Introduction

The occurrence of persistent mixed-phase clouds consisting of both supercooled liquid droplets and ice crystals in the Arctic lower troposphere has crucial impacts on the energy and hydrological budgets in the Arctic1. The clouds interact with longwave radiation throughout the year and downwelling shortwave radiation (sunlight) during polar day, and their properties and lifetime are strongly dependent on the amount of liquid water and ice. It is recognized that liquid-dominated clouds tend to have a higher reflectivity and longer lifetime than ice-dominated clouds2. Therefore, the phase composition and transition have important implications for the microphysical properties of mixed-phase clouds.

Climate change may compromise the nature of Arctic mixed-phase clouds. In general, the transition toward clouds with more liquid water and less ice is viewed as a robust feature of climate model responses to Arctic warming, because the formation of ice crystals in mixed-phase clouds may be suppressed in a warmer atmosphere3,4,5. This prevailing view appears to be simple and straightforward, but there are several potential sources and processes that may contribute to the phase transition of the clouds in a warming climate. For example, aerosols influence the cloud microphysics, but the mechanisms responsible for ice formation remain debated, partly due to lack of knowledge about the aerosol population capable of nucleating ice (i.e., ice-nucleating particles (INPs)) in the Arctic1,5,6,7. Primary ice production in mixed-phase clouds is thought to be induced by glaciation of a small subset of supercooled liquid droplets that contain efficient INPs8,9,10,11. Ice crystals formed via primary ice production and subsequent ice multiplication processes can grow more rapidly than supercooled liquid droplets at the expense of the droplets2. Hence, despite their lower amounts, the presence of INPs has the potential to largely reduce the liquid water content in mixed-phase clouds and make the clouds less reflective and short-lived, leading to more solar radiation reaching the Arctic surface. However, such aerosol-related processes that may influence the cloud-phase transition are not necessarily well represented in current climate models. This limits our ability to predict how Arctic mixed-phase clouds will change in response to climate change and vice versa.

The source and abundance of INPs and their spatio-temporal distribution in the Arctic atmosphere are still not well understood, but there is mounting empirical evidence that ice crystal formation in Arctic mixed-phase clouds over snow/ice-free areas could be activated in response to local/regional INP emissions12,13,14. The INP data in the Arctic are sparse, but recent field studies have indicated that the amount of INPs tends to show higher values in warm seasons, possibly due to the influence of natural aerosols released from snow/ice-free areas in the Arctic15,16,17,18,19,20,21. Such natural aerosols might include dust and organic materials originating from high-latitude oceanic and terrestrial sources15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33. It is important to note that natural aerosol emissions are projected to increase in response to rising surface air temperature (SAT) in the Arctic7,34,35. However, we still have a poor understanding of how INPs change in response to Arctic warming and then influence the phase transition scenario of Arctic mixed-phase clouds.

Here, we focus on the variation of INPs in the Svalbard region, which is recognized as one of the most climatically sensitive regions in the world. Satellite observations have shown the frequent occurrence of Arctic mixed-phase clouds over this region36. Importantly, recent studies have reported that the largest warming takes place over the Eurasian sector of the Arctic Ocean near Svalbard and Novaya Zemlya, which has been warming five to seven times faster than the global average37,38. There is thus an urgent need to improve our understanding of the links between Arctic mixed-phase clouds and climate over the Svalbard region. To evaluate possible impacts of aerosols on ice formation processes in the clouds, it is essential to understand the amount of INPs active under conditions relevant for mixed-phase clouds and the sources and processes that contribute to its variation. Our year-round measurements provide insights into the warming-driven changes in the population of natural aerosols serving as INPs in the Svalbard region.

Results

Variation of INPs

We have conducted continuous year-round measurements of INPs active in the immersion mode over the temperature regime down to −30 °C (i.e., under conditions relevant for mixed-phase clouds) at the Zeppelin Observatory located on Mt. Zeppelin in Svalbard (Fig. 1a and Supplementary Fig. 1). To quantify the number concentrations of INPs (NINP), we have analyzed the ice-nucleating ability of weekly aerosol samples using a cold-stage-based immersion freezing method16,39,40. In this paper, we present the INP data obtained during the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) project period41 from September 2019 to September 2020, which overlapped with the period of the Ny-Ålesund Aerosol Cloud Experiment (NASCENT) campaign42. The meteorological data at Ny-Ålesund located at the foot of Mt. Zeppelin show unusually cold conditions in March 2020 (Fig. 1b), possibly due to anomalously strong and persistent northerly winds associated with frequent cyclone events around the central Arctic43. Nevertheless, the long-term SAT data demonstrate an increasing trend over the last couple of decades and record the highest value in July 2020 (Fig. 1b).

