Complex reef architecture supports more small-bodied fishes and longer food chains on 1 Caribbean reefs

Coral community shifts towards reefs dominated by stress-resistant corals have contributed to rapid declines in the architectural complexity of reefs throughout the Caribbean. Complex reef architecture provides important niches and refuges for many reef fishes and thus widespread declines in reef complexity could have important consequences for the structure and function of fish assemblages. We explore the influence of reef architecture on fish assemblages by comparing the size and trophic structure of reef fishes along a 20 km-long 15-reef gradient of coral cover, coral species dominance and architectural complexity in Cozumel, Mexico. Our results show that reefs with high architectural complexity, in particular those dominated by robust Montastraea corals, supported fish assemblages with larger numbers of individuals in the smallest size classes (<20 cm) and longer food chains (higher mean trophic levels). The association between coral complexity and fish communities is highly size-structured and is greatest for smallest size classes. The greater abundance of both small fish and the key early life stages of larger fishes on more complex reefs suggests that architectural complexity may influence entire reef fish assemblages, even though larger fish are less dependent on reef complexity. Key reef-building corals such as Montastraea are thus likely to be disproportionately important for maintaining reef fish communities, and shifts in Caribbean coral communities may compromise fish recruitment and truncate food chains, reducing resilience and inhibiting reef recovery from degradation.


INTRODUCTION
Habitat architecture is known to profoundly influence the abundance and diversity of species through the provision of niches and diverse ways of exploiting environmental resources (e.g.MacArthur andMacArthur 1961, Tews et al. 2004).In many ecosystems, the architecture of the habitat is defined by the abundance and morphological attributes of certain foundation species.Many of these species, such as numerous overstorey trees, reef-building corals and kelp, have experienced or are experiencing widespread declines (Steneck et al. 2002, Gardner et al. 2003, Ellison et al. 2005).These declines commonly result in dominance shifts towards more stress-resistant species that often have rapid growth rates and are likely to provide fewer structural benefits (Ellison et al. 2005, Steneck et al. 2009).Understanding the influence of changes in the relative abundance of different foundation species on associated biodiversity can therefore provide valuable insights into the potential consequences of changing habitat structure resulting from environmental disturbances.
In tropical coral reef ecosystems, the architectural complexity of coral species is important for providing a wide variety of refuge spaces, such as crevices, holes and branches, which can influence the diversity and structure of reef-associated communities (Bell andGalzin 1984, Jones andSyms 1998).For instance, at small spatial scales (< 1 km), the recruitment of fish larvae depends in part on the availability of suitable refuges, and the postsettlement movements of fish recruits are largely influenced by the availability of microhabitat structures (Jones 1991, Tolimieri 1995, Caley and John 1996, Caselle and Warner 1996, Schmitt and Holbrook 1999, Johnson 2007).In juvenile and adult fishes, levels of competition and predation can also be influenced by the complexity of reef habitats, as more complex habitats may reduce competition by providing a greater spectrum of resources, and reduce predation by providing more prey refuges and/or reducing encounter rates between predators and prey (Holbrook and Schmitt 1988, Hixon and Beets 1993, Almany 2003, 2004).
Currently, coral reefs are changing very rapidly, particularly in the Caribbean where structurally complex reefs are becoming increasingly scarce, at least partly as a consequence of the loss of key reef-building, morphologically-complex corals (Alvarez-Filip et al. 2009a, 2011a, b).Such rapid declines in reef complexity are thought to have severe consequences for biodiversity and associated environmental services.The consequences for coral reef fishes are of particular concern because reef fisheries are extremely valuable and are an important source of protein and livelihood for many vulnerable coastal human communities (Newton et al. 2007, Allison et al. 2009).Declines in reef architectural complexity may therefore result in reductions in overall species richness, particularly of reef specialists, with potential long-term consequences for large-bodied species (Beukers andJones 1997, Graham et al. 2007).The implications of habitat-driven changes in fish assemblages are only just beginning to be understood.However, it is likely that the loss of some services generated by fish populations can have negative consequences for ecosystem functioning and human societies.For example, the reduction of large-bodied is likely to reduce the yield of local fisheries (Graham et al. 2007).The loss of key functional groups (e.g.herbivorous) may also be manifested in the reduction of resilience on coral reefs (Mumby et al. 2009).
As different species of coral vary in their vulnerability to environmental perturbations, the variety of natural and human disturbances can greatly impact the relative abundance of key coral species, potentially resulting in a differential loss of specific microhabitats.For example, in the Caribbean, rapid declines in live coral cover have been accompanied by a shift in communities towards reefs dominated by stress-resistant and less architecturally complex corals (e.g.Hughes 1994, Green et al. 2008).In this region, mid-water reefs are now dominated by stress-resistant corals such as Porites and Agaricia which provide consistently less architectural-complex reefs in comparison to Montastraea-dominated sites (Alvarez-Filip et al. 2011b).These shifts in coral community composition can potentially have severe consequences for associated reef fishes; hence there is a pressing need to understand the response of fish communities to the changes in coral community composition and associated loss of reef complexity.
Here we test whether more structurally-complex coral reefs support greater abundances or diversity of fish along a 15-reef gradient of architectural complexity in Cozumel, Mexico.As the architectural complexity of reef-building corals may be especially important for small-bodied fishes, including recruits and juveniles from larger species that rely directly on reef microhabitat features to settle and feed, we explore these relationships using both size-based and trophic-guild perspectives.

