United States EPA/600/R-03/018
Environmental Protection December, 2005
Agency
Final report on:
MAPPING THE SPATIAL EXTENT OF GROUND DUST AND DEBRIS FROM
THE COLLAPSE OF THE WORLD TRADE CENTER BUILDINGS
David B. Jennings, David J. Williams and Donald Garofalo
United States Environmental Protection Agency
Office of Research and Development
National Exposure Research Laboratory
Environmental Sciences Division
Landscape Ecology Branch
Environmental Photographic Interpretation Center
Reston, Virginia 20192
-------
NOTICE
This document is a Final report. It has been peer-reviewed and comments reconciled.
Mention of trade names and/or commercial products does not constitute endorsement
or recommendation for use.
-------
ABSTRACT
This report presents the results of an analysis into the geographical extent of the
ground dust/debris field produced by the collapse of the buildings at the World Trade
Center (WTC) on September 11, 2001. The study focused on an area within an
approximate eight mile radius from the World Trade Center. The study period ranged
from September 11, 2001 through September 13, 2001. This temporal limit is due to
an approximate two inch precipitation event on the morning of September 14, 2001.
Various remote sensing imaging sources (aerial photographic and satellite image data)
and analytical techniques (qualitative interpretive analysis and quantitative image
processing analysis) were utilized in this study.
Results include:
High-spatial-resolution aerial photographs from September 11, 2001 and
September 13, 2001 show distinct primary and secondary deposition along roadways,
parking lots and other ground areas in lower Manhattan that extend as far north as
Canal Street on the September 13 photographs. In the small area of Brooklyn covered
by the September 13 photographs, possible dust is observed on pier areas adjacent the
East River and directly south-southeast of the WTC area. An excavated area and
multiple mounds of material are also observed on the piers in the general vicinity of the
possible dust. The area of coverage for the high-spatial-resolution aerial photographs
was generally limited to the lower Manhattan area during the study period. This
limitation prevented a determination of ground dust/debris boundaries over the wider
-------
study area
• On one-meter IKONOS-1 satellite imagery from September 12, 2001
ground dust/debris was observed in lower Manhattan. However, the extreme oblique
perspective provided by the image precluded a boundary determination for the lower
Manhattan area. For those areas outside of lower Manhattan, ground dust/debris could
not be ascertained.
• Multi-temporal Landsat 7 ETM+ multispectral images (September 12,
2001 and August 27, 2001) and EO-1 Hyperion satellite hyperspectral data were used
to assess ground dust/debris over a wide geographic area.
A qualitative assessment of spatial patterns of high reflectance change, derived
from Landsat 7 temporal image ratio data, provided three distinct spatial patterns. One
generally coincided with the ground boundaries derived from September 11, 2001 aerial
photographs in lower Manhattan, south of Chambers Street. A second was related to
the WTC plume and a third was related to the eastern edges of shorelines (land/water
interface). Lower reflectance change, although prevalent throughout the study area,
showed no coherent spatial pattern for ground dust/debris.
The spatial patterns derived from spectral signature mapping of EO-1 Hyperion
data also showed a spatial coincidence with the ground dust/debris boundaries derived
from September 11, 2001 aerial photographs in lower Manhattan, south of Chambers
Street. An additional pattern was noted along the west side of Manhattan proceeding
from the WTC area in the south towards Central Park to the north.
iv
-------
The maps produced from the Landsat and Hyperion data do not assure the
presence or absence of ground dust/debris for any given pixel, but rather illustrate
spatial relationships from which ground dust/debris, per pixel, may be qualitatively
assessed. The spatial resolution provided by these sensors (30 meters) was not
sufficient for interpretive mapping of ground dust/debris boundaries, such as those
delineated using the high resolution aerial photographs.
-------
CONTENTS Page
NOTICE ii
Abstract iii
Introduction 1
General 4
Analysis 9
September 11, 2001 9
NYPD Aerial Photographs
Data 9
Methods 10
Results 11
September 12, 2001 13
Landsat 7 ETM+
Data 13
Methods 14
Results 16
EO-1 Hyperion
Data 18
Methods 18
Results 19
IKONOS-1
Data 20
Methods 20
Results 20
September 13, 2001 21
KAS Vertical Aerial Photographs
Data 21
Methods 21
Results 22
REFERENCES 25
TABLES
Table 1 Summary Statistics, per band, for Image Ratio brightness change values
derived from multi-date Landsat 7 ETM+ imagery 17
Table 2 Listing of Digital Imagery and Aerial Photographs used for Analysis 27
APPENDICES
Appendix A, Figures. Mapping the Spatial Extent of Ground Debris/Dust from the
Collapse of the World Trade Center Buildings A-1
Appendix B, Local Climatological Data, New York C. Park, NY (NYC), SEP/2001 ....B-1
Appendix C, Listing of Standard Operating Procedures (SOPs) C-1
vi
-------
INTRODUCTION
This report presents the results of an analysis into the geographical extent of the
ground dust/debris deposition produced by the collapse of the buildings at the World
Trade Center (WTC) on September 11, 2001. Specifically, the United States
Environmental Protection Agency's (USEPA) Office of Research and Development
(ORD) was tasked by the USEPA Region 2 Response and Recovery Operations Team
to evaluate remote sensing data to determine the geographical ground distribution of
the dust/debris produced by the collapse of the buildings at the WTC in New York City,
New York. The purpose of this mapping effort was to provide geographical boundaries
in support of USEPA Region 2 response and recovery operations (Final WTC Test and
Clean Program Plan, 2005).
The analysis encompassed an approximate 256 square-mile area centered on
the WTC site (Figure 1). Remote sensing data collected September 11, 2001 through
September 13, 2001 were utilized to investigate the spatial extent of the ground
dust/debris field. This temporal limit is due to an approximate two inch precipitation
event on the morning of September 14, 2001 (Appendix B; NCDC, 2001) that altered
the extent of the original ground dust/debris field.
