Trait-mediated leaf retention of atmospheric particulate matter in fourteen tree species in southern China

Particulate air pollution is a serious threat to human health, especially in urban areas, and trees can act as biological filters and improve air quality. However, studies on greening tree species selection are rare. We measured three particular matter adsorption metrics (PM2.5, PM2.5–10, and PM>10 captured per leaf area) and six functional traits for each of fourteen species and estimated their minimum light requirements based on field surveys. We found that shade-tolerant species captured more coarse particles (PM2.5–10) than light-demanding species. For traits, a strong negative correlation was found between photosynthetic capacity and adsorption capacity for all three PM size fractions, indicating that in comparison to acquisitive species, conservative species captured larger amounts of particles. Moreover, denser wood species and smaller leaves were more efficient in capturing large particles (PM>10), while species with “expensive” leaves (high leaf N or P) were more efficient in capturing fine particles (PM2.5), indicating that capturing large and fine particles was related to mechanical stability traits and leaf surface traits, respectively. Our results demonstrated that the metabolism (e.g., photosynthetic capacity) and chemistry (e.g., leaf N and leaf P) of leaves help explain species capacity to capture PM. We encourage future studies to investigate the ecosystem functions and stress tolerance of tree species with the same framework and trait-based methods.


Introduction
Developing countries such as China have experienced fundamental changes in recent decades, while gains in wealth have been achieved at a severe cost to the environment Xu et al. 2018). Air pollution is a major threat to people's health, particularly in urban areas (Przybysz et al. 2014;Yan et al. 2016), due to particulate matter (PM) emissions linked to traffic and industry (Simon et al. 2016). Coarse particles can cause lung diseases, while smaller particles, such as PM 2.5 , can be inhaled more deeply into the lungs (Beckett et al. 2000;Liang et al. 2016) and have been widely recognized as more harmful to human health. Air pollution is estimated to contribute to at least 5 million premature deaths each year (WHO 2006).
Vegetation can reduce particle concentrations and thereby improve air quality (Litschke and Kuttler 2008;Liu et al. 2013;Shi et al. 2017;Weerakkody et al. 2018a). A leaf is the main organ for particle deposition (Weerakkody et al. 2018b) and can even biodegrade or transform pollutants into less or nontoxic molecules with its habituated microbes and Responsible Editor: Elena Maestri endophytes . Recognizing what kind of plants are more powerful in capturing airborne PM is important but challenging because the retention capacity of leaves in terms of accumulating PM is influenced by a variety of factors, such as the concentration of atmospheric PM (Luo et al. 2020), days after rainfall , wind speed (Beckett et al. 2000), leaf stage (Nguyen et al. 2015), and sampling season , which are linked with surface moisture (Wang et al. 2013;Sun et al. 2020). Therefore, it is necessary to investigate what kind of trait could improve species' PM adsorption abilities (Janhäll 2015;Shao et al. 2019). For example, species with smaller leaves, more complex leaf structure, waxy leaves, and hairier leaves usually capture PM more efficiently (Beckett et al. 2000;Dzierżanowski et al. 2011;Weerakkody et al. 2017;Sun et al. 2018;Liu et al. 2019). These findings have important significance for selecting greening tree species, which is the primary task in the construction of urban forests.
In comparison to ecosystem functions, environmental stress tolerance is sometimes even more important in greening tree species selection Rathore 2018, 2019;Przybysz et al. 2021). However, there are very few studies investigating species' PM adsorption abilities in relation to species' environmental stress tolerance, such as shade tolerance and drought tolerance. Therefore, it is of great importance to test whether species with high PM adsorption abilities are fast-or slow-growing species and whether they are resource-demanding species. Because light is the major limiting resource in moist forests (Zhao and He 2016), we specifically sought to determine how PM adsorption abilities are correlated with the light requirements of tree species (question 1).