Field surveys
Cozumel is a continental island located 22 km off the east coast of the Yucatán Peninsula in the North-Western Caribbean.The island is approximately 46 km long on the north-south axis and 16 km wide (Fig. 1).Cozumel is surrounded by coral reefs, the most developed of which are on the western side of the island and primarily comprise shelf-edge reefs (mostly in the southwestern area), which typically have a complex framework formed by calcareous formations of several meters height, and barrier reefs, mostly in the central part of the island, which are less well-developed.Unlike many other Caribbean reefs that were originally primarily comprised of Acroporid species, Cozumel reefs have been built mainly by Montastraea (Muckelbauer 1990).The southwestern coast of Cozumel has been under official protection since 1980 (Alvarez-Filip et al. 2009b;Fig. 1), and while visitation and tourist activities are permitted, fishing is banned on the western coast of the marine reserve.

Description of fish assemblages and reef structure
Benthic composition, architectural complexity and reef fish assemblages were surveyed in 15 mid-depth reefs (10-15 m) within Cozumel's protected area between October 2007 and February 2008 (Fig. 1).The surveyed area in each reef (i.e.transects and the area between them) covered ca.2000 m 2 , and the 15 reefs sites were all at least 500 m apart.Between three and eight 30 x 5 m transects (150 m 2 ; number of transects per reef is given in Fig. 2) were surveyed on each reef, within the constraints due to logistics and weather conditions during the time available for study.In each transect, a plastic tape measure was unrolled as the diver moved forward, identifying all the conspicuous fishes that crossed the transect line.
Individuals were counted in an instantaneous manner to minimise double-counting and inflated density estimates (Ward-Paige et al. 2010).The fork length of each fish was estimated within 10 cm categories (≤ 10, 10 -20, 20 -30, 30 -40, 40 -50 cm), which is well within the precision possible with minimal training (Darwall and Dulvy 1996;Dulvy et al. 2004).The surveyor was previously trained in fish identification and size estimation, and has carried out similar surveys in the study area for several years.Count time was not standardized because this was dependent on fish abundance, diversity and subsequent habitat measures.All transects were haphazardly located within the reefs and surveyed between 09:00 and 15:00 h by the same observer.
Benthic community structure was characterized using the point intercept method (Loya 1972) in each 30 m transect once the fish census was complete.The percentage cover of each benthic component (live coral species, soft coral, turf algae, macroalgae, sponges, bare rock and sand) was identified every 25 cm, resulting in 120 estimates per 30 m transect.
Architectural complexity was measured with two complementary methodologies: a chain measure of local reef-relief and a larger-scale visual categorical estimate.For the small scale we used the rugosity index, which is the ratio of a length of chain moulded to the reef surface to the linear distance between its start and end point (Risk 1972).A perfectly flat surface would have a rugosity index of one, with larger numbers indicating more complex surfaces.A three-meter chain (0.7 cm link-length) was used to estimate reef rugosity five times evenly distributed along each 30 m transect, and the average was calculated as a measure of transect-level rugosity.The reefscape-scale visual assessment of architectural complexity consists of assigning the area surrounding each transect a score from 0 (flat surface) to 5 (exceptionally complex reef with numerous caves and overhangs) (Wilson et al. 2007).

