Quantitative evaluation of animals' diets is a powerful tool for understanding niche specifications in community dynamics (Krebs 1999). In addition, studying feeding habits of natural populations can serve as a basis for making recommendations for captive husbandry, especially for species with economic importance (Fachin-Teran et al. 1995). The feeding habits of Amazonian freshwater turtle species have been studied since the 1960s but have received more attention the past 2 decades (Medem 1964; Almeida et al. 1986; Fachin-Teran et al. 1995; Balensiefer and Vogt 2006; Caputo and Vogt 2008). However, the diet of some species in the wild, e.g., Mesoclemmys helliostemma, Mesoclemmys gibba, Platemys platycephala, is virtually unknown (Souza 2004).

Turtle assemblages are common in Amazonian habitats (Fachin-Teran et al. 1995) as well as in other tropical and temperate regions (Vogt 1981; Vogt and Guzman 1988; Bjorndal et al. 1997; Alcalde et al. 2010). In living together, species tend to overlap or to partition the food resources. Overlapping usually occurs when resource availability is high (Bjorndal et al. 1997; Godbold et al. 2009; Rocha et al. 2011). Partitioning mechanisms could occur when there are microhabitat segregation of the species, interspecific differences in feeding strategies, and interspecific differences in food organism preference (Vogt 1981; Souza and Abe 1998; Luiselli 2006). Although turtle diversity is high in the Amazon basin (Buhlmann et al. 2009), little is known about resource overlap or mechanisms involved in food partitioning in those species (Fachin-Teran et al. 1995).

Podocnemis unifilis (Podocnemididae) is widespread in the tropical lowlands of northern South America, occurring in Bolivia, Brazil, Colombia, Ecuador, French Guyana, Guyana, Peru, Surinam, and Venezuela (Pritchard and Trebbau 1984). Its feeding habits are known both for wild (Medem 1960; Almeida et al. 1986; Fachin-Teran et al. 1995; Portal et al. 2002; Balensiefer and Vogt 2006) and captive individuals (Malvasio et al. 2003), but no quantitative information is available for this species in Peru (Ferronato and Morales 2012).

Phrynops geoffroanus (Chelidae) is the South American turtle with the widest geographical distribution of all freshwater turtles in the region, ranging from the Colombian Amazon to southern Brazil and northern Argentina (Ernst and Barbour 1989), where they inhabit pristine areas as well-urbanized, often polluted rivers (Souza and Abe 2000; Ferronato et al. 2009; Piña et al. 2009; Martins et al. 2010). Phrynops geoffroanus exhibits a carnivorous feeding habit in captivity (Medem 1960; Molina 1990), but, in the wild, its diet varies according to the habitat within which it lives (Medem 1960; Fachin-Teran et al. 1995; Souza and Abe 2000; Dias and Souza 2005; Martins et al. 2010). To date, no diet records of P. geoffroanus from Peru have been published (Ferronato and Morales 2012).

Mesoclemmys gibba (Chelidae) is a lowland rainforest species that occurs in Brazil, Colombia, Ecuador, French Guyana, Guyana, Peru, Surinam, Trinidad, and Venezuela (Mittermeier et al. 1978; Pritchard and Trebbau 1984). The species' diet is only known from captive individuals (Fretey 1977; Mittermeier et al. 1978).

The goal of this study was to describe the feeding habits of wild P. unifilis, P. geoffroanus, and M. gibba from the Peruvian Amazon and to verify the degree to which dietary overlap occurs among these species. To our knowledge, this is the first quantitative dietary study of freshwater turtles from Peru and the first quantitative description of wild M. gibba diet for the Amazon basin.

METHODS

The study was conducted in the Santa Rosa de Chivis community (260 m above sea level, lat 10°20′13″S, long 74°58′33″W) and surrounding areas. The community is formed by an Asháninka ethnic group and is located in Puerto Bermúdez District, Pasco Department, central Peru. It is located in the buffer zone of the natural protected area “Bosque de Protección San Matías – San Carlos.” The average annual temperature is 25.4°C, and average annual rainfall is 3500 mm. The dry season extends from May to September, and the wet season is from October to April. The typical ecological formation in the area is Humid Tropical Forest (Gaviria 1981).

