Modeling Approaches to Quantify Leatherback Nesting Trends in French Guiana and Suriname
ABSTRACT
Nesting of leatherbacks in French Guiana and Suriname has been monitored for more than 30 years. Given the documented exchange of tagged females across the nesting beaches, leatherbacks found on the 6 principal nesting areas in French Guiana and Suriname are considered to be a single large nesting population. Despite more than 3 decades of work on this population, this population's status remains unclear. Here we describe the most recently available estimates of various life history parameters and describe the trend of the number of nests laid in the region over the past 36 years. Our analyses suggest that the trend of the whole population in French Guiana and Suriname is stable or slightly increasing over this time period. We strongly urge the continued monitoring of the population by the different research groups in the region so that future data sets will enable an accurate description of the status of this leatherback population.
The leatherback turtle, Dermochelys coriacea, is the most abundant of the 4 species of marine turtles that nest in French Guiana and Suriname. The other 3 species, in descending order of numbers of nests laid in the region, are the green turtle (Chelonia mydas), the olive ridley turtle (Lepidochelys olivacea), and the hawksbill turtle (Eretmochelys imbricata). Tagging data for leatherbacks have been collected since the early 1980s in the Guianas and have shown that exchanges occur between and among beaches in French Guiana and Suriname (Chevalier et al. 1998). This suggests that females nesting in these 2 countries belong to the same population. French Guiana and Suriname beaches are the largest remaining nesting grounds for leatherbacks in the world and host ca. 40% of the global population of breeding leatherback females (Spotila et al. 1996).
Despite observations and reports of large numbers of leatherbacks in French Guiana and Suriname over the past 30 years, leatherbacks are mentioned infrequently in literature published before 1950. Many historical authors have described the marine fauna of French Guiana but leatherbacks appeared for the first time in discussions of the region only in 1664 (Biet 1664). The next description of this turtle species was by Goupy des Marets in 1675 (Goupy des Marets 1675–1676, 1687–1690), who probably copied Biet's descriptions. Another 50 years passed before they were reported again, by Barrère (1741). In 1774, leatherbacks were described as being seen near the Kourou River estuary in French Guiana in a document that codified the regulation of marine turtles fishing activity (Anonymous 1774). The next historical document available with information about leatherbacks in the Maroni/Marowijne River region was published by Kappler (1881), nearly 100 years later. The 1941 and 1945 fisheries reports of Diemont and Geijskes (Schulz 1971) contained a number of observations on leatherbacks nesting in Suriname. Finally, the main nesting beaches were “discovered” in the late 1960s, and population surveys have been conducted on these beaches since then (Pritchard 1969; Schulz 1971). Daily nest counts on most beaches in Suriname have been conducted by Stinasu from 1969 to now. It is interesting to mention that Stedman (1796) wrote that in Suriname “the turtles are . . . generally distinguished by the names of calipee or green turtle, and carett” (this latter name refers to olive ridley turtle, Schulz 1971); he did not mention specifically the presence of leatherbacks.
Since 1967, work in this region has concentrated on counting nests (since 2001 combined with counting females on the beach) in Suriname and counting nests and female leatherbacks on the beaches in French Guiana; numerous reports and publications deal with these data (Pritchard 1969; Pritchard 1971; Schulz 1971; Schulz 1975, 1982; Reichart and Fretey 1993; Chevalier et al. 1998; Chevalier and Girondot 1998a; Chevalier and Girondot 1998b; Chevalier et al. 1999a; Hilterman 2001; Schouten et al. 2001; Hilterman and Goverse 2002, 2003; Girondot et al. 2006). However, some older reports lack a clear description of the methodology used to collect the data, the exact temporal window used (both in terms of hours and dates the beaches were patrolled), and the extent of the sampled area. Therefore, the quality of data varies greatly among sites and years. An initial attempt to analyze trends revealed a strong correlation between the number of nests observed on beaches in French Guiana and Suriname, near the Maroni estuary (Chevalier and Girondot 1998b), and demonstrated a temporal tendency for a higher proportion of nests laid in eastern Suriname compared with central Suriname (Chevalier and Girondot 1998b).
