Tropical reefs are among the most biologically rich ecosystems on earth providing income for millions of people through fisheries, tourism and coastal protection (Moberg and Rönnbäck, 2003). On tropical reefs scleractinian corals are largely responsible for the architectural complexity due to the deposition of calcium carbonate and thereby provide habitat, food, and refuge for many taxonomic groups. Over the past decades corals reefs have been degrading worldwide at unprecedented rates (Gardner et al., 2003; Bak et al., 2005; Pandolfi and Jackson, 2006). Deteriorating coastal water quality (Rogers, 1990; Williams et al., 2002; Vega-Thurber et al., 2014), overexploitation of herbivores (Jackson et al., 2001) and anthropogenic induced climate change (Glynn, 1996; Aronson et al., 2002a; Wilkinson and Souter, 2008) are generally associated with this ubiquitous decline in coral cover (Jackson et al., 2014). These stressors negatively impact coral resilience, impeding the ability of corals to cope with disease, bleaching or physical disturbance, which in turn affects their competitive position with respect to other reef organisms such as macroalgae or sponges (Vega-Thurber et al., 2014; Zaneveld et al., 2016). Aforementioned developments are suggested to cause shifts in coral communities from assemblages dominated by framework building corals to a dominance of more persistent and opportunistic non-framework building species (e.g., Green et al., 2008; Alvarez-Filip et al., 2011a; Darling et al., 2012; Perry et al., 2015).

Whether coral species are likely to become winners or losers in the changing reef environment depends strongly on their specific life-history traits (Knowlton, 2001; Van Woesik et al., 2011). In an attempt to predict what species are likely to be lost and what species may be more persistent, several studies have divided species over groups based on shared characteristics such as: way of propagation, growth rate and thermo-tolerance (e.g., Edinger and Risk, 2000; Murdoch, 2007). Recently, Darling et al. (2012) divided 147 coral species from tropical reefs all over the world over four main groups: competitive, weedy or ruderal, stress-tolerant and generalist. This grouping was based on 11 functional traits. In the early 1970s coral reefs were still generally dominated by stress-tolerant, competitive and generalist species. It is hypothesized, however, that coral communities are shifting toward species assemblages comprising mainly of phenotypically plastic weedy species and some persistent framework building (stress-tolerant, generalist) species (Darling et al., 2012; Grottoli et al., 2014; McClanahan et al., 2014; Perry et al., 2015). Since the frequency and intensity of events inducing coral mortality (e.g., bleaching and hurricanes) is predicted to increase, recovery and adaptation will become progressively more difficult for all coral species (Hoegh-Guldberg, 1999; Van Woesik et al., 2011; McClanahan et al., 2014). As a consequence, the contribution of corals in benthic reef communities will likely decrease even further and the most stress susceptible coral species may disappear entirely at many reefs (Knowlton, 2001; Hoegh-Guldberg et al., 2007). Predicting trajectories of coral community change, however, is extremely complex due to the numerous factors involved (Hughes and Tanner, 2000; Done et al., 2010). Additionally, shifts in coral communities may occur slowly and are therefore often difficult to distinguish (Pauly, 1995). There appears to be a consistent lack of long-term temporal quantitative data documenting the trajectories of different species in the coral community.

Here, we provide data on the coral community dynamics at reefs on the southern Caribbean islands of Curaçao and Bonaire. Over the course of the studied time series human populations in the Caribbean increased from 27 million in 1973 to approximately 43 million in 2015 (United Nations, Department of Economic and Social Affairs, Population, Division, 2015). As a consequence, the Caribbean basin experiences chronic decline of water quality and overexploitation of fish (Siung-Chang, 1997; Jackson et al., 2014). Due to climate change and increasing sea water temperature, Caribbean corals have been exposed to mass bleaching events in 1998 (Aronson et al., 2002a), 2005 (Wilkinson and Souter, 2008; Eakin et al., 2010) and 2010 (Alemu and Clement, 2014). Furthermore, White Band Disease greatly reduced the cover of Acropora spp. from the late seventies to the mid-eighties, which negatively impacted rates of calcification and architectural complexity of the shallow reef (Gladfelter, 1982; Aronson and Precht, 2001). Additionally, herbivory was greatly reduced due to the mass-mortality of the Diadema antillarum sea-urchin in 1983 (Bak et al., 1984), formerly important in controlling the densities of competitive algae (De Ruyter Van Steveninck and Bak, 1986).

