Exposure to PM_2.5 While Walking in the City Center ^†

Abstract

This study investigates personal exposure to fine particulate matter (PM_2.5) during walking commutes in Gliwice, Poland—a city characterized by elevated levels of air pollution. Data from a low-cost air quality sensor were compared with a municipal monitoring station and the Silesian University of Technology laboratory. PM_2.5 concentrations recorded by the low-cost sensor (7.3 µg/m³) were lower than those reported by the stationary monitoring sites. The findings suggest that low-cost sensors may offer valuable insights into short-term peaks in PM_2.5 exposure to serve as a practical tool for increasing public awareness of personal exposure risks to protect respiratory health.

1. Introduction

Physical activity is essential in preventing noncommunicable diseases in developed countries [1]. Compared to recreational physical activities, which require an individual to allocate time, utilitarian walking during commuting specifically enables physical activities with daily and necessary routines. Almost all commuter walking falls into the ‘moderate intensity’ category of exercise, with an energy expenditure from 3 to 6 kcal/kg/hour, on average 4 MET (metabolic equivalent of task) [2]. According to WHO, the average walking distance is 1.3 km/trip, and the reference duration is 168 min per person per week [3]. A model example is Sweden, where 19.2% of the working population meets the WHO’s global health recommendations, including moderate- to vigorous-intensity physical activity performed at least 30–60 min per day and more than 150 min of moderate-intensity activity per week [4]. People often walk short distances in large cities or combine walking with public transportation. Due to the proximity to traffic, pedestrians in large cities can be exposed to relatively high concentrations of air pollutants. The pollutant mixture can vary considerably and decrease sharply in concentration within short distances moving away from traffic [5]. Traditional methods of PM_2.5 measurements are costly and time-consuming, so there is a growing interest in low-cost air quality monitoring sensors, especially in public areas where people could be exposed [6].

This study presents the results of PM_2.5 exposure while walking short distances while commuting to work by using a low-cost sensor compared to stationary air quality monitoring stations located in the city of Gliwice, situated in the most polluted region of Poland. The principal aim is to address the data gaps in real-time PM_2.5 exposure to facilitate a comprehensive health risk assessment (HRA).

2. Methods

2.1. Study Area and Design

The study was conducted between January and November 2022 in Gliwice, which is recognized as a well-known European city with a high PM_2.5 concentration, ranking 8th among the 15 most polluted cities in the EU [7]. The study was designed to assess the real-time exposure of two participants to PM_2.5 while walking to the university campus. Participants were not assigned a specific walking route to the university but rather followed paths of their own choosing. Consequently, the environmental conditions along their routes may have varied considerably, potentially including areas with heavy traffic and construction activities in the city center, as well as limited green spaces. This reflects real-world variability in exposure and may have influenced the observed differences in personal PM_2.5 concentrations. The measurements were performed by a low-cost Flow 2 personal sensor (Plume Labs, Paris, France), which was attached to the strap of a backpack worn on the participant’s back. This positioning placed the device outside the direct breathing zone, which may have resulted in a slight underestimation of personal PM_2.5 exposure, particularly in situations involving wind direction or vicinity to local emission sources. Furthermore, limited airflow around the sensor due to its proximity to the body and partial obstruction by clothing or the backpack may have reduced its responsiveness to short-term fluctuations in pollutant concentrations. The device has a light-scattering laser counter that measures particulate concentrations and a metal oxide sensor that measures gas concentrations. The time interval is fixed at 60 s. An application on the phone enables data recording as graphs and levels that can be uploaded. According to the manufacturer, the correlation coefficient for PM_2.5 with the reference device AeroTrak 9306 Handheld Particle Counter was 87.3% to 97.0% (average 90.9%) [8]. Crnosija et al. [9] proved in their study that a Flow air quality sensor can be a helpful instrument for monitoring air quality, particularly in higher air pollution concentrations. Linear regression (R²) between flow sensors and the reference sensor Plantower A003 for PM_2.5 was 0.76.

Additionally, during the study, data from the nearest air quality monitoring station in Gliwice and the air quality laboratory at the Silesian University of Technology (SUT) campus were collected. The air quality monitoring station is an urban background station set up by the Polish Chief Inspectorate of Environmental Protection [10]. The SUT laboratory had a BAM1020 monitor for PM_2.5 measurements (MetOne Instruments, Grants Pass, OR, USA; accuracy exceeds the U.S. Environmental Protection Agency Class III PM_2.5 Forum on Environmental Measurements standards). Weather condition measurements, including temperature, relative humidity, pressure, wind speed, and direction, were measured by WS500 (Lufft, Fellbach, Germany, accuracy: ±0.2 °C, ±0.2% RH ± 0.5 hPa). Before the measurement series, the analyzers in the mobile laboratory were calibrated using certified standard gases.

