Field reproductive biology and breeding ecology parameters are poorly known for most of Brazilian turtle species of the Chelidae (Souza 2004; Souza et al. 2006). Reproduction can be considered a critical demographic parameter for management and conservation strategies when considering long-lived organisms such as turtles (Dunham et al. 1989; Congdon et al. 1994). Life-history traits such as clutch size, body size and age at maturity, reproductive effort, growth rate, survival, and population turnover are important for evolutionary approaches as well as for conservation assessments including population viability analyses (Burke and Gibbons 1995; Pike 2006; Famelli et al. 2012). Field data on chelonian breeding seasons and reproductive traits, including clutch and egg size, are usually gathered by means of direct detection of nests, X-rays of females (Gibbons and Greene 1979), or ultrasound techniques (Kuchling 1989).

A clear lack of information regarding reproductive behavior in the wild, including mating and nesting, is evident for most Brazilian chelids (Souza 2004). Because access to field reproductive data is difficult, the knowledge for some Brazilian chelonian species has been developed through study of captive animals (Molina 1989; Novelli and Sousa 2007), but it is recognized that some stereotyped behaviors may differ because captivity cannot provide the amount of space animals have in the wild (Novelli and Sousa 2007). Thus, the knowledge of reproduction and nesting of Chelidae in Brazil is still scarce, which makes it difficult to ascertain ecological and evolutionary histories (Souza 2004).

Understanding the factors shaping the reproductive traits and behavior of a species is critical for evolutionary ecology (Shine 2005) as well as for management and conservation (Hamann et al. 2010). While trade-offs can emerge from reproductive analyses (e.g., optimal clutch size according to female body size) at both local and geographical spatial scales (Iverson et al. 1993), natural history data (e.g., nesting behavior and nest success) can provide interesting data for management strategies. For turtles, these considerations can be even more important because many species throughout the world are threatened due to human activity (Rhodin et al. 2011).

Hydromedusa maximiliani (Mikan 1825) is a freshwater turtle endemic to the mountainous regions of the Brazilian Atlantic rainforest (Souza and Martins 2009), a worldwide recognized threatened hot spot (Myers et al. 2000). The species is considered Vulnerable by the International Union for Conservation of Nature (IUCN 2011). Preliminary studies suggest that the reproductive period of H. maximiliani (including mating behavior and hatching) is associated with the rainy season (Yamashita 1990; Guix et al. 1992; Souza 1995a, 2004; Souza and Abe 1997a; Famelli et al. 2011); during this period, higher activity was evident (as judged by encounter rate) and there were higher capture rates for hatchlings and juveniles. However, this hypothesis was based on mark–recapture techniques, which could be subject to biases in animal sampling.

We present data on egg and clutch size, female age and size at maturity, and inferred breeding season based on X-rays of H. maximiliani. These are the first substantial records of the species' reproductive biology in the wild and show that reproduction is strongly associated with local climatic characteristics.

METHODS

Fieldwork was carried out from September 2007 to December 2008 and from September to December 2009 at Parque Estadual Carlos Botelho, São Paulo State, southeastern Brazil (lat 24°15′00″S, long 47°45′00″W–48°10′00″W). The 37,000-ha area encompasses pristine Atlantic rainforest physiognomies and has a complex topography with sequences of ridges and valleys drained by hundreds of rivers and streams. For the present study, we selected some of the streams sampled in previous studies of the species, in an area of approximately 250 ha, totaling 5 km of water course. More detailed descriptions of the area and of the drainage system design can be found in Souza and Abe (1997a), Souza et al. (2002), and Souza and Martins (2009).

Turtles were hand-captured and marked by notching marginal scutes for individual identification (Cagle 1939) as part of a long-term project on the species' ecology. Individuals were sexed according to external characteristics such as plastron color, plastron concavity, and tail length (Souza 1995a, 1995b; Souza and Abe 1997a, 1998; Souza and Martins 2009), and traditional body measures were taken (straight-line midline carapace length [CL], carapace width [CW], midline plastron length [PL], plastron width [PW], and carapace height [CH]; all measures at a 0.01-mm accuracy). Body mass (BM) was taken with a spring scale to the nearest 5 g.

