
Concerns
Portability of the
Device
Instrument
Durability
Observation of the equipment, operating
procedures, and any equipment certifications
during the demonstration.
Review of device specifications,
measurement of key components, and
observation of equipment setup and tear
down before, during, and after the
demonstration.
Observation of the equipment design and
construction, and evaluation of any
necessary repairs or instrument downtime
during the demonstration.
The instrument appears to be easy to operate as a
stand-alone unit or in conjunction with the VAS software.
Training on the unit and the VAS software would be
recommended for first-time users. A laboratory or field
technician with a basic knowledge of chemistry and basic
computer skills could operate the equipment after a 1-day
training course.
No significant health and safety concerns were noted for
the PDV 6000. The main potential health and safety
concern observed during the demonstration was potential
exposure to fumes resulting from the reactions that
occurred during sample preparation. This observation
emphasizes the need for PPE such as safety glasses,
gloves, and possibly a laboratory apron when conducting
acid digestions.
The PDV 6000 was easily portable due to its compact
size and weight, and by the method by which it is
transported (i.e., the unit and instrument accessories are
transported within a briefcase size carrying case. The
unit was easily set up by MTI and was easily carried while
walking. The instrument can be characterized as truly
field portable.
The PDV 6000 control unit is well designed and
constructed for durability. The hard shell carrying case
used to transport the unit and ancillary equipment
provides adequate protection of electrodes and other
sensitive components.
Availability of the
Vendor
Instruments and
Supplies
Review of vendor website and telephone
calls to the vendor after the demonstration.
Per MTI, the PDV 6000 units are stocked in the U.S. and
available for purchase, rent, or lease within one week of
order placement through the new MTI U.S. operations.
Spare parts and consumable supplies can be added to
the original PDV 6000 order. Supplies not typically
provided by MTI (pipetters and laboratory scale) are
readily available from laboratory supply firms.
68
-------
Chapter 9
Bibliography
Anchor Environmental. 2000. Engineering Design
Report, Interim Remedial Action Log Pond Cleanup/
Habitat Restoration Whatcom Waterway Site,
Bellingham, WA. Prepared for Georgia Pacific West,
Inc. by AnchorEnvironmental, L.L.C., Seattle, WA. July
31, 2000.
Confidential Manufacturing Site. 2002. Soil Boring Data
from a Remedial Investigation Conducted in 2000.
Monitoring Technologies International Pty. Ltd. 2002a.
PDV 6000 - Portable Heavy Metals Analyzer,
Operation Manual, version 2.2.
Monitoring Technologies International Pty. Ltd. 2002b.
Voltammetric Analysis System - VAS User's Guide,
version 2.1.
Rothchild, E.R., R.R. Turner, S.H. Stow, M.A. Bogle, L.K.
Hyder, O.M. Sealand, H.J. Wyrick. 1984. Investigation
of Subsurface Mercury at the Oak Ridge Y-12 Plant.
Oak Ridge National Laboratory, TN. ORNL/TM-9092.
U.S. Environmental Protection Agency. 1994. Region 9.
Human Health Risk Assessment and Remedial
Investigation Report - Carson River Mercury Site
(Revised Draft). December 1 994.
U.S. Environmental Protection Agency. 1995.
Contaminants and Remedial Options at Selected
Metal-Contaminated Sites. July 1995. Washington
D.C. EPA/540/R-95/512.
U.S. Environmental Protection Agency. 1996. Test
Methods for Evaluating Solid Waste,
Physical/Chemical Methods, SW-846 CD ROM, which
contains updates for 1986, 1992, 1994, and 1996.
Washington DC.
U.S. Environmental Protection Agency. 1998.
Unpublished. Quality Assurance Project Plan
Requirements for Applied Research Projects, August
1998.
U.S. Department of Energy. 1998. Report on the
Remedial Investigation of the Upper East Fork of
Poplar Creek Characterization Area at the Oak Ridge
Y-12 Plant, Oak Ridge, TN. DOE/OR/01-1641&D2.
