Ecological aspects of Rhinoclemmys punctularia punctularia (Testudines: Geoemydidae) in two localities in northern Brazil

Rhinoclemmys punctularia punctularia is the only turtle of the family Geoemydidae with part of its range in Brazil. The species is semi-aquatic and widely distributed across the Amazon basin but is still poorly studied. We evaluated aspects of the population structure, habitat use, seasonal activity, and diet of the species during one year, from the beginning of the rainy season to the end of the dry season, in two headwater localities in the Tocantins and Tapajós river basins, in the state of Pará, Brazil. We captured 45 individuals, 14 in the Tocantins basin and 31 in the Tapajós basin. The size-class frequency distributions of captured individuals did not differ between the two samples. The turtles were most frequently captured in first-order streams, followed by lentic (third-order dammed streams) and second-order streams. The capture rate did not vary significantly between sampling periods (start and peak of the rainy season; start and end of the dry season). The trapping effort to capture a single individual was much higher in the Tapajós basin (1,215 trap-hours) than in the Tocantins basin (136 trap-hours). Overall, the low capture rates confirm the secretive nature of the species in Amazonia and its apparent rarity. The individuals analyzed in this study consumed mainly plants.


INTRODUCTION
Rhinoclemmys punctularia punctularia (Daudin, 1801) is currently the only species of the family Geoemydidae with part of its known distribution in Brazil (Rhodin et al. 2017).It is a relatively small-sized freshwater turtle with up to 260 mm carapace length and semi-aquatic behavior (Rueda-Almonacid et al. 2007).Its known geographic range comprises about 3,000,000 km 2 (Ferrara et al. 2017), from eastern Venezuela to the Brazilian states of Amazonas, Amapá, Roraima, Pará, Tocantins, and Maranhão (Dornas et al. 2011;Silva et al. 2011;2016;Ferrara et al. 2017).Although generally inhabiting rivers of the Amazon basin, individuals of this species were recently recorded in the Atlantic Forest, in the Brazilian states of Bahia, Rio Grande do Norte, and Rio de Janeiro.However, these records have been questioned and alternatively attributed to introduced individuals (Siciliano et al. 2014;Valle et al. 2016), to a relictual population (Siciliano et al. 2014), or to a new species (Ferrara et al. 2017), because of the discontinuity in the distribution of known populations (Valle et al. 2016).
The conservation status of Rhinoclemmys p. punctularia is considered to be as Least Concern in the Brazilian list of threatened fauna (Vogt et al. 2016) and by Rhodin et al. (2018), but its global status has not yet been officially determined by the International Union for Conservation of Nature -IUCN (https://www.iucnredlist.org/en).There is evidence of consumption of R. p. punctularia by indigenous people in the states of Amazonas and Roraima (Souza-Mazurek et al. 2000;Ferrara et al. 2017).Considering the rapid growth of human-induced impacts in the Amazon region (Phillips et al. 2008;Fagundes et al. 2018), it is important to obtain more information on the population structure and ecology of this species, specially in the more degraded areas of southeastern Amazonia.In this context, we sampled R. p. punctularia in two headwater localities in southeastern Brazilian Amazonia and analyzed their habitat use, population structure, seasonal activity, and diet.

