Broad-scale spatial distribution, microhabitat association and habitat partitioning of damselfishes (family Pomacentridae) on an Okinawan coral reef

Broad-scale fish spatial distribution

Underwater visual surveys were conducted at Sekisei Lagoon and Nagura Bay on the Yaeyama Islands, Okinawa, Japan from July to December 2019 (Figs. 1A, 1B). A total of 67 study sites (31 sites on exposed reefs and 36 sites on inner reefs) were established with an inter-site distance of ~2 km (Fig. 1C).

A 10-min time transect with a 5-m width was employed using SCUBA at each site, and all individuals of damselfish species on the time transect were recorded using a data collection board. Prior to observations, a 5-m reference tape measure was laid on the sea floor. Then, an observer (A.N.) checked the visual estimates of 5 m width (2.5 m width in each side). A portable GPS receiver was used to measure the length of each time transect. The average distance covered was 155.1 m ± 25.3 m standard deviation (minimum length = 101 m; maximum length = 234 m). The fish density of each damselfish species per each site (number of individuals per 500 m²) was determined from the number of individual fish and the length of the 10-min transect. Depth profiles were recorded using a dive computer at a recording interval of 60 s. Ten depth values were obtained for each site (one depth point per minute × for 10 min). The obtained values were averaged for each site and used for subsequent analysis. The water depths ranged from 3.4 m to 11.0 m (7.65 m ± 1.89 m standard deviation).

During the observation period, 47 species and one genus (Stegastes) were identified at the study site, and 26 species were observed at high densities (Fig. S1). Given the similar morphological traits (body size and coloration) of Stegastes spp. as well as that the observations should be conducted while swimming, species-level identification for Stegastes spp. was not be conducted. Thus, Stegastes spp. was excluded from the analysis. Consequently, the above-mentioned 26 species were selected for further analysis because the 26 species comprised of 98.01% of the damselfish population at the study site (Fig. S1), and the use of dominant species would show more robust results about the spatial distribution in relation to environmental characteristics. Broad-scale distributions of each species were displayed by bubble plots on the study map, in which the bubble size represents the fish density per 500 m².

Broad-scale substrate spatial variation

The substrate was recorded using a video camera (GoPro HERO5 Black), which was attached to the data collection board, from a top-down perspective along the transect at each site. Static images were extracted at 10-s intervals using QuickTime Player software (version 7.6) in the laboratory, yielding 61 static images for each site. The static images were used to calculate percentage coverage of each substrate. For each image, the substrate at the center of the static image was recorded. For example, if the number of points at a focal site was as “substrate A = 20, substrate B = 20, substrate C = 10, substrate D = 11”, the estimated coverage of each substrate at the focal site was calculated as “substrate A = 20/61 × 100 = 32.8%, substrate B = 20/61 × 100 = 32.8%, substrate C = 10/61 × 100 = 16.4%, substrate D = 21/61 × 100 = 18.0%”. The substrate was divided into 31 categories (Table S1): (1) staghorn Acropora, (2) branching Acropora, (3) bottlebrush Acropora, (4) corymbose Acropora, (5) tabular Acropora, (6) Pocillopora, (7) branching Isopora, (8) branching Millepora, (9) branching Porites, (10) other branching corals (other than genera Acropora, Isopora, Millepora and Porites), (11) foliose corals, (12) massive corals, (13) other live corals (encrusting corals and mushroom corals), (14) dead staghorn Acropora, (15) dead branching Acropora, (16) dead bottlebrush Acropora, (17) dead corymbose Acropora, (18) dead tabular Acropora, (19) dead Pocillopora, (20) dead branching Isopora, (21) dead branching Millepora, (22) dead branching Porites, (23) dead other branching corals, (24) dead foliose corals, (25) dead massive corals, (26) dead other corals, (27) soft corals, (28) rock (calcium carbonate substratum with lower substrate complexity than live corals), (29) coral rubble, (30) sand, and (31) macroalgae.

