Proceedings of PHYSMOD2003
International Workshop on Physical Modelling of Flow and Dispersion Phenomena
3-5 September 2003, Prato, Italy
WIND-TUNNEL SIMULATIONS TO ASSESS DISPERSION
AROUND THE WORLD TRADE CENTER SITE
William H. Snyder1'2, David K. Heist1, Steven G. Perry1,
Roger S. Thompson1 and Robert E. Lawson, Jr.1,2
Atmospheric Sciences Modeling Division/Air Resources Laboratory/NOAA,
MD-81, US EPA, RTP, NC 27711, USA
ABSTRACT
A wind-tunnel study was conducted of
dispersion from the site of the destroyed World Trade
Center (WTC) in New York City. A scale model of
lower Manhattan, including a scaled representation of
the rubble pile, was constructed. The first phases of
the study involved smoke visualization and
measurements of flow patterns with winds from the
west; the second phase involved the measurement of
dispersion patterns resulting from tracer releases
from the rubble pile. Neither the initial explosions
nor the collapses of the towers have been simulated
but, instead, dispersion from the smoldering rubble
pile was modeled for the time period around two to
six weeks after the catastrophe.
Notable features included; strong horizontal
recirculation patterns caused by a group of tall
buildings not directly downwind acting as a single
obstacle, vertical recirculation caused by a tall
upwind building resulting in "pumping" of
contaminants up the lee side to heights above the
building top, and consistent alignment of flow
directions with the street canyon axes at the lower
levels, tending toward free-stream values at the upper
elevations.
NOMENCLATURE
C	Concentration, ppm by volume
H	Length scale, m
U	Mean velocity, m/s
U0	Reference or free-stream velocity, m/s
Rc5 Street canyon Reynolds number
W	Street canyon width
INTRODUCTION
The recent destruction of the World Trade Center
(WTC) in New York City resulted in releases of large
amounts of gaseous and particulate matter. With the
large population of workers and residents in this area,
these events have elevated the need for reliable
models to predict concentrations of such
contaminants within complex urban areas. This, in
turn, magnifies the need for laboratory measurements
of flow and dispersion in complex urban settings for
the development and evaluation of such models. The
primary goal of this study was to obtain, in the
laboratory, a data base that may be used to develop
'On assignment to the National Exposure Research
Laboratory, U.S. Environmental Protection Agency.
2Now retired.
specific guidelines for estimating near- and mid-
range concentrations of smoke and pollutants emitted
from the WTC site and, more generally, for
comparison with estimates from computational fluid
dynamics (CFD) models.
MODELING DETAILS
A photograph of the 1:600 scale model of the
lower end of Manhattan Island installed in the
Meteorological Wind Tunnel (3.7m wide, 2.3m high,
18m long, more fully described by Snyder, 1979) is
shown in Fig, 1. It was constructed of rigid
poiyurcthanc foam mounted on a plywood base and
centered 250m east and 115m south of the center of
the WTC site. It encompasses all of the southern tip
of Manhattan Island.
A simulated atmospheric boundary layer (abl)
was developed using three Irwin (1981) spires and
roughness blocks (18mm high, 27mm square) with
25% area coverage. These blocks ended at roughly
the western edge of the Hudson River, which is about
1 km wide at this point. Measurements showed the
full-scale (600:1) equivalent of this abl would have a
depth of 1100m, a power-law index of 0.2, and a
roughness length of 0.4m full scale at the end of the
roughness blocks, which is expected to match
reasonably well with the built-up urban/suburban
area on the New Jersey (western) side of the river.
The free-stream velocity was set at 4.23 m/s,
providing a street-canyon Reynolds number (U0W/v)
of approximately 10,000 for the smallest street-
canyon width (35mm or 21m full scale). Independent
measurements of flow in idealized two-dimensional
street canyons suggested that Reynolds-number
independence would be achieved if Res exceeded
4200.
A plan view of the model is shown in Figure 2.
A roughly square array of 9 discrete tubes was used
to release effluent from the smoldering rubble pile so
as to simulate the distribution of emissions (smoke
for visualization or neutrally buoyant ethane as a
tracer for quantitative measurements of
concentrations with flame ionization detectors). A
large number (133) of sampling ports were installed
on the model surface to facilitate the measurement of
ground-level concentration distributions. Sampling
rakes on a Lraverse system allowed lateral and
vertical concentration profiles to be measured at
virtually any position around the city.