Theoretically, plant traits can track environmental changes and reflect the adaptive strategies of plants (Liu et al. 2021). Dust accumulation reduces the light availability of leaves ) and may inhibit chlorophyll biosynthesis (Chen et al. 2015) and stomatal conductance (Lewis et al. 2017), which is important for plant photosynthesis (Hetherington and Woodward 2003;Yu et al. 2018). Consequently, species with a greater ability to accumulate PM might need to be more shade tolerant from the perspective of evolution. We therefore hypothesized that species with high PM accumulation abilities should be able to adapt to poor light environments (hypothesis 1).
In forest ecology, there is a well-established trade-off between the survival rate in deep shade and the growth rate in bright light (Adler et al. 2014). Fast-growing species that rapidly acquire resources thrive in rich light environments, while slow-growing species that conserve resources are dominant in poor light environments (Lohbeck et al. 2013). This acquisitive-conservative trade-off points to the extremes of a continuum in plant design, and the position of a species along this continuum can be quantified by its functional traits (Poorter et al. 2008;Wright et al. 2010). For instance, an acquisitive species tends to have a high photosynthetic capacity, dark respiration rate, specific leaf area (SLA), and leaf nitrogen content (LNC) and hence a high growth rate. In contrast, conservative species tend to have thicker leaves and a higher wood density (WD), reducing the volumetric stem growth rate but facilitating leaf and stem protection and high survival rates (Chave et al. 2009). In this study, we tested whether species with high PM adsorption abilities are characterized by conservative or acquisitive functional traits (question 2).
Conservative species commonly have high WD and leaf thickness but a low SLA and photosynthetic capacity, suggesting a low growth rate strategy (Firn et al. 2019). Low growth rate species tend to produce durable leaves with long life spans (Adler et al. 2014). The tough leaves of these species may be associated with complex leaf surface structure and waxes, which could trap PM (Dzierżanowski et al. 2011). Therefore, we hypothesized that species with high PM accumulation capacities should be characterized by conservative traits (hypothesis 2).
If these two hypotheses were confirmed, then we would test the third question relating to which traits are good predictors of the PM accumulation capacities of species (question 3). The functional traits measured in this study, including photosynthetic capacity, leaf length, SLA, WD, LNC, and leaf phosphorus content (LPC), represent relatively easily measured characteristics that can be obtained for large numbers of species . Relationships among functional traits and PM adsorption capacities may reflect direct effects but could also result from trait coevolution (Reich et al. 2003). Therefore, we used structural equation modeling (SEM) to investigate the relationships among a suite of traits and PM adsorption capacities of different size fractions.
To test the first two hypotheses and establish linkages between plant functional traits and ecosystem functioning (PM adsorption in this study), we sampled leaves from 14 evergreen tree species (Fig. 1) and experimentally measured their PM adsorption capacities. We then quantified shade tolerance and measured six functional traits of these species and compared them with their PM adsorption abilities.

Study site
The study was conducted in a 50-ha dynamic forest plot in the Heishiding Provincial Nature Reserve (23°27′N, 111°52′E; altitude, 150-927 m), located in Guangdong Province, southern China. This region is covered by a subtropical evergreen broad-leaved forest and has a subtropical moist Fig. 1 Leaves of the fourteen tree species tested in the study. The species are ordered based on their total suspended particle accumulation capacity per leaf area 33611 Environmental Science and Pollution Research (2023) 30:33609-33623 monsoon climate. The annual precipitation is approximately 1700 mm, and the mean daily temperature is 19.6 °C.

