Various environmental, physiological, and behavioral factors combine during egg incubation and nest emergence periods to influence the survivorship and individual fitness of freshwater turtle hatchlings (reviewed in Baker et al. 2013). Among others, these include nest and egg temperatures, which influence embryonic development periods and vitality, and emergence timing, which influences risks of predation, thermal stress, and desiccation for neonate turtles (e.g., Plummer 2007; Baker et al. 2010, 2013). These early influences on hatchling survivorship in turn affect turtle population structures (e.g., Drake and Spotila 2002) and are thus of considerable importance in species persistence and in conservation management efforts (e.g., Doody et al. 2001; Nagle et al. 2004; Congdon et al. 2011; Gibbons 2013).

Despite its biological importance, our knowledge of hatchling emergence ecology for freshwater turtles mostly derives from laboratory studies and remains incomplete for most species (Spencer and Janzen 2011). Most investigations have centered on marine species, where thermal cues to nocturnal emergence are believed to function as a mechanism to reduce diurnal predation risks and avoid overheating on hot beach substrates (reviewed in Glen et al. 2006). Potential cues used by freshwater turtles are shared with those for marine species and include extrinsic factors such as critical thermal thresholds above which emergence does not occur (e.g., Hendrickson 1958; Lahanas 1982; Drake and Spotila 2002), temperature changes within the soil column (including rates of change and thermal gradients; e.g., Hays et al. 1992; Gyuris 1993), and precipitation (reviewed in Lovich et al. 2014), either as single factors or in combination (Moran et al. 1999; Doody et al. 2001). Less studied endogenous mechanisms, such as exit timing via internal biological clocks (e.g., Lindeman 1991; Moran et al. 1999; Salmon and Reising 2014), have also been implicated as emergence cues. To date, findings on emergence cues used by freshwater turtle hatchlings are disparate, with little consensus (Costanzo et al. 2008; Spencer and Janzen 2011).

Incomplete knowledge of these early life stages makes it difficult to holistically interpret and provide adaptive explanations for emergence parameters both for given species in local contexts (e.g., interplay of emergence schedules and local predator activity schedules) and for comparative investigations between species or across geographies (e.g., role of regional climates and differing predator communities; Doody et al. 2001). Much of the explanation for limitations in our current knowledge lies in the difficulty of obtaining the requisite data in field settings (e.g., Doody et al. 2001; Plummer 2007; Muldoon and Burke 2012). Further, many field study designs impose restraints on natural hatchling behavior (e.g., dispersal hindered by collection barriers) or yield imprecise data on emergence times and hatchling behavior as a result of deficient site-visitation schedules (Lovich et al. 2014).

Graptemys ouachitensis (Ouachita Map Turtle, Cagle 1953) occurs primarily in large to medium rivers as well as lentic reservoirs throughout the Mississippi River system. Near their northern range limit, the species occurs in Wisconsin in the Mississippi and Wisconsin rivers (Lindeman 2013). Over the past decade, a nesting population of G. ouachitensis in the Wisconsin River has been studied to better understand the nesting and reproductive biology of the species (Geller 2012a), but little is known about the hatchling emergence ecology of this species and, in general, many Graptemys species. Increased knowledge of G. ouachitensis hatchling emergence ecology may provide useful points of reference with sympatric congeners and other emydids in different geographies and aid comparative investigations into underlying dynamics. Similarly, increased knowledge of habitat-related thermal impacts on hatchling viability and sex ratios, along with studies on hatchling dispersal tendencies (see Geller et al. 2020), will help inform effective land-management strategies aiming to conserve turtle populations at these early life stages. The goal of this study was to investigate temperature-related influences on G. ouachitensis hatchling survival and nest emergence schedules using dedicated trail cameras and data loggers.

METHODS

Study Site and Timelines. — We studied the hatchling emergence ecology of a population of G. ouachitensis at a nesting site along the lower Wisconsin River within 10 km of Spring Green, Wisconsin (43°10′38″N, 90°04′02″W) from 2015 to 2017. The site is on a sand terrace approximately 52 m from the riverbank and is bordered on the north and west sides by dry-mesic hardwoods (primarily river birch [Betula nigra] and maple [Acer spp.]) with dense understories of shrubs (esp. Rhus radicans and Rhamnus cathartica), and on the east and south sides by drier, more open habitats with scattered trees (primarily oak [Quercus spp.] and ash [Fraxinus spp.]). Shade from trees variably affects much of the nesting area during morning and late afternoon periods. The nesting habitat consists of various xerophytic herbaceous vegetation covering approximately 20% of the surface, predominantly common ragweed (Ambrosia artemisiifolia), with the remainder being open sand. Vegetative cover on this site increases significantly as the season progresses, partly due to fertile sediments deposited by spring floods. This recurring seasonal growth threatens to render the site unusable for turtle nesting, so vegetation was set back manually in early May of each study year using a scuffle stirrup hoe and rake.

