Diatoms are one of the most ubiquitous and functionally diverse unicellular eukaryotic groups on the planet, contributing to around 20% of global primary productivity and 40% of oceanic primary production (Malviya et al., 2016). Across the global oceans, marine diatom abundance, species composition, and genus diversity exhibit biogeographical patterns in their spatial distributions (Rytter Hasle, 1976; Malviya et al., 2016; Tréguer et al., 2018). In marine waters, pelagic diatom speciation and biogeography is an evolutionary response to environmental selection rather than dispersal, as opposed to freshwater ecosystems (Cermeno and Falkowski, 2009; Cermeño et al., 2010). Along with the biogeography of diatom species, environmental heterogeneity shapes diatom functional traits and diversity over temporal and spatial scales (Key et al., 2010).

Although it is comparatively simple to study the distribution of phytoplankton species using environmental niche modelling (Brun et al., 2015; Barton et al., 2016), it is more difficult to understand the mechanisms involved in structuring the community as a whole (McGill et al., 2006). Many efforts have been made to map phytoplankton distributions based on their functional biogeochemical role using modelling approaches (phytoplankton functional types, PFTs) (Le Quere et al., 2005; Follows et al., 2007). However, only a few species are used to represent phytoplankton groups, which can lead to an oversimplification of their functions (Irwin and Finkel, 2017). One way to refine functional variability is by using trait-based approaches, which define species in terms of their ecological roles that capture key aspects of organismal functionality in the system (Litchman et al., 2007; Litchman and Klausmeier, 2008; Glibert, 2016). Trait-based research has become a novel approach to interpret the data and elucidate the patterns of species distributions (Violle et al., 2007). Thus, instead of using simple taxonomic ecology to explain species distribution, biogeography, seasonality and/or long-term trends, functional traits analysis has the potential to give a realistic prediction of a biological community's response to changes in the environment (McGill et al., 2006).

Trait-based analyses have been used extensively in terrestrial ecology (Cornelissen et al., 2003; Westoby and Wright, 2006; Lavorel et al., 2011). However, they have only recently been applied to freshwaters systems (Weithoff, 2003; Kruk et al., 2010; Schwaderer et al., 2011; Edwards et al., 2013b; Jamil et al., 2014; Zwart et al., 2015) and, far less frequently, to marine ecosystems (Barton et al., 2013a,b; Edwards et al., 2013a), to understand the distributional patterns and re-organisation of phytoplankton communities under climate change scenarios. Examples of functional traits associated with diatom species and morphotypes, which link to their seasonal and/or spatial occurrence, include: cell size (Alves-de-Souza et al., 2008), shape (Pahlow et al., 1997; Leonilde et al., 2017), morphological characteristics (Ligowski et al., 2012; Assmy et al., 2013b; Allen, 2014), chain formation (Pahlow et al., 1997), colony shape (Passy, 2002), attachment mechanisms to a hard substrate (Algarte et al., 2016), nutritional strategies (Edwards et al., 2015), and the level of cell wall silicification, some of which may confer resistance to grazing (Assmy et al., 2013b) or contribute to carbon export (Tréguer et al., 2018).

Trait-based analyses can be extremely insightful when applied to strong spatial and/or temporal environmental gradients, as well as habitats which are under severe climatic pressure. The Labrador Sea sector of the North Atlantic Ocean is an ideal location to apply trait-based approaches to marine diatom species biogeography due to the intense diatom-dominated spring blooms that occur in contrasting hydrographical zones that sub-divide the region into distinct ecological provinces (Fragoso, 2016; Fragoso et al., 2016, 2017). The surface layer of the Labrador Sea comprises blended waters of distinct origin: warmer and more saline (herein defined as Atlantic-influenced for simplicity) waters covering the central deep basin, and cooler and fresher waters (herein denoted as Arctic-influenced) covering the surrounding shelves and shelf margins, but often extending offshore (Head et al., 2013). While an ice-related (i.e., species that grow in ice-melt waters) and polar diatom community develops on the Labrador Shelf (along the Canadian margin of the Labrador Sea) as sea ice melts during spring, Arctic pelagic and neritic diatom species occur near the Greenland coast and subpolar oceanic species are common in North Atlantic waters of the central basin in late spring and early summer (Fragoso et al., 2016).

