Introduction

Many lakes in the Arctic are dilute, low productivity systems (Markager et al. 1999; Anderson et al. 2001). Primary production in these lakes is constrained in part by low nutrient concentrations (Miller et al. 1986). While lakes in this region were previously considered limited by phosphorus (P) alone (Schindler 1974; Rigler 1978; Gregory -Eaves et al. 2000), there is now evidence that nutrient limitation patterns vary spatially across Arctic landscapes. In addition to P-limitation, nitrogen (N) limitation as well as N and P co-limitation have been found in lakes across many regions of the Arctic (Levine and Whalen 2001; Ogbebo et al. 2009; Symons et al. 2012; Hogan et al. 2014). Overall, these patterns suggest variation in nutrient limitation among lakes within the same region as well as across the Arctic.

Rapid environmental change occurring in high-latitude regions may alter nutrient availability in lakes. Increases in warming are causing thawing of permafrost, which is rapidly releasing N and P into Arctic surface waters (Frey and McClelland 2009; Reyes and Lougheed 2015). A range of other processes can also influence nutrient availability, including additional catchment processes (e.g., altered hydrology, enhanced weathering) and enhanced atmospheric nitrogen deposition (Holtgrieve et al. 2011; Fritz and Anderson 2013; Reyes and Lougheed 2015). These may lead to changes in the supply of nutrients, which can ultimately alter the balance of nutrients available for phytoplankton communities in lakes.

In addition to nutrient limitation, light is a factor that limits algal production in Arctic lakes. Light availability is also important for determining algal community structure (Reynolds 1984), with fluctuating light availability affecting growth rates and interspecific competition (Litchman 1998). The changing climate in the Arctic can alter light availability in lakes in many ways. Increasing air temperatures have caused changes in the length of the ice-free season (e.g., earlier ice off) for many Arctic lakes (Surdu et al. 2014, 2016). A longer ice-free season can result in longer periods of higher light availability for photosynthesis. Climate can also alter water transparency, particularly via its effects on dissolved organic carbon (DOC) concentrations in lake ecosystems (Williamson et al. 1999; Weyhenmeyer and Karlsson 2009). DOC is a major regulator of water transparency (Williamson et al. 1999), affecting primary production in oligotrophic lakes (Karlsson et al. 2009). When it is present in high concentrations, it plays an important role in limiting whole-lake primary production (Ask et al. 2009; Cole 2009). There have been variable changes in DOC concentrations in Arctic surface waters over recent decades (Striegl et al. 2005; Anderson and Stedmon 2007; Saros et al. 2015), and this has likely altered lake water transparency, which affects the thermal structure of small Arctic lakes (Saros et al. 2016). Such changes in lake mixing and water transparency have important implications for light availability to phytoplankton.

Given the highly variable nutrient and light conditions in Arctic lakes, interactive effects between these resources are likely to shape how phytoplankton in these lakes respond to environmental changes. Varying effects of light and nutrient interaction heavily influence phytoplankton communities, since they have short generation times and respond quickly to environmental change. Experimental studies have demonstrated that under nutrient-rich conditions and low light availability, phytoplankton will compete for light, with species that are superior competitors at low light intensity succeeding (Litchman 2003; Kardinaal et al. 2007). In contrast, in poor nutrient conditions and higher light penetration, phytoplankton will compete for nutrients (Passarge et al. 2006). Nutrient uptake is also often dependent on light, with some nutrients requiring more energy to assimilate than others (Syrett 1981). This dependence often weakens with increasing nutrient limitation. Phytoplankton can synthesize more chlorophyll a to capture more light or adjust photosynthetic machinery and match the prevailing light environment (Falkowski and Raven 2007; Lepetit et al. 2012), but this requires more resources. Phytoplankton can also tune their photosynthetic machinery to maintain relatively high efficiency of photochemistry in photosystem II even when deficient in phosphorus or nitrogen (Silsbe et al. 2015). Clearly, interactions between nutrient and light availability play a key role in determining phytoplankton ecology in lakes.

