Resident areas and migrations of female green turtles nesting at Buck Island Reef National Monument, St. Croix, US Virgin Islands

Satellite tracking studies can reveal much about turtles’ spatial use of breeding areas, migration zones, and foraging sites. We assessed spatial habitat-use patterns of 10 adult female green turtles Chelonia mydas nesting at Buck Island Reef National Monument (BIRNM), US Virgin Islands (USVI; 17° 47.4’ N, 64° 37.2’ W) from 2011 to 2014. Turtles ranged in size from 89.0 to 115.9 cm curved carapace length (CCL) (x− ± SD: 106.8 ± 7.7 cm). The inter-nesting period for all turtles ranged from 31 July to 4 November, and sizes of the 50% core-use areas ranged from 4.2 to 19.0 km2. We observed consistency of inter-nesting habitat-use patterns, with all turtles using near-shore (<1.5 km), shallow waters (<–20 m depth) within approximately 10 km of Buck Island. Seven of the 10 turtles remained locally resident after the nesting season; 5 turtles (50%) established resident foraging areas around Buck Island, 2 established resident foraging areas around the island of St. Croix, and the other 3 (30%) made longer-distance migrations to Antigua, St. Kitts & Nevis, and Venezuela. This is the first empirical dataset showing limited migration and use of ‘local’ resources after the nesting season in the USVI by this unique management unit of green turtles. Five of the turtles had resident foraging area centroids within protected areas; thus, inter-nesting and foraging areas at BIRNM that overlap with human use zones present an important management concern. Delineating spatial areas and identifying temporal periods of nearshore habitat use can be useful for natural resource managers with responsibility for overseeing vulnerable habitats and protecting marine turtle populations.

. Coupled with recent advances in analytical modeling techniques such as switching state-space modeling (SSM; see Jonsen et al. 2003, 2005, 2006, 2007, Patterson et al. 2008, Hoenner et al. 2012), tracking data can provide an unprecedented window into behavioral modes of marine turtles (i.e.directed movement or migration, area restricted search or foraging; see Bailey et al. 2008, 2009, Hart et al. 2015).
Listed as Endangered by the International Union for the Conservation of Nature (Groombridge 1982, Groombridge & Luxmoore 1989, Seminoff 2004), and threatened under the US Endangered Species Act, Caribbean green turtles are part of the North Atlantic distinct population segment (DPS), yet little is known about the species in the US Virgin Islands (USVI).Analysis of mitogenomic haplotype frequencies from rookeries within the southern Greater Caribbean region indicates that green turtles at Buck Island Reef National Monument (BIRNM) in St. Croix warrant recognition as a distinct management unit that is part of the larger USVI stock (Shamblin et al. 2012).
Much reduced from historic numbers (Jackson 1997, McClenachan et al. 2006), green turtle population numbers in Caribbean rookeries are low, but recovering.Three separate long-term saturation nesting programs in St. Croix (The Nature Conservancy [TNC] at East End Marine Park [EEMP], National Park Service at BIRNM, and US Fish and Wildlife Service [USFWS] at Sandy Point) have documented increases in nesting green turtle numbers since the 1990s, but annual numbers of nesting individuals at each of these 3 main sites are only in the 20s (e.g. 26 in 2015 at BIRNM; C. Pollock pers.obs.).An increase in seagrass distribution at BIRNM (Ken dall et al. 2004a,b), federal protection for green turtles in the 1970s (NMFS & USFWS 1991), a decrease in poaching, and the removal of mongoose from BIRNM nesting beaches in the 1980s all provided in creased habitat security for green turtles at BIRNM.By 1995, the number of nesting green turtles at BIRNM had increased, followed by increased numbers at East End beaches (K.Lewis [TNC] pers.comm.).
Currently, critical habitat for green turtles is delineated in waters surrounding Culebra Island, Puerto Rico (NOAA 1998), and there is a pending proposal to list 11 DPSs of green turtles as endangered or threatened, with a revision of current listings (NMFS & USFWS 2015, March Federal Register).Understanding the specific movement ecology for different DPSs can help inform conservation efforts targeted to wards those population segments.Further, understanding green turtle movement patterns in protected areas is considered a priority for ongoing conservation efforts and Federal recovery plans (NMFS & USFWS 1991, Hart et al. 2013).However, green turtle spatial habitat use within BIRNM waters has not been assessed.
To assess spatial habitat-use patterns of green turtles nesting at BIRNM, our goals were to delineate zones used during inter-nesting periods, define any migration paths after the nesting season, and identify foraging sites where turtles remain resident.Where possible, we quantified spatial overlap of habitat use for individuals at inter-nesting areas and foraging sites.To glean possible reasons behind spatial selection of habitats and whether resident areas were selected by turtles of a specific size or experience level, we quantified ecological and spatial correlates such as the relationship between turtle size and bathymetry values at inter-nesting areas, characterized the habitats associated with foraging areas, and determined whether habitat overlap varied depending on nesting 'experience' (i.e.'neophyte' for first time nester vs. 're-migrant' status).Finally, we discuss green turtle habitat use in relation to nearby marine protected areas (MPAs).

