Phytoplankton dynamics in a changing Arctic Ocean

Changes in the Arctic atmosphere, cryosphere and Ocean are drastically altering the dynamics of phytoplankton, the base of marine ecosystems. This Review addresses four major complementary questions of ongoing Arctic Ocean changes and associated impacts on phytoplankton productivity, phenology and assemblage composition. We highlight trends in primary production over the last two decades while considering how multiple environmental drivers shape Arctic biogeography. Further, we consider changes to Arctic phenology by borealization and hidden under-ice blooms, and how the diversity of phytoplankton assemblages might evolve in a novel Arctic ‘biogeochemical landscape’. It is critical to understand these aspects of changing Arctic phytoplankton dynamics as they exert pressure on marine Arctic ecosystems in addition to direct effects from rapid environmental changes. Ongoing Arctic changes are impacting phytoplankton. This Review considers recent primary productivity trends and the environmental drivers, as well as how these are changing, that drive phytoplankton diversity in the region.

T he Arctic climate is in the process of irretrievably moving into a new state, forced by increases in atmospheric levels of carbon dioxide and other greenhouse gases 1,2 . As a result, the years 2010 to 2015 were the warmest five years on record in the Arctic, with a dramatic 5 °C increase in January since 1900 (ref. 1 ). This heat trapped by greenhouse gases triggers a cascade of feedbacks that collectively amplify Arctic warming 1 . Over the next two decades, an additional 4-5 °C increase is predicted in the autumn and winter across the Arctic in conjunction with the projected increase in greenhouse gases concentrations 1 . This expected temperature increase is more than twice that projected for the global average. Associated with the increased temperatures, the Arctic Ocean is experiencing radical modifications in its hydrographic properties (that is, freshwater, salt and heat content) and in its overall circulation. Driven by increasing inputs from multiple freshwater sources (that is, river inflow, net precipitation and melting sea ice and glaciers), the volume of freshwater in Arctic surface waters has increased by 8,000 km 3 (more than 11%) compared to the 1980-2000 period 1,3 . Below the surface over much of the Arctic Ocean, a strong halocline layer acts as a barrier separating Atlantic water heat from the cold and fresh upper Arctic waters 4,5 . Increased stability of the Arctic halocline, a potential consequence of climate change, would likely reduce both the vulnerability of sea ice to upward heat fluxes from the ocean interior and the vertical mixing of carbon and nutrients 6,7 .
At the atmosphere-ocean interface, the traditional Arctic icescape is changing largely due to atmospheric forcing and, to a lesser extent, winter ocean heat flux, particularly in the Eurasian Arctic sector 6 . Since the 1970s, the extent of summer sea ice has decreased by >40% (Fig. 1a) and its thickness by 65%, with first-year ice largely replacing the thick multiyear pack ice (Fig. 1b,c) 1, 8 . In addition to the increased open-water area and duration that amplify ice albedo feedbacks, sea ice is becoming increasingly fragmented and dynamic, resulting in drastically altered under-ice light fields that significantly impact both pelagic and sympagic (that is, sea-ice-associated) ecosystem dynamics.
Together, these climate-driven changes within the atmosphere, cryosphere and ocean have wide-ranging consequences for Arctic marine ecological dynamics, influencing productivity, species interactions, population mixing and pathogen and disease transmission [9][10][11] . To advance knowledge of these unique ecosystems in an era of rapid change, here we have synthetized recent developments over the last decade to study phytoplankton dynamics (the base of marine ecosystems) by examining current changes in primary productivity, phenology and assemblage composition, and how these changes could alter the ecology and biogeochemistry of the Arctic Ocean.