Camera Monitoring and Data Collection During Nesting and Preemergence Period. — Newly constructed turtle nests were located beginning in late May of each study year by daily, midafternoon, onsite review of data from 2 trail cameras (RECONYX^™, Inc., Holmen, WI) monitoring the nesting area using a laptop computer. Cameras were angled downward 30° within sheltering boxes mounted on poles at each end of the nesting area at heights of approximately 2.5 m, yielding a combined field-of-view (FOV) large enough to survey the entire site. Cameras were programmed to take continuous time-lapse (TL) images at 1-min intervals and provided multiple images of all nesting events. Efforts to monitor nesting activity and locate turtle nests ended ≥ 5 d beyond the last camera-documented nesting or transient turtle in early July of each study year, after which site visits were reduced to a biweekly schedule to check the integrity of the caged nests.

Up to 20 nests each study year (total n = 48) were protected by screen cages (56 cm in diameter, 31 cm high) made of 14-gauge, 5 × 10-cm wire mesh, with sides additionally wrapped with 2.5-cm chicken wire. Nest cages were placed over each nest and data loggers (if any, see below) and were secured to the substrate with lengths of stiff wire bent into U-shapes, functioning as staples. Half-meter-wide lengths of ½-inch hardware cloth were similarly secured to the substrate under the peripheries of the nest cages to prevent predators from digging under them, with excess screening inside the nest cages cut away to ensure unimpeded hatchling emergence. A plastic tag labeled with a nest identification code was attached to the top of each nest cage to ensure correct nest identity under camera data review. Monitored nests were selected from early, middle, and late periods within each annual nesting season in order to investigate the potential effects of nest construction timing on patterns of hatchling emergence. Each study nest's location was drawn on a printed photo of the relevant camera FOV and referenced by a unique identification code to track individual histories.

Distance from the substrate surface to the top-of-clutch measurements were obtained by carefully digging down through the sand above the nest chamber using nitrile gloves until the top surface of the uppermost egg was located. Bottom-of-clutch metrics were derived by removing only as much substrate alongside the egg cluster (at 2 locations) as necessary to locate the putative lowermost eggs without dislodging eggs from their initial positions and are, thus, not as well-defined as top-of-clutch metrics.

Soil temperatures at a subset of nests (total n = 22) were monitored with data loggers in each study year (Onset HOBO Pendants [UA-001-08]; Onset Computer Corp., Bourne, MA) programmed to collect temperature data to the nearest 1°C ± 0.5°C at 1-hr intervals from the day of oviposition through the hatchling emergence period. Data loggers were placed horizontally, ∼ 15 cm from each nest at approximate midclutch depth as determined by hand excavation of the associated nest cavity. In 2016 and 2017, additional Pendant data loggers were placed 2 cm below the surface, 15 cm from the nest, to log soil temperature gradients at each monitored nest. Reported temperatures at hatchling emergence are time-weighted interpolations between the respective hourly values before and after nest exit.

Two additional data loggers (Onset HOBO H08-032-08; Onset Computer Corp.) with solar shields were positioned on a metal t-post in a central location of the nesting area and recorded air temperatures to the nearest 1°C ± 0.01°C at 15-min intervals at heights of 1 m and 10 cm above the substrate, respectively. Data from the unit mounted at 1 m (“ambient air temperature”) were also used to calculate heating degree-days (number of days with daily mean values below 18.3°C) from April through September during all study years.

Camera Monitoring and Data Collection During Hatchling Emergence Period. — Observational data on hatchling emergence were collected from mid-August to early October via laterally moveable dedicated cameras (RECONYX models with either low-glow or no-glow infrared emissions, programmed to take continuous TL images at 1-min intervals) suspended over each nest (Fig. 1). Each camera was affixed to the underside of a ½-inch exterior plywood carrying plate (15.2 × 16.5 cm) and held at a height of approximately 1 m above the substrate by tightly strung support wires attached to metal rails at the periphery of the nesting area. Camera carrying plates were positioned slightly to the north of nesting cages to limit shading impacts and to reduce interference with vertical or near-vertical rainfall. Small wedges placed under each camera tilted camera FOVs southward to compensate for this displacement and to provide unhindered views of the nests within the cages. Cameras were pulled along the guide wires to the nest site perimeter for periodic maintenance (data card and battery change-outs every 10 d), thereby avoiding disturbance to the nest site such as vegetation trampling and installing camera support structures near nests.

View Figure

• Download as Powerpoint
• Download Figure

During the hatchling emergence period, rainfall duration data were primarily provided by camera evidence (starting and ending time-stamp data from images showing rainfall-generated substrate darkening and sand particle movements within dedicated nest camera FOVs). Rainfall amounts and supplemental meteorological data were obtained from the Lone Rock Airport, Sauk County, Wisconsin, approximately 9.7 km northwest of the study site.

Nests were excavated in early October to count hatched and unhatched eggs and to photograph each clutch. Unhatched eggs were dissected in the laboratory to quantify the number of eggs containing undifferentiated yolk and those with evidence of embryo development. Reported clutch sizes are the larger of either photo-documented emergent hatchlings or egg counts at nest excavation.

