Where Lakes Were Once Rivers: Contrasts of Freshwater Turtle Diets in Dams and Rivers of Southeastern Queensland
Abstract
We contrasted diets of three turtles (Elseya albagula, Myuchelys latisternum, Emydura krefftii) from free-flowing or impounded rivers in southeastern Queensland, Australia, to evaluate the effects of flow regulation. The turtle species encompassed the herbivorous, carnivorous, and omnivorous feeding guilds, respectively. The study design simultaneously considered ontogenetic dietary shifts and seasonal effects on prey availability. Relative to the river samples, diets for three turtles in impoundments were substantially reduced in prey abundance, species richness, and dietary breadth. Turtles with narrower dietary preferences in free-flowing rivers were more affected by dams than was the dietary generalist. For the most omnivorous turtle, qualitative differences were evident in diets from impoundments and rivers; yet considering its greater dietary breadth, quantitative effects on diet were minor. Diets of turtles in impoundments included fewer subaquatic plants and wind fallen fruits than did the diets of turtles in rivers. For the largely carnivorous turtle, fewer aquatic invertebrates were ingested in impoundments than in riverine habitats, and scavenging behavior was noted mainly in impoundments. Multidimensional scaling of the site characteristics identified dams or weirs with back-ups exceeding 20 km as being relatively similar in impacts on prey diversity. Canonical correspondence analysis identified factors of habitat alteration and turtle size as major determinants of underlying variance in the diets. The results suggest that turtle food webs are altered by river regulation. A general finding that turtle diets in impoundments were depleted of aquatic plants or macro-invertebrates pertains to other turtles of conservation importance.
River systems can be monitored reliably by biotic indicators that identify functional and structural integrity (Norris and Thoms 1999). Flow-regulated rivers show declines in aquatic complexity and biodiversity when river currents are replaced by still water conditions (Poff and Allan 1995; Abramovitz 1996; Ward 1998). The River Continuum Concept (VanNote et al. 1980) summarizes that flow-regulated rivers typically undergo increased turbidity and sedimentation, salinization, oxygen and thermal stratification, loss of riparian vegetation or aquatic macrophytes, dampened flow regimes, and modified thermal conditions (Calow and Petts 1994). Numerous before–after studies of flow-regulation effects in aquatic invertebrates and fish communities document declines in aquatic biodiversity (Doeg et al. 1987; Pardo et al. 1998; Gehrke et al. 1999; Gehrke and Harris 2001; Petr and Swar 2002; Lessard and Hayes 2003; Hayes et al. 2006; Käiro et al. 2011) or degraded biotic integrity (Davis and Simon 1995; Karr and Chu 1999). Such findings may also be expected for other aquatic vertebrates under conditions of flow regulation. Even so, comparative studies on the concurrent effects of dams on multiple river systems and species are rare.
Riverine communities contain ecological generalist and specialist turtle species; yet freshwater turtles are usually omitted or coarsely lumped as tertiary species in aquatic trophic webs (see Cousins 1996; Winemiller 1996). Such simplistic treatment of turtles as a single trophic group is untenable for modern community food web analysis, if for example 23 turtle species occupy northeast India (Baruah and Sharma 2010) and 17 turtle species coexist in the Mississippi River (Moll and Moll 2004). Dietary studies with Australian turtles include habitats as diverse as farm ponds, ephemeral wetlands, perennial rivers, seasonally fluctuating rivers, and oligotrophic lakes (Legler 1976; Legler and Cann 1980; Burbidge 1981; Georges 1982; Chessman 1983, 1984, 1986; Georges et al. 1986; Georges and Kennett 1989; Cann and Legler 1994; Kennett and Tory 1996; Welsh 1999; Flakus 2002). However, there are no studies that evaluate how a community of turtle species responds to successional changes of aquatic food webs from riverine to lake conditions (Bodie 2001). Stated simply, what do river turtles eat after a river is dammed? Turtle species with narrow dietary niches should be strongly responsive to quantitative changes in food availability whether from natural temporal variability or from human-induced impacts (Allanson and Georges 1999). With turtles that are intimately linked with specific habitat types, anthropogenic changes to river habitats may generate related changes of prey availability (Dodd 1990).
