Climate effects on nesting phenology in Nebraska turtles

Abstract A frequent response of organisms to climate change is altering the timing of reproduction, and advancement of reproductive timing has been a common reaction to warming temperatures in temperate regions. We tested whether this pattern applied to two common North American turtle species over the past three decades in Nebraska, USA. The timing of nesting (either first date or average date) of the Common Snapping Turtle (Chelydra serpentina) was negatively correlated with mean December maximum temperatures of the preceding year and mean May minimum and maximum temperatures in the nesting year and positively correlated with precipitation in July of the previous year. Increased temperatures during the late winter and spring likely permit earlier emergence from hibernation, increased metabolic rates and feeding opportunities, and accelerated vitellogenesis, ovulation, and egg shelling, all of which could drive earlier nesting. However, for the Painted Turtle (Chrysemys picta), the timing of nesting was positively correlated with mean minimum temperatures in September, October, December of the previous year, February of the nesting year, and April precipitation. These results suggest warmer fall, and winter temperature may impose an increased metabolic cost to painted turtles that impedes fall vitellogenesis, and April rains may slow the completion of vitellogenesis through decreased basking opportunities. For both species, nest deposition was highly correlated with body size, and larger females nested earlier in the season. Although average annual ambient temperatures have increased over the last four decades of our overall fieldwork at our study site, spring temperatures have not yet increased, and hence, nesting phenology has not advanced at our site for Chelydra. While Chrysemys exhibited a weak trend toward later nesting, this response was likely due to increased recruitment of smaller females into the population due to nest protection and predator control (Procyon lotor) in the early 2000s. Should climate change result in an increase in spring temperatures, nesting phenology would presumably respond accordingly, conditional on body size variation within these populations.

Earlier reproduction has the potential to severely disrupt an organism's life cycle. On the positive side, for temperate animals it might permit the production of additional clutches or broods, and neonates may have more time to feed and grow in the fall prior to their first winter (Carroll & Ultsch, 2007;Rhen & Lang, 1999;Schwanz & Janzen, 2008;Tucker et al., 2008). However, on the negative side, early reproduction might expose eggs or neonates to atypical or mismatched conditions, increasing mortality rates (Benard, 2015;Jara et al., 2019;Muir et al., 2012;Pike et al., 2006;Saino et al., 2011).
The potential impact of earlier nesting is especially complicated for species that exhibit temperature-dependent sex determination during development in the nest, as most turtles do (Janzen & Paukstis, 1991). Hence, understanding the impacts of climate on the reproductive phenology of turtles (and other organisms) is critical to conservation and management in the face of climate change, but also because turtles are among the most endangered organisms on the planet (Stanford et al., 2020).
We have been studying the reproduction and demography of turtle populations in western Nebraska since 1981 (Iverson, 1991;Iverson et al., 1997;Iverson & Smith, 1993). For this study, we sought to examine the effects of climate variables on the nesting phenology in two species: Common Snapping Turtle (Chelydra serpentina) and Painted Turtle (Chrysemys picta). Detailed descriptions of the reproductive biology of these two species at this site have been previously reported (Iverson et al., 1997;Iverson & Smith, 1993). Both species nest annually in late May to late June or early July, but the timing of nesting varies among years by as much as two weeks in Chelydra and over a month in Chrysemys (see below). That variation is likely related to variability in weather, but the specific climatic variables that drive nest timing, and how those variables might be changing over time, have not been studied at this site.
Preliminary data from our site suggested that cooler springs delayed the onset of nesting in turtles (date of first nest only ;Janzen et al., 2018). For this expanded study, we predicted that spring temperatures would be inversely correlated with the Julian date of the first nest produced each year and the mean date of that first clutch.
We also explored the potential effects of monthly precipitation and mean monthly maximum and minimum temperatures during the previous summer, autumn, and winter, when females are undergoing vitellogenesis of the clutch produced the following year (Rollinson et al., 2012). We hypothesized that warmer autumn conditions might contribute to more complete follicle development before winter, and hence, the production of earlier clutches the next spring. In addition, warmer conditions in winter (e.g., Mitchell et al., 2017) or spring (Edge et al., 2017;Janzen et al., 2018) were expected to advance nesting phenology during the following season. We also investigated whether body size, clutch size, or egg size affected nest timing, speculating that larger turtles or those with relatively large eggs or clutches might nest earlier in the season.
Finally, given the deep continental location of our study site, and the finding that climate change is generally progressing more rapidly in continental versus coastal North America (Loarie et al., 2009), we expected turtle nesting at our site to have advanced in time over the course of our long-term study. Furthermore, that advancement should be more evident than for populations farther east.

