The Influence of Biomass Burning on the Organic Content of Urban Aerosols

Abstract

This study examines the influence of biomass burning on the organic content of urban aerosols in Zagreb, Croatia, by analyzing anhydrosugars, elemental carbon (EC), organic carbon (OC), and water-soluble organic carbon (WSOC) in PM_2.5 and PM₁ fractions collected during different seasons of 2022. Seasonal trends showed that the highest average concentrations of PM_2.5 (27 µg m⁻³) and PM₁ (17 µg m⁻³) were measured during the winter and decreased in the spring, summer, and autumn, which is in accordance with the specific activities and environmental conditions typical for each season. Different sources of OC and WSOC were noticed across different seasons; levoglucosan (LG) was measured during the winter (1314 ng m⁻³ in PM_2.5 and 931 ng m⁻³ in PM₁), indicating that biomass that was mostly used for residential heating was the main source rather than the agricultural activities that are usually common during warmer seasons. The contribution of LG to PM was 5.3%, while LG contributed to OC by up to 13.4% and LG contributed to WSOC by up to 36.5%. Deviations in typical seasonal variability of LG/WSOC revealed more intense biomass burning episodes during the autumn and several times during the winter season. A back trajectories HYSPLIT model revealed a long-range transport biomass emission source. The levoglucosan-to-mannosan (LG/MNS) ratios indicated the burning of mixed softwood and hardwood during colder seasons and the burning of softwood during warmer seasons. Spearman’s correlation tests and principal component analysis showed a strong and statistically significant (p < 0.05) correlation between LG, PM, OC, EC, and WSOC only during the winter season, demonstrating that they had the same origin in the winter, while their sources in other seasons were diverse.

1. Introduction

In air pollution studies, biomass burning, along with traffic emissions, is consistently identified as one of the major pollution sources in southeastern Europe and the Western Balkans [1,2,3]. In these regions, biomass remains a widely used energy source for residential heating due to its affordability and widespread availability. However, the incomplete combustion of biomass—dependent on factors such as oxygen availability, biomass moisture content, and combustion temperature—results in the release of significant amounts of air pollutants into the atmosphere [4,5]. In addition to gases such as SO₂, NO_X, CO₂, and CH₄, particulate matter (PM) is one of the pollutants emitted in significant amounts during biomass combustion [6,7]. Generally, PM is recognized as one of the major threats to human health as it can be inhaled and, depending on the size, penetrate deep into the respiratory and cardiovascular systems [8,9]. During biomass and fossil fuel combustion, the emissions mostly consist of the PM_2.5 fraction, in which carbonaceous matter (CM) is one of the major components [10]. In addition to direct emissions from pollution sources, known as primary carbonaceous matter, it can also be generated through chemical reactions (e.g., oxidation and oligomerization) and condensation of particles formed by gaseous precursors. These processes typically occur during atmospheric aging, and due to its multi-step formation, it is classified as secondary carbonaceous matter [11].

CM can be divided into organic carbon (OC) and elemental or black carbon (EC or BC, respectively), depending on whether a thermo-chemical or optical property is being investigated [12]. EC is considered a primary aerosol, originating almost exclusively from pyrolysis during incomplete combustion processes, such as traffic emission, industrial activities, or biomass burning [13]. In contrast, OC can originate from both primary natural (e.g., vegetative detritus) and anthropogenic sources (e.g., vehicular emissions, wood burning, industrial processes, and cooking operations), referred to as primary organic carbon (POC), or/and secondary sources through a photochemical reaction between gas-phase precursors, resulting in secondary organic carbon (SOC) [14]. This variety of OC sources contributes to a broad spectrum of water-soluble (carboxylic acids, amino acids, amines, dicarboxylic acids, hydroxycarboxylic acids, carbohydrates, and their derivatives, humic-like substances (HULIS), etc. [15]) and water-insoluble organic compounds in aerosols (alkanes, alkanals, alkanones, waxes, proteins, plant fragments, polycyclic aromatic hydrocarbons (PAHs), n-alkanoic acids, etc. [16]). While the literature associates WSOC with SOC, it can be produced during biomass burning [17]. Moreover, a source apportionment study by Lopez-Caravaca et al. showed biomass burning was the primary contributor to WSOC levels during the cold season, whereas, in the summer, the largest fraction of WSOC was linked to secondary organic aerosol formation [18].

