Properties of Arctic liquid and mixed phase clouds from ship-borne Cloudnet observations during ACSE 2014

This study presents Cloudnet retrievals of Arctic clouds fr om measurements conducted during a three-month research expedition along the Siberian shelf during summer an d autumn 2014. During autumn, we find a strong reduction in the occurrence of liquid clouds and an increase for both mixed-p hase and ice clouds at low levels compared to summer. About 80% of all liquid clouds observed during the research cruise show a liquid water path below the infra-red black body limit of approximately 50 gm−2. The majority of mixed-phase and ice clouds had an ice water p ath below 20 gm−2. 5 Cloud properties are analysed with respect to cloud-top tem perature and boundary layer structure. Changes in these par ameters have little effect on the geometric thickness of liqui d clouds while mixed-phase clouds during warm-air advectio n events are generally thinner than when such events were absent. Clo ud-top temperatures are very similar for all mixed-phase cl ouds. However, more cases of lower cloud-top temperature were obs rved in the absence of warm-air advection. Profiles of liquid and ice water content are normalised with r espect to cloud base and height. For liquid water clouds, the 10 liquid water content profile reveals a strong increase with h eig t with a maximum within the upper quarter of the clouds followed by a sharp decrease towards cloud top. Liquid water content is lowest for clouds observed below an inversion dur ing warm-air advection events. Most mixed-phase clouds show a l iquid water content profile with a very similar shape to that o f liquid clouds but with lower maximum values during warm-air dvection. The normalised ice water content profiles in mixe dphase clouds look different from that of liquid water conten t. They show a wider range in maximum values with lowest ice 15 water content for clouds below an inversion and highest valu es for clouds above or extending through an inversion. The ic e water content profile generally peaks at a height below the pe ak in the liquid water content profile – usually in the centre o f the cloud, sometimes closer to cloud base, likely due to particl e sublimation as the crystals fall through the cloud. 1 https://doi.org/10.5194/acp-2020-56 Preprint. Discussion started: 4 February 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
on clear-sky periods) and liquid layer cloud boundaries by distributing the liquid using the scaled-linear adiabatic assumption, i.e. LWC increasing linearly with height from zero at cloud base (Albrecht et al., 1990;Boers et al., 2000). Typical errors in LWC are below 20% (Ebell et al., 2010). IWC is calculated from radar reflectivity and temperature using the method of Hogan et al. (2006), where the fractional error in IWC at 94 GHz is +55%/-35% between -10 and -20 C, rising to +90%/-47% 120 for temperatures below -40 • C. Note that an error in the calibration of the radar reflectivity of 1 dB would bias IWC by 15%.
The Cloudnet target classification Illingworth et al. (2007) has been used to separate between water clouds, ice clouds, and mixed-phase clouds on a profile-by-profile basis with a resolution of 30 s, and to identify cloud base and top heights. The original Cloudnet target classification for the three months of ACSE measurements is presented in Figure 2. The figure also shows fog periods as identified by a visibility of less than 1 km in the 10-min mean of the visibility sensor measurements 125 aboard Oden. The target classification reveals an unrealistically high occurrence of Aerosol, Aerosol & insects, and Insects during periods that were actually dominated by fog. Hence, visibility data have been used to re-classify some of the targets originally misidentified by Cloudnet into these categories below 500 m as fog. Cloud profiles are classified as mixed-phase if they show a cloud layer classified as Cloud droplets only but features precipitating ice below cloud base, or if a cloud layer contains regions of any combination of Ice only, Cloud droplets only and Ice & super-cooled droplets. Profiles of cloud fraction 130 per volume (Brooks et al. 2005) have been obtained using time-height sections of 30 min by 90 m height (3 height bins).
We use the estimates of the depth of the planetary boundary layer (PBL) provided by Sotiropoulou et al. (2016). They obtained PBL depths from the locations of the main inversions in the radiosonde temperature profiles following the methodology of Tjernström and Graverson (2009). A separation between coupled and decoupled boundary layers (Shupe et al., 2013;Sotiropoulou et al., 2014;Brooks et al., 2017) was performed by investigating the presence of an additional, weaker, tempera-135 ture inversion below the main inversion (Sotiropoulou et al., 2016). An absence of such an additional lower inversion defines coupled PBLs. Cloudnet retrievals within one hour of a sounding have been used in the investigation of the effects of (a) coupled and decoupled PBLs and (b) the location of the clouds with respect to the inversion (i.e. PBL top) on the observed cloud properties. To avoid oversampling of persistent clouds, we considered only one Cloudnet profile every 5 minutes, leading to at most 24 profiles for per sounding.

