Disturbances drive changes in coral community assemblages and coral calci ﬁ cation capacity

. Anthropogenic environmental change has increased coral reef disturbance regimes in recent decades, altering the structure and function of many coral reefs globally. In this study, we used coral community survey data collected from 1996 to 2015 to evaluate reef-scale coral calci ﬁ cation capacity (CCC) dynamics with respect to recorded pulse disturbances for 121 reef sites in the Main Hawaiian Islands and Mo'orea (French Polynesia) in the Paci ﬁ c and the Florida Keys Reef Tract and St. John (U.S. Virgin Islands) in the western Atlantic. CCC remained relatively high in the Main Hawaiian Islands in the absence of recorded widespread disturbances; declined and subsequently recovered in Mo'orea following a crown-of-thorns sea star outbreak, coral bleaching, and major cyclone; decreased and remained low following coral bleaching in the Florida Keys Reef Tract; and decreased following coral bleaching and disease in St. John. Individual coral taxa have variable calci ﬁ cation rates and susceptibility to disturbances because of their differing life-history strategies. As a result, temporal changes in CCC in this study were driven by shifts in both overall coral cover and coral community composition. Analysis of our results considering coral life-history strategies showed that weedy corals generally increased their contributions to CCC over time while the contribution of competitive corals decreased. Shifts in contributions by stress-tolerant and generalist corals to CCC were more variable across regions. The increasing frequency and intensity of disturbances under 21st century global change therefore has the potential to drive lower and more variable CCC because of the increasing dominance of weedy and some stress-tolerant corals.


INTRODUCTION
The growth of calcium carbonate (CaCO 3 ) structures is one of the most important functions of coral reefs in providing the ecosystem services that reefs afford to humanity (Kleypas et al. 2001, Edmunds et al. 2016, Courtney et al. 2017, Perry et al. 2018. Scleractinian corals typically account for the majority of CaCO 3 production on coral reefs (Montaggioni and Braithwaite 2009) and are therefore critical to the provisioning of shoreline protection, fisheries habitat, and tourism revenue services that coral reefs provide (Moberg and Folke 1999). However, the maintenance of these services is threatened by global and local environmental changes that have led to pantropical declines in coral cover (Gardner et al. 2003, Bruno and Selig 2007, Jackson et al. 2014) and shifts in coral community composition (Alvarez-Filip et al. 2013, Perry et al. 2015), which will likely prevent some coral reefs around the world from keeping up with 21st century sea-level rise (Perry et al. 2018).
The frequency and intensity of disturbances shape coral community structure (Connell 1978) with enhanced disturbance regimes typically driving a disproportionate loss of fast-growing, architecturally complex corals (e.g., Acropora spp.) in favor of slow-growing, massive (e.g., Porites lobata, Siderastrea siderea) and/or fastgrowing, early-successional colonizers (e.g., Agaricia spp.; Porites astreoides; Pocillopora damicornis; e.g., Loya et al. 2001, McClanahan and Maina 2003, McClanahan et al. 2009, Fabricius et al. 2011, van Woesik et al. 2011, Darling et al. 2013, Grottoli et al. 2014). These differences in coral traits have been further analyzed in the context of hypothesized coral life-history strategies, which describe consistent traits that characterize an evolutionary survival strategy (Grime 1977, Darling et al. 2012). The pioneering work by Darling et al. (2012) analyzed pantropical coral trait data to define four primary coral life-history strategies: (1) competitive: fast-growing, large-colony, broadcast spawning corals with branching/plating morphologies; (2) stress-tolerant: slow-growing, long-lived, highly fecund, broadcast spawning corals with massive/encrusting morphologies; (3) weedy: small-colony, brooding corals; and (4) generalists: corals with traits representing a mixture of those found in the previous three groups (e.g., stress tolerance and moderate growth rates). Shifts from reefs dominated by competitive to stress-tolerant and weedy corals can drive declines in reef-scale calcification as evidenced by modeled (Alvarez-Filip et al. 2013) and observed (Perry et al. 2015, Kuffner and Toth 2016, Lange and Perry 2019 decreases in net coral reef CaCO 3 production budgets; however, the impacts of shifting coral life-history strategies on reef-scale calcification through time and across broad geographic spatial scales remain to be rigorously characterized. In this study, we quantified coral calcification capacity (CCC) across a variety of disturbance regimes using a life-history strategies perspective (sensu Darling et al. 2012) for 121 reef sites surveyed from~1996 to 2015 across four focal regions in the Pacific (Main Hawaiian Islands and Mo'orea) and western Atlantic (Florida Keys Reef Tract and St. John). We tested the hypothesis that disturbances drive decreases in the contribution to CCC by competitive corals and increases in the contribution to CCC by weedy and stresstolerant corals owing to alterations of coral community composition as coral cover declines. Our results reveal how disturbance-driven shifts in the species composition of coral communities and overall coral cover have altered reef CCC trajectories at our sites in the Pacific and western Atlantic and how these changes may impact the critical reef-building function of coral reefs now and in the future.