Remote sensing data analyzed for this study included:
1) Scanned, natural color, oblique aerial photographs collected from hand-held
cameras on September 11, 2001 by New York Police Department (NYPD)
photographers on board NYPD helicopters (NYPD, 2002).
-------
2) Digital imagery collected from various satellite sensor platforms on September
12, 2001 to include IKONOS-1, Landsat 7 ETM+ and EO-1 Hyperion.
Additionally, Landsat 7 ETM+ image data collected on August 27, 2001, were
acquired as a change detection vehicle to provide a pre and post September 11, 2001
comparison of landscape condition.
3) Analog, black and white (B+W) panchromatic, vertical aerial photographs
collected on September 13, 2001 by Keystone Aerial Survey (KAS) using a standard 9"
x 9" frame aerial mapping camera.
The aerial photographic coverages from September 11, 2001 and September 13,
2001 provided the means for high-resolution interpretive mapping but were limited to the
general area of lower Manhattan. Satellite imagery from September 12, 2001 provided
a means of wide-area coverage of the area of interest for this report. In particular,
Landsat 7 was included to meet the specific request of analysis over the entire two-
hundred and fifty-six (256) square miles.
For this study, the dust/debris extent was defined by those boundaries produced
from the collapse of the WTC buildings and demarcated as light/dark boundaries on the
ground. For example, as determined from the aerial photographs, boundaries were
generally distinct along the road network due to the contrast between the dark-toned
road surfaces and the light-toned dust/debris. The dust/debris covered the road
surfaces to the point where road and cross-walk lines were obscured. On the opposite
side of the boundary some faint dust may have been present but not in a sufficient
quantity to obscure surface details such as road lines or cross-walks (Figure 2B). In
-------
contrast, it proved difficult to distinguish dust/debris boundaries atop light-toned roofs
due to a lack of tonal difference. Given the possibility of gradations of dust/debris
deposition on the landscape, areas of light dust/debris deposition may not be
detectable from the remote sensing data included in this report and therefore it is
possible that dust/debris may extend beyond the boundaries as delineated in this report.
In contrast, boundaries could not be interpretively mapped from medium spatial
resolution satellite data (Landsat 7, Hyperion). Mapping results from these raster data
were derived from image processing algorithms based on 1) pixel reflectance change
per multi-date Landsat images, and 2) mapping of Hyperion data to endmember spectra
from lower Manhattan. While these maps may have an association with dust/debris,
they do not assure the physical presence/absence of ground dust/debris for any given
pixel.
The report addresses the spatial extent of ground dust/debris deposition as an
aggregate (paper, pulverized concrete and wall board, larger building materials, etc.)
and is concerned with material that would change the tonal quality and spectral
signature of ground features in the area-of-interest. It does not analyze the chemical or
molecular composition of the ground dust/debris. Further, the analysis assumes no
spatial correlation between ground dust/debris boundaries, as mapped in this report,
and the dispersal of dust/debris into those areas above ground level.
The U.S. Environmental Protection Agency (EPA), Office of Research and
Development, Environmental Sciences Division, Landscape Ecology Branch,
Environmental Photographic Interpretation Center (EPIC) in Reston, Virginia prepared
-------
this report for the USEPA Region 2 Response and Recovery Operations Team, New
York, New York. The Figures in this report are located in Appendix A. Sources for the
aerial photographs and digital imagery, evaluated for this report, are listed in Table 1.
SPOT-4 and NYWCAP imagery, collected on September 12, 2001, were reviewed for
this report, however, they did not provide additional detail on ground dust/debris and
therefore were not included in this analysis.
Note: the New York State Office for Technology (NYSOFT) began contract
airborne overflight operations with the Keystone Aerial Survey overflight on September
13, 2001. No known airborne overflight collection operations occurred on September
14, 2001. Airborne operations for daily collection of 1-foot digital camera imagery as
well as Light Detection and Ranging (LIDAR) data of the WTC began on September 15,
2001.
GENERAL
This report was prepared using a standard methodology that included the
following procedures:
• data identification and acquisition,
• imagery analysis and mapping, and
• graphics and text preparation.
Data identification and acquisition included a search of government and
commercial sources of imagery encompassing the study area. Publicly available
airborne and satellite imagery collected over the WTC for the time period September 11,
2001 through September 30, 2001 were identified and data matching the temporal
-------
period September 11, 2001 through September 13, 2001 were acquired. United States
Geological Survey (USGS) topographic maps as well as a September 15, 2001
IKONOS-1 one-meter rectified image were acquired and used as reference data to
provide geographic context for the mapping of the dust/debris boundaries.
Analysis and mapping techniques applied to the various data were unique to
each imagery type and included, 1) monoscopic visual interpretation of digital images
viewed within a CIS environment, 2) stereoscopic visual interpretation of analog
diapositive photographs viewed on a backlit light table, and 3) digital image processing
utilizing pixel reflectance values. The individual data, methods and results of the
mapping of the ground dust/debris are specifically detailed in the Analysis section by
date and separate data type.
-------
The process of visual interpretive analysis involves the examination and
comparison of numerous components of the remotely sensed data. These components
include shadow, tone, color, texture, shape, size, pattern, and context of individual
elements of a photograph. The image interpreter identifies objects, features, and
"signatures" associated with specific environmental conditions or events. In a visual
interpretive context, "signature" refers to a combination of components or characteristics
that indicate a specific object, condition, or pattern of environmental relevance. The
professional and academic training, image interpretation experience gained through
repetitive observations of similar features or activities, deductive logic of the analysts as
well as collateral information are important factors employed in interpretive analysis.
When utilizing analog overlapping photographs, the analyst may utilize a stereoscope to
view a three-dimensional representation of the study area, thus providing the ability to
view the three dimensional relationships of landscape features.
