Diazotrophic Trichodesmium impact on UV–Vis radiance and pigment composition in the western tropical South Paciﬁc

. We assessed the inﬂuence of the marine diazotrophic cyanobacterium Trichodesmium on the bio-optical properties of western tropical South Paciﬁc (WTSP) waters (18–22 ◦ S, 160 ◦ E–160 ◦ W) during the February–March 2015 OUTPACE cruise. We performed measurements of backscattering and absorption coefﬁcients, irradiance, and radiance in the euphotic zone with a Satlantic MicroPro free-fall proﬁler and took Underwater Vision Proﬁler 5 (UPV5) pictures for counting the largest Trichodesmium spp. colonies. Pigment concentrations were determined by ﬂuorimetry and high-performance liquid chromatography and picoplankton abundance by ﬂow cytometry. Trichome concentration was estimated from pigment algorithms and validated by surface visual counts. The abundance of large colonies counted by the UVP5 (maximum 7093 colonies m − 3 ) was well correlated to the trichome concentrations (maximum 2093 trichomes L − 1 ) with an aggregation factor of 600. In the Melanesian archipelago, a maximum of 4715 trichomes L − 1 was enumerated in pump samples (3.2 m) at 20 ◦ S, 167 30 ◦ E. High Trichodesmium abundance was always associated with absorption peaks of mycosporine-like amino acids (330, 360 nm) and high particulate backscattering, but not with high Chl a ﬂuorescence or blue particulate absorption (440 nm). Along the west-to-east transect, Trichodesmium together with Prochlorococcus represented the major part of total chlorophyll concentration; the contribution of other groups were relatively small or negligible. The Trichodesmium contribution to total chlorophyll concentration was the highest in the Melanesian archipelago around New Caledonia and Vanuatu (60 %), and 565 nm exhibited lower correlations. These results, as well as differences in the PCA of BIOSOPE data, indicated that nL w variability in the greenish blue and yellowish green during OUTPACE was inﬂuenced by other variables associated with Trichodesmium presence, such as backscattering coefﬁcient, phycoerythrin ﬂuorescence and/or zeaxanthin absorption, suggesting that Trichodesmium detection should in-volve examination of nL w in this spectral domain.


Introduction
The ecological importance of filamentous diazotrophs (Trichodesmium spp. in particular) in the archipelago region of the western tropical South Pacific (WTSP) has been suspected for a long time (Dandonneau and Gohin, 1984;Dupouy et al., 1988Dupouy et al., , 1990Dupouy et al., , 1992. Trichodesmium spp. have to be taken into account for estimating the global oceanic nitrogen and carbon fluxes (Capone et al., 1997;Dutheil et al., 2018;Shiosaki et al., 2014). In the past decade, efforts have been made to extract abundances of different autotrophic groups from ocean color data (Blondeau-Patissier et al., 2014;Bracher et al., 2017). Other attempts have been made to get remote sensing estimates of the abundance and diazotroph activity of Trichodesmium at a global scale (Subramaniam et al., 2002;Westberry et al., 2005McKinna et al., 2011;Dupouy et al., 2011;McKinna, 2015). Satellite detection of Trichodesmium is facilitated when concentration at the sea surface is high, leading to a building of mat larger than a 300 m satellite pixel as these mats induce a high reflectance in the near infrared, a "red edge", which can easily be observed (Hu et al., 2010;Dupouy et al., 2011;McKinna et al., 2011;Gower et al., 2014;McKinna, 2015;Rousset et al., 2018). Detection becomes more difficult when Trichodesmium concentrations are at non-bloom or sub-bloom abundance, i.e., when colonies are distributed throughout the water column and mixed with other species. Using an empirical statistical approach, De Boissieu et al. (2014) determined that at sufficient concentration level, these filamentous diazotrophs could be distinguished from other groups. This complements empirical parameterizations that were used to derive the vertical distribution of different phytoplankton groups (micro-, nano-and picoplankton) using high-performance liquid chromatography (HPLC) diagnostic pigments and surface chlorophyll a concentration (Chl a) determined from space (Uitz et al., 2006;Ras et al., 2008;Brewin et al., 2011).
In order to validate Trichodesmium discrimination algorithms, and to improve the knowledge of the influence of Trichodesmium spp. on apparent (AOPs) and inherent (IOPs) optical properties of seawater, accurate field determinations of these properties are required. Among AOPs, normalized water-leaving radiance nL w (λ)(in µW cm −2 sr −1 ), the radiance that emerges from the ocean in the absence of atmosphere, with the Sun at zenith, at the mean Earth-Sun dis-tance (Gordon, 2005), is governed by two main IOPs (Mobley, 1994;Kirk, 1994): volume absorption (a(λ) in m −1 ) and volume backscattering (b b in m −1 ) coefficients. IOPs are controlled by the concentrations of optically active components in a volume of water, which include phytoplankton and colored detrital matter (CDM), the latter being composed of non-algal particulate matter (NAP) and chromophoric dissolved organic matter (CDOM). If AOPs are well related to phytoplankton pigments in Case I oceanic waters (Morel and Maritorena, 2001;Morel et al., 2007), this relationship might be modified by the presence of Trichodesmium (with moderate Chl a concentrations < 1 mg m −3 ). As summarized in Westberry and Siegel (2005), Trichodesmium displays unique optical properties that may allow their detection: (1) a strong absorption in the ultraviolet (UV) domain related to the presence of mycosporine-like amino acids (MAAs) (Subramaniam et al., 1999a;Dupouy et al., 2008), (2) a higher relative reflectance near 570 nm due to phycoerythrin fluorescence (Borstad et al., 1992;Subramaniam et al., 1999b) and (3) increased backscattering across all wavelengths caused by the change in refraction index of intracellular gas vacuoles (Borstad et al., 1992;Subramaniam et al., 1999b;Dupouy et al., 2008).
The WTSP between New Caledonia and the Tonga Trench is particularly rich in Trichodesmium colonies during summer (Dupouy et al., 1988(Dupouy et al., , 2000(Dupouy et al., , 2011Biegala et al., 2014), and this richness is further enhanced during the positive phase of the El Niño-Southern Oscillation (ENSO) in 2003 (Tenório et al., 2018). Using bio-optical measurements, this study aims (1) to describe several AOPs and IOPs of interest in the UV and visible domains of WTSP waters, as well as pigments, and abundance of all phytoplanktonic cells including large and smaller Trichodesmium colonies and picoplankton and (2) to determine the influence of Trichodesmium spp. on in situ measurements of ocean color, and absorption and backscattering coefficients. For this purpose, we used the same type of measurements made in the tropical oligotrophic ocean during the BIOSOPE cruise (Tedetti et al., 2007(Tedetti et al., , 2010.

