Quantifying Vegetation Effects on Near-Road Air Quality

This report, based on a study conducted by Cornell University, investigates the effects of vegetation on near-road air quality, specifically focusing on the dispersion of PM2.5. The research involved brief field campaigns in Queens, NY, near major highways, where high-resolution measurements of PM2.5 were taken along transects. The study found that the presence of trees can influence local pollution concentrations, sometimes increasing them due to aerodynamic effects that reduce dispersion. The report highlights that the polydisperse PM2.5 class poorly represented the behavior of discrete classes. The study also compared transects across lawns with and without trees, observing fewer spikes in PM2.5 concentrations but a more gradual decrease downwind of trees, indicating recirculation and reduced dispersion. The findings emphasize that simply planting trees might not always improve air quality, especially if the goal is to protect vulnerable populations. The report concludes by advocating for a mechanistic approach based on fluid dynamics to better understand and manage air quality in urban environments.

Quantifying the effect of vegetation on near-road air quality using
brief campaigns
Zheming Tonga
, Thomas H.Whitlow b, * , Patrick F.MacRaeb
, Andrew J. Landersc
,
Yoshiki Haradab
a Department of Mechanical and Aerospace Engineering,Cornell University,Gruman Hall,Ithaca,NY,USA
b Section of Horticulture,School of Integrative Plant Science,Cornell University,Room 23 Plant Science Building,Ithaca,NY 14853,USA
c New York State Agricultural Experiment Station,630 West North Street,Geneva,NY 14456,USA
a r t i c l e i n f o
Article history:
Received 18 August 2014
Received in revised form
18 February 2015
Accepted 20 February 2015
Available online 20 March 2015
Keywords:
Near-road air pollution
Trees
Dispersion
Aerodynamics
PM 2.5
a b s t r a c t
Many reports of trees'impacts on urban air quality neglect pattern and process at the landscape scale.
Here,we describe brief campaigns to quantify the effect of trees on the dispersion of airborne particu-
lates using high time resolution measurements along short transects away from roads.Campaigns near
major highways in Queens,NY showed frequent,stochastic spikes in PM2.5.The polydisperse PM2.5
class poorly represented the behavior of discrete classes.A transect across a lawn with trees had fewer
spikes in PM2.5 concentration but decreased more gradually than a transect crossing a treeless lawn. Th
coincided with decreased Turbulence Kinetic Energy downwind of trees, indicating recirculation, longer
residence times and decreased dispersion.Simply planting trees can increase localpollution concen-
trations, which is a special concern if the intent is to protect vulnerable populations.Emphasizing
deposition to leaf surfaces obscures the dominant impact of aerodynamics on local concentration.
© 2015 Elsevier Ltd.All rights reserved.
1. Introduction
There is a general consensus that proximity to major highways
increases the risk of adverse health effects caused by exposure to air
pollution (HEI, 2010). Roadside barriers, including vegetation, have
been shown to alter the dispersion of traffic emissions.If the
vegetative barriers consistently lower ground-levelair pollution
concentrationsin the near-road environment, they may be a
practical tool for reducing human exposure to air pollution along
populated roadways.
It is widely reported that trees intercept airborne particles
which are subsequently removed from the canopy by re-
suspension,by rain and leaf abscission (Dochinger,1980; Freer-
Smith et al., 2004; Nowak, 2002; Nowak et al., 2013). Using
empirical estimates of deposition velocities, these reports estimate
the total particulate removed by trees (typically PM10) at either city
wide or local scales.Calculations like these are often used to
advocate tree planting policieslike the numerous million-tree
programs across the US.However laudable these programs are,
the approach ignores the effects of distance from source and the
local aerodynamics around trees,how these affect dispersion and
ultimately local PM concentration, and provide no guidance for the
rational design of landscapes to improve local air quality.For this
purpose, a mechanistic approach based on fluid dynamics of
different particle sizes and the local turbulent flow field caused by
road-canopy configurations is needed.
Aerosol science has long known that particle dry deposition
velocity varies as a function of particle size,and ranges over three
orders of magnitude (Sehmel,1980; Seinfeld and Pandis,2006;
Slinn et al., 1978).This is because particles <0.001mm behave
more like gases, diffusing along concentration gradients to deposit
on surfaces,unlike particles >10mm whose deposition rates de-
pends on inertial impaction and gravitationalsettling.The local
turbulent flow field also plays a significant role in particle disper-
sion.A tree canopy consists of numerous elements such as leaves,
branches and trunks.When these elements interact with airflow,
the flow momentum is absorbed by both form and skin-friction
drag on the canopy,reducing mean flow velocity (Raupach and
Thom,1981).Larger scale turbulent eddies introduced by traffic
and the background atmosphere are broken down to smallscale
eddies by a tree canopy,causing a recirculation zone behind the
vegetation with elevated concentrations (Steffens et al., 2012; Tong
et al.,2011; Wang and Zhang,2009).
* Corresponding author.
E-mail address: [e-mail] (T.H.Whitlow).
Contents lists available at ScienceDirect
Environmental Pollution
j o u r n a lhomepage: w w w . e l s e v i e r . c o m / l o c a t e / e n v p o l
http://dx.doi.org/10.1016/j.envpol.2015.02.026
0269-7491/© 2015 Elsevier Ltd.All rights reserved.
Environmental Pollution 201 (2015) 141e149

Both experimental and numerical simulation studies have
investigated the effectof vegetation on PM concentration along
roads unbounded by buildings.Brantley et al.(2014) conducted a
field assessment of the effect of roadside vegetation on near-road
black carbon and particulate matter.They found that particle
counts in the fine and coarse particle size range (0.5e10mm aero-
dynamic diameter)were unaffected by vegetation.Baldaufet al.
