Connect with us

Climate News Live

Caribbean reefs of the Anthropocene: Variance in ecosystem metrics indicates bright spots on coral depauperate reefs

Caribbean reefs of the Anthropocene: Variance in ecosystem metrics indicates bright spots on coral depauperate reefs


Global Warming

Caribbean reefs of the Anthropocene: Variance in ecosystem metrics indicates bright spots on coral depauperate reefs

1 INTRODUCTION Human activities in the Anthropocene are altering ecosystems, causing significant differences from their predecessors in species composition, community structure, and ecological function. Examples of such altered ecosystems include regenerating forests dominated by non‐native trees, grassland, and shrubland systems replacing forests due to human alteration of fire regimes, and the ongoing tropicalization of temperate…

Caribbean reefs of the Anthropocene: Variance in ecosystem metrics indicates bright spots on coral depauperate reefs

1 INTRODUCTION

Human activities in the Anthropocene are altering ecosystems, causing significant differences from their predecessors in species composition, community structure, and ecological function. Examples of such altered ecosystems include regenerating forests dominated by non‐native trees, grassland, and shrubland systems replacing forests due to human alteration of fire regimes, and the ongoing tropicalization of temperate marine ecosystems (Hobbs et al., 2006; Vergés et al., 2014). Some ecosystem changes are so dramatic, persistent, and widespread that recovery to previous states appears unlikely, creating a pressing need to understand the ecological functions and ecosystem services provided by altered systems (Hobbs et al., 2014). However, evaluations of ecological function are often overlooked, with reductions in functions and services assumed to result from the shift to a new ecosystem state. This is true for coral reefs, diverse ecosystems that are undergoing dramatic alterations due to climate change, overfishing, disease, pollution, sedimentation, and food web disruption, as well as interactions among these stressors (Hughes et al., 2017). Reef building corals are the foundation species of tropical reefs, but many reefs have experienced dramatic losses of hard corals. In the case of the Caribbean, live coral cover has declined from 50%–60% to 10%–20% or lower in recent decades (Gardner, Côté, Gill, Grant, & Watkinson, 2003; Jackson, Donovan, Cramer, & Lam, 2014). As live coral declines, reefs lose topographic complexity and undergo major changes in community composition that can result in declining ecological functions and compromised ecosystem services (Alvarez‐Filip et al., 2011; Graham et al., 2014; Rogers et al., 2015). Yet, the full implications of these changes for the diverse range of potential ecosystem services provided by reefs—including food provision, coastal protection, tourism, recreational opportunities, and aesthetic and cultural value—have not been thoroughly investigated (Woodhead, Hicks, Norström, Williams, & Graham, 2019).

Numerous studies seek to better understand historical states of tropical reefs, address the drivers of coral declines, and consider how corals and coral reef ecosystems might be better conserved or restored (Hughes et al., 2017; Jackson et al., 2014; Pandolfi et al., 2003). Similarly, reef management and restoration initiatives often focus on coral cover. For example, some use coral propagation, gardening, and outplanting to slow or reverse the trajectory of decline (Ladd, Miller, Hunt, Sharp, & Burkepile, 2018), while others focus on protecting herbivorous fishes with the intent that increased grazing will facilitate coral recovery by removing the seaweeds that compete with corals (Steneck, Mumby, MacDonald, Rasher, & Stoyle, 2018). To meet recovery goals, these efforts should be continued, but it is also critical to understand the variance in ecosystem states and the ranges of services currently provided on “degraded” (i.e., coral depauperate) reefs. The few studies that assess a broad array of functions and services for low‐coral states tend to assume or predict an overall decline in services (e.g., Mumby et al., 2008; Pratchett, Hoey, & Wilson, 2014; Rogers et al., 2015), despite limited empirical data to support some of these expectations. As one example, many studies presume that reef‐based tourism will suffer following coral decline (e.g., Rogers et al., 2015; Yee, Carriger, Bradley, Fisher, & Dyson, 2015). However, many tourists may not distinguish living hard coral from dead coral, soft corals, or other benthic invertebrates, and may care more about non‐coral attributes like water clarity, fish diversity, and the presence of large fishes (Gill, Schuhmann, & Oxenford, 2015; Uyarra et al., 2005). Furthermore, while there is a growing body of research acknowledging that reefs of the future will likely lack historic analogs (Hughes et al., 2017), there is still a focus on hard corals as the sole foundation taxon. A better understanding of the ecosystem functions and services provided by reefs that have lost significant hard coral and may be dominated by other species (e.g., macroalgae, sponges, or gorgonians) will enable more effective management into the future.

Here we examine 10 ecosystem metrics, relating to different ecosystem functions and services, under varying levels of coral cover at 328 sites across the Caribbean. We use data collected by the Atlantic and Gulf Rapid Reef Assessment (AGRRA) between 2011 and 2015 at forereef sites that currently have or historically have had abundant hard coral (AGRRA, 2018). In contrast to coral reef research that tends to focus on mean responses to coral loss, we assess the range and variability in these metrics across reefs with low (<10%), intermediate (10%–20%), and high (>20%) coral cover. Few sites in our dataset have coral cover higher than 30%, thus we recognize that our “high” cover category is modest relative to many historical Caribbean reefs. Nevertheless, within the range of coral cover observed in the present‐day Caribbean, we test whether lower coral cover results in declining means and/or changes to the variability of ecosystem metrics among reefs.