Fig. 1: Continuous measurements of ice-nucleating particles (INPs) and meteorological parameters in Svalbard.
figure 1

a Location of the measurement sites. The map is based on satellite images in summer 2020 retrieved from TopoSvalbard (https://toposvalbard.npolar.no/). The position of Svalbard in the pan-Arctic environment is shown in Supplementary Fig. 1. b Changes in the monthly mean surface air temperature (SAT) at Ny-Ålesund over the period 2000–2020 and the MOSAiC project year. c Time series of the weekly mean number concentrations of INPs (NINP) active at a given temperature (T) at the Zeppelin Observatory during the MOSAiC project year. d Time series of the daily mean SAT and snow depth at the meteorological stations during the MOSAiC project year.

Full size image

The time series of the weekly mean NINP at the Zeppelin Observatory during the MOSAiC project year indicate the seasonality of NINP, especially in the higher temperature regime (Fig. 1c), and INPs active at temperatures above −15 °C were continuously detected from mid-April through to September. To clarify their seasonal trends, we divided the NINP dataset into four seasons and calculated the seasonal mean and median values (Fig. 2a). Note that in winter the seasonal mean NINP values were somewhat higher than the seasonal median NINP values in the higher temperature regime due to some high NINP cases found in December 2019 and January 2020. Nevertheless, the comparison of the seasonal values (Fig. 2b) illustrates that although the seasonal variations of the NINP values in the temperature regime of −30 °C were relatively small, those in the higher temperature regimes were obviously enhanced during warm periods. The seasonal trends of INPs measured during the MOSAiC project year are consistent with the results from our previous intensive measurement campaigns at the Zeppelin Observatory showing that the NINP active at temperatures above −25 °C in July 2016 were roughly one order of magnitude higher than those in March 201716.

Fig. 2: Seasonal trends of INPs at the Zeppelin Observatory during the MOSAiC project year.
figure 2

a Mean and median NINP data in autumn (September 2019 to November 2019), winter (December 2019 to February 2020), spring (March 2020 to May 2020), and summer (June 2020 to August 2020) calculated using the weekly data presented in Fig. 1c. The gray shadows represent the 5th and 95th percentiles. b Comparison of the mean and median NINP data.

Full size image

Relation between INPs and SAT

To investigate possible factors that might influence the variation of INPs active at relatively higher temperatures (i.e., highly active INPs), we utilized meteorological data measured at various locations in Svalbard. As shown in Fig. 1d, the SAT data measured at four sites in Svalbard exhibited a similar seasonal pattern during the MOSAiC project year. Overall, the SAT started to rise above 0 °C from mid-April through to September and recorded the highest value in late-July. It seems that the period when the SAT in Svalbard exceeded 0 °C is reasonably consistent with the timing of the detection of highly active INPs at the Zeppelin Observatory (Fig. 1c).

The snow depth data (Fig. 1d) indicate that the ground surface at the monitoring point in Ny-Ålesund was covered with snow from mid-October through to mid-May, whereas snow-free ground surfaces would already be distributed from mid-April in/around Longyearbyen located within the Inner Fjord Zone (the warmest area of Svalbard). The relationship between SAT and snow depth (Fig. 3a) indicates that the snow depth dramatically decreased when the SAT rose above 0 °C. This implies that conditions favorable for emissions of certain highly active INPs from snow-free ground surfaces might exist in Svalbard when the SAT exceeded 0 °C. On the other hand, the ground surfaces were almost always covered with varying amounts of snow (snow depth >0 cm) even at Longyearbyen when the SAT fell below 0 °C. In this regime, local/regional INP emissions would be very limited, because the ground surfaces in Svalbard should be almost entirely covered with snow, glaciers, and ice caps.

