Introduction

The green sea turtle (Chelonia mydas) faces multiple threats from human activities, including global warming, habitat loss, incidental capture in fishing gear and overharvesting (Donlan et al. 2010), and is currently listed as ‘Endangered’ on the International Union for the Conservation of Nature (IUCN) Red List of Threatened Species (Seminoff 2004). Green turtles have unique life cycles, spending the majority of their lives in shallow coastal areas foraging primarily on seagrass and algae (Bjorndal 1997), although more recent studies indicate significant omnivory in some populations (Esteban et al. 2020). They recruit to these foraging grounds at a mean size ranging from 40 to 50 cm CCL (Limpus 2008, Piovano et al. 2020), and show strong fidelity to very narrow foraging ranges as they grow to maturity (Makowski et al. 2006; Seminoff et al. 2002; Shimada et al. 2016). Even when green turtles begin breeding, which may involve long-distance migrations to breeding and nesting areas (Hays and Scott 2013; Luschi et al. 1996; Read et al. 2014), they return to their home foraging ground between breeding seasons (Shimada et al. 2020). Understanding the foraging ecology of green turtles in these habitats will contribute for implementing an effective conservation strategy. Whilst early studies on sea turtle foraging ecology were limited to investigating stomach contents of deceased animals (Bjorndal 1985; Garnett et al. 1985; Mortimer 1981; Ross 1985), more recent development of minimally invasive techniques, such as skin sampling for stable isotope analyses, have allowed for more hypothesis-driven and comprehensive assessments of green turtle foraging ecology.

Stable isotope analysis (SIA) involves the measurement of the ratio of different isotopes of the same element, which change in a predictable way as elements cycle through the environment (Peterson and Fry 1987). SIA, typically of carbon, nitrogen and sulphur, can be performed on plant and animal (e.g. skin) samples to investigate the interconnectedness of ecosystem components. Generally, the stable isotope ratios of these elements in consumers reflect the isotope ratios of their prey items, which can be traced down to the baseline isotope value of primary producers, adjusting for trophic discrimination, or enrichment between trophic levels (Fry 2006; Peterson and Fry 1987). Further, ratios of nitrogen isotopes can be used to predict the trophic position of consumers as they are enriched, via metabolic processes, between trophic levels in a predictable way (DeNiro and Epstein 1978; Hobson 1999). SIA has been used widely to study green turtle foraging ecology, including spatial and temporal differences in diet composition (Arthur et al. 2009; Burkholder et al. 2011; Lemons et al. 2011; Williams et al. 2014), ontogenic diet shifts (Arthur et al. 2008; Cardona et al. 2009; Reich et al. 2007), intrapopulation variation in feeding habitat (Hatase et al. 2006; Vander Zanden et al. 2013), and trophic niches (Di Beneditto et al. 2016).

Despite widespread use of SIA to investigate foraging ecology of C. mydas, very little is known about the foraging ecology of this species in New Caledonian waters. The Grand Lagon Sud is considered one of the major feeding grounds for C. mydas in New Caledonia (Read et al. 2020). This foraging ground supports adult green turtles that nest in locations such as the d’Entrecasteaux atolls (Chimmo 1856; Pritchard 1982), and the southern Great Barrier Reef (Limpus et al. 1992), as well as significant numbers of juvenile and sub-adult green turtles (Read et al. 2020). The Grand Lagon Sud also supports an independent genetic stock of nesting green turtles (Dethmers et al. 2006), including turtles that migrate from foraging grounds as far away as Australia (Dethmers et al. 2010; Jensen 2010; Read et al. 2014).

This study used SIA analyses of free ranging green turtles, complemented with the opportunistic examination of stomach contents of green turtles harvested by Indigenous communities for tribal ceremonies, to assess the foraging ecology of the Grand Lagon Sud foraging green turtle population. Specifically, differences in isotopic niche width between different foraging aggregations and age-classes, as well as relationships between stable isotope values and turtle size (a proxy for age), were investigated to understand the degree of omnivory, trophic diversity and potential ontogenetic diet shifts in different aggregations of foraging green turtles in this location. This study will facilitate the understanding of the foraging ecology and habitat use of green sea turtles in this regionally important area for green turtles, whilst promoting links between science and Indigenous culture in New Caledonia.

Materials and methods

Study site

This study was conducted in the Grand Lagon Sud (GLS) (Fig. 1) of New Caledonia (22º 30″ S, 167º 57ʹ E), which supports a major foraging population of C. mydas. GLS is a World Heritage site that is characterized by a large reef diversity and the presence of many endangered species (Menu and Hebert 2006). Sea turtles are a protected species in New Caledonia (No. 4-2009/APS), but permits to hunt C. mydas are granted to Indigenous communities for traditional ceremonies such as weddings, funerals and yam celebrations.

