Fig. 4 in Fig. 5 in Cymonomus curvirostris Sakai 1965

Comparative Feeding Ecology of Cardinalfishes (Apogonidae) at Toliara Reef, Madagascar Bruno Frédérich1,2,*, Loïc N. Michel2,*, Esther Zaeytydt1, Roger Lingofo Bolaya1, Thierry Lavitra3, Eric Parmentier1, and Gilles Lepoint2 1Laboratoire de Morphologie Fonctionnelle et Evolutive, AFFISH Research Center, Université de Liège, Belgique. E-mail: [email protected] (Frédérich); [email protected] (Zaeytydt); [email protected] (Bolaya); [email protected] (Parmentier) 2Laboratoire d’Océanologie, MARE Center, Université de Liège, Belgique. E-mail: [email protected] (Michel); [email protected] (Lepoint) 3Institut Halieutique et des Sciences Marines (IH.SM), Université de Tuléar, Madagascar. E-mail: [email protected] (Received 30 January 2017; Accepted 20 April, 2017; Published 12 May 2017; Communicated by Hin-Kiu Mok) Bruno Frédérich, Loïc N. Michel, Esther Zaeytydt, Roger Lingofo Bolaya, Thierry Lavitra, Eric Parmentier, and Gilles Lepoint (2017) Despite their importance in coral reef ecosystem function and trophodynamics, the trophic ecology of nocturnal fishes (e.g. Apogonidae, Holocentridae, Pempheridae) is by far less studied than diurnal ones. The Apogonidae (cardinalfishes) include mostly carnivorous species and evidence of trophic niche partitioning among sympatric cardinalfishes is still limited. The present study combines stomach contents and stable isotope analyses to investigate the feeding ecology of an assemblage of eight cardinalfishes from the Great Reef of Toliara (SW Madagascar). δ13C and δ15N of fishes ranged between -17.49‰ and -10.03‰ and between 6.28‰ and 10.74‰, respectively. Both stomach contents and stable isotopes showed that they feed on planktonic and benthic animal prey in various proportions. Previous studies were able to group apogonids in different trophic categories but such a discrimination is not obvious here. Large intra-specific variation in the stomach contents and temporal variation in the relative contribution of prey to diet support that all apogonids should be considered as generalist, carnivorous fishes. However the exploration of the isotopic space revealed a clear segregation of isotopic niches among species, suggesting a high level of resource partitioning within the assemblage. According to low inter-specific variation in stomach content compositions, we argue that the differences in isotopic niches could be driven by variation in foraging locations (i.e. microhabitat segregation) and physiology among species. Our temporal datasets demonstrate that the trophic niche partitioning among cardinalfishes and the breadth of their isotopic niches are dynamic and change across time. Factors driving this temporal variation need to be investigated in further studies. Key words: Apogonids, Stable isotopes, Isotopic niche, Diet, Western Indian Ocean. * Correspondence: Bruno Frédérich and Loïc N. Michel contributed equally to this work. Tel: +3243665133. Fax: +3243663715. E-mail: [email protected] BACKGROUND Trophic niche partitioning is a major axis of ecological diversification in reef fishes (Wainwright and Bellwood 2002). Trophic ecology of reef fishes has been broadly studied but, to date, most studies have focused on diurnal taxa (e.g. Pratchett 2005; Frédérich et al. 2009). Abundant nocturnal reef fishes include bullseyes (Pempheridae), soldierand squirrelfishes (Holocentridae), and cardinalfishes (Apogonidae) (Hobson 1965; Hobson 1972). Despite their importance for coral reef ecosystem function and trophodynamics (Harmelin-Vivien 2002), these groups remain less studied than other reef fish families. Diurnal fishes show a high level of trophic diversity