(2008) found that both a solid noise barrier and vegetation barrier
can reduce PM concentrations in their wakes when wind is from the
direction of the road. In general, these studies have shown decreased
concentrationsof ultrafine and coarse mode PM with limited
reduction measured for PM2.5 mass. Set€al€a et al. (2013) used passive
samplers to study the effect of urban park/forest vegetation on NO2,
anthropogenic VOCs and particle deposition in two Finnish cities.
They found that pollutant concentrations were often only slightly
lower under tree canopies than in adjacent open areas. Maher et al.
(2013) examined the impact of a line of young trees on indoor air
quality adjacent to a heavy traffic road, and a substantial reduction of
PM 10 was observed. Cavanagh et al. (2009) conducted a field study to
investigate the spatialattenuation ofPM10. Concentrations were
higher outside the forest than deep within the forest.
Other researchers have used physicalmodels in wind tunnels
and Computational Fluid Dynamics (CFD) models to simulate the
impact of vegetative buffers on roadside plume dispersion. Gromke
and Ruck performed a wind tunnel experimenton dispersion
processesof traffic exhaust in urban street canyons with and
without street trees (Gromke, 2011; Gromke and Ruck, 2009). Trees
reduced pollutant dispersion, thereby increasing particle residence
time and concentration.In the wind tunnel, street trees caused
localized concentration increases of 50% at some locations in the
canyon compared with the treeless case. This indicates that trees in
street canyons reduce air exchange with the ambient atmosphere.
Buccolieri et al.(2009) conducted both CFD and wind tunnel ex-
periments to study the aerodynamic effects of trees on pollutant
concentration in street canyons.Both approaches showed consid-
erably greater pollutant concentration near the leeward wall and
slightly lower concentration near the windward wallwhen trees
were present.Another CFD study compared CFD modeled results
and field measurements explore the effect of a near-road vegeta-
tion barrier on ultrafine particles (Steffens et al.,2012).The CFD
model was evaluated against the roadside measurements,and a
good agreement was observed (Hagler et al., 2012). They found that
increasing leaf area density (LAD) reduced ultrafine particle con-
centration,but the response was non-linear.Pugh et al.modeled
the effect of green walls on air quality in a street canyon.Using
deposition velocities from the literature, they calculated that green
walls could cause a 40% reduction for NO2 and a 60% reduction for
PM 10 (Pugh et al.,2012).
This brief review shows that vegetation can either decrease or
increase PM concentration,depending on the road-canopy config-
uration, particle size, and local flow field. The goal of this study is to
improve our understanding of the impact of vegetation on PM2.5
transport in the near road environment. We focus on PM2.5 because
it includes the particle sizes with the lowest deposition velocities
and is more closely linked to human mortality (EPA, 2009; Seinfeld
and Pandis, 2006). By strategically deploying multiple particle
counters and sonic anemometers,this approach achieves high
spatial and temporal resolution of PM2.5 concentration in discrete
particle size classes,and corresponding turbulence data.This is a
unique addition to the existing literature that provides empirical
data for detailed landscape scale modeling.
We posed 4 initiating hypotheses:
1. PM 2.5 concentrations will be reduced below ambient downwind
of tree canopies.
2. PM 2.5 concentration will decline more sharply along a transect
occupied by trees than an open transect
3. The effect of trees on PM2.5 concentration depends on wind
direction.
4. The effect of trees on PM2.5 concentration depends on particle
size.
2. Methods
2.1.Measurement approach
We used an observational approach to conduct a series of short-
term field campaigns exploring the spatiotemporalpatterns of
particulate matterdispersion across a large urban open space
(Dominici et al.,2014).We used portable monitoring instruments
(see below) to conduct a series of brief, intensive campaigns during
a 2 week period,lasting ca. 10 h each day during daylight hours,
capturing both morning and evening rush hours.This approach
resembles that of Spengler et al.(2011) in their study of ultrafine
particles in a neighborhood adjacent to a toll plaza. In comparison
with permanently located monitors, brief campaigns can be used in
public spaces where vehicles are not allowed, make efficient use of
instrumentation and labor,allow multiple locations to be moni-
tored in real time, are suited to addressing the effectiveness of
vegetated buffers atscales relevantto engineering and human
exposure, and permit sampling where permanent samplers cannot
be secured against vandalism. Importantly, small mobile sensors do
not impact local dispersion patterns and can monitor near the
ground where human exposure would occur.The tradeoff is in
terms of generalizability of the findings over long time periods and
varying air mass conditions.
2.2. Sample location
We selected Flushing Meadows-Corona Park,a 3.63 km2 com-
plex in Queens,New York City,USA (Map is shown in the Supple-
mentary Material(SM1), and relevantfeatures are described in
Table 1). The park is surrounded by the heavily trafficked Van Wyck
and Long Island Expressways (LIE),allowing us to select sample
locations to control for prevailing wind direction on any given
sampling day.Annual Average Daily Traffic (AADT) is 84,601 vehi-
cles/day on the Van Wyck and 138,406 vehicles/day on the LIE. Over
a 2-week mid-summer period when trees were in full leaf, we
sampled at three locations in the park when weather and wind
direction were suitable for testing our hypotheses.None of these
sites was deliberately designed to modify airflow or capture par-
ticles, yet each represents a landscape common in urban centers in
the eastern US, consisting of trees, lawns, and playing fields near a
highway. The park is separated from the highway right-of-way by a
continuous 2.4 m high chain link fence deliberately kept free of
vegetation, thus having essentially no impact on wind and particle
movement at the scale of our measurements (Details of the vege-
tation at each site are presented in Table 1 and SM2).