Broadly, one of the phytoplankton groups most strongly affected by interactive environmental conditions is Cyclotella sensu lato taxa, a group of diatoms. Changing environmental conditions in the Arctic have caused widespread changes in Cyclotella species since 1850 across Arctic lakes (Smol et al. 2005; Perren et al. 2009, 2012). These changes appear to be driven by climate change, but we are only beginning to understand the mechanisms underlying these changes. Several studies have reported that these species distribute along vertical habitat gradients of light and nutrient availability (Winder et al. 2008; Saros et al. 2012), and as a result, the vertical thermal structure of a lake is important as it controls these resources. Changes in lake thermal structure in small lakes (<5 km2) are linked to climate via its effect on mean lake temperatures as well as DOC loading (Fee et al. 1996; Kraemer et al. 2015). DOC strongly affects water transparency, which in turn determines the depth of the thermocline in small lakes (Snucins and Gunn 2000).

Watershed processes affect DOC loading as well as nutrient inputs, and can be highly variable across landscapes (Canham et al. 2004; Seekell et al. 2014). Paleolimnological studies from lakes of southwestern Greenland have reported low coherence across lakes in the timing of response of Cyclotella assemblages to similar climatic forcing (Perren et al. 2009; Law et al. 2015). These studies have attributed variation in catchment processes that influence delivery of nutrients into lakes as the driver of the varied response of Cyclotella assemblages during the mid-Holocene thermal maximum. Experiments in these Arctic lakes of Greenland demonstrated important effects of nutrients (none added or N+P) and light, and sometimes interactive effects between them, on several Cyclotella sensu lato taxa (Saros et al. 2014; Malik and Saros 2016), which may explain variation in species distribution patterns. These experiments did not separate the effects of different nutrients or nutrient limitation status. Given the variable nutrient limitation patterns across Arctic lakes, quantifying the effects of differing nutrient limitation patterns and their potential alterations to light responses of different taxa will help in clarifying the ecology of this group of phytoplankton that is common across Arctic lakes.

Here, we investigated whether the nutrient limitation status of a lake alters the responses of three common Cyclotella sensu lato taxa to light. To assess this, we collected source water with the natural phytoplankton assemblages from two lakes in southwest Greenland with different nutrient limitation status. The responses of these three taxa to differing light levels (low, moderate, or high) were tested under different nutrient limitation conditions. The distributions of these three taxa in relation to nutrient limitation and light availability were also assessed across 20 lakes in the area.

Methods

Taxonomy

The recent reclassification of Cyclotella sensu lato taxo had moved some species into the genera Puncticulata and Handmannia. The character that joins the lineage Lindavia is the presence of one or more rimoportulae on the valve face. We followed the latest classification of Cyclotella into the genus Lindavia (Nakov et al. 2015); hence, we use this genus where appropriate. We also follow the transfer of stelligeroid taxa into the genus Discostella (Houk and Klee 2004). We did not distinguish between Discostella stelligera and Discostella pseudostelligera but rather combined them together into one taxon, referring to them as D. stelligera.

Site description

Lakes near Kangerlussuaq, southwest Greenland (67°N, 55°W), were selected for this study (Fig. 1). The lakes around Kangerlussuaq vary in nutrient limitation status (Brutemark et al. 2006; Whiteford et al. 2016) and have abundant Cyclotella sensu lato taxa (Perren et al. 2009). These lakes thermally stratify shortly after ice off, which typically occurs sometime in June; stratification persists until mid-August. Thermal stratification is strongly affected by light attenuation, which in turn is controlled by DOC concentration and quality metrics as well as chlorophyll (Saros et al. 2016). In late June-early July of 2013, 21 lakes were surveyed to assess the distributions of these taxa with respect to nutrient limitation and light availability (Fig. 1).