Study areas
Our study was conducted in two localities in the state of Pará, Brazil: one in the municipality of Itupiranga, in a headwater area of the Tocantins River (henceforth Tocantins basin), and another in the municipality of Itaituba, in the headwater areas of the Tapajós and Jamanxin rivers (henceforth Tapajós basin) (Figure 1).The Euclidean distance between the two sampling areas is about 800 km, while the hydrologic distance is about 1,900 km.Except for dammed streams (see below), sampled streams were classified following Horton's scale modified by Strahler (Petts 1994), which ranks streams according to their size and branching hierarchy from the water source.In the Tocantins basin, we sampled 18 waterbodies (three waterbody types): two first-order, three second-order, and 13 third-order streams.In the Tapajós basin we sampled 17 waterbodies (five waterbody types): three first-order, three second-order, five third-order small streams, one sixth-order, and five lentic streams (Figure 2).These latter correspond to third-order streams dammed during the building of the federal highway BR-230 (known as the Transamazônica Highway, a former attempt of the federal government to promote regional development accross Brazilian Amazonia).VOL.52(4) 2022: 307 -314 ACTA AMAZONICA Sampling followed the general technical guidelines and criteria determined by the federal environmental agency (Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis -IBAMA) for the environmental licensing of the Marabá Hydroelectric Power Plant and São Luiz do Tapajós Hydroelectric Power Plant, which potentially impact the local biota in the Tocantins and Tapajós basins, respectively.In the Tocantins basin, we sampled each waterbody on three occasions, in 2010: at the peak of the rainy season (March) and at the beginning (June) and the end (September) of the dry season.In the Tapajós basin, we sampled each waterbody on four occasions: at the beginning of the dry season (June/July), end of the dry season (September/October) and beginning of the rainy season (November/December) in 2012, and at the peak of the rainy season (March/April) in 2013.

Turtle trapping and data collection
Turtles were captured using single-chamber funnel traps (1 m long × 0.45 m external diameter × 0.20 m entrance diameter; plastic mesh 5 × 1 mm).As it was a prospective study, designed for multi-species capture, we used fresh fish as bait (Balestra et al. 2016), placed inside a sealed plastic bottle with tiny holes, so that trapped animals had no access to the bait.On each sampling occasion we installed between 10 and 15 traps (depending on stream size) on the margins or in the center of each stream.In lotic streams, we installed the traps at an average distance of 50 m from each other over a 500-m stretch.Traps were operated for five continuous 24-h periods and checked daily, in the morning.Two individuals opportunistically found while moving on land were manually captured (Tapajós basin only).
We marked each captured individual with a combination of notches in the marginal scutes (Ferner 1979).Sex was ascertained by visual inspection of the tail.Individuals with a longer and thicker tail were classified as males (Balestra et al. 2016).Specimens that were too small for sex identification were classified as juveniles.For each individual, we measured straight-line carapace length (SCL), maximum carapace width (MCW), maximum plastron length (MPL), and maximum plastron width (MPW) with a 300-mm Vernier calliper to the nearest 0.1 mm, and obtained body mass with a 2.5-kg spring scale, to the nearest 5.0 g.Individuals captured in the Tocantins basin were kept in plastic boxes with water at ambient temperature for 24 hours to obtain feces samples for diet analysis.We also obtained a spontaneous regurgitation (considered to be stomach contents) from a single individual.Feces and regurgitation samples were preserved in 70% ethanol.After handling, all turtles were released at the point of capture.

Dietary assessment
The dietary samples (feces and regurgitation) were examined under a stereoscopic microscope.Food items were quantified and identified to the lowest possible taxonomic level, based on consultation with experts.Items were classified into plant material (leaf fragments, seeds, flowers, stems, and unidentified plant material), animal material (fish scales), periphyton (algae), and unidentified materials.The volume of each item was determined by water displacement, using 1-ml syringes with 0.01-ml graduations.
Due to the low number of dietary samples collected, we pooled the samples from all individuals (males, females, and juveniles, all captured in the same first-order stream in the Tocantins basin), from all three sampling periods.For each food category, we calculated the frequency of occurrence (percentage of the total number of samples in which a given food category occurred), numerical frequency (percentage of each food item relative to the total number of food items per category), and relative volume (percentage of the volume of each food category in relation to the total food volume) (Hyslop 1980;Marrero 1994).

Statistical analysis
We used a chi-square test to determine whether the sex ratio of turtles sampled from each basin differed from 1:1 (Zar 1996).We used a Shapiro-Wilk test to check if straight-line carapace length (SCL) followed a normal distribution, and a Kolmogorov-Smirnov test to compare the SCL distributions between basins.We used a Kruskal-Wallis test to assess if SCL differed between males and females in the Tapajos and Tocantins basins, and to test if within-basin abundance differed among sampling periods (three in the Tocantins and four in the Tapajós basin).In the Tapajós basin, we also used a Kruskal-Wallis test to assess if abundance differed among first-, second-, and third-order streams (Sokal and Rohlf 1995).For this analysis we pooled the data on number of captures in all third-order streams (dammed and not dammed).The adopted significance level for the statistical tests was 0.05.We  2 for details on the streams).This figure is in color in the electronic version.
ACTA AMAZONICA performed all statistical analyses using the "stats" package in the R software (R Development Core Team 2021).

