1. Introduction
Clouds are prevalent in the high-latitude postfrontal marine boundary layer (BL) despite the deep subsidence typically found following the passage of a cold front. To a first order, this cloud cover can be explained by surface-driven shallow convection. Shallow clouds are particularly common over the Southern Hemisphere (Mace et al. 2009; McFarquhar et al. 2021). Satellite imagery suggests that the areal fraction occupied by shallow clouds over the Southern Ocean, especially its southern part around Antarctica in the warm season, is considerably higher than most climate models predict (Trenberth and Fasullo 2010). This model bias is most pronounced in the cold sector of Southern Ocean cyclonic storms (Bodas-Salcedo et al. 2014), suggesting that most models underestimate the liquid fraction and overestimate ice-mediated precipitation growth in these shallow clouds (Choi et al. 2010). Mesoscale cellular convection over the Southern Ocean generally has cloud-top temperatures (CTTs) below freezing (e.g., Fletcher et al. 2016; Zaremba et al. 2020; Lang et al. 2021). Models disagree significantly about cloud-phase composition, including in shallow clouds over the Southern Ocean (Lenaerts et al. 2017; Atlas et al. 2020). The predominant cloud phase matters because it affects precipitation growth and fallout, cloud lifetime, and thus radiative properties of the aggregate cloud field: the introduction of ice particles in a shallow convective cloud containing supercooled liquid tends to accelerate precipitation, decrease the life span and cloud optical depth, and thus decrease the reflected shortwave radiation, a climate warming effect (H. Morrison et al. 2011). The abovementioned model bias has implications not just for the regional albedo, but also Southern Ocean baroclinic storm dynamics and the global climate system (Trenberth and Fasullo 2010). For instance, models that glaciate high-latitude mixed-phase clouds relatively rapidly produce more warming in CO2 doubling experiments (McCoy et al. 2014).
Satellite imagery indicates that shallow clouds in the postfrontal sector in the Southern Ocean are mostly convective (Naud et al. 2020), frequently display mesoscale cellular organization (Muhlbauer et al. 2014), and often produce light precipitation (Ahn et al. 2017). The properties and mesoscale organization of these clouds are the result of interactions between large-scale deep-tropospheric forcing (generally subsident), surface fluxes, BL circulations, turbulence, radiation, as well as aerosol, cloud, and precipitation processes. An understanding of these interacting processes requires dedicated in situ measurements, which are difficult to make given the remoteness of the Southern Ocean.
There is a dearth of observational constraints to improve the representation of ice initiation and multiplication in clouds, and this challenge is especially impactful in marine BL clouds over the Southern Ocean. This motivated a coordinated series of field campaigns in 2017/18, involving an instrumented aircraft, two ships, and an instrument deployment on a remote island in the Southern Ocean (McFarquhar et al. 2021). One of these was the Measurements of Aerosols, Radiation, and Clouds over the Southern Ocean (MARCUS) campaign aboard the Australian icebreaker Aurora Australis. This paper (Part I) examines postfrontal shallow clouds observed with the in situ probes and profiling remote sensors aboard this icebreaker during its voyages between Tasmania and the Antarctic coast in the austral summer of early 2018.
In recent years several studies have used in situ measurements to characterize the cloud-phase composition of supercooled shallow clouds over the Southern Ocean. Ahn et al. (2017) found that a significant fraction (39%) of the shallow clouds sampled by aircraft just south of Tasmania contains ice crystals, almost all in conjunction with liquid droplets. Four of their 20 flights were in postfrontal conditions, but they did not classify cloud-phase data as a function of synoptic condition. Huang et al. (2017) argued that ice particles are common in postfrontal mesoscale cellular convection over the Southern Ocean. They presented aircraft measurements of abundant ice particles in relatively warm marine BL clouds (CTTs around −9°C), again near Tasmania. D’Alessandro et al. (2019) report rather frequent (17%–37%) encounters of ice in aircraft cloud penetrations in the −10° to 0°C range in summer over the Southern Ocean, more frequently than time–space matched output from climate models. Some of this ice could have initiated at far lower temperatures higher in the cloud, but this study did not examine CTTs. The observation of numerous ice particles at relatively high temperatures may be due to Hallet–Mossop ice multiplication (secondary ice production) (Huang et al. 2017; Lang et al. 2021; McFarquhar et al. 2021).
Contrary to these in situ observations, spaceborne data, in particular from the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO), indicate that the tops of shallow clouds over the Southern Ocean contain or are dominated by supercooled liquid water (SLW) (Hu et al. 2010; A. E. Morrison et al. 2011; Muhlbauer et al. 2014; Huang 2017). Combined airborne lidar and radar observations over the Southern Ocean indicated that 92% of the subfreezing tops of postfrontal cold-sector clouds had supercooled liquid, and 75% were liquid-only (Zaremba et al. 2020). It should be noted that MODIS and CALIPSO cloud-phase estimation applies to the cloud-top region only and tends to underestimate the frequency of ice occurrence over the full cloud depth (e.g., Ahn et al. 2018). Liquid-dominated marine BL clouds have been observed to contain ice particles below cloud top (Rangno and Hobbs 1991; Crosier et al. 2011; Mace and Protat 2018). A comparison of the CloudSat–CALIPSO-based cloud-phase determination against a surface-based approach confirmed that thin supercooled cloud layers over the Southern Ocean are more commonly liquid dominated, down to −30°C (Mace et al. 2021a). Mace et al. (2021a) note that ice, when present in liquid clouds, is not observable with sensors based on visible light scattering: any ice hydrometeors present within a mostly liquid cloud are hidden by the much greater surface area of droplets that dominate visible scattering near cloud top (for spaceborne lidars) and cloud base (for surface-based lidars). Surface-based lidars have the advantage of being able to measure hydrometeors falling below cloud base (and still above the freezing level). Mace et al. (2021a) further found that ice at levels below the cloud top appears to be more common in cold-sector convection equatorward of the oceanic Antarctic polar front.
In this context, the present study examines the vertical structure of clouds and precipitation and cloud phase in postfrontal shallow convection over the Southern Ocean. In Part I, we use upward-looking remote sensing data collected during MARCUS between 52° and 67°S over Southern Ocean. The observed clouds were generally 0.5–2.5 km deep, and CTTs generally ranged between −18° and −8°C. The main objectives are to determine predominant cloud phase, and to examine the origin of ice in these clouds. Numerical simulations of these clouds are examined in Part II (Hu et al. 2023), to examine the ability of simulations with different microphysics schemes to reproduce the observed cloud-phase distribution.
Data sources and the analysis method are introduced in section 2. The case study results are presented in section 3. Section 4 analyzes other postfrontal shallow clouds during the MARCUS campaign. Section 5 discusses key questions. Section 6 presents a summary and the main conclusions.
2. Data sources and methods
a. The MARCUS dataset
The data used herein were collected in the MARCUS field program, conducted between 29 October 2017 and 25 March 2018 (the austral summer) (McFarquhar et al. 2021). During MARCUS, instruments from the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) user facility’s second Mobile Facility (AMF2) were installed on the ice breaker Aurora Australis during four voyages between Hobart, Australia, and the Australian Antarctic stations. This study focuses on data from the last two voyages, specifically the legs between Davis and Hobart and between Hobart and Macquarie Island.
The AMF2 obtained measurements of BL aerosols, vertical distributions of microphysical properties of clouds, in situ meteorology, eddy fluxes, and radiative fluxes in MARCUS. The instruments most important to this study are the balloon-borne sounding system, the profiling micropulse lidar (MPL), the passive microwave radiometer (MWR), the Vaisala ceilometer, the ship’s inertial navigation system, and the profiling Marine W-band (95 GHz, or 3 mm) ARM Cloud Radar (MWACR). MWACR reflectivity values were adjusted upward by 7.0 dB, based on intercomparison analyses by Kollias et al. (2019) and Mace et al. (2021b) against the well-calibrated W-band radar on CloudSat, plus a recommendation from the ARM team (S. Giangrande 2022, personal communication). This upward correction brings the MWACR reflectivity values closer to proximity GPM Ka and Ku radar reflectivity values during the main case examined herein (Fig. 1). Ku-band (13.6 GHz, or 2.1 cm) precipitation radar (KuPR) reflectivity values up to 30 dBZ are observed in rather close proximity to the ship (Figs. 1a,b), and average values are close to 20 dBZ. Note that the minimum detectable radar reflectivity value for the KuPR is 15.5 dBZ, plus the footprint size is rather large (5 km) (Liao and Meneghini 2022); therefore, many more benign or smaller precipitating clouds in the scenes of Fig. 1 are undetected by the KuPR. As will be shown below, MWACR reflectivity values around the times of these two overpasses are much lower, with mean and 95th percentile values around 3 and 10 dBZ. The W-band radar likely is incapable of characterizing the most intense echoes in these clouds. Mie scattering occurs for particles larger than ∼1 mm at W band. The higher KuPR reflectivity values, in essence, are due to Rayleigh scattering for larger particles, up to ∼7 mm at Ku band. The W-band backscatter cross section asymptotes for these larger particles (such as snow) and reflectivity becomes, to a first order, a function of number of particles instead of size. This is an important limitation we will consider for the stronger MWACR echoes. Another factor is the attenuation of W-band reflectivity by liquid water in the column. No further correction to MWACR reflectivity was made to compensate for this, because the vertical distribution of liquid water content (LWC) is unknown.
