The nesting ecology of Apalone spinifera in large North American rivers is largely unknown despite the wide distribution of the species in these naturally dynamic ecosystems. We describe the nesting locations, timing, behavior, and habitat of A. spinifera in relation to natural and anthropogenic factors in the Missouri River. Nesting followed annual peak river stage, mostly occurred in the afternoon when air temperatures were 25°–30°C, and did not occur when human activity was nearby. Apalone spinifera nested in June in a year with average discharge (2012), but nested 20 d later in a year with a large flood event (2011). During the average discharge year, 90% of nests were found on islands, but similar proportions of nests were found on island and mainland habitats during the flood year because many islands were inundated. Nests were mostly in mixed-gravel substrates where vegetation cover was sparse or absent. Depredation occurred only after the emergence of hatchlings (∼ 60 d after nesting) and more often on nests on the mainland than on islands. Emergence rates were ∼ 1.5 times higher in the average year than the flood year, and emergence rates were higher in mixed-gravel nests than in pure-sand nests in 2011. In artificial nests, incubation temperatures averaged ∼ 4.3°C higher in mixed-gravel than in sand substrates, and freezing temperatures in winter penetrated to depths greater than the mean egg chamber depth (7.5 cm) for up to 3 wks. Therefore, incubation might be accelerated in mixed-gravel substrates. Accelerated incubation would enhance reproductive success because freezing temperatures preclude hatchlings from overwintering in nests in our study area. Mountain snowmelt-driven hydrology, coupled with the onset of freezing temperatures in autumn, might create a temporal “runoff-freeze squeeze” that limits the successful reproduction of A. spinifera in some years. However, high runoff also scoured vegetation from shorelines where A. spinifera nested in subsequent years. Natural variation in annual discharge might therefore be crucial to conservation of A. spinifera in large rivers.
Understanding the spatial and temporal aspects of nesting is critical for conservation of turtles because nesting success can dictate population dynamics and persistence (Moll and Moll 2004; Ernst and Lovich 2009; Kuchling 2012). Nesting—including terrestrial oviposition, incubation, and hatchling emergence—is an essential yet vulnerable period in the life cycle of aquatic turtles. Nesting increases predation risk for adult turtles, eggs and hatchlings might be destroyed by predators, and eggs are vulnerable to inundation by flood pulses (Moll and Moll 2004; Kuchling 2012).
The nesting ecology of riverine turtles evolved in response to natural discharge regimes, and flood-dependent habitats are essential for the reproduction of riverine turtles (Moll and Moll 2004; Kuchling 2012). However, river modifications alter hydrological cycles and geomorphology, change the spatiotemporal availability of habitats, prevent migrations, fragment populations, and convert lotic habitats to lentic habitats (Hesse et al. 1989; Moll and Moll 2004; Tracy-Smith et al. 2012). For example, dams and channelization have reduced the extent of sand bars in the Lower Missouri River by 98%, thereby reducing potential turtle nesting habitat (Funk and Robinson 1974). Such river modifications can decrease annual nesting success by limiting the spatiotemporal availability and number of areas available for nesting (Bodie 2001; Moll and Moll 2004; Tracy-Smith et al. 2012). River modifications influence turtle diets (Tucker et al. 2012), growth (Snover et al. 2015), demography (Melancon et al. 2013), spatial ecology and population genetics (Bennett et al. 2010), and species richness (Vandewalle and Christiansen 1996). Moreover, human activity can disturb nesting and influence nest success (Garber and Burger 1995; Bolton 2007). Therefore, determining the influences of natural and anthropogenic factors on the nesting ecology of riverine turtles is crucial to facilitate conservation efforts for these species.
Softshell turtles (Apalone spp.) have inhabited the Missouri and Mississippi river basins since the Cretaceous Period (Holman 1973; Parmley 1992). Spiny softshells (Apalone spinifera) are the most widespread Apalone in North America (Webb 1962, 1973) and are relatively common in most large rivers of the central and eastern United States. Accordingly, their life-history strategies have evolved in the context of local environmental conditions, natural variation in river discharge, and disturbance events such as flood pulses (Webb 1962; Plummer and Shirer 1975; Moll and Moll 2004). Although A. spinifera ecology has been studied in lakes (Galois et al. 2002; Bolton 2007) and creeks (Plummer and Mills 2008), scarce peer-reviewed information exists concerning the ecology of A. spinifera in large river-floodplain ecosystems (Anderson et al. 2002; Barko and Briggler 2006) and the natural and anthropogenic factors that might influence them therein. Even less is known about the nesting ecology of A. spinifera at the northern extent of their range (Tornabene et al. 2017). To our knowledge, only 2 peer-reviewed studies have investigated the nesting ecology of northern A. spinifera populations and both were on Lake Champlain in Quebec, Canada, and Vermont, USA (45°N; Graham and Graham 1997; Galois et al. 2002). No peer-reviewed studies exist concerning A. spinifera nesting ecology in a large river in their northern range. However, a long-term (24-yr) study of this species is ongoing along the Thames River in southwestern Ontario (S.D. Gillingwater, pers. comm., February 2018).
