Coral Reefs (2000) 19:37±49
Ó Springer-Verlag 2000
H. V. Cornell á R. H. Karlson
Coral species richness: ecological versus biogeographical in¯uences
Accepted: 9 September 1999
Abstract Species richness in communities varies with habitat area, productivity, disturbance level, intensity of species interactions, and regional/historical eects. All of these factors in¯uence coral richness but their eects vary with spatial scale, position on the reef, and regional location. Species richness of corals along depth gradients shows a unimodal, hump-shaped curve that peaks at intermediate depths. Moreover, the peak of the curve is higher in regions with larger species pools. This ``regional enrichment'' of the local community appears in line transect samples as small as 10 m in length. The pattern suggests that ecological factors operating over scales of tens of meters and regional/historical factors operating over thousands of kilometers can both aect local richness. Regional factors probably include dierences in speciation relative to extinction rates among regions and proximity of local sites to richness hotspots. Plausible factors operating at the local scale are species interactions, disturbance, and productivity which combine in dierent ways to produce the unimodal pattern. Shallow areas support few species because extinction rates are high due to frequent disturbance or because of environmental extremes. In addition, high productivity encourages rapid growth and thus the potential for intense interspeci®c competition. In areas where branching acroporids are abundant, exclusion by these dominant competitors is possible. Deep areas may be depauperate because few species can tolerate the low light levels found there. Areas of intermediate depth have the richest communities because they are open for colonization by many species and because extinction rates are low. Several theories may explain this ``openness'' and species persistence:
H. V. Cornell (&) á R. H. Karlson Department of Biological Sciences, University of Delaware, Newark, DE 19716 USA e-mail: [email protected]
1. Occasional disturbance coupled with low growth rates results in glacially slow exclusion by the dominant competitor. 2. Aggregation of corals creates spatial variation in the intensity of competition and thus refuges from competition within a spatial landscape. Inferior competitors persist because they are superior at dispersal and refuge colonization. 3. Specialist predators focus on high-density juvenile populations near the parent, creating ecological space for colonization by non-prey. 4. Coral competitive abilities are roughly equal and recruitment into the community is a probabilistic event. The community thus exhibits random drift and exclusion is an extremely lengthy process. Based upon empirical evidence, these theories are listed in order of plausibility, but still need to be rigorously tested. Key words Competition á Dispersal á Disturbance á Diversity á Saturation
Introduction One of the most obvious and consistent patterns in ecology is that some habitats support more species than others. The factors responsible for such dierences in species richness have been a central concern of ecologists at least since the turn of the century. After such extensive consideration, it has become clear that these factors can be ecological or evolutionary, and that the time scales in which they operate can range over several orders of magnitude, from a matter of hours to millions of years. Similarly, they can operate over spatial scales ranging from a few square meters to thousands of square kilometers (Cornell and Lawton 1992; Ricklefs and Schluter 1993). Multi-scale in¯uences on species richness are inevitable because habitats are nested within much larger biogeographical regions having long evolutionary histories and comprising many dierent local environments. Species can arise, evolve, and disperse among
suitable habitats in the region, thus species richness at local and regional spatial scales are inextricably linked. Here we explore these multi-scale links in assemblages of reef-building coral species, which represent some of the richest communities in marine environments. We will review the various factors that in¯uence local richness, and argue that biogeography can be at least as important as local ecology in explanations for richness variation among dierent locations. Next we will discuss the eects of local spatial scale on the detection of regional in¯uences on local richness and propose guidelines for reef ecologists as to how best to sample coral richness. Finally, we will evaluate various models that have been proposed to explain why local communities are open to regional eects and discuss their relevance to reef processes.
