The leatherback sea turtle (Dermochelys coriacea) is morphologically and physiologically unique compared with all other species of sea turtles (Heppell et al. 2003; National Marine Fisheries Service and US Fish and Wildlife Service [NMFS and US FWS] 2013). It is the oldest living, largest, and most divergent species of sea turtle. It inhabits all of the world's oceans except the Arctic and Antarctic and undertakes long-distance migrations to tropical and subtropical nesting beaches (Reina et al. 2002; Bell et al. 2003; NMFS and US FWS 2013). In general, Atlantic leatherbacks produce an average of 6 clutches per season (nesting up to 11 times in 1 season) and have an average internesting interval of 9–10 d and an average remigration interval of 2–3 yrs (Boulon et al. 1996). This is unique when compared with hard-shelled sea turtle species that lay significantly fewer clutches per season (only 2–4) and have significantly greater internesting intervals (loggerhead turtle Caretta caretta 12–16 d, green turtle Chelonia mydas 10–14 d, and hawksbill turtle Eretmochelys imbricata 13–15 d; Miller 1997). Leatherbacks are also hypothesized to reach sexual maturity at an earlier age than most sea turtle species at 9–14 yrs (Zug and Parham 1996; Jones et al. 2011) versus 20–40 yrs for the hard-shelled species (Chaloupka and Musick 1997; Heppell et al. 2003). Leatherbacks produce a diminished number of eggs per clutch (80–85 on average [Garner and Garner 2010] vs. 112–130 for loggerhead, green, and hawksbill sea turtles [Miller 1997]) and have significantly lower hatch success at only 50%–55% on average compared with > 80% average for other sea turtle species (Miller 1997; Garner and Garner 2010; Rafferty et al. 2011). Reproductive parameters and nesting statistics, such as number of reproductive females in a given year, number of eggs and clutches laid, remigration interval, and hatch success, are vital to assessing sea turtle abundance and population trends (National Research Council 2010). Variations over time in these factors will impact the demographics for a given population and ultimately impact the rate of recovery. Because sea turtles exhibit interannual variation in nesting numbers and productivity (Broderick et al. 2001; Heppell et al. 2003), long-term evaluation of a defined population is the most effective way to evaluate population trends.

The population of leatherback sea turtles nesting at Sandy Point National Wildlife Refuge (SPNWR), St. Croix, US Virgin Islands, has been studied comprehensively for over 30 yrs. Flipper tagging of leatherbacks was initiated in 1977 by the Virgin Islands Department of Planning and Natural Resources (Dutton et al. 2005). In 1978, the US Fish and Wildlife Service designated critical habitat for leatherbacks on land at Sandy Point, and in 1979, the National Marine Fisheries service designated in water critical habitat for leatherbacks (Boulon et al. 1996). In 1981, an annual leatherback sea turtle research and conservation project was instituted, and since then it has provided long-term protection of this population, in conjunction with saturation tagging and nest management programs. Historical data show a population that initially had 20 or fewer individual nesting females annually. This low number was due in part to lack of protection for nesting females, eroding nesting habitat, and a high incidence of egg poaching (Boulon et al. 1996). It is estimated that prior to 1981, approximately 100% of all nests laid annually were lost to poaching activity, and before the initiation of the nest relocation program in 1982, an estimated 60% of all nests laid were lost due to erosion (Boulon et al. 1996). In 1984, the US Fish and Wildlife Service acquired Sandy Point as part of its National Wildlife Refuge system. Habitat protection, in conjunction with nightly patrols and nest management, afforded additional protection of nesting adults and eggs and resulted in a reduction in poaching from 100% to 0.0%–1.8% per year (Boulon et al. 1996; Garner and Garner 2010). The relocation of nests in danger of erosion or inundation (approximately 30%–65% of nests laid each year) decreased nest loss due to erosion to < 5% annually (Boulon et al. 1996; McDonald-Dutton et al. 2001; Garner and Garner 2010). As a result, the number of hatchlings produced annually increased exponentially, and population numbers increased to a record of 202 individual turtles in 2009 (Boulon et al. 1996; Dutton et al. 2005; Garner and Garner 2010). While a majority of the population exhibits high nest site fidelity to the beach at Sandy Point, interbeach and interisland movements of leatherbacks have been documented annually for a small number of turtles (Boulon et al. 1996; Dutton et al. 2005; Garner and Garner 2010). Leatherbacks from Sandy Point have been documented nesting on beaches north of Sandy Point on the west end of St. Croix as well as on south shore beaches (such as Ha'Penny and Manchenil) and occasionally on additional satellite beaches around the island (Boulon et al. 1996; Dutton et al. 2005; Garner and Garner 2010). Tag returns have also identified interchange of turtles between Sandy Point and Culebra, Vieques, mainland Puerto Rico, the British Virgin Islands, St. Kitts, Anguilla, and Dominica (Boulon et al. 1996; Dutton et al. 2005). This suggests that the St. Croix population is part of a larger, more regional population that includes neighboring beaches and islands and includes subgroups that exhibit greater fidelity to their particular nesting beaches (Boulon et al. 1996; Dutton et al. 2005, 2013).

