Long‐Term Incubation Duration Decline Indicates Climate‐Change Driven Feminization of Three Sea Turtle Species in Florida, USA

ABSTRACT

Climate change is altering the thermal environment of nesting beaches worldwide, threatening species with temperature‐dependent sex determination (TSD) such as sea turtles. While models have predicted feminization of primary sex ratios—that is, a progressive increase in the proportion of females—empirical, population‐scale evidence across multiple species remains rare. Here we present the first broad‐scale, multi‐species evidence of long‐term changes in incubation duration (ID)—used as a proxy for temperature and primary sex ratio—across genetically distinct Management Units (MUs) of loggerhead ( Caretta caretta ), green ( Chelonia mydas ), and leatherback ( Dermochelys coriacea ) turtles nesting in Florida, USA. We introduce a simple, scalable method to assess population‐level feminization trends by identifying directional shifts in ID distributions over time, avoiding the uncertainty of model‐based primary sex ratio estimates. Using data from over 110,000 clutches laid between 2001 and 2022, we document significant declines in ID, spatial variation in embryo mortality across MUs, likely associated with greater exposure to lethal incubation temperatures, and the presence of seasonal and geographic male‐producing refugia. These findings provide robust empirical evidence of increasing feminization, early signs of temperature‐related lethal effects in at least one region, and highlight the importance of MU‐scale, species‐specific monitoring. This study underscores the need to protect male‐producing beaches and early‐ and late‐season clutches, which may be disproportionately vulnerable or overlooked. Given the simplicity and accessibility of ID data, we encourage its broader use in sea turtle conservation and recommend applying our approach to detect climate‐driven trends in incubation conditions and potential feminization across other rookeries.

We analysed more than 110,000 sea turtle nests from three species across Florida to track climate‐driven changes in incubation duration, a simple proxy for hatchling sex. Our results reveal widespread shortening of incubation periods and thus increasing feminisation, but also identify geographic and seasonal refuges where more males are produced. This accessible, non‐invasive approach offers a powerful way to detect early climate impacts and prioritise conservation of key sites.

1. Introduction

Climate change represents a major conservation concern, from entire ecosystems to individual species (Parmesan 2006; Parmesan and Yohe 2003). Sea turtles have temperature‐dependent sex determination (TSD) (Yntema and Mrosovsky 1980, 1982), and the sex is determined during the middle third of incubation (Miller 1985; Yntema and Mrosovsky 1982). High incubation temperatures produce all females, low temperatures produce all males, and intermediate temperatures produce a mix of the two sexes (Mrosovsky 1994). Soon after its discovery, TSD was recognized as an additional vulnerability for the taxon, especially in a context of global warming (Davenport 1989; Mrosovsky 1994; Mrosovsky, Dutton, and Whitmore 1984; Mrosovsky and Provancha 1989), because climate change is expected to increase incubation temperature and, thus, may lead to the production of an insufficient number of males to maintain a viable population. If primary sex ratios (PSR; i.e., the sex ratio of offspring) are extremely female‐skewed, low male production (and subsequent reduction of male availability) may threaten long‐term population viability (Mrosovsky and Provancha 1992). Another concern of global warming is that incubation temperatures may become incompatible with embryo development, causing a reduction of hatchling production (Bladow and Milton 2019; Howard et al. 2014; Laloë et al. 2017; Matsuzawa et al. 2002; Santidrián Tomillo et al. 2014).

Since the 1980s, a growing number of studies have focused on feminization of hatchlings (i.e., a progressive increase in the proportion of females) as a potential threat, especially through modeling approaches aimed to predict future scenarios (Hawkes et al. 2007; Monsinjon et al. 2019; Tanner et al. 2019; Witt et al. 2010). Most studies estimating current primary sex ratios have reported high female biases (Hays et al. 2014), and predictions accounting for climate change forecasts suggest many sites will stop producing males soon (Hawkes et al. 2007; Monsinjon et al. 2019; Tanner et al. 2019). Such a dooming prediction is tamed by four considerations. First, empirical studies have shown that moisture has a masculinizing effect on embryos that can partially contrast the feminizing effect of high temperatures (Wyneken and Lolavar 2015); therefore, predictions based exclusively on environmental temperature may overestimate the problem. Second, theoretical and empirical evidence show that operational sex ratios (OSR; the sex ratio of adults reproducing in a breeding season) (Maurer et al. 2021) are less female skewed than primary sex ratios because males reproduce more often (even every year) and may mature earlier than females (reviewed by Hays et al. 2022). Third, females might respond with a range shift and nest in more climatically suitable areas (Abella Perez et al. 2016; Mainwaring et al. 2017) that will continue to produce male hatchlings. Fourth, turtles might respond with a phenological adaptation that is, shifting their nesting activity to cooler periods (Pike et al. 2006; Weishampel et al. 2010, 2004), although recent modelling indicates that even the most extreme phenological shift reported so far could only partially mitigate the problem (Laloë and Hays 2023) and that the phenological shift required to maintain present‐day primary sex ratio is too large to be feasible at many locations (Fuentes et al. 2024).

Empirical evidence of the impact of global warming on sea turtles is scarce because detecting such effects at a population level requires extensive sampling over decades. Several different approaches exist for obtaining primary sex ratio estimates, varying in the degree of difficulty and efficacy. The most accurate primary sex ratio estimates rely on direct observation of gonads via histology (Merchant Larios 1999; Wibbels 2003). To investigate global warming effects on the primary sex ratio trend (i.e., monitoring feminization of a population) using this methodology would require sacrificing many hatchlings per nest from many nests each year to encompass seasonal variation in the annual primary sex ratio and then repeating this sampling for many years. Such an approach is both laborious and unrealistic due to ethical and conservation reasons. Estimates obtained with other methods are all undermined by possible errors and biases, mainly sample bias. Despite the ethical concerns, a few studies still sacrifice a subset of hatchlings per clutch (n = 2 to 10 hatchlings/clutch) that are then sexed histologically (Godley et al. 2002; LeBlanc et al. 2012; Rebelo et al. 2012), but such a small sample may be easily biased and not provide an accurate sex ratio estimate of the entire clutch. Moreover, since the approach requires lethal take, it is usually applied to a small number of clutches (n = 10 to 40) that may not be a good representation of all the clutches laid across the nesting season.

