Origin of water-soluble organic aerosols at the Maïdo high-altitude observatory, Réunion Island in the tropical Indian Ocean

The tropical and subtropical Indian Ocean (IO) is expected to be a significant source of water-soluble organic aerosols (WSOAs), which are important factors relevant to cloud condensation nuclei and ice nuclei of aerosol particles. Current atmospheric numerical models significantly underestimate the budget of organic aerosols and their precursors, especially over tropical oceans. This is primarily due to poor knowledge of sources and the paucity of observations of these parameters considering spatial and temporal variation over the tropical open ocean. To evaluate the contribution of sources to 20 WSOA as well as their formation processes, submicrometer aerosol sampling was conducted at the high-altitude Maïdo observatory (21.1° S, 55.4° E, 2,160 m a.s.l), located on the remote island of La Réunion in the southwest IO. The aerosol samples were continuously collected during local daytime and nighttime, which corresponded to the ambient conditions of the marine boundary layer (MBL) and free troposphere (FT), respectively, from March 15 to May 24, 2018. Chemical analysis showed that organic matter was the dominant component of submicrometer water-soluble aerosol (~45 ± 17%) during the wet 25 season (March 15–April 23), whereas sulfate dominated (~77 ± 17%) during the dry season (April 24–May 24). Measurements of the stable carbon isotope ratio of water-soluble organic carbon (WSOC) suggested that marine sources contributed significantly to the observed WSOC mass in both the MBL and the FT in the wet season, whereas a mixture of marine and terrestrial sources contributed to WSOC in the dry season. The distinct seasonal changes in the dominant source of WSOC were also supported by Lagrangian trajectory analysis. Positive matrix factorization analysis suggested that marine secondary 30 OA dominantly contributed to the observed WSOC mass (~70%) during the wet season, whereas mixtures of marine and terrestrial sources contributed during the dry season in both MBL and FT. Overall, this study demonstrates that the effect of marine secondary sources is likely important up to the FT in the wet season, which may be responsible for cloud formation as well as direct radiative forcing over oceanic regions.


Introduction
The ocean surface is a major source of submicrometer aerosols, which play an important role in the atmospheric radiative budget because they determine the number of cloud condensation nuclei (CCN) and ice nuclei (IN). Marine-derived submicrometer organic aerosols (OAs) can affect the marine aerosol optical depth (AOD) as well as CCN and IN concentrations. These are particularly important over remote oceans, as these areas experience minimal influence from 5 anthropogenic emissions originating from terrestrial sources. In general, organic matter (OM) is concentrated in the sea surface microlayers relative to bulk seawater. OM is further concentrated in aerosols during the bubble bursting process, which produces primary submicrometer sea spray aerosols (SSAs) that are enriched in OM (O'Dowd and De Leeuw, 2007). Moreover, sea-to-air emissions of volatile organic compounds (VOCs) produced by marine microbial activity have the potential to form secondary OAs. Nevertheless, there is still a large uncertainty in the potential sources and formation processes of OA in the 10 marine atmosphere, leading to uncertainty in determining their climate impact.
The tropical Indian Ocean (IO) is an oceanic region with high primary productivity (Jayaraman et al., 1998;Langley DeWitt et al., 2013;Höpner et al., 2016) (Figure 1), where significant emissions of VOCs, including oxygenated VOCs (OVOCs), and OAs are expected. A number of previous studies have focused on aerosols over the northern IO, particularly around India (Chylek et al., 2006;Madhavan et al., 2008;Srinivas and Sarin, 2013). These studies have addressed the impact of 15 anthropogenic and land influences from Asia on the marine background. Conversely, the southwest IO is one of the few pristine regions in the global ocean. It is generally not affected by anthropogenic emissions originating from continental sources.
Moreover, the western IO has been recognized as a region in which phytoplankton blooms occur frequently (Kyewalyanga, 2016;Roxy et al., 2016). Consequently, it is suitable to investigate remote marine aerosol composition and its relationship to oceanic emissions (Mallet et al., 2018). The source apportionment of organic aerosols has not yet been investigated, particularly 20 for both marine and high-altitude sites that cover both the marine boundary layer (MBL) and lower free tropospheric (FT) conditions over tropical oceans in the Southern Hemisphere. This paper presents a 2-month study of chemical composition and stable carbon isotope ratios in marine aerosols obtained at a high-altitude observatory in La Réunion in the southwest IO over two seasons. The purpose of this study was to evaluate the contribution of marine/terrestrial sources to water-soluble organic aerosols and their formation processes in MBL and FT over 25 the tropical Indian Ocean.

