We determined concentrations of monosaccharide anhydrides, sugar alcohols, 2-methyltetrols, monomeric, and dimeric sugars in PM10 filter samples using ultra-performance liquid chromatography (UPLC) (Vanquish UHPLC, Thermo Fisher Scientific) in combination with Q Exactive™ Plus Orbitrap (Thermo Fisher Scientific) operated in the negative electrospray ionization (ESI) mode, with a resolution of 70 000 FWHM (full width at half maximum) at 200 Da.

We added an isotopically labeled internal standard to filter punches (2×1.5 cm2) which were submerged in precleaned tetrahydrofuran (THF) (2 mL) in separate screw-neck amber glass vials, which we subjected to ultrasonic extraction (30 min). We transferred the solute to a centrifuge tube using pipetting and repeated this step twice. Afterward, we evaporated the solute to 0.4 mL, spun it (10 min; 2000 rpm, revolutions per minute), and transferred and evaporated it to dryness in a screw-neck amber glass vial. The sample volume was redissolved in 0.25 mL precleaned THF/Milli-Q water (55:45) and placed in a Whirlmix before analysis. The extraction procedure was equal to that of Dye and Yttri (2005). We used two columns in series for separation (two 3.0 mm × 150 mm HSS T3, 1.8 µm; Waters Inc.), using isocratic elution (Milli-Q; 18.2 MΩ), and flushing with acetonitrile (high purity) at the end of the run. The Milli-Q water was purified using an EDS-Pak Polisher containing activated coal (Merck KGaA, Darmstadt, Germany) and a LC-Pak cartridge (Merck KGaA, Darmstadt, Germany) containing reversed-phase silica.

We identified all species based on the retention time and mass spectra of authentic standards, using isotope-labeled standards of levoglucosan, galactosan, mannitol, arabitol, trehalose, and glucose as recovery standards (Table S1 in Yttri et al., 2021). The limit of detection (LOD) was 1 to 3 pg m-3 for the monosaccharide anhydrides, 1 pg m-3 for the 2-methyltetrols, 4 pg m-3 for the sugar alcohols, 6 pg m-3 for the dimeric sugars, and 8 pg m-3 for the monomeric sugars.

We based the analysis of free cellulose on the saccharification of cellulose and subsequently quantified the glucose produced, following the method by Kunit and Puxbaum (1996). We switched the final detection of glucose from a photometric method to HPAEC PAD (high-performance anion exchange chromatography with pulsed amperometric detection), similar to Qi et al. (2020). We extracted filter aliquots with a citrate buffer (0.05 M citric acid) adjusted to pH 4 and added thymol to a final concentration of 0.05 % to prevent bacterial growth. We enhanced extraction by ultrasonic agitation. We added enzymes (Trichoderma reesei cellulase; Aspergillus niger cellobiase), which had been precleaned by ultrafiltration to reduce glucose blanks, for saccharification. We stopped saccharification (at 45 °C) after 24 h by heating the samples to 80 °C. We analyzed glucose on a Dionex ICS-3000 equipped with a CarboPac MA1 column, using a sodium hydroxide gradient reaching from 480 mM NaOH to 630 mM. We corrected the results with the free glucose contained in the samples.

We picked one–two filter samples for each month of the year from samples collected from 2017 to 2018 to capture the seasonal variability in the source composition (Table S3 in the Supplement). No valid solution was found for the samples on 5–13 December 2017 when using the Latin hypercube sampling (LHS) approach (Sect. 2.8), hence the low coverage for December (6 %) compared to September (77 %). We pooled two consecutive samples for the months of June, July, August, September, October, and December to meet the LOD (3 µgC) for EC. Moreover, we aimed for the front : back filter carbon content ratio > 3, but this criterion was not met for one of the samples in June, September, and October and for two samples in December.

We analyzed 14C–TC on both front and back filters, while 14C–EC was analyzed only on front filters. To measure 14C–EC, we used three circular punches (22 mm diameter) from the filter sample aliquot (16.6 cm2). We used the remaining front filter area (5.2 cm2) for 14C–TC analysis, along with an equivalent area of the back filter.