Fig. 3: Possible link of the near-surface conditions with the variation of INPs over Svalbard.
figure 3

a Relationship of the weekly mean snow depth with the weekly mean SAT at Ny-Ålesund and Longyearbyen. b Comparison of the weekly mean NINP at the Zeppelin Observatory with the weekly mean SAT at Ny-Ålesund for the period when local INP emissions from the ground surfaces were limited. c Same as (b), but for the period when local INP emissions from the ground surfaces were expected. The equations fitted to the data are summarized in Supplementary Table 1. The comparisons of the weekly mean NINP with the weekly mean SAT at Longyearbyen are presented in Supplementary Fig. 2 and Supplementary Table 2.

Full size image

Here we examined the correlation between highly active INP loadings and surface warming in Svalbard. Although the variations of the NINP values were not related to SAT under conditions when local/regional INP emissions in Svalbard would be limited (SAT < 0 °C, Fig. 3b and Supplementary Fig. 2a), there was a clear increase in the NINP values with increasing SAT under conditions when local/regional INP emissions would be expected (SAT > 0 °C, Fig. 3c and Supplementary Fig. 2b). Notably, highly active INPs tended to show more significant positive relationships with SAT (R2 > 0.5 for NINP active at temperatures above −20 °C) than INPs active at lower temperatures (R2 < 0.4 for NINP active at temperatures below −25 °C). This suggests that highly active INP loadings over Svalbard would be driven by surface warming and related changes in the near-surface conditions in Svalbard and its surrounding.

Relation between vegetation index and SAT

To evaluate the variation of the ground surface conditions in Svalbard, we analyzed the satellite-derived normalized difference vegetation index (NDVI) data (Fig. 4). It is expected that the NDVI values close to zero (0–0.1) indicate barren areas with less vegetation (e.g., glacial outwash plains) and the moderate values (0.2–0.4) represent vegetated areas with grasses, mosses, and lichens44,45,46. As illustrated in Fig. 4a, the positive NDVI values were detected from spring to autumn and peaked in summer. The higher values tended to be distributed in the Inner Fjord Zone (Longyearbyen and its surroundings) and Edgeøya located in southeast of the Svalbard archipelago, whereas the locally enhanced values were also detected in other areas including an area near Ny-Ålesund. Note that the satellite data did not fully cover the Svalbard region for the period from November 2019 to February 2020, but negative NDVI values would be expected to spread over almost all regions of Svalbard during this period due to the widespread presence of snow, glaciers, and ice caps. The percentages of areas (pixels) with positive NDVI values in the Svalbard archipelago showed the highest values in the summer, and those with NDVI above 0 and 0.2 accounted for about 35% and 15%, respectively (Fig. 4b), thus indicating that snow-free barren and vegetated areas were widely distributed in Svalbard. The areas with NDVI above 0.5 were rarely detected even in the mid-summer (~1%), probably because there are no forested areas in Svalbard. The percentages of the NDVI-positive areas were positively correlated with SAT when the SAT rose above 0 °C (Fig. 4c).

Fig. 4: Seasonal trends of normalized difference vegetation index (NDVI) values in Svalbard during the MOSAiC project year.
figure 4

a The distributions of monthly mean satellited-derived NDVI values. Note that the satellite data in this region were not available for the period from November 2019 to February 2020. b Time series of the percentages of NDVI-positive areas (pixels with positive NDVI values) in the Svalbard archipelago. c Comparison of the monthly mean vegetated areas in the Svalbard archipelago with the monthly mean SAT at Longyearbyen. The equations fitted to the data are summarized in Supplementary Table 3.