The Labrador Sea is also a region that is susceptible to climate change (Dickson et al., 2002; Yashayaev and Loder, 2016, 2017). Accelerated sea ice melt has intensified freshening inflow episodes in the Labrador Sea (Yashayaev et al., 2015), which may affect deep mixing and impact spring diatom bloom phenology and community composition (Zhai et al., 2013). The intrusion of the Pacific diatom Neodenticula seminae into North Atlantic waters, including the Labrador Sea, via the Northwest Passage is thought to be a consequence of intense Arctic sea ice melt during recent summers (Reid et al., 2007). Sea surface temperature has also increased during the last few decades in North Atlantic waters, including the Labrador Sea (Yashayaev et al., 2015), and warmer temperatures have promoted further biogeographical shifts that favour warmer-water plankton species in the North Atlantic (Beaugrand et al., 2002). Projections into the future show displacement of phytoplankton niches and diatom species, due to changes in oceanic circulation in the North Atlantic (Barton et al., 2016). Increasing our understanding of the biogeography of diatom functional traits in Arctic and Atlantic waters will further elucidate the re-organisation of phytoplankton communities as oceans continue to warm.

Here, a trait-based analysis was applied to understand the mechanisms driving diatom species' distributions during spring in the Labrador Sea. The trait analysis was based on gathering ecological information on the main diatom species from distinct bioregions of the Labrador Sea based on the literature. Quantitative and qualitative traits, with the latter converted to numerical variables, were analysed together to calculate the mean trait values of different diatom communities. This trait-based analysis allowed an ecological explanation of why certain diatom species thrive in Arctic or Atlantic waters, further contributing to the plankton biogeography typically observed in the Labrador Sea (Sathyendranath et al., 1995, 2009; Li and Harrison, 2001; Head et al., 2003; Platt et al., 2005; Fragoso et al., 2016, 2017).

The Labrador Sea receives inputs of both North Atlantic and Arctic waters, where marked frontal features shape the hydrography along the water masses boundaries (Yashayaev, 2007). In spite of the hydrographical contrast of water masses of Atlantic and Arctic origin found in the Labrador Sea, modification of these waters occurs to some degree, where strong mesoscale physical processes, such as eddies, in addition to riverine and ice melt input (from sea ice, glaciers, and icebergs) lead to mixing, particularly during spring and summer (Wu et al., 2007; Straneo and Saucier, 2008; Yebra et al., 2009).

Few typical Arctic diatom species, such as Bacterosira bathyomphala and Pauliella taeniata (von Quillfeldt, 2001), were found only in Arctic waters in this study. Rather, the majority of diatom species were not categorised solely as being indicators exclusively of Arctic or North Atlantic waters because of the mixing of water masses and the ubiquitous distribution of certain cold-water species in both water masses (e.g., Chaetoceros decipiens) (Tomas, 1997). The diatom species biomasses, conversely, varied strongly between the different water masses of Arctic and Atlantic origin, suggesting that environmental factors selected for the growth and maintenance of a species' population under distinct conditions. This implies that certain traits, common to the different diatom species that co-occur in a mixed community, shaped the biogeographical distribution of these taxa.

The community-weighted trait mean (CWM) approach used in this study assumes that functional traits occurring in a system result from the quantitative records (abundance or biomass) of the individuals carrying these traits (Frainer et al., 2017). Trait-based analysis using CWM is relatively novel in phytoplankton research (Kruk et al., 2015; Klais et al., 2017), perhaps because of the difficulty in identifying phytoplankton to a species level and finding relevant trait information in the literature. CWM was suitable for this study because of the high degree of taxonomic resolution (>70% of total diatom biomass were identified to species' level using the SEM) and trait information available for polar/subpolar diatoms. Other methods, such as modelling of key species are also useful in trait-based approaches because they investigate the mechanisms underlying the structure of a community, in addition to providing future predictions (Edwards et al., 2013a, 2016). However, because of their simplistic resolution, such model-based approaches may also lead to underestimation of other meaningful traits present in a community (Irwin and Finkel, 2017), by disregarding the multi-faceted functions of a species (Paganelli et al., 2012). The main advantage of using CWM when taxonomy (species) information is provided is that the whole (or at least the majority) of the community is assessed, giving an overall picture of the key functional traits of a community. Although considered a hypothetical approach, trait-based analysis based on CWM can be very insightful and can lead to further hypothesis-testing.

