Overwintering Ecology of Head-started Blanding s Turtles (Emydoidea blandingii) in an Artificial Wetland Complex

Overwintering Ecology of Head-started Blanding s Turtles (Emydoidea blandingii) in an Artificial Wetland Complex by Shannon Ritchie A thesis submitted in conformity with the requirements for the degree of Master of Science Ecology and Evolutionary Biology University of Toronto Copyright by Shannon Ritchie 2017

Overwintering Ecology of Head-started Blanding s Turtles (Emydoidea blandingii) in a Artificial Wetland Complex Shannon Ritchie Master of Science Ecology and Evolutionary Biology University of Toronto 2017 Abstract The Blanding's Turtle (Emydoidea blandingii), a species at risk, is expected to become extirpated in urban environments without anthropogenic mitigation. Rouge National Urban Park, along with the Toronto Zoo, is working to create wetland habitat, and is supplementing the existing population with head-starts; turtles raised for two years at the Zoo then released into the wild. The selection and availability of appropriate overwintering sites can minimize turtle mortality. This research examined environmental conditions influencing site selection by head-starts in an artificial wetland complex, 2014-2017. Temperature was determined to significantly influence site use by the turtles. Head-starts preferred cold water (2.16 o C ± 0.37), close to the shore (< 2.5 m). Head-starts were found to successfully overwinter in the anoxic areas (0.92 mg/l), however, melt events increased access to atmospheric oxygen, suggesting dissolved oxygen is not a significant influence in areas with minimal ice cover. The habitat and behaviour of the head-starts was found to be different than turtles in other overwintering studies. ii

Acknowledgments I would first like to thank my thesis advisors Dr. Nicholas E. Mandrak and Dr. Marc W. Cadotte, and my committee member Dr. Njal Rollinson, who provided me guidance and steered me in the right the direction. I would like to thank the Toronto Zoo of experts who were involved in the validation and field assistance for this research project including Dr. Andrew A. Lentini, Curator of Reptiles and Amphibians, and the Adopt-A-Pond team Crystal Robertson, Paul Yannuzzi, Chevaun Toulouse and Kassie McKeown. I would also like to acknowledge my project partners including Parks Canada and the Toronto and Region Conservation Authority. I am grateful for their invaluable support. My research was funded by the Toronto Zoo, Parks Canada and NSERC Discovery grants to N.E. Mandrak and M.W. Cadotte. I have profound gratitude to my husband James Jurrius; thank you for providing me with unfailing support, free labor and continuous encouragement throughout this project. This accomplishment would not have been possible without him. Finally, to Bob Johnson, Emeritus Curator of Reptiles and Amphibians at the Toronto Zoo, whose passion for the conservation of species was a great inspiration for my spirit and, without his vision, this project and the Toronto Zoo head-starts would not exist. iii

Table of Contents Table of Contents...iv List of Tables...v List of Figures...vi Part 1: Background...1 General Introduction...1 The Blanding s Turtle...2 Habitat Requirements...3 Wild Blanding s Turtles Population in Rouge National Urban Park...4 Toronto Zoo Blanding s Turtle Head-starts...6 Study Area Wetland...9 Winter Pond Ecology...10 Overwintering Ecology of Blanding s Turtle...12 Research Objectives...15 Significance...15 References...17 Part 2: Methods...28 Cohort Size Comparisons...29 Vegetation Type and Cover...29 Overwintering Locations...30 Spatial Analysis...32 Pond Frequency...33 Site fidelity...33 Logger Placement...33 Winter 2015-2016...34 Winter 2016-2017...35 Temperature Analyses...38 HOBO Loggers to ibutton Comparison...38 Winter Environmental Parameters...38 Ice Period...40 Discriminant and Spatial Analysis...40 Permitting and Permissions...41 Results...42 Cohorts Size Comparison...42 Vegetation Type...44 Spatial Analysis...45 Pond Frequency of Use...48 Site Fidelity...49 Temperature Comparison...50 Environmental Parameters...54 Discriminant Analysis...63 Discussion...65 General Conclusion...76 References...78 iv

List of Tables Table 1: Blanding's Turtle densities at various locations throughout North America... 5 Table 2: Summary statistics of the size of head-starts turtle s released... 43 Table 3: Kruskal Wallis test for significance with a post-hoc Dunn pairwise comparison with Bonferroni correction for size of head-start turtles released 2014, 2015 and 2016... 44 Table 4: Results of the Kruskal-Wallis test with a post-hoc Dunn pairwise comparison test with Bonferroni correction test for differences in the spatial parameters o turtles released 2014, 2015 and 2016.... 46 Table 5: Summary statistics for spatial parameters of turtles released..... 47 Table 6: Summary statistics of temperatures at turtle overwintering sites and randomized sites in winters 2015/2016 and 2015/2017. HOBO loggers were used at turtles overwintering sites 2015/2016 and at random sites 2015/2016 and 2016/2017. ibutton loggers where attached to turtle carapace during winter 2016/2017.... 51 Table 7: Summary statistics of mean temperatures, December to March 2017, taken from ibuttons attached to overwintering head-starts adjacent to HOBO temperature loggers... 54 Table 8: Summary statistics of environmental parameters for turtle overwintering sites and randomize sites for winters 2015/2016 and 2015/2017. Both the Shapiro-Wilk test of normality and Leven s test of homogeneity of variance (HOV) were significant (p = 0.05)... 55 Table 9: Results of Kruskal-Wallis test with a post-hoc Dunn pairwise comparison with Bonferroni correction for environmental parameters at turtle overwintering sites and randomized sites for winters 2015/2016 and 2015/2017..... 62 Table 10: Number of days where ice covered ponds in **** wetland from 2014 to 2017, and corresponding dates of melt events.... 62 Table 11: Discriminant Analysis Tests of Equality of Group Means. Wilk s Lambda, Chi-square statistic and significance with Bonferroni correction (p = <0.0.008).... 63 Table 12: Standardized Canonical Discriminant Function Coefficients... 63 Table 13: Discriminant Analysis Classification Results Predicted Group Membership.... 64 Table 14: Comparison of environmental parameters of Blanding s Turtle overwintering sites compared to the head-start overwintering at **** wetland..... 73 v

List of Figures Figure 1: General overview of study area in Rouge National Urban Park.... 10 Figure 2: Tracking Blanding s Turtle head-starts in November 2015 using radio telemetry. Turtle with transmitter attached to shell.... 31 Figure 3: Overwintering locations of Blanding s Turtle head-starts released in 2014 determined by radio-tracking.... 32 Figure 4: Data logger set up.... 34 Figure 5: Data logger locations in **** wetland winter 2015/2016.... 35 Figure 6: Data logger locations in **** wetland winter 2016/2017.... 36 Figure 7: Head-start with white transmitter and blue ibutton. Turtle with ibutton attached with epoxy and colored for camouflage to minimize risk of predation.... 37 Figure 8: Water measurements were taken using a handheld probe YSI 6600 EDS V2 Multiparameter Water Quality Sonde.... 39 Figure 9: Frequency of pond size in **** wetland.... 48 Figure 10: Overwintering locations of the three cohorts of Blanding s Turtle head-starts... 49 Figure 11: Results of Kruskal-Wallis test comparing temperatures at head-start overwintering sites (n = 27) and random sites (n = 11) in 2015/2016, and head-start overwintering sites (n = 23) and random sites (n = 30) in 2016/2017.... 52 Figure 12: Monthly averages of all HOBO loggers at randomized sites (orange) compared to ibuttons at overwintering sites (blue) in winter 2016/2017... 53 Figure 13: Substrate depth comparing head-start overwintering sites (n = 26) and random sites (n = 13) in 2015/2016, and head-start overwintering sites (n = 28) and random sites (n = 33) in 2016/2017.... 57 Figure 14: Water depth comparing head-start overwintering sites (n = 26) and random sites (n = 13) in 2015/2016, and head-start overwintering sites (n = 28) and random sites (n = 33) in 2016/2017..... 58 Figure 15: Ice thickness comparing head-start overwintering sites (n = 26) and random sites (n = 13) in 2015/2016, and head-start overwintering sites (n = 28) and random sites (n = 33) in 2016/2017.... 59 vi

Figure 16: Dissolved oxygen comparing head-start overwintering sites (n = 26) and random sites (n = 13) in 2015/2016, and head-start overwintering sites (n = 28) and random sites (n = 33) in 2016/2017.... 60 Figure 17: The ph comparing head-start overwintering sites (n = 26) and random sites (n = 13) in 2015/2016, and head-start overwintering sites (n = 28) and random sites (n = 33) in 2016/2017.... 61 Figure 18: The potential distribution of overwintering sites in the **** wetland complex based on discriminant analysis.... 64 Figure 19: Monthly averages of temperature o C and snow on ground (cm) data from Buttonville Airport for each head-start s winter 2014/2015, 2015/2016 and 2016/2017, and from 1986 to 2016... 68 vii

Disclaimer Please note that the locations and maps in this report have been altered to protect the sensitive, at risk species found in the study area. Please contact the author if you require additional information. Part 1: Background General Introduction Chelonian species worldwide are in decline (Environment and Climate Change Canada, 2016). Their long life span and delayed sexual maturity make populations highly sensitive to disturbances (Environment and Climate Change Canada, 2016). In Ontario, seven of the eight turtle species are either endangered, threatened or of special concern (OMSTARRT Draft Recovery Strategy, 2007). The Blanding's Turtle is listed as a threatened species both nationally and provincially (Environment and Climate Change Canada, 2016). Restoration projects, such as the creation of **** wetland in Rouge National Urban Park (RNUP), mitigate habitat loss by adding significant habitat, including overwintering, basking, foraging and nesting sites to an area with minimal wetland features (Ritchie and Kula, 2011). After more than a decade of monitoring Blanding's Turtles in what is now RNUP, the Toronto Zoo s Urban Turtle Initiative recommended supplementation of the Park's Blanding s Turtle population with head-started hatchlings, with their first release into the park in 2014. Headstarting most often involves the collection and incubation of eggs and the subsequent captive care of hatchlings, where they are raised until they reach a size that would deter most predation. The supplementation program is part of a comprehensive approach, including habitat creation and education within Park boundaries. Monitoring and assessing habitat used by these head-starts reveals additional information about population dynamics as well as the effectiveness of release strategies. It is not certain if the head-start turtles can successfully use the **** wetland for overwintering, or if this type of restoration is suitable for the complete life cycle requirements of Blanding s Turtle (J. Phillips, Toronto Zoo, pers. comm., 2015). This research project proposes to identify the locations of 1

overwintering sites used by head-started Blanding s Turtles, and to determine if overwintering success is related to site selection based on environmental conditions. The Blanding s Turtle In Canada, Blanding's Turtle (Emydoidea blandingii) is historically found in pocket populations throughout southern Ontario, in some parts of Quebec, and in isolated populations in Nova Scotia (Environment and Climate Change Canada, 2016). Blanding's Turtle has a characteristically bright yellow chin and a highly domed carapace with yellow specks that ranges 12-27 cm in length as adults (Gillingwater, 2015). The Blanding s Turtle has been recently reassessed as Endangered federally (COSEWIC, 2016) and, under current federal and provincial legislation, is listed as a Threatened species (COSSARO, 2006; Environment and Climate Change Canada, 2016). Blanding s Turtle populations are highly vulnerable to anthropogenic stressors, such as habitat degradation (loss, fragmentation and/or pollution), being hit on roads, and increased predator abundance, such as racoons, in urban areas (Rubin et al. 2001; OMSTARRT Draft Recovery Strategy, 2007, Environment and Climate Change Canada, 2016). In urbanized areas, such as Toronto, it is at a high risk of extirpation (OMSTARRT Draft Recovery Strategy, 2007). Blanding s Turtle is long lived, on average living 40 years, and there have been many reports of turtles over 70 years of age (Congdon et al. 1993; Ernst and Lovich 2009). Blanding's Turtles reach sexual maturity at 14 25 years of age depending on the region, with females typically reaching sexual maturity at age 19 in Canada (Standing et al. 1999; Congdon et al. 2001). Maturity is largely based on turtle size; Blanding s Turtles living in the southern extent of the range grow faster and, therefore, mature more quickly than those living farther north (Edge et al. 2009). This extended maturation period contributes to the species decline in urban areas where a high percentage of the reproducing adults is killed by cars when crossing roads (OMSTARRT Draft Recovery Strategy, 2007). Populations are sensitive to even small losses of breeding adults (Environment and Climate Change Canada, 2016). Blanding s Turtles can lay 3-20 eggs in one clutch, but typically lay between 8-15 eggs. Not all sexually mature females will nest every year (Standing et al. 1999). Nest digging and egg laying 2

typically commence in June and the eggs incubate in underground nest chambers for 50-75 days, before hatching in late August to October (Ernst and Lovich, 2009). Blanding s Turtle gender is determined by its egg incubation temperatures, with higher temperatures resulting in predominantly female hatchlings and lower temperatures resulting in predominantly male hatchlings (Gutzke and Packard, 1987). Habitat Requirements Blanding s Turtle is considered a landscape species because it requires large complex areas of various wetland types (Congdon et al. 2011). Blanding's Turtles nest, forage and overwinter in distinctly different habitats, with activity dependent on season and age of the individual (Pappas and Brecke 1992; McMaster and Herman 2000; Arata et al. 2007; Ernst and Lovich, 2009; Caverhill et al. 2011; Congdon et al. 2011; Mui et al. 2017). Compared to other freshwater turtles found in North America, the Blanding s Turtle is known to occupy a relatively large home range (Ross and Anderson, 1990; Rowe and Moll, 1991; Pipegras and Lang, 2000; Joyal et al. 2001; Rubin et al. 2001; Mui et al. 2017). Turtles will select both wetland and upland habitats in response to change in breeding or hibernating period (phenology), prey abundance, temperature, and seasonal water levels fluctuation (Pappas and Brecke 1992; McMaster and Herman 2000; Arata et al. 2007; Ernst and Lovich, 2009; Caverhill et al. 2011; Congdon et al. 2011; Tan, 2016; Mui et al. 2017). Projects aimed at restoring diverse wetland habitats designed for Blanding's Turtles are, therefore, beneficial for other turtle and wetland species. In the spring and summer, juvenile Blanding s Turtles prefer large wetland complexes, often with deep, organic bottoms with dense vegetative cover (Arata et al. 2007; Ernst and Lovich, 2009; Caverhill et al. 2011; Mui et al. 2017). Juvenile Blanding s Turtles prefer shallow water with available basking sites near shore or in aquatic vegetation (Pappas and Brecke 1992; McMaster and Herman 2000; Arata et al. 2007; Caverhill et al. 2011). In the summer, adults frequent deeper waterbodies and drought-resistant pools (Arata et al. 2007; Caverhill et al. 2011). Waterbodies can include marshes, creeks, slow-moving rivers, lakes, and ponds with firm organic bottoms (Ross and Anderson 1990; Arata et al. 2007; Caverhill et al. 2011). Habitat should also have accessible movement corridors to other wetlands and nesting sites (Arata et al. 2007; Caverhill et al. 2011). 3