-------
frreft«
IM
ligisrr. V V'lfiw flf seal? mrmri in vrimi tunnel, Ifiiifilnfi rtftft'rtStlrisfrnJ (Nttmrii fcwsl).
JR>)- (]»(.- ^fanitiilivi; nO'vv nli5ii>m«f)i>cijtj., ¦> b4fr-
ilpppltr jmgaanteter (L-DA) W3B iigcd si « ooirs of
[XiiiiLs littl^d wiiSiin v^riaJii sireet	spread
awinm! 13ie urea. The 1 ,Da wo* atrtitd thrionjjti
V^doWiSi tii ihi? jflooj; of l&e \vkid turniei so iii to
eliminate disturbances from the r Da probe as epeil
id flvoid ftuiUlfng	to I'ft? LPA lifie-of-
>igbl Thiti itrma iriitii umi alittwtid Ink; mtijaLicvnwni a!'
Figure I. Flan vitw of
Wbwital ciSrijfjxJlicnts of velocity. A sm^il mirror
prated lil 45" la tbt f,DA vitttvrntL ilirtsjlijs'i mi J
suppim-cd by sJisuJ^diutUftcr I'l'^ly twuflidirig Llirxi-uyli
the. window enahi^j rfoe nviditjiiri.'il pis^iiremeur fsf
vcrlicHl cuinpi.nieiVai ¦& well as veffl'sal ?raver&iiifc
Cwipft, I^S),
Eighl basic 'options were selected 3!m' LtM
fthfaSwviBSnlSi pru^cnun a crs>£» Krcliiin at" the
djlif.rtrni types (if khciil building topes©pLii^ jft the
rcriot; !£.£., low-riit buildings wflh mtfi»' Streets,
open spate surrounded by tall feiiiI uarrww
•catavLin siinscsuiided by call buttdimga eic.) and
Crtvjtii'injti a range uF-di-siansfls and diraetiM; fro-m t'ia
WTC fli.Lf, tiidi fid" these moasareinEii;: loea
-------
smoke up the iee side to the building top and even
above, resulting in what appeared to be a continuous
elevated release. Similar "pumping" action was
observed from buildings to the south and somewhat
downwind of the source. The results may be seen in
Figure 3, where a laser-light sheet illuminated a
horizontal plane just above one of the tallest
buildings (the Chase Manhattan Building). The
figure would appear to show 3 distinct "plumes".
The one on the right originated from the WFC (213m
high), the middle one from the Liberty Plaza Building
(236m), and the (somewhat more diffuse) one on the
left from the Chase Manhattan Building (274m).
These are the three tallest buildings in the vicinity of
the WTC site.
Figure 3, Visualization of smoke in a horizontal
plane just above the tallest building.
Another striking feature observed during the
visualization was that puffs of smoke, albeit weak,
were very frequently observed quite near the sidewall
of the tunnel opposite Wall Street (located southwest
of the WTC site). This was not-at-all expected, since
it would require a half-angle plume spread in excess
of 45°, and it appeared that it was the result of a
large-scale horizontal recirculation of the flow. The
large dense cluster of very tall buildings surrounding
Wall Street acted much like a single obstruction, and
the flow downwind from the WTC site was observed
to curl around the lee side of this obstruction.
Consequently, smoke carried by this recirculating
flow was observed to infiltrate the street canyons in
the vicinity of the financial district and even close to
the tunnel wall.
Flow vectors at two different elevations over the
city are shown in Figure 4. The low-level vectors
(9m elevation) show strong channeling within the
street canyons. Intermediate-level vectors (not
shown) align more closely with the free-stream
direction over the low-rise buildings but are strongly
