Foraging Behavior and Diet Preferences of a Released Population of Giant Tortoises in the Seychelles
Abstract
This study evaluates the potential impacts of the release of the giant tortoise, Dipsochelys arnoldi, to vacant habitat within the species' presumed historic range. Five individuals (3 males and 2 females) were released in December 2006 in the isolated Grande Barbe area of Silhouette Island, Seychelles. A comprehensive vegetation survey of all plant species within feeding height of the tortoises in Grande Barbe was conducted. These data, combined with daily feeding observations, were used to calculate feeding rates and diet preferences. Tortoises were observed to have a mean diurnal active time of 257 minutes per day exhibiting a mean feeding rate of 3.86 g of plant material consumed per minute of active time. Individuals were recorded feeding on 18 of 26 recorded plant species, 9 of which were selectively foraged. Species composition of plant species ingested differed between sexes. Males selectively foraged on 6 plant species and females selectively foraged on 7 species. Only 2 plant species were selectively foraged by both sexes.
Captive breeding and subsequent reintroduction has become an important conservation tool in recent decades (Snyder et al. 1996). However, in spite of the increased use of this highly publicized conservation technique the estimated global reintroduction success rate is only 11% (Stanley-Price and Sooray 2003).
A number of problems have been highlighted in the reintroduction of captive bred animals, such as domestication, loss of wild behavior, loss of resistance to disease, genetic drift, and high financial costs (Snyder et al. 1996).
An additional, and often overlooked potential problem with reintroduction of animals to within their historic range after local extinction or decline is the suitability of habitat, and the effect of the reintroduced population on the habitat. Habitat change may have led to the initial population decline, or the habitat may have changed in the absence of the species. It is essential for the long-term success of release programs that the habitat requirements of species, and impacts of species reintroductions on ecosystems, are understood.
Despite the historical presence of a species, the ecosystem impacts of a reintroduced population cannot be assumed to be beneficial. Large herbivores can play a keystone role in ecosystem dynamics. Upon release or relocation (natural or artificial) into previously unoccupied areas, the presence of large herbivores such as elephants has been found to have negative impacts beyond direct consumption on botanical species (Landman et al. 2008). Trampling, knock-on effects, and zoochory are some of the potential impacts of the introduction of a large herbivore (Landman et al. 2008).
The presence of introduced herbivores on islands can result in top-down control of vegetation community structure with preferred species declining and unpalatable plant species increasing in abundance, with the opposite trend observed in the absence of herbivores (Donlan et al. 2002). Giant tortoises have been compared to elephants in their ability to act as habitat engineers (Bourn et al. 1999), and their reintroduction should be monitored closely. Trampling, habitat modification, and apparent selective foraging have been recorded within the captive enclosures of giant tortoises (Gerlach 2006).
Historically, the Indian Ocean was home to about 15 endemic species of giant tortoises. As a result of overexploitation by settlers only one Indian Ocean giant tortoise species, the Aldabra giant tortoise (Dipsochelys dussumieri, also known as Aldabrachelys gigantea), exists in the wild today (Bourn et al. 1999), on the island of Aldabra (Samour et al. 1987).
In 1997, 2 species of giant tortoises, both thought to be extinct since about 1840, were claimed to have been found in captivity in the Seychelles. Preliminary genetic analysis carried out by the Nature Protection Trust of Seychelles suggested the presence in captivity of a few specimens of the Seychelles giant tortoise, Dipsochelys hololissa, and Arnold's giant tortoise, Dipsochelys arnoldi (Gerlach 1998). Whether these purported species are actually distinct from the Aldabra giant tortoise remains in question (Palkovacs et al. 2002, 2003). Silhouette Island in the Seychelles was identified as a suitable site for reintroduction of tortoises identified as D. arnoldi.
At least one species of giant tortoise was historically present on Silhouette Island before European exploitation, but it is unknown whether D. arnoldi was present because the remains discovered were unidentifiable to species (Gerlach and Canning 1998). The historical presence of a now-extinct large herbivore on Silhouette is enough justification for the release of D. arnoldi into the wild as a functional replacement of the native species of giant tortoise.
However, there is very little known of the ecology and behavior of D. arnoldi in captivity and, with the species not being observed in the wild for over 150 years, nothing is known about its ecology and behavior in the wild. The potential impacts of a released population of D. arnoldi on Silhouette Island are therefore unknown. The observation of habitat modification via trampling and apparent selective foraging within captive enclosures of D. arnoldi suggests it has the potential to modify the island habitat (Gerlach 2006).
