Spatial Ecology and Population Genetics of Northern Map Turtles (Graptemys geographica) in Fragmented and Continuous Habitats in Canada
Abstract
In riverine systems, the construction of locks and dams may present barriers to movement for aquatic organisms and effectively fragment otherwise continuous populations. We examined the spatial ecology and population genetics of northern map turtles (Graptemys geographica) in the Trent-Severn Waterway (TSW), Ontario, Canada. We hypothesized that the locks and dams on the TSW are fragmenting map turtle habitats and creating barriers to movement, and predicted that turtles from fragmented habitats have significantly reduced home range sizes and daily movements compared to turtles from unfragmented habitats, and that populations on either side of a lock and dam would be genetically distinct. Home ranges and average daily movements were smaller for turtles from fragmented habitats compared with continuous habitats, indicating that locks and dams are restricting mobility. However, populations in fragmented habitats showed none of the predicted genetic consequences of fragmentation (e.g., reduced heterozygosity and allelic richness, differentiation across a barrier). Genetic data may be reflecting historic population structure because only a few turtle generations have passed since the construction of the locks and dams about 100 years ago. Additionally, our genetic data may not be sufficiently robust to detect differentiation at this relatively small geographic scale. Our study highlights the conservation importance of examining multiple facets of a single potential threat, especially for long-lived species that may show differences in short- and long-term effects of fragmentation.
Habitat fragmentation is one of several leading causes of reptile declines worldwide (Gibbons et al. 2000; Klemens 2000). In a freshwater ecosystem, habitat fragmentation most often refers to the loss of connectivity, generally caused by the construction of dams that obstruct waterways (Morita and Yamamoto 2002; Xie 2003; Hanfling and Weetman 2006). Fragmentation causes contiguous populations to be split into smaller, more isolated units that occur in habitat patches. A decrease in population size can reduce the genetic variation of the population, and an increase in isolation reduces the influx of novel genetic material from nearby populations (Lande 1976; Frankham et al. 2002; Toro and Caballero 2005). An overall decrease in genetic variation reduces both individual fitness and the ability of a population to adapt to environmental changes (Avise 1995; Templeton et al. 2001; Reed and Frankham 2003; DiBattista 2008). Conservation strategies for populations living in fragmented areas must prioritize the maintenance of genetic variation, if those populations are to persist for the long term (Friar et al. 2000; Kuo and Janzen 2004).
Connectivity among habitat fragments is critical in the maintenance of genetic variation and the conservation of endangered species (Howeth et al. 2008; Neel 2008). As individuals migrate between habitat patches, their genetic material (assuming successful reproduction in the new population) is incorporated into the receiving population's gene pool. For a highly aquatic species, such as the northern map turtle (Graptemys geographica), movements between habitat patches are made largely by water (Roche 2002) and the construction of aquatic barriers, such as dams, could isolate populations. However, unlike fish, turtles are capable of making terrestrial movements, and therefore aquatic barriers cannot by default be considered impermeable to these species. It is therefore necessary to determine whether dams are creating barriers to movements for aquatic turtle species, and to determine the level of connectivity between fragmented aquatic habitats in order to implement the best conservation strategies for aquatic turtles.
In navigable waterways, elevation changes and areas of otherwise unnavigable waters are made passable by the construction of a lock. Locks provide passage to boats moving through the waterway by allowing periodic connections from one section of the waterway to the next. Reid (2005) demonstrated that redhorse (Moxostoma spp.) fish can successfully move through multiple locks during a single season, allowing movements between areas otherwise fragmented by a dam. The dispersal capacity of redhorse (upwards of 22 km; Hackney et al. 1968) is greater than that of most aquatic turtles; however, the ability of a lock to facilitate passage between habitat patches must be considered in terms of its potential conservation application for aquatic species other than fish.
A study of spatial ecology alone gives researchers a current snapshot of what individual turtles are doing at a single moment in time. Population genetics studies, on the other hand, may reflect historical processes and miss current trends. By combining multiple methods of study, conservationists and managers can make informed decisions about the health and status of populations. We examined both the spatial ecology and population genetics of northern map turtles in fragmented (areas with multiple locks and dams) and continuous habitats on the Trent-Severn Waterway (TSW), Ontario, Canada.
