Habitat Use and Diet of Astrochelys radiata in the Subarid Zone of Southern Madagascar
Abstract
Astrochelys radiata is one of the threatened tortoise species of Madagascar. We studied their home range size, use of shelters, and diet in Tsimanampetsotsa National Park. The goal was to identify suitable habitats and food chemistry. These data could serve as bases for future evaluations of the suitability of habitats under altered environmental conditions, such as degradation and climate change. The study was performed with the help of radio-tracking in two different types of vegetation: dry deciduous forest on sand and spiny bush on limestone between November 2007 and October 2008. Monthly range sizes were 2.3 ha in the littoral forest and 1.4 ha on the limestone massif when calculated with the kernel method. Based on the minimum convex polygon, the values were 1.0 and 0.5 ha, respectively. Monthly range size did not vary significantly between sexes, but range sizes were larger in the littoral forest than on limestone. A. radiata chose shelters under trees with large trunk diameters (> 15-cm DBH) and lower heights than trees in representative samples in the habitats. Plant items eaten and not eaten by this species were analyzed for primary and secondary plant chemicals. A. radiata consumed a wide variety of plants. Its diet was composed of leaves (91%), flowers (5%), and fruits (4%) from 109 plant species. During the dry season, high energy content was the most important factor for food selection by A. radiata.
Madagascar has a unique herpetofauna with some 400 species, of which more than 90% are endemic (Glaw and Vences 2007). This includes four threatened species of endemic terrestrial tortoises, all listed on the International Union for Conservation of Nature (IUCN) Red List for Endangered Species. In the past, the Radiated tortoise, A. radiata, has been abundant and was often found along roads in the southwestern and southern parts of Madagascar. As such, it has served as a symbol of Madagascar's south. Habitat destruction, illicit exploitation for food, for supposed medicinal purposes, and for the pet trade led to a reduction in its distribution by about 20% during the past 25 yrs (Nussbaum and Raxworthy 2000; O'Brien et al. 2003; Irwin et al. 2010). Tortoises are killed and consumed by the Vezo and Antanosy, who occupy the northeastern and southeastern limits of the species' range. In contrast, tortoises are taboo (fady) for the Mahafaly and Antandroy. This taboo seems to be largely responsible for the survival of the species on the Mahafaly and Karimbola plateaus (Nussbaum and Raxworthy 2000; Lingard et al. 2003). Apart from local consumption, the international pet trade also contributes substantially to the decline of the species in its natural habitats (Ganzhorn 2011). These activities resulted in the classification of A. radiata as “Critically Endangered” according to the IUCN Red List criteria (Leuteritz and Rioux Paquette 2008). Today, the tortoise's natural distribution is limited to xeric spiny forests from south of the Onilahy River to the Cap Saint Marie with isolated populations as far east as Andohahela National Park (O'Brien et al. 2003; Sage 2003; Leuteritz et al. 2005; Pedrono 2008).
Despite its status as “Critically Endangered” and flagship for southern Madagascar, more information is needed to arrive at sustainable management plans for A. radiata. Recent studies have focused on the distribution, harvest, reproductive ecology, population genetics, and phylogeny of this species (Durrell et al. 1989; Van der Kuyl et al. 2002; O'Brien et al. 2003; Leuteritz et al. 2005; Rioux Paquette et al. 2005; Le et al. 2006; Seui 2006; Fritz and Bininda-Emonds 2007; Rioux Paquette and Lapointe 2007; Paquette et al. 2007, 2010; Rakotondrainy 2008; Hammer and Ramilijaona 2009; Paquette and Lapointe 2009; Rasoma et al. 2010; Hammer 2013). Despite the large number of studies, little is known about the natural history of A. radiata, such as home range size, habitat, and food requirements in different types of habitats. Although possibly unsuitable habitat structures can be envisaged, chemical properties of food and their changes are not obvious; but food chemistry, as measured in the field, can provide hints of the nutritional requirements of the animals for proper ontogenetic development and successful reproduction (Nagy et al. 1998; Hazard et al. 2010). Foraging theory has stimulated research on the diets and foraging behaviors of a variety of animals. The basic assumption of foraging theory is the optimization of the energy budget (Schoener 1971; Pyke et al. 1977). However, food and consumers do not consist only of energy. Nutritional values, nutrient balancing, and water contents could explain the choice for food (Raubenheimer and Simpson 2004). In addition, many plants contain toxic compounds that render them impossible or very expensive to process (Karban and Myers 1989).