Description of fish assemblages
Fish assemblages were characterized using three univariate metrics that describe the overall fish community (total number of species, total number of individuals and total biomass) and four metrics that describe food-web structure (mean trophic level, trophic guilds and the slope and intercept of the size spectra).Although aggregating data has the disadvantage of overlooking species-specific responses, these metrics were selected as our aim was to identify community-wide functional effects of reef structure on energy flow and foodweb structure.
The three community metrics were calculated at the transect level and then averaged across transects to produce reef-level fish community metrics.Total fish biomass was calculated by converting the estimates of fish length to weight using species-specific length-weight conversions (Froese and Pauly 2009).
Mean trophic level was used as a proxy of the food chain length, and was calculated for each reef by averaging the trophic level (obtained from Fishbase; Froese and Pauly 2009) of each individual recorded on all transects within each reef.Although, in theory, food chain length could be measured as the trophic level of the largest apex predator, this is difficult to estimate in practice because such 'asymptotic' values are highly sensitive to census effort.
However, as the combination of the efficiency of energy transfer between trophic levels and size spacing between trophic levels (predator-prey mass ratio) tend to be constant across marine ecosystems (Barnes et al. 2010), it is expected that the mean trophic level of the community will be highly correlated to the maximum trophic level.As mean trophic levels were calculated consistently across reefs, this metric allows comparison of relative differences in food chain length among reefs.In addition to the mean trophic level, fishes were also grouped into six major trophic guilds (planktivores, herbivores, omnivores, invertivores, carnivores and piscivores) following Paddack et al. (2009); these guilds represent important differences in behavioral, functional and ecological responses to habitat variation.
In aquatic environments, organismal function varies more with the size of individuals rather than the identity of species, particularly for fishes that grow by several orders of magnitude during their life (Jennings 2005).Ontogenetic size change has profound ecological implications, because most life history and ecological characteristics, such as growth schedules, age at maturity, lifespan, reproductive output, density-dependence in juvenile survival and relative prey sizes are related to body size in marine fishes (Jennings et al. 1999, Denney et al. 2002, Barnes et al. 2010).The size-frequency distribution (size spectrum) of fish captures the size distribution of energy in the ecosystem and describes the flow and depreciation of energy toward larger individuals and species.The size spectrum represents, in effect, the total amount of energy in the ecosystem with the intercept reflecting the overall productivity or energy entering the community (Jennings 2005, Jennings et al. 2008, Wilson et al. 2010) and the slope representing the flow and depreciation of energy across fish size classes (Boudreau and Dickie 1992, Kerr and Dickie 2001, Jennings 2005).The size spectrum of the fish community on each reef was calculated by summing the total number of individuals in each of the five 10 cm length categories, up to 50 cm.Individuals larger than 50 cm were very rare (n = 3) and therefore were excluded from the analyses.The fish size spectrum was then described by the slope and intercept of a linear regression of this size frequency distribution (Dulvy et al. 2004).Body length data were log 10 (x + 1) transformed before analysis, and the midpoint of size classes was rescaled to zero to remove the correlation between slope and intercept (Daan et al. 2005).Steeper slopes indicate an increase in the relative abundance of small fish (a combination of recent recruits of larger bodied species and larger size classes of small species), a decrease in large fish abundance, or both, and this in turn can be disentangled by examining the relative contribution of each size class along the environmental or impact gradient (Dulvy et al. 2004).