There are at least 2 methods available to describe the status of a marine turtle population. The first is to use a population dynamics model (Spotila et al. 1996), and the second is to do trend analysis, incorporating current and historical records of the number of nests or females (Spotila et al. 2000). We will discuss both strategies as applied to leatherbacks in French Guiana and Suriname to estimate the trend of the nesting population in this region over the last 36 years.
Model Parameters
Nesting Area
Nesting by leatherbacks regularly occurs from the eastern end of the coast of French Guiana around Cayenne and extends westward to Braamspunt at the head of the confluence of the Commewijne and Suriname Rivers in Suriname. It encompasses approximately 600 km of coastline, with highly dynamic beaches that appear or disappear according to the displacement of large mud bank from Amazonian origin (Marchand 2003). The principal nesting areas are generally separated into 5 zones: the Matapica area in the central section of coastal Suriname, the Galibi Natural Reserve area in eastern Suriname near the border with French Guiana (Babunsanti and remote nesting beaches), the French beaches in and near the estuary of the Maroni and Mana Rivers (Yalimapo-Awala and Pointe Kawana), the western oceanic beaches of French Guiana (from Pointe-Isère nesting beach), and the Kourou and Cayenne area in eastern French Guiana (Fig. 1). Further east, nesting of leatherbacks in the extreme north of Brazil was reported in Cabo Orange National Park (Roberto 2004), although a recent survey by one of us (LP) found no sandy beaches in this region. Further west, the closest known nesting beach for leatherbacks is Shell Beach in Guyana. The relation between the nesting populations in Guyana, Trinidad, and Suriname/French Guiana is unclear. To date, there have been 2 exchanges of nesting females from Guyana to nesting beaches in Suriname and 3 exchanges of nesting females between the nesting beaches of French Guiana/Suriname and Trinidad, located northwest of Guyana (Hilterman and Goverse 2003).
Citation: Chelonian Conservation and Biology 6, 1; 10.2744/1071-8443(2007)6[37:MATQLN]2.0.CO;2
Nesting Seasons
In French Guiana, nesting leatherbacks have been shown to exhibit a bimodal nesting season, with a brief nesting season in December and January (“small nesting season”) and a long one from March to August (“normal nesting season”) (Chevalier et al. 1999c). Limited data suggest that nesting females are not segregated by season (Table 1). The scarcity of these observations reflects that field workers are generally not present on the beaches in December and January; to date, only 20 tagged females have been observed during the “small nesting season” compared with a total of 28,986 identified females for the normal nesting season in French Guiana (Chevalier et al. 1999c) (note that a same female can appear several times in the database because of tag loss).
Number of Nests per Female per Nesting Season
During the normal nesting season in French Guiana, females have been observed 1–11 times (note that not all females nesting in the region are observed each time they lay eggs on the beach). Females nest every 9 or 10 days. This repeatability can be used to estimate by interpolation the total number of nests laid between 2 extreme observation dates and generates a maximum estimate of 14 nests laid in 1 year. However, this methodology underestimates the true number of nesting events by females, because it implicitly assumes that the first observation of a female during the season is its first nest of the season. However, it is possible that earlier nesting events may not have been observed, and a similar situation exists for the last observed nest. To account for this, stopover methodology can be used, because it uses the total duration an animal remains in a given place to estimate the missing observations outside the first and last records (Schaub et al. 2001). By using stopover methodology, the mean total number of nests per female during the nesting season is calculated to be 8.3 ± 0.9 standard error (SE), with significant interannual variation (Rivalan 2003). Note that females seen only one time during the nesting season are not taken into account for this analysis. The significance of these “one time nesters” appears to be more complex than simply reflecting a low probability of observation (see below and also Hilterman and Goverse 2007).
Loyalty of Females to a Nesting Beach
When tag data from several beaches are analyzed together, it appears that females may switch nesting beaches from season to season. However, females are generally faithful to one nesting beach during a nesting season and only rarely move to another nesting beach for a single laying event (Table 2). Therefore, the females observed nesting only once during the nesting season in Yalimapo Beach could be “unfaithful visitors” rather than true “one time nesters”. However, this hypothesis remains untested. An alternative hypothesis is that the females observed only once may be new recruits in the population (Hilterman and Goverse 2007).