Fig. 1
figure 1

Map of the study locations near Kangerlussuaq, southwestern Greenland. Lakes studied in the survey are indicated by a different light levels and nutrient limitation. Circle (low light), triangle (moderate light), star (high light). Symbols in white represent P-limitation and in black represents N+P co-limitation. Lakes SS85 and SS67 are experimental lakes

Full size image

We used the ratio of dissolved inorganic nitrogen (DIN) to total phosphorus (TP) to identify nutrient limitation status, with DIN:TP>3.4 indicating P-limitation, ratios <1.5 indicating N limitation, and values from 1.5 to 3.4 suggesting co-limitation. This metric performs better in oligotrophic lakes than TN:TP (Bergström 2010). Out of these 21 lakes, 14 had DIN:TP indicative of co-limitation, six of P-limitation, and one of N limitation. Two of these 21 lakes were selected for experiments because of their different nutrient limitation status and presence of similar Cyclotella taxa. Lake SS67 had a DIN:TP of 3, suggesting N+P co-limitation, while Lake SS85 had a DIN:TP of six, indicating P-limitation. Both lakes had Lindavia bodanica (Grunow) Håkansson, Lindavia radiosa (Lemmermann) Håkansson and D. stelligera (Cleve and Grunow) Houk and Klee. While we found one lake with DIN:TP indicating N limitation, this lake had very low Cyclotella sensu lato populations, and only one species in common with the other lakes, so we chose to exclude an N-limited lake for those reasons. Both SS67 and SS85 had similar light conditions, with 1% depths of photosynthetically active radiation (PAR) of 8.7 (SS85) and 8.8 (SS67) m in June 2013.

Experiments

For each lake, experiments were established by collecting water samples from different depths of the photic zone using a van Dorn sampling bottle. Source water was filtered through 100-μm Nitex mesh to remove zooplankton grazers, and combined into a single acid-washed 5-L container. Three 50-mL sub-samples of this water were preserved with Lugol’s iodine to determine intial cell densities of diatoms. Initial nutrient concentrations in source water were also determined. TP, nitrate (NO3 −), ammonium (NH4 +) and soluble reactive phosphorus (SRP) were quantified using the persulfate digestion method, cadmium reduction method, phenate method, and ascorbic acid method (APHA 2000), respectively, on a Lachat QuickChem 8500 flow injection analyzer. Limits of quantification were 2 µg L−1 for all nutrients. Concentrations of NO3 − and NH4 + were summed to determine DIN. All nutrient analyses were conducted within four weeks of sample collection, with samples kept refrigerated in the dark until analysis.

Experiments were conducted in a growth chamber in the Kangerlussuaq International Science Station using a 3 × 2 factorial design, in which light (100, 60, or 20% of ambient) and nutrients (limiting nutrient(s) or no addition) were manipulated. Light levels were manipulated by inserting the flask into a window screen mesh pocket (resulting in 20% of ambient PAR), inserting into a mesh bag (resulting in 60% of ambient PAR), or leaving uncovered (resulting in 100% of ambient PAR). These percent transmissions were determined using a spectrophotometer (Varian Cary UV–VIS). We used a BIC radiometer (Biospherical Instruments Inc., San Diego, CA) to measure PAR intensity in the growth chamber. The full light intensity (100%) was 272 μEinsteins m−2 s−1 (μE m−2 s−1), 60% was 163 μE m−2 s−1, and 20% was 54 μE m−2 s−1. In the nutrient addition treatments, nutrients were added in the form of NaNO3 for nitrogen (N = 112 μg L−1) and NaH2PO4 for phosphorus (P = 31 μg L−1). Source water was transferred to 75-mL non-treated culture flasks, with each treatment established in triplicate. Flasks were incubated in the growth chamber for 7 days at a temperature of 15°C on a 14:10 h light:dark cycle. This temperature is representative of mid-summer epilimnetic water temperatures (Saros et al. 2016). After the incubation, all flasks were preserved with Lugol’s iodine.

