Keewaydin Island and Sanibel Island are coastal barrier islands located on the lower southwest coast of Florida (Figure 1; Sanibel – N 26.46481°, W 81.16714°, Keewaydin – N 26.04844°, and W 81.78139°). Sanibel Island is 21 km in length with a land area of 4,429 ha while the more linear Keewaydin Island is 12.5 km long and encompasses 526 ha. Eighty-five percent of Keewaydin is state-owned and is within Rookery Bay National Estuarine Research Reserve. Keewaydin is a relatively pristine island that can only be visited by boat, while the coastline of Sanibel is much more developed and readily accessible. The straight-line distance between the two sites is approximately 30 km.

Nighttime tagging surveys run from May 1 to July 31 on Sanibel Island and May 15 to July 31 on Keewaydin Island. Staff patrol these two beaches on ATVs from sundown to sunrise seven nights per week. We captured turtles during night patrols by temporarily corralling them after nesting in an adjustable plywood box. We tagged the trailing edge of front flippers with 681/C Inconel self-piercing sea turtle tags (National Band and Tag Company, Newport, KY, United States) and inserted subcutaneous passive integrated transponder (PIT) tags (Biomark FDX-B HPT12 PIT) in the right front flipper (NMFS-SEFSC, 2008). We recorded morphometric data and took biopsies for genetics and stable isotope analysis for each individual. We used Sirtrack K2G 575E KiwiSat 202B Argos platform terminal transmitters (PTTs) that were adhered according to Mansfield et al. (2009). We cleaned the carapace, prepped it with a drying agent and attached the transmitter using a two-component adhesive (DeWalt Pure50+™) followed by a two-part steel reinforced epoxy putty (Sonic Weld ^®). Additionally, we coated the transmitter and surrounding epoxy with antifouling paint. We released the turtles on site immediately after all attachment methods had dried or set. Sirtrack Argos PTTs were programmed by the manufacturer with a 24-h on/24-h off duty cycle to conserve battery life in 2017 and were continuously on in 2018 and 2019.

We used switching state space modeling (SSM) to characterize the movements of tagged turtles. The model was previously described in Jonsen et al. (2005) and has been applied to model movement of many marine animals (Bailey et al., 2009; Shillinger et al., 2010; Maxwell et al., 2011; Shaver et al., 2016; Iverson et al., 2020). Specifically, we applied a Bayesian hierarchical movement model with behavioral-state switching in the R package “bsam” (Jonsen et al., 2005, 2017; Jonsen, 2016; R Core Team, 2020) by calling JAGS (package rjags; Plummer, 2019) to run the Markov Chain Monte Carlo (MCMC) algorithm. The model was fit with the function “fit_ssm(),” using the “hDCRWS” specification (hierarchical first difference correlated random walk switching model); we applied a time step of 1 day. The MCMC parameters were set following Hart et al. (2020), using adaptive sampling for 7,000 draws, taking 10,000 samples from the posterior distribution, and thinning by 5 to reduce MCMC autocorrelation, resulting in 2,000 posterior samples for inference. The SSM estimates location and behavioral mode at regular time intervals, improving accuracy of tag data by accounting for positional errors and dynamics of animal movements. The SSM assigned either area restricted movement (ARS) or directed movement to a location based on the swimming speed and turning angle of the animal. The ARS movements corresponded to either inter-nesting or foraging behavior, and directed movement (transiting) corresponded to migration. Switching SSMs utilize raw track data to determine behavioral state in a bimodal context by providing a behavioral index between 1 and 2, referred to as a “b” value. Mode 1 (e.g., “migrating”) is represented by b values <1.5 and mode 2 (e.g., “resident/foraging”) is represented by b values >1.5 (Jonsen et al., 2005; Breed et al., 2009). We used the model to classify ARS and migratory behavior, determine the dates of migration, and identify the date of arrival at foraging grounds (i.e., when ARS behavior resumed post-migration).

To assess inter-nesting movements, we determined the mean distance traveled per day during inter-nesting for each turtle by calculating the distance traveled per day from the SSM points using the geosphere package (successive distance between points from 1 day to the next; Hijmans, 2019) in R (R Core Team, 2020), and then calculated the mean distance traveled per day for each turtle over their inter-nesting period. We calculated the inter-nesting interval (the number of days between nesting events) using recapture data when available, as telemetry locations that appeared on the beaches had high estimated error associated with points (i.e., spatial data were unreliable). Means were accompanied by ±SD.