The data for substrate availability obtained from the 67 study sites were used for the further analysis (see “Data preparation for CCA” section).

Microhabitat-scale substrate association of fishes

Additional underwater observations were conducted at 19 sites from November 2021 to January 2024 using SCUBA to clarify the microhabitat-scale substrate associations of the 26 damselfish species (Fig. 1D). Four 20 m × 2 m transects were established at each site. Then, the substrate on which fish individuals were initially associated was recorded. In minimizing the impact on fish behaviors, an observer (A.N.) recorded fish data while pulling a 20-m tape measure. For each transect, the 20-m tape measure was set at the center of the transect. Substrates beneath the tape measure were recorded by using a video camera (GoPro HERO5 Black). Then, in the laboratory, substrate images were extracted at 10-cm intervals and divided into the above-mentioned 31 categories for analysis.

Analyses for broad-scale spatial distribution

For each fish species, a generalized linear model (GLM) was applied to examine the significant difference in fish density between exposed and inner reefs using R statistical computing language (R Core Team, 2022). The objective and explanatory variables were fish density and reef type (i.e., exposed reefs or inner reefs), respectively.

To consider the data distribution for broad-scale spatial distribution of each fish species, average and variance of the number of fish individuals at the 67 sites were calculated. This revealed that variance was greater than average (9.27-fold - 281.35-fold). Since Poisson distribution assumed that average and variance is almost equal, negative binomial distribution was applied for further data analysis.

The GLM was performed by “glm.nb” function of “MASS” package. The data were assumed to follow a negative binomial distribution with a log-link function. Considering that the fish count data at each site were obtained from a 10-min survey, the length of each time transect varied among the 67 sites. Thus, fish data were analyzed with the “offset” option in the R package using the length of each time transect.

After performing the GLM, the degree of zero-inflation was examined by using “check zeroinflation” function of “performance” package. This procedure revealed that two out of 26 species (Neopomacentrus anabatoides and Chromis ternatensis) showed a significant zero-inflation of data distribution under the assumption of negative binomial distribution. Thus, additional GLM was performed by “zeroinfl” function of “pscl” package for these two species with the assumption of zero-inflated negative binomial distribution.

The relationship between the broad-scale spatial distribution of the 26 damselfish species and the 32 environmental characteristics (31 substrate categories plus depth) was analyzed as follows: (1) detrended correspondence analysis (DCA) was performed to examine the species response (linear or unimodal) to the environmental characteristics using CANOCO software (Ter Braak & Smilauer, 2002); (2) since the DCA revealed the unimodal responses of species against environmental characteristics, canonical correspondence analysis (CCA) was performed to clarify the relationship. In addition, to identify the environmental characteristics that strongly affect the spatial distributions of the 26 damselfish species, forward selection was applied using CANOCO software.

Data preparation for CCA

Prior to the CCA, principal component analysis (PCA) was performed to reduce the number of independent variables, thereby avoiding multi-collinearity among the above-mentioned 32 environmental characteristics. The PCA was performed using PRIMER software (version 6). The PCA provided the principal component scores for 67 study sites along with five PC axes. Among the five PC axes, three axes (PC 1, PC 2 and PC 3) showed greater contributions to explain the overall trends in the site-specific difference in environmental characteristics (PC 1 = 64.1%; PC 2 = 18.4%; PC 3 = 7.2%). Thus, these principal scores were used as environmental variables for the CCA. For fish data, fish density data were log (x + 1)-transformed.

Analyses for microhabitat-scale substrate association

To clarify the overall trends in species-specific differences in substrate associations, cluster analysis using the group average linkage method with the Bray-Curtis similarity index was applied. Cluster analysis was performed using PRIMER software (version 6).