Phytoplankton cells were enumerated by settling 20-ml sub-samples from each flask in Utermöhl-style chambers for 12 h, and examined with a Nikon TS-100 inverted microscope with ×400 magnification. Empty frustules were present in some treatments and these were not counted; this resulted in negative growth rates in some cases. Counts from each flask were converted to cell densities (cells mL−1) and averaged across the three replicates for each treatment. Three transects were counted per slide for D. stelligera, while the whole slide was counted for L. radiosa and L. bodanica because of their larger cell sizes.

To assess the distribution of species cell densities and growth rates, normal quantile–quantile plots were first examined for univariate non-normality and heterogeneity of variance. Univariate normality was tested by Shapiro-Wilk test and Bartlett Test was used to test homogeneity of variances. Growth rates and cell densities of all three species were log transformed for the N+P co-limited lake to achieve normal distributions. To assess the effects of light and lake nutrient limitation status on each species, two-way analysis of variance (ANOVA) was conducted on growth rate across light levels in only the control treatments (i.e., no nutrients added and therefore limited as indicated by the ambient DIN:TP). We used growth rates for this comparison to normalize for differences in absolute cell densities in the two different lakes. To assess the effects of light and nutrient enrichment on cell densities, two-way analysis of variance (ANOVA) was conducted on the three light by two nutrient (control or added) treatments. All tests were run using R (Version 0.98.501).

Distribution patterns

Distributions of the three Cyclotella sensu lato taxa were assessed across nutrient limitation patterns and light conditions in 20 lakes in the area (the one N-limited lake was excluded from this analysis because of the lack of experimental counterpart). Water was collected with a van Dorn horizontal sampling bottle. Epilimnetic samples were used to determine DIN:TP. DIN and TP were measured as described above. Water from the epi-, meta-, and hypolimnia of each lake were also collected and preserved with Lugol’s iodine for phytoplankton enumeration. These were quantified as described above.

To quantify the light conditions in each lake, water column profiles of PAR were measured with a BIC submersible profiling radiometer coupled with a deck radiometer (Biospherical Instruments Inc., San Diego, California). PAR profiles were used to calculate the depth of 1% PAR attenuation for each lake. We also determined the average depth of each lake from bathymetric maps created as described in Saros et al. (2016). We compared relative light environments across lakes using a ratio of the 1% PAR depth to average depth for each lake. Lakes with ratios of about one or less were placed in the low light category; while low, these lakes all had 1%PAR depths that were about the same as the average lake depth. The calculated ratios ranged from 0.9 to 4.2; we established three light categories across that gradient. Lakes with ratios from 1.2 to 2.0 were categorized as moderate light, while those with ratios >2.0 were considered high light. The high light lakes would have roughly 10% PAR remaining at the average lake depth, while the moderate category would be between 1 and 10% PAR at the average depth. We note that all lakes had sufficient PAR (1% of surface irradiance) for phytoplankton growth at the average depth; lakes in the high light category had 10% PAR or higher at the average depth, suggesting that ecologically relevant light gradients are captured across the lake categories.

Based on DIN:TP and 1%PAR:average depth, each lake had a nutrient limitation (P or co-limitation) and a light (low, moderate, high) status, resulting in six possible categories when these are combined. The vertical distributions of taxa within lakes added a second dimension to light relationships. For each lake, the abundances of each Cyclotella sensu lato taxon were ranked from 1 to 3 across the three lake strata (epi, meta-, and hypolimnia). The layer with the highest abundance received a rank of three, lowest abundance a rank of one. If two layers had the same highest abundance, both received a three and the lower abundance layer received a one (i.e., there was no rank of two in this type of case). For each species, the ranks for each stratum were then averaged across lakes in the same categories. For example, the ranks of L. bodanica in the epilimnia of all low light, N&P co-limited lakes were averaged. Because sample sizes in the different categories varied widely and sometimes only included one lake, statistical analyses were not conducted on these data.