Study area
The Oligotrophy to the UlTra-oligotrophy PACific Experiment (OUTPACE cruise, 18 February to 3 April 2015) was conducted on board R/V L'Atalante in the WTSP (Table 1, Fig. 1). In situ measurements and water sampling were performed at 15 stations along a 4000 km transect. This transect extended from the mesotrophic waters of the Melanesian archipelago (MA: SD1 to SD6) near New Caledonia and Vanuatu, to the Fijian archipelago between Fiji and Tonga (FI: SD7 to SD12), and to the eastern end in the hyper-oligotrophic waters of the South Pacific Gyre, east of Tonga Trench (SPG: SD13 to SD15). In addition, three longduration stations A, B and C were sampled during 7 days in each of these three regions (LDA in MA,LDB in FI and LDC in SPG;Fig. 1). General biogeochemical and hydrographic characteristics of the waters along this transect are described in Moutin et al. (2017).
2.2 Radiometric measurements and determination of nL w (λ), K d (λ) and Z10 %(λ)values At each station, two or more profiles of downward irradiance (E d (Z, λ) in µW cm −2 ) and upward radiance (L u (Z, λ) in µW cm −2 sr −1 ) were made around solar noon using a Satlantic MicroPro free-fall profiler equipped with OCR-504 downward irradiance and upward radiance sensors with UV-B (305 nm), UV-A (325, 340 and 380 nm) and visible (412, 443, 490 and 565 nm) spectral channels, as further described in Tedetti et al. (2010). The MicroPro profiler was operated from the rear of the ship and deployed 30 m away to minimize the disturbances of the ship. Surface irradiance (E s (λ)) in µW cm −2 was concomitantly measured at the same wavelengths on the ship deck using other OCR-504 sensors to take into account the short-time variations of cloud conditions during the cast. Surface and in-water radiometers were calibrated before the cruise. Mostly, cloudy sky conditions existed during the profiles (only a few acquisitions were made under clear skies), and at SD5 at 17:30-19:00, they were made under a heavy shower. SD3, SD4 and SD13 profiles were not available (night stations). Details of the casts can be found in Appendix A. For the nL w (λ) of the longduration stations, an average on 7 days was calculated as representative of the station, with coefficients of variation of 12-14 % at LDA, 6-9 % at LDB (without day 4) and 2.5 % at LDC. Diffuse attenuation coefficient for downward irradiance (K d (λ) in m −1 ) was determined using E d (Z, λ) and E s (λ) values (Appendix A). The first optical depth corresponding to the surface layer observed by the satellite ocean color instruments (Kirk, 1994) (Z 10 % (λ) in m) was extrapolated from K d (λ) and calculated as ln(10)/K d (λ). Determination of nL w (λ) was conducted from L u (Z, λ) values and diffuse attenuation coefficient for upward radiance (K L (λ) in m −1 ), within different depths according to stations and wave bands, then normalized by E s (λ) (see calculations in Appendix A). The K d (λ) data presented in this study are average values of two to three upward radiance casts (coefficient of variation < 8 % for each station concerned).

Water sampling
Seawater samples were collected during the noon casts at different depths using 12 L Niskin bottles for the determination of various parameters. For the determination of Chl a and particulate (phytoplankton + NAP) absorption coefficient (a P (λ)), samples were collected at depths corresponding to different percents of PAR (i.e., 75, 54, 36, 10, 1, 0.1 %) and filtered (288 mL for Chl a determination by fluorometry, and 2.25 L for a P (λ)) through 25 mm Whatman GF/F filters. Then, the filters were immediately stored in liquid N 2 (−196 • C) in Nunc ® cryogenic vials until analysis. Samples were also collected at all depths for liposoluble HPLC pigment analyses (see LOV laboratory data, OUT-PACE database, J. Ras). In addition, samples for HPLC pigments were taken in duplicate at surface and deep chlorophyll maximum (DCM) as part of a NASA satellite validation program. For this, 3 to 4.5 L of seawater was filtered onto 25 mm Whatman GF/F filters, which were further stored in liquid N 2 until analyses at NASA. Water-soluble pigment (phycoerythrin, PE) concentration was determined for the > 10 µm size fraction. Therefore 4.5 L of seawater was filtered onto 47 mm Nuclepore filters with pore sizes 10 µm and stored in liquid N 2 in Nunc ® cryogenic vials (PE > 10 µm). Filters were preserved (at −80 • C) until analysis in the laboratory (IRD French Polynesia). For the determination of picoplanktonic population abundances (Bock et al., 2018), water samples were fixed with paraformaldehyde (final concentration of 0.2 %) immediately after sampling, flash frozen in liquid nitrogen, and stored in liquid N 2 in Nunc ® cryogenic vials until analysis, and abundances at 5 m were selected for our study. For the determination of CDOM absorption, 200 mL of seawater was immediately filtered on Micropore filters of 0.2 µm pore size using Nalgene filtration units previously rinsed twice with HCL and stored in SCHOTT ® glass bottles, previously combusted (450 • C, 6 h) and rinsed twice with HCL. Pump samples (depth of 3.5 m) were also collected all along three transects in order to increase the frequency of both pigments and IOPs' surface measurements (Chl a, HPLC-NASA) in areas characterized by important Trichodesmium surface slicks: the "Simbada" transect, with 7 samples between SD3 and SD4 in the MA, the high-frequency HF1 transect (31 samples) in the MA near LDA, and the high-frequency HF2 transect in the FI near LDB (42 samples). Besides radiometric measurements and water sampling, in situ measurements were also performed for the determination of Trichodesmium spp. colonies and backscattering coefficients (see below).

FTL Tricho abundance: large Trichodesmium spp. colonies
The Underwater Vision Profiler 5 (UVP5), serial number Sn003, pixel size ca. 0.147 mm × 0.147 mm (Picheral et al., 2010) was coupled to the metal structure of the CTD. The device emitted flashes of red LED light that illuminated 0.95 L of water. Images of all particles within the illuminated area were recorded and analyzed in terms of abundances of defined size ranges. Objects larger than 30 pixels were saved and uploaded on EcoTaxa (http://ecotaxa.obs-vlfr.fr/prj/37, last access: 20 August 2018) and further determined on board as Trichodesmium colonies of fusiform-and roundshaped colonies of all sizes. From 190 074 objects recovered, 100 342 were identified on board as "fiber tricho-like Trichodesmium" (FTL Tricho ), i.e., all particles of Trichodesmium with fusiform-shaped (tuff form) and round-shaped (puff form) colonies from > 200 µm to 2-5 mm in size. FTL Tricho is assumed to be mostly Trichodesmium colonies with a possible risk that a small quantity of fibers could instead be diatoms chains. Contrary to a classical counting at the microscope, no abundance of free filaments is available, although these filaments represent often a significant part of the Trichodesmium assemblage (Carpenter et al., 2004). The FTL Tricho abundance is expressed in colonies per m −3 and measured at 5 m depth intervals (Picheral et al., 2010) providing FTL Tricho "vertical concentrations" at each cast. The FTL Tricho abundance at 5 m depth was generally underestimated compared to that at 10 m and 15 m depths (possibly due to smaller size of colonies). Therefore, the value at 10 m was selected as representative of the abundance of the surface layer. As different FTL Tricho abundance profiles were acquired during the day (from 1 to 5 depending on the station), a daily average of the 10 m FTL Tricho abundance was made. Daily average, maximum value of the day and the FTL Tricho abundance at noon (i.e., the nearest from the time of the Satlantic radiometric profile) were compared and showed no statistical difference. For the three long-duration stations, an average on 7 days of the 10 m FTL Tricho abundance and a variation coefficient were calculated. In order to estimate a trichome concentration, photographs taken with a Dino-Lite hand-held digital microscope covering the totality of the filtered surface on the GF/F filters ded-icated to absorption measurements were used. Colonies were first visually enumerated. The uncertainty on this colony and isolated filaments (essentially Katagnymene) visual enumeration was estimated at 10 %. The trichome concentration (L −1 ) was estimated using a constant number of 10 trichomes per colony as representative of an average of each size class and shapes.

