Final Report on the World Trade Center (WTC) Dust
Screening Method Study
August 17, 2005
Prepared By:
U.S. Environmental Protection Agency,
Office of Research and Development (ORD),
Research Triangle Park, NC and Washington, D.C.
and
U.S. Environmental Protection Agency, Region 2
New York, New York
The use of trade names does not imply endorsement and is for illustrative purposes only.
-------
ACRONYMS:
ATSDR Agency for Toxic Substances and Disease Registry
AVG Average
COPC Contaminant of Potential Concern
EPA U.S. Environmental Protection Agency
EPIC Environmental Photographic Interpretation Center
ERT U.S. EPA's Emergency Response Team
HEPA High Efficiency Particulate Air
LI Long Island
MMVF Man Made Vitreous Fibers
MQO Measurement Quality Objective
ND Non-Detect
NEIC U.S. EPA's National Enforcement Investigations Center
NERL U. S. EPA's National Exposure Research Laboratory
NJ New Jersey
NYCDOMH New York City Department of Health and Mental Hygiene
ORD U. S. EPA's Office of Research and Development
PM Particulate Matter
PM2.5 Particulate Matter Smaller than 2.5 microns
QAPP Quality Assurance Project Plan
SD Standard Deviation
SEM Scanning Electron Microscopy
USGS U.S. Geological Survey
WTC World Trade Center
-------
TABLE OF CONTENTS:
EXECUTIVE SUMMARY 4
I. INTRODUCTION AND BACKGROUND 5
II. METHOD DEVELOPMENT 6
III. METHOD VALIDATION STUDY 9
IV. RESULTS AND DISCUSSION 12
V. CONCLUSIONS 21
VI. REFERENCES 22
VII. CONTRIBUTORS 22
VIII. ACKNOWLEDGEMENTS 23
IX. APPENDICES 24
A: Quality Assurance Project Plan (QAPP) for the World Trade Center Screening
Method Study (Under Separate Cover)
B: Data gathered by U.S. EPA NERL during method development
C: Data gathered by U.S. EPA NERL post method development
D: Analytical method/protocol used during study
E: Report from USEPA contractor on Screening Method Study Results (including SEM
calibration data)
F: Statistical analysis and interpretation of test results
TABLES:
TABLE 1: Average, Standard Deviation and Range of Results 14
FIGURES:
FIGURE la: ORD-Modeled WTC Plume Dispersion 7
FIGURE Ib: EPIC Analysis of Deposition Boundaries 8
FIGURE 2: USGS Spiking Material Results 11
FIGURE 3: 4 Albany Spiking Material Results 12
FIGURE 4: Average Slag Wool in background and spiked samples 15
FIGURE 5: Average Slag Wool in background, spiked and impacted samples 16
FIGURE 6: Average of Elements of Concrete in background and spiked samples 17
FIGURE 7: Average of Gypsum in background and spiked samples 18
FIGURE 8: Map of the origin of samples analyzed 19
-------
EXECUTIVE SUMMARY
The September 11, 2001 attack on the World Trade Center (WTC) covered a large area with dust
and debris. To assist in determining if residual contamination exists in the indoor environment,
the U.S. Environmental Protection Agency (EPA) initiated a study to sample indoor
environments that may have been impacted by the WTC collapse. A critical component of this
study is determining whether sampled dust originated from the collapse of the WTC or instead is
urban dust originating from other sources. This report describes work performed to develop and
validate a screening method for indoor dust that can be used to determine whether dust sampled
is from the collapse of the World Trade Center towers.
Dispersion models, monitoring, photos, interviews, and satellite data were reviewed to discern
areas that were likely impacted by WTC emissions and those that were not (US EPA 2002;
2004). A total of 117 samples were collected from both impacted and non-impacted areas. A
subset of these samples were analyzed by EPA's National Exposure Research Laboratory
(NERL) and National Enforcement Investigations Center (NEIC), and United States Geological
Survey (USGS) to evaluate the slag wool levels in the dust and develop an analytical method.
The analytical method that was developed screens for three materials that are believed to be
present in large quantities in WTC dusts: slag wool, elements of concrete, and gypsum. This
method involves the use of Scanning Electron Microscopy (SEM) to determine the quantity of
each of the materials present.
Five commercial laboratories, along with the three above listed government labs, were recruited
to test the screening method. Thirty-two dust samples, consisting of both confirmed background
samples and a confirmed background dust spiked with varying amounts of confirmed WTC dust,
were sent out to the eight labs. The labs were provided the samples "blind". They did not know
which samples were background dust and which were non-impacted dust spiked with WTC dust.
In addition to the thirty-two samples, one of the five commercial laboratories also received
twenty-eight background samples to increase the available data characterizing background
locations.
The data reported by these laboratories indicated the following:
1) Five of the eight laboratories were able to reasonably measure the slag wool
concentrations in non-impacted dust spiked with confirmed WTC dust.
2) A substantial amount of variability in slag wool measurements was found within labs
and between labs. Despite this variability, slag wool measurements appear to be sensitive
enough to distinguish WTC dust (defined as 4 Albany) spiked at the 10% level from
background dust.
3) The levels of gypsum and elements of concrete in the spiked samples were
indistinguishable from the levels in the background samples. This suggests that, while
these components may have been elevated in dust samples collected near the WTC site in
September 2001 (as found by USGS in their studies on WTC dust), they are also
commonly found in the indoor environment and would not be useful as WTC signature
-------
components.
4) Analysis of samples during method development showed elevated levels of slag wool
in samples from several impacted locations compared to slag wool levels measured at
background locations.
I. INTRODUCTION AND BACKGROUND
The objective of this effort was to develop and validate a means of determining whether dust
sampled as part of EPA's planned sampling program contains residual contamination attributable
to the collapse of the WTC towers. The tested screening method is a critical component of the
sampling program as it will be used for two primary purposes: 1) to determine the geographic
extent of the dust remaining from the collapse impact, and 2) along with the results from
contaminants of potential concern (COPC) testing, to determine the need for a clean-up of the
sampled areas.
The USGS has published two reports that provided the basis for the initial hypothesis that a
WTC collapse signature is comprised of three marker components: slag wool, gypsum and
elements of concrete. The first report discusses the analysis and interpretation of indoor and
outdoor WTC dust samples collected near Ground Zero, days and weeks after September 11,
2001 (Meeker et al., 2005). From this work, we see that the WTC dust samples are dominated
by gypsum, concrete, and man-made vitreous fibers (MMVF), mainly slag wool. It is on the
basis of these key results that gypsum, elements of concrete, and slag wool were identified as
candidates for a WTC signature. The second report discusses the analysis of EPA supplied
samples taken from several indoor locations well outside of the WTC impacted area
(background). These samples were taken between September of 2004 and April of 2005. Slag
wool was absent from many of these background samples, but Lowers et al. (2005a) state that the
samples do have gypsum present, which they speculate might be due to the presence of wall
board in the sampled apartments. Because of the lack of slag wool in these samples, USGS
concluded that these samples did not contain WTC dust. USGS also concluded that perhaps slag
wool is the single most critical of the three WTC dust constituents when distinguishing WTC
dust from other common dusts.
Other studies also identified MMVF and gypsum as predominant components of WTC dust. In a
study of air and settled dust quality in apartments in Lower Manhattan, the Agency for Toxic
Substances and Disease Registry (ATSDR) and the New York City Department of Health and
Mental Hygiene (NYCDOMH) found significantly more MMVF and gypsum in samples taken
from Lower Manhattan apartments as compared to samples taken from apartments in areas above
59th Street (NYCDOMH/AT SDR, 2002). They also concluded that gypsum was seen at a higher
percentage level in the Lower Manhattan dust samples as compared to the comparison area
samples. In a comprehensive study of the composition of settled dust in the Deutsche Bank
building at 130 Liberty Street, RJ. Lee identified numerous hazardous contaminants that were
present in the dust at levels much higher than in background office buildings, and among those
substances identified in their "WTC signature" were mineral wool and gypsum (RJ. Lee, 2004).
-------
If the WTC building collapse signature components of slag wool, gypsum, and elements of
concrete are not present, then one could conclude that WTC building collapse dust is not present.
However, since these components might be present in typical New York City dust, and as slag
wool is a component of insulating materials in currently constructed buildings, it is possible that
a test might show them to be present even though WTC dust never impacted the sampled area.
A 'screening test' will, by its design, result in some fraction of such false positives (i.e. a
location without residual WTC dust that tests positive for the above components). However, an
appropriate 'screening test' would result in very few, if any, false negatives (i.e. a location with
residual WTC dust that tests negative for the above components).
II. METHOD DEVELOPMENT
Sample Collection
EPA acquired 117 dust samples during the time period of September 2004 to April 2005.
Twenty-one 'impacted' samples were taken by the EPA at two buildings that were part of the
Deutsche Bank complex located at 130 Liberty Street and 4 Albany Street. Both affected
buildings were uninhabited and slated for demolition. Fifty samples were taken from locations
well beyond the impacted zone (based on modeling, monitoring and photo analysis; these
samples are considered to be 'background' dust). Forty-six samples were taken from locations
that were possibly impacted, but were a bit farther from the WTC site than the known 'impacted'
samples. None of these forty-six samples were used in the method validation study, but several
were evaluated during both the method/protocol development phase and post-study. In addition,
one impacted sample was obtained from the USGS. This sample was a composite sample of
outdoor and indoor WTC dust collected in September of 2001.
A standard method utilizing a High Efficiency Particulate Air (HEPA) vacuum collector was
used by EPA to collect most bulk dust samples. Information on this method is provided in the
Quality Assurance Plan (QAPP) for this study (Appendix A). Some bulk dust samples were
collected from residential and commercial vacuum cleaner bags.
Modeling and satellite photography were used to determine sampling locations for the collection
of the 117 samples. Figures la and Ib (EPA 2002; EPIC 2004) are examples of modeling and
photographic analysis used to distinguish non-impacted or background locations. Figure la
shows ORD-modeled WTC Plume Dispersion on September 11, 2001 at 12 noon. The values
indicated by red are hourly PM2.5 concentrations (in |ig/m3) measured at pre-existing NJ and NY
State-operated PM monitoring stations in northern New Jersey and New York City. Red, orange,
and yellow shading represent most likely areas of plume dispersion (red = estimated dilution to
100th to 500th and dark blue = dilution to < one millionth of pollutant concentration at WTC
source). As seen in this figure, the plume very rapidly diluted to concentrations less than 1/1000
(which is the yellow area) of the initial source strength at Ground Zero. Figure Ib shows the
boundaries of collapse deposition debris as determined by aerial photographs. This photograph
was taken on September 13, and shows the four areas of "confirmed", "probable", "possible",
and "no dust" from the collapse. These areas were used in the determination of strata used in the
design for the overall sampling program.
-------
CALMET Surface Wind Field and CALPUFF Plume Dilution
(relative ta volume laurac at reoaveiy site>
v \ ^ \ V \. \ \ X - \
Figure la: ORD-modeled WTC Plume Dispersion on September 11, 2001 at 12 noon.
(Source: Exposure and Human Health Evaluation of Airborne Pollution from the World
Trade Center Disaster (External Review Draft). U.S. Environmental Protection Agency,
Washington, D.C., 2002.)
-------
This 15 a DRAFT rtorumnnl
and has not been -jupmvuu
fur publicalluti by Sic USEPA
II is for internal agency u
UirjiLuiiun only.
Mapping Results from September
13, 2001 aerial photographs
Confirmed Dust/Debris
Probable Dust/Debris
Possible Dust/Debris
Vehicle tracks and
possible dust
Excavation Area
Mounded Material
N
Figure 12. September 13, 2001. Image
mosaic of lower Manhattan and portions
of Brooklyn Points in black represent
areas where vehicle tracks and possible
dust vjere observed along wharf areas in
Brooklyn-
Figure lb: Display of boundaries of expected deposition based on analysis conducted by
EPA's Environmental Photographic Interpretation Center (Updated by EPIC from the
figure which appears in EPIC, 2004).
-------
Preliminary Analysis of Collected Samples for Slag Wool
Most of the collected samples were analyzed for slag wool content by the EP A's National
Exposure Research Laboratory (NERL) Scanning Electron Microscopy (SEM) Laboratory. This
analysis was performed as part of the EPA's development of a protocol for sample preparation
and analysis and for preliminary sample characterization. These samples were not analyzed for
elements of concrete or gypsum as an analytical method for these components had not yet been
developed. The data acquired during this method/protocol development effort are presented in
Appendix B. Caution should be used with these data as it was obtained while the method was
being developed. Post-study data acquired by NERL are also presented in Appendix C.
In evaluating the method development data acquired by NERL (Appendix B), there appears to be
a distinction between samples taken in impacted areas versus background samples. Eighteen of
the 21 samples from impacted areas had slag wool at concentrations of greater than 100,000 slag
wool fibers per gram of dust, with a range of 69,000 to 13,400,000, while all of the samples from
background areas had concentrations less than 100,000 fibers/gram, ranging from no slag wool
detected (in 12 of 47 samples) to 92,800 fibers/gram of dust.
Based on this preliminary work, the USGS, the EPA's Office of Research and Development
(ORD), the EPA's National Enforcement Investigations Center (NEIC), and experts five
commercial testing laboratories (denoted labs A-H in Appendix E), worked together to develop
an analytical method to identify the presence and concentration of the screening constituents (i.e.
slag wool, gypsum and elements of concrete) in indoor dust. This method was reviewed by the
WTC Expert Technical Panel's signature subcommittee and is presented in Appendix D. The
composition of this technical panel can be found at http://www.epa.gov/wtc/panel.
III. METHOD VALIDATION STUDY
Study Design
The basis for the WTC dust screening method discussed above is as follows: if a unit has been
impacted, those materials that are found in WTC dust will be found in the dust collected from the
unit. The materials under consideration are: 1) slag wool, 2) elements consistent with concrete
and 3) gypsum. The study described herein was intended to validate the WTC dust screening
method by demonstrating the following things:
1) that the above described materials are reasonable markers for WTC dust (by showing that
these markers distinguish WTC-laden dust from background dust);
2) that WTC dust at a diluted concentration can be distinguished from background; and
3) that the analytical method works well enough and is able to be carried out by enough
analytical laboratories to: 1) evaluate the above materials as markers and 2) distinguish
WTC dust from background dust.