Seasonal temperature trends influence the timing of nesting and emergence, nest success, and growth rates in temperate-region turtles (Packard et al. 1987; Plummer et al. 1994; Kuchling 2012). Turtles in their northern range have a truncated active season and limited degree-days for egg development and incubation (Cagle 1950; Christiansen and Moll 1973; Compton 1999; Kuchling 2012) and consequently might have lower nesting success, annual growth, relative clutch mass, and delayed maturity (Bobyn and Brooks 1994). Although the hatchlings of some turtle species can overwinter in the nest (Ultsch 2006), freezing temperatures influence survival and quality of hatchlings of Apalone spp. and other turtle species (Packard and Packard 1988; Ultsch 1989; Costanzo et al. 1995). Apalone hatchlings typically emerge from nests from late summer through autumn (Ernst and Lovich 2009). Although Minton (2001) suggested that some A. spinifera hatchlings might overwinter in nests, their sensitivity to freezing temperatures (Ultsch 1989; Costanzo et al. 1995) suggests an inability to do so in their northern range. Therefore, emergence of hatchling A. spinifera probably must occur before the onset of freezing temperatures in autumn in large rivers at northern latitudes.
Apalone spinifera in the Missouri River in Montana might be vulnerable to declines or extirpation because the population is isolated from downstream populations by Fort Peck Dam, influenced by upstream dam regulation, and at the northern range limit of the species (47°N). Apalone spinifera appears to be absent in a range gap in the Missouri River basin from the confluence with the Musselshell River in Montana downstream to central North Dakota (Dood et al. 2009; LeClere et al. 2009). Apalone spinifera populations may have existed here prior to the construction of 2 large dams and reservoirs, but no records have been published to our knowledge.
Management and conservation of vulnerable populations require baseline natural history information and an understanding of the relationships between the life history of the species and anthropogenic and natural factors. Our goal was to provide insights on river management and the conservation of A. spinifera in large river ecosystems. Our objectives were to document A. spinifera nesting ecology and habitats, and determine how natural and anthropogenic factors affect them, in a large river ecosystem.
We conducted our study on the Missouri River above Fort Peck Reservoir in central Montana. The 101-river kilometer (rkm) reach extended from the Judith River confluence (rkm 3194 measured from the confluence with the Mississippi River) downstream to the Fred Robinson Bridge (rkm 3093; Fig. 1). We selected this reach because catch rates of A. spinifera were highest there during surveys of the Missouri and Yellowstone rivers from 2004 to 2008 (Dood et al. 2009). This reach is part of a designated Wild and Scenic River and is located within the Upper Missouri River Breaks and the Charles M. Russell National Wildlife Refuge. Channel bottoms and shorelines are dominated by gravel, cobble, and occasional boulders in the upper reaches of the study site, but transition to predominately sand near the Fred Robinson Bridge (Richards 2012).
We define our study system as having a quasi-natural discharge regime. Several hydroelectric dams affect the hydrology of the Missouri River in our study area, and 3 upstream dams (i.e., Canyon Ferry, Gibson Dam on the Sun River, and Tiber Dam on the Marias River) have the most influence on runoff and habitat (Ramey et al. 1993; Gardner and Jensen 2011). Downstream of these dams, peak flows in spring have been reduced by 20%–30% and base flows have been increased (Ramey et al. 1993). However, our study area is considerably less hydrologically altered than downstream Missouri River reaches that have been converted to lentic settings, have highly altered flows, are channelized, or are influenced by a combination thereof (Pegg et al. 2003; Galat et al. 2005; Sanford 2007). Annual discharge is typically characterized by 2 peaks: a smaller peak occurs in early spring resulting from prairie runoff, and a larger peak occurs in early summer resulting from mountain snowmelt. Serendipitously, discharge differed substantially between 2011 and 2012 and afforded us the opportunity to examine its effect on nesting phenology, behavior, and success. In 2011, a large flood event occurred following record snowfall in the Missouri River headwaters and near-record rainfall in eastern Montana. During 2011, mean daily, annual, and peak discharges were higher than the historical (1935–2012) discharge means at the US Geological Survey flow gauge at the Fred Robinson Bridge (gauge 06115200 near Landusky, Montana; Fig. 2). In 2012, discharge was similar to historical averages. Peak discharge in 2011 was ∼ 4 times higher than in 2012.