U.S. Environmental Protection Agency. 2002a. Region
9 Internet Web Site, www.epa.gov/region9/index. html.
U.S. Environmental Protection Agency. 2002b.
Guidance on Data Quality Indicators. EPA G-5i,
Washington D.C., July 2002.
U.S. Environmental Protection Agency. 2003. Field
Demonstration and Quality Assurance Project Plan -
Field Analysis of Mercury in Soil and Sediment.
August 2003. Washington D.C., EPA/600/R-05/053.
Wilcox, J.W., Chairman. 1983. Mercury at Y-12: A
Summary of the 1983 UCC-ND Task Force Study.
Report Y/EX-23, November 1983.
www.Fishersci.com, 2003.
www.monitoring-technologies.com, 2003.
www.mti.com.au, 2003.
69
-------
Appendix A
MTI Comments
Site Demonstration Comments
The ITVR demonstration showed that the PDV 6000 can
produce data in the field that can confidently be used to
effectively manage on site investigations and remediation
projects. MTI is a small company and values the
opportunity to compare its method to a fully validated
reference method. MTI would like to thank the teams at
SAIC and the EPA for setting up the program and looking
after the vendors in such a professional way.
MTI completed all 197 samples in a 3-day working period,
2 days using 2 technicians and the third day using just 1.
This time would have been significantly reduced had the
PDV 6000 method been used to classify soil samples into
set concentration bands as against a more quantitative
analysis.
The PDV 6000 when used in the field is designed to
provide screening data and is not expected to generate
laboratory quality data over a wide concentration range.
The instrumentis calibrated using a single pointcalibration,
either by the method of curve comparison or standard
addition. The concentration selected to calibrate the PDV
6000 is equivalent to the pollutant action level for that site.
This concentration is simply achieved by varying the
volume of standard added to the analysis cup. The amount
of sam pie extract analysed can also be varied to reflect the
expected concentration range in the soil. This approach
has two benefits:
1. It is very easy to change the calibration range to
reflect changing pollutant concentrations in the
soil.
2. It gives the best accuracy around the important
action concentration without the need to generate
multipoint calibration curves and therefore allows
for significantly faster sample throughput.
The PDV 6000 field analysis procedure has been designed
to give the best accuracy around the site specific pollutant
concentration that drives the investigation or remediation.
For example, if the threshold for contamination is 5 mg/kg
the standard concentration and analysis procedure sets
5mg/kg as the midpoint of the calibration. This will provide
accurate results within a 1-10 mg/kg range (x10 range) for
the samples analysed. The PDV calibration curve is linear
over 2 orders of magnitude and therefore a single point
calibration is sufficient to cover 1 order of magnitude.
Table A-1 shows the volumes and analysis methods
recommended for the commonly encountered
contamination ranges.
This appendix was written solely by MTI. The statements presented in this appendix represent the developer's point of view and summarize
the claims made by the developer regarding the PDV 6000. Publication of this material does not represent EPA's approval or endorsement of
the statements made in this appendix; performance assessment and economic analysis results for the PDV 6000 are discussed in the body of
the ITVR.
70
-------
Table A-1. Calibration Standard and Analysis Information
Target Soil
Concentration (mg/kg)
0.5-5
2-20
10- 100
100-1000
Volume of sample
extract (ml) to add to
analysis cup
5
2
1
0.1
Volume of Hg
Electrolyte (ml) to add
to analysis cup
15
20
20
20
Deposit time
(seconds)
200
100
60
30-60
Hg calibration
concentration
(mg/ml)
0.025-0.075
0.01 -0.1
0.025-0.25
0.025-0.25
The above procedure could not be adopted for this ITVR
evaluation, as a quantitative result was wanted across a
very wide concentration range. The concentration ranges
encountered were:
Carson River: low x500 (0.001 - 0.5 ppm), mid x100 (0.5
- 50 ppm), high x200 (50 - >1000 ppm)
Puget Sound: low x10,000 (0.001 -10 ppm, highxSO (10 -
500 ppm)
Manufacturing Site: x200 (5 - 1000 ppm)
Oak Ridge: low x100 0.1 - 10 ppm), high x80 (10 - 800
ppm)
This meant that the procedure used to analyze the sample
deviated from the standard operating procedure for the
PDV 6000. The data generated however clearly shows that
the PDV 6000 method has similar precision and accuracy
to the laboratory method when the calibration is close to
the actual concentration of mercury in the sample. This
makes the PDV 6000 perfectly acceptable for managing
site investigations and remediations.
