The FAAM modified BAe 146-301 Atmospheric Research Aircraft (ARA) is fitted with a suite of aerosol, cloud microphysics, and remote sensing instrumentation, detailed by , , and amongst others. Measurements from these instruments are used here to investigate microphysical properties of clouds in the context of their environment. In this article, all data are expressed as ambient measurements, and number and mass concentrations are not corrected to standard temperature and pressure conditions.

Derived cloud top temperature from MODIS satellite retrievals and AVHRR visible satellite imagery are used to illustrate cloud spatial structure and distribution. Additionally, sea ice fraction from NASA's National Snow and Ice Data Center (NSIDC), derived from passive microwave brightness temperatures , and the approximate ice fraction from the Met Office Unified Model (MetUM) are used to contextualise the in situ observations.

Cloud layers were observed with the lidar during section 1 of flight B762 (Fig. ). A continuous layer was observed where cloud top descended from approximately 1900 to 1100 m with increasing latitude. Evidence of a second, lower-altitude cloud layer can be seen at high latitudes (500 m at 76.5∘ N) through breaks in the continuous layer. The 500 m cloud was not observed at lower latitudes along the flight path sampled (between approximately 73 and 73.5∘ N), as indicated by the surface echo measured during breaks in the cloud above. An intermittent, high-altitude cirrus layer (with optical depth ∼ 0.5) was seen at various levels from 3000 to 8000 m at higher latitudes on the approach to Longyearbyen. However, these data cannot be used to indicate the spatial extent of these structures as measurements were only made along the flight path.

The structure of the lower troposphere was sampled extensively in section 1 by the 11 dropsondes marked in Fig. . A summary of key information from each dropsonde is listed in Table . These measurements were collected approximately 2 h before and to the west of the in situ cloud observations of section 2; however, the dropsonde and lidar measurements provide a good indication of the structure of the atmosphere during this study.

Figure  indicates that the boundary layer structure varied with latitude. Figure b and d show vertical profiles of Θ and RH data obtained from dropsondes nos. 5 and 11, collected at latitudes comparable to the in situ aircraft runs of section 2. A double potential temperature inversion can be seen over the sea ice (no. 11), whereas the temperature profile over the ocean indicates that the boundary layer was well-mixed and coupled to the surface (no. 5). The double temperature inversion over the sea ice is mirrored by twin RH peaks measured at the corresponding altitudes (∼ 84 %, 500 and 1100 m, no. 11, Fig. d), suggesting the presence of two cloud layers <1500 m below a dry region above the boundary layer (∼ 1500–2000 m). At lower latitudes over the ocean (74 ∘N), a single, moist layer was observed between approximately 300 and 1200 m. A single temperature inversion was measured by the dropsondes at approximately 1300 m at this latitude. These cloud layers measured by the dropsondes are in accordance with the lidar data (Fig. ). However, these data are affected by a dry bias of approximately 15–20 %, where in-cloud RHs of 84 % in dropsonde no. 11 and 88 % in dropsonde no. 5 were measured. Dry biases have been observed in dropsonde data in previous studies e.g., and have been attributed to a slow response time at low temperatures . Sondes are particularly prone to these response issues when descending from a dry region into a cloudy region , which was the case for both of our considered dropsondes.

Near-surface temperatures sampled by the dropsondes were approximately 13 ∘C colder to the north over the sea ice (-16.4 ∘C at 76.8∘ N) than over the ocean to the south (-3.4 ∘C at 72.2∘ N, see Table ). Between the latitudes of the in situ measurements (dropsondes nos. 5 and 11), the difference in near surface temperature is approximately 6 ∘C.

Satellite imagery was examined to provide lateral context for the dropsonde and lidar measurements. Figure  displays AVHRR visible satellite imagery in panel a and the derived cloud top temperature from MODIS satellite data in panel b. The flight track of section 2 is overlaid to indicate the regions sampled with the aircraft. The high-altitude cirrus layer indicated by Fig.  can be seen with both of these data. This cirrus cloud was to the north-west of the main science region investigated, closer to Spitsbergen. At the locations sampled during the aircraft runs (Fig. b), there is no clear indication of a higher cirrus cloud layer, and the cloud top temperature is approximately -18 and -23 ∘C over the sea ice and ocean regions respectively.