Terrestrial tortoises are classified as herbivores or omnivores that can persist on low-quality diets (Woodbury and Hardy 1948; Zug 1993; Joshua et al. 2010). Nevertheless, some studies indicate unexpected specializations and feeding on a very limited subset of plants available to the animals (El Mouden et al. 2006), indicating some kind of preferences or limitations due to energy, mineral, or specific nutrient concentrations (Oftedal and Allen 1996; Nagy et al. 1998; Leuteritz 2003; Hazard et al. 2010).
The current anthropogenic pressures limit habitat availability. Grazing by livestock leads to habitat degradation (Ratovonamana et al., in press). Anticipated temperature increases threaten to reduce the suitable habitats for the species even further (Ratovonamana et al. 2011). These changes will result in changes in floristic, structural, and chemical vegetation characteristics. Thus, it is important to determine the species' present habitat and food requirements in habitats within the center of its distribution. These areas are assumed to represent suitable habitats for the species with population densities ranging from 27 to 5700 individuals/km2 (Leuteritz et al. 2005). Data from these habitats (such as compiled by Leuteritz 2003) could then be used in the future to investigate whether or not marginalized habitats would satisfy the requirements of this species in terms of habitat and food characteristics.
Therefore, the aims of this study were the following: first, to describe sources of variation in individual range size and the use of shelters in two different habitats; and second, we wanted to describe the floristic and chemical composition of the tortoise food in relatively undisturbed habitats. These data can be used as a baseline for comparisons of the habitat suitability in degraded areas or habitats modified otherwise. The results should identify the suitability of different habitats under scenarios of future habitat change.
METHODS
Study Site
The study was carried out in the Parc National de Tsimanampetsotsa, located in southwestern Madagascar (24°03′–24°12′S, 43°46′–43°50′E), 85 km south of Toliara at the western escarpment of the limestone Mahafaly Plateau. The study area is situated between the camp “Andranovao” (24°01.578′S and 43°44.238′E) and the Grotte de Mitoho (24°02.973′S and 43°45.095′E), 38 to 114 m above sea level (asl).
The vegetation of the study region is xerophytic and can be divided into three principal formations starting at the coastal plain and moving inland (Mamokatra 1999; Ratovonamana et al. 2011): 1) a formation resting on sand and thin reddish clays and occupied by Didierea madagascariensis (Didiereaceae) and a variety of Euphorbiaceae and Burseraceae, 2) an area at approximately 50 m asl and near the foot of the Mahafaly Plateau, dominated by sparse vegetation, with the exception of Salvadora angustifola (Salvadoraceae) in close vicinity to the soda lake (Lac Tsimanampetsotsa), and 3) an area on the limestone Mahafaly Plateau. Rising to 200 m asl, there is an abrupt shift to a spiny bush formation and the families Didiereaceae, Euphorbiaceae, and Burseraceae dominate. Trees on limestone are distinctly smaller in height and occur in much lower density, resulting in much reduced vegetation cover and biomass production than in the littoral forest. More details of the study site are provided by Hammer and Ramilijaona (2009), Andriatsimietry et al. (2009), Rakotondranary et al. (2010), Rasoma et al. (2010), Bohr et al. (2011), and Ratovonamana et al. (2011). The study was performed in the littoral forest formations on sand close to the lake (littoral forest) and in spiny bush on limestone (Fig. 1).