Picoplankton
Picoplankton population abundances were estimated by flow cytometry using a BD Influx flow cytometer (BD Biosciences, San Jose, CA, USA). Prochlorococcus (Proc), Synechococcus (Syn) and picoeukaryotes (Peuk) were enumerated using the red and orange fluorescence, while nonpigmented bacteria and protist groups were discriminated in a sample aliquot stained with SYBR Green I DNA dye, as described in Bock et al. (2018). Using a forward-scatter detector with the "small particle option" and focusing at 488 and 457 nm (200 and 300 mW solid state, respectively) laser into the same pinhole greatly improved the discrimination between the dim signal from Proc at the surface and background noise in unstained samples. Nanoeukaryotes (Neuk) were not further differentiated from Peuk. Cell abundances of Proc, Syn, Peuk and bacteria showed a vertical and uniform abundance distribution due to their mixing in the 0-30 m layer (Bock et al., 2018).

Chlorophyll a, phycoerythrin and pigment analyses
For Chl a determination by the fluorimetric method, filters were extracted with 5 mL methanol in darkness over a 2 h period at 4 • C and quantified using a Trilogy Turner fluorometer according to Le Bouteiller et al. (1992). HPLC pigment analyses on surface and DCM samples were performed according to the NASA protocol and provided monovinyl Chl a (MV-Chl a), divinyl Chl A (DV-Chl a), accessory chlorophyll, and photosynthetic and photoprotective carotenoids (Hooker et al., 2012). PE was extracted in 50 : 50 glycerol / phosphate buffer. Quantification of this pigment was obtained from the area below the fluorescence excitation curve, using a calibrating procedure previously described (Wyman, 1992;Lantoine and Neveux, 1997;Neveux et al., 2006). Furthermore, pigment ratios were also used to estimate the relative importance of pico-, nano-and micro-plankton in terms of Chl a using relative contributions of different accessory pigments divided by the sum of accessory pigments (Ras et al., 2008). The proportion of Proc to total Chl a (TChl a) was estimated from the DV-Chl a / TChl a ratio. It usually represents a high proportion due to its high abundance despite its small size (Grob et al., 2007).

Algorithms for Trichodesmium abundance estimates using pigments
As a true microscopic determination of Trichodesmium abundance was not carried out at each station during the OUT-PACE cruise, we used algorithms to derive trichome abundances from pigment concentrations (chlorophyll, zeaxanthin, PE > 10 µm) and flow cytometric cell counting. Using a constant PE concentration per trichome (196 pg trichome −1 ) and a constant Chl a per trichome (100 pg cell −1 ) as in Tenório et al. (2018), calculations of trichome concentration (L −1 ) could be done both from PE or Chl a > 10 µm, assuming that other autotrophic organisms have a negligi-ble contribution in this large size fraction. As Chl a > 10 µm was not available for OUTPACE, total MV-Chl a was used, which corresponds to the sum of Chl a from Syn and Trichodesmium, and all eukaryotic phytoplankton cells (pico-, nano-and microphytoplankton). MV-Chl a associated with Syn and Peuk was estimated at the surface using measured cell concentrations and the Chl a per cell values obtained on cultures grown under high light intensity conditions (Laviale and Neveux, 2011), i.e., 1.2 fg cell −1 for Syn and 10 fg cell −1 for Peuk (10 fg cell −1 = intermediate between the Chl a per cell of Micromonas and the Chl a per cell of Ostreococcus). Peuk included also Neuk counts. Neglecting the rest of phytoplankton (larger cells; see Tenório et al. 2018) was possible. Microplankton biomass other than Trichodesmium was not significant at OUTPACE (as at DIAPALIS; Tenório et al. 2018). MV-Chl a from Syn + Peuk including Neuk was then deduced from total MV-Chl a to obtain MV-Chl a associated with Trichodesmium. The Trichodesmium spp. abundance was also estimated from total zeaxanthin (TZea). For this, Zea per Proc cell (Zea-Proc) was determined in the area where Trichodesmium is absent and assuming a constant Zea concentration per Syn cell (Zea-Syn) determined on Syn cultures under high light intensity conditions (Laviale and Neveux, 2011). The Zea Trichodesmium was then deduced by subtracting Zea-(Proc + Syn) from TZea (Zea associated with chlorophytes being considered as negligible). Trichodesmium abundance was deduced from Zea concentration per colony found in Carpenter et al. (1993). We then compared estimations of Trichodesmium from these pigment algorithms to FTL Tricho abundance and trichome concentration estimated from visual counts.

Particulate and CDOM absorption, backscattering measurements
Light absorption spectra were measured directly with filters soaked in filtered seawater, by referencing them to an equally soaked empty filter. Measurements were done in a singlebeam Beckman DU-600 spectrophotometer with filtered seawater. Absorbance (optical density) spectra were acquired between 300 and 800 nm in 2 nm steps. To correct the pathlength amplification effect on filters, the optical density of the equivalent suspension (ODs) was obtained from the optical density on filter (ODf) as ODs = AODf + B(ODf) 2 with A and B coefficients determined by Mitchell et al. (1990) as used in oligo-to mesotrophic waters in the Pacific Ocean (Dupouy et al., 1997(Dupouy et al., , 2003. All spectra were shifted to zero in the infrared by subtracting the average optical den-sity between 750 and 800 nm. Optical densities were finally converted into the total particulate absorption coefficients (a P (λ) in m −1 ). The UV absorption by MAAs can be amplified by freezing of filters (Carreto and Carrignan, 2011). The a P (330) to a P (676) ratio was calculated as photoprotection index related to MAAs (330 nm: absorption maximum of shinorine) in total phytoplankton (676 nm, absorption maximum of Chl a), as in Ferreira et al. (2013). CDOM absorption spectra were measured on board with a 200 cm pathlength liquid waveguide capillary cell (LWCC, WPI) as described in Martias et al. (2018). A broad peak around 350 nm was visible in most of the CDOM spectra, except for LDB and SPG stations (not shown).
Backscattering coefficients were determined with the default correction (σ ) applied to compensate for the loss of photons absorbed by the medium between the instrument and the detection volume as described in  from a HydroScat 6 (HOBI Labs, Inc) at six wavelengths (412, 442, 510, 550, 620 and 676 nm). The particulate backscattering (b bp (λ) in m −1 ) was obtained by subtracting the backscattering coefficient of pure water, b bw (Morel and Maritorena, 2001). Due to an electrical shortage inside the instrument, only stations SD1-SD6 and LDA days 1-5 were available, and no stations were sampled after SD6. Backscattering coefficients of surface oligotrophic waters (SD13, LDC, SD14, SD15), which are supposed to depend deeply on TChl a according to Huot et al. (2008) for the southeastern Pacific, were deduced from Chl a using a look-up table of data obtained during DIAPALIS (DIAzotrophy in the Pacific on ALIS) cruises in the Loyalty Channel .

Statistics
Ocean Data View (sections Schlitzer, R., Ocean Data View, http://odv.awi.de, last access: 20 August 2018, 2016) was employed for the spatial representation of biogeochemical parameters over the vertical (0-150 m). The spatial interpolation/gridding of data was performed using Data-Interpolating Variational Analysis (DIVA). Principal component analyses (PCAs) were conducted on the basis of Pearson's correlation matrices using XLSTAT 2011.2.05.

