Southern Ocean Cloud and Aerosol data: a compilation of measurements from the 2018 Southern Ocean Ross Sea Marine Ecosystems and Environment voyage

Due to its remote location and extreme weather conditions, atmospheric in situ measurements are rare in the Southern Ocean. As a result, aerosol–cloud interactions in this region are poorly understood and remain a major source of uncertainty in climate models. This, in turn, contributes substantially to persistent biases in climate model simulations, numerical weather prediction models and reanalyses. It has been shown in previous studies that in situ and ground-based remote sensing measurements across the Southern Ocean are critical for complementing satellite data sets due to the importance of boundary layer and 5 low-level cloud processes. These processes are poorly sampled by satellite-based measurements which are typically obscured by near-continuous overlying cloud cover observed in this region. In this work we present a comprehensive set of ship-based aerosol and meteorological observations collected on the TAN1802 voyage of R/V Tangaroa across the Southern Ocean, from Wellington, New Zealand, to the Ross Sea, Antarctica. The voyage was carried out from 8 February to 21 March, 2018. Many distinct, but contemporaneous, data sets were collected throughout the voyage. The compiled data sets include measurements 10 from a range of instruments, such as (i) meteorological conditions at the sea surface and profile measurements; (ii) the size and concentration of particles; (iii) trace gases dissolved in the ocean surface such as dimethyl sulfide and carbonyl sulfide; (iv) and remotely sensed observations of low clouds. Here, we describe the voyage, the instruments, data processing, and provide a brief overview of some of the data products available. We encourage the scientific community to use these measurements for 1 https://doi.org/10.5194/essd-2020-321 O pe n A cc es s Earth System Science Data D icu ssio n s Preprint. Discussion started: 9 November 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
The Southern Ocean is the cloudiest region on Earth and is also distant from major anthropogenic sources of aerosol (Haynes et al., 2011). This makes the Southern Ocean an ideal environment for studying aerosol-cloud interactions (Krüger and Graßl, 2011;Fossum et al., 2018;Hamilton et al., 2014) and the role of marine aerosol in the radiation budget. The contribution of 65 marine aerosol to Earth's radiation budget is both direct through aerosol scattering and absorption, and indirect via cloud droplet activation and their subsequent influences on cloud radiative processes (Murphy et al., 1998;Mulcahy et al., 2008;McCoy et al., 2015;Fossum et al., 2018). Marine aerosol can be classified as primary or secondary in origin (Fossum et al., 2018).
Primary aerosols, such as sea spray, are directly injected into the atmosphere when breaking waves entrain air bubbles into the ocean surface, which subsequently form whitecaps and burst (Hultin et al., 2010;Salter et al., 2014). Secondary aerosols, 70 such as sulfate aerosols, are formed from the nucleation of sulfur-containing gases in a gas-to-particle conversion process. One of the main precursors of sulfate aerosol in the marine environment is dimethyl sulfide (DMS), a by-product of an enzymatic compound produced within phytoplankton (dimethylsulfoniopropionate, DMSP; Read et al., 2008;Fossum et al., 2018). DMS is the main natural source of atmospheric sulfur, with a global average of 28.1 Tg of sulfur being emitted annually from the oceans into the atmosphere in the form of DMS (Lana et al., 2011). When DMS is emitted into the atmosphere, it undergoes 75 a series of chemical reactions to form sulfur dioxide (SO 2 ), resulting in a typical lifetime of DMS in the atmosphere of 1-2 days (e.g. Chen et al., 2018). The SO 2 can then be further oxidised to form sulfuric acid, sulfate aerosol and methanesulfonic acid (MSA; e.g. Yan et al., 2020). Aerosol emitted into the atmosphere can grow in size via condensation and coagulation. The ability of any aerosol particle to serve as a nucleus for water droplet formation depends on its size, chemical composition, the local supersaturation, and meteorological conditions such as the cloud base updraft velocity (Rosenfeld et al., 2014). Aerosol 80 has a significantly different impact on cloud formation and evolution, depending on whether it acts as an ice nucleating particle (INP), a cloud condensation nuclei (CCN), or both.
Despite their significant influence on climate, clouds still represent the largest source of uncertainty in modern climate models with aerosol-cloud interactions being a major factor in this uncertainty (Myhre et al., 2013;Haynes et al., 2011). For example, Hyder et al. (2018) recently identified that 70 % of the sea surface temperature biases observed in model simula- and overestimate downwelling solar radiation at the ocean surface. This leads to excessive sunlight being absorbed by the 90 ocean (Trenberth and Fasullo, 2010;Kay et al., 2016;Hyder et al., 2018) and subsequent higher sea surface temperatures than observed (Bodas-Salcedo et al., 2012;Mechoso et al., 2016). Previous studies have also shown the importance of accurate mixed-phase cloud parameterisations over the Southern Ocean in climate models to properly simulate cloud radiative properties over the Southern Ocean (Lawson and Gettelman, 2014;Kay et al., 2016;Schuddeboom et al., 2019;Noh et al., 2019).
In this paper we present a new data set of atmospheric (cloud, aerosol and thermodynamic properties) and seawater measurements that were collected during the six-week Southern Ocean Ross Sea Marine Ecosystem and Environment voyage (TAN1802) from Wellington, New Zealand, to the Ross Sea, Antarctica, in 2018. ::: The :::: data :::: sets :::::::: presented :::: here :::: are ::::::: publicly 120 ::::::: available :: at : https://doi.org/10.5281/zenodo.4060237 :::::::::::::::::: (Kremser et al., 2020) : . Given the sparsity of data in the Southern Ocean region, this data set provides a valuable collection of atmospheric and underway measurements that can be used to better understand aerosol-cloud processes over the Southern Ocean. This paper includes a description of DMS and carbonyl sulfide (OCS) measurements as previous work has identified that DMS plays an important role as a sulfate aerosol precursor. Furthermore, DMS concentrations have a particularly large impact on model aerosol forcings, yet are poorly represented in climate models 125 (Hoffmann et al., 2016;Bodas-Salcedo et al., 2019). Although not strictly related to aerosol-cloud interactions, OCS is a greenhouse gas and an important source of stratospheric sulfate aerosol (Crutzen, 1976;Brühl et al., 2012;Kremser et al., 2016).
3. Investigate the importance of sea-salt and other primary aerosols as CCN.
4. Investigate the influence of local biogenic sulfur emissions to secondary aerosol abundance.
5. Measure boundary layer profiles of aerosol and thermodynamic properties through combination of lidar measurements 155 and radiosonde flights and evaluate coupling between surface measured aerosol and low-level cloud capping within the marine boundary layer.
6. Link aerosol and surface trace gas properties to surface water biogeochemistry.
All measurements that were made to address the research objective :::: these ::: six ::::::: research ::::: aims are summarised in Table 2a and :: 2b ::: and : a detailed description of the instrumentation and their measurements is given below.
AWS measurements were complemented by human weather observations, all-sky cameras, ceilometer, Mini-Micropulse 195 lidar, and micro rain radar measurements, which provided important information about visibility, sky conditions, clouds, cloud type, and the amount of precipitation or fog events. In addition, up to three daily regular radiosondes of type InterMet iMet-1-ABxn were launched throughout the voyage, as well as smaller balloons carrying Windsond radiosondes that were launched in synoptically interesting conditions, e.g. within low pressure systems (Sect. 3.1.1). An overview of all radiosonde releases during the voyage is provided in Table B1 and B2.
All meteorological data are available in netCDF format at UTC times and are provided with the data set accompanying this study. Section 4.1 below provides some detail about the meteorological conditions encountered during the voyage.

