Keystone seabird may face thermoregulatory challenges in a warming Arctic

Climate change affects the Arctic more than any other region, resulting in evolving weather, vanishing sea ice and altered biochemical cycling, which may increase biotic exposure to chemical pollution. We tested thermoregulatory impacts of these changes on the most abundant Arctic seabird, the little auk (Alle alle). This small diving species uses sea ice-habitats for foraging on zooplankton and resting. We equipped eight little auks with 3D accelerometers to monitor behavior, and ingested temperature recorders to measure body temperature (Tb). We also recorded weather conditions, and collected blood to assess mercury (Hg) contamination. There were nonlinear relationships between time engaged in different behaviors and Tb. Tb increased on sea ice, following declines while foraging in polar waters, but changed little when birds were resting on water. Tb also increased when birds were flying, and decreased at the colony after being elevated during flight. Weather conditions, but not Hg contamination, also affected Tb. However, given our small sample size, further research regarding thermoregulatory effects of Hg is warranted. Results suggest that little auk Tb varies with behavior and weather conditions, and that loss of sea ice due to global warming may cause thermoregulatory and energic challenges during foraging trips at sea.


Study system
Our study took place at a breeding population of little auks situated at Ukaleqarteq (Kap Höegh), East Greenland (70°44′N, 21°35′W).Little auks (~ 150 g) are the most abundant seabird in the high Arctic and breed in large colonies 36 .Both males and females incubate a single egg, and contribute equally to provisioning of the chick 37,38 .Little auks forage on copepods, and return to the colony from foraging sites at sea with prey items stored in a specialized gular pouch, which are then regurgitated for the chick 38 .Upon return to the colony, little auks can be captured and recaptured at or near the nesting sites in rock crevasses, facilitating fitting and retrieval of accelerometers and deployment of T b loggers.Mean blood Hg levels in little auks at Ukaleqarteq in recent years fall into the low risk range for toxicological effects.However, past research documents negative effects on reproduction 39,40 and body condition 14 , suggesting that despite relatively low Hg levels, there may be other effects on physiology and thermoregulation.Fieldwork adhered to the ASAB/ABS guidelines for use of animals in behavioral research, and was conducted in accordance with Greenlandic law.The Government of Greenland, Ministry of Environment and Nature, and Department of Fisheries, Hunting and Agriculture, approved research procedures and provided ethical clearance (permit: 2020-1006).The methodology and results of this study are reported in accordance with ARRIVE guidelines.

Deployment of T b loggers and accelerometers
During July 2020, we captured eight little auks outside nesting crevasses using a combination of lassos and noose carpets.We fit each bird with a T b logger (BodyCap Anipill Core Body Temperature Ingestible Tablet; BMedical; ± 0.2 °C; 1.7 g; 17.7 × 8.9 mm; ~ 1% of mass), a data logging system for gastrointestinal temperature recording.These capsules are designed for use in birds, and have been used by a number of past studies on wild and freeranging species 41,42 .Birds spontaneously ingested T b loggers placed within the beak.Loggers recorded abdominal temperature every minute for periods of ca. 30 h.We remotely downloaded data from loggers via telemetry when the bird was within ~ 1 m.T b loggers are assumed to be eliminated through defecation or regurgitation, and are not recovered.Focal individuals were simultaneously equipped with miniaturized triaxial accelerometers (Axy 4, Technosmart, 3 g, ~ 2% of mass), to record body acceleration.We attached accelerometers to the central breast at the level of the sternum using Tesa® tape.Accelerometers recorded data at a frequency of 50 Hz (50 readings per second).We marked birds with color rings to facilitate identification and recapture within ~ 4 days, upon which we retrieved the accelerometer.Deployment dates fell within nine days during the mid-late chick rearing phase [July 22 -30].All birds were breeding adults, but sex was not determined for this study, as the remaining volume of blood was conserved for other physiological assays.The thermoregulatory physiology of male and female little auks during the nestling stage is likely to be similar, as the sexes are monomorphic in coloration, overlap extensively in size (males may be slightly larger), and share equally in breeding duties 37,38 .Nevertheless, we recognize lack of knowledge of sex as a limitation to our work.A portable HOBO H21-USB weather station at the study site recorded weather conditions (every 1-min), including ambient temperature (T a ; °C), relative humidity (RH; %) and wind speed (V; m/s).Conditions measured at the weather station were taken as a proxy of environmental conditions at both the breeding site and at-sea foraging grounds (as a caveat, little auks can forage up to ~ 100 km from the colony 43 ).

