Southern Ocean latitudinal gradients of Cloud Condensation Nuclei

The Southern Ocean region is one of the most pristine in the world, and serves as an important proxy for the pre-industrial atmosphere. Improving our understanding of the natural processes in this region are likely to result in the largest reductions in the uncertainty of climate and earth system models. While remoteness from anthropogenic and continental sources is responsible for its clean atmosphere, this also results in the dearth of atmospheric observations in the region. Here we 5 present a statistical summary of the latitudinal gradient of aerosol and cloud condensation nuclei concentrations obtained from five voyages spanning the Southern Ocean between Australia and Antarctica from late-spring to early autumn (October to March) of the 2017/18 austral seasons. Three main regions of influence were identified: the northern sector (40-45S) where continental and anthropogenic sources added to the background marine aerosol populations; the mid-latitude sector (45-65S), where the aerosol populations reflected a mixture of biogenic and sea-salt aerosol; and the southern sector (65-70S), south 10 of the atmospheric Polar Front, where sea-salt aerosol concentrations were greatly reduced and aerosol populations were primarily biologically-derived sulfur species with a significant history in the Antarctic free-troposphere. The northern sector showed the highest number concentrations with median (25 to 75 percentiles) CN10 and CCN0.5 concentrations of 681 (388 839) cm−3 and 322 (105 443) cm−3, respectively. Concentrations in the mid-latitudes were typically around 350 cm−3 and 160 cm−3 for CN10 and CCN0.5, respectively. In the southern sector, concentrations rose markedly, reaching 447 (298 15 446) cm−3 and 232 (186 271) cm−3 for CN10 and CCN0.5, respectively. The aerosol composition in this sector was marked by a distinct drop in sea-salt and increase in both sulfate fraction and absolute concentrations, resulting in a substantially higher CCN0.5/CN10 activation ratio of 0.8 compared to around 0.4 for mid-latitudes. Long-term measurements at land-based research stations surrounding the Southern Ocean were found to be good representations at their respective latitudes i.e. CCN 1 https://doi.org/10.5194/acp-2020-1246 Preprint. Discussion started: 27 January 2021 c © Author(s) 2021. CC BY 4.0 License.

Both platforms host standard meteorological stations, deployed in duplicate, which run as part of the respective ongoing underway systems and whose data are used as part of the analyses in this manuscript. Underway data, together with their associated metadata, are publicly available (Facility, 2018;Symons, 2019a, b, c, d).

Aerosol measurements
0.3, 0.2% supersaturation. Similar quality control procedures were undertaken for this instrument as for the MARCUS data, however because this instrument was calibrated at the Droplet Measurement Technologies laboratory in Colorado, pressure corrections for the supersaturations were made which resulted in the actual measured supersaturations being 0.055% higher 135 than the set points (e.g., 0.5% was actually 0.555%). This was not the case with MARCUS data since calibrations were undertaken at sea level. Flows on both instruments were set to 0.5 Lmin −1 and checked regularly to ensure they remained within specification.
Throughout this study, the two supersaturations common to both campaigns were utilised in our data analysis -0.2 and 0.5% (referred to throughout this manuscript as CCN 0.2 and CCN 0.5 , respectively). It is noted however that recent aircraft 140 measurements in the region suggests that 0.3% is likely to be the best representation of actual environmental conditions in the region (Fossum et al., 2018;Sanchez et al., 2020;Twohy et al., 2020).
The full metadata record and measurement data for the MARCUS campaign are available at Kulkarni et al. (2018), however these data have not been filtered for exhaust contamination. An exhaust filtered and reprocessed dataset was undertaken specifically for this study and data are available at (Humphries, 2020). Fully processed and exhaust filtered data for CAPRICORN2 145 are available at (Humphries et al., 2020a).

