Six prescribed fires were conducted on two types of shrub species (Genista hispanica subsp. occidentalis (hereafter Genista) and Calluna vulgaris (hereafter Calluna)) to characterise the emissions (particulates and gases) from the burning process. Four quartz filters were sampled during Genista burning and two during Calluna burning (∼1000 m² were burned in each plot). The fires were carried out in La Cueta, León (NW Spain), within a protected natural area (“Natural park of Babia y Luna”) on 3 and 4 October 2016 (Fig. 1). Several sampling and monitoring instruments were used: (i) a low volume TECORA ECHOPM sampler operated at a flow rate of 38.3 L min⁻¹ to collect PM_2.5 onto quartz filters; (ii) Tedlar bags to sample smoke for further analysis; (iii) CO and CO₂ Combo IAQ Meter; and (iv) a weather station Geos N11 to record meteorological variables at 1.85 m above the ground, at 10 s intervals; (v) an infrared camera type Flir Systems to obtain the characteristics of the fires (mean rate of spread, mean temperatures and residence time). The sampling duration for each PM_2.5 filter was adjusted to the duration of the smoke plume generated by each prescribed fire and ranged from 6 to 15 min. Four PM_2.5 filter samples were collected during Genista burning and two during Calluna burning. Smoke samples for gaseous analysis were collected in Tedlar bags during the active combustion period, avoiding the entrainment of background air as far as possible. The PM_2.5 sampler, Tedlar bag sampling system, CO/CO2 monitor, meteorological station, and infrared camera were positioned at ground level approximately 10 m downwind of the fire front.

We conducted various chemical analyses to characterise the particulate matter and gases emitted during the burning:

Elemental analysis of Calluna and Genista fuel samples before burning was carried out by combustion at high temperatures, separation of the gases produced and detection by TCD (Dumas Method) with an accuracy of 0.2 %. The parameters analysed were total humidity, volatiles, ash at 550 and 815 °C, C, H, N, and S content, Low Calorific Value (LCV), and High Calorific Value (HCV). The analysis was carried out according to the current regulations (ASTM International, 2016, 2018, 2019; UNE, 1984a, b, 1995, 2010).

The parameter used to characterise combustion is Modified Combustion Efficiency (MCE). It assesses the completeness of combustion, taking into account that >90 % of the carbon combusted is emitted in the form of CO₂ and CO and <10 % of the carbon is emitted in the form of hydrocarbons and particulate carbon. The MCE, expressed as a percentage, is calculated as (Eq. 1): 1MCE(%)=CO2CO2+CO⋅100 If MCE value is higher than 90 %, it indicates that more than 50 % of the emissions were produced by flaming combustion, whereas if MCE is lower than 90 %, it suggests that more than 50 % of the emissions were caused by smouldering combustion (Ward and Hardy, 1991).

The emission factor (EF) is a parameter that relates the emission of a particular species to the amount of biomass burned. The carbon combusted during biomass burning is emitted as CO₂, CO, CH₄, non-methane hydrocarbons (NMHC), and particulate carbon (EC + OC). Therefore, the EF of a specie n, is estimated from the ratio of n mass concentration to the total carbon concentration emitted. Thus, EF is defined as the amount of a compound released per amount of dry fuel consumed, expressed in units of gkg-1, where the above ratio is multiplied by the mass percentage of carbon in the fuel (Reid et al., 2005) using Eq. (2). The term NMHC is less than 2 % and includes the sum of methanol, ethane, propene and acetylene, while the sum of CO₂ and CO is 8792 % of the total carbon emitted (Urbanski et al., 2008). The carbon mass fraction of the dry fuel used in Eq. (2) was obtained from the elemental analysis described in Sect. 2.2 and reported in Table 2. The %Cfuel values were 53.3 % for Calluna and 53.2 % for Genista. 2EFn=nCO2+CO+CH4+NMHC+EC+OC⋅%Cfuel

Particle morphology was analysed by scanning electron microscopy (SEM). PM₁₀ polycarbonate filter membranes were mounted on aluminium SEM stubs using conductive carbon tape and sputter-coated with a 1–2 nm thick Au/Pt layer to minimize charging effects. Sampling periods for SEM analysis were kept short and adjusted according to smoke-plume intensity to minimize filter overloading and particle overlap on the substrate, which is essential for reliable single-particle SEM/EDS analysis (Mamane et al., 2001).

