Study Area

We conducted the study in 2 streams in the vicinity of Chimichagua, department of Cesar, Colombia (lat 9.28°N, long 73.79°W). The region is tropical dry forest, which has a mean annual temperature above 24°C and annual precipitation between 700 and 2000 mm, with 1 or 2 dry seasons per year (Espinal 1985; Murphy and Lugo 1986). The 2 streams studied, the San Fernandera (SF) and Las Peñas (LP), were the only ones where the species was found to occur after an extensive search of 7 streams in the region. These 2 streams were slow-water creeks, with varying degrees of riparian vegetation, surrounded by pastures used mainly for cattle (Forero-Medina et al. unpubl. data, 2011). They had depths that varied with rainfall patterns during the year, usually within the range of 10 to 200 cm. LP is narrower, approximately 2–3-m wide, whereas SF varies in width, reaching 8 m at some points. The 2 streams are part of the same watershed and are connected during the wet season. During the driest months (December to February) LP can dry completely, whereas SF usually maintains small pools. The total length of the streams sampled was 2.3 km, with a mean width of 4.5 m.

Abundance

We conducted mark–recapture sampling from February 2008 to February 2009. Initially, sampling was conducted every month, but, after June 2008, it was conducted once every 2 months. Each sampling occasion lasted 3 consecutive days, 1 for LP and 2 for SF, because the SF stream was longer. The method for capturing turtles along the streams consisted in setting 2 parallel 10-m × 2-m seine nets (2.5-cm mesh size), which enclosed a 20-m portion of the stream and then directing the turtles into the nets by beating the water. The nets were then moved forward along the stream, and the process was repeated. After capture, each individual was measured for carapace length (CL; straight line in millimeters from the nuchal scale to the midpoint of the contact between marginals 12), weighed to the nearest 5 g, and marked by using a marginal scale notching system (Cagle 1939). Sex was determined by inspecting sexually dimorphic traits for the species; males have longer, thicker tails, narrower posterior plastral lobes, and narrower heads than females (Ernst and Barbour 1989). For individuals smaller than approximately 130-mm CL, sex was not clearly identified by this method, and these individuals were considered juveniles.

Because sampling effort was constant across events, we compared the total number of individuals caught in each sampling event in each of the 2 streams as a measure of the variation in the relative abundance of the population over time. We evaluated if there was a relationship between relative abundance (turtles captured per sampling occasion) of M. dahli and precipitation, by using data on total monthly precipitation from the weather station located at Chimichagua, Cesar. Data of monthly precipitation for the 1-year period of the study was provided by the Instituto de Hidrología, Meteorología y Estudios Ambientales. We also compared the mean size (CL) and body mass of the individuals captured during wet vs. dry months.

With the mark–recapture data, we estimated population sizes for the 2 streams for the year study by using the Jolly–Seber model for open populations (Jolly 1965; Seber 1965; Lebreton et al. 1992), assuming non constant parameters from sampling to sampling period, as described in Krebs (1999). Confidence intervals for population sizes were estimated according to Manly (1984). The 2 streams connect during the rainy season and are part of the same watershed, therefore, we considered them to be part of a single population. To estimate density, we calculated the area of the portion of the streams sampled during the wet season.

Home Range and Movements

We studied home range size and movement patterns by using radiotelemetry. For this, 8 adult individuals (4 males and 4 females) were equipped with very high frequency (VHF) radio transmitters (Model Telenax TXE-124G; frequencies 150.50–150.78). Each transmitter had a 30-cm external antenna and weighed approximately 12.5 g (Fig. 1). The transmitters were attached by using a nontoxic epoxy adhesive (Loctite E-00CL), which cures at room temperature. Adults from this species may weigh up to 1500 g, which means that the proportional weight of the transmitters was below 2% of body weight, less than the maximum recommended for freshwater turtles (Schubauer 1981). Animals with transmitters were tracked in the field by using a 3-element Yagi antenna and a receiver (model RX-TLNX). The turtles were followed simultaneously for 4 consecutive days before each trapping effort, monthly from May to August 2008, and bimonthly from August 2008 to April 2009. Individuals were not approached closely, but their position was estimated by using triangulation from at least 4 different locations.

View Figure

• Download as Powerpoint
• Download Figure

We estimated home range sizes by using the minimum convex polygon method (MCP) and by using fixed kernel density estimators to allow for comparison with studies of other chelonians. The MCP is the most common method for estimating home ranges and consists of constructing the smallest convex polygon encompassing all known or estimated locations for the animal (Hayne 1949). The kernel density estimator produces a utility distribution, which represents the probability that an animal will be in any part of its home range (Worton 1989; Seaman and Powell 1996; Powell 2000). For the fixed kernel estimation, we used 150 m as the smoothing factor, based on mean and maximum distances moved daily, as indicated by the telemetry data. We present the results as the 95% contour for the fixed kernel estimator. We did not eliminate spatial autocorrelation in the point data because it has been suggested that doing so reduces biological relevance of home range estimates (De Solla et al. 1999). We compared home range sizes between males and females for the wet months only by using 2-tailed t-tests, because data for males during the dry season were insufficient. We also determined if there were differences in home range sizes between the wet and dry season months by using 2-tailed t-tests.