Males were released after data collection. Juveniles (unknown sex) and female turtles were kept in captivity for 2 or 3 d and transported to a medical clinic for X-ray procedures and then released at points of capture. Animals were X-rayed with a portable 80 KV FNX X-ray machine. From September to December 2008 fieldwork was carried out 2 times a month with the purpose of identifying multiple nesting (clutch frequency) in gravid females (Gibbons and Greene 1979, Tinkle et al. 1981). Fieldwork was conducted monthly in 2009.

Previous studies reported that the growth rate of H. maximiliani was better explained by PL than by CL (Martins and Souza 2008); thus PL was used herein as an indicator of female sexual maturity. Sexual maturity is reached when individuals exhibit eggs in the reproductive system (Gibbons 1990). Age at sexual maturity was estimated using the von Bertalanffy H. maximiliani growth equation

(Martins and Souza 2008) to age (t) the smallest female PL that exhibited eggs in X-ray analyses. Thus, all females with a PL equal to or larger than the PL of the smallest female with eggs were considered mature, potentially reproductive females.

Egg length and width (EL and EW) were measured with a caliper to the nearest 0.1 mm directly from radiographs (Gibbons and Greene 1979; Guix et al. 1992), with posterior estimates of egg volume as EV  =  πxy²/6 where x and y are egg length and width, respectively (Vanzolini 1977). Clutch volume (CV) was defined by the total sum of the egg volumes of the same female. Egg shape was defined by the ratio of its length and width. Spherical eggs had ratios between 1.00 and 1.10, elliptical eggs between 1.11 and 1.99, and oblong eggs had lengths at least twice their widths (Miller and Dinkelacker 2008).

Linear regressions were performed on EL, EW, EV, and CV (dependent variables) and female body size (CL, CW, PL, PW, CH, and BM; independent variables) using log-transformed data (log₁₀). Correlation of clutch measurements with female body size were examined using log-transformed data (log₁₀), which improves linearity and homogeneity and facilitates comparisons with other studies (King 2000). The highest correlation detected was chosen to explain the relation between females through linear regression. Regression analyses were also conducted using log-transformed data (log₁₀). All statistical analyses were performed with the R program (R Development Core Team 2012). The Bonferroni adjustment for multiple comparisons was used to control for Type I error (Magnusson and Mourão 2005; Norman and Streiner 2008) by dividing our α level (0.05) by the number of comparisons (4), resulting in an adjusted significance level of 0.0125.

Analysis of variance (ANOVA) was used to compare biometric measurements from gravid females, nongravid females, and juveniles, with a post hoc Tukey test. Data were first checked for normality (Shapiro-Wilks test) and homoscedasticity (Bartlett test) with posterior adjustment of the original data through the Box-Cox transformation (Box and Cox 1964). Statistical analyses were performed with the R program (R Development Core Team 2012).

RESULTS

One hundred fourteen turtles were captured during the study period, represented by 24 males, 43 females (one of which was dead), and 47 juveniles. Turtles were intensively recaptured during all study periods. Of the 289 captures and recaptures recorded, 68 were males, 117 were females, and 104 were juveniles. Most captures and recaptures occurred during the rainy season, between October 2007 and February 2008 (n  =  154) and between October and December in 2008 and 2009 (n  =  94). The months with the lowest capture rates were the cold and dry season, June and July (n  =  20; Fig. 1). Individuals with typical hatchling characteristics (e.g., soft carapace) were captured in August, October, and November 2007 and 2008 and October and November 2011. These animals had a linear midline CL varying from 45.0 to 60.4 mm (49.8 ± 1.28 mm SE; n  =  13) and a midline PL varying from 30.0 to 43.4 mm (33.7 ± 1.11 mm SE; n  =  13).

View Figure

• Download as Powerpoint
• Download Figure

Forty-two females were X-rayed. Twenty-six gravid females (62%) were found during 3 consecutive months (September–November) in the 3-yr sampling period (Fig. 2). Of 11 females with eggs in 2007, 3 also had eggs in 2008. In 2009, 26 females were submitted for X-rays analysis and only 12 had eggs. The smallest gravid female had a PL of 100.2 mm (Fig. 3), characteristic of a 13-yr-old animal according to the von Bertalanffy equation. Although turtles were intensively radiographed, clutch frequency could not be detected since there was no difference in egg size among months. Although nesting has not been observed, consecutive X-rays analyses suggested that egg laying probably occurs from late spring to early summer, in November and December.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Juveniles were significantly smaller than adults for all biometric parameters (ANOVA, p < 0.05; Table 1). Gravid and nongravid females differed only in BM, which was 12% greater in gravid females on average (p < 0.05; Table 1).

