k *
Office of Research and Development
National Homeland Security Research Center
A Review of Biological Agent Sampling
Methods and Application to a Wide-Area
Incident Scenario to Characterize Time
and Resource Demands
EPA
United States
Environmental Protection
Agency
EPA/600/R-17/176 I August 2017
www.epa.gov/homeland-security-research
-------
This page left intentionally blank
-------
EPA/600/R-17/176
May 2017
A Review of Biological Agent
Sampling Methods and Application
to a Wide-Area Incident Scenario to
Characterize Time and Resource
Demands
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
-------
Acknowledgments
Contributions of the following individuals and organizations to this report are gratefully
acknowledged:
U.S. Environmental Protection Agency (EPA) Project Team
Worth Calfee, Ph.D. (Principal Investigator, EPA, NHSRC)
Sang Don Lee, Ph.D. (EPA NHSRC)
Timothy Boe (EPA NHSRC)
Leroy Mickelsen (EPA OLEM CMAD)
Francisco J Cruz (EPA OLEM CMAD)
US EPA Technical Reviewers of Report
Lukas Oudejans, Ph.D. (EPA ORD NHSRC)
Claire Dixon, Ph.D. (UK DEFRA GDS)
Sara Casey (UK DEFRA GDS)
US EPA Quality Assurance
Eletha Brady Roberts
Ramona Sherman
Eastern Research Group, Inc. (ERG)
-------
TABLE OF CONTENTS
Disclaimer iii
List of Tables iv
List of Figures v
Acronyms and Abbreviations vi
Executive Summary viii
1 Introduction 1
2 Historical Anthrax Incidents 3
2.1 Amerithrax 3
2.2 Sverdlovsk Anthrax Leak 4
2.3 Gruinard Island 5
2.4 Lessons Learned 5
3 Overview of Sampling Methods 6
3.1 Swab Sampling 8
3.2 Sponge Wipe and Gauze Wipe Sampling (Wipe Sampling) 13
3.3 Vacuum Sampling 18
3.4 Air Sampling 20
3.5 Aggressive Air Sampling 21
3.6 Robotic Floor Sampling 21
3.7 Summary of Sampling Methods and Data 23
4 Wide-Area Incident Hypothetical Scenario 23
4.1 Introduction 23
4.2 Methodology 24
4.2.1 Hypothetical Scenario 24
4.2.2 Assumptions 24
4.2.3 Sampling Strategy 29
4.3 Results 32
4.3.1 Outdoor 32
4.3.2 Indoor 32
4.3.3 Underground Transit System 33
4.4 Observations 33
5 Wide-Area Incident Analysis 34
5.1 Introduction 34
5.2 Sampling Models 34
5.2.1 VSP Modeling 35
5.2.2 Regression Modeling Results 36
5.3 Resource Demands 39
5.3.1 Resource Demands Analysis 42
5.3.2 Resource Demands Results 42
i
-------
6 Observations and Recommendations 47
Bibliography 49
APPENDIX A. Literature Search Source Criteria and Keywords
APPENDIX B. Literature Review Scoring Criteria
li
-------
DISCLAIMER
The U.S. Environmental Protection Agency, through its Office of Research and Development, funded and
managed the research described here under Contract #EP-D-11-006 to Eastern Research Group, Inc. It has
been subjected to the Agency's review and has been approved for publication. Note that approval does
not signify that the contents necessarily reflect the views of the Agency. Mention of trade names,
products, or services does not convey official EPA approval, endorsement, or recommendation.
Questions concerning this document or its application should be addressed to:
M. Worth Calfee, Ph.D.
U.S. Environmental Protection Agency
Office of Research and Development
National Homeland Security Research Center
109 T.W. Alexander Dr. (MD-E-343-06)
Research Triangle Park, NC 27711
Phone 919.541.7600 Fax 919.541.0496
in
-------
LIST OF TABLES
Table 1. Number of Sources That Contained Information and/or Data Relevant to Sampling
Method/Technology 6
Table 2. Number of Sources That Contained Information and/or Data Relevant to This Study 7
Table 3. Summary of Available Data for Swab Sampling 8
Table 4. Summary of Available Data for Sponge Wipe Sampling 13
Table 5. Summary of Available Data for Gauze Wipe Sampling 16
Table 6. Summary of Available Data for Vacuum Sampling 18
Table 7. Summary of Available Data for Air Sampling 21
Table 8. Summary of Available Data for Robotic Floor Sampling 22
Table 9. Outdoor Surface Area 25
Table 10. Indoor Surface Area 25
Table 11. Underground Surface Distribution 26
Table 12. Outdoor Sample Medium to Surface Type Assignment 27
Table 13. Indoor Sample Medium to Surface Type Assignment 27
Table 14. Underground Transit System Sample Medium to Surface Type Assignment 27
Table 15. Sampling Time per Sample Medium 28
Table 16. Cost per Sample Medium 28
Table 17. Available Response Personnel per Environment and Sample Medium 28
Table 18. Individual Laboratory Throughput Factors per Sample Medium 29
Table 19. Waste Generated per Sampling Medium 29
Table 20. Summary of Sampling Inputs for the Outdoor Environment 31
Table 21. Summary of Sampling Inputs for the Indoor Environment 31
Table 22. Summary of Sampling Inputs for the Underground Transit System Environment 31
Table 23. Outdoor Samples Sampling Results 32
Table 24. Outdoor Samples Laboratory Analysis Results 32
Table 25. Indoor Sampling Results 32
Table 26. Indoor Analysis Results 33
Table 27. Underground Transit System Sampling Results 33
Table 28. Underground Transit System Analysis Results 33
Table 29. Input Parameters Used for VSP Modeling 35
Table 30. Range of Values Used for VSP Input Parameters 35
Table 31. Model Statistics with False Negatives in the Indoor Environment 36
Table 32. Model Statistics with no False Negatives in the Indoor Environment 36
Table 33. Model Statistics with no False Negatives in the Outdoor Environment 37
Table 34. Model Statistics with no False Negatives in the Underground Environment 37
Table 35. Model Statistics with False Negatives in the Underground Environment 37
Table 36. Variable Sensitivity with no False Negatives in the Underground Environment Model 38
Table 37. Variable Sensitivity with no False Negatives in the Outdoor Environment Model 38
Table 38. Variable Sensitivity with no False Negatives in the Indoor Environment Model 38
Table 39. Variable Sensitivity when Accounting for False Negatives in the Underground Environment
Model 39
Table 40. Variable Sensitivity when Accounting for False Negatives in the Indoor Environment Model. 39
Table 41. Table of Factors Used in Resource Demand Estimation Spreadsheet 40
Table 42. Resource Demand Estimation Spreadsheet Outputs 42
Table 43. Effect of Varying the Sample Area Size and Number of Sampling Teams on Total Cost and
Time to Complete Sampling 43
Table 44. Effect of Varying the Sample Area Size and Number of Analysis Laboratories on Total Cost
and Time to Complete Analyses 43
IV
-------
LIST OF FIGURES
Figure 1. Comparison of Surface Areas 25
Figure 2. Total Cost and Total Time for Sampling and Analysis as a Function of Sample Area Size 44
Figure 3. Sampling and Analysis Time as a Function of Number of Available Resources at 50 ft2 Sample
Area Size 45
Figure 4. Sampling and Analysis Time as a Function of Number of Available Resources at 500 ft2 Sample
Area Size 46
Figure 5. Sampling and Analysis Time as a Function of Number of Available Resources at 1,000 ft2
Sample Area Size 47
v
-------
ACRONYMS AND ABBREVIATIONS
Ba Bacillus anthracis
BOTE Bio-Response Operational Testing and Evaluation Project
CARC Chemical-Agent-Resistant Coating
CDC Centers for Disease Control and Prevention
CFU Colony-Forming Unit(s)
CI Confidence Interval
cm Centimeter
DNA Deoxyribonucleic Acid
DST Decision Support Tool
DoD U.S. Department of Defense
DOE U.S. Department of Energy
EPA U.S. Environmental Protection Agency
ESA European Space Agency
FNR False Negative Rate
ft Feet
GIS Geographic Information System
HEPA High-Efficiency Particulate Air
HSPD Homeland Security Presidential Directives
HSRP Homeland Security Research Program
HVAC Heating, Ventilation, and Air Conditioning
in Inches
INL Idaho National Laboratory
LOD Limit of Detection
m Meter
MCE Mixed Cellulose Ester
MLI Multilayer Insulation
N/U Not Used
NASA National Aeronautics and Space Administration
NHSRC National Homeland Security Research Center
NPS National Planning Scenario
ORD Office of Research and Development
PBC Polyester-Bonded Cloth
PBST Phosphate-Buffered Saline including Tween
PTFE Polytetrafluoroethylene
PVC Polyvinyl Chloride
RE Recovery Efficiency
RFC Robotic Floor Cleaner
RH Relative Humidity
SE Standard Error
SD Standard Deviation
TTC Time to Completion
UK United Kingdom
VI
-------
(.un Micrometer
USAMRIID U.S. Army Medical Research Institute of Infectious Diseases
VSP Visual Sample Plan
WEST Waste Estimation Support Tool
Vll
-------
EXECUTIVE SUMMARY
A large-scale aerosol release of a persistent, disease-causing biological agent can result in contamination
of a wide area, and may require significant time and resources for recovery depending on the severity of
adverse health effects on the exposed population(s). Many unknowns are associated with characterization
and clearance sampling during response to a wide-area (including indoor, outdoor, and underground area)
biological incident. The biological agent and its characteristics, the release mechanism, amount of
contaminant released, and a plethora of environmental and meteorological factors are completely
separate, yet interconnected processes that greatly influence the extent and level of contamination.
Similarly, decisions related to the sampling strategy (i.e., sample medium, sampling area, spacing, etc.)
will affect the cost, time, amount of waste generated, and personnel (i.e., resource demand) required to
characterize and clear the contaminated area. The process of understanding how these elements influence
one another and contribute to the overall problem is referred to as a systems approach. To what degree
sampling and, more specifically, variations in the sampling strategy interact and contribute to overall
resource demand, following a wide area biological incident, is still largely unknown. To date, there have
been no attempts to model characterization sampling following a wide-area biological incident.
The objectives of this study were fivefold: (1) to review available facts and information related to
historical wide-area biological contamination incidents; (2) to conduct a literature review to generally
assess the current state of knowledge regarding current sampling methods that may be used for
characterization sampling following a wide-area biological incident, specifically focusing on Bacillus
anthracis (Ba) contamination; (3) to summarize available data obtained from the literature for a
predetermined set of data elements (i.e., performance metrics) for current sampling methods and
technologies, (4) apply appropriate data from the literature and/or from recent operational field studies to
inform an analysis of the application of wide-area sampling strategies in the context of National Planning
Scenario (NPS) #2, and (5) evaluate the impacts of (a) variable sampling areas, hot spot areas,
probabilities, and false negatives rates; and (b) variable resource availability; on the total estimated cost
and time to complete sampling and analysis activities in a hypothetical scenario.
Three recent incidents (see Section 2) have illustrated the need for enhancement and validation of
sampling methods and development of decontamination and remediation plans for a biological release. In
each of those incidents, contamination occurred in a vast range of structures and sizes. The large costs
associated with decontamination in these cases support the necessity of appropriate sampling methods to
accurately characterize contamination and inform appropriate response efforts. No one remediation
strategy is universally applicable.
A small variety of established sampling methods have been used, and are currently recommended, for
many biological agents, including Ba. Some methods focus on small spaces: swab sampling, sponge
sampling, gauze sampling, and vacuum sampling; others are more applicable to larger areas or other
applications beyond the initial response stage of an incident: air sampling, aggressive air sampling, and
robotic floor sampling. This study focused on gathering data on surface types sampled, recovery
efficiencies, and sample area sizes for wipe, sponge, swab, and vacuum sampling. These three data
elements were identified as being the most critical for the purposes of further analyzing the impacts of
sampling strategy for a hypothetical wide-area incident, including impacts on cost, time, resource
demand, etc. The literature revealed a diverse variety of data on surface types, recovery efficiencies, and
sample area sizes. Surface types sampled included stainless steel, carpet, plastic, wood, glass, brick,
wallboard, various other metals, concrete, and fabrics including cotton, rayon, and polyester. Sample
areas also varied greatly depending on the source. The sampling methods reviewed and the aggregated
data highlights the vast range of reported RE values for differing surface types, spore concentrations, and
(in some cases) deposition methods. Additionally, the data were collected under largely controlled
-------
conditions and there is the possibility of potential biasing of the RE values due to the individual collecting
the samples or conducting the sampling. Furthermore, the identified sampling methods have been
characterized on the selected surfaces, mainly focused on indoor surfaces. A wide-area biological incident
may require sampling of indoor as well as outdoor surfaces and possibly underground transit systems. It is
expected that the sampling in a wide-area incident may require sampling very different surfaces compared
to (or in addition to) the ones evaluated in the studies above. Only a small number of surface sampling
methods are well-documented with their sampling protocols. It is uncertain, if a wide-area biological
incident happens, how the impacted areas would be characterized and how much resources would be
required with the currently-available sampling methods.
A statistically defendable sampling plan was developed to characterize a hypothetical city following a
hypothetical biological release (National Planning Scenario (NPS) #2). Realistic spatial, infrastructure,
and surface media information was gathered to determine the extent of sampling and to match surface
material types at each sampling location to their appropriate sampling method. Visual Sample Plan (VSP)
and a custom spreadsheet were used to estimate the sampling criteria, the number of samples, and the
resources needed to take those samples for outdoor, indoor, and underground environments. According to
the hypothetical sampling plan, environmental characteristics, resource demand assumptions, and the
assumption that Ba spores will remain persistent and detectable within the environment; characterization
of an assumed two square mile area with a total surface area of 4.76 * 10s ft2 using approximately 1,600
sampling personnel and 250 laboratories will cost an estimated 15.3 billion dollars in total and may take
over 10 years to characterize. The total time for sampling may be up to 4 years, while the total analysis
throughput is estimated to take 10 years. Although this disconnect is rather improbable, the results do
lend themselves to the possibility that analysis efforts may eclipse sampling efforts. The total indoor
surface area is one order of magnitude greater than the outdoor area, which is to be expected for highly-
populated urban setup. Because of the increased surface area, the indoor environments represent almost
90% of the total cost and resource demand of all the environments combined.
Because of this, for wide-area biological incidents, it is recommended that a rule in/rule out approach be
taken when sampling indoors. If a positive hit is discovered through preliminary sampling, the building
should be considered contaminated unless the decontamination approach is costlier than further sampling
to define more discrete decontamination zones. This analysis assumed that all buildings within the
contaminated area would either be contaminated or need to be sampled thoroughly. Available field data or
a lines of evidence approach may be useful in reducing this requirement. It is likely the requirements set
forth by this environment are not practical and would likely exhaust local, state, and federal resources.
Furthermore, by the time characterization has concluded (i.e., 10 years later), it is possible that the
community and its economy will no longer be viable.
The process to develop a sampling strategy for wide-area biological incident response operations has up
to now been poorly defined and not systematic. Additionally, the options available in terms of sampling
techniques are often constrained by regulatory mandates. In situations where the sampling method is not
mandated by regulation(s), then it is necessary to analyze available data concerning both sampling
methods and sampling techniques before selecting the most appropriate method or technique.
Leveraging the capacity of VSP and the results of the hypothetical scenario, a predictive model for wide-
area sampling strategies by way of regression modeling was developed to model the number of samples
needed for the underground, outdoor, and indoor environments individually under a range of values for
VSP input parameters (sampling area, hot spot area, false negative rate, and probability). The relative
impact of the VSP input parameters on the resulting number(s) of samples was analyzed, and the
modeling results were combined with resource demand estimates to analyze the potential impacts on total
time (for sampling and analysis) and total cost impacts following a wide area biological incident. Based
on the results, the total cost is a function only of the sample area size (i.e., as the area covered by each
IX
-------
sample collected increases, the number of samples required for a given area of interest decreases) and not
on either the number of sampling teams collecting those samples nor on the number of labs available to
analyze those samples. However, the amount of time needed to both collect the samples and to analyze
them is a function of the number of sampling teams and analysis labs available. Once the number of
samples is determined, the total cost is fixed, and is independent of the number of sampling teams and the
number of labs. Once the total cost is determined based only on the number of samples required, the
impact of a variable number of available sampling teams and analysis labs was evaluated.
Several needs for future research were identified based on this study. First, a lines-of-evidence approach
to sampling design should be considered at the outset of the response planning phase of a wide area
incident for the eventual purpose of reaching clearance goals. Second, we currently lack operational
strategies to deal with such an incident, and cohesive operational information must be developed to help
inform the development of those strategies. Third, the current sampling methods used to characterize a
wide area would result in an overwhelmingly unrealistic number of samples that must be collected and
analyzed. Existing availability of resources (sampling personnel and analytical laboratories) would not be
able to handle such a large demand in even a reasonable amount of time. Different sampling methods
must be developed to characterize the extent of contamination in wide areas. Those novel sampling
methods must focus on larger sampling area sizes, be usable for outdoor surfaces, must require less time
and personnel to collect, and thereby reduce the overall sampling and analysis burden by increasing the
area characterized by each sample. Improvements to existing sampling methods with respect to recovery
efficiencies would largely be inconsequential to reducing the resource demand (total cost and time to
characterize) compared to improvements to sample area size (even using probabilistic sampling design
approaches). Fourth, there is a lack of understanding concerning the fate and transport of spores over long
periods of time. This understanding is critical when considering the amount of time potentially required to
conduct characterization sampling. Spore fate and transport must be understood for effective sampling
and remediation strategy development. Lastly, this study supports further consideration of a combined
sampling design approach using probabilistic and non-probabilistic sampling when characterizing a wide
area incident. Additional decision support tools are needed to help direct sampling efforts for wide area
incidents where time and cost considerations are critical factors.
x
-------
1 INTRODUCTION
This project supports the mission of the U.S. Environmental Protection Agency's (EPA) Office of
Research and Development's (ORD) Homeland Security Research Program (HSRP) by providing
information relevant to the decontamination of areas contaminated as a result of an act of terrorism.
Under Homeland Security Presidential Directives (HSPDs) 5, 7, 8, and 10, EPA, in a coordinated effort
with other federal agencies, is responsible for "developing strategies, guidelines, and plans for
decontamination of... equipment, and facilities" to mitigate the risks of contamination following a
biological agent contamination incident.
EPA's National Homeland Security Research Center (NHSRC) aims to help EPA address the mission of
the HSRP by providing expertise and products that can be widely used to prevent, prepare for, and
recover from public health and environmental emergencies arising from terrorist threats and incidents.