Observations were performed using a field-emission SEM (JEOL JSM-6335F) equipped with an Oxford Instruments X-Max energy-dispersive X-ray spectroscopy (EDS) detector at the Spanish National Center for Electron Microscopy (ICTS), Complutense University of Madrid (UCM). Multiple random SEM fields of view were inspected for each sample to ensure that the reported morphologies were representative of the overall particle population. This approach has been previously applied in similar studies (Coz et al., 2008).

Aerosol and gas sample characteristics depend on: (i) meteorological conditions, (ii) biomass properties, and (iii) combustion dynamics (Pio et al., 2008): i.

Calluna burnings were carried out at an altitude of 1534 m with a mean wind speed of 4.3±1.7 m s¹, a mean temperature of 17.7±0.6 °C and a relative humidity (RH) of 40.1±1.1 %. Meteorological variables were similar for Genista burnings (at 1395 m) with a mean wind speed of 3.6±1.8 m s¹, a mean temperature of 19.7±0.8 °C and a RH of 37.4±1.3 %. Meteorological conditions during the fires are therefore comparable. Table 1 summarises the recorded meteorological data.

The PM_2.5 concentrations registered in the smoke plume for the burnings of Calluna and Genista were 31.1±13.1 mg m⁻³ and 12.0±6.1 mg m⁻³, respectively. Carbonaceous constituents (TC) accounted for 48.6 % and 30.7 % of the mass of PM_2.5 from Calluna and Genista combustion. Concentrations of organic and elemental carbon during Genista fires were 2.05±1.51 and 0.60±0.41 mg m⁻³, respectively, while for the Calluna fires the levels were 9.94±2.97 and 1.40±0.98 mg m⁻³. It is noteworthy that the OC/EC ratio was more than twice as high for Calluna (8.4±3.8) compared to Genista (3.2±1.6). For Calluna, the sum EC+OC represented 28.1 % of PM_2.5, while for Genista it represented 32.9 %. These percentages are lower than those reported by Andreae (2019) for savanna (52.6 %) and temperate forest (61.9 %).

Net concentrations (concentrations measured in the smoke plume minus the corresponding background concentration) of gases obtained through high-resolution Fourier transform infrared (FTIR) analysis of combustion products collected from Calluna and Genista burnings are shown in Table 3. CO₂ was by far the most abundant gas emitted during combustion, with mean concentrations of 653 ppm for Calluna and 54 ppm for Genista. Of the total measured carbon emitted during combustion, carbon dioxide (CO₂) was by far the dominant component, representing 89 % in Calluna fires and 65 % in Genista fires. These results confirm that Calluna combustion was more efficient and dominated by flaming phases, while Genista fires exhibited incomplete combustion associated with smouldering conditions. The higher proportion of CO and CH₄ in Genista is consistent with lower modified combustion efficiency (MCE), as previously observed in low-intensity fires (Ward and Hardy, 1991).

Interestingly, despite its higher combustion efficiency, Calluna emitted more TOC than Genista. This may be attributed to intense flaming combustion, which enhances the volatilisation of organic compounds and the release of oxygenated hydrocarbons (Yokelson et al., 2013).