KuPR attenuation-corrected reflectivity (dBZ) (left) mapped at 1.5 km ASL and (right) in a vertical transect along the GPM track, which is shown as a line in the left panels. The overpasses at (top) 1344 UTC 24 Feb and (bottom) 0629 UTC 25 Feb 2018. The red star shows the location of the Aurora Australis.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
The key derived variables used in this study are listed in Table 1. Note that very low values of liquid water path (LWP) (<20 g m−2) are uncertain because the MWR used in MARCUS has an accuracy of about 20 g m−2 (Westwater et al. 2001; Crewell and Löhnert 2003). Several of the products in Table 1 are value-added products (VAPs), involving more quality control, separate measurements, and/or reanalysis data. We use the experimental ARM VAP THERMOCLDPHASE, a thermodynamic cloud-phase classification (section 2b). We also use air parcel back-trajectories from the MARCUS Environmental Parameters VAP dataset version 1.41.
MARCUS datasets used in this study. δt is the time resolution, and δz the height resolution (if applicable).
The advantage of the MARCUS dataset is that data were collected aboard a vessel traveling between Antarctica and Australia: these measurements represent unique offshore conditions over the full range of latitudes of the Southern Ocean. MARCUS data are not contaminated by terrain effects, unlike data collected in a sister campaign, the 2016–18 Macquarie Island Cloud Radiation Experiment (MICRE). However, while MICRE collected two years of data, the Aurora Australis spent rather little time crossing the Southern Ocean (Fig. 1 in McFarquhar et al. 2021).
b. Cloud and precipitation phase classification
The THERMOCLDPHASE VAP uses the multisensor method developed by Shupe (2007) to provide vertically resolved cloud hydrometeor phase classification from combined lidar, radar, microwave radiometer, and radiosonde measurements. Eight classes are distinguished, plus clear sky: cloud liquid, cloud ice, mixed phase, snow, rain, drizzle, liquid + drizzle, and unknown. The THERMOCLDPHASE algorithm is based on several other products listed in Table 1: ARSCLWACR1KOLLIASSHP, INTERPSONDE, MWRRET1LILJCLOU, and 30SMPLCMASK1ZWANG. Cloud liquid is identified by strong lidar backscatter intensity (or large backscatter gradient) and low depolarization ratio. When lidar signals are fully attenuated, radar reflectivity, mean radar Doppler velocity, and radar Doppler spectrum width measurements are used distinguish between liquid, ice, or mixed phase. Radar reflectivity and mean radar Doppler velocity measurements are further used to distinguish between small cloud particles (liquid, ice), medium size particles (drizzle), and large precipitating particles (rain and snow). In addition, ancillary measurements of liquid water path from the microwave radiometer and temperature profiles from radiosonde measurements are used to further constrain the classifications. To better compare with model simulations and aircraft in situ measurements, the THERMOCLDPHASE VAP also classifies the whole cloud layer thermodynamic phase as liquid, mixed phased, or ice following the method by Korolev et al. (2017). If the fraction of ice-containing hydrometeors (e.g., ice, snow, and mixed phase) in the cloud layer from cloud base to cloud top (μice) is smaller than 0.1, the cloud layer is determined as a liquid layer. If 0.1< μice < 0.9, it is a mixed-phased cloud layer; and if μice > 0.9, it is an ice cloud layer.
The classification thresholds and the use of constraints from ancillary measurements are based on our understanding of cloud hydrometeor physical properties (e.g., particle size distribution, shape, number concentration, falling velocity, freezing/evaporation temperatures) and from the literature (Eloranta 2005; Shupe et al. 2004, 2006). Quantitative assessment of the uncertainty of the classification algorithm is difficult due to the lack of a definitive validation dataset.
Sources of potential error include the temporal interpolation of 6-h soundings, the 20 g m−2 uncertainty associated with the LWP retrieval, and the thresholds employed in the classification algorithm. The first of these mainly affects clouds with temperatures that are close to important temperature thresholds at −40° and 0°C. Potential errors from the LWP uncertainty are the most significant for thin clouds, e.g., clouds with very little liquid water might be falsely classified as ice, or those with no liquid water might be falsely classified as being liquid or mixed phase. Finally, the classification thresholds were largely based on relationships from the literature (see Shupe 2007). Evaluations of the impacts of these thresholds on the THERMOCLDPHASE classification are encouraged.
c. Auxiliary data
The ECMWF reanalysis v5 (ERA5) dataset (horizontal resolution 31 km, time resolution 1 h) (Hersbach et al. 2020) is used for synoptic analyses and surface heat flux estimates. The sea surface temperature (SST) estimate is from the NOAA National Climatic Data Center Daily Optimum Interpolation Sea Surface Temperature (OISST) data (Huang et al. 2021). The mesoscale structure of the cloud fields near the Aurora Australis is analyzed using 1-km-resolution MODIS visible imagery from both Terra or Aqua overpasses, providing two daytime snapshots per day.
d. Analysis methods
1) Hydrometeor vertical velocity estimation
The MWACR was positioned on a stabilized platform, intended to minimize the effect of ship motion on Doppler measurements. As can be seen in Fig. 2, the Doppler velocity VDop in the ARM Data Archive (Table 1) contains a strong oscillatory component due to ship heave (up and down) and the rocking of the ship into and out of the wind. In most cases, the platform motion contamination is largely due to ship heave, which is aligned with the radial velocity. Contamination from platform roll and pitch can become large during strong crosswinds. Correcting such high-wind cases relies on measured roll/pitch angles, the sounding-estimated wind profile, and the time match between the ship attitude measurements and the Doppler measurements. In the example shown in Fig. 2, the velocity perturbations are nearly identical to the ship heave because the small roll and pitch result in a total angular deviation from vertical that is less than 1°. Here, a 2-s offset was applied to the heave values to achieve a better match with Doppler velocity. After subtracting the heave in the process described below, the wavelike contamination in the resulting hydrometeor vertical velocities is much reduced (see the red line in Fig. 2) but some contamination remains as a result of low time resolution (preventing a good match) and from assuming a constant sounding wind profile. A moving average (in time dimension) of 10 s is applied to reduce the remaining contamination. The platform motion contamination of Doppler spectral width cannot be removed; thus, this field is rather noisy, with columns of enhanced spectral width, mostly associated with accelerations in ship heave.
(a) Average (over all heights) vertical velocity perturbation (black) plotted with the ship’s heave velocity (green) between 1322 and 1325 UTC 24 Feb 2018. The corrected vertical velocity perturbation is shown in red. The perturbations velocities are departures from a 10-s running mean. (b) Absolute value of the original vertical velocity perturbation (black) plotted with MWACR’s deviation from vertical (green) caused by the roll and pitch.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
The correction carries an uncertainty related to accuracy of the assumed profiles of (u, υ). The wind profiles are obtained from the INTERPSONDE dataset (Table 1) whose time resolution is 1 min, but based on 6-hourly data. In theory, HVM can be partitioned between vertical air motion and hydrometeor fall speed (terminal velocity). We make no attempt to do so, but over long time periods (an hour or more), the mean vertical air motion can be assumed to be zero, and thus, the residual mean HVM can be interpreted as the mean (reflectivity-weighted) fall speed. The mesoscale cloud systems examined here are assumed to be large enough that this assumption holds, to a first order, i.e., we will interpret the mean HVM as the average hydrometeor fall speed.
2) Ice water content retrieval from profiling radar data
The ice water content (IWC) and LWP in these clouds is generally too high for the use of a lidar-radar IWC retrieval algorithm (Heymsfield et al. 2008). Retrievals using radar reflectivity alone have limitations as well, especially for higher IWC values (Matrosov and Heymsfield 2008; Heymsfield et al. 2016). Instead, we estimate IWC from two radar moments, MWACR radar reflectivity and HVM, following the algorithm in Deng et al. (2022), which is adapted from a W-band three radar moments method described in Deng and Mace (2006). The basic assumptions are that the reflectivity and reflectivity-weighted Doppler velocity are dominated by ice particles, and that these particles are not heavily rimed. (We will validate these assumptions later, based on the distribution of reflectivity and HVM values.) Furthermore, this retrieval algorithm requires an a priori choice of ice particle habit. Here, we assume aggregates, an assumption that results in the best validation of retrieved IWC against in situ measurements for the mixed-phase clouds examined in Deng et al. (2022). Deng et al. (2022) show that this retrieval works well in ice-dominated orographic clouds, through comparison against independent in situ IWC estimates. More than 100 h of collocated data from an airborne W-band radar and a Nevzorov probe were used.
The IWC retrieval does not use THERMOCLDPHASE. If the MWACR reflectivity is partially or entirely due to drizzle or raindrops, then the retrieved IWC overestimates the actual IWC. This is relatively rare, as will be shown below: only 1.7% of the clouds are classified as drizzle or rain at low levels (the −5° to 0°C layer), and none higher up, according the THERMOCLDPHASE.
The ice water path (IWP) is integrated from either the first level with good MWACR data [about 280 m above mean sea level (MSL)] or the freezing level, whichever is higher, to the top of the BL clouds, to obtain IWP. (The freezing level generally is below 280 m MSL in main case examined here, referred to as case 10.) MWACR reflectivity profiles indicate that for the cases examined here, cloud layers above the BL are rarely present. If they are, they are excluded from the IWP integration as this study focuses on the shallow convective clouds in marine BL. Temperature is not considered in the IWC retrieval.
3) Surface-driven thermal instability
4) Case selection
This study focuses on postfrontal BL convection. Time periods of interest are defined as those with BL thermal instability (Me > 0). We further require shallow clouds (MWACR echo top generally below 3.0 km MSL) and eliminate those periods where, according to MWACR time–height transects, shallow clouds are seeded by clouds aloft (resulting in a much higher first echo top). We also require the surface air temperature to be <+5°C and the lowest CTT for any contiguous MWACR-observed cloud system to be <−10°C, in order to focus on the colder clouds in MARCUS, where radar reflectivity may be ice dominated, and the retrieved IWC [section 2d(2)] is more likely to be representative. For a time period to be included in this analysis, we also require a minimum duration (at least 6 h). Two adjacent periods no more than 6 h apart are merged into a single one. In addition, the essential data from the MWR, MWACR, MPL, and radiosondes need to be available.