Factors which in¯uence species richness Ultimately there are just two processes which in¯uence local richness; the rate of addition of species to the community via speciation or colonization and the rate of species removal via local extinction and emigration. The rates of these opposing processes (given in terms of total numbers of species per time unit) can be fast or slow, and can ¯uctuate smoothly or in a punctuated manner over more than one time scale. Nevertheless, the dierence among these rates at any given time determines how the richness of the community is changing. When addition equals removal, local richness depends upon the average per species addition rate relative to the per species removal rate (analogous to the slopes of the colonization and extinction curves in the theory of island biogeography). If the per species addition rate is low and the per species removal rate is high, local richness will be low. Similarly, if the per species removal rate is relatively low, increasing the per species addition rate will increase local richness. There are numerous factors that aect addition and removal rates. Here we consider area, productivity and light, species interactions, disturbance, and history/biogeography. Area This ®rst and most obvious factor is sheer space. Larger habitats can support more species than smaller habitats as predicted by the now familiar non-linear species area equation, S = CAz where S is local richness, A is habitat area, and C and Z determine the slope and shape of the relationship (Preston 1960). In this context, species addition and removal rates become colonization rates into the habitat from a species pool and extinction rates of habitat residents, respectively (MacArthur and Wilson 1967). Species numbers increase with habitat area for at least three reasons (Connor and McCoy 1979): (1) larger habitats provide larger targets to potential colonists, increasing the per
species colonization rate, (2) larger habitats are more environmentally diverse, increasing the successful colonization and persistence of species specialized to dierent environments, and (3) larger habitats provide space for larger populations, reducing the per species extinction rate as predicted by the theory of island biogeography. In addition, species numbers at very small scales in space-restricted environments can be limited by the size of the neighborhood occupied by each individual, especially if the species are sessile (e.g., plants and corals). All of these possibilities assume that colonization and extinction are in balance, but even if they are not, species numbers can still increase with area. For example, continuous fringing reefs may become discontinuous over time by anthropogenic in¯uences, reducing colonization rates and increasing the extinction rates of resident coral species. Since per species extinction rates are determined in part by habitat fragment size, species number will drop faster on smaller fragments than larger ones. Smaller fragments may also be more distant from potential colonization sources, resulting in larger decreases in the per species colonization rate. Both eects can generate a speciesarea relationship, even though richness in the fragments may not have reached a new equilibrium. Since speciesarea relationships can occur under non-equilibrium as well as equilibrium conditions, habitat area is a robust predictor of species richness in many dierent taxa and environments. There are literally hundreds of examples from the literature (Connor and McCoy 1979), making it one of the most pervasive empirical patterns in ecology. Productivity and light A second factor conceptually close to habitat area is energy availability which is directly related to light availability for photosynthesis. Local habitats that absorb and transform more light into primary productivity can support higher numbers of species. In a recent literature review, Wright et al. (1993) showed that primary productivity and other energy-related factors correlated strongly with species richness in many dierent taxa: Fraser and Currie (1996) demonstrated this correlation for coral genera. Wright et al. developed the ``species energy theory'' to explain these correlations. According to this theory, available energy limits local richness by limiting the population sizes of resident species. As average population size decreases, per species extinction rates increase, and species number is reduced. The species energy theory is thus similar to the theory of island biogeography, with the main dierence being that increased energy is substituted for increased area as the explanation for reduced extinction rates (Tilman and Pacala 1993). Indeed, energy availability is commonly calculated as the product of productivity per unit area and habitat area. The species energy theory has been valuable in bringing an
