Thirteen Years of Turtle Capture–Mark–Recapture in a Small Urban Pond Complex in Louisiana, USA

Greenspaces are increasingly being designed into the planning of developed areas in urban and suburban spaces. Aydin and Cukur (2012) define greenspace as “a type of land use which has notable contributions to urban environments in terms of ecology, aesthetics or public health, but which basically serves human needs and uses.” Greenspaces exist on a continuum from intensely managed areas to wild places, and may include forests, shrublands, meadows, marshes, golf courses, suburban parks, and riparian areas (Guzy et al., 2013). Ponds, rivers, lakes, and reservoirs, sometimes referred to as bluespaces, are often components of greenspaces.

There are many noted environmental benefits of greenspaces (Davern et al., 2016; Rakhshandehroo et al., 2017; Twohig-Bennett and Jones, 2018). One specific benefit that greenspaces offer is a refuge for wildlife, which has been documented in amphibians (Semlitsch et al., 2007; Bogosian et al., 2012), birds (Chamberlain et al., 2007; Fuller et al., 2009), insects (McGeoch and Chown, 1997; Baldock et al., 2015), mammals (VanDruff and Rowse, 1986; Fabianek et al., 2011; Hale et al., 2012), plants (Schwartz et al., 2002), and reptiles (Bogosian et al., 2012; Guzy et al., 2013; Price et al., 2013).

The order Testudines, to which all turtles belong, is among the most imperiled major vertebrate groups in the world today (Rhodin et al., 2018). Of six significant threats to reptile biodiversity, habitat loss (degradation, fragmentation, or conversion) is generally recognized as the most pervasive and a leading cause for noted faunal declines (Sala et al., 2000), including in many turtle taxa (Gibbons and Stangel, 1999; Stanford et al., 2020). Urbanization results in habitat loss, degradation, and fragmentation through increased road networks, increased land use, and subsidized predators, which can negatively impact genetic structure (Rubin et al., 2001), demography (Garber and Burger, 1995; Lindsay and Dorcas, 2001; Aponte et al., 2003), and movements in turtle populations (Marchand and Litvaitis, 2004; Ryan et al., 2008). Direct mortality of turtles via vehicle roadway collisions, especially adult females on nesting forays, is a common occurrence in many urban turtle populations (e.g., Gibbs and Shriver, 2002; Steen and Gibbs, 2004; Aresco, 2005; Steen et al., 2006).

Semiaquatic turtles are a common component of many aquatic greenspace habitats and play an important role as omnivores, scavengers, predators, prey, and nutrient recyclers (Dupuis-Desormeaux et al., 2018). Energetic effects of semiaquatic turtles in a system can be profound, with turtle productivity and biomass often among the highest of all vertebrates (Iverson, 1982; Congdon et al., 1986; Lovich et al., 2018). Undoubtedly, some plants and animals, including turtles, cannot survive in heavily altered urban environments; however, synanthropic species often thrive in such urban greenspaces (Mitchell, 1988; Souza and Abe, 2000; Hays and McBee, 2010; Munscher et al., 2020). Our study focuses on a freshwater turtle capture–mark–recapture effort in an urban greenspace to understand baseline information such as species composition, survival, and individual growth rates. The basic demographic data gained can serve as a starting point for broader questions on urbanization effects and as a comparison to more natural populations.

We conducted our study in two urban man-made ponds within the city limits of Lafayette, the fifth most populous city in Louisiana (2020 Census population: 121,374; U.S. Census Bureau, 2020). The two ponds are in the University of Louisiana at Lafayette’s Research Park. One pond is adjacent to the U.S. Geological Survey Wetland and Aquatic Research Center (hereafter WARC), and the other is adjacent to the National Oceanic and Atmospheric Administration Estuarine Habitats and Coastal Fisheries Center (hereafter EHCFC). The ponds adjacent to the WARC (named the National Wetlands Research Center until 1 October 2015) and the EHCFC were built in 1992 and 1999, respectively, when the buildings were constructed (Pham et al., 2007). The two ponds are about 30 m from each other at their closest points, but do not share a water connection and are separated by two vehicle entranceways. The Lafayette region is on the eastern edge of historical coastal prairie habitat, but before the ponds and buildings were constructed the sites were open fields with shallow seasonal wetlands (Pham et al., 2007). A sidewalk, four-lane divided road, and a medium-density housing complex are located to the west of the ponds (Fig. 1).