Data analyses
Benthic data were averaged within transects and then within individual reefs, and the average percentages of cover of the benthic components were arcsine square root transformed.Exploratory analyses revealed strong positive covariation between the percentage cover of coral and both measures of architectural complexity (chain and visual estimation methods; table 1) across the 15 reefs.Hence, we calculated a combined index of "coral cover-complexity" from the first axis of a principal component analysis of these three variables.This first PCA axis explained 99.6% of the variance and comprised coral cover (coefficient = 0.11), chain rugosity (coefficient = 0.23) and the visual estimation of habitat complexity (coefficient = 0.97), such that positive PCA1 scores reflect high coral cover and high architectural complexity and the converse for negative scores.
The influence of different components of reef structure and benthic community on all fish assemblage metrics were first explored with hierarchical partitioning analysis (Mac Nally 2000, 2002).The metrics that were identified from these analyses as responding to variation in reef structure were then included in analyses (also using hierarchical partitioning) of the influence of different dominant coral types on the structure of fish assemblages.We restricted this second analysis to four categories: the three most dominant coral genera in Cozumel, Montastraea, Agaricia and Porites, and one category including all the other remaining coral genera.Previously, it has been shown that these coral genera differ in colony and size structure, and hence differ in their contribution to reef complexity, with the large and massive nature of Montastraea corals disproportionately increasing reef complexity (Alvarez-Filip et al. 2011b).Species from the genera Agaricia and Porites have variable morphologies (e.g.encrusting, branching).However, the most architecturally complex species of the both genera (Porites porites complex and Agaricia tenuifolia) had relative low cover during our study period.These two genera are therefore represented mainly by small encrusting and massive species (Alvarez-Filip et al. 2011b).
Hierarchical partitioning is a multiple-regression technique that calculates goodness of fit measures for all possible combinations of independent variables in linear models, in order to calculate the average contribution of each predictor (Chevan andSutherland 1991, Mac Nally 2000).Hierarchical partitioning provides both the independent and the joint effects of each predictor, with independent values representing the explanatory power associated with a single predictor and joint values representing the explanatory power of the parameter of interest in combination with other variables (Mac Nally 2002).In some situations, significant independent effects can be suppressed by antagonistic joint effects, in which case hierarchical partitioning technique converts antagonistic effects into negative joint contributions, effectively unmasking the independent contributions (Chevan and Sutherland 1991).Independent effects of the explanatory benthic variables on the fish assemblages were tested with 1000 permutations of the data matrix and significance was evaluated at the 5% level (Mac Nally 2002).We used hierarchical partitioning rather than traditional multiple regressions or other multivariate analysis, because this method deals well with multicollinearity of predictor variables (Mac Nally 2000, 2002).We conducted hierarchical partitioning within R using the hier.partpackage (R Development Core Team.2011; Walsh and Mac Nally 2008).Hierarchical partitioning does not calculate the direction of the effects of the explanatory variables, so these were derived from Pearson correlations.

Fish and habitat description
In Cozumel there was a gradient of increasing coral cover from the north to the south of the island (Fig. 2), and the reefs are largely dominated by Agaricia, Montastraea and Porites, although the surface cover of these genera varied considerably between reef sites with the northernmost and southernmost reefs having a particularly high relative cover of Montastraea species (Fig. 2).The major reef-building corals of the genus Acropora accounted only for 1.1% of the total cover of scleractinian corals in Cozumel.
A total of 10,965 individuals from 93 species of fish were recorded across the 15 reefs (appendix 1).The most common fishes were those that mainly feed on plankton and algae (appendix 1, Fig. 2), and there was a clear trend of decreasing abundance with body size; the smallest individuals (< 10 cm) represented 48.1 % of the total abundance while larger individuals (> 40 cm) accounted for only 0.7 % of total abundance.

Influence of benthic components on fish assemblages
Fish species richness and total fish abundance were both positively (but not significantly) associated with cover-complexity and negatively associated with the percentage cover of bare rock and sand (Fig. 3), and total fish biomass was unrelated to any of the benthic community components or structure (Fig. 3).However, much stronger links were apparent between reef structure and the trophic and size structure of fish assemblages.The food chain length (as indexed by mean trophic level) and the size spectra slope and intercept of fish communities were significantly related to increasing cover complexity but not to any of the other benthic components (Fig. 3).The intercept of the fish size spectra increased and the slope of the size spectra decreased with increasing cover-complexity (Fig. 3).While mean trophic level and size spectra slope were related only to cover-complexity, the size spectra intercept was also influenced by a joint contribution from different benthic components (Fig. 3), suggesting that, while the overall productivity of fish communities is influenced by several different benthic components, the food chain length is mainly driven by 'coral covercomplexity'.

