Inter-habitat variation in density and size composition of reef fishes from the Cuban Northwestern shelf

Movement and exchange of individuals among habitats is critical for the dynamics and success of reef fish populations. Size segregation among habitats could be taken as evidence for habitat connectivity, and this would be a first step to formulate hypotheses about ontogenetic inter-habitat migrations. The primary goal of our research was to find evidence of inter-habitat differences in size distributions and density of reef fish species that can be classified a priori as habitat-shifters in an extensive (~600km2) Caribbean shelf area in NW Cuba. We sampled the fish assemblage of selected species using visual census (stationary and transect methods) in 20 stations (sites) located in mangrove roots, patch reefs, inner zone of the crest and fore reef (12-16m depth). In each site, we performed ten censuses for every habitat type in June and September 2009. A total of 11 507 individuals of 34 species were counted in a total of 400 censuses. We found significant differences in densities and size compositions among reef and mangrove habitats, supporting the species-specific use of coastal habitats. Adults were found in all habitats. Reef habitats, mainly patch reefs, seem to be most important for juvenile fish of most species. Mangroves were especially important for two species of snappers (Lutjanus apodus and L. griseus), providing habitat for juveniles. These species also displayed well defined gradients in length composition across the shelf. Rev. Biol. Trop. 62 (2): 589-602. Epub 2014 June 01.

Movement and exchange of individuals among habitats is critical for the dynamics and success of reef fish populations.Many fish species move among different coastal habitats created by mangroves, seagrass beds, and coral reefs during their life cycles (Grol, Nagelkerken, Rypel & Layman, 2011).Consequently, conservation efforts to protect species and biodiversity are developing to provide protection of habitats including linking corridors for all life stages (Mumby et al., 2004;Sale et al., 2005).Recent research indicates that more consideration should be given to the ecological processes that occur along nursery-reef boundaries that connect neighboring ecosystems (Nagelkerken, Grol & Mumby, 2012).
In a recent review, Adams et al. (2006) classified coral reef fishes based on their ontogenic migration patterns.The authors define three groups: Group A) habitat specialists that use the same habitat at all life stages, Group B) habitat generalists which are not site-attached and use a variety of habitats, and Group C) ontogenetic shifters.The latter species switch habitats during their life time, such as the transition from larval to juvenile to maturing adults.These species in particular depend on ontogenetic habitat change from the backreef to the fore-reef (Adams & Ebersole, 2009).
The nursery function of back-reef habitats has been inferred by studying spatial and temporal patterns in the size distribution of juveniles and adults across such habitats (Mumby et al., 2004;verweij, Nagelkerken, Wartenbergh, Pen & van der velde, 2006;Dorenbosch, verbeck, Nagelkerken & van der velde, 2007).Size segregation among habitats could be taken as evidence for habitat connectivity and this would be a first step to formulate hypotheses about ontogenetic inter-habitat migrations.
The goals of these studies were to identify natural and anthropogenic factors that may influence fish assemblage structure.All of these and other studies across the Caribbean (Nagelkerken & van der velde, 2002;Beets et al., 2003;Aguilar-Perera & Appeldoorn, 2007) suggest a degree of habitat connectivity, but none was designed to specifically test the hypothesis of ontogenetic, habitat migrations, which probably occur across larger spatial scales of >100km 2 .
The primary goal of our research was to find evidence for habitat-related differences in size composition of a group of selected fish species considered as habitat shifters at the NW Cuban shelf.The coast in this region has a variable shelf width but a general profile from mangrove, seagrass to coral reef is found, which makes it an ideal area to study ontogenetic habitat shifts of reef fishes.We hypothesized that ontogenetic migrations occur, and then tested the predictions that inter-habitat differences in size distribution and density of reef fish species would be apparent.The specific predictions were: (1) fish density and size composition for selected species among habitats have significant differences; and (2) a gradient in length composition across the island shelf from mangrove to fore reef is found.