Biomass, chemically defined as lignocellulose, consists of different ratios of cellulose (40–50%), hemicellulose (15–25%), and lignin (15–30%) [19]. Suciu et al. proposed a detailed mechanism of cellulose and hemicellulose pyrolysis where the structures transform through different pathways, such as dehydration, fragmentation, oxidation, hydrolysis, polymerization, etc., after which the formation of mono- and oligomeric anhydrosugars occurs. The thermal degradation of crystalline cellulose, which consists of linearly linked β-(1 → 4)-D-glucopyranose units, forms levoglucosan (1,6-anhydro-β-D-glucopyranose), while the thermal degradation of amorphous hemicellulose produces mannosan (1,6-anhydro-β-D-mannopyranose) and galactosan (1,6-anhydro-β-D-galactopyranose) [20]. These three isomers belong to the group of carbohydrates called anhydrosugars. Since cellulose is more abundant than hemicellulose in plants [21], the pyrolytic production of levoglucosan is generally higher than that of its isomers, i.e., mannosan and galactosan. Recent studies have demonstrated the semi-volatile properties of anhydrosugars [22], highlighting the significant influence of factors such as air temperature, relative humidity, wind direction, etc., on their concentrations in airborne particles. Due to its abundance, relatively high stability (particularly during seasons with low ultraviolet radiation and low levels of hydroxyl radicals), and distinctive source characteristics, LG is widely recognized as a specific marker for biomass burning [7,23,24]. Additionally, due to the structural differences of hemicellulose across different types of biomass compared to cellulose, the levels of two less abundant isomers, namely, mannosan and galactosan (GA), fluctuate depending on the type of vegetation [25]. Such variations allow for the identification of the biomass type that has been burned. The LG/MNS ratio has been suggested as a useful indicator for differentiating between the burning of softwood and hardwood, while the ratio of LG/(MNS + GA) has been employed to distinguish between older wood (e.g., brown coal) and more recently harvested wood [26]. Additionally, this ratio can help identify wood combustion from other types of biomass burning, e.g., leaves, crops, grass, herbs, etc. [27]. Although these compounds have been found in lower-maturity coals, likely due to the presence of residual cellulose and hemicellulose, this source is negligible in countries where coal use is not predominant [28].

The ratio of levoglucosan to OC (i.e., LG/OC) is used to quantify the contribution of biomass to organic matter [29,30]. It enables one to calculate emission ratios (ERs) and emission factors (EFs), where ERs are expressed as the ratio of levoglucosan to total OC on a mass concentration basis (µg/µg OC) and EFs are defined as the ratio of the mass of levoglucosan to the total fuel burned (mg/kg), although EFs are mostly obtained in laboratory trials rather than in field studies [31]. These two parameters are used to estimate fire emissions and trace vegetation sources of fire tracers in the environment. For example, higher ERs are registered during coniferous forest fires compared to burning in deciduous forests and grasses [32].

The OC/EC ratio is used to provide information regarding the type of fuel burned in terms of biomass or fossil fuel [33] as well as to monitor the changes in aerosol composition due to different sources of both OC and EC [13]. Moreover, the OC/EC ratio is often used to estimate SOC since it is directly influenced by photochemical activity [34]. If SOC predominates, higher OC/EC ratios are typically observed, while minimum ratios are taken as representative of POC. Castro et al. obtained an OC/EC ratio of 1.1 in an urban background station and characterized it as the contribution of POC to OC, while all the higher values were assigned to SOC [35]. An elevated OC/EC ratio may also result from biomass burning sources, where higher levels of OC relative to EC can be attributed to significant local contributions from household heating, favorable meteorological conditions, and the influence of mixing layer height [36].

Due to the high solubility of anhydrosugars in water, the measurement of WSOC can be useful in separating fossil to non-fossil combustion sources, where the WSOC fraction is mostly dominated by a non-fossil source [37]. The comparison of LG to WSOC and OC, along with the estimation of biomass burning contributions to overall WSOC and OC levels [38], can serve as a tool for identifying mutual sources and detecting more intense biomass burning episodes through LG/WSOC and LG/OC ratios.

The objective of this study is to identify the contribution of biomass burning as a source of urban aerosols by determining the mass concentrations of certain organic compounds that can be used as tracers and diagnostic parameters. The levels of OC, EC, WSOC, LG, MNS, and GA were determined in different particle size fractions, i.e., PM_2.5 and PM₁, through different seasons. Diagnostic ratios, such as OC/EC, LG/OC, and WSOC/OC, were utilized to estimate the extent of biomass burning in airborne particles, while the type of biomass burned was assessed using the LG/MNS ratio. Unusual behavior in the diagnostic ratios was compared to a trajectory study using the NOAA HYSPLIT model.

