Study Site and Fieldwork

D'Arros Island (5°24.9′S, 53°17.9′E) is oval-shaped, 150 ha in area, 1.9 km long, and 1.0 km wide, with 5.0 km of coastline comprising virtually unbroken white coralline sandy beach. It hosts a population of 40–50 people. The turtle program has been a collaborative effort between interested Seychellois residents of D'Arros who conduct the beach surveys (JCC, NB, and others) and the senior author (JAM) who initiated, designed, and supervised the study. In the early months, turtle surveys were performed on a volunteer basis but eventually were compensated by over-time pay as work outside normal hours. Ensuring quality control in the data collection posed a challenge given the remote location of D'Arros (255 km from Mahé where JAM was primarily based) and given that the turtle workers had no previous formal scientific training. But this was successfully addressed from the outset by workers carrying a GPS device (Garmin Geko 201, Olathe, KS) during every survey. They used the device to track their own routes as they surveyed the beach and also to record waypoints to document the exact position of each turtle track encountered. The GPS data were regularly downloaded and emailed to JAM, and the data sheets were delivered to Mahé whenever transport was available.

For each turtle track encountered, the following data were recorded: 1) survey date and time; 2) GPS waypoint number; 3) track width in centimeters; 4) number of body pits dug (i.e., excavations made with the front flippers); 5) assessment of whether eggs were laid; and 6) estimated date when track was made. Based on the number of body pits dug and whether or not eggs were laid, each turtle track was assigned to one of the following three behavioral categories described by Mortimer and Bresson (1999): 1) “LAID”, eggs were laid after the turtle dug one or more nests; 2) “Did Not Lay (DNL)”, one or more body pits were dug but no eggs laid; and 3) “No Dig”, digging did not occur. The study was designed to minimize subjectivity in data collection. To that end, species identification was based on track width (with green turtles >100 cm wide). Once recorded, each track was marked by drawing a line through it in the sand at a point above the high tide line. Body pit counts distinguished emergences during which no digging occurred (No Dig) from those that involved digging (LAID and DNL). The criteria used to distinguish LAID from DNL tracks include the characteristics of turtle tracks described by Schroeder and Murphy (1999) but were actually based on the decades of field experience of the senior author (JAM) who provided intensive onsite training to the field personnel. Only after the first 11 months of the study (from October 2005 onward) were the data distinguishing the three track types considered reliable enough to include in the present analysis (see footnotes in Tables 2, 3).

To optimize data quality, we aimed to survey the island at least once a day (ideally twice a day during peak hawksbill season). This was not always possible, but records were kept of all surveys conducted whether or not tracks were encountered.

Data Extrapolation

Data extrapolations were performed for “missed days” when tracks were likely to have disappeared before they could be counted—i.e., when too many consecutive missed days occurred in combination with environmental conditions likely to erase turtle tracks (e.g., spring tides, strong wind, and heavy rain). Fortunately, tracks of both species usually remain visible for at least several days, especially during periods of neap tide; and the tracks of nesting green turtles are particularly resilient given that the adults weigh 3–4 times as much as adult hawksbills (Pritchard and Mortimer 1999). In light of these factors, during the present study, the following three types of track count extrapolations were performed.

First, for periods of up to 4 consecutive missed days, extrapolations were performed only: 1) for green turtles, when 3 or more consecutive survey days were missed during periods of spring tides but not neap tides; and 2) for hawksbills, when 2 or more consecutive survey days were missed regardless of tide. In such cases, extrapolations were made by averaging numbers of tracks recorded daily during the week before and after the period of missed days.

Second, for the 16 consecutive missed days between 23 December 2006 and 7 January 2007, missing data for 23–31 December and for 1–7 January were extrapolated separately by month, based on the following rationale. In the inner islands of Seychelles, Mortimer and Bresson (1999) estimated mean clutch frequency to be 4–5 clutches per season, with an average of 15.1 days between successive clutches. We assumed the same pattern for hawksbills in the Amirantes and extrapolated missing data by calculating mean daily track counts recorded during the period 15 days before and 15 days after the interval of missing data and then averaging the two figures. Hence, extrapolating mean daily nesting activity for the period 23–31 December involved averaging mean daily tracks recorded during 8–16 December (i.e., 15 days before) and during 8–15 January (i.e., 15 days after). Likewise, missing data for 1–7 January were extrapolated from track count data collected during 17–22 December and 16–22 January.