influenced by the taller buildings. The vectors at the
highest elevation (90 m) clearly align with the tree-
stream direction except in the near vicinity of the
very tallest buildings. This behavior is as expected.
Flow vectors in cross sections perpendicular to
two street canyons are shown in Figure 5. These
cross sections were located only 60m apart along the
same street (Church), which is oriented about 20° off
perpendicular to the free-stream wind direction, but
they were separated by a cross street (Murray - see
points "A" and "B" in Fig. 2). The two locations are
more or less on the borderline between areas of low-
and medium-rise buildings. In Figure 5a, located as
shown by point "A" in Fig. 2, the upwind building is
about the same height as the downwind one, and the
spiraling flow within the street canyon is quite clear.
The along-street component of the flow vector at the
9m elevation (shown in Fig. 4) is a little over twice as
large as the cross-street component (0.14U0
compared with 0,063Uo). The tall protrusion on the
downwind building is of small cross section and thus
has little effect upon the overall flow structure.
By contrast, Figure Sb, located as shown by Point
"B" in Figure 2, has a much taller upwind building
and, whereas spiraling How is not prominent here,
upwash on the lee side of the upwind building is very
strong over the full cross section. Continuity would
suggest a fairly significant aiong-street component
and, indeed, Figure 4 shows this to be 0,2Uo,
The surface concentration pattern is shown in
Figure 6, where the footprint appears to be strongly
skewed toward the south, both in terras of the
"centerline" of the plume and the large tail on the
south side. This is further evidence of the influence
of the dense cluster of tall buildings just southwest of
the WTC site as mentioned earlier (Wall Street area).
Finally, cross sections of concentration were
measured at 300, 600 and 1200m downwind of the
WTC site. These are shown in Figure 7, where
isoconcentration contours against a background of
the city skyline viewed from downstream provide
some indications about the plume size and behavior.
(The values of the isoconcentrations are
nondimensional, 100CUoFl2/Q, where C is the
measured concentration, H is a reference length scale
(150mm) indicative of, perhaps, an average building
height, and Q is the volumetric source How rate.)
At the 300m distance (Fig. 7a), the most notable
feature of the cross section is the two lobes of higher
concentrations on the sides, with a valley of lower
concentrations in the middle. This is rather clear
evidence of the "pumping" of effluent up the lee
sides of the World Financial Center and the Liberty
Plaza Building. The influence of the Chase
Manhattan Building has not been felt at this
downwind distance, since it is itself located very near
this plane. The plume is strongly asymmetrical, with
the north lobe being much wider and higher in
elevation than the south lobe. Also at this downwind
distance, the plume appears to be shifted slightly to
the north side of a line aligned with the free-stream
wind and through the center of the WTC site
(henceforth referred to as the centerline). The largest