Instrument descriptions
In addition to the instrumentation mentioned above, atmospheric measurements were conducted using a range of instruments, including a cavity ring-down spectrometer, cloud condensation nuclei counter, condensation particle counter, mobility particle 205 size spectrometer, optical particle counter, neutral cluster and air ion spectrometer, a filter sampler, tethered balloon, and an unmanned aerial vehicle (UAV). During rare clear sky conditions, aerosol optical depth (AOD) measurements were made using a hand-held sun photometer. The instrumentation and measurement techniques of each instrument are described below.
Furthermore, all data sets described here include some means of quality control and calibration procedures, which are also described below in the respective sections. Radiosondes are balloon-borne instruments that measure the vertical profile of temperature, relative humidity and pressure.
Altitude, wind direction and wind speed are calculated from the GPS location of the sonde. A total of 58 radiosondes of type InterMet iMet-1-ABxn (hereafter referred to as iMet) and 12 of type Windsond were released on a weather balloon during the 215 voyage (see Table B1 and B2). The iMet radiosondes were attached to 100 g Kaymont weather balloons and released two to three times per day at about 07:30, 00:00 and 19:30 UTC. The typical height reached by the balloons was between 10 and 20 km ASL. Of the total iMet radiosondes released, one failed right after launch, and one failed at 216 m ASL. In addition, two iMet radiosondes had faulty or intermittent relative humidity readings. No iMet radiosondes were released north of 58 • S or in unsuitable weather conditions, e.g. when wind speed was exceeding 35 kn or in high swell. In addition to the iMet radiosondes, 220 S1H3 Windsond radiosondes were launched sporadically throughout the voyage. The typical altitude reached by the Windsond radiosondes was 6 km. Five of the Windsonds were equipped with a second balloon to perform measurements during the descent, but only two descending profiles were successfully measured.

240
During the voyage, two UAV flights were performed when the observed wind speed was below 5 m s −1 . For the first flight, which took place on 4 March 2018, the UAV was equipped with an optical particle counter (OPC) of type Alphasense OPC-N2, a GoPro Hero4 camera and a customised radiosonde. The radiosonde was equipped with a SHT75 temperature and relative humidity sensor. Temperature can be measured between -40 • and +40 • C with an accuracy of 0.3 • C and a resolution of ±0.01 • C and relative humidity can be measured with an accuracy of 1.8 % and a resolution of 0.05 %. A customised radiosonde 245 was required to be deployed on the UAV (rather than using a standard sonde) as it needed to interface with the OPC-N2 sensor and data had to be transferred over radio to the ground station. The Alphasense OPC-N2 is an OPC designed to count ambient particulate and drizzle sized cloud droplets between 0.38-17 µm in size. Ambient air is drawn into the sensor by a small rotary micro-fan at a flow rate of about 1.2 L min −1 . The air enters the front of the device through a 6 mm orifice into an open optical cavity, where red laser light (around 650 nm) is incident on the incoming aerosol. Scattered light from the aerosol is collected 250 via an elliptical mirror and a dual-element photodetector. These measurements are used to determine the particle size and particle number concentration.

Helikite
Similar to the UAV flights, two helikite flights were conducted in suitable weather conditions subject to wind speed of below 5 m s −1 . For the first flight, the helikite was equipped with an iMet radiosonde and an OPC-N2, providing profiles of aerosol concentration, as well as temperature, pressure and humidity profiles. The second helikite flight had to be terminated shortly after launch as the weather conditions changed rapidly, resulting in no measurements.

265
The helikite comprised a large 6 m 3 balloon with a sturdy kite base. Lift can be achieved by inflating the balloon with helium and is aided by the additional lift of the kite. As a result of the large volume of the balloon, the total payload can be around 2 to 3 kg. The helikite was flown off the fantail and was anchored to an electric winch fitted with >1 km of high tensile strength Dyneema line. This system itself offers the opportunity to fly more expensive sampling equipment than typically deployed during a radiosonde flight where equipment is usually not recovered.

