Introduction

Animal behaviour is strongly influenced by predation risk (Sih et al. 1985; Lima and Dill 1990), with animals responding rapidly to a perceived threat by increasing predator-avoidance behaviours such as fleeing and sheltering (Lima 1987; Brown and Laland 2003; Ferrari et al. 2010). In addition to natural predation, predator-avoidance behaviours are substantially affected by human hunting practices in the terrestrial environment (both lethal and non-lethal) (Kilgo et al. 1998; Frid and Dill 2002; Matson et al. 2005; Stankowich 2008). Research has demonstrated that hunting primarily affects individual vigilance levels (Frid and Dill 2002; Ordiz et al. 2012). In this respect, flight initiation distance (FID; the distance predators can approach prey before the prey flees) of numerous terrestrial taxa has been shown to increase with hunting pressure (Thiel et al. 2007; Jayakody et al. 2008; Stankowich 2008; Reimers et al. 2009) and is higher outside areas protected from hunting than inside them (Frid and Dill 2002; Stankowich 2008).

Similar behavioural effects of human harvesting have been reported from marine systems, prompting suggestion that fishing may be inducing a widespread ‘exploitation-induced timidity syndrome’ (Arlinghaus et al. 2017). Research in coral reef ecosystems has found that spearfishing increases the FID of numerous coral-reef fish taxa (Feary et al. 2011; Januchowski-Hartley et al. 2013, 2015), for example, the FID of surgeonfish (Acanthuridae) and parrotfish (Scaridae) increasing along a gradient of spearfishing pressure in Papua New Guinea (Januchowski-Hartley et al. 2011). The effect of fishing on FID is typically investigated indirectly by comparing behaviour both inside and outside of areas protected from fishing (e.g., Marine-Protected Areas [MPAs]). This is done because directly measuring fishing effort and its effects is rarely achievable over suitable spatial and temporal scales. FID investigations to date have focused on a limited range of coral reef taxa (although see Cole 1994), despite apparent species-specific responses and the potential for ecological interactions and environmental factors to mediate behavioural responses (Nunes et al. 2018, 2019; Quadros et al. 2019; Stamoulis et al. 2019; Pereira et al. 2020). Expansion of FID investigations to additional species and other ecosystems that experience fishing pressure is, therefore, warranted, to inform management of potentially impactful activities (Samia et al. 2019). FID of fish is also influenced by body size, with larger individuals typically displaying greater FID (Gotanda et al. 2009; Feary et al. 2011; Samia et al. 2019), as well as shoal size and the presence or absence of a spear, the reported effects of which have been mixed depending on the study (Januchowski-Hartley et al. 2011, 2012; Tran et al. 2016). For example, Tran et al. (2016) found that FID of lined bristletooth (Ctenochaetus striatus) was significantly greater in the presence of a spear gun and varied depending on whether the encounter was inside or outside of an MPA. In contrast, Januchowski-Hartley et al. (2012) found that FID of parrotfishes differed between protection status but was not influenced by the presence or absence of a spear. The ecological context and species examined must be considered when attempting to understand the impacts of fishing on FID, along with the multitude of potential confounding influences.

There is still little understanding of the mechanisms responsible for FID responses to fishing pressure, with fishing practices expected to have both direct and indirect effects on FID. The act of spearfishing, which is an active form of fishing, could directly increase FID by inducing fear in individuals subject to ‘near misses’ (Sbragaglia et al. 2023). Fishing could also indirectly increase FID by exposing non-target individuals to visual and chemical fear cues from captured individuals (Chivers et al. 2002; Brown and Laland 2003). Such a mechanism is also more likely for line fishing, which is a more passive form of fishing that does not require underwater approach of fish. Although the majority of studies have focused on conspecific social learning, such learning also occurs between heterospecifics (Griffin 2004), with evidence that predator-avoidance behaviour can be rapidly passed between both closely related and phylogenetically distant coral reef fishes (Manassa et al. 2013). In this way, effects of fishing on predator-avoidance behaviour may extend beyond just the target species.

The southeast coastline of Australia provides a model system to examine the effects of fishing practices on the behaviour of temperate reef fishes. This region has one of the highest levels of recreational fishing pressure within Australia (both spear- and line fishing), with the most recent estimate of fishing effort totalling approximately 1.7 million fisher days (Murphy et al. 2020). The region also supports several areas that are protected from both spear- and line fishing practices. These areas are both well-established (old) and enforced (Turnbull et al. 2018).