Results
3.1 Distributions of nL w (λ), K d (λ) and Z10%(λ) Along the OUTPACE transect, nL w (λ) showed a large range of values and spectral shape (Fig. 2a). In the UV (305-380 nm), violet (412 nm) and blue (443 and 490 nm), nL w (λ) values were the lowest in the MA, increasing towards the SPG (SD14-SD15, LDC). For all the wavebands, with the exception of the yellowish green one (565 nm), nL w (λ) at SD14 and LDC was higher than the 90th percentile, and nL w (λ) values at SD9 and LDB were lower than the 10th percentile (Fig. 2a). Values of nL w (λ) in this violet-blue domain were similar to those measured in the most oligotrophic oceanic areas at the eastern part of the OUTPACE transect (Tedetti et al., 2010). For example, in the center of the SPG during the BIOSOPE cruise (20-30 • S, 142-126 • W), nL w (412), nL w (443) and nL w (490) reached up to 4.5, 4 and 2 µW cm −2 sr −1 , respectively, for TChl a concentrations < 0.022 mg m −3 and with a DCM at 180 m. The frontal station LDB at OUTPACE exhibited a peculiar spectrum with waters greener than all other stations (Fig. 2b). The low nL w corresponded to a high TChl a concentration of 0.43 mg m −3 formed by Trichodesmium and picoplankton on a surface physical front (Rousselet et al., 2018). Moreover, the GF/F filters used for absorption at these stations showed an orangeyellow color when observed under the Dino-Lite microscope. Such a color was not observed in the MA and is typical of small picoplanktonic cells such as Pro and Syn.
For all stations, K d (λ) decreased from the UV-B to UV-A spectral domain (Table 1). From the MA to the FI, K d (λ)(325) was high from SD1 to SD6, then decreased from SD7 to SD12, and showed a peak at LDB, and minimum at the SPG stations. During the 5-day long-duration stations, K d (325) variations (not shown) reflected those of TChl a with values decreasing from day 1 to 5 at LDA (0.11 to 0.09 m −1 ) and LDB (0.13 to 0.11 m −1 ) and remained stable at LDC (0.05 m −1 ). K d (PAR) showed the same tendency at LDA and LDB (0.028 to 0.023 m −1 ), and LDC (0.020 m −1 ). These typical low values of K d (PAR) in oligotrophic waters were associated with DCMs at 125, 165, 110 and 135 m and a TChl a concentration of 0.036, 0.045, 0.048 and 0.023 mg m −3 measured at SD13, SD14, SD15 and LDC, respectively. Such values are close to those found in the southeastern Pacific during the BIOSOPE cruise (08-35 • S, 142-73 • W) (Tedetti et al., 2007) and much lower than those reported for the oligotrophic waters of NW Mediterranean Sea (Sempéré et al., 2015). Maxima of Z10 %(380) (Table 1; Fig. 3) were found in the FI in the oligotrophic part of the transect (LDC and SD15, 100-120 m, for a TChl a concentration of 0.02 mg m −3 ) and were comparable to those reported for the clearest natural waters in SPG (Tedetti et al., 2007). Conversely, stations exhibiting the lowest Z10% (SD1, 40 m) were found in the MA and also at the frontal station LDB in the FI (DCM of 41 m, TChl a = 0.433 mg m −3 ). The first optical depth determined in the UV-visible varied from 13 m (LDB day 3) to 28 m (SD14).

FTL Tricho abundance derived from the Underwater Vision Profiler
The UVP5 FTL Tricho abundance showed a wide range of values along the transect SD1-SD15 ( Fig. 4a; Table 1). It was essentially concentrated in the upper 50 m, although some  (Table 1)).
colonies were still visible below. The maximum was obtained at SD1 (at 10 m, 7663 colonies m −3 ) and rapidly dropped to 2000 colonies m −3 at SD2 to stabilize between 200 and 500 colonies m −3 at the east of SD4. It progressively decreased from west to east. Still visible at SD5 (170 • E), it vanished at SD7, where the maximum of FTL Tricho abundance was located deeper and finally disappeared between SD8 and SD11. On the first day of LDB, an exceptional high value of 3700 colonies m −3 was observed. During the longduration stations, the average (CV) of 10 m Trichodesmium abundance was 1000 m −3 (35%) at LDA, 1726 m −3 (9%) at LDB and 2 m −3 (1 %) at LDC. FTL Tricho abundance allowed one to define three groups of stations, according to the log10 of abundance. The first group was composed of the stations SD1 to SD7, and included both LDA in the western MA and LDB in the FI (log10 > 2.8). The second group was composed of SD3, and SD8 to SD12 with medium concentrations (2 < log10 < 2.8). Finally, the third group contained the stations SD13, SD14, LDC and SD15 characterized by no or very low FTL Tricho abundance (log10 < 2).

Picoplankton abundance and influence on TChl a
Picoplankton predominance was typical of oligotrophic waters (Neveux et al., 1999;Buitenhuis et al., 2012;Bock et al., 2018). The Syn abundance was particularly high in the surface layer in the MA at SD3-LDA (> 22 × 10 3 cells mL −1 ) until the intermediate area of the Fijian basin. However, the Syn surface maximum was observed at LDB (> 100 × 10 3 cells mL −1 ), together with Proc abundance peaking at LDB with more than 9 × 10 5 cells mL −1 in the upper surface layer and a small Peuk abundance of 850 cells mL −1 . Note that the Peuk abundance was high (> 3000 cells mL −1 ) at the DCM only.

Chlorophyll a, PE and accessory pigments
HPLC pigment analyses revealed the occurrence of four major pigments identified as MV-Chl a, DV-Chl a, zeaxanthin and β-carotene. HPLC pigment concentrations from LOV were used since they were available for each station and depth (Fig. 4b-c). The 0-150 m section of zeaxanthin concentration, the main photoprotective carotenoid contained in all cyanobacteria (Syn, Proc + Trichodesmium), showed values > 0.15 mg m −3 in the 0-50 m layer and almost continuously from SDA to SD12. Furthermore, a strong maximum was observed at the frontal LDB (Fig. 4b). TChl a-LOV and TChl a-NASA (from a regression between only 5 m and DCM values) were highly correlated (TChl a-LOV = 0.81× TChl a-NASA; r 2 = 0.87, p < 0.05, n = 12 and zea− LOV = 0.71 × zea − NASA; r 2 = 0.88, p < 0.05, n = 12; p < 0.0001). However, this good relation was obtained on different bottle casts. TChl a section (LOV: Fig. 4c) showed  (Table 1), with a higher value (0.534 mg m −3 ) and a shallower DCM (52 m) at the frontal LDB. Surface PE > 10 µm values (indicative of Trichodesmium) showed two spots of high concentrations (Fig. 4e). The first spot is located in the western part of the MA (SD1 to SD5), and the second is located at LDB. PE > 10 µm was low in the central part of the transect (between SD6 and SD12) and was near 0 in the SPG.
Higher surface values of TChl a and PE > 10 µm at LDA and LDB (Fig. 4d, e) were obtained from pump samples and provided higher values than surface Niskin samples. DV-Chl a of Proc at the surface (Fig. 5a) tended to increase from west to east until a prominent maximum of 0.18 mg m −3 at the frontal LDB and showed minimum concentrations, higher to the east, in the SPG. It represented 22 % of TChl a in the MA, 39 % in the FI and up to 39 % in the SPG (and 45 % at LDB). The estimates of MV-Chl a in Trichodesmium populations (Chl a-Trich.) using pigment algorithm (see Sect. 2.6) were between 0.10 and 0.23 mg m −3 in the MA, around 0.03 mg m −3 in the FI, with a high value of 0.08 mg m −3 at LDB and lower than 0.02 mg m −3 in the SPG (Fig. 5a). Its contribution to TChl a (Fig. 5b) varied from 52 % in the MA (mean of % contribution from SD1 to SD7) and 30 % in the FI, and it was still 23 % of TChl a in the SPG (SD12-LDC). Its percent contribution at LDB was lower (31 %) because of a high contribution of DV-Chl a. Identical contributions were calculated either using LOV or NASA surface pigments. The contribution of Trichodesmium zeaxanthin followed roughly the same pattern, with a contribution of 53, 40 and only 3 % in the MA, FI and SPG, respectively. Note that the contributions to TChl a or zeaxanthin in the SPG are indicative only, as they are calculated on values < 0.03 mg m −3 . The zeaxanthin contribution was lower at SD1 and was somewhat higher between SD8 and SD11 than the contribution to TChl a (Fig. 5b).

