Elucidating the present-day chemical composition, seasonality and source regions of climate-relevant aerosols across the Arctic land surface

The Arctic is warming two to three times faster than the global average, and the role of aerosols is not well constrained. Aerosol number concentrations can be very low in remote environments, rendering local cloud radiative properties highly sensitive to available aerosol. The composition and sources of the climate-relevant aerosols, affecting Arctic cloud formation and altering their microphysics, remain largely elusive due to a lack of harmonized concurrent multi-component, multi-site, and multi-season observations. Here, we present a dataset on the overall chemical composition and seasonal variability of the Arctic total particulate matter (with a size cut at 10 μm, PM10, or without any size cut) at eight observatories representing all Arctic sectors. Our holistic observational approach includes the Russian Arctic, a significant emission source area with less dedicated aerosol monitoring, and extends beyond the more traditionally studied summer period and black carbon/sulfate or fine-mode pollutants. The major airborne Arctic PM components in terms of dry mass are sea salt, secondary (non-sea-salt, nss) sulfate, and organic aerosol (OA), with minor contributions from elemental carbon (EC) and ammonium. We observe substantial spatiotemporal variability in component ratios, such as EC/OA, ammonium/nss-sulfate and OA/nss-sulfate, and fractional contributions to PM. When combined with component-specific back-trajectory analysis to identify marine or terrestrial origins, as well as the companion study by Moschos et al 2022 Nat. Geosci. focusing on OA, the composition analysis provides policy-guiding observational insights into sector-based differences in natural and anthropogenic Arctic aerosol sources. In this regard, we first reveal major source regions of inner-Arctic sea salt, biogenic sulfate, and natural organics, and highlight an underappreciated wintertime source of primary carbonaceous aerosols (EC and OA) in West Siberia, potentially associated with the oil and gas sector. The presented dataset can assist in reducing uncertainties in modelling pan-Arctic aerosol-climate interactions, as the major contributors to yearly aerosol mass can be constrained. These models can then be used to predict the future evolution of individual inner-Arctic atmospheric PM components in light of current and emerging pollution mitigation measures and improved region-specific emission inventories.


Introduction
The rapidly changing Arctic environment is characterised by a substantial sea ice loss and land surface temperature increase at a rate of 0.5 • C per decade since the late 1970s (IPCC 2013(IPCC , 2021. This world's highest rate of warming is called Arctic amplification (Serreze andBarry 2011, Block et al 2019), and is driven primarily by the greenhouse gas (e.g. CO 2 ) forcing, the Planck feedback, and the snow and ice albedo effect (Pithan andMauritsen 2014, Hegerl et al 2019), as well as low-level clouds (e.g. Curry and Ebert 1992, Shupe and Intrieri 2004, Lubin and Vogelmann 2006, Tan and Storelvmo 2019. The warming is accompanied by changing inner-Arctic natural and anthropogenic particulate matter (PM, or aerosol) and precursor emissions due to environmental and socio-economic change (Kirkevåg et al 2013, Schmale et al 2018.
