Evidence of summer thermal stratification in extreme northern lakes

Arctic lakes are widely accepted to be highly sensitive sentinels of environmental change (Smol et al. 2005; Vincent et al. 2008b). Notwithstanding this recognition, and despite the prevalence of lakes and ponds in many high-latitude landscapes, most of what is understood about thermal stratification regimes in Arctic lakes is based on assumptions and extrapolation rather than empirical data. This is due, at least in part, to the logistical difficulties in accessing many remote High Arctic regions, as well as the short summer season during which almost all sampling takes place. Existing High Arctic limnological data are largely based on point sampling, with most sites visited only once, and profiles or time series are available only in very limited cases. As such, while baseline physicochemical data are available from lakes spanning much of the Canadian Arctic Archipelago (e.g., Antoniades et al. 2003; Keatley et al. 2007; Huser et al. 2022; Smol 2023), more detailed information such as water column profiles or time series data are almost entirely lacking for large swaths of the High Arctic (Klanten et al. 2023).

Much more is known about annual limnological cycles in lower Arctic sites, thanks to well-established research programs such as at Toolik Lake, Alaska (Hobbie and Kling 2014), the Søndre Strømfjord region of SW Greenland (Anderson et al. 1999), and the Cambridge Bay area of Nunavut (Potvin et al. 2022; Blackburn-Desbiens et al. 2023), that have greatly advanced our understanding of processes in high latitude aquatic ecosystems. However, there are pronounced climate gradients within the Arctic, with, for instance, 2500+ km, >14 degrees of latitude, and 8–9 °C difference in mean annual temperatures between the well-studied Alaskan sites and the High Arctic northern margin of Ellesmere Island and Greenland, that imply pronounced differences in numerous climatic forcings with important limnological implications, including ice cover.

Due to extreme cycles of air temperature and solar radiation, lakes in the High Arctic are snow- and ice-covered for all but a few weeks each year (Vincent et al. 2008b). Lake ice has a controlling influence on multiple limnological processes, including exchanges with the atmosphere, wind-induced mixing, and the attenuation of solar radiation and therefore photosynthesis, among others (Kirillin et al. 2012; Griffiths et al. 2017). The average thickness and duration of Arctic lake ice has diminished over the past several decades and is expected to further decrease over the coming century, leading to fundamental changes in aquatic ecosystem function across the Arctic (Brown and Duguay 2010; Prowse et al. 2011; Dauginis and Brown 2021; Woolway et al. 2022). Recent meta-analyses have suggested that warming-induced increases in the length of the open-water period will promote mixing regime shifts in many lakes, although most such studies include few if any High Arctic sites (e.g., no lakes > 60°N in the Western Hemisphere and none in the High Arctic climate zone; Woolway and Merchant 2019). The reported direct and indirect effects of warming in Arctic lakes have included more frequent and longer periods of stratification (Cadieux et al. 2017), shallowing (Wauthy and Rautio 2020) or deepening of thermoclines (Northington et al. 2019), the strengthening of stratification and development of bottom anoxia (Kraemer et al. 2015), and the breakdown of inverse stratification and, counterintuitively, the cooling of lake waters (Bégin et al. 2021). As such, in addition to a general lack of information on current lake thermal stratification regimes in the Arctic, and particularly those across the High Arctic, the implications of future warming for the stratification of these lakes are unclear.

Much of what is thought to occur on annual time scales in High Arctic lakes is derived from the pioneering studies of Meretta and Char lakes on Cornwallis Island, at 74.7°N, from 1969 to 1972 as part of the International Biological Program (Schindler et al. 1974a, 1974b; Kaff and Welch 1974). Other detailed limnological studies of High Arctic lakes have often focused on extreme, end-member lakes, including meromictic (e.g., Ludlam 1996; Van Hove et al. 2006) and perennially ice-covered ecosystems (Vincent et al. 2011; Bégin et al. 2021). Historically, limnological classifications have suggested that lakes >75° “adjusted” latitude (i.e., adjusted to consider both latitude and elevation; see Lewis (1983) for details) should be amictic, while those between 70° and 75° are either cold monomictic or cold polymictic (Hutchinson and Loffler 1956; Lewis 1983). Lewis’ (1983) revised classification placed the poleward limit for dimixis at 70°, while acknowledging that many perennially cold lakes may rise above 4 °C without stratifying. Here, we report on the presence of lakes that showed evidence of summer thermal stratification in 2022 far beyond these limits, at >82°N on northern Ellesmere Island.

Snow began to melt from the slopes adjacent to the three lakes on 29 May, when Clements Markham Inlet was also solidly covered by snow and ice (Fig. 2), and melting continued steadily until the terrain surrounding the lakes was mainly snow-free sometime between 6 and 12 June. The ice on lakes CM2, CM5, and CM6 began to melt in mid-June, with open water appearing around the margins on 19, 19, and 15 June, respectively, and the lakes became completely ice-free on approximately 22, 15, and 11 July, respectively, by which point the sea ice in Clements Markham Inlet had become fractured by melting (Table 1; Fig. 2).