This selection process leads to 12 suitable time periods (called cases) in MARCUS (Table 2). These cases correspond approximately to the cold-sector (postfrontal) periods in Table S9 in McFarquhar et al. (2021), which is based on synoptic analyses. Our study includes a few short cases that are not included in McFarquhar et al. (2021); their Table S9 only includes major frontal passages. During the first two ship voyages, prior to 13 January 2018, the MPL power (and thus the MPL signal penetration depth in cloud) were reduced, and the polarization was compromised. Thus, products such as THERMOCLDPHASE are less accurate. Therefore, we decided not to use cases 1–8 in this study. This study is based on just 108 h of postfrontal shallow clouds sampled by all relevant instruments during the third MARCUS voyage (Table 2).
Postfrontal shallow convective cloud periods during MARCUS used in this study. Only cases 9–12 are used in this study (highlighted in boldface). Case 10 is the main event (highlighted in italics).
3. Case study
a. Synoptic and mesoscale overview
The main case studied here is case 10, the longest-duration case in MARCUS lasting 75 h, with a relatively high average Me value, compared to the other cases (Table 2). This period does include a brief period where the echo top was slightly too high to be included in our dataset (although still below 4.0 km), and another brief period that was slightly too warm at the surface, but mesoscale shallow cloud fields, mostly open-cellular convection, are persistent during this 75-h period. Almost all sea ice around Antarctica had melted at the time of case 10, in late summer (Fig. 3). During the 75-h period, the Aurora Australis moved to the northeast from 60° to 53°S, remaining within an eastward-moving postfrontal cold sector (Fig. 3). MODIS visible imagery indicates mesoscale cloud fields (Fig. 4) near the ship at most times during this event. An open-cellular structure is evident at some times (e.g., Fig. 4c). During the first day of this case (23 February 2018), back trajectories of BL air (ending at 1 km above the ship) (Fig. 5) suggest that the air sampled by the Aurora Australis had been in the circumpolar maritime flow for at least 3 days (Table 3).
Synoptic map over Southern Ocean with ship track (bold black and white line) and ship location (cross) during case 10. Thin white contours indicate 850 hPa height and color fill shows local temperature anomaly from 30-yr average. The thin red contour indicates sea ice extent. Frontal boundaries are overplotted.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
MODIS satellite images during case 10. The track of the Aurora Australis during this case is shown with a red line, and its location at the time of the satellite image is marked by a red star. Frontal boundaries are overplotted.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
The 72-h back trajectories ending at a height of 1 km MSL during case 10. Colored lines indicate the height of air parcel. Black dots on the trajectories are 12 hourly time markers. Date and time (DD/HH Z) of parcel arrival in February 2018 is shown. The thin red contour indicates sea ice extent. Blue ship tracks indicate “maritime periods,” as defined in text. Red ship tracks indicate “CAO periods.”
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
The 72-h back-trajectories ending at a height of 1.0 km MSL at the location of the Aurora Australis during case 10. The distance (duration) refer to the location where (time when) the air parcel left Antarctica or the adjacent sea ice. Times and dates are in 2018.
We define this as a “maritime period” (ship track shown in blue in Fig. 5), with large open cells alternating with transverse precipitation bands (Fig. 4a). Near the end of the first day (∼2100 UTC 23 February), the tail end of an occluded front moved over the ship, with infrared (IR) CTTs around −19°C, according to data from the Visible IR Imaging Radiometer Suite (VIIRS) (not shown).
This occluded front passage heralds the arrival of air trajectories more directly from the Antarctic continent (Fig. 5): the residence time in the BL over open water was <72 h and the fetch from the ice edge was <2000 km during the second day (Table 3). We define this as a cold-air outbreak (CAO) period (shown in red in Fig. 5). An open cellular structure still prevailed around the ship during this period, but the cells were smaller (Fig. 4b). The last day, 25 February, saw both a maritime period, and a second brief CAO period (Fig. 5, Table 3), both with similar cloud structures (Fig. 4c). The passage of a warm front shortly after 0000 UTC 26 February (Fig. 3d) ended case 10.
Clouds were shallow during case 10, topping mostly around 2.0 km MSL, and CTTs of −13°C on average (Fig. 6a). They were spaced irregularly and typically far apart, suggesting a low cloud fraction (in the along-ship-track direction), qualitatively consistent with satellite imagery (Fig. 4). The distance scale shown on top of Fig. 6 is indicative only; it assumes the average ship-relative wind speed in the direction of the ship track (0–2 km mean wind, then averaged over 75 h). This suggests that the cloud transects are usually <50 km wide. One exception is a very wide cloud between 2000 and 0000 UTC 23 February, i.e., the tail end of the occluded front mentioned above, with cloud tops just over 3.0 km (Fig. 6a). This wide cloud system will be examined in detail below. Most but not all clouds produce precipitation (defined here as producing echoes reaching the lowest radar data level, ∼280 m MSL), although generally at light rates with an occasional heavier shower: the 90th percentile lowest-radar-gate reflectivity is only 8 dBZ. The light precipitation rate could generally not be estimated from the gauge on the ship, and there were periods of sea spray contamination. Some hydrometeors appear to be lofting (positive HVM), indicative of convection (Fig. 6b) (i.e., updrafts stronger than the particle fall speed). These convective particle ascents are weak and localized. The low Doppler spectral width values (<1 m s−1) suggest that turbulence generally is light (Fig. 6c). We recognize that Doppler spectral width is not a perfect surrogate for air turbulence, as it is affected also by ship motion [section 2d(1)] and by particle fall speed diversity, although the latter is small, as will be discussed below.
Time series of radar measurements and surface measurements during case 10. (a) Reflectivity, (b) hydrometeor vertical motion (HVM), and (c) spectral width from the MWACR profiling radar. White dotted lines in (a)–(c) are isotherms. (d) Surface wind speed and air temperature. (e) BL thermal instability (Me) and EIS. (f) Sensible heat (SH) and latent heat (LH) flux from ERA5. The date in February 2018 is shown below the time at the bottom. Ship latitude is labeled below. The distance at the top of the figure is intended to provide an approximate size of the convective cells and their spacing: time is converted to distance assuming the average ship-relative wind speed in the direction of the ship track (0–2 km mean wind, then averaged over 75 h). The red boxes in (a) indicate the location of the transects in Figs. 9 and 10.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
These clouds occurred in an environment with substantial thermal instability (Me around 3 K, Fig. 6e) and rather high surface sensible heat fluxes (Fig. 6f). Sensible and latent heat fluxes are from ERA5; these estimates are regional scale, and thus more meaningful than those estimated from aboard the ship (using eddy correlation data or a bulk aerodynamic formula). Surface fluxes were stronger when the winds were stronger (during the CAO period on 24 February, Fig. 6d) or when Me was larger (on 25 February). Surface winds were moderate, remaining below 11 m s−1.
During the entire case, and especially on the third day (25 February), θe decreased with height in the lowest 0.5–1.0 km (∼2–4 K km−1), according to 6-hourly rawinsonde data (Fig. 7). The virtual potential temperature (θυ) profiles in this layer are more uniform, especially in the later soundings (Fig. 8), indicating a well-mixed convective BL (Fig. 7). On the first day (23 February), θe was nearly constant with height to at least 3 km MSL (Fig. 7), that is above the prevailing cloud-top height (Fig. 6a). There was virtually no stable layer capping the BL on this day (Figs. 7 and 6e), explaining the variability in echo-top height and CTTs during this period. On the second day, the BL became capped, and on the third day, the cap strengthened (packed isentropes in Fig. 8). The estimated inversion strength (EIS), computed as in Wood and Bretherton (2006), rose steadily to 6 K (Fig. 6e). Toward the end of case 10, a sharp inversion had developed at 2.0 km MSL, on account of warm air advection above 2 km. At the same time, winds weakened, and the surface temperature increased (Fig. 6d) as the ship sailed into a region with higher SST, notwithstanding a brief period of shallow CAO conditions (Fig. 5).
Evolution of the vertical profiles of θe during case 10, based on 6-hourly radiosonde data.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
Time series of virtual potential temperature (K; contours) and along-track wind speed (m s−1; color fill) during case 10. The white line indicates the estimated BL height, based on radiosonde data, and the dashed black line is the lifting condensation level, based on surface meteorological data.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
b. Cloud and precipitation properties: Two cloud transects
1) Example 1
We now zoom into two cloud cells, identified with red boxes in Fig. 6. The first one, shown in Fig. 9, is a ∼150-km-wide cloud system observed between 2000 UTC 23 February and 0000 UTC 24 February. Note that, in the bottom panel (Fig. 9h), the ice fraction is defined as the ratio of the IWP to the total condensed water (IWP+LWP). The retrieved IWP [section 2d(2)] and MWR LWP are shown in Fig. 9g. The MPL lidar depolarization ratio (LDR) is shown in Fig. 9e. In Fig. 9h, the vertically averaged LDR is shown, averaged between the cloud-base height (derived from MPL) and the lidar extinction level or radar cloud top, whichever is lower.
Time series of multisensor measurements of the mesoscale cloud field traversed between 2000 and 2400 UTC 23 Feb. (a) MWACR reflectivity, (b) HVM, (c) spectral width, (d) MPL backscatter power, (e) LDR, (f) retrieved IWC, (g) cloud and precipitation phase from the THERMOCLDPHASE product. White dotted lines in (a)–(g) are isotherms. The red line in (e) is cloud-base height, estimated from MPL. (h) Retrieved IWP and radiometer LWP. (i) Ice fraction and vertically averaged LDR.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
The mesoscale cloud system shown in Fig. 9 includes a midlevel stratiform cloud associated with the tail end of an occluded front (section 3a). The presence of this midlevel stratiform cloud above small BL convective cells (Fig. 9a) may render this example less representative of the entire sample, but the estimated cloud phase (Fig. 9f) and the quantity of liquid (Fig. 9h) are remarkable, given the cold cloud tops as low as −20°C, generally consistent with other recent findings in this region (e.g., Zaremba et al. 2020). Cloud tops tend to be liquid, and there are some mixed-phase sections, but the majority of this cloud is classified as cloud ice (Fig. 9f).
