Large predators influence food web structure through consumptive and non-consumptive effects (Estes et al., 2011). However, the nature and strength of predator-prey interactions depends on a variety of factors, such as habitat type and structure, prey abundance and diversity, as well as selectivity by predators (e.g. Paine, 1992; Hixon and Beets, 1993). In the marine environment, there are a range of factors governing success of predation events that include removal or destruction of habitat through natural disturbance events (Syms and Jones, 2000), fishing pressure that selectively removes large predators first (Pauly et al., 1998), and the perceived threat of predation, leading to modification of prey behaviours (Lester et al., 2020). Fish predator-prey interactions are largely governed by size and gape, with most species being limited by the size of prey they can swallow, which can be > 50% of their own body size (Scharf et al., 1997). The diet of larger species of shark, for example, is less restricted than large teleosts as they are equipped with serrated dentition that facilitates the removal of parts of animals larger than themselves and they also generally have wider gapes than their teleost competitors (e.g. Lucifora et al., 2006; Braccini, 2008). The role of predation by piscivores on reef environments represents a top-down driver of fish community structure (Graham et al., 2003; Boaden and Kingsford, 2015; Hixon, 2015), which may indirectly influence the benthic environment (Madin et al., 2019). However, the importance and strength of top-down effects has been debated due to the complexities of ecosystem dynamics and poorly understood exploitation histories (Ruppert et al., 2013; Rizzari et al., 2015; Casey et al., 2017).

Establishing baselines of the trophic structure or establishing natural (i.e. in the absence of human impacts) trophic links of predatory marine fish communities is complicated due to these species being among the first to be harvested by commercial, recreational, and subsistence fisheries (Jennings and Polunin, 1997; Pauly et al., 1998; Pinheiro and Joyeux, 2015; Frisch et al., 2016), due to their high value and catchability. Furthermore, much of the work to date has focussed on describing predatory fish community structure in shallow waters around reefs (e.g. Frisch et al., 2014; Boaden and Kingsford, 2015; Bierwagen et al., 2019; Speed et al., 2019), although recent studies have also included mesophotic depths (> 50 m) (e.g. Lindfield et al., 2016; Asher et al., 2017).

Specific information on the diet of predatory fishes targeted by commercial fishing and their prey is often based on stomach content studies and in many instances limited to localised areas of a fishery (e.g. Salini et al., 1994; Overholtz et al., 2000). Even though the diet of most target species can be classified broadly as carnivorous, there is known intraspecific plasticity in feeding among regions (Farmer and Wilson, 2011), which complicates management at both the species and community level. Having a sound understanding of diet and trophic linkages is critical to developing ecosystem models and assessing how they respond to fishing pressure (Bascompte et al., 2005; Farmer and Wilson, 2011). Long-term studies that document ontogenetic or seasonal shifts in diet or dietary sources are rare (although see Holt et al., 2019). However, this level of information is essential to understand potential impacts that may occur across the lifecycle of a species, which is currently lacking for many seascapes (Abrantes et al., 2015). Therefore, our current understanding of the feeding ecology of large predatory fish species is limited to a restricted range of species in certain areas, which ultimately hinders effective management and conservation efforts (Farmer and Wilson, 2011).

The most common method of obtaining dietary information for fishes (and thus trophic role) is through examination of stomach contents and or intestinal tracts (Hyslop, 1980). Estimates of contributions or relative importance of dietary items using this approach range from counts of the number and percentage of prey items identified, to the gravimetric and volumetric calculation per prey item (See reviews by Hyslop, 1980; Baker et al., 2014; Amundsen and Sánchez‐Hernández, 2019). This method is usually lethal and only provides an immediate snapshot of the most recently consumed items (Takahashi et al., 2020). Furthermore, identification of partially digested items can be problematic (Baker et al., 2014), and in many cases (>30% of the time) piscivore stomachs are empty due to a variety of reasons, such as intermittent feeding and stomach eversion due to swim bladder expansion during capture (See review by Vignon and Dierking, 2011; Vinson and Angradi, 2011).