Blanding s Turtle nesting sites have specific soil, drainage and aspect requirements (Congdon et al. 2000). Nesting females most commonly choose sites with moist, but well-drained, sandy loam or organic soils with a southern or south-west aspect (Congdon et al. 2000). Like other Ontario turtle species, females have high nest-site fidelity and will return to the same area year after year (Congdon et al. 2000; Caverhill et al. 2011; Reid et al. 2016). In general, Blanding s Turtles use overwintering sites that have a soil depth that sits below the frost line so that turtles avoid freezing (Thiel and Wilder, 2010; Edge et al. 2009; Newton and Herman, 2009). Typically, overwintering sites are located in natural or artificial permanent wetlands, ponds, or marsh areas where there is enough unfrozen water under the ice in the winter to allow the turtle to move across the bottom (Edge et al. 2009; Newton and Herman, 2009; Thiel and Wilder, 2010; Markle and Chow-Fraser 2017). Wild Blanding s Turtles Population in Rouge National Urban Park In the Rouge National Urban Park (RNUP), the Blanding s Turtle population was found to mainly use cattail marshes in the spring, then ponds, open marsh habitats and rivers (Arata et al. 2007). Infrequently, RNUP Blanding s Turtles used drainage ditches, railway corridors, and hiking paths as corridors to move to adjacent wetlands or nesting sites (Arata et al. 2007). The Toronto Zoo has performed habitat evaluations and population viability analysis to validate the population viability (Ritchie and Kula, 2011; Lee, 2012). In 2011, all wetlands in RNUP were evaluated for their ability to support Blanding s Turtle, specifically examining suitability for an entire life cycle (i.e. availability of nesting sites, juvenile pools, adult foraging site, overwintering sites, habitat connectivity) (Ritchie and Kula, 2011). In 2012, a population viability analysis found that RNUP wetlands are able to support a greater number of Blanding s Turtles than the present population (Lee, 2012). Blanding s Turtles in the RNUP were also found to be less dense per hectare of wetland compared to other urban areas in North America (Lee, 2012) (Table 1), and that suitable and connecting habitat is minimal and heavily supported by restored wetland projects (Ritchie and Kula, 2011). 4

Table 1: Blanding's Turtle densities at various locations throughout North America (Lee, 2012) Location Density (individuals/ha) Reference Toronto (Rouge Marsh) 0.11 Arata et al. 2007 Toronto (Rouge Park) 0.04 Arata et al. 2007 Minnesota 0.35 Piepgras et al. 1998 Minnesota 0.47-1.45 Sajwaj et al. 1998 Wisconsin 27.5 Ross, 1989 Michigan 8.8-10.0 Congdon et al. 1986 Maine 3.9-5.9 Joyal, 1996 Missouri 55 Kofron and Schreiber, 1985 Blanding s Turtle presence in Rouge National Urban Park was confirmed in the 1980s (B. Johnson, Toronto Zoo, pers. comm., 2012). Turtle research in Rouge Park, now RNUP, has been conducted by the Toronto Zoo s Adopt-A-Pond (AAP) Wetland Conservation Programme since 1999, with radio-telemetry data first collected for Blanding s Turtles in 2005 (B. Johnson, Toronto Zoo, pers. comm., 2012). To analyze home ranges and movement patterns, in 2005 and 2006, five adult Blanding s Turtles were captured, radiotagged and tracked to the **** ****. Only one juvenile was captured and tagged, and was tracked to **** wetland (B. Johnson, Toronto Zoo, pers. comm., 2012). In 2007, the **** **** turtles were retired from the study, and tracking efforts were concentrated, north of the marshes, on Blanding s Turtles using the artificial **** wetland complex (B. Johnson, Toronto Zoo, pers. comm., 2012). Of the six known Blanding s Turtles historically tracked in RNUP, only one female Blanding's Turtle is known to have laid eggs (B. Johnson, Toronto Zoo, pers. comm., 2012). Through AAP research efforts, it was determined that the wild Blanding s Turtle population is inviably small, the RNUP contains minimal amounts of nesting and wetland habitat, nesting occurrences are minimal, and there is a precedence for restoration with head-starting to restore current population to sustainable levels (B. Johnson, Toronto Zoo, pers. comm., 2012). From 2009 to 2015, a juvenile living in the **** wetlands was re-captured and actively tracked, and a mature male turtle was captured, tagged and tracked in 2014-2015 (Yannuzzi, 2015). Both turtles successfully used the **** wetland as overwintering habitat for a number of years (Yannuzzi, 2015). Tracked turtles exhibited consistent patterns of dispersal from their overwintering sites in 5

May to their favoured basking and foraging sites (Arata et al. 2007; Yannuzzi, 2015). The juvenile showed fidelity to the same overwintering site, over multiple winters, but the older male turtle did not (Yannuzzi, 2015). Toronto Zoo Blanding s Turtle Head-starts The Toronto Zoo Blanding s Turtle head-start program incorporates management and stewardship goals outlined in the Ontario Multi-Species Turtles at Risk Recovery Team (OMSTARRT) draft Recovery Strategy for Species at Risk Turtles in Ontario (2007) and the National Recovery Plan for the Blanding s Turtle (2003). Protection of a population using the designation of protected areas alone has demonstrated to be insufficient to maintain a sustainable population (Bowne and Hecnar, 2007). This has resulted in a shift in protection targets from an ecosystem approach to a species approach that targets the protection of older adults, which highlights the need for more recruitment of mature reproducing turtles (Bowne and Hecnar, 2007). Nest success and hatchling survivorship is low in Blanding's Turtles mainly due to high predation rates. Predators, such as skunks, raccoons, coyotes, large frogs, crows, and the introduced opossum, will consume Blanding's Turtle hatchlings (Haskell et al. 1996; Herman et al. 2003; COSEWIC, 2005). In urban areas, predation of freshwater turtles is increased and can result in up to 100% mortality (OMSTARRT Draft Recovery Strategy, 2007). This increased predation leads to a chronic prevention of next-generation recruitment (Spinks et al. 2003; Seburn, 2007; Fordham et al. 2008). Compared to pristine landscapes, urban areas have a higher proportion of predators due to human influences such as the increase in food attractions (e.g. garbage, agricultural crops) (Mitchell and Klemens, 2000; Seburn, 2007; Fordham et al. 2008) and the removal of top predators due to restrictions on hunting and trapping (Barton and Roth, 2008; Prugh et al. 2009; Ritchie and Johnson, 2009). The mortality rate of turtles decreases as turtles grow larger. Larger turtles have harder and larger shells, which are more difficult for predators to break open and consume the internal soft tissues (Eckert et al. 1992; Congdon et al. 1993; Haskell et al. 1996; Heppel et al. 1996). Head-starting is the term for raising individuals in a controlled environment during their early and most vulnerable stage of life. Head-starting is used to increase population numbers that are 6

at risk of extirpation or extinction, such as in the Rouge National Urban Park (RNUP). Headstarting is a long-term initiative, requiring a large investment in staff, time and resources. However, head-starting can be a viable and effective method for increasing population stability in situations where egg/hatchling mortality rates are extremely high and cannot be addressed through traditional conservation measures (e.g. nest caging) and in situations where there are too few breeding adults to sustain juvenile recruitment (Frazer, 1992; Heppell et al. 1996; Environment and Climate Change Canada, 2012). Commonly, head-started hatchings are captivity raised until anywhere from their first week after hatching up to two years old (Murphy et al. 1994; Moore and Joosten, 2002; B. Johnson, Toronto Zoo, pers. comm., 2012; Burke, 2015). Each head-starting program has a number of unique factors, such as predation risk, climatic period, availability of resources, funding and in situ husbandry constraints, that determine the size suitable for release and type of release performed (e.g. soft vs. hard). (Murphy et al. 1994; Moore and Joosten, 2002; B. Johnson, Toronto Zoo, pers. comm., 2012; Burke, 2015). By eliminating egg mortality and skipping hatchling life in the wild, which has the highest mortality rates (Eckert et al. 1992; Congdon et al. 1993; Haskell et al. 1996), head-started turtles are expected to have a greater chance at surviving to adulthood. To be considered successful, a portion of head-started turtles are expected to become breeding adults, thus contributing to population recruitment (Eckert et al. 1992; OMSTARRT, 2007; Seburn, 2007). Head-starting turtles is a relative new conservation initiative with little research available on which method is most successful (Johnson, Toronto Zoo, pers. comm., 2012; Burke, 2015). Head-start projects require multiple releases of head-starts and long-term monitoring to determine success rates. A common limiting factor in assessing head-starting success is the long growth period before sexual maturity (Congdon et al. 1993). Most head-start projects are not old enough to evaluate reproductive success. However, there have been a number of successful nesting turtles reported in head-start projects involving the Western Pond Turtle (Actinemys marmorata) for west coast populations in the United States, and international programs with sea turtles including the Kemp Ridley's Turtle (Lepidochelys kempii) and Green Turtle (Chelonia mydas) (Eckert et al. 1992; Overtree and Collings, 2000). In addition, there are reports of successfully nesting, although rare, of head-started Wood Turtle (Glyptemys insculpta) (K. Bériault, OMNRF Parry Sound District, pers. comm., 2017). 7

A population viability assessment (PVA) was conducted in 2012 to determine how many juvenile turtles would be required to create a self-sustaining population of 150 adult Blanding s Turtles in RNUP (Lee, 2012). Based on the variables inputted into the PVA (i.e. demographic data from head-start Blanding s Turtle populations in Nova Scotia and wild Blanding s Turtle populations in Michigan), 40 head-start hatchlings with a 60% female-to-male ratio will be required to be released every year for the next 20 years to reach a stable target of 150 adults with at least 12 breeding pairs (Lee, 2012). Extending releases over 20 years ensures a higher probability of population persistence (Lee, 2012). The PVA will be revised at regular intervals to adaptively manage the Blanding s Turtle recovery program, as data specific to RNUP demographics will alter the outputs of the model (Lee, 2012). Since 2014, AAP has been supplementing the Blanding s Turtle population in the RNUP with head-started hatchlings as part of a comprehensive partnership project with the University of Toronto Scarborough, Toronto and Region Conservation Authority, Parks Canada, Environment Canada Habitat Stewardship Program, and the Ontario Ministry of Natural Resources and Forestry Species at Risk Stewardship Fund (J. Phillips, Toronto Zoo, pers. comm., 2015). This partnership includes habitat creation, road-mortality mitigation, stewardship education, and the ongoing research and monitoring of the Park s current population (J. Phillips, Toronto Zoo, pers. comm., 2015). All individuals were collected as eggs from a healthy population near Brantford, Ontario. In 2014, the first cohort of head-start Blanding s Turtles (n = 10) was released in the **** wetland, followed by the second cohort (n = 20) in summer 2015 and the third cohort (n = 36) in summer 2016. The **** wetland consist of six large (2257 to 8175 m 2 ) and deep ponds (>1.5 m), surrounded by multiple small ponds (<1000 m), and bordered by upland habitat and berms. The second largest pond (5802 m 2 ) is found on the east side of the complex. This pond is where the head-starts turtles were released and where the turtles are frequently found throughout the year. The eggs were incubated and raised at the Toronto Zoo s head-start turtle nursery for two years before being released. The head-starts were also conditioned for release to the wild in outdoor holding tanks from May to June of the release year (A. Lentini, Toronto Zoo, pers. comm., 2016). This process allowed the turtles to acclimate to natural temperature fluctuations 8