We aimed to determine if the predator-avoidance behaviour (FID) of temperate reef fishes is higher in areas where recreational fishing (both spear- and line fishing) is permitted than in areas protected from fishing, and whether this response is species-specific. We hypothesised that a greater effect of protection on FID would be observed for species that are more heavily targeted by recreational fishing, as has been observed for some species in other systems (Januchowski-Hartley et al. 2011; Sbragaglia et al. 2018). We used a mixed modelling approach to test the effect of fishing on FID while simultaneously controlling for the effect of body size and accounting for spatial variability in the response.

Materials and methods

Between April and July, 2012, the FID of temperate fishes was assessed in three areas along Australia’s New South Wales coastline: Port Stephens (32°42′19.79″S, 152°10′15.98″E), Sydney (33°49′58.82″S, 151°17′47.32″E) and Wollongong (34°35′28.55″S, 150°54′7.72″E). Protected sites within these areas were Fly Point Sanctuary Zone (termed Fly Point) in Port Stephens, Cabbage Tree Bay Aquatic Reserve (termed Cabbage Tree) in Sydney and Bushrangers Bay Aquatic Reserve (termed Bushrangers Bay) in Wollongong. Fished sites were Harasti’s Hole (Port Stephens), Long Bay (Sydney), and Shell Cove (Wollongong), each located near, but not directly adjacent to, the protected sites in each area. Protected sites were established in 1983, 2002, and 1982 for Port Stephens, Sydney, and Wollongong, respectively, and fishing within them by any method is prohibited. All locations were 10’s to hundreds of kilometres apart. Benthic habitat within both protected and non-protected sites was primarily composed of 0.5–1.5 m diameter boulders (urchin barren habitat; Underwood et al. 1991), at 1–3 m in depth, with areas ranging from 1050 to 9825 m2.

We focused on six fish species: two that are common targets for both spear and line fishers in NSW (yellowfin bream, Acanthopagrus australis [F. Sparidae] and luderick, Girella tricuspidata [F. Kyphosidae]), one species that is considered a secondary target for both spear and line fishers (red morwong, Morwong fuscus [F. Cheilodactylidae]), one species that is occasionally targeted by line fishers but protected from spearfishing (blue groper, Achoerodus viridis [F. Labridae]) and two species that are not targeted by either spear or line fishers (crimson-banded wrasse, Notolabrus gymnogenis [F. Labridae] and rock cale, Aplodactylus lophodon [F. Aplodactylidae]). Due to low availability in some regions, red morwong and rock cale could only be measured at Sydney and Wollongong, while blue groper could only be measured at Sydney and Port Stephens.

Flight initiation distance

Two observers (DAF and AMF) swam directionally on snorkel throughout reef habitat at each site to reduce the likelihood of inadvertently re-measuring the same individuals. Focal species that were foraging or moving slowly near the benthos, and that could be approached directly, were chosen for FID assessment separately by observers. Observers approached the chosen focal species in a horizontal swimming position at the fish’s depth, without a spear or other fishing apparatus. Benthic features where the fish was positioned were noted to precisely mark the individual’s position. When the individual started to flee, as indicated by an increase in speed, often accompanied by a change in direction, the observer dropped a weighted marker at their own position (following Feary et al. 2011). A second marker was placed at the benthic feature where the fish was located prior to flight initiation. The distance between the two markers was measured to the nearest cm using a tape measure and the size of the individual (total length, LT) was visually estimated to the nearest 5 cm. The latter value was considered the limit of reliable distinction between size classes at a range of distances. If the fish exhibited a change in behaviour that was not obviously a result of the approaching snorkeler (e.g., it was disturbed by another fish), the trial was abandoned. Starting points and the survey direction for each observer were haphazardly assigned prior to surveys, to avoid overlap of areas (and focal individuals) sampled between observers and to avoid repeatedly surveying the same areas.

Data analysis

The potential effects of protection from fishing (protected vs fished) and body size (LT) on FID were tested for each species using generalised linear mixed models (GLMMs; Bolker et al. 2009). Site was treated as a random effect nested within protection, because each site was either protected or fished. The most parsimonious combination of fixed effects (protection, size, protection × size) was identified using model selection, based on relative model fit and parsimony determined by Akaike’s Information Criterion (AIC). For the most parsimonious model, Wald tests were used to examine the null hypothesis that the coefficient = 0 for each parameter.

