Camera Monitoring and Data Collection

This study reports the camera-generated observations of nesting G. ouachitensis frequenting 2 nesting sites (referred to as Sites A and B) obtained during research on the development of an electric fencing strategy to reduce turtle nest predation (Geller 2012b). Sites were located using satellite imagery of the lower Wisconsin River corridor (http://www.weather.com/weather/map/interactive; accessed 25 October 2007) to remotely survey likely habitat, followed by on-site inspections to confirm previous nesting activity via destroyed nests. Both sites are on relatively level glacial outwash sands on the south bank of the Wisconsin River in Iowa County, Wisconsin, within 10 km of the town of Spring Green.

Site A is approximately 52 m from the river, bordered on the north and west sides by a mesic river birch (Betula nigra)–silver maple (Acer saccharinum) woodland and on the east and south sides by an open, dry-mesic oak (Quercus spp.)–ash (Fraxinus spp.) woodland. The nesting area (∼ 550 m²) consists of exposed habitat with bunch grasses (primarily switchgrass, Panicum virgatum) and xerophytic herbaceous plants (primarily lyre-leaved rock cress, Arabis lyrata; beach wormwood, Artemisia campestris; horsetail, Conyza canadensis; and peppergrass, Lepidium sp.) covering approximately 20% of the substrate, with the remainder being open sand. Site B (∼ 350 m²) is 15 m from the river, adjacent to an open sand beach. It is flanked to the north by sandbar willows (Salix interior) and to the east by open river birch upland. The nesting habitat is similar in species composition and percent open sand area to Site A.

Areas were continuously camera monitored from 10 May through 10 October during 2008, 2009, and 2011, although a mid-July flood in 2010 caused the loss of all nests and early termination of monitoring activities. June floods in 2008 and 2011 allowed for at least partial nest survival and continuation of site monitoring. Sites were visited 2 or 3 times a week to change the camera data cards, with efforts made to avoid travel in the camera-monitored areas.

Four digital trail cameras were deployed at each site (combinations of Reconyx Silent Image™ RM30 and Reconyx HC600 HyperFire™), yielding a combined field of view of approximately 232 m². Cameras were angled downward at approximately 30° within protective boxes mounted on poles at heights of approximately 2.5 m. Each pole was topped with an inverted 10-cm-diameter plastic funnel into which was inserted a sharpened dowel to discourage perching by American crows (Corvus brachyrhynchos). Cameras were programmed to continuously take both time-lapse (TL) images at 1-min intervals as well as sequential rapid-fire series when triggered by motion (MT). RM30 models took 0.3-megapixel (MP) black-and-white images, while the HC600 models took 3.1-MP color images. RM30 models used infrared (IR) 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 IR use (wavelengths > 900 nm; J. Thiner, Reconyx Inc., pers. comm., 2012). Both camera models incorporated accurate time and date stamp data, although the temperature stamp readings for the RM30 models tended to be higher than actual ambient air temperatures, especially under nonovercast conditions. Both models performed reliably throughout the 4-yr study period, without any noted failures or loss of data.

Daily temperature and precipitation data were obtained from National Weather Service records for the nearest locale (Lone Rock Airport, Sauk County; approximately 13 km northwest of study areas). Within-day temperatures were derived using camera stamp temperature data adjusted via correction curves based on on-site thermometer readings. On-site timing of rainfall events was provided by a funnel-driven, bicolored waterwheel enclosed within a Lucite-fronted housing placed within the peripheral camera field of view (Fig. 1). Time-stamped TL images indicating waterwheel movement delineated precipitation event time lines. These units were accurate to within 1–3 min of rainfall duration (sprinkles to heavier amounts), as determined by field tests and camera-visible nocturnal precipitation.