The Maïdo high-altitude observatory
The high-altitude Maïdo observatory (21.1°S, 55.4°E, 2,160 m a.s.l) is located on the remote island of La Réunion in the southwest IO (Baray et al., 2013). The observatory is affected by prevailing southeasterly trade winds in the MBL. The 30 meteorological field in that region is characterized by wet (typically from November to April) and dry seasons (from May to October). Cyclones can occur typically between November and May (Baray et al., 2013). Previous studies reported that the observatory is located in the MBL during daytime, and in the FT during nighttime (Baray et al., 2013;Guilpart et al., 2017).Thus, we aimed to obtain aerosol samples in daytime and nighttime conditions by using two identical aerosol samplers, as described in the following subsection. 35

Aerosol sampling
Submicrometer aerosol samples were collected at the Maïdo observatory during the period of March 15-May 24, 2018, in the framework of the OCTAVE (Oxygenated Compounds in the Tropical Atmosphere: Variability and Exchanges) project (e.g., Verreyken et al., 2020). The aerosol samplings were conducted continuously using two high-volume air samplers (HVAS; https://doi.org /10.5194/acp-2021-277 Preprint.  Cleves, OH, USA) attached to each. The samples were collected onto quartz-fiber filters at a flow rate of 1130 L min -1 with one sampler during daytime (0700-1800 in local time; LT) and the other sampler during nighttime (2200-0500 LT) by auto power supply.
In this study, we used analytical results obtained from the bottom stage of the impactor, which collected particles with 5 aerodynamic diameter (Dp) lower than 0.95 μm. Here, ambient aerosol particles collected at the bottom are referred to as submicrometer aerosol particles. The sample filters were typically exchanged every two to three days. The average volumes of the sampled air were 2098 m 3 and 1454 m 3 during daytime and nighttime, respectively.

Measurements of chemical parameters of water-soluble aerosols
The term water-soluble aerosols are defined as particles sampled on the filter and extracted with ultrapure water followed by 10 filtration through a syringe filter (Miyazaki et al., 2018). To determine the WSOC concentration of the submicrometer filter samples, a filter cut of 39.25 cm 2 was extracted with 15 mL ultrapure water using an ultrasonic bath for 15 min. The extracts were filtered through a 0.22 µm pore syringe filter and then injected into a total organic carbon (TOC) analyzer (Model TOC-LCHP, Shimadzu) (Miyazaki et al., 2018(Miyazaki et al., , 2020. To measure the stable carbon isotope ratio of WSOC (δ 13 CWSOC), another filter cut (27.24 cm 2 ) for each sample was acidified 15 to pH 2 with hydrochloric acid (HCl) to remove inorganic carbon prior to extraction (Miyazaki et al., 2018(Miyazaki et al., , 2020. The decarbonated filter samples were then dried under a nitrogen stream for approximately 2 h. WSOC was extracted from the filters in 20 mL of ultrapure water using the method described above to measure the WSOC concentration. The extracted samples were concentrated by rotary evaporation, and 40 μL of each sample was transferred to be absorbed onto 10 mg of precombusted Chromosorb in a pre-cleaned tin cup. The δ 13 CWSOC was then measured using a Flash EA 1112/continuous flow 20 carrier gas system (ConFlo)-isotope ratio mass spectrometer (Delta V, Thermo Finnigan).

Measurements of molecular tracer compounds
Another portion of the filter (58.9 cm 2 ) was extracted with dichloromethane/methanol to measure biogenic molecular tracers.

Meteorological parameters and FLEXPART backward trajectory 35
Water vapor mixing ratio values at the sampling site were derived from the automatic measurements of ambient temperature and relative humidity monitored by meteorological sensors (Vaisala, Helsinki, Finland) at the Maïdo observatory. To investigate air mass histories from the sampling site, ten-day backward trajectories were computed using the Lagrangian FLEXible PARTicle dispersion model, FLEXPART (Stohl et al., 1998;Pisso et al., 2019). These FLEXPART simulations https://doi.org/10.5194/acp-2021-277 Preprint. Discussion started: 1 June 2021 c Author(s) 2021. CC BY 4.0 License.
were driven with hourly ECMWF operational data at half a degree horizontal resolution and 137 vertical levels. The calculation was initialized at 00, 06, 12, and 18 UTC every day during the sampling period.

