Platypus predation has differential effects on aquatic invertebrates in contrasting stream and lake ecosystems
Study areas
The lotic exclosure experiment was conducted in Brogers Creek, a westerly flowing stream arising near the town of Nowra on the south coast of New South Wales, Australia (34°44′ S, 150°35′ E), an area with warm summers and cool winters. The stream winds through a steep valley surrounded by dairy farms, with riparian vegetation consisting of an undisturbed overstorey of river oaks (Casuarina cunninghamiana) with an understorey of sedges (Lomandra longifolia), introduced grasses and herbs. River oaks are the main source of litter input, dropping needle-like cladodes and small branches. The dairy farms contribute some organic matter as run-off, although no eutrophication was observed. The stream depth is 2 m maximum, but usually ≤ 1–1.5 m. The substratum is a mix of boulders, gravel, pebbles and cobbles, with silt and detritus in slow-flowing backwaters. A large population of platypuses was resident in the stream during the study, with over 78 individuals captured from August 1998 to August 2001, and individuals travelled its length while foraging44.
The lentic exclosure experiment was conducted in Lake Lea, a small (142 ha), shallow, relatively undisturbed sub-alpine dystrophic lake in north-western Tasmania (41°30′S, 146°5′E)45. Water depth is mostly 1–2 m, with one hole over 10 m deep. The lake substrate is mostly mud and sand, with some large areas of stone and rocky outcrops. Despite being relatively thinly vegetated, diverse macrophytes are present45 with extensive but patchy beds of submerged quillwort (Isoetes drumondii) that provide structural habitat and food for aquatic invertebrates. Platypuses and introduced brown trout (Salmo trutta) are the main vertebrate predators of invertebrates in the lake. We selected this lake as natural undisturbed freshwater lakes are rare in mainland Australia, and none have been studied with respect to platypus. Lake Lea, by contrast, has a large and well-studied population of platypuses32,33,34,46,47,48. Prior to our experiment, 52 individual platypuses were captured48. However, as Bethge48 did not sample the entire lake, the population probably exceeded 52 animals. The exclosure experiment was conducted in the north west of the lake, to avoid interference by anglers48.
Experimental design
Because of the paucity of exclosure experiments investigating the impacts of aquatic mammals on their prey, we briefly reviewed equivalent studies on terrestrial mammals to seek information on appropriate exclosure size, replication and design. Experiments excluding insectivorous mammals, although scant, have used sheet metal or nylon mesh as barrier materials, creating exclusion plots of 3 × 3 m40,41. These experiments used 3–4 exclusion plots and 3–4 control plots, and reported rapid increases in numbers of spiders40 and of large invertebrates41. In further experiments, Wise and Chen42 excluded all vertebrates from 50 m2 plots (n = 5 treatments, 5 controls), but detected no effect on densities of wolf spiders. A review of the effects of predator removal on terrestrial vertebrate prey found that 23 of 116 experiments used exclosures43. Of these, only 13 studies reported any replication, this ranged from 2 to 4 removal plots and an equal number of control plots in all cases43. The median size of plots was 2 ha, reflecting the larger spatial requirements of vertebrate compared to invertebrate prey. Despite the possibility that exclusion fences might affect prey, only two studies reported the use of procedural controls (i.e. sham fences)40,66, all others used open control plots to compare the effects of predator exclusion43.
Following this review, we ran an exclosure experiment in the stream from late summer through autumn 1999 and in the lake from late summer to autumn 2000. The experiments were designed to examine the impact of platypuses on the abundance, taxon richness and community structure of benthic invertebrates, as well as on sediment and epilithic algal biomass. We used three treatments in each of these two contrasting experimental systems: exclosure cages (− PLATYPUS) which prevented access by platypuses to the substrata; uncaged benthic areas (+ PLATYPUS) where platypuses had free access; and a procedural control to determine any cage effects (+ PLATYPUS control).