Surface water area of the two ponds is similar (WARC = 0.21 ha, EHCFC = 0.25 ha), and was calculated using Google Earth imagery in a geographic information system (ArcMap 10.8, ESRI). Both ponds receive their water from rainwater and separate well systems. Unlike the EHCFC pond, the WARC pond was built with a liner and has a greater variety of depths than the EHCFC pond. Whereas the EHCFC pond bottom is relatively flat, the WARC pond edges are generally sloped, with thinner, v-shaped passages. Since this study began in 2009, the EHCFC pond has remained open water, whereas the WARC pond is often largely or completely choked with vegetation. In March 2011, much of the vegetation in and around the WARC pond was removed, but the vegetation-choked state soon returned and continues to present day. Whereas the WARC pond bottom has always been very soft, likely because of the amount of decaying vegetation sitting on top of the liner, the EHCFC pond was relatively firm in early years of this study, but recently has become softer and filled with sediment. The deepest part of the WARC pond is approximately 2 m, but most of the pond is much shallower, with many areas less than 1.0 m. The majority of the EHCFC pond ranges from 1.0–2.0 m in depth.

We captured turtles for 5 days each May from 2009 to 2021. We used two primary methods to capture turtles: hoop traps (Lagler, 1943) and deep-water crawfish nets (Glorioso and Niemiller, 2006). Hoop traps are a passive capture technique, and therefore each year we set hoop traps on Sunday afternoon, checked them each weekday afternoon, and picked them up on Friday afternoon, giving five consecutive trap nights. Deep-water crawfish nets are an active capture technique, and each year we set crawfish nets and checked them every 20 min between 1700 and 1900 h (i.e., six total checks of all nets) for five consecutive trapping sessions with rare exceptions, usually because of inclement weather. In 2020, inclement weather forced us to reschedule our Thursday crawfish net sampling for Saturday, the only instance of five nonconsecutive trapping days in the study. The number and type of hoop traps used each year varied, generally with more traps used in later years. Total crawfish nets typically varied from 10 to 14 nets, sometimes spread equally between the ponds, but often with more in the WARC pond. We baited hoop traps predominantly with fish, whereas we baited crawfish nets with either raw poultry or beef spleen on all but 1 day. For specific information on the number and type of traps and baits used each year, see Glorioso (2022).

In 2009, we set hoop nets on Sunday afternoon and active crawfish net sampling followed that evening. We checked hoop nets Monday–Friday, but sampled using crawfish nets from Sunday to Thursday. In 2009, we released captured turtles after processing all hoop net captures before beginning crawfish net sampling. Because of the 2009 protocol, it was possible that we could catch a turtle captured in a hoop net again in a crawfish net on the same day. We changed the protocol after 2009, and from 2010 to 2021 we set hoop nets on Sunday afternoon, and checked them on Monday afternoon, followed by the first crawfish net sampling. We held all turtles captured with hoop nets until after crawfish net sampling, releasing all turtles from both capture methods at the end of crawfish net sampling into their pond of capture so that the same turtle could never be captured twice on the same day.

We identified each captured turtle to species and individually marked them with a unique three-letter identification by filing the marginal scutes (Cagle, 1939; Dorcas, 2005). We modified the marking scheme for Eastern Musk Turtles (Sternotherus odoratus) and Eastern Mud Turtles (Kinosternon subrubrum) to compensate for the reduced number of marginal scutes in kinosternid turtles. In some of the smallest individuals we still made marks by using nail clippers, but we also injected an 8.2-mm passive integrated transponder tag (PIT, model HPT8; 134.2 kHz, FDX) under the left bridge of the turtle into the inguinal cavity out of concern the marks would be unreadable if not recaptured for multiple years (Buhlmann and Tuberville, 1998).