We investigated these general principles in 3 sympatric turtles in 4 rivers of southeastern Queensland. We quantified the foods ingested by these turtles concurrently in different habitats (river or lake) and independent of influences of ontogeny or season. We investigated the effect of dams as a quasi-experimental design by inferring from multiple sites on rivers as control replicates and impounded lakes as unique treatments of variable magnitude. The approach substitutes space for time to deal with the impossibility of identical replicate sites for dams or rivers yet acknowledges that key environmental factors that operate at landscape scales may influence prey assemblages (Wang et al. 2003).
The study has three central questions related to prey assemblages ingested by turtles. First, are dietary effects of habitat modification (i.e., river impoundments) independent of inherent environmental flux or functional trophic differences among turtle species? If so, then conclusions about habitat alteration can be developed independently of temporal or ontogenetic influences on diets. Second, which dietary components are more affected by an altered flow regime? We test a prediction that dietary specialists are affected by changes of resource availability to a greater extent than generalists (Kennett and Tory 1996) in contrasts of chelid turtles having different foraging guilds. Third, we ask whether findings about flow-related changes on turtle diets are relevant to other turtle species of conservation concern. If so, then dietary preferences of turtles may identify critical habitat associations that warrant management under environmental flows or suggest tradeoffs when river habitats are impounded.
METHODS
Study Species
The Fitzroy, Kolan, Burnett, and Mary rivers of southeastern Queensland contain a host of existing or proposed water storages (Table 1) and the highest diversity of turtle species in Queensland (Georges and Adams 1996). Of seven species within the region, we focused on 3 common species (Emydura krefftii, Elseya albagula, and Myuchelys latisternum) found in both rivers or impoundments that early studies had respectively characterized as primarily omnivorous, herbivorous, or carnivorous feeding guilds (Legler 1976; Legler and Cann 1980; Georges 1982; Kennett and Tory 1996; Cann 1998; Allanson and Georges 1999; Armstrong and Booth 2005). The provisional feeding guilds (sensu Moll and Moll 2004) of earlier studies were retained as a working model, even though a species may occasionally consume almost every category of prey. Four other species were encountered during surveys but excluded from the present study because a habitat contrast was lacking (Chelodina longicollis and Chelodina expansa were rarely found in rivers or impoundments) or their endemic status to a single river precluded a cross-rivers comparison (Elusor macrurus and Rheodytes leukops). Descriptions and maps of all sampled sites are given elsewhere (Tucker 2000; Hamann et al. 2007).
Field Methods
Turtles were collected along 1–2-km reaches of river or lake sites during a regional assessment on effects of dams and weirs on turtle diversity (Tucker 2000). Capture methods included snorkelling, dip netting, seining, and baited traps. Each turtle was measured to the nearest 0.1 mm (straight carapace length) with calipers and weighed to the nearest 1 g on a digital balance. Turtles were marked individually with a numbered tag in the rear foot webbing and by a notched code on the marginal scutes. All turtles were returned to the water within a few hours after processing.
The 3 species attain different maximum sizes and degree of sexual size dimorphism. Size was entered as a continuous variable in the analysis, and opportunistic samples included small, medium, and large individuals of each species to represent immatures, males, and females, respectively. Individuals were sexed as males or females by their dimorphic tail sizes (Reed and Tucker 2012) or classed as immature if below a threshold at which sexual dimorphism was displayed in that species (Tucker 2000).
Experimental Design
The study has unbalanced replication of impoundments as a treatment factor (either lake or river sites) across 4 rivers as experimental units. Impoundments were created by weirs (< 5 m wall height, top spilling) or dams (> 5 m wall height, bottom spilling). Sample sites were spatially stratified to include several fluvial and impounded sections in the upper, middle, and lower portions of each river. We sampled 26 impounded lakes and 28 river sites for a representative set of sites that covered 40%–100% of the impoundments on each of the four rivers (Tucker 2000). Queensland Water Resources provided data on each dam's height, date of completion, the length of river back-up, and full-storage capacity.