throughout the day and exposed them to a typical photoperiod with ambient sunlight (Tuberville, 2008). Their diet was also modified to simulate conditions in the natural environment such as, by providing natural prey item including fishes and earthworms and by encouraging foraging through randomized placement of the food. (A. Lentini, Toronto Zoo, pers. comm., 2016). In 2015, a soft-release enclosure was used in-situ to acclimatize the turtles to their new wetland habitat for a period of seven days before being fully released in to the wild. The soft-release enclosures allowed for further acclimation at the natural wetland site (Fischer and Lindenmayer 2000; Sedon et al. 2007; Teixeira et al. 2007). In 2016, the release cohort was split into two groups with half being placed in the soft enclosures and half released using a hardrelease technique (use of no in-situ enclosures) to determine if there was a different in the hatchling s behaviour and survival (A. Lentini, Toronto Zoo, pers. comm., 2016). Additional assessments of both methods are still required and there was no significant difference between hard-release and soft-release turtles in total distance travelled (p = 0.39), rate of movement (p = 0.42) and home-range size (p = 0.37) (Tan, 2016). Each Blanding s Turtle was outfitted with a PIT tag for future identification and a portion of each cohort is radio tracked using an affixed radio transmitter (A. Lentini, Toronto Zoo, pers. comm., 2016). Of the 10 head-start turtles released into the **** wetland in 2014, only one successfully overwintered in the 2014/2015 winter season. Of the 20 head-start turtles released in 2015, all successfully overwintered in the 2015/2016 winter season. In the winter of 2016/2017, of the 29 head-start turtles tracked to their overwintering site (mix of cohort 2013/released 2015 and cohort 2014/released 2016), all successfully overwintered. Study Area Wetland Please note that parts of this section have been removed to protect the sensitive, at risk species found in the study area. Please contact the author if you require additional information. **** wetland (Figure 1) is an artificial 9 ha wetland complex found in Scarborough, Ontario. The wetland complex is composed of many shallow pools and larger ponds dug into upland habitat. Native tree planting and wildflower seeds are added annually to the landscape by the Toronto and Region Conservation Authority (TRCA) and Friends of Rouge Watershed (FRW). The natural seed bank of the surrounding area is a significant contributor to the current 9

vegetation in the wetland. This wetland has been determined to be a functioning wetland ecosystem, supporting a diversity of species including small fishes, large mammals, and a number of at-risk bird and reptile species (Ritchie and Kula, 2011). Figure 1: General overview of study area in Rouge National Urban Park. Google Maps 2016. **** wetland is part of a poorly connected wetland complex, which includes stream corridors such as Little Rouge Creek, Petticoat Creek and Little Rouge River. **** wetland was chosen as the release site for the head-start turtles because, although poorly connected, this complex contains -removed information- and is close to travel corridors, such as the Little Rouge River, used by the local Blanding s Turtle population (Ritchie and Kula, 2011). - removed information - This site was also selected for its known seasonal use and overwintering success of two Blanding Turtles (Yanuzzi, 2015) and containing large and diverse juvenile habitat and potential adult habitat (Ritchie and Kula, 2011). Winter Pond Ecology Winter pond ecology is an important aspect of this study. In the fall, as water cools it becomes dense, reaching its maximum density at 4 o C. At maximum density, the 4 o C water layer will stink to the bottom of the pond. As water temperature continue to cool the upper layers of water in a pond, each layer will sink until the underlying water profile becomes homogeneous in 10

density and temperature (Marchand, 2013). Cold winter air cools the top layer of pond water below 4 o C. Mixing of this top layer into the bottom layers is prevented as water (<4 o C) is less dense, resulting in the layer of colder water to float on the top, where it will eventually turn into ice (Marchand, 2013). An ice layer creates a cap for the pond, where the water underneath is expected to maintain the same 4 o C temperature and density throughout the winter, provided the pond is deep enough (Marchand, 2013). Organisms under the ice use hibernation or physiological acclimation to adapt to the overall decline in energy exchange in the pond ecosystem (Marchand, 2013). Organisms more freely move throughout the pond due to the loss of density stratification, and oxygen levels start to decline as the ice cap prevents the exchange of gases (Marchand, 2013). Animals and the decomposition of waste products by bacteria start to use the limited amount of oxygen found under the ice (Marchand, 2013). Cold water can hold more oxygen, however, in a pond, the bottom generally contains the warmest water and is where vegetation decomposition occurs, which depletes oxygen reserves further. Very low levels of oxygen are not uncommon at the bottom of ponds (Marchand, 2013). If oxygen depletion becomes severe enough, organisms, less tolerant to low oxygen conditions that cannot move out of the low oxygen zone, will die resulting in winter kill. More resistant organisms, such as turtles, switch to anaerobic respiration, which divert energy from the breakdown of carbohydrates to lactic acid, known as glycolysis (Marchand, 2013). Some plants, such as Elodea (Elodea canadensis), can provide oxygen under the ice (Marchand, 2013); however, this is depends on the amount of light that can penetrate through the ice if snow cover is absent. Some organisms, such as fishes, can migrate to areas better suited to their physiological needs, such as upper levels of the water column with higher oxygen levels (Marchand, 2013). Pond edges may also provide oxygen refuges because edge ice often breaks open first during warm periods. This can also be problematic as ice can freeze solid into the substrate, increasing the risk of tissue freezing, leading to the need to migrate into deeper, less oxygen, warmer water (Marchand, 2013). 11

Overwintering Ecology of Blanding s Turtle In areas where cold winters occur, overwintering sites or hibernacula are an important part of a turtle s lifecycle as they are ectotherms. As temperatures cool, the rate at which turtles can metabolize food, move and respire is reduced (Hochachka and Guppy 1987; Boutilier et al., 1997). In winter, turtles are more vulnerable to active, endothermic predators, such as mammals (Brooks et al. 1991; Rollinson et al. 2008) and are limited in their ability to acquire resources needed to stay active. Hibernation is both the best option to avoid predation and maintain metabolic homeostasis, slowing the depletion of their physiological reserves. The composition and availability of refugia and hibernacula found in Blanding s Turtle habitat significantly influences home-range size (Pieprag and Lang, 2000; Innes et al. 2008; Millar and Blouin-Demers, 2011; Rubin et al. 2015). Home ranges were found to overlap in hibernaculum wetlands, suggesting overwintering sites are an important component of wetland complexes (Walston et al. 2015). Despite the abundance of other wetland habitat, this suggests overwintering sites may have greater importance in maintaining population numbers at a local scale compared to others factors, such as turtle size, age and sex (Walston et al. 2015). Other factors may also be important in maintaining populations, including resource availability, water levels, landscape type and climatic conditions (Pieprag and Lang, 2000; Innes et al. 2008; Millar and Blouin-Demers, 2011; Walston et al. 2015). Very few overwintering studies have been conducted on Blanding s Turtle. Characteristics used to identify overwintering sites generally lack detail and tend to focus on coarse-scale observations, such as wetland type and dominant vegetation type (Ross and Anderson 1990; Piepgras and Lang 2000; Joyal et al, 2001; Newton and Herman, 2009; Markle and Chow- Fraser, 2014; Walston et al. 2015); however, these properties are thought to have little influence in overwintering selection (Newton and Herman, 2009). A few more detailed studies found Blanding s Turtle used overwintering sites with a soil depth that sits below the frost line, which reduces freezing risk (Edge et al. 2009; Newton and Herman, 2009; Thiel and Wilder, 2010). Typically, overwintering sites are located in natural or artificial permanent wetlands. Wild Blanding s Turtles have been found to overwinter in permanent wetlands including swamps, ponds, marshes, and rivers, and also in non-permanent vernal pools (Fowle, 2001; Arata et al, 2007; Edge et al, 2009; Newton and Herman, 2009; Thiel and Wilder, 2010; Caverhill et al, 12

2011). These areas must have enough unfrozen water to allow turtles to move across the bottom of the wetland (Edge et al. 2009; Newton and Herman, 2009; Thiel and Wilder, 2010). Generally, Blanding s Turtles are found in overwintering sites similar to Painted (Chrysemys picta) and Snapping (Chelydra serpentina) turtles, which are anoxic environments with temperatures 0-5 C for an approximately four-month period (Arata et al, 2007; Rollinson et al. 2008; Newton and Herman, 2009; Edge et al. 2009; Thiel and Wilder, 2010; Caverhill et al. 2011). Both Painted and Snapping turtles are known to overwinter successfully in some, but not all areas, of the study site (Arata et al, 2007; J. Phillips, Toronto Zoo, pers. comm., 2015). Hibernation sites must prevent the freezing of tissue and exposure to fluctuating temperatures. Fluctuating temperatures can influence the metabolic rate of many species of pond turtles, which affects physiological reserves, such as blood-oxygen content (Reese at al 2002; Reese et al. 2004; Dinkelacker and Costanzo, 2004; Jackson 2004). The metabolic rate of turtles, including respiratory function, is controlled by temperature, which is reduced in cold temperatures (Hochachka and Guppy 1987; Boutilier et al., 1997). Metabolic and respiratory acidosis in overwintering turtles occurs when high levels of carbon dioxide are created in the blood though anaerobic respiration. High levels of carbon dioxide decrease blood ph and can eventually cause death (Gregory 1982; Ultsch and Jackson 1982). An increased risk of metabolic acidosis occurs in turtles when their metabolic rate is increased due to warmer temperatures in conjunction with an insufficient supply of oxygen, or when a stable temperature environment becomes hypoxic, such as a waterbody with a prolonged ice cover (Gregory 1982; Ultsch and Jackson, 1982). Similar to the Painted Turtle, the Blanding s Turtle may use extra-pulmonary gas exchange to absorb oxygen in low concentrations at low stable temperatures (Gatten, 1980; Ultsch and Jackson, 1982; Jackson et al. 2004). Temperature and dissolved oxygen are both important. If a site becomes hypoxic during winter, Blanding s Turtles could lessen the impacts of metabolic and respiratory acidosis by moving to colder temperatures to reduce their metabolism. Colder environments would, therefore, be more desirable; however, temperatures that induce tissue freezing (< 0 o C) must be avoided (Hertbert and Jackson, 1985; Edge et al. 2009). 13

Wild Blanding s Turtles tend to use thermally stable habitats with low water temperatures (Edge et al. 2009; Newton and Herman, 2009; Thiel and Wilder, 2010). There is no evidence that overwintering site selection is based on dissolved oxygen levels (Edge et al, 2009; Newton and Herman, 2009); however, overwintering sites appear to have ubiquitously low oxygen (Edge et al. 2009; Newton and Herman, 2009) supporting the idea that Blanding s Turtles tolerate anoxia environments (Ultsch 2006). Therefore, it is expected that the head-starts can successful survive in the anoxic environment of **** wetland; however, their success rate suggests additional factors may influence this success. In natural wetlands, Blanding s Turtle overwintering locations are in thermally stable, shallow water and between 7 to 50 cm in depth (Edge et al, 2009). Studies have found that surface ice at Blanding's Turtle overwintering sites is typically not permanent through the entire winter (Edge et al. 2009; Thiel and Wilder, 2010). Open surfaces replenish dissolved oxygen and allow turtles an opportunity to breathe if required (Arata et al, 2007; Edge et al, 2009; Thiel and Wilder, 2010; Caverhill et al, 2011). In Nova Scotia, in both natural and artificial wetlands, Blanding's Turtle overwintering sites have been found in soil below the frost line (Newton and Herman, 2009) and organic bottoms with spring-fed sites allow for better temperature regulation (Edge et al. 2009; Newton and Herman, 2009). Common sites include wetlands that have enough unfrozen water to allow for movement (>10 cm) and maintain a dissolved-oxygen range between 2.8 mg/l and 11.3 mg/l (Newton and Herman, 2009). The ph of the site has also been linked to hibernacula density with Blanding s Turtles (Power at al. 1994); however, this was in an acidic bog wetland. It is difficult to determine the body position and location of a turtle under the ice or when buried in the substrate. Opportunistic observations of Blanding s Turtle in the winter have observed turtles near the ice-water interface in vegetation (Newton and Herman 2009; B.P Caverhill pers. comm. 2011), in mud and leaf litter (McNeil, 2002) and at the bank edge (Power, 1989). Blanding s Turtles have also been observed mid-depth in the water column or on top of, or buried in, the bottom (Sexton 1995; Sajwaj and Lang, 2000). Previous radio-tracking initiatives by the Toronto Zoo s Adopt-A-Pond Wetland Conservation Progamme (AAP) found that both wild and head-started Blanding s Turtles stay relatively close to the shoreline (+/- 2m), throughout the winter (J. Phillips, Toronto Zoo, pers. comm., 2015). 14

