Spatiotemporal Changes in Air Pollution within the Studied Road Segment

Abstract

Environmental protection is a pivotal element of sustainable development, both essential and indispensable in the construction of smart, green cities. Road transport contributes significantly to atmospheric pollution, accounting for as much as 25% of annual emissions within the European Union (EU). To combat the adverse effects of road transport, the EU has set targets to reduce greenhouse gas emissions from both passenger and commercial vehicles. Consequently, sustainable air pollution management has become a focal point for numerous researchers. This study continues the investigation into the distribution of air pollutants along a specific highway segment in Poland. The article addresses two primary research questions: first, the temporal and spatial variations in air pollution adjacent to a major highway in Poland, and second, whether emission levels exhibit significant annual differences and if there is a correlation between pollutant concentrations and the distance from the roadway. The findings offer valuable data on one of the principal substances polluting the air along EU transportation routes. Moreover, the analysis provides recommendations for future road infrastructure renovation projects and strategies to protect the public from harmful traffic-related pollutants, thereby supporting the development of green cities in accordance with sustainable development principles.

1. Introduction

With the advancement of urbanisation and the economy, the number of vehicles with internal combustion engines has been increasing, representing a significant proportion of the total vehicle fleet. Fuel combustion in these engines results in the emission of harmful pollutants such as particulate matter (PM) (PM₁₀, PM_2.5, PM₁) and gases like NxOy, SO₂, O₃, and benzo(a)pyrene [1]. PM₁₀ and PM_2.5 dust contribute to numerous respiratory diseases, but particularly dangerous are smaller dusts, including PM₁, which, due to their small size, are able to penetrate the human circulatory system. PM pollution generated by road traffic has become a global concern [2]. According to the European Environment Agency (EEA), approximately 25% of total harmful substance emissions originated from transport, with 71.7% of these emissions coming from road transport [3]. Air pollution resulting from road transport is considered carcinogenic and dangerous to our health [4,5]. National air quality monitoring stations typically rely on point measurements; however, visualising the issue on a spatial scale, considering temporal changes, would provide a more accurate depiction of the problem. Therefore, research on the characteristics of traffic-related emissions is crucial to ensure the quality of atmospheric air, both in large cities and in smaller regions. Such characteristics are essential reference points for preparing regional inventories, formulating preventive recommendations, and assessing the scale of environmental contamination impacts [5].

2. Justification of the Topic and Literature Review

Available literature demonstrates that there are methods for studying the dispersion mechanisms of traffic-related pollutants, which are most often based on simulations and field measurements [6]. Simulation-based methods create models linked to vehicle information, such as fuel type, average speed, fuel consumption, mileage, and vehicle age [7,8], as well as external factors like atmospheric conditions, urban conditions (bridges, towers, buildings), and vegetation [9]. Lu and others [10] utilised computational fluid dynamics (CFD) to study the dispersion of PM and observed that smaller particles are more susceptible to factors related to airflow. Hang et al. [11] conducted CFD simulations to investigate the flow and dispersion of traffic pollutants in an urban street canyon. Their research confirms that viaducts can enhance the dispersion of traffic pollutants at high speeds when there is only one source of pollution. An interesting perspective on the spatial dispersion of pollutants is presented by Choe et al. [12], which demonstrated that the distance from the main road, the standard deviation of building floor heights, and the average building floor height are the three most important urban form characteristics influencing the spatial variation of all pollutants. Jin et al. [13] used the COPERT model to inventory greenhouse gases from vehicles in street canyons. On the other hand, Du et al. [14] presented a model based on real data that combines traffic information with pollution concentrations. The authors used hourly data in creating models that link exhaust emissions to conduct temporal assessments and identify emission mitigation factors [15,16]. Maps or profiles are utilised to assess the impact of traffic range on the concentrations of harmful traffic pollutants [9,17]. Although there is a substantial body of work on studying the spatial dispersion of traffic pollutants, there is a lack of studies presenting changes both in time and space through mathematical analysis. Therefore, the aim of this study was to address the following research problems:

• Is there a significant difference in traffic pollutant emissions across the observed seasons (spring, summer, autumn, winter)?
• Are there changes in the concentration of traffic pollutants with increasing distance from the road?