The first of these three objectives was partially addressed in method development work, which
focused on slag wool. As described in the previous section, slag wool was found to be elevated
in locations deemed "impacted", while slag wool was not detected or detected at low
concentrations in "background" areas.
-------
Five independent laboratories and three government laboratories participated in this method
validation phase. One government laboratory analyzed only a small portion of the samples, but
this lab was critical in the method development. Each laboratory attended a two day session
during which the method was further developed and discussed, and the protocol was adapted to
suit each laboratory's equipment.
Following this session, the laboratories received dust samples consisting of both confirmed
background samples (10 samples plus duplicates for a total of 20) and confirmed non-impacted
dust spiked with varying amounts of confirmed WTC dust (6 spiked samples plus duplicates for
a total of 12). Specifically, a sample that was characterized and confirmed as non-impacted
(designated in Appendix B as NE Queens maid service) was split, and the splits were spiked at
levels of 1, 5, and 10% total mass with two different characterized and confirmed WTC dusts.
These spiked samples were then homogenized as documented in the QAPP for this study
(Appendix A). The two spiking dusts were 1) a composite sample of predominantly outdoor dust
collected in September of 2001 by USGS, and 2) dust collected by the U.S. EPA from the
Deutsche Bank building at 4 Albany Street in September of 2004. The 4 Albany Street building
borders the south side of the WTC complex. Six spiked samples were prepared for each
laboratory; these were split so that each laboratory received 12 spiked samples. Each laboratory
also received 10 non-impacted background samples that were also split, resulting in a total of 20
background samples. Thirty-two samples in all were sent for analysis to the eight labs.
In addition to the 32 samples, one of the five commercial laboratories also received 28
background samples to increase the available data characterizing background locations.
The labs were provided the 32 samples "blind"; they did not know which samples were pure
background dust, and which were the spiked dust. To ensure sufficient results for spiked
samples, the government laboratory that was only able to analyze a small portion of the samples
was asked to analyze only the 12 spiked samples. Again, they were not told the identity of these
samples (Lab C). The labs had five weeks to analyze all samples. The final data from all
laboratories, including the data for the additional 28 background samples, were reviewed,
evaluated and analyzed by the EPA and the EPA's prime contractor. This prime contractor's
from this analysis is presented in Appendix E.
Composition of Spiked Samples
The USGS performed an analysis of the spiked, homogenized samples prior to the samples being
sent to the labs. The measured levels were in the approximate range for the spiking percent (1, 5,
and 10%) based on the undiluted concentration level of each WTC dust and, in all but one case,
each percent level was fully distinguishable from the others (Figures 2 and 3). The variability in
the measured levels was expected due to the difficulty in homogenizing dusts that have large
particle size distributions, and the fact that components of WTC dust will vary within a sample
because of the nature of the source. Given these difficulties and the measurement results, these
dusts were determined to be reasonably homogeneous.
As seen in Figures 2 and 3, the level of slag wool differs between the two WTC dusts, with the
pure dust that was collected from 4 Albany Street in 2004 more than an order of magnitude
lower than the dust collected by the USGS in September of 2001. The pure dust from 4 Albany
10
-------
Street had slag wool levels at 500,000 fibers/gram of dust versus approximately 11,000,000
fibers/gram of dust for the USGS collected sample. There are likely explanations for this large
difference in slag wool levels. The USGS sample was a composite of multiple outdoor samples
and one indoor sample taken during September of 2001. The 4 Albany was an indoor sample
was taken three years post 9/11 in September of 2004. As this 4 Albany sample was taken
exclusively inside of a building, it was not only diluted by three years accumulation of urban
background dust, but was also characteristic of dust that had penetrated the shell of a building as
opposed to that deposited on the ground outside.
USGS Spiking Material
3
Q
5
o
1,400,000
1,200,000
1,000,000
800,000
600,000
400,000
200,000
10
% Spiking
Figure 2: USGS Spiking Material Results. Analysis was conducted by USGS prior to being
sent to labs for study. Pure dust averaged approx. 11,000,000 fibers/gram.
(Figure provided by USGS)
11
-------
4 Albany Street Spiking Material
(0
3
Q
•s
(0
a)
13
70,000
50,000
40,000
30,000
20,000
I
% Spiking
Figure 3: 4 Albany Street Spiking Material Results. Analysis was conducted by USGS
prior to being sent to labs. Pure dust averaged approx. 500,000 fibers/gram.
(Figure provided by USGS)
IV. RESULTS AND DISCUSSION
Development of Study Results
The final report from the prime contractor with all raw analytical and calibration data can be
found in Appendix E. A summary of the study results that includes the data from the 28
additional background samples analyzed by a single commercial laboratory is provided in Table
I, as well as Figures 4-7. A map of the origin of the samples analyzed during this study is shown
in Figure 8.
All background sample data used in Table I and Figures 4-7 are from the Greater NY City area.
Background samples taken in Research Triangle Park, North Carolina are not included as they
are not representative of NY City background dust. Data for all background sample results may
be found in Tables 3 and 4 of the Versar report in Appendix E. It should be noted that the
Research Triangle Park samples show higher slag wool levels than NY City area background
samples. This is due to the presence of slag wool containing ceiling tiles in the building
sampled. Note also that Table I indicates two average values for background slag wool. These
values reflect the inclusion and exclusion of two samples collected in New Jersey (NJ) and Long
Island (LI) that were extremely high in slag wool fibers, likely due to their insulation,
fireproofing or ceiling tiles. Based on these results it is likely that some false positive results will
occur in buildings with slag wool-based ceiling tiles, fireproofing or insulation. .
12
-------
Three of the commercial laboratories, designated as labs E, F and G, reported analytical data that
are not consistent with other five labs. Generally, these labs were not able to distinguish
differences between the three spiking levels. In addition, these labs did not meet the
measurement quality objectives (MQOs) for the spiked samples put forth in the QAPP for this
study (Appendix A Section A.7.1). Thus, the data from these three labs are not considered in the
results presented in Table I and Figures 4-7. The statistical analysis performed to make this
determination is presented in Appendix F. In addition, Lab H was not considered when
determining concrete and gypsum levels as their data were at least two times higher than the
sample average without these data (Table I and Figures 6 and 7).
In discussions with the commercial laboratories, it was determined that some labs did not have
the personnel or the equipment to perform the required analysis in the given timeframe, thus,
data quality became an issue. Additionally, labs that had less experience with slag wool analysis
felt that a clearer definition, in addition to that provided in the catalog developed by USGS in
Lowers et al., 2005b, of slag wool was needed to distinguish it from other mineral wools.
Finally, labs that were unable to automate the gypsum and concrete analysis expressed their
belief that the method was too long and complicated for accurate quantitative dust analysis. All
laboratory comments will be taken into consideration in when finalizing the protocol.
13
-------
Slag Wool
Average
(fibers/g
dust)
Elements of
Concrete
(% Area)
Gypsum
(% Area)
Background
(Greater NY Area)
AVG + SD
35,950 + 74,300
17,740+15,835*
Range of Samples
ND* - 369,230
ND* -60,000**
AVG + SD
15.6 + 5.7
Range of Samples
6-30.5
AVG±SD
9.5 + 3.4
Range of Samples
4-16.5
USGS Spiked (Collected
9/01)
1%
94,000 + 25,740
5%
452,510+100,640
70%
870,280 + 310,420
1%
20 + 6
5%
19 + 7
70%
16 + 2
7%
9 + 6
5%
7 + 3
70%
6 + 0.5
4 Albany Spiked
(Collected 9/04)
7%
17,270 + 7,880
5%
52,510 + 26,140
70%
88,540 + 18,300
7%
15 + 1
5%
18 + 4
70%
16 + 3
7%
9 + 4
5%
5+2
70%
7 + 2
• * *ND=Non Detect (Zero slag wool fibers)
• *Two extremely high values from NJ and LI removed.
Table 1: Avg, Standard Dev., and Range of Results for Background and Spiked Samples
(Data Summarized from Tables 1, 2, 3 and 4 of Appendix E).
14
-------
1000000
100000 -
(A
D
T3
E
TO
5
In
10000 -
O)
TO
1000
100
High Slag Wool
Values in NJ and LI
I
$
Background
Spiked Samples
Figure 4: Average Slag Wool (Fibers/Gram of Dust) in background and spiked samples.
(Data from Tables 3 and 4 Appendix E)
15
-------
10000000
1000000
I/)
3
•o
E
2
e>
"J2
0)
.a
iZ
"o
o
o>
(0
100000
10000
1000
100
* Background
A 4 Albany 1%
• 4 Albany 5%
• 4 Albany 10%
AUSGS1%
• USGS 5%
• USGS10%
Impacted
High Slag Wool
Values in NJ and LI
Background Impacted Spiked Samples
Figure 5: Average Slag Wool (Fibers/Gram of Dust) in background, spiked and impacted
samples. Impacted samples are locations that are shown in satellite pictures to have been
affected by WTC Collapse Dust. Slag wool results for impacted samples were derived
during method development and were not part of this method validation; they are provided
for comparative purposes. These impacted samples range from 0.1 to 1.6 miles from the
WTC site (see Figure 8 for sample origin location). Data from Appendix B (Impacted) and
Tables 3 and 4 of Appendix E (Background and Spiked).
16
-------
35
30
25 -
20
15
10 -
5
I
T
Background
Spiked Samples
Figure 6: Average of Elements of Concrete (% Area) in background and spiked samples.
(Data from Tables 1 of Appendix E)
17
-------
18
14
12
10
JL
0
o
I
t
Background
Spiked Samples
Figure 7: Average of Gypsum (% Area) in background and spiked samples.
(Data from Tables 2 of Appendix E)
18
-------
DESIGNATION
1,2
3,4
5,6
7
S
9
10
11,HS3(1),HSC2)
14,15
16
17J*V(1),MWC2)
16,1 9,20,21 ,22,23,C1-RTP1 +2
24,25,26
APS(1),AP5(2)
CMC(1),CMC(2)
WGS(1),WGS(2)
USC(1),USC(2)
MUNYC1 (1 )+(2) ,MUNYC2(1 >(2)
12,13
28
27
FP(1XFP(2)
LOCATION
Stony Brook, LI
West End Ave between 72nd and 73rd Streets
30th Ave between 21 st and 23rd, Queens
70th Street between 20th and 21 st Ave, Brooklyn
79th St between York and East End Ave, Manhattan
92nd Street between Columbus and CPW, Manhattan
Columbia Medical Center, W. 1 68th St, Manhattan
Teaneck, NJ
88th Street between Amsterdam and Columbia, Manhat
80th Street between Riverside and East End Ave, Ma
West End Ave between 1 05th and 1 06th Streets
Research Triangle Park, NC
Edison, NJ
Chittenden Avenue, Manhattan
Columbia Medical Center, W. 168th St, Manhattan
Nassau County, LI
Federal Courthouse, White Plains, NY
Northern Manhattan, Above 70th Street
Nassau County
NE Queens
Long Beach Island, NJ
Federal Courthouse, Central Islip, NY
Figure 8: Map of the origin of the samples analyzed during this study
(Reference Appendix D for sampling data).
19
-------
Discussion
Slag wool appears to be an indicator for WTC dust and can be distinguished from background
dust at all three spiking levels for the USGS dust and at the 10% level of the 4 Albany Street
dust. The 4 Albany Street dust is considered to be WTC impacted dust but as noted earlier, the 4
Albany dust likely had lower levels of slag wool due to the fact that it was an indoor dust that
was not sampled until three years after the WTC collapse.
Levels of gypsum and elements of concrete have no discernable relationship to the level of WTC
dust. There does not appear to be a distinguishable difference between levels of concrete and
gypsum in background dust and the samples spiked with WTC dust, despite USGS analysis of
WTC dust from 2001 (Meeker, 2005) showing elevated levels of these components. This is
likely due to the fact that while these components may seem high in WTC dust, they are also
high in general background dust as they are common building materials.
While method development (Appendix B and summarized in Section II above) work showed that
dusts from known impacted locations generally had slag wool levels above 100,000 fibers/gram,
several samples taken within this impacted zone and analyzed during method development
showed lower levels of slag wool. Two likely explanations can be offered for these results.
First, as the data in Appendix B was acquired during method development, it must be viewed as
such, and second, multiple cleanings of the inhabited areas since September 11, 2001 may have
removed residual WTC collapse contamination. The majority of these samples were taken in
fully inhabited buildings, from locations within the buildings that can be characterized as either
'accessible' or 'infrequently accessed' areas. These terms are described in the final draft EPA
sampling program, and they denote areas that are accessed by people over the course of time,
such as counter tops or rugs (accessible) or underneath furniture (infrequently accessed). For
this reason alone, it is encouraging that a substantial amount of the dust sampled in late 2004 and
beyond had high levels of slag wool.
While there was ample evidence of higher levels of slag wool associated with the WTC dust and
lower levels associated with background, there is high variability in slag wool measurements
within and between labs. Estimates of within lab relative standard deviations based on analysis
of duplicate samples of the 4 Albany Street data are 55%, 24% and 14% for the 1%, 5% and 10%
dilution levels, respectively. Estimates of between lab relative standard deviations based on the
4 Albany Street data are 64%, 70% and 29% for the 1%, 5% and 10% dilution levels,
respectively (looking at results from analysis of the same spike level samples by multiple labs).
Causes of the high levels of variability may include:
• Procedures to homogenize the spiked samples did not result in complete
mixing and distribution of fibers; they instead resulted in a 'reasonably'
homogeneous sample given the large size variation of the dust components.
• Components of both non-impacted/background and WTC dusts will vary
within a sample because of the inherent nature of the dust samples. Thus, the
samples received by the labs may vary in content.
• Operator experience with the target components appeared to be an issue -
post-study discussion indicated that labs representatives with less familiarity
with slag wool expressed a belief that further guidance as to its definition was
needed.
20
-------
• The variability in the mass of dust used for the analysis, as the protocol allows
for a range, not a specific mass, to be used. This range is essential due to the
extreme differences in slag wool levels possible between background and
spiked samples.
