Influence of Mixing Layer Height on Air Pollution in the City of Zagreb, Croatia

Highlights

What are the main findings?

• MLH strongly influences PM₁₀ and ∑PAH concentrations measured in Zagreb.
• Approximately 90% of PM₁₀ daily limit exceedances occurred during temperature inversion events, highlighting inversion duration as a key meteorological driver of pollution episodes.

What are the implications of the main findings?

• Atmospheric stability conditions are a key meteorological driver of urban pollution episodes.
• Incorporating MLH into statistical air quality assessment could improve the prediction of pollution events.

Abstract

The mixing layer height (MLH) is a key meteorological parameter governing the dispersion and accumulation of air pollutants in the atmospheric boundary layer. Over the Zagreb urban area, low MLH values are associated with stable atmospheric conditions and temperature inversions that inhibit vertical air mixing and promote pollutant accumulation near the ground. This study investigates the influence of MLH on concentrations of particulate matter with an aerodynamic diameter smaller than 10 µm (PM₁₀) and particle-bound polycyclic aromatic hydrocarbons (PAHs) using a five-year dataset of radiosonde measurements and air quality observations. Daily MLH values at 12:00 UTC were compared with simultaneous measurements of PM₁₀ and PAH concentrations. Inversion conditions (MLH < 1000 m) occurred on 44% of days, with approximately 80% of events recorded during the cold season. Elevated PM₁₀ and PAH concentrations were strongly associated with these stable conditions. Of 69 daily exceedances of the PM₁₀ limit value (50 µg m⁻³), 62 occurred during inversion events. Statistical tests confirmed significantly higher concentrations during stable atmospheric conditions (p < 0.05), with further increases observed during prolonged inversion periods. Empirical relationships between MLH and pollutant concentrations were developed and may support future machine-learning-based air quality forecasting. The results highlight atmospheric stability as a critical factor for urban air quality management.

1. Introduction

The harmful effects of air pollution, especially particulate matter (PM), have been recognized for several decades. Since the 1970s and 1980s, numerous epidemiological studies have demonstrated a strong association between PM exposure and adverse health outcomes, including cardiovascular and respiratory diseases and increased mortality [1,2]. It is now well established that not only PM concentrations but also the chemical composition of particles significantly influences their toxicity and environmental impact [3,4]. PM represents a complex mixture of organic and inorganic compounds, including metals, polycyclic aromatic hydrocarbons (PAHs), organic and elemental carbon, anhydrosugars, anions, cations, volatile organic compounds (VOCs), secondary organic aerosols (SOAs), etc. [5,6,7,8,9]. In the late 1980s, the World Health Organization (WHO) was the first to set air quality limit guidelines, offering health-focused recommendations that were subsequently adopted by the European Union, which established the EU legislation with legally binding air quality standards [10,11]. Currently relevant WHO air quality guidelines from 2021 [12] introduced stricter recommended levels for key air pollutants, as new evidence shows that pollution influences human health even at lower concentrations than previously thought. Consequently, EU legislation was updated in 2024, with the aim of bringing the prescribed air quality standards closer to the latest WHO guidelines, which will apply from 2030 [13]. In addition to the limit values for particulate matter (PM₁₀, PM_2.5), the EU legislation has set limit values for ozone, nitrogen dioxide, sulfur dioxide, carbon monoxide, and benzo(a)pyrene (BaP), as well as several metals (Pb, As, Cd, Ni). According to the Directives (EU) 2008/50/EC and 2004/107/EC, the daily limit of PM₁₀ is set to 50 µg/m³, not to be exceeded on more than 35 days in a year; the annual PM₁₀ limit is set to 40 µg/m³, while one of the main PAHs, BaP, has a target annual value of 1 ng/m³ [10,11]. The updated Directive (EU) 2024/2881 sets stricter PM₁₀ limits (45 μg/m³ daily and 20 μg/m³ annually) and aims to reduce air pollution impacts on health and ecosystems by 2030, in line with the Zero Pollution Action Plan [13].

Air pollution, which is usually more pronounced in urban areas due to higher emission intensities from traffic, residential heating, and population density [14,15,16,17,18], is strongly influenced by meteorological conditions that control pollutant dispersion, dilution, and accumulation [19]. Such changes in pollutant concentrations are closely linked to atmospheric stability, which governs the efficiency of vertical mixing and the removal of pollutants from the near-surface layer [20]. Stable atmospheric conditions are characterized by weak winds and reduced turbulence and are usually associated with temperature inversions [21]. During these inversions, air temperature increases with height, creating a layer that suppresses vertical mixing, limits pollutant dispersion, and leads to their accumulation near the ground [22]. Such conditions typically form during nighttime due to radiative surface cooling, or mostly during winter, when reduced solar heating limits turbulence, allowing them to persist for longer periods, especially under high-pressure systems [23]. To better understand these processes, it is essential to consider the atmospheric boundary layer (ABL), the lowest part of the atmosphere directly influenced by the Earth’s surface [24]. In air pollution studies, the lowest part of the ABL is often referred to as the mixing layer, and its height (mixing layer height, MLH) is a key parameter that controls pollutant dispersion, as it determines the effective vertical volume available for dilution [25]. Since stable conditions during winter coincide with increased emissions from residential heating and traffic, pollutant levels rise, and air quality standards are more often exceeded. The MLH is commonly determined from radiosonde observations, using vertical profiles of temperature, humidity, and wind. One of the most widely used methods is the bulk Richardson number (RiB) approach, which is simple and applicable to both observational and modeling studies [26]. The method defines the mixing height as the altitude where RiB exceeds a critical value, typically 0.25 [27,28]. The RiB at a specific height, z, is calculated using the potential temperature and wind speed at a given level and at the lower boundary of the ABL, typically the surface. Therefore, the gradient Richardson number (R_i) is defined as:

$$R_i={\displaystyle \frac{g}{{\theta }_{ave}}}{\displaystyle \frac{\frac{\partial \theta }{\partial z}}{{\left(\frac{\partial u}{\partial z}\right)}^2+{\left(\frac{\partial v}{\partial z}\right)}^2}}$$