Both artificial and natural sites have been found to be successful as overwintering locations for mature turtles (Newton and Herman, 2009). New sites and sites for juvenile turtles may be difficult to establish as Blanding's Turtles tend to return to previous wintering sites (Arata et al, 2007; Thiel and Wilder, 2010; Caverhill et al, 2011). Head-start turtles in this study have not had the opportunity to undergo natural selection in this area; therefore, successful overwintering sites are not yet well established. Research Objectives The objective of this study is to determine the characteristics of overwintering habitat used by head-started Blanding s Turtles in an artificial wetland, based on environmental parameters including ambient water temperature, dissolved oxygen content, substrate depth, water depth, cover, ph, ice thickness, ice cover period and soil type and depth. Overwintering sites were compared to randomized sites to determine if these environmental parameters influence overwintering habitat selection and success. Significance The recovery of the Blanding s Turtle has been mandated by the federal government under its responsibility to protect Canadian wildlife under the Accord for the Protection of Species at Risk Act, 2002. The recovery strategy for the Great Lakes/ St. Lawrence population, states that the conservation, management and restoration of habitat is designated as a high priority in the recovery of this species (Environment and Climate Change Canada, 2016). There is a long-term population and distribution objective to increase and maintain population abundance of the current Great Lakes/St. Lawrence population (Environment and Climate Change Canada, 2016). Head-starting is part of the broad strategies to meet these objectives and is currently deemed necessary in the recovery strategy for the Endangered Nova Scotia population to increase the population size found in Kejimkujik National Park (Environment and Climate Change Canada, 2012). Habitat loss through land conversion for agriculture and development and habitat degradation through urban influences are the highest risks to the survival of turtles in places such as, the 15

Greater Toronto Area (Environment and Climate Change Canada, 2016). Restoration projects, such as the creation of **** wetland, mitigate habitat loss by adding significant habitat, including overwintering, basking, foraging and nesting sites to an area with minimal wetland features (Ritchie and Kula, 2011). It is not certain if the head-start turtles can successfully use this site and if this type of restoration is suitable for the complete life cycle requirements of Blanding s Turtles (J. Phillips, Toronto Zoo, pers. comm., 2015). Blanding s Turtle critical habitat is identified as an area where a minimum of two Blanding s Turtles have been observed in the past 40 years, and/or one individual has shown site fidelity in the last 40 years, and/or suitable habitat for all or parts of a Blanding s Turtle complete life cycle (Environment and Climate Change Canada, 2016). Overwintering sites are identified as critical habitat because these sites are an essential part of the Blanding Turtle life cycle. There have been very few overwintering studies that focus on characteristics other than wetland type and vegetation composition (Ross and Anderson, 1990; Piepgras and Lang, 2000; Joyal et al. 2001; Walston et al. 2015; Markle and Chow-Fraser, 2017), and there are significant knowledge gaps in the habitat suitability of human-created wetlands (Reid et al. 2016). This research project identifies the locations of overwintering sites used by head-started Blanding s Turtles and determines if overwintering success is related to site selection based on environmental conditions. Insights gained through this research project would benefit future wetland restoration design and population supplementation programs, both required for the recovery of this species (Environment and Climate Change Canada, 2012; Environment and Climate Change Canada, 2016). 16

References Arata J., Cox, C., Ireland, D., Johnson B., Lawton, M., Marcks, H., Seabrooke, S., & Nelson, M. (2007). Turtle field research in the Rouge Park, 1999-2006: Snapping Turtle, Northern Map Turtle, and Blanding's Turtles. Toronto Zoo, Adopt-A-Pond Wetland Conservation Programme. Toronto, ON. Barton, B. T., & Roth, J. D. (2008). Implications of intraguild predation for sea turtle nest protection. Biological Conservation 141(8), 2139-2145. Bériault, K. (2017). Email Correspondence. Ministry of Natural Resources and Forestry. Parry Sound District, ON. Brooks, R. J., Brown, G. P., & Galbraith, D. A. (1991). Effects of a sudden increase in natural mortality of adults on a population of the common snapping turtle (Chelydra serpentina). Canadian Journal of Zoology 69(5), 1314-1320. Browne, C. L., & Hecnar, S. J. (2007). Species loss and shifting population structure of freshwater turtles despite habitat protection. Biological Conservation 138(3), 421-429. Boutilier, R. G., Donohoe, P. H., Tattersall, G. J., & West, T. G. (1997). Hypometabolic homeostasis in overwintering aquatic amphibians. Journal of Experimental Biology 200(2), 387-400. Burke, R. L. (2015). Head-starting turtles: learning from experience. Herpetology Conservation and Biology 10 (Symposium). Hofstra University. Hempstead. NY. 299-308. Caverhill B.P. (2011). Personal Interview. Toronto Zoo, Adopt-A-Pond Wetland Conservation Programme. Toronto, ON. Caverhill, B. P., Johnson, B., Phillips, J., & Nadeau, E. (2011). Blanding's Turtle (Emydoidea blandingii) and Snapping Turtle (Chelydra serpentina) habitat use and movements in the Oakland Swamp Wetland Complex, and their response to the Provincial Highway 24 Exclusion 17

Fence and Aquatic Culvert Ecopassage. Toronto Zoo, Adopt-A-Pond Wetland Conservation Programme. Toronto, ON. Congdon, J. D., Dunham, A. E., & van Loben Sels, R. C. (1993). Delayed sexual maturity and demographics of Blanding's Turtles (Emydoidea blandingii): Implications for conservation and management of long lived organisms. Conservation Biology 7(4), 826-833. Congdon, J. D., Kinney, O. M., & Nagle, R. D. (2011). Spatial ecology and core-area protection of Blanding s Turtle (Emydoidea blandingii). Canadian Journal of Zoology 89(11), 1098-1106. COSEWIC. (2005). COSEWIC assessment and update status report on the Blanding s Turtle (Emydoidea blandingii) in Canada. Committee on the Status of Endangered Wildlife in Canada. Environment and Climate Change Canada. Ottawa, ON. COSEWIC. (2016). COSEWIC assessment and update status report on the Blanding s Turtle (Emydoidea blandingii) in Canada. Committee on the Status of Endangered Wildlife in Canada. Environment and Climate Change Canada. Ottawa, ON. COSSARO. (2006). Species at Risk in Ontario list- Blanding's Turtle (Emydoidea blandingii). Committee on the Status of Species at Risk in Ontario. Queen s Printer. Ottawa, ON. Costanzo, J. P., Baker, P. J., Dinkelacker, S. A., & Lee, R. E. (2003). Endogenous and exogenous icenucleating agents constrain supercooling in the hatchling painted turtle. Journal of Experimental Biology 206(3) 477-485. Dinkelacker, S. A., Costanzo, J. P., Iverson, J. B., & Lee, Jr, R. E. (2004). Cold-hardiness and dehydration resistance of hatchling Blanding's turtles (Emydoidea blandingii): implications for overwintering in a terrestrial habitat. Canadian Journal of Zoology 82(4) 594-600. Eckert, S. A., Crouse, D., Crowder, L. B., Maceina, M., & Shah, A. (1994). Review of the Kemp s Ridley sea turtle headstart program. NOAA Technical Memorandum NMFS-OPR 3(11), 189. 18

Edge, C. B., Steinberg, B. D., Brooks, R. J., & Litzgus, J. D. (2009). Temperature and site selection by Blanding s Turtles (Emydoidea blandingii) during hibernation near the species northern range limit. Canadian Journal of Zoology 87(9), 825-834. Edge, C. B., Steinberg, B. D., Brooks, R. J., & Litzgus, J. D. (2010). Habitat selection by Blanding's turtles (Emydoidea blandingii) in a relatively pristine landscape. Ecoscience 17(1), 90-99. Environment and Climate Change Canada.(2012). Recovery strategy for the Blanding's Turtle (Emydoidea blandingii), Nova Scotia Population, in Canada. Species at Risk Act Recovery Strategy Series. Ottawa, ON. Environment and Climate Change Canada. (2016). Recovery Strategy for the Blanding s Turtle (Emydoidea blandingii), Great Lakes / St. Lawrence population, in Canada [Proposed]. Species at Risk Act Recovery Strategy Series. Ottawa, ON. Ernst, C. H., Barbour, R. W., & Lovich, J. E. (1994). Turtles of the United States and Canada Smithsonian Institution Press. Washington DC. Ernst, C. H., & Lovich, J. E. (2009). Turtles of the United States and Canada. Johns Hopkins University Press. Baltimore, MD. Frazer, N. B. (1992). Sea turtle conservation and halfway technology. Conservation Biology 6(2), 179-184. Fischer, J., & Lindenmayer, D. B. (2000). An assessment of the published results of animal relocations. Biological Conservation 96(1), 1-11. Fowle, S. (2001). Guidelines for protecting Blanding's Turtles and their habitats in Massachusetts. Natural Heritage and Endangered Species Program. Massachusetts Division of Fisheries and Wildlife. Westborough, MA. 19

Fordham, D. A., Georges, A., & Brook, B. W. (2008). Indigenous harvest, exotic pig predation and local persistence of a long lived vertebrate: managing a tropical freshwater turtle for sustainability and conservation. Journal of Applied Ecology 45(1), 52-62. Gatten, R. E. (1980). Aerial and aquatic oxygen uptake by freely-diving snapping turtles (Chelydra serpentina). Oecologia 46(2), 266-271. Gillingwater, S.D. (2015). Photo guide to the reptiles and amphibians of Ontario. St. Thomas Field Naturalist Club Inc. St. Thomas, ON. 28. Gregory, P. T. (1982). Reptilian hibernation. Biology of the Reptilia 13, 53-154. Gutzke, W. H., & Packard, G. C. (1987). The influence of temperature on eggs and hatchlings of Blanding's Turtles (Emydoidea blandingii). Journal of Herpetology 161-163. Haskell, A., Graham, T. E., Griffin, C. R., & Hestbeck, J. B. (1996). Size related survival of headstarted redbelly turtles (Pseudemys rubriventris) in Massachusetts. Journal of Herpetology 30(4), 524-527. Heppell, S. S., Crowder, L. B., & Crouse, D. T. (1996). Models to evaluate headstarting as a management tool for long lived turtles. Ecological Applications 6(2), 556-565. Herbert, C. V., & Jackson, D. C. (1985). Temperature effects on the responses to prolonged submergence in the turtle (Chrysemys picta bellii). II. Metabolic rate, blood acid-base and ionic changes, and cardiovascular function in aerated and anoxic water. Physiological Zoology 58(6), 670-681. Herman, T. B., Boates, J. S., Drysdale, C., Eaton, S., McNeil, J., Mockford, S. W., & Jones, C. (1999). National recovery plan for the Blanding s turtle (Emydoidea blandingii) Nova Scotia population. Environment and Climate Change Canada. NS. Hochachka, P. W., & Guppy, M. (1987). Metabolic arrest and the control of biological time. Harvard University Press. Cambridge, MA. 20

Innes, R. J., Babbitt, K. J., & Kanter, J. J. (2008). Home range and movement of Blanding's Turtles (Emydoidea blandingii) in New Hampshire. Northeastern Naturalist 15(3), 431-444. Jackson, D. C., Rauer, E. M., Feldman, R. A., & Reese, S. A. (2004). Avenues of extrapulmonary oxygen uptake in western painted turtles (Chrysemys picta belli) at 10 C. Comparative Biochemistry and Physiology Part A: Molecular & Integrative Physiology 139(2), 221-227. Johnson, B. (2012). Personal Interview. Toronto Zoo, Adopt-A-Pond Wetland Conservation Programme. Toronto, ON. Joyal, L. A., McCollough, M., & Hunter, M. L. (2001). Landscape ecology approaches to wetland species conservation: a case study of two turtle species in southern Maine. Conservation Biology 15(6), 1755-1762. Lee, R., Ritchie, S.D. Phillips, J., & Johnson, B. (2012). Rouge Park Blanding s Turtle Headstart Initiative. Toronto Zoo, Adopt-A-Pond Wetland Conservation Programme. Toronto, ON. Lentini, A. (2016). Personal Interview. Toronto Zoo, Adopt-A-Pond Wetland Conservation Programme. Toronto, ON. Marchand, P. J. (2013). Life in the cold: an introduction to winter ecology 3 rd Ed. University Press of New England. Lebanon, NH. Markle, C. E., & Chow-Fraser, P. (2014). Habitat Selection by the Blanding's Turtle (Emydoidea blandingii) on a Protected Island in Georgian Bay, Lake Huron. Chelonian Conservation and Biology, 13(2), 216-226. Markle, C. E., & Chow-Fraser, P. (2017). Thermal Characteristics of Overwintering Habitats for the Blanding s Turtle (Emydoidea blandingii) Across Three Study Areas in Ontario, Canada. Herpetological Conservation and Biology 12, 241-251. 21