Analysis of accelerometry and T b data
We used Igor Pro 8.04 (64-bit; WaveMetrics) to classify data on triaxial acceleration into behavioral states (see details in 15 ).In brief, we used k-clustering analysis applied to acceleration axes, followed by application of a custom-written script, which utilized output from the clustering analysis and surface temperature data.Behavioral states identified were: flying, diving, on the water surface, on sea ice, and at the colony.We proceeded to determine whether time spent on the water was part of a foraging bout (i.e.inter-dive interval), or represented time resting on the water.To this end, we determined the dive bout ending criterion, using R package diveMove 44 , which applies the methods of 45 and 46 for identification of behavioral bouts.Using the standard method of classification, based on absolute duration of behavioral bouts (i.e.inter-dive intervals), the bout ending criteria derived was 307.1 s.Consequentially, we ended diving bouts if time spent on the water exceeded this value, and classified these intervals as time resting on the water.Time resting on the water additionally encompassed intervals of time on the water that were not between dives.We combined time engaged in diving and inter-dive intervals into a single behavioral category, representing active foraging.Thus, final behavioral categories were: actively foraging (hereafter also "diving"), flying, at the colony, on sea ice, and resting on the water.For each T b measurement, we determined the corresponding behavioral state by aligning time stamps from the T b and behavioral (accelerometer) data in Microsoft Excel 16.16.27.

Mercury contamination: sampling and analysis
We obtained ~ 0.2-0.5 ml blood samples from the brachial vein after recapture of birds to retrieve accelerometers, which minimized stress during the experimental period.Blood samples were centrifuged for 10 min at 3500 rpm to separate plasma from red blood cells (RBCs), which were stored in 70% ethanol.After evaporation of ethanol, RBCs were freeze dried for 48 h and homogenized prior to analysis for total Hg (hereafter Hg) concentrations.Total Hg serves as a proxy for highly toxic MeHg, since most of the Hg in bird blood, feathers and eggs is MeHg 47 .Samples were analyzed in duplicate using an Advanced Mercury Analyser spectrophotometer (Altec AMA 254) at the Institute Littoral Environnement et Sociétés (LIENSs) 48 .The standard deviation between duplicates was < 10%.We used TORT-3 as a standardized reference material (CRM; Lobster Hepatopancreas Tort-3; NRC, Canada; [Hg] = 0.292 ± 0.022 µg g −1 dry weight (dw)) and performed a blank before initiating measurements on samples.The limit of detection for Hg and mean ± SD of Tort-3 measurements were 0.005 µg g −1 dw and 0.302 ± 0.004 µg g −1 dw (N = 8; replicates), respectively.

