Ecological geography of the hawksbill turtle (Eretmochelys imbricata) in the West Atlantic

Abstract

Understanding the geographic distribution of genetic diversity of imperiled species across all life history stages, and identifying the factors that shape those distributions, are key to maintaining long-term genetic diversity and the health of populations. This knowledge is particularly important for highly mobile marine organisms, whose extensive movements can obscure patterns of population structure. We substantially expand the genetic dataset for the critically endangered hawksbill turtle, Eretmochelys imbricata, in the West Atlantic, focusing on the southwest Caribbean. Our dataset comprises nearly 3,000 mtDNA control region sequences (740 bp) assigned to 60 haplotypes: 41 found in rookeries and 47 in foraging grounds, including 17 orphan haplotypes. The Panama metapopulation represents a major center of genetic diversity for hawksbills, with one of the highest recorded diversity values for the species (h = 0.749, π = 0.00782), nine endemic haplotypes, and four additional haplotypes that are endemic to the Southwest Caribbean. Rarefaction analyses indicate that a sample size of at least 100 is necessary to reveal true haplotype richness at most rookeries. Many-to-many mixed stock analyses, which incorporated rookery size and distance priors for 19 rookeries and 15 developmental foraging grounds, suggest that hatchlings from rookeries in the southwest Caribbean are distributed among multiple, widely-spaced foraging grounds across the West Atlantic. These results support a groups-to-soups analogy, in which genetic variability across foraging grounds represents a continuum of genetic diversity that can best be explained by a “current conveyor” model. The dataset shows that philopatry in hawksbills is not absolute, resulting in true biological dispersal and geneflow on local, regional, and ocean-basin scales, likely facilitated by dispersion during the epipelagic stage. The important contribution of oceanographic features to genetic variation at rookeries and foraging grounds is corroborated, as is the concept of oceanographic “dispersal shadows” that limit geneflow between rookeries. This study reinforces the assertion that all range states share responsibility for the recovery of the hawksbill, because foraging grounds, that are often at distant locations, are the source of future generations of reproductive adults. We also document significant movement by hawksbills between regional management units (RMUs) 29 and 30 in the West Atlantic. The Spanish version of the Abstract is available in Supplementary File 1.

1 Introduction

The critically endangered status of the hawksbill turtle (Eretmochelys imbricata) was corroborated by two reviews in recent decades (Meylan and Donnelly, 1999; Mortimer and Donnelly, 2007) and this species continues to be a major focus for sea turtle conservation (Hays et al., 2025). Like other marine turtles, the hawksbill has a long and complex life cycle and further improvement of the conservation outlook for this species requires a more complete understanding of the ecological geography (Carr, 1980) of the species. Better knowledge of how individuals make use of resources across all life history stages will inform range states how to optimally contribute to the recovery of this ecologically important and critically endangered species.

The life cycle of the hawksbill is similar to that of most other cheloniid sea turtles (Meylan et al., 2011). Females emerge onto beaches in tropical regions to lay relatively large clutches of eggs. Following hatching, hatchlings are dispersed during an epipelagic stage lasting several years, followed by settlement onto benthic foraging grounds. These are typically mixed-stock, developmental foraging aggregations, from which resident adults are frequently absent (Blumenthal et al., 2009; Meylan et al., 2011; Wood et al., 2013). Individuals from these sites make a final developmental migration to an adult foraging range, from which they make reproductive migrations after maturation. The mixing of individuals during the epipelagic stage and on developmental foraging grounds (FGs) leads to complex population genetics, which has been a long-standing challenge for studies of ecological geography.

The toolkit for understanding the ecological geography of marine turtles has expanded from tagging studies to a wide range of methods, including telemetry, stable isotope studies, genetics, and particle drift modeling. However, only molecular genetics provides sufficient empirical evidence to corroborate hypotheses about the extent to which individuals from different genetic populations travel to specific sites during their migrations.

Extensive genetic studies of hawksbills in the West Atlantic have made important contributions to conservation efforts (e.g., Bowen et al., 2007; Mortimer et al., 2007a, b). However, the full potential of these analyses will only be realized when all major rookeries and a broader selection of feeding grounds are sampled. The first studies of hawksbill conservation genetics in this region began with the analyses of a 384–480 bp segment of the mitochondrial genome (e.g., Bass et al., 1996; Bowen et al., 1996, 2007; Díaz-Fernández et al., 1999). However, Leroux et al. (2012) and others showed that a longer, 740 bp control region sequence (Abreu et al., 2006) reveals additional informative variation. The longer sequence suggests eleven distinct management units for the hawksbill in the West Atlantic (Leroux et al., 2012; Vilaça et al., 2013). At a regional scale, Wallace et al. (2023, 2025) recognized three RMUs (# 29–31) for Eretmochelys in the Atlantic: Northwest Atlantic (15 genetic stocks), Southwest Atlantic (two genetic stocks), and East Atlantic (one genetic stock). The longer sequence now represents the dataset of choice to evaluate the patterns and levels of connectivity among rookeries and feeding aggregations via mixed-stock analyses (e.g., Cazabon-Mannette et al., 2016; Pérez-Bermúdez et al., 2017; Labastida-Estrada et al., 2019). However, this approach is limited by the lack of data for all rookeries, and, for many-to-many evaluations, all relevant foraging grounds. Researchers using MSAs have emphasized the need to consider temporal sampling issues (Pérez-Bermúdez et al., 2017) and differences between life history stages (Velez-Zuazo et al., 2008; Wood et al., 2013; Proietti et al., 2014; Putman and Mansfield, 2015), along with the relative size of rookeries and the distances between rookeries and foraging grounds (Stahelin et al., 2022). MSAs for hawksbills are complicated by extensive exploitation, extinction, and recolonization at nesting sites. Historical exploitation resulted in the removal of hundreds of thousands—if not millions—of hawksbills for tortoiseshell, and nesting populations of this species worldwide were seriously depleted at some sites (Meylan and Donnelly, 1999). Harvest at nesting sites and random recolonization likely modified haplotype distributions. At present, very few rookeries have more than 1000 nests per year. However, as more rookeries move towards recovery, larger samples have become available, allowing increased geographic sampling for genetic studies.

After decades of conservation effort (Meylan et al., 2013) we are able to contribute a robust genetic dataset for four major nesting beach populations in Caribbean Panama that have not been analyzed previously. Additionally, we include new datasets from four widely spaced developmental FGs: Bermuda; Bonaire; the Upper Florida Keys, USA; and Bocas del Toro, Panama. The major objectives of this work were to: (1) document recently discovered genetic diversity that characterizes the hawksbill nesting metapopulation in western Caribbean Panama; (2) evaluate the level of differentiation between populations from four Panamanian nesting beaches and between Panama and other West Atlantic hawksbill rookeries; (3) describe the genetic variation at four previously undescribed FGs; (4) use MSAs to determine the level of connectivity between rookeries and FGs in the West Atlantic, with particular attention to the contribution by the Panamanian metapopulation; (5) hypothesize how the biology of this species explains the observed patterns of genetic variation. Our results were also used to examine the groups-to-soups concept for explaining variation in mixtures at FGs. The original hypothesis that thorough mixing of post-hatchlings from different populations during the epipelagic stage (Lahanas et al., 1998) would result in FGs being random samples from all rookeries (soups) (Engstrom et al., 2002). Blumenthal et al. (2009) refined this concept by considering the influence of currents and found that the contributions from some rookeries were broadly dispersed leading to “soup-like” FGs while other rookeries were regionally constrained and resulted in more “group-like” FGs. Labastida-Estrada et al. (2019) suggested that the “turtle soup” model could be explained by ocean current patterns that favor the connectivity of hawksbill rookeries to certain FGs and that FGs that look like groups “experience high levels of proximate or local recruitment.”

2 Methods

2.1 Study sites, sample collection, and life-history stages

The hawksbill genetic datasets presented here for the first time were collected via four research projects (see Supplementary Table S1 for sampling details). Samples from Panama were collected as part of ongoing studies of marine turtles in Bocas del Toro Province and the Comarca Ngäbe Buglé (“the Bocas Region”; Meylan et al., 2011, 2013). Samples from Bermuda were collected as part of the Bermuda Turtle Project’s ongoing studies (Meylan et al., 2011, 2022, 2024). Samples from the upper Florida Keys were collected by the Florida Fish and Wildlife Conservation Commission as part of the NMFS Sea Turtle Stranding and Salvage Network. Samples from Bonaire were collected by Sea Turtle Conservation Bonaire.

The four sampled Panama beaches are located in the Bocas del Toro region (Figure 1A). Samples were collected between 2002 and 2018 from tagged individuals and are unique. Nighttime patrols on ~47 km of beaches allowed estimation of movement by nesting females between beaches. We also accessed hawksbill tagging records from three projects in Costa Rica to identify females that moved between that country and Panama.

Figure 1

Based on size, tail length, laparoscopy, or necropsy, all individuals from FGs were considered immature. The Bermuda FG dataset was assembled from entrapment net, snorkel, and scuba captures, and strandings documented by the Wildlife Rehabilitation Center, Bermuda Aquarium Museum and Zoo (Meylan et al., 2011). The Bonaire FG samples were collected between 2003 and 2007 at multiple sites from live captured turtles that were individually tagged. The Florida Upper Keys FG samples were obtained from 48 stranded hawksbills collected between 12 and 26 Jan 2010 during a single cold-stunning event; samples were collected between Tavernier and Little Torch Key. The location, timing of recovery, and condition suggested these individuals were resident in the Florida Keys or adjacent Florida Bay. The Panama FG sample (n = 38) was limited to immature individuals from entanglement net captures, confiscations, or contributions from local citizens from ~70 km of Bocas del Toro coastline between Tobobe and Bocas del Drago.

Samples for genetic analyses consisted of whole blood or connective tissue (muscle or skin). Whole blood samples were taken from the cervical sinus (Owens and Ruiz, 1980) and preserved in lysis buffer (Engstrom et al., 2002). Tissue samples from Bonaire were taken from the margin of one flipper or right shoulder using a sterile 4 mm biopsy punch and preserved in 95% ethanol or salt-saturated 20% DMSO-20% EDTA buffer. Tissue samples from necropsied turtles consisted of 1 cc of muscle or skin preserved in SED buffer (Dutton, 1996), ethanol, or saturated salt solution (Proebstel et al., 1993).

2.2 Datasets and abbreviations

In addition to the novel datasets, we incorporated data from the published literature for 19 rookeries and 11 FGs (Supplementary Table S1). We excluded datasets from the East Atlantic (Monzón-Argüello et al., 2011; see below). We considered a rookery to be a construct that may include multiple nesting beaches. In some cases, we recombined rookery datasets from the literature into groups that differ from those used in previous publications. For Mexico, we included two rookeries from Labastida-Estrada et al. (2019), one for Campeche (MX.C) and one that combines four nesting beach populations in Yucatan and Quintana Roo (MX.Y). We treated the US Virgin Islands (USVI) as two rookeries: Buck Island (USVI.BI) and Sandy Point (USVI.SP; Hill et al., 2018). We included all Tobago nesting samples (Cazabon-Mannette et al., 2016) as a single rookery (TT) and all in-water samples as a single FG (TT.fg). We treated Brazilian populations as two rookeries: one for “Bahia” (BR.B; Lara-Ruiz et al., 2006) and one for “Pipa” (BR.P; Vilaça et al., 2013), with hybrids excluded. Thus, our initial dataset included the haplotype distributions from 23 possibly distinguishable rookeries and 15 foraging grounds (FGs) (Figures 1A, B; Supplementary Table S1).

We used standardized abbreviations (Figure 1; Supplementary Table S1). For rookeries, the first two to four capital letters indicate the country, additional letters indicate the location within that country. The lower-case suffix .grp is used for four Panama nesting beach populations that are treated as one metapopulation and a single rookery (PA.grp) (see below). Abbreviations for all FG datasets have a lower-case suffix, .fg. All are considered benthic developmental sites, and we use the abbreviation FG or FGs for these mixed stock aggregations of immatures. To our knowledge, no adult Eretmochelys foraging ground in the Atlantic has been genetically sampled.

2.3 Geographic considerations

We restricted our analyses to the Northwest and Southwest Atlantic (RMUs #29 and 30 of Wallace et al., 2023) from Ascension Island and Brazil, to Bermuda, and Campeche, Mexico (Figures 1A, B). We excluded the Eastern Atlantic rookery at Principe and Eastern Atlantic FGs (RMU 31), which are dominated by “EATL” haplotypes (Monzón-Argüello et al., 2011). The only site for which long haplotypes are available is a very small rookery on Poilão Island, Guinea-Bissau for which the sample size is seven (Madeira et al., 2025). Although eastern Atlantic stocks are considered to show a “high degree of isolation” (Monzón-Argüello et al., 2011), it should be noted that 14 FG samples in our study were assigned to haplotypes that are either known from an East Atlantic rookery or appear to be related to EATL haplotypes.