NHSRC's mission includes providing expertise and guidance on the selection and implementation of
decontamination methods and providing the scientific basis for a significant reduction in time, cost, and
complexity of decontamination and waste handling activities.
A large-scale aerosol release of a persistent disease-causing biological agent can result in contamination
of a wide area and may require significant time and resources for recovery depending on the severity of
adverse health effects on the exposed population(s). Contamination would likely be heterogeneously
deposited (i.e., light and patchy) over a wide area and might migrate to clean areas by way of
resuspension and weathering over a short period (e.g., days). Although information on the agent type and
possibly some of its characteristics may be available during the response, an arduous environmental
sampling effort to determine the identity, concentration, viability, and approximate location of the
contamination would likely follow. Such an effort would be both costly and time-consuming.
Many unknowns are associated with characterization and clearance sampling during the response to a
wide-area (including indoor, outdoor, and possible underground transportation areas) biological incident.
The biological agent and its characteristics, the weapon design, amount of contaminant released, and a
plethora of environmental and meteorological factors are completely separate yet interconnected
processes that greatly influence the extent and level of contamination. Similarly, decisions related to the
sampling strategy (e.g., sample medium, sampling area, spacing, etc.) will affect the cost, time, amount of
waste generated, and personnel (i.e., resource demand) required to characterize and clear the
contaminated area. The process of understanding how these elements influence one another and
contribute to the overall problem is referred to as a systems approach. To what degree sampling and, more
specifically, variations in the sampling strategy interact and contribute to overall resource demand
following a wide area biological incident is still largely unknown. To date, there have been no attempts to
model characterization sampling following a wide-area biological incident.
The objectives of this study were fivefold: (1) to review available facts and information related to
historical wide-area biological contamination incidents; (2) to conduct a literature review to generally
assess the current state of knowledge regarding current sampling methods that may be used for
characterization sampling following a wide-area biological incident, specifically focusing on Bacillus
anthracis (Ba) contamination; (3) to summarize available data obtained from the literature for a
predetermined set of data elements (i.e., performance metrics) for current sampling methods and
technologies, (4) apply appropriate data from the literature and/or from recent operational field studies to
inform an analysis of the application of wide-area sampling strategies in the context of National Planning
Scenario (NPS) #2, and (5) evaluate the impacts of (a) variable sampling areas, hot spot areas,
1
-------
probabilities, and false negatives rates; and (b) variable resource availability; on the total estimated cost
and time to complete sampling and analysis activities in a hypothetical scenario.
Literature was aggregated using a set of predefined keywords relevant to this study (Appendix A). Each
piece of literature was read, assessed, and documented based on several criteria (Appendix B). Literature
deemed at least moderately relevant to meeting the objectives of this study, according to the above
criteria, were then summarized and relevant data were extracted and compiled in Microsoft Excel
spreadsheets.
In addition to available literature, reports from field studies, planning documents, and guidance
documents were also reviewed for data and information that were relevant for the objectives of this study.
The Bio-response Operational Testing and Evaluation (BOTE) Project was a recent collaborative effort
between the Department of Homeland Security, the U.S. Environmental Protection Agency (EPA), the
Department of Defense (DoD), and the Centers for Disease Control and Prevention (CDC). The BOTE
Project was designed to test and evaluate the operational response to a Ba spore release in a moderately
sized (-8,000 ft2) building. The test building was a two-story unoccupied office building. Two of the
primary objectives of the study were to "demonstrate that biological sampling and analysis methods
evaluated in previous studies provide accurate characterization of Ba simulant concentration challenges
for detection/identification purposes" and "[c]ollect and analyze the results from the decontamination
study and perform a cost analysis of all aspects of the remediation approaches." [1]
Over 3,000 samples were taken during the BOTE project using a variety of sampling techniques. Surface
sampling methods included cellulose sponge-stick wipes, macrofoam swabs, vacuum socks, and
Versalon® Wipes. Air sampling methods included SKC BioSamplers®, Model 3314 UV-APSTM
Spectrometer, Dycor XMX/2L-MIL Aerosol Collection System, and Mattson-Garvin Model 220 Slit-to-
Agar Sampler. The BOTE Project represents the most current application of biological agent sampling
techniques in a field application. While the sampling campaign focused solely on a single building and its
immediate surroundings, this sampling campaign represented the most comprehensive study to date that
has investigated the implications of sampling for a wide-area event and how the sampling plan and
number of samples impacts resource requirements, cost, and time to conduct sampling, analysis,
decontamination, and waste generation.
In addition to the BOTE Project report, the New York City Department of Health and Mental Hygiene's
Environmental Response and Remediation Plan for Biological Incidents [2] was reviewed. This report
delineates the operational and technical assistance required to lead a multi-agency effort to prepare for
and remediate contaminated areas following an intentional or unintentional biological incident involving
the release of Ba. The report includes: a decision-making framework, along with roles and
responsibilities; operational tools and guidance for government/regulatory entities and stakeholders;
technical guidance, including decontamination approaches and tactical procedures, sampling strategies
and plans, and clearance criteria and procedures; and guidance for establishing technical working groups
and environmental clearance committees. The plan is structured to help users employ the correct actions
for assessment, decontamination, and other actions needed for a successful cleanup without regard to
incident type and how many structures are involved. The plan does not focus on an incident of any
particular size, but instead provides guidance that is scalable and can be applied to small-scale incidents,
wide-area incidents, or regional incidents that could involve multiple city blocks, hundreds of facilities,
and multiple response agencies.
This assessment report is structured in the following manner: Chapter 2 provides an overview of four
historical anthrax incidents. Chapter 3 presents an overview of current sampling methods and techniques
and summarizes the results of our literature review. Chapter 4 applies a combination of data identified in
the literature (if relevant), relevant data from the results of the BOTE Project, relevant planning
2
-------
information from the New York City Department of Health and Mental Hygiene, and best estimates for a
hypothetical exercise involving a biological attack on large urban area. Chapter 5 extends the hypothetical
analysis presented in Chapter 4 to evaluate and analyze a range of possible variables and their impacts on
the total cost and time to complete sampling and analysis activities. Chapter 5 also introduces and
discusses elements of an EPA-developed spreadsheet which allows such a resource demand and impacts
analysis to be performed. Chapter 6 discusses the conclusions reached as a result of this study.
2 HISTORICAL ANTHRAX INCIDENTS
Several recent incidents have illustrated the need for enhancement and validation of sampling methods
and development of decontamination and remediation plans for a biological release. These incidents
include the "Amerithrax" incident in 2001, the Sverdlovsk, former Soviet Union anthrax release in 1979,
and an incident leading to the cleanup of Gruinard Island, United Kingdom in 1981. Naturally occurring
incidents, such as drummer-related incidents in the United Kingdom, Danbury CT, and Durham NH, were
excluded from this summary of historical incidents as the extent of contamination in these was
comparatively small.
2.1 Amerithrax
In October 2001—at a time when the U.S. had recently become acutely aware of its vulnerability to acts
of international terrorism—a series of bioterrorism incidents involving Ba spores (the causative agent of
anthrax) occurred. These incidents, referred to as the "Amerithrax" case, were the first known U.S. cases
of anthrax-related bioterrorism and resulted in 22 cases of infection across four states and Washington,
D.C., five of which were fatal [3, 4],
The Ba spores were sent through the mail in letters that were addressed to two U.S. senators and members
of the media [3], As the first cases of anthrax were diagnosed, the possibility of contamination led to a
major disruption in the operations at various buildings including the Hart Senate Office Building in
Washington, D.C., and U.S. postal facilities in the capital area and Trenton, New Jersey [5],
Approximately 42 buildings were contaminated to some degree. The mean building size (for 11 buildings
for which data were available) was 23 million square feet. However, this figure may not be representative,
as contamination was limited to just a few rooms for most buildings [6].
Funded through its Superfund program, EPA spent approximately $27 million to remediate (sampling,
decontamination, waste management) 26 buildings in the Capitol Hill area. Of this total cost, EPA spent
$25 million to hire 27 contractors and three federal and state agencies to assist in cleanup activities. The
other $2 million covered EPA's personnel costs, mostly for the staff who supervised the contractors.
Cleanup took three months before Capitol Hill office buildings were allowed to reopen [7].
Before the cleanup, sample collection and other analytical methods were used to detect Ba in postal
facilities—but no validated sampling and cleanup methods were available. Thus, while testing in 286
postal facilities produced largely negative results, the accuracy of the sampling effort was unknown [3],
Without validation, sampling may have been based on false assumptions. There were many unknowns
during the sampling process, such as how many samples to take and, in general, which sampling method
was appropriate in a given situation.
At the time, EPA had not developed a guidance document for sampling procedures, though the U.S.
Postal Service did have a sample collection plan. Site-specific decisions by EPA and the CDC were
largely discretionary and based on discussions with facility members, reviews of facility floor plans, and
other observations that helped inform a targeted sampling method. Unlike a probabilistic sampling design,
3
-------
targeted sampling design does not provide a level of statistical confidence in sample test results. Without
this statistical certainty, there could be many reasons for negative results, other than an actual absence of
Ba spores: (1) samples not being collected from places where Ba spores were present, (2) the detection
limit of the sampling and analysis method being greater than the actual contamination level, (3) not
enough samples being collected, (4) not enough spores being recovered from the sample material during
laboratory analysis, and (5) analysis of the sample extract not detecting Ba spores [3],
During the decontamination of congressional buildings, EPA collected both air and surface samples using
three types of surface sampling methods (wet swabs and wipes for nonporous surfaces and high-
efficiency particulate air [HEPA] vacuuming for porous1 materials), and four types of air sampling.
Nationwide, approximately 120,000 samples were taken at a cost of approximately 8.7 million dollars [6],
Numerous decontamination methods were used, including fumigating (vaporous hydrogen peroxide®,
chlorine dioxide gas, paraformaldehyde), spray-based decontamination with liquids (Spor-klenz®,
chlorine dioxide, disinfecting with Sandia foam®), and physical removal of surface-bound particulates by
methods such as HEPA vacuuming [7].
In addition to a lack of standard procedures and validated sampling methods, there were logistical
challenges. Insufficient laboratory analytical capacity limited the number of samples that could be
collected and the speed at which samples could be processed [3, 4], CDC's Laboratory Response Network
and military laboratories such as the U.S. Army Medical Research Institute of Infectious Diseases
(USAMRIID) processed over 125,000 clinical specimens and 1 million environmental samples [5],
Questions surrounding the Amerithrax case remain. It was later determined that the spores used in the
attacks were of the Ames strain, a unique strain used in U.S. military research. This realization led to the
plausible conclusion that the attacks may not have been an act of international terrorism, but perhaps
perpetrated by a well-trained and educated scientist with access to a U.S. military laboratory [4],
2.2 Sverdlovsk Anthrax Leak
In April 1979, an unknown number of anthrax-related deaths occurred in the former Soviet Union in
Sverdlovsk (now Yekaterinburg). The reticent nature of the Soviet Union made a more thorough
international investigation of the incident nearly impossible, leaving epidemiology as the only tool to find
the source of the anthrax outbreak. One of the few sources of first-hand evidence of an anthrax outbreak
comes from the handwritten notes of Dr. Faina Abramova, a doctor who treated some of the Sverdlovsk
patients. The most widely accepted reports of the incident state that 96 people were infected, and 64 of
these cases were fatal [4],
The Soviet authorities blamed the outbreak on the consumption of infected meat [4, 5], Meat was
collected and environmental samples were taken. However, the extent of these efforts is unknown. Local
fire brigades reportedly washed trees and building exteriors in sections of the city where most cases
resided. Stray dogs were shot, and some unpaved streets paved [8-10], However, based on doctors' notes,
victims suffered from symptoms that were not representative of gastrointestinal anthrax but rather what
appeared to be pulmonary anthrax, a deadlier form of the disease [4]. Anthrax cases appeared up to six
weeks after the incident ~ exceeding the 1-10-day incubation period due to either resuspension or
extended incubation time due to low dose exposure [8, 9], Further investigations suggest that the outbreak
was caused not by infected meat but by an accidental release of Ba spores by way of the exhaust
ventilation system from a Ministry of Defense microbiological facility in Sverdlovsk [5, 11],
Debate over the source of Ba spores in Sverdlovsk continued for many years after the incident [5], In
1992, a team of American scientists were allowed to review autopsy materials and other evidence from
1 Porous surfaces are the surfaces that are permeable to water or fluid used for sampling, and nonporous surfaces are
not permeable to water or fluid used for sampling.
4
-------
the incident. While the KGB had confiscated many medical records during the time of the incident, based
on the available evidence, the team of scientists was able to conclude that the outbreak resulted from the
escape of aerosolized Ba spores from the Sverdlovsk microbiological facility [5], According to
epidemiological data, the contamination may have extended over 50 km and was over 1 km wide in some
areas [9], Due to the covert response to this release, very little cost or personnel information is available.
2.3 Gruinard Island
Located off the coast of Scotland, Gruinard Island was a site of biological weapon testing for Great
Britain in 1942 before World War II. Tests of Ba bombs were carried out, and experiments on the island
successfully produced anthrax among targeted sheep. The island was eventually quarantined due to the
high levels of soil contamination by Ba spores, and Great Britain terminated testing with its offensive
program between 1955 and 1965 [5], The extent of contamination was assumed to be the entire 520-acre
island [12],
Sampling was carried out annually between 1948-1968 and 1972. A survey carried out in 1979 took 125
core soil samples. After a full-scale decontamination effort, clearance sampling consisted of 78 core soil
samples, 58 of which had previously indicated the highest levels before decontamination and 20 selected
at random. Only three of the 58 points showed high contamination and three of the sites randomly
sampled showed evidence of Ba. All suspect locations were subsequently decontaminated with 5%
formaldehyde in 1986 [13], Further sampling took place six months later with no positive results. Forty
sheep were allowed to graze for five months for further verification [10],
The British government eventually initiated a more comprehensive cleanup of Gruinard Island in 1986.
Scientists used formaldehyde that was circulated through a custom-built irrigation system that dispensed
over 280 tons of formaldehyde, diluted in 2,000 tons of seawater, for approximately one year. Soil
samples were tested at various distances away from the center of the Ba bomb explosions [10, 12], The
cost of decontamination was estimated at £500,000 or $800,000 or approximately 1.5 million dollars
when accounting for inflation. It is not known to what extent sampling contributed to this total [14],
2.4 Lessons Learned
In the three historical incidents previously described, contamination occurred in a vast range of structures
and sizes. The Amerithrax 2001 bioterrorism case involved large mail distribution centers and office
buildings. Contamination from the Sverdlovsk leak may have extended over an industrialized urban area
50 km long and over 1 km wide. Additionally, the entire 520-acre island of Gruinard was assumed to be
contaminated because of experimental weapons testing. Although information is limited on the
Sverdlovsk incident, in both the Amerithrax and Gruinard Island cases, sampling for Ba spores informed
decontamination efforts, i.e., where to focus decontamination activities and whether decontamination was
effective. In the case of the Amerithrax incident, the lack of validated sampling methods led to unknown
accuracy of results, which in turn, makes it difficult to fully understand whether the appropriate
decontamination stringency was performed (too stringent or not stringent enough). Available information
on the cost of decontamination ranged from $1.5 million dollars for Gruinard Island to $27 million dollars
to decontaminate 26 buildings in the Capitol Hill area associated with the 2001 Amerithrax incident. No
one remediation strategy is universally applicable, and each situation will have to be confronted in a
unique manner [15], Enhancement of sampling methods and strategies may help to improve sampling
accuracy, ease of use, representativeness, and could reduce cost and time required for remediation
activities.
5
-------
3 OVERVIEW OF SAMPLING METHODS
A small variety of established sampling methods have been used and are currently recommended for
many biological agents, including Ba. Some methods focus on small spaces: swab sampling, sponge
sampling, gauze sampling, and vacuum sampling (swab and wipe sample processing procedures have
been validated by CDC) [16], Others are more applicable to larger areas or other applications beyond the
initial response stage of an incident: air sampling, aggressive air sampling, and robotic floor sampling.
As stated above, one of the objectives of the literature review was to summarize available data from the
literature for a predetermined set of data elements for current and established sampling methods. Data for
air sampling and robotic floor sampling technologies were also gathered. The data elements of most
interest initially included all the following:
• Surface type sampled
• Recovery efficiency (RE, %)
• Cost
• Time needed to take sample
• Sample area size (sampling area per sample)
• Number of personnel required
• Necessary equipment
• Special deployment
• Safety concerns
• Waste produced
• Ease of laboratory analysis.
The initial literature search identified 212 sources as being potentially relevant for this study. These
sources included many peer-reviewed and non-peer-reviewed documents, including journal articles,
government reports, books, book chapters, theses, websites, magazine articles, and other studies (e.g.,
from the National Academies). Initial review and evaluation eliminated (or temporarily eliminated) 38
sources, leaving 174 that were further evaluated and rated against predetermined criteria (i.e.,
applicability, accuracy, clarity, and uncertainty and variability). Each of the 174 sources was scored
against each of those criteria and the overall score was calculated. Based on the overall scores, each
source was classified as "unacceptable," "low quality," "moderate quality," or "high quality." A total of
90 sources deemed at least "of moderate quality," according to the above criteria, were then summarized
and relevant data were extracted and summarized in Microsoft Excel spreadsheets. A few sources were
identified as being lower quality, but did contain some information relevant to this study; their data were
also extracted when appropriate.
The relevant sources identified above covered one or more of the sampling methods of interest, including
the currently established sampling methods. Other sources focused on one or more "wide-area" emerging
sampling methods (i.e., air sampling, aggressive air sampling, robotic floor sampling) or one of the in-situ
methods (i.e., point detection, standoff, passive standoff). Many of the relevant sources can be categorized
as either laboratory/bench-scale/controlled studies. Table 1, below, summarizes the number of sources
reviewed and the sampling methods/techniques they discuss or evaluate.
Table 1. Number of Sources That Contained Information and/or Data Relevant to Sampling
Method/Technology
Current Accepted Methods
Gauze/wipe sampling
32
6
-------
Sponge sampling
15
Swab sampling
38
Vacuum sampling
25
In-Situ Detection Methods3
Passive standoff
1
Point detection
2
Point detection and identification
6
Standoff
7
Other methods
17
"Wide-Area" Methods
Air sampling
16
Robotic floor sampling
2
a Currently, the sensitivity and accuracy of in-situ detection
methods precludes their use in response operations where
contaminants are unknown. These detection devices may,
however, have utility at detecting/confirming agent from a
visible powder.
Not all sources reviewed included data identified as relevant for this study. For all sampling methods and
technologies identified, Table 2 summarizes the number of reviewed sources that contained each of the
data elements listed above.