Although the population has increased since the project's inception, the most dramatic increase occurred between 1997 and 2001 and is attributed partly to the increased hatchling production in the 1980s (Dutton et al. 2005; Garner and Garner 2010). This increase includes a significant number of neophytes recruited into the population each year and coincides with the proposed age of sexual maturity (9–14 yrs) for the hatchlings produced in the early years of the project (Dutton et al. 2005). Genetic data also support the theory that first-time nesters encountered more recently are indeed the offspring of turtles protected after the inception of the project (Dutton et al. 2005). The increased production of hatchlings at Sandy Point most likely benefited leatherbacks on nearby beaches within the regional metapopulation, such as Culebra, Vieques, mainland Puerto Rico, and other islands in the wider Caribbean (Dutton et al. 2005, 2013).

The drastic population increase at Sandy Point also occurred at a time when population statistics were more reliable and accurate due to the use of passive integrated transponder (PIT) tags. The number of neophytes in the population was likely inflated in the early years (between 1982 and 1992) because of low flipper tag retention (McDonald and Dutton 1996; Dutton et al. 2005). Only 50%–60% of the flipper tags applied were retained through the first remigration, and those that remained lasted only 3–4 yrs (McDonald et al. 1996; Dutton et al. 2005). The use of pineal spot photographs for individual identification helped mitigate the effects of flipper tag loss prior to the implementation of PIT tag use in 1992 (McDonald et al. 1996). Taking into consideration interannual variation in nesting numbers (based on an average 2–3-yr remigration interval), tagging and reproductive data obtained from 1995 to 2010 provide an additional 15 yrs of comprehensive and reliable data to assess the population and build on the 15-yr conservation summary provided by Boulon et al. (1996).

Comprehensive mark–recapture programs that tag every female and confirm repeated observations of nesting activity and productivity minimize the error caused by unobserved females and provide the most reliable estimates for population trends and analysis (Witherington et al. 2009). Changes in productivity (number of clutches laid and hatch success), recruitment rate, and remigration interval have been attributed to declining nest counts and population numbers in sea turtle populations (Hays et al. 2000; Witherington et al. 2009). Therefore, these parameters were hypothesized to be important factors contributing to the population dynamics observed at Sandy Point. The purpose of our study was to reevaluate these and other parameters for the leatherback population at Sandy Point National Wildlife Refuge for the first 30 yrs of the conservation project to improve estimates relevant to the recovery plan for leatherbacks in the Atlantic.

METHODS

Nesting Beach Monitoring

The nesting leatherback population at SPNWR served as the study group. From 1995 to 2010, nightly beach patrols were conducted following the data collection methods outlined in Boulon et al. (1996), which were the same ones used for the first 15 yrs of the project. Patrols were initiated annually beginning 1 April and continued until approximately 10 d after the last female leatherback nested (usually mid-July to August). The beach was patrolled nightly on foot during this time frame, starting around 2000 hrs until either 0500 hrs or until the last female had finished nesting.