According to the trait-based approach based on CWM, the ability to produce IBPs was one trait commonly found in diatom assemblages from Arctic waters, indicating that these species originated from sea ice. Examples of sea ice diatoms found in the water column in this study (though in low concentrations) are Navicula glaciei, Attheya septentrionalis, and Nitzschia frigida, with the latter forming arborescent colonies that help to anchor them under ice-floes (Olsen et al., 2017). Release of sea ice diatoms from melting ice, transfer of sympagic (ice-associated) algae to the pelagic community and accumulation in the water column has been assumed to be the primary causes for bloom development in ice melting regions (Syvertsen, 1991; Leventer and Dunbar, 1996; Ichinomiya et al., 2008), although most sympagic diatoms tend to sink to the seafloor, given that they are unable to adapt to ice free waters (Riaux-Gobin et al., 2011). As the Labrador Shelf is a seasonal ice zone (SIZ), where sea ice melts from spring to early summer (Wu et al., 2007), this explains the presence of sea ice diatoms that are also typically found as planktonic forms (e.g., Fragilariopsis cylindrus) and were present in high concentrations in Arctic waters in this study. Sea ice-melt also provides a haline-stratified layer, where species associated with the transition from sea ice to meltwater and pelagic habitats are suddenly exposed to high-light levels (Smith and Nelson, 1985). This implies that while sympagic diatoms are known to be adapted to low light, transitional (from ice to water column) species have a high plasticity to cope with rapid and relatively large changes in irradiance.

Fragilariopis cylindrus, a diatom species found in large abundances in the sea ice as well as in the water column has been shown to display high degrees of phenotypic plasticity to environmental fluctuations (Kropuenske et al., 2009, 2010; Petrou et al., 2010; Sackett et al., 2013), which suggest that this might be an underlying functional trait in this study. The whole-genome sequence of F. cylindrus shows a high number (25%) of divergent alleles that are differently expressed across variable conditions, and which might be involved in adaptation of this species to rapid environmental fluctuations (Mock et al., 2017). The high gene number of chlorophyll a/c light-harvesting complexes and proteins involved in photo-protection (Mock et al., 2017) may help to explain why F. cylindrus is successful in low and high-light conditions, such as the transition from sea ice to the water column. Other physiological traits, such as the production of photosynthetic carotenoids, UV-protection, and intracellular osmoregulation, have been suggested to be associated with species that produce IBP and could explain the success of ice-related algae in Arctic waters (Mock and Junge, 2007).

Diatom assemblages in Arctic waters also included species that were able to form heavily silicified resting spores. Resting spore formation is a strategy to overcome unfavourable environmental conditions commonly found in Arctic waters, such as low (Zhang et al., 1998) or high light (Oku and Kamatani, 1999), low temperatures (Durbin, 1978), and low nitrate (Kuwata and Takahashi, 1999), phosphorus (Oku and Kamatani, 1995), or iron (Sugie and Kuma, 2008) concentrations. High silicate to nitrate ratios (Si/N) are typical of Arctic waters, including those of the Labrador Shelf, particularly in pre-bloom conditions (> Si/N are found in Arctic waters as compared to Atlantic waters in winter) (Fragoso et al., 2017). These high Si/N ratios likely promote the occurrence of diatom species able to form resting spores, since formation often occurs when there is an excess of Si (relative to N) (Kuwata and Takahashi, 1990). The ability to produce resting spores is also a trait common among neritic diatom species (Garrison, 1981) and, as expected, its level of occurrence was negatively correlated with bathymetry in this study. As the winter progresses, the growth conditions become sub-optimal and resting spores are formed and passively incorporated into sea-floor sediments (and/or sea ice), to be re-introduced into the water-column during spring-time mixing (and sea ice melting) prior to bloom formation. Resting spores are often interpreted as a proxy for sea ice melting (Crosta et al., 2008), and are an important life-cycle strategy for planktonic species that thrive on sea ice melt (e.g., many Thalassiosira, Fragilariopsis, and Chaetoceros species) (Tsukazaki et al., 2013). This suggests that resting spore formation is a common trait in Arctic waters influenced by the seasonal presence of sea ice.

