Dimethyl sulfide cycling in the sea surface microlayer in the southwestern Pacific – Part 1: Enrichment potential determined using a novel sampler

. Elevated dimethyl sulfide (DMS) concentrations in the sea surface microlayer (SML) have been previously related to DMS air-sea flux anomalies in the southwestern Pacific. To further address this, DMS, its precursor dimethylsulfoniopropionate (DMSP), and ancillary variables were sampled in the SML and also subsurface water at 0.5 m depth (SSW) in different water masses east of New Zealand. Despite high phytoplankton biomass at some stations, chlorophyll a enrichment in the SML was low (enrichment factor (EF) < 1.06), and DMSP was enriched at one station with DMSP EF 15 ranging from 0.81 to 1.25. DMS in the SML was determined using a novel gas-permeable tube technique which measured consistently higher concentrations than with the traditional glass plate technique; however, DMS enrichment was also present at only one station, with the EF ranging from 0.40 to 1.22. SML DMSP and DMS were influenced by phytoplankton community composition, with correlations with dinoflagellate and Gymnodinium biomass , respectively. DMSP and DMS concentrations were also correlated between the SML and SSW, with the difference in ratio attributable to greater DMS loss 20 to the atmosphere from the SML. DMS in the SML did not significantly influence regional DMS emissions, with the calculated air-sea DMS flux of 2.28 to 11.0 µmol m -2 d -1 consistent with climatological estimates for the region. These results extend previous regional observations that DMS is associated with dinoflagellate abundance but indicate that additional factors are required for significant enrichment in the SML.