ecosystem perspective to community ecology. Nevertheless, the theory has been somewhat controversial (e.g. Tilman and Pacala 1993; Latham and Ricklefs 1993a, b; Francis and Currie 1998). For example, it remains to be seen whether the energy-richness correlation occurs because of the direct in¯uence of energy on species numbers or because energy correlates with other important factors such as area or habitat diversity. Moreover, richness does not always show a simple monotonic increase with energy. Waide et al. (1999) extensively reviewed the literature on the richnessenergy relationship. Energy was measured in terms of productivity, and the richness-productivity relationships were grouped into four types: monotonically increasing, monotonically decreasing, neutral, and unimodal (hump shaped) where richness at ®rst increases and then decreases with increasing productivity. The data were then cross-classi®ed by taxon (animal, plant) and ecological scale (within community, across community, and continental to global). The data were also examined across geographical scales but the patterns are similar to those across ecological scales and are not shown here. All four types of relationship occurred at nearly all scales by taxon classes, but the relative frequency of the relationships showed some notable trends (Fig. 1). In plant systems, positive relationships predominated at the largest scales whereas the relative frequency of unimodal relationships increased at within-and among-community scales. In animal systems, the trend was similar, but unimodal and positive relationships were more equally distributed at the within- and among-community scales. Coral systems seem to con®rm this trend from monotonically positive to unimodal relationships as scales get smaller. At regional/ oceanic scales, the relationship is usually monotonically increasing. For example, Fraser and Currie (1996) found a monotonic relationship for coral genera at a scale of 101±106 km2. At the local scale where energy availability is presumed to operate, the pattern is more frequently unimodal. Huston (1985) and Cornell and Karlson (1996) found unimodal relationships along depth/habitat gradients spanning scales of 101±102 m within individual reefs. Some of this pattern might simply be due to habitat selection by coral larvae during settlement. However, productivity on reefs generally declines with depth due to light extinction (see below) and may play an important role as well. Energy limitation on local richness may certainly be implicated where the relationship is ascending, but why does richness sometimes decline at higher values of energy availability? Species interactions The answer to this last question will vary for dierent systems, but in many cases may involve interspeci®c interactions, the third factor that can aect species addition and removal rates. As energy availability
increases, relative allocation among species can change due to shifts in competitive superiority or the ability to avoid attacks by natural enemies. Under these circumstances, population densities of some species may increase while others are driven extinct, reducing the number of coexisting species in the more energy-rich environment. Experimental manipulations of nutrient levels in diverse plant communities have demonstrated with dramatic consistency this negative eect of enrichment (and associated increases in productivity) on local richness (Gough et al. 1999). The idea that species interactions, mediated by local environmental conditions, can limit local richness has been central to modern theories of community regulation. According to this view, each habitat has a unique environment, and only those species from the pool that can coexist together in that environment will establish successfully. Dierences in local richness thus result from environmental variation and species composition among habitats. In the case of competition, population levels of resident species are limited by resource availability and physical conditions. Thus, given sucient time and the absence of perturbing outside in¯uences, local richness will saturate at a level determined by the number of limiting resources and physical factors. Tilman and Pacala (1993) provide an informative summary of the development of this model. The model can be made more general if limiting resources are broadly de®ned. For example, if resources include ways of avoiding natural enemies, then phenomena such as apparent competition (Holt 1977) can also play a role in saturation. At saturation, species are suciently similar in their resource use that ®ner subdivision of resources becomes increasingly dicult and new colonizations into the community must be accompanied by extinctions of current residents. Potential colonists may also simply be preempted by established populations. It is important to point out that the level of saturation can be aected by factors other than the number of resources. Colonization order of species into the habitat may aect the richness limit, and communities can exhibit multiple stable points representing dierent richness levels in similar abiotic environments (Drake 1990, 1991). Habitat heterogeneity can also aect the richness limit. Heterogeneity can increase local richness by providing opportunities for specialists to use the habitat in unique ways so that they avoid strong interactions with potential competitors. Rosenzweig (1995) discusses the eects of habitat heterogeneity at length. Other mechanisms by which heterogeneity can allow local richness to increase are discussed in a later section. Finally, sourcesink eects can allow species that would normally go extinct to persist in the community by means of constant migration from source areas (Pulliam 1988). Whether or not communities ever reach saturation as predicted by this model, there are clear recent demonstrations that species interactions have strong in¯uences on community structure and might exclude species from the community (see Lawton 1999 for a brief review).