Finally, it is noted that Table I indicates two average values for background slag wool. These
values reflect the inclusion and exclusion of two samples (and their duplicates) collected in New
Jersey (NJ) and Long Island (LI) that were extremely high in slag wool fibers, likely due to their
insulation, fireproofmg or ceiling tiles. Similarly, it was earlier noted that samples taken from a
North Carolina building due also to slag wool used in ceiling tiles were not included in the
interpretative analyses. Based on these results, it is likely that some false positive results will
occur in buildings with slag wool-based ceiling tiles, fireproofmg or insulation.
V. CONCLUSIONS
The interlaboratory results indicate that the better performing labs are capable of distinguishing
the difference between 1, 5 and 10% 4 Albany Street dust. Also, despite the high levels of within
sample and within lab variability, the method using slag wool appears to be sensitive enough to
distinguish 10% 4 Albany Street dust from background dust. Additional evaluation of the data
will be performed to further understand the variability. Measures will be taken (i.e. standards
will be sent regularly to each lab) during EPA's planned sampling program to evaluate the
accuracy and precision of the laboratories.
In summary, the data developed in this study support the following findings:
1) Five of the eight laboratories were able to reasonably measure the slag wool
concentrations in background dust spiked with confirmed WTC dust.
2) High levels of variability in slag wool measurements, both within labs and between
labs, were observed in the data. Despite this variability, the slag wool method appears to
be sensitive enough to distinguish WTC dust from background dust at the 10% level
(defined as, 4 Albany Street).
3) The levels of gypsum and elements of concrete in the spiked samples were
indistinguishable from the levels in the background samples. This observation suggests
that, while these components may have been elevated in dust samples collected near
September 2001, as found by USGS in their studies on WTC dust, they are also
commonly found in the indoor environment and would not be useful as WTC signature
components.
4) Analysis of samples during method development generally showed slag wool levels in
samples from impacted locations to be greater than slag wool levels in samples from
background locations.
21
-------
VI. REFERENCES:
Lowers, H.A., G.P. Meeker, and I.K Brownfield. (2005a) Analysis of Background Residential
Dust for World Trade Center Signature Components Using Scanning Electron Microscopy and
X-ray Microanalysis. U.S. Geological Survey Open File Report 2005-1073.
http://pubs.usgs.gov/of/2005/1073/
Lowers, H.A., Meeker, G.P., I.K.Brownfield. (2005b) World Trade Center Dust Particle Atlas:
U.S. Geological Survey Open-File Report 2005-1165. http://pubs.usgs.gov/of/2005/1165/.
Meeker, G.P., A.M. Bern, H.A. Lowers, and I.K. Brownfield. (2005) Determination of a
Diagnostic Signature for World Trade Center Dust using Scanning Electron Microscopy Point
Counting Techniques. .URL: http://pubs.usgs.gov/of/2005/1031/
U.S. Geological Survey Open File Report 2005-1031.
NYCDOHMH/ATSDR. (2002) New York Department of Health and Mental Hygiene and
Agency for Toxic Substances and Disease Registry. Final Technical Report of the Public Health
Investigation To Assess Potential Exposures to Airborne and Settled Surface Dust in Residential
Areas of Lower Manhattan. URL: http://www.epa.gov/wtc/panel/ATSDRFinal-report-
lowermanhattan-02.pdf Agency for Toxic Substances and Disease Registry, US Department of
Health and Human Services, Atlanta, GA.
RJ. Lee (2004). Signature Assessment 130 Liberty Street Property Expert Report WTC Dust
Signature. Prepared for: Deutsche Bank. May, 2004. RJ. Lee Group, Inc. 350 Hochberg Road,
Monroeville, PA. 15146.
US EPA (2004) Mapping the Spatial Extent of Ground Dust and Debris from the Collapse of the
World Trade Center Buildings, DRAFT in peer review, EPA/600/X-03/018, URL:
http://www.epa.gov/wtc/panel/backdocs.html
United States Environmental Protection Agency, Office of Research and Development,
Washington, D.C., July 2004, 35pp.
US EPA (2002) Exposure and Human Health Evaluation of Airborne Pollution from the World
Trade Center Disaster, Draft in peer review, EPA/ URL:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=54667 United States Environmental
Protection Agency, Office of Research and Development, Washington, D.C., October 2002
VII. CONTRIBUTORS TO THIS STUDY:
Principal Investigators:
Jacky Ann Rosati U.S. EPA, ORD
David Friedman U.S. EPA, ORD
Contributors (in alphabetical order):
Nancy Adams U.S. EPA, ORD
Gina L. Andrews U.S. EPA, ORD
Lara Autry U.S. EPA, ORD
22
-------
Amy Bern U.S. EPA, NEIC
Isabelle Brownfield USGS
Teri Conner U.S. EPA, ORD
Evan Englund U.S. EPA, ORD
Pat Evangelista U.S. EPA, Region 2
Henry Kahn U.S. EPA, ORD
Matthew Lorber U. S. EPA, ORD
Heather Lowers USGS
Mark Maddaloni U.S. EPA, Region 2
Lisa Matthews U. S. EPA, ORD
Greg Meeker USGS
Tim Oppelt U.S. EPA, ORD
Joachim Pleil U.S. EPA, ORD
Dennis Santella U.S. EPA, Region 2
Raj Singhvi U.S. EPA, Region 2, ERT
Stanley Stephanson U.S. EPA, Region 2
Shirley Wasson U.S. EPA, ORD
Steve Wilson USGS
Supporting Contractors:
Prime Contractors
Alion Scientific
Lockheed-Martin
Versar
Subcontractors
RJ Lee Group, Inc.
EMSL Analytical Inc.
MVA Scientific Consultants
Reservoir Environmental
MAS, Inc.
Other Contributors:
John Holland SEE U.S. EPA, ORD
VIII. ACKNOWLEDGEMENTS:
The U.S. EPA would like to acknowledge the U.S. General Services Administration (GSA),
NY/NJ Port Authority, the U.S. National Park Service, Deutsche Bank, and Columbia University
(K. Crowley) for allowing samples to be collected at their facilities.
In addition, the EPA would like to acknowledge the residents of NY and NJ who allowed us to
sample in their homes, and the maid services that collected vacuum cleaner bags for use in this
study. Finally, the EPA would like to thank the Sierra Club for helping to recruit sampling
locations.
23
-------
IX. APPENDICES
APPENDIX A: QUALITY ASSURANCE PROJECT PLAN FOR THE WORLD TRADE
CENTER (WTC) SCREENING METHOD STUDY
(Due to formatting - this document will be provided under separate cover)
24
-------
APPENDIX B: DATA ACQUIRED BY EPA NERL DURING METHOD
DEVELOPMENT
Samples Collected at Background Locations
Residential
West End Ave between 72nd and 73rd Streets, Manhattan
30th Avenue between 21st and 23rd St, Queens
E 79 Street between York and East End Ave, Manhattan
Chittenden Avenue, Manhattan
92nd Street between Columbus and CPW, Manhattan
80th Street between Riverside and West End Ave, Manhattan
Edison, NJ
Stony Brook, LI
70th Street between 20th and 21st Ave, Brooklyn
Teaneck NJ
Long Beach Island, NJ
West End Avenue between 105th and 106th Streets, Manhattan
Edison, NJ
88th Street between Amsterdam and Columbia, Manhattan
North East Queens (Maid Service)
Nassau County, Long Island (Maid Service)
Business
Port Authority Bldg, Port of Newark, NJ
slag wool fibers/
gram of dust
2.53E+04
5.47E+04
2.80E+04
2.26E+04
4.93E+04
1.53E+04
2.87E+04
2.42E+03
1.46E+04
O.OOE+00
1.79E+04
2.90E+04
4.09E+04
4.77E+04
O.OOE+00
O.OOE+00
1.77E+04
4.12E+03
8.35E+03
O.OOE+00
5.74E+03
O.OOE+00
O.OOE+00
O.OOE+00
5.37E+03
1.02E+04
1.27E+04
O.OOE+00
1.63E+04
6.43E+03
O.OOE+00
1.65E+04
O.OOE+00
O.OOE+00
1.95E+04
3.86E+04
3.45E+04
7.32E+04
5.09E+04
1.85E+04
6.60E+04
Average of Duplicates
(slag wool fibers/gram
dust)
2.20E+04
4.43E+04
25
-------
Columbia Medical Center, W168 St., Manhattan
Edison, NJ
Federal Courthouse, Quarropas St, White Plains
Federal Courthouse, Islip, Long Island
8.58E+04
O.OOE+00
1.33E+04
9.09E+04
9.28E+04
9.00E+04
Samples Collected at Known Impacted Locations
Business
290 Broadway, Manhattan
Broadway between Maiden Lane and John Street, Manhattan
Deutsche Bank Bldg, 130 Liberty Street, Manhattan
Deutsche Bank Bldg, 4 Albany Street, Manhattan
USGS Composite Sample Collected Sept 2001
Samples Collected at Locations with Unknown Impact
Residential
John Street between Gold and Pearl, Manhattan
South End Avenue between Albany and Liberty, Manhattan
River Terrace, Manhattan
40th Street between Tunnel Exit St and 2nd Ave, Manhattan
Orange Street between Henry and Hicks, Brooklyn
24th Street between 8th and 9th Ave, Manhattan
Montague between Montague Terrace and Hicks Street, Manhattan
Houston and Mulberry Streets, Manhattan
Business
Port Authority Bldg, Columbia St, Brooklyn
6.92E+04
8.81 E+04
1.64E+05
1.95E+05
8.35E+04
1.33E+05
2.79E+05
4.71 E+06
5.77E+06
6.60E+06
1.18E+07
1.22E+07
1.13E+05
2.06E+05
2.14E+05
2.25E+05
2.28E+05
2.78E+05
6.36E+05
1.67E+06
1.34E+07
1.26E+04
9.17E+03
O.OOE+00
2.91 E+03
1.11 E+04
3.32E+03
5.03E+03
6.30E+03
2.06E+05
9.89E+04
1.30E+05
1.94E+05
1.12E+04
3.06E+05
1.20E+05
6.19E+06
2.30E+05
26
-------
Governor's Island 5.07E+04
5.75E+05
8.79E+04
Varick Street, Manhattan 9.57E+04
Samples Collected Outside of NY City
Business
Research Triangle Park, NC 5.00E+04
8.96E+04
27
-------
APPENDIX C: DATA ACQUIRED BY EPA NERL POST-STUDY
Samples Collected at Background Locations
Residential
Composite -North East Queens (Maid Service)
Business
Port Authority - Port of Newark, NJ
Samples Collected at Impacted Locations
Business
Governor's Island
Port Authority Bldg, Columbia St, Brooklyn
1.06E+04
1.49E+04
9.77E+03
1.93E+04
6.39E+05
1.21E+06
1.22E+05
1.28E+04
28
-------
APPENDIX D: PROTOCOL USED FOR THE SCREENING METHOD STUDY
Protocol for Preparation and Analysis of Residential and Office Space
Dust by Polarized Light Microscopy and Scanning Electron Microscopy with
Energy Dispersive X-Ray Spectroscopy
June 27, 2005
Prepared by:
U.S. Environmental Protection Agency
National Enforcement Investigations Center/ National Exposure Research
Laboratory/National Homeland Security Research Center
Denver, CO and Research Triangle Park, NC
The use of trade names does not imply endorsement and are used for illustrative purposes only.
29
-------
Contents
1.0 Purpose 31
2.0 Scope/Application 31
2.1 Limitations of the Method and Future Considerations 31
3.0 Definitions 31
4.0 Summary of Method 31
5.0 Interferences 32
6.0 Safety 32
7.0 Apparatus and Materials 32
8.0 Reagents 33
9.0 Sample Storage 33
10.0 Quality Control 33
10.1 Calibration 34
11.0 Procedure 34
11.1 Weighing and Splitting 34
11.2 Ashing 35
11.3 Sieving 35
11.4 Preparation of Sample for Polarized Light Microscopy 35
11.4 Mounting Sample on SEM Sample Stubs 35
12.0 Analysis 37
12.1 Analysis by Polarized Light Microscopy 37
12.2 Analysis by SEM/EDS 35
12.2.1 Screening for Slag Wool 35
12.2.2 EDX Screening for Gypsum/Anhydrite 35
12.2.3 X-Ray Mapping for Gypsum 36
12.2.4 X-Ray Mapping for Ca-rich Particles 37
12.2.5 Particle Analysis for Gypsum and Concrete 37
13.0 Data Analysis and Calculations 38
14.0 References 39
15.0 Appendix 40.
30
-------
1.0 Purpose
This document describes sample preparation and analytical screening procedures for bulk samples of dust
collected from residential and commercial office environments. These methods are collectively referred to
as the protocol.
2.0 Scope/Application
The protocol describes polarized light microscopy (PLM) and scanning electron
microscopy (SEM) with energy dispersive spectrometry (EDS) to screen bulk dust
samples for mineral slag wool, particles consistent with concrete compositions, and
gypsum. The analysis methods include operating parameters and particle identification
criteria.
2.1 Limitations of the Method and Future Considerations
This protocol provides a means of analyzing for particles consistent with those found in dust
present after the collapse of the World Trade Center (WTC) in New York City. Components of
WTC Dust have been documented and catalogued by the U.S. Geological Survey Denver
Microbeam Facility and the images and characteristics shall be used in identification of particles
(1).
The x-ray mapping procedure in sections 12.2.3 and 12.2.4 and the calculations presented in
section 13.0 only determine the maximum percentage of non-gypsum, calcium-rich particles,
which may include non-concrete materials. The particle analysis procedure presented in section
12.2.5 is the preferred procedure for determining the percentages of gypsum and concrete particles
in the sample.
The x-ray mapping and image analysis procedure relies heavily on the thresholds for backscattered
electron images. Binary (particles white and background black) backscattered electron images
(BEI) should be used to reduce errors in setting thresholds in Photoshop
3.0 Definitions
1. PLM - Polarized Light Microscopy
2. SEM - Scanning Electron Microscope
3. EDS - Energy Dispersive Spectrometry
4. SEI - Secondary Electron Image
5. BEI - Backscattered Electron Image
6. Mineral Wool - lightweight vitreous fibrous material composed of rock wool and slag wool and used
especially for heat and sound insulation
7. Rock Wool - a man-made vitreous fiber (MMVF) component of mineral wool containing magnesium,
aluminum, silicon, and calcium. Sodium and potassium may also be present. Iron oxide is typically 3-
12% by weight.