Turtles were captured, twice a month, from July to November 2009. They were collected in 2 oxbow lakes (both approximately 1.7 m deep, 80 m long and 8 m wide) and in a section of the Azupizu River (10-m depth, 25-m width). We used fishing nets (nylon, mesh size 4.5 cm, 2 m deep, 40 m long), a trammel net (mesh size of the inner net was 7 cm between knots and 25 cm for the outer net, 2 m deep, 25 m long), and basking traps (nets under basking logs) to collect turtles in the oxbow lakes, and basking traps for the animals in Azupizu River. We collected the animals during the day, setting the nets and basking traps early in the morning and checking nets every 3 hrs and basking traps every hour. Turtles were identified and sexed based on their morphological characteristics by following Pritchard and Trebbau (1984) and Rueda-Almonacid et al. (2007), marked by shell notching (Cagle 1939), and weighted and measured with calipers (straight-line carapace length [SCL]). We designated all P. unifilis less than 13 cm in SCL, P. geoffroanus less than 20 cm (Souza and Abe 2000), and M. gibba less than 15 cm (Pritchard and Trebbau 1984) as subadults of unknown sex. Turtles were stomach-flushed (Legler 1977), and the items were preserved in 70% ethanol. The turtles were released within 3 hrs of initial capture in the same area where they were trapped.

Stomach contents were examined with a stereoscopic microscope, and the items were identified to the lowest taxonomic level possible. The total volume of each prey item was measured by water displacement by using graduate cylinders to the nearest 0.1 ml (Bjorndal et al. 1997). For each food group, we registered frequency of occurrence (percentage of turtles in which a given food item was found [%F]) and volume percentage (volume percentage of prey category in relation to all food items detected [%V]). The importance of each food item in the diet was quantified in each species by the index of relative importance (IRI) by integrating frequency of occurrence and volume: IRI  =  100 (Fo × Vi)/∑ (Fo × Vi), where F is the percentage frequency of occurrence and V is the percentage volume. Values near zero indicate low importance, and values near 100 indicate high importance (Bjorndal et al. 1997). Similarity or niche overlap among the species was measured by the simplified Morisita index (CH): CH  =  2 ∑ (P_ij × P_ik)/∑ (P²_ij) + ∑ (P²_ik), where j and k are categories to be compared, and P_ij, P_ik are the proportions represented by item i in such categories; values near zero indicate low similarity and values near 1, high similarity (Krebs 1999). We did not include sediment, a common finding in the stomach contents, in our analysis of IRI and CH because we considered it an incidental ingestion and not a food item; further details are depicted in the Results and Discussion sections.

We performed descriptive statistics for our results. We could not analyze our data into sex or life stage categories due to the small sample size.

RESULTS

We captured and stomach-flushed 20 P. unifilis (11 males, 4 females, and 5 juveniles), 10 P. geoffroanus (all juveniles), and 4 M. gibba (1 male and 3 juveniles). The SCL and mass (mean ± SD and range) were 15.9 ± 3.9 cm (8.4–22 cm) and 619.7 ± 382 g (90–1500 g), respectively, for P. unifilis; 14 ± 2.8 cm (9.5–18.5 cm) and 298.1 ± 158 g (101–575 g), respectively, for P. geoffroanus; and 14 ± 3.1 cm (11.7–18.6 cm) and 302.5 ± 166 g (190–550 g), respectively, for M. gibba. Most of the captures occurred during dry season and in one of the oxbow lakes.

Podocnemis unifilis diet was composed of a greater variety of items when compared with the other turtle species (Table 1). Podocnemis unifilis consumed 49 food items, including 32 animal and 17 plant items (Table 2). Although animal matter was represented by a greatest number of diet items (n  =  32 food types) and had a high %F (90%), it comprised a relatively low percentage of the overall food volume (3.9%) in the stomachs that we examined. Podocnemis unifilis consumed 16 species of fish (based on scales present in the stomachs contents; one turtle ingested a small and whole fish) and 12 species of insects, with a volume of only 1.6% and 2.3% (Table 1), respectively. In contrast, plant material corresponded to 62.9% of the volume and was found in 95% of the turtles (Table 1). Seeds from the Fabaceae (Leguminosae) family and bark were common items in P. unifilis stomachs, which yielded 70% and 85% of frequency, 28.3% and 18.8% of volume, and 42.78 and 34.57 of IRI, respectively (Table 2). Sediment also was commonly encountered, because it was present in 75% of the individuals sampled and made up 33.1% of the sample volume (Table 2). Sediment was mainly composed of mud and small rocks. Other frequent items were stem (65%F), unidentified plant matter (75%F), and Fabaceae leafs (80%F). Of the 49 food items consumed by P. unifilis, several of them showed a low frequency of occurrence in the stomach samples (Table 2).

