The transformation of Caribbean coral communities since humans

Abstract The mass die‐off of Caribbean corals has transformed many of this region’s reefs to macroalgal‐dominated habitats since systematic monitoring began in the 1970s. Although attributed to a combination of local and global human stressors, the lack of long‐term data on Caribbean reef coral communities has prevented a clear understanding of the causes and consequences of coral declines. We integrated paleoecological, historical, and modern survey data to track the occurrence of major coral species and life‐history groups throughout the Caribbean from the prehuman period to the present. The regional loss of Acropora corals beginning by the 1960s from local human disturbances resulted in increases in the occurrence of formerly subdominant stress‐tolerant and weedy scleractinian corals and the competitive hydrozoan Millepora beginning in the 1970s and 1980s. These transformations have resulted in the homogenization of coral communities within individual countries. However, increases in stress‐tolerant and weedy corals have slowed or reversed since the 1980s and 1990s in tandem with intensified coral bleaching and disease. These patterns reveal the long history of increasingly stressful environmental conditions on Caribbean reefs that began with widespread local human disturbances and have recently culminated in the combined effects of local and global change.

data since the 1980s were received directly from contributors or gleaned from peer-reviewed literature to construct the Global Coral Reef Monitoring Network (GCRMN) database that assessed trends in Caribbean reef benthic communities from 1970-2011 (Table S1; Jackson et al., 2014).
Temporal changes were tracked for 14 common coral taxonomic groups that were consistently recorded within each time bin (henceforth termed "species groups": A. cervicornis, A. palmata, Agaricia spp., Montastraea cavernosa, Colpophyllia spp., Pseudodiploria spp., Madracis spp., Meandrina spp., Millepora spp., Orbicella spp., branching Porites spp., Porites astreoides, Siderastrea spp., and Stephanocoenia spp.). To provide sufficient temporal resolution and ensure adequate sample sizes for assessing change over the full time series, data were grouped into eleven time bins: Late Pleistocene (~131,000-12,000 years ago, which encompasses the period prior to human settlement in the Caribbean region), Holocene (~9,100 years ago-1500 AD, which encompasses the prehistoric period prior to European contact), 1500-1959, 1960-1969, 1970-1979, 1980-1984, 1985-1989, 1990-1994, 1995-2000, 2001-2004, and 2005-2011. Bins were reduced to 5-year increments after 1980 (except for a 6-year increment for the most recent bin) due to the large increase in reef survey effort following the mass die-off of the urchin D. antillarum. The prehuman Pleistocene bin encompasses a period of high-magnitude fluctuations in sea level and climate (during which Acropora coral dominance persisted in the Caribbean; Pandolfi & Jackson, 2006), whereas the prehistoric Holocene and subsequent bins encompass a period with higher stability in climate and sea level than the Pleistocene (Khan et al., 2017). Please see Table A1 for a timeline of major events affecting Caribbean reefs within each time bin.
Qualitative data were included in the database if, in addition to presence/absence information for at least one coral species, the following information was also available: (a) age of fossil data or year of observation of modern data, (b) original source of data, (c) country and island, coastline, or reef site, and (d) water depth or reef zone.
Data were recorded at the survey level, with a survey constituting a unique combination of reef site, depth zone, and year. For surveys which included multiple replicates (i.e., transects or quadrats) at the same site, depth zone, and year/period, an overall value was computed for all replicates. Surveys constituted individual reef "sites" and in some cases encompassed more extensive areas such as entire reef tracts, bays, or banks.
We analyzed data from "reef crest" and "midslope" reef zones separately. Generally, the reef crest data spanned 0-6 m water depth and midslope data spanned between 6-20 m, as 6 m was the depth at which dominance typically transitioned from A. palmata to A. cervicornis in the semi-quantitative and quantitative data.
However, the reef crest/midslope zone delineation was made on a location-by-location basis by first considering water depth and, when available, additional environmental characteristics such as wave exposure and reef morphology. For some offshore locations with presumably higher water clarity, the cutoff was closer to 10 m. When a precise water depth was not available, we utilized Acropora species presence and/or dominance in addition to environmental characteristics to delineate between zones (Cramer, Jackson et al., 2020).
When not reported in the papers from which data were extracted, paleo water depths were determined using the procedure outlined in Cramer, Jackson et al. (2020). Surveys from backreef habitats, reef flats, and reef pavements were excluded from our database, as these reef zones are not the preferred environments for Caribbean Acropora (Milliman, 1969).
Change in coral community composition was assessed using coral species presence/absence data, extracted from species rankings, presence/absence, percent living coral cover, number of individual colonies, and percent weight or volume of coral fossils. Only species included in the 14 commonly occurring species groups were considered in our analyses. Data recorded as "Porites spp." were assumed to be branching Porites spp. because species within this complex are difficult to distinguish, were not consistently recognized as distinct until the 20th century (Jameson & Cairns, 2012), and were often recorded as a single species, "P. porites" in our database, whereas the morphologically distinct Porites astreoides was consistently recorded in our database and was therefore assigned to a separate category in our analyses. Data recorded as "Montastraea spp." were assumed to be Orbicella spp. because species within this complex are difficult to distinguish and were not recognized as separate species until 1992 (Knowlton et al., 1992). The morphologically distinct M. cavernosa was consistently recorded in our database and was therefore assigned to a separate category in our analyses. Pseudodiploria clivosa, Pseudodiploria strigosa, and Diploria labyrinthiformis were assigned to Pseudodiploria spp., as these species were placed within the same genus until 2012 (Budd et al., 2012).