Arctic primary production in a changing icescape
The reduction in sea-ice extent in the Arctic Ocean over the last few decades has resulted in both a longer phytoplankton growing season and increased open-water habitat for phytoplankton growth [12][13][14] . Consequently, phytoplankton blooms now begin earlier 13,15 and end later 16 in the year, and annual net primary production (NPP) in open waters over the entire Arctic Ocean increased 30% between 1998 and 2012 (ref. 17 ), the year that summer sea-ice extent reached its historical minimum. This increase in NPP might have been even larger if it were not for a concomitant increase in cloudiness and decrease in light availability over the Arctic 14 . Over this period, the largest increases in annual NPP were restricted to the interior Arctic shelves (Laptev, Kara and Siberian), where NPP increased 70-112% over the 15-yr study and sea-ice loss was most severe (4.2-5.4% yr -1 ). In contrast, NPP on the outflow shelves showed a much smaller increase (8%, Baffin) and even a decrease (-15%, East Greenland) in response to much smaller declines in sea-ice cover (0-0.9% yr -1 ) 17 .
A more recent satellite-based study 18 that was extensively validated using a large in situ database has shown that since reaching the minimum sea-ice extent in 2012, the rate of increase in open-water area in the Arctic Ocean has slowed considerably, decreasing from 88,000 km 2 yr -1 between 1998 and 2012 to only 620 km 2 yr -1 between 2012 and 2018 (Fig. 2a). In addition, the average ice-free period in the Arctic only increased from 147 to 152 open-water days (five days) between 2012 and 2018. Thus, neither the extent of open-water phytoplankton habitat nor the length of the growing season changed significantly during this time. Nevertheless, annual NPP in ice-free waters continued to increase at rate of 6.8 TgC yr -1 between 1998 and 2018, peaking at 391 TgC yr -1 in 2018 ( Fig. 2c). Notably, despite the flattening in the trend in open-water area between 2012 and 2018, the increase rate in annual NPP over the Arctic during that time period (13.5 TgC yr -1 ) was double that between 1998 and 2012 (6.4 TgC yr -1 ). This suggests that the factors controlling annual net primary production changed between the early and late phases of the 1998-2018 time series.
Whereas the increase in annual NPP between 1998 and 2012 was highly correlated to the decline in sea ice and an increase on open-water phytoplankton habitat, the increase in annual NPP between 2012 and 2018 was associated with increased surface chlorophyll a (Chl a) concentrations (Fig. 2b,d). Prior to 2012, concentrations of Chl a across the Arctic Ocean were relatively constant, increasing by only 0.002 mg m -3 yr -1 (or 0.27%). However, between 2012 and 2018, mean Chl a concentration in Arctic Ocean surface waters increased at a rate of 4.3% yr -1 , 16 times faster than the rate prior to 2012. The largest increases in Chl a between 1998 and 2018 were measured on the inflow shelves of the Chukchi Sea (26%) and Barents Sea (61%), and at the recently exposed shelf break of the Laptev Sea (Fig. 2d). If we assume that increases in Chl a concentration are indicative of increases in phytoplankton biomass, the latter of which cannot increase without additional nutrients, these results suggest that increased nitrogen (N) supply from either increased vertical mixing near the shelf break or advection from lower latitude waters may be responsible for the NPP increase in ice-free Arctic waters since 2012 (ref. 18 ). However, we cannot eliminate the possibility that changes in grazing pressure also may have impacted phytoplankton populations, and hence NPP [19][20][21][22][23] .

Biogeography shaped by changing environmental drivers
The complexity of Arctic biogeography makes it particularly challenging to fully comprehend the interplay of environmental drivers and their change and impacts on phytoplankton dynamics (Figs. 3 and 4). Here, we attempt to highlight at the regional scale (that is, between inflow, interior and outflow shelves and the central Arctic; following the terminology of Carmack et al. 24 ) how the    185 and tschudi et al. 186 ).
major environmental drivers will change and how new ones might emerge in a changing Arctic Ocean.