We used the Optimized Hot Spot Analysis in ArcGIS Pro 2.4.3 to identify areas of high use during inter-nesting and migration (Environmental Systems Research Institute [ESRI], 2019). The program uses inter-nesting and migratory data from the SSM and the Getis-Ord Gi* statistic to generate a fishnet grid of 18 km squares, assigning them a color based on statistical significance (Getis and Ord, 1992). We calculated bathymetry from the “hotspot” grid cells comprising the migratory corridor where higher proportions migrated. We layered the NOAA National Geophysical Data Center (2001) United States Coastal Relief Model Vol.3 – Florida and East Gulf of Mexico in ArcGIS Pro to estimate depth associated with each migration grid cell (CRM, NOAA National Geophysical Data Center, 2001; Environmental Systems Research Institute [ESRI], 2019).

Post-migration ARS points were considered “foraging” and we used the first date of these ARS points to delineate the turtle’s date of arrival to the foraging area. We corroborated these dates using cumulative distance plots (not shown Hart et al., 2021). To determine core use areas upon arrival at foraging grounds, SSM locations for each turtle were used for kernel density estimation (KDE) analysis for all individuals. Kernel density is a tool used in spatial ecology to identify areas of heavy use (i.e., core areas) within a home range boundary. We used the (the adehabitatHR package via Home Range App created by Cyril Bernard, SIE, CEFE-CNRS) for each KDE. We used 95% KDEs to represent the home foraging range and the 50% KDEs to represent core areas of activity at foraging sites.

Eight turtles had clear migrations identified by SSM (Figure 2), providing a well-defined end date for their nesting season and allowing us to characterize inter-nesting movements prior to initiation of migration. Turtle J was instrumented at her last nesting event of the season, which precluded any inter-nesting observations. Turtle A did not have a migration identified by SSM and therefore an end date for her nesting season was not defined, prohibiting the calculation of distance traveled per day. However, she clearly remained close to the Sanibel nesting beach throughout the tracking period. With the exception of this individual, all turtles were highly mobile between nesting events, exhibiting mean track lengths between 11 and 30 km/day during the nesting season (Table 2). For the eight turtles that had migrations identified by SSM, the mean distance traveled per day between the tagging date and the start of migration (i.e., the nesting season) was 21 km/day. Turtle C traveled the longest distance during the nesting season, logging 1,489 km between the tagging date and initiation of migration (June 16, 2017–August 22, 2017; Figure 2A). The mean water depth associated with inter-nesting movements was 6.3 m.

Six of the 14 turtles traveled between the nesting beach and a distinct in-water location ∼30 km west of Cape Sable, southwestern tip of Florida, between consecutive nesting events (Figure 3). These individuals traveled as far as 300 km roundtrip and spent 2–6 days at the site before returning to the nesting beach. Three of these tracks were documented in 2017, one in 2018, and two in 2019. Turtle E and Turtle G were satellite-tagged twice but only exhibited this behavior during one of the 2 years.

Logistical limitations prevent a true saturation tagging project (in which every turtle is observed and tagged) at each of the study sites. As such, the satellite-tracked turtles in this study were not observed during every nesting event on Sanibel and Keewaydin Islands. However, on the 10 occasions when a satellite tagged turtle was seen during two consecutive nesting events, the inter-nesting interval ranged from 9 to 14 days (mean = 11 days; n = 5 turtles).

Eight turtles had migration periods (Table 2). Two turtles (Turtle G in 2017 and Turtle F in 2019) had well defined foraging grounds but SSM did not delineate a clear migratory phase. Each turtle traveled south after their presumed final nesting event. A nearshore migratory hotspot was identified along the coast of southwest Florida (Figure 4). Initiation of migration to foraging areas ranged from July 6 to August 28. Turtle G and Turtle J were the only individuals that departed for their foraging grounds in July. For the other turtles, migration start dates ranged from August 5 to August 28. The average time spent migrating to the foraging grounds was short, at 4 ± 5.7 days (range 1–18). Migratory movements identified through SSM took place in shallow waters (mean depth was 14.6, 8.3, and 8.0 m in 2017, 2018, and 2019, respectively) off the southwest coast of Florida, with the exception of two turtles: one who had a migratory phase to and from the Dry Tortugas prior to settling in Florida Bay, and another who traveled beside the continental shelf while migrating to Florida’s east coast as far north as the Patrick Space Force Base in Brevard County.