2.4 Laboratory methods

For all sites other than Bonaire, sequence data were generated from 2008 through 2013 at the ICBR Genetic Analysis Core, University of Florida. DNA was extracted using Qiagen DNeasy Tissue Kits as per manufacturer’s protocols, and an ~850 bp segment of the mtDNA control region was amplified using PCR primers LTEi9 and H950 (Abreu et al., 2006) and sequenced on an automated DNA sequencer (Applied Biosystems model AB3730xl). After 2016, samples were processed by undergraduate genetics classes at Eckerd College, St. Petersburg, FL and an ~850 bp sequence was generated by Eurofins Genomics LLC. For the Bonaire samples, DNA was extracted using Qiagen DNeasy Tissue Kits as per the manufacturer’s protocol. Amplified fragments were sequenced using the PCR primers listed above via ABI Big-dye^® terminator chemistry on an automated station ABI 3130XL sequencer (Applied Biosystems). All high-quality readings were cropped to the standard 740 bp sequence.

2.5 Analytical methods

2.5.1 Sequence alignments

Sequences were assembled and aligned using Multalin (Corpet, 1988), SEQUENCHER 5.4.6 (Gene Codes), GENEIOUS version 8.2 (Kearse et al., 2012), or NCBI BLAST (Johnson et al., 2008) and checked manually. All sequences were compared to a set of all available Atlantic Ocean 740 bp sequences (Abreu-Grobois, unpublished). Sequences that matched established standards were assigned an EiA designation. Haplotype names previously used by Bass et al. (1996) and Díaz-Fernández et al. (1999) for shorter versions of most of these haplotypes are given in Supplementary Table S2. Sequences that did not match a standard were sequenced in both directions for confirmation before assigning a new haplotype designation. During our review of sequence data, we determined that EiA57 is identical to EiA86; the latter is a longer sequence and is used herein. The term orphan haplotype is used throughout the text to indicate haplotypes known from FGs but not from any rookeries.

2.5.2 Networks

We constructed haplotype networks using the Median-Joining algorithm (Bandelt et al., 1999) implemented in PopART. Since PopART eliminates sites with indels, which were interpreted as genuine variant sites, all gaps were re-coded as distinct bases and all multi-bp inserts were collapsed to a single distinct base.

2.5.3 Genetic diversity

Genetic diversity (h and π; Arlequin v 3.5.1.2; Excoffier and Lischer, 2010) was examined for the West Atlantic hawksbill rookeries for which 740 bp haplotypes were available from a sample of at least 15 individuals. These included 19 rookeries from the published literature plus four Panamanian nesting beach populations. A similar compilation was made for 11 published FG studies for which 740 bp haplotypes were available from a sample of at least 22 individuals, plus four FGs from this study.

To examine the impact of sample size on haplotype diversity at rookeries and FGs, we used the iNEXT R package version 3.0.1 (Chao et al., 2014; Hsieh et al., 2024); the number of each haplotype found at rookeries or FGs was used to extrapolate predicted haplotype diversities for all sites at a sample size of 250 for rookeries and 200 for FGs. Plots of the results were drawn using the R packages ggplot2, ggrepel, and cowplot (Wickham, 2016; Slowikowski, 2024; Wilke, 2024).

2.5.4 Population genetic structure

Pairwise genetic differentiation between West Atlantic hawkbill rookeries and the four Panamanian nesting beach populations, as well as 15 benthic FGs, was examined using pairwise F_ST values (conventional frequency-based) in Arlequin v 3.5.1.2 (Excoffier and Lischer, 2010). The significance of differences in the Fst values was determined through permutation p-value tests, with p-values below 0.05 considered to indicate significant differentiation. Adjusted p-values were computed using the Benjamini-Hochberg method to control the false discovery rate (Benjamini and Hochberg, 1995). In initial runs, the four Panama beach populations were assessed separately: Small Zapatilla Cay, Big Zapatilla Cay, Playa Larga, and Chiriqui Beach. Following the first round of comparisons, and in consideration of the geographic locations, documented movements by nesting females between two or more of our study beaches, the likelihood of founder effect after a bottleneck, and management considerations, the four Panama nesting beach populations were combined as PA.grp (see below) in additional analyses. A UPGMA tree for the West Atlantic hawksbill rookeries and four Panamanian nesting beach populations was constructed based on the Fst estimates.

2.5.5 Mixed stock analyses

To estimate connectivity among source rookeries and FGs, we performed a series of MSAs (Bolker et al., 2003) using the long (~740 bp) mtDNA sequences with the mixstock R package (Bolker, 2015) under the many-to-many option. We used the mcmc (Markov chain Monte Carlo) function in Mixstock R package, with a Gibbs sampler with 250,000 or 500,000 iterations, a burn-in of 5000, and a thinning rate of four to ensure a Gelman-Rubin criterion value of ≤ 1.1 that verifies chain convergence (Gelman and Rubin, 1992). Ninety-five-percent confidence intervals (CI) were output from the 97.5% and 2.5% quantiles of the posterior distribution.

The availability of long reads for 23 rookeries and 15 FGs allowed us to use a robust regional dataset. However, we made several important compromises to do so. We elected to leave out two rookeries with only short reads — Belize (n = 14; Bass et al., 1996) and Venezuela (n = 7; Bowen et al., 2007). We also eliminated DR.J (n = 15) from the final MSA runs due to a small sample size. In order to include Cuba, Doce Leguas (CU.DL), for which 70 samples of a 480 bp sequence are available, we assumed that all A(CU1) haplotypes are EiA01, and none are EiA61 (currently only known from Brazil); we preferred this assumption to leaving out all Cuban rookeries. We also excluded FG datasets for which only short sequences were available, including Inagua (Bahamas), the Dominican Republic, the USVI (Bowen et al., 2007), Turks and Caicos (Richardson et al., 2009), and three Cuban and one Puerto Rican datasets from Díaz-Fernández et al. (1999). However, more recently published datasets provided ~740 bp haplotypes for FGs in Cuba (Pérez-Bermúdez et al., 2017), Puerto Rico (Velez-Zuazo et al., 2008), and two rookeries in the USVI (Hill et al., 2018), and these data were used instead. We sought to increase the comparability among FGs by separating immatures from adults at sites where these life history stages were sampled together (Velez-Zuazo et al., 2008; Wood et al., 2013; Cazabon-Mannette et al., 2016; Pérez-Bermúdez et al., 2017). We did not include the single set of epipelagic hawksbills from Texas (Bowen et al., 2007).

We ran both rookery - (source) centric and foraging ground - (mixture) centric MSAs. Initial analyses included uniform (equal likelihood) and rookery size priors. Due to their tighter CI, we focused on foraging ground-centric runs for additional MSAs, and we treated PA.grp rookeries as a single unit. Additional MSA runs included a set that tested the utility of various distance priors following the method of Stahelin et al. (2022).

2.5.6 Rookery size prior and timing of rookery size estimate

Details about rookery size estimation for the West Atlantic are given in Supplementary Table S1. In order to use estimates for rookery sizes from an appropriate time period (when turtles sampled at an FG would have hatched; Pérez-Bermúdez et al., 2017), we employed estimates that were made approximately 10 years before the majority of FG samples were collected (2005 ± 6.3 yr). For rookeries first studied after the mid-1990s, we had to rely on the earliest estimate available. Published data for three rookeries (MX.Y, MX.C, and USVI.SP) were collected long after the majority of FGs were sampled. For these rookeries, we adjusted the estimated number of nests by consulting with colleagues with access to relevant data. For Doce Leguas, Cuba (CU.DL), Pérez-Bermúdez et al. (2017) suggest no significant increase in nesting between 1997 and 2009, so we retained the estimate from Díaz-Fernández et al. (1999). For Panama, we used data for the first year with a complete nesting season survey for each nesting beach population.

2.5.7 Distance prior and the use of current track distances

We used the Stahelin et al. (2022) method to incorporate distance priors into a set of MSAs. However, noting that the presence of strong currents may provide rapid transport of hatchlings away from natal beaches (Putman et al., 2010; Phillips et al., 2025), we did not assume that rookeries are more likely to contribute to closer FGs than distant FGs. Instead, we compared results for four minimum distance priors (1, >320, >520, and >1020 km) in combination with a rookery size prior. Following Stahelin et al. (2022), distances were estimated as the length of “probable paths” between hawksbill rookeries and FGs (Supplementary Figure S1) using available marine currents as vectors for transport under our four minimum distance scenarios. Individual distances between landmarks (Supplementary Table S3.1) were summed for the total paths between rookeries and FG (Supplementary Tables S3.2, S3.4). Covariates were calculated by transforming total distances into weighted proportions for input into the MSAs using a modification of the mixstock R package (Supplementary Tables S3.3, S3.5). Our comparative tests of distance priors included the small rookery dataset for DR.J and used an earlier dataset for AG.JB (Leroux et al., 2012), which does not differ significantly (Fst = 0.006) from a more recent, larger dataset used in other analyses (Levasseur et al., 2019).

2.5.8 Comparisons of MSAs, contributions to FGs, and the groups-to-soups analogy

We compared the reliability of different MSA runs using three criteria: the size of 95% CI (Blumenthal et al., 2009), consistency of results among runs (Day et al., 2024), and the ability of individual models to explain the total genetic diversity observed at sampled sites. To identify estimates of rookery contributions to FGs that differ from zero while minimizing the risk of accepting false positives, we tested Q 02.5 values of > 0.01, >0.005, >0.002, and >0.001 and compared results with empirical observations of small contributions over long distances to justify a conservative minimum “non-zero” Q 02.5 value. We explored the groups-to-soups analogy for FGs via the standard deviations for all rookeries to an FG (average contributions closer to the mean being more soup-like), the number of “non-zero” contributing rookeries, and the size of the single largest estimated contribution.

3 Results

3.1 Haplotypes

We assembled a dataset of 740 bp sequences for 2,894 individual hawksbills, including 1,650 from 20 West Atlantic rookeries and four Panama nesting beach populations, and 1,244 from 15 FGs (Supplementary Table S1). The combined dataset includes 60 confirmed haplotypes based on 59 polymorphic sites, 29 of which are reported for the first time (Supplementary Table S2).

Our work confirms previously reported but rare haplotypes: EiA33 (h), EiA35 (m), and EiA63 (F/PR1). Genbank accession numbers for examples of these haplotypes, new haplotypes, and longer reads for previously reported haplotypes, along with the identity of 12 likely invalid haplotypes, are given in Supplementary Table S4.

3.1.1 Haplotypes at rookeries and Southwest Caribbean endemics

Haplotype distributions for 20 West Atlantic rookeries and four Panama nesting beach populations are shown in Table 1 and Figures 1 and 2. Data for the Panama populations consists of sequences from 463 nesting females. Haplotype EiA11 dominates all four Panama nesting beach populations; the Southwest Caribbean (SWC) endemic, EiA02, is also common. EiA12 and EiA47 are common at Playa Larga (PA.PL) and Chiriqui Beach (PA.CH), while EiA84 is common at the Zapatilla Cays (PA.SZ and PA.BZ). The four Panama nesting beach populations host one or more of the Panama endemic haplotypes: EiA33, 35, 55, 58, 64, 77, 91, 94, 95 (Table 2). The Chiriqui Beach population has seven of nine endemic haplotypes, two of which are unique (EiA77 and EiA91); the Zapatilla Cays have four of nine, one of which is unique (EiA94); and Playa Larga has four of nine endemic haplotypes, one of which is unique (EiA64). In addition to the Panama endemics, three haplotypes that occur in Panama populations (EiA02, 65, and 84) are endemic to the SWC, i.e., they are only known from Panama, the Pearl Cays, Nicaragua, and/or Tortuguero, Costa Rica. Five additional haplotypes (EiA12, EiA30, EiA43, EiA47, and EiA63) are SWC “near-endemics” because they are known from relatively few records outside of the SWC (see discussion), with 50–96% of known occurrences in the SWC (Table 2).

Figure 2

Given the presence of EiA63, we considered the option that CO.CV should be included in the SWC group of rookeries. However, the Panama-Colombia Gyre, a critical oceanographic feature of the region (see discussion), does not extend to CO.CV, instead the Caribbean Current exerts a major influence (https://data.marine.copernicus.eu/viewer/expert?view). Furthermore, no other SWC endemic haplotype is present at CO.CV and EiA01 is a major contributor (41%) which is not the case in SWC rookeries (2%). This is reflected in the Fst values (CO.CV differs from PA.grp —Fst = 0.115; whereas Fst values for SWC rookeries average 0.043 ± 0.033).

3.1.2 Haplotypes at FGs

A total of 47 haplotypes were documented at 15 FGs (Table 3), including 17 orphans (Supplementary Table S4). Fourteen haplotypes from rookeries were not recovered from any FGs (Supplementary Table S4). However, two of these, EiA30 (CU4) and EiA33 (h), were identified among 540 bp sequences from FGs in Cuba and Puerto Rico, respectively (Díaz-Fernández et al., 1999). We were unable to incorporate these two records into our study (see Methods). EiA01 was reported from all FGs in our study, and for 46.2% of all individuals from FGs. EiA11 occurred at 14 FGs and was recorded for 16.9% of all FG individuals. Other haplotypes were recorded for less than 10% of FG individuals; 25 haplotypes were observed in five or fewer individuals at FGs.

