Diets and Resource Partitioning among Three Sympatric Gurnards in Northeastern Tasmanian Waters, Australia

Abstract

Dietary niches can support the coexistence of closely related sympatric species in marine systems, which can lead to the presence of greater abundances of those species that can potentially support their fisheries or greater abundances for other fish species that prey upon those species. Dietary relationships for three species of gurnard (Family Triglidae) that occur together in the benthic coastal environment of northeastern Tasmania, Australia (Red Gurnard Chelidonichthys kumu, Grooved Gurnard Lepidotrigla modesta, and Roundsnout Gurnard Lepidotrigla mulhalli), were examined for the presence of such dietary niches. The species are either fishery-important (Red Gurnard) or provide prey (Grooved Gurnard and Roundsnout Gurnard) for fishery-important species (e.g., Platycephalidae and Zeidae). Based on stomach content analyses, all three gurnards were shown to be bottom-feeding carnivores that consumed mainly benthic crustaceans, particularly decapods and amphipods, with teleosts also being important in the diets of only the larger Red Gurnard. Nonmetric multidimensional scaling ordination and multivariate analyses based on volumetric contributions of different prey taxa to the stomach contents revealed significant differences in dietary composition among all three species, implying a partitioning of food resources. Size-related and temporal changes in dietary composition were each significant among the three gurnards, but there were no interactions between body size and time. Principal components analysis of head and mouth morphology demonstrated that mouth protrusiveness was the dominant morphological difference among species, which may in part account for the niche partitioning observed from the stomach content analysis. Given the important role of gurnards in benthic food webs, these relationships will improve the specification of ecosystem-based fisheries models and their ability to predict the effects of environmental and anthropogenic perturbations.

Received June 29, 2016; accepted April 9, 2017

Understanding feeding relationships in sympatric fishes is of fundamental interest in studies of trophic ecology, particularly when species are closely related and morphologically similar. While studies have historically assumed that co-occurring, morphologically similar species would fulfill similar ecological niches (Webb 2000; Violle et al. 2011), sympatric species of fish often display resource partitioning as a strategy for reducing direct interspecific competition for available resources (Connell 1980; Ross 1986; Platell and Potter 2001). Such partitioning can occur through differences in food, habitat, and diel foraging activity and can play a major role in overall community structure and function (Ross 1986). Superimposed on such temporal and spatial resource use are ontogenetic (or purely size-related) changes in consumption of prey, where individuals of different sizes and/or ages within a species target different prey, further reducing the potential for direct intraspecific competition that may detract from the success of subsequent cohorts (Platell and Potter 1999, 2001; Lek et al. 2011; Sommerville et al. 2011). Such reductions in direct competitive processes provide a mechanism by which taxonomically similar fish species can co-occur.

Gurnards (Scorpaeniformes: Triglidae) are marine demersal fishes that commonly occur in tropical and temperate waters globally (Richards and Jones 2002). Gurnards comprise approximately 126 species within nine genera (Eschmeyer et al. 2016), with more than 33 species occurring in Australian waters (Gomon et al. 2008). Three sympatric species of gurnard—the Red Gurnard (Bluefin Gurnard) Chelidonichthys kumu, Grooved Gurnard Lepidotrigla modesta, and Roundsnout Gurnard L. mulhalli—are widely distributed throughout southern and eastern Australian waters from Western Australia to New South Wales, including Tasmania (May and Maxwell 1986; Rees et al. 1999; Gomon et al. 2008). Although several of the larger gurnards, such as the Red Gurnard, Painted Latchet Pterygotrigla andertoni, and Latchet P. polyommata, grow large enough to be marketed in Australia (Rowling et al. 2010:237–239), the three species of interest in our study (i.e., smaller Red Gurnard and the full size range of Grooved Gurnard and Roundsnout Gurnard) are generally regarded as fisheries bycatch from demersal trawling in Australian waters (Tuck et al. 2012), and there is no information on how bycatch-related issues may impact these species. Among the Triglidae, the Roundsnout Gurnard has been listed as an ecologically important species because of its high abundance and prominence in the diets of other commercial fishes, such as the John Dory Zeus faber and various flatheads (Platycephalidae) in southeastern Australian marine ecosystems (Bulman et al. 2001).

The diets of gurnards have been studied worldwide (e.g., Ross 1978; Moreno-Amich 1992, 1994, 1996; Terrats et al. 2000; Boudaya et al. 2007; Huh et al. 2007; Baeck et al. 2011), with interspecific partitioning of resources recorded among species in the northwestern Atlantic (Ross 1977), the Mediterranean Sea (Labropoulou and Machias 1998), southwestern Australia (Platell and Potter 1999), and the Cantabrian Sea (Lopez-Lopez et al. 2011). Despite their economic and ecological importance in southeastern Australia, little is known about the dietary habits of the species in this region. Coleman and Mobley (1984) and Bulman et al. (2001) described feeding guilds of fishes in the area, including Red Gurnard, Grooved Gurnard, and Butterfly Gurnard L. vanessa in southeastern Australian waters, but those authors did not explore the factors that minimize competition and facilitate sympatry, such as size-related dietary changes and morphological specialization. Furthermore, Park et al. (2017) recently compared diets and trophic positioning of Roundsnout Gurnard and Butterfly Gurnard but did not include the co-occurring Red Gurnard or Grooved Gurnard and did not examine for any morphological differences that potentially impact diet choice. Such information is crucial to defining functional roles of gurnards within coastal ecosystems (Wootton 1990; Brodeur and Pearcy 1992) and is essential for ecosystem-based models that have a strong trophic basis (Fulton et al. 2011; Smith et al. 2011a; Bulman et al. 2014).