View Figure

• Download as Powerpoint
• Download Figure

While all female Graptemys observed basking on offshore logs and found on land near the monitored areas were G. ouachitensis, nesting individuals that could not be confidently identified to species were removed from analysis except as noted. Only Graptemys that had images showing conspicuous post-orbital bars (sometimes via computer-magnified images) were considered G. ouachitensis. Supplemental field marks, such as degree of dorsal keel development, overall/head size (relatively large in G. geographica), and well-developed tomia (G. geographica), were often available in the camera record (mean ∼ 50 images per nesting female) to aid species identification (Fig. 2). As turtles were not marked, females producing more than 1 nest in a season may be represented multiple times in some analyses. Reported sample sizes reflect varying numbers of camera records available for each analysis, as influenced by degree of intervening vegetation, distance, and other variables.

View Figure

• Download as Powerpoint
• Download Figure

Statistical Analysis

Pearson's correlation coefficient (r) was used to investigate the association of temperature at the time of nest construction and the duration of the nesting event. The relationship of midday nesting to mean maximum temperature during the nesting season was examined using logistic regression (Agresti 2007), with the number of midday nests out of the total number of nests each year as the response variable and mean maximum temperature during the nesting season as the single predictor variable. The influence of precipitation on nesting activity was assessed using 2-tailed Fisher's exact tests by comparing the proportion of nests constructed within 24 hrs of previous rain to the proportion of calendar days in the nesting season with rain. Alpha was set at 0.05 for all statistical tests.

Utilization of Sites and Nesting Season Chronology

Graptemys ouachitensis was the predominant nesting species for both sites in all study years, constructing 98.7% of the nests for turtles identified to species level (n  =  79), with 1 G. geographica nest also being documented. Twenty-nine additional nests (26.9%; n  =  108 total nests for both sites combined), all Graptemys, could not confidently be assigned to species but are also likely to have been constructed by G. ouachitensis. Of 142 transient turtles (moving through but not nesting on monitored areas), 53.5% (n  =  76) were unidentified to species but likely G. ouachitensis, 36.6% (n  =  52) were known G. ouachitensis, 3.5% (n  =  5) were G. geographica, 4.2% (n  =  6) were Chelydra serpentina, and 2.1% (n  =  3) were Apalone spp. Some transient turtles made aborted nesting attempts.

The nesting season for G. ouachitensis was from late May through early July across all study years, although flooding events in 2008 and 2011 may have influenced seasonal time lines and numbers of nesting females in those years (Table 1). The annual onset of G. ouachitensis nesting was associated with spring temperatures, with warmer spring months preceding earlier nesting activity at both sites (Table 1). Site A was used by more nesting and transient turtles than Site B (15 vs. 14 in 2008, 43 vs. 19 in 2009, 56 vs. 12 in 2010, and 64 vs. 27 in 2011).

Table 1.  Ouachita map turtle (Graptemys ouachitensis) nesting parameters at 2 sites (combined) on the lower Wisconsin River, Iowa County, Wisconsin. (See text for details.)

View Fullscreen

The overall hourly distribution of G. ouachitensis nest initiations was bimodal for both sites when all study years were combined, with a primary peak in mid- to late morning (median 1012 hrs Central Daylight Time [CDT]) and a smaller, secondary peak in mid-afternoon (median 1621 hrs CDT; Fig. 3). Bimodality resulted from the apparent avoidance of nesting activity during the midday (1000–1600 hrs) on relatively warm days. Nesting occurred throughout the day when maximum daily temperatures were in the lower half of the observed range (21°–37°C), but was limited during midday when temperatures were in the higher end of the range (Fig. 4). The degree of nesting bimodality varied annually, with proportions of midday nesting being negatively associated with the mean maximum temperature during the nesting period (for G. ouachitensis, 50% midday nesting at 25.9°C mean nesting-day maximum; n  =  8 in 2008; 73% at 25.1°C, n  =  22 in 2009; 38% at 27.0°C, n  =  21 in 2010; and 30% at 32.4°C, n  =  27 in 2011; χ²  =  6.36, p  =  0.01, n  =  78). Thus, the relatively warm nesting season of 2011 exhibited the greatest degree of bimodality (Fig. 4).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Skies were sunny or partly sunny (i.e., not overcast) during 69% of the midday nestings (n  =  36) for G. ouachitensis and 70% of all Graptemys (n  =  54) during all study years. Parts of both sites had areas that retained morning shade from nearby trees (until ca. 0945 hrs and 1200 hrs in mid-June at Site A and Site B, respectively).

