Between January 2021 and March 2023, MAP-IO carried out a total of 16 campaigns aboard the Marion Dufresne, including six during which the aerosol instruments performed well, which are presented in this article (Fig. 1, Table 1). During these six campaigns, the spatial coverage of the Marion Dufresne extended from -10.65 to -60° S in latitude and from 31 to 83.2° E in longitude, thus encountering various environmental conditions (Fig. ). The start and end dates of the various campaigns and the instruments used are shown in Table 1.

The SWINGS campaign (Fig. 1; blue track) took place during the austral summer from 13 January to 8 March 2021. The aim was to collect CO₂, pH, and fish population density measurements at different latitudes and depths in the Southern Ocean. During this campaign, the Marion Dufresne route began along coastal regions passing south of Madagascar and southeast of Africa, then moved on to the open ocean, where it stopped at Crozet and then the Kerguelen Islands before returning to Réunion. The SCRATCH campaign (Fig. 1; red track) took place during austral winter and was carried out from 1 to 22 July 2021. The aim was to better characterize the region using a combination of biological and geological techniques. During this campaign, the vessel passed east of Madagascar and stayed in the northern part of the Mozambique Channel before going back to Réunion. The MAYOBS campaign (Fig. 1; green track) was carried out during austral winter from 13 September to 3 October 2021 as part of the monitoring of an underwater volcanic eruption that began around Mayotte in 2018. The route of the vessel was similar to the one during the SCRATCH campaign. The OP3 and OP4 campaigns (Fig. 1; orange and lime tracks) were carried out under the TAAF charter during austral spring from 28 October to 28 November 2021 and from 28 November to 30 December 2021, respectively. During these two campaigns, the Marion Dufresne took the same routes: Réunion–the Crozet Islands–the Kerguelen Islands–Amsterdam Island–Réunion. The OBSAUSTRAL campaign (Fig. 1; purple track) took place during the austral summer in 2023 from 18 January to 28 February. The objectives were the same as those of the SWINGS campaign, including the monitoring of seismic activity at ocean ridges and the identification of cetacean vocal signatures using hydrophones. The vessel's route was similar to the one during OP3 and OP4 with one exception: it went further to the south between the Crozet Islands and the Kerguelen Islands and further to the east between Amsterdam Island and Réunion.

Between 10 and 25° S, westward surface atmospheric and oceanic circulation prevails, characterized by southeast equatorial winds and the South Equatorial Current (SEC) (Fig. 1). In this latitude range, the average wind speed measured on board the Marion Dufresne is 6.7 ± 3.9 ms-1 (Fig. 2a), the average wave height is 2.5 ± 1.2 m (Fig. 2c), and the average nanophytoplankton abundance is 306.1 ± 90.2 cellscm-3 (Fig. 2c). Madagascar's high plateaus play a role in regional circulation by splitting the trade winds into two branches: a northeasterly flow, which notably affected the region of the SCRATCH and MAYOBS campaigns, and a southeasterly flow, which affected the beginning of the SWINGS campaign's route between Réunion and Madagascar. The oceanic circulation also divides into two branches at the northeastern tip of Madagascar, with one bypassing the island to the north (North East Madagascar Current, NEMC) and the other to the south (South East Madagascar Current, SEMC) (Fig. 1). Between the southern tip of Madagascar and 30° S, the wind direction progressively changes and becomes eastward when encountering strong westerlies typically extending from 35 to 60° S (Fig. 1). The SEMC flows toward the southeast coast of Africa, where it becomes the Agulhas Current (AC) (Fig. 1). The area extending southeast of Madagascar has been considered pristine in the literature (Fig. 1; blue-shaded area) . At 40° S, the AC is retroflected and goes eastward. It then becomes the South Indian Current (SIC) and flows northeastward (Fig. 1). The subantarctic front is located south of it. This current is characterized by a strong sea temperature and salinity gradient between the subtropical zone and the Antarctic zone, which marks the northern boundary of the Southern Ocean . The Marion Dufresne crossed the subantarctic front during the SWINGS, OBSAUSTRAL, OP3, and OP4 campaigns. In this area, phytoplankton blooms – driven by nitrate, phosphate, or iron water fertilization – were observed during the austral summer on the Crozet Islands (46° S, 51° E) and Kerguelen Plateau (49, 69° S) . Measurements of nanophytoplankton abundance were performed during SWINGS and OBSAUSTRAL, during the austral summer, and showed a clear signal of enhancement between 40–55° S (Fig. 2c). The polar front is further south between 55 and 60° S and marks the boundary between the warmer subantarctic water and the cold Antarctic water. The Southern Ocean is the roughest ocean on Earth due to the absence of land . Even in the austral summer, measured average wind speeds of 11 ms-1 and swells in excess of 3.5 m. During the SWINGS and OBSAUSTRAL campaigns, average winds also exceeded 10 ms-1, and average wave heights exceeded 5 m. Three major storms were documented during the SWINGS campaign in 2021 (Fig. 1a; black circles). The first one was located south of the Crozet Islands, the second one south of the Kerguelen Islands, and the third one north of the Kerguelen Islands. The maximum wind speed was 23, 33, and 27 ms-1, respectively. The predominant wind direction was from the northwest south of the Crozet Islands and the Kerguelen Islands and from the southwest north of the Kerguelen Islands (Fig. 2b). The maximum wave height was 14, 21, and 15 m, respectively (Fig. 2c). The nanophytoplankton abundance was lower (400–800 cellscm-3) along the storm tracks (Fig. 2d) due to a more intense mixing that deepened the ocean mixing layer and limited the light available for phytoplankton growth . It is important to note that the local wind direction observed during the campaigns is highly variable and differs from the general wind circulation presented in Fig. 1. Therefore, in order to conduct a thorough analysis of particular periods within the campaigns, it is imperative to pay close attention to the local measurements of wind. For more information on the climatology of the southwestern Indian Ocean and the Southern Ocean, see for example the studies by and .