Study sites
Sampling and tagging of green turtles occurred from 15 July to 30 September annually from 2011 to 2014 in the USVI at BIRNM (17°47.4'N, 64°37.2'W), which includes a 0.71 km 2 uninhabited island (Buck Island) located on the shallow St. Croix shelf (depth range approx.-15 to -20 m), 2.4 km northeast of the island of St. Croix (see Fig. 1).Buck Island is 1.82 km long and 0.8 km wide, and rises 103 m above sea level at the highest peak.With the expansion of the monument boundaries in 2001, the amount of submerged lands around Buck Island now covers 76.3 km 2 .BIRNM is a nesting and foraging area for loggerhead Caretta caretta, leatherback Dermochelys coriacea, hawksbill Eretmochelys imbricata, and green sea turtles.

Turtle capture and transmitter deployment
Nightly surveys were conducted from 19:00 to 05:00 h local time.Turtle interception and tagging followed methods similar to those of Hart et al. (2013); turtles were documented and fitted with transmitters using established protocols (NMFS-SEFSC 2008).Briefly, female green turtles were intercepted on the beach after nesting.Immediately after marking each turtle with Inconel and PIT tags, standard carapace measurements were recorded, in cluding curved carapace length (CCL).We adhered platform transmitter terminals (PTTs; Wildlife Computers SPOT5s; length × width × height: 7.2 × 5.4 × 2.4 cm, mass: 119 g in air) using slow-curing epoxy.Attachment materials were streamlined to minimize the epoxy footprint and potential buoyancy and/or drag effects on turtle swimming ability.Each tag was programmed to transmit 24 h d −1 .

Sea turtle tracking and switching state-space modeling
Location data were retrieved using Satellite Tracking and Analysis Tool (STAT; Coyne & Godley 2005) available at www.seaturtle.org.Location classes (LCs) 3, 2, 1, 0, A, and B were used to assess the position of the turtles during the tracking period, and served as the location data to characterize inter-nesting, foraging, and migration behavior of each turtle.We used switching state-space modeling (SSM; Jonsen et al. 2003, Patterson et al. 2008) as described in Jonsen et al. (2005) to assess the fine-scale behavioral modes of individual green sea turtle tracks that originally nested in BIRNM.Switching SSM techniques followed our previous studies (see Hart et al. 2013, 2014, 2015, Shaver et al. 2013, 2016 and the Supplement at www. int-res.com/ articles/ suppl/ n032 p089 _ supp.pdf for general information on this technique).Earlier applications defined a binary behavioral mode, categorized as either 'foraging' or 'migration' (Jonsen et al. 2005, 2007, Breed et al. 2009); however, since we tagged animals during the nesting season, behavioral modes predicted by the SSM algorithm were defined as either 'inter-nesting or foraging' or 'migration'.
We summarized data acquired until the transmitters stopped sending information or until the time of data synthesis: 13 November 2014.We used the SSM approach to determine the beginning and end date of the inter-nesting, migration, and foraging behavioral periods for each turtle.After assigning the beginning and end dates of each behavioral mode for each turtle track, we used the original satellite locations within those time periods for all further analyses.