Fig. 1
figure 1

Foraging locations where C. mydas were captured during this study

Full size image

Turtle capture

A total of 200 C. mydas were captured using the turtle rodeo method (Limpus and Reed 1985) in four areas of the GLS: Goro (n = 68), Isle of Pines (n = 26), Ouen Island (n = 47), and Mato/Uo Islands (n = 59), between August 2012 and March 2013 (Fig. 1). Goro is a town located on the northern part of the main island, and is in close proximity to an important seagrass meadow and a large river inlet. Ouen Island is located 1 km from the southern coastline of the main island. The Mato/Uo Islands are located approximately 16 km from the southern coastline of the main island and are in close proximity to the barrier reef. Isle of Pines is the largest island of the south province and is located 55 km to the south-east of the main island, surrounded by the barrier reef. All turtles were tagged with a titanium tag (Limpus 1992), their midline curved carapace length (CCL) was measured, and their sex determined (when possible) using external morphology or gonad examination. The turtles were released in the vicinity of their capture site following processing. In addition to the rodeo-captured turtles, 21 C. mydas that were caught and killed by tribes in the GLS for traditional ceremonies Goro (n = 13), Ouen Island (n = 3) and Isle of Pines (n = 5) between March 2013 and March 2014 were also used in this study. A small number of the turtles used for the ceremonies (n = 12) could not be measured precisely (due to access restrictions during tribal ceremonies), but an age-class was assigned based on an estimate of size.

The captured turtles were divided into four age-class categories, according to Limpus et al. (1994) and Limpus and Chaloupka (1997): new recruits (20–65 cm CCL, with a very white plastron and very low algae presence on the carapace), juveniles (25–65 cm CCL, with a yellow/discoloured plastron), sub-adults (65–90 cm CCL) and adults (> 90 cm CCL). The juveniles, sub-adults and adults were presumed to be long term residents at their capture sites.

Skin sample collection and stable isotope analysis

Epidermal (skin) tissue (~ 1 cm3) was collected from the hind flipper of all C. mydas (n = 200), using a sterilised scalpel, and immediately placed in 90% NaCl solution for storage (Arthur et al. 2008). Before stable isotopes analysis (δ13C and δ15N), samples were rinsed with distilled water and dried at 50 °C for 48 h, then ground into a fine powder using a mortar and pestle. The analysis was undertaken at the Griffith University Stable Isotope Analysis Laboratory using a sample preparation system (Sercon Europa EA-GSL), and an isotope ratio mass spectrometer (Sercon Hydra 20-22). This analysis provided δ13C and δ15N isotope values as a relative deviation (‰) from conventional standards (Pee-Dee Belemnite for carbon, atmospheric air for nitrogen), and expressed in delta (δ) notation in parts per thousand:’

$$\delta = \left( {\left[ {R_{{{\text{sample}}}} / \, R_{{{\text{standard}}}} } \right] - 1} \right) \times 1000$$

where Rsample is the ratio of heavy to light isotopes in the sample, and Rstandard is the isotope ratio of the corresponding standard. Bovine liver (n = 18) was used for quality control, with the standard deviation of the mean bovine liver samples assessing the precision of analysis (0.03‰ for δ13C and 0.13‰ for δ15N). Lipid extraction was not conducted on the skin samples as previous studies consider this process unnecessary for sea turtles due to the low lipid composition of sea turtle epidermis (Vander Zanden et al. 2012; Turner Tomaszewicz et al. 2017).

Stomach content collection and diet identification

Stomach contents were collected from the green turtles harvested by Indigenous communities (n = 21). The samples were collected as the turtles were being sacrificed, which was up to 5 days following capture. The collection and analysis of the stomach contents followed methods described by Garnett et al. (1985). Briefly, the entire stomach sample was removed from each turtle and homogenised before a sub-sample was collected and dried at 70 °C for 48 h. The samples were then further sub-sampled (~ 5% of the dry weight) into a petri dish for identification of individual diet items under a microscope, using methods described by Lanyon (1986). Animal matter was identified to phylum (Porifera or Cnidaria), algae to genus (or sometimes family), seagrass to species, and any diet items that could not be identified were assigned ‘unidentified’. In each stomach sample, each forage group (animal phylum, algal genus/family, seagrass species and ‘unidentified’) was weighed and the percentage of each forage group in each stomach sample was calculated.