The TSW is a 386-km navigational route that links Lake Ontario to Georgian Bay. It is comprised of 44 locks and over 100 dams and control structures (Angus 1998). Construction on the waterway was started in the early 1800s, and the waterway officially opened in 1920 (Angus 1998). The generation time of the northern map turtle is estimated at 20–25 years, with age at maturity likely occurring around 13–14 years (Roche 2002; Ernst and Lovich 2009), hence 3 or 4 generations of turtles have passed since the construction of the locks and dams. A significant reduction in fine-scale genetic structure has been noted over a period of 24 years in female red deer, Cervus elaphus (Nussey et al. 2005), a species with a generation time of approximately 7 years (Clutton-Brock et al. 1982). It is therefore not unreasonable to expect significant changes in the genetic structure of subpopulations even over a relatively short period of time. In addition, Carlsson et al. (1999) found that brown trout (Salmo trutta) showed significant genetic structuring over a relatively short (3 km) section of stream with no discernable physical barriers to movement; they hypothesized that historical geographic barriers had fixed specific spawning grounds for subpopulations of brown trout, regardless of whether those barriers continued to exist into the present day (Carlsson et al. 1999). Given that the TSW lock system was constructed to bypass navigational barriers, such as waterfalls and rapids, it is possible that genetic structuring along the waterway may represent historical barriers that are presently exacerbated by the construction of the locks and dams.
We hypothesized that the locks and dams on the TSW are fragmenting map turtle habitats and creating barriers to movement, and predicted that turtles from fragmented habitats have significantly reduced home range sizes and daily movements compared with turtles from unfragmented habitats. We also predicted that 1) genetic diversity is lower in fragmented reaches of the waterway compared to continuous reaches (unfragmented controls); 2) there is little to no gene flow between sections separated by a lock and dam; and 3) turtles from either side of a lock and dam are genetically distinct. The ultimate goal of this population genetics analysis was to understand the impacts habitat fragmentation may be having on the health of map turtle populations, and to determine whether map turtles on the TSW should be considered for conservation purposes as a single unit or whether there are distinct populations (i.e., designatable units, Green 2005) with unique conservation requirements (Moritz 1994; Palsboll et al. 2007).
METHODS
Surveys
Trapping and capture surveys were performed on the TSW from Peterborough (lat 44°12′N, long 78°24′W) to Trenton (lat 44°06′N, long 77°30′W), Ontario, from May to August 2007 and 2008. Three study locations were chosen along the waterway: Trenton (Locks 1–3) and Ranney Reach (Locks 9–11 and 12) were designated as fragmented study sites, and Otonabee (Rice Lake – Lock 19), an unfragmented site (Fig. 1). The fragmented sites were characterized by the presence of locks and dams spaced approximately 2 km apart. Turtles were captured with basking traps, by hand while snorkelling, or with dip nets. Basking traps were constructed using 1-cm copper tubing and a small seine net (Lucky Strike Bait Works Ltd, Peterborough, ON, Canada). The seine net was woven around the tubing, which was then formed into a circle. Polystyrene was attached to the tubing for floatation. The trap was then placed beneath logs and branches where turtles had been observed basking. Depending upon the situation, the trap was secured with fishing line, bungee cords, or staples. Dip nets had 1.8 m of extendable handle, and 33 by 38 cm baskets.
Citation: Chelonian Conservation and Biology 9, 2; 10.2744/CCB-0824.1
Radiotelemetry
Radiotransmitters (Holohil Systems Ltd, Carp, ON, Canada; Advanced Telemetry Systems [ATS], Inc, Isanti, MN, USA) were attached to turtles greater than 200 g in mass, so that the combined mass of the radio and attachment agents was less than 10% of the mass of the turtle. Holohil radios were attached to the turtles by drilling 2 small holes in the marginal scutes along the posterior carapace, and securing the radio with copper wire. Plumber's epoxy putty was then placed over the copper wire and around the radio for extra security and to streamline the radio onto the shell. ATS radios were attached by drying the carapace of the turtle completely, abrading a small section of the posterior carapace with fine grain sandpaper, and applying a 2-part 5-minute epoxy in thin layers to both the radio and the shell. Plumber's epoxy putty was then used to streamline the radio and for extra security.