270
13 The first flight of the helikite occurred midway through the voyage on 26 February 2018. Conditions were good with wind speeds less than 5 m s −1 . Due to the inexperience of the helikite operator, the helikite was flown in near neutral buoyancy, i.e. the lift provided by the balloon was near or equal to the weight of the payload. As a result, the only lift received during the flight was from the kite alone. Once the helikite left the slipstream of the Tangaroa :::::::: Tangaroa it rose slowly to an altitude of 260 m.
At this stage, the additional weight of the tethered string counter-balanced all lift. After sampling for around 45 minutes, the 275 system was reeled back in.

Ceilometer
During the voyage a ceilometer, which is a low-power lidar, made continuous measurements of the overlying atmospheric state. The ceilometer deployed on the voyage was a Lufft CHM 15k, which operated at an infrared wavelength of 1064 nm, with a maximum range of 15 km. The ceilometer was installed on the Gilson gantry behind the monkey island ( layers, and boundary layer height can be determined. As the emitted signal is strongly attenuated by thick clouds, it is often not possible to observe the middle or tops of clouds. On some occasions, the movements of the ship (pitch and roll) affected the ceilometer measurements when there were horizontally inhomogeneous clouds, producing a vertical filament structure in the backscatter.

290
The Mini-Micropulse Lidar (MiniMPL) is a sophisticated laser remote sensing system that provides continuous, unattended monitoring of the profiles and optical properties of clouds and aerosols in the atmosphere. A micropulse lidar (MPL) transmits laser pulses that scatter (reflect) off particles in the atmosphere. The MPL then measures the intensity of backscattered light using photon-counting detectors and transforms the signal into atmospheric information in real time. During the campaign, aerosol backscatter data were collected using the Sigma Space MiniMPL, which is a compact version of the standard MPL 295 described by Ware et al. (2016). The detection range of the MiniMPLis limited to the upper troposphere/lower stratosphere.
14 The MiniMPL is a dual-polarisation micropulse lidar operating at a wavelength of 532 nm at 2.5 kHz (pulse energy is 3-4 µJ). Laser light that is scattered back towards the instrument is collected by an 80 mm diameter receiver (Spinhirne et al., 305 1995;Campbell et al., 2002;Flynn et al., 2007). The vertical range resolution was set at 15 m during the ship campaign. A two-axis scanning head was mounted on top of the environmental enclosure containing the lidar, to provide variable-angle scanning throughout the voyage. Azimuth was fixed for observations (pointing outward from the side of the ship) and the scanning head was programmed with an elevation-only scanning routine that included the following angles: 0, 5, 10,15,20,30,40,45,50,60,70,80, 90 • elevation. The finer, 5 • elevation step was used near the horizon, and then 10 • steps from 20 • 310 to 90 • (zenith). An observation was also made at 45 • because it is convenient geometrically. At 0 • and 90 • , the observations were 12 minutes long, at other angles 6 minutes, resulting in the full scanning cycle taking 90 minutes. The elevation angle of each particular observation is recorded in the data file. Note that there were some instances during the campaign :::::: (overall :: 9 :::: days) : when a software failure caused the scanning system to not follow the programmed schedule.
The instrument ordinarily requires range-dependent calibration of backscatter in the form of dead time, overlap and after-325 pulse corrections, which account for the saturation of the photon counter, incomplete overlap of the outbound and inbound beams, and post-pulse reflections from the internal parts of the instrument, respectively. These were supplied by the manufacturer. An improved calibration was produced post-voyage, which addresses a technical issue with the manufacturer calibration (bit truncation of dead time polynomial coefficients) and a change in overlap which might have happened during transport and deployment of the instrument. The data product produced with the third-party mpl2nc software was calibrated with the 330 improved calibration and is supplied with the data set.
The CHM 15k ceilometer and Sigma Space MiniMPL measurements were both processed using the Automatic Lidar and Ceilometer Framework (ALCF, Kuma et al., 2020b). While ALCF was developed to provide a tool to evaluate clouds simulated by climate models or reanalysis data using ceilometer or MiniMPL observations, ALCF can be run independently of any model input to process ceilometer or MiniMPL observations. ALCF can ingest the raw measurements, transform backscatter profiles 335 to profiles comparable with different instruments, and output the results in netCDF format. ALCF is described in detail in Kuma et al. (2020b).

345
-Level 1: contains ALCF processed raw ceilometer and MiniMPL data sets (one file per day) in netCDF format. The data products ::::::::::: dataproducts included are time series of vertical backscatter profiles, backscatter standard deviation, cloud base height, cloud mask ::::::::: cloudmask, and lidar ratio. ::: The :::: data :::: were ::::::::::: sub-sampled :: to :: 5 min ::::::: intervals :::: with :: a :::::: vertical ::::::::: resolution :: of :: 50 m : ." : drop size distribution from the Doppler spectra. These drop size distributions can be integrated to derive rain rates even for very small amounts of precipitation, below the thresholds detectable by conventional rain gauges. The software supplied by the manufacturer completes all this processing and also makes estimates of other parameters, such as liquid water content. The temporal resolution of the measurements is 10 s. Measurements of snowfall using this instrument are more challenging because the particle backscattering cross sections depend on both their mass and shape, while terminal velocities relationship to particle 360 size depends on their projected area. In the case of snowfall, we use the method of Maahn and Kollias (2012) to process the raw data to derive radar reflectivity, velocity, spectral width and snowfall rate estimates. The radar was installed on the port side of the gallery beneath the bridge (Fig. 3).
Furthermore a record of whether or not the sun was obscured by clouds was produced by monitoring the image saturation over the solar disk. All-sky imagery, along with estimates of cloud fraction and sun obscuration obtained during this voyage were 385 primarily used for quality assurance and quality control (QA/QC) of other sky viewing observations such as the ceilometer and MiniMPL measurements (as described in, e.g. Wagner and Kleiss, 2016). Cloud fraction derived from the sky camera product is also useful for model evaluation and when combined with the raw imagery and ceilometer data it could potentially be used to classify cloud types as described in Huertas-Tato et al. (2017).

