Species richness may be similar across an ecosystem; however, discrete patterns of species abundance may characterize assemblages within the same ecosystem (DonnerWright et al. 1999; Moll and Moll 2004; Bluett et al. 2013). The dynamics of an assemblage are influenced by the ecological demands of specific taxa and environmental processes operating at geographic, regional, local, and temporal scales (Kennard et al. 2007; Bluett et al. 2011; Habel et al. 2016). However, at the broader scale, the extent to which changes in local assemblages contribute to global biodiversity loss is poorly understood (Dornelas et al. 2014). Environmental modification is one process that is driving changes to the composition and relative abundance of species within an assemblage (Vandewalle and Christiansen 1996; Smith et al. 2006; Habel et al. 2016). This process may cause shifts in abundance, resulting in either a positive or negative effect upon single or multiple species (Moll and Moll 2004; Browne and Hecnar 2007; Devictor et al. 2008). Thus, comparative studies of the composition of assemblages will allow for an evaluation of potential responses of taxa and assemblages to environmental changes over time. Implementation of conservation and management decisions are typically at the scale of local or regional ecosystems, thus knowledge of change within species assemblages may assist in evaluating the efficacy of conservation strategies and informing policy (Dornelas et al. 2014; Habel et al. 2016).

The specialist–generalist concept can be useful in identifying winners and losers in the current biodiversity crisis (Devictor et al. 2008; Clavel et al. 2011). Ecological generalist species are adaptable and extremely resilient and may thrive in human-altered environments, in some cases even more than they do under natural conditions (Eskew et al. 2010; Roe et al. 2011, Concepción et al. 2015). On the other hand, the more specialized a species, the more negative its spatial response to landscape fragmentation and disturbance, and thus the more prone to extirpation from anthropogenic factors (Kennett and Tory 1996; Devictor et al. 2008; Tucker et al. 2012; Concepción et al. 2015; Habel et al. 2016). Thus, assessment of the relative abundance of each may provide insight into degradation of the habitat or ecosystem (Clavel et al. 2011).

Freshwater turtles are present in many ecosystems throughout the tropic, subtropic, and temperate zones, where they occur in assemblages of up to 17 species (Moll and Moll 2004). Like other species (Wilson and Lawler 2008; Habel et al. 2016), turtles can be grouped according to their ecological tolerance and degree of specialization (Moll and Moll 2004). The relative abundance of generalists and specialists within a chelonian assemblage may shift in response to habitat alteration (Moll and Moll 2004; Browne and Hecnar 2007). For example, the loss of lotic habitat within impoundment waters was found to have favored Krefft's turtle (Emydura macquarii krefftii), a generalist, which had a greater relative abundance within impounded waters than the more specialized white-throated snapping turtle (Elseya albagula) and saw-shelled turtle (Myuchelys latisternum; Tucker 2000; Hamann et al. 2007; Tucker et al. 2012). Hence, abundance of a single species alone may not indicate an ecologically healthy ecosystem, as habitat alteration may not affect all turtle species equally (Moll and Moll 2000; Devictor et al. 2008; Dornelas et al. 2014).

Time-series monitoring, a tenet of adaptive management, is critical for understanding trends in riverine turtle assemblages, as the effects of environmental changes on long-lived species may take years to detect and overcome. Thus, given the length of time required for a change to become evident, it is critical that capture protocols are repeatable across time in order to detect temporal changes in species diversity and assemblage composition (Gibbons et al. 1997; Wallace et al. 2007; Dornelas et al. 2014; Habel et al. 2016). It is critical for management of species and their habitat, that trends in relative abundance of species within assemblages through space and time are identified. Assessing seasonal, temporal, and spatial variation in assemblages of cryptic and elusive species, such as freshwater turtles, can be challenging because of variable patterns in seasonal behavior and environmental conditions. These variables will determine selection of appropriate capture technique.

Multiple techniques are used to capture freshwater turtles. The biases inherent within each technique may lead to misrepresentation of certain species relative to their actual abundance (Breen and Ruetz 2006; Hubert et al. 2012; Tesche and Hodges 2015). For example, local environmental conditions can affect each method, which will thus vary in effectiveness based on the biology of available species (Cagle and Chaney 1950; Frazer et al. 1990; Gamble 2006; Wallace et al. 2007). Factors such as species biology, operator proficiency, and environmental conditions make it difficult to compare results across the various trapping methods, because the variation in the probability of capture among species (Cagle and Chaney 1950; Gibbons et al. 1997). However, when sampling protocols and methods are standardized within and among studies, these biases remain constant, and trends in turtle assemblages can be detected through space and time (Hill 2005).

