A total of 92 total suspended particulate (TSP) samples were collected during three Chinese Antarctic research expeditions and one Arctic expedition aboard the icebreaker R/V Xuelong. Sampling spanned a latitudinal range of approximately 160° (86° N to 75° S), encompassing polar and mid-latitude marine regions. The Antarctic samplings were conducted in October to November in 2019 (SP2019, 14 samples), November 2021 to March 2022 (SP2021, 23 samples), and October 2023 to April 2024 (SP2023, 15 samples), while the Arctic campaign occurred in July to September in 2021 (40 samples).

During the Antarctic campaigns, aerosols were collected using a high-volume air sampler (HVAS, TISCH Environmental, USA; flow rate: 1.2 m3min-1) equipped with pre-baked (500 °C, 24 h) Whatman quartz filters (20.3 cm × 25.4 cm; Whatman Ltd., UK). For Arctic sampling, a DIGITEL DHA-80 sampler (flow rate: 500 Lmin-1) with 14.2 cm diameter Whatman quartz filters was employed. Each sample represented a 48 h integrated collection period, corresponding to 2–4° latitude traversed during ship transits. To minimize contamination from ship emissions, a wind sector controller restricted sampling to air masses within 120° of the ship's heading. Filters were handled using nitrile gloves and masks to avoid potential contamination. Post-sampling, filters were folded with the collection surface inward, wrapped in pre-cleaned aluminum foil, sealed in polyethylene bags, labeled with sampling time and location, and stored at -20 °C. Detailed protocols followed established methodologies (Shi et al., 2021). Following expeditions, samples were transported to the laboratory under frozen conditions and maintained at -20 °C until analysis. The sampling locations for the Antarctic and Arctic campaigns are illustrated in Fig. 1.

Major ions were quantified through ion chromatographic analysis of water extracts of the aerosol samples. The extraction of filters in the laboratory followed protocols comparable to those described in the previous study (Shi et al., 2021). Prior to measurement, three-quarters of each filter was sectioned into small pieces using acid-cleaned Teflon-coated scissors and transferred into high-purity Milli-Q water (18.2 MΩ). The suspensions were subjected to ultrasonic treatment for 30 min, followed by continuous shaking at 120 rpm for 12 h to ensure thorough extraction of water-soluble components. The extracts were subsequently filtered through 0.22 µm polytetrafluoroethylene (PTFE) membranes prior to ion analysis. The concentrations of the main ions (NO3-, SO42-, Na+, NH4+, K+, and Ca2+) in the sample were determined by an ion chromatograph (AQ1100, RFIC, equipped with a CS12 column (2 mm × 250 mm) for cation analysis, AS11 column (2 mm × 250 mm) for anion analysis, Thermo Scientific, USA), and the eluents of cation and anion were 18.00 mM methylsulfonic acid (MSA) and potassium hydroxide (KOH), respectively. During sample analysis, the relative deviation of repeated assays (n = 5) of all ions is usually less than 5 %. We used the following formula to calculate non-sea salt SO42- (nssSO42-), non-sea salt Ca2+ (nssCa2+) and non-sea salt K+ (nssK+): 1[nssSO42-]=[total SO42-]-0.252×[Na+]2[nssCa2+]=[total Ca2+]-0.038×[Na+]3[nssK+]=[total K+]-0.037×[Na+] where 0.252, 0.038, and 0.037 in the above expressions are the ratios of SO42-/Na+ (Quinby-Hunt and Turehian, 1983), Ca2+/Na+ (Anonymous, 1997), and K+/Na+ (Keene et al., 1986) in the sea water, respectively.

OC and EC concentrations were determined using a Thermal/Optical Carbon Analyzer (DRI, Model 2001, Atmoslytic Inc., USA) following the IMPROVE protocol as implemented by Wu et al. (2024). OC and EC measurements were conducted only for aerosol filters collected during the 2021 Arctic and 2019 Antarctic cruises.

Aerosol ON and IN were simultaneously measured using the recently developed Aerosol Nitrogen Analyzer system, which enables sensitive quantification directly from filter samples without pretreatment. Detailed descriptions of the method are provided in Yu et al. (2021). Briefly, the method detection limit is 96 ngN. Because the detection limit scales inversely with the analyzed filter area, it can be readily lowered by analyzing a larger aliquot. In this study, 4–6 cm2 of filter material was typically analyzed for each sample, yielding a proportionally lower effective detection limit and ensuring stable and reliable quantification for low-concentration marine aerosol samples. Compared with traditional IC-based approaches, this analyzer provides a clear advantage by determining IN and ON simultaneously on the same filter aliquot, thereby avoiding the subtraction-based “difference method” (ON = TN - IN) and the associated uncertainty propagation when TN and IN are similar in magnitude.