We sampled year-round (September 1997 to October 1999) to account for temporal variability. The region's wet season extends from November to April and the dry season from May to October. The subtropical region includes 189,374 km2 of drainage, ranging from 5–400 m elevation and covers latitudes from 23–28°S. Western upland areas of the catchments typically receive 56–76 cm of annual rainfall, and eastern coastal areas collect 100–150 cm of annual rainfall. Thermal conditions varied across catchments from the upper eastern slopes of Great Dividing Range down to the coastal plains. Representative climatic data (Australian Bureau of Meteorology [www.bom.gov.au]) for upland areas are Injune (−25.84, 148.57; mean monthly maximum temperatures 20.0°–33.6°C; mean monthly minimum temperatures 3.2°–19.6°C) and for coastal areas are Rockhampton (−23.40, 150.50; mean monthly maximum temperatures 23.1°–32.5°C; mean monthly minimum temperatures 10.5°–22.5°C). At all collection sites, we recorded oxygen maxima and minima, water temperature maxima and minima, and conductivity. Impoundments in southeastern Queensland are typically stratified; therefore, seasonal changes in oxygen saturation and water temperature have larger ranges of variation in rivers than in lakes (Tucker 2000). Oxygen saturation was consistently lower in impoundments compared with rivers for maximum and minimum temperatures in all seasons (Fig. 1). Impoundments gained and lost heat more slowly than did rivers because of a larger thermal mass and strong thermal stratification (Fig. 1). In spring, water temperatures rose quickly in rivers that were shallow, less turbid, and well-mixed thermally.
Citation: Chelonian Conservation and Biology 11, 1; 10.2744/CCB-0906.1
Dietary Analysis
We tried to sample at least three small, medium, and large individuals of each species per site to account for sex differences or ontogenetic effects. Those objectives were usually attained even though E. albagula and M. latisternum were less abundant than E. krefftii at all sites (Table 2). However, any site-related bias was considered negligible inasmuch as all E. albagula and M. latisternum were sampled as they came to hand.
We obtained diet samples via stomach lavage (Legler 1977) within 1–2 hrs of capture such that any ingested bait was easily recognized and excluded from analysis. A small percentage of turtles yielded no sample and were interpreted as turtles with an empty stomach rather than as food that did not dislodge. Any “empty” individuals were excluded from the analyses. Stomach contents were transferred to labelled vials and preserved in 70% ethanol.
We viewed lavage samples under a stereo dissecting microscope. Trace amounts of sediment were believed to be ingested incidentally and were ignored. Food items were readily discriminated and grouped into functional plant or animal categories. Plant items included ribbon weed, leafy subaquatic plants, filamentous algae, wind-fallen fruits or nuts of riverine trees, wind-fallen leaves, bark fragments, and terrestrial grasses when banks flooded. Animal prey included freshwater sponges, terrestrial insects, aquatic insects, fish, amphibians, unidentified carrion, bivalves, gastropods, and crustaceans.
Statistical Analyses
We recorded the occurrence (presence/absence) and abundance (counts) of dietary items. We investigated the effect of habitat changes adjusted for season and size in a two-way ANCOVA conducted separately for the abundance of plant (pooled for 6 categories) and animal (pooled for 8 categories) dietary items. Habitat (river or lake) and season (wet or dry) were main effects after adjusting for size (carapace length) as a covariate. To normalize and stabilize the variance of the response variable, we transformed counts by 
to account for zero counts in any prey class.
Ordination
Standard factorial experimental designs (involving replication, randomization, and controls) were not ideal for this survey. Consequently, we evaluated gradients among the uniquely sized dams, on separate rivers, at different elevations, and in operation for varying periods of time. We used ordination techniques (PC-ORD 4.2) to reveal structural relationships among interrelated environmental factors that were hypothesized to influence the dietary ecology but where the factors were of unknown importance.
