Evaluation and Application of an Online Coupled Modeling System to Assess the Interaction between Urban Vegetation and Air Quality

Vegetation has always been an integral part of the urban scene, affecting ambient air quality through both direct and indirect ways: by enhancing the dry deposition process of air pollutants and by contributing to the formation of ozone due to the emission of biogenic volatile organic compounds (BVOC). In this study, hourly measurements of gaseous dry deposition velocities are used to evaluate the performance of two dry deposition modules. Based on verification against measurements, an online coupled modeling system (RBLM-Chem) is introduced to investigate the dry deposition process of air pollutants (both gas and particles), the diurnal and seasonal variation patterns, and the discrepancy between different vegetation species as well as to assess the role of urban vegetation in affecting local air quality under different greening scenarios. Results indicate that trees are generally more efficient in removing air pollutants than shorter vegetation (e.g., grass). Moreover, conifers exhibit higher dry deposition velocities than broadleaf trees in terms of annual average. The introduction of vegetation (either trees or grass) clearly raises the dry deposition velocity of air pollutants. The air pollutant that is most removed by urban vegetation in Suzhou is PM10, with an annual removal rate of 1484.5 t a. The current urban greening scenario within Suzhou contributes to a reduction in daily mean concentration of 8.1% (SO2), 7.1% (NO2), 5.6% (O3), 4.7% (PM10) and 4.4% (PM2.5) in summer, while the reduction in winter is 4.6%, 5.5%, 4.5%, 3.6% and 3.7%, respectively. The improvement in pollutant concentration can be strengthened by increasing vegetation coverage. Additionally, the peri-urban forest ecosystem plays a role in air quality improvement within an urban area. As for the effect of BVOC emissions, the emission from urban trees under 40% coverage results in the consumption of NOx (–3.2%) and the formation of O3 (2.3%).


INTRODUCTION
Air pollution has become a worldwide challenge affecting human health, quality of urban life and the development of economy, especially in developing countries (UNEP, 2002;Akimoto, 2003;UNEP, 2005).China, as the largest developing country, has been suffering from poor air quality caused by rapid urbanization and industrialization in the recent decades (Matsui et al., 2009;Zhang et al., 2012;Wang et al., 2014).For example, particulate concentrations such as PM 2.5 in Beijing and Shanghai were almost ten and six times the World Health Organization guideline values of 10 µg m -3 (annual average) (WHO, 2005), respectively (He et al., 2001;Ye et al., 2003).Air pollution problems in mega cities will remain to be one of the most urgent concerns in the upcoming decades in China (Chan and Yao, 2008).These cities need to come up with effective ways to control the air pollution problems and reduce the damages.
The air quality can be significantly improved by regulating the source of air pollutants in China.However, such improvement only works in a short term, that is, the air quality will plummet as soon as the control measures come to an end (Huang et al., 2013).Besides, the accomplishment of such strategies is inevitably at the expense of costly instruments or cutting down production scale.
Vegetation has always been an integral part of urban circumstances and has beneficial meteorological and environmental impacts.In many regions of America, tree canopy coverage can reach up to 20-40% on a city-wide basis (Oke, 1989).Urban vegetation could modify local climate by humidifying the ambient atmosphere and alleviating the urban heat island effect (Yu and Hien, 2006;Ng et al., 2012;Yang et al., 2015).
Vegetation can also affect urban environment through both direct and indirect ways.The direct environmental effects of urban vegetation is due to its important role as an intensive deposition sink for both gaseous and particulate air pollution thus providing an efficient way of removing pollutants from the urban atmosphere (Freer-Smith et al., 1997;Yang et al., 2008;Baumgardner et al., 2012).Vegetation removes gaseous air pollutants primarily by uptake through leaf stomata, whereas some species are removed by the plant surface.Once within the leaf, gases spread into intercellular spaces and may be dissolved in liquids to form acids or interact with inner-leaf surface (Nowak et al., 2000).Vegetation also ameliorates air quality by removing airborne particles.Four mechanisms are involved in the particle dry deposition process: gravitational settling, Brownian motion, impaction and interception (Beckett et al., 2000).Different planting scenarios, including different vegetation coverage and different plant species (conifer, broadleaf or grass), may result in quite distinct role in air pollution reduction.For example, a measurement conducted in Israel indicated that PM 10 concentrations between two comparable cities were reduced by about 5-20% as the tree coverage increased by between 19% and 25% (Freiman et al., 2006).The capacity of particle uptake by conifers is greater than broadleaf for their more complex foliar structure and smaller collection radii (Zhang et al., 2001;Freer-Smith et al., 2005).Besides, dry deposition velocity over forests is generally larger than over short vegetation (Flechard et al., 2011).
On the other hand, urban vegetation, especially urban trees, could also serve as a significant source of biogenic volatile organic compounds (BVOCs), which may indirectly contribute to the formation of ozone (O 3 ) and other secondary air pollutants (Curci et al., 2009;Goldstein et al., 2009).This is the indirect environmental effect of urban vegetation.
Atmospheric deposition, including dry and wet processes, is an indispensable component of the pollutant budget in the atmosphere-biosphere system.Although the dry deposition process is slower than wet deposition, it operates continuously over all surfaces while wet deposition occurs episodically.Hence the accumulated removal by dry deposition may be much more pronounced than wet removal (Ginoux et al., 2001).Dry deposition is one of the most important sinks for pollutants.It is responsible for a remarkable portion of the total (dry and wet) nitrogen deposition (e.g., 10-50%, Zhang et al., 2009;58%, Sparks et al., 2008).With respect to sulfur species, dry and wet deposition accounts for 56 and 44% of the annual total deposition over mainland China, respectively (Wang et al., 2004).Moreover, for coarse particles, dry deposition dominates their total removal (Zhao et al., 2003).
In-site measurements of dry deposition are extremely challenging and expensive, hence the derivation of dry deposition velocity always resort to numerical simulations and then paired with measured pollutant concentrations to determine the total deposition fluxes.Modeling studies on dry deposition fluxes of air pollutants within urban areas are relatively scarce.Moreover, conventional dry deposition schemes, e.g., the scheme introduced by Wesely (1989), are too simple for application in urban circumstances, which is actually a complex mosaic of diverse surface types and vegetation coverage.Accurate representation of deposition velocities, is prerequisite in the simulation of atmospheric pollutant concentration (Michou et al., 2005).Dry deposition scheme developed by Zhang et al. (2001Zhang et al. ( , 2003) ) comprehensively considered the impact of u * , RH, LAI and canopy humidity on deposition velocity.Wu et al. (2011) suggested that the performance of Wesely (1989) dry deposition scheme can be substantially improved by utilizing the parameterization scheme developed by Zhang et al. (2003, 'new-scheme' hereafter).
As stated above, urban vegetation could affect ambient air quality through both direct and indirect ways and this effect may vary by seasons, plant species, vegetation coverage as well as characteristics of air pollutants.Hitherto, quantitative study to project the potential air quality changes within urban areas under different urban greening scenarios (e.g., different vegetation coverage, different planting species, different distribution patterns of vegetation in different seasons) are relatively scarce, especially in Chinese cities.In this study, an online coupled modeling system is built up, wherein a regional boundary-layer model (RBLM) is used to provide meteorological fields and an atmospheric chemistry model (ACTDM) provides air pollutant concentrations.With the adoption of this new modeling system, a series of numerical cases are designed to quantify the likely impacts of different urban greenery scenarios on local pollutant concentrations, considering the role of urban vegetation as both enhanced deposition sinks and BVOC emitters.This study provide a numerical modeling tool for investigating the complex interactions between urban vegetation and urban air quality, and such attempts would be a meaningful scientific issue for more accurate prediction of air pollutant concentrations and of major interest to urban planners.