“Resource selection ratio” was applied to examine the substrate association (Manly et al., 2002) as:

$$w_i=o_i/{\pi }_i$$where w_i is the resource selection probability function, o_i is the proportion of the ith substrate that was used by a focal fish species, and π_i is the proportion of the ith substrate that was available in the study area (Manly et al., 2002). For multiple comparisons, Bonferroni Z corrections (Quinn & Keough, 2002) was used to calculate the 95% confidence interval (CI) for each w_i as:

$$\text{95}\%\text{\hspace{0.17em}CI}=Z_{a/2I}\sqrt{}[o_i(1-o_i)/(U_{+}{\pi }_i^2)]$$where Z_a/2I is the critical value of the standard normal distribution corresponding to the upper tail area of a/2I, a is 0.05, I is the number of substrate categories, and U₊ is the total number of individuals of the focal fish species. Substrates with w_i ± 95% CI above and below one indicate a significantly positive and negative (non-positive, not avoidance) association, respectively. Substrates with w_i ± 95% CI encompassing one showed no significant positive or negative association.

Degree of dependence on live or acroporid corals

In accordance with Pratchett et al. (2016), obligate coral dwellers are defined as species in which over 80% of the individuals are associated with live corals. In addition, the author of this study (A.N.) proposed two additional definitions for the degree of dependence on live corals as follows: (1) greater degree of dependence (over 60% of the individuals are associated with live corals) and (2) some extent of dependence (over 40% of the individuals are associated with live corals).

Dependence on acroporid corals is also regarded as an indicator of the susceptibility of damselfish species to the destruction of coral assemblages by mass coral bleaching events and outbreaks of crown-of-thorns starfish (Pratchett et al., 2008). The degree of dependence on live or acroporid corals was calculated as follows:

$$\begin{array}{rl}& \mathrm{D}\mathrm{e}\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{o}\mathrm{f}\mathrm{d}\mathrm{e}\mathrm{p}\mathrm{e}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}\mathrm{o}\mathrm{n}\mathrm{l}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{o}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{r}\mathrm{o}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{i}\mathrm{d}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{a}\mathrm{l}\mathrm{s}(\mathrm{\%})\\ & =[(\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{f}\mathrm{i}\mathrm{s}\mathrm{h}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{d}\mathrm{u}\mathrm{a}\mathrm{l}\mathrm{s}\mathrm{t}\mathrm{h}\mathrm{a}\mathrm{t}\mathrm{w}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{s}\mathrm{s}\mathrm{o}\mathrm{c}\mathrm{i}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{w}\mathrm{i}\mathrm{t}\mathrm{h}\mathrm{l}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{o}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{r}\mathrm{o}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{i}\mathrm{d}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{a}\mathrm{l}\mathrm{s})\\ & /(\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{o}\mathrm{b}\mathrm{s}\mathrm{e}\mathrm{r}\mathrm{v}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{d}\mathrm{u}\mathrm{a}\mathrm{l}\mathrm{s})]\times 100\end{array}$$

This index was calculated for each damselfish species using microhabitat-scale substrate association data.

Analyses of habitat partitioning using niche overlap index

To examine both broad-scale and microhabitat-scale habitat partitioning, Pianka’s index (Pianka, 1973) was applied as follows:

$$O_{jk}=O_{kj}={\displaystyle \sum }\text{}{p}_{ik}•p_{ik}/[\sqrt{(}{\displaystyle \sum }\text{}{p}_{ij}^2•{\displaystyle \sum }\text{}{p}_{ij}^2)]$$where O_jk and O_kj are the niche overlap indices between jth and kth species, respectively; pij and pik represent the proportions of the ith resource used by the jth and kth species, respectively. The values of this index range from 0 to 1 with greater values representing a greater niche overlap (habitat overlap) and vice versa. This index was calculated for both broad-scale spatial distribution and microhabitat-scale substrate associations. Then, the relationship between the broad-scale and microhabitat-scale niche overlap indices was plotted as a two-dimensional graph for a focal species to the other species (x-axis = broad-scale index, y-axis = microhabitat-scale index).