With effective monitoring, this reintroduction program can provide insights into management of captive breeding and reintroduction of populations of chelonian herbivores. It also provides the potential to evaluate herbivore reintroductions to manipulate vegetation structure and to replace a missing functional ecological guild in ecosystems.
The aim of this study was to produce a comprehensive and accurate description of the activity and diet preferences of a released D. arnoldi population and therefore to inform the potential long-term impact that a released population would have on vegetation characteristics.
METHODS
Study Area
Silhouette Island, Seychelles, was chosen as the site of the D. arnoldi breeding program. Silhouette is situated 20 km northwest of Mahe and at 20 km2 is the third largest island in the Seychelles archipelago. Because of its mountainous topography Silhouette has been the subject of minimal development by man and supports a human population of around 140 people. The island has been declared a biodiversity hotspot and lies within the Silhouette Marine National Park.
The captive breeding facility provided 6 adult D. arnoldi (1:1 sex ratio) with a 200-m2 pen containing natural vegetation, wallows, and various other natural features such as rocks and logs to provide shade and landscape diversity. Attempts to plant trees and shrubs resulted in them being eaten or trampled by the tortoises.
During their time in the captive breeding enclosure, the tortoises appear to have significantly altered the vegetation species composition within the pen (Gerlach 1998), suggesting selective foraging.
On 7 December 2006, the first steps were taken toward establishing a wild breeding population of D. arnoldi when 5 adult animals (3 males and 2 females) from the original captive population were transported by boat and released in an area of coconut forest in the Grand Barbe area of Silhouette. This area was chosen because of its relative inaccessibility and wide habitat range: freshwater marsh, coconut forest, mangrove, beach, rainforest, and river. It houses the largest area of seminatural marsh and woodland on the island. This area is only accessible for a limited number of weeks per year by sea and is also bordered on the inland side by mountainous rainforest consisting of 5 peaks, the highest being Mount Dauban (740 m), and a river, limiting dispersal and providing protection from poaching (Gerlach 2006). Mean annual temperature is 26.8°C, with mean annual rainfall of 2500 mm at sea level increasing to 5000 mm at higher altitudes (Fleischmann 1997).
The study area was an approximately 13.9-ha area adjoining the release site that encompassed the home ranges of the released population. Observations were made by JWP from 6 June 2007 to 28 June 2007, a period when southeast trade winds dominate, resulting in low rainfall and high winds.
The tree composition was primarily coconut (Cocus nucifera), breadfruit (Artocarpus altilis), and puzzle nut (Xylocarpus moluccensis), reflecting the semimanaged nature of the area.
The D. arnoldi population consisted of 3 adult males (M1, M2, M3) and 2 adult females (F4, F5). The exact age of the individuals was unknown. F5 was thought to be the youngest at approximately 45 years old. Individual tortoises were located at dusk the evening prior to their sampling by following tracks in the grass. Tortoises appeared to show a high degree of fidelity to sleep sites, and individuals were always located at dawn in the same spot that they had occupied the previous dusk. The individual tortoises were identified using a photographic key, and differentiated according to shell shape, size, and the presence of scars and shell damage.
Diet
The methodology of this study with regard to evaluation and statistical analysis of diet was based on a study of Testudo horsfieldii, the steppe tortoise (Lagarde et al. 2003).
First, the relative abundance of plant species present at the site was censused. Thirty quadrats (1 m2) were placed randomly along 5 random 300-m transects throughout the Grande Barbe site. Plant species were identified using keys provided by The Nature Protection Trust of the Seychelles. The number of plants of each species in each quadrat was multiplied by the mean aboveground biomass of 20 individual plant samples of each corresponding species in order to estimate plant species biomass. Each part (flower, stem, fruit, and leaf) of each recorded plant species was also weighed in order to ascertain the mean biomass (based on 20 randomly selected samples) of individual plant parts. These were weighed using a Kern CM 60-2N pocket balance (maximum = 60 g, d = 0.1 g) in the field to eliminate desiccation errors. For plant species that weighed above 60 g, the plants were dissected into smaller pieces, and the sum of the smaller pieces combined in order to ascertain the total aboveground biomass.
