Abundance and behaviour of detritivorous fish as indicators of reef degradation in the Pemba Channel, Tanzania

Coral reef degradation is changing benthic habitats globally, with increasing transitions from structurally complex coral frameworks to rubble- and sediment dominated substrates. This study examined how habitat structure influences the abundance and behaviour of two surgeonfish, Naso brevirostris and Naso hexacanthus, across a coral–rubble gradient around Pemba Island, Tanzania, using Baited Remote Underwater Video systems (BRUVs). Forty deployments were analysed to quantify relative abundance (MaxN), behavioural prevalence and intensity, and temporal behavioural variation during each BRUV deployment. Depth and benthic composition were also analysed as environmental predictors. Abundance showed significant species-specific responses to habitat condition. N. brevirostris was associated with structurally complex coral habitats, whereas N. hexacanthus was more abundant in degraded rubble-dominated areas. In contrast, behaviour remained consistent, with no significant differences in behavioural prevalence or intensity between species or habitat types. Temporal analyses revealed short-term variation in feeding and social behaviours linked to changes in resource availability and aggregation dynamics, rather than habitat condition. Depth influenced the benthic composition but was not a significant predictor of fish abundance or behavioural responses. The findings suggest that habitat degradation changes the distribution of detritivorous fishes, while their behaviour remains stable across habitat conditions. The contrasting abundance patterns challenge broader generalisations of detritivore responses to reef degradation. Stable behaviour despite changes in species distribution suggests that key detrital processing functions may persist despite ongoing habitat simplification

Coral reef ecosystems are globally recognised as centres of biodiversity and productivity, underpinning ecological functions that support fisheries, coastal protection, and biogeochemical cycling (Hughes et al., 2017). However, these ecosystems are increasingly threatened by a range of anthropogenic stressors, including extensive pollution, overfishing and extractive processes such as mining (Duarte et al., 2025). Human-induced climate change has added further pressures, including intensifying storm events and causing shifts in ocean conditions (Bhuyan et al., 2026). Rising global sea temperatures and ocean acidification have accelerated coral mortality and habitat degradation, reducing the structural complexity of reef habitats worldwide (Bhuyan et al., 2026; Hughes et al., 2017). Mass coral bleaching has emerged as one of the most pervasive drivers of degradation. Repeated bleaching events weaken coral resilience, reduce live coral cover, and promote transitions towards algal- and rubble-dominated states (Prada et al., 2017).

In many regions, including Tanzania, these global stressors are compounded by local destructive practices, such as dynamite or blast fishing, which involves using explosives to kill large numbers of marine species (Slade and Kalangahe, 2015). This unsustainable fishing method physically shatters reef frameworks and rapidly converts complex coral structures into flattened rubble fields (Slade and Kalangahe, 2015). Such disturbances accelerate ecological phase shifts and fundamentally alter the benthic substrate on which reef communities depend (Ainsworth and Mumby, 2015).

These shifts typically result in the replacement of structurally complex coral habitats with simplified substrates dominated by turf algae and rubble (Ainsworth and Mumby, 2015). This transformation is often accompanied by expansion of the epilithic algal matrix (EAM), a layer formed of turf algae, detritus, sediment, and microbial biomass. The EAM acts as a major reservoir of organic material on degraded reefs (Tebbett et al., 2025; Wilson et al., 2003). The proliferation of EAM and rubble substrates is widely regarded as a signal of ecosystem collapse, as it reduces habitat complexity and alters resource landscapes for reef fishes (Tebbett et al., 2025; Graham and Nash, 2013). Yet not all species respond negatively to these changes. Some functional groups, particularly detritivores, may benefit from increased detrital availability in degraded habitats (Wilson et al., 2003).

Different fishes play important functional roles on coral reefs, and their presence or absence can offer insight into ecosystem condition (Samoilys et al., 2025; Schiettekatte et al., 2022). Planktivorous species, for example, rely on zooplankton delivered by water movement, and their distribution and feeding success are closely tied to hydrodynamic conditions that shape plankton availability (Holland et al., 2021; Hamner et al., 1988). As a result, shifts in planktivore abundance can reflect broader changes in water-column processes and reef health (Samoilys et al., 2025). Detritivorous fishes play a key role in nutrient recycling and energy transfer by consuming organic detritus embedded within the EAM (Wilson et al., 2003). Contrary to traditional classifications that labelled many benthic feeders as herbivores, evidence shows that numerous species derive most of their nutrition from detrital fractions rather than algal tissue (Goatley and Bellwood, 2010). Grazer-detritivorous surgeonfish (family Acanthuridae) demonstrate this fine-scale selectivity, choosing to sample organic-rich detrital fractions while avoiding inorganic sediment (Goatley and Bellwood, 2010). As reefs degrade and detrital resources accumulate, detritivores often increase in abundance, particularly in rubble- and sediment-dominated habitats (Eggertsen et al., 2019). This has led to suggestions that detritivorous fishes may serve as useful indicator species of reef degradation (Samoilys et al., 2025).

