Animals gather in aggregations to reduce predation risk, facilitate mating, and access resources with patchy distributions (Krebs and Davies 1993; Alcock 2013). Turtles are known to aggregate at basking sites (Carr 1952; Lindeman 2013), in hibernacula (Brown and Brooks 1994; Litzgus and Mousseau 2004; Newton and Herman 2009), at nesting beaches (Alho and Pádua 1982; Doody et al. 2003; Bernardo and Plotkin 2007; Vogt 2008), at drinking sites (Doody et al. 2011), and at feeding sites (Dolbeer 1969; Bjorndal et al. 2000; Vogt 2008; Wood et al. 2013). These resources are not always stable and may shift geographically and over time. Our long-term study of a riverine ecosystem in northern Florida provided an opportunity to study how a geographic shift in food resources affected one wide-ranging herbivorous freshwater turtle.

In 2006, we initiated long-term capture–mark–recapture studies of the 11 native freshwater turtle species that inhabit the Santa Fe River (SFR) and its associated springs in northern Florida (Johnston et al. 2012, 2015, 2016). During the eighth year of our study, we observed a large aggregation of turtles dominated by Suwannee cooters (Pseudemys concinna suwanniensis) at Gilchrist Blue Springs Park (GBSP) in August–October 2013 and again in March–May 2014 (Figs. 1 and 2; Adler et al., in press). The highly conspicuous aggregations exceeded anything we had previously seen and were apparently unprecedented in the SFR ecosystem according to residents whose families have lived along the SFR for multiple generations (K. Davis, owner of GBSP, pers. comm., September 2013; M. Wray, owner of Ginnie Springs Outdoors, pers. comm., September 2013). In this article, we describe the aggregations, use data gathered while studying turtles in GBSP and throughout the SFR ecosystem during 2006–2016 to place the aggregations in context, and suggest a hypothesis that may explain why they occurred.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

STUDY AREA

The SFR is a 113-km tributary of the Suwannee River that originates in swamps near Lake Santa Fe in northern Florida. It begins as a tannin-stained blackwater stream in its upper reaches but receives substantial input of clear water from ≥ 45 artesian springs in its lower 37 km (Fig. 2). Johnston et al. (2016) provided a detailed description of the SFR and examined how the mosaic of habitats created by springs affects native freshwater turtles. In the upper reaches of the SFR, tannin-stained water inhibits growth of submerged aquatic macrophytes. Aquatic vegetation is limited to floating plants such as spatterdock (Nuphar advena), duckweed (Lemna sp.), water spangles (Salvinia minima), and introduced water hyacinth (Eichhornia crassipes). Spring water dilutes the tannins in the lower 37 km of the SFR and increases clarity of the river, especially during periods of little to no rainfall. Greater water visibility allows growth of patches of submerged aquatic macrophytes such as muskgrass (Chara sp.), carpet moss (Fontinalis sp.), introduced hydrilla (Hydrilla verticillata), introduced Indian swampweed (Hygrophila polysperma), water milfoil (Myriophyllum sp.), strap-leaf sagittaria (Sagittaria kurziana), and tapegrass (Vallisneria americana). Epiphytic cyanobacteria (Lyngbya wollei) and algae (Vaucheria sp.) are abundant during drought periods.

During periods of high rainfall, the entire 37-km spring-influenced section of the SFR temporarily becomes dark due to large volumes of tannin-stained water flowing downstream from the upper SFR. During 2006–2016, we conducted mark–recapture studies of turtles by snorkeling in the spring-influenced section of the SFR during periods of clear water. Our ability to conduct fieldwork was directly affected by the concentration of tannins in the water. Four periods (2–9 mo in duration) with water too dark to permit snorkeling occurred during 2006–2012. During 24–26 June 2012, Tropical Storm Debby brought 28.8–35.3 cm of rain to the upper SFR basin (Suwannee River Water Management District 2017a). This storm and several subsequent episodes of heavy rainfall resulted in tannic conditions lasting 34 mo (June 2012–May 2015) in the spring-influenced section of the SFR. Rainfall was higher than long-term (1932–2012) monthly averages throughout the SFR basin during 11 mo between February 2013 and September 2014 (Suwannee River Water Management District 2017b). The spring-influenced section of the SFR was flooded during June–August 2012, March–May 2014, and September 2014 (Suwannee River Water Management District 2017c). When water visibility returned to normal in May 2015, we observed that nearly all submerged aquatic macrophytes had disappeared and were replaced by a community of macroalgae dominated by Vaucheria sp.