Results

Experiments

Source water from Lake SS85, the P-limited lake, contained 4.2 cells mL−1 of L. bodanica, 68 cells mL−1 of L. radiosa and 47 cells mL−1 of D. stelligera. The source water from SS67, the N+P co-limited lake, contained lower cell densities overall, with 0.1 cells mL−1 of L. bodanica, 6 cells mL−1 of L. radiosa and 23 cells mL−1 of D. stelligera. Initial nutrient concentrations in Lake SS85 were 18 μg L−1 DIN and 3 μg L−1 TP (DIN:TP of 6), while in Lake SS67, DIN was 6 μg L−1 and TP was 2 μg L−1 (DIN:TP of 3).

Light affected Cyclotella growth rates differently depending on nutrient limitation status, with interactive effects of light and limiting nutrient of lakes (p < 0.001) resulting for all species (Table 1). In the P-limited lake, L. radiosa had positive growth rates under all light conditions (Fig. 2b), whereas L. bodanica and D. stelligera had positive growth rates only under low or low and medium light conditions, respectively (Fig. 2a and c). The highest growth rates under P-limitation were observed in medium light for L. radiosa and D. stelligera and in low light for L. bodanica.

Table 1 Results of two-way analysis of variance (ANOVA) for nutrient limitation status of lake and light effects on the growth rates of L. bodanica, L. radiosa, and D. stelligera in experiments
Full size table
Fig. 2
figure 2

Growth rates of different Cyclotella taxa under P-limited or N+P co-limited conditions with manipulated light levels (low (L), medium (M), or high (H)) indicated. a L. bodanica; b L. radiosa; c D. stelligera. Bars represents standard errors

Full size image

In the N+P co-limited lake, responses to light differed for all species. Lindavia bodanica growth rates were positive under all light conditions (Fig. 2a), whereas L. radiosa and D. stelligera had positive growth rates only under high or low light, respectively (Fig. 2b and c). The highest growth rates under N+P co-limitation were observed in high light for L. bodanica and L. radiosa and in low light for D. stelligera.

In both lakes, additions of the limiting nutrient(s) increased cell densities of Cyclotella taxa for most treatments. Final cell densities were higher when nutrient limitation was relaxed in the P-limited versus N+P co-limited assemblages; in general, they were more variable in all of the populations from the N+P co-limited lake (Fig. 3). In some cases, the addition of P changed the response to light compared to P-limited conditions; no change in response to light was apparent with nutrient enrichment of the N+P co-limited lake treatments.

Fig. 3
figure 3

Cell densities of different Cyclotella taxa from P-limited or N+P co-limited lake with manipulated light levels (low (L), medium (M), and high (H)) indicated. Results are shown for treatments without nutrient additions (Control) and with the limiting nutrient(s) added. a and d Lindavia bodanica; b and e Lindavia radiosa; c and f Discostella stelligera. The cells mL−1 for each species are depicted in each case, with bars representing standard error

Full size image

In populations from the P-limited lake, interactive effects of nutrients and light affected the cell densities of L. bodanica (p = 0.009; Table 2). L. bodanica cell densities were highest under low light for both control and P-addition treatments (Fig. 3a). The addition of P increased their cell densities under high light (approximately doubled) but not in low light treatments, resulting in an interactive effect. L. bodanica cell densities were also affected by independent effects of light (p = 0.001) but not by nutrients (p = 0.85; Table 2). Lindavia radiosa cell densities were highest under medium light in both control and P-addition treatments (Fig. 3b). Addition of P increased cell densities by 1.3 to 1.6 times but did not change the response to light. L. radiosa cell densities were affected by independent effects of nutrients (p = 0.001) and light (p = 0.01) only; there were no interactive effects between nutrients and light (p = 0.85; Table 2). D. stelligera cell densities were affected by interactive effects between nutrients and light (p < 0.001). The addition of P increased cell densities about two to six times compared to controls. Furthermore, P-addition changed their response to light, with highest cell densities under medium light in the control treatment compared to low light with P enrichment (Fig. 3c). D. stelligera cell densities were also affected by independent effects of light (p < 0.001) and nutrients (p < 0.001; Table 2).