The fingerlike reflectivity structure near cloud top often associated with updrafts and the high spectral width (possibly due to turbulence) near cloud top (Fig. 9c) are reminiscent of cloud-top generating cells, frequently observed at the top of mature stratiform cloud decks, e.g., those associated with occluded frontal systems (Rauber et al. 2015; Rosenow et al. 2014). The leading mechanism explaining these cells is cloud-top radiative cooling, resulting in small pockets of negatively buoyant, turbulent downdrafts surrounded by compensating updrafts (Keeler et al. 2016). Ice particles often initiate in these cloud-top generating cells and have been observed to locally increase reflectivity (Plummer et al. 2015; Alexander et al. 2021). Cloud-top generating cells commonly occur on top of stratiform cloud decks over the Southern Ocean as well (Wang et al. 2020; Zaremba et al. 2020; Alexander et al. 2021). Wang et al. (2020) used a combination of profiling radar and in situ probes to show that these generating cells are rather small and have more ice and liquid than surrounding cloud-top areas. In the case illustrated in Fig. 9, the generating cells are rather small (∼0.5–1.0 km wide) as well, and the reflectivity in the cells remains below −3 dBZ. A closer look of these generating cells between 2215 and 0000 UTC reveals little or no reflectivity enhancement compared to surrounding areas, and if present, such enhancement does not extend far below the cloud-top layer (not shown). Cloud tops generally are classified as liquid, followed by mixed phase, and cloud ice below that (Fig. 9g), but no snow. This suggests limited ice growth in these generating cells, even though the CTT falls within the temperature range listed in Plummer et al. (2015) and Wang et al. (2020) for ice initiation, and even though the relative humidity relative to ice was well above 100% in this layer. This is according to the 2330 UTC radiosonde data, in a layer between 2.9 and 3.4 km MSL (not shown). In sections where the MPL signal is not fully attenuated, strong lidar backscatter is evident in this layer (Fig. 9d), with little signal depolarization (Fig. 9e). The radiosonde balloon penetrated the leading anvil of this cloud system at this time (near x = 130 km, where x is distance shown on top of Fig. 9). The MWACR echoes are exceedingly weak (below −20 dBZ) in this region, yet the LWP is high (up to 500 g m−2), suggesting that the droplets are numerous, but they do not grow large, as evident from the low MWACR reflectivity and limited fallout from this anvil. Note that the MWACR reflectivity may be underestimated, due to attenuation by liquid, especially in this region.
There is ice present in this mesoscale system (Fig. 9f), but the ice fraction generally is below 0.5 (Fig. 9i), notwithstanding the presence of cloud-top generating cells. The primary ice initiation region appears to be the BL convective cells below the midlevel stratiform cloud, where the IWP is the highest (up to 800 g m−2) (Fig. 9h). These convective cells, marked in Fig. 9a, are associated with updraft velocities sufficient to loft hydrometeors (Fig. 9b), relatively high spectral width (Fig. 9c), and surface precipitation (Fig. 9a). The MWACR reflectivity peaks around 15 dBZ in these cells, implying a possibly much higher centimeter-wave radar reflectivity, as illustrated by the KuPR radar data collected on a GPM overpass over the ship some 14 h later (Fig. 1a). Their cloud phase is classified mostly as snow (Fig. 9g).
The midlevel stratiform cloud (best observed after the passage of the shallow convective cells, i.e., after ∼2125 UTC) appears to contain both liquid droplets and ice crystals, given the high radiometer LWP (Fig. 9h), and the relatively high mean LDR, between ∼0.05 and 0.22 (Fig. 9i). Pure liquid water clouds exhibit low LDRs (<0.02) because of the drops’ spherical shape, while pure ice particles tend to have higher values due to their irregular shape (typically 0.30 < LDR < 0.60) (Sassen 2005; Hu et al. 2006). In general, droplets tend to dominate the lidar scattering because they are far more numerous than ice crystals. Also, multiple scattering can increase the LDR some. Mace and Protat (2018) and Lang et al. (2021) used LDR > 0.03 for ice-dominated clouds, and LDR < 0.02 for liquid-dominated clouds over the Southern Ocean. The vertically averaged LDR values generally are above 0.03 in this case, and this variable correlates moderately with the ice fraction (Fig. 9i). The MWACR reflectivity of the midlevel stratiform cloud may appear low for ice and snow (∼−10 to −5 dBZ after 2125 UTC, Fig. 9a), but given a LWP around 400 g m−2, the path-integrated attenuation-corrected W-band reflectivity will be several decibels higher, depending on the vertical distribution of this liquid (e.g., Grasmick et al. 2022).
In summary, while this mesoscale cloud system contains much supercooled liquid and appears not very efficient in converting the liquid to ice, the dominant precipitation growth mechanism involves ice and snow.
2) Example 2
The second example of a mesoscale cloud field (Fig. 10) is more representative in size and depth of the entire case 10 (Fig. 6). This system was observed during a CAO period (Fig. 5) with fairly strong winds (Fig. 6d), and strong along-track wind shear in the lowest 1.5 km along the direction of the ship track (Fig. 8). This system contains a few cells of very high reflectivity (for a W-band radar) with precipitation generally reaching the sea. Three cells can be distinguished, labeled in Fig. 10 as cells 1, 2, and 3. These cells, and especially cell 3, contain no regions of hydrometeor ascent except near cloud top (Fig. 10b), rather high spectral width (Fig. 10c), very high LDR (Figs. 10e,i), and relatively high IWC (Figs. 10f,h). Their cloud phase is classified as snow in the reflectivity cores, surrounded by mixed phase (Fig. 10g), similar to the shallow convective cells in example 1 (Fig. 9g).
As in Fig. 9, but between 1305 and 1345 UTC 24 Feb.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
The question of the origin of the ice requires a description of cloud evolution, which cannot be discerned from these observations alone. Cores of hydrometeor lofting with high spectral width were observed in a few other cells (e.g., around 0800 UTC 23 February, not shown in detail). These are presumably young cells. The cloud structure in Fig. 10 suggests that cell 1 is younger (elevated high-reflectivity core and no precipitation reaching the lowest detectable level), cell 3 older (more rapid hydrometeor descent, highest reflectivity near the lowest level), and the downshear (Fig. 8) anvil on the right (between 16 < x < 32 km) could be the residue of decayed convection. In this interpretation, the ice observed in this anvil (Fig. 10f) would be of convective origin.
c. Composite analysis
The composite vertical structure of the postfrontal shallow convection in case 10 is shown in the form of contoured frequency by altitude diagrams (CFADs; Fig. 11). Surface precipitation is generally light, but a wide range of rates is observed, with reflectivity values near the surface ranging from extremely low values to over +10 dBZ (Fig. 11a). There is a rapid decrease in the frequency of very low reflectivity values in the lower BL, toward the surface, indicating evaporation or sublimation of small hydrometeors below cloud base. Yet the mean reflectivity increases down to low levels (∼600 m MSL, which is the average surface-based lifting condensation level; Fig. 8), suggesting that the larger hydrometeors generally grow down to that level. Many mesoscale cloud systems are largely contained in the upper BL, with no surface precipitation (e.g., right side of example 2, Fig. 10). Most systems are topped at 2.0 km MSL, except for the deeper system in example 1 (Fig. 9).
CFADs of MWACR (a) reflectivity, (b) HVM, (c) spectral width, and (d) retrieved IWC for case 10. Also shown are the profiles of average values (solid lines) and percentiles (dotted lines).
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
We refer to this cloud regime as convection but lofting of hydrometeors is rather rare and weak: just 13% of the CFAD pixels below 2.0 km MSL have positive HVM, and this lofting does not exceed 1 m s−1 (Fig. 11b). The lofting of hydrometeors is most common near cloud top, where the fall speed is lower, and also between 0.5 and 1.0 km height, where convective updrafts may be most intense (as illustrated in Fig. 9b). In general, hydrometeor vertical motions are benign (Fig. 11b) and the spectral width low (Fig. 11c), indicating weak convection. It should be noted that the spectral width distribution is broadened by ship motion [section 2d(1)]. The average downward hydrometeor velocity (bold line in Fig. 11b) is a measure of typical fall speed, since the average air vertical motion is zero for a sample as large as this one (75 h). This average particle fall speed increases only slightly toward sea level, to −0.8 m s−1 at the lowest radar level, indicative of unrimed snow or drizzle. Downward hydrometeor velocities exceeding 2 m s−1 are very rare, even near the surface, indicating that the high reflectivity occurrences (say, >10 dBZ, Fig. 11a) are not due to raindrops, but rather snow (possibly rimed snow). Ice, when present, is found in small mass concentrations at all levels in these clouds, not just in the anvils (Fig. 11d). A few profiles have much more ice; thus, the mean IWC is far larger than the mode.
Small IWP values are commonly encountered: 82% of all profiles with any ice have IWP values in the first bin (10–50 g m−2) (Fig. 12). High IWP values are rare: just 10% of the profiles have IWP values over 400 g m−2. High LWP values are more common; in fact, LWP values in the range of 100–500 g m−2 are an order of magnitude more common IWP values in that range (Fig. 12).
