571 2018 NORTHEASTERN NATURALIST 25(4):571–586 Characterizing Wood Turtle (Glyptemys insculpta) Populations at the Northwestern Periphery of the Species’ Range in Canada Jennifer Cross1, Robert Cross1,2, Derek Chartrand3, and Dean G. Thompson1,3,* Abstract - We report morphometric and demographic characteristics of 2 distinct populations of Glyptemys insculpta (Wood Turtle), at the northwestern periphery of the range in Canada, where they are designated as a species at risk. Our surveys of the 2 study watersheds (2012–2015) were assisted by a trained canine unit which was demonstrably more efficient than human crews in detecting Wood Turtles. We observed that both populations were large—214 and 114 uniquely marked individuals documented over time. We found no significant differences (P ≥ 0.05) in age structure, sex ratios, sexual size-dimorphism, body condition, number of observed mating attempts, or frequency and type of injuries between populations. We observed female-biased sex ratios in both populations (1:1.47 and 1:1.84, respectively) that were not attributable to sampling bias. Our data generally support the postulate of an inverse relationship between Wood Turtle body size and number of frost-free days or latitude. The general health of the 2 study populations was evidenced by the numerous large and reproductively mature individuals of both sexes, relatively high percentage of juveniles observed (average = 26%), and size-class frequency distributions that indicated sustained juvenile recruitment over several years in both watersheds. Our data suggest that high-quality forested watershed habitats, even at the northwestern extreme of the species range in Canada, can and do support large, healthy populations of Wood Turtles. Introduction The study of peripheral populations is considered to be important, particularly for conservation of species at risk (Koprowski et al. 2008). Populations at extremes of the range may exhibit demographic, genetic, density, and population-flux characteristics that differ from those of core populations (e.g., Holt et al. 2005, Kirkpatrick and Barton 1997, Williams et al. 2003). Better understanding of such differentials may provide insight into limiting factors, which can be used to develop conservation strategies that may help local populations to persist (Calder 1995, Walde et al. 2003). Glyptemys insculpta (Agassiz) (Wood Turtle) is a species at risk, listed internationally as endangered on the International Union for the Conservation of Nature (IUCN) Red List (Hilton-Taylor 2000, van Dijk and Harding 2011), as threatened at the national level by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC 2007), and as endangered in the province of Ontario by the Ontario Ministry of Natural Resources and Forestry (OMNRF 2016). Approximately 30% 1Algoma Highlands Conservancy, Sault Ste. Marie, ON P0S 1E0, Canada. 2US Forest Service, Sitka, AK 99835. 3Canadian Forest Service, Natural Resources Canada, Sault Ste. Marie, ON P6A 2E6, Canada. *Corresponding author - dean.thompson57@gmail.com. Manuscript Editor: Joseph Milanovich Northeastern Naturalist J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 572 of the global distribution of Wood Turtles is in Canada, wherein the range extends broadly to span the eastern provinces of Nova Scotia, New Brunswick, Québec, and Ontario (COSEWIC 2007). Within that broader geographic range, populations are generally patchy and discontinuous. A number of disjunct populations are known to exist within the boreal-forest transition ecozone, representing the northern extreme of the species’ range in North America. Although there have been numerous studies examining various aspects of Wood Turtle population status and ecology across their range in North America (reviewed by Ernst and Lovich 2009), Greaves and Litzgus (2009) identified only 6 previous studies (Daigle 1997, Greaves and Litzgus 2009, Harding and Bloomer 1979, Saumure 1992, Saumure and Bider 1998, Walde et al. 2003) on populations at the northern extent of the range for this species (~46°N latitude). In this paper, we report on the characteristics of 2 Wood Turtle populations occupying relatively undisturbed forested watersheds at the northwestern periphery of the range in Canada and which had not been previously described in the peer-reviewed literature. Our objectives were to characterize the demographic and morphological profiles of these independent populations and compare these characteristics to those of other northern populations. We also use the data to further examine postulates relating population characteristics to latitude and differential climatic regimes. Our study directly addresses recommendations for the initiation or expansion of studies on Canadian Wood Turtle populations (COSEWIC 2007). Finally, given the difficulties noted by Flanagan et al. (2013) associated with surveying for Wood Turtles, we also sought to compare the detection efficiency of a trained canine unit versus experienced human surveyors, with discussion regarding differences in derivative population metrics associated with these 2 methods. Field-site Description The study took place in 2 watersheds located in the district of Algoma, ON, Canada at approximately 47°N, 84°W and within the eastern boreal-forest transition ecozone. One of the main threats to Wood Turtle populations is poaching (van Dijk and Harding 2011); thus, we intentionally withheld the exact location of study areas as a protective measure, following the recommendation of Litzgus and Brooks (1996). Although it has been estimated that only 10% of the eastern boreal transition ecoregion remains as intact forest (Kavanagh et al., nd), our study areas may be considered relatively undisturbed compared to those supporting more southerly populations. Often, habitats in more southerly regions, although