Cavity ring-down spectrometer -Picarro
By the voyages nature, the ship did not always head into the wind. As a result, there were distinct times throughout the voyage when winds from the stern outpaced the motion of the ship and therefore the sampling line of air sampling instruments was 405 often exposed to exhaust from the ship. This problem was largely unavoidable, but the ship's measurements of wind-speed and heading combined with high precision measurements of carbon dioxide (CO 2 ) were used to identify contamination episodes.
Experience from previous voyages (e.g. Law et al., 2017) has shown that the Cavity Ring-Down Spectrometer (CRDS) is ideally suited to detect ship exhaust contamination. For this and other reasons beyond the scope of this paper, a CRDS (G2301, Picarro) was installed on the ship and operated continuously throughout the voyage. The CRDS was installed in an equipment 410 room off the middle lab (Fig. 3) measuring atmospheric mixing ratios of CO 2 and methane (CH 4 ) continuously at 1 Hz. Air for analysis by the CRDS was obtained from an inlet on the Forward Light Tower above the Bridge (∼20 m ASL) via an airline to the middle lab. Air was pumped down from the airline at about 2 L min −1 , of which 150 mL min −1 (determined by a mass flow controller) is used for analysis. Before the air from the airline was sampled by the analyser, it was dried to a dew point of 2-4 • C using a thermoelectric cooler and then dried further to a dew point between -30 and -40 • C using a back-flushed 415 Nafion dryer in which remaining water vapour in the air is transferred to the CRDS exhaust air that had been dried by passing it through a molecular sieve trap. While the Picarro instrument does measure the concentration of water vapour in the air, in this system the water vapour measurement was only used as a diagnostic indicator of system performance. Solenoid valves controlled by the Picarro were used to select either pre-dried air for analysis, or air from one of three reference tanks, plus a target tank, for system calibration. A calibration sequence was automatically run twice per day.

430
Throughout the voyage, the container laboratory, which housed the majority of the underway aerosol sampling instrumentation, was positioned behind the mid-ships exhaust (2 m ASL). To prevent exhaust air from contaminating the in situ measurements of ambient marine aerosol, ambient air was drawn from the mast of the R/V Tangaroa :: RV ::::::::: Tangaroa, through the conduit (Fig.   3) to the container laboratory, at a rate of 4.1×10 −2 m 3 s −1 . Size-dependent losses of particulate to conduit walls from anisokinetic sampling, gravity, turbulence, and diffusion are described in detail in Hartery et al. (2020b). The average transit time 435 for particulates through the 40 metre long common aerosol sampling conduit was <8 s. The inlet of the conduit was angled downwards to prevent the accumulation of precipitation within the inlet region.
while the SMPS instrument is slow. However, the disadvantage of the PCASP is that it can only measure particles larger than 100 nm.

470
The PCASP instrument recorded the number of observed particles in 30 particle size bins at a frequency of 1 Hz. While the PCASP measurement frequency is high, it is generally beneficial to integrate the PCASP over a period as long as 5 minutes to get better counting statistics and decrease the relative measurement uncertainty. As a result, the measurements in each size bin were block-averaged into 5 minute intervals in a post-processing stage. Between 9 February and 21 March 2018, there were a total of 12,000 5 minute intervals, throughout which the instrument recorded for a total of 11,400 intervals. Four additional 475 measures of quality control were implemented in the data post-processing chain, viz: 1. Using the mole fraction of CO 2 in a coincident sampling line, measured by the Picarro instrument, to screen the 1 Hz sub-samples for contamination by ship exhaust (Hartery et al., 2020b). For 11,118 of the 5 minute intervals with data, the mole fraction of CO 2 was less than 405 ppm and the sample was flagged as "clean air".
2. Using the relative wind direction measured by the sonic anemometers in a simple wind sector analysis. Measurements 480 that were taken when the relative wind direction from both the port and starboard anemometers were within 60 • of aftward were removed. All other samples were flagged as having come from a "clean sector". Out of the 11,118 clean air samples (i.e. not contaminated by ship exhaust) 9,986 were from clean sectors.
3. Calculating the standard deviation of the 1 Hz sub-samples within each of the 5 minute intervals (Hartery et al., 2020b).
Even for a steady concentration of particles, the number of particles counted within a given interval will vary according 485 to Poisson counting statistics; thus, the standard deviation of the 1 Hz samples within the 5 minute interval should be approximately equal to the square root of the measured concentration. However, if the standard deviation of the 1 Hz sub-samples was more than three times greater than the square root of the concentration, then the 5 minute sample was discarded. This additional measure removed 184 samples.
4. Removing observations in the first size bin, as the lower threshold of particle detection in this bin is not well defined 490 due to potential variations in the refractive index of the measured particle(s). Additionally, the 4 th and 5 th size bins were added together and redefined as a single bin, as the 5 th size bin was in between linear gain stages of the particle counter, which led to spuriously low counts.

515
The CCN-100 was calibrated by the manufacturer prior to the voyage and calibrated by the operator after the voyage. The calibration procedure followed the methodology described in Rose et al. (2008). Overall, the supersaturation of each stage was accurate to within 20 % of the set-value, e.g. the stated supersaturation of 0.3 % was accurate to within ±0.06 %.

Condensation Particle Counter
The total abundance of particles in the size range 0.01-3.0 µm was measured with a Condensation Particle Counter (CPC3010; 520 TSI) at a frequency of 1 Hz. Similar to the data processing procedure for the PCASP-100X, the raw data were screened for contamination by ship exhaust according to the coincident CO 2 mole fraction, the relative wind direction, and the standard deviation of the 1 Hz subsamples. The screened data were then averaged over 5 minute intervals and merged to the common date coordinate. On 1 March 2018, the laser beam dump became partially dislodged within the optical cavity of the CPC3010 and the operator was unable to resolve this issue at sea. As this led to spurious counts, the data following 1 March were 525 excluded from the data set.