In the framework of the MAP-IO program, the Marion Dufresne has been equipped with 19 measurement and remote sensing instruments, described in . Among the 19 instruments, 7 are dedicated to the study of aerosols and atmospheric gas. In Fig. 3, the top-left and top-right pictures present the aerosol and gas inlets located at the center of the vessel, upstream of the exhaust stack and 21 m above the sea level. The bottom picture of Fig. 3 shows the aerosol and gas acquisition system and analyzers located in the meteorological laboratory and mounted on a shock-absorbing table.

Aerosol inflow enters the same sampling line, which, after passing through a Nafion tube used to dry the aerosol, divides and directs the flow toward the different instruments. The distance between the inlets and the instruments (8 m) was carefully chosen to minimize aerosol losses along the sampling line. To minimize the loss of the largest particles (> 1 µm), three OPC-N3 sensors are installed directly outside on the main deck. A water-based condensation particle counter (CPC; MAGICTM-200/210) measures the particle number concentration within the size range of 5 nm to 2.5 µm using a condensational growth system. A scanning mobility particle sizer (SMPS; 4S) is used to measure the number size distribution of aerosols. It is composed of a differential mobility particle sizer (DMPS) and a CPC (MAGICTM-200/210). The DMPS takes in the aerosol flow composed of dry particles and cloud droplets, and the water of the latter is evaporated while penetrating the inlet. The aerosol population is first neutralized using an X-ray neutralizer (TSI) to provide the same electrical charge to all particles, then selected using their size-dependent electrical mobility and classified into 133 bins from 3 to 350 nm. The number of aerosols per bin is then determined by the CPC. The whole size range is scanned in 5 min.