| Data collection
We monitored nesting turtles that emerged from Gimlet Lake (41°45.24′N, 102°26.12′W), a shallow, sandhill lake on the Crescent Lake National Wildlife Refuge, Garden County, Nebraska, USA (see Iverson & Smith, 1993 for study site description) during 18 (Chelydra; Table 2) or 23 years (Chrysemys; Table 3). The primary nesting areas were monitored daily during the nesting season (May-July) from at least 06:00 to 22:00 hr by two to five observers. Turtles were weighed, measured (maximum carapace length and maximum plastron length), and marked after nesting. It was not possible to monitor both species for the entire nesting season for every year between 1986 and 2017, but data were available for most years (Tables 2 and   3). In some years, we also sampled Chelydra that nested at nearby Island Lake (41°43.95′N, 102°24.16′W).
For each species, we recorded the date in May or June each year that the first gravid female of each species emerged from the lake with the intention of nesting (i.e., gravid and attempted or completed a nest). Additionally, we calculated the mean nest date each year for all emergence dates (even if a nest was not completed that day). For females that failed to complete a nest when first sighted in a given year (e.g., if she was disturbed by Refuge personnel activities) and then nested on a subsequent night during the following several days (i.e., before she could produce a second clutch), her nest was scored as having been deposited on the night she was first observed constructing a nest.
Chelydra produced a maximum of one clutch per year at this site (Iverson et al., 1997), but some female Chrysemys produced at least three clutches per season (Iverson & Smith, 1993 the latter species, mean nest date refers only to the first clutch of the season. The end of production of first clutches for Chrysemys was estimated by assuming that at least ten days are required to produce a second clutch by a given female (though usually 12 or more days; Iverson & Smith, 1993) and noting the dates for females known to be depositing their second clutches. The daily frequency of nesting females in the interval between 11 days after the first nest date and the date of the first known second clutch was examined for a gap (or at least a greatly reduced nesting frequency) that was presumed to indicate the transition between first and second clutches. The last day for a first clutch was estimated to be the last day before that gap. We realize that this method in imprecise and likely excludes some females that produced their first clutches, while most of the population was producing second clutches, but our sample sizes should be large enough to minimize this potential bias.