more productive and with longer frost-free foraging periods, are cumulatively impaired by anthropogenic land uses such as agriculture, urban development, paved-road networks, and utility corridors. While our 2 study watersheds are not completely devoid of these factors, their remoteness relative to urban and industrial land uses, and regulatory control on forest management within the riparian zone, serve to minimize potential for direct or indirect anthropogenic stress on the study populations. A prescribed Area of Concern (AOC) had previously been delineated by the Ontario Ministry of Natural Resources and Forestry (OMNRF) within each of our Northeastern Naturalist 573 J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 2 study watersheds. The AOCs were established in accordance with the provincial regulations for Wood Turtle habitat protection under the 2007 Endangered Species Act of Ontario. For the purposes of this investigation, the specific areas of study in watersheds 1 and 2 were restricted to within the AOC bounding the main river channels. Thus, in this study, we did not consider tributaries beyond 500 m upstream of their confluence with main river channels. The 2 study areas are geographically isolated by a minimum distance of 20 km, inclusive of areas with significant topographic relief and with no direct aquatic connections. Through multiple years of relatively ad hoc surveying in both areas, there have been no documented instances of uniquely marked turtles moving between the watersheds, suggesting that the populations are independent. Overview maps of the 2 study watersheds (Fig. 1) show the distribution of all currently known Wood Turtles within their respective AOC. Both study areas are relatively remote from industrial pollution sources and from substantial urban development, although there is a small village and sparse rural development around the southern portion of watershed 1. The AOC established for watershed 2 is segregated into 4 separate reaches where Wood Turtle occurrences have been documented; AOC 3 is the most recently designated as a direct result of multiple occurrences documented through our study. Historically, both watersheds were extensively logged. Commercial forestharvesting operations, governed under relatively stringent forest-management planning and regulations, continues in both watersheds to date. As a result, each study area is bisected by a main forest haul-road and a network of secondary roads and trails which, although used historically for forest-resource extraction, are now more commonly used by recreationalists using all-terrain vehicles (ATVs). There is substantial use of the gravelled, main forest haul-roads by both industrial vehicles (principally logging trucks) and personal vehicles including trucks, cars, and ATVs, largely supporting those engaged in hunting and fishing in the region. A portion of each watershed and the associated AOCs are intersected by a power-line corridor or right-of-way. In the case of the study area in watershed 1, the corridor runs parallel to the main river channel through a large portion of the AOC. In comparison, the right-of way in watershed 2 is oriented essentially perpendicular to the river and comprises only a very small portion of the overall AOC. The roads and power-line corridors, as well as the village area near AOC 3, may be considered to represent the most likely anthropogenic risks to these 2 Wood Turtle populations, through either poaching, road mortality, facilitated predation, or the cumulative effects of these factors. Methods Turtle sampling We carried out the study in compliance with OMNRF permits and approved Animal Research Protocol #278 (2012–2015), which was reviewed annually by the provincial Wildlife Animal Care Committee. From May through June 2012 and 2013, we conducted intensive surveys on foot and by canoe along a ~43.5-km Northeastern Naturalist J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 574 stretch of river in watershed 1, which was known from previous surveys to contain Wood Turtles. Similarly, from May through June in 2013 and 2014, we extensively surveyed stretches of the main river channel associated with AOCs 2 and 3 within watershed 2, having a combined riverine length of 85.6 km. We captured turtles by hand as human crews or the dog and handler (see below) walked the river, riverbanks, ephemeral pools, and tributaries within 100 m of the main river channel. Typically, experienced crews of 2–4 persons conducted surveys for a minimum of 2 d per week in each watershed and over a period of 3 consecutive weeks during the Figure 1. The distribution of all known male, female, and juvenile Glyptemys insculpta (Wood Turtle) within the Areas of Concern (AOCs) of (1a) watershed 1 and (1b) watershed 2 in Algoma District, ON, Canada. Note that AOC3 was established as the direct result of new occurrences documented through this study. Northeastern Naturalist 575 J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 early portion of the active season (May/June). Throughout the active season from 2012 to 2015, researchers who were conducting GPS and radio-telemetry studies on selected individual turtles as part of a broader research study made opportunistic turtle captures (Thompson et al. 2018). Over the course of the study (2012–2014), we had the opportunity to work with a specifically trained canine unit, consisting of an experienced handler (M. Buckner, Ontario Ministry of Natural Resources and Forestry, Sault Ste. Marie, ON) and a dog (Rebel) on 28 discrete surveying days. On each of these days, a human survey crew (2–3 trained and experienced individuals) as well as the canine unit intensively searched the same selected areas for Wood Turtles. Surveyors conducted searches between approximately 1000 and 1600 ETZ and focused on areas within 100 m of