It is important to note that sample numbers for stomach contents were low, and the exact capture locations of the turtles used in the tribal ceremonies were difficult to know precisely. So, for example, a turtle used in an Ouen Island ceremony may not necessarily have been captured in the Ouen Island foraging grounds.

Statistical analyses

As food items typically consumed by green turtles were not collected for stable isotope analyses in this study, the skin stable isotope data were not used to infer diet composition. Literature stable isotope data for forage material and trophic enrichment in green turtles could have been used to assess diet composition via stable isotope mixing model analyses (e.g. Piovano et al. 2020). However, due to significant temporal and spatial variation in stable isotope values of basal resource item isotopic values (Montoya et al. 2002), even over very small geographic ranges (Gillis et al. 2018), as well as the wide range in trophic enrichment values for green turtles in the literature (Seminoff et al. 2006; Turner Tomaszewicz et al. 2017; Vander Zanden et al. 2012), diet composition assessments were deemed inappropriate for the data collected in this study.

Instead, stable isotope data from green turtles (skin samples) were used to compare trophic width-between sites (all age-classes combined) and between age-classes (all sites combined). Standard metrics of isotopic niche width were calculated from skin δ13C and δ15N measurements using the Stable Isotope Bayesian Ellipses in R (SIBER) package (Jackson et al. 2011) of R version 4.3.2. Briefly, the six metrics used are based on the position of samples in the δ13C–δ15N bivariate space (Jackson et al. 2011; Layman et al. 2007), and are as follows: (1) δ15N range (NR), to represent trophic diversity within each group (site or age class); (2) δ13C range (CR), a measure of basal food source variability—groups with multiple basal resources can be expected to have higher CR values; (3) standard ellipse area (SEA), a measure of isotopic niche width; (4) mean distance to centroid (CD), a measure of tropic diversity in the group; (5) mean nearest neighbour distance (MNND), a measure of diet variability derived from the level of clustering within a group (mean Euclidean distances to the nearest neighbour in the δ13C–δ15N bi-plot); and (6) standard deviation of nearest neighbour distance (SDNND), a measure of evenness that is less influenced by sample size than MNND.

SEAs were corrected for differences in sample sizes by using a Bayesian approach (Jackson et al. 2011). Bayesian SEAs were further compared probabilistically, both between sites and between age-classes, using the posterior Bayesian distributions of simulated ellipses. To further investigate potential ontogenetic shifts in diet of these green turtles, the relationships between the CCL (a proxy for age) and green turtle skin δ13C and δ15N values were assessed via a series of regressions. All statistical analyses were run in RStudio version 1.4.1106 (2009–2021 RStudio, PBC).

Due to low and variable replication of stomach contents samples, and the uncertainty in the precise locations of the foraging habitats of these turtles, no statistical analyses were performed on the stomach contents data. These data were, however, discussed in the context of the stable isotope data.

Results

Stable isotope summary

The majority of green turtles sampled in this study were juveniles (159, 80% of turtles sampled), with smaller numbers of the other age-classes (Table 1). The δ13C and δ15N values of green turtle skin samples ranged from − 19.3 to − 7.3‰ and 2.8–15.9‰, respectively, and there was generally more similarity in δ13I values within sites compared to within age-classes (Fig. 2, Table 1). However, both juveniles and adults had very similar mean δ15N values within their age-classes, regardless of which site they were foraging in.

Table 1 The δ13C (‰) and δ15N (‰) values of C. mydas from different life history stages collected from four different foraging locations in Grand Lagon Sud, New Caledonia
Full size table
Fig. 2
figure 2

Mean (SD) green turtle skin δ13C or δ15N values from the four foraging sites within the Grand Lagon Sud, New Caledonia (Goro, Isle of Pines, Ouen Island and Uo/Mato Islands), separated by different age-classes: new recruits (square), juvenile (circle), sub-adults (triangle) and adults (△)

Full size image

Niche width-between sites

The standard ellipse areas (SEAs) were similar for Goro, Isle of Pines and Ouen Island, but much smaller for Uo/Mato Islands (Fig. 3, Table 2). The mean overlap between SEAs, based on 10 simulations of their Bayesian posterior distributions, was relatively low (24–35%) and similar for all pair-wise site comparisons (Table 3). Goro (high NR) had the highest tendency for omnivory and highest trophic diversity, whilst Ouen Island had high levels of herbivory and lowest trophic diversity. Isle of Pines (high CR) had the highest diversity of basal food sources, which was much lower in Uo/Mato Islands. The CD and MNND values were similar between sites, indicating similar trophic ecology at all sites. However, higher SDNND values at Isle of Pines indicated that there was a less even distribution of these turtles in the tropic niche at this site.