Nesting Timing and Behavior
We documented nesting phenology and behavior by closely monitoring nesting events throughout the study reach. We used 3 techniques to locate A. spinifera nests and observe nesting behavior: 1) observing potential nesting habitat from a distance with a spotting scope while hidden in a blind, 2) extensively surveying all potential nesting habitats on foot, and 3) placing remote cameras at known and potential nesting habitats (only in 2012). We conducted visual searches of all potential nesting habitats within the study site at least every other week from May to September in 2011 and 2012. We located nests during walking surveys by observing nesting females, following distinctive tracks of adult or hatchling A. spinifera, finding characteristic depressions in the substrate, finding eggshells and fragments, or a combination thereof (Bolton 2007). We carefully dug with bare hands to the top of the nest chamber at substrate depressions to confirm the presence of eggs while limiting disturbance. The first and last observations of nesting in each year defined the beginning and end of the nesting period.
We unobtrusively observed female A. spinifera behaviors when they were spotted on land near nesting locations. Turtles were observed from a boat or portable blind on the opposite shoreline (typically > 100 m away) with binoculars or a spotting scope. We documented behavior associated with nesting along with reactions to passing motorized and nonmotorized (i.e., kayaks, canoes, or rafts) boats. We also documented the behavior of nesting turtles in 2012 using 14 remote cameras (Reconyx PC800 Hyperfire Professional IR, Reconyx, Inc, or Moultrie Game Spy I-65, EBSCO Industries, Inc) on 9 potential or known nesting sites (Appendix 1). Remote cameras had wide views of entire nesting sites, adjacent shorelines, and river channels. Reconyx cameras took photographs every 5 min from about 16 June to 23 July 2012 at 7 locations. Two locations with larger, nonoverlapping nesting sites had more than one camera; one location had 3 and the other had 2 nonoverlapping Reconyx cameras. Paired Moultrie cameras took photographs every hour (a limitation imposed by this camera type), the camera fields of view were overlapped by ∼ 2 m on the edges, and exposures were timed to obtain a photograph at each site every 30 min (Fig. 3). We examined photographs with Timelapse Image Analysis software v. 188.8.131.52 (Bolton 2007; Greenberg and Godin 2012), which made entering and coding each photograph (n = 83,957) for time, date, temperature, cloud cover, other taxa presence and tracks, and number of turtles and boats in view more tractable. We counted the number of turtles nesting in each photograph and confirmed the presence of eggs in nests after reviewing photographs. In each photograph, we defined nesting as a turtle 100% out of water that was > 1 m from shore, facing the shore, and exhibiting nesting behaviors (i.e., nest searching, digging, or laying eggs). We also counted the number of boats in view and categorized them as motorized or nonmotorized.
We determined macrohabitat characteristics at nesting sites (≥ 1 nest in a distinct location) in 2011 and 2012. We classified habitat type as mainland (on the lateral channel margins of the river) or island (within the confines of the river). We documented seral stages of nesting islands only in 2012 and described them as bare, pioneer, or mature. Bare islands had no vegetation, pioneer islands had patches of willow (Salix spp.) and young plains cottonwoods (Populus deltoides subsp. monilifera), and mature islands had stands of mature plains cottonwoods. During peak discharge, bare islands were almost entirely inundated, pioneer islands were not entirely inundated, and mature islands were inundated only along their shorelines. We measured dimensions of potential nesting habitats on islands in 2012 in situ using a metric measuring wheel and calculated areas. We used aerial imagery from 2011 to calculate the total area of islands nested on in 2012 in ArcMap 10.1 (ESRI, Redlands, CA). We defined shoreline geomorphology types of mainland sites as alluvial, bluff, cut bank, or slope and geomorphology types of islands as alluvial, cut bank, or slope. Alluvial banks were sparsely vegetated, relatively flat banks where recent alluvial deposition had occurred. Bluffs occurred where steep erosion-resistant cliffs were adjacent to the river channel. Cut banks occurred where friable riverbanks were eroded and nearly vertical. Slopes were less steep, stable, and typically vegetated.
We determined several microhabitat characteristics at nesting sites in 2011 and 2012. To mark the exact location of each nest, and monitor subsequent depredation and emergence, we drove two 30-cm pieces of rebar 20 cm into the substrate ∼ 0.5 m from the nest to mark its exact location. We categorized substrate as sand or mixed gravel (sand, fine gravel, and coarse gravel). We measured distance from a nest site to shore (m) with a metric measuring wheel. We measured elevation above water surface (m) and riverbank slope using a clinometer and a 2-m-long × 2.5-cm-diameter polyvinyl chloride pipe with 5-cm increment measurements (herein, habitat pole). We measured riverbank slope (°) where turtle tracks indicated the land-entry point. If no discernable turtle tracks remained, we measured slope at the suspected land-entry point, as determined by accessibility and vegetation cover. We measured depth at shore (m) 0.5 m from the shoreline using a habitat pole or boat-mounted depth sounder (Eagle Cuda 350, NAVICO). We measured maximum vegetation height (cm) with a habitat pole and visually estimated percent vegetation cover within 1 m of each nest. We measured depth to the top of the egg chamber for each nest (egg chamber depth; cm) in 2012 using a metric ruler after careful digging to verify presence of eggs. We also documented the presence of potential predators and other animals (e.g., cattle, birds, coyotes [Canis latrans], red foxes [Vulpes vulpes], raccoons [Procyon lotor]), and humans, by identifying tracks and sign at nesting sites.