Citation: Chelonian Conservation and Biology 12, 1; 10.2744/CCB-0909.1
Rainfall was measured with rain gauges. Rainfall in the region is highly seasonal and rarely exceeds 400 mm/yr, with most rains falling between December and February (Donque 1975). As a general classification, the year is divided into a wet season from December to April and a dry season from May to November, even though several years can pass without any measurable or significant rainfall. The region experiences considerable variation in daily mean temperatures, ranging between 17° and 34°C (Ratovonamana et al., unpubl. data, 2011). Temperature was measured with temperature loggers (Hydrochron IButton; Dallas Semiconducter, Dallas, TX), placed at 1.5 m on the southern (shady) side of a tree.
Radio-tracking and Study Period
For the estimation of range sizes, tortoises were fitted with radio-transmitters (Biotrack). Transmitters were fixed to the back of adult tortoises as described by Boarman et al. (1998). Between 4 and 9 animals of each sex were tracked simultaneously in the littoral forest on sand at the western base of the limestone massif, and on the limestone plateau itself. Animals were followed during the wet and in the dry seasons. The locations of the animals were recorded by triangulation every hour for 8 hrs/d. Global Positioning System (GPS) coordinates were taken for each record. Tortoises were tracked for 4 d/mo (dry season: October–November 2007 and October 2008; wet season: March–April 2008) (Fig. 2). The GPS locations within each of the two habitat types (littoral and limestone) were analyzed using the Home Range Extension in ArcView 3.3 (Rodgers and Carr 1998).
Citation: Chelonian Conservation and Biology 12, 1; 10.2744/CCB-0909.1
Estimation of Monthly Range Size
Monthly range sizes were estimated by the Adaptive Kernel Method (AKM) based on the 95% confidence interval and by the Minimum Convex Polygon (MCP) method. The AKM yields comparable results to the conventional MCP method but is more robust with respect to the biological independence of observations and tracking errors (Worton 1995; Kernohan et al. 1998; Boyle et al. 2009). The AKM gives an indication where the animal spends most of its time, while the MCP represents the conventional methods for range estimates. Some of the animals could be tracked over the whole study period. Others lost their tags and different animals had to be used to continue the study.
Cumulative range estimates were calculated for 10 to 150 all-day follows using the AKM to determine at how many fixes the range estimates reached an asymptote. Due to the application of the 95% confidence interval in the AKM, the size estimation of ranges fluctuates widely at smaller sample size but stabilizes quickly with increasing sample size. In the littoral forest, range size kept increasing with increasing number of fixes (Fig. 3). In the limestone area, range size remained more stable (Fig. 4).
Citation: Chelonian Conservation and Biology 12, 1; 10.2744/CCB-0909.1
Citation: Chelonian Conservation and Biology 12, 1; 10.2744/CCB-0909.1
In order to account for the different number of fixes per individual, we used only animals for which 30 fixes were completed per mo. We defined the monthly range of an individual by 30 fixes. This area does not represent the actual monthly home range but rather relative home range size that can be used for standardized comparisons. For animals followed for more than 1 mo/season, the monthly values were averaged. Thus, each individual entered the analysis only once per season.
Shelter
Radiated tortoises were less active during the cool and dry season and seemed to aestivate under conditions that are not well-defined yet. In case an animal was considered to be inactive and remained at a given spot for extended periods of time (up to 4 d during this study), we considered this spot to represent some kind of shelter. Most of these sites were located underneath trees. Occasionally animals were found in crevices. These were not considered any further. Each tree, where an individual was found sleeping, was identified and its diameter and height were measured. For comparisons, the same measurements were recorded in standardized vegetation plots established in the two types of habitat (Ratovonamana et al., unpubl. data).
Diet
The composition of the diet of A. radiata was recorded opportunistically by direct observation during the radio-tracking activities. When a tortoise was seen feeding, the observer identified food items with binoculars without disturbing the tortoise. Foraging was observed for about 15 min or until the tortoise stopped feeding. Thus, the feeding data are composed of varying numbers of observations from different individual tortoises.