The remaining measured gaseous species – including nitrogen oxides (NO and NO₂), ethylene (C₂H₄), ethane (C₂H₆), hydrogen fluoride (HF), and formaldehyde (CH₂O) – were present at concentrations typically below 1 % of the total gas mixture. Although their contribution to total carbon emissions was minimal, these compounds are atmospherically relevant due to their roles in ozone formation, secondary aerosol production, and potential health effects (Andreae, 2019). Notably, NO and NO₂ concentrations were slightly higher in Genista fires, while ethane, formaldehyde, and hexane showed significantly greater concentrations in Calluna fires, indicating differences in combustion conditions and fuel composition. The elevated levels of light hydrocarbons such as ethane and hexane in Calluna emissions point to higher pyrolysis activity and faster fuel volatilisation, typical of more intense flaming combustion (Urbanski, 2014).

Calluna / Genista concentration ratios were calculated for all measured gases. The largest differences were observed for CO₂ (12.1), hexane (6.7), CO (3.0), TOC (2.6), CH₄ (2.3), and ethane (2.5). These results reinforce the idea that plant species exert a strong influence on the gaseous composition of emissions, even under similar prescribed fire conditions. Such differences should be considered in emission inventories and atmospheric models for Mediterranean shrublands (Alves et al., 2010a; Andreae, 2019).

The emission factors (EF) of carbonaceous species: CO, CO₂, CH₄, C₂H₆, C₂H₄, C₃H₈, C6H14, CHOH, organic carbon (OC), elemental carbon (EC) and PM_2.5, for Calluna and Genista burning were calculated as indicated in Sect. 2.3 (Table 3). The EF_CO2 for Calluna (1703 g kg⁻¹) is similar to that obtained for an Amazonian forest clearing fire and higher than that obtained for Genista fires (592 g kg¹). Soares Neto et al. (2009) carried out a study in Mato Grosso (Brazil) in the deforestation arc and reported CO₂ EFs for different combustion stages of Amazonian Forest clearing fires: ignition (1631–1625 g kg⁻¹), flaming (1690–1741 g kg¹) and smouldering (1540–1548 g kg⁻¹). (Andreae, 2019) reported an EF_CO2 of 1660±90 g kg⁻¹ for savanna and grassland.

A similar pattern between the shrubs species was obtained for the values of EF_PM2.5 and EF_OC. The EF_PM2.5 values obtained for Calluna and Genista were 53.5 and 44.7 g kg⁻¹, respectively. Andreae (2019) in a compilation of emission factors from 370 studies across 121 species and biomass types, reported average EF_PM2.5 values of 6.7±3.3 g kg⁻¹ for savanna and grassland and 18.5±14.4 g kg⁻¹ for temperate forest. In addition, contrary to what was observed by Alves et al. (2010b), whom carried out an analysis of emissions from prescribed fires in a shrub-dominated forest with some pine trees in the Lousã Mountain (Portugal) in 2008, EF_PM2.5 was higher in prescribed fires here than in wildfires reported by Alves et al. (2010b), and EF_CO2 was either similar (Calluna) or lower (Genista), suggesting less efficient combustion in the latter.

The pattern of EF_CO, EF_CH4 and EF_EC is opposite to that of EF_PM2.5 (Fig. 2) with higher values for Genista than for Calluna. The EF_CO2 and EF_CO are explained by differences in combustion efficiencies, flaming in Calluna and smouldering in Genista fires. Typically, CO and PM_2.5 emissions increase as combustion efficiency decreases (Ward and Hardy, 1991). However, the higher EF_PM2.5 observed in Calluna fires (53.5 g kg⁻¹) suggests that intense flaming combustion may have enhanced the release of fine particles, possibly due to higher volatile content or fuel characteristics that promote aerosol formation during flaming. Other studies reported a clear correlation between MCE and EFs (Alves et al., 2010b; Soares Neto et al., 2009).

Regarding EF_CO, Bertschi et al. (2003) conducted a laboratory study to evaluate smouldering combustion of tropical wooded savanna and obtained EF_CO values similar to the of the present research (128–165 g kg⁻¹). Similarly, Yokelson et al. (2009) characterised emissions from deforestation and crop residue burning, reporting EF_CO values of 75±26 and 83±14 g kg⁻¹, respectively, lower than the values observed here (Calluna: 113 g kg⁻¹; Genista: 155 g kg⁻¹).