A weak low pressure system was present to the east during the sampling period; however, conditions were dominated by high pressure to the west, causing a northerly flow of air from over the sea ice. Cold-air outbreak conditions, with wind speeds of ∼ 10 ms-1 measured within the boundary layer, were maintained for the duration of the science period. Aircraft measurements were made through a band of cloud orientated in the N–S direction, influenced by this northerly flow. Back trajectory analyses (shown in the Supplement, Fig. S1) also show that the sampled air came from the north, having travelled from northern Canada and/or Greenland, depending on the period of interest.

Aerosol number concentrations measured by the various probes on board the aircraft are reported in Table . CPC particle number concentrations are greater at high altitudes (> 800 m) over both surfaces, as is the total concentration recorded by the PCASP over sea ice. The gradient in CPC concentration is greatest over the ocean, with a high-altitude measurement of over 4 times that measured at low altitude. These number concentrations are not observed to the same extent in the PCASP data.

Over the ocean, the number concentrations measured by the PCASP (both total and > 0.5 µm) and CAS-DPOL are approximately constant with altitude. Concentrations of large aerosol (which may act as ice-nucleating particles, DP > 0.5 µm) are approximately uniform with altitude over the ocean. Over the sea ice, a greater loading of coarse-mode aerosol is measured nearer the surface and this consistency with altitude is not observed.

In general, the number concentrations measured over the ocean are lower than over the sea ice. Figure  displays the size distributions from the PCASP and CAS-DPOL split into high- and low-altitude data. Small particles measured by the PCASP (0.1 µm < DP < 0.5 µm) reach a greater number concentration over the sea ice than over the ocean. Low-altitude CAS-DPOL concentrations over the sea ice are heightened with comparison to the high-altitude data. It is possible that swollen aerosol particles associated with the haze layer were being measured, enhancing the number concentration. Such particles may not be removed from these data using the CDP LWC ≤ 0.01 g m-3 out-of-cloud threshold applied here. Over the ocean, data from both instruments are more comparable; however, the low-altitude PCASP data show a greater loading across most sizes.

Non-refractory sub-micron aerosol composition measured by the AMS is shown in Fig. . Technical issues prevented continuous measurement over the ocean, with problems occurring during run 7. The measured nitrate mass loading remains low and consistent throughout. The sulfate loading is variable with altitude, especially over the transition region between ice and ocean. Higher mass loadings are measured at higher altitudes (∼ 1 µg m-3 at 1400 m). Increasing during the last SLR over the sea ice (run 5), the signal becomes highly variable over the sawtooth profile. Such variability is also observed in the organic and ammonium traces. PCASP particle number and SP2 black carbon (BC) mass observations follow the same trends throughout the science period, both mirroring the same sinusoidal pattern over the MIZ. Both signals are variable, with increases observed at high altitudes, but no distinct differences are observed between sea ice and ocean measurements.

Turbulent kinetic energy (TKE) and sensible and latent heat fluxes measured along the flight path are shown in Fig. . Approximate MetUM ice fraction is shown in the left-hand and middle columns. Over the sea ice, the sensible and latent heat fluxes and the TKE remain relatively constant at about 0 W m-2 and 0.5 m2 s-2 respectively. More variability is observed in these three parameters over the ocean. Sensible heat fluxes range from -20 to 0 W m-2 at low altitude over the sea ice, whilst substantially greater values of > 30 W m-2 are observed over the ocean, with > 100 W m-2 measured in some instances. A similar difference is observed with the latent heat fluxes with variable measurements of approximately > 50 W m-2 over the ocean, contrasting observations of ∼ 0 W m-2 over the sea ice. Low-altitude (∼ 350 m) TKE increases from approximately 0–0.5 to 1.5 m2 s-2 over the transition. TKE, sensible heat fluxes, and latent heat fluxes all increase and become more variable over the MIZ and ocean compared to the sea ice, with the greatest values typically observed at low altitude over the ocean.