Based on a prevailing paradigm in the coral reef literature, our expectation was that various ecosystem metrics would decline in mean value on low coral cover reefs, but we did not have strong a priori predictions for changes in variance. In contrast to these expectations, our results suggest that “bright spots”—sites where these metrics are comparatively high—can occur where coral cover is low. Given that Caribbean reefs have experienced earlier and often greater coral loss than other coral reef systems such as those in the Indo‐Pacific (Jackson, 1997; Roff & Mumby, 2012), it may be a harbinger of changes to come for other reefs, or coastal systems in general. Therefore, lessons learned in the Caribbean could facilitate greater preservation of bright spots in other regions or ecosystems. As the Earth faces accelerated ecosystem change, there is a need to explore the variability in ecosystem services provided by altered ecosystems. Insights arising from such efforts may allow management to enhance bright spots and limit loss of key services.

2 MATERIALS AND METHODS

2.1 Ecosystem metrics

There are a wide variety of ecosystem processes on coral reefs that contribute to the maintenance of biodiversity and the delivery of benefits to people (i.e., ecosystem services; reviewed by Moberg & Folke, 1999; Mumby et al., 2008; Yee et al., 2015). Services are rarely measured directly, but we identify metrics that can be calculated from available survey data and are related to a number of important coral reef ecological functions and ecosystem services (Table 1; Figure S1). These metrics are incomplete proxies for the functions and services we are aiming to assess (e.g., fish species richness as one attribute of value to SCUBA and snorkel tourists) and most of them are related to more than one important ecological function or ecosystem service (Figure S1). Nevertheless, collectively they provide a useful if imperfect indicator for potential ecosystem services and functions at a site. These metrics include:

  1. vertical relief, a component of ecosystem structure underlying a variety of services including habitat provision and predation refuge for fishes (Graham & Nash, 2013) and coastal protection (Ferrario et al., 2014);
  2. calcareous macroalgal cover, a key contributor to sand and sediment production including for adjacent beaches (Lidz & Hallock, 2000);
  3. coral recruit density, a proxy for coral recruitment rates, and also important for reef recovery potential (Hughes et al., 2019);
  4. coral calcification rate, which can contribute to growth of reef structure, sand production, coastal protection, and reef recovery potential (Ferrario et al., 2014; González‐Barrios & Álvarez‐Filip, 2018; Perry et al., 2013);
  5. mean coral species richness, which captures components of biodiversity as an endpoint unto itself and as an important supporting service (Moberg & Folke, 1999), including facilitating coral growth and survivorship (Clements & Hay, 2019) and influencing the value of dive tourism (Uyarra, Watkinson, & Côté, 2009);
  6. herbivorous fish biomass, as a partial indicator for herbivory, because herbivores, by suppressing macroalgal overgrowth of the reef, can facilitate coral recruitment, growth, and survivorship (Hughes et al., 2007; Mumby, 2006) and improve the aesthetic value of the reef;
  7. potential fishery value, assessed as the potential market value of fish on surveyed reefs, as one key component of benefits from fisheries (Mumby et al., 2008);
  8. density of large fish, which contributes to the value of dive tourism (Gill et al., 2015) and may increase fishery value as fishers often target and/or receive higher prices for larger fish (Graham, Dulvy, Jennings, & Polunin, 2005);
  9. parrotfish bioerosion rate, which is positively related to sand production and negatively related to net reef accretion (Hubbard & Dullo, 2016; Perry et al., 2013); and
  10. mean fish species richness, which captures components of biodiversity as an endpoint unto itself and acts as an important supporting service (Moberg & Folke, 1999), including enhancing reef resilience (Burkepile & Hay, 2008), influencing the value of dive tourism and the aesthetic value of the reef (Williams & Polunin, 2002) and supporting fisheries productivity (Duffy, Lefcheck, Stuart‐Smith, Navarrete, & Edgar, 2016).
TABLE 1.
Ecosystem metrics examined in this study. All data from AGRRA () unless otherwise specified
Metric Description
Vertical relief Measured as the difference, in cm, between the highest structure (coral or reef rock) and the lowest point in the underlying substratum within a 1m diameter (measured every 5 m and averaged over a 30 m transect)
Calcareous macroalgae cover Percent cover of erect calcareous macroalgae (e.g., Halimeda spp., Galaxaura spp., etc.)
Coral recruit density In individuals/m2, counting all hard coral colonies less than 4 cm in diameter (identified to genus)
Coral calcification rate In kg CaCO3 m−2 year−1, applying species‐specific calcification rates from González‐Barrios and Álvarez‐Filip (2018) to coral cover data from AGRRA surveys
Coral species richness The number of coral species encountered on each transect (10 m2), with transect densities averaged over a site
Herbivore biomass In g/m2, using species‐specific fish biomass data from AGRRA surveys and summing biomass estimates across all fish species known principally to be herbivores (as classified by Paddack et al., 2009; see their table S2)
Potential fishery value In USD/m2, using species‐specific fish biomass data from AGRRA surveys and multiplying those biomass numbers by species‐specific price estimates (shown in Table S1)
Density of large fish In individuals/m2, counting all fish greater than 40 cm in size
Parrotfish bioerosion rate In kg m−2 year−1, applying species‐ and size‐specific parrotfish bioerosion rates from Ruttenberg et al. (2019) to parrotfish recorded in AGRRA surveys
Fish species richness The number of fish species encountered on each transect (60 m2), with transect densities averaged over a site
  • Abbreviation: AGRRA, Atlantic and Gulf Rapid Reef Assessment.