Additional direct methods for estimating the primary sex ratio have been proposed. One approach uses laparoscopy for direct observation of gonads and requires rearing hatchlings in captivity until they reach 120 g, the minimum weight needed for this surgical procedure (Wyneken et al. 2007). More recently, blood sampling from live hatchlings has been used for hormone assays (Tezak et al. 2020). A promising new approach, developed specifically for hatchlings of the American alligator, relies on DNA methylation patterns in blood samples to accurately distinguish males and females (Bock et al. 2022). However, although sex‐specific methylation differences have also been described in sea turtles (Mayne et al. 2023; Venegas et al. 2016; Yen et al. 2024), a comparable non‐lethal methylation‐based sexing assay has not yet been developed for sea turtle hatchlings. Such approaches requiring laparoscopy or blood sampling are considered invasive and are impractical for large sample sizes. They are typically applied to a relatively small number of clutches, which increases the probability of misrepresenting the primary sex ratio of a large population across the entire nesting season. A non‐invasive method based on the hormone levels in the amnion and allantois remains in the eggshell has also been proposed (Gross et al. 1995). Another non‐invasive approach consists in the direct observation of the gonads of hatchlings that are found dead in the nest during post‐hatching inventories (Kaska et al. 2006; Schmid et al. 2008; Wibbels et al. 1999). Although it avoids the ethical concerns of invasive techniques, when applied at the nest level, this method may lead to inaccurate estimations of the primary sex ratio if no more than a few dead hatchlings are typically found in a clutch. However, in the case of estimating a primary sex ratio for an entire nesting beach, it may be possible to partially compensate for this problem by examining large numbers of clutches (Wibbels et al. 1999). Moreover, the sex ratio of dead hatchlings may not represent that of live hatchlings due to potential differential mortality of pre‐emergent hatchlings between sexes (Witt et al. 2010), a hypothesis currently untested.

Non‐invasive indirect methods for estimating primary sex ratio are less accurate (Fuentes et al. 2017) and involve determining the temperature inside of clutches during the thermosensitive period for sex determination, determining general environmental temperatures (e.g., sand, air, sea surface), or determining incubation duration (ID) of clutches (reviewed by Patrício et al. 2021). Air temperature data have provided indications of long‐term increases in sand temperatures at sea turtle nesting sites (e.g., Hays et al. 2003; Laloë et al. 2024). Clutch or sand temperatures also provided important indications (e.g., Lamont et al. 2020; Stokes et al. 2024). However, they require technical and financial efforts (e.g., using dataloggers over the whole nesting area placed in the sand or inside clutches) that may limit sample sizes, and thus the ability to extrapolate results to a population in certain areas and contexts. In contrast to these temperatures, which are environmental variables assumed to cause a biological effect, ID, like primary sex ratio, is itself a biological effect strongly influenced by the incubation temperature (Ackerman 1997; Godfrey and Mrosovsky 1997; Godley et al. 2001; Marcovaldi et al. 1997). It is also relatively easy to determine ID, especially where sea turtle nesting activity is monitored for other reasons. The validity of using ID as a proxy for the primary sex ratio has been confirmed by histological sexing of hatchlings from clutches with known ID (Godfrey and Mrosovsky 1997; Mrosovsky et al. 1999). Moreover, ID can be used retrospectively (Mrosovsky et al. 1999). The main drawback of using ID is that it reflects the temperature during the entire incubation period, while sex determination occurs only in the middle third of the incubation period (Miller 1985; Yntema and Mrosovsky 1982). Moreover, other factors such as moisture may be involved in sex determination (Wyneken and Lolavar 2015). These confounding factors may reduce the accuracy of primary sex ratios estimated for individual nests, especially in the range of values where both sexes are produced, with more reliable predictions of primary sex ratio at short or long ID corresponding to all‐female or all‐male primary sex ratios, respectively (Mrosovsky et al. 1999). For instance, Mrosovsky et al. (1999) reported a 10% error in primary sex ratio prediction near pivotal incubation duration (PID; the incubation duration at which a 1:1 primary sex ratio is produced). However, incubation duration can be a suitable method for population‐level estimates (Mrosovsky et al. 1999) where the individual nest is not relevant. In fact, in a population‐level study with a large sample size, for the law of large numbers, errors at individual nest level (such as more estimated males or females than actually produced) will tend to average out, leading to more accurate and reliable overall primary sex ratio.

Surprisingly, this approach has been little used to investigate long‐term trends in primary sex ratios, despite the existence of several long‐term monitoring programs holding incubation duration data at the appropriate temporal scale. Notably, assessing ID trends suggestive of primary sex ratio trends does not require knowledge of the precise incubation duration—primary sex ratio relationship at the local level. To our knowledge, no study has yet provided empirical, long‐term evidence of increasing feminization of the primary sex ratio or declining hatching success due to rising incubation temperatures at the population level. One study provided evidence of a feminization trend in immature green turtles ( Chelonia mydas ) foraging in Bermuda (Meylan et al. 2024) that is likely due to a change of the primary sex ratio of the source populations, although it may also be due to a change of population mixture at that foraging ground. Another study suggested the feminization of an Australian green turtle population by comparing contemporary immature and adult sex ratios at foraging grounds (Jensen et al. 2018), although differential dispersal and mortality cannot be ruled out. Two other studies observed a decrease of ID (indicating the possibility of an increasing feminization of the primary sex ratio) in loggerhead turtle clutches ( Caretta caretta ) laid in single nesting sites in Greece (Margaritoulis et al. 2022) and in North Carolina, USA (Reneker and Kamel 2016), although a recent study revealed a more ambiguous pattern as the monitoring period increases (Ware et al. 2025). A recent study by Colman et al. (2025) provided valuable long‐term data on incubation duration and modeled sex ratios for a small, isolated leatherback turtle rookery in Brazil, supporting localized evidence of feminization. Other long‐term studies based on ID were carried out in Brazil on Caretta caretta and Eretmochelys imbricata but did not find evidence of increasing female biases (Dei Marcovaldi et al. 2014; Marcovaldi et al. 2016). Broad‐scale, empirical evidence of feminization across multiple species and populations is still lacking.