RESULTS
We captured 14 individuals of R. p. punctularia (Figure 3) in the Tocantins basin and 29 in the Tapajós basin.Two additional individuals were found fortuitously during the fieldwork in the Tapajós basin, on two different occasions, while we were walking along forest trails, on the way to the trap sites.Other chelonian species were captured alongside Rhinoclemmys p. punctularia, in smaller numbers: Mesoclemmys raniceps (Gray, 1856), in both the Tocantins and Tapajós basins; Mesoclemmys gibba (Schweigger, 1812) and Kinosternon scorpioides (Linnaeus, 1766), in the Tapajós basin only.
The total capture effort in the Tocantins basin was 4,010 trap-hours (818 trap-hours in the peak of the rainy season, 1,149 trap-hours in the beginning of the dry season, and 2,043 trap-hours at the end of the dry season).In the Tapajós basin, the capture effort was 77,760 trap-hours (19,440 trap-hours in each sampling period).The mean capture rate was higher in the Tocantins basin: one individual per 286 trap-hours.In the Tapajós basin, the mean capture rate was one individual per 2,681 trap-hours (excluding the two individuals found on land).There were no recaptures during the study period.

ACTA AMAZONICA
In the Tocantins basin, all captures occurred in a firstorder stream (Table 2).In the Tapajós basin, 44% of captures occurred in first-order streams, 31.5% in third-order streams (dammed and not dammed), and 20% in second-order streams, while 4.5% of the turtles were moving on land.There was no significant difference in the number of individuals captured in each waterbody type in the Tapajós basin (H test statistic = 5.38; df = 4; P = 0.07).
We captured most individuals (64%) in the dry season (21 in the Tapajós basin and eight in the Tocantins basin; Figure 6), but there was no significant difference in capture rates among sampling periods, in both localities (Tapajós basin: H = 1.65, df = 3, P = 0.64; Tocantins basin: H = 0.53, df = 2, P = 0.76).The overall capture rate in the dry season was one individual per 1,851 trap-hours in the Tapajós basin and one individual per 399 trap-hours in the Tocantins basin.In the Tapajós basin, we obtained the lowest capture rate (one individual per 9,720 trap-hours) at the peak of the rainy season, while in the Tocantins basin we obtained the lowest capture rate (one individual per 2,043 trap-hours) at the end of the dry season.
We obtained fecal samples from 10 individuals in addition to one spontaneous regurgitation.The most frequently consumed food items (frequency of occurrence) were of plant origin, mainly seeds of Myrtaceae (91%) and Inga sp.(82%), in addition to Lecitidaceae flowers (36%).The greatest relative volume also corresponded to Myrtaceae seeds (43%) and Inga sp.seeds (30%).In the regurgitation, we detected items of animal origin (fish scales), which were not present in the fecal samples (Table 3).

