Chum dine with me: assessing the effects of wildlife tourism on non-target fish assemblages

Introduction

Wildlife tourism is popular worldwide (Trave et al. 2017; Patroni et al. 2018), with the variety of opportunities to view and interact with wildlife growing rapidly, partly owing to the global growth of protected habitats and increased accessibility to remote areas where wildlife occurs (Orams 2002). When managed sustainably, it can provide substantial socio-economic benefits by offering a non-consumptive use of wildlife and contributes up to ~US$45 billion in annual global revenue (Rizzolo 2021). Wildlife tourism also provides employment opportunities for the local community, supports innovative research (Buckley 2009), and raises awareness for conservation and education (Apps et al. 2018).

Tourists’ expectations for memorable moments and unique experiences often relies on up-close encounters with wildlife (Hausmann et al. 2018), leading to increased pressure for operators to provide consistent and prolonged interactions with wild animals (Orams 2002). As a result, tourism operators target species that naturally aggregate and often use an attractant to provide up-close viewings (Knight 2009). Food-based attractants are the most common form of provisioning for wildlife tourism (Meyer et al. 2022) and can have various impacts on target species (Orams 2002; Patroni et al. 2018), but also on non-target species (Rizzari et al. 2017; Clarke et al. 2022, 2023; Dennis et al., in review). Food-based attractants can have various effects on fishes, including increased abundance of dominant species (Milazzo et al. 2005; Feitosa et al. 2012; Prinz et al. 2020); decreased diversity (Ilarri et al. 2008); changed distribution, daily movements, and residency, by enticing fish to occupy areas nearer the food source (Rizzari et al. 2017; Wen et al. 2019; Clarke et al. 2022; Pini-Fitzsimmons et al. 2023); habituation behaviour (Cole 1994; Hémery and McClanahan 2005; Milazzo et al. 2005; Albuquerque et al. 2014; Wen et al. 2019); aggression towards conspecifics and other species (Milazzo et al. 2005; Ilarri et al. 2008; Cruz de Paula et al. 2018; Wen et al. 2019) including humans (Hémery and McClanahan 2005; Ilarri et al. 2008; Cruz de Paula et al. 2018); anxiousness (Geffroy et al. 2018; Wen et al. 2019); or decreased foraging on natural preys (Prinz et al. 2020). For example, bread is regularly used to feed fish in the Cook Islands and has resulted in decreased richness during feeding times with an increase in abundance of certain trophic groups, leading to potential species dominance and reducing evenness (Prinz et al. 2020). Similarly, a study in the Abrolhos Bank also noted an increase in fish abundance at fish feeding sites, with some species dominating the assemblages (Feitosa et al. 2012). These studies highlight the changes fish feeding can have on community structures and by extension ecosystem function.

Delivering unnatural food source to fish through food-based provisioning can also affect fish health and condition through increased risk of crateriform lesions in the stomach wall, hepatic carcinomas, acarid parasites, and papillomas (Brookhouse et al. 2013) because of overhandling by tourists, increased exposure to bacteria, viruses, and fungi, and increased parasite loads causing lesions. Fish in feeding areas often have higher levels of corticosterone and cortisol, the two major stress hormones (Geffroy et al. 2018). Understanding the implications and health risks of food-based provisioning during wildlife tourism is vital for management strategies and tourism regulations, as well as being required for marine protected area and conservation strategies because many areas and species targeted are of conservation importance.

While provisioning aims to attract target (i.e. focal) species, non-target species can also be attracted and feed on bait. For example, non-target rodents, rabbits, and hares consumed up to 98% of the feed used to entice deer and quail into the viewing or hunting area (Donalty et al. 2003). Non-target fish and rays also consume food used to attract sharks (Meyer et al. 2020; Clarke et al. 2022, 2023; Dennis et al. 2024). Fish that inhabit the area but do not directly feed on the bait and berley can also be affected, owing to increased predation risk from more aggressive predators (Milazzo et al. 2006), which creates food competition (Prinz et al. 2020), and theft of local resources (Cruz de Paula et al. 2018). Despite the potential influence of provisioning on non-target species, most studies assessing the impacts of wildlife tourism have predominantly focussed on target species, with only 7% of wildlife tourism studies assessing the effects on non-target species or the broader ecosystem (Trave et al. 2017). Management frameworks have also identified the need to account for the impacts of feeding on non-target species to ensure a comprehensive assessment of the sustainability of wildlife tourism industries (Higginbottom et al. 2003; Meyer et al. 2021).

