The Arctic Century Expedition took place from 5 August to 6 September 2021. Collocated measurements of atmospheric and marine physics and chemistry were conducted on the research vessel (RV) Akademik Tryoshnikov. Figure a shows the route of the expedition. The Arctic Century Expedition started and ended at the harbor in Murmansk, Russia (68.98° N, 33.09° E) and explored an extensive area in the Eurasian Arctic Seas, including rarely accessible locations in the Kara and Laptev Seas, and the archipelagos of Franz-Josef Land, Novaya Zemlya, and Severnaya Zemlya.

A comprehensive set of atmospheric aerosol sampling and measurement (online and offline) was conducted on board. The locations of the measurement set-ups are indicated in Fig. b, and instrumentation at each location is given in Table . On the 2nd deck, monitoring and sampling of ambient aerosol were conducted from an aerosol container laboratory set-up following the configuration described in . A combination of online INP measurements and sampling for offline INP analysis was used to quantify NINP. Additionally, the aerosol concentration and number size distribution were monitored continuously. On the top deck (6th deck), filter samples were collected for INP analysis after the expedition. Details on instrumentation, sample collection, and analysis are provided below.

From the aerosol container laboratory, ambient aerosols were collected 3 times a day, for 3 h each, into 15 mL ultra-pure water (W4502-1L, Sigma-Aldrich), using the high flow-rate impinger (Coriolis^® µ, Bertin Instruments, with a lower limit aerodynamic cut-off size of 0.5 µm) at a flow rate of 300 Lmin-1. Additional ultra-pure water (W4502-1L, Sigma-Aldrich) was constantly supplied to the sampling container during the operation of the impinger via a refilling system to compensate for evaporation loss. On the top deck, aerosol particles were collected onto 47 mm polycarbonate membrane filters (Whatman, 0.4 µm pore size) using a low volume sampler (LVS, Model DPA14, Digitel) with a 10 µm particulate matter (PM₁₀) inlet that excludes particles that are larger than 10 µm in diameter from being collected (samples are hereafter referred to as PM₁₀ filters). The LVS inlet was approximately 25 m above sea level (m a.s.l.). The operating flow rate was maintained at 38.3 Lmin-1 for 12 h sampling intervals, which was selected to be consistent with the PM₁₀ inlet design, ensuring the aerodynamic cutoff size of 10 µm was maintained. A trade-off was made between sampling flow rate (38.3 Lmin-1) and duration (12 h) to maximize total sampled volume, achieve suitable temporal resolution, and match the INP detection range. The impinger and LVS were located on the 2nd and 6th decks, respectively, with a vertical separation of over 10 m. Under stratified atmospheric conditions, especially near the ocean surface, the vertical variability of aerosol properties between different sampling heights cannot be entirely ruled out. However, given that both are in the boundary layer, we do not expect large differences in aerosol number and composition between the two sampling locations, except those introduced by the sampling method (impinger samples particles >0.5 µm aerodynamic diameter). In addition, to minimize contamination from ship exhaust, specific procedures were followed for the two sampling locations. On the 6th deck, the PM₁₀ filter sampling (LVS) was automatically paused whenever wind direction sensors indicated air flow from the ship's funnel. For impinger and HINC measurements on the 2nd deck, spikes in total particle number concentration measured by a CPC were used to identify exhaust plumes; these periods were flagged and removed from the dataset. These measures substantially reduce the influence of exhaust emissions on the reported INP data, although minor residual contamination cannot be fully excluded.

The impinger samples and PM₁₀ filters were stored at -20 °C on board, for transport, and after the campaign at the ETH laboratory until analysis. During the campaign, a total of 75 impinger and 50 PM₁₀ filter samples were collected.