Introduction to focal-region surveys and disturbance histories
The four focal regions analyzed in this study were selected because of publicly available datasets (Guest et al. 2018b) of coral reef benthic surveys (i.e., total and taxon-specific percent coral cover) that were conducted on a range of reef types (e.g., patch, spur and groove, fringing, barrier, and forereef) across a range of depths from 1 to 22 m. The surveys were conducted using standardized protocols at a total of 121 sites surveyed a minimum of three times over at least ten years from~1996 to 2015 (see Guest et al. 2018a supplementary materials for a complete discussion of sites and survey methods).The data from the focal regions overlap from 2005 to 2014, but the total length of the time series varied slightly, with the Main Hawaiian Islands survey data ranging from 1999 to 2014, Mo'orea from 2005, Florida Keys Reef Tract from 1996, and St. John from 1999 In addition to data availability, the focal regions were chosen to include a broad range of disturbance histories during the survey period (see Guest et al. 2018a; Fig. 1). Although we acknowledge that long-term press disturbances such as land-use change, fishing pressure, eutrophication, and ocean acidification can contribute to ecosystem change, inconsistencies in data availability and the indirect nature of press disturbances (i.e., they do not directly cause coral mortality) led us instead to focus on documented pulse disturbances in this study. All focal regions experienced recorded coral bleaching events during the study periods including 2002, 2004, and 2014 in the Main Hawaiian Islands (Jokiel and Brown 2004, Bahr et al. 2017, Rodgers et al. 2017; 2007 in Mo'orea (Adjeroud et al. 2018);1997, 1998, 2011 in the Florida Keys (Manzello et al. 2007, Wagner et al. 2010, Ruzicka et al. 2013; and 2005 in St. John (Miller et al. 2006, Edmunds 2013. While the before and after effects of hurricanes were not directly quantified for each site in this study, we considered hurricanes to potentially impact the focal regions if a Category 1-5 storm passed within 100 nautical miles of the following locations during the respective time series of the focal regions in this study: Hilo Harbor, Kahului Harbor, Honolulu Harbor, or Kal aheo (Main Hawaiian Islands); Pao Pao (Mo'orea); Big Pine Key or Dry Tortugas (Florida Keys Reef Tract); Fig. 1. Mean annual coral calcification capacity (CCC) over time partitioned by the dominant calcifiers relative to recorded disturbance events. Mean (AE95% confidence interval) annual CCC (kg CaCO 3 Ám À2 Áyr À1 ) is reported for scleractinian coral and hydrozoan taxa contributing ≥5% CCC across all reef sites within a focal region for at least one year of the time series with all remaining taxa pooled into the "Other" category. Note the differing scales on the vertical axis for each panel. Disturbances and icons in the legend are Bleaching, severe coral bleaching event; Disease, coral disease event; COTS, crown-of-thorns sea star outbreak; Cold, cold-water mortality event; and Hurricane, n number of recorded hurricanes/cyclones for the given year. Icons are courtesy of the Integration and Application Network, University of Maryland Center for Environmental Science (ian.umces.edu/ symbols/). and Virgin Islands National Park (St. John). Based on these search criteria in the NOAA hurricanes database (https://coast.noaa.gov/hurrica nes/), there were two recorded hurricanes in the Main Hawaiian Islands, no recorded cyclones in the search region for Mo'orea (but waves from Cyclone Oli impacted some reefs in 2010; Han et al. 2016, Adjeroud et al. 2018, seven recorded hurricanes in the Florida Keys Reef Tract, and seven recorded hurricanes in St. John (Fig. 1). Mo'orea additionally experienced a crown-ofthorns sea star (COTS), Acanthaster planci, outbreak from 2006 to 2010 (Adam et al. 2011, Pratchett et al. 2011, Han et al. 2016, Adjeroud et al. 2018. The Florida Keys Reef Tract also experienced a cold-water coral mortality event due to anomalously cold winter temperatures in 2010 (Kemp et al. 2011, Lirman et al. 2011, Ruzicka et al. 2013). In St. John, a prolonged period of coral disease occurred from 2005 to 2007 following the 2005 coral bleaching event (Miller et al. 2006, Edmunds 2013. As a result of these extensive, yet variable disturbance histories and dissimilar initial coral community compositions across sites, there is a large variation in coral cover and community compositions over time across the focal regions, which allowed us to evaluate the community-level drivers of CCC for a range of coral reefs across ocean basins between~1996 and 2015.