Trichome concentration
The Trichodesmium distribution at the surface deduced from visual counts, UVP5 and pigment algorithms showed grossly similar pattern (Fig. 6). However, at a given site, differences in trichome estimates could be observed according to the method used. These differences could be partly due to patchiness distribution of trichomes (and colonies) and no concomitant sampling of the different parameters. Nevertheless, significant linear correlations between trichome concentrations estimated from PE > 10 µm, or Chl a-trichome, or microscopic visual counts, and FTL Tricho abundance were observed (Fig. 7a). The relatively high slopes of the linear regressions (i.e., 675, 735, 529, as factors between large colonies and trichomes, from PE > 10 µm and Chl a, or visual counts respectively) are explained by the fact that FTL Tricho counted by the UVP5 represented only the number of the largest colonies of Trichodesmium (without true determination of trichome number by colony). The correlation between Chl a-trichome and our microscopic visual counts (r 2 = 0.74) was also significant (Fig. 7b). A maximum of 4715 trichomes L −1 was enumerated on the filter during the Simbada transect from pump samples between SD3 and SD4, at 20 • S, 167 30 • E with TChl a of 0.3 to 0.5 mg m −3 .

Backscattering and absorption coefficients, photoprotection index
All particulate backscattering spectra showed large troughs due to absorption maxima by particulate material in the blue (440, 480 nm) channels (Fig. 8a) as pigment absorption has been shown to influence backscattering intensity . At the most concentrated Trichodesmium stations (slick, SD1), the backscattering coefficient (b bp−H6 ) was twice higher (at 510 nm, 0.007 m −1 ) than in the stations where they were moderately present (SD2-SD6) (as a mean at 510 nm: 0.0025 m −1 ) compared to the value for pure water of 0.0013 m −1 at 510 nm (Morel et al., 2007). The slopes of the b bp−H6 spectra calculated without the blue channels but including the 676 nm channel (i.e., from 510 to 676 nm), as the latter was not biased by Chl a fluorescence  were in the range 0.0017 to 0.0022 m 2 nm −1 , typi-  cal of large cells. The TChl a-specific backscattering coefficient was higher in slicks (0.017 m 2 mg(Chl a) −1 ) and lower in SD3-LDA (0.006 m 2 mg(Chl a) −1 ) near the ones determined on colonies (Dupouy et al., 2008). The section from 0 to 150 m of b bp−H6 showed that the high backscattering co-efficient characterizes the 0-10 m layer in the MA (no data were collected after SD6; Fig. 8b). Typical spectra of particulate absorption for Trichodesmium-rich waters exhibit the two MAA absorption peaks at 330 and 360 nm with a much lower intensity for the 360 nm peak (Fig. 9a). These peaks are characteristic of Trichodesmium spectra (Dupouy et al., 2008), and their amplitude was largely enhanced by freezing (as found on dinoflagellates by Laurion et al., 2003;Carreto and Carrignan, 2011). This MAA absorption has been used in many studies to show the degree of photoprotection of phytoplankton against UV (Ferreira et al., 2013). They were not observed at the surface in low Trichodesmium concentrations (FTL Tricho abundance; Fig. 9b). Sections from 0 to 150 m of a P (330) and a P (440) (Fig. 9c) exhibit the impact of MAAs in the upper layer at 330 nm (a P (330) > 0.4m −1 ). At 442 nm, there was no increase on the surface layer by FTL Tricho and the highest values are rather linked to the DCM at 80 m (a P (440) > 0.2 m −1 ). A reasonable relationship ( Fig. 10a) was found between UVP-5 FTL Tricho abundance and a P (330) when considering the entire 0-150 m layer (FTL Tricho abundance = 0.43 × a P (330) −2.1, r 2 = 0.57, n = 120, p < 0.0001), also indicating that the freezing effect remained proportional to the concentration for our dataset.
The a P (330) / a P (676) ratio showed relatively high values (> 80) from 0 to 25 m; then it abruptly fell to 20 below 30 m depth (Fig. 10b). When considering the surface layer only (Fig. 10c), the MAA index was tightly related to Trichodesmium, except at some stations (SD10). Indeed, MAA pigments are also produced by other phytoplankton groups (Carreto and Carignan, 2011) when exposed to high nL w (UV) values. MAAs of other groups show generally only one peak at 320 nm as in the southeastern Pacific  or at 330 nm (large phytoplankton in Argentina's continental shelf; Ferreira et al., 2013). At OUTPACE, phytoplankton counts indicate that the contribution of other large phytoplankton (shown by the size index from HPLC pigment ratios) was low. Nevertheless, a discrepancy was observed around SD10, where high values of a P (330) / a P (676) ratio corresponded to low UVP5 FTL Tricho and visual counts could be explained by a higher concentration of other photoprotected non-cyanobacterial phytoplankton (the second peak at 360 nm was less pronounced), or a spatial heterogeneity in sampling. At LDB, the mixing with Proc decreased the photoprotection index drastically.

Relationship between AOPs and pigments
In the present study, Chl a was well correlated to all nL w (λ) ratios (nL w (λ) / nL w (565)) with r 2 varying from 0.79 to 0.83 (power regressions) with RMSE (not shown) ranging from 51 to 30 % from 305 to 490 nm for OUTPACE and from 36 to 23 % for BIOSOPE according to wavelength considered (Fig. 11). The relations between nL w (λ) and Chl a were different at OUTPACE than at BIOSOPE (Fig. 11). These good relations obtained even in the UV domain, where Chl a though absorbing at 380 nm does not show any absorption peak in the UV domain, were already observed in the southeastern Pacific during the BIOSOPE cruise, for equivalent ranges and attributed to the fact that CDM substances absorbing in the UV domain co-vary with Chl a (Tedetti et al., 2010). The reason why the relationships are different at 305 and 325 nm wavelengths but not other UV bands (e.g., 340, 380 nm) is probably related to the presence of Trichodesmium and other constituent co-varying with Chl and absorbing more at 305 and 325 nm than at longer UV wavelengths. It can be noted, for the same Chl a, that ratios are higher at OUTPACE than at BIOSOPE; that is, absorption would be lower in the 305 and 325 nm bands and this difference is stronger at high Chl a (rich stations in the upwelling at BIOSOPE, MA stations and LDB at OUTPACE). One possible reason is that CDOM and CDM may be higher in the coastal upwelling or Marquesas Islands waters than in the Trichodesmium-rich waters of OUTPACE.