The turbulence and AIMMS probes recorded vertical velocity throughout the science period. The turbulence probe suffered some icing effects during runs 7 and 8, whilst the AIMMS probe collected no data for run 8 due to a technical issue. Averaged PDFs from over the sea ice and ocean are shown in Fig. . The turbulence probe and AIMMS PDFs compare well. The sea ice PDF displays little variation, with the majority of measurements lying close to the mean value. Maxima and minima of the distribution are approximately ±1 m s-1. In comparison, the ocean PDFs are significantly broader, with more variability from the mean observed. Maxima and minima of the ocean PDF extend to ±2 m s-1 and the distribution is skewed toward updraughts.

The averaged data over sea ice point toward a low-altitude cloud with a low liquid water content (Fig. ). Ice crystal number concentrations are spatially variable within the cloud (Fig. ) yet they are consistent, suggesting only primary ice nucleation was active. The temperature within the cloud was between -18 and -20 ∘C (Fig. b), far below the range required for secondary ice production e.g..

Data from runs 2, 3, and 4 depict the typical structure of a single-layer Arctic mixed-phase cloud e.g.: a liquid layer at cloud top, with ice formation and aggregation below. Such processes are inferred from the relative quantities of ice crystals measured by the 2DS and CIP100 instruments (Fig. ), as the latter can measure much larger ice crystals than the former. In Fig. c, the ice mode is smaller at cloud top (Fig. c3, run 4) than cloud base (Fig. c1, run 2), indicating that ice nucleation may be occurring towards cloud top; however, this cannot be verified with these data and vertical mixing likely has an influence.

Run 5 was planned to be an above-cloud aerosol run; however, the 2DS and CIP100 instruments detect notable ice number concentrations to the western end of the run (Fig. ). This suggests that precipitation from above was being sampled, which could possibly be from the high-altitude cirrus layer observed closer to Spitsbergen (Figs.  and ). Additionally, RH data from the high-latitude dropsondes (Fig. c) indicate the possibility of a higher cloud layer (∼ 2000–3000 m) in this region. CPI ice crystal images during run 5 (Fig. ) also point towards the sampling of another cloud, as the imaged bullet rosettes typically form at higher, colder altitudes and aggregate as they descend. This ice precipitation was observed predominantly to the west of and was separate from the main cloud layer sampled over the sea ice, and thus was excluded from comparison with the ocean cloud layer.

Observed aerosol concentrations varied substantially with altitude over the sea ice. Results show elevated sulfate, PCASP, CPC, and BC measurements during run 5 (Table  and Fig. ), the latter of which is consistent with the Asian BC plumes identified during the ACCACIA campaign . These plumes contained an average BC mass loading of ∼ 27 ng s m-3 during flight B762, consistent with the low-altitude measurements here (Fig. ). Similarly, this mass loading is also equivalent to the annual median BC concentration measured at Zeppelin, Svalbard over the period 1998–2007 . PCASP data do not increase as much as the CPC data at high altitudes, suggesting either a pollution layer characterised by small particle sizes (3 nm < DP < 0.1 µm) or new particle formation at these heights.

More large particles are measured close to the sea ice surface, as shown by the > 0.5 µm PCASP and CAS-DPOL measurements (Table ). This suggests contributions of primary particle emissions from the surface. A large fraction of these low-altitude particles were sea salt , which could have been transported from the nearby ocean or lofted into the atmosphere by near-surface winds over polynyas or leads in the sea ice. Additionally, frost flowers could be a surface source of modified sea-salt aerosol ; however, their characteristic signature would not be detectable by the analysis presented by .

Observed ice crystal number concentrations vary little over the sea ice, with slight increases toward cloud base attributable to aggregation and precipitation out of the cloud. Using the D10 parameterisation, predicted INP concentrations were 0.9, 1.90, and 3.31 L-1 for aerosol particle number concentrations measured by the aircraft filters , PCASP, and CAS-DPOL respectively (listed in Table ). When using D10, the filter data produce the best agreement with the mean 2DS ice concentration in the mixed cloud layer. D10 predictions using the probe data overestimate with comparison to the mean observed in the cloud; an overestimation which can be explained by incorrectly assuming that all predicted INPs nucleate to form ice crystals. The large fraction of coarse-mode sea salt particles identified over the sea ice is unlikely to contribute to the INP population as these are inefficient INPs. Despite this, given the uncertainties in the parameterisation itself, these predictions do well to replicate the ice observed in the cloud. T13 was derived from forest ecosystem data and, therefore, represents an environment with plentiful biological aerosol particles that may act as INPs. T13 predictions are highly variable with these data, giving 0.07 L-1 with the filter data and 19.7 L-1 with the CAS-DPOL. This variability is not surprising, as small increases in an INP-active aerosol population – such as that used to derive T13 – would cause significant changes in the ice crystal number concentrations in the clouds. Despite this, the T13 parameterisation agrees well with the observed ice number concentrations when PCASP data are used for its evaluation (1.10 L-1).