| Statistical approach
We investigated whether nest deposition (first nests per season, mean nests, and nest dates by individuals) for Chelydra and Chrysemys was influenced by climatic predictor variables and whether they changed over time during our study period. Additionally, we assessed whether population-level measurements of body size and reproductive variables (carapace length, plastron length, female mass postnesting, mean egg mass, and clutch size) have changed during our study, as these variables can influence timing of nest deposition of an individual.
We analyzed our data two ways. First, we conducted least squares regression analyses to assess long-term trends in climate variables versus year and relationships between temperature and precipitation variables and time (Julian day of nest deposition and years) for the first nest of a season and mean nest dates. Significant p-values for regressions were conservatively adjusted for multiple comparisons by using a sequential Bonferroni correction to α (Holm, 1979). We conducted these analyses using Statview software (Abacus Concepts).
Second, we assessed relationships between climatic variables and the above-mentioned life history traits on nesting phenology (Julian day of nest deposition) of all nests. For these analyses, we fit linear mixed-effect models via maximum likelihood in R using the package "lme4" (Bates et al., 2015;R Core Team, 2017). We evaluated candidate models using the Akaike information criterion (AIC) where the level of importance was assessed by model weights (w) and overall ranking in the candidate set. For both species, we included carapace length as a fixed effect to account for an individual's growth and size over time.
For Chrysemys, we used the random effects of Female ID (identification) to account for individual variation and year to account for differences in sample sizes. And for Chelydra, we used year as the random effect, but not female ID because 76% of nests could not be associated with a female ID (see below).
To improve model convergence and determine relationships with nest foray dates, we z-standardized the continuous covariates for our mixed-effect model analyses. We examined relationships between our covariates and dropped one of two variables if their Pearson's correlation coefficient was >|.70|, with one exception; mean minimum and maximum air temperatures are correlated, but serve important, separate roles in regulating water temperatures and metabolic rates of freshwater turtles (however, minimum and maximum means for a given month were not included in the same models because of collinearity).
In total, we had 33 climatic variables for our analyses. Therefore, we did not develop multiple regression models (additive or interactive) to assess which combination of variables in a model was "best" at predicting nesting phenology. Instead, we sought to evaluate how each climatic variable influenced timing of nesting, including how a variable ranked in importance among all candidate variables, and the magnitude of its effect. We determined the covariate's predictive importance by inspecting conditional beta coefficient (β) estimates and their 95% confidence intervals (CI), with significance defined as CIs for a variable that did not overlap zero. We evaluated predicted values of our significant variables to assess relationships with Julian day of nesting using the "ggeffects" package in R (Lüdecke, 2018).

| Nesting summary
We tallied 705 nesting forays for Chelydra and individually identified 230 (23.6%) females, although many of the others were marked but eluded us after nesting by returning to the water before capture.
Of the known females, 45 were only recorded on a nesting foray once during the study, but the average number of years that known females emerged to nest was 2.3 (range 1-8). We found no differences in nesting dates, changes in nesting dates over time (years), or body size metrics between Gimlet and Island Lakes, except for clutch size, our top model, which was significantly smaller at Island Lake (Tables S1-S4). However, sample sizes were disproportionate, with 204 nests from Gimlet Lake and 58 from Island Lake, and likely influenced this result. Therefore, we merged datasets of both lakes for climatic variable analyses.
We also recorded 981 total forays for first nests for Chrysemys and associated 503 of those to known females (51.3%). Of the known females, 160 were identified on a first clutch nesting foray only once during the study, but the average number of years that identified females emerged on forays for first nests was 1.9 (range 1-10). As the study progressed, clutch size for Chrysemys remained unchanged, but there was a significant increase in carapace length of nesting females (Tables S3 and S4).
Julian day of nest deposition was highly variable among years (e.g., Figure 1).  Table 3). Date of the first nest and mean date of nesting within years were highly correlated (p < .0001) for both Chelydra (R = .82) and Chrysemys (R = .93), although the first nesting dates in a given year between these two species were not correlated (N = 17; R = 0.38; p = 0.14) nor were the mean dates (N = 14; R = 0.34; p = 0.24).

| Climate summary
Annual precipitation at this site averaged 43.3 cm between 1970 and 2017, and wet season (May-June) rainfall averaged 14.9 cm.
However, no measure of precipitation (monthly, seasonal, or annual) changed significantly with time over those 48 years (p > 0.17 for all regressions). In contrast, mean annual temperature at our study site has warmed at a rate of about 0.5°C per decade (Figure 2). Mean daily minimum temperatures for every month of the year except February and December increased significantly from 1970 to 2017 (Table 4).
However, mean daily maximum temperatures increased significantly only for January (p = 0.037), but only if no adjustment in that p-value was made for multiple comparisons (Table 4). Mean April-May temperature also did not change over that period (Figure 2), although mean September-October temperature increased significantly, by about 0.5°C per decade (Figure 2).