the stream edge, inclusive of all possible microhabitats including the stream itself. For each successive search area, the canine unit and human surveyor-teams randomly initiated their searches from opposite ends of the area or opposite sides of the stream to avoid potential observational biases that might occur if a group always searched a sector first or were commonly assigned to search sectors with higher probability of locating a turtle. Human crews and the canine unit exhaustively surveyed each search area. Surveyors ascribed each Wood Turtle found to either the canine or human search-team and entered the information into the database together with detailed observational data, including the microhabitat associated with the occurrence—river, open beach, short shrub, tall shrub, or mature forest. We summarized the data for each of the 28 d in which direct comparisons were possible to quantify search efficiencies (# finds/day) for the canine unit as compared to those of human search-teams. We used a triangular file to uniquely mark newly captured turtles by notching the marginal scutes of the carapace (Cagle 1939). We employed 305-mm digital calipers accurate to 1 mm (Neiko Tools, Taiwan) to measure and record the maximum carapace length (MCL) and maximum plastron length (MPL) for each turtle. Maximum plastron length as measured in this study using calipers, did not account for carapace convexity, nor plastron concavity (characteristic of mature males). We weighed individuals with a hanging digital scale accurate to 10 g (Berkley, Spirit Lake, IA). We considered turtles with MCL < 160 mm to be juveniles (Daigle 1997, Harding and Bloomer 1979, Lovich et al. 1990) and considered individuals with MCL > 160 mm as sexually mature (Walde et al. 2003). For turtles with MCL > 160 mm, we observed secondary sexual characteristics, such as plastron concavity in males and pre-cloacal tail length, to determine sex (Ernst and Lovich 2009). We palpated the inguinal leg pockets for shelled eggs to identify gravid females. We counted growth rings multiple times on both the plastron and the pleural scutes of the carapace, and recorded the mean value as a crude estimate of age. Previous studies (Brooks et al. 1997, Litzgus and Brooks 1998, Wilson et al. 2003) suggested that this method, despite being convenient and widely used in the literature, may often be inaccurate. We also recorded detailed descriptions of deformities and injuries of each turtle. We returned turtles to their original capture location after morphometric measurements were completed, a process that typically required ~0.75 h per individual turtle. Northeastern Naturalist J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 576 To estimate the density of Wood Turtles in the 2 watersheds, we measured the length (m) of the intensively surveyed main river channel using GIS techniques applied to high-resolution digital imagery of the study areas. We divided the total number of uniquely marked turtles observed in each watershed during the course of this study (2012–2015) by the appropriate riverine length and used that value to calculate density in terms of number of turtles per 100 m of river channel as recommended by other researchers (Daigle 1997, Greaves and Litzgus 2009). In the case of watershed 2, we excluded the river sections associated with AOC 1 and 4 (Fig. 1) from density calculations because they were not intensively surveyed as part of our study. Data handling and analysis We weighed and measured recaptured individuals; some individuals were sampled up to 19 times. However, within a field season, we used only the first recorded mass and length for each individual in calculations requiring a single measure. Complete morphometric measurements were not available for 7 individuals, therefore morphometric analysis is based on a total of 167 turtles—98 from watershed 1 and 69 from watershed 2 (Table 1). We compared mean male and female MCLs using the sexual dimorphism index (SDI) as described by Lovich et al. (1990), where a positive number indicates females are the lar ger sex: SDI = (mean MCL of larger sex) / (mean MCL of smaller sex) We used straight-line MCL rather than MPL to test for sexual size-dimorphism, thus avoiding potential bias associated with the formation of plastron concavity in mature males (Lovich et al. 1990). We made comparisons of mean length and mass in relation to sex and watershed using discrete Student’s t-tests. We used linear regression to examine the relationship between MCL and mass, and MCL and the number of frost-free days (FFD). We used residuals derived from linear regressions, comparing mass and MCL of non-gravid adult turtles, as an index to compare body condition between watersheds based on a 2-tailed Student’s t-test. We added mean MCLs to data compiled by Walde et al. (2003) from various published studies, to Table 1. Summary of the number of individually notched Glyptemys insculpta (Wood Turtle) captured from April 2012 to October 2015 and included in each analysis for 2 forested watersheds (WS1 and WS 2) of the Algoma District, ON, Canada. We excluded individuals with incomplete morphometry data due to injury or malformations from length calculations and gravid females (n = 7) from mass calculations. Mean ± 1 standard deviation of body mass (g), maximum carapace length (MCL; mm), and maximum plastron length (MPL; mm) for males, females, juveniles, turtles in watershed 1 (WS1), turtles in watershed 2 (WS2), and all turtles. Asterisk (*) indicates that values differed significantly between sexes (t-test; P < 0.05). Total n MPL (n) MCL (n) Mass (n) Mating events Male 49 192 ± 10 (49) 213 ± 13* (49) 1235 ± 198 (49) 18 Female 79 194 ± 11 (76) 200 ± 10 (76) 1177 ± 167 (69) 23 Juvenile 46 109 ± 36 (42) 122 ± 39 (42) 287 ± 206 (42) - WS1 102 189 ± 9* (73) 201 ± 11* (73) 1147 ± 