Data were explored before analysis using boxplots, Cleveland plots and scatterplots following the protocol of Zuur et al. (2010). The most suitable error distribution was selected for each model through observation of diagnostic plots (see below) and an AIC comparison of equivalent model structures employing the normal distribution with an ‘identity’ link, normal distribution with a log link, and gamma distribution with a log link. For all species, the gamma distribution with a log link performed best and was used for all subsequent modelling. Adherence to model assumptions was verified visually using standard model diagnostic plots, including residuals versus fitted values to examine homogeneity and a histogram or Q–Q plot of the residuals for normality.

Modelling was done in R (ver. 4.2.3, R Foundation for Statistical Computing, Vienna, Austria) using the glmmTMB function from the ‘glmmTMB’ package (Brooks et al. 2017). Diagnostics for these models were produced using the ‘DHARMa’ package (Hartig 2022).

Results

FID of all species was higher in fished areas than protected areas, irrespective of the degree of targeting by recreational fishers (Table 1; Fig. 1). However, the magnitude of the effect varied among species, with FID ranging between 2.4 and 1.4 times higher in fished compared to protected areas (Table 1; Fig. 1). The fishing effect was greatest for the two commonly targeted species, yellowfin bream and luderick (Table 1). The full model selection results for each species are presented in Appendix 1.

Table 1 Parameter estimates for the best models of flight initiation distance (FID) determined for six temperate fishes in NSW. Yellowfin bream and luderick are common target species, red morwong and blue groper are secondary targets, and crimson-banded wrasse and rock cale are not targeted by recreational fishers
Full size table
Fig.1
figure 1

Partial effects of protection level on flight initiation distance (FID, cm) from generalised linear mixed effects modelling for six temperate fishes in NSW: a yellowfin bream; b luderick; c red morwong; d blue groper; e crimson-banded wrasse; f rock cale. Dark grey panels indicates common target species, light grey indicates secondary targets, and white indicates species not targeted by recreational fishers. The effect of protection on FID depended on body size for blue groper (d); the dotted line indicates the protection effect while the solid line indicates the fished effect. Data points indicate raw data. Error bars indicate 95% confidence intervals. Note the y-axis does not extend to zero and varies among species

Full size image

FID increased with body size for all species except luderick (Table 1; Fig. 2). Each additional cm of length increased FID by a factor ranging between 1.02 and 1.05 (Table 1). There was no apparent pattern of effect size with the degree of targeting; the greatest effect was seen for red morwong (secondary target) and slightly smaller and similar effects were observed for the other species (Table 1). For blue groper, protection interacted with body size, such that the rate of FID increase with body size was greater in fished areas compared to protected areas (Table 1; Figs. 1, 2).

Fig. 2
figure 2

Partial effects of body size on flight initiation distance (FID, cm) from generalised linear mixed effects modelling for five temperate fishes in NSW: a yellowfin bream; b red morwong; c blue groper; d crimson-banded wrasse; e rock cale. No effect was identified for luderick. The relationship for blue groper depended on protection level; the dotted line indicates protected areas while the solid line indicates fished areas. Dark grey panels indicate common target species, light grey indicates secondary targets, and white indicates species not targeted by recreational fishers. Data points indicate raw data. Error bars indicate 95% confidence intervals. Note axes do not extend to zero and vary among species

Full size image

Discussion

This study demonstrates reduced predator-avoidance behaviour of both target and non-target temperate reef fishes inside areas protected from fishing. While common target species exhibited the greatest FID response to protection, partially supporting previous findings that FID increases with fishing intensity (Januchowski-Hartley et al. 2011; Sbragaglia et al. 2018), all six species in the current study were significantly affected, including two species not known to be targeted by fishers in southeastern Australia. The result for rock cale is particularly notable, because this species is both undesirable for fishers and herbivorous (Yiu et al. 2018), meaning even accidental line catches are uncommon. Fishing pressure for this species is therefore extremely low (A.M.F. pers. comm.), yet mean FID was 1.8 times higher in fished areas than protected areas. Our results are similar to a finding for a tropical surgeonfish (Ctenochaetus striatus) in French Polynesia, which also exhibited greater FID outside of MPAs despite not being harvested (Tran et al. 2016). However, differences in FID between protected and unprotected areas in that study were only observed in the presence of a spear, which our observers did not use. Our results contrast with those of Januchowski-Hartley et al. (2013) who found no increase in FID for a non-targeted family of fishes (Chaetodontidae) across an MPA boundary despite finding an effect for targeted species. Our results add to a growing body of research that supports complex effects of fishing on FID that are dependent on species and ecological context.

