Nest site selection by female freshwater turtles is well known to affect hatchling size, population sex ratio, and gender via temperature-dependent sex determination (TSD) (Bull 1980; Janzen and Paukstis 1991). Nest temperature and consequently TSD are directly related to location of nests under or away from vegetation (Weisrock and Janzen 1999). Much of the data supporting conclusions on nest placement by Malaclemys terrapin are based on observations of nests destroyed by predators (e.g., Feinberg and Burke 2003; Butler et al. 2004) because humans are not as adept in finding intact nests. In some areas, predation declines as distance from the edge of vegetation cover or water increases (Kolbe and Janzen 2002; Spencer and Thompson 2003; Marchand and Litvaitis 2004). In others, nest loss is high regardless of where they are on the landscape (Burke et al. 2005; Hackney 2010). Thus, predator behavior may influence our understanding of patterns of nest site selection and nest survival.

Although much is known of the population ecology, reproduction, and conservation of Malaclemys terrapin elsewhere in its range (e.g., Montevecchi and Burger 1975; Seigel 1980; Tucker et al. 1995; Roosenburg et al. 1999; Gibbons et al. 2001; Butler et al. 2004), comparable studies are few for populations in Virginia. Our study evaluates the selection of nest sites by female terrapins on Fisherman Island, Virginia, and contributes comparative data to our developing understanding of this estuarine reptile in the Chesapeake Bay region.

Methods

Study Site

Fisherman Island National Wildlife Refuge (FINWR) encompasses a 749-ha barrier island located at the mouth of the Chesapeake Bay in Northampton County, Virginia, at the southernmost tip of the Delmarva Peninsula (Fig. 1). This southernmost barrier island in the archipelago adjacent to the east coast of Delmarva is separated from the peninsula by Fisherman's Inlet, a 0.6-km-wide body of ocean water. It is surrounded by Spartina marsh on all but the Chesapeake Bay side of the island. The upland portion of the island is about 80% wooded and 20% open and sandy areas with scattered beach grass (e.g., Panicum spp.). Active dunes occur on the northeast and southeast portions of the island. Much of the upland portion of the island, including the study area, consists of stabilized dunes that support loblolly pine (Pinus teada) and hardwoods such as wax-myrtle (Myrica cerifera) and black cherry (Prunus serotina). These trees and shrubs (e.g., Amelanchier [serviceberry]) occur primarily in discrete clumps and serve as habitat islands (Fig. 2). Slater and Lamont (2000) and Belden and Field (2007) described the flora and vegetative communities of FINWR, indicating that 30% of the plants are invasive. US Route 13, constructed in 1964, bisects the upland portion of the island on the western side. Approximately 16.4 ha lie above the mean high tide line. Known predators are fish crows (Corvus ossifragus), laughing gulls (Larus atricilla), red fox (Vulpes fulva), raccoons (Procyon lotor), and North American racers (Coluber constrictor) (Burger 1977; Hackney 2010; Mitchell 2012).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Field Methods

We encountered numerous M. terrapin nests that had been excavated on FINWR during an inventory of the island's herpetofauna in 2006 (Mitchell 2012). We flagged nests dug by terrapins during the nesting season in June and July. All had been excavated by predators, most likely raccoons, which have reached high densities on the island (P. Denmon, pers. comm., May 2006) and are well known for their predation on turtle nests (Ernst and Lovich 2009). Nests destroyed by raccoons are identified by the presence of scattered egg shells, whereas crows and gulls remove eggs individually during oviposition and fly to another location to consume them (Burger 1977; Hackney 2010); we found none of the latter. Some nests could not be located precisely despite the presence of eggshells because the nest chamber was not visible and because predators apparently scattered the shells away from the nest sites (J.C. Mitchell, pers. obs., June 2006). We returned to these sites on 19 October 2006 when we obtained nest location coordinates using a Garmin eTrex Vista Cx® global positioning system unit and measured the location of each nest from the nearest patch of woody vegetation to the nearest 0.5 m. None of the woody plants or habitat islands had expanded measurably since spring (Fig. 2). We scored the area around each nest for one of the following microhabitat characteristics: tree canopy, shrub canopy, no canopy, bare sand, grass cover, and herbaceous (broad-leaved herbaceous plants) cover. We could not determine locations of nests that had escaped predation; intact nests are usually identified only during or just after oviposition (Feinberg and Burke 2003; Feinberg 2004). However, we assume that raccoons searched open areas with or without grass cover, as well as along and within the edges of the habitat islands, and that the majority, if not all, of the nests in our study area had been discovered and likely excavated (see Burke et al. 2005; Ernst and Lovich 2009).