A total of 19 female map turtles were outfitted with radios, 10 from the Trenton fragmented site and 9 from the Otonabee unfragmented site (Ranney Reach was not used for the telemetry study; see Fig. 1). Because male turtles from fragmented habitats had an average mass below the 200-g threshold dictated by our animal care protocols, only females were used in the spatial ecology study. Turtles were located 2 to 3 times per week using a Telonics receiver (Mesa, AZ, USA) and a 3-element Yagi antenna (Wildlife Materials International, Murphysboro, IL, USA), depending upon the weather, from time of first capture until the end of August in each of the 2 field seasons (2007 and 2008). Turtle location coordinates were determined using a handheld GPS receiver (Garmin eTrex Legend) to at least 6-m accuracy. Visual confirmation of the turtle was sought in all cases, although this was not possible when the turtle was in deeper water, and location was fixed by canoeing over the approximate position.
Of the 19 turtles outfitted with transmitters, we had sufficient location data (a minimum of 10 radio location points each) from 16 individuals for the spatial analyses (9 from Trenton and 7 from Otonabee). Nine of the 19 radioed turtles went missing for some period of time during the study (Table 1). The cause of these disappearances is generally unknown, although in at least 2 instances the radio had failed, and in another case the turtle had become trapped but was recovered alive in a water treatment facility near Trenton, Ontario. We suspect that at least 1 turtle went missing as a result of becoming trapped against the grate of a hydroelectric dam inflow, although this could not be confirmed. The most likely reason for radio loss is battery failure or movements of a turtle beyond the expected home range (turtles moved unexpectedly long distances at times, making a thorough search logistically difficult in unfragmented habitats). Map turtles shed their scutes continuously throughout the year, so there is little reason for concern about turtles retaining dead radios past the completion of this study because they will eventually shed them.
Spatial Ecology Analyses
Average daily movements (ADM) were calculated as the straight-line distance between consecutive radio-location points divided by the number of days between them, averaged for each season. Seasons were defined as Spring (May), Nesting (June), and Summer (July and August). We compared seasonal daily movements between females from unfragmented (N = 7) and fragmented (N = 9) sites using a 2-factor ANOVA. Comparisons were made using only 7 turtles from the unfragmented site and 9 turtles from the fragmented site because of the loss of some turtles (Table 1) before sufficient movement data could be obtained.
Because map turtles are highly aquatic, movements within the home range (excluding nesting excursions) are likely to be by water. Therefore, we used range length, defined as the shortest distance by water between the 2 farthest location points, as a proxy for home range. Range lengths were measured using ArcGIS 9.2 (ESRI 2006). We compared home range lengths (Student's t-test) between fragmented and unfragmented sites. Range length was estimated only for turtles with location data spanning at least 2 months (N > 10 radio location points) to ensure that at least half of the active season was included. Range lengths and daily movements were log-transformed to meet the assumption of normal data distributions.
Tissue Collection for Population Genetics
For the genetic analyses, trapping and capture surveys were expanded to include another fragmented site, Ranney Reach (Locks 9–11 and 12; Fig. 1), for a total of 2 fragmented sites and 1 unfragmented site. Fragmented sites were further subdivided to test for gene flow around a barrier. Trenton was divided into Sidney (Locks 1–2) and Glen Miller (Locks 2–3) reaches, while Ranney Reach was divided into Hague's Reach (Locks 9–10) and Campbellford Reach (Locks 10–11 and 12; Fig. 1). Tail tips were collected from captured turtles using sanitized nail scissors and were stored in individual vials in 95% ethanol at −20°C. A total of 109 turtles were sampled from all study sites (Sidney: N = 25, Glen Miller: N = 25, Hague's: N = 11, Campbellford: N = 12, and Otonabee: N = 36).
DNA Extraction and Amplification
DNA extraction was performed using a DNeasy blood and tissue kit (Qiagen) following their protocols, with the exception that tissue was digested overnight (instead of 2–3 hours as directed). Tissue was digested overnight because the hard epidermis of turtle tail tissue was more difficult for the proteinase K to digest than mammalian samples. It was also necessary to finely chop up the tail tip in order for the digestion agents to fully penetrate the sample.