Study area:
We included the spatial coverage assumed to be the normal range of reef fish species (e.g.Nemeth, Blondeau, Hezlieb & Kadison, 2007;Pina-Amargós et al., 2008), we studied an extensive area of the Caribbean shelf in Northwestern Cuba (Fig. 1).The shelf in the study area is broad with a distance from shore to the shelf-edge (200m isobath) ranging from 22 to 36km.Total surface of the study area was 584km 2 .Mangroves were present along the shore occupying an almost continuous fringe of ~26km (excluding the coastal indentations).Red mangrove (Rhizophora mangle) was dominant and its roots provide a shallow (<1m deep) habitat for fishes.The width of the aquatic habitat associated with mangrove roots was narrow (<5m) due to the small tidal range (<0.5m) in the area (Claro, Lindeman & Parenti, 2002).There is a bank-barrier reef near the edge of the shelf.Corals occupied an almost continuous and shallow (≤5m deep) reef crest 21km long and 2-5km wide.Dominant coral species were Porites astreoides, Millepora complanata and Acropora palmata.The fore-reef was narrow ranging from 500-900m from the crest to the 50m isobath.Coral diversity was relatively high for a Caribbean reef and dominant species were Siderastrea spp., Porites astreoides, Montastrea spp.and Millepora alcicornis (González-Díaz, González-Sansón, Alvarez & Perera, 2010).There is a wide back-reef lagoon between the crest and the mangrove fringe.The depth of the lagoon varied from 6-8m near the crest and mangrove to 15-20m in the central portion.At this deeper zone, there were many patch reefs which rose from the bottom 14-18m and 6-9m from the surface.The remaining lagoons were covered by sand and extensive patches of turtle grass (Thalassia testudinum).

Sampling methods:
We sampled four habitat types: mangrove (MA), patch reefs (PR), inner zone of the crest (IC) and fore-reef (FR).We located five sampling stations within each habitat type (Fig. 1, Table 1).Stationary visual censuses (Bohnsack & Bannerot, 1986) were conducted by two scuba divers at the PR, IC and FR habitats.Instead of 7m as recommended in the method, we used a radius of 5m (area of sampling unit ~79m 2 ) to address any potential visibility variability among sites.At mangrove roots, we used 40m long transects (measured with a rope) for visual censuses by two snorkelers.The average distance of observation within the root habitat was 2m from the edge of the mangrove roots towards land, for a total area of 80m 2 , which was equivalent to that of the stationary visual censuses.Ten censuses (five per each diver) were done at each station for each habitat type.This sampling protocol was repeated twice over one week in June and September 2009.
For the census, we targeted a priori a selected group of reef fish species considered habitat-shifters (sensu Adams et al., 2006) by several authors (Cocheret de la Moriniere, Pollux, Nagelkerken & van der velde, 2002;Dorenbosch et al., 2007).These species were from the families: Serranidae (genera Cephalopholis, Epinephelus and Mycteroperca), Lutjanidae, Haemulidae, Chaetodontidae, Scaridae and Acanthuridae, and Lachnolaimus maximus (Labridae) and Sphyraena barracuda (Sphyraenidae).The individuals of Scarus iseri and S. taeniopterus were usually indistinguishable in the field and we hereafter refer to them as S. iseri/taeniopterus.Following the approach of Nagelkerken, van der velde, Gorissen, Meijer, van't Hof & den Hartog (2000), we used 5cm bins to estimate fish body length to reduce differences in size estimation between observers.We used the criteria of Nagelkerken & van der velde (2002) to classify the sampled fishes as juvenile or adults.
For the inter-habitat comparisons in our study area, we did not transform the data to density in terms of number of individuals per unit area.Instead, we averaged the values of the ten censuses made at each station, and used this average as a density indicator expressed as the number of fish per 80m 2 (Bohnsack & Bannerot , 1986).It is a relative estimate of abundance that allowed for comparisons within this study.A hierarchical agglomerative cluster analysis was performed using the Bray-Curtis index as the similarity measure and clustering based on group average (UPGMA).SIMPROF test (Clarke, Somerfield & Gorley, 2008) was used for identifying significant clusters.Nonparametric multidimensional scaling (MDS) was employed for ordination of samples based in the same distance matrix as the cluster analysis.The combination of clustering and ordination analyses has been described by Clarke & Gorley (2006) as the most effective way to check the adequacy and mutual consistency of both representations.Contribution of each species to total dissimilarity between pairs of habitats was calculated using the SIMPER routine.Only species adding to at least 90% dissimilarity were retained for comparative analyses.A two-way crossed ANOSIM based on a priori classification of samples (habitats x date) using sites inside habitats as replicates was performed on Bray-Curtis dissimilarities calculated between pairs of species.All analyses were performed on log-transformed number of individuals per count using PRIMER 6.0 (Clarke & Gorley, 2006).
A two-way fixed effects factorial design ANOvA, using date of sampling and habitat as factors, was performed on the average number/80m 2 (fourth-root transformed) for each of the most abundant species separately.A Student-Newman-Keuls post-hoc test was used for the pairwise comparisons of averages.Correlation between sampling dates for the percentage of juvenile fish was tested using the Spearman's rank correlation coefficient.Significant correlation indicates that date of sampling had no effect on juvenile abundance.A Kruskal-Wallis test was used to test the significance of observed differences among habitats for the median juvenile percentage.A Post-hoc pairwise test using z-transformation was applied to discriminate differences between pairs of habitats.All tests were performed with a level of significance equal to 0.05 using STATISTICA 7.0 (Statsoft, 2006).

