Abundance, distribution, and origins of hawksbill sea turtles on the nearshore Atlantic reefs of the Florida Keys National Marine Sanctuary

Hawksbill sea turtles (Eretmochelys imbricata) representing multiple regional stocks are known to reside in SE Florida, and play an important functional role in the health of coral reef ecosystems. However, baseline data concerning the abundance and distribution of hawksbills in the Florida Keys are lacking. Between 2015-2024, over 280 hours of in-water snorkeling surveys on the shallow coral reefs (<10 m) in the Florida Keys National Marine Sanctuary (FKNMS) from Key Largo (25.4276N, 80.1246W) through Key West (24.4463N, 81.9268W) resulted in the hand-capture of 76 hawksbill turtles (mean straight carapace length 39.3 cm (range 19.3 - 63.5 cm). Growth rates among re-captured turtles averaged 6.3 cm yr -1 (range 2.5-10.2 cm yr -1, n = 12). Among the seventeen distinct Caribbean haplotypes identified via mtDNA, mixed-stock analyses revealed a significant but weakening bias toward representatives of Mexican rookeries coupled with indications of increasing diversity over time. It is apparent the shallow reefs of the FKNMS represent valuable entry points for post oceanic-stage hawksbills establishing themselves in the State’s neritic habitats and should be regarded as critical to sustaining the long-term presence of this highly imperiled species in Florida waters.

The State of Florida (USA) features diverse aquatic and semi-aquatic habitats that host numerous chelonian species, including up to five species of marine turtles (Witherington and Witherington, 2024). Located in the southernmost part of the State, the Florida Keys National Marine Sanctuary (FKNMS) was established in 1990 to protect and manage over 2, 900 square miles (7500 square km) of marine habitats surrounding the Florida Keys. The FKNMS features multiple habitat types, including mangroves, seagrass beds, mixed hard-bottom, and coral reef ecosystems. The expansive shallow coral reef habitats in the Florida Keys, including patch reef systems and spur-and-groove barrier reefs, attract a wide array of marine organisms, including hawksbill sea turtles (Eretmochelys imbricata).

Hawksbill turtle populations have been drastically depleted throughout the Caribbean basin, largely due to a long-term and intense commercial harvest to supply the global tortoiseshell trade (Meylan and Donnelly, 1999). As a result, they remain listed as “Critically Endangered” by the International Union for the Conservation of Nature (IUCN, 2026), and “Endangered” under the Endangered Species Act (Mortimer and Donnelly, 2008; NMFS & USFWS, 2013). Though Florida’s beaches have not historically served as nesting grounds for hawksbills, nearshore areas are known as developmental foraging habitat for juveniles and sub-adults (Meylan and Redlow, 2006; Eaton et al., 2008; Hart et al., 2013; Wood et al., 2013; Gorham et al., 2014; Herren et al., 2018). Given that oceanic currents are primarily responsible for hatchling and post-hatchling dispersal patterns in hawksbill turtles (Blumenthal et al., 2009), Florida’s aggregation is comprised of genetic representatives from several regional sub-populations (Wood et al., 2013). However, little is known about what happens to small juveniles upon their arrival in coastal habitats.

As is common among hard-shelled sea turtles, juvenile hawksbills undertake an ontogenic shift from pelagic to neritic habitats, followed by a fairly long period of time to maturity that can exceed 20 years (Musick and Limpus, 1997; Godley et al., 2008; Avens et al., 2021). Because mortality rates for eggs and hatchlings are disproportionately high, the improved survival probabilities of the individuals that transition through the juvenile and sub-adult life stages to maturity has increasingly large implications for future reproductive output (Salmon et al., 2018). Indeed, the rate of somatic growth is critically associated with the time required to achieve sexual maturity, and therefore is an important factor to consider for estimating population structure.

Hawksbill growth rates in the Western Atlantic experienced declines between 1997-2015, potentially due to the negative effects of rising sea surface temperatures on reef ecosystems (Bjorndal et al., 2016). Continued intensification of sea surface temperatures raises ongoing concern for hawksbill food and coral reef habitat availability (Huang et al., 2024). Thus, continued monitoring of key regions for hawksbill recruitment and growth is necessary for population status and growth assessments. Relative abundance, habitat use, growth rates, and baseline information concerning the spatial ecology of immature turtles are vitally important to inform overall recovery efforts (Wildermann et al., 2018), yet these data are often logistically difficult to obtain, particularly over large, meaningful spatiotemporal ranges.