Table 1.  Description of groups of food items found in the stomach flushings of Podocnemis unifilis (n  =  20), Phrynops geoffroanus (n  =  10), and Mesoclemmys gibba (n  =  4) in Santa Rosa de Chivis, Pasco Department, Central Peru.^a

View Fullscreen

Table 2.  Percentages of frequency of occurrence (F%) and volume (V%), and index of relative importance (IRI) of each prey item ingested by Podocnemis unifilis (n  =  20), Phrynops geoffroanus (n  =  10), and Mesoclemmys gibba (n  =  4), in Santa Rosa de Chivis, Departamento de Pasco, Central Peru.

View Fullscreen

Table 2.  Continued.

View Fullscreen

Phrynops geoffroanus ingested 23 food items, with 18 of these being animal items and 5 plant items. Sediment was present in one stomach content. Animal matter contributed to 61.4% of the volume and was present in 90% of the P. geoffroanus stomachs. Another frequent item was unidentified plant material, which was present in 90% of the animals, contributed to 29.4% of volume (Table 1), and showed an IRI of 39.49 (Table 2). Insects made up the bulk of the P. geoffroanus diet, with 13 species ingested from 7 orders (Odonata, Ephemerida, Hemiptera, Neuroptera, Coleoptera, Diptera, and Trichoptera), which contributed 54.1% of the sample volume (Table 1). Insects prey sizes ranged from approximately 10 to 40 mm. Libellulidae larvae was the most frequent item (80% of the stomach contents), which represented 37.5% of the volume, therefore, the most important item in their diet (IRI  =  44.78) (Table 2). Other frequent animal dietary constituents in P. geoffroanus stomachs included Ephemeridae larvae (50%F) and the gastropod Biomphalaria anaticus (60%F) (Table 2).

Mesoclemmys gibba consumed 18 diet constituents, with 12 animal and 6 plant items. Sediment was found in one stomach content. Animal and plant matter were present in the stomachs of the 4 M. gibba captured (100% of frequency) and represented 23.7% and 47.6% of the total volume, respectively (Table 1). Mesoclemmys gibba ingested 9 species of insects and 2 fishes (only scales were recovered), and the bulk of their diet was represented by unidentified plant material, which represented 28.8% of total sample volume (Table 2).

We observed low dietary overlap between P. unifilis and P. geoffroanus (CH  =  0.368), although, when considering the upper limit on tolerable niche overlap (< 0.4) (Bonino et al. 2009), diet compositions of M. gibba appeared to overlap with P. geoffroanus (CH  =  0.521) and P. unifilis (CH  =  0.456). The foods mutually consumed by M. gibba and P. geoffroanus included insects (Libellulidae, Tenebrionidae, Culicidae), a crustacean (Ostracoda), and plant material (Fabaceae leaf, unidentified plant material) (Table 2). The diets of M. gibba and P. unifilis overlapped in the presence of insects (Libellulidae), fishes (Leporinus sp., Hoplias malabaricus), and plant material (Fabaceae leaf, bark, unidentified plant material) (Table 2).

DISCUSSION

The feeding habits of P. unifilis from the Peruvian Amazon are primarily herbivorous, which is similar to the records of the species' diet in other regions of the Amazon basin. We found plant material in 95% of the P. unifilis sampled, and it accounted for 62.9% of the sample volume, whereas animal matter represented 3.9% of the sample volume. Fachin-Teran et al. (1995) demonstrated that P. unifilis in Guapore River in Brazil ingested 89.5% of the total volume of stomach contents in vegetal items, such as seeds, fruits, leaves, and stems. In addition, P. unifilis in Mamirauá Reserve in northern Brazil fed mainly on plant material (79.6% of the volume), with leaves, seeds, and fruits occurring most frequently (Balensiefer and Vogt 2006). Other qualitative records of P. unifilis diet (Medem 1964; Almeida et al. 1986; Portal et al. 2002; Malvasio et al. 2003) also underscore the importance of plant material in its diet (Balensiefer and Vogt 2006).

In this study, the plant item most important for P. unifilis diet was Fabaceae (Leguminosae) seeds (IRI  =  42.78). Fabaceae seeds are known for their high nutritional values, including elevated protein and fat content (Sotelo et al. 1995; Duranti and Gius 1997; Portal et al. 2002). It is possible that P. unifilis ingested these seeds due to their nutritional value and high caloric content to more efficiently reach their energetic and physiological requirements. There is evidence that marine turtles (e.g., Chelonia mydas) select food items with higher protein and nutrients levels in their diet (Godley et al. 1998; Brand-Gardner et al. 1999; Nagaoka et al. 2012).