Finally, Seychelles hawksbills show a clearly defined nesting season (Mortimer and Bresson 1999). Therefore, for those two seasons in which surveys began (i.e., 2004–2005) or ended (i.e., 2009–2010) partway through the season, we extrapolated missing track count data based on the distribution of nesting activity recorded during the completely surveyed seasons (i.e., 2005–2006, 2006–2007, 2007–2008, and 2008–2009). For example, during the 2004–2005 season, no surveys were conducted prior to November 2004. Using track count data recorded in 2005–2006, 2006–2007, 2007–2008, and 2008–2009, we calculated what percentage of total nesting emergences occurred during July to October each season, took an average of those four figures, and used that to extrapolate number of nesting emergences during July to October 2004. This type of extrapolation was not attempted for green turtles, which display more variable, and therefore unpredictable, interannual patterns of nesting seasonality.

Data Analysis

Total estimated numbers of tracks for each species were compiled by month and graphed linearly to illustrate both inter- and intra-annual patterns and variation in seasonality. The distribution of nesting between the two species was compared using the following procedure. The day of the year when each nesting emergence occurred (e.g., 1, 2, 3 [in January] … 365 [December 31]) was converted into a corresponding angle (between 1 and 360 degrees) and then summed for each species by the number of observations per angle over the 5-year period of observations. The distribution of summed angles was then analyzed using a Rayleigh test (Zar 1999) to determine whether it differed statistically from uniformity (that is, whether each species showed significant seasonal clustering). Because clustering was apparent, next we determined whether the seasonal pattern differed between hawksbills and green turtles as evidenced by the presence or absence of overlap between the 95% confidence limits to either side of the mean angles shown by the two species. The following parameters of seasonality defined by Lauret-Stepler et al. (2007) were also calculated for both species: 1) “season duration” as the interval that includes 90% of the total season's nesting activity, and for which the endpoints, “onset” and “end”, are the dates on which 5% and 95%, respectively, of nesting activity had occurred; and 2) “season median” as the date on which 50% of nesting activity between the onset and end occurred.

To estimate number of egg clutches produced by each species each season, we multiplied total estimated tracks by the proportion of recorded tracks classified as LAID. Then we estimated annual numbers of nesting females of each species by assuming an average clutch frequency per season of 4–5 for Seychelles hawksbills (Mortimer and Bresson 1999) and 5 for green turtles; the latter taken from saturation tagging data collected at Heron Island Australia (Limpus 2008) because no saturation tagging data are available for green turtles in the western Indian Ocean.

Total Estimated Tracks

During the entire 62-month survey period (between 1 November 2004 and 31 December 2009) the following incidents of consecutive missed days occurred during peak season: 1 day on 266 occasions; 2 days on 51 occasions; 3 days on 17 occasions; 4 days on 6 occasions; and 16 days on 1 occasion. Details by season and species are presented in Table 1. Extrapolations involving 1–4 missed days accounted per season for an average of 5.2 extrapolated green turtle tracks (SD  =  4.9, range  =  1–11, n  =  5) and 25.5 extrapolated hawksbill tracks (SD  =  17.1, range  =  6–54, n  =  6). For the 16 consecutive missed days (23 December 2006 to 7 January 2007), 81 hawksbill and 4 green turtle tracks were extrapolated. During those four hawksbill seasons surveyed year-round (2005–2006 to 2008–2009), average total estimated tracks during July to October was 12.2% (SD  =  3.0, range  =  10.0–16.5, n  =  4), and during January to June, 23.4% (SD  =  4.9, range  =  18.7–30.1, n  =  4). Based on those figures, 64 tracks were extrapolated for July to October 2004, and 118 for January to June 2010. Raw track count data and total estimated tracks for each species and season are presented in Tables 2–3. Estimated mean total annual track counts for green turtles was 179 (range  =  109–257, n  =  5), and for hawksbills, 528 (range  =  505–548, n  =  6).