Incubation Period and Emergence Timing and Behavior
After initially locating nests, we conducted walking surveys every 1–2 wks to document emergence and depredation, calculate incubation periods and emergence rates, and monitor nest success. We documented emergence, and located previously undetected nests, by observing characteristic exit holes and tracks created when hatchlings emerged. We detected nest depredation on known and unknown nests during walking surveys by observing signs of depredation—excavated nests with scattered eggs or eggshells, often with mammal tracks or scat nearby. We tallied nest depredation separately for islands and mainland nests and calculated annual, site-level, and island or mainland percent nest depredation. We defined nesting aggregations in macrohabitats as ≥ 2 nests within 25 m of each other.
We also monitored depredation and nestling emergence timing and behavior using remote cameras (PC800 Hyperfire Professional IR, Reconyx, Inc) placed on 10 individual nests from 24 July (directly following the nesting period) through 3 October 2012. Each camera took a photograph every 5 min, was pointed nearly straight down on a single nest, and had a narrow field of view. We visually scanned each photograph (n = 198,297) and photographs containing substrate disturbance indicative of emergence or predation were individually entered into a database and coded for number of A. spinifera hatchlings or tracks, time elapsed since first emergence, and time elapsed since previous emergence for each of the 10 monitored nests. We also coded for time, date, and temperature. We documented emergence when the nest opened, when hatchlings were in view, when hatchling paths (observable by substrate shifts from consecutive images) appeared, or a combination thereof. We estimated the minimum number of emerging hatchlings by counting the unique paths of emerging hatchlings that we observed. We defined the emergence period as the time span between the first and last observations of emergence in each year. We calculated mean incubation period for each nest from two estimates to minimize error because we did not immediately detect nests when eggs were laid. We calculated the first estimate by subtracting the earliest possible nesting ordinal day number from the earliest possible emergence ordinal day number. We calculated the second estimate by subtracting the latest possible nesting day number from the latest possible emergence day number. We calculated emergence rate as the number of nests that had hatchlings successfully emerge, divided by the total number of nests minus the number depredated in each year.
Incubation and Overwintering Substrate Temperatures
We measured temperatures in artificial nests during the incubation period in 2012 because we observed that nest success differed between substrate types in 2011. We predicted that nest temperatures would be higher in nests with gravel substrates, which might have accelerated incubation and caused higher nest success. We used temperature loggers (HOBO TidbiT v2, Onset Computer) to compare temperatures among 3 depths and between artificial nests in sand and mixed gravel. We buried temperature loggers at depths of 5, 12.5, and 20 cm inside 3 artificial nests of each substrate type (9 in each substrate type and 18 loggers total) because median egg chamber depth in 2012 was 11.9 cm (range = 7.3–17.9 cm). Substrates in artificial nests of each substrate type had similar composition among replicates (> 90% sand in nests in sand and ∼ 60% sand and 40% gravel in nests with mixed gravel). Within each site (i.e., sand or mixed-gravel nest sites), nests were ∼ 2 m from each other. The site with mixed gravel was directly across the river from the site with sand (∼ 200 m apart). Artificial nests were also < 2 m from actual spiny softshell nests at both sites, and canopy cover at all sites was similar (∼ 50% cover measured at 1600 hrs at both sites). We tied temperature loggers in each nest to a 1-m-long piece of rebar with a 25-cm length of tieable metal fishing leader (Tyger Leader Sporting); the rebar was hammered 0.9 m into the substrate. We located artificial nests with a handheld Global Positioning System. We deployed temperature loggers on 16 June 2012 and retrieved them on 30 September 2012; loggers recorded temperature every 10 min. Four temperature loggers (1 at 5 cm, 2 at 12.5 cm, and 1 at 20 cm) at the mixed-gravel site failed on 18 June 2012 at 2020 hrs when the site was apparently struck by lightning (indicated by cameras at the site).
We investigated the potential for A. spinifera hatchlings to overwinter inside the nest, and emerge the next spring, by determining substrate temperatures inside artificial nests during the overwintering period. We deployed temperature loggers in the same locations and following the same design as above with 3 depths in 3 artificial nests of each of the 2 substrate types. We deployed temperature loggers on 9 November 2011 and retrieved them on 1 June 2012. We characterized frequency, duration, and minimum temperatures of potential freezing episodes to assess their influence on overwintering embryos (PFEs; Churchill and Storey 1992a, 1992b; Costanzo et al. 1995). Potential freezing episodes occurred when soil temperatures were < −0.55°C, the estimated point at which turtle tissues reach their equilibrium freezing point (Churchill and Storey 1992a, 1992b; Costanzo et al. 1995).