Shark-based tourism is conducted across the globe and often involves provisioning activities (Cisneros-Montemayor et al. 2013; Araujo et al. 2017). In Australia, the shark-diving industry generates at least AU$47 million per year (Huveneers et al. 2017), with white shark tourism in South Australia generating the second-largest revenue after the snorkelling with whale shark industry off Ningaloo Reef, Western Australia. White shark, Carcharodon carcharias, cage-diving is a sought-after tourism experience, because of their roles as top predators, threatened conservation status, and low probability of seeing them naturally in the wild (Huveneers et al. 2017). This form of tourism occurs in five countries, namely, Australia, New Zealand, South Africa, the United States of America (USA), and Canada, with all but the USA using food-based attractants (i.e. bait and berley) to bring white shark in proximity of the underwater cages. Mexico also previously offered white shark cage-diving experiences, but this industry closed in 2023 as a result of permitting restrictions on operators (Sosa-Nishizaki 2023). Similarly, the South African cage-diving industry has become unreliable as a result of orca predations displacing white sharks from the known aggregations (Towner et al. 2022). As a result, this has likely increased demand on the New Zealand and South Australian cage-diving locations for both tourists and documentary productions. The Neptune Islands Group Marine Park is currently the only location where white shark cage-diving is permitted in Australia. Two licensed operators are permitted to use food-based attractants. Both operators use tethered pieces of southern bluefin tuna, Thunnus maccoyii (referred to as bait), and berley (minced T. maccoyii) to attract white sharks to the cage-diving vessels. This bait and berley are sustainably sourced from the local tuna fisheries and is part of the bioregion’s natural fauna. Despite the industry targeting white sharks, smaller particles of bait and berley disperse into the water column and are being fed on by non-target species, with potential implications on the health and behaviour of these individuals (Meyer et al. 2020; Dennis 2024). Diet analysis indicated that several fish species feed on the bait and berley, including trevally, Pseudocaranx spp., horseshoe leatherjackets, Meuschenia hippocrepis, and yellowtail kingfish, Seriola lalandi, alongside visual observations of behavioural change, for example, demersal species coming from ~25 m deep to the surface to feed or species being attracted to the boat even before berley is released (Meyer et al. 2020). Other studies have found that operations lead to changes in the daily residency and movements of yellowtail kingfish at the Neptune Islands, in addition to raised activity during operating times (Clarke et al. 2022, 2023). However, body condition of yellowtail kingfish at the Neptune Islands was comparable to individuals from other control sites of similar oceanographic conditions. This suggests that behaviour changes did not lead to overall changes in body condition or raised energy expenditure was compensated by feeding on the southern bluefin tuna used as bait. Of the eight species assessed by Meyer et al. (2020), four showed different diets between the tourism and control sites. Whether this translates to effects on the abundance and composition of fish assemblages remains unknown, but has been identified as a research priority to assess the sustainability and acceptability of this industry (Meyer et al. 2021).

Our study serves as the first assessment of whether the food-based provisioning occurring at the Neptune Islands affects fish assemblages. Specifically, we tested whether food-based provisioning alters the overall fish assemblage, increases abundance of species that feed on the bait and berley, or decrease the number of species that either do not feed on the bait or could be prey of species attracted to the bait. We also assessed the evenness of the assemblage which, along with abundance and number of species, provides an indication of whether the bait and berley leads to some species dominating the assemblage. We assessed fish at two sites with different amount of food-based provisioning being used and compared those to six reference sites that have no berley or bait input.