Condensation Nuclei
Number concentrations of condensation nuclei (aerosols) larger than 10 nm (CN 10 ) were measured continuously at 1 Hz on both platforms using condensation particle counters (CPC Model 3772, TSI Inc. Shoreview, MN, USA). The CPC draws sample air continuously through a chamber of supersaturated 1-butanol, which condenses and grows particles to super-micron sizes 150 where they are counted individually by a simple optical particle counter. For this study, the manufacturers default 50% counting efficiency (D-50) was used, and is defined at 10 nm. The sample flow rate is typically regulated by an internal critical orifice (MARCUS instrument was configured this way). However the critical orifice in the CAPRICORN2 instrument was replaced with a mass flow controller (MFC, Alicat Scientific Model MC 5SLPM) to ensure more accurate flow control, particularly in a marine environment, where the critical orifice can becomes quickly blocked with sea-salt aerosol. The MFC was calibrated impacting onto a vaporiser heated to 600 o C where non-refractory particles are vaporised and then ionised with electron impact ionisation. Ions are then directed into a time-of-flight mass spectrometer (0-400 amu) resulting in 1 Hz mass spectra.
Aerosol spectra are identified above background air by continuously switching between particle-free (through a HEPA filter) and sample air every 20 seconds. Aerosol spectra are used to calculate 10 min averages of sulfate, nitrate, ammonium, chlorine, methanesulfonic acid (MSA) and a grouped 'organics' class. It is important to note that because of the size selection and the 170 refractory nature of sea-salt, the actual concentrations reported for chlorine have yet to be calibrated to obtain a correction factor and values are only used in a relative manner in this manuscript. Ammonium nitrate and sulfate calibrations were run prior to and after the voyage, as is standard operating procedure. During CAPRICORN2, time integrated aerosol composition measurements were made alongside the on-line ACSM measurements using a PM1 size selective inlet (BGI SCC model 2.229, Butler, NJ, USA) and 47 mm quartz filters. Each filter 175 sampled for 1-2 days (20-48 hours) at a flow rate of 16.67 vLPM (required for the SSC) controlled by a MFC (Alicat Scientific Model MC 20SLPM). To prevent the filters being contaminated (and overwhelmed) by exhaust aerosol, the system was placed on a switching controller which ceased sampling when relative winds directions were between 90 o and 270 o and CN concentrations were above a threshold value. This meant the PM1 sampling system was switched on and off throughout the sampling period so that total volumes through each filter ranged between 14 to 26 m 3 . The instantaneous volumetric flow rate 180 from the MFC was recorded and totaled by a electronic flow totalizer (Amalgamated Instruments Co., model PM4-IVT-DC-8E, Hornsby, NSW, Australia).
Five field blanks were collected approximately weekly throughout the campaign. Field blanks involved carrying out the full filter change process with the sample pumps remaining switched off. After sampling, filters were enclosed in clean aluminium foil and frozen until they could be analysed post-voyage. The soluble ion concentrations were determined using high per-185 formance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) was measured at the CSIRO laboratories in Aspendale, Victoria. The filters were be extracted in 10 ml of 18.2 mΩ de-ionized water. The sample was then preserved using 1% chloroform. Anion and cation concentrations are determined with a Dionex ICS-3000 reagent free ion chromatograph. Anions are separated using a Dionex AS17c analytical column (2 x 250 mm), an ASRS-300 suppressor and a gradient eluent of 0.75 mM to 35 mM potassium hydroxide. Cations are separated using a Dionex CS12a column (2 x 250 190 mm), a CSRS-300 suppressor and an isocratic eluent of 20 mM methanesulfonic acid. All values reported in this manuscript are blank corrected.
Aerosol composition data measured by the Tof-ACSM and on the PM1 filters during CAPRICORN2 are available at Humphries et al. (2020a).

Platform exhaust 195
Removal of exhaust contaminated data is a critical step required before using any aerosol composition data from dieselpowered ship platforms. For aerosol data, exhaust signals are typically orders of magnitude higher than ambient data, given the strength and proximity of the source to measurement points. The engine age, fuel type, ship architecture (e.g., how the air flows around the ship and creates local eddies), relative locations of exhaust and air sampling inlet as well as the operations of the ship during measurements, also affect how much impact the exhaust has on the atmospheric measurements. MARCUS was undertaken aboard the Aurora Australis, an ice-breaker commissioned in 1989, powered by two Martsila medium-speed diesel engines (one 16V32D and one 12V32D, producing a total of 10 000 kW) and burning standard marine grade fuel oil. As shown in Figure A1, the ARM measurement container was located directly adjacent to the exhaust pipe of the ship, meaning that a large proportion of wind conditions were able to push exhaust into the sampling inlet. The MARCUS campaign was also supplementary to the usual resupply voyages of the Australian Antarctic Program, and consequently, the direction of the ship 205 was rarely optimal for atmospheric measurements -instead being more focused on swell and sea-ice conditions.
In comparison, the CAPRICORN2 voyage was undertaken aboard the RV Investigator, which was purpose built for marine science, and specifically incorporated atmospheric measurements into its design, resulting in an architecture optimised for minimal exhaust impact. The atmospheric sampling inlet is located as far forward on the vessel as possible (see Figure A1), resulting in a significant distance from the exhaust. The ship itself is powered by three nine-cylinder MaK diesel engines 210 coupled to a 690V AC generator, and burns automotive-grade diesel fuel (as opposed to standard marine grade fuel oil used by most vessels). During the CAPRICORN2 voyage, the ship was largely positioned to face directly into the wind, ideal for marine Conductivity, Temperature and Depth (CTD) measurements occurring during the voyage, and where possible, transits between marine targets were optimised for favourable wind conditions to maximise the collection of exhaust-free atmospheric data.