(1)

where z is height, θ_ave is average potential temperature, u is wind speed in the x-axis, v is wind speed in the y-axis, and g is gravitational acceleration. In practical applications using radiosonde data, the Richardson number is calculated using finite differences between discrete measurement levels. The bulk Richardson number is then expressed as:

$$R_{iB}={\displaystyle \frac{g(z_j-z_1)}{{\theta }_{ave}}}{\displaystyle \frac{({\theta }_j-{\theta }_1)}{{\left(u_j-u_1\right)}^2+{\left(v_j-v_1\right)}^2}}$$

(2)

In addition to radiosonde observations, remote sensing techniques such as lidar and ceilometer systems [29,30] are frequently used for continuous monitoring of MLH. These instruments estimate the MLH from vertical aerosol backscatter profiles, where strong gradients in aerosol concentration usually indicate the upper boundary of the mixing layer [25,31]. Microwave radiometers estimate MLH using continuous temperature and humidity profiles derived from atmospheric microwave emissions, while SODAR systems use acoustic wave propagation to detect turbulence structure and atmospheric stability [32,33]. Furthermore, aircraft observations and unmanned aerial systems can be used to directly measure vertical profiles of meteorological variables and pollutants within the boundary layer. In addition, numerical weather prediction and dispersion models simulate MLH evolution based on atmospheric stability, turbulence, and surface energy balance processes [34,35]. However, each method has specific limitations regarding temporal and vertical resolution, spatial representativeness, and performance under stable atmospheric conditions [25].

In Croatia, early investigations of MLH based on radiosonde data, as well as studies of surface-based and elevated temperature inversions, were conducted in Zagreb in the early 1990s by Lončar, who analyzed the period 1977–1982 [36]. However, only a limited number of studies examined the relationship between MLH and air pollution. Prtenjak et al. investigated a severe SO₂ pollution episode in Rijeka, Croatia, a coastal industrial area, by analyzing the roles of ABL structure, meteorological conditions, and emission sources, which showed that strong temperature inversions, very low MLH, weak winds, and local emissions were the key factors that led to extreme pollutant accumulation [37]. Previous studies of PM and PAHs in the region have reported pronounced seasonal variability, with elevated concentrations during the colder part of the year [38,39,40,41,42]. A strong negative correlation with temperature has been observed; commonly used meteorological parameters such as temperature, humidity, pressure, and wind speed alone were insufficient to fully explain the variability in pollutant concentrations, particularly during high-pollution episodes [43,44,45]. To the authors’ knowledge, quantitative relationships between MLH and particulate pollution, especially PM₁₀ and particle-bound PAHs, have not yet been comprehensively studied in Croatia. Moreover, such studies remain relatively scarce even at the European level, highlighting the novelty of the present research.

Zagreb, the capital of Croatia, frequently experiences wintertime air pollution episodes accompanied by fog formation, which reduces visibility and is often associated with unpleasant odors that raise increasing public concern [46]. The city is surrounded by elevated terrain, with mountains exceeding 1000 m in height to the north and southwest, and these local factors are known to influence boundary-layer dynamics and pollutant dispersion [47]. However, the quantitative relationships among atmospheric stability, MLH, and pollutant concentrations in Zagreb remain poorly understood, particularly for particle-bound organic pollutants such as PAHs.

The aim of this study was to examine how MLH influences PM₁₀ and particle-bound PAH concentrations in the urban atmosphere of Zagreb. Particular emphasis was placed on the occurrence and duration of temperature inversion conditions to assess their role in pollutant accumulation and in exceedances of the daily PM₁₀ limit value. To achieve this, a five-year dataset of air quality and radiosonde measurements was analyzed using statistical methods.

2. Materials and Methods

For the purposes of this study, data from the Ksaverska cesta air quality monitoring station were used, as it is one of the stations with the longest and most comprehensive sets of measurements. The station is located in Zagreb City, Croatia, at 45°50′07″ N, 15°58′41″ E, at a ground elevation of 167.2 m above sea level. The geographical position of Zagreb (Croatia), the sampling site, and its topographic map showing terrain elevation are shown in Figure 1. Zagreb is bordered to the north by the Medvednica mountain and to the southwest by the Žumberačko gorje, both exceeding 1000 m in elevation, while most of the city to the east and west lies at relatively low elevations, below 200 m above sea level. The monitoring station is part of the local and national air quality monitoring network and is categorized as an urban background station where PM₁₀ and PAH concentrations are measured continuously.

Figure 1. (a) Geographical position of Zagreb City in Croatia and the sampling site (source: Google Earth); (b) topographic map with elevation scale and terrain (source: www.topographic-map.com, based on OpenStreetMap data. © OpenStreetMap contributors, licensed under the Open Data Commons Open Database License (ODbL)).