Quantitative assessment of PM capturing ability
Samples were collected from three individuals for each of the 14 species. For each individual, 10-50 fully expanded mature leaves were collected at 2-5 m above-ground with a high branch scissor. Samples from the same tree were then put in a plastic bag, sprinkled with water, transported to the laboratory, and stored in a refrigerator at 4 °C until the laboratory experiment.
To ensure that all samples experienced similar dust conditions, we used a dust box, a dust grinder, and an air blower ( Fig. 2) to simulate a busy road environment, an approach similar to deposition chamber experiments (Hwang et al. 2011) or wind tunnel experiments (Burkhardt et al. 1995;Räsänen et al. 2013;Zhang et al. 2017). First, we cleaned the samples by immersing them with 250 ml distilled water and washed them with a no-hair-loss brush. Next, the cleaned leaves were hung on a rope installed in an experimental box (1.4 × 1.2 × 0.8 m, Fig. 2). Sixty grams of soil was ground to powder, sieved, and blown into the box to simulate a cloud of dust and left to set for half an hour. We then removed the dusted leaves and measured the PM retained on them.
The method for rinsing and weighing was similar to that described by Dzierżanowski et al. (2011). The filters and small beakers were dried for 60 min at 80 °C in a drying oven and then left in the weighing room for an hour to stabilize the humidity and obtain a constant weight (W 1 ). We immersed leaves with distilled water, cleaned them softly with a nylon brush, and then washed them with distilled water. In the next step, the rinse water was filtered sequentially using 10-μm and 2.5-μm filters (Whatman, UK, Types 91 and 42, respectively), and the eluent was poured into small beakers. The filters and beakers were oven dried at 80 °C for 8 and 12 h, respectively, and then left in the weighing room for an hour to obtain constant humidity and weight (W 2 ). Leaf areas (S) of the samples were measured using an Epson Expression 11000XL scanner (Epson, Inc., USA) and calculated using the Photoshop CC2019 software (Adobe Corp., USA). The amount of PM was expressed per unit area (cm 2 ) of one side of the leaf surface and calculated as (W 2 − W 1 )/S.

Shade tolerance estimation
The 50-ha plot (1000 × 500 m) was established following the ForestGEO (https:// www. fores tgeo. si. edu) protocol in 2011, and all free-standing individuals with diameters at breast height greater than 1 cm were tagged, mapped, measured, and identified to species. More than 270,000 stems belonging to 245 species were recorded in the first census. The understorey light availability was estimated using hemispherical photography at 2352 locations randomly distributed within the 50-ha plot (Zhao and He 2016), and light availability was then interpolated to grids of 10 × 10 m using ordinary kriging in the AUTOMAP package (Hiemstra et al. 2009) to estimate the environmental conditions where the individuals were located. The minimum light requirements (shade tolerance) of the species were estimated by calculating the 10th percentile of the distribution of juveniles (height < 4 m) in relation to light availability. There were at least 200 juveniles for each of the 14 species in the 50-ha plot.

Functional trait measurements
We measured six traits, including leaf length (LL), maximum rate of electron transport (J m ), specific leaf area (SLA), wood density (WD), leaf nitrogen concentration (LNC), and leaf phosphorus concentration (LPC). Leaf length was averaged from the samples collected for PM accumulation measurements. However, the other five traits were measured separately at other times on different sampled individuals. Shade-tolerant species with durable leaves tend to have lower J m , higher WD, and lower SLA than those of lightdemanding species with high leaf turnover rates (Walters and Reich 1999;Reich 2014).
J m represents the light saturated photosynthetic capacity per leaf area (Rascher et al. 2000). For J m measurements, a total of 1056 leaves from 335 individuals of the 14 species were sampled in the growing season (May-September) of 2014. A PAM fluorometer (PAM-2000; Walz, Effeltrich, Germany) was used with a leaf clip holder (2030-B) to measure the quantum yield of PSII. The gradient of photon flux density was set to 0,21,76,134,249,371,456,581,726,895,1267,1405,1790, and 2190 µmol/(m 2 ·s), and the measurement was processed under the automatic light-curve program (Feng et al. 2018). The Smith model (Smith 1937) was used to fit light-response curves with the measured photosynthetic data. J m was calculated from the fitted light-response curves.
SLA, LNC, and LPC were measured in a total of 464 individuals of the 14 species during September through November 2014. Initially, the leaf mid rib was removed; thereafter, a photograph was taken using an Epson Expression 11000XL scanner (Epson, Inc., USA) and later analyzed using the Photoshop CC2019 software (Adobe Corp., USA). Leaves were then oven dried at 70 °C for at least 48 h, the dry mass was measured, and the SLA was calculated. Leaf nitrogen and phosphorus contents were determined using the Kjeldahl acid digest method as detailed in Zhu et al. (2021).
Wood density was measured in a total of 280 individuals of the 14 species. For each species, 20 individuals were selected. For each selected tree, an outer crown twig of noncurrent year was harvested to measure wood density. All wood samples were wrapped with moist filter paper for transport to the laboratory. After removing the bark material, the wood sample was cut into a smooth column, and the volume was measured using water displacement. The wood dry mass was weighed after 72 h of oven treatment at 80 °C .

