The Spermonde Archipelago lies in central Indonesia within the Coral Triangle, between the southern arc of Sulawesi and the Makassar Strait in South Sulawesi Province. It comprises ca. 70 islands, of which around 50 are inhabited (Ferse et al., 2014, Figure 1 ). Coastal urbanization of Makassar started in the 17^th century during the Dutch colonial period (Knaap and Sutherland, 2004) and accelerated after the 1950s, driven by economic growth and migration (Hugo, 1982). The city then expanded into suburbs, prompting substantial land-use changes and land expansions along coastal areas (Surya et al., 2020, 2021). Since 2017, a massive land reclamation project has created 157 ha for the “Center Point of Indonesia,” featuring business centers, housing, public facilities, and green spaces surrounding riverine outflows (Langkoke et al., 2022).

The rapid development in Makassar and its satellite cities has caused high river contamination, and most of these effluents reach the Archipelago as runoff from major rivers in the Makassar and Maros districts ( Figure 1 ). Habitat alteration, including coastal reclamation projects, has substantially increased nearshore sedimentation (Langkoke et al., 2022). Recent studies indicate that the inshore waters are polluted (Faizal, 2022; Lestari et al., 2021; Retnaningdyah et al., 2019; Tahir et al., 2018), with several water quality variables exceeding Indonesian standards. Elevated nutrient concentrations also reach offshore, where eutrophic waters expanded tenfold – up to nearly 10 km from the mainland – between 1996 and 2020 (Faizal, 2022).

To examine inshore-to-offshore urbanization effects, eight study sites were selected with increasing distances from the mainland (perpendicular to the South Sulawesi mainland shoreline): Lae-Lae (1LL, 1 km from mainland), Samalona (2SA, 7 km), Barang Lompo (3BL, 17 km), Bonetambung (4BO, 22 km), Badi (5BA, 23 km), Lumu-Lumu (6LU, 31 km), Karang Kassi (7KS, 39 km), and Kapoposang (8KP, 62 km, Figure 1 ). All sites, except 7KS which is a shallow reef bank, are on fringing reefs around inhabited islands. We chose the northwest side of each island or reef bank for surveys and experiments because this side has better reef development than other sides (Plass-Johnson et al., 2015), except at 8KP, where transects were laid on the northeast side for logistical reasons. Site 8KP lies within the Kapoposang marine protected area (MPA), located in a traditional fisheries zone that allows small-scale fishing. Field data collection was conducted in August 2022.

Four 1×1-m PVC quadrats were placed at 4–7 m depth at each site ( Figure 2a ), except for 1LL, where they were positioned at 3 m due to poor visibility. Given the diverse habitat characteristics across inshore-to-offshore gradients, quadrats were placed with two considerations: relatively flat substrates (slope <30 degrees) and having representative turf algal cover, which was ranging from 5% at 8KP to 40% at 1LL. Nine overlapping ~35×35-cm close-up images of the area enclosed within the quadrat were taken from above the quadrat grid and merged into a 1x1-m image using the automatic photo merge feature in Adobe Photoshop 2023^®. Furthermore, half of each quadrat (0.5×1 m) was carefully covered with a weighted transparent plastic sheet to serve as a control area.

Turf algae sediment loads (g/m²) and composition (inorganic/organic) were measured in the treatment area using three 10-cm diameter PVC rings placed on turf algae mats ( Figure 2a ). Sediment, which had settled within turf algal and may include bound detritus, within each ring was carefully suctioned using four to eight 200-mL syringes without damaging the algae ( Figure 2b ). After collection, the rings were removed. The entire treatment area was vacuumed with a SCUBA tank-powered suction device to create a “sediment-free” area ( Figure 2c ). This manipulation aimed to compare fish herbivory on substrata with and without sediments/detritus following Tebbett and Bellwood (2019). A high-resolution video camera (GoPro Hero 3 or 6) was positioned one to two meters from each quadrat, with four markers on the quadrat edges and two at the center of each horizontal/vertical frame within view. Subsequently, the camera recorded continuously for at least one hour after careful removal of the plastic sheet and quadrat ( Figure 2d ). Fish herbivory observations occurred between 11:00 and 14:00, coinciding with peak feeding activity for many herbivorous fishes (Bonaldo and Bellwood, 2008; Khait et al., 2013).

In the laboratory, very fine sediments were removed by sieving samples through a 58 µm mesh, as these sediments frequently became suspended in the water column and could be expelled from the syringe tips during data collection, potentially compromising the reliability of sampling for this fraction. Sediments larger than 58 µm were washed three times with distilled water, then filtered through pre-weighed filter paper (Whatman 41, pore-size 25 µm) using a peristaltic vacuum pump. The samples were dried for one hour at 105°C in a drying oven to yield total particle loads, with weights measured using an analytical digital scale (Mettler Toledo ME204E). To separate inorganic and organic sediment fractions, the total particle load was combusted at 650°C for four hours to oxidize organic matter (Zhang and Wang, 2014). Weight loss upon combustion indicated organic sediment, while the residuals represented inorganic sediment.