3. Materials, Methods and Results

To address the research questions an unmanned aerial vehicle (UAV) system equipped with the MapAir (Poland, Gliwice) Sensor was used, which is equipped with a laser sensor for measuring the concentration of PM_1.0, PM_2.5, and PM₁₀ particles and semiconductor sensors of other substances from Winsen, Gravity, SGX Sensortech for measuring gaseous pollutants such as SO₂, NxOy, formaldehyde, volatile organic substances (VOCs) and ozone (O₃). The units of measurement depend on the type of pollutant. For PM, the unit of measurement is μg/m³, and the concentration values of other pollutants are measured in ppm. Additionally, the sensor was fitted with a Global Positioning System (GPS) locator, allowing precise determination of the measurement point (latitude, longitude, and altitude). The value of all pollutants detected by the Mapair sensor is modified by a normalization parameter, which causes the pollutants to be presented for the reference atmosphere (288.15 K and 1013.25 hPa), which then allows for direct comparison of pollutants from the sensors regardless of atmospheric conditions. This is done by taking data from the weather station closest to the sensor. This data is used to calculate the normalizing parameter. This parameter is refreshed every 5 min to ensure its accuracy and to maximize the responsiveness of the sensor. The sensor used was calibrated at the Institute of Chemical Processing of Coal in Zabrze. The measurement setup is illustrated in Figure 1.

The study area was located in the immediate vicinity of the important Polish highway A4, which accommodates approximately 200,000 vehicles daily. The area covers approximately 1 hectare. Due to spatial constraints related to the nearby infrastructure, measurements were taken up to 50 m from the source of emissions without attempting to locate the point of pollutant dissipation. Additionally, the selection of this site was decided by the presence of a national environmental monitoring station, which allowed for a comparative analysis of the results. The study area is presented in Figure 2 (yellow area) Figure 3 shows the carpet method of flying the drone and the places of recorded sampling.

Measurements were conducted over twelve months, separately for each season. Preliminary analysis and verification of the results were performed using cartographic mapping in the form of pollution maps. The generated maps, published in [18], allowed for the identification of pollutants with significantly variable concentrations within the measurement area for further analysis. After identifying significant measurement variations, suspended PM₁₀, PM_2.5, and PM₁ were selected for detailed analysis. The following research scheme was adopted:

• Checking the conformity of distributions with the normal distribution using the Lilliefors test.
• Determining whether the concentration distributions in different seasons differ using the Kruskal-Wallis test.
• Using the Wilcoxon test to identify specific concentration distributions that differ from each other.
• Employing Pearson’s correlation coefficient to check for a correlation between traffic pollutant concentration levels and distance from the road.
• Analyzing the significance of the obtained results.
• Drawing conclusions.

The study periods included the four seasons: spring, summer, autumn, and winter. Spring is the research period, where measurements took place in May, and the average temperature was 25 °C. Summer measurements were carried out in July, where the average temperature was 33 °C. Autumn measurements were carried out in November, and the average temperature was 16 °C. Winter measurements were carried out in January for an average temperature of 2 °C. All measurement days were selected so that the weather phenomenon “calm”, that is, a windless day. Selecting such days reduced the impact the impact of wind on the diffusion of pollutants.

3.1. Checking Whether Traffic Pollutant Concentrations Differ by Season

To address the first research question, “Is there a significant difference in traffic pollutant emissions over the course of annual observations?”, the first step was to check the normality of the distribution using the Lilliefors test. The Lilliefors test is a modification of the Kolmogorov-Smirnov test, which examines the distance between the empirical cumulative distribution function (CDF) and the theoretical normal distribution N(μ, σ²). The empirical CDF for an ordered sample x(1) ≤ x(2) ≤ … ≤ x(n) x(1) ≤ x(2) is defined as follows:

$$F_n\left(x\right)=\left\{\begin{array}{cc}0,& forx<x_{\left(1\right)}\\ {\displaystyle \frac{i}{n}},& for{x}_{\left(i\right)}\le x<x_{\left(i+1\right)},1\le i<n\\ 1,& forx\ge x_{\left(n\right)}.\end{array}\right.$$

(1)

The theoretical cumulative distribution function (CDF) of the normal distribution is defined by the formula:

$$F\left(x\right)={\int }_{-\mathrm{\infty }}^x{\displaystyle \frac{1}{\sigma \sqrt{2\pi }}}{e}^{-{\displaystyle \frac{(x-\mu {)}^2}{{2\sigma }^2}}}dx$$

(2)

where $\mu$—average of the population, σ—standard deviation, x—sample.