When sampled, the lakes had been ice-free for only ∼8, ∼13, and ∼18 days (Table 1), and there had been a total of 30.4, 73.3, and 98.7 thawing degree days between the ice-free date and the sampling date for lakes CM2, CM5, and CM6, respectively. Despite their high latitudes, the water columns of lakes CM2 and CM5 had well-developed positive thermal stratification, with maximum water temperatures of 8.3 and 11.7 °C and temperature differences of 3.4 and 6.7 °C between their surface and bottom waters, respectively (Fig. 3; Table 1). The third lake, CM6, which was both smaller and shallower, had monotonic profiles of all measured variables, with an isothermal water column at 11.6 °C (Fig. 3, Table 1). These maximum water temperatures were reached even though air temperature exceeded 10 °C on only 3 days in 2022 prior to sampling, only one of which exceeded the maximum water temperatures of lakes CM5 and CM6. The two lakes with thermal stratification also had chemoclines with higher specific conductivities in their hypolimnia, with bottom oxygen depletion that was moderate in CM5, very slight in CM2, and absent in CM6 (Fig. 3).

Schmidt stability was 68.6 J·m⁻² for CM2 (on 30 July) and 279.8 J·m⁻² and 208.8 J·m⁻² for CM5 (27 and 29 July, respectively); accordingly, the density gradient was stronger in the smaller, warmer CM5 (Fig. 3). Although very few data from Arctic sites are available for comparison, these stability values are within (CM2) or exceed (CM5) ranges calculated from Alaska and SW Greenland lakes (MacIntyre et al. 2009; Saros et al. 2016), suggesting at least comparable strength of stratification, despite the more extreme climate and ice cover regime of northern Ellesmere Island. Density gradients were driven by a combination of changes in temperature and solutes through the water column; the combination of lower temperature and higher TDS (as measured by specific conductivity) below the thermoclines of lakes CM2 and CM5 augmented the top–bottom density differences and would have helped to stabilize stratification. The relatively dilute surface waters likely result from low-conductivity inputs of snow and ice melt, and the resulting ionic gradient might have prevented complete spring mixing and homogenization of the water column (Ladwig et al. 2023). The short time between ice-out and summer stratification on the dates of sampling suggests that the stratification likely developed under ice as a result of strong solar heating through the ice rather than due to sensible heat transfer from the overlying air masses (Kirillin et al. 2021).

The latitude of the Clements Markham Inlet lakes placed them well beyond the widely accepted northern boundary for summer thermal stratification (i.e., ∼70°N), and in a zone where all lakes would be expected to be amictic or cold monomictic/polymictic (Lewis 1983). However, very few water column temperature profiles exist to ground truth such assumptions. Of these, the majority are from Svalbard, where well-studied Kongressvatnet shows evidence of summer stratification, while other lakes such as Linnévatnet and Revvatnet retain isothermal water columns during the ice-free period (Holm et al. 2012; Luoto et al. 2019; Tuttle et al. 2022). However, despite Svalbard's high latitude (∼77 to ∼80°N), it has a far milder climate than that of our study region, with mean annual air temperatures ∼9 °C higher than those of northern Ellesmere Island (Hanssen-Bauer et al. 2019). On Store Koldewey, at 76.1°N in northeast Greenland a single lake was shown to have a weak positive thermal gradient, with a 2.8 °C temperature change over a 70 m water column (Cremer et al. 2005), that again was in a region with a mean annual air temperature ∼6 °C higher than northern Ellesmere Island (Cappelen 2019). The authors suggested that the weak stratification was an ephemeral feature and attributed it to a recent period of calm, sunny weather (Cremer et al. 2005). Six other nearby lakes in the same study were found to be unstratified, with maximum top–bottom temperature differences of 0.5 °C (Cremer et al. 2005).

No lakes at similarly high Antarctic latitudes have shown any evidence of thermal stratification due to summer temperatures; however, due to the climatic severity and glaciological context at extreme southern latitudes most such lakes are either perennially ice covered or subglacial (Spigel and Priscu 1998). A notable exception is warm monomictic Deep Lake in the Vestfold Hills (68.6°S, 78.2°E). This polar desert lake is so hypersaline that it does not ice over in winter but mixes for several months and then thermally stratifies in summer (Ferris and Burton 1988). For lakes and melt ponds >70°S, thermal stratification may occur throughout all months of the year, but this is due to strong salinity gradients whose effects on density are far greater than those of temperature, and which prevent the mixing of water masses of different temperatures (Wait et al. 2006; Healy et al. 2006). For example, the waters of Lake Vanda in the McMurdo Dry Valleys (77.6°S, 161.7°E) rose from 4.5 °C in the dilute freshwater immediately under the ice to 23.5 °C at 68 m depth where the conductivity was three times that of seawater (Vincent et al. 1981).