Annual mean levoglucosan concentrations ranged from 0.335 to 0.919 ng m-3, which is comparable to the annual mean (0.680 ng m-3) reported for Zeppelin for March 2008 to March 2009 (Yttri et al., 2014). The interannual variability was 40 %, similar to major aerosol constituents such as OC and SO42-. In 2020, the annual mean was twice as high as the mean of the previous 3 years, with an increase attributed to elevated monthly means (∼ 2 ng m-3) in February, July, and October (Fig. 3).

Increased levels and peak concentrations of levoglucosan in the H season reflected RWC emissions, as shown by Yttri et al. (2014). Increased levels in July and August were not shown by Yttri et al. (2014), partly due to missing data, although impact from wild and agricultural fires was predicted by modeling. In the present study, increased levels in July and August were a hallmark of the levoglucosan time series, pointing to the importance of WF emissions. FLEXPART model transport of modeled BC emissions also showed a substantial influence of WF emissions for July and August (2017 to 2020) (Fig. S1). The levoglucosan concentration in the 2020 NH season was ∼ 3 times higher than the average of the 3 previous years, demonstrating a pronounced interannual variability in WF influence at Zeppelin. The levoglucosan to mannosan ratio (L / M) was lower for the NH season (4.8±1.2) compared to the H season (7.5±1.9) (Fig. 3; Table S6) and might reflect a shift from WF and agricultural waste burning in the NH season to RWC in the H season. Our findings correspond with L / M ratios below 5 in summer at Gruvebadet (Ny-Ålesund) (Feltracco et al., 2020). However, we did not observe the very high L / M ratios, occasionally exceeding 40, that have been attributed to emissions from crop residue burning in Asia during the spring.

Tracer-based LHS source apportionment found that BB was the primary source of EC in all but one sample (Tables S3 and S4). On average, 61±15 % of EC was attributed to BB, with this percentage varying by season, with 67±5 % in the NH season when EC levels were low and influenced by WF, and 57±18 % in the H season when RWC dominated. Our results showed a much higher BB fraction in the H season than Winiger et al. (2019) for the H season from 2012 to 2013 (36 % to 39 %), whereas it matched that of the AH period in 2009 (57±21 %) (Winiger et al., 2015). The BB fraction in NH season was slightly higher than the 58 % to 62 % range for the NH season in 2013 (Winiger et al., 2019). Notably, differences in the sample preparation and in 14C analytical protocol should be considered, along with interannual variability, to seek an explanation for the observed differences.

The weekly maximum BB fraction of EC in February 2017 (81 %) was somewhat lower than the extremely high (95 % to 98 %) daily BB fractions during AH at Zeppelin in 2009 (Winiger et al., 2015). Although no conclusive explanation was given for the extreme values reported by Winiger et al. (2015), it cannot be excluded that BB emissions can dominate for an entire week. eBCBB apportioned by PMF (Sect. 3.2.3) supports nearly exclusive (90 %) BB contributions for 24 h (Fig. S2) but not for an entire week (80 %) (not shown). Notably, Kongsfjorden has around 20 cabins and a few research stations, and wood is used for heating in these facilities when in use. Hence, emissions from these sources cannot be excluded. The weekly minimum BB fraction for January 2018 (21 %) was much lower compared to the lowest percentage reported by Winiger et al. (2015) (39 %). FLEXPART footprints for the February 2017 and the January 2018 samples were similar, covering northwestern Russia and northeastern Greenland (not shown) and providing no further insight to their extreme values, and thus “highlights the complexity of BC in the Arctic atmosphere, where the generally low BC levels may be strongly influenced by point sources or occasional combustion practices” (Winiger et al., 2015).

The eBC (sum of eBCBB and eBCFF) and EC time series were similar, with enhanced levels during AH, a small increase in midsummer, and a slight increase towards the end of the year. FF was the major fraction of eBC annually (70±2.7 %), in the H season (71±2.7 %), and in the NH season (67±6.7 %) (Table 3; Fig. 4).

