Environmental Pollution 194 (2014) 96e104 Contents lists available at ScienceDirect Environmental Pollution journal homepage: www.elsevier.com/locate/envpol Assessing the relationship among urban trees, nitrogen dioxide, and respiratory health Meenakshi Rao a, Linda A. George a, *, Todd N. Rosenstiel b, Vivek Shandas c, Alexis Dinno c a School of the Environment, Department of Environmental Science and Management, Portland State University, Portland, OR, USA Department of Biology, Portland State University, Portland, OR, USA c Nohad A. Toulan School of Urban Studies and Planning, Portland State University, Portland, OR, USA b a r t i c l e i n f o a b s t r a c t Article history: Received 13 January 2014 Received in revised form 16 May 2014 Accepted 12 July 2014 Available online Modeled atmospheric pollution removal by trees based on eddy flux, leaf, and chamber studies of relatively few species may not scale up to adequately assess landscape-level air pollution effects of the urban forest. A land use regression (LUR) model (R2 ¼ 0.70) based on NO2 measured at 144 sites in Portland, Oregon (USA), after controlling for roads, railroads, and elevation, estimated every 10 ha (20%) of tree canopy within 400 m of a site was associated with a 0.57 ppb decrease in NO2. Using BenMAP and a 200 m resolution NO2 model, we estimated that the NO2 reduction associated with trees in Portland could result in significantly fewer incidences of respiratory problems, providing a $7 million USD benefit annually. These in-situ urban measurements predict a significantly higher reduction of NO2 by urban trees than do existing models. Further studies are needed to maximize the potential of urban trees in improving air quality. © 2014 Published by Elsevier Ltd. Keywords: NO2 Land use regression Urban forest Health impacts 1. Introduction Epidemiological research has established that urban air pollutants such as NO2, PM2.5 and O3 can be detrimental to human health. An increase in the average air pollution in a city is correlated with an increase in cardiovascular disease, strokes and cancer (Brunekreef and Holgate, 2002; Dockery et al., 1993; Nyberg et al., 2000; Pope et al., 2002; Samet et al., 2000; Samoli et al., 2005). More recent epidemiological research has shown that the health impacts of air pollution are not uniform across a city. For example, numerous studies show a higher burden of respiratory problems close to major roadways (Brauer et al., 2007; Jerrett et al., 2008; McConnell et al., 2006; Ostro et al., 2001), which is not surprising as primary air pollutants levels are greatest near the source and decay rapidly away from it (Faus-Kessler et al., 2008; Gilbert et al., 2007; Jerrett et al., 2005). A meta-analysis by Karner et al. (2010) shows that air pollutants within cities decay rapidly within 200 m of the source, reaching background concentrations between 200 m and 1 km, creating strong air pollution gradients at short spatial scales within a city. * Corresponding author. E-mail address: georgeL@pdx.edu (L.A. George). http://dx.doi.org/10.1016/j.envpol.2014.07.011 0269-7491/© 2014 Published by Elsevier Ltd. To address the challenge of reducing human exposure to urban air pollution, then, we need to monitor or model air pollutants at a spatial resolution of 200 m or finer. To date, however, institutional observations, monitoring and modeling efforts have primarily focused on the regional and global scales. Active monitoring stations such as those in the US Environmental Protection Agency (US EPA) monitoring network, satellite observations, and atmospheric transport models provide air pollution data at the 10 km or coarser spatial scale. Chemical transport models such as CMAQ and WRFChem that could be used to model air pollutant levels at the intra-urban scale lack emissions inventories as well as model validation studies at this scale. Land use regression (LUR), a method that has been widely used by epidemiologists, (Hoek et al., 2008; Jerrett et al., 2005; Ryan and LeMasters, 2007) is well-suited to model the intra-urban variability of air. LUR combines measurements of air pollution and statistical modeling using predictor variables obtained through geographic information systems (GIS). The European Union (EU), for example, is currently using LUR in the ESCAPE project, which aims to model the intra-urban variability of several urban air pollutants (Beelen et al., 2013; Cyrys et al., 2012; Eeftens et al., 2012). Further, we need to understand the factors such as distance from source, terrain, deposition onto the urban forest, photochemical environment and local meteorology that affect the dispersion of air pollutants within a city at this highly local scale of M. Rao et al. / Environmental Pollution 194 (2014) 96e104 200 m. This information is critical for urban dwellers, planners and policy-makers seeking to create healthier cities. However, many of the factors affecting dispersion of air pollutants at these smaller spatial scales are not well understood; specifically, the role of vegetation in urban air pollution is poorly understood. Vegetation pays a complex role in the urban ecosystem, potentially contributing both positively and negatively to urban air pollution. For example, biogenic volatile organic compounds (BVOCs) emitted by trees react with urban NOx emissions to produce aerosols (a component of PM2.5) and ozone, both urban air pollutants regulated by the US EPA, the EU and the World Health Organization. In addition, several recent studies at the household and neighborhood scales have found an improvement in human health associated with urban greenery, particularly trees (De Vries et al., 2003; Donovan et al., 2011; Maas et al., 2006), although the explicit relationship between the urban forest, air pollution reduction and human health is not understood. While eddy flux, leaf and chamber studies clearly demonstrate the physiological potential for vegetation to remove air pollutants from the atmosphere (Fujii et al., 2008; Min et al., 2013; Sparks, 2009; Takahashi et al., 2005), landscape level studies show mixed results. For instance, Yin et al. found a 1e21% reduction in NO2 associated with park trees in Shanghai, China (Yin et al., 2011), while Setala et al. found no effect €la € et al., associated with NO2 and trees in Helsinki, Finland (Seta 2012). However, UFORE (I-Tree, 2011; Hirabayashi et al., 2012) the big-leaf model based on leaf and canopy level deposition studies, scaled to landscape levels, indicates that the urban forest reduces air pollution by < 1% (Nowak et al., 2006). Our goal for this study is to develop an urban, observationbased, predictive model of NO2 at the highly spatially resolved scale of 200 m and to assess the relative strengths of sources and sinks of NO2 in the urban environment, focusing especially on vegetation. Further, through application of this high resolution NO2 model, we can also estimate the economic value of the health benefits provided by trees through the reduction of NO2. Here, we focus on NO2, a strong marker for anthropogenic air pollution, as it can be measured accurately and simultaneously at a large number of sampling locations. 