41 b Fig. 1 The form of the productivity-diversity relationship at three dierent ecological scales. Individual bars are the percentage of published studies within each group that show a unimodal, positive, negative, or no relationship between productivity and diversity. n = the number of published studies in each group (From Waide et al. 1999). With permission, from the Annual Review of Ecology and Systematics, Volume 30, 1999
Disturbance The proposition that species interactions limit local richness assumes that communities normally exist at equilibrium with respect to the impact of interactions on population levels. In light of strong evidence for the countervailing eects of disturbance (Petraitis et al. 1989), it is apparent that communities often don't exist at equilibrium. If disturbances keep the community below equilibrium, extinction of species due to resource limitation need not occur because populations never reach densities necessary to generate strong interspeci®c interactions. Higher species richness can thus be maintained (Connell 1978). Disturbance can aect richness within the disturbed patch itself (Huston 1979) as well as within the larger landscape by generating a patchwork of disturbed areas of dierent ages and thus species compositions (Paine and Levin 1981). In the within-patch case, if disturbance is absent, exclusion or preemption due to species interactions prevails and local richness will be relatively low. At moderate disturbance, interaction-imposed extinctions are reduced and local richness will be relatively high. At high disturbance, populations can be driven to critically low levels, extinction will increase again and local richness will be relatively low. In the amongpatch case, the absence of disturbance will result in a homogeneous landscape of old patches where competitive exclusion and thus low richness prevails. High disturbance rates will generate a homogeneous landscape of young patches with high extinction rates and thus low richness once again prevails. However, intermediate disturbance should produce maximum heterogeneity of patch ages, maximum composition dierences among patches, and thus maximum richness across the landscape. In both within and among patch cases, a balance between the rate of disturbance and the rate of population recovery is reached at some spatial scale larger than the individual patch. Communities in such circumstances exhibit transient dynamics, and the disturbance rate relative to the rate of recovery is more important in predicting local richness than resource availability. In addition, local richness shows the same type of unimodal relationship with disturbance as it sometimes does with productivity. Petraitis et al. (1989) oer ample evidence that such unimodal relationships are common and widespread in diverse taxa (but see Hubbell et al. 1999 for a persuasive counter example at the within-patch scale in a tropical rainforest).
History/biogeography All of the above processes assume the existence of a species pool from which the local communities are drawn. Factors that characterize the local site (area, abiotic environment energy, interactions, disturbance) have some in¯uence on the number of species from the pool that can coexist within that site. Dierences in species richness arise in part because of the way these factors aect species addition and removal rates. But where does the pool come from in the ®rst place? Ultimately, it is generated by species additions and removals operating over larger spatial and temporal scales. The species that comprise the pool are distributed over large biogeographical regions that have had long geological histories. At these larger scales, species additions are mainly due to in situ speciation and to immigration from other regions. Species removals are mainly the result of wholesale species extinctions from the region. The relative importance of immigration, speciation, and extinction rates in setting regional richness will vary, depending upon the taxon and history of the region. The number and kinds of species that occur across the region are thus determined to a large degree by its unique historical development. As a consequence, regions with dierent histories can support dierent numbers of species, and this variation in regional richness can be re¯ected in the richness of the local community. That is, communities in richer regions can support a higher local richness (Cornell and Lawton 1992; Cornell and Karlson 1997). This higher local richness can occur either because the ancestral community from which it evolved was rich or because the community is open to colonization by species from elsewhere in the region (Brooks and McLennan 1991). Because regional and local richness are linked in this way, understanding community assembly will require more focus on macroevolutionary processes. Community structure is undoubtedly in¯uenced by both contemporaneous and macroevolutionary processes, thus, determining their relative in¯uence will be one of the most important challenges in community ecology.
