Impact of Biomass Burning on Arctic Aerosol Composition

Emissions from biomass burning (BB) occurring at midlatitudes can reach the Arctic, where they influence the remote aerosol population. By using measurements of levoglucosan and black carbon, we identify seven BB events reaching Svalbard in 2020. We find that most of the BB events are significantly different to the rest of the year (nonevents) for most of the chemical and physical properties. Aerosol mass and number concentrations are enhanced by up to 1 order of magnitude during the BB events. During BB events, the submicrometer aerosol bulk composition changes from an organic- and sulfate-dominated regime to a clearly organic-dominated regime. This results in a significantly lower hygroscopicity parameter κ for BB aerosol (0.4 ± 0.2) compared to nonevents (0.5 ± 0.2), calculated from the nonrefractory aerosol composition. The organic fraction in the BB aerosol showed no significant difference for the O:C ratios (0.9 ± 0.3) compared to the year (0.9 ± 0.6). Accumulation mode particles were present during all BB events, while in the summer an additional Aitken mode was observed, indicating a mixture of the advected air mass with locally produced particles. BB tracers (vanillic, homovanillic, and hydroxybenzoic acid, nitrophenol, methylnitrophenol, and nitrocatechol) were significantly higher when air mass back trajectories passed over active fire regions in Eastern Europe, indicating agricultural and wildfires as sources. Our results suggest that the impact of BB on the Arctic aerosol depends on the season in which they occur, and agricultural and wildfires from Eastern Europe have the potential to disturb the background conditions the most.


S1 FIGAERO-CIMS Operation and Mass Concentration Conversion
The iodide ions of the FIGAERO-CIMS for the ionization were produced by passing 2 LPM of dry nitrogen over a perm tube of methyl iodide and afterwards through a Polonium source.The pressures within the CIMS were as follows: 100 mbar in the ion molecule reaction chamber (IMR), 2 mbar in the first quadrupole (SSQ), 1e-2 mbar in the second quadrupole (BSQ), 1e-7 in the time of flight (TOF) region.The voltage difference between the end of the SSQ and the entrance of the BSQ was approx.3 V.
To determine the sensitivity for our instrument, calibrations with levoglucosan were conducted in the laboratory using the same voltage settings as during the campaign.For the calibration a solution of 0.0001 g of levoglucosan in 100 ml of acetone was prepared and three different volumes (2, 4,   and 6 µL and by that three different amounts of levoglucosan, i.e., 2 ng, 4 ng, and 6 ng) of this solution were deposited on the FIGAERO filter using a syringe (following the syringe method described in Ylisirniö et al. 1 ).Thereafter, the same heating procedure as during the field measurements was used, where a flow of 2.3 LPM of nitrogen was passed over the filter in different phases: In a ramping phase the temperature on the filter was gradually increased from ambient to 200°C within 20 min, followed by a soaking phase where the temperature at the filter was kept at 200°C for 20 min, before the temperature was cooled down to room temperature again within another 20 min.The signal was normalized to one million reagent ions, integrated over the ramping and soaking period to yield the units of counts, and background subtracted.The result of that calibration is shown in Figure S1 and yields a sensitivity of 23 ± 8 counts s -1 ppt -1 .The conversion of the deposited mass on the FIGAERO filter to a mixing ratio in ppt (Vlevo/Vtotal, where Vlevo is the volume of levoglucosan deposited on the filter, and Vtotal is the volume of air that passed over the filter during desorption) was done by using the ideal gas law (at T = 273.15K and p = 101325 Pa) with a volume Vtotal of 92 L (as a result of 20 min ramping + 20 min soaking time at 2.3 LPM).
The flows, instrumental settings and electric field strength of our instrument were largely similar to the settings in Lee et al. 2 and Lopez-Hilfiker et al. 3 , who reported sensitivities in the range of 19 to 26 counts s -1 ppt -1 .As a result of that, assuming similar uncertainties of the total mass concentrations as these two previous studies is a good approximation.The mass concentrations reported from the FIGAERO-CIMS were converted using a maximum sensitivity (cal) of 22 counts s -1 ppt -1 2 .The measured signal from the FIGAERO-CIMS in ion counts was converted to atmospheric mass concentration (conc) for each compound i in µg m -3 according to the following calculation: Where F is the FIGAERO-CIMS signal in ion counts, mmol is the molar concentration in mol/L (calculated via the ideal gas law p/(R*T) with ambient pressure p and temperature T and the ideal S4 gas constant R), Lin is the total flow in L going into the IMR chamber of the CIMS during the desorption, MW is the molecular weight in g/mol, htime is the duration of the desorption phase in seconds, partflow is the particle sampling flow in LPM during the particle collection time, and colltime is the duration of the particle collection in minutes.For E3 EBAS levoglucosan is not available.The mass concentrations for the event times are labelled (E1-E7) and highlighted in green stars, and the concentrations during the rest of the year is shown as black dots.The lines show the linear fits for the non-event data (solid black), the event data (solid green), the event data without E4 (dashed green), and the combination of the events and the non-events (dashed black).(c) Scatter plot of ACSM organic vs FIGAERO organic mass concentrations.