2.2 Dataset description

We used reef monitoring data from the AGRRA version 5 surveys, collected between 2011 and 2015 (AGRRA, 2018). We chose this dataset because it is relatively comprehensive in measuring a range of components of reef ecosystems. It also has broad coverage across the Caribbean surveying consistent reef habitats (e.g., forereef), and data are collected by scientists using a standardized protocol (available at www.agrra.org). Data were downloaded from the AGRRA website and were filtered to include only sites in forereef habitats between 3 and 20 m in depth that were surveyed with at least six fish transects and at least four benthic transects. This resulted in 328 sites across the Caribbean that were sampled between 2011 and 2015 (Figure 1).

image

Map of Atlantic and Gulf Rapid Reef Assessment survey sites included in this analysis and histograms of mean site‐level percent live coral cover within each ecoregion

The Atlantic and Gulf Rapid Reef Assessment surveys benthic organisms (crustose coralline algae [CCA], macroalgae, corals, sessile invertebrates) and ecologically and/or commercially important reef fish species, and documents attributes of reef condition (e.g., reef structure, coral health). Fish and coral taxa are identified to species (or occasionally to genera), while other components are grouped at higher taxonomic levels (e.g., sponges), with 93 fish taxa and 76 coral taxa reported across all surveys analyzed in this study. Coral recruits/juveniles (i.e., individuals <4 cm diameter) are counted and reef vertical relief (i.e., vertical height of the tallest coral or reef rock above the lowest point in the underlying substratum) is measured.

To reduce the influence of extreme outliers in the fish biomass data from AGRRA, we identified species that had individual observations (i.e., a single species, size and count record on an individual transect) greater than the 99th biomass quantile for all fish observations (n  = 54 species). For these species, we capped the biomass of each observation to the 99th quantile from all individual observations for that species in the entire dataset (Donovan et al., 2018). This led to our capping approximately 0.7% (660 out of 96,147) of the observations. These observations were not removed but were adjusted so that they are still large values without influencing the site averages as strongly as they would otherwise. In particular, this reduced the influence of data input errors and of rare, large groups of roaming or aggregating species, such as Sphyraena barracuda , which can constitute a large proportion of biomass in certain transects while not representing individuals that are permanent residents of the surveyed reef. We also calculated macroalgal cover (for Figure S5) for the AGRRA data by summing percent cover across calcareous macroalgae, fleshy macroalgae, and mixed macroalgae. Coral species richness and fish species richness were computed as species richness per transect, with transects measuring corals and fishes over a standardized area (10 and 60 m2, respectively). Transect richness values were then averaged for each site. This procedure standardized richness estimates to a fixed sampling effort and was necessary because sites varied in the number of transects surveyed.

To estimate parrotfish bioerosion rates, we applied species‐ and size‐specific estimates of bioerosion rates from Ruttenberg, Adam, Duran, and Burkepile (2019) to the AGRRA parrotfish observations at each site. These estimates were derived from field behavioral observations of 10 parrotfish species in the Florida Keys, United States, with the bioerosion rate for each fish of a given size class calculated as bite rate multiplied by bite volume, multiplied by the probability of leaving a scar, and by bulk density of reef carbonate, with the latter terms derived from the literature (Ruttenberg et al., 2019).

To estimate coral calcification rates, we applied species‐specific estimates of mean calcification rates (in kg CaCO3 m−2 year−1) from González‐Barrios and Álvarez‐Filip (2018; see their table 2) to mean cover estimates (in m2) from AGRRA. González‐Barrios and Álvarez‐Filip (2018) estimated calcification rates for the 47 most common scleractinian coral species in the Caribbean, covering most of the corals observed in the AGRRA coral dataset. Occasionally corals were recorded at the genus‐level in AGRRA; in such case we calculated an average calcification rate across all species in that genus (González‐Barrios & Álvarez‐Filip, 2018). The few species of corals present in AGRRA for which we did not have calcification estimates (<0.003 percent cover) were excluded from this analysis.

To determine the fishery value of the reef fish assemblages measured at the AGRRA sites, we used fishery statistics from Puerto Rico that provided an estimate of value in USD per pound of fish landed by species or by family. We extracted annual, island‐wide, ex‐vessel values for all species and families for which data were available from the period 2007 to 2011 from Matos‐Caraballo (2012). Because some groups were not valued in all years, we used the maximum annual value for each species (or for each family for species that lacked species‐specific values). We did not include species or families that did not appear in this dataset (Matos‐Caraballo, 2012); some of these omitted species are occasionally caught, but they have little or no value and are considered “trash fish” in Puerto Rico. While some of these species may have value on other Caribbean islands, the value is likely low, particularly since most of these species are small or unpalatable. We then used these value data (Table S1) to calculate the potential fishery value of the fish assemblage at each AGRRA site based on the biomass of each species. Ideally, we would have had price data from local and/or export markets in each location, but such data are non‐existent or buried in grey literature across the Caribbean. Puerto Rico is centrally located in the Caribbean, has an extensive and diverse reef fish fishery, and is unique in having a well‐documented time series of catch and value data.