The 3-km study area (Fig. 1) was divided into multiple sections, and each area was patrolled hourly (at a minimum) to ensure that all nesting turtles were observed, tagged, and recorded. Metal Monel or Inconel flipper tags were applied (type of flipper tag depended on availability; National Band and Tag Co, Newport, KY) to either the left or the right rear flipper of individual turtles. (Front flippers were tagged in the early years, but that method was halted because of extremely low tag retention.) Glass-encased microchips (AVID PIT tags, Avid Identification Systems Inc, Norco, CA) were injected into the shoulder muscle during successful oviposition events and were read using a handheld AVID scanner (Dutton and McDonald 1994). From the late 1980s through the early 2000s, pineal “pink spot” photos were also obtained and used for individual identification of turtles that had lost their flipper tags (McDonald and Dutton 1996). Every time a turtle was encountered, all data regarding nesting activity and turtle identification were recorded. If a turtle crawled onto the beach but returned to the water without attempting any nesting behaviors (i.e., body pitting or digging), it was documented as a “track only.” If body pitting, digging, or laying activities were attempted but no eggs were deposited, it was documented as a “dry run.” If an activity was missed during patrols, the turtle identification was labeled 999-999 (unknown). The track was evaluated, and if it was deemed a potentially successful nest, it was documented as a “probable lay” and verified on excavation after the appropriate incubation period. All nests in danger of erosion (based on annual sand cycles) or inundation or with standing water in the nest were relocated per standard protocols (Garrett et al. 2010) to more stable areas of the beach that resembled the original nest site (Boulon et al. 1996; Garner et al. 2005). Nests were relocated within 1 hr (on average) of oviposition and constructed in suitable habitat to species-specific shape and dimensions (Dutton et al. 1992). The location of all nests were triangulated using numbered stakes placed at 20-m increments along the vegetation line of the 3-km study site. Nests were excavated 2–3 d postemergence unless there was an immediate threat of predation (i.e., nests near the vegetation line were susceptible to cat and mongoose predation or if there was a known presence of dogs or predatory birds on the beach). A representative sample of relocated and in situ nests were chosen for excavation in all beach zones. Probable lays and nests that were not observed to hatch (i.e., no visible tracks) were also excavated to confirm potential nests and ensure that hatching data were not skewed toward observed emergences and away from nests with zero (and low) hatch success. On excavation, all nest contents were categorized to determine hatching and emergence success.

View Figure

• Download as Powerpoint
• Download Figure

Statistical Analyses

Nesting numbers and basic reproductive statistics were compiled at the end of each season from 1981 to 2002 and in real time from 2003 to 2010 (using the WIMARCS relational Sea Turtle Database). Data from the study site at SPNWR are reported from 1981 to 2010 and encompass the first 30 yrs of the annual saturation tagging, research, and conservation project.

The number of individual neophytes (untagged females who were first-time nesters) and remigrants (tagged females with a reproductive history) were counted and reported. Turtles that exhibited tag holes or flipper tag-related scars were considered unknown remigrants unless valid identification could be determined via pink spot photo-documentation or scar patterns. Turtles that were originally tagged at beaches other than SPNWR were also deemed new remigrants if they had not previously nested at Sandy Point. The average remigration interval (RI) for the population (sum of the RIs for each individual/total number of turtles) was calculated for each year. The remigration interval was the interval between consecutive nesting years (Hays 2000). The average number of nests laid (sum of the number of nests recorded per turtle/number of turtles) and average number of hatchlings produced (Boulon et al. 1996) per turtle (number of hatchlings produced/number of nesting turtles) were also calculated. Nonlinear regression analysis was conducted to establish trends in number of neophytes, remigrants, and total turtles observed annually as well as the average number of nests laid and average hatching success over the project duration. Taking into consideration a general 2-yr remigration interval for leatherbacks and a historical pattern of high output in odd years and low output in even years, an analysis of variance was conducted to determine if differences observed among even and odd years for the total number of turtles, number of neophytes recruited into the population, and number of remigrants observed were indeed significant (p < 0.05) and warranted further analyses of reproductive parameters segregated by odd- and even-year categories.

Hatching success was calculated annually as (hatched shells/total yolked eggs) (Boulon et al. 1996; Garner and Garner 2010). Nests lost due to or impacted by erosion were not included in hatch statistics. “Yolked eggs” included undeveloped, unhatched, and unhatched terms as defined by Miller (1999). Emergence success [hatched shells − (dead hatchlings + live in nest)/total yolked eggs] (Dutton et al. 2005; Garner and Garner 2010) was also calculated. “Live in nest” are hatchlings or live full-term pipped eggs that are found in the bottom of the nest and would not have emerged (Garner and Garner 2010). Dead full-term pipped eggs are considered “dead hatchlings” in the nest for the purpose of emergence success. Hatchlings (alive or dead) that are found in the neck of the nest during excavation were considered emerged (Garner and Garner 2010). Annual hatchling production was calculated as (average number of eggs per clutch (based on preinventory counts from relocated clutches) × number of in situ nests × in situ emergence success) + (average number of eggs per clutch × number of relocated nests × relocated emergence success).