Full size image

Possible sources and processes related to INPs

The changes in the relation between NINP and SAT linked to conditions near the surfaces in Svalbard (Figs. 3 and 4 and Supplementary Fig. 2) suggest a strong SAT dependence of the sources and processes that influence the variation of INPs in this region. It is assumed that the aerosol populations in the Arctic lower troposphere tend to be influenced by long-range transport of air masses from low-mid latitudes in cold periods and characterized by particles emitted or formed within high latitudes in warm periods6,47. This general view is also supported by measurements of black carbon (BC) mass concentrations and carbon monoxide (CO) mixing ratios at the Zeppelin Observatory during the MOSAiC project year, which showed higher values in the winter and spring and lower values in the summer (Supplementary Fig. 3a, c). We interpret this result as evidence that the air masses transported from low-mid latitudes contained higher amounts of specific tracers emitted during incomplete combustion processes of fossil fuel and biomass, such as BC and CO, and that there were no significant sources of their emissions in the Svalbard region. It is increasingly recognized that the roles of BC aerosols themselves as INPs in mixed-phase clouds would be almost negligible48,49, but we consider that the information on the variation of BC and CO would be useful for discussing the influence of long-range transported air masses that might contain certain other aerosols serving as INPs. The aerosol size distribution data indicated that the concentrations of fine mode particles (0.3–1 μm in diameter) tended to decrease in the summer and autumn (Supplementary Fig. 4), which would be because the air masses in this period were less influenced by long-range transported aerosols linked to Arctic haze events50. On the other hand, the magnitude of the seasonal variation of the concentrations of coarse mode particles (1–10 μm in diameter) was not significant as compared to that of fine mode particles, suggesting the possibility that coarse mode particles originating from high-latitude sources might be present in the warm seasons. As a result, the fractions of coarse mode particles increased in the summer and autumn, whereas the total aerosol volume concentrations (0.3–10 μm in diameter) showed somewhat higher values in the winter and spring (Supplementary Fig. 3b, d). We note that anomalously high BC and CO events were occasionally detected in the summer (e.g., late-July to early-August), possibly due to intrusions of polluted air masses from northern Eurasia (Supplementary Fig. 5), but their impact on the seasonal variation of atmospheric constituents were relatively small (Supplementary Fig. 3c, d). For this reason, we believe that the occurrence of such temporal intrusion events would not be the primary cause of the enhancements of highly active INPs, which were continuously detected from mid-April through to September (Fig. 1c), and that more likely sources and processes would exist in Svalbard and/or its surroundings during the warmer periods.

To address particle types that might serve as highly active INPs over Svalbard, we utilized weekly aerosol samples collected in the same manner as those for measuring INPs and examined the morphology, size, and elemental composition of single aerosol particles (0.3–10 μm in diameter) using scanning electron microscopy (SEM) with energy-dispersive X-ray (EDX) analysis. We focused on the SEM/EDX analysis of the samples collected in March 2020 and July 2020 (Fig. 5a and Supplementary Figs. 6 and 7) to highlight the characteristics of the aerosol populations during the periods when local/regional INP emissions would be limited and expected, respectively. We note that the number fraction of soluble salts might shift to smaller size ranges, because such salts might partially lose their water content or break up into small fragments on filter substrates due to their recrystallization during/after sampling. For example, sea salt particles that were commonly detected in both periods might be somewhat overestimated in smaller size ranges and underestimated in larger size ranges. Nevertheless, we rule out such salts as a candidate for the major sources of highly active INPs that typically consist of insoluble materials. The observed number, surface area, and volume size distributions (Fig. 5b) indicate that there were no significant differences in the concentrations of ambient aerosol particles between these periods, especially for coarse mode particles, while the concentrations of fine mode particles in March 2020 were somewhat higher than those in July 2020. Although the number concentrations of coarse mode particles were much lower than those of fine mode particles, coarse mode particles would contribute largely to the INP population, since insoluble particles that have the potential to serve as INPs are widely distributed in the coarse mode and the ice-nucleating ability of particles is often expressed in terms of the ice nucleation active site densities per unit surface area or unit mass/volume10,11.

Fig. 5: Seasonal characteristics of ambient aerosol particles at the Zeppelin Observatory.
figure 5

a Number fractions of the detected particles averaged for the periods of 1 to 28 March 2020 (left) and 5 to 25 July 2020 (right). The averaged values are calculated from the weekly datasets obtained during these periods (Supplementary Figs. 6 and 7) and the total numbers of the analyzed particles are summarized in Supplementary Table 4. The data are plotted against area-equivalent diameter (DSEM) obtained by means of single-particle scanning electron microscopy (SEM). The bottom panels show the number fractions of P-poor and P-rich particles in carbonaceous particles presented in the top panels. b Number (dN/dlogDOPS), surface area (dS/dlogDOPS), and volume (dV/dlogDOPS) concentration distributions of ambient aerosol particles averaged for the same periods as (a). The data are plotted against optical diameter (DOPS) measured by an optical particle sizer. c Exemplary SEM images of carbonaceous particles that look like primary biological organic aerosols collected for the period from 5 to 25 July 2020.