Introduction 25
Dimethyl sulfide (DMS), a trace gas mainly derived from dimethylsulfoniopropionate (DMSP) primarily produced by phytoplankton (Keller et al., 1989), is a natural aerosol precursor (Yu and Luo, 2010;Sanchez et al., 2018), and a potential regulator of climate. About 4 to 16% of DMS is ventilated to the atmosphere (Galí and Simó, 2015) and oxidized to non-sea salt sulfate aerosols and methane sulfonic acid, which subsequently contribute to formation and growth of cloud condensation nuclei (CCN). Condensation of water vapor on CCN leads to the formation of cloud droplets, with the resulting increase in 30 cloud reflectivity potentially reducing incoming solar radiation to the ocean and consequently decreasing phytoplankton growth and DMS emissions, as postulated by the CLAW hypothesis (Charlson et al., 1987). Although the CLAW hypothesis has been questioned, due to spatial and temporal decoupling of CCN and DMS emissions, and the identification of other CCN precursors (Quinn and Bates, 2011), it continues to be investigated to elucidate potential feedbacks of DMS emissions on climate. 35 DMS concentrations in the surface ocean fluctuate in response to variation in regional biology and physical controls (Stefels et al., 2007). DMSP concentration is influenced by phytoplankton community composition (Keller et al., 1989), bacterial processes (Curson et al., 2017), grazing (Wolfe et al., 1994), and physicochemical variables such as nutrient availability, light, salinity and temperature via DMSP and DMS cycling (Stefels et al., 2007). These factors may have a direct effect on DMSP 40 production and consumption, and also an indirect effect via their influence on plankton community composition (Stefels et al., 2007;Stefels, 2000). Variability in DMSP and DMS in the surface ocean is reflected in regional variation in DMS flux to the atmosphere. Generally, the air-sea flux is estimated from DMS concentration in surface waters (2 to 10 m), but there is evidence that processes within the sea surface microlayer (SML) may also affect the DMS flux (Walker et al., 2016). The SML is vertically less than 1,000 µm and connects the ocean to the atmosphere (Hunter, 1980). Biogeochemical cycling within the 45 SML may differ to that of the subsurface water (SSW) due to the concentration of biogenic material and exposure to high irradiance, both of which influence dissolved trace gas concentrations and flux to the atmosphere (Upstill-Goddard et al., 2003;Carpenter and Nightingale, 2015), and production of primary and secondary aerosols (Leck and Bigg, 2005;Roslan et al., 2010). DMS enrichment in the SML relative to the SSW has been reported, with an enrichment factor (EF) range of 0.6 to 5.7 (Yang et al., 2005a;Zhang et al., 2009;Walker et al., 2016;Yang, 1999). DMS enrichment is often associated with blooms of 50 certain phytoplankton groups, such as dinoflagellates and haptophytes (Yang, 1999;Matrai et al., 2008;Yang et al., 2009;Walker et al., 2016), whereas enrichment is often absent where diatoms dominate (Zhang et al., 2008;Matrai et al., 2008), except when present in high abundance (Yang et al., 2005a;Zhang et al., 2009). High DMS enrichment in the SML has also been reported in association with specific physical and meteorological conditions and may result in anomalously high air-sea DMS flux and discrepancies between observed and calculated DMS air-sea fluxes (Marandino et al., 2008;Walker et al., 55 2016).
A global DMS climatology model based on all reported observations (82,996 datapoints; (Wang et al., 2020)), shows a seasonal pattern, particularly in mid to high latitude regions (Kettle et al., 1999). The climatological average DMS concentration in the southwestern Pacific does not exceed 4 nmol L -1 , except during January and February when DMS concentration ranged 60 between 6 and 10 nmol L -1 . East of New Zealand, the Subtropical (STW) and Subantarctic (SAW) water masses meet at the Subtropical front (STF) along the Chatham Rise, where high phytoplankton production is often observed (Murphy et al., 2001;Chiswell et al., 2015). The STW north of the Chatham Rise is characterized by warm saline water and low phytoplankton productivity due to low nitrogen availability, whereas the SAW south of the Chatham Rise is fresher with high macronutrient concentrations but low productivity due to iron limitation (Boyd et al., 1999). Consequently, this region provides an ideal area 65 to determine the influence of variability in water mass properties on DMS and aerosol precursor production (Law et al., 2017).
During the SOAP (Surface Ocean Aerosol Production) voyage in the Chatham Rise region in 2012, DMSP and DMS distribution varied with phytoplankton composition and biomass, with elevated DMS concentrations relative to regional climatological estimates (Bell et al., 2015;Walker et al., 2016;Wang et al., 2020). DMS concentrations exceeded 20 nmol L -1 , resulting in an elevated DMS flux during a dinoflagellate bloom (Bell et al., 2015;Walker et al., 2016;Lizotte et al., 2017;70 Lawson et al., 2020), with two independent approaches (direct SML concentration measurement and indirect estimation from eddy covariance) indicating that DMS enrichment in the SML influenced air-sea flux (Walker et al., 2016).
The SOAP results also raised questions regarding how DMS enrichment is maintained in the SML, and the influence of the SML on DMS emissions. Sampling of the SML is challenging and existing techniques are not optimal for trace gas sampling. 75 The Garret screen (Garrett, 1965) has generally been preferred to the plate (Harvey and Burzell, 1972) for DMS sampling of the SML (Yang et al., 2001), although this may result in artefacts (Yang et al., 2005b;Walker et al., 2016), and underestimation of DMS concentration (Yang and Tsunogai, 2005;Zhang et al., 2008;Zemmelink et al., 2006;Matrai et al., 2008). However, Walker et al. (2016) used the plate and the Garret screen and found that the screen was overestimating DMS due to preconcentration of organic material in the mesh. To address this, a novel SML sampling technique using gas-permeable tube 80 to minimize DMS loss was deployed, and results compared to those obtained with the glass plate method during the Sea2Cloud voyage. The primary aim of this voyage was to examine the relationships between marine biota and aerosol formation, and so DMSP, DMS and ancillary variables were measured in the SML and SSW to estimate EFs, and establish the factors influencing DMS cycling and emission (see companion paper, Saint-Macary et al. (2022)). Estimation of DMS fluxes enabled reconciliation of regional estimates based upon empirical data (Bell et al., 2015;Walker et al., 2016) and climatology models 85 (Lana et al., 2011;Wang et al., 2020).