Digital image processing is the computerized application of algorithms to digital
imagery. The algorithms utilize the reflectance value and/or the spatial context of a
pixel to automate the process of interpretation and classification. However, while the
maps produced from these methods may have an association with dust/debris, there is
no assurance that the results are specifically due to the physical presence of ground
dust/debris. The ENVI (RSI, 2004) and ERDAS IMAGINE (Leica Geosystems, 2004)
image processing software packages as well as the ArcView CIS (ESRI, 2004)
software package were utilized for the processing and analysis of digital imagery used
in this report.
-------
Mapped boundaries are presented on transparent overlays, attached to the
figures, and discussed in the associated text. All mapped boundaries are approximate
and will be identified in the graphics and text according to the analyst's degree of
confidence in the evidence. A qualitative distinction is made between certain, probable,
and possible identifications. When the analyst believes the identification is
unmistakable (certain), no qualifier is used. Probable is used when a limited number of
discernible characteristics allow the analyst to be reasonably sure of a particular
identification. Possible is used when only a few characteristics are discernible, and the
analyst can only infer an identification.
The satellite images and the image mosaic produced from the September 13
aerial photographs were rectified, prior to acquisition for this report, to the Universal
Transverse Mercator (UTM) projection, Zone 18N, NAD 83 datum, units=meters. When
necessary for overlay analysis and reporting, un-rectified data were rectified to the
above projection. Rectification was completed via an image-to-image registration
technique utilizing previously rectified images for reference control points. The rectified
images presented in this report have not been evaluated for national map accuracy
standards (NMAS).
The image figures in this report have been reproduced from either the original
digital data (as acquired for this report) or from digital scans of analog data produced
"in-house" on a UMAX Mirage II digital scanner at a scan resolution of approximately
800 dots per inch (dpi). All figures in the report have been printed at 720dpi. Although
the reproductions allow effective display of the imagery and interpretive annotations,
-------
they may have lower resolution than the original imagery. Therefore, environmental
features identified from the original data and described in the text may not be as clearly
discernible on the prints in this report. The specific images included in this report were
chosen based on providing the reader with the best possible spatial context regarding
the in situ environmental conditions, however boundaries were analyzed and mapped
from many more photographs/images than were illustrated in this report. See Table 1
for a listing of photographs and images used for this report.
Remote sensing and geospatial processing Quality Assurance Standard
Operating Procedures (SOPs), as set forth in the Master Quality Assurance Project Plan
(MQAPP) under Remote Sensing Support Services Contract (RSSSC) 68-D-00-267,
were implemented for this report (Appendix C). Additionally, SOPs as set forth in the
Image processing and CIS software documentation were implemented in the
processing of the data. Due to the lack of ground truth data for the study area, a
quantitative error assessment of the mapping results could not be provided. However,
multiple remote sensing analysts have provided input into the analysis, results and
review of this study.
Mention of trade names or commercial products does not constitute endorsement
or recommendation for use.
8
-------
Analysis
This section is listed by date with data, methods and results reported for each
data type.
September 11, 2001
NYPD Aerial Photographs
DATA
Data from September 11, 2001 were provided by USEPA Region 2 in two
formats:
1) One-hundred and ninety-seven (197) digital aerial images in JPG compressed
image format were provided by USEPA Region 2. These data were produced
from hand-held photographic images collected by NYPD photographers onboard
NYPD helicopters on September 11, 2001. Metadata, such as camera type,
photograph time-tags, sequence of collection, photographic scale, photographers
name and scan resolution of the digital images were not provided, therefore the
specific parameters of these data cannot be conclusively determined. A relative
temporal reference frame was established based on the time of the collapse of
WTC buildings 2, 1 and 7 (approximately 9:50 am, 10:30 a.m. and 5:30 p.m.
respectively on September 11, 2001; FEMA, 2002). The identification and
condition of the WTC buildings, in the images, provided a general collection time
sequence. The images were collected at both low and high oblique perspectives
with the image foreground generally focused on lower Manhattan. This collection
characteristic generally limited the interpretable portion of the imagery (and the
-------
ground dust/debris boundary determination) to the lower Manhattan area. A
spatial index of the NYPD JPG images was not produced for this study.
2) NYPD oblique aerial photographs reproduced in the publicly available book
Above Hallowed Ground (NYPD, 2002). These photographs were also collected
by NYPD photographers onboard NYPD helicopters, however, these data were
not available to the USEPA as individual photographs or scanned images for this
report. Therefore, photographs from the book that were deemed necessary for
inclusion in this report were scanned at 800 dots per inch (dpi), converted to a
Tagged Image File Format (TIFF) format and utilized as figures in this report.
Despite the lack of attributable metadata, the NYPD photographs were the "best
available" aerial image data source collected on September 11, 2001. The image scale
of the data is variable due to the non-consistent flying height of the platform(s) as well
as the oblique nature of the photographs.
METHODS
The ground dust/debris boundaries were mapped by employing the combined
means of visual interpretive mapping techniques (Avery and Berlin, 2002; ASPRS,
1997) and digital mapping methods within a CIS. On the computer screen, a "dual-
view" mode was employed whereby the un-rectified aerial images were visually
analyzed in one view to determine the general placement of the dust/debris boundaries
while, in a second view, "heads-up" digital collection techniques were employed to map
the dust/debris vector boundary using the rectified mosaic in Figure 3 as a base map.
Visible boundaries delineated from September 11, 2001 data were based
10
-------
primarily on tonal variation (light/dark feature demarcation) along paved roadways and
parking lots as well as change analysis of pre and post collapse images of WTC
buildings 1 and 2 respectively.