Statistical analysis
We conducted statistical analyses in R 3.6.1 49 .We used generalized additive mixed effect models (GAMMs) in R package mgcv 50,51 to assess whether T b varied with behavior, environmental conditions, or time of day.For this model, we used each observation of T b .We used the corAR1 correlation structure in package nlme to account for temporal autocorrelation 52 , included individual ID and behavioral bout (nested within individual ID) as random effects, and incorporated two non-linear smooth terms.The first non-linear smooth term tested for non-linear variation between T b and time in each behavioral state.To this end, we used a cubic regression spline (specified as bs = "cr") with the degree of smoothness set to k = 50 (which minimized AICc and increase the R 2 relative to models with lower k).We used the "by" call within the smoothing function to test for unique non-linear relationships within each behavioral state.For the second non-linear smooth term, we used a cyclic cubic regression spline (specified as bs = "cc") to test for variation in T b with time of day.We also included two-way interactions between behavioral state and: (1) T a , (2) RH, (3) wind speed, and (5) Hg concentrations.We removed interactions with P values > 0.059 from models, followed by elimination of main effects above P > 0.05.We standardized continuous predictor variables to a mean of zero and standard deviation of one to facilitate interpretation of main effects when including interactions in models 53 .We used package emmeans 54 for pairwise comparisons between interaction terms (function emtrends) and differences in T b between behavioral states (function emmeans).
To more thoroughly explore differences in how T b changes with time when birds are engaged in different behaviors, we calculated change in T b (deltas) for each behavioral bout as: ΔT b = T b,end -T b,start ; where T b,end = T b at the last time point recorded in that behavioral state and T b,start = T b at the first time point recorded.We used a linear mixed effects model in nlme to compare ΔT b across behavioral states, while including the length of the time interval in the model.We extracted and plotted predicted values from models using function ggpredict within the ggeffects package 55 .In addition, to evaluate modulators of the effect of time of day on T b , we constructed GAMMs predicting T a and RH from time of day, using the same random structure as above.
We also assessed whether between minute variation in T b differed between behavioral states by calculating the absolute value of the difference between subsequent measurements of T b , and constructing models with the same structure described for mean T b .Values could not be calculated for time points at the beginning of behavioral intervals, so these rows were dropped from the analysis.

Predictors of little auk T b across behavioral states
Mean ± SD T b of little auks was 41.0 ± 0.55 °C and showed significant, but low magnitude differences between behaviors (Table 1; see  S2 for pairwise comparisons).The minimum adequate model predicting T b included non-linear relationships with time spent in each of the behaviors, with the exception of time spent on the water surface, for which the relationship was non-significant (Table 1).T b decreased with the amount of time spent diving up to ~ 19 min, from ~ 41.5 to 40.6 °C, after which predicted T b plateaued (Fig. 1a).There was also a more gradual decrease in T b after birds arrived at the colony up to ~ 90 min, from 41.9 to 40.9 °C (Fig. 1a).On the other hand, T b increased the longer birds spent flying to ~ 15 min, from ~ 41.3 to 41.6 °C°, before leveling off or declining slightly (Fig. 1a).T b also displayed an increase for the initial 30 min when birds were on the sea ice, from ~ 40.8 to 41.4 °C, after which T b leveled off or declined slightly (Fig. 1a). Figure 2  The overall ΔT b for diving was negative, with the 95% CI not overlapping zero (EMM ± SE [95% CI] = − 0.312 ± 0.026 [− 0.374, − 0.251] °C), and was lower than all other ΔT b (Table S3 for pairwise comparisons).The ΔT b for flying and on sea ice were positive, with Cis not overlapping zero (EMM ± SE [95% CI] = 0.146 ± 0.029 [0.077, 0.215] °C; 0.335 ± 0.045 [0.228, 0.443] °C), and were higher than all other ΔT b , with ΔT b for sea ice also greater than that of flying (Table S3 for pairwise comparisons).The overall ΔT b for at the colony and on the water were negative, and positive, respectively, but did not significantly differ from each other or zero (Estimate marginal mean ± SE = − 0.074 ± 0.044 [− 0.179, 0.031] °C; 0.0002 ± 0.038 [− 0.089, 0.089] °C, respectively).
With respect to environmental effects, there was a significant interaction between behavioral state and wind speed in predicting T b (Table 1).T b tended to increase with wind speed when birds were flying, but not in other behavioral states (Table 1; Fig. 1b; Table S4 for pairwise comparisons).In addition, T b was positively related to RH (Table 1; Fig. 1c), independent of behavioral state (Table S1).T a was unrelated to T b , either independently, or in interaction with behavioral state (Table S1).
Finally, the best model predicting T b included a non-linear effect of time of day on T b (Table 1; Fig. 3), with bimodal peaks in the late morning (~ 10:00) and around 20:00, and the lowest values in the early morning (~ 4:00 am) and around 16:00 (Fig. 3).When examining potential environmental effects underlying this relationship, we found that there was a non-linear, but unimodal, relationship between time of day and T a , with a peak around ~ 16:00 (edf = 7.33, F 8 = 175, P < 0.001; Fig. S8).Similarly, RH was non-linearly related to time, and showed a reverse pattern to T a , with a nadir around ~ 16:00 (edf = 7.30, F 8 = 203, P < 0.001; Fig. S9).