including herbivores, corallivores, detritivores, durophagous fishes, zooplankton Zoological Studies 56: 10 (2017) doi:10.6620/ZS.2017.56-10 1 feeders, fish predators, and omnivores (Randall 1967; Wainwright and Bellwood 2002). This variety contrasts the limited trophic diversity of nocturnal fishes, which are mainly carnivorous. Common prey items of nocturnal fishes are restricted to fish, zooplankton and mobile benthic invertebrates (Gladfelter and Johnson 1983; Marnane and Bellwood 2002; Wainwright and Bellwood 2002). Despite this apparent similarity of diet preferences, differences in the feeding ecology of nocturnal fishes can be highlighted. Food might be partitioned by taxon and prey size. For example, some holocentrids consume predominantly shrimps when others mainly eat crabs (Gladfelter and Johnson 1983). Variation in the timing of foraging and spatial niche partitioning has also been reported (Marnane and Bellwood 2002; Annese and Kingsford 2005). Cardinalfishes (Apogonidae) comprise 347 valid species (Eschmeyer et al. 2016), widely distributed in all tropical and warm temperate seas. They usually occur in coral and rocky reefs while some species inhabit seagrass meadows, soft bottoms and estuaries. Apogonids form a major component of reef fish assemblages, both in terms of species diversity and numerical abundance (Wainwright and Bellwood 2002). Most apogonids are carnivorous species feeding on benthic organisms, plankton and small fish (Vivien 1975; Chave 1978; Barnett et al. 2006; Marnane and Bellwood 2002). They can be segregated into two trophic groups: piscivores and generalists that feed on a range of benthic and planktonic crustaceans (Barnett et al. 2006). Stomach content analysis suggested overlap in many apogonids’ diet, and most studies failed to identify clear subgroups of planktonic and benthic feeders based on stomach contents (Barnett et al. 2006). However, Marnane and Bellwood (2002) found that some species foraged high in the water column at night, suggesting a diet relying more on planktonic prey. To date, most studies about trophic diversity of apogonids are based on stomach content analyses (Vivien 1975; Marnane and Bellwood 2002; Barnett et al. 2006). This method allows identification of prey with high resolution. However, it only gives a snapshot of the diet at sampling time (Hyslop 1980), while trophic processes can show high temporal variation. Gut content examination can also lead to over-estimation of poorly palatable and/or digestible items as it focuses on ingested food, but gives no information about whether this food is actually assimilated and exploited by consumers or not. These limitations reinforce the importance of trophic markers, such as the use of stable isotopes of nitrogen (δ15N) and carbon (δ13C). Stable isotope analysis has emerged as a powerful tool for tracing dietary sources, as the isotope ratios of a consumer are mostly driven by those of its food (Peterson and Fry 1987; Layman et al. 2012). This method provides an integrated measure of the dietary components over a much longer period of time than do gut contents. Although stable isotope analysis does not provide a detailed picture of dietary preferences, it gives an average estimate of an organism’s preferred diet that is much less subject to temporal bias (Layman et al. 2012). Recently, stable isotope compositions were also revealed as a powerful tool for assessing the trophic niche width of species and for identifying trophic specialists from generalists (Bearhop et al. 2004; Jackson et al. 2011). In the present study, our main objective was to compare the feeding ecology of eight species of Apogonidae at the Toliara Great Reef (SW Madagascar). Since these apogonids co-occur in the inner reef, we expected some degree of trophic partitioning among species to reduce competition (Schoener 1974). Specifically, we aimed (1) to characterize their diet; (2) to estimate their trophic niche size and potential overlap among trophic niches; (3) to study interannual variation in their feeding ecology. To achieve these goals, we combined stomach contents and stable isotope analyses. This dual approach was motivated by the complementarity between the two techniques and the potential of each one to compensate for the other technique’s caveats. Joint use of these techniques has already proven valuable to delineate feeding strategies among numerous consumers, including coral reef fishes (e.g. Frédérich et al. 2009; Layman and Allgeier 2012; Lepoint et al. 2016). To ensure robust quantitative estimates of isotopic niches, data were explored using advanced Bayesian tools such as the SIBER niche metrics (Stable Isotope Bayesian Ellipses in R; Jackson et al. 2011). MATERIALS AND METHODS Sampling Fishes were collected on the Great Reef of Toliara (SW Madagascar – 23.36°S, 43.66°E) during June 2011, July 2012 and July 2014. Every year, fishes were captured within the same area page 2 of 14Zoological Studies 56: 10 (2017) along the inner reef slope. A total of 182 specimens (Table 1) were sampled by scuba diving at depths ranging from 2 to 10 m after being anesthetized by a solution of clove oil. During the field campaigns conducted in 2011 and 2012, fishes were captured during the morning (i.e. between 7 and 9 a.m.) while fishes from 2014 were collected at dawn. After their capture, the fishes were brought to the surface and killed as quickly as possible by overdose immersion in MS-222. They were then placed on ice. Each fish was weighed and its standard length (SL) was measured to the nearest millimeter. Samples (± 1 cm3) of lateral muscle tissue of each fish were used for stable isotope analysis. The digestive tract was removed and conserved in 70% ethanol for stomach content analysis. Potential fish food items (i.e. zooplankton and benthic invertebrates) were taken from the fish collection site. The protocol for sampling food sources is fully detailed in Frédérich et al. (2009). In brief, mesozooplankton was sampled with a 250 mm mesh size net every sampling year, and small benthic invertebrates (e.g. amphipods, isopods, annelids…) were trapped using small light traps made by plastic bottles containing glow sticks (Michel et al. 2010) in 2011 and 2014 only. Sample sizes of these food sources and their mean isotopic values are provided as supplementary material (Table S1). Stomach content analysis After dissection, stomachs were opened and all dietary constituents were dispersed onto individual glass slides. All food items were identified using a Wild M10 binocular microscope. Animal prey were identified to the phylum, class or family and assigned to the planktonic or the benthic environment. Plant items were classified as either phytoplankton or fragments of benthic algae. Amorphous material (i.e. items lacking any identifiable features) was classified as unrecognized. In order to define the diet of every apogonid, we quantified food items as a percentage of occurrences and as a mean percent composition of each item in the stomach content (Hyslop 1980). Stable isotope analysis Samples of lateral muscle tissue and potential food sources were dehydrated for at least 48 h at 50°C before being ground into a homogenous powder using mortar and pestle. Inorganic carbon present in samples can be a source of bias for C stable isotope ratio analysis. Therefore, after grinding, samples