215
On both platforms, wind direction was found to be a poor parameter for filtering exhaust (Humphries et al., 2019), likely because of the eddies that form around the ship's superstructure and create differences between wind directions measured by meteorological instruments and those experienced at measurement height. Instead, we used differences in composition between the exhaust and clean background air to identify and remove exhaust influence. The exhaust was identified the same way on both platforms. A first pass was undertaken using the automated exhaust identification algorithm described in (Humphries 220 et al., 2019). This algorithm was designed to strike a balance between accurately removing obvious exhaust signals, but not being too over-zealous and unwittingly removing clean data. This balance results in the correct removal of about 95% of exhaust signals. Manual filtering is undertaken as the next step, and identifies any rapid increases in CN, black carbon, carbon monoxide or carbon dioxide concentrations, and in the case of the Aurora Australis, any drops in ozone resulting from titration from engine-produced nitrogen oxides. It is noted here that the CN data is relied on most heavily during this process, because 225 of its high time resolution and highest sensitivity to the exhaust signal, which typically changes by orders of magnitude relative to background data.
Because of the differences between the platforms and ship operations, the resulting proportion of clean data available for each campaign was significantly different. For MARCUS, only 11% of data was exhaust-free, resulting in approximately 500 hourly data points for CCN over the 129 days at sea. In contrast, over 86% of data during CAPRICORNII was exhaust free, 230 resulting in over 760 hourly data points from the 42 day campaign. An example time series of MARCUS CCN data and the amount of data removed by exhaust filtering is shown in Figure A2. Overlayed on Figure 1 ship tracks are the locations where exhaust-free data exists for the campaigns. Despite the significant data loss associated with exhaust contamination, the latitudinal coverage of the data is reasonable, with each of the 5 o latitudinal bins having over 110 data points in each, with the bins south of 60 o S containing over 450 data points (as shown in Figure 2).

Trajectory analyses
The HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) model (Draxler and Hess, 1998) was used to calculate air parcel trajectories. In this study, HYSPLIT was used to calculate back trajectories in order to evaluate source regional and vertical source locations for the various categories. Trajectories were calculated using the Global Data Assimilation System (GDAS) reanalysis (Kanamitsu, 1989). The model was setup to utilise 1 degree horizontal resolution reanalysis data and 240 vertical motion utilised the model vertical velocity method. Calculations utilised surface invariant geopotential, surface 10 m horizontal (U and V) winds, 2 m surface temperature, and U, V, W (vertical wind), temperature and humidity on pressure levels from 1000 to 20 hPa. Each trajectory calculation provided hourly three-dimensional air parcel locations for a total time-span of up to five days in order to limit uncertainty magnification. Trajectories were initiated at the ship's location for every hour of the cruise at heights of 10 m and represent heights above ground level.

245
Once calculated, trajectories were divided into 5 o latitudinal bins based on their starting location. Only trajectories corresponding to exhaust free aerosol data were used. For each set of latitudinally-binned trajectories, frequency plots were calculated by summing the number of times trajectories passed through a map, binned such that the horizontal resolution of the boxes was 0.5 o and with linearly spaced, 10 m vertical bins. Resulting plots were smoothed using the kernel-density estimations using Gaussian kernels (implemented with Python SciPy's guassian_kde function). The bandwidth was determined by 250 taking the average effective samples in each bin calculated using Scott's factor (Scott, 2015).
Total precipitation along each trajectory was calculated using ERA-5 reanalysis data (ECMWF, 2018). For each step in each trajectory, the precipitation at the time and location was retrieved from the reanalysis data, then the values for each step were summed, resulting in a single total precipitation value for each trajectory. As before, only trajectories corresponding to exhaust-free measurement periods were included in the analysis.

