Experimental reef communities persist under future ocean acidication and warming

Coral reefs are among the most sensitive ecosystems affected by ocean acidication and warming, and are predicted to shift from net accreting calcier-dominated systems to net eroding algal-dominated systems over the coming decades. Here we present a long-term experimental study examining the responses of entire mesocosm coral reef communities to acidication (-0.2 pH units), warming (+ 2°C), and combined future ocean (-0.2 pH, + 2°C) treatments. We show that under future ocean conditions, net calcication rates declined yet remained positive, corals showed reduced abundance yet were not extirpated, and community composition shifted while species richness was maintained. Our results suggest that under Paris Climate Agreement targets, coral reefs could persist in an altered functional state rather than collapse. Environmental monitoring: CPJ with assistance from KDB, EWB, LS, and TT. Coral calcication, bleaching, and survivorship, as well as rubble calcication: CPJ. Mesocosm calcication: CPJ with assistance from KDB, EWB, and LS. Coral colony genotyping: AC and RJT. Zooplankton:


Main Text
Many studies project that the combination of ocean acidi cation and warming will lead to the functional collapse of coral reef ecosystems at a global scale over the next few decades, driving major losses in biodiversity and ecosystem services [1][2][3][4][5] . Reefs are expected to shift from net calci cation to net carbonate dissolution sometime later this century when atmospheric CO 2 reaches 550-650 µatm, seawater pH declines by 0.1-0.15 units, and seawater temperature increases by 1-1.5 °C 4,6 . Likewise, corals are expected to be essentially extirpated from reefs in the next few decades (>99% decline) 7,8 once coral bleaching becomes an annual phenomenon. These projected future scenarios are typically based on either short-term laboratory perturbation experiments with few species that are then scaled up to longterm responses of complex communities in nature 3,4,9,10 or in situ observations of coral reefs that span natural gradients in chemistry or thermal stress [11][12][13][14] . Laboratory experiments, however, typically do not include natural ecological interactions among species which could affect community function, and some experiments may not have provided organisms with su cient time to fully respond to the treatments. Likewise, natural gradient studies do not re ect the intensity of heat stress expected later this century and often do not incorporate acidi cation and warming simultaneously (but see [15][16][17] ). An alternative approach to lab or natural gradient studies is the use of mesocosms which allow diverse communities to be exposed to future ocean levels of acidi cation and warming at the same time while preserving realistic ecological interactions and environmental conditions [18][19][20][21][22][23] . While no study can incorporate every aspect of future reef ecosystem composition and function, mesocosms provide some of the most complete experimental systems to test the hypothesis that sustained future ocean acidi cation and warming will induce the functional collapse of experimental reef communities. With eight coral species representing three divergent families (Acroporidae, Pocilloporidae, and Poritidae), both major evolutionary lineages of scleractinians (Complexa and Robusta), and all four of their major life history strategies (see Supplementary Information), along with the literally thousands of other species surveyed in our study (ranging from bacteria and archaea to many of the major animal and algal phyla), we present a long-term experimental examination of coral reef organismal and community responses under future ocean conditions. This dataset provides unprecedented insights into the likely responses of coral reef communities to ocean acidi cation and warming over coming years.
We conducted a two-year ecologically realistic outdoor ow-through mesocosm experiment at the Hawai i Institute of Marine Biology to examine the long-term responses of biologically diverse coral reef communities to chronic acidi cation and warming similar to levels projected to occur toward the end of this century given current commitments under the Paris Climate Agreement (conditions between Representative Concentration Pathways RCP 6.0 and RCP 8.5) 24 . Mesocosms were initially stocked with replicate communities of the regionally most common reef-building corals, each sourced from multiple locations around the island of O'ahu, as well as reef rubble, reef sand, algae, invertebrates, and sh (see Supplementary Information). We determined the multilocus genotype of each coral colony sampled to ensure results were not biased by inclusion of clones. Replicate clonal fragments (ramets) of each genetically unique coral colony (genet) were included in all four treatments. Over time, the mesocosms recruited a diverse assemblage of algae, invertebrates, and microbes (Fig. 1). These communities developed under one of four treatments with 10 mesocosms per treatment: control treatment (present-day pH and temperature), ocean acidi cation treatment (present-day temperature with acidi cation of -0.2 pH units relative to control), ocean warming treatment (present-day pH with elevated temperature of +2 °C relative to control), or combined future ocean treatment (both acidi cation of -0.2 pH units and elevated temperature of +2 °C relative to control). All mesocosms experienced natural diurnal and seasonal variation in temperature, chemistry, and irradiance (Table 1, Fig. 2, S1). After approximately two years of exposure, we assessed (i) coral bleaching, survivorship, recruitment, skeletal extension, metabolism, and energy reserves, (ii) net calci cation by the mesocosm communities, coral communities, and rubbleassociated communities, and (iii) community structure and species richness of the major functional groups, including benthic algae and invertebrates, coral-associated algal endosymbionts (Symbiodiniaceae), coral-associated microbes, and water column-associated microbes (Fig. S2). Table 1. Carbonate chemistry and temperature from the experiment. Data are daily mean values derived from weekly sampling at 1200 hr as well as monthly sampling every 4 hr over the diel cycle (see Supplementary Information) and are shown as mean±SD. The uncertainties associated with these values re ect daily and seasonal variability, as well as variability among replicate mesocosms in each treatment. The mean uncertainties among mesocosms on a given sampling day are provided in parentheses. Note that the variation among mesocosms is relatively small and most of the variation is explained by daily and seasonal uctuation of these parameters. See Fig. 2, S1 for additional environmental information. Corals in the treatments with elevated temperature (ocean warming and combined future ocean treatments) were exposed to severe 3 heat stress in successive years. These corals experienced temperatures at or above the nominal bleaching threshold for 3.5 months per year, during which they accumulated 24 Degree Heating Weeks (DHW) annually (Fig. 2). Many studies predict that this level of repeated annual bleaching stress should have been more than su cient to extirpate corals in our elevated temperature mesocosms 1,3,7,8 and that acidi cation should have exacerbated the heat stress 3 .
