Freshwater turtles have been drastically declining in the last decades and are certainly among the most endangered animal groups (van Dijk et al. 2000; Turtle Conservation Coalition 2011). Obviously, conservation strategies may be misguided if there is no careful observation of an organism's wild ecology. Concerning freshwater turtles, it is unfortunate that considerable ecological knowledge has been obtained for a relatively small number of species only, mainly inhabiting North America (Ernst and Barbour 1989; Frazer et al. 1991). Concerning African species, and despite a recent claim for high conservation concern (Luiselli 2008a), very few data are available, and only for a handful of species (Branch 2008), mainly Pelusios castaneus, P. niger, and Trionyx triunguis (Ernst and Barbour 1989; Gramentz 1994; Kasparek 1994; Luiselli 1998; Luiselli et al. 2006). Hence, there is a priority for field-based ecological studies of African freshwater turtles if we want to reduce their decline and extinction risks (Luiselli 2008a; Bombi et al. 2011).

From the ecological point of view, it is very interesting to test the dietary shifts that may occur in reptile species from tropical Africa depending on the alternation of wet vs. dry season. In general, it is well known that dietary habits of such different Afrotropical animals, including snakes (Luiselli 2001, 2006a, 2006b), tortoises (Luiselli 2006a), and birds (Siegfrid 1972; Dean et al. 2009), may change considerably as a seasonal response to overall desiccation of their natural environments. In theory, these seasonal dietary shifts should be particularly remarkable in species that inhabit temporary pools or water bodies that tend to dry out over the dry season. In this paper, we combine the necessity of providing wild-collected data on African freshwater turtles with the interesting ecological issue of exploring potential dietary shifts associated with seasonality in an inhabitant of temporary pools, by analyzing the diet of the African helmeted turtle (Pelomedusa subrufa). This Pelomedusidae species is widespread in sub-Saharan Africa (Ernst and Barbour 1989; Boycott and Bourquin 2008; Branch 2008), usually linked to ephemeral water systems and marshes (Branch 1998, 2008), and still very little studied in the wild. Recent genetic analyses indicate that the species as currently understood is in fact a complex of deep genetic lineages (Vargas-Ramírez et al. 2010; Wong et al. 2010; Fritz et al. 2011). Here we focus on the ontogenetic and seasonal shifts in the diet of 2 different populations of helmeted turtles from central Nigeria, with emphasis on their foraging strategies as shown through application of null model statistical analyses (Gotelli and Graves 1996). These analyses are very useful to explore resource utilization patterns (e.g., Friggens and Brown 2005). In addition, the data presented herein are particularly noteworthy because this is the first quantitative study of the wild diet of P. subrufa in the international literature, thus filling a crucial gap in the ecological knowledge of this widespread Pelomedusidae species (Boycott and Bourquin 2008).

Methods

Data were collected at 2 different study areas in Nigeria, West Africa: 1) Abuja (Federal Capital territory, geographic coordinates: 9°4′60″N, 7°31′60″E); and 2) Lokoja (Kogi State, about 15 km northeast of the metropolitan area; geographic coordinates: 7°48′0″N, 6°44′0″E). The study site in Abuja was characterized by several temporary ponds, mainly formed along the course of several streams, the principal ones being Maitama and Wuse. These streams sustain a considerable human population density in the Abuja metropolitan area but have relatively low levels of nitrate pollution (see Chiroma et al. [2007] for data on the water chemical analyses). The study site in Lokoja was also constituted by a series of ponds, which were located around the main axis of the River Benue. Both study areas are characterized by reduced water availability during the peak of the dry season, where several ponds become dry due to over 2 months of nearly total absence of rain. The climate of the study areas is tropical, with a wet season from May to September and a dry season from October to April. In Abuja, temperatures range from 18.45° to 36.05°C, with maxima in February and March and minima in July to August. Rainfall varies from 0.0 to 729.0 mm/mo, with peaks in August to September (650 to 750 mm/mo) and minima in January to February (0 mm/mo). In Lokoja, temperatures range from 19.50° to 35.90°C, with maxima in February to March, and minima in July to September. Annual rainfall is over 1230 mm, with a peak in September (250 mm/mo) and minima in January to February (1 to 3 mm/mo).

Data were collected during the years 2003 to 2008. Although data collection was rather opportunistic, we always recorded the dates of capture of each turtle and marked and measured each individual, and we were thus able to understand and reliably describe the diet composition of our study populations.

