Winter in temperate regions is the coldest period of the year, defined by temperatures remaining around or below freezing point (0 ◦C), for several months (Williams et al. 2015; Studd et al. 2021). Animals in these areas must avoid prolonged exposure to the cold, which can lead to freeze injury or death. Consequently, temperate species display a variety of adaptations to avoid the threats of cold exposure. While some species survive freezing temperatures through physiological adaptations such as freeze tolerance and freeze avoidance, others avoid freezing temperatures altogether by migrating or selecting habitats buffered against freezing (Storey and Storey 2017; Studd et al. 2021). Freshwater environments can offer a refuge from freezing temperatures in the winter for certain aquatic and semi-aquatic species including many amphibians and reptiles (Ultsch 2006; Jackson and Ultsch 2010; Storey and Storey 2017). Overwintering in water, however, may present a substantial physiological challenge for air-breathing vertebrates. Freezing temperatures allow ice to form on surface waters, which can force aquatic and semi-aquatic species to be submerged for several months without access to atmospheric oxygen (Greaves and Litzgus 2007; Edge et al. 2009; Hampton et al. 2017; Studd et al. 2021). Most species of freshwater turtles——including the eight species found in Canada——escape freezing winter temperatures by overwintering in liquid water under the ice (Ultsch 2006). Some species, including the painted turtle (Chrysemys picta (Schneider, 1783)) and the snapping turtle (Chelydra serpentina (Linnaeus, 1758)), can survive several months submerged in anoxic water (Ultsch 2006). Other species, however, cannot survive an entire winter without oxygen and rely on extrapulmonary routes to extract dissolved oxygen (DO) from the water (Ultsch and Cochran 1994; Reese et al. 2001, 2003). The northern map turtle (Graptemys geographica (LeSueur, 1817)) is such a species. Laboratory studies indicate that northern map turtles die from metabolic acidosis after 45 days in anoxic water at 3 ◦C, making normoxic conditions a requirement to support aerobic respiration throughout the winter (Crocker et al. 2000; Reese et al. 2001). Northern map turtles overwinter communally (Ultsch et al. 2000; Fig. 1B) and show fidelity to their overwintering sites (Graham et al. 2000). This aggregation behaviour may indicate that overwintering sites with sufficient oxygen for this species are limited, and a similar suggestion has been made for other freshwater turtle species such as the spotted turtle (Clemmys guttata (Schneider, 1792)); Ultsch and Jackson 1995; Litzgus et al. 1999; Reese et al. 2001; Ultsch 2006; Rasmussen and Litzgus 2010). Northern map turtles appear to be responsive (i.e., conscious and actively moving) in winter and have been documented to show locomotor activity at temperatures near 0 ◦C (Evermann and Clark 1916; Graham and Graham 1992). This behaviour is uncommon for freshwater turtles at this time of year and has otherwise only been observed in the wood turtle (Glyptemys insculpta (LeConte, 1830)); Greaves and Litzgus 2007). The extent of northern map turtle winter activity and the reasons for it remain unknown; however, locomotor activity in the winter may be related to oxygen consumption. For instance, overwintering smooth softshell turtles (Trionychidae spp. (Fitzinger, 1826)) do “push-ups”, which are hypothesized to help turtles shed the oxygen-depleted boundary layer and ventilate their skin surface (Plummer and O’Neil 2019). Winter activity may thus be necessary for anoxia-intolerant species to meet their demand for oxygen. Studying turtles under the ice is logistically challenging and our understanding of their overwintering activity in nature largely comes from direct but short-term field observations (e.g., Graham and Graham 1992; Crocker et al. 2000) and radio-telemetry studies conducted from the surface of the ice without direct observations (Litzgus et al. 1999; Greaves and Litzgus 2007). These approaches, while insightful, have a limited potential for documenting activity throughout the winter. In radio-telemetry studies, activity is inferred from changes in location so locomotor activity not resulting in a measurable change in location is not documented. Moreover, the frequency of movement captured by radio-telemetry studies is limited by the frequency at which animals can be tracked. Finally, manual radio-tracking is limited to the portion of the winter during which the ice is safe to walk on potentially excluding important shifts in activity during the shoulder seasons. Advancements in biologging technology allow researchers to record detailed activity patterns in free-ranging animals over extended periods of time (Wilson et al. 2006; Gleiss et al. 2011; Fossette et al. 2012; Wright et al. 2014; Brownscombe et al. 2018) and thus offer opportunities to record the fine-scale behaviour of turtles overwintering in the wild. Here, we used biologgers equipped with tri-axial acceleration, temperature, and pressure sensors to measure locomotor activity, as well as temperature and depth use of northern map turtles overwintering in the wild. Tri-axial accelerometers produce recordings of acceleration in three spatial dimensions (see Halsey et al. 2009; Gleiss et al. 2011), which can then be used to calculate the overall dynamic body acceleration (ODBA) of animals, a proxy for locomotor activity, and field metabolic rate (Wilson et al. 2006; Gleiss et al. 2011; Halsey et al. 2011). Our first objective was to quantify locomotor activity, water temperature, and depth use of the northern map turtle before, during, and after the period of forced submergence under the ice. Documenting locomotor activity, as well as temperature and depth use, provides information on how anoxia-intolerant turtles meet their metabolic requirements during a significant part of their life. Our second objective was to test for differences among demographic classes (i.e., sex and age) in overwintering behaviour. The northern map turtle shows pronounced sexual size dimorphism. In our study population, adult females weighed between 800 and 3880 g, whereas males never exceeded 400 g. We hypothesized that behaviours during the overwintering period would differ between demographic classes because of these marked size differences. Size differences between demographic classes have been observed to translate into inherent differences in metabolic rates and oxygen needs (Graham and Graham 1992). Graham and Graham (1992) measured oxygen consumption in one adult male and two adult females submerged at 3 ◦C and found the male to consume 3.7–4 times less oxygen in absolute terms than the females. According to these measurements, adult females (i.e., larger turtles) should deplete DO more rapidly in their surroundings than males (i.e., smaller turtles); therefore, we hypothesized that this would translate into reduced activity or a tendency for females to seek colder temperatures to passively reduce metabolic and oxygen needs.
