Ice Cloud Formation Related to Oceanic Supply of Ice‐Nucleating Particles: A Case Study in the Southern Ocean Near an Atmospheric River in Late Summer

This study investigated ice cloud formation associated with marine bioaerosols over the Southern Ocean (SO) using a combination of cloud particle sensor (CPS) sonde observations, satellite products, reanalysis data, and backward trajectory analysis. The CPS sonde detected ice clouds at temperatures higher than −10°C in the mid‐troposphere near an atmospheric river at high‐latitudes over the SO. Backward trajectory analyses indicated that a mid‐latitude air mass with a high concentration of atmospheric dimethylsulfide (DMS) in the atmospheric boundary layer (<1 km) arrived at the ice cloud formation layer over the high‐latitudes. The DMS in the boundary layer began to increase under high wave conditions, coincident with the highest chlorophyll‐a concentrations in the ocean. These results suggest that bioaerosols emitted from the ocean over the mid‐latitudes acted as ice‐nucleating particles for ice cloud formation over high‐latitudes.

• A cloud particle sensor sonde measured ice clouds at air temperatures higher than −10 °C in the mid-troposphere near an atmospheric river • Mid-latitude air mass with high dimethylsulfide (DMS) in the boundary layer reached the ice cloud formation layer over high-latitudes • Simulated high surface atmospheric DMS was associated with observed high chlorophyll-a concentration under simulated high wave conditions

Supporting Information:
Supporting Information may be found in the online version of this article. 10.1029/2023GL106036 2 of 10 CloudSat satellites, Adhikari et al. (2012) revealed that low-level cloud with a base height below 2 km dominates over the SO in all seasons with the maximum fraction in summer (Adhikari et al., 2012;Kawai et al., 2015).Sea ice cover reduces the release of both water vapor and heat from the ocean to the atmosphere, thereby causing a relatively low cloud fraction at low levels over the sea ice compared with that over the open ocean (Wall et al., 2017).
The microphysical properties of clouds (e.g., cloud phase and cloud particle number) are also important parameters for determining the surface heat budget.Liquid water clouds, which freeze under pure and very cold conditions (i.e., temperatures below −38°C), can reflect more shortwave radiation than is reflected by ice clouds.However, various aerosols (e.g., mineral dust, organic aerosols, and bioaerosols) that originate from local and remote areas generally act as ice-nucleating particles (INPs) for ice cloud formation at higher temperatures (Hoose & Möhler, 2012;McCluskey et al., 2018;Vergara-Temprado et al., 2018).In satellite retrievals, low-level mixed-phase clouds at relatively high temperatures of −15 to −5°C are observed during spring and summer (Listowski et al., 2019).Following comparison of the seasonal changes in the fraction of low-level mixed-phase clouds with changes in aerosols over coastal regions of Antarctica, Listowski et al. (2019) concluded that increase in bioaerosols (e.g., bacteria, fungal spores, pollen, and diatoms) related to high biological activity would contribute to low-level mixed-phase cloud formation.Additionally, in another satellite-derived product, the highest fractions of low-level ice clouds were found to occur at temperatures above −10°C, coincident with high chlorophyll-a concentrations in summer (Sato & Inoue, 2021).Phytoplankton blooms, which are a potential source of marine bioaerosols in the lower troposphere via sea spray (DeMott et al., 2015), can occur in the marginal ice zone during the seasons of sea ice melting (Arrigo et al., 2008;Taylor et al., 2013).Air sampling conducted onboard a ship revealed that marine bacteria dominate in the lower troposphere over the SO in summer (Uetake et al., 2020).Therefore, recent field campaigns targeting a link between marine bioaerosols and cloud formation have been conducted over the SO (Alexander et al., 2021;Mallet et al., 2023;McFarquhar et al., 2021;Sato et al., 2018).In contrast, secondary ice-production processes (e.g., rime splintering, collision fragmentation, and droplet shattering) can also increase ice crystal numbers (Hallett & Mossop, 1974;Scott & Hobbs, 1977;Vardiman, 1978).
Numerical climate models are valuable tools for the study of atmospheric circulation globally; however, there are several problems with the representation of microphysical and macrophysical cloud properties in such numerical models (Inoue et al., 2021a;Kuma et al., 2020;Lawson & Gettelman, 2014;Wall et al., 2017).Specifically, several studies have reported large biases in cloud radiative properties, which are related to errors in the transitions between the liquid and ice phases (Varma et al., 2020;Vergara-Temprado et al., 2018).Therefore, direct observation is required for comprehensive investigation of the atmospheric environment suitable for ice cloud formation.This study elucidated the relationship between ice clouds and marine aerosols over the SO using observational data obtained during an Antarctic cruise of the Japanese Research Vessel (R/V) Shirase.