The test statistic ${\mathrm{D}}_{\mathrm{n}}$ is defined by the equation:

$$D_n=\underset{1\le i\le n}{\max }\left|F_n\left(x_{\left(i\right)}\right)-F\left(x_{\left(i\right)}\right)\right|$$

(3)

In cases where the p-Value is less than the chosen significance level α = 0.05, the null hypothesis should be rejected in favour of the alternative hypothesis. The null hypothesis assumes the normality of the distribution, while the alternative hypothesis suggests a deviation from the normal distribution. The results of the test are presented in

Table 1. Lilliefors Test Results.

	Pollutant	Test Statistic Dn	p-Value
Spring	PM10	$2.95\times {10}^{-1}$	$<2.2\times {10}^{-16}$
	PM2.5	$2.83\times {10}^{-1}$	$<2.2\times {10}^{-16}$
	PM1	$3.08\times {10}^{-1}$	$<2.2\times {10}^{-16}$
Summer	PM10	$9.52\times {10}^{-2}$	$<2.2\times {10}^{-16}$
	PM2.5	$1.18\times {10}^{-1}$	$<2.2\times {10}^{-16}$
	PM1	$1.57\times {10}^{-1}$	$<2.2\times {10}^{-16}$
Autumn	PM10	$9.89\times {10}^{-2}$	$<2.2\times {10}^{-16}$
	PM2.5	$1.22\times {10}^{-1}$	$<2.2\times {10}^{-16}$
	PM1	$1.34\times {10}^{-1}$	$<2.2\times {10}^{-16}$
Winter	PM10	$4.36\times {10}^{-2}$	$1.78\times {10}^{-12}$
	PM2.5	$5.35\times {10}^{-2}$	$<2.2\times {10}^{-16}$
	PM1	$5.18\times {10}^{-2}$	$<2.2\times {10}^{-16}$

.

Since the p-Value for each concentration distribution is less than 0.05, the alternative hypothesis must be accepted, indicating a lack of conformity with the normal distribution. Therefore, the Kruskal-Wallis test, which does not require the assumption of normality, was used in the subsequent analysis [19]. The null hypothesis is ${\mathrm{H}}_0:{\mathrm{H}}_0:{\mathrm{F}}_1\left(\mathrm{x}\right)={\mathrm{F}}_2\left(\mathrm{x}\right)=\dots ={\mathrm{F}}_{\mathrm{k}}\left(\mathrm{x}\right)$—the distributions do not differ significantly, while the alternative hypothesis is H₁—there exist such $\mathrm{i},\mathrm{j}$ that ${\mathrm{F}}_{\mathrm{i}}\left(\mathrm{x}\right)\ne {\mathrm{F}}_{\mathrm{j}}\left(\mathrm{x}\right)$ indicating that the distributions differ significantly. To verify the hypothesis, the data should first be ranked $\left\{{\mathrm{x}}_{11},{\mathrm{x}}_{12},\dots ,{\mathrm{x}}_{1,\mathrm{n}1},{\mathrm{x}}_{21},{\mathrm{x}}_{22},\dots ,{\mathrm{x}}_{2,\mathrm{n}2},{\mathrm{x}}_{31},{\mathrm{x}}_{32},\dots ,{\mathrm{x}}_{\mathrm{k},{\mathrm{n}}_{\mathrm{k}}}\right\}$, where ${\mathrm{R}}_{\mathrm{i}\mathrm{j}}$ is the rank for the element ${\mathrm{x}}_{\mathrm{i}\mathrm{j}}$, $1\le \mathrm{i}\le \mathrm{k},1\le \mathrm{j}\le {\mathrm{n}}_{\mathrm{i}}$. For each group Xi, we obtained the sequence of ranks ${\mathrm{R}}_{\mathrm{i}}={\{\mathrm{R}}_{\mathrm{i}1}{,\mathrm{R}}_{\mathrm{i}2},\dots ,{\mathrm{R}}_{\mathrm{i}{,\mathrm{n}}_{\mathrm{i}}}\}$, where the average rank for the i-th group is calculated using the formula:

