Role of Plant Leaves in Removing Airborne Dust and Associated Metals on Beijing Roadsides

As the capital of China, Beijing is continuously exposed to high amount of airborne dust, thus it is necessary to find improvement methods. Taking advantage of phytoremediation, an ecological and friendly way to improve air quality, this study investigates the role of urban plant leaves in removing airborne dust and its associated metals by analyzing leaf samples of 32 plant species in autumn. Results showed that leaves could remove dust from 0.510 to 23.0 g m with an overall mean of 7.50 g m on Beijing roadside sites. Some species removed certain metals more efficiently than others. Leaves of Chaenomeles speciosa accumulated the highest Cd (9.48 μg g) and the highest Cr value (19.8 μg g) was observed for leaves of Sorbaria kirilowii. Both of the highest concentrations of Cu (34.1 μg g) and Fe (868 μg g) appeared for leaves of Sophora japonica, whilst the highest values of Mn (169 μg g) and Ni (18.7 μg g) were found for leaves of Rosa chinensis and Prunus cerasifera f. atropurpurea, respectively. Populus beijingensis accumulated the most Pb (6.57 μg g) and Populus tomentosa the most Zn (142 μg g). For multi-metal pollution, Metal Accumulation Index (MAI) values were calculated, and the highest values were observed in unwashed leaves of Amygdalus persica (387), washed leaves of Punica granatum (105) and leaf dust of Viburnum sargentii (6.46). Plant species with dust accumulation rate above the mean including Koelreuteria paniculata, Ulmus pumila, Syringa oblata, Malus micromalu, Weigela florida cv. Red Prince, Ailanthus altissima, Salix babylonica, Robinia pseudoacacia, Ligustrum × vicaryi, Euonymus japonicus, Prunus cerasifera f. atropurpurea, Magnolia denudata, and species with higher MAI values including Amygdalus persica, Magnolia denudata, Syringa oblata are suggested to be considered in future green belt planning in Beijing.


INTRODUCTION
Most developing nations have been influenced by atmospheric pollution, in terms of human health (Anderson et al., 2012), climate change (Kan et al., 2012) and loss of biodiversity (Lovett et al., 2009).With urban transport development, traffic-derived pollutants become an increasing problem (Walsh and Shah, 1997;Ning and Sioutas, 2010;Cao et al., 2015) and have been linked to respiratory and cardiovascular disease, birth and developmental defects, cancer (Hei, 2010) and so on.According to WHO (2003), the life expectancy of urban residents in strongly polluted areas could decrease by over one year, particularly for children and people with lung and heart disease.There are 800 000 deaths annually, which could be due to urban air 2006; Hofman et al., 2014).Escobedo et al. (2008) found that planting vegetation in urban areas of Santiago City, Chile was a cost-effective measure to capture air pollutants.On an annual basis, urban trees/shrubs are estimated to remove about 215,000 tons of PM 10 for the whole US (Nowak et al., 2006) and about 234 tons of PM 10 in Chicago alone (Nowak, 1994).In China, vegetation covering an area of 1639 ha at eight residential areas in Beijing could remove 2170 tons of dust (Zhang et al., 1997), and trees in the city center could remove 772 tons of PM 10 over a year (Yang et al., 2005).Studies from other Chinese cities showed that concentrations of PM decreased by 9.1% near a forest in Shanghai (Yin et al., 2011) and the amount of dust retained by trees was measured as 8600 t yr -1 in an area of 103 km 2 in Zhengzhou (Zhao et al., 2002).A study in Huizhou, Guangdong province found that foliar dust contains appreciable amounts of Cd, Cr, Cu, Pb, Zn, and S as 0.040 t, 1.63 t, 2.70 t, 1.84 t, 5.54 t, and 19.52 t respectively from a study area (Qiu et al., 2009).
Vegetation is one of the most commonly cited ecosystem methods for reducing atmospheric pollutants (Pataki et al., 2011), but the ability of different taxa to accumulate particulate matter and air-borne pollutants differs (Dzierżanowski et al., 2011;Escobedo et al., 2011).For example, particulate accumulation on leaves of 22 trees and 25 shrubs was studied in Norway and Poland, but only Pinus mugo and Pinus sylvestris, Taxus media and Taxus baccata, Stephanandra incisa and Betula pendula were able to capture particulate matter efficiently (Saebø et al., 2012).High degrees of traffic saturation (80%-90%) (Notes: The degree of traffic saturation (%) is a ratio of demand to capacity on approaching to a junction) have been observed in the Chinese capital city of Beijing, which promote the release of heavy metals from automobile exhaust onto road surfaces and into the ecosystem (Wang et al., 2012).As a result, Beijing is constantly facing high concentrations of airborne dust and there is a need to improve air quality.Previous studies mostly measured metals accumulated by plant leaves but rarely paid attention to the dust collected by leaves.In order to take advantage of the phyto-remediation method to improve air quality in urban areas of Beijing, this study investigates the role of the extant urban green plants in removing airborne dust and associated metals by their leaves.In autumn, leaf samples of 32 plant species were collected from heavy traffic roadsides, and the airborne dust collected by the leaves and its associated metals were analysed in order to identify those plant species that could well accumulate or remove dust and metals, and hence provide scientific guidance for urban vegetation planting.