In this study, dietary habits and intraspecific and interspecific differences in dietary composition were assessed for Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard in southeastern Australian waters. Specifically, we aimed to (1) quantify the overall diet of the three sympatric gurnard species; (2) determine size-related and seasonal changes in dietary composition within each species; and (3) link any differences in diet and trophic niches with subtle variations in head and mouth morphology of the three species. Quantification of diets, trophic niches, and morphological traits in sympatric gurnards can be used to support future trophic and ecosystem models that seek to quantify community interactions and better predict the effect of perturbations on community structure.

METHODS

Study area and sampling

Sampling was conducted in waters off northeastern Tasmania, Australia (40°15–42°20′S, 147°05–148°35′E; Figure 1). Samples were collected from numerous research trawls onboard the RV Bluefin (Australian Maritime College vessel) in July (winter), September (early spring), and November (late spring) 2012. Mean water temperatures in the study area ranged from 14.4°C in winter to 19.8°C in late spring (D. Hunt, Australian Maritime College, unpublished data). Fish were collected at depths between 30 and 40 m by using a demersal trawl with a 70-mm-mesh cod end and 16-m headline length towed at 3 knots (1.54 m/s). Immediately after capture, individuals were identified to species, frozen at −20°C, and kept frozen until processing, which occurred immediately after thawing in the laboratory. For each individual, TL was measured (nearest 1 mm), the stomach was removed, and the stomach contents were preserved for at least 24 h in a 70% solution of ethanol. These methods of preservation (i.e., freezing and preservation in 70% ethanol) did not reduce our ability to identify prey or to estimate the volumetric contributions of the various prey items (our unpublished data).

FIGURE 1. Location of the study area in northeastern Tasmanian waters, Australia. Samples of three gurnard species were collected via trawling within the area denoted by gray shading.

Stomach content analyses

Each stomach was cut open, and stomach fullness was visually estimated by assigning a score between 0 (empty) and 10 (fully distended with food) as a measure of feeding intensity. All prey were identified to the lowest possible taxon (typically the order or family level) by using a dissection microscope and taxonomic sources (e.g., Shepherd and Thomas 1982; Poore 2004). The contribution of each prey taxon to the total volume was visually assessed under a dissecting microscope. The percentage volumetric contribution of each prey taxon to the total volume of the stomach contents (%V) was estimated visually with the aid of a grid-marked Petri dish (Hynes 1950; Hyslop 1980).

Cumulative prey curves were constructed for each species to determine whether a sufficient number of stomachs was analyzed to describe the diet (Ferry and Cailliet 1996). To achieve this, the order of dietary data was randomized 10 times, and the cumulative number of new prey taxa was re-counted for each randomization. The mean number of prey taxa per stomach (±SD) was plotted against the number of stomachs analyzed, with the asymptote of the curve indicating that an adequate number of stomachs were studied. A curve was considered to asymptote if at least 10 previous values of the total number of prey taxa were in the range of the asymptotic number of prey ± 0.5 (Huveneers et al. 2007).

Dietary data were expressed as frequency of occurrence (%F = [A_i/N] × 100) and as a volumetric percentage (%V = [V_i/V_T] × 100), where A_i is the number of fish that consumed prey taxon i, N is the total number of fish examined (excluding those with empty stomachs), V_i is the volume of prey taxon i, and V_T is the total volume of prey taxa. Although the volumes of both unidentifiable crustaceans and sediment are shown in Table 1, these were not considered valid dietary categories and were not included in subsequent dietary analyses. Dietary niche breadth was calculated at the lowest possible taxonomic level by using Levin’s standardized niche breadth (B; Krebs 1989), calculated as B = [(∑P_ij)⁻¹ – 1] × (n − 1)⁻¹, where P_ij is the proportion of the diet consumed by predator i that is made up of prey j, and n is number of prey taxa. The index B ranges from 0 to 1; low values indicate a diet dominated by few prey taxa (i.e., specialist predator), and high values indicate a generalist diet.

TABLE 1. Percentage frequency of occurrence (%F) and mean percentage volumetric contribution (%V) of the prey taxa, with the highest taxonomic levels for those groups shown in bold italics, in the diets of Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard in northeastern Tasmanian waters. The number of stomachs with food and the mean stomach (±SE) fullness score are also presented.

	Red Gurnard		Grooved Gurnard		Roundsnout Gurnard
Taxon or variable	%F	%V	%F	%V	%F	%V
Anthozoa			1.2	<0.1
Polychaeta	23.0	3.8	14.0	1.8	30.5	5.9
Mollusca	6.7	1.1	4.7	0.6	19.0	1.9
Bivalvia	5.2	0.6	2.3	<0.1	19.0	1.9
Gastropoda	0.7	<0.1	1.2	<0.1
Cephalopoda	0.7	0.5	1.2	0.6
Crustacea	83.7	80.1	96.5	96.4	93.3	89.3
Copepoda					1.0	<0.1
Leptostraca	0.7	<0.1	8.1	0.5	18.1	1.5
Ostracoda	3.7	0.9	4.7	0.3	43.8	7.7
Mysidacea	3.7	0.4	2.3	0.1	10.5	0.4
Tanaidacea	4.4	0.3	20.9	3.7	25.7	1.8
Amphipoda	48.9	10.4	75.6	17.6	76.2	21.5
Cumacea	13.3	1.8	3.5	2.3	67.6	9.6
Isopoda	16.3	2.8	15.1	2.2	46.7	7.6
Decapoda	65.9	37.0	82.6	49.9	43.8	11.7
Caridea	59.3	27.3	72.1	44.4	22.9	2.9
Penaeoidea	0.7	0.1	2.3	0.4
Galatheoidea	3.7	0.5	3.5	0.7	1.0	0.5
Brachyura	21.5	9.0	24.4	4.4	35.2	8.3
Unidentified crustaceans	73.3	26.5	69.8	19.6	92.4	27.5
Teleostei	27.4	12.2	2.3	<0.1	1.0	0.8
Other material	30.4	2.8	17.4	1.1	43.8	2.0
Sediment	30.4	2.8	17.4	1.1	42.9	2.0
Algae	0.7	<0.1
Number of stomachs with food		135		86		106
Mean stomach fullness (±SE)		4.7 (0.2)		7.0 (0.2)		5.1 (0.2)