Nutrient supply and vertical mixing. Nutrient supply (primarily nitrate 25,26 and, to a lesser extent, silicate [27][28][29] and rarely iron 30 ) drives the biogeography and trophic status (that is, oligotrophic or eutrophic) of the Arctic Ocean, while light availability modulates the rate of NPP within each region 29,31,32 . These key factors (that is, nutrients and light) are in turn regulated by a complex interplay of processes altering stratification and mixing as well as the presence of sea ice and snow cover superimposed on seasonal and latitudinal controls on light availability 33 (Fig. 3). Winter nutrient inventories show a strong linear relationship with annual NPP via their role in regulating phytoplankton biomass, except for areas where wind-driven and topographically enhanced mixing (that is, driving resuspension and upwelling events) resupply nutrients during the growing season 31 . Locally, depending on the interplay between atmospheric forcing (that is, intensity, duration and direction of the wind stress) and the strength of vertical stratification, the injection of nutrients into surface layers can easily vary by two orders of magnitude across the Arctic Ocean 31,34,35 . Stratification is expected to increase in the Canada Basin 36 but decrease in other regions (for example, in the Eurasian sector 6,37 ; Fig. 4). On the other hand, the loss of sea ice is expected to increase winds and strong wind events by the end of the century due to reduced atmospheric stability resulting from increased temperature and turbulent fluxes 38 . These enhanced atmospheric forcings occur in all seasons but especially in autumn and winter (where they are expected to strengthen by up to 50% in the central Arctic and peripheral seas) 38 . Changing the balance between vertical stratification and atmospheric forcing will be subject to antagonistic and region-specific environmental drivers (that is, advection, mesoscale activity, atmospheric forcing and related processes, such as upwelling) and will ultimately alter phytoplankton dynamics.
Advection at the Arctic gateways. The Arctic marine biome is tightly connected to lower latitudes through the northern Pacific and Atlantic oceans. In inflow shelves (Fig. 3), Atlantic and Pacific waters flow northward through the European Arctic Corridor and the Bering Strait, respectively, carrying heat, nutrients and planktonic organisms to the Arctic Ocean. A twofold increase in North Atlantic current velocities over the last 24 years was recently revealed, explaining decadal variations in the spatial distribution of the coccolithophorid Gephyrocapsa huxleyi (previously called Emiliania huxleyi 39 ), a tracer for temperate ecosystems [40][41][42] . Bio-advection, rather than the previously assumed water temperature, may be the major mechanism responsible for poleward intrusions of southern species like coccolithophores 40 (and potentially the picocyanobacteria

Synechococcus 43 ). A similar increase in poleward advection of
Pacific waters through the Bering Strait suggests that the shrinking Arctic domain may be prone to intrusions of temperate species at both of the Arctic gateways 44 . However, as described previously, advective transport can result in a case of life or death in the Arctic Ocean depending on the ability of these invasive species to survive strong environmental gradients, low water temperature and long periods of darkness 45 . Advective processes alter mostly lower trophic levels that have a limited capacity for mobility (for example, phytoplankton, zooplankton and some fish), and consequently can impact entire marine ecosystems by shifting species distributions and modifying interactions at higher trophic levels [46][47][48] . Considering the role of bio-advection in ecological models (that is, trait-based and niche-based approaches) will be necessary to improve predictions of future ecosystem shifts in the context of climate change 49 . Not only are the inflow shelves shaped by advective processes, the outflow shelves (for example, the Canadian Arctic Archipelago-Baffin Bay complex and the East Greenland shelf) also respond strongly, and likely differently, to upstream changes in sea-ice dynamics and export, hydrographic structure and biogeochemical cycles 33,50 . In both northern Baffin Bay and the East Greenland shelf, changes in stratification and sea-ice dynamics (that is, extent and export) shape spring bloom phenology and nutrient inventories for the phytoplankton growing season 51,52 . On the East Greenland shelf, the magnitude of sea-ice export can explain spatial shifts in the boundaries between stratification regimes (that is, haline-versus temperature-based stratification 53 ) and different phytoplankton dynamics (for example, more intense and early phytoplankton blooms in haline-based stratification). However, strong haline-based stratification reinforced by increasing freshwater input in the North Water Polynya within northern Baffin Bay, a biologically productive oasis for marine mammals and birds, resulted in a sharp decline in phytoplankton biomass (especially centric diatoms) and NPP during 1999-2011 (ref. 54 ). In addition, early season nutrient consumption by phytoplankton blooms under the sea ice and/or at the sea-ice edge in waters that are advected into North Baffin Bay can contribute to this overall productivity decline along with a change in the algal community toward smaller cells 54 . Assessing the impacts of climate change on the inflow and outflow shelves clearly requires proper consideration of the connectivity between the northern Pacific and Atlantic Oceans and the central Arctic Ocean.