To complete the aerosol size distribution toward supermicron diameters, three optical particle counters (OPC-N3, Alphasense) are used to measure the number concentration and the size of particles within the size range of 0.35 to 40 µm. Using a calibration based on Mie theory, the OPC-N3 measures the light scattered by each particle of a sample airflow (sample flow rate of 210 mLmin-1) passing through a 658 nm laser beam. The particle size and concentration are determined according to the intensity of the scattered light and allow the classification of each particle into one of the 24 bins covering the measurement size range at a rate of about 10 000 particles s⁻¹. Generally, 100 % of the particles are detected at 0.35 µm and 50 % at 40 µm (Alphasense user manual, http://www.alphasense.com, last access: 1 September 2025). In our case, as the OPC-N3 sensors have no sampling line and are directly located outside, there is no loss of large aerosols, except on the walls of the instruments. To study the aerosol activation properties, a cloud condensation nuclei counter (CCN-100, DMT) is used to measure the number concentration of activated aerosols at different supersaturation levels (SS of 0.1 %, 0.2 %, 0.3 %, 0.4 %, 0.5 %, 0.6 %, 0.8 %, and 1.0 %), with the first two set for 15 min and the others for 5 min. Supersaturation is created by three temperature control zone columns mounted vertically. The aerosol sample enters at the top of the column and becomes progressively supersaturated with water vapor as it goes through the column. Activated droplets are then counted by an internal OPC and distributed into 20 size bins going from 0.75 to 10 µm. Gas measurements are realized by an NO_x analyzer (Teledyne N500 CAPS), an O₃ analyzer (HORIBA APOA-370), and a Picarro analyzer (Picarro G2401 CFKADS-2372) measuring atmospheric gas traces as CO, CO₂, CH₄, and H₂O (ppb). The gas inlets are located to the right of the aerosol inlet (Fig. 3b). Along with aerosol measurements, wind speed (ms-1), and wind direction (°), air temperature (°C), pressure (hPa), and humidity (%) are recorded by the Vaisala and Mercury meteorological stations with a sampling time step of 5 s and 1 min, respectively. Sea surface elevation (m) and ship position are also recorded by the inertial unit of the ship at a sampling time step of 1 s.

First, the data are filtered to remove all the measurements likely to be contaminated by the exhaust of the vessel. Observations of time series of the total number concentration and the relative wind direction concluded that measurements realized within the direction range of 90–225° were associated with high aerosol concentrations (> 5000 cm-3) lasting over time. A second filter – combining relatively low wind speed, vessel cruise speed, and gas concentration data – was applied in an attempt to remove local episodes of pollution. Thus, all data collected at wind speeds below 2 ms-1, which may have inhibited pollution plume dilution around the inlets, were removed. In addition, data were removed when the ship was stationary, thus eliminating contamination by maintenance activities on board (painting, rust removal). To complete the filter, data with corresponding high concentrations of NO and NO₂ (> 1 ppbv) were removed, with the latter being relatively low (< 1 ppbv) in remote marine environments. CO peaks in concentration were also identified following the semi-automatic method of detection described in . The use of different chemical tracers was not only useful when one of the analyzers did not work properly, but also useful for confirming the good agreement between NO_x and CO concentrations. A third step consisted of the quality control of the aerosol data. For SMPS data, the particle size distributions were filtered manually for each measurement to remove non-physical variations (e.g., local pollution undetected by the dynamic and chemical filters). The latter were noticeable by an abrupt and short increase in the concentration of aerosols over the entire range of diameters, known as “spikes”. SMPS data were quality controlled and corrected over the sampling periods using CPC data.

The original raw SMPS dataset consisted of 55 144 measurements of aerosols, which represent ∼ 192 d of measurements. After the first and second filtering steps, it consisted of 53 201 measurements. After the manual filtering step, 36 226 measurements were conserved, which represent ∼ 66 % of the original dataset. In total, 29 376 measurements were taken in 2021 and 6850 measurements in 2023, as shown in Fig. 1. The start and end dates of each campaign are shown in Table 1. The data collected in 2022 were not used in this study due to SMPS maintenance and calibration.

