The cruise took place from October to November 2020, encompassing field surveys at a total of 19 stations in the EIO, as shown in Figure 1 . SA section is situated along the equator, while the SB section is located at the region perpendicular to the equator. By using conductivity-temperature-depth (CTD) sensors (Sea-Bird Electronics Inc., Bellevue, USA) equipped with 1L Niskin bottles, seawater samples in 7 depths (generally in 5, 25, 50, 75, 100, 150, and 200 m) were collected. Temperature and salinity data were obtained from the seabird CTD sensor. 40 mL seawater for DMS was gathered from CTD slowly into an amber glass via an acid-cleaned Tygon tubing. 25 mL seawater for DMSPt and DMSOt was pretreated with 100 μL 50% sulfuric acid and 100 μL 25% hydrochloric acid, respectively, as well as 25 mL seawater filtered by gravitational pressure was collected for DMSPd and DMSOd analysis, and then severally treated with sulfuric acid and hydrochloric acid (Zhai et al., 2018). All DSC samples were sealed and stored in the dark at 4°C and analyzed immediately following their transportation to the lab. The seawater samples of Chl‐a and nutrients were filtered through Whatman GF/F glass fiber membranes (25 mm diameter, 0.7 μm) and immediately stored at -20°C until they were measured. Phytoplankton samples were loaded in 1L polyethylene bottles, fixed with 3% formaldehyde solution, and placed in shade.

Samples of DMS, DMSP, and DMSO were determined using the established method based on ion mobility spectrometry (Peng et al., 2020, 2022). The concentration of DMS was measured via gas stripping, while those of DMSP and DMSO were obtained indirectly by measuring DMS content. The frozen DMSO and DMSP samples were restored to room temperature and then purged with zero air to remove any present DMS. 500 μL and 200 μL of KOH (10 mol/L) solution was added to the DMSPt and DMSPd seawater samples, respectively. Then the samples were sealed and placed in the dark at 4°C for 24 h to ensure the DMSP completely converted to detectable DMS (Yang et al., 2016). Subsequently, the concentrations of DMSPt and DMSPd were indirectly obtained by measuring the content of DMS, and the content of DMSPp could be calculated from their difference. The DMSO sample was treated with 200 μL of 20% TiCl₃ solution, then sealed and placed in a constant temperature water bath at 55°C. After the complete reaction for 1 h, the DMSO in samples could be reduced to DMS and then determined (Kiene and Gerard, 1994). The DMSOp content could be defined as the difference between DMSOt and DMSOd.

According to the method of Hansen and Koroleff (1999), the nutrients in the seawater sample were determined via a Technicon AA3 autoanalyzer (Bran + Luebbe, Norderstedt, Germany). Silicates (DSI) could be determined by using the silicon-molybdenum blue method with a limit of detection (LOD) of 0.02 μmol/L (Isshiki et al., 1991). The concentrations of nitrate (NO₃ ^-) and ammonium (NH₄ ⁺) were determined through the cadmium copper column reduction method with a LOD of 0.01 μmol/L (Wood et al., 1967) and the sodium salicylate method with a LOD of 0.03 μmol/L (Verdouw et al., 1978), respectively. The concentrations of phosphates (DIP) and nitrite (NO₂ ^-) were determined using the phosphomolybdenum blue method with a LOD of 0.02 μmol/L (Taguchi et al., 1985) and the naphthalene ethylenediamine method with a LOD of 0.01 μmol/L (Wang et al., 2022a), respectively.

The filter membrane of Chl-a was placed into a 10 mL brown glass tube and then extracted with 5 mL acetone with a volume fraction of 90% in the dark at 4°C for 24 h. The content of Chl-a was measured by Turner-Designs Trilogy fluorometer (Sunnyvale, CA, USA) (Welschmeyer, 1994). For phytoplankton analysis, 1L samples were concentrated in a 100 mL sedimentation column for 24 to 48 hours. The identification and counting of the phytoplankton cells were conducted by inverted microscope at 400× (or 200×). The methods of Yamaji (1984), Jin et al. (1965), and Sun et al. (2002) were used to determine the species of phytoplankton (Wang et al., 2022b). The abundances of picophytoplankton and bacteria were enumerated via a flow cytometer (FCM, Becton-Dickinson Accuri C6) equipped with a laser emitting at 488 nm (Sgorbati, 2007; Wei et al., 2019).

DMS (P, O) horizontal and vertical distributions were depicted using Ocean data view 4. The aggregated boosted tree (ABT) analysis was applied to quantify the impact of environmental and biological factors on the DMS (P, O) concentrations by the “gbmplus” package with 500 trees for boosting in R. Generalized additive models (GAMs) was adopted to fit relationship between response and explanatory variables by the R package “mgcv”. The structural equation model (SEM) was employed to reveal the causal relationship of DMS (P, O) and key factors. Only when Supplementary Figure S1 was plotted, missing data were imputed using linear interpolation according to the characteristics of the data.