We used Coral Point Count with Excel extensions (CPCE; Kohler and Gill, 2006) to calculate the percentage of ten benthic categories (live corals, turf algae, macroalgae, soft corals, crustose coralline algae, cyanobacteria, sponges, other living organisms, recently dead coral, and sand) in each quadrat. Each benthic image was examined with 100 uniformly distributed data points, resulting in 50 points per treatment (with and without sediment removal). Benthic groups were classified by the topmost layer; e.g., dead coral covered by cyanobacteria was categorized as cyanobacteria. The turf algae category included short productive algal turfs (SPAT), long sediment-laden algal turfs (LSAT), and mixed turf algae assemblages. Differentiation was limited by image resolution, especially for images from inshore reefs. Only live corals, turf algae, and macroalgae data were analyzed due to their relevance to this study.

Herbivorous fish species are often classified into functional groups based on their contributions to herbivory (Ladds et al., 2018). We focused on five functional groups: detritivores, croppers, browsers, scrapers, and excavators, reflecting their distinct diets, feeding mechanisms, and ecological roles. Given the high diversity of herbivorous species recorded in the surveys (49 species), we followed the general functional group classification from www.datamermaid.org (accessed July 2023, Supplementary Table S1 ), which includes an extensive species list compiled from FishBase (Froese and Pauly, 2010). As herbivorous fishes may feed on various algae and microorganisms (Choat et al., 2002), potential overlap exists among the functional groups. Therefore, while functional groups are used as the unit of analysis, a closer look at individual species is included to provide deeper insights into their ecological roles in urbanized reef systems.

Due to poor water visibility at 1LL, small herbivorous fish (<5cm TL) were excluded at all sites to avoid identification bias. These small individuals contributed <1% of total recorded bites, making their impact negligible. The first and last 15 minutes of each video were discarded to eliminate diver disturbance effects during quadrat setup and recording. Herbivory rates were quantified over 30 minutes per video, totaling 14.5 hours across 29 quadrats at eight study sites. Four quadrats were filmed per site except at 2SA (n=3) and 3BL (n=2) due to camera malfunctions.

For each feeding fish, we recorded the species, functional group, estimated total length (TL, cm), bite count, and feeding location (treatment/control). These data were converted into mass-standardized bite rates, hereafter referred to as “herbivory rates” for simplicity, by multiplying the biomass of each fish (kg) by its bite count (Hoey and Bellwood, 2009; Plass-Johnson et al., 2015). Fish biomass was estimated via length-weight relationships (Kulbicki et al., 2005) using conversion parameters from FishBase (Froese and Pauly, 2010). Herbivory rates were expressed as “bites × kg/30 mins.” Fish species identification, TL, and bite counts were visually estimated from the videos by a trained observer (PLK) using the quadrat width as a reference.

We estimated reef rugosity and roving fish biomass within three 50 m transects at 7–9 m depth, slightly deeper than the quadrats (see 2.2), to minimize disturbance to fish herbivory during the experiment. Rugosity was measured using the chain intercept method (Hill and Wilkinson, 2004), where a 20 m chain was laid along the transect tape following the reef contour. A rugosity index (C) was calculated as C=1-d/l, where d is the horizontal distance the chain covered, and l is its fully extended length. A higher index indicates greater reef structure complexity. Roving fish surveys were conducted with underwater visual censuses (UVCs) along 5×50 m transects, with three replicates per site (Hill and Wilkinson, 2004). A trained observer (PLK) recorded detritivore, cropper, browser, scraper, and excavator fishes ≥5 cm TL within 2.5 m of either side of the tape. Fish identity and size (TL to the nearest cm) were recorded, and biomass was calculated using the length-weight conversion parameters described above.

We conducted a two-step analysis to test whether the composition of herbivory differed among sites and treatments. First, we used herbivory rate data from all five functional groups and performed a two-way analysis of similarity (ANOSIM, Clarke and Warwick, 1994) with treatments nested within study sites. Herbivory rates were fourth root-transformed, and Bray-Curtis similarity was used to measure the distance between herbivory rates on 9,999 permutations. The two-way ANOSIM showed significant differences in herbivory rates among sites (p-value <0.001, R=0.409, Supplementary Table S1 ) but no differences between treatments.