PM₁₀ daily samples were collected on Whatman QM-A quartz fiber filters (47 mm diameter) (Tisch Scientific, Cleves, OH, USA). Before and after sampling, each filter was placed in a Petri dish and conditioned in the chamber for 48 h under controlled conditions (20 ± 1 °C; 47.5 ± 2.5% RH). After 24 and 48 h of conditioning, the filters were weighed using Mettler Toledo MX5 and XP6/M microbalances (Greifensee, Switzerland) with a resolution of 10⁻⁶ g, equipped with electrostatic discharge control systems. Sampling was performed using a low-volume sampler (Sven Leckel LVS3, Sven Leckel Ingenieurbüro GmbH, Berlin, Germany) with an airflow rate of 55 m³/day. The sampler flow rate was verified every three months in accordance with EN 12341 [48]. Flow calibration was performed annually, or whenever the flow deviated by more than 5% from the nominal value of 2.3 m³ h⁻¹ during routine checks. The volumetric flow rate was continuously regulated by the pump frequency converter. PM₁₀ mass concentration was determined gravimetrically from the difference between post- and pre-sampling filter masses, in accordance with the standardized reference method EN 12341 [48]. PAH mass concentrations bound to the PM₁₀ fraction were determined by high-performance liquid chromatography with a fluorescence detector (HPLC, Agilent Infinity 1260, Agilent Technologies, Santa Clara, CA, USA) using an external standard method. Detailed sample preparation and PAH analysis in particulate matter were described by Godec et al. [49] and Jakovljević et al. [50]. Briefly, the extraction of PAHs was performed using a toluene/cyclohexane solvent mixture (7:3, v/v) in an ultrasonic bath for 1 h. The extracts were separated from undissolved particles by centrifugation (10 min at 3000 rpm) and evaporated to dryness under a gentle nitrogen stream at 30 °C. The residues were subsequently re-dissolved in acetonitrile. Analytical separation of PAHs was performed using a Zorbax Eclipse PAH analytical column (100 × 4.6 mm) (Agilent, Santa Clara, CA, USA). The mobile phase consisted of acetonitrile and water (60:40, v/v) and was delivered at a flow rate of 1 mL min⁻¹. Quantitative and qualitative determination of PAHs was performed using an external standard method and a commercial PAH standard (EPA 610 PAHs Mix, Supelco (Sigma-Aldrich), Bellefonte, PA, USA). Nine different PAHs were analyzed by this method (fluoranthene, pyrene, benzo(j)fluoranthene, benzo(a)pyrene, benzo(k)fluoranthene, dibenzo(a,h)anthracene, benzo(ghi)perylene, benzo(b)fluoranthene, and indeno(1,2,3-cd)pyrene); however, the results are shown as the sum of PAHs (∑PAHs).

The meteorological parameters, including average daily air temperature and mixing layer height (MLH) at 12:00 UTC, were provided by the Croatian Meteorological and Hydrological Service. The MLH values were derived from radiosonde soundings using the bulk Richardson number method [51] and performed continuously at the Maksimir meteorological station, located approximately 5 km southeast of the sampling site. Data analysis was based on a comprehensive five-year observational dataset spanning 2020–2024.

Air-mass transport pathways were analyzed using the NOAA (College Park, MD, USA) Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model, with meteorological input fields from the Global Forecast System (GFS) dataset (0.25° spatial resolution). Seventy-two-hour backward trajectories were generated at 6 h intervals, terminating at the sampling location at 1000 m above ground level (a.g.l.).

Descriptive statistical parameters, the Kruskal–Wallis H test (p < 0.05), post hoc multiple comparison tests, the Mann–Whitney U test, and mathematical fitting were conducted using Statistica software (TIBCO Software Inc., version 14.0.0.5).

3. Results and Discussion

3.1. PM₁₀ and ∑PAHs Mass Concentration

Figure 2 presents a time series of daily PM₁₀ (blue line, left axis, µg m⁻³) and ∑PAH mass concentrations (yellow line, right axis, ng m⁻³) over the period 2020–2024. The Shapiro–Wilk test indicated that PM₁₀ and ∑PAH concentrations were not normally distributed (p < 0.05). The average annual PM₁₀ concentrations during the measurement period ranged from 21 to 24 µg m⁻³, far below the annual limit value (40 µg m⁻³).

Figure 2. Daily mass concentrations of PM₁₀ (blue line) and ∑PAHs (yellow line) at the Zagreb urban background site in the measuring period 2020–2024. Red stars indicate reported Saharan dust intrusion events.

PM₁₀ concentrations exhibit frequent short-term peaks, with several episodes exceeding the daily limit value (50 µg m⁻³), particularly during winter. Both pollutants exhibit pronounced seasonal variability, with maximum concentrations in winter and lower concentrations in warmer months. Elevated values are most common between late autumn and early spring, whereas summer concentrations remain relatively stable and lower. Although several short-term PM₁₀ peaks were observed, the elevated concentrations recorded on 28 March 2020, 21 June 2021, 31 March 2024, and 21 June 2024 were most likely associated with long-range transport of desert dust, as indicated in Figure 2. These events were reported in the DHMZ meteorological bulletins [52,53,54,55,56] and further supported by HYSPLIT back-trajectory analysis (Figure S1).

