Selected Land Cover Factors as a Determinant of Differences in Particulate Matter Concentrations – A Case Study of Warsaw, Poland

The research was conducted in the area of Warsaw, Poland. Warsaw is the capital of Poland, a city with an area of 517 km² at 78–121 m ASL located at the Vistula river banks (GUS, 2024). The main sources of PM₁₀ emissions in Warsaw are individual heating, road traffic and industrial and construction activities.

Individual heating, especially during the heating season, is a significant source of PM₁₀ emissions, mainly due to the fact that not all of the city is connected to the district heating network and the use of solid fuels in domestic stoves and boilers is often supplemented by illegal burning of waste in them (Badyda et al., 2020). Several important national roads with highway or expressway standards run through Warsaw. Especially, Warsaw is crossed by two east-west routes, the only highway connections in the country, which cross the largest river in Poland, the Vistula, making the Grot Rowecki Bridge the section of road with the highest traffic intensity in the country (GDDKiA, 2021), and the this traffic intensity has a significant impact on PM₁₀ concentrations (Bihałowicz et al., 2023). As Poland is a developed but still developing country (Miłek, 2018; Raszka et al., 2021), there is also a lot of investment going on in Warsaw, which means that industrial processes and intensive construction work increase PM₁₀ concentrations in the air.

The meteorological data were downloaded from the Institute of Meteorology and Water Management National Research Institute Public Data (IMGW-PIB, 2024). The data used in the study were obtained at the Airport Meteorological Station, Warsaw Chopin Airport (World Meteorological Organization code 12375). The station provides comprehensive data regarding the climate and meteorological conditions in Warsaw due to the high intensity of air traffic, the list of measured parameters available in this database is provided in Tab. 1. In the study all parameters were used.

In Poland, air quality monitoring is carried out by the Chief Inspectorate for Environmental Protection (GIOŚ) through a network of reference stations, which form the foundation of the atmospheric assessment system. These stations are equipped with approved analysers, enabling continuous and precise measurements of concentrations of air pollutants. In Warsaw there are 5 stations measuring PM₁₀ concentrations, whose locations are provided in Fig. 2. The brief summary of data used in the study for each air quality stations is provided in Tab. 2 while the monthly average concentrations are for each station are provided in Fig. 1.

Figure 1.

The monthly average concentration of PM₁₀ at five air quality stations in Warsaw in 2022

Figure 2

Location of meteorological station, air quality stations (GIOŚ, 2023) on the background of land cover (GUGiK, 2025) within borders of Warsaw (Główny Urząd Geodezji i Kartografii, 2022) created using QGIS (QGIS Development Team, 2021)

It is noteworthy that all sites demonstrate a marked increase in PM₁₀ levels in March, which may be indicative of a transient event or phenomenon, such as dust intrusion or increased urban activities. This is followed by a significant decrease in PM₁₀ levels during the subsequent spring months. In contrast, winter levels remain moderate but tend to increase slightly in December, potentially reflecting increased heating emissions amid stagnant air conditions. Summer months generally exhibit lower PM₁₀ concentrations, particularly in July, likely due to enhanced atmospheric dispersion, although some areas register an increase in August. Finally, autumn is characterised by a gradual rise in PM₁₀ levels, reflecting the influence of reduced dispersion and possibly intensified residential emissions as temperatures drop. Seasonal variations in meteorological conditions, including temperature, precipitation, and wind patterns, have been demonstrated to exert a substantial influence on the concentrations of particulate matter (PM₁₀) in urban environments (Rogula-Kozłowska et al., 2014; Sówka et al., 2018). For instance, during the colder months (e.g., January, February, and December), increased heating demands and reduced dispersion of pollutants have been shown to result in elevated PM₁₀ levels, as evidenced by the elevated concentrations observed in Fig. 1. Conversely, the warmer months (e.g., June, July, and August) may experience lower PM₁₀ concentrations due to increased atmospheric mixing and precipitation, which can effectively remove particulates from the air.

The Polish Space Agency (POLSA), on demand of Poland’s Head Office of Geodesy and Cartography, has developed the POLSA land cover (POLSA LC) data set. This detailed 10-meter-by-10-meter raster, delineated in the ETRF2000-PL/CS92 coordinate system (EPSG: 2180), encompasses the entirety of Poland's territory, meticulously partitioned into 71,520 columns and 69,759 rows (GUGiK, 2025; POLSA, 2022). The land cover is determined through an automated classification process using multi-temporal Sentinel-2 imagery, random forest classifiers, and the S2GLC PL2020 methodology (Malinowski et al., 2020). A notable feature of the POLSA LC data is its yearly updates, with land cover information available for 2019, 2020, and 2021. This high temporal resolution facilitates the identification of rapid land cover changes, forest and water body/wetland area modifications, and the assessment of expanding anthropogenic zones that increase surface runoff.