It should also be noted that in this evaluation, all of the
samples had been carefully prepared to ensure they were
virtually identical for all the vendors. In the field, the issue
of sample homogenization is critical. The PDV 6000
procedure uses a 2 g sample that can be prepared on site
using a very quick mixing procedure from an initial 500 -
1000g sample. Two grams are considered to be the
minimum amount of sample that can be used to get a
representative field analysis for soil types that include,
clays, chalks and large particles of other porous stones.
The relevance of any field or laboratory analysis data will
only be good if the sampling is statistically meaningful and
sample size is an important factor in this. The smaller the
sample analyzed on site, the more samples that must be
analyzed to provide meaningful data.
Comments on SRM Precision and Accuracy
Interpretation
As discussed above, quantifying over wide concentration
ranges is not ideal for the PDV as a single calibration point
cannot cover the entire range. The PDV operators in this
situation therefore selected the mid point of the expected
sample concentration range to set the calibration to,
knowing that it would be less accurate at either end of the
range. It was hoped to initially screen the samples and
those that fell out of the calibration range could be re-
analyzed during the course of the evaluation. The
emergencies unfortunately prevented this from being done
for all of these samples. The data generated was still
submitted and MTI accepted that this would compromise
the statistical analysis of the results at either end of the
range.
Of the SRM sets analyzed, 57% of the target
concentrations were at the extreme limit of the sample set
and therefore out of the effective calibration range of the
PDV. Where the SRMs were within the calibration range of
the PDV, good agreement with the expected concentration
was achieved, together with reasonable precision. (lot# 48
%RSD 26.8, lot#49 %RSD 28.2, lot# 50 %RSD 38.2, lot#
62 %RSD 21.1, lot# 63 %RSD 34.8 and lot# 66 %RSD
26.6), which compares to an average 25% RSD for the
laboratory analysis. Not shown in the report is the data for
SRM lot# 46 (%RSD 20), which on extraction evolved large
amounts of Bromine. The targetvalue was 21 ppm and the
PDV method averaged 18 ppm over 5 samples analyzed
using the 10 to 100 ppm calibration range with a % RSD of
15.3. Sample lot# 44 also evolved bromine, but these were
reported as below detection limits using the 1-10 ppm
calibration range, even though the expected concentration
was 4.7 ppm. For this sample lot 5 ml_ of sample extract
were added to the electrolyte, compared to 1 ml_ for lot#
This appendix was written solely by MTI. The statements presented in this appendix represent the developer's point of view and summarize
the claims made by the developer regarding the PDV 6000. Publication of this material does not represent EPA's approval or endorsement of
the statements made in this appendix; performance assessment and economic analysis results for the PDV 6000 are discussed in the body of
the ITVR.
71
-------
46. The extra Bromine in the analysis cup could have
allowed the formation of the relatively insoluble HgBr2. It is
suggested that the evolution of bromine on site would
indicate that there was a serious problem with the soil in
the area and further investigation of the site would be
needed. The laboratory was unable to accurately detect
mercury in either of these brominated SRMs.
Lot# 52 was within the calibration range, but both the
laboratory and the PDV method reported significantly lower
concentrations than expected.
