Pollution monitoring using the leaf-deposited particulates and magnetism of the leaves of 23 plant species in a semi-arid city, Northwest China

We conducted a study of the leaf-deposited particles and magnetism of plant leaves in different functional areas (traffic areas, parks, and residential areas) in Lanzhou, China. The saturation isothermal remanent magnetization (SIRM) of the washed and unwashed leaves of 23 plant species (including evergreen shrubs, deciduous shrubs, deciduous liana species, and deciduous trees) at three sampling heights (0.5 m, 1.5 m, and 2.5 m) was measured. In addition, the mass of the leaf-deposited particles was measured using the elution-filtration method and the leaf morphological characteristics were determined by scanning electronic microscope (SEM) analysis. The results revealed significant differences in particle retention capacity among the 23 plant species, with evergreen shrub species at the heights of 0.5 m and 1.5 m having higher particle concentrations. Buxus sinica, Buxus megistophylla, Prunus cerasifera, and Ligustrum×vicaryi were the most effective plant species for accumulating particles. The SEM results showed that leaves with a relatively complex adaxial surface (such as deep grooves and protrusions) were more effective at accumulating particles. The SIRM of washed leaves, unwashed leaves, and leaf-deposited particles were significantly higher in traffic areas than in parks and residential areas. In addition, significant correlations were found between SIRM of unwashed leaves and leaf-deposited particles and the mass of leaf-deposited particles, and therefore the leaf magnetic properties effectively reflect levels of PM pollution under different environmental conditions. Overall, our results provide a valuable reference for the selection of plant species with high particle retention capacity that is suitable for urban greening and pollution mitigation.


Introduction
Human health and well-being are closely related to the quality of the urban environment. Approximately 55% of the global population lives in cities and the trend of increasing urbanization and the growth of cities will continue. However, poor urban air quality diminishes people's sense of well-being and increases health risks such as respiratory (Evans et al. 2014) and cardiovascular diseases (Wilker et al. 2013). Consequently, urban development is increasingly concerned with building an ecological and pleasant living environment. As one of the most important aspects of such efforts, urban vegetation plays a crucial role in optimizing the urban environment by regulating the microclimate, reducing noise, and mitigating particulate matter (PM) pollution (Baldauf 2017). The benefits of vegetation have been demonstrated in studies using model simulations (Gromke and Ruck 2009;Selmi et al. 2016), wind tunnels (Gromke and Ruck 2009;Wang et al. 2019), and field experiments (He et al. 2020;Shao et al. 2019;Yan et al. 2019). Urban green infrastructure, such as hedges, in both closed and open-road environments, significantly reduces pollutants such as particulate matter (Al-Dabbous and Kumar 2014), black carbon (Abhijith and Kumar 2019), and trace metals (Abhijith and Kumar 2019).
Although several studies have shown that trees in street canyons may cause a deterioration in air quality by reducing ventilation (Morakinyo et al. 2016;Yli-Pelkonen et al. 2017), appropriate urban greening vegetation can overall improve the urban environment.
Environmental magnetism, especially measurements of saturation isothermal remanent magnetization (SIRM), is widely used for monitoring of PM pollution (Baldacchini et al. 2017;Castanheiro et al. 2016;Chen et al. 2019;Hofman et al. 2016), which is present as leaf-deposited and leaf-encapsulated particles in vegetation (Jouraeva et al. 2002;Popek et al. 2013). However, previous studies have shown that there are significant inter-species differences in the efficiency of PM removal by plant leaves because of differences in vegetation type (Islam et al. 2012) and leaf morphology (Shao et al. 2019;Wang et al. 2013). Therefore, it is necessary to carefully investigate the particle retention capacity of the leaves of common urban greening plants, and to determine whether the particle retention capacity of leaves is consistent with the magnetic characteristics of the leaves of different plant species in different environmental conditions, in order to see which plant species performs best in biomagnetic monitoring.
In addition, external factors, for instance, meteorological factors (Luan et al. 2019;Xu et al. 2017) and the pollution intensity (Przybysz et al. 2014), have been shown to be critical factors affecting the retention of PM on the surface of leaves. In addition, leaf waxes can also incorporate different fractions of PM (0.2~100 μm) (Dzierzanowski et al. 2011). Consequently, the occurrence of PM on or in leaves can be tested by washing with water and chloroform and SIRM measurements of unwashed and washed leaves. The differences in the magnetism of washed and unwashed leaves help to understand the contribution of leaf-deposited particles to the magnetic signal. Such research can potentially help formulate urban greening management policies, such as determining the frequency and composition of green belts.
The specific aims of the present study were as follows: (1) to identify differences in the levels of PM pollution among areas of high traffic density, parks, and residential areas in Lanzhou, a major city and transport hub in Northwest China; (2) to explore the differences in particle retention and magnetic contamination of the leaves of different plant species; and (3) to evaluate differences in the magnetism of washed and unwashed leaves.

