How To Determine The Relationship Between Butterflyfish And Coral

Transcription

1 AQUATIC CONSERVATION: MARINE AND FRESHWATER ECOSYSTEMS Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5) Published online in Wiley InterScience ( DOI:.2/aqc.62 The use of butterflyfish (Chaetodontidae) species richness as a proxy of total species richness of reef fish assemblages in the Western and Central Pacific M. KULBICKI* and Y. M. BOZEC IRD } Universite de Perpignan, 668 Perpignan, France ABSTRACT 1. A new use of butterflyfishes (Chaetodontidae) is proposed for monitoring programmes, based on the relationship existing between the total species richness of reef fish assemblages and the species richness of butterflyfishes. These two variables are highly correlated and it is possible under certain circumstances to predict the total species richness of a reef fish assemblage from the butterflyfish species richness. 2. In the present study, the effects on this relationship of three regions (Tuamotu, Tonga, New Caledonia), eight islands, two reef types (barrier and fringing reefs) and coral cover were investigated based on 544 transects. The effect of time was also tested based on 8 transects (12 stations trimesters). 3. Coral cover had no statistically significant effect on the relationship; both slope and intercept of the relationship varied with region. 4. Islands were a significant factor in Tonga and New Caledonia, but not in Tuamotu. 5. On fringing reefs the correlations were higher and the linear regressions had flatter slopes and lower intercepts than on barrier reefs, while the correlations between total species richness and butterflyfish species richness were not influenced by time over a 3-month period. 6. The potential applications of this relationship for monitoring programmes are discussed on the basis of these results and on a power analysis relating butterflyfish diversity and the correlation level of this relationship. Copyright # 5 John Wiley & Sons, Ltd. KEY WORDS: species richness; monitoring; Chaetodontidae; Pacific; coral reef INTRODUCTION Chaetodontidae (butterflyfishes) are probably the most extensively sampled reef fish family in the Indo- Pacific to date. The purposes of these studies range from biological (diet and behaviour mainly) (e.g. Harmelin-Vivien and Bouchon-Navaro, 183; Sano et al., 184; Bouchon-Navaro, 186; Findley and *Correspondence to: M. Kulbicki, IRD } Université de Perpignan, 668 Perpignan, France. michel.kulbicki@univ-perp.fr Copyright # 5 John Wiley & Sons, Ltd.

2 S128 M. KULBICKI AND Y.M. BOZEC Findley, 18, 1; Harmelin-Vivien, 18; Motta, 18; Roberts and Ormond, 12; Cox, 14; Chabanet et al., 17; Lewis, 18), to ecological (especially relationships between butterflyfishes and their environment) (e.g. Bell et al., 185; Sano et al., 187; Bouchon-Navaro and Bouchon, 18; Fowler, 1; Roberts et al., 12; Cadoret et al., 1), or biogeographical (Blum, 18; Findley and Findley, 1). Butterflyfishes are also an increasing part of reef monitoring programmes at the national (e.g. Great Barrier Reef Marine Park Authority and Australian Institute of Marine Sciences monitoring programmes in Australia), regional (e.g. PROCFISH programme by South Pacific Commission) and international (e.g. Global Coral Reef Monitoring Network) levels. It has also been proposed that butterflyfishes might be used to monitor the ecological status of coral reefs (Reese, 181; Bouchon-Navaro et al., 185; Hourigan et al., 188; Roberts et al., 188; White, 188; Crosby and Reese, 16; Erdman, 17; O hman et al., 18; Khalaf and Crosby, 5; Temraz and Abou Zaid, 5; Samways, 5). Most of the interest in these fish stems from: (1) their relationship with coral, as a number of these species are coral-dependent; (2) the ease of identification; (3) the ease of censusing; and (4) their very wide geographical range. Despite the fact that their relationship to coral has been rather well studied (e.g. Reese, 177, 181; Bell and Galzin, 184; Bell et al., 185; Roberts et al., 188; White, 188; Cadoret et al., 15, 1; O hman et al., 18; Findley and Findley, 1), the use of these fish in general reef monitoring is often ill-defined and their use for reef ecological monitoring is open to debate (Erdman, 17; O hman et al., 18). For these reasons it is important to explore new applications for using these fishes. Bell and Galzin (184) found that several fish families, including butterflyfishes, as well as entire reef fish assemblages showed positive correlations with coral cover. Since there were good correlations between total reef fish species richness (SR) and coral cover and similar correlations between butterflyfish SR and coral cover, there should be links between total reef fish SR and butterflyfish SR. If such relationships exist and are highly significant, then butterflyfish SR could be used as a proxy of total reef fish SR. If this hypothesis proves to be correct under a wide enough range of circumstances, then the use of butterflyfishes might expand to include the spatial comparison of reef fish assemblages, monitoring of a reef over time, wider biogeographical studies, and a wider range of ecological studies than those for which they are currently used. To demonstrate that butterflyfish SR may in some circumstances be used as a proxy of total reef fish assemblage SR, this paper has four goals: 1. To demonstrate that the relationship between total reef fish SR and butterflyfish SR is highly significant and that the confidence intervals are sufficiently narrow to enable useful predictions of total reef fish SR from observed butterflyfish SR. 2. To analyse the variations of this relationship at various spatial scales. Kulbicki et al. (5) show that the relative importance of butterflyfishes to the total reef fish species pool is a function of the total number of species, so that the higher the number of species known on an island, the lower is the proportion of butterflyfishes. In addition these authors found that the biological and ecological characteristics of the butterflyfish regional species pool and of the butterflyfishes observed on a given island may not correspond. These findings suggest that the relationship between total reef assemblage SR and butterflyfish SR may be a function of the region, island and biotope. 3. To test whether the relationship between total reef fish assemblage SR and butterflyfish SR is sensitive to the level of coral cover. Since coral cover is known to influence both total reef assemblage SR and butterflyfish SR (Reese, 181; Harmelin-Vivien and Bouchon-Navaro, 183; Bell and Galzin, 184; Bouchon-Navaro and Bouchon, 18; Chabanet et al., 17; Cadoret et al., 1; Jones et al., 4), and because several species of butterflyfishes are obligate coral feeders with their diversity linked to coral cover, it will be useful to test whether the ratio of N chaet species of butterflyfishes to N tot species for the entire reef fish assemblage in an area with low coral cover is the same as in an area with a high coral cover. Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