Influence of Trichodesmium on the distribution of UV-visible nL w (λ)
To better assess the influence of Trichodesmium on the distribution of nL w (λ) values, the eight radiances measured during the southwestern Pacific OUTPACE cruise (this study) and the southeastern Pacific BIOSOPE cruise (2004) were statistically analyzed and compared. Figure 12a-d show the results of a PCA operated separately on nL w (λ) values and TChl a concentrations for the two cruises. In the western tropical South Pacific (OUTPACE), the first two principal components (PCs) represent 94 % of total variance (Fig. 12b). The graph of correlations between PCs and the variables (Fig. 12a) indicates that UV and visible nL w (λ) are distributed along the PC1 axis, with all radiances on the right side, except 565 nm. This first axis (81 % of total variance) indicates an effect of Chl a on nL w (λ), with all nL w (λ) being higher at low Chl a (blue waters) and lower at high Chl a (mesotrophic waters), except at 565 nm, where on the contrary nL w increases with Chl a. Oligotrophic stations are on the right side and mesotrophic stations on the left. PC2 represents 13 % of the total variance. The variables that have significant correlation with PC2 are nL w (565) (Chl a-rich wa- ters) and nL w (490) (Chl a-poor waters), both on the upper side of the PC2 axis. These different behaviors in nL w (565) and nL w (490) are significant compared to the sensitivity of the Satlantic instrument. A series of stations are positively linked to this PC2 axis (LDB4, SD1, SD2, LDA-2, SD7) while LDA-3 and LDA-4 are negatively linked to PC2. The relatively high correlation between PC2 and nL w (565), minimally influenced by Chl a, suggests that parameters other than abundance (e.g., size, type) might affect nL w (565) at the stations with sizeable PC2 values. In comparison, the first two PCs for the southeastern Pacific dataset (BIOSOPE) represent 96 % of the total variance, with 89 % for PC1 and only 7 % for PC2 (Fig. 12c). The main difference is that nL w (565) is no longer linked to PC2, only to PC1, and that for PC2 nL w (490) has an opposite behavior to that in the southwestern Pacific (correlation is negative instead of positive). At 490 nm, Chl a appears to explain most of the nL w variability. This could reflect the absence of Trichodesmium in the east-ern Pacific. Except for a few stations, the PC2 contribution is much lower; that is, variability is mostly described by PC1.

Contribution of other phytoplankton and filamentous cyanobacteria to optical properties for interpreting satellite Chl a imagery
The determination of Trichodesmium's influence on IOPs compared to other microorganisms and non-living particles in the sea is a main challenge. Indeed, previous models showed that absorption is governed by size and intracellular content (Bricaud et al., 1995(Bricaud et al., , 2004 and that the absorption by large Trichodesmium colonies suffers from a double package effect (in filaments and in colony; Subramaniam et al., 1999a, b;Dupouy et al., 2008). Absorption by MAAs was observed on disaggregated colonies rather than on intact colonies ( Fig. 3 in Subramaniam et al., 1999a), suggesting that a large fraction of MAAs is potentially present in sheaths or in the intracolony spaces. It has been shown that the highest a P (330) values in the upper layer, particularly in the western part of the MA, coincided with the highest FTL Tricho (for all stations) and that the correlation was significant (r 2 = 0.55, n = 120, p < 0.0001). Conversely, there was no correlation between a P (440) and FTL Tricho . This lack of correlation is striking as Trichodesmium contribution to TChl a is between 30 % and 60 % along the transect and somehow contradicts the good correlation between phytoplankton absorption coefficient at 440 nm and TChl a of colonies concentrated in tanks (Dupouy et al., 2008). TChl a measured from HPLC large volumes (4.5 L) or a P (440) (2.5 L) catches at a maximum one large FTL Tricho . Indeed, in order to get a representative Trichodesmium biomass or a P (440), it would be necessary to adjust filtered volumes to expected abundance (8 L, Tenório et al., 2018). This is not the case for a P (330) or a P (360) peaks, which are both sensitive to the presence of colonies on the filter and are therefore the best indicators of Trichodesmium abundance. It must also be noted that the high package effect of absorption by Trichodesmium colonies due to a double shadow effect of absorption of light inside the filament, and because of the stacking in a colony (Subramaniam et al., 1999a, b;McKinna, 2015), tends to lower the specific absorption a P (440). Similarly, it was also striking that the in situ red fluorescence signal in the upper layer at OUT-PACE is weak as already found on CTD profiles in the region despite a large abundance of FTL Tricho (DIAPALIS; Tenório et al., 2018). This can be attributed to a low red fluorescence of colonies in the upper layer. Also, the small volume (0.25 mL; Neveux et al., 2010) "seen" by the ECO FLNTU fluorometer of the CTD does not contain many colonies, and the response of large colonies to the blue excitation light is low comparatively to one of the numerous small picoplanktonic cells. Such low responses in absorption and fluorescence could lead to an underestimate of the biomass of Trichodesmium from optical remote sensing.
Backscattering in the ocean is influenced by small particles (< 0.5 µm) of mineral origin, bubbles and colloids rather than by marine living particles (Loisel et al., 2007;Stramski et al., 2008) but also by large Trichodesmium colonies or associated detritus (Dupouy et al., 2008). Recent studies in the open ocean indicate a greater contribution of phytoplankton- sized particles to b bp than theoretically predicted (Dall'Olmo et al., 2009;Brewin et al., 2012;Martinez-Vicente et al., 2013;Slade and Boss, 2015). In oligotrophic waters of the southeastern Pacific, absorption and backscattering coefficients are well related to TChl a with specific relationships (Morel et al., 2007;Huot et al., 2008;Bricaud et al., 2010). At 5 m depth, the OUTPACE H6-backscattering data (b bp (550)) were, on average for all stations, 2 times higher than the b bp (550) calculated from TChl a from the equation b bp = σ [Chl]β during the BIOSOPE cruise (Huot et al., 2008;Stramski et al., 2008). The particulate backscattering was enhanced in the presence of Trichodesmium with a value 2 to 5 times higher at 5 m than at 25 m depth with high Chlspecific backscattering coefficients (0.006 to 0.016 m 2 (mg TChl a) −1 ) and low backscattering slopes (as already shown in tanks; Dupouy et al., 2008). Nevertheless, the layer of the highest backscattering coefficient is situated above the 10 m FTL Tricho , and no relationship was found between vertical distributions of b bp and FTL Tricho , at least in the MA (SD1-SD6, LDA).

