Sampling site

This study was carried out at EAFIT University (Fig. 1), located in the south of Medellín, Colombia, which is the second-most populous city in the country. This city is the nucleus of the Aburra valley, an elongated depression with a total length of 60 km. It ranges in width between 3 to 10 km and is surrounded by mountains with elevation between 1300 and 1750 m asl (^{Hermelin 2007}). The main sources of atmospheric PM are industrial activities and high vehicle density. Air quality conditions diminish severely around March-April and October-November, the time of arrival of the intertropical convergence zone (^{Lopez-Restrepo et al. 2020}).

Fig. 1 Sample site. A) Medellín city B) Aburra’s Valley C) EAFIT university campus N 6º 11’ 57.77’’, W -75º 34’ 41.59’’. 

EAFIT University covers an area of 148 339 m² (^{EAFIT 2023}) and it is located between two main avenues characterized for heavy traffic. The area has an annual mean temperature of 22 ºC and annual mean rainfall of 1750 mm (^{Baca-Cabrera 2016}).

Species selection and collection

A total of 120 leaves were collected from six herbaceous plant species, belonging to three families, namely Marantaceae -Calathea rufibarba Fenzl and Calathea zebrina (Sims) Lindl, small herbs with bushy growth- Heliconiaceae -Heliconia psittacorum L.f., a medium size herb, and Heliconia rostrata Ruiz & Pav., a large herb- and Araceae -Philodendron sp. Schott and Dieffenbachia sp. Schott, both are large herbs. These 120 leaves correspond to 20 leaves collected from each species in two different periods, collecting 10 different individuals per species in each period. To guarantee that all leaves were in similar ontogenetic stages, and thus similar air-exposure time, the third leaf from the apex to the base was collected from each specimen. Furthermore, a photo from each plant was taken to assess the angle formed between the petiole and the main axis, measured using the ImageJ software (^{Schneider et al. 2012}). Immediately after collection, leaves were placed in hermetic plastic bags, such that PM loss was minimized during manipulation.

Samples were collected in two periods, March-April and October-December of 2017 that initially corresponded to rainy and dry seasons, respectively. However, due to climate abnormalities, it was not possible to obtain samples exposed to the dry season. Therefore, from now on, the “rainy and dry period” will be referred to as the “first and second period.” In the first period, C. rufibarba and C. zebrina were collected on March 18th, Dieffenbachia sp. and H. psittacorum on April 10th, and H. rostrata and Philodendron sp. on April 11th. The previous day before each of the three collection periods, it rained (Table SI). In the second period, C. rufibarba and C. zebrina were collected on October 24th, while H. psittacorum and H. rostrata on November 11th. The previous day before these two dates it rained. Philodendron sp. was collected on November 27th with three days of no-rain before sampling. Finally, Dieffenbachia sp. was collected on December 12th for which we do not have precipitation data (Table SI). The average monthly precipitation is reported in table SI. Climatic data were obtained from the Olaya Herrera climatic station, which is the closest station to EAFIT University (^{IDEAM 2018}).

Leaf-wash, PM weight, and trait measure

To determine PM₁₀ and PM_2.5 accumulated in leaves, we adapted the method described in ^{Dzierżanowski et al. (2011)}. The leaf surface was washed with distilled water and passed through a 10 µm mesh filter and then through a 2.5 µm mesh filter -both filters by Whatman. These filters were previously weighted on a TX323L analytical balance (Shimadzu, resolution 0.001g). After all the water dripped out, the filters were dried in an oven at 23 ºC until a constant weight was obtained. The filter final weight minus the filter weight before the leaves were washed was used to calculate the mass of PM₁₀ and PM_2.5. Leaf fresh mass weight and leaf dry mass weight were measured using an analytical balance -Shimadzu, resolution 0.001g-. To obtain leaf dry mass weight, the leaves were dried in an oven at 60 ºC until a constant weight was obtained. Further, leaf contours were scanned at 400 ppp resolution and analyzed using ImageJ software (^{Schneider et al. 2012}) to calculate leaf area in cm². Using fresh leaf area and dry weight, the specific leaf area (SLA) was also calculated (^{Perez-Harguindeguy et al. 2016}). To express PM deposition in µg/m², leaf area was transformed to m².

We also determined the presence of trichomes, epicuticles, and epidermal thickness. All of these procedures were performed with cross-section cuts observed under an optical microscope and measured using ImageJ software (^{Schneider et al. 2012}). The adaxial leaf surfaces of each species were photographed under an environmental scanning electron microscope (SEM) -Phenon G2Pro- operating in low-vacuum mode -8 kV-, to determine the relationships between dermal tissue, morphological traits, and PM deposition.