Empirical data on functional groups
Acquisitive species generally correspond to light-demanding pioneer species that are fast growing and characterized by acquisitive traits (Russo et al. 2021). We categorized the 14 species into acquisitive/conservative groups with reference to Flora of China (http:// www. eflor as. org/) and Zhou et al. (1999). Zhou et al. (1999) focused on the successional seral stages of species in the Heishiding Nature Reserve, in which species that reached maximum abundance by 35 years after clear cutting were considered early successional species (Feng et al. 2018). We classified early successional species as acquisitive species. Furthermore, species that were not included in Zhou et al. (1999) but were described as "pioneer species" in Flora of China were classified as acquisitive species. At the same time, species described as "mesophyte," "living in shady environment," "shade tolerant" in Flora of China, or reaching maximum abundance after 35 years of clear cutting were assigned to the group of conservative species.

Statistical analysis
One-way analysis of variance (ANOVA) was applied to test whether individual PM values differed among species. If the difference among the species was significant, then the significance of the differences between the mean values was tested using Tukey's honest significant difference test (HSD) at α = 0.05 with the function "HSD.test" in the R package "agricolae" (R Development Core Team 2016). Differences between acquisitive and conservative groups in capturing PM of different size classes per leaf area were assessed using Student's t-test following confirmation of normality using the Shapiro-Wilk test.
To test the relationship between shade tolerance and PM accumulation capacities, we examined the response of PM to minimum light requirements using simple linear regressions. To test the effect of functional traits on PM accumulation capacities, we also examined the response of PM to traits using simple linear regressions. To better meet the assumptions of normality and homogenous variances, PM, shade tolerance, and trait values, except for WD, were log transformed before analysis.
Because functional traits are correlated with each other, bivariate relationships alone provide limited information. We used a structural equation model with the function "psem" in the R package "piecewiseSEM." Initially, all possible paths were allowed in the model, and the complex model was further simplified by removing nonsignificant paths. This approach is useful for exploring the comparative strength of relationships. A bidirectional line was used to denote the equivalent models.

Relationships between shade tolerance and PM accumulation capacities
We found that all three PM accumulation metrics were negatively correlated with the minimum light requirements (shade tolerance) of the species, but only the relationship between PM 2.5-10 and shade tolerance was significant (R 2 = 0.72, P < 0.001, Fig. 4). C. carlesii, A. quinquegona, and X. hainanense were observed as the most shade-tolerant species, and they ranked third, sixth, and second in accumulating TSP per leaf area, respectively. M. chinensis and H. heptaphyllum were observed to be the most light-demanding species and the least effective plant species for capturing TSP per leaf area.

PM accumulation capacities of different functional groups
Nine of the 14 species were classified as conservative species, and five were classified as acquisitive species (Table 1). We found that conservative species captured larger amounts of PM 2.5 per leaf area (0.42 ± 0.21 g·m −2 ) than acquisitive species (0.23 ± 0.05 g·m −2 ; P < 0.05), whereas the two functional groups (acquisitive and conservative) showed nonsignificant differences in capturing PM 2.5-10 and PM >10 (P > 0.05).