Most alpine lakes already stratify for at least part of the summer and their stratification regimes may be influenced by different factors, including air temperature, inputs of cold glacial meltwater, or high winds (Peter and Sommaruga 2017; Dokulil 2022). Lake- and catchment-specific characteristics may therefore produce different trajectories in response to warming climates, often including increases in the strength, duration, and depth of stratification, and in some cases shifts from dimixis to warm monomixis due to the loss of winter ice cover (Preston et al. 2016; Råman Vinnå et al. 2021; Dokulil 2022). Paleolimnological studies of tropical Andean lakes (at 3140–3920 m asl) showed mid to late 20th century changes in diatom assemblages that were interpreted as ecological reorganizations in response to shifts from cold polymixis to greatly strengthened summer thermal gradients (Michelutti et al. 2015).

Two separate profiles were taken from lake CM5, on 27 and 29 July. Between the two dates, the position of the thermocline deepened from 2.8 to 3.5 m depth, while the temperature of the epilimnion declined from 11.7 to 10.5 °C, and Schmidt stability declined from 279.8 to 208.8 J·m⁻², indicating a weakening of the stratification due to the declining thermal gradient. Close linkages to temperature were suggested by low air temperatures between the 2 days of water column profiling (i.e., 27–29 July, average daily temperature (T): 1.6 °C) with calm winds; the first day of sampling was also preceded by several days of low temperatures (24–26 July average daily T: 2.5 °C). Similar changes in temperature and thermocline depth, which resulted from convection driven by heat loss at the lake surface during cool periods, were observed in Toolik Lake, Alaska (MacIntyre et al. 2009), which is consistent with the occurrence of such changes during a cold spell at Clements Markham Inlet. There are insufficient data to assess the persistence of stratification in the Clements Markham Inlet lakes beyond the sampling dates, and further profiles at regular intervals are needed to clarify the stability of their summer thermal regimes.

The absence of stratified High Arctic lakes reported in the literature to date may result from one of two factors: that poor spatial and temporal coverage of studies including water column profiles precluded their detection, or that they are a recent phenomenon due to pronounced high-latitude climate warming during the past several decades. While no summer stratification was observed in Char Lake between 1969 and 1972 (Schindler et al. 1974b), it has been suggested to occur more recently (Vincent et al. 2013). For instance, in 1970 and 1971, maximum open water temperatures in Char Lake were 3.0 and 4.5 °C, respectively, in isothermal water columns (and the lake did not lose its ice cover in summer 1972), while in late July 2011 the maximum epilimnion temperature was 10.2 °C and the lake was stratified to a depth of ∼14 m (Schindler et al. 1974b; Drevnick et al. 2013; Hudelson & Drevnick (personal communication 2023)). In Kongressvatnet, Svalbard, mean summer epilimnion temperatures increased by 2 °C between 1962 and 2010 along with warming air temperatures, reaching a maximum of ∼7 °C during the warm summer of 2007; these increases were accompanied by a strengthening of summer stratification (Holm et al. 2012). Immerk Lake, on Devon Island, was shown to be isothermal in summer 1962 (Michaud and Apollonio 2022), but no recent data are available to assess whether it may now stratify in summer. Similarly, Upper Dumbell Lake, ∼90 km to the ESE of our study region, reached a maximum surface temperature of 5 °C during July/August 1959, with a maximum gradient of 1.9 °C over a 33 m water column, but no more recent data exist as to its current thermal regime (Apollonio and Saros 2014). Ward Hunt Lake, meanwhile, now mixes in years in which it loses its summer ice, with no sign as yet of warm stratification (Bégin et al. 2021). Water temperatures there have, in fact, decreased due to heat loss to the atmosphere after the melting of perennial lake ice (Bégin et al. 2021). Similar cooling has been predicted to occur in solar-heated meromictic Lake A on Ellesmere Island with ongoing ice loss (Vincent et al. 2008a). In Toolik Lake, the relative strength of stratification was observed to be closely related to temperature, with more stable stratification developing in warmer years (MacIntyre et al. 2009). Therefore, while summer stratification is appearing and/or strengthening in some High Arctic lakes, for most there is insufficient data to assess any changes that may have occurred over time.

The northernmost land on Earth is only ∼110 km closer to the North Pole than the Clements Markham Inlet lakes, suggesting that there is essentially no latitudinal limit to summer thermal stratification, at least in the Northern Hemisphere. Given the importance of stratification for multiple physical and biotic limnological processes (Saros et al. 2016; Klanten et al. 2023), and the apparent inadequacy at high latitudes of existing classification schemes, there is a pressing need for more data from High Arctic lakes to better understand spatial patterns of summer stratification and how thermal structure may be responding to rapid regional warming.

We thank Parks Canada for their continued support of our research (this study was carried out under Parks Canada Research and Collection Permit QUT- 2022–42541), and to the Polar Continental Shelf Program for invaluable logistic support. We are grateful to K. Hudelson and P. Drevnick for information about Char Lake.