Table 2 Results of two-way analysis of variance (ANOVA) for light and nutrient enrichment effects on the cell densities of L. bodanica, L. radiosa, and D. stelligera in experiments
Full size table

In populations from the N+P co-limited lake, there were no interactive effects of nutrients and light for any of Cyclotella species (Table 2). L. bodanica cell densities increased when nutrients were added, but this did not alter their response to light (Fig. 3d). Cell densities of L. bodanica were affected by independent effects of nutrients (p < 0.001) and light (p = 0.02; Table 2), with highest cell densities in the high light treatment. Similarly, L. radiosa cell densities increased about three times when nutrients were added (Fig. 3e). Cell densities of L. radiosa were also affected by independent effects of nutrients and light (p < 0.001 in both cases; Table 2), with highest cell densities in the high light treatment. In contrast, D. stelligera cell densities were affected by nutrients only (p = 0.001; Table 2). Addition of nutrients almost doubled their cell densities, but it did not change their response to light (Fig. 3f).

Distribution patterns

Of the 20 lakes surveyed, 14 had DIN:TP indicative of N+P co-limitation and six of P-limitation (Table 3). The 1% PAR ranged from 5 to 19 m across the lakes. Three lakes were in the low light category, while ten were in moderate and seven in high. There was at least one lake in each of the six defined lake categories (Table 3). For comparison to the light intensities of the growth chamber, the average across all lakes of average PAR intensity of the mixed layer was 568 ± 23 μE m−2 s−1; PAR intensities at the transition between the epi- and metalimnia ranged from 61 to 293 μE m−2 s−1.

Table 3 Light, depth (average and maximum, Zmax), nutrient limitation, and species data from the twenty lakes surveyed
Full size table

L. bodanica was present in 13 of the lakes, L. radiosa in 10 lakes, and D. stelligera in 18 lakes. The distributions of each species varied with nutrient limitation status and light conditions. L. bodanica was generally found in lower light conditions in P-limited lakes (although the moderate light category was not consistent with this observation), and higher light conditions in N+P co-limited lakes (Fig. 4). L. radiosa was absent from the low light lakes in both nutrient limitation categories. Distribution patterns of this species across lake strata varied and did not show a strong trend other than the absence in low light lakes. D. stelligera was generally found in higher light conditions in P-limited lakes, and lower light conditions in N+P co-limited lakes.

Fig. 4
figure 4

Ranked abundances of each taxon in lake strata across the twenty survey lakes, with lakes categorized by nutrient limitation patterns and relative light availability as explained in the text. The number of lakes (n) in each category is indicated. Higher rank values indicate greater abundance of that taxon in that layer

Full size image

Discussion

Our results demonstrate that variation in the nutrient limitation patterns of lakes has important implications for the response of phytoplankton to changing environmental conditions. Specifically, we found that the nutrient limitation status of a lake altered the responses of three Cyclotella sensu lato taxa to light. This is consistent with previous research that has demonstrated complex, interactive effects of nutrients, and physical variables (e.g., light, temperature) on cell densities of L. bodanica and D. stelligera (Saros et al. 2012, 2014; Malik and Saros 2016). Our results advance understanding further by clarifying how specific types of nutrient limitation (P or both N+P) affect the growth rates and cell densities of three taxa under different light conditions, and reveal how variation in nutrient limitation patterns across Arctic lakes may affect species distribution patterns. With differing limitation patterns and ongoing alterations to Arctic lake nutrient dynamics, our results help to clarify seemingly disparate patterns in Cyclotella sensu lato taxa distributions.