Often, turtle surveys are carried out in an opportunistic, unsystematic manner with a focus on maximizing the capture rate (Sterrett et al. 2010). We would argue that, in those instances, the number of turtles captured is a snapshot measure of capture efficiency, which may be appropriate for presence/absence studies, but not for comparative studies. When dissimilar capture techniques are employed, the results can be skewed in comparative population studies (Gibbons et al. 1997).

Six species of freshwater turtles inhabit the Mary River, Queensland (Elusor macrurus, Elseya albagula, Emydura macquarii krefftii, Myuchelys latisternum, Chelodina expansa, and Chelodina longicollis). It is known that 3 of these species exist at a higher abundance than the other 3 species, and that all species have been captured throughout the length of the river (Limpus 2008). It is not known, however, if the turtle species assemblage significantly varies throughout the river as a result of local environmental conditions. In order to assess if there was spatial variation in turtle species assemblage throughout the river, we used a passive capture methodology that ensured that the probability of capturing individuals of a specific species remained constant over space and time.

METHODS

Study Site

Our study area encompassed approximately 180 km of the Mary River (Queensland, Australia), from Cambroon in the south to Tiaro in the north (Fig. 1). This subtropical river flows in a northerly direction for approximately 300 km. The majority of stream flow occurs throughout the austral summer, and early autumn months of January to April. While the occurrence and intensity of rainfall is irregular, the flow during the dry-season period of July to October is relatively stable (Pusey et al. 2004; Queensland Department of Natural Resources and Mines 2017). The low level of flow regulation and in-stream infrastructures within the main channel pose minimal obstruction to movement. The most significant in-stream barrier is a 2.9-m tidal barrage located 37 km downstream of the study area.

View Figure

• Download as Powerpoint
• Download Figure

The Mary River catchment supports one of the highest diversities (6 species) and endemicities (2 species) of freshwater turtles in Australia (Fig. 2; Cann 1998; Limpus 2008). The Mary River turtle (Elusor macrurus) and the white-throated snapping turtle (Elseya albagula) are river specialists (Flakus 2002; Thomson et al. 2006). Krefft's turtle (Emydura macquarii krefftii), a generalist, does not show any preference and can be found in lentic and lotic environments and man-made pools (Tucker et al. 2012). The saw-shelled turtle (Myuchelys latisternum) prefers the upper reaches and side channels of larger rivers (Freeman and Cann 2014). The broad-shelled turtle (Chelodina expansa) and the snake-necked turtle (Chelodina longicollis) have preference for shallow, ephemeral wetlands often remote from a river (Kennett et al. 2009).

View Figure

• Download as Powerpoint
• Download Figure

The study was confined to the freshwater lotic reaches of the river with the downstream study reach located upstream of the impounded waters. Four study reaches were selected, “upper”, “middle”, “mid-low”, and “lower”, with 60-km intervals between each consecutive reach. Reaches were characterized according to environmental features. The upper river reach (adopted middle thread distance [AMTD] 264 km from river mouth) is characterized by a generally narrow valley, low to moderate channel sinuosity, pools and runs interspersed by long reaches where the transient channel moves across a shallow braided bed (Brizga et al. 2004). The middle reach (AMTD 208 km) consists of highly sinuous meandering units, composed of longer pools intercepted with substantial lengths of riffle and glide habitat. The mid-low (AMTD 143 km) and lower (AMTD 86 km) reaches are typified by reduced stream slope, a highly sinuous channel, and long deep pools with occasional riffles. The streambed material was composed of sand, gravel, cobblestones, or bedrock and varied between sites. Five sampling sites, approximately 1 km apart, were located within each study reach. Accessibility, landholder agreement, and geomorphology dictated the final selection of each sampling site.

Sampling Methodology

A passive sampling technique, fyke netting, was used for this study (Fig. 3). Set nets, such as fyke nets, are commonly used as a standardized sampling method when monitoring fish and turtle populations (Vogt 1980; Hubert et al. 2012). Passive sampling involves the capture of the target species by a device that is not actively moved by humans or machines while the organisms are being captured, thus facilitating standardization of sampling gear (Hubert et al. 2012). Fyke nets are similar to a cylindrical fish trap (i.e., hoop net) that has a series of internal, funnel-shaped openings that enable turtles and fish to enter and minimize escapes. Unlike hoop nets, fyke nets have single, double, or even triple leaders that assist in guiding the animal into the hoop sections. They have been successfully used in large rivers, creeks, fast-moving water, small ponds, and large lakes (Vogt 1980; Breen and Ruetz 2006; Wallace et al. 2007; Larocque et al. 2012).