Bivariate relationships between functional traits and PM accumulation capacities
For the morphological traits, we found that PM 2.5-10 and PM >10 accumulation capacities were negatively correlated with leaf length (Fig. 5b, c), whereas nonsignificant correlations were observed between PM 2.5 accumulation capacity and leaf length (Fig. 5a). For the metabolism traits, photosynthetic capacity (J m ) was negatively correlated with PM accumulation capacity for all particle size fractions ( Fig. 5d-f).
For the economic traits, the relationships between specific leaf area (SLA) and PM 2.5 , PM 2.5-10 , and PM >10 accumulation capacities were nonsignificant (Fig. 6a-c). Wood density (WD) was positively correlated with PM >10 accumulation capacity (Fig. 6f), whereas it was not significantly  correlated with PM 2.5 and PM 2.5-10 accumulation capacities (Fig. 6d, e). For the leaf chemical traits, leaf nitrogen content (LNC) was positively correlated with PM 2.5 (Fig. 7a), whereas it was nonsignificantly correlated with PM 2.5-10 and PM >10 accumulation capacities (Fig. 7b, c). Similar patterns were found for leaf phosphorus content (LPC), where PM 2.5 accumulation capacity was positively correlated with LPC   (Fig. 7d), and the relationships between PM 2.5-10 and PM >10 and LPC were nonsignificant (Fig. 7e, f).

Path analysis
Among the six traits LNC, LPC, SLA, WD, leaf length, and J m , mainly J m , leaf length, and LPC influenced PM accumulation capacities. Leaf length decreased the PM >10 accumulation capacity, and LPC increased the fine particle (PM 2.5 ) accumulation capacity. The photosynthetic capacities of leaves (J m ), which may be related to cuticle thinness, decreased the PM retention capacities of all three size classes (Fig. 8).

Discussion
Leaf characteristics, including both leaf architecture and material composition, may influence the ability of a leaf surface to accumulate PM (Liang et al. 2016). In the present study, we found that suites of functional traits displayed a tight association with the PM accumulation capacity of leaf area for each species. We showed that species with a small leaf size (low LL), low photosynthetic capacity (low J m ), high input into construction (high WD), and a conservative nutrient use strategy (high LNC and LPC) increased in PM accumulation capacities. Our results indicate that conservative species (characterized by low J m and high WD) may have higher PM accumulation capacities than acquisitive species (characterized by high J m and low WD) and that species with high PM accumulation capacities are adapted to poor light environments.

Leaf size affected PM retention capacity
We found that smaller leaves captured more PM per unit leaf area than larger leaves, especially for coarse and large particles (Fig. 5a-c). The three most effective plant species in capturing TSP per leaf area, X. hainanense, C. carlesii, and C. concinna, had the shortest leaves among the 14 species (Table 1). In line with our results, Weerakkody et al. (2018b) found that smaller synthetic leaves accumulate more PM than larger synthetic leaves with the same surface characteristics. This correlation between leaf size and PM accumulation capacity was also supported by field surveys (Leonard et al. 2016) but not by the results of some other  (Saebø et al. 2012). Scanning electron microscopy images have shown that more particles are distributed near leaf edges (Weerakkody et al. 2018b); therefore, in comparison to broadleaved species, conifer species commonly capture larger amounts of particles (Freer-Smith et al. 2005;Liu et al. 2012).
However, broadleaved species have been reported to have higher leaf retained PM wash-off efficiency than conifer species (Luo et al. 2020), and PM wash-off events are important for the net removal of airborne PM ).