In the second step, we identified functional groups with different herbivory rates among sites. Kruskal-Wallis tests were used for detritivore herbivory rates due to the presence of only one species, while one-way ANOSIM was applied to cropper, scraper, and excavator rates. Browser herbivory rates were excluded, as they occurred at only one site. Treatment and control area data were averaged at the replicate level to avoid pseudoreplication, as the first analysis detected no effect of treatment. Herbivory rates were fourth root-transformed, and Bray-Curtis similarity was calculated before ANOSIM analysis. When statistical differences were observed, pairwise analysis and similarity percentages (SIMPER, Clarke, 1993; Clarke and Warwick, 1994) were used to identify sites with different herbivory rates and discriminating species driving the overall pattern. For both analysis steps, replicates with no herbivory data were removed to increase the analytical power, as the statistical methods were sensitive to zero values. Therefore, the analysis focused on sites where herbivory activity was recorded. While this approach provided a clearer picture of how the distance of the reef site to the mainland affects these rates, in some cases, this resulted in a low sample size.

Last, we employed multiple linear regression for detritivores and biota and/or environmental matching (BIO-ENV) analysis (Clarke and Ainsworth, 1993) for croppers, scrapers, excavators, and all herbivorous fish combined to determine the environmental and ecological factors influencing the herbivory rates. This analysis used six reef parameters (i.e., percentage of organic matter in turf algae sediments, turf algae, macroalgae, and live coral cover, reef rugosity, and total roving herbivorous fish biomass). We used the percentage of organic matter in turf algae sediment, calculated by dividing the organic sediment loads by total (inorganic + organic) sediment loads, considering that this variable is one of the best predictors of fish herbivory (Tebbett et al., 2024b). Due to differing replicate numbers, analyses were conducted in two steps: first, using the percentage of organic matter in turf algae, macroalgae, and live coral cover variables at the replicate (quadrat) level; second, using reef rugosity and UVC’s total fish biomass variables at the site level. For the second step, given the low sample size (between five and eight) and to increase statistical power, the biomass of roving herbivorous fish was aggregated across all functional groups to obtain total roving biomass, thereby reducing the number of independent variables.

Prior to analysis, multicollinearity among independent variables was assessed using Spearman’s rank correlation. Since no variables showed high collinearity (Spearman’s ρ < 0.8), all were retained. Herbivory rates were fourth-root transformed, while macroalgae and organic matter percentages were log-transformed to reduce skewness. After transformations, all variables were standardized to have a mean of zero and a standard deviation of one. Linear regression assumptions were checked using diagnostic tests, including Levene’s test for homoscedasticity, the Shapiro-Wilk test for normality, and QQ plots for residuals.

For the BIO-ENV analysis, a Bray-Curtis similarity matrix was calculated for herbivory rates of croppers, scrapers, excavators, and all herbivorous fish combined. BIO-ENV was then performed, applying Spearman’s rank correlation and a permutation test with 9,999 iterations to assess statistical significance. Directions were visualized through principal component analysis (PCA) when significant relationships were detected.

Given the exploratory nature of this study and the limited sample size, a significance threshold of p < 0.1 was used to identify potential patterns and relationships for further investigation. Some analyses (Kruskal-Wallis, linear regression, and PCA) were performed using R 4.2.3 (R Core Team, 2023), while other analyses (ANOSIM, SIMPER, and BIO-ENV) were done using PRIMER v7 (Clarke and Gorley, 2015).

Benthic reef condition showed spatial patterns along an inshore-offshore gradient ( Figure 3 ). Across the Spermonde Archipelago, percentage of organic matter within turf algae (mean ± SE = 29.0 ± 8.7 g/m²), reef rugosity (0.325 ± 0.009) and live coral cover (16.9 ± 3.3%) generally increased from inshore to offshore reefs, while inorganic and organic algal turf sediment loads (125 ± 64.7 g/m² and 14.6 ± 6.4 g/m², respectively), turf algae (22.8 ± 5.5%) and macroalgae (9.4 ± 4.5%) covers decreased. The nearest inshore site (1LL) was characterized as an algae-dominated reef, indicated by having 75% turf and macroalgal cover and 2% live coral cover, with Sargassum and Padina species thriving well at this site. In contrast, the offshore reefs (6LU to 8KP) displayed the lowest turf algae sediment loads (inorganic <1–14 g/m²; organic 2.7–3.7 g/m²), highest live coral cover (21–31%), and highest reef rugosity (0.298–0.407), indicating better overall reef condition compared to other study sites.

A shift in turf algae composition was observed from inshore to offshore reefs. The substrate at 1LL was dominated by silt, and long, dense, sediment-laden turf algae could be found covering dead corals ( Figure 3e ). Site 3BL ( Figure 3f ), in contrast, had a mixture of sand, live coral, and dead coral substrate, and the turf algae formed a dense assemblage with medium sediment loads. Meanwhile, mid-to-offshore reefs displayed a range of turf algae characteristics: from very sparse, short turf algae ( Figure 3g ) on relatively flat, smooth substrates to slightly longer, denser mixtures with cyanobacteria, small fleshy algae, or other microorganisms ( Figure 3h ) on rubbles or broken dead corals. Further, mid-to-offshore turf algae consistently had low sediment loads ( Figure 3a ).