Coral calcification capacity
We define CCC as an estimate of the annual reef-scale calcification rate (kg CaCO 3 Ám À2 Áyr À1 ) by a coral community. This metric is similar to the gross carbonate production terms calculated as part of complete CaCO 3 budgets, but these budgets also include CaCO 3 dissolution, bioerosion, and import/export (e.g., Chave et al. 1972, Stearn et al. 1977, Hubbard et al. 1990, Kleypas et al. 2001, Perry et al. 2012. Additionally, CCC does not include calcification by coralline algae, Halimeda, and benthic foraminifera, each of which have been measured to produce upwards of 2 kg CaCO 3 Ám À2 Áyr À1 (e.g., see Montaggioni and Braithwaite 2009 and references therein) and may contribute greater proportions of reef-scale CaCO 3 production in lower coral cover and/or bleached reef systems (Courtney et al. 2018). While these other calcifiers can be significant components of reef CaCO 3 budgets, the usage of CCC as an ecologically meaningful metric is supported by previous studies documenting that corals are typically the dominant coral reef CaCO 3 producers (Montaggioni and Braithwaite 2009), by observations of greater net calcification by hard corals relative to calcifying algae on a planar-area basis (Kuffner et al. 2013, Kennedy et al. 2017, and by the high degree of agreement between coral colony and reef-scale calcification (Courtney et al. 2016(Courtney et al. , 2017. Our measures of CCC utilized the best available calcification rate data from the literature because it was logistically not possible to directly quantify the local variability in calcification rates, which can be considerable (Kuffner et al. 2013), owing to genetic variability and the environmental drivers of coral calcification (e.g., Pratchett et al. 2015) over the broad spatial and temporal scales of this study. The application of literature-based rather than direct measurements of location-specific calcification rates is a common limitation of complete census-based CaCO 3 budget calculations but is necessary to produce rapid assessments of census-based reef-scale calcification (e.g., see Perry et al. 2012). Nonetheless, the CCC methodologies presented here allow for robust, retrospective analysis of the capacity for annual coral community calcification and provide valuable new insights into how shifting coral communities impact reef-scale calcification from data-limited historical time series.
To calculate temporal trends in CCC for our focal regions, we first estimated annual calcification rates of individual scleractinian coral and hydrozoan Millepora taxa (hereafter collectively referred to as "corals") as the product of published taxon-specific linear extension rates (cmÁyr À1 ), published taxon-specific skeletal densities (g CaCO 3 Ácm À3 ), and a growth form adjustment factor that estimates the void space between topographic features generated by complex (i.e., branching, plating, corymbose, digitate, columnar, foliose, sub-massive) morphologies following established methods (Morgan and Kench 2012, Perry et al. 2012, Guest et al. 2018a). All extension, density, growth form, and calcification rate data and references are summarized in Guest et al. (2018a : Tables S2 and S3). The contribution of each species to annual, reef-scale calcification was then determined by multiplying taxon-specific calcification rates (kg ❖ www.esajournals.org CaCO 3 Ám À2 Áyr À1 ) by their respective planar benthic cover (%) after Perry et al. (2012) and Guest et al. (2018a). Reef-wide calcification rates were calculated for each species where species-level benthic cover data were available (i.e., Main Hawaiian Islands, Florida Keys Reef Tract, St. John) and by genus for Mo'orea, where benthic cover was surveyed at the genus level. Annual CCC (kg CaCO 3 Ám À2 Áyr À1 ) for each reef site was then determined by summing the calcification (kg CaCO 3 Ám À2 Áyr À1 ) by all coral taxa. The average of CCC across all surveyed sites within a focal region for each given year was calculated to determine the changes in mean annual CCC through time. It is important to note, however, that not every reef site in the Main Hawaiian Islands focal region was surveyed each year in this study primarily due to funding limitations and this precludes any robust conclusions about changes in interannual CCC for this focal region. Guest et al. (2018a) previously demonstrated that median z-score of coral cover explained 79-87% of the variance in the median CCC within each of our focal regions. Here, we hypothesized that the remaining variance may be due to differences in coral community composition (i.e., CCC is the product of taxa-level coral cover and calcification rates summed for all taxa). We parsed the contribution of each coral taxon to mean annual CCC in each focal region to identify the dominant calcifying taxa, which were defined in this study as taxa that contributed at least an average of 5% to the total CCC across all sites in each focal region for the duration of the time series. To further visualize how the differing calcification rates by the dominant corals have the capacity to impact the variance in CCC within each focal region, hypothetical scaling of calcification rates for each of the dominant calcifying corals from 0% to 100% cover was superimposed on plots relating percent coral cover and CCC (Fig. 2). In effect, these hypothetical calcification trajectories provide a frame of reference for interpreting CCC of the communities at each reef site and year relative to calcification rates of the dominant calcifying corals for the given focal region.