The cumulative NINP as a function of freezing temperature from both online and offline measurements using HINC (T=-34 and -30 °C), impinger samples, and PM₁₀ filters (T≥-25 °C) is shown in Fig. . Across the assessed freezing temperatures, NINP spans 1 to 3 orders of magnitude, overlapping the parameterization derived from measurements in Svalbard except for impinger samples. The discussion on the difference in NINP measured across HINC, impinger samples, and PM₁₀ filters was detailed in Appendix . Figure  shows a comparison of our observed NINP from the Arctic Century Expedition to the range of NINP derived from midlatitude precipitation samples (, the range between the light magenta lines). Overall, NINP observed during the Arctic Century Expedition is up to 2 orders of magnitude lower compared to the NINP at mid-latitudes, especially at T<-15 °C. The significantly lower NINP at T<-15 °C, where mineral dust becomes a dominant source of INPs , indicates that the concentration of mineral dust INPs is lower over the Eurasian-Arctic Ocean compared to mid-latitudes. Smaller differences in NINP were found at T>-15 °C, in particular for T>-10 °C, where NINP from the Arctic Century Expedition aligned with the range reported by . For T<−10 °C, the variability in the impinger INP data and the decrease to the lower bound of the line could display a more pronounced trend than currently presented, as our measurements are limited by the lower LOD of the analytical method. For a comparison within different Arctic regions, Fig.  also includes ship-based INP observations in the summer over the Arctic Ocean from previous work (). Our offline NINP are similar to NINP reported by . However, the NINP measured with HINC at T=-34 and -30 °C were systematically lower than the low-temperature observations during this summer measurement campaign circumnavigating Svalbard from . A difference in the concentration of mineral dust INP could be the reason. reported an abundance of mineral dust from Greenland and adjacent Svalbard, following ice and snow melt (see for information on Arctic dust sources). In addition, our summertime NINP spectra over the Eurasian-Arctic Seas are largely consistent with recent year-long observations from the Central Arctic (; represented by orange crosses in Fig. ). reports that biological INPs dominate the Arctic spectrum year-round, with a significant seasonal peak in early summer (June–July) reaching concentrations up to 1.4 L-1 at -15 °C. While our late-summer observations (August–September) generally fall within the range of these Central Arctic measurements, they do not capture the extreme peaks observed earlier in the melt season. This discrepancy likely reflects the temporal shift in marine productivity and terrestrial runoff, which peaks prior to our late-summer campaign. Nevertheless, the overall consistency in NINP magnitudes and temperature-dependent trends across these two geographically distinct sectors suggests that the Eurasian Arctic Seas share a similar INP regime with the Central Arctic Ocean, characterized by regional emissions from the marginal ice zone and snow-free land surfaces. Other Arctic NINP measurements above -30 °C by , , , , corroborate our observations.

At most freezing temperatures, NINP from PM₁₀ filters was observed to be higher than that from impinger samples. This difference could be influenced by the larger volume of air processed by the impinger (approximately 54 m3) compared to the PM₁₀ filters (approximately 27.6 m3), which shifts the detectable range to lower NINP for the impinger samples. The minimum detectable NINP with DRINCZ decreases as the air volume increases. Therefore, the NINP detectable from impinger samples is lower, which is particularly relevant at the highest freezing temperatures. For T<-10 °C, the variability in the impinger INP data and therefore the difference to the lower limit of the line could potentially be even more than currently presented because our data are limited by reaching the lower limit of detection of our analysis method. Several other factors, listed below in no particular order, may also account for the differences in NINP measured from impinger and PM₁₀ samples.

Aerosols captured by the impinger are directly immersed in water, potentially reducing the IN activity of certain particles, such as mineral dust . Filter samples, due to their shorter water exposure before INP analysis, are less likely to undergo such degradation.

Compared to the offline measurements, NINP measured with HINC at -34 and -30 °C generally align better with the values from the PM₁₀ filters than those from impinger samples. The observed alignment of NINP from HINC with those from PM₁₀ filters could come from HINC also sampling particles with sizes below 2.5 µm , therefore covering the sub-0.5 µm particles like the PM₁₀ filters.

Overall, for T<-15 °C, we observe lower NINP in the Eurasian Arctic compared to the mid-latitude range. We propose that this is due to lower concentrations of dust. Some of the observed NINP values fall within the lower end of the range reported for biological INPs in continental mid-latitude regions (T<-15 °C). This supports the contribution of MBAs with ice-nucleating potential over the Eurasian Arctic Ocean, despite the overall NINP in our study being lower than typical mid-latitude values.