Under P-limitation, two species had their highest growth rates at moderate light and one species at low light. In contrast, under N+P co-limitation, two species had their highest growth rates at high light and one species at low light. There are many physiological mechanisms by which different types of nutrient limitation may affect the response of diatoms to light. The P-limited lake had three times more DIN than the co-limited lake. With a substantial proportion of the light-harvesting apparatus containing nitrogen (Larkum and Barrett 1983), the greater availability of nitrogen in a P-limited lake would enable maintenance of sufficient light harvesting under low-to-moderate light levels. In contrast, phytoplankton under high light conditions will have lower growth rates because P-limitation leads to a decrease in light saturated growth rates and increases susceptibility to photoinhibition (Litchman et al. 2003). Limitations by both N and P can cause a decrease in chlorophyll concentration per cell (Daley and Brown 1973; Porra and Grimme 1974; Litchman et al. 2003), raising light requirements. However, small cells have an advantage under light limiting conditions because light absorption per unit of chlorophyll is higher and internal shading by pigment is lessened (Finkel 2001; Finkel et al. 2004). To maintain photosynthesis in high light conditions, phytoplankton must maintain photoinactivation of PSII with repair. Cell size causes a trade off between PSII photoinactivation and susceptibility. In our experiments, D. stelligera is the smallest diatom; L. bodanica is the largest. In general, small centric diatoms are more susceptible to PSII photoinactivation and depend upon PSII repair in response to high light. They can repair cells faster when photoinactivated whereas large centric diatoms are less susceptible to photoinactivation and therefore incur lower costs to endure short-term exposures to high light (Key et al. 2010).

Nutrient limitation status had a strong effect on the response of L. bodanica to light. The growth rate of this species was highest at low light under P-limitation, and at high light under N&P co-limitation. This suggests that the vertical distribution of this species in a lake will depend on nutrient limitation status, with higher abundances predicted at deeper depths (lower light) in P-limited lakes and at shallower depths (higher light) in N&P co-limited lakes. This is generally consistent with the patterns we observed across these Arctic lakes, with L. bodanica being more abundant in deeper waters of P-limited lakes (although not in the moderate light lake surveyed), whereas in N&P co-limited lakes, it was more abundant in upper layers. This effect weakened slightly in the more transparent lakes, where deeper layers would also be well illuminated. These observations are consistent with distribution patterns at mid-latitudes. L. bodanica had higher cell densities in the hypolimnion of a P-limited alpine lake (Saros et al. 2012), and was most abundant during spring turnover (a period of lower average light intensities) in P-limited Sebago Lake in Maine, USA (Boeff et al. 2016). While we are unaware of results from N&P co-limited lakes, higher abundances of this taxon at shallower depths (high light) have been observed under N-limited conditions in subalpine lakes (Interlandi et al. 1999; Noble et al. 2013). Overall, the relationship of L. bodanica with light appears dependent upon nutrient availability, which will determine its vertical distribution in the water column.

The response of D. stelligera to light was also dependent on nutrient limitation status. The growth rate of this species was highest at moderate light under P-limitation, and at low light under N&P co-limitation. The distribution patterns of this taxon across the twenty lakes generally reflected the experimental results. This species was more abundant in the upper layers of P-limited lakes, and deeper or more evenly distributed across layers in N&P co-limited lakes. These patterns are consistent with those found at mid-latitudes. D. stelligera was more abundant in the epilimnia of P-limited alpine lakes (Saros et al. 2012), and small Cyclotella taxa were more abundant in Lake Tahoe during periods of higher N:P supply ratios and stronger vertical stratification (Winder and Hunter 2008). In Maine (USA), Boeff et al. (2016) found that D. stelligera bloomed in summer in the epilimnion of two P-limited lakes, while it was most abundant during spring turnover (lower light exposure) in Lobster Lake, which was N&P co-limited during spring. As with L. bodanica, the distributions of D. stelligera appear dependent on interactions between nutrient limitation and light conditions.