We conducted a site ordination with principle components analysis (PCA) to identify the 16 explanatory variables with respect to all sites (Table 3). We also conducted a canonical correspondence analysis (CCA) for a primary matrix of prey items as constrained to a second matrix of the 16 environmental variables for the different sites. We constructed separate models for the limited set of species with sufficient data for generalized linear modeling and used a strict version of CCA. We log transformed (log10[x + 1]) species abundance and omitted species that occurred on fewer than four sites because rare species are poorly described by ordination techniques (ter Braak 1995). We log transformed all quantitative (log[y + 0.01]) habitat variables to reduce the influence of extreme values. We examined Pearson correlation matrices of weighted r-values and inspected variation inflation factors to assess colinearity among explanatory variables.
Successional Analysis
We tracked dietary differences in aquatic successional states from lotic to lentic conditions. We calculated a dissimilarity matrix from mean prey abundances taken by the three species of turtles in rivers, weirs, and dams. Data were log10(x + 1) transformed and relativized by the row maximums before calculation of dissimilarity. An ordination by nonmetric multidimensional scaling (MDS) used the Sørensen (Bray–Curtis) distance measure with a random starting configuration. The probability that a similar final stress was obtained by chance was calculated from 1000 randomized runs of a Monte Carlo test. Stress (Kruskal's Stress Formula 1) was calculated from a plot of ordination space versus dissimilarity in the original dimension space. We reran the recommended number of axes (2) for a final solution (400 iterations, with a step length of 0.2) using the starting coordinates saved from the initial solution. A 2-D solution with stress of 1.31, and final instability value was reached after 176 iterations. The recommended 2-D solution was displayed by a 90° rigid rotation of the ordination space to produce a simple and intuitive interpretation that aligned well with axis 1.
RESULTS
Functional Grouping of Prey
We obtained dietary samples from 385 E. krefftii, 126 E. albagula, and 45 M. latisternum from all size categories (Fig. 2). A dietary summary was obtained for intraspecific and interspecific contrasts of habitat, season, and body size (Figs. 3–5). A two-way ANCOVA gave evidence that habitat effects (lake or river) and season (wet or dry) affected the dietary specialist turtles more so than the omnivorous E. krefftii. With the clearly omnivorous guild E. krefftii, there was no significant difference in plant prey consumed by habitat (F = 2.41, p > 0.05) but a significant difference in plant prey by season (F = 7.53, p < 0.01); no significant differences in animal prey consumed by habitat (F = 3.12, p > 0.05) or season (F = 1.92, p > 0.05) were observed. For the herbivorous guild E. albagula, there was a significant difference in plant prey consumed by habitat (F = 8.90, p < 0.01) and season (F = 4.11, p < 0.05); but no significant differences in animal prey consumed by habitat (a nonexistent contrast since no animals were consumed in impoundments by this species) or season (F = 2.29, p > 0.05) were found. For the primarily carnivorous guild M. latisternum, there was no significant difference in plant prey consumed by habitat (F = 0.03, p > 0.05) or season (F = 0.71, p > 0.05); a significant difference in animal prey consumed by habitat (F = 4.62, p < 0.05) was observed but not across seasons (F = 0.28, p > 0.05).
Citation: Chelonian Conservation and Biology 11, 1; 10.2744/CCB-0906.1
Citation: Chelonian Conservation and Biology 11, 1; 10.2744/CCB-0906.1
Citation: Chelonian Conservation and Biology 11, 1; 10.2744/CCB-0906.1
Citation: Chelonian Conservation and Biology 11, 1; 10.2744/CCB-0906.1
In stark contrast to the dietary specialists, the omnivorous E. krefftii displayed little dietary differences by size class or between sexes. The mostly herbivorous E. albagula, juveniles consumed more animal matter than did adults; there was no difference between sexes as herbivorous adults. In contrast, for the mostly carnivorous M. latisternum, males and females differed greatly in their dietary choice. Larger females consumed more plant matter, terrestrial insects or scavenged items than the medium-sized males that fed upon aquatic insects, gleaned from algal mats. A difference in diets between sexes often reflects a body-size difference in sexually dimorphic species. However, body size was not a consistent explanation for why the most sexually dimorphic turtle in the study (E. albagula, SDI = +1.41) had little difference in male–female diets, and a moderately dimorphic omnivore (E. krefftii, SDI = +1.26) had only a minor addition of bivalves to distinguish any dietary difference by sex. Overall it appears that sexual size dimorphism, if it was influential at all, was a factor only with the moderately dimorphic carnivore (M. latisternum, SDI = +1.28).
