Turtle populations may become extinct because of a combination of extrinsic (e.g., habitat destruction, over-harvesting) and intrinsic (e.g., limited dispersal ability) factors that can result in little or no gene flow among isolated populations. Thus, understanding the ecological and demographical patterns that drive population dynamics is fundamental for conservation and management efforts (Caughley 1994).

Dispersal is an important parameter in general (Macdonald and Johnson 2002) and for freshwater turtles in particular (Scribner and Chesser 2001; Souza et al. 2002b). Given their life history, freshwater turtles only leave water during the reproductive season when females look for nesting sites, or during the rainy season when migration over land may occur (Gibbons et al. 1990). Freshwater turtles may be quite sedentary (Souza and Abe 1997; Scribner and Chesser 2001; Souza et al. 2002a, b), resulting in a metapopulation structure in some areas where each river and stream harbors its own population (Souza et al. 2002a). This has important consequences for conservation because habitat fragmentation can alter dispersal ability (Macdonald and Johnson 2002).

Rivers and water basins form a naturally fragmented habitat in a complex hierarchical system. For aquatic organisms such as freshwater turtles, the landscape matrix encompassing the drainages is an important habitat characteristic. To disperse, suitable microhabitat such as riparian forests along river edges, inland lakes, or ponds that can be used for turtles may be needed (Bodie 2001; Gibbons 2003). However, intensive habitat fragmentation caused mainly by human action has changed the original landscape matrix throughout much of the species' original geographical distribution with the resulting habitats usually exhibiting both biotic and abiotic characteristics distinct from that original. Despite challenges imposed by these new human-created ecosystems, some organisms are able to survive in extensively modified habitats. Increased nutrient inputs and fewer natural predators in urban environments can benefit freshwater turtles (Souza and Abe 2000; Spinks et al. 2003). On the other hand, for those freshwater turtle species surviving in an urban habitat, overland dispersal may be hindered because of river channelization or may be dangerous because of road mortality (Bodie 2001; Aresco 2005; Gibbs and Steen 2005; Steen et al. 2006).

Geoffroy's side-necked turtle, Phrynops geoffroanus (Chelidae), has a broad geographical distribution in South America, ranging from the Colombian Amazon to southern Brazil and northern Argentina (Vanzolini 1994). Across its range, the species can also be found in large streams and lakes as well as in urban rivers (Souza and Abe 2000, 2001). Despite its common occurrence in disturbed habitats, the biology of P. geoffroanus is poorly understood in that environment. In an apparent paradox, although P. geoffroanus density in urban habitat can be as high as 230 turtles/ha of river, individual recaptures in these high-density areas during capture–recapture studies are not common (Souza and Abe 2000). We examined dispersal of P. geoffroanus at small time-scales (hours and days) by using mark and recapture techniques in an urban stream in central Brazil.

Methods. — Turtles were captured from August 2004 to February 2005 by fishnets extended between the banks (Souza and Abe 2001) of a nonchannelized 300-m stretch of the Anhanduizinho River at Campo Grande, Mato Grosso do Sul, central Brazil (ca. lat 20°28′S, long 54°40′W). Campo Grande has over 730,000 inhabitants. In Mato Grosso do Sul, as well as in every Brazilian federal state, there are environmental problems related to human habitation around large urban centers, and water pollution in such urban rivers represents the most important environmental impact. The typical natural vegetation throughout the region is the savanna-like vegetation known as cerrado. Given its geographical position, Campo Grande experiences severe droughts during the dry winter (March–August) and months of heavier rain in the summer (September–February). The Anhanduizinho, a stream that belongs to the Paraná River basin, is the main river of the city. It is relatively narrow (3−10 m) and shallow (0.5−3 m) with bottom mud. Pollution is evident by the presence of human sewage and domestic waste. In the nonchannelized parts, river banks are covered by grass and extend by 10−20 m with a slope around 20° to a channel.

Captured turtles were identified as either juvenile, adult male, or adult female according to external morphological characteristics (Molina 1989; Souza and Abe 2001). Turtles were measured (straight-line carapace length) with calipers (0.1 mm) or a metric tape (0.5 mm) and individually marked by notching marginal scutes (Cagle 1939). Before release, a small (6 × 3 mm), thin PVC plate, attached to a 3-m-long soft cotton line with small, uniquely colored Styrofoam balls at the distal end was glued to the turtle's supracaudal scutes. The balls allowed us to follow the animals during their aquatic dispersal (an analogous device to thread trailing; Breder 1934) by walking along the river banks. To preclude accidental drowning the cotton line would break if it became entangled in aquatic vegetation.

Dispersal behavior was followed for the first 4 hours after release and then after 24-, 48-, and 72-hour periods. At the river banks, dispersal distance was measured with a metric tape after locating individual turtles. This procedure allowed us to identify initial-hour dispersal distance as well as accumulated dispersal after the 3-day period. A χ² test was used to verify the independence of dispersal (up-or downriver) after turtle release. Kruskal-Wallis 1-way analysis of variance was performed to verify differences of dispersal rate of turtles among hour classes after release; the relationship between turtle size and accumulated dispersal was examined by simple linear regression (Zar 1999).