Table 1 Survey effort between November 2004 and December 2009. For each hawksbill (peak season only) and green turtle (entire year) nesting season the number of surveys conducted is indicated, as well as the total number of missed days (during which no surveys were conducted), and the numbers of consecutive missed days with frequencies indicated by parentheses.

View Fullscreen

Table 2 Assessment of green turtle nesting activity at D'Arros Island during those five years (2004–2005 to 2008–2009) in which surveys were conducted year-round. For each season, the following information is given: (a) raw track count data; (b) total estimated tracks after extrapolation; (c) proportion of each of the three main track types recorded; and estimates of (d) numbers of egg clutches laid and (e) numbers of nesting females per season.

View Fullscreen

Table 3 Assessment of hawksbill nesting activity at D'Arros Island during six nesting seasons between 2004 and 2009. Table indicates: (a) raw track count data; (b) total estimated tracks after extrapolation; (c) proportion of each of the three main track types recorded; (d) estimates of numbers of egg clutches laid; and (e) bracketed estimates of numbers of nesting females per season.

View Fullscreen

Seasonality of Nesting Activity

Figure 2 compares numbers of total estimated tracks made by hawksbill and green turtles at D'Arros Island during each of 62 months from November 2004 through December 2009. Intra-annual patterns of nesting seasonality are depicted by month in Figure 3 for each species and nesting season and also as an average of all seasons combined for each species. For five green turtle seasons (November 2004 to October 2009) and four hawksbill seasons (July 2005 to June 2009), Figure 4 shows total numbers of nesting emergences by day of the year. Green turtles nested during all months with a distinct peak in late March, whereas hawksbills nested almost exclusively between October and February with a distinct peak in early December. Both distributions were significantly different from uniformity (Rayleigh test: green turtle Z  =  19.6; hawksbill Z  =  1050; p < 0.001 for both species). Relative lengths of the r-vectors indicated much greater variability around the mean for green turtles (r  =  0.17) than for hawksbills (r  =  0.85). The two nesting distributions differed significantly from one another as there was no overlap between the 95% confidence intervals of the means (66.5°–102.1° for green turtles, 333.7°–337.0° for hawksbills; Fig. 4). The mean “season duration” and “season median” calculated were, for green turtles, 310 days (SD  =  16; n  =  5) and 24 April ± 18 days and, for hawksbills, 104 days (SD  =  14.1, n  =  4) and 7 December ± 2.6 days.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Nesting Success and Population Size

The frequency of track types recorded (from October 2005 onward) in our study at D'Arros was 0.32 No Dig, 0.15 DNL, and 0.53 LAID for green turtles and 0.35 No Dig, 0.08 DNL, and 0.57 LAID for hawksbills. For those tracks that involved digging behaviour (DNL and LAID) the average number of body pits recorded was 1.31 (SD  =  0.64, range  =  1–5, n  =  486) for green turtles and 1.15 (SD  =  0.44, range  =  1–6, n  =  1459) for hawksbills.

Estimated average numbers of egg clutches laid annually were 95 for green turtles (range  =  65–120, n  =  5) and 300 for hawksbills (range  =  277–318, n  =  6). In terms of average numbers of females nesting annually, this would be equivalent to 19 green turtles (range  =  13–24, n  =  5), assuming an average of 5 clutches per female per season (Limpus 2008) and 60–75 hawksbills (range  =  56–80, n  =  6), assuming an average of 4–5 clutches per season (Mortimer and Bresson 1999).