Short-lived particulate climate forcers are estimated to exert an annual-mean net-cooling direct radiative effect in the Arctic (Sand et al 2015). The observed long-term trends in the single scattering albedo at different Arctic stations indicate significant spatial variability (Collaud Coen et al 2020), which remains to be elucidated in terms of PM chemical composition to resolve the sources. The recent combined decline in (scattering) sulfate and (absorbing) black/elemental carbon (BC/EC) aerosols (Schmale et al 2021a), due to better emission regulations across the Arctic Council nations, might explain a non-negligible fraction of the Arctic surface warming (Acosta Navarro et al 2016, Dobricic et al 2019, Ren et al 2020. Still, there is a particular need to understand which sources reduce both absorbing aerosols (e.g. BC/EC) and long-lived greenhouse gases (Shindell and Faluvegi 2009, AMAP 2015, Najafi et al 2015, Lohmann et al 2020. Further, a significant aspect of the Arctic aerosol impacts on climate is their influence on the Arctic sea ice through interference with the Arctic clouds and radiation . Hence, the question of whether potential short-term mitigation actions can result in a climate benefit, i.e. net Arctic cooling (Sand et al 2015), can only be answered if also the highly variable Arctic aerosol-cloud effects (Liu and Key 2014) are better understood in terms of sign and magnitude. That is because the indirect effect is likely more important than the direct effect in the Arctic (Menon et al 2008, Struthers et al 2011, where the limited amount of available aerosol can constrain cloud formation (Mauritsen et al 2011). The post-Soviet industrial collapse is associated with a decrease in inner-Arctic anthropogenic emissions over the past decades , Laing et al 2013, Kyrö et al 2014, Sharma et al 2019. However, aerosol concentrations remain less monitored in the vast Siberian Arctic (Popovicheva et al 2017, 2019, Manousakas et al 2020, a region listed among the four global observational hot spots with limited spatial coverage (Kulmala 2018). Models simulate aerosolclimate effects with low confidence at high northern latitudes (i.e. large inter-model spread), owing to a lack of extensive, spatially-resolved observational constraints (Mann et al 2014, Arnold et al 2016, Sand et al 2017, Willis et al 2018, Petäjä et al 2020, Schmale et al 2021b. That can lead to biased regional or global warming estimates (IPCC 2013, Cohen et al 2014, Yang et al 2014, Cohen et al 2018. Knowing the pan-Arctic yearly aerosol loading and chemical composition is crucial, as their future changes will impact the radiative balance of the Arctic atmosphere, whereby the net cooling or warming effect at different sites and seasons remains uncertain (Wegmann et al 2018, Schmale et al 2021b. The aerosol-climate effects, i.e. direct absorption or scattering of solar radiation, the cloud condensation nuclei (CCN) and ice nucleation activity, and therewith impact on the indirect aerosol effect, are a function of the aerosol chemical composition, particle size, and mixing state (Nguyen et al 2017, Adachi et al 2021, because different components have different physicochemical properties. Those properties lead to different optical characteristics (absorbing and scattering potential) and different hygroscopicity or ice-nucleating abilities relevant to aerosol-cloudclimate interactions (Liu and Wang 2010, Lange et al 2019, Schneider et al 2021. In addition, compound volatility and acidity (e.g. Wang et al 2020) are critical for multiphase chemical reactions that ultimately determine many physicochemical properties , Tong et al 2021. While enhanced observational and modelling efforts for Arctic atmospheric BC and its deposition have been conducted (Qian et al 2014, Cho et al 2019, Winiger et al 2019, the Arctic organic aerosol (OA) mass and its importance relative to EC or inorganic aerosols are still poorly characterised (Uttal et al 2016). Major ions such as sulfate, nitrate, and ammonium, are routinely monitored at several Arctic stations. However, beyond those, not all the different key aerosol components and concentration trends are observed together and in sufficient detail to reduce uncertainties in aerosol-climate interaction simulations (Quinn et al 2007, Hirdman et al 2010. So far, many studies were only able to develop a limited scope, i.e. focus on a small number of sites, a limited number of seasons, or only specific components (Willis et al 2018).
Here we present an unprecedented dataset of speciated PM in the recent past from eight sites across all sectors of the Arctic. We have achieved that with a combination of major inorganic ion and EC measurements, as well as offline aerosol mass spectrometry (AMS) analysis (Daellenbach et al 2016) for OA quantification and the apportionment of its natural and anthropogenic sources (Moschos et al 2022). We emphasise how the different Arctic aerosol components behave relative to each other, how they change over a seasonal cycle and between stations in relative terms, and discuss the assets of the presented pan-Arctic aerosol dataset for simulating future Arctic climate.