Though the fine-scale mechanisms that influence the turtles’ ability to initially locate suitable nearshore habitats remain unknown, hawksbill turtles broadly appear to transition from the pelagic to the nearshore environments at sizes that range between 20–25 cm, and, strongly influenced by ocean currents, subsequently recruit to mixed-stock foraging aggregations (Meylan and Redlow, 2006; Blumenthal et al., 2009; Meylan et al., 2011). Recruitment site choices are highly consequential as each new arrival must negotiate through a suite of stochastic challenges that include, but are not limited to, predator avoidance, food availability, inter- and intra-species competition, and even acute exposure to novel parasites, contaminants and chemical compounds. New recruits that find themselves in the most favorable locations, either through random or adaptive processes (or some combination of the two), gain a competitive advantage at the outset (Diez and van Dam, 2002). It is well documented that hawksbills commonly associate with and select coral reef habitats (Meylan and Redlow, 2006; Selby et al., 2019; Matley et al., 2021), so it is likely that the new arrivals to Florida would preferentially settle in areas dominated by rocky, hard-bottom substrate. Due to its positioning at the interface between the Gulf of Mexico and the Atlantic Ocean, young turtles drifting southeast in the Loop Current within the Gulf of Mexico, and those drifting northwest in the Antilles Current from the Caribbean, have a good chance of encountering the FKNMS as they exit the Gulf through the Florida Straits (Wood et al., 2013). As such, we hypothesized that the Sanctuary may serve as a key entry point to SE Florida’s coastal reef system for hawksbill turtles representing numerous source rookeries, particularly those in Mexico’s Yucatan Peninsula. Additionally, we hypothesize the shallow reef habitats (<10m) serve primarily as temporary foraging grounds for small juveniles that subsequently gain access to deeper water as they transition to sub-adults. To address these hypotheses, we carried out a study conducted between 2015 and 2024, designed to document the abundance, distribution, growth rates, natal origins, and behavior of immature hawksbill turtles in the FKNMS. Our results provide strong evidence that a sizeable population of hawksbills are indeed recruited to, and reside within, this area.

Study site

Snorkel and/or scuba diving surveys were focused on the shallow (3–18 m; mean 7.3 m; median 6 m) offshore hardbottom habitats of the Southeast Florida Continental Reef Tract from Key Largo (25.379, -80.8443) through Key West (24.4457, -81.8268) (Figure 1). Additional surveys were conducted at artificial reef sites north of Key West (East and West Jetty) and near Long Key (Long Key Reef), and briefly on the Gulf side of the lower Keys (Snipes Creek and Content Keys). A total of 69 sites were surveyed at least once (Supplementary Table S1). Due to a combination of logistical constraints and our general understanding of hawksbill ecology, the survey effort was biased toward patches of observably moderate to high benthic rugosity that included spur and groove reefs, patch reefs, and mixed hard-bottom communities.

Sampling

Between May and October, 6-24–2015 and 9-15-2024, snorkeling and/or scuba diving surveys were conducted annually between to take advantage of seasonally favorable sea conditions in the Florida Keys. Surveys were not attempted when surface waves exceeded 1 m; most were conducted in 0.5 m seas or less. Upon accessing a selected site by boat, between two and six observers (four on average) simultaneously snorkeled through the habitat, generally in the direction of the prevailing current. In some cases, up to four observers were simultaneously towed by rope approximately 10 meters behind the boat in slow, semicircular patterns. Upon sighting a turtle, observers were instructed to notify other team members of its location by raising one hand into the air and, if necessary, await the arrival of at least one other observer to pursue and subsequently capture turtle by hand. Due to the varied survey configurations employed and because it is often difficult to detect hawksbills in such structurally complex habitats, most sites were surveyed repeatedly on a rotational basis both within and between each field season. To mitigate against sampling inconsistencies across individual sites and years, cumulative observation times were summed for three broad zones (upper, middle, and lower Keys) (Supplementary Table S1).

Each captured turtle was immediately transported to a nearby research vessel for processing. Straight and curved carapace lengths (nuchal notch to the tip of the longest supracaudal scute) were measured with Haglof™ tree calipers and flexible tailor’s tape, respectively. Curved measurements were not recorded if raised barnacles intersected the line of measurement. Each turtle was placed in a nylon mesh dive bag and weighed to within 1 kg (due to boat sway) with an Outmate™ 300kg digital handheld crane scale. To assure their future identification at a later capture date, turtles were photographed from multiple angles with special attention to injuries or anomalies prior to release at or near the capture location.