Site Ordinations
The PCA analysis among sampling sites based on environmental variables in a two-dimensional plot explained 56% of the total variance. Adding an additional two axes explained a total of 82% of the variance, but the rest of the variance was interpreted as noise. The first principal component (32% of variance) presented a linear function among correlated variables of floodplain hydrology, distance from river mouth, altitude, water temperature seasonal maxima, water temperature seasonal minima, catchment drainage area, and annual catchment discharge. The first axis was interpreted as reflecting proximity to the river mouth where higher water temperatures occur in the lower part of the catchment. The second axis (23% of variance) was based on the climatic variables of season, wet or dry season, water thermal maxima, and water thermal minima. It represented a temporal transition across seasons, with summer and winter at opposite ends of the gradient. Stated simply, catchment proximity from river mouth (elevation and distance), the characteristics of the dam (age, back-up, wall height), and water parameters (thermal or oxygen measurements, for maxima and minima) characterized structure within the dataset.
The 3 axes explained 63% of the cumulative variation in a readily interpreted context. The first axis primarily arrayed sites based on the water parameters to the right and the landscape level differences to the left. The second axis displayed a gradient from the dams (at the bottom) against the river sites (at the top), in relation to these underlying variables.
A biplot illustrated that elevation and distance from river mouth had greater bearing than a hydrological classification as influences on the water parameters. In these rivers, there appeared little need to include the hydrological classes. The lack of sharply defined endpoints represented a smooth hydrological gradient across the floodplain which was interpreted as slow-flowing water and gradual change across the catchments.
Dams created an opposite component loading to the gradient plotted for oxygen minima: in other words, larger dams were associated with hypoxic conditions (low oxygen). This effect was largely independent of season, confirming a pattern that bigger dams are consistently lower in oxygen saturation year-round, particularly the hypoxic conditions at the bottom of the water column.
Finally, the similar directionality on the second axis indicates a gradient of alteration from a river toward a lacustrine system as a dam increases in capacity, back-up, or age. Colinearity or redundancy is implicit along this axis, although a more subtle interpretation is that the local effects are similar whether from a very long back-up in a small dam, as a small back-up of a very old dam, or as a very young dam with massive coverage. In other words, co-occurrence of any two of the three factors (capacity, back-up, or age) guides a river's dynamics toward a new domain of ecological, hydrological, and physical processes.
A species ordination constrained to the environmental influences (CCA) showed that diets were fundamentally different among the sites. Diets of all three turtles showed little overlap between impounded sites and river sites (Fig. 6). The MDS illustrated distinctive vectors for the herbivorous, omnivorous, and carnivorous turtles in a succession from river to weir to dam (Fig. 7). Notably, the succession vector for the omnivorous E. krefftii was almost static, with insignificant changes along either trophic axis. The successional vector for the herbivorous E. albagula was aligned with an axis for plant material (Axis 2), and the successional vector for the carnivorous M. latisternum was aligned with an axis of animal foods (Axis 1). The changes for either of the dietary specialists occurred during a transition from river to weir, which was interpreted as a similar magnitude of effect as from a small weir to a larger dam.
Citation: Chelonian Conservation and Biology 11, 1; 10.2744/CCB-0906.1
Citation: Chelonian Conservation and Biology 11, 1; 10.2744/CCB-0906.1
DISCUSSION
The landscape level contrasts in the present study demonstrated that three turtle species at multiple trophic levels had different diets in natural river sites and impoundments. In addition to a qualitative difference in prey consumed, there was a net decrease in dietary diversity and breadth in impoundments. Different prey ingested at the artificial lentic and natural lotic sites in turn suggests that flow regulation may have affected prey availability. Moreover, the successional vectors in diets from the riverine to lacustrine habitats imply that dietary specialists were affected by impoundments more profoundly than a dietary generalist. A next step would be to measure prey availability to determine whether the observed declines in dietary diversity were caused by altered flow regimes. We recognize that as a logical premise to test, but this study was not designed to collect contemporaneous data on prey availability independent of the consumed prey. Thus, data were lacking to demonstrate cause and effect, even though associations of diet differences and habitat differences were clearly apparent.