We recorded 12,750 herbivorous fishes from 45 species across five families (Acanthuridae, Pomacanthidae, Pomacentridae, Labridae, and Siganidae) in 24 visual census transects at eight sites ( Supplementary Table S1 ). The average biomass of all herbivorous fish species ranged from 29.4 ± 22.7 (1LL) to 861.5 ± 202.1 (8KP) kg/ha, with an average of 389.5 ± 90.1 kg/ha ( Figure 4a ). These species were categorized by their feeding habits into detritivores (3 species), croppers (21 species), browsers (4 species), scrapers (13 species), and excavators (4 species, Figure 4 ). Fish biomass varied among functional groups, with browsers and detritivores showing the lowest biomass (31.9 ± 14.7 and 32.2 ± 14.6 kg/ha, respectively), while croppers and excavators exhibited the highest (138.9 ± 37.6 and 118.0 ± 34.8 kg/ha, respectively, Figure 4a ). Biomass generally increased offshore for all groups except scrapers, which decreased with distance from the mainland ( Figure 4 ). However, sites 6LU and 7KS showed reduced biomass for croppers, scrapers, and excavators, with browsers absent at 7KS ( Figure 4a ). A heatmap analysis ( Figure 4b ) revealed key species with high biomass within each functional group. For example, Siganus virgatus, S. vulpinus, and Pomacentrus species were major contributors among croppers, Ctenochaetus striatus among detritivores, and Chlorurus bleekeri among excavators. Almost half of the species recorded during the visual censuses (bold texts, Figure 4B ) fed within the quadrats. The site closest to the mainland (1LL) exhibited the lowest fish diversity and biomass, with only two cropper and one browser species (Dischistodus prosopotaenia, Siganus virgatus, and S. canaliculatus) present ( Figures 4a, b ). Similarly, browsers were not observed at all at 2SA.

A total of 25 herbivorous fish species from five functional groups were observed feeding within the quadrats: a detritivore (1 species), croppers (12 species), browsers (2 species), scrapers (8 species), and excavators (2 species, Figure 5a ). All species recorded performing herbivory were found roving in the surrounding reefs ( Figure 4b ), except for four cropper species ( Figure 5b , species names with asterisk). We recorded 1,538 individual fish taking 12,956 bites within the quadrats, ranging from 1 to 116 bites per fish. The mean herbivory rates across the Spermonde Archipelago were 27.1 ± 22.9 bites.kg.30 mins^-1, ranging from 0.6 (1LL) to 187.1 (7KS) bites.kg.30 mins^-1. Excavators had the highest average herbivory rate (53.7 ± 47.9 bites.kg.30 mins^-1), followed by scrapers (19.3 ± 6.5 bites.kg.30 mins^-1), and browsers with the lowest rate (1.2 bites.kg.30 mins^-1). High average herbivory rates by excavators were primarily driven by exceptionally high rates at 7KS, with values more than 25 times higher than those at other sites. At all sites except 7KS, scraper herbivory rates (<1–40 bites.kg.30 mins^-1) were generally higher than those of excavators (2-11 bites.kg.30 mins^-1). Herbivory rates for all functional groups typically increased from nearshore to offshore sites, though site-level variations were noted ( Figure 5a ). Cropper and excavator herbivory rates increased offshore, but cropper rates dropped at 6LU, while excavator rates showed a substantial increase at 7KS. Herbivory by detritivores, scrapers, and excavators was not observed at 1LL, 3BL (except scrapers), and 5BA, and browsers were only found feeding at 8KP.

The herbivory rates of all herbivore species differed significantly among sites (ANOSIM sites, p-value=0.001, R=0.409), but not between treatments ( Supplementary Table S2 ). Further analysis showed cropper and scraper herbivory rates contributed to site-specific differences (ANOSIM cropper, p-value <0.001, R=0.389; ANOSIM scraper, p-value <0.001, R=0.434, Supplementary Tables S3 , S4 ), while detritivore and excavator rates were similar across sites ( Supplementary Table S5 ). Pairwise analysis showed that several sites had different herbivory rate compositions. For example, cropper herbivory rates were significantly lower at 3BL and 5BA compared to 7KS and 8KP ( Supplementary Table S3 ), and scraper rates were higher at 7KS than at 2SA ( Supplementary Table S3 ).