The Florida peninsula is a good candidate for such studies because it hosts important nesting grounds for three sea turtle species and is in the subtropical zone; therefore, its sea turtle nesting sites are expected to be more affected by global warming than sites at higher latitudes (Pike 2014). Florida represents the main nesting ground of the loggerhead turtle ( Caretta caretta ) Northwest Atlantic (NWA) Regional Management Unit (RMU; assemblage of breeding stocks of the same species sharing marine areas) (Ceriani and Meylan 2017; Wallace et al. 2023), with seven Management Units (MU; genetically independent nesting stocks of adult females; Moritz 1994) identified (Shamblin et al. 2011; 2012) and is a major nesting ground for this species globally, with an average of 97,447 clutches laid annually (Ceriani et al. 2019). Florida is also an important nesting ground for the green turtle Chelonia mydas North Atlantic RMU (Seminoff et al. 2015; Wallace et al. 2023), with at least four MUs identified (Shamblin et al. 2020), and for the leatherback turtle ( Dermochelys coriacea ) Northwest Atlantic (NWA) RMU (Stewart et al. 2011; Wallace et al. 2023). From as early as 1989, a female‐skewed primary sex ratio has been estimated for loggerhead turtles nesting in Florida and, in particular, in central‐east and southeast Florida (Hanson et al. 1998; Heppell et al. 2022; Mrosovsky and Provancha 1989, 1992). Concerns have been raised that too few males are being produced due to climate change, and the few males produced may not be sufficient to ensure population viability (Heppell et al. 2022; Wyneken and Lolavar 2015). A few studies have reported mixed sex ratio estimates on some beaches in SW Florida (Foley et al. 2000; Schmid et al. 2008), but those studies took place in the early 1990s and early 2000s, respectively, and temperatures have increased since then because of climate change (Runkle et al. 2022); thus, they may no longer produce a mixed primary sex ratio. In a more northern NWA loggerhead turtle nesting area (North Carolina, USA), past and future primary sex ratio scenarios were investigated through sand and air temperatures and predicted future feminization; therefore, raising even more concerns for southern areas like Florida (Hawkes et al. 2007). Recently, using a modeling approach, it was suggested that even a limited production of males in some periods may be sufficient for maintaining long‐term population viability (Heppell et al. 2022). Even though concerns of an extremely female‐biased primary sex ratio and lack of loggerhead male production from Florida beaches were postulated over 30 years ago (Mrosovsky and Provancha 1989), empirical evidence is still lacking, and the topic has not been investigated for green turtles and leatherback turtles, despite the importance of these regional rookeries in the Northwest Atlantic and wider Caribbean region.

The present study aims to use data from long‐term monitoring efforts that include data needed to determine ID and hatching success to investigate possible effects of global warming on several MUs of the three sea turtle species nesting in Florida (USA) and to derive conservation implications. This is pursued through eight specific objectives: (1) to test the hypothesis of increasing incubation temperatures experienced by turtle egg clutches, that could affect the primary sex ratio; (2) to test the hypotheses of increasing embryo mortality due to a greater occurrence of lethal incubation temperatures or of decreasing fertility due to lower numbers of males produced in the past (these two effects are confounded in field observations); (3) from the first test, to provide indirect insights about the hypothesis of phenological adaptation; (4) to identify areas or (5) periods of the nesting seasons where or when a higher proportion of males is produced, representing conservation priorities; (6) to provide a baseline for the three species that could be used to evaluate future impacts from climate change; (7) to provide a replicable approach to make spatio‐temporal comparisons based on incubation duration data.

2. Materials and Methods

2.1. Study Area and Sub‐Areas

The Florida Fish and Wildlife Conservation Commission (FWC) oversees sea turtle nesting monitoring in Florida across 228 coastal monitoring units (‘beaches’). A standardized data collection methodology through nest inventory was initiated in 2001 on 16 beaches (Brost et al. 2015) and then expanded, with at least 167 (73%) of the beaches included since 2016.

For this study, these beaches were assigned to different sub‐areas depending on the turtle species. For the loggerhead turtle, Florida was divided into seven subareas corresponding to genetically distinct management units (MU) as identified by Shamblin et al. (2011, 2012): (1) northeast (NE), (2) central‐east (CE), (3) southeast (SE), (4) Dry Tortugas (DRTO), (5) southwest (SW), (6) central‐west (CW) and (7) northwest (NW) (Figure 1a). For the green turtle, Florida was divided into eight subareas. Four subareas corresponded to the MUs genetically identified by Shamblin et al. (2015, 2020): (1) central‐east (CE), (2) southeast (SE), (3) Key West National Wildlife Refuge (MARQ), and (4) Dry Tortugas (DRTO). The remaining coastal tracts of Florida (not covered by the aforementioned genetic studies) were arbitrarily divided into four additional sub‐areas: (5) northeast (U‐NE), (6) west (U‐W), (7) northwest (U‐NW), and (8) Monroe County (all beaches in Monroe County north of Key West, U‐Monroe) (Figure 1b). For the leatherback turtle, no sub‐area division was deemed necessary based on current genetic knowledge (Stewart et al. 2011; Wallace et al. 2023).

FIGURE 1.

Map of sea turtle management units (MUs) and subareas in Florida. Panel (a) shows the seven MUs for loggerhead turtles ( Caretta caretta ) as defined by Shamblin et al. (2011): NE, CE, SE, DRTO, SW, CW and NW. Panel (b) shows the eight subareas for green turtles ( Chelonia mydas ), with four corresponding to genetically defined MUs (CE, SE, MARQ and DRTO) and four additional subareas (U‐NE, U‐W, U‐NW and U‐Monroe). Leatherback turtles ( Dermochelys coriacea ) nesting in Florida has not been further sub‐divided. Map lines delineate study areas and do not necessarily depict accepted national boundaries.

2.2. Data Collection

Data were collected during the period 2001–2022 by a network of individuals (including representatives of conservation organizations; local, state, and federal government personnel; academics; and consultants) who hold a permit from the FWC allowing them to conduct research and conservation activities on sea turtles on behalf of FWC. Surveyors at each beach site conducted daily clutch counts during the sea turtle nesting season, marked new clutches, recorded disturbances to existing clutches, and conducted clutch excavation no sooner than 3 days after hatchling emergence following procedures detailed by FWC (Florida Fish and Wildlife Conservation Commission 2016) to determine hatching and emergence success. Surveyors documented and excavated either all clutches or a spatially and temporally representative sample of clutches (for details see Brost et al. 2015). At excavation, variables relevant for this study were: number of whole eggs, damaged unhatched eggs, pipped eggs (hatchlings that have breached the eggshell but were still inside the egg), and hatched eggs (from fragments: see Ceriani et al. 2021), the total of which provided an estimated clutch size.