The study design is the first to our knowledge that contrasts assemblages of riverine turtles in habitats of fluvial and lacustrine conditions. The response of multispecies assemblages to riverine habitat change certainly warrants more study (Poff and Allan 1995), especially given a lack of community studies for Australian turtles (for an exception in fluvial conditions, see Welsh 1999) in contrast with other tropical (Vogt 1981; Vogt and Guzman Guzman 1988; Moll 1990) or temperate assemblages of turtles (Williams and Christiansen 1981; Bjorndal et al. 1997; Lindeman 2000). Typically, studies of riverine turtles adopt a single species approach focused on limited stream reaches, a single watershed, or comparisons of two stream segments. However, no previous ecological studies of Australian turtles have investigated species assemblies across multiple catchments to determine large scale effects of flow stability, stream order, and environmental variables. If such landscape level factors remain unexamined, the extremes of seasonal variability in Australian rivers may confound direct comparisons to rivers elsewhere with more sustained flows.
Habitat diversity was related to species diversity in a turtle assemblage in the St. Croix River, Wisconsin, that was less diverse above a dam than below the dam (Donner-Wright et al. 1999). Habitat differences were noted between dammed and natural sites in studies of the western pond turtle (Actinemys marmorata) in the Trinity River, California, even though this species was considered a generalist that could occupy different types of waterways without difficulty (Reese and Welsh 1998). The western pond turtles preferred sites with deep pools and debris, and although the dammed sites did provide deep pools, the shape of these pools caused higher water velocities that reduced the availability of turtle habitats. The dams and water diversions “fragmented aquatic habitat directly by acting as a barrier to migration and indirectly by creating patches of unsuitable habitat” (Reese and Welsh 1998). In impoundments of Kentucky Lake, riverine species (Graptemys and Apalone mutica) tended to occupy deeper zones with current, whereas habitat generalists (Trachemys scripta) occupied shallower zones (Lindeman 2000). The spatial availability or lack of remnant aquatic habitat types may thereby influence the scope of prey available to foraging turtles.
In less productive habitats, turtles may be prey-switching when preferred foods are limited. For example, Ouachita map turtles (Graptemys ouachitensis) usually eat aquatic vegetation, but if conditions are associated with widely fluctuating water levels, they feed instead on terrestrial plants or other foods of terrestrial origin (Moll 1976). In similar circumstances within oligotrophic lakes, E. krefftii on Fraser Island and C. longicollis in Jervis Bay are reliant upon windfall foods of plant and animal origin (Georges 1982). Within the Amazon basin, many river turtles consume windfall items from the riparian zone (Vogt 2008).
Therefore, it is unsurprising to see seasonal changes of prey availability and prey switching by turtles that forage opportunistically. Our study recorded that the allochthonous inputs of terrestrial plant and animals to the aquatic food webs were greatest in the summer-wet season. The summer diet of T. scripta includes both plants and various animal prey, whereas the winter diet was composed largely of aquatic plants (Parmenter 1980; Lagueux et al. 1995). In contrast, our study indicated that E. krefftii ingested a greater percent of animal matter in autumn and winter and that E. albagula ingested animal matter only in autumn (Fig. 2). We cannot speculate whether seasonal differences in dietary composition related to locally fluctuating stocks of plant and animal prey, or responses to temperature or rainfall, because our study scope across multiple catchments did not survey food availability simultaneously.
A more dramatic difference emerged between wet and dry seasons than for thermal differences, perhaps because of the relatively benign winters in subtropical Queensland. Winter diets biased toward plants may result from a lack of animal food or thermal constraints on digestive efficiency at low metabolism, even if animal and plant foods persist through winter and can be assimilated as long as a turtle stays metabolically active. A slowed digestive function at cooler environmental temperatures would depress assimilation (Parmenter and Avery 1990). At subtropical latitudes, prey availability would appear more important than metabolic activity as an influence on dietary intake.