What factors in¯uence coral species richness? The coral reef is the paragon of a rich marine community. Many taxa express this high richness on reefs and the corals themselves are no exception. Some areas in the western Paci®c support as many as 500 hermatypic species, and up to 33 species can occur along a single 10 m transect (Moll 1986). Nevertheless, coral species richness is quite variable among reefs and among locations on the reef. Some of this variation is unpredictable in space and time, but corals often show a unimodal trend with reef depth and relative distance from the shoreline (Huston 1985; Cornell and Karlson 1996). Richness is generally low on shallow reef ¯ats that are relatively close to shore. It increases to a maximum on
not entirely a result of pool exhaustion since some species that can obviously survive at shallow depths are often not present in the samples. However, many coral species are not adapted to deeper water and reduced richness at depth at least in part may be the result of reduced pool size (Huston 1985). Pool size dierences among regions can also aect peak richness as is demonstrated by the Caribbean-Indian Ocean comparison. The environmental gradient along which the unimodal pattern occurs is complex, but the pattern oers a convenient opportunity to compare the various factors that control local richness. Area
Fig. 2 Unimodal relationship between coral species richness and shoreward position along the surface of the reef. Shoreward position generally follows the depth gradient on each reef. Data on local richness were compiled from 63 primary literature sources published from 1918 to 1993. Local richness was evaluated using quadrats, line transects, and point sampling units of various sizes. In order to account for variation due to sampling method and sample unit size, local richness was regressed against sample unit size for each of the three sampling methods. Residuals of local richness were then calculated by subtracting the predicted value (on the regression line) from the observed value for each data point. The residuals from each of the three regressions were then pooled and used as the dependent variable. Habitat code: 1 inner ¯at, 2 outer ¯at, 3 upper slope, 4 midslope, 5 lower slope. (From Cornell and Karlson 1996)
the upper to mid- slope between 15 and 30 m in depth and then gradually decreases as depth increases along the lower slope (Fig. 2). This unimodal pattern occurs on both lagoonal and seaward slopes and can occur along narrow as well as broad depth gradients. Moreover, overall richness among reefs in separate regions may dier but the relative richness along depth gradients is still unimodal. For example, reefs in the Chagos Archipelago in the Indian Ocean support over 3 times as many coral species as those in the West Indies in the Caribbean Sea (185 versus 50; Karlson and Cornell 1998). Yet the unimodal trend is apparent in both regions, with the main dierence being that the diversity peak at mid-depths is higher on the richer Chagos reefs (Huston 1985). Many other extant and fossil marine invertebrate groups show this unimodal pattern with depth, albeit over larger depth ranges (Rosenzweig and Abramsky 1993). The lower richness of corals at shallow depths is
The eects of area on coral richness at relatively small spatial scales can be determined by comparing species numbers in sample units of dierent sizes. That is, the sampling unit becomes the estimator of area. Corals have been sampled most extensively using quadrat and line-transect methods. The functional relationship between transect length and species number is dierent from that between area and species number, so it is not used here. We gathered data from 63 sources and regressed local richness on several independent variables including quadrat area, as well as depth and habitat type (Cornell and Karlson 1996). The multiple regression makes it possible to separate the eect of quadrat area from that due to environmental dierences. Quadrat area was a strong independent predictor of local richness. The result is, of course, not surprising as virtually every species assemblage studied shows this pattern. Even at large geographic areas and over geological time, area plays a role in determining coral richness (Chadwick-Furman 1996). In our own analysis, area eects were studied in quadrats ranging mainly from 0.25± 100 m2 (in one instance an area of 125420 m2 was sampled but not in a quadrat). Within this size range several factors might be responsible for the richness-area correlation, none of which have been studied in detail. If areas are particularly small (e.g. 0.1±1 m2) local richness might be limited simply by neighborhood eects which are characteristic of sessile species. In coral communities, individual corals take up substrate space and the number of species per area might be limited simply by the average size of individual corals, some of which easily span a meter or more in diameter. Habitat heterogeneity, on the other hand, is likely to be important in medium to large areas. In very large sampling units, the samples probably transgress habitat boundaries, and between-habitat heterogeneity may contribute to total richness. Even within a habitat/depth zone, heterogeneity is high on reefs due for example to disturbances that are patchy in space and time, producing sites that are at dierent successional stages. Patchy distribution of natural enemies and competitors, currents, and anthropogenic eects such as pollution, dredging, and over®shing may also play a role. Finally, pure spatial