GBSP is a 120-ha privately owned park located on the southern shore of the SFR approximately 30 km upriver from the confluence with the Suwannee River (lat 29°49′47.64″N, long 82°40′58.44″W, WGS84; Fig. 2; Johnston et al. 2016). Three large spring groups in the park (Blue Spring, Naked Spring, and Little Blue Spring) converge into a main run (approximately 1 m deep, 6–27 m wide) that flows 350 m north through floodplain forest and into the SFR. An elevated boardwalk follows the full length of the main run on its eastern edge (Fig. 3). The total area of aquatic habitat created by the springs and their network of runs is 0.9 ha. During 2006–2013, aquatic vegetation was abundant and dominated by H. verticillata, with native species such as spring-run spider lily (Hymenocallis rotata), red ludwigia (Ludwigia repens), pickerel weed (Pontederia cordata), S. kurziana, and V. americana restricted to the margin or downstream section of the main run. During August–October 2013, a large aggregation of P. c. suwanniensis selectively consumed almost all H. verticillata (Adler et al. in press). During the 2013–2014 winter, H. verticillata regrew rapidly and returned to its original abundance by March 2014. A second aggregation of P. c. suwanniensis occurred in GBSP (March–May 2014), and the turtles again selectively consumed almost all H. verticillata. The aquatic plant community was dominated by native species from June 2014 through 2016.

View Figure

• Download as Powerpoint
• Download Figure

METHODS

We conducted visual surveys of P. c. suwanniensis in GBSP by counting all turtles we observed in the clear water while walking down the boardwalk that parallels the main GBSP Run (Fig. 3). We surveyed the run in this manner approximately once per month during 2006–2016. We made our observations between 0800 and 0930 hrs on days when recreational activity in the park was relatively low. Employees at GBSP made daily observations of turtles in the park and notified us if they observed any obvious changes in turtle abundance or unusual turtle activity.

We gathered most of our data on P. c. suwanniensis throughout the SFR ecosystem by snorkeling and capturing turtles by hand (Johnston et al. 2011, 2016). During each snorkel survey, a group of 4–8 experienced snorkelers captured all turtles observed from midmorning to midafternoon (∼ 0900–1500 hrs), placed the turtles in canoes, and then returned to shore to measure and mark them prior to release. In 2006–2016, we conducted 34 surveys of a 20.6-km section of the SFR, extending 16 km upriver and 4.6 km downriver from GBSP. We also conducted 8 surveys in the Ichetucknee River, a major 9.7-km spring-run tributary, and 8 surveys in GBSP. When we sampled in GBSP, 2–3 people and a multifilament gill net (45.7 × 3.6 m; 0.1-m bar mesh) were positioned at the end of the spring run to temporarily prevent turtles from swimming out of the sampling area. Twenty snorkelers captured turtles during the survey in GBSP on 8 September 2013.

We measured straight midline carapace length (CL) and straight midline plastron length (PL) of each P. c. suwanniensis to the nearest 1 mm using aluminum tree calipers (Haglöf®, Långsele, Sweden). We weighed turtles < 5 kg to the nearest 1 g using a portable digital scale (Ohaus®, Pine Brook, NJ) and those > 5 kg to the nearest 10 g using a spring scale (Pesola®, Baar, Switzerland). We marked small (< 120 mm PL) turtles individually by filing or cutting notches in the marginal scutes and peripheral bones using a standard numbering system (Cagle 1939). We marked larger (> 120 mm PL) individuals with drill holes following the same system. We marked turtles captured in the Ichetucknee River by inserting passive integrated transponder (PIT) tags into the muscle and connective tissue between the plastron and pelvis lateral to the midline (Runyan and Meylan 2005). Turtles captured in the large sample at GBSP on 8 September 2013 were also temporarily marked on the carapace with a small (1–2-cm-diameter) dot of white, nontoxic, oil-based paint (563 Speedry, Diagraph, Marion, IL; Kornilev et al. 2012) to ensure that individuals previously measured and released were recognized by snorkelers and not repeatedly captured during the same sampling session. Females were palpated to determine the presence of eggs. We released all turtles at the capture site on the same day of capture.