RESULTS

Summary data for the number of individual turtles encountered annually as well as a summary of remigration data are presented for all years of the project, including the years (1977–1981) prior to saturation tagging (Table 1). Consistent beach patrols began in 1981, and these data are considered to some degree in the results, but they are not included in the subsequent figures due to an incomplete data set. Since the inception of the project (1981), a total of 1014 individual leatherbacks have been tagged. The number of turtles nesting annually at SPNWR between 1981 and 2010 ranged from 18 turtles (1986) to a project record of 202 individuals in 2009 (Fig. 2). The 2001 and 2007 seasons also exhibited high numbers with 186 and 193 individuals, respectively. The 2010 season recorded the second-lowest number of nesting individuals (94) since 2006. While the population numbers from 2001 to 2010 were generally stable, the dramatic increase observed in the years from 1997 to 2001 did not continue as projected (Garner and Garner 2010), and recovery slowed. The number of nesting individuals was lower in even years compared with that in odd years, with record highs recorded during odd years and lows recorded during even years (Fig. 2). The average number of turtles observed in odd years was significantly greater than in even years between 1997 and 2010 (p = 0.002), coinciding with an influx of neophytes into the population during this time frame.

Table 1. Leatherback remigration intervals to Sandy Point National Wildlife Refuge, St. Croix, US Virgin Islands, from 1977 to 2010 (from Garner and Garner 2010).

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

The remigration interval for the 30-yr study period ranged from 1 to 11 yrs, with the most common intervals being 2 and then 3 yrs (Table 1). The average remigration interval was observed to be relatively low throughout the 1990s (when PIT tagging began providing reliable identification; Fig. 3). It spiked in 2000 (3.17 ± 0.18 yrs [mean ± SD]), decreased through 2003 (2.34 ± 0.09 yrs), and then increased steadily to a record high (3.41 ± 0.20 yrs) in 2008 (Fig. 3). The number of observed 1-yr remigrants has increased over the course of the project, with 5 1-yr remigrants observed in 2010 (Table 1).

View Figure

• Download as Powerpoint
• Download Figure

The number of remigrants observed annually has significantly increased between 1982 and 2010 (R² = 0.76, p < 0.001). The number of remigrants observed annually in the Sandy Point population reached an initial peak of 96 turtles in 2001 and remained stable over the period 2001–2010 (Fig. 4). Odd-year remigrant numbers ranged from 96 to 133 individuals during this time frame, with an average of 115.6 ± 18.58 individuals. Maximum remigrant numbers were observed in 2007 and 2009 (with 136 and 133 turtles, respectively). Even-year remigrant numbers remained steady (ranging between 61 and 80 individuals, with an average of 68.4 ± 7.63 remigrants per year). The average number of remigrants observed during odd years was significantly higher compared with that in even years beginning in 1997 (p = 0.01). Because the total number of nesting females varied annually as well as significantly between even and odd years, the percentage of remigrants observed (number of remigrants/total number of nesting turtles) was also calculated. The percentage of the annual nesting population represented by remigrants increased steadily between 1982 (5.26%) and 2010 (73.4 %) (R² = 0.77) (Fig. 5).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

In contrast, the number of neophytes arriving to nest at SPNWR has remained steady overall (from 1982 to 2010; R² = 0.35, p = 0.01) (Fig. 4), but a gradual decline was observed after 2001 (Fig. 4). The 2001 season boasted a project record, with 90 individuals tagged (Fig. 4). After 2001, the number of neophytes recruited into the nesting population for a given year generally decreased through 2006 (R² = 0.77). While all even years remained low, even year numbers decreased from 2002 through 2006 (R² = 0.77). Odd years (2007 and 2009) showed a slight increase in numbers with 57 and 69 individual neophytes tagged, respectively. The 2010 season recorded the lowest number of neophytes in the last 10 yrs of the study period. In general, the average number of neophytes observed in odd years was also significantly greater than that in even years beginning in 1997 (p < 0 .01). Because the number of total nesting females varied annually as well as significantly between even and odd years, the percentage of neophytes observed (number of neophytes/total number of nesting turtles) was also calculated over the 10-yr period. The overall percentage of neophytes recorded decreased over this period (2001–2010) (Fig. 6). Odd years decreased 14.2% from 48.4% in 2001 to 34.2% in 2009. Even years decreased 10.5% from 39.1% in 2002 to 28.6% in 2010. The 2007 nesting season recorded the third-lowest percentage of neophytes in project history, with 29.53%, while 2008 and 2010 recorded the lowest percentages of neophytes in project history with 28.57% and 26.59%, respectively. Overall, the percentage of neophytes observed has decreased since the inception of the project (R² = 0.77) (Fig. 6).