The loyalty of female leatherbacks can also be investigated on the scale of different sections along a beach during consecutive nesting events. During the 1997 nesting season, we separated Yalimapo Beach into 2 sections of similar length (2 km): Awala in the east and Vigie in the west. We recorded the section in which a single female nested during 2 successive observations with less than 18 days between them, to ensure that only true successive nesting events were used. Each female was included only once in the analysis to ensure independence between observations. Of 473 pairs of successive observations, 89 females nested twice in a row in the Awala section, 180 females nested twice in a row in the Vigie section and 204 females nested once in Awala and once in Vigie. Based on these values, the probability of nesting in Awala was pA = 0.40 and in Vigie was pV = 0.60. If the choice of section was independent of the previous nesting section, the expected number of successive nesting events in Awala would be N.pA.pA = 77.12, in Vigie, it would be N.pV.pV = 168.12, and for switching between sections would be 2.N. pV.pA = 227.74. The difference between observed and expected is significant (χ2 = 5.14, df = 1, p = 0.02), reflecting a greater likelihood of consecutive nesting in the same section. Note that this effect disappeared if nest locations of the first and third nest of individual females were analyzed without information on the location of the second nest (n = 280, χ2 = 0.003, df = 1, p > 0.05). Therefore, although there appears to be some preference for beach section for successive nests, nest site selection across the nesting season is more complex.
Number of Years Between Nesting Seasons and Female Annual Survivorship
It is commonly reported that individual females of most sea turtle species, including leatherbacks, do not nest every year. This character of reproductive behavior of sea turtles complicates analyses of demographic variables, e.g., population size (Hays 2000). Moreover, female mortality outside the nesting season can bias estimates of the time females spend away from the nesting beach. A new capture-mark-recapture (CMR) methodology has been recently designed to more accurately describe this behavior. Indeed, in classical Cormack-Jolly-Seber CMR analyses, one of the implicit assumptions is that the probability of observing an individual will be independent of the individual's history. For leatherbacks and, more generally, for most marine turtles, this assumption is violated, because the probability that an individual will nest 2 years in a row is low but not zero (Girondot and Fretey 1996). The new CMR methodology is free of this assumption and was applied to leatherbacks. The majority of females return to nest in French Guiana 2 years after their prior nesting season, and the annual survivorship of adult females is at least 0.96 (Rivalan 2003).
Hatching Success and Sex Ratio
Hatching success (the number of eggs in a clutch that produce viable hatchlings that reach the water) for leatherbacks in French Guiana and Suriname is difficult to assess when nesting density is high, making it logistically challenging to follow the fate of all nests laid (Caut et al. 2006b). It is known that some leatherback nests on beaches in Suriname and French Guiana were lost to erosion (Mrosovsky 1983, 1997). At Yalimapo Beach, hatching success in 2001 was estimated for 48 nests to lie between 33.27% (3.37 SE) and 38.95% (3.51 SE) (Torres 2002), 35.9% (7.1 SE) for 10 nests laid in 2001 and 2002 (Maros et al. 2003) and 35.5% (1.9 SE) for 99 nests in 2002 (Caut et al. 2006a). Data from Suriname sometimes include the proportion of nests that were not found at the end of the incubation. These nests could be completely lost or position could have been lost. Then, we have estimated the range of possible values according to these 2 hypotheses, but the standard error cannot be calculated, because the raw data are not available. In 2000, hatching success of randomly marked nests, including that of nonemergent nests was 37.48% and 40.35%, respectively, at Samsambo and Matapica Beaches (Hilterman 2001). However, these data should be used with caution, because standardized methodology subsequently used was not still implemented in 2000 (M. Hilterman, pers. comm.). In 2001, hatching success was between 9.15% and 10.6% at Galibi Beach within the Maroni estuary, and between 38.9% and 52.7% in Matapica Beach (Hilterman and Goverse 2002). For this year, erosion was very low on the worked beach section and the missing nests were probably lost by the way of marking (M. Hilterman, pers. comm.). In 2002, the hatching success was 56% (2.96 SE) at Matapica and 25.8% (4.93 SE) at Babunsanti (Hilterman and Goverse 2003). Data are summarized in Fig. 2 and older studies on beaches in Suriname give a similar range of values for leatherback clutches, despite the differences in methodologies (Whitmore and Dutton 1985; Schouten et al. 1997).