RESULTS
Figures 2A through 2E are centered on the intersection of West street and
Warren street in lower Manhattan and present examples of clearly visible ground
boundaries produced by the WTC building collapse clouds. The figures are presented
here in order to clarify the definition of a dust/debris boundary and to illustrate how the
analysis and mapping proceeded in this report. Boundaries were generally distinct
along the road network due to the contrast between the dark-toned paved road surfaces
and the light-toned dust/debris. The dust/debris covered the road surfaces to the point
where road and cross-walk lines were obscured. On the opposite side of the boundary
some faint dust may have been present but not in a sufficient quantity to obscure
surface details such as road lines or cross-walks (Figure 2B). It proved difficult to
distinguish the dust/debris boundaries from light-toned rooftops due to a lack of tonal
difference. As stated in the introduction, boundaries derived from visual interpretive
analysis were based on clearly demarcated areas of contrast. Further, a distinction was
observed with respect to the ground dust/debris directly deposited by the collapse
clouds (WTC buildings 1 and 2 as shown on Figure 2A-2E) and secondary ground
dust/debris deposition that may have been due to post-collapse dispersal vectors such
as transportation and wind. This secondary condition is evident in Figure 3, along West
Street, where vehicular traffic has likely redistributed dust north of the original boundary
11
-------
(Stuyvesant School Bridge, Figure 2E).
Figure 3 is a rectified mosaic of the photograph displayed on pages 48 and 49 in
Above Hallowed Ground (NYPD, 2002). The photograph was collected at a low-oblique
perspective and provided a base map from which to present the boundary in lower
Manhattan on a single image. The ground dust/debris boundary was derived from the
analysis of multiple images, most of which were not presented in this report, and
reflected ground dust/debris boundary conditions between 10:30 am (post-collapse,
WTC buildings 1 and 2) and 5:30 pm (pre-collapse, WTC building 7) on September 11,
2001. The mosaic seam, observed in the center of Figure 3, is due to a lack of overlap
between the scanned images (pp. 48-49 in Above Hallowed Ground) and the oblique
characteristic of the aerial photograph.
No visible ground dust/debris boundaries could be determined for the limited
areas of Brooklyn, adjacent the East River, where coverage was available.
Results (Figure 3) show the northern limit (between North End Avenue and
Centre Street) of the ground dust/debris boundary at approximately Chambers Street.
However, variations in the northern boundary are observed 1) between Reade and
Duane streets along Greenwich Street, Hudson Street and West Broadway, and 2)
between Duane and Thomas Streets along Church Street. To the northeast, the
boundary appears to extend to the east from the vicinity of City Hall along a line
approximately parallel and south of the Brooklyn Bridge (Frankfort Street) as it meets
the Franklin Delano Roosevelt expressway (FDR). The southern boundary of the
ground debris field appears to extend to Battery Park and to the surrounding southern
12
-------
limits of Manhattan. The western boundary generally extends to the Hudson River.
The mapping of the eastern and southeastern boundaries (south of the
Brooklyn Bridge and east of Pearl and Water Streets) was problematic due to the limited
coverage of the area provided by the September 11, 2001 aerial photographs and the
plume emanating from "Ground Zero" which obscured the area. However, a dust/debris
boundary was delineated, for this day, based on the analysis of September 13, 2001
aerial photographs, where a "blanket" of ground dust/debris deposition - characterized
by light toned dust that completely covered the roadways and obscured white road and
cross-walk lines - was observed in the eastern and southeastern portions of lower
Manhattan. The deposition characteristics in this area were consistent with areas of
lower Manhattan that were impacted by primary deposition from the WTC building
collapses on September 11, 2001. Therefore, it is likely that the boundary, south of the
Brooklyn Bridge did extend up to the FDR and possibly to the raised FDR roadway as
well as the adjacent piers along the East River on September 11, 2001. However,
because of the limited coverage, no qualitative distinction of the ground dust/debris in
this area was attempted on this date. See the results for September 13, 2001 for a
more comprehensive assessment of ground dust/debris in this area.
September 12, 2001
Landsat 7 ETM+
DATA
13
-------
Landsat 7 ETM+ satellite data (path14/row 32) of September 12, 2001 included six
bands of multi-spectral data (bands 1-5, 7) ranging from the visible to mid-infrared at
28.5-meter spatial resolution and one band of panchromatic B+Wdata (band 8) at 14.5-
meter spatial resolution (USGS, 2004a). Additionally a spatially coincident Landsat 7
ETM+ image from August, 27, 2001 was acquired and the identical bands were
processed for analysis. These pre and post WTC building collapse datasets were
processed in order to determine patterns of reflectance change that would be consistent
with the deposition of ground dust/debris. These data also provided the best means for
analyzing ground dust/debris across the entire 256 mile2 study area (Figure 4, Figure 5).
METHODS
ENVI software was used to process the Landsat 7 ETM+ data. Steps involved in
the data processing chain for each date of data consisted of:
1) A geometric and radiometric correction to Level 1G was applied, prior to
acquisition by the USEPA, by the source vendor EROS Data Center (USGS, 2004b).
2) A top-of-atmosphere correction (TOA) was applied to transform the pixel
digital numbers (DN) to exoatmospheric reflectance values (Lillesand et al., 2004) and
to provide a scene-to-scene reflectance calibration.
3) A 256 mile2 subset centered on the World Trade Center site was
clipped from the original scenes.
4) A Neighborhood Analysis consisting of a 3x 3 pixel, mean spatial filter
(Lillesand et al., 2004) was applied to each pixel to compensate for an approximate one
pixel registration offset between the two dates of imagery. A low-pass, 3x3 mean filter
14
-------
was used in order to maximize large-area patterns in brightness and to aid in the
mapping of large area contiguous boundaries associated with the dispersal of the
ground dust/debris. All subsequent analysis was performed using the spatially filtered
pixel reflectance values.
