Declining reptilian populations have been a growing concern over the last couple of decades. One such declining species of concern, the Blanding’s turtle (Emydoidea blandingii), occurs as isolated populations in North American prairie-wetlands and is at risk of extirpation due to habitat loss and fragmentation, and increased predator (e.g. racoons, coyotes) populations due to supplemented resources in urban environments. To help mitigate declining populations, wildlife managers have invested in the conservation of this species through head-starting (i.e. reared in ex-situ) and juvenile release programs to augment wild Blanding’s turtle populations. However, much of their spatial and winter/thermal ecology is understudied, and data for juveniles, and juveniles reared in ex-situ is especially scarce, yet this information is imperative to understanding shortfalls and improving head-starting efforts in the future. In spring 2016 (RR₂₀₁₆) and 2017 (RR₂₀₁₇) we released a cohort (n=12 each year) of head-started juvenile Blanding’s turtles equipped with radio transmitters and temperature dataloggers into a prairie-wetland in the greater Chicago region, North America. Using ground-based radio telemetry, we determined seasonal movement areas (SMAs; spring, summer, and fall) and annual home ranges (AHRs) for both RR₂₀₁₆ and RR₂₀₁₇ cohorts via Kernel Density (KD) estimates. We also investigated the thermal characteristics of overwintering for both juvenile cohorts. We found that SMAs for the RR₂₀₁₆ cohort, but not for the RR₂₀₁₇ cohort, significantly differed across seasons for most SMA estimators. We also found that juveniles in both cohorts not only survived overwintering but also displayed similar overwintering phenology (i.e. initiation: October-November; termination: April) and temperature variation as Blanding’s turtles adults in other studies. Overall, our results indicate that head-started juvenile Blanding’s turtles may be able to acclimatize quickly to their natural environment post-release. Our study provides evidence of the efficacy of well-developed head-starting programs that aim to augment and preserve imperiled turtle populations.

Collection

Morphometric data 

Carapace length was measured using a pair of digital calipers. Mass was recorded via a digital animal weighing scale. Percent fat was calculated using the following formula from Newman et al. 2019 (see. Cann and Weber et al. 2020) fat % = 2.025 + (3.978 × 10^-4 × (CL³ ⁄mass)) − (1.152 × 10^-3 × mass), where CL = carapace length.

Morphometric data filename: Morphometrics_allCohorts

Spatial data

In quotes are from the manuscript itself:

Juveniles in RR₂₀₁₆ [Recently Released group 2016] were tracked once weekly from May 2016 to November 2016 for spatial analyses, and then bi-monthly as weather allowed to April 2017 for estimated emergence from brumation. Juveniles in RR₂₀₁₇ [Recently Released group 2017] were tracked two to three times weekly from May 2017 to November 2017 for spatial analyses, and then bi-monthly as weather allowed until April 2018 for estimated emergence from brumation. Juveniles in RR₂₀₁₆ were tracked further in tandem with juveniles in RR₂₀₁₇ until April 2018 if survival status allowed, though this data was excluded from RR₂₀₁₇ statistical analyses. Global positioning system coordinates were recorded in UTMs on Garmin (GPSmap 62sc/64st) devices. To provide a probabilistic estimate of SMAs and AHRs by scaling the boundaries to the area’s most frequently visited by the individual, we used the Geospatial Modeling Environment (Beyer 2015; version 0.7.4.0 in tandem with R 3.1.1+) to calculate kernel density SMAs and AHRs estimates (KD; i.e. non-parametric method to measure the probability of occurrence based on the density of points in similar areas). Variation in seasonal movement areas estimates were calculated for each individual juvenile released during each season (Table 1), approximately following the phenology of Blanding’s turtle activity in the adjacent state of Wisconsin, U.S.A. (Ross and Anderson 1990, Thiel and Wilder 2010), and for AHRs (i.e. total active season; static location of the juveniles during brumation were excluded from SMA and AHR calculations). We used the least squares cross-validation (LSCV) bandwidth following Seaman and Powell (1996) and Byer et al. (2017) with a cell size of two. Kernel density raster files were then converted to isopleths at 95, 90, and 50% confidence intervals to get a range of estimates (Fischer et al. 2013, Ghaffari et al. 2014, Drabik-Hamshare and Downs 2017). These intervals represent how likely it is for our tracked juveniles to be found in their respective isopleth SMAs and AHRs. For example, a SMA KD home range at 95% represents the entire range of an individual, or the area in which the animal spends 95% of its time; whereas an isopleth of 50% represents the core area of habitat where the animal spends 50% of its time. Isopleth files were then imported into ArcMap version. 10.3.1 (ESRI 2015) where we converted each isopleth to polygons for area calculation. Individual juveniles that did not complete the entire season or seasons of observation were removed from the mean calculation of the group (e.g. deaths, transmitter loss; see Cann and Weber et al. 2021 for more survivorship details).