Dominant calcifying corals and life-history strategies
The dominant calcifying corals were then characterized as competitive, stress-tolerant, weedy, or generalist sensu Darling et al. (2012) to evaluate whether shifts in the relative contributions of corals with different life-history strategies are predictive of changes in CCC over the time series in each of the focal regions of this study.

Statistical analyses
First, we tested whether differences in coral community composition between the focal regions significantly (P < 0.05) affected the relationship between percent coral cover and CCC for our focal regions (i.e., to test for significant differences in mean coral community calcification rates between focal regions). Linear mixed effects models were constructed to test whether slopes of CCC vs. coral cover were significantly different for each focal region. Percent coral cover was used as a fixed effect with random slopes and intercepts for each site to account for the fact that each reef site has varying coral community composition and therefore varying calcification rates for a given percent coral cover throughout the respective time series. Statistical models were fit using the R (R Core Team 2017) package nlme (Pinheiro et al. 2017) with model parameters estimated by maximum likelihood.
Second, we tested whether contributions by the dominant calcifying taxa to CCC shifted during the time series in each of the focal regions and whether these changes match the theoretical expectations for the four coral life-history strategies following disturbances (Darling et al. 2012(Darling et al. , 2013. To accomplish this task of evaluating longterm changes in the percent contribution of calcifying taxa to CCC, linear mixed effects models were constructed to test whether year was a significant (P < 0.05) predictor of percent contribution to CCC by each of the dominant calcifying taxa. The slopes of these models therefore indicate whether there were statistically significant mean increases (positive slope) or decreases (negative slope) in the contribution of the dominant calcifying corals to CCC for the duration of the time series for each of the focal regions. The models used the percentage of annual site-level CCC by each dominant calcifying genus as the response variable and included fixed effects for year and a random intercept and slope for each site using the R (R Core Team 2017) package nlme (Pinheiro et al. 2017) with model parameters estimated by maximum likelihood. Random ❖ www.esajournals.org intercepts and slopes for each site were included to account for the fact that each site had different initial percent contributions of dominant calcifying coral to CCC (random intercepts) and that each site had different disturbance histories resulting in variable responses in dominant calcifying coral contributions to CCC over time (random slopes). Because sites were repeatedly surveyed through time, a continuous autocorrelation structure (corCAR1 in package nlme; Pinheiro et al. 2017) was included in each model to account for temporal autocorrelation and the variable time interval between surveys in the Main Hawaiian Islands. While temporal fluctuations in CCC may be nonlinear at least for some reef sites and focal regions on interannual timescales (e.g., Fig. 1B), we utilized linear mixed effects models in this study because the purpose of this analysis was to focus on longer term mean trends in coral contribution to CCC. Notably, this focus on longer timescales prevents the inconsistent interannual sampling effort for the Main Hawaiian Islands from biasing conclusions regarding mean trends in coral contribution to CCC through time.

RESULTS
Mean of focal-region CCC (mean AE 95% confidence intervals) through time was generally higher for the Pacific regions than in the western Atlantic (Fig. 1). CCC for the Main Hawaiian Islands was consistently the highest of the four focal regions although the large interannual variability resulting from non-uniform sampling efforts (i.e., not all sites were surveyed in all years) and relatively larger, overlapping confidence intervals (Fig. 1A) preclude any formal conclusions about interannual changes in mean CCC in this location. In Mo'orea, mean CCC (kg CaCO 3 Ám À2 Áyr À1 AE 95% CI) declined following a combination of a crown-of-thorns outbreak (2006)(2007)(2008)(2009)(2010), coral bleaching (2007), and Cyclone Oli (2010) from 4.8 AE 0.5 in 2006 to a minimum of 1.5 AE 0.8 in 2012 and then increased to 2.5 AE 0.8 in 2015 (Fig. 1B). In the Florida Keys, mean CCC (kg CaCO 3 Ám À2 Áyr À1 AE 95% CI) decreased following coral bleaching in 1997/1998 from 1.7 AE 0.4 in 1996 to 1.0 AE 0.3 in 2000 and remained relatively stable through 2015 (Fig. 1C). Mean CCC (kg CaCO 3 Ám À2 Á yr À1 AE 95% confidence intervals) in St. John increased slightly from 0.6 AE 0.5 in 1999 to 1.1 AE 0.6 in 2005, decreased to 0.5 AE 0.3 in concert with the 2005 coral bleaching event and associated 2005-2007 disease outbreak, and remained approximately stable thereafter (Fig. 1D).
