Internesting and Postnesting Movements and Foraging Habitats of Leatherback Sea Turtles (Dermochelys coriacea) Nesting in Florida
ABSTRACT
We tracked 10 leatherback turtles by satellite from 2 Florida Atlantic Coast nesting beaches for a period ranging from 38 days to more than 454 days. Movement and foraging areas were often coastal, which contrasts with other satellite telemetry studies where leatherbacks are more pelagic. Using kernel home-range estimation we identified the primary internesting residence areas as well as Atlantic foraging areas. The primary internesting habitat was centered east–southeast of Cape Canaveral, Florida, from 2 to 60 km offshore and extending 215 km along the coast. Atlantic foraging areas were located primarily on the continental shelf from 30° to 50°N, and in an offshore area centered at 42°N, 65°W, as well as off Africa in the Mauritania upwelling. Seasonally, the location of these foraging areas changed, occurring on the North American continental shelf from March through November and off the shelf from December through February. One of the tracked turtles may have been killed with 17 other leatherbacks by coastal shrimp fishing located near the Georgia–Florida border. We illustrate how using remotely sensed data could be used to prevent such mortalities.
Leatherback turtles (Dermochelys coriacea) are among the widest-ranging large marine vertebrates in the world. Studies have demonstrated that this species may swim in excess of 10,000 km per year and may circumnavigate entire ocean basins (Eckert 1998, 2006; Hughes at al. 1998). Distribution of the species has long been considered to be temperate to north temperate with the species only entering tropical waters to breed (Pritchard 1976). Recent advances in satellite telemetry have enabled research on the movements and oceanic habitat use of leatherbacks (e.g., Keinath and Musick 1993; Eckert 1995, 1998; Hughes et al. 1998; Ferraroli et al. 2002, 2004; Hays et al. 2004; James et al. 2005a, 2005b). Results confirm that leatherbacks are primarily pelagic in their distribution and behavior. Additionally, in the North Atlantic Ocean, the species may make annual north–south migrations, entering tropical waters in nonbreeding years (Eckert 1998, 2006; James et al. 2005a, 2005b)
The recent elevation of leatherbacks' global status to Critically Endangered by the International Union for the Conservation of Nature and Natural Resources (IUCN 2004), and the realization that mortality occurring in coastal waters, and on the high seas, may be contributing to the endangered status of the species has resulted in more efforts to understand movements, migrations, and habitat use (Morreale et al. 1996; Eckert and Sarti 1997; Eckert 1998, 2006; Hays et al. 2004). One source of coastal mortality is shrimp fishing operations along the southeast United States which led the US National Marine Fisheries Service to institute mitigation measures in the form of time-area closures and gear modifications. Satellite telemetry studies of leatherbacks indicate that most turtles disperse far from coastal waters after nesting (e.g., Morreale et al. 1996; Papi et al. 1997; Eckert and Sarti 1997; Eckert 1998; Hays et al. 2004; Hughes et al. 1998; James et al. 2005a, 2005b) and reside outside of areas where shrimp fishing is common. However, most leatherback telemetry studies in the Atlantic have been from low-latitude nesting colonies (with the exception of South Africa) or from high-latitude foraging areas. Florida nesting beaches are continental and in close proximity to coastal shrimp fishing areas of Florida, Georgia, South Carolina, and North Carolina and are close to North Atlantic foraging areas (Shoop and Kenney 1992; James et al 2005a; Eckert 2006). It might be expected that Florida leatherbacks would transit through coastal shrimp fishing areas waters postnesting, and would thus be subject to incidental capture during that transit period. This study seeks to determine both internesting and postnesting movement patterns of leatherback turtles from east Florida nesting colonies. This information is used to delineate important Atlantic habitats for the Florida nesting population as well as to improve our understanding of whether the species moves through southeast United States coastal waters on their journey to high-latitude foraging areas.
METHODS
Satellite transmitters used in this study were Wildlife Computers (Redmond, WA) SSC3 platform transponder terminals (PTTs) and a Sirtrack (New Havelock, New Zealand) ‘Kiwisat'. For this study, both were configured for a 1-W output. Transmitter location and data uplink was via the ARGOS CLS satellite system (Argos 1996). The SSC3 measured 14 cm × 14 cm × 5 cm, weighed 1100 g in air, and had a sloped front and back to improve hydrodynamic performance. It was equipped with a microprocessor to collect and summarize sensor data as well as to regulate transmission intervals and data collection. The SSC3 collected data on dive depth, dive duration, time spent at specific depths, and time spent at specific temperatures. Each variable was summarized and reported in 4 6-hour frequency distributions (= 24 hours) when the turtle surfaced to breathe. Maximum dive depth in each 24-hour period and time spent near the surface (depth was user-designated) for 6 and 12 hours prior to transmission was reported. Transmitter performance including battery voltage and the number of transmissions made was also reported. Dive information will be summarized in a separate publication.