Contribution of Trichodesmium spp. to TChl a
All Trichodesmium abundance data, obtained from UVP5, pigments and flow cytometry data, or from visual counts showed the highest abundance in the western part of the MA and the lowest abundance in the SPG, with a high value at LDB. Trichome abundances estimated from pigment algorithms were in the same range as those enumerated by microscopy in the region (at 167 • E, 21 • S, DIAPALIS data; Tenório et al., 2018). The UVP5 counted the largest colonies of the Trichodesmium population, i.e., the upper part of the colony size distribution. The factor between UVP5-FTL Tricho and trichome concentrations visual counts depends on the number of isolated filaments, small colonies and on the number of trichomes per colony; this was defined as an "aggregation factor (AF)". This AF determined when using video recorders on Trichodesmium colonies varied between 400 for the highest to 50 for the lowest (Davis and McGillicudy, 2006;Guidi et al., 2012;Olson et al., 2015). The larger AF found here implies that the FTL Tricho represents a smaller proportion of total trichomes in the western tropical South Pacific than during other cruises or in other regions. The vertical distributions of UVP5-FTL Tricho and trichome concentration show that Trichodesmium populations of both sizes were concentrated below 0 and 30 m and scarce below 50 m, which contrasts with the Trichodesmium distribution inferred from the nifH gene still detected below 100 m (Stenegren et al., 2018). In the western tropical South Pacific, trichomes are generally found from 0 to 60 m (Tenório et al., 2018;Carpenter, unpublished data;Trichonesia 1 in 1998 cruise), the maximum abundance being found between 5 and 10 m depth, with a regular decrease of colonies from 5-15 to 60 m depth. Around 180 • E, Trichodesmium colonies were located deeper than in the Melanesian region. Nevertheless, there might be enough colonies below 20 m (less visible by the satellite) to produce mats episodically, when environmental conditions are favorable, as it was often observed south of Fiji with the CZCS ocean color sensor (Dupouy et al., 1992). On this cruise, visual counts could not detect deep green Trichodesmium colonies as those detected in the Coral Sea at 150 • E (Neveux et al., 2006).
Apart from Trichodesmium, Proc was the other dominant group impacting the Chl a biomass in two parts of the WTSP ocean: (1) the western part of the MA between New Caledonia and Vanuatu, also impacted by a large contribution of Trichodesmium, and (2) the eastern part of the transect (FI) which was more oligotrophic. LDB showed a dominance of both Trichodesmium and Proc, with TChl a proportions of Trichodesmium, Syn + Peuk, Microeuk, Neuk, and Proc of 25, 7, 1.4, 5 and 45 %, respectively.

The influence of Trichodesmium-CDM on
UV-visible water-leaving radiance OUTPACE and BIOSOPE data show that the southwestern and southeastern Pacific surface waters exhibited similar ranges of values for nL w (λ) and Chl a (0.02-0.58 and 0.02-1.3 mg m −3 , respectively). Apart from the "extreme" value of 1.3 mg m −3 recorded in the Peru upwelling (BIOSOPE), Chl a ranges were similar during the two cruises. The fact that nL w ratios were well related to TChl a in the UV domain as well as in the visible domain shows that a strong coupling exists between the UV-absorbing material and Chl a. The contribution of chromophoric detrital matter (CDM = CDOM + NAP) is the sum of the total colored detritus + dissolved absorption . During OUTPACE, a high CDOM amount was associated with Trichodesmium through the formation of CDOM (mainly MAAs) from the colony (Subramaniam et al., 1999a;Steinberg et al., 2004;Dupouy et al., 2008). MAAs identified by their strong UV absorption at 332 and 362 nm are mainly asterina-330 and shinorine, but also minor quantities of mycosporine-glycine, porphyra-334 and palythene-360; all are present in Trichodesmium (Carreto et al. in Roy et al., 2011). A complete analysis of the different components of CDM implies the determination of NAP after bleaching of the filters , but this is biased because of the incomplete degradation of phycoerythrins in the case of high cyanobacterial abundance. Note that the MAA absorption of live colonies is much lower than that of the frozen ones (Dupouy et al., 2008); therefore its impact is much lower on in situ nL w values, OUTPACE and BIOSOPE data differing only in two spectral bands, the yellow-green (nL w (565)) and the blue-green (nL w (490)), according to PCA results. The PC1 axis was linked to Chl a concentration for both cruises while the PC2 was linked to another optically active variable, independent of Chl a, especially for OUTPACE. The PCA shows that during OUT-PACE a significant correlation exists between PC2 and nLw (490) and nLw (565). The fact that the relationship with PC2 is weaker in the southeastern Pacific (and the PC2 contribution about twice smaller) means that these other optical components had little influence during the BIOSOPE cruise -there is practically no effect of PE or particles at high Chl a concentrations. Indeed, Huot et al. (2008) showed that backscattering measured during the BIOSOPE stations (between 41 • and 173 • W) was totally linked to Chl a. The relationship of nL w (490) to PC2 is more difficult to interpret due to its opposite behavior between the southwestern (OUT-PACE) and the southeastern (BIOSOPE) Pacific. One explanation would be that, in the presence of Trichodesmium, a higher backscattering is expected at all wavelengths (linked to a factor other than Chl a) and a PE fluorescence impacting nL w at 565 nm. The fact that nL w (490) is not correlated to TChl a in the same way as nL w (565) implies that the backscattering is not the only driving parameter and that another optical property impacts nL w (490). This could be the absorption effect by zeaxanthin (the major photoprotecting pigment, not totally correlated with Chl a as shown by the PCA) or by a different accessory pigment. The signification of PC2 was explored by performing a new PCA using phytoplankton group index (micro-, nano-and pico-) as additional variables, as well as other parameters (a P (330) and a P (565), PE > 10 µm, UVP5-FTL Tricho and zeaxanthin). New PC1 and PC2 axes explain 47 % and 18 % of total variance, respectively. PC1 is still linked to TChl a. PC2 represents the gradient between the stations influenced by cyanobacteria and those rich in micro-and nano-phytoplankton. The relationship between PC2 and cyanobacteria (rich in Chl b or/and DV-Chl b and zeaxanthin) was inverse to that with the other groups (micro-, nano-phytoplankton). At BIOSOPE, where nL w (490) is essentially a function of Chl a, PC2, the zeaxanthin effect would be negligible or totally linked with the one of Chl a. Our PCA indicated that the two wavelengths (490 and 565 nm) showed anomalous behavior. The latter were chosen by Westberry et al. (2005) to set an algorithm to globally map Trichodesmium high abundance with SeaW-iFS satellite data. The modeled spectrum of a Trichodesmium bloom at equivalent Chl a concentrations to those recorded in that study (0.5 mg Chl a m −3 ) showed higher magnitudes for nL w (490), nL w (510) and nL w (565) (Subramaniam et al., C. Dupouy et al.: Trichodesmium impact on UV-Vis radiance and pigments 1999b) with a nL w (510) greater than nL w (443). Moreover, the authors pointed out the difficulty of direct comparisons between modeled and measured UV-visible radiance due to the uncertainties in Chl a measurements or Trichodesmium abundance. Such spectral responses were not obtained at OUTPACE. This may explain why the model did not provide satisfactory results when applied around WTSP islands , where blooms are numerous as detected by TRICHOSAT particularly in summer (Dupouy et al., 2011). The reason might be that the radiance anomalies at 490 and 565 nm are different than expected versus TChl a or Trichodesmium concentrations, and particularly in the case of moderate abundance.