RESULTS
A total of 11 507 individuals of 34 species were counted in 400 censuses (Table 2).The ten most abundant species were Haemulon aurolineatum, S. iseri/taeniopterus, L. griseus, Ocyurus chrysurus, Sparisoma aurofrenatum, Haemulon plumieri, Sparisoma viride, L. apodus, Acanthurus coeruleus, and A. bahianus.These species accounted for >90% of individuals per census.Nine species occurred in at least two habitats (with averages >1/80m 2 ).H. aurolineatum was present in large numbers at a few sites of the PR habitat resulting in a high aggregation (61.5±29.5).The remaining species had lower densities and thus we could not perform reliable single-species quantitative analyses.We used the total assemblage data for exploratory multivariate analyses and analyses by species restricted to the nine species present in at least two habitats.
The assemblage composition varied notably among habitats (ANOSIM´s global R=0.752, p<0.001, 999 permutations), but was similar among sampling periods per habitat (global R=0.08, p=0.148, 999 permutations).All pairwise comparisons between habitats were also significant (p<0.005).The numerical classification and multidimensional scaling yielded diagrams those were highly consistent with ANOSIM results.Samples formed two well separated groups (SIMPROF Pi=16.36,p=0.001;Fig. 2).One group (A) included all reef samples and the other group (B) included all mangrove samples.Within the reef habitats, two subgroups were significantly separated (Pi=3.69,p=0.001): the A1 including patch reefs and the A2 fore-reef plus inner crest habitats.Mangrove stations showed a greater heterogeneity and split into four subgroups (B1-B2, B3, B4: Pi=2.83, p=0.03;B2-B3, B4: Pi=2.82, p=0.04 and B3-B4: Pi=3.96, p=0.024).Group B2 included mostly June, and group B4 included mostly September stations.Group B1 and group B3 included only one station in June and September, respectively.The nine species used for two factorial analyses made an important contribution to the dissimilarity among habitats (Table 3).Other species were also important because they were abundant in one habitat (H.aurolineatum) or were observed in just one habitat (Cephalopholis fulva in FR).Some rare species also made smaller contributions (>4%) to total dissimilarities (e.g., Cephalopholis cruentata, Epinephelus guttatus, Haemulon flavolineatum, and Sparisoma chrysopterum) in reef habitats but not mangrove roots (<1/80m 2 ).Haemulon sciurus was present in all habitats with a low average density (<1/80m 2 ) but densities in patch reefs and mangroves were more than twice the inner crest and fore reef habitats.
No significant differences in relative abundance were found between sampling periods for any species.Significant differences in fish densities were found between habitats in all cases (Table 4; Fig. 3).L. apodus showed significantly higher density at mangrove and patch reefs.Density of L. griseus in mangrove roots was the highest, followed by a lower value in patch reefs, but this species was not recorded in other habitats.O. chrysurus was not recorded in mangrove habitats and had higher densities in patch reefs with a progressive decrease towards the fore reef.H. plumieri, S. taeniopterus/iseri, S. aurofrenatum, S. viride and A. coeruleus showed no significant differences in density among reef habitats, and were mostly absent in mangrove roots.Density of A. bahianus was higher in fore reef and inner crest habitats, lower in patch reefs, and observed the lowest values in mangrove roots.
Juvenile proportions changed among habitats and species, but showed less variability between sampling dates (Table 5).Based in our results, we grouped juvenile species in three categories: a) species that were not present in mangrove roots; b) species present in all sampled habitats; and c) species that were more abundant in mangrove roots.For the three species that were not present in mangrove roots habitats, no clear trend in size composition was found.For A. coeruleus, the percentage of juvenile individuals was high (45-71%) in fore reef and inner crest habitats; while O. chrysurus was found in most habitats (62-94%).H. plumieri were mostly adults (68-88%) in all habitats.Additionally, four species (S. taeniopterus/ iseri, S. viride, S. aurofrenatum and Acanthurus bahianus) were present in all habitats but were scarce in mangrove roots.No clear trend was evident in three species of parrotfishes (S. taeniopterus/iseri, S. viride and S. aurofrenatum) and most of these species were juveniles (40-82%).A. bahianus individuals were larger in patch reefs with a low representation of juvenile fish (10-25%).In the rest of the habitats,  however, the percentages of juveniles for this species were greater (53-100%).L. apodus and L. griseus were the only species that were abundant in mangrove root habitats as juveniles, yet they were larger in body size at reef habitats.For this reason, we presented a detailed analysis of size distributions by habitat for these two species (Fig. 4).Juveniles of L. apodus were present in all habitats and their proportion decreased from mangrove to fore reef habitats (mangrove, 100%; patch, 40-48%; inner crest, 13-57%; fore reef, 0-31%).Individuals of L. griseus were almost only juveniles in mangrove root habitats (97-100%), while juvenile fish decreased notably in patch reefs (53-57%).This species was not observed in inner crest and fore reef stations.
The estimated percentage of juvenile fish per species per habitat showed a significant correlation between sampling dates (r s =0.787, p<0.001, n=30).Kruskal-Wallis test pooling all nine species data yielded a significant value indicating differences among habitats (H=18.04,p=0004, n=60).After a post-hoc pairwise comparison of the medians, mangrove habitat values were significantly higher compared to other habitats (not significantly different, Table 5).