The setting value of the threshold whether habitat partitioning was found or not was 0.5 based on Harmácková, Remesová & Remes (2019), indicating that O_jk < 0.5 and O_jk > 0.5 represent greater degree of habitat partitioning and habitat overlap, respectively (Fig. S2).

Overall trends in broad-scale spatial distribution

The map and GLM revealed the overall trends in the species-specific spatial distribution of the 26 fish species at the 67 study sites (Figs. 2–4, Table 1).

Three species (Abudefduf vaigiensis, Chromis atripes and Chromis vanderbilti) were found only in the exposed reefs (Figs. 2A–2C, Table 1). Seven species (C. margaritifer, C. ovatiformis, Chrysiptera rex, Pomacentrus lepidogenys, P. philippinus, P. vaiuli and Pomachromis richardsoni) showed significantly greater densities in the exposed reefs (Figs. 2D–2J, Table 1; p < 0.05 for all seven species).

Eight species (Amblyglyphidodon leucogaster, Chromis chrysura, C. viridis, Chrysiptera cyanea, Neopomacentrus anabatoides, P. coelestis, P. moluccensis and P. sp. 1) showed no significant difference in density between the exposed reefs and inner reefs (Fig. 3, Table 1; p > 0.05).

By contrast, six species (Amblyglyphidodon curacao, Chromis ternatensis, Dascyllus aruanus, D. reticulatus, Pomacentrus alexanderae, P. amboinensis,) showed significantly greater densities in the inner reefs (Figs. 4A–4F, Table 1; p < 0.05 for all six species). Two species (Chrysiptera parasema and Pomacentrus sp. 2) were found only in the inner reefs (Figs. 4G, 4H, Table 1).

Broad-scale spatial distribution in relation to environmental characteristics

The results of PCA revealed the relationship between the 32 environmental characteristics and the three PC axes (Fig. S3). PC 1 represents greater coverage of coral rubble and sand on the positive axis, and greater coverage of rock on the negative axis. PC 2 represents greater coverage of sand on the positive axis, and greater coverage of live corals (e.g., branching Acropora, bottlebrush Acropora, branching Millepora and other corals), dead corals (dead branching Acropora and dead other corals) and coral rubble on the negative axis. PC 3 represents greater coverage of live corals (e.g., branching Acropora, bottlebrush Acropora, branching Millepora, massive corals and other corals), dead corals (dead branching Acropora and dead bottlebrush Acropora) and macroalgae on the positive axis, and greater coverage of coral rubble on the negative axis.

The CCA revealed that the fish assemblages were primarily divided into three groups in relation to the environmental characteristics (Fig. 5). The first group consisted of 11 species (Abudefduf vaigiensis, Chromis atripes, C. chrysura, C. vanderbilti, C. margaritifer, C. ovatiformis, Chrysiptera rex, Pomacentrus lepidogenys, P. philippinus, P. vaiuli and Pomachromis richardsoni) that were found at the first and fourth quadrants of the CCA plot (quadrants with the minus direction of PC axis 1), indicating that these species primarily occurred in the exposed reefs with greater coverage of rock (Figs. S3A, 5). The second group consisted of nine species (Amblyglyphidodon curacao, A. leucogaster, Chromis ternatensis, Chrysiptera parasema, Neopomacentrus anabatoides, Pomacentrus alexanderae, P. moluccensis, P. sp. 1 and P. sp. 2) that were found in the second quadrant of the CCA plot (quadrant with the plus directions of PC axes 1 and 3), indicating that these species primarily occurred in the inner reefs with greater coverage of live corals (branching Acropora, bottlebrush Acropora, staghorn Acropora and branching Millepora), dead corals (dead branching Acropora and dead bottlebrush Acropora) and macroalgae (Figs. S3A, S3C, 5). The third group consisted of six species (Chromis viridis, Chrysiptera cyanea, Dascyllus aruanus, D. reticulatus, Pomacentrus amboinensis and P. coelestis) that were found at the third quadrant of the CCA plot (quadrant with the plus directions of PC axes 1 and 2), indicating that these species primarily occurred in the inner reefs with greater coverage of sand (Figs. S3A, S3B, 5).