The generation of POLSA LC data entails the classification of multi-temporal Sentinel-2 satellite images from the growing season (April to mid-October) with a cloud cover limit of 50%. The process entails the selection of 10 to 20 images per scene, distributed uniformly throughout the growing season, and the implementation of atmospheric correction using the Sen2Cor algorithm prior to classification by the Random Forest method (POLSA, 2024). The resulting classification data undergoes validation using a minimum of 15% of scenes within Poland's borders and a minimum of 30% of the total area analysed. A minimum of 10,000 samples are utilized in both the semantic and geometric layers of the map, ensuring an overall accuracy of at least 80% and 70% for individual classes, respectively. The land cover maps for the years 2019–2023 are accessible on the Portal, displaying the original 10-meter spatial resolution of the source imagery, categorized into 10 classes: Artificial surfaces, Cultivated areas, Broadleaf tree cover, Coniferous tree cover, Herbaceous vegetation, Moors and Heathland, Marshes, Peatbogs, Natural material surfaces, and Water bodies. The land cover of Warsaw is provided in Fig. 2 while the land cover in 1-km from air quality stations is provided in Tab. 3.

The analyses provided in this study were basing on the two main programs QGIS (QGIS Development Team, 2021) and Gretl (GRETL, 2024). Spatial analyses were performed in QGIS. Firstly the data about the land cover in study area was extracted to from the WCS service provided by Head Office of Geodesy and Cartography (GUGiK, 2025) to the local raster file. The locations of air quality stations were obtained from Chief Inspectorate for Environmental Protection (GIOŚ, 2023) while location of meteorological station was determined by satellite images of Warsaw Chopin Airport (GUGiK, 2022). For the layer containing air quality stations’ locations a zonal histogram was performed in QGIS attributing to each location the area of given land cover withing 1 km from the station. Hence, the land cover structure was assigned to each location, constant through the year. The air quality data were 5 timeseries of 365 days. The timeseries were stacked into one table creating 1825 rows table. The table was joined on the location with landcover structure table to each location a land cover structure was assigned.

The model for explaining the variability of PM₁₀ was developed based on meteorological and land cover data. Initially, the model was developed for all independent variables, i.e. parameters mentioned in Tab. 1 and Tab. 3. First, collinear and constant variables were excluded from the subsequent analysis. The GRETL software provided the statistical significance of all coefficients of the linear model. The next step was to drop variables with high p values, i.e. greater than 0.05. The parameters of the obtained model are provided in Tab. 4. The model was subjected to testing for the exclusion of “duration of fog” and “duration of wind >10 m/s”, as these parameters showed the lowest statistical significance. However, the elimination of these parameters from the model resulted in an increase in both the Akaike Information Criterion and the Bayesian Information Criterion, as well as a decrease in the R² value. The R² value of the model presented in Tab. 4 is 0.850.

As this is a simple linear model, the interpretation of the coefficients' values is straightforward. The coefficient with maximum daily temperature is 1.321±0.076 µg/m³·°C). It has thus been suggested that elevated temperatures may have a significant impact on the release and later lift of dust particles into the atmosphere. Maximum diurnal temperatures have been shown to have a considerable effect on PM₁₀ atmospheric concentrations, due to a number of complex physical and chemical processes. This observation is in agreement with previous works (Al-Hemoud et al., 2021). The following mechanisms can be plausible causes of increased concentrations of this particular pollutant in response to higher temperatures:

Surface drying and dust uplift. Elevated temperatures have been shown to cause moisture to evaporate rapidly from surfaces such as roads, fields and other areas. This phenomenon leads to an increased propensity for dust particles to be released into the atmosphere. The creation of dry conditions has been shown to result in an enhancement of dust particles present in the atmosphere, a phenomenon that is particularly evident in rural and urban areas with high traffic density. The presence of additional factors, such as construction or other types of human activity, can lead to a substantial increase in dust concentrations (Bukowski & Van Den Heever, 2022; Garcia-Carreras et al., 2021; Kim et al., 2013; Qi et al., 2023; Raupach & Lu, 2004).

Thermal inversion. It occurs when a warmer layer of air is positioned above a colder layer in close proximity to the ground. This results in the entrapment of pollutants, including PM₁₀, within the lower atmospheric layer due to their inability to rise and disperse. Such conditions are more prevalent in the early morning or late evening hours on warmer days. Cumulative pollution leads to localized increases in PM₁₀ concentrations, which can result in significant health concerns, particularly in densely populated areas (Lagmiri & Dahech, 2024; Nejad et al., 2023; Staehle et al., 2022; Yavuz, 2024).

Chemical reactions favouring the formation of secondary particulate matter. Secondary particulate matter is formed through chemical reactions that are favoured by higher temperatures. Nitrogen oxides and ammonia, which are by-products of various industrial and transport processes, react with heat and solar radiation to form secondary PM₁₀ particles. This process has the potential to increase the overall concentration of particulate matter in the air, and to alter the chemical composition of PM₁₀, which can have additional health consequences as different types of particles have different toxic properties (Deng et al., 2021; Thomsen et al., 2024).