| Coral life-history strategies
To infer potential environmental changes driving coral community shifts, we also tracked trends in coral life-history groups. We utilized a trait-based classification approach that grouped scleractinians into four life-history strategies separated primarily by colony morphology, growth rate, and reproductive mode . This configuration roughly follows Grime's arrangement of plant species into three basic life-history strategies: competitive species that maximize growth, stress-tolerant species that maximize survival, and ruderal or weedy species that maximize fecundity (Grime, 1977(Grime, , 2006. A fourth category, generalist species, represents a mixture of these strategies. However, because the generalist group was composed primarily of corals that are less common in the Caribbean and their presence and absence was not consistently noted in our dataset, this group was not included in our analyses. To add ecological context to these groupings, we collated from the literature taxon-specific, qualitative measures of additional life-history characteristics including sexual reproductive output (larval recruitment), asexual propagation via colony fragmentation, interspecific aggression, and susceptibility to disturbances such as sedimentation and bleaching (Table 1 and Table A2). Although larval recruitment represents the end-point of fecundity, fertilization, dispersal, and early postsettlement mortality and is not a life-history characteristic per se, this metric is closely linked to life-history strategy  and provided valuable ecological insight into observed community change. Our life-history groupings followed those determined via quantitative analyses in  and Hardt (2007), with the addition of the hydrozoan Millepora described below. This addition was based on qualitative similarities in ranked values of lifehistory traits described in Table 1; methods for assigning life-history trait rankings are described in Table A2.
The competitive life-history group included A. cervicornis, A. palmata, and Millepora spp. This group is distinguished by fast growth rates, large branching morphologies that can outcompete other corals for light and/or space, medium to high levels of aggression, a spawning mode of reproduction but low rates of sexual recruitment, high propensity for asexual reproduction via fragmentation, and low tolerance to disturbances such as sedimentation and thermal stress (Table 1 and Table A2). This combination of traits historically allowed Acropora corals to dominate shallow, high-energy reef environments prior to local and global anthropogenic stressors Jackson, 1992;.
Although not included in the previous analyses of coral life-history guilds, we included Millepora in the competitive category because of its Acropora-like ability to preempt space on reefs due primarily to fragmentation and fast growth and its high susceptibility to bleaching (Table 1 and Table A2, Dubé, 2016;. The stress-tolerant life-history group includes Colpophyllia natans, Pseudodiploria spp., Meandrina spp., Montastrea cavernosa, Orbicella spp., Siderastrea spp., and Stephanocoenia spp. This group is distinguished by slow to moderate growth rates, large and domed morphologies with higher ability to clear sediment and other particles and resistant to storm damage, a spawning mode of reproduction with low to moderate sexual recruitment, low to high interspecific aggression, and relatively higher tolerance for sedimentation and thermal stress (Table 1 and Table A2). Although some stress-tolerant corals have a higher susceptibility to bleaching, colony survival rates are typically high within this group (McClanahan & Muthiga, 1998).
This combination of traits historically allowed these species to persist and dominate in environments subject to frequent, low-magnitude disturbances such as sediment resuspension and temperature stress (Geister, 1977;Rutzler & Macintyre, 1982).
Because the three extant Orbicella species were until recently classified as Montastrea annularis (Knowlton et al., 1992), we assigned Orbicella spp. to the stress-tolerant category in accordance with the classification for M. annularis . However, this genera's historical dominance on midslope zones on many Caribbean TA B L E 1 Coral life-history groups and their defining traits. Rankings compiled from data references listed in Table A2. Growth rate is average linear extension rate, reproductive output is larval recruitment rate. Growth rate ranking computed from mean of all published values separated into bottom/middle/top third percentiles of mean (e.g., , 67-100 percentiles); ranges = 1. 1-4.0, 5.0-7.0, and 13.2-119.5 mm/year for slow, moderate, and fast growth, respectively. Interspecific aggression ranking computed from experimental results from  and qualitative ranking from synthesis of literature, and rankings for all other traits computed from qualitative ranking from synthesis of literature   suggests this coral also has characteristics of the competitive life-history group.
The weedy life-history group includes Agaricia spp., Madracis spp., branching Porites spp., and P. astreoides. This group is distinguished by lower-relief plating, foliose, branching, and domed morphologies with slow to fast growth rates, a brooding mode of reproduction that allows for rapid colonization at low population densities, generally high rates of sexual recruitment, high to low occurrence of asexual reproduction via fragmentation, low interspecific aggression, generally high susceptibility to bleaching, and high tolerance of sedimentation (Table 1 and Table A2). This combination of traits historically allowed these early-successional species to opportunistically and rapidly colonize open spaces cleared by highmagnitude acute disturbances  Hughes & Jackson, 1985).