$${\overline{\mathrm{R}}}_{\mathrm{i}}={\displaystyle \frac{1}{{\mathrm{n}}_{\mathrm{i}}}}{\sum }_{\mathrm{j}=1}^{{\mathrm{n}}_{\mathrm{i}}}{\mathrm{R}}_{\mathrm{i}\mathrm{j}}.$$

(4)

The test statistic is a measure of the deviation of the sample mean ranks from ${\displaystyle \frac{\mathrm{n}+1}{2}}$ and is described by the formula:

$$\mathrm{T}={\displaystyle \frac{12}{\mathrm{n}\left(\mathrm{n}+1\right)}}{\sum }_{\mathrm{i}=1}^{\mathrm{k}}{\mathrm{n}}_{\mathrm{i}}{\left({\overline{\mathrm{R}}}_{\mathrm{i}}-{\displaystyle \frac{\mathrm{n}+1}{2}}\right)}^2$$

(5)

The test statistic $\mathrm{T}$ follows a ${\chi }^2$ $\left(\mathrm{k}-1\right)$ distribution with degrees of freedom. The critical region is defined by:

$$\mathrm{W}=[{\chi }_{1-\mathsf{\alpha }}^2\left(\mathrm{k}-1\right),+\mathrm{\infty })$$

(6)

where ${\chi }_{1-\mathsf{\alpha }}^2\left(\mathrm{k}-1\right)$ is the 1 − α-th quantile of the ${\chi }^2$ distribution with $\mathrm{k}-1$ degrees of freedom.

When $\mathrm{T}\in \mathrm{W}$ at the significance level α, the null hypothesis H₀ should be rejected, and the alternative hypothesis H₁—that the distributions in the groups differ significantly—should be accepted. Conversely, when $\mathrm{T}\notin \mathrm{W}$ the null hypothesis should be accepted, indicating that the distributions in the groups do not differ significantly. The results of the Kruskal-Wallis test are presented in

Table 2. Kruskal-Wallis Test Results.

Pollutant	${\chi }^2$	p-Value
PM10	8352.1	$<2.2\times {10}^{-16}$
PM2.5	8317.3	$<2.2\times {10}^{-16}$
PM1	8257.9	$<2.2\times {10}^{-16}$

. The ${\chi }^2$ column is the value of the test statistic, and the p-Value column is the probability level of receiving the test result.

Referring to the test results (Table 2), it can be concluded that the distributions differ significantly depending on the period of the study. To determine which distributions differ from each other, the Wilcoxon test was used. The test is based on the differences between the values of the characteristics from the compared sets. It requires an interval scale and does not have assumptions regarding the distribution of the sample. For $d_i=y_i-x_i$ it is assumed that the differences d_i are independent, where each difference d_i comes from a population with an identical continuous distribution, symmetric about a common median. The null hypothesis tested is that the distributions of concentrations in different seasons are equal. The results of the Wilcoxon test are presented in

Table 3. Wilcoxon Test Results.

		Autumn	Summer	Spring
PM10	Summer	0	0	0
	Spring	0	$3.615\times {10}^{-15}$	0
	Winter	0	0	0
PM2.5	Summer	0	$0$	0
	Spring	0	$8.95\times {10}^{-11}$	0
	Winter	0	0	0
PM1	Summer	0	$0$	0
	Spring	0	$4.716\times {10}^{-7}$	0
	Winter	0	0	0

.

For each of the above comparisons, the null hypothesis should be rejected, indicating that the concentration distributions in different seasons differ significantly from each other. The results of the tests are confirmed by the box-and-whisker plots shown in Figure 4, Figure 5 and Figure 6.

