*Article* **The Temporal Variation of the Microclimate and Human Thermal Comfort in Urban Wetland Parks: A Case Study of Xixi National Wetland Park, China**

**Zhiyong Zhang 1, Jianhua Dong 2, Qijiang He <sup>2</sup> and Bing Ye 1,\***


**Abstract:** As an important part of the ecological infrastructure in urban areas, urban wetland parks have the significant ecological function of relieving the discomfort of people during their outdoor activities. In recent years, the specific structures and ecosystem services of urban wetland parks have been investigated from different perspectives. However, the microclimate and human thermal comfort (HTC) of urban wetland parks have rarely been discussed. In particular, the changing trends of HTC in different seasons and times have not been effectively presented. Accordingly, in this research, a monitoring platform was established in Xixi National Wetland Park, China, to continually monitor its microclimate in the long term. Via a comparison with a control site in the downtown area of Hangzhou, China, the temporal variations of the microclimate and HTC in the urban wetland park are quantified, and suggestions for clothing are also provided. The results of this study demonstrate that urban wetland parks can mitigate the heat island effect and dry island effect in summer. In addition, urban wetland parks can provide ecological services at midday during winter to mitigate the cold island effect. More importantly, urban wetland parks are found to exhibit their best performance in improving HTC during the daytime of the hot season and the midday period of the cold season. Finally, the findings of this study suggest that citizens should take protective measures and enjoy their activities in the morning, evening, or at night, not at midday in hot weather. Moreover, extra layers are suggested to be worn before going to urban wetland parks at night in cold weather, and recreational activities involving accommodation are not recommended. These findings provide not only basic scientific data for the assessment of the management and ecological health value of Xixi National Wetland Park and other urban wetland parks with subtropical monsoon climates, but also a reference for visitor timing and clothing suggestions for recreational activities.

**Keywords:** microclimate; human thermal comfort; outdoor thermal environment; public health; ecological services

#### **1. Introduction**

Due to the complex social process of rapid urbanization, approximately half of the global population lives in urban areas, and this percentage is still increasing [1,2]. Undeniably, modernized fundamental infrastructure has provided great convenience to human life. Compared to a century ago, people living in urban areas work efficiently and create high economic value. However, due to rapid urbanization, factors such as the expansion of urban areas, the use of concrete and asphalt, and the loss of natural resources and space have led to increased and decreased temperatures in urban areas in summer and winter, respectively, as well as reduced humidity, compared to rural areas. These phenomena are respectively known as the urban heat (cold) island effect and the urban dry island effect [2]. Furthermore, in the context of global climate change, extreme weather events, including urban waterlogging, dry winds, and persistent heat and cold waves, are frequently observed

**Citation:** Zhang, Z.; Dong, J.; He, Q.; Ye, B. The Temporal Variation of the Microclimate and Human Thermal Comfort in Urban Wetland Parks: A Case Study of Xixi National Wetland Park, China. *Forests* **2021**, *12*, 1322. https://doi.org/10.3390/f12101322

Academic Editors: Manuel Esperon-Rodriguez and Tina Harrison

Received: 27 July 2021 Accepted: 24 September 2021 Published: 28 September 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

in urban areas [3,4]. This can be attributed to the spatial heterogeneity and fragmentation of the urban environment. Human-induced urban spaces cause damage to the essential functions of natural ecosystems, including substance circulation, energy exchange, and information transfer. Indeed, urban ecosystems cannot regulate their temperature independently, as the concrete ground is directly exposed to sunlight, and air conditioners and other electrical devices are widely used [5]. This results in the incapability of the urban ecosystem to effectively exert its ecological function of adjusting the microclimate, leading to the growing discomfort of people during their outdoor activities [6].

The ecological infrastructure plays an important role in maintaining the integrity of the urban ecological structure. Due to the COVID-19 pandemic, people are required to maintain social distance. In this case, there has been more awareness of the significance of natural outdoor space and fresh air. As one of the most important ecological infrastructures in urban areas, wetland parks provide various ecosystem services, including climate regulation, flood control, aquifer recharging, water purification, carbon sequestration, and they act as habitats for plants and animals [7–9]. In recent years, the specific structures and ecosystem services of wetland parks have been investigated from different perspectives [10–12]. The human thermal comfort (HTC) of the environment is also an important component of ecosystem services and plays a vital role in improving the physical and mental health of visitors in wetland parks. Unfortunately, the HTC of urban wetland parks has rarely been discussed.

HTC, which is a concept that reflects humans' perception of the thermal environment, has been widely employed to evaluate the comfort level of the body in response to the environment [13,14]. The traditional method of the measurement of HTC is to use a series of meteorological factors (e.g., temperature, humidity, wind speed, radiation, etc.) to establish a comprehensive index by which to quantify the bioclimatic conditions for human. Over the past century, more than 100 indices have been developed and used to assess HTC in combination with meteorological factors [15]. In recent years, several models, such as the urban canopy model, Urban Tethys-Chloris (UT&C), have been proposed to evaluate the urban thermal climate. UT&C is a combination of an urban canyon scheme and an ecohydrological model, and considers air and surface temperatures, air humidity, and soil moisture, as well as the urban energy and hydrological fluxes in the absence of snow [16–18]. These indexes and models all provide the basis for assessing the human thermal sensation of the environment. Previous studies of microclimates and HTC focused on urban outdoor spaces, college campuses, street green spaces, and city parks [19–23]. However, most of these studies were conducted in the summer, and the monitoring periods tended to be short (typically 3–7 d). Additionally, some studies did not consider the effects of relative humidity and wind speed on HTC, and continual monitoring in the long term was not sufficient, leading to the ineffective presentation of the changing trends of HTC in different seasons and at different times.

Therefore, based on the annual monitoring of the microclimate features and HTC of Xixi National Wetland Park, China in different seasons, this study determines the period of time in a day when people can obtain the highest level of HTC during recreational activities, and offers suggestions for appropriate clothing. The results provide a reference for the design and management of ecology therapy activities in wetland parks.

#### **2. Target Site and Methods**

#### *2.1. Target Site*

In this study, Xixi National Wetland Park in China was selected as the target site. Xixi National Wetland Park (30◦15.39 N–30◦16.96 N, 120◦03.16 E–120◦04.94 E) is 16 km from the downtown area of Hangzhou, China. As a typical urban wetland park and China's first national wetland park, Xixi National Wetland Park was established by the State Forestry and Grassland Administration (formerly the State Forestry Administration) of China in 2005 [24]. It has a total area of 11.5 km2, and over 50% of its surface is covered by water. The park has a typical humid subtropical monsoon climate with high temperatures and

precipitation in summer, and low precipitation in winter. The average temperature is about 16.2 ◦C, and the average precipitation is about 1400 mm per year. Xixi National Wetland Park consists of three areas: the wetland ecological protection and cultivation area (east), the wetland eco-tourism leisure area (center), and the wetland ecological landscape conservation area (west). As shown in Figure 1, the wetland park site (WPS) and the control site (CS) are in the ecological protection and cultivation area and the downtown area of Hangzhou, respectively. When the positioning monitoring platform used in this study was installed, the destruction of vegetation around the site was minimized. After installation, the original vegetation was reconstructed. Excluding the presence of protective measures, such as a guardrail installed around the site, there is no man-made shade or open space to prevent the modification of wind flow and the limiting of heat dissipation. The CS is not shaded by nearby buildings, and is not affected by wind channelling, downdraft effects, etc. Thus, the microclimate characteristics of the regions represented by the two platforms are objectively reflected.

**Figure 1.** The wetland park site and the control site in Hangzhou, China.

#### *2.2. Methods*

#### 2.2.1. Monitoring and Data Sources

At each site, a monitoring platform was set up to monitor the air temperature, relative humidity, and wind speed, as shown in Figure 1. The air temperature and relative humidity were detected by a sensor (Vaisala HMP155, Vaisala, Inc., Woburn, USA), and the wind speed was measured by an anemometer (RM Young Wind Sentry Set03002-1, Campbell Scientific Inc., Logan, USA) 24 h per day.

The data collector (CR1000, Campbell Scientific Inc., Logan, USA) recorded data every 15 min and transmitted it to a remotely controlled computer via the MA8-L GPRS terminal. The description of measuring devices is provided in Table 1. In this study, data collected from January–December 2016 were used.

