Study Site

The study was conducted from 15 May to 13 October 2007 on the Santa Fe River (Alachua and Columbia counties, north-central Florida) in a 3.4-km stretch downstream from the US Highway 441 bridge (hereafter US 441; Fig. 2). We also collected data 4 km upstream from the US 441 bridge to the point where the river reemerges as a first-order spring at the “River Rise.” Most of the study site and the surrounding area is managed by the River Rise Preserve State Park and comprises a part of the largest publicly protected area along the Santa Fe River.

View Figure

• Download as Powerpoint
• Download Figure

Water depth varied throughout the study site from less than 0.2 m to more than 6.5 m but did not fluctuate substantially during our study. The current was slow and consistent throughout the study site except in a few riffles 450 m downstream from the Santa Fe Canoe Outpost, our point of entry onto the river (Fig. 2). The river width was ∼ 40 m throughout most of the river but varied locally from 25 to 85 m. Drought during the previous year had lowered the river level substantially, resulting in clear water with good visibility in contrast to the normally tannin-stained water. Lowered water levels (to 20-cm depth) 0.3 km and 3.6 km downstream from US 441 resulted in decreased recreational canoe and kayak traffic. Detailed information on the study site has been presented in Florida Department of Environmental Protection (2003) and Kornilev (2008).

Capture and Marking

We caught 50 Suwannee cooters (group A: 23 ♂, 25 ♀, 2 medium-sized juveniles) during 18–23 May 2007 and attached temperature-sensitive data loggers (iButtons; 9 × 11 mm, 3 g) on the right posterior marginal scutes. While attempting to retrieve iButtons between 23 July and 8 October 2007, we captured an additional 39 turtles (group B: 12 ♂, 20 ♀, 7 medium-sized juveniles). Group A turtles were initially caught in a ∼ 1-km stretch of the river starting 800 m downstream from US 441 (Fig. 2). Subsequent captures and recaptures occurred primarily in the first 2.5 km immediately downstream from US 441.

Turtles were captured by active pursuit while snorkeling or opportunistically by hand (Marchand 1942, 1945a, 1945b; Kramer 1995; Huestis and Meylan 2004), and standard morphometric data were taken in the field (straight-line carapace length [CL], plastron length, body mass, sex, life stage). Each individual was permanently marked by drilling marginal scutes (modified from Cagle 1939) and temporarily marked by painting a large unique number on each clean and dry dorsolateral side of the carapace using a nontoxic white oil-based paint marker (563 Speedry, Diagraph, Marion, IL). After processing, each turtle was released close to the point of capture, and the coordinates were obtained with a handheld GPS unit (Garmin eTrex, Garmin International, Olathe, KS; error: < 15 m).

Movement Observations

The relatively clear water allowed us to paddle the length of the field site and observe and identify turtles that were both basking and swimming. We further supplemented sampling by occasionally kayaking between US 441 and River Rise to detect dispersal beyond our primary study site. Transects were carried out 23 May–15 October, usually between 1100 and 1600 hours to maximize turtle observations.

We cautiously approached and examined every turtle we encountered and looked for the presence of drill holes, an iButton, or a painted number on the carapace. We made an effort to avoid disturbing turtles and inspected basking turtles using binoculars. On encountering a marked turtle, we recorded its approximate position using a handheld GPS, the time of day, whether it was swimming or basking, and its identity.

We used “the minimum direct distance over water between the 2 most distant points of [observation]” (Sexton 1959:137) to estimate the minimum linear aquatic home range and extent of long-term movements. This method was used previously in studies of 2 species of Pseudemys (Pseudemys nelsoni, Kramer 1995; P. c. suwanniensis, Jackson and Walker 1997). Visual observation points were combined with capture data, imported, and plotted on a 2007 satellite image into Google Earth (v. 4.2; Google Inc., Mountain View, CA). When viewing the image from an eye altitude of 200 ± 10 m, we obtained robust measurements of movements by manually tracing midriver distances between turtle observations. We used the minimum distance between observations to estimate movement between different days. In cases of more than one observation of the same individual on the same day, we calculated the minimum distance between observations that the turtle must have covered. We used only data from turtles that were positively identified and estimated home range only for individuals we observed at least 30 days after original capture.