5) Temporal image ratio change detection methods (Lillesand et al., 2004) were
implemented as a means for identifying those pixels that increased in reflectance
between the two dates of imagery. Temporal image ratioing is a digital pre-
classification change detection technique that compares coincident, per band pixel
reflectance values of the two rectified images (Jensen, 1996). In this study, the
temporal image ratio approach compared the after image (September 12, T2) with the
before image (August 27, T^ to compute a per-pixel reflectance change ratio (T2/Ti).
Ratio values > 1 represented positive reflectance change over time.
The temporal image ratio algorithm was applied to each individual band to create
seven separate change images. Although image ratio data for all bands were produced
and analyzed, for purposes of graphic reporting, only image ratio data from band 7 and
band 8 were illustrated for this report. Band 7 was the only band to approach an
increase in mean reflectance change between the two dates (0.98, Table 1), and its
spectral wavelength range (2.09 - 2.35 microns) was shown to have a strong
relationship with the WTC dust material (Clarke et al., 2001). Band 8 was included due
to the increase in spatial resolution (14.5 meters) which provided additional spatial
detail. Pixel values corresponding to no-change and decreases in reflectance were not
reported. While image ratio figures for bands 1-5 were not included here, their spatial
15
-------
results were similar to the patterns observed in bands 7-8 and discussed below.
RESULTS
The temporal image ratio results provided insight into the spatial pattern of
change over the whole study area and allowed general conclusions based on those
patterns. The ratio data do not assure the presence or absence of ground dust/debris,
for any given pixel, but rather denote those pixels that have greatly increased in
reflectance and thus may have an association with dust/debris deposition.
Figure 5 and Figure 6, based on results from band 8 and band 7, respectively,
utilized pixel values corresponding to a positive increase in image ratio data of > 2
standard deviations (SD). Data resident within this threshold (> 2 SD) represented
approximately 3% and 2% of the image ratio data in band 7 and 8 respectively and were
utilized to detect those areas most related to increases in reflectance. From Figures 5
and 6 the following spatial patterns were observed:
1) The area of lower Manhattan, bounded approximately by Chambers street to
the north, has substantially increased in reflectance between the two dates (Table 1).
The boundary is in general spatial agreement with the ground dust/debris boundary
mapped from the September 11, 2001 photographs (in orange on overlays). Table 1 is
a summary of image ratio statistics derived for bands 1-5, 7 and 8 and provides the
mean and SD for each band over the entire study area as well as a mean value, per
band, for the area of lower Manhattan encompassed by the September 11 dust/debris
boundary as determined from the NYPD photographs. Noted are the approximate 50%
increases in the mean ratio values in lower Manhattan as compared to the mean values
16
-------
for the entire study area.
Table 1. Summary statistics, per band, for Image Ratio brightness change values
derived from multi-date Landsat 7 ETM+ imagery.
* total area-of-interest = 256 miles2 centered on the WTC.
** 9/11 boundary as mapped from NYPD photos.
Landsat 7 ETM+:
Band Number
Band 1
Band 2
Band 3
Band 4
Band 5
Band 7
Band 8
Mean image ratio
value for total
area-of-interest *
0.73
0.75
0.76
0.87
0.9
0.98
0.81
St. Dev.of image
ratio value for
total area-of-
interest
0.13
0.17
0.21
0.26
0.32
0.33
0.23
Mean image ratio value
for area encompassed
within the 9/11
boundary (Lower
Manhattan) **
1.01
1.14
1.27
1.33
1.39
1.54
1.28
2) high reflectance change is noted along the eastern shorelines of land/water
interfaces throughout the study area. This pattern is possibly related to an east/west
directional mis-registration of the two images which would have a particular effect of
distorting reflectance ratios related to north/south trending linear features.
3) a large area of increased reflectance is associated with the plume extending
from the WTC area to the southwest. The increase is probably due to reflectance from
the plume and not necessarily associated with a change in ground reflectance due to
deposition of dust/debris.
4) clouds and shadow, present on August 27 (Figure 4), dominate the northern
17
-------
portions of the study area as well as portions of Queens, Brooklyn and New Jersey and
were a source of large reflectance change. The reflectance areas related to the clouds
and shadows, in proximity to lower Manhattan and the WTC plume, are mapped on
Figure 5 and Figure 6 overlays. Due to the presence of clouds/shadows on August 27,
no determination of ground reflectance change could be ascertained in these areas on
September 12.
EO-1 Hyperion
DATA
The hyperspectral dataset from the EO-1 Hyperion sensor (USGS, 2004c)
consists of 220 separate spectral bands covering the wavelengths from 0.4 -2.5 microns
at a spatial-resolution of 30 meters. The Hyperion data covered an approximate 4.7
mile swath across the center of the study area (Figure 7).
METHODS
ENVI software was used to process the Hyperion data. Steps involved in the
data processing chain consisted of:
1) removing noisy and defective spectral bands, by inspection, which reduced the
dataset to 154 bands.
2) correcting for atmospheric effects by radiative transfer modeling using the
software ACORN (Atmospheric CORrection Now; AIG, 2004).
3) noise reduction using principle component analysis (PCA).
4) identifying candidate regions in the data that contained materials of interest
(in this case, ground dust/debris from the WTC collapse). Ground based spectral
18
-------
signatures obtained by the USGS (Clark et al., 2001) were used to create a library of
signatures (endmember spectra) for material identification and mapping purposes. This
library was compared to spectral signatures in the Hyperion dataset and areas (pixels)
matching entries in the library were used in the mapping algorithms. Spectral
signatures were obtained just north of the WTC site and were used in the material
mapping algorithms.
5) Areas matching these endmember spectra were mapped throughout the entire
image. The Mixture Tuned Matched Filter (MTMF; Boardman, 1998) was used for this
analysis due to the fact that the spectra of materials in cluttered urban areas are a
mixture of many other materials at the pixel resolution of the Hyperion sensor (30
meters). The spectral features of the WTC dust were similar enough to common
building material found in urban areas to defeat the spectral feature algorithms. This
problem, combined with the noise inherent in data obtained from space, made it difficult
to spectrally differentiate ground dust/debris from the surrounding urban background
material.