Full size image

We posit that dust and organic materials released from nearby natural sources contributed to enhancing highly active INPs over Svalbard in the summer. Mineral dust particles were detected at the Zeppelin Observatory in both March 2020 and July 2020 (Fig. 5a and Supplementary Figs. 6 and 7); however, they would probably be influenced by different sources and processes. Considering snow/ice cover conditions in Svalbard and the Arctic (Fig. 1d and Supplementary Fig. 1) and model results showing that dust deposition in the Arctic from low-mid latitude sources peaks in winter and early spring51,52,53, we attribute the detection of mineral dust particles in March 2020 to their long-range transport from remote sources. It is expected that mineral dusts from desert areas in low-mid latitudes can typically act as dominant INP types at relatively low temperatures below about −20 °C10,11. By contrast, the origin of mineral dust particles detected in July 2020 would be linked directly to local/regional dust emissions in Svalbard and its surroundings. In this regard, we have confirmed that dusts originating from glacial outwash plains in Svalbard have a much higher ice-nucleating ability than typical mineral dusts from low-mid latitude sources in the temperature regime above −20 °C and can serve as INPs even at temperatures around −5 °C16. It is also important to note that the fractions of carbonaceous particles in the samples of July 2020 were found to be higher than those of March 2020 (Fig. 5a and Supplementary Figs. 6 and 7). Carbonaceous particles detected in July 2020 could be mainly attributed to natural organic aerosols of biological origin, such as primary biological organic aerosols (PBOAs) and biogenic secondary organic aerosols (BSOAs), released from nearby high-latitude sources16,33,54,55. For example, the morphological analysis demonstrated the existence of various types of PBOA-like particles, which look like microorganisms or vegetative detritus, in the samples of July 2020 (Fig. 5c and Supplementary Fig. 8) and they dominated the vast majority of carbonaceous particles in the coarse mode. As shown in Fig. 5a, we further confirmed that some carbonaceous particles in the coarse mode contained rich phosphorus (P), which has often been used as a marker of microorganisms56,57,58, and this is probably due to the existence of P-rich types of PBOAs. The lack of P-rich carbonaceous particles in the fine mode might be because other types of PBOAs (e.g., organic aerosols originating from P-poor microorganisms, vegetative detritus, or their associated products) and/or BSOAs released from nearby high-latitude sources accounted for the majority of the fine-mode carbonaceous aerosol populations. On the other hand, PBOA-like or P-rich carbonaceous particles were rarely detected in the samples of March 2020, probably because the major types of carbonaceous particles were replaced by BC, elemental carbon, or organic aerosols transported from remote anthropogenic sources in winter and early spring55.

Discussion

The results presented here provide evidence for the possible impact of warming-driven changes on the population of INPs active under conditions relevant for mixed-phase clouds over Svalbard. Based on continuous year-round measurements at the Zeppelin Observatory during the MOSAiC project period from September 2019 to September 2020, we find that the NINP, especially in the higher temperature regime, tended to increase with rising SAT when SAT rose above 0 °C and snow/ice-free ground surfaces appeared in the Svalbard region. We further confirm that dust and carbonaceous particles (most likely, organic materials) contributed largely to the population of ambient aerosols during the enhanced INP period, consistent with the results from an intensive measurement campaign in July 201616.

We rationalize the positive correlation between highly active INP loadings and SAT in Svalbard as the supply of natural aerosols such as dust and organic materials from snow/ice-free surfaces in high latitudes, especially in the Svalbard region. The influence of long-range transported aerosols from lower latitudes to the Arctic lower troposphere is rarely expected in the summertime Arctic6. On the other hand, the SAT rise in the Arctic drives the melting of snow, glaciers, and ice caps, leading to the widespread distribution of snow/ice-free areas (Figs. 1a and 4), which can increase the probability of natural aerosol emissions from Arctic terrestrial sources. For example, model simulations have shown that dusts released from snow/ice-free terrestrial areas of the Arctic, such as glacial outwash plains, are activated in warm periods and contribute largely to dust loadings in the Arctic lower troposphere16,51,52,53. In addition, it is likely that the SAT rise is an important factor affecting vegetation coverage and growth in this region. Time series of the satellite-derived NDVI data have indicated that the strong positive correlation between the NDVI and SAT in Svalbard (Fig. 4). Furthermore, field measurements of fluorescent aerosols (a proxy for PBOAs) at the Zeppelin Observatory have shown the enhancement of fluorescent aerosol number concentrations during the high NDVI and SAT period33, implying the possible link of certain organic aerosols with vegetation dynamics and microbes in terrestrial ecosystems of the Svalbard region.