8. Slag Wool - a man-made vitreous fiber (MMVF) component of mineral wool containing magnesium,
aluminum, silicon, and calcium. Sodium and potassium may also be present. Iron oxide is typically
less than 2% by weight.
9. HEP A - High-Efficiency-Particulate-Air Filter
4.0 Summary of Method
1. Weigh sample to nearest 0.0005 g.
31
-------
2. Split the sample, archive half and keep half for analysis.
3. Ash half of the sample for analysis.
4. Sieve the ashed sample to 150 jim.
5. Split the <150 um ashed portion. Archive three quarters of the sample. Keep one
quarter for PLM and SEM/EDS analysis.
6. Weigh the quarter and place it in enough isopropanol to get a 10-20 mg per mL
dilution. Apply an aliquot to a glass slide, let dry, and add 1.55 (or 1.605) refractive
index oil. Analyze by PLM for mineral wool.
7. Prepare a sample for SEM/EDS analysis using the same dilution prepared for PLM.
8. Apply an aliquot of the sample to an aluminum sample stub with a carbon adhesive
tab covered by a piece of polycarbonate filter (13-mm diameter or punched out of a
larger filter to fit the size of the stub).
9. Identify fibers by EDS and record the occurrence of fibers > 25 jim in length at 100 x
magnification to get a statistical representation of fiber compositions.
10. Prepare 10-fold dilution of the suspension from step 7 and apply an aliquot to a
polycarbonate/adhesive tab substrate affixed to an aluminum sample stub.
Alternatively, a lighter loading can be prepared by filtering the diluted suspension
through a 25-mm diameter, 0.4-|im pore size, polycarbonate filter and affix this to a
carbon adhesive tab affixed to an aluminum sample stub.
11. Collect x-ray maps of 10 fields at 500 x magnification for major elements, especially
Ca, S, and Fe and use Adobe Photoshop or similar software to determine the area
percent of gypsum and Ca-rich particles. Fe-rich particles may also be identified in
this step.
12. Perform particle analysis via computer-controlled SEM/EDX analysis.
5.0 Interferences
Interferences include possible contamination of samples by airborne dust or through improperly cleaned
glassware and sieves. Interferences are minimized by performing all procedures involving dry dust in a
clean room, cleaning countertops and glassware thoroughly before proceeding and placing particle-free
wipes on all working surfaces. To avoid cross-contamination, properly clean all glassware, sieves, and
tools between samples.
6.0 Safety
Respirable particles which may present a health hazard may exist in the sample. Bulk samples may release
respirable particles during handling. All procedures involving dry dust samples will be performed under a
negative flow High-Efficiency-Particulate-Air Filter (HEPA) hood. Samples handled outside of the HEPA
hood will be covered with aluminum foil or placed in sealed glass jars.
7.0 Apparatus and Materials
1. HEPA negative flow hood
2. Forceps
3. Kimwipes
4. Stainless steel spatula
5. Weighing paper
6. Programmable furnace [not required for validation study]
7. Ceramic crucibles with lids [not required for validation study]
32
-------
8. Analytical balance (accuracy to 0.0005 g)
9. Retsch ultrasonic sieve shaker (AS200 Basic), or similar [not required for validation study]
10. Sample sieves, 3-inch diameter (recommended), 150-um (100-mesh) opening, with lid and bottom pan
similar [not required for validation study]
11. SEM aluminum sample stubs
12. Conductive carbon adhesive tabs
13. Eppendorf pipette, 10-|aL capacity
14. Disposable pipette tips
15. 1 - 10 mL pipette
16. Glass vials for sonicating dust in isopropanol suspension (holds 10-mL volume)
17. Razor blade
18. Ultrasonic bath
19. 50 mL glass beaker
20. Polycarbonate filters (25-mm diameter, 0.4-um pore size)
21. Polycarbonate filters (13-mm diameter, 0.4-um pore size), or borer to cut larger filters to SEM stub
size
22. 11-mm diameter cork borer
23. Millipore filter apparatus for use with 25 mm filters
24. 125 mL Nalgene bottles
25. Hand-held vacuum pump
26. High-vacuum carbon evaporator with rotating stage
27. Glass etri dishes with lids
28. Adobe Photoshop Software, or similar
29. Glass petrographic slides
30. Glass cover slips
31. Polarized light microscope for mineral identifications
32. Scanning Electron Microscope with the following attributes:
a. Resolution: 5 nm (at 25 kV, WD= 10 mm - system dependent) or better
b. Accelerating Voltage: 10 to 20 kV
c. Minimum magnification range: 50x to 200,000x
d. SEI (secondary electron image)
e. BEI (backscattered electron image)
f. Energy dispersive x-ray detector and analyzer for EDS analysis
g. Ability to collect x-ray maps or particle analysis software (preferably both)
8.0 Reagents
1. Isopropanol, reagent grade [CAS No. 67-63-0]
2. 1.55 or 1.605 Refractive Index Oil
9.0 Sample Storage
Dust samples will be stored in an air-tight container, such as a sealed glass jar. Samples placed in reagents
will be labeled appropriately and stored according to laboratory safety standards. Samples prepared for
analyses will be stored in a protective container, such as a plastic case or covered etri dish, to prevent
contamination.
10.0 Quality Control
Quality control is implemented by thoroughly cleaning glassware and spatulas, keeping working surfaces
clean, and preventing cross contamination. During ashing, particles may be suspended if slow heating is
not achieved. Following the ashing program as outlined will minimize flashing, which can cause particles
to become airborne. Covered crucibles will be used to prevent contamination caused by flashing. Used
Eppendorf pipette tips and weighing papers will be discarded and new tips and papers will be used for each
33
-------
sample.
Duplicate samples shall be prepared to determine the precision of the analysis. In addition, sample blanks
shall be prepared. These blanks are checks for cross contamination during handling of the samples. Blanks
shall be prepared at the same time and in the same manner as samples.
10.1 Calibration
Calibration of the EDS system must be completed at least once at the beginning and again at the
end of each analytical session. Backscattered electron image (BEI) calibration should be
performed at the beginning of the session and anytime the backscattered image brightness and/or
contrast is adjusted.
EDS calibration for both qualitative and quantitative (not required by this method but could be
useful for identification of particle type) analysis is accomplished by the analysis of a polished
carbon-coated reference standard. The recommended material is USGS BIR1-G basalt glass
mounted in epoxy in a brass tube, polished, and carbon coated using a carbon evaporator (2, 3).
The calibration reference material should be analyzed at the same operating conditions to be used
for the analysis including beam current, accelerating voltage, working distance, detector dead
time, and sample tilt (= 0°). For BIR1-G the analysis should be performed with a beam size of 10-
20 |am or equivalent area raster. All calibration spectra will be saved with the corresponding data
set. The calibration data will be used for inter- as well as intra-laboratory comparisons. This
calibration is in addition to, and not a substitute for the normal EDS calibration recommended by
the EDS manufacturer which will be performed at regular intervals as specified by the EDS
manufacturer.
Backscattered electron detector calibration can be performed on the same BIR1-G material by
adjusting the detector brightness and contrast to achieve the following conditions. The epoxy on
the BIR1-G reference material will be at 0 in a 256 grayscale image and the brass mounting tube
will be at 256. The BIR1-G basalt glass should fall at approximately 130-140 gray scale units
11.0 Procedure
11.1 Weighing and Splitting
Weighing and splitting should be performed under a negative flow HEPA hood.
If the fan speed is set too high, loss of particles may occur. The fan speed may
need to be adjusted to prevent the loss of fine particles.
Obtain an analytical balance with an accuracy of 0.0005 g and preweigh a clean
piece of weighing paper. Transfer the dust from the sample vial to the weighing
paper and determine the weight of the dust. Split the sample with a clean razor
blade using the cone-and-quarter method. If there are large clumps of organic
fibers, such as hair or lint, temporarily remove the hair with a pair of forceps and
tap the forceps lightly with another tool over a piece of weighing paper to remove
fine particles. Center the fine fraction on the paper and split the sample into four
equal parts using a razor blade. Collect opposite corners (l/2 of the sample) for
analysis and archive the other half. Quarter the larger organic fiber bundles the
same way, keeping half to proceed to the ashing step and half for archival
purposes.
34
-------
Place the two quarters for ashing into a preweighed crucible. Weigh the split and
record the results.
11.2 Ashing
Place the ceramic crucibles containing the samples into a furnace.
The furnace program should proceed as follows:
1. Increase temperature by 1 °C/minute until sample reaches 250 °C.
2. Hold temperature at 250 °C for 4 hours.
3. Increase temperature by 1 °C/minute until sample reaches 480 °C.
4. Hold temperature at 480 °C (sufficient for decomposing organics) for 8 hours. Do not exceed
500 °C.
5. Shut off furnace.
6. Allow sample to cool before removing from furnace.
7. Weigh the ashed sample to the nearest 0.0005 g and record the result.
11.3 Sieving
Sieve the sample through a 150-jim sieve using a Retsch ultrasonic sieve shaker,
or similar. Three-inch diameter sieves are recommended to minimize sample loss
from particles being trapped in the sieve. The ultrasonic shaker will be operated
at 20-minute intervals at the following settings: 20, 40, 60, 70, 80, then back
down to 50 and 20. This will provide amplitudes ranging from 0 to 1.5 mm.
Transfer the large and small fractions to clean pieces of weighing paper and
weigh to the nearest 0.0005 g. Archive the fraction greater than 150-|im.
11.4 Preparation of Sample for Polarized Light Microscopy
Split the less than 150-um sample fraction using the cone and quarter method. Collect one corner
for analysis and archive the other three quarters. Weigh the quarter split to the nearest 0.0005 g
and place it into a glass vial. Make a suspension of 10-20 mg dust per mL of isopropanol. The
amount of isopropanol needed will vary depending on the amount of dust; the target dilution is 10-
20 mg per mL.
Cut an Eppendorf pipette tip with a razor blade to increase the opening to
approximately 1 mm.
Place the suspension in an ultrasonic bath for one minute, then remove the suspension from the
ultrasonic bath and shake it gently to suspend all particles. Collect a 10-|oL aliquot of the mixture
using an Eppendorf pipette with the modified tip and transfer to a glass slide. Prepare 4 such
slides. Allow them to dry, then add a drop of 1.55 (or 1.605) refractive index oil.
11.5 Preparation of Sample for SEM Analysis
Prepare the SEM substrate on aluminum stubs using 0.4-|im pore size
polycarbonate filters, carbon adhesive tabs. Using an 11 mm filter punch and
placing the filter between two filter separators, punch a circle the size of the
-------
carbon tab into the filter. Place carbon adhesive tab affixed to an aluminum stub
on the dull side of the 11-mm polycarbonate filter such that the shiny side of the
filter exposed. If available, a 13-mm diameter polycarbonate filter may be used in
place of the punched out 11-mm filter.
Collect a 10-|j,L aliquot of the mixture from the PLM sample preparation using
the Eppendorf pipette with the modified tip and transfer to a prepared
polycarbonate/adhesive tab substrate. This will yield a loading on a 12-mm SEM
stub of about 100-200 jig, which is a moderately heavy loading. Adjust the
number of aliquots as needed to obtain the target loading.
Prepare a 10-fold dilution of the above suspension to get a suspension of 1-2 mg
dust per mL of isopropanol. Sonicate the suspension in an ultrasonic bath for one
minutes. Remove the suspension and gently shake it to suspend all particles.
Wait one minute to allow the coarse particles to settle. Collect a 10-|j,L aliquot of
the suspended mixture using an Eppendorf pipette with the modified tip and
transfer to a prepared polycarbonate/adhesive tab substrate. This will yield a
loading on a 12-mm SEM stub of about 10-20 |j,g, which is a light loading.
Adjust the number of aliquots as needed to obtain the target loading.
Alternatively, prepare a lightly loaded sample using the filtration method as
follows: Use a Millipore filter apparatus for use with 25-mm filters for filtration.
Place a few drops of isopropanol on the fritted glass surface and place the 25-mm
polycarbonate filter (0.4-um pore size) on the isopropanol. Attach the top of the
apparatus and add a few milliliters of isopropanol to the filter so that no part of it
is exposed to air. Sonicate the suspension (diluted as described in previous
paragraph) in an ultrasonic bath for one minute. Remove the suspension and
gently shake it to suspend all particles. Wait one minute to allow the coarse
particles to settle. Collect 1 mL of the suspended mixture using a pipette and
filter it through the polycarbonate filter. Actual amounts for filtration will vary
based on sample loading. The goal is to have a loading on a 12-mm SEM stub of
about 10-20 |j,g, or about 5-10 percent area coverage, which is a light loading.
Adjust the volume of the aliquot to filter as needed to obtain the target loading.
Place the filter on a carbon adhesive tab on a standard SEM aluminum mount.
The filter needs to be completely flat on the SEM stub. This can be achieved by
forming the wet filter into a gentle U-shape using forceps and the side of the
forefinger, then placing the bottom curve of the filter onto the center of the carbon
adhesive tab and slowly releasing the sides so they lay flat. Trim the edges of the
filter using a razor blade.
After drying, coat the samples on the polycarbonate or polycarbonate/adhesive tab substrates with
carbon using a carbon evaporator with a rotating stage. Transfer the stubs to the SEM in a clean,
covered container.
12.0 Analysis
36
-------
12.1 Analysis by Polarized Light Microscopy
Polarized light microscopy will be conducted using the general techniques outlined in EPA
600/R93/116 (4). For this procedure, four slides (prepared as described in section 11.4) will be
analyzed. The fraction of fibers with refractive index greater than 1.55 (or 1.605) will contain
mineral wool, which includes both slag wool and rock wool, and possibly some E-type glass and
ceramic fibers. The fraction of fibers with refractive index less than 1.55 (or 1.605) will contain
primarily soda-lime glass fibers. For the validation study, numbers of fibers greater than and less
than 1.55 (1.605) refractive index will be counted. Dispersion staining and becke line techniques
may be used. Fiber point counting will be performed at 100 x magnification.
If more than 20 mineral wool fibers are found, continue counting and recording all of the fibers
above and below the index oil refractive index. Report both raw fiber counts per refractive index
category and number of fibers from each category per gram of sample. Continue on to step 12.2.1
to determine the ratio of slag wool to other fibers with refractive index greater than 1.55 (or 1.605)
using EDS as described below.