2. Materials and methods 2.1. Field campaign Our study area is the Portland Metropolitan Area, a mid-size urban area covering 1210 km2, with a population of ~ 1.5 million, located in the state of Oregon, in northwestern USA. It is situated at 45.52 N, 122.68 W, and has a temperate climate with relatively dry summers. Portland is home to Forest Park, one of the large forests within urban boundaries in the USA. Portland's urban forest is predominantly deciduous, with big-leaf maple, black cottonwood and Douglas fir constituting >50% of the urban forest, based on a public tree assessment (Portland Parks and Recreation, 2007). Two rivers flow within the city boundaries e the Willamette and the Columbia, with an active port on the Columbia. The Portland Metro area has some hilly terrain, especially west of the Willamette, with a maximum elevation of 387 m. NOx varies between summer and winter in Portland (summer and winter 2013 NO2 averages were 7.5 ppb and 11.4 ppb respectively). However, since we were interested in assessing the effect of vegetation on local NOx, we focused our earlier sampling in the summer. NO2 and NO were measured at 144 sites in the Portland Metro area using passive Ogawa samplers (only the NO2 results are presented here; NO results will be discussed in a future paper). Sites were chosen using a spatial allocation model coupled with a stratified random approach to encompass the spatial extent of the Portland Metro area and to capture the effect of roads, railroads and vegetation on ambient NO and NO2. A single passive Ogawa sampler, with an NO2 pad on one side and a NOx pad on the other, was placed at each site between 2 m and 3 m above ground. Controls were co-located at the Portland State University monitoring station, which actively monitors NO and NO2 using a calibrated chemiluminescent NOx monitor (Teledyne NOx Analyzer, Model T200). Lab and field blanks were also deployed to detect contamination during assembling the samplers or excess exposure during transportation. Samplers were placed in the field 23rd e 25th Aug 2013, and retrieved 3rd e 5th Sep 2013, for an approximate field exposure of 12 days. Samplers were analyzed in the lab on 6th Sep 2013 using the methodology outlined in the Ogawa manual (Ogawa & Co., USA, 2006) and corrected for temperature and relative 97 humidity based on measurements at the Portland State University air quality station. The field and lab blanks readings were all low (with an average reading of ~ 0.2 ppb NO2). The NO2 measured by five co-located Ogawa samplers (average 17.6 ± 0.5 ppb) was within 5% of the PSU chemiluminescence monitor ambient reading (average 16.9 ppb). (See Fig. S1 for map of sites and measured NO2). 2.2. Land-use regression (LUR) Briefly, LUR (Briggs et al., 2000; Hoek et al., 2008; Jerrett et al., 2005; Ryan and LeMasters, 2007) is a statistical modeling technique used to predict air pollutant concentrations at high resolution across a landscape based on a limited number of measurements of the pollutant of interest within the study area. Land use and land cover variables are extracted at each measurement site using a spatial analysis program and a regression model developed, with the air pollutant measurements as the dependent variable and the land use parameters as the independent variables. For this study, we constructed two LUR models. The first model, the sources and sinks model (SSM), was specifically developed to examine the relative strengths of sources and sink of NO2 in an urban environment. For the SSM model, we considered only those land use and land cover variables that were proxies for known urban sources and sinks of NO2. Land use and land cover proxies were identified based on a previous LUR model for Portland (Mavko et al., 2008), existing literature on LUR models (Beelen et al., 2013; Henderson et al., 2007; Hoek et al., 2008; Jerrett et al., 2005; Ryan and LeMasters, 2007), and knowledge of sources and sinks of urban NO2. In all, we identified four classes of roadways, length of railroads, industrial area, population, tree canopy area, and area with grass and shrubs as proxies for urban sources and sinks of NO2 (Table 1). The second model we developed was a predictive model (PM) to assess the health impacts of NO2. In this model, we wanted to have the best model fit (R2), and hence did not constrain the independent variables to be proxies for urban sources or sinks of NO2. The PM includes all the independent variables identified by the SSM and adds latitude and longitude to the regression variables. While latitude and longitude are neither sources nor sinks of NO2, these terms capture the spatial variability of the sources and sinks in the Portland Metro area, and hence improve the model fit. All spatial analysis was done using ArcMAP® 10.1 by ESRI. Table 1 summarizes the land use and land cover variables used in the study and the data source. In addition, elevation, latitude and longitude were associated with each site. Each land use and land cover variable was extracted for each of the 144 sites using spatial analysis in 24 circular buffers ranging from 50 m to 1200 m in 50 m increments. We considered using wind buffers as was done in the previous Portland LUR study (Mavko et al., 2008). However, we found that the average wind direction varied widely across our study area and could not be modeled using a single wind direction, as was done by Mavko et al., due the smaller spatial extent of their study area. In all, we extracted more than 200 land-use and land-cover variables. We did not consider area under grass and shrubs or latitude further in the model as neither showed any