Box plots are shown in Figure 4, Figure 5 and Figure 6. It is a form of graphical presentation of the distribution of a statistical feature. The width of the box corresponds to the value of the quarter range. There is a vertical line inside the rectangle defining the median value. The largest graph for winter shows the largest distribution of data, i.e., in our case, the concentrations of PM indicated. To provide a more detailed presentation of changes in traffic pollutant concentrations throughout the year, a model described by Equation (7) for PM₁₀, (8) for PM_2.5, and (9) for PM₁ was proposed for observation. The presented Equations (7)–(9) is a linear model, and the model parameters correspond to the season. Depending on the parameter/season, the value of a given variable, i.e., dust concentration, increases or decreases.

The trends are illustrated in Figure 7, Figure 8 and Figure 9, where x₁ = summer, x₂ = spring, x₃ = winter

y = 38.684 − 23.4817x₁ − 21.1945x₂ + 282.1549x₃

(7)

y = 31.1638 − 17.8853x₁ − 15.7035x₂ + 230.8787x₃

(8)

y = 22.0161 − 12.1287x₁ − 9.6042x₂ + 135.2486x₃

(9)

In Figure 7, Figure 8 and Figure 9, all measurements for each season are taken into account. There are less than 10,000 in total. The “steps” observed in Figure 4, Figure 5 and Figure 6 are caused by the variability in traffic intensity across different seasons. Emissions were highest in winter, followed by autumn, and lowest in summer. The higher emissions in winter are related to the fact that passenger cars consume more fuel on average during driving [20].

3.2. Determination of the Correlation between Traffic Pollutant Concentrations and the Distance from Their Source

To answer the second research question, “Are there changes in the concentration of traffic pollutants with increasing distance from the road?”, Pearson’s correlation coefficient was used. This coefficient determines the level of linear dependence between variables. A linear dependence means that an increase in one value causes a proportional change (increase or decrease) in the other value. Pearson’s correlation coefficient is defined by the formula [21]:

$$r={\displaystyle \frac{\sum _{i=1}^n{(x}_i-\overline{x})(y_i-\overline{y})}{\sqrt{\sum _{i=1}^n{(x}_i-\overline{x}{)}^2{\sum _{i=1}^n{(y}_i-\overline{y})}^2}}}$$

(10)

where ${\mathrm{x}}_{\mathrm{i}}$ and ${\mathrm{y}}_{\mathrm{i}}$ are the empirical values of the respective variables, while $\overline{\mathrm{x}}$ and $\overline{\mathrm{y}}$ are the mean values of the respective variables. The correlation coefficient ranges from 〈−1;1〉, where 1 indicates a perfect positive correlation, and −1 indicates a perfect negative correlation. Since the correlation coefficient is a statistic, it is necessary to check its statistical significance. When n > 100, it is assumed that the variable follows a normal distribution, and the test statistic $\mathrm{U}$ should be calculated. The test statistic is described by the equation [22]:

$$U={\displaystyle \frac{r}{\sqrt{1-r^2}}}\sqrt{n}.$$

(11)

At the significance level α, the critical region is defined by the formula:

$$W=\left(-\mathrm{\infty },{-u}_{1-{\displaystyle \frac{\alpha }{2}}}\right]\cup \left[u_{1-{\displaystyle \frac{\alpha }{2}}},+\mathrm{\infty }\right)$$

(12)

where ${\mathrm{u}}_{1-\frac{\mathsf{\alpha }}{2}}$ is the $1-\frac{\mathsf{\alpha }}{2}$ quantile of the normal distribution. The working hypothesis being tested is that the correlation between the variables does not differ significantly from zero. The confidence level is set at α = 0.05. If U ∈ W, the null hypothesis should be rejected, and the alternative hypothesis should be accepted.

In this part of the study, we examined whether there is a statistically significant correlation between the concentrations of selected traffic pollutants and the distance from their source. The results of the analysis are presented in

Table 4. Results of the Correlation Coefficient and Test Statistic.

