Introduction

Migration is an integral part of the life histories of numerous species. For long-lived animals, it often involves seasonal round-trip movements between foraging and breeding grounds (Dingle and Drake 2007). Sea turtles, for example, exhibit philopatry, returning to genetically distinct nesting areas, but have mixed foraging grounds (Jensen et al. 2013). Understanding the migration patterns of sea turtles is essential for comprehending habitat connectivity and implementing effective, long-term conservation measures. Protecting sea turtles at or near their nesting beaches has been instrumental to the recovery of previously declining nesting populations (Mazaris et al. 2017; Pritchard et al. 2022). However, reproductive females likely spend > 75% of their time in their foraging areas (discussed in Shimada et al. 2020), and these regions can have a significant influence on body size and fecundity (e.g. in loggerhead turtles Caretta caretta; Ceriani et al. 2015). Declines in the number of juvenile green turtles at foraging aggregations have been recorded (e.g. green turtles in Florida; Long et al. 2021) despite increasing trends in the nearby nesting female populations (Chaloupka et al. 2008). Since many foraging aggregations are composed of sea turtles from multiple nesting regions (Bjorndal and Bolten 2008), decreases in nesting populations due to threats in foraging areas (e.g., habitat degradation, fisheries, etc.) may not be detected by nesting beach monitoring for several years (e.g., Ceriani et al. 2019). Identifying and accessing foraging areas remains challenging, hindering the assessment of threats and prioritization of conservation efforts. Nonetheless, tools, such as satellite telemetry, have helped to identify key conservation areas by identifying crucial foraging habitats and migratory pathways in various ocean regions (e.g., Ferreira et al. 2021; Cerritelli et al. 2022; Lamont et al. 2023).

Stable isotope analysis (SIA) is a powerful complementary tool to satellite telemetry for understanding the spatial and foraging ecology of sea turtles (Ceriani et al. 2017; Bradshaw et al. 2017). SIA relies on variations in the ratio of heavy to light isotopes of common elements, which exhibit spatial and temporal variability, leading to natural and predictable isotope gradients within the environment (DeNiro and Epstein 1978, 1981). Analyzing animal tissues with slower turnover rates such as skin or muscle enables inferences to be made on the animal’s foraging location over a longer period of time, ranging from several weeks to months, depending on the species and tissue type (Hobson 1999; Rubenstein and Hobson 2004). This can reveal the isotopic values of the non-breeding foraging areas when samples are collected at breeding grounds, if the animal remains long enough in a non-breeding area to assimilate a local isotopic signature and moves between isotopically distinct areas (Hobson 1999; Rubenstein and Hobson 2004; Hobson and Wassenaar 2019). SIA provides coarser geographic information than satellite telemetry but allows for a larger sample size when the two techniques are combined (Newsome et al. 2007). Commonly used stable isotopes in sea turtle research are carbon (δ13C) and nitrogen (δ15N) (Ceriani et al. 2012; Seminoff et al. 2012; Vander Zanden et al. 2014). The δ13C of a consumers’ tissues reflects that of the primary producer, since there is minimal fractionation (~ 1‰) with successive trophic levels (DeNiro and Epstein 1978). δ13C has generally been used to identify location, because it decreases with increasing latitude, and benthic environments tend to be more enriched than pelagic (Rubenstein and Hobson 2004). δ15N is more indicative of trophic position (DeNiro and Epstein 1978, 1981) because the δ15N of a consumer is enriched by its diet. δ15N values increase when eating prey with higher trophic positions in the food web (DeNiro and Epstein 1981). Using SIA to determine foraging areas requires knowledge of the spatial distribution of isotopic values in potential areas, through large-scale isotopic maps or regional isoscapes (Graham et al. 2010; McMahon et al. 2013), although such studies are limited (but see Lorrain et al. 2015). Some sea turtle studies have bypassed this limitation by using isotopic values of tissue samples with lower turnover rates (i.e., skin has a 3‒4 month turnover in fast growing juvenile sea turtles; Reich et al. 2008) (Seminoff et al. 2007; Reich et al. 2008; Ceriani et al. 2014a) from satellite tracked females to characterize isotopically distinct regions. Isotopic values of satellite tracked females are then used to develop models to assign the likely origin of non-satellite tracked individuals that are sampled for SIA, allowing for larger sample sizes (Ceriani et al. 2012, 2014b; Seminoff et al. 2012; Bradshaw et al. 2017).