Cloud occurrence
Cloud occurrence probability distributions as a function of height are shown in Figure 3, both for total occurrence and partitioned into liquid, ice, and mixed-phase clouds for the entire ACSE campaign, and separated into summer and the autumn 155 seasons following Sotiropoulou et al. (2016). For completeness, the cloud fraction for all clouds, i.e. including those with a cloud-top height above 6 km for which only cloud base could be detected, is provided as dotted line.
In general, Figure 3 shows that clouds were more abundant below 4 km height during autumn than during summer. This is reflected in the lower tropospheric maxima of the mean cloud fraction of 0.28 and 0.74 in summer and autumn, respectively. In summer, there is a clear separation between height ranges dominated by liquid-water (< 1.2 km) and by ice clouds (> 1.2 km).

160
Mixed-phase clouds during summer were found at all height levels though their cloud fraction strongly decreased upwards of 0.5 km. Autumn showed a strong reduction in the occurrence of liquid clouds and an increase in both mixed-phase clouds and ice clouds at low levels. Ice clouds during autumn extended almost down to the surface, while low clouds during summer were predominantly liquid.
A statistical overview of top temperature, top height, bottom height, and geometrical thickness of the clouds observed during 165 ACSE is provided in Figure 4. The results refer to cloud layers (up to three allowed per profile) for which both cloud base and top could be clearly identified. The minimum cloud geometrical depth was defined by the radar range resolution of 31 m.
Again, the results were separated according to cloud phase and season. Average cloud top temperatures were 0 • C for liquid clouds, -10 • C for mixed-phase clouds, and -15 • C for ice clouds. Cloud top temperatures were slightly higher during summer and slightly lower during winter, though with a similar spread of values. The seasonal behaviour of cloud top and base heights 170 for liquid clouds differs from that of ice and mixed-phase clouds. Liquid clouds were relatively unchanged in vertical extent between summer and autumn, while both ice and mixed-phase clouds had lower top and base heights in autumn than in summer.
In general, the clouds observed during ACSE were rather shallow with a median (mean) geometrical thickness of 250 m (800 m). Liquid clouds were found to be thinnest during both seasons and with only a small variation between median (220 m) and mean (285 m) values. Mixed-phase clouds were the thickest with median depths of 750 m and 940 m in summer and winter, 175 respectively, with a similar mean value for both seasons. Ice clouds were slightly deeper in autumn, with a median (mean) geometric thickness of 250 m (730 m) compared to 220 m (570 m) in summer. It should be emphasised that these statistics are dominated by liquid clouds in summer and by mixed-phase clouds during autumn. 180 The frequency distribution of LWP in liquid water clouds during summer and autumn is shown in Figure 5a. While a negative LWP related to the retrieval error of 25-30 gm −2 (Turner, 2007) is clearly unphysical, these values cannot be excluded without biasing the statistics. Liquid water clouds during summer had a mean LWP of 37±59 gm −2 and median of 13 gm −2 . These 6 https://doi.org/10.5194/acp-2020-56 Preprint. Discussion started: 4 February 2020 c Author(s) 2020. CC BY 4.0 License.

