Diel cycle of sea spray aerosol concentration over vast areas of the tropical Pacific Ocean and the Caribbean Sea


 Ocean-atmosphere interactions such as sea spray aerosol (SSA) formation have a major role in the climate system, but a global-scale assessment of this micro-scale process is highly challenging. We measured high-resolution temporal patterns of SSA number concentration over the Atlantic Ocean, Caribbean Sea, and the Pacific Ocean covering 42,000 km of open ocean waters. We discovered a ubiquitous 24-hour rhythm to the number concentration, clearly seen for particle diameters > ~ 0.58 µm, with spikes at dawn and drops at dusk throughout the Pacific Ocean and Caribbean Sea, showing more than doubling of the SSA number concentration during the day than at night. No correlation with surface winds, atmospheric radiation, pollution nor oceanic physical properties were found. Instead, parallel diel patterns in particle sizes detected in near-surface waters, attributed to variations in the size of particles smaller than ~ 1 µm, point to microbial day-to-night modulation of bubble-bursting dynamics as the cause of the SSA cycle.

majority of the route. We discuss the origin of the aerosols and their approximate chemical signature; (ii) we explore how atmospheric and oceanic environmental factors affect N SSA_0.58µm and show they are not responsible for the diel cycle; (iii) we show that the identi ed N SSA_0.58µm diel cycles were accompanied by parallel diel patterns of the size of seawater particles; and (iv) we hypothesize that microbial processes at the ocean surface that in uence bubble-bursting dynamics are the most probable cause of the observed diel cycle in aerosol number concentration.

Diel pattern in SSA number concentration
We explored temporal patterns of the number concentration of SSA (N SSA ) and found a distinct 24-hour pattern in N SSA , clearly seen for particles with optical diameters above ~0.58 µm (N SSA_0.58µm ; Fig. 1 and see Fig. S1 for the diameter determination). We also found day-to-night ratios of N SSA_0.58µm > 1 on the vast majority of the route ( Fig. 1 and Fig. 2), with an average ratio of 2.3 (±0.6) over the tropical Paci c Ocean.
The diel N SSA_0.58µm changes follow a de nite pattern: from midnight to dawn the N SSA_0.58μm is stable, it begins rising ~06:00 MST (mean solar time; all data was converted from UTC to MST, see Methods), within 30 min after sunrise and begins returning to pre-dawn values at ~17:00 MST, with 18 (±357) L -1 at nighttime; Fig. S2D). Around Japan the cycle was identi ed only for a few days (Fig. S2C). However, high aerosol concentrations are likely to obscure the diel cycle in those regions. The Atlantic Ocean has generally higher background aerosol concentration due to high mineral dust load (2,11,15) and the regions near Japan and New Zealand were highly polluted (11). And, we also found an inverse relationship between the day-to-night ratio of the N SSA_0.58µm and the background aerosol concentration (taken as the nighttime concentration; see Fig. S4).
We determined the measured SSA to be of marine origin based on back trajectory analysis and scanning electron microscopy. The calculated back trajectories (using HYSPLIT 16,17 ) proved that the vast majority of the air masses spent at least 48 hours over the ocean (Fig. 1). In addition, for the period where clear N SSA_0.58µm diel patterns are seen in the western Paci c (Fig. 1A), we obtained an approximate chemical signature of aerosols collected on lters with average D > 0.58 μm. We used SEM-EDX analysis and a similar particle classi cation scheme as described in Laskin et al. 18 (Fig. 1B; a total of 15.5 days were analyzed between 3 -17 May, 2017, but for clarity only 11 days are shown. See Table S1 for the other days and Methods for the classi cation scheme). We found sea salt particles comprised between 50 -100% of the total particles by number. Up to 1000 km (May 9) away from Keelung, Taiwan, we found a noticeable depletion of chloride (Fig. S5), together with a lower sea salt fraction (50 -82%) and an increase in the presence of other metals (e.g., Al, Si, K, Ca, S) and sulfate. The chlorine depletion and the increase in sulfates suggests that anthropogenic pollutants (e.g., H 2 SO 4 , SO 2 ) were present in the AMBL, as SSA are known to react with them (15). After this period, the sea salt fraction comprised 84-100% of the total particles by number. Sulfates with no sodium and "Other" species were < 3% for the whole period. The N SSA_0.58µm diel cycles were also revealed in the lter counts; we counted a total of 4560 particles in 14 daytime lters and 3706 particles in 15 nighttime lters (see Table S1). We have not, however, observed signi cant day-to-night aerosol-class differences.