DESCRIPTION OF RBLM-CHEM MODEL
The RBLM-Chem model is developed on the basis of the Regional Boundary Layer Model (RBLM), with meteorological input provided by the RBLM model to online drive the Atmospheric Chemical Transport and Dispersion Model (ACTDM) for modeling the pollutant concentration field.Moreover, in this study a new dry deposition parameterization scheme has been implemented in RBLM-Chem for more accurate consideration of the impact of urban vegetation on pollutant dry deposition process.Details will be illustrated in the following sections.

Meteorological Model
In order to model regional air quality issues over a complex underlying surface within urban areas, it is necessary to develop a regional transport and dispersion modeling system to combine a meteorological model with an air quality model.In this study, the primary source of meteorological fields for input to the air quality model is generated by the Regional Boundary Layer Model (RBLM).RBLM is a regional scale (kilometers to hundreds kilometers) meteorological numerical weather prediction model which was developed for application in urban meteorological modeling.Detailed descriptions about this model can be found in Chen et al. (2009) and Yang et al. (2015).This model has been widely adopted in several urban meteorological modeling studies (Xu et al., 2002;Chen et al., 2009;Yang et al., 2015).

ACTDM Model
ACTDM (atmospheric chemical transport and dispersion model) is a regional atmospheric pollutant concentration prediction model involving online calculation of source emissions, transport, gas phase chemistry, aerosol modules, radiation and photolysis rates, and dry and wet deposition processes.A detailed description of ACTDM can be referred to Liu (2002).The ACTDM model could provide timedependent three-dimensional distribution of air pollutant concentration, including gaseous species such as SO 2 , NO x , O 3 , CO, PAN, and particulate matter such as PM 10 , PM 2.5 , sulfate, nitrate, ammonium, black carbon, organic carbon.Additionally, this model could also be applied to the forecast of atmospheric extinction coefficient and visibility derived from the modeled results of aerosol concentration and ambient relative humidity.ACTDM has been successfully applied in the operational forecast and modeling studies on multiple cities, and proved to exhibiting good performance (Jiang et al., 2001;Liu et al., 2009;Qian et al., 2013).
Currently, the dry deposition module incorporated in ACTDM is based on Wesely (1989) parameterization ('oldscheme' hereafter).In this study, the ACTDM model is further developed by integrating the more realistic and accurate dry deposition algorithm (i.e., the new-scheme) instead of the old-scheme.The new-scheme contains two sub-modules: one gaseous dry deposition module and one aerosol dry deposition module.Details of this updated dry deposition scheme will be described in the following section.