The Kiwisat measured 25 cm × 8 cm × 6 cm, weighed 750 g in air, and was half-teardrop shaped. This transmitter can be used to determine location and report surface temperature at the time of transmission.
Transmitters were attached to a flexible harness, which was placed on turtles during egg-laying. The harness was constructed of 3.75-cm flat webbing, silicon elastic, and stainless steel D-rings (modified from Eckert and Eckert 1986). The shoulder straps of the harness were encased in thick polyvinyl tubing to prevent chafing. Both harness and transmitter were coated with ULTRA-KOTE antifouling paint (Interlux Yacht Paint Inc., Union, NJ). No invasive manipulation of the turtle was required for the attachment of the harness.
Each turtle was measured (curved carapace length and width), tagged with a microchip passive integrated transponder tag (PIT tag) injected in the right shoulder muscle and with external inconel flipper tags to the rear flippers. Other data collected included time, date, location on beach (using a Magellan Colortrak global positioning receiver), and the time the turtle entered the water. In some cases a VHF radio transmitter was added to the harness to assist in relocating the turtle when she returned to nest on subsequent occasions.
Locations provided by ARGOS are not equally accurate. For each reported location, ARGOS calculates a measure of accuracy based on a number of variables, including uplink strength and number of uplinks. This measure of accuracy is reported as a location class (LC). LCs are classified as Z, B, A, 0, 1, 2, and 3, with 3 of highest accuracy and Z as lowest. For mapping, we filtered data to exclude aberrant locations using a filtering algorithm for PC-SAS Version 8.0 created by Dave Douglas (http://alaska.usgs.gov/science/biology/spatial/). The program uses 3 primary filtering algorithms, with a number of user-based parameters. The minimum redundant distance (MRD) algorithm allows the user to set a maximum distance from an initial location point beyond which locations are rejected. This filter is used primarily for data sets of slow-moving animals or with high-resolution locations. The distance-angle rate (DAR) algorithm evaluates locations based on the divergence from a path determined by 3 consecutive points and is best used for migrating animals or animals that move in distinct directions. User inputs for this algorithm include a limit on the plausible rate of travel and the acceptable angle of divergence from a trackline. There is also a hybrid filter that combines both MRD and DAR parameters. In all cases, the program never filters out LC 3 locations, and allows the user to set that limit lower. The program output consists of a number of data sets including, all locations (no filtering), MRD-filtered locations, DAR-filtered locations, and hybrid-filtered locations. It also provides data sets consisting of the single best locations per duty cycle from the MRD-, DAR-, and hybrid-filtered data sets. We provide our user-set parameters for the Douglas filter in Table 1. For location data in 2000 and 2001, we set the LC acceptance level at LC ≥ 2 and for 2002 we accepted LCs ≥ 1 because in 2002 reported location accuracy was improved. An important input to the Douglas filter is the maximum travel rate per hour. To calculate a logical value for this travel rate we used standard locations (LC = 1,2,3) that were more than 1 hour apart and calculated a 95% percentile (4.07 km/h) travel rate. This estimate eliminated the influence of outlier data and was comparable to 95% percentile of maximum swim speed for leatherbacks (Eckert 2002).
An advantage of using the Douglas filtering program is that it allowed us to retain locations which might have otherwise been rejected if we had simply filtered data based on LC > 0 (the minimum level at which ARGOS provides accuracy estimates). Without the Douglas filter program, in 2000 we would have used only 21% of all reported locations; in 2001 we would have used 33% of the locations and in 2002, 27% of the locations. Instead we were able to use 73%, 70%, and 64%, respectively, of the locations reported and significantly improved the resolution and sample size of each trackline.
Geographic information systems plotting software (ArcView 3.2, Environmental Systems Research Institute, Inc.) was used to map turtle movements and calculate high-use areas and movement pathways. To define important habitats for each turtle we calculated fixed kernel home ranges (postnesting) using Home Range Extention for ArcView, Version 0.9 (Rodgers and Carr 1998). To reduce temporal autocorrelation and sampling bias, data sets were filtered using the best single location per day hybrid-filtered data set (Dave Douglas, USGS) and the internesting period data were assessed separately from postnesting period data.