External tags were affixed to the trailing edge of each front flipper in the second scale distal to the axilla (National Tag and band Co. Style 681 Inconel for turtles >30 cm SCL; Style 1005–3 for turtles <30 cm SCL). PIT tags were implanted in the front right shoulder of each turtle (Biomark Series APT-12 for turtles >30 cm SCL; Series HPT8 10mm for turtles <30 cm SCL). Tagging sites were sterilized with 70% alcohol prior to application. Opportunistically recaptured individuals were re-measured and new photos were taken. If necessary, their tags were inspected and replaced.

Survey effort was first converted from total search time (minutes) to hours for each site over the entire study period. Search time was inclusive of all in-water survey time. Catch per unit effort (CPUE) was calculated at the sampling-event level as the number of turtles captured divided by sampling effort in hours. Sites were assigned to one of the three geographical regions: upper Keys (Carysfort – Alligator Reef), middle Keys (Long Key Reef – Looe Key), and lower Keys (American Shoal – Western Dry Rocks) (Figure 1; Supplementary Table S1). Site-level CPUE was subsequently calculated as the total number of turtles captured at each site divided by the total sampling effort (hours) at that site. Regional CPUE was calculated as the total number of turtles captured across all sites within a region divided by the total sampling effort across those sites. Thus, regional CPUE estimates were weighted by sampling effort, with sites contributing proportionally to the regional estimates according to the amount of effort sampled. Zero-capture observations were retained in all calculations to account for sampling events in which no turtles were caught. Only novel, untagged turtles were used to calculate CPUE to provide a metric that more directly reflects the recruitment of new individuals into the sampled sites throughout the study period. Excluding recaptures minimized the potential for individuals exhibiting site fidelity to be disproportionately represented in CPUE through repeated captures.

Growth rate

Summary statistics for turtle carapace size (SCLmax and CCLmax) and weight (kg) were calculated for the first capture of each hawksbill, excluding all recapture measurements to avoid pseudoreplication. The relationship between body mass (kg) and straight carapace length (SCLmax cm) was examined using a linear regression. Only turtles with both SCLmax and weight measurements were included in the analysis.

Growth was calculated as the change in SCLmax between successive capture events divided by the days elapsed between captures, and then standardized to an annual rate (cm/year). We fitted linear mixed-effects models to examine the relationship between hawksbill turtle growth rate and body size, while accounting for repeated measurements among individuals that were recaptured multiple times by including an individual-level random effect. Mean SCLmax between consecutive captures was assigned as the corresponding measurement for each growth interval. Using additional recapture data from Palm Beach County collected between 2004 and 2025 using the same measurement protocols and techniques as described above (n = 55) (Wood et al., 2017; Wood, unpublished data), linear mixed-effects models with individual-level random effects were also used to generate a more comprehensive growth curve for the larger Florida aggregation. Growth rate (cm/year) was modeled as a function of mean straight carapace length (SCLmax) using both a simple linear regression and a second-order polynomial regression (including SCL and SCL² terms). Prior to analysis, data were filtered to include only observations with a measurement interval of at least 365 days. Model selection was based on Akaike’s Information Criterion (AIC) and the best model fit was used for inference. All growth rate statistics were calculated in RStudio (R version 4.3.1 (2023-06-16), R Core Team, 2026).

Estimated age–size relationships were developed using a von Bertalanffy growth function parameterized with mark–recapture growth intervals (von Bertalanffy, 1938), including hawksbills from Monroe County and the greater Southeast Florida reef track between April 2004- May 2025 (Wood, unpublished data) to gain a comprehensive view of the regional population. Growth parameters were estimated from changes in SCLmax between recaptures using the Fabens formulation (Fabens, 1965). To account for individuals that contributed more than one recapture interval, the Fabens growth model was fitted using nonlinear mixed effects with turtle identity included as a random effect. Population-level estimates of asymptotic length (L∞) and growth coefficients (k) were modeled as fixed effects. To convert the growth model to an age-size relationship, the estimated (L∞) and (k) were used to calculate the age parameter (t o) by anchoring the curve to the assumed size of 27.9 cm SCLmax (SCL x)at age 3 (Age x), based on predicted age-size relationships calculated for Florida hawksbills in Avens et al. (2021):

Individual age estimates were subsequently calculated using the size ranges observed in the Florida Keys dataset using the inverse von Bertalanffy equation:

Estimated uncertainty around predicted SCL values was quantified using the standard deviation of residuals from the non-linear mixed-effects Fabens von Bertalanffy growth model.

