The littoral zone of polar lakes: inshore–offshore contrasts in an ice-covered High Arctic lake

In ice-covered polar lakes, a narrow ice-free moat opens up in spring or early summer, and then persists at the edge of the lake until complete ice loss or refreezing. In this study, we analyzed the horizontal gradients in Ward Hunt Lake, located in the Canadian High Arctic, and addressed the hypothesis that the transition from its nearshore open-water moat to offshore ice-covered waters is marked by discontinuous shifts in limnological properties. Consistent with this hypothesis, we observed an abrupt increase in below-ice concentrations of chlorophyll a beyond the ice margin, along with a sharp decrease in temperature and light availability and pronounced changes in benthic algal pigments and fatty acids. There were higher concentrations of rotifers and lower concentrations of viruses at the ice-free sampling sites, and contrasts in zooplankton fatty acid profiles that implied a greater importance of benthic phototrophs in their inshore diet. The observed patterns underscore the structuring role of ice cover in polar lakes. These ecosystems do not conform to the traditional definitions of littoral versus pelagic zones but instead may have distinct moat, ice-margin, and ice-covered zones. This zonation is likely to weaken with ongoing climate change.


Introduction
Lakes are sentinels of environmental change and integrate the conditions of their surrounding watersheds (Williamson et al. 2009). The shallow nearshore portion of a lake, the littoral zone, is the interface between water and land, and plays a key role in the whole lake ecosystem and its responses to change (Vander Zanden and Vadeboncoeur 2020). Dissolved and particulate nutrients enter the littoral zone by inflows from the watershed and influence primary production throughout the lake (Jones et al. 1998). In oligotrophic lakes, ecosystem productivity is the result of not only autochthonous processes but also heterotrophic production fuelled by allochthonous organic carbon entering the littoral zone from terrestrial sources (Ask et al. 2009). These inputs of detrital organic matter are a potential food source for zoobenthos (Solomon et al. 2011) and zooplankton (Rautio et al. 2011;Harfmann et al. 2019), and can stimulate bacterial production (Traving et al. 2017). The presence of macrophytes along with associated periphyton supports a productive inshore benthic habitat, especially in oligotrophic lakes (Vadeboncoeur et al. 2008). However, the abundance and composition of the biota in the littoral zone can be seasonally influenced by changing inflow conditions, and by the combined mechanical effects of suspended particles, wave action and ice that alter near-shore sediments and vegetation (Strayer and Findlay 2010), thereby affecting the whole lake ecosystem.
A variety of definitions have been proposed to set the boundary between inshore (littoral) and offshore (pelagic or limnetic) zones of lakes, and these have differed among fields of study such as hydrology, biology and geomorphology. The littoral zone can be broadly defined as the region where land processes and the aquatic ecosystem influence one another (Strayer and Findlay 2010), and more specifically as the area where at least 1% of incoming photosynthetically available radiation (PAR, 400-700 nm) reaches the bottom and allows macrophytes to grow (Peters and Lodge 2009). In High Arctic and Antarctic lakes, however, vascular plants are absent, and hardy mosses and microbial biofilms are the only phototrophs colonizing the sediments of inshore waters (Sand-Jensen et al. 1999;Mohit et al. 2017). These shallow waters freeze to the sediments in winter, and the habitats are subject to the additional stress of scouring by ice movements during melt-out.
Light availability in polar lakes is constrained more by ice cover and the extreme seasonality of solar variation rather than by the depth of the water column, further underscoring the difficulty of applying a temperate latitude definition of 'littoral zone' to high latitude aquatic ecosystems.
Arctic and Antarctic lakes are commonly covered by a thick ice cover that remains for most or even all of the year and that greatly reduces the annual input of light to the water column (Schindler et al. 1974). In early summer, the ice cover begins to melt at its edge, producing a narrow ice-free band of open water along the shore of the lake. This 'moat' region provides the only area of direct exchange with the atmosphere, allowing wind-induced turbulence, exposure to incident solar radiation and full gas exchange with the overlying air, along with input of water from the surrounding watershed (Priscu et al. 1996;Hall et al. 2017). Few studies, however, have examined these inshore features and processes in detail. The early inshore melting is driven by the higher input of solar energy per unit ice volume in these shallow depths, with positive feedback effects caused by the increasing amounts of resultant meltwater that heat up in the shallows. After the initial rapid melt-out towards deeper water, the central pan of floating ice may then rest in place for a long period of time, sometimes anchored on one side of the lake according to wind direction, and with a potentially strong influence on underwater light and photosynthesis (Belzile et al. 2001).
In the present study, we focused on the inshore-offshore gradients in Ward Hunt Lake in the Canadian High Arctic. This far northern lake is covered by thick ice for at least 11 months of the year, and in many years by multi-year ice that persists throughout summer (Paquette et al. 2015).
