Evaluation of headstarting as a conservation tool to recover Blanding's Turtles (Emydoidea blandingii) in a highly fragmented urban landscape

Abstract

Freshwater turtle populations are declining globally as a result of anthropogenic activities. Threats to turtles in urban areas are exacerbated by road mortality and subsidized predators, which can lead to catastrophic shifts in population size and structure. Headstarting is used as a conservation tool to supplement turtle populations that may otherwise face extirpation. A headstarting program began in 2012 to recover a functionally extinct population of Blanding's Turtles (Emydoidea blandingii) 26in Rouge National Urban Park (RNUP), Ontario, Canada. The original population included five adults and one juvenile turtle. From 2014 to 2020, 270 headstarted turtles were released. The population has been monitored annually since 2014 using visual-encounter surveys, radio-telemetry, and live trapping (from 2018 onwards). We used mark-recapture and radio-telemetry data to quantify abundance, survival, and sex ratio of the headstarted turtle population. Using a Jolly-Seber model, we estimated abundance to be 183 turtles (20 turtles/ha) in 2020. Estimated survival of headstarted turtles approached 89%, except for turtles released in 2019 when survival was 43% as a result of a known mass mortality event at the study site. Pre- and post-release sex ratios were not significantly different (χ2 = 1.92; p = 0.16), but shifted from 1:1.5 to 1:1 male:female post-release. Given that the headstarted turtles have not yet reached sexual maturity, it is unclear whether headstarted turtles will reach adulthood and successfully reproduce to maintain a self-sustaining population. Thus, to evaluate the success of the headstarting program, long-term monitoring is required.

Fig 1. Spatial overview of the Rouge National Urban Park in Toronto, Ontario, Canada.

The exact location of the study site is excluded to protect at-risk species from poaching. The map was created by compiling data available from https://geohub.lio.gov.on.ca/ on QGIS (3.20.1-Odense) [–32].

Fig 2. Simplified conceptual framework of the Jolly-Seber open population model used to estimate abundance and survival of headstarted Blanding’s Turtles in the Rouge National Urban Park.

The analysis included 260 released headstarted turtles, of which 139 had capture histories from mark-recapture and radio-telemetry surveys. Survival and detection parameters were indexed by time and by acclimation (acc), which had two levels: ‘New’, for newly released turtles and Exp. (experienced) after more than one year in the wild. Probability of entry (b) was indexed by time and release cohort. b₀ occurs prior to the first sampling occasion. RH_year refers to the size and year of each release. ${\widehat{N}}_t$ refers to population size in each year. Implementation of this model included partitioning of release cohorts into smaller groups according to telemetry duration so that detection probability, p, could be fixed at 1 on specific occasions when turtles were tracked.

Fig 3. Conceptual diagram of how survival, telemetry, and detection were represented in encounter histories and using parameter fixing of detection probability, p, for a subset of potential scenarios.

Scenarios are illustrated for a single Blanding’s Turtle release cohort over three sets of annual surveys. Hollow turtle symbols indicate an individual was not observed, a solid turtle symbol indicates that the turtle was observed. The red X indicates mortality, and the blue “wifi” symbol indicates a radio transmitter. Sampling by trapping and opportunistic detection were collapsed into one occasion per year with survival occurring during intervals between occasions. Telemetry observations of live individuals in May were considered live encounters during the corresponding annual occasion; otherwise, we considered telemetry observations during the rest of the year to occur during intervals, which are not represented directly in encounter histories. *Detection probability is conditioned on survival; ^†Jolly Seber models use only live encounter data, even when mortality is observed; ^§Encounter history is a binary sequence: 1 when an individual is observed alive, 0 when not observed; ^‡Conditional on being released, turtles are encountered according to the product of current detection probability and previous survival. Certain detection (p = 1) but no encounter, implies no survival over a previous occasion, which informs survival estimation.

Fig 4. Population size of headstarted Blanding’s Turtles in the Rouge National Urban Park estimated using the Jolly-Seber open population model.

The blue shading indicates the 95% confidence interval for population size.

Fig 5. Headstarted Blanding’s Turtle abundance in the Rouge National Urban Park based on release year.

The population size ($\widehat{N}$) increased each year but the number of turtles in each cohort decreased each year as a result of mortality. The total abundance in 2020 was 218 headstarted turtles.