Disturbances were categorized as follows: predation, poaching, eggs scattered by another turtle, inundation (water inside the egg chamber at the time of nest excavation), accretion (accumulation of sand above the egg chamber observed by the surveyor at the time of clutch excavation), partial or complete washout (some or all eggs lost due to erosion and water action that removed them), and wave wash‐over (i.e., a previous high tide line was located landward of the clutch or direct observation of water at or beyond the nest).

2.3. Data Analysis

Only in situ and undisturbed clutches were considered for analyses described in Section 2.3.1, while analyses described in Section 2.3.2 also included clutches that were washed over (see definition above). Incubation Duration (ID) was calculated as the number of integer days between the dates of the morning surveys when the nest and the first emergence were detected. To reduce erroneous data due to human error, records with ID and clutch size outliers in the 0.5% percentile were removed. All analyses were conducted by species and by sub‐area if sample size allowed. Bayesian Generalized Linear Models (BGLMs, hereafter ‘model(s)’) were run through the package brms (Bürkner 2017) in R (R Development Core Team 2024). Priors for regression coefficients were left at the default settings used in brms, which are weakly informative to regularize estimates and prevent overfitting. Model convergence was assessed via the Gelman–Rubin diagnostic (R‐hat < 1.01) and effective sample sizes (i.e., Bulk Effective Sample Size (Bulk_ESS) and Tail Effective Sample Size (Tail_ESS) > 400), suggesting adequate mixing of chains. Model fit was assessed via the Bayesian p‐value (the proportion of posterior predictive means that are greater than or equal to the observed mean), with a value close to 0.5 indicating good fit (Gelman et al. 1996). A predictor was considered meaningful if the 95% credible intervals (rounded at two decimal places) did not include zero.

2.3.1. Temporal Change of ID, Embryo Mortality or Infertility

To remove spatial effects, only the 16 beaches monitored during the entire period (2001–2022) were considered in the temporal analyses (see Table 1 in Brost et al. 2015 for beach details) and assumed to be representative of the MU or subarea where they were located. Only in situ and undisturbed clutches were included in the analyses to remove any potential confounding factor.

TABLE 1.

Incubation durations (IDs) at the beginning and end of the study period (2001–2022) for species by management units (MUs) or subareas (see Figure 1) for loggerhead and green turtles. Leatherbacks nesting in Florida have not been further subdivided into MUs or subareas. When a meaningful negative effect of year (decreasing trend) was detected by a Bayesian General Linear Model (BGLM), the predicted ID value is given for the first (2001) and last year (2022) of the study period. When the BGLM did not detect a meaningful effect of year (indicated by the asterisk), the mean IDs during the first 5‐yr period (2001–2005) and the last 5‐year period (2018–2022) are given. The latter were significantly different (Mann–Whitney test; see text for details).

MU or subarea	2001	2022	N
Caretta caretta
Northeast (NE)	55.6*	54.4*	326
Central‐east (CE)	54.2	51.2	2866
Southeast (SE)	51.5	49.9	6917
Southwest (SW)	61.3	59.3	2008
Central‐west (CW)	56.2*	54.8*	3451
Northwest (NW)	59.6	56.3	1648
Chelonia mydas
Central‐east (CE)	55.4	50.5	2233
Southeast (SE)	52.2	50.3	1458
Dermochelys coriacea	66.6*	65.3*	407

To test the hypothesis of increasing incubation temperature over time, ID was used as a proxy that is expected to decrease in such a case (Mrosovsky et al. 1999). For each species, a model was run, with ID as the response variable and Year and Subarea as the explanatory variables (ID~Year + Subarea). Then, a model was run on each subarea separately (ID~Year).

To test the hypothesis of increasing embryo mortality or of decreasing fertility over time, the proportion of whole (unhatched) eggs observed at nest excavation was used as a proxy that is expected to increase in both cases. Indeed, the two cases are difficult to distinguish (Phillott and Godfrey 2020) and are here confounded. For each species, a model was run with the proportion of whole unhatched eggs as the response variable and Year and Subarea as the explanatory variables (Proportion_whole_eggs~Year + Subarea). Then, a model was run on each Subarea separately (Proportion_whole_eggs~Year). Where Year had a meaningful effect, the ID or the proportion of whole unhatched egg values of the first (2001) and last (2022) years were predicted from the model through the predict function. In cases where Year had no meaningful effect, possibly because of a small sample size, a Mann–Whitney test (α = 0.05) was conducted between ID or the proportion of whole unhatched eggs values from the oldest (2001–2005) and most recent (2017–2022) 5‐year periods. In the case of increasing embryo mortality due to increasing incubation temperature, a negative correlation between the proportion of whole unhatched eggs and ID (higher mortality at higher temperatures with shorter IDs) was expected. To assess whether such an effect of temperature on mortality was detectable in the study context and to facilitate the interpretation of the above analyses, the proportion of whole unhatched eggs (assumed to represent a proxy for embryo mortality) and incubation temperature (ID) were tested for correlation through a Kendall correlation test (cor.test function), which is more robust in the present context with many ties represented by ID (because ID is an integer and multiple nests have the same ID value). This test was run on the entire dataset of loggerhead turtles (all beaches), not just those used for temporal analyses.