In all years, even in the past when the lake was covered by 4-m thick perennial ice, a moat of open water has extended around more than half the shoreline, persisting for several weeks each summer (Paquette et al. 2015;Bégin et al. 2020). We sampled along transects from the lake-edge moat to sub-ice waters in the middle of the lake to address the hypothesis that there would be an inshoreoffshore discontinuity in limnological properties that is directly linked to the ice cover. This study is a contribution to the project 'Terrestrial Multidisciplinary distributed Observatories for the Study of Arctic Connections' (T-MOSAiC), which places emphasis on system-level properties, including connectivity, gradients and discontinuities (Vincent et al. 2019).

Study site
Ward Hunt Lake (83°05.226'N; 74°08.721'W; WGS84 map datum), is Canada's northernmost lake and is covered by thick annual or multi-year ice. It is located on Ward Hunt Island, 6 km off the northern coast of Ellesmere Island (Nunavut, Canada; Supplementary Fig. S1), and has an area of 0.37 km 2 and a maximum recorded depth of 9.7 m. The region is influenced by a polar desert climate with a mean annual air temperature of -16.7°C and a July average temperature of 1. 7°C (2006-2018CEN 2020). Ward Hunt Lake is mainly fed by meltwater from snow patches in its 1.82 km 2 watershed, which flows into the lake via surface and subsurface water tracks (Paquette et al. 2020). The ice phenology and moat formation in Ward Hunt Lake were observed with time-lapse images captured by an automated camera on its western shore, at hourly intervals from 10:00 to 14:00 h each day. The camera was in function from 25 June to 31 October 2016, from 18 July to 7 November 2017 and from 25 May 2018 to 21 April 2019. The full dataset of images and the details about the camera installation are archived in the Nordicana D data repository (NEIGE 2020).

Physicochemical variables
In July 2015, four sites were sampled along three replicate transects that covered the inshoreoffshore gradient, extending out from the northwestern shore of the lake (Fig. 1). Samples from the inshore moat site (S1) were collected within 5 m of the shore. Samples from the ice-margin site (S2) were obtained in the open-water area within 2 m of the edge of the ice cover, which was located approximately 20 m from the shore. A sampling of the intermediate under-ice site (S3) was at a distance of about 100 m from the shore and corresponded to under-ice conditions approximately mid-distance between the deepest part of the lake and the shore. Samples at the central under-ice site (S4) were collected beneath the ice at this location of the greatest water column depth. Samples were collected between 14 and 21 July 2015. Oxygen, temperature, pH and conductivity profiles were measured using a DS5X (Hydrolab, Loveland, CO) and a 600QS probe (YSI, Yellow Springs, OH).
Light profiles from 280 to 720 nm were measured in July 2016 with a RAMSES ACC UV/VIS hyperspectral irradiance sensor (TriOS, Germany) from the four sites as shown in Figure   1. The diffuse attenuation coefficient (K d ) was calculated as: Water sampled for pigment analysis was filtered through 25 mm GF/F filters that were stored initially in the field at -20°C and then at -80°C until analysis. Microbial mats were also stored in the field at -20°C and then at -80°C until analysis. Pigments from the filters and the mats were extracted with methanol 95% and measured by high pressure liquid chromatography (HPLC) as in Bonilla et al. (2005). Four successive extractions were necessary to retrieve all the pigments from the microbial mats. Microbial mat samples contained abundant degraded chlorophylls and carotenoids. Unknown carotenoids were quantified using standard conversion factors from their closest relative according to their retention time and spectra. If the carotenoid was not closely related to a known standard, the ß,ß-carotene conversion factor was used, except for myxolglycosides, for which we used the myxoxanthophyll conversion factor. For unknown chlorophylls, the chlorophyll a (Chl a) conversion factor was used.
Additional lake water samples were fixed with a mixture of glutaraldehyde and paraformaldehyde for protist identification as in Lovejoy et al. (1993). Heterotrophic and autotrophic protists were counted and identified in sedimentation chambers (Utermöhl 1958) with an epifluorescence microscope at 400X (Axiovert 10, Zeiss, Germany). Zooplankton samples were collected using a 21 cm diameter net (63 µm mesh), fixed with formaldehyde (final concentration 4%) and then identified and counted by light microscopy at 100X magnification.

Statistical analyses
One way ANOVAs and Tukey HSD tests were performed on limnological variables

Moat dynamics
During our sampling in July 2015, the ice cover of Ward Hunt Lake had a mean thickness of Some of the images during the spring and fall period when the lake was completely ice-covered showed that the smooth moat ice was clear of snow, while snow covered the offshore lake ice (NEIGE 2020).

Physicochemical profiles
The vertical profiles of temperature, dissolved oxygen and conductivity showed pronounced changes along the inshore-offshore transect (Fig. 3). The shallow water column at S1 was fully mixed at 5.5°C, falling to 3.8°C in the surface waters at S2. At the ice-covered sites S3 and S4, a thin layer of cold meltwater below 1°C lay over the warmer water beneath (Fig. 3a), with the highest temperatures around 7.0°C, between 4 and 5 m in the water column. Oxygen concentrations were around 100% air equilibrium at S1, consistent with full exposure of the moat zone to windinduced mixing, while the below-ice sites S3 and S4 showed a highly stratified pattern with midwater column values up to 140% in the depth region 4 to 8 m (Fig. 3b). Specific conductivity values were low and homogenous at S1 (from 0.119 to 0.146 mS cm -1 ), while the offshore sites showed pronounced water column gradients, rising from 0.040 mS cm -1 in the dilute meltwater beneath the ice to 0.309 mS cm -1 at the bottom of the water column (Fig. 3c). The marginal ice site S2 showed vertical gradients in temperature, oxygen and conductivity, but these were much less pronounced than at ice-covered sites S3 and S4.