Genetics

A 4 mm skin biopsy sample was collected from the trailing edge of either rear flipper. Samples were preserved in Qiagen RNAprotect Tissue Reagent (cat no. 76106) and transferred on dry ice to the University of Florida Whitney Laboratory for Marine Biosciences. DNA was extracted from tissue samples (half of each 4 mm punch biopsies) using a Qiagen DNeasy Blood and Tissue Kit (Cat. No. 69504) per the manufacturer’s instructions. An 832 bp fragment of the mtDNA control region was then amplified using conventional PCR with the LCM15382 (forward) and H950 g (reverse) primers (Abreu-Grobois et al., 2006). PCR was conducted in a 20 μL reaction of 0.25 μL ExTaq DNA Polymerase (TaKaRa RR001), 2 μL 10X ExTaq Buffer, 1.6 μL 2.5mM dNTP mixture, 1.6 μL each of 10 μM forward and reverse primer, 10.95 μL nuclease-free water, and 2 μL template DNA with the following cycling conditions: 95 °C for 3 min, 40 cycles of 95 °C for 30 sec, 55 °C for 60 sec, and 72 °C for 30 sec, followed by a final 72 °C extension for 10 min. PCR products were analyzed on a 2% agarose gel and extracted using the Qiagen QIAquick Gel Extraction Kit (Cat. No. 28704) following the manufacturer’s instructions. Purified PCR products were sent to Plasmidsaurus, Inc. for sequencing using their Linear/PCR amplicon sequencing service, which utilizes long-read Oxford Nanopore Technology to cover the PCR product in a single read to an average read depth of over 1, 000 reads per PCR product.

For haplotype assignment, reference sequences were downloaded from the NCBI database. Reference and sample sequences were aligned separately using MUSCLE alignment on Geneious (v. R10), then trimmed to the shortest sample sequence length (730bp), starting from the 5’ end of the D-loop (Arantes et al., 2020). Sample sequences were matched to the reference haplotypes using Geneious Map to Reference and the alignment chart from Arantes et al. (2020). Untrimmed full length (832 bp) sequences were also analyzed using NCBI BLASTn (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastn&PAGE_TYPE=BlastSearch&LINK_LOC=blasthome) to verify trimming did not lose key markers for haplotype assignment.

Four chi-square tests using simulated p-values with 2000 replicates and a Benjamini & Hochberg p-value adjustment were conducted using R (v 4.4.1) and R Studio (v. 2024.04.2) to determine if haplotype distributions differed between each of the study sites and time periods: Key West (Gorham et al., 2014), Florida Keys (this study), Palm Beach (Wood et al., 2013), and Palm Beach 2015 - 2022 (Wood, unpublished data).

To estimate connectivity between source rookeries and the Florida foraging aggregation, Bayesian mixed stock analyses (MSA) were conducted using mitochondrial DNA (mtDNA) haplotype frequency data compiled from across the Northwest Atlantic (Meylan et al., 2026 and references within; Supplementary Table S2). Haplotype datasets included long (~740 bp) control region sequences representing major nesting populations throughout the Caribbean and Brazil.

MSAs were performed in R using the mixstock package (Bolker et al., 2007) under the many-to-many framework, implementing the Thomas–O’Hara (TO) model with a Gibbs sampler. Analyses were run using Markov chain Monte Carlo (MCMC) methods. Posterior summaries were extracted from the 2.5% and 97.5% quantiles to generate 95% credible intervals (CI).

The analysis focused on a biologically informed model incorporating ecological covariates of rookery size and distance rather than uniform priors. Rookery size was included as an informative prior to account for variation in nesting population productivity across the region. Biologically, rookery size acts as a proxy for total hatchling output, where larger nesting populations are assumed to contribute a greater absolute number of individuals into ocean currents. Number of nests was used for rookery size (Supplementary Table S3) (Meylan et al., 2026). Geographic distance between rookeries and the Florida foraging ground was incorporated as a second covariate. Distances between foraging grounds and hawksbill rookeries were determined using marine currents as vectors for transport. Distances between landmarks were summed for the total path between rookeries and foraging grounds (Supplementary Table S3). Following calculations from Stahelin et al. (2022), we transformed the estimated dispersal distances (D ij) between each rookery (j) and the Florida foraging aggregation (i) intro proportional weights to be used as informative priors. The effective inverse-distance weight (P) for each rookery-aggregation pair was calculated as.

Our final model used the modifier many-to-many model from Stahelin et al. (2022),

where Source contribution is based on the haplotype frequencies, Source size is the rookery information, and P is the inverse distance.