McMaster, N. L., & Herman, T. B. (2000). Occurrence, habitat selection, and movement patterns of juvenile Blanding s turtles (Emydoidea blandingii) in Kejimkujik National Park, Nova Scotia. Chelonian Conservation Biology, 2, 602-610. McNeil, J. (2002). Distribution, movements, morphology and reproduction in a population of Blanding s turtles in an unprotected landscape in southwestern Nova Scotia M.Sc. Thesis, Department of Biology, Acadia University, Wolfville, NS. McNeil, J. (2011). Recovery Strategy for the Blanding's Turtle (Emydoidea blandingii), Species at Risk Act Recovery Strategy Series. Environment and Climate Change Canada. NS. McNeil, J. A., Herman, T. B., & Standing, K. L. (2000). Movement of hatchling Blanding s turtles (Emydoidea blandingii) in Nova Scotia in response to proximity to open water: a manipulative experiment. Chelonian Conservation and Biology 3(4), 611-617. McNeil, J., Herman, T., & Bourque, G. (2012). Population Viability Analysis on Two Populations of Blanding's Turtles in Nova Scotia. Parks Canada. NS. Millar, C. S., & Blouin-Demers, G. (2011). Spatial ecology and seasonal activity of Blanding's Turtles (Emydoidea blandingii) in Ontario, Canada. Journal of Herpetology 45(3), 370-378. Mitchell, J. C. & M. W. Klemens. (2000). Primary and secondary effects of habitat alteration. in M. W. Klemens, editor. Turtle conservation. Smithsonian Institution Press, Washington, DC. Moore, A. T., & Joosten, S. (2002). NWRA Principles of Wildlife Rehabilitation: The Essential Guide for Novice and Experienced Rehabilitators. National Wildlife Rehabilitators Association. Clearwater, MN. Mui, A. B., Caverhill, B., Johnson, B., Fortin, M. J., & He, Y. (2017). Using multiple metrics to estimate seasonal landscape connectivity for Blanding s turtles (Emydoidea blandingii) in a fragmented landscape. Landscape Ecology 32(3), 531-546. 22

Murphy, J. B., Adler, K., & Collins, J. T. (1994). Captive management and conservation of amphibians and reptiles. Society for the Study of Amphibians and Reptiles. Ithaca, NY. Newton, E. J., & Herman, T. B. (2009). Habitat, movements, and behaviour of overwintering Blanding s turtles (Emydoidea blandingii) in Nova Scotia. Canadian Journal of Zoology 87(4), 299-309. OMSTARRT. (2007). Recovery strategy for species at risk turtles in Ontario. Unpubl. report, Ontario Multi-Species Turtles at Risk Recovery Team. Ontario Ministry of Natural Resources. Peterborough, ON. Overtree, L., Collings, G.R. (2000) Successful headstart methods for the western pond turtle, Clemmys marmorata, in the Kern River Valley, California. 77th Annual Meeting of the American Society of Ichthyologists and Herpetologists. Kern River Valley, CA. Pappas, M. J., & Brecke, B. J. (1992). Habitat selection of juvenile Blanding's turtles, Emydoidea blandingii. Journal of Herpetology 26(2), 233-234. Pappas, M. J., Congdon, J. D., Brecke, B. J., & Capps, J. D. (2009). Orientation and dispersal of hatchling Blanding s turtles (Emydoidea blandingii) from experimental nests. Canadian Journal of Zoology 87(9), 755-766. Phillips, J. (2015). Personal Interview. Toronto Zoo, Adopt-A-Pond Wetland Conservation Programme. Toronto, ON. Piepgras, S. A. (1998). Summer and seasonal movements and habitats, and home ranges, and buffer zones of a central Minnesota populations of Blanding s turtles. M.Sc. Thesis, Department of Biology University of North Dakota. Grand Forks, ND. Piepgras, S. A., & Lang, J. W. (2000). Spatial ecology of Blanding s turtle in central Minnesota. Chelonian Conservation and Biology 3(4), 589-601. 23

Power, T.D. 1989. Seasonal movements and nesting ecology of a relict population of Blanding s turtle (Emydoidea blandingii (Holbrook)) in Nova Scotia. M.Sc. Thesis, Department of Biology, Acadia University, Wolfville, N.S. Power T.D., Herman T.B., & Kerekes J. (1994). Water colour as a predictor of local distribution of Blanding s turtles, Emydoidea blandingii, in Nova Scotia. Canadian Field-Naturalist (108), 17-21. Prugh, L. R., Stoner, C. J., Epps, C. W., Bean, W. T., Ripple, W. J., Laliberte, A. S., & Brashares, J. S. (2009). The rise of the mesopredator. Bioscience 59(9), 779-791. Reese, S. A., Crocker, C. E., Jackson, D. C., & Ultsch, G. R. (2000). The physiology of hibernation among painted turtles: the midland painted turtle (Chrysemys picta marginata). Respiration Physiology 124(1), 43-50. Reese, S. A., Jackson, D. C., & Ultsch, G. R. (2002). The physiology of overwintering in a turtle that occupies multiple habitats, the common snapping turtle (Chelydra serpentina). Physiological and Biochemical Zoology 75(5), 432-438.Reid et al. 2016 - have high nest-site fidelity Reese, S. A., Ultsch, G. R., & Jackson, D. C. (2004). Lactate accumulation, glycogen depletion, and shell composition of hatchling turtles during simulated aquatic hibernation. Journal of Experimental Biology 207(16), 2889-2895. Ritchie, E. G., & Johnson, C. N. (2009). Predator interactions, mesopredator release and biodiversity conservation. Ecology Letters 12(9), 982-998. Ritchie, S., & Kula, M. (2011). Identifying wetland restoration opportunities for recovery of species at risk turtles in Rouge Park. Toronto Zoo, Adopt-A-Pond Wetland Conservation Programme. Toronto, ON. Rollinson, N., Tattersall, G. J., & Brooks, R. J. (2008). Overwintering habitats of a northern population of Painted Turtles (Chrysemys picta): winter temperature selection and dissolved oxygen concentrations. Journal of Herpetology, 42(2), 312-321. 24

Ross, D. A., & Anderson, R. K. (1990). Habitat use, movements, and nesting of Emydoidea blandingi in central Wisconsin. Journal of Herpetology 6-12. Rowe, J. W., & Moll, E. O. (1991). A radiotelemetric study of activity and movements of the Blanding's turtle (Emydoidea blandingi) in northeastern Illinois. Journal of Herpetology 178-185. Rubin, C. S., Warner, R. E., & Ludwig, D. R. (2001). Habitat use and movements of radio tagged Blanding's turtles (Emydoidea blandingii) in a suburban landscape. Chelonian Conservation and Biology 4(1), 136-141. Sajwaj, T. D., & Lang, J. W. (2000). Thermal ecology of Blanding s turtle in central Minnesota. Chelonian Conservation and Biology 3(4), 626-636. Sajwaj, T. D., Piepgras, S. A., & Lang, J. W. (1998). Blanding's turtle (Emydoidea blandingii) at Camp Ripley: critical habitats, population status and management guidelines. Report to Nongame Wildlife Office. Minnesota Department of Natural Resources. Brainerd, MN. Sarkar, S., Sengupta, A., Chaudhuri-Sengupta, S., & Maiti, B. (2007). Thyroid responses to altered photoperiod in the soft-shelled turtle Lissemys punctata punctata Bonnoterre. Acta biologica Hungarica 58(1), 11-19. Seburn, D. C. (2007). Recovery strategy for species at risk turtles in Ontario. Unpubl. report, Ontario Multi-Species Turtles at Risk Recovery Team. Ontario Ministry of Natural Resources, Peterborough, ON. Seddon, P. J., Armstrong, D. P., & Maloney, R. F. (2007). Developing the science of reintroduction biology. Conservation Biology 21(2), 303-312. Sexton, O. J. (1995). Miscellaneous comments on the natural history of Blanding s turtle (Emydoidea blandingii). Transactions of the Missouri Academy of Science 29, 1-13. 25

Smith, D., (2013). Email Correspondence. Kejimkujik National Park and National Historic Site, NS. Spinks, P. Q., Pauly, G. B., Crayon, J. J., & Shaffer, H. B. (2003). Survival of the western pond turtle (Emys marmorata) in an urban California environment. Biological Conservation 113(2), 257-267. Standing, K. L., Herman, T. B., & Morrison, I. P. (1999). Nesting ecology of Blanding's turtle (Emydoidea blandingii) in Nova Scotia, the northeastern limit of the species' range. Canadian Journal of Zoology 77(10), 1609-1614. Tan, T. (2016) Using movement patterns of headstart Blanding s Turtles (Emydoidea blandingii) to evaluate release techniques. MEnvSc Thesis, Department of Physical and Environmental Sciences, University of Toronto, Scarborough, ON. Teixeira, C. P., De Azevedo, C. S., Mendl, M., Cipreste, C. F., & Young, R. J. (2007). Revisiting translocation and reintroduction programmes: the importance of considering stress. Animal Behaviour 73(1), 1-13. Thiel, R. P., & Wilder, T. T. (2010). Over-wintering characteristics of west-central Wisconsin Blanding's Turtles, Emydoidea blandingii. Canadian Field-Naturalist 124(2), 134-138. Tuberville, T. D., Gibbons, J. W., & Balbach, H. E. (2009). Estimating viability of gopher tortoise populations. Engineer Research and development Centre Construction Engineering Research lab. Champaign, IL. Ultsch, G. R. (2006). The ecology of overwintering among turtles: where turtles overwinter and its consequences. Biological reviews 81(3), 339-367. Ultsch, G, R., and Jackson D, C. (1982). Long-term submergence at 3 C of the turtle, Chrysemys picta bellii, in normoxic and severely hypoxic water: I. Survival, gas exchange and acid-base status. Journal of Experimental Biology 96(1), 11-28. 26

Walston, L. J., Najjar, S. J., LaGory, K. E., & Drake, S. M. (2015). Spatial ecology of Blanding s turtles (Emydoidea blandingii) in Southcentral New Hampshire with implications to road mortality. Herpetological Conservation and Biology 10(1), 284-296. Williams, D. A., & Osentoski, M. F. (2007). Genetic considerations for the captive breeding of tortoises and freshwater turtles. Chelonian Conservation and Biology 6(2), 302-313. Yannuzzi, P. (2015). Blanding s Turtle Headstart Monitoring: 2015 Report. Toronto Zoo, Adopt-A- Pond Wetland Conservation Programme. Toronto, ON. 27

Part 2: Methods The objective of this study is to determine the characteristics of overwintering habitat used by head-started Blanding s Turtles in an artificial wetland, based on environmental parameters. Overwintering sites were compared to randomized sites to determine if there is a difference between environmental parameters where overwintering selection and success occurred compared to other areas of the wetland. For the three cohorts released June 2014, 2015 and 2016 (n = 9, 20, 29), release weight (g) and body measurements including maximum carapace length (mm), maximum plastron width (mm) were compared to determine if there were significant differences between physical attributes of cohort years. The turtles were tracked using a R410 Receiver (Advanced Telemetry Systems, Inc, MN) from September to December to their overwintering sites, determined when turtles were found in the same location (+/- 2m) for a three-week period in fall (Thiel and Wilder, 2010). The total distance that the turtles moved to the overwintering sites from September to second week in December, the distance from overwintering sites to the shoreline, and distance to logs present was measured by mapping logger waypoints using ArcMap 10.3 (ESRI, 2017). Type of dominant vegetation used by the head-starts as cover, substrate depth, and pond size and use of (i.e. pond frequency) was recorded. To determine temperatures at overwintering sites in the winter of 2015/2016, 27 HOBO Pendant Temperature/Light data loggers were placed at known overwintering locations of the head-start Blanding s Turtles: 19 data loggers were placed where the 2015 cohort overwintered; 1 data logger was placed where the 2014 cohort overwintered; 2 were placed where the wild Blanding s Turtles overwintered in the winter 2014/2015; and, 5 were placed where the 2014 cohort overwintered in the winter 2014/2015. An additional 13 data loggers were placed at random points in the ponds using an interspersion grid to determine if other areas of the wetland had similar water temperatures to sites where the turtles overwintered. ibutton (ibuttonlink, WI) temperature loggers were attached to the turtle shells in the winter 2016/2017 to determine if there is a difference between turtle temperatures compared to static stations. Four HOBO Pendant Temperature/Light brick data loggers were placed adjacent to overwintering turtles and 34 randomized sites to determine if areas of the wetland had similar water temperatures to sites where the turtles overwintered. 28