We used the total number of individuals captured during each survey in GBSP to estimate minimum density of P. c. suwanniensis in the 0.9-ha study area. The sum of body masses of all P. c. suwanniensis captured during each survey in GBSP provided a minimum estimate of biomass. To evaluate demographic composition of P. c. suwanniensis in each survey at GBSP, we categorized each individual into one of the following demographic groups following Johnston et al. (2016): unsexed juveniles (< 180 mm PL), juvenile females (180−295 mm PL), adult females (≥ 296 mm PL), and adult males (≥ 180 mm PL). We used Google Earth Pro to estimate distances marked turtles travelled via river between GBSP and capture sites before and after 8 September 2013. Visual boardwalk survey data gathered before, during, and after each aggregation are presented as mean ± standard deviation (minimum–maximum) number of P. c. suwanniensis per survey. Turtle taxonomy and common names follow the Turtle Taxonomy Working Group (2017).

RESULTS

GBSP Visual Boardwalk Surveys

During 86 surveys of the main GBSP Run (June 2006–July 2013), we observed an average of 15.1 ± 3.7 (8–25) P. c. suwanniensis. On 10 August 2013, we observed approximately 300 P. c. suwanniensis. We continued to observe at least 250 turtles per survey (maximum 351 on 3 September 2013) until 22 October 2013 (24 turtles). During November 2013–February 2014, counts returned to baseline (16 ± 4.1 per survey, 11–21, n = 4). On 14 March 2014, we counted approximately 200 turtles in the run. Our counts were at least 250 turtles per survey (maximum 362 on 4 May 2014) during April and May 2014. Turtle counts in all surveys during June 2014–December 2016 averaged 16.1 ± 4.3 (6–24; n = 31 surveys). During the 2013 and 2014 aggregation events, we observed a daily routine in which turtles from the SFR entered GBSP at dawn, consumed primarily H. verticillata for several hours, and then returned to the river.

Snorkel Surveys

During 7 snorkel surveys in GBSP between May 2006 and August 2012, we hand-captured 8 species, with an average of 5.4 species/survey and 32.3 individuals/survey (Table 1). The largest total sample was 48 turtles representing 5 species on 21 February 2010. On 8 September 2013, we hand-captured 496 turtles in GBSP representing 5 of the 8 previously documented species (Table 1). The numbers of Florida red-bellied cooters (Pseudemys nelsoni), loggerhead musk turtles (Sternotherus minor), common musk turtles (Sternotherus odoratus), and yellow-bellied sliders (Trachemys scripta) in the 8 September 2013 sample were within the range of the number of individuals captured in previous surveys at GBSP. However, the number of P. c. suwanniensis in the 8 September 2013 sample (n = 477) was 18.4 times greater than the maximum captured during any previous snorkel survey.

Table 1. Numbers of turtles hand-captured during sampling sessions in Gilchrist Blue Springs Park. Combined data from 7 sessions in 2006–2012 presented as mean (minimum–maximum) per session.

View Fullscreen

Pseudemys c. suwanniensis was the most abundant species in all GBSP surveys (Table 1). Density (530 turtles/ha) and biomass (2242 kg/ha) of this species on 8 September 2013 were vastly higher than the previous maximum values at GBSP (17.8 turtles/ha, 47.3 kg/ha). All demographic groups and all size classes between 52 and 363 mm PL were represented in the 8 September 2013 sample (Fig. 4). We captured all demographic groups in greater numbers on 8 September 2013 than the maximum in any previous survey (Table 2). We found no gravid females on 8 September 2013; however, we captured 21 gravid females in GBSP during April–July in previous years.

View Figure

• Download as Powerpoint
• Download Figure

Table 2. Numbers of unsexed juvenile, juvenile female, adult female, and adult male Suwannee cooters (Pseudemys concinna suwanniensis) hand-captured during sampling sessions in Gilchrist Blue Springs Park. Combined data from 7 sessions in 2006–2012 presented as mean (minimum–maximum) per session. Definitions of demographic groups follow Johnston et al. (2016).