Citation: Chelonian Conservation and Biology 6, 1; 10.2744/1071-8443(2007)6[37:MATQLN]2.0.CO;2
Leatherbacks, as all other species of marine turtles, exhibit temperature-dependent sexual differentiation, with warmer incubation temperatures producing more females and cooler temperatures producing more males (Rimblot et al. 1985; Wibbels 2003). The pivotal temperature (the constant incubation temperature that produces equal number of both sexes) for leatherbacks in French Guiana is close to 29.5° C (Chevalier et al. 1999b).
Quantification of Nesting
We used a mathematical model to render the global shape of nesting season, which, in turn, allowed us to estimate the missing nest count data and also to calculate the total number of nests laid during the nesting season, together with its standard error (Girondot et al. 2006).
Nesting seasons of marine turtles typically show a peak of nesting at the approximate middle of the nesting season. The number of nests at the extreme ends of the nesting season is usually low, generally less than 1 nest per week or even month in some cases. This type of pattern can be modeled by using the product of 2 sigmoid equations, the first one ranges from 0 to 1 and the second one ranges from 1 to 0. Therefore, the product of the equations describes a 0-1-0 pattern, if the transition of the first equation is observed at lower abscissa than the second. For the sigmoid equations, we used a modified form of the classical Verhulst equation (Verhulst 1846) that allows asymmetry to be set. The first-order derivative of this equation is similar to the Richards equation (Richards 1959):
The value of M(d) ranges from 0 to 1 with M(d) = 0.5 for P = d, d being the number of days since the starting date of the nesting season. The steepness of M(d) at P = d depends on S and K values. The value of M(d) increases when S is negative (i.e., beginning of the nesting season) and decreases when S is positive (i.e., end of the nesting season). Asymmetry around P is determined by a positive or negative K value.
The mathematical description of nesting season is therefore described as:
With M1(d) and M2(d) being different according to the sign of the S parameter (by convention, 1 is used for beginning of nesting season and 2 for the end). Thus, S1 is negative and S2 is positive and P1 < P2.
The parameter min is the basal level of nesting outside the nesting season and max − min is a scaling factor. Note that max is not the maximum of the function because (M1(d) · M2 (d)) can be lower than 1 at the peak of nesting season. The maximum can only be calculated numerically.
The curve was fitted to experimental data by the maximum likelihood method. For this purpose, we assumed that the error associated with day d was normally distributed with a standard deviation, σ = Exp(a.N′(d)c + b), where a, b and c are parameters that were also fitted. This function has the advantage of being strictly positive and monotonically increasing according to N′(d) for positive values of a and c. It also takes into account the observed heteroskedasticity (i.e., counts that are more dispersed at the peak of nesting season). Goodness of fit of the model was evaluated by using the determination coefficient (r2) between observed and estimated nest number.