Site and Sampling
Beijing is located in the semi-humid warm temperature zone (39°90'N, 116°32'E) and has a continental monsoon climate, with cold dry winters and wet warm summers.Five high-speed roads were built up inside and around the city to satisfy traffic demand.In total, 96 leaf samples from 32 plant species (three replicates for each species) were collected at 14 sites alongside the west 3 rd ring road with diurnal heavy traffic (Fig. 1, Table 1) on Oct. 28 th , 2014, when there was no rain prior for 17 days to ensure that plenty of dust was deposited on leaf surfaces.Healthy intact leaves were obtained from shoots toward the street line and with height at 2.5-3.0 m above ground level for megaphanerophyte, and at 2.0 m for small trees or shrubs.Leaf samples with the same distance to traffic center were selected as possible.After collection, they were put into paper bags carefully and transported to the laboratory.

Sample Treatment and Analysis Sample Subgroup and Leaf Area Measurement
All 96 duplicate leaf samples from 32 species were divided equally into two groups to form 192 leaf samples in total.Each of them was photographed with a ruler and the leaf area was measured under Adobe Photoshop software using the formula of leaf area = known area × leaf pixel value/target pixel.
Overall, about 300 to 500 cm 2 of leaves was measured and prepared for each group of all duplicates.

Sample Pretreatment, Digestion and Measurement
In the two groups of each duplicate sample of plant species, one group of leaves was kept unwashed and another was washed with 18.2 MΩ de-ionized water.Afterwards, both washed and unwashed leaf samples were dried in an oven at 70°C, weighed and cut into pieces, ground, and kept for further chemical analysis.Referring to Liu et al. (2014), about 0.2 g leaf powder of each sample was digested with a chromatographic class HNO 3 -H 2 O 2 solution (2 mL:1 mL) using a Microwave Digestion Instrument from CEM company.The digestion procedure was set as 3-min heating up to 80°C and kept for 15 min, then up to 190°C and kept for 15 min.After the temperature naturally dropped down to room temperature, the digested solution was diluted to 25 mL with 18.2 MΩ de-ionized water.Heavy metals (Cd, Cr, Cu, Fe, Mn, Ni, Pb, Zn) were detected by TAS-990 AAS (Beijing Puxi General Company).For sample calibration, a series of metal standards of known concentration was prepared using 1000 µg mL -1 stock solution, purchased from the National Center for Analysis and Testing for Nonferrous Metals and Electronic Materials.Samples of first-grade national standard material of shrubs and leaves (GBW 07603 (GSV-2)) with certified concentrations of elements from Institute of Geophysical and Geochemical Exploration, Chinese Academy of Geological Science were also analysed for quality assurance check.

Calculation and Data Analysis
Calculation was conducted by three ways: (1) Dust accumulation (g m -2 ) by plant leaves based on leaf area was calculated using unwashed leaf weight per square meter leaves minus washed leaf weight per square meter leaves.
(2) Heavy metal accumulation (µg g -1 ) in unwashed/washed leaves was calculated simply by the metal weight (µg) in unwashed/washed leaves over the unwashed/washed leaf weight (g).(3) Leaf dust metal accumulation (µg g -1 ) based     In order to estimate the overall metal accumulation ability, metals accumulation index (MAI) formula developed by Liu et al. (2007) and Monfared et al. (2013) was adopted to calculate MAI for each plant species including unwashed leaves, washed leaves and leaf dust, where N delegates the total number of metals analyzed and I j = x/δx is the sub-index for variable j, obtained by dividing the mean value (x) of each metal by its standard deviation (δx).Metal accumulation index identifies the total metal concentration in plants.In our study we have studied eight metals, therefore the value of N equals to 8 (Liu et al., 2007).

