Study Site

The study site was in the western Upper Peninsula of Michigan along a 1.5-km section of river in the Ottawa National Forest that is several kilometers from a paved road, with little to no recent human disturbances. No human activity other than our research team was observed during the entire 2009 and 2010 field seasons, suggesting that this population is relatively unaffected by human disturbance and is in an essentially pristine habitat. The meandering river had several sandy–gravel bars and cut banks suitable for nesting.

Experiment 1: Effects of Artificial Nest Treatment on Predation

Artificial nests (n = 224) were constructed on 9 nesting beaches between late May and mid-June, which is when nesting usually occurs in wood turtle populations in the northern part of their range (Walde et al. 2007). Artificial nests were constructed on 27 May (n = 40), 30 May (n = 44), 2 June (n = 32), 11 June (n = 36), 17 June (n = 40) and 23 June (n = 32) in 2009. The last 3 dates coincided with natural nest deposition.

One of 4 treatments was applied to each artificial nest: 1) water in which a female wood turtle had resided (hereafter turtle water); 2) soil disturbance; 3) turtle water plus soil disturbance; and 4) distilled water (control). Nests were placed in groups of 4 with 1 group receiving each of the 4 treatments. The groups were placed 2 m apart in a symmetrical arrangement with nest treatment randomized within the group. Nest groups were separated by at least 10 m. All artificial nests were constructed in areas where wood turtle nesting activity (e.g., nest digs and tracks) had been observed by J.L.R. or Ottawa National Forest personnel.

A female wood turtle was collected from the study area on 19 May 2009 and transported to Northern Michigan University to collect chemical cues for artificial nest treatments. The turtle was placed in a 76-l aquarium where it remained for 5 days. It was fed greens plus 1–2 earthworms per day. The aquarium contained distilled water 20 cm deep and a rock with a wooden platform for basking. A heating lamp was used to heat 1 side of the aquarium. The room, water, and basking site temperatures were 24°C, 12°C, and 30°C, respectively. The turtle was released at the site of capture after this procedure. Control water was obtained from another 76-l aquarium that contained no turtle, distilled water 20 cm deep, a rock, and a wooden platform. Turtle water and control water were placed in 473-ml sterile plastic freezer containers, labeled, and frozen until use. Turtle water was discolored and smelled strongly of turtle excrement.

Artificial nests with turtle water were created by applying 60 ml of turtle water from a sterile measuring cup to the top of undisturbed soil. Turtle water was poured so as to prevent soil disturbance. Artificial nests with soil disturbance were created by digging a 10-cm-deep hole with a trowel. Average nest depth, derived from depredated nests of wood turtles, is approximately 10 cm (Brooks et al. 1992). The surface was smoothed to mimic natural nests. Control treatments (with no turtle water or soil disturbance) received 60 ml of distilled water applied gently to the soil surface. Rubber gloves and rubber boots were worn when applying all treatments to minimize human scent. Artificial nests were marked with a wire-stemmed flag labeled with nest treatment, and positions were sketched on a site map. Studies report that nest markers have no effect on nest predation by mammalian predators (Tuberville and Burke 1994; Burke et al. 2005; Strickland et al. 2010). Artificial nests were checked for predation on 3 consecutive mornings following nest creation and recorded as depredated or nondepredated. In previous studies, nests were most commonly depredated 24–48 hrs following egg laying (Tinkle et al. 1981; Congdon et al. 1983). Nests were counted as depredated if digging was visible or the nest cavity was exposed and counted as nondepredated if no apparent disturbance occurred. When a nest cavity was exposed, it was left untouched.

A log-linear logit saturated model was used to identify significant differences between treatment effects using the program PASW Statistics (v. 18.0). The dependent variable was nest fate (depredated or nondepredated) and the independent variable was nest treatment. The logit approach was appropriate because one of the categorical variables was clearly dependent, and this method is more effective than traditional chi-square analysis when many cells have frequencies of zero. One limitation of the logit model is that it does not control for beach effects on the dependent variable. Thus, data were additionally analyzed using a randomized block experimental design to block for beach. Only one randomly selected data point per beach was used in the analysis, lowering sample size to n = 9. A Cochran's Q-test (Cochran 1950) was applied to determine nest treatment effects on nest predation, with beach as the blocking effect. To test for significant differences between treatments, a multiple comparisons test was applied (Marascuilo and McSweeney 1967). Similar results were found for both approaches mentioned above. Because beach did not have an effect on the variable being tested, only the logit model is presented.

A paired t-test was computed by PASW Statistics (v. 18.0) to determine mean differences in nest predation counts during the 2 periods of nest construction (27 May–2 June vs. 11–23 June). A Fisher's Exact Test was run to determine if nest treatment had an effect on the type of predator that dug up artificial nests. Predator tracks and scat were noted and identified whenever observed. To determine predator activity patterns, predator trails observed on separate beaches were counted on 5 days prior to nesting (i.e., turtles were not ovipositing) and on 5 days during the nesting period (i.e., turtles were ovipositing). These beaches were spaced along the 1.5-km section of river to reduce the likelihood of recounting tracks from the same individual. A paired t-test in PASW Statistics (v. 18.0) was used to compare track counts prior to and during the nesting period.