DISCUSSION
Our results suggest that Rhinoclemmys punctularia punctularia is an uncommon species in the sampled areasas it appears to be throughout its distribution area in the Amazon basin, despite its extensive range and the diversity of aquatic habitats it occupies (Vogt 2008;Ferrara et al. 2017).Rhinoclemmys p. punctularia is common in coastal lentic habitats along parts of the northern Brazilian coast (Siciliano et al. 2014).However, most captures during our study occurred in first-order streams, which are small, shallow, clear-water streams.Some of these ran through monodominant palm-tree stands of Mauritia flexuosa L.f. and Euterpe oleracea Mart.(regionally known as buritizais and açaizais, respectively), which are commonly found in the headwater areas in the Amazon region (Pezzuti and Chaves 2009).Streams in these areas are lotic but rarely have strong currents, differing considerably from the floodplain forests and lentic coastal lakes inhabited by R. p. punctularia on the coast of Pará (Wariss et al. 2012).They also differ from the restingas (coastal sand dune plains) where several individuals of an apparently new species of the same genus were found in Espirito Santo (Oliveira et al. 2018).
The non-normal size distribution of the turtles sampled in the Tocantins basin probably was due to the small sample size, with 89% of our sample corresponding to adult individuals.The capture of low numbers of very small or very large individuals is a common occurrence in freshwater turtle studies (Wariss et al. 2012;Brito et al. 2018;Butterfield et al. 2020;Folt 2020).Turtle samples are usually composed primarily of adult individuals due to the combination of low survival rates and the higher difficulty in capturing juveniles, which use habitats different from those typically used by the adults and have a more subtle behavior (Armstrong et al. 2018).The   ACTA AMAZONICA presence of relatively few juveniles in our samples needs to be further investigated to determine whether it is owed to a lack of recruitment, which may lead to population decline (e.g., Howell et al. 2019).
Although capture rates were higher during the dry season in both sampled river basins, the difference between seasons was not significant.Higher capture rates during the dry season were observed in a population of R. p. punctularia from Algodoal-Maiandeua Island, Pará (Wariss et al. 2012), and in a population of R. melanosterna (Gray, 1861) in the extreme northwestern region of Ecuador, near Colombia (Carr et al. 2014).Captures rates of other Amazonian freshwater turtles that live in large lakes, rivers, and seasonally flooded habitats, e.g., Podocnemis sextuberculata Cornalia, 1849 and P. unifilis Troschel, 1848, are also usually higher in the dry season, due to the aggregation of individuals in the remaining waterbodies.Besides, some species reproduce during the dry season and can aggregate in the vicinities of nesting sites (Fachín-Terán and Vogt 2004;Bernardes et al. 2014).Rhinoclemmys p. punctularia has a long, intermittent nesting period (March-April and September-December; Vogt 2008;Cunha et al. 2019), and smaller streams in the headwater areas are subject to less dramatic seasonal variation in water level.Both factors probably contributed to the fact that the distribution and activity rates of this turtle in our study areas (and therefore their capture probability) were similar throughout the year.
The data from fecal samples point to the exclusive consumption of plant items along the entire sampling period in the same stream, although the regurgitation suggests the consumption of fish, as prey or carrion.Rhinoclemmys p. punctularia is generally reported as omnivorous (Ernst and Barbour 1989;Vogt 2008;Ferrara et al. 2017), with a tendency to herbivory (Vogt 2008).Omnivorous turtles are opportunistic feeders, consuming items according to their availability, as the abundance of more nutritious plant materials (such as fruits and flowers) and animal prey varies in space and time.Plants are generally an important component of freshwater turtle diets, and some species consume a great amount of fruit, such as Rhinoclemmys funerea (Cope, 1975) (Moll and Jansen 1995), Mesoclemmys raniceps (Fachín-Terán et al. 1995), and Rhinemys rufipes (Spix, 1824) (Lima et al. 1997;Caputo and Vogt 2008).Seeds of many plant species, both terrestrial and aquatic, can resist digestion and be efficiently dispersed by turtles (e.g., Tol et al. 2017;Falcón et al. 2019), including species of Rhinoclemmys (Moll and Jansen 1995).Therefore, the high number of seeds in the fecal samples could be related to their lower digestibility (Hailey 1997), leading to an overrepresentation relative to other consumed items.It is possible that some food items of animal origin, which are easier to digest and harder to detect in the faeces, had been completely digested, thus being undetectable in the fecal samples (Caputo and Vogt 2008).The small fish scales present in the spontaneous regurgitation were possibly from fish attracted to and preyed upon inside the traps.The turtles captured during our study could have been attracted either by the fresh fish bait or by trapped live fish, which would support the notion that R. p. punctularia is omnivorous.
Rhinoclemmys p. punctularia occurred in syntopy with the chelids M. gibba and M. raniceps and the kinosternid K. scorpioides, being the most abundant species in all waterbodies sampled (except for stream 01, a first-order stream at the Tapajos River basin).In the Amazon River basin, R. p. punctularia occurred in sympatry with other three species of freshwater turtles, K. scorpioides, M. gibba, and Platemys Table 3. Dietary items identified in 10 fecal samples and one regurgitation of 11 individuals of Rhinoclemmys p. punctularia captured in a headwater area of the Tocantins River basin, in southeastern Brazilian Amazonia.Items marked with an asterisk (*) were identified in the regurgitation (fish scales occurred exclusively in this sample).N samples = number of samples containing the item; N items = total number of items found in all samples., 1792), in first-order streams and shallow pools amidst buritizal palm stands, and was the most abundant species in the sample (Vogt et al. 2009), as in our study.Niche overlap among syntopic species occurs along several dimensions (Pianka 1982), from which diet is the most important (Sabagh and Carvalho-e-Silva 2008).Although the species co-occurring with R. p. punctularia also are omnivores, K. scorpioides and M. gibba tend to carnivory (Vogt 2008;Ferronato et al. 2013), which likely reduces competition for food resources with R. p. punctularia, which tends towards herbivory (Vogt 2008;this study).