Regional setting
The Sea2Cloud voyage took place from the 16 to 28 March 2020 (austral autumn) onboard R/V Tangaroa in the Chatham Rise region (Figure 1a). The characteristics of the water masses sampled during this voyage and meteorological conditions are 90 summarized in Table 1. Six workboat deployments were carried out to sample the SML and SSW in different water mass types: STF at stations 1 and 2, SAW at stations 3 and 4, STW at station 5 (see Figure 1a, Table 1). Mixed water (Mixed) at station 6 was a composite of coastal and shelf water from Cook Strait and STW, with higher nutrient content than STW, as presented in Figure 1b. Local wind measurements were obtained from an Automatic Weather Station (AWS) located at 25.2 m above sea level above the bridge of the R/V Tangaroa, which was exposed to all wind directions (Smith et al., 2018). 95

Sampling of the SML
The SML and SSW were sampled from a workboat 0.5 to 1 nautical mile away from the R/V Tangaroa between 08:00 and 105 12:00 NZDT during periods when the wind speed was below 10 m s -1 (Table 1). Station 5-STW was sampled in the afternoon due to high wind speed in the morning (> 10 m s -1 ). DMS was sampled using a novel gas-permeable tube technique in which a 280 cm long loop of silicone tube (external diameter 2.41 mm, wall thickness 0.49 mm) was deployed on the sea surface.
The gas-permeable tube was filled with Milli-Q® water prior to deployment and closed by joining the two tube ends with a union. The gas-permeable tube was threaded through floating beads to ensure contact with the SML and deployed free-floating 110 upstream of the workboat. Once in contact with the SML, the technique relies upon diffusion of DMS through the gaspermeable tube membrane across the concentration gradient between seawater and Milli-Q® water. In theory at least 50% of the tube surface area is in contact with the SML and surface seawater, with the remainder exposed to the atmosphere. The gas-permeable tube was recovered after 10 minutes, with the Milli-Q® water withdrawn immediately using a syringe and stored in a chilly bin. SML sampling was carried out in duplicate at each station. 115 Prior to deployment in the field, the diffusion efficiency of the gas-permeable tube was determined in semi-controlled conditions using coastal seawater in Wellington, New Zealand, naturally elevated in DMS (range: 1.25 -16.88 nmol L -1 , average 4.94 nmol L -1 ). The calibration tank was continuously filled with seawater at a flowrate of 75 L min -1 , with a constant overflow to ensure that there was no SML formation; this approach resulted in a uniform and homogenous DMS concentration 120 in the tank for the gas-permeable tube floating to equilibrate with. The gas-permeable tube was filled with Milli-Q® water and placed on the surface of the seawater in the tank for 10 minutes, after which the Milli-Q® water was withdrawn into a syringe with no headspace whilst the gas-permeable tube remained in contact with the surface water. The 10-minute exposure time was pre-determined in laboratory experiments and represented the optimum time to achieve significant diffusion efficiency whilst reducing deployment time. The gas-permeable tube was then removed from the water and refilled with Milli-Q® water 125 and the experiment was repeated from 3 to 8 times. In addition, ambient seawater in the calibration tank was sampled at t0 and t+10min for each repetition. Between each repetition, samples were transferred to the laboratory for immediate analysis. The DMS diffusion efficiency was subsequently determined using Eq. (1): (1)

where [DMS]MQ is the DMS concentration measured in the Milli-Q® water at t+10min, and [DMS]tank is the averaged DMS 130
concentration between t0 and t+10min, measured in the calibration tank. The average D for 10 minutes exposure was 61% (± 10% standard deviation, n = 19) as determined over a 4-month period during which the seawater temperature range was similar to that during the Sea2Cloud voyage, at 12 -16 ℃. Further details of the gas-permeable tube technique are provided in Saint-Macary (2022). The average D was then applied to calculate the actual DMS concentration in the SML, [DMS]SML, using Eq.
(2): 135 where [DMS]MQ is the DMS concentration in the Milli-Q® water after 10 minutes of exposure in the SML.
A glass plate (Harvey and Burzell, 1972) and a sipper were also used for sampling of DMSP, DMS and ancillary variables in the SML. The sipper consists of a tube with multiple inlets that float on the sea surface. A syringe was used to slowly draw 140 SML water through the open inlets to sample for chlorophyll a (Chl a), phytoplankton composition and DMSP. The sipper external diameter was similar to the gas-permeable tube (2.2 and 2.4 mm, respectively), so enabling sampling of a similar SML thickness but larger SML water volume in a shorter period. DMSP and DMS were also sampled with the plate for method comparison only. The repeatability for DMS sampling with the gas-permeable tube and plate were calculated using the standard deviation (Eq. 3, (Olivieri and Faber, 2009)), based upon two duplicates at each sampling event: 145 with the standard deviation, N the total number of terms, terms given in the data, and the mean.