A separate mark–recapture study headed by G.J. was conducted concurrently in the first 1.1 km of the river starting at River Rise, 4 km upstream from US 441, which allowed us to extend our observations on turtle movement to 7.4 river km. Five monthly sampling sessions of more than 120 person-hours of snorkeling were conducted between May and September 2007; as many turtles as possible were hand captured, marked, and processed. The marking schemes were identical in both studies, and there was no duplication in numbering. However, because of the different goals of that study, few individuals were paint marked, and few exact coordinates of captures and observations were recorded. Therefore, we can report only anecdotal observations on long-distance movement and home range of these turtles.

Spatial Distribution and Abundance

We conducted several pilot transect surveys where we observed an unequal distribution of turtles across the length of the study site. Therefore, we divided the study site into 16 sections of varying lengths based on the following river characteristics that might be biologically important to turtles: major changes in depth (visually detected from the surface), compass orientation (which affects the amount of sunlight exposure during the day), and width.

We conducted 21 counts of both basking and swimming turtles on 13 days between 5 June and 7 July 2007 to quantify the spatial distribution of turtles and to assess relative abundance. Surveys were carried out by a single observer (Y.K.) in a kayak moving in the middle of the river. Thirteen surveys were made going downstream and 8 upstream. Surveys took 2–4 hours each since a variety of observations were being recorded. On encountering large aggregations of basking turtles, the observer slowed and used binoculars to count as many turtles as possible before approaching and possibly disturbing them. The start and finish time of surveys varied, with the earliest surveys begun at 0830 hours and the latest surveys finished at 2000 hours. Most surveys were carried out between 1100 and 1700 hours to maximize the number of turtles basking and increase detection by taking advantage of the best light conditions. Because Suwannee cooters reach sexual maturity around a CL of 190 mm for males and 275–300 mm for females (Jackson and Walker 1997; Huestis and Meylan 2004), instead of life stages, we defined 3 size classes: large (L; CL ≥ 200 mm), medium (M; 70–200 mm CL), and small (S; CL ≤ 70 mm). We recorded the number of individuals per river section in each size class (L, M, or S or only observed a head).

Habitat Assessment

We collected data on 5 environmental variables (Table 1) to examine macrohabitat characteristics that might influence turtle distribution. On 6 August 2007, we took 66 midriver depths at ∼ 50-m intervals by dipping 2 3.3-m interlocking PVC pipes graduated at 0.1 m, starting at US 441 and working downstream. During the turtle counts in June and July 2007, we used a digital thermometer to record water temperature (T_water) 20 cm below the surface in the middle of the river at the beginning of each section. Each section's minimum surface area (not hidden by overhanging trees) was determined based on 2007 satellite images using Google Earth. We visually approximated the major compass orientation for each section using the built-in compass tool in Google Earth.

Table 1 Physical parameters of the 16 study sections of the study site in the Santa Fe River, northern Florida. Area  =  river surface area (ha); depth  =  mean midriver depth (m); T_water  =  mean midriver water temperature at 20-cm depth; basking sites  =  number of counted basking objects in the section; direction  =  approximate downstream river flow using 360-degree compass orientation; total  =  combined counts for all turtles observed during the 21 sampling periods.

View Fullscreen

We located 132 known and potential aerial basking objects throughout the study site between 28 June and 12 July 2007. We measured most basking objects (length, width, type of object, angle, height above river surface, depth of water at basking site, distance from shoreline; Kornilev 2008) on which we observed turtles as well as objects that subjectively appeared to be suitable. Although this was not a complete count of basking locations, we attempted to include a representative sample in terms of type (logs, rocks) and position in relation to shoreline and sun orientation. The number of basking sites was relatively stable during the study; that is, it did not change because of changes in water level or the infusion or elimination of objects as a result of storms.

A multivariate regression analysis was performed in R (v. 2.6.2; R Development Core Team 2008) to identify the most important habitat predictor variables for relative abundance based on the raw counts of the various size classes of turtles. We used mean water depth, mean T_water, compass direction, section surface area, and number of basking sites counted as independent variables for each river section. Unless otherwise specified, α  =  0.05 for all statistical tests, and standard deviation is presented as ± 1 SD.