Previous modeling studies indicate that WF is the primary source of Arctic BC during summers (Stohl et al., 2013; McCarty et al., 2021). The 14C EC measurements support this for Zeppelin but not for other high Arctic observatories (Table 4; Winiger et al., 2019). In our study, 27 %–42 % of eBC was attributed to WF emissions in the NH season, which is lower than previous findings. Our estimate assumes that all eBCBB in the NH season comes from WF emissions (eBCWF) and from residential wood combustion emissions (eBCRWC) in the H season. Neither eBCWF nor eBCRWC dominated on a monthly basis (Fig. S3), although they came close in October 2017, July 2020, and October 2020, accounting for 46 % to 48 %. The annual contribution of eBCWF to eBC was estimated to be 5.4 % to 12 %, while eBCRWC contributed 20 % to 26 %, highlighting that RWC is a larger source of eBC compared to WF.

FLEXPART predicted an almost equal share of BC from BB and FF annually, whereas BCFF (56±3.6 %) dominated in the H season and BCBB (61±3.3 %) in the NH season (Table 3), similar to results found in Stohl et al. (2013). For a direct comparison, BCWF (and BCRWC) was calculated similarly from FLEXPART BCBB output as eBCWF from eBCBB by PMF; i.e., BCWF equals all BCBB in the NH season, whereas BCRWC equals all BCBB in the H season. Comparing this proxy BCWF with the FLEXPART-modeled BCWF provided a ratio of 0.97 to 1.09 for 2017 to 2020, indicating that the BCWF proxy is a sound approximation. With 16 % to 22 % of BC attributed to WF and 27 % to 36 % to RWC annually (Fig. 4; Table 3), FLEXPART concludes, in the same way as PMF, that RWC > WF but suggests higher percentages for WF and RWC fractions.

Neither PMF nor FLEXPART seem to fully reflect the predominant role of BC from WF above 50° N, which McCarty et al. (2021) suggest are larger than the emissions from anthropogenic residential combustion, transportation, and flaring combined. In 2020, 56 % of the BC emissions north of 65° N were attributed to open biomass burning by McCarty et al. (2021), whereas 12 % (PMF) and 22 % (FLEXPART) of (e)BC was attributed to WF at Zeppelin for 2020 in the present study. Spatial variability and vertical distribution of the emissions might explain part of the discrepancy, as might midlatitude emissions below 65° N, being less influenced by WF. Vertically resolved BC concentrations in the Arctic in spring and summer, based on aircraft measurements, show a decrease with increasing altitude (Jurányi et al., 2023), but this remains yet to be confirmed for BC from WF.

For the BB tracer levoglucosan, the fraction observed in the NH season (36 % to 64 %), corresponding to the WF fraction, was higher than seen for both eBCWF (17 % to 35 %) (PMF) and BCWF (32 % to 45 %) (FLEXPART). However, the degradation of levoglucosan during LRT and lack of representative (e)BC / levoglucosan ERs for a vast number of fuel categories, vegetation types, and combustion conditions implies considerable uncertainty in deriving the RWC / WF (e)BC split using this technique.