Once polygons for each isopleth were created and the areas estimated within ArcMap, two excel files were created to record all individual's home ranges according to the two estimators used:

Kernal Density Estimate filename: KDE_area_allCohorts

Minimum Convex Polygon filename: MCP_area_allCohorts

Temperature data

In quotes are from the manuscript itself:

Thermal characteristics of the brumation sites were assessed by placing a Thermochron iButton dataloggers (model DS1921G, Dallas Semiconductor) layered in a black rubber coating (i.e. for water damaged mitigation; Performix – Plasti-dip Brand®) on the carapace of each released individual. Carapace temperature (T_c) was measured in Celsius using the iButton dataloggers attached to a single carapacial scute adjacent to the radio-transmitter on each juvenile turtle also using 5-min epoxy (Milanovich et al. 2017) prior to release. Although it has been shown that coating Thermochron iButtons can have an effect on the temperature readings, the differences seen were relatively small at 0.0-1.3°C compared to those uncoated (Roznik and Alford 2012), therefore we followed methods of Akins et al. (2014) and Harden et al. (2015) to coat iButtons. T_c was logged at a rate of once every 60 mins day^-1 for the spring, summer, and fall months for RR₂₀₁₆. T­_c iButtons were collected and replaced after the following periods: May-August and September-October 2016. We changed the rate at which temperatures were logged for the 2016-2017 winter season (November-April), programming them to log once every 150 mins day^-1 until turtle emergence to ensure memory space for the entirety of the brumation period. We subsequently made the same data-logging rate change as well for the dataloggers attached to RR₂₀₁₇ in the spring, summer, fall, and 2017-2018 winter. Environmental temperature (T_e) was measured by using black rubber (Performix – Plasti-dip Brand®) coated iButtons placed in three vertically oriented PVC pipes randomly located throughout the wetland among emergent vegetation, excluding the open water habitat. iButtons were inserted at 15 cm intervals in notched groves on the PVC pipes to correlate with depths into the substrate of 45 cm, 30 cm, and 15 cm below substrate level, 0 cm at surface level of the muddy substrate/water, and 15 cm above surface level (ambient, above the water), allowing us to compare the T_c with T_e at or above surface level and various subsurface levels. T_e dataloggers for the wintering period were programed to log temperatures once every 150 mins day^-1 until spring or turtle emergence. Additionally, T_e loggers enabled us to quantify brumation site depth in the substrate after cross-comparison with T_c loggers (Currylow et al. 2013).

RR₂₀₁₆ filename: Temp_RR2016

RR₂₀₁₇ filename: Temp_RR2017

Note: Cohort 2016 and Cohort 2017 in any excel sheet correlate to RR₂₀₁₆ and RR₂₀₁₇ in the manuscript.

Processing

Morphometric data

We used Statistica to import the excel file containing all morphometric related data. We then summarized said data by month, cohort, and year data was collected. All turtle ID measurements were combined and then averaged for each month and each year. Cohorts (i.e. RR₂₀₁₆ and RR₂₀₁₇) were summarized separately.

Spatial data

We used Statistica to summarize and analyse Kernel Density and Minimum Convex Polygon estimator datasets. All turtle IDs within the same cohort were analyzed across each season for the three isopleth values used (i.e. 95%, 90%, 50%) in the Kernel Density estimates. Since there were no isopleths associated with the Minimum Convex Polygon estimate, this was done once.

In quotes are from the manuscript itself:

Differences in AHR [annual home range] KD [kernel density] home ranges of individual turtles across years, and with different number of locations, were examined using a mixed-effect ANOVA with year as fixed effects, individual turtle ID as the random effect, and the number of locations as the covariate. We used one-way ANOVAs to test if there was variation in SMA KD home ranges for each cohort separately. Tukey Multiple Comparison tests were calculated for significant ANOVAs.

Temperature data

For the two temperature dataset files corresponding to RR₂₀₁₆ and RR₂₀₁₇, the three random locations that were used to measure environmental temperature (T_e) labeled: Center_PVC, East_PVC, and West_PVC were consolidated into one T_e, and averaged for each day/time categorized by the depth of T_e (i.e. 15 cm below, 0 cm ground, 15 cm above, 30 cm above, and 45 cm above).

All turtle ID temperatures were consolidated into carapace temperature (T_c), and averaged for each day/time. 

In quotes are from the manuscript itself:

We used Statistica to graph all T_e (at the 5 different depths described above) and T_c values in line plots. This was done for each RR₂₀₁₆ and RR₂₀₁₇ temperature datasets.We used one-way ANOVAs to examine whether T_c mean values (dependent variable) for each cohort were significantly different compared to the five T_e mean values (independent variables) for the entire brumation period, and for each biweekly period throughout brumation (October through April; year depending on cohort). This analysis helped determine at what depth the juveniles were positioned for the duration of the brumation period, and during periodic periods during brumation. We conducted one-way ANOVAs to examine whether the two-week period prior to juvenile initiation into brumation, and the two weeks prior to spring emergence, were significantly different than T_e. Subsequent Tukey Multiple Comparison Post-Hoc tests were conducted for significant ANOVAs.

We used Microsoft 365 Excel to open and view the .csv files used. For all analyses done on the .csv files we used Statistica version 13.0 (TIBCO Software Inc., 2017) which is proprietary. An open-source alternative that can be used is R from the cran.r-project.org website.