Predictors of between minute variation in T b
Mean ± SD between minute variation in T b (|T b1 -T b1+1 |) of little auks was 0.09 ± 0.11 °C (range: 0-1.5 °C), and showed low magnitude, significant variation between behavioral states (Table 2; Table S5 for full model).The EMM (± SE) was highest when birds were foraging at sea (0.097 ± 0.008 [0.080, 0.113] °C), followed by flying (0.060 ± 0.008 [0.044, 0.077] °C), on the water (0.055 ± 0.011 [0.033, 0.078] °C), at the colony (0.041 ± 0.008 [0.024, 0.057] °C) and on sea ice (0.040 ± 0.007 [0.025, 0.054] °C), but the only significant differences were between diving and all other behavioral states (Table S6 for pairwise contrasts).There were significant non-linear relationships www.nature.com/scientificreports/ between time in the behavioral state and between minute variation in T b (Table 2); which involved early decreases with time in the behavioral state for all behaviors, before leveling off (Fig. 4a).This decrease was steepest when birds were resting on the water, diving, or in flight (all with a slope of ~ − 0.007 °C min −1 in the first ~ 15 min), and slightly more gradual at the colony (slope of ~ − 0.004 °C min −1 in the first ~ 25 min) and on the sea ice (slope of ~ − 0.002 in the first ~ 15 min) (Fig. 4a).
With respect to environmental effects, there was an interaction between wind speed and behavioral state in predicting between minute variation in T b (Table 2; Fig. 4b).Variation in T b increased with wind speed when birds were diving and flying but did not vary significantly with wind speed in the other behavioral states (Table S7 for pairwise comparisons).There was a positive correlation between T a and variation in T b (Table 2; Fig. 4c) independent of behavioral state (Table S5).When T a was also in the model, RH was unrelated to variation in T b (Table S5).However, when T a was removed, RH was negatively related to variation in T b (β ± SE = − 0.006 ± 0.002, T = − 3.42, P = 0.006), independent of behavioral state (Table S5).
Between minute variation in T b did not vary with time of day (Table S5).

Mercury contamination
There was no significant relationship between Hg contamination and T b (Table S1) or between minute variation in T b (Table S5).The mean ± SE of Hg concentrations in RBCs was 1.290 ± 0.031 µg g −1 dw [range: 1.030-1.746µg g −1 dw], which assuming 79% blood moisture content is equivalent to 0.271 ± 0.007 µg g −1 ww [range: 0.216-0.367µg g −1 ww], and falls within the low risk range for toxicological effects (0.2-1.0 µg g −1 ww 47 ;   www.nature.com/scientificreports/calculated for whole blood, but comparable to levels in RBCs).Table 3 lists the Hg concentrations in RBC of each focal bird in dw and ww equivalents.