Scanning Mobility Particle Size Spectrometer
The abundance of particles in the size range 0.020-0.50 µm was measured with a Scanning Mobility Particle Size Spectrometer (SMPS3936; TSI). The SMPS instrument sizes particles according to how mobile the particle is in air. The instrument measured the total abundance of particles passing through an electrostatic classifier (EC3080L; TSI) with a condensation particle counter 530 (CPC3772; TSI). For a specific voltage setting, only particles of a specific size and charge will pass through the EC3080L and be observed by the CPC3772. The instrument was set to observe the concentrations of particles at 32 logarithmicallyspaced voltage levels. The concentration at each voltage level was observed over a period of 10 s, with an additional 2 s purge between voltages. The instrument scanned through the 32 set voltages once every 6.4 minutes. As with previous counters, the coincident CO 2 mole fraction time series was used to screen the raw 0.1 Hz data for contamination by ship exhaust. After 535 screening, the concentration-voltage spectra were merged to a common 30 minute data coordinate. The inversion of the merged concentration-voltage spectra into concentration-diameter spectra was calculated in the post-processing stage, accounting for multiple-charged particles and diffusional losses to the bipolar diffusion charger within the SMPS (Stolzenburg, 1988). As with the PCASP-100X data, the processed particle size spectra were corrected to ambient conditions by applying the size-dependent loss corrections detailed in Hartery et al. (2020b). originating from a corona discharge to an equilibrium that is used to calculate the total particle concentration in a given size range. The size of ions generated by the corona discharge is around 2 nm, masking the atmospheric signal of this size of neutral 555 clusters. In addition to ion and particle measurements, each measurement cycle contains an offset measurement during which particles in the sample air are charged by a unipolar corona charger and electrically filtered for measuring the background level of the electrometers. The offset is used to evaluate noise levels and instrument functioning. Measures of quality control were implemented in that data post-processing chain. First, the mole fraction of CO 2 in a coincident sampling line, measured by the Picarro instrument, was used to screen the NAIS data according to the suggested filtering protocol outlined in Sect. 4.2.1.

560
Note that the filtering of the NAIS data differ from the filtering of pollution events for the PCASP and SMPS data, but the impact on the remaining measurements is negligible. Secondly, data above 15 nm were excluded from the final data set due to technical issues with one of the electrometers. Further quality control of the measurements was performed by following the data cleaning and quality check guidelines described in Manninen et al. (2016), which are mainly based on visually inspecting the measurements. Overall, 37 % of the measurements made were included in the QA/QC data set.
To avoid contamination, seawater samples were gently filtered by pumping the water, using a peristaltic pump, through a 25 mm GF/F filter. The filter was changed after every four injections. A calibrated volume of 5.84 mL of the filtered water was 610 transferred to a silanized glass chamber :::::: (sparge ::::: tower), which was fitted with a quartz frit and purged with zero-grade nitrogen (99.9 % pure). To prevent organic matter build-up the chamber and frit were cleaned daily. The gas-phase DMS sample was then dried by passing through Nafion® dryers and trapped on a teflon-lined Tenax® stainless steel trap at -20 • C for 5 minutes and purged at 110 • C for GC analysis (DB-megabore sulfur column, 70 m length, 0.530 megabore diameter and film thickness 4.30 µm). The detector sensitivity was checked daily using two temperature controlled VICI® permeation tubes, one filled with 615 methylethylsulfide (MES) and the other with DMS. MES was used as an internal standard, with samples doped during analysis to allow for correction of short-term changes in detector sensitivity, while the DMS permeation tube provided the external standard (Walker et al., 2016). On average over the duration of the voyage, the detection limit was 0.079 (± 0.016) pgSs −1 .

Mid-Infrared CAvity enhanced spectrometer -MICA
The MICA (Mid-Infrared CAvity enhanced spectrometer, which is a prototype of a commercially available ABB Los Gatos