The ecological consequences of structural complexity loss extend beyond species-specific abundance and behaviour to broader trophic restructuring. Reef degradation has been shown to alter trophic composition, with herbivores increasingly dominating while planktivores decline, reflecting bottom-up assemblage restructuring (Huang et al., 2023). At a global scale, coral reef food webs share congruent trophic pathways maintained by dietary specialists, yet these pathways are highly vulnerable to biodiversity loss when structural complexity declines (Pozas-Schacre et al., 2021). These findings highlight that habitat simplification cascades through trophic levels, reshaping community resilience and ecosystem functioning.

Environmental variables such as depth, substrate composition, and structural complexity are well-established drivers of reef fish distribution and community structure (Darling et al., 2017). Depth gradients influence light availability, hydrodynamics, and benthic composition (Pérez-Peris et al., 2023), while substrate complexity shapes refuge availability and feeding opportunities (Santano et al., 2021; Graham and Nash, 2013). Existing studies in the Pemba Channel, a deep coastal channel in northern Tanzania, have shown that variation in depth and benthic structure explains substantial spatial differences in reef fish assemblages, including predatory species (Osuka et al., 2022). Despite this, behavioural mechanisms underlying these patterns remain comparatively understudied, particularly for detritivorous taxa. Behaviour represents a mechanistic link between environmental variation and ecological outcomes. Changes in environmental conditions can shape foraging decisions, habitat use, and species interactions (Nanami, 2025). However, in situ behavioural data are challenging to obtain, particularly in areas that are difficult for SCUBA divers to access (Langlois et al., 2020). Baited Remote Underwater Video systems (BRUVs) have emerged as a powerful tool for quantifying both abundance and behaviour, providing standardised observations of species’ responses to exogenous food stimuli (Langlois et al., 2020; Cappo et al., 2004). They additionally offer several practical advantages: unlike other methods, they are nonextractive, can be deployed at a wide range of habitats and depths and are relatively inexpensive to operate (Whitmarsh et al., 2017). BRUVs elicit species-specific attraction, feeding, and competitive behaviours, offering insights into foraging flexibility and behavioural plasticity under different environmental conditions (Dorman et al., 2012; Whitmarsh et al., 2017). Behavioural metrics derived from BRUV footage, such as approach frequency, feeding intensity, and interaction rates provide a valuable complement to traditional abundance-based assessments (Whitmarsh et al., 2017). However, abundance estimates derived from BRUV footage can be biased, as attraction to bait varies among species and individuals, potentially inflating or underrepresenting true densities. Similarly, behavioural observations are stimulus-driven and may not replicate natural conditions, meaning results must be interpreted within these methodological constraints (Langlois et al., 2020).Although BRUVs studies have predominantly focused on predatory and scavenging species, their application to detritivorous fishes offers novel opportunities to examine foraging flexibility in response to altered resource contexts (Dorman et al., 2012).

Despite growing recognition of detritivorous fishes as key contributors to ecosystem functioning on degraded reefs (Wilson et al., 2003; Lin et al., 2021), the behavioural mechanisms underpinning their responses to habitat change remain poorly understood (Eggertsen et al., 2019). Most existing research has focused on patterns of abundance or biomass following coral loss, demonstrating that detritivores often increase in rubble- and sediment-dominated habitats (Eggertsen et al., 2019). However, these metrics provide limited insight into the processes driving such patterns. In particular, they do not reveal how altered benthic conditions influence foraging decisions, competitive interactions, or habitat use (Nanami, 2025; Dall et al., 2005).

This distinction is critical because increases in detritivore abundance may arise through different ecological pathways. One possibility is behavioural flexibility, whereby individuals adjust their feeding strategies and resource use in response to environmental change (Wong and Candolin, 2015). Alternatively, elevated abundances may reflect opportunistic exploitation of the greater detrital resources available within the EAM on degraded reefs (Goatley and Bellwood, 2010; Wilson et al., 2003). Understanding these behavioural responses is particularly important because behavioural change is often the earliest signal of how organisms respond to human-induced environmental alteration (Gunn et al., 2022).