Inter-nesting and foraging periods
We verified SSM behavioral modes both spatially and temporally using the satellite data.Inter-nesting periods occurred before migration to foraging grounds.Some turtles exhibited local foraging behavior around Buck Island, and also did not show a distinct migration period.Nesting activity data for this species on Buck Island indicates that the end of October is the end of the inter-nesting season, based on the last beach encounters for 9 nesting green sea turtles from 1995 to 2013.Therefore, a temporal cutoff of 31 October was applied to differentiate between inter-nesting and foraging behavior for the turtles that did not migrate.From the satellite data during the inter-nesting and foraging periods, we filtered out locations that represented movement speeds > 5 km h −1 , locations on land, and very distant spatial locations (>120 km from the nearest valid point).We also filtered out points associated with ocean water depths deeper than -200 m (neritic zone delineation).Green turtles have been found to stay in shallow waters during residency times (see Meylan 1995), remaining in mean depths shallower than -10 m in the Mediterranean (Hays et al. 2002) and the Gulf of Mexico (Hart et al. 2013), and shallower than -25 m in the Caribbean (Esteban et al. 2015).For all tracks in this study, the removed locations deeper than -200 m comprised less than 7.0% of available speed-filtered loca tions.For bathymetry, we used the ETOPO1 global relief model (bedrock, cell-registered, 1 arc-minute; Amante & Eakins 2009).

Characterization of inter-nesting and foraging areas
After assigning the tracking data to inter-nesting and foraging behavioral modes, we quantified internesting and foraging home ranges using kernel density estimation (KDE) and minimum convex polygon (MCP) analyses.To minimize autocorrelation of points, we generated mean daily locations (MDL) within each inter-nesting and foraging period in the software program R version 3.1.2(R Development Core Team 2014) from the filtered satellite locations.We used MDLs (when n ≥ 20) for KDE analysis and filtered satellite locations (when MDLs < 20) for 95% MCPs.
Kernel density is a non-parametric method used to identify one or more areas of disproportionately heavy use (i.e.core areas) within a home-range boundary (Worton 1987, 1989, White & Garrott 1990), with appropriate weighting of outlying observations.We used the software program R and the package 'adehabitatHR' (Calenge 2006) to calculate home range analyses.We applied the fixed-kernel leastsquares cross-validation smoothing factor (h cv ) for each KDE (Worton 1995, Seaman & Powell 1996).Following Walcott et al. (2012), we used 95% of points to create MCP polygons, as it is possible for a proportion of distant filtered locations to represent infrequent movements or explorations external to the home range (sensu Burt 1943, Rodgers & Kie 2011).When the standard deviations of the x and y coordinates were unequal (< 0.5 or >1.5), data were rescaled prior to home range calculations by dividing the coordinates by their standard deviation (following Seaman & Powell 1996).We used ArcGIS 10.2 (ESRI 2013) to plot the data and to calculate the area (km 2 ) within each kernel density contour (50 and 95%) and each MCP.The 95% KDEs were used to represent the overall home range, and the 50% KDEs represented the core area of activity (Hooge et al. 2001).
We tested location data for, and quantified site fidelity to, the in-water inter-nesting and foraging locations using the Animal Movement Analysis Extension for ArcView GIS 3.3 (ESRI 2002).Using Monte Carlo random walk simulations (100 replicates), we tested tracks during each inter-nesting and foraging period for spatial randomness against randomly generated walks (Hooge et al. 2001).We bounded the range for random walks from −200 to 0 m bathymetry to encompass all filtered locations, and smoothed the bounding polygon with a 250 m inland and 5 km seaward buffer to account for numerous small bays, and allow for the generation of random walks with points in close proximity to land (the 5 km seaward border was extended an additional 5 km for one turtle track: Turtle ID 7 inter-nesting).Tracks exhibiting site fidelity indicate movements that are more spatially constrained rather than randomly dispersed (Hooge et al. 2001); any tracks that failed site fidelity were not analyzed in the home range analyses.
We also calculated the centroid of each 50% KDE and MCP; if a 50% KDE included multiple activity centers, we calculated the centroid for the largest activity center.We extracted bathymetry depths for all centroids and the distance from each centroid to the nearest land.

