Study Sites and Time Lines

Two Ouachita map turtle (Graptemys ouachitensis Cagle 1953) nesting sites along the lower Wisconsin River within 10 km of Spring Green, Wisconsin (lat 43°10′38″N, long 90°04′02″W), were studied during 2013 and 2014. I estimated the date to start nest monitoring based on historical associations of annual nesting initiation and preseason air temperatures at these sites (Geller 2012a) and continued monitoring until > 10 d after the last documented nesting or transient turtle at each site (late May until mid- to late July). Both study areas are located on relatively flat, deep glacial outwash sands that remained as sand throughout the soil column (98% sand, 2% silt, < 1% clay under sodium hexametaphosphate-separation analysis). Xerophytic vegetation covered approximately 20% of the surface at site A and 10% at site B, with the remainder being open sand (see also Geller 2012a).

Experimental Plots

At site A, a 112-m² area exhibiting visually uniform vegetative composition and percent cover was chosen for study based on past knowledge of nesting activity. This area was subdivided into 4 adjacent, alternating areas of 6.1 × 3.7 and 6.1 × 5.5 m, designated as 2 unswept control plots and 2 swept treatment plots (Fig. 1). Similarly, at site B, a 157-m² area was subdivided into 4 adjoining control and treatment plots that were 10.6 × 3.7 m each (Fig. 1). Plot designations used in 2013 were interchanged at each site in 2014. Lightweight polypropylene cord was strung between opposing marker stakes at ground level to demarcate plot side boundaries and guide the sweeping treatments. Coarse debris and some of the silt deposited by spring floods were removed from both sites by raking prior to each year’s fieldwork. Preseason leveling of substrates was also performed as necessary to ensure adequate broom contact during sweeping.

View Figure

• Download as Powerpoint
• Download Figure

Nests constructed by turtles in control plots served to document natural predation rates, and these areas were not walked in or otherwise disturbed. In an attempt to simulate natural conditions, excavations made by predators at natural nests were not refilled. Eggshells on the surface outside of depredated nests were allowed to remain in situ in control plots and were swept in treatment plots (i.e., were ignored during treatment application).

Treatment plots underwent surface modification by dragging (hereafter “sweeping”) a 3-headed push broom assembly (individual broom heads 61.0 cm long with 7.6-cm medium-stiffness bristles, 6.0-kg total assembly weight) under light to moderate pressure while walking nonoverlapping transits. Sweeping affected the top ~ 1–2 cm of substrate on these sites under dry sand conditions. Treatment plots were swept daily from the beginning of monitoring periods to ≥ 7 d after the last documented nesting turtle (sweeping periods 29 May–18 July 2013 and 29 May–12 July 2014 at site A, 29 May–13 July 2013 and 29 May–12 July 2014 at site B). Each day, the treatment plots were swept in the opposite directions as the day before to prevent unidirectional displacement of surface materials. Sweeping typically began at around 1700 hrs each day, took approximately 5 min at each site, and was initiated at alternating sites on a daily basis. However, study sites were not visited (nor were treatments applied) during rainfall (i.e., more than light sprinkles), when rainfall was imminent, or when the sites were temporarily flooded (several days in June 2013).

Artificial Nests and Surface Facsimiles

Artificial nests and manufactured surface markings were used to investigate the cues used by raccoons to locate turtle nests. Artificial control nests (hereafter “unswept artificial nests”) were constructed by hand while wearing nitrile gloves by digging a 12–15-cm-deep cavity, approximating the typical depth of Ouachita map turtle nests (G.A.G., pers. obs.; Vogt 1980). Excavated materials were allowed to accumulate on the nearby surface during construction and then subsequently replaced using moderate pressure, approximating the actions of turtles during natural nest construction. In an attempt to replicate the surface appearance of natural nests, an approximately 31-cm-diameter circular area above each refilled cavity was smoothed over and then incised to a depth of approximately 1 cm with a random, swirling pattern using 4 spread fingertips. This marking produced representations of both the surface markings and soil texture differences found at natural nests (hereafter, collectively, “surface markings”; see also Vogt 1980; Fig. 2). These artificial nests were used to isolate the influence of turtle-related olfactory cues on predator excavation rates, as they lacked turtle and egg scents but retained proxies for potential surface marking and disturbed soil profile cues produced during natural turtle nest construction.