The Pemba Channel in northern Tanzania represents a particularly valuable system in which to examine these dynamics. The region supports some of the most diverse reef fish assemblages in the Western Indian Ocean, yet it has been subjected to extensive anthropogenic pressures, including overfishing, recurrent bleaching events and dynamite fishing (Slade and Kalangahe, 2015; Osuka et al., 2022). The Pemba Channel has a long history of dynamite fishing, particularly during the 1980s and 1990s when weak enforcement and rising fishing pressure made this destructive method widespread along the Tanzanian coast (Slade and Kalangahe, 2015). Around Pemba Island, repeated explosions shattered coral frameworks and produced extensive rubble fields that persist since loose rubble inhibits coral settlement and slows structural recovery (Kenyon et al., 2023; Grimsditch et al., 2009). Although blasting has declined, the surrounding areas of Pemba Island are now characterised by a high variety of coral reef conditions, with both healthy high complexity coral substrates and degraded areas with increased rubble and sediment (Grimsditch et al., 2009; Osuka et al., 2022). This patchwork of healthy and degraded habitats, particularly in shallow reefs (<20m), supports substantial variation in reef fish assemblages (Osuka et al., 2022). These habitat contrasts make the Pemba Channel an ideal system for examining how environmental gradients shape the distribution and behaviour of reef fishes.

This study aimed to investigate how habitat degradation influences the abundance and behaviour of two surgeonfish (family Acanthuridae, genus Naso) using BRUVs across a coral-rubble gradient on Pemba Island. It further explored the potential of Naso spp. to act as indicator species of reef degradation and the relative influence of environmental variables, particularly depth and substrate condition. To address these aims, the study tested three hypotheses: (1) Naso spp. abundance increases in degraded habitats relative to coral-dominated habitats, (2) Naso spp. feeding behaviour is more prevalent and intense in degraded habitats relative to coral-dominated habitats and (3) environmental variables such as depth and substrate complexity influence Naso spp. abundance and behavioural response.

Study context and video data source

This study utilised BRUV footage collected around the Pemba Channel, Tanzania (Figure 1), as part of a wider ecological survey investigating environmental drivers of reef fish assemblages (Osuka et al., 2022). The survey was conducted on the outer reefs along the western margin of Pemba Island, within the Pemba Community Conservation Area (PECCA). BRUV deployments were conducted across a range of depths and benthic habitat types, including hard coral, soft coral, rubble, sand, and mixed substrates. The original survey followed a standardised BRUV protocol, as per Langlois et al. (2020), deploying a single-camera system with a bait canister containing oily fish (families Scombridae and Carangidae) to create a consistent olfactory plume and attract fish. Each deployment recorded approximately 60 minutes of footage, alongside associated environmental data including depth and benthic composition. The survey was conducted during daytime hours over five days in November 2019.

A total of 79 BRUV deployments were available from the original dataset, representing a broad environmental gradient across the Pemba Channel (Figure 1).

Species selection and site shortlisting

This study focused on two surgeonfish species within the family Acanthuridae: Naso brevirostris (Spotted unicornfish) and Naso hexacanthus (Sleek unicornfish). These species were selected due to their relatively high occurrence across deployments, providing sufficient sample sizes for quantitative analysis. They were also chosen due to their ecological role in detrital feeding pathways (Eggertsen et al., 2019), and their suitability for behavioural observation in BRUV footage, directly relevant to the study objectives. Other detritivorous fish, such as parrotfishes (subfamily Scarinae) and rabbitfishes (family Siganidae), contribute to detrital processing (Eggertsen et al., 2019), but their low detectability in the recorded BRUV deployments limited the feasibility of robust behavioural comparisons in this dataset. Species identification was carried out using Lieske and Myers (1994).

A subset of videos was selected to enable detailed behavioural analysis. From the full dataset (n = 79), 40 deployments were shortlisted based on high MaxN values (defined as the maximum number of individuals of a species observed simultaneously within a single video frame). This ensured that behavioural observations were conducted under conditions where individuals were sufficiently abundant to display a range of ecologically relevant behaviours. MaxN is widely used in BRUV studies as a standardised measure of relative abundance, minimising the risk of repeated counting of individuals (Langlois et al., 2020).

Two deployments contained high MaxN values for both species and were therefore included in analyses for each species, resulting in a total of 42 behavioural recordings (21 per species). This targeted shortlisting approach maximised the likelihood of capturing meaningful behavioural variations while maintaining consistency in sampling effort across species.