The Western Indian Ocean (WIO) hosts globally important rookeries for Endangered green turtles (Chelonia mydas) (Seminoff et al. 2015), mostly on islands (Mortimer et al. 2020). Satellite telemetry has been used to investigate foraging areas in the region with nesting green turtles in the Chagos islands (Hays et al. 2020) and in the French territories (Dalleau et al. 2019). North-west Madagascar, South Madagascar, central Tanzania, and North and South Mozambique have been identified as green turtle foraging hotspots in this region, which is supported by satellite tracked green turtles within the Mozambique Channel (Dalleau et al. 2019). These modeled foraging hotspots had some overlap (such as along East Africa and Madagascar) with those used by satellite tagged females from the Chagos islands to their foraging grounds (Hays et al. 2020). One study so far has looked at adult green turtle diet in the WIO, with seagrass as the dominant prey item for non-breeding females (Stokes et al. 2019). Otherwise, very little is known about adult green turtle diet in this region (Esteban et al. 2020).

Despite their vast spatial range, green turtles in the WIO continue to face threats, including from fisheries interactions and direct take (Williams et al. 2016; Temple et al. 2018; van de Geer et al. 2022). In South-west Madagascar, it is estimated that at least 10,000 green turtles are caught and sold annually (Humber et al. 2011), emphasizing the importance of identifying and protecting critical foraging habitats in the region to ensure conservation efforts are effective and sustained. To do so, a sufficiently large sample size to ensure representation of the nesting populations is needed, which can be achieved with SIA.

The second largest green turtle rookeries in the WIO is Aldabra Atoll (> 15,000 clutches annually), which is a large, remote atoll in the southern Seychelles (Pritchard et al. 2022). Aldabra falls within the Southwest Indian Ocean regional management unit (RMU) for green turtles (Wallace et al. 2023). The RMU has an estimated 103,000‒144,000 green turtle clutches laid annually (Mortimer et al. 2020) and over 10% are laid at Aldabra (Pritchard et al. 2022). Therefore, there is a lot of conservation value to understanding what foraging grounds green turtles nesting at Aldabra use. In the WIO, SIA has been evaluated for co-occurring large marine megafauna surrounding La Reunion (including foraging green turtles) (Chandelier et al. 2023), but it has not been evaluated in the region for linking nesting green turtles to their foraging areas. Also, the Indian Ocean has been recognized as having a lack of SIA research for contributing to and advancing ecological knowledge and making effective conservation measures for sea turtles in the region (Pearson et al. 2017; Haywood et al. 2019).

Our overall aim was to investigate the use of bulk carbon and nitrogen stable isotope analyses in identifying the foraging areas used by green turtles nesting at Aldabra, within the broader WIO. Using preliminary information on foraging areas obtained by satellite tracking (Seychelles Islands Foundation, unpubl. data), we investigated (i) whether nesting green turtles at Aldabra utilize isotopically distinct foraging areas; and (ii) where they primarily forage.

Methods

Study site and sampling

Aldabra Atoll, managed by the Seychelles Islands Foundation, is a large raised atoll and UNESCO World Heritage site in the Western Indian Ocean (WIO; 9°25’S, 46°20’E; Fig. 1). It is ca. 34 × 14 km with a total land area of 155 km2 with four main islands that surround a large, mostly shallow lagoon (< 3 m) (Hamylton et al. 2018). There are 52 beaches (total beach length ca. 5.2 km) around the outer coastline of the islands which are used by nesting green turtles (Pritchard et al. 2022).

Fig. 1
figure 1

Foraging areas (FA)/end points of the 13 satellite tracked green turtles after nesting on Aldabra (shown in insert; research station/Settlement beach represented with a star) (SIF, unpubl. data). Green turtle numbers 1–5 coincide with the endpoints in Madagascar (FA2; filled triangles). Numbers 6–11 are endpoints along the East African coastline (FA1; filled circles) (Table S1, supplementary material). Green turtles 12 and 13 had no isotopic data, and green turtles 1 and 6 were not considered to belong to neither FA1 nor FA2 (filled boxes)