**Table 1.** The description of measuring sensors.


#### 2.2.2. Assessment Indicators of HTC

Table 1 reports the common indices of HTC used internationally, each of which has its own applicable weather conditions. By comparing various indices, Blazejczyk et al. confirmed that each index has a high correlation with each other (the lowest *R*<sup>2</sup> coefficient was 93.1), regardless of the inclusion of a radiation factor in the equation. Moreover, it was pointed out that non-radiation equations are used more widely in some climatic conditions [15]. In other words, each index demonstrates high consistency in the assessment of HTC. Therefore, in the application of bioclimatic indices for the assessment of HTC, the climatic conditions of the research area and people's thermal perception habits should be fully considered during index selection.

Similar to the indices listed in Table 2, the indices reported in Table 3 are the most widely used in China. In comprehensive consideration of the scope of application of each index and the seasonal dressing habits of the Chinese people, the human thermal comfort index (HTCI) and clothing thickness index (CTI) were employed to evaluate the human perception of the environment. The HTCI was focused on evaluating the comfort level of the body. Additionally, the CTI can serve as an indicator of clothing.


**Table 2.** The combined environmental variables and applicable weather conditions of indices.

#### **Table 3.** The most widely used indices in China.



**Table 3.** *Cont.*

\* *T* is the temperature (◦C), *RH* is the relative humidity (%), and *V* is the wind speed (m/s).

(1) The HTCI, proposed by Lu et al. [40], is a synthesis of temperature, relative humidity, and wind speed, and is calculated by the following Equation (1) [41,42]. Compared to the thermal humidity index (which is only applicable to hot climate conditions), the HTCI can better reflect the outdoor satisfaction level of humans in the four seasons, and thus it is widely applied.

$$\text{HTCI} = 0.6 \times (|T - 24|) + 0.07 \times (|RH - 70|) + 0.5 \times (|V - 2|), \tag{1}$$

where *T* is the temperature (◦C), *RH* is the relative humidity (%), and *V* is the wind speed (m/s). The HTCI is negatively related to the body perception of comfort, as presented in Table 4.

**Table 4.** The relationship between the HTCI and human body perception [41,42].


(2) The CTI is also a synthesis of temperature, relative humidity, and wind speed, and is calculated by the following equations [41,43].

If *T* ≤ 18 ◦C and *RH* ≥ 60%,

$$\text{CTI} = \frac{\left[1 + \frac{0.4(RH - 60)}{100}\right] \times 0.61(33 - T)}{(1 - 0.01165V^2)}.\tag{2}$$

If 18 ◦C < *T* < 26 ◦C or *T* ≤ 18 ◦C, and *RH* < 60%, or, if *T* ≥ 26 ◦C and *RH* < 60%,

$$\text{CTI} = \frac{0.61(33 - T)}{(1 - 0.01165V^2)}.\tag{3}$$

If *T* ≥ 26 ◦C and *RH* ≥ 60%,

$$\text{CTI} = \frac{\left[1 - \frac{0.4(RH - 60)}{100}\right] \times 0.61(33 - T)}{(1 - 0.01165V^2)}.\tag{4}$$

In these equations, *T* is the temperature (◦C), *RH* is the relative humidity (%), and *V* is the wind speed (m/s). Table 5 presents the relationship between the CTI and the clothing suggestion, which was adapted from the work of Gu et al. [41] and Zhu et al. [43].


**Table 5.** Division of the CTI and clothing suggestion.

#### 2.2.3. Data Analysis

Herein, spring refers to the months of March, April, and May, summer includes June, July, and August, autumn includes September, October, and November, and winter includes December, January, and February. The winter data considered in this study were collected in December, January, and February 2016.

When the observation data were obtained from the data collector, a quality check was carried out first. After removing the abnormal data, statistical analysis was performed. The temperature, relative humidity, and wind speed at the same time each day of the year were averaged to obtain microclimatic data at that time, from which the daily variations are plotted. After that, the temperature, relative humidity and wind speed in each month and season were averaged arithmetically, i.e., the microclimatic data of the month and season were obtained respectively. Finally, the daily, monthly, and seasonal variations of HTCI and CTI were calculated according to Equations (1)–(4). The significance of arithmetic mean values for all data sets was assessed using Student's *t*-test. Error bars were calculated with the s.d. function. The differences at *p* < 0.05 were considered statistically significant by One-way ANOVA. According to the Student's *t*-test, characters in the figure represent statistically significant differences compared with control (\* *p* < 0.05, \*\* *p* < 0.01 and \*\*\* *p* < 0.001). Data analysis was conducted using SPSS 19.0 software (IBM® SPSS® software, Armonk, NY, USA), and the data were plotted by GraphPad Prism 8 software (GraphPad Software, Inc., San Diego, CA, USA).

#### **3. Results**

#### *3.1. Seasonal, Monthly, and Daily Variations of the Microclimate*

#### 3.1.1. Seasonal, Monthly, and Daily Variations of Temperature

The seasonal variations of average temperature in both the WPS and CS exhibited a greater range in summer (27.5 and 29.7 ◦C) than in autumn (18.9 and 20.4 ◦C), spring (16.6 and 18.1 ◦C), and winter (7.0 and 8.5 ◦C) (Figure 2a). The monthly variation of the average temperature could be described as a unimodal curve, and in July and August, the peak values were reached in both the WPS and CS (Figure 2b). Regarding the daily variation, the average temperature exhibited a trend from decreasing to increasing, reached the highest point between 12:00 and 16:00, and then gradually decreased (Figure 2c,d and Supplementary Figure S1a–j). The seasonal, monthly, and daily variations of the temperature followed the natural laws of cool temperatures in the winter, warm temperatures in the summer, high temperatures during the day, and low temperatures at night. Moreover, the average temperature of the WPS in each season and month was significantly lower than that of the CS, and the percentage of reduction was between 6.30% and 20.83%. However, in terms of the daily variation, some interesting findings were discovered. Taking February as an example (Figure 2c), the average temperature of the WPS was higher than that of the CS between 11:45 and 16:30. This also occurred in March, January, and December (Supplementary Figure S1a,b,j). In other months, the average temperature of the WPS was lower than that of the CS in the 24-h cycle (Supplementary Figure S1c–i). Compared to the urban environment, the ecological infrastructures in the wetland park can reform the wind pattern and limit heat dissipation at noon in the cold season, so that more heat can be stored in the environment. Figure 2d presents the daily variation of temperature in July.

**Figure 2.** The seasonal, monthly, and daily variations of the average temperature in the WPS and CS. (**a**) The seasonal variation of the average temperature; different lowercase letters of the same color indicate results of a one-way ANOVA followed by Tukey's test (*p* < 0.05), and error bars represent the standard deviation. (**b**) The monthly variation of the average temperature. (**c**) The daily variation of temperature in February. (**d**) The daily variation of temperature in July. Note: Significant differences between the means of WPS and CS were determined using Student's *t*-test (\* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001).

#### 3.1.2. Seasonal, Monthly, and Daily Variations of Relative Humidity

The average relative humidity in both the WPS and CS differed with various seasons. It peaked in autumn (85.45% and 75.03%), followed by summer (83.34% and 68.74%) and spring (79.30% and 67.97%), and the lowest average relative humidity occurred in winter (77.08% and 65.32%). The average relative humidity of the WPS presented significant differences between the four seasons, while that of the CS exhibited no significant difference between spring and summer and had a significant difference with other seasons (Figure 3a). Regarding the monthly variation, the relative humidity was at a relatively high level in all months, there were negligible changes excluding February and August (Figure 3b). The daily variation of the average relative humidity first increased and then decreased, reached the lowest point at 14:40–16:00, and then increased gradually. In February, the average relative humidity of the WPS and CS was relatively close, between 11:45 and 16:30, while the percentage increase of relative humidity in the WPS was less than 10% that in the CS (Figure 3c). This was also observed in January, March, and December (Supplementary Figure S2a,b,j). In other months, the percentage increase of relative humidity was generally above 10% most of the day and overnight (Supplementary Figure S1c–i). Figure 3d presents the daily variation of the relative humidity in August. The average relative humidity of the WPS and CS exhibited significant differences on monthly and annual scales. Moreover, the average relative humidity of the WPS in each season and month was significantly higher than that of the CS, and the percentage increase was between 12.77% and 24.77%.