| Analyses of coral community change
To estimate the proportion of reef sites containing each coral lifehistory and species group in each time interval, we utilized binomial generalized linear mixed effects models that predicted the propor- To assess the effects of coral community change on regional diversity patterns, we tracked temporal changes in community dissimilarity. We utilized species presence/absence matrices to compute Jaccard's dissimilarity index (Jaccard, 1912

| Species community database
Coral species presence and absence data were compiled from 2,396 reef sites from 26 countries for the reef crest zone and 5,091 reef sites from 30 countries for the midslope zone ( Figure 1). Full community data (containing a 0 or 1 value for all 14 common coral species groups) were available for all time periods and were compiled from 1,569 reef sites from 24 countries for the reef crest and 3,207 sites from 27 countries for the midslope zone (Tables A3 and A4).

| Long-term change in coral community composition
Since the mid-20th century, shallow water reefs across the Caribbean The assessment of trends for individual species groups revealed that at both reef zones, the occurrence of competitive A. palmata and A. cervicornis declined significantly between the Pleistocene and Holocene periods. After the Holocene, the next significant decline in competitive Acropora corals occurred in the 1960s; the occurrence of these corals remained significantly lower than prehuman levels from this point forward (
Our time series suggests that the Caribbean-wide loss of However, the significant post-1980s declines in the occurrence in stress-tolerant and weedy corals that occurred at the reef crest (but not midslope) zone may reflect higher coral mortality from (a) anthropogenic bleaching in shallower zones that experience greater thermal stress Bridge et al., 2013) and/or (b) the die-off of the Diadema urchin that prefers shallower reef zones .