2.2.1.Hypotheses 1,2 and 4
Northeast of the Van Wyck Expwy, we sampled 2 parallel
transects along distance gradients from the road to test for differ-
ences in PM 2.5 transport across a lawn with scattered trees
compared with an adjacent open lawn (Fig. 1A). One particle
counter was located next to the highway while the other counters
were rotated among 3 points along the 2 transects every 15 min,
yielding a 45 min cycle. This was repeated throughout the day
(Table 2).
Z.Tong et al./ Environmental Pollution 201 (2015) 141e149142

2.2.2.Hypotheses 1 and 4
The 2 LIE sites were selected because the landscape on both
sides of the highway has roadside trees,open lawn and sports
fields, and in the case of LIE North, a patch of closed canopy forest.
On the south side of LIE (Fig. 1B), we located particle counters at 3
fixed points: at the highway edge, 12 m downwind of a line of trees,
and 52 m downwind of the trees in an athletic field. Measurements
were fully synchronized in time.
2.2.3.Hypotheses 3
North of the LIE, we sampled at 3 static locations: adjacent to the
highway shoulder,in a grassy field and under a forest canopy
(Fig. 1C). On the day we sampled, wind was from the north, upwind
of the highway. The three sites are referred to as Van Wyck East, LIE
South, and LIE North in the text.
2.3. Instrumentation
2.3.1.Particle counters
We measured atmospheric particulates using 3 Grimm Aerosol
Spectrometers (Model1.108) equipped with isokinetic probes to
reduce the effect of variation in wind speed. We monitored 15 size
classes between 0.3 and 20mm every 6 s. This approximates a
human resting inhalation rate and also allows us to observe con-
ditions corresponding to spikes in PM 2.5 concentration. In-
struments had been factory calibrated just prior to the summer
campaign. In addition, all 3 instruments were co-located for 60 min
each sampling day and readings from each instrument were
regressed against their average.These empiricalequations were
used to adjust readings to compensate for small variations among
the instruments.We approximated fine particulate matter (PM2.5)
as sum of all sizes from 0.3mm to 3.0 mm, and particle counts are
converted to mass by assuming that the particles are spherical and
using the conversion factor 1.4 g/cm3 (Armbruster et al., 1984;
Murakami et al.,2005).This underestimated the regulatory defi-
nition of PM2.5 because it excludes particles below the detection
Table 1
Site description; a) Measurements were taken on days when weather permitted and wind direction was appropriate for testing the 4 hypotheses; b) The unit of latitude
longitude is in decimal degrees. c) Canopy porosity was determined from hemispherical images taken beneath the canopy. More details are provided in the Supplemen
Material (SM2).d) Tree cover percentage was measured with a line intercept method from aerial photographs.e) Grass,bare soil,and pavement cover were measured by
quadrat method.Percentage of road pavement is not presented in this table,but can be found in SM2.A list of tree species is also provided in SM2.
Datea Latitudeb Longitudeb Porosityc % Treesd % Bare soil % Grasse
Van Wyck East Jun 7,2011 40.723 73.838 15.7% 44.1%(w/trees)
0%(no trees)
10.5%(w/trees)
3.5%(no trees)
89.5%(w/trees)
96.5%(no trees)
LIE South Jul 13/14/15,2011 40.741 73.841 9.8% 4.3% 3.1% 66.7%
LIE North Jul 12,2011 40.743 73.841 21.9% 82.5%(w/trees)
6.3%(no trees)
17.0%(w/trees)
22.5%(no trees)
71.9%(w/trees)
64.7%(no trees)
Fig. 1. Details of the sample points at the 3 sites. A) Van Wyck East,stations and 2,3,4
represent the vegetated transect,and 5, 6, 7 represent the open transect.Station 1
beside the road serves as a common reference point for both transects; B) LIE (Long
Island Expressway) South; C) LIE North; Wind roses are based on daily on site mea-
surements.The numbers on each figure indicate the sampling points.In the text,the
sites are referred to as Van Wyck East,LIE South,and LIE North.
Table 2
Average concentration and standard deviation (shown in parentheses) of PM2.5 at
various sampling stations for Van Wyck East and LIE South; Locations are indicated
by distances from the road. The averaging period for Van Wyck East site is the same
as the one used in the decay curves. The averaging period for LIE South is from 12:10
to 1:30 PM where the traffic and wind condition is most steady.
Open transect at Van Wyck EastRoadside 10 m 23 m 40 m
Average concentration [mg/m3
] 4.92(1.84) 3.90(1.33) 3.81(1.19) 3.75(1.27)
Vegetated transect at Van
Wyck East
Roadside 7 m 15 m 51 m
Average concentration [mg/m3
] 4.92(1.84) 4.36(0.94) 4.46(1.11) 4.17(0.81)
Vegetated Transect at LIE SouthRoadside 12 m 52 m
Average concentration [mg/m3
] 1.96(0.84) 1.77(0.74) 1.65(0.64)
Z.Tong et al./ Environmental Pollution 201 (2015) 141e149 143

limit of the instruments (0.3mm). It does, however,include the
range of particle diameters in so-called accumulation mode where
deposition is lowest.
2.3.2.Instantaneous wind speed/direction
We used four 3-D Gill sonic anemometers to measure the
instantaneous wind speed and direction at 1 Hz.These data were
used to generate wind roses for the sampling days and also tur-
bulent kinetic energy (TKE,see below and SM4).