Due to significant differences in cropper and scraper herbivory rates, we conducted SIMPER analysis to identify the primary discriminating species contributing significantly to group dissimilarity (noted by the highest average dissimilarity/Av.Diss in SIMPER results). Five cropper species (Astrosalarias homomelas, Dischistodus prosopotaenia, Pomacentrus burroughi, Siganus virgatus, and Zebrasoma scopas, Supplementary Table S6 ) and three scraper species (Scarus dimidiatus, S. quoyi, and S. rivulatus, Supplementary Table S7 ) primarily drove these differences. The distribution and herbivory activity of these discriminating species reflected environmental gradients from nearshore to offshore. For example, nearshore reefs were mainly influenced by the cropper D. prosopotaenia (1LL and 3BL) and the scraper S. quoyi (2SA), while the outermost site (8KP) exhibited unique discriminating species due to their high herbivory rates there, i.e., Z. scopas and S. dimidiatus. The remaining discriminating species mostly drove the differentiation of herbivory rates in the middle reef zone.

The composition of herbivory rates was affected by several reef variables that differed for each functional group ( Table 1 ). Reef rugosity was linked to detritivore herbivory rates (Multiple linear regression, p-value = 0.07). Overall, herbivory rates of the three functional groups and all herbivorous fish were consistently influenced by the percentage of organic matter in turf algae sediments (BIO-ENV results, p-value < 0.1, Table 1 , Supplementary Tables S8–S11 , Supplementary Figure S12 ). In addition, croppers were influenced by reef rugosity and roving fish biomass (BIO-ENVcropper, p-value = 0.08), scrapers by turf algal cover (BIO-ENVscraper, p-value = 0.02), and all herbivorous fish combined by macroalgae and live coral covers (BIO-ENVallherb, p-value = 0.02). The PCA results revealed distinct relationship directions between herbivory rates of different fish functional groups and environmental variables ( Supplementary Figure S12 ). For example, the percentage of organic matter in turf algae sediment was positively associated with excavator herbivory rates, particularly at 7KS, and negatively associated with cropper herbivory rates at 8KP.

Coral reef conditions in the Spermonde Archipelago vary with distance from the mainland. The inshore reef (1LL) showed high turf and macroalgal cover and elevated turf algae sediment loads, indicating poor reef health. Conditions gradually improved from 2SA to 5BA, with increasing rugosity, live coral cover, and decreasing turf algae sediment loads and macroalgal cover. Offshore reefs (6LU to 8KP) had higher hard coral cover and rugosity, suggesting healthier conditions. Roving herbivorous fish biomass generally increased offshore, peaking at middle-to-offshore sites, although roving scraper biomass decreased offshore. These findings align with earlier studies (see Edinger et al., 1998; Edinger and Risk, 2000; Plass-Johnson et al., 2018a; Teichberg et al., 2018), highlighting poor water quality from urbanization remains the primary factor shaping reef conditions and assemblages (Becking et al., 2006; Cleary et al., 2005; Girard et al., 2022; Kegler et al., 2017; Plass-Johnson et al., 2016; Polónia et al., 2015; Teichberg et al., 2018). Fish herbivory rates also varied across functional groups and distances from the mainland. Reefs closer to the mainland (1LL to 5BA) had lower herbivory rates, particularly for detritivores and croppers, suggesting higher susceptibility to algal overgrowth. Reefs further offshore exhibited increased herbivory, especially for scrapers and excavators. Several functional groups were absent in inshore reefs, and discrepancies were noted between fish species observed in the UVC surveys and those feeding in video recording. Such patterns are likely driven by poor water quality resulting from urbanization and island-based human activities, which shape fish herbivory composition through environmental filtering. This process selectively limits species capable of functioning under specific environmental conditions, e.g., high turbidity, while reducing the feeding activity of others (Bejarano et al., 2017).

Coastal urbanization factors affect herbivory composition differently across functional groups. Detritivore and excavator herbivory rates remained consistent across sites, suggesting stable roles in the urban Spermonde reef ecosystem. Although diet composition was not examined, past studies indicate many herbivores can adapt or alter their diets in response to environmental changes (Bonaldo and Bellwood, 2008; Gordon et al., 2016; Lin et al., 2021; Plass-Johnson et al., 2018b). For instance, Ctenochaetus striatus adjusts its diet based on coral cover and algal availability (Lin et al., 2021). Although detritivore herbivory stability is crucial for removing detritus and sediments settled within turf algae, the presence of only one detritivore species in a reef system impacted by urbanization and sedimentation highlights low functional redundancy and increased vulnerability to disturbance (e.g., Bellwood et al., 2003; Bonaldo and Bellwood, 2009). If C. striatus is functionally lost due to disease, predation, fishing, or severe environmental changes, it could lead to over-accumulation of detritus and sediments settled within turf algae, which thus can smother the corals. This detritus removal role may be fulfilled by other fish, such as parrotfish, known for their herbivore-detritivore function in sediment production and reworking on reefs (Perry et al., 2015). Excavators also maintained relatively high herbivory rates across spatial gradients, which may influence reef dynamics in complex ways. While excavators can help control algal growth and facilitate coral recruitment by clearing substrate (Bellwood and Choat, 1990; Bonaldo and Bellwood, 2009), they may also inadvertently damage small, newly settled corals or larger colonies during feeding (Huertas et al., 2021; Charendoff et al., 2023; Rempel et al., 2024). Their net effect is, therefore, context-dependent, shaped by spatial scale, coral resilience, and local conditions.

