Challenges and Potentials: Environmental Assessment of Particulate Matter in Spaces Under Highway Viaducts

Abstract

Under-viaduct space (UVS) is becoming an important solution to urban mobility problems, and the construction and use of high-density city center highways and elevated bridges are increasing, which has a negative impact on the UVS. Air pollution is a problem in these spaces, but research on air pollution in UVSs is lacking. To further study air pollution in UVS, this study selected a case area of a UVS in central Shanghai and investigated the spatiotemporal distribution patterns of air pollution and the influencing factors. We found that air pollution in the UVS is significantly higher than the background levels, and the higher the background levels, the greater the difference between the pollution of the UVS and the background. In terms of the impact factor, air pollution is highly correlated with the built environment and traffic flow. The research provides evidence of the exposure to air pollution in under viaducts spaces in the microenvironment.

1. Introduction

Urban elevated highways and viaducts play an important role in relieving traffic congestion, but they also gather a large amount of air pollution from traffic sources [1]. Some studies point out that motor vehicles are an important source of urban particulate matter pollution [2]; therefore, it is important to study the spatiotemporal characteristics of motor vehicle pollutant emissions for urban ambient air quality [3]. Meanwhile, as the construction of high-density cities in China gradually shifts towards urban regeneration, the under-viaduct space (UVS) is regarded by scholars as a potential public space or site for activities [4], and more and more UVSs have been modified into activity spaces for residents. However, only a few studies have analyzed the health effects of activities conducted in the UVS. In this study, the UVS is defined as the space identified via orthographic projection under the viaduct as well as any surrounding neighborhood areas significantly affected by the viaduct. The efficient and rational use of the UVS is highly significant, but because of the high degree of space confinement, the UVS has become an area where particulate matter pollution caused by traffic increases [5], resulting in poor environmental quality. Therefore, the exposure to human activities under viaducts is worth considering, and the impact of viaducts on the urban environment and the use of space around viaducts are thus receiving more and more attention in high-density cities.

1.1. Air Pollution Has Become an Important Threat to Human Health

Environmental air pollution has become a major urban stressor affecting the global health burden [6,7]. Air pollution alone constitutes the largest single environmental risk factor in Europe today, causing more than 430,000 premature deaths [8].

Studies comparing the burden of disease show that air pollution is the leading cause of disability-adjusted life years (DALYs) [9]. In the process, PM_2.5 contributes to a variety of human health problems [10]. Short- or long-term exposure to high PM_2.5 concentrations can elevate the risk of heart disease while potentially triggering respiratory diseases, asthma, and other diseases, and may also induce cancers, such as lung cancer, and increase the risk of mental disorders, such as major depression [11,12,13,14].

1.2. Air Pollution Is an Important Mode of Transportation Pollution

Exposure to air pollution is important for physical health, especially in urban areas where road traffic is a major source of air pollution [15]. Among the various common sources of air pollution in urban environments, road traffic is recognized as a major contributor to it [16]. For road traffic, exhaust emissions are the focus of current research, and the contribution of non-exhaust emissions to air pollution is still questionable and, therefore, not regulated at present [17]. It has also been shown that urbanization of an area causes a dramatic increase in traffic air pollution and is associated with a wide range of health disorders in people of different ages and genders [18].

Air pollution caused by traffic is a major health hazard for urban dwellers worldwide [19]. Traffic-related air pollution is strongly associated with different cardiovascular health disorders, including blood pressure and heart rate variability [20]. Particulate matter with a diameter less than or equal to 2.5 microns (PM_2.5) is an essential criterion for determining air pollution [21], and most PM_2.5 emissions in cities originate from transportation, particularly diesel trucks and buses [22]. In Toronto, motor vehicle-related emissions account for about 40% of PM_2.5 [23]. A previous study [24] found that the percentage contributions of motor vehicles to PM_2.5 in seven U.S. sites ranged from 20% to 76%. In London, the mean incremental traffic contribution to PM_2.5 near motorways can be as high as 50% [25]. PM_2.5 can also cause many adverse health effects [26]. Research [27] found that outdoor air pollution, mainly PM_2.5, causes 3.3 million premature deaths each year.

1.3. Air Pollution Problems Worse in the UVS (But Under-Researched)

Pollution from the combination of surface roads and viaducts can result in 1.6 to 4.5 times the amount of pollution compared with road surfaces alone [5]. Urban roads in the UVS have become priority areas for particulate pollution [28]. PM_2.5 concentration distribution in viaducts is attracting increasing attention, but relatively few relevant studies have been conducted. Most existing studies have focused on urban roads, such as highways, roadsides, intersections, and arterial roads (Table 1). Urban road spaces have high similarity with the UVS, so the experimental methods and results of these studies can be used as references. In the late 1990s, studies on the PM_2.5 concentrations of major roads became more frequent and found that proximity to roads increases both PM_2.5 concentrations and the background concentration (which refers to natural pollutant concentrations measured in areas remote from the source of pollution, reflecting the base level of pollution of the atmosphere) [29,30,31,32]. In addition, some studies found an attenuation effect of PM_2.5 concentration within 200 m of highways [33,34]. Similar results were verified at intersections, where PM_2.5 concentrations at the downwind curb of the intersection were significantly higher than background values and were also significantly higher during peak traffic hours [34,35]. Many studies have also focused on influencing factors, with traffic conditions, such as traffic volume; meteorological conditions, such as wind speed, local topography, and surroundings; barriers and noise walls all affecting the distribution of PM_2.5 in the road space [36,37].

