Climate change is predicted to increase average global temperatures by 0.3–0.9°C per decade. Polar regions are expected to experience the highest temperature increases (Christensen et al., 2007), with far-reaching effects on the Arctic (Wassmann and Reigstad, 2011). Furthermore, an acceleration of temperature increases is also anticipated in some regions (around the Antarctic Peninsula and near-coastal Antarctica) of the Southern Ocean (Christensen et al., 2007).

The ecological and physiological adaptations of phytoplankton to temperature influence their global distribution (Huertas et al., 2011; Thomas et al., 2012). The effect of temperature on phytoplankton growth has been widely observed (Raven and Geider, 1988; Coello-Camba et al., 2014, 2015). The thermal performance curves of these organisms usually follow a typical shape of being bound at extreme temperatures with a single intermediate mode (Angilletta, 2006; Figure 1). This mode represents the optimum temperature (T_OPT) for growth. The farther away the temperature moves from this optimum, the lower the growth rate. Phytoplankton communities account for most primary production in the oceans (Sakshaug and Holm-Hansen, 1984). These pelagic organisms have optimum temperatures for growth that are strongly related to mean annual temperatures at their geographical location, demonstrating a clear latitudinal trend in their T_OPT (Thomas et al., 2012), with higher T_OPT for growth at lower and hence warmer latitudes. Warming is therefore expected to cause poleward shifts in the range of species (Parmesan, 2006; Thomas et al., 2012) as they vary their biogeographic distribution to accommodate changes in ambient temperatures (Loarie et al., 2009; Burrows et al., 2011). Nonetheless, for species inhabiting polar oceans no “colder waters” will be available as refugia (Boyd et al., 2013). The high-latitude niches occupied by species with low T_OPT values will diminish, necessitating that these species acclimatize or adapt to higher temperatures to survive (Boyd et al., 2013). Besides this, an increase in the number of subpolar phytoplankton species moving into the polar region is likely to happen, particularly in the Arctic Ocean, where advection processes help to connect this ocean with waters at lower latitudes (Hunt et al., 2016). Under such conditions, competition will become a strong selection factor as the environment becomes better suited to incoming species from lower latitudes and resident communities will be challenged.

Although thermal performance curves have been described for several organisms (Loarie et al., 2009), few datasets presenting thermal curves for polar phytoplankton are available (i.e., Thomas et al., 2012). This may be a consequence of the remoteness of polar waters and of the limited number of polar phytoplankton species that have successfully been cultured in the laboratory.

The goal of this paper is to identify the thermal thresholds of phytoplankton in polar waters and to predict possible implications of increasing temperatures for growth and bloom development. To do so, we created a growth-vs.-temperature database using data from the literature and new experiments. Then, based on this database, we calculated the community T_OPT for growth, a parameter that represents a turning point beyond which temperature becomes a restrictive factor. We also calculated activation energies (E_a) to illustrate the steepness of the changes in growth rates above and below the optimum temperature and to infer the capacity for phytoplankton to respond to increased temperature in a warming polar sea. We then compared thermal thresholds with recent surface seawater temperatures (SSTs) in both polar oceans to evaluate how close phytoplankton populations are to their thermal limits. Finally, we modeled a future SST scenario to predict when temperatures will rise above the thermal thresholds of phytoplankton in polar waters. Temperature predictions will be especially relevant to the Arctic and to the spring bloom, a key process that accounts for the largest part of annual primary production (Leu et al., 2011) and that is tightly coupled with melting ice (Leu et al., 2011). Alterations in the timing and/or quality of this key peak in primary production may consequently alter Arctic food webs and possibly influence biogeochemical cycles.