Alternative distance thresholds were evaluated to assess the robustness of results. We evaluated minimum- distance model of 1, 320, 520, and 1, 020 km, following Meylan et al. (2026). The 1 km model allowed extremely short routes to dominate the inverse-distance weights, and the 1, 020 km model excluded some rookeries. Specifically, a minimum threshold of 320 km was used to define plausible dispersal pathways, as prior work on hawksbills demonstrated that models incorporating thresholds of >320 km and >520 km produced nearly identical estimates (Meylan et al., 2026). These threshold-based models yielded slightly narrower credible intervals (~10% reduction) and explained marginally more genetic diversity at foraging grounds compared to models using rookery size priors alone. Rare haplotypes were consolidated within mixstock to reduce model complexity. Model performance was evaluated based on (1) the width of posterior credible intervals, (2) consistency among model runs, and (3) the proportion of observed genetic diversity explained by each model. Contributions were considered non-zero based on conservative thresholds of the lower 95% credible interval (Q2.5), following previous approaches to minimize false-positive assignments.

A total of 76 out of 113 hawksbills observed by the team were hand-captured during 17, 245 minutes (287.4 hours) of in-water surveys. These resulted in 64 new encounters and 12 recaptures. Additionally, three previously tagged turtles were positively identified but not otherwise disturbed. Regional CPUE for novel, untagged hawksbills was 0.26 hr -1 in the upper Keys (Carysfort – Alligator Reef), 0.15 hr -1 in the middle Keys (Long Key Reef – Looe Key), and 0.19 hr -1 in the lower Keys (American Shoal – Western Dry Rocks) (Supplementary Table S1). The greatest site-level CPUE for each region was 0.59 hr -1 (6 turtles/10.2 hr) at Cheeca Rocks (Upper), 0.31 hr -1 (2 turtles/6.5 hr) at West Turtle Shoal (Middle), and 0.34 hr -1 (11 turtles/32.5 hr) at Western Sambo.

The average SCLmax for all Monroe County hawksbills was 38.5 cm (19.3 - 63.5 cm; n = 64) and CCLmax was 40.3 cm (21.5–60 cm, n = 62; Figures 2A, B). Body mass for hawksbills averaged 7.9 kg (1–32 kg; n = 57; Figure 2C). Body mass increased significantly with SCLmax (R² = 0.899; t = 22.08, p < 0.001, n = 57; Figure 2D).

Growth rate

The average interval between Florida Keys hawksbill recaptures was 470 days (63–825 days ± 217 SD, n = 12). Only one turtle was recaptured at a different location from the initial capture location, relocating 3.7 km from West Sambo to the adjacent Middle Sambo reef. The mean mid-point size of the turtles between respective captures was 39.1 cm SCL ± 7.2 SD. Overall, the average growth rate among the turtles sampled was 6.3 cm yr-1 2.5 ± SD (range 2.5-10.2 cm yr-1).

To assess growth across the entire SE Florida population (n = 67), we first tested mixed-effects models to determine the best fit. When compared to the linear model (AIC = 234.6; ΔAIC = 3.9), the second-order polynomial regression provided the best fit for the relationship between size and growth rate (AIC = 230.7). The final model indicated a significant nonlinear relationship between SCLmax and growth rate (marginal R² = 0.67, conditional R² = 0.83, X2(2) = 67.64, p < 0.001) (Figure 3). Both the linear term (β = −0.36, SE = 0.09, p < 0.001) and the quadratic term (β = 0.0019, SE = 0.00076, p = 0.015) were significant, indicating that growth rate gradually declines with increasing carapace length. Residual standard deviation was 0.91 cm/year.

The Fabens von Bertalanffy growth model for all hawksbills in Southeast Florida (Key West – Jupiter, FL) was fitted using 67 recapture intervals from 56 individual turtles. The median time at liberty was 702 days (range = 28–3, 302 days). The estimated asymptotic straight carapace length (L∞) was 75.83 ± 2.01 cm, and the growth coefficient (k) was 0.19 ± 0.02 year - ¹ (Figure 4). The residual standard deviation of the fitted model was 1.56 cm. The predicted age range model-based estimates for hawksbill turtles in Monroe County based on SCL range of 19.3 cm – 63.5 cm were approximately 2.1 ± 0.15 years to 10.2 ± 0.68 years old.

Population structure

A total of 17 distinct haplotypes were identified from 42 individuals from Monroe County representing the Florida Keys foraging aggregation, captured between 17 August 2016 and 15 September 2024. For comparative purposes, haplotypes were also identified for 44 individuals captured in Palm Beach County between 12 May 2005 and 15 August 2024 (Figure 5).

The results of the chi-square tests show significant differences between Palm Beach 2013 and Florida Keys 2014 haplotype distributions (Table 1). Significant differences in haplotype distributions were also found in Palm Beach 2013 compared to Palm Beach 2022. There is no significant difference in haplotype distributions across the two sites in the 2022 and 2025 data, and no significant difference in the Florida Keys 2025 and Key West 2014 distributions.