Of the four turtles that were tagged twice, there were no instances when migration data were identified in both seasons. Turtle F and Turtle E were not sending transmissions at the likely time of migration in 2019 and did not transmit long enough to determine foraging areas. Turtle G did not have migratory points defined by SSM in 2017. Turtle A did not migrate away from the coast of Sanibel either year (2017 or 2019) during the migratory period defined in this study and could be a resident turtle that forages and nests in the same area, but the dataset acquired during the presumed foraging period was too limited to make this assumption.

We tracked turtles to foraging grounds in the Marquesas Keys (n = 4), western Florida Bay (n = 4), and an area offshore Cape Sable (n = 2; Table 3 and Figures 5, 6). Turtles were tracked at resident foraging sites for 333–345 days (mean = 94 ± 90.5 days). The size of the 50% core use resident areas for each individual ranged from 8 to 904 km² (mean = 296 ± 309.3 km²; Figures 5, 7). In every case the foraging period ended when the PTT stopped transmitting rather than the turtle leaving the location, and thus is not a true reflection of time spent at the foraging grounds. Nonetheless, the transmitter for Turtle F was sending high quality data until August 12 of the following nesting season.

Four individuals (Turtle D, Turtle E, Turtle I, and Turtle J) used a previously unidentified foraging hotspot in the western Florida Bay region (see Figure 6). The centroid locations for core use areas were very close to the shoreline for turtles foraging in the Marquesas Keys (mean = 2.3, range 1.4–3.0 km) compared to turtles foraging in western Florida Bay (mean = 21.4 km, range 11.3–23.7 km). The mean water depth at centroids in these two sites was comparable (4.2 and 4.1, respectively). The distance to shore for Turtle G that used the Cape Sable residence area in two separate years was 25 km, considerably farther than the other two sites, and the water depth at the foraging centroid was slightly deeper at 5.5 m.

For the foraging centroids, the mean distance to the closest foraging centroid was 3.1 km (range 1.1–5.5 km) for turtles foraging in the Marquesas Keys and was 19.6 km (range 4.2–36.5 km) for turtles foraging in western Florida Bay. Turtle G and Turtle F were tracked to their foraging grounds twice, and both turtles exhibited fidelity to the same foraging areas, one off Cape Sable and the other at Marquesas Keys. Turtle G’s centroid locations were 4.7 km apart and Turtle F’s centroid locations were 1.6 km apart (Figure 6). The other two turtles that were tracked twice (Turtle A and Turtle E) did not have defined foraging grounds for both tracking years and thus we could not assess the extent of their interannual site fidelity.

State space modeling model output did not produce discrete inter-nesting, migration, and foraging phases; instead, there were numerous ARS and migration points intermingled throughout the entire tracking period. It is likely that these turtles were foraging to some extent during the nesting and migration periods. While it was previously believed that turtles do not forage while nesting (Limpus, 1973; Carr, 1975), there is mounting evidence that foraging occurs in the inter-nesting period when forage is available. Dive data from nesting green turtles near Cyprus suggests that foraging behavior may occur between nesting events (Hochscheid et al., 1999) and studies analyzing gut contents have confirmed that opportunistic foraging occurs at nesting sites (Tucker and Read, 2001; Hays et al., 2002). These inter-nesting foraging sites are typically located in close proximity to the nesting beach, as nesting turtles generally limit their movement in order to minimize energy expenditure (Cheng, 2009). However, it appears that the six turtles in our study which made the 300 km trip to Cape Sable area may be exhibiting intentional food-seeking behavior rather than opportunistic feeding. Future studies are needed to determine why turtles are using this site in between nesting events.