2. Materials and Methods

Particulate matter collection for two aerodynamic particle sizes, i.e., PM_2.5 and PM₁, was conducted in an urban background area of Zagreb, Croatia. Sampling was performed over 24 h cycles using quartz fiber filters (47 mm, Pallflex Tissue quartz 2500 QAT-UP (Pall Corporation, New York, NY, USA)). Blank quartz filters were pre-baked at 850 °C for 3 h to minimize the carbon content prior to sample collection. Low-volume samplers (Sven Leckel 47/50, Ingenieurbüro GmbH, Berlin, Germany) were used, operating at an ambient air flow rate of approximately 55 m³ per day. A total of 122 samples per fraction were collected, with around 30 samples taken during each season of 2022. PM_2.5 and PM₁ mass was determined gravimetrically using microbalances (Mettler Toledo MX5 and XP6/M, Giessen, Germany, with a resolution of 1 µg) and an electrostatic charge outflow system. Filters were conditioned for 48 h at a controlled temperature (20 ± 1 °C) and relative humidity (45–50% RH) before and after sampling. The samples were then portioned, sealed in aluminum foil envelopes, and stored at −20 °C until analysis.

The sampling site was located in the moderately populated district near extensive public green spaces and forests and about 3 km from Medvednica Nature Park. The area is surrounded by residential buildings and is close to a moderately trafficked road. In the continental region of Croatia, including Zagreb, seasonal activities significantly influence environmental conditions and pollutant sources. During the winter, increased traffic, road salting, and intensified residential heating are prevalent. Although there has been a natural gas network in the area for more than 30 years, many households still use wood for heating. In the spring, enhanced biological activities, including some agricultural activities. Summer is characterized by reduced urban traffic due to school breaks and vacations, while agricultural activities such as plowing, harvesting, and burning of crop residues, which predominantly occur in rural areas, are more frequent.

Meteorological data for Maksimir, Zagreb, approximately 3 km away from the sampling site, were obtained by the Croatian Meteorological and Hydrological Service and are presented in Table S1 of the Supplementary Materials.

The levels of organic carbon and elemental carbon in the PM_2.5 and PM₁ fractions were determined using the thermal–optical transmittance method (TOT) [39] on a Carbon Aerosol Analyzer from Sunset Laboratory Inc., Tigard, OR, USA by following the EUSAAR_2 protocol [40]. The EUSAAR_2 thermal–optical analysis protocol was developed through the EU-project EUSAAR to improve the accuracy of distinguishing between OC and EC. To ensure quality control, an inner standard, an external sucrose solution, and a cross-method procedure were used. Recovery results ranged between 98% and 102%, with a relative standard deviation (RSD) of less than 5% [41].

SOC and POC were determined using the following equations:

SOC = OC_exp − POC

(1)

POC = EC_exp × (OC/EC)_min

(2)

where OC_exp and EC_exp are experimentally determined mass concentrations of OC and EC and (OC/EC)_min is the minimal ratio of OC/EC mass concentrations in the season [36]. This approach is widely used despite its limitations and uncertainties, which arise from the assumption that SOC formation occurs when the OC/EC ratio exceeds its minimum value and due to the fact that the OC/EC ratio is dependent on meteorological conditions, seasonal emission patterns, and contributions from local sources [42]. Mass concentrations of anhydrosugars (LG, MN, and GA) in PM_2.5 and PM₁ fractions were determined by high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD). The analytical procedure is described in Sopčić et al. (2024) [43]. Briefly, a sample aliquot was dissolved in ultrapure water (18.2 MΩ cm², Smart2Pure3 UF/UV, Thermo Scientific Barnstead GenPure, Waltham, MA, USA) and extracted using an ultrasonic bath for 45 min at 30 ± 5 °C. The water-soluble part of the sample was separated from the filter by centrifugation at 3000 rpm for 10 min, transferred to a polypropylene vial, and analyzed using an ICS-6000 (Thermo Fischer Scientific, Waltham, MA, USA). Analytes were separated on a Dionex CarboPac MA1 analytical column (Thermo Fischer Scientific, Waltham, MA, USA) using a NaOH eluent and detected on a gold working electrode with a standard quadruple potential waveform. For proper analytical separation of LG, MNS, and GA, two methods were developed. The detection limit for anhydrosugars ranged from 0.9 to 6 ng m⁻³, while that for the method of recovery was 95–98%. For water-soluble organic carbon analysis, the same sample solution was employed. A specific volume was added to a blank quartz filter and analyzed using the same thermal–optical protocol with a Carbon Aerosol Analyzer from Sunset Laboratory Inc., USA.

To determine the relationships between organic compounds and their statistical significance, Statistica 14 software was used. Backward trajectories were used to identify the origin and transport paths of air masses arriving at the sampling site. A 24 h period was analyzed using the HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) model, accessed via the READY (Real-time Environmental Applications and Display System) platform provided by the NOAA (National Oceanic and Atmospheric Administration) Air Resource Laboratory, College Park, Maryland, USA. Trajectories were simulated for the days of interest at three altitude levels: 500 m, 1000 m, and 1500 m. The meteorological input for the model was obtained from the GFS (Global Forecast System) dataset, with data provided at a spatial resolution of 0.25°.