correlation with NO2 at any buffer size between 50 m and 1200 m. We also did not consider industrial area as it was very highly correlated with both railroads and major arteries. Our approach allowed us to build a parsimonious model for teasing out the relative strengths of primary sources and sinks of NO2 in the study region. For the land use and land cover variables that showed a correlation with NO2, we identified the appropriate buffer size for each land use variable based on correlation with NO2 (Clougherty et al., 2008; Henderson et al., 2007). Table 1 Land use/land cover variables used in LUR, data source and NO2 source/sink proxy. Land use/land cover variable Data source Proxy Freeways (length) AADT* (traffic volume) Major arteries (length) Arteries (length) Streets (length) Railroads (length) Industrial area (area) Population (number) RLIS 2012a NHPN 2007b RLIS 2012a RLIS 2012a RLIS 2012a RLIS 2012a RLIS 2012a RLIS 2012a (based on 2010 census) City of Portland, 2010 Traffic emissions Traffic emissions Traffic emissions Traffic emissions Traffic emissions Railroad emissions Industrial point sources Area sources Sink through deposition Area under tree canopy (area) Area under shrubs/ herbaceous cover (area) Elevation (height) Latitude City of Portland, 2010 Sink through deposition RLIS 2012a Measured/Google Earth Longitude Measured/Google Earth Potential sink (wind flow) Spatial variability of sources & sinks Spatial variability of sources & sinks *AADT: Annual average daily traffic. a Regional Land Information System, Metro Resource Data Center. b National Highway Planning Network. 98 M. Rao et al. / Environmental Pollution 194 (2014) 96e104 Fig. 1. NO2 quantiles, based on the predictive NO2 LUR model (PM) applied at a 200 m spatial resolution. We followed a cross-validation design in developing both regression models. The monitored data was divided into 5 sets. Four sets were combined at a time to create 5 sets of training data, with the excluded set kept aside for validation. A hierarchical nested regression analysis (Cohen et al., 2003) was also done on the SSM to estimate the direct and indirect contributions of each land use category to urban NO2. All statistical analyses were done in SPSS 19. We developed five spatial distributions of NO2; to estimate the effect of scale and tree canopy on respiratory health: (i) A “rural background” NO2 spatial distribution, assuming a uniform NO2 distribution of 0.1 ppb (estimated level in rural areas upwind of Portland) (Lamsal et al., 2013) across the Portland Metro area. (ii) A “regional” NO2 spatial distribution, based on the average Oregon DEQ NO2 measurement for the period of the study (7.5 ppb). (iii) A 200 m predictive NO2 spatial distribution across Portland Metro, generated by applying the predictive LUR model to points on a 200 m grid (the PM model) (iv) A 200 m sources-and sinks spatial distribution of NO2 across Portland Metro, generated by applying the sources-and-sinks model to points on a 200 m grid (the SSM model) (v) A 200 m resolution “no trees” map of NO2 in the Portland area modeling the NO2 levels in the absence of trees, generated by applying the sources-andsinks model without the sink term associated with tree canopy, to points on a 200 m grid (the SSM model without trees). 2.3. Respiratory health impact analysis Analysis of several asthma-related endpoints e including asthma exacerbation resulting in missed school among children aged 4e12; asthma exacerbation resulting in one or more symptoms among children 4e12; emergency room visits for asthma at any age; and hospital admissions (HA) for any respiratory condition among elderly persons aged 65 and over e to estimate incidence and economic valuation was done in The Environmental Benefits Mapping and Analysis Program version 4.0.35 (BenMAP) (U.S. Environmental Protection Agency, 2010). BenMAP is a Windows-based computer program developed by the US EPA that uses a Geographic Information System (GIS)-based approach to estimate the health impacts and economic benefits (or dis-benefits) occurring when populations experience changes in air quality. BenMAP comes with multiple built-in regional and national datasets, including health impact functions and baseline incidences, to facilitate health benefits modeling. It has been used to estimate the health impacts of urban air pollutants at the regional and national scales (Davidson et al., 2007; Fann et al., 2009; Hubbell et al., 2009, 2004). For this study, we estimated the incidence and valuation for four respiratory endpoints of NO2, using the health impact and valuation functions built into BenMAP (Table S1). We used Popgrid, a population allocation tool that comes with BenMAP to allocate the census population into age bins required by BenMAP (Abt Associates Inc., 2010). We estimated the respiratory health impacts of NO2 on the Portland population at two different spatial scales of assessment. At the city-scale, we used the “regional” NO2 distribution based on the Oregon DEQ NO2 measurements. For the highly spatially resolved scale, we used our 200 m predictive NO2 LUR model (PM model). Incidences of respiratory outcomes for both scales were assessed against an estimated rural background of 0.1 ppb NO2 (Lamsal et al., 2013). The respiratory health outcomes under these two different scale scenarios were compared to evaluate the role of scale in health impact assessments. Respiratory health benefits of trees due to reduction in NO2 associated with tree canopy were assessed by comparing the respiratory outcomes based on the 200 m NO2 distribution generated using the SSM LUR model with and without the sink term for tree canopy. 