Comparing PMF results to the few samples subjected to 14C measurements and source apportionment by tracer-based LHS showed that these two approaches were on opposite ends of the scale, with FLEXPART in between (Table 4). Radiocarbon measurements and LHS estimated a BB fraction twice as high as the PMF approach, but all three methods agreed on a higher BB fraction in the NH season than in the H season. Notably, BB and FF fractions of eBC derived from PMF were more aligned with those from radiocarbon measurements at Zeppelin in 2012 to 2013 (Winiger et al., 2019) and with fractions derived from levoglucosan measurements at Zeppelin in winter 2008 to 2009 (Yttri et al., 2014). However, interannual variability makes such a comparison indicative only. Consideration of methodological differences is essential. Crucial steps of 14C–EC measurements include preventing EC loss during OC removal and avoiding OC mixing with the minor EC fraction, thus impacting its modern vs. fossil fuel signature. The advancements in the analytical approach used in this study (Rauber et al., 2023) specifically aimed to improve these critical steps. Additionally, eBC derived from Aethalometer measurements provides no information on the age of carbon undergoing combustion but reflects the wavelength dependence of the absorption linked to the combustion condition (Garg et al., 2016). Consequently, the eBCFF factor obtained by PMF could also contain emissions from combustion of biofuels, observed as modern carbon by 14C measurements. Conversely, emissions from coal combustion might contribute to the eBCBB factor. Terms like liquid fuel instead of fossil fuel and solid fuel instead of biomass burning could be more appropriate, but we maintain the notation for comparability with 14C-based apportionment and Aethalometer model studies (Sandradewi et al., 2008). Applying the Aethalometer model approach with an Ångstrøm exponent pair of (1,2) (Sandradewi et al., 2008) and (0.9,1.68) (Zotter et al., 2017), yields annual eBCBB contributions of 32 % and 29 %, supporting our more novel PMF approach (27±14 %) (Table 4). Furthermore, we note that the Ångstrøm exponents obtained from the PMF (1,1.6) are consistent with the distribution of Ångstrøm exponents observed at various sites across Europe (e.g., Tobler et al., 2021). Despite the differences in the estimated contribution size from biomass vs. fossil fuel sources between 14C–EC measurements and eBC by PMF, important conclusions drawn are unaffected. For instance, both methods indicate the presence of residential wood combustion and wildfire-sourced black carbon at Zeppelin, with a higher relative BB contribution during the non-heating season.

Nearly exclusive (90 %) contributions to eBC were seen for both eBCBB and eBCFF for periods of 24 h (Fig. S2; upper left panel). This corresponds with 24 h 14C–EC data from Zeppelin dominated (EC fbb> 95 %) by contemporary carbon (Winiger et al., 2015) and 14C–EC data dominated (EC fff> 95 %) by fossil carbon observed at other high Arctic sites (Winiger et al., 2019). Exclusive contributions were most frequent for eBCFF and seen for 1.1 % of the dataset compared to 0.1 % for eBCBB. Hence, with a few exceptions, eBCBB and eBCFF co-appear.

Annual mean concentrations of sugars and sugar alcohols were 1 to 2 orders of magnitude lower at Zeppelin compared to Birkenes, reflecting the modestly vegetated Arctic and that PBAP mainly have a local origin (Samaké et al., 2019). This contrasts with the factors for 2-methyltetrols (≤ 5), which are secondarily formed species with a stronger regional character but might also relate to the temperature sensitive high flux of biogenic volatile organic compounds for Arctic vegetation (Kramshøj et al., 2016). Higher levels of primary biological organic aerosol (PBOA) at Gruvebadet (50 m a.s.l.), 1 km south of Ny-Ålesund, compared to the Zeppelin Observatory (472 m a.s.l.) (Moschos et al., 2022), indicate a local contribution associated with the more verdant lower altitude areas. However, maximum concentrations of sugars and sugar alcohols were observed for the LRT episode 22–27 July 2020 (Sect. 3.6.1), explaining 29 % of the annual sugars and sugar alcohols loading. We are left speculating about the LRT fraction of PBAP vs. that of local origin, but LRT likely makes a larger contribution to PBAP in the Arctic than for more vegetated southerly biomes.

All species experienced a modest increase in June, coinciding with the onset of the growing season, but evolved differently after that, suggesting a mixture of sources, highlighting the importance of measuring a broad spectrum of PBAP tracers. Arabitol and mannitol were elevated throughout summer before successively declining towards the end of the year, and fructose and glucose started decreasing immediately after the peak level in July, whereas trehalose experienced comparable levels from July to November. Snow cover can be decisive for PBAP levels (Yttri et al., 2007a, b) and probably more so for the non-forested Arctic. However, our data do not explicitly demonstrate an influence of the snow cover, e.g., the seasonality of trehalose (and cellulose; Sect. 3.4.2).

