The nesting process has important fitness consequences for female turtles (Spencer and Thompson 2003). Females are particularly vulnerable to predation at this time (Tucker et al. 1999; Spencer 2002), and eggs and neonates also suffer predation during incubation and nest emergence (Congdon et al. 1983; Yerli et al. 1997; Burke et al. 1998; Escalona and Fa 1998; Janzen et al. 2000a, 2000b). Microclimatic conditions within the nest, such as humidity and temperature, also affect embryo survival (Packard et al. 1991; Burger 1993; Resetarits 1996) and influence sex (Janzen and Paukstis 1991; Spotila et al. 1994), incubation period (Yntema 1978; Packard et al. 1987), and the size, morphology, yolk reserves, growth, physiology, and performance abilities of neonates (Burger 1991; Packard et al. 1993; Bobyn and Brooks 1994; Tucker et al. 1998a; Steyermark and Spotila, 2001; Filoramo and Janzen, 2002; Janzen and Morjan, 2002).

Given the apparent fitness consequences that nest-site selection implies, we predicted that female Trachemys callirostris should nest nonrandomly with respect to specific environmental variables. Other studies of freshwater turtles have shown that females often oviposit in sites that differ in various ways from randomly selected points, suggesting the existence of microhabitat preferences on their part (Wilson 1998; Janzen and Morjan 2001; Kolbe and Janzen 2002). Here, the existence of preferences for specific nesting-site locations was defined statistically as the oviposition of eggs disproportionately in sites that differed from randomly selected points within a given constrained nesting area (Wilson 1998).

Various hypotheses have attempted to explain female nesting-site preferences. Some researchers have suggested that females in species with temperature-dependent sex determination select nest locations so as to manipulate progeny sex ratios to consistently, or alternatively, produce males or females (Vogt and Bull 1982; Roosenberg 1996; Spencer 2002). Other researchers have hypothesized that females prefer to nest in areas shielded from nest predators, parasites, or environmental extremes, thereby maximizing embryo survivorship (Burger and Montevecchi 1975; Burger 1976; Gochfeld 1979; Bodie et al. 1996; Spencer and Thompson 2003). Finally, others have argued that females restrict their nesting activities to areas near the shoreline to minimize their own predation risks while on land, at the expense of ovipositing in locations that are suboptimal (Spencer 2002; Spencer and Thompson 2003). In practice, it is difficult to differentiate among the sex ratio, embryo survivorship, and female survivorship hypotheses (Schwarzkopf and Brooks 1987; Spencer and Thompson 2003), given that nest-site selection in many species probably represents a trade-off between the costs and the benefits of these multiple, sometimes contradictory, factors (Misenhelter and Rotenberry 2000; Wood and Bjorndal 2000; Spencer and Thompson 2003).

Different habitat characteristics could serve as important proximate cues to females during the nesting-site selection process (Weisrock and Janzen 1999). Vegetative cover is one of the most studied characteristics, given its association with nest temperature (Burger and Montevecchi 1975; Vogt and Bull 1984; Janzen 1994; Bodie et al. 1996; Roosenberg 1996; Wilson 1998). Slope and soil surface temperatures also have been proposed as important cues in nesting-site selection (Burger and Montevecchi 1975; Stoneburner and Richardson 1981; Schwarzkopf and Brooks 1987; Wood and Bjorndal 2000).

Advancing on work begun by Bernal et al. (2004) that described the basic nest characteristics of T. callirostris in the Mompos Depression of Colombia, the objective of this study was to evaluate whether females in the same area nest randomly or whether they demonstrate preferences for particular locations within the general nesting area.

METHODS

The Mompos Depression is the largest wetland in Colombia (CVS 2002), located in the Caribbean lowlands and bounded by the Cauca River to the west, the San Jorge River and Ayapel wetlands to the east, the Magdalena River to the northeast, and the Ayapel highlands to the south (Garramuño 2001; Fig. 1). Mean ambient temperature is 29°C but may attain extremes of 35°C. The annual precipitation varies between 1000 and 2000 mm (Fundescala 1997; Turbay et al. 2000), with 2 dry seasons occurring from December to mid-April and from July to August. During the 2 rainy seasons, the region is subject to periodic flooding of the principal rivers that pass through it (Peña 1993).