If less than 20 mineral wool fibers are found on each slide, count the number of slag wool fibers
using SEM/EDS and report as number of fibers per gram of sample.
12.2 Analysis by SEM/EDS
12.2.1 Screening for Slag Wool
Operating conditions for the JEOL 6460-LV SEM are 15 kV, 0.5-5-nA beam current, 10-
mm working distance (system dependent), and zero degree tilt.
Place the more concentrated sample deposited directly on the polycarbonate/adhesive tab
substrate into the SEM. Use the backscattered electron mode at lOOx magnification to
quickly distinguish carbon fibers from inorganic fibers (carbon fibers may be visible, but
not as bright in a BEI). Identify all inorganic fibers over 25 um in length (smaller fibers
cannot be reliably detected at the lOOx operating magnification). When an inorganic fiber
is found, identify the composition of the particle by EDS. Slag wool is the primary fiber
of interest. Record all inorganic fiber results as number of fibers for each fiber type.
For the samples with high fiber loading, as determined by PLM as described in section
12.1, count fibers per type until a statistical representation of the ratios of fiber
compositions in the sample is achieved. Report the ratio (by fiber number) of slag wool
fibers to total MMVF fibers corresponding to the high RI. Use this ratio to correct the
total number for high RI fibers counted by PLM to number of slag wool fibers present.
For the samples with low fiber loading, as determined by PLM as described in section
12.1, scan the entire stub to determine the number of fibers per type. Report the slag
wool fiber results as the number of slag wool fibers/gram of sample.
12.2.2 EDS Screening for Gypsum/Anhydrite
Place the more concentrated sample deposited directly on the polycarbonate/adhesive tab
substrate in the SEM. Choose a random field at lOOx magnification and perform an EDS
analysis on the entire field. Look for the presence of sulfur in this field. If sulfur is
present, continue to Section 12.2.3 or 12.2.5 for analysis of gypsum and concrete by
mapping or particle analysis. If it is not present, repeat the analysis on another random
field. If sulfur is still not present, mark the sample as non-detect (ND) for sulfur.
37
-------
12.2.3 X-Ray Mapping for Gypsum
Place a more dilute sample, deposited directly on the polycarbonate/adhesive tab
substrate or prepared by filtration, in the SEM. Collect binary backscattered electron
images (particles white and background black, shadow off) and secondary electron
images for 10 non-overlapping, random fields at 500 x magnification. Collect x-ray
maps for Na, Mg, Al, Si, S, Ca, and Fe at each of these fields. Fields containing MMVF
will not be used for this analysis. Operating parameters for the SEM are the same as
those for analyzing slag wool. Acquisition parameters for x-ray mapping using the
NORAN System Six Software are time constant 14 (mapping mode, 11333 cps), 10-20 %
deadtime, 256 x 256 image resolution, 20 second frame time, and 100 frames collected
(about 40 minutes total acquisition time). Secondary electron images will be used for
reference only. Save all of the maps and electron images in TIFF format.
Open the backscattered electron image and the Ca and S x-ray maps in Adobe Photoshop.
Make sure that all of the element maps are the same size and resolution by choosing
Image Size from the Image Menu and changing the pixel size or the resolution as needed.
The presence of gypsum can be determined by overlapping the Ca and S maps.
Perform the following functions in Adobe PhotoShop. (A macro is in development to
perform the following functions to decrease user time and human errors in adjusting the
threshold.)
1. Convert each of the three images to grayscale (Image —> Mode —> Grayscale).
2. Perform an auto contrast and brightness on each image and map to increase the scale
of colors (Image —> Adjustments —> Auto Levels).
3. Threshold each element map, Ca and S (do not analyze the backscattered electron
image at this time), by going to the Image Menu and choosing Adjustments —>
Threshold. Adjust the threshold to 128. The background will be black and the
particles white.
4. Invert the image (Image—*- Adjustments —^Invert) to make the background white and
the particles black.
5. Copy the S map and paste it over the Ca map in a separate layer in the file and
change the opacity (located in the Layers window) to 50 % for the S map layer. The
black areas are gypsum/anhydrite.
6. Display a histogram of the image in expanded mode by selecting the Histogram tab
on the Navigator Window (or under the Image Menu in some versions of
Photoshop). Place the cursor over the line for the black area and record the
percentile for the black area. This is the percentage of particles containing Ca and S
in the entire field.
NOTE: If a binary backscattered electron image is obtained during data collection, then
steps 7-11 may be deleted. The Invert function will, however, need to be applied to make
the particles black and the background white before continuing to step 12.
7. Begin analysis of the backscattered electron image. Select the particles by going to
the Select Menu and choosing Color Range. Go to the selection pulldown menu and
choose Highlights.
8. Fill the selection with black by going to the Edit Menu —> Fill and choosing black
from the color pulldown menu.
9. Select the inverse areas by going to the Select Menu and selecting Inverse.
10. Fill the selection with white by going to the Edit Menu —> Fill and choosing white
from the color pull down menu.
11. Deselect the area by clicking on the image.
12. Perform the Threshold and Histogram functions for the backscattered electron image
38
-------
as outlined in 3 and 6. Record the histogram result for the backscattered electron
image.
Determine the area percent of gypsum by performing the calculations in Section 13.0.
12.2.4 X-Ray Mapping for Ca-Rich Particles
Analysis of components of concrete will be performed on the same fields as the
gypsum/anhydrite analysis. At this time, only a method for the determination of the area
percent of Ca-rich particles is presented. See Section 2.1 for discussion.
Perform the following steps on the Ca x-ray map Tiff file in Adobe Photoshop:
1. Convert the Ca x-ray map to grayscale (Image —> Mode —> Grayscale).
2. Perform an auto contrast and brightness on the map to increase the scale of colors
(Image —> Adjustments —> Auto Levels).
3. Threshold the Ca map by going to the Image Menu and choosing Adjustments —>
Threshold. Adjust the threshold to 128. The background will be black and the
particles white.
4. Invert the image (Image—*- Adjustments —^Invert) to make the background white and
the particles black.
5. Display a histogram of the image. Place the cursor over the line for the black area
and record the percentile for the black area. This is the area percent coverage of
particles containing Ca in the entire field.
Determine the maximum area percent coverage of non-gypsum, Ca-rich particles by
performing the calculation in Section 13.0.
12.2.5 Particle Analysis for Identification of Gypsum and Concrete.
Place the more dilute sample, deposited directly on the polycarbonate/adhesive tab
substrate or prepared by filtration, in the SEM. Particle analysis will be used to identify
gypsum and concrete particles.
Perform particle analysis at 500 x magnification. All other operating parameters for the
SEM are the same as those used to analyze for slag wool (Section 12.2.1). A binary
backscattered electron image should be used in particle analysis mode. Particle analysis
parameters should be set to analyze all particles in the field greater than 0.5 um and to
separate touching particles. For particles greater than 5 um, scan the entire particle; spot
analysis is adequate for smaller particles. The x-ray spectrum and counts for all particles,
and an image of particles > 20 um long, will be recorded and saved. Other particle
parameters to be reported will include the maximum, minimum, and average diameters,
the aspect ratio, and area of each particle.
It will be necessary to review data collected by automated software to ensure data
integrity. An Excel spreadsheet, in conjunction with images and x-ray data, may be used
for this purpose. Particles should be sorted into one of three categories: Ca-S (gypsum),
Ca-rich, and Other. Aid in identification of particles may by facilitated by referencing
the U.S. Geological Survey's WTC Dust Particle Atlas (1). A particle classification
protocol will be developed based on the data from the validation study.
The number of particles analyzed will be determined using the results of the validation
study. For the study, the area percent of each component should be within 10% relative
error or better. Typically, data for 1000 - 1200 particles should be acquired.
39
-------
Results for particle analysis will be recorded as area percent gypsum and area percent concrete
particles for each field and average area percent for the each component in the sample.
13.0 Data Analysis and Calculations
Table 60 To determine the concentration of slag wool in fibers/gram, perform the
following calculations:
Determine the number of fibers withRI > 1.55 (or 1.605):
# fibers identified -^ mg of sample on slide x 1000 = fibers/gram on slide
Determine the percentage of fibers with the composition of slag wool with RI > 1.55 (or 1.605):
Fibers/gram on slide x # fibers identified as slag wool = fibers slag wool/gram on slide
Total number of fibers identified by EDS with RI > 1.55 (or 1.605)
Back calculate to the number of fibers per gram of the original sample:
Fibers slag wool/g on slide x g after sieving x g sample after ashing = Total f/g of sample
g before sieving x g sample before ashing
Table 61 To determine the area percent of gypsum/anhydrite from the x-ray mapping
procedure, perform the following calculations:
Determine the area percent of gypsum/anhydrite in each field of view.
% of black area in Ca-S map overlay x 100 = area % gypsum
% of black area in BSE image
Calculate the average percentage of gypsum/anhydrite for the sample.
(area % gypsum)^ + (area % gvpsum)f7 + ... = Avg. area % gypsum
number of fields
Table 62 To determine the maximum area percentage of Ca-rich particles, which includes
concrete particles, from the x-ray mapping procedure, perform the following calculations:
Determine the area percent of non-gypsum Ca-rich particles in each field of view:
(% black area Ca map) - (% black area Ca-S map) = % non-gypsum Ca-rich particles
% black area on BSE image
Calculate the average percentage of non-gypsum Ca-rich particles for the sample:
(area % Ca-rich particles)^ + (area % Ca-rich particles'ln + ... = Avg. area % Ca-rich particles
number of fields
Table 63 Calculate the area percent for gypsum and concrete by summing the areas of
each particle in for each particle type and dividing by the total area analyzed:
area gypsum 1 + area gypsum 2+ ... x 100 = area percent gypsum (do likewise for concrete)
total area analyzed
Rules for concrete and gypsum classification are currently being developed.
40
-------
14.0 References
1. Lowers, Heather A., Meeker, Gregory P., Brownfield, Isabelle K., 2005. World Trade Center
Dust Particle Atlas: U.S. Geological Survey Open-File Report 2005-1165. On the web at
http://pubs.usgs.gov/of/2005/1165/.
2. Meeker, G.P., Taggart, J.E., and Wilson, S.A., 1998. A Basalt Glass Standard for Multiple
Microanalytical Techniques. Proceedings: Microscopy and Microanalysis 1998. Microscopy
Society of America.
3. A polished and carbon coated calibration reference sample of BIR1 -G may be obtained by
contacting Stephen Wilson, U.S. Geological Survery, MS 973, Denver Federal Center, Denver,
CO, 80225, swilsontgiusgs.gov.
4. Perkins, R.L. and Harvey, B.W., 1993, TEST METHOD: Method for the Determination of
Asbestos in Bulk Building Materials, EPA/600/R-93/116.
41
-------
15.0 Appendix: DATASHEETS
Determination of Slag Wool Fibers in Dust- PLM with Dispersion Staining
Sample ID: Project:
Analyst:
Circle One: Original Duplicate Triplicate Date:
General Sample Appearance:
Homogeneous?:
Structure #
Y
Rl Fluid
1.55
1.605
Dispersion Staining
>RI
RI
-------
SEM Sheet
Reference ASTM - D5755-03
Report Number: _
Sample Number:.
File Name:
Sample Description:
Preparation Date:
Analysis Date:
Computer Entry Date:
Sample weight:
Dilution Volume:
Volume Aliquot:
Magnification:
By:
By:
By:
grams
mL
uL
X
Structure*
Field #
Fiber Type Length (Microns) Width (Microns)
Image
EDS
43
-------
-------
APPENDIX E: REPORT FROM THE U.S. EPA CONTRACTOR ON THE SCREENING
METHOD STUDY
Versar
6850 Versar Center
Springfield, VA 22151
Ms. Jacky Rosati
US Environmental Protection Agency
E-305-03 109 T.W. Alexander Drive
Research Triangle Park, NC 27711 July 21, 2005
Dear Ms. Rosati:
Attached is a preliminary report based on analytical data thus far received, for dust
samples collected primarily in the New York City area. Most of the samples were taken in areas
that, it is believed, were not affected by particulate matter generated during the World Trade
Center (WTC) collapse (i.e., background samples). Some of the samples were spiked with one
or the other of two dusts that are believed to have originated from the WTC collapse. The
analytical protocol was developed by the government, specifically for this project, and was
modified as the project developed. The purpose of the testing was to determine if the spiked
background dusts could be distinguished from those samples that were not spiked.
Three parameters were measured to make this determination: (1) slag wool fiber content;
(2) calcium-rich particle content; and (3) gypsum particle content.
The analytical data indicate that:
• With respect to calcium-rich particles and gypsum particles, spiked samples cannot readily be distinguished
from background samples.
• With respect to slag wool content in the samples spiked with the first of the two WTC dusts, spikes at the
10% level may be statistically identifiable as WTC-contamination, although spikes at or below the 5% level
are probably not identifiable.
• With respect to slag wool content, samples spiked with 5% and 10% of the second of the two WTC dusts
are easily identifiable as WTC-contaminated. Even at the 1% spike level, samples may be statistically
identifiable.
The attached preliminary report will explain the above conclusions in more detail. However, it must be
noted that all of the analytical data from the eight laboratories that performed the analysis has not yet been
received. Nevertheless, it is believed that the above conclusions will not likely change once those additional
data are incorporated.
Sincerely,
Stephen M. Schwartz, P.E., Q.E.P.
Project Manager
45
-------
Versar
Preliminary Report
of
Analysis of New York City Area Dust Samples
Purpose:
The objective of this study is to determine if New York City area dusts that are contaminated with
varying levels of dusts known to originate from the collapse of the World Trade Center (WTC) can be
distinguished from background dusts that are believed not to be contaminated with WTC dusts.
Project Summary:
In the initial portion of the testing, 10 dust samples from New York City areas that are believed
not to be contaminated with dusts originating from the collapse of the WTC were used. These are
referred to as the first set of background samples. An additional background dust sample was spiked at
1, 5, and 10 percent levels (by weight) with dust believed to have originated from the WTC collapse.