Nesting Seasonality

Our study is the first to document the seasonal patterns of nesting activity for sea turtles in the Amirantes Group of Seychelles. Like the hawksbill season of the inner islands of Seychelles (Mortimer and Bresson 1999) that at D'Arros Island is remarkably consistent from year to year, with a peak during the months of October through February. At D'Arros, during the four completely monitored hawksbill seasons (2005–2006 to 2008–2009), 94.0% (SD  =  3.4, range  =  89.2–97.1) of nesting emergences were recorded between 1 October and 31 January compared to 88.0% during the same period at Cousin Island (Mortimer and Bresson 1999). The peak coincides with the northwest monsoon rainy season in both the inner islands (Mortimer and Bresson 1999) and the Amirantes. During the 5 green turtle seasons at D'Arros Island for which year-round data are available (2005 to 2009), nesting showed a bimodal pattern with a dominant peak during February to April and a secondary peak in July and August. But seasonal peaks were inconsistent from year to year, and in any given year nesting could be either trimodal, bimodal or unimodal (Figs. 2, 3).

Seasonality at D'Arros (05°25′S, 53°18′E) needs to be considered in the context of patterns documented for green turtle rookeries elsewhere in the western Indian Ocean region—i.e., Aldabra (09°24′S, 46°20′E), Glorieuse (11°33′S, 47°17′E), Mayotte (12°59′S, 45°11′E), Tromelin (15°33′S, 54°31′E), and Europa (22°21′S, 40°21′E) islands. Although nesting occurs year-round at all of these islands, peak nesting occurs in different months of the year depending on locality. Like D'Arros, Aldabra atoll shows inconsistent bimodal nesting peaks during the months of February to May and May to September (most dominant) (Gibson 1979; Frazier 1984; Mortimer 1988; J.A. Mortimer and Seychelles Islands Foundation, unpubl. data). Nesting peaks reported elsewhere include those at Glorieuse, February to July (Lauret-Stepler et al. 2007); Mayotte, April to September (Bourjea et al. 2007); and Tromelin and Europa, November to February (Lauret-Stepler et al. 2007). These data indicate that peak nesting occurs later in the year at rookeries located farther south of the equator.

Population Size and Trends

Historical records indicate that large numbers of green turtles once nested in the Amirantes. In 1770, the log of L'Heuve du Berge reported that the men came ashore at night on an island in the Amirantes, probably Poivre (Fig. 1), and took 32 turtles (Fauvel 1980). These would have been green turtles because hawksbills nest diurnally in Seychelles (Mortimer and Bresson 1999). In contrast, in the early 1980s, Mortimer (1984a) estimated that fewer than 30 female green turtles nested annually at any island in the Amirantes Group. During the 19th century, hawksbills were heavily hunted for their shell throughout Seychelles. According to Dupont (1929), in about 1850, an unnamed individual acquired Desroches, Poivre, and D'Arros Islands principally for the capture of hawksbills, and when the supply of hawksbills decreased, he sold all three Amirantes islands.

Frazier (1984), based largely on his own observations made during brief visits, estimated that in the early 1970s a maximum of 10 hawksbill and 5 green turtle females nested annually at D'Arros Island and St. Joseph atoll combined. This contradicts information obtained by J.A. Mortimer (unpubl. data) in the early 1980s during interviews with fishermen, turtle hunters, and laborers who had worked at D'Arros and St. Joseph in the 1970s, and who claimed that significantly more turtles nested in the 1970s but that most were killed. Therefore, Frazier may have underestimated nesting numbers because he came at a time when most females were being slaughtered early in the nesting season and did not leave tracks on the beach. The situation for turtles improved somewhat in the mid-1970s when the current management of D'Arros began to actively discourage turtle exploitation. Although these efforts undoubtedly saved some turtles, complete protection was impossible at that time given the isolation of the islands, the large expanse of territory involved, and the fact that no national legislation protected turtles at D'Arros/St. Joseph until 1994 when the Seychelles Government enacted the Wild Animals (Turtles) Protection Regulations (1994), which banned all killing of sea turtles nationwide. In the early 1980s based on track counts and interviews, Mortimer (1984a) estimated that 25–40 hawksbills and 5–20 green turtles nested annually on D'Arros and reported that they were being hunted. But Mortimer (like Frazier) may also have underestimated the size of the nesting population if her informants underreported numbers of turtles killed. When there is intense exploitation of nesting females, track count surveys provide an inadequate index of population size. In the early 1980s, during peak hawksbill season, Mortimer conducted 6 surveys at D'Arros: on 19 and 24 November 1982; 5 December 1982; and 9, 19, and 23 December 1983. Her data indicated a total of 20 tracks produced over a period of 19 days (i.e., 1.1 tracks per day). In another survey conducted 13 years later, on 17 November 1996, her data indicated 9 tracks over 7 days (i.e., 1.3 tracks per day) but also included evidence of ongoing turtle exploitation (Mortimer 1998). In contrast, during 2004–2009, at the same time of year (11 November to 23 December), we recorded an average of 6.2 tracks per day (SD  =  3.6, range  =  0–19, n  =  244). Thus, regardless of whether the nesting population has increased, decreased, or remained stable since the 1980s, reproductive activity has certainly increased with implementation of protective measures under the DRC turtle conservation program. The resulting enhanced reproductive output can be expected to manifest itself as an increase in nesting numbers within the next 25 years, once the large numbers of hatchlings produced in recent years have attained adulthood (Mortimer and Donnelly 2008). Such recovery of a previously overexploited population has been documented at Cousin Island where turtles have enjoyed protection since 1968 (Mortimer and Bresson 1994; Allen et al. 2010).