Most studies on viaducts concentrate on street canyons [38,39], while fewer focus on the horizontal distribution of PM_2.5 in the UVS. Viaducts have been found to increase PM_2.5 concentrations in street canyons, particularly on the ground. Li et al. [40] utilized a mobile monitoring method to measure the space beneath three viaducts with different types of barriers in Xi’an, China, and found that PM_2.5 concentrations have a single-peak decay trend in locations with noise barriers and combined barriers, meaning that they initially increase and then decrease to background levels within 300 m. In Wuhan, China, Yin et al. [41] monitored fourteen UVSs and nine potential environmental parameters, such as temperature, to study spatial and seasonal differences in particulate matter such as PM_2.5. The authors also found that PM_2.5 concentrations were related to the surrounding environment and that PM_2.5 concentrations in the UVS were higher compared with surrounding roads in winter and vice versa in summer. Their study contributes to understanding the factors influencing PM_2.5 concentrations in the UVS, but it lacks spatial analysis at the microscopic scale. In summary, the realistic distribution pattern of PM_2.5 in the planar dimension and the influence mechanism of parameters remain unclear; further spatial and temporal studies are needed to guide the assessment and judgment of spatial reuse.

Table 1. Summary of research findings on PM_2.5 concentration in road space.

Research Object		Monitoring Methods	Instrument	Results	Sources
Road	Highway	Mobile monitoring	Laser Photometer-TSI Sidepak AM510 (TSI Incorporated, Shoreview, MN, USA)	Within 200 m, PM2.5 concentrations usually increase closer to the highway. PM2.5 increases with the onset of peak hours and is strongly influenced by temperature changes in colder months.	Patton et al., 2014 [32]
			Tufts Air Pollution Monitoring Laboratory (Tufts University, Boston, MA, USA)	PM2.5 levels were highest in winter and lowest in summer and fall, higher on weekdays and Saturdays compared with Sundays, and higher during morning rush hour compared with later in the day.	Padró-Martínez et al., 2012 [41]
		Fixed monitoring	Continuous Dichotomous Ambient Air Monitor (Thermo Fisher Scientific, Waltham, MA, USA); 1405-DF Tapered Element Oscillating Microbalance (Thermo Fisher Scientific, Waltham, MA, USA)	The reduction in urban traffic helps to reduce the concentration of particulate matter pollution.	Azhari et al., 2021 [42]
			R&P Sequential Air Samplers (Thermo Fisher Scientific, Waltham, MA, USA)	Particulate concentrations decay exponentially with increasing distance within 100–150 m from the road and return to background levels within a few hundred meters from the road.	Clements et al., 2009 [33]
			Continuous Dichotomous Ambient Air Monitor (Thermo Fisher Scientific, Waltham, MA, USA); 1400AB Tapered Element Oscillating Microbalance (Thermo Fisher Scientific, Waltham, MA, USA)	Wind speed, prevailing wind direction, daily cycle of the atmospheric boundary layer, and traffic density are the main influencing factors.	Charron & Harrison, 2005 [43]
	On-Road and Roadside	Fixed monitoring	Continuous Dichotomous Ambient Air Monitor (Thermo Fisher Scientific, Waltham, MA, USA); Tapered Element Oscillating Microbalance (Thermo Fisher Scientific, Waltham, MA, USA)	Higher PM2.5 concentrations were observed in Madrid during the cold season.	Kassomenos et al., 2014 [44]
		Mobile and fixed-site monitoring	Desert Research Institute (DRI, Reno, NV, USA) portable samplers (Desert Research Institute, Reno, NV, USA); Laser DUSTTRAK 8530 Spectrometers (TSI Incorporated) (TSI Incorporated, Shoreview, MN, USA)	PM2.5 concentrations on the road are highly correlated with nearby roadside PM2.5 concentrations in winter and summer. Seasonal effects must be considered, and short-term exceedances of particulate matter concentrations may occur in early spring.	Cheng et al., 2010 [45]; Lozhkina et al., 2016 [46]
		Fixed monitoring	EPA Air Quality System (AQS) (United States Environmental Protection Agency, Washington, DC, USA)	There was an overall decreasing trend in PM2.5 concentrations with increasing distance from the roadside. The incremental PM2.5 decreased by 75% between 5 m and 30 m from the road.	Mukherjee et al., 2020 [47]
	Intersection	Mobile monitoring (high-flow gravimetric personal samplers)	High-Flow Personal Sampler (HFPS) (SKC Inc., Eighty Four, PA, USA)	PM2.5 concentrations are higher in the morning, decrease at noon, rise in the early afternoon, and are even higher in the evening. This phenomenon is related to traffic activity in the surrounding area.	Kaur et al., 2005 [34]
		Fixed monitoring	Laser Photometer-TSI SidePak AM510 (TSI Incorporated, Shoreview, USA); Portable Air Sampling Instruments (BGI, Boston, MA, USA)	Roadside concentrations were higher than those simultaneously monitored at urban sites away from the road. Average particulate matter concentrations at downwind roadside stations are higher than at background stations. PM2.5 concentrations during peak traffic hours are higher than during off-peak hours.	Wang et al., 2018 [37]; Kinney et al., 2011 [48]
	Arterial	Fixed monitoring (near-road continuous air monitoring station [CAMS])	Data from Near-Road Continuous Air Monitoring Station (CAMS) (TCEQ, Austin, TX, USA)	PM2.5 concentrations are usually elevated close to roads. Traffic-related PM2.5 concentrations are usually highest at night, followed by the morning peak, evening peak, and midday, respectively.	Askariyeh et al., 2020 [49]
			Laser Photometer-TSI DRX DustTrak monitor	PM2.5 quality is more related to regional sources and meteorological conditions, with a limited role in traffic volume.	Kendrick et al., 2015 [50]
	Street canyon	Mobile monitoring	Light-Scattering Laser Nephelometer (TSI Incorporated, Shoreview, MN, USA); ES-642, Metone Inc. (Met One Instruments Inc., Grants Pass, OR, USA)	Street canyons have a significant impact on the accumulation of PM2.5 concentrations on Hong Kong’s roads.	Rakowska et al., 2014 [51]
		Fixed monitoring (three heights: 0.20m, 1.0m and 2.60m)	Particle Spectrometer (DMS500) (Cambustion Ltd., Cambridge, UK)	The concentration decreases exponentially with increasing canyon height.	Kumar et al., 2008 [52]
Viaduct	Street canyon	Fixed monitoring	Laser DustTrak (TSI Incorporated, Shoreview, MN, USA)	Particle mass concentrations in street canyons are negatively correlated with height, and particle mass concentrations in street canyons are negatively correlated with temperature.	Feng et al., 2015 [38]
			Portable Laser Aerosol Spectrometer Dust Monitor(Model Grimm 11-A) (GRIMM Aerosol Technik, Ainring, Germany)	Viaducts increase PM2.5 concentrations in street canyons and significantly affect airflow fields.	Zhi et al., 2020 [39]
	UVS	Mobile monitoring	Portable Laser Aerosol Spectrometer Dust Monitor(Model Grimm 11-A) (GRIMM Aerosol Technik, Ainring, Germany)	PM2.5 concentrations decay exponentially near the viaduct without any obstacles.	Li et al., 2021 [40]
		Mobile and fixed-site monitoring	Laser Photometer-TSI Sidepak AM520 (TSI Incorporated, Shoreview, MN, USA)	PM2.5 concentrations in the UVS are lower in summer than in the adjacent road environment but higher in winter. Meteorological factors, green structure, viaduct structure, and surrounding built environment have a significant impact on PM2.5 concentrations.	Yin et al., 2021 [28]

