Methods: Study Sites

We conducted the study at 5 sites in 3 river drainages in northeastern Oklahoma (Fig. 1). The easternmost site was located on the Spring River upstream of its confluence with the Neosho River. The headwaters of the Spring River originate in the Ozark Highlands ecoregion (Woods et al. 2005) and are characterized by high bluffs, high water clarity, and cobble and gravel substrate. Two of the study sites were located in the Verdigris River watershed. One consisted of 6.3 km of the main channel located 19.0 km above Oologah Reservoir, and the second consisted of 4.8 km of Big Creek, a tributary of the Verdigris. Both flow through the Central Irregular Plains ecoregion (Woods et al. 2005) and, in comparison to the Spring River, are turbid, slow flowing, and have substrate ranging from bedrock to fine sediment. Finally, we sampled the Caney River and a tributary named Pond Creek, located approximately 50 km west of the Verdigris in the Cross Timbers ecoregion (Woods et al. 2005). Flow is nearly stagnant for most of the year in both the Caney and Pond Creek because the dam forming Hulah Reservoir stabilizes water levels. This dam has also increased sedimentation, resulting in a predominantly muddy substrate. All of the study sites have narrow riparian buffers and are predominantly surrounded by agricultural land used for cattle and hay production. All of the stretches of rivers sampled were within 0.3–8.9 km of public boat ramps.

View Figure

• Download as Powerpoint
• Download Figure

Data Collection

We conducted turtle trapping surveys from 15 May until 11 August in 2011 and 14 May until 7 August in 2012. Each site was surveyed during 4 intervals each year on an approximately weekly rotation. Deviations from our schedule resulted from occasional inclement weather, as well as efforts to avoid interfering with fishermen during holidays in May and July each year. Each afternoon, we set 6–14 (x̄ = 9.4) 0.9-m single-throat hoop nets with 2.5-cm mesh, orienting the traps in the stream channel so that the funnel of the trap was completely submerged but enough of the trap remained above water that trapped turtles could access the surface. When available, we baited traps with fresh fish caught with a gill net or incidentally captured in funnel traps. When no fresh fish were available, we baited traps with canned sardines.

We used a random-number table to select daily trap locations from a pool of 100 locations per site that we had previously identified as favorable for securing a trap. We set traps between 1300 and 1800 hrs and checked traps the following morning. Emydid and kinosternid turtles were assigned unique scute notch codes using a numbering method adapted from Cagle (1939). We used a cordless rotary tool to create V-shaped notches. We injected passive integrated transponder tags into chelydrid and trionychid turtles to facilitate future identification. We examined all turtles for evidence of traumatic injury, including damaged or missing limbs, missing marginal scutes, scarring to the carapace or plastron, and disfigurement to the head. Some injuries were the obvious result of propeller strikes, consisting of several long clean scars evenly spaced along the carapace, whereas others were of ambiguous origin. We photographed all detected examples of traumatic injury for future reference.

To quantify boat traffic, we recorded each boat equipped with an outboard motor that was observed while setting and pulling traps in 2012. We made these observations during roughly the same period each day. We then divided the number of boats by the number of hours spent at each study site to generate a boat traffic index.

Data Analysis

We used linear regression to correlate total combined injury rates for all species with boat traffic intensity at each site. We divided the carapace into 4 quadrants and performed χ² tests of independence to analyze the distribution of shell damage between the anterior and posterior halves and between the left and right halves. We performed the same tests to analyze limb damage. Finally, we used a χ² test of independence to compare injury rates among the turtle species. We separated turtles by species and compared injury rate within each species to the expected injury rate for that species based on its representation within the overall turtle community.

Results: Injury Locations

Shell injuries were unequally distributed among quadrants of the body (χ²₃ = 9.18, p = 0.027; Table 1). Turtles exhibited scarring on the left and right sides of the body at nearly equal frequencies (left = 134, right = 132; χ²₁ = 0.015, p = 0.90). However, turtles were 1.4 times more likely to exhibit scarring on the posterior half than the anterior half (anterior = 112, posterior = 154; χ²₁ = 6.63, p = 0.010).

Table 1. Anatomical distribution of injuries to freshwater turtles (n = 3284). Values indicate the number of turtles with injuries in a given quadrant of the body and (proportion of injured turtles exhibiting damage to that region).

