Colonization and usage of an artificial urban wetland complex by freshwater turtles

Study site

• Tommy Thompson Park (TTP)—Toronto, Ontario (43°37′4″N 79°20′33″W)

The City of Toronto, Canada’s largest metropolitan area, is built on a series of watersheds that cut along a north–south axis ending in Lake Ontario. Toronto is historically rich in wetland habitat that has been gradually lost due to urbanisation. In the 1960s, the Toronto Port Authority built a five km long breakwater (the Leslie Street Spit) into Lake Ontario to create additional lands for port infrastructure in conjunction with the opening of the St. Lawrence Seaway. As the Spit grew in size, native and non-native plants began to colonize the land, and it became an important stopover habitat for migrating birds. By 1991, large colonies of cormorants, gulls, and herons nested at the Spit.

Toronto and Region Conservation Authority was tasked to develop and implement a master plan for a public park on the Spit, leading to the creation of TTP. One component has been natural area enhancement, and TRCA has focused efforts on creating a varied wetland complex, to improve aquatic habitat including features such as aquatic vegetation and structural habitat (logs, root balls, stones, gravel beds, sandy beaches, etc.). TRCA continues to tweak wetland design at TTP using an iterative adaptive management process that relies on active monitoring.

Description and history of the individual water bodies at TTP

At TTP, there are three types of created water bodies forming the wetland complex: ponds, embayments, and cells (Fig. 1). Ponds are small and unconnected stand-alone bodies of water. There are two main ponds in the wetland complex, Triangle and Goldfish ponds. Goldfish pond is the smallest pond and was accidentally created through asphalt disposal in the late 1970s. It measures approximately 500 m² (0.05 ha). TRCA has not undertaken any restoration projects in Goldfish pond and it has been allowed to naturalize without any intervention. Triangle pond was constructed between 1975 and 1978 as a trial confined disposal facility and it covers 0.8 ha. TRCA capped the facility in 1997 and subsequently created a wetland by adding structural habitat (large stones, rocky shoals, woody materials), as well as aquatic and riparian vegetation in 1998.

The TTP embayments were created as coastal wetlands. Embayment A covers 6.25 ha and has benefited from two restoration projects; (a) 2004–2005 to create spawning shoals on the east side of the embayment; (b) 2009 to create backwater refuge and coastal wetland along the west and south shorelines. Embayment B covers 19 ha and has received two restoration projects; (a) 1995–1997 to create northern pike (Esox lucius) spawning channels in the sheltered backwater area to the east of the embayment, as well as structural habitat in the main embayment; (b) 2011 to improve and enhance the coastal wetland conditions along the shoreline of the main embayment. Embayment C covers 24 ha and has seen three restoration projects; (a) 1996–1998 to create structural, wetland, and riparian habitat on the west portion of the embayment; (b) 1998 to create pike spawning channels on the southeast shoreline; (c) 2010 to improve coastal wetland habitat along the east shoreline along with additional structural habitat enhancements throughout the embayment. Embayment D covers seven ha. In 2012–2014 a coastal wetland restoration project isolated the embayment from the lake through the construction of a berm and the installation of a fish and water level control structure. Embayments A, B, and C are open to the lake, with backwaters and fish spawning channels, while Embayment D is connected to the lake via a fish gate with seven cm vertical grates, impeding the travel of large invasive common carp (Cyprinus carpio) into the wetlands (French, Wilcox & Nichols, 1999).

Cells were created as confined disposal facilities to store dredged material from the lower Don River and Keating Channel and eventually capped and transformed into wetlands. Cells are also connected to Lake Ontario via fish gates. Cell-1 was filled to capacity in 1985 and capped. TRCA created 11 ha of terrestrial and aquatic habitat in Cell-1, including a 7.7 ha coastal hemi-marsh between 2003 and 2006. A fish and water level control structure was installed in 2012. Cell-2 covers 10 ha and was filled to capacity in 1997. Cell-2 was capped and a coastal hemi-marsh wetland was created in 2015–2017 by adding structural habitat and vegetation. TRCA expected to install fish and water level control structures in late 2017–2018 and to plant aquatic vegetation into 2018. Cell-3 measures 32 ha and is still accepting dredged material. TRCA does not expect to transform it into a wetland for some time, and hence this cell is currently devoid of emergent vegetation or structural habitat suitable for turtles.

Survey methods

We studied the turtle population using three main methods: a capture-mark-recapture (CMR) methodology, very high frequency (VHF) tracking, and shore-based visual basking surveys.