In the surface layer, concentrations of DMS, DMSPt, and DMSOt varied from 0.07 to 7.37 nmol/L, 0.14 to 9.17 nmol/L, and 0.15 to 3.32 nmol/L, respectively ( Figures 2A–C ). DMS and DMSPt showed similar distribution patterns, gradually increasing from 14°N to 6°N and then decreasing from 6°N to 14°S ( Figures 2A, B ). DMSOt, however, was primarily concentrated between 10°N to 5°S ( Figure 2C ). The highest concentration of DMS (6.56 nmol/L) and DMSPt (7.20 nmol/L) was observed at station EQ-07, while that of and for DMSOt at EQ-04 (3.06 nmol/L) and E87-18 (3.32 nmol/L) near the equator. This pattern was linked to the Wyrtki jets (WJs) within 2° of the Equator (Wang, 2017), which transported high DMS (P, O) surface seawater from west to east, resulting in higher concentrations from 85°E to 88°E (Wei et al., 2019). Perpendicular to the equator, the highest DMS and DMSPt values were 7.37 nmol/L and 9.71 nmol/L at station E87-28, respectively, correlating with high Chl-a concentrations near Sri Lanka ( Figure 2D ). These findings align with previous studies indicating a positive correlation between DMS, DMSP concentrations, and Chl-a in coastal regions (Zhang et al., 2014; Li et al., 2015; Zhai et al., 2020).

The sea-to-air exchange flux of DMS was estimated using the stagnant film model and relevant empirical equations (Saltzman et al., 1993; Nightingale et al., 2000). The fluxes ranged from 3.44 to 329.03 μmol m^-2 d^-1, with an average of 79.76 μmol m^-2 d^-1. The highest fluxes, 329.03 μmol m^-2 d^-1 at station E87-28 and 265.99 μmol m^-2 d^-1 at station E80-10 showed a general weakening trend from west to east and north to south, consistent with the surface DMS distribution ( Figures 2A , 5A ). DMS, temperature, and wind speed based on data collected between 1987 and 2001 of the eastern Indian Ocean (data retrieved from the DMS Database: https://saga.pmel.noaa.gov/dms/) were analyzed to determine if there was a transient change of DMS concentration and sea-to-air flux ( Figures 5B, C ). DMS concentration with no significant correlation to the sampling site ( Figures 5D, E ). While the sampling site’s impact was less pronounced for fluxes ( Figures 5G, H ), there was a peak in sea-to-air flux around 15°N -20°N ( Figure 5G ). Both DMS concentration and fluxes varied seasonally but with different trends, probably due to wind speed ( Figures 5F, I ). DMS concentration increased monotonically from June to November ( Figure 5F ). However, the fluxes increased from June to August and decreased from September to November ( Figure 5I ), which aligns with the previous study indicating DMS emissions elevated during the Central Indian Ocean summer (Hulswar et al., 2022). The monthly mean fluxes in these six years (1987, 1998, 1999, 2000, 2001, 2020) were shown in Supplementary Table S1 .

Spearman’s correlation analysis elucidated the relationship between DMS (P, O) and environmental/biological factors throughout the water column. Temperature showed a significant positive correlation with DMS (P), while NO₃ ^-, DIP, and DSI were negatively correlated. Chl-a, Pro, and Syn all correlated positively with DMS (P). However, no factors were significantly correlated with DMSOt ( Figure 6A ). These results aligned with those for different groups in Supplementary Figure S1 . An aggregated boosted tree (ABT) analysis quantified the relative impacts of different parameters on the DMS (P, O), indicating temperature and DSI as the most significant for DMS, and NO₃ ^- for DMSPt and DMSOt. Pro significantly affected the concentrations of DMS, DMSPt, and DMSOt, with latitude also having a notable effect on DMSOt (23.33%). In addition, Syn significantly affected DMS contents. Bacteria and Chl-a were important biological factors affecting DMSPt, while Chl-a and PEuks for DMSOt ( Figures 6B–D ). The top four important parameters of the ABT model would be fitted nonlinearly using generalized additive models (GAMs).

Generalized additive models (GAMs) confirmed temperature, DSI, Pro, and Syn as strong predictors of DMS (all P < 0.05). DMS concentration increased with rising temperature and decreasing DSI ( Figures 7A, B ). An increased abundance of Syn and Pro was beneficial for higher DMS concentrations ( Figures 7C, D ). The interaction model indicated that DMS concentration increased significantly after 23°C and decreased with rising DSI ( Figure 8A ). The interplay between Pro and Syn suggested that maximum DMS contents occurred at their highest abundance, with Syn contributing more to DMS concentration than Pro ( Figure 8B ).