Full size image

A subset of nesting females was encountered and sampled for stable isotope analysis (SIA) on Settlement Beach during night patrols and morning track count surveys in three periods; from March to October 2020 (n = 39), March to September 2021 (n = 51) and January to April 2022 (n = 39), hereafter referred to by the year of sampling. Curved carapace length notch-to-tip (CCLn-t) was recorded (Bolten 1999), and as part of a long-term monitoring program, unique metal tags were applied to both front flippers (Titanium Turtle Tags, Stockbrands Pty Ltd) (Balazs 1999) after the green turtle started covering her eggs. In 2020, epidermis (0.3 cm2) was sampled from the soft skin on the trailing edge of the rear flipper (originally sampled for a separate genetic study), while in 2021 and 2022, the top epidermal layer (‘skin sample’; 0.3 cm2) was sampled from the neck (between the neck and front flipper). All skin samples were collected with sterilized razor blades. In 2020, samples were originally stored in 99.8% ethanol (ETOH) for 372–567 days, then a small piece was subsampled and transferred to 75% ETOH for ca. 9 months until analyzed. In 2021 and 2022, skin samples were stored in iodized salt.

Additionally, in 2022, 13 green turtles were satellite tagged (SIF, unpubl. Data). Eleven of those satellite tagged green turtles were also skin sampled for SIA; therefore the 2022 skin sample total for SIA (n = 39) had 28 non-satellite tagged green turtle tissues and 11 satellite tagged green turtle tissues. For the 13 satellite tagged green turtles, Wildlife computers Splash 10-224D (n = 5), Lotek Kiwisat K237E (n = 5), and Telonics SeaTrkr units (n = 3) were applied to the green turtles’ carapaces using standard procedures (Coyne et al. 2008): nesting females were approached after laying or while returning to the sea after an unsuccessful nesting attempt, and were restrained in a wooden box. The carapace was cleaned and sanded, then cleaned with acetone. Epoxy resin (DeWalt Pure 150-Pro) was applied to the highest, flattest part of the carapace along the vertebral scutes and to the bottom of the tag (also roughened with sandpaper and cleaned with acetone). The tag was attached to the carapace, additional epoxy applied, and smoothed along the shell. Green turtles were released once the epoxy had dried, generally 2–3 h after the tag was applied (Fig. S1, supplementary material).

The three different tag types might theoretically affect the length of transmissions. However, no such evidence exists (Hart et al. 2021), and for our aim, the tags just need to transmit long enough to identify their wide foraging areas, which could be achieved for all tags.

Sample preparation and SIA

Samples were prepared for SIA following standard procedures (Ceriani et al. 2014b, 2015). Each sample was put in successive distilled water baths to remove NaCl, then surface epidermis and any dermal tissue underneath was separated. The epidermis was dried in the drying oven overnight (> 12 h) at 60 °C. Each tissue was cut into several small pieces, and lipids were extracted using an accelerated solvent extractor (Model 200, Dionex) with petroleum ether. Each batch underwent 5 min of heating followed by 5 min of static purging, three times.

Subsamples of prepared tissues (0.2–0.6 mg) were weighed with a microbalance and sealed into sterile 3 mm x 5 mm tin capsules and analyzed for % carbon, % nitrogen, ẟ13C, and ẟ15N values. Analyses were performed at the Marine Environmental Chemistry Laboratory at the University of South Florida College of Marine Science (St. Petersburg, FL, USA), where samples were converted to N2 and CO2 using a Carlo-Erba NA EA1108 Elemental Analyzer (Thermoquest Italia, S.p.A., Rodano, Italy) and analyzed with a continuous flow isotope ratio mass spectrometer (Delta PlusXP, Thermo Finnigan, Bremen, Germany). Stable isotope ratios were expressed in conventional notation as parts per thousand (‰) according to the following equation:

$$\delta X= \left[\right(\frac{Rsample}{Rstandard})-1] \times \text{1,000}$$

where X is 15N or 13C, and R represents the corresponding ratios of heavy to light isotopes (15N:14N and 13C:12C) in the sample and international standard, respectively. Raw measurements were calibrated relative to VPDB (δ13C) and AT-Air (δ15N) with certified reference materials NIST 8574 (δ13C = + 37.63 ± 0.10‰, δ15N = + 47.57 ± 0.22‰, N = 9.52%, C = 40.81%, C: N [molar] = 5.0) and NIST 8573 (ẟ13C = -26.39 ± 0.09‰, δ15N = -4.52 ± 0.12‰ N = 9.52%, C = 40.81%, C: N [molar] = 5.0). Estimates of analytical precision were obtained by replicate (n = 58) measurements of an internal laboratory reference material (NIST1577b Bovine liver, ẟ13C = -21.69 ± 0.14‰, δ15N = 7.83 ± 0.16‰, %N = 9.95 ± 0.22%, %C = 48.04 ± 0.90%, C: N [molar] = 5.63 ± 0.08) and yielded a precision (reflecting ± 1 SD) ≤ 0.13‰ for ẟ13C and ≤ 0.11‰ for δ15N.

Tissue storage treatment comparison

Due to concerns in the literature about preservation method and effect on isotopic values (Barrow et al. 2008; Bradshaw et al. 2017), we compared tissue treatments, following Bradshaw et al. (2017). Twenty-two of the 39 skin samples collected in 2022 were large enough to be subdivided in three parts with a razor blade. They were placed into three treatments: salt, 75% ETOH, 99.8% ETOH. Tissues (between individuals) were in treatments from 105 to 151 days. A Friedman test was used to analyze the repeated treatments.

Satellite tagged green turtles and foraging areas

To designate if the green turtle’s reached their foraging ground, apparent minimum distances and time differences between successive locations were calculated and the ratio between the two was used to calculate the apparent minimum speeds between locations. These were used to filter out locations with speeds > 5 km h− 1. Displacement plots were made with difference in time and straight-line distance between the first location and successive locations, and the plots were visually inspected for migratory stage and foraging areas (Cerritelli et al. 2022). For our purpose, we needed to identify broad foraging areas rather than the exact foraging location of each individual. We considered a green turtle had reached their foraging area once their displacement gradient effectively reached zero. For the 13 satellite tagged green turtles, two main foraging areas were identified but two green turtles migrated elsewhere (one to South Madagascar, one to Mozambique). The displacement gradient of the green turtle that last transmitted in Mozambique did not reach zero, but due to the general pattern of Aldabra satellite tagged green turtles not switching between the East African coastline and Madagascar, Mozambique was considered her general foraging area. The 12 other green turtles were considered to have reached their respective general foraging areas (Fig S2; Table S1, supplementary material).

Data analyses

All analyses were performed in R (v4.3.0; R Core Team 2023) using RStudio (v1.2.5; Posit team 2022). The relationship between δ15N and δ13C and size or latitude and longitude was explored using the Pearson correlation coefficient. The relationship between year (2021, 2022) and δ15N and δ13C was investigated with a Wilcoxon rank sum test. Data were tested for normality and equal variance (Shapiro-Wilk and Levene’s tests), and since the results were not significant, a multivariate analysis of variance (MANOVA) with Pillai’s trace test and post-hoc Games-Howell (multiple comparison tests) were used to determine differences in isotopic values and foraging areas.

Two methods were used to group foraging areas and assign non-satellite tagged green turtles to a foraging area: (1) Ward’s hierarchical cluster analysis (Ward and Hook 1963), and (2) discriminant function analysis (DFA). For the first, the package “stats” (R Core Team 2023) was used with function “hclust” and method “ward.D2” using a Euclidean distance. For DFA, the package “MASS” (Venables and Ripley 2002) with function “lda” was used to take the stable isotope values of the training dataset (satellite tagged green turtles) to make predictions, using “predict()”, for the untagged green turtles. Function “lda” uses the ratio of between-class variance to within-class variance to ensure the best class separability (Venables and Ripley 2002). Nine of the 13 satellite tagged individuals were used as the training set with two foraging areas: FA1 (East Africa: n = 5) and FA2 (northern Madagascar; n = 4). The other four satellite tagged individuals were not included because two did not have skin tissues taken and two went to areas other than the two main areas above (a sample of one for each of these additional areas was too small for the analysis). The more conservative posterior probability ≥ 0.8 was employed since there is little information available about green turtles and stable isotopes in the region; this threshold gives a fourfold improvement in odds for assignment compared to random allocation (Wunder 2012) and has been used in similar studies (Haywood et al. 2020; Pfaller et al. 2020). Equal priors were used since the satellite tagged sample was too small to confidently use the observed proportions as priors. A leave-one-out cross-validation for individual assignment to a foraging area was carried out. The proportions of green turtles frequenting the two foraging areas were estimated through two approaches: (i) as the proportions of assigned individuals and (ii) as the average posterior probabilities of all individuals (Ceriani et al. 2023). Size (CCLn-t) was compared between the two groups with Welch’s t-test (incl. satellite tagged females).