2.3.2. Degree of Female‐Bias Primary Sex Ratio and Areas and Periods With Higher Male Production

To provide insights about the current level of male production and to identify areas or periods of the nesting seasons where or when a higher proportion of males is currently produced, clutches laid during the period 2016–2022 were considered. During this period the highest number of beaches (167 [73.2%] to 180 [78.9%] depending on the year) were monitored. Therefore, this dataset provides a more comprehensive spatial coverage. Considering that the frequency and timing of wash‐over events might affect incubation temperature, and thus incubation duration and possibly primary sex ratio, washed‐over clutches were included in the analysis. Clutches were categorized into three groups: (A) clutches with female‐only ID (FID) at which 0% of males are produced (ID range where males were never observed), (B) clutches with ID between FID and pivotal ID (PID, where 50% of sexes are produced), and (C) clutches with ID > PID, where a majority of males is produced. Such an approach is more robust than estimating exact primary sex ratio values, especially considering the non‐linear relation between temperature and primary sex ratio (e.g., Abreu‐Grobois et al. 2020). PID values were provided as such by original articles and rounded to the nearest integer, while integer FID values were extracted from figures or tables of those articles, as follows. For loggerhead turtles, FID was considered to be ≤ 51 days as determined by (1) field incubation duration in the USA (Godfrey and Mrosovsky 1997) and (2) incubation duration in the laboratory (i.e., until hatching) under high moisture conditions (Wyneken and Lolavar 2015), then converted to a field FID (i.e., until emergence instead of hatching) by adding 4 days (Godfrey and Mrosovsky 1997). Loggerhead PID was considered as 62 days (USA; Godley et al. 2001). No equivalent data are available from the study area for green and leatherback turtles. Therefore, FID and PID values were taken from the nearest populations with available data, under the assumption that incubation duration–primary sex ratio relationships are similar across populations of the same species. This assumption is partially supported by the broadly similar temperature–primary sex ratio relationships observed within species (Wibbels 2003). However, as the aim was not to estimate precise primary sex ratio values but rather to assess relative spatio‐temporal differences in likely male production, minor inter‐population differences are unlikely to affect the main conclusions. For green turtles, FID was considered to be ≤ 52 days, the lowest value observed in the Atlantic (Ascension Island: Godley et al. 2002; Suriname: Mrosovsky, Dutton, and Whitmore 1984). PID was conservatively (i.e., maximum ID with fewer males than females) considered as 60 days, which is the highest value among those observed in the Atlantic (Suriname, 59–60 days: Godfrey et al. 1996; Mrosovsky, Dutton, and Whitmore 1984). For leatherback turtles, FID was considered to be ≤ 62 days and PID was conservatively considered as 66 days, which is the highest value among those observed in Suriname (Godfrey et al. 1996; Mrosovsky, Dutton, and Whitmore 1984).

To estimate the degree of female‐biased primary sex ratio in Florida, the total number of Florida clutches belonging to each ID group (A, B or C) was estimated. For this extrapolation, the proportions observed in the study sample in each subarea were applied to the total number of clutches laid in that area (source: FWC‐FWRI Statewide Nesting Beach Survey Program database; Table S1).

3. Results

3.1. Temporal Change of Incubation Duration (As a Proxy of Incubation Temperature)

ID values were available for a total of 44,621 clutches of the three target species on 16 beaches monitored during the period 2001–2022. Of these, 27,957 (62.7%) clutches were retained for the analyses after removing clutches not left in situ, those that experienced disturbance, and those with ID values considered as outliers. The models showed a meaningful effect of MU on ID in both loggerhead turtles (n = 22,454) and green turtles (n = 4587); therefore, models were run for each MU separately. Year had a meaningful effect in four loggerhead turtle MUs (CE, SE, SW, NW), with ID decreasing by 1.8–3.3 days over time (Figure 2). Moreover, ID values from the other two MUs (CW and NE) were also significantly lower during the most recent 5‐yr period than during the first 5‐year period (Mann–Whitney test; CW n = 3451; NE n = 395) with differences of 1.2–1.4 days. Year had a meaningful effect in two green turtle MUs (CE, SE) (Figure 3) with ID decreasing by 2.3–5.0 days, whereas the other three MUs (U‐NE, U‐NW, U‐W) had too few data for an analysis. In leatherback turtles, no meaningful effect of Year on ID was detected by the model (n = 916; Figure 4). However, ID values were significantly lower during the most recent 5‐year period than in the first 5‐year period (Mann–Whitney test; n = 407) with a difference of 1.3 days. The predicted ID values for years 2001 and 2022 or the mean ID values during the first and last 5‐year periods for each MU are provided in Table 1.

FIGURE 2.

Annual distribution of incubation duration (ID) of clutches laid in Florida by the loggerhead sea turtle ( Caretta caretta ) by management unit (MU) (see Figure 1): NW (n = 1648), NE (n = 854), CW (n = 8161), CE (n = 2866), SW (n = 2008), SE (n = 6917). Blue line: linear regression. The upper dashed line indicates the pivotal incubation duration (PID, at which both sexes are produced in equal numbers). Clutches above this line are expected to produce a majority of males. The lower dashed line indicates the upper limit of the female‐only incubation duration range (FID). Clutches with an ID below this line are expected to produce females only. Clutches with an ID between the two dashed lines are expected to produce a majority of females. Boxplots: median, 50% percentile (black line), 25%–75% (box) and 95% percentile range (whiskers).

FIGURE 3.

Annual distribution of incubation duration (ID) of clutches laid in Florida by the green sea turtle ( Chelonia mydas ) by management unit (MU) or subarea (see Figure 1): CE (n = 2035), SE (n = 1363). The other MUs/subareas are not shown because of low sample size. See Figure 2 for description of symbols.

FIGURE 4.

Annual distribution of incubation duration (ID) of clutches laid by the leatherback sea turtle ( Dermochelys coriacea , n = 916). See Figure 2 for the description of symbols.

3.2. Temporal Change of Embryo Mortality or Egg Infertility

Proportions of unhatched eggs were available for a total of 57,903 clutches of the three target species laid in Florida on the 16 beaches monitored during the period 2001–2022. Of these, 33,487 (57.8%) clutches were retained in the analyses after removing clutches not left in situ, those that experienced disturbance, and those with clutch size values considered as outliers.

Models showed a meaningful effect of MU on the proportion of unhatched eggs in both loggerhead turtles (n = 26,547) and green turtles (n = 5588); therefore, models were run for each MU separately. No meaningful effect of Year on the proportion of unhatched eggs was detected by models in any MU of loggerhead (Figure 5) and green turtles (Figure 6). In loggerhead turtles, proportions of unhatched eggs during the first and most recent 5‐year period were not significantly different in four MUs (CE, n = 1608; NE, n = 374; SW, n = 1136; NW, n = 1120), while during the most recent 5‐year period, they were lower in CW (n = 3925) and higher in SE (n = 4826). In green turtles, proportions of unhatched eggs were not significantly different in CE (n = 1419), while they were higher during the most recent 5‐year period in SE (n = 607) and UW (n = 718). The other MUs had too few data for analysis. In leatherback turtles, the model showed a meaningful effect of Year on the proportion of unhatched eggs, with the proportion of unhatched eggs decreasing over time (n = 1352, Figure 7). ID and the proportion of unhatched eggs of loggerhead turtles were negatively correlated (Kendall's test; Tau = −0.145; p < 0.001), supporting the above use of the proportion of unhatched eggs to investigate temperature‐dependent embryo mortality.

FIGURE 5.