3. Results

3.1. Particulate Matter and Carbonaceous Content

Mass concentrations of PM_2.5, PM₁, and their carbonaceous content (i.e., OC, EC, WSOC, LG, MNS, and GA) collected in different seasons are summarized in Tables S2 and S3. A comparison of PM_2.5 and PM₁ mass concentrations across different seasons is illustrated in Figure 1. According to the 2022 annual report from the Institute for Medical Research and Occupational Health [44], the average annual mass concentrations for PM_2.5 and PM₁ at this sampling station were 14.4 ± 10.1 µg m⁻³ and 10.8 ± 7.4 µg m⁻³, respectively. It is worth mentioning that the average annual value of PM_2.5 levels was below the annual limit value of 25 µg m⁻³ set by the relevant EU Directive but higher than 5 µg m⁻³, which is recommended by the 2021 World Health Organization (WHO) guidelines [45]. Both fractions demonstrated a similar seasonal trend, with peak concentrations observed in the winter, followed by a decline through the spring and summer, reaching the lowest values in the autumn. Such a trend can be attributed to specific activities and environmental conditions typical of each season. In Zagreb and in the continental part of Croatia in general, winter and spring are characterized by low temperatures (Table S1), salting of roads, more intensive residential heating, and increased traffic, which contribute to elevated PM levels. On the other hand, during the summer, urban traffic decreases due to school breaks and vacations, which is reflected in the lower PM levels. The lowest levels in the autumn are most likely influenced by frequent and intense rainfall, which likely facilitated the removal of PM from the atmosphere to the ground. Seasonal differences in mass concentrations were statistically significant, as determined by the Kruskal–Wallis analysis (p < 0.05). A strong correlation was observed between PM₁ and PM_2.5, with R² = 0.905 and a slope of 0.585, indicating that approximately 60% of PM_2.5 consisted of particles smaller than 1 µm (Figure 1).

The overall average OC concentrations during the sampling period were 5.4 ± 2.2 µg m⁻³ in PM_2.5 and 3.8 ± 1.5 µg m⁻³ in PM₁. Seasonal variations of OC concentrations in both fractions had the same trend as the PM concentrations, decreasing sequentially from the winter, spring, and summer to the autumn. Such a pattern was also in accordance with meteorological conditions where, at lower temperatures, the condensation of organic compounds onto pre-existing particles is enhanced, leading to increased OC mass concentrations and higher OC content in PM. The observed concentrations were slightly higher compared to OC levels measured in 2020 at the same measuring site (average OC of 4.7 ± 3.4 µg m⁻³ in PM_2.5 and 3.8 ± 2.5 µg m⁻³ in PM₁) and much lower compared to those measured in 2011–2013, when the average level was 9.4 µg m⁻³ [46]. OC levels were also higher than those reported for OC at an urban site in Helsinki (3 µg m⁻³ in PM_2.5 and 2.5 µg m⁻³ in PM₁) [10,47] but comparable or slightly lower than the values reported for Athens, i.e., 6.3 µg m⁻³ (winter) and 2.9 µg m⁻³ (summer) [48]. In the urban background sites of the Lombardy region in Italy, OC in PM₁₀ was between 4.5 and 9.7 µg m⁻³ [49].

The average EC concentrations over the sampling period were 0.9 ± 0.7 µg m⁻³ in PM_2.5 and 0.7 ± 0.5 µg m⁻³ in PM₁. EC exhibited the same seasonal trend as OC. The average EC levels observed in this study (0.3 µg/m³ in the summer and 1.3 µg/m³ in the winter) were either lower (Amsterdam (1.9 µg/m³, summer), Barcelona (2.6 µg/m³, winter)) or comparable to levels reported in Ghent, Belgium (0.8 µg/m³ in the summer and 1.2 µg/m³ in the winter) [50]. As EC is frequently employed as an indicator of traffic emissions [51], our findings indicate that the influence of traffic at the urban background site in Zagreb is less pronounced compared to other urban background sites in Europe. The noticeably lower summer values reflect reduced traffic density typically observed during summer vacations, in contrast to winter when street traffic is considerably higher. The contribution of EC to PM ranged from 3% to 11% in both fractions, with the highest values observed in the autumn, followed by winter, spring, and summer (Figure 2 and Figure 3), suggesting a minimal contribution of traffic emissions to the overall aerosol content [50]. The OC/PM ratio was significantly higher than the EC/PM ratio, which is consistent with findings from previous studies [52]. The mean OC/PM ratios in PM₁ and PM_2.5 were 0.35 ± 0.07 and 0.34 ± 0.07, respectively, aligning with other studies [41,53]. While the contribution of OC to PM showed modest seasonal variation (ranging from 31 to 38% in both fractions) (Figure 2 and Figure 3), statistical analysis revealed significant differences (p < 0.05) between spring, summer, and winter, as well as between winter and autumn. These differences suggested variations in the contributions of different sources across seasons.