Microhabitat-scale substrate association

Species-specific variations in microhabitat-scale substrate associations were observed (Figs. 6–9, S4–S7). The cluster analysis revealed that the 24 species could be divided into six groups, and the remaining two species (Pomacentrus amboinensis and Neopomacentrus anabatoides) had unique patterns in terms of substrate association (Fig. 10).

Group A is comprised of 12 species (Abudefduf vaigiensis, Chromis atripes, C. vanderbilti, C. chrysura, C. margaritifer, C. ovatiformis, Chrysiptera rex, Pomacentrus coelestis, P. philippinus, P. vaiuli, P. sp. 1 and Pomachromis richardsoni) (Figs. 6, 7, 10) that showed a significant positive association with rock (p < 0.0016, Table 2 except for Pomacentrus sp. 1).

Group B is comprised of two species (Chrysiptera cyanea and Pomacentrus sp. 2) (Figs. 8A, 8B, 10) that showed a significant positive association with coral rubble and negative association with rock (p < 0.0016, Table 3).

Group C is comprised of two species (Dascyllus aruanus and D. reticulatus) (Figs. 8C, 8D, 10) that showed significant positive associations with corymbose Acropora and Pocillopora (p < 0.0016, Table 3).

Group D is comprised of three species (Chromis ternatensis, Pomacentrus lepidogenys and Chrysiptera parasema) (Figs. 8E–8G, 10) that showed a significant positive association with branching Acropora (p < 0.0016, Table 3).

Group E is comprised of three species (Chromis viridis, Pomacentrus amboinensis and P. alexanderae) (Figs. 9A–9C, 10) that showed significant positive associations with branching Acropora, branching Isopora and branching Porites (p < 0.0016, Table 4).

Group F is comprised of two species (Amblyglyphidodon curacao and Am. leucogaster) (Figs. 9D, 9E, 10) that showed positive associations with staghorn Acropora and branching Millepora, as well as a negative association with rock (p < 0.0016, Table 4).

For the remaining two species, Pomacentrus amboinensis showed a significant positive association with branching Millepora (p < 0.0016, Fig. 9F, Table 4), whereas Neopomacentrus anabatoides showed significant positive associations with branching Porites, foliose corals and dead branching Porites (p < 0.0016, Fig. 9G, Table 4).

Dependence on live or acroporid corals

Among the 26 species, nine species (Dascyllus aruanus, D. reticulatus, Chromis ternatensis, Chrysiptera parasema, C. viridis, Pomacentrus moluccensis, P. alexanderae, Amblyglyphidodon curacao, and Am. leucogaster) were categorized as obligate coral dwellers (>80% of the individuals are associated with live corals; Figs. 8C–8E, 8G, 9A–9E).

Three species (Chromis ternatensis, Chrysiptera parasema and Amblyglyphidodon leucogaster) showed a greater degree of dependence on acroporid corals (>60% individuals were associated with acroporid coral) (Figs. 8E, 8G, 9E). Another three species (Dascyllus aruanus, D. reticulatus and Pomacentrus lepidogenys) also showed some extent of dependence on acroporid corals (>40% individuals were associated with acroporid coral) (Figs. 8C, 8D, 8F).

By contrast, 10 species (Abudefduf vaigiensis, Chromis atripes, C. chrysura, C. ovatiformis, C. vanderbilti, Chrysiptera rex, Pomacentrus coelestis, P. philippinus, P. sp. 2 and Pomachromis richardsoni) showed a lower degree of dependence on live corals (less than 16% individuals were associated with live corals) and acroporid corals (less than 10% individuals were associated with acroporid coral).