Table 2. Number of Sources That Contained Information and/or Data Relevant to This Study
Data Element
Number of Sources
Sample area size
54
Necessary equipment
42
Ease of laboratory analysis
42
Recovery efficiency, RE (%)
40
Surface type
32
Time needed to take sample
26
Cost
7
Special deployment
7
Number of personnel
6
Safety concerns
6
Waste produced
6
After further evaluating the volume of data extracted and the percentage of coverage for each of the
identified data elements in Table 2, this study was further defined to focus on the following data elements
for wipe, sponge, swab, and vacuum sampling:
• Surface type
• RE (%)
• Sample area size.
These three data elements were identified as being the most critical for the purposes of further analyzing
the impacts of sampling strategy for a hypothetical wide-area incident, including impacts on cost, time,
resource demand, etc. The analysis itself is discussed in Chapter 4. An overall summary of the data
7
-------
identified for surface type, RE, and sample area size for each sampling method appears below. General
observations are also noted and discussed.
Only those sources where at least one data point was available for: (1) surface type, (2) RE, and (3)
sample area size are included in the summary below. In the summary, data are presented as they were
found in the reference material. For example, for references that contained multiple RE ranges for a single
surface type: the summary below presents each individual range presented in the reference as opposed to
combining the ranges for the single surface type. Additionally, where a single range of results for multiple
surface types were found, the data are transcribed as such below.
Due to the large number of sources reviewed, the volume of data available, and the relative inconsistency
in the format of the data, it is not possible to comprehensively summarize all data for the three key data
elements. Variation in how the data were reported, the laboratory and/or test conditions, and the purposes
for which the data were obtained and/or the study objective(s) further complicate any effort to standardize
the results of the literature search. The following sections are intended to provide a general qualitative
overview of the current state of the science related to the various sampling methods and technologies
discussed.
For all the established sampling methods in the sections that follow, the literature reveals a diverse variety
of data on surface types, recovery efficiencies, and sample area sizes. Surface types sampled included
stainless steel, carpet, plastic, wood, glass, brick, wallboard, various other metals, concrete, and fabrics
including cotton, rayon, and polyester. In some cases, recovery efficiencies were reported with associated
errors or confidence limits and in a few cases only concentrations (measured in colony-forming units, or
CFU, for an area) were available. Sample areas also varied greatly depending on the source. A summary
of the data for swab sampling is presented in Table 3, wipe sampling in Tables 4 (sponge stick) and 5
(gauze), and vacuum sampling in Table 6. Sources listed in Tables 3 through 6 reported data for RE and
sample area size.
3.1 Swab Sampling
Swab sampling is appropriate for small, smooth, nonporous, hard-to-reach areas, and complex surfaces
less than 26 cm2 [17], Sampling conducted on rough or porous surfaces or over an extended area could
damage the swab. A sterile solution is applied to moisten the collection medium and the swab is wiped
over the sampling area in an overlapping vertical "S" pattern. After an initial pass, the swab is rotated and
another "S" pattern is made horizontally. Finally, the swab is rotated again and a diagonal "S" pattern is
performed over the sampling surface [16], Of the literature reviewed, 10 sources contained identified data
for surface type sampled, RE, and sample area size [18-27], Table 3 summarizes data from those sources
on surface type sampled (in alphabetic order), RE, sample area size, and number of samples (where
available) for swab sampling. Note that, in most cases, RE values in the reference material were presented
as a percentage; however, in a limited number of instances, RE was presented as a fractional value.
Fractional values are identified via footnotes in Table 3, as well as subsequent sampling summary tables.
Please refer to the list of acronyms and abbreviations at the beginning of this document for acronym
definitions.
Table 3. Summary of Available Data for Swab Sampling
Source ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area
Size
Nonporous Surfaces
[24]
Aluminum, V2A steel, multilayer
insulation (MLI) foil, Kapton,
Teflon
19.8-73.7
25 cm2
8
-------
Source ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area
Size
[20]
Chemical-agent-resistant coating
(CARC)-painted steel
Liquid-deposited spores—cotton: 47.0,
Dacron: 42.5, rayon: 43.6, macrofoam:
55.7.
Aerosol-deposited spores—cotton: 51.9,
Dacron: 57.6, rayon: 53.1, macrofoam:
51.5.
10 cm2
[24]
Glass
42.5-89.1
10 cm2
[24]
Glass
42.1-92.7
10 cm2
[24]
Glass
68.6-73.5
5 cm2
[20]
Glass
Liquid-deposited spores—cotton: 88.7,
Dacron: 82.1, rayon: 87.5, macrofoam:
89.1.
Aerosol-deposited spores—cotton: 62.4,
Dacron: 64.9, rayon: 65.2, macrofoam:
61.2.
10 cm2
[19]
Glass
52.2 ±8.3
317 cm2
[24]
Glass, stainless steel, polycarbonate,
vinyl/vinyl tile
51.5-75.5
10 cm2
[19]
Metal file cabinet
39.5 ±5.3
317 cm2
[24]
Monitor
0.5-2.7
25 cm2
[24]
Petri dish
35.9-66.1
25 cm2
[19]
Plastic seat
31.2 ±1.5
317 cm2
[20]
Polycarbonate
Liquid-deposited spores—cotton: 74.9,
Dacron: 83.4, rayon: 75.4, macrofoam:
88.3.
Aerosol-deposited spores—cotton: 65.1,
Dacron: 71.9, rayon: 68.9, macrofoam:
75.5
10 cm2
[24]
Directly inoculated stainless steel
99.9
6.25 cm2
[24]
Stainless steel
31.7—49.1
10 cm2
[24]
Stainless steel
0.1-43.6
25.8 cm2
[24]
Stainless steel
35.5-15.6
25 cm2
[24]
Stainless steel
15.8-83.1
26 cm2
[18]
Stainless steel
100-1,000 CFU/cm2: 0.395 (range 0.151-
0.659)3
10,000-100,000 CFU/cm2: 0.429 (range
0.146-0.747)a
Overall: 0.414 (range 0.146-0.747)3
25 cm2
9
-------
Source ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area
Size
[25]
Stainless steel
Nylon-flocked swab with phosphate -
buffered saline including 0.02% [vol/vol]
Tween 80 (PBST) extraction solution and
vortexing: 45.5 ± 1.2 (SE = 2.74)
Nylon-flocked swab with water extraction
solution and vortexing and sonication:
49.0 ±1.9 (SE = 4.18)
Cotton swab (standard National
Aeronautics and Space Administration
(NASA) protocol): 13.2 ± 1.2 (SE = 2.32)
Recovery of Ba from stainless steel:
Nylon-flocked-swab with PBST extraction
solution and vortexing: 19.7
Nylon-flocked-swab with water extraction
solution and vortexing and sonication:
19.6
Cotton swab (standard NASA protocol):
6.5
25 cm2
[21]
Steel
3 CFU/100 cm2: 3.4
30 CFU/100 cm2: 6.5
200 CFU/100 cm2: 5.0
103 cm2
[22]
Steel
31.7—49.1
10 cm2
[27]
Steel
Cotton Swab:
Dry, no extraction: 0.5
Pre-moistened, no extraction: 4.7
Dry, vortex: 8.0
Pre-moistened, vortex: 41.7
Dry, sonication: 6.9
Pre-moistened, sonication: 13.6
4 in2 (2x2 in)
[27]
Steel
Macrofoam Swab:
Dry, no extraction: 0.7
Pre-moistened, no extraction: 6.3
Dry, vortex: 11.9
Pre-moistened, vortex: 43.6
Dry, sonication: 12.7
Pre-moistened, sonication: 17.7
4 in2 (2><2 in)
[27]
Steel
Rayon Swab:
Dry, no extraction: 0.1
Pre-moistened, no extraction: 1.0
Dry, vortex: 4.4
Pre-moistened, vortex: 11.5
Dry, sonication: 4.5
Pre-moistened, sonication: 8.5
4 in2 (2x2 in)
[27]
Steel
Polyester Swab:
Dry, no extraction: 0.1
Pre-moistened, no extraction: 2.0
Dry, vortex: 2.1
Pre-moistened, vortex: 9.9
Dry, sonication: 1.4
Pre-moistened, sonication: 11.2
4 in2 (2x2 in)
[23]
Teflon
75.20
6,451.6 cm2
10
-------
Source ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area
Size
[20]
Vinyl tile
Liquid-deposited spores—cotton: 49.0,
Dacron: 62.2, rayon: 58.3, macrofoam:
72.0.
Aerosol-deposited spores—cotton: 60.3,
Dacron: 68.7, rayon: 60.2, macrofoam:
67.0
10 cm2
[19]
Vinyl tile
41.1 ±3.5
317 cm2
[18]
Painted wallboard
100-1,000 CFU/cm2: 0.355 (range 0.030-
0.887)3
10,000-100,000 CFU/cm2: 0.456 (range
0.072-0.808)a
Overall: 0.405 (range 0.030-0.887)a
25 cm2
[19]
Wood laminate
28.6 ±6.6
317 cm2
Nonporous and Porous Surfaces
[24]
Stainless steel, carpet
3.4-14.0
103 cm2
[24]
Plastic, wood, cotton
0.2-5.5
1 cm2
[24]
Plastic, oak wood, polyester
upholstery fabric, carpet
0.6-6.6
104.04 cm2
Porous Surfaces
[21]
Carpet
3 CFU/100 cm2: 12
30 CFU/100 cm2: 14
200 CFU/100 cm2: 12
103 cm2
[19]
Finished concrete
0.8 ±0.1
317 cm2
[19]
Nylon cushion
3.9 ±0.1
317 cm2
[26]
Vectran fabric
Foam-tipped swab: 1.4
Nylon-flocked swab: 5.9
Puritan swab 3600: 2.4
Puritan swab 3655: 5.3
Whatman FTA card applicator: 6.6
25 cm2
a RE reported as fractional values in this source.
G.F. Piepel et al. summarized 20 laboratory studies to evaluate collection, storing/transporting,
processing, and analyzing samples from surfaces contaminated by Ba, including swab, wipe, sponge, and
vacuum samples of porous and nonporous surfaces. The authors identified knowledge gaps and suggested
areas for future studies. Some of the studies used methods that are out-performed by other methods, were
limited in their scope, and addressed repeatability uncertainty but not reproducibility uncertainty. In their
summary, Piepel et al. suggested future studies should address false negative rates (FNRs) and quantify
reproducibility uncertainty, not just repeatability uncertainty [24],
A study by Estill et al. tested swab, wipe, and vacuum sampling techniques with low, medium, and high
surface loadings of Ba Sterne spores. Sampling techniques were tested on steel and carpet. In terms of
CFU/sampled area, wipes had the lowest limit of detection (LOD) on stainless steel, and swabs had the
lowest LOD on carpet. The authors noted that LOD estimates should be interpreted with caution due to
limitations of small sample sizes, less than ideal surface concentrations, and reaerosolization of spores
during sampling. Furthermore, the recovery efficiencies and LOD values from the study should not be
considered definitive as precision and confidence intervals should also be considered. Lastly, the authors
11
-------
note that the results do not consider the variance introduced by the use of multiple technicians to collect
samples and that precision improved with increasing surface concentration Estill, Baron [21],
Edmonds et al. studied whether differences existed in swab sampling recovery efficiencies for glass, steel
painted with chemical-agent-resistant coating (CARC), polycarbonate, and vinyl tile and demonstrated
that recovery of liquid-deposited spores differs significantly from recovery of dry aerosol-deposited
spores in most instances [20],
Brown et al. evaluated the CDC-recommended swab surface sampling method in terms of RE and LODs
from stainless steel and painted wallboard using rayon swabs. After determining a mean extraction
efficiency of 75.6% with a standard deviation of 11.8%, spore material specifically prepared for aerosol
suspension and inhalation was deposited on test coupons using a dry aerosol deposition system. Spores
were deposited using both low and high surface loadings. Spores were collected from the coupons using
rayon swabs wetted with sterile deionized water. Spores were also collected from reference coupons
seeded alongside the sample coupons by placing the reference coupons into containers of buffer solution.
Spores from the swabs and reference coupons were extracted via sonication. For stainless steel, the
average recovery efficiency was 41.4% with a standard deviation of 16.7%. For painted wall board, the
average recovery efficiency was 40.5% with a standard deviation of 23.2%. LOD was reported to be 25
CFU per sample area (25 cm2) for both stainless steel and painted wallboard. The authors noted no
statistical difference in recovery efficiencies between stainless steel and wallboard but indicated a high
difference in recovery efficiency variability between the two coupon types [18],
L.R. Hodges et al. evaluated the recovery of spores by pre-moistened macrofoam swabs and vortex
processing. Testing was conducted by five analysts from two laboratories. RE for all analyses ranged
from 38.0% to 49.1% [22],
Buttner et al. tested a sponge (the Sped-sponge; Nasco, Fort Atkinson, WI) and a macrofoam swab (the
SWKit - currently named the Sample Collection and Recovery Device; ASD BioSystems, Danville, VA)
for their sampling efficiencies. Overall efficiencies ranged from 0.7% to 52.2%, depending on the
sampling method and the surface material. During laboratory trials, the swab was more efficient than the
sponge by a factor of approximately 1.1 to 2.5 for five of the seven surface types. Efficiency for both
methods was highest on glass surfaces and lowest for finished concrete. Test chamber trials supported
laboratory results that the swab is generally more efficient than the sponge. Efficiency for both methods
was highest on the metal file cabinet, vinyl tile, and wood laminate surfaces. Computer monitor screens
and finished concrete had the lowest efficiencies [19],
In a study by Probst et al., ten different sampling methods were tested for their efficiency at recovering B.
atrophaeus spores from Vectran fabric, a material used for spacecraft airbags. The most efficient method
was a wipe-rinse technique with a foam-spatula protocol (13.2% efficiency). Compared with the standard
wipe sampling technique used the by European Space Agency (ESA), the foam-spatula protocol
performed better with an efficiency of 41.1 % while the standard method had an efficiency of 13.9%.
Multiple experiments were conducted with different spore concentrations and different types of Bacillus
species and different methods of inoculating surfaces (spotting versus aerosolization) resulting in
differences in recovery efficiencies. Sampling efficiency was found to depend largely on sampling
material and the surface type being sampled [26].
A study by Rose et al. tested cotton, macrofoam, polyester, and rayon swabs for their efficiency in
recovering Ba spores from nonporous 2/2-in steel coupons. Two swab preparation methods (pre-
moistened and dry) were also tested. The authors found that, when vortexed, pre-moistened macrofoam
and cotton swabs had the highest recovery efficiencies with mean efficiencies of 43.6% (standard
deviation [SD] 11.1%) and 41. 7% (SD 14.6%), respectively. Mean RE values for pre-moistened and
12
-------
vortexed polyester and rayon swabs were 9.9% (SD 3.8%) and 11.5% (SD 7.9%), respectively. The
authors noted that various factors can contribute to poor recovery efficiencies: sampling material, target
organism to be cultured, variations in sampling surface, and differences in personnel who are collecting
and processing samples. The authors add that wide variation in the results from this study demonstrates
the overall low efficiency and precision of swabs used for surface sampling but add that swab sampling
may sometimes be the best available method [27],
Probst et al. stated that the National Aeronautics and Space Administration's (NASA's) standard cotton
swab sampling protocol has not been improved upon in decades. To determine if there is a more efficient
swab type and sampling method, the authors tested two protocols for a new type of nylon-flocked swab.
These novel protocols were found to recover three to four times more B. atrophaeus spores than the
standard NASA method, with recovery efficiencies of 45.4% and 49.0% compared to a recovery
efficiency of 13.2% for the standard method. The recovery efficiencies of the protocols were also
relatively high for seven different Bacillus species, including Ba Sterne (a RE of 20%). Sampled surfaces
include 5 x 5-centimeter stainless steel, carbon-fiber-reinforced plastic, roughened carbon-reinforced
plastic, and Vectran fabric type A. The authors noted that variation in recovery efficiencies depended on
the roughness of the sampled surface. The highest recovery efficiencies were measured from sampling on
carbon-fiber-reinforced plastic and stainless steel. Certain Bacillus species are also sampled at a higher
efficiency, possibly due to different physiochemical adhesive properties of the spores (e.g., Ba has an
exosporium while B. atrophaeus does not). Other factors that may affect RE include choice of extraction
method. While cotton swabs may have recovery efficiencies as high as 41.7% based on recent studies,
they may have significant shortcomings. As organic material, their inherent DNA content may lead to
false positives. Improved sampling techniques will contribute to efforts in both spacecraft cleanliness
control and the detection of Ba contamination [25],
3.2 Sponge Wipe and Gauze Wipe Sampling (Wipe Sampling)
Wipe sampling is appropriate for small nonporous surfaces between 645 cm2 and 949 cm2 using sponge
wipes or gauze wipes [17], These CDC-developed and -recommended procedures for sponge and gauze
sampling and modified procedures are described by the EPA [28], Wet or dry samples can be collected;
wipes are generally used for smooth surfaces. Wipe sampling is appropriate for determining the extent
and location of contamination and for screening particular items. It is also used for determining
decontamination effectiveness [16],
When collecting wet samples, a sterile solution is applied to moisten the collection medium, and the wipe
is moved over the sampling area in an overlapping vertical "S" pattern. After an initial pass, the wipe is
rotated, and another "S" pattern is made horizontally. Next, the wipe is rotated again, and a diagonal "S"
pattern is performed over the sampling surface. Finally, as opposed to swab sampling, the surface is then
wiped around the perimeter.
Seven of the sources that were reviewed contained data for surface type sampled, RE, and sample area
size for sponge wipe sampling [19, 21, 24, 26, 29-33], Table 4 summarizes data on surface type sampled,
RE, and sample area size for sponge wipe sampling.