3.2 mtDNA haplotype network

The minimum spanning network for 60 different 740 bp haplotypes from 19 rookeries, four Panama nesting beach populations and 15 FGs (including 17 orphans) (Figure 3) reveals that haplotypes are assignable to three clades. Clades 1 and 3 have a star-shaped structure around the common haplotypes EiA01 and EiA11, respectively, and have broad geographic distributions. Clade 2 is centered on EiA41 and is restricted mainly to the Gulf of Mexico (GOM). Nineteen of 23 haplotypes known from Panama (83%), including all nine Panama endemics, are part of clade 3, which centers around EiA11. Nine new haplotypes differ from EiA11 by one to four base substitutions. However, the second most common haplotype, EiA02, is part of clade 1 and differs from EiA01 by a single substitution. The 17 orphan haplotypes are spread among clades 1 and 3 identified above and the EATL haplotype network.

Figure 3

3.3 Genetic diversity (h and π) and genetic structure

3.3.1 Genetic diversity at rookeries

Haplotype (h) and nucleotide diversity (π), along with the number of haplotypes from the 19 rookeries and four Panama nesting beach populations, are shown in Table 4. Other than the DR.J rookery, the Panama and SWC rookeries have the highest haplotype diversity among samples from the West Atlantic. Haplotype diversity in the SWC varies from 0.531 to 0.750 with an average of 0.641 ± 0.091, whereas rookeries outside the SWC vary from 0.0000 to 0.848 (DR.J) with an average of 0.427 ± 0.206. The number of haplotypes observed at rookeries in the SWC is, on average, more than twice (9.2 ± 4.35) as high as elsewhere in the West Atlantic (4.2 ± 1.37). The very large numbers of haplotypes for SWC populations were further corroborated by rarefaction curves (Figure 4A). After controlling for sample size, the predicted number of haplotypes indicated that three of four sampled Panama beaches are the most diverse currently known. In addition, rarefaction analysis indicated that a sample size of at least 100 is necessary to reveal true haplotype richness at most rookeries.

Figure 4

Hsieh et al., 2024). See Supplementary Table S1 for sampling details.

3.3.2 Genetic diversity at FGs

On average, haplotype and nucleotide diversity, as well as the number of haplotypes, were higher for the 15 FGs (Table 5) than at rookeries (Table 4). These measures are lowest for FGs at the beginning (upstream end) of a “current conveyor system” in which hatchlings are added to prevailing major currents as they move in one direction and pass additional rookeries. In this case, the “conveyor system” consists of the South Equatorial Current to North Brazil Current to Guyana Current to Caribbean Current to Yucatan Current to Loop Current to Gulf Stream). All three measures of diversity increase significantly towards the downstream end (Supplementary Figure S2). The number of haplotypes increases from ASC.fg (n = 4) to BM.fg (n = 23), and these findings were corroborated by rarefaction curves (Figure 4B). The lowest rarefaction trajectories are observed at ASC.fg and BR.C.fg in the southeast, and the highest are observed at FGs along the Florida Current and Gulf Stream (CU.JR.fg, BM.fg, and three FL FGs). Intermediate trajectories are observed at sites within the Caribbean (MX.QR.fg, TT.fg, BQ.fg, and CAY.fg). MX.PX.fg, which is outside of the main trend of regional currents, also shows a low trajectory. PR.MI.fg, with a high trajectory, stands out as an outlier within the Caribbean to this general pattern.

3.3.3 Genetic structure of rookeries

Our expanded analysis confirms previous observations (Leroux et al., 2012) that nearly all hawksbill rookeries in the West Atlantic can be considered distinct genetic populations. Of the 242 pairwise comparisons (groups excluded), only 16 (6.6%) did not differ significantly (Supplementary Tables S5.1, 5.2). Rookeries that were not significantly different fit two general categories. One set is dominated by EiA11 (67.7–74.6%), has small sample size, and few endemic or rare haplotypes (e.g., DR.S and TC). These populations were not well differentiated from others that fit that category. In contrast, the second set, the four Panama populations, have large sample size (n = 70–194) and multiple endemic haplotypes (Tables 1, 2; Figure 2B). The four Panama populations show an unexpected relationship, with Playa Larga and Chiriqui Beach (separated by more than 55 km) showing no significant difference but differing significantly from the two Zapatilla Cays populations (Supplementary Table S5.3) located geographically between them (Figure 2B). The Zapatilla Cays populations are not genetically differentiated, but this was expected as they are separated by less than 1.4 km and 6–12% of nesting females used both cays in recent nesting seasons.

The four Panama populations are part of a set of SWC populations that exhibit shared endemism (Table 2), a predominance of haplotype EiA11, and reduced genetic differentiation (Supplementary Table S5.2). While the average Fst value for pairwise comparisons for all rookeries in the study is 0.364 ± 0.246) that for six SWC beaches, including four from Panama, is 0.035 ± 0.32. However, a phylogram of 19 West Atlantic rookeries and the four Panama beaches (Figure 5) shows that the Panama beaches do not cluster together, but rather that PA.PL plus PA.CH cluster with NI.CP and CR.T. The two Zapatilla beaches cluster with other rookeries dominated by EiA11 (TC, USVI.BI, DR.S, and BB.W).

Figure 5

Monitoring at the four Panama study beaches (Meylan et al., 2013) revealed occasional but regular movement of nesting females between beaches both within and between nesting seasons. At least 45 females have been observed on more than one Bocas Region beach over a 20-year period (not including movements between the two Zapatilla Cays). An examination of the number of females that moved relative to the distance between beaches showed a strong negative relationship (Supplementary Figure S3). We also recorded three international movements of females that nested in Costa Rica that were later seen nesting in Panama, two at Chiriqui Beach and one at the Zapatilla Cays.

The absence of strong differentiation between the four Panama nesting beach populations and their very recent rapid growth in the last 20 years, combined with their shared endemic and near-endemic haplotypes (Table 2), and the documented movement of females between monitored beaches, suggested that further analyses were best approached by treating Panama as a single “rookery.” The best explanation for the difference between the Zapatilla and the other two rookeries is likely to be the result of founder effect at these beaches following a significant bottleneck (Meylan and Donelly, 1999). Thus, for MSAs and most other analyses and discussions, the four sampled beaches occurring within a ~100 km stretch of westernmost Caribbean Panama are considered a single “metapopulation” for which we have used the combined Panama rookery dataset (PA.grp).

3.3.4 Genetic structure of FGs

Fst values (Supplementary Tables S5.1–5.5) revealed less differentiation among benthic developmental FGs than among rookeries. In many cases, well-sampled FGs that are quite distant from one another did not differ significantly (e.g., BM.fg and PR.MI.fg, 0.001; BQ.fg and MX.QR.fg, 0.001; FL.UK.fg and PA.BT.fg, 0.003). However, three FGs located in southern Florida differed significantly (Fst 0.033 – 0.109), as did two Brazilian FGs (Fst = 0.030). Comparisons of genetic structure of FGs and adjacent rookeries typically showed significant differences: PA.BT.fg and PA.grp (0.081); PR.MI.fg and PR.MI (0.163); MX.PX.fg and MX.C (0.210); MX.QR.fg and MX.Y (0.551). However, there were exceptions: TT.fg does not differ from TT (0.005) and BR.C.fg does not differ from BR.P (0.011).

Although our Fst tables were calculated with EATL haplotypes included, the MSA runs were run without them. Based on recent confirmation of EiA48 and 49 from Guinea Bissau (Madeira et al., 2025) we can report that at least two FGs have a proportion of their input from EATL sources not used in MSA analyses. For BR.I.fg the estimated EATL input is just under 10%, for ASC.fg it is just over 13%.

3.4 Results of mixed stock analyses

3.4.1 Identifying non-zero contributions

Using foraging ground-centric MSAs, we compared four different lower limits for 95% CI values (Q 02.5) generated by MSA runs to identify the smallest value that can be used to indicate a non-zero contribution to an aggregation with the lowest risk of accepting false positives (Supplementary Table S6). We found that a minimum Q 02.5 value > 0.002 was the most useful non-zero cutoff. On average, the use of larger cutoffs did not sufficiently explain the total contribution to individual FGs (32.6–51.8% for > 0.01; 48.0–59.1% for > 0.005); a smaller cutoff (> 0.001) suggested that nearly three-quarters of rookeries were contributing to some FGs (higher chance of false positives). For foraging ground-centric runs, the use of > 0.002 as a “non-zero” cutoff explained an average of 65.3% (uniform priors) and 75.5% (rookery size prior) of aggregation genetic diversity and indicated that, on average, about seven rookeries contributed significantly to each FG. Furthermore, the > 0.002 minimum is supported by empirical evidence. EiA63, present in Bermuda, is known from two possible sources with Q 02.5 values of 0.0018 and 0.0019 (with a rookery size prior) for Bermuda; thus, at least one of these sources must be present. Similarly, EiA23 has been observed in BR.C.fg and its most likely source is MX.Y, the contribution for which has a Q 02.5 value of 0.0018 at BR.C.

3.4.2 Foraging ground-centric runs with uniform and rookery size priors

Many-to-many MSA foraging ground-centric runs with uniform and rookery size priors (Figures 6, 7; Supplementary Tables S7.1, 7.2) revealed that a single or multiple dominant source rookeries with mean contribution estimates > 0.20, were found in just six of 15 FGs under uniform priors and eight of 15 FGs with a rookery size prior. For foraging ground-centric runs, the use of a rookery size prior reduced the average CI by 13.5%; the use of this prior also explained more of the genetic variation at FGs than uniform priors (Supplementary Table S8). CIs were smallest for FGs at which contributions from genetically distinctive rookeries dominate (e.g., BR.C.fg, dominated by Brazilian rookeries; FL.PB.fg, and MX.PX.fg, dominated by Mexican rookeries). FGs with larger CIs were dominated by less genetically distinctive rookeries and, in most cases, had more contributing rookeries (e.g., ASC.fg, BM.fg, PR.MI.fg). The largest contribution to PA.BT.fg came from PA.grp; however, this rookery group also made significant contributions at nine distant and widespread FGs (BM.fg, BQ.fg, CAY.fg, CU.JR.fg, FL.PB.fg, FL.UK.fg, MX.QR.fg, PR.MI.fg, and TT.fg; Supplementary Table S7.2). These foraging ground-centric runs produced CIs that were 25% (uniform priors) and 33% (rookery size prior) smaller than the CIs obtained in rookery-centric runs (see rookery-centric results below).

Figure 6

Figure 7

3.4.3 Foraging ground-centric runs with a distance prior

Foraging ground-centric MSA runs using a modification of the distance prior method of Stahelin et al. (2022) and four minimum travel distances (1, >320, >520, and >1020 km; Supplementary Table S8), revealed CIs that were narrowest and very similar (0.095–0.097) when averaged across all FGs for rookery size plus the three shortest distances (1, >320, and >520 km). However, a 1 km minimum resulted in unrealistic results for TT.fg that were not consistent with the biology of the species, other runs in our study, or previous studies (e.g., Cazabon-Mannette et al., 2016). Results for both >320 and >520 km (with rookery size) are nearly identical to results obtained using a rookery size prior (Table 6) only but have slightly (~10%) smaller average CIs and accounted for slightly more genetic diversity at FGs. These two runs gave remarkably similar results (Supplementary Table S9) and for 13 of 15 FGs, they suggested that the same two rookeries were the largest contributors, and they agreed on 87% of all non-zero contributors to FGs. The largest difference between the two runs was observed for PR.MI.fg and PA.BT.fg, for which the largest (nearby) contributor at the >320 km minimum distance was not a significant contributor at the >520 km distance.

3.4.4 Rookery-centric results

Rookery-centric runs (Supplementary Tables S10.1, S10.2; Supplementary Figure S4) resulted in larger average CIs (rookery size prior, avg. = 0.1693 ± 0.039; uniform priors avg. = 0.1674 ± 0.030) than foraging ground-centric runs (rookery size prior, avg. = 0.1134 ± 0.028; uniform priors avg. = 0.1255 ± 0.023). Contributions to rookeries in rookery-centric MSAs include the category “unknown.” On average, this contribution differed little between runs with uniform and rookery size priors (0.095 ± 0.062, 0.097 ± 0.112, respectively). For the run with a rookery size prior, the “unknown” contribution provided between 3.9% and 51.7% of total contributions. For seven of 19 rookeries, the “unknown” category fit our criterion for a “non-zero” contribution and for three important rookeries—MX.C, PR.MI, and PA.grp—the “unknown” category was the largest contributor.

3.4.5 Groups-to-soups

Three parameters were used to evaluate the groups-to-soups analogy for FGs and suggested that this model works best as a continuum (Supplementary Table S11; Figure 8). A smaller SD for all rookery contributions to an FG suggests that contributions are more similar in size (soup-like), whereas a larger SD suggests greater variation among contributions (more group-like). For foraging ground-centric MSA runs, with a rookery size prior only or rookery size plus a medium distance (>320 or >520 km) priors, the SD of mean contribution varied from a low of 0.037 (BM.fg, rookery size only) to 0.172 (BR.I.fg, rookery size plus 520 km). The more soup-like FGs, with low SD, also have the highest number of rookeries making “non-zero” contributions and the largest single contribution to soup-like FGs is small. FGs with high SD have the smallest number of “non-zero” contributions to that FG, and the largest single contribution was often quite large (Figure 8).