Given both target and non-target species were affected in the current study, the mechanisms responsible for increased predator-avoidance behaviour in response to fishing may be direct, indirect, or both. Species subject to targeted fishing may develop heightened predator responses via ‘near misses’, yet non-target species are much less likely to experience such direct encounters. However, individuals of non-target species may have indirectly learned to avoid humans in fished areas through social transference of behaviour from target species (Chivers et al. 2002; Brown and Laland 2003; Askey et al. 2006). There is a wealth of research (predominantly within terrestrial and freshwater literature) showing that social facilitation of alarm cues and/or predator recognition can have a substantial impact on the response of predator-naïve individuals (Brown and Laland 2003; Griffin 2004; Fernö et al. 2006). Socially transmitted information on predator presence or identity is an important mechanism for changing a fish’s behaviour; such socially transmitted information is relatively widespread in teleost species (Brown and Laland 2003). Therefore, predator-naïve individuals could be alerted to the presence of potential predators via their associations with other experienced conspecifics (Magurran and Higham 1988) or heterospecifics (Krause 1993).

While socially transferred fear responses have obvious short-term benefits for fish, such as escape from novel predators, longer-term exposure to such cues may have a net negative effect on fitness. This may be particularly true for non-target species, because fishing poses minimal risk, and heightened predator-avoidance behaviour is, therefore, unlikely to reduce total mortality. This behaviour may, however, suppress essential foraging or reproductive behaviours which would reduce fitness if sustained. Given that recreational fishing is a sustained pressure in populated areas, it is likely that any negative effects of fishing on the feeding or reproductive behaviour of non-target species would be maintained through time. Further investigation of the potential negative consequences of heightened predator-avoidance behaviour in response to fishing is, therefore, warranted.

The patterns in FID observed for target species in the current study may have been influenced by differences in abundance between fished and protected areas. Although the effect varies among species, locations and times, numerous studies have identified higher abundances of fished species in ‘no-take’ protected areas compared to fished areas in southeastern Australia, including the protected sites investigated in the current study (Kelaher et al. 2014; Harasti et al. 2018a, 2018b; Turnbull et al. 2018). Specifically, the two common target species (yellowfin bream, luderick) and one secondary target species (red morwong) in the current study were found to be more abundant in protected areas than fished areas (Harasti et al. 2018a; Turnbull et al. 2018). Fish have been shown to be more tolerant of approach when in larger schools (Stankowich and Blumstein 2005), in contrast to other taxa (e.g., reptiles, birds) which have greater flight initiation distances when in larger groups (Ydenberg and Dill 1986; Stankowich and Blumstein 2005). Hence, the lower FID found for yellowfin bream, luderick and red morwong in protected areas in the current study may be due in part to their greater abundance in those sites relative to fished areas and the likely larger schools or closer proximity resulting from this. Interestingly, a previous study found no difference in the abundance of blue groper between Cabbage Tree Bay Reserve (one of our protected sites) and fished areas (Turnbull et al. 2018). The lower FID found for this species in protected areas in the current study may not, therefore, be related to differences in abundance.

The impact of protection on FID found in the current study was surprising, given the small size of each protected area surveyed (Fly Point: 0.14 hectares; Cabbage Tree: 20 hectares; Bushrangers Bay: 3.88 hectares) and the large home-ranges of numerous species. For example, both yellowfin bream and luderick undertake substantial daily, monthly, and yearly coastal movement, encompassing 10–100 s of km (Gray et al. 2012; Pollock 1982). Such movement would be predominantly within areas open to fishing. Therefore, differences in FID response between protected and fished areas for these fishes may have resulted from a rapid switch in predator vigilance. Although there is still little information on the time it takes for fish to show a change in behaviour from being naïve to showing an aversion to fishers, evidence from freshwater systems indicates fish can become wary of fishing practices within days of exposure (Askey et al. 2006). In fact, angling pressure (within freshwater lakes) can have substantial and rapid impacts on the ‘catchability’ of fishes, with sustained fishing effort resulting in quick and substantial drops in catch rates over a 30 day period (Askey et al. 2006; Cooke et al. 2013). Future research should attempt to examine the spatio-temporal dynamics of fishing-induce predator-avoidance behaviour, to understand the relative impacts on mobile versus sedentary species within a ‘patchwork’ of protected areas.