RESULTS
As with the Landsat 7 reflectance change results, we found a spatial coincidence
between the Hyperion results and the results derived from the September 11, 2001
aerial photographs in lower Manhattan south of Chambers Street (Figure 7, Figure 8).
An additional pattern was noted along the west side of Manhattan proceeding from the
WTC area in the south towards Central Park to the north.
The mapped areas do not assure the presence or absence of ground dust/debris
19
-------
for any given pixel, but rather indicate the likelihood that a Hyperion pixel is spectrally
similar in the signature library - here, ground dust/debris from the WTC building
collapses in lower Manhattan.
IKONOS-1
DATA
Rectified digital imagery data were collected by Space Imaging's IKONOS-1
satellite (Space Imaging Inc., 2004) at nominal four-meter multispectral spatial
resolution and nominal one-meter panchromatic spatial resolution. These data were
subsequently merged to produce a multi-spectral dataset at nominal one-meter
resolution. The imagery covered approximately 50% of the study area (Figure 9 ) and
included portions of Manhattan, Brooklyn, Staten Island and New Jersey. However, the
low collection angle of acquisition of the sensor (approximately 32 degrees) produced a
low-oblique image, rather than the standard near-vertical image, and served to degrade
the spatial resolution and spectral integrity of the data. In the lower Manhattan area this
caused extreme building layover/shadow on the imagery and obstructed
ground/roadway visibility.
METHODS
The imagery was analyzed, using visual interpretive methods, from within the
ArcView CIS environment.
RESULTS
Although a boundary map was not produced from the IKONOS-1 data, some
areas of dust/debris were visible in lower Manhattan. Probable light-toned dust/debris
20
-------
was visible along West Street extending north from the WTC area to approximately
Canal street. Additionally, ground dust/debris was visible at discrete points (Figure 10)
south of Canal Street along Church Street, along West Broadway, east of Centre Street
in the vicinity of the approaches to the Brooklyn Bridge and in a parking lot located
between Pearl and Water Streets to the south of the Brooklyn Bridge.
The distinctive light-toned demarcation of dust/debris, evident in lower
Manhattan, was not observed in areas outside of lower Manhattan.
September 13, 2001
KAS Vertical Aerial Photographs
DATA
Data from September 13, 2001 consisted of 78 B+Wfilm diapositives
(transparencies) of 9" x 9" vertical and oblique aerial photographs of the lower
Manhattan collected at a scale of 1:6,000 (KAS, 2004). Time of collection was
approximately 9:00a.m. eastern daylight time (edt). The areal coverage of the
photographs (Figure 11) was limited to lower Manhattan and a small portion of Brooklyn
west of Hicks Street and adjacent the East River.
METHODS
Visual stereoscopic interpretive techniques were employed using a Richards 500
zoom stereoscope backlit light table to determine the dust/debris boundaries.
Stereoscopic viewing provided the ability to visualize vertical as well as horizontal
spatial relationships of landscape features. As with the September 11, 2001 data,
21
-------
"heads-up" digital collection methods were implemented to map the boundaries.
Determination of boundaries generally relied upon the analysis of tonal and texture
variation, along with stereoscopic visualization, to delineate clearly defined light/dark
edges associated with light toned dust/debris on dark toned road pavements.
RESULTS
A probable change in the northern ground dust/debris boundary, in lower
Manhattan, occurred between September 11, 2001 and September 13, 2001. The
probable northern boundary on September 13 extended approximately to Canal Street
with other possible areas of dust extending beyond Canal Street (Figure 12). Multiple
possibilities exist for the apparent change in the northern dust/debris boundary. First,
WTC building 7 collapsed in the intervening time period which may have caused a wider
distribution of dust than that delineated for the September 11, 2001 dust/debris map.
Second, a transportation vector may have brought about a redistribution of dust. This
scenario is most evident along West Street which was apparently an ingress/egress
route for WTC operations. And third, an atmospheric vector due to a shift in wind
direction to the north-northwest on September 13, 2001 (NCDC, 2003). The new lower
Manhattan boundary area, existing between Chambers and Canal Streets, is the outer
limit of an irregular and patchy distribution of dust/debris that appears to have
accumulated along roadways and curbs in high traffic zones. Throughout this area,
vehicle tracks are visible in the dust on the roads. This is particularly evident along
Canal Street in proximity to a probable heavy equipment staging area (Figure 13).
22
-------
A possible area of ground dust/debris is observed in the Brooklyn pier areas,
adjacent the East River, directly east of Governors Island and south-southeast of the
WTC (Figures 12, 14, 15). The dust appears to be limited to pier areas along the East
River, west of Columbia Street, and south of Congress Street. The southeastern limits
could not be determined due to a lack of photographic coverage south of Sedgewick
Street. The dust appears to lightly coat the paved areas of the wharfs (light-toned lines
are visible) and distinct vehicle track patterns are observed in the dust (Figures 14, 15).
The vehicle track patterns, however, are not observed on the roadways in the
surrounding area, such as Columbia Street or the Brooklyn-Queens Expressway.
Additionally, an excavation area with adjacent crane and multiple mounds of material
are observed in the general vicinity of the possible dust (Figure 15). A specific
association between the observed dust, the excavation activity and the mounded
material could not be ascertained.
In addition, due to improved photographic quality and coverage of the September
13 photographs in comparison the September 11, 2001 photographs, a qualitative
distinction was made regarding the ground dust/debris in the southeastern portions of
lower Manhattan along the FDR and areas east along the river. Possible and probable
ground dust/debris was now noted along the raised portions of the FDR and on the
piers extending into the East River (Figure 12). These areas were obscured on the
September 11, 2001 NYPD photographs and no qualitative distinction of the ground
dust/debris was attempted form the September 11, 2001 photographs.