View Figure

• Download as Powerpoint
• Download Figure

The net dimensions, mesh size, size of mouth, and funnel openings were made specifically to capture a wide range of turtle sizes, ranging from hatchlings (35 mm) to adults (436 mm). The cylindrical section of the net was 4 m in length (from mouth to cod end) and included 4 × 0.9-m-diameter aluminium hoops with 2 internal funnels, each having a fixed opening of 0.4 m. The entire fyke net was covered with 20-mm (stretched) mesh made from 2-mm braided cord. A polystyrene float was placed in each compartment to provide space for captured turtles to surface for air (Larocque et al. 2012). The cod end of the net was secured to a 1.8-m metal post as an additional measure to maintain access to the surface for trapped animals. Two leaders extended from either side of the external opening of the hoop net to form a “V” (Fig. 3a). Each leader was 10 m long with a drop of 1.2 m. The ends of each leader were secured with metal posts, set approximately 7–10 m apart, with the end of 1 leader set on the water's edge. A float line ran along the top of each leader. To ensure nets sat firmly on the riverbed, 1-m lengths of 8-mm galvanized chains were randomly clipped along the mouth and bottom of both leaders to reduce the possibility of individuals passing beneath the net (Fig. 3b).

The nets were set facing upstream parallel to the riverbank within the vicinity of a riffle. Here the water flow and depth were conducive to the physical dimensions of the set nets. Sampling sites were excluded from below riffles due to higher flow rates, which created an unsafe environment for researchers. To standardize sampling protocols, each net was set in the same location and for the same time period on subsequent sampling episodes. The nets were not baited and sampling relied on the turtles encountering and entering the net. Each morning, each net was checked, the bycatch released, litter removed, and the turtles processed.

Sampling occurred in 2 discrete seasons: austral spring (September–October) and autumn (March–April) over a 2-yr period (2015–2016). Throughout these seasons, there is a reduced likelihood of flood events, thus minimizing the risk of a sampling episode being compromised by a rise in the river height. Nonetheless, minor flooding occurred in April–May 2015, which delayed the completion of initial autumn sampling until June 2015. Five nets were set over 4 consecutive nights within each of the 4 sampling reaches (a total of 320 trap nights, i.e., 20 nets set within each of the 4 sampling reaches over 4 sampling periods).

Turtle Processing

Species were identified by head and plastron characteristics (Cann 1998). Sex was determined by dimorphic tail sizes in all species (McDiarmid et al. 2012). Although carapace length is not an absolute indicator of reproductive status, assigning a constant straight carapace length (SCL; minimum) allowed for individuals of all species to be assigned an age class. Sex of individuals of all species with an SCL of < 150 mm could not be determined with confidence; thus all were considered juveniles, with no discrimination between species (Thomson et al. 2006). The number of male, female, and juvenile turtles of each species captured within each reach was recorded.

Data Analysis

A nonparametric multivariate analysis of variance, PERMANOVA—R (version 3.4.1) package Vegan—was conducted to assess for the influence of river reach, season, and year, as well as the interactions between these variables (Anderson 2001; R Development Core Team 2016). The model was run using 200 random permutations. Post hoc pair-wise comparisons were run using an F-test. The Hellinger transformation was used to offset the zeros in the model before applying the Bray-Curtis index. Dissimilarity matrices for the PERMANOVA were calculated using the Bray-Curtis index on transformed (square root) abundance values. Rarely captured species were removed from the PERMANOVA analysis (Clarke and Gorley 2006).

RESULTS

A total of 782 individuals were captured during the 2 yrs of sampling. Elusor macrurus was most frequently trapped, followed by E. albagula, E. m. krefftii, M. latisternum, C. expansa, and C. longicollis (Table 1). Although the number of trapped turtles across each of the 4 reaches was similar, species relative abundance varied between study locations (Fig. 4). In the lower reach, the abundance of E. albagula was greatest but their numbers gradually decreased moving upstream. The abundance of E. m. krefftii and E. macrurus both peaked within the middle reach, with a reduction in abundance up- and downstream of the middle reach. These species were at their smallest abundance in the lower reach. The upper reach had highest number of M. latisternum.