During these episodes, elevated PM₁₀ concentrations were influenced not only by local emissions and boundary-layer dynamics but also by transboundary mineral dust transport originating from the Sahara and the Karakum Deserts. In contrast, the highest PAH concentrations occurred during winter stagnation episodes. The extreme PAH value observed on 14 January 2020 (90.2 ng m⁻³) was likely associated with stable anticyclonic conditions, weak atmospheric mixing, temperature inversions, and intensified residential heating emissions [53]. Similarly, elevated PAH concentrations recorded on 27–28 January 2024 were probably linked to persistent high-pressure conditions, weak winds, sub-zero morning temperatures, and dry weather following the cold outbreak in mid-January, all of which favored pollutant accumulation near the surface [54].

Figure 3a shows the average seasonal PM₁₀ concentrations over the five-year study period. Winter averages ranged from 27 to 33 µg m⁻³, spring levels from 17 to 24 µg m⁻³, summer values from 14 to 19 µg m⁻³, and autumn values from 22 to 26 µg m⁻³. Seasonal PM₁₀ concentrations exhibited some interannual variability over the five-year period, with slightly lower winter values observed in later years; however, no consistent long-term trend was evident across the seasons.

Figure 3. Average seasonal mass concentrations of (a) PM₁₀ and (b) ∑PAHs at the Zagreb urban background site in the measuring period 2020–2024.

Similar PM₁₀ patterns were reported in earlier studies conducted at the same site; however, the concentrations were slightly higher. Mean winter concentrations of 39 µg m⁻³ and summer concentrations of 21 µg m⁻³ were reported for 2008 [57], 37 µg m⁻³ in winter and 28 µg m⁻³ in summer for 2009 [45], and 44 µg m⁻³ in winter and 19 µg m⁻³ in summer for 2010 [49]. The observed levels at our site were compared with those at other sites in Croatia (Table 1). In eastern Croatia, somewhat higher values were obtained at an urban industrial site in Slavonski Brod, where winter levels were 41 µg m⁻³ and summer values were 16 µg m⁻³, and comparable values were measured at an urban background site in Vinkovci with 34 µg m⁻³ during winter and 20 µg m⁻³ during summer [40].

Table 1. Comparison of winter and summer PM₁₀ mass concentrations at different sites in Croatia.

	PM10/µg m−3		∑PAHs/ng m−3
year	winter	summer	winter	summer		Study site
2008	39	21	7.4	0.3	urban background	Zagreb [57]
2009	37	28	15.1	0.9	urban background	Zagreb [45]
2010	44	19	21.7	0.4	urban background	Zagreb [49]
2020	35	15	16.5	0.4	urban background	this study
2021	31	19	14.9	0.5	urban background	this study
2022	30	16	14.9	0.4	urban background	this study
2023	27	19	11.1	0.2	urban background	this study
2024	30	17	13.1	0.2	urban background	this study
2010	47	19	48.5	0.8	rural industrial	Delnice [49]
2019	-	16 *	/	0.9 *	rural coastal	Šibenik [44]
2021/2022	34	20	25.0	0.6	urban industrial	Vinkovci [40]
2024	14	17	/	/	rural background	Polača [58]

* spring and summer.

In southern Croatia, in Polača, a rural background site near the coast, the average annual PM₁₀ concentration was 14 µg m⁻³ in 2024, with fewer seasonal differences but generally higher summer levels (17 µg m⁻³) than winter levels (12 µg m⁻³) [58]. Additionally, at the central Adriatic coastal site in Šibenik, reported PM₁₀ levels were 16 µg m⁻³ from February to July of 2019 [44]. In western Croatia, at a rural industrial site in Delnice, winter values of 47 µg m⁻³ and summer values of 19 µg m⁻³ were reported for 2010 [49].

From Figure 2, it is evident that ∑PAH concentrations also exhibit a strong seasonal pattern, with sharp winter peaks that mostly coincide with periods of increased PM₁₀. During these episodes, ∑PAH levels reach up to 25–90 ng m⁻³, while remaining close to zero throughout summer. The average annual ∑PAH concentrations ranged from 4.1 ng m⁻³ to 6.6 ng m⁻³. The average seasonal values of ∑PAHs are shown in Figure 3b.

The highest average ∑PAH concentrations were consistently observed in winter, with values ranging from approximately 11.1 to 16.5 ng m⁻³, depending on the year. Concentrations decrease significantly in spring, ranging from roughly 1.3 to 6.2 ng m⁻³, and reach their lowest levels in summer, with values from 0.2 to 0.5 ng m⁻³. In autumn, ∑PAH concentrations increase again, typically ranging between 3.4 and 7.0 ng m⁻³. As shown in Table 1, observed ∑PAH values were lower than levels reported at the same site in 2010 (winter: 21.7 ng m⁻³, and summer: 0.4 ng m⁻³), and much lower compared to winter levels (48.5 ng m⁻³) and summer levels (0.8 ng m⁻³) measured in Delnice in 2010, although those measurements included a total of 10 different PAHs [49]. Comparable levels were reported at the central Adriatic coastal site in Šibenik, with 0.9 ng m⁻³ from February to July of 2019 [44]. In the urban area, the PAH values vary depending on the type of station; at urban industrial sites, reported values were the highest, reaching 44.7 ng m⁻³ during the colder seasons and 1.3 ng m⁻³ during the warmer seasons. The urban background site recorded 24.9 ng m⁻³ during the colder part and 0.6 ng m⁻³ during the warmer period in 2022 [24]. As seen from Figure 2, high PAH levels are coincident with higher PM₁₀ mostly during wintertime, which most likely suggests a common origin, e.g., increased residential heating and traffic emissions, but also the accumulation and persistence of semi-volatile PAHs in the lower temperature conditions [59,60]. In contrast, the consistently low summertime concentrations of both PM₁₀ and PAHs most likely indicate reduced emissions and enhanced atmospheric dispersion and photochemical degradation under warmer, well-mixed conditions [61]. Overall, the time series highlights the dominant role of seasonal combustion sources and meteorological conditions in shaping particulate pollution levels. Therefore, further analysis focuses on evaluating the effect of stable atmospheric conditions on elevated PM₁₀ and ∑PAH concentrations.