To investigate whether any size-related trends existed in the diets of Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard, volumetric dietary data for each species were aggregated into successive 10-mm TL intervals, and the mean %V data for each prey taxon in each 10-mm length-class for a given predator species were subjected to a hierarchical cluster analysis based on a Bray–Curtis similarity matrix (Clarke and Gorley 2006). As a result, all three gurnard species were divided into two size-groups (small or large). The small length-class was defined as 158–189 mm for Red Gurnard, 139–169 mm for Grooved Gurnard, and 103–159 mm for Roundsnout Gurnard; the large length-class was 190–267 mm for Red Gurnard, 170–197 mm for Grooved Gurnard, and 160–207 mm for Roundsnout Gurnard (Supplementary Figure S.1 available separately online).

Any seasonal changes in diet were visually represented by aggregating the dietary data for each of the small and large individuals of the three species into three periods that represented Australian seasons: winter (July), early spring (September), and late spring (November). Mean %V of the various prey taxa in the diets of individuals from different length-classes in different seasons was calculated for each gurnard species.

Multivariate analyses on dietary data

Because volumetric data are considered to best represent the relative importance of each prey taxon, especially when different-sized prey are ingested (Hyslop 1980), subsequent analyses were performed using volumetric data for each prey taxon. Prior to such analyses, dietary data for each of the species were first randomly sorted into groups that contained three to five individuals within each length-class in each seasonal group (depending on the sample size of that group), and the averages of %V for each prey taxon were determined for each of the resultant groups. If the number of individuals in a resultant length-class/seasonal group was less than two individuals, the group was not included in subsequent multivariate analyses (e.g., the winter group for Grooved Gurnard and the early spring group for Roundsnout Gurnard were excluded). Averages thus represented the dietary samples that were used for all subsequent analyses. Such randomization and subsequent grouping of volumetric data were designed to reduce the number of prey taxa in the samples with zero values, thereby increasing the effectiveness of multivariate analysis (De Lestang et al. 2000; Platell and Potter 2001), and were used because many of the stomachs contained only a few (mostly 1–5) of the 21 dietary categories recorded in the overall diets. Volumetric data were then square root transformed to avoid any tendency for the main dietary components to be excessively dominant.

Bray–Curtis similarity matrices were constructed for each gurnard species and visualized via nonmetric multidimensional scaling (nMDS) ordination (Platell and Potter 2001; Clarke et al. 2006). The matrices were then subjected to a series of two-way (intraspecific) and three-way (interspecific) permutational multivariate ANOVAs (PERMANOVAs) to assess whether there were significant effects of species (three levels), length-class (two levels), and season (three levels) as well as their two-way and three-way interactions. The PERMANOVA is a nonparametric distance-based ANOVA that uses permutation procedures to test hypotheses; it assigns components of variation (COVs) of differing magnitudes to the main factors and any two-way or three-way interactions between combinations of main factors included in the chosen comparison. The larger is the COV, the greater is the influence of a particular factor or interaction term on the structure of the data (Anderson et al. 2008; Linke 2011). The COV attributable to a fixed factor in a given model was considered in terms of the sum of squared fixed effects (Anderson et al. 2008).

Two-way crossed analysis of similarity (ANOSIM) was used to test for any significant differences in dietary composition among the three species with respect to size or season based on the same factors used in the PERMANOVA (see above); the magnitude of the ANOSIM R-statistic was used to indicate the relative importance of any such differences (Clarke et al. 2014). Global R-values from the ANOSIM to the verified similarities (distance) within defined groups vary between 0 and 1. An R-value of zero represents no difference in the average similarity among and within groups, whereas an R-value of 1 indicates that in terms of composition, all samples within each group are more similar to each other than to samples from any other group (Clarke and Gorley 2006). For among-species comparisons of diet, the ANOSIM average R-value () was computed for species versus each of the other two factors (length-class and season) combined. In cases where ANOSIM detected a significant difference, pairwise ANOSIM comparisons were then used to determine which comparisons between species or between length-classes/seasonal groups of each species showed significant differences.

Similarity percentage (SIMPER) analysis was employed to determine which prey taxa typified the diets of particular species and which prey made the greatest contributions to any dietary dissimilarities between species that were identified by both PERMANOVA and ANOSIM. The SIMPER analysis gives the percentage of similarity or dissimilarity between levels of factors and for specific levels of factors.

All analyses were performed using routines in the PRIMER version 6 multivariate statistics package (www.primer-e.com) and the PERMANOVA+ add-on module (Anderson et al. 2008).