Atmospheric and oceanic environmental factors
The size distribution and number concentration of the bubbles and SSA created by a breaking wave are controlled mainly by the wind speed 1 . We found the N SSA _ 0.58µm increases with wind speed for both the Paci c and Atlantic Oceans ( Fig. 3A and 3B), consistent with previous studies 1,2 . However, the pristine Paci c Ocean daytime N SSA_0.58µm consistently exceeded the night ones, compared at similar wind speeds (Fig. 3A). Further, wind speed anomalies, for the days the diel cycles were detected, showed no clear diurnal pattern (see Fig. S6). This suggests that the wind is a key player in the process of marine aerosol emission but it doesn't cause the observed diel cycles. Other atmospheric and oceanic environmental factors are known to affect the production and number of SSA, but we found no evidence that they drive the observed N SSA_0.58µm diel cycle. Links between the RH, air temperature, and atmospheric stability to the N SSA_0.58µm diel cycle were explored (see details in the Methods and Fig. S6,   Fig. S7, and Fig. S8), and were ruled out as the causes of the N SSA_0.58µm diel pattern. Rain was shown to suppress the N SSA_0.58µm cycle (Fig. 1A). No link between the intensity of photosynthetically active radiation and the daytime N SSA_0.58µm was found (Fig. S9). We also discarded secondary organic aerosol production as the cause of the diel cycle since the produced aerosol by this mechanism is much smaller in diameter (< 0.1 µm). Nor could we nd any signi cant link between physical changes within the AMBL and the N SSA_0.58µm cycle. While there is a diurnal signal in the AMBL height, it has less than 30% variations 19 and it cannot account for the variations we see in N SSA_0.58µm .
Within the ocean surface, changes in salinity and sea surface temperature (SST) can also affect SSA production 20 . However, salinity in the tropics has typical diurnal anomalies of only 0.005 psu 21 , and we did not nd strong diurnal salinity changes (Fig. S10). As for SST, even though diurnal variations are known 22 , the rate of temperature change is much weaker and can be observed later in the day ( 22 and Fig. S10) compared to the observed increase in N SSA_0.58µm . Additionally, we observed the N SSA_0.58µm cycle in conditions of wind speeds above 10 m s -1 (Fig. 1A and Fig. S8), where SST diurnal changes are expected to be less than 0.25 K 22 , and in morning overcast conditions (Fig. S8B) when no signi cant changes in SST are expected. Therefore, we can conclude that neither salinity nor SST cause the N SSA_0.58µm cycles.
These ndings suggest that neither atmospheric nor oceanic environmental factors can explain the N SSA_0.58µm diel cycle, and while the wind speed affects the N SSA_0.58µm , it does not drive the observed diel cycles. Furthermore, the presence of the N SSA_0.58μm diel cycle in the presence of anthropogenic pollutants and in clean conditions (Fig. 1A,B and Fig. S5), implies anthropogenic and continental sources are also not the cause. This suggests that the diel cycle may be triggered by changes in the ocean surface itself.
Diel patterns in the size of seawater particles The identi ed N SSA_0.58µm diel cycles in the lower atmosphere were accompanied by distinct diel cycles in the measurements of near-surface ocean light attenuation wavelength dependence of c p (expressed as particle size index γ; Fig. 1C). The index is found via a power-law t to the wavelength-dependent c p 23,24 : where c p (λ 0 ) is c p at a reference wavelength λ 0 and γ is the spectral slope of c p . Variability of γ indicates changes in the median particle size and is most sensitive to particles in the range of 0.22 -20 µm 23,25 , with smaller γ associated with larger median particle size (similar to the Ångström exponent for aerosols). Figure 1C shows that the N SSA_0.58μm diel cycle correlates well with the values of γ. Note that γ reached a maximum value (i.e., minimum mean diameter) right after sunrise when the N SSA_0.58μm began to increase, and a minimum value (i.e., maximum mean diameter) before sunset when N SSA_0.58μm were decreasing (Fig. 1C). We, therefore, calculated the rate of change of γ (∂γ/∂t (hr -1 ); see Methods) for concurrent days (92 days in total) where a N SSA_0.58μm diel cycle was also detected (Fig. 4A,B). We found a parallel behavior, with a continuous decrease in the mean particle diameter at nighttime (∂γ/∂t (hr -1 ) > 0) when N SSA_0.58μm were lowest, and a continuous increase (∂γ/∂t (hr -1 ) < 0) between 07:00 and 17:00, λ λ 0 when N SSA_0.58μm were highest. In the Paci c Ocean, where the N SSA_0.58μm diel cycles occurred, most γ values were above 0.8 (Fig. S11A) reaching 1.4 and having a strong latitudinal dependence (Fig. S11B).