View Figure

• Download as Powerpoint
• Download Figure

Artificial treatment nests (hereafter “swept artificial nests”) were constructed in the same manner as unswept artificial nests but were subsequently swept over within a several square-meter pattern. These were used to supplement unswept artificial nest investigations on the presence of disturbed soil profiles without turtle-related olfactory cues and, by comparison, to isolate the effect of the elimination of surface markings on excavation rates.

“Top-layered” artificial nests were constructed using nitrile gloves by first collecting the top 0.5–1.0 cm of surface material from an approximately 12-cm-diameter area and placing it in a clean plastic container. Cavities were then constructed in the same manner as standard artificial nests but with the excavated materials collected into another clean plastic container rather than deposited on the nearby surface. This material was then lightly mixed, carefully replaced within the cavity using moderate pressure, and topped off with the previously collected surface material and only as much similar substrate collected nearby (upper 0.5–1.0 cm within 0.75 m) as needed to produce a similar appearance with surrounding surfaces. These artificial nests retained potential subsurface soil profile disturbance olfactory cues beneath thin layers of local surface material exhibiting little or no visual or olfactory differences with nearby substrates.

Additional series of artificial nests were constructed in 2014 to investigate the influence of cavity depth on raccoon detection. Along with newly created surface facsimiles (see below), these were cavities constructed using nitrile gloves at depths of 2, 3–4, 6–7, and 12–15 cm in the same manner as top-layered artificial nests but with all of the excavated materials sequestered into just 1 clean plastic container and then carefully replaced and covered with local surface materials, as before. Additionally, these cavities were made by working within a 30-cm length of 77-mm (inside diameter) PVC pipe with referencing depth markings to better standardize excavation depths and cavity dimensions across different moisture gradients (e.g., artificial nests made in very dry substrates are otherwise not as tightly constructed as those made in more moist conditions). These “graded-depth” series were designed to present foraging raccoons with different signal strengths of soil-related olfactory cues concealed beneath thin layers of local surface material, as in the top-layered nests.

Finally, I also constructed “surface facsimiles” that were only imitations of the surface markings left at natural map turtle nests at the study sites, made as described above as components of standard artificial nests using fresh gloves but without the underlying cavity (Fig. 2). These were used to evaluate the presence of this visual cue, alone, on excavation rates, isolated from potential turtle olfactory and soil cavity disturbance cues.

All of the above artificial nests either were positioned 0.25 m interior to plot boundaries or opportunistically located nearby and were constructed at irregular intervals (typically 1–5 d) after nesting began at a given site. While some artificial nests were constructed singly and well separated from others, in most cases different types of these artificial nests were constructed in groupings 1–3 m apart (hereafter “close arrays”), such that if 1 element was excavated, predator proximity was confirmed. Close arrays thus reduce chances for “false negatives” resulting from a simple lack of nearby predators while still allowing for meaningful comparisons relating to detectability (after Strickland et al. 2010). Typically, small, locally obtained natural sticks were laid on or stuck into the ground within 1–2 m of individual artificial nests at varying compass directions to provide references to nest locations. Artificial nest fates were tracked for the first 4 d after construction, well within typical map turtle nest predation time lines at these sites (Geller 2012b). Undepredated nests experiencing significant (≥ 10 mm) rainfall within the first 2 nights after construction were removed from analysis except as noted, as significant precipitation may influence nest survival (e.g., Carr 1952; these sites, Geller 2012b). Soil moisture levels for each site were qualitatively assessed during daily site visits and artificial nest construction using subjective classes (very dry to thoroughly wet) based on visual, weight, and formability characteristics.