Results. — Forty-two turtles were captured and released (11 adult males, 12 adult females, and 19 juveniles). Because resighting rate was unequal in each time class after release, the number of observations in our analyses varied (Table 1). The mean distance traveled by turtles, estimated by accumulated resightings of the colored Styrofoam balls was 41.7 m after the 72 hours from release (SD 32.0; range, 0.0–115.0 meters; n = 40 turtles). Considering that 10 m is the mean width of the river in this stretch, species home range size averaged 420 m² with a maximum of 1150 m² (or 0.04 to 0.12 ha). After the first 4 hours, turtles exhibited a marginally significant dispersal downstream (χ² = 3.38; p = 0.07; n = 58 captured–resighted turtles). However, after the 3 final periods after release (24, 48, and 72 hours) the movement was predominantly upstream (χ² = 10.0; p < 0.005; n = 40 captured–resighted turtles).

Table 1. Dispersal of 42 Phrynops geoffroanus in central Brazil, according to 7 time classes, reproductive status, and body size (straight-line carapace length). Blank space indicates those turtles not localized in a given time. Values are expressed as mean ± SD (range; sample size).

View Fullscreen

Turtle dispersal distance was significantly different among the 7 time classes after release (H = 58.6; p < 0.001; n = 161 captured–resighted turtles), with most of the distance traveled occurring during the first 3 hours after release (Table 1). This kind of dispersal pattern was also verified for adults (H = 28.8; p < 0.001; n = 84 captured–resighted turtles) and juveniles (H = 32.1; p < 0.001; n = 77 captured–resighted turtles) if they were considered separately in analyses.

Because very few turtles (6) were encountered at 72 hours, this period was excluded from regression analysis. Larger turtles showed greater dispersal than smaller ones after 24 hours (y =−32.24 + 0.32x; p = 0.001; r² = 0.29; n = 31 encountered turtles; Fig. 1A) and after 48 hrs (y =−19.89 + 0.27x; p = 0.02; r² = 0.15; n = 35 encountered turtles; Fig. 1B), but the low determinant coefficients reflect the weak linear relationship.

View Figure

• Download as Powerpoint
• Download Figure

Discussion. — The estimated home range size for P. geoffroanus in our study (0.04 to 0.12 ha) is smaller than that observed for another Brazilian chelid species, the red side-necked turtle, Rhinemys rufipes (Magnusson et al. 1997). In the Amazonian region, the small area (around 0.8 ha) used by R. rufipes was thought to reflect the species' diet, mainly palm fruits because palm trees were abundant in the area (Lima et al. 1997; Magnusson et al. 1997). On the other hand, Souza and Abe (2000) hypothesized that if feeding behavior influences home range and dispersal of freshwater turtles, then P. geoffroanus from urban habitats should have smaller home ranges than that exhibited by R. rufipes given the great nutrient input in anthropic habitats. Our findings are consistent with that hypothesis.

Feeding resources are important environmental constraints for species (Taylor et al. 2005). In urban ecosystems, P. geoffroanus can take advantage of domestic waste that is continually discharged into the river as an extra feeding resource (Souza and Abe 2000). Thus, a typical polluted urban river can be described as having constant food supplementation. Organisms being provided supplementary feeding can reduce their home range size (Boutin 1990). However, some species' life history traits, such as foraging behavior, metabolic rate, or dominance status in hierarchical systems, can affect this pattern by increasing or decreasing the home range area (Hansen and Closs 2005; Taylor et al. 2005). Phrynops geoffroanus is a widely foraging species and the aggressive behavior detected among captive individuals suggests a well-defined hierarchical structure in populations (Molina 1992). The few available field records of P. geoffroanus diet report that the species can change feeding habit according to habitat characteristics. Turtles can switch from a fish-, crustacean-, and mollusk-based diet in nonpolluted areas (Fachin-Teran et al. 1995; Dias and Souza 2005) to one dominated by Chironomidae larvae in polluted urban rivers (Souza and Abe 2000). Our study reinforces the hypothesis that home range size can be mediated by food resources. However, comparing dispersal behavior of P. geoffroanus from populations across distinct human-impacted and natural environments could provide interesting results on this subject.

The short-term dispersal pattern detected for P. geoffroanus was similar to other freshwater turtle species with a great variance in the amount of movement with time after release (Kramer 1995; Magnusson et al.1997). Movement was accentuated in the first hours after release and progressively decreased after 2 days. Escape is a common behavior employed by freshwater turtles after animals have been captured (Hayes 1989) and the dispersal characteristic exhibited could reflect the stress involved in handling animals. Available records on dispersal of Brazilian chelids report sedentary behavior for some species. Capture–recapture studies showed that although Hydromedusa maximiliani had a mean daily dispersal of 2 m, Ranacephala hogei can move up to 15 m/d (reviewed in Souza 2004). However, in trailing experiments, H. maximiliani can employ a dispersal rate as high as 45 m/d after released.

In urban rivers, management practices such as flow regulation and channel modification may have effects on freshwater turtles. Such effects can include changes in population structure, increased population fragmentation processes, or reduced population turnover, by reduction of sandbars and beaches available for nesting or nest inundation and failure (Bodie 2001). Although P. geoffroanus is a common species in urban ecosystems, the real effects of the human practices on the turtle population dynamic is unknown. Hatchlings and juveniles of P. geoffroanus are frequently detected and captured in the nonchannelized study areas, suggesting that the population may persist. However, turtle mortality on highways is an important factor on turtle demography (Aresco 2005; Gibbs and Steen 2005; Steen et al. 2006) and, although not measured, P. geoffroanus are frequently found killed by vehicles on the highways along the river course. The urban watercourses may provide suitable habitat for P. geoffroanus, but recent studies demonstrate that terrestrial habitats also must be considered in population management of freshwater turtles (Bodie 2001; Gibbons 2003). Efforts must be required to assess the land dispersal behavior for a species living in a predominantly urban habitat.