Observational Data
The R/V Shirase conducted the 64th Japanese Antarctic Research Expedition (JARE64) cruise in the SO during November 2022 and March 2023 (Figure 1).During this cruise, we conducted occasional cloud particle sensor (CPS) sonde (Meisei Electric Co., Ltd., Japan) observations over the SO.The CPS uses a near-infrared laser (typical wavelength: 790 nm) as a linearly polarized light source.The CPS sonde with a dipole sensor and two photodetectors that are placed at angles of 55° and 125° to the direction of the source light can measure the vertical structures of clouds (e.g., total particle count, particle phase, and particle size) at 1 s intervals during its ascent (Fujiwara et al., 2016).The two detectors receive scattered light through slits placed in front of them.The detector at 125° receives only light polarized perpendicularly to the light source.The particle signal voltage from the two detectors (I 55 and I 125p ) is 0-7.5 V (resolution: 0.03 V).To distinguish cloud ice from cloud water, the degree of polarization (DOP), defined as DOP = (I 55 − I 125p )/(I 55 + I 125p ) by Fujiwara et al. (2016), is used.When the DOP is <0.5, the particle is most likely to be ice (nonspherical particles), based on field campaigns in cold regions (Inoue et al., 2021b).In contrast, in many cases, when the DOP is >0.5, the particle is water (spherical particles).
To estimate the total count of cloud particles (/L), we corrected the original number of particles using a method proposed by Inoue et al. (2021c).This investigation used the cloud phase information (i.e., liquid, ice, or mixed), the cloud particle count, and the signal voltage (I 55 ) for cloud particle size obtained by the CPS.Additionally, a regular radiosonde (RS-11G; Meisei Electric Co., Ltd., Japan), which can obtain basic meteorological profiles (e.g., pressure, temperature, relative humidity, and wind speed), was connected to the CPS sonde.

Reanalysis and Satellite Chlorophyll-a Data Set
This study used two reanalysis data sets: the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis v5 (ERA5; Hersbach et al., 2020) and the Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA-2; Gelaro et al., 2017).The ECMWF ERA5 data set includes atmospheric and surface oceanic parameters (e.g., sea surface temperature and wave height).This investigation used atmospheric (mean sea level pressure, horizontal wind speed, temperature, and specific humidity) and oceanic (significant height of combined wind waves and swell) parameters at 1 hr intervals.Comparison of significant wave height data from ERA5 with in situ observations revealed that wave height is well reproduced in ERA5 (Hošekov et al., 2021).
For atmospheric river (AR) analysis, we calculated the integrated water vapor (IWV: kg/m 3 ) and saturated IWV (IWV sat ) from vertical profiles of specific humidity and temperature between 900 and 300 hPa using ERA5 data with 16 pressure levels: where q is specific humidity (kg/kg), T is air temperature (K), q s is the saturated specific humidity (kg/kg) corresponding to T at each pressure level according to the Clausius-Clapeyron relation and g is gravitational acceleration (=9.8065 m/s 2 ).To identify the AR over the SO, we used the threshold described by Gorodetskaya et al. (2014): where IWV sat,mean is the zonal mean IWV sat along each latitude, IWV sat,max is the maximum value of IWV sat along the same latitude, and AR coef (=0.2) is a coefficient determining the relative strength of an AR.Additional details are provided in Gorodetskaya et al. (2014).
The modern-era retrospective analysis for research and applications, version 2 (MERRA-2) is a global atmospheric reanalysis data set produced by the Global Modeling and Assimilation Office of the National Aeronautics and Space Administration.In addition to standard meteorological observations, aerosol observations (aerosol optical depth) from ground and satellites observations are assimilated in MERRA-2.Major atmospheric aerosols including sulfate (e.g., dimethylsulfide; DMS), sea salt aerosols, dust, black carbon and organic carbon in MERRA-2 are simulated with a radiatively coupled version of the Goddard Chemistry, Aerosol, Radiation, and Transport model (Chin et al., 2002).DMS emission are based on monthly climatology of oceanic DMS concentration (Chin et al., 2002).To investigate the vertical profile of aerosol parameters, this study used the mixing ratio of DMS with 3 hr fields at horizontal resolution of 0.625° × 0.5° and 72 sigma vertical levels up to 0.01 hPa.
In this study, 8-day mean chlorophyll-a concentration data, available on a 0.1° × 0.1° latitude/longitude grid from the MODIS-Aqua satellite, were used to investigate the presence of phytoplankton over the SO.Full details of the data have been reported in a previous study (Hu et al., 2012).