| Climate effects
Based on our mixed model analysis, variation in the nesting date by year for Chelydra was best explained by mean May minimum temperatures ( For all Chelydra nests (mixed model analyses), eight of our climatic variables significantly influenced nest deposition (Tables S5 and S6).
Notably, an increase in mean May minimum (6.1 to 9.6°C) and mean

| Life history effects
Body size and reproductive variables (carapace length, plastron length, clutch size, and spent mass of a postnesting female) were all significantly inversely correlated with nesting dates (Tables S9 and S10

| Temporal effects
Least squares analyses of first and mean nesting dates for both    (Tables S3 and S4). However, for Chrysemys, mean annual carapace length decreased over time ( Figure 6). In contrast, our mixed model analysis suggested that carapace length in Chrysemys increased over time (Tables S3 and S4), although the latter results are complicated by the uneven annual sample sizes (Table 3) and the clear trend of an increase in body size over the last third of the study (Figure 6) during the years with large sample sizes (Table 3). Furthermore, the relationship between body size and nest date over the full study period in Chrysemys was confounded by population-level demographic changes due to variation in female mortality and nest survivorship ( Figure 6; see discussion for details). At our site, warmer springs also advanced nesting in Chelydra, as did increased mean December maximum temperatures. The mechanisms driving this pattern likely operate through thermoregulation or local food chain productivity (Schwanz & Janzen, 2008). Increased local environmental temperatures during the winter and spring presumably permit earlier emergence from hibernation, increased metabolic rates (e.g., via basking), and accelerated vitellogenesis, ovulation, and egg shelling, all of which would drive earlier nesting (Mitchell et al., 2017;Obbard & Brooks, 1987). Similarly, an increase in local food chain productivity due to increased temperatures could also provide more resources necessary to speed up reproductive demands, although this mechanism is probably secondary to thermoregulatory affects. However, nest timing in Chrysemys was not strongly influenced by spring temperatures, as predicted, but rather, delayed by warm temperatures in the fall and winter. Rollinson et al. (2012) demonstrated that snapping turtles complete vitellogenesis primarily by the end of the previous fall, whereas

| D ISCUSS I ON
for Chrysemys the process occurs both in the fall and spring (see also Callard et al., 1978).

TA B L E 5
Spring climate variables correlated with nest date for two turtle species at Crescent Lake National Wildlife Refuge, Garden County, Nebraska which presumably lag air temperatures. Interpreting differences between the two species is further complicated by the much higher propensity of Chrysemys to bask aerially compared to Chelydra.
Indeed, basking in Chrysemys may explain why nest timing in that species was so much less predictable by air temperatures than in