Filter sampling and measurements
Filter sampling and subsequent offline measurements have three main advantages: (a) extending the spatial and temporal coverage of the sampled aerosol, as filters are routinely collected at many stations; (b) collection of sufficient aerosol loading to increase the signal-to-noise ratio of subsequent analyses; (c) possibility to analyze the total aerosol including the coarse size fraction (>2.5 µm). We provide an overview of the sampling sites in table 1, including the station acronyms, coordinates and altitude, duration of the sampling, and polar night/midnight sun periods. We collected PM 10 (particulate matter with an equivalent aerodynamic diameter of less or equal to 10 µm) and total suspended particulate (TSP) matter, i.e. total aerosol without a defined size cut, on quartz fiber filters at the different sites following procedures described elsewhere (Moschos et al 2022). We then measured the samples with various offline techniques as outlined below. The typical filter-composite sample time resolution is weekly (more variable at BAR), except for ALT (bi-weekly) and TIK (∼3 d).
We quantified elemental and organic carbon (Sunset OC-EC analyzer-based) by thermal-optical analysis using the EUSAAR-2 protocol (Cavalli et al 2010), except for ALT (ECT9; Huang et al 2021) and UTQ (NIOSH 5040 protocol). We measured major water-soluble inorganic ionic components (SO 4 2− , NO 3 − , Cl − , NH 4 + , Na + , K + , Ca 2+ , Mg 2+ ) by ion chromatography (Jaffrezo et al 1998); ion data for BAR, TIK, and UTQ were presented in Popovicheva et al (2019), Manousakas et al (2020), and Moffett et al (2020), respectively. We measured major ions in samples from ALT during 2018 at two different laboratories, and the measurements and seasonal trends were reproducible (major ions sum: slope = 1.15; R 2 = 0.98; intercept = 0). For all samples, we calculated the ionic balance (in nEq m −3 ; nEq: nano-equivalents) considering all major inorganic ionic components. We calculated the non-seasalt ( where the value of 0.252 represents the mass ratio of sulfate-to-sodium in seawater, and 1.47 accounts for the presence of anions and cations other than Na + and Cl − in sea salt (Moffett et al 2020). Negative nss-sulfate concentrations were calculated for 2% of the samples but were not significantly different from zero (within measurement errors); we have not considered these samples for calculating the NH 4 + /nsssulfate and OA/nss-sulfate ratios (outliers). Further, the Cl − /Na + mass ratio was greater than that of seawater (1.8) in only 2 % of the samples, but this value never exceeded 2.1. Our calculations do not include (Na + -containing) mineral dust, which may become important in long-range episodic air mass transport events or locally (Groot Zwaaftink et al 2016). That is because mineral dust is a complex (and variable) mixture of oxides and carbonates and consists of elements that our techniques could not quantify, and none of the measured water-soluble ions can be considered unique tracers for dust. While Ca 2+ has been used as a tracer to identify air masses influenced by dust, we could not retrieve accurate estimates of dust contribution to PM due to the variability in the Ca 2+ /total_dust_PM. Our yearly-average Ca 2+ concentration of 43 ng m −3 at ALT is comparable to earlier results for the same station (Sharma et al 2019).
As detailed in Moschos et al (2022), we utilised a high-resolution long time-of-flight Aerodyne aerosol mass spectrometer (L-ToF-AMS) with electron impact ionisation, for the bulk chemical characterisation of the pan-Arctic OA fraction upon water extraction. We used the AMS OA mass spectra as inputs in positive matrix factorisation (PMF) analysis and the total OA mass was apportioned to different primary vs. secondary and natural vs. anthropogenic source components (Moschos et al 2022). Here we focus on the AMS-PMF-based total OA mass Table 1. Arctic station acronyms, coordinates, altitude (meters above sea level, m asl), polar night and midnight sun periods, sampling period, size cut, and the number of filter sample composites. (OC and heteroatoms mass, WSOC-based), the total anthropogenic OA (Anthr-OA), and methanesulfonic acid-related OA (MSA-OA). We hereafter refer to PM as the sum of the sea salt, ammonium, nitrate, nss-sulfate, EC, and total OA mass concentrations.