Gaseous Dry Deposition
Velocity of gaseous dry deposition is determined on the basis of Zhang et al. (2003).In this scheme, dry deposition velocity (V d ) can be defined as the reciprocal of an ensemble of resistance: where R a is the aerodynamic resistance, R b is the quasilaminar sub-layer resistance, and R c is the bulk surface resistance.Expressions for R a and R b can be found in many earlier studies (e.g., McRae, 1981;Erisman et al., 1994).
Besides, uncertainties are relatively small in R a and R b .R c is the most complex and important portion of the total resistance and discrepancy between different models mainly comes from the computation of R c .Detailed parameterization of R c can be refered to Zhang et al. (2003).

Particulate Dry Deposition
The parameterization of particle dry deposition velocity is considered as a function of particle density and size as well as ambient meteorological variables.This scheme is considered to be relatively comprehensive.It has included deposition mechanisms such as gravitational settling, Brownian diffusion, impaction, turbulent transfer, particle rebound as well as hygroscopic growth under wet conditions.Detailed descriptions of the parameterization of particle dry deposition velocity can be found in Zhang et al. (2001).

Study Area
Given the purpose of this study, Suzhou City is used as a case study.It is located in the middle of the Yangtze Delta River (YRD) which is one of the most developed and booming areas in China (see Fig. 1).The population of Suzhou is about 2.4 million.Suzhou endures a typical subtropical marine climate under the influence of the East Asia monsoon and the four seasons are quite distinct.Suzhou has a relatively high-rise morphology with an average building height of about 19 m.Besides, it is also an environmentally friendly city for its relative high level of urban greening.The average per capita green space area of Suzhou is 14.9 m 2 (Statistics Bureau of Suzhou City, 2011).The most common tree species applied in the study area belong to deciduous broadleaf species which account for about 88.2% of the total and the rest belong to coniferous species (Qian, 2005).Suzhou is representative of East Asia metropolises for its monsoon climate, rapid economic and urbanization progress as well as endeavor to introduce more vegetation within urban areas.

Satellite Data Acquisition and Analysis
This study employs Landsat5 satellite observation to determine the spatial distribution properties of land use types and vegetation coverage within the model domain.Landsat5 satellite data is characterized by its high spatial resolution (25 m × 25 m).The satellite data used in this work is obtained on 09/21/2010 with path 119, row 38 covering Suzhou City.The study area is cut out from the satellite image, as shown in Fig. 1(b).
The model domain includes 95 × 95 grid points, with resolutions of 1 km × 1 km.Four cities are involved in the study area, which is Suzhou, Wuxi, Changshu and Kunshan, respectively.Suzhou is located in the middle of the domain and centered at 31.26°N, 120.63°E (see Fig. 1(b)).The area of Suzhou is about 576 km 2 .The land use types within the domain include urban, tree, grass, water and cropland.Spatial distribution pattern of vegetation coverage (VC) could be derived from the Landsat5 satellite observations.As the resolution of the model grid is 1 km × 1 km, each grid contains 40 × 40 (1 km/25 m = 40) subgrids.Hence, the vegetation coverage is defined as the ratio of tree and grass subgrids within each grid.The mean VC of Suzhou is 21.1% involving tree coverage of 18.2% and grass coverage of 2.9%.Most vegetation is concentrated in the urban fringe area (VC ≈ 30%-40%), while in the densely populated central area VC is lower (< 10%).

Observation Data Used in this Study
Hourly measurements of NO y (NO, NO 2 , HNO 3 and  (2005).Flux data were omitted when turbulence intensity is too low (i.e., when friction velocity < 0.2 m s -1 ), which lead to approximate 18% and 21% of the data being omitted for O 3 and NO y (Wu et al., 2011).Given the hourly measured pollutant concentrations and deposition fluxes, dry deposition velocity (V d ) can be computed as where C(z) is the pollutant concentration at reference height z and -F denotes downward deposition flux (upward flux is positive).
Hourly meteorology data for the study area in 2014 is obtained from six weather stations (Suzhou, Changshu, Kunshan, Dongshan, Wujiang and Taicang, as shown in Fig. 1(c)).Hourly air pollution data including SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 concentrations in the study area for the same year is obtained from Suzhou, Changshu, Kunshan, Wujiang and Taicang sites.These data are used for validation of the model performance of meteorological factors and pollutant concentrations.Additionally, hourly meteorology and air pollution data obtained from Suzhou station are adopted as input for running the new-scheme so as to investigate the seasonal and diurnal variation features of air pollutants dry deposition processes within Suzhou areas.