2.3 Analyses

We examined whether lower coral cover results in declining means and/or changes to the variability of ecosystem metrics among reefs (Figure 2). We calculated 10 metrics related to different aspects of reef functions and services (Figure S1), using the AGRRA data described above (Table 1). For many analyses, we binned sites into three categories of live coral cover: (a) low: <10% cover; (b) medium: 10%–20% cover; and (c) high: >20% cover. These categories were defined based on the distribution of coral cover across all of the sites in the dataset (Figure 1); although 20% is still quite low relative to historical coral cover for many sites in the region, mean coral cover across all our sites was 14.36% (n  = 328; ±10%), and about a quarter of the sites (n  = 79) have coral cover greater than 20%. We confirmed the robustness of our results to these categories by examining patterns using an alternative set of coral cover thresholds (Figure S7). Our qualitative conclusion—that most metrics can be high at sites with low coral cover—did not appear to be sensitive to the coral cover categories that we used. We also show the relationships between each metric and coral cover on a continuous scale (Figure S8). We do not focus our analysis around the continuous results because we were most interested in examining variability in metrics across different levels of coral cover rather than trying to identify predictive relationships. Furthermore, coral cover explained a very low percentage of the variability for most of these metrics (Figure S8).

image

A priori hypotheses about different ecosystem patterns observed on high and low coral cover reefs. (a)–(d) Shifted distributions relative to a high‐coral state (gray). At low coral cover, relative values may have equal mean and increased variance (a; blue), decreased mean and equal variance (b; red), decreased mean and decreased variance (c; yellow), or decreased mean and increased variance (d; green). (e) Distributions from (a) to (d) as boxplots. These alternative hypotheses are not a complete set, as some services could also potentially increase with declining coral cover. Furthermore, we represent hypotheses based on comparisons of high and low coral cover for simplicity, although our empirical analyses focused on three coral cover categories to better capture variation in coral cover across the Caribbean

We also binned sites by low (<40 cm), medium (40–80 cm), and high (>80 cm) vertical relief. These categories were defined based on the distribution of vertical reef relief across all sites in our dataset. Using these definitions, about a third of the sites had low vertical relief and a little less than a quarter had high relief. These categories allowed us to evaluate, for example, whether relict structure from dead coral is associated with higher ecosystem metrics at sites with low live coral cover. We categorized sites based on protection status using the information provided in AGRRA for a subset of the sites (n  = 219) that noted whether surveys were conducted within marine protected areas (MPAs) to evaluate if protection status could confound our results. Lastly, we examined patterns separately for just the Western Caribbean and the Bahamian ecoregions sampled in 2011. These represented the two region‐year combinations for which we had the largest sample sizes, allowing us to evaluate the role of coral cover in the absence of regional and interannual variability. For example, we did not test for an association of the response variables with physical gradients like sea surface temperature or productivity; examining patterns at the ecoregional scale should partially control for variation in these physical factors.

We evaluated how each of the 10 metrics (Figure 3; Table 2) was associated with a variety of predictor variables using linear mixed effects models (LMERs) in the R package “lme4” (Bates, Maechler, Bolker, & Walker, 2015). We assumed a Gaussian family (normal distribution) for response variables in all models. For data that needed to be transformed to meet assumptions of normality (all variables except density of large fish, fish species richness, and coral species richness), we used an inverse hyperbolic sine transformation. Estimates from the linear model were then back‐transformed in order to calculate effect sizes. We treated ecoregion and year as categorical random effects, depth as a continuous fixed effect, and coral cover category (low, medium, high) as a categorical fixed effect. For the LMER examining MPA status (Figure S2), MPA status (yes or no) was included as a fixed effect in addition to the effects included in the main model (i.e., ecoregion, year, depth, coral cover). For the LMERs focused on individual ecoregions sampled in a single year (Figures S3 and S4), we removed ecoregion and year as random effects.

image

Distribution of 10 metrics, by three categories of mean percent live coral cover. High coral cover is defined as >20%, medium is 10%–20%, and low is <10%. Data from AGRRA (2018; number of sites for high, medium, low coral cover: a–g, i, j, n  = 79, 120, 129; h, n  = 65, 104, 103). Lines represent individual data points. P‐ value from linear mixed effects model results (Table 2) shown in parentheses after panel title, with results of Tukey’s post hoc test shown by letters next to boxes (coral categories sharing a letter are not significantly different from one another)

TABLE 2.
Results of linear mixed effects models (LMER) and generalized linear mixed effect models (GLMER) testing the effect of coral cover categories (low, medium, high) on the 10 metrics; and results of chi‐squared tests examining whether sites in the 90th or 75th percentile for each metric are distributed in coral cover categories proportional to the number of sites in each category. Shading corresponds to significance level (blue) and effect size (green)