Data Collection and Analysis

Data on turtle nesting and natural nest predation was obtained by monitoring control and treatment plots with 4 digital trail cameras at each site (1 RECONYX™ HC600 HyperFire™, 3 RECONYX Silent Image™ RM30; RECONYX, Inc, Holmen, WI). RM30 models used infrared wavelengths peaking at approximately 850 nm for nighttime images, with illuminators that glowed in the visible red when activated, while the HC600 models did not produce visible light during nighttime use (wavelengths > 900 nm; J. Thiner, RECONYX Inc., pers. comm., May 2012). HC600 cameras took continuous 1-min time-lapse (TL) images and motion-triggered (MT) series. However, RM30 cameras, programmed with a more limited schedule to address the possible but unlikely influence of red wavelength illumination on nest detection by predators (see Geller 2012b), took continuous TL images at 5-min intervals and MT series from 0500 to 2100 hrs. Cameras were angled downward 30°–35° in sheltering boxes mounted on poles at heights of approximately 2.5 m, yielding a combined field of view (FOV) of approximately 232 m² at each site (Fig. 1). Camera data cards were changed out 2 or 3 times a week, and batteries were exchanged weekly. Both camera models performed reliably throughout the study period with no noted camera failures.

Although natural nests were not marked in situ, the elevated camera positions provided accurate nest delineation within FOVs and allowed each nest’s location to be drawn on a printed photo of the relevant camera FOV and referenced by a unique identification code to track individual histories. Artificial nest locations, mostly not within camera FOVs, were similarly depicted on hand-drawn maps on the day of their construction and had their corresponding survival status recorded during daily site visits. Camera data were also used to assess the degree to which raccoons appeared to follow the routes made by nesting turtles. This was done by drawing both pre- and postnesting turtle locations (from 1-min TL images) from unswept nests on printed photos of the relevant camera FOV as well as the paths that raccoons took before and after depredating these nests (from multiple images in close succession derived from MT image series) and then visually assessing the degree of route overlap.

Precipitation was recorded daily using on-site rain gauges, with the timing of rainfall determined by a funnel-driven waterwheel device (Geller 2012a) placed within 1 camera’s peripheral FOV at each site.

Reported results are from both sites and years combined, unless otherwise noted. Predation rate comparisons between unswept and swept treatments for all natural and artificial nests were performed using 2-tailed Fisher’s exact tests with Minitab 16 software (Minitab Inc, State College, PA) with the statistical significance of derived p-values set at α = 0.05. Fisher’s exact tests were used to compare these binomial proportions, as the result probabilities are exact (rather than estimates yielded by χ² tests) and are considered more accurate than χ² tests when expected values under the null hypothesis are small (McDonald 2014).

Predation Rates on Natural Nests

All but 1 of 20 unswept and all 16 swept natural map turtle nests were depredated by raccoons; the 1 unswept nest from 2014 that was not depredated was still intact at the end of the study period on 20 July. There was no difference in predation rates between unswept and swept natural nests (2-tailed Fisher’s exact test, p = 1.00). For both years combined, natural nests also included 7 treatment area nests depredated before sweeping and not included in treatment nest predation metrics as well as 2 additional excluded depredated nests with uncertain construction and treatment application time lines. All map turtle nests are believed to have been constructed in graded to thoroughly moist soil columns under dry to moist surface conditions. All natural nests were depredated within 24 hrs of construction, and none experienced precipitation before depredation in either study year.

Comprehensive MT image series were available to fully delineate the paths raccoons took before and after depredating 2 unswept turtle nests. Review of these camera data indicated that raccoon paths were different than paths made by the nesting turtles, as inferred from 1-min TL images showing their pre- and postnesting positions within the relevant FOV (e.g., Fig. 3). In both cases, the routes of raccoons and turtles showed no apparent areas of concurrence, and there were no images of raccoons at known turtle positions.

View Figure

• Download as Powerpoint
• Download Figure

Predation Rates on Artificial Nests

Paralleling predation rates on natural nests, excavation rates were high for both unswept (20 of 20) and swept (18 of 19) standard artificial nests in close arrays where at least 1 element was predator disturbed and did not differ (2-tailed Fisher’s exact test, p = 0.49; Table 1). Artificial nests were excavated both when there was limited moisture within the soil column, resulting in little or no obvious postconstruction color/luminance contrasts or moisture differences with surrounding surfaces (11 of 11 unswept and 9 of 10 swept standard artificial nests), and when there was a moderate increase in soil moisture with column depth, resulting in greater differences in surface appearance and representing soil conditions more typical of those encountered in natural nesting situations (8 of 8 unswept and 8 of 8 swept standard artificial nests). The difference in soil column moisture thus appeared to have no effect on excavation rates for either unswept or swept artificial nests; however, how long these surface differences persisted as possible nest location cues to foraging raccoons is unknown. Other unswept and swept artificial nests constructed in thoroughly moist to wet soils were also excavated in both close array (2 of 2) and nonarray (8 of 14) settings.