Frequency distribution of LWP and IWP in case 10. Only occurrences above threshold values are counted (LWP > 20 g m−2 and IWP > 10 g m−2). Note the logarithmic frequency scale.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
We now examine the frequency of the dominant cloud and precipitation phase in three temperature layers: from the freezing level (or sea level, whichever is lower) to −5°C layer, the −5° to −10°C layer, and the −10°C to cloud-top layer (Table 4). The bottom layer is most important in terms of precipitation growth processes because it typically contains more precipitation particles, whereas the top layer typically contains more cloud particles (water or ice).
Frequency (%) of the dominant THEROCLDPHASE phase during MARCUS. Three layers are distinguished. N is the total number of columns in each layer (excluding clear sky). The phase “cloud liquid” is denoted as CLIQ, cloud ice as CICE, mixed phase as MIXP, and unknown as UNKN. RAIN stands for the phases rain, drizzle, and liquid + drizzle combined. Percentages in each row add up to 100.
In the upper cloud layer (−10°C to echo top), cloud ice dominates, followed by cloud water, in case 10 and in all four cases (Table 4). The main change down toward the freezing level (−5° to 0°C) is the increased frequency of snow (11%) and mixed-phase hydrometeors (23% in case 10). Many of the cloud profiles are not precipitating, as evident from Figs. 9 and 10, and from the number of columns in each of the three layers (Table 4). We define “liquid dominated” those profiles that have mostly (>50%) CLIQ or RAIN in the 0° to −5°C layer, and “ice dominated” those profiles that have mostly CICE, MIXP, or SNOW in the same layer. In case 10, 37% of the profiles are liquid dominated, and 63% are ice dominated. For all four cases combined, the ice-dominated fraction increases slightly (66%). While cloud droplets and high LWP values are common (Fig. 12, Table 4), the warm cloud growth process (collision–coalescence) does not dominate precipitation production: fewer than 2% of the profiles are dominated by drizzle/rain above the freezing level.
We now compare the vertical cloud structure for ice- versus liquid-dominated precipitation profiles (Fig. 13). Again, we use the 0° to −5°C layer only, since it contains proportionally more precipitation particles and we are interested in the dominant precipitation growth mechanism. Reflectivity is 2–5 dB higher for the ice-dominated profiles, compared to the liquid-dominated profiles, for case 10, and for all four cases examined here (Fig. 13a). Reflectivity is dominated by the largest scatterers, so this is an indication that the upper-size ice particles are larger than the upper-size liquid drops. The hydrometeors in the liquid-dominated profiles mostly are drizzle-size drops, given the typically low reflectivity (Fig. 13a) and low fall speed (Fig. 13b): a mean fall speed of 1.1 m s−1 (observed at the lowest radar level) suggests a typical drizzle diameter of 300–400 μm (Foote and du Toit 1969; Yu et al. 2016). Ice particles do not fall out faster than drizzle below 1 km MSL on average (Fig. 13b): with an average fall speed of up to 1.0 m s−1 at the lowest radar level, the ice particles probably are hardly rimed. The small average difference in HVM and spectral width (Fig. 13c) indicates that HVM cannot be used as an indicator for hydrometeor phase at any level.
Contrast between liquid-dominated (blue) and ice-dominated (red) profiles, in terms of MWACR (a) reflectivity, (b) HVM, (c) spectral width, and (d) retrieved IWC for all 4 cases (108 h). The solid lines are averages, and the dashed lines are the 10th and 90th percentiles.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
The spread in HVM, shown in terms of its standard deviation, is small in either cloud phase, generally less than 2.0 m s−1 (spread between the 10th- and 90th-percentile lines in Fig. 13b). In general, a large spread in vertical velocity and a large spectral width are indicative of convective motions. Neither are significantly higher in the ice-dominated profiles (Figs. 13b,c). In other words, the presence of ice in these shallow postfrontal clouds cannot be attributed to convective activity. The relations shown here are simultaneous; our argument would be stronger if we had data to relate convective motions to time-lagged ice presence.
Ice-dominated profiles are proportionally more common during the two so-called CAO periods in this case study (Fig. 5), compared to the two maritime periods (71% versus 55%), even though cloud tops tend to be slightly shallower (not shown). The higher fraction of ice profiles during CAO periods is consistent with a higher average reflectivity and faster downward HVM in the lowest 1 km, compared to the maritime periods (not shown).
4. Characteristics of postfrontal shallow clouds in MARCUS
Compared to case 10, the three other cases with adequate quality data (9, 10, and 12) are much shorter (Table 2), just 33 h combined. On average, these three cases have slightly less thermal instability (lower Me, Table 2), lower reflectivity below 1.2 km with most profiles echo-free at lower level (Table 4), and proportionally more cloud ice and snow, especially near cloud top (Table 4).
While this cloud regime generally is referred to as mesoscale cellular convection (e.g., Wood and Hartmann 2006; McCoy et al. 2017; Lang et al. 2021), perusal of all MARCUS time–height transects (similar to Figs. 9 and 10) suggests that some cloud systems (or parts thereof) appear more convective, while others appear more layered (stratiform). Take the example in Fig. 10, for instance: the left portion (0 < x < 16 km) appears convective, in terms of isolation and magnitude of the reflectivity, while the right portion (16 < x < 34 km) appears stratiform. We classify all cloud cells objectively, following basic principles of isolation and intensity (e.g., Steiner et al. 1995) applied to time–height transects of radar reflectivity. Convection requires a core at least 10 dB stronger than the cloud-mean reflectivity, and a peak intensity of at least +5 dBZ. This partitioning yields 99 h of convective clouds and 9 h of stratiform clouds in all four cases combined. Not surprisingly, convective clouds are characterized by a much higher mean reflectivity than stratiform clouds (Fig. 14a). They also have more intense vertical drafts (10th and 90th percentiles in Fig. 14b), and higher spectral width (Fig. 14c), but less ice than stratiform clouds (when ice is present).
As in Fig. 13, but contrasting convective (red) against stratiform (blue) cloud regions.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
A well-established feature of the Southern Ocean is the Antarctic polar front. As mentioned in the introduction, Mace et al. (2021a) found ice to be relatively more common in postfrontal shallow convection equatorward of this front. Most of the cold-sector shallow clouds in our 108-h-long MARCUS dataset occurred poleward of the climatological (monthly mean) location of this oceanic front (Freeman and Lovenduski 2016). The latitude of this front ranged from 51° to 55°S, depending on the cruise. The sample collected equatorward of the polar front, about 12 h, is probably too small for the comparison in Fig. 15 (between high and low latitudes) to be statistically significant. Mesoscale cellular convection equatorward of the front tends to be deeper (Fig. 15a), have a larger spectral width (Fig. 15c), slightly stronger vertical drafts (dotted lines in Fig. 15b), and higher particle fall speed near the surface (Fig. 15b), yet less ice at low levels (Fig. 15d), and thus probably more drizzle precipitation, than those poleward of the front. In short, the small MARCUS dataset is unable to confirm the satellite-based cloud-phase climatology in Mace et al. (2021a).
As in Fig. 13, but contrasting clouds equatorward (red) and poleward (blue) of the oceanic Antarctic polar front.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
5. Discussion
We have shown that postfrontal mesoscale cloud fields in the high-latitude Southern Ocean contain much supercooled liquid water, and that according to the THERMOCLDPHASE, precipitation growth mostly is driven by cold-cloud processes in the presence of ice. The majority (66%) of individual cloud profiles is classified as ice dominated (i.e., cloud ice, snow, or mixed phase) in the −5° to 0°C layer (Table 4). Precipitation tends to be quite light, and a large fraction of the cloud area is not precipitating at the surface. The dominance of cloud ice at all levels may disagree with in situ cloud observations in D’Alessandro et al. (2021), who find mostly supercooled liquid between −20° and 0°C in summertime shallow clouds over the Southern Ocean, although their observations are mostly farther north. We find cloud liquid to be usually dominant near cloud top, even though CTTs usually are below −10°C (e.g., Figs. 9 and 11), consistent with Zaremba et al. (2020) over the Southern Ocean and with observations elsewhere (e.g., Rauber and Tokay 1991).
The analysis has not uncovered clear environmental or cloud characteristics that control the abundance of ice in these mostly liquid clouds. In general, IWC is controlled by primary ice initiation, secondary ice production (ice multiplication), and the various ice growth processes. Ice multiplication processes (such as rime splintering) are likely to occur as long as primary ice is present in these clouds, given the relative abundance of liquid water and the observed temperature range (Hallett and Mossop 1974). One limiting factor may be the apparent scarcity of drizzle (Table 4) (Field et al. 2017; Luke et al. 2021).
Given the presence of both liquid-dominated and ice-dominated regions in these cloud systems, we now address the question of ice initiation and growth mechanisms in these clouds. Recall that the seeder-feeder mechanism (high clouds dropping ice crystals into these shallow clouds) was eliminated a priori in the selection of suitable cases. The MWACR radar is sufficiently sensitive to detect even small ice crystals falling into the shallow clouds, although the signal may be attenuated in some regions of high LWC in the shallow clouds. We examine three factors: CTT, strong vertical drafts, and ice nucleating particles (INPs).
Cloud-top temperature (i.e., the lowest temperature in these relatively well-mixed clouds) is considered first, because ice initiation parameterizations primarily are temperature dependent (e.g., Fletcher 1962; Cooper 1986; Meyers et al. 1992). Figure 16a shows the distribution of cloud profiles as a function of local CTT, which often is higher than the lowest CTT of an entire cloud (as is evident, e.g., in Fig. 9 or Fig. 10). So, while the case selection assumed a −10°C entire-cloud CTT threshold [section 2d(4)], locally the CTT can be higher. We define the ice profile fraction as the fraction of cloud profiles classified as ice dominated (cloud ice, mixed phase, or snow) in the −5° to 0°C layer. The ice profile fraction increases with decreasing CTT (Fig. 16a), consistent with in situ aircraft data over the Southern Ocean (Fig. 4 in D’Alesandro et al. 2021). The local cloud depth (vertical distance between MPL-determined cloud base and MWACR-determined cloud top) (Fig. 16b) is another good predictor of ice dominance: ice growth processes are more important in these cold-sector marine clouds when they are deeper. These findings are consistent with other studies (e.g., Mace et al. 2021a).