Plant Samples
We collected plants at the time when tortoises were found feeding on any given item. We collected parts of the plant that were identical to the part consumed. These could be the stems, leaves, flowers, and fruits. For comparisons, we collected plant items that were available to the animals but that we never saw to be consumed by A. radiata. During the wet season, we collected only 2 species that we did not see to be consumed by the tortoises. Due to rapid changes in phenology during the end of the wet season, we were unable to assign items as “nonfood” and collect them at the end of the wet season because they were gone by then, and plants had changed their properties in such a way that it seemed impossible to use these items for comparisons between “food” and “nonfood” during the wet season. Each plant species was collected at the time when it was seen to be eaten. Each sample was used only once and only for the analysis of samples either during the wet or during the dry season. Plants were identified within an associated botanical PhD Thesis (Ratovonamana et al., unpubl. data). One hundred fifty-five samples were collected and analyzed. From each sample, 5 g of dry material was collected. Plant species were identified in the field at least to family level. Further determinations were made in the herbarium of the Parc Botanique et Zoologique de Tsimbazaza, Antananarivo, and the herbarium of the Direction des Resources Forestières et Piscicoles ( = FO.FI.FA), Antananarivo.
Chemical Plant Analyses
We analyzed samples from plants eaten and plant not eaten by the tortoise for the following dietary components: water, neutral detergent fiber (NDF), acid detergent fiber (ADF), ash, nitrogen, soluble protein, sugar, total phenolics condensed tannins and the presence of alkaloids. Plant samples were dried in the sun or in a drying oven, ground to pass a 2-mm sieve, and dried again overnight at 50°–60°C prior to analyses. The water content was measured in the field as the difference between fresh weight and dry weight of the sample. Samples were analyzed for NDF and ADF (Goering and Van Soest 1970; Van Soest 1994) as modified according to the instructions for use in an ANKOM FIBER ANALYZER. Hemicellulose was calculated as the difference between NDF and ADF. After analyses of fiber components the residue was combusted at 600°C for 4 hr to measure the ash content. Total nitrogen was determined using the Kjeldahl procedure. Multiplying N by 6.25 can convert total nitrogen to crude protein. Concentrations of soluble sugar were determined as the equivalent of galactose after acid hydrolization of the 50% methanol extract. This measurement correlates well with concentrations obtained with enzymatic analyses of glucose, fructose, and galactose (Ganzhorn and Tomaschewski, unpubl. data). The digestible energy content was based on the conversion factors for protein and carbohydrates to energy as follows: 1 g crude protein = 1 g carbohydrates = 4.1 kcal = 17.15 KJ (Schmidt-Nielsen 1979). Total energy content of 1 g of a plant sample was calculated as (% crude protein + % soluble sugar + % hemicellulose) × 17.15 KJ/100. Analyses of total phenolics (Folin and Ciocalteau 1927) and condensed tannins (measured as equivalents of quebracho tannin with the butanol method; Oates et al. 1977) were based on water extracts (Bollen et al. 2004; Stolter et al. 2009). We performed three different qualitative tests for alkaloids (Dragendorff, Wagner, Mayer; Cromwell 1955). Alkaloids were assumed to be present in a sample if two or three of the tests showed a positive reaction. If only one reagent indicated the presence of alkaloids, we assumed that this represented a false positive reaction. Biochemical analyses were carried out at the Institute of Zoology, Department of Ecology and Conservation at Hamburg University.
Statistical Analyses
Data were tested for normality with the help of Kolmogorov-Smirnov 1-sample test. Square-root transformation was applied to the range data prior to analyses of variance (ANOVA) or t-tests to improve normality. If residuals deviated from normality we applied nonparametric tests. Tests were performed with the help of SPSS (1999). When the same data set was used for multiple pairwise comparisons, p-values were adjusted through Bonferroni correction.