Habitat partitioning among multiple species in two spatial scales

Fourteen species (Amblyglyphidodon curacao, Am. leucogaster, Chromis chrysura, C. viridis, Chrysiptera cyanea, Ch. parasema, Dascyllus aruanus, D. reticulatus, Neopomacentrus anabatoides, Pomacentrus amboinensis, P. coelestis, P. vaiuli, P. sp. 1 and P. sp. 2) showed relatively greater habitat partitioning (O_jk < 0.5) with other species at the broad-scale and/or microhabitat-scale (Fig. 11, Table S2).

For the remaining 12 species, greater habitat overlaps were found within four fish groups, each consisting of three fish species (Fig. 12, Table S3): (1) Abudefduf vaigiensis, Chromis atripes and C. ovatiformis (Figs. 12A–12C, Table S3); (2) Chromis margaritifer, C. vanderbilti and Chrysiptera rex (Figs. 12D–12F, Table S3); (3) Chromis ternatensis, Pomacentrus alexanderae and P. moluccensis (Figs. 12G–12I, Table S3); and (4) Pomacentrus lepidogenys, P. philippinus and Pomachromis richardsoni (Figs. 12J–12L, Table S3). Chromis ovatiformis also showed habitat overlaps with Pomacentrus philippinus (Figs. 12C, 12K, Table S3).

Species distribution based on two spatial-scale approaches

Broad-scale analysis revealed that 11 species were primarily found in the exposed reefs with a greater coverage of rock. Among these, microhabitat-scale analysis revealed that 10 species (Abudefduf vaigiensis, Chromis atripes, C. chrysura, C. vanderbilti, C. margaritifer, C. ovatiformis, Chrysiptera rex, Pomacentrus philippinus, P. vaiuli and Pomachromis richardsoni) showed positive associations with rock. Since Froese & Pauly (2024) has shown that the fish length of these 10 species is less than 20 cm, small holes and fine-scale uneven surfaces of rock in the exposed reef can provide refuge space. Ticzon et al. (2012) and Nanami (2021) have suggested that a rocky surface inherently provides uneven surfaces and large holes, and it creates complex physical structures. The substrate complexity provided by rock increases the density of groupers and parrotfishes (Ticzon et al., 2012, Nanami, 2021), although the degree of complexity is lower than that of live corals. Thus, it is suggested that some damselfish species are associated with complex physical structures provided by non-coralline substrates as habitats and refuge spaces. One exception was Pomacentrus lepidogenys, which was positively associated with live corals (branching Acropora and branching Millepora) and dead corals (dead staghorn Acropora, dead branching Acropora and dead branching Millepora). This suggests that P. lepidogenys selectively utilized live and dead corals as habitat, although these substrates were not abundant in the exposed reef.

Broad-scale analysis also revealed that nine species were primarily found in the inner reef with greater coverage of live corals, dead corals and macroalgae. Microhabitat-scale analysis revealed that seven species (Amblyglyphidodon curacao, Am. leucogaster, Chromis ternatensis, Chrysiptera parasema, Neopomacentrus anabatoides, Pomacentrus alexanderae and P. moluccensis) showed positive associations with live corals (including both acroporid and non-acroporid corals) that have complex physical structures. Corals with complex structures form suitable refuge spaces for damselfishes (reviewed in Pratchett et al. (2016)), because such structures reduce fish mortality caused by predation (Almany, 2004). Two species (N. anabatoides and P. alexanderae) also showed positive associations with dead corals, suggesting that the complex structures provided by dead corals are utilized as habitats or refuge spaces to a certain extent. The exceptions were two species (Pomacentrus sp. 1 and sp. 2). Pomacentrus sp. 1 showed no significant association with any substrate in the microhabitat-scale approach, although this species was associated with a greater coverage of live corals, dead corals and macroalgae in broad-scale approach. Pomacentrus sp. 2 showed significant positive associations with dead staghorn Acropora and coral rubble. Although broad-scale analysis revealed that these nine species were primarily found in the inner reefs with a greater coverage of macroalgae, microhabitat-scale analysis showed that no individuals were associated with macroalgae. This suggests that these nine species primarily occur in the inner reefs with live and dead corals, where macroalgae also occur.