Study area
Lanzhou City is located in Northwest China (Fig. 1), at the convergence zone between the Tibetan Plateau, Inner Mongolia Plateau and the Chinese Loess Plateau. Lanzhou is situated in a basin surrounded by mountains, at an average elevation of 1500 masl. Lanzhou is a typical Chinese city located in the semi-arid area with low annual average wind speed (~1.24 m·s −1 ) and an annual precipitation of less than 400 mm. Air pollution in urban Lanzhou is mainly characterized by a mixture of suspended dust, soot, volatile organic compounds (VOCs), and motor vehicle exhausts. The annual concentrations of PM 10 and PM 2.5 exceed the 2nd grade standard of the state, China (70 μg·m −3 and 35 μg·m −3 ), and the quantity of dust fall is 12.81 ton km −2 month −1 in 2016 (Lanzhou Municipal Environmental Protection Bureau 2016). High concentrations of heavy metal in dust fall were observed, which were mainly contributed by anthropogenic sources (Li et al. 2020). According to the statistical yearbook of Lanzhou, Lanzhou had a population of > 3 million inhabitants. The total number of civilian cars in the city was 898,800 by the end of 2016, and the road traffic density (total number of motor vehicles/total length of roads) is approximately 750 cars·km −1 , causing serious traffic congestion. Traffic pollution has become one of the important factors affecting environmental pollution in Lanzhou. Moreover, due to the influence of valley topography, the urban inversion layer occurs in approximately 80% of the year, and when it does, it is long-lasting and sometimes lasts during the whole day, which is not conducive to the diffusion of atmospheric pollutants.

Sampling
In order to compare the PM contamination and magnetic characteristics of the surface of leaves under different environmental conditions, leaf samples were collected from three different functional areas ( Fig. 1): (I) traffic areas, where the distance between the sampling sites and main road was less than 5 m; (II) parks, where samples were collected from the central area that was at least 1 km far away from the road, to minimize the impact of traffic; and (III) residential areas, hospitals, and university campuses, with a high frequency of various human activities and many domestic emissions.
In August 2016, leaf samples from 23 plant species were collected (SI Table 1). They included 1 deciduous liana species, 2 deciduous shrub species, 4 evergreen shrub species all sampled at a height of 0.5 m above the ground surface, 1 evergreen shrub species, 4 deciduous shrub species, and 3 deciduous tree species all sampled at the height of 1.5 m above the ground surface, and 8 deciduous tree species sampled at the height of 2.5 m above the ground surface. The plant species were divided into 7 types depending on plant type (shrubs and trees; deciduous (D) and evergreen (E) species) and the height of sampling above the ground surface: 0.5 m-lianas-D, 0.5 m-shrubs-D, 0.5 m-shrubs-E, 1.5 m-shrubs-D, 1.5 mshrubs-E, and 2.5 m-trees-D. The oldest leaves in the newest twigs were chosen, and mature and undamaged leaves were collected from the outer canopy. All leaves were sampled from the side facing to the road in traffic areas and any of the sides in residential areas/parks. In order to reduce random error, three subsamples were collected for each species, and the number of leaves for each subsample depends on the size of the leaf. The weather conditions during sampling were dry (there was no precipitation during the 15-day preceding sampling) and the air was still and clear. Disposable plastic gloves were used for sampling and they were changed after each sample to avoid cross-contamination. Leaves were tightly packed together using cling film to prevent movement and were then placed in a plastic pot of 8 cm 3 volume in the field. All leaf samples were refrigerated at 5°C prior to transport to the laboratory. In total, 601 sampling sites were chosen; 4056 leaves were selected for SIRM and mass of leaf-deposited particles measurement (SI Table 1).