Figure  shows the correlations between NINP and parameters related to particle size (see Table  for the correlation parameters). Among the investigated freezing temperatures, a significant, moderate correlation was exclusively observed at T=-34 °C for all parameters. At this temperature, the notable positive correlations between NINP, particle concentration, and surface area across all sizes indicate that as the freezing temperature decreases and approaches the homogeneous freezing temperature, a considerable fraction (1 in 102) of all particles become IN active. Conversely, at higher temperatures, NINP demonstrates a mostly insignificant correlation with bulk aerosol properties (except the small negative correlation with S>0.5). This decoupling can be attributed to INPs constituting a small fraction of ambient aerosols – only about 1 in 106 particles are IN-active at -15 °C. A recent study observed even lower NINP at -15 °C in Svalbard in spring, summer, and autumn (9×10-9–7.4×10-6 L-1), demonstrating that lower concentrations than what we report are possible. The significant but weak anti-correlation between NINP and S>0.5 is unexpected, pointing towards higher NINP when the particle number concentration is low, which is the case when INPs come from a local source that has not been diluted by transport. This suggests the contribution is local at warmer temperatures where bioaerosols are IN-active. Additionally, the high abundance of sea salt particles, which contribute significantly to total surface area but are largely inactive as INPs, could further dilute the relationship between NINP and surface area concentrations. Moreover, the presence of mixed aerosol types with varying ice-nucleating efficiencies can further obscure any direct scaling with bulk surface area. However, we note that the small dataset size limits a more comprehensive interpretation. Nevertheless, our findings in the maritime Arctic environment challenge the established parameterization of ambient NINP based on specific aerosol sizes and surface areas (e.g., ).

To further investigate the nature and source of INPs, the elemental composition of the ambient aerosol was determined from the impinger samples. Figure  exhibits the correlation analysis between NINP and multiple representative elements in the bulk aerosol samples. Statistically significant correlations were not found for NINP at -10 and -15 °C, likely due to the small ice active fraction of particles. In general, NINP is poorly correlated with the concentration of sea salt, which is expected since the soluble Na and Mg containing sea salt particles typically do not act as immersion INP. Despite the weak correlations at -10 and -15 °C, NINP increases with the concentration of AlSiCa (indicator for mineral dust), indicating an overall weak contribution of dust or covariant INP species over the Arctic Ocean. Additionally, a significant but weak correlation was found between NINP at -20 °C and phosphorus concentrations, and moderate but not significant correlations are observed between NINP and sulfur concentrations. In the biogeochemical cycle in the Arctic Ocean with minimal anthropogenic influence, phosphorus plays a crucial role as a nutrient for marine biology. Similarly, marine-released sulfur, often derived from dimethyl sulfide (DMS), a byproduct of marine phytoplankton and microbial metabolism (e.g., ), suggests that the appearance of marine biological activity, such as phytoplankton blooms, could enrich the INP population . While directional trends between NINP and elemental tracers (e.g., AlSiCa, P, and S) are broadly consistent with potential mineral dust and marine biogenic influences, the generally weak to moderate correlations and limited statistical significance at several temperatures suggest that bulk elemental composition alone cannot fully resolve the diversity of INP sources. This is due to the chemical and physical complexity of the ambient aerosol population that are influenced by multiple potentially correlated factors in the Eurasian Arctic, where marine, terrestrial, and biogenic particles frequently co-emit and mix during transport. Furthermore, highly effective INPs constitute only a small fraction of total aerosol mass and are not adequately represented by bulk tracer signals. To disentangle the overlapping terrestrial and marine contributions to INPs, particularly in the remote and dynamic Arctic oceanic environment, future field studies would benefit from the use of more specific molecular tracers (e.g., molecular organics, biochemical tracers like DNA and lipids, and isotopic ratios), single-particle analyses, larger sample sizes, and multivariate data approaches, the latter of which are not possible with the limited dataset we have from this single cruise.

Figure  shows the geographical variability of NINP measured along the ship track with HINC and PM₁₀ filters. Additional maps for NINP measured from the impinger samples are shown in Fig.  in Appendix . Generally, there is a tendency that at higher latitudes, lower NINP are observed at all temperatures, except PM₁₀ filters collected near the Severnaya Zemlya, highlighting the influence of nearby terrestrial sources. NINP vary with sea ice coverage, with the highest NINP often occurring over the ice-free ocean or marginal ice zone (MIZ), particularly near land, and less so within the dense ice pack. Figure a categorizes INP temperature spectra by sea ice cover. Observations at T>-12 °C over the ice pack showed NINP below the LOD. T-test results indicate that at T≥-20 °C, the mean NINP is systematically, but not significantly higher over the ice-free ocean than over the ice pack and higher in the MIZ than on the ice-free ocean. Mean NINP at -30 and -34 °C are significantly different over the ice-free ocean compared to within the ice pack or the MIZ. Figure b summarizes the activated INP fraction at selected temperatures based on the locations of measurements. For aerosols in the ice-pack region, despite being less numerous, they may possess higher intrinsic IN efficiency, particularly at cold temperatures (T<-20 °C). Similarly, at warmer temperatures (T≥-20 °C), the activated INP fraction in the MIZ is notably higher than in the open Ocean, suggesting a lower total aerosol background and a higher relative abundance of warm-temperature INP, likely driven by marine biological activity or wave breaking at the MIZ. Potential sources include biogenic INPs originating from marine biota, such as phytoplankton exudates , augmented by nutrient influx from the warmer, saltier Atlantic waters entering the Arctic Ocean during ice-melting seasons (a process known as Arctic Atlantification, ), and mineral dust input from river runoff and thawing permafrost . An additional potential source of INP in the MIZ that has previously been reported is sea ice algae growing at the marginal or seasonal ice zone and in open leads .