Statistical Analyses

Originally, we planned to use a χ² test for 2 independent samples (Siegel and Castellan 1988) to examine whether the frequency of oviposition sites (nests) differed with respect to canopy (open, shrub, and tree) and ground cover type (open, grass, and herbaceous). However, the number of categories within each of these habitat variables produced a contingency table with expected frequencies less than 1 in some cells, which is a violation of an assumption of the χ² test (Siegel and Castellan 1988). We therefore combined adjacent categories to increase the expected frequencies in various cells (Siegel and Castellan 1988) and conducted a 2 × 2 χ² test (2-tailed) to examine whether the frequency of oviposition sites differed when canopy cover was either absent or present, and when ground cover was either absent or present.

Because low expected cell frequencies precluded us from using all categories of canopy and ground cover in a 2-sample test, we conducted a separate 1-sample χ² test for each habitat variable. We tested whether the observed frequencies of destroyed nests differed among the 3 classes of ground cover (open, grass, and other herbaceous material) and, separately, whether observed nest site frequencies differed with respect to canopy type (open, shrub, and trees). Because the same data were analyzed twice in these analyses, we reduced our critical level of α to 0.025 according to Bonferroni's inequality rule (Snedecor and Cochran 1980). Last, we asked whether the distribution of nests in relation to the edge of the nearest patch of woody vegetation was different from a uniform distribution using a Kolmogorov-Smirnov 1-sample goodness-of-fit test (Siegel and Castellan 1988).

Results

Female Malaclemys terrapin used the patchily vegetated sand dune area on the island (Fig. 2) extensively for selection of nest locations. The number of destroyed terrapin nests (48) found in the open vs. under canopy, and in the absence vs. presence of ground cover was statistically similar (Table 1: χ²  =  0.2765, df  =  1, p > 0.05, 2-tailed). However, with respect to ground cover, 56% of the nests were in the open, 40% in grass cover, and 4% among other herbaceous material. Similarly, we found 37% of nests in areas without canopy, 48% under shrub cover, and 15% under tree canopy. When each of these 2 factors (ground cover and canopy) were analyzed separately with 1-sample χ² tests, there were significant differences between observed and expected frequencies among the 3 categories of ground cover, but not those of canopy. More destroyed nests than expected were found in the open with no ground cover, whereas fewer nests were observed in herbaceous material (other than grass) (Fig. 3a: χ²  =  20.375, df  =  2, p < 0.001, 2-tailed). In contrast, there were no significant differences between observed and expected frequencies of the 3 types of canopy (Fig. 3b: χ²  =  5.75, df  =  2, p > 0.05, 2-tailed). Further collapsing these data to examine whether differences in nest site selection occurred in the absence (18 nests) vs. the presence of canopy (30 nests) likewise revealed no statistical difference (binomial test, z  =  −1.59, p  =  0.1118, 2-tailed).

View Figure

• Download as Powerpoint
• Download Figure

Table 1. Observed frequency of Malaclemys terrapin oviposition sites found in the open vs. under canopy and in the absence vs. the presence of ground cover. Numbers in parentheses are expected values.

View Fullscreen

Observed distances of destroyed nests from the edge of the nearest patch of woody vegetation (Fig. 2) were significantly different from a normal distribution (Fig. 4). Most were placed at the edge or within 0.5 m inside the woody patch.