Fig. 3
figure 3

The δ13C and δ15N values of green turtles foraging within the Grand Lagon Sud, New Caledonia: Goro (black circles), Isle of Pines (red triangles), Ouen Island (green pluses) and Uo/Mato Islands (blue crosses). Lines depict the standard ellipse (the 40% confidence interval) for the site-specific isotopic niches

Full size image
Table 2 Summary of niche metrics for C. mydas foraging at four sites within the Grand Lagon Sud, New Caledonia (Goro, Isle of Pines, Ouen Island and Uo/Mato Islands)
Full size table
Table 3 Overlapping SEA (%) between the foraging grounds of the C. mydas foraging within the Grand Lagon Sud (isotopic niche overlap)
Full size table

Niche width-between age-classes

SEAs were similar in size for new recruits, juveniles and adults, and much larger for sub-adults (Fig. 4, Table 4). Similar to the site comparisons, the mean overlap between SEAs, based on 10 simulations of their Bayesian posterior distributions, was relatively low (22–35%) and similar for all pair-wise age-class comparisons (Table 5). New recruits and juveniles (high NR) had the highest tendency for omnivory, while sub-adults and adults had high levels of herbivory and low range of trophic levels. New recruits (low CR) had the lowest diversity of food sources, which was much higher in the other age-classes. The CD and MNND values were similar between new recruits, sub-adults and adults, indicating similar trophic diversity and redundancy in these age-classes. However, juveniles had lower MNND and SDNND values, indicating this age-class has greater trophic redundancy, and a more even distribution of individuals in their isotopic niche.

Fig. 4
figure 4

The δ13C and δ15N values of green turtles foraging within the Grand Lagon Sud, New Caledonia, separated by age-class: new recruit (black circles), juvenile (red triangles), sub-adult (green pluses) and adult (blue crosses). Lines depict the standard ellipse (the 40% confidence interval) for the age-class-specific isotopic niches

Full size image
Table 4 Summary of niche metrics for C. mydas foraging within the Grand Lagon Sud, separated by age-class
Full size table
Table 5 Overlapping SEA (%) between the age-classes of the C. mydas foraging within the Grand Lagon Sud (isotopic niche overlap)
Full size table

The regression analyses revealed significant but weak positive relationships between CCL and δ13C at Goro (p < 0.05, slope = 0.10, R2 = 0.11). In addition, significant but weak negative relationships were identified between CCL and δ15N for Goro (p < 0.05, slope = − 0.14, R2 = 0.11) and Uo/Mato Islands (p < 0.05, slope = − 0.12, R2 = 0.07) (Fig. 5). There were no other significant relationships between CCL and δ13C or δ15N.

Fig. 5
figure 5

Regressions between CCL (a proxy for age) and δ13C and δ15N values for C. mydas foraging within the Grand Lagon Sud, New Caledonia: Goro, Isle of Pines, Ouen Island and Uo/Mato Islands. Slope shown for significant regressions. Upper and lower curves indicate 95% Confidence Interval

Full size image

Stomach contents

The per cent contribution of different forage types in stomach contents varied greatly, both within and between sites (Table 6), although some trends could be observed. The stomach contents of green turtles foraging in Goro contained, on average, 90.4% algae, predominantly the red alga, Hypnea sp., which was present in all 13 stomach contents at an average of 44% of the total stomach contents. Green algae (mostly Caulerpa sp. and Ulva sp.) were present in 12 out of the 13 Goro green turtle stomach content samples, with an average of 28% each. Only four of the 13 Goro green turtle stomach content samples contained animal matter, although when they did, it was usually a substantial proportion (mean = 15%). Only one of the 13 Goro green turtle stomach samples contained seagrass, just 9.5% of Cymodocea serrulata. In contrast, the stomach contents of green turtles foraging in Ouen Island contained, on average, 82% seagrass (67% Halodule uninervis and 15% Cymodocea serrulata) with the remainder mostly animal matter. Finally, the stomach contents of two green turtles foraging in Isle of Pines contained 100% seagrass, mostly Cymodocea serrulata (69%), and stomach contents of the remaining three turtles contained 100% algae (or unidentified forage), mostly green and red species.

Table 6 Per cent composition (%) of C. mydas stomach contents analysed from foraging grounds in the GLS (New Caledonia)
Full size table

Discussion

The stable isotope niche width analyses presented here show that green turtles foraging in the Grand Lagon Sud (GLS) of New Caledonia have diverse foraging ecology, both within and between sites. The Uo/Mato Islands had the smallest isotopic niche width (smallest SEA), which combined with a relatively low diversity of food sources (low CR value, Table 2), indicates that green turtles foraging at Uo/Mato Islands have higher fidelity to a narrow range of foraging resources, and may therefore be more sensitive to potential habitat change at this site (Di Beneditto et al. 2016). However, it is important to note that this site included only one age-class (juveniles), which all seem to have similar trophic ecology in this region (low MNND value, Table 4). The remaining three sites had larger niche widths, and higher diversity of food sources, also reflecting the higher diversity of age-classes present, but possibly also indicating more resilience to future habitat impacts.