We examined normality of all variables using histograms, Q–Q plots, and Shapiro-Wilk tests. Homoscedasticity was examined using Levene, Brown-Forsythe (BF), or Bartlett's tests. We used nonparametric statistics for parameters not meeting normality or variance assumptions, despite efforts at data transformation. We compared microhabitat variables of nesting sites between years 2011 and 2012 to investigate annual variation in habitat use in relation to variable discharge using Mann-Whitney (MW) U-tests. We used Pearson's product-moment correlation coefficient to investigate the relationships between nest site areas and numbers of nests, and total island areas and numbers of nests. We examined the differences in incubation-period substrate temperatures between substrate types and among depths using generalized linear mixed models (GLMM) fit by restricted maximum likelihood. We implemented these models to nest parameters for depth within each individual artificial nest for each substrate type as a random effect per our experimental design, to account for spatial autocorrelation if it occurred, and because they are robust to missing data. We used a backward-elimination model-selection approach and compared models with Akaike's Information Criterion (AIC) using ΔAIC and Akaike weights (Kenneth et al. 2002). We investigated autocorrelation with autocorrelation and partial autocorrelation function plots, included autocorrelation structure into the model if necessary, and tested the goodness of fit of the autocorrelation model with a likelihood ratio test (Zuur et al. 2009). We compared frequency, duration, and minimum temperatures of PFEs among depths within each substrate type using Kruskal-Wallis (KW) 1-way analysis of variance tests and between substrates at each depth using MW tests. We completed multiple comparisons following significant KW tests with pairwise Wilcoxon rank-sum tests with Bonferroni's confidence interval adjustment. We conducted all analyses in Program R v3.0.2 (R Core Team 2013) with α = 0.05.
Nesting Timing and Behavior
Timing of nesting was later and shorter, and we located fewer nests, in 2011 than in 2012 determined by direct observations and remote cameras. Nesting occurred after peak discharge in both years (Fig. 2). In 2011, nesting began on 4 July when river discharge was ∼ 1050 m3/sec and ended on 23 July (duration, 17 d). We observed female turtles on inundated islands unsuccessfully attempting to nest, as soon as mid-June (before river stage receded) in 2011. In 2012, nesting began on 14 June (20 d earlier than in 2011) when discharge was ∼ 455 m3/sec and ended on 7 July (duration, 24 d). We observed that A. spinifera in our system did not nest when river discharges were greater than ∼ 1100 m3/sec. We located 25 nests in 2011 and 97 in 2012 (Appendix 1). We do not believe that nest detection differed between years because all mainstem, side channel, and island shorelines were extensively surveyed for nests in both years by multiple observers.
We documented 1 complete nesting process (from start to finish) in 2011 and 7 in 2012 at 6 different locations by direct observation (n = 5 observations) or with remote cameras (n = 3; Fig. 3). The nesting process consisted of 4 major steps: 1) the female searched the shoreline, probing the substrate with her tubular snout and engaging in superficial digging with her front legs at several locations (i.e., test digs); 2) digging actively with front legs; 3) repositioning, then digging with back legs; and 4) oviposition of eggs, burial with alternating strokes of the back legs, and smoothing over the mound of substrate created by nesting with her plastron. Mean nesting duration was 66 min and ranged from 39 to 127 min. Mean number of test digs per female was 1 and ranged from 0 to 3 test digs. We documented 10 incomplete nesting events; 6 events stopped after nest site searching (Step 1) and 4 events stopped after digging with front legs (Step 2).
From our remote camera data, we observed that nesting occurred during warm and sunny days and did not occur simultaneously with human activity in photographs. Remote cameras captured 482 instances of A. spinifera nesting behavior and 208 instances of human activity during the nesting period in 2012. Nesting activity in 2012 began on 16 June and peaked on 19 June. Diel timing of nesting was roughly bimodal, began at 0800 hrs, peaked at 1000 hrs and 1700 hrs, and ended at 2000 hrs. Nesting activity generally occurred when temperatures were 25°–30°C (54% of nesting photographs), when no precipitation was falling (99%), when cloud cover was 0%–25% (43%), when no other animals were present on shore (90%), and when no humans were in view (100%). Additionally, during the nesting period, we only observed turtles basking while humans were in view in 3 images (< 1%). Human activity largely involved nonmotorized boats (76% of human activity photographs), but motorized boats were also observed (24%).
Nest Site Aggregation, Fidelity, and Macrohabitat Characteristics
Apalone spinifera nested communally with some interannual fidelity to nesting sites. About one-third of the nests we located were in aggregations in both 2011 (32%) and 2012 (28%). We located more than one nest at 62% of nesting sites in 2011 and 87% of sites in 2012. The average number of nests located per site was 3 in 2011 (range = 1–12) and 5 in 2012 (range = 1–15).