Sampling of the SSW
For DMSP and DMS sampling of the SSW, a Teflon tube was deployed with the inlet at a depth of approximately 0.5 m by a system of ropes and fishing weights. Fifty mL of SSW were withdrawn using a syringe and collected in an amber bottle leaving 150 no headspace. For larger volumes for other ancillary variables in the SSW, a bottle was immersed to 0.5 m below the surface and filled with seawater. To avoid SML contamination, the bottle was immersed with its lid on, then opened and closed in the SSW before recovery. For each variable the EF was calculated by dividing the concentration in the SML by its concentration in the SSW.

155
The CTD was launched between 10:00 and 12:15 following SML sampling, except at 5-STW when the CTD was deployed before the SML sampling at 07:00. Six depths from 5 to 150 m were sampled with 12 L Niskin bottles, although only the results from 5-m depth are discussed in this paper. For DMS sampling from the CTD casts, the water was overflowed by gravity by at least 100% into amber bottles and then sealed with no headspace.

DMSP and DMS analytical system 160
For DMS measurements, water from the amber bottles was withdrawn in plastic Terumo® syringes. The samples were injected through a 25-mm glass microfiber filter (GF/F) into a 1-mL loop, before transfer to a silanized sparging tower, where the sample was sparged for 5 minutes with nitrogen (N2) at a flow rate of 50 mL min -1 . Nafion® dryers removed the water vapor from the gas samples before DMS preconcentration at −110 ℃ on a Tenax® trap. The trap was then heated to 120 ℃ to release the DMS onto an Agilent Technology 6850 Gas Chromatography coupled to an Agilent 355 Sulfur Chemiluminescent 165 Detector (GC-SCD). The daily sensitivity and detection limit of the detector were confirmed using VICI® methyl ethyl sulfide and DMS permeation tubes. The average detection limit during the voyage was 0.14 (± 0.03) pgS sec -1 . For total DMSP measurements, 20 mL glass vials were filled with seawater and 2 pellets of sodium hydroxide added before gas-tight sealing the vials, which were stored at ambient temperature in the dark. DMSP was analysed one day after sampling using the same semi-automated purge and trap system followed by GC-SCD, as described above. A wet standard calibration curve was made 170 daily from a stock solution of DMSP diluted in Milli-Q® water, with calibration concentrations ranging from 0.1 to 95 nmol L -1 . These were decanted into 20 mL gas tight glass vials, hydrolysed with 2 pellets of sodium hydroxide and then injected into the sparging unit and processed as with the samples.