Annual distribution of proportion of whole unhatched eggs of clutches laid in Florida by the loggerhead sea turtle ( Caretta caretta ) by management unit: NW (n = 2008), NE (n = 948), CW (n = 9278), CE (n = 4400), SW (n = 2069), SE (n = 7844). Blue line: linear regression.

FIGURE 6.

Annual distribution of proportion of whole unhatched eggs of clutches laid in Florida by the green sea turtle Chelonia mydas by management unit/subarea: UW (n = 939), CE (n = 3254), SE (n = 1304). The other MUs/subareas are not shown because of low sample size. Blue line: linear regression.

FIGURE 7.

Annual distribution of proportion of whole unhatched eggs of clutches laid in Florida by the leatherback sea turtle ( Dermochelys coriacea , n = 1352). Blue line: linear regression.

3.3. Areas and Periods With Higher Male Production

ID values were available for a total of 116,977 clutches of the three target species laid in Florida during the recent period 2016–2022. Of these, 96,313 (82.3%) clutches were retained in the analyses after removing clutches not left in situ, those that experienced disturbance except for wash‐over, and those with ID values considered as outliers.

In all loggerhead MUs, most clutches had ID values suggesting primary sex ratios were completely (100%) or mostly (> 50%) female biased, with SE producing the lowest proportion of males and NW and SW the highest proportion (Figures 8 and 9). In green turtles, at least one MU (U‐NW) is likely to produce a majority of males, while the others (especially SE and DRTO) would mostly produce females (Figures 8 and 9). However, when the relative abundance of clutches is considered, clutches incubating in CE Florida likely produced the largest number of males of both species (Figure 10). When the proportion of the estimated total clutches in the three ID groups is compared among the three species, loggerhead and green turtles show a similar pattern, while leatherback turtles probably produce a higher proportion of males than the other two species (Figure 11). The observed ID values by species and beach are provided in Figures S1–S3.

FIGURE 8.

Distribution of incubation duration (ID) of clutches laid in Florida during the period 2016–2022 by three sea turtle species ( Caretta caretta , n = 81,907; Chelonia mydas , n = 12,142; Dermochelys coriacea , n = 2258) and by management unit (MU) or subarea when applicable (see Figure 1). Caretta caretta MUs: NE (n = 9529), CE (n = 6470), SE (n = 34,751), SW (n = 8032), CW (n = 19,942), NW (n = 2963). Chelonia mydas MUs/subareas: U‐NE (n = 612), CE (n = 3033), SE (n = 6417), DRTO (n = 419), U‐W (n = 1443), U‐NW (n = 218). The upper dashed line indicates the pivotal incubation duration (PID, at which both sexes are produced in equal numbers). Clutches above this line are expected to produce a majority of males. The lower dashed line indicates the upper limit of the female‐only incubation duration range (FID). Clutches with an ID below this line are expected to produce females only. Clutches with an ID between the two dashed lines are expected to produce a majority of females. Boxplots: median, 50% percentile (black line), 25%–75% (box) and 95% percentile range (whiskers).

FIGURE 9.

Proportion of Caretta caretta and Chelonia mydas clutches laid in Florida during the period 2016–2022 by management unit or sub‐area (see Figure 1) and divided into three groups of primary sex ratio (PSR) according to their incubation duration (ID) values. (A): Clutches with an ID that indicates a female‐only PSR (female‐only incubation duration, FID). (B): Clutches with an ID that indicates a female‐biased PSR (ID < PID, pivotal incubation duration). (C): Clutches with an ID that indicates a male‐biased PSR (ID > PID).

FIGURE 10.

Number of Caretta caretta and Chelonia mydas clutches laid in Florida in the period 2016–2022 by management unit or subarea (see Figure 1) and divided into three groups of primary sex ratio (PSR) according to their incubation duration (ID) values. (A): Clutches with an ID that indicates a female‐only PSR (female‐only incubation duration, FID). (B): Clutches with an ID that indicates a female‐biased PSR (ID < PID, pivotal incubation duration). (C): Clutches with an ID that indicates a male‐biased PSR (ID > PID).

FIGURE 11.

Proportion of sea turtle clutches by species (Cc: Caretta caretta ; Cm: Chelonia mydas ; Dc: Dermochelys coriacea ) laid in Florida during the period 2016–2022 and divided into three groups of primary sex ratio (PSR) according to their incubation duration (ID) values. (A): Clutches with an ID that indicates a female‐only PSR (female‐only incubation duration, FID). (B): Clutches with an ID that indicates a female‐biased PSR (ID < PID, pivotal incubation duration). (C): Clutches with an ID that indicates a male‐biased PSR (ID > PID).

In loggerhead and leatherback turtles, more males are likely produced in the first part of the nesting season, while in green turtles, more males are likely produced during both the first and the last parts of the nesting season (Figure 12). Results at the MU level are provided for loggerhead and green turtles in Figures S4 and S5, respectively.

FIGURE 12.

Seasonal distribution of incubation duration (ID) of clutches laid in Florida by three sea turtle species by 10‐day periods (labelled by the last day of the period) during the period 2016–2022. Number of clutches by species was 81,757 for Caretta caretta , 12,127 for Chelonia mydas , and 1821 for Dermochelys coriacea . See Figure 8 for symbols.

4. Discussion

This study provides the first empirical evidence of increasing incubation temperatures over time for three sea turtle species sharing the same nesting area (Florida). This increase likely affects the primary sex ratio and may also have some lethal effects. While a few localized studies have documented trends in incubation duration, this is the first to assess climate‐driven changes across multiple sea turtle species nesting in the same region, using large‐scale, long‐term data from genetically distinct MUs. This direct evidence from a biological indicator such as incubation duration (ID) complements both local and global analyses of air and sand temperatures (Laloë et al. 2024), collectively supporting the conclusion that climate change is already affecting sea turtle sex ratios—an impact that is projected to intensify in the future (Laloë et al. 2014). In this respect, this study establishes a robust empirical baseline for long‐term monitoring and adaptive conservation planning.

4.1. Evidence of Increased Incubation Temperature Possibly Affecting Sex Ratio and Natality

Our findings indicate that the ID of clutches for all three species has decreased over the past 22 years by 1.2–5 days, depending on the species and MU, and an increase in incubation temperature is the most plausible explanation. While both temperature and humidity influence ID, temperature is the primary driver, with higher incubation temperatures leading to shorter incubation durations (Ackerman 1997). Such a decrease can be compared to a general temperature increase at sea turtle nesting sites worldwide (Laloë et al. 2024).