**Figure 3.** The seasonal, monthly, and daily variations of the average relative humidity in the WPS and CS. (**a**) The seasonal variation of the average relative humidity; different lowercase letters of the same color indicate results of a one-way ANOVA followed by Tukey's test (*p* < 0.05), and error bars represent the standard deviation. (**b**) The monthly variation of the average relative humidity. (**c**) The daily variation of relative humidity in February. (**d**) The daily variation of relative humidity in August. Note: Significant differences between the means of WPS and CS were determined using Student's *t*-test (\*\*\* *p* < 0.001).

#### 3.1.3. Seasonal, Monthly, and Daily Variations of Wind Speed

The wind speed in both the WPS and CS exhibited seasonal variation, and the wind speed in the WPS was significantly lower than that in the CS. The highest wind speed in the WPS occurred in winter (0.11 m/s), followed by summer (0.08 m/s) and spring (0.07 m/s), and the lowest wind speed occurred in autumn (0.02 m/s). The wind speeds in spring and summer were not substantially different, but recorded a large difference compared with winter and autumn (*p* < 0.05). The highest wind speed in the CS occurred in summer (0.21 m/s), followed by spring (0.14 m/s), and the lowest wind speed occurred in winter and autumn (0.12 m/s), which demonstrated a significant variation as compared with spring and summer (*p* < 0.05) (Figure 4a).

The monthly wind speed in the WPS and CS displayed fluctuations. In general, the wind speed in the WPS was relatively high in December, January, February, March, and June, while that in the CS was relatively high in February, March, June, July, and August, and the dispersion of the wind speed value in the CS was relatively large (Figure 4b). The daily variation curve for each month reveals that the wind speed in both the WPS and CS was the highest between 14:00 and 16:00 with a single peak. The daily variation of the wind speed in the CS was large in each month, whereas that in the WPS exhibited small fluctuations between April and November (Supplementary Figure S3c–i) and large fluctuations in other months (Supplementary Figure S3a,b,j). Figure 4c,d presents the daily variation of wind speed in February and August.

**Figure 4.** The seasonal, monthly, and daily variations of the average wind speed in the WPS and CS. (**a**) The seasonal variation of the average wind speed; different lowercase letters of the same color indicate results of a one-way ANOVA followed by Tukey's test (*p* < 0.05), and error bars represent the standard deviation. (**b**) The monthly variations of the average wind speed. (**c**) The daily variation of wind speed in February. (**d**) The daily variation of wind speed in August. Note: Significant differences between the means of WPS and CS were determined using Student's *t*-test (\*\*\* *p* < 0.001).

#### *3.2. Seasonal, Monthly, and Daily Variations of the HTCI*

#### 3.2.1. Seasonal and Monthly Variations of the HTCI and Human Body Perception

As shown in Figure 5a, the HTCI in both the WPS and CS demonstrated certain seasonal variations. Apart from summer, when the HTCI in the WPS was significantly lower than that in the CS (*p* < 0.01), the HTCI in the WPS was higher than that in the CS in the other seasons (*p* < 0.001). From the seasonal perspective, significant variations of the HTCI in the WPS between all seasons was observed (*p* < 0.05). The largest variation occurred in winter (12.17), which could be classified as a very uncomfortable feeling. The lowest variation occurred in summer (4.34), which could be classified as a very comfortable feeling. The variations in spring (6.38) and autumn (5.49) had a middle rank, which could be classified as a comfortable feeling. In the CS, the largest variation occurred in winter (10.90), which could be classified as a very uncomfortable feeling, the lowest variation occurred in autumn (4.26), which could be classified as a very comfortable feeling, and the variations in spring (5.09) and summer (4.86) had a middle rank, which could be classified as a comfortable feeling. The difference between spring and summer was not substantial, but their difference with winter and autumn was significant (*p* < 0.05).

The monthly HTCI in the WPS and CS exhibited a "W"-shaped trend. The value gradually decreased from January to May, slowly increased from June to July, decreased again from August to September, and then eventually increased. Apart from July and August, when the HTCI in the WPS was significantly lower than that in the CS (*p* < 0.001), the HTCI values in the WPS in all other months were larger than those in the CS. The WPS would have given people a very uncomfortable feeling in December, January, and February, an uncomfortable feeling in March and November, a comfortable feeling in April, July, August, and October, and a very comfortable feeling in May, June, and September (Figure 5b).

**Figure 5.** The seasonal and monthly variations of the HTCI in the WPS and CS. (**a**) The seasonal variation of the HTCI; different lowercase letters of the same color indicate results of a one-way ANOVA followed by Tukey's test (*p* < 0.05), and error bars represent the standard deviation. (**b**) The monthly variations of the HTCI, and error bars represent the standard deviation. Note: Significant differences between the means of WPS and CS were determined using Student's *t*-test (\*\* *p* < 0.01, \*\*\* *p* < 0.001).

#### 3.2.2. Daily Variation of the HTCI

As shown in Figure 6, the HTCI displayed different trends of daily variation over the 12-month observation period.

In January, February, March, April, October, November, and December, the HTCI in both the WPS and CS exhibited an increasing trend, then a decreasing trend, and finally rebounded. Two relationships existed between the WPS and CS in terms of the magnitude of the HTCI. Take February as an example. The HTCI in the WPS was lower than that in the CS during the period of 9:45–17:00, while it was higher than that in the CS in other periods (Figure 6b). The same relationship occurred in January, March, April, and December. In contrast, in October and November, the daily HTCI values in the WPS were higher than those in the CS in all time periods.

In May, the HTCI in the WPS exhibited a trend of increasing, decreasing, then increasing again, whereas the HTCI in the CS exhibited a "W"-shaped trend. During 12:45–15:30, the HTCI in the WPS was lower than that in the CS, whereas it was higher than that in the CS in other periods (Figure 6e).

In June and September, the HTCI in the WPS exhibited a "W"-shaped trend, while the HTCI in the CS presented a trend of decreasing followed by increasing and reached its peak at around 14:00, after which it slowly decreased (Figure 6f,i). The HTCI in the WPS was lower than that in the CS during the daytime. For example, in June, the HTCI in the WPS was lower than that in the CS during 9:45–20:00, but was higher than that in the CS in other periods (Figure 6f).

In July and August, the HTCI in both the WPS and CS exhibited a trend of increasing then decreasing, and reached its peak at around 14:00, whereas the HTCI in the WPS was lower than that in the CS in almost all time periods (Figure 6g,h).

**Figure 6.** The daily variation of the HTCI between (**a**) January and (**l**) December in the WPS and CS.

#### 3.2.3. Daily Variation of the Human Body Perception

As shown in Figure 6, in January, February, and December, the WPS and CS generated very uncomfortable or uncomfortable feelings all day long (Figure 6a,b,l). In March, the CS generated a very uncomfortable or uncomfortable feeling all day long, while the WPS generated a comfortable feeling during 12:45–15:30 but a very uncomfortable or uncomfortable feeling in other periods (Figure 6c). In April, the WPS generated a comfortable or very comfortable feeling during 8:30–22:15, but an uncomfortable feeling in other periods, while the CS generated a comfortable or very comfortable feeling all day long (Figure 6d). In July, the WPS generated a comfortable or very comfortable feeling all day long, while the CS generated an uncomfortable feeling during 11:00–17:30 but a comfortable or very comfortable feeling in other periods (Figure 6g). In August, the WPS generated an uncomfortable feeling during 13:45–15:00, but a comfortable or very comfortable feeling in other periods, while the CS generated a very uncomfortable or uncomfortable feeling during 11:00–18:15, but a comfortable or very comfortable feeling in

other periods (Figure 6h). In November, the WPS generated a comfortable feeling during 11:45–16:30, but a very uncomfortable or uncomfortable feeling in other periods, while the CS generated a comfortable feeling during 10:15–20:30, but an uncomfortable feeling in other periods (Figure 6k). In May, June, September, and October, the WPS and CS both generated a comfortable or very comfortable feeling all day long (Figure 6e,f,i,j).