Behavioural analysis framework

Defining behaviour

Behavioural analysis was conducted using a structured ethogram developed specifically for this study. Behaviours were categorised into functional groups relevant to feeding ecology (Choat et al., 2002) and BRUV interactions and defined clearly (Table 1) to minimise subjectivity and ensure consistency. The ethogram categories were established collaboratively by the authors (AL & KOE) prior to annotation, providing a shared framework for scoring behaviours.

Time frame selection

For each video, the focal species was observed during three standardised 5-minute time windows (Table 2), selected to represent distinct phases of bait availability and fish aggregation. The use of these windows aimed to investigate potential temporally dynamic fish responses to exogenous food.

Handling early MaxN events

In some deployments, MaxN occurred within the first 10 minutes of footage. In these cases, the defined T 1 window overlapped with T 0, compromising independence between time periods. To maintain consistency and avoid replication, such deployments were analysed using only T 0 and T 2. These cases were recorded and treated consistently throughout the dataset.

Behavioural scoring

Within each time window, the behaviour of the focal species was recorded and assigned a behavioural intensity score. Behaviours were scored on a scale of 1–3:

1- behaviour observed rarely or minimally within the 5-min window

2- behaviour observed occasionally or at moderate expression within the 5-min window

3- behaviour observed frequently, sustained, or involving multiple individuals within the 5-min window

This scoring system complemented presence–absence data by allowing differences in not only the occurrence but also capturing variation in how frequently or vigorously behaviours were expressed.

To minimise observer bias, each behavioural category was accompanied by predefined criteria for scoring, agreed upon by the authors prior to analysis, providing a consistent definitional framework and all videos were analysed by a single observer, ensuring consistency across observations. MaxN values were initially determined by author (KOE) and subsequently confirmed independently by another (AL), constituting a cross-check of abundance data. Ambiguous behavioural cases were flagged during annotation and resolved through author discussion, ensuring that scoring decisions were not made unilaterally.

Environmental data and habitat characterisation

Environmental variables were incorporated to assess how habitat conditions influenced behavioural responses of surgeonfish to exogenous food stimuli. Two key environmental predictors were used: depth and benthic habitat composition.

Depth (m) was recorded for each BRUV deployment during the original survey and treated as a continuous variable. Depth is a well-established driver of reef fish assemblage structure and behaviour, influencing factors such as light availability, habitat composition, (Scott et al., 2022) and the dispersal of bait plumes.

Benthic habitat composition was derived from the original survey dataset which quantified the percentage cover of major substrate categories (hard coral, soft coral, seafans, sponge, halimeda, macroalgae, seagrass, rubble, sand). Percentage cover was visually estimated from each BRUV video recording in the original survey, and for the 40 shortlisted sites these values were subsequently extracted for data analysis (Appendix 1). These variables describe the structural characteristics of reef habitats and are known to influence the distribution and feeding ecology of fishes (Choat et al., 2002).

Data analysis

Habitat structure analysis

To identify the main gradients in substrate composition across sites, principal components analysis (PCA) was applied to benthic habitat variables using Euclidean distance in PAST v.4 (Hammer et al., 2001). The first principal component (PC1) captured the dominant axis of variation and was therefore retained for subsequent analyses. PC1 scores were used to classify BRUV sites into two habitat categories: coral sites (negative PC1 values, n= 19 sites) and degraded sites (positive PC1 values, n=21 sites).

Abundance analyses

MaxN values for N. brevirostris and N. hexacanthus were compared using Wilcoxon rank-sum tests for each species. A two-way negative binomial Generalised Linear Model (GLM) with Species, Habitat, and their interaction was then fitted to assess combined effects on abundance. Model residuals were inspected to confirm assumptions of dispersion and fit.

Behavioural analyses

Behavioural prevalence was calculated as the number of occurrences of each behaviour across all analysed time intervals, expressed relative to the total number of observation intervals. Prevalence data was analysed as a categorical presence-absence measure for each behaviour for statistical testing. Behavioural intensity was scored on a 1–3 ordinal scale.

Differences in behavioural prevalence between habitats and species were tested using Fisher’s exact tests due to the categorical nature of the data (McDonald, 2014). Behavioural intensity differences were assessed using Wilcoxon rank-sum tests, comparing ordinal data between two groups.

Temporal behavioural analysis quantified behavioural prevalence and intensity across the three standardised time intervals (Table 2). Temporal changes in behaviour (intensity and prevalence) were analysed using repeated-measures non-parametric tests (Friedman tests). Stationary behaviour was not observed in any of the 42 BRUV deployments and was therefore excluded from subsequent statistical analyses.

Environmental driver analysis

A simple linear regression was conducted to examine the relationship between depth and the first principal component (PC1). Statistical significance was assessed using both the parametric t-test and a permutation test to confirm robustness to distributional assumptions.