3 BUTTERFLYFISH SPECIES RICHNESS IN WESTERN AND CENTRAL PACIFIC S12 4. To analyse whether time variations are important, and in particular, whether they are more important than spatial ones. In order to be useful in monitoring programmes, this relationship has to be stable through time for a given area. Finally the conditions and limits of use of these first findings and the circumstances in which these relationships were found to be of interest for developing proxies of total reef fish assemblages, will be discussed. METHODS The data analysed in this study came from various experiments conducted across the Pacific using the same sampling procedures. Three regions were visited } Tuamotu (French Polynesia), Tonga, and New Caledonia. They are at least km distant from one another and have decreasing numbers for total species numbers (Table 1). Several islands were sampled within each of these regions. All the data analysed came from Underwater Visual Censuses (UVC) transects. These transects were 5 m long and placed in such a way that the habitat covered was as homogeneous as possible. For each transect the following parameters were recorded: biotope (fringing or barrier reef), coral cover (as percentage of bottom cover), number of butterflyfish species (butterflyfish SR), and number of all other fish species which could be observed (total SR). The sampling design is given in Table 2. All transects were included in one general linear model (GLM) model: SR total ¼ X þ C þ F þ X F þ e Number of species Table 1. Total reef fish diversity and butterflyfish diversity of the three regions studied New Caledonia to Tonga: km to Tuamotu: 5 km Tonga to New Caledonia: km to Tuamotu: 3 km Total Chaetodontidae % Chaetodontidae Tuamotu to New Caledonia: 5 km to Tonga: 3 km Table 2. Number of transects according to reef type (fringing or barrier), region (New Caledonia, Tonga or Tuamotu) and islands Fringing Barrier Total New Caledonia 253 Main Island 88 6 Uvea Atoll 6 Tonga 5 Tongatapu 5 22 Hapai 15 1 Vavau Tuamotu 16 Kauehi 64 Marokau 64 Nihiru 68 Total Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

4 S13 M. KULBICKI AND Y.M. BOZEC where X is butterflyfish SR, C is coral cover and F is the factor which characterizes the transect; F will be expressed as Region Island Reef-type with Reef-type crossed with Island and Region, whereas Island is nested in Region. If X is significant then total SR and butterflyfish SR are linked by a linear relationship; and if F is significant then the intercept of at least two combinations of the factors making F are significantly different; and if X F is significant then the slope of at least two combinations of the factors making F are significantly different. To test the separate effects of region and reef type on the relationship between total SR and butterflyfish SR a covariance analysis restricted to Tonga (all islands) and New Caledonia (Main Island only) was conducted according to the model: SR total ¼ SR butterflyfish þ Region þ Reef-type þ Region Reef-type þ e The effects of time variations were tested only for the fringing reefs of New Caledonia s largest island. This analysis was performed on 12 stations (S) in which reef fish assemblages were sampled every three months over a period of 3 months, with one sampling period skipped due to bad weather conditions ð12 ¼ 8 transectsþ. Coral cover was not included in this analysis as this variable did not change significantly within stations over the time period tested. The effect of time was tested in two steps. First, whether there was a relationship between total SR and butterflyfish SR for all stations and within stations through time. The covariance model was: SR total ¼ X þ S þ X S þ e Second, if time changed this general relationship by applying a second model: SR total ¼ X þ T þ X T þ e in which T is the trimester. RESULTS Spatial comparisons The simple correlation between total reef fish SR and butterflyfish SR was highly significant (r 2 ¼ :61 for N ¼ 534, p5 8 ). The correlations between total SR (r 2 ¼ :7, p5 4 ) with coral cover and butterflyfish SR (r 2 ¼ :23, p5 4 ) with coral cover were also highly significant, but less so. A multiple regression relating total reef fish SR to both butterflyfish SR and coral, however, indicates that coral cover was not a significant factor once butterflyfish SR is taken into account. (total r 2 ¼ :61; F coral ¼ 3:7, p coral ¼ :6; F SRbutterflyfish ¼ 67, p SRbutterflyfish 5 8 ). These correlations do not take into account the fact that several factors play a significant role as indicated by the result of the GLM (1) in Table 3. This model confirms the highly significant correlation between total SR and butterflyfish SR and the nonsignificant role of coral cover once butterflyfish SR was taken into account. In addition, it shows that region, island and reef type play significant roles in these relationships. Each factor was examined separately and then the various interactions by a covariance analysis and graphically. First, each region yielded a significantly different slope and intercept (Figure 1), with Tonga having the highest values and Tuamotu the lowest. In order to avoid differences between regions due to unequal proportions of each reef type within each region, only barrier reefs were considered for this figure (as they were common to all three regions), but similar findings were found also for fringing reefs between Tonga and New Caledonia (Table 4). Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