Our results appear to indicate that certain types of dust and organic aerosols detected during the enhanced INP period at the Zeppelin Observatory have very high ice-nucleating ability and serve as efficient INPs in the higher temperature regime ranging from about −20 °C to −5 °C. For example, given that dusts derived from glacial outwash sediments in Svalbard (Svalbard dusts) can act as INPs even at temperatures around −5 °C16, dusts released from local/regional terrestrial sources might contribute largely to the population of atmospheric INPs during the enhanced INP period over Svalbard. We note that Svalbard dusts contain small amounts of organic materials. In this regard, we have confirmed that the very high ice-nucleating ability of Svalbard dusts is mostly attributed to the presence of such organic components rather than mineral components and that the ice-nucleating ability of the mineral components themselves are not very different form that of typical mineral dusts from desert areas in low-mid latitudes16. Although it is technically difficult for the SEM/EDX analysis to quantify the presence of organic matter internally mixed with mineral dust particles, we suppose that dust emissions from nearby terrestrial sources might consist of both mineral and organic components and account for large amounts of mineral particles and certain amounts of carbonaceous particles detected during the enhanced INP period at the Zeppelin Observatory. In addition, the release of certain organic aerosols from tundra vegetation (e.g., grasses, mosses, lichens) might also contribute to the detected carbonaceous particles. Recent work showed that the periods of fluorescent aerosol loadings were almost consistent with those of highly active INP loadings at the Zeppelin Observatory33, suggesting the possible contribution of PBOAs to the population of atmospheric INPs active in the higher temperature regime. It should be noted here that our SEM images offer direct evidence for the existence of a wide variety of PBOA types (Fig. 5c). Given that only specific types of biological INPs are expected to indicate a warmer onset freezing temperature10,11,59, we assume that the ice-nucleating ability of the detected PBOA-like particles varies among the particle types. To our knowledge, the ice-nucleating ability of PBOAs derived from Arctic ecosystems has not yet been well explored. Further studies are therefore necessary to identify what types of organic materials can indeed serve as highly active INPs in the Arctic atmosphere.

While our research in Svalbard sheds light on the existence of organic materials that likely originated from Arctic terrestrial ecosystems during the enhanced INP period, many previous studies so far have suggested the possible importance of marine organic aerosols emitted from oceanic ecosystems as a significant source of Arctic INPs22,23,24,25,26,27. Even though the Svalbard region is surrounded by open water and the aerosol population seems to be influenced by air masses traveled over the surrounding oceans throughout the year (Supplementary Fig. 1), our results indicate that the highly active INP loadings linked to surface warming occurred only when snow/ice-free ground surfaces were found in the Svalbard region, suggesting the importance of local/regional terrestrial sources. A recent study reported the variation of fluorescent aerosol number concentrations linked to ambient air temperatures over the Bering and Chukchi Seas in summer and explained that the fluorescent aerosol loadings could be attributable to PBOA emissions from terrestrial sources rather than oceanic sources, based on the backward trajectory analysis60. These results imply that oceanic emissions might not be a primary cause of the INP and/or PBOA loadings linked to surface warming in the summertime Arctic, at least in locations close to terrestrial sources. However, the possibility remains that marine organic aerosols serving as INPs might be partly influenced by episodic oceanographic events (e.g., phytoplankton blooms25, high waves13) and that oceanic sources might play an important role for INP loadings in particular locations and seasons.

Since the Svalbard region has been warming anomalously faster than the Arctic average over the recent decades37,38, this region may be regarded as a natural laboratory for studying the climatic sensitivity of aerosols, clouds, and their interactions to Arctic warming. The link we see between the NINP and SAT suggests that the timing and magnitude of the release of highly active INPs from snow/ice-free areas in the Arctic could be mediated by the interactions between potential aerosol sources and surface warming. Although our analysis was based on a year-round record during the MOSAiC project period, evidence is mounting for a robust summer temperature dependence of accelerated ice loss, permitting the appearance of newly exposed landscapes across Svalbard over the recent decades61. The multi-decadal summer temperature rise also leads to increased vegetation cover, biomass, and productivity in Svalbard46. These facts lead to speculation that the probability of the emissions of natural aerosols serving as highly active INPs from Arctic terrestrial sources has increased, while the magnitude of future emissions remains uncertain. It is also important to note that SATs in Svalbard occasionally soared to ~0 °C and then snow/ice-free surfaces temporally appeared even in mid-winter during the MOSAiC project year, at least in/around Longyearbyen located within the inner fjord zone of Svalbard (Fig. 1d). Given the fact that the warming trends in winter are more serious than those in summer and exceed 2 °C per decade in the Svalbard region37,62, we suppose that the appearance of snow/ice-free areas in the wintertime Arctic will become increasingly common in the next decades, which will further increase the probability of natural aerosol emissions from terrestrial sources and alter the annual cycle of Arctic INPs.