Simple linear regressions were also fitted separately for each species with MaxN as the response variable and either PC1 or depth as the predictor. A multiple regression model including both PC1 and depth was fitted for each species. Standardised coefficients and p-values were used to evaluate the contribution of each predictor.

A binomial Generalised Linear Model (GLM) was used to test whether behavioural prevalence varied with depth. An ordinal logistic regression was also used to assess whether behavioural intensity varied with depth.

All statistical analyses were completed in RStudio (R Core Team, 2025).

Habitat structure

PCA revealed clear multivariate separation of benthic habitat types across survey sites. PC1 represented the dominant coral–rubble gradient, explaining 51.77% of total variance, while PC2 accounted for 19.95% (Figure 2). Sites with negative PC1 scores corresponded to coral-dominated habitats, whereas positive PC1 scores corresponded to degraded rubble-dominated habitats.

Abundance

Abundance (MaxN) differed significantly between habitats and species. A two-way negative binomial GLM detected a highly significant Species × Habitat interaction (p < 0.001).

Species-specific Wilcoxon rank-sum tests supported this pattern, with N. brevirostris significantly more abundant on coral sites (W = 334, p = 0.00098) and N. hexacanthus significantly more abundant on degraded sites (W = 134.5, p = 0.0466; Figure 3).

Behavioural prevalence and intensity

Fisher’s exact tests showed no significant species differences in behavioural prevalence (all p > 0.05), and Wilcoxon rank-sum tests found no significant differences in behavioural intensity between N. brevirostris and N. hexacanthus (all p > 0.05). A summary of all behaviours by species is shown in Figure 4.

Behavioural patterns did not differ between coral and degraded habitats for both species. Fisher’s exact tests showed no significant differences in behavioural prevalence for any behaviour (all p > 0.05), and Wilcoxon rank-sum tests indicated no significant differences in behavioural intensity between habitats (all p > 0.05; Figure 5).

Temporal behaviour

Friedman tests revealed significant temporal changes in the prevalence of three behaviours (Figure 6): solitary behaviour (χ² = 12.944, p = 0.001546), direct bait feeding (χ² = 11.045, p = 0.003996), and intraspecific competition (χ² = 9.1599, p = 0.01026). All other behaviours showed no significant change in prevalence across time (all p > 0.05).

Friedman tests also revealed a significant temporal change in the intensity of solitary behaviour (χ² = 9.2497, p = 0.009805), while all other behaviours showed no significant temporal change in intensity (all p > 0.05).

Environmental

Depth was positively associated with PC1, indicating that deeper sites were increasingly characterised by rubble-dominated substrate. The linear model (Figure 7) showed a significant positive slope (a = 0.1385 ± 0.0500 SE, t = 2.77, p = 0.0085), with an intercept of 20.02 ± 1.61 SE. 95% confidence intervals confirmed the robustness of this relationship. The model explained 16.5 % of the variance (r² = 0.165), confirming a moderate relationship between depth and habitat composition. Both the parametric (p = 0.0085) and permutation tests (p = 0.0084) indicated that the association was statistically significant.

Linear models were used to examine relationships between abundance (MaxN) and environmental variables (depth and PC1) for both species (Figure 8).

For N. brevirostris, the model with depth alone showed a significant negative relationship (slope = –0.254 ± 0.107 SE, t = –2.36, p = 0.023). The model with PC1 alone also showed a significant negative relationship (slope = –0.149 ± 0.031 SE, t = –4.80, p < 0.001), with abundance decreasing as habitats became less structurally complex. When both predictors were included, PC1 remained highly significant (slope = –0.136 ± 0.034 SE, t = –4.00, p < 0.001), while depth was no longer significant (p = 0.365). The combined model explained 38 % of the variance in N. brevirostris abundance (adjusted r² = 0.35).

For N. hexacanthus, the model with depth alone showed no significant relationship (slope = +0.235 ± 0.167 SE, p = 0.17). The model with PC1 alone revealed a significant positive relationship (slope = +0.180 ± 0.051 SE, t = 3.53, p = 0.001), with abundance increasing in low-complexity habitats. In the combined model, PC1 remained significant (slope = +0.176 ± 0.056 SE, t = 3.13, p = 0.003), while depth was not (p = 0.876). The model explained 24 % of the variance (adjusted r² = 0.20).

The binomial GLM revealed no significant relationship between behavioural prevalence and depth (all behaviours p > 0.05). Ordinal logistic regression showed no significant effect of depth on behavioural intensity (p = 0.064).