Table 4. Summary of Available Data for Sponge Wipe Sampling
Document
ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area
Size
Nonporous Surfaces
[29]
Ceramic tile
48.90
645.16 cm2
[24]
Glass
74.3
32.49 cm2
13
-------
Document
ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area
Size
[19]
Glass
47.3 ±1.0
930 cm2
[31]
Glass
48.40
100 cm2
[33]
Glass, clean,
high relative
humidity (RH)
0.333 (95% CI: 0.219-0.447; SE=0.055; SD=0.270)
100 cm2
[33]
Glass, dirty, high
RH
0.547 (95% CI:0.367-0.727; SE=0.087; SD=0.426)a
100 cm2
[33]
Marble, clean,
highRH
0.436 (95% CI: 0.354-0.518; SE=0.041; SD=0.318)a
100 cm2
[33]
Marble, clean,
lowRH
0.482 (95% CI: 0.393-0.570; SE=0.044; SD=0.342)a
100 cm2
[33]
Marble, dirty,
highRH
0.685 (95% CI: 0.514-0.855; SE=0.085; SD=0.656)a
100 cm2
[33]
Marble, dirty,
lowRH
0.595 (95% CI: 0.481-0.709; SE=0.057; SD=0.434)a
100 cm2
[19]
Metal file cabinet
44.8 ±0.6
930 cm2
[29]
Plastic panel
9.8
645.16 cm2
[24]
Stainless steel
18-31
929 cm2
[24]
Stainless steel
46.1-77.9
645.16 cm2
[24]
Stainless steel
24.4-32.4
645.16 cm2
[24]
Stainless steel
32.3
645.16 cm2
[24]
Stainless steel
26.8
645.16 cm2
[24]
Stainless steel
36.3
645.16 cm2
[24]
Stainless steel
26
645.16cm2
[29]
Stainless steel
48.1
645.16 cm2
[32]
Stainless steel
RE, % (SE): 32.4 (4.4), 24.4 (2.8), and 30.1 (2.3) for the 1-, 2-,
and 4-loglO inoculum levels, respectively
645.16 cm2
[24]
Stainless steel,
ceramic tile,
vinyl/vinyl tile
12.5-75.5
645.16 cm2
[33]
Stainless steel,
clean, high RH
0.18 (95% CI: 0.126-0.233; SE=0.026; SD=0.145)a
100 cm2
[33]
Stainless steel,
clean, low RH
0.797 (95% CI: 0.676-0.917; SE=0.059; SD=0.344)a
100 cm2
[33]
Stainless steel,
dirty, high RH
0.198 (95% CI: 0.146-0.249; SE=0.025; SD=0.133)a
100 cm2
[33]
Stainless steel,
dirty, low RH
0.968 (95% CI: 0.872-1.065; SE=0.047; SD=0.285)a
100 cm2
[21]
Steel
3 CFU/100 cm2: 31
30 CFU/100 cm2: 22
200 CFU/100 cm2: 18
929 cm2
[29]
Painted Wood
25.5
645.16 cm2
[19]
Plastic seat
18.1 ±4.2
930 cm2
[19]
Wood laminate
11.4 ±0.7
930 cm2
[29]
Vinyl tile
25.6
645.16 cm2
14
-------
Document
ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area
Size
[29]
Faux leather
30.3
645.16 cm2
[24]
Faux leather,
painted wood,
plastic light over
panel
4.0-76.9
645.16 cm2
Porous Surfaces
[24]
Carpet
21-120
929 cm2
[21]
Carpet
3 CFU/100 cm2: 120
30 CFU/100 cm2: 21
200 CFU/100 cm2: 23
929 cm2
[19]
Finished concrete
0.7 ±0.2
930 cm2
[19]
Nylon cushion
11.3 ± 1.8
930 cm2
[26]
Vectran fabric
Inoculation by spray diffuser: 3.0
Inoculation by spray gun: 18.7
Inoculation by spotting: 13.2
400 cm2
[26]
Vectran fabric
Ba Sterne: 0.3
B. atrophaeus: 13.2
B. megaterium 2c 1: 5.1
B. safensis: 0.5
B. thuringiensis E24: 5.4
400 cm2
[26]
Vectran fabric
Foam-spatula protocol on Vectran fabric: 13.2
Nylon-flocked spatula protocol on Vectran fabric: 4.2
SpongeSicle (cellulose tip material) protocol on Vectran fabric:
3.4
400 cm2
a RE reported as fractional values in this source.
Rose et al. reported a cellulose sponge wipe-processing protocol tested by nine laboratories. Steel
coupons were inoculated with Ba Sterne spores, then sampled with cellulose sponges. Mean percent
recovery varied by level of inoculation. The efficiencies of many sampling methods (gauze wipes, foam
spatula, and BiSKit; Quicksilver Analytics, Inc., Abingdon, MD) have been evaluated but not sufficiently
to provide a validated method. The authors claim that the sponge wipe-processing method was validated
by this controlled study [32],
A study by Krauter et al. evaluated the sponge-wipe method in a series of tests. Factors that were
evaluated were the effects of contaminant concentrations on RE, the effects of surface materials on RE,
FNR, LOD, and the uncertainties of all these values. The authors stated that a nearly linear dependence
was found between surface roughness (i.e., vertical deviation of surface measured in |_im) and RE, with
smoother surfaces having higher recovery efficiencies. FNRs corresponded with RE data, also
demonstrating that smoother surfaces resulted in higher recovery efficiencies and lower FNRs. The
authors also speculated that the reasons for differences between the results of this study and the results of
others include the use of different surrogate spores and the lack of a mixed microbial culture. The results
of the study by Krauter et al. show linear dependence of RE and FNR on surface roughness, with
smoother surfaces resulting in higher mean REs and lower FNRs. The authors contend that these findings
may have implications for field sampling, suggesting that field technicians might sample smooth surfaces
and take quick measurements of the roughness index to estimate RE, FNR, and LOD [29],
A study by Lewandowski et al. evaluated a polyester swab, a macrofoam sponge wipe, and a foam spatula
in terms of recovering B. atrophaeus and Pantoea agglomerans from 20 by 20 cm (400 cm2) glass and
15
-------
stainless steel surfaces. Aerosolized spores were used in the study because they are more representative of
deposition during previous incidents than spores that are deposited via spore suspension. The coupons
were divided into 10 by 10 cm (100 cm2) quadrants, so only one quadrant on one plate was swabbed to
avoid cross-contamination. Two quadrants were swabbed in experiments that tested vertical and
horizontal surfaces. Using a macrofoam sponge wipe was more efficient than a polyester swab at
recovering spores deposited by aerosol. Median RE for foam spatulas was 9.9% for B. atrophaeus spores
on glass. Mean RE was 48.4%. The RE of the foam spatula was dependent on the concentration of spores
aerosolized. Dilutions of samples were plated and incubated for 24 hours or for one to three days. The
authors calculated RE relative to the theoretical number of spores that are deposited on a surface. They
note the lack of a method to accurately measure the number of spores that settle on a surface in an
experiment, but the foam spatula appears to be a reliable indicator of the level of B. atrophaeus
aerosolized spore-contaminated surfaces [31],
Einfeld et al. discuss the evaluation and recovery of B. atrophaeus spores from grime-treated and clean
surfaces in a controlled chamber study and outdoor surfaces including stainless steel, glass, marble, and
concrete using wipe and vacuum sampling methods. Testing was conducted using both low- and high-
humidity conditions, and the results show that spore recovery from grime-coated surfaces is the same as
or better than spore recovery from clean surfaces. Statistically significant differences between method
performance for grime-coated and clean surfaces were observed in only approximately half of the
chamber tests conducted [33],
Of the literature reviewed, eight sources contained data for surface type sampled, RE, and sample area
size for gauze wipe sampling [23, 24, 26, 32, 34], Table 5 summarizes data on surface type sampled (in
alphabetic order), RE, and sample area size for gauze wipe sampling.
Table 5. Summary of Available Data for Gauze Wipe Sampling
Docum
ent ID
Surface Type
Recovery Efficiency
(RE, %)
Sample
Area Size
NonPorous Surfaces
[24]
Metal
11.3-18.5
10,000 cm2
[24]
Monitor
2.6-4.2
25 cm2
[34]
Painted wallboard
100-1,000 CFU/cm2: 0.325 (range 0.081-0.574)
10,000-100,000 CFU/cm2: 0.252 (range 0.081-
0.566)
Overall: 0.285 (range 0.081-0.574)
25 cm2
[24]
Directly inoculated stainless steel
99.90
25 cm2
[34]a
Stainless steel
100-1,000 CFU/cm2: 0.312 (range 0.153-0.509)
10,000-100,000 CFU/cm2: 0.392 (range 0.167-
0.674)
Overall: 0.346 (range 0.153-0.674)
25 cm2
[24]
Stainless steel
30.80
646.16 cm2
[24]
Stainless steel, glass, marble
18.0-96.8
100 cm2
[24]
Stainless steel, painted wallboard
25.2-39.2
25 cm2
[23]
Teflon
90.40
1.3 m2
Nonporous and Porous Surfaces
[24]
Painted wallboard, stainless steel,
vinyl/vinyl tile, wood
39.50
900 cm2
16
-------
Docum
ent ID
Surface Type
Recovery Efficiency
(RE, %)
Sample
Area Size
[24]
Plastic, oak wood, polyester
upholstery fabric, carpet
1.4-7.9
104.04 cm2
[24]
Plastic, wood, cotton, cloth
0.2-6.6
1 cm2, not
recorded
Porous Surfaces
[24]
Heating, ventilation, and air
condition (HVAC) filter
14.5-19.9
100 cm2
[26]
Vectran fabric
Polyester Spec wipe 7 wipers, 115-0043 (ESA
standard) from Vectran fabric: 0.5
Polyester Vectra Alphasorb TX 1050 from Vectran
fabric: 0.5
400 cm2
a The data presented in [34] were referenced by
32], which is also included in the bibliography presented and the
end of this document.
Studies containing relevant data on gauze wipe sampling and not previously discussed in the above
sections for another sampling method are summarized below.
Brown et al. evaluated extraction efficiencies, recovery efficiencies, and LODs for nonporous surface
sampling (stainless steel and painted wallboard) using polyester-rayon blend wipes and the current CDC
wipe collection method. The surrogate used in this study was B. atrophaeus. Spore material specifically
prepared for aerosol suspension and inhalation was deposited on test coupons using a dry aerosol
deposition system. Spores were deposited using both low and high surface loadings. Spores were
collected from the coupons using sterile polyester-rayon blend gauze wipes moistened with sterile
deionized water. Spores were also collected from reference coupons, seeded alongside the sample
coupons, by placing the reference coupons into containers of buffer solution. Spores were extracted from
the wipes and reference coupons via sonication. The average recovery efficiency for the stainless steel
coupons was reported to be 34.6% with a standard deviation of 12.2%. The LOD for stainless steel was
reported to be 90 CFU per unit of sample area. The average recovery efficiency for the painted wallboard
coupons was reported to be 28.5% with a standard deviation of 15.2%. The LOD for painted wallboard
was reported to be 105 CFU per unit of sample area. While the surface area evaluated in this study was 25
cm2, the investigators assumed "that the number of CFU required for detection is independent of the
sample surface area and primarily a function of recovery efficiency." The results of the study indicated
statistically significant lower recovery efficiencies of spores using the wipe method for painted wallboard
versus stainless steel, though both are considered nonporous surfaces. The authors indicated the
difference in recovery efficiencies is likely due to differences in surface texture and physiochemical
adhesive properties between the surface types. The authors also presented several causes for the large
variability in the results for each coupon type including errors in sampling mechanism, sample collection,
and sample processing. The authors also indicated that variability in surface deposition and the
incomplete removal of spores from reference coupons could also contribute to overall variability in the
results. [34],
A study by Sanderson et al. describes the relative effectiveness of sampling using dry swabs, wet swabs,
vacuum socks, and wipes. The results are presented as median and range of CFU/cm2, but relative
recovery efficiencies could not be determined because the initial concentrations of spores on the sampled
surfaces were not known. This study evaluated the results of wet and dry swab (sterile rayon—
noncotton), wipe (sterile rayon gauze pad), and vacuum sock surface sampling of nonporous surfaces at a
U.S. postal facility contaminated with Ba. The study compared detection percentage and spore
concentration for the collection methods and materials used, but not collection efficiencies (since the
17
-------
initial concentrations of the spores on the various surfaces were unknown). The results of the study
suggest that dry swabs should not be used for Ba sampling: wipes are preferred for nonporous surfaces
with light dust, and vacuum socks for heavily dusted surfaces [35],
Kirschner and Puleo tested polyester-bonded cloth (PBC) for its efficiency in sampling large areas. The
authors determined it was most effective on surfaces less than 0.74 m2 (8 ft2) in area. PBC was found to
have an RE of 90.4% versus 75.2% for cotton. The focus of this study is on sampling spacecraft surfaces.
Advantages of the PBC over cotton swabs include the higher release of spores by the PBCs (i.e., cotton
swabs tend to retain organisms), and a higher surface area sampled per unit. The bioassay of 6,451.6 cm2
(1,000 in2) requires 250 cotton swabs for full coverage. Alternatively, only one PBC would be needed for
the same area and coverage. The maximum effective sampling area ranges between 0.74 and 1.49 m2, but
for standardization, this area should be reduced to 0.74 m2. In terms of time needed to take samples, an
experienced team can collect approximately one cotton swab per two-minute period. The same team
should be able to sample an unobstructed 0.74 m2 area in the same amount of time. While this wipe-rinse
technique was designed for spacecraft hardware surfaces, its efficiency may prove useful in other settings
[23],
3.3 Vacuum Sampling
As opposed to swab, wipe, and gauze sampling, vacuum collection can be used on porous surfaces as well
as nonporous surfaces and where the surface may not be entirely smooth or flat. Furthermore, vacuums
can be used over wider areas to collect bulk samples. Vacuum sampling uses a HEPA filter on the exhaust
and a sample collection sock or cassette attached to the inlet of the vacuum collection hose. A specialized
sampling vacuum, such as an Atrix Omega (Atrix International, Inc., Burnsville, MN) is used to collect
samples [16],
Of the literature reviewed, five sources contained identified data for surface types sampled, RE, and
sample area size for vacuum sampling [21, 24, 33, 36, 37], Table 6 summarizes data on surface type
sampled (in alphabetic order), RE, sample area size, and number of samples (where available) for vacuum
sampling.
Table 6. Summary of Available Data for Vacuum Sampling
Document ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area Size
Nonporous Surfaces
[33]
Marble, clean, high RH
0.121 (95% CI:0.067-0.171;
SE=0.012; SD=0.069)a
100 cm2
[33]
Marble, clean, medium RH
0.119 (95% CI: 0.067-0.171;
SE=0.056; SD=0.154)a
100 cm2
[33]
Marble, dirty, high RH
0.170 (95% CI: 0.142-0.199;
SE=0.014; SD=0.085)a
100 cm2
[33]
Marble, dirty, medium RH
0.100 (95% CI: 0.056-0.144;
SE=0.021; SD=0.123)a
100 cm2
[36]
Painted wallboard
100-1000 CFU/cm2: 0.245
(range 0.041-0.577)a
10,000-100,000 CFU/cm2:
0.250 (range 0.204-0.295)a
100-100,000 CFU/cm2: 0.248
(range 0.035-0.577)a
100 cm2
[24]
Stainless steel
3.7-5.5
929 cm2
18
-------
Document ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area Size
[36]
Stainless steel
100-1,000 CFU/cm2: 0.321
(range 0.062-0.55 l)a
10,000-100,000 CFU/cm2:
0.231 (range 0.091-0.414)a
100-100,000 CFU/cm2: 0.289
(range 0.062-0.551)a
100 cm2
[21]
Steel
3 CFU/100 cm2: 5.5
30 CFU/100 cm2: 4.7
200 CFU/100 cm2: 3.70
929 cm2
Nonporous and Porous Surfaces
[24]
Ceiling tile, painted wallboard,
stainless steel, vinyl/vinyl tile,
wood
4.4
900 cm2
[24]
Marble, concrete
10.0-19.7
100 cm2
[24]
Stainless steel, painted
wallboard, carpet, concrete
16.4-36.1
100 cm2
Porous Surfaces
[37]
Carpet
Vacuum sock (slow): 25.1
Vacuum sock (fast): 15.2
2787 cm2
[24]
Carpet
3.7-6.3
929 cm2
[24]
Carpet
0.20-0.48
400 cm2
[24]
Carpet
2.26-1.69
400 cm2
[21]
Carpet
3 CFU/100 cm2: 6.3
30 CFU/100 cm2: 3.7
200 CFU/100 cm2: 4.7
929 cm2
[36]
Carpet
100-1,000 CFU/cm2: 0.361
(range 0.057-0.650)a
10,000-100,000 CFU/cm2:
0.229 (range 0.136 0.366)a
100 100,000 CFU/cm2: 0.282
(range 0.057-0.650)a
100 cm2
[37]
Concrete
Vacuum sock (slow): 10.2
Vacuum sock (fast): 11.7
37 mm mixed cellulose ester
(MCE) filter cassette: 48.7
2787 cm2
[36]
Concrete
100-1000 CFU/cm2: 0.216
(range 0.011-0.627)a
10,000-100,000 CFU/cm2:
0.164 (range 0.059-0.3 ll)a
100-100,000 CFU/cm2: 0.189
(range 0.011-0.627)a
100 cm2
[33]
Concrete, clean, high RH
0.165 (95% CI: 0.144-0.187;
SE=0.011; SD=0.082)a
81 cm2 circular coupon
[33]
Concrete, dirty, high RH
0.197 (95% CI: 0.170-0.225;
SE=0.014; SD=0.104)a
81 cm2 circular coupon
[24]
HVAC filter
2.42
309.68 cm2
[24]
HVAC filter
1.16-3.52
400 cm2
19
-------
Document ID
Surface Type
Recovery Efficiency (RE, %)
Sample Area Size
[37]
Upholstery
Vacuum sock (slow): 4.2
Vacuum sock (fast): 9.1
3M™ Forensic filter: 1.4
2787 cm2
aRE values reported as fractional values in this source.
Calfee et al. performed a comparative evaluation of vacuum-based methods using vacuum socks, 37 mm
mixed cellulose ester (MCE) filter cassettes, 37 mm polytetrafluoroethylene (PTFE) filter cassettes, and
forensic evidence filters to collect samples from carpet, concrete, and upholstery [37], For the vacuum
sock filters, both a slow and fast pace were used to collect samples. The surrogate used in the study was
B. atrophaeus. Spore material was deposited on the sampling coupons using a metered dose inhaler.
Spore material was deposited simultaneously on stainless steel reference coupons which were
subsequently sampled with pre-moistened wipes. The wipe recovery results from the stainless steel
coupons were used to normalize the recoveries of the vacuum based sampling methods. The 37 mm MCE
method resulted in significantly higher relative recoveries than the other studied methods when sampling
concrete and upholstery. For the concrete coupons, the 37 mm MCE filter had a relative recovery of 1.242
as compared to the next highest value, 0.33 for the forensic filter. For the upholstery coupons, the 37 mm
MCE filter had a relative recovery of 0.350 as compared to the next highest value, 0.232 for the vacuum
sock used at a fast pace. On carpet, the vacuum sock used at a slow pace had a significantly higher
relative recovery of 0.641 as compared to all other studied methods, except the 37 mm MCE filter with a
relative recovery of 0.474. The authors speculated that the MCE filter demonstrated higher recovery
efficiencies because it likely has fewer spores tightly bound to the collection media as compared to the
vacuum sock and forensic filter. The authors provided the time required for sampling and analysis for
each method and concluded that sample collection speed and ease of laboratory analysis should be
considered, in addition to recovery efficiency, in selecting a sampling method [37],
Another study by Brown et al. evaluated vacuum filter sock sampling in terms of recovery efficiency and
LODs from nonporous surfaces (stainless steel and painted wallboard) and porous surfaces (carpet and
bare concrete) [36], The surrogate used in this study was B. atrophaeus. Spore material specifically
prepared for aerosol suspension and inhalation was deposited on test coupons using a dry aerosol
deposition system. Spores were deposited using both low and high surface loadings. Spores were
collected from the coupons using 100% polyethylene vacuum filter socks. Spores were also collected
from reference coupons, seeded alongside the sample coupons, by placing the reference coupons into
containers of buffer solution. Spores were extracted from the vacuum filter socks and reference coupons
via sonication. The average recovery efficiency for the stainless steel coupons was reported to be 28.9%
with a standard deviation of 13.8%. The average recovery efficiency for painted wallboard was reported
to be 24.8% with a standard deviation of 14.5%, the average recovery efficiency for carpet and bare
concrete was reported to be 28.2% and 18.9%, with standard deviations of 13.4% and 14.1%,
respectively. The LOD for stainless steel and carpet was reported to be 105 CFU per unit of sample area.