First, the CCN-100 data at 0.2 % and 0.4 % supersaturation were treated separately, and only the data recorded on the plateau (measurements realized approximately 3 min after a supersaturation change according to our experience) were averaged over 5 min to make them comparable to SMPS data. Assuming that aerosols are internally mixed and that larger particles are activated preferentially before the smallest ones due to the curvature effect, we can determine the hygroscopicity of aerosols at the activation diameter by calculating activation diameters and deriving the hygroscopicity parameter at both supersaturation levels. Activation diameters were then calculated using the total number of cloud condensation nuclei given by the CCN-100 and the number concentration of aerosols measured in each bin by the SMPS. The number of aerosols per bin was integrated from the largest diameter to the smallest until it matched the number of CCN. The activation diameter corresponds to the smallest diameter. The hygroscopicity parameter, kappa–Köhler, was derived from the previously determined activation diameters as follows : 1κ=4A327Dd3ln⁡2Sc,2A=4σs/aMwRTρw, where A is composed of σs/a = 0.072 Jm-2, the surface tension of pure water; Mw = 0.018 kgmol-1 is the molecular weight of water; R = 8.314 Jmol-1K-1 is the perfect gas constant; T is the activation temperature in K; and ρw = 997 kgm-3 is the density of water. Dd is the activation diameter in nm calculated for a given supersaturation Sc. Note that hygroscopic particles have theoretical kappa values between 0.5 and 1.4. summarized several κ values derived from CCN measurements. κ values are 1.28 and 0.9 for NaCl and H₂SO₄, respectively. When the Na+ ion is associated with other compounds, its κ values decrease and range between 0.8 and 0.88. Ammonium sulfate has κ values between 0.61 and 0.67. Organic compounds have κ values between 0.01 and 0.4, and hydrophobic aerosols, such as black carbon, have values of 0.

Back trajectories provide additional information to understand the origin of the air mass and thus help to better understand the observed aerosol size distribution and properties. For this purpose, the FLEXible PARTicle (FLEXPART) version 10.4 model was used from the vessel's position.

The FLEXPART model is part of the multi-scale offline Lagrangian particle dispersion models (LPDMs), which have been developed to simulate transport and turbulent mixing of aerosols and gases in the atmosphere. The model can be run for forward or backward simulations. In backward mode, the location where particles are released, called a “receptor”, is defined within a longitude–latitude–altitude cell . For this study, ERA5 meteorological fields with a spatial resolution of 0.5° × 0.5° and time resolution of 1 h were used as input, and 5 d back trajectories were run each hour of the campaign at the location of the vessel with a particle release between 0 and 50 m altitude. To analyze the dataset according to the air mass origin, the area covered by the vessel was divided into three subdomains related to the main types of air masses. Continental air masses were identified as such when their residence time over the area north of 40° S and west of 50° E was the greatest. Air masses were classified into the subtropical Indian Ocean group when their residence time over the area east of 50° E and north of 40° S was the greatest. Thus, southern Indian Ocean air masses were identified as such when their residence time over the area south of 40° S was the greatest. This classification led to 15 385 receptors in the southern Indian Ocean group (with a spatial extension of 20–60° S and 35–80° E), 5495 receptors in the subtropical Indian Ocean group (with a spatial extension of 20–50° S and 55–80° E), and 4048 receptors in the continental group (mostly located north of Madagascar and around Réunion). This classification also resulted in a mixed zone where all three types of air mass are present in the middle of the subtropical Indian Ocean. This classification resulted in 6273, 18 647, and 9049 SMPS data points (0.02–0.35 µm) and 2305, 13 727, and 5554 OPC-N3 data points (0.41–5.85 µm) used in the calculation of the average aerosol size distribution of the continental, southern Indian Ocean, and subtropical Indian Ocean air masses, respectively.

A case of pristine air mass observed on 19 and 20 February 2023 is shown in Fig. 6 and is associated with the low NSMPS observed along the easternmost transect of OBSAUSTRAL (Fig. 4a; label 1). During this period, the air masses arriving at the ship's position evolved over the subtropical Indian Ocean, far from the continents. The height of the back trajectory shows that the air mass remained in the oceanic boundary layer, i.e., below 1000 m, for the 3 d prior to the measurement (Fig. 6a to d). During this period, the aerosol size distribution is very stable, with a clearly visible Aitken mode around 50 nm and a weak and broad accumulation mode between 150 and 300 nm. This large accumulation mode mean diameter and the lack of mode evolution indicate that the air mass is aged (Fig. 6e). NSMPS and NCCN remained low, between 150–250 and 40–120 cm-3, respectively (Fig. 6f). Throughout the day, the activation diameter is between 60 and 140 nm and lies between the Aitken mode and the accumulation mode in a very low concentration zone close to the measurement noise. The κ values obtained range from 0.1 to 0.5 at 0.2 % SS and from 0.1 to 0.3 at 0.4 % SS. Specifically, when NSMPS and NCCN concentrations are exceptionally low, it has been observed to induce a decrease in κ. This decrease occurs without any discernible pattern in the particle size distribution or in the origin of the air mass. This situation illustrates a case where the calculation of κ based on the combination of two instruments exhibits an important sensitivity to low concentrations. Consequently, the resultant value should be considered with caution.

