Brazil has about 10% of the biodiversity of freshwater turtles of the world, 65% of which can be found in the Amazon region (Mittermeier et al. 2015). The species in the family Podocnemididae are the most abundant turtles in the region. Populations have been devastated since the colonial invasion by Portuguese traders in the late 1800s, and exploitation continues today to a lesser extent for local use and trade (Gibbons et al. 2000; Vogt 2008). The most threatened species are the giant South American river turtle (Podocnemis expansa), classified as critically endangered; the yellow-spotted sideneck (Podocnemis unifilis), classified as threatened; and the six-tubercled sideneck (Podocnemis sextuberculata), classified as vulnerable (Turtle Taxonomy Working Group 2014).

Podocnemis unifilis has an extensive distribution in South America, inhabiting the Orinoco and Amazon basins of Venezuela, Colombia, Ecuador, Peru, French Guiana, Guyana, Suriname, northern Brazil, and northern Bolivia (Rueda-Almonacid et al. 2007; Vogt 2008). The species is a generalist, inhabiting large and small rivers, lakes, marshlands, and flooded forests, in all types of water: black water, white water, and clear water (Vogt 2008). They have an herbivorous diet, consuming fruits, seeds, roots, algae, grasses, and leaves of other plants; food and habitat are not considered to be limiting factors for this species in the Brazilian Amazon Basin (Fachín-Terán et al. 1995; Balensiefer and Vogt 2006). The life cycle of turtles inhabiting the Amazon River basin is directly related to the hydrological cycle of the region. The annual variation in water levels, with seasonal flood pulses, permits the turtles to exploit a wide range of habitats as water levels rise 6 to 8 m or more above dry-season levels, allowing turtles to feed within the flooded forests (Junk et al. 1989; Leite 2010). During the rising water levels, turtles of all sizes leave the rivers and lakes and enter the flooded forests. The complex structure in the forests provides more hiding places from predators and opens access to a vast banquet of seeds, fruits, leaves, and periphyton to feed on (Fachín-Terán 1992; Vogt 2008). After the waters recede in the dry season, females migrate to nesting areas. Some females remain in the lakes and nest in the clay banks of the lakes near the forest edge, while others migrate to the river and nest in high coarse-sand beaches (Fachín-Terán et al. 2006; Vogt 2008).

The Rio Trombetas Biological Reserve (REBIO Rio Trombetas) is the only conservation unit in Brazil created to protect species of Podocnemis. However, the main threats are still increasing due to harvest and trade of these species on the black market in urban centers, illegal fires in the forest, and increased boat traffic in the region. These activities have hindered local conservation actions and generated local conflicts since the REBIO Trombetas became a full protection unit (Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis [IBAMA] and STCP Engenharia de Projetos Ltda [STCP] 2004; IBAMA 2009).

Understanding movements of turtles and their habitat use throughout the year is necessary to establish effective management and conservation programs for a species. Home range is defined as the total area an animal utilizes in its routine daily activities, such as feeding, copulating, basking, and parental care (Burt 1943; Jewell 1966). The size of the home range may vary according to the size of the animal, sex, feeding patterns, and availability of resources (Slavenko et al. 2016) and is not constant for a species over its lifetime. Overlapping of home ranges among individuals in the same population often occurs, and movement patterns and social organization of a population may change over the years (Gibbons 1986). The purpose of the present study was to determine the extent to which P. unifilis travels away from the nesting beaches in order to make recommendations for management and determine priority areas to help protect this species from poaching at the REBIO Trombetas.

METHODS

Study Area. — The study was conducted in the 385,000-ha Rio Trombetas Biological Reserve (REBIO Trombetas; lat 1°22′19 ″S, long 56°51′30 ″W), municipality of Oriximiná, in the northeastern part of the state of Pará, Brazil. The REBIO is in the Amazon Basin between the Trombetas and Acapu rivers. The Rio Trombetas is a left-margin, clear-water tributary of the Amazon River, extending 800 km toward the Venezuelan border (IBAMA and STCP 2004). The REBIO is replete with numerous lakes and channels flowing into the rivers. In addition to the aquatic habitats, there are 7 high, extensive coarse-sand beaches that are inundated during the rainy season but exposed for nesting by the 3 species of Podocnemis that occur there in the dry season. The presence of these beaches and the high number of nesting turtles were the impetus for creating the reserve in 1979.