Ten primers, designed for bog turtle (Glyptemys muhlenbergii) microsatellite loci by King and Julian (2004), were tested to determine their ability to successfully amplify our map turtle microsatellites. Amplification and specificity of polymerase chain reaction (PCR) products were assessed with 1% agarose gel electrophoresis. Gels were stained with ethidium bromide for visualization under UV light and photographed using an Alpha Innotech FluorChem 8900. Specificity was assessed visually by the clarity of the bands (fuzziness indicates low precision), and by the number of bands seen (multiple bands indicate primers are attaching at more than just the target site on the DNA). Genetic structure was then assessed with the 5 microsatellite loci that amplified with adequate specificity (GmuD16, GmuD87, GmuD90, GmuD93, and GmuD121; Table 2). PCR amplifications were set up in 10-µl volumes, each containing 100–250 ng of DNA, 1× reaction buffer, 25 mM MgCl2, 10 mM dNTP, 10 mM of each primer, and 1.5 U Taq polymerase. All forward primers were labelled with a fluorescent dye (FAM, HEX, or TET). Amplification was performed in an Eppendorf Mastercycler, programmed with an initial step of 94°C for 2 minutes, followed by 33 cycles of 94°C for 45 seconds, 58°C for 45 seconds (annealing), and 72°C for 90 seconds (elongation). A final extension step of 72°C for 5 minutes was performed to ensure completion of amplification. The PCR products were then separated and visualized on a 3730 DNA Analyzer (Applied Biosystems) using a GeneScan 500 Rox size standard (Applied Biosystems) at MOBIXLab (Hamilton, ON, Canada). The alleles were then sized with Peak Scanner version 1.0 (Applied Biosystems).
Population Genetics Analyses
Standard measures of genetic diversity (observed heterozygosity, expected heterozygosity, and allelic richness [average number of alleles per locus]) were calculated for each population using GENEPOP 4.0 (Rousset 2008). Because of the variation in sample sizes per population, a regression analysis was used to test whether sample size had a significant effect on the number of alleles and the observed heterozygosity using STATISTICA 6.1 (StatSoft 2002). Mann-Whitney U-tests were used to determine if there were significant differences in measurements of genetic diversity between fragmented and unfragmented populations. Significance was set at α = 0.05.
Tests for deviations from Hardy-Weinberg equilibrium and for linkage disequilibrium were done in GENEPOP 4.0 (Rousset 2008). Presence of null alleles was tested using Micro-Checker (van Oosterhout et al. 2004). Population differentiation was measured in Arlequin 2.000 (Schneider et al. 2000) using pairwise FST values (Weir and Cockerham 1984) for each pair of populations. Only 4 loci were used in this analysis, since one locus (GmuD87) did not meet the assumption of Hardy-Weinberg equilibrium. Significance of FST values was determined by the proportion of 10,000 permutations leading to an FST value larger or equal to the observed value (Rousset 2008), with α = 0.05. To reduce the amount of type I error, a sequential Bonferroni correction was applied to the test for significance of FST. However, this application greatly reduces statistical power; therefore, following Mockford et al. (2007), we report both corrected and uncorrected values. Isolation by distance (IBD) was tested by measuring the correlation between genetic distance (FST) and log geographic distance (Rousset 2008). Geographic distance was defined as the distance by water between 2 sites, measured in ArcGIS 9.2 (ESRI 2006). We also measured the correlation between genetic distance and number of barriers (locks and dams). Significance of correlations was determined using a Mantel test implemented in the Isolation by Distance, Web Service (Jensen et al. 2005).
The program STRUCTURE (Pritchard et al. 2000) was used to determine the most likely number of populations of map turtles in the total stretch of the TSW from Trenton to Peterborough, Ontario. STRUCTURE employs a Bayesian methodology to determine the level of genetic structure in the data set independent of sampling location. To estimate the number of populations (K), 5 independent runs of K = 1–8 were carried out at 500,000 Markov chain Monte Carlo repetitions following a burn-in of 500,000 repetitions. Again, Locus GmuD87 was excluded from this analysis as it did not meet the assumption of Hardy-Weinberg equilibrium. The most probable value of K was determined as the value of K that showed the maximum increase in likelihood, Ln[Pr(X | K)], over successive increases in K (Evanno et al. 2005; Davis et al. 2008).