Nests were defined as totally successful if all eggs produced emergent hatchlings (i.e., those successfully exiting the nest chamber), partially successful if they produced ≥ 1 emergent hatchling, and failed if no eggs produced emergent hatchlings (after Nagle et al. 2004). Reported numbers of emergent hatchlings for each nest are the larger number derived from either the direct count generated by camera images or from conservative counts of eggshells believed to represent successful hatching events (few, large shell fragments with “entire” edges; relatively nondiscolored, typically pliable and leathery). We defined the time to first emergence for a given nest as the time interval between nest construction and the first appearance of a hatchling outside of the exit hole, emergence duration as the interval between the first and last emergence events for a given nest, and the emergence period as the interval between the first and last hatchling emergence events for all nests within a given season (after Baker et al. 2010, 2013). We defined per capita emergence success as the rate of hatchlings successfully emerging from initial eggs (after Holcomb and Carr 2011) and considered emergence to be synchronous if all the hatchlings for a given nest exited within the same 24-hr period (i.e., not necessarily within the same calendar day) and asynchronous if this interval exceeded 24 hrs. Hatchling emergence was considered to be diurnal if it occurred in the daytime, including dawn and dusk twilight periods, and nocturnal if it occurred in the nighttime between the hours of evening and morning astronomical twilight (i.e., when the sky is no longer illuminated). Reported sample sizes reflect varying numbers of camera records available for analysis, as influenced by camera position, intervening vegetation, and other variables.

Statistical Analysis. — We used a variety of statistical tests to examine the nesting ecology of G. ouachitensis. For count response variables, including clutch size, number of emergent hatchlings, and number of initial nest emergences per day, we used Poisson models with a log-link function. In most cases, we treated year as a random effect to allow estimation of an overall mean across years and to quantify (albeit imprecisely) the amount of annual variation in nest characteristics. For categorical response variables, including nest success (fully successful, partially successful, or unsuccessful), we used softmax models (Kruschke 2015). For binary outcomes, including per-capita emergence success, probability of synchronous emergence, and probability of initial emergence, we used binomial models with a logit-link function. For positive continuous variables such as time to initial emergence and interval between hatchling emergence times, we used gamma regression models (Hobbs and Hooten 2015) with an identity link function. Finally, for continuous data that could fall anywhere on the number line, such as temperatures, we used models based on the normal distribution. For all models, we selected priors to be vague (Appendix 1). In analyses of the influence of temperature and precipitation on probability of first emergence or number of nests with first emergence per day, we included a random effect of year to account for varying duration of the emergence period (defined as each day from the first emergence to the last first emergence) and the different number of successful nests each year.

We analyzed all Bayesian models with Markov-chain Monte Carlo (MCMC) algorithms implemented in JAGS 4.3.0 (Plummer 2017) and called from Program R 3.5.1 (R Core Team 2018) using the package “jagsUI” (Kellner 2019). We ran most models on 5 independent chains of ≥ 20,000 iterations each after discarding 1000 adaptive phase iterations and an additional ≥ 9000 burn-in iterations. Minimum effective sample sizes were always > 1000. We assessed convergence with the Gelman-Rubin diagnostic (Gelman and Rubin 1992) and visual inspection of history plots. We assessed fit by simulating observations from the fitted model and comparing residuals from the simulated data to residuals from the observed data with a Bayesian p-value (Kéry 2010). We present MCMC settings, numerical convergence diagnostics, Bayesian p-values, and minimum effective sample sizes for each model in Appendix 2. All models converged (all < 1.1) and fit (all Bayesian p-values > 0.10), and we report posterior distributions as mode (95% highest posterior density interval) unless otherwise indicated.

We examined emergence times using circular statistics (Pewsey et al. 2013). We calculated circular means, mean resultant lengths (a measure of dispersion around the mean, where 1 is highly concentrated and 0 is widely dispersed, although not necessarily uniform; Pewsey et al. 2013), and various tests of uniformity: Rayleigh's test for uniformity against a unimodal distribution, and the Kuiper, Watson, and Rao spacing tests that more generally test for uniformity in the presence of multimodal distributions (Pewsey et al. 2013). We also tested whether the mean emergence time was the same across years using the bootstrap version of Watson's nonparametric test of a common mean direction (Pewsey et al. 2013) with 9999 bootstrap samples of the data.

RESULTS

Nesting Seasons and Clutch Sizes. — Nesting seasons spanned 32 to 41 d during the 3 study years, although flooding in 2015 and 2017 affected site availability to nesting turtles and may have influenced seasonal timelines (Table 1). The overall mean clutch size for all 3 study years was 7.2 eggs (SD = 2.27, range = 3–13 eggs, mode = 8, n = 48 nests; Table 1).

Table 1. Nesting season parameters and clutch morphometrics for a Graptemys ouachitensis population along the lower Wisconsin River, Wisconsin, USA.

View Fullscreen

Nest Morphometrics. — The overall mean soil surface to uppermost egg depth of the clutches was 94.1 mm (SD = 15.47, range = 59–125 mm, n = 50). Overall mean nest depth at the lowermost found egg (see “Methods”) was 132.5 mm (SD = 11.38, range = 107–153 mm, n = 42). Overall mean depth from top to bottom of clutches was 40.1 mm (SD = 11.38, range = 13–63 mm, n = 42; Table 1).

Nest Success and Hatchling Emergence Success. — For all 3 study years, 40.6% of nests protected from aboveground predation and not exposed to flooding events were totally successful, 43.8% were partially successful, and 15.6% were failed attempts (n = 32). Overall per capita emergence success for noninundated nests was 60.1% (n = 233 initial eggs). Nests exposed to periods of submergence during nesting season floods in 2015 and 2017 had reduced levels of success (overall 38.5%, n = 13) relative to nonflooded nests or those only briefly (≤ 1 d) in saturated soils in those same years (86.7%, n = 15; Fisher's exact test, p = 0.0163, n = 28), and no flooded nests were completely successful.