Table 1. Predation rates (no./total no. and % in parentheses) on natural Graptemys ouachitensis nests and artificial nests in depredated arrays at 2 sites along the lower Wisconsin River, Iowa County, Wisconsin. See text for additional details on artificial nest construction.

View Fullscreen

As with natural nests, raccoons were believed to be the sole excavator of artificial nests at both sites during this study, as no other mammalian predators were seen on camera except for 1 transient Virginia opossum (Didelphis virginiana) at site A in 2014, and all mammal tracks noted around and sometimes inside excavated artificial nests were those of raccoons. Also like natural nests, almost all disturbed unswept and swept artificial nests with well-defined survival time lines were excavated within 24 hrs of construction (32 of 33), with no difference in ratio (2-tailed Fisher’s exact test, p = 0.43, n = 77, for natural nests vs. standard artificial nests), and all were excavated without experiencing intervening precipitation. The only depredated standard artificial nest surviving more than 24 hrs was depredated within 48 hrs of construction.

Top-layered artificial nests, putatively requiring the olfactory detection of reordered soils beneath thin layers of relatively unmixed surface material for their discovery, were also excavated at high rates (8 of 10) by raccoons in both dry and moist soil columns (Table 1). Although only coarsely assessed due to small sample sizes, the excavation rates of artificial nests within the similarly visually inconspicuous graded-depth arrays showed a positive association with cavity depth, with the deeper 2 cavity depths in the series disturbed by raccoons significantly more frequently (12 of 24) than the shallower 2 cavity depths (5 of 26; 2-tailed Fisher’s exact test, p = 0.04, n = 50; Table 2).

Table 2. Predation rates (%, sample sizes in parentheses) on graded-depth artificial nests in depredated arrays for 2014 at 2 sites along the lower Wisconsin River, Iowa County, Wisconsin. See text for details.

View Fullscreen

Surface facsimiles experienced significantly lower excavation rates (5 of 19) than artificial nests with standard depth cavities (46 of 49; 2-tailed Fisher’s exact test, p < 0.001, n = 68; Table 1). For all types of artificial nests with a manufactured cavity and in all soil conditions, the excavations made by raccoons were single, on-target efforts, without any evidence of uncertainty as to their locations even when visual surface cues appeared absent to a human observer.

Effect of Sweeping Treatment on Natural Nest Predation Rates

Removal of the surface markings produced by map turtles during nesting forays and nest construction by sweeping the nesting areas with a broom was ineffective in reducing nest predation in this study: all nests that were swept were depredated. This result indicates that visible cues are not necessary for raccoon detection of newly constructed turtle nests and suggests the alternate possibility that olfactory cues are of ultimate importance to nest-foraging raccoons. In the present study, MT camera image series consistently documented the rather deliberate nest-searching behavior of foraging raccoons with noses held within a few centimeters of the ground, implying a primarily olfactory-based searching method. These findings also suggest that olfactory rather than visual signal persistence determines the effective nest predation window for raccoons given their particular sensory abilities (typically the first few days following nest construction; see Galois 1996 and Geller 2012b). These results are consistent with the results of Burger (1977), who found extensive diamondback terrapin (Malaclemys terrapin) nest predation when obvious visual cues were absent due to the effects of wind and rain, as well as those of Feinberg and Burke (2003), who similarly reported high levels of diamondback terrapin nest predation (92%) even though the nests were typically inconspicuous (Burke et al. 2005).