#### *3.3. Seasonal, Monthly, and Daily Variations of the CTI*

#### 3.3.1. Seasonal and Monthly Variations of the CTI and Clothing Suggestions

As shown in Figure 7a, the CTI values in the WPS were all higher than those in the CS in all four seasons (*p* < 0.01). The CTI in the WPS and CS exhibited significant seasonal variations (*p* < 0.05), with winter displaying the highest values (17.14 and 15.49). This belongs to level IV, so a sweater in addition to a thin coat or thin cotton coat is recommended for outdoor activities. Summer had the lowest variations (3.21 and 1.94). This belongs to level II, so a long-sleeve shirt is recommended for outdoor activities. Spring (10.50 and 9.32) and autumn (9.03 and 7.93) ranked in the middle. These belong to level III, so a shirt in addition to a jacket or suit is recommended for outdoor activities.

**Figure 7.** The seasonal and monthly variations of the CTI in the WPS and CS. (**a**) The seasonal variations of the CTI; different lowercase letters of the same color indicate results of a one-way ANOVA followed by Tukey's test (*p* < 0.05), and error bars represent the standard deviation. (**b**) The monthly variations of the CTI. Significant differences between the means of WPS and CS were determined using Student's *t*-test (\*\* *p* < 0.01, \*\*\* *p* < 0.001).

From the monthly perspective, the CTI in the WPS and CS demonstrated a decreasing and then increasing trend, with July having the lowest variation (−2.06 and 0.68). The CTI values in the WPS were all higher than those in the CS in each month (*p* < 0.01). In January, the CTI in the WPS belonged to level V, so a thick sweater in addition to a wool coat or down jacket or other winter clothing is recommended for outdoor activities. In the CS, the CTI belonged to level IV, so a sweater in addition to a thin coat or thin cotton coat is recommended for outdoor activities. In February, March, November, and December, the CTI values in the WPS and CS belonged to level IV, so a sweater in addition to a thin coat or thin cotton coat is recommended for outdoor activities. In April, May, and October, the CTI values in the WPS and CS belonged to level III, so a shirt in addition to a jacket or suit is recommended for outdoor activities. In June and September, the CTI values in the WPS and CS belonged to level II, and a long-sleeve shirt is recommended for outdoor activities. In July and August, the CTI values in the WPS belonged to level II, and a long-sleeve shirt is recommended for outdoor activities. In the CS, the CTI values belonged to level I, so short-sleeved summer clothing is recommended for outdoor activities (Figure 7b).

#### 3.3.2. Daily Variation of the CTI and Clothing Suggestions

As shown in Figure 8, the daily variations of the CTI in the WPS and CS in each month demonstrated an increasing and then decreasing trend, followed by a slowly increasing

trend. In February, March, and December, the CTI values in the WPS were larger than those in the CS in certain time periods (Figure 8b,c,l). Taking March for example, the CTI in the WPS was lower than that in the CS during the period of 10:15–15:30, whereas it was higher in other periods (Figure 8c). Apart from February, March, and December, the CTI values in the WPS were higher than those in the CS in all periods in all other months (Figure 8a,d–k).

**Figure 8.** The daily variation of the CTI between (**a**) January and (**l**) December in the WPS and CS.

Clothing suggestions are further discussed based on the daily variation of the CTI. In January, the daily variation of the CTI in both the WPS and CS was reduced from level V to level IV before increasing to level V (Figure 8a). In February, the daily variation of the CTI in the WPS reduced from level V to level IV before increasing to level V, while that in the CS stayed in level IV over the whole day (Figure 8b). In March, the daily variation of the CTI in both the WPS and CS was reduced from level IV to level III before increasing to level IV (Figure 8c). In April, the daily variation in the WPS was reduced from level IV to

level III before increasing to level IV, while that in the CS increased from level III to level IV before reducing to level III (Figure 8d). In May, the WPS stayed at level III over the whole day, while that in the CS was reduced from level III to level II before increasing to level III (Figure 8e). In June and September, the daily variation of the CTI in the WPS was reduced from level III to level II before increasing to level III, while that in the CS stayed at level II over the whole day (Figure 8f,i). In July, the daily variation of the CTI in the WPS was reduced from level II to level I before increasing to level II, while that in the CS increased from level I to level II before reducing to level I (Figure 8g). In August, the daily variation of the CTI in both the WPS and CS was reduced from level II to level I before increasing to level II (Figure 8h). In October, both the WPS and CS stayed at level III over the whole day (Figure 8j). In November, both the WPS and CS stayed at level IV over the whole day (Figure 8k). Finally, in December, the daily variation of the CTI in the WPS increased from level IV to level V before reducing to level IV, while the CS stayed at level IV over the whole day (Figure 8l).

#### **4. Discussion**

#### *4.1. Ecosystem Services of Urban Wetland Parks in Terms of Microclimate Improvement*

This study explored the microclimate and HTC of urban wetland parks, and the seasonal, monthly, and daily variations of the microclimate, HTCI, and CTI were quantitatively analyzed. The results demonstrate that the WPS displayed evident seasonal and monthly changes in the microclimate, suggesting that the microclimates of urban wetland parks are affected primarily by the overall water and heat conditions and maintain similar changes as the regional climate [44]. HTCI was studied on an annual scale, which resulted in a relatively larger range fluctuation. This phenomenon of large ranges on a larger space–time scale has been mentioned in other studies by Vinogradova (2021) [45] and An et al. (2021) [46]. Therefore, in order to obtain a perception of a certain scenario, we also mainly studied the monthly and daily variations of HTCI. The ecological benefits exhibited reduced temperature, increased humidity, and decreased wind speed in summer and other hot months, which confirms that urban wetland parks could provide good ecological services in terms of mitigating the heat island effect and dry island effect.

The findings have been supported by other studies [16,47–50]. The main reason for these is that, compared with gray infrastructure (impermeable concrete and buildings), green infrastructure, especially trees and water bodies, can absorb solar radiation, reduce the land surface temperature, and reduce the increase in vapor caused by solar radiation [17,51,52]. Moreover, ecological infrastructures that are reliant on vegetation can reduce the temperature via shading and increase heat fluxes, because they prevent the solar radiation from reaching the surface and through evapotranspiration to form low-temperature areas under canopies or in grasslands) [18,53]. Additionally, the natural branching configuration of plants can directly block the wind, produce a wind barrier effect, and can reduce the wind energy via the swinging of branches, thereby reducing wind speed (see Figure 9).

**Figure 9.** Schematic diagrams of the (**a**) shading, (**b**) transpiration, and (**c**) wind barrier effects of plant communities.

Some interesting findings based on the further analysis of the daily variation in each month were also presented. It was found that urban wetland parks have a warming effect in the midday of the cold months, while they can effectively reduce fluctuations in the daily variation of the wind speed in warmer months. The lack of vegetation and the thermal conductivity of large, impervious surfaces in urban environments result in faster heat loss during colder months [54–56]. Additionally, urban forests will also modify the airflow, thereby causing local strong winds, and the wind chill benefit is obvious during colder months. However, the ecological infrastructures in urban wetland parks have complex spatial structures, and therefore do not have the smooth, reflective planes of artificial facilities. Thus, when solar radiation reaches living organisms such as plants, it can be trapped by plants or diffuse, thereby affecting the energy exchange between vegetation patches [57] (see Figure 10). Moreover, the ecological infrastructures in urban wetland parks also offset a portion of the wind energy via branches and leaves, thereby reducing the daily variation of the wind speed. Therefore, urban wetland parks exert a certain heat preservation effect in the midday period of cold months, rather than a real warming effect.

**Figure 10.** A diagram of the absorption and diffuse reflection of solar radiation by plants.

#### *4.2. HTC Features of Urban Wetland Parks*

HTC is a parameter based on temperature, relative humidity, and wind. The data collected and analyzed in this study indicate that the HTCI in the WPS presented significant seasonal variation. In spring, summer, and autumn, the HTCI in the WPS was at a very comfortable and comfortable level, respectively, whereas in winter, it was at a very uncomfortable level. However, the urban wetland park was found to have a better comfort level than the urban environment only in summer (July and August).

The daily variation of the HTC in each month was also discussed. Xixi National Wetland Park can provide a comfortable environment throughout the day in May, June, July, September, and October, as well as at midday in March, April, and November. A comfortable environment was not available at noon in August or throughout the day in winter (December, January, and February). However, compared with the urban environment, Xixi National Wetland Park was found to have significantly improved comfort in the midday hours of winter.