3.2. Mixing Layer Height (MLH) Characteristics

Figure 4 shows the MLH values derived from radiosonde measurements over the five-year study period. A data gap between 1 April and 25 July 2021 was caused by instrument maintenance. In total, MLH data were available for 1669 days. The time series indicates that the mixing layer is lowest during the cold seasons, with 22% of days when the MLH values were below 500 m. Statistical analysis using the Kruskal–Wallis test confirmed significant daily and seasonal variability in MLH (p < 0.05).

Figure 4. Mixing layer height (MLH) trend (12:00 UTC) in Zagreb City over the five-year period (2020 to 2024).

At most meteorological stations, radiosoundings are performed twice daily at standard synoptic times (00:00 and 12:00 UTC), as was the case at the Maksimir station in Zagreb. Although radiosonde measurements are widely used for determining MLH [25,31,35], this method has several limitations, including low temporal resolution (12-h intervals), limited vertical resolution for boundary-layer studies, smoothing effects due to sensor response time, and the fact that each sounding represents only a snapshot of the atmospheric boundary layer structure [25]. The MLH values measured at 00:00 UTC were examined; however, no seasonal dependence was observed, and these data were therefore excluded from further analysis. Such results could be explained by radiative cooling of the surface at night, which leads to the formation of a stable boundary layer [20]. During this process, the air near the ground becomes cooler than the air above, suppressing vertical air motion and limiting turbulent mixing. As a result, the mixing layer height remains consistently low during nighttime throughout the year [25,62].

After sunrise, surface heating induces turbulence and enhances vertical mixing, causing the mixing layer to grow and typically reach its maximum height in the early afternoon. This behavior is reflected in radiosonde measurements at 12:00 UTC [25,63]. During the cold season, however, reduced solar heating limits turbulence, so the mixing layer often remains relatively low even during daytime. For this reason, MLH values measured at 12:00 UTC were used as an indicator of atmospheric stability and, consequently, the temperature inversion conditions.

Based on the positions and elevations of the mountains surrounding Zagreb City and the measurement site shown in Figure 1, the MLH below 1000 m is considered sufficient to detect temperature inversion conditions. Analysis of MLH frequency below 1000 m showed that these conditions occurred on 44% of days in the five-year dataset (Figure 5a). Inversion conditions were most common during the cold seasons (winter and autumn), as shown in Figure 5b. Inversions occurred in 35% of cases, representing nearly 80% of all inversion events during winter and autumn. This ratio represents a total of 738 days in a five-year period. On the other hand, during spring and summer, inversions were present in only 9% of cases. Similar results were obtained by Lončar in Zagreb over the 5-year period from 1977 to 1982, which showed a very high frequency of inversion conditions, particularly during nighttime (around 90% of cases), while mixing layer heights were significantly lower during the cold season and higher during the warm season. The study confirmed pronounced diurnal and seasonal variability in atmospheric stability, with stable conditions and more persistent inversions occurring mainly in winter [36].

Figure 5. Histogram of MLH (a) during the five-year period, (b) in different seasons.

3.3. Influence of Mixing Layer Height on PM₁₀ and ∑PAH Concentrations

To prevent bias due to missing radiosonde data, the statistical analysis was limited to matched observations, including only days with simultaneous PM₁₀, PAHs, and MLH measurements. Descriptive statistics of the dataset are presented in Table 2. PM₁₀ concentrations exhibited clear seasonal variability, with the highest values observed during winter (31 µg m⁻³) and autumn (23 µg m⁻³). In contrast, concentrations during spring and summer were significantly lower, averaging 18 µg m⁻³ and 17 µg m⁻³, respectively. Out of 1669 days, the daily PM₁₀ limit value was exceeded on 69 days, predominantly during the cold season: 46 times in winter and 17 times in autumn, indicating the strong influence of seasonal emissions and meteorological conditions. A similar seasonal pattern was observed for ∑PAHs, with the highest concentrations recorded in winter (14.101 ng m⁻³) and autumn (5.096 ng m⁻³), while significantly lower values were observed during spring (2.701 ng m⁻³) and summer (0.333 ng m⁻³).

Table 2. Descriptive statistics of PM₁₀ and ∑PAH mass concentrations in different seasons in the observed period.

Season	N	PM10/µg m−3		∑PAHs/ng m−3
		CAVG/µg m−3	N (C > 50 µg m−3)	CAVG/ng m−3
Winter	438	31	46	14.101
Spring	385	18	4	2.710
Summer	397	17	2	0.333
Autumn	449	23	17	5.096
All groups	1669	22.5	69	5.776

N—number of days, C_AVG—average mass concentration, N (C > 50 µg m⁻³)—number of days when the daily concentration exceeded the limit value of 50 µg m⁻³.

The Kruskal–Wallis H test confirmed statistically significant seasonal differences in both PM₁₀ and ∑PAH concentrations (p < 0.05). Post hoc analysis further confirmed significant differences between individual seasons.