The overall nesting duration mean (time to excavate nest cavity, deposit eggs, and cover site) for G. ouachitensis was 42.1 min (range 20–73 min, SD  =  12.7, n  =  72), although Graptemys nesting duration varied inversely with air temperature at the time of nest construction (Pearson's r  =  −0.372, n  =  97, p < 0.001; Fig. 5) as estimated from recalibrated camera stamp temperature data (see Methods).

View Figure

• Download as Powerpoint
• Download Figure

Influence of Precipitation on Nest Initiation

Available data provided no evidence that G. ouachitensis nesting activity was associated with whether rainfall (as a single factor) had occurred within the previous 24 hrs, as results were not significant for either individual or combined years (all-year 2-tailed Fisher's exact tests, p  =  0.651, n  =  76; Table 2). This pattern was also exhibited by the overall nesting Graptemys community as well as when transient (putatively nesting) turtles were included (all-year 2-tailed Fisher's exact tests, p  =  0.889, n  =  102, and p  =  1.000, n  =  207, respectively). These analyses excluded nests constructed within 3 d after flood recedence in 2008 and 2011, as females may have altered behaviors because of egg retention during flooding periods (see below). Most nests were constructed more than 24 hrs after previous rain (53.9%, n  =  102).

Table 2.  Effect of precipitation within previous 24 hrs on Ouachita map turtle (Graptemys ouachitensis) nest construction by year for both sites combined.

View Fullscreen

Only 2 G. ouachitensis nests were constructed during precipitation. In both instances, the females continued during light to heavy rainfall and completed nesting under relatively warm although cooling conditions (from ∼ 29.4°C to 23.9°C and 22.8°C to 20.0°C, respectively).

Impact of Flooding on Nesting Activity and Nest Survival

Flooding occurred at the nesting sites in 3 of the 4 study years: 8–18 June 2008, 17–27 July 2010, and 25–29 June 2011. Numerous recently destroyed nests (∼ 32) and several turtles in the process of nesting were present on and near the study areas at the first site visits 1 or 2 d after the 10-d flood in 2008. While the period immediately postflood was not monitored because of access difficulties, many nests were thus constructed during that time. Although smaller in scale, similar increases in turtle presence and nesting activity were also noted after the 4-d flood in 2011.

The long inundation periods (∼ 10 d) of both the 2008 and the 2010 floods are believed to have caused complete mortality of all nests constructed prior to flooding. The continuous 2008 flood caused the mortality of 0.75- and 2.5-d-old nests. The episodic 2010 flood resulted in approximately 6 d of water coverage (over 2 periods) at Site B and approximately 3.5 d of coverage (over 3 periods) at Site A. After the 2010 flood, hatchlings inside the eggs at both a 56-d-old nest at Site A covered by water for a total of 81 hrs (and in saturated sand for additional ∼ 12 hrs) and a 51-d-old nest covered by water for 14 hrs (and in saturated sand for ∼ 62 hrs) were found to be dead on inspection. In contrast, the shorter 2011 flood was not known to cause the loss of any nests (although most experienced only saturated sand), including 2 17-d-old nests at Site B continuously covered with water for 4 d and in saturated sand for approximately 12 additional hours.

Utilization of Sites and Nesting Season Chronologies

The G. ouachitensis nesting seasons in this study were within the mid-May to early July interval described by Vogt (1980) for G. ouachitensis and G. pseudogeographica nesting in Wisconsin and, similarly, had earlier onsets in years with higher spring temperatures. The correlation of nesting onset with increased environmental temperatures appears to be a widespread pattern for North American emydids (Aresco 2004), presumably having influential effects on vitellogenesis of follicles (Congdon et al. 1983, 1987) and shell deposition (Vogt 1980).