2.3.3.Hemispherical canopy porosity and tree cover
Porosity is a property ofvegetation that correlates wellwith
downwind velocity, turbulence and particle deposition (Heisler and
Dewalle, 1988; Li et al.,2010; Loeffler et al., 1992; Raupach et al.,
2001).We estimated porosity to characterize the canopy density
of the trees closest to our downwind monitoring stations for each of
the three sites using a technique modified from Kenney (1987). We
used a digital camera (5 megapixel resolution; Nikon Coolpix 5700)
equipped with a fisheye lens (Nikon FC-E9) to take hemispherical
images beneath the canopy. Images were rendered in high contrast
black and white in Photoshop®
, white and black pixels were tallied,
and porosity was calculated as the % white pixels on the image. An
example of the hemisphericalimage is provided in the Supple-
mentary Material.We also estimated tree cover using line in-
tercepts perpendicular to the highway and tabulating the distance
below the drip lines of the trees as a percent of the total distance
between the highway and the particle counters.
2.4. Data analysis
2.4.1.Temporal variation
Koniographs,analogous to hydrographs used by hydrologists,
were used to show fine scale temporal variation in concentration at
the 6-s sampling frequency of the aerosol spectrometers (Whitlow
et al., 2011). This sampling rate approximates the human inhalation
rate, hence exposure to short term concentration spikes.
2.4.2.Return period
PM concentrations averaged over periods ranging from days-
years,while useful for regulatory purposes,eliminate fine scale
patterns that are usefulfor quantifying human exposure risk at
temporal scales relevant to daily activities,especially physical ex-
ercise. Risk is probabilistic and contingent on many environmental
factors, many of which are beyond our control. Recognizing that air
pollution events are stochastic over time, resembling flood events,
we used the Gumbel Method (Gumbel, 1941; Whitlow et al., 2011)
to calculate return period of PM2.5 events of any observed magni-
tude during each day's set of observations. Return period estimates
the magnitude of the highest concentration occurring during a
given period.This statistic is analogous to the familiar 100-year
flood, which expresses the probability ofa flood of an observed
magnitude occurring in any given year. In our usage here, the time
scale is in minutes instead of years. Further, like the 100-year flood
plain, it also characterizes a specific location.The calculation of
return period is provided in the Supplementary Material (SM3).
2.4.3.Turbulence Kinetic Energy (TKE)
We calculated TKE for time intervals when the wind direction
and speed were relatively steady for each day.The sampling fre-
quency of the sonic anemometer is 1 Hz,which though not rapid
enough to capture turbulence in the dissipation range, does capture
most of the energy containing eddies.The calculation ofTKE is
provided in the Supplementary Material (SM4).
3. Results and discussion
3.1.Experiment 1: Van Wyck East
The high resolution of the 6-s sampling frequency shows the
nearly instantaneous stochastic variation of PM2.5 concentration in
the roadside environment.Sampling location 1 in Van Wyck East
site adjacent to the highway displays most variable PM2.5 concen-
tration data (Fig.2a),showing frequent spikes above background,
corresponding to passage of especially “dirty” vehicles. Because our
spectrometers cannot detect particles <0.3mm, these spikes are not
caused by primary tailpipe emissions butare either secondary
particles or particles re-suspended from the road surface and lofted
by the turbulent wakes of vehicles (Fig. 3a). Observations across the
vegetated transectshowed far less variation,higher mean con-
centrations, and large spikes in concentration were absent (Fig. 2b),
while the open transect had concentration spikes (Fig.2c).
Return period plots reveal that along the open transect, samples
at 10,23 and 40 m from the road are virtually indistinguishable
from each other (Fig.3a).In comparison,samples on the transect
with trees show differences in magnitude of events at frequencies
exceeding ca. 1 min (Fig.3b).
While trees attenuated concentration spikes,the transect with
trees had higher average PM2.5 concentrations (SM5,Table 2).
Koniographs of the 15 min average concentrations corresponding
to the sampling intervals at the different points on the transect
show the pattern more clearly than the 6-sec koniographs (SM5).
Roadside PM2.5 varied over the course ofthe day in relation to
traffic conditions.Noticeably greater concentration was observed
during the morning and afternoon rush hours, when traffic was low
speed, stop and go (Tong et al., 2000). This flow pattern is typically
more polluting than the steady-speed driving modes,and gener-
ates more total emissions (Tong et al.,2011).
Fig. 2. Koniographs atVan Wyck East at 6 s sampling resolution.a) station 1 at
roadside; b) roving among station 2,3, and 4 (7 m, 15 m,51 m away from the road)
along the vegetated transect,c) roving among station 5,6, and 7 (10 m,23 m, 40 m
away from the road) along the open transect.
Z.Tong et al./ Environmental Pollution 201 (2015) 141e149144

All else being equal,concentration should decrease exponen-
tially with distance from source.To de-confound the effect of dis-
tance from the effect of trees, we fit exponential curves to data from
a period of relatively constant traffic conditions and roadside PM2.5
concentration between 11:30 AM to 3:30 PM (Fig. 4 Top, SM6). Each
point is the average of ca. 800 observations. The polydisperse PM2.5
size class shows the expected exponential decline,with the open
transect showing a steeper decline than the transect with trees
(Fig. 4 Top). Two parameters quantify the difference in decay rates.
The decay constants in the equations fitted to the data are 0.029
(with trees) and 0.204 (open), a 7-fold difference. The half distances
(the distance required to attain half of the initial concentration) are
12.74 m (with trees) and 3.46 m (open) (Table 2). Both parameters
indicate that particle concentration decreases ata substantially
higher rate along the open transect.
Note that the R2 for PM2.5 along transect with trees was much
lower than that for the open transect, indicating a poorer fit (SM6).