3. Results and discussion 3.1. LUR models Two LUR models were developed based on the NO2 measured at 144 sites (Fig. S1) and the adjoining land use. The SSM fit estimated M. Rao et al. / Environmental Pollution 194 (2014) 96e104 NO2 using only land-use proxies for urban sources and sinks of NO2. The PM added longitude (as X_DIST) to the SSM to account for the spatial variability of land use within the Portland Metro area (Fig. 1). The average adjusted R2 (across 5 training models) and average RMSE (across 5 validation models) was 0.70 (2.6 ppb) and 0.80 (2.2 ppb) for the sources and sinks (SSM) and predictive models (PM), respectively. The R2 for the models is consistent with published R2 values, ranging from 0.50 to 0.90, while the RMSE is on par with the lowest measured RMSE values (1.4e34 ppb) (Hoek et al., 2008). The coefficients of the two LUR models are similar for roadways and area sources, but differ for railroads, tree canopy and elevation. This is an indication that roadways and population are relatively evenly distributed across the Metro area, while railroads, trees and elevation show a strong spatial gradient across the city, consistent with the local geography. Based on the regression equation (Eq (2)) ambient NO2 in Portland in the absence of considered land use sources and sinks is 7.7 ppb. Ambient NO2 levels increase by 1.1 ppb for every 100 000 vehicle kilometers traveled (annually) on freeways within a 1.2 km buffer of the site. NO2 levels increase by an additional 0.65 ppb for each kilometer of major arteries within 500 m of the site, and 1.7 ppb NO2 for each kilometer of arteries within 350 m. However, ambient NO2 levels decrease by 0.57 ppb for every 10 ha of trees (20% tree cover) within 400 m of a site (See S2 for descriptive statistics of model predictor variables). Sources and sinks model (SSM). 3.2. Scale & health impact assessment Based on the 200 m predictive NO2 spatial distribution in BenMAP (on an annualized basis, i.e., assuming comparable levels of NO2 over the year, and assuming the same association of reduced NO2 with trees across all seasons), we estimate 140 370 excess cases annually of asthma exacerbation in 4e12 year-olds, valued at $30 million (2013 USD), over a rural background NO2 background of 0.1 ppb. Further, we estimate 384 incidents (annually) of ER visits due to NO2-triggered asthma and 423 incidents of hospitalization in the elderly due to respiratory problems triggered by NO2. Altogether, the four NO2 triggered end-points considered here result in an economic cost to society of roughly $46 million (2013 USD). Assessing the same respiratory health impacts, using the same population distribution, but using the regional NO2 level for Portland for the summer sampling period instead, we estimate 105 819 excess cases of asthma exacerbation in 4e12 year-olds annually, 280 excess ER visits and 296 excess hospitalizations among the elderly attributable to urban NO2. Altogether the four endpoints considered at this scale result in an estimated $34 million (2103 USD) cost to society (Table 2). From this analysis, it is clear that the spatial resolution of the NO2 estimates used to assess the health impacts makes a significant difference in the magnitude of predicted health outcome. Specifically for Portland, the 200 m scale predictive NO2 model results in an estimate 30e45% higher than the number of incidences and economic cost as the uniform regional value. Even keeping in mind NO2 i ¼ 9:4 þ 1:2 Â 10À8 *FWY AADT1200;i þ 5:0 Â 10À4 *MAJ ART500;i þ 1:8 Â 10À3 *ARTERIES350;i þ 1:7 Â 10À8 *STREETSðPOPÞ800;i þ 1:5 Â 10À3 *RAILS250;i À 3:5 Â 10À3 *ELEVATIONi À 8:4 Â 10À6 *TREES400;i Adj R2 ¼ 0.70, validation RMSE ¼ 2.6 Predictive model (PM). 99 (1) that different health impact and valuation functions will result in different health and economic cost estimates, the critical point e that the incidence and valuation increase significantly for Portland when using the 200 m map of NO2 e holds. NO2 ðiÞ ¼ 7:7 þ 1:1 Â 10À8 *FWY AADT1200;i þ 6:5 Â 10À4 *MAJ ART500;i þ 1:7 Â 10À3 *ARTERIES350;i þ 1:8 Â 10À8 *STREETSðPOPÞ800;i þ 1:0 Â 10À3 *RAILS250;i À 1:0 Â 10À2 *ELEVATIONi þ 1:4 Â 10À5 *ðELEVATIONi Þ2 À 5:73 Â 10À6 *TREES400;i þ 1:1 Â 10À4 *X DISTi (2) Adj R2 ¼ 0.80, validation RMSE ¼ 2.2. Where: NO2(i) e NO2 ppb, at site (i) FWY_AADT1200,i e freeway (m) in 1200 m, weighted with AADT MAJ_ART500,i e major arteries (m) in 500 m ARTERIES350,i e arteries (m) in 350 m STREETS(POP)800,i e streets (m) in 800 m, weighted by the population RAILS250,i e railroads (m) in 250 m ELEVATIONi e elevation (ft) TREES400,i e tree cover (m2) in 400 m X_DISTi e distance from center of city (in m), along EeW axis For a mid-size city like Portland with relatively clean air and US criteria pollutants below US standards, for an urban air pollutant like NO2, which is related to relatively mild respiratory health outcomes, the single regional value for NO2 underestimates the annual respiratory health impacts on the order of $10 million (2013 USD). The majority of EPA monitors are focused on assessing areawide air quality, and are required to be sited to minimize near-road influences (Ambient Air Quality Surveillance, 2007). Thus, we can expect underestimation of air pollutants to hold for other air pollutants such as PM2.5 whose health impacts include mortality, neurological, pulmonary, and cardiovascular diseases, particularly in larger cities that show greater variation in intra-urban 100 M. Rao et al. / Environmental Pollution 194 (2014) 96e104 Table 2 Comparison of incidence and valuation of respiratory problems attributable to NO2 at the 200 m and regional scale. All valuations are in 2013 USD. Health impact Incidence estimate 200 m LUR NO2 Economic valuation (in $1,000,000) 200 m LUR NO2 Incidence estimate regional NO2 Economic valuation (in $1,000,000) regional NO2 Asthma exacerbation, missed school days (4e12 years) Asthma exacerbation, one or more symptoms (4e12 years) Emergency room visits, asthma (all ages) HA, all respiratory (65 and older) 47,918 140,370 384 423 5.55 29.70 0.16 10.58 36,239 105,819 280 296 4.20 22.39 0.12 7.43 Estimated cost $45.99 million distribution of urban air pollutants. This disparity between the highly resolved spatial scale and a single regional representative value is likely to grow even more when the cumulative health impacts of all urban air pollutants are taken into account. Thus, this analysis emphasizes the need for spatially (and temporally) resolved air pollutant data to more accurately assess the health, social, and economic costs of urban air pollution. 