HYSPLIT Model
To investigate the origin of air masses arriving at observation points, we used the National Oceanographic and Atmospheric Administration's Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (version 5.2.0) (Rolph et al., 2017;Stein et al., 2015).Global Forecast System (GFS) data (version 15) on a 0.25° × 0.25° latitude/longitude grid and Global Data Assimilation System (GDAS) data on a 1° × 1° latitude/ longitude grid (Kanamitsu, 1989) were also used as meteorological conditions for the HYSPLIT model.The details are shown in Stein et al. (2015).The HYSPLIT model has been used in many earlier studies to classify the source of aerosols over the SO (e.g., Kurita et al., 2016;Neff & Bertler, 2015;Uetake et al., 2020).Given the estimated residence time of bacteria in the atmosphere, 3-day was considered a period suitable for evaluating bacterial sources (Uetake et al., 2020).Therefore, we calculated 3-day backward trajectories for air masses at characteristic heights determined from the CPS sonde observations.To reduce the uncertainty in the trajectory results, backward trajectory analyses were conducted using 27 ensemble members for each data set (i.e., GFS and GDAS).

Results
An area of low pressure generated over the Indian sector of the SO on 20 February 2023 subsequently crossed the SO.On 22 February 2023, this area of low pressure was near the location of the R/V Shirase (Figure 1).
To the east of the low, a corridor of moisture, associated with an AR, stretched from the mid-latitudes to the high-latitude region.During the eastward movement of the R/V Shirase from Japan's Syowa Station (69.0°S, 39.6°E) to the Totten Glacier, a CPS sonde launch was conducted near the Antarctic coastal region at 1300 UTC on 22 February 2023.Figure 2 shows the vertical profiles of air temperature, water vapor mixing ratio, wind speed, wind direction, and cloud information (e.g., cloud phase, cloud size, and cloud particle count) obtained by the CPS sonde.Relatively high air temperatures of above −10°C with several inversions were observed in the lower and middle troposphere (<4.5 km; Figure 2a).The water vapor mixing ratio peaked between the heights of 3 and 3.5 km (up to 4 g/kg) with a moisture inversion at the height of approximately 3 km (Figure 2b).The water vapor mixing ratio in the middle troposphere was higher than that in the lower troposphere.Although the relatively high temperatures and the water vapor mixing ratio were reasonably well reproduced in ERA5, biases in the temperature and water vapor mixing ratio were evident in the lower and middle troposphere, particularly underestimation of the water vapor mixing ratio in ERA5 (Figures S1a and S1b in Supporting Information S1).
The peak in the water vapor mixing ratio with the moisture inversion in the mid-troposphere during the AR event was reported by Gorodetskaya et al. (2020).A strong northerly wind prevailed in the troposphere (Figure 2c), indicating that the southward advection of warm and moist air associated with the AR caused the moisture inversion and the relatively high temperatures.
The CPS sonde measured two cloud layers (first layer cloud between surface and 1.1 km and second layer cloud between 2.3 and 6.8 km) below the height of 7 km at 1300 UTC on 22 February 2023 (Figures 2d and 2e).A relatively high DOP (>0.5) with a lower voltage (i.e., a signal for a smaller particle) was observed below the height of 1.1 km, meaning that small spherical particles (i.e., liquid water cloud) dominated in the lower troposphere.In contrast, the upper cloud layer exhibited different cloud characteristics compared with those of the lower cloud layer.Above the height of 5 km, even though the lower DOP value with high and low voltages indicated the presence of large and small nonspherical particles (i.e., snow and ice cloud), liquid water cloud dominated up to the height of 6.8 km.Below 5 km, the DOP values of <0.5 with various voltages indicated that ice clouds dominated with snowfall.The presence of ice clouds between the heights of 2.2 and 5 km under higher temperatures (i.e., above −10°C) shows that INPs could contribute to ice cloud formation in this layer (Figures 2a and 2d).The mid-tropospheric ice clouds were observed with a relatively high water vapor mixing ratio at heights above 3 km (Figures 2b and 2d), suggesting that the southward advection of warm and moisture air associated with the AR influenced the mid-tropospheric ice cloud formation.In the mid-troposphere, large nonspherical particles (I 55 is 7.5 V) were detected that corresponded to snow (Figure 2d); however, visual observation conducted on the deck of the R/V Shirase did not detect snowfall at the surface during the time of the CPS sonde observation.The relatively dry layer