Different taxonomic assemblages and ecosystem dynamics
Among the six modules identified in the co-occurrences network, two confirmed the link between transparency and low export. Module C was related to the summer stratification community. At this time of the year, new nutrient inputs in the ecosystem were scarce and taxa were typical of a recycling community. It means that nutrients are recycled by coprophagy, particle-feeding and throughout the trophic network which helps to maintain zooplankton biomass in the upper mixed layer during periods of intense stratification of the water column (Kiørboe, 1997). Images analyses further revealed that the community tended to be more transparent when module C was dominant which could be explained by the presence of low carbon content organisms (Siphonophorae, Salpida). The taxa in module C fit in these ecological niches. Oithona sp. biomass is negatively correlated to the vertical flux of faecal pellets carbon (Svensen & Nejstgaard, 2003) and they also are also known to reduce the vertical flux of faecal pellets (González & Smetacek, 1994).Paracalanus aculeatus , a small copepod typical of low production ecosystems which helps in nutrients recycling, was also present (Paffenhöfer et al., 2006). The abundance of the recycling community and the presence of organisms known to have a low carbon content explain the reduced carbon export during summer stratification. Module C contained five Gastropoda taxa, three of which do not have any shell reducing their dry weight content (volume:C ratios 20 to 100 times those of non-gelatinous organisms; Kiørboe, 2013) and contribute to the overall transparency of the community (Alldredge, 1984). Found commonly in the samples, Atlanta selvagensis has a transparent shell (de Vera & Seapy, 2006). Along with Thliptodon sp., they are strong recyclers which feed on other mesozooplankton, including Cymbuliasp. (personal communication of Amy Maas; Mifsud, 2001). Further strengthening the link between opacity and low carbon export,Cymbulia sp. and Peracle reticulata are particle-feeders associated with recycling communities. Module F, although relatively less abundant than module C, was also negatively correlated with carbon export. It included Doliolida, Larvacea or Bryozoa taxa whose centroids are on the transparent and low export side of the morphospace (PC2 > 0, Supp. Figure III), further linking transparency and recycling community. Again, it contained taxa indicating a recycling community. For example, Halocypris sp. are predominantly detritivores (Angel, 1983) and play a role in the recycling of organic matter (Castellani & Edwards, 2017). In addition to having the same transparency signal as module C, module F was also linked to a community with more visible appendages (PC3 < 0, Supp. Figure III),e.g. , Bryozoa larvae can use their cilia to capture particles (Strathmann, 2006).
Despite the above, an opacity-export signal could have been expected for two other modules but was not measured in this study. Module A contained almost exclusively taxa known to migrate, and using the co-occurrences network, we identified these taxa potentially interacting (Berry & Widder, 2014) that we would have missed otherwise. Euphausidae, except for the genus Stylocheiron, , are in general strong migrators (Brinton, 1962; Roe et al., 1984) and known to contribute to the diel community changes at BATS station (Blanco-Bercial, 2020). Siphonophorae and Salpidae also realise diel vertical migrations (Andersen et al., 1992; Lüskow et al., 2019). DVM was shown for several genus of ostracods (McHardy and Bary, 1965; Ursella et al., 2018; Kaeriyama and Ikeda, 2002). Even though DVM is a process known to significantly contribute to the biological pump (Steinberg et al., 2008), no significant positive correlation was found between carbon export and module A relative abundances. There are several possible explanations for this discrepancy. Euphausidae migration ranges are usually larger than 200 meters (Andersen & Sardou, 1992), meaning that sediment traps might have missed their contribution to the overall export. A possible explanation for DVM is the predator-evasion hypothesis, in which large or pigmented organisms can hide at depth during the day to be less visible while transparent organisms can stay in surface waters (Johnsen, 2014; Sha et al., 2021). However, no correlation was found between migrators and opacity, which might be explained by the non-amplification of certain taxa (see Quantitative property of metabarcoding). For example, Amphipoda are common vertical migrators in the Sargasso Sea (Steinberg et al., 2000). Individuals were part of the morphological dataset and very opaque in a large majority (PC2 < 0, Supp. Figure III), however they were not present at all in the molecular dataset, leading to a potential DVM signal discrepancy. Finally, Module D matched with the zooplankton bloom community when the peak of biomass usually occurs (Madin et al., 2001; Steinberg et al., 2001). This community was composed of organisms more elongated without visible appendices, e.g. , Larvacea, Chaetognatha (PC3 > 0, Supp. Figure III). Module D relative abundances were positively and strongly correlated to carbon export; however, no opacity signal was found. It might be due to the gelatinous plankton taxa it contained (five Oikopleura sp., four Doliolidae and one Salpidae). Indeed, by their ability to bloom and/or to produce high quantity of faecal pellets, they can play an important role in carbon export (10-79% of the 200-m POC flux for salps, Stone and Steinberg, 2016), but by being highly transparent (PC2 > 0, Supp. Figure III; Johnsen and Widder, 1998), their abundance might have skewed the expected opacity-export signal.
Imaging data were obtained from formalin-fixed organisms through the Zooscan process. Grey colour levels and shape might then not exactly represent how organisms looked when alive, however, it is known that the position of copepods preserved in formalin remain stable after an initial shrinkage period (Sampei et al., 2009) which is much shorter (in the order of days) than the time gap between sampling and scanning in our study. Because more than 1500 organisms were scanned per size-class and sample, we believe that shape-related morphological descriptors, such as Size or Elongation (PC1 and PC3, Figure 1), were robust at the scale of the community. Similarly, all the samples were stored under exactly the same conditions, following the same protocols, and therefore underwent the same potential degradation. There is still the fact that fixation affects each taxonomic group differently (e.g.gelatinous vs. arthropods). The morphological descriptors linked to the transparency of the organisms are however inter-comparable since we are analysing trends and changes in the community, therefore the patterns found would then be reliable.