To explore the possible reasons forthis, we fitted exponential
curves to the largest and smallest, more nearly monodisperse
particle sizes included in our estimates of PM 2.5, namely
0.3e0.4 mm and 2.0e3.0mm (Fig.4 Bottom).As before,concentra-
tions of both size classes declined more steeply across the open
transect than across the transect with trees. In contrast with
0.3e0.4mm class, 2.0e3.0mm class for both transects declined more
sharply due to greater settling velocity. These findings reaffirm the
well-known fact that larger particles within the polydisperse PM2.5
class behave differently than smaller ones even though they are
lumped for regulatory purposes (Seinfeld and Pandis, 2006).
3.2. Experiment 2: LIE south
This experiment investigated the effect of a line of trees near the
road. Observations were made using instruments at static locations
and were synchronized,thereby accounting for fine scale differ-
ences in emission sources and weather.It provides valuable infor-
mation to study instantaneous spike attenuation and turbulence
dissipation along the transect.There was no open transect for
comparison at the site.
Although many concentration spikes at the roadside were
closely correlated with spikes at stations distant from the road,
non-synchronous spikes were also observed at these distant loca-
tions (Fig. 5). Further analysis shows that particles in the
1.6e2.0mm and 2e3mm size classes accounted for 63% of the total
mass of particles comprising the spikes in contrast with the average
proportion for the entire period of 36%.Taken together,these ob-
servations indicate that the road was not the only source of PM2.5
and that different sources have characteristically different size class
signatures. Variation in PM2.5 concentration was lower at station 2
behind the vegetation 12 m from the roadway (Table 2).A com-
parison of decay curves for PM2.5, 0.3e0.4 mm and 2e3 mm (Fig.6,
SM7) shows that the concentration of2e3 mm particle declined
more sharply than 0.3e0.4mm particles. The average concentration
of 0.3e0.4mm particles downwind of the line of trees was essen-
tially the same as beside the road despite being 12 m more distant.
Return period plots for the three different size classes illustrate
the importance of both landscape location and particle size in
determining the exposure risk (SM8).For events with return pe-
riods <4 min,PM 2.5 behaves as might be expected: more distant
locations experience lower concentrations.For events recurring
less frequently than 4 min,however,the field station had higher
concentrations than the station behind the trees (station 2).Inter-
estingly,for very frequent events (<0.3 min),concentrations of
particles between 0.3 and 0.4mm were highest at the 12 m station,
downwind of trees.
3.2.1.Recirculation zone
Wind measurements upwind and downwind of trees provides
insight into the local turbulent flow field. At the station downwind
of a tree, reduced TKE was observed at both Van Wyck East and LIE
South (Fig. 7a). TKE followed the same general pattern at both sites.
It was lowest immediately downwind of trees and greater further
downwind and in the open for all three days and exceeded the TKE
at the roadside station for two out of three days.Large scale back-
ground turbulence due to atmospheric instability was the likely
cause.This finding agrees with many other studies showing that
porous vegetation barriers reduce mean wind speed,break down
larger scale upwind turbulent eddies,and create a recirculation
zone downwind (Kaimal and Finnigan, 1994; Steffens et al.,2012).
The particle data shows that the PM2.5 concentration along the
vegetated transects decays considerably slower and has less vari-
ation than across the open transect.The recirculation zone slows
particle dispersion and sometimes results in higher local concen-
tration than the open transect (Figs.4 and 6). In contrast,solid
barrier generally reduces concentration immediately downwind by
deflecting the flow field upward (Baldauf et al., 2008; Hagler et al.,
2012). A barrier comprised of trees with low hemispherical porosity
and a canopy extending to the ground might have a similar effect.
Our estimates of porosity beneath tree canopies range from 10 to
Fig. 3. Event return period plot at the Van Wyck East site, a) open transect, station 5, 6, 7; b) vegetated transect, station 1, 2, 3, 4; Monitoring stations are shown by distanc
the road.
Z.Tong et al./ Environmental Pollution 201 (2015) 141e149 145

22%.Wind breaks of this density have been found to have a recir-
culation zone extending downwind between 8 and 10 times their
height (Heisler and Dewalle, 1988).
3.3. Experiment 3: LIE north
The third experiment showed the impact of wind direction
relative to the highway emission source. Though we were unable to
collect continuous measurements for all3 locations during the
entire period due to equipment problems,when all 3 instruments
were operating, extreme spikeswere absent beneath the tree
canopy and in the open field (Fig.8, SM9). North winds, that is
north from the highway, had an overriding effect regardless of tree
cover. However, random spikes were still observed beside the road,
indicating that vehicle-induced turbulence can cause extreme
events (Wang and Zhang, 2009). Beneath the canopy there were no
Fig. 4. Top) Normalized decay curves of PM2.5 concentration from the edge of the highway to the field based on 4 measurement locations from a period where the traffic condition
is steady (11:30 AM to 3:30 PM) at Van Wyck East site.Concentrations are normalized by the average concentration at station 1 (roadside).Bottom) Normalized decay curve for
0.3e0.4mm and 2e3mm size fractions along the open and vegetated transects at Van Wyck East site. Curves for both the open and vegetated transects were forced to decay to
same ambient background at station 7 (Van Wyck East) along the open transect.
Fig. 5. Koniograph of PM2.5 concentration for LIE South at 6 s sampling resolution.Measurements at three static stations are shown,a) roadside,station 1; b) 12 m,station 2; c)
52 m,station 3; Traffic and wind conditions were most steady during the selected period from 12:10 to 1:30 PM.