In the Fiji to New Zealand leg, γ values were around 1.0, but no cycle was detected (Fig. S11C). In the Atlantic Ocean, though there were diel cycles in γ (Fig. S11A), its values were below 0.8, indicating a larger mean particle size and thereby suggesting the presence of larger planktonic species than in the Paci c Ocean.
Similar diel cycles of γ in the open ocean have been previously documented in [24][25][26] . Changes in γ can be attributed to several factors: cell growth, division, and aggregation, selective changes in particle size or concentration due to a balance between primary production and loss due to grazing and viral pressure, or changes in the refractive index of the cell population, which is related to their carbon content. In the Equatorial Paci c such changes were attributed primarily to phytoplankton growth and division 27 . During daytime cells photosynthesize, x inorganic carbon and accumulate carbohydrates and lipids which are respired during the night 28 , subsequently cells will generally divide. We found the daytime increase is associated with cell growth or aggregation of pico-phytoplanktonic populations, as the γ changes measured during N SSA_0.58μm diel cycles (in the western Paci c) can be attributed to variations in size of particles smaller than ~1 µm (see Methods and Fig. S12), and we also see a daily increase in particulate organic carbon (Fig. S10A). The decrease in mean size at night can be due to selective grazing, cell division or virus-induced lysis 29,30 .

Mechanism hypothesis
Areas with high chl-a concentrations showed the lowest day-to-night N SSA_0.58µm ratios, and the N SSA_0.58µm diel cycles were mainly observed in areas with low chl-a (Fig. 2). These oceanic oligotrophic regions are typically dominated by cyanobacteria 31,32 .
During photosynthesis eukaryotic phytoplankton and cyanobacteria can secrete extracellular polymeric substances (EPS). EPS are a diverse array of large molecules which forms a major component of the dissolved organic carbon pool in the ocean, and have been implicated in the formation of bio lms and marine snow 33 . Recently, in a different context, EPS released by bacteria were observed to increase bubble lifetime, thereby dramatically decreasing their lm thickness, and yielding more numerous and transportable droplets at burst than those produced by clean bubbles (see Fig. 1 in 34 ). These new ndings are essential as SSA formation is directly related to lm drops formed by the fragmentation of the thin uid cap lm 35 .
Consequently, we conjecture that the diel changes in the N SSA_0.58μm are controlled by microbial processes in the ocean surface. Such processes whether at the near-surface water-AMBL interface, or in the upper several meters of the ocean through bubble scavenging of the excreted EPS from bacteria, may affect bubble bursting dynamics, changing the number and size of the emitted droplets and therefore N SSA_0.58µm . The AMBL typically mixes within an hour. A characteristic time is in the order of 20 min for a 500 m AMBL (1). Hence, signi cant changes in the production of droplets can be expected to happen within a similar timeframe in which we observed the transient increase of N SSA_0.58µm . Although we do not pinpoint a direct role of marine bacteria affecting bubble-bursting dynamics in the open ocean surface, recent studies have shown a possible role for bacteria modulating ocean surface properties 33,36-39 and bubble-bursting dynamics 34,40,41 .
We presented in-situ evidence that the number concentration of SSA with optical diameter D >~ 0.58μm (at RH < 40%) have distinct 24-hour cycles over oligotrophic waters, with an average of 2.3 ±0.6 times higher number concentrations during daytime than at night. Our results show inverse correlation between the ambient aerosol concentration and the magnitude of the observed cycle, suggesting that over areas with high aerosol concentration (e.g., with a strong contribution from long range transport of continental aerosol) the cycle is present but masked. Additionally, the gradual emergence of the diel cycle for larger diameters (Fig. S1) suggests that the longer lifetimes of smaller aerosols (the lifetime of aerosols is inversely proportional to their size) also contributes to masking the cycle, particularly for the small sizes.