Back-trajectory analysis to identify source regions of aerosol components
Back-trajectories (BTs) show the air mass history (origin and atmospheric transport paths) and thus can provide information on the geographic location of potentially advected emissions at large geographical scales. We performed BT analysis to assess potential source locations of individual Arctic aerosol components over the entire sampling period at each station. We calculated the trajectories backward for 10 d (every 6 h) using the HYSPLIT4 model with meteorological data from the Global Data Assimilation System with one-degree resolution. We weighted the calculated BTs with the time series of each component using the concentration-weighted trajectory (CWT) approach to localise air parcels responsible for high concentrations at the receptor site (Rai et al 2020). We used the Igor-based user interface ZeFir (Petit et al 2017) to calculate component-specific CWT maps separately for each specific station. The CWT value of a particular grid cell (i, j) (latitude, longitude) is a measure of the source strength of a grid cell to a receptor site and is determined as follows: responding to the arrival of BT l, τ ijl is the number of trajectory segment endpoints in grid cell (i, j) for back trajectory l divided by the total number of trajectory segment endpoints for back trajectory l (i.e. residence time of a trajectory in each grid cell), and L is the total number of back trajectories over the entire period for one station. Since the temporal resolution is low, enlarging the size of the input dataset, which is a novel tool available in ZeFir, allowed to take more BTs into account and improve their statistical representativeness (Petit et al 2017). For instance, when the filter composite time resolution was weekly on average, all trajectories arriving at the station (with a frequency of four times a day) for 7 d back in time since the sampling end (i.e. 7 d temporal extension of the input data) were considered for each sample. We merged results from the CWT-based BT analysis for each Arctic station (without prior normalisation) to highlight component-specific pan-Arctic hot spot source regions, i.e. associated with the highest mass concentrations across all stations, with greater accuracy than would be possible with single-site only results (Han et al 2007, Petit et al 2017, Moschos et al 2022. We combined N-site trajectory analyses (multi-site CWT, MS-CWT) based on the following notation: MS − . TIK was not considered for the BT analysis here, as only a few samples were available with PM measurements, and hence a robust result was less likely to be obtained.

Results and discussion
The major aerosol components, in descending order of pan-Arctic relative contributions to PM 10 and TSP mass (figure 1), are sea salt, secondary (nss) sulfate, OA, ammonium, EC, and nitrate. Considering all measurements, the mean PM mass is 1.6 µg m −3 (1 st and 3 rd quartiles, Q 1 -Q 3 = 0.68-2.1 µg m −3 ). We note that the general composition (and climate effects) of PM 1 and PM 2.5 might differ. We present individual component yearly mass concentrations in figure S1 (available online at  stacks.iop.org/ERL/17/034032/mmedia) for the different stations. The seasonal cycles of PM (figure S2) reveal that mass concentrations in spring, and in certain cases also during the polar night (winter) period (e.g. at ALT and BAR; figure 1), are usually higher than in summer (midnight sun period). By contrast, biogenic emissions contribute significantly to the higher abundance of PM in summer vs. winter at PAL (Moschos et al 2022). Figures 2 and 3 show the relationships between different inorganic and carbonaceous aerosol components and their fractional contributions to PM, as well as their seasonality and spatial variability. We display the BT analysis results for major Arctic aerosol components in figure 4. The findings shown in figures 1-4 are discussed in the following for each chemical component individually. Sea salt is the major contributor to PM (figure 1) with a mean pan-Arctic contribution to PM (figure 2) of 39% (Q 1 -Q 3 = 19%-58%) and an annual mean concentration (figure S1) of 530 ng m −3 (Q 1 -Q 3 = 150-700 ng m −3 ). At PAL, TIK, and VRS, where the OA and/or nss-sulfate contributions are elevated, the fraction of sea salt is less dominant, i.e. contributing on average 24% of the PM (Q 1 -Q 3 = 8%-29%). Nss-sulfate contributes on average 30% (Q 1 -Q 3 = 17%-41%) to PM across the stations (figure 2). The corresponding annual mean concentration (figure S1) of 470 ng m −3 (Q 1 -Q 3 = 160-630 ng m −3 ) is similar to reported values at ALT and UTQ (Quinn et al 2002, although these values have been decreasing over the past Figure 2. Fractional contributions of Arctic aerosol components to PM (left), and relationships between aerosol components (right), with the temporal variability shown as box and whisker plots (line and box: median and IQR; squares and whiskers: mean and range within the 5 th and 95 th percentiles; diamonds: outliers) for the different sites (x-axes; acronyms in table 1). EC data from GRU were not available. Major ion data from UTQ (hatched boxes) were available only for the summer period, and only a few samples were available from TIK (hatched boxes).