We further consider that our findings also have important implications for the projected cloud-phase feedback in the Arctic atmosphere. The current view holds that ice crystal formation in Arctic mixed-phase clouds will be suppressed as the isotherms shift upwards in a warmer atmosphere and that the clouds are in transition towards a state with more liquid water and less ice3,4,5. Mixed-phase clouds over the Svalbard region occur frequently at altitudes from about 0.5 to 6 km in summer and 0.5 to 3 km in other seasons36. In this regard, our measurements on Mt. Zeppelin revealed that relatively high number concentrations of highly active INPs found in warm periods could extend at least to the lower part of mixed-phase clouds (~0.5 km). Recent model simulations have also indicated that high-latitude dusts dominate the atmospheric dust loading from the ground level to the 700 hPa pressure level (~2.5 km) in the summertime Arctic52,53, suggesting that INPs released from high-latitude terrestrial sources have the potential to be transported to sufficiently high altitudes where they could affect the microphysics of Arctic mixed-phase clouds. However, the impact of Arctic INPs on current and future aerosol-cloud-climate interactions in the projected cloud-phase transition remains highly uncertain. Our results suggest the possibility that the supply of highly active INPs from high-latitude terrestrial sources will increase in response to the projected surface warming, and thus this effect needs to be considered in climate models to improve our understanding of the phase transition scenario of Arctic mixed-phase clouds.

Methods

Sampling and measurements at the Zeppelin Observatory

Continuous year-round measurements of atmospheric constituents have been performed at the Zeppelin Observatory (78.9°N, 11.9°E, 474 m above mean sea level) located on Mt. Zeppelin near Ny-Ålesund, Svalbard (Fig. 1a). Aerosol properties are sampled in conditions of low relative humidity (<30%) due to the strong temperature gradient between ambient and indoor temperatures, although active drying is not used for the inlet systems at the Zeppelin Observatory33,63.

Weekly sampling of ambient aerosols (Sunday (00:00 UTC) to Saturday (23:59 UTC)) was performed continuously and sequentially using a 10-line Global Sampler (GS-10N, Tokyo Dylec Corp.) connected with a PM10 common inlet. Each sample has been collected on a precleaned Whatman Nuclepore track-etched polycarbonate membrane filter (47 mm in diameter and 0.2 μm in pore size) supported by a filter cassette screen (part no 59-005147-0010, Thermo Fisher Scientific) and mounted in a NILU inline filter holder system at a typical flow rate of 3 L min−1 at standard temperature and pressure (STP, 0 °C and 1 atm) conditions. After collecting ambient aerosol samples, the filters were placed into 50 mL sterile centrifuge tubes, transported to the laboratory of the National Institute of Polar Research in Japan (typically within 4 months), and stored in a freezer at −20 °C until analysis. In this study, we focused on the samples collected during the period from 1 September 2019 to 3 October 2020, and utilized the half of the samples (i.e., half-cut exposed filters) for measuring the NINP and the remaining samples for other objectives (e.g., SEM/EDX analysis). In addition, ambient aerosol number, surface area, and volume concentrations (0.3 to 10 μm in optical diameter) and BC mass concentrations at STP were monitored using an optical particle sizer (OPS Model 3330, TSI Inc.) and a continuous soot monitoring system64,65 connected with the same PM10 common inlet, respectively. We also referred to CO data that have been monitored in the framework of the Integrated Carbon Observation System (ICOS) and are available at the EBAS database (https://ebas-data.nilu.no/).