Dust Accumulation by Plant Leaves
Various dust retention abilities can be found for different tree species even under identical environmental conditions (Freer-Smith et al., 1997;Chai and Han, 2002).Similarly, our result shows that, per square meter, leaves could remove 0.510 to 23.0 g dust with the average of 7.50 g for all 32 species (Fig. 2 and Table 2) on Beijing roadside sites.Species 24 (Koelreuteria paniculata) and 2 (Ulmus pumila) accumulated the highest amount of dust as 23.0 g m -2 and 22.4 g m -2 respectively, significantly higher than those of any other species.Leaves with surface roughness, leaf hairs, raphe, mucilage or oil, and those with short petioles absorb more dust based on previous study (Freer-Smith et al., 1997), which may explain the higher leaf dust capture rates in species 24 and 2 due to mucus secretion and coarse leaf surface.The third highest dust accumulation rate (20.5 g m -2 ) was found in species 32 (Syringa oblata), which is again significantly higher than those of the remaining species, except 30 (Malus micromalu).Species 30 accumulated dust of 15.3 g m -2 , which has no significant differences with dust collection rates in species 25 (Weigela florida cv.Red Prince, 11.9g m -2 ), 20 (Ailanthus altissima, 11.1 g m -2 ), 12 (Salix babylonica, 10.0g m -2 ), 21 (Robinia pseudoacacia, 9.14 g m -2 ) and 29 (Ligustrum × vicaryi, 9.10 g m -2 ).The lowest dust accumulation of 0.510 g m -2 occurred in leaves of species 10 (Amygdalus persica).Dust accumulation of species 1 (Sophora japonica), 2 (Ulmus pumila), 11 (Populus tomentosa), 12 (Salix babylonica), 14 (Ginkgo biloba), 17 (Fraxinus chinensis) and 24 (Koelreuteria paniculata) in Jinan city, Shangdong province of East China (Sun et al., 2015) was studied using artificial dust blowing method, and the results indicated a leaf dust accumulation sequence from high to low as for species 17, 12, 24, 11, 2, 14 and 1, which is totally different from the actual sampling results in our study, showing an order of 24 > 2 > 12 > 11 > 17 > 1 > 14.For the most common investigated species (2,12,17,20,21,24), the leaf dust capture rates at our sites are 15.1, 13.5, 8.2, 5.2, 7.6 and 15.9 times higher than those measured from Xingxiang, a southern city of China in autumn, implying heavier airborne dust pollution at the Beijing roadside (Table 2, Zhang et al., 2013), where heavy traffic always occurs.

Heavy Metal Accumulation Heavy Metal Accumulations in Leaves of Different Plant Species
Accumulation rates of metals Cd, Cr, Cu, Fe, Mn, Ni, Pb and Zn in both washed and unwashed leaves are illustrated for all plant species in Fig. 3 and Table 3, and for individual plant species in Figs.4(a)-4(h) and Table S1.