Table 1. Summary of research findings on PM_2.5 concentration in road space.

Research Object		Monitoring Methods	Instrument	Results	Sources
Road	Highway	Mobile monitoring	Laser Photometer-TSI Sidepak AM510 (TSI Incorporated, Shoreview, MN, USA)	Within 200 m, PM2.5 concentrations usually increase closer to the highway. PM2.5 increases with the onset of peak hours and is strongly influenced by temperature changes in colder months.	Patton et al., 2014 [32]
			Tufts Air Pollution Monitoring Laboratory (Tufts University, Boston, MA, USA)	PM2.5 levels were highest in winter and lowest in summer and fall, higher on weekdays and Saturdays compared with Sundays, and higher during morning rush hour compared with later in the day.	Padró-Martínez et al., 2012 [41]
		Fixed monitoring	Continuous Dichotomous Ambient Air Monitor (Thermo Fisher Scientific, Waltham, MA, USA); 1405-DF Tapered Element Oscillating Microbalance (Thermo Fisher Scientific, Waltham, MA, USA)	The reduction in urban traffic helps to reduce the concentration of particulate matter pollution.	Azhari et al., 2021 [42]
			R&P Sequential Air Samplers (Thermo Fisher Scientific, Waltham, MA, USA)	Particulate concentrations decay exponentially with increasing distance within 100–150 m from the road and return to background levels within a few hundred meters from the road.	Clements et al., 2009 [33]
			Continuous Dichotomous Ambient Air Monitor (Thermo Fisher Scientific, Waltham, MA, USA); 1400AB Tapered Element Oscillating Microbalance (Thermo Fisher Scientific, Waltham, MA, USA)	Wind speed, prevailing wind direction, daily cycle of the atmospheric boundary layer, and traffic density are the main influencing factors.	Charron & Harrison, 2005 [43]
	On-Road and Roadside	Fixed monitoring	Continuous Dichotomous Ambient Air Monitor (Thermo Fisher Scientific, Waltham, MA, USA); Tapered Element Oscillating Microbalance (Thermo Fisher Scientific, Waltham, MA, USA)	Higher PM2.5 concentrations were observed in Madrid during the cold season.	Kassomenos et al., 2014 [44]
		Mobile and fixed-site monitoring	Desert Research Institute (DRI, Reno, NV, USA) portable samplers (Desert Research Institute, Reno, NV, USA); Laser DUSTTRAK 8530 Spectrometers (TSI Incorporated) (TSI Incorporated, Shoreview, MN, USA)	PM2.5 concentrations on the road are highly correlated with nearby roadside PM2.5 concentrations in winter and summer. Seasonal effects must be considered, and short-term exceedances of particulate matter concentrations may occur in early spring.	Cheng et al., 2010 [45]; Lozhkina et al., 2016 [46]
		Fixed monitoring	EPA Air Quality System (AQS) (United States Environmental Protection Agency, Washington, DC, USA)	There was an overall decreasing trend in PM2.5 concentrations with increasing distance from the roadside. The incremental PM2.5 decreased by 75% between 5 m and 30 m from the road.	Mukherjee et al., 2020 [47]
	Intersection	Mobile monitoring (high-flow gravimetric personal samplers)	High-Flow Personal Sampler (HFPS) (SKC Inc., Eighty Four, PA, USA)	PM2.5 concentrations are higher in the morning, decrease at noon, rise in the early afternoon, and are even higher in the evening. This phenomenon is related to traffic activity in the surrounding area.	Kaur et al., 2005 [34]
		Fixed monitoring	Laser Photometer-TSI SidePak AM510 (TSI Incorporated, Shoreview, USA); Portable Air Sampling Instruments (BGI, Boston, MA, USA)	Roadside concentrations were higher than those simultaneously monitored at urban sites away from the road. Average particulate matter concentrations at downwind roadside stations are higher than at background stations. PM2.5 concentrations during peak traffic hours are higher than during off-peak hours.	Wang et al., 2018 [37]; Kinney et al., 2011 [48]
	Arterial	Fixed monitoring (near-road continuous air monitoring station [CAMS])	Data from Near-Road Continuous Air Monitoring Station (CAMS) (TCEQ, Austin, TX, USA)	PM2.5 concentrations are usually elevated close to roads. Traffic-related PM2.5 concentrations are usually highest at night, followed by the morning peak, evening peak, and midday, respectively.	Askariyeh et al., 2020 [49]
			Laser Photometer-TSI DRX DustTrak monitor	PM2.5 quality is more related to regional sources and meteorological conditions, with a limited role in traffic volume.	Kendrick et al., 2015 [50]
	Street canyon	Mobile monitoring	Light-Scattering Laser Nephelometer (TSI Incorporated, Shoreview, MN, USA); ES-642, Metone Inc. (Met One Instruments Inc., Grants Pass, OR, USA)	Street canyons have a significant impact on the accumulation of PM2.5 concentrations on Hong Kong’s roads.	Rakowska et al., 2014 [51]
		Fixed monitoring (three heights: 0.20m, 1.0m and 2.60m)	Particle Spectrometer (DMS500) (Cambustion Ltd., Cambridge, UK)	The concentration decreases exponentially with increasing canyon height.	Kumar et al., 2008 [52]
Viaduct	Street canyon	Fixed monitoring	Laser DustTrak (TSI Incorporated, Shoreview, MN, USA)	Particle mass concentrations in street canyons are negatively correlated with height, and particle mass concentrations in street canyons are negatively correlated with temperature.	Feng et al., 2015 [38]
			Portable Laser Aerosol Spectrometer Dust Monitor(Model Grimm 11-A) (GRIMM Aerosol Technik, Ainring, Germany)	Viaducts increase PM2.5 concentrations in street canyons and significantly affect airflow fields.	Zhi et al., 2020 [39]
	UVS	Mobile monitoring	Portable Laser Aerosol Spectrometer Dust Monitor(Model Grimm 11-A) (GRIMM Aerosol Technik, Ainring, Germany)	PM2.5 concentrations decay exponentially near the viaduct without any obstacles.	Li et al., 2021 [40]
		Mobile and fixed-site monitoring	Laser Photometer-TSI Sidepak AM520 (TSI Incorporated, Shoreview, MN, USA)	PM2.5 concentrations in the UVS are lower in summer than in the adjacent road environment but higher in winter. Meteorological factors, green structure, viaduct structure, and surrounding built environment have a significant impact on PM2.5 concentrations.	Yin et al., 2021 [28]