Chelydra.
We were surprised to find that increased precipitation in July was correlated with a delay in nesting in Chelydra, over ten months later. Although previous studies have examined the impact of precipitation on nest timing during the nesting season (see review in Czaja et al., 2018), no study has examined the effects on nest timing of precipitation outside the nesting season. Increased precipitation in July at our site was correlated with colder mean daily July maximum temperatures (R = .38; p = .007; N = 48 years), but average July temperatures were not related to nest timing in Chelydra. We can therefore only speculate that increased precipitation in July delays nesting the following year by slowing vitellogenesis, perhaps via its effect on lowering water temperatures. The importance of April precipitation in delaying nesting in Chrysemys was surprising.
We suspect that high precipitation in April reduces basking opportunities and decreases water temperatures, both of which would be expected to delay nesting in Chrysemys, which must complete vitellogenesis in the spring (Rollinson et al., 2012).
Our analyses revealed that larger female Chelydra and Chrysemys tended to nest earlier in the season than smaller females. Because clutch size and egg size are correlated with body size in both species (Iverson et al., 1997;Iverson & Smith, 1993), early nests included more and bigger eggs. Earlier nesting by larger female turtles has previously been reported for Graptemys geographica in Pennsylvania (Nagle & Congdon, 2016). These results suggest that the size class distribution of a population can impact its nesting phenology, complicating phenology comparisons across years, populations, and species.
Although mean body size of nesting female Chelydra did not vary over time at our study site (R = −.46; p = .07), mean annual carapace length of nesting female Chrysemys did decrease significantly F I G U R E 3 First Julian nesting date of Chelydra serpentina from 1993 to 2017 at Crescent Lake National Wildlife Refuge, Garden County, Nebraska, in response to mean May minimum daily temperature (TOP) and mean December maximum daily temperature (BOTTOM). For May temperatures, regression is statistically significant before and after sequential Bonferroni adjustment (p = .0006; see Table 4); for mean Julian nesting date y = −1.899x + 250.642, R 2 = .412, and p = .004. For December temperatures, regression is statistically significant before but not after sequential Bonferroni adjustment (p = .0041; see Table 4); for mean Julian nesting date y = −1.260x + 177.987, R 2 = .311, and p = .0162  Figure 6). In conjunction with our overwinter physiology studies (e.g., Costanzo et al., 1995), we began sporadic protection of Chrysemys nests in 1993, followed by rigorous protection of every located nest commencing in 1999 and continuing through 2017. This effort flooded the nesting population with small, primiparous females in the mid to late 2000s (e.g., see sample sizes in Table 3). In addition, the local raccoon ( Climate change over the last five decades has produced warmer temperatures overall at our site, with the greatest impact being a noticeable increase in nighttime minimum temperatures (Table 4).
It was also our subjective impression that nighttime skies grew increasingly hazy over the study period, and although the cause(s) are not yet clear, the increasingly cloudy skies and warmer nights were likely related. Despite the significant overall warming at our site over the past several decades and a clear inverse relationship between spring temperatures and nesting timing, nesting phenology in at least Chelydra has not changed between 1993 and 2017 at our site. This is likely at least in part a reflection of the fact that spring temperatures at our site have not changed over that period (Table 5; Figure 3), even though annual temperatures have (Figure 2). The meteorological reasons for this spring difference are not yet evident. In any case, should spring day-time temperatures eventually warm at our site, the nest phenology of at least these two turtle species in western Nebraska will likely be affected.
Of the 38 studies summarized in Table 1  Regretfully, water temperatures were not recorded during our study, since they might be expected to be better predictors of nesting dates (e.g., as sea surface temperatures have been for marine turtles; Table 1). However, even those data would be complicated by differences in habitat use by our study species. In our experience, Chelydra seems to occur in shallow (warmer?) water and does limited aerial basking, whereas Chrysemys seem to inhabit deeper water and exhibits extensive aerial basking (see also Ernst & Lovich, 2009).
Similarly, more detailed analyses of temperatures (water and air) beyond simply monthly means (especially during the spring and fall temperature windows that are most highly correlated with nest timing) might clarify the mechanism for the relationship between temperature and nest timing more precisely (e.g., see Edge et al., 2017;Schwanz & Janzen, 2008).
It is also possible that the inability to detect a change in the nesting phenology at our site, as well as many other sites in Table 1, could be due to the stalling in increasing global mean surface temperatures from the late 1900s through the 2000s known as the "climate change hiatus" (e.g., Kosaka & Xie, 2013). Steady rather than increasing temperatures during that period could explain the lack of statistically significant change in nesting phenology in studies including data collected during that period, although this hiatus is not evident in our climate histories (Figures 2-4).
As previously noted, only 7 of 23 studies that evaluated turtle nesting phenology over time have documented that nest dates have advanced over recent decades. Clearly, the collection of more data is necessary before we can generalize that climate change has altered nesting phenology in nonmarine turtles. Part of the problem is that such studies depend on demanding long-term studies. For example, of the 38 studies reviewed in Table 1, only three field sites (Ontario, Canada, and Illinois and Nebraska, USA) have recorded nesting histories exceeding 20 years (see also Janzen et al., 2018). Thus, our ability to detect long-term changes in nest timing in many turtle populations may be constrained by sample size, speaks to the value of long-term studies, and argues for the continuation of those currently in place.
Furthermore, among sea turtles, most of the nesting phenology research done to date has focused on only two species (with complicated migratory cycles), while most of the work done on nonmarine turtles has focused on Chelydra and Chrysemys (