decades (Sharma et al 2019, Ren et al 2020, Schmale et al 2021a. While the nss-sulfate concentrations are largely spatially homogeneous (figure S1) and are thus likely related to regional processes, we observe slightly higher mean values at the Eurasian sites GRU and PAL (∼600 ng m −3 ), and additional nss-sulfate sources might exist at TIK (1600 ng m −3 ). The majority of sulfate, typically 80%-90%, is of non-sea-salt origin (figure 2) and associated with the oxidation of SO 2 (i.e. sulfate is secondary), which is of natural (emissions from marine biota) or anthropogenic (fuel combustion) origin, whereas crustal contributions are small (Udisti et al 2016). The highest nss-sulfate concentrations (∼1.5 µg m −3 ; figure 4) are observed in the second half of the Arctic haze period (figure S2) and are associated with air masses arriving from the larger Eurasian region (figure 4), similar to secondary anthropogenic-dominated organics (Moschos et al 2022). Hence, the pan-Arctic nss-sulfate is likely predominantly anthropogenic, consistent with past observations at GRU in spring (Udisti et al 2016) and yearly and long-term sulfur isotope measurements at ALT Barrie 1993, Norman et al 1999). Based on the BT analysis results, marine source regions of biogenic nss-sulfate can be inferred ( figure 4).
The source regions of NH 4 + (figure 4), which exhibits relatively high (and comparable) concentrations at ALT, BAR, PAL, and highest at TIK (figure S1), are similar to those of nss-sulfate. That indicates the frequent co-emission and conversion of their precursors to ammonium (bi-)sulfate, at least during the peak haze (spring) period (figure S2). Ammonium measurements at such low concentration levels (figure S1; Q 1 -Q 3 = 15-100 ng m −3 ) can be prone to positive artefacts and hence are associated with substantial uncertainties (Xu et al 2020). Nevertheless, we derive a pan-Arctic mean ammonium/nss-sulfate ratio (figure 2) of 0.15 (median: 0.12; Q 1 -Q 3 = 0.06-0.21), indicating that ammonium is not sufficiently present to neutralise nss-sulfate (∼0.4 in terms of equivalent ratio). When taking all measured major inorganic ions into account, the pan-Arctic ratio of cations-to-anions is 0.9-1.0 (table S1), resulting in a neutral PM 10 and TSP aerosol. An exception is the station GRU, where the PM might be slightly acidic in many samples (average cations/anions <0.75; table S1) if experimental errors are not accounted for. The significant presence of Na + , originating mainly from sea salt particles, contributes substantially to the near-equal abundance of anions and cations. That indicates that the ammonium-to-nss-sulfate mass ratio might not always represent the PM acidity in the Arctic but likely only relates to the potential neutrality/acidity of fine-mode particles. The nitrate contribution is rather low (figures 1 and 2), as expected for ammonia-poor acidic or long-range transported (fine-mode) aerosols (Kerminen et al 2001, Bauer et al 2007, Kim et al 2014, Kakavas et al 2021. Total OA contributes on average 22% (Q 1 -Q 3 = 11%-28%) to PM (figure 2), which is often comparable to the relative contributions of sea salt and nss-sulfate (figure 1). The hotspot source regions for long-range transported total OA (figure 4), dominated by the highest concentrations