This study provides novel empirical evidence for habitat partitioning and behavioural resilience in detritivorous fishes across a coral-rubble gradient. These findings highlight species-specific responses to reef degradation and clarify how structural complexity shapes community assembly, with implications for understanding resilience and monitoring reef health.

Habitat structure drives abundance

Habitat structure emerged as the dominant driver of Naso spp. abundance in this study, aligning with extensive evidence that structural complexity underpins reef fish distributions and biomass at global scales (e.g. Graham and Nash, 2013; Wilson et al., 2007). Rather than a uniform detritivore response to degradation, the two focal species exhibited contrasting patterns along the coral–rubble gradient. N. brevirostris was significantly more abundant on structurally complex, coral-dominated substrates, whereas N. hexacanthus increased in abundance on rubble and sand-dominated habitats where detritus accumulates more predictably. This habitat partitioning was further supported by a highly significant Species × Habitat interaction. The divergent patterns of the two species highlight the need for species-specific interpretation rather than guild-level generalisations. This trend has been observed in several studies of Surgeonfishes, where acanthurids exhibit strong species-specific spatial distributions across environmental gradients (Cheal et al., 2012; Nanami, 2025).

The contrasting abundance patterns could be explained by niche partitioning rather than a simple degradation effect. Although both species are detritivorous, with similar functional roles, they occupy opposite ends of the substrate gradient, suggesting that they partition resources or habitat to reduce competition (García-Seoane et al., 2025). Studies of damselfishes have demonstrated that even closely related species segregate along fine-scale microhabitat axes, such as coral type, height and exposure, despite sharing similar diets (Hata and Kato, 2004). Similarly, parrotfishes have been found to partition resources along multiple niche axes including space, foraging mode and time, generating high functional diversity within a single functional group (Adam et al., 2015). This mirrors the abundance patterns observed in this study between two closely related detritivorous surgeonfish.

Biochemical studies further support this interpretation. Stable isotope and fatty acid analyses revealed that herbivorous and detritivorous fishes that appear functionally similar often occupy distinct trophic niches, assimilate different organic fractions, and rely on different foraging pathways (García-Seoane et al., 2025; Brandl et al., 2019). Species could partition these components in ways that are not clear when using broad functional labels such as “detritivore”. Thus, the divergent abundance responses of N. brevirostris and N. hexacanthus could reflect underlying trophic specialisation and habitat association.

However, it is also important to note that the use of oily fish bait in BRUV deployments can influence species detectability, as carnivorous and omnivorous taxa are more reliably attracted to scent plumes than detritivores or herbivores (Whitmarsh et al., 2017; Langlois et al., 2020). While surgeonfish species have been found to respond to bait odours (Delgado-Pech et al., 2020), attraction is not ubiquitous, and detectability may therefore vary among individuals and contexts. This recognised constraint of BRUV methodology should be considered when interpreting abundance patterns. Nonetheless, the consistent detection of Naso spp. across deployments indicates that the observed abundance trends remain ecologically informative, though absolute distributions may not be fully represented.

Additionally, while the PC1 axis captured major structural differences, it remains a relatively broad measure of habitat complexity. Finer-scale metrics, such as rugosity, fractal dimension, or profile, may reveal microhabitat preferences that further differentiate the two species (Knudby and LeDrew, 2007). However, the strong and opposing abundance responses observed here provide evidence that habitat structure drives abundance through species-specific ecological mechanisms.

Behavioural stability across habitats

Despite the clear habitat associations observed in abundance, there was no statistically significant difference in behavioural prevalence or intensity between the two Naso species. Across both species, the most prevalent behaviours were suspension feeding, exploratory movements, and schooling, while direct bait feeding and interspecific interactions were comparatively rare. Similarly, behavioural prevalence and intensity did not significantly differ between coral and degraded habitats, indicating that substrate composition did not alter the expression of the analysed behaviours. Contrary to our initial expectations, Naso species demonstrated similar behavioural responses regardless of substrate composition change, suggesting core function was maintained irrespective of structural complexity.

Behavioural stability between the two species indicates that niche partitioning in this system is driven predominantly by spatial segregation rather than by differences in behavioural strategies. Both N. brevirostris and N. hexacanthus expressed comparable behavioural patterns across habitats, despite occupying distinct positions along the structural gradient. Similar patterns have been documented in other reef fish assemblages, where species coexist through differential habitat use while maintaining broadly similar behavioural roles (Hata and Kato, 2004; Adam et al., 2015). This consistency suggests that the functional roles performed by each species are conserved across habitat types, reflecting a degree of behavioural flexibility. Adaptive responses to degradation have also been documented in other reef fishes, with dietary shifts serving as a mechanism of persistence (Clever et al., 2026). DNA metabarcoding studies reveal browsers switching from anthozoans on healthy reefs to annelids on degraded reefs, mediating the effects of degradation (Clever et al., 2026).