For Hangzhou, China, which has a subtropical monsoon climate, compared with other climatic factors, humidity has a greater impact on the comfort perception of the human body throughout the year. Particularly, in winter, the "cold and wet" climate features are very unfavorable to people staying outdoors [58]. In the cooler months, urban wetland parks can increase outdoor comfort in the warm midday hours. In August, compared with the urban environment, the urban wetland park was found to provide a better comfort level in the midday hours, but the high temperature and high humidity are not sufficient to make a person feel comfortable. Hence, based on the comparison between the comfort levels provided by the urban wetland park and urban environment, in an urban wetland park with a subtropical monsoon climate, July and August provide the largest improvements of HTC.

#### *4.3. Clothing Suggestions*

According to the definition and purpose of the CTI, although Xixi National Wetland Park is in a subtropical climate area, the residents are suggested to wear heavy winter clothes when they go out in the cold season due to the restriction of high air humidity, especially at night. Recreational activities with accommodation are not recommended in winter, and specific protective measures should be taken to make visitors feel comfortable.

When visitors go to Xixi National Wetland Park in summer, they will feel much more comfortable wearing cool summer clothes. In August, they are suggested to take good protective measures (e.g., to wear a sunshade hat and sun-protective clothing) and enjoy their activities in the morning, evening, or at night, not at midday.

In this study, the HTCI and CTI were synthetically calculated by climate factors. While both indexes have had numerous applications since their conception, they both indirectly reflect the human body perception and provide clothing suggestions. Hence, in future research, determining the direct perceptions of volunteers and further combining these data with climate data to carry out an evidence-based study will be conducive to research on the relationship between the ecological environment and human health.

#### **5. Conclusions**

The increase of gray infrastructure (e.g., concrete buildings, hard pavements, and metal materials) and the decrease of ecological infrastructure (e.g., greenbelts, wetlands, and water bodies) change the underlying structure of the urban ecological environment, thereby affecting the ecosystem services in urban areas [52]. In this study, the effect of urban wetland parks on HTC over one year was quantified, and the results indicate that urban wetland parks can mitigate the heat island effect and dry island effect (by reducing the temperature, increasing the humidity, and reducing the wind speed) in summer, thereby exhibiting a good ecological function. More importantly, urban wetland parks can provide ecological services at midday during winter to mitigate the cold island effect, thereby exerting a certain heat preservation effect. Additionally, urban wetland parks were found

to exhibit their best performance in improving HTC during the daytime of the hot season (June, July, August, and September, and especially the whole day in July and August) and the midday period of the cold season (December, January, February, and March). However, improvements in other months (especially in October and November) were not authenticated by the data analyzed in this study. Finally, based on the findings of this study, it is suggested that citizens should take good protective measures and enjoy their activities in the morning, evening, or at night, not at midday in hot weather. Moreover, extra layers are suggested to be worn before visiting urban wetland parks at night in cold weather, and recreational activities involving accommodation are not recommended.

In urban planning, more green space (plants) and blue space (water bodies) should be introduced, and the effect of impervious surfaces on the land surface climate should be reduced to create a microclimate conducive to human health. Administrators and policymakers should consider detailed management and strategies in parks and should plan indoor and outdoor activities for visitors to induce the most comfort and relaxation.

Finally, the HTCI is based upon a thermal stress index that does not account for radiation, which is difficult to comprehensively characterize from a microclimatic perspective. Meanwhile, like other empirical indexes, the HTCI is also intrinsically unable to take into account metabolic rate and clothing insulation. These are the certain limitations of this study. In addition, because each thermal index has its own applicable scope, it is suggested that, in addition to focusing on the research object and its thermal environment condition, the selection of the index and the formulation of a monitoring program should take into account the clothing insulation effect, metabolic activity changes, and the spatial heterogeneity of temperature [37].

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/f12101322/s1, Figure S1: The daily variations of the average temperature, Figure S2: The daily variations of the average relative humidity, Figure S3: The daily variations of the average wind speed.

**Author Contributions:** Conceptualization, Z.Z. and B.Y.; writing—original draft preparation, Z.Z. and J.D.; writing—review & editing, Z.Z. and Q.H.; supervision, B.Y. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research is supported by the Fundamental Research Funds for CAF (CAFYBB2019ZC008).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data is contained within the article and Supplementary Materials.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **The Nature and Size Fractions of Particulate Matter Deposited on Leaves of Four Tree Species in Beijing, China**

**Huixia Wang 1, Yan Xing 2, Jia Yang 1, Binze Xie 1, Hui Shi 1,\* and Yanhui Wang 3,\***

<sup>1</sup> School of Environmental and Municipal Engineering, Xi'an University of Architecture and Technology, Xi'an 710055, China; wanghuixia@xauat.edu.cn (H.W.); yangjia1802@126.com (J.Y.); xiebinzejianda@126.com (B.X.)

<sup>2</sup> Shaanxi Environmental Monitoring Centre, Xi'an 710054, China; xy18792833799@126.com

<sup>3</sup> Institute of Forest Ecology, Environment and Nature Protection, Chinese Academy of Forestry, Beijing 100091, China

**\*** Correspondence: shihui@xauat.edu.cn (H.S.); wangyh@caf.ac.cn (Y.W.)

**Abstract:** Particulate matter (PM) in different size fractions (PM0.1–2.5, PM2.5–10 and PM>10) accumulation on four tree species (*Populus tomentosa*, *Platanus acerifolia*, *Fraxinus chinensis*, and *Ginkgo biloba*) at two sites with different pollution levels was examined in Beijing, China. Among the tested tree species, *P. acerifolia* was the most efficient species in capturing PM, followed by *F. chinensis*, *G. biloba*, and *P. tomentosa*. The heavily polluted site had higher PM accumulation on foliage and a higher percentage of PM0.1–2.5 and PM2.5–10. Encapsulation of PM within cuticles was observed on leaves of *F. chinensis* and *G. biloba*, which was further dominated by PM2.5. Leaf surface structure explains the considerable differences in PM accumulation among tree species. The amounts of accumulated PM (PM0.1–2.5, PM2.5–10, and PM>10) increased with the increase of stomatal aperture, stomatal width, leaf length, leaf width, and stomatal density, but decreases with contact angle. Considering PM accumulation ability, leaf area index, and tolerance to pollutants in urban areas, we suggest *P. acerifolia* should be used more frequently in urban areas, especially in "hotspots" in city centers (e.g., roads/streets with heavy traffic loads). However, *G. biloba* and *P. tomentosa* should be installed in less polluted areas.

**Keywords:** air pollution alleviation; accumulation on leaves; PM2.5; encapsulated particles; urban trees

#### **1. Introduction**

In Beijing, the capital of China, with rapid economic development, urbanization, and industrialization, ecological problems are becoming increasingly prominent. Air pollution, especially particulate matter (PM), has become one of the most severe problems over the past several decades [1]. Studies have shown that PM has recently been ranked fifth among the major risk factors threatening human health globally, and thus is first among environmental risks [2,3]. Particles with a diameter less than 10 μm (PM10) can cause premature mortality, accelerated atherosclerosis, lung cancer, heart disease, asthma, preterm birth, mutagenicity and DNA damage, and inflammatory responses [4–6]. Nevertheless, fine particles (particles with diameter less than 2.5 μm, PM2.5) are more toxic and more strongly associated with human health effects than coarse particles (particles with diameter between 2.5 and 10 μm) [7]. Therefore, the air quality standards for PM10 and PM2.5 in China are set to 75 and 35 μg/m<sup>3</sup> (annual mean), and 150 and 70 μg/m<sup>3</sup> (daily mean), respectively [8]. However, urban Beijing's PM10 and PM2.5 concentrations are much higher than the national standards [9]. Thus, reducing PM concentrations, especially PM2.5, is considered one of the most significant tasks related to environmental protection in urban areas. The government has taken measures to control pollutant sources (e.g., adjusting the industrial structure and promoting energy-saving technology). Meanwhile, massive

**Citation:** Wang, H.; Xing, Y.; Yang, J.; Xie, B.; Shi, H.; Wang, Y. The Nature and Size Fractions of Particulate Matter Deposited on Leaves of Four Tree Species in Beijing, China. *Forests* **2022**, *13*, 316. https://doi.org/ 10.3390/f13020316

Academic Editors: Manuel Esperon-Rodrigue and Tina Harrison

Received: 17 January 2022 Accepted: 11 February 2022 Published: 15 February 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

afforestation is believed to be an additional and helpful measure to alleviate air pollution by filtering and adsorbing PM through forest crown/leaves.