The analyzer integrates a thermal aerosol carbon analyzer and a chemiluminescence NOx analyzer. Aerosol samples collected on quartz fiber filters were thermally evolved under a programmed 6-step temperature protocol (150, 180, 300, 400, 500, and 800 °C) in a 1 % O2/99 % He carrier gas. The evolved materials were catalytically oxidized to CO2 and nitrogen oxides (NOy), with the C signal monitored via methanator-FID detection and the N signal recorded through chemiluminescence after converting NOy to NO. The C signal assists in differentiating IN and ON components, as ON aerosols produce both C and N signals while the IN fraction only yields a N signal. The programmed thermal evolution facilitates separation of aerosol IN and ON due to their distinct thermal characteristics. Specifically, IN and ON discrimination is achieved by jointly interpreting the C and N thermograms: ON is identified by co-evolving C and N signals across the temperature steps, whereas IN is characterized by N-only evolution without a corresponding C signal. The separation of overlapping thermal features and the subsequent quantification of IN and ON were achieved through multivariate curve resolution (MCR) of the C and N thermograms, implemented using the USEPA positive matrix factorization (PMF) model (version 5.0).

To study air mass origins, air mass backward trajectories have been calculated using the Hybrid Single-Particle Lagrangian Integrated Trajectories (HYSPLIT) model with meteorological fields from the National Oceanic and Atmospheric Administration (NOAA) air resources laboratory GDAS database. 5 d backward trajectories were calculated in order to reveal the history of the air masses arriving at the sampling site (Stein et al., 2015). Each trajectory originated at the vessel's real-time position with an arrival height of 20 m, capturing boundary layer transport while minimizing local ship influence. Given that the ship was continuously moving and each sample integrates air masses over approximately 2–4° of latitude, we applied a nested strategy to account for spatiotemporal variability. For the initial characterization of the entire dataset, a representative sampling location was defined for each sample using the average latitude and longitude of its start and end positions, with backward trajectories simulated at 6 h intervals anchored to this midpoint to identify dominant air–mass categories (Fig. 3). In Antarctica, to precisely investigate the influence of sea ice on ON in the Southern Ocean and Antarctic marginal regions (Sect. 4.2), a targeted high-resolution analysis was performed on this subset of samples. For each Antarctic sample, the actual cruise track was equally divided into 48 points corresponding to the hourly intervals of the 48 h sampling period, and a 120 h backward trajectory was calculated for each of these 48 coordinates (Fig. S3a and b in the Supplement).

To determine whether the backward trajectories of the MABL samples were mainly influenced by the open ocean, sea-ice-covered regions, or the continental area, we calculated the time-weighted residence-time ratios of air masses over sea ice (RS), open ocean (RO), and the continental area (RC) using the following equation: 4RS(RO or RC)=∑i=1NS(NO or NC)×e-ti120∑i=1Ntotal×e-ti120 where Ntotal denotes the total number of trajectory endpoints; NS, NO, and NC represent the numbers of endpoints located over sea ice, the open ocean, and the Antarctic continent, respectively. ti is the backward-trajectory time (in hours), and ti/120 is a time-weighting factor (Zhou et al., 2021). This factor accounts for air–mass dispersion during transport and aerosol removal by particle deposition; thus, areas with longer backward trajectory times exert a weaker influence on the sampling point location compared to nearby areas. Accordingly, higher values of RS, RO and RC indicate greater influences from sea ice, the open ocean, and the Antarctic ice sheet, respectively.

Potential Source Contribution Function (PSCF) analysis was implemented to identify source regions of ON observed during the sampling period (Ashbaugh et al., 1985). A higher PSCF value indicates a greater potential source contribution to the receptor site. In our study, the PSCF domain was established within a grid cell encompassing all backward trajectories. The cruises were discretized into 1° latitude × 1° longitude grid cells. The PSCF value for cell ij was calculated as: 5PSCFij=∑mij∑nij where, nij = total trajectory endpoints within cell ij; mij = subset of endpoints associated with aerosol component concentrations exceeding the 75th percentile of cruise measurements. To mitigate uncertainty in cells with sparse trajectory density, a latitude-dependent weighting function (W) was applied: 6W=1.0 when nij>N20.8 when N1<nij<N20 when nij<N1 where nij is the number of trajectories passing for each cell in the study period and N1 = 60 × cos(latitude), and N2 = 300 × cos(latitude). The cosine factor is used to account for the changing grid cell size with varying latitude.