The absence of habitat-linked behavioural differences could be attributed to the feeding ecology of Naso spp. Detritus itself is not a homogeneous resource; its nutritional quality, particle size, and organic composition vary across substrates (Tebbett et al., 2017). Detritivorous fishes rely on broad, flexible foraging strategies, that are weakly tied to fine-scale structural features (Brandl and Bellwood, 2014; Mihalitsis and Wainwright, 2024). Such flexibility allows detritivores to maintain feeding and competitive behaviours across a range of substrate types, provided that detrital resources remain available. The consistency of feeding and movement behaviours in this study indicates that both species maintained their core functional roles irrespective of habitat type. This behavioural stability suggests that key ecological functions, including detritus removal, sediment turnover, and nutrient recycling, may be preserved even as structural complexity declines (Clever et al., 2026; Wilson et al., 2003).

The dominance of suspension feeding and exploratory movements is consistent with generalist foraging strategies that enable Naso spp. to exploit food resources across a range of habitat conditions rather than relying on highly specialised feeding tactics (Brandl and Bellwood, 2014; Mihalitsis and Wainwright, 2024). Suspension feeding reflects the trophic flexibility described for reef fishes exploiting suspended food resources (Mihalitsis and Wainwright, 2024), while exploratory movements are consistent with behavioural plasticity that facilitates resource acquisition under spatially heterogeneous conditions (Nanami, 2025). The persistence of these behaviours across both coral and degraded habitats suggests that Naso spp. maintain stable core foraging behaviours while adjusting habitat use in response to changing resource availability a pattern consistent with resilience strategies documented in reef fishes (Clever et al., 2026). The decoupling between abundance patterns and behavioural expression aligns with broader evidence that habitat change often alters species distributions before it affects behavioural repertoires. Behaviour tends to be more conserved than habitat use, with species shifting spatially to maintain preferred foraging conditions rather than modifying core behaviours (Brandl and Bellwood, 2014; Samoilys et al., 2025). While this ecological explanation is consistent with the observed patterns, it is important to consider the extent to which methodological constraints may also contribute to the absence of detectable behavioural differences. Behaviour was quantified using BRUV footage, which analysed short, stimulus driven observational windows, not directly mimicking natural conditions (Cappo et al., 2004; Langlois et al., 2020). Additionally, the prevalence and intensity metrics used provide coarse behavioural categories and may not detect subtle variation captured in other behavioural analyses that use specific metrics including bite rate (Ebrahim et al., 2020) or detailed time budget analysis (Altmann, 1974). Together, these constraints suggest that the absence of behavioural differences should be interpreted cautiously, as the methods used may not have been adequately sensitive to detect subtle or context-dependent variation. Additionally, while BRUVs can be valuable for documenting species distributions and broad behavioural categories, they do not replicate natural conditions and may artificially generate certain behaviours, limiting the reliability of behavioural inferences (Langlois et al., 2020).

Temporal dynamics

Although behaviour remained broadly stable across habitats, several behaviours exhibited short-term temporal variation throughout BRUV deployments. Significant temporal changes were detected in the prevalence of solitary behaviour, direct bait feeding, and intraspecific competition for both species. There was additionally significant temporal change in the intensity of solitary behaviour. These findings indicate that behavioural activity was influenced more strongly by immediate ecological conditions during deployments than by broader habitat structure.

These patterns suggest that observed behavioural fluctuations could be driven by short term ecological dynamics rather than by underlying habitat condition. Peaks in feeding and competitive behaviours around the MaxN interval are typical of BRUV deployments (Harvey et al., 2013; Whitmarsh et al., 2017). These dynamics are shaped by changes in bait plume dispersion, local flow conditions, and transient social aggregations around the camera unit, which influence fish arrival and interaction rates (Langlois et al., 2020). Coral reef environments are inherently dynamic systems in which behavioural interactions fluctuate over relatively short temporal scales in response to changes in local resource availability, fish density, and social interactions (Nanami, 2025; Brandl and Bellwood, 2014). Feeding opportunities on reefs are spatially and temporally variable, particularly for detritivorous species exploiting suspended or particulate organic matter within the water column and EAM (Lin et al., 2021).