As expected, underwater light conditions varied greatly across the transect. At the ice-free site S1, 54.1% of the total incoming irradiance reached the bottom at 2 m depth, while at the icecovered site S4, only 26.5% entered the top of the water column immediately beneath the ice, falling to 5.5% at 9.4 m (Fig. 4). The ice cover allowed a slightly greater proportion of ultraviolet radiation (UVR; 33.3% penetration) compared to PAR (27.1%); UVR at 2 m depth in the inshore moat site S1 was 48.0% of incoming UVR, but undetectable at the bottom of the water column at S4. The downward irradiance spectra and the K d at 2.0 m showed a similar pattern at S1 and S4, with the least attenuation between 400 and 600 nm (Fig. 4). The upward irradiance spectra were strikingly different between the two sites. Wavelengths between 600 and 700 nm were reflected by the bottom in the inshore moat site (S1) and upwelling irradiance values at 2 m were higher than the upwelling values at the surface (Fig. 4). In S4, upward spectral irradiance peaked at 573 nm just below the ice (2.0 m) and near the bottom at 9.1 m ( Fig. 4).
Maximal values of the diffuse extinction coefficient (K d ) were below 350 nm and over 600 nm in both the inshore and the central sites (Fig. 4). K d was lower for wavelengths above 600 nm in the moat than in the under-ice zone. The apparent K d of the ice (not corrected for reflection) increased toward longer wavelengths. The total K d between 300 and 720 nm was 0.31 m -1 for the 2 m water column at S1 falling to 0.20 m -1 for the upper 2 m of the water column at S4, indicating 36% less attenuation by water under the ice. The ice cover had an apparent K d of 0.64 m -1 and the water column between 2 and 9.7 m had a K d of 0.20 m -1 . However, non-algal particle and algal particle absorption values were lower inshore and CDOM absorption was almost identical in the two sites ( Fig. 5a), suggesting that scattering affected the observed variations in light attenuation. Absorbance peaks at 439 nm and 674 nm associated with Chl a, and at 625 nm associated with phycocyanin, a characteristic cyanobacteria pigment, were more pronounced in the central site (S4).

Inshore-offshore concentration gradients
Two patterns of change along the inshore-offshore transect were observed in the measured chemical and biological properties. The first pattern consisted of an abrupt drop at S2 in the ice margin site, where Chl a and phytoplankton were less concentrated and total nitrogen more concentrated (Fig. 6), and where DOC, bacteria, and mixotrophic protists tended to be less concentrated. The second pattern was a shift of concentrations between the ice-free moat zone (S1,

Cluster analysis of all variables
According to the cluster analyses ( Fig. 7), there was a clear distinction between the openwater moat (S1 and S2) and offshore ice-covered (S3 and S4) zones. The intermediate and central sites (S3 and S4) generally clustered together, suggesting more homogenous conditions under the ice. The ice-margin site was distinct relative to the others according to the limnological parameters, algal pigments in the water column and rotifer assemblages (Fig. 7).
According to their clustering, the nanoplankton communities observed by microscopy were less separated by sites than other groups. The inshore (S1) and intermediate (S3) sites were more similar to each other, with higher concentrations of Snowella sp., Micractinium sp., and Tabellaria sp. The central (S4) and ice-margin (S3) sites also showed similarities, with higher concentrations of Tetraedron sp. and Synura sp (Supplementary Table S1).
An unusual feature of Ward Hunt Lake was that there was no evidence of crustacean zooplankton in any of our net samples, consistent with a previous mid-summer sampling of the lake (M. Rautio, unpublished observations Table S3). Diadinoxanthin is associated with diatoms, peridinin with dinoflagellates, and myxol-glycosides with cyanobacteria (Roy et al. 2011).
Pigments extracted from the benthic mats at S1 were distinct from benthic mat pigments in the ice-covered zone, with higher concentrations of scytonemin and reduced scytonemin, both signature pigments of UV-exposed cyanobacteria (Bonilla et al. 2005;Roy et al. 2011). The offshore benthic mats additionally contained bacteriochlorophyll c, which is generally associated with green sulfur bacteria that grow under anoxic conditions (Supplementary Table S4).