While it is known that SSA formation originates from the interaction of wind and waves, with bulk oceanic properties (i.e., SST, salinity, chl-a) affecting it (2), here we show that there is a concomitant daily mechanism that modulates aerosol concentration in a 24-hour rhythm. While we do not provide (or possess) direct measurements of near-surface water microbial processes, the parallel increase of the mean particle size within the ocean surface during the day, driven by photosynthetic growth and secretion of extracellular polymeric substances, points towards a possible link between microbial processes at the ocean surface and the N SSA_0.58µm cycle.
The discovery of the diel cycle in N SSA_0.58µm opens many new questions for future studies to elucidate the mechanism underlining this phenomenon and the direct impact of marine biological processes on the physical properties of the surface ocean, and the link to aerosol uxes and properties. Moreover, on a larger scale, the connection to cloud and rain properties and consequently energy uxes and climate.

Schooner Tara
Measurements were conducted aboard the R/V Tara during the rst year of the Tara Paci c expedition [12][13][14] . The R/V Tara is a 36 m long, 10 m wide aluminum hull schooner with two 27 m long masts, equipped with a meteorological station (Station Bathos II, Météo France) measuring air temperature, relative humidity, and pressure. The station is located on the stern around 7 m above sea level, the wind speed and direction are measured at the top of the mast, ~27m above sea level (asl), and a thermosalinometer (Sea-Bird Electronics SBE45 MicroTSG) measures sea surface temperature (SST) and salinity with its main water entrance located about 0.5-3m under the sea surface (depending on ocean conditions). The intensity of Photosynthetically Active Radiation (PAR; wavelengths between 400 and 700 nm) was measured next to the meteorological station by a QCR-2150 (Biospherical Instruments Inc.). The meteorological station recorded frequencies are listed in Table S2. The SST and salinity were measured at 0.1 Hz and processed to 1 min averages. The PAR is analyzed to 1 min average of 1 Hz measurements.

Continuous aerosol instrumentation and inlet
A detailed description of the aerosol instrumentation during the expedition can be found in Flores et al. 2020 14 . In short, an optical particle counter (OPC; EDM-180 GRIMM Aerosol Technik Ainring GmbH & Co. KG, Ainring, Germany), for continuous aerosol size distribution measurements (from 0.25 -32 µm, sorted into 31 bins), and a custom-made aerosol lter system consisting of four 47mm lter holders and one vacuum pump (Diaphragm pump ME 16 NT, VACUUBRAND BmbH & Co KG, Wertheim, Germany) were installed aboard R/V Tara. Two separate inlets, located next to each other, were constructed out of conductive tubing of 1.9 cm inner diameter and a funnel (allowing the collection of all diameters) and mounted on the rear backstay of Tara. For the Atlantic Ocean measurements, from Lorient, France to Miami, U.S.A., the inlet was installed half way up the backstay (~15m asl) and after Miami, the inlet was relocated to the top of the backstay (~27m asl).
The OPC measures single particles at 683 nm and it was calibrated at the refractive index of polystyrene latex spheres. It collects the scattered light using a wide-angle collector optic at a mean scattering angle of 90°; the optical design smoothes out Mie scattering resonances and reduces the sensitivity to particle shape. A Na on dryer was installed before the OPC, which reduced the sampled air relative humidity to below 40 % 14 . The ow through the OPC was 1.2 liters per minute and it produced a particle size distribution every 60 seconds.
The lters from the custom-made system were changed, in general, twice a day, collecting aerosols for periods of at least 12 hours. The lter holder for the analysis presented here contained 0.8 µm polycarbonate lters (ATTP04700, Millipore) that were stored at room temperature in PetriSlide dishes preloaded with absorbent pads (Millipore, PDMA04700) to keep the lters dry while stored. The ow through the lter was 30 litter per minute for the lters analyzed here.