Z.Tong et al./ Environmental Pollution 201 (2015) 141e149146

concentration spikes (Fig. 8). The proportion of the PM population
falling in different size classes was affected by upwind sources and
hence wind direction plays a big role. When wind passed over open
park space upwind of the highway, the proportions of each particle
size were similar among all sample locations (Fig.9 a,b,c).The
0.3e0.4 mm and 0.4e0.5 mm classes accounted for nearly 75% of
PM 2.5 (Fig. 9 a,b,c). In contrast, south of the LIE with wind from the
north,the percentage of these size fractions dropped to ca.40% of
Fig. 6. Normalized decay curves for PM2.5 concentration, 0.3e0.4mm particle and 2e3mm particles as a function of distance from the edge of highway for the corresponding period
at LIE South site.The concentration of large particles declines more steeply with distance than does the concentration of small particles.The polydisperse class PM2.5 shows an
intermediate decay rate.The poorer fit for the small particles corresponds to the relatively higher concentration observed downwind of trees.
Fig. 7. Average turbulent kinetic energy (TKE) calculated as a function of distance away from the road at a) Van Wyck East,b) LIE South.At Van Wyck East,measurements were
taken from a station upwind of vegetation (4 m from road), and a station downwind of vegetation (53 m from the road) at two heights (2 m and 3 m above the ground). At L
three stations are shown as,roadside,downwind of vegetation barrier (12 m from road),and far field (52 m from road).
Fig. 8. Konigragh for LIE North site.“Roadside” is station 1.“Inside canopy” is location 2.The Open Field is location 3.The wind direction was from the park to the highway.A
problem with the monitor in the open field prevented data collection from the entire day.
Z.Tong et al./ Environmental Pollution 201 (2015) 141e149 147

PM 2.5 (Fig.9 d,e,f).
This is approximately equal to the contribution of larger parti-
cles (2e3 mm and 1.6e2 mm). It is informative to compare the
particle size distribution during concentration spikeswith the
average size distribution excluding spikes. All spikes exceeding two
standard deviations above the average are analyzed for the LIE
North and South sites. The pie chart displays a very different picture
in contrast with the proportion computed over the entire period at
roadside station (SM10). At LIE North (upwind of highway), greater
proportion of large particles (1.6e3mm) was observed,which is
probably due to wear products that are re-suspended from the
pavement by vehicle-induced turbulence (Wang and Zhang, 2009).
At LIE South site (downwind of highway),spikes monitored had a
higher proportion of small particles (0.3e0.4mm) when the wind
blew across the road. These particles are likely to include secondary
particles aggregating from tailpipe emissions as well as particles re-
suspended from the road.
4. Conclusions
Brief monitoring campaigns investigated the dispersion of par-
ticulates with high temporal and spatial resolution next to highway
sources. Only 2 of our hypotheses were supported by our findings,
namely the wind direction in relation to source and particle size
affect PM2.5 concentration. The remaining hypotheses dealt directly
with the positive effects that trees are expected to have on local air
quality.Neither was supported by the evidence:concentrations
were higher downwind of trees,leading to less steep decay rates.
Decomposing PM2.5 into discrete size classes ranging from 0.3mm
to 3 mm showed that the physical behaviors of each size class are
very different along both open and vegetated transect. Fitting PM2.5
concentration to exponential decay curves reveals that small par-
ticles (0.3e0.4mm) have lower extinction coefficients and longer
half distances in the air column than do larger particles (2e3mm).
Although the deposition behavior of different sizes of particles are
well known in the atmospheric science, this is typically ignored or
overlooked in the planning and design of green space intended to
mitigate air pollution.
An especially interesting finding is that the presence oftrees
between a source and measurementlocation reduced the fre-
quency and intensity ofconcentration spikes while at the same
time increasing average concentration.The net effect was that
concentrations along transects with trees declined less sharply
than along open transects. This accompanied decreased TKE
downwind of trees,indicating recirculation,longer particle resi-
dence times and decreased dispersion.In sum,aerodynamics,not
surface area available for deposition,controls local PM2.5 concen-
tration. This has direct implications for designing landscapes
intended to mitigate air pollution: downwind recirculation zones
must be dimensioned and we must avoid locating sensitive uses in
these zones.The laws of diffusion state that higher ambient con-
centrations result in a steeper concentration gradient,one of the
“driving forces” controlling deposition rate.Higher concentration
therefore means thatdeposition is enhanced in the downwind
recirculation zones while at the same time increasing human risk.
Our empirical findings comply with theory and echo what has been
called “the green paradox,” underscoring the need for multi-scale
approaches (Vos et al.,2013).
Differences between transects with and withouttrees were
apparent only when wind was from the direction of the road, hence
the effect trees have on particle concentration is contingent on
wind direction. The mass proportion of each particle size class from
the roadside concentration spikes is very different from their pro-
portion of the average concentration.The proportion of small
particles (0.3e0.4 mm) was greater downwind of the highway,
probably due to the effects of traffic.Upwind of the highway,the
proportions of each particle size were similar across sample loca-
tions, showing that wind direction overrides the traffic impact,
distance from source and canopy cover.
PM 2.5 is often reported as the daily average at a city scale.To
better understand health impacts,however,particle size distribu-
tions, temporal and spatial resolution of PM 2.5 may be more
informative. Smaller particles in PM2.5 will be inhaled more deeply
and are more toxic than larger particles of the same composition
(Ferin, 1994; Ferin et al., 1992). Studies also show that acute
exposure to traffic related PM was associated with elevated levels in
exhaled nitrate and nitrite,an early biological response that may
precede respiratory symptoms (Ezzati et al.,2000; Van Vliet et al.,
2013).Establishing links between spatiotemporal variation in PM
and human responses warrants far greater emphasis.