Turtle inter-nesting and foraging days per grid cell
To depict the inter-nesting and foraging locations used by the turtles over time, we calculated the number of turtle inter-nesting days and foraging days in grid cells (2 × 2 km).For each turtle track containing days in either inter-nesting or foraging mode (regardless of whether a KDE or MCP was calculated), we determined the number of days that each turtle was recorded in each grid cell (turtle days) using all filtered satellite locations, and summed the number of turtle days across all turtles for each grid cell.We derived mean bathymetry in each 2 km grid cell and examined associations between the turtle days and bathymetry for each inter-nesting and foraging period.

Core area space-use sharing
We calculated the amount of home range overlap for the core-use areas (50% KDEs) during inter-nesting and foraging using the 'adehabitatHR' package in R (Calenge 2006, R Development Core Team 2014) to quantify the extent to which green turtles share their home ranges.The utilization distribution overlap index (UDOI) is a non-directional joint measure that is a function of the product of the 2 utilization distributions (UDs), a modification of Hurlbert's (1978) E/E uniform statistic to allow for continuous spatial UDs, and is the most appropriate statistic for measuring space-use sharing between animals.The UDOI is zero for 2 UDs without overlap, and 1 for uniformly distributed UDs with complete overlap; however, UDOI can be >1 for non-uniformly distributed UDs with a high degree of overlap, indicative of higher than normal overlap relative to uniform space use (Fieberg & Kochanny 2005).
We calculated UDOI space-use sharing for all turtles during inter-nesting (n = 10 turtles, 45 pairs), because KDEs were calculated for each turtle.Of the 5 turtles for which foraging KDEs were determined, we calculated UDOI space-use sharing for the 3 turtles that foraged in the same area around Buck Island (n = 3 pairs).Four turtles had foraging MCPs; one was located near Antigua, but for the others we calculated the distance between same-region centroids.This included 2 turtles foraging SW of St. Croix and 1 turtle foraging in the Buck Island area (we determined the mean distance from this MCP centroid to the 3 KDE centroids).We also determined the level of temporal overlap (days) that same-region turtle pairs had during foraging.For inter-nesting, we conducted a non-parametric Mann-Whitney rank sum test to determine whether there was a difference in spaceuse sharing between turtle pairs with and without neophytes in order to examine whether newly nesting turtles shared inter-nesting habitat with experienced nesters any differently than experienced pairs shared habitat.We classified any turtles that were 1 st time nesters at BIRNM as neophytes.

Ecological and spatial correlates
For the inter-nesting periods, we conducted linear regression analyses to determine whether there was a relationship between turtle size (i.e.CCL) and the size of the 50% KDE, the bathymetry values at the 50% KDE, and the bathymetry values at the MDLs; additionally, we conducted linear regression analyses to examine whether there was a relationship between the number of tracking days and the size of the 50% KDE using SigmaPlot (Systat Software 2012).For turtles foraging around Buck Island and St. Croix (n = 6), we determined the benthic habitat type located at the foraging centroids using NOAA benthic habitat maps of Puerto Rico and the USVI (Kendall et al. 2001).