containing carbonates (zoobenthos) were placed for 24 h under a glass bell with fuming HCl (37%; Merck, for analysis quality) to eliminate calcareous material. Measurements were performed using an IsoPrime100 isotope ratio mass spectrometer (Isoprime, UK) coupled to a vario MICRO cube C-N-S elemental analyzer (Elementar Analysensysteme GMBH, Italy) for sample transformation and automated analysis. Isotopic ratios were expressed using the δ notation (‰) (Coplen 2011). Certified Reference Materials (CRM) were IAEA-N1 (ammonium sulphate, δ15N = 0.4 ± 0.2‰; mean ± SD) for nitrogen and IAEA-C6 (sucrose; δ13C = -10.8 ± 0.5‰; mean ± Table 1. List of the studied species. N, number of specimens; SL, standard length. The percentage of specimens in which stomach was empty is provided Year Species NSize range (SL, mm) Empty stomach (%) 2011 Ostorhinchus cookii 9 47.1 – 69.5 66.7 Ostorhinchus cyanosoma 15 37.5 – 46.8 60 Pristiapogon fraenatus 15 52.4 – 68.3 20 Pristiapogon kallopterus 5 46.1 – 57 0 2012 Ostorhinchus cookii 14 48 – 79 28.6 Cheilodipterus quinquelineatus 252 – 69 50 2014 Ostorhinchus aureus 25 54.1 – 91.2 24 Ostorhinchus cyanosoma 18 41.9 – 49.9 22.2 Pristiapogon fraenatus 27 53.7 – 89.6 55.6 Pristiapogon kallopterus 15 62.6 – 93.8 26.7 Taeniamia fucata 25 65.2 – 76.7 24 Zoramia leptacantha 12 34.5 – 39 0 page 3 of 14Zoological Studies 56: 10 (2017) SD) for carbon. Both CRM are calibrated against the international references Vienna Pee Dee Belemnite and atmospheric air for carbon and nitrogen, respectively. Standard deviations on replicate measurements of a randomly selected fish muscle sample (one replicate measurement every 15 analyses) were less than 0.3‰ for both δ13C and δ15N. Data treatment and statistics To determine whether apogonids differed in their diet composition, Bray-Curtis similarity coefficients between individuals were computed using relative stomach content composition (percentage) data. These coefficients were subsequently used to perform hierarchical clustering. The non-parametric ANOSIM test (analysis of similarity) was then used to statistically test differences in stomach contents among species. This test provides an output R-value and a P-value stating about its significance. R-value is supposed to vary between 0 and 1, and R-value > 0.5 suggests divergence between groups (Clarke and Warwick 2001). Additionally, the null hypothesis of no difference in global foraging tactics (expressed as the percentage of planktonic or benthic animal prey found in stomach contents) among species was tested using a one-way ANOVA followed by post-hoc multiple comparison tests (Tukey test). The percentages were arcsinesquare root transformed before the analysis to meet the normality assumption (Shapiro-Wilk’s test, after transformation). When relevant, interspecific and/or interannual differences in δ13C and δ15N were tested using hypothesis-based comparison procedures. D’Agostino & Pearson normality tests revealed that several datasets did not follow a Gaussian distribution. Non-parametric procedures (Mann-Whitney U test when 2 groups were compared, Kruskal-Wallis one-way analysis of variance followed by Dunn’s post-hoc test when 3 groups or more were compared) were therefore applied. Linear regressions were applied to examine trends of stable isotope composition with fish size (SL) and with stomach contents (% of zooplankton) (Frédérich et al. 2010). The ANOSIM, ANOVA, Mann-Whitney U test, Kruskal-Wallis test and associated post-hoc tests, and linear regressions were performed using the statistical software PAST (Hammer et al. 2001). For fish groups with N ≥ 5, isotopic niche parameters were computed using the SIBER package (Version 