5 BUTTERFLYFISH SPECIES RICHNESS IN WESTERN AND CENTRAL PACIFIC S131 Table 3. Major results of the model (1) relating total SR to butterflyfish SR and coral cover according to region, island and reef type. : :5 > p > :1; : :1 > p > :1; : p5:1 Estimate Standard-error t-value Intercept *** Butterflyfish SR *** Coral cover Region : Tonga * Region : Tuamotu *** Island : Hapai Island : Kauehi Island : Marokau Island : Ouvea ** Island : Tongatapu *** Reef : fringing *** Region : Tonga: Reef : fringing ** Island : Hapai : Reef : fringing Island : Tongatapu : Reef : fringing *** 1 New Caledonia Tonga Tuamotu 5 15 Figure 1. Relationship between total SR and butterflyfish SR for the barrier reefs of the three regions, without taking island into account. Each point represents a transect. : all observed species with butterflyfishes excluded. The plain line is for New Caledonia, the dashed line for Tonga and the dotted line for Tuamotu. The analysis of the effect of reef types indicated that this factor has a very significant effect on the relationship between total SR and butterflyfish SR (Tables 3 and 4). This effect was not the same between regions and between islands (Table 3). Differences between regions (Table 5) were examined using a covariance analysis limited to Tonga and New Caledonia where both reef types were sampled. The covariance analysis confirmed the highly significant differences between regions for this factor. Differences within regions were also significant (Table 5) as illustrated by Figure 2. Within Tonga there were significant Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

6 S132 M. KULBICKI AND Y.M. BOZEC Table 4. Major parameters of the regressions within cells of model (1) which make a significant contribution to the model (e.g. the islands from Tuamotu are not given separately as they are not significantly different from one another). N: number of transects; r=correlation coefficient; p=probability of accepting H (there is no linear relationship between total SR and butterflyfish SR) Region Island Reef-type N r p Intercept Slope New Caledonia Ouvea Barrier Main Island Barrier Main Island Fringing Tonga Hapai Fringing Vavau Fringing Vavau Barrier Tongatapu Barrier Tuamotu All Barrier reef Table 5. Covariance analysis restricted to Tonga (all islands) and New Caledonia (Main Island only) testing the differences in the relationship between total SR and butterflyfish SR according to region and reef type, the model used being: SR total ¼ SR butterflyfish þ Region þ Reef-type þ Region Reef-typeþe Degrees of Estimates F Probability (> t ) freedom Intercept Butterflyfish SR Region Reef-type Region*Reef-type Error Tonga 1 New Caledonia Barrier Fringing 5 15 Number of Butterflyfish species Barrier Fringing 5 15 Figure 2. Relationship between total SR and butterflyfish SR for the two reef types according to region. Slopes and intercepts are significantly different in each case. Each point represents a transect. : all observed species with butterflyfishes excluded. The dashed lines correspond to fringing reefs, the plain lines to barrier reefs. differences for this relationship between and within islands as well (Table 3; Figure 3) according to reef type. Within each region, islands were not necessarily a significant factor. In particular, in Tuamotu, the three islands sampled showed no difference in either slope or intercept (Tables 3 and 4), whereas in Tonga and Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