For DMSPt, Pro, bacteria, Chl-a, and NO₃ ^- were strong predictors (all P < 0.05). DMSPt concentration increased with a rising abundance of Pro, bacteria, and Chl-a, but decreased with increasing NO₃ ^- ( Figures 7E–H ). Surprisingly, in the interaction of Chl-a and NO₃ ^-, DMSPt concentration decreased with increasing Chl-a ( Figure 8C ). The interaction between Pro and bacteria showed a unique peak in DMSPt concentration at around 50 bacteria ( Figure 8D ). The DMSPt concentration first increased and then decreased with rising bacteria abundance, while it continuously increased with Pro abundance. The highest DMSPt was observed when bacteria were around 50 and Pro abundance was at its maximum. If both X and Y axes were zero, axis Z still had intercepted, indicating the presence of other vital microorganisms affecting DMSPt in the EIO ( Figures 8C, D ). Based on the ABT results, we proceeded to explore the effect of Pro and Syn interaction on DMSPt. Syn had a greater impact on DMSPt than Pro ( Figure 8E ), and DMSPt decreased with increasing Chl-a in the interaction of Chl-a and other biological factors ( Figures 8F–H ).

Latitude, NO₃ ^-, and Pro were significant indicators for DMSOt in GAMs ( Figures 7I–K , all P < 0.05), while the relationship between DMSOt and Chl-a could not be fitted (R² < 0, data not shown). DMSOt concentration enhanced as the location approached the northern latitude ( Figure 7I ) and had a broad peak around 5 μmol/L NO₃ ^- ( Figure 7J ). The correlation between Pro and DMSOt was monotonically negative ( Figure 7K ). The interaction between NO₃ ^- and latitude was complex, with a single peak in DMSOt concentration at 5°N and 10 μmol/L NO₃ ^- ( Figure 8I ). Lower Pro abundance near the equator favored increased DMSOt ( Figure 8J ).

In surface seawater, the DMS concentrations observed in this study were within the range reported for open sea areas ( Table 1 ), yet the mean value was marginally higher. The mean DMSPt concentration resembled that in the Western Pacific Ocean during the survey period 2009.10.9 to 10.24 but was lower compared to other regions. The mean DMSOt value aligned with those in the Arabian Sea, but was lower than in the Western Mediterranean and Western Pacific Ocean. Notably, even in identical sea areas and seasons, the range and mean value of DMS (P, O) exhibited slight variations due to differences in specific sampling locations and timings, as exemplified by studies in two different periods in the Western Pacific Ocean in Table 1 . The temporal and spatial variability of the DMS (P, O) might correlate with distinct microbial community structures in various oceanic regions (Shenoy and Kumar, 2007).

Vertically, DMS and DMSPt concentrations were primarily found in water columns shallower than 75 m, whereas higher DMSOt concentrations occurred at depths deeper than 75 m. Influenced by the west-to-east Wyrtki jets, elevated DMS (P, O) values were primarily observed near the equator between 87°E and 90°E in the SA section. In the SB section, DMS (P, O) distribution was influenced by the BBR and SEC. The greater precipitation than evaporation in the Bay of Bengal resulted in a low salinity layer north of 8°N ( Supplementary Figure S2H ) (Rao and Jayaraman, 1968; Sengupta et al., 2006), which stimulated phytoplankton production ( Figure 4H ). Concurrently, the SEC transported high-nutrient and high-Chl-a seawater from east to west south of 5°S ( Figure 4H ; Supplementary Figures S2I–K ), leading to lower DMS (P, O) values (Gao et al., 2021). However, due to missing data from some sampling sites south of the equator, biomass and DMS (P, O) concentrations were likely underestimated ( Figures 4E–H ).

Temperature emerged as the most critical environmental factor affecting DMS concentration, in line with previous studies (Boyd et al., 2013; Guo et al., 2022). The metabolic processes of phytoplankton, including those involving enzymes essential for DMS production and metabolism, are temperature-dependent (Gao et al., 2017). Notably, the impact of temperature on DMS was modulated by the interaction between temperature and DSI ( Figures 7A , 8A ). DSI indirectly regulated DMS concentration by influencing the biomass of Pro and Syn. The significant negative correlation between DSI and these organisms suggested that high DSI concentrations inhibited their growth ( Figure 6A ), corroborating findings from Wang et al. (2022c). DMSPt content increased with intensifying nitrogen limitation, as illustrated in Figure 7H . Previous research highlighted the role of the transamination reaction in the DMSPt synthesis pathway, allocating nitrogen to new amino acids (Dacey et al., 1987; Hanson et al., 1994; Colmer et al., 1996; Gage et al., 1997; Curran et al., 1998). Therefore, abundant DMSPt could conserve nitrogen in cells under nitrogen-limited conditions (Stefels, 2000), particularly in the oligotrophic waters of the EIO. Meanwhile, NO₃ ^- was identified as the most influential environmental factor for DMSOt through ABT analysis ( Figure 6D ) and SEM ( Figure 9A ). The optimal NO₃ ^- concentration for DMSOt production was around 5 μmol/L, as indicated in Figure 7J , with an interaction effect of NO₃ ^- and latitude observed at 10 μmol/L ( Figure 8I ). These results suggest an optimum NO₃ ^- concentration for DMSOt production; relevant Laboratory evidence is needed to validate the result.