Measuring the Nesting Trend in French Guiana and Suriname
The nesting peaks for different seasons varied slightly (Fig. 3). Nest counts corrected by using this methodology were coded with a quality index of 1, whereas nests counts taken from the literature that had no reported precision concerning the method of data collection were coded with a quality index of 0.6. It is important to note that this quality index is a relative value that measures the quality of data and, therefore, a value of 1 does not indicate that the number of nests is known without error. A temporal trend in the relative proportions of nests in eastern and central Suriname has been already demonstrated (Chevalier and Girondot 1998b). We expanded on this finding with more recent data and modeled this proportion by using a logistic model with the year as a cofactor. We used least-square criteria between the angular transformations (2. Asin(p0.5)) of the logistic model and the proportion of nests at eastern Suriname relative to the total nest number in Suriname (Fig. 4a). Similarly, we expanded upon the relation between nest counts in Yalimapo and eastern Suriname (Babunsanti in Fig. 1), as previously described in Chevalier and Girondot (1998b) (Fig. 4b). However, whereas this relation was still observed when using general linear modeling, the 3 most recent points were clearly outliers based on Tukey's biweight (Press et al. 1992). Therefore, the strong relation between Yalimapo and eastern Suriname was not observed for the data from the most recent years and the fitted equation (nests at Yalimapo = 5.69 nests at Babunsanti) could not be used after 1999. This recent change in the relation is probably related to the development of new large areas of nesting habitat at the western edge of this region (Kolukumbo and Samsambo Beaches, see the Suriname Marowijne estuary–oceanic beaches column in Appendix 1). The determination coefficient, r2, multiplied by the quality of the reference count was then used as an index of quality for the estimates generated from this relation. For example, if we used the second relation (r2 = 0.67) with nest counts at Babunsanti with a quality index of 0.75, the result would be coded with 0.67 × 0.75 = 0.5025 quality index.
Citation: Chelonian Conservation and Biology 6, 1; 10.2744/1071-8443(2007)6[37:MATQLN]2.0.CO;2
Citation: Chelonian Conservation and Biology 6, 1; 10.2744/1071-8443(2007)6[37:MATQLN]2.0.CO;2
For the beaches in and around Irakumpapy, there is a lack of strong historical information. Partial information is available in various unpublished reports and one of us (MG) has visited these beaches each year since 1985. However, the available information is not extensive enough to precisely establish the nest numbers for each entire nesting season. Instead, by using available information, we constructed the most plausible temporal series on these beaches but assigned a quality index of 0. A similar situation exists for beaches in the Cayenne region. Whereas, nesting occurred in this region before 1984, all the sandy beaches there disappeared from 1985 until 1988 when new beach reappeared in the eastern French Guiana. At that time, nests were regularly reported to one of us (MG) in this region, as well as in Kourou. Nesting activity has been actively recorded from 1999 in eastern French Guiana. However, only partial information is available and the quality indices from these estimates were coded as 0.5. In 2002, nightly nest counts were available for all principal nesting beaches of Suriname (De Dijn, pers. comm.), except for Galibi area where data were partly based on nightly observations of PIT (passive integrated transponder)-tagged females on the beaches (Hilterman and Goverse 2003). By using these data, it was possible to correct for partial information on all the beaches. Therefore, the quality indices for the 6 main regions were coded as 1 for 2002. The synthesis of the available nest data is summarized in Appendix 1. The quality index for total nests laid per year is the sum of confidence indices for each of the 6 main nesting areas (Fig. 5a).
Citation: Chelonian Conservation and Biology 6, 1; 10.2744/1071-8443(2007)6[37:MATQLN]2.0.CO;2
Model Application
Population Dynamics Model
The life history parameters discussed above can be used to create a population dynamics model (Spotila et al. 1996). However, the lack of detail for specific variables only serves to reduce the confidence of the results of this exercise. For example, the life histories of juveniles and subadults are completely unknown. Also, the variation of many of the estimates of the parameters discussed the previous sections is unknown but could have a major impact on the outcome of a population dynamic model, at least enough to warrant their consideration. For example, previous studies have suggested that for sea turtles, large juveniles and subadults survivorship are the main factors driving population dynamics (Heppell 1998), whereas, the adults possess the highest reproductive value per individual (Laurent 1993). However, if adult survivorship is nearly constant over time while hatchling success is highly variable, as might be expected, then population dynamics will also be influenced by the latter factor (Gaillard et al. 1998). Finally, the covariation or compensation among the different life-history parameters is largely unknown. For example, the relation among nest density, hatching success, and primary sex ratio production as described above may be only one of many complex relations among life-history characters (Girondot et al. 2002). In particular, there appears to be a positive relation between number of nests laid in a season and the number of years that females remained away from the nesting beach before this (Rivalan 2003). This effect is probably mediated by the amount of energy a female can allocate to future reproduction. We also expect non-independence between leatherback survivorship and the number of years between 2 nesting seasons, if the risks are different while turtles are in front of or near the nesting beach, where there are concentrated driftnet fisheries (Ferraroli 2003), and while they are in other areas of the ocean (Ferraroli et al. 2004). However, data are lacking to adequately test these relations.