RESULTS
Spatial Ecology
Females in fragmented habitats had significantly smaller average home ranges (1.53 ± 0.31 km) than those in unfragmented habitats (8.51 ± 1.59 km; t16 = 6.2, p < 0.001). Both season (F2,39 = 4.4, p < 0.05) and habitat category (i.e., fragmented versus unfragmented; F1,39 = 17.6, p < 0.0005) had significant effects on daily movements of female map turtles. A Tukey HSD post hoc test revealed that summer movements by females in fragmented habitats (ADM = 76.1 ± 10.2 m/d) were significantly smaller than those made by their unfragmented counterparts (ADM = 277.6 ± 50.0 m/d; Fig. 2).
Citation: Chelonian Conservation and Biology 9, 2; 10.2744/CCB-0824.1
Population Genetics
All 5 microsatellite loci were polymorphic, with the number of alleles per locus ranging from 2 to 9. Neither number of alleles (r2 = 0.024, F2,3 = 0.45, p = 0.51) nor observed heterozygosity (r2 = 0.0089, F2,3 = 0.16, p = 0.69) were correlated with sample size. No significant presence of null alleles was detected at any locus. Habitat fragmentation did not have an effect on allelic richness or heterozygosity (Table 3).
No significant linkage disequilibrium was detected. Departures from Hardy-Weinberg equilibrium were found in 2 populations (Otonabee and Ranney) after sequential Bonferroni correction for multiple comparisons, both at locus GmuD87. Pairwise FST estimates ranged from −0.039 to 0.044 (Table 4). Four out of 10 pairwise population comparisons were statistically significant prior to Bonferroni correction; with the correction, 1 of 10 remained significant (Table 4). Pairwise FST values were lowest between sections separated by a lock and dam (−0.039 between Hague's and Campbellford, 0.010 between Sidney and Glen Miller; Table 4). A significant pattern of isolation by distance was found using a Mantel test with 10,000 randomizations (r = 0.65, Z = 0.33, p < 0.05). Genetic differentiation was not correlated with the number of barriers between sites (r = 0.45, Z = 1.60, p = 0.099). For the STRUCTURE analysis, no value of K showed an increase in likelihood above K = 1, indicating that the model that best explains the genetic data from the sampled turtles is a single, panmictic population (Fig. 3).
![Figure 3. Average cluster membership (mean Q max, where Q is the proportion of ancestry for each individual assigned to a putative population; dark triangles; y-axis on the left) and likelihood {Ln[Pr(X | K)]; open circles; y-axis on the right} plot for STRUCTURE analysis of northern map turtles (Graptemys geographica) on the Trent-Severn Waterway, Ontario, Canada using 5 runs at each K from 1 to 8. Ln[Pr(X | K)] is the natural log of the posterior probability (Pr) of the data (X) for a given number of clusters (K). A high mean Q max value (> 0.8) indicates that a large proportion of the individual turtles were assigned successfully to a single putative population, and a similar likelihood among all 5 runs indicates a greater precision in the model's ability to assign individuals to a particular population.](/view/journals/ccab/9/2/i1071-8443-9-2-185-f03.png)
![Figure 3. Average cluster membership (mean Q max, where Q is the proportion of ancestry for each individual assigned to a putative population; dark triangles; y-axis on the left) and likelihood {Ln[Pr(X | K)]; open circles; y-axis on the right} plot for STRUCTURE analysis of northern map turtles (Graptemys geographica) on the Trent-Severn Waterway, Ontario, Canada using 5 runs at each K from 1 to 8. Ln[Pr(X | K)] is the natural log of the posterior probability (Pr) of the data (X) for a given number of clusters (K). A high mean Q max value (> 0.8) indicates that a large proportion of the individual turtles were assigned successfully to a single putative population, and a similar likelihood among all 5 runs indicates a greater precision in the model's ability to assign individuals to a particular population.](/view/journals/ccab/9/2/full-i1071-8443-9-2-185-f03.png)
![Figure 3. Average cluster membership (mean Q max, where Q is the proportion of ancestry for each individual assigned to a putative population; dark triangles; y-axis on the left) and likelihood {Ln[Pr(X | K)]; open circles; y-axis on the right} plot for STRUCTURE analysis of northern map turtles (Graptemys geographica) on the Trent-Severn Waterway, Ontario, Canada using 5 runs at each K from 1 to 8. Ln[Pr(X | K)] is the natural log of the posterior probability (Pr) of the data (X) for a given number of clusters (K). A high mean Q max value (> 0.8) indicates that a large proportion of the individual turtles were assigned successfully to a single putative population, and a similar likelihood among all 5 runs indicates a greater precision in the model's ability to assign individuals to a particular population.](/view/journals/ccab/9/2/inline-i1071-8443-9-2-185-f03.png)
Citation: Chelonian Conservation and Biology 9, 2; 10.2744/CCB-0824.1
DISCUSSION
Spatial Ecology
The locks and dams on the TSW are restricting movements for northern map turtles. Average daily movements and home range lengths were significantly smaller in areas bounded by locks and dams compared to those for turtles from the continuous habitat. Therefore, the locks and dams on the TSW appear to have fragmented northern map turtle habitats.