Trees are a significant element in a city's landscape and are the most effective vegetation types with regards to reducing PM [10]. This is mainly due to the fact that trees have extensive leaf areas [11], and the structure of tree crowns changes the turbulence of air movement above and within tree canopies [12]. McDonald et al. [13] estimated that a 3.7–16.5% increase in tree cover in West Midland, UK, could reduce PM10 concentrations by 10%. Reductions in total suspended particles (TSP), PM10, PM7, PM4, PM2.5, and PM1 associated with an increase in canopy cover have been reported in China, the United States, Chile, and Israel [14–17], indicating a direct positive effect on air quality by removing PM2.5 by urban trees. If a massive plantation was employed to decrease airborne PM2.5, the differences in PM2.5 accumulation among tree species should be considered and the most efficient species should be recognized. Some studies compared the PM2.5 capturing ability of different plant species. Räsänen et al. [18] investigated the efficiency (*C*p) of *Pinus sylvestris*, *Betula pubescens*, *Tilia vulgaris*, and *Betula pendula* leaves to capture PM2.5 using simulation method (i.e., NaCl particles). They found that *C*<sup>p</sup> is influenced by leaf structures (e.g., leaf size, wettability, stomatal density, and leaf hair density). However, simulation studies are different from field investigations. Sæbø et al. [11] compared the PM accumulation on leaves of 47 species, including 22 trees and 25 shrubs, in Norway and Poland. They found a species-related difference in PM accumulation in both countries. The abilities of leaves to accumulate PM and its size fractions could be attributed to leaf morphology (e.g., leaf hair density, leaf roughness, and wax content) [6,19–21].

The detriments of PM on human health are primarily determined by particle size. Thus, the present study examined plants accumulating PM in two aspects. First, the amount of PM and its size fractions (PM0.1–2.5, PM2.5–10, and PM>10) deposited on leaves in two contrasting urban environments were quantified. Second, the influences of anatomical/physiological leaf characteristics (e.g., stomatal density, stomatal size, single leaf area, wettability) on PM (i.e., PM, PM0.1–2.5, PM2.5–10, and PM>10) accumulation abilities were investigated. Four tree species (*Populus tomentosa*, *Platanus acerifolia*, *Fraxinus chinensis*, and *Ginkgo biloba*) were selected as test species due to their prevalence in urban and suburban environments, widespread in temperate regions, and also since they are recommended for extensive plantation in Beijing. The findings of this study can provide the impetus for using urban trees to improve air quality, and provide guidance for the work of urban planners and those involved in environmental protection.

#### **2. Materials and Methods**

#### *2.1. Plant Materials and Experimental Sites*

Four tree species, *P. tomentosa*, *P. acerifolia*, *F. chinensis*, and *G. biloba*, were selected for this study at Beijing Botanical Garden (Site 1, located in Haidian District, 39◦59 29.66 N, 116◦12 40.25 E, upwind of Beijing) and Huangcun (Site 2, located in Daxing District, 39◦42 45.13 N, 116◦19 08.44 E, downwind of Beijing) (Figure 1). The sampling plants grow in the center of the garden or near a busy road with a traffic density of ~5680 cars/h at Site 1 and Site 2. The two sites showed different PM2.5 and PM10 concentrations, as measured at the nearest monitoring station operated by the Beijing Municipal Ecological and Environmental Monitoring Center [1].

**Figure 1.** Location of the study sites.

#### *2.2. Sampling Procedure*

Leaf sampling was conducted on October 1 and 3, 2014 at Site 1 and Site 2, respectively, when there was no previous rainfall for more than a week. For each species, five individual trees under good growth conditions were selected. At Site 2, the distance between the sampled trees and the road center was about 10 m. Thus, the surrounding environment of each sample tree was similar. Small branches with mature and healthy leaves were cut from four dimensions (N, S, E and W) at 2–6 m above ground at each site and for each species. After cutting, small branches bearing leaves were placed in labeled ziplock bags, transported to the laboratory, and analyzed as soon as possible.

#### *2.3. Analysis of PM*

For each plant species at Sites 1 and 2, three batches of leaves were initially prepared. For each batch, 20–30 pieces for *P. acerifolia*, or 30–80 pieces for *P. tomentosa*, *F. chinensis*, and *G. biloba* were selected. The leaves were hand washed using a brush, with 200 mL of ultrapure water (ELGA, High Wycombe, Buckinghamshire, UK). The hemi-surface leaf area was measured using Image J software (Version 1.46; Wayne Rasband, National Institutes of Health, Bethesda, ML, USA) after scanning (HP Scanjet 3570c, HP Inc., Palo Alto, CA, USA). For the filtration procedure, membranes with pore size of 10, 2.5, and 0.1 μm were first soaked in ultrapure water for 2 h and then dried at 105 ◦C in a drying chamber for 6–8 h to remove soluble impurities. The filters were then put in a balancing chamber for at least 24 h to stabilize. Every membrane was pre-weighed before filtration using a balance with 0.1 mg accuracy (SI-114, Denver Instrument Co., Arvada, CO, USA). The washing solution was hand-shaken for several minutes to re-suspend all washed particles before filtration. Washing solution was then pumped through membranes with pore size of 10, 2.5, and

0.1 μm successively. The filtration was carried out using a 47-mm glass filter funnel with stopper support assembly (Millipore Corp., Bedford, MA, USA) connected to a vacuum pump (SHB-III; Greatwall Scientific Industrial and Trade, Co., Ltd., Zhengzhou, China). Three fractions of PM were collected on the filters: (i) PM>10 (particles intercepted by membrane with pore size of 10 μm, large), (ii) PM2.5–10 (particles intercepted by membrane with pore size of 2.5 μm, coarse), and (iii) PM0.1–2.5 (particles intercepted by membrane with pore size of 0.1 μm, fine). Loaded filters were subsequently dried for more than 24 h at 40◦C, stabilized in the weighing room for 30 min, and then re-weighed. Consequently, the pre-weight was subtracted from the post-weight to calculate the mass of PM deposited on leaves in every size fraction of each washed sample. The resulting weight was finally divided by leaf area. At this moment, we obtained the total weight of deposited PM per unit leaf area for each sample, and also for PM0.1–2.5, PM2.5–10, and PM>10.

The deposited PM0.1–2.5, PM2.5–10, PM>10, and PM per unit green land was calculated by multiplying PM per unit leaf area and leaf area index (LAI). The LAI values were 2.13, 3.18, 2.67, and 2.52 for *P. tomentosa*, *P. acerifolia*, *F. chinensis*, and *G. biloba*, respectively, using a LAI-2000 (LI-COR., Inc., Lincoln, NE, USA) at Site 1.

#### *2.4. Analysis of Leaf Surface Characteristics*

Field emission scanning electron microscopy (FESEM, Quanta 200 FEG; FEI Company, Hillsboro, OR, USA) was used to determine leaf surface microstructures. Six pieces (three for upper side and three for lower side) of air-dried samples (about 5 mm × 5 mm) for each species at each site were cut from the center of the leaves, and coated with a thin layer of gold-palladium using a precision etching coating system (Model 682, Ga-tan Co., Ltd., Pleasanton, CA, USA). Stomatal density (per mm2) was determined by calculating the number of stomata in ten FESEM images with a magnification of ×500. Particle frequency on stomata of each species was counted from 10 FESEM images with a magnification of ×1000. Fifty stomata were randomly selected to measure stomatal length, width, and aperture. Leaf roughness on upper and lower sides was evaluated on a subjective scale (1= relatively smooth, and 5 = very rough) [11] using FESEM images with a magnification of ×1000.