The LOD for painted wallboard was reported at 120 CFU per unit of sample; bare concrete was 160 CFU
per unit of sample area. The authors noted a high variability in recovery efficiency between all the surface
types and discussed sources of potential error. The authors indicated that no statistically significant
difference was detected when comparing recovery efficiencies between porous and nonporous surfaces,
indicating this was a departure for normal trends for other sampling methods such as swabs or wipes [36],
3.4 Air Sampling
Air sampling methods are not always available for the initial first response or characterization of
biological contamination but may be appropriate for helping determine the existence of contamination in
the air or rapidly identify contaminated indoor areas. To perform air sampling, a sampling pump is used
20
-------
to draw air through dry filters, impingers, or impactors [16], Dry filter sampling is generally used for
biological agent detection and filter types include gelatin, MCE inert, and Teflon filters.
Data from three sources were identified as being relevant, primarily for RE and sample rate. RE for air
sampling is simply the RE from the media after collection and does not include capture/collection
efficiency. These data are summarized in Table 7.
In one study, water-soluble tape was used in place of swabs to measure reaerosolization of spores from
polyvinyl chloride (PVC) floor and carpet in an environmental chamber. The study also showed water-
soluble tape is a viable replacement for swabs [38]. Another study measured filter materials and
extraction methods that are best suited for environmental sampling of Ba using B. subtilis var. niger
endospores (also known as B. globigii) as a surrogate. MCE and PTFE filters combined with vortexing
and shaker extraction yielded the best performance for the filter collection and extraction of B. globigii
spores [39],
Table 7. Summary of Available Data for Air Sampling
Document ID
Recovery Efficiency (RE, %)
Sample Rate
[38]
99999a
50 liters/min of air flow rate
[39]b
Physical collection efficiency by filter
type: MCE (3 (im): 97.6;
polytetrafluoroethylene (1 |im): 94.2;
polytetrafluoroethylene (3 |im): 63.6;
gelatin (3 |im): 97.9; gelatin (3 |im):
94.3; polycarbonate (3 (im): 61.4
4 liters/min of air flow rate for inhalable aerosol
samplers
a The source presents the results of a study to determine reaerosolization of spores from flooring surfaces. However,
the authors indicate the sampling efficiency of the apparatus used in the study has previously been shown to be
99.999%for B. atrophaeus spores. The original source is cited as Parks SR, Bennett AM, Speight SE, Benbough JE.
1996. An assessment of the Sartorius MD8 microbiological air sampler. J. Appl. Bacterid. 80: 529-
534. http://dx.doi.Org/l1 1365-2672..1.996.tb03252.x. last accessed March 2017.
b The data presented in Document ID 38 were referenced by Document ID 2, which is also included in the
bibliography presented at the end of this document.
3.5 Aggressive Air Sampling
Aggressive air sampling has been used to confirm a negative finding of lead dust and asbestos
contamination, either as part of a public health investigation, during clearance after decontamination for
an area known to be contaminated, or at Superfund sites with where air is a pathway of exposures. While
air sampling has been used for asbestos and Ba, there is limited information on sensitivity and impacts
from environmental conditions [40], Aggressive air sampling involves (1) vigorous agitation of surfaces
in a space (using leaf blowers, for example) to aerosolize particles and (2) use of high-volume air
samplers to acquire and concentrate aerosolized materials for analysis [16],
3.6 Robotic Floor Sampling
One innovative sampling technique is the use of robotic floor cleaners (RFCs). These devices may be
used as a force multiplier and an alternative to exposing personnel to contaminated areas. A recent EPA
study evaluated three vacuum-based cleaning robots, one wipe-based robot, and one wet vacuum-based
robot for their sampling efficiency on nonporous surfaces (laminate and tile). EPA also evaluated the
vacuum-based RFCs on carpet to assess efficiency on a nonporous surface. The results showed that
sampling via RFC is a viable option—in some cases, the RFCs were as effective as currently
recommended methodologies. The sampling efficacy of a wet-mop RFC was comparable to the currently
used wet-wipe method on nonporous surfaces with low spore loadings. Vacuum RFCs performed
21
-------
comparably to vacuum sock sampling. Overall, these technologies show promise for future sampling. The
investigators compared RFC test results to currently used surface sampling methods (vacuum sock for
carpet and sponge wipe for laminate). The results showed that the average sampling efficacies for three
vacuum based RFCs on carpet were 26%, 162%, and 92% of vacuum sock sampling efficacy,
respectively. On laminate, five RFCs including vacuum, wet mop, and wet vacuum, average sampling
efficacies were 8%, 11%, 2%, 62%, and 32% of sponge wipe sampling efficacy, respectively. The authors
concluded that some robotic cleaners were as efficient as the currently used surface sampling methods for
B. atrophaeus spores on these surfaces [41],
A subsequent study evaluated two RFCs to explore their effectiveness at collecting samples widely
dispersed in a small area (approximately 4.5 square meters) and located in a small hot spot covering 2%
of the overall area within that same small area. A vacuum-based robot was used to assess effectiveness on
a carpet surface while a wipe-based robot was used on a laminate floor surface. The effectiveness was
compared to the effectiveness of the currently-used sampling methods, including vacuum sock and
sponge wipe. This study helped demonstrate that cleaning robots have various benefits for wide-area
sampling over the currently used sampling methods: fewer samples (i.e., one per deployment), composite
samples (fewer samples requiring analysis), higher chance to detect hotspots, and less risk of personnel
exposure [42], Data from these two studies are summarized in Table 8 below. The first study [41]
reported recovery efficiencies relative to sponge wipe and vacuum sampling recovery efficiencies and the
second absolute spore concentrations (CFU/area).
Table 8. Summary of Available Data for Robotic Floor Sampling
Source ID
Surface Type
Spore Loading Concentration
Recovery
Sample
Area Size
Nonporous Surfaces
[41]
Laminate
~lxl06 on 30.5 x 30.5 cm
Comparative Recovery3
Robot Rl: 8.1
Robot R2: 10.9
Robot R3: 2.4
Robot R4: 61.7
Robot R5: 31.9
71.5 cm2
[42]
Floor laminate
Hotspot scenario: 104 CFU/cm2 (but
only loaded onto approximately 2% of
the entire area of the room)
Widely dispersed scenario: 101
CFU/cm2
Hotspot scenario: 1.2><107 and
2.1xl06 CFU
Widely dispersed scenario:
2.8xl03and4.9xl02CFU
Entire test
surface: 4.4
m2;
inoculated
area: 1.8
m2
22
-------
Source ID
Surface Type
Spore Loading Concentration
Recovery
Sample
Area Size
Porous Surfaces
[41]
Carpet
~lxl06 on 30.5 x 30.5 cm
Comparative Recovery13
Robot Rl: 25.8
Robot R2: 161.5
Robot R3: 91.9
71.5 cm2
[42]
Carpet
Hotspot scenario: 104 CFU/cm2 (but
only loaded onto about 2% of the
entire area of the room)
Widely dispersed scenario: 10
CFU/cm2
Hotspot scenario: 1.4><105and
5.8x10s CFU
Widely dispersed scenario:
5.0xl03and3.2xl02CFU
Entire test
surface: 4.4
m2;
inoculated
area: 1.8
m2
a Source data presented are recovery efficiencies as compared to a sponge wipe sampling method on laminate
surface.
b Source data presented are recovery efficiencies as compared to a vacuum sock sampling method on carpet
surface.
3.7 Summary of Sampling Methods and Data
The sampling methods reviewed and the aggregated data detailed above highlight the vast range of
reported RE values for differing surface types, spore concentrations, and (in some cases) deposition
methods. Additionally, the data were collected under largely controlled conditions and there is the
possibility of potential biasing of the RE values due to the individual collecting the samples or conducting
the sampling. Furthermore, the identified sampling methods have been characterized on the selected
surfaces, mainly focused on indoor surfaces. A wide-area biological incident may require sampling of
indoor as well as outdoor surfaces and possibly underground transit systems. It is expected that the
sampling in a wide-area incident may require sampling very different surfaces compared to (or in addition
to) the ones evaluated in the studies above. Only a small number of surface sampling methods are well-
documented with their sampling protocols. It is uncertain, if a wide-area biological incident happens, how
the impacted areas would be characterized and how much resources would be required with the currently-
available sampling methods.
Some of the major gaps in our knowledge of sampling methods and capabilities to conduct wide-area
sampling currently include;
• Prediction of sampling performance (RE) in real-world settings based on laboratory studies
• Lack of method performance data (RE) for many/most outdoor surface types
• Prediction of analytical method accuracy and sensitivity in real-world settings, particularly
outdoor samples
• Detailed field-use protocols for all sampling methods are not available
• Inconsistent laboratory test parameters and conditions, resulting in inability to adequately
compare sampling method performance across studies
4 WIDE-AREA INCIDENT HYPOTHETICAL SCENARIO
4.1 Introduction
For this scenario, a statistically defendable sampling plan was developed, based on a hypothetical city.
Realistic spatial, infrastructure, and surface media information was gathered to determine the extent of
sampling and to match surface material types at each sampling location to the appropriate sampling
23
-------
method. The most appropriate sampling method for each surface material was chosen based upon the
findings of the sampling method review (Chapter 3). The resulting data, along with data gathered during
the literature survey portion of this study (Chapter 2) were loaded into a custom spreadsheet that
estimated the total resource demand required for characterization. Sections below describe the
methodology, inputs, and results of this study.
4.2 Methodology
4.2.1 Hypothetical Scenario
For this study, NPS #2 was used to simulate a biological attack by way of aerosolized Ba spores.
According to NPS #2, a truck, concealing an improvised spraying device, covertly aerosolizes
approximately 100 liters of a wet-fill Ba spore slurry (109 CFU/mL) into a densely packed urban area.
Approximately 1%, or 1012 CFU, of Ba spores are dispersed across two square miles.
To bind the event to some spatial context, a plume from a prior wide-area biological exercise was used
[43], The plume served as a reference point for surface and infrastructure characteristics for a city of
comparable size and for plotting sampling points. The area of interest is further described in the following
sections.
4.2.2 Assumptions
4.2.2.1 Environment
This study used a hypothetical city: a very large, densely populated urban area with a number of high-rise
buildings and a highly developed underground transportation system. A tangible city was used to derive
the spatial, infrastructure, and surface media information, using a combination of EPA and geographic
information system (GIS) tools and estimates. This study also used a plume derived from an earlier
biological exercise to define the area of contamination. The outputs from this model were converted to a
shapefile (i.e., polygon) for use in GIS software. Using the GIS software, a half-mile buffer was applied
to the plume to represent an area of uncertainty. Using the oval-shaped polygon, surface area and
characteristics were extrapolated from intersecting outdoor, indoor, and underground geospatial data. The
total estimated surface area for the outdoor, indoor, and underground environments is approximately 2,
15, and 0.5 square miles, respectively. Figure 1 shows ato-scale comparison of total ground/floor surface
area across the three environments. Note that the indoor surface area is approximately nine times greater
than the outdoor surface area, which would likely be the case for most densely populated urban areas. The
following sections describe the spatial and temporal boundaries for the outdoor, indoor, and underground
areas.
24
-------
Underground Surface Area
Outdoor Surface Area 0
L
Indoor Surface Area
Figure 1. Comparison of Surface Areas
4.2.2.1.1 Outdoor
Table 9 shows the inputs used to represent the surface area between buildings and the total outdoor
surface area. These values were obtained using EPA's Waste Estimation Support Tool (WEST) [44],
Table 9. Outdoor Surface Area
Surface Area
Total (m2)
Building footprints (estimated
roof area)
2.93 x 107
Surface area between buildings
2.82 x 107
Total outdoor surface area
(excluding vertical surfaces)
5.75 x 107
4.2.2.1.2 Indoor
Table 10 shows the inputs used to represent the building footprint (i.e., first floor surface area) and the
total indoor floor surface area. These values were generated using EPA's Incident Waste Assessment &
Tonnage Estimator (I-WASTE) [45] and Waste Estimation Support Tool (WEST) [44],
Table 10. Indoor Surface Area
Surface Area
Total (m2)
Building footprint
2.93 x 107
Total indoor floor
surface area
4.16 x 10s
2 Miles
0 5
I
25
-------
4.2.2.1.3 Underground Transit System
Table 11 shows the inputs used to represent the surface area for the underground transit system. These
values represent a best guess estimate based on publicly available datasets for a city of similar size.2
Table 11. Underground Surface Distribution
Surface Area
Total (m2)
Track area
1.03 x 105
Station area
7.53 x 104
Total area
1.78 x 105
4.2.2.2 Sampling Meth ods According to Surface Type
The characteristics of the surface being sampled determine the sampling method. Collection efficiencies
can vary greatly across material types. To better model the resource requirements associated with each
sampling medium, the distribution of surface types for all three environments were estimated using
WEST, I-WASTE, and empirical observations for outdoor, indoor, and underground surfaces,
respectively. Each surface type was assigned an appropriate sampling medium. The number of samples
attributed to each sample medium was calculated based on the proportion of the surface being sampled.
Sampling methods for each surface type were assigned based on guidance in EPA's Comprehensive
Biological Tactical Guidebook [16],
For this study, bulk sampling was assigned to surfaces such as trees and grass. Nonporous surfaces were
assigned a sponge wipe sampling technique because the sponge wipe method has been validated by CDC
and is applicable to the surface types identified in I-WASTE and WEST. Porous surfaces such as carpet
were assigned the vacuum sampling technique [16], It is acknowledged that air techniques, such as high-
volume air samplers, may have utility in large outdoor release operations. However, air sampling methods
have yet to be well characterized for outdoor sampling of Ba. Accordingly, was not considered in the
current analysis. In addition to sampling, laboratory throughput was considered and estimated according
to sample media type. The following sections describe the processes used to determine surface medium
and sample type assignment.
4.2.2.2.1 Outdoor Surface Characteristics
The distribution of outdoor surfaces was determined using WEST. WEST uses a built-in image
classification algorithm that identifies surface types for the purposes of estimating waste streams
following radiological incidents. This algorithm outputs an estimated distribution for outdoor surface
media using aerial imagery [44]. Aerial imagery from a city analogous to the hypothetical city described
in section 4.2.2.1 (i.e., a large, densely populated metropolitan area) was used to derive outdoor surface
media that are commonly found in cities of similar size. The results are shown in Table 12. Although
nonporous surfaces requiring the use of the sponge wipe sample type may also constitute a significant
portion of the outdoor surface types, WEST cannot identify these surface types. As a result, only soil and
vacuum sample types were assigned to outdoor surfaces.
2 See https://www.gc.ciHXY.edii/Page-Elements/Academics-Research-Centers-Initiatives/Centers-atKl-
Institutes/Center-for-Urban-Research/CUNY-Mapping-Service/Proiects/NYC-Subwav-and-Bus-data-in-GIS-format.
(last accessed March 20.1.7). The shapefiles were loaded into GIS and were used to estimate the dimensions and
surface area of the length of subway tunnel that intersected with the plume.
26
-------
Table 12. Outdoor Sample Medium to Surface Type Assignment
Surface Type
Surface Distribution
Sample Type
Trees
5%
Soil sample (at base of tree)
Grass
1%
Soil sample
Buildings
51%
Vacuum
Asphalt
22%
Vacuum
Concrete
21%
Vacuum
4.2.2.2.2 Indoor Surface Characteristics
For indoor surfaces, I-WASTE was used to estimate the distribution of indoor floor material based on the
number of buildings, square footage, and occupancy distribution. Due to the sheer number of indoor
materials that I-WASTE outputs, surface media were broken down according to porous and nonporous
materials. The results are shown in Table 13.
Table 13. Indoor Sample Medium to Surface Type Assignment
Surface Type
Surface Distribution
Sample Type
Porous (i.e., carpet)
90%
Vacuum
Nonporous (i.e., marble, ceramic tiles)
10%
Sponge Wipe
4.2.2.2.3 Underground Transit System
Methods for estimating underground surfaces are less explored than the other surfaces. Therefore, surface
estimates for underground transit systems were based on empirical observations. Concrete is assumed to
account for a large portion of the surface media, with the trains contributing a relatively small amount of
metal and carpet surfaces. The distribution of surfaces expected to be present in underground transit
systems are shown in Table 14.
Table 14. Underground Transit System Sample Medium to Surface Type Assignment
Surface Type
Surface Distribution
Sample Type
Concrete
90%
Vacuum
Metal
9%
Sponge Wipe
Carpet
1%
Vacuum
4.2.2.3 Resources
Resource demand based on time, cost, personnel, throughput, and waste were studied to better understand
the implications of the prescribed sampling plan. Each of these factors is defined below, along with its
assigned value.
4.2.2.3.1 Time
The amount of time required to return an area to normalcy following a biological incident is critical. If the
chosen sampling strategy requires an arduous number of samples, the characterization process may take
years or more. For this study, time is defined as the duration needed to collect an individual sample. This
value is based on the type of sample medium. The inputs used for this study were derived from the BOTE
study [1] and are shown in Table 15.
27
-------
Table 15. Sampling Time per Sample Medium
Sample Medium
Time
Unit3
Soil sample
0.07
team hours/sample
Sponge/gauze/wipe
0.08
team hours/sample
Vacuum
0.13
team hours/sample
a "Team hours" is the total number of hours it takes a
single team to collect the sample. A single team
consists of three people.
4.2.2.3.2 Cost
Barring the social and political impacts of a biological incident, cost is one of the primary drivers of
recovery. Cost is a direct derivative of the sampling plan, number of samples, and personnel. There is
very little costing information specific to wide-area characterization, especially for wide-area incidents.
Table 16 shows the assumptions for the sample medium costs, excluding sample collection labor.