In PM_2.5, the highest levels of LG were observed during the winter season, with a concentration of 1314 ± 726 ng m⁻³, decreasing in the spring (439 ± 186 ng m⁻³), autumn (69 ± 65 ng m⁻³), and summer (20 ± 14 ng m⁻³). MNS and GA followed the same seasonal pattern, while their concentrations were significantly lower, with GA often being below the detection limit. A similar seasonal trend was observed in PM₁, with average LG mass concentrations reaching 931 ± 562 ng m⁻³ in the winter, 414 ± 289 ng m⁻³ in the spring, 57 ± 49 ng m⁻³ in the autumn, and 8 ± 15 ng m⁻³ in the summer. The relative contribution of LG in total anhydrosugars exceeded 80% throughout the entire measuring period. Figure 4 presents the linear regression of anhydrosugars in PM_2.5 and PM₁, showing a strong correlation (R² = 0.983). Additionally, the data indicate that approximately 80% of the total anhydrosugars were associated with the PM₁ fraction. The observed seasonal trend was different from the one measured in 2020 in the same station where the levels of LG decreased from the winter, autumn, spring, and summer [43]. The results of this study indicate that biomass use for residential heating was more prominent in both winter and spring compared to the autumn. Such behavior corresponds to the lower average daily temperatures recorded during the spring sampling period compared to the autumn (Table S1). Comparable average winter LG levels were observed at the same site in 2020 [43] and at an industrial zone near Venice, Italy [36]. However, the concentrations were lower than those reported in Ioannina, Greece, during periods of intense residential wood burning [54]. Overall, the measured concentrations at our site were higher than those typically reported in other European cities, where levoglucosan levels generally remained below 1 µg m⁻³ [55,56,57,58,59,60]. The average contribution of LG to PM was 0.2% in the summer, 1.1% in the autumn, 2.3% in the spring, and 5.3% in the winter for the PM_2.5 fraction. In the PM₁ fraction, the content of LG was 0.04% in the summer, 1.3% in the autumn, 2.9% in the spring, and 5.4% in the winter [61]. Such portions were higher than the one obtained by Benetello et al. in the PM_2.5 fraction of an urban area (1.2 to 1.4%) [36] and slightly lower than the one registered during the heating seasons in different studies, ranging up to 6.9% [43,62]. A value of 10.5% was registered in Greece during severe winter residential wood-burning events [54]. It is worth mentioning that levels of anhydrosugars can also be influenced by meteorological conditions due to photo- and chemical oxidation as well as their semi-volatile property, which can introduce some uncertainties into the measured concentrations. The results of this study show that, despite Zagreb’s natural gas infrastructure, biomass continues to be extensively utilized for residential heating. This trend is likely driven by the increasing cost of natural gas and more environmentally friendly energy alternatives. For this reason, future actions for improving air quality should focus on (1) the promotion and stimulation of the transition of household heating to alternative energy sources, thus avoiding the use of fossil fuels; (2) improving the energy efficiency of buildings; (3) replacing old stoves with new ones that have minimal emissions of pollutants; and (4) dealing with agricultural waste by mulching and composting rather than burning.

3.2. Source Diagnostic Based on Elemental and Organic Carbon

Linear regression between OC and EC showed a positive correlation, with varying coefficients of determination (R²) across different seasons (Figure 5). These varying coefficients suggest that different sources contribute to the presence of OC and EC in aerosols across seasons. Moreover, the weak correlations indicate that, even in the same season, OC and EC most probably originate from distinct sources and follow different formation mechanisms. This also suggests that, in addition to primary sources, a portion of OC is produced from secondary sources as well. The greatest variability in sources appears when comparing the spring and winter seasons to the summer and autumn (Figure 5). Seasonal differences in aerosol formation and composition are influenced by the unique activities and environmental conditions associated with each season and location. Although the site is located in an urban area, its proximity to residential homes and forested regions leads to biomass combustion for both residential heating and vegetation-clearing activities. Due to different biomass burning activities throughout the year, our aim was to investigate the seasonal pattern of organic species.

Although there was no clear boundary between the emission sources of EC and OC, the OC/EC ratio can provide insights into the type of fuel being combusted, with higher OC/EC ratios typically associated with biomass burning [63,64] and lower ratios linked to fossil fuel combustion [33]. In some studies, this ratio has also been used to identify the type of biomass burned; for example, Schmidl et al. reported OC/EC ratios between 2.6 and 5.7 for residential wood burning using Austrian biofuels [65], while McDonald et al. found ratios of 3.9 for softwood and 7.9 for hardwood burned in a woodstove [66].

The highest average OC/EC ratios (Tables S2 and S3) in both fractions were observed in the summer (12.2), followed by decreasing values in the winter (9.0), spring (5.6), and autumn (4.7). These values were notably higher than those reported by Viana et al., who obtained OC/EC ratios in the range of 3.1–4.7 during the winter and 2.6–3.5 during the summer at urban background sites in Amsterdam, Barcelona, and Ghent [50], or those reported for an urban industrial site in Venice, Italy, with a value of 4 in the winter and 1.7 in the summer [36]. Similar seasonal trends with comparable OC/EC values (7.7–15.0) were reported at a rural background site in Croatia [67].