Our efforts to find studies on optimal temperatures for growth of phytoplankton in polar areas demonstrate that, until now, few datasets describing thermal curves for polar phytoplankton have been published despite their relevance to warming trends currently observed in polar ecosystems. Limitations of the compiled dataset include few temperature treatments, which contributes to uncertainty in the identification of the optimum temperatures for growth in some cases. Moreover, there is a limited number of polar phytoplankton species that have been isolated and successfully cultured in the laboratory, forcing the use of natural communities to test the responses to temperature. This implies that, in some experiments, phytoplankton growth rates could have been constrained due to natural nutrient limitations. Besides this, the geographical distribution of the available data is biased toward some particular regions of the Arctic and Southern Oceans (Figure 2). More specifically, most of the Arctic studies on this topic come from the Svalbard Archipelago (as also pointed by Wassmann et al., 2011), an area influenced by the inflow of the West Spitsbergen Current, and the Canadian Arctic-Baffin Bay area. In the Southern Ocean, most data have been obtained next to the Antarctic Peninsula and the Ross Sea. A seasonal bias can also be found in general in polar research, as the access to study areas is often difficult in the coldest seasons and regions.

Among all the phytoplankton groups studied here, genera such as Thalassiosira spp., Chaetoceros spp., Fragilariopsis spp., and Phaeocystis spp. have been usually reported as dominant organisms in the phytoplankton standing stocks associated with spring blooms in the Arctic (Barber et al., 2015). These genera are similar in the Antarctic.

The thermal thresholds determined here do not describe the physiology of these phytoplankton groups, but reflect their ecological preferences. This way, we can depict the actual responses to warming of the phytoplankton groups and autotrophic communities inhabiting polar oceans as a whole. The optimum temperature for growth represents a turning point beyond which temperature becomes a restrictive factor for organisms. By conditioning growth, temperature changes could trigger substantial alterations in phytoplankton responses and therefore in the whole polar ecosystem. The dataset compiled here agrees with the description of Thomas et al. (2012) of lower T_OPT for phytoplankton species living at high latitudes (Figure 3). We observe that the relatively low T_OPT values found (Table 1) agree with the low mean annual SST in polar regions: from 1.8°C in the Arctic (Payne et al., 2011) to −0.2°C near the Antarctic Peninsula (Etourneau et al., 2013). For some phytoplankton, T_OPT varied between different estimates (i.e., Micromonas spp. T_OPT: 3.02–8°C). This may suggest that there is genotypic plasticity in responses to temperature that may reflect different adaptations of the strains. This plasticity could be associated with geographical differences of the strains as Micromonas showed a higher T_OPT in the Canadian Arctic than did the strains from the European Arctic area that showed the lowest values (Lovejoy et al., 2007; Coello-Camba et al., 2015). According to our results, the T_OPT estimated for the phytoplankton community from the Southern Ocean (5.2°C) was lower than that estimated for the Arctic phytoplankton community (6.5°C). This difference may be due to the limited connectivity of the polar waters of the Southern Ocean, as the Antarctic Convergence creates an oceanographic barrier that limits exchanges with warmer subpolar areas (Hasle, 1969). This lack of connectivity may result in a higher proportion of cold-adapted species and endemisms (Balech, 1970) that pull the T_OPT average down.

The unimodal growth-vs.-temperature curves of both Arctic and Antarctic phytoplankton communities were characterized by left skewness (Dell et al., 2011), as reflected by the higher E_a calculated above than below the optimum (Dell et al., 2011). The E_a values for the fall traits fell within the confidence intervals and were close to the mean value determined by Dell et al. (2011), who performed a global analysis of the thermal E_a of various organisms. The E_a for the rise traits obtained here were low; nevertheless, they agreed with the mean value of 0.50 (±0.18) eV estimated by the same authors for unicellular organisms (Dell et al., 2011). According to this left skewness, phytoplankton living in polar oceans is expected to be more sensitive in responding to increased than to decreased temperatures, as the decrease in the growth rates above T_OPT is sharper than the increase in these rates below T_OPT (Thomas et al., 2012).

In the Arctic Ocean, summer SSTs have already surpassed the thermal thresholds observed here for phytoplankton. In 2013 at 70°N, temperatures remained above 6.5°C during the first 2 weeks of August, reaching a maximum value of 7.2°C on the 12 August 2013 (Figure 6). In contrast, during the last century, temperature increases in the Southern Ocean have been less severe (Christensen et al., 2007), with the highest warming described in the vicinity of the Antarctic Peninsula (Chapman and Walsh, 2007). At present, maximum SSTs in this area (80–50°W, 75°S-Antarctic Convergence) reach up to 4.2°C in January, 1°C below the T_OPT for growth that we measured here for Antarctic communities.

