Is coral richness related to community resistance to and recovery from disturbance?

Introduction

A large body of recent and classical ecological research has focused on how biodiversity influences the stability of communities (May, 1973; McCann, 2000; Pimm, 1984; Hughes & Stachowicz, 2004; Steiner et al., 2006; Cardinale et al., 2012). Despite this interest, there is still no consensus on whether or why diversity influences stability. This may be due in part to the multitude of ways stability has been defined (Grimm & Wissel, 1997; McCann, 2000; Pimm, 1984). Here, we define community stability as the ability to maintain a given state regardless of perturbation, invasion, or extinction (McCann, 2000). Mechanistically, stability is made up of two components: resistance—the degree to which a community changes in response to a disturbance, and recovery or “resilience”—the rate of return to pre-disturbance conditions (Grimm & Wissel, 1997; Ives & Carpenter, 2007; May, 1973; McCann, 2000).

Species richness (the number of species within an area) is thought to influence community stability via several mechanisms (Ives, Cardinale & Snyder, 2005). For example, species rich communities are more likely, due to chance alone, to include greater functional diversity, and redundancy of function within functional groups—the insurance hypothesis (McNaughton, 1977; Tilman & Downing, 1994). The inclusion of species that can tolerate or benefit from disturbance could increase community resistance (McCann, 2000; Tilman & Downing, 1994; Yachi & Loreau, 1999). Additionally, positive feedbacks among species via facilitation may be more common in diverse communities, which may enhance post-disturbance recovery (Naeem, 1998; Naeem & Li, 1997; Yachi & Loreau, 1999).

Understanding the role of species richness in community stability has been a focus of a number of field studies and meta-analyses (Duffy et al., 2001; Ives & Carpenter, 2007; Stachowicz, Bruno & Duffy, 2007; Stachowicz et al., 2008; Cardinale et al., 2012). Some experimental studies in terrestrial plant communities have found a positive relationship between species richness and stability (Ives & Carpenter, 2007; Tilman & Downing, 1994). For example, more diverse plant communities appear to be more resistant to disturbances including drought (Tilman & Downing, 1994). Manipulations of marine plant and invertebrate community richness have found that diversity is positively related to resistance and recovery (Stachowicz, Bruno & Duffy, 2007). Likewise, Boyer, Kertesz & Bruno (2009) experimentally demonstrated that species-rich algal communities were more resistant to disturbances than species-poor communities in estuarine ecosystems of eastern North Carolina. One of the few large scale tests of the diversity–stability concept was a study of boreal forest communities in north-western Quebec (Grandpre & Bergeron, 1997), which similarly found that greater species-richness conferred more resistance to changes in community composition and abundance following a disturbance.

We tested the hypothesis that species richness is related to community stability (resistance and recovery) in coral reef ecosystems. Coral communities are affected by a large range of natural and anthropogenic disturbances including storms, disease outbreaks, coral bleaching due to ocean warming, dynamite fishing, and predator outbreaks (e.g., Acanthaster planci or Crown-of-Thorns Starfish (COTS)). A recent study of coral-community recovery found that disturbance type, e.g., “physical” vs. “biological” disturbances, and other reef characteristics had little effect on recovery rate; although, recovery was influenced by post-disturbance coral cover, geographic region, and reef management (Graham, Nash & Kool, 2011). We expanded on this work by asking whether coral richness is related to resistance to and or recovery from disturbances at a large spatio-temporal scale using coral cover as a metric of reef state. We hypothesized that coral communities in species rich regions would exhibit greater resistance and recovery in response to disturbances. We synthesized the results of 41 field studies documenting post-disturbance changes in coral cover across a natural gradient of coral richness.

Methods

Results

There was substantial variation among sites in predicted coral richness (24–552 species, Fig. 1), coral cover loss (Fig. 1), and recovery rate (Fig. 2). The full linear regression model testing the null hypothesis that coral species richness was not related to (sqrt transformed) post-disturbance coral cover loss (decline) found that all covariates (including richness; p = 0.79) except pre-disturbance coral cover (BLCC, p < 0.0001) were unrelated to decline (Fig. S2, Table 1, Model 1, Adj. R² = 0.63, see all model results and R code in the Statistical Supplement). A simpler model (Model 2, Adj. R² = 0.60) that included three covariates (depth, richness, and BLCC) with a similar fit indicated that the richness effect was marginally significant (p = 0.03). There was no relationship (Adj. R² = 0.01, p = 0.50) between predicted coral richness and initial cover, thus it is unlikely that collinearity between these two covariates led to the putative effects of richness on resistance. When we removed sites (two potential outliers) with extreme decline and high diversity, the relationship between relative decline and predicted species richness was still significant (in the reduced model with three covariates).

The full recovery model (Table 1, Model 4 in the Statistical Supplement, Adj. R² = 0.27) was the best fit and indicated that none of the covariates were related to recovery rate. Predicted coral richness in particular appeared a very poor predictor of coral recovery rate (p = 0.56).