The low concentration SRM results from the initial analysis
using a 1-10 ppm calibration range did not detect any
Mercury and only a trace for those samples with 0.6 ppm
and higher. The exception was samples from lot#38 that
showed good peaks. The analysis was then repeated using
the very low level method. The mercury concentrations
using the low level method were typically reported as up to
5 times the expected concentration. For these samples
(Lot#35, 37 and 57) 10 ml_ of sample and 10 ml_ of
electrolyte were used with a 120-180 second deposit time.
These samples were run sequentially, using the same
batch of electrolyte. The initial matrix blank indicated no
mercury. The sample analysis however all showed a
significant peak at the Mercury position. This shows
mercury was in the analysis cup. The results for lot# 2 and
35 which are below the detection limits of the PDV indicate
approximately 40 parts per billion of mercury was in the
analysis cup. This is equivalentto 1.6 ppm in the soil. All of
these samples stood in the extraction bottle for over 24
hours before analysis, which is not in the usual SO P, where
analysis is required within a few hours of the extraction.
The better variability in %RSD between the SRMs and the
field samples could be due to the chemical form of the
mercury in the sample. The field samples have been in
contact with the Mercury for several decades and as such
the mercury can form tighter associations with the matrix.
SRMs due to their method of production are often highly
processed with very small particle sizes and rarely contain
elemental Hg or Hg in amalgams. The PDV 6000
extraction procedure is less effective for these forms of Hg
which were present in the field samples. Variability within
the field sample matrix, even though some processing had
occurred would also vary the extraction efficiency.
The results obtained from the SRMs shows that the PDV
6000 field method gives acceptable precision and accuracy
when used in accordance with its designed operating
parameters.
Comments on the Precision and Accuracy
Interpretation of Field samples
The PDV 6000 field extraction is the main cause of
variability between the results within each sample lot. This
extraction method is not time or temperature controlled and
cannot be expected to give absolutely consistent or 100%
extraction efficiency. It is expected that approximately 85%-
100% of the Mercury will be extracted during the initial acid
digest and peroxide addition, where temperatures will
reach 80 °C. The addition of water and electrolyte does not
completely neutralize the solution, as residual acidity is
needed to keep any mercury extracted in solution. This
means further extraction may occur over time. It was
however decided to keep the soil in the extraction solution
before analysis and not to decant sample to a separate
container in order to reduce the waste generated in the
field and reduce cross contamination of the low
concentration range samples.
In some cases the samples were re-analyzed at the end of
the day using the correct calibration range, which means
thatthe sample remained in the extraction solution for over
5 hours. The combination of a more accurate calibration
and longer extraction time will have given the much wider
%RSD observed for many of the sample batches as the
statistics will include results from both sets of analysis.
For example, lot# 13 from the Manufacturing Site gives an
average of 15 ppm for 5 samples when the actual
concentration is 5.5 ppm. However, 2 of those samples
were re-analyzed using the lower concentration calibration
range and gave an average of 5.15 ppm with a %RSD of
17.8. Sample lot# 32, also from the manufacturing site,
gives an average concentration from 5 samples of 875.5
ppm with a %RSD of 40. If the one high value reported is
removed from this set of results, the average value is 700
ppm with a %RSD of 3. The expected result was 650 ppm.
It is likely thatthe high sample result was genuine because
the laboratory also found one sample in this batch with
more than twice the concentration of the other samples.
This appendix was written solely by MTI. The statements presented in this appendix represent the developer's point of view and summarize
the claims made by the developer regarding the PDV 6000. Publication of this material does not represent EPA's approval or endorsement of
the statements made in this appendix; performance assessment and economic analysis results for the PDV 6000 are discussed in the body of
the ITVR.
72
-------
Forsoilsamples above 1,000ppm, the extraction efficiency
will be lower depending on the competing effects of other
compounds in the matrix neutralizing the acid and the
physical form of the Mercury. Elemental Mercury and
amalgams are the most difficult to dissolve.