The Amazon Basin has 4 seasons, which are related to the rising and falling of the level of the rivers. In the REBIO Trombetas, the water level is at its lowest in October and November. The rains begin in late November, and the river begins to rise, flooding the forests, and continues rising as it rains more and more, leveling off with the height of the water in the flooded forests and reaching its peak in March and April. As the rains diminish in intensity, the water levels gradually recede until the forest is no longer flooded by late August, when the turtles begin congregating to migrate to the nesting beaches. Rather than a distinct dry season (given that there is rain every day in the region), the year is divided into 4 seasons according to the fluctuations in water level (Fig. 1):

View Figure

• Download as Powerpoint
• Download Figure

1. Vazante (falling water level), 18 July–17 October 2013, water level lowered nearly 5 m in 91 d, and 22 July–18 October 2014, water level lowered 5.4 m in 89 d.

2. Seca (lowest water level), 18 October–7 December 2013, water level rose only 0.87 m in 51 d.

3. Enchente (rising water level), 8 December 2013–14 April 2014, water level rose 5.31 m in 128 d.

4. Cheia (full, water level rising gradually), 15 April–21 July 2014, water level increased by 2.12 m in 98 d.

For our study, we designated the dry season, when the river was at its lowest with little fluctuation in water depth (seca) combined with the river level falling (vazante), from the middle of July to the first week of December (142 d). The rainy season was designated as the combination of the period when the water was rising (enchente) and high (cheia) from the first week of December to the second week of July (226 d).

Field Methods. — We captured 10 adult males and 10 adult females from September 2013 to January 2014, when the turtles were concentrated in the river and lakes due to the lower water level. Four females (PU08, PU09, PU10, and PU12) were captured at nesting sites; the remainder of the turtles were caught in trammel nets (130 × 3 m; 40-cm-mesh external webbing and 11-cm internal webbing). Nets were checked every 2 hrs to avoid turtle drownings and to remove fish and other animals captured to avoid attracting black caiman (Melanosuchus niger; Vogt 1980).

Each turtle was individually marked with a combination of rectangular notches cut with a hacksaw in the marginal scutes (Cagle 1939), measured, and sexed, and a transmitter was fixed to the carapace. Straight-line maximum carapace length (CL) was measured with a caliper to the nearest 1 mm and mass determined with a Pesola spring scale to the nearest 1 or 100 g, depending on the size of the scale used. Sex was determined by external dimorphic characters: males retain the characteristic bright yellow head spots, and their precloacal tail length is longer. Minimum adult size was set at 20 cm CL in males and 30 cm in females (Vogt 2008). The 10 females had a mean CL of 321 ± 18 mm SD (range, 300–351 mm), and the 10 males a mean CL of 230 ± 22 mm SD (range, 208–260 mm); females had a mean mass of 3820 ± 658 g SD (range, 3100–5000 g), and males had a mean mass of 1373 ± 328 g SD (range, 950–1900 g).

VHF transmitters (MP2; AVM Instrument Co) of 164 or 165 MHz were attached between the third and fourth costal scutes with Tubolit. The antenna of the transmitters was fixed to the marginal scutes of the carapace with Tubolit so that it would not tangle or cause problems with the turtle's movements within the flooded forest. The battery life of the transmitters was 18–24 mo. The mass of the transmitters together with the adhering material was less than 5% of the mass of the turtles, about 130 g, in order to not affect their movements (Jacob and Rudran 2004). All the transmitters were on individual frequencies, and each turtle was given a code according to the order of capture (PU01–PU20). Before releasing the turtles back at their points of capture, they were kept in fiberglass tanks to allow the potting material to dry thoroughly for 24 hrs.

The turtles were monitored directly from September 2013 to September 2014, except during February. A minimum period of 24 hrs was set between the register of consecutive locations to enable the data to be independent for statistical analyses (Swihart and Slade 1985). The areas monitored were dependent on the movements of the turtles.

Transmitter signals were located using a Yagi 3-element antenna, connected to an R2000 receiver (ATS Instrument Co.). The range of signal reception was about 1 km. The model of the receiver allowed for 8-sec intervals for procuring the 20 different frequencies to locate each turtle that was being monitored (Jacob and Rudran 2004).

Locations were made by homing in on the signal until it was strong enough to make visual observation of a marked turtle (Jacob and Rudran 2004). When it was not possible to directly observe a turtle, the antenna was disconnected from the receiver, and the point where the receiver measured the signal most intensely was marked as the locality of the turtle (Jacob and Rudran 2004). We ensured that our traveling in a boat locating the turtles with transmitters did not disturb them and affect their natural behavior (Famelli et al. 2016). We recorded the date, time, geographic position (UTM, datum WGS1984) with a GPS (Garmin GPSmap 76Cx), habitat description, and other pertinent observations about the proximity to other turtles and species of plants in the area.