Magnetic measurements
Unwashed and washed (washing procedures are shown in the "Mass of leaf-deposited particles" section) leaf samples were magnetized in a direct current (DC) field of 1 T with an MM PM 10 pulse magnetizer. SIRM was measured with an AGICO JR6 magnetometer and the resulting values normalized according to total leaf area (TLA (m 2 )). The remanence of unwashed leaves was recorded as SIRM u (×10 −6 A), which represents the total remanence (including leaf-deposited and encapsulated particles, and the leaves themselves). The remanence of washed samples was recorded as SIRM w (×10 −6 A), which represents the magnetic signal of leaf-encapsulated particles and the leaves themselves. In general, the SIRM w of the leaves was negligible. Consequently, the difference of SIRM uw (SIRM u-w = SIRM u − SIRM w ) between SIRMu and SIRMw is assumed to represent the magnetic signal of leaf-deposited particles (Hofman et al. 2014). The SIRM was measured in the Key Laboratory of Western China's Environmental Systems (Ministry of Education), Lanzhou University, China.

Mass of leaf-deposited particles
The mass of leaf-deposited particles was measured within 1 day after sampling, by the elution-filtration method, as follows: After SIRM u measurement, the leaves were removed from the pots and the cling film that were rinsed to make sure the adhered particles were removed, and then were placed in a beaker with 200 ml of distilled water; they were then washed with a non-depilatory soft brush for~1 min to ensure that particles on the surface of the leaves were transferred to the water. And then the beaker was placed into an ultrasonic cleaner (KQ-500DE) for~2-3 min to further ensure that leafdeposited particles were removed, and the leaves were then Fig. 1 The study area of Lanzhou City and sampling locations. The inset map shows the location of Lanzhou in China. The dots indicate the sampling points in the considered functional areas carefully removed with tweezers. The suspension was then agitated with the ultrasonic cleaner (KQ-500DE) for 10-20 min to make it homogeneous, and then, were sieved using a nylon sieve (1 mm) to remove large impurities and then were filtered using quartz filters (Whatman, filter size: 0.2 μm), and dried and weighed (W 1 ) using an electronic balance (PT-104/ 35S, accuracy of 0.01 mg). The filters with adhering particles were then oven-dried at 80°C for 24 h and reweighed (W 2 ). The mass of the particles deposited on the surface of the leaves was calculated as W 2 − W 1 .
After the washing, leaves were wiped using a kitchen towel, and the leaf area (LA (cm 2 )) of each leaf was then measured using the following procedure. First, leaves were scanned using a scanner (CanoScan LiDE120) and coordinate paper, and the Image J image processing software (https://imagej.en. softonic.com/) was then used to calculate LA. The particle retention capacity was defined as the weight of particles deposited per unit leaf area (g·m −2 ), which was calculated as: where W 2 = total mass of filter and particles, W 1 = mass of filter, and TLA = total leaf area of each subsample (m 2 ) which is the summed area of all individual values of LA.

SEM observations of adaxial and abaxial leaf surface morphology
In order to explore the relationship between the adaxial and abaxial leaf surface characteristics and particle retention capacity of different plant species, leaves were scanned using a field emission scanning electron microscope (SEM) (Apreo S, Thermo Fisher SCIENTIFIC, USA). Two pieces of rounded leaf near the center of the laminae were cut, 0.7 cm in diameter, and the adaxial surface was marked using a marker. The cut samples were then dried using a vacuum freeze-drier for 48 h to ensure that the original shape of the stomata on the leaf surfaces was maintained. The samples were then placed on a metal slide with double-sided tape and coated with a layer of gold using a high vacuum sputter coater to increase electrical conductivity and to improve optical transmission. For each sample, the images were scanned at different magnification levels (×250, ×500, ×1000).

Data analysis
Kolmogorov-Smirnov test was used to verify the normality of data. And then a nonparametric test which does not assume that data follow a specific distribution was performed to determine whether there were significant differences between the mass of leaf-deposited particles, SIRM u , SIRM w , and SIRM u-w within different functional areas (traffic areas, residential areas, and parks), life forms (trees, shrubs, and lianas), and life habits (evergreen and deciduous). The Kruskal-Wallis one-way ANOVA (k-sample) was used for multiple comparisons. In order to aid the election of plant species with maximum dust retention ability for urban greening management, hierarchical cluster analysis was used to classify plant species, and the clustering method is "between-groups linkage," with the mass of leaf-deposited particles as the variable. Data analysis was performed using SPSS (v.22.0, SPSS Inc., Chicago, IL, USA) and Origin (v.2018, OriginLab, USA).