Capture-mark-recapture

We conducted a CMR study in 2016 and 2017, trapping turtles using hoop nets, basking traps, or seine nets. CMR (Lecren, 1965) is a robust method to determine population densities and trends (Pollock et al., 1990). We used one metre diameter three-ring hoop nets (no.15 net with 6.25 cm mesh from Champlin Net Company, Jonesville, LA, USA), basking traps (manufactured in-house, varying in size from 1 to 2 m², all with 6.25 cm mesh size) and a 30 m seine net (manufactured in-house, with 6.25 cm mesh). The size of the mesh precluded capturing very small turtles or juveniles, so our demographic information is skewed toward adults. Juveniles were captured opportunistically by hand. We baited the hoop traps with a variety of food, including frozen fish, canned cat food, canned sardines, and frozen chicken, depending on availability. Bait was placed in perforated bags, hung near the back of the traps at such a height that the bait was in the water. The trap was three-quarters submerged in water, in order to always have air available to trapped animals. Captured turtles were removed from the traps and processed on shore, where we recorded weight with a digital scale, body morphology with digital vernier calipers (plastron length, straight carapace length, carapace width, body depth, precloacal length, foreclaw length) and sex (if possible). For MP turtles, we identified males based on the elongated foreclaws and long tail, with the cloaca situated well beyond the limit of the carapace, compared to shorter foreclaws and tails in females. For SN, we used the relative length of the posterior lobe vs. the base—cloaca tail measurement, and by gentle rocking to encourage penile eversion. For Blanding’s turtles, males have a concave plastron and longer tail. We notched the marginal scutes of the shells to give each turtle a unique identifying number (Cagle, 1939). After processing, we returned turtles to the general area of the capture trap. We trapped for 25 days during the months of June and July 2016 and for 24 days in May and June 2017, using up to 20 hoop traps (13 traps in 2016 and 20 traps in 2017), four basking traps and one seine net (2016 only). The seine net was only used 1 day in 2016 (one SN captured) and was unavailable in 2017 and the basking traps had little success in 2016 and were not used in 2017.

We trapped in each water body for a 1–2-week period, depending on water body size and success of trapping. Using a CMR protocol in successive years also allowed us to observe turtles moving from one water body to another. As part of our invasive species protocol, we removed all captured Red-Eared Sliders (Trachemys scripta elegans) and transferred them to an adoption center (Little RES Q, Pefferlaw, ON, Canada).

We used previously recorded locations of visually surveyed turtles to decide in which water body we should trap. Staff and volunteers had reported many turtle sightings in both ponds, in all embayments and in Cells 1 and 2. Although Cell-2 had previously recorded turtle activity, it was under further restoration during the time of this study and was unsuitable for trapping. Consequently, we trapped in seven distinct water bodies (see Table 1): Goldfish pond (2016/2017), Triangle pond (2016/2017), Cell-1 (2016/2017), Embayments A (2016), B (2016), C (2017), and D (2016/2017). Wetlands that produced very few turtles in 2016 were not re-sampled in 2017.

Statistics

1. Population estimates: We applied the Chapman correction for small sample sizes to a simple two-sample Lincoln–Petersen index to estimate population size (Chapman, 1951). We estimated the population size for the two main turtle species at TTP where N = [(K+1)(n+1)]/(k+1); N is the estimated population, n is the individuals captured in 2016, K is the captured population in 2017, and k is the population recaptured in 2017 that was marked from a previous capture.

2. Turtle sizes: Straight-line carapace length is reported using the mean ± 1 standard deviation, with ranges reported as minimum–maximum.

3. Species representation: We tested for uniformity of individual species distribution through the various water bodies using a Chi-square test.

4. Sex ratio: We performed a binomial test (non-parametric two-sided) to determine if the sex ratio of the two main turtle species was unbiased. The resulting probability, p, is two-tailed and is significant at 95% confidence level if the resulting p is smaller than 0.025 or greater than 0.975.

VHF tracking

We affixed a 12 g VHF transmitter (model R185; Advanced Telemetry Systems, Isanti, MN, USA) to certain individuals of turtle species considered endangered or species at risk in Ontario (Blanding’s and SN). Using a water epoxy, we glued the transmitters to the back of the carapace, generally offset to one side, with the antenna trailing. We only attached transmitters to turtles >300 g (transmitter weight <5% turtle body weight). In 2016, we aimed to get an even number of males and females and we able to track eight male and eight female SN and three female Blanding’s turtles. After an initial data investigation, we noticed that the female SN were travelling greater distances than the males. Therefore, in 2017 we became more interested in the movement of females, and consequently added new transmitters only to new female SN. We also removed the old VHF transmitters (most of which had depleted their battery power) from any recaptured male SN. As a result, in 2017, we replaced the transmitters on all three female Blanding’s turtles, on four recaptured female SN, added five new female SN to the study, and removed old, non-functional transmitters from three recaptured male SN. We continued to track the remaining five male SN until we lost their VHF signals.