DISCUSSION
The presence of adult fish in all sampled habitats was consistent with Sheaves (2005) and Chittaro et al. (2005) conclusions that the term 'nursery area' applied for mangroves and seagrass habitats should be used with discretion.Based in our results, we consider the term 'adult habitat' applied for coral reefs should be used also with discretion.As Gillanders, Able, Brown, Eggleston & Sheridan (2003) discuss, it is important to measure the life history and specific movements of individuals, if we were to better understand the location and function of nursery habitats for marine organisms.Supporting former ideas, we found adults of scarid species in mangrove roots, and juvenile of all most abundant fish species in reef habitats.
Our data did not allow a quantitative approach to define nursery areas (sensu Beck et al., 2001) or essential juvenile habitats (Dahlgren et al., 2006;Layman et al, 2006).Instead, we discuss juvenile habitats as a qualitative labeling for those places where a relative high percent (>40%) of individuals were classified as juveniles.Generally, juveniles were relatively abundant in reef habitats and most abundant in mangrove roots.Nagelkerken et al. (2000) showed that fish had a preference for specific nursery habitats, but that most species used multiple nursery habitats simultaneously.The value of nursery habitats is species specific and a combination of habitats is used by high densities of juveniles (Aguilar-Perera, 2004).Our results for A. bahianus and A. coeruleus corroborate earlier findings.Nagelkerken et al. (2000) found that juveniles of Acanthuridae were restricted to the shallow water biotopes (mangrove, notches in fossil reef rock, seagrass bed and fossil reef boulders), whereas adults were found on the reef.In contrast, Cocheret de la Moriniere et al. (2002) reported that A. bahianus was not observed in mangroves.Aguilar-Perera & Appeldorn (2007) found a similar trend for A. coeruleus and A. bahianus which were almost absent from mangroves in their study.
Habitat and size distribution patterns of Lutjanidae in our study agree with those reported by other authors.Nagelkerken et al. (2000)  found that L. apodus and L. griseus were more abundant in mangrove habitats and that L. apodus showed an ontogenetic shift to deeper coral reefs.Cocheret de la Moriniere et al. (2002), Chittaro et al. (2005), and Aguilar-Perera & Appeldoorn (2007) found that L. apodus and L. griseus showed strong preferences for mangroves over seagrass beds at all size-classes.
Juvenile yellowtail snapper (O.chrysurus) prefer seagrass over mangrove roots and other habitats (Nagelkerken et al., 2000;Aguilar-Perera & Appeldoorn, 2007), but they have been recorded in both habitats (Cocheret de la Moriniere et al., 2002).Huijbers, Nagelkerken, Debrot & Jongejans (2013) found that the contribution of the coral reef as a nursery habitat for O. chrysurus was minimal.We found relatively high abundances of juveniles of this species in all reef habitats except mangrove roots.Mumby et al. (2004) reported that juveniles of the white grunt, H. plumieri, were common in mangrove habitats.Appeldoorn, Recksiek, Hill, Pagan & Dennis (1997) presented length-frequency distributions that indicated an offshore ontogenetic migration.Along the Northwest coast of Cuba, we observed no white grunt in mangrove roots, and the length composition in reef habitats did not suggest an ontogenetic migration, because proportions of juveniles were similar in all reef habitats.Aguilar-Perera & Appeldoorn (2007) reported that mean densities of juveniles were significantly higher in seagrass relative to mangroves and coral reefs.
Our results for species in the family Scaridae indicated that these species were found at all stages in all reef habitats.Aguilar-Perera & Appeldoorn (2007) found significant differences among habitats for S. aurofrenatum and S. viride juveniles, but not for S. iseri juveniles.The preference of S. aurofrenatum and S. iseri for reef habitats has also been reported by others (Adams & Ebersole, 2002;Chittaro et al., 2005).
In summary, we found significant differences in densities and size compositions of fishes among reef and mangrove habitats of Northwestern Cuba.These suggest a species-specific, differential use of coastal habitats during ontogeny.Adults were found in all habitats.Reef habitats, mainly patch reefs, seem to be very important for juvenile fish of most species in our study area.These results differ notably from the main trend found in the Caribbean, where studies report that most or all nursery fish species studied herein had the highest juvenile densities in mangrove areas only (Nagelkerken, 2009).In our study area, mangroves were especially important for the two snapper species (L.apodus and L. griseus) which were also the only species that showed well defined gradients in length composition from mangrove to the fore reef.