Current and past observations in the Archipelago (Plass-Johnson et al., 2015) show browsing by Naso species mainly at reefs further offshore (i.e., 5BA and 8KP, with observations in this study restricted to the latter). This contrasts with the Great Barrier Reef, where browsers like Naso unicornis significantly remove macroalgae in nearshore areas (Hoey and Bellwood, 2011; Lefévre and Bellwood, 2011). However, Plass-Johnson et al. (2014, 2015) also recorded other species, classified in different functional groups in this study, feeding on offered macroalgae, such as Siganus virgatus, S. corallinus, Zebrasoma scopas (croppers), Scarus flavipectoralis (scraper), and Bolbometopon muricatum (excavator), indicating dietary flexibility across species. Although these species may help curb macroalgae, observing browser herbivory only at 8KP remains concerning, given browser species were present at all sites except 2SA and 7KS ( Figure 3b ). Browsers can be selective, often favoring specific macroalgae even when less abundant (Kelly et al., 2016), and are influenced by the chemical properties of plants (Meyer et al., 1994). For example, N. lituratus avoids green algae like calcified Halimeda spp (Meyer et al., 1994), which are more common on inshore reefs of the Spermonde Archipelago. Inshore macroalgae may also contain chemical compounds from polluted runoff, potentially deterring browsers from feeding on them. Additionally, intensive and destructive fishing throughout the Archipelago may negatively impact herbivorous fish and herbivory. In a global review, Edwards et al. (2014) reported a 50% drop in herbivorous fish biomass in fished areas, with browsers most affected.

A few species drove differences in cropper and scraper herbivory rates. Their distributions differed across the spatial gradient, likely influenced by ecological factors such as habitat differences, food resource availability, and environmental conditions. For instance, the territorial damselfish D. prosopotaenia was among the five cropper species distinguishing 1LL from all sites except 8KP. This farmerfish defends its territory against other herbivorous fishes (Hoey and Bellwood, 2009; Plass-Johnson et al., 2018b), ensuring a consistent food supply despite environmental changes.

Furthermore, S. rivulatus was a key driver of scraper herbivory rate differences in the middle-to-offshore Archipelago (4BO and 7KS). This species demonstrates feeding flexibility across turf algae sediment sources, although it favors a high-quality epilithic algal matrix (EAM, sensu Wilson et al., 2003) with fine sediments and high nutritional content (Bonaldo and Bellwood, 2008; Gordon et al., 2016). The middle-to-offshore Archipelago (4BO and 7KS), less affected by coastal pollution, sedimentation, and oceanic waves, likely provides ideal conditions for S. rivulatus, with 7KS having the highest organic-to-inorganic sediment ratio in turf algae. Despite high S. rivulatus herbivory rates, this species was not observed in the UVCs, possibly due to methodological differences in measuring herbivory versus roving biomass and local conditions during data collection. Noises from frequent boat traffic at 4BO and blast fishing at 7KS, which were heard multiple times during data collection, may have affected fish movement and feeding patterns. Moreover, some species may be susceptible to diver presence and only be detected using video observations (Mallet et al., 2014).

The surgeonfish Z. scopas distinguished 8KP from other sites, where its herbivory was uniquely recorded, consistent with findings from Plass-Johnson et al. (2015). This species is likely sensitive to urban influence and, therefore, was most abundant and active feeding at the furthest site from the mainland. Its high herbivory rates at 8KP may stem from a flexible diet and capacity to exploit diverse microhabitats and lower-quality food resources (Robertson et al., 1979). In the competitive environment of 8KP, with high roving herbivore biomass and intense herbivory, Z. scopas thrive perhaps by consuming algae that other species do not efficiently consume. This adaptability enables Z. scopas to maintain high feeding rates despite competition. Preserving species with such dietary flexibility is vital for ecological balance and for reducing the risk of phase shifts to algal dominance in urbanized reefs like the Spermonde Archipelago.

Our study provides preliminary insights into environmental factors influencing herbivory rates across functional groups. For detritivores, herbivory rates were higher on lower-rugosity reefs. Syafruddin, Estradivari et al. (in review) reported increased sediment accumulation on flatter substrates in the Spermonde Archipelago, potentially favoring detritivore herbivory. For croppers, scrapers, excavators, and all herbivorous fish combined, herbivory rates were consistently associated with the percentage of organic matter within turf algae sediments, although the nature and strength of these associations varied across groups ( Supplementary Figure S12 ), underscoring the importance of food quality. Other variables also mattered: cropper herbivory was correlated with reef rugosity and roving fish biomass, scraper rates with turf algal cover, and combined herbivory with macroalgal and live coral cover. These findings highlight the complex dynamics shaping herbivory in urban reefs. However, the limited sample size necessitates cautious interpretation, and further studies with larger datasets are needed to confirm these patterns.