The highest OC/EC ratios were observed during the summer, most likely due to increased SOA formation, which was in accordance with enhanced photochemical oxidation and higher temperatures (Table S1). The ratio of 12.2 in the summer can be compared to those observed by Tian et al. during wheat straw combustion experiments in a chamber [68]. Despite the high ratios, the low concentrations of biomass burning tracers (LG, MNS, and GA) observed during the summer suggest that another source of organic carbon, unrelated to biomass burning, dominates airborne particles during this period. The POC and SOC portions within the OC for both PM fractions are presented in Figure 2 and Figure 3. In the PM_2.5 fraction, the seasonal SOC contribution was highest in the winter and decreased in the summer, autumn, and spring, while in PM₁, its contribution decreased sequentially from the winter to the summer, spring, and autumn. The calculated SOC values did not follow the same trend as the OC/EC ratio. These results were most likely influenced by generally lower (OC/EC)_min ratios in the winter (2.3) compared to the summer (4.2), driven by lower EC levels, reflecting a reduced contribution from primary sources, such as traffic and biomass burning. A similar behavior, with higher SOC values in the winter than in the summer, was reported by Ma et.al. (2024), where such a pattern was explained by static atmospheric conditions, lower temperatures, and higher humidity, which promoted the phase transition of semi-volatile and intermediate volatile organic compounds (SVOCs/IVOCs) from gas to particle form [69]. However, estimations of POC and SOC mass concentrations serve to assess potential uncertainties mainly on the assumption that primary OC contributions are negligible and that the OC/EC primary ratio remains constant for each season. If contributions of primary non-combustion sources, such as biogenic OC, are significant, the calculated SOC concentrations may be overestimated [70].

The PM₁ fraction showed similar contributions from carbon sources during the summer and winter. However, compared to PM_2.5, differences emerged in the spring and autumn, where the dominance of POC and SOC was reversed. Particle size distribution in urban environments is influenced by diverse sources, such as traffic emissions, residential heating, cooking, and secondary aerosol formation. Traffic mainly contributes fine and ultrafine particles, while biomass burning and cooking produce a broader range of particle sizes [71]. Residential heating releases coarser particles during colder months, and secondary organic aerosols form across various sizes, particularly in humid and low-temperature conditions [72,73]. Nevertheless, to define how the different sources were distributed across different particle sizes, additional studies of source apportionment should be carried out.

3.3. Water-Soluble Organic Carbon and Anhydrosugars in Biomass Burning Source Determination

To distinguish biomass and fossil fuel contributions to OC in PM, water-soluble organic carbon was analyzed, despite its potential origins from various sources, such as biogenic emissions, VOC oxidation, soil resuspension, and fuel combustion [7,74].

Both WSOC and OC followed the same seasonal pattern (Tables S2 and S3), with concentrations increasing from the autumn to the summer, spring, and winter. Linear regression showed a strong correlation between WSOC and OC in PM_2.5 (R² = 0.88) and PM₁ (R² = 0.75). The results indicated that WSOC and OC were generally higher in colder seasons, likely due to increased biomass burning during the colder months. A pronounced seasonal variation in WSOC content in OC was observed in both particle size fractions (Figure 6a), indicating seasonal differences in source contributions. The average WSOC/OC ratio was lower in PM_2.5 (0.49) than in PM₁ (0.56). Similar seasonal patterns were reported by Du et al., with a WSOC/OC ratio of 0.61 in the summer at an urban residential site, averaging 0.4 in other seasons [75]. In Asia, values between 0.4 and 0.55 are usually observed at sites highly affected by biomass burning sources [76,77]. In Sweden, rural sites had a ratio of 0.61, while urban sites had a ratio of 0.48 [37]. Lower ratios were recorded at urban background sites in Barcelona (0.33–0.34) and Ghent (0.40–0.42) and a traffic site in Amsterdam (0.34), with higher ratios consistently observed in the summer compared to the winter [50]. Increased WSOC/OC ratios in the summer are attributed to enhanced photochemical oxidation, which increases water solubility by forming oxygenated functional groups during warmer months [78].

In contrast, the WSOC/OC ratio observed in the PM₁ fraction followed a different variation, with the highest ratio in the autumn (0.72), followed by summer (0.64), spring (0.44), and winter (0.43). Some studies on PM₁ reported less pronounced seasonal fluctuations but still found the ratio to be lowest in the winter (0.41), which was attributed to a higher contribution of water-insoluble organic compounds during that period [18]. These findings suggest that during the autumn, there may have been an episode of enhanced WSOC emissions, primarily associated with particles smaller than 1 µm. In addition to photochemical activity, elevated biogenic emissions could also promote the production of WSOC during warmer months, whereas winter sources are more dominated by water-insoluble organic matter from biomass and fossil fuel combustion [79].