The clarity of the present study’s findings indicating the predominant importance of olfactory cues in nest location by raccoons at these sites contrasts somewhat with previous efforts to delineate underlying causal factors. Soil disturbances inherent in the construction of natural nests and many types of artificial nests produce both visual and olfactory signals that have led to difficulties in resolving which are used as nest location cues by raccoons. Thus, while Burke et al. (2005) concluded that soil disturbance was likely an important nest location cue for raccoons in their artificial nest studies (various types, all smoothed over postconstruction), they suggested that disturbances at natural nests may be detected visually, although they did not test for visual detection directly. In other experiments with artificial nests (inverted soil without eggs), Strickland et al. (2010) implied that raccoon predation was associated with visible surface soil disturbance, without discussion of potentially co-occurring soil odor cues from their manufactured nest cavities. Finally, while Wirsing et al. (2012) noted that both the greater visual and soil- and egg-related olfactory sign left at nests by snapping turtles (Chelydra serpentina) may have caused higher raccoon predation rates than was observed at the more cryptic nests of painted turtles (Chrysemys picta), the relative importance of these potential cues was not assessed.

Insights from Artificial Nest Experiments

The high excavation rates on both unswept artificial nests retaining surface markings (100%, n = 20) and those without them (83%, n = 42; swept and top-layered artificial nests and the standard 12–15-cm-deep cavities of the graded-depth series) show that raccoons can locate and will excavate manufactured cavities lacking both the nest excursion tracks and olfactory cues from female turtles and the surface markings that remain after nest construction. Other research has also demonstrated that turtle-related cues are not necessary for raccoons to investigate artificial nests (e.g., Wilhoft et al. 1979; Burke et al. 2005). Significantly, camera data from each of 5 nest predation events with comprehensive MT images at these Wisconsin River sites (present study and Geller 2012b; all nests depredated within 24 hrs and without intervening precipitation) show that, at least in the context of these nesting areas, raccoons do not appear to regularly follow the residual visual or olfactory trails produced by nesting turtles in locating natural nests, in contrast to earlier suggestions (e.g., Congdon et al. 1983). Similarly, although they may have been visually detected by foraging raccoons, surface facsimiles were excavated relatively infrequently when compared to artificial nests with cavities. Under the assumption that the facsimiles were adequate representations of postnesting markings, this result lends further evidence that raccoons at these sites do not primarily associate turtle nest presence with visible surface signals. Indeed, the excavation rate of surface facsimiles reported here may actually be inflated due to my misinterpreting occasional bird or rabbit dust-bathing marks for raccoon disturbance. Interestingly, surface facsimiles were typically only shallowly excavated when disturbed, suggesting that raccoons terminated excavation behaviors when they tactilely or olfactorily sensed the lack of an underlying cavity.

Combined with the high predation rates observed on both upswept and swept natural nests, the present findings indicate that the operative cues used by nest-foraging raccoons are olfactory and related to disturbed soil profiles, as this was the only component common to all types of frequently depredated natural and artificial nests and was responsible for high predation rates regardless of type. Contrarily, the presence or absence of visual cues to underlying cavities, as well as olfactory cues of turtle- and turtle egg–related origin, did not affect predation rates. Higher raccoon excavation rates on deeper nests within the graded-depth artificial nest series also suggest that the signal strength of these soil disturbance olfactory cues may be positively associated with cavity depth, although it should be noted that the greater volumes of manipulated soil in deeper-constructed cavities may have, in itself, resulted in increases in cue signal strength aside from depth related influences per se.

Potential Olfactory Cues in Disturbed Soil Profiles

Among other potential olfactory cues that may result from disturbed soil profiles are increases in aerosolized compounds produced by soil microbes, perhaps in particular the human-sensible odor of the actinomycete metabolite geosmin (Lindbo et al. 2012). Actinomycetes are widespread in soil ecosystems, are well represented in dry habitats such as those often used by nesting turtles, and increase in density with depth within the top several centimeters of soil profiles (for an example in a semiarid habitat, see Fierer et al. 2003), properties consistent with the well-drained soils and higher excavation rates on deeper artificial nests found in this study. Although speculative, as microbial populations and their associated compounds were not assessed, the use of depth-related microbial cues by nest-foraging raccoons, rather than moisture gradients (suggested by Wilhoft et al. 1979), may account for the high rates of artificial nest detection across a broad range of soil moisture conditions, including those for which the soil profiles appeared both thoroughly dry and moist during their construction. Although these particular artificial nests displayed limited or no surface or subsurface differences in soil moisture with surrounding materials, they did have the underlying soils remixed within the manufactured cavities like natural nests, potentially resulting in olfactorily sensible releases of geosmin into the air from within the entire column, including that from soils formerly at lower depths potentially containing higher initial geosmin concentrations.