During observations, the number of plant parts of any given plant species were recorded. Then, using the previously calculated mean biomass of each plant part of each plant species, the total plant biomass consumed was calculated.
Only plants < 1 m tall or with edible parts < 1 m above ground were recorded. Anything taller than this was deemed to be above the tortoises' reach. One meter was the maximum reach of the largest individual (M3), ascertained by measuring the height from the ground of suspended food items taken. Fallen leaves and fruit from trees and other taller plants were also recorded.
Cocus nucifera and Cymbopogon sp. were within reach of the tortoises but the leaves were too large to ingest. For these species 10-cm2 samples were used to calculate biomass as these were estimated to be suitable sizes for ingestion. When larger pieces of these leaves were consumed, the area was estimated in multiples of 10 cm2, and these portions were recorded as individual leaves. The term biomass refers to biomass of plant material within the tortoises' reach.
Plant species were classified as graze, browse, or opportunistic (fallen fruit and leaves) in terms of how the tortoises fed upon them. The proportion that each feeding mode made up of the diet was calculated as the mass of any one of the categories consumed divided by the total mass consumed.
Sampling events began before sunrise and ended when the individual took up a sleep position at dusk. Individual tortoises were located the evening prior to their sampling in order to minimize search time on the day of sampling. Tortoises were always found in the same spot at dawn which they occupied the previous dusk. However, because no nocturnal sampling occurred, nocturnal activity cannot be completely ruled out.
Active time was taken as the total time from initial arousal postsunrise until movement into a sleep spot position at dusk, minus all inactive periods during the sampling event. An individual was judged to be inactive when not actively foraging or showing any obvious movement. All individuals exhibited some activity during each sampling event. Each individual was subject to a total of 48 hours of observation with the exception of F5, which was subject to 60 hours of observation. F5 was chosen randomly from the females to be subject to supplemental observation in order to minimize the difference in total observation time between males and females because of the additional male present.
Pilot observation suggested that human presence within 5 m resulted in disturbance; therefore all observations were made using binoculars from greater than 5 m but less than 15 m.
Statistical Analysis
The proportion of any given plant species in a tortoise's diet (pi), based on aboveground biomass, was calculated as the percentage of estimated fresh vegetation consumed for the species (g) divided by the total estimated fresh matter consumed for all species (g). A pi was calculated for each plant species consumed over the entire sampling period for each individual, sex group, and the total population. Taxonomic availability of plant species (qi) was calculated as the percentage of fresh mass (g) of each species of plant relative to the total fresh mass (g) of plant material recorded during the vegetation survey of the site. Hunter's index (Hi = pi/qi; Lagarde et al. 2003) was used to determine whether plant species were preferred (Hi > 1) or avoided (Hi < 1). Hunter's index, unlike χ2 tests, takes into account spatial heterogeneity of the vegetation structure and interindividual diet heterogeneity (Lagarde et al. 2003).
Simpson's reciprocal index (1-D) (Magurran 2005) was calculated to assess the diversity of species consumed by the tortoises, where D = ∑{(ni[ni − 1])/(N[N − 1])}, the probability that 2 randomly selected 1-g plant samples of tortoise diet belong to the same species, where ni = the number of individuals of the ith species and N = the total number of individuals. Simpson's reciprocal index can vary from 0 to 1, increasing with diversity. Evenness (E) was calculated as (1/D)/S, where S = species richness (Magurran 2005), and can vary from 0 to 1, increasing with evenness. The evenness value when used traditionally (i.e., assessing the diversity of a site), indicates whether there is an even distribution of species based on abundance values. Our analysis used total biomass values of each consumed species. This indicates the degree of specialization in the tortoises diet (i.e., a low evenness value indicates that the majority of fresh mass consumed by the tortoises came from a small proportion of the total number of plant species consumed).
Feeding rate was calculated as the mass of plant material eaten by each individual per daily active time. The biomass of plant material eaten was estimated as the number of plant parts consumed from any species during active time multiplied by the mean biomass of that plant part, as ascertained from the vegetation census.
Two sample t-tests were used to compare male and female values for total mass consumed, mean feeding rate, number of species consumed, mean activity time per day, mean foraging time per day, and 1-D and E values. T-tests were conducted using Minitab 13. Probability value was set at p ≤ 0.05.
