Environmental Pollution 201 (2015) 141e149

Contents lists available at ScienceDirect

Environmental Pollution journal homepage: www.elsevier.com/locate/envpol

Quantifying the effect of vegetation on near-road air quality using brief campaigns Zheming Tong a, Thomas H. Whitlow b, *, Patrick F. MacRae b, Andrew J. Landers c, Yoshiki Harada b a b c

Department of Mechanical and Aerospace Engineering, Cornell University, Gruman Hall, Ithaca, NY, USA Section of Horticulture, School of Integrative Plant Science, Cornell University, Room 23 Plant Science Building, Ithaca, NY 14853, USA 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

a b s t r a c t

Article history: Received 18 August 2014 Received in revised form 18 February 2015 Accepted 20 February 2015 Available online 20 March 2015

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 particulates 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. This coincided with decreased Turbulence Kinetic Energy downwind of trees, indicating recirculation, longer residence times and decreased dispersion. Simply planting trees can increase local pollution concentrations, 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.

Keywords: Near-road air pollution Trees Dispersion Aerodynamics PM2.5

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-level air pollution concentrations in 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 resuspension, by rain and leaf abscission (Dochinger, 1980; FreerSmith 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 policies like the numerous million-tree programs across the US. However laudable these programs are, the approach ignores the effects of distance from source and the * Corresponding author. E-mail address: [email protected] (T.H. Whitlow). http://dx.doi.org/10.1016/j.envpol.2015.02.026 0269-7491/© 2015 Elsevier Ltd. All rights reserved.

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 10 mm whose deposition rates depends on inertial impaction and gravitational settling. The local turbulent flow field also plays a significant role in particle dispersion. 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 small scale 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).

142

Z. Tong et al. / Environmental Pollution 201 (2015) 141e149

Both experimental and numerical simulation studies have investigated the effect of 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 (0.5e10 mm aerodynamic diameter) were unaffected by vegetation. Baldauf et 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 concentrations of ultrafine and coarse mode PM with limited €l€ reduction measured for PM2.5 mass. Seta 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 PM10 was observed. Cavanagh et al. (2009) conducted a field study to investigate the spatial attenuation of PM10. Concentrations were higher outside the forest than deep within the forest. Other researchers have used physical models 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 experiment on dispersion processes of 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 experiments to study the aerodynamic effects of trees on pollutant concentration in street canyons. Both approaches showed considerably greater pollutant concentration near the leeward wall and slightly lower concentration near the windward wall when trees were present. Another CFD study compared CFD modeled results and field measurements explore the effect of a near-road vegetation 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 concentration, 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 PM10 (Pugh et al., 2012). This brief review shows that vegetation can either decrease or increase PM concentration, depending on the road-canopy configuration, 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. PM2.5 concentrations will be reduced below ambient downwind of tree canopies.

2. PM2.5 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 shortterm field campaigns exploring the spatiotemporal patterns of particulate matter dispersion 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 monitored in real time, are suited to addressing the effectiveness of vegetated buffers at scales relevant to 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 complex in Queens, New York City, USA (Map is shown in the Supplementary Material (SM1), and relevant features 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 vehicles/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 particles, 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 vegetation 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 differences in PM2.5 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) 141e149

143

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 and longitude is in decimal degrees. c) Canopy porosity was determined from hemispherical images taken beneath the canopy. More details are provided in the Supplementary 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

Van Wyck East

Jun 7, 2011

40.723

LIE South LIE North

Jul 13/14/15, 2011 Jul 12, 2011

40.741 40.743

Longitudeb

Porosityc

% Treesd

% Bare soil

% Grasse

73.838

15.7%

73.841 73.841

9.8% 21.9%

44.1%(w/trees) 0%(no trees) 4.3% 82.5%(w/trees) 6.3%(no trees)

10.5%(w/trees) 3.5%(no trees) 3.1% 17.0%(w/trees) 22.5%(no trees)

89.5%(w/trees) 96.5%(no trees) 66.7% 71.9%(w/trees) 64.7%(no trees)

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 (Model 1.108) equipped with isokinetic probes to reduce the effect of variation in wind speed. We monitored 15 size classes between 0.3 and 20 mm every 6 s. This approximates a human resting inhalation rate and also allows us to observe conditions corresponding to spikes in PM2.5 concentration. Instruments 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 empirical equations 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 0.3 mm 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 definition of PM2.5 because it excludes particles below the detection 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 East Roadside

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 measurements. 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.

10 m

23 m

40 m

Average concentration [mg/m ]

4.92(1.84) 3.90(1.33) 3.81(1.19) 3.75(1.27)

Vegetated transect at Van Wyck East

Roadside

Average concentration [mg/m3]

4.92(1.84) 4.36(0.94) 4.46(1.11) 4.17(0.81)

3

7m

15 m

51 m

Vegetated Transect at LIE South Roadside

12 m

Average concentration [mg/m3]

1.77(0.74) 1.65(0.64)

1.96(0.84)

52 m

144

Z. Tong et al. / Environmental Pollution 201 (2015) 141e149

limit of the instruments (0.3 mm). 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 turbulent kinetic energy (TKE, see below and SM4). 2.3.3. Hemispherical canopy porosity and tree cover Porosity is a property of vegetation that correlates well with 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 hemispherical image is provided in the Supplementary Material. We also estimated tree cover using line intercepts 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.

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 concentration data (Fig. 2a), showing frequent spikes above background, corresponding to passage of especially “dirty” vehicles. Because our spectrometers cannot detect particles

Quantifying the effect of vegetation on near-road air quality using brief campaigns.

Many reports of trees' impacts on urban air quality neglect pattern and process at the landscape scale. Here, we describe brief campaigns to quantify ...
2MB Sizes 0 Downloads 8 Views