Liquid-water clouds
values were similar during autumn with a mean of 41±54 gm −2 and median of 20 g/m2. Both distributions peak at a LWP around 10 gm −2 . In summer a small number of clouds (less than 1% of all cases) had a LWP in excess of 400 gm −2 while in 185 autumn the maximum LWP was approximately 495 gm −2 . These high values of LWP are generally related to frontal passages.
Almost no seasonal difference in the LWP distributions is apparent in the cumulative frequency curves in Figure 5a. The curves also show that in summer and autumn 76% and 72%, respectively, of liquid clouds were below the infra-red black body limit of approximately 50 gm −2 (Tjernström et al., 2015). If the black body limit is set to 30 gm −2 (Shupe and Intrieri, 2004), the occurrence rates are reduced to about 67% in summer and 60% in autumn. These clouds were therefore often semi-190 transparent to long-wave radiation; hence, long-wave cooling and the resulting turbulence generated in cloud, as well as the surface downwelling radiation, will be very sensitive to small changes in LWP. Figure 5b shows the distribution of cloud-top temperature for liquid-water clouds during summer and autumn. Summer liquid clouds were warmer than those in winter. Their cloud top could be warmer than 15 • C but were never found to be colder than -15 • C. A closer look at the data revealed that all the cloud-top temperatures above 10 • C were the result of a period of In autumn, liquid cloud-top temperatures rarely exceed 0 • C with observed values as low as -25 • C. The maximum of cloud-top temperature occurrence rate shifts from 0 • C in summer to -5 • C in autumn. In addition, cloud-top temperatures for autumn also show a broader distribution with a long tail towards low temperatures than those in summer.

Mixed-phase clouds
The LWP frequency distribution for mixed-phase clouds presented in Figure 6a is similar to that for liquid-only clouds in Figure 5a though with a broader shape. Summer had more cases of high LWP and fewer cases of low LWP than autumn.
For both seasons, the peak occurrence was at around 10 gm −2 . The mean and median values, however, are higher than for liquid-only clouds, with summer values of 98±94 gm −2 and 72 gm −2 , respectively; in autumn the corresponding values are 205 34±44 gm −2 and 21 gm −2 . The cumulative distributions in Figure 6a show that, with infrared-black body limit of 50 gm −2 (30 gm −2 ), 41% (31%) and 76% (60%) of the clouds during summer and autumn, respectively, had LWPs below this limit. The same general relationships of higher median LWP in mixed-phase clouds compared with liquid-only clouds is consistent with the observations during SHEBA (Shupe et al., 2006).
In contrast to LWP, there is little difference in the frequency distributions for IWP in the mixed-phase clouds observed in 210 either summer or autumn ( Figure 6b). The majority of clouds had an IWP below 20 gm −2 with mean and median values in summer of 34 and 7 gm −2 , respectively, and in autumn of 32 and 9 gm −2 .
During summer, IWC was lowest in clouds with a low cloud top height and highest for clouds with tops between 3.0 and 4.0 km and cloud-top temperatures of -8 • C to -17 • C (not shown). During autumn, the lowest values of IWC were observed for clouds with top heights in the range from 2.0 to 3.0 km. Cold clouds with cloud top temperatures between -15 • C and during summer and autumn had very low IWC; < 0.1 gm −3 . Mean (median) values were 0.0156 gm −3 (0.0025 gm −3 ) and 0.0087 gm −3 (0.0016 gm −3 ) during summer and autumn, respectively.
The frequency distribution of cloud-top temperature in Figure 6c again shows a different behaviour for clouds during summer and autumn. During summer, the tops of mixed-phase clouds were generally warmer than in autumn with a maximum at 0 • C 220 to -2.5 • C. However, they were always colder than liquid-only clouds during the same season. During summer, cloud-top temperature could be as low as -30 • C though they were mostly warmer than -5 • C. Autumn had a bi-modal distribution of cloud-top temperature, which could be the result of precipitating (T top >-10 • C) versus non-precipitating clouds (T top <-10 • C) (Westbrook and Illingworth, 2011). Very few mixed-phase clouds showed cloud-top temperatures above 0 • C (these were cases related to warm-air advection events where the cloud top extended into the warmer air above) or as low as -35 • C. In general, 225 mixed-phase cloud top temperatures were up to 5 • C colder during autumn than during summer.