View Figure

• Download as Powerpoint
• Download Figure

Discussion

Female M. terrapin on FINWR, the only barrier island in the mouth of the Chesapeake Bay, nested in sandy areas with no vegetation, in areas with variable densities of grass and herbaceous cover, and under cover in patches of woody vegetation that form habitat islands. Our haphazard searches for reptiles (Mitchell 2012) and observations by Hackney (2010) showed that terrapins nest throughout the island. The portion of the island we studied had a high density of nests as revealed by the number of readily apparent excavations and egg shells. We believe there was no bias in our search patterns, which included nonvegetated open areas and under vegetation along habitat edges. Our results underscore the need to understand predator search behavior and how it may influence our ability to obtain an accurate picture of where females nest in different habitat types.

We could not determine locations of nests that had escaped predation because we were not present during or immediately after oviposition (Feinberg 2004). However, we assume that raccoons searched open areas with or without grass cover, as well as along and within the edges of the habitat islands, and that the majority, if not all, of nests had likely been discovered and excavated. Hackney (2010) reported nearly complete nest loss on Fisherman Island in 2007, although she based her conclusion solely on predator-destroyed nests. Of the nests we located, fewer than expected were found in areas with herbaceous ground cover relative to those found in open sandy areas and in patches of grass. Most of the nests, however, occurred at the edge of, or slightly within, islands of woody vegetation that afforded variable amounts of shade and thus variation in temperatures. In a study conducted on Fisherman Island after 2006, the best-supported models revealed that occurrence of open sand, distance to marsh, and distance to road were the most important landscape and microhabitat variables describing M. terrapin nest locations (Hackney 2010). However, these results revealed little insight into microhabitat selection for nest sites.

Female M. terrapin use a wide variety of habitat types for nesting throughout their range. The pattern of nest placement by M. terrapin on Fisherman Island we describe appears to differ from most other sites reported in the literature. For example, half of the females nesting at Sandy Neck Beach in Cape Cod, Massachusetts, constructed nests on nonvegetated dunes and at Brigantine National Wildlife Refuge in New Jersey, half nested on vegetated dunes with < 75% shrub cover and 5%–25% grass cover (Burger and Montevecchi 1975; Burger 1977; Auger and Giovannone 1979). Feinberg and Burke (2003) documented nesting at Gateway National Recreational Area, New York, on the beach, in dunes with mixed grasses, in shrub-dominated habitats, and on trails. A river population in northeastern Rhode Island nested primarily in sparsely vegetated areas with 0% to < 5% vegetation (Goodwin 1994). Terrapins in Patuxent River, Maryland, a tributary of the Chesapeake Bay, nested on narrow sandy beaches, in open areas away from shade, and in agricultural fields ± 200 m from shore (Roosenburg 1994). Farther south, terrapins on a small island near the Intracoastal Waterway in northeastern Florida nested near the high tide line on the beach and in sandy, sparsely vegetated areas (Butler et al. 2004). And on Merritt Island, Florida, females nested on dikes, sandy roads, and banks of roads and wetlands (Seigel 1984). The pattern of nest placement on Fisherman Island appears to be most similar to that described by Roosenburg (1996) for a Patuxent River population where most nests were placed in open sun and partial shade. “Sparsely vegetated” is the most commonly used descriptor of the microhabitat in which M. terrapin females choose to put their nests; however, that may not be accurate for all nesting areas.

Our results may, in part, reflect the search strategy used by raccoons on Fisherman Island. Nests in open sandy areas without vegetative cover may be more difficult for them to find than those associated with grasses or woody vegetation. Raccoons are often reported to occur more frequently along habitat edges where predation pressure on bird and turtle nests is high (e.g., Wilcove 1985; Temple 1987; Dijak and Thompson 2000). Larivière (2003), however, argued that evidence is lacking for preferred use of edges by predators, including raccoons. The turtle-nest–searching behavior of raccoons has not been well studied but would provide insights into whether they preferentially search along habitat edges or if they search areas based on experience or discovery by olfaction. Do they miss nests located in open areas without vegetation? Several studies of turtle nest predation suggest they do not miss much (Congdon et al. 1994; Mitchell and Klemens 2000; Feinberg and Burke 2003; Hackney 2010).