Figure 8

Using the above criteria, there were few examples of strong linkages between rookeries and FGs. Only four of 15 FGs have one rookery contribution that was estimated to be >50%, and only one of these FGs (BR.I.fg) had a single rookery accounting for >75% of total contributions. For eight FGs, the largest contribution from a single rookery represented less than one-third of total contributions to that FG. The strongest linkages (resulting in the most group-like FGs) were observed between the two BR FGs and the two BR rookeries, between CAY.fg and TT, between two Florida FGs and MX.Y, and between MX.PX.fg and MX.Y. For the nine most soup-like FGs, between six and 11 source rookeries provide a “non-zero” contribution to that FG in the rookery size prior run, and the single largest contributing rookery provides an estimated 11.8–26.5% of the total aggregation.

The rookery-centric results suggest that, in most cases, individuals from rookeries are dispersed to multiple FGs in roughly equal proportions. For 15 of 19 rookeries, less than 25% of foraging is concentrated at any one of the FGs studied. The exceptions to this general pattern include the results for the Brazilian and Mexican rookeries. These rookeries relied mostly on in-country FGs. Additionally, a close relationship between TT and CAY.fg was reflected in both the rookery-centric and foraging ground-centric analyses. Hawksbills from the PA.grp appear to be dispersed principally to the local FG (PA.BT.fg) and, additionally, to unknown FGs.

4 Discussion

4.1 Haplotype diversity

Although Bowen et al. (2007) suggested that most of the mtDNA diversity of hawksbills in the Caribbean had been identified as of 2007, the widespread use of a longer sequence in recent studies has markedly increased the number of recognized haplotypes. With the addition of the Panamanian nesting beach populations and four previously unpublished FGs, we identified a total of 41 haplotypes that are informative in the assessment of the ecological geography of the hawksbill in the West Atlantic (Table 1). Three previously orphan haplotypes (EiA33, 35, 58) can now be attributed to a source; however, the frequency of orphan haplotypes and the number of haplotypes known from rookeries that are unknown at FGs suggest that much remains to be learned about the number and distribution of hawksbill haplotypes in the West Atlantic. The trajectories of the rarefaction curves, particularly for FGs (Figure 4B), suggest that sample sizes as large as 200 may still be insufficient to capture total haplotype diversity.

The four Panama nesting beach populations reported here host a total of nine endemic haplotypes (EiA33, 35, 55, 58, 64, 77, 91, 94, 95) that are all part of clade 3 (Figure 3). Three of these (EiA33, 35, 58) were previously considered to be orphans from FGs in Cuba, Puerto Rico and Florida (Díaz-Fernández et al., 1999; Velez-Zuazo et al., 2008; Wood et al., 2013). There are no published records for the other six. These nine haplotypes and three others (EiA02, 65, and 84) are endemic to the Southwest Caribbean (SWC); among these endemics, only EiA02 is a part of haplotype clade 1.

Five haplotypes (EiA12, 30, 43, 47, 63) found in Panama were identified as SWC “near-endemics” as they are known from few records outside of the SWC (Table 2). EiA12 was observed in 28 nesting females from Panama. It was first described from a nesting female from Belize (Bass et al., 1996), and subsequently detected at Tortuguero, CR (n = 5; Troëng et al., 2005) and Guadeloupe (n = 1; Leroux et al., 2012). Thus, 33 of 35 rookery records (94%) for EiA12 are from the SWC. EiA30 was detected in two samples from Chiriqui Beach and one from Tortuguero, CR, and was first described from the rookery at Doce Leguas, Cuba, (n = 1, Díaz-Fernández et al., 1999). Thus, three of four records (75%) are from the SWC. EiA43 was previously reported from 16 females from Nicaragua; we have added three records from Panama. It has also been reported from DR.J (n = 4), PR.MI (n = 3), and TT (n = 1); thus, 19 of 27 (70%) known occurrences are in the SWC. EiA47 was observed in 20 females from Panama nesting beach populations. It was previously reported from Tortuguero (n = 6, Troëng et al., 2005) and Jaragua, DR (n = 1; Carreras et al., 2013) and was recently discovered in Antigua (n = 1; Levasseur et al., 2019). Thus, 26 of 28 records (93%) are from the SWC. EiA63 is known from just two occurrences in nesting beach populations—one at Playa Larga, Panama, and one at Cabo de la Vela, Colombia.

4.2 Genetics at rookeries

4.2.1 Diversity and divergence among rookeries

Genetic diversity at sea turtle rookeries results from founding haplotypes, autochthonous evolution, genetic drift, and true biological dispersal of reproductive adults that results in geneflow into established rookeries (“leakage” of Bowen et al., 1989; natal dispersal of Shamblin et al., 2020). Our dataset confirms previous observations that a small number of common haplotypes dominate hawksbill rookeries, but local endemics and less common haplotypes lead to individual variation that can be quite localized. In contrast to findings for this mitochondrial control region marker in green turtles (e.g., van der Zee et al., 2019), hawksbill rookeries show a “curiously mixed pattern” (Leroux et al., 2012) where relatedness is patchy, there is occasionally strong similarity between distant rookeries, and few proximate rookeries are similar to one another (Figure 5). Structuring of hawksbill rookeries over time has not been detected (Velez-Zuazo et al., 2008; Leroux et al., 2012; Levasseur et al., 2019).

The distribution of the major haplotypes for hawksbills across West Atlantic rookeries is unexpected. Instead of regional domination by common haplotypes, as seen in Chelonia, the distributions of the two most common Eretmochelys haplotypes cross geographically (Figure 2). The most discordant distribution occurs in Barbados, with BB.L fixed for EiA01 and BB.W dominated by EiA11. Only rookeries in the Gulf of Mexico (GOM) are geographically structured; they are dominated by haplotypes from clade 2 (Figure 3), which are extremely rare at rookeries outside the GOM. At certain rookeries, either EiA01 or EiA11 appears to have been established for long enough that novel derivatives have arisen, producing haplotypes that differ by one or two base pairs (e.g., EiA01 in Brazil, EiA11 in Panama).

The complexity of genetic structuring at hawksbill rookeries may result in part from extinction events due to heavy exploitation for the tortoiseshell trade and subsequent stochastic recolonization (Browne et al., 2010). The dominance of a single common haplotype at some rookeries may be evidence for recolonization of extinct populations (e.g., Guadeloupe, dominated by EiA09; Cuba and Barbados Leeward, dominated by EiA01).

The accumulation of multiple, closely related endemic haplotypes in Panama can be considered evidence of an older, historically substantial population (Grant and Bowen, 1998; Helfman et al., 2009). The large number of endemic haplotypes, all satellites around EiA11, indicates that these variants stem from mutations in local populations—rather than true dispersal (geneflow) from other populations. Nine of these haplotypes differ from EiA11 by one bp; two others differ by two bp; and four more appear to be derived from two other common SWC haplotypes, EiA09 or EiA47. High levels of endemism and low levels of outward geneflow have likely led to higher relative genetic diversity in the SWC (Table 4; Figure 4A). On average, SWC rookeries had more than twice as many haplotypes as other rookeries (even when sample size was considered). Both haplotype and nucleotide diversity were significantly higher for SWC rookeries than for rookeries outside of the SWC (t-test, h, p = 0.040; π, p = 0.002).

4.2.2 Causes of diversity and endemism for Panama and SWC rookeries

The high genetic diversity in the Panama and SWC rookeries may result from both intrinsic and extrinsic factors. There is evidence that the metapopulation has existed for a long time (see above), and an obvious oceanographic feature—the Panama-Colombia Gyre (Andrade and Barton, 2000)—has likely promoted isolation of SWC rookeries while it facilitated geneflow via true dispersal among SWC rookeries. This gyre circulates in the southwestern-most Caribbean, south of the Caribbean Current (Figure 9). It rotates counter-clockwise, southeast along the coast, where the six SWC study populations are located, and then recirculates to the north and west. This gyre, combined with less-than-absolute site fidelity for hawksbills (Supplementary Figure S3; Hill et al., 2018; Levasseur et al., 2021; this study), could easily result in stochastic mixing along the coast from Nicaragua to Panama. The relative isolation of the SWC from the Caribbean Current is also supported by studies using drifters to resolve travel patterns of marine turtles (Blumenthal et al., 2009; Proietti et al., 2014). Very few drifters from the eastern Caribbean and Greater Antilles were captured by this gyre as they crossed the Caribbean Sea. Instead, drifters generally moved northwest with the Caribbean Current and exited this sea via the Yucatan Current. This is in marked contrast to drifters that started from Nicaragua and Costa Rica, which circulated in the Panama-Colombia Gyre for some time before leaving via the Caribbean and Yucatan currents. Thus, the combined age and relative oceanographic isolation of the SWC rookeries could explain the high level of endemism in the region.

Figure 9

Gaspar et al. (2022). The major “current conveyor system” (see text) in this region consists of the South Equatorial and North Brazil currents (not shown), Guyana, Caribbean, Yucatan, Loop, and Florida currents, and the Gulf Stream.

4.2.3 Geneflow, dispersal shadows, and rookery diversity

As “natal homing is not absolute” (Finn et al., 2016), there is occasional haplotype “leakage” (Bowen et al., 1989). This is revealed by geneflow via true biological dispersal, i.e., sweepstakes events in which a nesting female has moved to a rookery other than the one at which she hatched. True dispersal can be detected when a haplotype known to be common at one or more rookeries occurs out of place given the known distribution of haplotypes among rookeries. A dispersal (geneflow) shadow will occur where nesting populations are subject to very little leakage and thus maintain their genetic character. True dispersal and the geneflow that results are most easily detected in these dispersal shadows.

It is important to note that the term “dispersal” is regularly misused in the sea turtle literature to describe the movements of hatchlings away from nesting beaches (e.g., Blumenthal et al., 2009; Putman and Mansfield, 2015; Phillips et al., 2025). Dispersal is “the suite of behaviors that results in the movement of individuals away from their natal population to a different breeding population” (Buoro and Carlson, 2014) and it results in geneflow. Since hatchling sea turtles will most likely return to their natal beach to reproduce, this is migration (developmental migration) and not true biological dispersal as it will normally not result in geneflow. “Dispersion” is a better term for what happens to hatchlings in oceanic environments (Hays et al., 2010; Carreras et al., 2013; Scott et al., 2017). Shamblin et al. (2020) appear to have recognized this problem when they suggested the term “natal dispersal” for true dispersal.

Distinguishing dispersion of hatchlings during the epipelagic stage from dispersal between populations is important, because true dispersal leads to geneflow and homogenization of nesting populations (Arantes et al., 2020). However, it is possible that true dispersal could be a byproduct of developmental dispersion, so the two processes must be distinguished. We agree with Phillips et al. (2025) that post-hatchlings are regularly found in neritic waters (Meylan et al., 2011), so the term “oceanic stage” is not entirely appropriate. However, we feel that it would be inappropriate to use their proposed term “dispersal stage” since true biological dispersal is not occurring at this stage. Thus, we encourage the use of the epipelagic stage for this portion of the life cycle.

True dispersal by nesting females that has resulted in geneflow is an important contributing factor for rookeries in the SWC. Our data suggest that geneflow has been greater among nesting beach populations within the SWC than between outside rookeries and the SWC. Of nine Panama-endemic haplotypes, five were found in more than one nesting beach population. Of four SWC endemics, two were found in more than one nesting beach, and four of five near-endemics were found in multiple SWC populations. The evolution of a novel haplotype on one SWC beach and then subsequent dispersal to other beaches is the best explanation for the distribution of nine haplotypes on SWC beaches. Geneflow of this kind is explained by known movements of females between Panama beaches (Supplementary Figure S3) and between Costa Rica and Panama. Telemetry data reported by Dunbar et al. (2025) show that females from some nesting beaches in the SWC become entrained in the Panama-Colombia Gyre during the inter-nesting interval, which is likely to promote movement between nesting beaches within the SWC.

Geneflow within the SWC is in contrast to geneflow into the SWC from outside. For the SWC, our study only revealed three likely cases of geneflow from non-SWC rookeries, all on Panama beaches: EiA01 (eight occurrences), EiA20 (one occurrence), and EiA23 (one occurrence). All three of these “outside” haplotypes were also found in the Panama FG sample (n = 38); thus, one explanation for these true dispersal events is that individuals from distant rookeries that would normally return home via natal homing failed to do so and instead nested on beaches near their developmental habitat. Collectively, these data suggest that the SWC rookeries lie in a dispersal shadow.

A second case for a dispersal or geneflow shadow exists for rookeries in the GOM that lie westward of the main axis of the Yucatan and Loop currents which pass into the central Gulf. Dispersion into the western Gulf, and especially into the Bay of Campeche, is likely to be rare. This would explain the absence of geneflow of any outside haplotypes into the rookeries at MX.C and MX.Y. A developmental aggregation within the Gulf (MX.PX.fg) is made up mostly of local genotypes, with only five of 43 sampled individuals likely to have originated from outside of the Gulf (Table 3). This is in marked contrast to MX.QR.fg, which lies on the main axis of the dominant current from the Caribbean and largely comprises non-Mexican haplotypes. The existence of a dispersion shadow in the western GOM and Bay of Campeche is corroborated by a set of 112 tracks of epipelagic sea turtles in the GOM (Phillips et al., 2025). Only three of these tracks entered the western GOM and none reached the Bay of Campeche.