Ideally,trees could be located in landscapes to reduce the PM
dose humans receive. Dose is probabilistic, contingent on location,
timing,duration,traffic,and aerodynamics,among other variables
like regional air mass.If we intend to design landscapes to reduce
exposure to air pollution,then a simple “more trees are better”
approach is inadequate.Failure to account for the functional
Fig. 9. Mass proportion of each particle size class as shown in the legend.a,b,c) represent stations at LIE North site: roadside (station 1),inside canopy (station 2),and open field
(station 3) respectively.d,e,f) represent stations at LIE south site: roadside (station 1),behind vegetation barrier (station 2),and open field (station 3).
Z.Tong et al./ Environmental Pollution 201 (2015) 141e149148

mechanismsmay not achieve the desired outcomesand may
actually increases local pollution concentrations, with concomitant
negative effects on human health.For example,PM 2.5 concentra-
tion downwind of the highway was consistently higher (10% at
25 m) along the transect with trees than the open transect (Fig. 4).
The magnitude of this local increase is 40e200 times the calculated
reduction in PM2.5 by trees in 10 US cities (Nowak et al., 2013). How
does this spatially distributed effect translate into the monetary
value of the services attributed to urban trees? Ultimately,all hu-
man exposure is local. Designing landscapes to reduce exposure to
air pollution should start with operational definitions of biophysical
outcomes that can be empirically verified.For example,such defi-
nitions could include a specific half distance for the decay in con-
centration in a specific size class, or the total distance needed to de-
couple the concentration from roadside influence. This will require
both empirical field study and CFD modeling of the pollution gra-
dients and flow fields around trees and other landscape features.
Attention should shift from the current paradigm based on depo-
sition to the overriding influences of landscape aerodynamics and
dispersion. Trees define zones of recirculation, which in turn cause
greater deposition because they increase localpollution concen-
trations,while at the same time increase the dose received by
humans in these zones.This finding is especially important if we
are serious about using green infrastructure to improve localair
quality and hence,human health. As Nowak has been quoted,
“We're not going to be able to plant our way out of this problem”
(Kessler,2013).
Acknowledgments
We are grateful for the support from the following grants: USDA
(CREES) grants 2001-38875-10702 and 2004-38875-02190; USDA
(Forest Service) 05-DG-11244225-228CRIS0190768; and for
thoughtful comments from K. Max Zhang, Frederick Cowett, Aaron
Match, Hannah George and 2 anonymous reviewers.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.envpol.2015.02.026.
References
Armbruster,L.,Breuer,H., Gebhart,J., Neulinger,G., 1984.Photometric determina-
tion of respirable dust concentration without elutriation ofcoarse particles.
Part.Part.Syst.Charact. 1,96e101.
Baldauf,R.,Thoma,E.,Khlystov,A., Isakov,V., Bowker,G.,Long,T.,Snow,R.,2008.
Impacts of noise barriers on near-road air quality. Atmos. Environ. 42,
7502e7507.
Brantley,H.L.,Hagler,G.S.W.,Deshmukh,P.J.,Baldauf,R.W.,2014.Field assessment
of the effects of roadside vegetation on near-road black carbon and particulate
matter.Sci.Total Environ.468e469, 120e129.
Buccolieri,R., Gromke,C., Di Sabatino,S., Ruck,B., 2009. Aerodynamic effects of
trees on pollutant concentration in streetcanyons.Sci. Total Environ. 407,
5247e5256.
Cavanagh,J.-A.E.,Zawar-Reza,P.,Wilson, J.G.,2009.Spatial attenuation of ambient
particulate matter air pollution within an urbanised native forest patch.Urban
For.Urban Green.8, 21e30.
Dochinger, L.S., 1980. Interception of airborne particles by tree plantings. J. Environ.
Qual.9, 265e268.
Dominici,F., Greenstone,M., Sunstein,C.R.,2014.Particulate matter matters.Sci-
ence 344,257e259.
EPA, 2009. Integrated Science Assessment for Particulate Matter. US Environmental
Protection Agency,Washington,DC.
Ezzati, M., Saleh,H., Kammen,D.M., 2000. The contributions of emissions and
spatialmicroenvironments to exposure to indoor air pollution from biomass
combustion in Kenya.Environ.Health Perspect. 108,833e839.
Ferin, J., 1994.Pulmonary retention and clearance ofparticles.Toxicol.Lett. 72,
121e125.
Ferin, J., Oberd€orster, G., Penney, D., 1992. Pulmonary retention of ultrafine and fine
particles in rats.Am. J. Respir.Cell Mol. Biol. 6, 535e542.
Freer-Smith, P.H., El-Khatib, A.A., Taylor, G., 2004. Capture of particulate pollution by
trees: a comparison of species typicalof semi-arid areas (Ficus Nitida and
Eucalyptus globulus) with European and North American species. Water, Air, Soil
Pollut. 155, 173e187.
Gromke,C.,2011.A vegetation modeling concept for building and environmental
aerodynamics wind tunneltests and its application in pollutant dispersion
studies.Environ.Pollut. 159,2094e2099.
Gromke, C., Ruck, B., 2009. On the impact of trees on dispersion processes of traffic
emissions in street canyons.Bound.Layer Meteorol. 131, 19e34.
Gumbel,E.J., 1941.The return period of flood flows.Ann. Math. Stat. 12, 163e190.
Hagler, G.S.W., Lin, M.-Y., Khlystov, A., Baldauf, R.W., Isakov, V., Faircloth, J.,
Jackson,L.E., 2012.Field investigation ofroadside vegetative and structural
barrier impact on near-road ultrafine particle concentrations under a variety of
wind conditions.Sci.Total Environ.419,7e15.