Cadmium (Cd)
In total 192 leaf samples from 32 plant species, including both unwashed and washed treatment, could accumulate Cd from 0.520 to 9.80 µg g -1 with the average of 4.71 µg g -1 (Table 3, Fig. 3), which is higher than the Cd normal value of 2 µg g -1 in plants (Hajar et al., 2014).The accumulation  rates of Cd from this study are also higher than that in plants sampled from Yan'an, a small city in western China, with a range of 0.14-3.52µg g -1 (Hu et al., 2014) and from the roadside of Florence, Italy with mean Cd values of 0.09-0.12µg g -1 in Quercus ilex leaves (Ugolini et al., 2013).
In this study, species 9 (Chaenomeles speciosa) accumulated the highest Cd of 9.48 µg g -1 in its unwashed leaves, which is significantly higher than that of any other plant species (Table S1,Figs. 4(a)), presumably due to its rich leaf oleanolic and other organic acid contents, as previous study showed that the contents of acetic and citric acids were significantly correlated with the content of Cd in leaves (Sun et al., 2006).Species 32 (Syringa oblata) had the second highest Cd accumulation rate of 8.74 µg g -1 , which is again significantly higher than those of other plant species.Leaves of species 31 (Sorbaria kirilowii) and 29 (Ligustrum × vicaryi) also accumulated a significantly higher amount of Cd (8.14 and 8.06 µg g -1 ), in comparison with those in leaves of the remaining species excluding species 28 (Cornus alba, 7.80 µg g -1 ).The accumulation rates for species 30 (Malus micromalu), 27 (Kerria japonica) and 26 (Magnolia denudata) were 7.42, 7.41, 6.81 µg g -1 respectively, and they are all significantly higher than those in the remaining 81.2% species (Table S1, Figs.4(a)).For same species, the differences of metal Cd accumulation rates between unwashed and washed leaves were various, and were only significant for species 1 (Sophora japonica), 3 (Prunus cerasifera f. atropurpurea), 7 (Populus beijingensis) and 24 (Koelreuteria paniculata).
Metal concentrations in leaves from this study were compared with results from previous studies in Table 4.The Cd concentrations in all investigated species at our sites were lower than the general levels of Cd (> 10 µg g -1 ) in plants (Markert, 1993), but exceeded the plant food toxic maximum limit of 0.5 µg g -1 in China (MPHPRC, 2012), suggesting that roadside leaves are unfit for human food or medicine usage.Cd concentration in leaves of Robinia pseudoacacia (species 21) from an industrial city Denizli, Turkey was 3.70 µg g -1 (Çelik et al., 2005), which is lower than the 5.73 µg g -1 measured in this study.Species of Ailanthus altissima (20), Ligustrum × vicaryi (29), Sophora japonica (1), Fraxinus chinensis (17), Robinia pseudoacacia (21) and Ulmus pumila (2) accumulated Cd of 5.45, 8.06, 3.18, 3.79, 5.73 and 3.08 µg g -1 in unwashed leaves and 4.83, 7.70, 2.93, 3.40, 5.50 and 2.66 µg g -1 in washed leaves respectively from this study.They are all much higher than those of 2.28, 1.96, 1.59, 1.59, 1.10, 1.36 µg g -1 in the same species leaves from the Yan'an city (Hu et al., 2014), indicating heavier Cd pollution in the capital city with frequent heavy traffic.Combustion of fossil fuels (coal and petroleum), incineration of municipal solid waste, vehicle tire wear, and combustion of vehicle lubricating oil can all contribute to airborne Cd in urban area.

Cr
All 192 plant leaf samples, including both unwashed and washed treatments from 32 species, showed accumulation rates for Cr ranged from 2.01 to 22.7 µg g -1 with an average of 6.35 µg g -1 (Table 3, Fig. 3).The current results were very similar to the Cr normal range of 0.006-18 µg g -1 in plants (Hajar et al., 2014), but much higher than those in the investigated plants leaves from Yan'an where Cr ranged as 0.021-0.952µg g -1 with averages from 0.097 to 0.365 µg g -1 (Hu et al., 2014).At the roadside of Florence Italy, the mean Cr values of 1.3-3.1 µg g -1 in Quercus ilex leaves (Ugolini et al., 2013) fall within the range of this study but towards the lower end.

Cu
For all 192 unwashed and washed leaf samples, Cu ranged from 10.6 to 47.3 µg g -1 with the mean of 19.5 µg g -1 (Table 3, Figs.3), which is roughly within the Cu normal range of 0.4-45.8µg g -1 in plants from Malacca of Malaysia (Hajar et al., 2014) but towards the higher end.The copper concentrations in plants from Yan'an ranged as 0. 58-9.97 µg g -1 with averages of 2.94-7.38 µg g -1 (Hu et al., 2014), which were much lower than those in leaves from this study.In addition, the mean Cu values of 15.3-19.8µg g -1 in Quercus ilex leaves from the roadside site of Florence, Italy (Ugolini et al., 2013) were close to the mean value in this study.In regard to individual plants, species 1 (Sophora japonica) accumulated the highest Cu of 34.1 µg g -1 in its unwashed leaves, which was significantly higher than those from any other plant species.Species 16 (Paulownia fortunei) and 21 (Robinia pseudoacacia) also collected higher amounts of Cu as 32.1 and 27.8 µg g -1 respectively, and each was significantly higher than those in the leaves of other 87.5% and 59.4% species.Ulmus pumila (2), Prunus cerasifera f. atropurpurea (3), Euonymus japonicus (6), Populus beijingensis (7), Chaenomeles speciosa (9), Populus tomentosa (11), Ginkgo biloba (14), Viburnum sargentii (15), Koelreuteria paniculata (24) and Magnolia denudata (26) showed significantly higher Cu values in unwashed leaves than that in washed ones as expected.