Analysis of INP number concentrations

To quantify the NINP at STP conditions, the ambient aerosol samples collected at the Zeppelin Observatory were analyzed using the Cryogenic Refrigerator Applied to Freezing Test (CRAFT) system39. The basic procedures for measuring the number concentrations of atmospheric INPs using the CRAFT system are essentially the same as those reported previously16,40. After preparing a suspension of the collected particles in Milli-Q purified water (18.2 MΩ cm), the particle-containing water droplets (typically, 49 or 98 droplets) with a volume of 5 μL were placed on an aluminum plate coated with a thin layer of Vaseline (petroleum jelly). Then, the temperature was lowered at a cooling rate of 1 °C min−1 in the CRAFT system, and the number fractions of the droplets frozen and unfrozen were counted every 0.5 °C. If all droplets made of the initial suspension were frozen before reaching −30 °C, then the initial suspension was diluted and the additional freezing experiments with the diluted droplets were conducted. Since few pure water droplets freeze at temperatures above −30 °C in the CRAFT system39,40, we consider that artifacts related to freezing of INP-free droplets would have negligibly small effects on these freezing experiments. Based on the results of the freezing experiments, the NINP at STP conditions were determined over a temperature range down to −30 °C. In this study, the minimum limit of detection was set to ~0.001 L−1 for the samples collected during the period from 1 September 2019 to 11 January 2020, and ~0.0005 L−1 for those collected during the period from 12 January to 2 October 2020, respectively.

Electron microscopy analysis

The size, morphology, and elemental compositions of individual ambient aerosol particles (0.3 to 10 μm in area-equivalent diameter) collected on a precleaned Whatman Nuclepore track-etched polycarbonate membrane filter (47 mm in diameter and 0.2 μm in pore size) were automatically analyzed using SEM (SU 3500, Hitachi High-Technologies Co.) and EDX (X-Max50, Horiba Ltd.) system16. To detect a sufficient number of particles over this size range, the automated SEM/EDX analysis was performed twice per filter with different settings (e.g., magnification) and more than 350 particles were analyzed in the fine (0.3 to 1 μm) and coarse (1 to 10 μm) modes, respectively (Supplementary Table 4). To classify the particle types based on the elemental composition analysis, the weight percentages of elements were measured, and we took the value over five standard deviations (σ) for each element obtained from the EDX spectra as an appropriate threshold above which a particle would contain the element as a major component. The analyzed particles were then classified into one of the following five major categories: (1) sea salt particles (Na: >5σ, Si: ≤5σ); (2) mineral-sea salt mixture particles (Na: >5σ, Si: >5σ); (3) mineral particles (Na: ≤5σ, Si: >5σ); (4) carbonaceous particles that indicated a stronger C signal than the polycarbonate filter substrate (for example, see Supplementary Fig. 8); or (5) others. These categories are almost the same as those used in previous work16, while metallic particles that were rich in metal elements but contain little or no Si were categorized as others, because the category of mineral/metallic particles was not adopted in this study. Carbonaceous particles were further grouped in P-poor (P: ≤5σ) and P-rich (P: >5σ) types. We also manually checked the SEM images and EDX spectra of the classified particles and confirmed that the above classification would be appropriate.

Dataset for near-surface conditions

The meteorological data at Ny-Ålesund, Verlegenhuken, Longyearbyen (Svalbard Lufthavn), and Hornsund in Svalbard were analyzed using the hourly dataset provided by the Norwegian Centre for Climate Services. The seasonal snow/ice percent coverage data over the northern high-latitude region were analyzed using the monthly mean values derived from the Clouds and the Earth’s Radiant Energy System (CERES) Single Scanner Footprint (SSF) 1 deg dataset. The CERES snow/ice coverage data are based on the National Snow and Ice Data Center (NSIDC) Near-Real-Time SSM/I-SSMIS EASE-Grid Daily Global Ice Concentration and Snow Extent product overlaid on the permanent snow and ice map from the International Geosphere–Biosphere Programme (IGBP). The vegetation index data in Svalbard were analyzed using the monthly mean NDVI values derived from the Terra Moderate Resolution Imaging Spectroradiometer (MODIS) dataset at a spatial resolution of 0.1° × 0.1°.

Backward trajectory analysis of air mass origin

The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model66 was used for calculating 5-day backward trajectories for air masses arriving at a height of 500 m above mean sea level over the Zeppelin Observatory. For this study, the trajectories that ended over the observatory were calculated at 3-h intervals and then the trajectory frequency in each 1° × 1° coordinate grid cell (=100 × number of endpoints per grid cell/total number of endpoints) was obtained.