Alternative methods for assessing dietary information and trophic role of fishes include molecular approaches such as analysis of stable isotopes of the consumer tissue (Post, 2002), fatty acid profiles (e.g. Bierwagen et al., 2019) and more recently, DNA identification of prey items obtained from stomach content metabarcoding (e.g. Casey et al., 2019; Takahashi et al., 2020). Of these, the analysis of stable isotopes is the most common approach used in dietary and trophic relationship studies. Stable isotope ratios of nitrogen N¹⁵:N¹⁴ ‘δ¹⁵N’ and carbon C¹³:C¹² ‘δ¹³C’ in consumer tissues are often used in ecological studies to infer dietary niche, with δ¹⁵N enrichment providing an indication of an organism’s trophic level (Perkins et al., 2014). In comparison, δ¹³C can provide geographic and trophic information on nutrient residence because of the predictable variation in δ¹³C values with latitude and among different primary production types (e.g., coastal, benthic, and planktonic) (Bird et al., 2018). Stable isotopes can be used to estimate trophic positions and body length relationships (Speed et al., 2012; Frisch et al., 2014), establish spatial foraging patterns and migrations (Ménard et al., 2007; Ferreira et al., 2017), monitor dietary shifts between seasons, life stages, or life styles (e.g. benthic vs pelagic) (Nakamura and Sato, 2014; Park et al., 2020), and can be modelled to estimate prey source contributions (Frisch et al., 2014). White muscle is the most commonly used tissue for analysis of assimilated isotopic source material (usually δ¹³C and δ¹⁵N), which provides a resolution or turnover rate (‘isotopic half-life’) of several months for large predatory fish (e.g. Nelson et al., 2011). More recently, fish fin tissue has been trialled as a non-lethal method, with generally a good correlation between isotope ratios and muscle tissue, although turnover rates can vary compared to muscle tissue (Sanderson et al., 2009; Willis et al., 2013; Galván et al., 2015; Matley et al., 2016). While the assessment of assimilated stable isotopes in predators can provide a range of inferred information, they lack the fine taxonomic resolution that can be achieved with stomach content analysis or use of DNA. Some of the limitations of direct sampling of fish stomachs or indirect molecular techniques, such as stable isotope analysis, can be avoided by using these approaches together in the same study (Matley et al., 2018; Nielsen et al., 2018). This has the potential to offer greater insights compared to the use of any of these methods independently and can resolve trophic interactions at scales necessary for ecosystem-based management (Levin and Lubchenco, 2008).

Here, we used stomach and gastrointestinal tract content in conjunction with stable isotope analyses to describe the trophic structure of predatory fishes of high importance to commercial fisheries in north-western Australia. An important aspect of our study was the ability to sample a community of commercially important predatory fishes in a large (77km x 40 km) fishery management zone on a tropical shelf that had been protected from commercial fishing for more than a decade. This allowed us to examine the trophic roles of these species in an environment where the impacts of removal by fisheries on trophic structure was minimised. We focused on five common species of snapper (Lutjanidae), emperors (Lethrinidae), and grouper (Epinephelidae) that together constitute >25% of the North Coast Demersal Resource, with an annual value of $12 million AUD (Newman et al., 2018b). Our study examined the following questions: a) How is the predatory fish community structure reflected in terms of stable isotopes δ¹⁵N and δ¹³C, and is there evidence of niche overlap among species?; b) How do isotope signatures vary between sampling periods – is there a seasonal shift?; c) Is there evidence of an increase in δ¹⁵N with the body length of fish, implying that larger individuals feed higher in the food web?; d) Does fin tissue act as a suitable non-lethal proxy for muscle tissue for a key fishery target, the red emperor (Lutjanus sebae)?; and e) Is the isotopic composition of fish reflected in their diet obtained from stomach and gastrointestinal tract content analysis?

There was evidence of separation of the five predatory species in isotopic space, suggesting differences in dietary sources and trophic structure, which was partially confirmed by items present in the stomach and GIT for some species. The range of δ¹⁵N‰ between the smallest species in our study (L. punctulatus) and the largest (E. multinotatus) was within the range of trophic discrimination factors (TDF) that indicate likely predation within this group of predators (e.g. Wyatt et al., 2010; Matley et al., 2016; Canseco et al., 2021). Indeed, E. multinotatus was observed to predate on both L. punctulatus and E. areolatus while in the traps during this study (pers. obs.). It is possible however, that other factors not explored in our study may have influenced TDFs of these predatory fish species, such as water temperature, tissue types, growth rates, and prey type (Canseco et al., 2021).