Conclusions
The OUTPACE cruise in the WTSP from 158 • E to 160 • W provided a unique set of simultaneous measurements of nL w (λ) in the UV and visible domains, pigments and Trichodesmium and picoplanktonic cell abundance along the whole transect during a late summer bloom. Trichodesmium abundance given by the UVP5 (FTL Tricho , i.e., largest colonies) with an AF of 500-700 with trichome concentration by different methods decreased from west to east and occupied the 50 m upper layer of the ocean from the MA to the FI. Such an AF between large colonies and trichome concentration is indicative of aggregation processes and is specific to all cameras towered or lowered in the ocean. Trichodesmium abundance was also well correlated with the absorption peak of MAAs, i.e., a P (330) and the photoprotection index (a P (330) / a P (676)), useful parameters to quantify the latter. The weak CTD-Chl a fluorescence and blue absorption observed in rich Trichodesmium waters tend to underestimate Prochlorococcus abundance if used on profilers or ocean color remote sensing while carrying out the backscattering (high coefficient, spectral troughs) trace surface aggregations. Along the 165 • E-170 • W transect, Trichodesmium together with Prochlorococcus represented the major part of TChl a (a mean of 40 % for the whole transect at the surface, as the other groups were negligible). Trichodesmium contribution to TChl a was the highest (60 % TChl a) in the western part of the Melanesian archipelago (around New Caledonia and Vanuatu) and regularly decreased to the east, in the vicinity of the islands of Fiji, to reach a minimum in the South Pacific Gyre stations, where the Prochlorococcus contribution to TChl a was higher. Profiling Trichodesmium abundance from 0 to 150 m with a UVP5 allowed the detection of colonies deeper south of Fiji, which may produce mats more episodically than at 170 • E. In the WTSP, the relationship between nL w and Chl a was generally similar to that found in the eastern tropical Pacific. In particular, radiance ratios were related to TChl a in the visible, and the UV domain was interpreted as a strong coupling between the UV-absorbing CDM and Chl a. The nL w values were strongly correlated to Chl a except in the greenish blue and yellowish green (490 and 565 nm). These results, as well as differences in the PCA of BIOSOPE data, suggested that nL w variability in the WTSP was influenced by other variables associated with Trichodesmium presence, namely a high specific backscattering coefficient, phycoerythrin fluorescence, and/or zeaxanthin absorption (related to phytoplankton group size). These wavelengths (490 and 565 nm) are often chosen in Trichodesmium detection algorithms. While detecting Trichodesmium mats (above surface) with the "red edge" is possible with MODIS (Rousset et al., 2018), the change in the UV-visible radiance detected during OUTPACE at moderate Trichodesmium concentrations is essential to assess true nitrogen fixation rates in the WTSP as it addresses the general case where colonies are homogeneously distributed over the first optical depth. The use of a hyperspectral profiler to better define the radiance changes linked to Trichodesmium and the development of an instrument to detect the whole Trichodesmium population, including smaller colonies or isolated trichomes, are both required.
C. Dupouy et al.: Trichodesmium impact on UV-Vis radiance and pigments 5265 Appendix A: AOP measurements and processing For in-water sensors, the full width at half maximum (FWHM) of the channels was 2 nm for 305, 325 and 340 nm, and 10 nm for 380, 412, 443, 490 and 565 nm. For in-air sensors, the FWHM of the channels was 2 nm for 305, 325 and 340 nm, 10 nm for 380 nm, and 20 nm for 412, 443, 490 and 565 nm. The MicroPro free-fall profiler was operated from the rear of the ship and deployed 20-30 m away to minimize the shadowing effects and disturbances of the ship. Surface irradiance (E s (λ), in µW cm −2 ), which is equivalent to the downward irradiance just above the sea surface, (E d (0+, λ)), was simultaneously measured at the same channels on the ship deck using other OCR-504 sensors to account for the variations of cloud conditions during the cast. Details of cast measurements are as follows. Rejection was the case at SD6 (second profile), during the long-duration stations LDC (second profile day 1, second profile day 2, first profile day 3, second profile day 5) and LDA (first profile day 5), LDB (second profile day 3) an LDC (second profile day 1, second profile day 2, second profile day 5). In total, all stations were characterized by at least one to two profiles and sometimes three profiles. Only two values of nL w (λ) at 305 nm (SD5 and SD14) showed some suspicious radiometric values among the 30 nL w profiles. E d (λ) was taken from the OCR HyperPro values from 400 to 700 nm and then integrated using the formula (Tedetti et al., 2007, Eq. 1) where E d , PAR(Z) is the downward irradiance in the spectral range of PAR at depth Z (quanta cm −2 s −1 ), λ is the wavelength (nm), h is the Planck constant (6.63.10 −34 J.s), c is the speed of light in the vacuum (3.108 m s −1 ) and E d (Z, λ) is the downward irradiance at depth Z (in µW cm −2 ). Downward attenuation coefficient was determined in accordance with their Eq. (2), where E d (0−, λ) is the downward irradiance beneath the surface. Because of the wave-focusing effects leading to fluctuations in in-water irradiance near the surface, irradiance data of the first meters were omitted from the calculation, and E d (0−, λ) was theoretically computed from deck measurements as in their Eq. (3), where α (0.043) is the Fresnel reflection albedo for irradiance from sun and sky. The diffuse attenuation coefficient for upward irradiance was determined from the slope of the linear regression of the logtransformed upward radiance versus depth in accordance with the equation between L u (Z 1 , λ) and the upward radiance L u (Z 2 , λ) (µW cm −2 sr −1 ] at depths Z 1 and Z 2 (m), respectively (Tedetti et al., 2010). As for K d (λ), the depth interval within the upper water column used for the K L (λ) determination was chosen from a visual examination of each log-transformed profile and was typically 10, 15, 20, or 30 m, depending on the stations and wave bands. The determination coefficients (r 2 ) of the K L (λ) calculation were > 0.98. Water-leaving radiance L w (λ) in µW cm −2 sr −1 was then derived (their Eq. 2) where L u (0−, λ) is the upward radiance beneath the sea surface computed by extrapolating L u (Z, λ) to the sea surface from K L (λ) and Eq. (1), t (0.975) is the upward Fresnel transmittance of the air-sea interface, and n (1.34) is the refractive index of water. Normalized waterleaving radiance (nL w (λ) in µW cm −2 sr −1 ) was determined (equation 3 in Tedetti et al., 2010) by dividing the waterleaving radiance L w (λ) by E s (λ) the surface irradiance and multiplying by F 0 (λ) the solar irradiance at the top of the atmosphere, at the mean Earth-Sun distance (µW cm −2 ). F 0 (λ) data in the ranges 305-340 and 380-565 nm were used from Thuillier et al. (1997Thuillier et al. ( , 1998 as in Tedetti et al. (2010).
Author contributions. CD and RF acquired optical and pigment measurements aboard the R/V L'Atalante during the OUTPACE cruise and were responsible for the analysis of the data and elaboration of the manuscript. MM, BC and MT processed the SAT-LANTIC UV-Vis data and helped in the use of statistics tests. MP acquired and processed the UVP5 data aboard the R/V L'Atalante during the OUTPACE cruise, and LG and FL helped in analyzing the data. JN provided methods for extracting Chl a and zeaxanthin contributions of Trichodesmium. SD acquired the flow cytometry aboard the R/V L'Atalante during the OUTPACE cruise. RS supervised the use of the SATLANTIC UV-Vis radiometer. MR provided phycoerythrin data and ensured high-quality figures. All authors greatly improved the manuscript.
Competing interests. The authors declare that they have no conflict of interest.
Special issue statement. This article is part of the special issue "Interactions between planktonic organisms and biogeochemical cycles across trophic and N 2 fixation gradients in the western tropical South Pacific Ocean: a multidisciplinary approach (OUTPACE experiment)". It is not associated with a conference.