Influence of coral identity on size and trophic structure of fish assemblages
The abundance of the two smallest fish size classes (< 20 cm) was significantly greater on reefs with greater cover of Montastraea, whereas other coral genera had no significant effect on the abundance of fishes from any size categories (Fig. 4).The positive relationship between the abundance of small fishes and scleractinian corals suggests that the pattern indicated by the size spectra analyses (i.e. higher intercept and steeper slope on reefs with greater coral cover-complexity; Fig. 3) is likely to result from greater densities of the smallest size class (<10 cm) on complex reefs.The association between fishes and coral cover and complexity was highly size-dependent.Initially the smallest fishes (< 10 cm) appear to be most dependent on the presence of the most architecturally-complex corals (i.e. Montastraea), and this relationship decays for increasingly larger size classes, suggesting a strong ontogenetic detachment from scleractinian corals.Fishes in the second size-category (10 to 20 cm) were also more abundant in the presence of Montastraea, but the presence of other types of corals also appear to influence their abundance, given the considerable positive contribution of the joint effects to the variation in abundance of this size-class.Although abundance of fish in size classes greater than 20 cm were not significantly related to the cover of any coral genera, those in size classes 20 -30 and 30 -40 cm tend to be positively associated with the cover of all coral genera while the largest fish (40 -50 cm) tend to be negatively associated with coral cover (Fig. 4).
Although recruit and juvenile phases were not distinguished from adult phase in our surveys, our results suggest that Montastraea corals may indeed favor the presence of the early-stages (recruits) of larger species of fishes.While the number of small fish species (maximum fork length, L max < 20 cm) was not significantly correlated with the cover of Montastraea corals (Size class <10 cm: r = -0.36,P = 0.18; Size class 10-20 cm: r = 0.45, P = 0.09), small individuals of species with larger maximum body size (L max > 20 cm) were significantly more abundant on sites with greater cover of Montastraea (Size class <10 cm: r = 0.56, P = 0.03; Size class 10-20 cm: r = 0.66, P < 0.01).
The abundance of carnivores, omnivores and herbivores all increased significantly with increasing cover of Montastraea corals (Fig. 5) but the cover of Porites, Agaricia or other coral species did not significantly influence the abundance of fish species of any trophic group.