phenomena such as patchy recruitment, aggregation, and local neighborhood dispersal may also contribute. Disturbance and other factors that destroy corals can also increase extinction rates. Since small disturbances are likely to be more frequent than larger ones, and since small populations are more likely to be driven extinct by such factors, smaller areas are likely to suer higher per species extinction rates. They may also provide smaller targets for dispersing propagules, thus reducing per species colonization rates. Both factors can reduce local richness. In the absence of direct evidence supporting these mechanisms, little can be de®nitively said about their relative in¯uence. However, we suspect that heterogeneity will be a dominant factor in the richness-area correlation because of its ubiquity on reefs and its importance in other systems. Productivity and species interactions Area is clearly not the only factor in¯uencing coral richness since equal-sized samples from dierent depths support dierent numbers of species. In his review of coral diversity patterns, Huston (1985) posits that productivity and species interactions are the ultimate factors responsible for the unimodal richness trend with reef depth. Huston points out that most corals contain endosymbiotic algae and are effectively primary producers. Light is thus the primary factor which determines coral growth rate and productivity. Light levels diminish rapidly with reef depth and thus produce a monotonic reduction in productivity and growth along the depth gradient. There is also a detectable shift from autotrophy to heterotrophy with depth. The tendency toward heterotrophy seems to increase with increases in average polyp size (Porter 1976) and species with large polyps are more abundant on the deep reef (Huston 1985). Thus, corals on the deep reef probably tend to rely more on plankton than sunlight as an energy source. However, there is some evidence that food in the form of plankton also declines with depth (Huston 1985). Richness of corals clearly decreases with depth from the mid-slope down to the lower slope, and this decrease is likely to be related in some sense to productivity. Light levels become so low that autotrophic species drop out of the assemblage, thus the number of species that can maintain a positive growth rate declines with depth. If species energy theory explains the gradient then the total population level summed over all species (and thus average density per species) should also decline with depth. These declines should increase the probability of local population extinctions and thus declines in richness. Common sense suggests that this is the case, but there are few data to support the idea. Some plant taxa, which have lifestyles similar to corals, may not show this pattern. In some plant systems, there is an inverse relationship between productivity and the density of plant stems even
though richness declines with declining productivity (Tilman and Pacala 1993). Tilman and Pacala use this observation and related information to argue that species-energy theory does not explain latitudinal gradients in tree species richness. However, it is still an open question whether the theory applies over a smaller spatial scale where the resident ¯ora has a common evolutionary history. In any event, the theory needs to be explicitly tested for corals. In shallow water, productivity is high. In nutrientrich waters high productivity can lead to turbidity, reduced light levels and, consequently, reduced coral richness. In less turbid situations, light availability increases. Although high light levels can actually inhibit growth in some cases, coral growth rates can be very rapid for many species. Huston (1985) suggests that high light levels encourage intense interspeci®c competition, rapid approach to competitive equilibrium, and consequent low richness as a result of competitive exclusion. According to this model, pool size, average population size and thus richness increases with light from the lower slope up to the mid-slope but then competition becomes intense, reducing richness again in the expected way. Huston (1985) and Lang and Chornesky (1990) provide ample evidence that corals and other sessile species on the reef surface compete, but evidence that richness is reduced by competitive dominance is circumstantial at best. This evidence comprises observations of monospeci®c stands, high cover by corals and other organisms, and increases in cover by some coral species, particularly branching corals, at the expense of others on the upper reef over relatively short time intervals (e.g. Colgan 1987; Brown and Suharsono 1990). However, there is also evidence that some corals, once they become established, are dicult to dislodge. Direct contacts between colonies sometimes result in aggressive responses and countermeasures that result in standos and competitive coexistence (Karlson 1999). Strong inferences regarding competitive exclusion will thus require observed increases in cover, reciprocal changes in relative abundance, and direct observation of species disappearance at individual sites over extended time periods. The latter should not be exceedingly dicult if growth and exclusion are as rapid as they are hypothesized to be in shallow water. Competition-induced extinctions should be easier to document for sessile corals than for more mobile organisms because encounters, overgrowth and population reduction can be directly observed. In spite of some long-term monitoring of speci®c reef patches, such competition±induced extinctions have not been extensively documented (see Lang and Chornesky 1990; Karlson 1999). Moreover, many upper reef sites that don't show high cover still support low numbers of coral species relative to mid-depth zones. These sites suggest that competitive exclusion is not always necessary to generate low richness in shallow water.