Materials and methods

Study sites

The South Australian white shark cage-diving industry is restricted to the Neptune Islands Group (also known as the Ron and Valerie Taylor) Marine Park, which encompasses two island groups located ~70 km south of Port Lincoln, namely, North Neptune Islands Group (35.3375°S, E 136.1199°E; hereafter North Neptunes) and South Neptune Islands Group (35.3375°S, E 136.1199°E; hereafter South Neptunes) situated 11 km south of North Neptunes (Fig. 1). At these two island groups, two companies use a near-constant plume of food-based attractants, i.e. berley (minced southern bluefin tuna, Thunnus maccoyii, products, such as blood, offal, and oil) and tethered baits (gills and stomachs, or small ~3–5 kg body sections of T. maccoyii), to attract white sharks into the view of tourists (Huveneers et al. 2018). White shark cage-diving primarily (~85% of the time) occurs at North Neptunes, owing to it being closest to Port Lincoln, having reliable sightings historically, and to suitable anchoring locations in most weather conditions (Meyer et al. 2020). Because of the small size of the Neptune Islands, we could not consider any accessible areas to be unaffected by provisioning activities and, thus, chose reference locations at other island groups. This allowed us to develop a study with two levels of provisioning (high at North Neptunes, and low at South Neptunes), along with multiple reference locations that can be used to assess the natural variability in fish assemblages and how those assemblages compare to potentially affected assemblages at the Neptune Islands.

Fig. 1.

Map showing the study sites and their respective marine bioregions. Esri Service Layer Credits: Esri, TomTom, Garmin, The Food and Agriculture Organization of the United Nations, National Oceanic and Atmospheric Administration, The United States Geological Survey.

Fish assemblages were also recorded at the following six reference offshore islands with similar habitat and oceanographic conditions (Fig. 1), where commercial shark cage-diving is prohibited: Dangerous Reef (34.8156°S, 136.2125°E); Liguanea Island (34.9895°S, 135.6214°E); The Pages (North and South) (35.7771°S, 138.2919°E); Flinders Island (33.7166°S, 134.4842°E), and Pearson Island (33.9617°S, 134.2663°E). All locations, except for North and South Pages, are within the Eyre Bioregion of South Australia. The Pages (North and South) are within the Coorong Bioregion, but they are also temperate offshore islands with habitat structure similar to that in all other locations (i.e. macroalgae-dominated reefs and seagrass meadows) and have breeding populations of pinnipeds similar to those in the Neptune Islands (Goldsworthy and Page 2009). All locations should therefore contain similar assemblages of fishes and allow for comparisons of abundance and diversity to determine the influence of feeding from cage-diving operations.

Results

Assemblage structure

Each offshore island had different assemblage structures and were typically most similar to geographic neighbours (Fig. 2). We found that North and South Neptunes had significant overlap in assemblages, along with their closest neighbour Liguanea Island. Pearson Island and its closest neighbour Flinders Island also had a significant overlap. Dangerous Reef situated further north of the Neptunes and being more sheltered within the Spencer Gulf had a relatively distinct assemblages compared to other locations which had more overlap. Sites furthest to the east, i.e. North and South Pages, also had significant overlap and were the most distinct from the Neptunes. These were the only two islands sampled within in a separate bioregion from that of Neptunes. We observed significant differences between provisioning and reference areas for the fish assemblage across all habitat types (Table 1), but all island groups were equivalently distinct from each other for reef locations (Fig. 3). Because of lower numbers of deployments in other habitat types, e.g. sand, similarity values are not shown.

Fig. 2.

CAP ordination plot showing differences in fish assemblages among locations and habitat types. Choice of m = 30. Mis-classification error = 52%. Trace and delta P = 0.001.

Table 1.PERMANOVA results for the differences in fish assemblages across impact (North and South Neptunes vs other locations), location (fixed, eight levels) nested within impact, and habitat type (fixed, four levels).

Factor	d.f.	MS	Pseudo-F	P
Year (covariate)	1	9752.2	4.36	0.001
Impact	1	53,197	23.77	0.001
Habitat	3	45,051	20.13	0.001
Location (impact)	6	11,398	5.09	0.001
Impact × habitat A	2	11,467	5.12	0.001
Location (impact) × habitat	9	5576	2.49	0.001
Residual	323	2238.2

Year was included as a covariate, unique permutations ranged from 996 to 999. Significant values are shown in bold. Pairwise tests for the interaction of impact and habitat were significant (P = 0.001) for all habitat types (except ‘other’ which did not run).

^A Not all habitats were present at all locations.

Fig. 3.

SIMPER dissimilarity values showing differences between location pairs for reef-only deployments (larger number represent more dissimilar pairs). Green indicates highly significant (P < 0.001), whereas yellow indicates significance of between P = 0 001–0.05. Grey indicates that no test was performed because of low sample size.