3.3. Relative strength of sources and sinks LUR models have been extensively used to capture the intraurban variability of urban air pollutants in epidemiological studies that focus on establishing health impact functions for urban air pollutants (Cyrys et al., 2012; Eeftens et al., 2012; Henderson et al., 2007; Jerrett et al., 2005). To our knowledge, LUR models have not been used to examine the relative strength of sources and sinks of urban air pollutants within cities. To examine the relative strengths of sources and sinks of NO2, we look at the SSM from several different perspectives (Table 3). The standardized regression coefficients (betas) in the SSM show that the freeway traffic volume within 1.2 km of a site has the strongest association with local NO2. One standard deviation increase in freeway traffic volume corresponds to a 0.4 standard deviation increase in NO2 levels at a site. Roadways, taken together, have a strong association with NO2 e a combined beta of 0.7. The association of trees and elevation with NO2 is much weaker, corresponding to a reduction in local NO2 levels by 0.2 and 0.15 standard deviations for each standard deviation increase in tree canopy and elevation respectively. Another way to examine the relative strengths of NO2 sources and sinks is to see how much each source (sink) increases (decreases) the background NO2. The modeled urban NO2 background, that is the NO2 level in the absence of LUR sources and sinks, is 9.4 ppb. The average traffic volume value across the 144 sites of 120 Â 103 vehicle-km increases the ambient NO2 by16% (1.5 ppb) over background. Roadways, considered together, on average across the 144 sites, increase NO2 levels by 25% (2.4 ppb) over background. Trees, on average, reduce the NO2 by about 15% of background (1.4 ppb). From a planning perspective, it is also important to know the range of change in NO2 levels associated with a specific source or sink that can be encountered in the city. Based on the SSM and the $34.14 million range of values observed for the land use variables in the Portland Metro area (Table 3), we find that freeway traffic can increase NO2 from 0 to 7.9 ppb above the 9.4 ppb background. Roadways taken together can increase NO2 from 0 to 11.8 ppb; railroads can increase NO2 from 0 to 3.3 ppb; while tree canopy can decrease NO2 from 0 to 4.2 ppb from the urban background. We further used hierarchical nested regression analysis (Cohen et al., 2003) to address the question of whether the reduction in NO2 seen with elevation and tree canopy could be attributed simply to the absence of sources, as sites at high elevation and with dense tree cover are less likely to have high traffic volume roads, railroads and area sources in the vicinity. The hierarchical analysis shows that about 30% of the total reduction in NO2 associated with elevation is directly due to elevation, while 38% is related to increased tree canopy at higher elevation, 12% due to fewer railroads, 14% due to fewer area sources and 6% due to lower road traffic. Similarly, 56% of the reduction in NO2 associated with trees is directly associated with tree canopy, while 14% is due to fewer area sources and 30% due to fewer roadways and lower traffic volume. LUR analyses using ordinary linear regression are not able to disentangle these effects. An assessment of the relative strengths of sources and sinks of air pollutants within a city is important and relevant data for determining optimum air pollution strategies. Chemical transport models, which are used extensively to evaluate mitigation policies at the regional scale, do not encode local dispersion phenomenon such as the urban canyon effect, the role of trees in changing wind flow, or deposition of air pollutants to the urban forest. Another shortcoming is the lack of emissions inventories and validation at the intra-urban scale. LUR is a technique that can be readily used to both inform policy and chemical transport model adaptation to the intra-urban scale. Of course, it is necessary to develop LUR models for diverse cities, collecting and analyzing the data using a standard methodology, along the lines the ESCAPE project has undertaken for determining health impact functions for the European population. 3.4. Role of vegetation Based on the SSM LUR model for NO2 in the Portland Metro area, every 10 ha of trees within 400 m of a site is associated with a 0.57 ppb reduction in NO2 at the site in summer. Of this reduction, Table 3 Comparison of relative strengths of sources and sinks of urban NO2 based on land use proxies. FWY_AADT1200 MAJ_ART500 ARTERIES350 STREETS(POP)800 RAIL250 ELEVATION TREES400 Std beta (avg of 5 training models) % of background NO2 on average NO2 contrib (ppb) based on avg land-use values for 144 sites NO2 contrib (ppb) based on max land-use values for 144 sites 0.414 0.121 0.189 0.222 0.164 À0.148 À0.179 15.8% 4.3% 5.6% 6.6% 5.6% À10.8% À14.7% 1.5 ppb 0.4 ppb 0.5 ppb 0.6 ppb 0.5 ppb À1.0 ppb À1.4 ppb 7.9 ppb 2.0 ppb 1.9 ppb 2.7 ppb 3.3 ppb À4.0 ppb À4.2 ppb M. Rao et al. / Environmental Pollution 194 (2014) 96e104 101 Fig. 2. Quantile map showing the modeled reduction of NO2 (as percentage of background) attributable to tree canopy, based on the NO2 SSM LUR model, with map of tree canopy above for comparison. 102 M. Rao et al. / Environmental Pollution 194 (2014) 96e104 Table 4 Incidence and valuation of potential respiratory health benefits due to reduction in NO2 attributable to tree canopy. Valuations are in 2013 USD. Health impact (Annualized) Reduced incidence due to trees Valuation of reduced incidence (in $1,000,000 USD) Asthma exacerbation, missed school days (4e12 years) Asthma exacerbation, one or more Symptoms (4e12 years) Emergency room visits, asthma (all ages) HA, all respiratory (65 and older) 7380 21,466 54 46 0.85 4.55 0.03 1.16 Potential estimated respiratory health benefit due to trees 56% is directly associated with trees, while 14% is due to the absence of area sources and 30% due to fewer roads and lower traffic volumes associated with treed area. Fig. 2 maps the reduction in NO2 associated with the presence of trees (as % of the background) in the Portland Metro area. The reduction in NO2 ranges from <1% of background (<0.1 ppb) to a maximum of 45% of background ( ~ 4 ppb). Not surprisingly, the greatest reduction in NO2 is in Forest Park, a 2092 ha forest within the Portland Metro area, while the least reduction is in the industrial areas in North Portland, which have very little tree cover. Our model shows a correlation between reduction in the urban air pollutant NO2 and trees. However, without being able to relate this reduction to known mechanisms through which trees affect NO2, i.e., through wet and dry deposition, changing airflow, accelerating chemical transformation, we cannot discount unknown factors associated with trees themselves. Nevertheless, it is instructive to estimate the potential health benefits associated with trees due do this