An additional background sample was spiked at 1, 5, and 10 percent levels with a second dust sample
that is believed to have originated from the WTC collapse. Therefore, a set of 16 samples was
generated:
• 10 different background dusts
• 3 samples, each consisting of one background dust sample spiked with one source of WTC
dust at 1, 5, and 10% levels
• 3 samples, each consisting of one background dust sample spiked with a second source of
WTC dust at 1, 5, and 10% levels
Initially, 32 samples were sent to each of eight analytical laboratories (three U.S. government, and
five private). The 32 samples consisted of two identical sets (i.e., duplicates) of the 16 samples
discussed above. The private laboratories did not know that there were duplicate samples. Further,
they did not know which, if any, of the samples contained WTC spikes.
Subsequently, a second set of 28 different background samples was analyzed to obtain a better
understanding of the variability of background dusts. These 28 samples were sent to only one of the
five private laboratories.
It was ultimately agreed that each of the laboratories would perform the following
three Scanning Electron Microscopy-based (SEM) analyses on each of the
samples they received (see Methodology and Data Analysis section):
• Slag wool fiber content (in number of fibers per gram of dust). Slag wool was a significant
component of the WTC insulation material.
• Calcium-rich particle content (in area percent concentration in the SEM field). Such particles
are assumed to be indicative of cement/concrete-like particles.
• Gypsum particle content (in area percent concentration in the SEM field). Such particles are
assumed to be indicative of "dry wall" (i.e., gypsum-containing wall board).
Conclusions:
-------
A number of conclusions can be drawn from the analytical results thus far
obtained. It is not expected that data that are subsequently received will
substantially change these conclusions. It must be noted that there are several
caveats that affect the quality of the data. Those are discussed later in this report.
2. With respect to calcium-rich particles and gypsum particles, spiked samples
cannot readily be distinguished from background samples.
Tables 1 and 2 present the analytical data thus far available for calcium-rich and
gypsum content respectively. Analysis was performed using SEM and x-ray
mapping (XRM) techniques. The shaded areas represent the samples spiked
with 1, 5, and 10% WTC dust. The others areas are background samples.
Sample designations followed by "(1)" and "(2)" are duplicate samples.
(Samples received by the laboratories had random identification numbers, so
that the laboratories did not know if any samples were duplicates, nor did they
know if any samples contained WTC dust.) In addition, Table 3 is the
analysis of a subsequent 28 background samples, analyzed by only laboratory
"B". Analysis of calcium-rich and gypsum particles for this sample set is
shown on Table 3.
The average of all background samples (including the second set of 28 samples) for
calcium-rich particles is 22.3 area percent, with a high value of 66.5% and a
low value of 4.2%. The average for the spiked samples is 20.7%, with the
highest value being 25.9%. The 1, 5, and 10% spiked samples do not show
any trend with respect to calcium-rich particle content (i.e., they do not show
any increase as the spike level increases).
The average of all background samples (including the second set of 28 samples) for
gypsum particles is 11.7 area percent, with a high value of 56.5% and a low
value of 0.1%. The average for the spiked samples is 9.3%, with the highest
value being 32.8%. The 1,5, and 10% spiked samples do not show any trend
with respect to gypsum particle content.
3. With respect to slag wool content in the samples spiked with the first of the
two WTC dusts, spikes at the 10% level may be statistically identifiable as
WTC-contamination, although spikes at or below the 5% level are probably
not identifiable.
Table 4 presents all the analytical data thus far available for SEM slag wool fiber
analysis (as the number of slag wool fibers per gram of dust). The shaded
areas represent samples that are spiked at the 1, 5, and 10% levels with WTC
dust. Table 3 also presents additional slag wool fiber background-only sample
data (next to last column). It can be seen from Figure 1 that for those spiked
samples designated as "DB" that at the 5% spike level, the slag wool
concentrations probably do not exceed one standard deviation above the
average slag wool background concentration (including the Table 3
47
-------
background data). However, at the 10% spike level, the slag wool
concentration typically exceeds one standard deviation (see Figure 2), but
never exceeds two standard deviations above the average background sample
concentration. The average background concentration is about 27,400 fibers
per gram. The standard deviation is about 40,100 fibers per gram.1
It should be noted that there is a trend showing a clear increase in slag wool fiber
concentration from the 1% to the 10% spike level (see "DB" sample shaded
area on Table 4). However, the numerical values of those concentrations, as
noted above, are still less than two standard deviations above the average
concentration.
4. With respect to slag wool content, samples spiked with 5% and 10% of the
second of the two WTC dusts are easily identifiable as WTC-contaminated.
Even at the 1% spike level, samples may be statistically identifiable.
The slag wool content data for the samples spiked with the WTC dust shown in Table
4 as "USGS" are easily identifiable. As can be seen in Figures 4 and 5,
samples spiked with the USGS WTC dust at the 5 and 10% levels are
essentially all more than two standard deviations above the average
background sample concentration. (Average plus two standard deviations
would be about 108,000 fibers per gram.2) At the 1% spike level though,
WTC dust is more difficult to identify because the slag wool concentrations
are mostly between one and two standard deviations above the average
background sample (see Figure 3).
5. With respect to slag wool content, clearly, there is a large difference between
the two WTC dust spikes used. In the "DB"-spiked samples, as noted above,
it is expected to be more difficult to determine a significant slag wool fiber
concentration difference from background. The "USGS"-spiked samples
clearly had significantly more slag wool fiber content than the "DB" samples.
6. Examining Tables 1, 2, and 4 and the Figures, it can be seen that the analyses
for the duplicate samples rarely replicate one another. However, the variation
between duplicate sample values (i.e., intralab) is about half of the variation
between individual laboratory values (inteHab).3
1 Background concentration data for this analysis excluded several samples that were known to have high
slag wool content, specifically the Cl-RTP samples (see Table 4), and samples C2,3,4,5,6 (see Table 3).
2 Ibid.
3 For slag wool fiber analysis, the average difference between the analyses of duplicates (i.e., intralab
differences) is about 50% of one standard deviation of the between-laboratories analyses (i.e., inteHab
differences). For both calcium-rich and gypsum particle analysis the average intralab difference is 20% of
the interlab difference.
48
-------
Methodology and Data Analysis:
The analytical protocol was developed specifically for this project by one of the
government laboratories, and modified by all laboratory participants at a meeting held
for that purpose. All laboratory participants held weekly conference calls as the
analytical program was proceeding to discuss general issues with the protocol.
Additional modifications were made to the protocol based on those conference calls.
The original protocol included analysis by Polarized Light Microscopy (PLM), so
data are also available for PLM analysis. The PLM analyses were curtailed because it
became obvious that PLM could not adequately differentiate between fiber types.
Further, total fiber concentrations were also determined, both by PLM and SEM
methods, but those data are not presented in this report.
Caveats:
There are a few factors that may contribute to data uncertainty. Nevertheless, it is
unlikely that these factors will alter the above major conclusions. Some of these factors
are as follows:
1. As noted earlier, not all of the analytical data have been received.
2. Dust samples were collected by several methods. Evaluation of the
sampling methodology was not part of the study.
3. To determine fiber concentration, fibers were counted using an SEM.
Different laboratories diluted samples to different levels before counting,
introducing some variability of results.
4. Laboratory equipment capabilities and personnel skills varied.
49
-------
TABLE 1: SEM X-Ray Mapping - Calcium-Rich Area Percent
Sample
Designations
AP5(1)
AP5(2)
CMC(1)
CMC(2)
HS3(1)
HS3(2)
WGS(1)
WGS(2)
MW(1)
MW(2)
DB1%(1)
DB1%(2)
DB5%(1)
DB5%(2)
DB10%(1)
DB10%(2)
C1-RTP(1)
C1-RTP(2)
USGS1%(1)
USGS1%(2)
USGS5%(1)
USGS5%(2)
USGS10%(1)
USGS10%(2)
USC(1)
USC(2)
FP(1)
Laboratory Letter Codes
A
B
23.4
22.4
27.7
34.1
10.3
17.8
22.8
19.9
12.2
12.0
18.2
13.1
13.4
20.5
14.4
12.6
13.2
16.2
16.5
21.0
14.8
14.6
17.0
17.9
12.3
9.4
13.0
C
15.6
23.0
18.9
26.6
26.4
D
20.4
22.1
21.8
21.5
14.1
13.3
13.2
16.3
14.2
10.9
14.0
14.1
16.1
12.9
15.2
14.9
11.3
11.9
17.4
11.1
14.2
16.2
15.8
12.1
11.1
9.5
11.6
E
14.4
16.8
6.7
20.4
6.4
14.8
13.9
5.7
12.6
8.3
18.3
15.4
10.8
4.1
8.6
10.8
7.5
5.6
12.4
5.7
11.9
10.7
10.9
9.8
19.5
6.6
5.9
F
11.6
9.8
7.9
10.1
15.3
6.5
7.7
7.4
7.6
5.7
8.9
9.4
9.0
7.5
8.0
8.1
6.1
4.2
7.6
7.4
6.7
8.3
8.9
8.3
5.4
7.1
6.2
G
30.7
39.6
55.1
38.9
29.1
44.0
58.4
49.3
55.9
40.0
50.7
57.1
43.5
42.4
H
45.8
48.9
60.4
55.2
63.1
49.7
53.4
52.3
46.3
49.8
50.2
52.2
49.0
39.6
40.6
48.8
66.5
61.0
43.2
41.9
53.6
51.8
40.0
45.1
46.1
40.3
70.9
Location Key:
Chittenden Avenue, Manhattan
Columbia Medical Center, W68th Street, Manhattan
Teaneck, NJ
Nassau County, LI
West End Ave Between 105th and 106th Streets, Manhattan
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
Research Triangle Park, NC
Research Triangle Park, NC
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
Federal Courthouse, White Plains, NY
Federal Courthouse, Central Islip, LI
-------
FP(2)
MUNYC1(1)
MU NYC 1(2)
MUNYC2(1)
MUNYC2(2)
10.5
25.4
19.7
20.4
17.6
10.3
17.1
14.0
19.9
14.4
10.3
15.2
31.9
27.7
13.6
8.2
6.3
8.0
8.9
7.3
55.6
45.4
61.5
39.6
36.6
56.8
57.8
Northern Manhattan, Above 70th Street
Northern Manhattan, Above 70th Street
Samples spiked with WTC dust, at 1, 5, and 10% levels are shaded. All others are
background samples.
51
-------
TABLE 2: SEM X-Ray Mapping - Gypsum Area Percent
Sample
Designations
AP5(1)
APS (2)
CMC(1)
CMC(2)
HS3(1)
HS3(2)
WGS(1)
WGS(2)
MW(1)
MW(2)
DB1%(1)
DB1%(2)
DB5%(1)
DB5%(2)
DB10%(1)
DB10%(2)
C1-RTP(1)
C1-RTP(2)
USGS1%(1)
USGS1%(2)
USGS5%(1)
USGS5%(2)
USGS10%(1)
USGS10%(2)
USC(1)
USC(2)
Laboratory Letter Codes
A
B
8.0
20.3
4.3
6.9
5.9
14.9
2.9
6.1
3.8
5.4
7.2
7.1
7.3
6.1
6.5
5.0
8.5
8.7
6.3
5.4
7.7
2.5
6.3
4.8
4.8
6.2
C
13.8
3.4
8.7
15.2
9.8
D
14.4
11.3
4.8
3.0
9.2
11.0
5.4
4.7
7.0
5.3
5.7
5.2
5.5
5.5
7.8
4.8
9.7
8.2
5.8
4.1
5.7
4.1
7.1
4.8
5.2
4.2
E
0.9
1.8
0.2
1.0
0.3
2.3
0.2
0.2
0.2
0.1
3.0
1.1
0.7
0.1
0.5
0.6
0.2
0.3
1.0
0.2
0.9
0.5
1.2
0.7
1.2
0.2
F
2.5
1.6
1.1
1.0
5.7
1.5
0.4
0.3
0.7
1.1
0.6
1.3
1.2
1.6
1.0
1.9
1.3
0.8
0.9
0.9
1.1
2.4
1.1
1.4
0.7
2.4
G
34.1
31.3
26.1
30.8
44.0
29.0
19.0
23.2
22.0
29.1
25.7
24.5
24.9
H
26.1
33.7
22.4
17.6
42.9
40.5
42.2
39.1
37.6
41.6
28.0
30.0
24.3
28.9
27.0
28.5
53.4
50.4
29.4
29.2
29.3
21.7
30.9
32.8
27.1
32.4
Location Key:
Chittenden Avenue, Manhattan
Columbia Medical Center, W68th Street, Manhattan
Teaneck, NJ
Nassau County, LI
West End Ave Between 105th and 106th Streets, Manhattan
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
Research Triangle Park, NC
Research Triangle Park, NC
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
Federal Courthouse, White Plains, NY
52
-------
FP(1)
FP(2)
MUNYC1(1)
MU NYC 1(2)
MUNYC2(1)
MUNYC2(2)
11.6
4.4
10.5
3.0
5.5
4.2
5.4
6.1
9.2
5.5
6.1
6.0
0.3
0.6
1.2
1.4
9.2
0.7
1.2
1.5
0.9
1.0
2.5
1.8
24.5
26.8
31.0
56.5
40.0
24.1
26.3
30.8
29.5
Federal Courthouse, Central Islip, LI
Northern Manhattan, Above 70th Street
Northern Manhattan, Above 70th Street
Samples spiked with WTC dust, at 1, 5, and 10% levels are shaded. All others are
background samples.