164* (70) 25 WS2 72 199 ± 10 (52) 211 ± 13 (52) 1281 ± 179 (48) 16 All 174 193 ± 10 (125) 205 ± 13 (125) 1201 ± 182 (118) 41 Northeastern Naturalist 577 J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 determine if our data supported the inverse linear relationship between FFD and body size. We conducted statistical analyses in the freeware statistical software program R (Version 3.2.3; 2015-12-10; https://www.r-project.org). Results Of the 174 individuals captured from 2012 to 2015, 79 were females, 49 were males and 46 were juveniles (Table 1). Of the total, we observed 102 in watershed 1 and 72 in watershed 2, and included 61 and 59 individuals, respectively, which had not been marked in a study conducted ~10 y previously (Wesley 2006). Estimated densities of Wood Turtles were markedly greater in watershed 1 (0.49/100 m), compared to watershed 2 (0.13/100 m). Wood Turtle populations in both watersheds showed female-biased sex ratios (1:1.47 and 1:1.84, respectively), and the sex ratio did not differ significantly between watersheds (χ2 = 0.51, df = 1, P = 0.77). Although growth-ring counts are not considered an accurate means of estimating age, particularly for older, slow-growing turtles, the maximum discernable number of rings on carapace scutes that we observed was 27 in watershed 1 and 25 in watershed 2 (Table 2). The juvenile-to-adult ratios were 1.00:2.64 for watershed 1 and 1.00:3.00 for watershed 2, and did not differ significantly from each other (χ2 = 0.13, df = 1, P = 0.72). Overall, juveniles made up 26% of the sampled populations (watershed 1 = 27%, watershed 2 = 25%; Table 1). Table 3. Summary of canine versus human search-team efficiencies for capture of Glyptemys insculpta (Wood Turtle) in northern forested watersheds of the Algoma District, ON, Canada. Year of study 2012 2013 2014 Total Comparative days 13 11 4 28 Canine-team finds 44 10 12 66 Human-team finds 1 4 2 7 # days of no canine-team finds 2 6 0 8 # days of no human-team finds 12 7 3 22 Table 2. Carapace growth rings (mean ± SD) for female (F), male (M), and juvenile (J) Glyptemys insculpta (Wood Turtle) captured from April 2012 to October 2015 in 2 forested watersheds of the Algoma District, ON, Canada. Number of growth rings Watershed Sex n* Mean (± 1 SD) Minimum Maximum 1 F 43 15.9 ± 1.1 10 27 1 M 25 16.9 ± 1.4 12 24 1 J 24 5.4 ± 1.2 2 13 2 F 30 17.7 ± 1.2 7 25 2 M 15 16.8 ± 1.8 9 23 2 J 18 6.3 ± 1.3 2 9 *Of the 174 total individuals captured and measured for morphometric data, only 155, as shown in this table were used in growth ring analysis. The excluded individuals were either old adult turtles with heavily worn carapaces, or very young juveniles with indistinct growth rings. Northeastern Naturalist J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 578 Comparative results for canine versus human search-team efficiencies, spanning 3 study years (2012–2014) and involving 2 watersheds, are summarized in Table 3. Over the 28 d for which directly comparative data were available, the canine search-team accounted for 41% (n = 72) of the total turtle finds and an even higher proportion (56%) of the juveniles observed. The MCLs of captured turtles varied from 56 mm to 242 mm (Fig. 2). We found sexual dimorphism in adults with MCL greater than the 160 mm threshold; males had significantly greater mean MCLs than females (t = 1.97, df = 123, P < 0.001; Table 1). The mean MCL of adults in watershed 1 (mean = 201 mm, SD = 11, n = 73) was significantly less than that of adults in watershed 2 (mean = 211 mm, SD = 13, n = 52) (t = 1.98, df = 123, P < 0.001; Table 1). The overall calculated SDI values were -1.05 and -1.08 for populations in the 2 respective study watersheds, with an overall value of -1.06, indicating that males are slightly larger than females. The MPLs of captured turtles ranged from 50 mm to 218 mm. We detected no significant differences in mean MPLs between males and females (t = 1.98, df = 123, P = 0.53; Table 1). Paralleling the trend for MCL measurements, the mean MPL of adults in watershed 1 (mean = 189 mm, SD = 9, n =73) was significantly less than that observed in watershed 2 (mean = 199 mm, SD = 10, n = 52) (t = 1.98, df = 123, P < 0.001; Table 1). We compared body mass at first capture for adult males and non-gravid females (Table 1). Despite the sexual dimorphism observed in MCL, adult males and females did not differ in mass (t = 1.98, df = 116, P = 0.08; Table 1). However, Figure 2. Maximum carapace lengths (MCL), represented by 10-mm increments, for female (n = 76), male ( n = 49), and juvenile (n = 42) Glyptemys insculpta (Wood Turtle), captured from April 2012 to October 2015 in 2 forested watersheds of the Algoma District, ON, Canada. Northeastern Naturalist 579 J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 adults from watershed 2 weighed significantly more than adults from watershed 1 (t = 1.98, df = 116, P < 0.001; Table 1). As expected, we detected a significant relationship between mass and MCL in both males (r2 = 0.76, F= 148.38, df = 1,48, P < 0.001; Fig. 3) and females (r2 = 0.68, F = 141.53, df =1,68, P < 0.001; Fig. 3). The relationship between mass and MCL was significant for adults within watershed 1 (r2 = 0.629, F = 115.07, df = 1,68, P < 0.001) and watershed 2 (r² = 0.617, F= 74.209, df = 1,46, P < 0.001), but we found no significant difference between the body-condition index of the 2 populations ( t = 1.98, df=116, P = 1.00). During the period of investigation, we recorded a total of 30 mating events involving 41 different individuals, with 70% (21) of events occurring from mid- August to early October, and 50% (15) of events occurring at mid-day (1100 to 1400 ETZ). The smallest mating females observed in watershed 1 and 2 were 172 mm and 186 mm, respectively. The smallest mating males were relatively larger than females at 183 mm and 