	Pollutant	Correlation Coefficient	Test Statistic Value	p-Value
PM10	Spring	0.0878	3.4665	0.0005
	Summer	0.0188	1.1356	0.2562
	Autumn	0.0409	2.0159	0.0439
	Winter	−0.0859	−4.4006	$1.123\times {10}^{-5}$
PM2.5	Spring	0.1379	5.4772	$5.036\times {10}^{-8}$
	Summer	0.0217	1.3046	0.1921
	Autumn	0.0349	1.7195	0.0856
	Winter	−0.0884	−4.5292	$6.18\times {10}^{-6}$
PM1	Spring	0.1584	6.3101	$3.634\times {10}^{-10}$
	Summer	0.0541	3.2585	0.0011
	Autumn	0.0342	1.6844	0.0922
	Winter	−0.0621	−3.1777	0.0015

.

Based on Table 4 and a confidence level of α = 0.05, it can be concluded that only two results do not show a correlation. These are PM_2.5 and PM₁₀ for the summer period. The remaining results, based on the conducted analysis, show a correlation. When the correlation coefficient is positive, it means that the concentration of harmful substances increases with the distance from the road; when it is negative, the concentration decreases with the distance from the road. The graphical presentation of changes in concentration, depending on the distance from the road for the four seasons, along with the fitted line, is shown in Figure 10, Figure 11 and Figure 12. The red line in the figures (Figure 10, Figure 11 and Figure 12) indicates the fit line.

Only in one of the observed seasons was it indicated that the concentration decreases with increasing distance from the road. This phenomenon can be explained by the elevated humidity in the winter period. Particles near the emission source become saturated with water, reducing their ability to diffuse in the space and causing them to “settle” closer to the source. This behaviour of the particles is normal and is similarly explained in the formation of “smog”. Similar relationships were observed by the authors in [23].

To provide a more accurate depiction and representation of the spatial dispersion of harmful substances, profiles were created for seven distance points, taking into account the average concentration values for each measurement band. The points presented are average concentration values of individual PM at designated distances from the road. The line connecting these points is a trend line, the purpose of which was to show the tendency of changes in the selected section. The profiles were obtained from actual field measurements. The profiles are shown in Figure 13.

The obtained profiles are consistent with the statistical analysis. For spring and autumn, the concentration of harmful substances increased with the distance from the road. Summer showed no correlation, while in winter, the concentration decreased, which aligns with the described phenomenon of particle homogenization. In the summer, there was “chaos” for larger particles such as PM_2.5 and PM₁₀, which is related to the lowest humidity, preventing particles from “settling” and causing significant fluctuations in concentration with increasing distance from the road. PM₁ showed a correlation due to its very small size, which does not hinder its dispersion. Despite the short measurement section of less than 50 m from the road, an increase in concentration with distance was observed for most measurement points. This situation was also expected due to the high traffic volume, which emits additional air currents, causing the dispersion of polluted particles. The 50-m section was also determined by Polish building regulations, which allow residential infrastructure to be situated at this distance from highways.

Future research is planned on another measurement section to compare results and determine the point of pollutant dissipation.

4. Discussion of Results

Comparison of Measurement Results with Those of the Main Inspectorate of Environmental Protection

The selection of the measurement location is associated with the presence of a Main Inspectorate of Environmental Protection (GIOŚ) station, which allows for comparison of the obtained results. Additionally, the city under study (Katowice, Poland) has two GIOŚ stations. Station 1 is located in the immediate vicinity of the studied highway, and Station 2 is located in the city centre. Comparing the results from both stations also allows for the assessment of whether urban pollution has a significant impact on the measurements. The average measurement values from the UAV and GIOŚ Stations 1 and 2 are presented in

Table 5. Average Measurement Results from UAV and GIOŚ Stations.

	Pollutant	UAV Data [μg/m3]	GIOŚ 1 Data [μg/m3]	GIOŚ 2 Data [μg/m3]
PM10	Spring	17.48	29	21.8
	Summer	15.94	15.6	12.3
	Autumn	38.67	84.5	21.1
	Winter	320.5	55.6	20.5
PM2.5	Spring	15.45	15.2	9.2
	Summer	13.24	12.3	6.9
	Autumn	31.16	29.3	15.1
	Winter	262.04	25.0	15.9
PM1	Spring	12.41	Not measured	Not measured
	Summer	9.85
	Autumn	22.01
	Winter	157.26

.