1.4. Review of Monitoring and Analysis Methods

Fixed monitoring and mobile detection are two types of effective methods for measuring air pollution in urban environments. These methods have been tested by numerous scholars and proven successful; however, fixed measurement has limitations, such as the cost of equipment and maintenance for fixed-site air quality monitoring stations (AQMS) (Aclima In-corporated, San Francisco, CA, USA) [53], which can make it challenging to accurately reflect particulate matter distribution over short distances [54]. Mobile monitoring technology can help to overcome these limitations and provide a spatial map of population exposure in the microenvironment with higher spatiotemporal resolution [55]. Specific methods include platforms, such as instrumented cars and vans [53,56]; public transportation, such as trains and trams [55,57]; bicycles [58]; and backpack walking measurements [59,60,61]. In this study, backpack walking measurements were used to directly assess personal exposure to particulate matter in the traffic microenvironment.

Possible methods of analyzing particulate matter include spatial analysis, temporal analysis, and correlation analysis of influencing factors. Spatial analysis can be divided mainly into two categories: differences in particulate matter distribution in different road spaces [45,59,62] and the continuous distribution characteristics of particulate matter in larger spaces, such as cities and districts [59,63], with insufficient attention given to small-scale horizontal space. Temporal analysis usually focuses on the differences in particulate matter concentrations during different seasons, such as winter and summer [46], working and non-working days [51], and different periods of the day [32,34,42]. The correlation analysis of impact factors generally includes meteorological factors, such as temperature and humidity; road factors, such as traffic volume; built environment factors, such as distance from the main road and height-to-width ratio [28].

An increasing number of studies show a growing concern about the combined exposure to particulate matter and its relationship to urban demographics and social aspects [16]. Some studies point to integrated environmental health assessment as one of the biggest challenges for the next decade [64]; therefore, further research in the context of combined exposure to air pollution is needed to improve the understanding of their interconnections in urban environments.

Research on small-scale UVS is scarce. Many current studies focus on statistical analysis, with limited exploration of the spatial layout and potential for built-up space reuse. To address these gaps, the research goals of this paper are as follows: (1) to expand the scope of study and analyze the spatial and temporal distribution characteristics of particulate matter under the viaduct and within a block or so (about 250 m) of the surrounding space; and (2) to enrich the research objectives, investigate the factors affecting particulate matter in the UVS and further propose planning countermeasures for spatial reuse.

2. Data and Methodology

2.1. Study Area

The study area is the UVS of Tianmu Middle Road (31°14′32.7′′ N, 121°27′45.2′′ E), located on the south side of the intersection of Hengtong Road and Gonghexin Road in Jing’an District, Shanghai, China. As an international metropolis with a high population density and elevated roads throughout the city, Shanghai is representative in its exploration of the quality of the urban spatial environment. The city has conducted myriad practices and explorations in the use of the UVS and has issued relevant policies and standards to guide its development and utilization. The study area is currently designed as outdoor basketball courts and other fitness activities space, with high traffic flow in the surroundings and a large number of people exposed to the viaduct environment on a daily basis. The surrounding built environment is diverse; therefore, the area is of great practical importance in evaluating the fine particulate matter environment under the viaduct.