In the stream, we selected a large pool, ~ 100 m in length, bounded upstream and downstream by 10–20 m long riffles, and installed four mesh cages to exclude platypuses (− PLATYPUS treatment). As noted above, this level of replication is similar to, or greater than, that in most terrestrial exclosure experiments. All cages (1.2 m × 1.2 m, 30 cm high) were constructed of brown plastic Nylex® garden mesh (mesh dimension 5 × 5 cm). Five extra holes, 5 × 10 cm, vertically aligned, were cut in the mesh on all sides and at the top of the cages. These holes, and mesh size, while excluding platypuses, allowed access by invertebrates and fish, including adults of larger fish in the system—Australian bass (Macquaria novemaculeata), long-finned eels (Anguilla reinhardtii), and short-finned eels (Anguilla australis). As judged by the free movement of leaf litter and detritus in water through the cages, the cages had minimal or no effect on water current velocity. Four additional mesh cages of the same dimensions were installed as procedural controls (+ PLATYPUS control) but had 25 × 25 cm holes in the sides and top. These cages allowed free access by platypuses yet still approximated any influence of the cage structure on movements of platypuses, fish, and invertebrates. Plastic mesh was used to prevent any possible interference with platypus electroreception during feeding49. In addition to the mesh cages, four open, uncaged plots the same dimensions as the cages were marked on the open stream bed to serve as open treatments (+ PLATYPUS).
The cage mesh was secured to the substrate using metal stakes and rocks. To simulate this disturbance for all treatments, including the open treatment, rocks were similarly displaced. Cages were placed at the downstream end of the pool where current velocity was minimal, at least 2 m from the stream edges to avoid any systematic differences in current velocity due to the stream banks. Although treatments were confined to broadly the same area, and thus were exposed to similar environmental conditions, we stratified the placement of cages in water depths of 0.45–1.25 m to ensure more representative sampling of the environment. We also placed cages with opposite corners in line with stream flow to minimise leaf litter accumulation on the upstream edge. Treatment plots were ≥ 3 m apart; as the benthic prey of platypuses was expected to be largely sessile, this separation was considered sufficient to avoid spatial confounding. Within these constraints, cages were set in random locations, with assignment to treatment made at random. A single post driven into the substrate was used to mark locations of the + PLATYPUS treatment replicates.
Within each treatment replicate a sediment trap consisting of a plastic tube 10 cm high, with a 4.5 cm diameter opening, was fixed vertically to a stake ~ 20 cm inside the downstream corner of the cage. Sediment traps were used to collect benthic sediments disturbed and suspended by platypus foraging activities or other disturbances. Also, a pre-conditioned terracotta tile (20 cm2) was placed in the middle of each cage, or in the case of the + PLATYPUS treatment, about 20 cm upstream of the sediment tube/marker post to determine if platypuses had any direct or indirect effects on epilithic algae. If platypuses suppress algal-grazing herbivorous invertebrates, it is likely that algal abundance would vary differentially between treatments on the artificial tile substrates. Tiles were preconditioned by leaching them in the river for six weeks prior to the experiment, and any accumulated algae were removed before deployment.
The exclosure experiment in the lake was similar to that conducted in the stream, except that six replicates of each treatment were used rather than four. This increased statistical power to detect any treatment differences, given that the lake was expected to have lower invertebrate biomass compared with the stream. Treatment plots were again ≥ 3 m apart, set up on sites where the substrate was firm enough to support the cages, and treatments allocated randomly. Platypuses are larger in Tasmania than on mainland Australia, but still much smaller than the holes in the procedural control cages and thus able to readily pass through them. Brown trout (Salmo trutta) in the lake are 0.6–1 kg, but rarely reach this size (https://www.ifs.tas.gov.au/ifs/IFSDatabaseManager/WatersDatabase/lake-lea), so individuals could readily pass through all the exclosures.
Both experiments ran for six weeks before invertebrate sampling took place. Six weeks was deemed long enough for any potential effects of platypus foraging to be detected, especially as the late summer to autumn study period is when male platypuses attain their greatest body mass and condition and could be expected to forage most intensively29,44. Conversely, a more prolonged experimental period would have seen increasing damage to the exclosure structures from both water flow and human interference. We did not repeat the experiments in winter through spring to avoid disturbance to the platypus breeding season44. However, there is little or no seasonal variation in the composition of aquatic invertebrates between seasons, at least in the stream system29. This may suggest that similar results could be obtained at other times, although further experiments are needed to confirm this. At least 14 platypuses were known to have moved through the experimental stream pool over the study period, with some individuals visiting the open and control treatments, based on capture and radiotracking data44. In comparison, only five Australian bass were captured during extensive net sampling during the same period, suggesting that, during the course of the experiment, platypus abundance exceeded that of the most abundant large predatory fish in the pool44. Platypuses were probably present in much greater numbers in the pool than those identified, as platypuses in this system have large and overlapping linear home ranges44, and numbers were not monitored continuously during the experiment.
Ideally the experiments would have been replicated in multiple streams and lakes to increase the power and generality of our results, and to have been run across different seasons, but this was not logistically possible. We therefore interpret our results with caution and note that our conclusions are restricted to the sites and seasons that were studied.
