Figure 4. The ice-active site density (ns) for aerosol sampled during the cruise (see Figure 2 for key to colours) compared to n sparameterisations for desert dust [Ullrich et al. , 2017], Icelandic dust [Sanchez-Marroquin et al. , 2020] and for aerosol from the North Atlantic [McCluskey et al. , 2018b] that is also consistent with values for aerosol over the Southern Ocean [McCluskey et al. , 2018a].
3.2 Ice-nucleating particle concentrations above the surface mixed layer
Eight flights with a balloon-borne size-resolved aerosol sampler were conducted while the Oden was at the ice station (see SI for flight details). This sampler, known as the selective height aerosol research kit (SHARK) was lofted to a defined height using a tethered balloon system operated from the sea ice and aerosol samples were collected into multiple size bins from below 0.25 µm to 10 µm with an additional stage for particles larger than 10 µm (with a poorly defined upper limit) [Porter et al. , 2020]. During these flights, we used a live link to the on-board temperature and humidity measurements to ensure that we sampled above the surface mixed layer and thus in air decoupled from the surface, but within the boundary layer (i.e. in the cloud mixed layer). Hence, the flights occurred at 390 m to 600 m altitude while the ship was on the ice-floe station and we sampled for 3 to 6 hours. In addition, we paused sampling when the RH was more than 80 % to avoid sampling biases associated with hygroscopically swollen aerosol and we also avoided sampling while the SHARK was enveloped in cloud.
Given the surface mixed layer is often decoupled from the rest of the boundary layer, these measurements in principle allow us to compare INP concentrations within and above the surface mixed layer. The values ofT [INP]=0.1 are shown in Figure 2b, whereas the INP spectra are shown in Fig SI2 for the INP summed across all the particle size bins. It should be borne in mind that, for practical reasons, the sampling durations on the ship and on the SHARK were not the same, however it is still possible to draw conclusions from this comparison. There is evidence that there are substantial differences between the INP concentrations in the surface mixed layer compared to above it. For example, on the 5th and 8th September theT [INP]=0.1 (summed across all sizes) was around −18 to −19 °C above the surface mixed layer, whereas it was below –26 °C within it. However, on the 20th August,T [INP]=0.1 was around –23 °C above the surface mixed layer, but –14 °C within it. On all three days, radiometer and radiosonde temperature profiles confirmed that the surface mixed layer was decoupled from the rest of the boundary layer (Table S2). In contrast, on 13th September, the surface mixed-layer was mainly coupled and the INP concentrations within and above the surface mixed layer were similar. However, on two other days (23rd August and 10th September) the activity in the surface mixed layer and above were similar even though it was decoupled. Overall, out of the eight SHARK samples collected above the surface mixed layer, there was one SHARK sample that had much lower ice-nucleating activity than that in the surface mixed layer, three samples with higher activity, three with similar activity and one that was ambiguous (due to both samples being close to the baseline of detection). This is consistent with the air at the surface sometimes being coupled to the cloud mixed layer, allowing transport of aerosol throughout the boundary layer, but at other times the measurements at the surface are not representative of those above the surface mixed layer.
The size-resolved INP activity (T [INP]=0.1) is also shown in Figure 3b. In many locations around the world, supermicron aerosol dominate the INP population [Porter et al. , 2020]. However, contrary to what might be expected, the smallest size ranges of < 0.25 μm contributed the most INPs on five out of the eight flights, with the 2.5 to 10 μm and 0.5 to 1 μm bins both contributing the most on one flight each. Inspection of the corresponding INP spectra associated with each bin (Fig SI2) revealeds that the particles < 0.25 μm made a pronounced contribution to the INP population on the 23rd August and the 8th and 9th September. The other flights produced data mainly in the baseline for all sizes.
The coarse mode (>2.5 µm diameter) has a relatively short lifetime in the Arctic boundary layer, being removed effectively by wet scavenging processes [Leck and Svensson, 2015]. Hence, it is perhaps not so surprising that the fine mode aerosol (< 0.25 μm) appears to be so important in this region for the INP population. While INPs are typically thought of as being the larger particles in a size distribution [Mason et al. , 2016; Porter et al. , 2020], there are INPs that fall into the < 0.25 μm size range that are also very active. For example, film droplet aerosol resulting from wave breaking are produced in a range of sizes centred around 100-200 nm and are often rich in organic material [O’Dowdet al. , 2004] that is known to include small ice-nucleating entities [Schnell and Vali, 1975; Wilson et al. , 2015]. Alternatively, ice-nucleating macromolecules from terrestrial biological sources internally mixed with other aerosol particles might fall into this size range [O’Sullivan et al. , 2016; O’Sullivan et al. , 2015; Pummer et al. , 2015] and it has been proposed that fungal material, some of which is known to act as an INP [O’Sullivanet al. , 2015], can fragment to form nanoparticles [Lawleret al. , 2020].