Subscribe to the newsletter news

We hate SPAM and promise to keep your email address safe

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *


Global Warming

Caribbean reefs of the Anthropocene: Variance in ecosystem metrics indicates bright spots on coral depauperate reefs

1 INTRODUCTION Human activities in the Anthropocene are altering ecosystems, causing significant differences from their predecessors in species composition, community structure, and ecological function. Examples of such altered ecosystems include regenerating forests dominated by non‐native trees, grassland, and shrubland systems replacing forests due to human alteration of fire regimes, and the ongoing tropicalization of temperate…

Caribbean reefs of the Anthropocene: Variance in ecosystem metrics indicates bright spots on coral depauperate reefs

1 INTRODUCTION

Human activities in the Anthropocene are altering ecosystems, causing significant differences from their predecessors in species composition, community structure, and ecological function. Examples of such altered ecosystems include regenerating forests dominated by non‐native trees, grassland, and shrubland systems replacing forests due to human alteration of fire regimes, and the ongoing tropicalization of temperate marine ecosystems (Hobbs et al., 2006; Vergés et al., 2014). Some ecosystem changes are so dramatic, persistent, and widespread that recovery to previous states appears unlikely, creating a pressing need to understand the ecological functions and ecosystem services provided by altered systems (Hobbs et al., 2014). However, evaluations of ecological function are often overlooked, with reductions in functions and services assumed to result from the shift to a new ecosystem state. This is true for coral reefs, diverse ecosystems that are undergoing dramatic alterations due to climate change, overfishing, disease, pollution, sedimentation, and food web disruption, as well as interactions among these stressors (Hughes et al., 2017). Reef building corals are the foundation species of tropical reefs, but many reefs have experienced dramatic losses of hard corals. In the case of the Caribbean, live coral cover has declined from 50%–60% to 10%–20% or lower in recent decades (Gardner, Côté, Gill, Grant, & Watkinson, 2003; Jackson, Donovan, Cramer, & Lam, 2014). As live coral declines, reefs lose topographic complexity and undergo major changes in community composition that can result in declining ecological functions and compromised ecosystem services (Alvarez‐Filip et al., 2011; Graham et al., 2014; Rogers et al., 2015). Yet, the full implications of these changes for the diverse range of potential ecosystem services provided by reefs—including food provision, coastal protection, tourism, recreational opportunities, and aesthetic and cultural value—have not been thoroughly investigated (Woodhead, Hicks, Norström, Williams, & Graham, 2019).

Numerous studies seek to better understand historical states of tropical reefs, address the drivers of coral declines, and consider how corals and coral reef ecosystems might be better conserved or restored (Hughes et al., 2017; Jackson et al., 2014; Pandolfi et al., 2003). Similarly, reef management and restoration initiatives often focus on coral cover. For example, some use coral propagation, gardening, and outplanting to slow or reverse the trajectory of decline (Ladd, Miller, Hunt, Sharp, & Burkepile, 2018), while others focus on protecting herbivorous fishes with the intent that increased grazing will facilitate coral recovery by removing the seaweeds that compete with corals (Steneck, Mumby, MacDonald, Rasher, & Stoyle, 2018). To meet recovery goals, these efforts should be continued, but it is also critical to understand the variance in ecosystem states and the ranges of services currently provided on “degraded” (i.e., coral depauperate) reefs. The few studies that assess a broad array of functions and services for low‐coral states tend to assume or predict an overall decline in services (e.g., Mumby et al., 2008; Pratchett, Hoey, & Wilson, 2014; Rogers et al., 2015), despite limited empirical data to support some of these expectations. As one example, many studies presume that reef‐based tourism will suffer following coral decline (e.g., Rogers et al., 2015; Yee, Carriger, Bradley, Fisher, & Dyson, 2015). However, many tourists may not distinguish living hard coral from dead coral, soft corals, or other benthic invertebrates, and may care more about non‐coral attributes like water clarity, fish diversity, and the presence of large fishes (Gill, Schuhmann, & Oxenford, 2015; Uyarra et al., 2005). Furthermore, while there is a growing body of research acknowledging that reefs of the future will likely lack historic analogs (Hughes et al., 2017), there is still a focus on hard corals as the sole foundation taxon. A better understanding of the ecosystem functions and services provided by reefs that have lost significant hard coral and may be dominated by other species (e.g., macroalgae, sponges, or gorgonians) will enable more effective management into the future.

Here we examine 10 ecosystem metrics, relating to different ecosystem functions and services, under varying levels of coral cover at 328 sites across the Caribbean. We use data collected by the Atlantic and Gulf Rapid Reef Assessment (AGRRA) between 2011 and 2015 at forereef sites that currently have or historically have had abundant hard coral (AGRRA, 2018). In contrast to coral reef research that tends to focus on mean responses to coral loss, we assess the range and variability in these metrics across reefs with low (<10%), intermediate (10%–20%), and high (>20%) coral cover. Few sites in our dataset have coral cover higher than 30%, thus we recognize that our “high” cover category is modest relative to many historical Caribbean reefs. Nevertheless, within the range of coral cover observed in the present‐day Caribbean, we test whether lower coral cover results in declining means and/or changes to the variability of ecosystem metrics among reefs.