True dispersal and geneflow out of the SWC was also detected. All examples involve haplotypes we describe as SWC near-endemics. Three of these have clearly leaked out: EiA12 with single occurrences in Belize and Guadeloupe; EiA30 with a single occurrence in Cuba; and EiA47 with a single occurrence at DR.S. The case for the fourth haplotype, EiA43, is less certain. EiA43 is best considered an SWC near-endemic; 19 of 27 observations (70%) occurred at Nicaragua and Panama rookeries (Table 1).

4.3 Foraging grounds

4.3.1 Developmental migrations and the developmental habitat hypothesis

The results of our study fit a model of developmental migration, in which dispersion of hatchlings to benthic foraging grounds occurs over a wide area of the Caribbean and West Atlantic. Our dataset expands the opportunity to estimate the range over which developmental migrations occur. Hawksbills from the Panama rookery make “non-zero” contributions to ten FGs, from Bermuda and Florida to Tobago (Supplementary Figure S5). There is one record of the SWC endemic EiA02 in Brazil (BR.I.fg; Table 3), but no evidence for transatlantic developmental migration from the SWC. Although hawksbill hatchlings from Panama beaches are dispersed widely across the West Atlantic during their epipelagic stage, reproductive adults appear to reside in a much more restricted area. Unpublished tag returns and satellite telemetry (P. Meylan, A. Meylan, C. Ordoñez, D. Evans and Sea Turtle Conservancy) suggest Panama nesting females reside in an area from Belize to the Pedro Cays (Jamaica), and south along the coast of Nicaragua.

4.3.2 The role of currents and gyres in dispersion

Currents clearly play a key role in the lives of marine organisms (Gaspar et al., 2006). However, recent studies of the “lost years” of the sea turtle lifecycle suggest that post-hatchlings do not simply drift with major currents, but that swimming is also important (Witherington, 2002; Mansfield et al., 2014, 2021; Phillips et al., 2025). Given this observation, in combination with the effects of variation in major currents over time (Ascani et al., 2016), a simple distance prior (Lahanas et al., 1998; Nishizawa et al., 2013) may be of limited value in MSAs. However, compared to a rookery size prior alone, use of a distance prior that incorporated “probable paths” (Stahelin et al., 2022) along with a rookery size prior, produced a slightly smaller average CI and accounted for slightly more genetic diversity at FGs (Supplementary Table S8). With the addition of minimum distance priors of 320 and 520 km, average CI decreased by 12.8 and 11.9% respectively over the use of a rookery size prior alone. The small difference between these two runs is due to the fact that there are very few alternate “probable paths” when the minimum distance is increased by 200 km. The average distance along “probable paths” between all rookeries and FGs (as used in this study) is 1835 ( ± 3334) km.

Our dataset shows that both the number and diversity of haplotypes at FGs increase downstream along a “current conveyor system” made up of the major currents of the West Atlantic region (Supplementary Figure S2). However, looping currents like the Panama-Colombia Gyre, eddies that spin off of the Loop Current in the GOM and the Gulf Stream, variation in regional currents (https://data.marine.copernicus.eu/viewer/expert?view), and the slow rotation of the Sargasso Sea all offer possibilities for post-hatchlings to circulate in the same region—rather than move in a linear fashion. Outliers to this downstream pattern include PR.MI.fg and MX.PX.fg. The higher-than-expected haplotype diversity at PR.MI.fg and its genetic similarity to BM.fg (Fst = 0.001) could be explained by input from multiple currents. The lower-than-expected haplotype diversity at MX.PX.fg is undoubtedly due to the fact that it is in the same dispersal shadow as MX.C described above.

The general idea of a “conveyor system” model is that hatchlings are added to the system as major currents move in one direction and pass additional rookeries. Nishizawa et al. (2013) demonstrate this effect in Japan, where the North Equatorial Current delivers hatchlings from Pacific rookeries to southern Japanese FGs and the Kuroshio Current mostly delivers hatchlings from the Japanese Ogasawara Rookery to northern Japanese FGs. Proietti et al. (2014) suggest why there are so many haplotypes at BM.fg in the Atlantic, the last sampled site along the West Atlantic conveyor system. Figures in that paper show that drifters that start near any of the Caribbean hawksbill rookeries eventually pass into the Gulf Stream and out into the Atlantic, or else move into the Antilles Current and out into the Atlantic. Brazilian endemic haplotypes EiA32 and EiA61 (three individuals) were found in Bermuda and most likely arrived via the Antilles Current, rather than the major current system described above. The drifter paths in Figure 4b of Proietti et al. (2014) show that some drifters from the vicinity of both Brazilian rookeries followed the Antilles Current towards the Sargasso Sea and Bermuda; those that took a more northerly route were in the vicinity of Bermuda after two years. We also note the recovery of a juvenile tagged in Brazil in Bermuda (Santos et al., 2019). A conveyor system also fits the thesis of Proietti et al. (2014) that Brazilian FGs have lower genetic diversity because they have fewer, less diverse source rookeries. The same is likely to be true for Ascension Island. The conveyor system model also aligns with studies suggesting that hawksbills from the GOM, Caribbean, and Central America become entrained in the Loop Current and then encounter suitable benthic habitat along the Florida coast (Meylan and Redlow, 2006; Gorham et al., 2014; Ramirez et al., 2023).

Hatchlings from Mexican beaches may represent a special case. Drifter particles originating on Mexican beaches spend long periods within the GOM, but some do exit via the Loop Current (Blumenthal et al., 2009; Proietti et al., 2014). This explains the predominance of Mexican haplotypes among epipelagic hawksbills stranded along the Texas coast (Bowen et al., 2007). Our results agree that some hatchlings are retained, but some do leave. The endemic Campeche haplotype, EiA39, was found at two FGs in Florida and in Bermuda. MX.C is estimated to contribute 5.9% of the aggregation in Bermuda, 6.3% at CU.JR.fg, and 11.1%, 22.2%, and 4.2% at three aggregations in Florida.

Although downstream dispersion may predominate, there are multiple cases of haplotype dispersion against prevailing currents. SWC and Mexican endemic haplotypes occur at very low frequencies in the eastern Caribbean and in Brazil. Coastal counter-currents such as the Panama-Colombia Countercurrent (Andrade et al., 2003) may contribute to eastward travel against the prevailing currents in the Caribbean. Both Bass et al. (2006) and Meylan and Redlow (2006) argue that FGs at the confluence of currents are more diverse. This could also be an alternative explanation to the conveyor system for the high diversity at Bermuda and Florida, which could receive input from both the Gulf Stream and Antilles currents.

4.4 Mixed stock analyses

The use of MSAs to assess connectivity between sea turtle nesting populations and FGs is subject to important caveats (e.g., Bowen et al., 2007), some of which are particularly relevant for hawksbills. Firstly, there are large errors associated with contribution estimates due to shared common haplotypes among rookeries. For hawksbills in the West Atlantic, EiA01 and EiA11 predominate at 20 of 24 rookeries. Second, hawksbills were heavily exploited at rookeries followed by sporadic and stochastic recolonization. Third, not all rookeries and FGs have been sampled; some important sites remain unsampled. Fourth, sites included in previous MSAs based on short reads (Belize rookery, Inagua and Turks and Caicos FGs) cannot be used in our MSAs, reducing completeness. Fifth, the use of a distance prior requires additional testing. Modification of the method of Stahelin et al. (2022) in this study proved useful. However, recent publications (e.g., Phillips et al., 2025) suggest that the dispersion behavior of hatchlings, including changes in swimming ability, is more complex than these models presume. Sixth, there is the additional complication that currents may vary over time (Ascani et al., 2016; Scott et al., 2017). Circular currents, areas of low current velocity, and counter-currents could retain epipelagic-stage individuals and keep them out of major current transport systems for variable periods. Finally, although temporal changes at rookeries may not be an issue for MSAs (Velez-Zuazo et al., 2008), recent work makes it clear that—for Chelonia mydas—genetic structure at some FGs is changing as rookeries recover (van der Zee et al., 2019; Meylan et al., 2024). We can expect the same for hawksbills in the Caribbean.

Given these caveats, our criteria for comparing MSA runs, and results of MSAs with four different distance priors (Supplementary Table S8), we hereafter focus on results of foraging ground-centric MSAs with a rookery size prior only or rookery size plus minimum >320 km and >520 km distance priors (Supplementary Figures S6, S7). The results of MSA runs incorporating distance priors were similar to those using the rookery-size prior alone (Supplementary Table S8), with the former producing slightly smaller average CIs and accounting for slightly more genetic diversity at FGs (Supplementary Figure S7).

4.4.1 Foraging ground-centric MSAs

Our results strongly support previous observations that there is less genetic divergence among FGs than among rookeries (Bowen et al., 2007). This is undoubtedly due to the degree of mixing that occurs during the epipelagic phase before post-hatchlings move to a benthic developmental site (Lahanas et al., 1998). A high degree of mixing should result in “soups”; little or no mixing should result in “groups” (Engstrom et al., 2002; Blumenthal et al., 2009). This analogy appears to be useful when considered as a continuum between more group-like to more soup-like FGs. We found only limited evidence for distinct groups at hawksbill FGs.

We are not aware of any published methods to quantitatively measure the groups-to-soups continuum that can be observed graphically in the pie charts in Figure 6 (e.g., MX.PX.fg vs. BM.fg) and can be quantified using the standard deviation of average estimated contributions from MSAs. In more group-like FGs there is greater variation of observations relative to the mean; in soup-like FGs, all contributions deviate less from the mean. Rarefaction curves (Figure 4B) provide additional evidence that some FGs are more soup-like (BM.fg, PR.MI.fg), whereas others are more group-like (ASC.fg, BR.C.fg, MX.PX.fg).

Some FGs along the major current conveyor system are likely to receive input from many upstream rookeries (MX.QR.fg, CU.DL.fg, PR.MI.fg, BM.fg). However, there are also FGs to which epipelagics are less likely to be dispersed, including MX.PX.fg because it is in a current shadow and ASC.fg and BR.C.fg because they are upstream of most rookeries. However, the conveyor system does not fully explain groups and soups. The FGs at CAY.fg, FL.PB.fg, and FL.KW.fg are far downstream along the main current system but are dominated by input from a small number of rookeries. CAY.fg is dominated by input from TT and two Florida FGs are dominated by input from Mexico (specifically, MX.C). Perhaps, these represent cases where epipelagics from specific rookeries have reached the appropriate size to make the transition to benthic developmental habitat when they reach these sites, leading to more group-like aggregations.

4.4.2 Rookery-centric MSAs

Rookery-centric approaches in our study provided novel insights into both genetically studied and unstudied (unknown) FGs for individual rookeries. Rookery-centric MSAs support a pattern of long-distance connectivity and, for most rookeries, a large number of FGs contributing to a given rookery. However, rookeries with maximum estimated contributions to a single FG (> 0.20) were located within the same subregion as the FG (BR.B and BR.P to Brazilian FGs; MX.C and MX.Y to MX.PX.fg; PA.grp to PA.BT.fg; TC to CU.JR.fg). The inclusion of an unknown category in rookery-centric runs provided additional evidence that much remains to be learned about connectivity for hawksbill populations in the West Atlantic; for three major rookeries (PA.grp, PR.MI, MX.C) the largest proportion of dispersion was to as-yet-unknown sites.

Long-distance travel by immature hawksbills is well known (Marcovaldi and Filippini, 1991; Meylan, 1999; Bellini et al., 2000; Grossman et al., 2007; Santos et al., 2019). The results of our rookery-centric MSAs suggest that this is not anomalous, but a regular part of hawksbill life history. They suggest that the PA.fg is the most significant host for AG.JB; BM.fg is most significant for BB.W, CO.CV, and USVI.BI; TT.fg is most significant for NI.CP; and MX.QR.fg is most significant for USVI.SP. These findings indicate that—far from being a nonmigratory species (Witzell, 1983), hawksbills in the West Atlantic regularly use resources at far-off developmental sites.

The regular, long-distance travel implied by the rookery-centric MSAs (Supplementary Tables S10.1, S10.2) is corroborated by recaptures of immatures tagged in developmental habitat and then later seen on distant nesting beaches. Bjorndal et al. (2008) reported a hawksbill tagged at Inagua, Bahamas that nested nine years later in Tobago; Ordoñez et al. (2010) reported a hawksbill tagged at Monito, Puerto Rico that nested multiple times at Playa Larga, Panama. In addition, a juvenile tagged at Atol dos Rocas, Brazil (BR.I.fg) nested at BB.L (Santos et al., 2019) and the recapture in Grenada of a hawksbill tagged as an immature in Bermuda (Meylan et al., 2011) suggests developmental migration towards a likely nesting beach. There is also an unpublished record of a juvenile observed regularly at FL.PB.fg from 2005–2010 that nested at MX.Y in 2022 (Larry Wood, unpublished.).