The relationship between pollutant concentrations and MLH is shown in Figure 6, indicating that the highest concentrations of both PM₁₀ and ∑PAHs occur at lower MLH values. Since a higher frequency of low MLH was observed during colder periods (Figure 5), these results suggest that limited vertical mixing plays a key role in pollutant accumulation. In contrast, during the warm season, higher mixing layers and enhanced turbulent exchange promote more efficient pollutant dispersion. Higher temperatures generally enhance daytime vertical atmospheric mixing, whereas cooler conditions suppress turbulent exchange and favor the accumulation of pollutants within the near-surface layer, leading to elevated PM₁₀ concentrations [64]. In addition to reduced atmospheric mixing under stable conditions, this seasonal pattern may also be associated with increased emissions during the cold season, particularly from traffic and residential heating [40,49]. Furthermore, precipitation contributes to the removal of airborne particles through wet deposition and scavenging processes, especially during rainfall events [65].

Figure 6. Mass concentrations of (a) PM₁₀ and (b) ∑PAHs versus MLH during the observed period.

A comparison of average daily concentrations between the full dataset and inversion conditions (MLH < 1000 m) is presented in Table 3. Both PM₁₀ and ∑PAH concentrations were higher during inversion conditions, with ∑PAHs showing a particularly strong response, reaching nearly twice the average values of the full dataset (9.917 ng m⁻³ compared to 5.776 ng m⁻³). PM₁₀ concentrations also increased, although less markedly (27.5 µg m⁻³ vs. 22.5 µg m⁻³). Notably, approximately 90% of PM₁₀ daily limit exceedances occurred during inversion events, demonstrating the dominant role of atmospheric stability in pollution episodes. Studies conducted by Vardoulakis et al., which included urban background sites of Athens (Greece) and Birmingham (UK), showed that high PM₁₀ concentrations are also closely associated with unfavorable dispersion conditions, such as low wind speeds and temperature inversions, particularly during the cold season [66]. Similarly, Holst et al. demonstrated that reduced mixing layer height and stable atmospheric stratification significantly enhance PM₁₀ accumulation in Baden-Württemberg (Germany) in both rural and urban backgrounds [67].

Table 3. Mean daily PM₁₀ and ∑PAH concentrations and the number of PM₁₀ daily limit exceedances in the full dataset and during temperature inversion events.

	N	PM10		∑PAHs
		CAVG/µg m−3	N (C > 50 µg m−3)	CAVG/ng m−3
Full dataset	1669	22.5	69	5.776
Inversion	738	27.5	62	9.917

N—number of days, C_AVG—average mass concentration, N (C > 50 µg m⁻³)—number of days when the daily concentration exceeded the limit value of 50 µg m⁻³.

To further investigate the impact of atmospheric stagnation, pollutant concentrations were analyzed by inversion duration: one-day, two-day, and ≥3-day (Table 4). Both PM₁₀ and ∑PAH concentrations increased progressively with the duration of inversion events. During one-day inversions, PM₁₀ concentrations remained relatively low (17.4 µg m⁻³), with only two exceedances of the daily limit value, while ∑PAH concentrations remained modest (4.45 ng m⁻³), whereas multi-day inversion events were associated with substantially higher pollutant levels and the majority of exceedances. Results showed two-day inversions were associated with higher PM₁₀ (22.5 µg m⁻³) and ∑PAH levels (7.15 ng m⁻³), accompanied by a greater number of PM₁₀ daily limit exceedances (N = 7). The highest concentrations were observed during prolonged inversions lasting three or more days, when PM₁₀ averaged 26.2 µg m⁻³ and ∑PAHs averaged 12.95 ng m⁻³. In particular, prolonged inversions (≥3 days) accounted for 53 exceedance days, indicating that persistent stable conditions allow continuous emissions to accumulate near the surface.

Table 4. Mean daily PM₁₀ and ∑PAH concentrations and PM₁₀ daily limit exceedances for different durations of temperature inversion events.

Duration	N	PM10		∑PAHs
		CAVG/µg m−3	N (C > 50 µg m−3)	CAVG/ng m−3
1-day inversions	136	17.4	2	4.452
2-day inversions	47	22.5	7	7.147
≥3-day inversions	63	26.2	53	12.951

N—number of days, C_AVG—average mass concentration, N (C > 50 µg m⁻³)—number of days when the daily concentration exceeded the limit value of 50 µg m⁻³.

These results indicate that both the occurrence and duration of temperature inversion events are key meteorological drivers of elevated PM₁₀ and ∑PAH concentrations. Longer-lasting inversions suppress vertical mixing and dispersion, allowing continuous emissions from traffic and residential heating to accumulate near the surface. The stronger response of PAHs compared with PM₁₀ suggests that combustion-related pollutants are particularly sensitive to atmospheric stability. This is likely due to their semi-volatile nature and tendency to partition onto particles at low temperatures. Overall, the results suggest that inversion duration is a key meteorological driver of pollution episodes, with multi-day inversions posing the greatest risk for elevated particulate concentrations and regulatory exceedances. Similar results were reported by Mustapić et al., who studied PM levels (PM₁, PM_2.5, PM₁₀) in urban areas, using unmanned aerial vehicles equipped with sensors to focus on traffic-related pollution in Zagreb. Results show that particle concentrations are highest during morning traffic peaks and, contrary to expectations, do not consistently decrease with altitude, highlighting the importance of meteorological conditions in pollutant distribution [68]. A study of a severe PM pollution episode in Santiago, Chile, investigated the role of atmospheric stagnation over complex terrain. It showed that strong temperature inversions, shallow boundary layer heights (often below 200 m), and weak wind conditions significantly reduced atmospheric ventilation, leading to the accumulation of pollutants near the surface [69]. A recent study conducted in the urban environments of Athens investigated the seasonal and diurnal variability of the MLH and its relationship with urban air pollution. The results showed that pollutant concentrations, including PM_2.5, black carbon, and NOx, are strongly negatively correlated with MLH, with the highest levels occurring under stable atmospheric conditions and shallow mixing layers [29].