View Figure

• Download as Powerpoint
• Download Figure

A total of 23 “new” remigrants (turtles that had been previously tagged) were documented between 1995 and 2010 (Table 1). This includes turtles originally tagged in St. Kitts, Anguilla, Culebra, mainland Puerto Rico, and Panama (one turtle). The origins of some new remigrant turtles was not determined and thus remain unknown. Thirty-five new remigrants were also documented from 1982 to 1995 (Table 1). Prior to 1995, Boulon et al. (1996) reported tag returns from turtles who were tagged on other St. Croix beaches as well as other islands, including Culebra, Vieques, mainland Puerto Rico, the British Virgin Islands, and Anguilla. The number of unknown remigrants decreased after 1995 (3 yrs after implementation of PIT tags).

The total number of known nests deposited (relocated + in situ) for the first 30 yrs are presented (Fig. 7). The number of nests laid annually ranged from a low of 72 nests in 1986 to a project high of 988 nests in 2001 (Fig. 7). The total number of nests laid annually increased overall from 1982 to 2010 (R² = 0.64). However, the average number of observed nests laid per individual decreased (Fig. 8). The average number of nests produced per turtle annually was analyzed during the time frame from 1992 to 2010, when PIT tags were utilized and individual identification was more reliable. During these years, a steady decline in the average number of nests laid was observed (R² = 0.84) (Fig. 8). The highest average was observed in 2003 with an average 6.17 ± 2.30 nests per turtle. In spite of this, the average number of nests laid annually per turtle decreased overall to a record low of 3.60 ± 2.16 nests per turtle in 2010.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Overall hatching success also varied over the 30-yr period from a record high of 67.80% ± 20.31% in 1991 to a project low of 40.28% ± 23.20% in 2005. Mean overall hatching success for the first 30 yrs of the project was 58.50% ± 7.75%. Hatching success in the 1990s was generally high (> 65%) but consistently dropped below 60% beginning in 2001 (Fig. 9). Hatching success decreased between 2001 and 2007 (including a record low 40.28% in 2005), then increased between 2008 and 2010 (hovering around 55%). Mean hatching success declined overall throughout the project (Fig. 9). Relocated versus in situ nest hatching success varied annually and is also reported from 1982 to 2010 (Fig. 10). As Boulon et al. (1996) reported previously, between 1982 and 1994, relocated nest hatching success was higher than in situ success for 2 yrs (1982 and 1986); no difference was observed in 1988, and in situ hatching success was significantly higher than relocated hatching success for the remaining 9 yrs of that time period (Fig. 10). For the additional 15 yrs documented in this study (1995–2010), in situ nests exhibited significantly greater hatch success than relocated nests for all years (p < 0.01) (Fig. 10).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Hatchling production ranged from approximately 2200 (1982) to 44,325 (2001) turtles produced annually. Overall production generally increased since the project's inception (1981–2010) but slowed in the last decade (Fig. 11). Because of variations in annual nesting numbers, hatchling production was subsequently evaluated based on population size for a given year. This resulted in the average number of hatchlings produced on a per turtle basis annually and steadily increased from 1982 through 1990 (R² = 0.70) with an average of 327 hatchlings produced per turtle (Fig. 12). Levels remained similar through 1996 before dropping precipitously through 1999 (Fig. 12). The average hatchlings per turtle subsequently increased from 1999 through 2003 to a moderate level (approx. 200–250). It remained significantly lower from 2004 to 2010 (< 200 annually with an average of 157.14 ± 25.54) compared with the 1990–1996 time frame (with an average 310.7 ± 13.19 hatchlings per turtle, p < 0.01).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

Interannual variation in nesting numbers is common in sea turtle populations; therefore, long-term data are necessary to truly evaluate population trends. The nesting population of leatherbacks at Sandy Point, St. Croix, has been comprehensively studied for over 30 yrs and presents the most complete data set for nesting leatherback sea turtles in the Caribbean (NMFS and US FWS 2013). The population at Sandy Point showed a significant increase in the first 20 yrs of the project, increasing from less than 20 nesting turtles in 1981 to 186 turtles in 2001 (Dutton et al. 2005; Garner et al. 2005), which was a record for that time. Population growth through 2001 was estimated at approximately 13% per year (Dutton et al. 2005), with the annual growth rate (1986–2004) estimated at 10% per year (confidence interval [CI]: 7%–13%). Nesting also increased at other local Caribbean rookeries during this time frame. The populations at Culebra Island and mainland Puerto Rico increased from 9 nests in 1978 to 469–882 nests from 2000 to 2005 (NMFS and US FWS 2013). Average annual growth rate for the Puerto Rican population (1978–2005) was estimated to be 10% (CI: 4%–12%) (NMFS and US FWS 2013). Additionally, nesting numbers in the British Virgin Islands increased between 1986 and 2006 with the number of nests increasing from 0 to 6 in the 1980s to 35 to 65 in the 2000s (with an estimated growth rate of 20% from 1994 to 2004 [NMFS and US FWS 2013]).