Observations of Gaseous Dry Deposition and Evaluation of its Modeled Results
Figs. 2 and 3 compares the hourly modeled V d (O 3 ) and V d (NO y ) by old-scheme and new-scheme against observations.The observed and modeled V d values are all conducted at the HFEMS site as described above.In addition, Table 1 presents the statistical results of the comparisons.Following Charusombat et al. (2010), model results of the   ) and V d (NO y ) are evaluated here using descriptive statistics of the fractional bias (FB) and degree of agreement (DoA).
In general, the modeling performance of the new-scheme exhibits obvious improvement compared to the old-scheme for higher DoAs and smaller FBs.Additionally, Fig. 4 shows the annual mean diurnal variation of observed and modeled V d (O 3 ) by the two dry deposition schemes.It can be seen that the new-scheme generally well captured the diurnal pattern of V d (O 3 ) throughout the day except for a slightly overestimation in some nighttime hours, whereas the oldscheme considerably underestimates the rising of V d (O 3 ) in the morning as well as the peak value (~0.3 cm s -1 ) at noon.
For particulate pollutants, measurements are scarce on its deposition velocity.Hence, in this study, documented data from relevant literature are employed to test the validity of model predictions.As shown in Figs.5(a)-5(c), modeled size dependence of particle V d is investigated on grass, broadleaf and coniferous trees, respectively, and compared with documented measurement data.The solid line stands for the modeling V d under intermediate wind conditions (i.e., u * = 44 cm s -1 ), while the shadow represents modeling V d ranges from low wind conditions (i.e., u * = 11 cm s -1 ) to strong wind conditions (i.e., u * = 117 cm s -1 ).Scatters in the figure stand for the documented data from relevant literatures.Generally speaking, a good agreement can be found between measurements and modeling results, especially for larger particles (e.g., D p > 1 µm), while modeling results may overestimate the V d for submicron particles to some extent.Particle V d is strongly size-dependent as a V-shape trend as indicated by the modeling result.Different mechanism dominates the deposition process of particles in different size ranges.For fine particles (D p < 0.1 µm), Brown diffusion is the main deposition mechanism and its importance decreases as particle size increases.Impaction and interception are important for particles of size 2-10 µm.Gravitational settling is only effective for larger particles (D p > 10 µm).Whereas for particles of size 0.1-2 µm, none of the above four mechanisms is evident, hence particles within this size range possess the lowest V d values (see Figs. 5(a)-5(c)).

Impacts of Urban Vegetation on Pollutants Dry Deposition Velocities and Fluxes
As stated above, the modeling performance of atmospheric dry deposition has been evaluated using hourly measured data at HFEMS.In this section, we will investigate the potential impacts of urban vegetation on air pollutants dry deposition velocities and fluxes through the combination of hourly measured meteorology data, pollutant concentrations and modeled V d values in Suzhou.Here, hourly meteorology data (including air temperature, wind speed, relative humidity and solar radiation) of Suzhou in 2014 are obtained from Suzhou meteorological bureau.Besides, hourly concentrations of SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 for the same year are from the environmental monitoring site in Suzhou.
Urban vegetation could directly improve local air quality  throughout the year except for summer.Summer correspond to the full-leaf period for broadleaf forest, hence V d values over broadleaf trees are likely to exceed that over conifers.Moreover, dry deposition process of particles (e.g., PM 10 and PM 2.5 ) also exhibit clear diurnal variation patterns (figure not shown) similar as gaseous species.Fig. 7 shows the average monthly variation of V d (O 3 ) and V d (PM 10 ) over three categories of vegetation surfaces.Gaseous dry deposition over vegetation surfaces exhibits evident seasonal variation tendency.It generally peaked in summer and leveled off thereafter, coinciding with the evolution of air temperature, solar radiation and LAI.With respect to comparison between different vegetation species, V d (O 3 ) over conifers is larger than that over broadleaf trees and grass during most of the year except for summer.In summer, O 3 uptake is most strengthened over broadleaf trees compared to conifers and grass, whereas in terms of annual mean, V d (O 3 ) is the largest over conifers, followed by broadleaf trees, and V d (O 3 ) over grass is the generally the minimum among the three vegetation species considered.
Monthly average V d (PM 10 ) exhibits similar seasonal variation pattern as gaseous species (see Fig. 7(b)).Nonetheless, V d (PM 10 ) over conifers is always larger than that over broadleaf trees, even in summer, followed by broadleaf trees and grass.This is because: (1) LAI for conifers remain at a relatively high level all year round, whereas broadleaf species have leafless periods in late autumn and winter.(2) Conifers, with finer and more complex foliage structures, have larger Stokes numbers, which may subsequently result in larger efficiency in impaction collection.Thus, conifers are shown to be more efficiency in capture particles (i.e., higher V d ) than broadleaf trees.Similarly, V d (PM 2.5 ) is largest over conifers throughout the year, compared to broadleaf trees and grass, but with smaller magnitude than PM 10 particles (figure is not shown).This is because the deposition process of larger particles (> 2 µm, in diameter) could be accelerated under the joint contribution of gravitational settling, impaction and interception.Whereas for finer particles, especially those within size range 0.1-2 µm, all four deposition mechanisms are less effective than PM 10 particles, hence result in lower V d values.