To investigate the variability in NINP with airmass origin, 2 d backward trajectories starting from the boundary layer at the ship's location during observation periods were calculated. Figure a and b show the trajectories colored according to the observed NINP at T=-34 °C and T=-15 °C, respectively, at sea level. For both temperatures, the highest NINP were captured when the ship approached the northern coast of Novaya Zemlya, where the air masses originate from the western Siberian coast near the estuary of the Pyasina River. Local terrestrial sources from Novaya Zemlya or long-range aerosol transport may be significant contributors to these high-NINP cases. reported a similar air mass origin for the highest NINP active at temperatures above -15 °C. They measured in the central Arctic in 2018 during the summer season, suggesting the Novaya Zemlya region can be a strong Arctic INP source. Previous studies have reported that the shallow seas off the Siberian coast and Arctic archipelagos are heavily affected by fluvial discharge rich in organic matter, silt, clay, and nutrients . Thawing permafrost in the summer enhances the mobilization of soil, nutrients, and active microbes into rivers and runoff into the ocean . Aerosolization of silt, clay, and soil components at the water-air interface may contribute to the NINP at T=-34 °C . Elevated marine biological productivity due to increased nutrient availability from fluvial input could have enhanced NINP at T=-15 °C. The INPs are likely locally generated dust and biological particles is further supported by the chemical composition time series shown in Fig.  (Appendix ), where at the corresponding ship location around 3 September, high concentrations in all categorized compositions were observed. Further evidence that local sources are mainly contributing to the NINP at -34 °C, comes from moderately elevated NINP when the ship approached Franz-Josef Land and Severnaya Zemlya (Fig. a). In these locations, air masses often originated from the central Arctic with high sea ice coverage, where the INP population is scarce, indicating local influences may dominate the INP source over long-range transport. Similarly, at T=-15 °C, a few high INP occurrences coincided with air masses originating from the MIZ or ice packs, where the ship was located over the open ocean close to Severnaya Zemlya (Figs. b). Local aerosol emissions from the ocean and the island likely contributed to the high INP population at this temperature. This finding is corroborated by extended 7 d back trajectories (see Fig.  in the Appendix). Notably, elevated NINP observed near Novaya Zemlya at both -34 and -15 °C coincide with air masses characterized by prolonged residence over the western Siberian coast – a hotspot acting as a potent source for both mineral dust and biological particles as discussed previously. This regional terrestrial contribution is further substantiated by the vertical transport history of the pathways (Figs.  and in the Appendix). Throughout the 7 d transport period, the air masses associated with the elevated NINP events near Novaya Zemlya remained predominantly confined to the lower troposphere, traveling below 900 hPa while traversing the Western Siberian coast. This persistent low-altitude transport indicates that the air parcels were actively coupled with the surface boundary layer, facilitating the effective entrainment and transport of regional mineral dust and biogenic particles to the marine boundary layer over the Eurasian-Arctic Seas. Conversely, trajectories showing low NINP predominantly originated from aloft or spent minimal time within the continental boundary layer, confirming that surface-level source interaction is the primary driver of the observed NINP enhancements.