We determined the sex of turtles by examining secondary sexual characteristics such as elongated foreclaws and tails in males. We took standard measurements of all turtles, including straight-line carapace length (SCL) along the midline, straight-line plastron length along the midline, maximum width, and maximum height using tree calipers to the nearest millimeter. We measured body mass to the nearest gram using a digital scale. We attempted to age turtles by counting lines of arrested growth (annuli) on the plastron when possible. We typically used the abdominal scutes to count annuli, but sometimes used other plastral scutes if they were more apparent. We photographed the carapace and plastron of each individual captured each year to assist in identification discrepancies and aging turtles.

We summarized data on captures of uniquely marked turtles into capture history matrices for the 13 annual sampling occasions 2009–2021. For survival analysis, the matrix showed whether each individual was captured (1) or not (0) each year, whereas for growth analysis, the matrix showed the carapace length if the turtle was captured and NA (missing data) otherwise. We modeled overall survival for the two species with >50 individuals captured and modeled growth for the one species with >50 individuals measured on two or more occasions.

We used the live recaptures procedure in Program MARK (White and Burnham, 1999) to model survival using the Cormack–Jolly–Seber open population mark–recapture model. We fitted four alternative models that varied according to whether survival and recapture probabilities were either constant or fully time dependent. We conducted model selection based on Akaike’s information criterion (AIC; Burnham and Anderson, 2002).

We modeled growth using a sex-specific hierarchical version of the von Bertalanffy growth model (Armstrong and Brooks, 2013) fitted using OpenBUGS 3.2.3 (Spiegelhalter et al. 2014). Under this model, the growth rate and asymptotic size parameters not only differ between sexes, but can also vary among individuals. We found that we needed to use mildly informative priors (Banner et al., 2020) for individual standard deviation in asymptotic size and mean asymptotic size of males and females to obtain sensible results. However, the use of informative priors only means that parameters were constrained to plausible ranges.

We used the growth model to generate posterior distributions for ages of individuals where age was not inferred from annuli. Therefore, we used Bayes’ theorem to infer each turtle’s most likely initial age based on its size when first measured, its subsequent growth rate, and the prior expectation based on the age distribution of the population (Armstrong and Brooks, 2014). For the prior age distribution, we used a negative binomial distribution based on the constant annual survival rate estimated from this study, that is, NB(1, 1 − S) where S is the annual survival probability. We used the cut function in OpenBUGS to prevent this age estimation from affecting growth parameters, similar to Armstrong and Brooks (2014).

We had 574 total captures of 251 individuals of five species from 2009 to 2021 (Table 1). The most abundant species was Trachemys scripta elegans (Red-Eared Sliders), comprising 74.7% of all captures and 67.7% of all individuals, followed by Sternotherus odoratus (Eastern Musk Turtles) with 20.0% of all captures and 23.1% of all individuals. Interestingly, we captured only 28 individual S. odoratus in the first 12 yr combined, but captured 30 new individuals in 2021. We captured Chelydra serpentina (Common Snapping Turtles) and Kinosternon subrubrum hippocrepis (Mississippi Mud Turtles) infrequently, along with a single Sternotherus carinatus (Razor-Backed Musk Turtle).

We had more recaptured individual T. scripta from previous years than new individuals in 8 of 12 yr. We captured individual T. scripta marked the first year of sampling in every subsequent year and likewise for S. odoratus, except for 2012, when we did not capture any S. odoratus (Tables S1 and S2). We captured more turtles using crawfish nets than hoop traps, especially considering we only trapped with crawfish nets for 2 h daily, whereas hoop traps were continuously set for 5 days (Table S3).

The constant survival and time-varying capture probability model was supported with >96.0% of the AIC weight for T. scripta, whereas the constant survival and constant capture probability model was supported with >99.9% of the AIC weight for S. odoratus. Trachemys scripta had an apparent annual survival of 0.79, whereas S. odoratus had an apparent annual survival of 0.89 (Table 2).