For comparison, parameterisations based on mineral dust data were considered. N12 and D15 were evaluated with the dust size distribution and number concentration of dust particles > 0.5 µm, derived from the aircraft filters (Table ). As discussed by , this dust loading is likely under-represented as a result of the analysis technique and collection efficiency issues. Evaluations of N12 with measured, double, and triple dust loadings all compare well with the run 2 ice data, illustrating a lack of sensitivity to this input. However, N12 agrees best with observations when the measured dust concentrations from the filters are used. Agreement with observations is poor when applying the D15 parameterisation. This parameterisation was developed to simulate the high nucleating efficiencies of mineral dusts at temperatures < -20 ∘C; therefore, our data are at the upper limit of its applicability. Given the good agreement of N12 with our observations, at temperatures well within the range represented by the parameterisation, it can be speculated that the ice number concentrations observed over the sea ice may be explained by the dust loadings present.

Over the sea ice, the liquid water content is low (∼ 0.03 g m-3, Fig. a) and the mean droplet radius is small (∼ 4–5 µm, Fig. ), with mean droplet number concentrations of 110 ± 36 cm-3. It is possible that the cloud layer interacted with aerosol from above via entrainment processes, as suggested by the presence of numerous small cloud droplets; with more CCN available, more cloud droplets can form . Mean droplet number concentrations from the CDP (Fig. ) are consistent with both the high- and low-altitude PCASP number concentrations measured (Table ), further suggesting that the ambient aerosol mixed with the cloud layer from above and below.

The vertical velocity PDFs (Fig. ) and TKE (Fig. ) suggest that the boundary layer over the sea ice is stable with relatively little mixing compared to downstream. Little variability and low values are observed in the measured TKE, sensible heat flux, and latent heat flux (Fig. ). The potential temperature profiles from the dropsondes (Fig. a, b) also indicate weakly stable stratification with very strong temperature inversions, emphasising that boundary layer mixing was inhibited over the sea ice. A lack of substantial vertical air motions may explain the low LWC; by the WBF mechanism, the ice crystals act as a sink for vapour in the cold temperatures observed (∼ -20 ∘C). This causes the suppression of the liquid phase via a suppressed supersaturation and cloud droplets remain small.

When transitioning from sea ice to ocean, both cloud depth and cloud base height increase (Fig. ). The most significant change in cloud microphysics over the transition is in the liquid phase, where the liquid water content and mean droplet size increase from 0.1 to 0.4 g m-3 and 5 to 10 µm respectively. Mean Ndrop peaks at 145 ± 54 cm-3 during profile 7, and falls to 120 ± 33 cm-3 during profile 8. This decrease in Ndrop is accompanied by an increase in Reff towards cloud top, from 8 to 10 µm, indicating enhanced collision coalescence within the deeper, ocean-based cloud layer (Fig. ).

A cloud with high Ndrop over the sea ice would reflect incident solar radiation efficiently ; however, as the sea ice surface itself is highly reflective, the net impact of the cloud on the sea ice radiative interactions is difficult to interpret. As the sea ice gives way to the MIZ, the increased droplet number concentration, mean droplet effective radius, liquid water content, and cloud depth suggest that the cloud optical thickness and albedo also increase . With increased optical and geometrical thickness, upwelling LW radiation from the surface can be trapped and incident solar radiation can be scattered more efficiently. Radiative predictions are hindered by the weak solar heating experienced during the Arctic spring; however, it is likely that these properties would cause a net warming at the surface due to the increasing LW influence from the ocean, and could potentially enhance the melting processes of the nearby sea ice . Over the ocean, droplet number concentrations decrease throughout the cloud, whilst the mean droplet effective radii increase. With fewer, larger, cloud droplets, the cloud over the ocean may not be as efficient as scattering solar radiation as the sea ice cloud, if their microphysical properties are compared irrespective of their environments. However, the thick cloud over the ocean would act to significantly increase the net albedo of the ocean regime, whilst the net impact of the sea ice cloud is unclear.