To calculate mean daily rates of travel, we multiplied the hourly rates by 24 and calculated a daily mean. Because distance between surfacings provided by satellite telemetry does not account for subsurface distance it cannot be used to calculate actual swim speed, but nonetheless an estimate of geographic distance over time has value in understanding turtle behavior (Eckert 2002).
RESULTS
Ten satellite transmitters were deployed on 10 nesting leatherbacks in Florida from 2000 to 2002. Seven transmitters were deployed at the Archie Carr National Wildlife Refuge (ACNWR), Melbourne Beach, and 3 at Juno Beach. All equipment was deployed during the peak of the leatherback nesting season, late May through early June (Table 2). The first leatherback tracked from a Florida nesting beach was equipped with a Sirtrack Kiwisat PTT on 19 May 2000 in the ACNWR. In 2001, 5 more females from the ACNWR were tracked with Wildlife Computers SSC3 PTTs. In 2002, 4 SSC3 PTTs were deployed, 3 at Juno Beach, and 1 in the ACNWR.
Tracking duration ranged from 38 days to more than 454 days; 2 animals from the 2002 nesting season were still transmitting as of 1 June 2003. Although it is known that these turtles continued to be tracked after that time, the data are held by another investigator and were not made available to the authors of this study. Number of transmissions ranged from 9383 to > 57,215 (Table 3). Of the 10 turtles tracked, the fates of 5 are known. DC1, originally telemetered in 2000, returned to the nesting beach in 2003 without the harness and transmitter. DC6 was equipped with a transmitter in 2001 and only tracked for 4 months. Its transmitter had over 44,000 transmissions logged at deployment, and over 76,000 when it stopped with a battery voltage below operational levels (6.98 V). It is clear that the batteries had expired on this transmitter. In 2003, DC6 nested at Canaveral National Seashore in Florida without the harness and transmitter. DC7 was still transmitting as of 1 June 2003, as was DC10. Subsequent information is unavailable to the authors on the fate of these turtles. The carcass of DC8 washed up on a New Jersey beach a few weeks after its transmitter ceased transmissions; the harness and transmitter were not on the turtle. Expiration of tracking records of the other 5 turtles could have been because of battery or electronic failure, loss of the attachment harness, or mortality of the turtle.
Internesting movements were recorded for 9 of the 10 turtles (Fig. 1). Twenty of those intervals were from turtles nesting at ACNWR and 5 intervals from turtles nesting at Juno Beach. Turtles generally moved north from their nesting beaches before returning to nest. For turtles that were tracked over more than 1 internesting period, the pattern was rarely similar to the previous period, so we treat each period as an independent event. The kernel 95% utilization distribution (UD) of the internesting habitat for these 10 turtles delineates a region centered to the east–southeast of Cape Canaveral 2–60 km from shore and extending approximately 100 km to the north and south (Fig. 2). As delineated by smaller UDs, turtles spent most of their time in close proximity to their nesting beaches.
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[239:IAPMAF]2.0.CO;2
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[239:IAPMAF]2.0.CO;2
At the conclusion of the nesting season, all turtles left the immediate area of their beaches and moved north (Fig. 3). Four of the turtles (DC1, DC4, DC8, DC9) remained on the North American continental shelf throughout the duration of tracking (range 48–213 days) whereas the other 6 moved out into the Atlantic for at least a portion of their records (range 38–454 days) (Fig. 4). Of the 6 that moved off the shelf, DC5 crossed the Atlantic Ocean, spent a minimum of 53 days between the Cape Verde Island and Mauritania, and was finally recorded 300 km off the coast of Liberia (454 days since deployment). DC2 moved north of Cape Breton, Nova Scotia, Canada, then south to 12°N, 52°W where her signal ceased (262 days). DC3 was only tracked briefly (38 days) but seemed to be moving along the same route as DC5. The final 2 turtles (DC7 and DC10), spent identical amounts of time in pelagic waters (44%) and eventually returned to the continental shelf.