Increased human activity at higher temperatures. Warmer days are often conducive to increased industrial and construction activity and an increase in traffic. This results in higher emissions of PM₁₀ particles, which are released into the atmosphere from various sources. In urban areas, where the concentration of industrial activity and traffic is high, elevated temperatures can precipitate severe smog and air pollution, engendering health risks for residents, particularly those afflicted with respiratory diseases (US EPA, 1995).

A 1°C higher minimum temperature results in a decrease in PM₁₀ concentrations of approximately 1.894±0.095 µg/m³. This may mean that warmer nights may favour lower atmospheric dust concentrations.

Lower emissions from local sources. On warmer nights, especially in cities, there may be less traffic and industrial activity. Reduced transport activity and smaller industrial operations lead to lower atmospheric dust emissions. The combination of reduced generation of pollutants and enhanced dispersion can lead to a decline in PM₁₀ concentrations. During warmer periods, particularly when temperatures do not decline significantly, residents can opt to reduce the use of heating appliances. This reduced burning leads to a decrease in emissions into the atmosphere (Jandacka et al., 2024; Salva et al., 2023; Senyel Kurkcuoglu & Zengin, 2021).

The urban heat island effect. In urbanized areas, higher night-time temperatures may be a result of the heat island effect, where cities retain heat more effectively than surrounding rural areas. This effect can lead to better air mixing, which in turn can promote a reduction in pollutant concentrations (Rao, 2014; Wang et al., 2022; G. Yang et al., 2021).

Rainfall has been shown to have a significant impact on air quality, with each hour of rainfall resulting in a decrease of 0.695 ± 0.092 µg/m³·h in the 24-hour average PM₁₀ concentrations. This phenomenon can be attributed to the action of rainfall, which serves as an air purification mechanism by removing particulate matter from the atmosphere. It has been demonstrated that each hour of snowfall leads to a decrease in PM₁₀ concentrations of approximately 1.08±0.14 µg/m³. In a manner analogous to rain, snow also contributes to the removal of pollutants from the air. The duration of snowfall exerts a significant impact on atmospheric PM₁₀ concentrations, with numerous benefits for air quality. The following are some of the key phenomena through which prolonged snowfall reduces particulate concentrations: Rainfall and snowfall exerts a substantial influence on air quality. Prolonged rainfall is effective in removing PM₁₀ particles from the atmosphere. Raindrops collect particles from the air and transport them to the ground, a process known as “pollution flushing”. This leads to a significant reduction in the concentration of particulate matter in the air. Rainfall increases the humidity of the air, which promotes the agglomeration of dust particles. High humidity causes dust particles to combine with water to form larger aggregates that fall to the ground. High atmospheric humidity can lead to increased condensation of particles, which results in them becoming heavier and falling to the ground, thus reducing their concentration in the air. During rainfall, traffic and other activities that can generate PM₁₀, such as construction or industrial work, are often reduced, thereby leading to a decrease in the amount of particulate matter emitted into the atmosphere. Longer rainfall events, especially those of moderate intensity, are more effective in removing pollutants from the air because their effects persist over a longer period of time. Repeated rain cycles can lead to a continuous reduction in PM₁₀ concentrations. (Maboa et al., 2022; Olszowski, 2016; Widziewicz et al., 2017; Zhou et al., 2020).

The duration of haze has been shown to have a significant impact on PM₁₀ concentrations, with an increase of approximately 0.55±0.25 µg/m³ observed for every hour of haze. The accumulation of dust is promoted by fog in the lower layers of the atmosphere, leading to an increase in PM₁₀ concentrations of around 0.940±0.059 µg/m³ over a longer period of haze. Haze and fog have been found to cause dust to settle in the atmosphere. The duration of fog/haze has a significant impact on air quality, and its effects on PM₁₀ concentrations can be understood in conjunction with visibility. Moreover, Fog often accompanies calm atmospheric conditions, limiting natural air movement. Low wind speeds make pollutants, including PM₁₀, less likely to disperse, leading to their accumulation in the lower layers of the atmosphere (Majewski et al., 2021, 2024; Maurer et al., 2019; Won et al., 2020).

The decline in PM₁₀ concentrations of approximately 0.59±0.23 µg/m³ for each hour of wind speeds ≥ 10 m/s can be attributed to several key phenomena associated with the effects of intense winds. These winds contribute to the dispersion of PM₁₀ particles in the atmosphere. Due to the high air velocity, the particles are lifted and transported over greater distances, causing the dust concentration in a given area to decrease. The particles are aerated and dispersed, and in the case of strong winds, dust particles can be transported to areas where their concentration is lower (Chalvatzaki et al., 2012; Yang et al., 2024).