7 BUTTERFLYFISH SPECIES RICHNESS IN WESTERN AND CENTRAL PACIFIC S133 Vavau Hapai 1 1 Barrier Fringing Barrier Fringing Figure 3. Relationship between total SR and butterflyfish SR for the two reef types according to island in Tonga. Slope and intercept are significantly different for Hapai, only the intercept is significantly different for Vavau. Each point represents a transect. All Species : all observed species with butterflyfishes excluded. The dashed lines correspond to fringing reefs, the plain lines to barrier reefs. 1 New Caledonia 1 Tonga Ouvea Main Island 5 15 Hapai Vavau Tongatapu 5 15 Figure 4. Relationship between total SR and butterflyfish SR for different islands within a region, reef type being restricted here to barrier reefs. Slopes and intercepts are significantly different between Ouvea and Main Island; intercepts are different between Hapai and Vavau or Tongatapu, whereas slopes are different between Tongatapu and Vavau. Each point represents a transect. : all observed species with butterflyfishes excluded. The dashed lines correspond to either Main Island or Vavau, the plain lines to either Ouvea or Hapai and the dotted line to Tongatapu. New Caledonia, there were differences (Tables 3 and 4). In order to illustrate such differences between islands within a region, the reef type has to be similar; e.g. for barrier reefs within New Caledonia and Tonga (Figure 4). In particular, the relationships were different for each island. The general model indicated that reef type is an extremely significant factor (Tables 3 and 4), but did not make it possible to say how fringing and barrier reef stations are different in their intercepts and slopes as this factor interacts with region and island. Comparison of the slopes and intercepts of each significant cell Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

8 S134 M. KULBICKI AND Y.M. BOZEC 1 Barrier 1 Fringing y = 2.5 x R = y = 4.47 x R = Figure 5. Relationship between total SR and butterflyfish SR for the two reef types, without taking into account region or island. Each point represents a transect. : all observed species with butterflyfishes excluded. (Table 4) suggests that the intercepts are higher and slopes more gentle for barrier reefs than for fringing reefs. Grouping all barrier reef stations versus all fringing reef stations (without taking into account region or island) confirms this trend (Figure 5), with the relationship more significant for fringing reefs than for barrier reefs. Temporal comparisons The analysis of time series data was a two-step process. First, significance of the relationship between total SR and butterflyfish SR was determined using model (2). This analysis indicated that the relationship between total SR and butterflyfish SR was highly significant when all transects were considered and that there were significant differences between stations for this relationship (Table 6). Second, the influence of time on this relationship was tested and no significant global effect was found (Table 7). On a case by case basis, there was a significant effect of trimester seven; i.e. the slope or intercept of the relationship between total SR and butterflyfish SR did not change from one trimester to the next (Figure 6; Table 7), except for trimester seven. The exception represented by the trimester 7 was linked to the recruitment of many small reef species during that period (March). Power analysis In order to give an indication of the predictive potential of the relationships between total SR and butterflyfish SR, a power analysis was performed on the data from Tonga and New Caledonia (Figure 7). For Tonga the power was lower than for New Caledonia, but this was related to the sampling effort which was nearly three times larger for New Caledonia. The top curves (Figure 7(a)) give the variations in predicted number of species. It shows in particular that the predicted range for the fringing reefs of New Caledonia was narrower than for the barrier reefs in that region, but it is the opposite for Tonga. However, these top curves do not take into account the number of observed butterflyfishes. Indeed, if the number of butterflyfish species is low, and even if the confidence interval is narrow, it may represent a sizeable amount of the total SR. This is illustrated by the bottom power curves (Figure 7(b)), with a somewhat counterintuitive result. The power for the New Caledonian fringing reefs was lower than for the barrier reefs of that region, despite a much better correlation between total SR and butterflyfish SR (Table 4). This drop in power was linked to the low numbers of butterflyfish species observed on these fringing reefs (5.8 species on average) compared to barrier reefs (7.7 for Ouvea, 8.2 for the Main Island). For Tonga the best Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

9 BUTTERFLYFISH SPECIES RICHNESS IN WESTERN AND CENTRAL PACIFIC S135 Table 6. Covariance analysis of model (2) testing the effect of stations on the relationship between total SR and butterflyfish SR Degrees of Estimates MS F Probability (> t ) freedom Intercept Butterflyfish SR Station Error Table 7. Covariance analysis of model (3) testing the effect of time (trimesters) on the relationship between total SR and butterflyfish SR Degrees of Estimates MS F Probability (> t ) freedom Intercept Butterflyfish SR Trimester Error power curve in species number (Figure 7(a)) was for the barrier reef which also has the highest diversity of butterflyfishes (8. species on average, versus 6.3 to 7. for fringing reefs). Therefore the order of the power curves for Tonga stays the same for the two types of curves (Figures 7(a) and (b)). DISCUSSION Butterflyfishes, especially coral feeding species, have been repeatedly proposed as indicators of the ecological status of coral reefs (Reese, 181; Hourigan et al., 188; Crosby and Reese, 16; Crosby and Reese, 5; Khalaf and Crosby, 5; Temraz and Abou Zaid, 5; Samways, 5). They have a number of qualities essential to indicators for management purposes } they are easy to recognize, they are Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