Results

Sample preservation

For the comparison of skin sample storage method experiment for the salt, 75% ETOH and 99.8% ETOH, δ15N values differed between the three treatments (df = 2, p = 0.006; Table S2, supplementary material), while δ13C values did not (df = 2, p = 0.23). These effects, in addition to the magnitude of variation between the treatments being greater than the mass spectrometer’s precision, led to the exclusion of the 39 samples collected in 2020 from further analyses. Further support for excluding the 2020 samples includes that the samples were exposed to two different preservatives at two different times (99.8% ETOH then into 75% ETOH), and the skin tissue samples for SIA from 2020 were taken from the trailing edge of the rear flipper, while the samples from 2021 to 2022 were taken from the neck. Whether SI skin values differ among regions of the body has not been investigated in the literature. Therefore, only the 90 green turtles sampled in 2021 and 2022 and preserved with the same method (iodized salt) were analyzed.

General summary

Stable isotope values were 5.7–14.0‰ (mean 9.6 ± 1.7‰) for δ15N and from − 19.4 to -6.1‰ (-11.8 ± 3.50‰) for δ13C (n = 90). The average green turtle size was 109.2 ± 5.3 SD cm CCLn-t (range 95.7–128.0 cm; n = 90). No relationship was found between green turtle size and δ15N or δ13C (p > 0.05). In the 11 satellite tracked green turtles with isotopic values, δ15N decreased (t = -3.52, df = 9, p = 0.007) and δ13C increased (t = 2.38, df = 9, p = 0.041) with longitude (from west to east), while no relationship was detected with latitude for δ15N (t = 0.43, df = 9, p = 0.68) or δ13C (t = 0.47, df = 9, p = 0.65). There was no difference between year and δ15N (W = 1104.5, p = 0.37) and δ13C (W = 1009, p = 0.91).

Foraging areas

Hierarchical cluster analysis showed foraging areas in two clear foraging area clusters. One cluster (Cluster 1) included satellite tagged individuals who went to Tanzania (n = 3), Somalia (n = 2), Mozambique (n = 1), South Madagascar (n = 1), and North Madagascar (n = 1). The other cluster (Cluster 2) included three satellite tagged individuals that went to northern Madagascar (Figs. 1 and 2; Fig. S3, supplementary material). Two main foraging areas were identified: Somalia and Tanzania (Foraging area 1; hereafter ‘FA1’) and northern Madagascar (Foraging area 2; hereafter ‘FA2’). Only one satellite tagged green turtle, that migrated to north-east Madagascar was assigned to the ‘incorrect’ cluster (Cluster 1 instead of 2) while two green turtles assigned to Cluster 1 went to South Madagascar and Mozambique. Assuming that Cluster 1 and 2 indicate FA1 and FA2, respectively, of the total 90 green turtles sampled for SIA, 47 (52.2%) would be assigned to FA1 and 43 (47.8%) to FA2.

Fig. 2
figure 2

The isospace of untagged and satellite tagged green turtles nesting at Aldabra when split into two clusters (Cluster 1 = blue; Cluster 2 = red; Ward clustering). Satellite tagged individuals are shown with bold dots, with the number representing the foraging area group each green turtle was assigned to based on the three migration patterns detected by satellite tracking (1 = East Africa: Tanzania/ Somalia, 2 = northern Madagascar, 3 = other: South Madagascar + Mozambique). Labels indicate the region the satellite tagged green turtles went to: Mozambique (Moz), Madagascar (Mad), Tanzania (Tan)