CONCLUSIONS
Rhinoclemmys p. punctularia, the only geoemidid occurring in Brazil, is one of the least known Amazonian freshwater turtles, from a biological point of view.At our study sites, in the headwaters of the Tocantins and Tapajós basins, capture rates of individuals of the species were higher in first-order, lotic streams.However, even higher capture rates have been reported for lentic coastal habitats in other parts of the species' distribution area in Brazil, which points to the need of additional studies on habitat use by R. p. punctularia throughout its range.The occasional finding, during our study, of two individuals wandering on land, possibly while crossing interfluves, also points to the need of studies addressing movement between aquatic habitats.Although spatially restricted and based on a relatively small sample, our preliminary findings on the diet of the species point to plant material as the most frequently consumed food items, which should also be investigated further, in distinct habitats and sites as well.

Figure 1 .
Figure 1.Study areas of Rhinoclemmys p. punctularia in the state of Pará, in the southeastern Brazilian Amazon, detailing the sampling sites in headwater regions of the Tocantins and Tapajós river basins.This figure is in color in the electronic version.

Figure 2 .
Figure 2. Examples of streams where individuals of Rhinoclemmys p. punctularia were captured in headwater regions of the Jamanxim and Tapajós rivers (Tapajós River basin) and the Tocantins River basin, Pará state, Brazil.A -stream 2; B -stream 4; C -stream 5; D -stream 8 (see Table2for details on the streams).This figure is in color in the electronic version.

Figure 3 .
Figure 3. Adult female of Rhinoclemmys p. punctularia captured in stream 8 (Table 2) in the Tocantins River basin in Pará state, Brazil.A -general view; B -detail of the head.This figure is in color in the electronic version.

Figure 4 .
Figure 4. Number of males, females, and juveniles of Rhinoclemmys p. punctularia captured in headwater areas of the Tocantins and Tapajós river basins in Pará state, Brazil.

Figure 5 .
Figure 5. Frequency distribution of straight-line carapace length (SCL) interval classes of Rhinoclemmys p. punctularia captured in headwater areas of the Tocantins and Tapajós river basins in Pará state, Brazil.

Table 2 .
Number of individuals of Rhinoclemmys p. punctularia (N) captured in each of the headwater streams sampled in the Tapajós and Tocantins river basins in Pará state, Brazil, and characteristics of each stream.(%) = percentage of the total number of captures.Stream order follows Horton's scale modified by Strahler.

Figure 6 .
Figure 6.Number of Rhinoclemmys p. punctularia individuals captured in headwater streams in the southeastern Brazilian Amazon in each of three (Tocantins River basin) and four (Tapajós River basin) sampling periods.Sampling periods were beginning of the rainy season, peak of the rainy season, beginning of the dry season and end of the dry season.

Table 1 .
Morphometrics of male, female, and juvenile Rhinoclemmys p. punctularia captured in headwater areas of the Tocantins and Tapajós river basins, state of Pará, Brazil.Values are the mean ± standard deviation; N = sample size.