Mass of leaf-deposited particles for different plant types
The mass of leaf-deposited particles among different plant types, growth habits, and at different heights are compared in Fig. 2 and SI Table 2. There are significant differences between the seven distinguished plant types (p < 0.0001). Across all plant types, the mass of particles deposited on the evergreen shrubs (0.5 m-shrubs-E and 1.5 m-shrubs-E) are significantly higher than for the other plant types (p < 0.05), being 2.47 and 2.17 times higher than that for 1.5 m-trees-D which has the lowest mass. In addition, within the evergreen shrubs, the mass of particles of 0.5 m-shrubs-E is higher than that for 1.5 m-shrubs-E, but the difference is not statistically significant (p > 0.05). In addition, there are substantial differences in the mass of leaf-deposited particles within the same plant types (Fig. 3). Within plant type 0.5 m-shrubs-D, the mass for Prunus×cistena is 10 times higher than that for Berberis thunbergii (p < 0.0001). Within the type 0.5 m-shrubs-E, Buxus sinica had the highest mass of leaf-deposited particles of all 23 plant species examined in the study, with mass much higher than those for the three other species in the plant type (p < 0.05). Within this plant type, the species can be ordered as follows: B u x u s s i n i c a > B u x u s m e g i s t o p h y l l a > Ligustrum×vicaryi > Rosa chinensis. Within plant type 1.5 m-shrubs-D, the mass of leaf-deposited particles decreases in the sequence of Syzygium aromaticum > Amygdalus triloba > Amygdalus persica > Lonicera maackii; however, there are no significant differences in mass of leaf-deposited particles among the four species. Within plant type 1.5 m-trees-D, the mass of leafdeposited particles decreases in the sequence of Malus×micromalus Makino > Cerasus yedoensis > Sophora japonica f. pendula; however, there were no statistically significant differences in mass of leafdeposited particles among the four plant species (p > 0.05). Within the plant type 2.5 m-trees-D, the mass of leaf-deposited particles decreases in the sequence of Robinia pseudoacacia > Fraxinus chinensis > Ulmus pumila > Sophora japonica > Salix matsudana > Ailanthus altissima > Ginkgo biloba > Salix babylonica. The values for Robinia pseudoacacia, Fraxinus chinensis, Ulmus pumila, and Sophora japonica are significantly higher than those for Salix babylonica (p < 0.05), although the differences within the plant types are not statistically significant (p > 0.05).

Mass of leaf-deposited particles between different functional areas
The concentration of leaf-deposited particles of each species varies between different functional areas (traffic areas, residential areas, and parks) (Fig. 4). Across all plant types, the mass of leaf-deposited particles in traffic areas is higher than in residential areas and parks. However, there are no statistically significant differences between the three areas for 0.5 mliana-D, 0.5 m-shrub-D, 1.5 m-shrub-E, and 1.5 m-tree-D (p > 0.05). For the three other plant types, the mass of leafdeposited particles in traffic areas are significantly higher than those from residential areas and parks (p < 0.0001), while there are no significant differences for those from residential areas and parks (p > 0.05).

Differences in leaf magnetism between different plant types
The SIRM values of the unwashed leaves (SIRM u ) are significantly different (p < 0.0001) among the seven considered plant types, which is similar to the results for the mass of leaf-deposited particles. In detail, the SIRM u of the 0.5 mshrub-E type is significantly higher (p < 0.0001) than that for the other plant types, and the lowest values are in the 1.5 m-tree-D type (Fig. 5). The SIRM w of 0.5 m-shrub-E type is significantly higher than 0.5 m-shrubs-D type (p < 0.0001), 1.5 m-shrubs-D, and 1.5 m-trees-D (p < 0.05). The variations in the SIRM of leaf-deposited particles (SIRM u-w ) among the seven plant types are consistent with the results for SIRM u . In addition, both SIRM u and SIRM u-w are higher for evergreen shrub species at the heights of 0.5 m and 1.5 m (p < 0.05). Similar to the particle mass of leaf-deposited particle results, the values of SIRM u , SIRM u-w , and SIRM w among plant species differ even within the same plant species. In detail, among the 23 plant species, the highest values of SIRM u and SIRM u-w occur for Buxus sinica and the lowest values for Berberis thunbergii (Fig. 6 a, c). Among the seven plant types, there is a near perfect correlation between the SIRM u and SIRM u-w results and mass of leaf-deposited particles (Table 1).

Differences in leaf magnetism between different functional areas
The variations in SIRM u , SIRM u-w , and SIRM w of each plant type in different functional areas are consistent with the particle weight results. The values for areas of high traffic density are consistently higher than those for residential areas and parks (Fig. 7), and the differences are statistically significant (p < 0.05) (SI Table 3), especially for 0.5 m-shrubs-E, 1.5 m-shrubs-D, 1.5 mtrees-D, and 2.5 m-trees-D. However, no statistically significant differences (p > 0.05) are observed between parks and residential areas across all plant types (SI Table 4). In addition, the variations of SIRM u , SIRM uw , and SIRM w are mutually consistent for the seven plant types (Fig. 7).