The study area is located in the most prosperous center of Shanghai, on the north side of the intersection of the North–South viaduct and Suzhou Creek (Figure 1). The north side is close to the intersection of arterial roads, the south side is only separated from Suzhou Creek by a branch road, and the east and west sides of the base are urban roads. The viaduct has 12 lanes and heavy traffic, while the ground level has two lanes on either side of the space underneath the viaduct. The sports field is located under the main bridge of the North–South viaduct and consists of four basketball courts, which are sheltered on the east and west sides by the overhanging eaves of the viaduct deck. On the east side of the site, there is a ramp departure point, and the degree of screening is less than on the west side. During weekday peak hours (7:00–9:00 and 17:00–19:00), slow and congested traffic is common, while road traffic is relatively smooth at other times. In summary, the study area was selected for an analysis of the particulate environment assessment of the UVS due to its good regional location, site shape, traffic characteristics, and particulate matter sources. Meanwhile, the UVS at this site has a variety of spatial elements that cover the vast majority of spatial situations encountered by viaducts in urban centers, making it highly generalizable and of significant research interest.

2.2. Data Preparation

2.2.1. Mobile Site Measurements

The mobile monitoring route covered the UVS and surrounding spaces within 250 m on both sides. The starting point of the route was a reference point used to measure background levels, which had no large pollution sources, such as factories or parking lots, within 50 m of it and no more than 3000 vehicles passing through per day [65]. Meanwhile, 32 measurement points were set up along the route in sequence, and peripatetic measurements were performed at each point using handheld instruments. The clocks of all real-time instruments were checked for synchronization prior to measurement. Points 3, 4, and 15–26 were located in the projected space of the viaduct, specifically indicated as the UVS-P. The remaining 18 points were peripheral spaces not covered by the viaduct set up to investigate the attenuation effects of particulate matter, specifically designated as the UVS-NP. A continuous wall interface occurred on the east side of the viaduct, with green space in front of the interface (Figure 2). The selection of this monitoring route was based on the following needs:

(1)

Since the site had different built environments in the north–south and east–west directions, the monitoring points were set up homogeneously in the UVS and in the road space around the viaduct, respectively, to measure distribution characteristics.

(2)

The viaduct had differences in the morphology of its eastern and western branches, so double rows of monitoring points were set up in the UVS to evenly reflect the particulate matter distribution and the various influencing factors in different spatial patterns. It was confirmed in advance that no industrial sources were adjacent to the study area.

2.2.2. Data Collection

Previous studies [28,55,63] recommend conducting environmental monitoring for more than 10 days, including both working and non-working days, to ensure the accuracy of the results. As such, our experiment was conducted over a 12-day period in winter (25 February–8 March 2022) (weather details are in Supplementary Table S1), encompassing both working and non-working days, with four different time segments (8:00, 11:00, 15:00, and 18:00) per day. At each time segment, we walked through each point with our instruments to take measurements. To prevent the backpack walking monitoring experimenters’ fatigue from affecting the results, each experiment was kept to a total duration of less than 30 min [66], and each measurement point was monitored for 20 s [67]. The portable sensors used to monitor environmental elements, such as PM_2.5 concentrations and meteorological parameters, were deployed at a height of approximately 1.5 m above the ground to ensure accurate correspondence with human activities [65].

Considering the specifics of the environment under the viaduct, experimental data collected in this study included geographical data and particulate matter data (PM_2.5 concentrations), and information about the factors that influence particulate matter (temperature and humidity, distance from the viaduct, distance from the river and traffic volume above and below the viaduct). It should be noted that in 2022, the vast majority of motor vehicles in Shanghai were still internal combustion vehicles rather than electric vehicles, which only produce non-exhaust emissions so that the exhaust emission indicator can represent the changes in air pollution caused by traffic volume.

Table 2. Summary of instruments and influencing parameters.

Type	Parameter	Description	Instrument and Model	Sampling Resolution
Geographic Information	GPS positioning data	Location of each data point (m)	Unistrong handheld GPS instrument (Unistrong, Beijing, China)	1 s
Particulate Matter Data	PM2.5 concentration	20-s average data at the sample sites (μg/m3)	PM: SidePak Aerosol Monitor AM520 Air particle test instrument (TSI Incorporated, Shoreview, USA)	1 s
Influence Factor	Traffic volume (TV)	Traffic volume data of the overall environment (up and down the viaduct; vel/h)	Mobile phone recording 5 min of video	-
	Distance to viaduct/Distance to river (DD)	Ratio of distance to the viaduct and distance to river	Unistrong handheld GPS instrument (Unistrong, Beijing, China)	1 s
	Temperature (TP)	Average value of temperature (°C)	HOBO temperature and humidity recorder (Onset Computer Corporation, Bourne, MA, USA)	1 s
	Humidity (HM)	Average value of humidity (%ph)

Note: The SidePak Aerosol Monitor AM520 Air particle test instrument has an expanded uncertainty of ±15% and a coverage factor of k = 2, which is based on typical values for similar portable aerosol monitoring equipment used in air quality studies. These values correspond to a 95% confidence level.

shows the various instruments used for data collection. PM_2.5 concentration data were collected using the PM: SidePak Aerosol Monitor AM520 (TSI Incorporated, Shoreview, MN, USA) air particulate experiment instrument, which is designed to provide accurate mass concentrations of particulate matter at 1 s intervals. To ensure the validity of these data, the PM_2.5 monitor was calibrated at the factory and further validated at an outdoor site in Shanghai prior to the study according to the configured standard method by performing daily zero calibration through the zero-calibration filter to ensure that the instrument reads zero in the absence of aerosols in the field. Location and distance data were collected using Unistrong handheld GPS (Unistrong, Beijing, China) instruments. This study analyzed the relationship between PM_2.5 concentrations and each of the influencing elements and also evaluated the suitability of exercising at the site using all of these data.