The ambient temperatures experienced within the cloud layers (Fig. ) remain colder than required for secondary ice production and warmer than the homogeneous freezing threshold throughout. The observed 2DS ice crystal concentrations are consistently low – approximately 0.5–1.5 L-1 – throughout the transition, again indicating that only primary ice nucleation was active. Higher ice concentrations are observed at high altitudes (up to 2 L-1 at ∼ 1350 m) over the sea ice, suggesting that some precipitation was again measured from a higher cloud layer. This agrees with the conclusion from run 5, as both this profile and run 5 were conducted close together (1400 m) and to the west (∼ 26.5–27 ∘E) of the main science region. Though some seeding from above is probable in both run and profile 5, the ice concentrations measured during the SLRs over sea ice and ocean – and the remaining profiles over the MIZ – suggest that this is not the case for the majority of these data.

The sulfate, organic, ammonium, and BC loadings vary almost sinusoidally with altitude over the transition region, with maxima reached at the peaks of the sawtooth profile (Fig. ). These peaks occur above the cloud layer, as indicated in Fig. . These species commonly act as CCN in the atmosphere and could be acting to increase the cloud droplet number concentration across the MIZ. There is no evidence that the organic species influences the ice phase, as the former varies significantly whereas the latter remains approximately constant. The observed increases in mass loading are small, as are the increases in number concentration measured by the PCASP (Fig. ); therefore it is unlikely that they are the cause of the significant microphysical changes observed.

Measured surface heat fluxes – both sensible and latent – become more variable and increase over the transition to ocean, with greatest values measured at low altitudes (< 1000 m). The approximate MetUM ice fraction indicated in Fig.  mirrors the transition of this variability, with little deviation from 0 W m-2 measured when sea ice is present and significant variability developing over the transition to open water. Similar changes are observed in the TKE data, where increased turbulence is induced as the air mass moves over the broken sea ice and comparatively warm ocean.

Between sea ice and ocean, the most prominent microphysical difference is in the liquid phase. The observed cloud is deeper, with a mean LWC and droplet number concentration of ∼ 0.3 g m-3 and 63 ± 30 cm-3 respectively (Fig. ).

Observed ice crystal number concentrations within the oceanic cloud are similar to those measured over the sea ice (0.5–1 L-1) and the D10 INP predictions are in reasonable agreement (∼ 1–2 L-1, Table ). T13 predictions are more variable with the input data, resulting in 2.66 and 0.34 L-1 using PCASP and CAS-DPOL data respectively. As discussed previously, T13 is particularly sensitive to the number concentration of aerosol particles used as input, with little difference in Naerosol causing significantly different predictions (Table ). The CAS-DPOL prediction is in better agreement with the ice number concentrations observed in this case than the PCASP prediction. The ice phase measurements are consistent with altitude; however, there is an increase at the base of the mixed cloud layer (∼ 700–900 m). This increase could be due to sedimentation of ice crystals or low sampling statistics at these altitudes. However, the presence of dendritic ice crystals (Fig. S2) combined with an ice concentration of ∼ 5 L-1, a mean temperature of approximately -16 ∘C, and an enhanced ice crystal mode in the size distributed data (Fig. c4) suggests that ice-ice collisions may have taken place at cloud base . Dendritic crystals are susceptible to break up, and have been shown to fragment due to air velocity alone . The ambient temperatures were too low to suggest secondary ice formation via the Hallett–Mossop pathway , but it is possible that some crystal fragmentation due to collisions enhanced the observed primary ice concentration at these altitudes.

Probe icing was an issue during run 7 and this effect can be seen in the sharp increase in the CIP100 and corresponding decrease in the 2DS ice number concentrations in Fig. c2. Run 8 was intended to be a low-altitude, below-cloud run; however, precipitation particles were again observed. Contrasting run 5, this precipitation is related to the observed ocean cloud layer sampled above during runs 6 and 7: images from the CIP100 (Fig. ) indicate that these particles are large and rimed, inferring an interaction with liquid droplets within the mixed layer above.