Migration periods
We summarized the primary migration periods that represented movement away from the inter-nesting area to the foraging grounds; these periods occurred after inter-nesting and directly before the foraging period.After the migration periods were assigned, they were verified both spatially and temporally using the satellite location data.We filtered out satellite locations during the primary migration periods representing movement speeds > 5 km h −1 , locations on land, and very distant spatial locations (>120 km from the nearest valid point).We quantified the number of days in the primary migration period, the straight-line migration distance, the migration path distance (when applicable), and the depth along migration paths.Some turtles in our study did not exhibit a migration path (e.g.turtles that foraged locally near Buck Island), and therefore only the straight-line migration distance was quantified, calculated as the distance between inter-nesting and foraging centroids.For Turtle ID 8, the mean center of the filtered foraging satellite locations was quantified and used for the straight-line distance calculation, because this turtle did not exhibit site fidelity and therefore no centroid was calculated.

Turtle size and tracking duration
We tagged 10 adult female green turtles over a 4 yr period between 2011 and 2014 (Table 1).Turtles ranged in size from 89.0 to 115.9 cm CCL (Table 1).In a total of 1681 tracking days across all turtles, individual turtle tracking durations ranged from 100 to 372 d (Table 1).Turtles were a mix of neophytes (n = 2) and re-migrants (n = 8), with varied nesting histories since 1995 (see Table S1 in the Supplement at www. int-res.com/articles/ suppl/ n032 p089 _ supp.pdf).

In-water inter-nesting areas
We obtained SSM results for all 10 turtles (see Fig. S1, Table S2 for example SSM prediction paths and model parameters).We tagged turtles early enough in each nesting season to track them through internesting periods, and all showed site fidelity (proportion of tracks that were more constrained than random movement paths > 99.0099 for all 10 turtles; Table 1) and comprised enough MDLs (≥20) to conduct KDE analyses (Table 1).One individual (Turtle ID 8) was also observed on 13 September 2013 nesting on property that TNC monitors at the East End of St. Croix (i.e.EEMP).The 10 inter-nesting periods totaled 760 d across all turtles, and ranged from 21 to 92 d (Table 1).The inter-nesting period across all turtles (regardless of year) ranged from 31 July (x − ± SD of first inter-nesting date: 7 August ± 8.2 d, n = 10) to 4 November (x − ± SD of last inter-nesting date: 21 October ± 16.0 d, n = 10).In total, we obtained 624 MDLs for inter-nesting KDE analyses (Table 1).Mean size of the 50% core-use areas during inter-nesting was 7.3 km 2 (range: 4.2 to 19.0 km 2 ); mean size of the 95% KDE areas during inter-nesting was 35.8 km 2 (range: 20.5 to 91.3 km 2 ); and distance to the nearest land from the 50% KDE centroids (Fig. 1) ranged from 0.0 to 1.4 km (Table 1).Bathymetry values (i.e. a proxy for water depth) at the 50% KDE centroid locations ranged from −17.0 to -1.0 m (Table 1).
The grids of the turtle days showed that high-use areas during inter-nesting periods were concentrated around Buck Island (Fig. 2).There was a weak but significant association between turtle days and bathymetry in the grid cells (Spearman's rank correlation, r S = 0.34, n = 253, p < 0.0001).

Core area space-use sharing
During inter-nesting, across all pairs of internesting females, UDOI ranged from 0 to 0.18 (x − ± SD: 0.04 ± 0.05, n = 45; Table S3 in the Supplement), where greater UDOI indicates greater space-use sharing between turtle pairs.Temporal overlap across all pairs of internesting females ranged from 0 to 91 d (x − ± SD: 19.2 ± 32.5, n = 45).Results of the non-parametric Mann-Whitney rank sum test indicated there was no difference in UDOI between turtle pairs with and without neophytes (Mann-Whitney U-test, U = 182.5,n 1 = 22, n 2 = 23, p = 0.169, where n 1 = pairs with neophytes and n 2 = pairs without neophytes).