Based on a prevailing paradigm in the coral reef literature, our expectation was that various ecosystem metrics would decline in mean value on low coral cover reefs, but we did not have strong a priori predictions for changes in variance. In contrast to these expectations, our results suggest that “bright spots”—sites where these metrics are comparatively high—can occur where coral cover is low. Given that Caribbean reefs have experienced earlier and often greater coral loss than other coral reef systems such as those in the Indo‐Pacific (Jackson, 1997; Roff & Mumby, 2012), it may be a harbinger of changes to come for other reefs, or coastal systems in general. Therefore, lessons learned in the Caribbean could facilitate greater preservation of bright spots in other regions or ecosystems. As the Earth faces accelerated ecosystem change, there is a need to explore the variability in ecosystem services provided by altered ecosystems. Insights arising from such efforts may allow management to enhance bright spots and limit loss of key services.

2 MATERIALS AND METHODS

2.1 Ecosystem metrics

There are a wide variety of ecosystem processes on coral reefs that contribute to the maintenance of biodiversity and the delivery of benefits to people (i.e., ecosystem services; reviewed by Moberg & Folke, 1999; Mumby et al., 2008; Yee et al., 2015). Services are rarely measured directly, but we identify metrics that can be calculated from available survey data and are related to a number of important coral reef ecological functions and ecosystem services (Table 1; Figure S1). These metrics are incomplete proxies for the functions and services we are aiming to assess (e.g., fish species richness as one attribute of value to SCUBA and snorkel tourists) and most of them are related to more than one important ecological function or ecosystem service (Figure S1). Nevertheless, collectively they provide a useful if imperfect indicator for potential ecosystem services and functions at a site. These metrics include:

  1. vertical relief, a component of ecosystem structure underlying a variety of services including habitat provision and predation refuge for fishes (Graham & Nash, 2013) and coastal protection (Ferrario et al., 2014);
  2. calcareous macroalgal cover, a key contributor to sand and sediment production including for adjacent beaches (Lidz & Hallock, 2000);
  3. coral recruit density, a proxy for coral recruitment rates, and also important for reef recovery potential (Hughes et al., 2019);
  4. coral calcification rate, which can contribute to growth of reef structure, sand production, coastal protection, and reef recovery potential (Ferrario et al., 2014; González‐Barrios & Álvarez‐Filip, 2018; Perry et al., 2013);
  5. mean coral species richness, which captures components of biodiversity as an endpoint unto itself and as an important supporting service (Moberg & Folke, 1999), including facilitating coral growth and survivorship (Clements & Hay, 2019) and influencing the value of dive tourism (Uyarra, Watkinson, & Côté, 2009);
  6. herbivorous fish biomass, as a partial indicator for herbivory, because herbivores, by suppressing macroalgal overgrowth of the reef, can facilitate coral recruitment, growth, and survivorship (Hughes et al., 2007; Mumby, 2006) and improve the aesthetic value of the reef;
  7. potential fishery value, assessed as the potential market value of fish on surveyed reefs, as one key component of benefits from fisheries (Mumby et al., 2008);
  8. density of large fish, which contributes to the value of dive tourism (Gill et al., 2015) and may increase fishery value as fishers often target and/or receive higher prices for larger fish (Graham, Dulvy, Jennings, & Polunin, 2005);
  9. parrotfish bioerosion rate, which is positively related to sand production and negatively related to net reef accretion (Hubbard & Dullo, 2016; Perry et al., 2013); and
  10. mean fish species richness, which captures components of biodiversity as an endpoint unto itself and acts as an important supporting service (Moberg & Folke, 1999), including enhancing reef resilience (Burkepile & Hay, 2008), influencing the value of dive tourism and the aesthetic value of the reef (Williams & Polunin, 2002) and supporting fisheries productivity (Duffy, Lefcheck, Stuart‐Smith, Navarrete, & Edgar, 2016).
TABLE 1.
Ecosystem metrics examined in this study. All data from AGRRA () unless otherwise specified
Metric Description
Vertical relief Measured as the difference, in cm, between the highest structure (coral or reef rock) and the lowest point in the underlying substratum within a 1m diameter (measured every 5 m and averaged over a 30 m transect)
Calcareous macroalgae cover Percent cover of erect calcareous macroalgae (e.g., Halimeda spp., Galaxaura spp., etc.)
Coral recruit density In individuals/m2, counting all hard coral colonies less than 4 cm in diameter (identified to genus)
Coral calcification rate In kg CaCO3 m−2 year−1, applying species‐specific calcification rates from González‐Barrios and Álvarez‐Filip (2018) to coral cover data from AGRRA surveys
Coral species richness The number of coral species encountered on each transect (10 m2), with transect densities averaged over a site
Herbivore biomass In g/m2, using species‐specific fish biomass data from AGRRA surveys and summing biomass estimates across all fish species known principally to be herbivores (as classified by Paddack et al., 2009; see their table S2)
Potential fishery value In USD/m2, using species‐specific fish biomass data from AGRRA surveys and multiplying those biomass numbers by species‐specific price estimates (shown in Table S1)
Density of large fish In individuals/m2, counting all fish greater than 40 cm in size
Parrotfish bioerosion rate In kg m−2 year−1, applying species‐ and size‐specific parrotfish bioerosion rates from Ruttenberg et al. (2019) to parrotfish recorded in AGRRA surveys
Fish species richness The number of fish species encountered on each transect (60 m2), with transect densities averaged over a site
  • Abbreviation: AGRRA, Atlantic and Gulf Rapid Reef Assessment.