RESULTS
The total population of 5 tortoises was observed feeding on 19 of the 27 different plant species available to them (Table 1; Fig. 1). Only 26 available plant species were present; however, Tacca leontopetaloides was treated as 2 separate species during data analysis depending on whether the specimen was adult or juvenile, as this appeared to impact on selectivity toward this species. The foraging strategy was found to consist of 57.9% opportunistic feeds (fallen fruit and leaves), 41.37% graze, and 0.68% browse. The population selectively foraged upon 9 species: Boerhavia repens, Brachiara umbellata, Lanana camera, Tridax procumbens, Artocarpus altilis, Tacca leontopetaloides, Loposchoenus hornei, Turnera angustifolia, and an unidentified species. These 9 species formed a greater proportion of the tortoises' diet than their availability would predict, based on the Hunter's index (Table 1).
Citation: Chelonian Conservation and Biology 8, 1; 10.2744/CCB-0728.1
The results of a correlation analysis carried out to test whether there was a correlation between diet selectivity (Hunter's index) and taxonomic availability based on biomass (qi) showed a nonsignificant relationship (correlation coefficient of 0.02) as did a correlation analysis of diet selectivity and plant species abundance (correlation coefficient of −0.003).
The population exhibited a relatively high diversity of species consumed (1-D = 0.593), but the evenness value (E = 0.121) for species consumed was low. The diet varied in the relative biomass of species consumed, with A. altilis and T. procumbens dominating the diet (Table 1; Fig. 1).
Males were observed feeding on 17 plant species (Table 2), with females feeding on 11 species (Table 3). The males foraged upon 9 plant species that the females did not: Stachytarpheta urticifolia, Epipremnum aureum, Lanana camera, Cocus nucifera, Tacca leontopetaloides, Tabernaemontana coffeoides, Ochrosia oppositifolia, Xylocarpus moluccensis, and Cymbopogon sp. The females foraged on 2 species that the males did not: Boerhavia repens and Brachiara umbellata.
Males consumed a slightly higher diversity of species (1-D = 0.491) than females (1-D = 0.431) and also consumed a slightly lower evenness distribution of species (E = 0.116) than the females (E = 0.121), suggesting that males are slightly more selective than females. The additional female sampling event did not result in an increase in the number of plant species consumed. Males (Table 2) and females (Table 3) showed preference for 6 and 7 plant species, respectively, with only 2 species preferred by both sexes: T. procumbens and T. angustifolia. Mean values for total mass consumed, feeding rate, and species richness consumed were all calculated for both sexes.
The mean foraging rate for the population was 3.86 g/min of active time, with a mean daily active time of 257 min, and a mean time spent foraging per day of 175.2 min (67.93% of mean active time) (n = 5 individuals, 21 days, 15,120 min) (Table 4). There was no significant difference between males (n = 3) and females (n = 2) for total mass of mean plant matter consumed (T = 1.73, df = 2, p = 0.23), feeding rate (T = 1.76, df = 2, p = 0.22), species richness consumed (T = 0.77, df = 2, p = 0.52), diversity (1-D) of species consumed (T = −0.63, df = 2, p = 0.59), and evenness (E) of species consumed (T = −0.66, df = 1, p = 0.63). Statistical comparisons were similarly nonsignificant upon exclusion of M2, a male with values substantially different from the other 2 males (Table 4) and who occupied an isolated forest habitat patch.
DISCUSSION
The data collected during this study indicate clearly that these tortoises are strictly herbivorous. No observations (n = 21 animal-days) were made of the consumption of any animal life. It is likely that some invertebrates present on the plant material were consumed but not intentionally. These tortoises are strict grazers. Females were never observed browsing and males seldom did so. These results contradict those found in captivity (Gerlach 1999). In captivity, D. arnoldi, when given the choice always chose browsing over grazing. This phenomenon in captivity may be due to curiosity as browsing opportunities were not a regular feature of the enclosure. It is also possible that the lack of browsing in the released population is due to the plant species present naturally or due to the human management of the release site. There is a lack of shrubbery, with trees dominated by coconut (C. nucifera), A. altilis, and X. moluccensis, all of which are tall species lacking edible parts accessible to the tortoises. Tortoises are also opportunist in terms of scavenging fallen fruit and leaves from A. altilis. Although there were no observations made of the foraging of fruit other than A. altilis, the fruit of X. moluccensis were found in fecal samples.