Discussion

We did not detect an effect of region, disturbance type, or depth on absolute loss in coral cover due to disturbance (Table 1, Statistical Supplement). There was a weak positive relationship between predicted species richness and coral loss (Fig. 1). However, a richness effect (p = 0.03) was only detected in the reduced model (Model 2), from which maximum depth, region, and disturbance type were removed (Statistical Supplement). In the full model with all five covariates (Model 1), the richness effect was not significant (p = 0.79).

Concordant with several previous studies (Graham et al., 2008; Selig & Bruno, 2010; Selig, Casey & Bruno, 2012), reefs with greater initial cover lost the most coral following disturbance (Fig. 1, Statistical Supplement). This could be due in part to the dominance of high cover reefs in the western Pacific and Indian Ocean by plating acroporid corals, that are especially sensitive to both physical and biological disturbances (Darling et al., 2012), i.e., not due to a cover effect per se, but instead to the relationship, both among reefs and regions, between coral cover and species composition. Acroporids are among the most sensitive coral species to most of the disturbances known to cause declines in coral cover, e.g., thermal stress, disease, predation, and physical disturbance (Bruno et al., 2001; Darling et al., 2012; Willis, Page & Dinsdale, 2004). Therefore, reefs dominated by plating and branching acroporids–the highest cover reefs–are likely to be especially susceptible to disturbance. Although high taxonomic richness may confer greater functional diversity and redundancy, this also implies a greater number of specialists that may not be well adapted to disturbance or stressful conditions (Dobzhansky, 1950; Schluter, 2000) and may partially explain our results. In fact, Côté & Darling (2010) argued that managing for coral cover could counter intuitively make reefs more susceptible to disturbances including climate change.

Unlike several previous experimental studies in marine systems (Allison, 2004; Reusch et al., 2005; Williams, 2001), our mensurative study found that the rate of post-disturbance recovery was not significantly related to taxonomic richness (Table 1, Fig. 2). The large observed variance in recovery rate among sites and studies (up to 12.5% per year) appears to be due to factors not included in our model and unrelated to estimated species richness.

Graham, Nash & Kool (2011) reported that coral recovery varied somewhat among regions, with the highest observed recovery rates in the western Pacific, the most species rich region, (Roberts et al., 2002). Despite that finding, in our analysis coral richness was a poor predictor of recovery rate (Fig. 2); although, the low richness reefs of the eastern Pacific recovered slowly, recovery of Caribbean reefs with similar richness was greater and similar to that seen in the more diverse western Pacific and Indian ocean (Fig. 2). This finding could be due in part to a publication biases towards studies that document recovery and the fact that we searched for studies documenting acute disturbances; whereas, chronic perturbations are also responsible for Caribbean coral losses. Furthermore, studies conducted in the eastern Pacific and included in our analysis were mostly conducted on reefs dominated by slow-growing, massive Porites. Given the current state of the Caribbean (Bruno et al., 2009) there is no doubt that coral cover on most Caribbean reefs has not recovered from various natural and anthropogenic disturbances over the last several decades (Roff & Mumby, 2012). Hence, our results suggest that low-diversity Caribbean reefs have the potential to recover (in terms of coral cover) quite rapidly (e.g., Idjadi et al., 2006), but for a variety of reasons and factors outside the scope of our study, they rarely do (Roff & Mumby, 2012).

The majority of reefs in our study were monitored for less than eight years and few were monitored long enough to recover fully (Fig. S3), given observed rates of recovery. Of the 23 reefs for which we had both initial post-disturbance and peak-year coral cover, only five (all monitored for at least 11 years after disturbance) recovered to pre-disturbance levels of cover. If recovery (coral cover increase) is indeed non-linear over time, we could have seen different results, e.g., a significant richness effect if the component studies were conducted longer. To a degree, our results could also be an artifact of the loss of Acroporid corals from Caribbean reefs in the 1980s, due primarily to disease but in some locations to storms and predation (Aronson & Precht, 2006). Our earliest Caribbean recovery study began in 1987 (Edmunds, 2002), after Acroporid populations had crashed regionally (Aronson & Precht, 2006). Had Acropora cervicornis and A. palmata been present on Caribbean reefs when the datasets used in our analyses were collected, we might have seen more rapid recovery on low diversity reefs because these corals were amongst the fastest growing corals in the Caribbean region. However, we may also have seen even less resistance to disturbance due to their high coral cover and vulnerability to various stressors.

It is possible that systematic variation in disturbance intensity could obscure a richness effect on coral community resistance. For example, disturbance intensity may have co-varied with coral richness or peak coral cover. Although quantification of the intensity of disturbance is beyond the scope of this study, it would, at least in theory, be possible. Our study also addresses a very large scale richness gradient but ignores smaller scale variation, e.g., among reefs or zones (Huston, 1985). Furthermore, we used predicted richness values from known biogeographic ranges (Veron, 2000; Veron et al., 2011), which should reflect patterns of richness at this scale, but can make richness and region hard to tease apart and cannot account for local site-level differences due to factors such as reef zone, disturbance history, or micro-climates. Diversity of fish assemblages may also be critical for coral recovery dynamics (Bellwood et al., 2004), which was not considered in this study. Future analyses with empirical diversity data of corals and fish, controlling for disturbance intensity and other drivers of coral loss and recovery, could improve our understanding of the influence of diversity on community stability in coral reef ecosystems.

Supplemental Information