Home range is defined as the area used by an individual organism to survive and reproduce (Burt 1943). Relatively small home ranges can be the result of 1 of 2 causes: either the individual happens to be residing in a particularly rich habitat matrix where all of its survival and reproductive needs are met in a small area (Beasley et al. 2007), or barriers are preventing its movements throughout the matrix (Dyer et al. 2002). In the latter case, the organism must make do with a reduction in habitat quality, which will likely reduce the health, longevity, and potential fitness of the organism (Martin-Smith and Laird 1998; Pozo-Montuy and Serio-Silva 2007).
The overall impact of a reduced average home range length for map turtles on the TSW is difficult to assess given the variation in reported range lengths for female map turtles across the species' range. Range lengths vary considerably, from 1.2 km (Pluto and Bellis 1988) to 12.5 km (Flaherty 1982). For populations living on lakes or larger river systems, home ranges are sometimes reported as areas (ha) and not range lengths (see Gordon and MacCulloch 1980; Carrière 2007), rendering a comparison to the TSW range lengths difficult. Even if a paper reports range length, the variety of methods used to obtain this estimation (visual recapture vs. radiotelemetry; straight line vs. distance by water) makes comparison difficult. Although the mean home range length for female map turtles in the fragmented study site (1.5 km) was not outside those reported for other populations of map turtles, it was significantly smaller than that for the turtles in the control site on the TSW.
Increasing connectivity between habitat patches will allow map turtles to move between them, reducing the likelihood of mortality from a lack of accessible resources. On the TSW, although the dams and locks restrict movements, they do not appear to present impermeable barriers to movements by turtles. Three of 12 marked turtles from the fragmented site, 2 females and 1 male, were found in the next lock section upstream from where they were originally caught. In a related study, we noted average home range of male map turtles in unfragmented habitats was 6.18 ± 1.50 km, and while males were not included in this analysis, it is likely that their movements are reduced in fragmented areas as well (Bennett 2009). It is possible in the cases of the females that some of these movements were made overland during nesting; however, we suspect that at least 2 were made through locks.
A female from the section between locks 2 and 3 was found swimming near lock 3 on 19 June 2008. By 27 June, she had moved into the section between locks 3 and 4. She stayed in the lock 3 channel, an area bounded by concrete walls around 1 m in height, for about 2 weeks, before she was rediscovered between locks 2 and 3 on 10 July 2008. She had sustained injury to the posterior part of her plastron, possibly caused by being caught between the lock gates, or against the hull of a boat (the injury was consistent with being crushed against something solid). The lockmaster had also reported seeing a turtle with a radiotransmitter floating in the lock itself around this time. Therefore, we conclude that this female had passed through a lock at least once, if not twice, during her trips between sections.
Population Genetics
Northern map turtles on the TSW showed none of the predicted genetic consequences of habitat fragmentation, despite the fact that the spatial analyses indicated that the dams presented barriers to movements. Similar levels of genetic diversity were observed between unfragmented and fragmented populations, and turtles on either side of a putative barrier (i.e., a dam) showed no genetic differentiation. STRUCTURE analysis revealed that the best explanation of the genetic variation in the sampled turtles is a single, panmictic population between Peterborough and Trenton, Ontario, a finding that does not support our hypothesis that locks and dams are currently detectable barriers to gene flow in this population of map turtles.