Positive Matrix Factorization (PMF)
Positive Matrix Factorization (PMF) (Paatero and Tapper, 1994) was used to identify and characterize possible sources of the observed WSOC during the study period. In this study, EPA PMF 5.0 (Norris et al., 2014)  period. In this study, the wet season was defined as the first half of the sampling period, whereas the latter half was defined as 15 the dry season reflecting the significant difference in the water vapor mixing ratios between wet (8.7 ± 2.6 g kg −1 ) and dry seasons (6.4 ± 1.4 g kg −1 ). Moreover, water vapor mixing ratio values during daytime (9.3 ± 2.7 g kg −1 and 8.1 ± 2.5 g kg −1 for wet and dry seasons, respectively) were significantly higher than those observed during nighttime (7.4 ± 0.9 g kg −1 and 5.4 Moreover, they reported distinct diurnal variations in the water vapor mixing ratio with averages of 9.7 ± 2.4 g kg −1 during the day (1100-1700 LT) and 6.4 ± 2.9 g kg −1 at night (2300-0500 LT) during the one-year period, showing that the Maïdo observatory indeed located both in the MBL and FT during the day and at night, respectively. The observed levels of the water vapor mixing ratio and their seasonal changes in this study are in good agreement to those reported by Guilpart et al. (2017).
Therefore, the results presented in this study confirm that the observatory was located in the MBL during daytime, whereas it 25 was in the FT at night.  the southeast of the island were frequently transported to the observatory, where this transport pathway was explained by the strong trade wind which is commonly observed in the dry season. In contrast, no significant corresponding increase in the WSOC concentrations during this period (Figure 4a) was observed, suggesting that the contribution of the volcanic eruption to the WSOC mass was insignificant.

Seasonal variations of mass fractions and concentrations of submicrometer water-soluble aerosol
Bromide is used here as a tracer of marine primary emissions, whereas MSA is formed by the oxidation of dimethyl sulfide 10 (DMS) emitted from the marine surface and is used as a tracer of marine secondary production. Substantially larger concentrations of bromide and MSA were also observed in the MBL (Figure 4c and 4d). The average bromide concentration in the wet season was twice as large as that in the dry season (Table 1), reflecting higher biological productivity in the wet season. The temporal variation of the concentration of MSA is similar to that of WSOC, suggesting that the dominant source of WSOC is similar to MSA. The concentrations of 2-methyltetrol, an oxidation product of isoprene, were typically larger in 15 the MBL than in the FT in the wet season (Figure 4e), although their temporal variations are generally different from those of WSOC with exceptions of a few samples. The contributions of terretrial sources are further discussed in the following sections.

Isotopic characterization of WSOC and FLEXPART backward trajectories
The isotopic composition of aerosol carbon has been used successfully to determine the contributions of marine and terrestrial sources to aerosol carbon mass found in the remote marine atmosphere (e.g., Cachier et al., 1986). In particular, the WSOC-20 specific stable carbon isotope (δ 13 CWSOC) provides robust tools for the source apportionment of aerosol WSOC in the marine atmosphere (e.g., Miyazaki et al., 2016). Figure 5 shows the temporal variations of δ 13 CWSOC in the submicrometer aerosols during the entire period. Based on previous studies (e.g., Cachier et al., 1986;Turekian et al., 2003), here we assume that δ 13 CWSOC > −24‰ indicates WSOC was mostly originated from sea water, whereas δ 13 CWSOC < −24‰ indicates WSOC mainly affected by terrestrial sources. In the wet season, 87% of the data in the MBL and 83% of the data in the FT, respectively, 25 showed the δ 13 CWSOC larger than −24‰, with averages of −23.2 ± 1.0‰ (MBL) and −23.5 ± 2.5‰ (FT). In particular, the larger concentrations of WSOC (e.g., > 300 ngC m −3 in MBL; Figure 4a) corresponded to a higher δ 13 CWSOC (> −24‰). The results suggest that marine sources contributed significantly to the WSOC mass under both the MBL and FT conditions during the wet season. In contrast, the average δ 13 CWSOC in the dry season were −24.4 ± 2.5‰ and −25.0 ± 1.4‰ in the MBL and FT, respectively, where 33% (MBL) and 33% (FT) of the data showed δ 13 CWSOC > −24‰. 30 To estimate the relative contributions of marine and terrestrial OC sources to the observed WSOC, a mass balance equation (e.g., Turekian et al., 2003) was applied assuming a δ 13 C value of −21.5‰ for marine OC (Turekian et al., 2003;Miyazaki et al., 2010), and −28‰ for terrestrial OC (e.g., Cachier et al., 1986). Our calculations indicate that marine sources contributed ~74% and ~69% in MBL and FT, respectively, during the wet season. The estimated contributions of marine sources are reduced to ~55% and ~46% in the MBL and in the FT, respectively, during the dry season, suggesting that the WSOC mass 35 was attributed to both marine and terrestrial sources with similar fractional contributions in the MBL and FT during the dry season. season (e.g., Mallet et al., 2018). The Lagrangian trajectory analysis showed that the majority of air parcels in the wet season was transported over the southern Indian Ocean. This is consistent with the results of the isotopic analysis and suggested that the majority of submicrometer WSOC originated from the sea surface during the wet season. During the dry season, some portion of the trajectories passed over Southern Africa in addition to the southern Indian Ocean, indicating some influence from the land surface in addition to the marine source. This is also consistent with the results from the isotopic analysis of 5 WSOC, which suggest the influence of both land and ocean surface.