at BAR and PAL (up to 1.5 µg m −3 ; figure S1), are collocated with urban/industrial centers in Europe and West Siberia, as well as biogenic emission-related regions, such as the Atlantic Ocean and the boreal forests in North-East Europe and central Siberia. That is because OA is a diverse mixture of multiple terrestrial or marine natural and anthropogenic source components in summer and winter (Moschos et al 2022). The mean pan-Arctic OA/nss-sulfate ratio (figure 2) is 0.81 (median: 0.63; Q 1 -Q 3 = 0.37-1.02). Increased values of 1.20 (median: 0.99; Q 1 -Q 3 = 0.75-1.63) are observed at BAR and TIK (figure 2) as well as in general in summer and early autumn (figure 3) when anthropogenic influence is at a minimum (Anthr-OA/OA in figure 3). Similar to our finding, a mean value of the modelled OA/sulfate mass ratio of 0.95 (median: 0.77; inter-model range: 0.25-2.0) has been reported in an evaluation and inter-comparison of . The color scales indicate concentrations corresponding to major distant source regions ('long-range' probability heat maps). For nss-sulfate, ammonium, EC, and water-soluble OA, 10-day back-trajectories were calculated with 3 km threshold altitude (PAL: 5-day BTs for nss-sulfate and OA), while for sea salt 5-day BTs with 1.5 km altitude were used. Since major ion data from UTQ winter, spring and autumn were not available and sea salt exhibits the highest concentrations at UTQ across all measurements (figure S1), the major summertime source regions of this PM component at UTQ were processed and presented separately (figure S3).
OA in global (AeroCom phase II) models for the year 2006 (Tsigaridis et al 2014). Our OA/nss-sulfate ratio at the different sites and months can be used to constrain the OA budget and seasonality at different Arctic sectors, because sulfate is routinely measured at most observatories included in this study and thus the nss-sulfate budget could be approximated.
The MSA-OA/nss-sulfate ratio (figure 2) peaks in summer (figure 3), reaching up to 0.6 at BAR, GRU, and VRS. Specifically, 6% and 13% of the pan-Arctic values in April-September are >0.4 and >0.2, respectively. Despite being strongly dependent on temperature and the occurrence of blooms (Yu et al 2021), we expect the variability in this ratio to be largely affected by different degrees of anthropogenic perturbation (Kerminen et al 1997). Hence, this ratio can be used -indirectly-to distinguish between anthropogenic and natural (marine biogenic) sulfate at different Arctic sectors. The maxima across the three stations (BAR, GRU, and VRS) are comparable to previously reported values at ALT and UTQ . At the same time, the relatively lower maximum summertime value of 0.068 at PAL in May (figure 2), which indicates some Arctic spatial variability in this ratio, is similar to the value of 0.090 reported for July at Kevo (Laing et al 2013), which is located also in the Finnish Arctic. This lower ratio for PAL, indicating a smaller influence of marine MSA in comparison to other stations, can be attributed to its distance to the ocean (>200 km), as well as the predominance of more local aerosol sources (Hellén et al 2020) and the advection of air masses of terrestrial origin from lower latitudes (as indicated from the BT analysis) containing nss-sulfate.