Hatchling Emergence Timelines. — All hatchling emergences in this study were during the fall, with no instances of overwintering in the nest. The mean time to first emergence (interval between nest construction and first hatchling emergence) in 2015 was 91.0 d (SD = 6.7, range = 72.4–92.4 d, n = 8), in 2016 was 76.6 d (SD = 5.4, range = 68.0–85.9 d, n = 15), in 2017 was 89.1 d (SD = 10.5, range = 76.8–104.9 d, n = 5) and differed between study years based on parameter estimates from gamma regression. Model-based differences in mean time to first emergence were 11.2 (5.0–17.0) d longer in 2015 than in 2016, 0.43 (–7.3–8.3) d shorter in 2015 than in 2017, and 11.8 (4.5–18.7) d shorter in 2016 than in 2017. Time to first emergence was not associated with clutch size based on the credible interval for the effect of clutch size on time to emergence from the gamma regression model overlapping zero.

The times of day hatchlings emerged from nests were not distributed uniformly (Rayleigh test statistic = 0.3791, p = 0.017; Kuiper test statistic = 2.186, p < 0.01; Fig. 2). Most first hatchling emergence was diurnal, with an overall mean time of emergence of 1141 hrs and annual mean times of first emergence between 0930 and 1500 hrs (n = 28; Appendix 3). Density plots and rose diagrams of times of first emergence appeared bimodal for all years combined, 2015, and 2016, with a mode between 0800 and 1100 hrs and a secondary smaller mode between 1800 and 1900 hrs, whereas emergence times in 2017 were unimodal and centered near 0930 hrs (Fig. 2A). Emergence times of all hatchlings followed a similar pattern (n = 114; Fig. 2B). The mean time of day at which hatchlings emerged did not differ among years (Watson's test statistic = 5.97, p = 0.261).

View Figure

• Download as Powerpoint
• Download Figure

Emergence duration (interval from first to last emergence event for nests with > 1 hatchling) for all nests varied widely (overall mean = 21.9 hrs, SD = 37.0, range = 0.03–121.3 hrs, n = 19; Appendix 3), while that for only synchronous nests (all emergence events within 24 hrs) with entirely diurnal emergence was less variable and averaged approximately 1 hr (overall mean = 1.0 hrs, SD = 0.52, range = 0.28–1.85 hrs, n = 9). Synchronous hatchling emergence was exhibited for most nests during this study, although in some cases emergence from the nest spanned > 1 calendar day. The probability of hatchlings emerging synchronously from a given nest was 0.70 (0.53–0.85), but was not affected by clutch vertical dimension, a proxy for thermal differences in development rates between eggs at the top and bottom of a clutch. The median interval between successive hatchling emergence events within the same calendar days (to reduce the effect of nocturnal inactivity) was 10.5 min, but was highly variable (overall SD = 64.0, range = 0–326 min, n = 72; Appendix 3). Based on gamma analysis of variance (year effects) or single-variable regression (all others), overall emergence duration was not related to year (coefficient for 2016 [β₂₀₁₆] = 0.54 [–46–47]; β₂₀₁₇ = 0.046 [–46–46]), depth to the middle of the clutch (β_cd = 0.010 [–0.014–0.034]), clutch vertical dimension (β_cvd = –0.008 [–0.034–0.015]), or the number of emergent hatchlings (β_eh = 0.024 [–0.082–0.14]). Overall, single exit holes were 1.9 (0.77–5.1) times more likely than multiple exit holes; asynchronous nests were 3.3 (0.34–36) times more likely to have multiple exit holes than synchronous nests.

Influence of Temperature on Time to First Emergence and Hatchling Survival. — Time to first emergence was negatively associated with mean temperatures at midclutch depth and positively associated with the number of degree-days (Fig. 3). For every 0.8°C increase in mean temperature at midclutch, the time to first emergence was 0.93 (0.89–0.97) times as great (Fig. 3). For every 14-d increase in degree-days (i.e., days with mean temperatures below 18.3°C) time to first emergence increased 1.06 (1.03–1.1) times (Fig. 3). For every day later nests were constructed in nesting seasons, the time to first emergence was 0.996 (0.992–1.00) as long (Fig. 4). However, nests constructed later in nesting seasons encountered yearly varying numbers of relatively cool days during late August and September; therefore, they did not necessarily have higher overall mean temperatures than those constructed earlier (Pearson's r = –0.65; 95% CI = –0.96–0.35; pvalue = 0.17).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

The overall mean midclutch temperature during time to first emergence periods for completely successful nests was 26.0°C (SD = 0.73, range = 25.1°C–26.9°C, n = 6) and for nonflooded, partially successful nests was 25.8°C (SD = 0.80, range = 24.9°C–27.1°C, n = 7); the difference is not statistically significant (t = 0.47, p = 0.6471, df = 11). The lack of emergence dates for failed, nonflooded nests complicates efforts to derive their overall mean nest temperatures during these periods, but some failed nests typically had midclutch temperatures 0.1°C–2.9°C lower than successful nests on the same day. However, this was not always the case, nor did this pattern of relatively lower temperatures necessarily hold for standardized periods (e.g., first 40 d of incubation). Late-term hatchling remains were sometimes found inside eggs from failed and partially successful nests with the lowest mean midclutch temperatures during their respective periods (i.e., 24.6°C during the first 67 d of incubation in 2016; 24.9°C and 25.7°C during entire time to first emergence periods in 2015 and 2016, respectively).