Aerosol data were not available for a significant fraction of the ocean component of the flight. Despite this, it can still be seen that the mass loadings of all AMS-measured species are low in this region (Fig. ). The measured BC loading is more variable over the ocean than the sea ice or MIZ, varying from approximately 10 to 100 ng m-3. This variability is mirrored by the particle number concentration measured by the PCASP. The heightened BC loadings are consistent with the monthly average reported by for Feb/Mar at Zeppelin station, Svalbard (∼ 80 ng m-3). There is also consistency between the high- and low-altitude measurements from the CAS-DPOL and PCASP (> 0.5 µm) over the ocean, suggesting a constant vertical profile of large aerosol in this region. Aerosol number concentrations were found to be lower in general over the ocean than over the sea ice. No aircraft filters were exposed over the ocean; therefore, the composition of the coarse-mode aerosol could not be established.

From Fig. , the atmosphere is notably warmer over the ocean than over the sea ice. There is a steady increase in the boundary layer potential temperature measured by the dropsondes from north to south. The changes in the Θ profiles are most prominent in the boundary layer, with less variability observed > 1500 m. The near-surface temperature inferred by the dropsondes increases by 13 ∘C between the most northerly and southerly latitudes sampled. Over the transition from sea ice to ocean, the inferred temperature difference at the surface was approximately 6 ∘C. The ocean surface was, therefore, significantly warmer than the sea ice, and this sharp temperature gradient affected the both structure of the boundary layer and any clouds that formed within it.

The broader vertical wind PDFs over the ocean (Fig. ) suggest increased turbulence levels and mixing within the boundary layer. This is in agreement with Figs.  and : the surface temperature increased with transition from the frozen sea ice to the warm ocean, thus increasing sensible heat and latent heat fluxes from the surface. At low altitude over the ocean, both of these measures routinely exceeded 20 W m-2, whilst they remained consistent at approximately 0 W m-2 over the sea ice. Measured TKE was significantly higher at low altitude over the ocean than over the sea ice, suggesting a much more turbulent boundary layer over the open water. The contrast between the observed cloud microphysics over sea ice and ocean is most likely due to these increased fluxes and induced turbulent motions. The warmer, more turbulent boundary layer promotes the efficient collision-coalescence (and subsequent growth via sustained supersaturations) of cloud droplets, promoting a higher mean droplet effective radius and lower number concentration (Fig. ). A consistent source of heat and moisture to the BL, enhanced turbulence (Fig. ), a deeper cloud layer, and a greater liquid water content (Fig. ) allowed rimed snowflakes to form which precipitated from the cloud. This precipitation will act to deplete the liquid in the cloud, potentially leading to cloud break up further downstream.

The sea ice cloud has little in common with the single-layer stratocumulus case observed during M-PACE . Conversely, the microphysics observed here agrees better with observations made during the M-PACE cirrus case study: the high-altitude, predominantly liquid, cloud layer observed below the cirrus cloud had a similar LWC to that measured here. Low ice number concentrations were identified in this layer, as they were found to be connected to the cirrus cloud above it through precipitating snow. In addition, this cloud layer was at a similar temperature (approximately -20 ∘C) to the sea ice cloud observed here. The close proximity to the cold sea ice surface could be causing this cloud to behave like a high-altitude, decoupled cloud: with little-to-no surface sensible and latent heat fluxes affecting the cloud (with little variability from 0 W m-2 observed, Fig. ), the resultant microphysics may evolve as it would higher up in the troposphere. During ISDAC, the springtime single-layer MPS observed over the sea ice had a similar mean LWC to our sea ice cloud 0.1 ± 0.13 g m-3,, further suggesting this is a common observation of MPS over sea ice.

Our measurements show that the cloud downstream over the ocean was deeper than the sea ice cloud, agreeing with the observations of . In contrast to the sea ice case, our ocean observations compare well with the ocean-based, single-layer stratocumulus observed during M-PACE . Low droplet concentrations (46 ± 30 cm-3) and a comparable mean LWC (0.19 ± 0.12 g m-3) were measured during M-PACE and, coupled with our observations, this suggests that such properties are common amongst ocean-based single-layer MPS. Through a comparison between M-PACE and ISDAC data, also concluded that this larger mean LWC during M-PACE was caused by moisture fluxes from the ocean below.