Ecological and spatial correlates
We did not find any significant relationships between turtle CCL and either the size of the 50% KDE (linear regression, r 2 = 0.01, F 1, 4 = 0.003, p = 0.87) or the bathymetry values at the 50% KDE centroid (linear regression, r 2 = 0.01, F 1, 8 = 0.11, p = 0.75), or between the number of tracking days and the size of the 50% KDE (linear regression, r 2 = 0.04, F 1, 8 = 0.30, p = 0.60).Thus, turtle size was not predictive of inter-nesting habitat selection.Distances to the nearest land from the 50% KDE foraging centroids (Fig. 3) ranged from 0.7 to 1.6 km, and bathymetry values at the 50% KDE centroid locations ranged from −23.0 to −1.0 m (Table 2).Distances to the nearest land from the MCP foraging centroids ranged from 0.5 to 1.6 km, and the bathymetry values that corresponded to the foraging MCP centroids ranged from −13.0 to -6.0 m (Table 2).

Turtle foraging days per grid cell
In addition to the area around Buck Island, where we observed high use by turtles during the internesting period, turtles were frequently located off the SW coast of St. Croix (Fig. 2).The number of turtle days per grid cell was weakly but significantly associated with bathymetry (Spearman's ρ = 0.31, p < 0.0001).Core area space-use sharing There were 5 turtles for which foraging KDEs were calculated, 3 of which foraged in the same region (near Buck Island).These 3 turtle pairs were all re-migrants, and the amount of UDOI space-use sharing for these turtle pairs (n = 3 turtles, n = 3 pairs) ranged from 0.09 to 0.14 (x − ± SD: 0.12 ± 0.02, n = 3; Table S3), where greater UDOI indicates greater space-use sharing between turtle pairs.Temporal overlap for these 3 pairs of foraging females near Buck Island ranged from 0 to 138 d

Ecological and spatial correlates
The benthic habitats associated with foraging centroids for turtles foraging near Buck Island and St. Croix consisted of submerged seagrass vegetation, coral reef and colonized hardbottom, and unconsolidated sand sediments (Table 2).Of the 9 turtles that had a foraging KDE or MCP (n = 5 KDEs, n = 4 MCPs), 5 turtles had foraging centroids located in protected or managed areas, including EEMP, USVI and Los Roques National Park (LRNP), Venezuela.The other 4 turtles had foraging centroids that were not located in protected areas (Table 2).