The observed temporal peaks therefore likely reflect natural short-term shifts in ecological conditions and social dynamics rather than stable habitat-associated behavioural differences. Such patterns are consistent with previous studies demonstrating that reef fishes exhibit considerable behavioural flexibility, adjusting feeding effort and interaction rates in response to transient environmental stimuli while maintaining relatively stable functional roles across habitats (Brandl and Bellwood, 2014; Samoilys et al., 2025). Alongside the lack of strong habitat effect, significant temporal variability supports the idea that both species may be relatively resilient to small-scale environmental heterogeneity at the behavioural level.

Evaluation of Naso species as indicators of degradation

The divergent habitat associations observed between N. brevirostris and N.hexacanthus suggest that broad functional classifications, such as detritivorous Acanthuridae, are insufficient for monitoring reef health. Samoilys et al. (2025) posit that effective bio-indicators must exhibit high specificity and a predictable response to environmental change. These results indicate that family-level or functional-level indicators may mask species-specific partitioning across environmental gradients (Samoilys et al., 2025).

Furthermore, the assumption that detritivores increase solely due to increased resources on degraded reefs is an oversimplification. While rubble fields aggregate benthic detritus within the EAM, complex coral matrices can also act as “biological sieves” that concentrate high-quality organic matter within sheltered crevices (Lin et al., 2021). Additionally, environmental fluctuations, including storm-induced resuspension, can also increase detrital availability within healthy coral zones, equalizing resource landscapes (Madin et al., 2014). This suggests that the distribution of these species may be more driven by the accessibility of detritus and the requirement for structural refuge than by a net increase in food biomass following degradation (Tebbett et al., 2017). Ultimately, N. hexacanthus emerges as a robust abundance-based indicator for degraded habitats, given its significant association with simplified substrates.

Increased abundance of this species in low-complexity sites positions this species as a valuable indicator for future assessments of structural loss, reflecting a clear biological response to the simplification of reef landscapes (Samoilys et al., 2025; Graham and Nash, 2013). However, the lack of significant behavioural variation across the coral-rubble gradient suggests these species do not provide sensitive behavioural indicators.

Relative influence of environmental drivers

Substrate composition significantly influenced the abundance of both species, whereas depth was not a significant driver of abundance or behaviour. Structural complexity (PC1) emerged as the dominant predictor of abundance, overriding the influence of depth in all combined models. For N. brevirostris, the initial negative correlation with depth was secondary to the effect of PC1. Similarly, N. hexacanthus abundance was driven solely by PC1, increasing significantly in simplified habitats.

The positive association between depth and PC1 explains the observed spatial patterns: deeper sites were more likely to be degraded, rubble and sand-dominated environments. However, the multivariate results confirm that Naso species respond directly to structural composition, rather than being constrained by depth itself. Consequently, the results suggest that loss of reef complexity (PC1) is an important factor driving detritivore community assembly, irrespective of the depth at which degradation occurs.

While habitat composition significantly mediated abundance, behavioural prevalence and intensity remained stable across both the PC1 and depth gradient. This suggests that while environmental conditions dictate where these species occur, they do not fundamentally alter how they interact with resources. This behavioural conservatism reflects the observed stability of core functional behaviours, such as feeding and schooling, despite substantial variation in habitat structure and species abundance.

As coral reefs continue to lose structural complexity, understanding how individual species reorganise within changing seascapes is essential for predicting the future of reef functioning. Rising sea temperatures, recurrent mass bleaching, and ocean acidification are expected to accelerate coral mortality and further reduce reef structural complexity. This study shows that closely related detritivores respond to habitat degradation in different ways, emphasising that species-level ecological strategies shape how communities restructure under disturbance. Despite these contrasting spatial patterns, both Naso species maintained stable behavioural roles, suggesting that key detrital processing functions may persist as coral cover declines. Such behavioural conservatism indicates resilience, with ecological processes like sediment turnover and nutrient recycling likely to remain intact despite habitat simplification. This demonstrates that some species can persist in rapidly changing reef ecosystems by adjusting their spatial distributions while retaining the functional behaviours, such as sediment turnover and nutrient recycling, that support ecosystem stability. More broadly, these findings emphasise the value of combining species-specific habitat associations with behavioural data when anticipating how reef communities will alter under continued degradation. The limits of behavioural flexibility must be recognised, as expected escalating disturbance may exceed species’ adaptive capacity. Integrating species-specific behavioural data into projections of reef futures therefore provides critical insight into how ecological functions will be sustained or lost under scenarios of ocean warming and repeated bleaching events. This study highlights the need for management approaches that protect not only biodiversity, but also the processes that sustain reef productivity in an era of accelerating coral loss.