We compared the movement parameters “mean step length” and “net displacement” between the rainy and the dry season by using 2-tailed t-tests for independent samples. Based on the precipitation pattern during the study period (Fig. 2), we considered as dry months those that received less than 100 mm of total precipitation (December, January, February). The remaining months were referred to as “wet months,” although the rainfall pattern is more complex than this. Rains start around March, with a decrease in June–July, called the “veranillo  =  little summer,” and then a seasonal rainy period from August to November. Step length was defined as the distance between consecutive locations, which provided a proxy for movement. Mean step length within sampling events corresponded to the mean distance traveled between successive measures in a single sampling occasion. This corresponded to movements completed within a time frame of 4–12 hours. Mean step length between sampling events corresponded to the mean distance traveled from the last location in a sampling event to the first location in the consecutive sampling event. This corresponded to movements completed in a time frame of 1 or 2 months. Step length between sampling events could correspond to either the same season (wet, dry) or the transition between wet and dry seasons. Therefore, we considered them separately in the analyses. Net displacement was the absolute distance between the initial location and a given location. We conducted all spatial analysis by using ArcMap 9.3 (ESRI 2010) and the application Hawth's Tools for ArcGIS (Beyer 2004).

View Figure

• Download as Powerpoint
• Download Figure

Abundance

During the study period, 41 M. dahli individuals were captured, 21 in stream SF and 20 in stream LP. There were 19 recaptures, which represents 32% of the total captures. The adult sex ratio was not significantly different from 1∶1 (n  =  27, 14 females, 13 males). The total number of individuals captured for each sampling event varied during the year (Fig. 2). Usually, no turtles or very few individuals were captured during the dry period, from December to February. The mean body mass for turtles captured during the dry season was 363 g, whereas the mean body mass for the wet season was 595 g, this difference being significant (2-tailed t-test, p  =  0.01; wet season, n  =  40 and dry season, n  =  15). The mean CL also was lower in the dry season (137 mm) than in the remaining year (168 mm), although this difference was not significant (p  =  0.06; wet season, n  =  27 and dry season, n  =  10).

There was considerable variation in the number of captures in each month when the water level was high (March to November). Thus, there was no significant correlation between the number of individuals captured in one month and the total precipitation in that same month for either stream (Pearson correlation: n  =  9; p  =  0.45 for SF; and n  =  9; p  =  0.30 for LP). According to the Jolly–Seber model, the estimated population size for the 2 streams varied during the year from 16 turtles (95% CI, 7–30) to 175 turtles (95% CI, 32–298). The lowest and highest population estimates were for April and June, respectively. When considering the area of the streams sampled at high water, the resulting densities ranged from 16 turtles/ha in April to approximately 170 turtles/ha for the maximum population size estimated in June.

Home Range and Movement Patterns

We recorded 223 locations for the 8 turtles with radiotransmitters. The average number of locations per individual was 28 ± 13 SD. Two turtles were lost after the second sampling event and were not located again, another was located only during the wet season, and 5 were located in both wet and dry seasons. The estimated home ranges for the year-long study varied from 1.6 to 30.8 ha according to the MCP and from 9.2 to 22.5 ha according to the fixed kernel estimator (Table 1). There was no significant difference in home range sizes between sexes for the wet months when using MCP (n  =  8 for wet and dry; p  =  0.63) or the fixed kernel density estimator (n  =  8 for wet and dry; p  =  0.99). We were not able to compare home range sizes between sexes for the dry period or the year period, because there were only data for 2 males during the dry months. However, we present the results for all the home ranges estimated with the location data (Table 1).

Table 1. Estimated home ranges by season for 8 individuals of Mesoclemmys dahli studied by using very high frequency (VHF) radio telemetry.