Disturbance and natural enemies
Disturbance is another major player in the richness decline in shallow water. Disturbance is more variable than productivity. Light always extinguishes with depth but disturbance depends on factors which vary from reef to reef and among sites on particular reefs (Huston 1985). Nevertheless, disturbance can have a strong in¯uence on richness where it occurs. Reef ¯ats and crests that are disturbed by cyclones and storms often have sparse coral cover and low richness (Connell et al. 1997). Diseases, predators, and land-based pollution can also act as disturbance agents (Aronson et al. 1998, Miller and Hay 1998; Edinger et al. 1998), and may have similar eects on richness. If the disturbances are suciently frequent, colonies on the upper reef may grow no faster than those in deeper water because the former are injured more often. If growth rates approach 0 or become negative, population levels of many species will be kept low and others will become extinct. The absence of competitive contacts in such circumstances rules out competitive exclusion as a cause for low richness in these frequently disturbed areas. As depth increases, disturbance generally declines and richness increases. Huston posits that low rates of disturbance coupled with reduced competition for space due to slow coral growth rates and reduced populations of other competitors (e.g. algae and other sessile organisms) maintain the peak diversities at mid-depths. According to this scenario, competitive exclusion is prevented by the very slow approach to competitive equilibrium and by occasional disturbances. However, corals do not always grow more slowly in deeper water. For example, many species on the upper reef have massive or encrusting life forms which grow relatively slowly whereas some mid-depth species are branching forms with rapid radial extension rates. Even some species that form plates in deeper water grow more rapidly than some other species in shallow water. Moreover, even if competition among coral species is less intense in deep water, competition between corals and other species may not be. For example, sponges and algae can kill more coral colonies than many coral-coral interactions. Given these exceptions to the competition gradient with depth, the role of competition needs to be treated cautiously until it is carefully tested. Reduced disturbance may certainly permit more species to persist in deeper water, and slower growth, if corals in fact grow more slowly on average with depth, may forestall equilibrium. However, occasional disturbances may not be required to prevent competitive exclusion. A considerable body of theory predicts that large numbers of species can coexist at equilibrium in hypothetical communities if realistic assumptions are made about the characteristics of local environments and the biology of the resident species (e.g. Cornell and Lawton 1992; Cornell and Karlson 1997). An attempt will be made to evaluate this body of theory for coral assemblages in the discussion.
The factors that in¯uence coral richness in a given region have a strong historical component. One measure of historical eects on community structure is the degree to which regional factors aect local richness. The relative importance of regional and local eects can be independently evaluated by means of regression models (Karlson and Cornell 1998). In order to keep the models simple, we combined local eects into two proxy variables, depth of the site and habitat type (inner and outer reef ¯at, upper, mid, and lower slope). These variables subsume much of the local environmental variation known to correlate with local species richness. Regional variables included species and generic richness in the region, distance from the local site to the equator and to the nearest high-diversity region, and the average age of genera in the region. Table 1 is a summary of the regression coecients for local species richness plotted against just regional species richness. The values of these coecients re¯ect the relative impact of a major regional variable on local richness in the various regions. It comes as no surprise that local coral richness is sensitive to both local and regional variables. In the Indo-Paci®c, local and regional variables account for 61 and 39% of the explained variation in local richness, respectively. Similar results were obtained in the very speciose central Indo-Paci®c where regional richness ranged from 244 to 411 species. In less speciose regions of the Indo-Paci®c (19±88 species), local richness was extremely sensitive to regional eects which accounted for fully 95% of explained variation in local richness. Clearly, historical eects as manifested in these regional variables (mainly pool size and proximity) can have a strong eect on coral community structure. The results imply that the local environment can modify local richness, but it does not set its upper limit. Put another way, local richness might generate a unimodal curve when local richness is plotted against site depth and habitat, but the average height of that curve above the abscissa will vary depending upon the characteristics of the region in which the community is imbedded. This Table 1 Slope of the regression of log-transformed local richness on log-transformed regional richness in various biogeographical regions. Local richness was evaluated using quadrats, line transects and point sampling units of various sizes. In order to account for variation due to sampling method and sample unit size, residuals of local richness from sampling eort regressions were pooled and used as the dependent variable in the regressions below. S = range of species richness, N = number of studies. (Modi®ed from Karlson and Cornell 1998) Region
Slope (mean 1SE)
Entire Indo-Paci®c Species-rich Indo-Paci®c Species-poor Indo-Paci®c Western Atlantic
0.27 0.24 0.60 0.06
19±411 244±411 19±88 19±50
42 18 11 21
0.04** 0.05* 0.06** 0.07 n.s.