statistical reduction of NO2 as a way to assess the viability of trees as a mitigation strategy. We used a BenMAP simulation to estimate both the incidence and economic valuation of the decrease in respiratory problems attributable to reduction in NO2 by trees (Table 4). The potential annual respiratory health benefit associated with trees in Portland due to reduction in NO2 is approximately 21,000 fewer incidences and 7000 fewer days of missed school due to asthma exacerbation for 4e12 year-olds; 54 fewer ER visits across people of all ages; and 46 fewer cases of hospitalization due to respiratory problems triggered by NO2 in the elderly. The economic value of these health benefits is approximately $7 million (2013 USD). Our findings that trees are associated with a significant reduction in NO2 is not unique: previous epidemiological LUR models include a term for trees or green spaces with a negative sign (Dijkema et al., 2011; Gilbert et al., 2005; Kashima et al., 2009; Mavko et al., 2008; Novotny et al., 2011). Here, however, we show for the first time e in our understanding e that this landscape-level reduction in NO2 associated with trees in Portland is large enough to make a discernible contribution to improved human respiratory health. Estimates using the big-leaf model UFORE indicate, however, that the urban forest in Portland removes only approximately 0.6% of atmospheric NO2 through deposition and foliar uptake (Nowak et al., 2006), which suggests that refinement of these deposition and uptake values are required or that other mechanisms may be the dominant mechanisms for landscape-level NO2 removal by trees in our study. In general, UFORE models of air pollution removal for US cities show <2% air pollutant removal by trees, leading urban ecologists (Pataki et al., 2011) to question the efficacy of urban tree plantings in mitigating air pollution. Our observations of the large reduction in NO2 associated with trees at the landscape level and the magnitude of the associated health impacts serve to highlight the need to understand, quantify and model the mechanisms through which trees impact urban air pollution at the landscape level, so we can effectively incorporate trees into the urban environment, balancing their benefits and disbenefits. One avenue to explore in understanding the role of the urban forest in air pollution mitigation is to determine whether the $ 6.59 million current generation of UFORE significantly underestimates the potential for tree-associated reduction of NO2 in urban areas. If the big-leaf model is indeed correct, then possibly other, less seasonal, mechanisms may dominate landscape level NO2 reduction, and we may find landscape-level winter reduction of NO2 to be roughly on par with our observed summer reduction. Eventually, speciesspecific measures of NO2 removal by intact urban canopies would provide the necessary foundation for a metabolically informed rational design of urban forest canopies, where key tree species are planted intentionally to maximize local NO2 removal, while taking into consideration the complex role of the urban trees in ozone production, allergen production, effect on local wind dispersion. Trees are an integral part of many urban environments, and hence, in an era of increasing global urbanization, it becomes even more important to understand the various mechanisms through which trees, of diverse forms and functions, are associated with reduced NO2, and potentially other more harmful urban air pollutants such as PM2.5. 3.5. Summary and conclusions We live in a rapidly urbanizing world e more than half of the world's population lives in cities today, and more than two-thirds will live in urban areas by 2050 (United Nations, 2011). An unintended consequence of increasing urbanization is an increase in anthropogenic emissions due to increased human activity, which in turn means more people are exposed to air pollution, potentially leading to reduced life expectancy, reduced productivity and a decrease in quality of life for urban dwellers (Straf et al., 2013). Due to the geographic variation in the distribution of air pollutants in a city, the health impacts are not uniform and tend to be increasingly borne by susceptible and socially disadvantaged urban populations (Clougherty and Kubzansky, 2009; Clougherty et al., 2007). Our study demonstrates the need to monitor or model air pollutants at a highly local scale in order to correctly assess the health impacts of urban air pollutants and to address social equity issues. Our study is further suggestive of the potential of an urban forest to reduce the air pollutant NO2 and hence provide health benefits on the order of millions of dollars (on an annualized basis) due to reduced incidence of respiratory problems. It emphasizes the need to resolve the NO2 conundrum so urban planners and urban foresters can better understand if and how trees may be more effectively incorporated into urban designs for healthier cities. Acknowledgments The authors gratefully acknowledge support from the US Forest Service Award #2011-DG-11062765-016. We would also like to thank the numerous volunteers that helped with the field campaign. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.envpol.2014.07.011. M. Rao et al. / Environmental Pollution 194 (2014) 96e104 References Abt Associates Inc, 2010. BenMAP environmental benefits mapping and analysis program user's manual appendices. U.S. Environ. Prot. Agency. Ambient Air Quality Surveillance, 40 C.F.R. pt 58 2007. Beelen, R., Hoek, G., Vienneau, D., Eeftens, M., Dimakopoulou, K., Pedeli, X., Tsai, M.Y., Künzli, N., Schikowski, T., Marcon, A., Eriksen, K., Raaschou-Nielsen, O., Stephanou, E., Patelarou, E., Lanki, T., Yli-Tuomi, T., Declercq, C., Falq, G., , M.J., Dedel _ _ A., Stempfelet, M., Birk, M., Cyrys, J., von Klot, S., N ador, G., Varro e, _ R., Mo € lter, A., Lindley, S., Madsen, C., Cesaroni, G., Ranzi, A., Gra˛ulevi ciene, €mer, U., Kuhlbusch, T., Badaloni, C., Hoffmann, B., Nonnemacher, M., Kra Cirach, M., de Nazelle, A., Nieuwenhuijsen, M., Bellander, T., Korek, M., €mgren, M., Dons, E., Jerrett, M., Fischer, P., Wang, M., Olsson, D., Stro Brunekreef, B., de Hoogh, K., 2013. Development of NO2 and NOx land use regression models for estimating air pollution exposure in 36 study areas in Europe e the ESCAPE project. Atmos. Environ. 