The EC relative contribution is on average only 2% of the PM (Q 1 -Q 3 = 1%-3%) across the Arctic  figure S1), are collocated with West Siberian regions associated with intense gas flaring activity in winter, similar to those found for primary anthropogenic organics (Moschos et al 2022). Our multi-site observation agrees with previous single-site studies using trajectory statistics or other transport model calculations. These have attributed winter/spring-high BC levels at ALT, BAR, TIK, UTQ, or ZEP to high-latitude Eurasian (rather than North American or South Asian) source regions (Polissar et

Conclusion
Knowing the overall aerosol chemical composition in the Arctic is not only the gateway to differentiating natural from anthropogenic contributions, but also understanding the impact the fast-changing Arctic environment has on the atmospheric chemical composition, and resulting climate feedbacks on the composition, which in turn affect aerosol-radiation and aerosol-cloud interactions. The presented Arctic dataset contains unprecedented comprehensive chemical information from eight Arctic stations for 2014-2019, including individual and comparable anthropogenic and natural component contributions to PM 10 and TSP dry mass. Our pan-Arctic yearly analyses, including the relative contribution of Arctic organics to PM, their importance compared to secondary inorganics and total carbonaceous aerosols, as well as the ionic balance and seasonal cycle of anions and cations ( figure S4; table S1), provide added value compared to earlier studies (e.g. Barrie 1995, Willis et al 2018. We show that the major primary (sea salt) and secondary (nss-sulfate) inorganic aerosol components typically dominate the present-day Arctic PM 10 and TSP mass, especially in winter and at the most remote stations. We demonstrate that the higher PM in spring vs summer (figure S2) is attributed to the seasonal cycle of inorganic components, whereas organics exhibit less of a seasonal cycle in absolute terms (figure S2), as discussed in Moschos et al (2022). The increased abundance of sea salt in winter is attributed mainly to blowing snow, whereas the increased abundance of nss-sulfate in spring is attributed to the Arctic haze phenomenon. The spatially and seasonally variable component ratios provide a broader assessment compared to singlecomponent or single-site/season reporting, and hence useful insights for policymakers regarding the yearly relative importance of major aerosol components and sources across the Arctic land surface. We note that a biomass-burning signature might not necessarily be observed at the surface sites investigated here but further aloft (Stohl 2006).
We further examined possible links to regional and distant sources: due to long-range transport of air masses, aerosol particles from the continental Arctic (Eurasia) spread over thousands of kilometers and can influence the radiative balance and the Arctic climate. New insights become available for Siberia from where most of the inner-Arctic air pollution originates. That is illustrated by the high levels of anthropogenic aerosol components (nss-sulfate, EC, Anthr-OA) measured at BAR and TIK combined with the component-specific BT analysis. Our observational effort, extending beyond the traditionally considered BC and sulfate, points to areas that require new measurements (e.g. design of targeted fieldwork) and can support informed policy decisions towards targeted emission reductions (e.g. gas flaring in West Siberia). In addition, models have been struggling to represent the climate change-sensitive natural Arctic aerosol components, such as sea salt and biogenic sulfate/organics. Our holistic observational approach reveals their present-day major source regions for the first time from a pan-Arctic perspective, providing vital information for models to ingest to simulate future Arctic change.
The reported observational results are of general interest to researchers not only focused on Arctic aerosol processes but also on improving model representations of climate-relevant aerosols at the poles and globally. The findings on speciated natural and anthropogenic aerosol components, including their spatial and seasonal variability and source regions, will be crucial in guiding Arctic climate model evaluation, even though the model spatial resolution remains coarser than that of ground-based point observations. Specifically, the presented complete dataset can assist in reducing uncertainties in Arctic aerosol-climate interaction simulations by constraining the aerosol acidity and hygroscopicity (including the bulk kappa value), carbonaceous aerosol radiative effects, the mass concentration of natural vs. anthropogenic components, and the OA contribution as a function of the better-known nss-sulfate loading. Correlations between components might indicate their mixing state, allowing to constrain radiative effect calculations for Arctic aerosol particles consisting of various components with distinct properties, e.g. OA-coated sea salt in sea spray aerosols (Kirpes et al 2019) or internally mixed EC/OA/nss-sulfate in haze particles. While information on the CCN number is not available, model vs. measurement comparisons of the mass concentration of different species and their seasonality is very helpful to validate modelling efforts extending to the entire Arctic and for different seasons. Our dataset is a crucial, timely addition to the upcoming Arctic Monitoring & Assessment Programme (AMAP) report on 'Impacts-of-SLCFs-on-Arctic-Climate-Air-Quality-and-Human-Health' , and can further assist in interpreting optical and other polar measurements, e.g. from the year-long MOSAiC expedition in the central Arctic Ocean (Shupe et al 2020), in light of the complex aerosol chemistry. It also provides the basis to predict the future evolution of inner-Arctic atmospheric PM, considering rapid environmental changes and emission control measures, by utilizing component-specific decadal trend analyses and applying compound ratios. At the same time, the long-term pan-Arctic aerosol historical trends (Schmale et al 2021a) and vertical distribution, including in-cloud processes (Creamean et al 2021), remain to be elucidated.