Distribution of cloud profiles (blue) during all four MARCUS cases as a function of (a) CTT, (b) cloud depth, (c) vertically averaged spectral width (1-min averaged), (d) maximum HVM (low-pass filtered to exclude wavelengths < 30 s and < 500 m vertical) in the vicinity (within 500 s), (e) Me values, (f) sensible heat flux, and (g) liquid water path. The connected red markers in each panel indicate, for each bin, the fraction of cloud profiles classified as ice dominated in the −5° to 0°C layer.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
Second, ice may be generated in convective updrafts, due to high supersaturation (e.g., Lawson et al. 2015), as well as in turbulent cloud-top generating cells (e.g., Plummer et al. 2015). Therefore, we examine the maximum HVM (Fig. 16d), LWP (Fig. 16g), and also spectral width (Fig. 16c), which, to a first order, we treat as a measure of turbulence intensity. [Spectral width can be a measure of hydrometeor fall speed diversity as well, but large fall speed values are extremely rare (Fig. 11b).] Ice initiation and snow growth take time; thus, IWC may not correlate with the instantaneous updraft speed. Rather, a cloud region of high IWC (such as cell 3 in Fig. 10f) may be the result of an earlier convective updraft. Lack of time-resolved data compels us to look for convective activity in the vicinity (we use 500 s, or roughly 5 km) (Fig. 16d).
The MARCUS dataset does not provide evidence that ice is more common near strong updrafts (or more correctly, pockets of hydrometeor ascent) (Fig. 16d), in areas with high LWP (Fig. 16g), nor in areas of high spectral width (Fig. 16c). Environmental factors that influence the likelihood or intensity of shallow convection, i.e., Me (Fig. 16e) and surface sensible heat fluxes (derived from ERA5) (Fig. 16f), are not good predictors of the fraction of profiles classified as ice dominated either.
That leaves us with the INP concentration. Open-faced aspirated filters were exposed to the ambient air for 23–50-h time periods in MARCUS (McFarquhar et al. 2021, supplement). The immersion freezing INP concentration as a function of temperature was inferred from the collected filters using an ice spectrometer (e.g., McCluskey et al. 2018). Filters flagged as possibly contaminated by ship exhaust are not used. INP concentrations are generally quite low in the postfrontal BL over Southern Ocean (Fig. 17) compared to other oceans (McCluskey et al. 2018), and they were very low (below 1 m−3) in the cases examined here. Given the long duration of the INP samples, many samples are severely “contaminated” by air outside the postfrontal cold-sector environment, as shown in Fig. 17. Nevertheless, the few available samples suggest that the fraction of ice-dominated cloud profiles increases with increasing INP concentrations (Fig. 17).
Scatterplot of INP number concentration against the fraction of cloud profiles classified as ice dominated (in the −5° to 0°C layer) for individual INP filters in all available cases. For each case, the case number (refer to Table 2), the number of hours in the case with INP sampling, and the percentage of the INP filter exposure time that fell during the case are shown. Each point is color coded by its 90th-percentile CTT.
Citation: Journal of the Atmospheric Sciences 80, 5; 10.1175/JAS-D-21-0243.1
6. Conclusions
This observational study examines the vertical structure of postfrontal (cold-sector), shallow mesoscale cloud systems as observed with profiling W-band radar, polarization lidar and microwave radiometer during the shipborne MARCUS campaign. Satellite imagery indicates that an open-cellular cloud structure is common. Four cases with good data are identified, 108 h in total, mostly poleward of 52°S and the Antarctic polar front. Case selection is based on thermal instability relative to the underlying sea surface (Me > 0), radar-defined cloud-top height (below 3.0 km MSL, with no ice particles falling from aloft), and cloud-scale CTT (<−10°C). Back-trajectories indicate that the sampled air masses typically had at least 3 days (3000 km) since emerging from the Antarctic ice dome, so they were generally well adjusted to the underlying sea surface, with surface sensible and latent heat fluxes generally positive but below 100 W m−2.
The BL clouds in these four cases generally top out at ∼2.0 km above MSL, have local CTTs around −20° to −6°C, are precipitating, and have a surface air temperature ranging between +1° and +5°C. Precipitation rates generally are light, but a small fraction (0.7%) of the profiles have near-surface reflectivity values over 15.0 dBZ, which approaches the upper limit of W-band reflectivity. One exceptionally long (75 h) case is examined in detail. The cloud and precipitation phase is obtained from the THERMOCLDPHASE product. The maximum IWC is estimated from the MWACR reflectivity and hydrometeor vertical velocity, assuming that the radar echoes are due to ice crystals. Our main findings are as follows:
-
The majority of cloud profiles (66% for the four cases combined) is classified as ice dominated (cloud ice, mixed phase, or snow) in the −5° to 0°C layer.
-
Convective vertical motions are quite weak, with only ∼13% of the hydrometeors carried upward.
-
Most clouds are precipitating lightly, although only over a small fraction of the cloud area, with a mean radar reflectivity increasing downward and peaking near cloud base at rather low values, about +2 dBZ for the liquid-dominated profiles (indicating mostly drizzle), and ∼+5 dBZ for the ice-dominated cloud profiles. Hydrometeors fall out at an average speed of just 1 m s−1 near cloud base, with little variation.
-
Cloud-top temperature, cloud depth, and INP concentration are good indicators of ice presence.
-
Measures of convective activity, such as maximum proximity updraft strength, turbulence intensity, and environmental thermal instability parameters, appear to be poor indicators for ice presence and ice-mediated precipitation growth. Cloud-top generating cells are found on top of many cloud systems, but they generally do not trigger ice crystal growth to the extent observed elsewhere. The fact that convective activity and turbulence are poor indicators for ice presence in these clouds may be attributable to the very low INP concentration, generally below 1 m−3.
In Part II (Hu et al. 2023), the main case examined here (10) is numerically simulated, to assess how well the macroscale and vertical structure can be reproduced, and how well various cloud parameterizations capture the observed cloud-phase distribution.
Acknowledgments.
This work was supported by Department of Energy Atmospheric Systems Research Grant DE-SC0018927. Thanks to Saisai Ding, Greg M. McFarquhar, and Roger Marchand for the production of the back-trajectories in MARCUS (EPVAP v.1.41). Thank you to Roger Marchand for insightful discussions contributing to this manuscript, to Paul DeMott and Thomas Hill for their insight into the INP data collected in MARCUS, to Gerald Mace for pointing to the GPM KuPR data, and to three anonymous reviewers for their in-depth reviews.
Data availability statement.
The data used in this study were obtained from the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Data Archive at https://adc.arm.gov/. The ERA5 data were obtained from https://cds.climate.copernicus.eu, the SST data from https://psl.noaa.gov, and the MODIS data from https://search.earthdata.nasa.gov.
REFERENCES
Ahn, E., Y. Huang, T. H. Chubb, D. Baumgardner, P. Isaac, M. de Hoog, S. T. Siems, and M. J. Manton, 2017: In situ observations of wintertime low-altitude clouds over the Southern Ocean. Quart. J. Roy. Meteor. Soc., 143, 1381–1394, https://doi.org/10.1002/qj.3011.
Ahn, E., Y. Huang, S. T. Siems, and M. J. Manton, 2018: A comparison of cloud microphysical properties derived from MODIS and CALIPSO with in situ measurements over the wintertime Southern Ocean. J. Geophys. Res. Atmos., 123, 11 120–11 140, https://doi.org/10.1029/2018JD028535.
Alexander, S. P., G. M. McFarquhar, R. Marchand, A. Protat, É. Vignon, G. G. Mace, and A. R. Klekociuk, 2021: Mixed-phase clouds and precipitation in Southern Ocean cyclones and cloud systems observed poleward of 64°S by ship-based cloud radar and lidar. J. Geophys. Res. Atmos., 126, e2020JD033626, https://doi.org/10.1029/2020JD033626.
Atlas, R. L., C. S. Bretherton, P. N. Blossey, A. Gettelman, C. Bardeen, P. Lin, and Y. Ming, 2020: How well do large-eddy simulations and global climate models represent observed boundary layer structures and low clouds over the summertime Southern Ocean? J. Adv. Model. Earth Syst., 12, e2020MS002205, https://doi.org/10.1029/2020MS002205.
Bodas-Salcedo, A., and Coauthors, 2014: Origins of the solar radiation biases over the Southern Ocean in CFMIP2 models. J. Climate, 27, 41–56, https://doi.org/10.1175/JCLI-D-13-00169.1.
Choi, Y.-S., C.-H. Ho, S.-W. Kim, and R. S. Lindzen, 2010: Observational diagnosis of cloud phase in the winter Antarctic atmosphere for parameterizations in climate models. Adv. Atmos. Sci., 27, 1233–1245, https://doi.org/10.1007/s00376-010-9175-3.
Cooper, W. A., 1986: Ice initiation in natural clouds. Precipitation Enhancement: A Scientific Challenge, Meteor. Monogr., No. 43, Amer. Meteor. Soc., 29–32, https://doi.org/10.1175/0065-9401-21.43.29.
Crewell, S., and U. Löhnert, 2003: Accuracy of cloud liquid water path from ground-based microwave radiometry 2. Sensor accuracy and synergy. Radio Sci., 38, 8042, https://doi.org/10.1029/2002RS002634.