Comparison of Air Pollution Removal by Different Urban Vegetation Species
According to Nowak (1994), the removal amount (Q) of a particular air pollutant in a certain time period (T) can be calculated as: where F is the dry deposition flux of the pollutant and L is the total area of the region considered.Dry deposition flux (F) of air pollutants is determined not only by its dry deposition velocity, but also by its mass concentration in the air.The leading species of air pollutant in Suzhou is PM 10 , followed by PM 2.5 , O 3 , NO 2 and SO 2 , respectively.The concentration of PM 2.5 , NO 2 and SO 2 exhibits quite similar seasonal patterns with maximum values occurred in winter and leveled off thereafter (figure is not shown).On the contrary, the concentration of O 3 peaked in summer and declined dramatically in winter.This is mainly because the production of O 3 is strongly dependent on ambient temperature and solar radiation to maintain relevant photochemical reactions active.
Figs. 8(a)-8(e) shows the monthly dry deposition fluxes of major air pollutants (O 3 , NO 2 , SO 2 , PM 10 and PM 2.5 ) over different vegetation surfaces, respectively.Additionally, annual dry deposition fluxes over different land use categories (LUCs) are listed in Table 2.The results indicate that annual dry deposition fluxes is greatest for PM 10 particles, while for gaseous species only, relative removal is greater for O 3 than NO 2 and SO 2 which is closely associated with its concentration in the atmosphere.PM 10 particles is always the dominant pollutant in Suzhou in terms of monthly average concentration, hence result in the largest dry deposition fluxes.The concentration of PM 10 peaked in winter and late spring, whereas V d (PM 10 ) peaked in summer and troughed in winter, hence no clear seasonal variation tendency occurred in the monthly dry deposition fluxes of PM 10 (Fig. 8(d)).Dry deposition flux of PM 2.5 exhibits similar variation tendency as PM 10 with a smaller magnitude due to lower concentration and V d .
For gaseous pollutant, the largest dry deposition flux of O 3 occurred in the early summer and lowest in winter as a result of its similar seasonal cycle for both dry deposition velocity and concentration.No clear seasonal cycle exists in the dry deposition flux of NO 2 and SO 2 (see Figs. 8(b)-8(c)) since its concentration maximizes in winter and minimizes in summer whereas its dry deposition velocity just the opposite (i.e., maximizes in summer and minimizes in winter).
In terms of comparison between different LUCs, conifers show the highest efficiency in removing major air pollutants, except for O 3 , due to longer time of foliage retention.For O 3 , its annual dry deposition flux is the largest over broadleaf trees under the joint contribution of maximum V d as well as concentration in summer.Grass surface is less efficient in removing air pollutants compared to tree canopies.Moreover, it should be noted that all vegetated LUCs (i.e., grass, broadleaf and coniferous canopies) exhibite evidently larger dry deposition fluxes than urban concrete surface (see Table 2).That is, the presence of vegetation within urban areas could substantially improve air quality by enhancing the uptake of air pollutants.For example, annual dry deposition flux of O 3 is largest over broadleaf trees (7.02 g m -2 a -1 ), folowed by conifers (6.53 g m -2 a -1 ) and grass (4.84 g m -2 a -1 ), which means an increment of approximately 3.6, 3.2 and 2.1 fold compared to urban concrete surface, respectively.

Annual Air Pollutant Removal Attributed to Different Species of Urban Vegetation in Suzhou
Based on Landsat 5 images, the relative area of different vegetation species and urban concrete surfaces are listed in Table 3. From which we can see that, Suzhou has 121.6 km 2 covered by vegetation surfaces which amounts to 21.1% of the study area, of which 92.6 km 2 , 12.4 km 2 and 16.6 km 2 is covered by broadleaf trees, conifers and grass, respectively.Annual removal of major pollutant species over different LUCs is also shown in Table 4.The air pollutant that is most reduced is PM 10 and the reduction amounts to 1484.5 t a -1 due to urban vegetation in Suzhou.Besides, annual removal of SO 2 , NO 2 , O 3 and PM 2.5 can be up to 257.0, 386.4,811.4 and 281.7 t a -1 , respectively.Namely, total annual removal by urban vegetation contribute approximately 48.5%, 50.0%, 53.7%, 29.1% and 25.3% to the annual total dry deposition flux of SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 within Suzhou City, respectively.
By comparing the air pollution removal rates in Suzhou with that in other mega cities in China (see Table 4), urban vegetation generally exhibits a higher efficiency in removing PM 10 which is also the dominant pollutant species in most parts of China.It should be noted that the V d values adopted in the study of Beijing and Guangzhou are all typical values documented in relevant literatures, e.g., Lovett (1994) and Nowak et al. (1998).In this study, hourly V d values are produced according to observed meteorological data which is more closed to the realities.It should be noted that particles captured on the plant surface could be resuspended, especially under long-duration dry conditions and high wind speed.For instance, Nowak et al. (2013) estimated that approximately 23% of fine particles deposited on vegetation surface could be resuspended when wind speed reached 13 m s -1 , and this ratio could be up to 76% based on the modeling study by Schaubroeck et al. (2014).The representation of the resuspension mechanism is oversimplified in the dry deposition scheme employed in the present study, with no dependence upon synoptic conditions like air humidity and wind speed, hence lead to certain uncertainties in the modeling of the "net" removal of particles by urban vegetation.Future studies should focus on better representation of the resuspension mechanism and the complex interactions between dry deposition and resuspension.