DISCUSSION
Through 4 yr of satellite tracking, we gained insight into the movement and habitat-use patterns of adult female green turtles nesting at BIRNM, which represents a unique management unit (see Shamblin et al. 2012), during breeding, migration, and foraging time periods.Our study presents key information on the BIRNM green turtle rookery, supports previous findings on spatial use from 3 other Caribbean studies (Blumenthal et al. 2006, Esteban et al. 2015, Becking et al. 2016), and our robust SSM approach accurately quantifies time periods of migration and residency at foraging and inter-nesting sites.
We observed consistency of inter-nesting habitatuse patterns over the 4 study years, with all turtles using near-shore, shallow waters (shallower than -20 m depth) within approximately 10 km of Buck Island.Green turtle inter-nesting habitat in Dry Tortugas National Park (Hart et al. 2013) and other Caribbean locations was also in near-shore, shallow waters (Lesser Antilles, Esteban et al. 2015;Cayman Islands, Blumenthal et al. 2006), although some exceptions have been found for turtles traveling between nesting sites and swimming through deep waters (e.g.mean depth −2940 m for 1 Cayman Islands turtle; Blumenthal et al. 2006, see also Becking et al. 2016).
Neophyte or first-time nesters may show different patterns during inter-nesting than established remigrants.For example, experienced re-migrant hawksbills were the only turtles to select distant resident areas during inter-nesting in Barbados (Walcott et al. 2012).However, we did not find differences in internesting resident areas between green turtle neophytes and re-migrants; 2 of our tagged females (Turtle IDs 7 and 8) were neophytes, and their habitat-use patterns mirrored those of the other 8 turtles.One exception was that Turtle ID 7 showed variability in nesting beach selection and nested at a nearby TNC study site at EEMP (6.5 km straight-line distance away), whereas all other nesting events for tagged turtles occurred at BIRNM.This indicates that nesting site fidelity within a single nesting season may vary for some BIRNM green turtles.Flexibility in nesting-site use was also observed in the Cayman Islands, where 1 of 7 green turtles shifted her nesting site within a season (Blumenthal et al. 2006).Continued satellite-tracking and nest-monitoring at BIRNM and neighboring beaches could help determine if both neophytes and re-migrants share the same level of flexibility in site-use within a nesting season.
Ours is the first empirical dataset showing limited migration and use of 'local' resources in the USVI by green turtles.We found that 7 of 10 turtles did not migrate, supporting recent findings of plasticity in migration and selection of local foraging sites in Dry Tortugas National Park and the Florida Keys National Marine Sanctuary (Hart et al. 2013) where 9 of 11 turtles showed year-round residency, in the Indian Ocean (Whiting et al. 2007), where all 6 turtles migrated to shallow foraging grounds within 40 km of the nesting beach, and elsewhere in the Caribbean (Esteban et al. 2015) where 2 of 3 green turtles foraged within 50 km of their original nesting grounds.Not all turtles in our study remained near nesting grounds to forage, however; some migrated long distances (e.g. up to 694 km straight-line distance to Venezuela waters; Turtle ID 7).This was also the case for turtles studied in the Lesser Antilles, where 1 green turtle that did not forage locally migrated 607 km straight-line distance (Esteban et al. 2015) and 4 turtles studied in Bonaire migrated 198 to 3135 km away after nesting (Becking et al. 2016).In a study of green turtles in the Cayman Islands, all 7 tagged green turtles traveled 520 to 856 km straightline distance to foraging locations (Blumenthal et al. 2006).Of all these studies, neither Bonaire nor the Cayman Islands had locally foraging post-nesting green turtles despite both areas supporting resident juvenile greens (Blumenthal et al. 2006, Stapleton et al. 2014, Becking et al. 2016).Local habitat may not be particularly suitable for adults in these locations.In the Cayman Islands, legal and illegal sea turtle fisheries that mainly target larger turtles (Bell et al. 2006) could also play a role, potentially capturing resident turtles more often than those that migrate away.Fishing for turtles is not legal in US Gulf of Mexico waters, St. Eustatius and St. Maarten (Lesser Antilles) or the USVI (Bräutigam & Eckert 2006), where turtles were found to forage locally after nesting.
Similar to the inter-nesting period, foraging centroids for BIRNM nesters were also close to shore (at most 1.6 km away) and in relatively shallow waters (up to −23 m depth).This finding is similar to that of Esteban et al. (2015), where the 2 green turtles that migrated to the Lesser Antilles remained resident there in shallow nearshore foraging areas (within approximately 5 km of land).
The grid analysis revealed high-use zones for BIRNM green turtles to the northeast of St. Croix during inter-nesting, especially surrounding Buck Island.During foraging periods, high-use grid cells were to the northeast and southwest of St. Croix.Grid cells to the northeast of St. Croix lay directly in the path of commercial and recreational boats traveling to BIRNM.Inter-nesting and foraging areas at BIRNM that overlap with human use zones present an important management concern, especially because some individual adult female turtles are resident year-round in this relatively small area.The area from Green Key Marina (on the mainland of St. Croix) to Buck Island is a major human thoroughfare for boating, so boat strikes of turtles are a potential concern.Given that we found some overlap of satellite-tagged green turtle home ranges with human use areas at BIRNM, the density of turtles at these sites may be higher than we can demonstrate from satellite tracking alone.Investigation of these foraging sites to determine resource condition and to further quantify turtle numbers at these sites would be valuable for informing future management actions.
While some turtles remained resident, 3 of our turtles migrated to other countries, supporting the need for international cooperation in conservation of these turtles (e.g.Blumenthal et al. 2006, Becking et al. 2016).We tracked BIRNM nesting green turtles on their migrations to Antigua, St. Kitts & Nevis, and Venezuela, and only one of these ended up in a protected area (Venezuela's LRNP).Exploitation of sea turtles in Antigua and St. Kitts & Nevis is not completely prohibited, although it is restricted (Bräutigam & Eckert 2006).Given the genetic distinctness of BIRNM green turtles, continued long-term protection of the nesting beach and protection of the in-water inter-nesting and foraging sites delineated here may improve the likelihood of population recovery.
2013, observed by The Nature Conservancy (TNC) at East End Marine Park (EEMP), mainland St.