Full size image

The isotopic signatures of the nine satellite tagged green turtles assigned to FA1 and FA2 were different (MANOVA, Pillai = 0.85, approx. F = 17.39, num df = 2, den df = 6, p = 0.003) for both δ15N (p < 0.001) and δ13C (p = 0.044; Fig. 3). The DFA results assigned 100% of the nine satellite tagged green turtles with a posterior probability ≥ 0.8 and seven (77.8%) were assigned correctly. The two incorrectly assigned satellite tagged green turtles went to NE Madagascar (FA2, green turtle #5) and Tanzania (FA1, green turtle #8). For green turtles with unknown foraging areas, 73 of 81 (90%) were assigned to a foraging group with a posterior probability ≥ 0.8. Forty-five green turtles were assigned to FA1 (East Africa) and 28 to FA2 (northern Madagascar). For the whole dataset of satellite tagged and non-satellite tagged females with a posterior probability ≥ 0.8 (n = 82), 50 (61.0%) were assigned to FA1 and 32 (39.0%) to FA2. The two satellite tracked green turtles with isotopic values that went to minor foraging areas and the two that were treated as unknowns (i.e., excluded from the training dataset) were both assigned to FA1 with the DFA. The alternative approach based on the average posterior probabilities of all individuals (n = 90) estimated that 58.9% of sampled green turtles foraged in FA1 and 41.1% in FA2. The size of green turtles with posterior probabilities ≥ 0.8 was similar at the two FAs (Welch’s t-test, t = 0.6, df = 57.2, p = 0.5; n = 82).

Fig. 3
figure 3

(a) δ15N and (b) δ13C distribution (median and quartiles), for satellite tagged green turtles based on their assigned foraging area (FA): (1) East Africa (Tanzania/ Somalia; n = 5); (2) northern Madagascar (n = 4)

Full size image

Discussion

We identified two main foraging areas, the East African coast (Tanzania and Somalia) and northern Madagascar, and two other potential foraging areas that should be investigated (Mozambique and South Madagascar). Our results indicate that the majority of the sampled green turtles nesting at Aldabra forage in one of these two main areas, with slightly more foraging along the East African coast. Although green turtle populations are recovering in the region after protection of major nesting areas (Lauret-Stepler et al. 2007; Mortimer et al. 2020; Pritchard et al. 2022), threats remain in their foraging areas, including poaching and fisheries bycatch (Bourjea 2015; Temple et al. 2018). The similar contribution of both foraging areas to the Aldabra nesting population highlights the importance of these large regional areas, and further investigation would determine if finer differentiation is possible. Worldwide, only a few studies on green turtles have used satellite telemetry combined with SIA to assign likely foraging areas (e.g., Hatase et al. 2006; Bradshaw et al. 2017), and this is the first such published study in the Western Indian Ocean.

Our sample size of satellite tagged females was small, but they covered a large geographic range (ca. 2800 km latitudinally) when migrating from Aldabra to their foraging areas. Other studies that tracked females from nesting rookeries to foraging areas have also found Mozambique and the East African coast to be key areas for foraging sea turtles. Loggerhead turtles (Caretta caretta) nesting in iSimangaliso Wetland Park in southeast Africa generally migrate along the coast of Mozambique or cross the Mozambique Channel to Madagascar and travel north up the western coast of Madagascar (Robinson et al. 2018). Green turtles nesting in the Chagos Islands forage widely in Seychelles, the East African coast (Mozambique to Somalia), North Madagascar, Tromelin and the Maldives (Hays et al. 2020). Foraging area modeling for green turtles from oceanic island rookeries in the Mozambique Channel identified the coastal waters of northern Madagascar and Africa, bordering the north Mozambique channel to be the most used (Dalleau et al. 2019), and the two main foraging areas identified for green turtles leaving Aldabra overlapped with the predicted hotspot area. Further SIA research linking nesting rookeries and foraging grounds will improve these models to give a better understanding of spatial use and population dynamics in this region.

We found no difference in body size between green turtles using the two foraging areas. Previous studies in other ocean basins have found mixed results: some found size differences based on neritic/oceanic foraging areas or latitude (Zbinden et al. 2011; Ceriani et al. 2014b; Özdilek et al. 2023), while others did not (Hatase et al. 2002; Tucker et al. 2014). The lack of size difference in our study could suggest that the sampled green turtles have a similar foraging strategy across their geographic range and/or that the foraging habitats are similar.