2.2 Dataset description

We used reef monitoring data from the AGRRA version 5 surveys, collected between 2011 and 2015 (AGRRA, 2018). We chose this dataset because it is relatively comprehensive in measuring a range of components of reef ecosystems. It also has broad coverage across the Caribbean surveying consistent reef habitats (e.g., forereef), and data are collected by scientists using a standardized protocol (available at www.agrra.org). Data were downloaded from the AGRRA website and were filtered to include only sites in forereef habitats between 3 and 20 m in depth that were surveyed with at least six fish transects and at least four benthic transects. This resulted in 328 sites across the Caribbean that were sampled between 2011 and 2015 (Figure 1).

image

Map of Atlantic and Gulf Rapid Reef Assessment survey sites included in this analysis and histograms of mean site‐level percent live coral cover within each ecoregion

The Atlantic and Gulf Rapid Reef Assessment surveys benthic organisms (crustose coralline algae [CCA], macroalgae, corals, sessile invertebrates) and ecologically and/or commercially important reef fish species, and documents attributes of reef condition (e.g., reef structure, coral health). Fish and coral taxa are identified to species (or occasionally to genera), while other components are grouped at higher taxonomic levels (e.g., sponges), with 93 fish taxa and 76 coral taxa reported across all surveys analyzed in this study. Coral recruits/juveniles (i.e., individuals <4 cm diameter) are counted and reef vertical relief (i.e., vertical height of the tallest coral or reef rock above the lowest point in the underlying substratum) is measured.

To reduce the influence of extreme outliers in the fish biomass data from AGRRA, we identified species that had individual observations (i.e., a single species, size and count record on an individual transect) greater than the 99th biomass quantile for all fish observations (n  = 54 species). For these species, we capped the biomass of each observation to the 99th quantile from all individual observations for that species in the entire dataset (Donovan et al., 2018). This led to our capping approximately 0.7% (660 out of 96,147) of the observations. These observations were not removed but were adjusted so that they are still large values without influencing the site averages as strongly as they would otherwise. In particular, this reduced the influence of data input errors and of rare, large groups of roaming or aggregating species, such as Sphyraena barracuda , which can constitute a large proportion of biomass in certain transects while not representing individuals that are permanent residents of the surveyed reef. We also calculated macroalgal cover (for Figure S5) for the AGRRA data by summing percent cover across calcareous macroalgae, fleshy macroalgae, and mixed macroalgae. Coral species richness and fish species richness were computed as species richness per transect, with transects measuring corals and fishes over a standardized area (10 and 60 m2, respectively). Transect richness values were then averaged for each site. This procedure standardized richness estimates to a fixed sampling effort and was necessary because sites varied in the number of transects surveyed.

To estimate parrotfish bioerosion rates, we applied species‐ and size‐specific estimates of bioerosion rates from Ruttenberg, Adam, Duran, and Burkepile (2019) to the AGRRA parrotfish observations at each site. These estimates were derived from field behavioral observations of 10 parrotfish species in the Florida Keys, United States, with the bioerosion rate for each fish of a given size class calculated as bite rate multiplied by bite volume, multiplied by the probability of leaving a scar, and by bulk density of reef carbonate, with the latter terms derived from the literature (Ruttenberg et al., 2019).

To estimate coral calcification rates, we applied species‐specific estimates of mean calcification rates (in kg CaCO3 m−2 year−1) from González‐Barrios and Álvarez‐Filip (2018; see their table 2) to mean cover estimates (in m2) from AGRRA. González‐Barrios and Álvarez‐Filip (2018) estimated calcification rates for the 47 most common scleractinian coral species in the Caribbean, covering most of the corals observed in the AGRRA coral dataset. Occasionally corals were recorded at the genus‐level in AGRRA; in such case we calculated an average calcification rate across all species in that genus (González‐Barrios & Álvarez‐Filip, 2018). The few species of corals present in AGRRA for which we did not have calcification estimates (<0.003 percent cover) were excluded from this analysis.

To determine the fishery value of the reef fish assemblages measured at the AGRRA sites, we used fishery statistics from Puerto Rico that provided an estimate of value in USD per pound of fish landed by species or by family. We extracted annual, island‐wide, ex‐vessel values for all species and families for which data were available from the period 2007 to 2011 from Matos‐Caraballo (2012). Because some groups were not valued in all years, we used the maximum annual value for each species (or for each family for species that lacked species‐specific values). We did not include species or families that did not appear in this dataset (Matos‐Caraballo, 2012); some of these omitted species are occasionally caught, but they have little or no value and are considered “trash fish” in Puerto Rico. While some of these species may have value on other Caribbean islands, the value is likely low, particularly since most of these species are small or unpalatable. We then used these value data (Table S1) to calculate the potential fishery value of the fish assemblage at each AGRRA site based on the biomass of each species. Ideally, we would have had price data from local and/or export markets in each location, but such data are non‐existent or buried in grey literature across the Caribbean. Puerto Rico is centrally located in the Caribbean, has an extensive and diverse reef fish fishery, and is unique in having a well‐documented time series of catch and value data.