Summary
Winter in temperate regions is characteristically the coldest period of the year. Species in these regions adapt to freezing temperatures with physiological or behavioural mechanisms to mitigate the threats of cold exposure. For aquatic species, taking refuge under the ice minimizes the risk of experiencing freeze injury. The northern map turtle (Graptemys geographica (LeSueur, 1817)) is one species that overwinters under the ice of lakes and rivers. Here, we observed the behaviour of free-ranging map turtles at a known overwintering site throughout an entire winter using biologgers equipped with tri-axial acceleration, temperature, and depth sensors. We observed that map turtles maintain localized locomotor activity at the overwintering site continuously during the winter. The extent and patterns of locomotor activity and habitat use varied between adult females, adult males, and juvenile females. Adult females were observed at the shallowest depths, coldest temperatures and moved the least, whereas juvenile females were observed at the deepest depths, warmest temperatures and moved the most. All groups remained at temperatures near freezing (0.98–1.39 °C) and at average depths ranging from 1.34 to 1.7 m. These behavioural patterns are consistent with a strategy to survive the winter while remaining aerobic and likely reflect differences in physiological demands.
Methodology
Study Site
This research was approved and conducted in accordance with the Canadian Council on Animal Care Guidelines as administered by Carleton University, Protocol No. 113287, Permit No. 1096574. We conducted this study in Lake Opiniconin Ontario, Canada (44◦55 90N, 76◦32 80W). Lake Opinicon is a medium-sized (7.9 km2) shallow lake, averaging 2.5 m in depth, with a maximum depth of approximately 11 m (Feng et al. 2019). It does not stratify during the winter months. The surface of the lake typically remains frozen from late December to early April, apart from the lake’s eastern end, where it connects to the Rideau Canal, and two creeks at the southeastern end of the lake (Feng et al. 2019). It is estimated that there are over 1500 northern map turtles located within Lake Opinicon (1.9 turtles/ha; Bulté et al. 2010). Overwintering sites were previously identified along the shoreline of an island in Lake Opinicon using radio-telemetry (Carrière et al. 2009) and monitored as part of a mark-recapture study since 2004. In this study, we defined the overwintering period as the time between ice-on and ice-off at the overwintering site because it corresponds to the period that atmospheric oxygen is unavailable to turtles. Using two time-lapse cameras (TimelapseCam, Wingscapes, Birmingham, Alabama, USA) to take daily pictures of the water surface in the vicinity of the communal hibernaculum, we estimated this period to be 19-Dec-2020 to 27-Mar-2021. We defined two additional time frames within our study period to delineate times without ice-coverage: pre-ice in the fall (24-Oct-2020 to 18-Dec-2020) and post-ice in the spring (28-Mar-2021 to 12-Apr-2021). Between October 16th and 21st 2020, we captured 9 juvenile females, 17 adult females, and 14 adult males by snorkeling in the vicinity of the communal overwintering site. Juvenile females were included in the study because their body size is midway between adult females and adult males yielding three demographic classes for comparative purposes (i.e., adult males, adult females, and juvenile females). On each turtle, we fitted a biologger (22 mm × 45 mm × 8 mm, 11 g in air; Axy-5, TechnoSmArt, Guidonia Monticello, Italy) and a radio transmitter (15 × 8.2 mm, 1.5 g in air; NanoTag, Lotek, Newmarket, Ontario, Canada) to permit the recapture and retrieval of the biologgers. The biologgers and transmitters were taped together and epoxied to the right posterior edge of the carapace (Figs. 1A and 1B). The biologgers recorded acceleration (i.e., locomotor activity), water temperature (±2 ◦C), and depth (±5 cm). Acceleration was measured at sample rate of 10 Hz with an 8-bit resolution. Biologgers recorded on a schedule to preserve battery life to obtain measurements throughout the entire winter. Measurements were taken each day between 0:00 and 04:00 h and between 07:00 and 17:00 h. Turtles were relocated between the months of April and October 2021 using radio-telemetry and recaptured to remove the tags. Tags were removed within 12 h of capture and turtles were released back to where they were captured. Thirty-seven of the 40 tagged turtles (92.5%) were recovered in the spring and summer following their overwintering. Three turtles were not recovered, and four faulty biologgers did not yield data leaving a sample size of 13 adult females (209–250 mm maximum carapace length; 1076–1675 g), 9 juvenile females (142–185 mm maximum carapace length; 357–1388 g), and 11 adult males (126–141 mm maximum carapace length; 222–302 g).