Table 1. Biometric parameters of gravid and nongravid females and juveniles of Hydromedusa maximiliani X-rayed from September 2007 to December 2008 and September to December 2009 in Parque Estadual Carlos Botelho, southeastern Brazil. CL, straight-line midline carapace length; CW, carapace width; PL, midline plastron length; PW, plastron width; CH, carapace height (all measurements at ± 0.01-mm accuracy); BM, body mass (taken with a spring scale to the nearest 5 g). Values (mean ± SE) with the same letter in rows did not differ significantly (post hoc Tukey test, p < 0.05).

View Fullscreen

Clutch size varied from 1 to 3. Most X-rayed females (n  =  23; 55%) had 2 eggs. EL varied from 33.1 to 45.4 mm (40.3 ± 0.44 mm SE; n  =  53) and EW varied from 19.9 to 25.0 mm (22.3 ± 0.19 mm; n  =  53). Estimated EV varied from 7789.3 to 13,090.0 mm³ (10,521.1 ± 279.2 mm³ SE; n  =  53). Eggs were elliptical, with an average length∶width ratio of 1.81∶1 (Fig. 4). Egg position in the oviduct varied according to egg number. The pointed end of each egg was oriented caudally in females with 1 or 2 eggs (Fig. 4B–C), whereas in the only female with 3 eggs, the eggs were positioned orthogonal to the median line of the body (Fig. 4D). There were positive and significant relationships between female body size and clutch characteristics (r² ranged from 0.54 to 0.764; p < 0.01; Fig. 5). However, mean EL did not show correlation with any measures of females (r²  =  −0.006 for CW and 0.137 for PW). CH was significantly correlated only with CV (r²  =  0.626). Body size showed significant correlation with mean EW (r²  =  0.690) and CV (r²  =  0.703), but was not correlated with the other 2 eggs measures (r²  =  005 for mean EL, 0.211 for mean EV). The highest correlation coefficient was obtained for CL and CV (r²  =  0.764; Fig. 5).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

Seasonality in reproduction for most turtle species is associated with periods of favorable habitat conditions for energy acquisition, nesting, incubation, and hatching (Kennett and Georges 1990; Fagundes and Bager 2007; Lescano et al. 2008; Mocelin et al. 2008; Ferreira-Júnior 2009), whereas local habitat characteristics can induce variation in body size of turtles hatching in natural nests (Packard et al. 1993). For H. maximiliani, male and female activity is more frequent from October to March (Souza and Abe 1997a). In the study site, higher activity (as indicated by higher capture rates) of H. maximiliani was observed between October and February. Our data corroborate previous studies suggesting that reproduction is associated with the rainy season (Guix et al. 1992; Souza 2004; Souza and Abe 1997a; Souza and Martins 2009). Preliminary observations on H. maximiliani gravid females tracked with thread bobbins resulted in up to 70-m overland movements far from the nearest stream, suggesting that these turtles could be engaged in nesting activity (S. Famelli, pers. obs., 2009).

Data on captures and recaptures of H. maximiliani suggest that nesting and laying occur in January and February (Souza 1995a; Souza and Abe 1997a, 1997b). In this study, gravid females were detected in November and December. It is therefore feasible to suppose that part of the species' reproductive cycle (nesting and egg laying) is restricted to a 2- or 3-mo period during the rainy season. This life-history trait evidence is in accordance with earlier studies that found that reproductive timing was regulated by local characteristics such as rainfall, which create small shallow forest ponds that will be used for refuge and as feeding habitat by hatchlings (Guix et al. 1992; Souza 1995a, 2005; Souza and Abe 1997a; Souza et al. 2006).

Data on clutch frequency are difficult to obtain for some reptiles, requiring circumstantial evidence such as detecting nesting females or knowing that females are carrying eggs again (Tinkle et al. 1981). Clutch frequency and nesting success can provide important additional information, which may affect population viability over the years (Famelli et al. 2012).