Shade-tolerant species captured larger amounts of PM per area than light-demanding species
The stress tolerances of species (e.g., drought tolerance or shade tolerance) are sometimes even more important than the functions of species when we design vegetation barriers for air pollution abatement (Barwise and Kumar 2020). Species must adapt to the shade conditions created by the deposited dust on leaves, tall buildings, and other vegetation. In comparison to light-demanding species, shade-tolerant species have been shown to better survive under shaded conditions, but they have lower maximum photosynthetic rates under high light conditions (Valladares and Niinemets 2008). We found a significantly negative correlation between minimum light requirements and PM 2.5-10 retention capacities for the species. Lightdemanding species, e.g., M. chinensis and H. heptaphyllum, accumulated less PM than shade-tolerant species, such as C. carlesii and A. quinquegona. Moreover, lightdemanding species usually have higher photosynthetic capacities than shade-tolerant species. Therefore, the finding that photosynthetic capacity (J m ) was significantly negatively correlated with the PM retention capacity of the species for all particle size classes (Fig. 5d-f) confirmed that species with high PM accumulation abilities are adapted to poor light environments. In this study, we showed that J m is a good predictor and that it alone predicted 45-64% (for different PM size classes) of interspecific variation in PM accumulation. Compared with other commonly used traits, such as hairiness, roughness, and waxiness of the leaf surface, photosynthetic capacity (J m ) is objective and easily measured.

Relationships between economic traits and PM accumulation capacities
For leaf economics, an axis exists from "cheap" tissue investment and fast returns to "expensive" tissue investment and slow returns (Wright et al. 2004). Species with "expensive" economic traits, such as low specific leaf area (SLA), tend to have a long leaf lifespan and low relative growth rates but high survival rates (Adler et al. 2014). These species produce durable leaves with thick cuticles (Kitajima 1994) where interactions between PM and leaves occur. We thereby expected that "expensive" leaves should have high PM accumulation capacities, and a negative relationship between SLA and PM deposition rate had been reported by a previous wind-tunnel experiment (Chiam et al. 2019). Unexpectedly, nonsignificant correlations were found between SLA and all three PM size fractions in this study . Species with high SLA (e.g., C. subulatum, H. heptaphyllum, and I. chinensis) were found to have similar PM capturing capacities to species with low SLA (e.g., M. seguinii, M. breviflora, and L. litseifolius). This result probably occurred because we used the species mean SLA value to characterize all individuals of a species. However, intraspecific SLA variation can arise from both phenotypic plasticity and genetic diversity (Siefert et al. 2015). In comparison to SLA, some other leaf traits may influence PM accumulation on leaf surfaces more directly. For instance, a comprehensive study on 96 plant species showed that PM accumulation on leaf surfaces is influenced by leaf trichomes (Muhammad et al. 2019), leaf wettability, and epicuticular wax structure types (Muhammad et al. 2020).
We found that species with denser wood (e.g., X. hainanense, M. seguinii, and S. superba) accumulated larger amounts of large particles (PM >10 , Fig. 4f). Species with high wood density (WD) represent a "slow" life history strategy (Wright et al. 2010); i.e., they have a limited potential growth rate but have low mortality risk under shaded conditions. A negative relationship between WD and leaf size was found in this study (Fig. 6) and has also been reported in some previous studies (Wright et al. 2007). This scenario occurs probably because species with denser wood usually have lower hydraulic conductivity and thereby cannot support large leaves as species with soft wood (Chave et al. 2009). Wet leaves more easily retain PM; however, denser wood species with lower hydraulic conductivity were found to have higher PM accumulation capacity (Fig. 6f). The correlations between WD and PM >10 , as the path analysis demonstrated (Fig. 8), were the result of a complex correlation network in the trait space.
We found that leaf N and P concentrations were both positively correlated with PM 2.5 accumulation capacities (Fig. 7a, d). Species with high N and P concentrations, such , and maximum rate of electron transport (J m ) influence species abilities to capture fine (PM 2.5 ), coarse (PM 2.5-10 ), and large (PM >10 ) particulates in the 14 species tested in this study. The model was tested with Fisher's C statistic, test statis-tic = 11.12 with 14 model degrees of freedom and P = 0.676 (indicating close model-data fit). Path coefficients were calculated based on standardized values. Solid lines represent positive paths, and dashed lines represent negative paths. Asterisks denote significance: ***P < 0.001; **P < 0.01; *P < 0.05. Nonsignificant paths were excluded from the model as C. subulatum, showed a greater potential to capture PM 2.5 . N and P are the two most limiting elements to plants and have specific functions in leaves. N is important for enzymatic activity, and P affects protein synthesis. Greater leaf N and P concentrations may enhance metabolic activity and facilitate the formation of complex leaf surface structures.
Except for thicker cuticles for adsorbing PM, trees with "expensive" leaves (low SLA) and denser wood may have higher structural stability (Reich et al. 1991) and sway less in wind, which may reduce the resuspension of PM back into the air (Chiam et al. 2019) and further improve their PM capturing efficiencies.