-------
Table 3: New York City Background Dust Samples
Sam
pie
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
EPA Sample ID
HS1 -06-01**
HS1 -06-02
AP2-07-01
AP2-07-02
AP3-08-01
AP3-08-02
HS2-09-01
AP4-10-01
AP7-14-01
CMC-17-01
HS3-18-01
WGS6557
WGS5826-1
PT152W88
PT152W88-2ndFI
CY321W80
MW924WEAve
C2**
C3
C4
C4 (no date)
C5
C6
N-01S
SEM XRM
Calcium-
Rich (area
%)
5.8
16.3
12.4
9.9
10.5
17.3
25.6
13.1
17.5
30.5
14.3
10.2
18.1
17.5
15.8
14.7
21.1
7.9
16.8
13.7
17.5
16.7
10.0
12.3
Gypsum
(area %)
5.8
4.0
8.5
7.2
4.7
7.6
9.6
11.5
6.0
10.4
9.0
4.2
6.2
8.4
5.6
12.2
10.1
6.1
3.4
3.8
8.8
7.2
10.9
7.8
SEM (Heavy Loading)
Slag Wool
(fibers/g)*
35,565
230,769
32,432
7,692
12,500
<3,636
7,605
42,857
3,333
4,651
11,858
34,826
15,564
17,021
19,305
30,888
8,097
46,703
170,309
160,772
488,372
74,236
280,762
369,231
Total
Fibers
(fibers/g)
104,603
523,077
113,514
130,769
212,500
21,818
22,814
485,714
23,333
23,256
71,146
44,776
54,475
46,809
42,471
34,749
28,340
102,890
321,696
227,760
790,698
148,472
415,039
523,077
Particle Count
Slag
Wool
9
15
6
2
2
0
2
3
1
1
3
7
4
4
5
8
2
11
18
24
21
17
23
24
Total
Fibers
25
34
21
34
34
6
6
34
7
5
18
9
14
11
11
9
7
24
34
34
34
34
34
34
Stony Brook, LI
West End Ave between 72nd and 73rd Streets
30th Ave between 21st and 23rd, Queens
70th Street between 20th and 21st Ave, Brooklyn
79th St between York and East End Ave, Manhattan
92nd Street between Columbus and CPW, Manhattan
Columbia Medical Center, W. 168th St, Manhattan
Teaneck, NJ
Nassau County, LI
Nassau County LI
88th Street between Amsterdam and Columbia, Manhattan
88th Street between Amsterdam and Columbia, Manhattan
80th Street between Riverside and East End Ave, Manhattan
West End Ave between 105th and 106th Streets
Research Triangle Park, NC
Research Triangle Park, NC
Research Triangle Park, NC
Research Triangle Park, NC
Research Triangle Park, NC
Research Triangle Park, NC
Edison, NJ
54
-------
25
26
27
28
Aver
age
Stan
dard
Devi
ation
Coeff
.Of
Varia
nee
Nevins Ct
E Curtis Ave**
LBI
Mixture
16.0
7.9
7.3
19.1
14.9
5.5
0.4
9.6
9.0
16.3
11.8
8.1
3.0
0.4
<4,367
5,173
<3,636
7,194
84,709
128,759
1.5
91,703
24,138
61,818
35,971
168,837
200,808
1.2
0
2
0
2
21
7
17
10
* A fiber count of one fiber was used to calculate the analytical sensitivity for non-
detects.
** Internal laboratory duplicates were run on these samples. The result shown is the
average of the two duplicates ("<" samples were assumed to be 0).
Edison, NJ
Edison, NJ
Long Beach Island, NJ
NE Queens
55
-------
TABLE 4: SEM - Slag Wool Fiber Count/Gram of Sample
Sample
Designations
AP5(1)
AP5(2)
CMC(1)
CMC(2)
HS3(1)
HS3(2)
WGS(1)
WGS(2)
MW(1)
MW(2)
DB1%(1)
DB1%(2)
DB5%(1)
DB5%(2)
DB10%(1)
DB10%(2)
C1-RTP(1)
C1-RTP(2)
USGS1%(1)
USGS1%(2)
USGS5%(1)
USGS5%(2)
USGS10%(1)
USGS10%(2)
USC(1)
USC(2)
FP(1)
FP(2)
Laboratory Letter Codes
A
non-det.
non-det.
16,393
5,900
12,232
5,747
34,826
72,562
67,797
104,575
84,746
246,914
98,039
600,000
1,218,855
73,394
18,519
B
3,663
<3636
3,448
<3875
7,299
7,692
34,221
10,753
18,939
3,717
10,909
17,422
29,197
25,271
66,421
77,778
159,011
173,585
109,091
83,032
404,332
343,284
840,231
1,366,470
56,025
41,199
18,051
16,470
C
5,451
9,133
32,385
33,646
74,837
57,644
50,293
50,160
364,813
531,277
521,212
D
non-det
6,980
11,800
9,620
19,000
18,600
26,400
18,100
18,700
31,800
29,900
27,300
50,800
35,800
113,000
95,100
269,000
165,000
119,000
104,000
681,000
146,000
1,620,000
238,000
91,800
40,700
16,300
31,800
E
<249
<667
<282
309
<286
<667
<256
6,990
1,320
893
<2,000
3,770
31,000
6,900
108,000
20,400
168,000
21,900
366,000
18,700
227,900
191,000
1,410,000
271,000
33,700
7,890
1,100
3,920
F
<500
500
<4,500
667
2,750
5,060
1,630
<30,500
1,000
<45,500
1,920
12,500
1,700
14,700
7,000
34,100
38,000
160,000
79,800
79,500
433,000
197,000
629,000
372,000
15,600
48,400
12,400
30,500
G
2,470
13,910
5,780
6,100
<6,320
7,370
9,480
3,520
13,630
18,080
7,650
1,320
6,230
13,040
12,900
25,210
84,650
39,930
9,200
25,370
66,450
73,330
144,120
33,040
<3,230
3,540
11,920
<1,181
H
<7,386
<7,698
<7,241
<6,289
<7,576
34,813
16,077
18,399
17,301
<9,497
15,924
16,038
107,143
70,472
114,638
96,696
188,088
318,143
90,992
137,363
672,926
347,904
734,767
413,153
29,268
74,212
28,249
25,489
Location Key:
Chittenden Avenue, Manhattan
Columbia Medical Center, W68th Street, Manhattan
Teaneck, NJ
Nassau County, LI
West End Ave Between 105th and 106th Streets, Manhattan
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
4 Albany Street Spiked into NE Queens background dust
Research Triangle Park, NC
Research Triangle Park, NC
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
Federal Courthouse, White Plains, NY
Federal Courthouse, Central Islip, LI
56
-------
MUNYC1(1)
MUNYC1(2)
MUNYC2(1)
MUNYC2(2)
10,840
41,298
7,220
3,745
28,777
48,507
14,400
20,200
66,500
45,500
14,900
1,960
1,390
24,200
13,100
<22,300
17,800
30,500
<2,545
<1 ,228
<12,453
2,330
6,803
41,118
123,106
59,473
Northern Manhattan, Above 70 Street
Northern Manhattan, Above 70 Street
Samples spiked with WTC dust, at 1, 5, and 10% levels are shaded. All others are background samples.
For data analysis purposes:
• Non-det = Non-detect - zero slag wool fibers were noted in the sample.
• <# indicates that the value was less than the detection limit of the respective laboratory. When this result was reached, the value was divided
by the square root of 2.
57
-------
Table 5: SEM - Slag Wool Fiber Count
Sample Designations
AP5(1)
AP5(2)
CMC(1)
CMC(2)
HS3(1)
HS3(2)
WGS(1)
WGS(2)
MW(1)
MW(2)
DB1%(1)
DB1%(2)
DB5%(1)
DB5%(2)
DB10%(1)
DB10%(2)
C1-RTP(1)
C1-RTP(2)
USGS1%(1)
USGS1%(2)
USGS5%(1)
USGS5%(2)
Laboratory Letter Codes
A
0
0
3
1
2
1
7
8
12
16
15
20
15
99
B
1
0
1
0
2
2
9
3
5
6
3
5
8
7
18
21
45
46
30
23
112
92
C
1
2
7
7
12
12
9
11
62
D
0
3
5
4
8
8
11
8
8
14
13
12
22
16
48
42
116
72
54
47
194
65
E
0
0
0
1
0
0
1
7
6
2
0
4
7
8
13
9
16
30
27
23
25
21
F
0
1
0
0
1
3
1
0
1
0
1
2
1
2
2
3
4
4
11
4
>20
19
G
2
6
4
5
0
4
6
3
6
22
7
1
6
11
10
25
22
17
11
22
64
27
H
0
0
0
0
0
4
2
2
2
0
2
2
6
10
13
12
24
37
10
15
43
39
Location Key:
Chittenden Avenue, Manhattan
Teaneck, NJ
Nassau County, LI
4 Albany Street Spiked into NE Queens background
dust
4 Albany Street Spiked into NE Queens background
dust
4 Albany Street Spiked into NE Queens background
dust
4 Albany Street Spiked into NE Queens background
dust
4 Albany Street Spiked into NE Queens background
dust
4 Albany Street Spiked into NE Queens background
dust
Research Triangle Park, NC
Research Triangle Park, NC
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
58
-------
USGS10%(1)
USGS10%(2)
USC(1)
USC(2)
FP(1)
FP(2)
MUNYC1(1)
MUNYC1(2)
MUNYC2(1)
MUNYC2(2)
181
6
3
1
7
45
45
13
11
5
5
2
1
8
13
124
129
450
105
39
18
7
14
6
9
28
20
38
19
6
6
3
4
4
2
3
24
19
16
13
4
2
1
3
0
3
3
18
9
0
2
2
0
0
0
0
1
41
49
3
8
3
3
1
5
13
7
USGS Dust Spiked into NE Queens background dust
USGS Dust Spiked into NE Queens background dust
Samples spiked with WTC dust, at 1, 5, and 10% levels are highlighted in yellow.
59
-------
ADDENDUM TO VERSAR REPORT: SEM CALIBRATION DATA
MEMORANDUM
TO: Jacky Rosati
CC: David Friedman
FROM: Stephen Schwartz
DATE: Augusts, 2005
SUBJECT: BIR-1G Sample Analyses
Identical mounted and polished reference samples, each designated BIR-1G, were
sent to each of the five private laboratories participating in the analyses of dust samples
from New York City and elsewhere. The samples were analyzed by Scanning Electron
Microscopy/Energy Dispersive X-ray Spectrometry (SEM/EDX) to determine their
elemental content. The purpose of the study was to determine the variation within and
between each of the laboratories, and to assess their ability to identify elements using this
technology.
Each of the five laboratories analyzed their BIR-1G sample between 4 and 11
times, as convenient (there was no requirement for a specific number of analyses). The
average elemental concentration data for each laboratory is presented in the attached
table. For calcium (Ca), magnesium (Mg), silicon (Si), and oxygen (O)4, which
constitute over 80% by weight of the elemental composition, the standard deviation
within each laboratory, for each element, was typically much less than 10% (i.e., the
coefficient of variation). Likewise, the coefficient of variation between laboratories for
Ca, Mg, Si, and O, as shown on the attached table, was also much less than a 10%. (The
graphic presentations of the EDX spectra within and between laboratories also appear to
be extremely similar.
Therefore, it can be concluded that each of the laboratories was easily able to
achieve excellent precision, by SEM/EDX, in quantifying the elements that were present
in larger concentrations.
4 Some of the laboratories reported results as the weight percent of the elemental oxides, specifically: Na2O,
MgO, A12O3, SiO2, CaO, TiO2, FeO, K2O, and MnO2. Oxide values were converted to individual elemental
values (e.g., A12O3 is about 53% Aluminum, and 47% oxygen by weight).
60
-------
AVERAGE ELEMENTAL CONCENTRATION REPORTED FOR BIR-1G SAMPLES (Weight Percent of Sample)
Lab
A
B
C
D
E
F
G
H
AVERAGES
Standard Dev.
(% of
Average)
Sodium
0.44
1.14
1.84
1.14
61.40
Magnesium
5.02
5.82
5.10
5.31
8.33
Aluminum
8.11
8.75
7.88
8.25
5.48
Silicon
21.60
24.76
22.20
22.85
7.35
Calcium
9.66
7.68
10.30
9.21
14.83
Titanium
0.68
0.52
0.64
0.61
13.58
Iron
9.68
5.56
8.37
7.87
26.73
Potassium
NR
0.00
0.03
0.02
141.42
Manganese
NR
0.00
0.21
0.11
141.42
Oxygen
44.84
45.78
43.43
44.68
2.65
NR - Not Reported
61
-------
APPENDIX F: STATISTICAL ANALYSIS AND INTERPRETATION OF TEST
RESULTS - LABORATORY QUALIFICATION
Slag wool fiber content as a discriminator for residual WTC contamination in
indoor dust sample: Interpretation of multi-laboratory test results
Introduction
Eight laboratories were each challenged with a number of blinded dust sample aliquots to
determine number of slag wool fibers per gram. These samples included background
dusts from various locations in NYC and a series of samples of common household dust
spiked with different levels of WTC collapse dust collected in 2001 or in 2004. The
purpose of this endeavor was to assess whether or not the method developed can be used
by qualified laboratories to discriminate between WTC and non-WTC impacted dust
samples.
In the following discussion, individual laboratories are evaluated and ranked for validity
and precision, and then the top performers are further evaluated as groups to determine
the expected confidence level for the slag wool content of any individual and randomly
assigned)sample.
Laboratory Qualification
Validity: Assessment of validity was conducted by analysis of a series of spiked samples
where the expected response ratios are known. The challenge samples consisted of a large
volume of non-impacted background dust collected in 2004 from locations in Northeast
Queens over ten miles from the WTC site. This dust was subsequently spiked with 1, 5,
and 10% WTC dusts by weight using either bulk collapse dusts collected in September
2001 immediately following the disaster (designated as USGS dust), or nominally
undisturbed dusts collected in 2004 in the abandoned Deutsche Bank (DB) complex that
borders the south side of the WTC complex (designated as 4 Albany dust).
Using units of (# slagwool fibers)/(gram of dust), preliminary analyses showed a mean
value of 12,200,000 (s.d. 1,697,056) for USGS dust, 579,667 (s.d. 173,782) for 4
ALBANY dust, and a nominal background level of 7,190. Based on these data, the
expected values of slope expressed as [(# slagwool fibers)/(gram dust)]/[% spike level]
are 121,928 and 5,725, respectively for USGS and 4 Albany spikes. Specifically, each lab
was furnished two samples each of 1, 5, 10% spikes from both 4 Albany and USGS series
for a total of 12 spiked samples plus a series of 20 additional background samples
collected from random locations all over the greater NYC area.
Scatterplots and linear least squares regressions were constructed for each lab and for
each of the two spike series. Preliminary inspection showed no apparent violations of
underlying assumptions required for regression analysis (primarily homogeneity of
variance); as such no lognormal transformation was performed. Using a forward selection
strategy, it was found that a higher order polynomial model does not statistically improve
the linear fit; this is expected as the sample set is designed as a linear progression. Also, a
simpler but more general "runs" test for each linear regression confirmed these results.