The foraging ground-centric MSA for the Palm Beach and Key West areas, incorporating both rookery size and distance as priors, revealed a pattern of connectivity across the Northwest Atlantic. Mixed stock analyses revealed that contributions to both the Palm Beach and Florida Keys foraging aggregations were dominated by a small number of source rookeries, with some consistent patterns observed between historical and contemporary datasets. In the 2013 dataset (Wood et al., 2013), Mexico, Yucatan was the dominant contributor (mean = 0.34, 95% CI: 0.10–0.45), followed by Mexico, Campeche (mean = 0.27, 95% CI: 0.15–0.56). Additional contributions from Puerto Rico, Mona Island were moderate (mean = 0.11, 95% CI: 0.07–0.21), while all other rookeries contributed 5% or less (Figure 6). In the current Palm Beach dataset, Mexico–Yucatán had the highest contribution (mean = 0.259, 95% CrI: 0.158–0.376), followed by Panama (mean = 0.127, 95% CrI: 0.029–0.246) and Mexico–Campeche (mean = 0.102, 95% CrI: 0.021–0.242). Posterior comparisons supported a statistically lower contribution from Mexico–Campeche currently, compared to 2013 (mean difference = −0.166, 95% CrI: −0.428 to −0.005; posterior probability of a decrease = 0.979). The apparent differences for the remaining Palm Beach sources were not clearly supported because their posterior differences in contribution intervals included zero values (Supplementary Figure S1).

For the Florida Keys aggregation, Mexico–Yucatán had the highest contribution in the 2014 dataset (mean = 0.273, 95% CrI: 0.150–0.389), followed by Mexico–Campeche (mean = 0.125, 95% CrI: 0.029–0.318). In the current Florida Keys dataset, Panama had the highest contribution (mean = 0.186, 95% CrI: 0.069–0.312), followed by Mexico–Yucatán (mean = 0.143, 95% CrI: 0.062–0.241) and Mexico–Campeche (mean = 0.101, 95% CrI: 0.014–0.277). However, Panama was not greater than either Yucatán (mean difference = 0.042, 95% CrI: −0.122–0.198) or Campeche (mean difference = 0.085, 95% CrI: −0.138–0.262). Comparison of the two Florida Keys datasets showed a greater contribution from Panama in the current dataset (mean difference = 0.153, 95% CrI: 0.011–0.293; posterior probability of an increase = 0.983). All other rookery-specific contrasts overlapped zero, indicating they were not significantly different (Supplementary Figure S1). Contributions from Mexico, Campeche remained moderate (mean ≈ 0.10), while other rookeries continued to show low contributions with broad uncertainty (partially due to widely geographically distributed natal haplotypes).

Across both foraging grounds, Mexican rookeries represented major contributors across time periods. However, the relative dominance of these sources decreased in the more recent years, accompanied by increased contributions from rookeries further south, like Panama (Figure 7). Evidence for differences between datasets was source- and aggregation-specific rather than indicative of a general decline in nesting populations. Overall, the MSA indicated aggregation-level connectivity between Florida foraging grounds and source rookeries throughout the Caribbean and Gulf of Mexico. The combined estimated contribution from Brazilian rookeries was 0.058 (95% CrI: 0.008–0.147) for the current Florida Keys aggregation and 0.059 (95% CrI: 0.008–0.147) for the current Palm Beach aggregation. These estimates represent probabilistic contributions to each aggregation and do not assign Brazilian natal origins to individual turtles.

Determining the whereabouts at all life stages of migratory species broadens our understanding of habitat use and informs research and conservation measures. However, the logistical challenges associated with accessing marine turtles during their developmental stages have limited our ability to understand dispersal patterns, identify recruitment sites, measure relative abundance, and examine processes of growth and development. Given the slow pace at which marine turtles mature, long-term studies are often required to gather biologically significant information (van Dam and Diez, 1998a; Blumenthal et al., 2009a; Hart et al., 2013; Strindberg et al., 2016).

Though previous studies have described specific hawksbill aggregations in Florida, very few sustained multi-year surveys have yielded details of growth and/or long-term residency (Hart et al., 2012; Gorham et al., 2014; Wood et al., 2013). The contrast in median size between the aggregation encountered by Gorham et al. (2014) near Key West (21-4 - 69.0 cm mean 44.1 cm) and that encountered by Wood et al. (2013) in the Palm Beach aggregation 35.7–83.9 cm (mean = 56.6 cm) suggested that the more shallow (<10m) reef habitats of the FKNMS may be an entry point for small juveniles arriving as new recruits from offshore environments, while the more northern/deeper reefs served as secondary developmental grounds for primarily larger juveniles and subadults. Although no turtles tagged in the FKNMS have been positively identified in Palm Beach County, our results confirmed the prevalence of smaller juveniles on nearly all of the shallow barrier reefs of the SE Florida Continental Reef Tract from Key West through Key Largo. Though there are likely additional recruitment sites still to be identified, these particular habitats within the Sanctuary clearly contribute to the long-term sustainability of Florida’s hawksbill populations, and at a smaller scale, the regional populations they represent.