View Figure

• Download as Powerpoint
• Download Figure

This study was conducted on Leon Island in the Pijiño wetland complex (9°17' N, 74°24' W) in Angostura county of the Magdalena Department, approximately 1 hour by boat to the northeast of the city of Mompos. The 12-hectare island is private property used for raising cattle and swine. Most of the vegetation is low shrub. The shoreline is composed of a mud flat created by the marked fluctuations in water levels each year, where the predominant vegetation is water hyacinth (Eichornia crassipes).

Trachemys callirostris (previously Trachemys scripta callirostris) is a recently recognized species (Seidel 2002) composed of 2 subspecies restricted to Colombia and Venezuela, respectively. In Colombia, Trachemys callirostris callirostris occurs from the western Gulf of Urabá throughout the wetland systems of the Sinú, San Jorge, and Magdalena drainages in the entire northern area of the country, except the Guajira desert (Castaño-Mora 2002). Trachemys callirostris is perhaps the most heavily exploited turtle in Colombia, for both direct consumption and sale (Castaño-Mora 1997) and is classified as NT (nearly threatened) in the Colombia Red List (Castaño-Mora 2002).

Trachemys callirostris callirostris females in this area predominantly oviposit nocturnally. Nests were located during the day by following tracks or by detecting other evidence of nesting activity. We recorded microhabitat characteristics of nests and at randomly selected locations on the island from 6 February to 15 May 2003. We considered the potential nesting area as being the ring of dry soil surrounding the perimeter of the island, beginning where the wet mud ended (hereafter, the shoreline), because no nests were encountered there, and ending at 20 m inland of this point, because more inland nesting was rare (Medem 1975; Bernal et al. 2004). Within the potential nesting area, we selected 80 random points as follows: from an arbitrary point on the shoreline, we defined a 20-m transect perpendicular to it, then walked 15 seconds along the shoreline before defining the origin of a second transect and so on until 20 transects were laid out that covered the majority of the perimeter of the island. In each transect, we selected 4 distances of the 20 possible meter-unit distances by using a random number table to define 80 random locations for study.

For both the actual nests and randomly selected locations, we recorded the exact distance to the shoreline and classified each site into 1 of 2 distance categories: 0–5 m and > 5 m. Vegetative cover was categorized by placing a 1-m² quadrat with 10-cm subdivisions over each nest or randomly selected location and recording for each subdivision whether it contained primarily grass, water hyacinth, or shrubby vegetation. Each m² was then classified as belonging to one of these categories based on the dominant coverage in its subdivisions or as mixed in cases where no one coverage was present in more than 75% of the subdivisions. Soil type was classified by using the methodology of Casanova (1991), yielding 3 main categories of dirt, clay, and sand substrates, and the possible combinations of them (dirt-clay, dirt-sand, and clay-sand).

Twenty three of the 80 randomly selected locations had distances to the shoreline greater than the maximum distance recorded for an actual nest in this study (14.5 m) and were thus excluded from the analyses. Also, no nests were encountered in sites that were predominantly bare substrate, so this category was removed from the analyses. We used χ² goodness of fit tests to compare the number of nests in each distance class, the number of nests located in each vegetative class (use by turtles), and the number of randomly selected locations present in each vegetative class (availability to turtles). Then we repeated the last 2 analyses (use and availability) after having removed the most abundant vegetative class from consideration. The same approach was used in the analysis of soil types, as well as the comparison of the number of nests encountered in the dirt soil type vs. all other soil types pooled. Finally, a χ² goodness of fit analysis was used to inspect for an association between soil type and vegetative cover. Heterogeneity χ² tests were used to compare use vs. availability of vegetative covers and soil types.