The low values of δ¹³C (-18 to -15‰) for all species were indicative of a food chain driven by benthic productivity (France, 1995), as has been observed for mesophotic fish communities in the East/Japan Sea (e.g. Park et al., 2020). Similarly, studies that have assessed δ¹³C values of primary producers, consumers, and particulate organic matter in shallow coastal areas of North-Western Australia provide reference ranges for planktonic sources between -24 to -19‰ and benthic producers predominantly ranging between -16 and -7‰ (Vergès et al., 2011; Vanderklift et al., 2016). A reliance on benthic production has also been previously observed for large predatory fish species (reef sharks) in shallower, coastal waters of North-western Australia (e.g., Speed et al., 2012). Similarly, studies focussing on emperors and groupers from other regions such as the Seychelles also found comparable δ¹³C values for E. multinotatus and L. sebae to our study (Sardenne et al., 2017) and one study of P. maculatus on the Great Barrier Reef estimated that 89% of its δ¹³C was derived from benthic production (Frisch et al., 2014).

Both δ¹³C and δ¹⁵N differed for each species between sampling periods, suggesting that there were seasonal differences or potentially a dietary shift over the three-month period between sample collection. Unfortunately, only limited numbers of stomachs were collected in the first trip (primarily L. sebae), so a comparison of GIT contents between sampling periods was not possible for most species. However, Lutjanus sebae caught at Location 1 during the first trip when δ¹⁵N was lower, had a predominance of crustacean prey items in guts, whereas teleosts were a much greater part of the diet in the second trip, when δ¹⁵N in diets was higher. This may indicate a shift in diet between seasons for this species or the changes in availability of prey. Alternatively, there might also be strong seasonal differences in the source of primary productivity in these food webs, as occurs in other fisheries (Abrantes et al., 2015). Indeed, seasonal enhancement of chlorophyll a concentrations at Ningaloo Reef in North-western Australia have been shown to be driven by an acceleration of the Leeuwin Current (Rousseaux et al., 2012), and fluxes in particulate organic matter composition across a reef have been observed through wind-driven upwelling (Wyatt et al., 2013). Unfortunately, a detailed assessment of productivity was beyond the scope of the current project. Seasonal differences in stable isotope ratios can also be related to a range of other variables, such as changes in light, temperature, salinity, fluctuations in food availability, or variation in feeding due to the life stage of consumer, sex, or physiology (Hyndes et al., 2013). Given the restricted range of length distributions of the species we sampled (adult fishes), life stage is unlikely to have been a major factor contributing to the observed differences in our study. Furthermore, a considerable change in growth rate between sampling periods (separated by three months) for adult fish was deemed unlikely to have been a major influencing factor on stable isotope values of these species. Indeed, a controlled feeding trial of medium to large sized coral groupers (Plectropomus leopardus) found that most ¹⁵N incorporation was driven by metabolism, rather than growth (Matley et al., 2016).

The δ¹³C ‰ for most species also varied between locations, which suggests that these demersal species are limited in their foraging range ( Supplementary Figure 1 ). Restricted movement of L. sebae was confirmed in a concurrent telemetry study, which showed no dispersal of tagged individuals between the two study locations over a six-month period (Meekan et al., 2021). Other telemetry studies of tropical demersal fishes have also found limited movement, home ranges, and activity spaces of tagged individuals (e.g. Zeller, 1997; Taylor and Mills, 2013; Currey et al., 2014). Despite this tendency for restricted and repeated fidelity to small areas, longer migrations have also been observed, possibly due to reproduction, as is the case for Lethrinus nebulosus (e.g. Babcock et al., 2017a). The individuals in our study were primarily adults, however, juveniles of L. punctulatus (< 20 cm TL) are known to spend their early life in inshore macroalgal habitat before migrating offshore (Babcock et al., 2017b), where presumably they also undergo changes in diet. This species had the highest enrichment of δ¹³C, which may be a residual signature from inshore feeding, although there was no evidence of smaller individuals having higher δ¹³C values than larger individuals ( Supplementary Figure 4 ) as might be expected if this was a result of ontogenetic changes in habitats. High overlap of both δ¹³C and δ¹⁵N between locations for L. punctulatus suggests that they may have a less specialised, and less location-specific diet compared to the other species, which was also reflected in their stomach and GIT content (i.e., all groups of prey were represented).