To determine how environmental parameters differed at overwintering sites to randomized sites measurements were taken from January to March using a handheld probe YSI 6600 EDS V2 Multiparameter Water Quality Sonde (Yellow Springs, OH). Timing of the water measurements corresponded to an ice layer thick enough to support the researcher s weight and included dissolved oxygen, turbidity, temperature, ph and water depth and ice thickness. Ice melt was also recorded during site visits. Width of melt from the shoreline or vegetation was determined using a meter stick. Length of ice melt period was determined by recording the number of days when pond water was open to the atmosphere to when ice reformed at edge in correspondence to below 0 o C temperatures. To determine if a combination of environmental parameters influences overwintering selection, a discriminant analysis (DA) was used to determine if the location of overwinter sites could be predicted using the environmental parameters. Cohort Size Comparisons For the three cohorts released June 2014, 2015 and 2016 (n = 9, 20, 29), release weight (g) and body measurements including maximum carapace length (mm), maximum plastron width (mm) were taken prior to release. These physical attributes were measured to determine if there were significant differences between cohort years. If the parameter was normally distributed before or after log transformation based on the Shapiro-Wilk test, exhibited homogeneity of variance using a Levene s test, and had no outliers, then a one-way ANOVA with a post-hoc pairwise comparison least-significant difference (LSD) test with Bonferroni correction was conducted. If there were outliers, the data were not normal after log transformation if required, or variances were significantly different between groups, a non-parametric Kruskal-Wallis test was used to test significance and a post-hoc Dunn pairwise comparison with Bonferroni correction was conducted. Vegetation Type and Cover Dominant vegetation type at each site was identified using Chadde (2012) during logger placement in (November). Dominant vegetation type was also obtained by reviewing Toronto Zoo Adopt-A-Pond Wetland Conservation s summer technician field information from 2014 to 2017. Type of vegetation used as cover was recorded when a head-start was visually seen with 29

the cover or when radio-tracking pinpointed a head-start in/under a cover type ± 1 m. Cover included emergent vegetation, logs, vegetation debris, and root systems. Overwintering Locations Overwintering sites of the Blanding s Turtle (Emydoidea blandingii) head-starts were determined through a partnership with the Adopt-A-Pond Wetland Conservation Programme (AAP) at the Toronto Zoo. AAP radio-tracked 10-12 turtles from each cohort released from 2014 to 2016, to a maximum of 35 turtle tracked annually (Figure 2). The turtles were tracked using a R410 Receiver (Advanced Telemetry Systems, Inc, MN) three times a week from May to August, when they are most active, once a week from September to November, and once a month from December to April. In 2014, two wild Blanding s Turtles were also tracked to their overwintering sites. Prior to release, a 9.0 mm HPT9 Passive Integrated Transponder (PIT) tag (Biomark) is inserted subcutaneously into the left hind-leg pocket of all juvenile turtles (J. Phillips, Toronto Zoo, pers. comm., 2015). PIT tags were used to permanently mark Blanding's Turtles for future identification and population studies (J. Phillips, Toronto Zoo, pers. comm., 2015). All released turtles were also uniquely marked using a notch system (Cagle, 1939) via triangular wedges cut into the marginal scutes. Turtles were tracked with a R1680 model radio-transmitter weighing 3.6 g (Advanced Telemetry Systems, Inc, MN). Transmitters were fixed following the standard methods outlined in Standard Turtle Handling and Research Practices and Protocols (2013). Transmitters were equipped with short, flexible antennas and weigh 5% of the turtle s mass (Olfert et al. 1993). Transmitters were attached to the posterior portion of the carapace using PC-7 or JB Waterweld epoxy putty and coloured using a marker to match the device more closely the general colour of the carapace to minimize predation risk. Locations of the head-starts were determined by either using radio telemetry or by direct observation, and GPS coordinates were recorded. Turtle location by radio telemetry was determined when the receiver s gain was set to its most minimal level and the signal could be audibly identified when the antenna was placed vertically over the expected head-start turtle. 30

This technique is considered highly accurate, because the head-starts are found by hand directly under the vertical antenna when using this method for re-capture. Turtles are treated in accordance with the Canadian Council on Animal Care Guide to the care and use of experimental animals (Olfert et al. 1993). The winter hibernation season and sites were determined when radio-tracked turtles were found in the same location (+/- 2m) for a three-week period in fall (Figure 3) (Thiel and Wilder, 2010). Generally, the overwintering season for the AAP head-starts begins in late October in southern Ontario. However, due to the unseasonably warm weather experienced in October and November 2015 and 2016, the overwintering season/site selection was determined to be end of mid-november in 2015 and end of December 2016. Figure 2: Tracking Blanding s Turtle head-starts in November 2015 using radio telemetry (left). Turtle with transmitter attached to shell (right). 31

Please note that the locations and maps in this report have been altered to protect the sensitive, at risk species found in the study area. Please contact the author if you require additional information. Figure 3: Overwintering locations of Blanding s Turtle head-starts released in 2014 determined by radio-tracking. Each point is labelled with the turtle unique notch ID. Blue points represent a head-start turtle from the 2014 cohort. Yellow points represent a wild Blanding s Turtle. Spatial Analysis Distance from overwintering sites to the shoreline, and distance to logs present was measured by mapping logger waypoints taken in December 2015 and December 2016 on DigitaGlobe satellite imagery (06/01/2015) and using ArcMap 10.3 (ESRI, 2017). The total distance that the turtles moved to the overwintering sites was determined using the difference in distance between waypoints from the first week in September to second week in December. After second week of December all but two turtles stayed within their overwinter location +/-1m. 32

Pond Frequency Pond frequency is defined as the number of head-starts using each pond per winter. Pond frequency was determined based on the waypoints of turtle overwintering locations. Ponds available to overwinter was also compared based on visual observation of water levels in each pond during the fall of each field seasons. Site fidelity Site fidelity is defined as a head-start using the same area of a pond to overwinter in consecutive winters. This was determined using waypoints of turtle overwintering locations. Logger Placement To determine the water temperatures of overwintering and random sites within the wetland temperature and light data were logged using the HOBO Pendant Temperature/Light data loggers UA-002-64 (Onset, MA). HOBO loggers were used at turtles overwintering sites in 2015/2016 and at random sites in 2015/2016 and 2016/2017. ibutton loggers where attached to turtle carapace during winter 2016/2017. The data loggers recorded temperature ( C) from November to April. Each data logger was programmed to record 3500 readings (3.5Kb); therefore, approximately 1,750 readings of temperature were made during deployment. HOBO loggers were certified to be accurate ± 0.53 C with temperatures between 0 to 50 C, and ibuttons ± 1.00 C from -5 to 26 C. For the purpose of this study, the overwintering period for the Blanding s Turtles was determined to be 182 days, which resulted in a maximum rate of 8 readings per day or one reading every 3 hours in a 24 h period. The location of loggers placed beside overwintering turtles was adjusted to match change of position over a three-week period until turtles were found in the same location (+/- 2m) (Thiel and Wilder, 2010). The data loggers were fixed to a brick, the same height as the head-starts, using a zip tie, and placed on top of the pond substrate (Figure 4). A nylon rope was tied around each brick and secured to the shore to ensure the data loggers could be retrieved easily in the spring (Figure 4). All locations were referenced using a hand-held GPS (GARMIN Rhino). 33

Figure 4: Data logger set up. Right to left: HOBO Pendant Temperature/Light data loggers UA-002-64 (Onset) zip tied to bricks; bricks tied to string and secured by stake to shoreline for easier retrieval in spring. Winter 2015-2016 To determine temperatures at overwintering sites in the winter of 2015/2016, 27 HOBO Pendant Temperature/Light data loggers were placed at known overwintering locations of the head-start Blanding s Turtles: 19 data loggers were placed where the 2015 cohort overwintered; 1 data logger was placed where the 2014 cohort overwintered; 2 were placed where the wild Blanding s Turtles overwintered in the winter 2014/2015; and, 5 were placed where the 2014 cohort overwintered in the winter 2014/2015. To determine if other areas of the wetland had similar water temperatures to sites where the turtles overwintered, an additional 14 data loggers were placed at random points in the ponds using an interspersion grid. An interspersion grid was created following the procedures in the OMNRF Ontario Wetland Evaluation System (OWES). Interspersion grids use equal-sized squares overlaid onto a map of the wetland complex. In OWES, these grids are used to measure the presence of ecotones or edges that exist between different, distinct vegetation communities, for example, between wetland and upland vegetation. The number of times the grid lines intersect with a different habitat boundary provides an indication of habitat diversity (OMNRF, 2013). As outlined in the OMNRF OWES manual, the grid was created by determining the widest portion of the wetland complex and a line was drawn between these two points to represent the center line of the interspersion grid (Figure 5). The length of line was determined and divided by 12, which made 50 m 2 grid squares. This value represents the length of the sides of one square used to generate of the grid. This method creates a system of equal-sized squares over the wetland map. Each intersection of the grid was numbered and, with the use of an online random number generator (www.random.org/integers), 34

data logger locations were selected. If a random number was found at an intersection on upland habitat, another number was generated until a point was found within a waterbody. Winter 2016-2017 Please note that the locations and maps in this report have been altered to protect the sensitive, at risk species found in the study area. Please contact the author if you require additional information. Figure 5: Data logger locations in **** wetland winter 2015/2016. The blue points represent 19 data loggers where the 2015 cohorts overwintered; yellow points represent 2 data loggers placed where the wild Blanding s Turtles overwintered in 2014/2015; green points represent 5 data loggers placed where the 2014 cohort overwintered in 2014/2015; and white points represent 14 data loggers placed in at random points in the ponds throughout **** wetland. To determine if there is a difference between the actual turtle overwintering temperatures compared to the brick being used to mimic overwintering turtles, ibutton (ibuttonlink, WI) temperature loggers were attached to the turtle shells in the winter 2016/2017. HOBO Pendant Temperature/Light brick data loggers were placed adjacent to overwintering turtles and 34 randomized sites throughout **** wetland to determine if areas of the wetland had similar water temperatures to sites where the turtles overwintered. An interspersion grid was used to identify 35

these randomized locations using a smaller grid size (5 m 2 ) to provide a more distributed spread of these points. All randomizes brick placements took place the first week of November 2016 (Figure 6). Please note that the locations and maps in this report have been altered to protect the sensitive, at risk species found in the study area. Please contact the author if you require additional information. Figure 6: Data logger locations in **** wetland winter 2016/2017. The green stars represent four HOBO and ibutton data loggers where turtles with ibutton overwintered; green circles represent 24 turtles with ibuttons overwintered, green stars resented the orange pentagons represent 34 HOBO data loggers placed randomly throughout **** wetland using an interspersion grid. In the first week of September 2016, 28 radio tracked head-starts were tracked for recapture. Of the 36 originally tracked over the 2016 summer season, two were found dead as a result of predation, and five transmitters without turtles were found removed from the turtle shell either by a failure of the epoxy or by predation (Tan et al. 2016). To determine the external temperature of the head-starts at their overwintering sites, 28 head-starts were recaptured and an ibutton (ibuttonlink, WI) temperature logger attached to their shell on the opposite side to their transmitter (Figure 7). All captured turtles were measured and weighted. The ibuttons were waterproofed prior to attachment with two coats of liquid plastic (Plasti Dip International, MN), 36

demonstrated as an effective waterproof method in other studies (Edge et al. 2009; Paterson et al. 2012; Roznik and Alford, 2012). The ibuttons were attached using the same procedure to affix transmitters to the turtle shells outlined in Standard Turtle Handling and Research Practices and Protocols (2013). The ibutton and epoxy were coloured black for camouflage to minimize risk of predation. Both the transmitter and ibutton with epoxy together was approximately 10 g, which is approximately 5% of a + 200 g head-start turtle. Prior to the ibutton deployment, six ibuttons, three with dip and three without, were tested at above and below freezing temperatures to determine if the dip changed the logged temperature result. The dip was found to not significantly alter the temperature reading with a difference of <1.0 o C. This is similar to other studies where temperature difference was small (<1.3 o C) (Paterson et al. 2012; Roznik and Alford, 2012). All ibuttons where coated and attached in such a way that any effects would be equal in all ibuttons. The ibuttons were programmed to record readings from October to April at a rate of one reading every 3 hours in a 24 h period for a total of 8 readings per day, the same sampling schedule as HOBO Pendant Temperature/Light data loggers. Figure 7: Head-start with white transmitter and blue ibutton (left). Turtle with ibutton attached with epoxy and colored for camouflage to minimize risk of predation (right). 37

Temperature Analyses Average winter temperatures were calculated for each site from December 8 to March 31 of each year to examine differences between turtle overwintering sites vs. random sites for 2015/2016 and 2016/2017, and separately between site years 2015/2016 and 2016/2017. If the parameter was normally distributed based on the Shapiro-Wilk test after log transformation if required, expressed homogeneity of variance based on Levene s test, and had no outliers, a one-way ANOVA with a post-hoc pairwise comparison least significant difference (LSD) test with Bonferroni correction was conducted. If there were outliers, the data were not normal, or variances where significantly different between groups, a non-parametric Kruskal- Wallis test was used to test significance and a post-hoc Dunn pairwise comparison with Bonferroni correction was conducted. HOBO Loggers to ibutton Comparison To compare the temperatures taken by all HOBO loggers to those taken by all ibuttons in 2016/2017, a paired two-sample T-test assuming unequal variance was used. A non-parametric Mann-Whitney t-test was used to examine differences in average temperatures where HOBO loggers were placed beside overwintering turtles with ibuttons in 2016/2017; Sites 116, 122, 126, due to the data being not normality distributed. Winter Environmental Parameters To determine if overwintering sites differed from randomized sites at each data logger and turtle points, measurements were taken from January to March using a handheld probe YSI 6600 EDS V2 Multiparameter Water Quality Sonde (Yellow Springs, OH). Timing of the water measurements corresponded to an ice layer thick enough to support the researcher s weight. Dissolved oxygen (DO), turbidity, temperature, ph and depth were measured once per week, between 10:00 and 16:30, by creating a hole in the ice with a hatchet or hand auger (Figure 8). DO and temperature was measured at two points in the water column: 10 cm above the substrate and, 10 cm below the ice. However, due to the shallow nature of the wetland ponds (>0.3 m), caused by a severe drought in the summer 2016, there were few points where turtles 38