4.5 Importance of sample size

Our results illustrate the importance of sample size for accurately assessing the ecological geography of the hawksbill using the mtDNA control region marker. Large sample sizes from Panama nesting beach populations and from the Bermuda FG reveal a relationship between sample size, endemic haplotypes, and long-distance connectivity; although the contribution is small (~4%). Because we have large sample sizes the CI falls entirely within the “non-zero” range (0.0022 to 0.0192). Rarefaction curves for both rookeries and FGs (Figure 4) suggest that, at nearly all sites, additional haplotype diversity remains undiscovered. This is likely due to endemic haplotypes that occur at low frequencies in rookeries and are unlikely to be observed at FGs. The large number of orphan haplotypes observed must be due to the existence of unstudied nesting beaches and small sample size for others.

The rarefaction trajectories observed for 15 FGs (Figure 4B) suggest that even 200 samples are not sufficient to capture total haplotype diversity. The Bermuda aggregation is represented by 190 samples collected over 26 years (1993–2018). Yet, rarefaction curves for this and most other FGs have still not flattened. This may be due, in part, to recovering rookeries that did not previously contribute to these sites but now produce hatchlings in numbers sufficient to allow rare haplotypes to be detected. In addition, variation in current patterns from year to year may result in haplotypes being differently dispersed over long timeframes. An excellent example of the importance of sample size is the discovery of the haplotype EiA83 at AG.JB when the sample size was expanded to 250 individuals (Levasseur et al., 2019). This haplotype and EiA20 were not detected in a smaller sample (n = 72) from this site (Leroux et al., 2012).

4.6 Conservation considerations

4.6.1 Conservation importance of Panama and rookeries in the SWC

Much of the haplotype diversity known for hawksbills in the West Atlantic is largely restricted to SWC rookeries. Seventeen of 41 haplotypes known from rookeries are endemic or nearly endemic to the SWC. This high level of genetic diversity exists despite historically high levels of exploitation of hawksbills in this region.

We elected to treat the four study beaches in Panama as one “rookery”, although we recognize their complex relationship to beaches in Costa Rica and Nicaragua. We think that this treatment is best because it promotes a regional approach to the conservation and recovery of hawksbills in Panama. There is good evidence that Caribbean Panama was a major contributor to hawksbill numbers historically; however, because of intensive exploitation (Meylan et al., 2013) that peaked in the 1960’s, concentrated nesting sites in western Caribbean Panama were decimated. The Chiriqui Beach population is estimated to have suffered a decline of 98% (Meylan and Donnelly, 1999). Tagging studies along the Bocas region coast indicate increases in nesting populations (Meylan et al., 2013; unpublished data, P. Meylan, A. Meylan, C. Ordoñez, A. Gonzalez Hooker, and G. Castillo) and reveal that reproductive females shift their nesting sites between beaches on a rare but regular basis. Hence, there is a good opportunity for true dispersal of hawksbills along this coast. We see indications that the nesting populations in the Bocas region are not at equilibrium but are in the process of recovery. The importance of Caribbean Panama nesting beaches for West Atlantic hawksbills is clear, given increasing numbers of nesting females, high genetic diversity, and reasonable expectation for continued protection of nesting beaches in the Parque Nacional Marino Isla Bastimentos and the Damani-Guariviara RAMSAR site, which includes Chiriqui Beach.

4.6.2 General conservation considerations

The decimation of hawksbills by the end of the last century (Meylan and Donnelly, 1999) is reflected in the perception that the species is a dispersed, solitary or “non-colonial” nester (Hendrickson, 1980; Witzell, 1983). While it is true that hawksbills will nest in small numbers where other sea turtle species will not, nesting has been reported to reach densities as high as 660 nests/km (Salm et al., 1993).

Ehrenfeld (1981) suggested that one limitation of marine turtle conservation is that you cannot change the biology of the species you are trying to conserve. In the case of the hawksbill, it is likely that the species’ biology has promoted resilience. Contrary to early reports, nesting can be both concentrated and dispersed and occurs on a wide variety of beach types. Although hawksbills make regular use of wide, high-energy beaches preferred by other species (e.g., green turtles at Tortuguero; leatherbacks at Chiriqui Beach), they also use very small pocket beaches and will climb across root mats and stony beaches to nest. This option for cryptic nesting at dispersed sites likely led to the survivorship of a diversity of haplotypes in Panama despite extensive harvesting where nesting was concentrated. The nesting season in Panama is also very long. Most nests are laid from May to October but have been observed in all months. Individuals nesting outside of the main season were more likely to escape detection during the years of concentrated exploitation. Panama is not the only location where the genetic diversity of hawksbills has survived a bottleneck. Campbell (2014) makes this point about the Dominican Republic, citing the work of Carreras et al. (2013), that the loss of genetic variability during population reductions appears to have been relatively low at these remnant nesting beaches.

The wide dispersion of epipelagics to distant developmental sites, the soup-like nature of many sites, and the long growth period spent at those sites (Avens et al., 2021) means that many individuals can be in the pipeline to maturation for long periods. Individuals at different sites are likely to experience different growth rates (Ramirez et al., 2023) and levels of local harvesting (Campbell, 2014). The high rate of domestic recaptures of juvenile hawksbills (León and Diez, 1999; Wood et al., 2013) reflects strong site fidelity over long periods during this development stage. Thus, long-term conservation at FGs is key to maintaining healthy rookeries. Our work shows that many FGs are contributing to distant rookeries; therefore, protecting turtles at any developmental FG is not just protecting local nesting populations but also populations that may be thousands of kilometers distant.

The results of our study suggest a small but significant adjustment to the Regional Management Units (RMUs) for the hawksbill in the Atlantic. RMUs 29 (Northwest Atlantic) and 30 (Southwest Atlantic) (Wallace et al., 2023; Hays et al., 2025) should be shown to overlap. Multiple lines of evidence support this adjustment. Our results indicate significant contributions from the rookery BR.B (in RMU 30) to FGs in Bermuda and Tobago (in RMU 29; Supplementary Table S7.2). There is also evidence that juveniles from RMU 29 use developmental habitats in RMU 30. Haplotypes EiA23 and 24 are only known from MX.Y and DR.J (as well as unpublished rookeries on the Yucatan Peninsula; Anahí Martínez Arenas, unpublished data), yet both are known from BR.C.fg (Proietti et al., 2014). Haplotypes EiA02, endemic to the SWC, and EiA28, endemic to Tobago, have been observed at BR.I.fg (Vilaça et al., 2013). Additionally, a juvenile tagged at Atol dos Rocas, Brazil was later observed nesting at BB.L (Santos et al., 2019) which suggests that the significant relationship between BB.L and FGs in Brazil is not just an artifact of the presence of only EiA01 at this rookery.

The increase in nesting observed in Panama is likely due in large part to the implementation of CITES, the curtailment of the international tortoiseshell trade, along with protection of major nesting sites. When our studies began in Panama in the 1970’s, tortoiseshell was selling for around $50 per pound at the first transaction. One turtle could be worth >$200. Once the international trade stopped, demand was greatly reduced, and nesting populations began to recover. In Panama, it has taken 30 years for nesting densities to return to a level where it is clear that the hawksbill is a communal nester. In order to more thoroughly understand the impact of CITES and recovery efforts across the West Atlantic, funds should be made available to increase the number of sampled rookeries and FGs in the region. Rookeries located up-current from the sites with the largest percentage of orphans (e.g., MX.QR.fg) could provide the greatest potential to identify the sources of the 17 orphans documented by our study. Each of the Panama nesting beach populations—and in fact all SWC nesting areas—have at least one unique haplotype, which suggests that surveys of additional SWC beaches could help to resolve the origin of remaining orphans. In Caribbean Panama, these beaches include Escudo de Veraguas, an island 17 km off of Chiriqui Beach; Playa Roja, a beach 17 km NW of the western limits of Chiriqui Beach; and Bluff Beach on Isla Colon near the town of Bocas del Toro. The beaches in Caribbean Nicaragua are also likely to be a source of additional orphan haplotypes.

Significant work remains to be done, sampling additional rookeries and FGs, and resampling well-known FGs in order to fully exploit the mtDNA control region marker. Unsampled rookeries located 300–500 km upstream of FGs with orphan haplotypes might prove to be most rewarding. Funding agencies should be attentive to this need as we look for ways to allow all range states to contribute to the conservation of this critically endangered species in the most optimal way possible. Moreover, future genomic sequencing will be required to resolve rookery connectivity at finer scales and to elucidate the role of males, thereby advancing our understanding of the ecological geography of hawksbills throughout the West Atlantic and across their broader range.

References

• 1

  AbreuA.HorrocksJ.FormiaA.DuttonP.LeRouxR.Vélez-ZuazoX.et al. (2006). “ New mtDNA D-loop primers which work for a variety of marine turtle species may increase the resolution of mixed stock analysis,” in Book of Abstracts. Proceedings of the 26th Annual Symposium on Sea Turtle Biology and Conservation. Eds. FrickM.PanagopoulouA.ReesA. F.WilliamsK. ( International Sea Turtle Society, Athens, Greece), 179. Available online at: http://www.nmfs.noaa.gov/pr/pdfs/species/turtlesymposium2006_abstracts.pdf.

  • Google Scholar

• 2

  AndradeC. A.BartonE. D. (2000). Eddy development and motion in the Caribbean Sea. J. Geophys. Res.: Oceans105, 26191–26201. doi: 10.1029/2000JC000300

  • CrossRef
  • Google Scholar

• 3

  AndradeC. A.BartonE. D.MooersC. N. (2003). Evidence for an eastward flow along the Central and South American Caribbean Coast. J. Geophys. Res: Oceans108.C6. doi: 10.1029/2002JC001549

  • CrossRef
  • Google Scholar

• 4

  ArantesL. S.VargasS. M.SantosF. R. (2020). Global phylogeography of the critically endangered hawksbill turtle (Eretmochelys imbricata). Genet. Mol. Biol.43.2 (2020), e20190264. doi: 10.1590/1678-4685-gmb-2019-0264

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 5

  AscaniF.Van HoutanK. S.Di LorenzoE.PolovinaJ. J.JonesT. T. (2016). Juvenile recruitment in loggerhead sea turtles linked to decadal changes in ocean circulation. Global Change Biol.22, 3529–3538. doi: 10.1111/gcb.13331

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 6

  AvensL.RamirezM. D.GosheL. R.ClarkJ. M.MeylanA. B.TeasW.et al. (2021). Hawksbill sea turtle life-stage durations, somatic growth patterns, and age at maturation. Endangered Species Res.45, 127–145. doi: 10.3354/esr01123

  • CrossRef
  • Google Scholar

• 7

  BandeltH.-J.ForsterP.RöhlA. (1999). Median-joining networks for inferring intraspecific phylogenies. Mol. Biol. Evol.16, 37–48. doi: 10.1093/oxfordjournals.molbev.a026036

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 8

  BassA. L.EpperlyS. P.Braun-McNeillJ. (2006). Green turtle (Chelonia mydas) foraging and nesting aggregations in the Caribbean and Atlantic: impact of currents and behavior on dispersal. J. Hered.97, 346–354. doi: 10.1093/jhered/esl004

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 9

  BassA. L.GoodD. A.BjorndalK. A.RichardsonJ. I.HillisZ. M.HorrocksJ. A.et al. (1996). Testing models of female reproductive migratory behaviour and population structure in the Caribbean hawksbill turtle, Eretmochelys imbricata, with mtDNA sequences. Mol. Ecol.5, 321–328. doi: 10.1111/j.1365-294X.1996.tb00323.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 10

  BelliniC.SanchesT. M.FormiaA. (2000). Hawksbill turtle tagged in Brazil captured in Gabon, Africa. Mar. Turtle Newslett.87, 11–12. http://www.seaturtle.org/mtn/archives/mtn87/mtn87p11.shtml

  • Google Scholar

• 11

  BenjaminiY.HochbergY. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc Ser. B Stat. Method.57, 289–300. doi: 10.1111/j.2517-6161.1995.tb02031.x

  • CrossRef
  • Google Scholar

• 12

  BjorndalK. A.ClovisT.ReichK. J.AlkinsG.EliazarP. J.BoltenA. B. (2008). Juvenile hawksbill tagged in the Bahamas nests in Tobago. Mar. Turtle Newslett.122, 10–11. http://www.seaturtle.org/mtn/archives/mtn122/mtn122p10.shtml

  • Google Scholar

• 13

  BlumenthalJ. M.Abreu-GroboisF. A.AustinT. J.BroderickA. C.BrufordM. W.CoyneM. S.et al. (2009). Turtle groups or turtle soup: dispersal patterns of hawksbill turtles in the Caribbean. Mol. Ecol.18, 4841–4853. doi: 10.1111/j.1365-294X.2009.04403.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 14

  BolkerB. (2015). Mixstock: mixed stock analysis in R ( R package version 0.9.9). https://github.com/bbolker/mixstock

  • Google Scholar

• 15

  BolkerB.OkuyamaT.BjorndalK.BoltenA. (2003). Sea turtle stock estimation using genetic markers: accounting for sampling error of rare genotypes. Ecol. Appl.13, 763–775. doi: 10.1890/1051-0761(2003)013%5B0763:STSEUG%5D2.0.CO;2

  • CrossRef
  • Google Scholar

• 16

  BowenB. W.BassA. L.Garcia-RodriguezA.DiezC. E.van DamR.BoltenA.et al. (1996). Origin of hawksbill turtles in a Caribbean feeding area as indicated by genetic markers. Ecol. Appl.6, 566–572. doi: 10.2307/2269392

  • CrossRef
  • Google Scholar

• 17

  BowenB. W.GrandW. S.Hillis-StarrZ.ShaverD. J.BjorndalK. A.BoltenA. B.et al. (2007). Mixed-stock analysis reveals the migrations of juvenile hawksbill turtles (Eretmochelys imbricata) in the Caribbean Sea. Mol. Ecol.16, 49–60. doi: 10.1111/j.1365-294X.2006.03096.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 18

  BowenB. W.MeylanA. B.AviseJ. C. (1989). An odyssey of the green sea turtle: Ascension Island revisited. Proc. Natl. Acad. Sci.86, 573–576. doi: 10.1073/pnas.86.2.573

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 19

  BrowneD. C.HorrocksJ. A.Abreu-GroboisF. A. (2010). Population subdivision in hawksbill turtles nesting on Barbados, West Indies, determined from mitochondrial DNA control region sequences. Conserv. Genet.11, 1541–1546. doi: 10.1007/s10592-009-9883-3

  • CrossRef
  • Google Scholar

• 20

  BuoroM.CarlsonS. M. (2014). Life-history syndromes: Integrating dispersal through space and time. Ecol. Lett.17, 756–767. doi: 10.1111/ele.12275

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 21

  CampbellC. L. (2014). Conservation Status of Hawksbill Turtles in the Wider Caribbean, Western Atlantic and Eastern Pacific Regions (Virginia USA: IAC Secretariat Pro Tempore). Available online at: www.iacseaturtle.org.