In addition, the broad-scale analysis revealed six species were associated with sites that have a greater coverage of sand, yet microhabitat-scale analysis revealed contrasting results in an association with coral and rock. Five species (Chromis viridis, Chrysiptera cyanea, Dascyllus aruanus, D. reticulatus and Pomacentrus amboinensis) and one species (P. coelestis) showed significant positive associations with live corals (including acroporid and non-acroporid corals) and rock, respectively. This finding suggests that these six species are associated with substrates that have complex physical structures, which are surrounded by sandy bottom. Considering that sandy bottom areas are primarily found in the inner reef, it is suggested that the six species prefer complex physical structure in the inner reefs.

These results suggest the importance of multi-scale approaches in examining the spatial distribution patterns of damselfishes. In particular, six species (Chromis viridis, Chrysiptera cyanea, Dascyllus aruanus, D. reticulatus, Pomacentrus amboinensis and P. coelestis) showed differences between broad-scale and microhabitat-scale analyses. Since damselfishes are closely associated with substrates with fine structures as habitats and as refuge spaces, broad-scale analysis might not detect precise aspects of substrate associations. By contrast, microhabitat-scale analysis may not detect landscape-level spatial distributions (exposed reefs or inner reefs). Thus, in addition to considering a broad-scale approach (e.g., exposed reefs vs. inner reefs), a microhabitat-scale approach (precise categorization of substrates) should be considered to examine the spatial distribution patterns in small-sized fishes with greater dependence on substrates.

Dependence on live corals

Nine of the 26 species (Dascyllus aruanus, D. reticulatus, Chromis ternatensis, C. viridis, Chrysiptera parasema, Pomacentrus moluccensis, P. alexanderae, Amblyglyphidodon curacao, Am. leucogaster) were categorized as obligate coral dwellers in the present study (>80% of the individuals are associated with live corals). The results were similar to the results of Pratchett et al. (2016), with exception of three species (Amblyglyphidodon curacao and Am. Leucogaster and Neopomacentrus anabatoides). In this study, two species (Amblyglyphidodon curacao and Am. leucogaster) and one species (Neopomacentrus anabatoides) were categorized as obligate and facultative coral dwellers, respectively. By contrast, Pratchett et al. (2016) categorized these species as facultative and obligate coral dwellers, respectively. This difference might be due to geographical variations in damselfish behavior and the species composition of live coral assemblages.

The degree of dependence on live corals provides some insights about the effects of coral assemblage degradation on damselfish assemblage structures. In particular, the dependence on acroporid corals might be an effective indicator for estimating the effects of mass coral bleaching events and outbreaks of crown-of-thorns starfish on the decline of damselfish populations (Pratchett et al., 2008, 2009; Coker, Wilson & Pratchett, 2014). The degree of dependence on acroporid corals for three species (Chromis ternatensis, Chrysiptera parasema and Amblyglyphidodon leucogaster) was over 60%, and the degree for the other three species (Dascyllus aruanus, D. reticulatus and Pomacentrus lepidogenys) was over 40%. These results suggest that the populations of these six species might be negatively impacted to some extent after the loss of acroporid corals. By contrast, since the degree of dependence on acroporid corals for the other 20 species was less than 40%, the negative impact on the population of the 20 species might be lower. For the remaining 10 species, these species might be more resilient to the loss of acroporid corals, since the degree of dependence on acroporid corals was less than 10% for the 10 species.