overwintered deep enough (>15 cm) to take both measurements. As a result, dissolved oxygen measured 10 cm above the substrate only was used for comparison in the statistical analysis. Where possible, DO recorded 10 cm below the ice was used to gain insight into DO stratification in deeper ponds relative to temperature. Figure 8: Water measurements were taken using a handheld probe YSI 6600 EDS V2 Multiparameter Water Quality Sonde. Water depths were determined using the sonde s depth measuring application and verified with a meter stick. Ice thickness was recorded using a meter stick. Ice condition was also recorded including if the ice was covered in snow, if the ice was opaque or clear, and if the ice was melting. Substrate depths were taken in October 2016 using a meter stick, and substrate type was determined using the procedure in the Ontario Ecological Land Classification (ELC). Pond surface area (m 2 ) was measured using DigitaGlobe satellite imagery (06/01/2015) using ArcMap 10.3. Winter environmental averages for each site were calculated based on measurements taken January to March of each year and used to examine differences in winter averages of site parameters between turtle overwintering sites and random sites for 2015/2016 and 2016/2017 combined, and separately between years 2015/2016 and 2016/2017. If the parameter was normally distributed based on to Shapiro-Wilk test, after log transformation if required, expressed homogeneity of variance using a Levene s test, and had no outliers, then a one-way ANOVA with a post-hoc pairwise comparison least significant difference (LSD) test with Bonferroni correction was conducted. If there were outliers, the data were not normal or variances where significant different between groups, a non-parametric Kruskal-Wallis test was 39

used to test significance and a post-hoc Dunn pairwise comparison with Bonferroni correction was conducted. Turbidity data were removed from this study due to equipment error as the data were inconsistent and ranged far above what would be expected for this wetland system (+500 NTUs). Ice Period Ice melt was recorded during site visits. Width of melt from the shoreline or vegetation was measured using a meter stick. Ice melt was measured at each turtle s overwintering site and visually compared to the rest of the pond to determine the ponds estimated melt width. Length of ice melt period was determined by recording the number of days when pond water was open to the atmosphere to when ice reformed at edge in correspondence to below 0 o C temperatures. Ice was considered reformed when ice was found frozen up to the shoreline/around emergent vegetation, was able to support slight pressure of a foot pushing on the ice surface (this was when the ice was approximately when it was 1 cm thick), and no visual areas were seen open to the atmosphere. When air temperatures fell below 0 o C, sites were visited at least every three days to determine if ice reformed. Differences in ice melt between each pond surveyed was also noted including location of melt (e.g. North side), size differences, colour differences (e.g. cloudy or clear), and difference in melt and re-freeze timing. Discriminant and Spatial Analysis To determine if a combination of environmental parameters influences overwintering selection, a discriminant analysis (DA) was used to determine if the location of overwinter sites could be predicted using the environmental parameters. Using Inverse Distance Weighted (IDW) interpolation, the DA equation was used to generate a map of potential overwintering location using ArcMap 10.3 (ESRI, 2017). 40

Permitting and Permissions All permits required for the research activities were acquired including OMNRF Endangered Species Act Protection or Recovery permit, OMNRF Wildlife Collection permit; Parks Canada Research and Collection permit, the Toronto Zoo and University of Toronto animal welfare reporting system. The study area, and most of the Rouge Park is in transition to becoming fully managed under Rouge National Urban Park. Access to the study area crosses three jurisdictions of land management. Property permissions where obtained from the City of Toronto, Parks Canada and The Toronto Region and Conservation Authority. 41

Results Cohorts Size Comparison The physical condition of three cohorts released summer of 2014, 2015 and 2016 were found to have carapace length ranging 78.4-108.9 mm, a plastron width ranging 44.0-63.1 mm, and weights ranging 86.2-225.0 g (Table 2). The parameters were found to be normally distributed and expressed homogeneity; however, due to the presence of outliers, a non-parametric Kruskal- Wallis test was used to test significance with a post-hoc Dunn pairwise comparison test with Bonferroni correction. The cohort released in 2014 was found to have a significantly smaller plastron (50.2 mm ± 3.5) and weight (131.4 g ± 26.6) compared to the cohort released in 2016 plastron (54.8 mm ± 3.6) and weight (166.1 g ± 26.9). There was no significant difference in the size of the 2015 cohort and the 2014 and 2016 cohorts (Table 3). 42

Table 2: Summary statistics of the size of head-starts turtle s released in 2014, 2015 and 2016. Both the Shapiro-Wilk test of normality and Levene s test of homogeneity of variance (HOV) were significant level (p = 0.05) Cohort Release Date 2014 2015 2016 N 10 21 36 Carapace Length (mm) Shapiro-Wilk normality 0.93 0.91 0.99 P(normal) 0.86 0.71 0.50 Min 78.40 85.00 81.30 Max 98.30 102.30 108.90 Mean 90.05 95.46 95.81 Std. error 2.06 1.07 1.05 Variance 42.30 24.25 39.47 Stand. dev 6.50 4.92 6.28 Outliers no yes no Levene's test 1.00 P(HVO) 0.372 Plastron Width (mm) Shapiro-Wilk normality 0.96 0.96 0.98 P(normal) 0.75 0.49 0.60 Min 44.10 48.30 47.80 Max 54.60 58.60 63.10 Mean 50.19 53.80 54.81 Std. error 1.10 0.68 0.60 Variance 12.18 9.71 13.07 Stand. dev 3.49 3.12 3.61 Outliers no no yes Levene's test 0.14 P(HVO) 0.869 Release Weight (g) Shapiro-Wilk normality 0.95 0.91 0.97 P(normal) 0.62 0.06 0.49 Min 86.20 108.20 112.00 Max 166.70 185.10 225.00 Mean 131.41 154.28 166.11 Std. error 8.40 4.58 4.48 Variance 705.97 440.39 722.10 Stand. dev 26.57 20.99 26.87 Outliers no no yes Levene's test 0.72 P(HVO) 0.490 43

Table 3: Kruskal Wallis test for significance with a post-hoc Dunn pairwise comparison with Bonferroni correction for size of head-start turtles released 2014, 2015 and 2016 Kruskal Wallis Carapace Length (mm) 5.61 0.061 P Dunn Pairwise 2014-2015 2014-2016 2015-2016 Plastron Width (mm) 10.72 0.005-17.10, p = 0.067-22.78, p = 0.003-5.68, p = 0.865 Release Weight (g) 11.54 0.003-14.89, p = 0.140-23.23, p = 0.003-8.35, p = 0.356 Vegetation Type The wetland complex ponds were found to have similar vegetation and soil characteristics. Broadleaf Cattail (Typha latifolia) and dense pockets of invasive Common Reed (Phragmites australis) where common throughout the wetland complex. Dominant submergent vegetation consisted of Common Hornwort (Ceratophyllum demersum), Canada Waterweed (Elodea canadensis) and Muskgrass Algae (Chara spp). White Pond Lily (Nymphaea odorata) and invasive Yellow Iris (Iris pseudacorus) were abundant in the east side ponds, specifically the large pond into which the head-starts were released. The ponds situated on the west side tended to be shaded by stands of Speckled Alder (Alnus incana) and willows (Salix discolor and Salix eriocephala). Deep organic soil was absent in all wetlands. Soil was silty clay on top of a compressed clay base, which was created during the wetlands construction. Prolonged periods of warm winter air temperatures above freezing (>5 days above freezing) resulted in a 1 to 2 m wide melt around all pond edges and emergent vegetation. During these melt events, a thick layer of ice remained floating in the centre of the ponds, which refroze to thinner edge ice when temperature returned to below freezing. The north side ponds of the complex were found to have larger edge melt compared to the south side ponds. During the winter, the treed west side ponds maintained a thicker and more intact layer of ice compared to the east side of the complex. In the 2016/2017 field season, four visual observations of live head-start turtles were seen buried in vegetation close to the surface where edge melt occurred. 44

The head-starts used a variety of cover to hibernate in or under, including clumps of cattails, alder root systems and floating logs. Radio-tracking indicated head-starts tended to favour edges (< 2.5 m from the shoreline) and did not move to the deep centres of ponds. In winter 2015/2016, head-starts spread out across the entire wetland complex, compared to their release location, using both large and small ponds to overwinter. The head-starts did not hibernate together (within 2 m of each other), with the exception of three pairs. Summer 2016 was dry and hot, which caused the majority of the ponds in the wetland complex to dry up or become shallow, warm and stagnant environments. Much of the aquatic vegetation perished in the dried ponds and shallow warm water. In fall 2016, ponds in the complex had very little vegetation and were mostly composed of barren silty clay bottoms. Sparse emergent vegetation pockets, such as cattails or shrubs, where found around the edges, which trapped leaf litter and other decaying vegetation. In the majority of cases, the head-starts overwintered in these pockets close to the edge of the pond. During the period when turtles moved to their overwintering spot (September 1 to second week in December), the ponds in the complex contained substantially low water levels. Turtle overwintering spots were restricted to the larger complex ponds. Specifically, the large east side pond, where the head-starts were released, supported 14 of the 28 overwintering turtles, under submerged brush. Many of the head-starts where found hibernating together (within 2 m of each other). Spatial Analysis The three cohorts released in the summers of 2014, 2015 and 2016 showed no significant differences in any of the spatial parameters between years (Table 4). Total distance moved to their overwintering sites from September 2 to the first week of December ranged 13.87-332.66 m, distance to of the overwintering sites to the shoreline ranged 0.26-11.49 m, and distance of the overwintering site to cover (i.e. being found under logs or in a root system) ranged 0.0-17.43 m (Table 5). 45

Table 4: Results of the Kruskal-Wallis test with a post-hoc Dunn pairwise comparison test with Bonferroni correction test for differences in the spatial parameters of turtles released 2014, 2015 and 2016. Kruskal Wallis P Dunn Pairwise 2014-2015 2014-2016 2015-2016 Distance to Cover (m) 2.017 0.365 Distance to Shoreline (m) 5.541 0.063 No significant difference between cohorts released 2014, 2015 and 2016 Total Distance Moved (m) 0.576 0.750 46

Table 5: Summary statistics for spatial parameters of turtles released in 2014, 2015 and 2016. Both the Shapiro-Wilk test of normality and Levene s test of homogeneity of variance (HOV) were significant (p = 0.05). Cohort Release Date 2014 2015 2016 N 9 21 28 Distance to Cover (m) Shapiro-Wilk normality 0.81 0.93 0.91 P(normal) 0.02 0.16 0.02 Min -2.00-2.00-2.00 Max 0.87 1.25 0.88 Mean 0.04-0.44-0.13 Std. error 0.30 0.22 0.15 Variance 0.78 1.03 0.63 Stand. dev 0.89 1.01 0.79 Outliers yes no no Levene's test 2.22 P(HVO) 0.118 Distance to Shoreline (m) Shapiro-Wilk normality 0.97 0.81 0.96 P(normal) 0.92 0.00 0.28 Min -0.12-2.00-0.59 Max 0.79 1.02 1.06 Mean 0.36 0.12 0.40 Std. error 0.10 0.14 0.07 Variance 0.09 0.41 0.14 Stand. dev 0.30 0.64 0.37 Outliers no yes no Levene's test 1.09 P(HVO) 0.343 Total Distance Moved (m) Shapiro-Wilk normality 0.84 0.95 0.98 P(normal) 0.07 0.34 0.93 Min 1.74 1.14 1.23 Max 2.26 2.30 2.52 Mean 1.90 1.80 1.86 Std. error 0.06 0.06 0.06 Variance 0.03 0.08 0.08 Stand. dev 0.17 0.29 0.29 Outliers yes yes no Levene's test 1.25 P(HVO) 0.296 47

Pond Frequency of Use Please note that parts of this section have been removed to protect the sensitive, at risk species found in the study area. Please contact the author if you require additional information. The wetlands consist of six large (2257to 8175 m 2 ) and deep ponds (>1.5 m), running through the centre of the complex. - Removed information - Between these six large ponds, there are 39 small (17 to 1492 m 2 ) shallowly dug ponds (<1.5m). Water levels were highest in March and April, when snow melt and rain accumulates, and shallowest in August when precipitation is minimal and evapotranspiration is highest. This wetland complex is situated on table land, where water is predominantly supplied by rain or snow melt. The majority of ponds available to head-start turtles in the **** wetland are under 1000 m 2. Turtles overwintering in 2015/2016 were found more frequently in ponds <1000 m 2. Turtles overwintering in 2016/2017 were found more frequently in ponds >5000 m 2. In summer 2016, the six largest ponds where able to hold water during a prolonged dry spell, whereas the majority of the smaller ponds dried up (Figure 9). Figure 9: Frequency of pond size in **** wetland (blue) during spring high water levels. Size of ponds used by head-starts to overwinter in 2015/2016 (red) and 2016/2017 (green). 48

Site Fidelity Ponds were used repeatedly for overwintering by different head-start turtles; however, site fidelity of individual turtles did not occur, with the exception of turtle with notch code 1-9, which repeatedly overwinter in the same small pond (Figure 10). Fidelity was defined as when a turtle overwintered within 2 m of its previous year s overwintering site. The majority of the turtles tracked over two consecutive overwinters were found to have overwintered in different ponds then previous winters. Please note that the locations and maps in this report have been altered to protect the sensitive, at risk species found in the study area. Please contact the author if you require additional information. Figure 10: Overwintering locations of the three cohorts of Blanding s Turtle head-starts released in 2014 (green), 2015 (yellow) and 2016 (red). 49