2.0; Jackson et al. 2011) for R (R Development Core Team 2015). SIBER was used to generate bivariate standard ellipses that represent core isotopic niches of consumers. Areas of these ellipses were estimated using correction for small sample size (SEAC, Jackson et al. 2011). Areas of the ellipses associated to each species were also estimated using Bayesian modelling (SEAB, 106 iterations), and direct pairwise comparisons of SEAB were performed. Model solutions were presented using credibility intervals of probability density function distributions. Pairwise comparisons were considered meaningful when probability of occurrence exceeded 95%. RESULTS Stomach contents A large proportion of the 182 examined stomachs were found to be empty (N = 55; Table 1). For the five species captured in 2011 and 2012, the numbers of stomachs containing prey were very low. Moreover, very little prey material was found in the non-empty stomachs. Therefore, only frequency of occurrences of all dietary categories was investigated for species sampled in 2011 and 2012 (Table 2). On the other hand, both frequency of occurrences and percentages of composition for each prey category were calculated for species collected in 2014, where sufficient amounts of prey material were observed (Tables 2 and 3). Generally speaking, apogonids showed a carnivorous diet. The eight species mainly fed on zooplankton (copepods, crustacean larvae, polychaete larvae and chaetognaths) and small benthic invertebrates (amphipods, small decapods and harpacticoid copepods; Table 2). A significant amount of unrecognized prey items was present in all studied species. Planktonic copepods and decapods were the most recurrent prey found in stomachs. Algae were never ingested, except by few individuals (N = 3) of Ostorhinchus aureus in 2014. Small fishes were observed in the stomach of Pristiapogon kallopterus in both 2011 and 2014. Temporal variation in the type of prey selected was observed. For example, O. cyanosoma, P. fraenatus and P. kallopterus were used to feed on harpacticoids in 2014 but not in 2011. Planktonic copepods were largely encountered in the stomachs of P. fraenatus in 2014 only. The ANOSIM test performed on percentage composition of stomach contents of fishes page 4 of 14Zoological Studies 56: 10 (2017) sampled in 2014 was significant (P < 0.001), suggesting that interspecific differences were present in stomach content composition. However the ANOSIM R statistic was very low (R = 0.23), suggesting that the “species” factor was not a major driver of stomach content similarity because the composition of stomach contents greatly varied within each apogonid species (Table 3). The ANOVAs revealed different feeding strategies among apogonids (zooplankton: F = 6.313, d.f. = 5,81, P < 0.001; zoobenthos: F = 3.089, d.f. = 5,81, P = 0.013). Pristiapogon kallopterus consumed significantly less planktonic prey than the other species. Zoramia leptacantha foraged less in the Table 2. Frequency of occurrence (%) of all dietary categories in the eight studied species of apogonids. For species collected during different sampling campaigns, results are shown for every year. C. = Cheilodipterus, O. = Ostorhinchus, P. = Pristiapogon, T. = Taeniamia, Z. = Zoramia C. quinquelineatus O. aureus O. cookii O. cyanosoma P. fraenatus P. kallopterus T. fucata Z. leptacantha Prey category 2012 2014 2011 2012 2011 2014 2011 2014 2011 2014 2014 2014 Planktonic animal preys Copepods 0 79 0 0 33.3 71.4 0 58.3 0 18.2 31.6 100 Crustacean larvae 0 26.3 0 0 0 28.6 7.7 8.3 0 0 57.9 91.7 Polychaete larvae 100 0 0 18.2 33.3 0 0 0 0 0 0 0 Chaetognaths 05.3 0 0 0 0 0 0 0 0 0 0 Benthic animal preys Harpacticoid