View Fullscreen

The 8 September 2013 sample included 134 P. c. suwanniensis that had previously been marked. Seventy-seven marked individuals had previously been captured in GBSP and/or in the SFR within 2 km of GBSP. Forty had previously been captured in the SFR 2–16 km upriver from GBSP, and 10 had previously been captured 2–5 km downriver from GBSP.

During 2014–2016, we recaptured 122 marked P. c. suwanniensis that had been captured in the 8 September 2013 GBSP aggregation. Seven individuals were recaptured in GBSP, 49 in the SFR within 2 km of GBSP, 36 in the SFR 2–16 km upriver from GBSP, and 18 in the SFR 2–5 km downriver from GBSP. We also recaptured 19 individuals in the Ichetucknee River 25–30 km downriver from GBSP.

DISCUSSION

The present study provides another example of the benefits of long-term ecological research. The data we collected during the 7 yrs prior to and 3 yrs after the 2013 and 2014 aggregation events allowed us to examine this aggregation phenomenon in context and evaluate its causes. Large groups of P. c. suwanniensis have been reported previously from Florida. Knight (1871) noted that groups occurred in bayous near the mouths of streams near Tallahassee, and Carr (1952) reported seeing hundreds of P. c. suwanniensis in the grassy flats at the mouth of the Suwannee River. Both of these reports are anecdotal and neither provided insights into the functional significance of the aggregations. Data provided by Jackson (1970) suggest a density of 741 P. c. suwanniensis/ha in Fanning Spring in the Suwannee River basin. The aggregation was dominated by juveniles, suggesting Fanning Spring functioned as a nursery.

Johnston et al. (2016) suggested that GBSP also functioned as developmental habitat for P. c. suwanniensis. During 2006–2012, GBSP was dominated by unsexed juveniles, and females were frequently observed nesting in the surrounding uplands. However, our data and observations during 2013 and 2014 suggest the aggregation events were unrelated to nesting or an influx of unsexed juveniles. The 8 September 2013 sample included all demographic groups and only 3.1% were unsexed juveniles (Table 2; Fig. 4). No female in the 8 September 2013 sample was gravid. The last nesting of which we are aware in 2013 occurred during July. Suwannee cooters are known to nest from late March through early August (Jackson and Walker 1997). We did not capture and examine turtles in the March–May 2014 aggregation, but the demographic composition appeared to be similar to the first aggregation. Only 1 female was observed nesting at GBSP during March–May 2014 (L. Matthews, pers. comm.).

We suggest that the herbivorous P. c. suwanniensis gathered at GBSP to forage based on our observations of their feeding behavior during the aggregations (Fig. 1). Furthermore, both aggregation events ended after almost all H. verticillata was consumed, clearly illustrating that the aggregations were associated with the presence of food in GBSP. Although abundant aquatic vegetation dominated by H. verticillata occurred in GBSP every year since the beginning of our study, we never observed an aggregation prior to 2013. The proximate cause of the aggregation events likely involves other factors in addition to the presence of food in GBSP.

We hypothesize that hydrological changes in the SFR basin contributed to the turtle aggregations at GBSP. Water levels in the Floridan Aquifer have declined since the 1950s, which has led to reduced spring flow at several springs in the SFR basin (Knight 2015; Johnston et al. 2016). The most severe declines in flow occurred in springs at relatively high elevations. These springs are located upstream from GBSP and contribute clear water that helps to dilute tannic water originating from the upper SFR. With reduced spring flow, there is less water to achieve the dilution effect in the lower reaches of the SFR, especially during periods of heavy rainfall. An extreme drought in May 2012 caused several springs to stop flowing. When Tropical Storm Debby arrived in late June 2012, heavy rainfall in the upper SFR basin resulted in a substantial increase in tannic water flowing down from the upper reaches. It overwhelmed the capacity of springs to dilute the river water and contributed to the unusually long tannic period (34 mo) in the lower 37 km of the SFR. The consequent loss of most submerged aquatic macrophytes in the SFR due to insufficient sunlight penetration was probably the stimulus that led the herbivorous P. c. suwanniensis to seek food in GBSP. It was the only habitat in the lower SFR we observed with abundant submerged aquatic vegetation in 2013 and 2014. Anecdotal observations of P. c. suwanniensis feeding on terrestrial grasses along the shore of the SFR near GBSP during August 2013 (M. Wray, owner of Ginnie Springs Outdoors, pers. comm., September 2013) further suggest that turtles were experiencing and responding to a period of low food availability.