We found considerable inter-annual variability in the relative importance of the Cape Sable area. The Cape Sable area was used all 3 years, but by different individuals each year, and never by the same animal twice. Seagrass beds are prevalent in this area and it is likely that these turtles used this site for feeding. Environmental conditions, including ocean currents, can influence turtle movements, but the use of this site each season suggests local conditions are favorable for these adult green turtles. The distribution and abundance of seagrasses in south Florida vary annually (Fourqurean et al., 2001), and the quality of the forage resources at this location may influence energy reserves available for turtles in the next non-breeding seasons.

Inter-nesting intervals lie within the general range reported for green turtles at other sites, with 9–18 day intervals between nests (Carr et al., 1974; Mortimer and Portier, 1989; Broderick et al., 2002; Van de Merwe et al., 2009; Hart et al., 2013). Cooler water temperatures can lengthen the number of days between nesting events for green turtles (Hays et al., 2002), but this phenomenon does not apply to the turtles in this study due to the consistently warm water temperatures in the Gulf of Mexico during the nesting season (Turner et al., 2017). Transiting behavior between nesting events was highly variable within and among individuals, but collectively the turtles exhibited more nomadic behavior than previously reported for nesting green turtles. Studies on inter-nesting behavior in green turtles suggest that turtles generally remain close to their nesting beach (Mortimer and Portier, 1989; Hays et al., 1999; Troeng et al., 2005; Blanco et al., 2013; Hart et al., 2013, 2017; Esteban et al., 2015; Chambault et al., 2016), although individual exceptions have been noted (i.e., Esteban et al., 2015; Chambault et al., 2016; Mettler et al., 2020). Overall, the individuals in our study had greater displacement in between nests than has been previously reported for green turtles, but they all used relatively shallow Gulf waters (mean 6.3 m) in between nesting events.

The mean migration distance using a compilation of tracks from adult green turtles worldwide was estimated at 806 ± 602 km (Hays and Scott, 2013). However, highly variable migration distances have been recorded for green turtles both among study sites and also within individuals from the same study site. The variation in migration distances for individual turtles nesting on a single beach can be over 1,000 km (Luschi et al., 1998; Godley et al., 2002; Hays et al., 2020). On a global scale, the range of distances reported for migration from nesting beach to foraging grounds can be less than 100 km (Whiting et al., 2007; Hart et al., 2017), while distances up to 4,619 km have been recorded in the Indian Ocean (Hays et al., 2020) and 5,278 km along the coast of French Guiana, Suriname, and Brazil (Baudouin et al., 2015).

Our findings are consistent with migration data for turtles nesting within the Gulf of Mexico at Dry Tortugas (5–282 km; Hart et al., 2013) and Buck Island (0.2–694 km; Hart et al., 2017), and also in the Indian Ocean (33–37 km; Whiting et al., 2007). Turtles in these studies traveled less than 200 km to shallow foraging areas after the breeding seasons. It has been hypothesized that these variable migration patterns may reflect the oceanic drift that individuals experienced in their early years as hatchlings or young turtles (Scott et al., 2014). The hatchling drift scenarios for Keewaydin, Sanibel, and Dry Tortugas are similar due to the relatively close proximity of the three sites within the Gulf of Mexico, and post-hatchling turtles from these beaches would likely encounter the same locations along their routes, possibly accounting for the similarity in consistent migration behavior among individuals at the three sites.

The mean migration depth in 2017 was deeper (14.6 m) compared to 2018 and 2019 (8.3 and 8.0 m, respectively). Turtle E migrated southbound along the east coast of Florida in 2017, using much deeper waters in the Atlantic Ocean. Migrations within the Gulf of Mexico were shallow and generally close to the shoreline. Some turtles in this study briefly visited the Cape Sable site along their migration route, and others remained in the area for longer periods of time before initiating migration. Foraging during migration has been reported in sea turtles (Baudouin et al., 2015; Shimada et al., 2020) and these animals may be maintaining foraging site fidelity to the Cape Sable area while en route to their final destinations.

The geographic locations of the three hotspots (Cape Sable, Florida Bay, and Marquesas Keys) determined by the model provided novel information on the foraging grounds for green turtles that nest on the west coast of Florida. Four turtles used the Marquesas Keys as foraging grounds. Turtle F returned to the same location in both 2017 and 2019 and remained there for 345 days in 2017, implying strong site fidelity to this area. High levels of fidelity to foraging grounds is common for both adult and juvenile green turtles (Broderick et al., 2007; Siegwalt et al., 2020).