Abundance and distribution

Our findings further demonstrate the well-established association between hawksbills and highly rugose reef habitats. Turtles were regularly encountered on steep spur-and-groove reef formations, which predominate on the seaward edge of the reef line in the lower Keys, and to a lesser degree in the upper Keys near Key Largo. Though prey availability does not always dictate hawksbill abundance or distribution (Leon and Bjorndal, 2002; Rincon-Diaz et al., 2011), the relative lack of reef development and lowered benthic diversity in the middle Keys is likely partly responsible for the comparatively low encounter rate in this area (Precht and Miller, 2007). It is noteworthy that when evading capture, the juveniles are adept at avoiding detection by immediately attempting to conceal themselves within caves, crevices, and/or the complex tunnel systems that underlie the most rugose parts of the reef. Aided by their slender shape and defensive armoring, predator avoidance/defense remains especially important at this life stage, so access to protective refuges may strongly influence their habitat preferences (Salmon et al., 2018). As observed in other juvenile hawksbill populations in the region (Wood et al., 2017; Chevis et al., 2017; Selby et al., 2019), the turtles recaptured in this study appeared to at least temporarily exhibit moderate site fidelity, likely indicating a preference for familiar structures that can serve as protective retreats. Indeed, the ongoing decline in reef rugosity throughout the Caribbean (Alvarez–Fillip et al., 2009) raises concerns for hawksbill habitat availability, as well as how these reef-dependent turtles will alter their foraging and movement ecology.

Although widely varying capture methodologies and sampling strategies across studies makes direct comparisons of relative abundance nearly impossible, the encounter rates reported here are comparatively low among published studies in the region, including in Palm Beach County (see Wood et al., 2013). The relative proportion of the region’s juveniles that annually recruit specifically to the reefs of the FKNMS reflects a combination of fluctuating trends in regional reproductive output and randomly-shifting mechanisms of oceanic dispersal, so it is reasonable to assume that their abundance in the Sanctuary likely fluctuates over time as these two factors gradually move in and out of phase. Continued long-term monitoring will be required to further develop links between these large-scale regional factors and local population trends.

Of particular interest is the stark contrast between the abundance of hawksbills noted by Gorham et al. (2014) and this study at a 2 km long, 10 m-wide derelict rock rubble structure located at the edge of the northwest channel of Key West. Known locally as the “West Jetty”, the configuration of the boulders has created a high-relief, benthically diverse refuge in relatively shallow water (3–4 m) that attracts a broad range of marine life. Hawksbills at this site alone represented over 50% of the captures in the refuge between 2001 and 2011 (Gorham et al., 2014), yet our team encountered only one hawksbill there (which was, incidentally, tagged by another research team) during multiple surveys conducted in 2016, 2018, 2019, and 2023. Though there are no direct data available on predation rates of juvenile sea turtles by goliath grouper (Epinephelus itajara), they are known to prefer habitats with high structural complexity and to consume juvenile sea turtles (Collins, 2014; Randall, 1968, L. Wood, pers. obs.). Our team noted the presence of goliath grouper at West Jetty in both 2016 and 2023, and periodically on the high-relief sites throughout the Refuge. Based on long-term survey data, goliath grouper occurrence in the Florida Keys markedly increased between 2005-2010, gradually declined until 2018, then increased again through 2023 (Harper, 2024). The relative isolation of the West Jetty structure and the similarities in habitat preference between these species may have concentrated small juvenile hawksbills and goliath grouper at this location. Though no direct evidence is available, the potential for increased predation pressure on small turtles by large grouper may have partly contributed to the decline of hawksbills encountered at this site.

Growth rate

The growth rates reported in this study gradually decreased with increased carapace length (Figure 3), and are comparable to those reported near Key West by Gorham et al. (2014). Combined with healthy body condition indices (BCI) among this aggregation (mean 1.15, range 1.03-1.39) (Stacy et al., 2023), our results suggest that prey availability on the barrier reefs of the FKNMS is sufficient to fuel growth rates among juvenile hawksbills that generally align with regional estimates of time-to-maturity trajectories. Recapture data from captive-reared juvenile hawksbills that were released at reef sites near Key West concurrent with this study period showed similar rates of growth to the wild-caught turtles presented here, indicating that FKNMS reef habitats are also able to support the reintroduction of head-started, rehabilitated hawksbills (Boyd and Wood, 2025).