Ancillary variables
For Chl a analysis, 250 mL of seawater was filtered onto a 25-mm GF/F filter, and then stored at −80 ⁰C until analysis. Chl a 175 was extracted in 90% acetone, measured and compared with Chl a standards by spectrofluorometry using a Varian Cary Eclipse fluorometer, with an accuracy of 0.5 nm at 541.2 nm. An acidification step was used to correct for pheophytin interference (10200 PLANKTON).
Phytoplankton community structure was determined for cells >5 µm using a Flowcam (Fluid Imaging Technologies Inc). A 180 sample of 250 mL of seawater was filter concentrated using a 47-mm diameter 3-µm polycarbonate filter to 10 mL final volume and stored at 4 ℃ until analysis. One mL of 25 times concentrated seawater sample was run through a 80-µm depth Field of View flow cell (FC80FV) at 0.050 mL min -1 and 20 frames per second, with an imaging efficiency of 61.9 ± 2%. Images were taken using a 10× objective on AutoImage mode. Total run time for each sample was 20 min. Between 4-SAW and 5-STW, the sample volume and flow rate were increased to 2 mL at 0.100 mL min -1 , with an imaging efficiency of 32.7%, due to the 185 high abundance of large diatoms (e.g. Chaetoceros sp.). Images were classified into cell size and class groupings using VisualSpreadsheet v4.16.7 software, by size category (<10 µm; 10 to 20 µm; 20 to 50 µm and >50 µm), and the results given as total phytoplankton biovolume of each size class.
For microscopic analysis of phytoplankton community composition, 500 mL of seawater was preserved at 1% (final 190 concentration) Lugol's iodine solution, with samples stored at room temperature in the dark. Phytoplankton community composition and cell numbers for phytoplankton >5 µm were determined using optical microscopy, following the method described in Safi et al. (2007) and references herein. Briefly, 100 mL subsamples were settled for 24 hours and the supernatant then carefully syphoned with 10 mL transferred to Utermohl chambers and resettled (Edler and Elbrächter, 2010). Where possible, all abundant organisms were identified to genus or species level before being counted. Phytoplankton biovolume 195 estimates were calculated from the dimensions of each taxa and approximated geometric shapes (spheres, cones, ellipsoids) initially following Olenina (2006). The biovolumes were subsequently used to calculate cell carbon (mg C m -3 ) using equations from the literature; Olenina (2006) and Montagnes and Franklin (2001) for diatoms, and Olenina (2006) and Menden-Deuer and Lessard (2000) for dinoflagellates and nanoflagellates. Menden-Deuer and Lessard (2000) was also applied to other low biomass unidentified groups referred to as small flagellates. 200

DMS air-sea flux calculation
The DMS air-sea flux, F, was calculated using the gas transfer flux equation (Liss and Merlivat, 1986), following Eq. (4): with H the Henry's law solubility coefficient for DMS (Dacey et al., 1984), [ ] dissolved DMS concentration, [ ] DMS concentration in the atmosphere, and , the gas transfer coefficient. The latter was calculated using the NOAA 205 COARE gas transfer (COAREG) version 3.6 algorithm (Fairall et al., 2003;Fairall et al., 2011) and parameterized in terms of local wind speed scaled to 10 m height, as described in Bell et al. (2015). The gas transfer velocity was adapted for DMS using the Schmidt number (Sc) calculated using local temperature (T) in ℃ (Saltzman et al., 1993)

Statistical analysis 215
The Shapiro test was used to verify the normality of variable distribution. For the non-normally distributed variables Spearman's rank correlation was carried out and for the normally distributed data a Pearson test was applied. Linear correlation was considered significant where the coefficient of correlation (rho and r for Spearman's rank and Pearson tests, respectively) was higher than 0.5 and p value was lower than 0.05.

Comparison of plate and gas-permeable tube
The repeatability of SML sampling techniques is generally not reported, although this is critical, particularly as the width and presence of the SML is inherently patchy and heterogenous (Frew et al., 2002;Ribas-Ribas et al., 2017). The repeatability of the plate and gas-permeable tube were similar although the plate had a smaller interquartile range (plate median 6%; interquartile range 6% n = 6; gas-permeable tube median 10%; interquartile range 17% n = 6; Figure 2). The repeatability 225 determined for DMS using the gas-permeable tube was subsequently applied in the current study to identify a significance threshold, with no significant difference between SML and SSW DMS when EF was within 0.90 -1.10. with the sipper may reflect that this method samples some water from immediately below the SML, whereas the plate only withdraws the organic layer associated with the SML (Harvey and Burzell, 1972;Cunliffe and Wurl, 2014).

Correlations between variables
For all station data the Pearson test identified that DMSP concentration, and diatom biomass in the SML were significantly correlated to their respective concentrations in the SSW (r = 0.95; p < 0.01 for DMSP, and r = 0.92; p = 0.03 for diatoms). The 305 SML DMS concentration presented in this section was obtained from the gas-permeable tube and was not normally distributed.
In addition, DMSP and DMS were correlated in both the SML and SSW (Spearman's rank test in the SML, and Pearson test in the SSW; Table 2 and Table 3). The SML DMSP concentration was also correlated with SML dinoflagellate biomass (Pearson test, Table 2). The Spearman's rank test established that Chl a and DMS in the SML correlated to their respective concentrations in the SSW (rho = 0.99; p < 0.01 for DMS, and rho = 0.94; p = 0.02 for Chl a), and DMS concentration in the 310 SML also correlated with SML Chl a concentration, the 20-50 µm fraction (Spearman's rank test; Table 2) and the biomass of the dinoflagellate Gymnodinium (rho = 0.95; p = 0.05; Spearman's rank test; Suppl. Info.). In the SSW, 20-50 µm and >50 µm size fractions correlated with the Chl a concentration (Table 3). The correlations from this current study were all positive.