S2 Kappa Calculation
The hygroscopicity parameter κ was calculated based on the ACSM data, following the Zdanovskii-Stokes-Robinson mixing rule 4,5 .By this rule, κ is composed of an inorganic and organic contribution, where the volume fractions (quotient of mass fraction to density) of the inorganic compounds (εinorg) are multiplied with the hygroscopicity of the inorganic compounds (κinorg), and similarly the organic contribution is the product of its volume fraction (εorg) and the hygroscopicity (κorg): κ = εinorgκinorg + εorgκorg.The inorganic contribution is composed of the neutral inorganic compounds ammonium sulfate (density: 1.769 g cm -3 6 , kappa: 0.61 7 ), ammonium nitrate (density: 1.720 g cm -3 6 , kappa: 0.67 7 ), ammonium bisulfate (density: 1.780 g cm -3 6 , kappa: 0.91), and sulfuric acid (density: 1.830 g cm -3 6 , kappa: 0.9 8 ).Their mass fractions (and by that the volume fractions) were calculated based on the inorganic compounds measured by the ACSM mass concentrations of sulfate, nitrate, and ammonium, by using the ion pairing scheme by Gysel et al. 6 .For the organic contribution, the measured organic mass concentration from the ACSM was used, with a density of 1.3 g cm -3 9 , and κorg = 0.07 8 .
Table S1.Ratio of the mass concentrations (in μg m -3 ) of particles smaller 180 nm (PM0.18) to those smaller 708 nm (PM0.708) at two different densities (ρ1 = 0.7 g cm -3 and ρ2 = 3 g cm -3 , range of densities used by FIDAS) for the individual events (E1-E7) and during the rest of the month that was not an event (NE).The mass concentrations were calculated from the DMPS number size distributions in the size range 5-708 nm.The numbers in brackets show the relative contribution of PM0.18 to PM0.708.   , who calculated a detection limit of 0.012 Mm -1 for the absorption coefficient (at a time resolution of 1-2 h) for the same instrument type we used, which S11 corresponds to 1.1 ng m -3 (using a MAC of 10.6 m 2 g -1 at 1-2 h resolution), or 0.8 ng m -3 at our time resolution of 2.

S8 Severity of the Fire Year
To assess the severity of the fires in the investigated BB events in 2020, the fire radiative power (FRP) of the events were compared to the year 2020 and to the long-term FRP from 2001-2020.
The fire regions were grouped in four areas (North America: 125°W to 71°W, 42°N to57°N; South Siberia: 55°E to 130°E, 40°N to 55°N; Black Sea: 25°E to 48°E, 40°N to 53°N; North Siberia: 65°E to 150°E, 60°N to 70°N), in accordance with the potential fire source areas attributed to the BB events.The definition of the areas and the comparison of the FRP of the year 2020 and the events in 2020 to the median of 2001-2020 is shown in Figure S7.The year 2020 shows more intensive fires in all areas during the entire year, which is in agreement with McCartney et al 14 .In addition, our events also show higher FRP throughout the year, indicating more intensive fires when compared to the multiyear fire activity.

Figure S1 .
Figure S1.Calibration of the FIGAERO-CIMS using three different amounts (diamonds) of

Figure S3 .
Figure S3.Schematic of the setup at the Zeppelin Observatory.The DMPS, MAAP, and Figure S4.Definition of the BB events, exampled of E3.(a) Time series of levoglucosan and

Figure S5 .Figure S6 .
Figure S5.Hygroscopicity parameter κ for the BB events and the rest of the year (non-events),

Figure S7 .
Figure S7.Back trajectories and fire activity for the individual BB events.The figure is similar to

Figure S8 .
Figure S8.Mean precipitation along the 27 ensemble trajectories for all the individual BB events.

Figure S9 .
Figure S9.(a) Definition of the fire regions (colored boxes) and location of the Zeppelin

Table S2 .
11mits of detection (LOD) at 2.5 h time resolution, fraction of data points (DP) below LOD, and yearly means and standard deviations of the species considered.The measured LODs were converted to the corresponding LOD at 2.5 h time resolution by taking the conversion stated in Fröhlich et al.11:   = LOD measured √(  /  ), where tmeasured is the time *LOD as determined by Asmi et al.

Table S3 .
p-values corresponding to Figure2.Chemical composition event vs. non-events.Values highlighted in shaded red are significant (p < 0.05).

Table S4 .
p-values corresponding to Figure3.Molecular-level chemical composition event vs.
non-events.Values highlighted in shaded red are significant (p < 0.05).

Table S7 .
p-value corresponding to Figure7.Kappa values events vs. non-events.The value highlighted in shaded red is significant (p < 0.05).