View Fullscreen

Upon pooling data for males and females, there was no significant difference in home range size between the wet and the dry months when considering the MCP (p  =  0.71) or the fixed kernel estimator (p  =  0.26). For some individuals, the range for the dry months occurred in spatially different areas than the range for the wet months (Fig. 3). However, home ranges generally overlapped across seasons and within and between sexes (Fig. 4). Kernel density estimators produced larger home range estimates than MCP, and disjoint ranges when locations were far from each other (Table 1; Fig. 3).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

The mean step length within sampling events was 68 m for the wet season and 70 m for the dry season, with no significant difference (p  =  0.94). The mean step length between sampling events was 175 m for the wet season and 268 m for the dry season, however, there was no significant difference (p  =  0.33). The mean step length that corresponded to the transition from wet to dry seasons was 287 m, which was not significantly higher than the 175 m mean step between sampling events for the wet season (p  =  0.28). This is due to the high variation in movements during the transition from wet to dry seasons, with some turtles moving only 13 m from one season to another. However, it is clear from the data that the highest individual movements recorded were conducted either during the transition from wet to dry season (895 m, 426 m) or during the dry season (1053 m) (Fig. 4). From December 2008 to February 2009, one female made the greatest movement recorded between sampling events (1053 m) (Fig. 4). The net displacements ranged from 117 to 1304 m from the initial point of first capture and release. The highest net displacement corresponded to 2 turtles for which data on the whole year cycle were obtained, and the last locations were recorded during the dry season.

Most locations during the wet season were in the water or very close to the streams. Because detection of the turtles was done by using triangulation, it was not possible to classify every location as aquatic or terrestrial. The error in the triangulation was higher than the width of the stream and the distance of most locations to the water. However, in the dry season, locations were farther from the stream (Fig. 4c, d).

Abundance

Although there was no significant correlation between the number of individuals captured in each month and the total precipitation in the same month, our results indicated that there were few turtles in the streams during the dry season. The lack of correlation occurred because there also was wide variation in the number of captures when the water level in the streams was high. These results, together with the radio telemetry data, indicate that, during the dry season, many individuals leave the streams, especially in LP, which dries completely. This agrees with previous reports that indicate that M. dahli buries on land during dry months and aestivates (Medem 1966; Rueda-Almonacid et al. 2004). However, not all turtles exhibited this behavior. Some individuals stayed in the water in the SF stream, where some pools remained.

The mean body mass for turtles captured during the dry season was significantly lower than for turtles captured during the wet season, which suggests that the largest individuals tended to leave the water during the dry months, whereas smaller ones remained in the pools. The mean CL was smaller for individuals captured during the dry season, but this difference was not significant. So, another possibility is that turtles lose weight during the dry season, perhaps due to a reduction in accessible resources. During the wet season (March to November) turtles tended to be in the water, but their abundance was not related to rainfall.

Our density values were low when compared with other studies of South American chelids. Estimates for Hydromedusa maximiliani were 193 turtles/ha in mountain streams (Souza and Abe 1997). Martins et al. (2009) estimated the population size of H. maximiliani on their study area to be 235–318 turtles, but no density was specified. Densities for Hydromedusa tectifera may be as high as 219 turtle/ha in some streams in Córdoba, Argentina (Lescano et al. 2008). In urban and polluted rivers in Brazil, densities of Phrynops geoffranus can be even higher, reaching 230 turtles/ha (Souza and Abe 2000). However, caution should be taken when comparing these estimates, because they come from different methods and environmental conditions. Information for species from the genus Mesoclemmys is limited. Silva de Brito et al. (2009) reported for Mesoclemmys vanderhaegei an estimated population size of 12–36 turtles in each km section of “small” streams sampled during a 3-month period, but no density values were given.

Previous work with M. dahli in the department of Córdoba indicated that the species is locally abundant in that part of its range (Rueda-Almonacid et al. 2004). In that study, density values were reported by using the number of captured individuals per sampling event instead of estimated population sizes. They reported approximately 8–18 individuals captured for each kilometer of stream sampled, which resulted in a density of 20–60 turtles captured/ha (Rueda-Almonacid et al. 2004). We can use the number of individuals captured to estimate density and make the data comparable with Rueda-Almonacid et al. (2004), who used a similar capture method. The highest number of turtles that we captured in the 2.3 km of stream sampled was 12. This value results in a maximum density of 10 turtles captured/ha, lower than the densities reported in streams for Córdoba. In water reservoirs in Córdoba, Rueda-Almonacid et al. (2004) reported even higher densities of 500 turtles captured/ha. Mesoclemmys dahli apparently is locally more abundant in Córdoba than in Cesar, which may be a result of slight differences in sampling effort between the 2 studies. It also may be related to the populations in Cesar being at the periphery of the species' geographic distribution (Forero-Medina et al. unpubl. data, 2011). Peripheral populations of many animals present lower abundances than populations that are closer to the center of the geographic range (Gaston 2003), which has importance for conservation efforts of an endemic species such as M. dahli. First, if one has to create one single protected area for the survival of the species, then populations in Córdoba may be more suitable, because they have higher abundances and potentially a higher chance of persistence (Gaston 2003). However, if the objective is the preservation of the species throughout its range, then peripheral populations like the ones in Cesar will also require protection, because they are more vulnerable. In any case, rare species such as M. dahli will need protection at multiple sites, to spread the risk of extinction and ensure that sufficient individuals are secure (Gaston 2003).