Our data show that hawksbill nesting numbers have been relatively consistent over the six seasons monitored (Table 3, Fig. 2), whereas green turtles show more interannual variation (Table 2, Fig. 2). Both populations appear to have been stable since 2004, but long-term monitoring will be needed to ascertain population trends. We are confident in the accuracy of our estimates of total tracks and also of our No Dig track counts. However, a possible source of error is that some DNL tracks may have been mistaken for LAID tracks, in which case our figures for actual numbers of egg clutches laid and annual numbers of nesting females would be inflated.

Conservation Issues and the Way Forward

The 1994 legislation, which outlawed killing of sea turtles and eggs and prohibited trade in sea turtle products, was of particular benefit to hawksbills. The government reinforced the law with a World Bank/GEF sponsored Artisan Training and Compensation Programme (Ministry of Foreign Affairs Planning and Environment 1994) and by destroying its stockpile of raw shell (Mortimer 1999), thereby effectively ending the domestic trade and the legal export of hawksbill shell (tortoiseshell). Nevertheless, demand for green turtle meat continued after 1994 along with the exploitation of green turtles, albeit at much reduced levels from that documented in the early 1980s (Mortimer 1984). Green turtle nesting populations have almost disappeared from the inner islands where > 99% of the human population resides (Mortimer 1984, 2004); therefore, hawksbills are sometimes killed there for their meat, which is salted to pass as green turtle meat. Meanwhile, the Amirantes Group located only 250–400 km from Mahé provides a more proximate source of green turtle meat for poachers than do the larger green turtle populations (Mortimer 1984) of the Southern Islands (some 750–1000 km southwest of Mahé). Evidence of systematic turtle poaching at D'Arros /St. Joseph was apparent during the years after the legislation passed (Mortimer 1998) and very occasionally in recent years. But during the 5 years that our turtle conservation and monitoring program has been underway, turtle poaching has virtually ceased at D'Arros/St. Joseph.

Elsewhere in the Amirantes Group, programs similar to ours have been implemented by the Island Conservation Society (ICS), which is the NGO with a mandate to manage conservation programs in the government-owned outer islands. The ICS initiated turtle conservation and management programs at Alphonse, Bijoutier, and St. Francois Islands in 2006, and at Desroches Island in 2009; and both of these ICS programs are ongoing. The DRC and ICS have agreed on a Memorandum of Understanding to promote collaboration and information exchange among the various islands of the Amirantes Group in the realm of conservation and research. We recommend that the D'Arros program continues in its present form but with a greater emphasis on tagging nesting females, which would provide data to further refine our population estimates and also help us document interisland movements, thus demonstrating the extent to which nesting turtles are a resource shared among the islands of the Amirantes Group and beyond.