Fatty acids
The highest total fatty acid (FA) concentrations were measured in the central site (S4) samples of seston and zooplankton (means of 479 and 290 µg/mg dry weight, respectively; Fig.   8a; Supplementary Table S5). Microbial mats contained the lowest concentrations (mean 25 µg/mg dry weight), especially in the inshore (S1) and central (S4) sites. Inshore organisms contained more polyunsaturated fatty acids (PUFA) than in the ice-covered zone (sites S3 and S4; Fig. 8b-f), but the distribution differed statistically only for microbial mats; due to the small sample sizes, no statistical tests were done for the mosses or chironomids. According to the PCoA performed with fatty acid assemblages ( Fig. 9), pelagic (seston) and benthic (microbial mats and mosses) food sources were different from each other but showed only slight differences between the central and inshore moat sites. Seston fatty acid assemblages were characterized by higher concentrations of C22:6n-3 that is a bioindicator for phytoplankton (Kelly and Scheibling 2012). The distribution of the assemblages in the biplot suggests a strong differentiation in zooplankton diet between the icefree moat and under-ice zones (Fig. 8). The zooplankton fraction (>63 µm) from the central site (S4) shared a similar fatty acid profile to the seston, suggesting their main food source was located in the water column. Zooplankton from the inshore moat site (S1) contained higher concentrations of C16:1n-7, a bioindicator for diatoms but also for cyanobacteria (Taipale et al. 2015) that were both common in the microbial mats, suggesting their diet is from diverse benthic phototrophs in these mats. The fatty acid assemblages in Metriocnemus larvae were more heterogenous, but all were closely related to the microbial mats, indicating that these biofilms may provide food resources for these benthic consumers in the central as well as inshore sites.

Moat development
An ice-free peripheral moat is a common feature of Antarctic and Arctic waterbodies (Adams et al. 1989;Miller and Aiken 1996), and is a distinct within-lake environment that can persist for many weeks each summer. The formation of the moat is initially rapid as a result of higher solar fluxes per unit volume in the shallow ice-free zone, the positive feedback effect of the resultant meltwater, the arrival of inflows and additionally, as indicated here, by reduced snow cover and albedo. Our observations of less snow retention over the inshore ice-free zone are consistent with those of Adams et al. (1989), who attributed this to the wind removal of snow from smoother moat ice and the accumulation of snow on rougher, slightly higher, multi-year ice.
Once the moat of open water is formed, its area then remains more stable for a prolonged period (Welch 1973), with the thermal inertia of the thicker offshore ice and deeper water column acting as a brake on the subsequent rate of ice loss. As we observed, however, there can be large variations in moat extent among years. The width of the moat of Ward Hunt Lake as estimated by the automated camera images extended from less than 5% to 25% (in September 2018 and August 2016, respectively) of the total lake area. In 2015, the width was around 20 m, similar to the moat of Lake A on nearby Ellesmere Island as measured in 2007(Tomkins et al. 2009). This moat extent is larger than in the perennially ice-capped lakes of the McMurdo Dry Valleys of Antarctica, where the summer moats account for around 3% of total lake area (Wharton et al. 1986;Priscu et al. 1996). In these south polar lakes, as at Ward Hunt Lake, there can be large variations among years; for example, Lake Fryxell had up to 11% of open water in the 2013-2014 summer, associated with higher air temperatures (Wayt et al. 2017). In the most extreme Antarctic lakes such as Lake Untersee, no moat has been observed and the lakes remain completely sealed throughout the year (Faucher et al. 2020).

Physicochemical variables
Given the high transparency of its waters, the entire expanse of Ward Hunt Lake might be considered as a "littoral zone" according to the definition of Peters and Lodge (2009), based on light penetration to at least 1% irradiance at the bottom. However, the distinct characteristics imposed by the presence of ice results in marked, discontinuous inshore-offshore differences, from the fully exposed moat zone to the shaded ice-covered zone. The horizontal extent of the inshore open-water zone is therefore a more meaningful descriptor of spatial differences in limnological properties of Arctic lakes that are ice-covered for most of the year than traditional definitions of the littoral zone.
The patterns observed in Ward Hunt Lake underscore the importance of ice cover as the main driver of spatial variation, with the ice edge as the key determinant of horizontal structure. The physicochemical profiles under the ice varied little among sites (Fig. 3). The ice cover allowed inverse thermal stratification along with chemical stratification by preventing exposure to the atmosphere and wind-induced mixing. Ice has a low thermal conductivity, limits heat loss in the atmosphere and allows solar heating of the water column to temperatures well above those of the overlying air: Ward Hunt Lake had under-ice water temperatures up to 7°C, yet air temperatures averaged around 1.7°C in July (Bégin et al. 2020). This greenhouse effect has been modelled in other ice-covered lakes of the Arctic (Vincent et al. 2008) and Antarctica (Obryk et al. 2019).
Complete loss of ice cover during warm summers allows mixing of the water column and ventillation of heat to the atmosphere, and water temperatures can then drop to near freezing (Schindler et al. 1974;Doran et al. 1996;Bégin et al. 2020). The vertical structure that was lost in Ward Hunt Lake in 2016 was reinstated in summer 2017 beneath the first-year ice cover, but with lower temperatures (Bégin et al. 2020). The presence of ice cover hence allows a stable and relatively warm environment to develop, which may stimulate primary production (Markager et al. 1999) and microbial food web activity (Wrona et al. 2006). However, these effects may be offset by the influence of ice in reducing light availability, mixing and nutrient supply.