RESULTS
Monthly Ranges
Based on the AKM, monthly home-range sizes varied between 0 ha (animals remaining stationary over ≥ 4 d) and 11 ha (Table 1). Range estimates based on minimum convex polygons (MCP) were smaller than estimates based on the adaptive kernel method and varied between 0 and 5.25 ha (Table 2). The range sizes estimated by the two methods are very closely correlated and the relationship can be expressed by the following regression: Size (in ha, as measured by MCP) = 0.44 × size (in ha, as measured by AKM) − 0.04 (R2 = 0.91, p < 0.001, n = 55).
Two animals remained almost stationary during the dry season and moved only within a few square meters within the month. Both did not remain at the same spot but changed positions. These low values occurred during the dry season. However, the largest monthly ranges were also recorded during the dry season, with 2 animals moving between 5 and 11 ha (Tables 1 and 2).
According to ANOVA, the interaction between the independent variables “sex”, “season”, and “habitat” was significantly related to monthly range size (F = 4.96, p = 0.03). Further analyses were hampered by low sample size. In pairwise comparisons of the pooled data, monthly ranges were larger in the littoral than in the limestone habitat (Tables 1 and 2; t = 2.08, p = 0.04 and t = 2.33, p = 0.02 for the AKM and MCP, respectively). Individuals differed substantially in their ranging activities in the wet as well as in the dry season. Some males had very large monthly ranges and larger ranges than females (Fig. 5), but monthly ranges did not differ significantly between female and male A. radiata and did not differ between the two seasons even though individuals differed substantially in their ranging activities in the wet as well as in the dry season (t < 1.2, not significant for all comparisons). Range overlap was not calculated because we could not track all animals that used the area.
Citation: Chelonian Conservation and Biology 12, 1; 10.2744/CCB-0909.1
Shelter
Twenty-seven tree species were recorded as shelters used by tortoises at Tsimanampetsotsa NP (Table 3). Nineteen species occurred in the littoral and 11 species in the limestone habitat. Five species of tree were used as shelter at both type of habitat (Alluaudia comosa, Commiphora mahafaliensis, Euphorbia stenoclada, Operculicarya decaryi, and Tamarindus indica). Sheltering trees in the littoral habitat were taller than in the limestone habitat (Table 4).
In the littoral forest, trees used for shelter had larger diameters at breast height (DBH) but were lower in height than trees found in the vegetation plots (DBH: t = 19.30, df = 254, p < 0.01; height: t = 8.81, df = 250, p < 0.01; Table 4). In the limestone area, the DBH also differed between used trees and trees in the vegetation plot (t = 7.22, df = 178, p < 0.01) but height differences did not show a significant difference between trees providing shelters and trees in the vegetation plot (t = 1.74, df = 178, p = 0.08).
Food Composition
Of the 155 samples collected, 70.3% were eaten and 29.7% had not been seen to be eaten by A. radiata. The food samples consisted of bark (0.9%), fruits (3.7%), flowers (4.6%), and leaves (90.8%), with 35.8% of the samples coming from herbaceous species, 30.3% trees, 24.7% shrubs, and 9.2% liana (Appendix 1).
The number of animals seen feeding varied systematically with time of day and season (Fig. 6). During the rainy season, most tortoises were feeding in the early morning. Fewer animals fed after 1000 hrs. Between 1200 and 1300 hrs, nearly all animals were resting in the shade. Feeding started again in the afternoon until 1800 hrs. Feeding observations also varied seasonally, with fewer animals seen feeding during the dry season.
Citation: Chelonian Conservation and Biology 12, 1; 10.2744/CCB-0909.1
Comparisons of the chemical composition between food items and nonfood items were only possible for the dry season. During the wet season, the phenological characteristics of food items changed quickly. During this time of the year, it was difficult to define items not eaten and to compare them with items that were eaten; therefore, we did not collect a sufficient number of items classified as “not eaten” during the wet season to allow comparisons of the chemical composition of food and nonfood items. During the dry season, items consumed by the tortoises contained more digestible energy than items not seen to be eaten (Mann-Whitney U-test: Z = 2.89, p = 0.004; Table 5). The content of hemicellulose was also higher in food than in nonfood items, though this difference was not significant statistically (Z = 1.65, p = 0.10). No other chemical component differed significantly between these 2 types of plant items.