The Air–mass Exposure to Chlorophyll a (Chl a) index (AEC) serves as a quantitative metric to assess the influence of marine biogenic emissions on a target region through air mass transport (Blazina et al., 2017; Choi et al., 2019). This approach is grounded in the well-established correlation between ocean surface phytoplankton biomass and marine biogenic emissions, particularly dimethyl sulfide (DMS), where Chl a concentration acts as a robust proxy for phytoplankton abundance (Siegel et al., 2013). The AEC index estimates the integrated exposure of an air mass to oceanic DMS source regions along its trajectory by accounting for both spatial distribution of Chl a and atmospheric vertical mixing dynamics (Zhou et al., 2023).

For each trajectory point, Chl a concentrations (Chla_i) were obtained from satellite remote sensing products (Aqua-MODIS, OCI algorithm; 8 d composite, 4 km × 4 km resolution; https://oceancolor.gsfc.nasa.gov/l3/, last access: 31 October 2025) within a 20 km radius to reduce the influence of missing/cloud-contaminated pixels and pixel-scale noise, while remaining small enough to preserve local marine biological variability relevant to each trajectory point. The 20 km radius approach has been widely adopted in previous studies to mitigate the uncertainty of trajectory endpoints and ensure robust matching with satellite data coverage in previous research (Park et al., 2018; Zhou et al., 2021, 2023). Trajectory endpoints over Antarctica, sea-ice-covered areas, or at pressures < 850 hPa were assigned Chl a = 0 because air masses at these altitudes are generally decoupled from local ocean surface biological activity (Zhou et al., 2023). Points without valid Chl a data were excluded. The AEC for a single trajectory was computed as: 7AEC=∑i=1120Chlai×e-ti120n where ti denotes time backward along the trajectory (hours), and n is the total number of valid trajectory points. The time points when the air mass passed over the continent or regions covered by sea ice were assigned a zero chlorophyll value. To ensure robustness, trajectories with n < 90 (75 % of 120 h data points at hourly resolution) were discarded. For each sample, the final AEC value was derived from the arithmetic mean of all valid trajectories during the sampling period (Yan et al., 2024).

In this study, remote sensing data are utilized to illustrate the spatiotemporal distribution of sea ice concentrations (SICs) in the Southern Ocean. For regional-scale visualization of sea-ice extent (SIE) and SIC variability, we used the Sea Ice Index (Version 3) distributed by the National Snow and Ice Data Center (NSIDC) (Fetterer et al., 2017), which is derived from passive-microwave observations from DMSP SSM/I and SSMIS sensors (Cavalieri et al., 1997).

Sea-ice concentrations used here are derived from daily gridded passive-microwave SIC products, which provide all-weather coverage and are widely used for polar sea-ice monitoring. The SIC of each sample is calculated using the following formula: 8SIC=∑i=1NsSICiNs where SICi represents the average sea ice density at the endpoint of the specified track. Ns represents the total number of trajectory endpoints located on the sea ice area. For each trajectory endpoint, the SIC value was extracted by collocating the endpoint latitude/longitude and the corresponding day with the daily SIC grid; the SICi for each sample was then calculated as the mean SIC across all sea-ice-covered endpoints (Ns). Sea ice concentration data are from the AMSR2 dataset (Version 5.4. University of Bremen, Germany. Index of /amsr2/asi_daygrid_swath/s3125).

ON in the MABL primarily originates from two main source pathways: marine emissions and long-distance continental transport. Marine sources include primary ON, predominantly associated with sea-spray particles enriched in biological material from the ocean surface microlayer, and secondary ON. The latter not only derives from marine precursors such as alkylamines that react with acidic species (Altieri et al., 2016; Facchini et al., 2008), but also significantly involves the atmospheric oxidation of marine-derived biogenic volatile organic compounds (BVOCs). Specifically, isoprene and monoterpenes emitted from the ocean can react with hydroxyl (OH) or nitrate radicals (NO3) to form secondary organic nitrates (Fisher et al., 2016; Ng et al., 2017). Additionally, direct sea-to-air emissions of light alkyl nitrates produced photochemically in the surface water contribute to the MABL ON pool (Chuck et al., 2002). Continental sources involve the long-range transport of organic emissions – including combustion byproducts, soil- and vegetation-derived compounds, and biomass burning aerosols. It is important to note that these continental inputs include both ON formed directly over land and ON produced from continental precursors during transport (Duce et al., 2008; Li et al., 2025). This transport can significantly influence remote ocean regions (Cape et al., 2011; Jickells et al., 2013).