Under-ice Videography
The accelerometry data indicated important locomotor activity during the ice on period. To qualitatively assess the extent of activity during this period, we installed an underwater camera (Eyoyo Underwater Fishing Camera) in January 2022 at the hibernation site. The camera was attached to a 1.1 kg diving weight and sat at approximately 15 cm above the substrate. The camera cables were fed through a 2.5 cm diameter, 30 cm long galvanized steel pipe. We visited the camera 11 times between 22-Jan-2022 and 13-Mar-2022 and made recordings on each occasion.
Environmental Monitoring
We installed six DO and temperature loggers (aquaMeasure DO, InnovaSea Systems Inc.) along a 500 m transect running parallel to the shoreline in the vicinity of the overwintering site. Each logger was stationed individually for a total of six measurement sites at water depths ranging from 1.8 to 4.6 m and recorded DO saturation (±5%) and temperature (±0.2 ◦C) every 2 h between 7-Nov-2020 and 19-Apr-2021. Each logger was approximately 1.5 m off the bottom of the lake. Additionally, on three separate days (5-Dec-2020, 13-Feb-2021, 20- Feb-2021), we measured DO saturation and temperature at 7–13 random locations using the same loggers. These measurements were also taken near the bottom of the lake, at depths ranging from 2.1 to 7.6 m. Measurements were used to create temperature and DO profiles of the overwintering site to observe how these parameters changed over time during the study period. Locomotor activity Locomotor activity was calculated as ODBA following the methods described in Brownscombe et al. (2018). Tri-axial acceleration data consisted of acceleration (g) in three axes (Ax = surge, Ay = heave, Az = sway with respect to the attachment location on the right posterior edge of the carapace), which was then corrected using a 2 s box smoother to remove static acceleration (i.e., gravity) from the dynamic acceleration (Shepard et al. 2008). The optimal smoothing interval was determined for 10 Hz based on the methods described in Shepard et al. (2008). Values of total daily ODBA were then obtained using these corrected values to produce the sum of absolute dynamic acceleration from all three axes (Ax, Ay, Az). These values were then observed over time throughout the study period to determine whether total daily ODBA during the overwintering period (19-Dec-2020 to 27-Mar-2021) was greater than 0g, which would confirm locomotor activity occurred. This was done using ggplot2 package to create graphs in R Studio (Wickham 2016). A background acceleration measurement was taken with an accelerometer left untouched for a 24 h period under the same configuration as the deployed biologgers. This produced a total ODBA of 0.02g, confirming that when no movement occurred, ODBA is near 0g.
Statistical Methods
To assess external influences on locomotor activity during the ice-on period, a linear mixed-effects model was fitted using the lmer function from the nlme package in R (Bateset al. 2015) with ODBA as the response variable where demographic class, water temperature, depth, and DO were considered as predictor variables, and turtle ID was used as a random variable. Water temperature and depth from the biologgers were included as measurements that are generally considered to influence locomotory activity (i.e., ODBA) and were measured simultaneously by the biologgers. Demographic class was included because of our hypothesis that the size difference between northern map turtles would influence locomotor activity. Measurements from the DO loggers were also initially included because it was expected locomotor activity might be driven by this species trying to meet their oxygen requirements under the ice. Model selection using Akaike Information Criterion (AIC) was used to produce a final model; the best model was selected based on an AIC difference greater than 2. This model was then followed up with an ANOVA to explore the differences among the significant predictor factors, and finally a Tukey’s post hoc test using the glht function from the multcomp package. Linear models were used to test the hypothesis that behaviour during the overwintering period differs between demographic classes. Model assumptions were checked following those described in Zuur et al. (2010). A linear model was fit using water temperature as the response variable and demographic class as the predictor variable to test our hypothesis that behaviours would differ between these classes. Similarly, a second linear model was fit with depth use as the response variable to determine whether depth use differed between the three demographic classes. Both linear models were fit using the aov function. Using the glht function from the multcomp package, Tukey’s post hoc tests were used to further explore the significant differences among the sexes (Hothorn et al. 2009). This same process was repeated for the pre-ice, ice-on, and post-ice periods to allow us to also test for behavioural differences in our demographic classes during these different time frames. p values < 0.05 were deemed statistically significant. All values have been reported as mean values ± standard deviation unless otherwise indicated. All analyses were conducted in R (4.1.2) via R Studio (2021.09.1).