During the wet season, food items had lower concentrations of phenolics than during the dry season (Z = 3.71, p < 0.001). The chemical composition of food items did not differ for any other component between the wet and the dry season.
DISCUSSION
Documentation of habitat use patterns may help to understand the habitat requirements of a species and thus may contribute to land management decisions for conservation purposes. To this aim, we investigated the monthly range sizes of the Radiated tortoise as a function of dry and humid season, vegetation formation, and sex. Ranges were larger in the littoral forest than on the limestone plateau. There were no significant differences in the size of monthly ranges, neither between seasons nor between males and females in this study, even though the ranges of male A. radiata were slightly larger than those for females during both seasons. The lack of seasonal differences is unexpected because Radiated tortoises are less active during the dry than during the wet season (Hammer and Ramilijaona 2009; Rasoma et al. 2010), as are Pyxis a. arachnoids, a sympatric species (Walker et al. 2007). Ranges overlapped extensively between males and females. According to the conventional interpretation of mating systems as derived from ranging data, this suggests a promiscuous mating system (Ostfeld 1990). In this system it is to be expected that males have larger home ranges than females, in order to look for mating opportunities (e.g., McRae et al. 1981; Barrett 1990). This should be true at least during the mating season (November and December).
It was also unexpected that ranges were larger in the dry deciduous forest on sand than in the less productive limestone area. Density estimates of Radiated tortoises indicated higher densities in the dry deciduous forest on sand than on limestone (Hammer and Ramilijaona 2009; Rasoma et al. 2010). Also, herbaceous food availability is higher in the dry deciduous forest on sand than on limestone (Ratovonamana et al., in press). These differences in densities should be reflected in smaller ranges on sand (where the densities are higher) than on limestone. Yet, this was not the case.
The high variation of monthly range sizes within and between individuals seems to be common in tortoises (McMaster and Downs 2009). This might reflect mixed strategies of time and energy optimization (Schoener 1971). Animals could either maximize food intake by searching food over large surfaces or they could simply remain inactive or aestivate and save energy. Based on the very variable ranges of the different tortoises, we assume that these different strategies are not specific for certain habitats, sex, or a certain season. We did not quantify food availability in the littoral forest as compared with the vegetation of the limestone plateau. But vegetation cover, especially in the herb layer, is much lower on limestone than on sand. The smaller monthly ranges of A. radiata on limestone indicate that the animals have to be considered time minimizers that do not invest energy in searching for food over large areas when food is scarce.
Shelters used by A. radiata were between rocks, most often at the base of large trees. In areas where the tortoises' ranges overlapped, the same shelter could be used by several tortoises. The most important factor for choosing a shelter seemed to be related to the shadow provided by overarching trees. In both habitats (limestone and littoral), A. radiata chose the base of trees with large DBH but low heights. Plant species associated with shelters most frequently were Commiphora mahafaliensis, Gyrocarpus americanus, and Opuntia stricta. All of them were well-represented in the littoral habitat. Tortoises were found associated with the first species mainly during the dry season. It grows as a bundle of stalks and provided protection against the heat while all other trees lost their leaves. The second species was used during the humid season. It is also an important food species, and thus tortoises might combine the need for shelter with a nearby food resource. O. stricta was used for food and shelter in both seasons. Roots and stumps may decrease the ability of predators (e.g., dogs) to dig tortoises out of their shelter. Shelters may also protect the animals against the sun and high temperatures and thus reduce the loss of water (McGinnis and Voigt 1971; Bulova 2001). Conservation of water and energy are important for tortoises (Henen et al. 1998). Tortoises used several different shelters, occupying them from several minutes to several weeks (Riedle et al. 2008). Daily and seasonal movements among shelters may be influenced by variation in shelter microclimate because tortoise may chance shelters if they are not suitable for thermoregulation (Bulova 2001). The variability observed in activity, the ranging pattern, and the impression that shelters are important for thermoregulation suggest a strategy to optimize energy expenditure through behavioral variation (Huey and Tewksbury 2009).