196 mm, respectively. The mean MCL of mating individuals (mean = 210.3 mm, SD = 8.6, n = 41) was significantly greater than the mean MCL in the population (mean = 205.1 mm, SD = 13.0, n = 125) (t = 1.97, df = 164, P = 0.02). We also observed significant sexual dimorphism (t = 2.02, df = 39, P = 0.01) in mating individuals; the mean MCL of mating males (mean = 214.0 mm, SD = 8.2, n = 18) was greater than that of mating females (mean = 207.4 mm, SD = 7.9, n = 23). Figure 3. Relationship between body mass (g) and maximum carapace length (MCL; mm) in non-gravid female (n = 69) and male (n = 49) Glyptemys insculpta (Wood Turtles) captured from April 2012 to October 2015 in 2 forested watersheds of the Algoma District, ON, Canada. Data fitted with linear regressions for males (long dash-dot line; y = 12.858 - 1497.8) and females (dotted black line; y = 13.249x - 1471.7). We excluded gravid females (n = 7) from this analysis. Females: r2 = 0.68, F1,68 = 141.53, P < 0.001; Males: r2 = 0.76, F1,48 = 148.38, P < 0.001. Northeastern Naturalist J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 580 The relationship between mean MCLs and number of frost-free days in our study fit well within the regression data previously compiled by Walde et al. (2003); the relationship including our data was statistically significant (F ig. 4). We noted signs of injury (e.g., toe, limb, and tail amputation) on 83 turtles (48%), many of which exhibited more than 1 injury. Seventy-three turtles (42%) were missing at least part of their tail, 23 (13%) were missing 1 or more toes, and 12 (7%) had a limb amputation (only 1 turtle had 2 limbs missing). We found 4 dead adult turtles (3F, 1M) over the course of the study. Necropsies performed on the 2 recently dead turtles indicated that predation was not the cause of death for either turtle. Discussion During the study period (2012–2015), we acquired detailed morphometric and demographic data from a total of 174 individual Wood Turtles captured from 2 independent populations occurring at the northwestern periphery of the species’ range in Canada. We observed 102 and 72 uniquely marked individuals in watersheds 1 and 2, respectively; the majority of these observations were individuals that had not been previously marked. The number of uniquely marked individuals observed Figure 4. Comparing results from the present study (located in the Algoma District, ON) to results compiled by Walde et al. (2003) for the relationship between the mean number of frost-free days and the mean maximum carapace lengths (mm) for male (n = 12) and female (n = 12) Glyptemys insculpta (Wood Turtle). Values from the present study are depicted with hollow symbols. Data were fitted with linear regression lines for males (long dash-dot line; y = -0.32x + 224.75) and females (dotted black line; y = -0.27x + 224.11). Males: r2 = 0.845, P < 0.001; females: r2 = 0.829, P < 0.001 (Walde et al. 2003). Males: r2 = 0.75, F1,10 = 29.55, P < 0.001; females: r2 = 0.72, F1,10 = 25.29, P < 0.001 (present study combined with Walde et al. 2003). Northeastern Naturalist 581 J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 during the course of this study indicates that both populations are quite large and are well within the upper values (31–191) recorded for 6 other populations at the northern limit of the range (≥46°N) (Greaves and Litzgus 2009, Harding and Bloomer 1979, Ross et al. 1991, Saumure and Bider 1998, Smith 2002, Walde et al. 2003). We point out that our surveys were not exhaustive, as constraints on personnel and financial resources limited surveying to the main stream channel and excluded tributaries, some of which are known to contain suitable Wood Turtle habitat. Further, the mixed survey methods that we employed in our study, involving both human survey-crews and a trained canine team, prohibited use of the data to estimate population size using mark–recapture methods. However, combining the number of uniquely marked individuals from our study with that conducted previously by Wesley (2006), indicates that totals of 214 and 114 individuals have been uniquely marked in watersheds 1 and 2, respectively. In a separate aspect of our study, we tracked 24 adult turtles in each watershed over a 3-y period using integrated telemetry techniques. During this period, we observed only 2 deaths among this subset of turtles, suggesting an annual adult mortality rate of (2/24)/3 = 0.03 for watershed 1, and (0/24)/3 = 0 for watershed 2. This result suggests limited mortality was likely to have occurred amongst the total 214 and 114 uniquely marked individuals documented in these study areas through time, and thus are the most reasonable estimates of population sizes. We recommend future surveys inclusive of all tributaries with potentially suitable habitat be conducted in these 2 watersheds using a canine unit to maximize detection efficiency and with the specific purpose of estimating the size of these 2 populations. In all years of study, the canine unit was markedly more efficient at detecting Wood Turtles than even well-experienced human crews. The dog and handler were particularly adept at finding juveniles, accounting for 56% of the juvenile finds overall. These results suggest that for Wood Turtle populations occurring in complex forested habitats similar to those studied here, human surveys are likely to underestimate the size of Wood Turtle populations overall and the proportional composition of juveniles as well. Such site conditions may be typical of the northern range extent in Canada. Some previous studies suggest that such potential bias exists where human surveys may have a higher probability of visually detecting females basking or nesting on open beach areas (Daigle 1997, Niederberger and Seidel 1999). Of the total finds made by the canine unit, only 7 (10.6%) occurred in