72, 10e23. http://dx.doi.org/ 10.1016/j.atmosenv.2013.02.037. Brauer, M., Hoek, G., Smit, H.A., de Jongste, J.C., Gerritsen, J., Postma, D.S., Kerkhof, M., Brunekreef, B., 2007. Air pollution and development of asthma, allergy and infections in a birth cohort. Eur. Respir. J 29, 879e888. http:// dx.doi.org/10.1183/09031936.00083406. Off. J. Eur. Soc. Clin. Respir. Physiol. Briggs, D.J., de Hoogh, C., Gulliver, J., Wills, J., Elliott, P., Kingham, S., Smallbone, K., 2000. A regression-based method for mapping traffic-related air pollution: application and testing in four contrasting urban environments. Sci. Total Environ. 253, 151e167. Brunekreef, B., Holgate, S.T., 2002. Air pollution and health. Lancet 360, 1233e1242. http://dx.doi.org/10.1016/S0140-6736(02)11274-8. Clougherty, J.E., Kubzansky, L.D., 2009. A framework for examining social stress and susceptibility to air pollution in respiratory health. Environ. Health Perspect. 117, 1351e1358. http://dx.doi.org/10.1289/ehp.0900612. Clougherty, J.E., Levy, J.I., Kubzansky, L.D., Ryan, P.B., Suglia, S.F., Canner, M.J., Wright, R.J., 2007. Synergistic effects of traffic-related air pollution and exposure to violence on urban asthma etiology. Environ. Health Perspect. 115, 1140e1146. http://dx.doi.org/10.1289/ehp.9863. Clougherty, J.E., Wright, R.J., Baxter, L.K., Levy, J.I., 2008. Land use regression modeling of intra-urban residential variability in multiple traffic-related air pollutants. Environ. Health 7, 17. http://dx.doi.org/10.1186/1476-069X-7-17. Cohen, J., Cohen, P., West, S.G., Aiken, L.S., 2003. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. Lawrence Erlbaum Associates. Cyrys, J., Eeftens, M., Heinrich, J., Ampe, C., Armengaud, A., Beelen, R., Bellander, T., Beregszaszi, T., Birk, M., Cesaroni, G., Cirach, M., de Hoogh, K., De Nazelle, A., de _ _ A., Dimakopoulou, K., Eriksen, K., Galassi, C., Vocht, F., Declercq, C., Dedel e, _ R., Grivas, G., Gruzieva, O., Gustafsson, A.H., Hoffmann, B., Gra˛ ulevi ciene, €mer, U., Lanki, T., Lozano, P., Madsen, C., Iakovides, M., Ineichen, A., Kra €lter, A., Mosler, G., Nieuwenhuijsen, M., Meliefste, K., Modig, L., Mo Nonnemacher, M., Oldenwening, M., Peters, A., Pontet, S., Probst-Hensch, N., Quass, U., Raaschou-Nielsen, O., Ranzi, A., Sugiri, D., Stephanou, E.G., Taimisto, P., Villani, S., Wang, M., Brunekreef, B., Hoek, G., 2012. € vi, E., Tsai, M.-Y., Vasko Variation of NO2 and NOx concentrations between and within 36 European study areas: results from the ESCAPE study. Atmos. Environ. 62, 374e390. http://dx.doi.org/10.1016/j.atmosenv.2012.07.080. Davidson, K., Hallberg, A., McCubbin, D., Hubbell, B., 2007. Analysis of PM2.5 using the environmental benefits mapping and analysis program (BenMAP). J. Toxicol. Environ. Health. A 70, 332e346. http://dx.doi.org/10.1080/ 15287390600884982. De Vries, S., Verheij, R.A., Groenewegen, P.P., Spreeuwenberg, P., 2003. Natural environments e healthy environments? An exploratory analysis of the relationship between greenspace and health. Environ. Plan. A 35, 1717e1731. http:// dx.doi.org/10.1068/a35111. Dijkema, M.B., Gehring, U., van Strien, R.T., van der Zee, S.C., Fischer, P., Hoek, G., Brunekreef, B., 2011. A comparison of different approaches to estimate smallscale spatial variation in outdoor NO2concentrations. Environ. Health Perspect. 119, 670e675. http://dx.doi.org/10.1289/ehp.0901818. Dockery, D.W., Pope III, C Arden, Xu, X., Spengler, J.D., Ware, J.H., Fay, M.E., Ferris Jr., Benjamin G., Speizer, F., 1993. An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med. 329, 1753e1759. Donovan, G.H., Michael, Y.L., Butry, D.T., Sullivan, A.D., Chase, J.M., 2011. Urban trees and the risk of poor birth outcomes. Health Place 17, 390e393. http:// dx.doi.org/10.1016/j.healthplace.2010.11.004. Eeftens, M., Beelen, R., de Hoogh, K., Bellander, T., Cesaroni, G., Cirach, M., _ _ A., Dons, E., de Nazelle, A., Dimakopoulou, K., Eriksen, K., Declercq, C., Dedel e, _ R., Heinrich, J., Hoffmann, B., Falq, G., Fischer, P., Galassi, C., Gra˛ulevi ciene, €lter, A., Jerrett, M., Keidel, D., Korek, M., Lanki, T., Lindley, S., Madsen, C., Mo dor, G., Nieuwenhuijsen, M., Nonnemacher, M., Pedeli, X., RaaschouNa Nielsen, O., Patelarou, E., Quass, U., Ranzi, A., Schindler, C., Stempfelet, M., , M.J., Vienneau, D., Stephanou, E., Sugiri, D., Tsai, M.-Y., Yli-Tuomi, T., Varro Klot, S. Von, Wolf, K., Brunekreef, B., Hoek, G., 2012. Development of land use regression models for PM(2.5), PM(2.5) absorbance, PM(10) and PM(coarse) in 20 European study areas; results of the ESCAPE project. Environ. Sci. Technol. 46, 11195e11205. http://dx.doi.org/10.1021/es301948k. Fann, N., Fulcher, C.M., Hubbell, B.J., 2009. The influence of location, source, and emission type in estimates of the human health benefits of reducing a ton of air pollution. Air Qual. Atmos. Health 2, 169e176. http://dx.doi.org/10.1007/s11869009-0044-0. 103 Faus-Kessler, T., Kirchner, M., Jakobi, G., 2008. Modelling the decay of concentrations of nitrogenous compounds with distance from roads. Atmos. Environ. 42, 4589e4600. http://dx.doi.org/10.1016/j.atmosenv.2008.01.073. Fujii, E., Manager, P., Lawton, J., Barnes, D.E., Spada, N., Trails, B.C.S., 2008. Breathe California of Sacramento- Emigrant Trails Health Effects Task Force Removal Rates of Particulate Matter onto Vegetation as a Function of Particle Size. Gilbert, N.L., Goldberg, M.S., Beckerman, B., Brook, J.R., Jerrett, M., 2005. Assessing al, Canada, with a landspatial variability of ambient nitrogen dioxide in Montre use regression model. J. Air Waste Manag. Assoc. 55, 1059e1063. Gilbert, N.L., Goldberg, M.S., Brook, J.R., Jerrett, M., 2007. The influence of highway traffic on ambient nitrogen dioxide concentrations beyond the immediate vicinity of highways. Atmos. Environ. 41, 2670e2673. http://dx.doi.org/10.1016/ j.atmosenv.2006.12.007. Henderson, S.B., Beckerman, B., Jerrett, M., Brauer, M., 2007. Application of land use regression to estimate long-term concentrations of traffic-related nitrogen oxides and fine particulate matter. Environ. Sci. Technol. 41, 2422e2428. http:// dx.doi.org/10.1021/es0606780. Hirabayashi, S., Kroll, C.N., Nowak, D.J., 2012. I-tree Eco Dry Deposition Model Descriptions. Hoek, G., Beelen, R., de Hoogh, K., Vienneau, D., Gulliver, J., Fischer, P., Briggs, D., 2008. A review of land-use regression models to assess spatial variation of outdoor air pollution. Atmos. Environ. 