Arctic-related diatoms also shared morphological traits, regardless of their phylogenetic relationship, in this study and elsewhere (Ichinomiya et al., 2008; Arrigo et al., 2010, 2014; Riaux-Gobin et al., 2011; Tsukazaki et al., 2013; Johnsen et al., 2018), such as similar cell and colony shape and low cellular surface-to-volume ratio (S/V). The morphological shapes in Arctic diatoms resembled rectangular prism cells and ribbon-shaped chains (e.g., Fragilariopsis cylindrus, F. oceanica, Fossula arctica, Pauliella taeniata, Navicula septentrionalis, N. granii, N. vanhoeffenii), as well as cylindrical cells and thread-like chains (e.g., most Thalassiosira spp., Porosira glacialis). The “bulky” shape of these diatoms contributed to the low S/V ratios, such as those observed in Coscinodiscus centralis and C. concinnus, Porosira glacialis, and Bacterosira bathyomphala; all species typically found in Arctic waters. Large centric diatoms, with low S/V ratios, also have lower optical absorption cross-sections, which protects them from visible and ultraviolet damage during the high-light levels (Key et al., 2010) typically observed in well-illuminated, ice-melt and shallow stratified Arctic waters (Fragoso et al., 2016, 2017).

Large and heavily silicified diatoms, such as those found in stratified Arctic waters in this study, would tend to sink faster, particularly in non-turbulent waters (Margalef, 1978). However, to avoid sinking, healthy diatoms (> 200 μm³ in cell volume = c.a. 8 μm equivalent spherical diameter) have developed trade-off strategies to encourage positive buoyancy, such as cellular vacuoles filled with low-density ions (Woods and Villareal, 2008), light-weight fatty acids (Duerksen et al., 2014), and extracellular mucous material (Assmy et al., 2013a). Another strategy is the presence of cellular protrusions, such as setae, spines, and long processes, which improves the cell's ability to increase its shear rate (and slow its sinking). In our study, the presence of cellular protrusions was positively correlated with stratification, suggesting that these may slow-down sinking rates of large, dense cells (Nguyen et al., 2011). Arctic diatoms also control their buoyancy in order to mine deeper waters for essential nutrients under low nutrient concentrations, similar to those observed in this study. These diatoms can be categorised as “S-strategists” (stress-tolerant), given that they are large-celled species with low S/V ratios, slow growth rates and are adapted to high-light conditions (Alves-de-Souza et al., 2008).

The Atlantic-related diatom species also shared similar traits among themselves, such as large S/V, MDL, high optimum growth temperatures, and many were solitary forms. Diatoms from Atlantic waters, however, showed less similarity in their CWM (average similarity = 60%, SIMPER analysis) compared to those in Arctic waters (75%). This may be due to high environmental variability, such as may occur during the transition from un-stratified to thermally-stratified waters, as typically occurs in the Atlantic waters of the central Labrador Sea between May and June (Fragoso et al., 2016, 2017). In our study, large S/V and MDL were correlated, and in North Atlantic waters diatom assemblages were characterised by species resembling needle-like cells, such as Cylindrotheca closterium, Thalassiothrix longissima, Thalassionema nitzschoides, and Pseudo-nitzchia granii. Elongated cell shapes have been suggested to improve a cell's ability to maximise nutrient uptake efficiency and light harvesting in a turbulently mixed environment, since it may increase the spin of cells and maintain concentration gradients across the diffusive boundary layer (Smetacek, 1985; Pahlow et al., 1997; Nguyen et al., 2011). Atlantic waters were also less well-stratified (with the exception of a few stations where Rhizosolenia hebetata f. semispina dominated) and had deeper mixed layers, which may favour species which are able to cope with turbulent environments. These diatoms are typical “R-strategists” (disturbance-tolerant), which are species that grow under high nutrient concentrations and are often elongated in cell shape, despite their large dimensions (high S/V), conferring on them the advantage of effectively harvesting light under high mixing (low average light) conditions (Alves-de-Souza et al., 2008).