3.3 Correlation between INP concentrations and dimethyl sulfide, equivalent black carbon and aerosol surface area
To investigate possible sources of the INPs we detected over the central Arctic Ocean, we have correlated the ice-nucleating activity of the aerosol with: i) dimethyl sulfide (DMS), a product of marine biological activity, particularly in the marginal ice zone (MIZ, the transitional zone between open sea and dense ice) [Leck and Persson, 1996]; ii) equivalent black carbon (eBC), based on aerosol absorption at 637 nm; and iii) aerosol surface area, derived from size distribution measurements. We present the time series for aerosol particle surface area, DMS and eBC concentrations, as well as the Pearson’s rcoefficient between ice-nucleating activity and each quantity in Figure 3a.
DMS is found in the marine atmosphere, originating from the metabolites of some marine algae [Leck and Persson, 1996; Lohmann and Leck, 2005]. Hence, the presence of DMS indicates that an air mass has origins in a location rich in biological activity, which may also be expected to correlate with marine biological INP sources. DMS is thought to be relatively short-lived in the atmosphere, with a lifetime on the order of 1-3 days [Kerminen and Leck, 2001; Khan et al. , 2016]. Therefore, it is a useful indicator for the interaction of air masses with the MIZ at the outer edge of the pack ice region, and possibly the open leads or melt ponds within the pack ice if they were producing DMS at that time.
The concentration of DMS during the cruise was highest in the outbound 24-hour MIZ station (2nd-3rdAugust), where the ship was close to open water, but was variable whilst in the pack ice (Figure 3), and remained relatively low in the inbound MIZ station (19th September). The data in Fig. 2 clearly shows that there is no obvious correlation between DMS and INP activity (r = 0.15) suggesting that MIZ marine biogenic sources exerted little influence on the measured INP concentrations.
Equivalent BC (eBC) is a quantity derived from aerosol absorption and is the equivalent black carbon mass concentration needed to produce the observed absorption. Other aerosol types such as dust, brown carbon or other organic aerosol might also produce absorption, thus potentially contaminating the small signal we observed. However, absorption by BC is much stronger at 637 nm than other materials, hence the signal is most likely dominated by BC. BC is produced through a range of combustion processes, including biomass burning, wildfires and fossil fuel combustion, which are all remote from the central Arctic. Other potential contributors to the absorption signal, such as dust or brown carbon are also remote from the central Arctic. Rigorous procedures were in place to ensure that BC (and other aerosol) from the ship stack did not affect measurements (see methods for details). Therefore, eBC is used here as an indicator of long-range transport. The literature indicates that BC is a relatively ineffective ice nucleator under mixed-phase cloud conditions [Adams et al. , 2020; Chen et al. , 2018; Schill et al. , 2020; Vergara-Temprado et al. , 2018a], hence we would not necessarily interpret a positive correlation as an indication of ice nucleation by BC. However, combustion processes are thought to be a source of ice-nucleating aerosol, even if BC itself is not an effective INP [Barry et al. , 2021; Jahn et al. , 2020; Umo et al. , 2015]. Thus, a correlation between BC and INP concentrations would indicate that aerosol particles transported along with BC from outside the central Arctic Ocean nucleate ice. Wildfires around the Arctic are a potential source of BC, and we note that during the cruise there were persistent Siberian wildfires, as can be seen using the NASA Worldview satellite imagery tool (e.g. clear skies on the 14th August reveal widespread fires; https://worldview.earthdata.nasa.gov/). There is also industry, shipping and mining along the Arctic coast of Russia, but the role of gas flaring is thought to be a particularly important source of BC [Stohl et al. , 2013].
The overall correlation between eBC and INP concentration (r = 0.65) is much stronger than for DMS. In fact, the eBC concentration appears to track the INP concentration in Figure 3 until the 27th August, after which the INP concentrations stay relatively low whilst the eBC remains highly variable. The decoupling of eBC and INPs later in the campaign indicates that distant sources of BC are not always related to distant sources of highly active INPs.
The surface area concentration of the bulk aerosol follows a similar trend to the eBC concentrations, but with a slightly weaker correlation with the INP activity (r = 0.52). This indicates that the variability in INP concentrations at the North Pole is not simply driven by aerosol surface area, rather that some specific component(s) of the aerosol population are ice-active and these particles are most likely associated with specific sources at latitudes further south than the MIZ.
3.4 Trajectory analysis of aerosol collected in the central Arctic
Backward trajectories from the sampling location near the surface are presented in Figure 5a. We only show points that are in in the boundary layer and for a maximum of 7 days. The figure shows a clear relationship between the origin of the aerosol – considered here as the boundary layer points along the trajectories - and the measured INP concentrations. The origin of the air with the most active INPs is around the Russian Arctic coast including the Barents, Kara and Laptev Seas. Out of the 30 filter runs, those filters with the highest INP activity (the top 20 % of filters; −9 °C ≥T [INP]=0.1 > −13.5 °C) sampled air masses originating over the Barents and Kara Seas. The next seven highest (23 %; −13.5 °C ≥ T [INP]=0.1> −22 °C) filters, in terms of INP activity, sampled air originating from over the Laptev and East Siberian Seas. The next six filters with lower INP activity (20 % of filters; −22 °C ≥T [INP]=0.1 > −25 °C) sampled air that originated off the eastern coast of Greenland from over both the pack ice and open ocean. The 11 filters with the lowest INP activity (bottom 37 % of filters; −25 °C ≥ T [INP]=0.1≥ −30 °C) all sampled air which mostly originated from the pack ice adjacent to North America (also see Figure SI3).