Head and mouth morphology analyses

Head and mouth morphology was characterized as described by Wainwright and Richard (1995) and Platell and Potter (2001), with measurements made using calipers to the nearest 0.1 mm. Morphological measurements consisted of head length (snout tip to operculum), premaxilla length (the bottom corner to the middle of the top lip), mouth height (greatest gape height with mouth opened fully), jaw-lever extensions (open upper jaw, open lower jaw, and closed lower jaw), mouth width (greatest gape width with mouth fully opened), and dentary length (the bottom corner of the lip to the middle of the top lip). Head and mouth measurements for each fish were expressed as a ratio of the individual’s TL (i.e., because all measurements were positively correlated with TL), and the data were then subjected to principal components analysis (PCA) to determine whether head and mouth morphology differed among species and varied with TL. A PERMANOVA using non-averaged data was then performed on the Euclidean distance matrix for normalized head and mouth variables to determine whether mouth morphology differed significantly among species and/or between length-classes and whether there was a significant species × length-class interaction. The PCA and PERMANOVA were conducted using PRIMER version 6 and the PERMANOVA+ add-on module (Anderson et al. 2008).

RESULTS

Overall Dietary Composition

The stomach contents of 146 Red Gurnard (160–267 mm TL), 90 Grooved Gurnard (139–197 mm TL), and 113 Roundsnout Gurnard (103–207 mm TL) were examined. Percentages of empty stomachs were 7.5% for Red Gurnard, 4.4% for Grooved Gurnard, and 7.9% for Roundsnout Gurnard. Mean stomach fullness was similar between Red Gurnard (mean fullness score = 4.7) and Roundsnout Gurnard (5.1) and was slightly greater for Grooved Gurnard (7.1; Table 1).

Cumulative prey curves attained asymptotes after about 114 stomachs for Red Gurnard, approximately 72 stomachs for Grooved Gurnard, and approximately 70 stomachs for Roundsnout Gurnard (Figure 2). Thus, our sample sizes were sufficiently large to allow us to confidently describe the diets of these fishes in our study area.

FIGURE 2. Cumulative prey curves (prey taxa per stomach; ±SD) for Red Gurnard Chelidonichthys kumu, Grooved Gurnard Lepidotrigla modesta, and Roundsnout Gurnard L. mulhalli sampled from northeastern Tasmanian waters.

In terms of both %V and %F, crustaceans made an overwhelmingly important contribution to the diets of Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard (Table 1). For both Red Gurnard and Grooved Gurnard, caridean decapods were the most important of the identifiable crustaceans, occurring in 59.3% and 72.1%, respectively, of all stomachs and contributing 27.3% and 44.4%, respectively, to the total dietary volume. Amphipods also made substantial contributions to the diet volume of Red Gurnard (%V = 10.4%) and Grooved Gurnard (17.6%) and an even greater contribution for Roundsnout Gurnard (21.5%), in which these crustaceans were the most important of the identifiable crustacean prey. The Red Gurnard was the only species to ingest substantial amounts of teleosts, comprising 12.2% of the total dietary volume. Anthozoans, polychaetes, mollusks, and algae each contributed less than 6% to the total dietary volumes for each species (Table 1). Sediment also occurred frequently in the diets of the three species (between 17.4% and 43.8%); however, the volumetric contributions of sediment were low (%V < 2.8%).

Dietary niche breadth differed among the three species, being lowest for Grooved Gurnard (B = 0.149), greatest for Roundsnout Gurnard (0.362), and intermediate for Red Gurnard (0.249).

Intraspecific Trends in Dietary Composition

Red Gurnard

When the volumetric dietary data were examined for each length-class by season, the small length-class of Red Gurnard fed mainly on amphipods, carideans, and crabs during winter and late spring, the last of which declined in importance, while carideans increased in volumetric contribution during early spring (Figure 3a). Ostracods only contributed moderately (%V = 13.7%) to the winter diets of small Red Gurnard. Diets consumed by larger Red Gurnard in Australian winter comprised mainly carideans, crabs, and teleosts, with the contribution of carideans becoming greater during both early spring and late spring (Figure 3b). Teleosts were also an important contributor during late spring. Dietary niche breadth was higher for smaller Red Gurnard (B = 0.422) than for larger specimens (0.228).

FIGURE 3. Mean percentage volumetric contributions of the different dietary components in winter, early (E) spring, and late (L) spring for (a), (b) small and large length-classes of Red Gurnard; (c), (d) small and large length-classes of Grooved Gurnard; and (e), (f) small and large length-classes of Roundsnout Gurnard in northeastern Tasmanian waters. Numbers above each column represent the total number of individuals in each length-class. Each of the other prey groups included small prey taxa or unidentifiable prey as follows: other invertebrates included anthozoans and mollusks; other decapods included penaeoids and galatheoids; other crustaceans included copepods, leptostracans, mysids, tanaids, and unidentifiable crustaceans; and other materials included sediment and algae.

On the nMDS ordination plot, dietary samples for Red Gurnard displayed discrete groups according to both season and length-class (Figure 4a). In terms of season, although there was some overlap, the seasons were relatively discrete from each other. In terms of length-class, small fish during late spring and winter tended to be separated from larger fish (Figure 4a).

FIGURE 4. Nonmetric multidimensional scaling ordination of the dietary composition constructed from Bray–Curtis similarity matrices that employed volumetric diet contributions during three seasons for small (open symbols) and large (filled symbols) length-classes of (a) Red Gurnard, (b) Grooved Gurnard, and (c) Roundsnout Gurnard. Each point represents the mean volumetric data for three to five randomly selected individuals from each season and size-class.