Fe
Metal Fe concentrations in this study ranged from 116 to 926 µg g -1 (Table 3, Fig. 3) for all washed and unwashed leaf samples, which were lower than and mostly fell outside the typical range of 640-2486 µg g -1 in plants (Hajar et al., 2014), but the mean value of 389 µg g -1 is similar to the mean values of 376.8-534.5 µg g -1 in Quercus ilex leaves from the roadside of Florence, Italy (Ugolini et al., 2013).
The highest Fe accumulation rate occurred in unwashed leaves of species 1 (Sophora japonica, 868 µg g -1 ), which is significantly higher than those in leaves of other species except 7 (Populus beijingensis, 776 µg g -1 ) and 2 (Ulmus pumila, 762 µg g -1 ), although the latter two showed significantly higher Fe values than that from other 81.2% and 75.0%species respectively.Species 9 (Chaenomeles speciosa) and 15 (Viburnum sargentii) gave a collection rate of 676 and 675 µg g -1 respectively, and they were significantly higher than those in leaves of other 62.5% species (Table S1,Figs. 4(d)).Iron concentrations, measured in a previous study in leaves of Avicennia schaueriana (up to 332.7 µg g -1 ) and Laguncularia racemosa (up to 300.9 µg g -1 ) collected from mangrove areas with high particulate iron pollution (Arrivabene et al., 2015), were still lower than those obtained at the current roadside sites.There are various inorganic elements in atmospheric aerosols, among which iron is one of the highest content (Li et al., 2005).As one of the principal elements in the Earth crust, the high iron values detected in this study may be partly due to atmospheric contamination from fossil fuel combustion, apart from soil dust.

Mn
The results for Mn from all 192 plant leaf samples revealed a range of 16.7-195 µg g -1 and a mean of 58.1 µg g -1 (Table 3, Fig. 3), which well covered the normal range of 15-100 µg g -1 in plants (Hajar et al., 2014), but were much lower than the mean values of 333.9-405.4µg g -1 in Quercus ilex leaves at the roadside site of Florence, Italy (Ugolini et al., 2013).

Ni
Ni concentrations measured from all leaf samples including both unwashed and washed treatment ranged as 2.49-25.4µg g -1 with an average of 7.38 µg g -1 (Table 3, Fig. 3), and most species showed higher values in comparison with the Ni normal range of 0.1-3.7 µg g -1 in plants (Hajar et al., 2014).
Comparison between different species showed that species 3 (Prunus cerasifera f. atropurpurea) accumulated the highest Ni concentration of 18.7 µg g -1 in its unwashed leaves, which is significantly higher than those in other species but remarkably lower than that collected by leaves of hyperaccumulating plants (plants with adaptations that allow them to take up and tolerate immense quantities of metals), with up to 3700 and 8100 µg g -1 Ni from Marivan of Iran, where high Ni concentration (1350 µg g -1 ) was also observed in soil (Ghaderian et al., 2007).Species 1 (Sophora japonica) and 4 (Platanus acerifolia) also showed significantly higher Ni concentrations (15.6 and 14.8 µg g -1 ) than those in leaves of the remaining species.In addition, Ni concentrations in species 16 (Paulownia fortunei, 11.4 µg g -1 ) and 15 (Viburnum sargentii, 10.8 µg g -1 ), ranked after 4, were also significantly higher than those in leaves of other remaining 62.5% and 59.4% species respectively.