Egg incubation time for H. maximiliani was previously estimated to be 250–300 d (Souza 1995a; Souza and Abe 1997a). In the present study, juveniles with a mean CL of 50 mm linear were captured from August to November 2007. Hatchling CL exhibits a positive relationship with egg size in turtles (Ewert 1979). Because the largest H. maximiliani egg size was 45.4 mm, juveniles were expected to have hatched some days later. Assuming hatching occurs in September and October (Souza 1995a; Souza and Abe 1997a), H. maximiliani egg incubation time could be up to 60 d longer than previously reported.

Reproductive parameters, including clutch size, can vary across a species' geographical range for freshwater turtles (Tinkle 1961; Iverson 2010). Clutch size is a life-history trait important for management programs, especially for species with wide geographic ranges (Gibbons and Tinkle 1969; Litzgus and Mousseau 2006). Although H. maximiliani is restricted to Brazilian Atlantic rainforest mountainous regions, local populations experience distinct climatic variations along a northeast–south gradient (Souza and Martins 2009). A phylogeographic analysis revealed northeastern and southern population clades (Souza et al. 2003), so local population adaptation may occur, which in turn could be reflected in distinct reproductive features, including clutch size.

Hydromedusa maximiliani exhibits late sexual maturity, at between 9 and 10 yrs of age (Martins and Souza 2008). Determining exact sexual maturity is an important issue in conservation and management programs (Wilson et al. 2003). The fact that gravid and nongravid females did not exhibit significant differences in body size suggests that at least some nongravid females could be sexually mature (Gibbons 1990). Nevertheless, gravid females were heavier than nongravid females, a difference that could be due to the mass of eggs.

Knowledge of South American chelid reproductive patterns is limited (Bujes 1998, 2008; Souza 2004; Fagundes and Bager 2007; Brito et al. 2009). Variation in egg size and shape may be associated with clutch size and female anatomy (Ewert 1979; Moll 1979; Godfray et al. 1991; Iverson and Ewert 1991). Hydromedusa tectifera, which is a larger-bodied species than H. maximiliani, reaching 300 mm in CL, has a mean clutch size of 11 eggs (Fagundes and Bager 2007). An increase in egg size usually is associated with a small clutch size. For turtle species with clutches as small as 1, the eggs can be larger (Smith and Fretwell 1974; Ewert 1979). However, given the low frequency of female H. maximiliani having 1 or 3 eggs, it was not possible to examine intraspecific variation in clutch size.

Trade-offs among clutch size, parental care, and egg predation can be addressed as evolutionary pressures shaping species reproductive behavior (Godfray et al. 1991). The lower predation risk may reflect a larger energetic investment devoted to the development of a few individuals as well as an increase in hatchling body size (Smith and Fretwell 1974). Analyses on reproductive pattern for Brazilian Chelidae showed that small species usually have small clutch sizes (< 5 eggs) and build shallow (up to 10 cm) nests (Souza 2004). For species inhabiting Amazonian forests habitats, eggs may be laid directly on forest litter (e.g., Platemys platycephala; Pritchard and Trebbau 1984) or between tree roots (e.g., Mesoclemmys gibba; Mittermeier et al. 1978). At least one nest of M. gibba was situated 8 m from a small river in Colombia (Mittermeier et al. 1978), whereas hatchlings of H. maximiliani can be found in flooded areas along the stream banks, where dead leaves accumulate (Guix et al. 1992; Souza 2005). Apparently, for small forest species such as H. maximiliani, the nests are built close to water, in a simple shallow groove camouflaged by dead leaves with the eggs close to the surface. The small clutch size of H. maximiliani can be related to a low egg predation risk given the highly complex habitat structure that could hide the nest. However, nest architecture remains unknown for the species.

Although apparently well-established, population viability analyses showed that this studied local population can be negatively impacted and prone to extinction according to environmental and demographic stochasticity (Famelli et al. 2012). This illustrates the importance of distinct approaches and techniques for management programs of this turtle species, considered Vulnerable by the IUCN (2011). Radiography should be an effective technique to investigate some aspects of reproductive biology of turtles, especially of vulnerable species, since animals are not sacrificed, and of species whose nests are hardly found in nature because of their reproductive strategy. Nevertheless, noncalcified eggs cannot be detected, introducing biases in results, so other methods (e.g., ultrasonography, measurement of sexual hormone levels) could be used in association with X-rays.