Ratios of SIRM u-w /SIRM u and SIRM w /SIRM u
For all plant species, the mean SIRM u-w /SIRM u ratio is 68%, range from 47 to 89%. The value varies in plant species, and the low values are found for Ailanthus altissima, Salix babylonica, Robinia pseudoacacia, Sophora japonica, Salix matsudana, Ginkgo biloba, and Ulmus pumila in plant type 2.5 m-tree-D, and Malus×micromalus Makino, Cerasus yedoensis, and Sophora japonica f. pendula in plant type 1.5 m-tree-D (SI Fig. 1). The SIRM u-w /SIRM u values are significantly lower (p < 0.0001) in parks (63.60%) and residential areas (63.53%) than in traffic areas (70.79%). Correlations between leaf magnetism and the mass of leaf-deposited particles Significant correlations were found between the mass of leafdeposited particles and SIRM u and SIRM u-w (r = 0.79~0.91) in parks, residential areas, and traffic areas, as shown in Table 2, with stronger correlations between the mass of leafdeposited particles and SIRM u-w than for SIRM u . In addition, for SIRM u-w , the correlation coefficient for traffic areas (r = 0.91) is higher than for residential areas (r = 0.87) and parks (r = 0.87). Nonsignificant correlations were found between the mass of leaf-deposited particles and SIRM w (r = 0.18~0.43); however, finite SIRM w values can still be observed for washed leaves.

Leaf surface morphology and distribution of particles
The average concentration of particles of 0.5 m-shrubs-E is greater than that of 1.5 m-shrubs-E; however, this is not the case for every species within the plant species, which may be caused by differences in leaf morphological characteristics. Buxus sinica at the height of 0.5 m has the highest particle retention capacity (SI Fig. 2a). The abaxial surface of this species has protrusive stomata with a large opening and deep surrounding grooves where particles are captured. In addition, particles were also observed in the folds between stomata (SI Fig. 2b). For Buxus megistophylla, the particles are heterogeneously existed on both adaxial and abaxial surfaces (SI Fig.  2c, d). The area around the stomata comprises a sink area on the abaxial surface where particles cluster together and agglomerations are also found (SI Fig. 2d). The structure of the abaxial surface of Ligustrum×vicaryi is complex because of dense protruding strips where sporadic fine particles were found (SI Fig. 2f). For Rosa chinensis, the adaxial and abaxial surfaces are relatively smooth, where only a few small particles were observed on the adaxial surface (SI Fig. 2g), and large particles (with a diameter about 20 μm) were sporadically scattered around the stomata (SI Fig. 2h). Many particles were observed in the dense, deep stripe-like grooves on both adaxial and abaxial surfaces of Prunus cerasifera (SI Fig. 3a, b); however, particles were largely absent from the areas in between the stomata.
Four deciduous plant species, those at the heights of 0.5-1.5 m, had a slightly greater particle retention capacity: e.g., the liana species Parthenocissus quinquefolia at 0.5 m, the adaxial surface of which has numerous large particles and deep stripe-like grooves where a lot of particles were observed (SI Table 1 Correlations between SIRM u , SIRM u-w , and the mass of leaf-deposited particles of seven plant types. S u SIRM of unwashed leaves, SIRM u-w the difference between SIRM u and SIRM w , M mass of leaf-deposited particles, D deciduous (n = 361), E = evergreen (n = 240). a represents p < 0.0001, and b represents p < 0.05  Fig. 5a). The adaxial surfaces of the shrub species Amygdalus triloba and Syzygium aromaticum at the height of 1.5 m have protrusions with deep recesses, where large and fine particles were observed (SI Fig. 6a-d); however, the protrusions in Syzygium aromaticum are relatively smooth and only a few particles were observed (SI Fig. 3c). As is evident from SI Fig. 5 and SI Fig. 6, although stomata, folds, protrusions, and tiny grooves are also present on the abaxial surfaces of the abovementioned seven species of deciduous plants, few particles were found on the surfaces.  In the case of deciduous tree species (1.5 m-trees-D and 2.5 m-trees-D) , the particles are also trapped on surfaces (SI Fig. 9c) and the cavity left by the breaking of the leaf tissue.