Along with an overall increase in nesting numbers during the first 20 yrs (1981–2001) at Sandy Point, a concomitant increase in hatchling production was also observed (with the number of hatchlings produced increasing from 2000 to over 44,000 (Dutton et al. 2005; Garner et al. 2005). The annual population growth rate (approximately 13% per year) observed since the early 1990s was not ascribed to increased survival probability of adults because survivorship was determined to remain high and constant during this period at approximately 89.3% minimal survival rate for nesters annually (Dutton et al. 2005). The population increase was instead associated with increased hatchling production. Nest protection (from terrestrial predators) and relocation programs began in the early 1980s and relocated (on average) 30%–40% of the clutches laid annually at Sandy Point (Dutton et al. 2005; Garner and Garner 2010). This program resulted in significantly increased hatchling production and a subsequent population boost due to neophyte recruitment approximately 12–14 yrs later, within the approximate duration to reach sexual maturity (Zug and Parham 1996; Dutton et al. 2005; Jones et al. 2011).

In spite of the interannual variability typical in sea turtle populations, the Sandy Point nesters showed a steady increase through 2001. Trends for the subsequent 10 yrs (2001–2010) have not previously been reported. The population numbers during this decade are generally stable; however, the dramatic increase observed in the years from 1997 to 2001 was halted, recovery slowed, and signs of a potential decline were noted. Odd years in the last decade of the study period did not continue the exponential increase observed from 1991 to 2001, and the 200-nester threshold was not breached until 2009, which was later than predicted (Dutton et al. 2005; Garner and Garner 2010). This decade also yielded record-low nesting numbers during even years, with 2006 and 2010 exhibiting the lowest numbers recorded since 1998 (with 92 and 94 turtles, respectively). A decline in nesting numbers during this time frame has been reported by additional Caribbean populations in spite of the previously observed annual increases prior to the early 2000s. Since 2004, nesting has declined in Culebra, reaching a record low of 5 turtles in 2012 (NMFS and US FWS 2013). Troëng et al. (2004) noted a potential decline in leatherback nesting at multiple Caribbean beaches along the coast of Central America, including Tortuguero, Pacuare, and Gandoca beaches. Between 1995 and 2006, Tortuguero exhibited an overall decline of 67.8% (Troëng et al. 2004). Gandoca beach also showed a slow decline from 1990 to 2004 with nest numbers decreasing after 2000 (Troëng et al. 2004; Chacón-Chaverri and Eckert 2007). Declines at these locations may be attributed to low nest site fidelity and a shift in nesting location (Troëng et al. 2004; Chacón-Chaverri and Eckert 2007; NMFS and US FWS 2013). While there is a large degree of movement by individual females between nesting beaches in the Caribbean region of Central America (Troëng et al. 2004; Chacón-Chaverri and Eckert 2007), that does not appear to be the case for Sandy Point turtles. Some exchange occurs between Sandy Point and other western Caribbean beaches; however, it appears to be limited. Only 58 unknown remigrants were reported nesting at Sandy Point during the first 30 yrs (1981–2010), and reports of Sandy Point turtles nesting at regional beaches on other islands were also limited and sporadic during this time frame. Tag returns and genetic data from mitochondrial DNA analysis further support the theory that the St. Croix population is a demographically independent group with high nest site fidelity (Dutton et al. 2013; NMFS and US FWS 2013); therefore, other explanations for the low nesting numbers in the past decade must be considered.

The interannual variability in nesting numbers observed over the years in sea turtle populations is generally associated with varying remigration intervals (Hays et al. 2000; Broderick et al. 2001); thus, the decreased nesting numbers observed between 2001 and 2010 may be an artifact of varying remigration intervals during the limited time frame of 10 yrs. However, in context with the 30-yr history of the project, the average remigration interval at Sandy Point appears to be increasing, with a project high of 3.41 observed in 2008. Remigration intervals vary for each individual nester and are based on reaching a nutritional threshold for reproduction and migration (Hays et al. 2000; Broderick et al. 2001). The ability to reproduce in a given year is directly linked to foraging ground productivity (and effective exploitation of these areas by individuals) and ultimately affects population numbers annually at nesting beaches (Hays 2000). Environmental stochasticity, climate change, and sea surface temperatures can directly influence the foraging capacity of sea turtle species and ultimately the remigration interval (Wallace et al. 2006). Remigration intervals, when increasing consistently on an annual basis over multiple years, may negatively impact population productivity and result in a low number of annual nesters (Witherington et al. 2009), such as that observed at SPNWR during multiple seasons after 2001.