Modeled Changes in Major Pollutants Concentrations for Different Urban Greening Scenarios
Urban vegetation could affect the concentrations of air pollutants in two different ways: firstly, by enhancing dry deposition hence removing pollutants from the atmosphere;  and S2, respectively.The observation data of major air pollutants used here are obtained from five environmental monitoring sites (Suzhou, Wujiang, Kunshan, Taicang and Changshu, respectivily) located in the modeling domain.In addition, relevant statistical measurements are also applied in this study for the evaluation of the chemical model, including normalized mean bias (NMB), index of agreement (IoA), factor of two (FAC) and correlation coefficient (R).
The statistical results are given in Table 5.In general, the model captures relatively well the diurnal variations of the pollutant concentrations.The statistical results show that the NMBs between observation and simulation data are within the range of 8%-31%.The IoAs of the modeled results are generally over 0.6.The percentage of modeled values lying between 0.5 and 2 folds to observation values (i.e., FAC) is within the range of 0.65-0.9.The correlation coefficients (R) between observational data and model predictions are mostly > 0.46.As a result, Table 5 demonstrates that the model performance on the simulation of air pollutants concentration is reliable.
On the basis of the validated model, series of scenarios are conducted to assess the impact that changes to the distribution of urban vegetation would have on air quality, in both summer and winter seasons.The setups for all the cases are summarized in Table 6.S_base (W_base) is designed as a base case with the actual distribution pattern of urban vegetation in Suzhou in summer (winter).S0 (W0) is an extreme case with completely no vegetation in Suzhou city under summer (winter) conditions.Hence, the impact of current greening level on moderating local air quality in summer (winter) could be investigated by comparing the modeling results of S0 (W0) against S_base (W_base).Moreover, trees within Suzhou are all turned into coniferous species, i.e., S1 (W1), to assess the sensitivity of the modeled pollutant concentration to changes in tree species in summer (winter).S0, S4 and S5 are differentiated by their preassigned percentage of tree coverage within the urban area, i.e., 0, 20% and 40%, respectively.Comparisons between these cases would reveal the different impactions on air quality due to the alternation of tree coverage.Effects of periurban ecosystems on the air quality of Suzhou are modeled by altering the LUCs in the rural area (see Fig. 1) from cropland to tree plantings.S2 (W2) and S3 (W3) represent changing the rural cropland into broadleaf and coniferous trees in summer (winter), respectively.S6 is running with consideration of BVOC emissions only under 40% tree coverage, i.e., not consider the impaction of trees on dry deposition processes.All the cases are presented in Table 6.
Modeled average diurnal variation of pollutant concentrations are shown in Fig. 9.Diurnal cycles of SO 2 , NO 2 , PM 10 and PM 2.5 exhibit evident bi-modal shape corresponding to the morning and evening rush hours.Whereas for O 3 , its concentration peaked at noon and leveled off thereafter as products for photochemical reactions.Concentrations of major air pollutants are generally higher in winter (Fig. 9(b)), except for O 3 , than those in summer (Fig. 9(a)) because of the energy-consumption activities for heating as well as shallower boundary layer in winter.The concentration of O 3 is much higher in summer as a result of higher air temperature and more active photochemical reactions.the VOC/NO x ratios, while the high concentration of NO x within urban areas always limited O 3 production potentials.Aerosol particles (PM 10 and PM 2.5 ) mainly gathered within urban areas, with peak concentrations reach 110 (PM 10 ) and 75 (PM 2.5 ) µg m -3 .
As discussed in the previous sections, the presence of vegetation within the urban circumstances could substantially enhance the uptake of gaseous and particulate pollutants (i.e., the dry deposition process), thus will lead to a reduction of air pollutant concentrations.Figs.10(f)-10(j) illustrated the role of urban vegetation in alleviating air pollutant concentrations, i.e., the difference between the modeled concentration values in the base case and the no-vegetation case (S_base-S0).Compared to the no-vegetation scenario (S0), urban vegetation could obviously reduce the urban air pollution concentrations in summer, and due to the impact of advection and transportation, air pollutants in the suburban areas could also be alleviated to a certain extent.In the following discussions, we mainly use the relative differences of the modeled results as compared to the relevant base cases (S_base, W_base, S0 or W0).Data shown in Table 7 are relative percentage changes of daily mean concentrations compared to reference cases.As shown in the table, concentrations of major air pollutants (SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 ) in the summer no-vegetation cases (S0) all exhibit different degrees of increment compared to the summer base case (S_base).That is, current urban greening scenario already has evident effect on air quality amelioration in summer.Due to the enhancement effect on dry deposition process of air pollutants, urban vegetation could result in reductions in the concentrations of SO 2 (8.1%),NO 2 (7.1%), O 3 (5.6%),PM 10 (4.7%), and PM 2.5 (4.4%), respectively, in summer.
In winter (see Table 7, figure not shown), when the urban vegetation factors are taken into consideration (W_base), there will also be certain reductions in the concentration of SO 2 (4.6%),NO 2 (5.5%), O 3 (4.5%),PM 10 (3.6%), and PM 2.5 (3.7%), respectively, as compared to the no-vegetation scenario (W0).By comparison, the role of urban vegetation in improving ambient air quality is more evident in summer than in winter, because winter is corresponding to the leafless period for deciduous species, hence the cleansing effect of air pollutant would be limited.
Expanding tree coverage within urban area would directly enhance the removal of air pollutants.Nonetheless, from the urban-planning perspective, implementation of vegetation within high-density urban circumstance to 50% coverage or even higher is either difficult or impractical (Ng et al., 2012).To address this issue, some studies discussed the potential impact of peri-urban ecosystems (e.g., national park or forest) on urban air quality (Baumgardner et al., 2012).In this study, four sets of numerical cases (S2, S3, W2, W3) were conducted to explore the cleansing effect of peri-urban forests in summer and winter, and to compare the difference between broadleaf and coniferous species.In summer, the introduction of peri-urban broadleaf forest (S2) would also contribute to the amelioration of urban air quality, the concentration of SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 would be reduced by 10.1%, 8.4%, 12.6%, 3.3% and 2.8%, respectively, compared to the control run (S_base).The alternation of coniferous species (S3) would also lead to the improvement of urban air quality, and the difference between these two rural-greening scenarios (S2 and S3) is quite small.However, As discussed above, urban vegetation could directly affect air quality by enhancing the dry deposition process of air pollutants.Besides, BVOC emissions from vegetation, especially trees, could also act as precursors of secondary air pollutants, hence may contribute to the formation of O 3 via photochemical reactions between BVOC and NO x .In this study, the BVOC-emission behavior is modeled according to Guenther et al. (1997).Based on the comparison result between sensitivity runs (see Table 7), BVOC emissions from the 40% tree coverage scenario (S6) may result in a consumption of NO x (-3.2%), meanwhile, a formation of O 3 (2.3%),compared to no-vegetation scenario (S0).