Some previous studies also showed the associations of fish with substrates that have less complex physical structures (Wilson et al., 2008). For example, some damselfish species (Chrysiptera rollandi, Dischistodus melanotus and Neoglyphidodon nigroris) showed significant positive associations with coral rubble in the Great Barrier Reef (Wilson et al., 2008). The results of the present study also showed that two species (Chrysiptera cyanea and Pomacentrus sp. 2) showed a significant positive association with coral rubble. This suggests that some damselfish species utilize the fine-scale space provided by coral rubble, and these species might have some degree of resilience to the degradation of live corals.

Habitat partitioning

Among the 26 species on the Okinawan coral reef, 14 species showed a greater degree of habitat partitioning at broad and/or microhabitat-scales. Several species showed differences in their spatial distribution between the exposed and inner reefs. For example, Chrysiptera cyanea and Pomacentrus sp. 2 showed similar patterns in microhabitat association but different patterns in the broad-scale spatial distribution. By contrast, several species (e.g., Amblyglyphidodon curacao and Pomacentrus sp. 2) showed different patterns in microhabitat associations among the species but no clear differences in broad-scale spatial distribution. These results suggest that both broad-scale environmental variation and microhabitat-scale substrate diversity provide diverse habitats at the present study site, supporting species coexistence via habitat partitioning.

By contrast, the remaining 12 species showed a certain degree of habitat overlap at the broad and/or microhabitat-scales. For example, Abudefduf vaigiensis and Chromis atripes showed greater habitat overlap at both scales (Pianka’s indices were over 0.97 in both spatial scales). The size variations in crevices and holes on the rocky surface might be the main factor supporting the coexistence, because the fish length of Abudefduf vaigiensis has been found to be greater than that of Chromis atripes (Froese & Pauly, 2024). More details about the precise aspects of the differences in architectural structures within rocky surfaces should be examined to explain such patterns of species coexistence. Another reason for the species coexistence is the prey item differences between the two species, as the former and latter species are a benthic organism feeder and a plankton feeder, respectively (Pratchett et al., 2016).

Chromis atripes also showed greater habitat overlap with C. ovatiformis, and both species are plankton feeders (Pratchett et al., 2016). However, Leray et al. (2018) showed prey item partitioning between two species of plankton feeders (Dascyllus flavicaudus and Chromis viridis) in the lagoon of Moorea, when the species composition of plankton assemblages in the digestive tract was precisely identified. Thus, precise identification of prey item plankton should be conducted, and prey item difference might be also found between the two species. All 12 species showed some niche overlaps, indicating similar broad-scale spatial distribution, microhabitat-scale substrate association or similar prey item categorization. However, as discussed above, more precise substrate identification and/or prey item identification might clarify niche partitioning among the 12 species in a more precise manner.

Other ecological processes (e.g., inter-specific competition, presence of con-specific individuals and stochastic processes of larval settlement) also promote species coexistence among damselfishes (Sweatman, 1983, 1985; Munday, Jones & Caley, 2001; Munday, 2004; Eurich, McCormick & Jones, 2018a, 2018b; Ebersole, 1985). These ecological processes might be among the factors that maintain the coexistence among the several species with greater habitat overlaps in this study. Comprehensive examinations including various ecological factors should be considered to explain the precise mechanisms that are responsible for the coexistence among damselfish species.

Marine protected areas to conserve damselfish species diversity

Various ecological aspects (e.g., diverse habitat) should be considered to establish effective marine protected areas (Kelleher, 1999; Green, White & Kilarski, 2013). This study revealed the species-specific broad-scale spatial distribution and microhabitat-scale habitat association of damselfishes in an Okinawan coral reef. Based on the results, several ecological factors should be incorporated to achieve effective MPAs to conserve damselfish species diversity as follows: (1) both exposed reefs and inner reefs should be protected; (2) diverse substrate types including various live coral species as well as non-coralline substrates should be protected; and (3) species-specific responses to habitat degradation should be more precisely clarified.