areas classified as open beach, thus discounting the possibility that female-biased sex ratios observed in this study are an artifact of sampling bias. The significant female-biased sex ratios found in our study populations were consistent with previously published results in some cases (Harding and Bloomer 1979, Ross et al. 1991, Walde et al. 2003), but less than the mean of 1:1.97 calculated for 17 studies as detailed by Greaves and Litzgus (2009). Overall, our results clearly support the use of trained canine units in future survey efforts for Wood Turtles, aimed at either discovering new populations, assessing the extent of habitat use by known populations, or to provide more accurate data for estimating population sizes, and adult:juvenile and female:male sex ratios. The use of Northeastern Naturalist J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 582 trained canines would be particularly advantageous in cases where populations may be large, occur in complex, heavily vegetated habitats, or where the capture efficiency of human crews is unknown and seasonally variable. The demographic and morphometric characteristics of our study populations, which occur at essentially the same latitude and within the same climatic zone, were similar to one another. However, turtles captured in watershed 2 were, on average, greater in mass and length than those from watershed 1, perhaps reflecting differences in general habitat quality. These data provide evidence that peripheral populations fit well within the generalized description of Wood Turtles as medium-sized, forest-dependent, freshwater species (Litzgus and Brooks 1996, Smith 2002). As observed by others (e.g., Smith 2002), we also found sexual dimorphism in MCL, with males being slightly larger than females. The mean MCL of our turtles was equal to or greater than those in 6 other populations from areas approximating the northern range limit in Ontario, QC, Canada, and Michigan (Daigle 1997, Greaves and Litzgus 2009, Harding and Bloomer 1979, Saumure 1992, Saumure and Bider 1998, Walde et al. 2003), supporting the contention that mature adults of northern populations are typically larger than those in the south (Brooks et al. 1992). Sexual dimorphism was not as evident in terms of body mass, and the mean MCLs of mature adults in our study populations are near the upper range of variation expected for Wood Turtles in the north. Their relatively large size may reflect the high quality of habitat broadly available in our study sites, reduced thermal stress owing to moderating climatic effects of nearby Lake Superior, and the ubiquitous availability of open beaches and deep undercut river banks. Where such critical habitat elements are readily available, core-activity areas can be quite limited, with reduced energy expenditures required to move between different microhabitats to meet basking, nesting, or over-wintering needs. Reduced energy expenditures associated with locomotion could allow for greater energy conversion into body mass. Supporting this postulate, we have reported a detailed assessment of movement and probabilistic habitat-use by adult Wood Turtles that demonstrates multi-year site fidelity to small core-activity areas (Thompson et al. 2018). We observed a reasonable fit of our data to the multi-study, negative relation between MCL and number of frost-free days (Walde et al. 2003). However, inclusion of our data in this regression resulted in an overall lower r2 value because both males and females in our study areas were comparatively larger in size than those in other studies with similar numbers of frost-free days (i.e., roughly equivalent foraging-period time frames). Thus, our data supports the general postulate that Wood Turtles in regions with fewer frost-free days take longer to reach maturity as a consequence of reduced time in which high-quality forage is available (Walde et al. 2003). However, it also suggests that body size may vary with habitat quality, thus inducing greater variation about the regression reflecting the average trend. We suspect this to be the case for our 2 study sites, where general availability of high-quality habitat may result in relatively larger size and mass of Wood Turtles compared to other populations at the same latitude and with essentially the same period available for foraging, although differential mortality and recruitment among differing sites may also be contributing factors. We suggest that all of these Northeastern Naturalist 583 J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 possible explanatory factors are likely to be positively influenced by the specific protection and mitigation measures imposed through establishment of AOCs about our study sites. While the maximal number of growth-ring counts we observed suggest a moderately young population, we concur with others (e.g., Brooks et al. 1992) who consider growth rings to be an unreliable basis for estimating age, particularly for older turtles. We are more confident about using less equivocal measures of carapace length or body mass to discuss age-class demographics. Both of our study populations are characterized by relatively high proportions of juveniles (26%) as compared to the average of 19.6%, which can be calculated for 6 other studies of northern populations based on data summarized by Greaves and Litzgus (2009). The size-class frequency distributions we observed indicate sustained juvenile recruitment over several years, which we interpret as an indication of healthy populations. Mean adult size for our populations, based on either mass or MCL, fall well within the range of natural variation for Wood Turtles at this latitude (Greaves and Litzgus 2009). Although body size of adults in watershed 2 was greater than that of adults in watershed 1, both populations consisted of numerous