Turtles nested on islands more commonly in 2012 than 2011, and the number of nests on islands was related to area of potential nesting sites in 2012. Only about half (44%) of nests were on islands in 2011, but almost all nests were on islands in 2012 (90%; Appendix 1). In 2012, A. spinifera nested most often on pioneer islands (70%), but also nested on mature (29%) and bare (1%) islands. Size and shape of islands and island vegetation cover were visibly altered after the 2011 flood event (Fig. 4). Numbers of nests at nesting sites were highly and positively correlated with areas of potential nesting sites in 2012 (r = 0.87, p < 0.001), but were not correlated with total island areas (r = 0.19, p = 0.54). Moreover, potential nesting areas were not correlated with total island areas (r = 0.02, p = 0.96). Although A. spinifera commonly nested on alluvial shorelines in both 2011 and 2012 (Table 1), they nested at the base of sloped (steeper) shorelines more often in 2011 than in 2012.
Microhabitat Characteristics at Nesting Sites
Nesting habitats were higher, had steeper slopes, and had deeper proximal water in 2011 than in 2012 (p < 0.01; Fig. 5 and Table 2). Median elevation (0.8 m), depth at shore (0.1 m), and slope (10°) in 2011 were 1.6, 1.7, and 10 times greater than in 2012 (0.5 m, 0.06 m, and 1°). However, median distance to shore (distance), percent vegetation cover, and maximum vegetation height at nests were not different between 2011 and 2012 (p > 0.34). Substrates at nesting sites were similar between years; only 2 nesting sites were predominately sand, whereas 16 sites were predominately mixed gravel overall (Appendix 1). Most A. spinifera nests were in mixed-gravel substrates in both 2011 and 2012 (84% and 93%), but some also nested in sand (16% and 7%).
Hatchling Emergence, Nest Success, and Nest Depredation
Emergence timing and emergence rates differed between 2011 and 2012, but incubation periods were similar in duration. Emergence occurred between 4 and 27 September (24 d) in 2011 and from 11 August to 6 September (27 d) in 2012; emergence began 24 d later in 2011 than in 2012. Emergence rates were 1.6 times higher in 2012 (95%) than in 2011 (60%). Emergence rates were nearly 3 times higher in mixed-gravel (73%) than in sand (25%) substrates in 2011. However, emergence rates were similarly high in both mixed-gravel (96%) and sand (88%) substrates in 2012. Mean incubation periods were 5 d longer in 2011 (63 d) than in 2012 (58 d) and minimum incubation period was 8 d longer in 2011 (44 d) than in 2012 (36 d).
We observed 64 hatchling tracks and 10 hatchlings emerging from 9 of 10 nests monitored with remote cameras in 2012 (Appendix 2); no emergence occurred at 1 nest. Hatchlings emerged from 14 to 31 August 2012 at 1900–2100 hrs (65% of hatchlings) and 0200 hrs (15%), mostly when temperatures were 25°–30°C (53%). Hatchlings usually emerged alone or in small groups. Mean group size was 2 (range = 1–6). They emerged en masse on a single occasion. A mean of 8 turtles emerged from each nest (range = 1–17 hatchlings, tracks, or both). Mean time between observed (photos taken at 5-min intervals) successive emergence events from the same nest was 52 min (range = 0–860 min), but mean time from observed first to last emergence events from the same nest was 227 min (range = 0–875 min).
Nest depredation occurred during or after emergence of hatchlings in both years and occurred more often at mainland sites than at island sites. We observed nest depredation on nests from 4 to 17 September in 2011 and 14 August through 13 September in 2012. We located 34 previously undetected nests by observing signs of depredation in 2011 and 2012. We observed depredation on known nests next to known nonemerged nests. Depredation of previously identified nests occurred a mean of 62 d (range = 55–72 d) after nesting in 2011 and 60 d (range = 48–61 d) after nesting in 2012. Coyotes were the primary nest predators as judged by scat and tracks near depredated nests, but red foxes were also common within our study site. Percentage of depredated nests was similar in 2011 (40%) and 2012 (37%). However, percentage of depredated nests at mainland sites (57% in 2011 and 90% in 2012) was ∼ 3 times higher than at island sites in both years (18% in 2011 and 31% in 2012). In 2011, 25% of nesting sites were depredated and 67% of nesting sites were depredated in 2012. Mean nest depredation at the site level was 32% (0%–100%).
Incubation and Overwintering Nest Substrate Temperatures
Temperatures were higher in artificial A. spinifera nests in mixed gravel than in sand during the incubation period. Model comparisons (against the null model, and additive and interactive models) suggested that substrate type was the most influential explanatory variable of incubation-period nest temperatures (cumulative weight = 0.81; Appendix 3). Temporal autocorrelation occurred, and we accounted for this using an autoregressive (AR) correlation structure (φ = 0.93, L = 34326, p < 0.001). Mean temperatures in mixed-gravel nests were 4.3°C (95% CI = 3.3°–5.4°C) higher than in sand nests (GLMM; t4 = −7.83, p = 0.001; Fig. 6), and temperatures were more variable in sand than in mixed gravel (BF; F = 8.25, p < 0.01).