Using the average measurement values from UAV, GIOŚ1, and GIOŚ2, the relative measurement error (error relative to UAV measurements) can be determined as follows:

$$∆x=\left|x-x_0\right|$$

(13)

where x—is the exact value and x₀—is the measured value. The absolute error is defined as:

$$\delta ={\displaystyle \frac{\left|x-x_0\right|}{x}}·100\mathrm{\%}$$

(14)

The results of the measurement error of UAV relative to GIOŚ 1 (Δ1 i Δ2) and GIOŚ 1 relative to GIOŚ 2 (Δ1 i Δ2) are presented in

Table 6. Measurement Error Δ1 and Δ2.

		Δ1	Δ2
PM10	Spring	65.91%	24.83%
	Summer	2.13%	21.15%
	Autumn	118.52%	75.03%
	Winter	82.65%	63.13%
PM2.5	Spring	1.62%	39.47%
	Summer	7.09%	43.90%
	Autumn	5.96%	48.46%
	Winter	90.45%	36.40%

.

Analyzing the results obtained and presented in Table 6, conclusions can be drawn regarding the relatively large measurement error. GIOŚ stations perform measurements with hourly averaging, taking measurements every 10 min at only one fixed measurement point. In contrast, the drone takes measurements every 4 s during a one-hour flight, resulting in a significantly larger number of measurements and, consequently, greater accuracy. Comparing the measurements performed by the GIOŚ stations, the mechanism of operation is the same. Referring to both Table 5 and Table 6, it can be observed that the results from the city centre (GIOŚ 2) are significantly lower than those from GIOŚ 1 station located in the immediate vicinity of the highway. This clearly indicates that the traffic on the highway emits a significant amount of air pollutants, which, as the EU has shown, accounts for as much as 25% of the total annual air pollution emissions [3].

5. Conclusions

Year after year, road transport emits a significant share of traffic pollutants into the atmosphere. Due to the ease of diffusion of gas and particulate matter, this results in a deterioration of air quality globally. Therefore, it is essential to illustrate the scale of the problem resulting from the generation of harmful traffic-related pollutants. The available measurement methods rely on point-based stations, and the obtained results are averaged with low accuracy and precision, depending on a single measurement point rather than surface measurements.

The presented literature review highlights numerous studies on the presentation, mapping, modeling, and profiling of traffic-related pollutants emitted by road traffic. However, few studies are based on the mathematical description of spatiotemporal changes and even fewer are compared with other national measurement methods. The aim of this work stemmed from two research questions: “Is there a significant difference in traffic pollutant emissions over the course of annual observations?” and “Are there changes in traffic pollutant concentrations with increasing distance from the road?” Addressing the first question using available statistical tests, it was observed that emissions were highest in winter and lowest in summer. This phenomenon is justified and related to the higher fuel consumption during driving in winter. Regarding the second research question, both the presented Pearson correlation results and averaged profiles indicate a relationship between concentration values and distance from the road. Only two results did not show a relationship during the same measurement period. This phenomenon is influenced by the air humidity discussed in the main section. A noted regularity in the observations is the negative correlation for all observed pollutants in winter, meaning that concentration decreases with increasing distance, which is also influenced by the air humidity supported by the literature. The conducted research had spatial limitations related to the road infrastructure, but these are justified in the context of Polish building regulations. Future studies are planned to include another measurement section, allowing measurements at a greater distance from the road to find the dissipation point. However, it will not be possible to refer the results to the GIOŚ stations, as in Poland, only one station is located in the immediate vicinity of a highway.

In summary, the UAV platform used in the study, equipped with integrated measurement apparatus, provides a valuable source of information on the spatiotemporal changes of one of the significant sources of air pollution on a global scale. The obtained observations will allow for drawing conclusions regarding the construction of protective screens, similar to those used for other environmental pollutants such as noise and vibrations.