Twelve species representing nine genera of corals out of the 84 species and 50 genera surveyed in this study were identified as the dominant calcifying taxa (i.e., calcified mean ≥5% of total CCC across all sites and all years for the respective time series) across the four focal regions with Acropora, Montipora, Pavona, Pocillopora, and Porites spp. contributing the most to CCC in the Pacific and Acropora, Millepora, Montastraea, Orbicella, Porites, and Siderastrea spp. in the western Atlantic (Fig. 1, Table 1). The dominant calcifying species were further characterized by life-history strategies sensu Darling et al. (2012) to evaluate the respective changes of the dominant calcifying corals with respect to life-history strategies (Table 1).
Although CCC generally increases with increasing coral cover (Guest et al. 2018a), the remaining variability in that relationship is a direct result of the differing calcification rates of the dominant calcifiers for each reef site at each point in time (Fig. 2; Guest et al. 2018a). In the Main Hawaiian Islands, which had sites with the highest coral cover out of any of the focal regions, CCC vs. coral cover data for each site surveyed in each year generally fall along the hypothetical trajectories predicted for Montipora spp. or Porites lobata corals ( Fig. 2A) with a mean (AE95% CI) CCC vs. coral cover slope of 0.19 AE 0.02 kg CaCO 3 Ám À2 Áyr À1 Á% coral cover À1 (P < 0.001, Fig. 3). Coral cover in Mo'orea was also high relative to most western Atlantic reef sites. However, whereas the annual site-level CCC vs. coral cover data for some sites in Mo'orea appear to follow the hypothetical trajectory predicted for a mixed Montipora/Acropora community, most sites appear to follow the Porites calcification trajectory (Fig. 2B) with a mean (AE95% CI) CCC vs. coral cover slope of 0.14 AE 0.02 kg CaCO 3 Ám À2 Áy À1 Á% coral cover À1 (P < 0.001, Fig. 3). In the Florida Keys Reef Tract, where most reef sites have comparatively lower coral cover, the annual site-level CCC vs. coral cover data follow the hypothetical trajectories predicted for the Orbicella annularis complex/ Montastraea cavernosa or Acropora palmata (Fig. 2C) with a mean (AE95% CI) CCC vs. coral cover slope of 0.17 AE 0.01 kg CaCO 3 Ám À2 Áy À1 Á% coral cover À1 (P < 0.001, Fig. 3). Lastly, in St. John the site-level CCC vs. coral cover data suggest that nearly every site follows the hypothetical trajectories predicted for the Orbicella annularis complex/Montastraea cavernosa (Fig. 2D) with a mean (95% CI) slope of 0.11 AE 0.01 kg CaCO 3 Ám À2 Áy À1 Á% coral cover À1 (P < 0.001, Fig. 3) owing to the dominance of the Orbicella annularis complex in this focal region (Fig. 1D).
To assess whether changes in the contribution by each of the dominant calcifying taxa to CCC were consistent with life-history strategy expectations of shifting coral cover following disturbances (Darling et al. 2012(Darling et al. , 2013, long-term changes of percent contribution by each of the dominant calcifying corals to total CCC across all sites for the duration of the time series were evaluated. In this analysis, positive slopes indicate mean increases in the percent contribution of the respective coral to CCC within the focal region over the time series whereas the opposite is true for corals with negative slopes and non-significant slopes indicate no detectable change in coral contribution to CCC (Fig. 4

DISCUSSION
Here, we have used CCC to retroactively estimate CaCO 3 production by coral communities in historical datasets (Guest et al. 2018a) to explore the ecological drivers of CCC across the Pacific and western Atlantic over an approximately 20yr period. Our results suggest that coral bleaching (Mo'orea, Florida Keys, St. John), coral disease (St. John), and a crown-of-thorns sea star outbreak (Mo'orea) were the dominant pulse disturbances that drove reductions in CCC over this period through a combination of reduced coral cover and shifting coral community compositions (Figs. 1, 2, 4). In the Main Hawaiian Islands, the absence of significant, widespread recorded pulse disturbances (e.g., the 2002 and 2004 bleaching events were minor and the reefs recovered; Bahr et al. 2017) and the large interannual CCC variability with overlapping 95% confidence intervals preclude any similar conclusions about the role of disturbance in modulating CCC there. However, widespread bleaching in Hawai'i was recorded in 2014 and 2015 at the conclusion of the time series in this study (Bahr et al. 2017, Rodgers et al. 2017, and it caused reduced reef-scale calcification observed for at least K ane'ohe Bay in the Main Hawaiian Islands (Courtney et al. 2018). These findings agree with the growing CaCO 3 budget and † LHS for species from the respective genera described in Darling et al. (2012). ‡ LHS for all Montipora spp. from Darling et al. (2012). § LHS inferred from Brown and Edmunds (2013). ¶ LHS for all species of Orbicella annularis complex described in Darling et al. (2012).