In comparison to other tropical regions (e.g., the Pacific), the Caribbean has relatively few coral species (Veron, 2014). As a consequence, certain ecologically relevant traits of specific species may be lost more rapidly in the Caribbean compared to other regions (Nyström, 2006). In many Caribbean sites the cover of corals providing complex 3D reef structure, mostly Orbicella spp. (= Montastraea, sensu Budd et al., 2012) and Acropora spp. was already strongly reduced over the past decades (Van Duyl, 1985; Aronson and Precht, 2001; Bruckner and Bruckner, 2006). The calcification balance on many of these reefs may currently have shifted toward net erosion which will result in the reduction of architectural complexity (Alvarez-Filip et al., 2009; Kennedy et al., 2013; Perry et al., 2013). This may drastically alter reef functioning and lead to loss of biodiversity as well as a reduction in important ecosystem services for coastal human population such as wave protection and food security (Moberg and Rönnbäck, 2003; Alvarez-Filip et al., 2009, 2011b; Graham and Nash, 2013; Newman et al., 2015).

On degrading Caribbean reefs, phenotypically plastic corals such as Madracis spp., Porites spp., Siderastrea spp. or Agaricia spp., are expected to replace species controlling reef structure like Acropora spp. and Orbicella spp. (Pandolfi and Jackson, 2006) and become dominant in coral communities (Jackson, 2001; Aronson et al., 2002b; Green et al., 2008; Perry et al., 2015). These species are, however, often small and generally not capable of fulfilling the important ecological functions provided by large framework building species (Bellwood et al., 2004; Nyström, 2006; Alvarez-Filip et al., 2013; Perry et al., 2013). Moreover, it may be expected that in the long run opportunistic weedy species (Darling et al., 2012) will also disappear since no coral species appears to be effectively insensitive to anthropogenic impact (McClanahan et al., 2014). We aim, on the basis of images from 18 time series collected at five reef sites in Bonaire and Curaçao, and covering up to 40 years, to (1) determine temporal and spatial changes in dominant coral species, (2) link observed trajectories of the coral community to the life-history strategies of large framework building species and opportunistic weedy species, (3) estimate the relative loss of carbonate production due to changes in coral species composition and cover over time, and finally (4) discriminate between the winners and losers on these Caribbean coral reefs.

Since the early 1970s the coral communities of Curaçao and Bonaire have changed significantly in coral cover and species composition. Total cover has declined and colonies have generally become smaller on both the shallow (10–20 m) and upper-mesophotic reef (30–40 m). In particular, initially dominant framework building species such as Orbicella spp., M. cavernosa and S. siderea, lost significant cover, a trend that occurs throughout the wider Caribbean region (Bruckner and Bruckner, 2006; Edmunds and Elahi, 2007; Jackson et al., 2014). Concurrently, the relative cover of some phenotypically plastic so-called weedy species (sensu Knowlton, 2001) moderately increased since 1973. The observed shift also took place at the reference site Awa Blancu (AB) which is located far upstream and relatively distant from anthropogenic disturbances along the coast of Curaçao (Figure 1), but less distinct.

Overall, reef degradation was less pronounced at Awa Blancu (Figure 2). A noticeable difference with the shallow reef of sites I-IV was that Orbicella, over time, remained the dominant genus at AB with still relatively high cover by 2014 (13% and 11% of the total quadrat at 10 and 20 m respectively). Moreover, the coral communities at AB remained dominated by framework building species throughout the times series. This is in contrast to sites I-IV, where opportunistic species have become dominant. The latter suggests that coral communities along the leeward coast near urban areas and outlet bays (I-IV) receive additional stress on top of stress factors acting on a larger scale (i.e., regional or worldwide), at least down to 20 m depth. Since the start of our time series both Curaçao and Bonaire have experienced significant human population expansion (50% increase on Bonaire since 2001) as well as increased tourism (CBS Curaçao 2015, http://www.cbs.cw; CBS Dutch Caribbean 2015 www.rijksdienstcn.com). Both have consequently led to extensive development of the coastal area. Factors associated with coastal development such as pollution, turbidity and sedimentation negatively impact coral recruitment and survival by smothering corals or reducing the ability of corals to cope with diseases, bleaching or spatial competition (e.g., Meesters et al., 1992; Nugues and Roberts, 2003; Fabricius, 2005; Jackson et al., 2014; Vega-Thurber et al., 2014). Similar to many tropical regions, sewage water on Curaçao and Bonaire is generally insufficiently treated or even largely untreated before it is discharged (Buth and Ras, 1992; Lapointe and Mallin, 2011). Indeed, along the coast of Curaçao clearly higher nutrient concentrations have been found at sites I-III compared to Fuik bay, which is located near Awa Blancu (Gast et al., 1999; Govers et al., 2014). Elevated nutrient levels have also been found at Karpata (IV) on Bonaire during measurements in 2012 and 2013 (Slijkerman et al., 2014). The observed spatial variation in coral community degradation may therefore reflect the impact of persevering stress factors associated with coastal urbanization around more densely populated areas. These results suggest that implementation of thorough local conservation and management strategies, such as the implementation of properly working sewage treatment plants, will benefit local reef health and retard reef degradation. In addition, stronger recreational diving regulation may be useful since approximately 75% of the stay-over tourists on Bonaire are divers (IVM report 2013; http://www.rijksdienstcn.com), undoubtedly having a local impact on the reef (e.g., Hawkins et al., 1999; Lamb et al., 2014). Comparable data are lacking for Curaçao. It is clear that supporting data on specific effects of the various local stressors will be essential for proper implementation of conservation strategies, but at present such data is very limited or even absent.