The number of nesting remigrants appeared to be constant to slightly increasing at Sandy Point during the study period, thus supporting the observed trend of high adult survivorship (0.893; CI: 0.87–0.92) described by Dutton et al. (2005). Therefore, adult survivorship does not appear to be directly related to any observed increase or decrease in population numbers at Sandy Point. High adult survivorship has been noted for other nesting populations (Florida = 0.889, French Guiana = 0.91) (Rivalan et al. 2005; Stewart et al. 2014), with the annual survival rates in the Atlantic generally estimated to be higher than that of Pacific turtles (Playa Grande = 0.65, Papua New Guinea = 0.85) (Dutton et al. 2005; Rivalan et al. 2005; NMFS and US FWS 2013). Troëng et al. (2004) suggests that the increased adult survivorship observed in the Atlantic is due to decreased fisheries interaction and hypothesizes that there may be less overlap between leatherback habitat (i.e., foraging area, nesting habitat, and migratory corridors) compared with Pacific turtles.

Although adult survivorship has been analyzed for multiple nesting populations, data regarding survivorship of hatchlings and juveniles are extremely limited. Increased mortality of early life stages could account for the observed decrease in number and percentage of neophytes observed annually at Sandy Point. Spotila et al. (1996) estimated the first-year survival rate (hatching through year 1) of the global population to be approximately 0.06. Eguchi et al. (2006) estimated the annual juvenile survival rate of St. Croix leatherbacks to be 0.63. Assuming a time period of 9–13 yrs to reach sexual maturity, Eguchi et al. (2006) also determined the survival rate from hatchling to first year of reproduction for Sandy Point females to be between 0.004 and 0.02. Changes in survivorship of earlier life stages may significantly impact recruitment (Dutton et al. 2005; Witherington et al. 2009). Troëng et al. (2004) notes that the Atlantic longline fisheries increased the number of hooks set in the north and tropical Atlantic during the 1990s, possibly increasing juvenile and adult mortality rates and ultimately impacting recruitment to nesting beaches. Marine pollution has also increased over the past 30 yrs and may also impact mortality of the pelagic life stages, as ingestion or entanglement in marine debris can be fatal (NMFS and US FWS 2013; Nelms et al. 2015).

Additionally, climate change represents a serious threat to sea turtles. Overall, leatherback movements are correlated with the distribution of meso-zooplankton in the Atlantic (Fossette et al. 2010). The distribution and availability of this prey could be altered as changes in water temperature, ice cover, salinity, carbon dioxide and oxygen levels, and ocean currents alter the range and abundance of plankton in the marine environment (NMFS and US FWS 2013). The increase in sea temperatures is expected to result in a shift in leatherback foraging areas, with habitat expanding into higher latitudes (James et al. 2006; McMahon and Hays 2006; NMFS and US FWS 2013). Nutritional status and the ability to obtain prey will directly impact the remigration interval. Additionally, it will potentially impact the time it takes for a juvenile to reach sexual maturity. If maturity is delayed, then recruitment is also delayed. Therefore, the lows observed over the past 10 yrs may be due to delayed or decreased recruitment in addition to an increasing RI (Witherington et al. 2009). This hypothesis is supported by the decreased number and percentage of neophytes observed annually at Sandy Point. The percentage of the population represented by neophytes has decreased steadily from 94.7% in 1982 to 26.6% in 2010. Tag retention accounts for a large drop in the percentage of observed neophytes during the early years. However, the percentage of neophytes observed annually continued to decrease post-1995, when PIT tags had been implemented for 3 yrs, the project had been conducted for 15 yrs, and remigrants were consistently and reliably identified. The number of neophytes spiked in 1997 and is attributed in part to the nest relocation effort and subsequent increase in hatchling production. However, the number of neophytes observed annually was steady after 1997 to decreasing (2001–2010), and the percentage of the population represented by neophytes continued to decrease annually through 2010. This decrease occurred in spite of an overall increase in hatchling production. Additionally, the recruitment of hatchlings to neighboring islands on reaching sexual maturity may also contribute to the observed decline in neophyte numbers on St. Croix. The Sandy Point population shows high nest site fidelity, and Dutton et al. (1999, 2007) suggest that significant mitochondrial DNA differentiation between Atlantic rookeries could mean that natal homing in leatherbacks is quite precise (NMFS and US FWS 2013). However, the Sandy Point population could potentially serve as a source population for nearby rookeries, such as mainland Puerto Rico. If this is the case, decreased productivity at Sandy Point could impact the future nesting numbers at other local beaches. Comprehensive monitoring and continued genetic studies at nearby beaches (such as Culebra, Vieques, and mainland Puerto Rico) will help determine if a local shift in nesting beach use is occurring and to what extent Sandy Point is contributing to nesting numbers there.