The composition of sugars and sugar alcohols at Zeppelin (Table 1) and Birkenes (Table S5) varied, reflecting different biomes. Glucose was the most abundant sugar, regardless of the season, at Zeppelin. At Birkenes, glucose dominated only in winter, while arabitol and mannitol were more prominent in summer. Trehalose levels were comparable or slightly higher than arabitol and mannitol at Zeppelin but lower at Birkenes. Samaké et al. (2020) showed how only a few genera of fungi and bacteria were responsible for the sugar and sugar alcohol containing PBAP in PM10 filter samples at a rural site in France and that these were associated with leaves rather than soil material. This strong association between sugars and sugar alcohols and vegetation likely explain the very low levels of these PBAP tracers at Zeppelin compared to Birkenes. Samaké et al. (2020) point to the fungus Cladosporium sp. when explaining ambient aerosol levels of arabitol, mannitol, and trehalose, as do Yttri et al. (2007b) for Birkenes. The annual mean mannitol to arabitol ratio was comparable between Zeppelin (1.1±0.5) and Birkenes (1.0±0.0) (Table S6) and to values reported for the Nordic countries (Yttri et al., 2011a). Mannitol and arabitol were highly correlated in the NH season (r2=0.983) when levels were elevated, and the mannitol to arabitol ratio variability was minor, suggesting one common source dominating. However, four samples with a mannitol to arabitol ratio ≥ 3 in the April to May transition could indicate influence from another source. Mannitol is considered the most abundant naturally occurring polyol, which is present and produced in a wide range of living organisms (Tonon et al., 2017) and accounts for 25 % of the dry weight of macroalgae for certain parts of the year (Horn et al., 2000); however, our data for Zeppelin suggest that fungal spores are decisive for arabitol and mannitol present in the Arctic aerosol. Assuming all mannitol was associated with fungal spores, their carbon content contributed 0.5±0.2 % to OC annually when applying the lower OC / mannitol ratio (5.2) of Bauer et al. (2002), whereas the highest monthly mean was seen for September (1.5±1.2 %). The contribution reached 5 % for only two of the weekly samples. Using the higher OC / mannitol ratio (10.8) would double these estimates.

Glucose is a building block of natural dimers and polymers and a ubiquitous primary molecular energy source and thus an important PBAP. Small amounts of glucose are present in RWC emissions (Nolte et al., 2001) and are increased in air masses influenced by forest fire smoke (Medeiros et al., 2006). Notably, 9 of the 10 samples highest in glucose were also highly increased with respect to levoglucosan and were all collected in the NH season (Table S8), demonstrating WF as an important source of glucose brought to the Arctic by LRT. A largely similar finding was made for the other sugars and sugar alcohols. Previous studies do not seem to link fungal-related sugars and sugar alcohols (arabitol, mannitol, and trehalose) with WF emissions (e.g., Table 5 in Medeiros et al., 2006) or with RWC emissions; e.g., levoglucosan and sugar alcohols end up in different factors in PMF studies (Waked et al., 2014; Yttri et al., 2021). This might partly be due to lack of correlation between levoglucosan and sugar alcohols for an entire dataset. Indeed, low correlations between levoglucosan and sugar alcohols (rNHseason2<0.423; rHseason2<0.056) were obtained considering the entire dataset for Zeppelin, although the data presented in Table S8 clearly demonstrate a connection between WF and sugar alcohols.

We estimated a 7 %–15 % contribution of PBAP to OC annually, using an OC to PBAPTracers emission ratio (ER) of 14.6±2.1 (Groot Zwaaftink et al., 2022), derived from measurements in the boreonemoral zone (Yttri et al., 2021), keeping in mind that such an ER would be site specific.