RESULTS

We located 86 nests during searches around the island. The greatest distance of a nest from the shoreline was 14.5 m, with a mean distance of 3.5 m. There were significant differences in the proportion of nests located in the 2 distance intervals (χ² = 26.8, df = 1, p < 0.001), with 67 nests (77.9%) located within the first 5 m (Fig. 2). There were significant differences in the proportion of vegetative cover types available on the island (χ² = 20.92, df = 3, p < 0.001); but, after eliminating the grass category, which was the predominant vegetative cover (Fig. 3), there were no differences in availabilities among the other 3 types (χ² = 2.35, df = 2, p > 0.05). There also were significant differences in the proportion of nests oviposited under the different vegetative covers (χ² = 35.77, df = 3, p < 0.001). Upon eliminating the hyacinth category from consideration, which was the predominant vegetative cover selected by nesting females, there were no significant differences in the use of the remaining categories (χ² = 2.35; df = 2; p > 0.05). Obviously, with grass as the predominant vegetative cover and hyacinth as the preferred vegetative cover type, the difference in availability and use of the different vegetative covers was significant (χ² = 9.22, df = 3, p < 0.001, Fig. 3).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

The relative availability of soil types on the island did not differ significantly (χ² = 7.35, df = 5, p > 0.05, Fig. 4), but there were significant differences among soil types in terms of their use for nesting (χ² = 153.48, df = 5, p < 0.001). Upon eliminating from consideration the dirt soil type where the majority (65.12%) of nests were located, there still were differences in the proportion of use of the remaining soil types (χ² = 19.33, df = 4, p < 0.001), as well as differences when use of the dirt soil type was contrasted to the use of the remaining 2 soil types pooled (χ² = 85.98, df = 1, p < 0.001, Fig. 4). There were significant differences between the availability and the use for nesting of the differing soil types (χ² = 16.29, df = 5, p < 0.001, Fig. 4).

View Figure

• Download as Powerpoint
• Download Figure

Water hyacinth was associated with the dirt soil type. In 78% of the cases where a randomly selected point or actual nest was located in the hyacinth vegetative type, it also occurred in the dirt soil type (χ² = 26.79, df = 1, p < 0.001), whereas no other vegetative type occurred predominantly in that soil type.

To evaluate whether the apparent preference by nesting females for the hyacinth vegetation type was an artifact of its association with the dirt soil type, or vice versa, we combined the vegetative cover and soil-type categories to compare availability and use of the 4 new categories (hyacinth on dirt soils, hyacinth on other types of soils, other vegetation types on dirt soils, and other vegetation types on other soil types). Apparently, females preferred both the hyacinth vegetation type per se and the dirt soil type per se, with the additive effects of these preferences explaining the strong tendency to nest where these 2 variables co-occurred (χ² = 26.03, df = 3, p < 0.001, Fig. 5).

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

The data on availability and use of vegetative cover and soil types revealed that the T. callirostris nests were not randomly distributed within the general area used for nesting, suggesting that females were actively selecting sites with specific characteristics for ovipositing. The previous study in this site, by Bernal et al. (2004), found that 82% of the nests that year were located beneath herbaceous vegetation (hyacinth or grass), consistent with the results of this study in which 65% of the nests were located under the same vegetative-cover types. Medem (1975) also reported 34 T. callirostris nests from the Totumo wetland in Bolivar Department, Colombia, as being covered by grass, shrub, or second-growth vegetation.

Selecting sites covered by vegetation for nesting might reduce thermal stress to the nesting females, as well as lower the detectability of nests to natural predators, human hunters, or their domesticated animals (Wilson 1998; Bernal et al. 2004). Also, nests under vegetation might experience less variation in temperature or humidity compared with those oviposited in exposed sites (Bodie et al. 1996; Wilson 1998; Bernal et al. 2004). This preference on the part of T. callirostris females differs from that reported for other tropical Trachemys species that tend to nest in relatively open areas that receive direct sunlight for at least part of each day (Moll and Legler 1971). This may be a response to the higher mean nest temperatures prevalent in this region of Colombia (Restrepo et al., unpubl. data, 2003) in comparison with those documented for other tropical Trachemys nesting sites (Llanos de Caño Negro, Costa Rica, mean 26.2° ± 1.4°C SD with a range of 25.5°C to 28.2°C, Cabrera et al. 1996; Juan Mina, Panama, with a range of 22°C to 30°C, Moll and Legler 1971). Apparently, female nest-site selection behavior varies geographically in response to prevailing climatic conditions (Gibbons 1983).