Diatoms at several stations of Atlantic waters were also weakly silicified compared to the majority of Arctic-related species (Figure 2). Such low cell silicification may relate to the low pre-bloom Si/N ratios characteristic of Atlantic waters, as observed more generally in the North Atlantic (Fragoso et al., 2017). Weakly silicified species include Leptocylindrus mediterraneous and Ephemera planamembranacea, both of which bloom in North Atlantic waters (Le Moigne et al., 2015; Fragoso et al., 2016).

Diatom species optimum growth temperatures in this study were correlated positively with in situ water temperature in both Arctic and Atlantic water masses. However, several diatom species appeared to be restricted to cold, Arctic-influenced waters, whereas others that were more abundant in Atlantic water were also found in Arctic waters, although in lower concentrations (e.g., E. planamembranacea and Pseudo-nitzschia granii). Generally, temperate and polar diatom species have wider thermal tolerances than tropical species and are able to survive under a wide range of environmental conditions (Thomas et al., 2012), which suggests that temperature may have little influence in shaping diatom species distribution in the Labrador Sea. Rather, other factors such as sea ice-melt phenology may serve to confine Arctic diatoms in the Labrador Sea, of which many are sea ice species and/or able to grow in highly stratified, ice melt-regions.

Changes in the Arctic are increasingly evident, including rising temperatures, sea ice decline and increasing freshening of surface waters (Anisimov et al., 2007). Several ecological questions have thus arisen, including whether “the Arctic will be the New Atlantic” if sea ice-melt persists (Yool et al., 2015). Many studies have revealed changes in phytoplankton community composition as a consequence of climate change, such as the intrusion of Atlantic species into high Arctic waters (Hegseth and Sundfjord, 2008). In this study, several phytoplankton species appeared to be restricted to Arctic-influenced waters, whereas some other species, abundant in North Atlantic waters, were also found in Arctic waters but in lower concentrations (e.g., Ephemera planamembranacea). This suggests that species abundant in Arctic-related waters have a more restricted distribution, potentially indicating a greater dependence of certain species to a sea ice or ice-influenced habitat, and well-illuminated and strongly stratified waters.

Sea ice is one of the largest and most extreme habitats in polar oceans, comprising up to 13% of Earth's surface at its maximum (Mock and Junge, 2007). Predictions show that the Arctic will have a total ice free summer by mid-twenty-first century (Wang and Overland, 2012). Loss of sea ice coverage indicates loss of species habitat, with potential losses of diversity and reductions in diatom blooms associated with sea ice melting (Yool et al., 2015). Our results have shown that diatoms from Arctic waters had more similar traits (and consequently less functional diversity) compared to Atlantic diatoms, indicating that the species from the Arctic shared unique characteristics that enable them to thrive in these waters. Traits that are linked to sea ice or Arctic phytoplankton species, such as ability to survive under extreme cold and harsh environments will also tend to decline, implying a great risk to seeding algae and bloom initiation. Diatoms that rely on sea ice to hibernate as resting spores and germinate under stratified conditions will be affected as sea ice cover declines.

In this study, we found that morphological and physiological traits are strongly linked to spring diatom biogeography in Arctic and Atlantic waters. This demonstrates that a functional trait evaluation is a valuable approach to understand changes in phytoplankton communities. Potential applications include examining species succession during seasonal events, such as spring blooms, in long-term time-series data (Barton et al., 2016), or during future environmental perturbation (Tortell et al., 2002).

GF designed and led the study, collecting and analysing the data. IY and EH provided part of the data. GF wrote the initial draft of the paper and GF, AP, and DP supervised the project and together with GJ edited subsequent drafts. All authors provided critical comments on the final manuscript.