Discussion
By employing advanced biologging approaches, we demonstrate that the T b of a free-ranging seabird is sensitive to variation in activity patterns and environmental conditions, and suggest thermoregulatory challenges that may arise under climate change scenarios.In particular, results suggest that the evolving nature of the Arctic cryosphere, especially sea ice extent and coverage, may not only alter foraging conditions for little auks 14,56,57 , but also affect thermoregulatory dynamics during foraging trips at sea.There was no evidence that Hg contamination modifies the T b of little auks, but our results regarding potential toxicological effects on thermoregulatory capacity are limited by a small sample size.Thus, further research is called for in this area.Changes in T b with activity were generally as expected.T b declined while little auks were foraging in cold waters.This decline may facilitate aerobic capacity and limit heat loss, but also reflects thermoregulatory challenge 34,58,59 .In addition, declines in T b while foraging, and high variability in T b , may arise from ingestion of cold prey.Indeed, ingestion-linked declines in T b have been used to identify feeding events 60 , but our data lacked resolution to achieve this end.In another Alcid, Brünnich's guillemots (Uria lomvia), T b declined over sequential diving bouts 35 .However, this decline occurred during periods of resting on the water.During dives themselves, T b increased and the peripheral temperature declined 35 .This pattern contrasts to T b declines during diving observed in some penguin species 59,61,62 , and may reflect peripheral vasoconstriction and high wing beat frequency that generates heat 35 .Our data could not separate changes in T b during dives and inter-dive intervals.Thus, a similar dynamic could be occurring in little auks.
Also as expected, T b was highest when little auks were flying, although the estimated marginal mean was not significantly different than when birds were on sea ice, at the colony, or on the water.Furthermore, T b increased during flight before leveling off, reflecting heat generated by intense physical activity.Indeed, little auks have high wing loadings and flight costs (~ 7.24 × BMR 32 ), which is expected to generate substantial amounts of heat and elevate T b .
T b was also high when little auks arrived at the colony.T b then declined before leveling off.This decrease in T b may reflect decreases following commuting flights between foraging sites and the colony, with a decline in diet-induced thermogenesis (i.e.heat production associated with digestion, also referred to as specific dynamic action 63 ), also potentially contributing.T b of little auks at the colony was also relatively invariable, which could suggest that the colony serves as a thermal refuge for little auks.During the current study, we did not detect evidence that little auks were subject to thermal stress on land.The upper critical temperature (UCT) of little auks is ~ 20°C 64 , a temperature not reached during the relatively cool 2020 breeding season.However, an air temperature of 20 °C was exceeded twice during July 2021, and captured birds were observed to rapidly exhibit sign of overheating (unpublished data).The operative temperature of little auks may be further elevated by solar radiation off the rocks, especially given their black plumage coloration 9 .Thus, with ongoing climate change, heat stress at the colony may eventually pose a threat, especially in the context of stress from predation pressure, which can activate flight-fight responses, or social interactions, which have been shown to elevate T b in a range of animal species [65][66][67] .Indeed, due to their Arctic habitat and amphibious live style, little auks have evolved a low thermal conductance, which conserves heat and energy under cold conditions, but creates a challenge for heat dissipation 64 .As predicted, little auks' T b increased substantially on sea ice following declines during diving episodes, before leveling off, suggesting that loss of sea ice as a resting substrate may elevate thermoregulatory costs, negatively  www.nature.com/scientificreports/affecting energy balance.T b was initially low when little auks first emerged on the sea ice, which could reflect that birds exit the water when T b falls below a threshold, triggering birds to cease foraging activity 35 .For the first 30 min, T b on sea ice increased by an average of ca.0.02 °C min −1 , while during foraging T b dropped by an average of ca.−0.05 °C min −1 for the first 19 min.Thus, the rate of T b loss while foraging was greater than the rate of gain on ice, but given that the average ice bout lasted ca. 25 min and the average foraging bout ca. 9 min, time spent on ice would allow for recovery of T b .On the other hand, there was no significant relationship between time resting on the water and T b , and although positive, the ΔT b of birds resting on water surface was not significantly different than zero, suggesting little potential for recovery of T b after foraging bouts.Furthermore, the lowest variation in T b occurred when birds were on sea ice, and the highest during diving behavior, suggesting that resting on sea ice plays an important role in allowing birds to restore and maintain normothermic temperatures after thermally challenging foraging bouts.
In the context of climate change, sea ice loss may have energetic and thermoregulatory implications, as birds are forced to instead rest on the water, which has ~ 25 × higher thermal conductivity than air 33,68 .In auks, compression of air space in feathers while diving reduces insulative properties, further facilitating heat exchange with the environment 69 .The costs of resting both in the air and on the water may be reduced by warming temperatures 70 .However, the thermal neutral zone, outside of which metabolic rate must be elevated to maintain T b , has been shown to be considerably narrower for seabirds resting on the water relative to when resting in air, with metabolic rate also increasing more steeply below the lower critical temperature (LCT) in some species 71 .For example, in Brünnich's guillemot, the LCT in air and water are 2 and 16 °C, respectively 71,72 , and the rate of increase in metabolic rate below the LCT is substantially greater in water than in air (0.60 versus 0.17 W × kg −1 °C−1 ) 71 .Similarly, in another small diving seabird, the Cassin's auklet (Ptychoramphus aleuticus), the LCT in air and water were 16 and 21 °C, respectively, and resting metabolic rate was 25% higher in water than in air 73 .Although, comparable data for the little auk is unavailable, these species are in the same family (Alcidae) as the little auk, and share a similar ecology.