relatively large and thus reproductively mature adult turtles. The mean MCL value calculated from our observations exceeded the 95th percentile range of variation from studies of either northern or southern populations, and lends further support to the previously described non-linear relationship between body size and latitude (Greaves and Litzgus 2009). MCL and body mass are strongly and directly correlated; thus, the same general assessment is likely to pertain to body mass. The relatively large body mass of individuals observed in our study may allow turtles to acquire and retain greater energy reserves conferring a survival advantage during periods of inactivity and low forage-availability. As noted by Greaves and Litzgus (2007), overwintering success may be a particularly important factor for Wood Turtles at the northern extreme of their range. Body sizes of individuals observed mating were greater than the mean body sizes in the populations. Furthermore, mean MCL among mating individuals was greater for males than females, supporting a sized-based dominance hierarchy (Kaufmann 1992) in which larger males are better able to find females and also avoid being displaced during copulation. A lesser size-based dominance hierarchy in females combined with an overall female-biased sex ratio (M:F = 1:1.61) may suggest that the active mating population of turtles in both watersheds includes a greater proportion of females. The density of Wood Turtles in watershed 1 (0.49 turtles/100 m) was almost 4-fold greater than that in watershed 2 (0.13 turtles/100 m). Both values are substantially less than the value reported for another northern Ontario population (1.3; Greaves and Litzgus 2009) and bracket the value of 0.2 turtles/100 m reported for a population at approximately the same latitude in Quebec (Walde et al. 2003). Although our data generally support the postulate of an inverse relationship between latitude and Wood Turtle population density (Greaves and Litzgus 2009), the degree of variation observed in density estimates for populations occurring at essentially Northeastern Naturalist J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 584 the same latitude, together with varying methods of reporting densities, erodes our confidence in such a relationship. Certainly, we concur with Greaves and Litzgus (2009) that standardization in reporting densities is required to make meaningful comparisons, and that the number of individuals per 100 m of river-channel length is a simple, reasonable basis for this purpose. The proportion of injured turtles in our study populations (48%) was less than that reported by Brooks et al. (1992), who observed a 60% injury rate in a more southerly population occurring in a protected park in central Ontario. Although our overall percentage of injuries was higher than the 35% reported by Walde et al. (2003), few of our observed injuries (7%) were actual amputations. The large proportion of adult turtles with injuries in our populations clearly indicates substantial predation pressure. This finding leads us to question whether predation, either natural or facilitated by anthropogenic features such as roads, ATV trails, and power line corridors in these systems, may be a significant factor affecting long-term population viability. However, we note that despite 4 y of intensive field work in the 2 watersheds, we documented only 4 cases of adult mortality, none of which could be directly attributed to either natural predation or anthropogenic activities. In the 2 of 4 cases where causality could be reasonably attributed, necropsy results suggested bacterial infection of the testes in a male in 1 case. In the second case, we found the dead female turtle, with radio-transmitter still functioning, submerged within a riverine log jam only a few days after she had been observed alive and healthy. In this case, we suspect that the turtle drowned as a result of being pinned against the log jam in a high river-flow storm event, which occurred in the time frame between the 2 observations. We were unable to attribute cause of mortality in the 2 remaining cases. Overall, our results demonstrate that these 2 Wood Turtle populations occurring at the northwestern periphery of the range in Canada are comprised of numerous, large-sized, reproductively mature males and females, with sex ratios biased towards females. In addition, the populations each exhibit sustained juvenile recruitment over several years, as evidenced by size-class frequency distributions. There was no evidence of mortalities attributable to predation or anthropogenic activities over 4 y of intensive study. Taken together, our findings support the conclusion that these populations of Wood Turtles, occupying prime forested riverine habitat at the northwestern periphery of the species’ range in Canada, are currently healthy and stable, and that regulatory controls imposed on these systems appear to be effective in their protection. Acknowledgments We sincerely thank N. Hanes, L. Laundriault, I. Langis, Dr. J.D. Litzgus, J. Rouse, and J. Sicoly for technical assistance and/or guidance on various aspects of the project. We thank M. Briel, M. Comrie, C. Ginou, C. Hopwood, P. McBay, A. Sulpizio, P. Tuarze, and S. Waite for assisting with field research. Special thanks to the Ontario Ministry of Natural Resources and Forestry’s Canine Unit, M. Buckner and dog “Rebel”, for their invaluable assistance in detecting Wood Turtles. We also appreciate the anonymous reviewers, whose comments Northeastern Naturalist 585 J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 helped improve and clarify aspects of this manuscript. Funding for this project was provided by Great Lakes Power Transmission, Algoma Power Inc., Hydro One Networks Inc., and a grant from the Ontario Species at Risk Stewardship Fund. Literature Cited Brooks, R.J., C.M. Shilton, G.P. Brown, and N.W.S. Quinn. 