Fachin-Teran et al. (1995) and Balensiefer and Vogt (2006) identified 30 and 15 species of plants, respectively, in P. unifilis stomach contents, whereas we found only 7 species. Perhaps the lower variety of plant species found in Peruvian turtles is due to the lesser number of sample sites, lesser types of habitat, or smaller sample size when compared with these previous dietary studies of Amazonian turtles (Fachin-Teran et al. 1995; Balensiefer and Vogt 2006). Nonetheless, our results still demonstrate that P. unifilis diet consists mostly in vegetal materials.

Sediment, represented mainly by mud, was a common component of P. unifilis diet in Peru, which accounted for 75% of frequency and 33.4% of sample volume. This likely is evidence that, at our study site, P. unifilis consumes foods that are closely associated with the riverbed and perhaps in the process, sediment is incidentally consumed. Balensiefer and Vogt (2006) also reported the presence of sediment in P. unifilis stomach contents (15% sample volume) and suggested that this may be due to incidental ingestions associated with bottom feeding or that turtles could be purposefully consuming substrate to obtain trace minerals not available in plants (Moll and Legler 1971; Beyer et al. 1994). Due to the elevated frequency and volume found in this investigation, we suggest further studies on the role that sediments play in freshwater turtle digestive processes.

The presence of a small whole fish and several large scales in the stomach contents indicates that P. unifilis can take live fish as well eat them as carrion, which also was observed by Fachin-Teran et al. (1995). Aquatic insect larvae also were part of the diet in many of the turtles in this study, which could have been eaten by active predation or neustophagia or been consumed incidentally along with plant matter (Belkin and Gans 1968; Balensiefer and Vogt 2006; Alcalde et al. 2010).

Although P. unifilis is considered primarily an herbivorous species due to the high amounts of plant matter it ingests (Belkin and Gans 1968; Malvasio et al. 2003; Balensiefer and Vogt 2006), analysis of our results suggests that it is a generalist species because we found animal items to be a common component of its diet. The consumption of animal matter by P. unifilis occurred with a large number of individuals (90%F), but in substantially lower volume (3.9%V). Similar results were obtained across the range of P. unifilis (Fachin-Teran et al. 1995; Balensiefer and Vogt 2006), which suggests additional evidence of diet mixing and possible benefits in digestive additive effects in freshwater turtles (Bjorndal 1991).

In this investigation, P. geoffroanus had a more onmivorous diet than that of P. unifilis, with animal matter and plant materials constituting 61.4%V and 32.4%V, respectively. Our results show some similarity with the diet of P. geoffroanus in another locality within the Amazon basin, where plant items and animal matter also were consumed (Fachin-Teran et al. 1995), but some of the animal groups consumed were different between our study site (arachnida, gastropods, bivalvia, and ostracoda) and that in the Guapore River (anurans, shrimp, crabs) (Fachin-Teran et al. 1995). Likewise, we found the proportion of animal:plant matter to be 2∶1 vs. 1∶1 (48% of volume each) found by Fachin-Teran et al. (1995). Contrasting with our results of animal and plant matter consumption, free-living P. geoffroanus were primarily carnivorous in Colombia (Medem 1960) and in Parana River in Brazil (Dias and Souza 2005), and the species fed mainly on Chironomidae larvae and pupa in Brazilian urban rivers (Souza and Abe 2000; Martins et al. 2010).

Although results of some studies suggest that plant material is incidentally ingested by P. geoffroanus (Souza and Abe 2000; Dias and Souza 2005; Martins et al. 2010), this was not the case at our study site because plant matter was found in 90% of the turtles and contributed to more than 32% of the sample volume. Souza and Abe (2000), Dias and Souza (2005), and Martins et al. (2010) showed that plant material was substantially less frequent (28.1%, 10.7%, and 2.6%, respectively) and occurred in much lower volumes (1.7% in Souza and Abe [2000], 1.2% in Martins et al. [2010]). We instead observed a high frequency of occurrence and high portion of the volume represented by plant material. Fachin-Teran et al. (1995) also found that plant matter, represented by seeds and fruits of legumes (35.6%V), was one of the principal dietary groups consumed by P. geoffroanus in the Guapore River in Rondonia State in Brazil.