Invertebrate, algal, and sediment sampling
Invertebrates were sampled by day in both systems using a Brooks suction sampler (Brooks67 (33 cm2 sampling area). Although 33 cm2 is relatively small, pilot studies suggested that this area would yield sufficient invertebrates to allow robust tests of our hypotheses. However, because we also expected small-scale spatial variation in the invertebrates, we took three sub-samples of invertebrates in each replicate cage. Suctioning for each sample took 60 s, with the sampler held firmly over the substrate. Samples were then preserved separately in 70% ethanol and transported to the laboratory for identification.
Invertebrates were sorted from the detritus under × 6–× 40 magnification, counted, and identified to genus where possible68. Exceptions, due to taxonomic impediments, were fly larvae of the families Chironomidae and Tipulidae, aquatic mites (Acarina), worms (Oligochaeta), flatworms (Dugesiidae), and members of the beetle family Scirtidae. Invertebrates were assigned to a trophic group (detritivore, herbivore, omnivore, predator) using published accounts36,50 and following our previous work29,44. These assignations are approximate as diets can vary between instars and locales. However, the categories were considered to be broadly useful in determining functional roles51 and thus for elucidating the role of platypuses in predator–prey and potentially trophic cascades in the study ecosystems. Leaf litter detritus from the stream samples was retained, dried and weighed, but these data are not presented as allochthonous leaf litter was not common in the lake, thus preventing direct comparisons44.
At the conclusion of both experiments, six weeks after exclosure establishment, algae were vigorously brushed from the tiles, washed into vials using stream or lake water and preserved using 2% Lugol’s iodine solution. In the laboratory, algae were filtered onto pre-weighed 0.45 μm filterpaper, dried at 60 °C to constant weight, and weighed to 0.0001 g. Sediment traps were collected and the material was transferred to a pre-weighed drying dish and dried to constant weight at 60 °C. The material was then weighed to 0.01 g precision.
Data analysis
All analyses focused on comparisons between the three treatments (i.e., + PLATYPUS, − PLATYPUS and + PLATYPUS control) within each experiment, but separately between the lotic and lentic systems. Two sets of analyses were undertaken for the two ecosystem datasets. Firstly, univariate comparisons were carried out to identify differences among means for the abundance and taxon richness of invertebrates and invertebrate trophic groups (hypotheses 1 and 2), algal biomass and sediment mass (hypotheses 3 and 4). Secondly, multivariate analyses were carried out to explore possible shifts in composition of the invertebrate community as a whole among treatments, separately in both systems (hypothesis 2). We did not formally compare the datasets observed in the lentic and lotic systems (hypothesis 5), but instead compared the effect sizes arising from the manipulation of platypus in each system.
Univariate data were subjected to Cochran’s test for homogeneity of variances52. Due to heterogeneity, invertebrate data were √-transformed, then analysed using a nested (hierarchical) one-factor analysis of variance (ANOVA), with treatments fixed, and replicates nested within treatments52. We took this approach to quantify replicate-within-treatment variance, rather than losing information by averaging across samples52. Algae and sediment data were also √-transformed and compared among treatments using a one-factor ANOVA. Tukey’s multiple comparison test was performed on each pair-wise comparison to identify sources of difference between treatments. Univariate tests were conducted using SYSTAT version 9 and Statistica 13.
The multivariate invertebrate community dataset was analysed using PRIMER, version 5. Community composition within each treatment was first assessed using a Bray–Curtis dissimilarity matrix. This distance measure is widely used in ecological studies, and is considered to be robust53,54 and useful in determining the underlying structure of biological communities. The matrix was then subjected to non-metric multidimensional scaling (nMDS), providing an ordination where the distance between samples reflects relative similarity in species composition. Data were square root transformed to down-weight the effects of the most common taxa and maintains the effects of the less common taxa29,55. An analysis of similarity (ANOSIM routine, PRIMER ver. 5) was performed on the dissimilarity matrices to test for differences between treatments. This permutation test uses a randomisation approach to generate significance levels to test a priori hypotheses about differences between groups of samples54,55. The SIMPER (Similarity Percentages) sub-routine in Primer ver. 555 was used to examine the contribution of each taxon to the average dissimilarity between all pairs of inter-group samples. This test does not have a statistical hypothesis-testing framework, but is useful in data exploration to indicate which ‘taxa’ are principally responsible for differences between a priori defined groups that differ in matrix structure55. SIMPER was used to determine which trophic groups contributed to dissimilarities between the + PLATYPUS and − PLATYPUS treatments. More