Source apportionment of WSOC by PMF
To further investigate the possible sources of the submicrometer WSOC under different conditions, a PMF analysis was performed as described in 2.6. Figure 7 illustrates each factor profile calculated by the PMF. The PMF resolved six interpretable factors, which were characterized by the enrichment of each tracer compound. Factor 1 (F1) was characterized 10 by the large contribution of MSA (⁓50%). Consequently, it is referred to here as "marine SOA." In fact, previous cruise measurements showed that in the southern IO, the sea-to-air emission of DMS is more active than that of other oceanic regions (Sciare et al., 1999) and that DMS was the most abundant VOC measured in the atmosphere (Colomb et al., 2009). The oceanic regions mentioned in these previous studies overlap with the possible oceanic source region shown in Figure 6a and 6b.
Factor 2 (F2) is characterized by sea salt components, such as sodium (66%) and magnesium (78%). Moreover, bromide also 15 contributed significantly to F2, which is thus referred to here as "marine primary aerosol." Because Factor 3 (F3) is dominated by sulfate (68%), it is defined as "the sulfate-dominated" source. Factor 4 (F4) is characterized by the dominant contributions of 2-methyltetrol and n-nonacosan- 10-ol. Miyazaki et al. (2019) identified n-nonacosan-10-ol in forest aerosols, suggesting that they originated mostly from plant waxes and could be a tracer of primary biological aerosol particles. Consequently, F4 is referred to here as "terrestrial biogenic sources." Although Factor 5 (F5) was difficult to attribute to a specific source, given 20 the possibility that nitrate is associated with terrestrial sources with smaller contributions of marine tracers, F5 was labeled here as "terrestrial sources." Similarly, Factor 6 (F6) is dominated by ammonium (50%) with a mixture of tracers of marine and terrestrial sources. F6 is referred to here as "mixture of marine and terrestrial sources" as a possible source category of WSOC. in the MBL and FT. The average contributions of each PMF-derived factor to the WSOC mass are also summarized in Figure   9. A distinct temporal shift of the dominant source of WSOC was apparent from the wet season to the dry season in both the MBL (Figure 8a) and FT (Figure 8b). On average, marine SOA dominantly contributed to the WSOC mass (~66% -70%) in both the MBL and FT during the wet season (Figure 9a and 9b). On average, terrestrial biogenic sources, which are mainly based on the contribution of 2-methyltetrol, accounted for 16% of the WSOC mass in the MBL during the wet season. 30 Specifically, the contribution of terrestrial biogenic sources was more than 40% of the WSOC mass around April 1 and April 19 in the MBL. Previous cruise measurements of VOCs suggested oceanic emissions of isoprene in the southern IO during austral summer (December) (e.g., Colomb et al., 2009). However, the data exhibiting large contributions of terrestrial biogenic sources mentioned above showed a lower δ 13 CWSOC < −24‰ (Figure 5b), supporting the validity of the definition of the PMF factor as terrestrial biogenic sources rather than marine biogenic origin. These biogenic sources are attributable to local 35 terrestrial biogenic emissions of VOCs on La Réunion Island, followed by the upward transport along the slope of the island particularly in daytime (Verreyken et al., 2020). It is noted that F1 had also large contributions of oxidation products of αpinene (i.e., pinic acid, pinonic acid, and 3-MBTCA; Figure 7), which is also attributable to local terrestrial biogenic emissions of VOCs during the transport from the ocean to the observatory. However, the dominance of marine SOA as a source of WSOC in the wet season (Figure 8) is consistent with the δ 13 CWSOC measurements, supporting the validity of the definition of F1 and 40 that the contribution of α-pinene SOA from local biogenic sources to the WSOC mass was small in this case. Mixtures of marine and terrestrial sources significantly contributed to the WSOC mass during the dry season, where they accounted for 61% and 47% of the WSOC mass in the MBL and FT, respectively (Figure 9c and 9d). These results point to the importance of marine SOA up the FT during the wet season, which is attributed to the high oceanic productivity in this region (Zhou et al., 2018), as well as to significant vertical transport of air during this season.