Additionally, the EF_CO and EF_CH4 values for both shrubs are higher than those typically reported for savanna and grassland, and closer to those found in temperate forests. The EF_EC values for Calluna (1.85 gkg-1) and Genista (3.32 gkg-1) also exceed the averages reported for savanna and grassland (0.53±0.35 gkg-1 and temperate forest (0.55±0.36 g kg⁻¹) (Andreae, 2019).

The remaining carbonaceous species – ethane, ethylene, hexane, and formaldehyde – showed higher EFs in Genista than in Calluna, reflecting enhanced pyrolysis under smouldering conditions (Andreae, 2019; Yokelson et al., 2009). Although their absolute values were low (<7 g kg⁻¹), their atmospheric relevance lies in their contribution to ozone formation, secondary aerosol production, and toxicological impacts (Alves et al., 2010a; Bertschi et al., 2003).

A more regionally specific comparison is provided by the experimental Mediterranean shrubland wildfires reported by Garcia-Hurtado et al. (2013), who obtained emission factors of 1257±40 gkg-1 for CO₂, 453±28 gkg-1 for CO, 46±12 gkg-1 for CH₄, 3.4±1.39 gkg-1 for PM_2.5, 1.4±0.34 gkg-1 for OC, and 0.051±0.028 gkg-1 for EC. Compared with those values, EFCO₂ in the present study was higher for Calluna (1700 gkg-1) and lower for Genista (592 gkg-1), which is consistent with the different combustion efficiencies observed for the two fuels. In contrast, EFCO and EFCH₄ were substantially lower in both prescribed fires than in Garcia-Hurtado et al. (2013), consistent with the strong smouldering contribution reported in their experiments. Conversely, EFPM_2.5, EFOC, and EFEC were markedly higher in the present study, particularly for PM_2.5 and EC. These differences highlight the influence of fuel type, combustion phase, fire behaviour, and sampling strategy on emission factors, even within Mediterranean shrubland ecosystems.

As observed for gaseous compounds, most water-soluble ions in PM_2.5 from Calluna combustion – with the exception of Cl⁻ and Na⁺ – were present at higher mass fractions than those emitted by Genista (Fig. 3). Overall, water-soluble ions represented 6.9 % of the PM_2.5 mass in Calluna and 4.5 % in Genista. The elevated ionic content in Calluna-derived particles may reflect a higher ash yield from this fuel. Similar proportions were reported by Alves et al. (2010b) for prescribed shrubland fires in Portugal. Although potassium (K⁺) is often used as a tracer of biomass burning, no clear correlation was observed between K⁺ and either organic carbon (OC) or elemental carbon (EC) in this study.

Na⁺, SO42-, Mg²⁺, Cl⁻ and Ca²⁺ were the most abundant water-soluble inorganic ions emitted during the burning of Genista, in that order. In contrast, the same ionic species dominated Calluna emissions, but with a different ranking: Cl⁻, Na⁺, SO42-, Mg²⁺ and Ca²⁺ (Andreae et al., 1998). reported that Cl, SO42-, K⁺ and NH4+ were the predominant ions in savanna fires (similar to Calluna), while Falkovich et al. (2005) indicated that the predominant ionic species in Amazon basin fires were SO42-, NH4+, K⁺, NO3- and Cl⁻.

The K⁺ mass fraction (0.21±0.04 wt % for Calluna and 0.32±0.20 wt % for Genista) was very close to the values reported for fires in a Mediterranean shrubland (Alves et al., 2010b), but lower than typical values for savannas and more comparable to those observed in temperate forests. Laboratory studies have shown a wide variability in K⁺ fractions, ranging from <1 % to 24 % (Andreae, 2019).