42, 7561e7578. http://dx.doi.org/ 10.1016/j.atmosenv.2008.05.057. Hubbell, B.J., Hallberg, A., McCubbin, D.R., Post, E., 2004. Health-related benefits of attaining the 8-Hr ozone standard. Environ. Health Perspect. 113, 73e82. http:// dx.doi.org/10.1289/ehp.7186. Hubbell, B., Fann, N., Levy, J.I., 2009. Methodological considerations in developing local-scale health impact assessments: balancing national, regional, and local data. Air Qual. Atmos. Health 2, 99e110. http://dx.doi.org/10.1007/s11869-0090037-z. I-Tree, 2011. I-tree Eco User's Manual. http://dx.doi.org/10.1007/SpringerReference_ 28001. Jerrett, M., Arain, A., Kanaroglou, P., Beckerman, B., Potoglou, D., Sahsuvaroglu, T., Morrison, J., Giovis, C., 2005. A review and evaluation of intraurban air pollution exposure models. J. Expo. Anal. Environ. Epidemiol. 15, 185e204. http:// dx.doi.org/10.1038/sj.jea.7500388. Jerrett, M., Shankardass, K., Berhane, K., Gauderman, W.J., Künzli, N., Avol, E., Gilliland, F., Lurmann, F., Molitor, J.N., Molitor, J.T., Thomas, D.C., Peters, J., McConnell, R., 2008. Traffic-related air pollution and asthma onset in children: a prospective cohort study with individual exposure measurement. Environ. Health Perspect. 116, 1433e1438. http://dx.doi.org/10.1289/ehp.10968. Karner, A.A., Eisinger, D.S., Niemeier, D.A., 2010. Near-roadway air quality: synthesizing the findings from real-world data. Environ. Sci. Technol. 44, 5334e5344. http://dx.doi.org/10.1021/es100008x. Kashima, S., Yorifuji, T., Tsuda, T., 2009. Application of land use regression to regulatory air quality data in Japan. Sci. Total Environ. 407, 3055e3062. http:// dx.doi.org/10.1016/j.scitotenv.2008.12.038. Lamsal, L.N., Martin, R.V., Parrish, D.D., Krotkov, N.A, 2013. Scaling relationship for NO2 pollution and urban population size: a satellite perspective. Environ. Sci. Technol. 47, 7855e7861. http://dx.doi.org/10.1021/es400744g. Maas, J., Verheij, R.A., Groenewegen, P.P., de Vries, S., Spreeuwenberg, P., 2006. Green space, urbanity, and health: how strong is the relation? J. Epidemiol. Community Health 60, 587e592. http://dx.doi.org/10.1136/jech.2005.043125. Mavko, M.E., Tang, B., George, L.A., 2008. A sub-neighborhood scale land use regression model for predicting NO(2). Sci. Total Environ. 398, 68e75. http:// dx.doi.org/10.1016/j.scitotenv.2008.02.017. McConnell, R., Berhane, K., Yao, L., Jerrett, M., Lurmann, F., Gilliland, F., Künzli, N., Gauderman, J., Avol, E., Thomas, D., Peters, J., 2006. Traffic, susceptibility, and childhood asthma. Environ. Health Perspect. 114, 766e772. http://dx.doi.org/ 10.1289/ehp.8594. Min, K.-E., Pusede, S.E., Browne, E.C., LaFranchi, B.W., Wooldridge, P.J., Cohen, R.C., 2013. Eddy covariance fluxes and vertical concentration gradient measurements of NO and NO2 over a ponderosa pine ecosystem: observational evidence for within canopy removal of NOx. Atmos. Chem. Phys. Discuss. 13, 12437e12484. http://dx.doi.org/10.5194/acpd-13-12437-2013. Novotny, E.V., Bechle, M.J., Millet, D.B., Marshall, J.D., 2011. National satellite-based land-use regression: NO2 in the United States. Environ. Sci. Technol. 45, 4407e4414. http://dx.doi.org/10.1021/es103578x. Nowak, D.J., Crane, D.E., Stevens, J.C., 2006. Air pollution removal by urban trees and shrubs in the United States. Urban For. Urban Green 4, 115e123. http:// dx.doi.org/10.1016/j.ufug.2006.01.007. Nyberg, F., Gustavsson, P., Jarup, L., Bellander, T., Berglind, N., Jakobsson, R., Pershagen, G., 2000. Urban air pollution and lung cancer in Stockholm. Epidemiology 11, 487e495. Ogawa, Co, USA, I., 2006. NO, NO 2, NOx and SO2 Sampling Protocol Using the Ogawa Sampler *. Ostro, B., Lipsett, M., Mann, J., Braxton-Owens, H., White, M., 2001. Air pollution and exacerbation of asthma in African-American children in Los Angeles. Epidemiology 12, 200e208. Pataki, D.E., Carreiro, M.M., Cherrier, J., Grulke, N.E., Jennings, V., Pincetl, S., Pouyat, R.V., Whitlow, T.H., Zipperer, W.C., 2011. Coupling biogeochemical cycles in urban environments: ecosystem services, green solutions, and misconceptions. Front. Ecol. Environ. 9, 27e36. http://dx.doi.org/10.1890/090220. 104 M. Rao et al. / Environmental Pollution 194 (2014) 96e104 Pope, C.A., Burnett, R.T., Thun, M.J., Calle, E.E., Krewski, D., Thurston, G.D., 2002. Lung Cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. JAMA J. Am. Med. Assoc. 287, 1132e1141. Portland Parks & Recreation, 2007. Portland's Urban Forest Canopy: Assessment and Public Tree Evaluation (Portland). Ryan, P.H., LeMasters, G.K., 2007. LUR_review_Ryan2006.pdf. Inhal. Toxicol. 19, 127e133. Samet, J.M., Dominici, F., Curriero, F.C., Coursac, I., Zeger, S., 2000. Fine particulate air pollution and mortality in 20 U.S. Cities, 1987e1994. N. Engl. J. Med. 343, 1742e1749. Samoli, E., Analitis, A., Touloumi, G., Schwartz, J., Anderson, H.R., Sunyer, J., Bisanti, L., Zmirou, D., Vonk, J.M., Pekkanen, J., Goodman, P., Paldy, A., Schindler, C., Katsouyanni, K., 2005. Estimating the exposureeresponse relationships between particulate matter and mortality within the APHEA multicity project. Environ. Health Perspect. 113, 88e95. http://dx.doi.org/10.1289/ehp.7387. €l a €, H., Viippola, V., Rantalainen, A.-L., Pennanen, A., Yli-Pelkonen, V., 2012. Does Seta urban vegetation mitigate air pollution in northern conditions? Environ. Pollut. 183, 104e112. http://dx.doi.org/10.1016/j.envpol.2012.11.010. Sparks, J.P., 2009. Ecological ramifications of the direct foliar uptake of nitrogen. Oecologia 159, 1e13. http://dx.doi.org/10.1007/s00442-008-1188-6. Straf, K., Cohen, A., Sammet, J., 2013. Air Pollution and Cancer, IARC Scientific Publication 161, International Agency for Reseach on Cancer. World Health Orgnanization, Lyo, FR. Takahashi, M., Higaki, A., Nohno, M., Kamada, M., Okamura, Y., Matsui, K., Kitani, S., Morikawa, H., 2005. Differential assimilation of nitrogen dioxide by 70 taxa of roadside trees at an urban pollution level. Chemosphere 61, 633e639. http:// dx.doi.org/10.1016/j.chemosphere.2005.03.033. United Nations, 2011. World Urbanization Prospects the 2011 Revision. U.S. EPA (US Environmental Protection Agency), 2010. Environmental Benefits Mapping and Analysis Program (BenMAP). Yin, S., Shen, Z., Zhou, P., Zou, X., Che, S., Wang, W., 2011. Quantifying air pollution attenuation within urban parks: an experimental approach in Shanghai, China. Environ. Pollut. 159, 2155e2163. http://dx.doi.org/10.1016/ j.envpol.2011.03.009.