below the base of the ice cloud (height: 1.5-3 km) would have led to redistribution of INPs through snow sublimation (Figures 2b and 2d).The ice cloud between 600 and 700 hPa under the relatively high temperature conditions (above −10°C) was well reproduced in ERA5, except for the ice cloud at 700 hPa (Figures S1c and S1e in Supporting Information S1).In contrast, the height of the liquid water cloud layer between 500 and 600 hPa in ERA5 was lower than that determined above 500 hPa from the CPS observations.A notable feature is that ice clouds were detected below the height of 5 km under higher temperatures (above −10°C), whereas liquid water clouds were observed at heights above 5 km under colder temperature conditions (i.e., below −25°C) (Figures 2a, 2d, and 2e).
The CPS observations suggest that INPs from local and/or remote regions could contribute to ice cloud formation under relatively high temperatures.To investigate the source of the air mass at the heights of the ice and liquid water cloud layers, we conducted 3-day backward trajectory analyses for initial heights of 4.5 km (i.e., the ice cloud layer) and 6.5 km (i.e., the liquid water cloud layer) above ground level (hereafter, BT4.5 and BT6.5, respectively) (Figure 3).The small ensemble spread in the trajectory tracks of BT4.5 revealed that the air mass at the height of 4.5 km, transported to the coastal region of Antarctica at the time of the CPS observation, originated near southern Africa (Figure 3a).The air mass that reached the ice cloud layer at the time of the CPS observation time passed through the atmospheric boundary layer (<1 km) and over an oceanic area with high chlorophyll-a concentration on 20 February 2023 (Figures 3a and 3b).The relatively large ensemble spread in the trajectory tracks of BT6.5 revealed that the air mass at the height of 6.5 km at the time of the CPS observation also originated near southern Africa (Figure 3a).Although trajectory tracks of BT6.5 also have a relatively large ensemble spread in height, the ensemble mean air mass that arrived at the liquid cloud layer at the time of the CPS observation passed above the atmospheric boundary layer (>1 km) during the period of the backward trajectory (Figure 3c).These trajectory analyses suggest that surface conditions on 20 February 2023 influenced the air mass at the height of 4.5 km on 23 February 2023.Additionally, similar results for both BT4.5 and BT6.5 were obtained when performing backward trajectory analyses using different meteorological data (GDAS), even with relatively large ensemble spread in the trajectory tracks and heights (Figure S2 in Supporting Information S1).
To investigate the atmospheric and oceanic environments at the positions of the air masses during the backward trajectory period, we assessed the atmospheric and oceanic parameters at the closest grid point for each air mass 10.1029/2023GL106036 6 of 10 position using MERRA-2 reanalysis data sets (Figure 4).The time-height sections of DMS for BT4.5 and BT6.5 are shown in Figures 4a and 4b, respectively.The DMS values for BT4.5 gradually increased in the atmospheric boundary layer after 20 February 2023 (Figure 4a).By following the calculated backward trajectory of the air mass, it was established that the relatively high DMS for BT4.5 extended to the mid-troposphere after 1200 UTC on 21 February 2023, coinciding with the presence of the observed ice cloud layer detected by the CPS sonde.In contrast, extension of high DMS for BT6.5 from the surface to the upper troposphere, including the liquid water cloud layer, was not found during the backward trajectory period (Figure 4b).
Sea spray represents a source of atmospheric aerosols under high wave and strong wind conditions (DeMott et al., 2015;Wilson et al., 2015).The strong wind associated with the area of low pressure contributed to gradual increase in wave height over the SO at the air mass positions from 20 February 2023 (Figure 4c).On 20 February 2023, wave height exceeding 3 m was found at the air mass positions, coincident with high chlorophyll-a concentrations (Figure 4d).Chlorophyll-a concentration, which is used to estimate phytoplankton numbers, can be an indicator of the quantity of marine bioaerosols in the lower troposphere (Richert et al., 2019;Williams et al., 2016).The DMS in the boundary layer began to increase under the conditions of high wave heights and high chlorophyll-a concentrations, indicating that marine bioaerosols were emitted into the lower troposphere via sea spray (Figures 4a, 4c, and 4d).These results suggest that marine biogenic material, including bacteria emitted with DMS under high wave conditions, could have contributed to ice cloud formation in the mid-troposphere at the time of the CPS observation.