To determine home range size, we used 3 methods: fixed-kernel (FK) with 95% and 50% of the locations and minimum convex polygon (MCP) with 100% of the locations, which are expressed in hectares, and the linear home range (LHR), which is expressed in kilometers. Estimates of FK and LHR home ranges were calculated based on combined locations for rainy and dry seasons, while estimates of MCP were calculated based on locations both for combined seasons and for separate seasons for turtles that had more than 3 locations in each season in the latter case. The FK represents each location graphically, with small round hills—the kernels—which are functions of the probable density. This method permits the detection of multiple centers of activity through time and excludes areas that are not visited frequently by the turtles, yielding a more accurate description of the space used (Worton 1989; Jacob and Rudran 2004). “Activity center” was used to designate a fraction of the home range with the greatest frequency of locations. The estimated area by FK50% was considered the activity center in the distributions. The degree of detail in this estimator is determined by the smoothing parameter (h), which controls the variation of the contours of the density of space utilization by each animal (Worton 1989; Jacob and Rudran 2004; Downs and Horner 2008). The h value was calculated by using least-squares cross validation (Silver-man 1986; Downs and Horner 2008). The home range size calculated by the FK estimator was not used for individuals that had 15 or fewer locations because it is sensitive to the intensity of use (Seaman et al. 1999; Famelli 2013). For this reason, 4 females (PU10, PU17, PU18, and PU19) and 2 males (PU11 and PU14) did not have their home ranges estimated by this method.

The MCP consists of uniting the most extreme points (outliers) in the distribution of the locations needing at least 3; in this way, the smallest possible polygon can be admitted without concavities (Mohr 1947). This method is susceptible to outliers and does not permit the use of information about the intensity of habitat use, usually acquired from the internal locations (Jacob and Rudran 2004). This method allowed for the comparison of home ranges between the seasons because the majority of individuals had fewer than 15 locations in each season.

In order to be able to compare our study with other studies, we calculated for each individual the minimum direct distance (direct line assuming unidirectional movements) between the most distant location points that define the LHR (Morales-Verdeja and Vogt 1997; Fachín-Terán et al. 2006; Bernhard 2010; Bernhard and Vogt 2012). The home ranges MCP100%, FK95%, FK50%, and LHR were calculated with the ArcGis 9.2 software by the Hawth tools extension. Representative maps were created from the same software 10.2.2 (ESRI 2006, 2014).

To test for a difference between the sexes in home range as estimated by FK95%, we used a Student t-test after verifying that the data were normal and had equal variances. We used the Kruskal-Wallis test (H-test) for a comparison by season and sex of home range as estimated by MCP 100% (nonparametric data).

The degree of overlap between individuals was estimated using FK95%. We also estimated the degree of overlap using FK50% in zones of intensive use (activity centers). The comparisons were made among individuals of the same sex and between the sexes using the ArcGIS software version 10.2.2 (ESRI 2014). The proportions (%) of overlapping were compared using the χ² test. Statistical tests were completed using the program R version 3.1.2 (R Development Core Team 2014). For all cases, differences were considered significant when p < 0.05.

RESULTS

A sampling effort of 3.58 m²/hr was made using trammel nets. A total of 490 locations (mean 25.8 ± 18 locations per turtle) were obtained from monitoring 9 females and 10 males during the monitoring period. One female (PU20) was never located again after her release in January 2014. We found 57% of the locations during the dry season and 43% during the rainy season. The descriptive statistics of the different estimation methods are separated by sex in Table 1, and home ranges according to the FK method are illustrated in Figs. 2 and 3.

Table 1. Estimated home ranges by different methods. MCP100% = minimum convex polygon based on 100% of locations; FK95% = fixed-kernel based on 95% of locations; FK50% = fixed-kernel based on 50% of locations; LHR = linear home range. Values expressed as means ± SD with ranges in parentheses.

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

There was no difference in estimated home range between the sexes using FK95% (t₈ = 0.25; p = 0.81; n = 13). The mean ± SD home range size of females using MCP100% was 198 ± 264 ha during the dry season and 25 ± 25 ha during the rainy season, while means ± SDs for males were 408 ± 564 ha during the dry season and 225 ± 391 ha during the rainy season. There was a significant statistical difference in the size of MCP between seasons and sex (H = 5.02; p < 0.05; n = 30), explained by the difference between female wet-season and male dry-season home range sizes.