Isotopic niche overlap was high (>89%) between L. sebae and P. maculatus, suggesting they are targeting prey within similar trophic levels ( Supplementary Figure 2 ). Unfortunately, we were unable to confirm this finding due to P. maculatus having mostly empty stomach and GIT contents. However, L. sebae did have a high proportion of teleosts in the gut samples (60%), confirming their reliance on other species of fish as a source of nutrients, as has been observed previously (Salini et al., 1994). Similarly, diet studies on coral reefs have also confirmed that P. maculatus are primarily piscivorous (Kingsford, 1992; St John, 1999) and tend to feed on benthic and midwater prey, compared to other species within the genus Plectropomus, which feed on planktonic prey (Matley et al., 2018). The largest (E. multinotatus) and smallest (L. punctulatus) species observed in our study had minimal niche overlap (≈9%), which was unsurprising given the difference in body size (Max TL 100 cm and 38.4 cm respectively) (Bray and Gomon, 2021; Newman et al., 2021), although there is currently little information available for the diet of either of these species (Farmer and Wilson, 2011). Dietary information from stomach and GIT content analysis was limited for E. multinotatus, with 20 of 21 gut samples being empty. This is likely due to a lifestyle of infrequent feeding, which is thought to be common in large piscivores (Vinson and Angradi, 2011), but may be a consequence of stomach eversion due to barotrauma when fish are raised from depth. Indeed, both stomach vacuity (being empty due to infrequent feeding) and regurgitation have been reported in coral reef grouper and snapper by other studies (Vignon and Dierking, 2011).

It is important to note here that while we observed minimal overlap in isotopic niche space for some species, this may not necessarily equate directly to trophic niche differences. Indeed, trophic niche can be considered synonymous to isotopic niche only when it is primarily driven by consumer-resource interactions (Yeakel et al., 2016), which may indeed be the case with our study, although without spatially and temporally relevant environmental measurements that may affect assimilated isotope values, it is difficult to assess. Isotopic niche extent has been shown to be dependent on sampling scale (Reddin et al., 2018), with fine-scale sampling having lower environmental heterogeneity. Given the limited spatial and temporal scale of sampling in our study, environmental heterogeneity is unlikely to have had a major influence on isotope values, which provides some confidence in comparisons of isotopic and trophic niche spaces.

Frequency distributions of body length were consistent across both sampling periods for all species, although there was a slight, albeit non-significant difference for P. maculatus, likely as a result of a low sample size. The size range for all species primarily included mature individuals, which are targeted by the commercial trap and line fishery in the region. Therefore, any differences in isotopic signatures observed between trips was unlikely to be a result of differing distributions or life history stage between trips. Body length was found to have a positive relationship with δ¹⁵N, which likely reflects increased size (e.g. St John, 1999) or trophic level (e.g. Costa, 2009; Dalponti et al., 2018) of prey items consumed. Despite a positive relationship between fin and muscle tissue for both δ¹³C and δ¹⁵N in L. sebae, there was no evidence of a difference in the enrichment of δ¹⁵N between the different tissue types for this species. However, this was not true for δ¹³C, which was significantly more enriched in fin tissue compared to muscle tissue, a pattern that has also been observed in other species of finfish (e.g. Willis et al., 2013) and sharks (e.g. Matich et al., 2010), although see Jardine et al. (2011). The closely matching values of δ¹⁵N for both tissues in L. sebae provides some evidence that using this isotope from fin tissue clips is a good proxy for muscle tissue. This may not be the case with other species, and any tissue corrections should be tested on a species-specific basis (Willis et al., 2013).