copepods 0 36.8 0 0 0 35.7 0 16.7 0 9.09 0 83.3 Amphipods 0 10.5 0 0 16.7 21.4 7.7 16.7 0 0 5.3 0 Ostracods 021.1 0 0 0 0 0 0 16.7 0 0 0 Decapods 100 26.3 100 81.8 66.7 28.6 84.6 66.7 100 90.9 63.2 16.7 Gastropods 05.3 0 0 0 0 0 0 0 0 0 8.3 Polychaetes 0 0 0 0 0 0 0 16.7 0 0 0 0 Algae 0 15.8 0 0 0 0 0 0 0 0 0 0 Fish 0 0 0 0 0 0 0 0 0 9.1 0 0 Fish scales 05.3 0 9.1 0 0 0 0 16.7 0 5.3 0 Eggs 0 0 0 0 0 7.1 0 16.7 0 9.1 0 0 Unrecognized 100 100 100 100 100 100 100 100 100 90.9 100 83.3 Table 3. Mean percent composition of all dietary categories in the six species of apogonids collected in 2014. Results are presented as Mean (%) ± SD. O. = Ostorhinchus, P. = Pristiapogon, T. = Taeniamia, Z. = Zoramia Prey category O. aureus O. cyanosoma P. fraenatus P. kallopterus T. fucata Z. leptacantha Planktonic animal preys 53.3 ± 33.5 57.9 ± 38.0 28.8 ± 32.9 11.6 ± 26.6 59.3 ± 39.2 82.8 ± 13.3 Copepods 35.7 ± 31.9 43.0 ± 37.2 27.4 ± 33.1 11.6 ± 26.6 16.3 ± 29.2 54.0 ± 22.9 Crustacean larvae 16.7 ± 30.9 14.8 ± 25.3 1.4 ± 4.8 0.0 43.0 ± 41.6 28.8 ± 21.9 Polychaete larvae 0.0 0.0 0.0 0.0 0.0 0.0 Chaetognaths 0.9 ± 3.8 0.0 0.0 0.0 0.0 0.0 Benthic animal preys 40.2 ± 33.3 35.0 ± 35.6 63.5 ± 34.3 70.2 ± 43.3 39.0 ± 40.3 17.2 ± 13.3 Harpacticoid copepods 19.2 ± 32.8 11.3 ± 18.9 5.6 ± 14.8 3.0 ± 10.1 0.0 16.3 ± 13.8 Amphipods 6.3 ± 23.1 10.2 ± 27.4 10.4 ± 29.1 0.0 1.1 ± 4.6 0.0 Ostracods 8.5 ± 18.9 0.0 0.0 0.0 0.0 0.0 Decapods 5.6 ± 11.0 13.5 ± 28.8 44.7 ± 41.4 67.2 ± 42.2 37.9 ± 40.0 0.7 ± 1.7 Gastropods 0.6 ± 2.5 0.0 0.0 0.0 0.0 0.3 ± 0.9 Polychaetes 0.0 0.0 2.8 ± 6.5 0.0 0.0 0.0 Algae 4.3 ± 10.5 0.0 0.0 0.0 0.0 0.0 Fish 0.0 0.0 0.0 9.1 ± 30.2 0.0 0.0 Fish scales 2.3 ± 9.8 0.0 0.0 0.0 1.8 ± 7.6 0.0 Eggs 0.0 7.1 ± 26.7 7.8 ± 19.0 9.1 ± 30.2 0.0 0.0 page 5 of 14Zoological Studies 56: 10 (2017) benthic compartment than P. kallopterus and P. fraenatus (Tables 3 and 4). No significant variation was shown for the other species (Table 4). Stable isotopes Zooplankton δ13C was comparable in 2011 and 2012, but less negative in 2014 (Fig. 1A). Carbon isotopic composition of zoobenthos was variable according to sampling year (Fig. 1A), but it was more 13C-enriched than zooplankton in both 2011 and 2014. The difference between mean δ13C of the two food items varied drastically according to sampling year, as it was 5.73‰ in 2011 but only 0.97‰ in 2014. δ15N of food items did not seem to follow a consistent temporal variation pattern, and was quite comparable for both food items in every sampling year (Fig. 1A). Carbon isotopic composition of cardinalfishes was spread over a large interval (Fig. 1A), with values ranging from -11.14 ± 0.45‰ (Ostorhinchus cookii in 2011; mean ± SD) to -16.97 ± 0.37‰ (Taeniamia fucata in 2014; mean ± SD). Nitrogen isotopic composition also showed considerable dispersion (Fig. 1A), as values ranged from 6.61 ± 0.19‰ (O. cookii in 2011; mean ± SD) to 10.14 ± 0.61‰ (P. kallopterus in 2014; mean ± SD). This isotopic variability was partly related to the sampling year, as δ13C and δ15N seemed to shift towards more negative and higher values throughout time, respectively. However, speciesspecific trends were also present. For example, δ15N of P. kallopterus and P. fraenatus were identical in 2011 (Mann-Whitney test: U = 19, P = 0.113; Fig. 1A), but P. kallopterus showed significantly higher δ15N than P. fraenatus in 2014 (Mann-Whitney test: U = 72.5, P < 0.001; Fig. 1A). In 2011, δ13C of fishes showed significant interspecific variation (Kruskal-Wallis test: H = 34.12, P < 0.001). Post-hoc multiple comparisons (Dunn’s test) showed that two groups were present (Fig. 1A): one composed of P. fraenatus and O. cookii, and another composed of P. kallopterus