Table 16. Cost per Sample Medium
Sample Medium
Cost
Unit
Soil sample
$25
$/sample
Sponge/gauze/wipe
$20
$/sample
Vacuum
$29
$/sample
4.2.2.3.3 Personnel
For wide area incidents, the number of personnel participating in characterization may decrease its
duration; however, the number of skilled personnel available for sampling will be limited. For each of the
three environments, the total number of available response personnel is based on planning documentation
for a city of similar size [2], Personnel numbers are shown in Table 17.
Table 17. Available Response Personnel per Environment and Sample Medium3
Sample Medium
Teams
Per sonnel/T earn
Total Personnel
Outdoor
Soil sample
50
3
150
Sponge/gauze/wipe
100
3
300
Vacuum
N/U
N/U
N/U
Indoor
Soil sample
N/U
N/U
N/U
Sponge/gauze/wipe
150
3
450
Vacuum
150
3
450
Underground
Soil sample
N/U
N/U
N/U
3 The authors acknowledge that the sampling teams would not be limited to collecting one type of sample, but were
arranged in this manner for evaluation purposes. The number of personnel assigned to a particular sampling medium
does affect the Time to completion (TTC) but does not affect the overall labor costs. Due effort should be made to
distribute personnel so that TTC is somewhat equal for all sampling media.
28
-------
Sponge/gauze/wipe
50
3
150
Vacuum
25
3
75
N/U - Not used
4.2.2.3.4 Throughput
Laboratory throughput is defined as the total amount of time needed to process all samples. Throughput
capacity varies depending on the sampling medium, analysis method used, equipment availability,
laboratory personnel, and laboratory operating time. Single laboratory throughput information used for
this study was modeled after data provided by the Interim Consequence Management Guidance for a
Wide-Area Biological Attack document and the New York City Department of Health and Mental
Hygiene Environmental Response and Remediation Plan for Biological Incidents [2, 46]. The total, factor
based, single-laboratory throughput for each sample medium used in this study is shown in Table 18. For
this study, the total number of laboratories available for analysis was assumed to be 250 based on EPA
input.
Table 18. Individual Laboratory Throughput Factors per Sample Medium
Sample Medium
Individual
Laboratory
Throughput
Unit
Soil sample
10
samples/day
Sponge/gauze/wipe
100
samples/day
Vacuum
10
samples/day
4.2.2.3.5 Waste
Waste is generated immediately in an emergency response. Therefore, it is important to plan for waste
early on. Most waste is commonly associated with decontamination. However, sampling should also be
considered as an implicating factor. Waste from sampling is undoubtedly the least studied of all the above
factors. The waste inputs used for this study were based on information collected from the BOTE study
[1] and are shown below in Table 19.
Table 19. Waste Generated per Sampling Medium
Sample Medium
Waste
Unit
Soil sample
1
lb/sample
Sponge/gauze/wipe
0.31
lb/sample
Vacuum
0.31
lb/sample
4.2.3 Sampling Strategy
During the early phase of the incident, the boundary (i.e., plume) of the contamination can be estimated
using intelligence, dispersion modeling, best guess estimates based on epidemiological data, or a
combination of all sources. Screening by way of environmental sampling is used to determine the
identity, concentration, viability, and approximate location of the contamination [46]. Once the evidence
suggests a wide-area release of an infectious biological agent has occurred, characterization sampling
should begin. Characterization is defined as an initial estimate of the environmental extent, concentration,
and characteristics of a given contaminant. Its ultimate purpose is to determine whether, and to what
extent, decontamination is needed. From this information, a decontamination strategy can be derived.
A software tool such as DOE's Visual Sample Plan (VSP) can be used to ensure that a statistically robust
sampling plan is developed. VSP is a decision support tool for determining the number and location of
29
-------
environmental samples based on a variety of sampling goals. Sampling clearance criteria, methods, and
geographical locations are typically defined by the sampling plan well before characterization takes place
[47, 48], The resulting information can be used to develop a sampling plan for use by decision-makers
and help define the problem so that sampling objectives can be determined for characterization,
decontamination configuration, and clearance [49].
For the purposes of this study, VSP was used to develop a sampling strategy. The polygons (i.e.,
shapefiles) used to describe the outdoor, indoor, and underground scenarios in the above analysis were
also used as sampling maps in VSP. Within VSP, there are several sampling strategies (sampling goals)
that can be used for characterization sampling. Given the likelihood of the contaminant being deposited in
patchy spots across a wide area, the "locate a hot spot" goal was used for all three environments. This
goal is used to statistically determine the probability of finding a hot spot of a given size and shape.
Tables 17, 18, and 19 further describe the VSP hot spot sampling design; associated statistical
assumptions; and general guidelines for estimating the number of samples for the outdoor, indoor, and
underground environments.
The inputs to VSP include the following4:
• Sampling Area: The area of the location being sampled for each sample collected;
• Hot Spot Area: The local contiguous area that has concentrations that exceed a threshold value;
• Probability: Probability of a sampling location being collocated with a hot spot; and
• False Negative Rate: Probability each contaminated sample will not be detected.
As previously discussed, Krauter etal. demonstrated that FNRs corresponded with RE data where
smoother surfaces resulted in higher recovery efficiencies and lower FNR [29], Furthermore, the
developers of VSP note the following:
Standard statistical formulas assume that the overall FNR = 0 when calculating 1) the number of
samples required to achieve the desired confidence for a characterization or clearance sampling
goal, and 2) the uncertainty and confidence associated with a characterization or clearance
decision using a specific sampling approach implemented following a contamination incident.
When FNR = 0, the formulas account only for the uncertainty in results associated with the
specific type of statistical or hybrid sampling approach being used. However, the overall FNR is
affected by anything in the sampling process that might yield a false negative, including (i) the
RE of a sampling method (e.g., swab, wipe, or vacuum).5
For the purposes of our analysis, we are assuming that the RE is a function of FNR given that VSP
aggregates anything that affects the sampling process into the FNR.
The choice of inputs is somewhat arbitrary, and in no way reflects EPA policy or even likely strategies for
a real biological incident. However, the choice of inputs does represent a conservative approach to sample
strategy development for a wide area biological incident and is based on currently accepted clearance
levels, common sampling strategies, and available technologies: a first guess at a sampling strategy that
might be used based on expertise of EPA response personnel. Many factors can change and/or reduce the
inputs as well as the number of estimated samples (e.g., on-site conditions, learning through sampling
4 See http://vsp.pnnl.gov/docs/PNNL-23211.pdf. (last accessed March 2017)
5 See http://www.pn.nl.gov/maiti/pnblications/external/techn.ical reports/PNNL-20910.pdf. p.2. (last accessed
March, 2017)
30
-------
results, multiple lines of evidence approaches6, etc.). This analysis does, however, provide an
understanding of how these variables interact and their impacts on the overall resource demand with
regards to a systems approach. Therefore, the results of this study are applicable to almost any sampling
strategy and can help optimize the response.
Table 20. Summary of Sampling Inputs for the Outdoor Environment
Input
Value
Probability of detection
95%
Grid pattern
Square
Area of hot spot7
10 ft2
Total area to sample
5.75 x 107ft2
Spacing between samples
3 meters
Optimum number of samples (VSP output)
3.63 x 106
Table 21. Summary of Sampling Inputs for the Indoor Environment
Input
Value
Probability of detection
95%
Grid pattern
Square
Area of hot spot
10 ft2
Total area to sample
4.16 x 10s ft2
Spacing between samples
3 meters
Optimum number of samples (VSP output)
2.63 x 107
Table 22. Summary of Sampling Inputs for the Underground Transit System Environment
Input
Value
Probability of detection
95%
Grid pattern
Square
Area of hot spot
10 ft2
Total area to sample
1.92 x 106 ft2
Spacing between samples
3 meters
Optimum number of samples (VSP output)
1.21 x 105
6 According to DOE's Multiple Lines of Evidence, " 'lines of evidence' is the ability to identify factors that are
influencing the possible contamination of a given decision area and then convert this knowledge into probabilities
estimating the likelihood of each area still being unacceptable" [50],
7 A 10 ft2 hot spot area was selected based on sampling criteria as defined by the BOTE exercise and best judgment
estimates.
31
-------
4.3 Results
Working with the inputs described above, VSP and a custom spreadsheet were used to estimate the
sampling criteria, the number of samples, and the resources needed to take those samples. The results are
reported by environment type (i.e., outdoor, indoor, and underground) in the following sections.
4.3.1 Outdoor
The results for the outdoor area are shown in Tables 23 and 24. The total number of samples for an area
of 5.75 x 107ft2 with a hot spot area of 10 ft2 is approximately 3,630,000. Based on the total number of
samples and the prescribed sampling medium, it may cost approximately 1.9 billion dollars and require
1.4 years to characterize.
Table 23. Outdoor Samples Sampling Results
Total Samples
Sample Type
Number of
Samples
Sampling
Cost
($)
Labor Cost
($)
Sampling
Time
(team hours)
3,630,000
Soil sample
219,370
5,484,260
6,449,489
15,356
Sponge/gauze/wipe
N/U
N/U
N/U
N/U
Vacuum
3,410,630
98,908,259
181,922,984
433,150
Table 24. Outdoor Samples Laboratory Analysis Results
Total Samples
Sample Type
Number of
Samples
Analysis Cost
($)
Labor Cost
($)
Analysis
Time
(laboratory
hours)
3,630,000
Soil sample
219,370
57,913,781
29,615,002
175,496
Sponge/gauze/wipe
N/U
N/U
N/U
N/U
Vacuum
3,410,630
982,261,329
515,005,072
2,728,504
4.3.2 Indoor
The results for the indoor area are shown in Tables 25 and 26. The total number of samples for an area of
4.16 x 10s ft2 with a hot spot area of 10 ft2 is approximately 26,300,000. Based on the total number of
samples and the prescribed sampling medium, characterization may cost approximately 13.4 billion
dollars and require 8.7 years.
Table 25. Indoor Sampling Results
Total Samples
Sample Type
Number of
Samples
Sampling
Cost
($)
Labor Cost
($)
Sampling
Time
(team hours)
26,300,000
Soil sample
N/U
N/U
N/U
N/U
Sponge/gauze/wipe
2,630,000
52,600,000
90,577,200
215,660
Vacuum
23,670,000
686,430,000
1,262,557,800
3,006,090
32
-------
Table 26. Indoor Analysis Results
Total Samples
Sample Type
Number of
Samples
Analysis Cost
($)
Labor Cost
($)
Analysis
Time
(laboratory
hours)
26,300,000
Soil sample
N/U
N/U
N/U
N/U
Sponge/gauze/wipe
2,630,000
628,570,000
310,340,000
210,400
Vacuum
23,670,000
6,816,960,000
3,574,170,000
18,936,000
4.3.3 Underground Transit System
The results for the underground transit system area are shown in Tables 27 and 28. The total number of
samples for an area of 1.92 * 106 ft2 with a hot spot area of 10 ft2 is about 121,000. Based on the total
number of samples and the prescribed sampling medium, characterization may cost about 1.4 million
dollars and require 34 days.
Table 27. Underground Transit System Sampling Results
Total Samples
Sample Type
Number of
Samples
Sampling
Cost
($)
Labor Cost
($)
Sampling
Time
(team hours)
121,000
Soil sample
N/U
N/U
N/U
N/U
Sponge/gauze/wipe
10,890
217,800
375,052
893
Vacuum
110,110
3,193,190
5,873,267
13,984
Table 28. Underground Transit System Analysis Results
Total Samples
Sample Type
Number of
Samples
Analysis Cost
($)
Labor Cost
($)
Analysis
Time
(laboratory
hours)
121,000
Soil sample
N/U
N/U
N/U
N/U
Sponge/gauze/wipe
10,890
2,602,710
1,285,020
871
Vacuum
110,110
31,711,680
16,626,610
88,088
4.4 Observations
Based on the results of the hypothetical scenario described above, the following observations and
conclusions can be made:
• According to the prescribed sampling plan, environmental characteristics, and resource demand
assumptions, characterization with a total surface area of 4.76 * 10s ft2, will cost an estimated 15.3
billion dollars in total and may take over 10 years to characterize.
• For all three environments combined, the total time for sampling may be up to 4 years, while the total
analysis throughput is estimated to take 10 years. Although this disconnect is rather improbable, the
33
-------
results do lend themselves to the possibility that analysis efforts may eclipse sampling efforts. In
practice, the actual sampling rate (e.g., samples collected per day, etc.) will likely be dictated by
laboratory throughput in part due to sample hold time limitations.
• The total indoor surface area is one order of magnitude greater than the outdoor area, which is to be
expected for a highly-populated urban setup. Because of the increased surface area, the indoor
environments represent almost 90% of the total cost and resource demand of all the environments
combined. Because of this, for wide-area biological incidents, it is recommended that a rule in/rule
out approach be taken when sampling indoors. If a positive hit is discovered through preliminary
sampling, the building should be considered contaminated unless the decontamination approach is
costlier than further sampling to define more discrete decontamination zones. This analysis assumed
that all buildings within the contaminated area would either be contaminated or need to be sampled
thoroughly. Available field data or a lines of evidence approach may be useful in reducing this
requirement
• It is likely the requirements set forth by this environment are not practical and would likely exhaust
local, state, and federal resources. Furthermore, by the time characterization has concluded (i.e., 10
years later), it is possible that the community and its economy will no longer be viable.
5 WIDE-AREA INCIDENT ANALYSIS
5.1 Introduction
The process to develop a sampling strategy for wide-area biological incident response operations has up
to now been poorly defined and not systematic. Additionally, the options available in terms of sampling
techniques are often constrained by regulatory mandates. In situations where the sampling method is not
mandated by regulation(s), it is then necessary to analyze available data concerning both sampling
methods and sampling techniques before selecting the most appropriate method or technique [51],
The purpose of this chapter is fourfold: 1) to leverage the capacity of VSP outputs to build a predictive
model for wide-area sampling strategies by way of regression modeling, and 2) using the hypothetical
scenario described in Section 4.2.1, model the number of samples needed for the underground, outdoor,
and indoor environments individually under a range of values for VSP input parameters (sampling area,
hot spot area, false negative rate, and probability); 3) evaluate the relative impact of the VSP input
parameters on the resulting number(s) of samples; and 4) use the modeling results (for the number of
samples required) in conjunction with resource demand estimates (based on a review of the literature and
the BOTE study) to analyze the potential impacts on total time (for sampling and analysis) and total cost
impacts following a wide area biological incident. The results of this chapter and the accompanying
developed spreadsheet tool will assist responders in evaluating the impacts of multiple variables
associated with a range of sampling techniques, and help inform the selection of an appropriate sampling
technique based on an informed evaluation of those variables according to the project's goals and
requirements.
5.2 Sampling Models
The following sections describe the approach used in developing several models for determining the most
appropriate number of samples based on a given sampling area, hot spot area, probability, and false
negative rate for all three environments (i.e., underground, outdoor, and indoor), using data generated by
VSP.
34
-------
5.2.1 VSP Modeling
A preliminary assessment of the ability of VSP to model wide area incidents showed instability due to
software memory limitations. Because of this, regression modeling was used to expand the limitations set
forth by VSP. The equations derived by regression modeling can be used to predict the number of
samples using a simple spreadsheet without interacting with VSP. Additionally, since the results are
generated and stored within a spreadsheet, a more extensive analysis of the data can be performed.
Furthermore, by using this approach, the relative impact that the predictors (i.e., sampling area, hot spot
area, probability, false negative rate) have on the target variable (i.e., number of samples) can be
analyzed.
The data needed for the regression analysis were generated by automating VSP. An Automation (i.e.,
OLE Automation) method was used to remotely access and manipulate VSP using VBscript (i.e., Visual
Basic Scripting Edition). The input parameters (Table 29) are sampling area (SA), hot spot area,
probability, and FNR, and their associated range of inputs referenced by VSP is shown in Table 30. Two
data sets were derived from VSP for each environment; a separate dataset for a range of false negative
rates and another for a range of sampling areas (VSP does not allow sampling area and false negative rate
to be varied simultaneously).
Table 29. Input Parameters Used for VSP Modeling
VAP Parameter
Meaning
Sampling Area (SA)
The area of the location being sampled
Hot Spot Area
The local contiguous area that has concentrations that
exceed a threshold value
Probability
Probability of a sampling location being collocated with a
hot spot
False Negative Rate
Probability each contaminated sample will not be detected
Table 30. Range of Values Used for VSP Input Parameters
Variable
Scenario
False
Negative Rate
(%)
Sampling
Area3 (ft2)
Probability
(%)
Hotspot Area
(ft2)
SA
Indoor
-
50-1,000
50-99
50-1,000
SA
Outdoor
-
50-1,000
50-99
50-1,000
SA
Underground
-
10-1,000
50-99
10-1,000
FNR
Indoor
1-8
-
50-95
55-1,000
FNR
Underground
1-10
-
50-99
4-1,000
a The sampling area value represents the area of the location being sampled on a per sample basis. For example, a
value of 50 ft2 for a vacuum indicates that the vacuum nozzle is passed over 50 ft2 for every sample collected. For a
fixed area of interest, as the sampling area increases, the number of samples decreases.
The resulting data sets from VSP were each analyzed using Eureqa8, totaling six datasets comprised of
over five million data points. Eureqa is a genetic programming-based symbolic regression mathematical
software tool for detecting equations and hidden mathematical relationships in scientific data. Eureqa
identified and ranked regression models that best corresponded with the input data. Equations that were
8 www.nutonian.com
35
-------
most representative of the training set or highest relevance to the VSP results are discussed in the
following section.
5.2.2 Regression Modeling Results
Equations 1-5 show the regression models for false negatives and assume no false negatives for the
underground, outdoor, and indoor environments, respectively. Limitations of VSP prevented the
generation of sufficient data to perform a regression modeling analysis for the outdoor environment with
false negatives. These equations were identified by Eureqa to have the highest relevance to the VSP
results. Tables 31-35 show the model performance statistics for those equations.