Harmonized, consistent, and comparable yearly measurements, similar to those presented here, should be expanded in the future, e.g. to non-coastal stations closer to the Arctic Circle. While a further decline in anthropogenic-dominated emissions will 'support the implementation of effective measures to reduce local air pollution in Arctic communities' , as recommended by AMAP, concurrent aerosol radiative forcing estimates remain uncertain and should be defined more clearly at different Arctic sectors and seasons. Such estimates can only be accurate if atmospheric organics and sea salt are considered, and not only BC and sulfate as is typically the case.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https://doi.org/10.5281/zenodo.5179984  Bishop, and David Oaks, and the Ukpeaġvik Iñupiat Corporation, specifically Walter Brower and Jimmy Ivanoff for sample collection and field assistance. The authors would also like to thank the Baylor University Center for Reservoir and Aquatic Systems Research for access to the instrumentation for ion chromatography analysis. Financial support is also acknowledged by the Danish Environmental Protection Agency and the Danish Energy Agency with means from MIKA/DANCEA funds for environmental support to the Arctic region (Project Nos. Danish EPA: MST-113-00-140; Ministry of Climate, Energy, and Utilities: 2018-3767), iGOSP project, and the Graduate School of Science and Technology, Aarhus University; Villum Foundation is gratefully acknowledged for financing the establishment of VRS; Thanks to the Royal Danish Air Force and the Arctic Command for providing logistic support to the project; Christel Christoffersen, Bjarne Jensen, and Keld Mortensen are gratefully acknowledged for their technical support. The observations at ZEP are funded by the Norwegian Environment Agency and are part of the national monitoring program. J L J acknowledges LABEX OSUG@2020 (ANR-10-LABX-56) for funding analytical instruments.

Author contributions
IeH, J S, U B, and V M conceived the study. V M analyzed the data. IeH, J S, A S H P, and U B supported and supervised the research. V M, IeH, J S, and U B interpreted the results. V M, J S, and IeH prepared the figures and wrote the manuscript. L H is the scientific authority for filter sampling, the measurements of EC/OC at ALT (2015ALT ( -2018, and the data processing/QAQC. W Z performed EC/OC analysis and logistic support for filter sampling at ALT. S S provided the ion chromatography data from ALT 2015-2018. O P coordinated filter sampling and provided ion data from BAR and TIK. S B, G C, M S, R T took care of aerosol samplings at GRU and RT is responsible for the RIS (Research in Svalbard) 3693 project involving multiple aerosol studies at GRU. M V set up and took care of the filter sampling system at PAL. R J S and C E M organized and managed the collection at UTQ. C E M performed ion chromatography for UTQ. A M coordinated the filter sampling at VRS. JKN has analyzed and quality assured the OC/EC data from VRS. H S is the scientific head of VRS and the PI on the research projects that funded the collection of the filters. K E Y and W A A are responsible for the collection of aerosol filter samples at ZEP. J L J contributed with measurements of chemical species on filters using ion chromatography, HPLC-PAD, and HPLC-MS. All co-authors commented on the manuscript.

Conflict of interest
The authors declare that they have no conflict of interest.