L. radiosa responded positively to moderate-to-high light under different nutrient limitation treatments, suggesting that light conditions in general, regardless of nutrient limitation status, are important for this species. The growth rate of this species was highest at moderate light under P-limitation, and at high light under N&P co-limitation. The positive effects of moderate-to-high light in both cases suggest that this species should be more abundant in the epi- and metalimnia of lakes. In the twenty survey lakes, we found that this species was absent from lakes in the low light category, while it generally varied in abundance in lake strata of the moderate and high light lakes. Other recent incubation experiments in Arctic lakes confirm these results, indicating that this species had the highest growth rates under high light conditions (Saros et al. 2014; Malik and Saros 2016). In lakes at mid-latitudes, this species was most abundant in the epilimnion of Piburger See, an alpine lake in Austria with high water transparency, and was correlated with the lower range of DIN:TP ratios for this lake, which varied from P to N&P co-limitation (Tolotti et al. 2012). In a eutrophic lake from Northern Ireland and in a German reservoir, L. radiosa disappeared as eutrophication occurred and re-appeared as nutrient enrichment declined (Rippey et al. 1997; Horn et al. 2011), suggesting positive effects of increased water transparency on this species.

The complex effects of nutrient limitation status and light availability on the distributions and abundances of various Cyclotella sensu lato taxa may contribute to the high degree of spatial and temporal variability in Cyclotella patterns in some Arctic lakes. As noted above, paleolimnological studies in southwest Greenland revealed low coherence in the timing of response of Cyclotella assemblages across lakes to similar climatic forcing (Perren et al. 2009; Law et al. 2015), and attributed this to variation in catchment processes that influence delivery of nutrients into lakes. Our results support these conclusions, and further suggest that the role of light and water transparency may also be important. Changes in light exposure and water transparency can occur with variation in the length of the ice-free season as well as in DOC concentrations and quality, all of which are, at least in part, climate-driven alterations to lake ecosystems. Phytoplankton production in southwest Greenland lakes is tightly linked to seasonal changes in nutrients and light availability associated with winter ice cover and open water conditions (Whiteford et al. 2016). With a sub-set of lakes in our study also investigated by Perren et al. (2009) who conducted a “top–bottom” assessment of sedimentary diatom assemblages, we examined whether our results provide additional context for the patterns in sediment records. We compared wherever possible; not all lake categories and taxa are covered, and we further note that sedimentary diatoms are expressed as percent relative abundance, so changes in one taxon affect relative abundances of others. In Perren et al. (2009), the core tops represented recent (~late 1990s) assemblages, while bottoms were estimated to pre-date 1850, hence the tops captured assemblages from a warmer period, and showed diverse responses across lakes of the same taxa. In the two high light, P-limited lakes in common between the two studies (SS56 and SS85 (called SS1478 in Perren et al. 2009), D. stelligera increased by 10–30% from the bottom slice to the top, whereas in the one moderate light, P-limited lake (SS2/SS1616), there was no change in its relative abundance. This is consistent with a higher light association under P-limitation for this taxon. The response in co-limited lakes was always positive: an increase of 30% in the low light lake (SS32), 20% (SS57) or 40% (SS68) increase in the moderate light lakes, and ~2% (SS66), 10% (SS67) or 40% (SS1341) increase in the high light lakes. This variability in the effects of light under N&P co-limitation was also apparent in our experiments. L. radiosa increased in all three high light lakes in common between the two studies (SS56, SS67, SS85/SS1478), whereas in the moderate light lakes, it only increased in one (SS57) and declined in the other two (SS2/SS1616, SS68).

Our results help to clarify some of the complex distribution patterns of planktonic diatom taxa found in lakes of many areas around the Arctic as well as at lower latitudes. They also underscore the conclusion that climate-driven changes in phytoplankton distribution patterns in Arctic lakes are complex, owing at least in part to factors influencing lake nutrient conditions and water transparency. Given the rapid changes occurring in nutrient and organic material delivery by watersheds in Arctic landscapes (Townsend-Small et al. 2011; Schuur et al. 2015), understanding ecological responses to such changes will be critical to assess future changes to species distribution patterns in oligotrophic Arctic freshwaters.