Genetic diversity is reduced by fragmentation by the combination of a decrease in population size (Lacy 1987; Frankham 1996; Tallmon et al. 2002) and an increase in isolation among fragments (Frankham et al. 2002; Anderson et al. 2004). The impact of genetic drift increases with the number of generations since fragmentation (Steinberg and Jordan 1997). Therefore, the lack of expected effects of fragmentation (loss of genetic variation, low gene flow) found in our population could be the result of a large population size (which reduces the effects of genetic drift), a short time since fragmentation, or effective migration through the locks reducing the potential for inbreeding, or any combination of these factors.
Map turtle population sizes were larger than expected on the TSW (Bennett et al. 2009), and fragmentation by the locks and dams has only been present for approximately the last 100 years (Angus 1998). The age at which map turtles become sexually mature is unknown, although it has been estimated to be 13–14 years (Ernst and Lovich 2009). The generation time of other freshwater turtles at northern latitudes ranges from 11 years (Chrysemys picta, Turnquist and Langdon 2007) to 25 years (Chelydra serpentina, Congdon et al. 1994; Glyptemys insculpta, Tessier et al. 2005; Clemmys guttata, COSEWIC 2004) or 31 years (COSEWIC 2008). Given the estimated age at maturity of map turtles, the generation time is likely to be about 25 years, which is in keeping with the estimate (> 20 years) put forth by Roche (2002). Therefore, only 3 to 4 generations have passed since the completion of the construction of the locks. Unexpectedly large population sizes, coupled with a relatively short time since fragmentation, may be limiting our ability to detect recent barriers to gene flow.
Alternatively, locks and dams may not present impermeable barriers to map turtles because we have evidence that turtles migrate around or through the locks. As stated above, over the course of this study, there were 4 observed incidents of turtles moving through or around a lock and dam. In 3 cases, movement was most likely made through the lock itself when the gates were opened to allow boats to move through. At times, turtles appear to congregate by the lock gates, and there have been reports of turtles eating zebra mussels (Dreissena polymorpha) directly off the gates themselves (A. Kidd, pers. comm., 2008). It is possible for a turtle to swim into an open gate and be accidentally passed through the locks. As well, lock masters have reported seeing map turtles in the lock proper, supporting the observation that turtles will pass through with boats. Finally, during nesting season, females will move onto land to lay their eggs, and it is possible that a turtle or its hatchlings might return up or downstream of the lock section from where they originated.
Significant differentiation, after Bonferroni correction, was found between 1 population pair: Otonabee (unfragmented) and Glen Miller (fragmented), 2 of the most distant study sites on the waterway. IBD is the pattern expected among geographically separated subpopulations, with individuals from more distant areas being more genetically distinct (Wright 1943). When considering the genetic structure of turtle populations in different river and stream drainages, IBD is used as a null hypothesis (Souza et al. 2002; Spinks and Shaffer 2005). For example, Souza et al. (2002) found that genetic variation in Hydromedusa maximiliani (Brazilian snake-necked turtle) was more correlated with river systems than with geographical distance. Souza et al. (2002) suggested, therefore, that each river system be considered a separate management unit for the conservation of H. maximiliani. If locks and dams were fragmenting map turtle populations on the TSW, we would have expected genetic distance to be more correlated with the number of locks and dams, rather than geographic distance. Because the IBD model was supported by our data, and genetic distance was not correlated with the number of locks and dams, we conclude that the locks and dams on the TSW do not appear to have had an impact on the genetic structure of the resident map turtle population, although this does not preclude the detection of future effects when a greater number of generations have passed. Because turtles are long-lived (Roche 2002; Litzgus 2006; COSEWIC 2008) contemporary impacts (such as those observed in our spatial analyses) may not manifest at the genetic level for several decades or centuries.
The population pairs with the lowest FST values were those separated by a single lock and dam, and populations on either side of a barrier were not genetically distinct. As discussed previously, this lack of an expected effect can be explained by the combination of long generation time coupled with a relatively short time since fragmentation. Current FST values may therefore reflect historic population structure rather than current patterns. Alternatively or additionally, locks and dams may not be impermeable barriers. The level of gene flow observed between turtles separated by a lock and dam may be attributable to these types of movements, in turn supporting the idea that the locks play a key role in connecting fragmented aquatic habitats, which has been demonstrated in a fish species, the redhorse (Moxostoma spp., Reid 2005). However, because of the relatively short time since fragmentation (ca. 100 years), we cannot infer that migration is occurring based solely on the genetic data.