Visual Observations and Captures

Of the 50 turtles from group A, 8 (6 ♂ and 2 ♀) were never positively resighted. We recaptured 12 individuals (3 ♂, 8 ♀, 1 juvenile; 4 individuals were recaptured 2 times and 1–3 times). Of the 39 turtles from group B, 12 (3 ♂, 8 ♀, 1 juvenile) were never positively resighted; we recaptured 8 individuals (2 ♂, 6 ♀, 1 female recaptured twice). Excluding original captures, we acquired 528 locations of turtles from both groups through visual observations and recaptures. We positively identified 69 individuals during 375 sightings; 67 of these sightings were a second or third observation for the same day.

Linear Home Range and Long-Term Movements

We estimated the linear home range for 26 turtles (6 ♂, 19 ♀, 1 medium-sized juvenile) observed more than 30 days after initial capture. The mean distance between initial capture and last observation (recapture or positive resighting) was similar for both sexes (♂: 484 m, ♀: 625 m; Table 2). Females showed a much wider range of distances between initial capture and last observation as well as slightly greater variation and generally larger home ranges than males (♂: 200–1600 m, ♀: 800–2800 m).

Table 2 Mean distance between initial capture and final observation, based on mark–recapture and visual observation. Days  =  mean number of days between first capture and last observation; SD  =  one standard deviation.

View Fullscreen

We observed pronounced individual variation in patterns of long-distance movements and changes in centers of activity, even though turtles had been handled similarly. Some individuals had strong site fidelity, highlighted by basking on the same log throughout a period of several weeks. Several turtles were consistently positively identified. For example, female #1113 (CL  =  325 mm) was observed 8 times in 17 days basking on the same log. Female #1115 (CL  =  348 mm) was observed postcapture 7 times in 19 days within a 15-m range; on 2 subsequent observations, she moved 200 m upstream and then returned to her preferred log within 17 days. Female #1173 (CL  =  360 mm) moved 450 m upstream after being captured, but within 5 days she returned to the original capture location and was observed on 4 occasions on the same log during the next 16 days. Following some short-distance movements, she was 6 m from where originally captured on her last observation 50 days after capture.

Site fidelity might not be limited to basking sites. At a location where we rarely saw turtles between 0900 and 1700 hours, Y.K. opportunistically caught an unmarked large male with a distinctive stubbed tail on 3 occasions (7 and 11 June, 3 July 2007). Captures occurred between 1900 and 2015 hours, all within 10 m of one another, in the shallow (< 25 cm) gravel rapid between sections 2 and 3 (Fig. 2). Whether the turtle had been feeding on nearby copious submerged vegetation or was in proximity to a nighttime refuge is unknown.

To illustrate the variety of turtle behaviors, we describe the movement patterns of 6 large turtles (4 ♀, 2 ♂) captured initially in the concurrent study in the first 1.1 km downstream from River Rise and later detected downstream from US 441. Interestingly, 4 of these turtles were caught originally during the same sampling session on 7 May 2007. In total, 109 Suwannee cooters were captured 168 times at the River Rise site, but none of these turtles were captured originally downstream of US 441. Since capture locations were only approximately known, we calculated minimum possible distances; movement data for these individuals are not included in the results presented in Table 2. Although these turtles might have moved longer distances than most turtles from groups A and B, we consider such types of behaviors and responses typical for adult Suwannee cooters from this area.

• A female (#82, CL  =  303 mm) caught within the first 200 m downstream from River Rise was recaptured 37 days later in the same area. However, 10 days later, we observed her 1850 m downstream from US 441, > 4.7 km from the initial capture site. She was located 58 days later only 115 m upstream from the previous observation.

• A female (#38, CL  =  350 mm) caught in 2006 close to River Rise was found 13 months later more than 4 km downstream. During the 31 days she was observed in 2007, the longest distance she moved between observations was 325 m, and the distance from original capture to last observation was 187 m. Even though she was captured 4 times in 2007, retained for 24 hours on 1 occasion, and observed on 3 additional days, she stayed within a shallow area with rapid flow. Limited movement might have been induced by poor health since she had an overall sickly appearance, heavy leech load, and an injury to the lower jaw.

• A female (#74, CL  =  374 mm) was captured at River Rise and recaptured 72 days later at the River Rise headspring. She was recaptured 72 days later after a downstream movement of at least 4.4 km. After she was released, she was resighted the same day 1.5 km farther downstream.