Isotopic patterns

This is the first study from the Western Indian Ocean using carbon and nitrogen bulk stable isotopes to infer foraging grounds of nesting green turtles, therefore we do not have regional comparisons. We found general low isotopic structure for Aldabra nesting green turtles in comparison to sea turtle species in other regions, such as loggerhead turtles in the northern hemisphere (Ceriani et al. 2012; Vander Zanden et al. 2015a). Although a latitudinal gradient in δ13C has been documented elsewhere (e.g. Southern Ocean; Quillfeldt et al. 2005), and even within the pelagic realm of the Western Indian Ocean (Lorrain et al. 2015), we found no latitudinal variation in δ15N and δ13C for green turtles that migrated from Aldabra to their potential foraging grounds. This may represent a limit of this approach in the Western Indian Ocean with this species, and possibly the wider Indian Ocean.

The two satellite tracked green turtles that went to areas (Mozambique and South Madagascar green turtles) other than the two main foraging areas identified (FA1 and FA2) and their isospace values suggest that other minor foraging areas of Aldabra nesting green turtles may exist in the region. Identifying more isotopically distinct areas with a larger sample size of individuals that are both satellite tracked and sampled for SIA would facilitate more accurate, fine-scale assignment of non-tracked green turtles to specific foraging grounds, allowing more precise identification of critical foraging habitats. Additionally, there would be higher confidence in assigning green turtles to specific foraging grounds, increasing the reliability of the inferences.

No other published studies have used both SIA and telemetry to infer post-nesting green turtle foraging areas in the Western Indian Ocean, so no direct comparisons can be made within the region. However, there have been similar studies on different taxa using coastal areas in the region. For example, whale shark foraging areas were found to isotopically overlap (δ13C and δ15N) in Mozambique and Tanzania, but were isotopically different between Mozambique and Qatar (Prebble et al. 2018). Prebble et al. (2018) compared whale sharks with goose barnacles and two species of tuna in the region (from Lorrain et al. 2015; Sardenne et al. 2016), and found that the δ13C and δ15N values generally increased with decreasing latitude, but the δ13C trend was less clear. The enriching latitudinal trend of δ15N was proposed to be driven by the δ15N baseline in the local environment. Our study, however, did not find any relationship with isotopic values and latitude. The lack of δ13C pattern in our study concurred with Lorrain et al.’s (2015) study on tuna, as well as results from loggerhead turtles in southeast Queensland, Australia, where there was no correlation between latitude and δ13C or δ15N. This was attributed to wide variation in individual diet, and to the southwest dynamic current patterns and environmental factors of the Pacific Ocean, which prevented the formation of unique isoscapes (Coffee et al. 2020). This emphasizes the need to investigate each region separately.

We found longitudinal variation of δ13C and δ15N that few studies have reported; with δ13C becoming more enriched and δ15N becoming more depleted with increasing longitude (from west to east). In the South Atlantic Ocean, variations in δ13C and δ15N were detected longitudinally (with lower values for both in the west compared to the east) and vertically in deep-sea copepods (higher δ15N with increased depth) and were attributed to regional differences in hydrography and sea surface temperature (Laakmann and Auel 2010). The oceanography of the Western Indian Ocean is dynamic: the Mozambique Channel and East African coast have strong currents and eddies: the South Equatorial Current flows east to west across the Indian Ocean and diverges north and south off northern Mozambique. The northern split becomes the East African Coast Current, where Tanzanian reefs and islands have similar coral communities to those in the northern Mozambique Channel (Obura 2012), while the southern split forms the Mozambique Channel, where various eddies and dipoles (100–300 km across) along the Mozambique coast drive primary production by bringing dissolved nutrients towards the surface (Schott et al. 2009; Obura et al. 2018). Additionally, upwelling events occur off NW Madagascar (Pripp et al. 2014). The complex oceanography of the WIO (Schott et al. 2009) is likely contributing to the patterns seen in our study.

Overall, further research is needed into the isotopic ranges of the foraging areas identified and other potential confounding or influencing factors, such as dietary shifts, regional prey, and individual feeding strategies (Thomson et al. 2018). Additionally, green turtle foraging ecology is strongly related to sea surface temperatures (Esteban et al. 2020), therefore individuals at lower latitudes may consume different prey, such as more animal matter, than green turtles at warmer sites (Esteban et al. 2020). The variation in δ13C and δ15N at the base of the food web can also vary across small spatial gradients (Post 2002; Magozzi et al. 2017). Contrastingly, foraging grounds of nesting green turtles in the Mediterranean show several geographically distinct areas with the same isotopic profile (Bradshaw et al. 2017). While the isotopic patterns of the WIO are not yet clear, analyzing additional isotopes or trace elements may help delineate the carbon and nitrogen, and identify more distinct sites (Bradshaw et al. 2017; Pearson et al. 2019).