When comparing the ratios in both fractions, it was observed that the WSOC/OC ratio was higher in PM_2.5 than in PM₁ only during the winter season. According to the literature, lower winter temperatures also favor the condensation of organic compounds onto pre-existing particles, leading to increased OC mass concentrations and a higher OC fraction in larger particles [80]. This suggests that, during winter, WSOC compounds may be more associated with larger particles that can originate either from aggregation processes due to higher humidity or by increased emissions of larger particles from biomass burning [81].

The mass concentration trends of LG and WSOC, illustrated in Figure 6b, showed elevated values during both the spring and winter, corresponding to the lower air temperatures in these seasons compared to the warmer summer and autumn months (Table S1). These findings suggest that domestic wood combustion contributes to WSOC content during colder periods. However, in the warmer seasons, despite low LG levels, WSOC concentrations did not decrease substantially, indicating the presence of alternative WSOC sources and suggesting a shift in dominant contributors during these periods.

The WSOC/OC and LG/WSOC ratios exhibited different seasonal patterns, as shown in Figure 7. The LG/WSOC ratio progressively increased from the summer (0.011) to the autumn (0.128) and spring (0.183), with the highest average values during the winter (0.365). A similar trend was observed for the contribution of LG to OC, with values of 0.6%, 3.4%, 7.3%, and 13.4% for the summer, autumn, spring, and winter, respectively. The LG/OC ratio in atmospheric aerosol ranges up to 20%, depending on the characteristics of the site. In Hungary, an LG/OC ratio of 18% was reported during the winter in urban aerosol [82], while in Salzburg and Graz, Austria, a ratio of 2–3% was recorded at sites with different characteristics [83]. For comparison, significantly lower contributions of LG (up to 5%) were reported in Graz and Salzburg during the winter season [83]. In the Tuscany region, the annual contribution of LG varied between 0.04% and 9.8% [23].

The strong correlation between LG and OC during the winter season (R² = 0.902) indicated that biomass burning was a major source of organic carbon at the urban background site. In other seasons, the correlation between LG and OC was significantly weaker: R² = 0.397 in the spring, R² = 0.101 in the summer, and R² = 0.065 in the autumn. The correlation between LG and WSOC was generally weaker (R² = 0.058–0.525), suggesting that biomass burning was not a major contributor to the water-soluble content of PM in this case. Instead, the source of WSOC during the summer was most likely related to compounds formed through enhanced photochemical oxidation, which aligns with our findings for the SOC ratio. The contribution of SOC to OC for the summer season was 60% in PM_2.5. Potential precursors for secondary organic compounds included VOCs [80], water-soluble HULIS, which contributed significantly to the water-soluble fraction of PM [84], and amino acids and proteins introduced by biological activities and emissions [85].

As shown in Figure 7, the LG/WSOC ratio exhibited four distinct peaks during the measurement period: one in the autumn and three in the winter. This pattern suggested more intense biomass burning episodes, deviating from typical seasonal variability. The elevated autumn peak may be attributed to agricultural activities such as the burning of crop residues, while the winter peaks were most likely linked to increased residential wood burning. Noticeable deviations from typical variability were also observed in the LG/PM ratio, particularly during the winter, across ten distinct days throughout the season. However, the most pronounced deviations occurred on two consecutive days at the very end of the measurement period. To assess whether these elevated LG/PM ratios were affected by long-range transport, meteorological data (temperature and wind direction) and back trajectory simulations using the NOAA HYSPLIT model were analyzed (Figure 8). An average temperature drop from 6 °C to 0 °C within a two-day period most likely resulted in intensified residential heating, contributing to the observed increase in PM_2.5 (from 16 to 43 µg m⁻³) and PM₁ (from 13 to 25 ug m⁻³) concentrations. Back trajectories that can potentially explain elevated values and give the most reasonable evidence based on the possible long-range transport and chemical stability are shown in Figure 8a, where air mass circulation covered an approximate area of 200 km, crossing the border shared with Bosnia and Herzegovina, then moving toward Gorski Kotar—a rural mountain area where wood is commonly used for residential heating—before returning in the direction of the measurement site. This effect was much less pronounced on other days with an elevated LG/PM ratio, even on days with lower temperatures (e.g., −3 °C). On these days, back trajectories indicated air masses arriving from the north (Figure 8b), passing through the north continental region of Croatia, where an extensive gas distribution network likely reduced the reliance on wood for heating. This conclusion is further supported by wind rose data, which show that the highest LG mass concentrations were observed when WSW wind was most prevalent (Figure 9).