Influence of Temperature on Hatchling Emergence Timelines. — Temperature affected both the likelihood of first hatchling emergence and the number of first hatchling emergences. For every 3.6°C increase in mean daily air temperature (at 1 m above the surface), the probability the first hatchling emergence would occur increased 3.1 (95% credible interval = 1.5–6.9) times (Fig. 5). Thus, days with first hatchling emergence were warmer than other days within yearly emergence periods (mean for days with first hatchling emergences = 22.9°C [18.2°C–28.0°C]; mean for days without emergences = 19.4°C [12.4°C–25.3°C]; difference = 3.4°C [1.9°C–5.0°C]), although most first diurnal emergence events (70.8%, n = 24) occurred before daily maximum temperatures were attained. The overall mean air temperature at the time of all diurnal first emergences was 24.7°C (range = 15.7°C–30.1°C, n = 24 events), although individual nests experienced varying amounts of shade during emergence events. Mean air temperatures were lower at the onset of first emergences during nocturnal periods (mean = 18.5°C, range = 14.1°C–22.0°C, n = 4 events). The difference between diurnal and nocturnal temperatures at first emergence was 6.2°C (95% credible interval = 2.1°C–10.4°C).

View Figure

• Download as Powerpoint
• Download Figure

Daily temperatures at midclutch depth were typically at their minima early in diurnal periods (mode = 0800 hrs), and at their maxima during the afternoon (mode = 1700 hrs), considering hatchling emergence periods in all study years. Midclutch temperature profiles lagged behind those for temperatures at 2 cm below the surface at nest locations by approximately 1 to 2 hrs. Temperatures at midclutch depth were cooler than those 2 cm below the surface from midmornings (ca. 0900 and 1000 hrs) until late afternoons (ca. 1800 and 1900 hrs) and, conversely, warmer than at 2-cm depths during the remaining period from evening until the following morning (Fig. 6).

View Figure

• Download as Powerpoint
• Download Figure

The overall midclutch mean temperature at first hatchling emergence during diurnal periods was 24.0°C (SD = 3.4, range = 18.9°C–32.6°C, n = 13 nests; Fig. 7) and varied between years (2015 mean = 26.1°C, range = 18.9°C–32.6°C, n = 4; 2016 mean = 23.7°C, range = 21.8°C–24.5°C, n = 5; 2017 mean = 22.1°C, range = 21.0°C–23.5°C, n = 4). Diurnal first hatchling emergence events for given nests and those on subsequent days (collectively hereafter, “all-d first emergence events”) occurred during periods of both increasing and decreasing nest temperatures depending on whether these events took place during the morning or late afternoon, respectively. However, most events occurred during increasing midclutch temperatures soon after the morning reversal of the nighttime cooling trend (72.4%, n = 29; e.g., Fig. 6). Six additional first emergence events that occurred when morning nest temperatures were not yet increasing, but when temperatures were rising at 2 cm below the surface, are not included in this metric although preemerged hatchlings were likely positioned above the eggshell remnants (e.g., for G. nigrinoda; Lahanas 1982) in uncertain, although probable warming, intermediate thermal environments. The mean interval between the early morning low temperature and the first emergence for a given nest was approximately 1 hr (SD = 1.08, range = 0.2–3.5 hrs, n = 8). When emergence occurred over more than 1 d, emergence that began during diurnal hours usually terminated before dark and only resumed again after sunrise on following days (88.9% of instances, n = 9).

View Figure

• Download as Powerpoint
• Download Figure

During the morning, nocturnal-to-diurnal thermal transition period, the first emergence event for a given nest occurred when nest-level temperatures were increasing at a mean rate of 0.6°C/hr within the hour of emergence (SD = 0.51, range = 0.2°C–1.4°C, n = 8 such nests with temperature data), 2-cm depth temperatures were increasing faster, at a mean rate of 1.6°C/hr (SD = 1.05°C, range = 0.2°C–2.5°C, n = 5), and the mean temperature difference between nest levels and 2 cm below the surface was 0.6°C (i.e., in the presence of an increasing temperature gradient; SD = 0.55°C, range = 0.0°C–1.4°C, n = 5). However, given hatchlings' likely intermediate depth positions (see above), temperature change rates within the hour of emergence at actual hatchling depth were likely between these values.

Contrarily, all-d first emergence events and those of subsequent individual hatchlings within nocturnal periods (n = 5 and n = 9, respectively) were during periods of decreasing temperatures, in accord with consistent patterns of temperature declines during nighttime periods (e.g., Fig. 6). During the only nocturnal first hatchling emergence event with temperature data in this study (at 0400 hrs), the midclutch temperature was 20.4°C during a period of decreasing midclutch temperature (–0.3°C/hr), with a nest depth–to–2-cm temperature gradient of –1.2°C.