The most important item in P. geoffroanus diet in Peru was Libellulidae larvae (IRI  =  44.78) with the presence in most of the stomach contents of the turtles (80%) and represented the majority portion of the sample volume (37.5%). Fachin-Teran et al. (1995) also found Libellulidae larvae to be a very frequent item in P. geoffroanus stomach contents in a similar Amazonian habitat. We are not sure at this time if Libellulidae larva is a preferred item or its elevated consumption is a reflection of high availability of this resource in our study area. Future studies should focus on evaluating resource availability and diet preferences in Amazonian freshwater turtles.

Fishes were commonly consumed by P. geoffroanus in Colombia (Medem 1960) and in northern Brazil (Fachin-Teran et al. 1995). In the present study, we recorded scales of fish in only one turtle, and it was probably eaten as carrion. The gastropod Biomphalaria anatinus has been found in 60% of the stomachs in our study site. Gastropods were also recorded but to a lesser extent in the diet of P. geoffroanus in polluted urban rivers (1.8%F and 2.6%F) (Souza and Abe 2000; Martins et al. 2010) and in Parana River (17.9%F) (Dias and Souza 2005). Our results are additional evidence of the opportunistic and omnivorous feeding habits of P. geoffroanus because the species diet drastically shifts according to the habitat within which it lives (Medem 1960; Fachin-Teran et al. 1995; Souza and Abe 2000; Dias and Souza 2005; Martins et al. 2010).

Captive M. gibba are known to feed almost exclusively on animal prey (Pritchard and Trebbau 1984). Fretey (1977) reported that captive individuals fed on live frogs, worms, and pieces of red meat, and Mittermeier et al. (1978) also observed that captive M. gibba consumed fish, meat, insects, dog food, and new-born mice, but sometimes would ingest plant matter (leaves of Colocacia sp). Our investigation shows that wild M. gibba is an omnivorous species, because animal and plant matter were ingested by all individuals captured (23.7%V and 47.6%V, respectively). Mesoclemmys gibba has eaten insects, fish, and crustaceans, in addition to unidentified plant matter, bark, leaves, stems, and algae, plant matter was more representative by frequency and sample volume. The presence of sediment in one stomach, which represented 28.6% of the volume, shows that the species can feed in the bottom of the oxbow lakes. In Rio Negro Basin, in northern Brazil, M. gibba has been found to feed almost exclusively on palm fruits during part of the year, which could be due to a more scattered distribution of aquatic animal prey when the forest was flooded (R.C. Vogt, pers. obs. in Caputo and Vogt 2008). We suggest future investigation on ontogenetic and seasonal variation in M. gibba diet to better understand its feeding habits.

Food partitioning has been observed in Neotropical turtle communities, in different systems, and with various species (Vogt and Guzman 1988; Fachin-Teran et al. 1995; Alcalde et al. 2010). Some reasons for partitioning could be due to specialized feeding habits of some species (e.g., Mesoclemmys raniceps is a mollusk specialist and Chelus fimbriatus is a fish specialist [Fachin-Teran et al. 1995]) in addition to differences in foraging strategies and microhabitat use (Vogt and Guzman 1988). In the present study, we observed low dietary overlap between P. unifilis and P. geoffroanus (CH  =  0.368). Although both species tend to have a generalist diet, P. unifilis fed on a large variety of items compared with P. geoffroanus, and the former preferred seeds and bark, contrary to P. geoffroanus, which fed largely on Libellulidae larvae and unidentified plant material.

Dietary overlap was observed in our investigation between P. geoffroanus and M. gibba (CH  =  0.512) and P. unifilis and M. gibba (CH  =  0.456). The 3 turtle species shared some food items in our study site (insects: Libellulidae; plant material: Fabaceae leaf, unidentified plant material), but only crustaceans (Ostracoda) and some insects (Tenebrionidae and Culicidae) were found in diets of both P. geoffroanus and M. gibba. Fish (Leporinus sp. and Hoplias malabaricus) and bark were overlapped by P. unifilis and M. gibba. However, the food overlap results could be partially inflated because we captured few M. gibba (n  =  4), and a food item eaten by one individual represents 25% of frequency of occurrence. We suggest future investigation, with a larger sample size to verify the degree of niche overlap between these species in Amazonian habitats.

Our results demonstrate that the 3 freshwater turtles have generalist feeding habits in consuming both plant and animal matter, although they differ in some of the items consumed. This pattern fits with other descriptions of turtle diets, where, although there is a tendency of these species to consume specific items, they also are opportunists by taking advantage of local availability of food items (Souza 2004). To our knowledge, this is the first quantitative diet study of Peruvian freshwater turtles, in addition is the first quantitative description of M. gibba diet for the Amazon basin.