2.3. Data Preprocessing and Analysis Methods

2.3.1. Data Preprocessing

This study retained high PM_2.5 concentrations (several hundred μg/m³ or higher) to represent real measurements in time and space [55,68]. During preprocessing, data affected by unusual weather conditions, such as strong winds and heavy rain, were removed and no monitoring was performed during the first 2 min after the instrument was switched on to reduce errors. The corresponding average measurements of PM_2.5 concentrations, temperature, and humidity were then calculated for each monitoring site by taking the arithmetic mean of the 20-s samples. Traffic volume data were obtained by averaging these measured data over a 5-min period at the beginning and end of each measurement to obtain two-way hourly motor vehicle throughput. To address the differing time scales, we directly correlate values by ensuring each factor captures broader trends, such as extrapolating traffic volumes from shorter counts and averaging PM_2.5 over time. This approach does not affect the regression model or the overall correlation results but requires careful interpretation of regression coefficients. GPS data were imported into GIS software for subsequent analysis. The final sample consisted of 1188 geographical coordinates, 1188 PM_2.5 concentration readings, 1188 temperature and humidity readings, 36 traffic volume readings, and 66 readings of the distance from the viaduct or river.

2.3.2. Analysis Method

After obtaining monitoring data, this study conducted descriptive, spatial, temporal, and correlation analyses with influencing factors for PM_2.5 concentration intensity. First, a significance test was performed using SPSS Statistic 25.0 software to determine the overall distribution of PM_2.5 and to assess the presence or absence of attenuation in the surrounding area. The Kriging interpolation technique was used to map the spatial distribution of PM_2.5 and to examine individual exposure within the site [68,69,70,71], which can be expressed as:

$${\mathrm{Z}}^{\mathrm{*}}\left(s_0\right)=\sum _{i=1}^n{\lambda }_i\mathrm{Z}\left(s_i\right)$$

(1)

where ${\mathrm{Z}}^{\mathrm{*}}\left(s_i\right)$ denotes the observation at point I, ${\lambda }_i$ is the unknown weight of the observation at point I; $s_0$ is the point to be estimated, and n is the total number of observations.

The temporal analysis focused on the characteristics of PM_2.5 concentrations during working versus non-working days, as well as peak versus non-peak hours. Finally, a backward multiple stepwise regression model was used to quantify the relative contribution of influencing elements to changes in PM_2.5 concentration intensities in winter, and the method is considered effective in environmental studies [72,73,74]. The prediction model can be expressed as:

$$Y=a+b_1{X}_1+b_2{X}_2+b_3{X}_3+...+b_n{X}_n (n \le 5)$$

(2)

where $Y$ denotes PM_2.5 concentration; a is a constant; $b_1$, $b_2$, …, $b_n$ denote the coefficients of the independent variables, and $X_1$, $X_2$, …, $X_n$ are impact factor parameters.

In the specific comparative and correlations analysis of PM_2.5 concentrations, the relative exposure concentration (REC) was used to reduce the confounding effect of ambient concentration differences [68,75,76]. A positive concentration difference indicates that the PM_2.5 concentration in the UVS exceeds the background concentration and vice versa. The $REC$ is defined as:

$$REC=C_{\mathit{e}\mathit{x}\mathit{p}\mathit{o}\mathit{s}\mathit{u}\mathit{r}\mathit{e}}-C_{\mathbf{b}\mathit{a}\mathit{c}\mathit{k}\mathit{g}\mathit{r}\mathit{o}\mathit{u}\mathit{n}\mathit{d}}$$

(3)

where $C_{\mathit{e}\mathit{x}\mathit{p}\mathit{o}\mathit{s}\mathit{u}\mathit{r}\mathit{e}}$ is the actual concentration measured at the monitoring site, and $C_{\mathit{b}\mathit{a}\mathit{c}\mathit{k}\mathit{g}\mathit{r}\mathit{o}\mathit{u}\mathit{n}\mathit{d}}$ is the background concentration monitored at the background site.

In order to avoid the multicollinearity problem, which should be prevented in regression modeling [77,78], the variance inflation factor ($VIF$) among the predicted structures in the model needs to be determined as a way to account for the multicollinearity diagnosis. In general, a $VIF$ value < 10 indicates the absence of multicollinearity problems. For the selected features, the $VIF$ can be evaluated by performing a linear regression on the features. In this case, the $VIF$ is typically calculated as:

$${VIF}_p={\displaystyle \frac{1}{1-R_p^2}}$$

(4)

where ${VIF}_p$ represents the $VIF$ of feature p and $R_p^2$ is the coefficient of determination when performing a multiple linear regression between feature p and the other features.

3. Results

3.1. Results of PM_2.5 Concentrations in the Under-Viaduct Space

3.1.1. Statistical Analysis of PM_2.5 Concentrations in the Under-Viaduct Space

Figure 3a objectively demonstrates that sample data for PM_2.5 concentrations in the UVS are higher than the overall background values. The background PM_2.5 concentrations are the measured values at the starting point of the route, with a mean value of 86.2 μg/m³. The UVS-P PM_2.5 concentrations are directly influenced by the viaduct, with a mean value of 107.0 μg/m³, while the UVS-NP PM_2.5 concentrations represent the monitored values at peripheral points not covered by the viaduct projection, with a mean value of 99.9 μg/m³. The mean UVS-P PM_2.5 concentration is the highest, approximately 7.1% higher than the mean UVS-NP PM_2.5 concentration and 24.1% higher than the mean background PM_2.5 concentration.