While bimodal diel nesting has been documented for several chelonians (e.g., for Chelydra serpentina [Hammer 1969; Congdon et al. 1987] and for Glyptemys insculpta [Walde et al. 2007]), it does not appear to have been previously reported for any Graptemys, although the reproductive biology of several species in this genus remains poorly known (Ernst and Lovich 2009). Variations in reported nesting times for a given species may reflect differences in nesting season temperatures during particular study time lines (Christens and Bider 1987), as demonstrated here by bimodal nesting patterns occurring only on relatively warm days and years because of apparent avoidance of high midday temperatures. Reduction of turtle nesting activity during both relatively cold (e.g., for Chelydra [Iverson et al. 1997]) and warm periods (e.g., for Chrysemys [Congdon and Gatten 1989] and for Malaclemys [Seigel 1979; Feinberg and Burke 2003]) has been noted previously (see also Jackson and Walker 1997).

Vogt (1980) reported that midday nesting by Graptemys on relatively open Mississippi River islands was common only under overcast conditions, which he attributed to depressed midday temperatures—a suggestion supported here by findings of midday nesting being restricted largely to relatively cool days. In contrast, however, the majority of midday nesting on these sites occurred under sunny or partly sunny sky conditions, possibly a result of cooler microclimates due to mid- to late-morning shade.

The duration of nesting events for G. ouachitensis in this study is generally similar to that reported for G. nigrinoda (mean ∼ 1 hr; Lahanas 1982) and G. flavimaculata (mean 30.5 min; n  =  10; Moore and Seigel 2006). While comparative data for other Graptemys are scarce, nesting durations appear to be generally shorter than those commonly reported for most North American emydids (e.g., mean 97 min for Chrysemys picta [Rowe et al. 2005]; mean 131 min for Glyptemys insculpta [Walde et al. 2007]). The short nesting durations of both Graptemys spp. and Malaclemys terrapin (e.g., mean 25 min; Feinberg and Burke 2003) may reflect the relative ease of nest construction in sandy substrates versus the wider substrate selection (some rocky) of many other emydids (P. Lahanas, pers. comm., December 2011). A negative association between nesting duration and temperature, presumably related to thermally affected metabolic rates, has been reported previously (e.g., Congdon et al. 1983; Christens and Bider 1987).

Influence of Precipitation on Nest Initiation

Rainfall within the 24 hrs previous to nesting did not appear to be associated with nest construction activity in this study. However, potential multifactor (e.g., temperature and precipitation) and less proximate associations (e.g., rainfall more than 24 hrs previous) were not tested. Variable responses to rainfall on the day of construction have been described among species and locations, with some reporting positive associations (e.g., Jackson and Walker 1997; Burke et al. 1998) and others finding no apparent relationship (e.g., Aresco 2004). Bowen and Janzen (2005) noted that factors other than rain on the day of nesting are implicated in construction timing, as many turtles nest on days without rainfall.

While behaviors promoting nesting before or during significant rainfall are presumably under positive selection pressure (e.g., Strickland et al. 2010) due to the potential to reduce predation rates via subsequently diminished nest location cues (e.g., Carr 1952; these sites, Geller 2012a), this is less probably the case for nesting after rainfall, which is less likely to increase nest survival (Bowen and Janzen 2005). Among other suggested benefits of nesting in association with precipitation are thermal stress reduction on nesting females (e.g., for Pseudemys suwanniensis; Jackson and Walker 1997) and facilitation of nest construction in compact substrates (Hammer 1969; Seabrook 1989; Doody et al. 2003). These factors are perhaps of limited influence on the present study areas, as their sand substrates do not require softening by precipitation to facilitate nest excavation (although some retained moisture is necessary to maintain cavity integrity; Vogt 1980; Doody et al. 2003) and turtle nesting activity appears to be of relatively short duration on thermally ameliorated sites.