-------
4508000
4507500
O)
c
IE
*:
o
4507000
4506mooo
&wmf%Q^4
on	rk*

n m
Building
heights (m)
0-50
51-100
101-150
U/Uo=1
Reference vector
583500	584000
Easting
584500
z = 90 m
4508000
/?
? X/ /-i Q
a-
Gn
\"
-------
10D
«. V
-50
-25
x, m
1S0|-
(b)
E -
f-T
•fi $ *r
If 4 4
f M
¦f i A 	
4» ),
~ V V-
i,i I i 			1	i	,				
-SO -25 0 25 50
x, m
Figure 5. Flow vectors in street canyons.
concentration, indicated by the circled star, however,
is on the centerline.
At 600m downwind, two iobes are stiil evident,
but are more diffuse. The more notable feature is the
lateral shift of the plume from a northward bias to a
southward bias; the maximum measured
concentration is located over 100m south of the
centerline. The maximum concentration is shown
elevated above ground level at this distance, but it
may have been lower; no measurements were made
at lower elevations here, because buildings were in
the way.
At 1200m downwind, only the north lobe is
pronounced; a roughly horizontal concentration
contour on the south side suggests that the middle
and south plumes observed in Fig. 3 have essentially
merged at this downwind distance. Even more
noticeable here is the strong spread of the plume to
the south side, with the "10T'-contour reaching only
170m to the north of the centerline, but 440m south
of the centerline. The maximum concentration is
located about 240m south of the centerline, indicating
an angular shift of the plume by approximately 11°
from a line directly downwind of the source. Again,
we believe this is an indication of the recirculation
caused by the dense cluster of tall buildings in the
vicinity of Wall Street (centered at approximately
500m south of the centerline through the source). At
this distance, the maximum concentration is elevated
approximately 80m above ground level. The dots
forming vertical lines in this part of the figure
indicate the locations where concentrations were
measured.
CONCLUSIONS
These measurements show obviously very
complex flow patterns over the densely packed
buildings in lower Manhattan, both on the micro-
level of the street canyons (spiraling flow, upwash
and downwash behind buildings, and channeling
within the canyons) as well as on the macro-level of
the large-scale recirculation effected by the dense
cluster of Wall-Street Buildings acting as a single
obstacle.
Flow visualization showed that three of the
tallest buildings surrounding the WTC site caused
strong transport of contaminants up their tee sides,
with results that looked like "chimneys" outputting
smoke plumes above their tops. The World Financial
Center was actually upstream of the WTC site, so
that effluent was first entrained into the building
wake, then transported to the building top. The
Chase Manhattan Building was well off to the south
and well downwind of the source, but nevertheless
displayed similar behavior.
Flow measurements over two slreet-canyon cross
sections in the same street showed quite different
behavior, In one case, strong spiraling flow was
observed, but in the other, longitudinal flow along the
street with strong upwash on the lee side of the
upwind building and over the full street width was far
more prominent. Flow vectors at various positions
around the city gave fairly strong and consistent
indications of flow channeling within street canyons
at low levels, tending systematically toward the free-
stream direction just above the tops of the adjacent
buildings.

-------
Figure 6. Surface concentration pattern over lower Manhattan.

-------
400
400
Chase Manhattan
Building Liberty Plaza
x Building
•;®1
World Financial
Center
*1
-600 -400 -200	0	200 400
Cross Wind Distance (m)
600
KT 4
-600 -400 -200	0	200 400
Cross Wind Distance (m)
600
: (c)
n
l
-600 -400 -200 0	200
Cross Wind Distance (m)
400
600
Figure 7. Plume cross sections at downwind distances of: a) 300m, b) 600m, and c) 1200m. View
looking upstream against skyline of city. Shorter buildings indicated in lighter colors.

-------
Concentration measurements showed a highly
distorted plume downwind that was clearly
influenced by the "pumping" action of the tall
buildings and by the blockage effect of the dense
cluster of tall buildings surrounding Wall Street. The
locus of maximum concentrations did not follow the
free-stream wind direction, but rather deviated by an
angle in excess of 10° from a line pointing directly
downwind. Further, the lateral distributions showed
a bifurcation of the upper levels of the plume (high
concentrations on the two sides with lower
concentrations in the middle) that clearly resulted
from the upwash behind the tall buildings.
Upon completion of the study, an extensive data
set containing all of the three-dimensional mean
velocity and turbulence data as well as concentration
measurements will be made available for the
development and evaluation of CFD and other types
of models.
DISCLAIMER
This research has been supported by the US
Environmental Protection Agency. It has been
subjected to agency review and approved for
publication. Mention of trade names or commercial
products docs not constitute endorsement or
recommendation for use.
ACKNOWLEDGMENTS
We acknowledge with thanks the generous
support of Dr. Bruno Pagnani and Messrs. Ashok
Patel and John Rose.
REFERENCES
Irwin, H.P.A.H. 1981 The Design of Spires for Wind
Simulation. J. Wind Engr. Indus. Aerodyn,, 7, 361-
366.
Snyder, W.H. 1979 The EPA Meteorological Wind
Tunnel: Its Design, Construction, and Operating
Characteristics. Rpt. No. EPA-600/4-79-051, Envir.
Prot. Agcy.. Res. Tri. Pk., NC, 78p.
Snyder, W.H. & Castro, LP. 1998 The Yaw-
Response of Hot-Wire Probes at Ultra-Low Wind
Speeds. Meas. ScL & Tech, 9, 1531-1536,

-------