• A female (#81, CL  =  368 mm) was caught originally at River Rise and recaptured 87 days later at least 3.4 km downstream. Four days later, she was observed about 1.5 km downstream but within 4 days had turned around and moved > 2.1 km upstream. She was recaptured at the River Rise headspring 35 days later and was observed 24 days later about 1.1 km downstream. Three observations made in 2 hours during 1 day showed a downstream displacement of 830 m.

• A male (#91, CL  =  290 mm) caught within the River Rise study area was recaptured at least 4.8 km downstream 97 days later. After 8 more days, we observed the turtle 650 m farther downstream. Two later observations during the next 17 days were made less than 100 m apart. Therefore, the longest distance between observations was more than 5.5 km, and the distance from capture to last observation was at least 4.9 km.

• Another long and rapid dispersal we observed was of an unidentified male originally marked upstream at River Rise site and observed 7 days later downstream in our field site. His exact original capture location was unknown, but he moved at least 4.8–5.9 km.

We observed only 3 turtles with an iButton upstream from US 441 despite making more than 15 kayak trips to River Rise throughout the study. An unidentified adult female was observed on 26 and 27 June 2007 at the same basking log 1.2 km upstream from US 441. Based on all original capture points, she had moved upstream 2.2–3.0 km 35–40 days after capture. An adult female (#1145, CL  =  341 mm) was caught 111 days after the original capture, 0.5 km upstream from US 441. We could not determine whether these observations pertained to the same or different individuals.

Distribution by Habitat

Overall, the greatest turtle concentrations were in areas around basking sites that allowed numerous turtles to bask simultaneously. Several sections of the river had very similar surface areas but strikingly different abundances; for example, #8 (high abundance) and #10 (low abundance), #3 (low) and #15 (high), and #4 (low) and #13 (high; Fig. 3). Large individuals were observed more than medium and small ones throughout the study site and in general showed widespread distribution. Assuming identical detectability, small turtles seemed to be concentrated in certain sections (e.g., #6 and #14) and were uncommon or absent in most others. Our experience suggests that detectability of all size classes of basking turtles was consistently very high throughout the study.

View Figure

• Download as Powerpoint
• Download Figure

The distribution of turtles of different size classes along the river was likewise affected by macrohabitat characteristics. The correlation matrix from a multivariate regression analysis of macrohabitat characteristics showed very similar results across size classes, but total counts correlated slightly better than individual size classes (Table 3). Therefore, we present only results based on the total number of turtles counted per section. In the complete model, only the number of basking sites was significant (t  =  2.778, p  =  0.02); the model's adjusted R² was significant (R²  =  0.662; F_{5, 10}  =  6.889, p < 0.01). The number of basking sites and depth were the best predictors for turtle abundance according to independent stepwise, forward elimination, and backward elimination models. For the best predictors model, the number of basking sites was significant at α  =  0.001 (t  =  4.637, p < 0.01), depth was significant at α  =  0.05 (t  =  2.713, p  =  0.018), and the model's adjusted R² was highly significant (R²  =  0.728; F_{2, 13}  =  21.09, p < 0.01).

Table 3 Correlation coefficients between river section's macrohabitat variables and turtle counts.

View Fullscreen

River Rise is > 30 m deep with a diameter around 40 m with mostly vertical limestone walls except for a 5 × 10-m terrace at a depth of around 9 m. Turtles resting on the terrace have been captured with little effort since they usually do not try to swim away. Many individuals were chased for 5 minutes around the Rise. They swam continuously at a depth of around 3–7 m to prevent capture, generally swam along the side of the spring, and were not observed diving deeper into the spring. In addition, we frequently observed courting at various water depths and recorded turtles feeding at or below the surface in shallow (< 1.5 m) locations even though the distribution of bottom-growing vegetation was limited based on depth and water clarity. While snorkeling, we noted P. c. suwanniensis escaping directly toward and hiding under submerged logs and debris at depths of 2–9 m but never deeper than ∼ 10 m.

Surface area and compass direction were not good predictors of turtle numbers, although we initially hypothesized they might be. Water depth was correlated strongly but nonlinearly with T_water and was selected as the preferred predictor; however, this positive relationship was due to the specific study site characteristics and cannot be extrapolated to other locations.