Influence of Precipitation on Hatchling Emergence Timelines. — Overall, 42.5% of all-d first emergence events had ≥ 1 mm rainfall in the previous 24 hrs (n = 40 events). Rainfall sometimes occurred during nocturnal emergence (overall, 35.7%; n = 14 total events), but was not observed during diurnal events (n = 100 total events). Although many first emergence events thus occurred without rainfall within the previous 24 hrs, the amount of rainfall on the day of first emergence was positively related to both the probability of first emergence and the number of first emergences (although 95% credible intervals for coefficients overlapped zero; Fig. 8A). For every 9-mm increase in precipitation, the odds of at ≥ 1 first emergence event increased by 1.4 (0.77–2.9) times (Fig. 8A). In contrast, rainfall on the previous day decreased the probability of first emergence by a factor of 0.56 (95% credible interval = 0.17–1.1; Fig. 8B). Significant rainfall within 24 hrs of diurnal emergence did not appear to influence diel timing of all-d first emergence events, because 66.7% (n = 9) of those that experienced ≥ 5 mm of preemergence precipitation still emerged before 1200 hrs, similar to the 69.2% (n = 26) of morning emergences for those that had lesser amounts, or none (Fisher's exact test, p = 1.0000, n = 35).

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

Temperature Impacts on Embryo Survival. — We found that warmer days and higher nest temperatures decreased time to first emergence timelines. The association of faster embryonic development rates with increased temperatures (Ewert 1979, 1985) has been well-documented across turtle taxa (e.g., for Emydoidea,Gutzke and Packard 1987; for Macrochelys,Holcomb and Carr 2011). The 82.3-d overall mean time to first emergence period in this study was longer than that reported for other Graptemys nests (mean = 72.2 d for natural G. nigrinoda nests, Lahanas 1982; and mean = 76 d for artificial G. ernsti [as G. pulchra] nests, Shealy 1976), likely due to variable periods of shade from nearby trees on this small, interior site and/or the seasonally cooler climate at our upper Midwest study site relative to that of the southern United States.

Despite the importance of thermal regimes on embryo survival, published reports of mean nest temperatures for fall-emerging Graptemys are scarce and fragmentary. For example, Vogt (1980) reported (low) mean nest temperatures of 22.5°C and 21.4°C during mid- to late incubation for shaded (R. Vogt, pers. comm., 2018) G. pseudogeographica and G. ouachitensis nests in Wisconsin (at 10 and 14 cm below surface, respectively; total n = 4). Bull (1985), at that same study site, found mean temperatures of approximately 24.3°C–29.3°C for (mostly) G. ouachitensis nests during weeks 4–5 of development (approximated from Bull 1985, fig. 1B); this compares to our value of 26.3°C (range = 23.5°C–27.8°C, n = 13) for successful, nonflooded nests during this same interval. Additional fragmentary records from more southern Graptemys populations include approximately 29.2°C–29.5°C mean G. nigrinoda nest depth temperatures on typical, sparsely vegetated habitats in mid-August (Lahanas 1982), and 27.3°C for 10 G. oculifera nests monitored for up to 60 d after construction (Jones 2006). To our knowledge, the only previous reports of entire time to first emergence period temperatures for natural Graptemys nests appear to be 25.9°C (n = 4) and 28.7°C (n = 6) for G. flavimaculata nests on riverbanks and sandbars, respectively, but emergence success for these nests was not reported (Horne et al. 2003). Although direct comparisons are thus difficult, our overall mean nest temperatures of 26.0°C (n = 6) and 25.8°C (n = 7) for completely and partially successful nests, respectively, appear somewhat lower than those reported for other fall-emerging Graptemys.

Low incubation temperatures are often associated with reduced turtle embryo survivorship (e.g., for G. ouachitensis, Freedberg et al. 2001), in accord with indications of nest failure and relatively low mean nest-depth temperatures (ca. 25.3°C for failed nests) in this study. However, this temperature, as a single factor, may not be solely responsible for nest failure because it is above the thermal minima for embryonic development in Graptemys (below 23°C, Vogt and Bull 1984). Both Vogt (1980) and Freedberg et al. (2004) reported only limited lack of development or embryo death of eggs artificially incubated at 25°C, although it should be noted that daily and seasonal environmental variation within natural nests may produce disparate results from controlled laboratory settings (e.g., Bull 1985; Brennessel 2006). One of 2 failed nests and 2 partially successful nests with relatively low midclutch mean nest temperatures in this study (ca. 24.6°C for first 67 d of incubation; 24.9°C and 25.7°C, respectively) contained dead late-term embryos, similar to artificially incubated Apalone mutica eggs maintained at 26°C, which not only exhibited late-term embryo mortality (50% within 2 wks of hatching, n = 68), but also whole-body trembling and reduced motor skills (Janzen 1993). These examples suggest that, under some circumstances, sublethal low temperatures can allow embryo development to late stages, but then some factor, either extrinsic or intrinsic, prevents effective hatching (e.g., Spencer and Janzen 2011).

The low nest incubation temperatures we observed may promote a largely male-biased hatchling sex ratio. Laboratory studies have documented only male Graptemys hatchling production under constant incubation temperatures of 23°C–28°C and only females from 30.5°C to 35°C, with both sexes developing within intermediate temperatures (Bull and Vogt 1979; Bull et al. 1982). Bull (1985) found natural nests producing greater proportions of male hatchlings had lower mean temperatures and variances during weeks 4–5 of incubation than did those producing more female hatchlings; all of our successful, nonflooded nests with temperature data during this interval (mean 26.3°C, variance range = 0.83°C–3.76°C, n = 13) fall within the parameters for a male-biased sex ratio using this metric (Bull 1985).