None of the functional groups showed a clear relationship between herbivory rates and roving fish biomass, except for croppers. This may reflect our limited sample size, which could have obscured potential patterns or local factors shaping these relationships. Visual observations revealed fewer fish feeding within quadrats than recorded in visual censuses, particularly in the middle zone (3BL to 6LU) where fish biomass peaked, yet herbivory rates varied or were even absent. While higher fish biomass often predicts greater herbivory, this relationship can be non-linear or non-existent due to factors such as species and size composition (Kindinger et al., 2024), habitat features (Jones and Andrew, 1990), coral cover, reef complexity (Williams et al., 2001, Verges et al., 2011), local fishing pressures, environmental conditions (Edwards et al., 2014; Plass-Johnson et al., 2015), and management (Humphries et al., 2014). Environmental filtering may prevent species from feeding under certain conditions despite their presence (Bejarano et al., 2017). In the Spermonde reef system, heavy fishing, destructive fishing methods, intense coastal anthropogenic activities (e.g., boat traffic), and urban influences from the mainland likely shape fish grazing behavior and herbivory across the Archipelago.

Unexpectedly, we found no significant differences in herbivory rates between sediment-cleared and control areas, suggesting that settled sediment within turf algae did not directly or immediately affect herbivory at our experimental scale. This contrasts with earlier studies reporting substantial herbivory increases (up to 225%) following sediment removal (Bellwood and Fulton, 2008; Goatley and Bellwood, 2012). Several factors may explain this discrepancy. First, sediment loads within turf algae in the Archipelago were generally below the 250 g/m² threshold at which herbivory is inhibited (Tebbett et al., 2024b), except at 1LL ( Figure 3a ). This implies that most naturally occurring sediment loads were insufficient to suppress herbivory; hence settled sediment removal had minimal impact. Second, turf algae assemblages might influence outcomes. For example, in Fiji, a mixture of cyanobacteria in turf algae reduced herbivory by up to 50% (Ford et al., 2021). While we observed inshore-offshore variations in turf algae mixtures, our photo quadrats lacked the resolution to confirm assemblage diversity. Habitat characteristics likely also influenced the results. Inshore reefs like 1LL had silt-dominated substrates with patchy corals and macroalgae ( Figure 3e ), making sediment clearance challenging without disturbing silt. Offshore reefs ( Figures 3g, h ) had mixed rubble substrates with holes and encrusting organisms, complicating sediment removal. Previous studies recommend sampling flat, smooth surfaces free of sediment-retaining pits or macroalgae (see Purcell, 2000; Tebbett et al., 2017), but standardizing surfaces across habitats was unfeasible. These challenges probably limited sediment removal efficacy, explaining the lack of observable effects on herbivory.

Our findings revealed associations between specific benthic and fish community characteristics and herbivory rates; however, unobserved factors also likely shape these patterns. This is evident from unexpected absences or reductions in herbivory: no detritivore, scraper, or excavator herbivory at 1LL, 3BL (except scrapers), and 5BA; no browser herbivory except at 8KP; and anomalously low rates at 6LU compared to nearby 7KS. The lack of recorded herbivory at the nearshore site 1LL aligns with its degraded reef and poor water quality. However, notably low herbivory at 3BL, 5BA, and 6LU, despite herbivorous fish presence at these sites ( Figure 3a ) and in video footage, is noteworthy. While earlier studies documented herbivory at these sites when macroalgae were offered (Plass-Johnson et al., 2014, 2015), this study focused on natural feeding. Thus, no herbivory observed in certain quadrats or reefs does not necessarily indicate a complete lack of herbivory across the entire reef. Rather, selective feeding behavior and localized factors, including anthropogenic pressures, appear to affect fish distribution and feeding activity.