An incidental finding of our study was that isotopic values differed depending on sample treatment. Although this led to the exclusion of one year of samples from our analyses, it is an important finding to inform how preservation methods (salt, 75% ETOH, and 99% ETOH) affected bulk isotope nitrogen values. A similar experiment on green turtles in the Mediterranean (using 96% ETOH and 70% ETOH) found no difference between treatments (Bradshaw et al. 2017), but our results indicate that the tissue preservation technique (e.g., salt or different concentrations of ethanol) can significantly influence the results. Therefore, where different storage methods are used, a comparison of the tissues preserved with different methods should be done to ensure confidence in the results.

Method limitations

Several limitations of our study warrant consideration, including the relatively small sample size of green turtles, which may prevent a fuller understanding of foraging area contribution, or not include all foraging areas used by green turtles from Aldabra’s rookery. The dynamic nature of Western Indian Ocean oceanography means that additional satellite tracking combined with SIA throughout the region is needed to determine whether more distinct foraging areas exist.

As green turtles were tracked after nesting, we assumed that the isotopic values of samples were representative of foraging area prior to the nesting season as has been a commonly used assumption in other sea turtle studies (Ceriani et al. 2012, 2023; Seminoff et al. 2012; Pajuelo et al. 2012). The isotopic turnover rate of skin is not known for green turtles, but for captive, fast-growing juvenile loggerhead turtles (Caretta caretta) it is ca. 3–4 months (Reich et al. 2008). Using ectotherm predictive equations (from Vander Zanden et al. 2015b); Prior et al. (2016) estimated the δ13C of large juvenile and adult green turtle skin samples to reflect their diet > 9 months before sampling. This strongly supports the likelihood that isotope ratios reflect foraging ground diets. We also assumed that green turtles had fidelity to their foraging areas, which many other studies have found (Evans et al. 2019; Pilcher et al. 2020; Shimada et al. 2020). Studies using both satellite telemetry and SIA of skin tissues to infer sea turtle post-nesting foraging areas have had either empirical evidence of site fidelity or have assumed site fidelity based on other studies (Ceriani et al. 2012; Vander Zanden et al. 2015a; Bradshaw et al. 2017; Haywood et al. 2019).

SIA will reflect a combination of signatures from different locations in sea turtles with low fidelity; therefore, if the sampled green turtles exhibit low foraging area fidelity, the SIA results may represent a composite of isotopic signatures from multiple foraging areas they have visited, potentially leading to misidentification of specific foraging locations. This emphasizes the need to verify if green turtles from this region have high site fidelity; there is also a need to know more about the regional isotopic structure of primary producers/sea turtle prey items, including precise information on the identified and other potential foraging areas, to either support or refute the findings here.

Conclusions and recommendations

The regional structure of stable isotopes from foraging areas of green turtles nesting at Aldabra is less clear than in other regions, but the combined SIA/satellite-tracking approach indicated that these green turtles migrate to at least two main foraging areas after nesting. More distinct and/or additional foraging areas could potentially be determined with larger sample sizes from Aldabra, and also from other potential foraging grounds (e.g., Mozambique and southern Madagascar) which highlights the need for international cooperation. Understanding the spatial resolution of SIA for sea turtles in this region would help evaluate the use and practicality of SIA. The lower cost and larger sample size that SIA allows, compared to satellite tracking, is beneficial for understanding the relative importance of these foraging grounds, for managing critical habitats, and for identifying threats to green turtles in this region, which are not well known. For further SIA research in this region, including additional tracers may be necessary (Bradshaw et al. 2017; Haywood et al. 2019). Additionally, research on the isotopic values of primary producers, such as seagrasses, in sea turtle foraging areas is currently lacking and would provide valuable insight into isotopic regional patterns. Continued research on green turtles at Aldabra, and other green turtle rookeries and foraging areas in the region, will shed further light on green turtle connectivity and will help inform effective conservation actions for foraging regions, where green turtles spend the majority of their lives.