High concentrations of oxygen were recorded between 4 and 9 m, indicating net autotrophy in the water column. At the bottom of the central under-ice site (S4), the oxygen concentration was around equilibrium (Fig. 3), suggesting a balance between gain and loss processes. The bottom waters of Ward Hunt Lake are well oxygenated in July, but reach anoxia at the beginning of winter that persists until spring (Bégin et al. 2020). In contrast, frequent mixing in the ice-free moat zone likely keeps the water and the microbial mats well oxygenated until freeze-up, which translates into a lower proportion of anaerobic bacteria (Mohit et al. 2017). Concentrations of oxygen above saturation (near 200% in Ward Hunt Lake) are well documented in ice-covered lakes of high latitudes (Wharton et al. 1993;Ludlam 1996). Wharton et al. (1993) suggested two main sources for high oxygen concentrations: oxygenic photosynthesis and oxygen exclusion from newly forming ice when water from inflowing streams freezes in contact with the bottom of the ice cover.
Vertical gradients in conductivity were steeper in the ice-covered sites (Fig. 3), probably due to the cold meltwater derived from the ice cover being less concentrated in ions, and remaining at the surface without mixing with the rest of the water column because of its greater buoyancy at near-zero temperatures (Bergmann and Welch 1985). Conductivity in the water column increased along the inshore-offshore gradient, possibly due to inputs from the sediments (including weathering of catchment mineral particles), from ions released by ice formation in the preceding fall and from microbial decomposition processes. The water tracks on the western shore of the watershed are the main source of freshwater for the lake (Paquette et al. 2017). As also seen in the present study, the ionic composition of freshwaters channelled through water tracks on the western slope of Ward Hunt Lake is dominated by bicarbonate and calcium (Paquette et al. 2020). The accumulation of ions in Ward Hunt Lake could further be accelerated by the absence of mixing of low-conductivity ice-melt water with the underlying water column before it reaches the outflow, as was observed in P&N Lake on the west coast of Hudson Bay (Bergmann and Welch 1985).
As the result of the ice cover, the moat and ice-covered zones differed markedly in terms of the quantity of solar energy reaching the top and bottom of their water columns (Fig. 4). In nearby Lake A, the K d is in these wavelengths was up to 53% lower in snow-free ice than in water, but 43% higher in the PAR waveband (Belzile et al. 2001). The penetration of green light at the bottom of the moat and ice-covered zones (notably in the range of 500 to 575 nm) observed in Ward Hunt Lake is generally common in lakes since maximal absorption for Chl a is at 640 and 405 nm and for Chl b at 620 and 440 nm, with carotenoids and CDOM also absorbing light in blue-green wavelengths and water itself absorbing strongly at the red end of the spectrum (Wetzel 2001).
There were pronounced differences in upward spectral irradiance between the moat and icecovered zones of the lake. In the upper waters of the central under-ice site (S4), the maximal irradiance was observed between 500 and 575 nm, as for downward irradiance (Fig. 4). At the bottom, there was a shift towards a maximum at 563 nm, which indicated a selective reflection of light from the bottom at green wavelengths, and selective absorption of light at higher and lower wavelengths by microbial mat pigments. In the inshore moat site (S1), upward irradiance at wavelengths above 570 nm was higher than upward irradiance at the surface (Fig. 4), which was likely due to pigments reflecting in the orange and red part of the spectra, and confering the colour of the mats (Supplementary Fig. S2b). The inshore mats had higher concentrations of cyanobacterial pigments including 4-ketomyxol 2'-fucoside, β,β-carotene, echinenone and the UVabsorbing pigment scytonemin (Supplementary Table S4). High concentrations of carotenoids associated with cyanobacteria were previously observed in Ward Hunt Lake's inshore microbial mats, and their taxonomic analysis indicated a high proportion of the scytonemin-rich genus Dichothrix (Villeneuve et al. 2001). This is also in accordance with observations from the mats in Lake Hoare, Antarctica, where cyanobacterial pigments associated with high carotenoid concentrations were mainly concentrated in moat mats (Hawes and Schwarz 1999). Moreover, the light reaching the bottom of the Ward Hunt Lake inshore moat site is reduced in wavelengths lower than 400 nm, probably as a result of absorption by CDOM (Fig. 5). The lake watershed is rich in carbonates (Trettin 1991), and the resultant light coloured rocks confer a high reflecting potential to the rock-coated moat zone (Supplementary Fig. S2a; Wetzel 2001).
The UVR attenuation coefficients (K d ) at 320, 340 and 380 nm for the first 2 m at the surface of Ward Hunt Lake were respectively 0.60, 0.47 and 0.32 m -1 in the inshore moat site (S1) and 0.63, 0.62 and 0.63 m -1 in the central site (S4; Fig. 4). These lie well below values recorded in the surface 5 m of the nearest lake, meromictic Lake A, located 20 km southwest of Ward Hunt Lake, where K d was 1.65, 1.33, and 0.80 m -1 at the same wavelengths. These higher values at Lake A are likely due to higher CDOM absorbance, with its a CDOM values between 2 and 3 m -1 at 320 nm The higher K d spectral values in the central site (S4) of Ward Hunt Lake for the UVR portion of the spectra are coherent with higher absorption by algal and non-algal particles (Fig. 5). In 2015, the central site contained lower concentrations of nanoplankton (S1; Supplementary Table S1).