The 3BL, 5BA, and 6LU islands are among the most densely populated in the Spermonde Archipelago, each hosting 1,900-4,700 people on 8–21 ha islands (BPS, 2022). These communities significantly impact coral reef ecosystems through fishing, boat traffic, and waste discharge. Most coastal communities are fishers, targeting marine biota indiscriminately, including herbivorous fishes like parrotfish (Tresnati et al., 2019, 2020). Many fishers also engage in illegal, destructive fishing practices using explosives, cyanide, and bottom trawls (Munsi et al., 2017; Suarthawan et al., 2022), due to weak enforcement, mild sanctions, and black-market access for materials and fish products (Radjawali, 2012; Renggong et al., 2022; Zaelany, 2019). While destructive fishing is more common at outer, remote, or uninhabited islands (Suarthawan et al., 2022), it also occurs around inhabited islands (Jompa et al., 2005; Zaelany, 2019; ES, unpublished data), though its frequency, distribution, and severity are unknown. During data collection at 3BL, 5BA, and 6LU, we frequently heard blasts from nearby reefs and encountered small-scale fishing near transects. At 5BA, our site was adjacent to a boat mooring area where fishers often fished while awaiting their boats. Local pressures, including human predation, noise pollution from blast fishing and boats, and untreated waste discharge, likely disrupt herbivorous fish distribution, density, and behavior, causing patchy feeding or avoidance behaviors (Ferrier-Pagès et al., 2021; Januchowski-Hartley et al., 2011; McCormick et al., 2018). Impacts on 3BL and 5BA may be amplified compared to 6LU because of closer proximity to the mainland, where urbanization effects are more pronounced and may deter effective feeding. Further research is needed to disentangle these factors and inform strategies for maintaining key ecological functions in reef ecosystems.

Distinct coral reef conditions and herbivory patterns across a spatial gradient from the urban center and functional groups underscore the need for targeted management. From an area-based management perspective, nearshore reefs (1LL to 5BA) require intensive interventions to mitigate urban impacts and support herbivore communities. This is crucial as many herbivorous fishes are sensitive to sediment loads and turf algae composition. Effective strategies should focus on improving water quality and protecting vital habitats to maintain herbivorous fish diversity and functionality. Integrating urban runoff controls, sedimentation management, and broader coastal zone measures can address both land-based and marine stressors, sustaining reef resilience under growing urban pressures.

For mid-to-offshore reefs, improving the newly established Liukang Tupabiring Pangkep marine protected area (MPA, 63,407 ha, established in 2022, Figure 1 ) and revitalizing over a dozen community-based conservation areas (Glaser et al., 2010, 2021; Jompa et al., 2023) is essential for habitat protection. Implementing Kapoposang MPA (50,000 ha, established in 1996) for offshore reefs appeared effective in maintaining diverse herbivorous fish populations, although better reef conditions may also reflect better water quality, which can strongly influence benthic composition (Fabricius et al., 2005; De’ath and Fabricius, 2010) and herbivorous fish biomass (Russ et al., 2015). This synergy is evident at 8KP, within Kapoposang MPA, where higher coral cover and rugosity coincide with lower turf algal and macroalgal covers, diverse fish species, and high herbivory rates. In urban Spermonde reefs, MPAs can play a crucial role by providing opportunities to integrate active restoration and sediment management strategies and buffering additional human impacts. MPAs also protect fish species and habitats resilient to sediment-rich and turbid conditions while supporting fish larval supply to degraded reefs. Studies indicate that outer zone reefs, including Kapoposang Island, benefit from a more substantial larval supply from other regions due to the Indonesian Throughflow (Reuter et al., 2021; Treml et al., 2012), enhancing connectivity with other reef systems in the Archipelago.

Protecting diverse herbivore functional groups is essential for reef resilience to withstand ongoing disturbances (Bellwood et al., 2004, 2006; Donovan et al., 2023), especially since coastal urbanization stressors affect fish herbivory differently across functional groups (e.g., lack of herbivory by detritivores, scrapers, and excavators in some inshore-to-mid reefs, and consistently low herbivory by croppers across all sites). Moreover, many large-bodied herbivorous fishes, such as Scarus and Chlolurus species, are heavily fished and over-exploited in the Archipelago (Fatihah et al., 2021; Tresnati et al., 2019, 2020). Additionally, illegal and destructive fishing methods, including bombs and cyanide, further devastate live coral and degrade reef structures, disproportionately reducing herbivorous fish biomass and herbivory rates (Edwards et al., 2014; McManus et al., 2000). Such impacts compromise reef resilience and increase the risk of ecosystem shifts towards algae-dominated states, hindering coral recovery and compromising overall reef health.

Maintaining browser, scraper, and excavator populations is particularly critical in nearshore reefs (2BA to 5BA) to prevent or reverse phase shifts to algal dominance. While all herbivorous species are functionally important for reef health, conservation efforts focusing on these specific functional groups may be most effective in preserving the ecological balance of nearshore reefs. It is important that the quota-based fisheries regulation (Government Regulation No. 11/2023), recently introduced by the national government, is adopted and effectively implemented by the provincial government. This regulation, which includes fishing zones, quota systems, and vessel and landing port management, is crucial for promoting sustainable fisheries in the Archipelago. However, it is essential to ensure the inclusion of fisheries management for herbivores, particularly those targeted by fishers, in this regulation. Besides this, strengthening enforcement, applying heavier sanctions against illegal and destructive fishing practices, and facilitating social approaches to encourage environmentally friendly fishing gear are paramount for sustaining the overall fish population, including key herbivores.