A geosmin cue could also help explain the commonly observed reduction in turtle nest predation following substantial precipitation (e.g., Bowen and Janzen 2005) because after rainfall, a widespread flush of geosmin into the atmosphere occurs as infiltrating rain displaces the air between soil particles (Lindbo et al. 2012). This phenomenon may reduce the preexisting olfactory gradient in geosmin concentration between nests and surrounding substrates, with larger rainfall amounts possibly being more effective in reducing predation due to percolating to greater, geosmin-containing depths. Similarly, the typical association of decreased predation risks with nest age may result from the declining signal strength of volatized geosmin from postconstruction substrates as natural actinomycete population gradients rebuild within nest cavities.

Although similarly tentative, species-related differences in levels of volatized geosmin produced during nest construction may be part of the complex of factors that influence differential nest predation risk among turtle species. For example, the typically lower predation rates of painted turtle nests compared to those of snapping turtles (mean ~ 36% vs. mean ~ 77%, respectively; reviewed in Wirsing et al. 2012) may reflect lower amounts and concentrations of geosmin released from the smaller and shallower nest cavities of painted turtle nests than that produced from the much larger excavations made by snapping turtles (e.g., nest depth range 9.9–10.4 cm vs. 7–20 cm or more, respectively; Ernst and Lovich 2009), paralleling and refining the suggestion of Wirsing et al. (2012) that increased visual and olfactory signals at snapping turtle nests rendered them more likely to be detected by predators.

Given that geosmin has been shown to be ecologically involved in other biological systems (e.g., guiding glass eels, Anguilla anguilla, to freshwater during migration [Tosi and Sola 1993]; identifying unsuitable breeding and feeding sites in Drosophila [Stensmyr et al. 2012]) and is, indeed, recognized by humans as the smell of newly exposed soil, variations in its signal strength make it an intriguing candidate as an olfactory nest location cue and worthy of further study.

Management Implications

While study results suggest that management attempts at reducing olfactory nest location cues may have potential in reducing raccoon and other likely olfactory-based mammalian predation (e.g., striped skunks, Mephitis mephitis; Galois 1996) on newly constructed turtle nests, the development of sensory-based nest protection methods based on concealing nest locations or by repelling predators with chemical deterrents should proceed cautiously. Future research along these lines will need to consider the potential impacts different approaches may have not only on the turtle nests themselves but also on nesting and transient turtles and hatchlings. For example, in some of the first work to appear involving the use of olfactory-based repellents to reduce nest predation rates, Lamarre-DeJesus and Griffin (2013) noted that, while apparently effective at reducing coyote (Canis latrans) nest predation levels on loggerhead nests (Caretta caretta), the possible deleterious effects of their habanero pepper powder treatments on hatchlings were unknown and required further study.

Due attention must also be paid to treatment effects on nesting habitats. Daily broom sweeping decreased standing vegetation on the xeric, low-productivity soils at these study sites over the course of each nesting season (G.A.G., pers. obs.). While potentially more effective in reducing predation rates, more aggressive soil disruption than that used in this study or the use of chemical masks or deodorizers that have long persistence time lines or that influence soil fertility, pH, and so on may be ecologically unsuited for many areas of currently optimal habitat or may require a spatially or temporally staggered implementation. Preferred methods would have, at most, only nonaccumulating, short-term impacts on both biotic and abiotic components, including nesting and nest incubation environments, plant cover extent and species composition, and soil profile structure and olfactory characteristics.

Olfactory-based approaches will likely have little utility in countering the often substantial proportions of nest loss attributed to visually oriented avian predators (e.g., fish crows, Corvus ossifragus, in the southern United States) and may ultimately not produce greater nest protection results than those provided by currently existing methods focused on predator exclusion; for example, traditional nest screening of individual nests (see Lamarre-DeJesus and Griffin 2013) and, in some contexts, electric fencing for nesting areas (e.g., for reducing raccoon predation on map turtle nests; Geller 2012c).