However, Chl a concentrations were higher in the central site compared to the inshore moat site, along with higher total pigment concentrations. Hence, cells could be less abundant but cellular pigment concentrations higher as a result of photoacclimation to the low light environment (Thompson et al. 1991). Ice cover on the central site of Ward Hunt Lake was thinner at the moment of irradiance measurements in 2016 and optical measurements in 2017 (average of 180 and 151 cm, respectively) compared to the inshore-offshore gradient sampling performed in 2015 (218 cm), and this may have contributed to some of the interannual differences in plankton and pigments.

Inshore-offshore gradients
Limnological variables followed two distinct patterns along the inshore-offshore gradient: a pronounced change at the ice margin (S2) that contrasted with the three other sites, and contrasting conditions between the ice-free and ice-covered sites (S1 and S2 versus S3 and S4). These two patterns highlight the role of ice margin as a zone of disjunction in the surface limnological properties of the lake. According to the first pattern, there was a clear decrease in concentrations of DOC, Chl a and mixotrophic protists in the ice-margin site (Fig. 6b,d,i). The lower concentrations observed in the ice margin site could be the result of convective upwelling. Lowdensity cold water released from the melting ice cover may circulate from under the ice towards the shore (Bochaton 2017). Given the homogenous physical conditions in the inshore moat site (S1), convective upwelling initiated by the warming of its water in the inshore site, as observed in Lake N2 in Alaska (Cortés and MacIntyre 2020), seems unlikely. Total nitrogen concentrations, contrary to most of the other variables, was higher at the ice margin (Fig. 6a), possibly as a result of dissolved and particulate N release from the melting ice face. However there were lower concentrations of DOC, bacteria, Chl a, autotrophic protists and pigments, suggesting lower primary production.
The second observed pattern included the change of concentration in viruses and rotifers between the ice-covered and the ice-free zones (Fig. 6). Viruses had lower concentrations inshore and in the ice margin, and their increase under the ice co-occurred with higher concentrations of DOC, Chl a and bacteria (Fig. 6b,d,g). Viruses have the potential to short-circuit the classical food web by attacking primary producers and releasing carbon that is only available to smaller organisms in the lower part of the food web (Wilhelm and Suttle 1999). The virus to bacteria ratio (VBR) ranged from 1.8 to 3.8, which is in the lower range of the Antarctic freshwater environments described by Säwström et al. (2007). Low VBRs suggest low bacterial mortality due to viral infection and is correlated with low bacterial abundance (Wommack and Colwell 2000;Yager et al. 2001). Lower concentrations of viruses in the ice-free zones could be attributable to their vulnerability to decay via sunlight (Suttle and Chen 1992). The pattern in rotifer communities was completely the opposite, where higher concentrations were observed in the moat zone (Fig. 6k) along with a higher abundance of picocyanobacteria and autotrophic protists (Fig. 6f,h). The central site (S4) was dominated by K. hiemalis that feeds preferentially on bacteria and detritus. The inshore moat site (S1) was dominated by high abundances of species that feed on larger particles such as Ascomorpha sp. that prefers dinoflagellates, Rhinoglena sp. that feeds on chrysophytes and A. priodonta that is predatory (Supplementary Table S2; Pourriot 1977).
Some variables did not clearly follow the two patterns depicted above. Autotrophic protists were higher in the inshore moat and dropped consistently in the ice margin to stay low under the ice (Fig. 6h). Higher concentrations of phototrophs are likely associated with higher light availability, and the drop at the ice margin could be related to higher grazer concentrations (heterotroph protists and rotifers; Fig. 6j,k). The most abundant rotifer in the ice-margin site, Rhinoglena sp., feeds on small unicellular algae and on detritus (Ruttner-Kolisko 1974). Higher concentrations of bacteria observed in the central site (S4) could be associated with a lower predation rate by heterotrophs or higher concentrations of DOC. However, genomic analyses of the microbial mats from the moat zone at our site S1 detected heterotrophic ciliates as the dominant eukaryotes, suggesting high microbial food web activity (Mohit et al. 2017).
Consistent with our hypothesis, there were clear discontinuities at the ice-margin site, indicating the decisive effect of ice cover in separating habitat properties between inshore and offshore parts of the lake. Lake ice margins have some characteristics in common with the edge of the sea ice in the Arctic Ocean, which is also undergoing transition from perennial to seasonal ice conditions associated with climate warming (Polyakov et al. 2012). The limnological variables at sites 3 and 4 of Ward Hunt Lake were closely similar, as was ice thickness. The ice-covered area of polar lakes thus has the potential to maintain a relatively uniform conditions, in contrast to the much more variable conditions of the moat zone. In the marginal ice zone of the Arctic Ocean, mismatches have been observed between phototrophs (ice algae and phytoplankton) and heterotrophs due to the earlier ice break up (Søreide et al. 2010). Such mismatches may also occur in Ward Hunt Lake where the phytoplankton phenology has been observed to change with complete loss of ice cover (Bégin et al. 2020).