10 S136 M. KULBICKI AND Y.M. BOZEC TRIMESTRE: 1 TRIMESTRE: 2 TRIMESTRE: TRIMESTRE: 4 TRIMESTRE: 5 TRIMESTRE: TRIMESTRE: 7 TRIMESTRE: 8 TRIMESTRE: Figure 6. Relationship between total SR and butterflyfish SR for 12 fringing reef stations in New Caledonia (the numbers on the charts represent the stations identification numbers), each station being sampled every trimester over 3 months (nine replicates). Trimester 7 has a significantly higher intercept than the other trimesters (Table 7). easy to detect and census, they are site attached, and they are probably long-lived (for example, Chaetodon larvatus could reach 14 years according to Zekeria (3)). The use of these fishes as indicators of the ecological status of reefs is, however, a source of debate (Erdman, 17; O hman et al., 18). These fishes are currently sampled in a number of monitoring programmes (Kulbicki et al., 5), but so far, there is no specific use for the data collected. In most instances, butterflyfish diversity or abundance is correlated with coral cover or diversity. However, these programmes as well as a large number of other field works have generated a wealth of data on their distribution (e.g. Bouchon-Navaro, 181; Bell et al., 185; Findley and Findley, 18, 1; Fowler, 1; Roberts et al., 12; Cadoret et al., 15, 1), behaviour (e.g. Bouchon-Navaro, 186; Driscoll and Driscoll, 188; Roberts and Ormond, 12; Cox, 14) and relationship with environmental variables (e.g. Bell and Galzin, 184; Bouchon-Navaro et al., 185; Jennings et al., 16; Chabanet et al., 17; O hman et al., 18; Zekeria, 3). The present study shows that these fishes may, in a number of circumstances, be good proxies of the species richness of the entire fish assemblage. Total species richness has a number of uses in ecology, e.g. exploring issues related to the resistance, resilience and stability of communities (Peterson et al., 18) diversity biomass relationships and production potential (Kulbicki et al., 4), biogeography (Hillebrand and Blenckner, 2). Our Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

11 BUTTERFLYFISH SPECIES RICHNESS IN WESTERN AND CENTRAL PACIFIC S NEW CALEDONIA 45 TONGA Delta Total SR 3 Ouvea Barrier Fringing Main Island Delta Total SR Hapai-Fringing Vavau-Fringing Tongatapu-Barrier 15 (a) Probability level (x) Probability Level (x) 7 NEW CALEDONIA 7 TONGA Delta Total SR as % 5 3 Barrier Fringing Main Island Delta SR Total as % 5 3 Hapai-Fringing Vavau-Fringing (b) Ouvea Probality Level (x) Tongatapu-Barrier Probability Level (x ) Figure 7. Power analysis indicating for New Caledonia and Tonga the expected variations in predicted values of total SR from butterflyfish SR for increasing probability levels (probability of rejecting H that the confidence interval based on the predicted value contains the true value). The y-axis represents the width of the confidence interval for the predicted values for the mean observed butterflyfish SR. Top graphs, indicate results as number of species; the lower graphs indicate results as a percentage of the mean total SR. discussion focuses on the quality and limits of butterflyfishes as such proxies, recalls the precautions necessary in using butterflyfishes for this purpose, and examines some potential applications. Large-scale factors Total reef fish SR is highly correlated to butterflyfish, however, this correlation varies according to many factors, in particular, region, island, and reef type. This strongly suggests that these factors should be taken into account for any use of this relationship. In particular, it will be especially important not to compare heterogeneous areas using the same relationship as this could lead to serious errors. To illustrate this, a new observation of seven butterflyfish species on a transect results in a prediction of 48 reef fish species on that transect for the fringing reefs of New Caledonia, but 6 species on the barrier reefs of the same island. Another important point is that this relationship will be best used in areas where the observed number of butterflyfish species is high, since it is at the highest densities that the power of this relationship is best. Also, regional diversity is not necessarily a good predictor of local diversity, as indicated by Findley and Findley Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