Data handling and manipulation was performed with Microsoft Excel SP-2; statistical
analyses were performed with SAS 9.1.3 XP-Pro (proc rsreg/lackfit, proc reg, proc
62
-------
mixed, and proc univariate); graphing, ANOVA, and various other statistical results were
performed or verified with GraphPad Prism 3.03. The linear regression results are given
in Table 1; a summary of SAS proc rsreg/lackfit results are given in Table 2.
Table 1: Summary of linear least squares regression results.
Lab
A
B
C
D
E
F
G
H
DB spikes (4
slope
8126
6541
6554
8538
6936
1523
1630
9695
Albany)
95% Cl
{+-'-)
1760
971
691
1515
3632
1288
610
2645
sig slope
p-value
0.0099
0.0025
0.0007
0.0049
0.1288
0.3027
0.0559
0.0215
r2
0.8421
0.9190
0.9573
0.8881
0.4769
0.2588
0.6404
0.7705
USGS spikes
(2001 dusts)
slope 95% Cl
(+/-)
124500
113300
52380
91340
74240
46370
7750
49510
488
23460
5914
58230
49810
14020
4607
21790
sig slope
p-value
0.0025
0.0085
0.0030
0.1918
0.2104
0.0298
0.1678
0.0855
r2
1.0000
0.8537
0.9632
0.3608
0.3571
0.7321
0,4144
0.5634
note: n =3
Table 2. Summary of "lack of fit" tests
Lab
A
B
C
D
E
F
G
H
AsB.C.andD
A,B,C:DandH
r2
linear
0.8421
0.9190
0.9573
0.8881
0.4769
0.2588
0.6404
0.7705
0,7550
0.7027
r2
quad
0.0590
0.0564
0.0001
0.0706
0.0197
0.0445
0.0069
0.1455
0.0032
0.0023
p-value
linear
0.0149
0.0018
0.0038
0.0040
0.1904
0.3386
0.1018
0,0135
0.0001
0.0001
pvalue
quad
0.2732
0.0791
0.9441
0.1087
0.7547
0.6913
0.8240
0.1070
0.6038
06473
Based on these summaries the linear model is appropriate. Laboratories A and B
demonstrate excellent performance across the board: each has r2 values > 0.80, significant
positive slopes with p < 0.05, and slopes with the expected magnitude. We caution that
Lab A only has three points for the USGS spike results (yellow highlights, Table 1).
Fields highlighted in blue indicate potential problem areas. If only the 4 Albany spike
series are considered, then Labs C and D can be added to the preferred performer group.
This is reasonable because the range covered here is more likely to reflect the range of
concern for unknown samples. Although the Lab H results demonstrate a lower r2 value
and a larger 95% Cl for slope, this is caused by a single outlying point. As such, there is
no reason to exclude Lab H from the analysis. Because Labs E, F, and G fail in more than
one category in both spiked data sets, they are not included in the remaining study
analysis.
63
-------
From the above discussion, we can construct two groups of laboratories based on
the estimated validity of their results: "Best", consisting of Labs A, B, C, and D and
"very good" consisting of the best group plus Lab H. In the following series of figures,
the individual and composite linear regression results for the groups are demonstrated
graphically.
120000 -
100000 -
80000 -
60000 -
40000 -
20000 -
Spiked Samples - 4 Albany
"Best" Group
o LabB
D Lab A
A LabC
» LabD
0 1
34567
% spike level
10 11
64
-------
Spiked Samples - 4 Albany
"Best" Group Combined
(Labs A, B, C, and D)
120000 -
100000 -
Slope: 7440+-903
r2: 0.7550
80000 -
60000 -
40000 -
20000 -
0123456
% spike level
8 9 10 11
120000 -
100000 -
80000 -
60000 -
40000 -
20000 -
O LabB
D Lab A
A Lab C
» LabD
* LabH
Spiked samples - 4 Albany
"Very Good" Group
(Labs A, B, C, D and H)
4567
% spike level
8 9 10 11
65
-------
120000 -
100000 -
Spiked samples - 4 Albany
"Very Good" Group Combined
(Labs A, B, C, D and H)
Slope: 7891 +-970
r2: 0.7027
80000 -
60000 -
40000 -
20000 -
0123456
% spike level
8 9 10 11
120000 -
100000 -
D LabF
O Lab G
O LabE
Spiked samples - 4 Albany
"Outlying" Group
(Labs E, F and G)
80000 -
60000 -
40000 -
20000 -
1 2
456
% spike level
66
8 9 10 11
-------
Spiked samples - 4 Albany
"Outlying" Group Combined
(Labs E, F and G)
120000 -
100000 -
80000 -
60000 -
40000 -
20000 -
Slope: 3363+-1402
r2: 0.2644
0 1
4567
% spike level
1IT 11
Precision: Up until this point, validity has been assessed only with those samples for
which there is some prior knowledge of content. For assessing precision, however, one
can use all of the samples (including unknowns) because each laboratory received
aliquots of the same set of 32 samples. Furthermore, the sample structure is such that
these 32 samples are comprised of 16 paired samples allowing within laboratory
precision estimates as well. Although there are a number of statistical options for
proceeding, an analysis of variance (ANOVA) and intra-class correlation coefficients
(ICC) are pragmatic for these circumstances as samples and laboratories are used in
groups. Preliminary analyses of "within" and "among" laboratory results indicate that the
underlying distributions (considering all 32 sample results) are not normal based on the
Shapiro-Wilk (S-W) test, and that natural log transformation of the data should used to
perform analysis of variance. The only exception is the USGS data set where only three
pairs of samples are reported and thus the natural space numbers did not require
transformation. Table 3 shows the results for the ICC analyses within laboratories, and
also for the groups (Labs A, B, C, D) and (Labs A, B, C, D, H) aggregated. The variance
components and p-values for the S-W normality test are also given. The lower part of
Table 3 gives the aggregated results for the background samples only; Laboratories A and
C did not contribute to these statistics but it is expected that they would perform
similarly.
67
-------
Table 3. Summary statistics for intra-class correlation coefficients.
All available Pairs
IntraClass Correlation Calculations: from SAS proc mixed
Log Space data
Lab
A
Aft
B
C
D
E
G
H
A,B,C,andD
A.B.C.DandH
A natural space
All NYC Backgrou
n obs
6
6
32
10
31
32
32
32
32
80
112
Int
7.6570E-01
1.3310E+09
2.7493E+00
1.9B70E+00
1.3965E+OQ
5.5988E+QQ
2.8559E+OQ
1.Q962E+00
2.0647E+OQ
2.0491E+00
2.0B20E+DQ
Res
5.491QE-01
2.1025E40S
1.7470E-01
7.3050E-01
2.7S6QE-Q1
1.7150E+00
1.3172E+00
6.7860E-01
3.2810E-Q1
3.0250E-01
2.6740E-01
ICC
0,5824
0.8S36
0.9403
0.7312
0,6337
0,7655
0,66-44
0.6176
0.8629
0,8714
0.8787
S-W p-value
0.2786
0,7679
0.'059
0_£p_42
0,8847
0.6012
0.6057
0.6571
0.1571
0.3296
0.1540
id Pairs
IntraClass Correlation Calculations; from SAS proc mixed
Log Space data
Lab*
A.B.C.andD
A,B,C,DandH
n obs
36
54
Int
6.4570E-01
6.9800E-01
Res
3.5070E-01
3.4290E-01
ICC
0.6480
0.6706
S-W p-value
0,2657
0.2571
*Laboratories A and C did not report paired New York City background data
From this exercise, we see that all of the individual laboratories demonstrate reasonable
ICCs (generally above 0.6). Furthermore, the laboratory groups chosen to demonstrate
good validity show ICCs greater than 0.87 when all data are considered. When only the
New York City background samples are analyzed, the ICCs are somewhat lower. These
results can be interpreted to mean that about 35% of the variance is attributable to
variability in the pooled laboratory analyses, and the remainder to true differences among
the background samples.
As a further assessment of inter-laboratory precision, the between laboratory ANOVA
shows no reason to reject the null hypothesis (Ho = no difference, in natural log space)
among Laboratories A, B, D, and H. Laboratory C was left out of this analysis because
they reported no background data at all.
68
-------
Evaluation of Unknown Samples
In the previous section we qualified a group of laboratories for measurement of
unknowns based on spiked samples (validity), and comparative precision measures based
on ICC and ANOVA. We now assume that these laboratories are statistically similar and
combine their spike results into a single response graph. Based on these results, we
calculate 95% confidence intervals and 95% prediction bands as illustrated in the figures
below.
140000
to
Spiked samples - 4 Albany
Labs A, B, C, D, and H combined
2 3 4 5 6 7 8 9 10 11 12
140000-
120000-
E 100000-
CO
in
i_
-------
The major effort here is to estimate the performance of the aggregate laboratory group
(A, B, C, D, and H) with respect to the group of samples from the greater New York City
area designated as "background" or "non-WTC impacted". The composite behavior of
these samples is illustrated below with respect to the analytical laboratories. The graph
indicates no consistent (high or low) percent bias from the cross laboratory means. This
confirms the conjecture made earlier that these laboratories are statistically similar. We
caution that Laboratory C did not provide any background data at all and could not be
directly included here, however, it is assumed that it would behave like the others.
Sample # vs percent difference by labs
(Slag Wool Fibers per gram dust)
NYC Background only
Lab A
LabB
LabC
LabH
1 234567
8 9 10 11 12 13 14 15 16 17 18
sample #
The next step is to assess how an individual (presumably unknown) dust sample assay
relates to the amount of spiked 4 Albany dust percentage. Given the graph and underlying
statistics of the above figure entitled "Spiked Samples - 4 Albany, Labs A, B, C, D and H
Combined), one can calculate the x-value in % spiked 4 Albany equivalent and the 95%
confidence interval for the prediction for any unknown sample measurement from any
laboratory. This is essentially the use of the prediction band graph above in reverse. As
such the prediction of "x" and the CI take the following form:
Xpredicted = (Ybar - a)/(t>)
CI = Xpredlcted ± [t(RSE)/b] * (1/m + 1/n + [(Ybar - ybar)2/(b2(n-l)sx2)}1/2
70
-------
where Ybar is the mean laboratory measurement, a and b are the intercept and slope of the
regression, t is the critical t-value for n-2 degrees of freedom, RSE is the residual
standard error, m is the # of replicate measurements, n is the number of calibration points,
ybaris the mean of the regression y data, and sx is the standard deviation of the x values of
the regression data.
The reported slag wool results for the background samples can now be interpreted. Table
5 shows the results for each background sample measurement across all participating
laboratories as a prediction of the percent equivalent 4 Albany spike level and half of the
95% confidence interval associated with the measurement. There are a total of 63
measurement results in the table.
Table 5: Results for each background sample across all participating laboratories as a
prediction of the percent equivalent 4 Albany spike level and ± 95% confidence interval.
Lab* A B D H
Sample Percent a+_ Percent CI +. Percent CI+- Percent CI+-
#
1 -1.35 5.43
2
3 -1.35 5.43
4
5 0.72 5.29
6
7 -0.61 5.38
8
9 0.20 5.33
10
25 7.95 5.21
26
27 0.99 5.28
28
29 0.02 5.34
30
31 3.88 5.18
32
*Laboratory C was left out of this analysis because they reported no background data.
We note that negative entries above are only statistical constructs. Of the 63 background
measurements in this table, 7 (or about 11%) exceed the 4 Albany 5% spike level; 2 of
the 63 measurements exceed the 10% 4 Albany spike level. If the upper confidence
limits are considered, 42 out of 63 (67%) exceed the 5% spike level and 7 of 63 (11%)
71
-0.89
-1.03
-0.92
-1.01
-0.43
-0.38
2.98
0.01
1.05
-0.88
5.75
3.87
0.93
0.73
-0.44
-0.88
2.29
4.79
5.40
5.41
5.40
5.41
5.37
5.36
5.20
5.34
5.28
5.40
5.17
5.18
5.28
5.29
5.37
5.40
5.22
5.17
-1.35
-0.47
0.14
-0.14
1.05
1.00
1.99
0.94
1.02
2.68
10.28
3.80
0.71
2.68
0.47
1.21
7.07
4.41
5.43
5.37
5.33
5.35
5.28
5.28
5.23
5.28
5.28
5.21
5.31
5.18
5.30
5.21
5.31
5.27
5.18
5.17
-0.69
-0.66
-0.71
-0.79
-0.68
3.06
0.68
0.98
0.84
-0.50
2.35
8.05
2.23
1.88
-0.49
3.86
14.25
6.18
5.38
5.38
5.39
5.39
5.38
5.20
5.30
5.28
5.29
5.37
5.22
5.21
5.22
5.24
5.37
5.18
5.63
5.17
-------
exceed the 10% spike level. For instance for sample 25 at lab A the percent equivalent of
the fiber measurement is 7.95% and the upper confidence limit is 7.95% + 5.21% =
13.16%.
As a further exercise, we calculated the same statistics for data within only one laboratory
(choosing Laboratory B as the example); these results do not include scatter in the
regression from the other qualified laboratories. Here we find some improvement: we see
3 of 18 values (16.7%) exceed the 5% 4 Albany dust level and 0 of 18 values exceed the
10% 4 Albany dust level. For the upper confidence levels, 6 of 18 values exceed the 5%
4 Albany dust level and 1 of 18 values exceed the 10% 4 Albany dust level.
Conclusions
The conclusions are based solely on the analytical data provided from the laboratory test
and a few analyses of the 100% WTC spike samples. From validity estimates based on
expected slopes and data scatter of WTC spiked samples, five of eight laboratories (A, B,
C, D, and H) were used for further analysis. Intra-class correlation coefficients (with
natural log transformation) for individual labs and for the group of five demonstrate
similar and reasonable values (>0.7) when all available data are considered. One-way
ANOVA analysis of Laboratories A, B, D, and H results provides no evidence to reject
the null hypothesis (that the results are from the same distribution). Laboratory C was
not included here because of insufficient reported data but, based on spike sample
statistics, it is likely that that they too would fall into this category.
Under the practical constraints that the five laboratories are used at random with one
analysis per unknown sample, we cannot expect statistical discrimination at the 1% or 5%
4 Albany spike equivalent level because the upper 95% prediction bounds exceeds the
5% spike equivalent level across the board. Reasonable discrimination is possible at the
10% 4 Albany spike equivalent level because the lower bound on 10% equivalent
measurements is approximately equal to the mean at 5% Albany spike equivalent level.
72
------- |