DISCUSSION
The identity, cover and complexity of coral reefs have profound implications for the fundamental processes structuring coral reef fish communities.Here, we show that coral cover and reef architectural complexity positively influence the abundance of small fishes and presumably the food chain length and energy flow through the unfished Caribbean fish assemblages of Cozumel, Mexico.Across reefs varying greatly in coral species composition, coral cover and complexity, greater abundances of small-bodied reef-associated fishes are supported on 'healthier' more architecturally-complex reefs, in particular those dominated by robust Montastraea corals.However, the community-wide dependence of fishes upon complex reef structures declines for larger-bodied fish size classes.This suggests that complex reef structures are most important for maintaining reef fish communities mainly by facilitating the presence of key early-life stage individuals and small-bodied fishes that are in turn preyed upon by larger fishes (Caley andJohn 1996, Beukers andJones 1997).
In contrast to size-based analyses, the variation in species richness, abundance and biomass across reef sites was weakly and non-significantly explained by 'coral covercomplexity' and other benthic components.One possible explanation for this is that the spread and quantity of data was insufficient to detect such sample-size-dependent relationships (see also Jennings et al 2001;Dulvy et al 2008).In this context, our analyses suggest that an aggregated size-based food web view of coral reef ecology may have more power to detect variation in fish communities, and therefore provide more insights into coral reef community structure, function and energy flow than the classical terrestrial species-based functional perspective.
The positive influence of 'coral cover-complexity', and particularly of the genus Montastraea, on reef fishes is most evident in small-bodied individuals and probably at small spatial scales.This suggests that the complexity provided by reef-building corals is an important mediator of the energy entering the reef system and the productivity of the reef-fish community.These results are consistent with experimental and observational analyses showing that, at small spatial scales (< 1 km), the recruitment of fish larvae depends in part on the availability of suitable refuges, and the post-settlement movements of fish recruits are largely related to the availability of microhabitat structures (Jones 1991, Tolimieri 1995, Caselle and Warner 1996, Schmitt and Holbrook 1999, Johnson 2007).The steepness of the size spectra on more complex reefs has been attributed to the depleted abundances of apex predators.For example, fishing can change the size distribution of fish communities by directly reducing the number of large-bodied fishes that in consequence will result in prey release (Dulvy et al. 2004, Graham et al. 2005, Wilson et al. 2010).However, Cozumel reefs have been under protection for 30 years (Alvarez-Filip et al. 2009b), and have relatively high abundances of apex predators (relative to other Caribbean reefs; Knowlton and Jackson 2008).Therefore the steepening of the slope in the size-spectrum, in combination with the elevated intercept, is likely to result from a stronger association of small-bodied fishes with 'coral cover -complexity' rather than an effect of fisheries exploitation of larger fish.
The dependence of reef fishes on complex reef structures declines gradually with increasing body length, which suggests an increase in home range size of larger-bodied species and individuals, and the inevitable increasing use of adjacent reef-associated habitats.This observation coincides with the general theory that home range size scales positively with body size, and the degree of exclusivity of habitat use declines with increasing body size due to the metabolic scaling of energy acquisition (Jetz et al. 2004).Small-bodied fishes are strongly related to coral complexity, and particularly to cover of Montastraea, probably because coral colonies from this genus develop many holes and crevices that may function as shelter for small fishes.For example, these corals are known to attract high numbers of damselfishes of the genus Stegastes that commonly recruit and 'garden' algae in these corals (Precht et al. 2010).Indeed, the small-scale pre-and-post settlement processes of reef-fishes are more likely to be explained by the presence of Montastraea than by other corals such as those from the genus Porites, likely due to the more suitable microhabitat characteristics provided by Montastraea (Tolimieri 1995).Our results are also consistent with previous studies showing that, at patch scales, Montastraea-dominated reefs are the most species-rich habitats on Caribbean reefs and have the greatest intra-habitat variability (Harborne et al. 2008).The abundance of fish of body lengths 10 -30 cm seem to be influenced by the overall cover of reef-building corals, not just Montastraea corals.This may be a consequence of increases in the home range size with increasing body size, as larger Caribbean reef fishes tend to also increase in mobility (Kramer and Chapman 1999), and are therefore less likely to be associated with specific attributes of individual coral colonies.The reduction in dependence on reef complexity of large-bodied fishes suggests that factors such as the availability of prey (Hixon and Beets 1993) and/or reef conditions at larger scales (e.g. the effect of protection at Marine Reserve scales; Kramer and Chapman 1999) or connectivity to adjacent habitats (Mumby et al. 2004;Harborne et al. 2008) may be stronger influences on the abundance of these adult fishes.
The increase in mean trophic level with 'coral cover -complexity', in addition to the strong and positive influence of Montastraea on higher-trophic level (especially carnivorous) fish, suggests that reefs with greater architectural complexity support longer food chains with more predator-dominated fish communities compared to flatter reefs with lower coral cover.
The ratio of the abundance of high-trophic guilds (carnivores and piscivores) to low-trophic guilds (herbivores and planktivores) is indeed significantly correlated to 'coral covercomplexity' (r = -0.50;P = 0.03) indicating that relatively flat reefs are largely dominated by low-trophic level fish.However, we caution that the species-based trophic levels used here do not take account of ontogenetic diet shifts, and do not entirely describe trophic relationships (e.g.some small-bodied zooplanktivores fishes can have a trophic categorization similar to some large predators; Froese and Pauly 2009).A more definitive test of our findings would require individual-level estimates of trophic level using stable isotopes.
The cover of Monstastraea corals also appears to positively influence the abundance of fishes of lower trophic guilds such as omnivores and herbivores.The omnivorous fishes recorded in this study are commonly recognized as reef-dependent species that feed on reef benthic flora and fauna (appendix 1; Froese and Pauly 2009).The greater number of refuges for small invertebrates and benthic algae provided by the massive structures of Montastraea may therefore support a greater number of food items for these fishes (e.g.Idjadi and Edmunds 2006).The relationship between herbivorous fishes and Montastraea cover is more complex; the complex structures and ramets of Montastraea may support greater algal diversity and biomass (Mumby et al. 2005, Piñón-Gimate andCollado-Vides 2008), and thus be attractive to herbivorous fishes.In addition, some herbivorous fish such as Sparisoma viridie and Stegastes planiforms are common grazers of Montastraea colonies (Mumby 2009, Precht et al. 2010) and may therefore also be attracted by the live tissue of these corals.
The ongoing decline in reef architectural complexity throughout the Caribbean may have profound effects on fish communities (Alvarez-Filip et al. 2009a, Paddack et al. 2009).
In particular, our findings suggest that recent declines in the overall density of fish communities (Paddack et al. 2009) may be in part a consequence of ongoing declines in the cover of Montastraea in reefs through the entire region (Hughes and Tanner 2000, Bruckner and Bruckner 2006, Steneck et al. 2009).Our study suggests that changes in the abundance of Montastraea populations could have significant and direct detrimental effects on the abundance of small-bodied fish, with subsequent impacts on the abundance of adult fishes through failures in recruitment and reductions on the transfer of energy through the food web (Graham et al. 2005, Graham et al. 2007, Wilson et al. 2010).This may result in bottlenecks limiting the flow of energy and production of fish assemblages and fisheries yields.
Montastraea historically ranked in importance with Acropora palmata and A. cervicornis in overall contribution to Western Atlantic reef structure (Jackson 1992).
Acroporids have now almost vanished from Caribbean reefs and important functional attributes have been lost with the demise of these highly complex coral structures.Both acroporid species are now listed as Critically Endangered by the International Union for the Conservation of Nature (Bruckner 2003, Aronson andPrecht 2006).More recent declines in the abundance of Montastraea populations and increases in the relative abundance of stressresistant coral species are further compromising the capacity of reefs to provide habitat for reef-associated biodiversity (Green et al. 2008, Steneck et al. 2009, Alvarez-Filip et al. 2011b).Our findings indicate that these recent changes in coral composition on Caribbean reefs may result in very little high quality recruitment habitat for fish species.The importance of these fish populations in sustaining thousands of livelihoods across the region highlights the urgent need to identify and implement means of preserving and improving reef habitat quality throughout the Caribbean.