Air-sea flux 325
Average wind speeds over the previous 12 h ranged from 3.79 to 8.19 m s -1 for the workboat sampling. The air-sea flux was calculated over the 12 h prior to sampling the SML as the SML structure and near-surface mixing would be influenced by winds over a longer preceding period than instantaneous winds. Average DMS fluxes were 3.68 µmol m -2 d -1 (range: 2.45 -6.96 µmol m -2 d -1 ) for of FSML, and 5.32 µmol m -2 d -1 (range: 2.49 -11.56 µmol m -2 d -1 ), with generally higher DMS fluxes recorded at higher wind speeds combined with higher DMS concentrations as expected (Table 4). Air-sea flux was also 330 calculated using DMS concentration at 5 m depth (F5 m) and compared with the SML and SSW fluxes to examine the influence of depth on calculated flux. Although FSML and F5 m exhibited differences across workboat stations, average F5 m 3.87 µmol m -

Discussion
From a regional perspective, the Sea2Cloud results contrast with previous studies (Law et al., 2017;Walker et al., 2016), with lower DMS concentrations encountered in SSW, and SML DMS enrichment at only one of the six stations. Furthermore, Chl 340 a enrichments in the SML were low, contrary to that reported in other studies (Yang et al., 2009;Zhang et al., 2008;Zhang et al., 2009). Enrichment of biogeochemical variables, such as Chl a, DMSP and DMS, in the SML has often been observed during a phytoplankton bloom in the underlying water (Nguyen et al., 1978;Yang et al., 2005a;Zhang et al., 2009;Walker et al., 2016); however, it appears that the major diatom bloom of 4.3 µg L -1 Chl a at 2-STF, which exceeded the maximum Chl a concentrations recorded during the previous SOAP voyage (2.8 µg L -1 ; (Lizotte et al., 2017)), was insufficient to generate 345 Chl a, DMS or DMSP enrichments in the SML. These contrasting regional results (Bell et al., 2015;Walker et al., 2016;Lizotte et al., 2017) suggests non-optimal conditions for DMS and Chl a enrichment in the SML in the area of investigation in this time of the year. SML DMSP concentration was primarily influenced by dinoflagellate biomass, as indicated by the positive correlation between 350 these variables (Table 2). This is consistent with previous observations, in which DMSP enrichment in the SML was attributed to phytoplankton composition (Yang and Tsunogai, 2005;Zemmelink et al., 2006), particularly when dinoflagellates were dominant (Yang, 1999;Matrai et al., 2008;Yang et al., 2009). However, DMSP was not enriched in the SML during the SOAP voyage, despite the high dinoflagellate biomass (C. Law, pers. comm), and SML enrichment only occurred at one station in the current study where the ratio of dinoflagellate to diatoms was the lowest (5-STW, 0.2, Figure 3b). The correlation between 355 DMSP and dinoflagellates was high in both SML and SSW in the current study, but only significant in the SML, indicating that specific factors enhance this relationship in the SML. DMSP production increases under oxidative stress (Sunda et al., 2002), and so light stress may be a co-factor that enhances DMSP production by dinoflagellates in the SML.
The complexity of DMS cycling often precludes identification of the main drivers of DMS production, and this is particularly 360 so in the SML where loss of DMS to the atmosphere obscures potential relationships with conservative properties such as Chl a and phytoplankton group (Stefels et al., 2007;Bürgermeister et al., 1990;Townsend and Keller, 1996;Turner et al., 1988).
Indeed, only one study has previously reported a correlation between enrichment of Chl a and DMS in the SML (Yang and Tsunogai, 2005). However, DMS concentration in the SML was correlated to both the Chl a and 20-50 µm size fraction in the current study (Table 2). During SOAP, high DMS EF and concentrations were associated with a dinoflagellate bloom (Walker 365 et al., 2016) with Gymnodinium and Gyrodinium being the most abundant genera, in addition to Ceratium and small flagellates (C. Law, pers. comm., Suppl. Info. Figure S3). In both SOAP and the current study, SML DMS was significantly correlated with Gymnodinium (Spearman's rank test; rho = 0.95; p = 0.05 and rho = 0.76; p = 0.02, respectively). The relationship between DMS and dinoflagellate is consistent with dinoflagellate being a source of DMSP, but also DMSP conversion to DMS may potentially be enhanced by other factors. For example, copepod grazing on Gymnodinium is reported to influence DMS 370 concentration (Dacey and Wakeham, 1986). Moreover, during senescence, dinoflagellates release gel-like compounds that accumulate in the SML (Jenkinson et al., 2018), altering the physical properties of the SML and influencing gas exchange (Wurl et al., 2016). Consequently, dinoflagellates affect DMSP and DMS both directly and indirectly in the SML.
DMS loss is expected to be more rapid in the SML due to its proximity to the atmosphere. However, other processes such as 375 elevated photo-oxidation of DMS in the SML may also be part of DMS removal processes in the surface ocean (Saint-Macary et al., 2022). The DMS maximum in the SSW, relative to the SML and 5 m depth (Figure 4a) may reflect a combination of near-surface stratification and elevated DMS ventilation at the surface. This is in contrast to the observations of Walker et al.