Figure 7 shows the evolution of the aerosol particle size distribution (Fig. 7e), NSMPS concentrations (Fig. 7f), and CO concentrations (Fig. 7h) between 16 and 19 July 2021. The origin of the air mass is also represented by the height of the back trajectory over the period (Fig. 7a to d). During this period, CO and NSMPS concentrations remained high (> 50 ppb and > 2000 cm-3, respectively), showing that the area was generally affected by residual continental pollution. The activation diameter remained high (∼ 128 nm at 0.2 % SS and ∼ 92 nm at 0.4 % SS), and κ values were less than 0.19. Several peaks of NSMPS > 6000 cm-3 were also measured periodically and are associated with a significant increase in CO concentrations of 10 to 15 ppb, indicating that the air mass was affected by anthropogenic pollution. The aerosol size distribution shows that these NSMPS peaks are associated with an increase in the number of aerosols in the Aitken mode (30 to 40 nm) and the appearance of a new nanoparticle mode located between 5 and 30 nm. The back-trajectory analysis clearly shows the passage of the air mass at about 800 m a.g.l. over the urbanized region of Majunga, northwest Madagascar. The temporal evolution of the mixing boundary layer (MBL) thickness over Madagascar, corrected for transport time, is shown in Fig. 7h. There is a clear correlation between the MBL thickness and the CO and NSMPS peaks: each concentration peak is associated with an MBL thickness greater than 800 m. This means that only the upper part of the polluted boundary layer over Madagascar can reach the ship by subsidence. This can only happen in the afternoon when the boundary layer is sufficiently developed. This situation shows that the high variability in aerosol concentration between 800 and 8000 cm-3 observed in the northern Mozambique Channel (Fig. 4a; label 2) can be explained by a residual polluted air mass that evolves with the diurnal evolution of the MBL over Madagascar. This important aerosol number has a small impact on NCCN. Moderate NCCN observed range from 100 to 500 cm-3 and are attributed to elevated activation diameter values (> 100 nm) that exceed the residual continental pollution mode. The resulting κ is low, ranging from 0.06 to 0.1 at 0.4 % SS and from 0.1 to 0.2 at 0.2 % SS. This indicates an air mass composed primarily of hydrophobic aerosols.

The third period highlighted corresponds to a case of primary aerosol emission observed during the SWINGS campaign between 2 March 2021 at 00:00 UTC and 6 March 2021 at 00:00 UTC. Three periods can be identified. On 2 March, prior to 06:00 UTC, the air mass moved from the west into a storm zone and was characterized by ERA5 winds that exceeded 18 ms-1 (Fig. 8a). The vessel was affected by this storm, with measured swells exceeding 14 m and wind speeds reaching 25 ms-1. The air mass was also under the influence of rain, which limited the aerosol concentration due to washout, particularly prior to 03:00 UTC (∼ 500 cm-3) (Fig. 8a, d, e). The activation diameter was observed to range between 60 and 100 nm, with significant fluctuations in hygroscopicity, as indicated by variations in κ, ranging from 0.2 to 1 (Fig. 8f). During the second period, from 2 March at 18:00 UTC to 3 March at 06:00 UTC, a decline in wind and wave height conditions was recorded (Fig. 8b, g). On 2 March at 18:00 UTC, the ship moved approximately 300 km northwestward. The measured wind and wave height remained elevated, and the air mass continued to be exposed to strong winds exceeding 20 ms-1 over the past 24 h. According to the ERA5 analyses, this air mass was not affected by rain. This explains why the aerosol concentrations measured on the Marion Dufresne exceeded 3000 cm-3 despite the slight decrease in wave height and wind intensity. Aerosol hygroscopicity increased overall but remained variable, with kappa ranging from 0.2 to 1 (Fig. 8f). The third period begins at 06:00 UTC on 3 March. SMPS data show the persistence of a pronounced Aitken mode of around 30 nm and a concentration of approximately 2000 cm-3 (Fig. 8d, e). The formation of a second mode is also observed, whose size increases over the period and reaches a size characteristic of the accumulation mode of around 100 nm. During this period, the Marion Dufresne maintained a northwestward course, away from the storm center. According to the ERA5 data, on 3 March at 18:00 UTC, the air mass remained unimpacted by rain and had experienced moderate winds of approximately 10 ms-1 the previous day (Fig. 8c). Observations from the Marion Dufresne indicate that the sea remained rough until at least 06:00 UTC on 4 March, with wave heights ranging between 6 and 10 m (Fig. 8g). Although the total aerosol concentration remained largely stable during this period, it is noteworthy that as the air mass aged, an accumulation mode emerged (probably through coagulation processes), resulting in a continuous decline in κ from 0.7 on 3 March at 06:00 UTC to 0.2 on 5 March at 00:00 UTC (Fig. 8f). As a partial conclusion, the findings indicate that the aerosol concentration is determined by the equilibrium between various factors, including advection, local primary production, and loss through precipitation. It is important to note that the maximum concentration of aerosols is not necessarily attained during periods of peak wind speed and wave height.