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[239:IAPMAF]2.0.CO;2
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[239:IAPMAF]2.0.CO;2
Use of UDs set to 50% and 95% provided best visualization of travel pathways and high-use areas for all but 3 of the turtles (DC3, DC5, and DC9) (Fig. 4). The tracking records for DC3 and DC9 were too short (38 and 48 days) to provide enough postnesting location data for kernel home-range analysis. The 50% UD assessment overestimated DC5′s high-use areas, so it was reduced to 25%.
Travel areas delineated by the 95% probability for all turtles formed a broad distribution in the western Atlantic, extending to 52°W in higher latitudes (35°–45°N) within 50 km of the Atlantic coast of North America (Fig. 4). The 50% probability (25% for DC5) describes important areas where Florida leatherbacks spent increased amounts of time (Fig. 4). These included continental shelf areas from 30°–50°N and offshore areas centered at 42°N, 65°W, as well as an area between the Cape Verde Islands and the African coast. The average postnesting time over the shelf for all 10 turtles was 63%.
To determine whether there were seasonal effects to the distribution of high-use areas by the turtles we plotted the 25% utilization distributions by month, which illustrates the core high- use areas for these turtles. Turtles remained in coastal areas (or on the continental shelf) for 3 of the seasons and moved to offshore areas in winter (Fig. 5).
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[239:IAPMAF]2.0.CO;2
Rates of travel varied during the time each turtle was tracked (Fig. 6). Six of the 9 turtles increased travel rates rapidly upon leaving the nesting areas and then slowed in areas where they appeared to reside for extended periods. In some cases this pattern of rapid travel punctuated by slow periods was repeated throughout their records. We evaluated travel rates within each of the areas delineated by the 95% and 50% UDs. Mean travel rate within the 50% UDs (25.2 km/d, 19.0 SD, n = 871) was lower than rates in the 95% UDs (48.8 km/d, 29.8 SD, n = 493) areas (Mann-Whitney test, p < 0.0001).
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[239:IAPMAF]2.0.CO;2
DISCUSSION
This study describes the movements of leatherback turtles from nesting beaches in Florida using satellite telemetry. In contrast to other studies that satellite-tracked leatherbacks from nesting beaches, the Florida turtles generally did not move across oceanic environments, and in fact spent a large portion of their time on the continental shelf of North America.
That the turtles resided over the continental shelf for much of the first year after nesting, means that they were within US and Canadian jurisdiction where protection for the species is higher than in international waters, or in some other national jurisdictions. Such protection could conceivably enhance recovery of the Florida nesting population, which is currently growing (Witherington and Koeppel 1999).
By using travel rates and kernel home-range analysis we delineated important high-use areas for Florida nesting leatherbacks. During the internesting period, Florida leatherbacks resided primarily just to the east–southeast of Cape Canaveral and off their nesting beaches bounded to the east by the western edge of the Gulf Stream. It can be inferred that the region east–southeast of Cape Canaveral is a critical internesting area for Florida leatherbacks, because turtles from both the geographically close Melbourne Beach colony and the more distant Juno Beach colony spent much of their internesting period within the area. Management actions or plans for the Florida leatherback nesting population need to take into account this common-use area as well as the waters directly adjacent to the nesting beaches.
At the completion of the nesting season all turtles traveled directly north, with the majority (6 of 10) moving parallel to the coast and west of the Gulf Stream while the others moved within or across the Gulf Stream and eventually off the continental shelf into pelagic waters. For turtles traveling north along the coast, the central coast of South Carolina and Cape Lookout, North Carolina, were 2 areas in which leatherbacks moved especially close to the coast. Generally most turtles moved offshore while moving around Cape Hatteras, North Carolina, but in some cases they would return close to the coast on the north side of the cape, or spend extended periods in that region.
Distinct postnesting high-use or residence areas are apparent from both the kernel estimates and from slowed rates of travel for turtles that moved away from the eastern Florida nesting colonies. It is likely that these are foraging areas. Such foraging areas are delineated in the near-shore coastal waters of Georgia, South Carolina, North Carolina, Virginia, Maryland, and Nova Scotia. Other foraging areas included a region south of Georges Bank centered at about 65°W and 41°N, as well as an area in the central Atlantic centered at 10°N and between the Cape Verde Islands and Mauritania. The latter area has also been the destination for leatherbacks tracked from the southern Caribbean (Eckert 1998, 2006; Ferraroli et al. 2002, 2004) and is known as a region of elevated oceanic productivity called the Mauritania upwelling (Mittelstaedt 1991). Although the Mauritania upwelling likely represents an important foraging area for Atlantic leatherback turtles, for Florida leatherbacks it is clear that North American continental shelf waters are exceptionally important.