In general, we have little information on factors that are essential to the construction of an efficient population dynamics model for leatherbacks. In the meantime, only one option remains to describe the status of the nesting population in French Guiana and Suriname: the description of trends of either females or their nests over time. Although the number of females is a preferred measure of population size, it is not directly measurable in Suriname or French Guiana. To estimate the number of females nesting per year on a scale greater than 10 years, we would require the use of a CMR model that can take into account transient nonrandom emigration, multisite modeling and tag-loss correction (Rivalan et al. 2005). Given that such a model is not currently available, we must focus our effort on analyzing the trend of a variable that is measurable and for which data exist: the number of nests laid per year.
Trend of Leatherback Nesting Activity in French Guiana and Suriname
The growth rate calculated from these 36 years is positive (r = 1.8 × 10−2, Fig. 5b). One thousand series have been generated by parametric bootstrapping, where the probability of the occurrence of a year was set as proportional to the quality index of that year. For each series, the growth rate was calculated (Fig. 5c) and was always positive. Therefore, by using these trend analyses, nesting seems to increasing at a low rate in French Guiana and Suriname. However, given the uncertainty of many of the data used to construct this temporal series, a more conservative conclusion is that leatherback nesting has been stable in French Guiana and Suriname over the past 36 years.
Conclusions
By using the largest amount of data available to date to understand the nesting trends of leatherbacks in French Guiana and Suriname, we found that nesting activity for this species is stable or slightly increasing. This conclusion is in contrast with previous studies on this species in this region (Chevalier et al. 1999a). We suggest that the recent discrepancies in describing the status of this population of leatherbacks is related to various forces, including the following: a) most sea turtle populations are automatically assumed to be “endangered” even in the face of contrary evidence (Mrosovsky 2003; Hays 2004); b) too much attention has been given to the exceptional nesting years of 1988 and 1992, which have then been inappropriately used as baselines for the population; c) the general focus of research in French Guiana and Suriname has been to uncover the cause of an assumed population decline (Chevalier et al. 1999a; Girondot et al. 2002; Ferraroli et al. 2004). Given the dramatic decline of leatherback populations in Pacific Ocean (Spotila et al. 2000), we agree that caution should be used when assessing the status of leatherback populations in the Atlantic Ocean. However, in the case of the leatherback nesting population in French Guiana and Suriname, much information remains to be collected and reported before we can fully understand the trend of this species. Therefore, we urge all local associations and organizations working on sea turtles in the region to share data for comprehensive analyses, because, as seen in this review, it is a necessary and essential step toward generating a realistic and representative view of the trend for leatherbacks in Suriname and French Guiana.
Map of French Guiana and Suriname with the principal nesting areas highlighted (see text).
Hatching success in French Guiana and Suriname according to the year of experiment. Matapica and Samsambo are Atlantic Surinamese nesting beaches, Babunsanti and Yalimapo are nesting beaches within Maroni estuary. Error bars are SE, but it cannot be calculated for 2000 season.
Example of fitted daily nest counts during leatherback nesting season from Yalimapo Beach. Envelopes are ±2 SD.
Relations between the number of nests deposited on different beaches in French Guiana and Suriname. A: The temporal change in the proportion of nests deposited in Babunsanti region (eastern Suriname) relative to the total number deposited in Suriname. B: Correlation between number of nests deposited on Babunsanti region (eastern Suriname) and those laid on French Guiana beaches within the Maroni estuary. Data from 2000 to 2002 were excluded as outliers (see text).
Leatherback nest counts in French Guiana and Suriname per year (B) with the quality index (range 0–6) associated with each estimate (A). Rate of population growth is given as r and the finite rate of increase is given as λ. In panel C, the growth rates (r) were derived from bootstrapping (1000 repetitions), while taking into account the quality index for each year.