(2016) in the same region who reported the opposite effect, with high DMS enrichment in the SML. The latter may have arisen from an optimal combination of factors: (i) a dinoflagellate bloom supporting elevated DMSP and resulting DMS production 380 (Walker et al., 2016), (ii) favourable meteorological conditions i.e. very low wind speeds (Law et al., 2017), that limited nearsurface mixing and led to (iii) near-surface stratification (Smith et al., 2018). Although near-surface temperature measurements were not obtained during the current study, wind speeds were generally higher than during SOAP, indicating higher mixing and reduced potential for near-surface stratification, although sole observation of DMS enrichment occurred at the station with the highest wind speeds (see Table 4). Contrasting near-surface DMS gradients have been reported in a stratified salt pond 385 (Zemmelink et al., 2006) and coastal water under calm meteorological conditions (Zemmelink et al., 2005), with respective increases and decreases in DMS concentration to the surface. The key factor determining DMS enrichment or depletion in the SML in these studies was irradiance, which stimulated DMSP production via the phytoplankton antioxidant response in the salt pond (Zemmelink et al., 2006), and DMS photo-oxidation in the stratified coastal water (Zemmelink et al., 2005).
Consequently, consideration of the physical controls in addition to biogeochemical processes is required to explain DMS 390 enrichment in the SML (assessed in a companion paper; (Saint-Macary et al., 2022)). An additional factor influencing enrichment may be the presence of surfactant, which can act as a barrier to gas transfer (Broecker et al., 1978;Goldman et al., 1988;Pereira et al., 2016). Surfactant, measured in mg L -1 TX-100 equivalents (Sigma Aldrich, TritonX 100), was enriched at half of the stations (3-SAW, 4-SAW and 6-Mixed; T. Barthelmess, pers. comm.), one of which showed DMS enrichment in the SML, although there was no correlation between surfactant and DMS, in terms of concentration or enrichment. 395 The current study also highlighted variation in sampling efficiency of different methodological approaches for determining DMS enrichment in the SML. The higher DMS concentrations obtained with the gas-permeable tube relative to the glass plate may reflect that the water sample in the gas-permeable tube is less exposed to air during the sampling procedure than with techniques such as the plate, screen and rotating drum (Yang, 1999;Zhang et al., 2009;Matrai et al., 2008;Zemmelink et al., 400 2006). Loss to the atmosphere is generally not accounted for in other SML studies (Zemmelink et al., 2005;Yang et al., 2001).
Although DMS is potentially lost with the gas-permeable tube, as the upper surface is exposed to the atmosphere; however, this is minimised by smearing of the SML over the tube by surface turbulence, and gas loss is accounted for by the diffusion efficiency correction (see Method). When sampled with the plate and the screen the EF DMS was shown to be affected by environmental conditions and sampling thickness (Yang et al., 2001). As the plate samples a thinner layer than the gas-405 permeable tube (nominally 20-150 µm (Cunliffe et al., 2013) and 1.21 mm, respectively), this may also result in a lower DMS concentration, depending on the SSW concentration. However, the plate samples the organics and bacteria of the SML, which may induce in vitro reactions in the sample bottle prior to analysis that may affect DMS concentration, whereas these are excluded with the gas-permeable tube. Another advantage of the gas-permeable tube is that it eliminates exposure of the water sample to high light, as with the plate and screen, so avoiding stress-induced responses and cell lysis. Patchiness of the SML 410 (Frew et al., 2002;Ribas-Ribas et al., 2017) is an issue that will decrease the reproducibility of all SML sampling techniques, but the larger surface area of the gas-permeable tube may decrease this variability. Yet, despite the increased effectiveness of the permeable tube technique for dissolved gases, the results indicate that DMS is not significantly enriched in the SML, in contrast to other studies that have used the plate and screen (Nguyen et al., 1978;Yang, 1999;Yang and Tsunogai, 2005;Yang et al., 2005a;Yang et al., 2001;Zhang et al., 2009;Walker et al., 2016;Zemmelink et al., 2006). Excluding the methodological 415 shortcomings detailed here, this anomaly may reflect differing environmental conditions between studies; however, environmental conditions are rarely reported, and only a few have considered DMS fate in the SML (Zemmelink et al., 2006;Zemmelink et al., 2005;Matrai et al., 2008;Walker et al., 2016). Consequently, it is difficult to draw conclusions as to whether previously reported DMS enrichments are artefacts, which limits the identification of the factors responsible for DMS enrichment. 420 DMS air-sea flux was calculated using the COARE algorithm, which was originally developed and tested based upon a representative depth of 5 m for surface waters (Huebert et al., 2004); consequently, this approach may be less appropriate for application to the SML, where conditions are not as homogenous as water at 5 m (Frew et al., 2002;Ribas-Ribas et al., 2017).
Regardless, the calculated fluxes based upon three different depths were consistent, and also low relative to previous regional 425 measurements during the SOAP campaign, in which DMS flux reached 100 µmol m -2 d -1 (Bell et al., 2015;Walker et al., 2016). The large difference in flux between SOAP and the regional climatological estimate Lana et al. (2011) may reflect the high DMS concentration in the dinoflagellate bloom during SOAP; the lower DMS concentrations and emission during the current study reflect differing phytoplankton community composition and surface ocean dynamics, but also potentially different process rates (Saint-Macary et al., 2022). 430