The dietary composition of Red Gurnard differed significantly with length-class and season (PERMANOVA: P = 0.002 and 0.003, respectively), but there was no significant interaction between these two factors (P = 0.055; Table 2). The COVs were similar for length-class and season (Table 2). Two-way crossed ANOSIM showed that the dietary composition for Red Gurnard differed with both length-class and season (P = 0.003 and 0.001, respectively) and that values of were also similar ( = 0.524 and 0.436 for length-class and season, respectively). Pairwise ANOSIM tests further demonstrated that these seasonal differences were strongest between early spring and winter. The SIMPER analysis emphasized that amphipods and carideans were present in relatively greater volumes in the diets of smaller Red Gurnard, whereas the opposite trend was true for teleosts. The diet consumed by Red Gurnard in late spring was distinguished from winter and early spring diets by containing greater volumes of amphipods and lesser volumes of teleosts.

TABLE 2. Differences in the mean percentage volumetric contributions of the various prey taxa in the stomach contents for Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard based on mean squares (MS), pseudo-F-ratios, and components of variation (COVs) for a series of permutational multivariate ANOVA tests employing Bray–Curtis similarity matrices (asterisks indicate significance at P ≤ 0.05).

Source	df	MS	Pseudo-F	COV	P
Red Gurnard
Length-class	1	3,147	4.167	17.900	0.002*
Season	2	2,217	2.935	16.810	0.003*
Length-class × Season	2	1,286	1.704	14.336	0.055
Residual	23	755		27.482
Grooved Gurnard
Length-class	1	1,540	3.239	11.874	0.006*
Season	1	7,179	15.099	29.796	0.001*
Length-class × Season	1	673	1.417	7.742	0.228
Residual	13	475		21.805
Roundsnout Gurnard
Length-class	1	1,635	3.576	15.861	0.008*
Season	1	2,505	5.477	20.909	0.001*
Length-class × Season	1	989	2.162	15.066	0.068
Residual	18	457		21.384

Grooved Gurnard

The diets of Grooved Gurnard in the small length-class were particularly dominated by carideans and contained smaller amounts of amphipods during early spring, with the former decreasing and the latter increasing in late spring (Figure 3c). For larger Grooved Gurnard, nearly 70% of the dietary volume in early spring comprised carideans, but their contribution declined to 24% in late spring, whereas amphipods and crabs contributed 13–15% in late spring (Figure 3d). Dietary niche breadth for the small length-class (B = 0.254) was higher than that calculated for the large length-class (0.084).

The nMDS ordination plot showed that the early spring diets of Grooved Gurnard were clearly separated from late-spring and winter diets; within the spring seasons, samples for the two length-classes showed more separation in late spring than in early spring (Figure 4b). The PERMANOVA showed that dietary composition for Grooved Gurnard also differed with length-class and season (P = 0.006 and 0.001, respectively) and that there was no interaction between body size and season (Table 2). The COV value was far greater for season than for length-class. Two-way crossed ANOSIM, on the other hand, showed that dietary composition differed significantly with season ( = 0.958, P = 0.001) but not with length-class ( = 0.231, P = 0.102). The SIMPER analysis revealed that the diets consumed by smaller Grooved Gurnard contained greater volumes of amphipods, whereas the large length-class was typified by a greater volume of carideans. The dietary composition of individuals sampled in early spring was distinguished from that of fish examined in late spring by greater abundances of carideans.

Roundsnout Gurnard

The diets of the small length-class of Roundsnout Gurnard were dominated by amphipods in both winter and early spring. However, substantial amounts of isopods were also ingested in winter but not early spring, while the contribution of cumaceans increased from winter to early spring, and a small amount of carideans was recorded only in early spring (Figure 3e). Five different types of crustacean (ostracods, amphipods, cumaceans, isopods, and crabs), each contributed between 9% and 14% of the total diet volume for larger Roundsnout Gurnard in winter (Figure 3f). In early spring, however, amphipods became more important in the diets of the large length-class, contributing 39.0% to the total dietary volume. Although other crustaceans (mostly unidentified crustaceans) constituted nearly half of the diet, amphipods, cumaceans, and isopods collectively contributed 36.5% to the dietary volume in late spring. Niche breadth was similar between the small length-class (B = 0.362) and the large length-class (0.343) of Roundsnout Gurnard.

The nMDS ordination plot of the dietary composition for Roundsnout Gurnard revealed a clear separation between the early spring diet and the winter diet (Figure 4c). In terms of length-class, the samples for small and large Roundsnout Gurnard were intermingled during early spring, whereas the single point for the small length-class in winter was positioned outside the large group of samples from the large length-class.

Similar to the results for both Red Gurnard and Grooved Gurnard, the dietary composition of Roundsnout Gurnard significantly differed with both length-class and season (PERMANOVA: P = 0.008 and 0.001, respectively), and there was no significant interaction between length-class and season (Table 2). Two-way crossed ANOSIM also showed that composition of the Roundsnout Gurnard’s diet differed significantly with length-class and season (P = 0.043 and 0.001, respectively), with the -statistic being greater for season ( = 0.833) than for length-class (0.474). According to SIMPER analysis, amphipods and cumaceans were more important in the diets of the small length-class of Roundsnout Gurnard, whereas the reverse was true for isopods, ostracods, and crabs in the diets of large fish. In terms of season, amphipods typified the Roundsnout Gurnard diet in early spring, together with isopods, amphipods, crabs, cumaceans, and ostracods in the winter.

Interspecific Comparisons

The nMDS ordination depicted a clear difference in dietary composition among the three species (Figure 5). The data points for Red Gurnard and Grooved Gurnard overlapped slightly on the left of the plot, whereas the data points for Roundsnout Gurnard lay on the right and were clearly separated from those of the other two species (Figure 5).