The Cl⁻ / EC ratios were 3.4±2.0 for Genista and 4.2±2.1 for Calluna, while the K⁺ / EC ratios were 0.104±0.071 and 0.241±0.045, respectively. The presence of K⁺ relative to EC supports the influence of biomass combustion, as water-soluble potassium is widely used as a tracer of biomass-burning aerosol (Andreae, 2019; Yu et al., 2018). Together with the observed Cl⁻ levels, this suggests the possible presence of potassium-containing salts such as KCl internally mixed with carbonaceous particles. These species may act as cloud condensation nuclei (CCN), contributing to regional indirect radiative forcing in Mediterranean regions (Alves et al., 2010a; Petters et al., 2009).

However, it should be noted that comparisons across studies may be affected by multiple factors, including: (i) biome characteristics of each site; (ii) variability in combustion efficiencies or (iii) sampling (ground-based measurements, mast, distance to the fire, PM inlets or different devices).

The ionic composition and particle-phase characteristics reported here may also support the refinement of next-generation air quality models. Recent AI-based tools for simulating PM_2.5 transport and chemical aging from prescribed fires (Liao et al., 2025) highlight the importance of comprehensive field emission datasets like those obtained in this study.

Scanning electron microscopy (SEM) analysis showed that PM samples collected during prescribed burning episodes were dominated by particles with irregular morphologies and complex aggregated structures, typical of biomass burning aerosols (Fig. 4). The particles shown in Fig. 4 are representative of the dominant morphologies observed across the analysed samples. A minor fraction of the particulate matter could be tentatively identified as soot-like carbonaceous agglomerates; however, these structures were frequently partially or completely obscured by thick surface coatings, preventing the clear observation of their characteristic open fractal geometry when present.

The particle population was largely dominated by spherical and sub-spherical particles (Fig. 4a–g), resembling so-called tar balls (Pósfai et al., 2004), with sizes below 100 nm to several cents of nanometers. The presence of tar balls is commonly associated with the condensation of semi-volatile organic compounds and the rapid cooling of combustion vapours in fire plumes. Qualitative inspection of multiple SEM fields of view indicated that these spherical and compact morphologies accounted for most of the observed particles. These particles exhibited smooth to slightly textured surfaces and sizes predominantly in the submicrometer range. Several particles displayed hollow or shell-like morphologies (Fig. 4h–i), suggesting rapid gas-phase condensation and subsequent volatilization or restructuring, possibly during atmospheric transport; however, changes can occur during the analysis. In addition, irregularly shaped and porous aggregates were observed, likely corresponding to internally mixed particles composed of organic material, ash residues, and inorganic inclusions. These mixed particles frequently showed thick coatings and poorly defined boundaries between phases, indicating extensive internal mixing and potential atmospheric aging (Li et al., 2011).

Energy-dispersive X-ray spectroscopy (EDS) revealed a dominant contribution of carbon and oxygen, together with variable amounts of potassium, silicon, calcium, aluminium, and iron, often forming part of the coated mixed particles. The frequent presence of potassium- and calcium-containing inclusions within the aggregates further supports the strong influence of biomass combustion, as these elements are well-established tracers of wood and vegetation burning emissions. The predominance of spherical, compact, and internally mixed particles with thick organic coatings is expected to influence both the optical properties and hygroscopic behaviour of the aerosol. Such coatings can modify particle refractive index and morphology, by enhancing light absorption through “lensing effects” when carbonaceous cores are present and increasing the particles' optical thickness (Bond et al., 2013; Moffet and Prather, 2009). Additionally, the internal mixing of inorganic salts (K, Ca) with organic matter suggests a transition toward higher hygroscopicity during atmospheric aging (Coz et al., 2008), by changing their effective chemical composition and surface properties, thereby promoting water uptake at elevated relative humidity and enhancing cloud condensation nuclei (CCN) activity. These effects are likely to play an important role in controlling the radiative and microphysical impacts of aerosols from biomass burning during transport and aging in the atmosphere (Carrico et al., 2010; Pósfai et al., 2004).