Figure c and d display 2 d backward trajectories classified by the overpassed types of surface during air mass history. In addition to the western Siberian coast, air masses containing moderate to high NINP also passed over ice-free ocean and land. In contrast, air mass trajectories passing over the ice pack or MIZ contained lower NINP. One factor could be the absence of aerosolization via wave breaking and bubble bursting where there is sea ice. A correlation analysis between NINP and the residence time of air masses over different types of surface is shown in Fig. , with correlation coefficients provided in Table  in the Appendix. Negligible to no correlations were found between NINP and the percentage of time that trajectories spent in the boundary layer over the four different types of surface (i.e., ice-free ocean, land, MIZ, or ice pack), suggesting no clear pattern emerges, which surface types are most important for high NINP. The missing correlation between surface types along trajectories and NINP along the ship track indicates that local INP sources are important in shaping INP population in the Eurasian-Arctic Ocean. From the combination of all these observations, it can be concluded that long-range transport generally plays a less important role in the concentration of INPs over the Eurasian-Arctic Ocean during the summer season than local emissions.

In this section, two selected periods of low and high NINP during the campaign are discussed in detail. The analysis includes measurements of aerosol size, elemental analysis, and air mass history. The periods are selected based on measured NINP at T=-20 °C, where NINP were in the bottom and top 25 % quartiles for the low and high NINP period, respectively, and where parallel aerosol characterization was available (see Fig.  in the Appendix for NINP time series). The ship's location during the two periods and back trajectories of the sampled air are shown in Fig.  (the extended 7 d back trajectories are shown in Fig.  in the Appendix). The low NINP period occurred when the ship sailed northeast of Franz-Josef Land at the edge of the MIZ, with the majority of air parcels originating from the northeastern Arctic ice pack, where the sea surface is mostly covered by ice and lacked terrestrial and marine aerosol sources. In contrast, during the high NINP period, the ship was sailing along the western coast of Novaya Zemlya in the ice-free ocean, with air parcels rich in SSA, dust, and biogenic particles arriving from the west Siberian coast, suggesting INP influence from both local source and long-range transport. Similar patterns have also been observed in the Alaskan Arctic , where lower NINP were linked to air masses originating from the central Arctic and regions with high sea ice coverage, while higher INP episodes corresponded to transport from southerly, mid-latitude regions. During this time, there is a substantial presence of melt ponds up to 20 %–30 % of the sea ice surface , the peak primary productivity has already passed by late summer , leading to low local sources of INP coming from the ice-covered region.

Figure a shows the average particle size distribution during the time windows of the two periods (see particle size distribution time series in Appendix Fig. ). Notably, particle concentrations were 3 orders of magnitude higher across all size ranges during high NINP. Particularly, aerosol particles with diameters larger than 5 µm were only detected during the high INP period and could be responsible for the elevated NINP measured from impinger and PM₁₀ filters. For particles below 2.5 µm captured by HINC, the high concentration in this size range could have caused the high NINP at -30 and -34 °C. Despite larger particle concentrations, Fig. b reveals consistently lower activated INP fraction during high NINP event at all selected temperatures, which could be caused by the dominance of ice-inactive aerosols, being enriched to a higher proportion in the air mass, such as sea salt particles. Figure c shows the abundance of tracer elements detected in impinger samples to support the case study. Sea salt concentrations were over 100 times higher in the high NINP sample compared to the low NINP sample, likely due to less saline seawater and ice covering the sea-air interface in the low NINP case. Additionally, the average wind speed at the sea surface was much higher (18 to 24 ms-1) in the high NINP period than during low NINP (3 to 6 ms-1) due to the passage of a cyclone during the high INP case. Since the concentration of local sea spray aerosols strongly depends on wind-induced wave breaking and bubble bursting , a heavier aerosol load, carrying IN-active MBA can be expected during high NINP. The higher phosphorous and sulfur contents during high NINP support the hypothesis of fluvial-marine transport of nutrients from meltwater runoff adjacent to Novaya Zemlya. Furthermore, approximately 25 times more AlSiCa (indicator of mineral dust) was detected in the high NINP sample compared to the low NINP sample. Albeit long-range transport of mineral dust cannot be ruled out, the local dust sources likely dominate as indicated by the concurrent elevated mineral dust and sea salt concentrations (see Figs. c, and in the Appendix). The local dust sources can include re-suspended dust previously deposited at the ocean surface , aerosolization of muddy water surrounding the island, or wind-blown dust from the Novaya Zemlya coast due to high wind speed. In summary, differences in aerosol concentrations and the different air parcel origin led to a higher INP concentration during high winds from a southeasterly direction. The variation in strength of local island or ocean sources, as well as long-range transport of INP from the Siberian coast, could explain the difference in NINP for the two periods.