FIGURE 5. Nonmetric multidimensional scaling ordination of the dietary composition constructed from Bray–Curtis similarity matrices that employed the overall volumetric diet contributions for the three study species (Red Gurnard Chelidonichthys kumu, Grooved Gurnard Lepidotrigla modesta, and Roundsnout Gurnard L. mulhalli) in northeastern Tasmanian waters. Each point represents the mean volumetric data of three to five randomly selected individuals from each species.

Three-way PERMANOVA revealed that dietary composition was significantly related to species, length-class, and season (Table 3). There also were significant two-way interactions for species and season, but no other significant two-way and three-way interactions among those factors. The COV was the greatest for species (Table 3). Two-way crossed ANOSIM tests using data for species versus the other two factors combined (i.e., length-class and season) showed that dietary composition for the three species was significantly different overall ( = 0.742, P = 0.001). Pairwise ANOSIMs revealed that the diets of Red Gurnard differed significantly from those of the two Lepidotrigla species (Grooved Gurnard: = 0.259, P = 0.015; Roundsnout Gurnard: = 0.909, P = 0.001) and that there was a very marked difference between the Lepidotrigla species’ diets ( = 0.981, P = 0.001).

TABLE 3. Interspecific comparison of the mean percentage volumetric contributions of various prey taxa to the stomach contents of Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard based on mean squares (MS), pseudo-F-ratios, and components of variation (COVs) for a series of permutational multivariate ANOVA tests employing Bray–Curtis similarity matrices (asterisks indicate significance at P ≤ 0.05).

Source	df	MS	Pseudo-F	COV	P
Main effects
Species	2	7,861	13.624	25.487	0.001*
Length-class	1	3,196	5.538	12.304	0.002*
Season	2	4,845	8.398	19.834	0.001*
Interaction
Species × Length-class	2	1,134	1.965	9.964	0.053
Species × Season	2	1,529	2.650	13.346	0.003*
Length-class × Season	2	970	1.684	8.507	0.097
Species × Length-class × Season	2	1,078	1.869	13.695	0.053
Residual	57	577		24.021

The SIMPER analysis indicated that the dissimilarity between species ranged from 50.7% to 62.1%, with 12 prey taxa contributing more than 90% to the dietary dissimilarity among the three species. The main typifying prey taxa for Red Gurnard and Grooved Gurnard were similar (e.g., carideans and amphipods), whereas those for the Roundsnout Gurnard included amphipods, cumaceans, and isopods. The most-observed prey items that contributed to the dissimilarity between Red Gurnard and the two Lepidotrigla species were carideans and teleosts; the diet items that most often contributed to dissimilarity between Grooved Gurnard and Roundsnout Gurnard were carideans and cumaceans (Table 4).

TABLE 4. Prey taxa identified by similarity percentage analysis as typifying (gray boxes) the dietary composition for each gurnard species (Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard) in northeastern Tasmanian waters and distinguishing (unshaded boxes) the diet composition among the species. Prey taxa with a contribution of over 10% to similarity or dissimilarity are presented. An asterisk indicates that the percentage contribution of a prey taxon is greater for the species listed in the column heading than in the row heading.

Species	Red Gurnard	Grooved Gurnard	Roundsnout Gurnard
Red Gurnard	Carideans
	Amphipods
	Teleosts
Grooved Gurnard	Carideans	Carideans
	Teleosts*	Amphipods
	Crabs*
	Amphipods
Roundsnout Gurnard	Carideans*	Carideans*	Amphipods
	Teleosts*	Cumaceans	Cumaceans
	Cumaceans		Isopods
	Amphipods

Head and Mouth Morphology

The PCA of standardized head lengths and mouth measurements showed that the first and second principal components (PC1 and PC2) explained 53.6% and 21.2% of the total variation, respectively (Table 5). The resultant PCA plot clearly showed separation among the three species along the first axis (Figure 6). Thus, the data points for Red Gurnard lay in the middle, while the data points for Grooved Gurnard and Roundsnout Gurnard lay toward either positive or negative PC1 values (i.e., to the right or left of the Red Gurnard points; Figure 6). Furthermore, TL was highly correlated with the PC2 axis, indicating that the points for the three species progressed from the smaller to larger individuals negatively along the PC2 axis (Figure 6).

FIGURE 6. Plots of principal component axis 1 (PC1) versus axis 2 (PC2) from principal components analysis of standardized head length and mouth measurements for individual Red Gurnard Chelidonichthys kumu, Grooved Gurnard Lepidotrigla modesta, and Roundsnout Gurnard L. mulhalli. Together, PC1 and PC2 explained 74.8% of the total variation. The vector represents Pearson’s product-moment correlation, and the circle shows a correlation of 1.0.

TABLE 5. Eigenvalues and percentage of variation explained by the first five principal component (PC) axes for standardized head and mouth measurements of Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard. Eigenvectors for each of the eight measurements on each PC are shown.

Variable	PC1	PC2	PC3	PC4	PC5
Eigenvalue	3.96	1.56	0.56	0.48	0.30
Variation explained (%)	53.6	21.2	7.6	6.5	4.0
Eigenvectors
Head length	0.463	–0.088	0.401	−0.269	0.161
Premaxilla length	0.144	0.043	0.008	0.018	−0.180
Mouth height	0.385	0.783	−0.365	−0.244	0.173
Open bottom jaw protrusion	0.433	−0.430	−0.669	−0.015	−0.305
Closed bottom jaw protrusion	0.308	−0.260	−0.100	0.437	0.749
Open top jaw protrusion	0.509	−0.153	0.440	−0.244	−0.200
Mouth width	0.236	0.319	0.234	0.728	−0.238
Dentary length	0.139	0.013	−0.035	0.294	–0.402

Variables with the highest loadings (eigenvectors) on PC1 were head length and open top jaw and bottom jaw protrusions. Mouth height and open bottom jaw protrusion had the highest loadings on PC2 and PC3, respectively (Table 5). The results demonstrated that head length and the protrusiveness of both the upper and lower jaws were lowest in Roundsnout Gurnard, moderate and comparable in Red Gurnard, and greatest in Grooved Gurnard. The particularly high and positive eigenvector for mouth height on PC2 emphasized that the smaller specimens of the three species had the larger vertical gapes relative to body size (Table 5).