Land-ocean continuum. Through the land-ocean continuum, peripheral environmental pressures play an increasing role in the functioning of Arctic marine biogeochemical cycles. In particular, increased inflow of glacial meltwater and freshwater from fjords and river mouths (Fig. 3) will dramatically alter coastal biogeochemical cycles. Outflow from fjords impact near-shore nutrient inventories (phosphorus, silicate and iron) 55-58 and can even fuel large summer phytoplankton blooms 21 . Interestingly, in the Greenlandic fjords, two types of glaciers (that is, marine-and land-terminating glaciers) have contrasting effects on phytoplankton productivity. Rising subsurface meltwater plumes originating from marine-terminating glaciers trigger upwelling of nutrient-rich deep water, which supports high summer phytoplankton productivity 59 . On the other hand, fjords with land-terminating glaciers or anomalously strong freshening events lack such upwelling processes and result in increased stratification, which lowers productivity 59,60 . As a result, future transitions from marine to terrestrial glaciers, and the associated increases in stratification intensity, will noticeably alter local productivity with potentially cascading effects on trophic levels. The increase in pan-Arctic river inputs, driven by an intensifying Arctic water cycle, is also expected to affect marine coastal biogeochemical cycles. However, while nutrient supply by rivers can be locally important, it does not seem to support a significant fraction of Arctic NPP 31 . For example, field observations and modelling at the mouth of the Mackenzie River in the Beaufort Sea have shown that inorganic N from rivers was contained and entirely consumed within a few tens of kilometres of delta outlets 61-64 , creating potential near-coastal biological hotspots. On the other hand, large river outflow will intensify the freshwater stratification and light-absorbing properties (that is, mainly via coloured dissolved organic matter and, to a lesser extent, particles 65,66 ). However, recent findings suggest that it does not appear to decrease phytoplankton productivity by limiting wind-driven shelf-break upwelling and decreasing light transmission 18 . Shelf-break waters have been shown to harbour massive phytoplankton blooms 67,68 and are likely to form an ' Arctic Green Belt' 64 as they become increasingly sea-ice-free and are exposed to atmospheric forcing 50 .
Wildfire and Arctic aerosol deposition. In recent years, an unexpected increase in the frequency of wildfire events has been observed in summer in boreal forest 69 and tundra ecosystems at a pan-Arctic scale 70 . Although wildfires are common in the northern hemisphere, they are becoming particularly unusual by their latitude, duration and intensity 69,70 . Such extreme wildfire activity has been clearly correlated to high temperatures and dry conditions, and increased atmospheric disturbances (that is, thunderstorms) 71 . Increased wildfire activity in these areas can release large amounts of carbon (C) and N to the Arctic atmosphere 72,73 and, by deposition, could alter sea-ice optical properties and, potentially, marine biogeochemical cycles (Fig. 3). Light-absorbing particles (including black C) is well recognized to be an efficient absorber of solar radiation 74 , and its deposition onto, or incorporation into, snow and sea ice can decrease the surface reflectance and increase melt rates [74][75][76][77] . More uncertain is whether wildfire-derived aerosols can be used as a new source of nutrients (mainly N species) in the summer, when N limits the Arctic phytoplankton NPP. Estimates of atmospheric nutrient deposition are scarce in the Arctic Ocean, but some global modelling budgets suggest that N and phosphorus inputs are low [78][79][80] . Further evaluation is clearly needed to quantify and incorporate these intensifying climate-driven changes in Arctic biogeochemical models (Fig. 4).