Turtles commonly exhibit an ontogenetic association of body size and food preference. Dietary divergence among different size classes promotes coexistence and decreases competition by expanding the array of dietary niches (Tucker et al. 1995; Souza and Abel 1998). Short-necked chelid turtles are gape-limited predators; thus, large individuals eat a wider range of prey sizes than do small turtles. In addition, species that are carnivorous as juveniles tend toward omnivory as they grow (Clark and Gibbons 1969; Moll 1976; Parmenter 1980). Ontogenetic shifts in foraging often relate to energetic efficiency, because there are marginal benefits for a large-sized turtle to take small prey unless it is of high energy return, of exceptional abundance, or energetically cheap to obtain. A more basic view is that turtles are carnivorous for as long as they need to gain essential nutrients, especially calcium, to bolster somatic growth (Parmenter and Avery 1990). Further studies may resolve whether intersexual preferences in trophic position, dietary preference, or habitat differences account for these observed differences.
Conservation Implications
Surveys at more than 50 sites across central Queensland catchments found that impoundments had reduced turtle biodiversity relative to the flowing rivers (Table 2). The larger the impoundment, the more severe this effect was. New studies are warranted to delve beyond our results with questions for why river turtles do poorly in impoundments.
Our study found differences in dietary indices, but it remains to be answered whether these were by direct cause and effect of habitat alteration on food resource availability. Aquatic insect larvae do rely on shallow riffle habitat that declines after impoundments (Choy 1998). Macrophytes are severely reduced or lost when impoundments are created and when their water levels fluctuate. Flooding washes away macrophyte beds, and turbid flows smother recovering macrophytes. Construction of impoundments results in initial loss of food plants as shallow areas upstream are inundated. For example, Duivenvoorden (1998) reported that macrophytes died and decomposed within six weeks of construction of an impoundment in central Queensland. After the water level upstream became constant, new subaquatic plants established after a further six to nine weeks, but these plants died when the water level fluctuated again. Windfall fruit from riparian trees are important in diets of river turtles (Moll and Moll 2004), but fruiting trees that did not tolerate root inundation died after impoundment (Tucker 2000).
Separate from the diet study are questions that remain about other postimpoundment effects. Postimpoundment declines in water quality may be affecting a preferred habitat of turtles with cloacal respiration, which is flowing water with high levels of dissolved oxygen throughout (Clark et al. 2008, 2009). Impoundments may also develop conditions favoring increased predation on juveniles. Declines in water quality reduces the time that turtles with cloacal respiration can spend diving for food and increases exposure of juveniles to predators when they surface frequently to breathe (Storey et al. 2008; Clark et al. 2009). This is likely to be compounded by a lack of protective cover associated with reduced macrophyte density in dams. Dams can also reduce the water quality downstream, because they often release poorly oxygenated water, increase sediment, and cause bank erosion through flow regime changes (Kingsford 2001).
Foraging studies provide a basic step to understanding higher-level trophic interactions in aquatic food webs and in some cases may illustrate an organism's response to changes of environmental or habitat conditions. Consequently, dietary studies for animals with narrow prey specializations can guide management efforts that seek to maintain essential prey resources. In future studies, a stable isotope characterization of diets would undoubtedly yield advanced insights that were unavailable from stomach lavages. In addition, the dietary diversity revealed by the present study may be a challenge to resolve in isotopic mixing models when both C3 and C4 plants are represented (Fry 2006).
The MDS successional analysis revealed that an ecological generalist was less affected by a changed aquatic food chain, which is consistent with a pattern that E. krefftii prevails even in habitats that are oligotrophic (e.g., window lakes on barrier islands, see Georges [1982]) or unproductive for extended periods (e.g., Coopers Creek in the arid zone of Queensland, see Cann [1998]). In contrast, E. albagula and M. latisternum retained a narrow trophic profile within the same foraging guild in both fluvial and impounded habitats. It is uncertain whether such dietary shifts were adaptive or simply making the best of adverse conditions.