Based on a combination of growth models and trace element data, Avens et al. (2021) estimated the ages of juvenile hawksbills ranging from 17.6–35 cm to be between 1.4–4 years old and concluded that most recruit to neritic habitats by three years of age. The total time spent in the epipelagic environment by any individual could be strongly influenced by its position within the oceanic current system in which it became entrained, and the distance traveled, speed, and random variabilities in current systems may dictate how long it takes to locate suitable nearshore environments. Nonetheless, at the growth rates observed here, a small juvenile (~25 cm SCL) arriving in S. Florida likely reaches the upper limit of the size-range typically observed on the shallow reefs in FKNMS (~50 cm) within four years (Figure 4). Hypothetically, if it were to remain in Florida waters and grow at the combined rates observed in juveniles and sub-adults in the Keys and Palm Beach County (Wood et al., 2013, Wood, unpublished data), it would likely attain the estimated minimum size at sexual maturity (69.5 - 78.8cm, see Avens et al., 2021) in approximately 15–20 years, predicted from the growth model.

Population structure

The hawksbills in the FKNMS have proven to represent a mix of genetically distinct Caribbean and Gulf of Mexico sub-populations that is prone to change over time. Consistent with the results presented here, prior studies by Wood et al. (2013) and Gorham et al. (2014) also identified multiple source rookeries within hawksbill populations residing in both Palm Beach County and Key West, respectively. In all cases, haplotypes known to be associated with rookeries in the Yucatan Peninsula predominated in the samples, yet significant differences in haplotype frequencies between the Key West and Palm Beach aggregations (particularly the relative lack of EiA01 in the Palm Beach cohort) led Gorham et al. (2014) to suggest differential recruitment may have occurred between the two sites, which are separated by over 350 km. However, our updated results from sites from Key West through Key Largo detected increased diversity over time, which coupled with the decreased contributions from Mexican rookeries, led to higher similarities between the relative haplotype frequencies between the FKNMS and Palm Beach aggregations (Table 1). Currently, there is no known association between any particular source rookeries and the eventual patterns of distribution along Florida’s SE coast.

The relative decrease in the proportion of Yucatan representatives (EIA23, EiA39, EiA41) may be partly the result of the lag effect from low nesting densities recorded from 2005–2010 on several nesting beaches in the Peninsula (Lopez-Castro et al., 2022). It is also noteworthy that the Deep Water Horizon oil spill occurred in the Gulf of Mexico in 2010; that event may have disproportionately reduced survivorship among neonates coincidentally occupying the Gulf during this period of their early development (Putnam et al., 2014; Wallace et al., 2017; Frasier et al., 2020). Nonetheless, the continued predominance of Mexican natal haplotypes in our samples is likely a consequence of hatchlings becoming entrained in the loop current and its various permutations, which would eventually provide some of them access to the FKNMS as the current exits the Gulf of Mexico through the Florida Straits. Likewise, the westward flow of the Caribbean Current likely transports a mixture of hatchlings from various distant Caribbean populations (Stahelin et al., 2022). For those on the verge of recruitment, South Florida represents among the last opportunities to exit these current systems before being drawn into the North Atlantic by the Gulf Stream, particularly for those arriving from the Caribbean. Although these mechanisms are thought to aid in the broad dispersal of juveniles, they may also displace some individuals considerable distances from their source rookeries to which they, given their highly-developed natal homing kills, will presumably have to return as young adults (Lavasseur et al., 2019; 2020).

Ultimately, the shallow reefs of the FKNMS serve as important ecological gateways that will strongly influence the capacity for Florida to sustain future hawksbill populations, and by extension continue to contribute to the regional recovery of this species. Once established, the propensity for immature hawksbills to temporarily occupy and potentially defend fairly small areas of core habitat use (van Dam and Diez, 1998a; Cuevas et al., 2007; Wood et al., 2017) may influence patterns of settlement, underscoring the need to maintain and protect sufficient resource availability on the reefs of FKNMS for as many new arrivals as possible. In turn, the positive effects of spongivory by hawksbill turtles facilitate a positive feedback loop by helping maintain species diversity within the reef community (Hill, 1998; Leon and Bjorndal, 2002). The results presented here provide only a recent snapshot concerning hawksbill abundance and population structure within the Sanctuary, but still highlight the relevance of SE Florida to the regional recovery of this species and provide valuable baseline data for future studies.