General Pattern of Hatchling Emergence. — Reported hatchling emergence times for Graptemys are mixed, with some being largely diurnal (i.e., putatively for G. geographica in Pennsylvania, Nagle et al. 2004), some within both diurnal and nocturnal periods (i.e., for G. pseudogeographica in Wisconsin, Vogt 1981), and some being largely nocturnal (i.e., within the first 3 hrs after sunset for G. oculifera and G. pulchra [population now considered G. pearlensis] in Louisiana and Mississippi, Anderson 1958; ca. 1 hr after dark for G. nigrinoda in Alabama, Lahanas 1982). All hatchling emergence in this study was during the fall, in accordance with Vogt (1980).

Predation risk reduction has long been considered a prominent influence on hatchling emergence timing for both marine and freshwater chelonians (e.g., Mrosovsky 1968; Burger 1976). Hatchlings emerging during daylight hours, like the G. ouachitensis in our study, are more likely to encounter visually based corvids and other avian predators, rather than mammals, while making overland movements to shelter or nearby waterbodies. Contrarily, hatchlings that emerge nocturnally are more exposed to concurrently active raccoons (Procyon lotor) and other mammalian predators whose foraging has a large olfactory-based component (e.g., Galois 1996). At our study site, yearly recurring increases in raccoons during the hatchling emergence period represent a secondary nest predation peak, while the presence of American crows (Corvus brachyrhynchos) and other potential avian predators does not change from low midsummer values (Geller 2012b). Our results and the diurnal emergence pattern also found by Nagle et al. (2004) in central Pennsylvania for G. geographica differ from nocturnal emergence schedules for Graptemys in more southern ranges where potentially higher levels of avian predation (e.g., from fish crows [Corvus ossifragus], Jones 2006) and excessive diurnal temperatures (see below) may select against diurnal emergence.

While speculative, the largely diurnal emergence schedules observed in this study may be a locally or regionally adapted response to reduce raccoon depredation, in parallel with explanations of crepuscular or nocturnal hatchling emergence in avoiding avian predators (e.g., for sea turtles, Hendrickson 1958 and Mrosovsky 1968; for Apalone mutica,Plummer 2007). Given that nest site fidelity has been established for several chelonians, including Graptemys (Freedberg et al. 2005), the development of locally adapted responses to prevailing predation pressures is theoretically plausible, although the comparative studies of emergence schedules and predator activity across turtle ranges necessary to adequately test this hypothesis are not yet available.

Another factor potentially enabling diurnal emergence at our study site may be relatively cool substrate temperatures compared with those at more southern locations, due to periods of shade from nearby trees and seasonally cooler climate. Excessive diurnal substrate temperatures are believed to inhibit nest emergence in sea turtles (e.g., Hendrickson 1958; Moran et al. 1999) and may be largely responsible for the nocturnal emergence of Graptemys hatchlings in southern locales (e.g., see Lahanas 1982). In these contexts, predation risks may have less influence on emergence timing than does temperature. Daytime emergence may be a default strategy for Graptemys whenever thermal constraints are not present or when predation risks are not excessive given the largely diurnal patterns of both adults and emergent hatchlings in this genus (Ernst and Lovich 2009; G.A.G., pers. obs.; respectively).

Influence of Temperature and Rainfall on Time to First Emergence. — Most first hatchling emergence in this study occurred in the morning, soon after temperatures at the nest level and near the surface began to rise after nighttime lows, increasing thermal regimes may serve as reliable proximate cues to the start of diurnal periods. Assuming preemergent hatchlings are at resting positions awaiting adaptive cues before emergence (e.g., Gibbons and Nelson 1978; Gyuris 1993; Plummer 2007), changing nest environment temperatures are likely detected at these locations at the time, rather than via awareness of gradients (temperature differences between the surface and subsurface), per se. Both rates of temperature change and thermal gradients can be resolved into temperature change models of hatchling emergence because both describe thermal energy moving through the soil column imparting a temperature change at any given depth; hatchlings may respond to rates of temperature change or simply overall temperature trends over longer periods (after Doody et al. 2001). Although less frequent, some first emergence events also occurred during periods of declining soil temperatures later in diurnal periods, and during the nighttime, showing that hatchling response was not absolute to a given directionality of temperature change.

First emergence events in this study occurred over a wide range of moderate nest-depth temperatures and without evidence of critical thermal thresholds, based on results of binomial regression and inspection of the emergence histogram (Fig. 7; after Moran et al. 1999). This suggests that the particular temperatures encountered during the typical morning emergence periods, in themselves, neither inhibited nor prompted hatchling emergence for G. ouachitensis on this site. Our observations thus align with other research in support of a temperature change model of hatchling emergence, rather than those based on absolute temperature thresholds, which, for example, may passively trigger nocturnal emergence in sea turtle hatchlings by declining to a point where hatchlings are physiologically able to coordinate muscle movements (e.g., Moran et al. 1999; Drake and Spotila 2002). Regarding freshwater turtles for example, Doody et al. (2001), discounting a thermal threshold model, found the nocturnal emergence of Carettochelys insculpta to be associated with declining temperatures of ∼ 0.2°C/hr to 1.2°C/hr at nest depth (nest depths and temperature change rates similar to this study). However, within Graptemys, there was evidence for thermal inhibition of G. nigrinoda hatchling emergence on a warmer nesting area than the present study, in Alabama, which largely limited emergence to nocturnal periods. Temperature change and temperature threshold models are not mutually exclusive, and elements of each may be present in some contexts (Doody et al. 2001; Plummer 2007).