In the Mompos Depression, relatively few nests were oviposited under grass (15%), given its abundance (51%), indicating nonpreference of this coverage, whereas hyacinth was preferred (52.3% of all nests oviposited under a cover that comprised only 17.7% of the total in the area, Fig. 3). However, in a previous year, Bernal et al. (2004) failed to document preferential nesting for hyacinth vs. grass vegetative covers. This difference may have been a result of the exceptionally high temperatures during our study and the fact that nests under hyacinth experience incubation temperatures that are lower than those of nests under other vegetative types (Restrepo et al., unpubl. data, 2003). There were no differences among vegetative types in terms of the other variables that we considered (hatching success rates, predation or parasitism rates, incubation periods, hatching success rates, hatchling sizes or sex ratios; Restrepo et al., unpubl. data, 2003).

We also documented that nesting females preferred to nest in the dirt soil type, especially but not exclusively when it was associated with the hyacinth vegetative cover. We detected no other associations between the other vegetative covers and soil types. Selecting for specific soil characteristics may be related to the possibility or the ease with which appropriate nest chambers may be constructed there or also related to the relation between grain size and water potential of the substrate. Flexible-shelled turtle eggs such as those of T. callirostris benefit from incubating in humid substrates in terms of the developmental rate of the embryos and size of the resultant neonates (Congdon and Gibbons 1990; Tucker et al. 1998b).

Although we documented a preference on the part of the nesting females for dirt soils, Moll and Legler (1971) argued that Trachemys in Panama nested irrespective of soil type, except that they avoided muddy areas. In a previous year, Bernal et al. (2004) found nests in our study site predominantly in mixed dirt-sand soils. However, the previous study was conducted in a year with considerable flooding, whereas our study was conducted during an El Niño year, with an exceptionally severe dry season. This suggests that female nest-site preferences not only vary geographically but also temporally, depending upon climatic variation, indicating a flexibility on the part of the females similar to that documented in green sea turtles, Chelonia mydas (Bjorndal and Bolten 1992).

We found 77.9% of all nests within the first 5 m from the shoreline, concordant with the reports of Bernal et al. (2004) of finding the majority of nests on Leon Island that year within the first 6 m and the study by Medem (1975) that reported a mean nest distance from the shoreline of 4.5 m. Also, Zenteno and Bouchot (2001) reported a mean nest distance to the shoreline of 3.5 m for Trachemys venusta in Mexico, despite the fact that many nests located there were lost to natural predators. In contrast, in an area with little human presence in Juan Mina, Panama, Moll and Legler (1971) documented a mean nest distance to shoreline of 50 m for T. venusta, with some nests located almost 400 m from the shoreline. Heavy hunting pressures on nesting females in some areas may have led to a tendency to confine nesting activities to the immediate vicinity of the shoreline, as has been shown experimentally for Emydura macquarii, where females increased their mean nest distance to the shoreline by 10 m in response to the removal of natural predators from some islands (Spencer and Thompson 2003).

Many freshwater turtles appear to actively select their nest-site locations, but decisions are probably influenced by complex trade-offs related to various fitness components, and it seems that the costs and benefits of different options may vary both geographically and temporally for any given species. Unfortunately, studies that merely characterize the microhabitat characteristics of sites where turtle nests occur, without also quantifying the general availability of the different characteristics in the overall nesting area, are insufficient. Only by rigorously documenting actual nest-site location preferences will a complete understanding of these complex fitness interactions be possible.