Turtles were collected by random searching along suitable microhabitats in the study areas. Turtles were captured by hand or by using nets and traps of indigenous design also used for capturing snakes (Luiselli 1998). Sex was determined by examining the tail external morphology; very small individuals without obvious external sexual characteristics were removed from analysis. Straight-line carapace length was recorded for each turtle captured, and each individual was individually marked by notching marginal scutes. Food items were obtained by fecal pellet analysis and by stomach flushing (Legler 1977; Luiselli 1998). This latter method was not applied systematically to all individuals because it proved to be time consuming and stressful for the animals in the field, especially with smaller individuals. Fecal samples were obtained from animals that defecated upon handling or in small containers where they were temporarily and individually stored immediately after capture. No turtles were purposely killed or damaged during the study process. Feces were stored in 10% formalin.

For analysis, each fecal sample was placed in a shallow dish of water, and the presence of different food items was noted. A dissection microscope and reference collections were used to classify the food remains. The percentage of turtles that ate a given food type compared with the total number of fecal samples was used for the analysis because it was difficult to count the total number of food items in feces.

In order to explore differences based on body size (carapace length), individuals were subdivided into the following size classes: males smaller than 8 cm, males 8.1 to 15 cm, males larger than 15 cm, and females smaller than 15 cm and larger than 15 cm. Only 2 size classes were used for females because they reached significantly larger sizes than males (Akani et al., 1994–2005), and the number of small females captured was very low at both study areas. Hence, in order to compare size classes pooling sexes together we grouped all the individuals in 2 further size classes, smaller and larger than 15 cm.

Random vs. nonrandom utilization of the prey resource by the various categories of turtles and by season, were explored by calculating Pianka's (1986) overlap formulas for all pairs of turtle categories and for the same category from different sites. We randomized the original species utilization matrices from which Pianka's overlap was calculated by shuffling the original values among resource states. We used 2 randomization algorithms (RA2 and RA3) of Lawlor (1980) because they are particularly robust for niche overlap studies (Gotelli and Graves 1996). RA2 tests for structure in the generalist–specialist nature of the resource utilization matrix by conserving guild structure but destroying observed niche breadth (Gotelli and Graves 1996). RA3 tests for guild structure by conserving niche breadth for each species but destroying guild structure manifested by the resource utilization matrix's zero structure (Gotelli and Graves 1996). For each pair of species, 30,000 random Monte Carlo permutations were generated. This amount ensured that algorithm biases were avoided (Lehsten and Harmand 2006). Niche overlap values were calculated for each of these randomly generated matrices, and turtle category-pair summary statistics were computed (Friggens and Brown 2005). Actual overlap values were then compared to the distributions of expected values. Structure was assumed when p_obs<exp ≤ 0.05 (Gotelli and Graves 1996). In all cases, equiprobable resource use was assumed a priori in the analyses. Apart from null models, all other statistical tests were 2-tailed, with alpha set at 0.05, using Statistica (StatSoft, Inc, 2001) software.

Results

Overall, analyzable diet contents were obtained from 138 individuals in Abuja and 102 individual turtles in Lokoja (Table 1). Respectively, another 37 and 39 individuals were processed, but either they did not defecate or their fecal contents were not identified. Recaptured animals were not processed again. The summarized dataset for the 2 study areas is given in Table 1. In both sites the diet consisted mainly of fish, tadpoles, and invertebrates, but small-sized terrestrial vertebrates such as mice, birds, and frogs were also regularly consumed (Table 1). Identified preyed rodents all belonged to the family Muridae. In Abuja, 5 were Hylomyscus sp., 1 was Mus minutoides, 1 was Rattus rattus, and 2 were Hybomys sp. In Lokoja, we identified 3 Hylomyscus sp., 1 Hybomys sp., and 2 Lemniscomys sp. Among identified bird remains, there were 2 Charadrius pecuarius, 1 Quelea quelea, and 2 Passer griseus in Abuja, and 2 P. griseus, 1 Anthreptes collaris, 1 Hylia plasina, and 3 Ploceus sp. (probably P. nigricollis). All these birds were small to medium-sized Passeriformes quite common at the study areas, thus suggesting that helmeted turtles forage opportunistically upon passerine birds when available. Eaten lizards were all Scincidae (3 Panaspis sp. and 3 Mabuya affinis in Abuja; 5 M. affinis in Lokoja), apart from a juvenile Agama agama that was preyed upon in Lokoja. As for snakes, all 7 individuals preyed upon in Abuja were Natriciteres sp. (i.e., small marsh snakes of the family Colubridae); whereas, in Lokoja 4 were Natriciteres sp., 2 were newborn Afronatrix anoscopus (Colubridae), and 2 were Typhlops sp. (Typhlopidae).