2.2. Data Analysis and Statistical Treatment

Based on each participant’s time and travel patterns, the speed ranges for each route mode were determined. In order to separate the measurement data into paths according to the recorded location, an algorithm was created in a Python version 3 application for walking at a speed < 10 km/h. The algorithm validated the continuity of the measurements with the geographical location within 5 min forward and backward of the time under study. The initial time was set to 5 min from the surveyed time, and then the algorithm looked to see if another measurement occurred within 10 min. If a location has been recorded in the data, the algorithm writes the location to the track and checks the following location. If no measurement has occurred within 10 min, the entire path with stored locations is saved in the database. Measurements of pollutant concentrations recorded by the flow sensor were then assigned to each path, and measurement data from the monitoring station and the mobile laboratory were collected for the same periods.

Data on pollutant concentration route analyses, including mean, maximum, and minimum values and standard deviation, were calculated using Python version 3. At the same time, statistical tests were carried out with package Statistica 13.1 (StatSoft) at the significance level p < 0.05, and results were presented by using MS Office 365, 2019 (Microsoft Corporation, King County, WA, USA), and the R programming language. No correction related to relative humidity was applied to the Flow 2 data. Additionally, outliers were not removed during data preprocessing and are presented in Figure 1 to illustrate the full range and variability of the recorded PM_2.5 concentrations.

The similarity of the distribution of the analyzed variables for the heating and non-heating seasons to the normal distribution was examined using the Shapiro–Wilk test, which indicated a significant deviation from normality in both cases. Therefore, the non-parametric Mann–Whitney U test, which does not require the assumption of normality in data distribution, was employed for further comparative analyses of pollutant concentrations between-subject effects.

3. Results and Discussion

Table 1. General information on walking during the heating and non-heating season, including number, time, distance of trips, and environmental conditions [11].

Parameter	Heating Season	Non-Heating Season
Investigated routes, N	42	74
Speed, km/h	4.1 ± 2.0	4.5 ± 2.6
Commuting time, min	18.9 ± 13.7	20.7 ± 11.6
Distance, km	0.2 ± 0.3	0.3 ± 0.5
Temperature, °C	10.3 ± 5.2	17.0 ± 5.2
Relative humidity, %	56.9 ± 21.2	67.9 ± 30.8
Wind speed, m/s	1.8 ± 0.9	1.2 ± 0.8
Atmospheric pressure, hPa	998.1 ± 7.3	976.2 ± 9.1

reports general information on the monitoring campaign during the heating and non-heating seasons, including trip time (min) and distance for each mode of transport and environmental conditions. Data presents average value ± standard deviation.

During the non-heating season, average temperatures were about 17 °C, while during the heating season, the temperatures dropped considerably (around 10 °C), reflecting cooler conditions typical of winter. Atmospheric pressures during the heating season are higher in colder weather. Relative humidity during the non-heating season dropped from 68% to about 57% during the heating season. The wind speed was lower during the non-heating season and increased noticeably (up to about 1.8 m/s) during the heating season.

Figure 1 presents average concentrations of PM_2.5 measured by individual low-cost sensors, the SUT laboratory, and the air monitoring station. Both stationary measurements show a statistically significant difference between the heating and the non-heating season (p < 0.05). At the same time, the low-cost sensor did not reveal significant differences.

Figure 1. Box plots of PM_2.5 exposure concentrations modes during non-heating and heating seasons. Boxes represent the 25th to 75th percentile, the central dark line is the median, bars outside the box are the min-max, the cross is the mean value, and the circles are the outliers.

Figure 1. Box plots of PM_2.5 exposure concentrations modes during non-heating and heating seasons. Boxes represent the 25th to 75th percentile, the central dark line is the median, bars outside the box are the min-max, the cross is the mean value, and the circles are the outliers.

The average concentration during the non-heating season by low-cost sensor was 7.3 µg/m³ (median 2.3 µg/m³) and ranged from 2.0 to 95.4 µg/m³ indicating greater variability in comparison to the heating season, where the average is 7.3 µg/m³ (median 2.6 µg/m³) and the distribution of values from 2.0 to 44.7 µg/m³ possibly due to weather conditions, local sources, or movement patterns of the individual. The concentrations were significantly lower than those recorded at the stationary sites. At the monitoring station during the heating season and non-heating season, the concentrations were 29.1 ± 11.5 µg/m³ and 16.0 ± 12.2 µg/m³, respectively. During both seasons, WHO’s recommended 24-h exposure limit of 15 µg/m³ [12] was exceeded. At the SUT lab, PM_2.5 average concentrations were 20.1 ± 17.3 µg/m³ and 12.3 ± 8.2 µg/m³ during the heating and non-heating seasons, respectively. They should show better air quality and not exceed WHO guidelines during the non-heating season.