Fig. 2 .
Fig. 2. Grids of inter-nesting days for 10 adult female green sea turtles Chelonia mydas (top panel) and foraging-days for 7 adult female green sea turtles (bottom panel).All turtles were satellite-tagged on Buck Island, US Virgin Islands

96Fig. 3 .
Fig. 3. Foraging and migration for adult female green sea turtles Chelonia mydas satellite-tagged on Buck Island, US Virgin Islands.Kernel density estimations (KDEs) for 5 turtles are shown with hatched blue lines (95% KDE) and blue areas (50% KDE); open circles represent 50% KDE centroids.Minimum convex polygons (95% MCPs) are shown by red polygons for 4 turtles; red diamonds represent 95% MCP foraging centroids.The rectangles labelled A, B, C, in the upper left panel show the figure extents for the other 3 panels: (A) 4 turtles foraging near Buck Island (Turtle IDs 2, 4, 6, and 9) and 2 turtles (Turtle IDs 3 and 10) foraging southwest of St. Croix; the boundary for Buck Island Reef National Monument is represented by the black line; (B) 1 turtle foraging near St. Kitts & Nevis Islands (Turtle ID 1) and 1 turtle (Turtle ID 5) foraging southwest of Antigua; and (C) 1 turtle foraging near Los Roques National Park (LRNP), Venezuela (Turtle ID 7).Both KDEs and MCPs were merged for visualization purposes.Migration paths from inter-nesting grounds to foraging grounds for 3 turtles are shown in the top left panel.Turtle ID 1 had a migration path to St. Kitts & Nevis (orange triangles), Turtle ID 7 had a migration path to LRNP (green circles), and Turtle ID 9 had a brief migration path to and from Buck Island, thereafter foraging locally (purple square)

Table 1 .
Croix Turtle identification, size (curved carapace length; CCL), and satellite-tracking dates for 10 female green sea turtle Chelonia mydas nesters tagged at Buck Island in the US Virgin Islands from 2011 to 2014, including information on inter-nesting and migration.Inter-nesting areas presented here were comprised of at least 20 mean daily locations; variables include information regarding the timing, duration, area, depth, and distance to shoreline for each inter-nesting period.Periods are given as mm/dd/yy.FINLs: filtered inter-nesting locations; MDLs: mean daily locations; KDE: kernel density estimation; centroids were derived from 50% KDEs; na: not applicable.Site fidelity proportions (p) were all > 99.0099; turtle tracks with p > 95.00 indicate movement paths that are more constrained than random movement paths.
NE: not estimated because centroid had positive value

Table 2 .
Foraging area characteristics for 10 satellite-tracked female green sea turtles Chelonia mydas that originally nested on Buck Island, US Virgin Islands.Foraging kernel density estimation (KDE) areas presented here were comprised of at least 20 mean daily locations (MDLs), and minimum convex polygons (MCPs) were comprised of less than 20 MDLs; variables include information regarding the timing, duration, area, water depth, distance to shoreline, and foraging area characteristics for each foraging period.Centroids were derived from 50% KDEs or 95% MCPs.Foraging periods are given as mm/dd/yy.