Secondary formation of marine-derived WSOC and implications 5
It is possible that the aging of marine POA and subsequent formation of more oxidized OA significantly contributed to the observed WSOC mass. Mallet et al. (2018) presented an 8-year satellite dataset of the distribution and variability of marine aerosols over the Southern IO, which included the current aerosol sampling site. They suggested that aerosols are mainly confined below 2 km above sea level and are dominated by sea salt over the Southern IO. However, the mass fraction of sea salt in the submicrometer particles observed at the Maïdo observatory (2,160 m a.s.l) was insignificant, which resulted in a 10 significantly lower contribution from marine PA (Figures 8 and 9). Therefore, the current results indicate that the aging of marine POA is insignificant, and the contribution of oxidation of VOCs from the sea surface to the WSOC mass up to the lower FT is likely more important.
Model results by Brüggemann et al. (2018) indicated that, especially in tropical regions with low POA concentrations, additional SOA from oxidation of photochemically produced VOCs contributes up to 60% of additional OA mass, such as 15 over the IO. In summary, the results of the current study highlight the importance of marine biogenic SOA up to the lower FT, a process missing in climate models. Current models typically consider only marine POA (i.e., SSA) from the sea surface to represent the OA burden in tropical "pristine" oceanic regions. The impacts of marine SOA up to FT aerosols lead to changes in the microphysical and optical properties of aerosol particles. Model calculations (Zhu et al., 2017) suggested that the contribution of SOA to radiative forcing will increase substantially in the future even if the increase of SOA burden is slight 20 and without considering the combined effects of changes in marine SOA. The current results may have important implications for understanding the climate effects of aerosols in these oceanic regions.

Conclusions
In this study, the origins of WSOC in submicrometer aerosols were investigated based on continuous ambient aerosol sampling 25 at the Maïdo observatory in La Réunion in the southwest Indian Ocean. OM was the dominant component of the submicrometer water-soluble aerosol (~46 ± 10%) in the MBL during the wet season, whereas sulfate dominated (~77 ± 19%) during the dry season. Our estimate using the stable carbon isotope ratio of WSOC showed that, on average, for the wet season, marine sources accounted for ~74% and ~69% of the WSOC mass in MBL and FT, respectively. Conversely, marine sources contributed ~55% and ~46% in MBL and FT, respectively, of the WSOC mass during the dry season, suggesting that the 30 WSOC mass was attributed to both marine and terrestrial sources in the MBL and FT during that season. The significant seasonal difference in the dominant source of WSOC between the two seasons was also supported by Lagrangian trajectory analysis.
The PMF analysis suggested that marine secondary OA was a dominant contributor to the observed WSOC mass (~70%) during the wet season, whereas mixtures of marine and terrestrial sources accounted for 61% and 47% of the WSOC mass in 35 the MBL and FT, respectively. Overall, this study demonstrates that emissions of biogenic VOCs from the ocean surface followed by the formation of secondary OA are likely important up into the FT during the wet season, when marine biological activity and vertical transport are more significant. These characteristics are responsible for subsequent cloud formation as well as the direct radiative forcing over this oceanic region. Data from measurements are available upon request from the authors.

Author contributions.
The overall measurements and analysis of the aerosol samples were performed by SAS, ET, and YM. SAS and YM wrote the manuscript. JB, RV, and TS coordinated the project and the aerosol samplings at the research sites. HF, JB, AC, OM, and YM 5 made aerosol samplings. FLEXPART simulations were performed by BV, SE, and JB.

Competing interests.
The authors declare that they have no conflict of interest.