Our observations suggest the turtles did not gather at GBSP in response to flooding. Although the SFR was flooded during the March–May 2014 aggregation, it was below flood stage during the August–October 2013 aggregation. We witnessed 6 flood events in 2006–2016 during which no aggregations occurred.

Our recapture data provide further evidence that the aggregation phenomenon was not a localized event. Turtles in the 2013 aggregation originated from a section of the SFR that is at least 20 km long. Some marked turtles swam 16 km downstream to reach GBSP; others swam at least 4 km upstream. The 19 turtles recaptured in the Ichetucknee River 25–30 km from GBSP during 2014–2016 further suggest that the concentration of individuals at GBSP represented a large portion of the P. c. suwanniensis population in the SFR basin. The long-distance movements we observed in this study build on a recent report by Johnston et al. (2017) of a P. c. suwanniensis that moved 260 km to and from a site in the SFR near GBSP and the mouth of the Suwannee River. Our study provides a context that may explain why P. c. suwanniensis move such long distances.

We do not know which environmental cues the turtles may have used to locate the concentrated food source at GBSP. Turtles downriver from GBSP may have used olfaction to detect chemical cues emitted by H. verticillata, but the olfaction hypothesis cannot explain how turtles upriver would have detected this food source. We hypothesize that turtles upriver may have moved out of their normal home ranges in search of food and opportunistically encountered GBSP. Alternatively, some type of communication among turtles may have enabled the upriver turtles to locate the food source at GBSP. Ferrara et al. (2014) demonstrated that acoustically mediated social behavior occurs in giant South American river turtles (Podocnemis expansa), and we cannot rule out the possibility that P. c. suwanniensis and other wide-ranging riverine turtles also communicate this way. If we observe another large aggregation of P. c. suwanniensis in our study area in the future, we hope to apply some of the techniques used by Ferrara et al. (2014) to examine the possible role of acoustic communication.

The aggregation of P. c. suwanniensis individuals at GBSP was dominated by adult females. In the 8 September 2013 sample, 50.3% of all individuals were adult females and the adult sex ratio was 0.72 male:1 female (Table 2). These data differ substantially from those of the general SFR population (Johnston et al. 2016). Of 1226 P. c. suwanniensis hand-captured in the SFR and its springs, 23.4% were adult females and the adult sex ratio was 1.39 males:1 female (Johnston et al. 2016). We hypothesize that the disproportionate relative abundance of adult females in the GBSP aggregation reflects the greater nutritional requirements of that demographic group relative to all others. Because of their larger body sizes, adult females may also have a greater ability to travel long distances in a riverine ecosystem to access localized food resources.

Hydrilla verticillata is an invasive, introduced species that grows rapidly (up to 2 cm/d), smothers native aquatic macrophytes, and clogs waterways (Langeland 1996). However, our observations show that it provided ecological benefits to P. c. suwanniensis during a period of food scarcity. We do not know where the turtles would have foraged or if survivorship would have been affected if H. verticillata had not been abundant in GBSP during the 34-mo tannic period in the lower SFR. We suggest future research to examine how H. verticillata dynamics affect P. c. suwanniensis and other native turtles. Species known to consume H. verticillata include the North American snapping turtle (Chelydra serpentina) (Johnston and Suarez 2012), P. nelsoni (Bjorndal et al. 1997), P. peninsularis (Bjorndal et al. 1997), and T. scripta (Bjorndal and Bolten 1993). Future studies should also examine the comparative nutritional content of H. verticillata and the variety of native macrophytes in the SFR and its springs.

We are unaware of any published report of a freshwater turtle feeding aggregation similar to those we observed in GBSP. Our observations more closely resemble the high-density aggregations of green sea turtles (Chelonia mydas), which gather to forage on turtle grass (Thalassia testudinum) pastures and then disperse when this food resource is depleted (Bjorndal et al. 2000). Unfortunately, the extraordinary aggregation of P. c. suwanniensis at GBSP may be an early symptom of major ecological changes occurring in the SFR basin, particularly the effect of declining aquifer levels and associated reduction in spring flows. Loss of spring habitats and submerged aquatic macrophytes could severely impact the SFR P. c. suwanniensis population.