Continuous water measurements
The R/V Tara was equipped with an ocean surface ow-through autonomous sampling system, similar to the one installed during the Tara Oceans Expeditions, to measure sea surface physical and bio-optical properties as described in 15 . The inline system consisted of a Sea-Bird Electronics SBE45 MicroTSG for measurements of sea surface temperature (SST) and salinity and an AC-S spectrophotometer (WET Labs, Inc.) measuring hyperspectral particulate absorption (a p ) and particulate attenuation (c p ) with a ~4 nm resolution, and an ECO-BB3 (WetLabs Inc.) set in a BB-box of ~4.5 L measuring particulate backscattering at three wavelength (470 nm, 532 nm and 650 nm), altogether mounted in an autonomous setup described in Dall'Olmo et al. 42 and Slade et al. 26 . The size range of the measured particles is > 0.2 µm, but contribution of particles > 20 µm is assumed negligible. Particulate Organic Carbon (POC) concentrations were computed from c p 43 and chlorophyll-a concentrations were estimated from the particulate absorption line height 15 . Additionally, a particle size index (γ), an estimate of the mean particle size in the ocean near-surface waters, was calculated using the wavelength-dependency of c p and that its spectral shape can be approximated as a power law (see main text).

Air mass back trajectory analysis
The presented 48 hours back trajectories in Fig. 1 were calculated using the NOAA's HYSPLIT atmospheric transport and dispersion model 16,17 . They represent the average trajectories of the 'Ensemble option' that were calculated based on an endpoint at 250 m height. We chose the 'Ensemble option' to have a better representation of where the air masses were coming from. We did not use a lower starting height as the minimum height for the optimal con guration of the ensemble is 250 m.

Diameter determination of the diurnal cycle
To obtain an estimation for the minimal diameter impacted by the changes in the diurnal emission, we analyzed each bin from the OPC. We took the data only in the Paci c Ocean and when Tara was at least 100 km away from land (islands included). The OPC has 31 bins for measurements between 0.25 to 32 µm, in Fig. S1 we show the box plot analyses for 16 bins from the OPC, up to 3.0 µm. Fig. S1 shows there is no diurnal difference for the diameters below 0.58 µm, whereas, for larger diameters the cycle is clear.
However, there is an exponential decrease in concentration from the smaller diameter channels to the 0.58-0.65μm channel, suggesting the diel cycle might be masked in the smaller diameters due to higher concentrations.

De nition of the diel cycle
First, all the data was converted to mean solar time (MST) using the equation: where UTC is the Coordinated universal time, and Lon is the longitude in degrees (west < 0, east > 0). MST assumes there is no day to day variation in the UTC of solar noon at a location, and for our dataset it is a good assumption, mainly since the majority of measurements were taken along the tropics.
Each 24 h period was analyzed independently. For a day to be considered to have a diel cycle an increase in concentration had to be observed between 06:00 and 07:00 and a decrease after 17:00, with a greater N SSA_0.58µm observed during daytime (see above for the diameter determination) during daytime. For this, we divided the day into four periods, from 00:00 to 05:00 (dawn), 07:00 to 11:59 (morning), 12:00 to 17:00 (afternoon) and from 19:00 to 23:59 (night). The counts (per xed liter volume) measured by the particle counter can be assumed to follow a Poisson distribution, therefore their standard deviation is , where µ is the mean. Hence, for a day to be considered to have a diel cycle the two following conditions had to be met: µ morning > µ dawn + σ dawn and µ night < µ afternoon -σ afternoon Figure S2 shows the places where the diel cycle was detected using this de nition.
Chlorophyll-a along Tara's route The chlorophyll-a concentration along Tara's route was calculated using the AC-S 15 and to approximate the chl-a concentrations when the AC-S was not functioning, we used the level 3 SNPP-VIIRS satellite monthly data maps. For each month, we used Tara's hourly location to rst extract a 0.2 x 0.2 degree area for each point, then this area was averaged to get a corresponding chl-a concentration at each point. Finally, a 24-hour average was taken along Tara's route. Figure 2B shows the satellite calculated chl-a concentration, and the in situ chl-a inferred from AC-S measurements.
Daytime and nighttime SSA 0.58μm concentration vs Wind speed In order to understand the role of wind speed in the N SSA_0.58µm cycle, we separated the Paci c data (for days when a cycle was detected) into daytime (07:00 -17:00) and nighttime (19:00 -05:00) periods, and binned the total aerosol counts of D > 0.58µm into 2 m s -1 bins ( Fig. 3a; data within 100 km from continental coasts and Japan was not used to avoid pollution artifacts). There were between 3612 to 23136 events per bin used. The Atlantic Ocean data was binned into 4 m s -1 bins for comparison. There were between 949 to 5056 events per bin used.