View Fullscreen

The distribution of limb injuries mirrored observations of shell damage. Captured turtles were equally likely to exhibit damage to left versus right limbs (left = 21, right = 21), but were 2.4 times more likely to have damaged or lost a hind limb than a forelimb (anterior = 12, posterior = 29; χ²₁ = 7.05, p = 0.008).

Effects of Boat Traffic

The average boat traffic rate calculated over the summer 2012 trapping season varied among sites. The Spring River exhibited the highest boat traffic rate (1.18 boats/hr). The Verdigris River and its tributary Big Creek had moderate rates, with 0.61 boats/hr and 0.48 boats/hr, respectively. The Caney River had a much lower rate of 0.10 boats/hr, and its tributary Pond Creek had the lowest rate (0.03 boats/hr).

There was a significant correlation between a site's boat traffic and the rates of traumatic injury among resident turtles (r² = 0.99, p = 0.003; Fig. 2). The average injury rate among the turtles of all species at each site mirrored the boat traffic rates, with a rate of 19% on the Spring River, 15% on the Verdigris River, 12% on Big Creek, 6% on the Caney River, and 6% on Pond Creek.

View Figure

• Download as Powerpoint
• Download Figure

Injury Patterns Among Species

In order of abundance, the 9 species that were captured included slider turtles (Trachemys scripta), Ouachita map turtles (Graptemys ouachitensis), spiny softshell turtles (Apalone spinifera), false map turtles (Graptemys pseudogeographica), common snapping turtles (Chelydra serpentina), alligator snapping turtles (Macrochelys temminckii), river cooters (Pseudemys concinna), common musk turtles (Sternotherus odoratus), and smooth softshell turtles (Apalone mutica). Not all species occurred at all sites, but species richness was consistently high, ranging from 5 to 8 species/site.

Injuries were unevenly distributed among species, both in absolute numbers and in proportion to each species' representation in the turtle communities (χ²₈ = 30.99, p = 0.0001; Fig. 3). Because we captured several species in low numbers, the expected injury rates were less than 5 for P. concinna, S. odoratus, M. temminckii, and A. mutica. Therefore, we also compared the observed injury rates to expected rates among just the 5 species that met the minimum expected value assumption inherent in χ² tests (χ²₄ = 24.30, p < 0.0001). Among all captures (n = 3284), the percentage of individuals that bore evidence of traumatic injury was 9%. We observed low injury rates in 4 species: A. spinifera (3%; n = 325), A. mutica (0%; n = 5), S. odoratus (0%; n = 6), and M. temminckii (0%; n = 49). Four species had intermediate injury rates: T. scripta (9%; n = 2263), P. concinna (7%; n = 14), G. pseudogeographica (7%; n = 92), and C. serpentina (7%; n = 54). Finally, G. ouachitensis, which was the second most commonly caught species, had an injury rate considerably higher than the community average (13%; n = 476).

View Figure

• Download as Powerpoint
• Download Figure

Discussion

Our results demonstrate a clear correlation between levels of boat traffic and the proportion of injured turtles. Importantly, however, these proportions reflect only those individuals that were not killed in an encounter with a boat. It is possible that the total impact of boats is considerably higher than what we recorded.

The average overall injury rate we observed (9%) was consistent with rates reported in earlier studies, although the methods used varied. Northern map turtles (Graptemys geographica) were injured at rates of 8% and 4% in the St. Lawrence River and Lake Opinicon (Ontario, Canada), respectively (Bulté et al. 2009). Diamondback terrapins (M. terrapin) have been reported to have an injury rate of 11% in the Kiawah River in South Carolina; 8% of the injuries consisted of limb loss and only 3% of individuals had shell damage (Cecala et al. 2008). Significantly higher injury rates have been found in other studies of M. terrapin, including 16% in the Everglades (Hart and McIvor 2008), 20% on the Patuxent River in Maryland (Butler et al. 2006), and 12%–17% on Long Beach Island, New Jersey (Burger and Garber 1995). Higher boat injury ranging from 17% to 20% were also found in G. geographica on the Trent-Severn Waterway in Ontario, Canada (Bennett and Litzgus 2014). These higher rates are consistent with the injury rates at the most heavily trafficked site and the lower rates are consistent with the overall average rate that was observed. Although many of the injuries were consistent with propeller strikes, many others were ambiguous. We instead grouped together all nonsuperficial injuries together, with the caveat that not all of the injuries observed were necessarily the result of boat collisions. Pond Creek, with its extraordinarily low boat traffic rate, may be representative of the general level of injury (6%) that can be expected in waterways that lack boat traffic but are still subject to a suite of other peripheral human activities.