12 S138 M. KULBICKI AND Y.M. BOZEC (1) and by Kulbicki et al. (5), therefore countries with high regional diversities of butterflyfishes such as Indonesia, the Philippines (Findley and Findley, 1) or New Caledonia (Kulbicki et al., 5) may not have the highest observed local butterflyfish diversities, and therefore the relationship between total SR and butterflyfish SR in these places may have a lower power than would may be expected by just considering regional diversity. Similarly, within a country, this relationship will be best used in biotopes with high butterflyfish diversities, e.g. on barrier reefs rather than on fringing reefs. It is notable that some factors are influential but not others. Regional differences could stem from the relationships existing between meta-communities and local communities (Hillebrand and Blenckner, 2), the SR of local communities usually being well correlated at the level of the meta-community. Table 1 and Kulbicki et al. (5) indicate that the relative importance of butterflyfishes to the regional metacommunity varies according to the region. It would be interesting to test whether the effect of the region on the total SR butterflyfish SR relationship could be predicted to some extent from the number of species found within the region. It could be expected that the larger the relative contribution of butterflyfishes to regional diversity, the steeper will be the slope of the relationship between total SR and butterflyfish SR, as confirmed by this present study, with Tuamotu having the steepest slope and New Caledonia the flattest (Table 4). Additional regions, however, would ideally need to be tested to validate such an hypothesis. Differences between islands may be expected as the diversity of the reef fish meta-community at the island level is a function of factors such as island size, island type or the connectivity with other islands (Bellwood and Hughes, 1; Kulbicki et al., 4). The role of these factors on the relationship between total SR and butterflyfish SR could be tested and even probably be predicted from a statistical model; however, as for the effect of regions, this would require more islands than have been tested so far. The effect of reef type was significant in the two cases where it could be tested formally, New Caledonia and Tonga (Table 3). Fringing reefs offered steeper slopes, lower intercepts and stronger correlations than barrier reefs. In both cases (Tonga and New Caledonia), the number of known species was higher on barrier reefs than fringing ones. Therefore, flatter slopes and lower intercepts on fringing reefs would be expected. As only the intercept agrees with what is expected, other factors have to come into play in this relationship. One possible explanation for the steeper-than-expected slope and the stronger correlation on fringing reefs could be a higher habitat diversity or heterogeneity on fringing reefs than on barrier reefs (this is at present being tested by satellite image analysis). Coral cover and time In contrast, two factors, time and coral cover were not found to be significant in the relationship between total SR and butterflyfish SR. This suggests that comparisons of within-an-area transects with different coral covers are valid, and since this relationship seems stable through time (at least at the scale tested), it could be of some assistance in programmes intended for monitoring changes of species diversity of reef fish communities. Most studies show that coral is a significant factor for both total SR and butterflyfish SR (Reese, 181; Harmelin-Vivien and Bouchon-Naarro 183; Bell and Galzin, 184; Bouchon-Navaro and Bouchon, 18; Chabanet et al., 17; Cadoret et al., 1). One would expect butterflyfishes to be relatively more diverse in coral rich areas, as in the Pacific a large number of them are coral feeders (Kulbicki et al., 5) whereas only a low number of other reef fish species are directly linked to coral for food. Consequently, coral cover should influence the ratio between butterflyfish SR and total SR. The lack of effect, therefore, of coral cover on the relationship between total SR and butterflyfish SR seems at first surprising (Table 3). A potential explanation is that coral cover plays globally a similar role for butterflyfish species and other species and is probably more important as providing shelter than as a direct or indirect food source. There is no direct evidence for this, but Jones et al. (4) demonstrated that coral-associated and non-associated fish changed in a similar manner in reef areas where coral cover was affected. Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

13 BUTTERFLYFISH SPECIES RICHNESS IN WESTERN AND CENTRAL PACIFIC S13 The lack of effect of time on the relationship between total SR and butterflyfish SR suggests that over the timescale of the study (3 months), the number of butterflyfish species and total reef species varied in the same way with either both butterflyfishes and total reef fish following similar pulses in SR (except for trimester 7), or neither of them changing during this time period. This means also that the relative detectability of butterflyfish species and other species on these reefs remained the same. This is important, as many factors play on fish detectability (Kulbicki, 18), in particular turbidity. This latter factor changed, often drastically, from one sampling to the next owing to storms, calm weather or nutrient inputs (and consequent algae development). This lack of time effect suggests that the relationship between total SR and butterflyfish SR is very robust. Applications The potential uses of this relationship between total SR and butterflyfish SR are numerous. Being able to recognize with a high degree of confidence all the fish species visible on a reef requires long training, as the number of species involved is usually very high (often several hundred on a single reef). In contrast, recognizing and counting butterflyfishes can be learned in a relatively short time. Therefore, the use of butterflyfishes for getting a proxy of the SR of the total fish assemblage could be very important for a number of monitoring programmes, in particular those which rely on volunteers or personnel who only dive occasionally. Total SR is useful to numerous ecological models as it is an important component of ecological parameters such as stability, resilience and resistance of communities to perturbations (Peterson et al., 18). Therefore being able to get some estimate of this total SR, or at least some relative measure, can be very valuable. The power analysis indicates, however, that even under the best conditions the total SR estimates obtained from butterflyfish SR should be considered more as an indicator than a precise measure of total SR. Thus, if an acceptable confidence interval on predictive values of total SR is one of less than % of the estimated total SR (i.e. if the estimated total SR was species, the confidence interval would cover from 48 to 72 species), the best probability of achieving this in the present study is.25 (i.e. 25% chance that the confidence interval does not include the true value of total SR). Therefore unless the correlation between total SR and butterflyfish SR is exceptional, it would be unwise to expect estimates of total SR which could be used directly in ecological models. The usefulness of this relationship is probably highest in studies where relative values are sufficient. Among the possible applications in this perspective, this technique may increase the power of spatial or temporal surveys. Reef fishes have very patchy distributions and diversity may change rather drastically within a reef over short distances. Counting all the fish species requires not only a high degree of expertise but is also time-consuming and only a restricted number of stations can usually be sampled. Butterflyfish counts are easy and fast and can be done by people with little training, thus allowing much higher spatial (or temporal) coverage during a survey. The values of these butterflyfish counts could be inserted into models as complementary data to complete counts and thus increase at little cost the power of the models. Another possible use is in looking at butterflyfish historical data; there are many monitoring programmes that have incidentally accumulated butterflyfish data. Once the relationship between total SR and butterflyfish SR for an area is established, these data could be useful to give a general indication of the spatial or temporal trends in total SR. Ecological and biogeographical applications of this type of datamining exercise could also be worth considering. The impact of fishing on the relationship between total SR and butterflyfish SR needs to be considered. In most of their geographical range, butterflyfishes are not targeted by fishermen; there are, however, exceptions, in particular in southeast Asia where they are caught both for food and for the aquarium fish trade. As long as fishing habits remain the same within any area, there are few reasons for this relationship to change. The introduction of new fishing practices or a drastic increase in fishing effort with a shift of target species, however, would certainly result in a change of this relationship. The same would probably be Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