The observed general decline in cover and abundance of corals not only occurred among reef building species, but included also most opportunistic species (e.g., P. porites, Mycetophyllia spp., A. agaricites and Madracis spp.). A few species, in particular M. mirabilis, A. lamarcki and P. astreoides, however, appear to be more resilient and their decrease in cover and abundance was less distinct (Figure 4, Table S2). This is confirmed by observations throughout the Caribbean (Green et al., 2008; Anderson et al., 2014; Perry et al., 2015). Nevertheless, these species seem to decline when the overall coral cover in the quadrat was substantially reduced. Like many coral species, M. mirabilis can not easily settle on loose substratum such as sand or rubble (Bak and Criens, 1981; Nagelkerken et al., 2000). Presumably, M. mirabilis relies on substratum provided by other species after mortality to settle or expand. Indeed, we found an increase in cover of M. mirabilis in quadrats where large colony building species (e.g., Orbicella complex) used to be dominant but died and left available hard substratum. Contrastingly, in quadrats where larger species were not initially present the cover of M. mirabilis remained low or even declined (e.g., site I, 10 m). This implies that survival of this species is all but certain. According to predictions (e.g., Darling et al., 2012), several larger framework building species such as S. michelini and M. meandrites indeed appear to be relatively persistent (stress-tolerant sensu Darling et al., 2012) and were not readily lost over time. Though their average cover declined in all quadrats, they continued to contribute to total quadrat cover by 2014 (Figure 3). D. strigosa even newly appeared in several shallow quadrats and seems relatively stress-tolerant (Figure 3, Table S2). If, however, the observed declining trend continues, it is likely that even these species will eventually disappear. Also, it remains to be seen whether the apparently opportunistic weedy species will be sufficiently tolerant to cope with the generally expected increase in frequency and severity of catastrophic events (Van Woesik et al., 2011; McClanahan et al., 2014). Historic records (e.g., van Woesik et al., 2012) show that even opportunistic species can become regionally extinct, as a consequence of large-scale climate disturbances. At the present rate of decline (Figure 3A) total coral cover on the reef slopes of Curaçao and Bonaire is expected to have dropped below 1% by 2030.

Evidently, the decline in colony size and shift toward dominance of smaller colonies is largely caused by a decline in abundance of species that potentially form large colonies such as the Orbicella spp. (Figure 5). This was most apparent on AB at 20 m which was originally dominated by many large Orbicella colonies (Table S3). The observed shift in colony size of Orbicella, A. lamarcki and most other species is largely a result of partial mortality of large colonies. Coral growth or tissue regeneration may be impeded by increased spatial competition between corals and other benthic organisms such as algae or benthic cyanobacterial mats (Hughes, 1994). Meesters et al. (1997a) describe two types of lesion in coral tissue: the so-called type I, where the lesion is completely enclosed by tissue; and type II, most frequently observed, where the lesion is located at the edge of the colony. The observed increase in the ratio of colony outline and live surface area of coral colonies (Table 3) could be an indication of the incapability of the coral to reclaim substratum after such a type II lesion (Meesters et al., 1997b). Partial mortality will generally result in a larger loss of surface area than colony outline, thus leading to a higher shape-ratio. A possible effect on colony size distribution caused by increased recruitment can be neglected since no new recruits were present in 2014. M. mirabilis may at present be the sole species to show recruitment within the quadrats. Fragmentation, however, makes it difficult to distinguish actual new M. mirabilis recruits. Nevertheless, the relative increase in abundance of opportunistic species in the past decade has contributed to a higher density of small colonies (Knowlton, 2001; Darling et al., 2012).