Comparison of dust retention capacity of different plant species
There were significant differences in the mass of leafdeposited particles of the 23 plant species analyzed in this study (p < 0.0001). For example, the mass of leaf-deposited particles Buxus sinica, with the highest mass, was 31.7 times higher than that for Berberis thunbergii which had the lowest weight (p < 0.0001). The results of hierarchical cluster analysis (Fig. 8a) show that final cluster members were divided into four clusters, with the dust retention capacity decreasing in the following sequence: I (median: 4.84 g·m −2 , n = 70) > II (median: 1.70 g·m −2 , n = 194) > III (median: 0.94 g·m −2 , n = 280) > IV (median: 0.38 g·m −2 , n = 57) (p < 0.0001) (Fig. 8b).

Differences in leaf-deposited particles between plant species
The results indicate significant differences in particle retention capacity among the studied plant species, which relate to differences in life form and growth habits. As shown in Fig. 3, the mass of leaf-deposited particles of the evergreen plant species, except for Rosa chinensis, are higher than for deciduous plant species at the heights of both 0.5 m (Prunus×cistena and Berberis thunbergii) and 1.5 m (Amygdalus persica, Syzygium aromaticum, Lonicera maackii, and Amygdalus triloba), indicating evergreen plant species have higher dust retention capacity than deciduous at both heights in the present study. Consequently, those evergreen plant species with high dust retention capacity are good choices for intercepting particles from the road and mitigating the air pollution at the shorter layer of green infrastructure. Simultaneously, other researchers have also noted that evergreen plants have a higher particle retention capacity than deciduous plants, because of their longer exposure times (Lehndorff et al. 2006) in winter time when the leaves of deciduous plants are absent.
In addition to the difference between evergreen and deciduous trees, it is notable that the mass of leaf-deposited particles for most of the shrub species were higher than those of trees. This phenomenon has been interpreted as a result of height where the leaves are collected; in urban environments, particle density is higher than at larger heights, so these shrubs are exposed to higher concentrations. The effects of height can also be between these species. Dzierzanowski et al. (2011) found that lower-growing shrubs such as Spiraea japonica had a greater ability to capture PM than trees, based on a survey of eight plant species. It is possible that lowergrowing plants are more susceptible to trapping particles derived from splashed soil than is the case for trees (Saebø et al. 2012). For Hedera helix, collected from different heights on a living wall in Hanover, there was a high degree of variability of the same sizes of PM on leaf surfaces (He et al. 2020). Large PM (diameter >10 μm) decreased substantially with increasing height, from 0.5 to over 2 m; however, for coarse (2.5-10 μm) and fine particles (1-2.5 μm), the variations were slight. In addition, the effect of sampling height on roadside plant species was reported among different plant species in different areas. For example, Hofman et al. (2013) found the SIRM of the leaves of Platanus×acerifolia decreased with increasing sampling height in a street canyon in Ghent (Belgium). Chen et al. (2019) found that more particles were retained by the leaves of Juniperus formosana below the height of 3.4 m than above that height in an open-road environment in Lanzhou, and the SIRM of leaf-deposited particles was lower at the heights of 3.6-5.4 m than below 3.4 m, possibly because the leaves at lower position are more influenced by dust suspension caused by vehicle traffic or other human activities. In addition, large particles are less likely to be transported sufficiently high above the ground surface to be deposited on leaves. However, in another study, no significant differences were observed in the particle amount at the heights of 0.25 m, 1.5 m, and 2.5 m for Hedera helix on living walls near a traffic road in the Netherlands (Ottelé et al. 2010).
In addition to the foregoing reasons, leaf characteristics are also an important factor affecting the quantity of leafdeposited particles on different plant species. In this study, the characteristics of the adaxial and abaxial leaf surfaces and the distribution of particles on these surfaces were observed by SEM. These results indicated that evergreen leaves with more complex structures and rougher surfaces capture more particles, which was also found in the common garden experiment of Muhammad et al. (2019) on 96 tree, shrub, and climber species. As shown in SI Figs. 1-8, for all plant species, leaves with dense protrusions or deep recesses in the protrusions and with stripe-like deep grooves appear to be more efficient at accumulating particles than leaves with simpler surface characteristics. The deep grooves between the protruding guard cells and blades result in a rough surface which promotes particle retention. There was no evidence of a relationship (r = 0.08, p = 0.71) between stomatal density and particle retention capacity for 23 plant species in this study, similar to the results of Muhammad et al. (2019) and Shao et al. (2019). The reason why this has not been tested more widely is that not all tree species retain large amounts of PM on the abaxial surface of leaves-where the stomata are for most species present and the leaf surface structure of different tree species is significantly different, making it difficult to determine the major factor(s) influencing particle retention capacity.