The PERMANOVA on the standardized head and mouth measurements for the different length-classes of the three gurnard species also showed significant differences with species and length-class (P = 0.001), but the mean square was nearly three times greater for species than for length-class, indicating that species was a far more important factor than size (Table 6). A significant species × length-class interaction was also observed.

TABLE 6. Interspecific comparison of head and mouth morphology among Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard based on mean squares (MS), pseudo-F-ratios, and components of variation (COVs) for a series of permutational multivariate ANOVA tests employing Euclidean distance matrices (asterisks indicate significance at P ≤ 0.05).

Source	df	MS	Pseudo-F	COV	P
Species	2	213.94	46.744	2.152	0.001*
Length-class	1	61.50	13.438	0.987	0.001*
Species × Length-class	2	34.11	7.425	1.143	0.001*
Residual	305	4.58		2.139

DISCUSSION

Although the three sympatric gurnard species in the present study each fed largely crustaceans, multivariate analyses of the different crustacean, teleost, and other prey types demonstrated that food resources were significantly partitioned among Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard in coastal waters of northeastern Tasmania. All three species were benthic carnivores that mainly consumed (based on volume) epibenthic crustaceans, including carideans, amphipods, crabs, cumaceans, and isopods. The dominant types of epibenthic crustacean prey, however, differed among the three species. Carideans, amphipods, and crabs were most prominent in the diets of Red Gurnard and Grooved Gurnard, with the contribution of carideans and amphipods being higher in the Grooved Gurnard’s diet than in the Red Gurnard’s diet. Amphipods, cumaceans, and crabs were the most important contributors to the diets of Roundsnout Gurnard, but the dietary contribution of carideans was far lower for this species than for the other two gurnard species. Thus, although the three species shared similar prey taxa, differential exploitation of prey was evident.

The diets reported here for the three gurnard species are comparable to those previously reported for the same species in this region (Park et al. 2017) and other areas and for gurnards studied elsewhere. Epibenthic crustaceans dominated the diets of Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard in southeastern Australia (Coleman and Mobley 1984; Bulman et al. 2001). Park et al. (2017) also found that these crustaceans dominated the diets of Roundsnout Gurnard and Butterfly Gurnard L. vanessa (except larger individuals of Butterfly Gurnard) in northeastern Tasmania. Such a focus on epibenthic crustaceans was also recorded for Grooved Gurnard and Australian Spiny Gurnard L. papilio from southwestern Australia (Platell and Potter 1999); the gurnard Chelidonichthys (Aspitrigla) cuculus, Large-Scaled Gurnard L. cavillone, and Streaked Gurnard Chelidonichthys (Trigloporus) lastoviza from the Cyclades and Dodecanese Islands (eastern Mediterranean; Terrats et al. 2000); Longfin Gurnard C. obscurus and Streaked Gurnard from the Gulf of Gabes, Tunisia (Boudaya et al. 2007); and Spiny Red Gurnard C. spinosus and Redbanded Searobins L. guentheri from southeastern Korea (Huh et al. 2007; Baeck et al. 2011).

All three of our study species consumed teleosts, but the relative contribution of teleosts to the diet differed among the species; the teleost contribution to the diets of Red Gurnard was more substantial than contributions to the other two species’ diets. Teleosts have been considered a fundamental feeding resource for some gurnard species (Moreno-Amich 1992, 1994; Colloca et al. 1994), and the relative proportion of teleosts in the diet often increases with increasing body size (Lopez-Lopez et al. 2011). Although the dietary contribution of teleosts was greatest for the large length-class of Red Gurnard, teleosts always comprised extremely low volumes in the diets of the two Lepidotrigla species, suggesting that teleosts were not important in their diets.

The proportion of epibenthic crustaceans in the diets of Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard indicates that these three species forage close to or just above the surface of the substratum (see also Platell and Potter 1999). This is consistent with other gurnard species that use chemosensors on their six modified pectoral rays to detect benthic prey while foraging (Roberts 1978; Finger 1982). However, the relatively low contribution of polychaetes and sediment to the diets in this study suggests that these gurnards feed just above the substratum rather than within it (see also Platell and Potter 1999).

Intraspecific Changes in Diet

Size-related shifts observed in the diets of the three gurnards suggested intraspecific resource partitioning within each species. These include shifts from epibenthic prey to more mobile teleosts for Red Gurnard; from amphipods to carideans for Grooved Gurnard; and from amphipods, isopods, and cumaceans to various benthic invertebrates for Roundsnout Gurnard as the predators grew larger. Such size-related diet changes are common in fish species and are usually related to maximizing energy intake as factors limiting predation change (Gerking 1994) and reducing direct intraspecific competition for resources (Langton 1982; Chizinski et al. 2007; Barnes et al. 2011).