  • Google Scholar

• 22

  CarrA. (1980). Some problems of sea turtle ecology. Am. Zoologist20, 489–498. doi: 10.1093/icb/20.3.489

  • CrossRef
  • Google Scholar

• 23

  CarrerasC.GodleyB. J.LeónY. M.HawkesL. A.RevueltaO.RagaJ. A.et al. (2013). Contextualising the last survivors: Population structure of marine turtles in the Dominican Republic. PloS One8, e66037. doi: 10.1371/journal.pone.0066037

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 24

  Cazabon-MannetteM.BrowneD.AustinN.HaileyA.HorrocksJ. (2016). Genetic structure of the hawksbill turtle rookery and foraging aggregation in Tobago, West Indies. J. Exp. Mar. Biol. Ecol.485, 94–101. doi: 10.1016/j.jembe.2016.09.002

  • CrossRef
  • Google Scholar

• 25

  ChaoA.GotelliN. J.HsiehT. C.SanderE. L.MaK. H.ColwellR. K.et al. (2014). Rarefaction and extrapolation with Hill numbers: a framework for sampling and estimation in species diversity studies. Ecol. Monogr.84, 45–67. doi: 10.1890/13-0133.1

  • CrossRef
  • Google Scholar

• 26

  CorpetF. (1988). Multiple sequence alignment with hierarchical clustering. Nucleic Acids Res.16, 10881–10890. doi: 10.1093/nar/16.22.10881

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 27

  DayJ.HallJ.RoseK.Vinette HerrinK.MarchD.PittO.et al. (2024). Mixed stock analysis identifies natal origins of green turtles at foraging grounds in southeastern Australia. Front. Mar. Sci.11. doi: 10.3389/fmars.2024.1346932

  • CrossRef
  • Google Scholar

• 28

  Díaz-FernándezR.OkayamaT.UchiyamaT.CarrilloE.EspinosaG.MárquezR.et al. (1999). Genetic sourcing for the hawksbill turtle, Eretmochelys imbricata, in the northern Caribbean region. Chelonian Conserv. Biol.3, 296–300.

  • Google Scholar

• 29

  DunbarS. G.EvansD. R.EggersL. R.BergmanQ. D.FonsecaL. G.PaladinoF. V.et al. (2025). Inter-nesting area use, migratory routes, and foraging grounds for hawksbill turtles (Eretmochelys imbricata) in the Western Caribbean. PloS One20, e0317778. doi: 10.1371/journal.pone.0317778

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 30

  DuttonP. H. (1996). “ Methods for collection and preservation of samples for sea turtle genetic studies,” in Proceedings of the International Symposium on Sea Turtle Conservation Genetics ( NOAA Tech Memo. NMFS-SEFSC), 396. Available online at: https://repository.library.noaa.gov/view/noaa/8453.

  • Google Scholar

• 31

  EhrenfeldD. (1981). “ Options and limitations in the conservation of sea turtles,” in Biology and Conservation of Sea Turtles. Ed. BjorndalK. ( Smithsonian Institution Press, Washington, D.C), 457–463.

  • Google Scholar

• 32

  EngstromT. N.MeylanP. A.MeylanA. B. (2002). Origin of juvenile loggerhead turtles (Caretta caretta) in a tropical developmental habitat in Caribbean Panama. Anim. Conserv.5, 125–133. doi: 10.1017/S1367943002002184

  • CrossRef
  • Google Scholar

• 33

  ExcoffierL.LischerH. E. (2010). Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows. Mol. Ecol. Resour.10, 564–567. doi: 10.1111/j.1755-0998.2010.02847.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 34

  FinnS. A.ThompsonW. P.ShamblinB. M.NairnC. J.GodfreyM. H. (2016). Northernmost records of hawksbill sea turtle nests and possible trans-Atlantic colonization event. Mar. Turtle Newslett.151, 27. http://www.seaturtle.org/mtn/archives/mtn151/mtn151-7.shtml

  • Google Scholar

• 35

  GasparP.CandelaT.ShillingerG. L. (2022). Dispersal of juvenile leatherback turtles from different Caribbean nesting beaches: A model study. Front. Mar. Sci.9. doi: 10.3389/fmars.2022.959366

  • CrossRef
  • Google Scholar

• 36

  GasparP.GeorgesJ. Y.FossetteS.LenobleA.FerraroliS.Le MahoY. (2006). Marine animal behaviour: neglecting ocean currents can lead us up the wrong track. Proc. R. Soc B. Biol. Sci.273, 2697–2702. doi: 10.1098/rspb.2006.3623

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 37

  GelmanA.RubinD. B. (1992). Inference from iterative simulation using multiple sequences. Stat. Sci.7, 457–472. doi: 10.1214/ss/1177011136

  • CrossRef
  • Google Scholar

• 38

  GorhamJ. C.ClarkD. R.BresetteM. J.BagleyD. A.KeskeC. L.TraxlerS. L.et al. (2014). Characterization of a subtropical hawksbill sea turtle (Eretmochelys imbricata) assemblage utilizing shallow water natural and artificial habitats in the Florida Keys. PloS One9, e114171. doi: 10.1371/journal.pone.0114171

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 39

  GrantW. A. S.BowenB. W. (1998). Shallow population histories in deep evolutionary lineages of marine fishes: insights from sardines and anchovies and lessons for conservation. J. Hered.89, 415–426. doi: 10.1093/jhered/89.5.415

  • CrossRef
  • Google Scholar

• 40

  GrossmanA.BelliniC.FallabrinoA.FormiaA.MbaJ. M.MbaJ. N.et al. (2007). Second TAMAR tagged hawksbill recaptured in Corisco Bay, West Africa. Mar. Turtle Newslett.116, 26. http://www.seaturtle.org/mtn/archives/mtn116/mtn116p26.shtml

  • Google Scholar

• 41

  HaysG. C.FossetteS.KatselidisK. A.MarianiP.SchofieldG. (2010). Ontogenetic development of migration: Lagrangian drift trajectories suggest a new paradigm for sea turtles. J. R. Soc Interface.7, 1319–1327. doi: 10.1098/rsif.2010.0009

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 42

  HaysG. C.LaloëJ. O.SeminoffJ. A. (2025). Status, trends and conservation of global sea turtle populations. Nat. Revs. Biodivers.1, 119–133. doi: 10.1038/s44358-024-00011-y

  • CrossRef
  • Google Scholar

• 43

  HelfmanG. S.ColletteB. B.FaceyD. E.BowenB. W. (2009). The diversity of fishes: biology, evolution, and ecology. (Chichester, England: Wiley-Blackwell).

  • Google Scholar

• 44

  HendricksonJ. R. (1980). The ecological strategies of sea turtles. Am. Zool.2, 597–608. doi: 10.1093/icb/20.3.597

  • CrossRef
  • Google Scholar

• 45

  HillJ. E.KingC. M.StewartK. R.PaladinoF. V.DuttonP. H. (2018). Genetic differentiation of hawksbill turtle rookeries on St. Croix, US Virgin Islands. Chelonian Conserv. Biol.17, 303–308. doi: 10.2744/CCB-1293.1

  • CrossRef
  • Google Scholar

• 46

  HsiehT.MaK.ChaoA. (2024). iNEXT: Interpolation and Extrapolation for Species Diversity. ( R package version 3.0.1). Available online at: http://chao.stat.nthu.edu.tw/wordpress/software_download/.

  • Google Scholar

• 47

  JohnsonM.ZaretskayaI.RaytselisY.MerezhukY.McGinnisS.MaddenT. L. (2008). NCBI BLAST: a better web interface. Nucleic Acids Res.36, W5–W9. doi: 10.1093/nar/gkn201

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 48

  KearseM.MoirR.WilsonA.Stones-HavasS.CheungM.SturrockS.et al. (2012). Geneious Basic: an integrated and extendable desktop software platform for the organization and analysis of sequence data. Bioinform28, 1647–1649. doi: 10.1093/bioinformatics/bts199

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 49

  Labastida-EstradaE.Machkour-M’RabetS.Díaz-JaimesP.Cedeño-VázquezJ. R.HénautY. (2019). Genetic structure, origin, and connectivity between nesting and foraging areas of hawksbill turtles of the Yucatan Peninsula: A study for conservation and management. Aquat. Conserv.29, 211–222. doi: 10.1002/aqc.2999

  • CrossRef
  • Google Scholar

• 50

  LahanasP. N.BjorndalK. A.BoltenA. B.EncaladaS. E.MiyamotoM. M.ValverdeR. A.et al. (1998). Genetic composition of a green turtle (Chelonia mydas) feeding ground population: evidence for multiple origins. Mar. Biol.130, 345–352. doi: 10.1007/s002270050254

  • CrossRef
  • Google Scholar

• 51

  Lara-RuizP.LopezG. G.SantosF. R.SoaresL. S. (2006). Extensive hybridization in hawksbill turtles (Eretmochelys imbricata) nesting in Brazil revealed by mtDNA analyses. Conserv. Genet.7, 773–781. doi: 10.1007/s10592-005-9102-9

  • CrossRef
  • Google Scholar

• 52

  LeónY. M.DiezC. E. (1999). Population structure of hawksbill turtles on a foraging ground in the Dominican Republic. Chelonian Conserv. Biol.3, 230–236.

  • Google Scholar

• 53

  LerouxR. A.DuttonP. H.Abreu-GroboisF. A.LagueuxC. J.CampbellC. L.DelcroixE.et al. (2012). Re-examination of population structure and phylogeography of hawksbill turtles in the Wider Caribbean using longer mtDNA sequences. J. Hered.103, 806–820. doi: 10.1093/jhered/ess055

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 54

  LevasseurK. E.StapletonS. P.FullerM. C.QuattroJ. M. (2019). Exceptionally high natal homing precision in hawksbill sea turtles to insular rookeries of the Caribbean. Mar. Ecol. Prog. Ser.620, 155–171. doi: 10.3354/meps12957

  • CrossRef
  • Google Scholar

• 55

  LevasseurK. E.StapletonS. P.QuattroJ. M. (2021). Precise natal homing and an estimate of age at sexual maturity in hawksbill turtles. Anim. Conserv.24, 523–535. doi: 10.1111/acv.12657

  • CrossRef
  • Google Scholar

• 56

  MadeiraF. M.CatryP.RebeloR.BarbosaC.RegallaA.PatrícioA. R. (2025). Status and biology of hawksbill turtles at Poilão Island, Guinea-Bissau: a key stronghold for the species in West Africa. Biodivers. Conserv. (2025) 1–19. doi: 10.1007/s10531-025-03084-z

  • CrossRef
  • Google Scholar

• 57

  MansfieldK. L.WynekenJ.LuoJ. (2021). First Atlantic satellite tracks of ‘lost years’ green turtles support the importance of the Sargasso Sea as a sea turtle nursery. Proc. R. Soc B. Biol. Sci.288, 20210057. doi: 10.1098/rspb.2021.0057

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 58

  MansfieldK. L.WynekenJ.PorterW. P.LuoJ. (2014). First satellite tracks of neonate sea turtles redefine the ‘lost years’ oceanic niche. Proc. R. Soc B. Biol. Sci.281, 20133039. doi: 10.1098/rspb.2013.3039

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 59

  MarcovaldiM. A.FilippiniA. (1991). Trans-Atlantic movement by a juvenile hawksbill turtle. Mar. Turtle Newslett.52, 3. http://www.seaturtle.org/mtn/archives/mtn52/mtn52p3.shtml

  • Google Scholar

• 60

  MeylanA. B. (1999). International movements of immature and adult hawksbill turtles (Eretmochelys imbricata) in the Caribbean region. Chelonian Conserv. Biol.3, 189–194.