Temperature predictions are especially relevant with respect to the Arctic spring bloom, a key process that accounts for the largest part of annual phytoplankton primary production (Lavoie et al., 2009). Leu et al. (2011) presented a conceptual model that described a close coupling between melting ice and primary production in the Arctic Ocean, where pelagic algal blooms require a considerable amount of thinning and broken-up ice. On the basis of Leu's model, Wassmann and Reigstad (2011) added predictions for future Arctic phytoplankton blooms that include earlier bloom development as the ice cover shrinks faster in spring (Wassmann and Reigstad, 2011). Increasing thermal stratification could prevent nutrient supplies during the post-bloom period, softening summer production peaks or even making them disappear (Wassmann and Reigstad, 2011). In addition, in those Arctic regions where strong wind-driven mixing events can weaken the strong vertical stratification, new fall blooms may appear (Ardyna et al., 2014), shifting toward a double bloom production mode, more typical in temperate waters.

Here we implemented both conceptual models by adding recent seasonal SST variability corresponding to year 2013 (http://www.ncdc.noaa.gov). At present, when spring blooms develop after the ice melts, the average temperature remains below 1°C (Figure 6), except at 70°N where there is no ice cover and the main bloom occurs between April and May. At 75°N, a weak and slow surface stratification in spring (Wassmann and Reigstad, 2011) delays the beginning of the bloom by a few weeks. At the heavily ice-covered latitudes of 80 and 85°N, the bloom begins even later in the year, after the ice melts closer to summer. In July 2013, temperatures increased faster at 70°N to a maximum of 7.2°C in August. This seasonal warming became less intense toward the pole, with a maximum SST of −0.4°C at 80°N. Finally, beyond 85°N where sea ice persists throughout the year, the temperature gradient is nearly absent and SST remains stable below −1°C (Figure 6).

We estimated the seasonal variability of SST in the Arctic Ocean by 2100 by assuming a 3°C increase over time, a rather conservative prediction taking into account that some regional climate models predict increases of up to 8°C by the end of this century (Collins et al., 2013). As in most future scenarios, our predictions are characterized by steeper temperature gradients with strong ice break-up processes. Predictions indicate that temperature increases would be stronger at the lowest latitudes. Spring temperatures are expected to remain below 1°C at 85°N, while average temperatures would rise higher at 70°N or 75°N as we approach 2100. In fact, throughout the year, conditions at 75°N will be similar to the present-day situation at 70°N, with thermal stratification becoming increasingly stronger at the lowest Arctic latitudes (Leu et al., 2011; Wassmann and Reigstad, 2011). Besides this, predicted SSTs at 70°N are expected to exceed the T_OPT all summer long by the end of this century and during the first half of August at 75°N.

As the seasonal ice melts, temperatures will increase more abruptly, although our predictions for 2100 do not anticipate that SST will exceed the T_OPT of Arctic phytoplankton growth in spring (Figure 6) when the bloom for primary production begins. Nonetheless, increased temperatures may lead to subsequent increases in the growth rates of phytoplankton (Figures 4, 5), resulting in a fast increase in blooming biomass. As a consequence, nutrient consumption may also be accelerated, giving way to a more rapid decrease at the end of the bloom (Figure 6). However, under this low emission future forecast, we expect that the T_OPT threshold would not be exceeded in spring. Polar phytoplankton species should therefore be able to grow properly and dominate the spring blooms (Figure 6). If we also consider the predicted increase in pCO₂ by 2100 (Meehl et al., 2007), the acceleration of the Arctic spring bloom could be even larger as a consequence of the fertilizing effect of pCO₂. This could compensate present pCO₂ limitation in those areas where nutrients are still available and temperatures remain below 6°C (Holding et al., 2015). This dependence of the fertilizing effect of increased pCO₂ on temperature adds to the temperature threshold of 6.5°C predicted here for the growth of Arctic phytoplankton communities.