Experiment 2: Effects of Nest Distance from River on Nest Predation

Artificial nests were constructed at 3 distances from the river to examine the effects of this variable on nest predation. These distances were ≤ 5 m (close), 7–10 m (medium), and ≥ 20 m (far). Owing to the nature of the nesting habitat, close nests were generally in sand and far nests were generally in vegetation, although none were under a tree canopy. Nests were considered to be in sand habitat if 50% or more sand was observed in 0.5-m² plots. Transects (a line of 3 nests, 1 at each distance) were created perpendicular to the river in areas that had evidence of turtle nesting activity and nest predators.

Sixteen transects separated by a minimum of 10 m were constructed on 22 May–10 June 2010. For each nest, vegetation was cleared if necessary and a 10-cm-deep hole was dug with a trowel. One chicken (Gallus gallus) egg was placed in the hole and sand was replaced. Sand plots were placed around each nest to record tracks for predator identification (Elbroch 2003). Sand plots were constructed of 4 pieces of cardboard 23 cm in length arranged in a square to surround each nest. Wet sand was placed 1 cm deep on top of the cardboard to record predator tracks. Rocks were placed at the 4 corners to secure sand plots in place. Strickland et al. (2010) found that researcher-associated cues (e.g., rocks, flags) did not attract or deter turtle nest predators. Nests were checked on the 3 subsequent mornings for predation.

A Pearson's chi-square test with a continuity correction was run to determine effects of distance from the river on artificial nest predation (n = 16). The frequency of depredated and nondepredated nests was compared for each level of the distance variable (close, medium, far). A simultaneous Bonferroni correction test was applied to the alpha value for post hoc comparisons between groups (Neu et al. 1974). All analyses were performed using PASW Statistics (v. 18.0) software with significance assessed at α = 0.05.

Experiment 1: Effects of Artificial Nest Treatment on Predation

In total, only 8% (18/224) of all simulated nests were depredated. Twenty percent (11/56) of nests with soil disturbance were depredated and 12% (7/56) of nests with soil disturbance plus turtle water were depredated. No artificial nests with turtle water alone or with distilled water control were depredated. Nest predation was significantly higher in the soil disturbance condition than in the turtle water condition (Z₃ = 2.299, p = 0.02) and distilled water control (Z₃ = −2.299, p = 0.020). Nest predation was not significantly different between soil disturbance condition and the condition with both soil disturbance and turtle water (Z₃ = 0.512, p = 0.318). Nest predation was not significantly different between the condition with both soil disturbance and turtle water and the turtle water condition (Z₃ = −1.928, p = 0.054) or the distilled water control (Z₃ = −1.928, p = 0.054).

Tracks of raccoon, river otter (Lontra canadensis), and American mink (Neovison vison) were found near depredated nests. Nine artificial nests were disturbed by raccoon, 8 by river otter, and 1 by American mink. If just raccoon and river otter are considered, then nest treatment did not have a significant effect on the type of predator that depredated the nest, but sample sizes were small (Fisher's Exact Test p = 0.134). Tracks of white-tailed deer (Odocoileus virginianus), gray wolf (Canis lupus), coyote (Canis latrans), black bear (Ursus americanus), and bobcat (Lynx rufus) were also seen on the river banks, but there were no tracks of these species near the nests.

Mean number of predator tracks per day on beaches was significantly greater during the nesting period when turtles were ovipositing (mean = 8.6 ± 1.14 SD, n = 5) than during the prenesting period, before turtles started ovipositing (mean = 1.2 ± 1.3 SD, n = 5) (t = −10.752, df = 8, p < 0.001) in 2009. The mean number of artificial nests depredated was also significantly greater during the nesting period than during the prenesting period (mean = 9 vs. 3, n = 18) (t = −2.449, df = 16, p = 0.026).

Experiment 2: Effects of Distance from River on Nest Predation

Forty-eight percent (23/48) of artificial nests in this experiment were depredated, with depredation of 69% (11/16) of close nests, 56% (9/16) of nests at medium distances, and 19% (3/16) of nests far from the river. These predation rates were significantly different among groups (Pearson's χ²₂ = 8.68, p = 0.013, n = 16). Post hoc comparisons of groups revealed a significant difference between predation at close and far nests (p = 0.002; Bonferroni correction p = 0.008) but no significant differences between predation of close and medium nests (p = 0.212) or medium and far nests (p = 0.050). Depredation usually occurred on the first day after nest creation (78%, 18/23). No nests were depredated on day 2 and 22% (5/23) were depredated on day 3. Artificial nests created in nonvegetated areas were depredated the most. Specifically, there was a significant difference in predation between nests made in open sand (16/20) and vegetated areas (7/28; Pearson's χ²₁ = 12.02, p = 0.001). However, this variable was confounded with distance from the river. Raccoon, river otter, and white-tailed deer tracks were detected on sand plots. Eighty-seven percent (20/23) of artificial nests were depredated by raccoon and 4% (1/23) were depredated by river otter. Nine percent of nest predators were unidentified due to rain (n = 1) or tracks of multiple species on sand plots (n = 1; white-tailed deer and raccoon).