14 S1 M. KULBICKI AND Y.M. BOZEC true for other types of perturbations such as pollution. This may narrow the range of uses of this relationship and accentuates the need for good baseline studies. In the present study, the total SR was taken to be all fish species that could be visually identified along the transect. For most studies and monitoring programmes which use total species counts, this total is represented by a restricted list of families and genera. This approach is dictated by the fact that many species are very difficult to count, usually because of their behaviour (nocturnal, pelagic or cryptic species in particular). This should be kept in mind if in the future one wishes to use total SR to butterflyfish SR relationships from different data sources. The concept of using butterflyfishes to estimate other parameters of the reef fish assemblage needs to be refined and further tested. It may also be the source of other applications. In particular, analyses currently in progress suggest that much better correlations can be found if, instead of total SR, the relationship between butterflyfish SR and some specific groups of fish, such as territorial and sedentary species, or herbivorous ones, is used. Similarly, taking into account the behaviour, size or diet of butterflyfishes may assist in improving the predictions of total SR. The combined use of butterflyfish SR and abundance is also promising for proxies of total reef fish density and biomass estimates. REFERENCES Bell JD, Galzin R Influence of live coral cover on coral-reef fish communities. Marine Ecology Progress Series 15: Bell J, Harmelin-Vivien M, Galzin R Large scale spatial variation in abundance of butterflyfishes (Chaetodontidae) on Polynesian reefs. Proceedings of the Fifth International Coral Reef Symposium, Tahiti; Vol. 5: Bellwood DR, Hughes TP. 1. Regional-scale assembly rules and biodiversity of coral reefs. Science 22: Blum SD. 18. Biogeography of the Chaetodontidae: an analysis of allopatry among closely related species. Environmental Biology of Fishes 25(1 3): 31. Bouchon-Navaro Y Quantitative distribution of the Chaetodontidae on a reef of Moorea island (French Polynesia). Journal of Experimental Marine Biology and Ecology 55: Bouchon-Navaro Y Partitioning of food and space resources by chaetodontid fishes on coral reefs. Journal of Experimental Marine Biology and Ecology 3: 21. Bouchon Navaro Y, Bouchon C. 18. Correlations between chaetodontid fishes and coral communities of the Gulf of Aqaba (Red Sea). Environmental Biology of Fishes 25: 47. Bouchon Navaro Y, Bouchon C, Harmelin-Vivien M Impact of coral degradation on a Chaetodontid fish assemblage (Moorea, French Polynesia). Proceedings of the Fifth International Coral Reef Symposium, Tahiti; Vol. 5: Cadoret L, Legendre P, Adjeroud M, Galzin R. 15. Répartition spatiale des Chaetodontidae dans diffe rents secteurs re cifaux de l ıˆ le de Moorea, Polynésie Franc-aise. Ecoscience 2(2): Cadoret L, Adjeroud M, Tsuchiya M. 1. Spatial distribution of Chaetodontidae fish on coral reefs of the Ryukyu Islands, southern Japan. Journal of the Marine Biological Association of UK 7: Chabanet P, Ralambondrainy H, Amanieu M, Faure G, Galzin R. 17. Relationships between coral reef substrata and fish. Coral Reefs 16: 3 2. Cox EF. 14. Resource use by corallivorous butterflyfishes in Hawaii. Bulletin of Marine Sciences 54(2): Crosby MP, Reese ES. 16. A manual for monitoring coral reefs with indicator species: butterflyfishes as indicators of change on Indo-Pacific reefs. Office of Ocean and Coastal Resource Management, National Oceanic and Atmospheric Administration. Silver Spring, MD; 45 pp. Crosby MP, Reese ES. 5. Relationship of habitat stability and intra-specific population dynamics of an obligate corallivore butterflyfish. Aquatic Conservation: Marine and Freshwater Ecosystems 15(S1): S13 S25. Driscoll JW, Driscoll JL Pair behaviour and spacing in butterflyfishes (Chaetodontidae). Environmental Biology of Fishes 22(1): Erdman M. 17. Butterflyfish as bioindicators } a review. Reef Encounter 21: 7. Findley JS, Findley MT. 18. Circumtropical patterns in butterflyfish communities. Environmental Biology of Fishes 25(1 3): Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)