a
Trophic level is based in diet of adult organism followingPaddack et al. (2009).In synthesis, planktivores, feed primarily on macro-and micro-zooplankton; herbivores are fishes that include detritus, turf algae and macroalgae in their diet; omnivores, feed on a mix of animal and plant material; invertivores, feed primarily on invertebrates associated to both hard and soft bottoms; carnivores feed on both mobile invertebrates and fishes; and piscivores feed almost only on fishes.b size class 1 = ≤ 10 cm; size class 2 = 10 -20 cm; size class 3 = 20 -30 cm; size class 4 = 30 -40 cm; size class 5 = 40 -50 cm.

Fig. 2 .
Fig.2.Mean (± SE) reef rugosity, cover of Agaricia, Montastraea, Porites and all other coral genera, and abundance (individuals per transect) of six major trophic guilds on the 15 surveyed reefs.Reefs are ordered from north to south (see Figure1).The total number of transects is provided in brackets.

Fig. 3 .Fig. 4 .
Fig. 3. Results of hierarchical partitioning analyses of the contribution of reef complexity and the cover of reef benthic components to the community and food-web structure of reef fishes.Bars show the independent (black = statistically significant at P  0.05, grey = not significant at P > 0.05) and joint (white) effects, and +/-indicates the direction of each relationship.Joint effects are either additive (above the line) or antagonistic (below the line) to the independent effects

Table 1 .
Associations between the percentage cover of different benthic components of the reefs and reef architecture.Values on the lower left of the table are Spearman-correlation coefficients and values in the upper right of the table (in italics) are P values.Visual = visual estimate of reef complexity.