Conclusions and Discussion
Using CPS sonde data, satellite data, reanalysis data, and backward trajectory analyses, this study investigated the occurrence of mid-tropospheric ice clouds at higher temperatures during an AR event.Near the AR on 22 February 2023, ice clouds were detected in the mid-troposphere at temperatures above −10°C, whereas liquid water clouds were detected at temperatures below −20°C in the upper troposphere.The 3-day backward trajectory analyses showed that an air mass, which was located near southern Africa on 19 February 2023, traveled through the boundary layer and over the SO with high chlorophyll-a concentration to reach the mid-troposphere 10.1029/2023GL106036 8 of 10 over the observation point.As the air mass crossed the mid-latitude SO with high chlorophyll-a concentration, the DMS in the boundary layer increased under the high wave conditions.These results suggest that marine aerosols emitted from the mid-latitude SO promoted ice cloud formation in the mid-troposphere over the CPS observation point.
Generally, non-sea-salt sulfate originating from DMS acts as cloud condensation nuclei for liquid water cloud formation (Charlson et al., 1987).However, atmospheric DMS is also recognized as an indicator of marine biogenic material.The chlorophyll-a concentration, which is used to estimate phytoplankton numbers in the surface ocean (Richert et al., 2019;Williams et al., 2016), also has high values associated with high atmospheric DMS in the boundary layer under high wave conditions.Additionally, a relationship exists between phytoplankton concentration and bacteria numbers (Richert et al., 2019;Williams et al., 2016).At high temperature conditions, bacteria act as INPs for ice cloud formation (Hoose & Möhler, 2012).Therefore, high atmospheric DMS in the boundary layer and high chlorophyll-a concentration at the ocean surface suggest that sea spray under high wave conditions would cause emission of bacteria into the atmosphere with the DMS originating from phytoplankton.The results derived from the backward trajectory analyses and the DMS data obtained from the MERRA-2 reanalysis data set indicate that although ice crystals produced by secondary ice-production processes would be observed in the middle troposphere, in this study case, the marine bacteria acted as INPs for ice cloud formation in the mid-troposphere.Several climate models did not reproduce the ice cloud formation under higher temperatures (Zelinka et al., 2020).Therefore, developing knowledge of the transport of marine bioaerosols could help improve our understanding of ice cloud formation under relatively high temperature conditions.
Using CPS observation data and reanalysis data, our investigation revealed the potential impact of the supply of marine-derived INPs on ice cloud formation.In this case study, occurrence of ice cloud in the mid-troposphere under relatively high temperatures was observed with a relatively high water vapor mixing ratio related to the AR, indicating that surface marine material suitable for promoting ice cloud formation was transported by the AR from the mid-latitudes to the mid-troposphere at high-latitudes.A peak in the water vapor mixing ratio in the mid-troposphere was observed during the AR event, but this is not always the case (Gorodetskaya et al., 2020).Although increased emission of marine aerosols under dominant cold advection from the high-latitudes can contribute to high-latitude ice cloud formation (Sato & Inoue, 2021), a strong AR is an important feature for ice cloud formation under higher temperature conditions at high-latitudes.This study did not provide evidence of atmospheric bacteria numbers in the mid-troposphere because of lack of shipboard bacteria sampling.However, in the high-latitudes of the Northern Hemisphere, an ice cloud observed at a temperature above −10°C was found to have high concentration of organic carbon under high wave conditions (Inoue et al., 2021b).For the high-latitudes and SO in the Southern Hemisphere, further scheduled meteorological observations of aerosols and clouds by R/V Shirase during 2022 and 2027 will provide the opportunity to find further evidence of the impact of marine bioaerosols on ice cloud formation over the SO.