Rate of change of γ (∂γ/∂t) As mentioned above, γ is an indicator of the size distribution among particles (< 20µm in diameter) in the ocean surface. From Fig. 1 we see γ decrease at daytime (i.e. the sizes of the plankton increase) and increase over nighttime. To quantify the intensity and timing of this change over a full day, we calculated the rate of change of γ. First, to ll in data gaps that correspond to periods when the AC-S was measuring ltered seawater for calibration purposes (normally shorter than 30 minutes), we did a linear interpolation. Data gaps larger than 30 minutes were not interpolated. Then, each continuous segment was smoothed applying a low-pass digital lter with a pass band frequency of 18 hours. Then the rate of change ∂γ/∂t (hr -1 ) was calculated. Finally, Fig. 4B shows a box plot analysis of the days where a N SSA_0.58µm diel cycle was found and there was at least 23 hours of the AC-S data.
Diurnal cycle near Niue Island when Tara was anchored Similar to Fig. 1 in the main text, a diel cycle of N SSA_0.58µm was detected while Tara was anchored near Niue Island (19°03′14″S 169°55′12″W). Figure S3 shows a diurnal cycle of γ, increasing during nighttime (smaller particle mean diameter) and decreasing during daytime (bigger particle mean diameter).

Box plot analysis for three different legs
Similar to the analysis shown in Fig. 3, we performed box plot analysis for the days the cycle was not detected in the Paci c Ocean (Fig. S4A), for the Atlantic Ocean transect (Fig. S4B), for the tour around Japan (Fig. S4C), and for the Fiji -New Zealand leg (Fig. S4D) Day to nighttime ratio vs. aerosol concentration To explore the relationship between the diel cycles and the aerosol concentration, we quanti ed the day to nighttime concentration ratio for D > 0.58 μm vs. the total (using all the bins from the OPC) nighttime aerosol concentration. For this purpose, after converting the data to mean solar time and taking every 24hour period as independent, we rst averaged the total nighttime concentration (from 19:00 to 05:00), next we took a 5-day running average, and nally the data was binned into equally number bins from low to high concentration. A 5-day running average was also taken for the day to nighttime ratio. Figure S4 shows the inverse relationship between the day to nighttime concentration ratio and the aerosol loading. The analysis was also done using the daytime and a 24hr concentration average, no signi cant difference was found.
Scanning electron microscope with Energy disperse X-ray analysis Using Scanning Electron Microscopy with energy-disperse X-ray analysis (SEM-EDX) and a similar particle classi cation scheme as described in Laskin et al. (2012) 18  To perform the SEM-EDS analysis, we used a Zeiss Sigma500 SEM with a Bruker XFlash®-6|60 Quantax EDS detector, and the Bruker ESPRIT feature software package for automatic particle detection and chemical classi cation in EDS.
The SEM was set at a working distance of about 7.5mm (±0.1), an accelerating voltage of 8.0kV, an aperture size of 60μm, and a magni cation of 2000. The backscatter detector was used to acquire the images. For each lter four images, covering a total of 2471 μm 2 surface area, were taken and each particle above a minimum area of 0.08 μm 2 was counted and an EDS spectrum acquired. After the acquisition of the images and EDS spectra, we took only the particles that had an average diameter greater than 0.58 μm and for each of their corresponding EDS spectra, the method described in 44 was used to calculate the mass percent of each detected element. We excluded C from the mass percent calculation since the lters were made of polycarbonate. Following the mass calculation, particles containing sodium above 0.01 mass percent ([Na] > 0) were rst separated from those without sodium. The Na containing particles with more sodium than any other detected element (besides Cl) were denoted "Sea-salt". The rest of Na containing particles were subdivided into two classes: "Metals with Na" if [Na] <  Table S1) we counted a total of 7247 particles and 4560 with D ≥ 0.58 μm. In the 15 nighttime ones we counted a total of 5894 particles and 3706 with D ≥ 0.58 μm. We had between 80 to 781 particles per lter. Figure S5A shows the SSA 0.58μm counts per litter calculated using the SEM images (particle count and area imaged) and the total air sampled. Figure 5SB shows histograms of the chlorine mass percentage found in the particles per lter. Between May 4 and May 9 we see a noticeable Cl depletion, suggesting the atmospheric marine boundary layer in this region had anthropogenic pollutants.