Fig. 1 .
Fig. 1.Study area.Locations of sampling stations are black and white circles.MA: mangrove, PR: patch reefs, IC: inner crest, FR: Fore reef.Numbers indicate replicates within each habitat.

Fig. 2 .
Fig. 2. Numerical classification (top panel) and MDS diagram (bottom panel) of samples.Numbers identify replicates.Envelopes indicate groups and subgroups of cluster analysis.

Fig. 4 .
Fig. 4. Length composition of fish from the mangrove habitat for two species of Lutjanids.ML=Length at maturation.MA: mangrove, PR: patch reefs, IC: inner crest, FR: Fore reef.

TABLE 1
Location, habitat type, and depth of stations sampled along the Northwestern coast of Cuba in 2009 (Figure 1)

TABLE 2
Fish species observed and their average density (D)±1 standard error (number/80m 2 ) in four habitat types of Northwestern Cuba in summer and fall, 2009

TABLE 3
The overall average dissimilarity (AD) and single species contributions (%) to total dissimilarity between habitat pairs in four habitat types of Northwestern Cuba in summer and fall, 2009 FR: fore reef, IC: inner crest, PR: patch reefs, MA: mangrove roots.The cut off for low contributions was 90%.Species codes are presented in table 2. 1 Species was not found or density was below the cutoff %.

TABLE 4
Two-factor ANOvA results for the most abundant species in NW Cuban shelf.Species codes are presented in table2 Fig. 3. Average number of individuals/80m 2 (±1 standard error) for selected species.Similar letters are not significantly different (SNK test).

TABLE 5
Percentage of juvenile individuals based on body length<reported maturation length (ML) in four habitat types of Northwestern Cuba in summer and fall, 2009.Species codes are presented in table 2 1Species not present.