635
Cavity Output Spectroscopy (OA-ICOS, Baer et al., 2002;O'Keefe et al., 1999;Paul et al., 2001). Air samples are internally pumped through a 305 mm long and 51 mm diameter cavity at a mass flow rate of about 6×10 −6 kg s −1 with the cavity pressure regulated to 80 hPa. The beam of a quantum cascade laser (QCL) ramped over the wavenumber range 2050.2-2051.2 cm −1 is coupled into the cavity, the light exiting the cavity on the opposite side is collimated onto a HgCdTe photodiode.
Two highly reflective dielectric cavity mirrors allow for an effective path length of approximately 1000 m. Trace gas mixing 640 ratios are retrieved from infrared spectra online using manufacturer Los Gatos software. In addition, raw spectra are saved every 15 seconds to allow for consistency and quality checks of the recorded data.
For the TAN1802 voyage, MICA was deployed in the temperature controlled aerosol container laboratory, alternating measurements of the marine boundary layer and the surface ocean at intervals of 10 minutes for air and 50 minutes for water using a fully autonomous setup that consists of a pump, switching valves and a spray-head seawater equilibrator (Lennartz et al.,645 2017). The intake of the airline was located at 20 m ASL at the starboard forward mast on the monkey island (Fig. 3). Seawater from about 5 m depth was supplied to the equilibrator at a flow rate of 2-3 dm 3 min −1 . To ensure that concentrations remain at near equilibrium, the gas phase was constantly recirculated between the equilibrator headspace and MICA. A filter (PallAcro, 0.7 µm) was placed directly in front of the MICA inlet to remove particles and droplets. Teflon was used for all tubing, and materials known to cause OCS contamination such as rubber, were avoided. From gas phase mixing ratios in the equilibrated 650 air, dissolved concentrations were calculated using Henry's law constants.
Time series of observed wind speed, pressure, relative humidity, temperature, sea surface temperature, and radiation along the complete voyage track are shown in Fig. 7. The vessel reached the Southern Ocean region on day five of the voyage 685 (14 February 2018). The drop in air temperature and pressure when entering the Southern Ocean is clearly visible in Fig.   7(c, g). Over the Southern Ocean, air temperatures observed ranged mainly between +1 • C and -2 • C with a minimum of -7 • C (Fig. 7), with observed sea surface temperatures remaining around 0 • C. The median air and sea surface temperatures throughout the time spent in the Southern Ocean were -1.4 • C and -0.3 • C, respectively. The observed median wind speed at 10 m in the Southern Ocean was 9 m s −1 (interquartile range of 5.96), and the maximum wind speed at 10 m recorded in the 690 Southern Ocean was 26 m s −1 . The wind direction over the Southern Ocean corresponding to strong winds was mostly south and south/west as indicated by the wind barbs in Fig. 7(a). The southernmost latitude reached during the voyage was 73 • S. Figure 8 shows example temperature and relative humidity profiles between the ground and 17.5 km as measured by radiosondes, which were released south of 60 • S. Fog events associated with moist air trapped near the surface by low-level temperature inversions are visible on 15 February and 5 March 2018 in the radiosonde data. The tropopause is also clearly 695 visible in the temperature profiles at around 11 km (15 February), dropping to between 8 and 8.5 km further south. Above the pronounced tropopause lies the stable and dry stratosphere with temperatures around -50 • C.
Cloud observations are dominated by periods of complete cloud cover. In such cases, the all-sky camera and the ceilometer agree well with each other due to the lack of spatial variability (Fig. 9). When lower cloud fractions are observed and when 710 there is spatial variability, the agreement between the ceilometer derived cloud fractions and the camera is reduced compared to events with complete cloud cover. This is due to the limited area that the ceilometer uses to compute cloud fraction compared to the camera system. The former uses a time weighted average of cloud occurrence to infer the spatial cloud fraction, essentially assuming that the spatial variability at a given moment is equivalent to the temporal variability over the preceding period.
Furthermore, the difference can be also caused, in part, by the different geographical region sampled by the ceilometer and a 715 sky camera (directly at zenith versus all-sky).
Using the cloud base height product derived from both the ceilometer and MiniMPL raw data using the ALCF tool (Kuma et al., 2020b), it is possible to look at the frequency of cloud occurrence binned by height :: in :::: 200 m ::::::: intervals, : as shown in An example of the backscatter ratios measured by the two lidar instruments is shown in Fig. 12, wherein we demonstrate the differences in sampling between the two instruments. The MiniMPL was scanning over a range of elevation angles (see Sect. 3.3). The scans at lower elevation angles would saturate at a lower altitude due to the higher effective air mass being measured.

735
Thus the periodic structure observed in the MiniMPL shown in the upper panel of Fig. 12, while the ceilometer, which did not have elevation scanning functionality and only measured in the zenith direction, shows a more continuous time-series.
This particular day (3 March 2018), with its nearly unbroken cloud signal around 1 km, is representative of the overall cloud statistics from the voyage. The initial 2 hours (00:00-02:00 UTC) show surface level cloud or fog (Fig. 12). From 02:00 to 18:00 UTC, low level cloud between 1 and 1.5 km is present. At 18:00 UTC, in addition to the low level cloud, a higher cloud 740 layer at 5 km is observed along with probable precipitation as it descends to 2 km by the end of the 24 hour period being shown. A challenge with measurements of this type is that there may also have been other high cloud layers throughout the day but they were not seen through the saturated low level cloud layer.
Precipitation was monitored throughout the voyage, but with relatively low occurrence throughout. Figure 13 displays the radar reflectivity, vertical velocity and spectral width for a range of altitudes over one 24 hour period collected near 71 • S 745 derived using the scheme detailed in Maahn and Kollias (2012). Figure 13 also displays snowfall estimates at the surface derived from the MRR-2 data. Note that the corresponding in situ precipitation measurement device on R/V Tangaroa ::: RV :::::::: Tangaroa was not sensitive enough to snowfall to measure these very small rates of accumulation. The diagonal structures identified in Fig. 13 between approximately 19:00 UTC on 16 February and 01:00 UTC 17 February 2018 at altitudes above 2 km in the radar reflectivity are related to fall streaks, which represent the movement of precipitation towards the surface. The 750 upward and downward motions observed in Fig. 13(b) are a distinctive characteristic of snowfall. 36 4.2 Particle size distributions and cloud condensation nuclei

CO 2 measurements for identifying contamination events
Throughout the voyage the mole fractions of atmospheric CO 2 were measured continuously by using a Picarro CRDS (Sect. 3.7). While the sampling line of the Picarro was separate to the particulate sampling line, it was in close proximity (within 5 m).

755
Contamination from ship exhaust from the rear of the ship would have been sufficiently well-mixed in the turbulent air around the ship superstructure to affect both sampling lines. The use of CO 2 measurements together with wind speed and direction measurements are often used as a reliable method to identify periods of contamination in the air sampled by the sampling inlet for all aerosol measurements performed in the aerosol container lab.
Five minute mean CO 2 measurements for the entire voyage are shown in Fig. 14. Following initially high values : an :::::: initial 760 :::: high :::: value : at the start of the voyage, due to proximity to land :::::::::: (Wellington), atmospheric CO 2 concentration rapidly decreased to close to the baseline value of 403 ppm, which was observed at NIWA's Baring Head atmospheric station at the time of the voyage. This baseline value is consistent with the voyage being conducted within the Southern Ocean/Antarctic source region for air selected for baseline analysis at Baring Head (Brailsford et al., 2012).
A large number of brief episodes of high CO 2 concentration (to >500 ppm) are apparent in the CO 2 data set shown in generators being blown back towards the airline intake above the bridge in certain wind conditions. During the voyage the DPS was operated during Deep-Towed Imaging System (DTIS) deployments. Two tests were used to identify these exhaust contamination events in the Picarro CO 2 data, i.e. CO 2 measurements were deemed as pollution events if: 1. the CO 2 standard deviation of the 5 minute mean was greater than 0.1 ppm, AND 770 2. the CO 2 5 minute mean was more than 0.1 ppm above the calculated 50-point median filter that was applied to the CO 2 5 minute mean data.
In Fig. 14, these exhaust contamination conditions are indicated in red, while data considered as good are indicated in blue.
The exhaust contamination tests are effective in identifying all of the data points attributable to exhaust contamination, at the expense of including a small number of points that may be considered good data.