There was an increased likelihood of hatchling emergence to occur within 24 hrs of higher rainfall amounts and higher ambient temperatures, but most hatchling emergence in this study took place without precipitation at the time or within 24 hrs of previous rain. Some early studies considered rainfall, rather than temperature, as the primary stimulus for freshwater hatchling turtle emergence, with many emphasizing a functional role in substrate softening (e.g., Hartweg 1944; Hammer 1969; Moll and Legler 1970). More recently, Lovich et al. (2014) also noted the probable role of precipitation in allowing hatchling emergence from hard soils during drought years. As in most studies, however, our results suggest that rainfall is not necessary for nest exit, nor, as a single factor, is it strongly associated with emergence activity (e.g., Gibbons and Nelson 1978; Baker et al. 2013; Gibbons 2013; for Graptemys, Lahanas 1982). Limited association of hatchling emergence activity and precipitation may be likely in cases where nesting substrates are sandy and friable and present little hindrance to emergence, and when the nests themselves do not have hard “nest plugs” (Burger 1976; DePari 1996), both of which apply in the present case.

Our finding that the probability of first hatchling emergence was more likely on days with increased amounts of rainfall parallels that of Tucker (1999), who noted that the largest numbers of Trachemys scripta hatchlings caught at drift fences were after periods of rainfall, especially when concurrent with rising daytime temperatures in the spring. Burger (1976) also found hatchling activity to be associated with relatively warm days for Malaclemys terrapin, but did not find an association with precipitation. The interaction of both minimum daily air temperature and rainfall amounts on the day before emergence was important in a study by Nagle et al. (2004), who found the greatest number of emerged G. geographica hatchlings in the morning at temperatures ≥ 12°C following ≥ 0.5 cm of rain. Our finding that both ambient air temperature and precipitation were positively related to the probability of emergence and to the number of nests with first emergences largely agrees with an emerging consensus on the influence of temperature and significant rainfall on emergence of emydid hatchlings.

Emergence Synchrony. — Until this study, Graptemys hatchling emergence had been reported as mostly or completely asynchronous (Lahanas 1982; Seigel and Brauman 1994; Baker et al. 2010, 2013). These records contrast with our finding of largely synchronous emergence for G. ouachitensis. However, whether this disparity results from general difficulties in deriving data on this life history aspect (Doody et al. 2001) or represents an actual difference within Graptemys, is unclear.

Our novel observations suggest a potentially important area of future research. In contrast to sea turtles that utilize synchronous, mass emergence from nests to lower the predation risk by effect of predator swamping (e.g., Santos et al. 2016) and as a mechanism to facilitate exit from deep nests (Carr and Hirth 1961), the cost–benefits for most freshwater species are less established. Recent literature reviews indicate that evidence in support of proposed fitness benefits of synchronous emergence in freshwater turtles, largely reduced predation risks, is scarce (Colbert et al. 2010; Baker et al. 2013; Riley et al. 2020; however, see Tucker et al. 2008). In fact, hatchling emergence asynchrony may be a better predation reduction strategy in the typical case when generalist predators (e.g., raccoons) that use prey-switching foraging methods based on relative prey abundance are predominant (Ims 1990; Testa 2002). Colbert et al. (2010) suggest that synchronous hatching (facilitating synchronous emergence) may be an ancestral trait in chelonians, but its presence today may not be strictly tied to predator avoidance.

Propensity for asynchrony also may be explained by differential conditions within the nest chamber that influence embryonic development rates within a clutch (e.g., thermal gradients increase development rates for upper eggs relative to those at the bottom), with most asynchrony expected for shallow nests containing multiple layers of eggs (Houghton and Hays 2001; Colbert et al. 2010). Thus, the preponderance of synchronous emergence in this study may result from relatively uniform conditions within nests due to limited temperature differentials between eggs typically arrayed in only 2 or 3 layers (Lindeman 2013), in accord with our not finding an association of emergence duration with vertical clutch dimension in these generally small clutches (mean = 7.2 eggs). However, the reason that asynchrony is more common for the even smaller clutches of G. flavimaculata (4.7 eggs/clutch, n = 134 and mean surface-to-uppermost egg depth of 92 mm, n = 50; Horne et al. 2003) and G. nigrinoda (mean 5.5 eggs/clutch, n = 8; Lahanas 1982) remains unclear.

Intriguingly, the newly discovered ability of some freshwater turtle hatchlings to produce sounds before emergence (Ferrara et al. 2012), including widespread clicking and other sounds by preemergent G. ouachitensis (Geller and Casper 2019), raises the possibility of acoustic communication in coordinating hatching and emergence synchrony. Further investigations are needed on the taxonomic extent of sound production by hatchling chelonians and the possible functions these sounds may have in emergence timing. Despite the theoretical importance of varying degrees of synchrony on hatchling survivorship, much remains to be learned about the factors that shape hatchling emergence patterns in freshwater turtles.