We observed a clumped distribution of hatchlings in the section of the Santa Fe River downstream from River Rise, but we never observed hatchlings within the initial 0.5 km even though adults were abundant. Hatchlings were commonly observed farther downstream. Even though hatchling turtles were not uniquely marked, we repeatedly observed similar numbers and sizes of hatchlings on the same logs as previously detected. Strong site fidelity is to be expected because hatchlings are weaker swimmers than adults and it would be energetically expensive and dangerous for them to move up and downstream through areas inhabited by large predators such as alligators and gars. The clumped hatchling distribution is probably related to proximity to successful nesting sites and the presence of shallow water and vegetation rather than to other river or microhabitat characteristics.

A shift in habitat use was observed, both on a daily and a monthly basis. Certain sections repeatedly had observable turtles mostly during specific times during the day; for example, turtles were often observed feeding in the afternoon in section 4 but were rare during earlier hours. Large aggregations of up to 30 large individuals were commonly observed in shallow vegetation mats composed predominantly of water pennywort (Hydrocotyle sp.), small duckweed (Lemna valdiviana), water spangles (Salvinia minima), and subsurface nonnative hydrilla (Hydrilla vericillata) and parrot feather (Myriophyllum aquaticum). Turtles were observed surface basking and feeding on such floating vegetation but not on mats dominated by the nonnative water hyacinth (Eichhornia crassipes).

Home Range, Movements, and Distribution

The extent to which turtles occupy a habitat is largely dependent on the availability of food resources, mates, shelter from adverse conditions and predators, nesting sites, and access to locations that meet specific biophysical requirements, such as thermally optimal basking sites (Boyer 1965). Such resources usually occur patchily, whether in terrestrial or in aquatic habitats. In terrestrial chelonians, habitat use is usually nonlinear through time, with home ranges covering freely accessible, mostly large 2-dimensional landscapes. Assuming that resources are available, movements need not involve wide-ranging frequent travel across the home range, and home ranges, while extensive, may not be very large in diameter.

Riverine species, however, are confined by the linear extent of habitat and live in a 3-dimensional world. Like terrestrial species, resources may not be evenly spaced or confined within proximate sections of river. In addition, fluctuations in river characteristics, such as width, depth, temperature, flow rates, and visibility, further serve to fragment patchy resources. Therefore, it should not be surprising that river-dwelling turtles frequently travel extensive linear distances between feeding, nesting, basking, and shelter locations (e.g., Vogt 1980; Dodd et al. 1988; Pluto and Bellis 1988; Dreslik et al. 2003). Our results confirm that Suwannee cooters do so as well.

On the Santa Fe River, linear home range and movement patterns are extremely variable, as our examples illustrate. The variance of these patterns seems to be explained by individual differences among turtles rather than by sex or size differences or habitat characteristics of the river, such as depth, vegetation beds, or the positioning of basking sites. Similar conclusions have been obtained in several other studies on large emydines. For example, nesting female Suwannee cooters in the Wakulla River, Florida, typically had strong nest site fidelity (within a 200-m segment), although they sometimes nested > 1.7 km away from previous nesting sites (Jackson and Walker 1997). Some individual Trachemys scripta elegans have preferred basking sites, whereas others do not (Cagle 1944). We observed similar results in our study.

In other aquatic turtle species (e.g., T. s. scripta; Schubauer et al. 1990), males have greater home ranges than females, and gender might influence linear home ranges on the Santa Fe River. At our study site, it appears that there might be a marked difference between the linear home ranges of large and small individuals, but our data are not sufficient to determine if this result is sex or size based. In the few other studies of P. concinna movements, Jackson and Walker (1997) noted that larger P. c. suwanniensis might have greater home ranges than smaller individuals in western Florida, and in Illinois, male P. concinna had greater home ranges and daily movements than females (5.3 vs. 4.9 ha; Dreslik et al. 2003). In all studies of P. concinna movements thus far (see also Buhlmann and Vaughan 1991; Sterrett et al. 2008), however, total sample size has been < 9, making generalizations premature.