In practice however, soils containing over 310 ppm are
usually considered highly contaminated and the PDV and
laboratory methods both gave similar results for samples
in this range.
The field samples that theoretically should have had less
than 1 ppm in (Lot# 2,5,11) were detected as more than
1 ppm by the PDV 6000 method. In every case the raw data
shows a significant peak at the mercury position. These
samples were again exposed to the extraction liquid for 24
hours and were also analyzed using the very low level
method. Lot#1 was only analyzed at the 1-10 ppm range
and showed no mercury present. Sample lot#4 was
analyzed as part of a separate batch of sam pies, and gave
an average concentration of 0.28 ppm compared to the
laboratory 0.11 ppm average. Sample lot#6 out of 5
samples were analyzed in the same batch as lot#4 and
showed less than 0.5 ppm in the sample. The raw data
gives an average of 0.46 for 4 samples for the values
above 0.25 ppm estimated and the laboratory estimated
0.23 ppm. The other 3 samples from the set gave an
average concentration of 0.19 ppm. These results were not
submitted in the data set as the standard run with these
samples was accidentally deleted while calculating the
results. The estimated concentrations are based on a
comparison with a standard run using different analysis
times.
The results from sample lots# 1, 4 and 6 indicate that the
values obtained for the other low concentration samples
may be contamination in the reagents, possibly the
electrolyte used to dilute the initial extract.
Comments on the Calculated MDL and PQL
The lowest quantifiable concentration of mercury detected
by the PDV 6000 is 2.5 micrograms per liter. The dilution
effect of the soil extraction procedure and the volume of
the extract added to the analysis cup would mean 2.5 ppb
is equivalent to 0.1 mg/kg in the soil. At this level of
sensitivity, sample or reagent contamination during
extraction or handling may cause an erroneous result.
The very low concentration samples were analyzed using
different calibration ranges. The reported results are from
the 0.1-1 ppm calibration range. The initial results at the
1-10 ppm range did not detect Mercury above 1ppm,
indicating all samples were below 1 ppm. The subsequent
analysis for samples on lot# 2,5, 11, 35, 37 and 57 using
the same batch of extraction reagents gave results greater
than 1 ppm. Unfortunately, no blank was analyzed for the
extraction reagent and contamination from this part of the
procedure cannot be ruled outdue to the close proximity of
samples and extraction reagents in this trial.
All of the low level sam pies showed Mercury peaks that are
significantly higher than the 2.5 ppb peak. This does
indicate that Mercury was present in the analysis solution,
but the source is not clear. The end result is that in this trial
the true MDL could not be determined from the data set.
The PQL estimated in the report is also slightly high.
The sample lots analyzed in the low level range have
concentrations that jump from 0.6 ppm to 4.75 ppm. This
large jump makes it difficult to calculate a PQL from the
data at anything less than 4 ppm. If the estimated
contribution by contamination is removed from the results
obtained and the results from sam pie lots# 4 and 6, a PQL
of around 1 ppm is obtained. This is in line with MTI's
reported field PQL of between 0.5 and 1 ppm.
Comments on the Economic Analysis
When evaluating the costs of an instrument it should also
be taken into consideration any other analyses that can be
carried out with it. The PDV 6000 can be used to analyze
soil and water samples for many toxic heavy metals,
including As, Cd, Cr, Cu, Hg, Ni, Pb, Sb, Sn, Tl, Te and Zn
with low part per billion detection limits. In most cases the
extraction procedure is the same forall metals. This means
only a few extra reagents such as standards would be
required in order to analyze a much wider suite of metals.
For sites where several metals a re present, the per sam pie
cost of the PDV would be considerably reduced.
This appendix was written solely by MTI. The statements presented in this appendix represent the developer's point of view and summarize
the claims made by the developer regarding the PDV 6000. Publication of this material does not represent EPA's approval or endorsement of
the statements made in this appendix; performance assessment and economic analysis results for the PDV 6000 are discussed in the body of
the ITVR.