| The role of local and global human stressors
Although climate change is currently imperiling reef ecosystems globally , the early timing of the initial declines  -19591960-196919701980-19891990-19941995-1999Holocene Pleistocene 1500-19591960-196919701980-19891990-19941995-1999 (Knutson et al., 2006;Sheppard & Rioja-Nieto, 2005), and warming-related coral bleaching was not observed until the late 1980s (Glynn, 1993). Our recent analysis of long-term trends in the dominance of A. cervicornis and A. palmata showed that initial declines in the 1950s and 1960s were unrelated to regional or global stressors (i.e., anthropogenic temperature stress or hurricane exposure; Cramer, Jackson et al., 2020).
Instead, the early timing of initial changes in Caribbean coral communities implicates long-standing local stressors such as fishing and land-based pollution. However, the paucity of long-term data on fishing effort/reef fish abundance or reef water quality precludes a quantitative assessment of the role of these activities in recent Caribbean-wide reef ecosystem change. Consequently, despite the well-established relationships between hermatypic coral persistence and abundant herbivorous reef fish populations and low-sediment, low-nutrient waters (Cramer et al., 2017;Fabricius, 2005;Hughes et al., 2007;Randall, 1961), historical fishing, and land clearing have been largely ignored in most analyses of Caribbean coral declines (Abelson, 2019). Fortunately, a few longer-term datasets on water quality at various Caribbean reefs provide valuable insights into the role of land-based runoff in coral community change. An analysis of seawater and macroalgae nitrogen content since the 1990s from the Florida Keys implicates land-based nutrients from agriculture and development in the decades-long coral declines within that reef tract (Lapointe et al., 2019). Studies based on historical and paleontological data also suggest that early reef ecosystem declines in Barbados and Panama may be attributed to increases in coastal runoff from historical land clearing for agriculture (Cramer et al., 2012;Cramer, O'Dea et al., 2020;Lewis, 1984). Last, the increase at both reef zones in the occurrence of Millepora and Siderastrea, corals that are particularly tolerant of high sedimentation and high turbidity conditions , strongly indicates that declining water quality is a major driver of coral community change across the Caribbean.
Although initial declines in Acropora predate regional disturbances, subsequent changes bear a clear imprint of anthropogenic climate change. For instance, while the occurrence of stress-tolerant and weedy corals is significantly higher today than the prehuman period, increases generally leveled off or reversed beginning in the late 1980s. The slowdown of increases in these corals is likely a response to the rapid explosion in benthic macroalgae following the die-off of the keystone herbivore D. antillarum (Jackson et al., 2014) and increases in bleaching-related mortality from anthropogenic temperature stress (Eakin et al., 2010). The more marked declines in Agarcia compared to branching Porites and P. astreoides at the reef crest likely reflect the relatively higher sensitivity of Agaricia to thermal stress: Agaricia experienced widespread bleaching episodes during thermal anomalies in the 1980s and 1990s (Aronson et al., 2000;Gates, 1990;Lasker et al., 1984). The post-1980s declines/plateaus in stress-tolerant and weedy species shown in this study are also a reflection of the increasingly frequent epizootics affecting these corals over the past 2-3 decades and that are linked to a combination of local and global anthropogenic stressors (Vega Thurber et al., 2020).
In contrast to the early transformation of Caribbean coral communities following the initial loss of Acropora in the 1950s/1960s, more recent changes since the 1980s/1990s demonstrate the heightened effects of local stressors and climate change acting on reefs simultaneously. Although our study suggests that White Band Disease was not the cause of initial Acropora declines, it confirms that it has unequivocally contributed to the loss of this genus: the second significant Acropora decline observed in our time series in the early 1980s immediately followed the first instances of this disease reported in the late 1970s (Gladfelter, 1982). Landbased runoff has been shown to exacerbate coral bleaching and disease (Bruno et al., 2003;Lapointe et al., 2019;Wiedenmann et al., 2013), suggesting that reef eutrophication played a role in the emergence of these morbidities. Similarly, the region-wide plateaus/declines in stress-tolerant and weedy corals we observed since the 1980s/1990s reveal that local and global stressors are making Caribbean reef environments less suitable for those corals with the hardiest of life-history strategies. Indeed, recent monitoring efforts have documented declines in several stress-tolerant taxa from bleaching and disease that were initiated two decades ago (Edmunds & Elahi, 2007;Harvell et al., 2007) and show that several stress-tolerant species are currently rapidly succumbing to the highly lethal Stony Coral Tissue Loss Disease that does not affect Acropora (Precht et al., 2016;Weil et al., 2009;van Woesik & Randall, 2017). Monitoring efforts are also documenting declines in weedy corals such as Agaricia due to recent Caribbeanwide bleaching events (Walton et al., 2018). Finally, thermal stress and algal overgrowth are causing recruitment failure in Caribbean coral species regardless of life-history guild (Arnold et al., 2010;Randall & Szmant, 2009