The wettability of leaf surface was evaluated by contact angle (CA) with distilled water. CAs were determined on upper and lower sides at room temperature using a goniometer (Kino SL200A, KINO Industry CO. Ltd., Somerville, Boston, MA, USA). For every species at every site, thirty pieces (about 5 mm × 5 mm, fifteen for upper side and fifteen for lower side) were cut from the middle of each leaf next to the main vein. Then, these pieces were attached to a glass plate with double-sided tape. A 3-μL water droplet was made with a capillary tube and carefully applied to the leaf surface. A photograph of the profile of each water droplet resting on the leaf surface was taken with a charge-coupled device equipped with a camera within 30 s after placing the water droplet. The digital photographs were downloaded, and CAs were determined using computer software (CAST2.0, KINO Industry CO. Ltd., Somerville, Boston, MA, USA). Then, the mean values were calculated.

Measurements of specific leaf area (SLA) and single leaf area, leaf length, leaf width, and petiole length were made on the same batches as were used for analysis of PM. After scanning, 30 leaves were randomly selected for measurements of single leaf area, leaf length, leaf width and petiole length using Image J software. The batches were dried in a drying chamber at 80 °C for at least 24-h and weighed to produce a value of dry weight. The dry weight and measured leaf area of the batches were used to calculate SLA as cm2/g (dry weight).

#### *2.5. Data Analysis*

Statistical tests were performed with Minitab 16 software (Minitab Ltd., Shanghai, China). One-way analysis of variance was undertaken to estimate the differences in PM accumulation and its size fractions (PM0.1–2.5, PM2.5–10, and PM>10) among different species at each site. The main effects of tree species, different sites, and their interaction on

accumulation of PM on leaves and its size fractions (PM0.1–2.5, PM2.5–10, and PM>10) were tested with a two-way analysis of variance. The relative importance of measured leaf characteristics (stomatal density, stomatal length, stomatal width, stomatal aperture, CA upper side, CA lower side, single leaf area, SLA, leaf length, leaf width, petiole length, roughness upper side, roughness lower side) on the amount of PM and its size fractions was evaluated using principal component analysis (PCA) and partial least squares regression. A given effect was assumed to be significant at *p* < 0.05.

#### **3. Results**

#### *3.1. Differences in PM and Its Size Fractions among Species*

The total PM deposited on leaves was significantly different among four species at each site (*p* < 0.001, Table 1). The amounts of PM0.1–2.5, PM2.5–10, PM>10, and PM ranged from 3.9–14.2, 5.7–41.2, 80.0–109.1, and 89.6–164.5 μg/cm2; and 9.3–25.0, 15.9–51.7, 98.7–389.5, and 123.9–466.2 μg/cm2, at Site 1 and Site 2, respectively. Among the four species, *P. acerifolia* had the greatest accumulated PM, followed by *F. chinensis*, *G. biloba*, and *P. tomentosa*. The mass of PM0.1–2.5, PM2.5–10, and PM>10 accumulated on leaves also showed significant differences among tree species (*p* < 0.001, Table 1), except for PM>10 at Site 1 (*p* = 0.363, Table 1). For all species, PM>10 and PM0.1–2.5 made up the greatest (66.4–89.2%) and smallest (4.4–10.0%) proportion of accumulated PM, respectively (Table 1).

The deposited PM0.1–2.5, PM2.5–10, PM>10, and PM per unit green land were 8.3, 45.2, 18.7, and 16.1 μg/cm2 (PM0.1–2.5); 12.1, 131.0, 29.6, and 21.7 μg/cm2 (PM2.5–10); 170.4, 346.9, 249.1, and 209.7 μg/cm<sup>2</sup> (PM>10), and 190.8, 523.1, 297.4, and 247.5 μg/cm<sup>2</sup> (PM), for *P. tomentosa*, *P. acerifolia*, *F. chinensis*, and *G. biloba*, respectively.

#### *3.2. Differences in PM and Its Size Fractions among Sites*

The PM deposited on leaves was significantly higher at Site 2 than that at Site 1 (*p* < 0.001, Table 1), with an increase of 40%, 180%, 40%, and 50% for *P. tomentosa*, *P. acerifolia*, *F. chinensis*, and *G. biloba*, respectively (Table 1). The corresponding increase at Site 2 compared with Site 1 was 20%, 260%, 30%, and 40% for PM>10, 180%, 30%, 120%, and 110% for PM2.5–10, and 140%, 80%, 120%, and 70% for PM0.1–2.5. The ratio of PM0.1–2.5/PM0.1–10 was 0.38 at Site 1 and 0.36 at Site 2. Both are lower than that in ambient air, which was 0.82 at Site 1 and 0.69 at Site 2 during the growing season (May to October, 2014).

#### *3.3. Morphological Structure of Leaf Surfaces*

Table 2 and Figure 2 present the leaf surface structural properties of four studied tree species. *P. acerifolia* had the largest single-leaf area, followed by *P. tomentosa*, *F. chinensis*, and *G. biloba*. In terms of leaf wettability, all of the analyzed species, except the lower side of *G. biloba*, had a mean CA less than 90◦ (i.e., they were wettable) [22].

In the FESEM study, epicuticular wax was observed in tubular form (*G. biloba*, Figure 2j–l), or wax film (*P. tomentosa*, *P. acerifolia*, *F. chinensis*, Figure 2a–i). Wrinkled cuticles were observed on the lower side of *P. tomentosa* (Figure 2b,c), both surfaces of *P. acerifolia* (Figure 2d–f), lower side of *F. chinensis* (Figure 2h–i). The upper side of *F. chinensis* (Figure 2g) and *G. biloba* (Figure 2j). Stomata of the investigated species were either level with epidermal cells (*P. tomentosa*, Figure 2b,c), sunken (*F. chinensis*, Figure 2h,i, *G. biloba*, Figure 2k–l), or slightly elevated (*P. acerifolia*, Figure 2e,f). *P. tomentosa* had smaller stomata than the other species. However, the greatest stomatal aperture occurred on the foliage of *P. acerifolia*.


*Forests* **2022** , *13*, 316

Leaf length (cm) Leaf width (cm) Petiole length (cm)

Roughness (upper side)

Roughness (lower side)

n.d. indicates variables that were not found.

 10.1 ± 1.6

 9.3 ± 0.9

 8.6 ± 1.0

 18.2 ± 2.3

 19.9 ± 1.0

 7.7 ± 1.1

 9.8 ± 1.5

 4.6 ± 0.6

 1.6 ± 1.1

 4.5 ± 0.6

 6.2 ± 1.0

 4.5 ± 1.7

 7.8 ± 1.3

 7.1 ± 1.0

 5.2 ± 0.9 13342334

34453445

 14.7 ± 1.3

 15.5 ± 1.7

 4.8 ± 1.2

 8.6 ± 0.9

 3.9 ± 0.4

 1.2 ± 0.8

 3.4 ± 0.6

 4.3 ± 1.2

 3.4 ± 1.1

**Figure 2.** Scanning electron microscopy images of *Populus tomentosa* (**a**–**c**), (**d**–**f**), (**g**–**i**), and (**j**–**l**). **a**, **d**, **g**, **j**, **m**, and **n**, the upper side; **b**, **e**, **h**, and **k**, the lower side; **c**, **f**, **i**, and **l**, stomata. Particulate matter embedded in leaf epidermis (**m**: *Fraxinus chinensis*; **n**: *Ginkgo biloba*). Symbols indicate examples of stomata (triangle) and particulate matter (arrow).

#### *3.4. Encapsulation of PM*

PMs within cuticles were observed for *F. chinensis* (Figure 2m) and *G. biloba* (Figure 2n), but not for *P. tomentosa* and *P. acerifolia.* Encapsulation of PM within cuticles was not so common using FESEM observation. No visible damage was observed to either cuticle or epidermal cell. Wax encapsulated particles had diameters less than 6 μm, which was dominated by PM2.5 (>90%).

#### *3.5. The Effects of Leaf Structure on PM Accumulation*

Stomatal aperture, stomatal width, leaf length, leaf width, stomatal density, and CA were crucial predictors for PM accumulation and its size fractions (Table 3, Figure 3). Among the 13 dependent variables, the sum of PC#1 and PC#2 was 71.8% (Figure 3). An increase in CA predicted a decrease in PM accumulation, while increased stomatal aperture, leaf length, leaf width, stomatal width, and stomatal density predicted an increase in PM accumulation (Table 3).