We evaluated the overlap among 13 individuals with home ranges estimated by FK95% (Table 2). The mean number of other turtles an individual overlapped with was 6 ± 1 SD (1–9). Mean overlap was 14% ± 17% SD (0.02%–81%). Overlap between females (20% of all possible pairs) was less frequent than among males (57% of all possible pairs), with a significant difference between the sexes (χ²₁ = 2.73, p < 0.05). We also found 19 overlapping areas of use for male–female pairs (48% of all possible pairs). Females used a mean of 15% ± 16% SD (0.02%–63.1%) of overlapping male areas, while males used a mean of 18% ± 20% SD (0.9%–70.5%) of overlapping female areas.

Table 2. Descriptive statistics (means ± SD, ranges in parentheses) of proportion of overlapping home ranges estimated by FK95% and FK50% (activity centers) between September 2013 and September 2014 at the REBIO Trombetas, Pará, Brazil. n = number of overlaps.

View Fullscreen

All 5 females had their home ranges overlapped by at least 2 males and a maximum of 2 females. Two females, PU07 and PU08, overlapped with 6 and 5 males, respectively. The male PU16 (second-smallest turtle in the study, 208 mm CL and 1100 g) did not have his area overlapped by any females, only with the male P13, the smallest turtle in the study (230 mm CL and 950 g).

In the zones of intensive use (activity centers) estimated by FK50%, we observed 9 individuals overlapping (Table 2; Fig. 3). The mean of the proportions of overlap was 5.3% ± 5.6% SD (0.1%–23.3%), and individual turtles had their activity centers overlapped by 4 ± 2 SD (1–6) other turtles. There were 2 areas where the activity center overlaps were concentrated (Fig. 3): one with a greater proportion of overlapping activity centers during the rainy season and the other with a lower proportion of overlap during the dry season.

DISCUSSION

The selection of the best method to estimate home range size in studies of radio tracking depends on the manner of movements of the animals studied, sample size, and the biological questions pondered (Seaman et al. 1999; Harless et al. 2010; Famelli 2013). Sometimes it is better to use a group of different estimators to more easily interpret the biological meaning hidden in the data collected (Kenward et al. 2001).

The LHR represents the linear use of space and has been documented in a diverse number of studies such that it has come to be used as a pattern for comparison among species and different habitats used by the same species. It is particularly useful in describing areas of use for species that are restricted to shallow water along the shorelines of ponds, lakes, or rivers (Ouellette and Cardille 2011). This method is favored for studies of aquatic fauna in small streams and rivers, where the points represented by the longest distance represent the relatively linear habitat use of the animal being studied. However, in dendritic systems, this estimate tends to hide the real value and can hide information, resulting in an underestimation of the home range (Ouellette and Cardille 2011). The MCP100% method is a linkage method that permits the evaluation of all of the zones that establish a point within areas with intense activity; however, it may overestimate areas used (Kenward et al. 2001). The fixed-kernel (FK95%) method is based on the density of the locations and establishes areas of use effectively for the individuals studied (Kenward et al. 2001). The estimators appear to be substantially different in their mathematical assumptions, and the methods of calculation produce estimates that are not distinguishable statistically (Litzgus and Mousseau 2004). However, analyses of the estimates show variation among individuals and differential use of their home ranges.

We observed a mean LHR of 2.53 km for females and 3.57 km for males of P. unifilis. These estimates were much lower than those found by Moreira and Vogt (1990) for P. expansa, who reported 10 adult females migrating more than 65 km downstream in the Trombetas River, passing Porto Trombetas and possibly reaching the Amazon River. Podocnemis sextuberculata was estimated to have an average LHR of 29.8 km in females (Fachín-Terán et al. 2006), and Bernhard (2010) reported an average of 28.0 km for adult female P. erythrocephala. (Souza 2012) monitored subadults of P. expansa in the Trombetas River and recorded mean LHRs of 6.9 km for females and 3.2 km for 1 male. The values of LHR are thus quite variable within the Podocnemididae regardless of their body size. As previously mentioned, this estimator may hide information in the use of space in areas where there are many channels and meanders. In our study, smaller areas were well represented by the LHR; however, when analyzing areas of greater size, it did not function well in that a large quantity of information was hidden. Home range size estimates were greater with MCP100%, as would be expected, because it overestimates the areas that are actually utilized but includes the corridors that link the areas of intensive use estimated by FK95%. The MCP100% estimator showed that males had larger home ranges (mean = 474.42 ha) than females (mean = 183.21 ha). In contrast, using FK95%, estimated home ranges of males (mean = 74.51 ha) were not different from those of females (mean = 86.90 ha). This result shows that these estimators cannot be compared ecologically. Even though males wandered over larger distances (larger MCP), they appeared to use the habitat in a less intense way, using less of the area within that larger home range compared with females.