Numerous potential freezing episodes (PFEs) occurred in both nesting substrates, reached temperatures down to −20°C, and lasted for up to 24 d during winter 2012–2013. Potential freezing episodes occurred in both substrate types (mixed gravel and sand) and at all depths (5, 12.5, and 20 cm; Appendix 4). Frequency, duration, and minimum temperature of PFEs during the overwintering period differed among depths within each substrate type, and differed between substrates (p ≤ 0.05; Fig. 7). More PFEs occurred in mixed-gravel than in sand substrate at each depth. However, PFEs were most common at the 5-cm depth in both substrate types. Median durations of PFEs only differed between substrates at 12.5 cm where duration was 3 times longer in sand. The longest PFE occurred at 12.5 cm in sand and lasted for nearly 24 d. The coldest PFE was at 5 cm in mixed gravel and reached −20.1°C.
Nesting behavior and habitat use of A. spinifera in our population were similar to those of populations throughout North America (see Ernst and Lovich 2009). Apalone spinifera in the Missouri River nested around June, selected habitats to increase nesting success, were susceptible to nest depredation, and were sensitive to natural and anthropogenic disturbances. However, annual differences in timing, magnitude, and duration of peak discharges regulated the nesting of A. spinifera by influencing spatiotemporal availability of A. spinifera nesting habitats in the quasi-natural, large, Great Plains river we studied. Differences in discharges also influenced the timing and duration of nesting and incubation periods that regulated emergence rates and nest success.
We observed that peak discharge, followed by declines in discharge and exposure of shoreline nesting habitats, influenced movements of A. spinifera to nesting habitats and the initiation of nesting behaviors. Female A. spinifera moved toward nesting sites in complex habitats with multiple islands, side channels, and exposed shorelines before and during peak discharge (Tornabene et al. 2017). In both years, nesting began after peak discharge as decreasing river stage exposed nesting habitats, and ended soon thereafter, as described for other riverine turtles (Moll and Moll 2004; Ernst and Lovich 2009; Kuchling 2012). Although some species may lay more than one clutch of eggs each year (Kuchling 2012), we did not see evidence of multiple clutches or nesting activity prior to the decline in discharge (from peak) in either year. We observed fewer available nesting habitats, that most islands were inundated with water, and that many nests were found aggregated at available sites during the nesting season in 2011 when discharge was higher and longer than average. Therefore, a discharge threshold that limits the availability of exposed nesting habitats might determine when and where A. spinifera can nest each year.
Apalone spinifera nested primarily in mixed-gravel substrates in our study site, converse to most observations of A. spinifera and other riverine turtles that nest in sand (e.g., Doody 1995; Graham and Graham 1997; Bodie and Semlitsch 2000; Moll and Moll 2004; Moler 2006). However, A. spinifera in rivers in Ontario also nested in gravel and sand mixtures, despite the availability of pure-sand substrates (S.D. Gillingwater, pers. comm., February 2018), which suggests that nesting in mixed-gravel substrates may be common at the northern extent of the range of this species. Temperatures in mixed gravel were higher and less variable than in sand, which maximized degree days and reduced thermal variation. Variability of incubation temperatures can negatively influence physiology, morphology, and behavior of hatchlings (Miller et al. 1987; Packard et al. 1987; Packard and Packard 1988; Janzen 1993). However, mixed-gravel substrates were also much more common than sand in our study area. Moreover, although we are confident in our ability to detect nests, detection of nests could have been lower in sand than mixed-gravel substrates. Wind, rain, and other disturbances (e.g., human and animal) can quickly obscure A. spinifera tracks, test digs, and hatchling emergence holes in fine substrates.
Depredation of nests occurred only when hatchlings began to emerge from nests, and coyotes were the primary nest predator. Contrary to many observations of turtle nest depredation occurring immediately after nesting (Ernst and Lovich 2009; Kuchling 2012), depredation of nests did not begin until after the emergence of hatchlings. We observed depredation on known emerged nests adjacent to known nonemerged nests, which suggests that the location and timing of depredation events in our system are related to visual (e.g., soil disturbance and scampering hatchlings) or olfactory (e.g., turtle urine or mucus) cues produced by emerging hatchlings (Hamilton et al. 2002; Burke et al. 2005; Strickland 2008). The lower island-specific depredation rates, and later depredation, we observed may have been because coyotes had difficulty accessing (i.e., swimming to) some islands during high discharge. Even so, coyotes were observed swimming across the river to islands, locating about half of the nesting sites, and sometimes depredating all nests at a site in both years.