Abundance

Total abundance

We found the average total abundance at North Neptunes (high provisioning; 72 ± 5.7 individuals per deployment) and South Neptunes (low provisioning; 71 ± 7.5) to be significantly higher than that of reference areas (no provisioning; 37–57 ± 4.6–7.1) (Table 2; Fig. 4).

Table 2.PERMANOVA results showing the differences in total abundance, species richness, and evenness (J′) across impact (North and South Neptunes vs other locations), location (fixed, eight levels) nested within impact, and habitat type (fixed, four levels).

Factor	d.f.	Total abundance		Species richness		Evenness (J′)
		Pseudo-F	P	Pseudo-F	P	Pseudo-F	P
Year (covariate)	1	43.1	0.001	0.1	0.815	25.2	0.001
Impact	1	18.5	0.001	3.1	0.072	2.3	0.107
Habitat	3	10.0	0.001	30.9	0.001	2.8	0.573
Location (impact)	6	1.5	0.191	8.7	0.001	0.9	0.042
Impact × habitatA	21	1.4	0.255	0.04	0.961	2.3	0.42
Location (impact) ×habitat	9	0.4	0.888	3.2	0.001	1.1	0.011
Residual	323

Year was included as a covariate, unique permutations ranged from 991 to 999. Significant values are shown in bold.

^A Not all habitats were present at all locations.

Fig. 4.

Boxplot showing differences among locations for total abundance of individuals observed on BRUVS deployments. The dashed lines show the mean value for North Neptunes (red) and South Neptunes (light red).

Key species

Large schools (~50–200 individuals) of trevally, Pseudocaranx spp., were observed sporadically at most sites, with smaller schools being more frequently observed at the Neptunes and The Pages (Fig. 5). Overall, trevally, Pseudocaranx spp., were observed on 73% of deployments, with a total of 5316 individuals being observed. We observed no significant differences in the abundances of trevally, Pseudocaranx spp., between provisioning and reference locations (Table 3).

Fig. 5.

Boxplots showing differences among locations for abundance for trevally, Pseudocaranx spp. (top), and horseshoe leatherjacket, Meuschenia hippocrepis (bottom), observed on BRUVS deployments from all habitat types. The dashed lines show the mean value for North Neptunes (red) and South Neptunes (light red).

Table 3.PERMANOVA results showing the differences among abundances of key species observed with year as a covariate, testing impact (North and South Neptunes (provisioning) vs other locations (reference)), location (fixed, eight levels) nested within impact, and habitat type (fixed, four levels).

Factor	d.f.	Pseudocaranx spp.		Meuschenia hippocrepis		Notolabrus tetricus		Caesioperca rasor
		Pseudo-F	P	Pseudo-F	P	Pseudo-F	P	Pseudo-F	P
Year (covariate)	1	27.3	0.001	21.1	0.001	1.3	0.258	26.6	0.001
Impact	1	0.8	0.38	109.8	0.001	0.1	0.786	41.6	0.001
Habitat	3	0.9	0.438	33.4	0.001	21.2	0.001	6.3	0.005
Location (impact)	6	1.5	0.182	13.1	0.001	0.5	0.762	4.0	0.005
Impact × habitatA	2	1.7	0.183	1.5	0.214	3.4	0.036	6.5	0.004
Location (impact) × habitat	9	0.8	0.493	0.4	0.886	0.3	0.962	0.3	0.949
Residual	323
Provisioning vs reference
Infralittoral reef		–	–	–	–	1.07	0.267	6.6	0.001
Soft sediment		–	–	–	–	3.4	0.001	0.3	0.806
Seagrass		–	–	–	–	0.9	0.336	0.3	0.783

Unique permutations ranged from 992 to 999. Significant values are shown in bold.

^A Not all habitats were present at all locations.

Two sites had high abundances of horseshoe leatherjacket, Meuschenia hippocrepis (up to 76 individuals per BRUVS), namely, North Neptunes (high provisioning site) and Liguanea Island (reference site) (Fig. 5). The high abundance at North Neptunes contributed to the significant difference observed between provisioning and reference locations (Table 3). All other sites had comparable abundances of M. hippocrepis.

All sites had similar abundances of blue-throat wrasse, Notolabrus tetricus, with abundances ranging from 0 to 38 individuals per replicate. We observed no significant differences between reference and provisioning locations for reef and seagrass areas but did observe significantly higher numbers of N. tetricus in soft sediment provisioning sites than in reference locations (Fig. 6, Table 3).