ON concentrations in the CEA region were the highest among all study regions, with air masses spending 22.6 % of their 5 d trajectories over continental areas (Fig. 3b). A significant correlation between ON and crustal elements such as nssCa2+ (r = 0.75, p < 0.01; Fig. 4) likely suggests the influences of continental transport of particles on the ON levels in this region (Xiao et al., 2017). A significant correlation between ON and the anthropogenic tracer EC (r = 0.81, p < 0.01; Fig. 4) indicates that fossil fuel combustion and biomass burning are important ON sources (Shubhankar and Ambade, 2016; Wu and Yu, 2016). Similarly, the robust association between ON and nssK+ (r = 0.78, p < 0.01; Fig. 4), a tracer of biomass burning, also supports contributions from agricultural and residential biomass burning (Song et al., 2018). Despite the high absolute ON concentrations, the relatively low ON/TN ratio (21.1 %) likely reflects disproportionately elevated IN emissions from intensive human activities, particularly NH3 volatilization from agriculture and vehicular NOx emissions (Pavuluri et al., 2015). This interpretation aligns with emission inventories that identify the CEA as a global nitrogen pollution hotspot, where ON is co-emitted or formed from precursors that share common sources with EC and other combustion-related pollutants, originating from incomplete combustion and industrial processes (Deng et al., 2024).

The SATO region exhibits intermediate level of ON concentrations (mean = 23.4 ± 18.0 ngm-3), lower than those influenced by anthropogenic activities in CEA but higher than in polar regions. In this region, ON shows a significant positive correlation with nssCa2+ (Fig. 4; r = 0.76, p < 0.01), suggesting that terrestrial mineral inputs (e.g., dust) influence ON levels, rather than purely marine sources. In addition, backward trajectory analysis showed that samples affected by continental air masses have significantly higher ON concentrations than those exposed solely to marine air (Fig. 3c), suggesting the influences of continental sources. However, ON does not exhibit significant correlations with nssK+ or with EC (p > 0.05), indicating that combustion emissions may not be the primary drivers. These findings suggest that the variability of ON in the SATO region results from a mixture of marine-terrestrial interactions, primarily modulated by episodic terrestrial mineral influence rather than continuous marine emissions. Notably, this region displays an elevated ON/TN ratio (Fig. 2), primarily due to its very low IN levels – approximately 85 % lower than in the CEA region – which amplifies the relative contribution of ON within TN.

In the AO region, ON concentrations were slightly lower than in the SATO region and significantly lower than in the CEA region. In this area, ON exhibited a significant positive correlation with Na+, which suggests the sea salts inputs (Fig. 4; r = 0.43, p < 0.01), and also showed significant correlations with nssK+ (r = 0.61, p < 0.01). Its correlations with nssCa2+ (r = 0.42) and EC (r = 0.42) were weaker but still significant (p < 0.05). These patterns suggest that ON in the AO region may originate not only from primary sea-salt aerosols but may also be linked to biomass burning. However, the backward trajectory analysis shows no significant difference in ON concentrations between air masses influenced by continental sources and those transported solely over the ocean (Fig. S1b; independent samples t-test, p = 0.16), likely suggesting the limited role of terrestrial inputs in this region. Unlike the SATO region, where ON showed no correlation with AEC (Fig. S2b; p > 0.05), ON concentrations in the AO exhibited a strong positive correlation with the AEC (Fig. S2a; p < 0.01), suggesting that marine biological activity is a key driver of ON variability in this region (Creamean et al., 2022). Collectively, these results demonstrate that the AO region is primarily governed by marine processes, with ON derived from both sea-spray organic enrichment and biogenic aerosol precursors, while terrestrial influences remain secondary (Nøjgaard et al., 2022).