We tracked each individual turtle to a specific water body on a weekly basis during the active period and once monthly over the winter. Hibernacula locations were determined in the fall by locating turtles that had buried themselves under vegetation, logs, stumps or in the mud and reconfirming these locations in the winter. The weekly time period will necessarily underestimate the movement of turtles throughout the wetland complex but we used it as a baseline indication of the minimum habitat use and of exploratory movement at the study site.

Visual surveys

Staff and volunteers at the park also conducted regular visual surveys (using binoculars) of basking turtles from shore and reported results back to the authors. Visual surveys offer a continuous data set starting in 2004. Staff and volunteers also reported opportunistic observations of turtle nests throughout the 2016–2017 study period.

DNA analysis

We collected blood samples from the Blanding’s turtle by caudal venipuncture and stored these samples on FTA cards (Whatman Inc., Little Chalfont, UK), for later DNA analysis to understand the likely source of these individuals, which could have re-colonized TTP from a nearby population or introduced to the site by humans after collection at a distant source site. DNA was extracted using standard methods for nucleated blood (Smith & Burgoyne, 2004) and genotyped using 11 microsatellite loci. These included loci GmuB08, GmuD16, GmuD21, GmuD28, GmuD55, GmuD87, GmuD88, GmuD93, GmuD107, and GmuD121 (King & Julian, 2004), and loci Eb17 and Eb19 (Osentoski et al., 2002). Polymerase chain reaction followed the recipe and thermocycling parameters detailed in Davy, Bernardo & Murphy (2014). We ran the TPP samples alongside a positive control DNA from a Blanding’s turtle genotyped previously in Davy, Bernardo & Murphy (2014) to ensure that the microsatellite genotypes obtained were accurate and comparable to existing genotype data. The new genotypes were compared to an existing database for Ontario Blanding’s turtles (Davy, Bernardo & Murphy, 2014) and used Geneclass2 (Piry et al., 2004) to assign or exclude populations of origin for the new samples. We used Bayesian methods of assignment (Rannala & Mountain, 1997), and Monte–Carlo re-sampling simulation algorithm (Paetkau et al., 2004), with 1,000 simulated individuals and α = 0.05, and then ran the analysis a second time using frequency-based assignment (Paetkau et al., 2004).

Permits

This study was conducted with the approval of York University’s Animal Care Committee (YUACC#2016-16W- and #2017-16W-R1) and under the Ontario Ministry of Natural Resources (Wildlife Scientific Collector’s Authorization numbers # 1083601 and 1085922) and Endangered Species Act Permit for Species Protection or Recovery (AU-B-007-16 and AU-B-006-17).

Trapping surveys

We captured 126 turtles (97 new and 29 recaptured) over 496 trap-days (catch-per-unit-effort, CPUE, 25.4%) in 2016 and 103 turtles (50 new and 53 recaptured) over 514 trap-days (CPUE 20.0%) in 2017 (Table 2). Three of the four expected native turtle species were represented in our trapping: MP, SN, Blanding’s, and the introduced Red-Eared Slider (see Table 2). Based on the trapping results we estimated the number of individual MP turtles within TTP at 140 turtles ± 11 (n = 69, K = 71, k = 35) and the number of SN at 35 turtles ± 3 (n = 20, K = 24, k = 14). Neither MP turtles (df = 6, n = 105) = 143.067, p < 0.001) or SN (df = 7, n = 30) = 23.867, p < 0.001) were uniformly distributed across the individual capture locations. The proportion of MP to SN in the whole wetland complex was 3.5:1 but varied by individual wetlands from 1:12 to 28:1. The male:female sex ratios of both MP (45:44, p = 0.584) and SN (12:15, p = 0.351) did not statistically differ from the expected 1:1 ratio if the demographics of each species are taken as a whole over the entire TTP wetland complex. However, MP turtle sex ratios deviated from the expected 1:1 in two wetlands, where in Triangle pond was male skewed, but just below significance (1.7:1, p = 0.970) and Embayment D was significantly female skewed (1:3.5, p = 0.015).