An important consequence of reduced coral cover is the loss of reef structural complexity and associated loss of biodiversity, but also coastal protection and human food security (e.g., Alvarez-Filip et al., 2009; Newman et al., 2015). In many of the studied quadrats, coral cover has approached or dropped below the 10% level. This point has been suggested as the threshold after which carbonate accretion will be surpassed by erosion and architectural complexity is being lost (Perry et al., 2013). Following this rationale, nine out of the 16 main quadrats would currently be in an eroding state (Table 3). Indeed, the estimated carbonate production has decreased considerably since 1973 in virtually all quadrats, largely attributed to the general reduction of coral cover and in particular the loss of large calcifying species such as Orbicella spp. (Perry et al., 2015). The relative decline in carbonate production was highly comparable across depths, with the exception of 10 m. At 10 m depth the degree by which carbonate production changed over time was generally lower and largely driven by the temporal dynamics in cover of the rapidly calcifying M. mirabilis. High M. mirabilis cover may compensate some of the reduction in production but it cannot be considered an important framework builder able to compensate for the loss of structural complexity provided by large calcifying species (such as Orbicella spp. and Acropora spp.) (Alvarez-Filip et al., 2011a). Moreover, opportunistic rapidly calcifying species such as M. mirabilis are typically absent at greater depths.

Diversity indices may reflect a gradient of decreased disturbance with greater depth. According to the intermediate disturbance theory highest community diversity is found at intermediate disturbance (Connell, 1978). Over the depth range studied, highest diversity occurred at 20 m depth both in 1973 and 2014, which could be indicative of intermediate disturbance level. At 10 m depth, high disturbance likely impacted all but the most resistant species such as M. mirabilis or P. astreoides, resulting in a low diversity. Both natural (e.g., hurricanes) and anthropogenic disturbances (e.g., eutrophication and run-off) generally have the highest impact on the more shallow reef. Particularly the anthropogenic stress has intensified over the past decades likely putting more stress on the most shallow reef zones (<20 m). The generally low diversity of coral communities at greater depth (>20 m), in particular at 40 m, might indicate that these reefs experience lower levels of disturbance (e.g., declined water quality, reduced herbivory and hurricanes). Following the intermediate disturbance theory, reefs with low disturbance are characterized by dominant competitors (Connell, 1978). Indeed, the deeper reefs of Curaçao and Bonaire were dominated by A. lamarcki consistently between 1973 and 2014. In addition, low light intensities at greater depth likely support a generally lower diversity of species. Although, the reefs at greater depths may initially have experienced less disturbance (Bak et al., 2005) an obvious degrading trend was nevertheless observed, in particular since approximately 1990. A decline in cover and abundance of coral, including A. lamarcki may suggest that stress on these reefs has recently been increasing. Lapointe and Mallin (2011), for instance, found evidence for land-based sewage inputs reaching the deeper reef (18 m) near urban areas on Curaçao. Nutrient measurements on reefs below 18 m are virtually absent, but eutrophication may be reaching greater depths as well. The observed deep community dynamics of the recent years may, however, not be linked (anymore) to the absence of disturbance, but simply be the outcome of one strong competitor (A. lamarcki) and a generally limited pool of species at greater depths. The absence of weedy species such as M. mirabilis here may impede a shift to a more opportunistic coral community as seen on the shallow reef. Such species are simply unable to settle at these depths due to factors such as low light intensity. As yet no new opportunistic candidate has settled and with A. lamarcki, still dominant though steadily declining as well, the prospect for recovery of mesophotic coral communities on reefs appears to be very limited.

In conclusion, we found that coral communities of Curaçao and Bonaire have seen dramatic changes since 1973. Overall coral communities were heavily degraded by 2014. Phenotypically plastic and opportunistic species have increased and M. mirabilis, A. lamarcki, P. astreoides and D. strigosa may at present be considered very modest winners among the Caribbean coral assembly. These species, however, are unlikely to remain winners in a compromised habitat where suitable substratum for settlement is lost and disruptive events may become more frequent and severe, nor will they be able to provide the same structural complexity as the species from the near past (e.g. Orbicella spp.). The reefs at mesophotic depths (30–40 m) do not escape the general degradation of the coral community. Degradation was less evident away from urban areas. This suggests that local management, such as the provision of proper sewage treatment and run-off prevention, can have a positive effect on the general health and survival of coral reef communities. Although, the upsurge of some opportunistic species may provide a glimmer of hope for coral communities proper conservation and management will be essential to keep coral species from turning collectively into losers.

Data collection: RB, EM, and GN; Data analysis: DD, EM, RB, and GN; Interpretation of the data: DD, EM, FV, and RB; Drafting work: DD; Critical revision: EM, FV, RB, and GN; Final approval: EM and FV; Agreement to be accountable for all aspects of the work: DD, EM, FV, RB, and GN.