Atmospheric Diurnal anomalies
We calculated diurnal anomalies for the air temperature, relative humidity (RH 7m ), and wind speed (U 27m ), in the Paci c Ocean for two scenarios: 1) the days where a diel cycle in N SSA_0.58µm was detected and 2) when there was no diel cycle in N SSA_0.58µm (Fig. S6). The average between midnight and 05:00 is the baseline for each variable. To avoid continental in uence, this analysis was done only in the open ocean and near the Paci c islands except Japan and Fiji.
The air temperature and relative humidity anomalies show no discernible differences between days where a cycle was detected (Fig S6; panels AirT_a, RH_a) and when there was no cycle (FigS6; panels AirT_b, RH_b). This implies that even though both atmospheric variables have a diurnal signature, their changes, especially in relative humidity, cannot explain the diurnal patterns seen for N SSA_0.58µm .
Finally, the wind speed anomaly analysis (Fig. S6, WS_a, WS_b) also does not show sharp changes at sunrise or sunset; there might be isolated cases, but no consistent pattern.
Effect of atmospheric variables on the SSA 0.58μm diurnal cycle Rain, relative humidity, air temperature, and atmospheric instability can have an effect on the production, growth, transport and removal of SSA. Here we explain why these variables do not explain the N SSA_0.58µm diel cycle.
First, rain suppressed the N SSA_0.58µm cycle ( Fig. 1A and Fig. S5A), as it is a known washout mechanism of aerosols. Second, there are several indications against RH as a major driving factor underlying the detected diurnal cycle in SSA 0.58µm concentration. For example, for a given SSA, its dry diameter is around ¼ of its diameter at formation (1). Hence, if RH variations were the cause, we expect to see the diurnal patterns in all sizes, and especially at smaller diameters, but this is not the case (Fig. S1).
Additionally, the N SSA_0.58µm diel cycles were observed in days with and without daily variations in air temperature and RH (Fig. S7). In addition, RH and air temperature diurnal anomalies have similar trends for days when the N SSA_0.58µm diel cycles was observed and when it was not (Fig. S6). Finally, the atmospheric stability that in uences the transport of aerosols from the ocean surface upward does not explain the diurnal cycle either. Firstly, under most atmospheric conditions, concentrations of SSA with D dry < 10 µm are well mixed in the marine boundary layer, showing little variation with height (1), hence a change in stability conditions will most likely not cause a change in N SSA_0.58µm . In addition, during the Taiwan -Fiji transect, the cycle appeared in three distinct atmospheric states: with clear skies at low wind speeds, with overcast conditions and with trade cumulus throughout the day (see Fig. S8).
Especially, that a cycle is observed even if the morning is overcast (Fig. S8b), and that there are cycles when there is no air temperature variability (Fig. S7), infers that most likely atmospheric stability does not play a signi cant role in the observed diel cycles. Therefore, we conclude that the atmospheric variables cannot explain the observed N SSA_0.58µm diel cycle.
Photosynthetically active radiation (PAR) vs daytime to nighttime ratio of concentration By de nition, solar radiation drives diurnal cycles. Therefore, we explored links between the intensity of solar radiation, measured by the average daytime photosynthetically available radiation (PAR), and the average daytime number count of SSA 0.58µm to determine if the intensity of solar radiation has a measurable effect on the total amount of SSA 0.58µm . We considered the nighttime (background) concentration by calculating the ratio of daytime to nighttime number concentration for the same day and plotted it against the average PAR (Fig. S9). No clear correlation between PAR and N SSA_0.58µm was found. Furthermore, examples of days with similar PAR that showed different daytime N SSA_0.58µm can be seen in Fig. 1B and Fig. S8. This analysis suggests the N SSA_0.58µm diel cycle is not caused directly by changes in solar radiation, but that there is a parallel mechanism.