Linear home range estimators for aquatic turtles must take into account the 3-dimensional aspect of their habitat. Characteristics at the river surface such as river width, presence of basking logs, and vegetation mats undoubtedly influence movements and subsequent home ranges, especially for basking herbivorous turtles (Marchand 1945b). However, the extent to which these variables alone influence distribution is unclear. Indeed, cryptic factors unseen to surface observers may play an important role in how turtles use riverine habitats. Compared to this study, 2 radio-tracked P. concinna in West Virginia were not observed deeper than 2 m, although the maximum water depth was only 3 m; non–radio-tracked turtles also were concentrated in water less than 2 m deep (Buhlmann and Vaughan 1991). Together, these observations suggest that river cooters are most likely to use the uppermost 1–2 m of the water column, although they are familiar with deeper habitats and position themselves to be in close proximity to underwater shelter should escape be necessary.

In conjunction with depth, water clarity might influence turtle movement patterns and spatial distribution. Clarity influences food plant distribution by allowing sunlight to penetrate into the water column. Submerged vegetation grows at greater depths in Florida's clear-water rivers than in blackwater streams. In clear waters, turtles also are able to see distant predators, such as mammals and alligators. Our results suggest a slightly larger female linear home range in the blackwater Santa Fe River than the home ranges of 200–600 m reported in the clear-water Wakulla River (Jackson and Walker 1997). Home range could be inversely proportional to food resource availability, which is likely lower in the blackwater Santa Fe than in the clear-water Wakulla (D. Jackson, pers. comm.).

Abundance of turtles appeared to be related to availability of nonstable resources, such as basking locations, and the location and type of such sites influenced individual habitat preferences. For example, in March 2008 after the main study was completed, some of the basking logs and vegetation mats were no longer available because water depth had increased slightly (∼ 0.3 m). Turtles were then observed in much greater numbers at locations where previously they were scarce (e.g., sections 2 and 3). Basking site availability plays a key role in motivating turtles to inhabit a specific area (Cagle 1944, 1950; Cagle and Chaney 1950; this study). Our subjective impressions were that turtle size and basking site size were correlated, with larger turtles choosing wider sites. A similar correlation was noted by Boyer (1965), who found that turtles most often emerged on sites no less than two-thirds of their body width.

In addition, spatial distribution and movements could be influenced by social interactions resulting from the presence or absence of other turtles around basking sites. Boyer (1965) provided evidence that turtle aggregation might play a role in basking site selection even when other conditions were the same. Aggregation might be beneficial by increasing the ability of the group to spot predators or increasing mating opportunities (Flaherty 1982; Flaherty and Bider 1984). Another possibility is that after a turtle is observed by conspecifics to use a basking site, other turtles assume it is of sufficient quality and congregate to utilize the resource. We did not observe any intraspecific aggressive interactions.

Our results raise questions as to what extent capture and handling affect the behavior of turtles. Although most turtles were released immediately after marking and were manipulated in a similar manner, the movement data suggest pronounced individual variation. Recent studies have failed to reach a consensus about the impacts of handling on stress levels and behavioral changes in chelonians (Cabanac and Bernieri 2000, Glyptemys insculpta; but see Pike et al. 2005; Kahn et al. 2007, Gopherus polyphemus). For aquatic turtles (including P. c. suwanniensis), Marchand (1945b:77) reported that an unspecified number of “marked turtles were common in… a distance of about [8 km] from the point of release”; he attributed such movements to dispersal due to stress from prolonged capturing and abnormal density due to release of captured individuals in one location. Suwannee cooters are long lived, have extended home ranges and high adult survivorship, and may not be dependent on limited resources and therefore might exhibit a strong response to disturbance even if predation risk in the form of handling was small (Gill et al. 2001; Beale and Monaghan 2004). Studies on stress hormones should help resolve this question.

The conservation of river biota in northern Florida is complicated by an extensive array of influences involving water and land use and underground recharge. Unlike some terrestrial communities, rivers cannot be conserved in a small-patch framework. What influences part of the river can influence much of it, extending the need for long-term management of entire drainage basins. This interrelationship is especially apparent when the protection of large riverine turtles is included in management plans (Moll and Moll 2004; Jackson 2005). The temporally varying habitat use and long-distance movements by Suwannee cooters in the Santa Fe River illustrate the linear extent that some aquatic biota utilize in order to maintain populations. In order to preserve turtle communities, it is necessary to think in kilometers or, better yet, dozens of kilometers. It is only by understanding spatial use in relatively undisturbed areas that normal baselines for conservation can be met.