Statistical analysis

To evaluate differences in the deposition of PM10 and PM2.5 among species, we analyzed variance (ANOVA) using species as the main factor. We transformed the data using log10+1 to meet the normality criteria for these variables. Subsequently, we made pairwise comparisons using Tukey criteria. ANOVA was performed only with the data from the first sampling period because species were measured under different dates in the second sampling period due to logistic problems. Therefore, comparisons among species for this period were not possible. However, we performed a t-test to evaluate differences in the PM deposited between each species’ first and second sampling periods. With the aim to understand how leaf traits were associated with PM deposition we performed Pearson correlations. A Kruskal Wallis test was performed to find significant differences among species traits. All statistical analyses were performed using the R statistical package 3.6.2 (^{RCT 2022}) using command aov, lsmeans, t.test, and cor.test.

PM deposition in different species

We observed that species showed differences in PM_2.5 and PM₁₀ deposition in both sampled periods (Table I). In the first period, there were significant differences between species in PM₁₀ and PM_2.5 deposition -F = 7.301, p-value < 0.005 and F = 19.47, p-value < 0.005, respectively- (Fig. 2). H. psittacorum, C. rufibarba and C. zebrina showed significantly high deposition of PM₁₀ and PM_2.5, while Dieffenbachia sp. and H. rostrata showed low deposition of PM₁₀. In the case of PM_2.5, H. psittacorum and C. zebrina showed significantly higher deposition than the other species. In the second period, although higher quantities of PM were deposited on the species compared with the first period, those differences were not statistically significant, as shown in table I. The variability in the results of both periods shows that deposition of PM in herbaceous plants is a complex phenomenon that requires constant monitoring to make better predictions.

Fig. 2 Results of PM₁₀ (top) and PM_2.5 (bottom) deposited for six herbaceous species in the first sample period. Calruf: Calathea rufibarba, Calzeb: Calathea zebrina, Helpsi: Heliconia psittacorum, Helros: Heliconia rostrata, Phisp: Philodendron sp. and Diesp: Dieffenbachia sp. measured in 2017-2018. Different letters represent significant differences between species. The error bars indicate the 95% confidence interval, the lower and upper ends of the box represent the 25th and 75th percentiles respectively, the line within the box represents the median, and outliers are represented as dots. 

Leaf traits associated with PM deposition

Leaf traits that could be associated with PM deposition significantly differed between species (Table II). In terms of leaf area, C. rufibarba, C. zebrina, and H. psittacorum had small leaf area compared with H. rostrata, Dieffenbachia sp., and Philodendron sp. which had big leaf areas. As expected, lowest fresh weight corresponds with species that had the smallest leaf area. SLA was significantly larger for C. rufibarba, Dieffenbachia sp. and C. zebrina compared with the other species. Values for leaf epidermis thickness were significantly high for C. rufibarba, C. zebrina, and H. psittacorum. The angles between leaf petioles and the main axis did not show significant differences among species. Besides, we observed that several species had special features on the leaf surface such as trichomes in C. rufibarba, epicuticles in H. psittacorum, H. rostrata and C. rufibarba, and concave epidermal cells in C. zebrina (Table II and Fig. 3).

Fig. 3 Scanning electron microscope micrographs of leaf epidermal surfaces of A) Calathea zebrina: arrangement of epidermal concave cells. B) Calathea rufibarba: trichomes and epicuticle. C) Heliconia psittacorum: smooth epidermal cells with irregular epicuticle deposition. 

Not all traits were correlated with PM deposition. Leaf area -Pearson coefficient: -0.46 and -0.48 to PM₁₀ and PM₂₅ respectively, table III- and SLA -Pearson coefficient: -0.20 and -0.30 to PM₁₀ and PM₂₅ respectively, table III- correlated with PM deposition, whereas leaf epidermis thickness, leaf fresh weight, and leaf angle did not correlate with PM (Table III). We found that species with large leaf areas such as H. rostrata and Dieffenbachia sp., did not retain high amounts of PM_2.5 and PM₁₀, suggesting that large leaf area is not an important trait in PM deposition (Fig. 4). Conversely, we observed that species with small leaf areas such as C. rufibarba, H. psittacorum and C. zebrina retain higher amounts of PM_2.5 and PM₁₀ (Fig. 4). We also found a weak but significant -p-value = 0.001- positive correlation between the deposition of PM_2.5 and SLA.

Fig. 4 Correlation between PM₁₀ (left) and PM_2.5 (right) and leaf areas. Calruf: Calathea rufibarba, Calzeb: Calathea zebrina, Helpsi: Heliconia psittacorum, Helros: Heliconia rostrata, Phisp: Philodendron sp. and Diesp: Dieffenbachia sp. 

Leaf traits associated with PM deposition

According to our results, there were three species that had high amounts of PM deposition. H. psittacorum was one of the species that showed the highest PM₁₀ deposition in the first sample period - PM₁₀ = 0.040 ± 0.026 µg/m² - and in the second sampling period - PM₁₀ = 0.158 ± 0.096 µg/m², PM_2.5 = 0.039 ± 0.030 µg/m². - (Table I and Fig. 2). It was followed by C. rufibarba in the first and second sample period - PM₁₀ = 0.037 ± 0.019 µg/m², PM_2.5 = 0.027 ± 0.015 µg/m² and PM₁₀, = 0.168 ± 0.057 and PM_2.5= 0.037 ± 0.022 µg/m², respectively. Finally, C. zebrina also deposited high PM₁₀ and PM_2.5 concentrations - PM₁₀ = 0.018 ± 0.007 µg/m² and PM_2.5= 0.008 ± 0.005 µg/m²- during the first period (Table I). Interestingly, H. psittacorum, C. rufibarba, and C. zebrina presented epidermal traits such as epicuticle deposition, trichomes, and concave epidermal cells (Fig. 3). These differences compared with the other three species -Heliconia rostrata, Philodendron sp., Dieffenbachia sp.- suggest that leaf surfaces vary among herbaceous species and that these variations are important in their role in the PM deposition.