Cluster analysis of the different biological variables confirmed the strong spatial differences.
Rotifer communities clustered distinctly in the ice-free and ice-covered zones, and the communities of the inshore and ice-margin sites were in separate branches, while those in the central and intermediate sub-ice sites were merged (Fig. 7c). A clear distinction between fatty acid assemblages of zooplankton between open-water and ice-covered sites was also observed (Fig.9).
Our results suggest that under the ice, zooplankton diet is more likely to be composed exclusively of seston, whereas in the inshore moat site microbial mats or epibionts associated with mosses, with their higher concentrations of PUFAs (Fig. 8), could contribute to a larger extent to zooplankton diet. Sea ice has been identified as a rich ecosystem where algae can contain higher levels of fatty acids than phytoplankton with large proportions of PUFAs (Falk-Petersen et al. 1998), but lake ice lacks the nutrient-rich brine channels of sea ice and would seem to provide a much less suitable habitat for algal growth. Little is known, however, about the biomass and productivity of ice-associated algae in lakes (Hampton et al. 2015).
A clear distinction according to the presence of the ice cover in Ward Hunt Lake was also observable in the limnological variables, pigments of the water column and PUFAs in chironomids ( Fig. 7a,e, 8 and 9), underscoring the homogeneity of response of these variables under the ice cover. The clear pattern in pigments was not reflected in the cluster of protist communities, which included autotrophs, heterotrophs and mixotrophs (Fig. 7b), nor between under-ice and moat seston fatty acid assemblages. (Fig. 8). The central sub-ice site generally contained more heterotrophs than the inshore moat site S1, mainly small-sized (<5 µm) non-flagellated heterotrophs (Supplementary Table S1). The inshore site was characterized by the highest concentrations of autotrophs, likely due to light availability, especially diatoms, colonial cyanobacteria (Snowella sp.) and the chlorophyte Micractinium sp. Mixotroph protists were generally more abundant in the intermediate site, especially Dinobryon sp., Ochromonas sp. and Cryptomonas sp. Additional under-ice light measurements would help to determine if ice degradation in this site could have reduced light availability, as white ice attenuates more light than candled ice (Welch et al. 1987).
Lower light availability could give a competitive advantage to mixotrophic species (Jones 2000).
Our discovery of abundant tubes in the benthos, most likely produced by chironomids, suggests a sustained presence of these dipterans in the Ward Hunt Lake ecosystem and their likely participation in food web dynamics. Their migration from the tubes in the benthos, swimming up through the water column and then passing through fissures in the candled ice, is consistent with previous observations in the Arctic (Oliver 1964). The tube concentrations measured here (14 000 m -2 ) is at the upper range for lakes elsewhere, and likely results from numerous seasons of accumulation. Chironomid larvae have been observed in densities between 70 and 11 000 ind. m -2 (Hölker et al. 2015), and in laboratory experiments, they can cover up to 48% of the surface of artificial substrates (Pringle 2004). Although we were able to collect only a small number of chironomid individuals, the fatty acid analyses indicated that there was considerable variability in chironomid diet among mats, plankton and mosses. Despite their lower fatty acid concentrations compared to seston, microbial mats and mosses may be a good food source as they have a high proportion of PUFAs, especially ω3 and ω6 in the inshore moat site, which was reflected in larvae (Figs. 8 and 9; Supplementary Fig. S3), and the benthic mats could also subsidize the central subice food web (Mariash et al. 2014). Chironomids display a remarkably large variety of modes of feeding, including by filtering, gathering, scraping, shredding, engulfing and piercing, and they are opportunistic in the type of prey they consume (Armitage et al. 2012). Food web analyses in Alaskan shallow lakes also suggested that chironomids may be closely linked to methane cycling, as their diet can be substantially composed of methanotrophic bacteria (Hershey et al. 2006).
Methanotrophs could be a source of food for chironomids in Ward Hunt Lake as they are abundant in the central sub-ice mats (Mohit et al. 2017). Further analyses could be undertaken on samples from the lake using adducts of monounsaturated fatty acids specific to methanotrophic bacteria and mixing models adapted to fatty acid assemblages (Virtue et al. 1996;Galloway et al. 2014;Wauthy and Rautio 2020) to disentangle and quantify the contribution of food sources.
One of the key aspects of the moat environment to examine in the future is its hydrodynamic regime. At the time of the first studies on Ward Hunt Lake, the 4-m thick lake ice had been found to extend to the lake floor, and it was assumed that the moat region along the western shore constituted the entire liquid water of the lake, with a short residence time before discharge at the outlet (Villeneuve et al. 2001). Subsequent studies with Ground Penetrating Radar, revealed a large area of the lake that extended to depths well below 4 m, as well as a thinning of the ice cover (Paquette et al. 2015). The latter would have enhanced the potential exchange of moat water with the offshore waters beneath the ice, including via density flows from water tracks entering the moat. Nonetheless, the persistent thick ice likely acts as an impediment to offshore exchange, and the moat water would have a shorter residence time than the main body of the lake, contributing to the distinctive limnological characteristics observed here.