The SEM not only elucidated the causal relationships among different microorganisms and DMS (P, O), but also clarified the interconversion process among DMS, DMSPt, and DMSOt ( Figure 9A ). The total effects of phytoplankton (-0.141) and bacteria (-0.196) on DMS were negative, whereas those of picophytoplankton (0.429), DMSPt (0.32), and DMSOt (0.19) were positive. These findings suggest that DMS is mainly produced by picophytoplankton, DMSPt cleavage, and DMSOt reduction, yet is predominantly consumed by phytoplankton and bacteria in the EIO. The direct effect of picophytoplankton (0.55) on DMSPt was positive, while the indirect effect was negative (-0.033), consistent with observations that most picophytoplankton take up DMSPt for intracellular storage, and certain species cleave DMSP. The overall effects of bacteria and phytoplankton on DMSPt were positive (0.05) and negative (-0.11), respectively, indicating that bacteria contribute to the DMSPt source, whereas phytoplankton primarily depletes it. The overall effect of picophytoplankton on DMSOt was negative (-0.206), exceeding that of bacteria (-0.15). In addition, the indirect positive effect of phytoplankton on DMSOt (0.0088) suggests that phytoplankton are biological producers of DMSOt. Moreover, the total positive effect (0.1288) of DMSPt on DMSOt demonstrates that DMSPt oxidation is a significant source of DMSOt in the EIO. (The calculation procedure of direct effect, indirect effect, and total effect was shown in Supplementary Table S2 ).

The SEM results are consistent with GAM’s conclusions and provide insights into the possible sources and sinks of DMS (P, O) in the EIO ( Figure 9B ). DMS primarily originates from picophytoplankton production, followed by DMSPt cleavage, and DMSOt reduction. Phytoplankton and bacteria act as DMS sinks. For DMSPt, aside from bacteria, macroalgae, angiosperms, and some corals are also important sources (Shaw et al., 2022). Picophytoplankton serves as a significant DMSP sink through assimilative storage, while the relatively low proportion of phytoplankton ( Supplementary Figure S3 ) consumes DMSP as a carbon and sulfur source (Simo, 2001). The primary sources of DMSOt are dominated by DMSPt oxidation, with picophytoplankton and bacteria as the main consumers.

The sea-to-air fluxes of DMS in various sea areas are compiled in Supplementary Table S3 in Supplementary Material . The DMS fluxes of 79.76 μmol m^-2 d^-1 in this study exceeded those in other seas, attributed to higher wind speeds ranging from 2.4 to 14.5 m/s, averaging 8.58 m/s in our survey areas. Although our wind speeds were lower than those reported by Zhai et al. (2020), the lower DMS concentration led to lower sea-to-air flux in the Northern South China Sea. The monthly mean fluxes changing in this investigation area were different from those of global monthly mean fluxes but were similar to those of the Northern Hemisphere reported by Wang et al. (2020), which increased from February to August and decreased from September to November, with a peak in August ( Supplementary Table S1 ). This is mainly because of the higher fluxes in this study area located in the Northern Hemisphere ( Figure 5G ). The survey regions cover 6.29 × 10⁶ km², accounting for 8.91% of the total EIO and 1.65% of the global ocean area, respectively. Missing data, including fluxes of April, May, and December were imputed using spline interpolation, and the monthly mean fluxes over the 12 months were summed to obtain the annual fluxes (1.846 Tg S yr^-1) ( Supplementary Table S1 ). If this annual flux is used to evaluate the contribution of DMS emissions in this study area to global DMS emissions, it accounted for 9.17% of the global DMS annual fluxes based on reported by Wang et al. (2020) (20.12 Tg S yr^-1) and occupied 6.81% based on reported by Hulswar et al. (2022) (27.1 Tg S yr^-1). DMS is a known major source of cloud condensation nuclei (CCN), and increased atmospheric DMS can enhance CCN formation, potentially amplifying the albedo effect and mitigating the greenhouse effect (Charlson et al., 1987). The emission flux in the Indian Ocean thus has an important impact on global climate change. Although the survey area and months are limited and cannot represent the monthly flux variation of the whole eastern Indian Ocean, we hope that the results of this study will provide data support for further assessment of monthly climatology of DMS fluxes for the entire Indian Ocean.