775
All particle measurements described below were screened according to contamination events using the method described here or in their respective sections above.

785
The quality assured measurements from all of the aerosol instruments operated throughout the voyage are shown in Fig. 16a-e. In Fig. 16a, the time series of the PCASP-100X observations is shown. Note that the concentrations in each size range have been normalized by the log width of the size bin. A notable instrument artefact within the PCASP-100X measurements is the persistent local peak in concentrations between 0.5-0.6 µm. Similar to the lack of particles observed in the 5 th size bin (see Sect. 3.8.1), this is likely a result of gain stitching errors between the multiple linear amplifiers the PCASP uses to detect 790 particles across such a broad range of sizes. The user may choose to exclude this size bin in further analysis.
In Fig. 17, the median particle concentration size spectrum measured by the PCASP-100X, SMPS, and NAIS is shown for the whole voyage. This spectrum can be used to compare particle concentration measurements between the various particle counters. Overall, there was reasonably good agreement between the particle size distributions measured by the PCASP-100X and SMPS 3936. However, on average the PCASP reported 1.6 times as many particles in the 100-300 nm range as the SMPS, 795 and it is recommended that the SMPS data are used in this size range.
In Fig. 17, it appears that there is significant disagreement between the SMPS3936 and the NAIS in the 10-15 nm particle size range. However, this is most likely a result of additional deposition of these particles within the sampling conduit and inefficient transmission through the SMPS itself. The NAIS measurements, which were conducted from the mast of the ship, are likely more accurate in this size range. The SMPS data for particles smaller than 20 nm are available, but should be 800 interpreted with caution.
The highest DMS concentrations were measured in the Eastern Ross Sea, in the transect between Iselin Bank and Scott Island, 875 with a maximum concentration of 27 nmol L −1 (Fig. 6).
The DMS sea-air flux estimates (F DM S ) were derived by applying the COARE gas exchange coefficient for DMS to the DMS gradient at the ocean surface (∆DM S): were k DM S,COARE is the gas exchange coefficient for DMS. The sea-air DMS concentration difference ∆DM S is equivalent 880 to: where H DM S is the temperature dependent dimensionless Henry's law solubility coefficient for DMS (Dacey et al., 1984), DM S w is the measured DMS concentration in seawater and DM S a is the DMS concentrations in air. The transfer velocity k DM S,COARE was calculated using the NOAA COAREG version 3.6 algorithm (Fairall et al., 2003(Fairall et al., , 2011Blomquist et al., 885 2006) and parameterized in terms of local wind speed scaled to 10 m height as described in Bell et al. (2015). The transfer velocity k DM S,COARE was then adapted for DMS using the Schmidt number for local seawater temperature and salinity at 6.0 m depth (Saltzman et al., 1993). For the flux calculations the DM S a concentrations were set to zero as the atmospheric concentration is negligible compared to the concentrations in the ocean surface (ppt to nmol L −1 ).
As shown in Fig. 22 the transfer velocity is strongly dependent on wind speed. There is a positive correlation for the data     Fig. 6(d).
MICA observations for the period 16 February to 1 March 2018, the time period without significant interruptions in either air or seawater sampling, are shown in Fig. 24. For OCS, atmospheric mixing ratios remain nearly constant around 500 ppt and 905 dissolved concentrations vary between 20 and 60 pmol dm −3 . OCS is nearly always supersaturated and follows a characteristic diel cycle of a photochemically produced gas. Within the region sampled between 16 February to 1 March, uncalibrated fluorescent dissolved organic matter (fDOM) data from an in-line sensor show diel variability but low spatial variability during this period (data not shown). fDOM refers to the fraction of CDOM (chromophoric dissolved organic matter) that fluoresces.
As expected, Fig. 24 shows a relationship between OCS concentration and irradiance, as CDOM is the main precursor to OCS 910 photoproduction (Ferek and Andreae, 1984). Besides photoproduction, wind speed (red line in Fig. 24) is a key driver to the observed variability of integrated daily fluxes (grey bars and numbers in Fig. 24). Daily fluxes are derived using the sea-air gas exchange parameterisation of Nightingale et al. (2000), and were integrated from 12:00 pm to 12:00 pm UTC. In the cold sub-Antarctic waters, the strongly temperature dependent OCS hydrolysis (Elliott et al., 1989) becomes slow with a lifetime of several days, and sea-air exchange becomes the dominant OCS removal process in the surface seawater. This explains the 915 observed behaviour of dissolved OCS concentration with high supersaturation, only building up at low to moderate winds when photoproduction is greater than removal, and high OCS fluxes often coincide with lower seawater concentrations on windy days. Observations from the TAN1802 voyage will be used to assess whether the behaviour of OCS in the Southern Ocean is adequately represented by a state of the art photochemical model (Lennartz et al., 2017). A specific model setup forced with high resolution observations made during the cruise will help to improve and fine tune the model.