Factors directly affecting reproductive productivity, such as number of nests laid per turtle and average hatch success, will impact hatchling production. Hatchling production has been linked to the drastic population increase within the first 20 yrs of the project (Dutton et al. 2005), and recent decreases may result in a continued decline in future recruitment and delayed population recovery. Hatch success at SPNWR decreased from 2001 through 2007 and reached a historical low in 2005. Average number of hatchlings produced annually per turtle also decreased in the 2000s compared with the 1990s. Abnormal erosion patterns observed at Sandy Point may account for some of these results (Garner and Garner 2010) as well as variations in sand properties, bacterial load, annual precipitation, and temperature. Previous studies conducted at SPNWR could not pinpoint the specific biotic or abiotic factors in the nest environment responsible for the decreasing hatch success observed in this population (Garrett et al. 2010). However, Santidrian Tomillo et al. (2015) found that local precipitation significantly impacted hatch success at Sandy Point, with hatching success increasing with increased precipitation. While the air temperature did not significantly impact hatching success at Sandy Point, the best generalized additive model for hatching success at Sandy Point included both air temperature and precipitation (rain accumulated during the month eggs were deposited as well as 2 mo prior) (Santidrian Tomillo et al. 2015). In their analysis, Santidrian Tomillo et al. (2015) also reported a declining trend in hatching success (1980–2012) and attributed it, at least in part, to a concomitant decrease in precipitation over time. It was also projected that the Sandy Point population was susceptible to a greater rate of temperature increase (+5.4°C) due to global warming (through 2100) compared with Caribbean populations at Playa Grande and Pacuare (Santidrian Tomillo et al. 2015). Results projected a trend toward slightly drier conditions, thus potentially exacerbating declines in hatching success in the future for this population.

Additional research suggests that intrinsic maternal influences also contribute to hatching success. Garrett et al. (2010) reported interclutch variation in hatching success, with hatching success declining with increased clutch number on St. Croix. Rafferty et al. (2011) found that multiple maternally derived factors impacted hatching success, including maternal identity, reproductive experience (neophyte vs. remigrant), and interclutch interval. Maternal investment in the egg (via allocation of energy and nutrients) may play a significant role in why some mothers have better success than others (Rafferty et al. 2011). Perrault et al. (2012) correlated maternal physiological parameters with hatching and emergence success in Florida leatherbacks. These included gamma globulin protein, red blood cell count, alkaline phosphatase, blood urea nitrogen, calcium, the calcium-to-phosphorous ratio, carbon dioxide, cholesterol, creatinine, and phosphorous. Perrault et al. (2011) also positively correlated levels of selenium and the selenium-to-mercury ratio in the liver with hatching and emergence success in leatherbacks. Complex interactions between environmental and maternally derived factors are likely responsible for the decrease in hatching success observed at Sandy Point and deserve further investigation.

The number of nests laid per turtle at Sandy Point also decreased from 2001 through 2010 and decreased in general over the previous 20 yrs. The number of nests produced by an individual female is likely linked to nutritional status and maternal health as well (Rafferty et al. 2011). A decreased number of nests produced per female, in conjunction with decreased hatching success rates, resulted in the decreased hatchling productivity numbers observed between 2001 and 2010 in spite of some years exhibiting increased nesting numbers. Long-term decreases in hatchling productivity will be detrimental to population recovery, with effects delayed for 9–14 yrs, the minimum interval assumed to be required for sexual maturation.

Decreased productivity (the number of nests laid, hatch success, and hatchlings produced per turtle), in association with the increased remigration intervals and the decreased recruitment rate observed throughout the project, appear to have impacted the recovery of the Sandy Point population through the 2000s. Although the number of remigrants and adult survivorship are generally stable, these same individuals are taking longer to return to reproduce and producing fewer nests and hatchlings over time. The robust historical data set at Sandy Point provides the opportunity to test and model how the observed changes in reproductive parameters will impact the population in future generations. Further documenting and analyzing the demographic changes over time will provide insight into the long-term population dynamics and survival rates for St. Croix leatherbacks and help ascertain if the decline is the precursor to a potential population demise.