The location of western Atlantic foraging areas changed based on the time of year. In spring, summer, and fall, these areas were located on the North American continental shelf for most turtles. However, in winter, foraging areas were located off the continental shelf.
This study also showed areas where Florida leatherbacks may be in contact with coastal fisheries in US Atlantic waters. Although the evidence is circumstantial, it is likely that one of the 10 turtles was killed in the coastal shrimp fishery. DC4 had established residency in an area 60–80 km off the coast of southern Georgia and northern Florida. On 28 October 2001, this turtle moved from its residence area to within 10 km of the coast. On 11 December we lost its signal, which to that point we had consistently received uplinks multiple times per day. To date, DC4 has not been seen again, and between 4 November 2001 and 10 December 2001, 17 leatherbacks were stranded in northeast Florida, representing a dramatic increase in leatherback strandings for this area. The mortality of these turtles was suspected to be because of incidental capture in shrimp trawl fisheries (Federal Register 2001). It is possible that DC4 was among the turtles killed by shrimp fisheries in these near-shore waters, which would explain the sudden cessation of transmissions.
The sudden movement of DC4 toward shore is intriguing because it coincided with a rapid cooling of near-shore waters. Temperature sensors on the PTT carried by DC4 noted that on 27 October she spent more that 80% of her time in waters 30°–32°C. On 29 October the turtle spent more than 80% of her time in waters 28°–30°C, and by 30 October in waters 26°–28°C. Sea surface temperature (SST) satellite imagery (http://coastwatch.noaa.gov/) confirmed that water temperatures near shore were cooler than offshore at this time. Air surveys of the area confirmed both elevated numbers of leatherbacks and jellyfish (Federal Register 2001). It is possible that near-shore oceanographic changes signified by cooler coastal waters also led to increased jellyfish concentrations and attracted DC4 and other leatherbacks to the area.
This observation has significant management implications. If remotely sensed data such as coastal SST imagery can be used to predict the movements of leatherbacks into areas of potential interaction with fisheries, mitigation measures can be implemented rapidly. In this particular example, the unusual stranding event (with 17 leatherbacks) took place between 4 November and 10 December. The US National Marine Fisheries Service issued emergency mitigation regulations effective December 14 (Federal Register 2001) in response to those strandings. However, the movement data of DC4 indicate that it had already moved into near-shore area by 30 October. It can be considered that other leatherbacks were behaving similarly (because they were caught and killed during November). Thus, turtles were exposed to shrimp fishery interactions more than a month before emergency regulations were implemented. If such mitigation regulations had been put into place based on the movements of this telemetered turtle, or by detection of the change in oceanic conditions, instead of waiting for strandings to occur (and to be reported by the stranding network), this mortality event might have been avoided.
Internesting movements of 9 leatherback turtles as determined by satellite telemetry. Twenty-nine intervals were monitored for the 9 turtles who nested more than 1 time (after transmitter deployment). Turtles usually moved north immediately after nesting from the 2 nesting beaches (Archie Carr National Wildlife Refuge and Juno Beach).
Kernel-estimated home-range utilization distributions (KHRE) of internesting locations for 9 female leatherback turtles tracked by satellite telemetry from Archie Carr National Wildlife Refuge and Juno Beach, Florida.
Postnesting movements of 10 leatherback turtles from 2 nesting beaches on the eastern coast of Florida (Archie Carr National Wildlife Refuge and Juno Beach) as determined using satellite telemetry.
Postnesting season kernel-estimated 50% and 95% utilization distributions (UD) for female leatherback turtles DC1, DC2, DC4, DC6, DC7, DC8, and DC 10. Kernel estimate was set to 25% and 95% UD for DC5 because it provided a better fit to this turtle's data. Postnesting tracking duration was too short for DC3 and DC9 to use kernel home-range estimation analysis methods.
Monthly postnesting high-use areas as delineated by the 25% kernel home-range utilization distributions (UDs) for 8 female leatherback turtles. We used the 25% kernel estimate because it delineated the core high-use areas better than a 50% UD. Months were grouped by season to better illustrate seasonal influences on leatherback movements in the North Atlantic Ocean.
Travel rates of 10 leatherback turtles during the duration in which they were tracked by satellite. To ease visualization of the trend in travel rate over time, a Lowess smoothing function (robust locally weighted regression) was fit to the data.