We found that some epidermal traits, such as epicuticle deposition, trichomes, and concave epidermal cells, were associated with PM deposition. The above is supported by the study of ^{Sæbø et al. (2012)}, which showed that the amount of epicuticle in the leaves is directly proportional to the capacity to accumulate PM. ^{El-Khatib et al. (2011)} also found that species having wax rings retain the highest amounts of PM10 among the species studied. However, authors such as ^{Dzierżanowski et al. (2011)} state that the potential for particle accumulation lies in the chemical composition of the epicuticle rather than the amount of the epicuticle and the structure of the epicuticular layer. Additional studies are required to determine whether PM accumulation in H. psittacorum and C. rufibarba is due to the amount of epicuticle, chemical composition, and arrangement because this was beyond our scope.

Our results suggest that trichomes are an important trait that play a role in the deposition of PM. Findings obtained by ^{Kim et al. (2020)}, who assessed the effects of Tillandsia usneoides trichomes on PM deposition, found differences in the performance of the species PM₁₀ deposition when trichomes were removed compared to the species with the trichomes present, the latter depositing higher amounts of PM₁₀. Besides, ^{Kwak et al. (2019)} support the relevance of trichomes to PM deposition; they demonstrated that leaf surfaces with trichomes had enhanced PM deposition compared to smooth leaf surfaces. Additionally, ^{Shao et al. (2019)} showed that PM deposition was improved when trichomes were present and associated with other leaf microstructures, such as epicuticle, presence of small chambers, stomatal and leaf roughness.

Leaf shape seems to be an important trait related with PM deposition. In our study, the three species that retained the most PM had lanceolate- or ovate-shaped leaves (Table I and Table II). ^{Corada et al. (2020)} reported that lanceolate and ovate leaves facilitate PM deposition. In addition, ^{Leonard et al. (2016)} showed that species with lanceolate-shaped leaves can accumulate more PM than obovate- and elliptic-shaped leaves, which supports our results. This was also shown by ^{Weerakkody et al. (2018a)}, who studied PM captured by green wall plants and concluded that plants with needle-shaped leaves and small leaf areas accumulated more PM.

Finally, in our study, species with small leaf areas tend to deposit higher amounts of PM compared to species with large leaf areas -Dieffenbachia sp., H. rostrata, and Philodendron sp.-. This can be verified in our negative correlation between leaf areas and atmospheric PM deposition (Table III and Fig. 4). The above was also found by ^{Weerakkody et al. (2018b)}, who showed a significant negative relationship between leaf size and PM deposition; the smaller the leaf size, the greater the PM deposition. ^{Weerakkody et al. (2018b)} explained this by the edge effect generated by the larger leaf perimeter/surface area ratio of small leaves. Moreover, smaller leaves have a thinner resistance boundary layer, allowing more contact between air pollutants and the leaf surface (^{Murray 1979}, ^{Chen et al. 2017}). This could explain the low PM deposition of H. rostrata, which presented an irregular epicuticle on the leaf surface but had the second-largest leaf area among our species.

PM deposition capacities of ornamental herbaceous plants

Urban greening is an important strategy for the regulation of air pollutants. Herbaceous plants, as shown in this study, promote air pollutant deposition, and therefore offer the possibility of air quality improvement. This is because they can be placed very close to motor vehicle traffic, maximizing the capture of air pollution (^{Weber et al. 2014}, ^{Janhäll 2015}). In addition, herbaceous plants can complement trees by depositing resuspended or washed-off particles from their canopy (^{Weber et al. 2014}).

Low vegetation can be easily adapted to complex urban architectural designs by expanding the possibilities of urban design to vertical structures and green walls. The integration of high and low vegetation according to urban street canyons, traffic density, wind flow, and other local meteorological conditions could be an important strategy for air pollution mitigation policies. Nevertheless, to encourage the use of herbaceous vegetation for air quality improvement, more research is needed, such as those related to allergenic and biogenic volatile organic compounds.

In our study, the three herbaceous species that most retained PM -H. psittacorum, C. rufibarba, and C. zebrina- presented small leaves areas, with lanceolate or oval shapes and micromorphological epidermal traits like trichomes and epicuticle deposition compared to the other species. These characteristics influenced PM depositions. These species are potentially appropriate to be used in gardens, thus improving the air quality in urban environments. To date, this is the first study to focus on the PM deposition capacity of ornamental herbaceous species used in urban environments in Colombia.