Temperature Comparison Comparing Sites Loggers Winter 2015/2016 Vs 2016/2017 In winter 2015/2016, of the 40 HOBO loggers places at randomized sites (n = 13) and head-start overwintering sites (n = 27), 38 were used to determine average site temperature between December and March. One logger failed over the winter (Site 40) and one logger was lost due to an anchor-site failure (Site 25). The mean temperatures based on the HOBO loggers at overwintering sites ranged 2.57-6.51 o C with a mean of 4.41 ± 1.01 o C. The mean temperatures based on the HOBO loggers at random sites ranged 3.75-6.54 o C with mean of 5.30 ± 0.92 o C. In winter 2016/2017, of the 28 turtles with ibuttons, 23 were used to determine the average overwintering site temperature immediately adjacent to the turtle between December and March. Failure of ibuttons occurred over winter for Sites 123, 146 and 149, and two turtles, Site 150 and 155, where unable to be retrieved during ibutton collection in April 2017. Mean head-start overwintering temperature ranged 1.56 _ 2.91 o C with a mean of 2.16 ± 0.37 o C. In winter 2016/2017, 38 HOBO loggers placed at randomized sites (n = 34) and at sites adjacent to headstarts (n = 4). Thirty loggers were used to determine mean winter temperature for December to March. Four loggers failed over the winter (Sites 102, 110, 123, 127), three where lost due to anchoring-string failure (Sites 107, 121, 129), and one was removed by beaver activity and found on land attached to its anchoring string in spring (Site 133). The mean overwintering temperatures ranged f3.75-6.92 o C with a mean of 5.20 ± 0.85 o C. (Table 6) A non-parametric Kruskal-Wallis test with a post-hoc Dunn pairwise comparison test with Bonferroni correction was used to test differences due to the presence of outliers, non-normally distributed data, and/or unequal variances (Table 6). There were significant differences between mean winter temperatures (p = <0.001). Head-starts in 2016/2017 (mean 2.16 ± 0.37 o C) were in significantly colder temperatures than head-starts at overwintering sites in 2015/2016, random sites 2015/2016 and random sites 2016/2017 (p = <0.001) (Figure 11). 50

Table 6: Summary statistics of temperatures at turtle overwintering sites and randomized sites in winters 2015/2016 and 2015/2017. HOBO loggers were used at turtles overwintering sites 2015/2016 and at random sites 2015/2016 and 2016/2017. ibutton loggers where attached to turtle carapace during winter 2016/2017. Both the Shapiro-Wilk test of normality and Levene s test of homogeneity of variance (HOV) were significant (p = 0.05). Year 2015/2016 2016/2017 Loggers ( o C) Turtle Random Turtle Random Overwintering Overwintering N 27 11 23 30 Shapiro-Wilk 0.98 0.94 0.97 0.98 P(normal) 0.87 0.54 0.74 0.87 Min 2.57 3.75 1.55 2.57 Max 6.51 6.54 2.91 6.51 Mean 4.40 5.30 2.16 4.40 Std. error 0.19 0.28 0.08 0.19 Variance 1.02 0.84 0.13 1.02 Stand. dev 1.01 0.92 0.36 1.01 Outliers no no no no Levene's test 15.90 P(HVO) <0.001 51

Figure 11: Results of Kruskal-Wallis test comparing temperatures at head-start overwintering sites (n = 27) and random sites (n = 11) in 2015/2016, and head-start overwintering sites (n = 23) and random sites (n = 30) in 2016/2017. Letters represent results of the Dunn test with Bonferroni correction. Temperatures with the same letter are not significantly different (p = <0.01) Box-whisker plot shows medians (horizontal line), Q1 and Q2 (25 th and 75 th percentiles, bottom and top of boxes, respectively), Q3 + 1.5 (Q3 -Q1) and Q1 1.5 (Q3-Q1) (bottom and top whiskers, respectively). 52

Comparing HOBO vs. ibutton Temperature Logger Winter 2016/2017 Using a paired two-sample T-test assuming unequal variance the 2016/2017 temperatures recorded by HOBO loggers were significantly warmer compared to temperatures recorded by ibuttons attached to head-starts (t = -3.1; p = 0.02) (Figure 12). The Shapiro-Wilk test confirmed normality (df = 4; F = >0.8; p = >0.055). Figure 12: Monthly averages of all HOBO loggers at randomized sites (orange) compared to ibuttons at overwintering sites (blue) in winter 2016/2017. Lines represent standard deviations. Due to the presence of outliers and data not normally distributed, a non-parametric, and/or unequal variances Mann-Whitney t-test was used on sites 116, 122, 126 where HOBO loggers were placed directly adjacent to overwintering head-starts with an ibutton. HOBO loggers were significantly warmer by >2 0 C than the ibutton loggers (M.W. < - 21.2; p = <0.001) (Table 7). 53

Table 7: Summary statistics of mean temperatures, December to March 2017, taken from ibuttons attached to overwintering head-starts adjacent to HOBO temperature loggers at Sites 116, 122 and 126. Shapiro-Wilk normality test was significance (p=0.05). Site 116 Site 122 Site 126 Logger ( o C) ibutton HOBO ibutton HOBO ibutton HOBO N 914 914 914 914 914 914 Transformation (eg. log, subtracted mean, removed trend or convert to ranks) Did not improve Shapiro-Wilk 0.8 1.0 0.9 1.0 0.8 0.9 P(normal) 2.74E-32 8.07E-17 3.28E-27 5.71E-14 5.65E-29 3.12E-17 Min 0 1.7 0.5 0.9 0 1.8 Max 9.5 9.3 10.0 10.1 9.5 9.2 Mean 2.2 4.4 2.3 4.8 2.4 4.5 Std. error 0.1 0.1 0.1 0.1 0.1 0.1 Variance 4.3 3.1 3.1 3.6 4.3 3.2 Stand. dev 2.1 1.7 1.8 1.9 2.1 1.8 Mann- Whitney U -22.6-25.3-21.2 P <0.001 <0.001 <0.001 Environmental Parameters A non-parametric Kruskal-Wallis test significance with a post-hoc Dunn pairwise comparison with Bonferroni correction was used to test to test for differences in the environmental parameters between overwinter and randomized sites (Table 8). Substrate depth and water depth were transformed using the log function to improve normality 54

Table 8: Summary statistics of environmental parameters for turtle overwintering sites and randomize sites for winters 2015/2016 and 2015/2017. Both the Shapiro-Wilk test of normality and Leven s test of homogeneity of variance (HOV) were significant (p = 0.05) Year 2015/2016 2016/2017 Loggers Turtle Random Turtle Random N 26 13 28 33 Substrate (cm) Transformation log log log log Shapiro-Wilk normality 0.88 0.95 0.93 0.97 P(normal) 0.01 0.59 0.07 0.40 Min 0.71 0.88 0.40 0.40 Max 1.71 1.48 1.48 1.71 Mean 1.06 1.19 1.00 1.14 Std. error 0.05 0.05 0.06 0.05 Variance 0.07 0.03 0.09 0.10 Stand. dev 0.26 0.18 0.30 0.31 Outliners yes no no no Levene's 1.13 P(HOV) 0.340 Water Depth (cm) Transformation log log log log Shapiro-Wilk normality 0.93 0.97 0.92 0.98 P(normal) 0.07 0.94 0.05 0.81 Min 0.00 1.00 1.34 1.03 Max 1.96 1.99 1.89 1.84 Mean 1.32 1.53 1.53 1.51 Std. error 0.08 0.08 0.02 0.03 Variance 0.18 0.08 0.02 0.03 Stand. dev 0.43 0.28 0.12 0.18 Outliners yes no no no Levene's 8.30 P(HOV) < 0.001 Ice Thickness (cm) Shapiro-Wilk normality 0.95 0.88 0.90 0.96 P(normal) 0.27 0.07 0.01 0.30 Min 11.43 13.97 16.00 13.72 Max 31.12 27.94 37.08 37.59 Mean 21.68 20.24 25.43 25.07 Std. error 0.99 1.05 1.31 1.16 Variance 25.24 14.45 47.76 44.35 Stand. dev 5.02 3.80 6.91 6.66 Outliners no yes no no Levene's 5.14 P(HOV) 0.002 55

Table 8: Continued Dissolved Oxygen (mg/l) Transformation Did not improve Not needed Not needed Not needed Shapiro-Wilk normality 0.87 0.94 0.98 0.96 P(normal) 0.00 0.44 0.80 0.29 Min 0.92 3.55 1.78 1.71 Max 11.25 12.70 11.63 15.69 Mean 4.54 7.28 6.71 6.99 Std. error 0.63 0.70 0.45 0.54 Variance 10.41 6.36 5.75 9.50 Stand. dev 3.23 2.52 2.40 3.08 Outliners no yes no yes Levene's 1.05 P(HOV) 0.374 ph Shapiro-Wilk normality 0.96 0.97 0.96 0.96 P(normal) 0.40 0.83 0.27 0.29 Min 1.26E-08 9.12E-09 2.00E-08 1.67E-08 Max 4.07E-08 2.57E-08 6.87E-08 6.61E-08 Mean 2.62E-08 1.68E-08 3.99E-08 3.71E-08 Std. error 1.55E-09 1.44E-09 2.28E-09 1.69E-09 Variance 6.21E-17 2.70E-17 1.46E-16 9.41E-17 Stand. dev 7.88E-09 5.20E-09 1.21E-08 9.70E-09 Outliners no no yes yes Levene's 15.90 P(HOV) <0.001 Of the 40 sites sampled in 2015/2016, 39 were surveyed throughout the winter; Site 13 had ice into the substrate throughout all survey attempts and was removed from the study. Of the 62 sites sampled in 2016/2017, 61 were surveyed throughout the winter; Site 140 had ice into the substrate throughout all survey attempts and was removed from the study. 56

Substrate depth at randomized sites ranged 2.54-30.50 cm with a mean of 15.32 ± 1.51 cm in 2015/2016 and 10.06 ± 1.98 cm in 2016/2017. Substrate depth at overwintering sites ranged 5.08-50.08 cm with a mean of 11.53 cm ± 1.80 in 2015/2016 and 13.70 cm ± 2.03 in 2016/2017. There were no significant differences in mean substrate depths between randomized and overwintering sites, and between winters (Table 9) (Figure 13) Figure 13: Substrate depth comparing head-start overwintering sites (n = 26) and random sites (n = 13) in 2015/2016, and head-start overwintering sites (n = 28) and random sites (n = 33) in 2016/2017. Letters represent results of Dunn test with Bonferroni correction. Substrate depth with the same letter are not significantly different (p = <0.01) Box-whisker plot shows medians (horizontal line), Q1 and Q2 (25 th and 75 th percentiles, bottom and top of boxes, respectively), Q3 + 1.5 (Q3 -Q1) and Q1 1.5 (Q3-Q1) (bottom and top whiskers, respectively). 57

Water depth at randomized sites ranged 10.10-98.10 cm with a mean of 34.25 cm ± 1.92 cm in 2015/2016 and 32.46 cm ± 1.53 in 2016/2017. Water depth at overwintering sites ranged 1.00-91.65 cm with a mean of 20.74 cm ± 2.67 in 2015/2016 and 34.17 cm ± 1.33 in 2016/2017. There were no significant differences in mean water depths between randomized and overwintering sites, and between winters (Table 9) (Figure 14) Figure 14: Water depth comparing head-start overwintering sites (n = 26) and random sites (n = 13) in 2015/2016, and head-start overwintering sites (n = 28) and random sites (n = 33) in 2016/2017. Letters represent results of Dunn test with Bonferroni correction. Water depth with the same letter are not significantly different (p = <0.01) Box-whisker plot shows medians (horizontal line), Q1 and Q2 (25 th and 75 th percentiles, bottom and top of boxes, respectively), Q3 + 1.5 (Q3 -Q1) and Q1 1.5 (Q3-Q1) (bottom and top whiskers, respectively). Outliers represented as circle outside boxes. 58

Ice thickness at randomized sites ranged 13.97 37.59 cm with a mean of 20.24 cm ± 3.80 cm in 2015/2016 and 25.72 cm ± 6.66 in 2016/2017. Ice thickness at overwintering sites ranged 11.43 37.08 cm with a mean of 21.68 cm ± 5.02 in 2015/2016 and 25.43 cm ± 6.91 in 2016/2017. There were no significant differences in mean water depths between randomized and overwintering sites, and between winters (Table 9) (Figure 15) Figure 15: Ice thickness comparing head-start overwintering sites (n = 26) and random sites (n = 13) in 2015/2016, and head-start overwintering sites (n = 28) and random sites (n = 33) in 2016/2017. Letters represent results of Dunn test with Bonferroni correction. Ice thickness with the same letter are not significantly different (p = <0.01) Box-whisker plot shows medians (horizontal line), Q1 and Q2 (25 th and 75 th percentiles, bottom and top of boxes, respectively), Q3 + 1.5 (Q3 -Q1) and Q1 1.5 (Q3-Q1) (bottom and top whiskers, respectively). Outliers (circles points outside boxes) and values further than 3 times the box height from the box are shown as stars 59