For all plant species in the present study, no relationship (p > 0.05) between leaf area and the mass of leaf-deposited particles (and magnetic properties) was observed, which was also the case in the study of Saebø et al. (2012).
Across all of the plant species, SIRM w contributed~11-54% of SIRM u , which indicates that magnetic particles are also captured by leaf wax, and the results are in line with the research of Hofman et al. (2014), who revealed that particles contribute 38% of the leaf SIRM signal. Muhammad et al. (2020) also indicated that washed leaf SIRM can be a good indicator to determine the effectiveness of a plant species in particle immobilization on a survey of 96 plant species in Antwerp, Belgium. In general, the waxy layer promotes the accumulation of particles (Burkhardt 2010;Jouraeva et al. 2002;Kaupp et al. 2002;Sabin et al. 2006), but leaf surface with waxes showed degradation, such as natural aging and weathering. Those changes of structure relating to waxes influence the deposition of particles onto leaf surface (Neinhuis and Barthlott 1998;Watanabe 2015). However, the amount of PM deposited on the leaf surface and trapped within the waxy layer is debated. For instance, Xu et al. (2018) reported that the proportions of particles deposited on the leaf surface and in the wax of 17 species in Beijing were 65% and 35% PM, respectively. The results of Saebø et al. (2012) showed that these proportions differed significantly between species: e.g., Betula pendula accumulated 82.6% of PM in the wax, while the respective values for Fagus sylvatica and Stephanandra incisa were 25% and 28%, which were the most effective species in terms of PM deposition. With regard to the size of trapped particles, epicuticular wax may be more effective in capturing fine PM (1-2.5 μm) (He et al. 2020;Viecco et al. 2018), andHe et al. (2020) reported that almost 65% of the PM in leaf wax consisted of fine PM (1-2.5 μm). The above results show that the capture effect of wax on PM is interspecific dependence that might be caused by wax structure which has influence on the leaf wettability that relate to the deposition of particles on the leaf surface (Muhammad et al. 2020).
According to the SEM results, for most of the plant species, particles are captured on the adaxial surfaces with more complex structures, such as deep stripe-like grooves, protrusions, and the cavity left by the breaking of leaf tissue, and consequently high stomatal density, which mainly occurs on the abaxial surfaces, may not be a decisive factor. For a few of plant species, e.g., Buxus sinica, Buxus megistophylla, and Prunus cerasifera, particles also can be found on the abaxial surfaces. In conclusion, leaf characteristic is a comprehensive factor that has influence on the deposition of particles, and researches indicated that leaf size, wax (Jouraeva et al. 2002;Muhammad et al. 2019), drop contact angle (Muhammad et al. 2019;Wang et al. 2013), and hair (Saebø et al. 2012;Muhammad et al. 2019) can affect particle retention capacity.
Influence of pollution sources on the leaf-deposited particles and leaf magnetism Significant differences in the mass of leaf-deposited particles and leaf magnetism in different functional areas were found. The concentration of magnetic minerals was mainly determined by the environmental conditions where the plant was located, and that more magnetic minerals were present within and on leaves in traffic areas than in residential areas and parks. By comparing the magnetic characteristics of unwashed leaves, leaf-deposited particles, and washed leaves in the traffic areas with those in parks and residential areas, it was found that the variations in amount of magnetic material retained on the surface and inside of the leaves were significantly related to the intensity of human activities, especially road traffic. Previous researches indicated that magnetism of leaf near the road was affected by magnetic particles produced by both vehicular fuel combustion (Harrison et al. 1997) and the abrasion of vehicle brakes and tires (Pant and Harrison 2013), which were generally considered to be important sources of PM pollution in urban area. Traffic-related PM usually contains abundant trace metals (e.g., Pb, Zn, Cu) that are associated with Fe particles (Baldacchini et al. 2017), which contribute to the leaf magnetism.
Since the sampling points in parks and residential areas are relatively distant from roads (>1 km), the mass of leafdeposited particles in these areas was only~40% of that in traffic areas; in addition, the values of SIRM u , SIRM u-w , and SIRM w were also lower than in the traffic areas (Fig. 7). And the more significant correlations between the mass of leafdeposited particles and SIRM u and SIRM u-w in traffic areas (r = 0.84 and 0.91, respectively, p < 0.0001) than those in parks (r = 0.84 and 0.87, respectively, p < 0.0001) and residential areas (r = 0.79 and 0.87, respectively, p < 0.0001) ( Table 1) were observed in the present study. The results indicate that leaves collected from different functional areas are both affected by the differences in local sources of atmospheric PM pollution and the distance to the pollution sources. The results were in line with previous studies; for instance, increased amounts of magnetic particles were found to be deposited on the leaves of roadside birch trees (Hofman et al. 2017;Maher et al. 2013). In addition, clear evidence has been found that pollutants from road traffic decrease with increasing distance from the road (Kardel et al. 2012;Zhang et al. 2006).