Fig. 6.

Boxplots showing differences among locations for abundance of blue-throat wrasse, Notolabrus tetricus (left), and barber perch, Caesioperca rasor (right), observed on BRUVS deployments in reef and sand habitats (seagrass and other habitats are not shown because of low deployment numbers and non-significance). The dashed lines show the mean value for North Neptunes (red) and South Neptunes (light red).

North and South Neptunes frequently had large schools of barber perch, Caesioperca rasor, whereas numbers of individuals were often lower in reference sites and very few individuals were observed at North and South Pages (Fig. 6). The range of individuals observed was from 0 to 77 across all sites. We found a significantly higher abundance of C. rasor in the provisioning sites than at reference sites for reef habitats only (Table 3), which was likely driven by the presence of sporadic large schools at the Neptune Islands. There was no significant difference in the abundance of C. rasor between North and South Neptunes (P = 0.22).

Diversity

Species richness

We found that species richness at North Neptunes (high bait and berley input) was similar to that at South Neptunes (low provisioning) and reference areas (no provisioning) with an average of 13 ± 0.4 taxa per deployment, compared with 12 ± 0.7 and 11–18 ± 0.5–1.5 respectively (Fig. 7). There was no significant difference between North and South Neptunes compared with the reference areas (Table 2).

Fig. 7.

Boxplots showing differences across locations for species richness and evenness. The dashed lines show the mean value for North Neptunes (red) and South Neptunes (light red).

Evenness

Evenness was also comparable across sites, with North Neptunes (high provisioning) and South Neptunes (low provisioning) similar to reference areas (no provisioning), with an average J of 0.72 ± 0.01 and 0.71 ± 0.7, compared with 0.66–0.85 ± 0.02–0.04 respectively (Fig. 7). There was no significant difference between North and South Neptunes compared with the reference areas (Table 2).

Discussion

The use of food-based attractants during wildlife tourism can have a multitude of effects on target species (Trave et al. 2017; Patroni et al. 2018), yet, the impacts on non-target species are often unexplored and difficult to assess. At the Neptune Islands, non-target fish consume the bait and berley used by the South Australian white shark cage-diving industry (Meyer et al. 2020) and this affects fish growth (Dennis et al., in review), daily movements, and residency (Rizzari et al. 2017; Clarke et al. 2022, 2023), prompting an assessment of the effects of this industry on the fish assemblage, abundance, and diversity. Although our study was not specifically designed to test for the effects of provisioning, it serves as an initial assessment of fish communities at temperate offshore islands and can be used to monitor potential changes in the future. Overall, we did not find any evidence suggesting that bait and berley affected fish assemblages at North Neptune Islands. Each offshore island assessed had unique fish assemblages, with those at North Neptunes not being more distinct than other locations. Total fish abundance was slightly higher at North Neptunes but species richness and evenness were comparable to those at other islands. Slight differences in abundance were observed for the individual species analysed, with horseshoe leatherjacket, Meuschenia hippocrepis, having higher abundances at the provisioning locations and barber perch, Caesioperca rasor, and blue-throat wrasse, Notolabrus tetricus, having higher abundances in soft sediment at provisioning sites than at reference locations. Therefore, individual species might be affected by the cage-diving industry. However, the abundance of M. hippocrepis at North Neptunes was similar to that in one of the reference sites, Liguanea Island, indicating that the higher abundance of this species may not be linked to provisioning.