Ocean acidification. Some of the fastest rates of ocean acidification have been recorded in the Arctic Ocean due primarily to the higher capacity of cold water to absorb CO 2 , but also due to inputs from river run-off and ice melt and the inflow of naturally low pH waters from the Pacific 81 . However, the severity of acidification is not homogeneous across the Arctic Ocean, resulting in high temporal and spatial variability [82][83][84] (Fig. 4). Overall, Arctic and subarctic phytoplankton communities seem to be relatively resilient to ocean acidification, with no significant change in productivity and little change in species assemblages under enriched scenarios up to

Fig. 4 | Synthesis of selected observed regional environmental changes in the Arctic Ocean. For each region, environmental changes and impacts for
Arctic phytoplankton are shown. For environmental changes, yellow and green colouring indicates an increase and decrease, respectively. the confidence level (+) refers to the confidence in attributing observed changes for phytoplankton dynamics. Question marks (?) indicate that the evaluation has not yet been assessed at the regional scale, or that there is insufficient evidence for the evaluation.

A new non-traditional Arctic phenology
The growth of phytoplankton at high latitudes is generally thought to begin in open waters of the marginal ice zone (MIZ) once the highly reflective sea ice 93,94 retreats in spring, as solar elevation increases and surface waters become stratified by the addition of sea-ice melt water [95][96][97][98] (Fig. 5a-c). The largest and most well documented of these blooms was observed in the Chukchi Sea beneath fully consolidated sea ice 67,68 . Physiological characteristics of the phytoplankton beneath the ice on the Chukchi shelf in early spring indicate that they were 'primed' to bloom once light levels became high enough to support net photosynthesis 116 . At their peak, these under-ice phytoplankton blooms (UIBs) reached Chl a concentrations in excess of 20 mg m -3 and extended from the ice-water interface to a depth of 25-70 m and from the ice edge to >100 km into the ice pack. The algal biomass associated with these features rivalled that of the most productive ocean ecosystems on Earth 67 . The observation of intense UIBs beneath fully consolidated sea ice has necessitated a re-examination of many aspects of Arctic marine ecology and biogeochemistry. Because these UIBs are invisible to satellite sensors, recent estimates of annual NPP in waters where UIBs develop may be at least an order of magnitude too low 67 , indicating that seasonally ice-covered waters on Arctic continental shelves have the potential to support vastly higher rates of NPP than has been attributed to them in the past 14,17,101 . In addition, the presence of UIBs shift maximum nutrient consumption and NPP to earlier in the season and farther into the ice pack relative to MIZ blooms. This is important because the timing and location of NPP and associated zooplankton grazing can directly influence the partitioning of organic C between the upper water column and benthic communities, and thus the efficiency of the biological pump and C sequestration 117 . Furthermore, in productive Arctic waters where diatoms dominate, food webs tend to be short and even small changes in NPP export pathways can have large cascading effects on higher trophic level organisms 118 . Therefore, it is important to determine how export rates of newly fixed C and the degree of benthicpelagic coupling will respond to a shift in location and timing of peak rates of NPP from the MIZ 119,120 to deeper within the ice pack.