Effect of boundary-layer structure
Here we investigate the effects of PBL structure on the observed clouds. The PBL top is defined as the height of the strongest temperature inversion (Brooks et al., 2017)   structure and large-scale circulation. We separate between liquid and mixed-phase clouds observed above, below, and extending into the inversion during WAA and non-WAA conditions as well as for coupled and decoupled PBLs. Cases of liquid and mixed-phase clouds in decoupled PBLs during WAA events were rare (N<100) in the ACSE data set, and thus, not considered 235 here. Liquid clouds showed little difference in mean and median cloud thickness. However, they do show a clear difference in the frequency distribution of cloud-top temperature with respect to WAA and non-WAA conditions. Mixed-phase clouds during WAA were generally thinner than during non-WAA. The deepest mixed-phase clouds were found for non-WAA and for decoupled PBLs. No difference is found in the thickness (Figure 7b) and cloud-top temperature (dotted line in Figure 7d  The profiles of IWC in mixed-phase clouds are distinctly different from those of LWC. They show a wide range in maximum values with lowest IWC close to 0 gm −3 for clouds below the inversion and highest values of 0.25 to 0.75 gm −3 for clouds above or extending through the inversion. Note that these are also the geometrically thinnest and thickest clouds, respectively 255 ( Figure 7). The IWC profile generally peaks at a height below the peak in the LWC profile -usually in the centre of the cloud but sometimes closer to cloud base, likely due to increasing particle sublimation as the crystals fall.
During non-WAA, liquid clouds below the inversion (i.e. with cloud top at or below PBL top) showed no statistically significant difference in LWP (two-sample t-test, p < 0.05) for coupled and de-coupled PBLs, with mean values of 24±62 gm −2 (median of 6 gm −2 ) and 22±41 gm −2 (median of 8 gm −2 ), respectively (not shown). For clouds below the inversion in coupled 260 PBLs, 90% of cases showed LWP below 50 gm −2 while this number slightly decreases to 88% for clouds below the inversion in decoupled PBLs. This behaviour is consistent with the observations reported in Sotiropoulou et al. (2016).