Evaluation of the applicability of biomagnetism monitoring in arid and semi-arid areas
The occurrence of atmospheric pollution on the leaf surfaces was confirmed by the SIRM u and SIRM u-w values, as illustrated in Fig. 9. The correlations between the mass of leafdeposited particles and SIRM u and SIRM u-w values decrease in the following sequence, with the correlation coefficients in parentheses: I > II = III > IV. These results indicate that both Fig. 9 Scatter plots of the mass of leaf-deposited particles and SIRM u and SIRM u-w for the four clusters of plant species. SIRM u = SIRM of unwashed plant leaves, SIRM w = SIRM of washed plant leaves, SIRM u-w = the difference between SIRM u and SIRM w . a-d Clusters I-IV (n = 70, 194, 250, and 87), respectively. The solid lines represent linear regression lines fitted to all values, "solid" for SIRM u , "dashed" for SIRM u-w SIRM u and SIRM u-w are good tools for PM monitoring, which have also been demonstrated in previous studies (Hofman et al. 2017); however, the effectiveness of the method varies between different plants. The results of this study have demonstrated that plant species with a higher particle retention capacity are more suitable for monitoring particulate pollution; however, these plant species with great particle trapping capacity are mainly the shorter species which are potentially more vulnerable to road traffic-derived pollutants which may result in leaf damage (Vinit-Dunand et al. 2002), for example by inhibiting photosynthesis and growth and causing leaf senescence and stomatal damage, and more works should be done about which plant species suit for being planted in the highly polluted areas or which plant species can be used as indicators for degree of urban air pollution (Du et al. 2007). Therefore, the results provide a useful reference for urban greening management. From the point of view of mitigating particulate pollution, plant species Buxus sinica, Buxus megistophylla, Prunus×cistena, Prunus cerasifera, Ligustrum×vicaryi, and Robinia pseudoacacia are good choices in arid and semi-arid urban like Lanzhou, which is heavily polluted by PM. Simultaneously, urban greening managers should program the frequency of cleaning these roadside plants reasonably to restore their dust retention ability and reduce the damage of leaf surface particles to leaves.

Conclusions
We have characterized the particle retention capacity and degree of magnetic contamination of 23 plant species (lianas, shrubs, and trees) in areas of varying land use in Lanzhou City. In urban environment, the mass of leaf-deposited particles and leaf magnetism are higher in traffic areas than those in parks and residential areas, indicating that pollution sources had significant influence on leaf-deposited particles. The leaves of evergreen shrub species at heights of 0.5 m and 1.5 m above the ground were found to have the greatest particle retention capacity. Leaves with relatively complex surfaces with features such as deep grooves and protrusions had a significantly higher particle retention capacity. The amount of particles retained by the leaves is the sum of the particles on the adaxial and abaxial surfaces; however, for most of the plant species, the rough adaxial surfaces are the main area of accumulation of particles, and consequently high stomatal density, which mainly occurs on the abaxial surfaces, may not be a decisive factor. Significant differences in the magnetic contamination of washed and unwashed leaves were observed, indicating that the cleaning process can effectively remove parts of particles deposited on the leaf surfaces. Significant correlations were found between the mass of particles deposited on leaves and SIRM u-w for plant species with different particle retention capacities, indicating that such species with high dust retention capacity are more suitable for biomagnetic monitoring; e.g., from the point of view of mitigating particulate pollution, those plant species, for instance, Buxus sinica, Buxus megistophylla, Prunus×cistena, Prunus cerasifera, Ligustrum×vicaryi, and Robinia pseudoacacia, are good choices for urban greening vegetation in arid and semi-arid urban in all the considered species, because of their high dust retention capacity. In addition, in order to improve air quality, different combinations of plant types (trees, shrubs, climbers, and herbs) should be explored, together with esthetics aspects, in order to formulate an optimum planting and urban management policy in Lanzhou.