Size-related changes in dietary composition have previously been reported in the diets of other triglids (e.g., Ross 1978; Platell and Potter 1999; Huh et al. 2007; Park et al. 2017), but they have not been consistent for all triglids. For example, the diets of Leopard Searobin Prionotus scitulus demonstrated a shift from planktonic and epifaunal prey for small individuals to infaunal prey for larger fish (Ross 1978), whereas Spiny Red Gurnard consumed the same type of prey but included larger individuals of those prey items in their diet as predator body size increased (Huh et al. 2007). Such changes in diet may be related to the onset of maturity, but assessing this possibility was outside the scope of the present research. It is therefore apparent that size-related shifts in diet have not been recorded for all gurnard species or may be region specific. Similar to our study, Morte et al. (1997) reported that two gurnard species (Tub Gurnard Chelidonichthys [Trigla] lucerna and Longfin Gurnard) off the coast of Spain (Gulf of Valencia) showed little size-related changes in diet. Similarly, Redbanded Searobins from southeastern Korea showed no size-related change in diet with respect to prey type, but the number of prey items consumed increased with increasing body size (Baeck et al. 2011).

“True” seasonal changes should be viewed conservatively, but interannual seasonal changes could not be assessed here due to the lack of seasonal replication. However, aggregation of sampling dates into seasonal categories for the two length-classes did show interesting patterns that may be indicative of seasonal changes. Marked differences in dietary composition for Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard were observed among the three seasons (Australian winter, early spring, and late spring). Among the identifiable prey taxa, consumption of teleosts by larger Red Gurnard increased in winter and early spring and decreased in late spring, whereas Grooved Gurnard and Roundsnout Gurnard ingested their highest volumes of carideans and amphipods in early spring regardless of size. Stagioni et al. (2012) and Platell and Potter (1999) also reported seasonal differences in the consumption of prey items for Tub Gurnard and two Lepidotrigla gurnards (Grooved Gurnard and Australian Spiny Gurnard). These seasonal differences probably reflect changes in the abundance and availability of prey items, but the lack of information on seasonal abundance of prey taxa in this study makes it difficult to present reasonable hypotheses for the observed seasonal dietary differences in our study species.

Resource Partitioning among the Three Gurnard Species

Partitioning of food resources is thought to occur when species that overlap in their distribution show differential predation to minimize competition for food (Krajewski et al. 2006). The sympatric gurnard species in the present study displayed a low competition potential in terms of their food resources. Thus, multivariate analyses of dietary composition revealed that although the three species generally consumed similar types of prey, their diets were significantly different and therefore provided a low niche overlap and evidence for some temporal and size-related resource partitioning among the species. Resource partitioning also has been observed among other gurnard species (Platell and Potter 1999; Terrats et al. 2000; Lopez-Lopez et al. 2011).

When resource partitioning is identified within and among species, it is common to see slight morphological changes that may reflect the differing functional roles of those species (Ross 1986). Differences in dietary composition among our study species may thus be reflected in differences in their mouth morphology. Our data showed that mouth protrusiveness at a given body length of Roundsnout Gurnard was smaller than that in Red Gurnard and Grooved Gurnard. Thus, the highly protruded and wider mouths of Red Gurnard and Grooved Gurnard may enable these individuals to more efficiently extract prey from the surface of the substrate. Jaw protrusion has been hypothesized to increase attack velocity and thus improve capture success (Motta 1984; Westneat and Wainwright 1989; Norton 1991), and it appears to be more related to where or how they feed than to the robustness of the prey itself.

Because the three gurnard species coexist in similar depths over at least part of the depth range, such differences in diet may partly reflect their use of differing microhabitats, which may house different suites of potential prey and/or support other differences in foraging behavior (see also Lek et al. 2011). This is difficult to determine in the present study because a large demersal trawl was used for the collection of gurnards. However, microhabitat use has been inferred in other studies of gurnards. For example, Labropoulou and Machias (1998) observed that two sympatric gurnards, the Large-Scaled Gurnard and Streaked Gurnard, exploited different habitats containing different prey. In terms of dietary niche breadth, the index B was considerably higher for Roundsnout Gurnard than for Grooved Gurnard, and the value for Red Gurnard was intermediate; these results indicate that the Roundsnout Gurnard is a more generalist feeder than the other two species. Such differences in feeding strategy or behavior among sympatric species constitute one of the key characteristics of resource partitioning within and among species (Ross 1986; Platell and Potter 2001; Smith et al. 2011b).

The present study yields important insights into the diets and resource partitioning among—and thus the functional roles of—three sympatric gurnard species in northeastern Tasmanian waters. Stomach content analyses indicated that Red Gurnard, Grooved Gurnard, and Roundsnout Gurnard were all associated with the benthic food web, feeding mainly on abundant epibenthic and benthopelagic species. Although epibenthic crustaceans were most commonly consumed, teleosts also were frequently ingested by larger Red Gurnard. Some temporal, spatial, and ontogenetic partitioning of resources was observed within and among the three species. Because these gurnards consistently consume benthic crustaceans, they were categorized within the range of benthic secondary and/or tertiary consumers in the southeast Australian marine ecosystem (Davenport and Bax 2002). Thus, their ecological roles as potential predators may be important in the benthic ecosystem of these waters. A lack of samples covering wider regions and/or all possible size ranges, along with a lack of seasonal replication, imposed limitations on our ability to describe the absolute diets consumed by the gurnard species throughout their entire life histories. Investigation of dietary habits of the three species is important for providing baseline data to improve trophic and ecosystem modeling in southeastern Australian waters.

ACKNOWLEDGMENTS

This research was funded by the University of Newcastle, the Department of Biological Sciences at Macquarie University, and the Environ-Ecological Engineering Institute, South Korea. We are indebted to the crew of the RV Bluefin for their assistance in sample collection. Samples were collected as bycatch from University of Tasmania under Animal Ethics Permit Number A0011023.