Mixed-phase clouds in the same situation (non-WAA, below inversion) showed LWP behaviour for coupled and de-coupled
PBLs opposite to that of liquid clouds. We find a statistically significant difference (two-sample t-test, p < 0.05) with mean values of 33±57 gm −2 (median of 13 gm −2 ) and 52±63 gm −2 (median of 32 gm −2 ), for coupled and de-coupled PBLs, re-265 spectively (not shown). For clouds below the inversion in coupled PBLs, 76% of cases showed LWP below 50 gm −2 while this number decreased to 64% for clouds below the inversion in decoupled PBLs. Interestingly, mixed-phase clouds below the inversion in decoupled PBLs were slightly thinner than in coupled PBLs (Figure 7b) while little difference was found in their respective profiles of IWC (Figure 8c). Arctic Ocean. ASCOS cloud fractions were obtained following Shupe (2006). The profiles of total cloud fraction are very similar in shape but show a generally lower cloudiness from ACSE. Note that while the profiles represent roughly the same 285 period of the year, the actual observations have been performed at different locations and in different years. Nevertheless, the resemblance in the shape of the total cloud fraction profile indicates the usefulness of relating Arctic observations to each other; particularly given their scarcity. For the comparison of cloud fraction, we need to keep in mind that the upper measurement height during ACSE was restricted to 6 km by instrument settings. This constrains all cloud fractions to zero at and above 6 km, as we only consider clouds for which a cloud top has been detected below this height. The total cloud fraction for all clouds 290 including those with undetected top heights, i.e. top heights above 6 km, is given by the grey dashed line for reference.
The cloud-fraction profile for liquid-only clouds during ACSE generally resembles the profiles derived from ASCOS measurements. However, the occurrence of liquid-only clouds was much lower during ACSE, except for the frequent fog periods in the lowermost height bins during the summer months. The occurrence of ice and mixed-phase clouds during ACSE also appear to be quite similar to those obtained from ASCOS. Considering that most of the clouds with undetected tops are likely to be ice 295 clouds and that the shape of the cloud-fraction profile for mixed-phase clouds during ACSE resembles that of ASCOS, Figure   9 shows that the height from which ice clouds are the dominant cloud type was about 1 km lower for ACSE than for ASCOS.
The monthly total cloud fraction of 95% in July, 74% in August and 97% in September as observed during ACSE can also be put into the context of previous studies. Shupe (2011b) compared observation from surface land sites (Figure 2) in Atqasuk (ceilometer, microwave radiometer), Barrow (ceilometer, radar, micro-pulse lidar, microwave radiometer, Atmospheric Emitted 300 Radiance Interferometer), Eureka (radar, high spectral resolution lidar, micro-pulse lidar, microwave radiometer, Atmospheric Emitted Radiance Interferometer), and the SHEBA project (ceilometer, radar, microwave radiometer, Atmospheric Emitted Radiance Interferometer). For July to September, they present a total cloud fraction of 92% to 98% at Barrow and Sheba.
Lower values of 80% to 85% are given for Atqasuk, while increasing from 65% in July to 80% in August and September at Eureka. Zygmuntowska et al. (2012)  values of 75% to 80% in July, 80% to 87% in August, and 84% to 90% in September. For all clouds, ACSE observations of more than 90% during July and September are mostly in line with the high cloud fractions observed during SHEBA (Shupe,310 2011b).
Cloud fractions of 60% to 90% as observed at Eureka (Shupe, 2011b) and for the Arctic region (Zygmuntowska et al., 2012;Mioche et al., 2015) suggest that the ACSE finding of a total cloud fraction of 74% in August is well within the range of values one would expect for the Arctic region. However, it should be noted that spaceborne data sets provide better spatial coverage than ground-based measurements during ACSE, and thus, are more representative of average conditions. When comparing the 315 fraction of mixed-phase clouds observed during ACSE to the multi-year (2007 to 2010) CALIPSO/CloudSat data set analysed by Mioche et al. (2015) it is apparent that the ground-based ACSE observations during July with a mixed-phase cloud fraction of 51% are in general agreement with the data from spaceborne instruments. However, ACSE observations of 33% during August and 80% during September show significantly lower and higher, respectively, fractions of mixed-phase clouds than the satellite record. This is probably the result of natural variability combined with the effect of comparing local measurements 320 during ACSE to area averaged results from satellite. Considering the fraction of mixed-phase clouds at Barrow, Eureka and SHEBA (Shupe, 2011b), ACSE findings are in line with SHEBA values of around 50% during July and around 85% during September. However, the ACSE mixed-phase cloud fraction of 33% during August is much lower than the SHEBA observation of around 80% (see Figure 2 in Shupe (2011b)). The lower August mixed-phase cloud fraction during ACSE does, however, resemble the findings for Barrow and Eureka (Shupe, 2011b). 325 Figure 10 compares the connection between the fraction of ice-containing clouds and cloud-top temperature for clouds observed during ACSE with those reported by Zhang et al. (2010) and Bühl et al. (2013). These previous studies combine measurements with cloud radar and aerosol lidar from space and ground, respectively. As in this study, they analyse clouds on a profile-by-profile basis. However, Zhang et al. (2010) and Bühl et al. (2013) focused on mixed-phase clouds at mid-latitudes.
While they find that about 50% of all clouds are mixed-phase at a temperature of about -10 • C, the ACSE observations reveal 330 that in the Arctic a mixed-phase cloud fraction of 50% is reached already at -2 • C. Previous studies suggest that almost all non-cirrus clouds with cloud top temperatures below -20 • C are mixed-phase at mid-latitudes. In the Arctic, this is the case already for warmer cloud-top temperatures of -12 • C.; though ice-containing cloud fractions for clouds with top temperatures below -18 • C to -25 • C were found to be lower than at mid-latitudes for ACSE observations during autumn. Figure 11 puts the ACSE observations of LWP and IWP for clouds during summer and autumn into the context of the earlier 335 observations of SHEBA and ASCOS. ACSE LWP frequency distributions -though different for summer and autumn -do not resemble the previous observations, having a wider distribution with less well defined peak. The ACSE observations of IWP closely follow the ASCOS frequency distribution, although with larger values in the tail. There was quite a substantial part of the ASCOS ice drift during which mixed-phase stratocumulus clouds dominated, that may bias ASCOS LWP statistics high.
In addition, air mass transit time is known to be an important factor in boundary layer structure and hence cloud properties.

340
The fact that SHEBA and ASCOS have been further away from open water than ACSE means that air mass transit time is a factor controlling the cloud properties observed.