Nine of the 26 plant species available were selected preferentially (above their relative biomass) by the tortoises. The preference shown for these 9 plant species (Tacca leontopetaloides, Artocarpus altilis, Tridax procumbens, Boerhavia repens, Brachiara umbellata, Lanana camera, Lophoschoenus hornei, Turnera angustifolia, and an unidentified species) as indicated by Hunter's index confirms that D. arnoldi is a selective forager. The results of a correlation analysis support the statement that D. arnoldi do not simply selectively forage those plant species that are present in high biomass or high abundance. Of the 8 plant species available at highest biomass (A. atilis, Stenotaphrum dimidiatum, S. urticifolia, T. procumbens, T. leontopetaloides, X. moluccensis, Wedelia trilobata, Mariscus dubious), all except 2 species (A. atilis and T. procumbens) were avoided (Table 1). Of the 8 most abundant species (S. urticifolia, M. dubious, T. procumbens, Ipomea sp., X. moluccensis, W. trilobata, Euphobia urta, and Paspalidium germinatum), all except one species (T. procumbens) were avoided (Table 1). With the exception of T. procumbens, which exhibits both high taxonomic availability and abundance, and A. altilis, which exhibits high taxonomic availability (Table 1), each selectively foraged species contributed less than 2% of the overall biomass and plant species abundance. Factors other than biomass availability and individual abundance determine food selection.
In general, it is more energetically expensive to feed upon large numbers of small nutrient-poor plant species than larger nutrient-rich ones (Parsons et al. 1994), providing that energetic output while feeding is similar. Whether this is actually the case for the Grande Barbe population of D. arnoldi is dependent on the biochemistry of the plant species. However, biochemical information is lacking for the majority of plant species recorded during this study. Of the 9 selectively foraged plant species, T. procumbens was the most preferred (having the highest Hunter's index) overall. Why was this species selectively foraged? It has a high energetic value and has an above-average energy conversion rate in rabbits (Omoikhoje et al. 2006). Its high energy conversion rate may be the factor that makes it one of only 2 species selectively foraged by both male and female tortoises. Another preferred species, A. atilis, was only available to tortoises as fallen fruit and leaves from the trees. This species provides a large and succulent energy-rich fruit (Obasuyi and Nwokoro 2006), but the leaves were highly desiccated. In addition to selectively foraging on A. altilis fruit, tortoises also appeared to actively seek out the desiccated leaves. The consumption of desiccated leaves is counterintuitive as they are likely to be low in nutrients (Glander 1978). However, the breakdown of carbohydrates into less complex sugars during the desiccation process (Obasuyi and Nwokoro 2006) may be a reason for the animals to eat these leaves. Another species, T. leontopetaloides, was preferred in spite of the fact that, when used in traditional medicines, it needs to be leached several times to make human consumption safe (Quah 2003). At Grande Barbe, despite adult T. leontopetaloides being of a suitable size for consumption, only juvenile plants were fed upon. Juvenile plants may simply be easier to process or more nutritious, or they may contain lower toxin levels than adults (Glander 1978).
Nutritional data on the other 6 selected species – Boerhavi repens, Brachiara umbellata, Lanana camera, Lophoschoenus hornei, Turnera angustifolia, and the unidentified species – are limited. These species may be particularly nutritious or may carry high levels of valuable vitamins or minerals. Biochemical analysis of plant species would be required to test this. Alternatively, it could be argued that the bias toward these species may simply be due to the animals being present in patches where these species were relatively abundant. However, as the tortoises exhibited relatively large ranges and covered all the areas where vegetation was sampled, this is unlikely (Fig. 2). In addition, the areas where vegetation was sampled were representative of the habitat used by the population.
Citation: Chelonian Conservation and Biology 8, 1; 10.2744/CCB-0728.1
Male Aldabra giant tortoises, due to their larger size, have a higher metabolic rate than females (Hughes et al. 1971). It is highly likely that a similar difference exists in the closely related D. arnoldi. Males are therefore likely to require a higher net energy intake than females.
Mating, nesting, and oviposition behaviors were observed during the study. It is possible that during this time there was a dietary shift in females in order to compensate for the high energetic output involved in egg production (e.g., increasing calcium uptake) (Shine 1980), nesting behavior, and the loss of feeding time due to nesting behaviors and oviposition (Brent-Thomas et al. 1999). The energetic costs of breeding in chelonians are generally lower in males than in females (Boglioli et al. 2003). However, the increased activity required to locate, pursue, and mate with females likely puts extra energetic stress on the males. Males may therefore also shift diet in the breeding season to maximize energy intake.