S = 5960*FNR/(HA*sin(2.3M + P)) + (2.467e7 + 41.67 *P3 + 2.079e5
*sin(2.384 + .P) + 1.812 *FNR*P3 + 1.9e4*P*cos(41.67*P))/i£4 Eq.l
where S = number of samples, dimensionless
SA = sampling area, ft2
FNR = false negative rate, %
HA = hot spot area, ft2
P = probability, %
Table 31. Model Statistics with False Negatives in the Indoor Environment
Mean Absolute Error
981.87148
Mean Square Error
4523080
R2 Goodness of Fit
0.99981571
Correlation Coefficient
0.99991905
Maximum Error
32765.938
S = 308.9*sqrt(&4) + -1683*P/&4 + 1.086e6*P/SA2 + 2.659e7
/(HS*0.9335 +&4*sqrt(&4 -26.93)) -5513 - tan(308.9*sqrt(&4)) - 4.685*&4 Eq.2
Where, S = number of samples, dimensionless
SA = sampling area, ft2
HA = hot spot area, ft2
P = probability, %
Table 32. Model Statistics with no False Negatives in the Indoor Environment
Mean Absolute Error
77.149587
Mean Square Error
53011.12
R2 Goodness of Fit
0.99965262
Correlation Coefficient
0.99982948
Maximum Error
5121.8342
S = 0.05987*SA + 5.448e4/(&4 - 24.77) + 4.378e7*P/(2.828e4 + 150.5
*HA + SA2 + 4.239*&4*HA + P*SA2) - 99.24 Eq.3
36
-------
Table 33. Model Statistics with no False Negatives in the Outdoor Environment
Mean Absolute Error
12.133013
Mean Square Error
1864.24
R2 Goodness of Fit
0.9993602
Correlation Coefficient
0.99968813
Maximum Error
1510.3916
5 = 3.541 + -6.385e4/SA2 + 1.484e5*sqrt(P/(SA4 + 7.232*HA*SA2)) +
2.232e5*SA2/(SA4 + 7.232*HA*SA2) - 0.003847*&4 Eq.4
Table 34. Model Statistics with no False Negatives in the Underground Environment
Mean Absolute Error
3.7782199
Mean Square Error
242.60252
R2 Goodness of Fit
0.99746774
Correlation Coefficient
0.99882424
Maximum Error
842.55387
S = (1.906e4*Probability + 1.345e4*FNR + 0.1514*exp(0.1547*P + 0.1162*FNR) -
FNR*P*tan(1.095*P))/HA Eq.5
Table 35. Model Statistics with False Negatives in the Underground Environment
Mean Absolute Error
111.86019
Mean Square Error
595210.75
R2 Goodness of Fit
0.99967441
Correlation Coefficient
0.99983964
Maximum Error
79767.245
5.2.2.1 Variable Sensitivity
Eureqa features the ability to show how changes in each variable affect the generated models. For the
purposes of this section, sensitivity is defined as the relative impact that a predictor has on the target
variable (i.e., number of samples). The variable sensitivity of the selected best fit models is discussed in
the following sections.
5.2.2.1.1 No False Negatives Errors
When varying sampling area, hot spot area, and probability, the sampling area has a relatively higher
impact on the number of samples versus hot spot area and probability for all three environments (i.e.,
underground, outdoor, and indoor). Although this phenomenon remains constant across all three
locations, the significance of this impact decreases with increasing surface area. Tables 36, 37, and 38
show the variable sensitivity for the underground outdoor and indoor environments with no false
negatives, respectively.
37
-------
Table 36. Variable Sensitivity with no False Negatives in the Underground Environment Model
Variable
Sensitivity9
% Positive10
Positive
Magnitude11
% Negative12
Negative
Magnitude13
Sampling Area
4.9421
0%
0
100%
4.9421
Hotspot Area
0.51549
0%
0
100%
0.51549
Probability
0.073084
100%
0.073084
0%
0
Table 37. Variable Sensitivity with no False Negatives in the Outdoor Environment Model
Variable
Sensitivity
% Positive
Positive
Magnitude
% Negative
Negative
Magnitude
Sampling Area
2.7538
0%
0
100%
2.7538
Hotspot Area
0.094395
0%
0
100%
0.094395
Probability
0.032768
100%
0.032768
0%
0
Table 38. Variable Sensitivity with no False Negatives in the Indoor Environment Model
Variable
Sensitivity
% Positive
Positive
Magnitude
% Negative
Negative
Magnitude
Sampling Area
2.7736
3%
0.011193
97%
2.8636
Hotspot Area
0.094389
0%
0
100%
0.094389
Probability
0.032678
100%
0.032678
0%
0
5.2.2.1.2 With False Negatives Errors
Conversely, when varying the false negative rate, hot spot area, and probability for the underground and
indoor environments, the hot spot area has a higher relative impact (17.085) on the number of samples
versus probability (0.38083) and false negative rate (0.070417) for the underground scenario (i.e., smaller
environment). For the indoor environment (i.e., larger environment) the hot spot area also has a higher
relative impact (2.03) on the number of samples versus probability (0.35668) and false negative rate
(0.059827). These inverse outcomes (compared to the results where there were no false negatives) may be
due to the increase in the area of interest, percentage (probability) versus a set range of integers (i.e.,
hotspot size), and the statistical method used to account for false negatives. Notably, the false negative
rate has relatively low impact on the number of samples. However, the models were less sensitive to these
variables as the area of interest increased. Tables 39 and 40 show the variable sensitivity for the
underground outdoor, and indoor environments with false negatives, respectively.
9 The relative impact within this model that a variable has on the target variable.
10 The likelihood that increasing this variable will increase the target variable.
11 When increases in this variable lead to increases in the target variable, this is generally how big the positive
impact is.
12 The likelihood that increasing this variable will decrease the target variable.
13 When increases in this variable lead to decreases in the target variable, this is generally how big the negative
impact is.
38
-------
Table 39. Variable Sensitivity when Accounting for False Negatives in the Underground
Environment Model
Variable
Sensitivity
% Positive
Positive
Magnitude
% Negative
Negative
Magnitude
Hotspot Area
17.085
0%
0
100%
17.085
Probability
0.38083
100%
0.38083
0%
0
False Negative Rate
0.070417
100%
0.070417
0%
0
Table 40. Variable Sensitivity when Accounting for False Negatives in the Indoor Environment
Model
Variable
Sensitivity
% Positive
Positive
Magnitude
% Negative
Negative
Magnitude
Hot Spot Area
2.03
0%
0
100%
2.03
Probability
0.35668
100%
0.35668
0%
0
False Negative Rate
0.059827
100%
0.059827
0%
0
5.3 Resource Demands
As discussed above, when comparing the impacts of (1) varying false negative rates on the number of
samples, and (2) varying the sampling area on the number of samples, variations in the sampling area had
a greater relative impact on the required number of samples than did variations in the false negative rate.
To evaluate the additional impacts that resource demands might have on the total time to complete
sampling and analysis and on the total cost, a Microsoft Excel spreadsheet was developed. The
spreadsheet combines the regression equations generated as described above for the number of samples
(based only on a variable sampling area) with a variety of different numerical factors for various
resources that would be needed following an event. These inputs and their associated "default" factors are
listed in Table 41.
39
-------
Table 41. Table of Factors Used in Resource Demand Estimation Spreadsheet
Component
Default Value
I nit
Basis
Sampling Plan
Probability
95
%
EPA
Hotspot Area
10
ft2
EPA
Sample Area Size
1
ft2
EPA
Sampling
Number of Available Teams for Sampling
500
teams
Estimated
Personnel per Sampling Team
3
persons/team
Assumed
Time to Collect Samples
Soil Sample
0.07
team
hours/sample
Estimated
Sponge Sample
0.08
team
hours/sample
BOTE; Adjusted Time Per Sample, pg.
H-34 [1]
Gauze/Wipe Sample
0.08
team
hours/sample
Estimated
Vacuum Sample
0.13
team
hours/sample
BOTE; Adjusted Time Per Sample, pg.
H-34 [1]
Sampling Team Hours per Shift
5
team
hours/shift
Assumed
Sampling Team Shifts per Day
1
team
shifts/day
Calculated
Sampling Hours per Day
2500.0
sampling
hours/day
Calculated
Sampling Personnel Hours per Day
7500.0
person
hours/day
Calculated
Sampling Team Labor Cost
420.00
$/hour/team
BOTE; Team Makeup, pg. H-33 [1]
Sampling Personnel Labor Cost
140.00
$/hour/person
Calculated
Sampling Material Cost
Soil Sample
25.00
$/sample
Estimated
Sponge Sample
20.00
$/sample
BOTE; Cost Equations, pg. H-6 [1]
Gauze/Wipe Sample
20.00
$/sample
Estimated
Vacuum Sample
29.00
$/sample
BOTE; Cost Equations, pg. H-6 [1]
Analysis
Number of Available Laboratories for Analysis
250
laboratories
EPA
40
-------
Component
Dol'ii 11 It \ ill no
I nil
liiisis
Time to Analyze Samples
Soil Sample
10
samples/day
EPA and Estimated based on IBRD +
NYC
Sponge Sample
100
samples/day
EPA and Estimated based on IBRD +
NYC
Gauze/Wipe Sample
100
samples/day
EPA and Estimated based on IBRD +
NYC
Vacuum Sample
10
samples/day
EPA and Estimated based on IBRD +
NYC
Analysis Laboratory Hours per Day
24
hours/day/
laboratory
Assumed
Analysis Hours per Day
6000.0
analysis
hours/day
Based on a network of 250 laboratories
Analysis Labor Cost
Soil Sample
135.00
$/sample
Estimated
Sponge Sample
118.00
$/sample
BOTE; Analytical Costs, pg. H-5 [1]
Gauze/Wipe Sample
118.00
$/sample
Estimated
Vacuum Sample
151.00
$/sample
BOTE; Analytical Costs, pg. H-5 [1]
Analysis Material Cost
Soil Sample
264.00
$/sample
Estimated
Sponge Sample
239.00
$/sample
BOTE; Analytical Costs, pg. H-5 [1]
Gauze/Wipe Sample
239.00
$/sample
Estimated
Vacuum Sample
288.00
$/sample
BOTE; Analytical Costs, pg. H-5 [1]
41
-------
The spreadsheet calculates/estimates the values listed in Table 42 (primary outputs shown in bold):
Table 42. Resource Demand Estimation Spreadsheet Outputs
Total Number of Samples
Total Required Sampling Time (team h)
Time to Complete Sampling (days)
Total Sampling Labor Cost ($)
Total Sampling Material Cost ($)
Total Required Analysis Time (laboratory h)
Time to Complete Analyses (days)
Total Analysis Labor Cost ($)
Total Analysis Material Cost ($)
Total Cost ($)
Total Time to Completion (days)
Limiting Time Factor
There are three separate regression equations for the number of samples based on sampling area, hot spot
area, and probability (one each for the indoor, outdoor, and underground environments). Furthermore,
different sampling methods/techniques are assumed for each of the different environment types based on
the types of surfaces in each environment (see Section 4). The total number of samples for each
environment type is distributed according to the percentage of the surfaces in each environment for which
a different sampling technique might be used. For example, if 364 samples are needed for the indoor
environment, and 10% of the indoor surfaces will be sampled using sponge sampling and 90% will be
vacuumed, then there will be 36 sponge samples and 328 vacuum samples. This procedure is repeated for
the number of samples for each of the three environment types, and the resource demands are calculated
for each sampling technique in each environment type. Those resource demand estimates for each
environment type are then totaled giving the primary outputs listed in Table 41 above.
5.3.1 Resource Demands Analysis
As Table 41 shows, default values were selected for the resource demand factors based largely on existing
data and information from EPA reports and field studies, assumptions, and educated estimates. The three
main variables evaluated for this analysis were: (1) sample area size, (2) number of available sampling
teams, and (3) number of available analysis laboratories. Sampling area was varied across the narrowest
range of values used as inputs to the VSP modeling runs for the three environment types, the number of
available sampling teams was varied from 100 to 500 teams, and the number of available analysis
laboratories was varied from 100 to 500 laboratories.
5.3.2 Resource Demands Results
Assuming all variables are fixed except for the sample area size, number of sampling teams, and number
of laboratories, the results of varying those three remaining variables on total cost and number of days
required for sampling and number of days required for analysis are presented in Tables 43 and 44,
respectively.
42
-------
Table 43. Effect of Varying the Sample Area Size and Number of Sampling Teams on Total Cost
and Time to Complete Sampling
Sample Area Size =
50 ft2
Sample Area Size =
500 ft2
Sample Area Size =
1,000 ft2
Sampling
Teams
Total Cost
($)
Sampling
Days
Total Cost
($)
Sampling
Days
Total Cost
($)
Sampling
Days
100
81,470,347
39
928,441
0.45
217,320
0.10
200
81,470,347
20
928,441
0.22
217,320
0.052
300
81,470,347
13
928,441
0.15
217,320
0.035
400
81,470,347
9.8
928,441
0.11
217,320
0.026
500
81,470,347
7.8
928,441
0.089
217,320
0.021
Table 44. Effect of Varying the Sample Area Size and Number of Analysis Laboratories on Total
Cost and Time to Complete Analyses
Sample Ai
50
•ea Size =
Ft2
Sample Area Size =
500 ft2
Sample Area Size =
1,000 ft2
Analysis
Laboratories
Total Cost
($)
Analysis
Days
Total Cost
($)
Analysis
Days
Total Cost
($)
Analysis
Days
100
81,470,347
49
928,441
0.56
217,320
0.13
200
81,470,347
24
928,441
0.28
217,320
0.065
300
81,470,347
16
928,441
0.19
217,320
0.044
400
81,470,347
12
928,441
0.14
217,320
0.033
500
81,470,347
9.8
928,441
0.11
217,320
0.026
Based on the results presented in Tables 43 and 44, the total cost is a function only of the sample
area size (i.e., as the area covered by each sample collected increases, the number of samples required for
a given area of interest decrease) and not on either the number of sampling teams collecting those samples
nor on the number of laboratories available to analyze those samples. However, the amount of time
needed to both collect the samples and to analyze them is a function of the number of sampling teams
available and the number of analysis laboratories available. Once the number of samples is determined,
the total cost is fixed, and is independent of the number of sampling teams and the number of laboratories.
For a range of sample area sizes from 50 to 1,000 ft2, Figure 2 illustrates the impact on both total cost and
the total time for both sampling and analysis combined, but reflects the limiting time value (i.e., if
sampling takes 20 days and analysis takes 30 days, then the total time to completion is 30 days).
43
-------
Sampling Area (ft2)
-~-Total Cost ($) -"-Total Time (days)
Figure 2. Total Cost and Total Time for Sampling and Analysis as a Function of Sample Area Size
Once the total cost is determined based only on the number of samples required, the impact of a
variable number of available sampling teams and analysis laboratories was evaluated. Figures 3 through 5
present the total times to complete sampling and analyses based on the available number of sampling
teams and the number of available analysis laboratories for three increments of sample area sizes, 50 ft2,
500 ft2, and 1,000 ft2, respectively.
44
-------
Number of Sampling Teams / Number of Labs
¦•-SamplingTime -"-AnalysisTime -*-Total Cost
Figure 3. Sampling and Analysis Time as a Function of Number of Available Resources at 50 ft2
Sample Area Size
45
-------
Number of Sampling Teams / Number of Labs
-~-SamplingTime -"-AnalysisTime -*-Total Cost
Figure 4. Sampling and Analysis Time as a Function of Number of Available Resources at 500 ft2
Sample Area Size
46
-------
Number of Sampling Teams / Number of Labs
-SamplingTime -"-AnalysisTime -*-Total Cost
Figure 5. Sampling and Analysis Time as a Function of Number of Available Resources at 1,000 ft2
Sample Area Size
6 OBSERVATIONS AND RECOMMENDATIONS
Several observations can be made based on the results of the above analysis of a sampling campaign
following a hypothetical wide-area incident that can potentially impact indoor, outdoor, and underground
transportation areas. Primarily, the study demonstrated that a wide area incident response will require
extended time and resources if a probabilistic sampling approach is utilized with high probability
requirements (i.e., >90%) and currently available sampling methods.
The current study used an accepted and commonly used tool (VSP) for developing a probabilistic
sampling design and followed accepted guidance for sampling strategy/methods. VSP may have some
limitations when applied to a wide-area sampling design. These limitations include: memory limitations
when running large scenarios, and uncertainty as to how VSP calculates total numbers of samples,
especially for indiscriminate surface areas. Further, EPA uses different terminology than VSP. Many of
the VSP inputs and the terminology used (e.g., false negative rate, probability, etc.) are used
inconsistently within the operational response community, and the primary users of VSP may have a
relatively limited understanding of the statistical significance of the VSP inputs, the VSP model, and
probabilistic design in general. Many, if not the majority, of bench-scale studies reviewed for this report
use different terminology (e.g., recovery efficiency), and there seems to be a lack of understanding of the
47
-------
correlation or relationship between parameters quantified in a laboratory and operationally significant
parameters used as inputs for sampling design.
Additionally, variation in recovery efficiency (and thus detection probability) by surface material types
(e.g., porous, nonporous, soil) are not accounted for in VSP which thus cannot be evaluated in the context
of wide area characterization. Additional information must be gathered on the distribution of surface
types within the study area before appropriate sampling methods can be prescribed. Other widely used
EPA tools such WEST and IWASTE and other GIS tools can be useful for wide area sample plan
development by assisting responders in determining or estimating the presence and distribution of surface
material types for both structures and open areas.
Further, due to the large area of potential contamination, the variety of indoor and outdoor surface types,
and uncertain and variable environmental conditions, probabilistic sampling design (and statistical design
in general) may not be appropriate for a wide area incident and likely would not be used for
characterization of a wide area incident. In addition, the current study used a plume derived from an
earlier biological exercise to define the area of contamination; however, recognition of a covert release
through illness would likely required a different strategy and could result in a much larger initial area of
interest.
Considering the observations noted above, several needs have been identified. First, a lines of evidence
approach to sampling design should be considered at the outset of the response planning phase of a wide
area incident for the eventual purpose of reaching clearance goals. Second, we currently lack operational
strategies to deal with such an incident, and cohesive operational information must be developed to help
inform the development of those strategies.
In addition, the current sampling methods used to characterize a wide area would result in an
overwhelmingly unrealistic number of samples that must be collected and analyzed. Existing availability
of resources (sampling personnel and analytical laboratories) would not be able to handle such a large
demand in even a reasonable amount of time. Different sampling methods must be developed to
characterize the extent of contamination in wide areas. Those novel sampling methods must focus on
larger sampling area sizes, be usable for outdoor surfaces, and must require less time and personnel to
collect samples. For example, "native air samplers," such as HVAC filters, bus/train/car cabin air filters,
and ambient air quality samples may be used as potential sampling resources. As the above analysis has
demonstrated, improvements to existing sampling methods with respect to recovery efficiencies would
largely be inconsequential to reducing the resource demand (total cost and time to characterize) compared
to improvements to sample area size (even using probabilistic sampling design approaches).
Also, there is a lack of understanding concerning the fate and transport of spores over long periods of
time. This understanding is critical when considering the amount of time potentially required to conduct
characterization sampling. Spore fate and transport must be understood for effective sampling and
remediation strategy development. Lastly, this study supports further consideration of a combined
sampling design approach using probabilistic and non-probabilistic sampling when characterizing a wide
area incident. Additional decision support tools are needed to help direct sampling efforts for wide area
incidents where time and cost considerations are critical factors.
48
-------
1.
2.