Crosier, J., and Coauthors, 2011: Observations of ice multiplication in a weakly convective cell embedded in supercooled mid-level stratus. Atmos. Chem. Phys., 11, 257–273, https://doi.org/10.5194/acp-11-257-2011.
D’Alessandro, J. J., M. Diao, C. Wu, X. Liu, J. B. Jensen, and B. B. Stephens, 2019: Cloud phase and relative humidity distributions over the Southern Ocean in austral summer based on in situ observations and CAM5 simulations. J. Climate, 32, 2781–2805, https://doi.org/10.1175/JCLI-D-18-0232.1.
D’Alessandro, J. J., G. M. McFarquhar, W. Wu, J. L. Stith, J. B. Jensen, and R. M. Rauber, 2021: Characterizing the occurrence and spatial heterogeneity of liquid, ice and mixed phase low-level clouds over the Southern Ocean using in situ observations acquired during SOCRATES. J. Geophys. Res. Atmos., 126, e2020JD034482, https://doi.org/10.1029/2020JD034482.
Deng, M., and G. G. Mace, 2006: Cirrus microphysical properties and air motion statistics using cloud radar Doppler moments. Part I: Algorithm description. J. Appl. Meteor. Climatol., 45, 1690–1709, https://doi.org/10.1175/JAM2433.1.
Deng, M., J. French, B. Geerts, S. Haimov, L. Oolman, D. Plummer, and Z. Wang, 2022: Retrieval and evaluation of ice water content from an airborne cloud radar in orographic wintertime clouds during SNOWIE. J. Atmos. Oceanic Technol., 39, 207–221, https://doi.org/10.1175/JTECH-D-21-0085.1.
Eloranta, E. W., 2005: High spectral resolution lidar. Lidar: Range-Resolved Optical Remote Sensing of the Atmosphere, C. Weitkamp, Ed., Springer Series in Optical Sciences, Vol. 102, Springer, 143–163.
Field, P. R., and Coauthors, 2017: Secondary ice production: Current state of the science and recommendations for the future. Ice Formation and Evolution in Clouds and Precipitation: Measurement and Modeling Challenges, Meteor. Monogr., No. 58, Amer. Meteor. Soc., https://doi.org/10.1175/AMSMONOGRAPHS-D-16-0014.1.
Fletcher, J. K., S. Mason, and C. Jakob, 2016: A climatology of clouds in marine cold air outbreaks in both hemispheres. J. Climate, 29, 6677–6692, https://doi.org/10.1175/JCLI-D-15-0783.1.
Fletcher, N. H., 1962: Physics of Rain Clouds. Cambridge University Press, 386 pp.
Foote, G. B., and P. S. du Toit, 1969: Terminal velocity of raindrops aloft. J. Appl. Meteor., 8, 249–253, https://doi.org/10.1175/1520-0450(1969)008<0249:TVORA>2.0.CO;2.
Freeman, N. M., and N. S. Lovenduski, 2016: Mapping the Antarctic polar front: Weekly realizations from 2002 to 2014. Earth Syst. Sci. Data, 8, 191–198, https://doi.org/10.5194/essd-8-191-2016.
Grasmick, C., B. Geerts, J. R. French, S. Haimov, and R. M. Rauber, 2022: Estimating microphysics properties in ice-dominated clouds from airborne Ka–W-band dual-wavelength ratio reflectivity in close proximity to in situ probes. J. Atmos. Oceanic Technol., 39, 1815–1833, https://doi.org/10.1175/JTECH-D-21-0147.1.
Hallett, J., and S. C. Mossop, 1974: Production of secondary ice particles during the riming process. Nature, 249, 26–28 https://doi.org/10.1038/249026a0.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Heymsfield, A. J., and Coauthors, 2008: Testing IWC retrieval methods using radar and ancillary measurements with in situ data. J. Appl. Meteor. Climatol., 47, 135–163, https://doi.org/10.1175/2007JAMC1606.1.
Heymsfield, A. J., S. Y. Matrosov, and N. B. Wood, 2016: Toward improving ice water content and snow-rate retrievals from radars. Part I: X and W bands, emphasizing CloudSat. J. Appl. Meteor. Climatol., 55, 2063–2090, https://doi.org/10.1175/JAMC-D-15-0290.1.
Hu, Y., Z. Liu, D. Winker, M. Vaughan, V. Noel, L. Bissonnette, G. Roy, and M. McGill, 2006: Simple relationship between lidar multiple scattering and depolarization for water clouds. Opt. Lett., 31, 1809–1811, https://doi.org/10.1364/OL.31.001809.
Hu, Y., S. Rodier, K. Xu, W. Sun, J. Huang, B. Lin, P. Zhai, and D. Josset, 2010: Occurrence, liquid water content, and fraction of supercooled water clouds from combined CALIOP/IIR/MODIS measurements. J. Geophys. Res., 115, D00H34, https://doi.org/10.1029/2009JD012384.
Hu, Y., Z. J. Lebo, B. Geerts, Y. Hu, and Y. Wang, 2023: Vertical structure and ice production processes of shallow convective postfrontal clouds over the Southern Ocean in MARCUS. Part II: Modeling study. J. Atmos. Sci., 80, 1307–1327, https://doi.org/10.1175/JAS-D-21-0272.1.
Huang, B., C. Liu, V. Banzon, E. Freeman, G. Graham, B. Hankins, T. Smith, and H.-M. Zhang, 2021: Improvements of the Daily Optimum Interpolation Sea Surface Temperature (DOISST) version 2.1. J. Climate, 34, 2923–2939, https://doi.org/10.1175/JCLI-D-20-0166.1.
Huang, Y., 2017: Observations and simulations of cloud thermodynamic phase over the Southern Ocean. Ph.D. thesis, Monash University, 179 pp., https://doi.org/10.4225/03/58b4eb811b885.
Huang, Y., T. Chubb, D. Baumgardner, M. deHoog, S. T. Siems, and M. J. Manton, 2017: Evidence for secondary ice production in Southern Ocean open cellular convection. Quart. J. Roy. Meteor. Soc., 143, 1685–1703, https://doi.org/10.1002/qj.3041.
Jensen, M., S. Giangrande, T. Fairless, and A. Zhou, 2017: Interpolated sonde (INTERPOLATEDSONDE): ARM Mobile Facility (MAR) Hobart, AUS to Antarctic coast—Resupply ship Aurora Australis; AMF2 (M1). Subset used: 31 October 2017–24 March 2018. ARM Data Center, accessed 10 January 2020, https://doi.org/10.5439/1095316.
Keeler, J. M., B. F. Jewett, R. M. Rauber, G. M. McFarquhar, R. M. Rasmussen, L. Xue, C. Liu, and G. Thompson, 2016: Dynamics of cloud-top generating cells in winter cyclones. Part II: Radiative and instability forcing. J. Atmos. Sci., 73, 1529–1553, https://doi.org/10.1175/JAS-D-15-0127.1.
Kollias, P., B. P. Treserras, and A. Protat, 2019: Calibration of the 2007–2017 record of atmospheric radiation measurements cloud radar observations using CloudSat. Atmos. Meas. Tech., 12, 4949–4964, https://doi.org/10.5194/amt-12-4949-2019.
Kolstad, E. W., T. J. Bracegirdle, and I. A. Seierstad, 2009: Marine cold-air outbreaks in the North Atlantic: Temporal distribution and associations with large-scale atmospheric circulation. Climate Dyn., 33, 187–197, https://doi.org/10.1007/s00382-008-0431-5.
Korolev, A., and Coauthors, 2017: Mixed-phase clouds: Progress and challenges. Ice Formation and Evolution in Clouds and Precipitation: Measurement and Modeling Challenges, Meteor. Monogr., No. 58, Amer. Meteor. Soc., https://doi.org/10.1175/AMSMONOGRAPHS-D-17-0001.1.
Kyrouac, J., and S. Springston, 2017: Meteorological measurements associated with the Aerosol Observing System (AOSMET): ARM Mobile Facility (MAR) Hobart, AUS to Antarctic coast—Resupply ship Aurora Australis; AMF2 (M1). Subset used: 29 October 2017–24 March 2018. ARM Data Center, accessed 10 January 2020, https://doi.org/10.5439/1025153.
Lang, F., Y. Huang, A. Protat, S. C. H. Truong, S. T. Siems, and M. J. Manton, 2021: Shallow convection and precipitation over the Southern Ocean: A case study during the CAPRICORN 2016 field campaign. J. Geophys. Res. Atmos., 126, e2020JD034088, https://doi.org/10.1029/2020JD034088.
Lawson, R. P., S. Woods, and H. Morrison, 2015: The microphysics of ice and precipitation development in tropical cumulus clouds. J. Atmos. Sci., 72, 2429–2445, https://doi.org/10.1175/JAS-D-14-0274.1.
Lenaerts, J. T. M., K. Van Tricht, S. Lhermitte, and T. S. L’Ecuyer, 2017: Polar clouds and radiation in satellite observations, reanalyses, and climate models. Geophys. Res. Lett., 44, 3355–3364, https://doi.org/10.1002/2016GL072242.
Liao, L., and R. Meneghini, 2022: GPM DPR retrievals: Algorithm, evaluation, and validation. Remote Sens., 14, 843, https://doi.org/10.3390/rs14040843.
Luke, E. P., F. Yang, P. Kollias, A. M. Vogelmann, and M. Maahn, 2021: New insights into ice multiplication using remote-sensing observations of slightly supercooled mixed-phase clouds in the Arctic. Proc. Natl. Acad. Sci. USA, 118, e2021387118, https://doi.org/10.1073/pnas.2021387118.
Mace, G. G., and A. Protat, 2018: Clouds over the Southern Ocean as observed from the R/V Investigator during CAPRICORN. Part I: Cloud occurrence and phase partitioning. J. Appl. Meteor. Climatol., 57, 1783–1803, https://doi.org/10.1175/JAMC-D-17-0194.1.