No significant difference was found between males and females in regard to overall biomass consumed or feeding rate. This indicates that the extra energetic input likely needed by the males is not obtained by an increase in consumption of plant material but is achieved by selectively consuming high-energy plants.
Male and female D. arnoldi selectively forage on similar numbers of species (6 and 7 respectively) and their diet had similar diversity (0.491 and 0.431, Simpson's reciprocal index) and evenness (0.116 and 0.121, Simpson's evenness) scores. However, male and female diet preferences differed. Only 2 preferred species are shared between sexes: T. procumbens and T. angustifolia. Higher male energetic requirements may explain diet differences, and their preference for A. atilis, which forms 65% of their intake. Fruits of this species yield a high mass of high-energy food (Obasuyi and Nwokoro 2006). Four of the five individuals' home ranges overlapped considerably (Pemberton, unpubl. data), ruling out the possibility that the animals show fidelity to small exclusive patches that may contain only particular plant species (Fig. 2). The sex differences in plant selectivity therefore likely resulted from the different selective pressures acting on male and female life histories. If the difference in diet preferences between the sexes is independent of breeding condition and is present all year round, then it is likely due to differences in metabolic rate between males and females.
One of the key impacts of the release of herbivores, particularly relatively large ones, into areas where they have not been present is that on the vegetation structure (Veblen et al. 1992). In the captive environment this species can modify vegetation structures in a relatively short period of time (Gerlach 1998). However, these results were based upon a much higher population density in captivity than in the release site. In the wild similar impacts would be expected to take a longer period of time, particularly given the small size of the released population.
One issue is that the dominance of grazing in the Grande Barbe population may be due to the lack of available browse (i.e., they may preferentially browse given the choice [Gerlach 1999]). Another question that follows is whether they are responsible for the lack of browse within the study site (i.e., whether they removed the low shrubbery during their 6 months in the environment).
The lack of available browse in the study site likely does have an impact on the apparent lack of browsing behavior exhibited by D. arnoldi. However, data on the vegetation structure of Grande Barbe prior to the release of the population (Royo and Simpson 2001) indicates that there has not been a significant alteration in the vegetation structure during the first 6 months of release.
It is most likely that the lack of browse is due to artificial management of the site, which is maintained by wardens for volunteers.
Within the release site there is browse available for D. arnoldi to forage upon, which they appear not to, indicating that the species is primarily a grazer. However, anecdotal evidence of browsing in captivity when given the choice has been presented (Gerlach 2006).
There are 3 possible explanations for this: that in captivity the irregular and novel experience of having browse available results in D. arnoldi browsing out of curiosity; that in captivity there is a lack of adequate graze available within the enclosures, resulting in D. arnoldi browsing when the opportunity is presented; or finally that there is a lack of suitable browse available within the release site. Given that captive D. arnoldi are fed with plant species that are available at the release site, the presence of browsing in captivity is most likely due to a lack of adequate graze within the captive enclosures.
Assuming D. arnoldi is a grazing species, significant impact on the ground and herb layers is expected. Two of the 9 plant species selectively foraged by the group of D. arnoldi are confirmed invasive species, T. procumbens and T. angustifolia. These 2 species make up 13.5% of the ground biomass of the site. One likely beneficial aspect of the release of D. arnoldi on the vegetation structure of the release site is a reduction of these 2 species via heavy predation, as these species produced the highest and fourth highest Hunter's index values, respectively, with T. procumbens making up 36.9% of the tortoise diet. However, it is unclear whether this increased predation would allow an increase in native flora. It is likely that through heavy grazing of particular alien species and moderate grazing of native species that the overall plant biodiversity and potentially, in turn, overall biodiversity of the site would increase in line with the intermediate disturbance hypothesis (Connell 1978).
The diet of the Grande Barbe Dipsochelys arnoldi population, showing the relative percentage of the species in the diet.
Map of Grande Barbe showing dominant vegetation structure and ranges of the Dipsochelys arnoldi population. M1, M3, F4, and F5 were each observed in each of the differing areas of vegetation structure within the border of their combined home range. (Map by the Nature Protection Trust of the Seychelles, 2008).