In the SO region, ON concentrations were the lowest among all regions (mean = 12.0 ± 7.1 ngm-3), yet the ON/TN ratio was relatively high (27.8 ± 11.0 %). Backward trajectory analysis indicates that air masses predominantly originated from the open ocean and Antarctic continent (Fig. 3d), with minimal anthropogenic influence. ON here exhibited a significant positive correlation with Na+ (Fig. 4; r = 0.43, p < 0.01), but no significant relationships with nssK+, nssCa2+ or EC. While long-range transport events may deliver stable continental tracers like EC to this remote region, the lack of correlation between ON and these markers suggests that continental inputs are not the primary driver of ON variability. This pattern, combined with the association with Na+ suggests that primary sea-salt emissions are an important pathway for ON in the SO atmosphere (Matsumoto et al., 2022), likely through the incorporation of marine-derived organic matter into sea-spray aerosols. Meanwhile, the absence of associations with terrestrial tracers further supports the notion that ON in this remote region is likely influenced more significantly by natural marine processes rather than continental or anthropogenic sources.

Sea-ice and open-ocean environments create distinct conditions for the production and emission of ON. While sea ice restricts direct air–sea exchange, it hosts specialized microbial communities and accumulates organic matter within brine channels. During melt and ice-edge retreat, this organic material is released into waters characterized by high primary productivity (Arrigo et al., 2008). This biological intensification enriches the surface microlayer and supplies precursors for aerosolization via sea spray and secondary formation (Dall'Osto et al., 2017; DeMott et al., 2016; Galgani et al., 2016; Wilson et al., 2015).

Along the Antarctic coast, we classified samples into two groups based on air–mass histories: open ocean (OO), influenced almost exclusively by open-ocean trajectories, and sea ice (SI), with air masses residing over sea ice for extended periods. SI samples exhibited significantly higher ON concentrations and ON/TN ratios than OO samples (p < 0.001; Mann–Whitney U test; Fig. 5a and b). Multiple lines of evidence point to sea-ice–associated biological processes as the driver of these enhancements: (1) Strong positive correlations of ON with sea-ice concentration (SIC; r = 0.86, p < 0.01) and with air–mass exposure to chlorophyll a (AEC; r = 0.91, p < 0.01) in the SI group indicate that both ice cover and associated biological activity elevate ON (Fig. 5c and d); (2) PSCF analysis identifies high-probability source regions (PSCF > 0.8) over sea ice and its marginal zone for SI samples (Fig. S3c), consistent with an ice-edge origin; and (3) In contrast, ON shows no significant correlation with Na+ (r = -0.22, p > 0.05) or with IN (p > 0.05) for SI samples (Fig. S4a and b), suggesting that primary sea-salt emissions and purely abiotic inorganic pathways are not the dominant contributors.

These observations support a mechanistic pathway whereby organic matter released from sympagic (ice-associated) communities during melt enriches the surface microlayer and is transferred to the atmosphere via sea spray as ON-rich particles (DeMott et al., 2016; Wilson et al., 2015). Concurrently, a portion of this organic nitrogen is rapidly microbially degraded to volatile alkylamines (e.g., methylamines) (Taubert et al., 2017), which then form aminium salts through acid–base reactions with marine emissions-derived acids (e.g., H2SO4, MSA), contributing to both ON and IN in SI conditions (Brean et al., 2021; Dawson et al., 2012; Fitzsimons et al., 2023). This process results in the formation of both organic (amine salts, contributing to ON) and inorganic nitrogen aerosol species (NH4+ and NO3-), which explains their elevated levels in the SI group samples (Fig. S5). The elevated ON/TN ratios in SI samples (31.0 %) relative to OO samples (20.8 %) further indicate a greater fractional contribution of ON under sea-ice influence (Fig. 5b), consistent with reported releases of organic species from the sympagic ecosystem during melt (Jang et al., 2023; Mirrielees et al., 2024; Yan et al., 2020).

For OO samples, PSCF hotspots (PSCF > 0.8) shift toward the offshore Southern Ocean (Fig. S3d), in line with trajectories dominated by open-ocean air masses. The positive association between ON and oceanic residence time (r = 0.66, p < 0.01; Fig. S6) suggests that, as sea-ice influence diminishes, ON variability becomes increasingly governed by open-ocean biological processes and long-range marine aerosol transport.

Overall, these results establish the ice-edge/sympagic environment as an important regulator of Antarctic aerosol ON. Sea-ice dynamics modulate both the magnitude (higher ON and ON/TN) and sources (biogenic enrichment and amine-driven secondary formation) of ON, underscoring the need to represent sea-ice–associated processes in polar atmospheric chemistry and climate models.