chemistry-based net ecosystem calcification literature showing that coral bleaching events and other disturbances can significantly reduce reefscale calcification rates (e.g., Kayanne et al. 2005, Perry et al. 2008, DeCarlo et al. 2017, Januchowski-Hartley et al. 2017, Perry and Morgan 2017, Courtney et al. 2018, Lange and Perry 2019. Mean CCC in Mo'orea showed signs of recovery following disturbance-induced CCC declines (Fig. 1B). Analogous recoveries in reefscale calcification have been observed in the Seychelles (Januchowski-Hartley et al. 2017) and K ane'ohe Bay, Hawai'i (Courtney et al. 2018), suggesting reefs have the capacity to recover CaCO 3 production rates in the absence of continued disturbances. However, it is important to note that reduced CCC rates for many of the lower percent coral cover reefs in this study (Fig. 2) may be exceeded by rates of CaCO 3 bioerosion and dissolution, highlighting the growing need for monitoring of these destructive processes on coral reefs to better predict the future of coral reef CaCO 3 structures under anthropogenic and climatic change (Kleypas et al. 2001;Perry et al. 2008, Andersson and Gledhill 2013, Eyre et al. 2018, van Woesik and Cacciapaglia 2018. Indeed, recent work has already documented net dissolution (Muehllehner et al. 2016) and rapid rates of bioerosion ) across the Florida Keys Reef Tract and loss of seafloor elevation occurring for reef ecosystems in the Pacific and western Atlantic (Yates et al. 2017). Nonetheless, a CCC approach is valuable for estimating changes in coral community calcification because it incorporates variability in both coral cover and coral community composition. For example, there were significant differences in the relationship between mean CCC and percent coral cover across focal regions (i.e., non-overlapping 95% confidence intervals in Fig. 3). St. John had the lowest CCC vs. % coral cover slope of all the focal regions indicating a more slowly calcifying coral community (i.e., dominated by Orbicella annularis complex; Fig. 2D) while the Main Hawaiian Islands had the highest slope (Fig. 3) due to an abundance of more rapidly calcifying Montipora spp. (Fig. 2A). The slope of the CCC vs. percent coral cover relationship for Mo'orea Fig. 3. Mean slopes of the relationship between coral calcification capacity (CCC) and percent coral cover across all sites and years for each focal region. Gray points represent the mean slope, wide gray lines represent standard error, and narrow black lines represent 95% confidence intervals. was significantly higher than St. John but lower than the Main Hawaiian Islands (Fig. 3) due to a mixture of reef sites dominated by rapidly calcifying Montipora/Acropora and more slowly calcifying Porites spp. (Fig. 2B). Similarly, the Florida Keys Reef Tract had a slope of the CCC vs. percent coral cover relationship that was not significantly different from the Main Hawaiian Islands or Mo'orea, but greater than St. John (Fig. 3) owing to a mixture of reef sites dominated with calcification by either Millepora alcicornis and Acropora palmata or Orbicella annularis complex and Montastraea cavernosa (Fig. 2C). Thus, while Guest et al. (2018a) found coral cover explained much of the variance in CCC within each focal region, the differing slopes of the CCC vs. percent coral cover relationships between focal regions presented here highlight the importance of incorporating coral community composition into reef-scale calcification studies.