Experiment 1: Effects of Artificial Nest Treatment on Predation

Results suggest that chemical, visual, or tactile cues associated with soil disturbance, rather than chemical cues from turtles, were the primary indicator for nest location by predators in this study. These results are similar to those for painted turtles (Chrysemys picta; Strickland et al. 2010) and ornate box turtles (Terrapene ornate; Bernstein et al. 2015), for which soil disturbance was a primary cue in turtle nest location by mammalian predators. Additionally, findings by Geller (2015) suggest that raccoons use olfactory cues in soil disturbance to locate nests of Ouachita map turtles (Graptemys ouachitensis).

Only 20% (11/56) of artificial nests that had soil disturbance were depredated. This low rate may reflect an incomplete set of cues used by nest predators in this experiment. Still, it seems likely that soil disturbance may be a primary determinant of nest predation, although cues such as turtle tracks and egg-derived chemicals may also be important. The absence of eggs in artificial nests in this study (as a reward for digging) could also explain the low depredation rate of artificial nests. This hypothesis is supported by the lower proportion of depredated artificial nests in experiment 1 (8%) compared to experiment 2 (48%), in which chicken eggs were present in the artificial nests. However, Strickland et al. (2010) and Bernstein et al. (2015) noted that eggs buried in artificial nests did not affect predator behavior. Another possible explanation for low depredation rates is that, because some trials were conducted prior to the onset of natural nesting, predators may not yet have focused on searching for this food resource. Trials were conducted 27 May through 23 June, but nesting was observed only from 14 June to 24 June. Twice as many artificial nests were depredated during the latter half of the trials (6 vs. 12). Conversely, Wilhoft et al. (1979) found that simulated turtle nests made in June were no more vulnerable than nests made in April.

In 2009, raccoon and river otter tracks were observed on nesting beaches before nesting activity started, and predator trails were observed on beaches more-often during the nesting season than at any other time during the summer. In 2010, there was no indication that predators foraged on beaches until 2 June, 11 days after nesting began. Early May was warm in 2010, which may have promoted early egg development and oviposition.

Both raccoons and river otters were able to locate artificial nests without turtle-derived chemical cues, as suggested by other studies of raccoon predation on turtle nests (Wilhoft et al. 1979; Hamilton et al. 2002; Strickland et al. 2010; Bernstein et al. 2015). River otters are not commonly reported as turtle nest predators. Their diets are specialized for piscivorous feeding, but they feed opportunistically on invertebrates, berries (Sheldon and Toll 1964), and overwintering adult snapping turtles (Chelydra serpentina; Brooks et al. 1991). Otters also feed on turtle eggs during nesting season, as observed in the present study, although they fed primarily on crayfish based on informal scat analysis. Nests with only turtle water were never detected by predators. However, this does not mean that predators did not detect the chemical cue applied. It is possible that turtle water deposited on artificial nests did not mimic the authentic scent left by egg-laying turtles and predators did not associate this smell with turtle nests. However, other studies have also shown that applying chemical cues, such as turtle urine or water taken from a source that turtles live in, do not affect nest predation (Hamilton et al. 2002; Strickland et al. 2010). It is possible that the application of turtle water mimicked a rain event and masked any chemical cues associated with soil disturbance. It is noted that rain prompts turtle nesting activity and nest depredation may decrease after heavy rainfall (Bowen and Janzen 2005). Wood turtle populations with decreased numbers may benefit by wildlife managers simulating an evening rain event (e.g., via sprinklers) on nesting beaches during peak nesting season. This may help erase soil disturbance cues that attract mammalian predators and could decrease predation rates. It is unclear whether visual, tactile, or chemical cues associated with soil disturbance attracted predators, but these seem to be most important in eliciting predator response.

Experiment 2: Effects of Distance from River on Nest Predation

Most predation on artificial nests occurred within 1 day after nest construction, as reported for natural nests (Tinkle et al. 1981; Congdon et al. 1983). Nest predation was greater closer to the river than further away, as has been reported for other turtle species (Congdon et al. 1987; Marchand et al. 2002; Spencer and Thompson 2003). Additional studies on habitat edge effects on turtle nest predation support these findings (Temple 1987; Marchand and Litvaitis 2004). Higher predation closer to the river is expected, as predators follow river banks while foraging (Cagle 1949), as observed in the present study. However, distance from water is confounded with vegetation density. Thus, nests in vegetated areas were depredated less and may have been more difficult to find.

The present study contrasts with other similar experiments (Burke et al. 2005; Strickland et al. 2010; Bernstein et al. 2015) in that raccoon populations were not artificially high, the study area was largely undisturbed by humans, and the most common nesting turtle was the wood turtle. Future direct comparisons between populations in relatively pristine habitats and those with more human activity could have important conservation implications with regard to the effects of human disturbance. Additionally, our results suggest that future studies seeking to investigate predation rates of wood turtle populations by using artificial nests may not need to apply turtle-derived chemical cues, as they seem to have little influence on predator behavior. However, the addition of eggs to artificial nests may increase predator search efforts and more accurately reflect predation rates.