Pb
Lead accumulation rate for all leaf samples ranged from as little as below detection limit (0.0540 µg g -1 DW) to 11.0 µg g -1 with a mean of 2.09 µg g -1 (Table 3, Fig. 3), which were close to the normal value of 3 µg g -1 (Hajar et al., 2014) and the typical concentrations of less than 10 µg g -1 (Kabata-Pendias and Pendias, 2001; Padmavathiamma and Li, 2007).Similarly, the Pb mean values of 3.3-3.7 µg g -1 in Quercus ilex leaves at the roadside site of Florence, Italy (Ugolini et al., 2013) were falling within the range of our study.
Similarly, species Sophora japonica (1), Platanus acerifolia (4), Populus tomentosa (11), Ginkgo biloba (14) and Koelreuteria paniculata (24) accumulated Pb of 2.46, 3.19, 0.823, 1.90 and 1.33 µg g -1 respectively, which were slightly lower than those measured previously at other roadside sites of Beijing for the same species as 5.69, 2.65, 2.91, 3.92 and 3.49 µg g -1 respectively (Liu et al., 2007).It may indicate that either there is less Pb pollution at the 3 rd ring road than those at other sites, or the government policy for eliminating for Pb in gasoline since 2000 has worked in reducing Pb pollution.

Zn
The accumulation rates of Zn for all leaf samples ranged from 13.7 to 246 µg g -1 with a mean of 47.4 µg g -1 (Table 3, Fig. 3), which is similar to the Zn normal range of 1-160 µg g -1 in plants (Table 4; Hajar et al., 2014), but much higher than that found in Yan'an city, which showed a range of 2.83-12.57µg g -1 for all investigated plant species (Hu et al., 2014).Mean Zn values of 40.8-42.8µg g -1 in Quercus ilex leaves at the roadside site of Florence, Italy were more or less the same as the mean of this study (Table 4; Ugolini et al., 2013).Species 11 (Populus tomentosa) collected the highest amount of Zn as 142 µg g -1 in its unwashed leaves, which is significantly higher than those in leaves of any other species, except species Sorbaria kirilowii (31) and Amygdalus triloba (19).The latter two species accumulated Zn of 123 and 113 µg g -1 , which were significantly higher than those in leaves of other remaining 90.6% and 87.5% species.Moreover, Zn concentrations of plant species 1, 2, 3, 6, 7, 9, 14, 15, 20 and 26 were significantly higher in unwashed leaves than washed leaves (Table S1, Fig. 4(h)).The highest Zn level was also observed in washed leaves of species 11 (131 µg g -1 ), but the subsequent higher values occurred in species 24 (Koelreuteria paniculata) and 12 (Salix babylonica) with accumulation rates of 80.7 and 78.8 µg g -1 , which were significantly higher than those in leaves of all other species except species 31 (Sorbaria kirilowii, 64.2 µg g -1 ).

Heavy Metal Accumulation Based on Leaf Dust for Different Plant Species
Metal contents based on per gram dust collected on leaves were calculated for the 96 pairs of leaf samples, and the results were illustrated in Table 5 for the values of maximum and mean of all plant samples and in Table 6 for the averages of individual species.Comparison was also made with previous studies at both national and international sites in Table 5.
In Huizhou of Guangdong province, a southern city of China, similar Cd concentrations (6.2-12.8µg g -1 ) but higher Pb values (434.0-512.0µg g -1 ) (Table 5; Qiu et al., 2009) were observed comparing with the results from this study, where levels of Cd and Pb were ranged as BDL-37.1 µg g -1 and BDL-237 µg g -1 .Concentrations of most metals investigated in this study are higher than that measured in previous studies for Cu, Fe, Pb, Zn from Urban Vienna, Cu, Fe, Mn, Ni, Pb, Zn in Debrecen (Hungary) and Cu, Ni, Pb, Zn from Belgrade, Serbia & Montenegro.
Metal concentrations in leaf dust were also compared with metal contents in soil samples.The results show that much higher mean levels of Cd (9.45 µg g -1 ), Cu (88.0 µg g -1 ) and Zn (308 µg g -1 ) were found in our leaf dust than those detected in the roadside soil (Chen et al., 2010).On the contrary, the mean values of Cr, Ni and Pb were lower in the leaf dust than in the soil (Table 5).The above results may indicate that leaf dust accumulation is much more efficient   et al., 2006), which is lower than those in both leaf dust in this study and in the roadside soil of Beijing in the previous study, presumably due to less traffic in the suburban area.Heavier Cd pollution in traffic area of Beijing was also reported in previous study, in which 127 urban soil samples collected from six areas in Beijing, indicating that Cd was mainly from traffic sources (Xia et al., 2011).Meanwhile, Cr and Pb values of 68.94 and 36.81 µg g -1 measured in the suburban Daxing were higher than those in the leaf dust from this study and the roadside soil from the previous study.