As the temperature of polar seawater increases, the principle of competitive exclusion implies that more thermo-tolerant phytoplankton groups (i.e., species or ecotypes) would out compete cold-water phytoplankton groups with lower growth T_OPT in polar waters. Some phytoplankton species can exhibit phenotypic plasticity to temperature (Reusch and Boyd, 2013). Among cold-water communities, extremely low temperatures have forced specific cellular and physiological adaptations (Morgan-Kiss et al., 2006). Phytoplankton can adapt to temperature under stable conditions (Huertas et al., 2011). The low interannual variability in temperature of these extremely cold niches has helped that adaptation but environmental increases in temperature above the regular values is limiting the capacity of low-temperature phototrophs to adapt to large increases in SSTs (Morgan-Kiss et al., 2006; Bonebrake and Mastrandrea, 2010; Huertas et al., 2011). This way, under future increased temperatures, a compositional change of communities from less to more thermo-tolerant groups rather than local adaptive evolution would be expected.

According to this principle, once seawater temperatures in a warmer polar ocean surpass the T_OPT of resident communities, these high-latitude niches will be opened up for incoming subpolar temperature-tolerant species. However, studies addressing the plasticity of phytoplankton species in polar waters to temperature changes are still limited, so predictions on community changes remain a challenge. Nonetheless, toward the end of this century, SSTs are expected to remain low during winter, with ice formation (Serreze et al., 2007) and limited light availability for primary production (Berge et al., 2015). Under these unfavorable conditions, many polar phototrophic plankton are able to persist by following different strategies such as facultative heterotrophy, storage and utilization of energy reserves at a reduced metabolic rate and formation of resting spores (Palmisano and Sullivan, 1982; Zhang et al., 1998). Incoming temperate species may not have the potential to survive such long periods of darkness (Peters, 1996; Peters and Thomas, 1996), and thus there are uncertainties in predicting their ability to offset polar species during the polar night. Polar species may then retain a competitive advantage surviving during the coldest months, and they will likely maintain their key role in seeding the subsequent spring polar bloom, as long as their optimum temperature for growth is not exceeded in spring. The income of subpolar phytoplankton species is more likely to happen in the Arctic Ocean, where advection processes help in connecting the polar ocean with waters at lower latitudes. Increased temperatures would favor the growth of the newly arriving groups that are already adapted to higher temperatures. In the Southern Ocean, atmospheric and oceanic patterns result in higher isolation (Hunt et al., 2016). This way, Southern Ocean phytoplankton communities might live in a narrower environmental temperature range, but they are less influenced by the influx of species from lower latitudes.

Future predictions describe a longer ice-free period in the Arctic Ocean and in some areas of the Southern Ocean, where ice will melt faster in spring and consequently make phytoplankton blooms begin earlier. According to our results, we expect that warming may lead to more intense but shorter spring bloom peaks in the Arctic Ocean. These alterations in the timing of blooms and ultimately in food quality could disrupt the delicate balance among trophic levels (Søreide et al., 2010) with implications for polar trophodynamics (Boyd et al., 2013).

The thermal threshold defined here for Arctic communities is already exceeded in August at the lowest Arctic latitudes. In future scenarios, this potential threat to phytoplankton will affect increasingly higher latitudes and move closer to the blooming season, extending into July and September. In fact, if we consider the RCP8.5 high-emission scenario described in the 2013 IPCC report, we can expect the T_OPT to be exceeded in spring at 70°N before reaching the year 2100. The literature on species thermal performance suggests that once the T_OPT is exceeded, the growth of organisms is constrained. Instead, if we consider a low-emission scenario, an intense increase in spring SST is not expected to take place during this century.

At present, polar studies are still quantitatively and geographically restricted, likely constraining our comprehension of the effects of global warming in polar oceans. Also, the temperature range of some of the growth-vs.-temperature curves included in this study were not wide enough to properly accommodate the optimum value, and nutrient limitation could constrain the growth results from some experiments. We highlight the importance of an increase in the number of species studied from the polar areas. More investigation is required to understand how predicted changes in environmental factors such as temperature, as addressed here, may influence polar phytoplankton in the decades to come.

AC and SA were responsible for running all experiments, collecting data from the literature, performing data analysis and for writing and editing the manuscript.