Surface Ocean variables anomalies Similar to the atmospheric variables anomalies, we calculated the rate of change (∂POC/∂t) for particulate organic carbon (POC) and diurnal anomalies for chlorophyll a, salinity, and SST for days where the cycle was detected and not in the Paci c Ocean (Fig. S10). The average between midnight and 05:00 is the baseline for each variable. To avoid continental in uence, this analysis was done only in the open ocean and near the Paci c islands except Japan and Fiji.
The Chl a and salinity anomalies analysis does not show any diurnal changes. The POC shows a decrease during nighttime and an increase during daytime both when we detected a N SSA_0.58µm diel cycle and when we did not. The daytime increase in POC can be attributed to photosynthetic growth or particle aggregation. Similarly, in the SST anomaly, for both cases, when cycles were detected (Fig. S10, SST_a) and not detected (Fig. S10, SST_b), we see a diurnal signature. However, here we see no difference from the base line up to 09:00, and only a gradual increase from 10:00. Depending on wind conditions, there can be a few degrees difference between SST at the skin of the ocean surface and SST at 0.5-3m depth, and expect to have a stronger diurnal cycle near the skin of the ocean (18), but we don't expect to have a sharp change at sunrise or sunset.

Particle size index γ in different parts of the ocean
To understand the differences in γ and how it might be related to SSA 0.58µm production, we calculated the average γ in 24 hour cycles. Figure S11 shows three different scenarios: Fig. S11A shows the average γ measured in the Atlantic Ocean and in the Paci c Ocean when diurnal cycles in N SSA_0.58µm were detected. Fig. S11B shows the average γ measured during the Keelung -Fiji leg separated by different latitude ranges and the average γ while Tara was anchored near Niue Island. Finally, Fig. S11C shows the average γ measured in the transect from Fiji to New Zealand. This shows the latitudinal dependence of γ at low latitudes and that in the Atlantic Ocean there is a larger mean particle size, which suggest the presence of larger planktonic species than in the Paci c Ocean.
Contribution of small, ~<1 µm particles, to γ variations In order to estimate what are the daytime γ changes associated with, we calculated the contribution of small particles to the size changes observed in γ. To do so, we calculated the backscattering (b bp ) to total particulate scattering (b p ) ratio (b bp :b p ) at λ = 532 nm.
where c p is the particulate attenuation and a p the particulate absorption. Since smaller particles have a higher b bp :b p , the b bp :b p can serve as a proxy for the contribution of small (~<1 µm particles) to the bulk particle size index γ variation. b bp :b p and γ can be taken as two independent variables since the former is the ratio of the amplitudes of the backscattering and the total scattering at a single wavelength (here 532 nm), while the latter refers to the spectral shape of c p .
Since the b bp measurements with the ECO-BB3 sensor have low signal to noise ratio due to its sensitivity to bubble in the water line and accumulation of particles in the sensor, the b bp :b p was rst averaged for one hour periods and then smoothen with a ve hour moving average. Figure S12A shows the cooccurrence of a diel cycle in γ and the b bp :b p in the western Paci c Ocean between 15°N 137°E and 17.6°S 177.4°E. where diel cycles in N SSA_0.58µm were also observed. Twardowski et al. 45 showed that while for γ < 0.8 the backscattering ratio is mostly affected by changes in refractive index of particles, for γ > 0.8 the size of particles become a major contributor to the backscattering ratio. Therefore, the observation of a diel cycle of the backscattering ratio synchronized with γ and daylight in these oligotrophic waters, suggests that the diel cycle in ocean particle size is mainly due to changes in size of pico-phytoplanktonic populations. Pronounced diel cycles of pico-phytoplanktonic populations (e.g., cyanobacteria) in the Equatorial Paci c, have been previously shown 27,46,47 . Figure S12B shows the same as the shaded area outlining the standard deviation. The largest day-to-night ratios are in areas with low chla concentration, i.e., in oligotrophic ("blue") waters. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 3
Dependence of NSSA_0.58µm on wind speed. Box plots of the NSSA_0.58µm vs. wind speed, binned at 2 ms-1 for the Paci c Ocean (panel A) and by 4 ms-1 for the Atlantic Ocean (panel B) where data collected further than 100 km away from land was used. The day and night data are offset for clarity. The y-axis scale is different for the two panels. While the expected increase of aerosol concentration with wind speed is indeed observed, no relation between the 24-hour cycle and wind speed is found.