1992. Body size, age distribution, and reproduction in a northern population of Wood Turtles (Clemmys insculpta). Canadian Journal of Zoology 70:462–469. Brooks, R.J., M.A. Krawchuk, C. Stevens, and N. Koper. 1997. Testing the precision and accuracy of age estimation using lines in scutes of Chelydra serpentine and Chrysemys picta. Journal of Herpetology 31:521–529. Cagle, F.R. 1939. A system of marking turtles for future identification. Copeia 1939:170–172. Calder, W.A. 1995. An extralimital broad-tailed hummingbird in winter: Disoriented or harbinger of change? Journal of Field Ornithology 66:522–530. Committee on the Status of Endangered Wildlife in Canada (COSEWIC). 2007. COSEWIC assessment and update status report on the Wood Turtle, Glyptemys insculpta, in Canada. Committee on the Status of Endangered Wildlife in Canada, Ottawa, ON, Canada. vii+42 pp. Available online at http://publications.gc.ca/collections/collection_2008/ec/ CW69-14-1-2008E.pdf. Accessed 12 July 2015. Daigle, C. 1997. Size and characteristics of a Wood Turtle, Clemmys insculpta, population in southern Québec. Canadian Field Naturalist 111:440–445. Ernst, C.H., and J.E. Lovich. 2009. Turtles of the United States and Canada, 2nd Edition. The Johns Hopkins University Press, Baltimore, MD. 840 pp. Flanagan, M., V. Roy-McDougall, and G. Forbes. 2013. Survey methodology for the detection of Wood Turtles (Glyptemys insculpta). Canadian Field Naturalist 127:216–223. Greaves, W.F., and J.D. Litzgus. 2007. Overwintering ecology of Wood Turtles (Glyptemys insculpta) at the species’ northern range limit. Journal of Herpetology 41:32–40. Greaves, W.F., and J.D. Litzgus. 2009. Variation in life-history characteristics among populations of North American Wood Turtles: A view from the north. Journal of Zoology 279:298–309 Harding, J.H., and T.J. Bloomer. 1979. The Wood Turtle, Clemmys insculpta: A natural history. Bulletin of the New York Herpetological Society 15:9–26. Hilton-Taylor, C. (Compiler). 2000. 2000 IUCN Red List of Threatened Species. Gland, Switzerland. 61 pp. Holt, R., T. Keitt, M. Lewis, B. Maurer, and M. Taper. 2005. Theoretical models of species' borders: Single species approaches. Oikos 108:18–27. Kaufmann, J.H. 1992. The social behavior of Wood Turtles, Clemmys insculpta, in central Pennsylvania. Herpetological Monographs 6:1–25. Kavanagh K., L. Gratton, M. Davis, S. Buttrick, N. Zinger, T. Gray, M. Sims, and G. Mann. (undated). Eastern forest-boreal transition. World Wildlife Federation. Washington, DC. Available online at http://www.worldwildlife.org/ecoregions/na0406. Accessed 27 January 2016. Kirkpatrick, M.A., and N.A.H Barton. 1997. Evolution of a species’ range. American Naturalist 150:1–23. Koprowski, J.L., S.R.B. King, and M.J. Merrick. 2008. Expanded home ranges in a peripheral population: Space use by endangered Mt. Graham Red Squirrels. Endangered Species Research 4:227–232. Northeastern Naturalist J. Cross, R. Cross, D. Chartrand, and D.G. Thompson 2018 Vol. 25, Issue 4 586 Litzgus, J.D., and R.J. Brooks. 1996. Status of the Wood Turtle, Clemmys insculpta, in Canada. Committee of the Status of Endangered Wildlife in Canada (COSEWIC),.Canadian Wildlife Federation. Ottawa, ON, Canada. Litzgus, J.D., and R.J. Brooks. 1998. Testing the validity of counts of plastral scute rings in Spotted Turtles, Clemmys guttata. Copeia 1998:222–225. Lovich, J.E., C.E. Ernst, and J.F. McBreen. 1990. Growth, maturity, and sexual dimorphism in the Wood Turtle, Clemmys insculpta. Canadian Journal of Zoology 68:672–677. Niederberger, A.J., and M.E. Seidel. 1999. Ecology and status of a Wood Turtle (Clemmys insculpta) population in West Virginia. Chelonian Conservation Biology 3:414–418. Ontario Ministry of Natural Resources and Forestry (OMNRF). 2016. Species at risk in Ontario list. Available online at https://www.ontario.ca/environment-and-energy/speciesrisk- ontario-list. Accessed 17 August 2015. Ross, D.A., R.K. Anderson, C.M. Brewster, K.N. Brewster, and N. Ratner. 1991. Aspects of the ecology of Wood Turtles (Clemmys insculpta) in Wisconsin. Canadian Field- Naturalist 195:363–367. Saumure, R.A. 1992. Clemmys insculpta (Wood Turtle). Size. Herpetological Review 23:116. Saumure, R.A., and J.R. Bider. 1998. Impact of agricultural development on a population of Wood Turtles (Clemmys insculpta) in southern Québec, Canada. Chelonian Conservation Biology 3:37–45. Smith, K.A. 2002. Demography and spatial ecology of Wood Turtles (Clemmys insculpta) in Algonquin Provincial Park. M.Sc. Thesis. University of Guelph, Guelph, ON, Canada. Thompson, D.G., T. Swystun, J. Cross, R. Cross, D. Chartrand, and C.B. Edge. 2018. Fineand coarse-scale movements and habitat use by Wood Turtles (Glyptemys insculpta) based on probabilistic modeling of radio- and GPS-telemetry data. Canadian Journal of Zoology 96:1153–1164. DOI:10.1139/cjz-2017-0343. van Dijk, P.P., and J. Harding. 2011. Glyptemys insculpta. IUCN Red List of Threatened Species Version 2011.2. Available online at http://www.iucnredlist.org. Accessed 22 March 2015. Walde, A.D., J.R. Bider, C. Daigle, D. Masse, J.C. Bourgeois, J. Jutras, and R.D. Titman. 2003. Ecological aspects of a Wood Turtle, Glyptemys insculpta, population at the northern limit of its range in Québec. Canadian Field Naturalist 117:377–388. Wesley, P.A. 2006. Local- and regional-scale habitat selection by Wood Turtles (Glyptemys insculpta) in Ontario. M.Sc. Thesis. University of Guelph, ON, Canada. Williams, C.K., A.R. Ives, and R.D. Applegate. 2003. Population dynamics across geographical ranges: Time-series analyses of three small-game species. Ecology 84:2654–2667. Wilson, D.S., C.R. Tracy, and R. Tracy. 2003. Estimating age of turtles from growth rings: A critical evaluation of the technique. Herpetologica 59:178–194.