Results
Despite the significant removal of data due to exhaust in the MARCUS campaign, the utilisation of the CAPRICORN2 campaign data, which occurred at the same time in the same regional area, meant that division of the data into 5 o latitudinal bins resulted in sample numbers high enough to calculate robust statistics in each bin, while having reasonable latitudinal resolution.
In Figure 2, violin plots show measurements distribution in each latitudinal bin for each of CCN 0.2 , CCN 0.5 and CN 10 (full 260 statistics for each bin are presented in Table A1 and latitudinal gradients for each voyage for CCN 0.5 and CN 10 are presented in Figure A3 and Figure A5, respectively). The highest concentrations (means of 169, 322 and 681 cm −3 for CCN 0.2 , CCN 0.5 and CN 10 , respectively) for all parameters are, unsurprisingly, observed in the northern-most bin, 40-45 o S which is closest to the coast of Tasmania, Australia, resulting in increased continental and anthropogenic influence. Moving south, CN 10 concentrations appear to be stable (300-400 cm −3 ) from 45 o S to 65 o S. This isn't the case with CCN at both supersaturations, Watch station, being representative of a globally significant region, so although the location is a reasonable distance from the 275 ship measurements, we use them here with confidence. In Figure 2 non-baseline selected CCN data, and generally agree well with ship data in the respective latitudinal bin. Cape Grim baseline data appear to be lower than ship-based measurements in their respective latitudinal bin, which may be explained by ship data not being filtered for baseline criteria, and are consequently likely to include some level of continental/anthropogenic 285 influence. Non-baseline CN 10 data from Cape Grim are higher than those measured on the ship, and this is likely because of the influence of fine-mode aerosol emissions from the metropolitan region of Melbourne, as well as emissions from Tasmania, both of which can influence Cape Grim measurements in non-baseline conditions. Curiously, ship-based CCN data are similar to non-baseline Cape Grim data, and significantly higher than baseline data (which is actually similar to the higher latitude bins), which suggests ship-based measurements were influenced significantly by continental sources while measuring at these 290 latitudes, a result confirmed by trajectory analyses presented later in the manuscript.
Most striking in the latitudinal distribution is the statistically significant increase in all aerosol parameters in the southernmost bin along the Antarctic coastline (p .001 compared to 60-65 o S bin). While most pronounced in CCN 0.5 (mean concentrations increase by 50% compared to mid-latitudes with 30% increases for both CCN 0.2 and CN 10 ), corresponding changes are also apparent in the CCN/CN ratio, wind speed, and more significantly in precipitation, as shown in Figure A7. To explore this 295 apparent change in composition further, we examined more closely the CAPRICORN2 data which provided both real-time, and filter-based aerosol composition data. Latitudinal aerosol composition data from this voyage is shown in Figure 3 alongside binned wind speed and precipitation data. , whose relative contribution to aerosol mass (as measured by the ToF-ACSM) was around 60-70% at lower-latitudes, but increased to its maximum in the southern-most bin, reaching over 80%. We note here that while trends from the filters are similar to those observed on from the ToF-ACSM, the differences observed as likely the result of different sampling 305 techniques: the ToF-ACSM measuring non-refractory aerosol composition, while the filters are analysed for soluble ions. The increases in CCN ratio could be driven by a stronger source of sulfur precursors (sulfate and MSA derived from DMS) emitted from enhanced phytoplankton near the Antarctic continent ( Figure 1), but are likely to also be driven by a significant drop in precipitation which would preferentially scavenge CCN compared to other aerosols. Chloride, which is used as a proxy for sea spray aerosol, is observed to be dominant at lower latitudes (and varies proportionately to wind speed, Figure A9) but 310 reaches its minimum in the high latitude bin . This significant reduction in the high-latitude bin is consistent with the combined effect of decreased wind speeds and the occurrence of sea-ice covering the ocean surface, resulting in a substantially lower source strength which outweighs the reduced precipitation sink. Interestingly, comparison of the distributions of CCN with the sulfur and chloride composition measurements suggests that while sea-salt aerosol contributes an important baseline to CCN numbers, the variability, and in particular the vast population of CCN at high latitudes, is driven by sulfur-based aerosols.