Mace, G. G., Q. Zhang, M. Vaughan, R. Marchand, G. Stephens, C. Trepte, and D. Winker, 2009: A description of hydrometeor layer occurrence statistics derived from the first year of merged CloudSat and CALIPSO data. J. Geophys. Res., 114, D00A26, https://doi.org/10.1029/2007JD009755.
Mace, G. G., A. Protat, and S. Benson, 2021a: Mixed-phase clouds over the Southern Ocean as observed from satellite and surface based lidar and radar. J. Geophys. Res. Atmos., 126, e2021JD034569, https://doi.org/10.1029/2021JD034569.
Mace, G. G., and Coauthors, 2021b: Southern Ocean cloud properties derived from CAPRICORN and MARCUS data. J. Geophys. Res. Atmos., 126, e2020JD033368, https://doi.org/10.1029/2020JD033368.
Matrosov, S. Y., and A. J. Heymsfield, 2008: Estimating ice content and extinction in precipitating cloud systems from CloudSat radar measurements. J. Geophys. Res., 113, D00A05, https://doi.org/10.1029/2007JD009633.
McCluskey, C. S., and Coauthors, 2018: Observations of ice nucleating particles over Southern Ocean waters. Geophys. Res. Lett., 45, 11 989–11 997, https://doi.org/10.1029/2018GL079981.
McCoy, D. T., D. L. Hartmann, and D. P. Grosvenor, 2014: Observed Southern Ocean cloud properties and shortwave reflection. Part II: Phase changes and low cloud feedback. J. Climate, 27, 8858–8868, https://doi.org/10.1175/JCLI-D-14-00288.1.
McCoy, I. L., R. Wood, and J. K. Fletcher, 2017: Identifying meteorological controls on open and closed mesoscale cellular convection associated with marine cold air outbreaks. J. Geophys. Res. Atmos., 122, 11 678–11 702, https://doi.org/10.1002/2017JD027031.
McFarquhar, G., and Coauthors, 2021: Observations of clouds, aerosols, precipitation, and surface radiation over the Southern Ocean: An overview of CAPRICORN, MARCUS, MICRE, and SOCRATES. Bull. Amer. Meteor. Soc., 102, E894–E928, https://doi.org/10.1175/BAMS-D-20-0132.1.
Meyers, M. P., P. J. DeMott, and W. R. Cotton, 1992: New primary ice-nucleation parameterizations in an explicit cloud model. J. Appl. Meteor., 31, 708–721, https://doi.org/10.1175/1520-0450(1992)031<0708:NPINPI>2.0.CO;2.
Morrison, A. E., S. T. Siems, and M. J. Manton, 2011: A three-year climatology of cloud-top phase over the Southern Ocean and North Pacific. J. Climate, 24, 2405–2418, https://doi.org/10.1175/2010JCLI3842.1.
Morrison, H., and Coauthors, 2011: Intercomparison of cloud model simulations of Arctic mixed-phase boundary layer clouds observed during SHEBA/FIRE-ACE. J. Adv. Model. Earth Syst., 3, M05001, https://doi.org/10.1029/2011MS000066.
Muhlbauer, A., I. L. McCoy, and R. Wood, 2014: Climatology of stratocumulus cloud morphologies: Microphysical properties and radiative effects. Atmos. Chem. Phys., 14, 6695–6716, https://doi.org/10.5194/acp-14-6695-2014.
Naud, C. M., J. F. Booth, K. Lamer, R. Marchand, A. Protat, and G. M. McFarquhar, 2020: On the relationship between the marine cold air outbreak M parameter and low-level cloud heights in the midlatitudes. J. Geophys. Res. Atmos., 125, e2020JD032465, https://doi.org/10.1029/2020JD032465.
Plummer, D. M., G. M. McFarquhar, R. M. Rauber, B. F. Jewett, and D. C. Leon, 2015: Microphysical properties of convectively generated fall streaks within the stratiform comma head region of continental winter cyclones. J. Atmos. Sci., 72, 2465–2483, https://doi.org/10.1175/JAS-D-14-0354.1.
Rangno, A. L., and P. V. Hobbs, 1991: Ice particle concentrations and precipitation development in small polar maritime cumuliform clouds. Quart. J. Roy. Meteor. Soc., 117, 207–241, https://doi.org/10.1002/qj.49711749710.
Rauber, R. M., and A. Tokay, 1991: An explanation for the existence of supercooled water at the top of cold clouds. J. Atmos. Sci., 48, 1005–1023, https://doi.org/10.1175/1520-0469(1991)048<1005:AEFTEO>2.0.CO;2.
Rauber, R. M., and Coauthors, 2015: The role of cloud-top generating cells and boundary layer circulations in the finescale radar structure of a winter cyclone over the Great Lakes. Mon. Wea. Rev., 143, 2291–2318, https://doi.org/10.1175/MWR-D-14-00350.1.
Rosenow, A. A., D. M. Plummer, R. M. Rauber, G. M. McFarquhar, B. F. Jewett, and D. Leon, 2014: Vertical velocity and physical structure of generating cells and convection in the comma head region of continental winter cyclones. J. Atmos. Sci., 71, 1538–1558, https://doi.org/10.1175/JAS-D-13-0249.1.
Sassen, K., 2005: Polarization in lidar. Lidar: Range-Resolved Optical Remote Sensing of the Atmosphere, C. Weitkamp, Ed., Springer Series in Optical Sciences, Vol. 102, Springer, 19–42.
Shupe, M. D., 2007: A ground-based multisensor cloud phase classifier. Geophys. Res. Lett., 34, L22809, https://doi.org/10.1029/2007GL031008.
Shupe, M. D., P. Kollias, S. Y. Matrosov, and T. L. Schneider, 2004: Deriving mixed-phase cloud properties from Doppler radar spectra. J. Atmos. Oceanic Technol., 21, 660–670, https://doi.org/10.1175/1520-0426(2004)021<0660:DMCPFD>2.0.CO;2.
Shupe, M. D., S. Y. Matrosov, and T. Uttal, 2006: Arctic mixed-phase cloud properties derived from surface-based sensors at SHEBA. J. Atmos. Sci., 63, 697–711, https://doi.org/10.1175/JAS3659.1.
Sivaraman, C., D. Flynn, L. Riihimaki, and J. Comstock, 2020: Cloud mask from micropulse lidar (30SMPLCMASK1ZWANG): ARM Mobile Facility (ANX) Andenes, Norway; AMF1 (main site for COMBLE) (M1). Subset used: 11 February–1 June 2020. ARM Data Center, accessed 1 October 2020, https://doi.org/10.5439/1508389.
Steiner, M., R. A. Houze Jr., and S. E. Yuter, 1995: Climatological characterization of three-dimensional storm structure from operational radar and rain gauge data. J. Appl. Meteor., 34, 1978–2007, https://doi.org/10.1175/1520-0450(1995)034<1978:CCOTDS>2.0.CO;2.
Trenberth, K. E., and J. T. Fasullo, 2010: Simulation of present-day and twenty-first-century energy budgets of the Southern Oceans. J. Climate, 23, 440–454, https://doi.org/10.1175/2009JCLI3152.1.
Wang, M., and K. Johnson, 2017: Marine W-band ARM Cloud Radar, Active Remote Sensing of Clouds (ARSCLWACR1KOLLIASSHP). ARM, https://doi.org/10.5439/1498736.
Wang, Y., and Coauthors, 2020: Microphysical properties of generating cells over the Southern Ocean: Results from SOCRATES. J. Geophys. Res. Atmos., 125, e2019JD032237, https://doi.org/10.1029/2019JD032237.
Westwater, E. R., Y. Han, M. D. Shupe, and S. Y. Matrosov, 2001: Analysis of integrated cloud liquid and precipitable water vapor retrievals from microwave radiometers during the Surface Heat Budget of the Arctic Ocean project. J. Geophys. Res., 106, 32 019–32 030, https://doi.org/10.1029/2000JD000055.
Wood, R., and C. S. Bretherton, 2006: On the relationship between stratiform low cloud cover and lower-tropospheric stability. J. Climate, 19, 6425–6432, https://doi.org/10.1175/JCLI3988.1.
Wood, R., and D. L. Hartmann, 2006: Spatial variability of liquid water path in marine low cloud: The importance of mesoscale cellular convection. J. Climate, 19, 1748–1764, https://doi.org/10.1175/JCLI3702.1.
Yu, C.-K., P.-R. Hsieh, S. E. Yuter, L.-W. Cheng, C.-L. Tsai, C.-Y. Lin, and Y. Chen, 2016: Measuring droplet fall speed with a high-speed camera: Indoor accuracy and potential outdoor applications. Atmos. Meas. Tech., 9, 1755–1766, https://doi.org/10.5194/amt-9-1755-2016.
Zaremba, T. J., R. M. Rauber, G. M. McFarquhar, M. Hayman, J. A. Finlon, and D. M. Stechman, 2020: Phase characterization of cold sector Southern Ocean cloud tops: Results from SOCRATES. J. Geophys. Res. Atmos., 125, e2020JD033673, https://doi.org/10.1029/2020JD033673.
Zhang, D., 2017: MWR retrievals (MWRRET1LILJCLOU): ARM Mobile Facility (MAR) Hobart, AUS to Antarctic coast—Resupply ship Aurora Australis; AMF2 (M1). Subset used: 21 October 2017– 23 March 2018. ARM Data Center, accessed 10 January 2020, https://doi.org/10.5439/1027369.
Zhang, D., and M. Levin, 2019: Thermodynamic cloud phase (THERMOCLDPHASE), version 1.4.6. ARM Data Center, accessed 6 December 2022, https://doi.org/10.5439/1871014.