In impoundments, turtle foraging habitats become strongly restructured by controlled flows that are seasonally irregular in amplitude and period from a river's historical flow. Flow alterations during dam operations may depress or extinguish local distributions of riverine-adapted food resources. However, flow regulation creates other influences apart from potential effects on turtle habitats or diet that must be disentangled, such as a loss of nesting and basking habitats and population fragmentation (Tinkle 1959; Dodd 1990; Moll 1990; Reese and Welsh 1998). A key question for turtles having a narrower dietary niche concerns their long-term response to changes in the aquatic food web. For example, will food availability, nutrition, and assimilation after flow regulation be adequate to sustain population health and reproductive functioning?
Our study findings are also relevant and transferable to turtle species with similar trophic specializations or habitat preferences. Endemic species of conservation concern in Queensland include E. albagula, E. macrurus, and R. leukops with a riverine habitat preference or diets gleaned from fluvial zones (Legler and Cann 1980; Flakus 2002; Thomson et al. 2006; Tucker et al. 2001). Elseya albagula has strong ecological similarities in foraging modes and habitat preferences (Armstrong and Booth 2005), as confirmed by this study, to Elusor and Carettochelys insculpta (herbivores in a fluvial habitat), and M. latisternum mirrors the foraging habits of R. leukops, Myuchelys georgesi, and Myuchelys purvisi (carnivorous gleaning in a fluvial habitat). River habitats occupied by these species of conservation concern would be clearly affected by postimpoundment declines in aquatic macrophytes (Brock and Casanova 1997) or aquatic invertebrates (Brittain and Salveit 1989).
Integrated catchment management is a contentious issue, with escalating demand for new water storages across southeast Queensland. The regional planning of new dams will serve the irrigation needs of major agricultural and economic producers such as sugar cane, cotton, cattle, or citrus crops. Consequently, flow-regulated rivers will increasingly affect a broad span of biological and physical variables. Although river responses to flow regulation are rapid and transparent in aquatic species with short turnover times, far less is known about effects that water storage structures might have on long-lived freshwater turtles. It is vital to monitor habitat changes as they occur, because turtles respond slowly to environmental change. Gradual and subtle demographic effects in a population may not manifest themselves until many years after an impoundment is created. Such time-lags and concerns demand a better understanding of effects of impoundments on freshwater fauna. The empirical data presented here for three river turtles imply that ecological specialists are more directly impacted by trophic changes following loss of lotic habitats than are trophic generalists.
Maximum and minimum ranges of seasonal changes in oxygen saturation (upper panel) and water temperature (lower panel) for impounded (solid lines) and unimpounded (dotted lines) aquatic habitats in southeast Queensland rivers.
Size composition of 3 turtles (Emydura krefftii, Elseya albagula, Myuchelys latisternum) sampled in southeast Queensland rivers.
Effects of habitat type (impounded or unimpounded) on the percentage composition of the diets for Emydura krefftii (top), Elseya albagula (middle), and Myuchelys latisternum (bottom).
Wet–dry season shifts on the percentage composition of the diets for Emydura krefftii (top), Elseya albagula (middle), and Myuchelys latisternum (bottom).
Effects of body size on the percentage composition of the diets for Emydura krefftii (top), Elseya albagula (middle), and Myuchelys latisternum (bottom).
The Principal Components Analysis indicates features that yielded underlying structure in from an environmental matrix of 16 factors (upper left). Canonical correspondence analysis shows nonoverlap of turtle diets constrained by the matrix of 16 environmental factors, separating mainly along axis 1. Impoundments (grey shaded) separate to the left and riverine conditions (black shaded) separate to the right for Emydura krefftii (EK), Myuchelys latisternum (ML), and Elseya albagula (EA).
Multidimensional scaling with successional analysis identifies dietary shifts from (1) riverine to (2) weirs to (3) dams for 3 turtles. Line connecting points H1-H3 mark the herbivorous Elseya albagula diet changing in plant material along Axis 2; line connecting points C1-C3 mark the carnivorous Myuchelys latisternum diet changing for animal material along Axis 1; line connecting points O1-O3 are close together and indicate the omnivorous Emydura krefftii diet as essentially unchanged.