315
To further understand the source regions of the observed latitudinal changes, we calculated the trajectories for each hour of exhaust-free data during the five voyages. In Figure 4, these trajectories, split into the six latitudinal bins based on each  mechanisms. The CCN/CN10 ratios for 0.2% and 0.5% supersaturation are shown in A) and B), respectively; the major aerosol chemical components from the ToF-ACSM in C); the total soluble ions from PM1 filters, shown in D); E) shows the wind speed measured onboard the vessel; and F) shows the total precipitation calculated using ERA5 reanalysis data along the backward trajectory for each measurement.
Note that on plot C, the sulfate is split to the left axis to enable better visibility of trends of other components.
trajectory's end location, are shown as density plots. As expected, the northern-most bin shows influence primarily from the marine boundary layer upwind of the measurements, and significant influence from both Tasmania and the more heavily is not so clear, suggesting that the same source as mid-latitudes is driving CCN concentrations at this supersaturation (presumably sea-salt). This would suggest that aerosols arising from anthropogenic and continental sources are less hygroscopic than sea salt, which is consistent with what is expected from the literature (e.g. Swietlicki et al., 2008).
The mid-latitude observations are consistent throughout a large range of latitudes, being dominated by sea-salt and sulfurbased aerosols ( Figure A9). Given the lack of any land-masses in this region, the primary aerosol sources are driven by wind-355 produced sea-salt, and secondary aerosol formation typically resulting from both local and long-range transport of aerosol precursors emitted from biological sources, chiefly DMS from phytoplankton. The dependence of sea-salt aerosol concentrations on wind speed and precipitation is striking, being directly and inversely proportional, respectively. This relationship breaks down in the high latitude bins where sea-ice cover impacts on the wind mechanism of sea-salt aerosol production.  Measurements have been made in other parts of the Antarctic sea ice (e.g. Davison et al., 1996;Fossum et al., 2018;Schmale et al., 2019). Typically these sectors do not show the step changes observed in the East Antarctic sea ice measurements, and instead are reasonably well represented by continental measurements (e.g. Asmi et al., 2010;Hansen et al., 2009;Hara et al., 2011;Ito, 1993;Järvinen et al., 2013;Koponen et al., 2003;Pant et al., 2011;Samson et al., 1990;Virkkula et al., 2009;Weller 370 et al., 2011;Hara et al., 2020)). During a campaign around the Antarctic Peninsula in summer 2015, Fossum et al. (2018) observed two distinct air-masses: those coming from continental Antarctica and those from the marine region north of the polar front. They found that, despite the differing composition of the two air-masses which reflected observations described Contrary to the West Antarctic region, the region of the East Antarctic coast included in this study is not well represented by measurements on the Antarctic continent itself, a phenomenon driven by well defined air-mass transport which isolates the continent from the sea ice region (Humphries et al., 2016). This result was true for springtime measurements, and further work by the same authors (currently unpublished), suggests that the meteorology that leads to this phenomenon may break 385 down both during summertime and around the Antarctic Peninsula. Since the majority of measurements in the region occur in summer and at either Antarctic stations or at lower latitudes, the East Antarctic coastal region remains one of the more poorly represented regions of the world.
The step change in aerosol properties at this high-latitude bin is consistent with the crossing of the Antarctic atmospheric polar front, as first described by Humphries et al. (2016) and observed by both Alroe et al. (2020) and Simmons et al. (2020).

390
While values observed in this manuscript are in line with those previously observed across the polar front (Alroe et al., 2020;Simmons et al., 2020), this data set adds confidence that the change observed in these previous studies is an enduring phenomenon across a wider range of East Antarctic longitudes and seasons. While the definition of the Antarctic polar cell is traditionally understood in terms of climatological averages, it is evident that a very real-time boundary exists that can be seen in the atmospheric composition observations, despite being very difficult to identify in variables by which the front is defined (i.e. meteorological). Because of this, and to create a distinction from the traditional defined meteorological front, we introduce a new term to define it here as the Atmospheric Compositional Front of Antarctica (ACFA), which represents the northern boundary of the region that extends south to approximately the Antarctic coastline -a region we term the Antarctic Sea Ice Atmospheric Compositional Zone (ASIACZ). It is important to note that while the aerosol properties in the ASIACZ is not captured by surface measurements on the Antarctic continent, nor those in the Southern Ocean mid-latitudes, trajectory 400 studies suggest that airmasses from this region travel both north and south, typically above the boundary layer (Humphries et al., 2016), making this an important region of exporting aerosols and precursors from a highly biologically productive region to other regions. This could help reconcile the predicted missing aerosol source in the wider region.
The ACFA is known to vary in time and space, and can be advected by the synoptic-scale meteorology. This is evident from the individual voyage latitudinal plots ( Figures A3 and A5) where the increases in the southernmost bin latitudes differ 405 depending on the voyage and location. This movement can even occur within a single voyage, as evidenced clearly from the CAPRICORN2 voyage data ( Figures A3 and A8), where the increase in CCN occurred at approximately 64 o S during one crossing at 150 o E, and 62.5 o S during the 140 o E transect. During this voyage, we also tried to intentionally cross the front while sampling south along the 132 o E meridian. However we were unable to locate the front even when when travelling further south ( 65 o S) of the ACFA's latitude just days before.