Table 1. Comparison of relative species abundance from 3 studies conducted on the Mary River (Queensland, Australia) that used different survey methods.

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

Three significantly discrete turtle assemblages were identified over the 4 sampled reaches using PERMANOVA (Tables 2 and 3). The assemblages within the lower and mid-low reaches were similar (F_3,80 = 1.6, p = 0.19), but significantly different from the middle reach (lower reach: F_3,80 = 11.34, p < 0.01; mid-low: F_3,80 = 4.03, p < 0.01), and the upper reach was significantly different from the other sampled reaches (lower: F_3,80 = 9.02, p < 0.01; mid-low: F_3,80 = 4.73, p < 0.01; middle: F_3,80 = 3.04, p < 0.05). The assemblages were not significantly altered by either the season or the year of capture (Season: F_1,80 = 1.05, p = 0.38; year: F_1,80 = 1.77, p = 0.17).

Table 2. Results from PERMANOVA pair-wise comparisons to test for significant differences in turtle assemblages between river reaches.

View Fullscreen

Table 3. Results from 3-way PERMANOVA test showing the influence of covariables on the turtle assemblage.

View Fullscreen

Juveniles were 6% of the overall number of captured turtles (Fig. 5). The upper reach yielded 67% (n = 33), while the lower reach contained only 4% (n = 2) of juveniles captured. Four species (E. macrurus, E. albagula, E. m. krefftii, M. latisternum) were represented in the upper reach, 3 species each in the middle (E. macrurus, E. m. krefftii, M. latisternum) and mid-low reaches (E. macrurus, E. albagula, E. m. krefftii), with 2 species (E. albagula, M. latisternum) found in the lower reach. However, an insufficient total number of juveniles were captured within each reach to undertake statistical analysis of species assemblage differences.

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

This study revealed significant changes in the relative abundance of the 6 turtle species along the length of the Mary River, in Queensland, Australia. Although all species were captured throughout the river, different reaches of the river served as hot spots for particular species. The spatial variation in species assemblages was consistent through time, illustrating the robustness of the sampling method for defining freshwater turtle assemblages. Only a few studies have previously investigated riverine turtle assemblages along a river gradient (DonnerWright et al. 1999; Bluett et al. 2013), but we argue that this current method is a quick and effective method to assess turtle community health. Other studies have investigated the relationship between environmental gradients (DonnerWright et al. 1999) and stream order (Bluett et al. 2013) on species richness or abundance. Like Bluett et al. (2013), we showed that turtle assemblages vary along the river continuum.

Although it was beyond the scope of the study to identify the ecological processes (e.g., competition or specific environmental gradients) that are driving the composition of assemblages (like DonnerWright et al. 1999), our study showed an association between variation in abundance of specific species and broad geomorphological features, such as stream reaches. Vannote et al. (1980) found that the abundance of numerous aquatic organisms, as well as community assemblages, vary predictably along the entire length of river systems. For example, a longitudinal study of the Meuse River in Europe found a gradual shift from a macroinvertebrate assemblage dominated by insects to a community dominated by crustaceans and molluscs (Usseglio-Polatera and Beisel 2002). Our study also detected variation in species assemblages along the river gradient and the significance of specific reaches for individual species. For example, E. albagula was most numerous in the lower reach, suggesting the significance of this reach for this species. Importantly, this spatial difference in assemblage was consistent among seasons and years.

This spatial variation of species abundance varied predictably within the known habitat requirements of individual species. It is indicative of variation in stream morphology, species niche requirements, anthropogenic influences, and exotic competitors (Morrison et al. 2006). Elseya albagula tends to prefer slow moving deep pools, and feeds on filamentous algae and crustaceans foraged from the muddy vegetated shallow margins of deep-water pools (Micheli-Campbell et al. 2017). Hence, it was expected that this species would be more abundant in the lower and mid-low reaches, where the pools are deeper and longer. In contrast, E. macrurus has a significantly larger, linear home range, frequents riffle zones, and prefers different food items, such as bivalves, gastropods, and aquatic insects that are found within rocky riffles (Micheli-Campbell et al. 2017). The abundance of this species in the middle reach suggests that this reach has the most appropriate ratio of pool–riffle sequences, and therefore more food sources. Myuchelys latisternum prefers deep to shallow pools typified in the upper reaches and side channels of rivers, and is chiefly carnivorous, occasionally feeding upon vegetative material (Freeman and Cann 2014). Accordingly, this species was most abundant in the upper reaches. Emydura macquarii krefftii is known to inhabit a wide range of natural and man-made water bodies, and has an omnivorous diet consisting of filamentous algae, sponges, and terrestrial insects (Hamann et al. 2008; Limpus 2008; Wilson and Lawler 2008). This is the only generalist species of the Mary River. The abundance of this species followed a similar distribution pattern to E. macrurus, and this was unexpected given it has the least habitat specialization. It was anticipated that given its generalist nature, the relative abundance of E. m. krefftii would be consistently high across all reaches. This suggests that the quality of the habitat has not declined to such an extent to select this species over the more specialized ones. The two Chelodina species (i.e., C. expansa and C. longicollis) prefer lentic habitats, and thus the low capture rates observed were expected (Kennett et al. 2009; Bower and Hodges 2014), although their nondetection within a reach does not imply they were genuinely absent.