**Table 3.** Regression coefficient (B) and standardized regression coefficient (Beta) of the partial least squares regression model for the factors affecting leaf particulate matter capturing of four tree species.


**Figure 3.** Loading plot of principal component analysis of the tree species with 13 dependent variables.

#### **4. Discussion**

#### *4.1. Differences in PM Accumulation among Species*

The plant species showed significant differences in PM accumulation among tree species and at different sites. In a study conducted by Sæbø et al. [11], they found that *Pinus mugo*, *P. sylvestris*, *Taxus media*, *Taxus baccata*, *Stephanandra incise*, and *B. pendula* showed higher PM accumulation. *Acer platanoides*, *Prunus avium*, and *Tilia cordata* showed lower PM accumulation. Zhang et al. [23] compared foliar PM retention and its size fractions of five plant species, *P. acerifolia* showed higher amount of PM accumulation ability than other species.

The differences in PM accumulation among species should be used for species selection during afforestation. However, tree size and LAI are important in determining the amount of PM accumulated per unit of land area, which is an index for PM retention ability and efficiency [24,25]. The relatively higher LAI of *P. acerifolia* further increased its PM removal potential, suggesting that this species could be used suitably as PM filters. The trees in Beijing could be damaged by air pollution; thus, the actual LAI was lower than healthy trees, which would influence the amount of accumulated PM. Among the investigated tree species, *P. acerifolia* is tolerant of air pollution, *G. biloba* and *P. tomentosa* are sensitive [26]. Therefore, we suggest that *P. acerifolia* should be used more in urban areas, especially in "hotspots" of PM pollution (e.g., the middle of the city, edges of streets/roads with heavy traffic loads). However, *G. biloba* and *P. tomentosa* should be installed in less polluted areas.

The contribution of PM in different size fractions to total PM on leaves decreased with the decreasing of PM diameter. Terzaghi et al. [27] found that large, coarse, and fine particles accounted for 87–95%, 5–12%, and 0.1–0.5% of the total PM for *Cornus mas*, *Acer pseudoplatanus*, and *Pinus pinea*. A study taken along a busy road in UK found the amount of PM captured on leaves increased with decreasing particle diameter [28]. These varied deposition level of PM with different aerodynamic diameter probably reflects the different aerodynamic properties of PM and their interactions with different leaf characteristics. Meanwhile, this difference can be ascribed to two groups of factors. The first group include PM concentration and compositions, plant species, and filtering materials (filter paper or membrane). The second group are mainly the used methods (measured mass or mass derived from particle number). Particles were assumed to be spherical and particle density equal for each size class [27].

#### *4.2. Differences in PM Accumulation between Sites*

More PM was found on the foliage of plants grown at heavily polluted site (Site 2) than less polluted site (Site 1). Deposition of pollutants depends on deposition velocity and pollutant concentration [29,30]. Micrometeorological conditions have been demonstrated to strongly impact on deposition dynamics, which may be partly explain the differences in PM accumulation between the two sites in our study. Furthermore, the deposition of PM on leaves depends on PM diameter. Deposition of PM with sizes of 0.1–1 μm is influenced by Brownian diffusivity and Stokes' law, and which is independent of size. For PM with diameter large than 1 μm, Stokes' law dominates the process of deposition, and which depends mostly on particle size. Larger particles will, therefore, deposit faster than smaller ones, either by sedimentation under the influence of gravity or by turbulent transfer resulting in impaction and interception [31], leading to a lower ratio of PM0.1–2.5/PM0.1–10 in Site 2 than the less polluted site.

In the present study, we observed a relatively higher percentage of fine (0.1–2.5 μm) and coarse (2.5–10 μm) size fractions in Site 2 than those in Site 1, while the large fractions were in lower percentages at Site 2. This finding seems to contrast with the results of Przybysz et al. [32], who obtained a relatively high rate of fine particles (6.2–9.8%) on plants at a rural site. Here, the lower ratio of PM0.1–2.5/PM0.1–10 accumulated on leaves, compared with that in ambient air, may be caused by three factors. First, some particles, elements, or ions are dissolved in water during the washing process [33]. Second, during the process of filtration, the membrane with pore sizes of 10 μm and (or) 2.5 μm may intercept some

particles with diameters less than 10 μm and (or) 2.5 μm after saturation [34]. Third, the leaf cuticle encapsulated some particles, especially tiny ones [27].

#### *4.3. Importance of Leaf Structure for PM Accumulation*

Urban trees accumulating PM is a complex and poorly characterized process which is influenced by many factors such as leaf surface micro-, and macro-structure (e.g., leaf wettability, stomatal density, leaf area, leaf roughness, and the shape and amount of trichomes) [11,23,24,35]. Research has shown that leaves with small size, low CAs, low stomatal density, high stomatal conductance, and higher amount of leaf hairiness can capture more PM [11,18,19,21,31]. In this study, we also found that high CAs decreased the amount of captured PM. The contact area between a particle and the underlying leaf surface is considerably reduced on surfaces with high CAs. Consequently, the physical adhesion forces between particles and leaf surfaces are reduced, leading to a lower PM accumulation [19]. However, we also found that increased stomatal aperture, density, and width increased the amounts of captured PM. Large stomatal density, aperture, and width could result in increased transpiration which can make particles more deliquescent. And as a consequence, deposition rates increase [36]. Transpiration of water through stomata can cool leaf surfaces, but increase PM deposition by thermophoresis [18], which may also partly explain the higher ability of leaves to capture PM under more polluted conditions than conparatively less polluted areas.

We found that *P. acerifolia*, the species with the largest leaves, had the highest PM accumulation among the investigated species. This is in opposition to the findings of previous studies, in which the authors found small leaves increased the number of captured particles [18,28]. According to Nobel [37], particles in the air could more easily collide with small leaves than large and flat leaves, which have thicker boundary layers. This contrasting finding may be caused by the microstructure of *P. acerifolia* leaves, which had a rough surface that can influence the boundary layer [38].

Some PM with diameters less than 6 μm (mainly PM2.5) were encapsulated in cuticles of leaves of *F. chinensis* and *G. biloba*, but not for *P. acerifolia* and *P. tomentosa*, by FESEM observation. These results suggest that the potential of PM embedded in wax layer depends on the quantity of wax as well as the composition and structure of the epicuticular wax layer; these are species-specific characteristics [39]. Terzaghi et al. [27] found that particles with diameters less than 10.6 μm were encapsulated into cuticles. The amount of encapsulation and the capacity of leaves to capture PM changed over time. They attributed this to the degradation of cuticular waxes, from a perfect wax crystal to an amorphous one. Dzierzanowki et al. [ ˙ 39] demonstrated that large particles appeared mainly on the leaf surface rather than in the wax layer of some plant species. The encapsulated particles always had small diameters and could not be easily washed off during rain events or dislodged by wind. If most of the accumulated PM is immobilized which can be considered to be beneficial in the planning of PM phytoremediation. However, if the washed- or blown-off PM is considered to be filter cleaning, leaving the leaves ready for additional deposition. This process may result in underestimating the PM removal effect [23]. The dynamics of deposition, including the amounts of PM washed-off by rain and blown-off by wind, need further investigation.

#### **5. Conclusions**


(iii) Leaf structures affect PM accumulation and its size fractions. Large leaves, along with low stomatal aperture, width, and density, as well as low CA, all resulted in increased PM capture.

**Author Contributions:** Conceptualization, H.W., H.S. and Y.W.; methodology, H.W. and Y.W.; validation, H.W.; investigation, H.W., Y.X., J.Y. and B.X.; resources, H.W. and Y.X.; data curation, H.W. and Y.X.; writing—original draft preparation, H.W., J.Y. and B.X.; writing, review and editing, H.W., H.S. and Y.W.; visualization, H.W., J.Y. and B.X.; supervision, H.S. and Y.W.; project administration, H.W.; funding acquisition, H.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Scientific Research Program of the Education Department of Shaanxi Provincial Government, grant number 20JS082.

**Acknowledgments:** We are grateful to Wenyan Yang, Tsinghua University, for assistance with the field emission scanning electron microscopy. We also are grateful to the Key Lab of Northwest Water Resources and Environment Ecology of Ministry of Education for providing the goniometer.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