and O. cyanosoma. The latter group had more negative δ13C than the former, suggesting that, in 2011, zooplankton was more important in the diet of P. kallopterus and O. cyanosoma than in the diet of P. fraenatus and O. cookii. In 2014, significant variation in δ13C was also present (Kruskal-Wallis test: H = 109.9, P < 0.001). Three groups were present (Dunn’s multiple comparison tests: P < 0.05 in each case): one composed of both species of Pristiapogon (P. fraenatus and P. kallopterus), one composed of both species of Ostorhinchus (O. cookii and O. aureus), and one composed of T. fucata and Z. leptacantha (Fig. 1A). δ13C decreased, and contribution of zooplankton to diet therefore presumably increased, when going from the first to the third group (Fig. 1A). Bivariate standard ellipses of all fish groups were markedly separated (Fig. 1B). This suggests that all cardinalfishes occupy distinct isotopic niches. The only niche overlap found was between P. fraenatus and P. kallopterus in 2014, and it was very small (0.04‰2, i.e. 8.46% of the SEAC of P. fraenatus and 5.05% of the SEAC of P. kallopterus; Fig. 1B). Moreover, there was no overlap between standard ellipses associated with different years for fishes sampled in more than one period (O. cookii, O. cyanosoma, P. fraenatus and P. kallopterus; Fig. 1B). Isotopic niche width was quite variable, with SEAC values ranging from 0.13‰2 (P. kallopterus in 2011) to 0.89‰2 (P. kallopterus in 2014; Fig. 2). Pairwise comparisons of modelestimated ellipse areas (SEAB) suggested niche width differences among species were robust, as many relative probabilities exceeded 95% (Table 5). Trends for species sampled in successive years were not consistent. Standard ellipse area of P. kallopterus showed a drastic increase from 2011 Table 4. Results from Tukey multiple comparisons tests when using data on planktonic (below the diagonal) and benthic animal preys (above the diagonal). Significant results are highlighted in italics and marked with an asterisk. O. = Ostorhinchus, P. = Pristiapogon, T. = Taeniamia, Z. = Zoramia O. aureus O. cyanosoma P. fraenatus P. kallopterus T. fucata Z. leptacantha O. aureus 0.999 0.602 0.239 0.999 0.747 O. cyanosoma 0.999 0.357 0.106 10.925 P. fraenatus 0.466 0.303 0.989 0.467 0.045* P. kallopterus 0.012*0.005*0.587 0.158 0.007* T. fucata 0.984 0.999 0.140 0.001*0.856 Z. leptacantha 0.402 0.579 0.005* < 0.001*0.818 page 6 of 14Zoological Studies 56: 10 (2017) to 2014 (SEAB, 2011 < SEAB, 2014 in 99.94% of model solutions; Fig. 2 and Table 5). On the other hand, the isotopic niche of O. cyanosoma showed a width decrease from 2011 to 2014 (SEAB, 2014 < SEAB, 2011 in 100% of model solutions; Fig. 2 and Table 5). Probabilities of standard ellipse area differences in O. cookii (SEAB, 2011 < SEAB, 2012 in 87.58% of model solutions) and P. fraenatus (SEAB, 2011 < SEAB, 2014 in 11.21% of model solutions; Fig. 2 and Table 5) were inferior to 95%, suggesting no meaningful temporal trend in niche width in these taxa. Finally, there was no relation between SEAC and the size range of sampled fishes (linear regression analysis, R2 = 0.15, P = 0.24), suggesting that the Fig. 1. (A) Mean values (± SD) of δ13C (‰) and δ15N (‰) of cardinalfishes. (B) Isotopic niches of cardinalfishes. Points are individual measurements, and solid lines represent the bivariate standard ellipses associated to each group Species and sampling years are represented by different symbols and colors, respectively. (A) (B) page 7 of 14Zoological Studies 56: 10 (2017) isotopic niche size was not related to variation in size ranges of the studied species (Table 1). Linear regression analyses revealed that the variation