73
-------
Method Improvements that Have Been Implemented
The evaluation gave MTI a wonderful opportunity to
compare its method against a fully approved method on
samples that are genuine field samples. This information
has highlighted some improvements that have now been
incorporated into the product and analysis procedure.
The PDV instrument has had a software modification that
allows greater sensitivity to be achieved. This gives an
improved signal to noise ratio and will generate better
peaks for quantification. This will improve the instrument
repeatability.
For the very low detection limits, the longer analysis times
will allow any Mercury coming from contamination to
contribute to the total mercury result to a significantly
greater extent. Due to the volumes of extract used,
electrolyte blanks can only performed once on the bulk
electrolyte and not for each sample. The analysis
procedure has been modified to run bulk electrolyte blanks
more frequently to exclude contamination in this reagent as
a source of error.
The VAS software has been modified to automatically
calculate the soil concentration directly. This removes the
need to transfer data to another program for the final data
generation and saves several hours of labor.
This appendix was written solely by MTI. The statements presented in this appendix represent the developer's point of view and summarize
the claims made by the developer regarding the PDV 6000. Publication of this material does not represent EPA's approval or endorsement of
the statements made in this appendix; performance assessment and economic analysis results for the PDV 6000 are discussed in the body of
the ITVR.
74
-------
Appendix B
Statistical Analysis
Two separate hypothesis tests were used to compare the
referee laboratory samples to the vendor tested samples.
This appendix details the equations and information for
both of these statistical analyses. For purposes of this
appendix, we have chosen to call the test comparing
sample populations using a separate calculation for each
sample lot the "hypothesis test," and the statistical
comparison of the entire sample set (all 32 separate
sample lots) analyzed by the vendor and the laboratory the
"unified hypothesis test," also known as an "aggregate
analysis" for all of the sample lots.
Hypothesis Test
A hypothesis test is used to determine if two sample
populations are significantly different. The analysis is
performed based on standard statistical calculations for
hypothesis testing. This incorporates a comparison
between the two sample populations assuming a specified
level of significance. For establishing the hypothesis test,
it was assumed that both sample sets are equal.
Therefore, if the null hypothesis is rejected, then the
sample sets are not considered equal. This test was
performed on all sample lots analyzed by both MTI and the
referee laboratory. H0 and Ha, null and alternative
hypothesis respectively, were tested with a 0.01 level of
significance (LOS). The concern related to this test is that,
if two sample populations have highly variable data (poor
precision), then the null hypothesis may be accepted
because of the test's inability to exclude poor precision as
a mitigating factor. Highly variable data results in wider
acceptance windows, and therefore, allows for acceptance
of the null hypothesis. Conclusions regarding this analysis
are presented in the main body of the report.
To determine if the two sample sets are significantly
different, the absolute value of the difference between the
laboratory average XL and the vendor average xv is
compared to a calculated u. When the absolute value of
the difference is greater than u, then the alternate
hypothesis is accepted, and the two sets (laboratory and
vendor) are concluded to be different.
To calculate u, the variances for the laboratory data set
and the vendor data set are calculated by dividing their
standard deviations by the number of samples in their data
set. The effective number of degrees of freedom is then
calculated.
Where:
f = effective number of degrees of freedom
VL = variance for the laboratory results
nL = number of samples for the laboratory
data set
Vv = variance for the vendor results
nv = number of sam pies for the vendor data
set.
The degrees of freedom (f) is used to determine the
appropriate "t" value and used to calculate u at the 0.01
level of significance using the following:
75
-------
Unified Hypothesis Test
For a specified vendor, let Y,j be the measured Hg
concentration for the f replicate of the ith sample for
/=1,2,...,l and)= 1,2,...,Ji. LetX^= log(Y^), where log is the
logarithm to the base 10. Define x,log to be the average
over all log replicates for the Ith sample given by:
Where x M is approximately a chi-square random variable
with (1-1) degrees of freedom:
5 = 7
-'
tog
z'log
ilog
-J
'
log
and
Denote the estimate of the variance of the log replicates for
the Ith sample to be:
-1
Now for the reference laboratory, let Y',y be the measured
Hg concentration for the/1 replicate of the ith sample for
/ =1,2,...,!' and j = 1,2,. ...J',. Denote the reference
laboratory quantities X',j, x/, and s'2 defined in a manner
similar to the corresponding quantities for the vendor.