Another possible explanation for the lack of genetic structure detected in the TSW population of map turtles are the low levels of polymorphism and the number of microsatellite loci used for the analyses. Alleles per locus ranged from 2 to 9, which is low compared to some population genetics studies on turtles (6–21 alleles per locus for Glyptemys insculpta, Tessier et al. 2005), but comparable to other studies (5–13 for Malaclemys terrapin, Hauswaldt and Glenn 2005; 1–11 for Emydoidea blandingii, Mockford et al. 2007). Many studies employ microsatellites as a molecular marker in part because of their high variability, which allows for a finer geographic scale of analysis (Banks et al. 2005; Mockford et al. 2007; Davis et al. 2008). Turtle DNA, however, is notoriously conservative (Avise et al. 1992; Amato et al. 2007), making population genetics analyses difficult on small spatial and short temporal scales. Low levels of polymorphism reduce the amount of genetic information available to test hypotheses, resulting in a coarse rather than fine-scale analysis. It is therefore possible that the combination of these factors (few loci, low polymorphism) has resulted in a false negative in our study, especially in light of the spatial ecology findings that showed restricted movements in the fragmented sites. Future work should consider the applicability of other highly variable molecular markers such as single nucleotide polymorphisms and restriction fragment length polymorphisms, developing markers designed specifically for Graptemys, and focus on obtaining large (N > 20) sample sizes from all suspected subpopulations.
Conclusion
While spatial ecology data gathered using radiotelemetry clearly support the hypothesis that locks and dams are fragmenting map turtle habitats on the TSW and thus restricting movements, population genetics analyses failed to indicate any of the negative consequences expected in fragmented populations. However, effects of short-term perturbations to populations of long-lived organisms may be overlooked if the investigation techniques do not take temporal scale into account. For example, contemporary impacts may take multiple generations to manifest at the genetic level. Management agencies should therefore consider multiple lines of investigation when creating conservation protocols for long-lived species such as turtles. Although map turtles appear to be doing well on the TSW, ongoing changes to the waterway, such as a decrease in the number of lockages per season (Panel on the Future of the Trent-Severn Waterway 2007), represent potential concerns for future generations. Because turtles are able use the locks as travel corridors between reaches, fewer lockages will mean less gene flow. While we concede that our population genetics analyses would benefit from a larger dataset, there are very few published articles on the genetics of map turtles (Graptemys spp.), and none that examine exclusively the northern map turtle. Future work should focus on finer scale genetic analyses in order to gain a better baseline understanding of aquatic turtle population genetic structure along waterways.
Study sites used for assessment of northern map turtle (Graptemys geographica) movement patterns and population genetics on the Trent-Severn Waterway, Ontario, Canada. The fragmented study sites (Trenton and Ranney Reach) are circled, and the unfragmented control site (Otonabee) is bounded by a rectangle. Fragmented sites were further subdivided to test for gene flow around a barrier. Trenton was divided into Sidney (Locks 1–2) and Glen Miller (Locks 2–3) reaches, while Ranney Reach was divided into Hague's Reach (Locks 9–10) and Campbellford Reach (Locks 10–11 and 12). White chevrons represent the locations of locks and dams.
Average daily distances moved (ADM; ± standard error) of female northern map turtles (Graptemys geographica) by season from Otonabee (unfragmented, N = 9) and Trenton (fragmented, N = 9) sites on the Trent-Severn Waterway, Ontario, Canada. ADMs were calculated as the straight line distance between consecutive radio-location points divided by the number of intervening days. Seasons were defined as Spring (May), Nesting (June), and Summer (July and August).
Average cluster membership (mean Q max, where Q is the proportion of ancestry for each individual assigned to a putative population; dark triangles; y-axis on the left) and likelihood {Ln[Pr(X | K)]; open circles; y-axis on the right} plot for STRUCTURE analysis of northern map turtles (Graptemys geographica) on the Trent-Severn Waterway, Ontario, Canada using 5 runs at each K from 1 to 8. Ln[Pr(X | K)] is the natural log of the posterior probability (Pr) of the data (X) for a given number of clusters (K). A high mean Q max value (> 0.8) indicates that a large proportion of the individual turtles were assigned successfully to a single putative population, and a similar likelihood among all 5 runs indicates a greater precision in the model's ability to assign individuals to a particular population.