HEI, 2010. Traffic Related Air Pollution: a Critical Review of the Literature on
Emissions,Exposure and Health Effects.Health Effects Institute.
Heisler,G.M., Dewalle,D.R.,1988.Effects of windbreak structure on wind flow.
Agric.Ecosyst.Environ.22e23, 41e69.
Kaimal, J.C., Finnigan, J.J., 1994. Atmospheric Boundary Layer Flows: Their Structure
and Measurement.
Kenney,W., 1987.A method for estimating windbreak porosity using digitized
photographic silhouettes.Agric.For.Meteorol.39, 91e94.
Kessler,R., 2013.Green walls could cut street-canyon air pollution.Environ.Health
Perspect. 121,a14.
Li, X.-X., Britter,R., Koh, T.,Norford,L., Liu, C.-H.,Entekhabi,D., Leung,D.C.,2010.
Large-Eddy simulation of flow and pollutant transport in urban street canyons
with ground heating.Bound.Layer Meteorol. 137, 187e204.
Loeffler,A.E., Gordon, A.M., Gillespie,T.J.,1992.Optical porosity and windspeed
reduction by coniferous windbreaks in Southern Ontario.Agrofor. Syst.17,
119e133.
Maher,B.A.,Ahmed,I.A.M.,Davison,B.,Karloukovski,V.,Clarke, R.,2013.Impact of
roadside tree lines on indoor concentrationsof traffic-derived particulate
matter.Environ.Sci.Technol.47, 13737e13744.
Murakami,M., Nakajima,F.,Furumai,H., 2005.Size- and density-distributions and
sources of polycyclic aromatic hydrocarbons in urban road dust.Chemosphere
61,783e791.
Nowak,D.J.,2002.The Effects of Urban Trees on Air Quality.USDA Forest Service,
Syracuse,NY.
Nowak, D.J., Hirabayashi, S., Bodine, A., Hoehn, R., 2013. Modeled PM2.5 removal by
trees in ten U.S. cities and associated health effects.Environ. Pollut. 178,
395e402.
Pugh, T.A.M., MacKenzie, A.R., Whyatt, J.D., Hewitt, C.N., 2012. Effectiveness of green
infrastructure for improvement of air quality in urban street canyons.Environ.
Sci.Technol.46, 7692e7699.
Raupach, M.R., Thom, A.S., 1981. Turbulence in and above plant canopies. Annu. Rev.
Fluid Mech. 13,97e129.
Raupach,M.R.,Woods,N., Dorr,G.,Leys,J.F.,Cleugh,H.A.,2001.The entrapment of
particles by windbreaks.Atmos.Environ.35, 3373e3383.
Sehmel, G.A., 1980. Particle and gas dry deposition: a review. Atmos. Environ. (1967)
14,983e1011.
Seinfeld,J.H., Pandis, S.N., 2006. Atmospheric Chemistry and Physics:from Air
Pollution to Climate Change.Wiley-Interscience.
Set€al€a, H., Viippola, V., Rantalainen, A.-L., Pennanen, A., Yli-Pelkonen, V., 2013. Does
urban vegetation mitigate air pollution in northern conditions? Environ. Pollut.
183, 104e112.
Slinn, W.G.N.,Hasse,L., Hicks, B.B.,Hogan,A.W., Lal, D., Liss,P.S.,Munnich, K.O.,
Sehmel, G.A., Vittori, O., 1978. Some aspects of the transfer of atmospheric trace
constituents past the air-sea interface.Atmos.Environ.(1967) 12,2055e2087.
Spengler,J., Lwebuga-Mukasa,J., Vallarino, J., Melly, S., Chillrud, S., Baker, J.,
Minegishi,T.,2011.Air Toxics Exposure from Vehicle Emissions at a US Border
Crossing: Buffalo Peace Bridge Study. Research report (Health Effects Institute),
pp. 5e132.
Steffens,J.T.,Wang, Y.J., Zhang,K.M., 2012.Exploration ofeffects ofa vegetation
barrier on particle size distributions in a near-road environment.Atmos.En-
viron.50, 120e128.
Tong,H.Y.,Hung, W.T.,Cheung,C.S.,2000. On-road motor vehicle emissions and
fuel consumption in urban driving conditions.J. Air Waste Manag.Assoc.50,
543e554.
Tong,Z.,Wang,Y.J.,Patel,M., Kinney,P.,Chrillrud, S.,Zhang,K.M., 2011.Modeling
spatial variations of black carbon particles in an urban highway-building
environment.Environ.Sci.Technol.46, 312e319.
Van Vliet, E.D.S.,Asante,K., Jack, D.W., Kinney, P.L.,Whyatt, R.M., Chillrud, S.N.,
Abokyi,L., Zandoh,C., Owusu-Agyei,S.,2013.Personalexposures to fine par-
ticulate matter and black carbon in households cooking with biomass fuels in
rural Ghana.Environ.Res. 127,40e48.
Vos, P.E.J., Maiheu, B., Vankerkom, J., Janssen, S., 2013. Improving local air quality in
cities: to tree or not to tree? Environ.Pollut. 183, 113e122.
Wang, Y.J., Zhang, K.M., 2009. Modeling near-road air quality using a computational
fluid dynamics model.Environ.Sci.Technol.43, 7778e7783.
Whitlow, T.H., Hall, A., Zhang, K.M., Anguita, J., 2011. Impact of local traffic exclusion
on near-road air quality: findings from the New York City “Summer Streets”
campaign.Environ.Pollut. 159,2016e2027.
Z.Tong et al./ Environmental Pollution 201 (2015) 141e149 149