At the Ukaleqarteq study site, even in an exceptionally warm year (2021) with very low sea ice coverage (mean ± SE: 0.04 ± 0.02%), sea surface temperature within the foraging range (mean ± SE: 5.89 ± 0.38 °C) 15 remained well below the LCTs in water reported for other Alcids (see above).Thus, even in warming oceans, little auks resting on the water may need to elevate their metabolic rate relative to when resting on ice, which, in lieu of compensatory changes, could elevate daily energy expenditure 74 , induce birds to return to the colony sooner, limiting time for energy acquisition, or force higher feeding rates.In addition, we recently demonstrated that high SST is associated with elevated daily energy expenditure in Ukaleqarteq little auks, a phenomenon associated with increased flight costs, as birds appeared to fly further to reach foraging grounds 75 .Longer flights may be motivated by the higher quality of lipid-rich copepods associated with colder ocean temperatures, but also by the opportunity to use sea ice as a resting substrate 15,75 .Loss of sea ice as a substrate for resting, foraging, and movement has demonstrated effects on energy balance in many sea ice-dependent species [76][77][78] .For instance, polar bears (Ursus maritimus) and narwhal (Monodon monoceros) show three-fourfold increases in locomotory costs in association with sea ice declines 78 .
T b of little auks was also sensitive to environmental conditions.However, mean T b was not related to T a .It is possible that a non-linear relationship could exist between T a and T b , which we could not capture given our modeling approach.On the other hand, variation in T b increased with T a across behavioral states, which could indicate that these cold-adapted birds face increasing challenges maintaining stable T b at higher temperatures, although, again, ambient temperature did not exceed the UCT of little auks in this study.T b also increased with RH across behavioral states.As RH rises, capacity for evaporative heat dissipation decreases, resulting in increases in T b or elevated thermoregulatory costs to maintain optimal T b 79,80 . In contrast, T b tended to increase with wind speed when birds were in flight and decreased with wind speed when birds were diving.High winds have been associated with increased energetic costs of flight for many avian species with a flapping flight mode 81 , whereas T b while foraging in cold waters could be further reduced by high winds via enhanced thermal conductance and heat loss.As for mean T b , wind speed and behavioral state interacted to predict between minute variation in T b .Specifically, between minute variation in T b increased with wind speed when birds were flying and diving, while varying little with wind speed in the other behavioral states.A possible explanation for these results is that little auks have difficulty maintaining thermal stability when foraging in turbulent seas and flying in challenging conditions induced by higher winds.In the context of climate change, results suggest that alterations in RH and T a may have implications for T b regulation that are independent of behavioral state.On the other hand, changes in wind patterns may have especially high costs during active periods, with increases in storm events associated with climate change potentially elevating energy expenditure.As a caveat, weather conditions measured at the colony were taken as a proxy of conditions experienced by birds across behavioral states.Thus, given that little auks can forage up to ~ 100 km from the colony, results regarding effects of weather conditions on T b should be interpreted with caution.Unfortunately, we did not have access to off-shore weather data on a fine temporal scale, and since birds were not GPS-tracked, we also had no way of knowing their precise locations during foraging trips.
We found no evidence for a relationship between Hg contamination and T b .However, our effective sample size for testing the relationship was low.In addition, little auks have lower Hg levels than many seabirds species that feed at higher trophic levels.Thus, our results regarding the relationship between Hg contamination and T b are preliminary, and further research is needed in this area, perhaps utilizing a different species with higher contamination levels.To our knowledge, there is currently no study documenting a link between Hg concentrations and T b in free-ranging animals.However, laboratory studies have demonstrated hypothermic responses to Hg exposure, for instance, in the mouse (Mus musculus) 82 .Hypothermic responses to contamination are hypothesized to reduce the toxicity of the chemical in the body 30 , but could create challenges for survival in dynamic thermal environments.We also observed weak, non-significant negative relationships between Hg www.nature.com/scientificreports/levels and variation in T b , which is inconsistent with the hypothesis that contaminated bird have more difficulty maintaining stable T b .Finally, there was a non-linear, bimodal relationship between time of day and T b .The highest values occurred in late morning and at night and the lowest in early morning and late afternoon.The pattern in T b observed did not parallel daily cyclicity in T a and RH, suggesting that it cannot be explained solely by diel variation in weather patterns.However, despite the fact that little auks in our population breed under 24-h of daylight, the pattern may reflect a combination of the timing of maximum solar radiation exposure, diel activity patterns, and/or underlying circadian rhythmicity in T b independent of activity.A past study on little auks found a regular rhythm of population attendance at the population level, perhaps linked to variation in predation pressure, which provides some foundation for expecting that T b could also display cyclic variation.However, this same study found little circadian rhythm in activity of individual little auks 83 .In contrast to mean T b , between minute variation in T b did not correlate with time of day.