Table 1 Diet composition of Pelomedusa subrufa at 2 study areas in Nigeria, by sex, body size, and season. Food items are shown in terms of absolute frequency. For fungi and plant materials presence only is indicated (+). M, males; F, females.

View Fullscreen

Unfortunately, it is not possible from fecal analysis to separate active hunting of terrestrial vertebrates from scavenging; hence, although Pelomedusa subrufa is documented to drown birds (Branch 2008), it remains unknown whether this behavior is exceptional or normal. Fungi and aquatic plant or algae remains were found in only a few individuals, and at least for the latter it was not clear whether they were ingested secondarily by turtles, e.g., while extracting an animal prey from a tangle of plants or algae.

Variations in diet composition between wet and dry season were remarkable at both sites, while the main trends were similar between sites (Fig. 1). For instance, fish, tadpoles, and frogs were eaten essentially during the wet season; whereas, small mammals, birds, lizards, snakes, and terrestrial invertebrates were eaten during the dry season. Overall, there was a clear trend for turtles to forage upon aquatic prey in wet season and upon terrestrial prey in dry season at both sites (Fig. 2).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

The interseasonal (wet vs. dry season) food niche overlap between turtles of similar body size and sex, captured at the 2 study areas, is given in Table 2. All the considered size categories showed a clear dietary shift between wet and dry season (particularly the smallest males and the largest females) except for the largest males at Abuja, which showed a relatively high interseasonal diet overlap (Table 2). At Lokoja, the mean overlap value was less than that found at Abuja (0.195 vs. 0.425, respectively).

Table 2 Interseasonal (wet vs. dry season) food niche overlap between individual turtles of the same body size and sex, captured at the study areas.

View Fullscreen

At both sites, a similar pattern, although different by season, emerged from food resource overlap analyses among sexes and body size categories. During the wet season, the observed overlap values were higher than those generated by randomization, and very often they were higher than expected if the resource use was not affected by both sex and body size category interactions (Table 3). During the dry season, the observed overlap value between turtles with body size smaller than 15 cm and larger than 15 cm was significantly lesser than expected (Table 3); whereas, no statistical significance was attained in any other overlap comparison (sex and other body size categories).

Table 3 Synoptic table of null model results comparing overlap degrees among size categories and sexes using both RA2 and RA3 randomization algorithms. Obs. overlap, observed mean of Pianka's index values. Size, comparison among all the size classes (see Methods for the single class range); < 15 vs. > 15, comparison between individuals smaller and larger than 15 cm (sexes pooled); Sex, comparison between males and females (size classes pooled). Significant differences between observed and expected (mean of randomized matrices) overlap values are shown in bold.

View Fullscreen

Discussion

This study revealed several patterns of dietary relationships of an Afrotropical turtle, Pelomedusa subrufa, that had been previously little known in the wild. Our data clearly showed that both males and females of any size category and at both sites were predominantly carnivorous. Although detailed data on P. subrufa wild diet are not available, this overall result is not unexpected because this turtle was considered to be carnivorous in field books (Branch 1998) as well as in the scientific literature (Ernst and Barbour 1989; Boycott and Bourquin 2008; Branch 2008), with Rödel (1997, 1999) documenting tadpole predation even by hatchlings. In addition, predominant carnivory was also observed in other Pelomedusidae species from Nigeria, including Pelusios niger and P. castaneus (Luiselli 1998; Luiselli et al. 2004). However, Pelusios species are more linked to main river tracts and permanent water bodies (Ernst and Barbour 1989; Luiselli et al. 2006), whereas Pelomedusa subrufa is found mainly in temporary and ephemeral ponds, with estivation/brumation and dispersal to different waterholes being the terrestrial activities of this species (Branch 2008). In southern Nigeria, Pelusios species sometimes occur in sympatry with helmeted turtles (Luiselli et al. 2006), but in these cases Pelusios species have a different diet composition compared with Pelomedusa subrufa (Luiselli et al. 2004).