It is important to further note that the CCC approach reveals some sites with relatively high CCC for a given coral cover and high variance in the CCC vs. percent coral cover relationship Fig. 4. Mean percent change in coral calcification capacity (CCC) by dominant calcifying corals in each focal region. Each data point describes the coral-specific change in the annual % contribution to CCC over time (i.e., slope represents mean change in contribution to CCC per year) as derived from linear mixed effects models within each focal region, indicating the slope (colored by life-history strategy from Table 1: red, competitive; green, weedy; blue, stress-tolerant; gray, generalist; white, more than two life-history strategies), standard error (wide gray lines), and 95% confidence intervals (narrow black lines). Positive slopes AE 95% above the dashed zero line indicate annual increases in % of CCC vs. year for the respective genus, slopes AE 95% overlapping the dashed zero line indicate no detectable change in % of CCC vs. year for the respective genus, and negative slopes AE 95% below the dashed zero line indicate decreases in annual % of CCC vs. year for the respective coral. across sites and over time within each focal region (Fig. 2), which is consistent with observations from CaCO 3 budget assessments for reefs across the western Atlantic by Perry et al. (2013). Most importantly, some reef sites within the Main Hawaiian Islands, Mo'orea, and Florida Keys Reef Tract focal regions reflect growth trajectories dominated by faster growing competitive corals (i.e., higher CCC per coral cover points plot along the hypothetical Acropora and Montipora growth trajectories, Fig. 2 ABC), suggesting those sites may have either experienced fewer disturbances than neighboring sites, locally resisted them, or recovered following disturbances (i.e., are oases sensu Guest et al. 2018a). Although hypotheses regarding why some reefs have escaped, resisted, and/or recovered from disturbances remain to be empirically tested, the use of CCC presented here serves as a potential means of quantifying the reef-growth potential of coral communities to improve our understanding and projections of future coral reef calcification trajectories.
We further explored changes in CCC by investigating the dominant calcifying corals that contributed at least 5% of the total CCC across the time series in this study (Fig. 1). These species represent potentially useful targets for management and restoration as they may be most important in maintaining reef-scale calcification in the future. However, the relative contributions of many of these taxa to CCC shifted over the time series in this study. For example, the dominant calcifying taxa Montipora capitata, Montipora sp. (Mo'orea), Millepora alcicornis, Porites astreoides, Siderastrea siderea, and Porites porites increased in relative contributions to CCC across the focal regions (Fig. 4, Table 1), suggesting they are becoming increasingly important reef builders in the 21st century. Conversely, Acropora palmata, Acropora spp. (Mo'orea), Montastraea cavernosa, Orbicella annularis complex, and Porites compressa decreased in mean contributions to CCC across the focal regions over the time series in this study (Fig. 4, Table 1) and are, therefore, likely to make diminishing contributions to future reef building.
Dominant calcifying corals were classified by life-history strategies to test whether disturbance events drove decreases in competitive corals' and increases in weedy and stress-tolerant corals' (Darling et al. 2012(Darling et al. , 2013 contributions to CCC in this study (Fig. 2, Table 1). Following the major coral bleaching, crown-of-thorns, and disease-disturbance events, the competitive Acropora spp. corals decreased their contributions to CCC across the focal regions in this study, and in the western Atlantic, the weedy Millepora alcicornis (Florida), Porites astreoides (Florida), and Porites porites (Florida and St. John) corals increased their contributions to CCC. These changes are congruent with Darling et al.'s (2012Darling et al.'s ( , 2013 expectations for competitive and weedy corals under increased disturbance regimes. Although stress-tolerant corals are predicted to increase in abundance on disturbed reefs (Darling et al. 2012), the generalist/stress-tolerant Orbicella annularis complex corals decreased their contributions to CCC throughout the time series across both the Florida Keys and St. John focal regions. It is important to note, however, that O. annularis complex was disproportionally affected by white plague disease in St. John (Miller et al. 2009), highlighting the importance of considering disease susceptibility in the framework of shifting coral communities and life-history strategies. Furthermore, Orbicella spp. have been disproportionately affected by disturbances in the Florida Keys over the last 20 yr (Ruzicka et al. 2013;Toth et al. 2014Toth et al. , 2019, which suggests that the characterization of these species as stress-tolerant may warrant reconsideration. Lastly, mean contributions to CCC by stress-tolerant Montastraea cavernosa decreased and Siderastrea siderea increased across the Florida Keys (Siderastrea siderea CCC decreased somewhat À0.18 AE 0.16 albeit not significantly at P = 0.269 in St. John), which, respectively, contradicts and agrees with hypothesized stress-tolerant life-history expectations on disturbed reefs for these corals (Darling et al. 2012(Darling et al. , 2013. Thus, while the changing contributions to CCC by competitive and weedy corals in this study appear to agree well with hypothesized life-history expectations for disturbed reefs, the shifting contributions of generalist and stress-tolerant corals to CCC are more variable. Some of these discrepancies could in part be due to the widespread coral disease in St. John and/or other more localized drivers testing the tolerance of coral communities to disturbances and warrants further investigation to improve future projections of coral reef community structures and CCC.