Our methodology yielded different relative abundances of individual species when compared to previous studies that employed alternative sampling techniques (i.e., snorkeling, dip netting, and visual surface sightings; Table 1). Elusor macrurus was the most abundant species in our study (0.38 across reaches); however, a snorkeling study reported a relative abundance of only 0.29 (Thomas 2007; Table 1), while a study utilizing multiple capture methods (snorkeling, dip netting, puddling, and visual surface sightings) reported a relative abundance of only 0.18 (Limpus 2008; Table 1). While this comparison may suggest the influence of capture methods on detection probability, the variation in capture methods and sampling protocols precludes rigorous comparisons. The snorkeling study was a single-episode methodology across multiple seasons, and presumably capture method was affected by operator and local river conditions (Thomas 2007). For example, the abundance of a highly mobile species like E. macrurus may have been underestimated due to their visibility being inhibited by presence of dense beds of macrophytes, turbidity, and the size and depth of pools, as well as difficulty in hand capturing (Thomas 2007), whereas the abundance of a less mobile species, such as E. albagula, may have been underestimated in our study as it is more likely to be captured by snorkeling, but less likely to be captured in a set net. The habitat generalist, E. m. krefftii, was most abundant when multiple techniques were used: snorkeling, cathedral traps (telescoping, vertical, cylindrical nets), seine and dip nests, and muddling (Limpus 2008); thus using a single capture method, such as in the present study, may have underestimated its abundance.

Most juveniles were captured in the upper reach. Other studies have also found the upper reaches to yield the highest capture rate of juvenile turtles (DonnerWright et al. 1999). The proportion of juvenile turtles was low for all species (varied from 0.02 to 0.13), with exception of M. latisternum (0.37). Typically, very few studies have captured juvenile turtles in high numbers in the water (Hamann et al. 2008; Pike et al. 2008; Tesche and Hodges 2015). This likely reflects the population dynamic of freshwater turtles in general, but also may be due to young turtles being cryptic and less mobile (Micheli-Campbell et al. 2013, 2017). Previous snorkeling studies in the Mary River captured a greater proportion of juvenile turtles (0.125; Thomas 2007) than were captured in the present study (0.062). While we accept that differences in capture method may be partially responsible for the disparity, the 50% decline is worrisome and worthy of future investigation.

Survey results were consistent between yearly sampling episodes. Life history characteristics of freshwater turtles, such as longevity, minimal migration, and a limited home range would have minimized variation. This consistency of results is expected as biases remained constant due to repetition of sampling methods, protocols, and locations.

CONCLUSIONS

The objective of the present study was to assess the relative species abundance over space and time, and thus identify longitudinal variation in turtle assemblages in the Mary River. While we agree that the results of our survey may not significantly reflect the absolute species abundance, it consistently identified discrete turtle species assemblages. Our intent was not to conduct an absolute abundance study, where the results may lead to inferences about the population status of multiple species. Thus, it was not critical that the most effective trapping method for individual species was employed. The advantage of our technique and protocol is that future studies will be able to identically replicate our sampling methodology, and thus identify trends in relative species abundance. These trends may be used to alert natural resource managers to shifts in conditions and ecological health of the river. The Mary River catchment has been affected by human activities (such as gold mining, vegetation clearing, sand and gravel extraction, water extraction, and introduction of exotic plants and animals) for over 150 yrs (Brizga et al. 2004). Hence, this current assemblage data set is unlikely to coincide with pre-European settlement assemblages, but rather reflects individual species responses to habitat modification.