Conclusions
The abrupt inshore-offshore changes that we observed in many limnological properties at the ice margin of Ward Hunt Lake underscore the importance of ice cover in the regulation and The ice edge shifts only slowly in position after the initial early season melting, but this is subject to climate-determined variations among years that in turn will influence the spatial organization of the lake, including its food webs. As atmospheric temperatures are increasing, the thickness, area and duration of the ice cover and the overlying snow depth on polar lakes are decreasing, including towards more frequent complete loss of ice. This will make the moat zone become a shorter-lived feature and may lessen the extent of inshore-offshore differences in the limnology of high latitude lakes. show the sampling locations in the ice-free moat zone (S1, the inshore moat site; S2, the icemargin site) and in the ice-covered zone (S3, the intermediate sub-ice site; S4, the central sub-ice site). The red arrows represent the inshore-offshore transect.        sites. Arrows represent the contribution of the five most influential fatty acids for the distribution of samples. *Length of the C16:1n-7 arrow was divided by 3. Figure 1. Sampling along the inshore-offshore transect in Ward Hunt Lake. The black arrows show the sampling locations in the ice-free moat zone (S1, the inshore moat site; S2, the ice-margin site) and in the ice-covered zone (S3, the intermediate sub-ice site; S4, the central sub-ice site). The red arrows represent the inshore-offshore transect.
1534x635mm (72 x 72 DPI) Figure 2. Images of Ward Hunt Lake captured by an automated camera located on its western shore. The full record of images and video sequences is archived in NEIGE (2020).
1534x1693mm (72 x 72 DPI) Figure 3. Profile gradients along the inshore-offshore transect of Ward Hunt Lake: a) temperature, b) specific conductivity and c) dissolved oxygen. Profiles were recorded on 13 July (c and d) and 19 July (a and b), 2015. Oxygen concentrations in the lake are expressed as % air-equilibrium. S1 was the inshore moat site, S2 the ice margin site, S3 the intermediate sub-ice site, and S4 the central sub-ice site.
101x190mm (300 x 300 DPI) Figure 4. Irradiance profiles at the inshore moat site (S1, 15 July 2016) and the central under-ice site (S4, 14 July 2016) of Ward Hunt Lake. Total irradiance values are for downward irradiance integrated across wavebands, expressed as percent incident downward irradiance (note the logarithmic scale). Downward irradiance (DI) is expressed as percent of downward irradiance recorded just below the surface (JBS) of the open water at S1, or just below the ice (JBI) at S4; upward irradiance (UI) is expressed as percent upward irradiance at JBS (S1) or JBI (S4). The diffuse attenuation coefficients (K d ) are for the upper 2 m of the water column at each site, and for the ice-cover at S4, uncorrected for reflection.
184x127mm (300 x 300 DPI) Figure 5. Spectral absorption coefficients for Ward Hunt Lake water samples. The coefficients are for: a) CDOM, a CDOM (λ); b) non-algal particles, a NAP (λ); and c) algal particles, a φ (λ). The samples were from the upper water column at sites S1 and S4. The shading represents ±SE for triplicate samples. . Cluster analysis of limnological variables, biological communities and pigments along the inshoreoffshore transect in Ward Hunt Lake, July 2015. Dendrograms represent Ward clustering performed on the following distance matrices: standardized Euclidian distances for limnological variables and Bray-Curtis distances for nanoplankton, rotifers and pigment assemblages. Points filled with gray denote sites covered by ice.
184x101mm (300 x 300 DPI) Figure 8. Fatty acid methyl esters (FAME) in samples from Ward Hunt Lake at the inshore moat (S1, in white), intermediate under-ice (S3, in grey) and central under-ice (S4, in black) sites: a) total FAME concentration in µg/mg dry weight; proportion of polyunsaturated fatty acids (PUFA) in b) microbial mats, c) mosses, d) chironomid larvae, e) seston, and f) zooplankton, expressed as percent of total FAME in their respective zones. Values are means for triplicates with SE, except for inshore mats and chironomids (n=2) and central chironomid larvae (n=1). Results of the ANOVA comparisons of total FAME: in food web compartments in panel a) F 4,22 =5.75, P=0.02; in mats in panel b) F 2,6 =16.85, P=0.003; in seston in panel e) F 1,5 =2.77, P=0.16; and in zooplankton in panel f) F 1,5 =2.65, P=0.20. Chironomid larvae were not tested due to insufficient numbers of samples.
184x65mm (300 x 300 DPI) Figure 9. Principal coordinate analysis (PCoA) of fatty acids in the Ward Hunt Lake food web. Samples were from the inshore moat (S1), intermediate under-ice (S3) and central under-ice (S4) sites. Arrows represent the contribution of the five most influential fatty acids for the distribution of samples. *Length of the C16:1n-7 arrow was divided by 3.