Variations in fish assemblages and species diversity among locations is more likely to be due to natural geographical differences among locations (Gaston et al. 2008). Separate offshore islands have their own environmental and oceanographic conditions, as well as different stressors and conservation measures (Wantiez et al. 1997; Velmurugan 2008), which makes it challenging to disentangle which factors are affecting fish assemblages. For example, the no-take Sanctuary Zone at North Neptunes may have affected the fish assemblage. However, recreational fishing pressure at this island prior to the Sanctuary Zone implementation in 2014 was very low, because this island is ~70 km from the closest boat ramp (Port Lincoln) and requires medium to large vessels that typically do not target reef fishes but focus on pelagic fishes such as southern bluefin tuna, Thunnus maccoyii. Before-after-control-impact (BACI) designs can be an ideal experimental design in ecology, especially those with multiple control locations (Underwood 1994), but as cage-diving started at the Neptune Islands in the early 1980s, information about fish assemblages prior to wildlife tourism is not available. The influence of geographic variation on fish assemblages is highlighted by the similar fish communities at the two geographically closest locations, namely, North and South Pages. Fish assemblages at North and South Pages are also distinct compared with other locations, which can be explained by this island group being in a different bioregion (Coorong bioregion) from the other locations (Eyre bioregion). Similarly, Dangerous Reef also had high dissimilarity values, potentially owing to geographic differences between this island and others (smaller, with less reef habitat). Comparing multiple reference locations helps account for these differing environmental, anthropogenic, and oceanographic conditions at each island group, and allow us to account for natural variation in fish assemblages when assessing whether fish communities at the Neptune Islands have been affected by the bait and berley used when cage-diving.

Large schools of trevally (up to ~3000 individuals), Pseudocaranx spp., can be observed actively feeding on the bait and berley next to the cage-diving boats (Dennis et al. 2024). It has been hypothesised that the large number of Pseudocaranx spp. may act as a buffer species by feeding on bait and berley at the surface, limiting access and consumption of the bait and berley by other species nearer the benthos (Meyer et al. 2020). This was suggested on the basis of a diet study, showing that only Pseudocaranx spp. and yellowtail kingfish, Seriola lalandi, had fatty acid signatures that reflected the consumption of the bait and berley (Meyer et al. 2020) at North Neptunes. In contrast, some benthic species also had fatty acid signatures indicating that the berley may reach to lower levels of the water column at South Neptunes where lower numbers of Pseudocaranx spp. are present at the cage-diving vessel. On the BRUVS, we observed similar abundances of Pseudocaranx spp. among all sites and were not able to detect the large schools dominating assemblages around the cage-diving vessels at North Neptunes. The lack of differences observed might be due to the conservative measure used to quantify abundance (MaxN), which is known to underestimate large abundances because of screen saturation (Stobart et al. 2015; Whitmarsh et al. 2018). As such, MaxN may not be sensitive enough in high-density environments to detect increases in species abundance beyond a maximum observable threshold (Whitmarsh et al. 2018). This would result in a hyper-stable relationship, whereby catchability decreases as true site abundance increases (Kilfoil et al. 2017; Whitmarsh et al. 2018). However, we did not observe large schools of Pseudocaranx spp. that would have saturated the field of view. Our benthic-BRUVS might, therefore, not adequately sample the large Pseudocaranx spp. schools typically seen around the surface at the cage-diving vessels. Although Pseudocaranx spp. are usually reef-associated or bentho-pelagic species, they have a tendency to school mid-water at North Neptunes and are more abundant on pelagic-BRUVS at North Neptunes than at South Neptunes and Dangerous Reef (S. K. Whitmarsh, unpubl. data). This highlights the need for more pelagic-based sampling to encompass pelagic or mid-water species (Rees et al. 2015, 2018; Clarke et al. 2019). However, presence of white sharks at the tourism site makes pelagic sampling challenging because of increased likelihood of interactions with sharks for pelagic-BRUVS than benthic-BRUVS. Further research on the Pseudocaranx spp. at Neptune Islands has been conducted recently, assessing their abundance by using mark–recapture methods, along with the fine-scale movement of individuals, and physiological aspects (Dennis 2024; Dennis et al. 2024). The swim pass method was assessed to be the best method to estimate Pseudocaranx spp. abundance (Dennis et al. 2024). This method lured Pseudocaranx spp. with bait past a camera. The fine-scale movement study also showed that the space use of Pseudocaranx spp. was much smaller in the presence of operators and that most Pseudocaranx spp. were occupying the upper 5 m of the water column (near the operator vessels). These results highlight the difficulty of assessing the Pseudocaranx spp. population by using traditional methods such as BRUVS or trawling, which, despite showing promise in other studies (e.g. Osgood et al. 2019; French et al. 2021), were not effective at capturing the Pseudocaranx spp. in the present study because of the modified behaviour of this species in the presence of operator vessels.