Home Range and Movement Patterns

Home ranges should be defined for a specific time interval, because they can be dynamic over time (Powell 2000). Year-long home ranges for M. dahli varied from 1.6 to 30.8 ha (MCP) or 9.2 to 22.5 ha (kernel estimator) (Table 1). Home ranges estimated by the 95% fixed kernel density estimator were usually higher than estimates from the MCP (Table 1). We believe that the kernel method better represents the area used by individuals with small ranges. For individuals with large ranges, the estimated home range sizes were similar for MCP and kernel density estimators (Table 1). Both methods are commonly used for estimating ranges of turtle species (Morrow et al. 2001; Litzgus et al. 2004; Donaldson and Echternacht 2005).

Studies of home ranges for South American chelids are limited. The range estimates from better-studied species, such as river turtles, cannot be compared with ranges of species such as M. dahli, which inhabit smaller streams and creeks, and avoid major river systems (Rueda-Almonacid et al. 2007). Results of studies of Phrynops geoffroanus, which inhabits small streams in southeastern Brazil, reported a 0.12-ha home range, estimated during a 72-hour period (Souza et al. 2008). The researchers suggested that the small range size was due to the food provided by polluted habitats. However, this range would clearly be expected to be larger for longer sampling efforts and is not comparable with our home ranges estimated for a 1-year period. A long-term study of Phrynops rufipes, which inhabits small, closed-canopy, black-water streams in the Amazon (Caputo and Vogt 2008), indicated that it has a linear home range of 1–2 km and may remain in this range for long periods (Magnusson et al. 1997).

Our study indicated that M. dahli uses areas of approximately 7–16 ha most of the year (March to November) and that, during the dry season, some individuals may expand these areas considerably or move to other sites. Future studies will reveal whether these individuals shift their ranges or return to the same areas used during the wet season. Although there were no significant differences in home range size or mean movements between the dry and wet period, it was clear that the greatest movements and a higher frequency of terrestrial movements occurred during the dry season. This, as well as the evacuation of the drying stream LP, suggests that movements and use of space are partly correlated with rainfall or with other environmental cues, as has been found for other chelids (Roe and Georges 2008).

Home range areas in our study included the streams where turtles were originally captured, the associated riparian vegetation surrounding these streams, and the pastures and sparse vegetation contiguous to the riparian vegetation, as well as small ponds and other water bodies located within these pastures. The aquatic habitat is where animals spend most of the time and where they forage for food, as suggested by what is known about the species' diet (Medem 1966, Rueda et al. 2007). Most of the items found in the feces during this study were mollusks, although this could be due to mollusk shells remaining in their stomachs more than other diet items. Though the aquatic vs. terrestrial locations could not be determined because of location error constrains, it was evident that the turtles remained close to the water during the wet season, and terrestrial movements were more common during the dry season. In captivity, individuals have been observed to spend considerable time outside the water but always close to the shore (de la Ossa-Velasquez 1998).

The riparian vegetation provides refuge for individuals, especially during the dry months, when turtles can be found under dead tree roots and vegetation. Turtles also use pastures and areas between streams and temporal pools, as in Córdoba (Rueda-Almonacid et al. 2004). Riparian vegetation and surrounding areas also may be important for nesting. Although very little is known about nesting sites for this species, studies in Córdoba reported a nest located 1 km away from the nearest stream (Rueda-Almonacid et al. 2004). It has been suggested that, in Córdoba, home ranges are larger during the rainy season because, during the dry season, animals tend to aestivate (Rueda-Almonacid et al. 2004). We did not find a significant difference between home ranges for the dry vs. wet season. Our results indicated that the largest movements were exhibited either during the transition from wet to dry season or during the start of the dry season. Some turtles moved very little during the dry season as well. This confirms that some, but not all, turtles are active and searching for new places during dry months. These individuals may move relatively long distances. Individuals that were tracked during the full-year cycle were farther from the initial site than individuals tracked for less than a year, which suggests that some turtles may shift their ranges from year to year.

A low abundance coupled with increasing pressure on their habitat from deforestation and continuous burning make the populations of M. dahli in Chimichagua vulnerable. We cannot yet determine population trends because there are no historical data for the species. Therefore, continuous monitoring of the population is essential. Studies on the reproductive ecology of the species are also urgently needed. When considering that it is an endemic species with a restricted range on the Caribbean coast of Colombia, we recommend protection of core and peripheral populations in the geographic range. Conservation areas for the species will need to include streams of occurrence, associated riparian vegetation, and surrounding buffer areas, all of which are commonly used by the species.