Comparisons of PM_2.5 exposure differences across studies are generally difficult due to a range of different designs, including differences in distance and duration of trips, the characteristics of the study site, and the use of air quality monitors. Our results focus on short walks, which agrees with the results of Bongiorno et al. [13], showing that most walking trips last from 4 to 17 min and cover distances from 0.3 to 0.9 km. To date, quite a few studies on pedestrian exposure have also been conducted. For example, in Asian cities, the average exposure to PM_2.5 during walking was from 27 to 278 µg/m³ [14,15,16,17,18,19,20]; in America, from 12 to 42 µg/m³ [21,22]; and in Europe, from 10 to 78 μg/m³ [23,24]. Our results are in the lower range of results, which may be because the participants were not required to walk along the roadway. They moved along sidewalks, including urban green areas. Additionally, positioning the low-cost sensor outside the direct breathing zone could partially explain the discrepancies observed between personal exposure measurements and data recorded by air monitoring stations. As the latter reflect average ambient concentrations at a fixed height and location, without accounting for microenvironmental conditions, physical activity, or the specific position of the sensor relative to the human breathing zone. Personal monitors, especially when not worn in the immediate breathing zone, may capture lower or delayed peaks in pollutant levels compared to ambient stations, particularly in highly dynamic urban environments where exposure can vary significantly over short distances and timeframes.

4. Conclusions

We evaluated exposure to PM_2.5 during the heating and non-heating season by two methods, i.e., a low-cost sensor and two stationary measurements: an urban background monitoring station and a laboratory located on the university campus. Statistical analysis revealed that the low-cost sensor did not show significant seasonal differences in measured concentrations, whereas both stationary reference measurements indicated statistically significant variations between the seasons (p < 0.05). Specifically, average PM_2.5 concentrations recorded by the low-cost sensor were identical in both seasons (7.3 µg/m³), with median values of 2.6 µg/m³ during the heating season and 2.3 µg/m³ during the non-heating season. In contrast, average PM_2.5 concentrations recorded at the monitoring station were 29.1 ± 11.5 µg/m³ (heating) and 16.0 ± 12.2 µg/m³ (non-heating), and at the university campus laboratory: 20.1 ± 17.3 µg/m³ and 12.3 ± 8.2 µg/m³, respectively. These values show clear seasonal variation and often exceeded the WHO 24-h exposure guideline of 15 µg/m³. These findings may imply that while the stationary reference measurements at the university campus and the urban background station recorded notable seasonal variations, the low-cost sensor might not have been sensitive enough to identify seasonal variations in PM_2.5 concentrations. Low-cost sensors often have lower accuracy, sensitivity, and stability than reference instruments. They may struggle to detect subtle seasonal variations due to humidity interference, calibration drift, or limited detection capabilities. While fixed-site monitors provide valuable background information, they may not accurately reflect personal exposure during dynamic activities such as walking. Variability in personal exposure, especially during the non-heating season (ranging from 2.0 to 95.4 µg/m³), underscores the influence of local sources, microenvironments, and individual movement patterns. The relatively low values observed in this study are likely due to sensor placement (on a backpack, outside the breathing zone), which may have contributed to slight underestimations of actual inhaled exposure and the route characteristics. The walking routes led from two different parts of the city to the university campus, which served as the common destination point. The campus itself is situated in close proximity to the city center but does not fall within its most central, densely urbanized zone. Although sections of the routes may have passed through areas classified as part of the city center, which has a more complex pollution environment with traffic emissions, construction dust, and other sources that could have short-term peaks and mask seasonal trends, the primary area of interest in terms of exposure assessment was the university campus and its immediate surroundings. In contrast, the background and university stations might better reflect regional pollution patterns, making seasonal changes more detectable. However, both may be more influenced by heating-related emissions during the winter, leading to clearer seasonal patterns.

Summing up, our findings highlight the importance of selecting appropriate measurement methods for air quality assessments. If the goal is to track seasonal changes, reference-grade instruments at well-chosen locations are more reliable than low-cost sensors, which may still be valuable for general trends but require careful validation. However, low-cost sensors play an increasingly important role in individual air quality monitoring, offering real-time, localized insights into personal exposure to pollutants like PM_2.5. Their significance can be outlined particularly in raising awareness about air pollution.

However, the study has limitations that must be acknowledged. The dataset is based on a small number of participants (N = 2), which limits the generalizability of the findings. Moreover, the accuracy of the low-cost sensor may be influenced by environmental factors such as humidity, and no correction was applied for this variable. Future research involving a larger and more diverse sample, standardized sensor placement closer to the breathing zone, and calibration or correction for environmental conditions is recommended to further validate and expand on these results.