CONCLUSIONS
In this study, we mainly focused on investigating the dry deposition process of air pollutants (both gas and particles), the diurnal and seasonal variation patterns, the discrepancy between different vegetation species, and the role of urban vegetation in affecting local air quality under different greening scenarios.First, we evaluated the abilities of two sets of parameterization schemes to model gaseous and particulate dry deposition velocities (V d ) against observations.Hourly meteorological observation data was employed as input for offline driving the dry deposition scheme to investigate the diurnal and seasonal variation patterns of both gaseous and particulate air pollutants and to compare the efficiency of different vegetation species in removing air pollutants, i.e., analyzing dry deposition fluxes over broadleaves, conifers, and grass, respectively.Finally, an online coupled modeling system was introduced, with the RBLM model providing meteorological fields, the ACTDM model providing chemical concentration fields, and the updated dry deposition scheme accounting for the deposition process of air pollutants.With the adoption of this new system, the potential effects of various urban greening scenarios on air quality have been comprehensively investigated, considering their roles as both enhanced deposition sinks and BVOC emitters.
The performance of the Wesely (1989) scheme can be improved by adopting the more comprehensive and realistic Zhang et al. (2001Zhang et al. ( , 2003) ) scheme, which takes u * , RH, wind speed, LAI and canopy wetness into consideration.Our modeling results indicate a pattern in the time series of the deposition process that peaks in the daytime, levels off thereafter, and is higher in summer than winter.Trees are more efficient in removing air pollutants than shorter vegetation (e.g., grass) for larger leaf areas as well as more beneficial toward creating turbulent air movement due to their morphological structures.Moreover, conifers generally exhibit higher dry deposition velocities than broadleaf trees, in terms of annual average, because of their longer foliage retention and greater efficiency in capturing airborne particles.With respect to urban planning, the introduction of vegetation (either trees or grass) could clearly raise the dry deposition velocity of air pollutants.For example, annual dry deposition fluxes of SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 would be increased by 3.5-, 3.1-, 3.2-, 1.3-and 0.64fold, respectively, over conifers compared to urban concrete surfaces.For current distribution patterns of urban vegetation within Suzhou, the most removed air pollutant is PM 10 , which possesses an annual removal rate of 1484.5 t a -1 .Furthermore, SO 2 , NO 2 , O 3 , and PM 2.5 are removed at the annual rate of 257.0, 386.4,811.4 and 281.7 t a -1 , respectively.Urban vegetation contributes approximately 48.5%, 50.0%, 53.7%, 29.1% and 25.3% to annual total dry deposition fluxes of SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 , respectively, in this area.
Urban vegetation could affect air quality through two different ways: 1) by enhancing dry deposition so as to directly remove air pollutants from the atmosphere, and 2) by emitting biogenic volatile organic compounds (BVOC) that could act as precursors of secondary air pollutants, hence indirectly contributing to the formation of O 3 and other photochemical species.Our modeling results indicate that the average reduction in the daily mean concentration of SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 is 8.1%, 7.1%, 5.6%, 4.7% and 4.4%, respectively, in summer, and 4.6%, 5.5%, 4.5%, 3.6% and 3.7%, respectively, in winter, due to the current urban greening scenario in Suzhou.The improvement in urban air quality is more pronounced in summer than in winter for larger leaf areas and more favorable meteorological conditions.Whereas changing urban tree species from broadleaves to conifers, given identical coverage, would make little difference to pollutant concentration in summer, a conifer-greening scenario could further reduce the concentration of SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 by 1.9%, 2.0%, 2.3%, 2.6% and 1.7%, respectively, in winter.The improvement in pollutant concentration strengthens with increasing vegetation coverage.By raising tree coverage from 0 to 40%, the concentration of SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 could decline by 9.7%, 11.6%, 14.0%, 5.5% and 4.0%, respectively.Increasing vegetation coverage within urban areas, especially high-density megacities, is often costly and unfeasible, yet the introduction of peri-urban forest ecosystems would also contribute to higher air quality.Our modeling data found that broadleaf and coniferous tree species in peri-urban forests meliorated the urban air quality at similar levels in summer, but conifers were better at cleansing the urban atmosphere in winter, resulting in a reduction of 7.7% (SO 2 ), 12.2% (NO 2 ), 17.2% (O 3 ), 8.3% (PM 10 ) and 7.9% (PM 2.5 ) in concentration, respectively.Urban vegetation can also affect urban air quality through the emission of BVOC, which could act as precursors of secondary air pollutants.Based on this study, if the tree coverage reaches 40% within an urban area, the BVOC emissions could result in the consumption of NO x (-3.2%) and the formation of O 3 (2.3%).
The interaction between urban vegetation and air quality has been quantified comprehensively with the adoption of an online coupled modeling system, yet some limitations still exist.Uncertainties that arose from the oversimplified representation of the complex interaction between dry deposition and resuspension (as stated above) need to be fully explored in future studies.In addition, only two 10day periods were selected as summer and winter typical cases in the modeling analysis of different greening scenarios.Further studies should simulate a longer time scale, e.g., one year, to fully explore the effect of urban vegetation on air quality under various synoptic conditions throughout the year.