Figure 9 shows the evolution of the aerosol concentration (NSMPS, NCCN, and size distribution) and ship tracks superimposed on a cloud mask based on satellite brightness temperature (cloud areas correspond to brightness temperatures below 282 K; e.g., ) on 11 September 2021. The satellite images in Fig. 9a–d show that the ship was in a cloudy area until 07:00 UTC (11:00 LT), then passed through a clear-sky area between 09:00 and 15:00 UTC, before returning to a cloudy area. Before 08:00 UTC (∼ 12:00 LT), SMPS measurements clearly show two modes at 30 and 130 nm, associated with an NSMPS of 500 cm-3. The measured NCCN is low: 178 cm-3 at 0.2 % SS and 221 cm-3 at 0.4 % SS. κ is between 0.1 and 0.4. A sharp increase in NSMPS is observed at 09:00 UTC, reaching 4000 cm-3 at 13:00 UTC. The SMPS size distribution depicts the formation of new particles at 09:00 UTC (5 nm), which is associated with classic banana-shaped growth (condensation–coagulation) that is characteristic of nucleation events and references cited. In addition, the formation of new particles is concomitant with the transition from cloudy to clear skies. The activation diameters at 0.2 % SS and 0.4 % SS are larger than the size of the nucleation mode. During the nucleation process, the Aitken mode demonstrates growth from 30 to 80 nm, which exceeds the activation diameter at 0.4 % SS. Thus, the NCCN at 0.4 % SS increases from 220 cm-3 (08:00 UTC) to 350 cm-3 (10:00 UTC), while the NCCN at 0.2 % SS remains relatively stable throughout the day. κ at 0.2 % of SS increases at 0.46. This increase in hygroscopicity is likely associated with the condensation of semivolatile oxidized species on the Aitken mode, which undergoes a simultaneous evolution from a median diameter of 50 to 90 nm. This nucleation situation observed during OP3 (Fig. 4a; label 4) demonstrates the possibility of a rapid increase in the number of marine aerosols in pristine areas without being associated with strong increases in CCN concentration.

Figure 10 shows the average aerosol size distributions as a function of air mass origin calculated by FLEXPART using the method described in Sect. 3.3. Table 2 presents the corresponding geometric parameters for each mode fitted to a log-normal distribution.

Regardless of the origin of the air mass, we can highlight the Aitken, accumulation, and coarse modes. The mean diameters of the Aitken, accumulation, and coarse modes are 33, 120, and 1.7 µm, and the standard deviations are 1.73, 1.52, and 1.5, respectively.

For the coarse mode, the mean aerosol number concentration is the highest for air masses originating in the southern Indian Ocean (4.5 cm-3) and the lowest for continental air masses (1.17 cm-3). These concentrations correspond to the presence of strong regular winds and swells in the southern Indian Ocean that generate significant primary emissions.