'Borealization' of Arctic phytoplankton phenology. Given the ongoing changes in phytoplankton phenology, the 'polar' paradigm of a single annual phytoplankton bloom and subsequent sedimentation of organic matter needs to be revised. Not only has the start of the growing season been profoundly accelerated, but the end of the growing season has been markedly delayed (Fig. 5) 121 . This finding supports an ongoing borealization of the Arctic Ocean and represents a paradigm shift with respect to Arctic Ocean NPP that is likely to impact both ecosystems and biogeochemical cycles in unpredictable ways. For example, a new phenological feature, a secondary (usually autumn) phytoplankton bloom, is developing in parts of the Arctic Ocean 16,[122][123][124][125] . Although secondary blooms were already prevalent in sea-ice-free regions of the Eurasian Arctic sector influenced by Atlantic waters 42,126 , they are now developing in seasonally ice-covered regions of the Arctic Ocean due to concomitant delayed freeze-up (greater light availability) and increased exposure of the sea surface to wind stress (wind-driven mixing brings nutrients to the surface 29,127 ). However, these secondary blooms are necessarily restricted to low Arctic latitudes where stratification is not sufficient to inhibit wind-driven mixing (Fig. 5c), and the growing season is long enough to allow phytoplankton growth during any vertical mixing events (Fig. 5b,c). In a changing Arctic Ocean, the spring bloom will undoubtedly remain the major annual NPP event for C export to higher trophic levels and sequestration in the deep ocean and sediments 121,127,128 . However, changes in phytoplankton phenology or additional pulses of phytoplankton NPP may alter the food web structure and lead to major ecosystem level changes in an environment where consumers must make the most of the short productive period before the long winter sets in 16 .
The borealization of the Arctic phytoplankton phenology will also be accompanied by a longer period of regenerated production and subsurface Chl a maxima (SCM), particularly in low productivity waters (for example, in the Western Arctic Ocean; Fig. 5). Widespread N deficiency in surface waters drives the seasonal persistence of SCM layers and phytoplankton C biomass in several reg ions 29,64,129,130 . The contribution of these layers to NPP is possibly higher in the Arctic Ocean than in thermally stratified waters of the subtropical gyres due to a combination of extreme acclimation to low light and a shallow nitracline 31,131 . Unsurprisingly, the deepening (related to the length of the growing season and the seasonal N consumption) and the steepness (associated to the intensity of the vertical stratification) of the nitracline determines the depth, biomass, productivity and assemblage composition of SCMs 64,132,133 . For the deepest SCMs in late summer, even shifts from autotrophic to heterotrophic communities have been revealed due to severe light limitation 132 . In the context of current sea-ice loss, the potentially increased role of SCM layers on biogeochemical fluxes remain to be quantified directly, both regionally and at the pan-Arctic scale 31,127 .

Arctic phytoplankton assemblage structure
Since the beginning of the nineteenth century, known Arctic and subarctic phytoplankton species have increased dramatically from 115 (ref. 134 ) to 1,874 (ref. 135 ) in 2011, and ultimately reaching 2,241 taxa in the latest updated pan-Arctic taxonomic inventory in 2017 (ref. 136 ). Here, our goal is not to describe this expanded understanding of Arctic phytoplankton diversity but rather to delineate the ecological niches of phytoplankton ranging from bloom-forming to ultra-oligotrophic adapted species in the changing Arctic environment (Fig. 6).
The prevailing view of Arctic Ocean phytoplankton assemblages is usually based on large diatom blooms (centric or pennate forms) in nutrient-rich waters supporting a cascade of higher trophic levels. What is less recognized is that some Arctic regions (for example, the Canada Basin) are as oligotrophic and unproductive as subtropical gyres 12 . The availability of N (mainly nitrate), which shapes the overall biogeography and regional trophic status of the Arctic, also drives large-scale shifts from diatom-to picoeukaryote (mostly flagellates)-dominated systems 29,64,137,138 (Fig. 6). This does not mean, however, that in less productive areas no diatom-dominated spring blooms can develop (except probably in perennial stratified Arctic waters), but only that these will be less intense and likely shorter (Fig. 5). In warmer, stratified and oligotrophic Arctic surface waters with a longer sea-ice-free season, the picoeukaryotic systems (in particular, the endemic prasinophyte Micromonas polaris) [139][140][141] are strongly suspected to rely on mixotrophy; that is, via osmotrophic and/or phagotrophic processes. Such an adaptive strategy of picoeukaryotes could increasingly be favoured over strictly phototrophic phytoplankton as sea ice continues to decline 138 . In addition, such a scenario could considerably increase the risk of the proliferation of potentially bloom-forming, harmful or toxic species currently    128,188 . b,c, Potential changes in production and duration of the spring bloom due to early productive season and the occurrence of under-ice phytoplankton blooms at low (b) and high (c) Arctic latitudes 67,68,109,114,189 . In low (b) and high (c) Arctic latitudes, the presence and extent of sea-ice algae blooms could be affected by drastic changes in sea-ice dynamics 103,107,190 . With a longer ice-free season, an increase in the importance of SCMs (in both low and high Arctic latitudes) and secondary/autumn blooms (only in low Arctic latitudes, when wind-driven mixing is strong enough to erode the vertical stratification and that replenishes sun-lit surface waters in nutrients) in the annual NPP estimates is expected 16 135,140,[142][143][144] , potentially affecting already fragile marine Arctic ecosytems 145 . Diatoms, which commonly exhibit seasonal succession from pennate to centric species due to different light requirements 146,147 , are not the only bloom-forming taxa observed in the Arctic Ocean (Fig. 6). Two other types of phytoplankton blooms have been reported, dominated either by the prymnesiophyte Phaeocystis pouchetii or by coccolithophores (Fig. 6). P. pouchetii blooms have long been observed in the Eurasian Arctic 148,149 but are relatively recent phenomena in Labrador fjords 150,151 , Baffin Bay 108 and under sea ice 109 . Silicate limitation can likely explain more favourable growth conditions for P. pouchetii compared to diatoms 108,152 and the co-occurrence or succession between diatoms and P. pouchetii 153 . Since the early 1990s, silicate concentration has decreased by 20% (and nitrate by 7%) in inflowing Atlantic waters due to natural multi-decadal changes in surface circulation and decreased depth of winter convection at lower latitudes 42,154,155 . Thus, as Atlantic Ocean silicate supplies continue to decline, P. pouchetii blooms may become more common in eastern Baffin Bay waters and perhaps at higher latitudes of the Eurasian Arctic sector, while diatom-dominated blooms will become less frequent. Coccolithophores, considered as new intruders to the Arctic Ocean and sentinel taxa in temperate waters, have already been discussed in detail in the previous subsection titled ' Advection at the Arctic gateways' .
Understanding overwintering strategies (including mixotrophy, dormancy and cyst or spore formation), dark survival and recovery among phytoplankton taxa are also urgent matters to resolve, particularly with ongoing environmental changes (for example, increasing water temperature and sea-ice loss). Cyst and spore production has been reported in polar phytoplankton taxa but remains limited to a few diatom and dinoflagellate species 147,156 . Arctic diatoms have also been reported to survive long periods of darkness and resume rapid growth as soon as light becomes available due to their ability to downregulate their metabolism and to maintain the main components of their photosynthetic machinery (PSII, pigments and Rubisco) at low light and temperature 157,158 . On the other hand, mixotrophy seems to be a widespread strategy among dinoflagellates and other photosynthetic flagellate taxa, keeping them active throughout the polar night 139,[159][160][161] . With higher temperatures in the future, photosynthetic flagellate taxa that rely on heterotrophy could have an advantage over diatoms that exhaust stored metabolic resources more quickly 140,162 during the polar night 140 . The taxon-specific survival traits between diatoms and flagellates could shape their geographical distribution in high latitude regions and thus the community structure during the preconditioning period of the spring bloom, thereby impacting both the phenology and magnitude of the bloom 163 .