Foraging theory has stimulated research on the diets and foraging behaviors of a variety of animals. The basic assumption of foraging theory is the optimization of the energy budget (Schoener 1971; Pyke et al. 1977). However, food and consumers do not consist only of energy. Nutritional values, nutrient balancing, and water contents could explain the choice for food (Raubenheimer and Simpson 2004). In addition, many plants contain toxic compounds that render them impossible or very energetically expensive to process (Karban and Myers 1989).
Terrestrial tortoises are classified as herbivores or omnivores that can persist on low-quality diets (Woodbury and Hardy 1948; Zug 1993; Joshua et al. 2010). Nevertheless, some studies indicate unexpected specializations and feeding on a very limited subset of plants available to the animals (El Mouden et al. 2006), indicating some kind of preferences or limitations due to energy, mineral, or specific nutrient concentrations (Nagy et al. 1998; Hazard et al. 2010). In particular, desert tortoises in North America (Gopherus agussizii) avoid food rich in potassium (Oftedal and Allen 1996). The need to excrete surplus minerals may be relevant in southwestern Madagascar because the groundwater tastes bitter, suggesting high mineral concentrations in the system. These high concentrations should be reflected in the chemical plant composition. Unfortunately we do not yet have any information on the mineral content of plants of this region.
In our study, A. radiata fed on a large number of different plant species and food types. The protein concentration of food items is of prime importance for animals (White 1993), but we did not find any indication for food selection on the basis of nitrogen or protein concentrations. The average nitrogen concentrations of the food and nonfood items were within the range of protein concentrations of food items found to be consumed by other tortoises; therefore, we assume that A. radiata did not need to discriminate based on nitrogen concentrations (Nagy et al. 1998; Hazard et al. 2009, 2010). In contrast to nitrogen concentrations, the caloric value of food items of the tortoises was higher in food than in nonfood items. We could not control for food availability or assess food digestibility, so our interpretations must be limited. Nevertheless, the results are consistent with the current thinking in studies on wildlife feeding and nutrition (Robbins 1993) and optimal foraging theory, which assume that animals optimize energy intake (Pyke et al. 1977). The impact of secondary substances is more difficult to interpret because effects of these components can be complex (Foley and Moore 2005; Wallis et al. 2012). Condensed tannins have been identified as one of the most important groups of secondary substances that prohibit nutrient extraction from food stuff in a quantitative way (Robbins et al. 1987; Cork and Catling 1996; DeGabriel et al. 2009). But neither phenolics, nor condensed tannins or alkaloids, differed between food and nonfood items during the dry season. The concentrations of these substances could have been below the threshold above which they become effective feeding deterrents, or phenolics and tannins can also be used to counteract gut parasites (Glander 1982; Huffman 2001).
In conclusion, the data from the ranging pattern and the chemical food composition indicate that A. radiata optimizes its energy budget. We were unable to arrive at a comprehensive answer about the factors that limit the animals with our approach to analyzing the two energy components (ranging and food) as separate questions, but an optimal foraging approach would be most promising for future studies.
Study sites for Astrochelys radiata in different types of habitat: littoral forest on sand (upper pictures); and spiny bush on limestone (lower pictures; photos by J. Rasoma).
Monthly mean temperature (right axis; black dots) and mean rainfall per month (left axis; bars) between 2006 and 2009; modified from Ratovonamana et al. (2011).
Change of home range size with increasing number of radio-telemetry fixes of four individual tortoises (Astrochelys radiata) in the littoral habitat based on the Adaptive Kernel Method.
Change of home range size with increasing number of radio-telemetry fixes of four individual tortoises (Astrochelys radiata) in the limestone habitat based on the Adaptive Kernel Method.
Monthly male and female Astrochelys radiata home-range size in the littoral zone (left) and in the limestone habitat (right).
Diurnal variation in feeding records of Astrochelys radiata during in the wet and the dry season.