Sample data for particulate matter concentrations in the UVS varied significantly, with a maximum daily mean of 167.1 μg/m³ corresponding to a background concentration of 131.0 μg/m³ and a minimum daily mean of 52.2 μg/m³, corresponding to a background concentration of 46.3 μg/m³. This study found that the background concentration had a “magnifying effect” on the PM_2.5 concentration in the UVS: the higher the background concentration value, the greater the difference between the UVS and the background value. The differences between daily data in Figure 3b are also similar, reflecting a consistent distributional trend, except for 26 and 28 February, when the variance of these data is relatively large. The notable variation in PM_2.5 concentrations in Shanghai is likely influenced by changing meteorological conditions and human activities. Additionally, the UVS may contribute to the accumulation of pollutants, making it more difficult for them to disperse efficiently.

3.1.2. Spatiotemporal Map

At the overall spatial distribution level, estimates of PM_2.5 concentrations at unmonitored sites were calculated via interpolation using GIS software. Figure 4 presents the 12-day map of mean PM_2.5 concentrations, which vary considerably between spatial sites. Differences in the concentrations between monitoring sites are as high as 15.4 μg/m³, accounting for up to 15% of the mean concentration of all sites. A discernible trend of significant attenuation from the viaduct to the east and west over a distance of approximately 210 m along the road exists. The areas of high mean PM_2.5 concentrations occur to the north of the UVS-P, particularly at the intersection with the main road on the north side, with all values above 107 μg/m³. This indicates higher levels of particulate matter exposure and more severe pollution in these areas. Two areas of low PM_2.5 concentrations were identified: one in the green space between the viaduct and the buildings to the east, where a continuous wall of about 2 m in height between the building and the green space sharply reduces the particulate matter concentration to about 99 μg/m³, and the other in the space along the river, with a clear reduction seen, particularly in the section covered by the viaduct, where the mean PM_2.5 concentration drops to 98 μg/m³, a much lower value than in the north. This shows that urban water bodies can reduce PM_2.5 concentrations in nearby areas. Additionally, the map of standard error indicates that the interpolation errors are generally lower than the instrument’s expanded uncertainty, suggesting that the spatial estimation uncertainties introduced by the Kriging method are within acceptable bounds relative to the measurement uncertainty of the PM_2.5 instrument.

At the overall temporal level, Figure 5 shows that PM_2.5 concentrations are generally lower on non-working days than on working days. Overall, PM_2.5 concentrations at different times of day also show regular variations. The peak traffic periods tend to have the highest PM_2.5 concentrations of the day, especially in the evening peak period, while PM_2.5 concentrations are relatively low in the morning peak and off-peak periods. Morning peak values are sometimes even lower than midday off-peak values, presumably because of differences in traffic volume and the deposition of PM_2.5. This phenomenon is more pronounced on working days or when background PM_2.5 concentration is high. The monitoring results are reliable despite the instrument’s extended uncertainty of ±15% and a coverage factor of k = 2. These data are averages of multiple measurements, which helps reduce uncertainty. In addition, it can be seen that most of the time, the difference between UVS-P, UVS-NP, and background at the same time is more than 15%, suggesting that the influence of the instrument’s error is quite small.

On this basis, two typical days (27 February and 4 March) with similar meteorological backgrounds and traffic volumes are selected for a detailed analysis of REC distribution characteristics for each time period between working and non-working days. Figure 6a shows that the overall REC values for the three periods on the typical non-working day are small. Figure 6b indicates that on a typical weekday, the difference in REC between the two peaks and the midday period is greater. The trends of increasing REC toward viaducts and major roads and decreasing REC toward water bodies can be clearly seen in the morning and evening peaks. Two further findings are that (1) the distribution of PM_2.5 concentrations during a single period on a typical non-working day is more homogeneous, and REC is smaller compared with the typical working day; and (2) the range of high PM_2.5 concentrations on the typical working day has a wider impact, almost significantly affecting the area within 240 m of the viaduct.

3.1.3. Results of Influencing Factors

A multiple linear regression model (backward method) was used to investigate the effect of different influencing parameters on the distribution of REC in the UVS during winter. The VIFs of all variables were less than 10, which was within the acceptable range (

Table 3. Regression analysis of influencing parameters. Beta^b = standardized coefficient; PM_2.5_TV = traffic volume; PM_2.5_DD = ratio of distance to viaduct and distance to river; PM_2.5_TP = temperature; PM_2.5_HM = humidity.

Variables	p-Value	Betab	VIF	R2	Total (Adjusted) R2
PM2.5_TV	0.000	1.213	3.356	0.343	0.573
PM2.5_DD	0.000	0.225	1.188	0.228
PM2.5_TP	0.013	−0.298	3.689	0.106
PM2.5_HM	0.041	0.121	1.752	0.143

). All influencing variables have low p-values (<0.05), indicating a statistically significant relationship between the predictor variables and the corresponding PM_2.5 concentrations. The standardization coefficient (beta coefficient) shows the relative contribution of different parameters to changes in REC. Furthermore, the adjusted R-squared of 0.573 indicates that the model has good explanatory power for these data.

REC is sensitive to changes in traffic, the built environment, and meteorological factors, such as temperature and humidity. Among them, traffic volume has the greatest influence on REC, followed by temperature, with beta coefficients of 1.213 and −0.298, respectively. The ratio of distance to the viaduct and distance to the water body is significantly correlated with REC. This suggests that sports venues should be located as close as possible to water and away from viaducts and that people should choose times of high temperature, low humidity, and low traffic volume to exercise.