Summary and Conclusions
We present remote-sensing observations of Arctic clouds conducted during a three-month cruise in the Arctic Ocean along the Russian shelf from Tromsø, Norway, to Barrow, Alaska, and back. Observations with ceilometer, Doppler lidar, cloud radar 345 and microwave radiometer were made within pack ice, open water, and the marginal ice zone. The Cloudnet retrieval has been applied to investigate cloud properties with special emphasis on the effects of cloud-top temperature and boundary layer structure. The data set has been split into summer and autumn based on a change in the lower tropospheric mean temperature observed from radiosoundings (Sotiropoulou et al., 2016).
The ACSE data set reveals a strong reduction in the occurrence rate of liquid clouds and an increase for both mixed-phase 350 clouds and ice clouds at low levels during autumn compared to summer. Ice clouds during autumn extend almost down to the surface, while low clouds during summer are predominantly liquid. In addition, it was found that liquid clouds vary little in their vertical extent between summer and autumn, while both ice and mixed-phase clouds have lower top and base heights in autumn than in summer.
About 74% of all liquid clouds observed during ACSE show LWP below the infra-red black body limit of approximately 355 50 gm −2 . This means that the majority of the observed Arctic liquid water clouds have long-wave radiative properties that are highly sensitive to small changes in LWP. In general, the frequency distribution of LWP shows little variation for mixed-phase and purely liquid clouds. Nevertheless, summer shows more cases of high LWP and fewer cases of low LWP and the mean and median values are higher for mixed-phase clouds. The majority of clouds had an IWP below 20 gm −2 with summer (autumn) mean and median values of 34 and 7 gm −2 (32 and 9 gm −2 ), respectively.

360
Whether the PBL structure was coupled or decoupled, and the occurrence of warm air advection had little effect on the geometric thickness of liquid clouds. In contrast, mixed-phase clouds during WAA are generally thinner than for non-WAA.
The deepest mixed-phase clouds are found for non-WAA and for decoupled PBLs.
Cloud-top temperatures for all mixed-phase clouds during non-WAA are between 0 • C and -30 • C. This range is reduced to 0 • C to -20 • C for mixed-phase clouds during WAA.

365
For liquid water clouds, the normalised profile of LWC reveals a strong increase with height with a maximum between 0.03 and 0.19 gm −3 within the upper quarter of the clouds followed by a sharp decrease towards cloud top. LWC is lowest for clouds observed below the inversion during WAA. Most mixed-phase clouds show a LWC profile with a very similar shape to that of liquid clouds with lower maximum values during WAA than during non-WAA.
The normalised profiles of IWC in mixed-phase clouds look different from that of LWC. They show a wider range in 370 maximum values with lowest IWC for clouds below the inversion and highest values for clouds above or extending through the inversion. Note that these correspond to the thinnest and thickest clouds, respectively. The IWC profile generally peaks at a height below the peak in the LWC profile -usually in the centre of the cloud but also closer to cloud base and likely due to more particle sublimation as the crystals fall.
Unsurprisingly, it was found that liquid-water clouds during summer show the highest cloud-top temperatures, which can 375 exceed 15 • C but don't go below -15 • C. As documented in Tjernström et al. (2015Tjernström et al. ( , 2019, ACSE cloud-top temperatures above 10 • C correspond to a period of strong warm air advection that occurred at the beginning of August 2015. As a consequence, the frequency distribution of cloud-top temperature observed during summer resembles that derived from Cloudnet observations at mid-latitudes (Bühl et al., 2016). In autumn the top temperatures of liquid clouds rarely exceed 0 • C with observed values as low as -25 • C. The maximum of cloud-top-temperature occurrence rate shifts from 0 • C in summer to -5.0 • C in autumn.

380
During summer, the tops of mixed-phase clouds are generally warmer than in autumn with a maximum just below 0 • C. However, they are always colder than liquid-only clouds during the same season. During summer, cloud-top temperature can be as low as -25 • C though they are mostly warmer than -10 • C. Autumn reveals a bi-modal distribution of cloud-top temperature corresponding to precipitating (T top >-10 • C) versus non-precipitating clouds (T top <-10 • C).