23
-------
A more comprehensive ground dust/debris mapping effort in Brooklyn could not
be accomplished due to a lack of photographic coverage east of Hicks Street.
24
-------
REFERENCES
AIG (Analytical Imaging and Geophysics LLC). 2004. ACORN, version 4. 44509
Arapahoe Ave., suite 100, Boulder CO. URL: http://www.imspec.com/ (last
accessed 29 November, 2005).
ASPRS (American Society of Photogrammetry and Remote Sensing). 1997. Manual of
Photographic Interpretation, 2nd edition. Ed., Warren R. Philipson. ASPRS,
Bethesda, MD., pp.689.
Avery I.E. and G. L. Berlin. 2002. Fundamentals of Remote Sensing and Airphoto
Interpretation. Prentice Hall, Upper Saddle River, NJ., pp. 472.
Boardman, J.W. 1998. Leveraging the High Dimensionality of AVIRIS Data for
Improved Sub-Pixel Target Unmixing and Rejection of False Positives: Mixture
Tuned Matched Filtering. AVIRIS Airborne Geosciences Workshop Proceedings
(JPL Publication 97-12).
Clark, R., R. Green, G. Swayze, G. Meeker, S. Sutley, T. Hoefen, K. Livo, G. Plumlee,
B. Pavri, C. Sarture, S. Wilson, P. Hageman, P. Lamothe, J. Vance, J.
Boardman I. Brownfield, C. Gent, L. Morath, J. Taggart, P. Theodorakos, and M.
Adams. 2001. Environmental Studies of the World Trade Center area after the
September 11, 2001 attack. U. S. Geological Survey, Open File Report OFR-01-
0429. URL: http://pubs.usgs.gov/of/2001/ofr-01-0429/ (last accessed 29
November, 2005).
ESRI (Environmental Systems Research Institute). 2004. ArcView CIS, version 3.3.
380 New York Street, Redlands, CA. URL: http://www.esri.com/ (last accessed
29 November, 2005).
FEMA (Federal Emergency Management Agency). 2002. World Trade Center Building
Performance Study, FEMA 403, URL:
http://www.fema.gov/librarv/wtcstudv.shtm (last accessed 29 November, 2005).
Jensen, J. 1996. Introductory Digital Image Processing. Prentice Hall, Upper Saddle
River, NJ., pp. 318.
KAS (Keystone Aerial Survey). 2004. URL: http://www.kevstoneaerialsurveys.com/
(last accessed 29 November, 2005).
25
-------
Leica Geosystems. 2004. ERDAS IMAGINE, version 8.6. 2801 Buford Highway, Suite
400, Atlanta, GA. URL: http://gis.leica-geosystems.com/ (last accessed 29
November, 2005).
Lillesand, T., R. Keifer and J. Chipman. 2004. Remote Sensing and Image
Interpretation. John Wiley and Sons, Hoboken, N.J., pp. 763.
NCDC (National Climate Data Center). 2001. Daily Climate summary, Central Park,
New York, N.Y.
NYPD (New York Police Department). 2002. Above Hallowed Ground. Viking press,
New York, N.Y., pp.192.
RSI (Research Systems Institute). 2004. ENVI (The Environment for Visualizing
Images), version 3.6. 4990 Pearl East Circle, Boulder, CO. URL:
http://www.rsinc.com/ (last accessed 29 November, 2005).
Space Imaging Inc. 2004. 12076 Grant Street, Thornton, CO 80241. URL:
http://www.spaceimaging.com (last accessed 29 November, 2005).
USGS (United States Geological Survey). 2004a. URL:
http://landsat7.usgs.gov/index.php(last accessed 29 November, 2005).
USGS (United States Geological Survey). 2004b. URL:
http://eosims.cr.usgs.gov:5725/DATASET DOCS/landsat7 dataset.html#1 (last
accessed 29 November, 2005).
USGS (United States Geological Survey). 2004c. URL:
http://eo1.usgs.gov/hyperion.php (last accessed 29 November, 2005).
Final WTC Test and Clean Program Plan. 2005. URL: http://www.epa.gov/wtc/panel
(last accessed 29 November, 2005).
26
-------
Table 2. Listing of Digital Imagery and Aerial Photographs used for Analysis
Image Associated
Source Fiaures
Landsat 7 ETM+
NYPD
Landsat 7 ETM+
EO-1 Hyperion
IKONOS-1
SPOT-4
NYWCAP
KAS
IKONOS-1
4
2-3
5-6
7-8
9-10
*
*
11-15
**
Date of
Acquisition
08-27-01
09-11-01
09-12-01
09-12-01
09-12-01
09-12-01
09-12-01
09-13-01
09-15-01
Original
Scale/GSD
28.5m
Variable
28.5m
30m
1m/4m
20m
Variable
1:6,000
1m/4m
Data type
as supplied to EPA
MS Imagery
Color Photographs
MS Imagery
HS Imagery
MS Imagery
MS Imagery
Color Imagery
B&W Photographs
MS Imagery
NYPD: New York Police Department
NYWCAP: New York Wing Civil Air Patrol
KAS: Keystone Aerial Survey
MS: Multi-spectral
HS: Hyper-spectral
GSD: Ground Sample Distance
*The SPOT-4 and NYWCAP imagery, collected on September 12, 2001, were reviewed
for this report, however, they did not provide additional detail on ground dust/debris and
therefore were not included in this analysis.
** The IKONOS-1 imagery, collected on September 15, 2001, was used only as base
data for "heads-up" mapping of the September 11, 2001 and September 13, 2001
photographs.