| Challenges with assessing long-term trends
To track Caribbean coral community change prior to and since the arrival of humans, we used data from multiple sources, including uplifted fossil reefs, reef matrix cores, qualitative historical data, and underwater survey data, which may have led to uneven detection of particular coral taxa across different data types. For example, we observed conspicuously lower within-country community dissimilarity observed within the fossil versus nonfossil time bins, likely due to the greater time-averaging within the former (Figure 4).
We also found significant differences in the occurrence of multiple coral taxa between the Pleistocene and Holocene, including declining occurrence of A. cervicornis at the midslope zone and increased occurrence of a number of stress-tolerant and weedy taxa at both zones (  Figures 2 and 3). These differences could reflect declining rates of sea level rise during the late Holocene, which would favor increased dominance of more slowly growing species with massive colony forms (Hongo, 2012)-a trend we observe in our data from the midslope zone (Figure 3b). The relatively low occurrence of several species with massive colony growth forms in the Holocene period could also be due to the difficulty of sampling these colony types in the narrow-diameter Holocene reef matrix cores that comprise much of the data from this time bin. Lower A. cervicornis occurrence in the Holocene compared to the Pleistocene could also be a result of inaccurate paleodepth estimates and/or underestimation of A. cervicornis abundance from the Holocene cores. However, A. palmata dominance increased slightly at the reef crest between the Pleistocene and Holocene, demonstrating that there was no bias against sampling this coral in the Holocene reef cores (Cramer, Jackson et al., 2020). Despite these discrepancies between the fossil time bins, Pleistocene coral communities were generally similar to those in the historical and early modern period (Figures 2 and   3), in agreement with other studies from the Caribbean Sea that showed remarkable comparability in Pleistocene and modern coral communities despite variation in growth rates among coral species, the higher degree of time-averaging in fossil assemblages, and possible transport and mixing of fossil material (Jackson, 1992;. Importantly, when data from the Pleistocene and Holocene periods are combined, overall trends in occurrence of coral functional and species groups are largely identical to those with these time periods separated: Declines in Acropora occurrence first occurred in the 1960s, followed by increases in stress-tolerant and weedy species in the 1970s and 1980s (Table A5, Figures A1 and A2).
Although abundance data allow for a more robust assessment of community composition than presence/absence data, the exceptionally broad temporal, taxonomic, and geographic scales covered in this study necessitated the utilization of the latter. We recognize that occurrence does not equal abundance; although we found that the current occurrence of stress-tolerant and weedy corals is higher than that from the prehuman and historical periods, the total abundance of living coral has declined by 50%-80% across the Caribbean since the initiation of quantitative surveys in the 1970s (Gardner et al., 2003;Jackson et al., 2014). However, ecological studies show that, on large spatial scales, trends in occurrence (proportion of sites occupied) are correlated with trends in abundance (MacKenzie et al., 2005;Weber et al., 2004), suggesting that the long-term trends shown in this study are reliable proxies of qualitative trends in relative abundance. The occurrence trends observed in this study also correspond to recent trends in absolute abundance: The increasing occurrence of some stress-tolerant and weedy corals (Agaricia, P. astreoides, branching Porites) over the past four decades in the midslope zone corresponds to trends in percent living cover from modern localized surveys of Caribbean coral communities

| Conservation implications
The anthropogenic transformation of Caribbean coral communities into their novel configuration has widespread consequences for reef ecosystem functioning. First, the loss of competitive Acropora and stress-tolerant Orbicella corals represents a massive simplification of reef architectural structure and loss of carbonate production that will likely compromise the ability of Caribbean reefs to keep pace with anthropogenic sea level rise (Alvarez-Filip et al., 2009;Perry et al., 2014). Second, the loss of habitat complexity has the potential to reduce the diversity, biomass, and abundance of reef fish communities, the fisheries productivity of reefs, and the diversity of coral-associated invertebrates (Cramer et al., 2012;Paddack et al., 2009;Richardson et al., 2018;Rogers et al., 2014). Third, coral community turnover has reduced the recovery potential of these reefs by selectively removing coral species with a spawning mode of reproduction and high larval dispersal rates (Acropora and Orbicella) and replacing them with species with a brooding mode of reproduction and low larval dispersal rates (Agaricia and Porites), limiting the ability of relatively intact reefs to re-seed degraded ones (Knowlton, 2001).
(e) (f) (d) TA B L E A 1 Time bins included in the analysis of long-term change in Caribbean coral communities and significant events affecting reef environments and detection of ecological change by researchers. Timeline sourced from Jackson et al. (2014) and Cramer, Jackson et al. (2020) 48 52, 53, 56, 57, 58, 59 Stephanocoenia spp.