Metal Accumulation Index (MAI) of Different Species
To estimate the overall metal accumulation ability of individual plant species, metal accumulation index (MAI) was separately calculated for unwashed leaves, washed leaves and leaf dust.The results were shown in Fig. 5 and Table 7, with the MAI values listed in decreasing order in the table.For unwashed leaves, species 10 (Amygdalus persica) showed the highest MAI value of 387, followed by species 26 (Magnolia denudata, 35.2) and 32 (Syringa oblata, 28.1), whilst species 13 (Parthenocissus tricuspidata) exhibited the lowest MAI value of 6.35.It implies that Amygdalus persica (10) is the best multi-metal removing species.Regarding to the washed leaves, plant Punica granatum (22) had the highest level of MAI (105), followed by species 18 (Populus canadensis, 59.8) and 30 (Malus micromalu, 50.3), and the lowest value occurred in species 25 (Weigela florida cv.Red Prince, 13.0), which indicate that leaves in species 22 were most toxic and in species 25 were the safest for consumption.In addition, Viburnum sargentii (15) showed the highest MAI value of 6.46 in the dust per gram leaves, followed by species 10 (Amygdalus persica, 5.94) and 2 (Ulmus pumila, 5.72), whereas the lowest value was found in species 30 (Malus micromalu,0.450).

CONCLUSIONS
This study suggests that some plant species are suitable for planting at heavy traffic roadside of urban areas in Beijing, based on the investigation of 32 plant species for their ability to accumulate airborne dust and associated heavy metals.
To target multiple metal pollution problems such as urban roadside site, MAI value can be a better choice for selecting efficient metal removing plant species.Species Amygdalus persica (10), with the highest MAI value in unwashed leaves and the second highest value in leaf dust, was found to be the best choice for mutiple metal accumulation.Species Magnolia denudata (26) and Syringa oblata (32) can also be selected due to their higher than average MAI values in unwashed leaves.Plants Punica granatum (22), Populus canadensis (18) and Malus micromalu (30) can be used based on metal reduction in washed leaves, whilst species Viburnum sargentii (15), Amygdalus persica (10), Ulmus pumila (2), Prunus cerasifera f. atropurpurea (3) and Magnolia denudata (26) may be considered based on MAI values for leaf dust.

Fig. 2 .
Fig. 2. Dust accumulations by leaves of different plant species (Note: error bar means standard deviation).

Fig. 4 .
Fig. 4. Heavy metal accumulations in unwashed and washed leaves of different plant species (Notes: ** : Refers to significant difference between metal concentration in unwashed and washed leaves for same plant species at 0.01 level.* : Refers to significant difference between metal concentration in unwashed and washed leaves for same plant species at 0.05 level).

Fig. 5 .
Fig. 5. Metal Accumulation Index in unwashed, washed leaves and dust per gram leaves for different plant species.

Table 1 .
Sites with plant species No. information.
calculated with Microsoft Office Excel 2007.One-way analysis of variance was carried out with SPSS17.0.In addition, Post Hoc Multiple Comparisons were applied to obtain the significance levels of differences between plant species.

Table 2 .
Dust accumulations by leaves of different plant species (g m -2 leaf).

Table 3 .
Heavy metal accumulation in all plant leaves.

Table 4 .
Metal concentrations in leaves of this study and previous studies (µg g -1 leaf).
Note: NA: not available; BDL: below detection limit.

Table 5 .
Comparison between metal concentration of leaf dust in this study and those of leaf dust and soil in previous studies (µg g -1 dust or soil).

Table 6 .
Heavy metal concentrations in leaf dust for individual species (mean ± Std.Error of Mean).

Table 7 .
MAI in unwashed leaves, washed leaves and leaf dust for different plant species (Ranked high to low).