920
Besides OCS, MICA also measured CO and CO 2 with spikes related to contamination by the ship's exhaust are removed from the data set. Atmospheric CO mixing ratios are, on average, 27 ppb. which is 10-20 ppb lower than expected even for the pristine air in this region (e.g. Novelli et al., 1998). While we can not irrevocably rule out an artifact, we found no indication in the raw spectra or during calibrations for a measurement error beyond the 10 ppb accuracy. Dissolved CO concentrations in the nM range agree with earlier CO measurements in the Southern Ocean (Williams and Bainbridge, 1973;Swinnerton and 925 Lamontagne, 1974;Bates et al., 1995;Wingenter et al., 2004). CO is also photochemically produced from CDOM (Wilson et al., 1970;Stubbins et al., 2006), but the low amplitude of the diel cycle and the sustained high supersaturation ratios of 10-80 even on days with high wind and moderate irradiation suggest significant production mechanisms in addition to photochemical production. Atmospheric CO 2 mixing ratios were close to 400 ppm throughout the cruise, which agrees with the Picarro measurements (Sect. 4.2.1) within uncertainties.
Two data files are provided for the MICA data: (i) atmospheric and dissolved OCS, CO and CO 2 concentrations at 2 minute temporal resolution and (ii) OCS sea-air flux at 1 hour temporal resolution (see Table A1). Note that OCS observations from TAN1802 have already been included in a long term global data set of ship based observations of atmospheric and dissolved OCS published by Lennartz et al. (2020).

935
Ground-based and ship-based measurements of cloud and aerosol properties over the remote Southern Ocean are sparse such that satellite-based measurements are the primary source of data in this region. However, satellite-based measurements are inherently limited in their utility in several ways, e.g., while CCN concentrations can be indirectly estimated, they can not be accurately determined from satellite-based measurements. As a result, many questions can only be addressed using in situ and remote sensing ground-based and/or ship-based measurements that observe the atmosphere from below. Incomplete under-940 standing of aerosol-cloud interactions over the Southern Ocean leads to a misrepresentation of aerosol and clouds processes in climate models. Such misrepresentations are manifested as biases in the representation of precipitation and radiation by climate models over the Southern Ocean.
A comprehensive description of meteorological, aerosol, cloud and precipitation measurements, made using a suite of sensors on board the New Zealand research vessel Tangaora during a six-week voyage over the Southern Ocean during February 945 and March 2018, has been presented above. These ship-borne measurements are an important supplement to satellite-based measurements as they provide data on low-level clouds and aerosol composition in the marine boundary layer that cannot be inferred from satellite-based measurements alone. As such, the ship-borne measurements can be used to investigate some of the processes that lead to biases in climate model representations of cloud-aerosol interactions that would otherwise not be amenable to diagnosis from satellite-based measurements alone. When both satellite-and surface-based measurements 950 are used in conjunction with model studies, the synoptically varying vertical structure of Southern Ocean boundary layer and clouds, as well as variability of sources and sinks of CCN, aerosols, and the role of local biogenic sources can be investigated.

Code availability
The mpl2nc source code to convert raw MiniMPL data files to netCDF files is available at https://github.com/peterkuma/ mpl2nc. The ALCF open source command line tool for processing of automatic lidar and ceilometer (ALC) data is availble 955 at https://alcf-lidar.github.io/. The tool to convert micro rain radar data into netCDF format is available from https://github. com/peterkuma/mrr2c. The COARE gas exchange algorithm to calculate the transfer velocity for sea-air flux estimates can be obtained from the NOAA ftp server (ftp1.esrl.noaa.gov//BLO/Air-Sea/bulkalg/cor3_6). A Matlab script that can be used to run the COARE code is available in the ReadMe file that is provided with all data from the DAS. Open-source software to convert native radiosonde data into netCDF format is available at https://github.com/peterkuma/rstool.

Data availability
The TAN1802 voyage measurements described in this study are publicly available in netCDF format from Zenodo at https: //doi.org/10.5281/zenodo.4060237 (Kremser et al., 2020). These are packaged in a set of product ZIP archives by instrument and processing level (see also Appendix A: Data products overview Table A1. Overview of data products available from the Zenodo TAN1802 data archive for different processing levels, i.e. level 0: Raw (unformatted) data, level 1: Raw data formatted into netCDF format and quality controlled as described in the main text of the paper, and   Author contributions. All co-authors contributed data from one or more instruments and provided relevant figures and material for the manuscript. PK participated in the organisation of the voyage and deployment of instruments, performed observations during the voyage, post-processed a part of the data set, and developed the mpl2nc, mrr2c, rstool and ALCF software packages. SH maintained and ran the 970 aerosol instruments during the voyage and prepared all aerosol data sets (except from the NAIS instrument). MP prepared the NAIS data, and together with KS prepared the required material for the manuscript. KS shipped the NAIS from France and installed the NAIS on the Tangaroa prior to the voyage and performed remote quality checks of the data during voyage. JM participated in the voyage, and prepared and quality controlled the CO2 measurements. AM was responsible for collecting DMS and OCS measurements during the voyage. AS-M was in charge of the QA/QC of the dissolved DMS measurements, and together with MH and CSL provided the figures and material for the 975 paper. MH led the collaborative proposal for the aerosol-cloud component of the voyage and calculated the DMS fluxes. RQ provided the MiniMPL instrument for the voyage and processed the data, AG developed the allskypi system and software and prepared the allskypi data.
STL and MvH provided, prepared and quality checked the MICA instrument assembly and prepared MICA related data. AMcD prepared the rain radar data. IS and CF took part in the processing and calibration of the MiniMPL data. TH, PDeM, and CH analysed and provided the INP data. GG designed and build the particle sensing AlphaSense radiosonde equipment used for measurements with the UAV. SP prepared 980 and installed the meteorological equipment, such as ceilometer, micro rain radar, Brinno sky cameras and provided logistical support. SK wrote the manuscript with contributions from all co-authors.
Competing interests. The authors declare that they have no conflict of interest.