2.3 Analyses

We examined whether lower coral cover results in declining means and/or changes to the variability of ecosystem metrics among reefs (Figure 2). We calculated 10 metrics related to different aspects of reef functions and services (Figure S1), using the AGRRA data described above (Table 1). For many analyses, we binned sites into three categories of live coral cover: (a) low: <10% cover; (b) medium: 10%–20% cover; and (c) high: >20% cover. These categories were defined based on the distribution of coral cover across all of the sites in the dataset (Figure 1); although 20% is still quite low relative to historical coral cover for many sites in the region, mean coral cover across all our sites was 14.36% (n  = 328; ±10%), and about a quarter of the sites (n  = 79) have coral cover greater than 20%. We confirmed the robustness of our results to these categories by examining patterns using an alternative set of coral cover thresholds (Figure S7). Our qualitative conclusion—that most metrics can be high at sites with low coral cover—did not appear to be sensitive to the coral cover categories that we used. We also show the relationships between each metric and coral cover on a continuous scale (Figure S8). We do not focus our analysis around the continuous results because we were most interested in examining variability in metrics across different levels of coral cover rather than trying to identify predictive relationships. Furthermore, coral cover explained a very low percentage of the variability for most of these metrics (Figure S8).

image

A priori hypotheses about different ecosystem patterns observed on high and low coral cover reefs. (a)–(d) Shifted distributions relative to a high‐coral state (gray). At low coral cover, relative values may have equal mean and increased variance (a; blue), decreased mean and equal variance (b; red), decreased mean and decreased variance (c; yellow), or decreased mean and increased variance (d; green). (e) Distributions from (a) to (d) as boxplots. These alternative hypotheses are not a complete set, as some services could also potentially increase with declining coral cover. Furthermore, we represent hypotheses based on comparisons of high and low coral cover for simplicity, although our empirical analyses focused on three coral cover categories to better capture variation in coral cover across the Caribbean

We also binned sites by low (<40 cm), medium (40–80 cm), and high (>80 cm) vertical relief. These categories were defined based on the distribution of vertical reef relief across all sites in our dataset. Using these definitions, about a third of the sites had low vertical relief and a little less than a quarter had high relief. These categories allowed us to evaluate, for example, whether relict structure from dead coral is associated with higher ecosystem metrics at sites with low live coral cover. We categorized sites based on protection status using the information provided in AGRRA for a subset of the sites (n  = 219) that noted whether surveys were conducted within marine protected areas (MPAs) to evaluate if protection status could confound our results. Lastly, we examined patterns separately for just the Western Caribbean and the Bahamian ecoregions sampled in 2011. These represented the two region‐year combinations for which we had the largest sample sizes, allowing us to evaluate the role of coral cover in the absence of regional and interannual variability. For example, we did not test for an association of the response variables with physical gradients like sea surface temperature or productivity; examining patterns at the ecoregional scale should partially control for variation in these physical factors.

We evaluated how each of the 10 metrics (Figure 3; Table 2) was associated with a variety of predictor variables using linear mixed effects models (LMERs) in the R package “lme4” (Bates, Maechler, Bolker, & Walker, 2015). We assumed a Gaussian family (normal distribution) for response variables in all models. For data that needed to be transformed to meet assumptions of normality (all variables except density of large fish, fish species richness, and coral species richness), we used an inverse hyperbolic sine transformation. Estimates from the linear model were then back‐transformed in order to calculate effect sizes. We treated ecoregion and year as categorical random effects, depth as a continuous fixed effect, and coral cover category (low, medium, high) as a categorical fixed effect. For the LMER examining MPA status (Figure S2), MPA status (yes or no) was included as a fixed effect in addition to the effects included in the main model (i.e., ecoregion, year, depth, coral cover). For the LMERs focused on individual ecoregions sampled in a single year (Figures S3 and S4), we removed ecoregion and year as random effects.

image

Distribution of 10 metrics, by three categories of mean percent live coral cover. High coral cover is defined as >20%, medium is 10%–20%, and low is <10%. Data from AGRRA (2018; number of sites for high, medium, low coral cover: a–g, i, j, n  = 79, 120, 129; h, n  = 65, 104, 103). Lines represent individual data points. P‐ value from linear mixed effects model results (Table 2) shown in parentheses after panel title, with results of Tukey’s post hoc test shown by letters next to boxes (coral categories sharing a letter are not significantly different from one another)

TABLE 2.
Results of linear mixed effects models (LMER) and generalized linear mixed effect models (GLMER) testing the effect of coral cover categories (low, medium, high) on the 10 metrics; and results of chi‐squared tests examining whether sites in the 90th or 75th percentile for each metric are distributed in coral cover categories proportional to the number of sites in each category. Shading corresponds to significance level (blue) and effect size (green)

Subscribe to the newsletter news

We hate SPAM and promise to keep your email address safe

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Top Stories

To Top