3.4. Cold-Season Statistical Analysis

Additional analyses were performed using the cold-season dataset, as approximately 80% of temperature inversion events during the five-year period occurred in autumn and winter (Figure 5b). Restricting the analysis to this period minimizes the influence of seasonal emission variability and reduces the effect of temperature-dependent PAH volatilization at higher temperatures [59,60]. The cold-season dataset, including only winter and autumn, was therefore used to compare PM₁₀ and ∑PAH concentrations during inversion (INV) and non-inversion (NON-INV) conditions. As mentioned, the MLH threshold of <1000 m was selected based on the local topography of the Zagreb area, which is surrounded by elevated terrain exceeding 1000 m. Under such conditions, a mixing layer below this altitude indicates that air masses are confined within the basin-like environment, reducing ventilation and favoring pollutant accumulation. Therefore, MLH < 1000 m was considered a suitable indicator of stable atmospheric conditions and inversion events in this study. Conversely, non-inversion conditions were considered when MLH was above 1000 m. The dependence of pollutant concentrations on MLH during the cold season is shown in Figure 7. Similar trends to those observed for the full dataset are evident, confirming that elevated concentrations are primarily associated with low MLH values and inversion conditions.

Figure 7. Mass concentrations of (a) PM₁₀ and (b) ∑PAHs versus MLH during the cold season.

Statistical parameters for inversion and non-inversion days during the cold season are presented in Table 5.

Table 5. Mean daily PM₁₀ and ∑PAH concentrations and the number of PM₁₀ daily limit exceedances during temperature inversion days (INV) and non-inversion days (NON-INV).

	N	PM10			∑PAHs
		CAVG/µg m−3	SD	N (C > 50 µg m−3)	CAVG/ng m−3	SD
INV	580	30	16.3	59	12.026	10.6442
NON-INV	307	20	10.7	4	4.853	7.0475
All Groups	887	26.7	15.4	63	9.543	10.1416

N—number of days, C_AVG—average mass concentration, N (C > 50 µg m⁻³)—number of days when the daily concentration exceeded the limit value of 50 µg m⁻³, SD—standard deviation.

Both PM₁₀ and ∑PAH concentrations were significantly higher during inversion conditions compared with non-inversion days. This effect is particularly evident in the number of exceedances: of the 63 PM₁₀ daily limit exceedances recorded during the cold season, 59 occurred during inversion events. In addition, the greater sensitivity of PAHs to atmospheric stability observed in this study is consistent with findings reported by Guzmán-Torres, where particle-bound PAH concentrations showed a clear dependence on mixing height and inversion conditions [70]. Furthermore, Klingner et al. emphasized that meteorological factors, especially mixing conditions and atmospheric stability, play a crucial role in PM₁₀ variability, often outweighing the impact of emission changes [71].

The statistical significance of INV and NON-INV differences was confirmed using the Mann–Whitney U test (Table 6), which showed that concentrations of both PM₁₀ and ∑PAHs were significantly higher during inversion conditions (p < 0.05). These findings reinforce that atmospheric stability and a lower mixing layer height primarily influence pollutant buildup during the cold season.

Table 6. Mann–Whitney U test analysis with cold-season dataset.

	N (INV)	N (NON-INV)	CAVG (INV)	CAVG (NON-INV)	Z	p
PM10/µg m−3	580	307	30	20	−4.6386	<<0.05
∑PAHs/ng m−3	580	307	12.057	4.853	−6.4201	<<0.05

N—number of days, C_AVG—average mass concentration, INV—temperature inversion days, NON-INV—non-inversion days.

Largeron et al. [22] obtained similar results when investigating the relationship between persistent wintertime temperature inversions and PM₁₀ air pollution in Alpine valleys. These results showed that long-lasting inversions, driven by high-pressure systems and warm-air advection aloft, significantly increased atmospheric stability and trapped pollutants near the surface, which led to severe pollution episodes. A strong correlation between inversion strength and PM₁₀ concentrations was demonstrated, indicating that local meteorological conditions, rather than long-range transport, are the primary driver of winter pollution events [22].

To further describe the relationship between MLH and pollutant concentrations, the data were fitted using logarithmic, exponential, and quadratic functions (Figure 8). The parameters of the fitted functions are presented in Table 7. Although the correlation coefficients (R²) were relatively low due to the high variability of pollutant concentrations, a consistent inverse relationship between MLH and pollutant levels was observed. The relatively weak correlations also indicate that, in addition to MLH and atmospheric stability, pollutant concentrations are influenced by other meteorological factors, including precipitation, wind speed, temperature, relative humidity, and synoptic circulation patterns. These parameters affect pollutant dispersion, deposition, chemical transformation, and atmospheric removal processes. For example, precipitation contributes to wet deposition and scavenging of airborne particles [72,73], while stronger winds enhance atmospheric ventilation and pollutant dispersion [74,75]. Nevertheless, the fitted functions provide a useful approximation of the relationship between atmospheric mixing conditions and pollutant concentrations and may serve as input variables for machine-learning models aimed at predicting air pollution levels [76,77,78].