Our findings of a decreasing ID trend align with those from a recent study on leatherback turtles nesting in Brazil (1988–2021; Colman et al. 2025), as well as two long‐term studies on loggerheads: one at Laganas Bay, Greece (1984–2021; Margaritoulis et al. 2022), and one at Bald Head Island, North Carolina, at the northernmost extent of the species in the Atlantic (1991–2015; Reneker and Kamel 2016). However, the rate of decline in ID was more pronounced at Bald Head Island, with an annual decrease of 0.28 days/year from 1991 to 2015, compared to a maximum of 0.16 (NW) or 0.14 (CE) days/year from 2001 to 2022 in Florida. While similar comparisons of annual ID decline rates in Greece (1984–2021) are not possible, the differences between beaches in Florida and North Carolina suggest that the effects of climate change—manifested as rising incubation temperatures—may be more pronounced at higher latitudes, where larger proportions of male hatchlings are typically produced (Hawkes et al. 2007; Mrosovsky 1988). This hypothesis is further supported by a recent global analysis of trends in air temperature, which found that sea turtle nesting sites in temperate zones are warming nearly twice as fast as those in tropical regions, despite regional variability (Laloë et al. 2024). Other three broad‐scale, long‐term studies found no significant trends in ID for loggerheads (Fuller et al. 2013; Marcovaldi et al. 2016) or hawksbills (Dei Marcovaldi et al. 2014). Although various factors can influence primary sex ratio, temperature is the primary determinant. The decrease in ID observed in Florida over the 22 years of the study likely reflects a shift toward a more female‐biased primary sex ratio in loggerheads, green turtles, and leatherbacks.

A critical question is whether increasing temperature may also increase embryonic mortality. Our findings suggest that it does. In both loggerhead and green turtles, the southeast MU showed an increasing trend in the proportion of whole unhatched eggs and the lowest predicted ID value for 2022. High temperatures and dry substrates are unfavorable for egg development (Packard et al. 1987; Santidrián Tomillo et al. 2014), and cumulative exposure to elevated temperatures has been shown to increase embryonic mortality in loggerhead and green turtle clutches in southeast Florida (Bladow and Milton 2019). The combination of increasing proportion of unhatched eggs and decreasing ID in southeast Florida may represent an early sign of temperature‐related lethal effects in Florida's warmest MU. This may be a warning that temperatures are approaching the lethal range (Hays et al. 2017), potentially leading to stronger effects in the future, even in other MUs.

4.2. Evidence of Insufficient Male Production

The ambiguous results regarding the proportion of whole unhatched eggs—with higher, lower or unchanged proportions depending on the individual MU—and evidence of multiple paternity from studies conducted at individual beaches on loggerheads nesting in Florida (Lasala et al. 2018; Silver‐Gorges et al. 2024) suggest that fertility is not currently a widespread issue. However, multiple paternity has not been investigated in leatherbacks and green turtles, nor specifically in southeast Florida for any species—the region where our study identified an increase in the proportion of unhatched eggs in loggerhead and green turtles. Notably, southeast Florida hosts the largest proportion of loggerhead (Ceriani et al. 2019) and leatherback (Stewart et al. 2011) turtles nesting in the state, making potential fertility issues in this region particularly concerning.

Time lag in sea turtle reproduction is an important concept to consider because adults reproducing today were born approximately 20–40 years ago, the estimated age at maturity for loggerheads, green turtles, and leatherbacks in the Northwest Atlantic (Avens et al. 2015, 2020; Goshe et al. 2010). Although the primary sex ratio may have been less female‐skewed when these adults were born, limited studies from that period already indicated an extreme female bias in loggerhead hatchling production on critical nesting beaches in Florida. For instance, Mrosovsky and Provancha (1989, 1992) reported a primary sex ratio ranging from 87% to 99.9% female at an important beach in central‐east Florida during the 1989–1991 seasons, while Hanson et al. (1998) predicted 100% female production in 92.5% of clutches examined in 1997 at a major southeast Florida beach. Despite this long‐standing female bias in loggerhead primary sex ratio, fertility does not appear to be a significant issue across Florida. Unfortunately, no similar data are available for green turtles and leatherbacks because there have been no studies on multiple paternity or primary sex ratio for these species in Florida.

A crucial question is whether the current primary sex ratio is jeopardizing the minimum number of adult males needed when today's hatchlings reach maturity. Theoretical and empirical evidence indicate that the current operational sex ratio (OSR)—the sex ratio of adults reproducing in a breeding season—is less female‐skewed than the primary sex ratio would suggest. This is likely due to male turtles' breeding behaviors, such as mating with multiple females, reproducing more frequently (potentially annually), and possibly maturing earlier than females (Hays et al. 2022). These behaviors may be sufficient to compensate for the highly female‐skewed primary sex ratio, even in the future.

Moreover, while our results show a female‐skewed primary sex ratio for all three species in Florida, many clutches likely still produce mixed sexes, and the resulting males may be sufficient to maintain viable populations. A recent modeling simulation of the NWA loggerhead population demonstrated that even a few years of higher male production (i.e., primary sex ratio of 70% female: 30% male) could significantly delay population collapse, even with a rapid increase in the female‐biased primary sex ratio (Heppell et al. 2022). Our findings indicate that such male‐producing years occurred within the study period. Additionally, genetic studies suggest that loggerhead females can breed with males from other Management Units (MUs) (Bowen et al. 2005), and it is reasonable to assume that green turtle and leatherback females may exhibit similar behaviors. Future studies on male breeding behavior, particularly better quantification of polygyny and regional male‐mediated gene flow, are critical for understanding the operational sex ratio (OSR) and, consequently, long‐term population viability.

Based on our results, leatherback turtle clutches appear to produce a more balanced primary sex ratio, suggesting that this species may be less at risk of insufficient male production compared to loggerheads and green turtles. Lastly, although a highly female‐skewed primary sex ratio is observed, high temperatures can reduce hatching success, particularly in clutches producing all females, thereby lessening the female bias in the secondary sex ratio (the sex ratio of hatchlings that successfully emerge) (Santidrián Tomillo et al. 2014).