Cellulose was the most abundant organic tracer analyzed (annual mean concentration of 1.6±0.4 ng m-3), but levels were much lower than in rural areas of continental Europe (annual mean of 16.3–284 ng m-3) (Sánchez-Ochoa et al., 2007; Brighty et al., 2022), likely due to the sparse vegetation at Svalbard. The highest monthly means were seen for June followed by October, but there was no pronounced seasonality for cellulose as seen for the other PBAP tracers (Sect. 3.4.1). This corresponds with findings made by Sánchez-Ochoa et al. (2007), who pointed to a minor seasonality “with higher winter levels than expected”, and that of Puxbaum and Tenze-Kunit (2003), who associated increased cellulose levels in spring with “seed production and repulsing of other cellulose containing plant material” and “production of leaf litter” in the fall. High wind speed might be a driving force for generation and entrainment of cellulose containing aerosol particles that is more pronounced in winter, and particularly in the harsh Arctic climate, but possibly limited by snow cover. In the recent study by Brighty et al. (2022), a clear seasonality was shown, with increased levels in the summer and fall at French and Swiss rural sites.

Size distribution measurement of cellulose is limited and inconclusive, with the highest concentrations reported both for the fine (Puxbaum and Tenze-Kunit, 2003) and the coarse mode (Yttri et al., 2011b; Brighty et al., 2022). Lack of comparable seasonality between nearby sites indicates that local sources dominate (Brighty et al., 2022), but with a certain fraction associated with fine aerosol, LRT is a possibility. Cellulose did not correlate with other PBAP tracers or levoglucosan, corresponding to the findings by Brighty et al. (2022), but this does not exclude co-emission (see Sect. 3.4.1). A minor fraction (0.08 %) of RWC emissions was attributed to cellulose in a combustion study by Schmidl et al. (2008), but we found no strong connection between the samples highest in cellulose and levoglucosan, whereas we did for the other PBAP tracers and levoglucosan, or between cellulose and the other PBAP tracers (Table S8). The lack of resemblance between cellulose and other PBAP tracers and BB aerosol should be explored further.

Cellulose (here ng C m-3) made a 1.0±0.3 % contribution to OC annually, corresponding to the lower range reported for rural background sites along an east–west transect across Europe (0.7 %–3.9 %) (Sánchez-Ochoa et al., 2007) but substantially lower compared to French (3.2±2.4 %) and Swiss (5.9±4.4 %) rural background sites (Brighty et al., 2022).

The contribution of plant debris (here ng C m-3) was estimated from cellulose (Puxbaum and Tenze Kunit, 2003; Yttri et al., 2011a, b) as a 2.0±0.6 % contribution to OC annually and thus somewhat higher than for fungal spores (0.5 %–1.1 %). On a monthly basis, 4 % to 6 % contributions were observed in all seasons. Weekly samples (n=23) with a high (5 % to 12 %) plant debris contribution were associated with low OC levels (mean 53 ng C m-3; 22th percentile). Plant debris and OC were correlated (r2=0.707), suggesting that plant debris is a driver of observed OC levels at low concentrations. We did not observe a similar feature for fungal spores.

In the transition from July to August 2020, CA levels were high, with peak concentrations from 22–27 July (2172 ng C m-3 for OC and 59 ng C m-3 for EC). These levels were the highest in 4 years of observations, explaining 21 % of the annual OC loading but only 7 % of EC. However, levels were still lower than the record-high concentrations (3.5 µgCm-3 for OC and 0.24 µgCm-3 for EC) observed in the April to May transition 2006, caused by emissions from wild and agricultural fires in Eastern Europe (Stohl et al., 2007). All tracers (except cellulose) experienced maximum values during this episode, but 2-methyltetrols, glucose, fructose, and arabitol were the most elevated when compared to the long-term annual mean and to the enhancement seen for OC. The FLEXPART footprint clearly shows an influence from WF in the Khanty–Mansi district (western Russia) (Fig. 6), corroborating to the high levoglucosan concentration (6.0 ng m-3). Source apportionment by PMF attributed 55 % of eBC to BB, whereas FLEXPART calculated 62 %, with WF (95 %) as the totally dominating fraction. Flaring was the dominating fossil fuel source category according to FLEXPART, explaining 58 % of BC from fossil fuel sources.