Souza (2012) found mean values of 13 ha in the areas calculated with MCP100% and 370 ha using FK95% for P. expansa subadult females, both of which are much higher than our home range estimates for adult females of P. unifilis. The only male studied by Souza had a 344-ha area estimated using MCP100% and a 79-ha area estimated using FK95%, similar to what we found for males of P. unifilis. Within the group of freshwater turtles that have been studied, the size and the form of the home range have many variables and may be influenced by the dimensions of the habitat available (Rowe and Moll 1991; Ely 2008).

The significant differences in home range sizes between the sexes and seasons that have been observed in other turtle species could be related to the movements of the females associated with nesting behavior (Litzgus and Mousseau 2004) or males in pursuit of females (Morreale et al. 1984). Our data show that males were utilizing open areas in the flooded forest more intensely than females, while females were found more often in the lake, which is novel information for the species. Flooded forests are important as sources of fruits and seeds for turtles, and their conservation depends on the protection of these areas (Vogt 2008).

The FK95% estimator presents more rigorous estimates of the proportion of overlap between individuals and yields smaller home ranges than the MCP100% estimator, which raises the question of whether each overlap in area is temporal overlap (Jacob and Rudran 2004). However, it was not possible to verify overlap among all possible pairs of turtles monitored because we analyzed the overlaps of home ranges estimated by FK95%, and therefore only the interaction of 13 individuals was studied. These individuals all overlapped at least 1 other individual with the maximum overlapping 9 individuals. Overlap studies allow the investigation of the interaction among individuals, which is difficult to detect in the field (Burt 1943; Famelli 2013). The species studied showed some social interaction, which can be represented by the aggregation of individuals observed basking along the riverbank in the present study. This behavior is observed in both males and females; however, juveniles tend to group separately (Vogt 2008).

We observed that one of the smallest turtles (PU16; male, 220 mm CL) was found to be restricted to 1 area and overlapped only with another small individual (PU13; male, 208 mm CL). It could be that larger turtles select different habitats with a greater abundance of resources and do not have a tendency to overlap individuals using smaller areas. PU16 also did not overlap with the area of any female; perhaps it was not yet sexually mature and had no interest in searching for females. This behavior suggests the necessity of making more in-depth studies of the behavior of these turtles at different life history stages to verify intraspecific interactions. Perhaps with the use of crittercams, more aspects of social behavior can be documented (Marshall 1998).

The frequency of overlap in females was not as great as that in males. However, there was more interaction between the sexes, with each female overlapping at least 2 males, corroborating the polyandry shown genetically for this species (Fantin et al. 2008). It is important to note that larger males (> 250 mm CL) were present in the backwaters near the nesting beaches along with one of the females being monitored (PU12) and after this period migrated to the flooded forest, where there was overlap with 2 other females being monitored, although these 2 females nested on the clay banks of the Lago Jacaré, not on the sand beach of the river. These males also had the largest home ranges, where we noted an abundant resource base, and thus they probably interacted with many more turtles as well as the ones that we monitored.

The analysis of the overlap of activity centers estimated by FK50% is extremely important for the establishment of conservation strategies in the management plan for this species in the REBIO. The higher density of location points defines the activity centers, and their overlap should indicate which areas are the most important within the REBIO to designate as priority areas for protection. We were able to identify 2 critical areas for the maintenance of this population where there was a frequent overlap of activity centers: one of these areas was identified during the dry season, composed of small overlaps among 7 turtles near the nesting areas on the clay banks within the Lago Jacaré, and the other area proposed was identified in the rainy season and was an open area in the flooded forest where there was a concentration of overlap areas of larger size among 6 turtles. In summary, we show that it is important to protect the coarse-sand nesting beaches of rivers, the nesting areas along the margin of the lake, and the savannas with white sand during the dry season. In the rainy season, the flooded forest areas are the localities with the major intensity of use; thus, it is necessary to protect them along with the canals that link them to the lake.