Fast-growing Acropora cervicornis and Acropora palmata corals were once widespread, dominant calcifiers in the western Atlantic, but declined in the late 1970s and 1980s primarily due to white band disease and hurricanes (Aronson and Precht 2001, Kuffner and Toth 2016) and are no longer major CaCO 3 producers in this region (Perry et al. 2015). However, Darling et al. (2019) found that competitive corals are still the dominant corals on many reefs across the Indo-Pacific, which suggests they are still major CaCO 3 producers across the region. This may in part be due to greater functional redundancy of competitive coral life-history traits (i.e., faster calcification) in the Pacific relative to the western Atlantic Toth 2016, McWilliam et al. 2018). Regardless, the declines in contributions of competitive corals to CCC following disturbances in this study imply future projected decreases in overall CCC under predicted increases in coral bleaching and disease-disturbance frequencies and intensities (e.g., Donner et al. 2005, Randall and van Woesik 2015, van Hooidonk et al. 2016. The findings of this study agree with previously observed declines in Acropora corals owing to coral bleaching and disease in recent decades (e.g., Aronson and Precht 2001, Loya et al. 2001, Perry and Morgan 2017.
The relative contribution to CCC by weedy corals may continue to increase as has been observed by the current dominance of nonframework building coral calcification in the western Atlantic (Perry et al. 2015). However, observed fluctuations in weedy coral cover following disturbance events Edmunds 2013, Darling et al. 2013) could drive corresponding interannual fluctuations in CCC. This suggests reefs with increased weedy CCC may, therefore, maintain lower and/or less stable CCC through time (Alvarez-Filip et al. 2013). Nevertheless, reefs with a veneer of weedy corals may experience lower rates of physical loss of CaCO 3 by rendering the reef framework less accessible to destructive grazing and bioerosion Toth 2016, Toth et al. 2018). The observed increase in contributions of some of the stress-tolerant corals in this study to CCC will likely decrease CCC for a given coral cover due to their generally slower calcification rates. Moreover, the greater bleaching resistance hypothesized for many of these stress-tolerant corals (e.g., Loya et al. 2001, McClanahan and Maina 2003, McClanahan et al. 2009, van Woesik et al. 2011, Darling et al. 2013) and current dominance of stress-tolerant corals across much of the Indo-Pacific (Darling et al. 2019) suggest stress-tolerant corals may act to stabilize CCC under increased thermal stress (e.g., Januchowski- Hartley et al. 2017, Ryan et al. 2019. Further research should therefore be conducted to rigorously quantify coral and reef-scale rates of calcification, bleaching/disease resistance, and recovery to better predict future warming-induced changes in CCC and coral reef CaCO 3 structures (e.g., Swain et al. 2016, Januchowski-Hartley et al. 2017, Courtney et al. 2018, Mizerek et al. 2018, Ortiz et al. 2018, van Woesik et al. 2018, Gouezo et al. 2019. As many coral reefs around the world continue to decline in coral cover and shift in coral species compositions under increasing frequencies of coral bleaching and other disturbances (Gardner et al. 2003, Donner et al. 2005, Bruno and Selig 2007, Jackson et al. 2014, van Hooidonk et al. 2016, metrics that evaluate historical reef condition from coral community data such as CCC may prove useful for understanding previous changes and projecting future coral reef CaCO 3 structures and functions. Likewise, evaluation of the ecological drivers of CCC in the context of coral life-history strategies and coral reef disturbances could easily expand reef monitoring programs to include estimates and projections of CaCO 3 production. The future of CCC for any given reef site relies on the reduction of the magnitude and frequency of coral reef disturbances through both global (i.e., reduction of greenhouse gas emissions that drive ocean warming and acidification) and local (e.g., reduction of land-based pollution, overfishing, and habitat destruction) efforts. global-scale ocean change: the coral reef example" supported by funding awarded to P. J. Edmunds, R. D. Gates, and I. B. Kuffner with a Powell Fellowship to J. R. Guest. T. A. Courtney and A. J. Andersson additionally received funding from the National Science Foundation. We thank P. J. Edmunds and J. Miller for improving this manuscript with their helpful comments and the Powell Center for their support of the working group. We thank K. Rodgers, E. Brown, R. Ruzicka, M. Colella, J. Miller, A. Atkinson, and M. Feely for providing the raw data used in this study that are now published in a USGS data release https:// doi.org/10.5066/f78w3c7w. All R code, and associated data files used in this analysis are publicly available on GitHub at https://github.com/traviscourtney/Powe llCCC. We further thank the reviewers and editor for improving this manuscript with their helpful comments and suggestions. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. The authors declare no conflict of interest. TAC and AJA designed the study, TAC analyzed the data, and TAC wrote the first draft of the manuscript with input from AJA and comments from all authors.