Significant differences are observed in the submicron modes between different types of air masses. The average aerosol number concentration is 830 cm-3 for the Aitken mode and 240 cm-3 for the accumulation mode. Continental air masses have concentrations of 1100 cm-3 for the Aitken mode and 490 cm-3 for the accumulation mode. Similarly, the cleaner air masses from the subtropical Indian Ocean exhibit the lowest average concentrations, with 400 and 160 cm-3 for the Aitken and accumulation modes, respectively. Significant differences are also observed for the mean diameter. Air masses from the southern Indian Ocean, which are more influenced by primary emissions, have the smallest diameters: 29 nm for the Aitken mode and 97 nm for the accumulation mode. Larger mean diameters are observed in the continental and subtropical groups of the Indian Ocean. For the Aitken mode, the mean diameter is 45 and 37 nm. For the accumulation mode, the mean diameter is 120 and 143 nm. The Aitken mode of the continental group has the largest average diameter, and there is no clear separation in terms of number concentration between this mode and the accumulation mode. This distinction is particularly noticeable when compared to air masses from the subtropical Indian Ocean group. This can be explained by the pollutant load (gases and aerosols) of the continental group, which favors growth through condensation and coagulation. In contrast, air masses from the subtropical Indian Ocean group are generally cleaner and therefore have less opportunity to grow.

Figure 11a, b, and c show the evolution of κ values at 0.2 % SS and 0.4 % SS and Δκ as a function of local wind speed, local nanophytoplankton abundance, and the three types of air masses. The difference in κ indicates the evolution of small-particle hygroscopicity between the two activation diameters (∼ 100 nm at 0.2 % SS and ∼ 80 nm at 0.4 % SS). Nanophytoplankton abundance values are higher for the southern Indian Ocean group, with a mean value of 884.8 ± 166.9 cellscm-3 compared to a mean value of 267.4 ± 29.6 cellscm-3 for the subtropical Indian Ocean group. The separation between the three types of air masses is clearly visible. The highest average κ values are observed for air masses from the subtropical Indian Ocean, the lowest for continental air masses, and intermediate values for air masses from the southern Indian Ocean. These results are consistent with previous observations of more polluted and hydrophobic air masses in the continental class, cleaner and more hydrophilic air masses in the subtropical Indian class, and intermediate air masses influenced by primary emissions (NaCl and organic matter) in the southern Indian Ocean class.

The effect of wind speed on hygroscopicity is clearly visible at both supersaturation levels for air masses from the subtropical Indian Ocean. κ values increase from 0.17 to 0.4 at 0.4 % SS (from 0.28 to 0.58 at 0.2 % SS) for wind speeds between 3 and 11 ms-1. This result is consistent with the lower abundance of nanophytoplankton in this group, which makes the κ values higher than those obtained for continental and southern Indian Ocean air masses, bringing the κ values closer to those of NaCl. Δκ also increases from 0.1 to 0.18 between 3 and 9 ms-1.

For air masses from the southern Indian Ocean, at 0.2 % SS, an increase in the average κ values from 0.33 to 0.5 is observed between 9 and 17 ms-1. At 0.4 % SS, the κ value is stable with wind speed (∼ 0.2). The effect of the primary organic emissions in reducing the hygroscopicity, which counterbalances the increase in hygroscopicity caused by the NaCl emissions, is therefore clearly noticeable. This effect is particularly pronounced at 0.4 % supersaturation.

Δκ is higher for air masses from the southern Indian Ocean than for those from the subtropical Indian Ocean and increases with increasing wind speed between 9 and 17 ms-1. Thus, a more pronounced decrease in hygroscopicity is observed for smaller aerosols, with primary emissions in regions where phytoplankton is abundant. This observation suggests the presence of a greater amount of organic matter on the surface of aerosol particles smaller than 100 nm compared to aerosols in the accumulation mode. This is consistent with the findings of , , and , who observed that organic surfactants tend to accumulate on the surface of aerosols. Thus, for an identical NaCl / organic mass ratio, the proportion of organic matter on the surface will be greater for smaller aerosols.