Hatchling emergence occurred ∼ 2 mo after nesting, as reported for other A. spinifera populations (Doody 1995; Ernst and Lovich 2009), when river discharge declined to baseflow and air and water temperatures began to decrease. Hatchling A. spinifera emerged from a single nest over minutes, hours, or days, as in Apalone mutica (Plummer 2007). Most A. spinifera hatchlings emerged at night. Emergence of some turtle species at dusk and night is initiated by decreasing diel temperatures or a temperature threshold following the onset of darkness after eggs have been pipped (Bustard 1967; Mrosovsky 1968; Hays et al. 1992; Gyuris 1993; Plummer 2007); these two temperature cues are probably interrelated (Moran et al. 1999; Doody et al. 2001). Most hatchling A. spinifera navigated to and entered water immediately after emergence, as do some other freshwater turtles (Anderson 1958; Burke et al. 2000).
Mountain snowmelt–driven hydrology, coupled with the onset of freezing temperatures in autumn, may create an occasional temporal “runoff-freeze squeeze” that limits timing of nesting and nest success. During the flood year, nesting habitats were inundated for most of the nesting season and nesting was delayed by nearly 3 wks. We also observed lower nest success in the flood year when the onset of freezing temperatures probably influenced nest temperatures. Hatchlings of this species are not tolerant of freezing temperatures in the nest (Costanzo et al. 1995), and successful overwintering of A. spinifera hatchlings has not been observed in a long-term Ontario study (S.D. Gillingwater, pers. comm., February 2018). In November 2010, we discovered a nest in sand that contained dead A. spinifera embryos, which demonstrated that hatchlings can incur mortality if they do not emerge before winter. The absence of A. spinifera in the Missouri River basin from near Mandan, North Dakota, upstream to above Fort Peck reservoir (Dood et al. 2009; Ernst and Lovich 2009; LeClere et al. 2009), where sand substrates dominate (Bramblett and White 2001), may be related to the runoff-freeze squeeze. However, more research is needed to understand the relationships between hydrology, substrate, and nesting success and the distribution of A. spinifera in large, northern rivers.
Human activity can affect timing and location of nesting of turtles, including A. spinifera (Garber and Burger 1995; Bolton 2007). Spiny softshells were extremely wary on land; we observed that A. spinifera abandoned nesting efforts when they became aware of human presence. Human activities can delay nesting, cause females to move to habitats concealed from human activity, decrease nest success, or all the aforementioned (Garber and Burger 1995; Bolton 2007). Spiny softshells on Lake Erie, Ontario, did not return to partially completed nests following disturbance, which often resulted in nest failure and led to restrictions on human activity near critical nesting sites (Bolton 2007).
Apalone spinifera populations in the Upper Missouri River are probably especially sensitive to disturbances (e.g., the runoff-winter squeeze and human recreation) because they inhabit a dynamic ecosystem, are disjunct from downstream populations, and are at their northwestern range limit. Although variation in annual discharge regulated the timing of nesting and nest success, flooding also scoured vegetation from shorelines where A. spinifera nested in subsequent years. Conserving natural river discharge patterns that maintain critical nesting habitats, while also limiting influences of human activity, might help facilitate the persistence of this sensitive population of A. spinifera and other riverine turtle populations throughout North America.
Funding for this project was provided by NorthWestern Energy, the US Fish and Wildlife Service (US FWS), the US Bureau of Reclamation (BOR), and Montana Fish, Wildlife & Parks (MTFWP). This study was conducted under the auspices of Montana State University Institutional Animal Care and Use protocols #2009-19 and #2012-24, MTFWP scientific collection permits #2009-056 and #2010-050, and US FWS special use permit #10-012. We thank J. Dullum from US FWS, R. Bolton from the University of Guelph, J. Chaffin and J. Peters from the US Bureau of Land Management, L. Hanauska-Brown from MTFWP, and D. Trimpe and J. Kucera from BOR for funding and logistical support. We thank M. Plummer, J. Rotella, and E. Kenison for comments on this manuscript. We also thank A. Dood, C. Jensen, C. Manning, M. Mayfield, D. Moore, R. Richards, T.D. Ritter, and Z. Wenderott for invaluable logistical support and help with data collection or analysis. We also thank the Lewistown, Montana, US FWS and MTFWP employees for shuttling vehicles and other logistical support; specific thanks to R. Potts and J. Fox. The Montana Cooperative Fishery Research Unit is jointly sponsored by the US Geological Survey, MTFWP, Montana State University, and US FWS. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government.
This is an open access article published under the Creative Commons CC-BY-NC-SA license (https://creativecommons.org/licenses/by-nc-sa/4.0/), which means the article may be reused with proper attribution for non-commercial use. Any remix or transformed version of the content must be distributed under the same license as the original.
Handling Editor: Joshua R. Ennen