Unpublished satellite telemetry data indicate that turtles nesting on the east coast also use the Marquesas Key as a foraging site (Bagley et al., 2021, Department of Biology, University of Central Florida, written communication, August 19, 2021). The Eastern Quicksands area is adjacent to the Marquesas Keys and provides important foraging grounds for green turtles that nest along the central Atlantic coast of Florida. Tag returns revealed that these turtles also nest in Quintana Roo, Mexico and Tortuguero, Costa Rica (Bresette et al., 2010), demonstrating that females from multiple nesting populations are using the area as their foraging grounds. Aggregations of foraging adults from multiple nesting beaches have also been reported in other areas such as Dry Tortugas National Park (Naro-Maciel et al., 2017), the Great Barrier Reef (Limpus et al., 2009; Jones et al., 2018), and Nicaragua (Bass et al., 1998). The source rookeries for the green turtles nesting on the west coast of Florida are unknown; genetic analyses of samples collected from nesting females will be required to assess relatedness to Florida east coast turtles.

Four turtles in this study used areas in western Florida Bay area as foraging grounds. The two turtles that foraged closer to shore (Turtle I and Turtle J) maintained much smaller core use areas compared to the turtles that used slightly deeper water into the Gulf of Mexico (Turtle D and Turtle E). This behavior may be a function of food availability, with higher density forage potentially located in shallower waters. One turtle tracked from Dry Tortugas used the Florida Bay area (Hart et al., 2013), and several turtles from an Atlantic coast nesting beach also foraged at this location during the non-nesting season (Bagley et al., 2021, Department of Biology, University of Central Florida, written communication, August 19, 2021), suggesting there could be individuals from multiple rookeries mixing at this foraging ground.

Foraging areas for the turtles in this study were located in relatively shallow habitats (2–6 m) that are likely to support seagrass beds. This is consistent with previous studies on green turtles nesting in Florida. The Eastern Quicksands was identified as a high-density adult green turtle foraging area by the Inwater Research Group (Herren et al., 2018). The depth of the waters at this location (3.1–6.0 m) are similar to those reported in this study and contain a bottom of shifting sand areas interspersed with seagrass beds of Thalassia testudinum and Syringodium filiforme, supporting a density of 58.5 green turtles/km². Seagrasses are widespread and conspicuous in the marine waters surrounding the southern tip of the Florida peninsula (Schomer and Drew, 1982; Fourqurean et al., 2001; Carlson and Fourqurean, 2016). Future work to determine if the Florida Bay and Cape Sable sites identified in this study contain similar foraging habitat would be worthwhile.

In addition to its importance in the nesting season, the Cape Sable site is a previously unidentified foraging hotspot and core use area. One turtle tracked from Dry Tortugas also used the same areas after nesting (Hart et al., 2013), putting this location on the map, but it was not characterized as a true hotspot prior to this study. A comprehensive study that mapped seagrasses in South Florida revealed exceptionally high biomass of T. testudinum at this exact location (Fourqurean et al., 2001), but more research is needed to confirm current benthic cover. Juvenile green turtles have been reported near this area (Hart and Fujisaki, 2010), and adult turtles may remain faithful to this site as a known high-quality foraging habitat. Future work at this site that includes a combination of underwater video (e.g., Hart et al., 2010), direct observations, and benthic sampling would be valuable.

The results of this study are particularly relevant from a conservation standpoint because of the noteworthy contribution of these foraging grounds to Florida’s green turtle population. The high concentration of green turtles foraging in the neritic habitat in the Florida Keys and Florida Bay region indicates that conservation efforts focused in this region could be beneficial for these imperiled turtle populations. The additional knowledge provided by this study about the convergence of sea turtles found at these important at-sea sites can inform resource management strategies intended to protect turtles nesting in many different areas. Concerted management and restoration plans focused on protecting the turtles in these areas from threats posed by fisheries and watercraft interactions could augment plans in place that protect them from nesting habitat degradation.