Figure 8. Mathematical fitting of (a) PM₁₀ and (b) ∑PAH mass concentrations versus MLH during the cold season.

Table 7. Parameters of fitted functions relating PM₁₀ and ∑PAH mass concentrations to MLH.

Fitting Function	Exponential	Logarithmic	Quadratic
	$\mathit{y}=\mathit{a}·{\mathit{e}}^{\mathit{b}\mathit{x}}$	$\mathit{y}=\mathit{c}·\mathbf{ln}\mathit{x}+\mathit{d}$	$\mathit{y}=\mathit{e}·{\mathit{x}}^2+\mathit{f}·\mathit{x}+\mathit{g}$
PM10 vs. MLH	a = 30.647 b = −0.0004 R2 = 0.1463	c = −4.757 d = 56.695 R2 = 0.1139	e = 4·10−6 f = −0.0178 g = 37.111 R2 = 0.1489
∑PAHs vs. MLH	a = 14.041 b = −0.001 R2 = 0.2367	c = −4.006 d = 36.699 R2 = 0.1413	e = 4·10−6 f = −0.0161 g = 20.362 R2 = 0.1935

3.5. Zagreb-Specific Characteristics of MLH Influence on Air Pollution

The present results highlight several characteristics specific to the Zagreb urban environment that modulate the relationship between MLH and air pollution. Zagreb is situated in a basin-like area surrounded by elevated terrain, particularly the Medvednica mountain to the north and higher terrain to the southwest, which can reduce atmospheric ventilation and favor pollutant accumulation during stable meteorological conditions. Frequent wintertime temperature inversions, weak winds, and persistent anticyclonic episodes further suppress vertical mixing and contribute to prolonged stagnation events. Under such conditions, pollutants emitted from residential heating and traffic remain confined within a shallow mixing layer, leading to elevated PM₁₀ and PAH concentrations and frequent exceedances of air quality limit values. The results further indicate that inversion duration plays an important role in pollution intensity in Zagreb, since multi-day inversion events were associated with the highest pollutant concentrations and the majority of PM₁₀ exceedances. Compared with many coastal or more ventilated European urban areas, Zagreb appears particularly sensitive to wintertime atmospheric stability due to the combined effects of local topography, seasonal combustion emissions, and reduced boundary-layer development during the cold season. These findings emphasize the importance of including MLH and inversion persistence in future air quality forecasting and management strategies for Zagreb.

4. Conclusions

This study examined the relationship between mixing layer height (MLH) and concentrations of PM₁₀ and particle-bound polycyclic aromatic hydrocarbons (PAHs) in Zagreb using a five-year dataset (2020–2024). The results indicate a strong influence of atmospheric stability on urban air pollution levels. Both PM₁₀ and PAH concentrations showed marked seasonal variability, with peak levels in winter and autumn. MLH values below 1000 m, indicative of temperature inversion conditions, occurred on approximately 44% of days, mainly during the cold season. Pollutant concentrations were significantly higher during inversion conditions than on non-inversion days. Average ∑PAH concentrations during inversion events were nearly double those of the full dataset (9.917 ng m⁻³ vs. 5.776 ng m⁻³), while PM₁₀ concentrations also increased, but to a lesser extent (27.5 µg m⁻³ vs. 22.5 µg m⁻³). Furthermore, about 90% of PM₁₀ daily limit exceedances occurred during inversion conditions. The duration of inversion events was a key factor influencing pollution episodes; both PM₁₀ and ∑PAH concentrations increased progressively with the length of inversion periods, reaching the highest concentrations (26.2 µg m⁻³ and 12.951 ng m⁻³, respectively), and exceedances during multi-day inversion events. These findings suggest that prolonged atmospheric stagnation facilitates the accumulation of pollutants emitted from local sources near the surface. Elevated concentrations coincide with increased combustion emissions from residential heating and traffic, common in urban areas, combined with meteorological conditions that limit atmospheric dispersion. In contrast, summer concentrations remained relatively low due to reduced emissions, enhanced atmospheric mixing, and increased photochemical degradation of pollutants. Additional analysis of cold-season data confirmed the significant impact of atmospheric stability on pollutant levels. Statistically significant differences in PM₁₀ and ∑PAH concentrations between inversion and non-inversion days highlight the importance of stable atmospheric conditions in pollution episodes. The results also reveal several characteristics specific to the Zagreb urban environment. Due to the basin-like topography surrounding the city, frequent wintertime temperature inversions, weak ventilation conditions, and enhanced residential heating emissions during the cold season, Zagreb is particularly susceptible to pollutant accumulation under shallow mixing-layer conditions. Multi-day inversion events were identified as especially critical for air quality deterioration since prolonged atmospheric stability allows continuous accumulation of locally emitted pollutants near the surface. These findings demonstrate that, in addition to emission intensity, local boundary-layer dynamics and inversion persistence play a crucial role in determining wintertime pollution episodes in Zagreb. Overall, these results emphasize the critical role of MLH and atmospheric stability in controlling particulate pollution in Zagreb. Based on the results of this study, the question is whether reducing emissions can generally help achieve the new air quality goals set by the WHO and the EU Commission. Incorporating the MLH and inversion persistence into statistical air quality assessment could improve the prediction of pollution episodes and support more effective urban air quality management strategies in Zagreb.