t-tests: **P < 0.00001, *P < 0.0001, n.s. = not signi®cant
was hinted at previously in a comparison of Caribbean and Indian Ocean reefs. The dominant eects of regional variables in the depauperate Indo-Paci®c are not unexpected since communities containing fewer species ought to be more open to larger scale in¯uences. In the speciose Indo-Paci®c, the reduced in¯uence of regional eects could imply that these richer communities are ®lling up and are thus less sensitive to pool eects. This hypothesis needs to be tested more rigorously with a comparative analysis of coral communities ranging across the entire Indo-Paci®c richness gradient, utilizing a standardized hierarchical sampling design. In contrast to the Indo-Paci®c, local richness in the western Atlantic was totally insensitive to the regional variables. Depth and habitat accounted for all the explained variation in the regression model. However, this pattern does not arise because Atlantic communities are saturated. It is most likely due to historical events that are not detected by the regression models. In the Pliocene, when the Atlantic and Indo-Paci®c provinces became separated, major extinctions of mostly stenotopic species generated a highly depauperate eurytopic fauna. A high proportion of the fauna occurs at most sites in the Caribbean (Liddell and Ohlhurst 1988), thus regional variables have little detectable impact on local communities.
Scale effects The above analysis was based upon data compiled from the literature that combined samples of local communities at many dierent spatial scales. We controlled for scale eects by regressing local richness against the size of the sampling unit. The analyses were then carried out on the residuals of the richness-sample size regression. Although the analysis detected substantial regional effects on local richness, it provided no information on the appropriate local scale at which regional eects should be studied. Regional eects should be large at very large local scales (e.g., Fraser and Currie 1996). But large local scales may exceed the boundaries of local communities and confound within- and between-habitat diversity. Regional eects may not be detectable at all at very small scales because only a few individuals will be circumscribed by the sample. The literature data can be used to determine the threshold at which regional eects on local richness start to show up. We used regression models to evaluate the sensitivity of local richness to regional richness at four spatial scales, two from quadrat samples (1 m2 and 5 m2) and two from line transects (10 m and 30 m; Karlson and Cornell 1999b). Regressions of local on regional richness at all four scales are shown in Table 2. Local richness was sensitive to regional richness at all scales in at least some of the analyses (simple regressions on regional richness or stepwise multiple regressions with depth and habitat data included). However, regressions on the 10-m line transects were consistent and strong regardless
Table 2 Coecients of simple regressions of local on regional richness at four sample unit sizes. Both variables are logtransformed and standardized. (From Karlson and Cornell 1998, 1999a, Karlson 1999, and unpublished analyses) Sampling unit
1-m2 quadrats 5-m2 quadrats 10-m line transects 30-m line transects
0.15 0.27 0.81 0.02
1.61 n.s. 3.80* 21.32** 0.28 n.s.
114 187 244 179
2.2% 7.1% 65.1%