If in general our data mirror previous information on P. subrufa diet, a more in-depth data inspection revealed original and unexpected patterns as for 1) between-site, 2) ontogenetic, and 3) seasonal diet variations. Interestingly, all dietary patterns observed in our study were consistent between sites, thus suggesting that P. subrufa may actually behave quite constantly in terms of foraging strategy, at least in this region of West Africa. Because of the strong consistency between sites, for reasons of clarity we here discuss our data as if they came from a single population. Considering the whole dataset, our analyses highlighted 2 contrasting patterns: 1) sexes highly overlapped during both wet and dry season, evidencing no differentiation in diet strategy and spectrum between males and females; 2) body size clearly affected diet composition of turtles, but only during the dry season. Indeed, by the dry season the smaller (and presumably younger) individuals fed primarily on invertebrates and small vertebrates; whereas, larger (presumably older) ones showed a preference for homeotherm vertebrates. This evidence of food resource partitioning among individuals of different size and age when resource availability is limited can be interpreted as, but is not necessarily, an ontogenetic dietary shift related to the need for reducing intraspecific competition pressure for food. Alternatively, it may reflect a “tipping point” in size, at which turtles become large enough to successfully prey on a “novel” category of available prey of a consistent size, such as mice or small birds, which have a very consistent adult size.

In terms of seasonal variations, our dietary data are of the greatest interest. To begin with, we clearly demonstrated that P. subrufa responded to increasing dryness (which would mean a very scarce availability of water during the peak of the dry season) by shifting from a diet based on aquatic organisms to a diet based on terrestrial organisms, including small vertebrates. For instance, we confirmed several cases of predation upon birds, rodents, lizards, and snakes. All vertebrate species eaten by turtles in this study are abundant in the field, and hence we consider that the foraging strategy of helmeted turtles towards small vertebrates is merely opportunistic. We suppose that these animals are captured while going to water, with turtles ambushing them from hiding places in the mud. This strategy is analogous to that exhibited by Nile crocodiles (Crocodylus niloticus) in East African savannas (Cott 1961; Games 1990) and has already been described for P. subrufa (Ernst and Barbour 1989; Boycott and Bourquin 2008). It is also reported that these turtles may exhibit gregarious hunting behavior in these cases (Diagne and Devaux 2009). However, the novelty of our study is that this ambush strategy may not be occasional in these turtles but may be the main foraging strategy during a low food availability period, hence being crucial for these turtles to survive over a potentially dangerous phase of life. It is likely, however, that these turtles consume their terrestrial food items in water because it has been reported that they cannot eat outside water (http://www.pelomedusa.com/Diet.html). In addition, it is well known that P. subrufa may estivate when the desiccation of the water bodies where it lives is complete (Stuart and Meakin 1983; Dumont 1987; Ernst and Barbour 1989; Branch 1998).

Our null model analyses revealed that, during the wet season (when turtles fed on mainly aquatic prey) we observed higher than random overlap among pairs of turtle categories, this showing aggregation patterns over the use of some specific prey resources. Higher than random overlaps are often associated with unlimited availability of food sources in turtles and snakes (Luiselli 2006a), so that all animals may forage extensively on that available resource. On the other hand, during the dry season (when turtles fed on terrestrial organisms), there was a less than expected overlap observed among turtle size categories (less than 15 cm vs. larger than 15 cm), thus suggesting that food niche partitioning, guided by limited resource availability, as also previously documented in freshwater turtles (Luiselli 2008b), or size limitations on foraging tactics, had occurred in this case. Obviously, it is not certain that drinking birds are a limited supply; instead, it is possible that successful hunting is limited by 1) the need for a certain minimum size of the turtle (because a small turtle cannot grab, pull, or drown a small terrestrial homeotherm) and 2) relatively inefficient hunting strategies by turtles (which are not specialized predators of small terrestrial vertebrates). In any case, small homeotherms are likely difficult to access for turtles during dry season months. These main patterns (i.e., agglomerated use of an unlimited food supply during the wet season and partitioning of limited/difficult to access food supply during the dry season) are also indirectly confirmed by the fact that density explosions of small aquatic vertebrates and invertebrates (for instance, tadpoles and small fish, which are common prey for turtles) occurred at the onset of the wet season in Nigeria (Luiselli 2006b; Akani et al., unpubl. obs.).