Accumulation of different size fractions of PM
PM of different sizes has varying effects on human health (Przybysz et al. 2014), with the smallest fraction considered to be the most hazardous. Particles > 100 μm in diameter have not been measured in many studies (Przybysz et al. 2014;Xu et al. 2017;Cai et al. 2017;Aguilera Sammaritano et al. 2021). However, large particles near busy urban roads are not harmless, especially when they contain large amounts of heavy metals such as mercury and lead (Liu et al. 2012;Aguilera Sammaritano et al. 2018). We did not exclude PM >100 in the washing solutions, but the ground soils used in this experiment were sieved through a metal sieve (retention 300 μm) before being blown into the experimental box. Large particles (PM >10 ) made up the greatest mass proportion (48.9-75.3%) of total suspended particles, followed by the coarse fraction (7.8-35.6%) and then the fine fraction (9.7-36.1%). We found that the accumulation of large particles on leaves was related to leaf morphology (leaf length), while the accumulation of fine particles was related to leaf chemistry. The mechanisms behind this result need further research. We hypothesize that large particles may mainly be captured by interception, while fine particles may mainly be captured by waxes or deposited in the microstructures (e.g., grooves), which may be related to LNC and LPC. Fine particles are not easily removed, which may be a possible reason for the poor correlation between morphology representing structural stability and fine particle accumulation capacities.
We would like to note that leaves in our experimental dust box might have different mechanical or physiological traits from leaves on field trees. Experimental investigations (e.g., deposition chamber experiments and wind tunnel experiments) and field surveys have their own advantages. Experimental investigations can control environmental factors such as the PM concentration of air and wind speed (Burkhardt et al. 1995;Räsänen et al. 2013;Zhang et al. 2017), while field surveys can maintain the mechanical and physiological characteristics of plant leaves (Aguilera Sammaritano et al. 2018). We encourage future studies to investigate relationships between plant traits and leaf PM accumulation abilities with both experimental investigations and field surveys to comprehensively understand the mechanisms of PM deposition by plant species.
Trees stand regularly occur along roads in many cities, and therefore, quantifying individual-level PM accumulation capacity is needed. We encourage future studies to integrate leaf-level PM accumulation capacity with leaf density and crown size to obtain whole-plant PM accumulation capacity and link it with functional traits of species.

Conclusion
PM in different particle size fractions captured per leaf area differed among species. C. concinna, X. hainanense, and C. subulatum were efficient species in capturing PM and might be suitable species for the roadside plantation. We recommended easily measuring traits such as light saturated photosynthetic capacity per leaf area (J m ) and leaf length to determine the PM accumulation abilities of plant species.
Our results provide encouraging evidence of correlations between species functional traits and abilities to capture airborne particulates, including PM 2.5 , PM 2.5-10 , and PM >10 . J m was negatively correlated with all three PM size fractions, indicating that conservative species accumulated larger amounts of PM than acquisitive species. Denser wood species and smaller leaves captured more large particles (PM >10 ), while "expensive" leaves (high leaf N or leaf P) captured more fine particles (PM 2.5 ), indicating different mechanisms for capturing particles of different sizes. Although the correlations between economic traits (specific leaf area and wood density) and PM deposition were weak, our findings highlight the great potential of functional traits as a tool for linking species' efficiency at air purification with their growth strategies and environmental stress tolerances.