15 BUTTERFLYFISH SPECIES RICHNESS IN WESTERN AND CENTRAL PACIFIC S141 Findley JS, Findley MT. 1. Global, regional, and local patterns in species richness and abundance of butterflyfishes. Ecological Monographs 71(1): 6 1. Fowler AJ. 1. Spatial and temporal patterns of distribution and abundance of chaetodontid fishes at one tree reef, southern GBR. Marine Ecology Progress Series 64: Harmelin-Vivien M. 18. Implications of feeding specialisation on the recruitment processes and community structure of butterflyfishes. Environmental Biology of Fishes 25(1 3): 1 1. Harmelin-Vivien M, Bouchon-Navaro Y Feeding diets and significance of coral feeding among Chaetodontid fishes in Moorea (French Polynesia). Coral Reefs 2: Hillebrand H, Blenckner T. 2. Regional and local impact on species diversity } from pattern to processes. Oecologia 132: Hourigan TF, Tricas TC, Reese ES Coral reef fishes as indicators of environmental stress in coral reefs. In Marine Organisms as Indicators, Soule DF, Kleppel GS (eds). Springer-Verlag: New York; Jennings S, Boulle DP, Polunin NVC. 16. Habitat correlates of the distribution and biomass of Seychelles reef fishes. Environmental Biology of Fishes 46: Jones GP, McCormick MI, Srinivasan M, Eagle JV. 4. Coral decline threatens fish biodiversity in marine reserves. Proceedings of the National Academy of Sciences of the USA 1(21): Khalaf M, Crosby MP. 5. Assemblage structure of butterflyfishes and their use as indicators of Gulf of Aqaba benthic habitat in Jordan. Aquatic Conservation: Marine and Freshwater Ecosystems 15(S1): S27 S43. Kulbicki M. 18. How acquired behaviour of commercial reef fish may influence results obtained from visual censuses. Journal of Experimental Marine Biology and Ecology 222: Kulbicki M, Bozec YM, Green A. 5. Implications of biogeography in the use of butterflyfishes (Chaetodontidae) as indicators for Western and Central Pacific areas. Aquatic Conservation: Marine and Freshwater Ecosystems 15(S1): S S126. Kulbicki M, Labrosse P, Ferraris J. 4. Basic principles underlying research projects on the links between the ecology and the uses of coral reef fishes in the Pacific. In Challenging Coasts. Transdisciplinary Excursions into Integrated Coastal Zone Development, Visser LE (ed.). Amsterdam University Press: Amsterdam; Lewis AR. 18. Effects of experimental coral disturbance on the population dynamics of fishes on large patch reefs. Journal of Experimental Marine Biology and Ecology 23: 1 1. Motta PJ. 18. Dentition patterns among Pacific and Western Atlantic butterflyfishes: relationship to feeding ecology and evolutionary history. Environmental Biology of Fishes 25(1 3): O hman MC, Rajasuriya A, Svensson S. 18. The use of butterflyfishes (Chaetodontidae) as bio-indicators of habitat structure and human disturbance. Ambio 27: Peterson G, Allen CR, Holling CS. 18. Ecological resilience, biodiversity and scale. Ecosystems 1: Reese ES Coevolution of corals and coral feeding fishes of the family Chaetodontidae. Proceedings of the Third International Coral Reef Symposium, Miami; Vol. 1: Reese E Predation on corals by fishes of the family Chaetodontidae: implications for conservation and management of coral reef ecosystems. Bulletin of Marine Science 31(3): Roberts CM, Ormond RFG. 12. Butterflyfish social behaviour, with special reference to the incidence of territoriality: a review. Environmental Biology of Fishes 34: 7 3. Roberts CM, Ormond RFG, Shepherd ARD The usefulness of butterflyfishes as environmental indicators on coral reefs. Proceedings of the Sixth International Coral Reef Symposium, Townsville; Vol. 2: Roberts CM, Shepherd ARD, Ormond RFG. 12. Large-scale variation in assemblage structure of Red Sea butterflyfishes and angelfishes. Journal of Biogeography 1: Samways MJ. 5. Breakdown of butterflyfish (Chaetodontidae) territories associated with the onset of a mass coral bleaching event. Aquatic Conservation: Marine and Freshwater Ecosystems 15(S1): S1 S7. Sano M, Shimizu M, Nose Y Changes in structure of coral reef fish communities by destruction of hernatypic corals: observational and experimental views. Pacific Science 38(1): Sano M, Shimizu M, Nose Y Long-term effects of destruction of hermatypic corals by Acanthaster planci infestation on reef fish communities at Iriomote island, Japan. Marine Ecology Progress Series 37: Temraz TA, Abou Zaid MM. 5. Distribution of butterflyfishes (Chaetodontidae) along the Egyptian Red Sea coast and its relation to coral health. Aquatic Conservation: Marine and Freshwater Ecosystems 15(S1): S5 S7. White AT Chaetodon occurrence relative to coral reef habitats in the Philippines with implications for reef management. Proceedings of the Sixth International Coral Reef Symposium, Townsville; Vol. 2: Zekeria ZA. 3. Butterflyfishes of the Southern Red Sea: ecology and population dynamics. PhD thesis, Groningen (NL); 1 pp. Copyright # 5 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. 15: S127 S141 (5)