3.
4.
5.
6.
7.
8.
9.
10
11
12
13
14
15
16
17
18
19
BIBLIOGRAPHY
U.S. Environmental Protection Agency, Bio-Response Operational Test and Evaluation (BOTE)
Project; EPA/600/R-13/168; EPA: Washington, DC, 2013.
New York City Department of Environmental Health and Mental Hygiene, Environmental
Response and Remediation Plan for Biological Incidents; NYC DOHMH: New York (NY), 2015.
United States Government Accountability Office, Anthrax detection: Agencies need to validate
sampling activities in order to increase confidence in negative results; GAO-05-251; GAO:
Washington (DC), 2005.
Stone Bahr, E.L. Biological Weapons Attribution: A Primer. Master's Thesis, Naval Postgraduate
School, Monterey, CA, 2007.
Martin, J.W., G.W. Christopher, and E.M. Eitzen. History of biological weapons: from poisoned
darts to intentional epidemics. In Medical Aspects of Biological Warfare; Z.F. Dembek, Ed.;
Office of The Surgeon General US Army Medical Department Center and School Borden
Institute: Washington, DC,2007.
Schmitt, K. and N.A. Zacchia. Total Decontamination Cost of the Anthrax Letter Attacks.
Biosecur Bioterror. 2012,10 (1), 98-107.
U.S. General Accounting Office, Capitol Hill Anthrax Incident: EPA's Cleanup was Successful;
Opportunities Exist to Enhance Contract Oversight; GAO-03-686; GAO: Washington, DC,
2003.
Dembek, Z.F., M.G. Kortepeter, and J.A. Pavlin. Discernment Between Deliberate and Natural
Infectious Disease Outbreaks. Epidemiol. Infect. 2007,135 (3), 353-371.
Guillemin, J. The 1979 Anthrax Epidemic in the USSR: Applied Science and Politcal
Controversy. Proceedings of the American Philosophical Society. 2002,146 (1), 18-36.
Sharp, R.J. and A.G. Roberts. Anthrax: the Challenges for Decontamination. J. Chem. Technol.
Biotechnol. 2006,57 (10), 1612-1625.
Tucker, J.B. Biological Weapons in the Former Soviet Union: an Interview with Dr. Kenneth
Alibek. The Nonproliferation Review. 1999, Spring-Summer, 1-10.
Manchee, R.J., M.G. Broster, J. Melling, R.M. Henstridge, and A.J. Stagg. Bacillus anthracis on
Gruinard Island. Nature. 1981, 294 (19), 254-255.
Mierzejewski, J.B., M. Decontamination of soil after bacterial warfare experiments on Gruinard
Island. PrzegladEpidemiologiczny. 1991,45 (3), 197-205.
Hutchison, R., Weapons of Mass Destruction: The No-Nonsense Guide to Nuclear, Chemical and
Biological Weapons Today. George Weidenfeld & Nicholson: London, UK, 2003.
Pottage, T., E. Goode, S. Wyke, and A.M. Bennett. Responding to Biological Incidents—What are
the Current Issues in Remediation of the Contaminated Environment? Envrion. Int. 2014, 72,
133-139.
U.S. Environmental Protection Agency, Comprehensive Biological Tactical Guidebook (Draftj;
EPA: Washington, DC, 2015.
Centers for Disease Control and Prevention. Surface Sampling Procedures for Bacillus anthracis
Spores from Smooth, Non-Porous Surfaces, http ://www .cdc .gov/n iosh/topics/em res/su rface-
sampling-bacillus-anthracis.html (accessed May 11, 2017).
Brown, G.S., R.G. Betty, J.E. Brockmann, D.A. Lucero, C.A. Souza, K.S. Walsh, R.M. Boucher,
M.S. Tezak, M.C. Wilson, T. Rudolph, H.D. Lindquist, and K.F. Martinez. Evaluation of Rayon
Swab Surface Sample Collection Method for Bacillus Spores from Nonporous Surfaces. J. Appl.
Microbiol. 2007,103 (4), 1074-1080.
Buttner, M.P., P. Cruz, L.D. Stetzenbach, and T. Cronin. Evaluation of Two Surface Sampling
Methods for Detection of Erwinia herbicola on a Variety of Materials by Culture and
Quantitative PCR. Appl. Environ. Microbiol. 2007, 73 (11), 3505-3510.
49
-------
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Edmonds, J.M., P.J. Collett, E.R. Valdes, E.W. Skowronski, G.J. Pellar, and P.A. Emanuel.
Surface Sampling of Spores in Dry-Deposition Aerosols. Appl. Environ. Microbiol. 2009, 75 (1),
39-44.
Estill, C.F., P.A. Baron, J.K. Beard, M.J. Hein, L.D. Larsen, L. Rose, F.W. Schaefer, 3rd, J.
Noble-Wang, L. Hodges, H.D. Lindquist, G.J. Deye, and M.J. Arduino. Recovery Efficiency and
Limit of Detection of Aerosolized Bacillus anthracis Sterne from Environmental Surface
Samples. Appl. Environ. Microbiol. 2009, 75 (13), 4297-4306.
Hodges, L.R., L.J. Rose, A. Peterson, J. Noble-Wang, and M.J. Arduino. Evaluation of a
Macrofoam Swab Protocol for the Recovery of Bacillus anthracis Spores from a Steel Surface.
Appl. Environ. Microbiol. 2006, 72 (6), 4429-4430.
Kirschner, L.E. and J.R. Puleo. Wipe-Rinse Technique for Quantitating Microbial Contamination
on Large Surfaces. Appl. Environ. Microbiol. 1979, 38 (3), 466-70.
Piepel, G.F., B.G. Amidan, and R. Hu. Laboratory Studies on Surface Sampling of Bacillus
anthracis Contamination: Summary, Gaps and Recommendations. J. Appl. Microbiol. 2012,113
(6), 1287-304.
Probst, A., R. Facius, R. Wirth, and C. Moissl-Eichinger. Validation of a Nylon-Flocked-Swab
Protocol for Efficient Recovery of Bacterial Spores from Smooth and Rough Surfaces. Appl.
Environ. Microbiol. 2010, 76 (15), 5148-5158.
Probst, A., R. Facius, R. Wirth, M. Wolf, and C. Moissl-Eichinger. Recovery of Bacillus Spore
Contaminants from Rough Surfaces: a Challenge to Space Mission Cleanliness Control. Appl.
Environ. Microbiol. 2011, 77 (5), 1628-1637.
Rose, L., B. Jensen, A. Peterson, S.N. Baneijee, and M.J. Srduino. Swab Materials and Bacillus
anthracis Spore Recovery from Nonporous Surfaces. Emerging Infect. Dis. 2004,10 (6), 1023-
1029.
Tufts, J.A., K.M. Meyer, M.W. Calfee, and S.D. Lee. Composite Sampling of a Bacillus
anthracis Surrogate with Cellulose Sponge Surface Samplers from a Nonporous Surface. PLoS
One. 2014, 9 (12), 1-16.
Krauter, P.A., G.F. Piepel, R. Boucher, M. Tezak, B.G. Amidan, and W. Einfeld. False-Negative
Rate and Recovery Efficiency Performance of a Validated Sponge Wipe Sampling Method. Appl.
Environ. Microbiol. 2012, 78 (3), 846-854.
Krauter, P.A., G.F. Piepel, R. Boucher, M. Tezak, B.G. Amidan, and W. Einfeld. False Negative
Rate and Other Performance Measures of a Sponge-Wipe Surface Sampling Method for Low
Contaminant Concentrations; SAND2011-3395; S.N. Laboratories: Albuquerque, NM and
Livermore, CA, 2011.
Lewandowski, R., K. Kozlowska, M. Szpakowska, M. Stepinska, and E.A. Trafny. Use of a Foam
Spatula for Sampling Surfaces after Bioaerosol Deposition. Appl. Environ. Microbiol. 2010, 76
(3), 688-694.
Rose, L.J., L. Hodges, H. O'Connell, and J. Noble-Wang. National Validation Study of a
Cellulose Sponge Wipe-Processing Method for Use after Sampling Bacillus anthracis Spores
from Surfaces. Appl. Environ. Microbiol. 2011, 77 (23), 8355-8359.
Einfeld, W., R.M. Boucher, M.S. Tezak, M.C. Wilson, and G.S. Brown. Evaluation of Surface
Sampling Method Performance for Bacillus Spores on Clean and Dirty Outdoor Surfaces;
SAND2011-4085; S.N. Laboratories: Albuquerque, NM and Livermore, CA, 2011.
Brown, G.S., R.G. Betty, J.E. Brockmann, D.A. Lucero, C.A. Souza, K.S. Walsh, R.M. Boucher,
M. Tezak, M.C. Wilson, and T. Rudolph. Evaluation of a Wipe Surface Sample Method for
Collection of Bacillus Spores from Nonporous Surfaces. Appl. Environ. Microbiol. 2007, 73 (3),
706-710.
Sanderson, W.T., M.J. Hein, L. Taylor, B.D. Curwin, G.M. Kinnes, T.A. Seitz, T. Popovic, H.T.
Holmes, M.E. Kellum, S.K. McAllister, D.N. Whaley, E.A. Tupin, T. Walker, J.A. Freed, D.S.
Small, B. Klusaritz, and J.H. Bridges. Surface Sampling Methods for Bacillus anthracis Spore
Contamination. Emerging Infect. Dis. 2002, 8 (10), 1145-1151.
50
-------
36. Brown, G.S., R.G. Betty, J.E. Brockmann, D.A. Lucero, C.A. Souza, K.S. Walsh, R.M. Boucher,
M.S. Tezak, and M.C. Wilson. Evaluation of Vacuum Filter Sock Surface Sample Collection
Method for Bacillus Spores from Porous and Non-Porous Surfaces. J. Environ. Monit. 2007, 9
(7), 666-671.
37. Calfee, M.W., L.J. Rose, S. Morse, D. Mattorano, M. Clayton, A. Touati, N. Griffin-Gatchalian,
C. Slone, and N. McSweeney. Comparative Evaluation of Vacuum-Based Surface Sampling
Methods for Collection of Bacillus Spores. J. Microbiol. Methods. 2013, 95 (3), 389-396.
38. Paton, S., K.-A. Thompson, S.R. Parks, and A.M. Bennett. Reaerosolization of Spores from
Flooring Surfaces To Assess the Risk of Dissemination and Transmission of Infections. Appl.
Environ. Microbiol. 2015, 81 (15), 4914-4919.
39. Clark Burton, N., A. Adhikari, S.A. Grinshpun, R. Hornung, and T. Reponen. The Effect of Filter
Material on Bioaerosol Collection of Bacillus subtilis Spores Used as a Bacillus anthracis
Simulant. J. Environ. Monit. 2005, 7 (5), 475-480.
40. U.S. Environmental Protection Agency, Systematic Evaluation of Aggressive Air Sampling for
Bacillus anthracis Spores; EPA 600/R-13/068; EPA: Research Triangle Park, NC, 2013.
41. Lee, S.D., M.W. Calfee, L. Mickelsen, S. Wolfe, J. Griffin, M. Clayton, N. Griffin-Gatchalian,
and A. Touati. Evaluation of Surface Sampling for Bacillus Spores Using Commercially
Available Cleaning Robots. Environ. Sci. Technol. 2013, 47 (6), 2595-2601.
42. Lee, S.D., M. Calfee, L. Mickelsen, M. Clayton, and A. Touati. Scenario-Based Evaluation of
Commercially Available Cleaning Robots for Collection of Bacillus Spores from Environmental
Surfaces. Remediation. 2014, 24 (2), 123-133.
43. U.S. Department of Homeland Securty, Wide Area Recovery and Resiliency Program (WARRP)
Integrated Program Plan; 2.3.0; DHS: Washington, DC, 2011.
44. U.S. Environmental Protection Agency, Waste Estimation Support Tool and User Guide
EPA/600/B-14/235; EPA: Washington, DC, 2014.
45. U.S. Environmental Protection Agency. Incident Waste Decision Support Tool (I-WASTE DST)
Version 6.4. http://www2.ergweb.com/bdrtool/login.asp (accessed May 10, 2017).
46. Raber, E., R. Kirvel, D. MacQueen, A. Love, M. Dombroski, T. McGrann, J. Richards, C.
Melius, T. Bunt, and W. Hibbard. Interim Consequence Management Guidance for a Wide-Area
Biological Attack; LLNL-TR-484706; L.L.N. Laboratory: Livermore, CA, 2011.
47. U.S. Environmental Protection Agency, Guidance on Choosing a Sampling Design for
Environmental Data Collection for Use in Developing a Quality Assurance Project Plan (EPA
QA/G-5S); EPA/240/R-02/005; EPA: Washington, DC, 2002.
48. U.S. Environmental Protection Agency, Guidance on Systematic Planning Using the Data
Quality Objectives Process (EPA QA/G-4); EPA/240/B-06/001; EPA: Washington, DC, 2006.
49. Matzke, B.D., J.E. Wilson, L.L. Newburn, S.T. Dowson, J.E. Hathaway, L.H. Sego, L.M.
Bramer, and B.A. Pulsipher. Visual Sample Plan Version 7.0 User's Guide; PNNL-23211; P.N.N.
Laboratory: Richland, WA, 2014.
50. Amidan, B.G., A. Venzin, and L. Bramer. Multiple Lines of Evidence; PNNL-24245; P.N.N.
Laboratory: Richland, WA, 2015.
51. U.S. Environmental Protection Agency, Validation of U.S. Environmental Protection Agency
Environmental Sampling Techniques that Support the Detection and Recovery of
Microorganisms; FEM Document Number 2012-01; EPA: Washington, DC, 2012.
51
-------
APPENDIX A. Literature Search Source Criteria and Keywords
Criteria for Sources of Information
¦ Information should come from sources that are considered recognized, reputable, and credible
¦ Information sources may include nationally and internationally recognized scientific, technical, or
response organizations
¦ Information can come from written text, publications, reports, subject-matter experts, and internet
sites
¦ Information sources can include (but are not limited to):
o Peer-reviewed journals, scientific manuals, and other scientific publications
o Federal, state, and local agency web sites or publications
o University web sites or publications
o Professional societies and organizations web sites or publications
o Recognized international scientific/environmental organizations
o International government web sites and publications
o Military web sites and publications
o Industry providers of equipment and materials (i.e., vendors)
o Conference proceedings.
Literature Search Strategy and Keywords
("Environmental Monitoring" OR "Environmental Microbiology" OR "Sample collection" OR "Sample
recovery" OR "Sample processing" OR Sampling OR "Sampling techniques" OR "Sampling procedure"
OR "Sample preparation" OR ("Environmental sample" OR "Environmental sampling") OR ("Surface
sample" OR "Surface sampling") OR ("Particle sample" OR "Particle sampling") OR ("Gross sample"
OR "Gross sampling") OR ("Air sample" OR "Air sampling") OR ("Soil sample" OR "Soil sampling")
OR ("Composite sample" OR "Composite sampling") OR ("Grab sample" OR "Grab sampling") OR
(PCR OR "Polymerase chain reaction") OR "Biological detection" OR "Limit of detection" OR
Detector* OR ("Detection methods" OR "Detection method") OR "Field detection")
AND
(("Biological agent" OR "biological agents") OR ("Bio-agent" OR "bio-agents") OR ("Bioagent" OR
"bioagents") OR "Biological weapons" OR "Bacterial aerosol" OR Bioaerosol* OR "Biological aerosol"
OR "Spores" OR "Bacillus anthracis" OR "Bacillus anthracis Ames" OR "Bacillus anthracis Sterne" OR
"Bacillus cereus" OR "Bacillus megaterium" OR "Bacillus atrophaeus" OR "Bacillus subtilis" OR
"Bacillus thuringiensis" OR "Geobacillus stearothermophilus")
AND
("Biological incident response" OR Bioterrorism OR "Biohazard release" OR "Biological warfare" OR
"Field demonstration" OR "Field exercise" OR "Site characterization" OR "Site clearance" OR
simulation)
To search PubMed, the following MeSH search segment was added:
A-l
-------
(("Bacillus anthracis/isolation and purification" OR "Spores, Bacterial/isolation and purification" OR
"Spores/isolation and purification") AND ("Environmental Monitoring/instrumentation" OR
"Environmental Monitoring/methods"))
A-2
-------
APPENDIX B. Literature Review Scoring Criteria
Process for Conducting Reviews
After identifying and locating relevant articles and other information from appropriate sources using
predefined keywords (see Appendix A). The reviewers read, assessed, and documented pertinent articles.
To standardize this process, a Literature Assessment Questionnaire was used to document the overall
quality of literature. Each article was evaluated using a Likert scale (i.e., (1) Poor - (5) Excellent) based
on four criteria: applicability and utility, accuracy, clarity and completeness, uncertainty and variability,
soundness, and evaluation and review. Articles were scored according to the presence of these criteria.
Table 1 shows the rubric for tallying articles.
Table B-l. Rubric for Tallying Articles
Overall Rating
Description
25—30
High quality article. Article shall be recorded and summarized accordingly
19—24
Moderately high quality article. Article shall be recorded and summarized
accordingly
10—18
Lower quality article but with some useful information. Article shall be
recorded and summarized accordingly
<10
Unacceptable/Do not use
Articles that score greater than or equal to 10 are deemed at least moderately relevant and shall be
recorded and summarized accordingly; however, articles scoring less than 10 shall be discarded. For each
source deemed as at least moderately relevant, the reviewer shall conduct an article summary. EPA shall
receive a synthesized report containing the main points presented by all of the pertinent articles reviewed.
Literature Assessment Questionnaire
Relevant articles were defined as those crucial to answering research questions pertaining to current
sampling methods for wide-area biological incidents. The reviewers considered the following criteria:
¦ Applicability and Utility: The extent to which the work not only addresses the area of inquiry
under consideration but also contributes to its understanding; it is germane to the issue at hand.
The extent to which the information is relevant for the intended use
¦ Accuracy: The extent to which data are consistent with accepted knowledge in the field or, if not,
the new or varying data are explained within the work. The degree to which data fit within the
context of the literature and are intellectually honest and authentic.
¦ Clarity and Completeness: The degree of clarity and completeness with which the data,
assumptions, methods, QA, and analyses employed to generate the information are documented.
¦ Uncertainty and Variability: The extent to which variability and uncertainty (quantitative and
qualitative) related to results, procedures, measures, methods, or models are evaluated and
characterized.
B-l
-------
¦ Soundness. The extent to which the scientific and technical procedures, measures, methods, or
models employed to generate the information is reasonable for, and consistent with, the intended
application.
¦ Evaluation and Review: The extent of independent verification, validation, and peer review of
the information or of the procedures, measures, methods, or models.
B-2
-------
vvEPA
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Office of Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
------- |