The minimal overlap (< 35%) of the SEAs between sites indicates that green turtles foraging at these four foraging sites have distinct isotopic niches. This is supported by the stomach contents analyses, which illustrated that Goro turtles feed predominantly on algae, Ouen Island turtles feed predominantly on seagrass, and Isle of Pines turtles have a mixture of both seagrass and algae (Table 6). Taking a closer look at individual sites, high NR values (Table 2) and negative correlations between δ15N and CCL at both Goro and Uo/Mato Islands (Fig. 5) indicated that there is a higher tendency for omnivory at these sites. This is supported by the dominance of new recruits and juveniles at these sites, and the higher tendency of these two age-classes for omnivory (high NR values; Table 4). In addition, although there were no stomach contents analyses for Uo/Mato Islands, the turtles sampled at Goro showed significant levels of animal matter in their stomach contents.

Green turtles are known to sometimes consume large quantities of animal matter, mainly macroplankton (e.g. jellyfish), especially when these animals are in abundance, and when abundance and quality of seagrasses and algae are low (Burkholder et al. 2011; González Carman et al. 2014; Lemons et al. 2011). The ingestion of animal material by green turtles was originally hypothesised to be incidental (Brand-Gardner et al. 1999; Mortimer 1981), although more recent studies have reported voluntary ingestion (Arthur et al. 2007; González Carman et al. 2014; Williams et al. 2014). Regardless if voluntary or involuntary, a certain amount of animal material is frequently found in stomach contents of green turtle populations globally, and it could be a more important diet source to this species than currently understood. Due to the high digestibility of animal material compared to plant material, the percentages found are usually low compared to seagrass or algae (Arthur et al. 2007; Heithaus et al. 2002). Certainly, the SIA and the stomach content data in this study indicate that green turtles in New Caledonia feed at more than one trophic level, and with significant consumption of animal matter, thus supporting the hypothesis for an omnivorous diet.

The tendency of new recruits and juveniles towards omnivory (high NR values and negative regression between δ15N at Goro and Uo/Mato Islands), in particular, supports findings from Moreton Bay (Australia) in which new recruits had significantly higher δ15N, compared to the older life stages that had been in the foraging ground for longer periods (Arthur et al. 2008). This was explained by an ontogenetic shift in diet, where oceanic juveniles feed mainly on a range of mostly carnivorous food items, compared to when they recruit to coastal habitats and switch to a diet of primarily herbivorous food items. This is further supported in the present study by the lower NR values in sub-adult and adult age-classes (Table 4), which implies high herbivory in these age-classes. Using SIA of different scute layers, this ontogenetic shift in diet was also found, in green turtles recruiting to coastal habitats in the Bahamas (Reich et al. 2007).

Globally, the diet of neritic foraging green turtle populations can vary greatly. The diet of some populations is dominated by one particular food source, such as seagrass (e.g. Green Island, Australia; Fuentes et al. 2006), algae (e.g. Moreton Bay, Australia; Brand-Gardner et al. 1999), or jellyfish (e.g. Samborombón Bay, Argentina; González Carman et al. 2014). Other green turtle populations (e.g. Fiji; Piovano et al. 2020), including the populations studied in GLS here, have more diverse diets, feeding on different proportions of a range of food items. From a management perspective, it is therefore important to understand the resilience of foraging areas to impact, and the environmental and anthropogenic processes that may drive the abundance and quality of food resources. Data from this study suggest that, even within a relatively small geographic area like the GLS, each of the distinct green turtle foraging aggregations, and even different age-classes within these aggregations, are supported by different food webs, and must therefore be managed individually.

Conclusion

This is the first study to investigate the feeding ecology of green turtles in New Caledonia. Combined SIA and stomach content analyses revealed that, even within a relatively small geographic area of the Grand Lagon Sud, the trophic ecology of green turtles differed significantly between foraging locations and age-classes, suggesting customised management is required at each location. This study therefore provides important information on the foraging ecology of green turtles in New Caledonia and ramifications for how to approach conservation of the species in this region. Further, this collaboration with the different tribes of the Grand Lagon Sud is also the first step towards the integrated management of green turtles as a resource and could be used as an example for the rest of the South Pacific.