Mean droplet number concentrations varied from 110 ± 36 cm-3 over the sea ice, to 145 ± 54 cm-3 over the MIZ, to 63 ± 30 cm-3 over the ocean. These concentrations were variable with altitude (Figs. , , ) and also varied substantially with longitude over the sea ice (Fig. ) and ocean (Fig. ) where such data were available. These values are consistent with the first ACCACIA spring case reported by but not the second: as also concluded by the authors, their second spring case was subjected to a higher aerosol loading which enhanced the droplet number concentration of the cloud. Consistency between the liquid phase in this study, the spring case 1 from , and the MPS observations reported by other Arctic studies e.g. suggest that droplet number concentrations of ∼ 150 cm-3 or below are common amongst Arctic mixed-phase single-layer stratocumulus in the transition seasons. Mean droplet effective radii over the sea ice are comparable to previous springtime Arctic studies e.g. 5.7 µm,, whilst the larger effective radii measured over the ocean agree better with observations of autumnal Arctic single- and multilayer clouds e.g. approximately 10 and 8–13 µm respectively,. These observations again suggest that larger droplet sizes may be a common occurrence in ocean-based clouds, whilst small droplets are common in clouds over sea ice, regardless of season.

The ice phase is approximately constant across the transition from sea ice to ocean. Again, these measurements agree well with the springtime ACCACIA cases presented by : the ice concentrations are variable and can reach up to ∼ 10 L-1 (thought to be due to crystal fragmentation here), yet they are low on average (0.5–1.5 L-1) throughout the mixed cloud layer. The sea ice cloud observed here would fall into the Type IV category established by , as it was characterised by droplet concentrations of > 100 cm-3, small droplet effective radii, and only a few ice crystals per litre of air (Figs. , , and ). These findings are consistent with the classification of clouds observed during ISDAC, as discussed by . The ocean cloud borders on the Type V category, with larger droplet sizes, mean droplet number concentrations < 100 cm-3, and precipitation developing; however, the observed ice number concentrations are still in better agreement with the Type IV criteria. One could postulate that the continued development of the ocean cloud over the warm surface, with further growth of even larger cloud droplets that might subsequently freeze, could allow the cloud layer to evolve into a Type V cloud, with more ice and less liquid. The microphysical characteristics of these clouds may be more susceptible to cloud glaciation and break up via the WBF mechanism. The M-PACE clouds were categorised as Type V e.g., with higher ice crystal mean of 2.8 ± 6.9 L-1 for mixed-phase single-layer stratus, and lower droplet number concentrations than reported here. From these differences, one might infer that the M-PACE clouds were simply further developed than those observed in this study, or that there may be some influence from either different geographical aerosol sources, minor secondary ice production as suggested by, or seasonal dependencies. Results from ISDAC may address the geographical hypothesis, as ice crystal and cloud droplet concentrations of approximately 0.5–1.5 L-1 and ∼ 150 cm-3 were observed over broken sea ice during the early spring at Barrow, Alaska April 2008,. These concentrations are comparable to our sea ice observations; however, the ISDAC clouds were much warmer, with cloud top temperatures ranging from -15 to -12 ∘C. The microphysical consistency between these clouds observed in different locations may suggest that a similar source of INPs is influencing these clouds, or that the ice phase is not highly sensitive to variability in aerosol properties between different locations. Variability in droplet number between different measurement locations can be more easily explained via pollution events, for example, such as that observed by .

The Arctic Study of Tropospheric Aerosol and Radiation (ASTAR-2004) campaign also made cloud observations in the vicinity of Svalbard; however, much higher ice crystal concentrations (up to 50 L-1) were observed May 2004,. This phenomenon was explained by Hallett–Mossop secondary ice production. Ice enhancement due to crystal collisions was inferred at cloud base over the ocean here, but the temperature was consistently too cold to allow for secondary ice via the Hallett–Mossop pathway. The lack of dominating secondary ice in the Arctic clouds studied here is again consistent with , , and , leading to the conclusion that the single-layer MPS present in the Arctic during early spring are typically too cold for this phenomenon, irrespective of their geographical location. Primary ice nucleation was found to be solely responsible for the ice in the clouds examined here, whilst secondary ice formation has been found to play a greater role in the late spring and summer .