Mean asymptotic SCL for male and female T. scripta was 165.3 and 227.1 mm, respectively (Fig. 2). Growth rates were similar between the sexes until about 5 yr of age, when male growth rate slowed compared to females. Males were estimated to grow less than 1 cm per year after 15 yr, whereas this was after 24 yr in females. Most individuals of unknown age were larger individuals, and models suggested all but one of these females were over 10 yr old, with the largest (242 mm SCL) estimated to be over 29 yr old (Fig. 3). Age estimates ranged up to 21 yr old for our largest unknown-age male (190 mm SCL).

Of 429 T. scripta captures, we captured 226 in the WARC pond and 203 in the EHCFC pond, with a similar distribution of individuals and recaptured individuals (Table S4). We captured only 3 of 115 total captures of 58 individual S. odoratus in the EHCFC pond. The 14 individual C. serpentina were evenly split between the two ponds, with recaptures only in the EHCFC pond. We captured seven of eight individual K. subrubrum from the WARC pond, and we captured the lone S. carinatus in the EHCFC pond.

The SCL of T. scripta at first capture in the WARC pond was significantly smaller than the mean SCL in the EHCFC pond (χ²₁ = 4.662, P = 0.031; Table S5). We captured over 75% of T. scripta individuals <100 mm SCL in the WARC pond. In contrast, the EHCFC pond had not had a new individual <100 mm SCL since 2015 until a single capture in 2021. Over 96% of S. odoratus individuals, a small species, was captured in the WARC pond as well as the two smallest C. serpentina, and the only ones <150 mm SCL.

Of the 93 T. scripta we recaptured at least once, 78.5% remained in the same pond at each capture, with 45.2% and 33.3% remaining in the WARC and EHCFC ponds, respectively. We recaptured 15 (16.1%) T. scripta in a different pond than their original capture. Based on our captures, five individual T. scripta (5.4%) moved from their original pond to the other pond, and then back again to the original pond.

The apparent annual survival of 0.79 for T. scripta in this study aligns closely with other studies of T. scripta from both natural and altered environments. In natural Carolina bays in South Carolina, Frazer et al. (1990) noted survival rates of 0.81 for females and 0.84 for males in 4+-yr-old turtles. At three locations in a Superfund site (i.e., contaminated with hazardous materials) in Oklahoma, Hays and McBee (2010) noted survival rates between 0.74 and 0.86. At five sites in metro Charlotte, North Carolina, Eskew et al. (2010) noted survival rates between 0.73 and 0.93. Lastly, in a northeast Mississippi farm pond, Parker (1996) found survival rates between 0.71 and 0.86.

The apparent annual survival of 0.89 for S. odoratus in this study is slightly higher than 0.84–0.86 reported by Mitchell (1988) in Virginia, and much higher than the 0.46 and 0.68 reported by Hrycyshyn (2007) for females and males, respectively, in Florida. High apparent annual survival at this site is likely because of the seemingly static and ideal habitat for the species within the WARC pond, where nearly all the population occurs. Also, the pond’s isolation from other suitable areas likely limits or eliminates migration to and from the pond complex in this highly aquatic species. In addition, though there are avian and mammalian predators in the area, there are no large aquatic predators such as Alligator mississippiensis (American Alligators) in the pond complex. One female S. odoratus captured as an adult (94 mm SCL) in our first year in 2009 has been captured in every subsequent year in the 13-yr study except 2012, when zero S. odoratus were captured. This female was captured multiple times within the trapping week in 6 of 13 yr. The individual’s SCL has been 99–100 mm since 2013, and her body mass has fluctuated between 160–180 g over the last 9 yr.

Many studies have used annuli to estimate turtle ages, but the validity of this method has been questioned (Wilson et al., 2003). However, for T. scripta in the temperate zone, as in our study, both Cagle (1946) and Gibbons (1970) directly confirmed a one-to-one relationship between a major growth period and the deposition of one annulus per year. Stone and Babb (2005) tested the annual growth line hypothesis in T. scripta in Oklahoma and found that turtles were reliably aged to about 4 yr of age. In addition, Stone and Babb (2005) found that the annuli count of 100 of 106 (94%) T. scripta was consistent with their known age. In our study, 58 of 65 (89.2%) T. scripta assigned an age based on annuli had an estimated age of 4 yr or less. Our result aligns with those of Sexton (1959) and Gibbons (1983), who stated that there is a direct correlation between known age in years and major scute growth zones (i.e., annuli) in subadult and immature turtles, but growth slows or ceases at maturity, making it difficult to recognize annuli, particularly in turtle species that shed their scutes.