The most common identifiable prey items based on stomach and GIT analyses were fish parts (scales and bones) followed by crustacean parts (limbs, claws, carapaces), as has also been previously identified for a species of snapper in north-west Australia, the saddle tail snapper (Lutjanus malabaricus) (Takahashi et al., 2020). We found that a high proportion of stomachs were empty, or contained the bait used in our fish traps. For this reason, the GIT contents were usually more informative than the stomach contents, although the drawback with this approach is that by the time prey items had progressed through the stomach, only hard components remained intact. The contents from GIT are therefore biased towards hard materials, particularly those found in fish bones and scales (e.g., collagen), and crustacean exoskeletons (e.g., chitin), and other calcium carbonate structures (e.g., mollusc shells). Several individuals, particularly from the larger species (E. multinotatus and P. maculatus), had amorphic organic matter (AOM) present in their GIT, which may have been partially digested fish remains. Future studies could adopt the use of DNA metabarcoding to help provide further resolution to identify the remains of this prey (e.g. Casey et al., 2019; Takahashi et al., 2020). Furthermore, we also advocate the collection of gut and molecular samples across all species and trips in future work, as this would provide enough data to support the potential diet shift, we observed for L. sebae. Additionally, collection of isotope samples from primary producers would provide N_base, which would allow the calculation of trophic positions of predatory fish, enable mixing models to estimate the source of dietary carbon and nitrogen, and also give an indication of seasonal fluxes that could be used to help better interpret findings.

The coarse level of dietary information obtained in our study, along with community trophic structure inferred from stable isotopes are required to help form a complete picture of food-webs, which are often lacking for many important fishery species and are needed for the development of ecosystem based models to assist management (Abrantes et al., 2015). Of particular importance to modelling, is the understanding of trophic pathways and associated implications for the removal of predators from an ecosystem (Farmer and Wilson, 2011), such as the group of species of commercially important fish that were the focus of the current project. For example, some of the smaller to mid-sized fish in our study such as L. punctulatus and L. sebae had large percentages of crustaceans in their diet, while trophic discrimination factors of δ¹⁵N among these predators with some of the larger sized species such as E. multinotatus suggested predation within this guild. Diet varies among and within species targeted by fishers and the impact of fishing on prey and predator communities will depend on the species and quantity of fish removed (Farmer and Wilson, 2011). Therefore, our findings will be of importance to current approaches to fishery management that rely on protection via static closures and bag and size limits of fish.

Our multi-facetted approach to defining trophic structure of a demersal predatory fish community identified differences in the isotopic niches occupied, although there was of overlap in diet for some of these species (e.g., L. sebae and L. punctulatus). The low values of δ¹³C in fish indicated a benthic driven food-web, which fluctuated through time, either due to environmental or biological drivers. Location-specific δ¹³C signatures in all species were indicative of animals with a restricted foraging range, which is consistent with our understanding of the limited movement patterns in demersal predatory fishes. A positive relationship with δ¹⁵N and body length was also consistent across species and provided evidence of inter-guild predation between the larger (E. multinotatus and P. maculatus) and smaller species (e.g., L. punctulatus and E. areolatus). Similarities of δ¹⁵N between muscle and fin tissue for L. sebae provided strong evidence for the use of fins as a non-lethal technique for sampling isotopes, although we were unable to compare multiple tissues for all species. Examination of both stomach and GIT contents for each species was informative for some species and partially confirmed patterns observed using stable isotopes, although the high proportion of empty stomachs limited comparisons. Our study highlights the complex spatial and temporal trophodynamics among predatory fish communities in a demersal shelf ecosystem and provides information that could contribute to ecosystem-based models to assist in the management of this commercially important guild.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by The University of Western Australia Animal Ethics Committee.

CS, MM, MB, MP, and BT collected data used in the manuscript. All authors assisted in conceptualising the idea for the manuscript. CS, MM, MB, MP, DM, BT, LT, JS, and SN made substantial contributions to drafting the manuscript, searching the literature, and interpreting results. All authors participated in drafting the final version of the manuscript and gave final approval of the version submitted and any revised versions.

The data for this study were collected as part of the Australian Institute of Marine Science’s Northwest Shoals to Shore Research Program, supported by Santos as part of their commitment to better understanding Western Australia’s marine environment.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.