We report MP and SN size and mass for the wetland complex and by specific wetland in Table 3. We also captured three mature female Blanding’s turtles (straight carapace length 215, 241, and 230 mm, and mass of 1,356, 2,200, 1,885 g, respectively). Blood samples were successfully collected from two of these females.

We report on turtle density and biomass (all species) as well as species-specific density and biomass for MP turtle and SN for the most productive wetlands in Table 4. Note that because of the small size of Goldfish pond, density, and biomass per hectare are artificially inflated.

Our first method of quantifying movement through the site was based on the locations of recaptured turtles, and provided a minimum value of exploratory movement. Out of the 35 recaptured MP turtles, seven (20%) switched water bodies between captures, while only one (7%) recaptured SN was first captured in a different location.

We also captured 52 fish of 10 species as by-catch, including 37 common carp (C. carpio) (see Table 5). Of note, we had not captured common carp in 2016 in Embayment-D, but carp were able to migrate into the embayment in 2017 as a consequence of record high lake levels in spring. In general, we were able to release by-catch unharmed. However, there were a few occasions where fish caught in the mesh of the hoop traps had drowned, or had been mortally injured and/or partially consumed by trapped or nearby SN.

Habitat use by Snapping and Blanding’s turtles

Snapping turtles: We monitored a total of 21 SN (eight males and 13 females), and 14 of these turtles (three males and 11 females) were observed visiting more than one wetland. Of the 11 SN that were initially captured in a wetland and then explored other water bodies (eight females and three males), 10 returned to their initial capture point for overwintering. The mean minimum distance travelled (straight line distance) was 0.5 km (0–2.5 km) for males and 0.76 km (0–2.6 km) for females. One female that was captured on land (in July 2017) left TTP to overwinter at an island, 1.6 km away. We tracked the turtles to hibernacula from mid-September to end of October (Fig. 2). However, a few SN in Embayment-D relocated in early November 2016, likely due to sudden drop in the water level caused by the demolition of a beaver dam located inside the fish gate. All of the SN overwintered in shallow water (less than one m of water above the hibernaculum). Overwintering sites included: under vegetation mats (12), under logs, root balls and stumps (10), in mud flats with no vegetation above (four), at the edge or inside of beaver lodges (two), and under artificial structures (a wooden dock and a concrete slab) (see Table 6).

Blanding’s turtles: We tracked three female Blanding’s turtles in 2016 and in 2017. All Blanding’s turtles explored at least three wetlands and all three returned to overwinter in the same small pond. The most active Blanding’s turtle traveled a minimum distance of 2.9 km over the summer of 2016 and 3.2 km in 2017. These turtles overwintered together in both study years, in an area measuring approximately 10 m² located in the deepest section of Goldfish pond (>1.5 m deep). The Blanding’s turtles returned to their overwintering ponds by mid-September and remained active in that pond until mid-November of each year.

Visual surveys

Opportunistic basking surveys conducted by staff and citizen scientists (Table 7) identified the presence of two further species: a Northern Map turtle (G. geographica) in Cell-2 and a suspected Spiny Softshell turtle (Apalone cf. spinifera) in Cell-1 and in Cell-2.

Turtle nests were opportunistically discovered throughout the study period, either by observing a nesting turtle or by discovering a nest site that had been depredated. During our study period, we came across five SN actively laying eggs along the gravel shoulder of the main spine road or in gravel side roads and trails. We did not observe any MP turtles actively laying eggs in a nest. Nesting activity was first observed in mid-June and as late as early August. We also discovered a number of other nest sites by finding dug out holes with scattered empty shells at the surface, presumably depredated. We did not identify any of the depredated nests to species.

Results of the DNA analysis

Microsatellite locus Eb19 did not amplify, and loci GmuD55 and GmuD93 produced genotypes for the positive control that did not match those in the reference data from Davy, Bernardo & Murphy (2014). These three loci were excluded from further analyses. Neither turtle was strongly assigned to any previously profiled population: probability of assignment was <0.65 in all cases for the Bayesian analysis, and <0.45 in all cases for the frequency-based analysis. This result suggests that the two TTP Blanding’s turtles originated from a population that was not represented in the reference data (including samples from the proximate Golden Horseshoe population), and that these assignment results should only be used to infer genetic similarity, but not genetic origin. The Bayesian analysis assigned the first individual most strongly to the Kawartha Highlands, and the second most strongly to Parry Sound District (the eastern shore of Georgian Bay).