3.2. Exposure Analysis of Human Activities

For the observation of activities, a combination of manual observation and a panoramic camera was used. In the study area, there are activities such as sightseeing, leisure, and sports, which can be divided into nine categories: scenery viewing (SV), strolling (S), standing and talking (ST), sitting and resting (SR), fitness dancing (FD), jogging (J), playing basketball (PB), exercising (E), and fishing (F). The sample was divided into three age groups: 42% in the older age group (>65 years old), 21% in the adult group (18–64 years old), and 37% in the children and adolescents group (5–17 years old) (Figure 7). The distribution of people’s activities has a certain regularity in time and space.

The mean value of the number of people exercising at all moments on non-working days was significantly higher than on working days, which were 274 and 198, respectively. During the day, sports activities such as FD, PB, and E accounted for 75%, 69% of the total at 8:00 and 18:00, and most of the participants were children, adolescents, and the elderly, while sightseeing and leisure activities such as SV, S, and J accounted for 75% of the total at 15:00, and most of the participants were adults and the elderly. In terms of space, 87% of strenuous activities such as FD, PB, and E took place at UVS because of the basketball court under the viaduct.

Children and older people are the most exposed groups to pollution risks, especially because children have a higher respiratory rate per minute than adults, so they have the highest PM_2.5 deposition during exercise; therefore, this sensitive group should minimize the time of exposure to air pollution. The areas where these sensitive groups conduct their activities coincide with areas of high air pollution, i.e., UVS-P and surrounding areas. Also, the timing of physical activity for most of the population overlaps with the morning and evening peaks, which can greatly increase exposure; therefore, from the point of view of distribution and activities of specific groups of people, the UVS is not suitable for the establishment of sports venues and is not suitable for the activities of groups of people.

4. Discussion

This study collected PM_2.5 concentration data in the UVS in central Shanghai based on mobile monitoring. Based on that, spatial patterns and influencing factors were analyzed. The results showed that (1) PM_2.5 concentration in the UVS is 19.5% higher than background values, with a slow decay from the viaduct to the sides; (2) traffic volume shows a significant positive correlation with PM_2.5 concentration; (3) PM_2.5 is influenced by meteorological elements and the built environment.

This work found and confirmed the spatiotemporal distribution patterns of PM_2.5 concentrations in the UVS. The spatial patterns findings are as follows: (1) PM_2.5 concentrations in the UVS-P are higher than the background case, consistently the highest values in the study area, and the higher the background levels, the greater the difference between the concentrations in the UVS and the surrounding area. Meanwhile, PM_2.5 concentrations have some attenuation distance in the USP-NP, which is consistent with previous studies [28,32,41,79,80], and (2) pollution levels along the river area are significantly lower, and the water bodies contribute to the reduction in PM_2.5 concentrations, which aligns with the study finding that lake wetlands can significantly reduce particulate matter concentrations within 300 m [81]. Furthermore, urban open spaces, green spaces, and roadside continuous walls can partially offset air pollution from viaducts [48,82] and effectively reduce PM_2.5 concentration attenuation distances [28], suggesting that pollution can be optimized and mitigated through appropriate design measures. The temporal patterns findings are as follows: (1) PM_2.5 concentrations in the UVS and its surrounding space are significantly higher on working days than on non-working days, with traffic flow mainly influencing PM_2.5 concentrations [83], and (2) PM_2.5 concentrations in UVS and its surrounding space are essentially higher during morning and evening peak hours than during other times of day. In particular, PM_2.5 concentrations are more pronounced when the background concentration itself is higher, which is consistent with existing studies [34].

This study also found that the enclosed environment of the viaduct significantly increases PM_2.5 concentration intensity in the UVS, which is significantly and positively correlated with traffic volume. Humidity is also positively correlated with PM_2.5. In comparison, a negative correlation exists between temperature and PM_2.5 concentrations. Additionally, the concentration of PM_2.5 is more sensitive to changes in the built environment, as its formation and dispersion are often influenced by urban buildings, floor area ratio, and vegetation [84].

According to this study, we recommend not placing open-air sports fields under highway viaducts or in spaces close to the UVS to prevent PM_2.5 from harming the health of sensitive people. Considering the influence of various factors, exercise spaces around highway viaducts should preferably be close to water bodies, and people should be guided to exercise in places with appropriately high temperatures, low humidity, and low traffic flow to mitigate the adverse effects of PM_2.5. Space for sports under existing highway viaducts can be transformed into enclosed indoor spaces, and mitigation measures can be taken for spaces under highway viaducts where transformation is difficult. First, they can be enhanced by planting a green barrier around the activity space or can be converted into indoor spaces by adding spatial barriers to the road or greenery. Second, the airflow under viaduct spaces can be maximized by using new materials, such as acoustic panels and mitigation measures for the viaducts [55], as well as enhancing the microclimate design of the UVS.

5. Conclusions

This study provides an intuitive understanding of the changes in particulate matter in the UVS microenvironment and reveals spatiotemporal characteristics and influencing factors of particulate matter pollution in a typical UVS in a high-density city through a field experiment with high accuracy. This provides theoretical support for the scientific and effective assessment of the health performance of UVS, as well as the subsequent reduction in air pollutants in high-density urban UVS environments.

It should be noted that this work selected only 12 days in winter to conduct the experiment, which lacks comparability to the summer season or a full year’s worth of data, and the limited timeframe may have impacted the correlation analysis. Second, as the dispersion of airborne particulate matter is influenced by aerodynamic mechanisms, the dispersion of polluting particulate matter in the UVS is the result of a combination of meteorological factors. Although this study avoided the effects of strong winds and rain, it did not specifically consider wind environmental factors, such as speed and direction. In the future, more factors, such as the types of vehicles passing by, should be incorporated.