In addition to the five captured species, we observed two other turtle species in the ponds. On several occasions both during and outside of trapping weeks we and other co-workers spotted at least two different large female Apalone spinifera (Spiny Softshells) in the EHCFC pond. Apalone spinifera are often trapped in baited hoop nets, but they have eluded capture thus far in this study. In 2016 we confirmed a hatchling Pseudemys concinna (River Cooter) in the WARC pond, and have since seen at least one large adult female basking in the WARC pond. Pseudemys concinna is primarily herbivorous and not frequently trapped using baited hoop traps and crawfish nets (Dreslik, 1997).

Though we are uncertain whether turtles naturally colonized these wetlands over the years, we have anecdotal evidence of release of turtles into the ponds. We witnessed someone attempting to release a box turtle (Terrapene sp.) during our sampling, but we intervened by picking up the box turtle before it entered the pond. We suspect the lone S. carinatus we captured was a released turtle, as these ponds are not a typical habitat for this species. We were also made aware of an instance in 2013 where a person released an adult male Chrysemys dorsalis (Southern Painted Turtle) into the EHCFC pond.

It is apparent that the WARC pond is preferable to small size classes of turtles. Over 67% of T. scripta <100 mm SCL in the WARC pond were first captured in the last 5 yr, indicating either more successful reproduction in recent years or a better ability to capture younger size classes. For T. scripta, the young turtles likely prefer the WARC pond because of its abundant emergent and floating vegetation, which offers both a food source and protection from predators. For S. odoratus and C. serpentina, the thick layer of soft mud and detritus at the bottom of the WARC pond is likely preferable for these bottom dwellers compared to the firmer EHCFC pond with less organic matter.

Our recapture data indicate that most turtles prefer one pond or the other under the assumption turtles are not moving between ponds at other times of year outside the May capture period. The two ponds are only 30 m from each other, but are separated by two vehicle entranceways. The entranceways have high vertical curbs that may serve as barriers to some turtles. We have on occasion observed turtles struggling to get over the curb, and it is possible the curbs could direct them toward the four-lane road and put them in danger. That said, we opportunistically documented few instances of direct mortality at any time of year as we frequently come and go from our offices and observe the roads for live or dead turtles. It is clear, however, that at least some turtles successfully move between ponds.

The urban greenspace of our study site seems to be good habitat for the turtles that occur there, with survival estimates for our two most abundant species similar or higher than those reported by other studies. Trachemys scripta seems to be especially prolific, with successful annual nesting despite the human-altered landscape around the ponds. Despite capturing S. odoratus in all but 1 yr, we rarely captured juveniles and were concerned about their recruitment. Then, in 2021, we doubled the number of S. odoratus individuals captured, with most being juveniles. As our study is ongoing, more precision is expected with our T. scripta growth and survival estimates as we follow some of our known-age turtles into adulthood and larger sizes. In addition, we may be able to perform a growth model for S. odoratus provided we continue to catch juveniles like we did in 2021 and recapture them in subsequent years.

The authors thank L. Muse, J. Hefner, C. Hillard, R. Kidder, C. Eaglestone, L. Elston, S. Godfrey, N. Jennings, B. Maldonado, P. Vanbergen, A. Hartmann, S. Jones, T. Thigpen, G. Griffard, T. Glorioso, and M. Baldwin for field assistance. They are grateful to the directors of both facilities who allow this work each year. Animals were captured under Louisiana Department of Wildlife and Fisheries Scientific Collecting Permits issued annually. All handling of animals was conducted in accordance with approved U.S. Geological Survey (USGS) IACUC protocol FY2008-1 and updated USGS/WARC/GNV 2019-07. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. This is contribution number 884 of the USGS Amphibian Research and Monitoring Initiative (ARMI).

Supplementary data associated with this article can be found online at http://dx.doi.org/10.1670/22-083.S1.