Fig. 1 .
Fig. 1.(a) Location of the study area in eastern China (as shown by the red box); (b) Latsat5 satellite image of the study area.Four cities are located in the study area as marked in the figure involving Suzhou (S.Z.), Wuxi (W.X.), Changshu (C.S.) and Kunshan (K.S.).The green box indicate the scope of sensitivity test area considered in the following section; (c) Land use types in the model domain.The land use data are obtained from Landsat5 observations (brown: urban; blue: tree; green: grass; white: water; yellow: farmland).The dots mark the location of weather stations: 1-Suzhou; 2-Changshu; 3-Kunshan; 4-Dongshan; 5-Wujiang; 6-Taicang.

Fig. 2 .
Fig. 2. Time series of observed and modeled O 3 dry deposition velocity by old-scheme and new-scheme (the shaded area indicate rain occurred that day).

Fig. 4 .
Fig. 4. Comparison of annual mean diurnal variation of observed and modeled V d (O 3 ) by different schemes in 2000.

Fig. 5 .Fig. 6 .
Fig. 5. Modeled size-dependence of particle V d over (a) grass (b) broadleaf and (c) coniferous trees for particle density taken as 1500 kg m -3 depositing under the same conditions (u * between 11 and 117 cm s -1 ), compared with documented measurement data.

Fig. 7 .
Fig. 7. Modeled monthly average dry deposition velocity of (a) O 3 and (b) PM 10 over different underlying surfaces (the dashed line corresponding to annual mean V d over different surfaces).

Fig. 9 .
Fig. 9. Modeled average diurnal variation of the concentrations of SO 2 , NO 2 , O 3 , PM 10 and PM 2.5 over 10 days for the base case (a) S_base (b) W_base.

Table 1 .
Statistical results of the modeled and observed V d (O 3 ) and V d (NO y ).

Table 2 .
Annual dry deposition fluxes of air pollutants over different LUCs.

Table 3 .
Annual air pollutant removal attributed to different species of urban vegetation (t a -1 ) and its relative contribution in Suzhou in 2014.

Table 4 .
Comparison of the annual air pollutant removal rate in Suzhou with other mega cities in China (g (m 2 a) -1 ).

Table 5 .
Statistical comparison of the modeled pollutant concentration with the observation data.

Table 6 .
Configuration of all the study cases.

Table 7 .
Modeled Changes in pollutant concentrations for different greening scenarios.