Conclusions
T b of little auks fluctuated according to behavioral state and environmental conditions, which likely aids animals in optimizing energy balance while performing essential behaviors in complex environments.Although this plasticity is predicted to facilitate energy balance in the face of climate change, the dynamic nature of T b regulation also suggests that changing environmental conditions may significantly alter energy balance, or the behavioral and energetic strategies that must be adopted to achieve energetic homeostasis.Our data suggests that little auks use sea ice as a thermal refuge, resting on this substrate to allow T b to rebound after submersion in cold water and ingestion of cold prey items.If sea ice decreases due to warming temperature, thermoregulatory costs are forecast to increase as birds are instead required to rest on the water surface, which may force restructuring of foraging strategies.No relationship was found between T b of little auks and Hg concentrations, but our results are preliminary, and we call for more research on the effects of chemical contaminations on T b , especially in interaction with other environmental stressors.
shows a trace of T b through time for one focal individual.See Fig. S1-S7 for other birds.

Figure 1 .
Figure 1.Relationships between T b (°C) of little auks predicted from the GAMM and (a) time within the behavioral state (min), (b) wind speed (m/s), and (c) Relative humidity (%).Shaded regions show 95% CIs.

Figure 2 .
Figure 2. Body temperature (T b ) through time color coded with respect to behavioral state for one little auk (LIAK20EG19) from the Ukaleqarteq, East Greenland, population.Note rebounds in T b when on sea ice following declines while diving in cold Arctic waters.T b also generally increases with time flying and declines with time at the colony.

Figure 3 .
Figure 3. Relationship between T b (°C) of little auks predicted from the GAMM and time of day.Shaded regions show 95% Cis.

Table 1 .
Minimum adequate GAMM for body temperature (T b ; °C) in little auks as a function of behavioral state, weather conditions, and time.a Relative to diving.

Table 2 .
Minimum adequate GAMM predicting between minute variation in body temperature (|T b1 -T b1+1 |) (°C) in little auks as a function of behavioral state and weather conditions.

Table 3 .
Hg concentrations µg g −1 in dw and ww measured in the RBCs of focal birds breeding at Ukaleqarteq, East Greenland.