Assumptions: Assume that the vendormeasurements, Y,y,
are independent and identically distributed according to a
lognormal distribution with parameters u,and o2. That is,
X,y= log(Y,y) is distributed according to a normal distribution
with expected value u,and variance o2. Further, assume
that the reference laboratory measurements, Y',j, are
independent and identically distributed according to a
lognormal distribution with parameters u',and o'2.
The null hypothesis to be tested is:
HQ : /jj = fj'j + 5, for some S and i = I,..., I
against the alternative hypothesis that the equality does not
hold for at least one value of /.
The null hypothesis H0 is rejected for large values of:
Zi-i =
i-i
i-1
1 pool
Critical values for the hypothesis test are the upper
percentile of the chi-square distribution with (1-1) degrees
of freedom obtained from a chi-square table.
Results of Unified Hypothesis Test for MTI
SAIC performed a unified hypothesis test analysis to
assess the comparability of analytical results provided by
MTI and those provided by ALSI. MTI and ALSI both
supplied multiple assays on replicates derived from a total
of 31 different sample lots, be they field materials or
reference materials. The MTI and ALSI data from these
assays formed the basis of this assessment.
The statistical analysis is based on log-transformed
(logarithm base 10) data and uses a chi-square test for
equality of MTI and ALSI population means for given
sample lot. Equality of variances is assumed.
Initially, the null hypothesis tested was that, on average,
MTI and ALSI would produce the same results within a
given sample lot. This hypothesis is stated as
H10: (MTI Lot log mean) = (ALSI Lot log mean)
H10 was strongly rejected in that the chi-square statistic
was 941.12, which exceeds the upper 99th percentile of the
chi-square distribution with 31 degrees of freedom having
a value of 52.19.
The null hypothesis was rejected in part because MTI
results tended to exceed those from ALSI for the same
sample lot. To explore this effect, the null hypothesis was
revised to included a bias term in the form of
76
-------
H20: (MTI Lot log mean) = (ALSI Lot log mean) +(delta),
where delta is a single value that does not change from
one sample lotto another, unlike the lot log means. H20
was rejected strongly in that the chi-square statistic was
699.19, which exceeded the upper 99th percentile of the
chi-square distribution with 30 degrees of freedom with a
value of 50.89. In this analysis, delta was estimated to be
0.355 in logarithmic (base 10) space, which indicates an
average upward bias for MTI of 100355=12.265 or about
127%.
For both hypotheses, the large values of the chi-square
test statistics summarize the disagreement between the
MTI and ALSI analytical results. Furthermore, a review of
the statistical analysis details indicates that the overall
Table B-1. Unified Hypothesis Test Summary Information
Hypothesis Tota! SamP|e Excluded Lot DF
discordance between MTI and ALSI analytical results
cannot be traced to the disagreement in results for one or
two sample lots.
Summary information on these analyses is provided in
Table B-1. The p-value can be considered as a
significance level. This is a calculated value and usually
when one sets a p-value (e.g., 95% confidence level which
translates to a p-value of 0.05), this value is used to test
the level of significance for com parison. As noted in Table
B-1 the p-value is calculated from the test statistics and
therefore it can be seen that because the p-value is so
small (< 0.000000) the two sample populations are
considered to be non-equivalent and hence the large chi-
square value.
pool
Delta
Chi-square
P-value
H,,
31
31
None
None
31
30
0.02645
0.02645
0.0000
0.3546
941.12
699.19
0.000000
0.000000
77
------- |