27
-------
APPENDIX B: Local Climatological Data, New York C. Park, NY (NYC), SEP/2001
§ I
1 2
SEPTEMBER 2001
LOCAL CLIMATOLOGICAL DATA
NOAA, National Climatic Data Center
"•EMPERA'URE f
< o
6
I DEG DAYS
BASE 65
23 F =
gm 5 o
LU h ' UJ O
3> I O
WEATHER
NEW YORK C.PARK, NY
CENTRAL PARK OBSERVATORY (NYC)
Lat: 40°47'N Long: 73°58'W Elev (Ground): 158 Feet
Time Zone: EASTERN WBAN: 94728 ISSN #:0198 3601
PREC P"~K" ON PRESSURE
IINCHPo. IINUtri.UFHt,.
0700 1300 2400 2400
LSI LSI LSI LSI
5 S
& Uj
12
WIND
SPEED - MPI!
DP - ~6No OF DE9REE3
MAXIMUM
5 SEC 2 MIN
Q I Q
I£V, - -<_•» MAXIMUM TEMP _ 11
| THUtdERS'ORMsj
ooo
'll*!
rnrnfrt
SUNSHINE, CLOUD. &
i t-EA LEVEL
MAX MUM
M N MUM
M N MUM -EMP ^ JJ
M'NMUM~EMP i 0 : •
I IEAVY FOG . 1
PREC
PREC
3NOV.
ON _ 0 01 NCI I
PREC'P'"ATGN £ 0 10 NCH .
SNOWFALL *: 1.0 NCH
B-1
-------
HOURLY PRECIPITATION
(WATER EQUIVALENT IN INCHES)
NEW YORK C.PARK, NY
SEPTEMBER 2001 NYC WBAN # 94728
LJU
i
01
02
03
04
05
Qfi
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
! ' 2400
FOR HOUR (LST) ENDING AT j^ FOR HOUR (LST) ENDING AT uJ ^ g i? LST
1
0 10
0.38
0.03
0.11
2
0.05
0.12
3
0.30
0.07
4
0.13
0.16
0.03
5
0.17
0.04
I
6 7
0.16 0,19
T |
0.01 j
8
0.35
i ii- > - : i - i < c i z ~ *
•< I I Q 5 £ t » -i?
9 10| 11 12 p 13 14 15| 16 17 18 1 19 20 21 22 23 24 » 5 1 | $
i ior II1 01 0.10
!02 I °2 °-°°
:03 ' °3 OOD
O.OSi 04 0.08 0.01 1 T 1 04 0 14
JQ5 I I °5 °'OQ
J06 06 0.00
I 1 07 I 07 0.00
:08 II 08 0.00
109 I i 09 0.00
J10 0.41] 0,17^ 0.01 1 0.57' 10 1.16
1 11 II 11 0.00
!12 ' 12 ooo
I 113 . I 13 0.00
0.20' 0.06! 0,02 0,03 i 14 0,02' 0,01 T I 14 1,90
I 115 ' 15 000
i
1 16 I 1S °'°°
117 ! 17 0.00
118 I 18 0-0°
I !19 ' 19 O'^O
0.04 0.05 i 20 016 003; 0.07J 0.12; T T 0.24; 0.02 0.01 008 001 20 0.63
1 21 II 21 0.36
!22 II 22 0.00
!23 II 23 0.00
i24 ! I 001 003 24 0.04
0.011 001 ! 0.01 J25 II 25 °-41
326 II 26 0.00
i27 ! 27 0.00
I '28 I T I 28 T
!29 I.I . 29 0.00
i30 T 0.03! 0,011 003 0.111 0,04, 010: 0.03' 001 T 30 0.36
MAXIMUM SHORT DURATION PRECIPITATION (See Note)
_. _ ' Note ; The sum of the hourly totals is qivsn when
, Time Period (Minutes) 5 10 15 20 30 45 60 80 100, 120 150 180 it differs from the daily tote' NWS does not edit
: Preaoitatar, (Inches) 26 37 .41 .44 57 .57 .57 .58 58 ' 60 .73 '' .84 ' ASOShouriy vaiuesoutmayeditdaiiyandmonthiy
1 ' .... .... i ; tola's. Hourly. dai:y, and monthly totais are printed
EndmgDate 14 10 10 10 10 10 10 10 10 14 14 14 as reported by the ASOS site
Ending Time (Hour/Win) LjOp^9_JI£[8_ 1421 i 1836 1843 1843! 1843 1 843 i Jt843J_j:)84|J 0252 |_0322J
Date and time are not entered for TRACE amounts
NOAA, National Climatic Data Center
PAGE 2
B-2
-------
-------
APPENDIX C: Listing of Standard Operating Procedures.
The following remote sensing and geospatial Quality Assurance Standard
Operating Procedures (SOPs), as set forth in the Master Quality Assurance Project Plan
(MQAPP) under Remote Sensing Support Services Contract (RSSSC) 68-D-00-267,
were implemented for this report.
SOP Number SOP Title
1. RS-04-97-01-R11 Data Searches
2. RS-04-98-08-R4 Digital Imagery Acquisition
3. RS-05-98-08-R5 Maintaining Project Files
4. RS-06-97-07-R7 Inventory Control
5. RS-05-97-06-R9 Standard Category 3 Report Writing and Basic
Measurements
6. RS-05-97-07-R9 Standard Category 3 Site Analysis Quality Control
7. RS-06-97-03-R7 Producing Category 3 Reports
8. RS-05-97-02-R9 Photointerpretation: Hazardous Waste Sites
9. RS-05-97-04-R6 Photointerpretation: Land Use/Land Cover
10. RS-05-98-11-R7 Production of Digitally Scanned Images
11. RS-08-98-09-R4 Geometric Rectification of Scanned Aerial Photos
12. RS-09-99-01-R2 Mosaicking of Digital Imagery
13. RS-08-98-08-R3 CIS Coverage Quality Control
14. RS-08-98-05-R5 CIS Metadata Documentation
C-1
------- |