Summary and conclusion
The current study presents the first application of a more robust sampling technique for trace gases in the SML, and identified higher DMS concentrations relative to the standard SML sampling technique of the plate (Figure 4b). However, DMSP and DMS were generally not enriched in the SML, with significant enrichment of both species observed at only one of six stations, and low Chl a enrichment despite sampling of different water masses, phytoplankton biomass and community composition. 435 However, relationships were apparent between DMSP, DMS, dinoflagellate biomass and the genus Gymnodinium biomass, suggesting that SML DMS and DMSP production may be enhanced in the presence of dinoflagellates. These observations complement the results from a previous study in the same region indicating that an optimal combination of physical and biological conditions are required for DMS enrichment in the SML. The calculated DMS air-sea fluxes were consistent with regional estimates in the Lana et al. (2011) and Wang et al. (2020) climatology models and, indicate that DMSP and DMS 440 cycling in the SML do not significantly influence regional air-sea DMS flux. These results raise questions about the significance of DMS enrichment in the SML and also how this can be maintained at the ocean interface where loss to the air dominates, and so emphasises the need for DMS process studies in the SML (Saint-Macary et al., 2022).

Acknowledgment.
We would like to thank Antonia Cristi and Wayne Dillon for their help during the Sea2Cloud campaign, and Karen Thompson for Flow Cytometry analysis. This research was supported by NIWA SSIF funding to the Ocean-Climate 445 Interactions Programme. We would also like to thank the support and expertise of the Officers and Crew of the R/V Tangaroa.