410
By investigating latitudinal gradients across the parts of the Southern Ocean, this work raises some important objectives for future work. A significant motivations for this work is to better inform and reconcile the radiation biases arising from poor representation of clouds in climate and earth system models. Hence relating these observations to recent cloud observations is important, and this work is well underway. Mace (2020) analyzed MARCUS and CAPRICORN2 data and found gradients in cloud droplet number concentrations in reasonable agreement with the gradients in CCN concentrations identified here.

415
Interestingly, Mace (2020) observed a bimodal distribution in cloud properties poleward of 62.5 o S. While one mode displayed properties of marine clouds from farther north, the second showed relatively high cloud droplet numbers and low effective radii.
The bimodality was inferred to be associated with changes in air mass properties (such as CN, CCN and aerosol chemistry), identified in previous work (Humphries et al., 2016), in case study events described in Mace (2020) and in further analysis of CAPRICORN 2 data currently underway. In particular these changes in air mass properties included those identified sys-420 tematically in this work (i.e. high CCN and CCN/CN and high sulfate and MSA indicative of biogenic aerosol sources) in the southern sector. The potential sensitivity of the cloud properties to these biogenic aerosol sources suggest a strong feedback with biological and photochemical activity in the region, an issue that warrants further and extensive investigation.
Further studies should also address the transition across the ACFA. A more detailed assessment of the datasets used in this manuscript is currently underway that focuses on the transition and the chemical and physical changes that occur in gases, 425 aerosols and clouds in this region, and will be described in a future publication. The comprehensive, continuous measurements aboard the RV Investigator also provide a perfect opportunity for understanding transition as the vessel frequently undertakes voyages into this region -albeit primarily limited to the summertime.
These conclusions are all based on the sector of the Southern Ocean between Australia and East Antarctic, and only valid for late spring and summer measurements. It is possible that different patterns may be observed in other sectors of the Southern 430 Ocean that are influenced by disparate continental influences, such as those around Africa and South America. Hence, it is important that future observation campaigns investigate these regions. While important circumpolar measurements such as those undertaken by Schmale et al. (2019) give an insight into this variability, campaign measurements are limited in their duration, and generally undertaken during the summertime (e.g. Humphries et al., 2016;Fossum et al., 2018). To avoid biases that may arise by applying these conclusions to other seasons, long-term measurements, such as those undertaken at Cape 435 Grim, Cape Point South Africa (Labuschagne et al., 2018) and the RV Investigator are needed across other longitudes of the Southern Ocean. These include remote sites such as Macquarie Island (measurements included here were limited to just two years) and research platforms that frequent the ASIACZ, such as the soon-to-be commissioned RVS Nuyina. Long term year round atmospheric aerosol data sets reveal important seasonal and annual variability, and the processes that contribute to this variability. Hence these data will be critical for ensuring reduced biases in modelling efforts. The ASIACZ region was not represented by either of these stations, and previous work suggests that measurements at research stations on the Antarctic continent are not reflective of this spatially significant region. Further measurements are important to 455 capture the spatial, seasonal and inter-annual variability across the different latitudes, as well as the longitudinal variability that is likely when investigating the Southern Ocean regions around Africa and South America.
Data availability. Datasets are available at relevant citations within the manuscript.  Figure A1. Ship schematics of both the Aurora Australis (top), used for the MARCUS campaign, and the RV Investigator (bottom), used for the CAPRICORN2 campaign. Contamination of samples by the ship's own exhaust is the primary driver in the removal of data, with only 11% (500 hours) and 86% (760 hours) of data remaining after exhaust filtering for MARCUS and CAPRICORN2 campaigns, respectively.
Exhaust contamination is driven by the proximity of the measurements to the exhaust, the age and cleanliness of the engine, together with ship operations during voyages and whether these operations align the ship with favourable wind directions that push the exhaust away from the sampling inlet. All these factors contributed to the high contamination of MARCUS data relative to CAPRICORN2 data.