Future Arctic Ocean and open questions
It is evident that the shift of the Arctic Ocean to a new atmospheric, cryospheric and oceanic state is resulting in profound and widespread changes in phytoplankton dynamics. Despite the many examples of the effects of this new Arctic 'biogeochemical landscape' on phytoplankton productivity, phenology and assemblage composition, foreseeing the consequences of intensifying climate change remains difficult. Model predictions can be particularly relevant to identify key climate change-mediated multi-stressors 164,165 and to better portray the future Arctic Ocean 166,167 . However, the taxonomic complexity of phytoplankton assemblages within diverse Arctic marine ecosystems remains difficult to model and parameterize correctly due to multiple little-known factors affecting growth 168 . For example, no model has been able to reproduce the observed changes in phytoplankton phenology (for example, the occurrence of under-ice and secondary/autumn blooms) due to unique temperature-and light-dependent growth as well as metabolism of Arctic phytoplankton communities 116

Fig. 6 | Major phytoplankton taxa and its environmental drivers in the Arctic
Ocean. the schematic diagram delineates the main Arctic phytoplankton taxa driven by specific environmental drivers (that is, nutrient status, light and mixing regime, and water temperature). N and Si indicate nitrogen and silicate availability, respectively. Warm/stratified versus cold waters is related to hydrographic properties of the water masses. the different branches separate the phytoplankton taxa: the shift from diatom-to flagellates-based system (that is, high to low N) 29,64,137,191 , the intra-specific taxa shift from pennate to centric diatoms (that is, low to high light) 107,146,147,192,193 , the shift from diatoms to (1) colonial prymnesiophytes, namely P. pouchetii (that is, high-to-low Si, cold waters), and to (2) coccolithophores (that is, high-to-low Si, advected warm/stratified waters) 28,40,108,150,194 , and the changes in phytoplankton taxa in the flagellates and picoeukaryotes 64,138,142 . Note the species E. huxleyi have recently reassigned based on genomic analyses to the coccolithophorid G. huxleyi 39 . exemplified by recent observations that have challenged the perception of the polar night, which has been shown to be a more active period for marine ecosystems and intense biogeochemical fluxes than previously recognized 158,[171][172][173] . Almost all recent winter expeditions have revealed unexpected scientific breakthroughs 172,173 , and much remains to be done to document the key role of the polar night as a true continuum between the autumn and spring, and to better integrate it into the Arctic biogeochemical cycles. Through the mosaic of environmental drivers influenced by climate change presented here, it is difficult to highlight particular 'sentinel' regions to be monitored in the near future: the interior shelves are mostly shaped by the land-ocean continuum, inflow and outflow shelves by advection, and the central Arctic by likely irreversible sea-ice loss. With the continued shrinking of the Arctic marine biome, however, multiple questions will need answers to better anticipate changes in the biogeochemical cycles of the central Arctic and predict its future. It is not yet possible to make definitive predictions about whether the central Arctic will become a new oasis or desert within the Arctic marine biome. However, we suspect that the shelf break, acting as a 'green belt' , could effectively supply inorganic and organic matter to the strongly stratified, but soon-to-be sea-ice-free in summer, central Arctic Ocean 64,174 . The key question is whether or not the disappearance of the summer sea-ice barrier will enhance atmospheric forcing and increase the efficiency of mixing on C cycling and nutrient exchanges.
As for the Arctic gateways, the importance of Pacific and Atlantic inflows that fundamentally shape Arctic biogeochemical cycles highlight the strong connectivity of the Arctic Ocean to the global ocean. In this Review, we mainly focused on the importance of potential changes in inflows from subarctic latitudes that carry heat, nutrients and planktonic organisms towards the Arctic Ocean. The reverse, also important but not addressed here, is the role of a changing Arctic Ocean on subarctic latitudes and the global circulation and climate. The initiation of industrial-era decline in subarctic Atlantic Ocean primary productivity, for example, seems to coincide with the onset of Arctic Ocean surface warming 175,176 . More importantly, the increasing long-term influx of freshwater into northeastern subarctic Atlantic surface waters may be implicated in driving the industrial-era Atlantic Meridional Overturning Circulation (AMOC) decline and contributed to this primary productivity decline over the late nineteenth and twentieth centuries [176][177][178][179][180] . Continued weakening of the AMOC, as projected for the twenty-first century 181,182 , may therefore result in further productivity declines, with important ramifications for future atmospheric C drawdown 183 and northern Atlantic fisheries 184 in the subarctic Atlantic Ocean 176 .