  • Google Scholar

• 61

  MeylanA. B.BrostB.ConradL. J.DenisonS. H.FlahertyD. B.GrayJ. A.et al. (2024). Feminization of a mixed-stock foraging aggregation of immature green turtles (Chelonia mydas), 1975–2018. Mar. Biol.171, 11. doi: 10.1007/s00227-023-04320-2

  • CrossRef
  • Google Scholar

• 62

  MeylanA. B.DonnellyM. (1999). Status justification for listing the hawksbill turtle (Eretmochelys imbricata) as critically endangered on the 1996 IUCN Red List of Threatened Animals. Chelonian Conserv. Biol.3, 200–224.

  • Google Scholar

• 63

  MeylanP. A.HardyR. F.GrayJ. A.MeylanA. B. (2022). A half-century of demographic changes in a green turtle (Chelonia mydas) foraging aggregation during an era of seagrass decline. Mar. Biol.169, 74. doi: 10.1007/s00227-022-04056-5

  • CrossRef
  • Google Scholar

• 64

  MeylanP. A.MeylanA. B.GrayJ. (2011). Ecology and migrations of sea turtles 8. Tests of the developmental habitat hypothesis. Bull. Am. Mus. Nat. Hist.357, 1–70. doi: 10.1206/357.1

  • CrossRef
  • Google Scholar

• 65

  MeylanA. B.MeylanP. A.OrdoñezC. O. (2013). Sea turtles of Bocas del Toro Province and the Comarca Ngöbe-Buglé, Republic of Panamá. Chelonian Conserv. Biol.12, 17–33. doi: 10.2744/CCB-0948.1

  • CrossRef
  • Google Scholar

• 66

  MeylanA.RedlowA. (2006). “ Eretmochelys imbricata—Hawksbill Turtle,” in Biology and Conservation of Florida Turtles. Ed. MeylanP. A.Lunenberg, Massachusetts, USA: Chelonian Research Foundation, 105–127. Chelon. Res. Monogr.

  • Google Scholar

• 67

  Monzón-ArgüelloC.LoureiroN. S.DelgadoC.MarcoA.LopesJ. M.GomesM. G. (2011). Príncipe island hawksbills: Genetic isolation of an eastern Atlantic stock. J. Exp. Mar. Biol. Ecol.407, 345–354. doi: 10.1016/j.jembe.2011.07.017

  • CrossRef
  • Google Scholar

• 68

  MortimerJ. A.DonnellyM. (2007). Marine Turtle Specialist Group 2007 IUCN Red List Status Assessment Hawksbill Turtle (Eretmochelys imbricata). Available online at: http://www.iucnredlist.org/documents/attach/8005.pdf.

  • Google Scholar

• 69

  MortimerJ. A.DonnellyM.MeylanA. B.MeylanP. A. (2007b). Critically endangered hawksbill turtles: molecular genetics and the broad view of recovery. Mol. Ecol.16, 3516–3517. doi: 10.1111/j.1365-294X.2007.03432.x

  • CrossRef
  • Google Scholar

• 70

  MortimerJ. A.MeylanP. A.DonnellyM. (2007a). Whose turtles are they, anyway? Mol. Ecol.16, 17–18. doi: 10.1111/j.1365-294X.2006.03252.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 71

  NishizawaH.NaitoY.SuganumaH.AbeO.OkuyamaJ.HirateK.et al. (2013). Composition of green turtle feeding aggregations along the Japanese archipelago: Implications for changes in composition with current flow. Mar. Biol.160, 2671–2685. doi: 10.1007/s00227-013-2261-1

  • CrossRef
  • Google Scholar

• 72

  OrdoñezC.MeylanA. B.MeylanP. A.PetersonI.DiezC. E.van DamR. P. (2010). Hawksbill tagged as a juvenile in Puerto Rico found nesting in Panama 15 years later. Mar. Turtle Newslett.127, 25. http://www.seaturtle.org/mtn/archives/mtn127/mtn127p25.shtml

  • Google Scholar

• 73

  OwensD. W.RuizG. T. (1980). New methods of obtaining blood and cerebrospinal fluid from marine turtles. Herpetologica36, 17–20.

  • Google Scholar

• 74

  Pérez-BermúdezE.UrquiolaA. R.Rebordinos-GonzálezL.Espinosa-LópezG.Lee-GonzálezI.PetricB. L. J. (2017). A new approach to mixed stock analysis that informs on the management and conservation of hawksbill in the wider Caribbean: the case of the legal fishery in Jardines del Rey, Cuba. Int. J. Mar. Sci.7, 316–343. doi: 10.5376/ijms.2017.07.0033

  • CrossRef
  • Google Scholar

• 75

  PhillipsK. F.PutmanN. F.MansfieldK. L. (2025). New insights on sea turtle behaviour during the ‘lost years’. Proc. R. Soc B. Biol. Sci.292, 20242367. doi: 10.1098/rspb.2024.2367

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 76

  ProebstelD. S.EvansR. P.ShiozawaD. K.WilliamsR. N. (1993). Preservation of nonfrozen tissue samples from a salmonine fish Brachymystax lenok (Pallas) for DNA analysis. J. Ichthyol.32, 9–17.

  • Google Scholar

• 77

  ProiettiM. C.ReisserJ.MarinsL. F.Rodriguez-ZarateC.MarcovaldiM. A.MonteiroD. S.et al. (2014). Genetic structure and natal origins of immature Hawksbill Turtles (Eretmochelys imbricata) in Brazilian waters. PloS One9, e88746. doi: 10.1371/journal.pone.0088746

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 78

  PutmanN. F.BaneJ. M.LohmannK. J. (2010). Sea turtle nesting distributions and oceanographic constraints on hatchling migration. Proc. R. Soc B. Biol. Sci.277, 3631–3637. doi: 10.1098/rspb.2010.1088

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 79

  PutmanN. F.MansfieldK. L. (2015). Direct evidence of swimming demonstrates active dispersal in the sea turtle “lost years. Curr. Biol.25, 1221–1227. doi: 10.1016/j.cub.2015.03.014

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 80

  RamirezM. D.AvensL.MeylanA. B.ShaverD. J.StahlA. R.MeylanP. A.et al. (2023). Dietary plasticity linked to divergent growth trajectories in a critically endangered sea turtle. Front. Ecol. Evol.11. doi: 10.3389/fevo.2023.1050582

  • CrossRef
  • Google Scholar

• 81

  RichardsonP. B.BrufordM. W.CalossoM. C.CampbellL. M.ClearveauxW.FormiaA.et al. (2009). Marine turtles in the Turks and Caicos Islands: Remnant rookeries, regionally significant foraging stocks, and a major turtle fishery. Chelonian Conserv. Biol.8, 192–207. doi: 10.2744/CCB-0871.1

  • CrossRef
  • Google Scholar

• 82

  SalmR. V.JensenR. A. C.PapastravrouV. A. (1993). Marine Fauna of Oman: Cetaceans, Turtles, Seabirds and Shallow Water Corals. (Gland, Switzerland: IUCN), 66.

  • Google Scholar

• 83

  SantosA. J. B.BelliniC.BortolonL. F. W.OuterbridgeB.BrowneD. C.SantosA.et al. (2019). Long-range movements and growth rates of Brazilian hawksbill turtles revealed by flipper tags. Chelonian Conserv. Biol.18, 75–81. doi: 10.2744/CCB-1343.1

  • CrossRef
  • Google Scholar

• 84

  ScottR.BiastochA.AgamboueP. D.BayerT.BoussambaF. L.FormiaA.et al. (2017). Spatio-temporal variation in ocean current-driven hatchling dispersion: Implications for the world’s largest leatherback sea turtle nesting region. Divers. Distrib.23, 604–614. doi: 10.1111/ddi.12554

  • CrossRef
  • Google Scholar

• 85

  Sequencher^® version 5.4.6 DNA sequence analysis software (MI USA: Gene Codes Corporation, Ann Arbor). Available online at: http://www.genecodes.com.

  • Google Scholar

• 86

  ShamblinB. M.HartK. M.MartinK. J.CerianiS. A.BagleyD. A.MansfieldK. L.et al. (2020). Green turtle mitochondrial microsatellites indicate finer-scale natal homing to isolated islands than to continental nesting sites. Mar. Ecol. Prog. Ser.643, 159–171. doi: 10.3354/meps13348

  • CrossRef
  • Google Scholar

• 87

  SlowikowskiK. (2024). ggrepel: Automatically Position Non-Overlapping Text Labels with ggplot2 ( R package, version 0.9.6). Available online at: https://CRAN.R-project.org/package=ggrepel.

  • Google Scholar

• 88

  StahelinG. D.HoffmanE. A.Quintana-AscencioP. F.ReuscheM.MansfieldK. L. (2022). Incorporating distance metrics and temporal trends to refine mixed stock analysis. Sci. Rep.12, 20569. doi: 10.1038/s41598-022-24279-2

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 89

  StringellT. B.ClerveauxW. V.GodleyB. J.PhillipsQ.RangerS.RichardsonP. B.et al. (2015). Protecting the breeders: research informs legislative change in a marine turtle fishery. Biodiversity Conserv.24, 1775–1796. doi: 10.1007/s10531-015-0900-1

  • CrossRef
  • Google Scholar

• 90

  TroëngS.DuttonP. H.EvansD. (2005). Migration of hawksbill turtles Eretmochelys imbricata from Tortuguero, Costa Rica. Ecography28, 394–402. doi: 10.1111/j.0906-7590.2005.04110.x

  • CrossRef
  • Google Scholar

• 91

  Trujillo-AriasN.AmorochoD. F.López-ÁlvarezD.Mejía-LadinoL. M. (2014). Relaciones filogeográficas de algunas colonias de alimentación y anidación de la tortuga carey (Eretmochelys imbricata) en el Pacífico y Caribe Colombianos. Boletín Investigaciones Marinas y Costeras-INVEMAR43, 159–182. doi: 10.25268/bimc.invemar.2014.43.1.39

  • CrossRef
  • Google Scholar

• 92

  van der ZeeJ. P.ChristianenM. J.NavaM.Velez-ZuazoX.HaoW.BérubéM.et al. (2019). Population recovery changes population composition at a major southern Caribbean juvenile developmental habitat for the green turtle, Chelonia mydas. Sci. Rep.9, 14392. doi: 10.1038/s41598-019-50753-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 93

  Velez-ZuazoX.RamosW. D.van DamR. P.DiezC. E.Abreu-GroboisA.McMillanW. O. (2008). Dispersal, recruitment and migratory behavior of a hawksbill sea turtle population. Mol. Ecol.17, 839–853. doi: 10.1111/j.1365-294X.2007.03635.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 94

  VilaçaS. T.Lara-RuizP.MarcovaldiM. A.SoaresL. S.SantosF. R. (2013). Population origin and historical demography in hawksbill (Eretmochelys imbricata) feeding and nesting aggregates from Brazil. J. Exp. Mar. Biol. Ecol.446, 334–344. doi: 10.1016/j.jembe.2013.06.004

  • CrossRef
  • Google Scholar

• 95

  WallaceB.BandimereA.Abreu-GroboisA.AcostaH.AkitiJ.AkomediM.et al. (2025). Updated global conservation status and priorities for marine turtles. Endangered Species Res.56, 247–276. doi: 10.3354/esr01385

  • CrossRef
  • Google Scholar

• 96

  WallaceB. P.PosnikZ. A.HurleyB. J.DiMatteoA. D.BandimereA.RodriguezI.et al. (2023). Marine turtle regional management units 2.0: an updated framework for conservation and research of wide-ranging megafauna species. Endangered Species Res.52, 209–223. doi: 10.3354/esr01243

  • CrossRef
  • Google Scholar

• 97

  WestlakeN. (2024). Population genetics of green (Chelonia mydas) and hawksbill turtles (Eretmochelys imbricata) in Grenada, West Indies. University of Plymouth, Plymouth, UK. Available online at: https://pearl.pymouth.ac.uk.bms-these/447/. Master’s Thesis.

  • Google Scholar

• 98

  WickhamH. (2016). “ Programming with ggplot2,” in ggplot2: Elegant Graphics for Data Analysis ( Springer International Publishing, Cham), 241–253. Available online at: http://dx.doi.org/10.1007/978-3-319-24277-4_12.

  • Google Scholar

• 99

  WilkeC. (2024). cowplot: Streamlined Plot Theme and Plot Annotations for ggplot2 ( R package, version 1.1.3). Available online at: https://CRAN.R-project.org/package=cowplot.

  • Google Scholar

• 100

  WitheringtonB. (2002). Ecology of neonate loggerhead turtles inhabiting lines of downwelling near a Gulf Stream front. Mar. Biol.140, 843–853. doi: 10.1007/s00227-001-0737-x

  • CrossRef
  • Google Scholar

• 101

  WitzellW. N. (1983). Synopsis of biological data on the hawksbill turtle, Eretmochelys imbricata (Linnaeus 1766) FAO fish. Synop, Vol. 137.

  • Google Scholar

• 102

  WoodL.HardyR.MeylanP.MeylanA. (2013). Characterization of a hawksbill turtle (Eretmochelys imbricata) foraging aggregation on a high-latitude reef community in southeastern Florida, USA. Herpetol. Conserv. Biol.8, 258–275.

  • Google Scholar