The differential injury rates that we observed among species may be partially explained by differences in habitat use. Species that typically move throughout their habitat by walking along the bottom, such as M. temminckii and S. odoratus (Berry and Shine 1980), would presumably be less likely to encounter boats than more vagile species that occupy more of the water column and make frequent use of the center of the river channel. The less pelagic species would be expected to be injured less often, although if frequent movement throughout the water column is the primary contributor to injury risk, C. serpentina might be expected to exhibit similarly low injury rates. Other species, such as A. spinifera and A. mutica, might be more likely to escape injury as a result of greater swimming speed. In addition, some of the injury rates were likely poorly representative due to small sample sizes (A. mutica, P. concinna, and S. odoratus). Other studies with larger sample sizes have found low boat-related injury rates of 4% (Bancroft et al. 1983) and 2% (Bennett and Litzgus 2014) in S. odoratus whereas injury rates observed in P. concinna have included 10% in a cache of illegally harvested turtles discovered near Cedar Key, Florida (Heinrich et al. 2012), 15% on Rainbow Run in Marion County, Florida, 9% at Manatee Springs State Park, and 8% at Fanning Springs State Park, both in Florida, suggesting P. concinna typically has an injury rate consistent with the average rates found in other emydids (Heinrich et al. 2012).

The most puzzling departure from the average is the high injury rate found in G. ouachitensis, which presumably exhibits similar behavior to the other emydid species in this study. It is possible that G. ouachitensis is, for some reason, more likely to survive a collision with a boat. It is also possible, however, that the species is more susceptible to harm from increased boat traffic. This appears to also have been the case in painted turtles (Chrysemys picta), which in a longitudinal study showed an increase in injury over time (along with a concomitant increase in recreational boating), that was not exhibited in G. geographica and T. scripta (Smith et al. 2006). It is possible that some species are especially prone to injury from boats due to behavioral patterns such as a higher percentage of time spent basking at the water surface. If this is the case, a thorough understanding of the use of habitat by various species could be very instructive in determining what conservation measures would be most effective at minimizing boat–turtle interactions. At-risk species known to be particularly susceptible to boat collisions, for instance, could benefit from speed limit restrictions.

Regardless of species, a majority of injuries were to the posterior carapace and limbs. This differs from an earlier study of Malaclemys terrapin that found limbs to be missing at roughly equal rates (Cecala et al. 2008). The concentration of injuries to the posterior half of the carapace and the hind limbs could be indicative of 2 things. First, turtles might simply be more likely to survive an injury to the posterior half of the body because an injury to the anterior half is more likely to include a head injury. Turtles exhibiting limb loss have been shown to have significantly reduced survivorship (Harding 1985), but little is known about whether the specific limb lost has any importance on survivorship. The loss of one or both forelimbs, which are often used in feeding, may be more detrimental than the loss of a hind limb. Another possibility is that a large proportion of turtles that sustain injuries are attempting to dive when the injuries are incurred, suggesting that the turtles are attempting to escape the oncoming boat but—in stereotypical turtle fashion—are simply too slow to do so. This has been shown to be the case for green turtles (Chelonia mydas), which successfully avoided contact with slow boats (4 km/hr) 15 times more successfully than when they fled from fast boats (19 km/hr; Hazel et al. 2007). If this is the case in freshwater turtles as well, the implementation of speed restrictions in areas with at-risk turtle communities would help to alleviate some of the negative impacts of recreational boating on those communities.

Although our study provides useful information about the effects of high boat traffic levels on an entire turtle community and indicates differences in impacts among species, it only provides an indirect measure of the full effect that boat traffic may have. Additionally, beyond the very direct effects of boat collisions, the increased physiological stress induced by high levels of boating activity can cause dramatic differences in overall shell condition due to increased bacterial and fungal infection rates (Selman et al. 2013). As studies like ours continue to demonstrate the harm that boat usage has on aquatic turtle communities, the need for conservation measures such as speed limits and protected areas becomes increasingly apparent.