Both of the nests constructed during rainfall in this study occurred under relatively warm temperatures. In contrast, Vogt (1980) noted cessation of Graptemys nesting activity during rainfall and suggested that falling temperatures were the cause. Such inhibitory effects likely result from the slowing of female metabolism and nesting activity due to reduced available thermal energy (e.g., Congdon et al. 1987; Jackson and Walker 1997) and represent further examples of the pervasive influence of heat availability on chelonian reproductive biology.

Impact of Flooding on Nesting Activity and Nest Survival

A pronounced effect of the 2008 nesting season flood—and, to a lesser extent, the shorter 2011 flood—was the notable pulse of nesting activity soon after flooding ceased. The elevated numbers of turtles and recently depredated nests observed at the first visits following flooding was unique. While some turtles had nested prior to flooding, others, including possible second nesters, apparently held their eggs until conditions improved. The ability to retain eggs in response to unfavorable conditions has been documented previously (reviewed in Buhlmann et al. 1995), although references specific to flooding influences appear scarce.

In this study 17-d-old Graptemys embryos survived 4.5 d of water and saturated sand nest coverage, while embryos ≤ 2.5 d old were killed by 10 d of inundation. This finding of decreased survival with longer periods of submergence parallels the experimental results of Plummer (1976) for 1- to 12-d-old Apalone mutica eggs and Kam (1994) for 19-d-old Pseudemys nelsoni eggs; both found limited or no survival beyond 6 d of submergence. The complete mortality of older (> 50 d) embryos covered with water and in saturated sand for 3–4 d presumably reflects their increased oxygen demands (Kam 1994) and indicates that short-duration floods later in reproductive seasons are likely to cause greater mortality of wild turtle nests, on average, than those occurring soon after nest initiation.

Overview of Camera Utility and Performance

Multiyear, whole-season monitoring using trail cameras (> 100,000 total camera-hours) yielded a broad record of observational data relevant to turtle nesting dynamics on these sites with minimal disturbance to subject animals. While comprehensive ecological studies of turtles will continue to require extensive on-site fieldwork (e.g., home range, morphological, and dietary studies), this methodology can likely be of value in producing general nesting and nest predation dynamics data (e.g., Geller 2012a) for many taxa, especially those that nest in easily camera-surveyed aggregations. Closer-placed cameras in modified arrays may similarly enable the collection of otherwise difficult-to-obtain data in studies with more specific focus (e.g., nestling emergence time lines and initial dispersal directions).

The use of cameras also has the potential to enhance turtle conservation efforts. For example, in some settings, the early evening review of daily camera data taken on sites used by diurnally nesting species could increase the efficiency and overall success of nest screening efforts by delineating nest locations, reducing the need for repeated on-site surveys and the associated disturbance risks to nesting turtles.

While some turtles have distinctive morphologies (e.g., Chelydra), continuing improvements in trail camera image quality (e.g., increased megapixels and signal-to-noise ratios) will facilitate identification of similar-appearing turtle species (e.g., Apalone spp. and Pseudemys spp.). Positioning cameras close to subject animals or likely points of passage (e.g., nesting area concentrations and travel corridors), as well as where vegetation is sparse, also enhances species identification and the study of specialized behaviors. Mounting cameras on structures in angled-down positions increases the ability to demarcate observed phenomena (e.g., locations of turtle nests).

Time-lapse images are necessary to capture turtle data, as turtles, because of ectothermy and generally slow movements, seldom activate camera PIR motion sensors. Shorter time-lapse intervals (1 min in this study) enhance the temporal and overall yield of collected data (including ability to identify turtles as to species) but entail increased camera battery and data card space use. Motion-triggered image series, especially from cameras that can produce video or videolike sequences (e.g., Reconyx RapidFire™ Near-Video™), can enhance the interpretation of turtle and nest predator movements and other behaviors by providing serial, contextually linked records of monitored events. Cameras producing IR emissions devoid of visible wavelengths may be preferable to those using wavelengths within the visible spectrum because of reduced chances for influences on predator presence and behavior. Such cameras are also less conspicuous targets for vandalism or theft, an important consideration for field studies in publicly accessible areas.