4.3. Evidence of Adaptation

Since this study directly investigated real clutches laid by females, either through spatio‐temporal sampling or by including all clutches laid, rather than relying on modeling, it accounts for potential behavioral adaptations by female sea turtles in response to climate change, such as shifts in nesting timing (phenological shifts: Weishampel et al. 2010; Weishampel et al. 2004), range shifts (Abella Perez et al. 2016; Hochscheid et al. 2022), and microhabitat selection (nest‐site choice: Kamel and Mrosovsky 2006). However, the observed decrease in incubation duration suggests that these adaptations may not be occurring or are insufficient to counteract the effects of rising temperatures. It remains possible that stronger adaptive responses could be triggered by more extreme environmental changes in the future.

Phenological changes in nesting distribution have been reported at two important sites in CE Florida for both loggerheads and green turtles (Pike et al. 2006; Weishampel et al. 2010, 2004), suggesting that both species begin nesting earlier (in cooler months) and that green turtle nesting season extends longer, at least at one of the two sites—the Archie Carr National Wildlife Refuge (Weishampel et al. 2010). The change in nesting phenology has been attributed to increasing sea surface temperature and may represent a response to climatic warming that could ensure some production of male hatchlings, although recent modeling suggests phenological shifts alone will likely not be sufficient to counteract future projected impacts of climate change (Fuentes et al. 2024). Future studies should revisit the nesting phenology of sea turtle nesting in Florida by expanding the number of years included (previous studies ended in 2008) and expanding the geographic scope to all MUs.

4.4. Implications for Conservation and Management

Results indicate that the largest proportion of male loggerhead and green turtles is produced in NW Florida. However, when the relative abundance of clutches is considered, CE Florida produces the greatest number of males for both species. The beach‐level results (Figures S1–S3) show several individual beaches where male production may be particularly high, and these beaches represent a conservation target for supporting male production.

Our study found both an inter‐annual and inter‐MUs variability in incubation duration and thus estimated primary sex ratio. Both inter‐annual and inter‐beach variability in primary sex ratio have been identified in the only other three wide‐scale, long‐term monitoring studies of primary sex ratio in Brazil (Dei Marcovaldi et al. 2014; Marcovaldi et al. 2016) and Cyprus (Fuller et al. 2013). More males are produced at the beginning of the nesting season, as already reported elsewhere (Dei Marcovaldi et al. 2014; LeBlanc et al. 2012; Mrosovsky, Dutton, and Whitmore 1984; Mrosovsky, Hopkins‐Murphy, and Richardson 1984). Specifically, loggerhead and leatherback clutches laid before May 29 and May 19, respectively, seem to be particularly valuable for male production. In the case of green turtles, clutches laid early in the season (prior to May 29th) and late in the season (after September 6) appear particularly valuable for male production. All sea turtle clutches are protected by federal and state law in Florida. However, minimization measures in Florida designed to reduce potential impacts from activities such as beach cleaning, beach events, seawall construction, and certain sand placements are only required during the official sea turtle nesting season (March 1 to October 31 in Brevard to Broward County; May 1st to October 31st in other areas of the state). Given that the periods identified as critical for sustaining male hatchling production extend beyond the official nesting season, it is essential to consider expanding minimization measures to cover these periods.

4.5. Methodological Approaches

Incubation duration (ID) has been proposed long ago as a simple, inexpensive, and effective method to estimate population‐level sex ratios (Mrosovsky et al. 1999). However, only a few studies have investigated long‐term trends (≥ 10 years) in primary sex ratio using ID, with most focusing on loggerheads (Fuller et al. 2013; Marcovaldi et al. 2016; Reneker and Kamel 2016) and one focusing on hawksbills (Dei Marcovaldi et al. 2014). This limited use is unfortunate, as ID is routinely collected by sea turtle nesting programs worldwide, offering an easy and cost‐effective way to compare primary sex ratio across broad spatial and temporal scales. Although further validation of the ID–sex ratio relationship at the MU level would be valuable, the approach remains useful for comparative purposes, especially given the logistical and permitting constraints that often preclude direct validation.

However, caution is needed when converting ID to primary sex ratio, as the relationship between the two is uncertain, especially in the range where both sexes are produced (Mrosovsky et al. 1999). Because of this uncertainty, particularly at intermediate ID values, this study avoided direct conversion and instead classified ID values into biologically meaningful categories (i.e., female‐only, mixed but predominantly male, and mixed but predominantly female) to conservatively assess trends in primary sex ratios. This binning approach enables more robust spatio‐temporal comparisons using a parameter that is already widely available from routine monitoring. Recent studies have continued to rely on thermal models and temperature‐sex curves to estimate primary sex ratios from reconstructed nest temperatures (e.g., Colman et al. 2025; Fuentes et al. 2024; Monsinjon et al. 2019), but these approaches require intensive data collection, rest on multiple assumptions, and, thus, result in high uncertainty on the precise primary sex ratio estimates. In contrast, the method adopted here is simple, scalable, and easily implemented, avoiding uncertain conversions while still providing a conservative indicator of broad‐scale shifts in primary sex ratio.

The ID‐based approach could be complemented by other methods. For instance, deceased hatchlings and embryos represent an ethically accessible and spatially diverse source of material for gonadal histology. Although such samples are often too small to estimate the primary sex ratio at the clutch level, they can enhance beach‐ or region‐level assessments when combined with ID data.

5. Conclusions and Recommendations

Since the rate of future sand temperature increase on nesting sites is expected to exceed that of past warming (Laloë et al. 2014), effects like those reported here are likely to become more pronounced in the near future. Monitoring such effects and assessing their impact on sea turtle populations will require standardized and comparable methodologies. Although incubation duration (ID) is subject to uncertainty at intermediate values (Mrosovsky et al. 1999) and should not be used to estimate primary sex ratio at the individual nest level, conservation programs worldwide that monitor clutch fate are encouraged to use long‐term ID datasets to establish a much‐needed retrospective baseline, as originally proposed by Mrosovsky et al. (1999), and to track future changes.

Author Contributions

Simona A. Ceriani: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, visualization, writing – original draft. Paolo Casale: conceptualization, formal analysis, methodology, visualization, writing – original draft.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: gcb70559‐sup‐0001‐DataS1.docx.

Ceriani, S. A. , and Casale P.. 2025. “Long‐Term Incubation Duration Decline Indicates Climate‐Change Driven Feminization of Three Sea Turtle Species in Florida, USA .” Global Change Biology 31, no. 10: e70559. 10.1111/gcb.70559.

Data Availability Statement

The data that support the findings of this study are openly available in Zenodo at http://doi.org/10.5281/zenodo.17243227.