Figure 1 .
Figure 1.Map of integrated water vapor (color shading: kg/m 2 ) with mean sea level pressure (gray contours: hPa) and averaged wind speed between 300 and 900 hPa (gray vectors: m/s) from ERA5 at 1300 UTC on 22 February 2023.Hatched area indicates the atmospheric river (AR) defined by the AR algorithm.Black line shows the track of the R/V Shirase during JARE64.Purple dot indicates the position of launch of the cloud particle sensor sonde on 22 February 2023.Cyan square shows the position of Japan's Syowa Station.

Figure 3 .
Figure 3. (a) Map of the 8-day mean chlorophyll-a concentration (mg/m 3 ) during 18-26 February 2023.Red and blue lines show 3-day backward trajectories of air masses arriving at the height of 4.5 km (BT4.5;blue) and 6.5 km (BT6.5;red) at the time of the cloud particle sensor (CPS) observation for the ensemble mean (thick) and 27 ensemble members (thin) with Global Forecast System meteorology on a 0.25° × 0.25° latitude/longitude grid.Plots show positions of air masses at 0000 UTC on 20 February (dots), 21 February (triangles), and 22 February 2023 (squares) for each ensemble member.Time series of the height (km) above ground level of the air masses arriving at the height of (b) 4.5 km and (c) 6.5 km at the time of the CPS observation for the ensemble mean (thick) and 27 ensemble members (thin).The 3-day backward trajectories were computed using National Oceanographic and Atmospheric Administration's hybrid single-particle Lagrangian integrated trajectory model.

Figure 4 .
Figure 4. Time-height sections of the 27-member ensemble mean dimethylsulfide (DMS:×10 −11 kg/kg) at the air mass position of (a) BT4.5 and (b) BT6.5 with Global Forecast System meteorology data during the backward trajectory period.Blue and red contours show the height (km) of (a) BT4.5 and (b) BT6.5 for the ensemble mean (thick) and 27 ensemble members (thin).Time series of (c) simulated wave height (m) from ERA5 and (d) observed chlorophyll-a concentration (mg/m 3 ) from MODIS-Aqua satellite at the air mass position of BT4.5 (blue) and BT6.5 (red) for the ensemble mean (thick) and 27 ensemble members (thin).