of carbon or nitrogen isotopic compositions was size-related in most of the apogonids (0.22 ≤ R2 ≤ 0.90; Fig. 3). However, size range varied greatly among species samples (Table 1) and that could have impacted the results of linear models. For example, Z. leptacantha was the only species for which the isotopic compositions were unexplained by body size but its size range was the smallest from all studied species (size range = 5.5 mm; Table 1). In P. fraenatus, P. kallopterus and T. fucata, there was a strong, Table 5. Pairwise comparisons of standard ellipses areas of cardinalfishes estimated through Bayesian modeling (SEAB). Each cell contains the relative probability (%) that the standard ellipse of the fish group listed as line is smaller than the standard ellipse of the fish group listed as column, based on 106 model runs. Values highlighted in italics and marked with an asterisk are probabilities higher than 95%. O. = Ostorhinchus, P. = Pristiapogon, T. = Taeniamia, Z. = Zoramia O. cookii 2011 O. cyanosoma 2011 P. fraenatus 2011 P. kallopterus 2011 O. cookii 2012 O. aureus 2014 O. cyanosoma 2014 P. fraenatus 2014 P. kallopterus 2014 T. fucata 2014 Z. leptacantha 2014 O. cookii 2011 - 98.67* 99.38* 10.89 87.58 89.28 13.04 96.53* 99.77* 96.83* 10.35 O. cyanosoma 2011 1.33 -65.45 0.25 10.10 6.62 0.00 22.31 79.16 25.28 0.00 P. fraenatus 2011 0.62 34.55 -0.14 5.12 2.50 0.00 11.21 66.24 13.17 0.00 P. kallopterus 2011 89.11 99.75* 99.86* - 98.21* 98.51* 70.07 99.46* 99.94* 99.48* 62.31 O. cookii 2012 12.42 89.90 94.88* 1.79 -49.78 0.40 76.20 97.96* 78.48 0.46 O. aureus 2014 10.72 93.38 97.50* 1.49 50.22 -0.15 80.83 99.25* 82.73 0.27 O. cyanosoma 2014 86.96 100.00* 100.00* 29.93 99.60* 99.85* - 99.99* 100.00* 100.00* 39.16 P. fraenatus 2014 3.47 77.69 88.79 0.54 23.80 19.17 0.01 - 95.63* 53.85 0.00 P. kallopterus 2014 0.23 20.84 33.76 0.06 2.04 0.75 0.00 4.37 -5.38 0.00 T. fucata 2014 3.17 74.72 86.83 0.52 2.04 17.27 0.00 46.15 94.62 -0.03 Z. leptacantha 2014 89.65 100.00* 100.00* 37.69 99.54* 99.73* 60.84 100.00* 100.00* 99.97* - Fig. 2. Boxplots of model-estimated bivariate standard ellipse area (SEAB). Dark, median and light grey boxes are respectively the 50%, 75% and 95% credibility intervals of the probability of density function distributions of the model solutions, and black dots are the modes of these distributions. Red dots represent the standard ellipse areas computed using a frequentist algorithm adapted for small sample sizes (SEAC). page 8 of 14Zoological Studies 56: 10 (2017) Fig. 3. Relationship between body size (SL, mm) and isotopic values (A: δ13C; B: δ15N) in studied apogonids. Only significant relationships are illustrated. 48 54 60 66 -12.0 -11.5 -11.0 -10.5 37.5 40.0 42.5 45.0 -14.0 -13.6 -13.2 -12.8 42.5 45.0 47.5 50.0 -15.00 -14.75 -14.50 -14.25 44 48 52 56 -14.4 -14.2 -14.0 -13.8 64 72 80 88 -14.5 -14.0 -13.5 -13.0 (A) (B) 50 60 70 80 90 8.4 8.8 9.2 9.6 48 56 64 72 6.25 6.50 6.75 7.00 36 39 42 45 48 6.6 7.2 7.8 8.4 60 70 80 90 7.2 8.0 8.8 9.6 10.4 60 70 80 90 8.4 9.0 9.6 10.2 10.8 64 68 72 76 8.4 9.0 9.6 10.2 R2 = 0.61 P = 0.01 R2= 0.32 P = 0.03 R2 = 0.22 P = 0.04 R2 = 0.83 P = 0.04 R2 = 0.44 P = 0.006 R2 = 0.31 P = 0.004 R2 = 0.69 P = 0.008 P = 0.04 R2 = 0.86 P = 0.0001 R2 = 0.90 P = 0.0001 R2 = 0.46 P = 0.001 δ15N δ15N δ15N δ15N δ15N δ15N R2 = 0.29 δ13C δ13C δ13C δ13C δ13C O. cookii (2011) O. cyanosoma (2011) O. cyanosoma (2014) SL (mm) P. kallopterus (2011) P. kallopterus (2014) O. aureus (2014) O. cookii (2011) O. cyanosoma (2011) P. fraenatus (2014) P. kallopterus (2014) T. fucata (2014) SL (mm) SL (mm) SL (mm) SL (mm) SL (mm) SL (mm) SL (mm) SL (mm) SL (mm) SL (mm) page 9 of 14Zoological Studies 56: 10 (2017)