Although diagnostic ratios such as OC/EC, LG/OC, LG/WSOC, and WSOC/OC have been utilized to estimate the extent of biomass burning in particulate matter in the air, these ratios can be influenced by various factors, including combustion type, combustion conditions, and meteorological conditions. That is why there could be some uncertainties regarding the levels of this compound and, consequently, the diagnostic ratios.

3.4. Statistical Analysis of Source Contribution Study

To confirm the relationship between the biomass burning tracer, i.e., levoglucosan and the carbonaceous species EC, OC, and WSOC in PM_2.5 and PM₁, their intercorrelation through different seasons was analyzed using Spearman’s correlation tests. Tables S4 and S5 present the correlation matrix for carbonaceous compounds in both fractions through different seasons. The results confirmed notable seasonal variations in the correlations, indicating the influence of various sources on the measured levels of PM and carbonaceous species.

The correlations observed during the winter season were statistically significant and very strong between PM and WSOC, EC, OC, and LG in both fractions, suggesting that they shared the same predominant source. A moderate correlation was observed between WSOC and EC, confirming that during the winter, WSOC originates from secondary sources, in addition to primary sources. Weaker correlations were observed for all the other seasons. During the spring, a very strong correlation was noted only between PM_2.5 and OC, while EC, OC, and LG exhibited a moderate correlation. Although LG levels were elevated (439 ng m⁻³) during the spring compared to the autumn and the summer, the absence of strong correlations with OC or WSOC suggests that biomass burning is not the dominant source of OC. However, a moderate correlation was observed between LG and EC in both fractions, indicating some contribution of biomass burning to EC formation. During the summer and autumn, seasonal correlations between compounds were much weaker. A strong correlation was observed between PM and OC, while the correlation between OC and WSOC was also moderate; in addition, the correlation between OC and EC was weak and even negative between OC and LG. These results indicate that the main sources of OC were not related to biomass burning. PCA analysis (Figure S1) confirmed the observed correlations: during the spring, summer, and autumn, LG and EC were grouped together, while OC, EC, and PM were grouped separately. In contrast, during the winter season, all compounds were grouped together, indicating a shared dominant source of aerosols.

3.5. Wood Type Prediction

Given that LG is primarily derived from cellulose, while MNS and GA originate from hemicellulose, the ratios of LG/MN and LG/(MNS + GA) are frequently employed to identify the specific type of biomass being burned. In the literature, the ranges reported for softwood combustion range from 2.6 to 10 [86], while higher ratios, ranging from 13 to 24, are associated with hardwood combustion [83], 25 to 50 for the combustion of herbaceous materials, above 40 for the combustion of crop residues [65], and 31 to 90 for lignite combustion [28]. The average LG/MN ratios obtained in this study during the autumn and winter were 11.7 and 11.5, respectively, indicating the burning of softwood and hardwood, which could have come from residential heating during the winter and from crop residue burning during the autumn season. In the spring and summer seasons, the ratios were 5.9 and 9.2, respectively, as previously reported for softwood combustion.

4. Conclusions

This study investigated the impacts of biomass combustion on the composition of urban aerosols by analyzing anhydrosugars, WSOC, EC, and OC in two particulate matter fractions, namely, PM_2.5 and PM_1. Significant seasonal variations of both PM_2.5 and PM₁ concentrations suggest different sources in different seasons, with the highest average levels observed during the winter and the lowest in the autumn. Seasonal activities, including increased traffic and residential heating in the winter, elevated biological activity during warmer seasons, and varying meteorological conditions influenced the aerosol composition. The diverse sources of OC and EC were evident across seasons. The results indicate complex interactions between primary emissions and secondary formation processes, showing that OC/EC ratios alone are unreliable indicators of increased SOA. Biomass burning activities were detected year-round; however, the levels of anhydrosugars suggested that biomass burning was primarily related to residential heating rather than agricultural activities, which are more typical during warmer seasons. Distinct seasonal trends in LG and WSOC mass concentrations confirmed the presence of alternative WSOC sources and shifts in dominant contributors. More intense biomass burning episodes disrupted typical seasonal LG/WSOC trends, particularly in the autumn and several times during the winter. Meteorological analysis, including temperature trends, wind patterns, and back trajectory modeling, revealed that elevated LG/WSOC ratios often occurred on days when air masses circulated through regions with high biomass usage, suggesting that LG levels were also influenced by long-range transport. The combustion of a mixture of softwood and hardwood was predominant during the autumn and winter, while softwood combustion was more common in the spring and summer. Spearman’s correlation tests and PCA analyses highlighted strong correlations and likely common sources of LG, OC, EC, WSOC, and PM during the winter, whereas different sources dominated in other seasons. Given the significant role of biomass burning in overall air pollution based on these findings, there is a need for further source apportionment studies to better understand the diverse origins and formation mechanisms of organic species contributing to PM across seasons.