The plume transport time from the source region to the Zeppelin Observatory was short, at fewer than 7 d for 67 % of eBC observed at Zeppelin 25 to 27 July (Fig. S3), whereas on average only 30 % of the observed eBC reaches the Arctic station after such short time. This might have contributed to the high level of 2-methyltetrols, which are indicated to have short atmospheric lifetimes (Yttri et al., 2021), in addition to the arguments raised in Sect. 3.3. Certain PBAP, such as fungal spores (Bauer et al., 2002; Yttri et al., 2007b), are small enough to be transported over long distances, even between continents (Prospero et al., 2005), and pyroconvection might bring larger-sized PBAP to altitudes that enable LRT (e.g., Groot Zwaaftink et al., 2022). PBAP contributed 14 % of OC, using the OC to PBAPTracers ER by Groot Zwaaftink et al. (2022).

The high CA level coincided with a prolonged period (24 to 29 July) of high temperatures, which were unprecedented since temperature measurements were initiated at Zeppelin (1998), caused by intrusions of warm air masses from Siberia; i.e., T> 10 °C for 111 consecutive hours, mean T=14.5±1.5 °C, and TMax= 18.2 °C. A disproportionally strong warming of the Arctic compared to the midlatitudes could create an important pathway of pollution to the Arctic (Stohl et al., 2007), and as for the LRT episode in spring 2006 (Stohl et al., 2007), emissions from WF at lower latitudes were essential in the deterioration of Arctic air quality also in July 2020.

Air masses with a history over southwestern Russia and eastern, central, and northern Europe (including Scandinavia) (Fig. 7) increased the OC (549 ng C m-3) and EC (52 ng C m-3) concentrations at Zeppelin to levels corresponding to 14 % of their annual loading. Source apportionment by the LHS suggested that BSOA (57 %) and BB (32 %) dominated CA (Table S4). Certain PBAP tracers (arabitol, mannitol, and trehalose) were enhanced beyond that of OC, reflecting the seasonal peak in fungal spores, but PBAP contributed only 3 % to CA (Table S4). An upper estimate of 13 % was obtained using the OC to PBAPTracers ER by Groot Zwaaftink et al. (2022). PMF apportioned 47 % of eBC to BB, comparing well with FLEXPART (51 %) (ascribing 62 % of BB to WF) and LHS (58 %). FLEXPART apportioned the majority of BC from FF combustion to traffic (55 %). The mean ambient temperature during the episode was enhanced compared to the long-term mean, as seen for all three episodes described.

This episode (Fig. 8) was studied by Groot Zwaaftink et al. (2022), combining surface and remote sensing observations and transport model simulations to understand its origin and development, whereas we in the present study focused on its carbonaceous aerosol content. Briefly, the EC level (78 ng C m-3) was the highest in 4 years of observations, whereas the OC level (818 ng C m-3) was much lower than observations made during the July 2020 (Sect. 3.6.1) episode, explaining 15 % and 13 % of the annual EC and OC loading, respectively (Fig. 9). Levoglucosan (5.0 ng m-3) was the only organic tracer elevated beyond that of OC, supporting FLEXPART calculations pointing to WF emissions in Ukraine and southern Russia as being one of two major sources of air pollution for this episode. Source apportionment of eBC by PMF indicated an almost equal share of eBC from BB (52 %) and FF combustion (48 %), as does FLEXPART (BB = 57 % and FF = 43 %), with the majority of BB attributed to WF (72 %), while traffic was the major FF category (52 %). Mixing with mineral dust emissions from central Asia en route caused a mineral dust level of 1.9–2.6 µgm-3, likely explaining the presence of carbonate (20 ng C m-3; 100 ng CO32- m-3). Before entering the Arctic, the polluted air masses deteriorated the air quality in a large part of northern Europe, giving PM10 levels around 100 µgm-3 and the same aerosol particle chemical signature as described for Zeppelin. These levels violate EU air quality guidelines, which have daily mean limit values for PM10 of 50 µgm-3.