Meuschenia hippocrepis is also regularly observed feeding on bait and berley at the surface where cage-diving occurs (Meyer et al. 2020). Although the natural diet of M. hippocrepis is mostly based on algae and sponges (Jones 1992; Rodgers et al. 2013), it is known to be attracted to fish-based baits and is commonly caught as by-catch in the commercial rock-lobster fishery (Brock et al. 2007; Rodgers et al. 2013). It is, therefore, likely that this species is an opportunistic scavenger that will consume carnivorous baits when available. In addition to observations of M. hippocrepis feeding on bait and berley, biochemical analyses reflected the consumption of the bait and berley at South Neptunes (Meyer et al. 2020). Similar changes in the diet of a benthic feeder in relation to tourism provisioning has previously been shown in threadfin butterfly fish, Chaetodon auriga (Prinz et al. 2020). This species is normally corallivore, feeding on live coral (Cole and Pratchett 2014), but consumed bread during provisioning activities. The abundance of M. hippocrepis followed the expected trend for a species affected by wildlife tourism provisioning, with the highest abundance recorded at North Neptunes, followed by South Neptunes. However, high abundance of M. hippocrepis was also observed at Liguanea Island, and to some extent at Pearson Island, with large schools being sighted at several other locations. Therefore, the bait and berley use at North Neptunes might be affecting the abundance of this species, but not to the point of reaching unnatural levels.

Barber perch, Caesioperca rasor, also showed a response, with significantly higher abundance in reef habitats at the Neptune Islands Group than at other offshore islands. This species schools in large numbers over rocky outcrops and feeds on zooplankton (Gomon et al. 2008). This species has not been observed to feed directly on the bait and berley near the cage-diving vessels. Their increased abundance could be explained by either natural variation and habitat suitability at the Neptune Islands, because they are known to form large schools in other areas (e.g. Wilson’s Promontory in Victoria) (Turner and Norman 1998), or the uptake of bait and berley by other species could potentially increase food availability for C. rasor, therefore increasing their abundance. Trevally, Pseudocaranx spp., consume zooplankton as part of their natural diet (Meyer et al. 2020) and if these individuals are instead consuming bait and berley, there may be less competition for zooplankton, freeing up resources for other species such as C. rasor. Research into the diet of C. rasor at the Neptunes Islands may provide insight into whether they are directly feeding on bait and berley.

Provisioning at the Neptune Islands may also be affecting the behaviour of species, with higher abundances of blue-throat wrasse, Notolabrus tetricus, but only on soft sediment habitats. Provisioning activities may therefore be encouraging species to spend time in untypical areas, because N. tetricus was rarely sighted on soft sediment habitats at the reference sites. Provisioning can change space use and result in individuals inhabiting areas close to the food source that may be outside their usual habitat (Wen et al. 2019; Clarke et al. 2022; Pini-Fitzsimmons et al. 2023). Habitat heterogeneity at the Neptune Islands may have also influenced this result because reef was the primary target for most reference locations, leading to BRUVS deployments on soft sediment being mostly incidental and likely close to reef. Future research should generate detailed seafloor maps such as those derived from multibeam echo-sounders to determine habitat distributions and patch size. This information is critical to better understand the influence of habitat on demersal fish assemblages and how this may have affected our results.

Conclusions

Our study serves as a first look into whether food provisioning associated with the use of bait and berley by the white shark tour operators influences fish assemblages at the Neptune Islands by examining assemblage structure, species abundances, and diversity using demersal BRUVS. We found no evidence that the South Australian white shark cage-diving industry impacted demersal fish assemblages substantially, but it likely contributed to overall high fish abundance, driven by changes from a small number of species. Fish assemblages were distinct among island groups, but the variation between the Neptune Islands Group and the reference locations was similar to that among reference locations, supporting that food provisioning does not markedly alter the structure of the Neptune Islands fish community. Large schools of Pseudocaranx spp. that are commonly observed around the shark-cage diving vessels were not detected on our benthic BRUVS in high abundances, indicating that there may be other effects from food provisioning that we could not quantify. Overall, while some effects were observed, food provisioning at the Neptune Islands does not appear to alter demersal fish assemblages, nor increase the abundance of bait-feeding species beyond natural variability, or decrease species diversity at an ecosystem-wide level. This suggests that the current level of provisioning has limited observable ecological impacts on the demersal fish community and highlights that the management regulations may be suitable for non-target fish assemblages to ensure the sustainability of the white shark cage-diving industry.