As expected, many of the corals that bleached severely subsequently died during the trials, with up to 13% of the corals in the heated treatments remaining pale or bleached at the end of the experiment, compared to less than 2% under the control and ocean acidi cation treatments (Fig. S3, Table S1). Contrary to projections of near total mortality 7,8 , coral survivorship was reduced by only 35% in the heated treatments compared to the present-day temperature treatments, with no evidence that acidi cation affected either survivorship or the incidence of bleaching (Fig. S3, Table S1). Among the survivors, the extent of paling, however, was 14-18% higher in the ocean acidi cation, ocean warming, and combined future ocean treatments relative to the control (Fig. S3, Table S2). Indeed, some individual corals in this study bleached and died, whereas others bleached annually yet survived to the end, and still others never bleached at all. These responses differed by coral genus. All individuals of both species of Pocillopora paled or bleached, though some survived, whereas responses within the other genera were more variable and some individuals within each of the three Montipora species and the three Porites species exhibited all of these response patterns. Coral species richness declined under warming, primarily due to the loss of one or both Pocillopora species from many of the heated mesocosms (Table S3). Some corals are capable of increasing zooplankton feeding rates to survive and recover from bleaching 25 . Although we used un ltered sea water and shes in the mesocosms were fed to simulate natural zooplankton import to the communities (see Supplementary Information), live zooplankton densities in the mesocosms were at the lower end of values typically reported for reefs (Fig. S4), making our results potentially conservative. Among the surviving corals, horizontal skeletal extension was reduced 22% in the heated mesocosms (ocean warming and combined future ocean treatments) relative to those under present-day temperatures (control and ocean acidi cation treatments), but was unaffected by acidi cation (Fig. S5, Table S3). In contrast, total organic carbon uxes increased under warming (Table S3, Fig. S4). Coral photosynthesis, respiration, and energy reserves, however, did not differ signi cantly among those individuals that survived the treatments (Fig. S5, Table S3). While severe, this 35% reduction in survivorship and 22% decline in skeletal extension among the survivors is far less than the >99% loss of corals projected for reefs 7,8 . The discrepancy between our data and these projections may re ect some of the challenges associated with scaling up prior results. In particular, few projections quantitatively account for the substantial variation in bleaching tolerance we observed within and among species or the biotic interactions which could modulate these outcomes 26 .
Large numbers of one coral species (Pocillopora acuta) recruited into the mesocosms, undoubtedly due to spawning of the adult corals housed in them 18,27 , but recruitment rate was unaffected by any of the treatments 27 . Similarly, we observed widespread spawning of another coral species (Montipora capitata) across all four treatments (CPJ pers. obs. in June 2018), though it was infeasible to assess reproductive output or gamete quality at the time of release. Hence, many of the corals grew from the size of prereproductive juveniles to reproductive adults (and completed a substantial fraction of their life cycles) over the course of the experiment. If coral individuals that are resistant to bleaching or resilient to annual bleaching (like some of those in this study) proliferate in the future, then they could help to offset some of the negative effects of ocean warming on coral abundance predicted for coral reefs. Indeed, coral communities in Hawai'i already appear to be mounting adaptive responses to climate change with bleaching and mortality occurring at higher temperatures and after longer exposures than reported 50 years ago 17,28 .
The calci cation rates measured in the control mesocosm communities were very similar to those measured on the nearby reefs 29 , indicating that the mesocosms adequately replicated the processes involved in community calci cation. Net calci cation of the mesocosm communities (sometimes referred to as net community calci cation, NCC, or net ecosystem calci cation, NEC, in other studies) declined in all treatments relative to the control, with the largest decline under the combined future ocean scenario ( Fig. 3, Table S3). The 19-24% reduction in mesocosm calci cation attributable to acidi cation is lower but roughly similar to the 30% reduction measured on an experimentally acidi ed reef at 30 , further illustrating the e cacy of our approach to simulate the natural system. Nevertheless, all communities continued calcifying. Even under the combined future ocean treatment, reef community calci cation was positive, albeit at only 56% the rate of control reef communities. At present-day rates of calci cation, however, few reefs are expected to accrete fast enough to be able to keep up with sea level change and many future reefs may become submerged as the oceans rise 31 . Acidi cation does not by itself kill corals but rather tends to inhibit their skeletal growth by an average of 15-20% 10 , which may compromise their competitive abilities in nature 11,12,15 . Acidi cation, however, had no effect on net coral community calci cation or skeletal extension within this study (Fig. 3, Fig. S5, Table S3). While these results differ from many prior laboratory experiments, both ex situ and in situ studies have found that some corals can maintain normal calci cation rates under lower pH 12,13,17,[32][33][34] . Further, irradiance and water ow are both known to affect coral responses to acidi cation 19 . We conducted this experiment using natural sunlight (attenuated by shade cloth to ambient levels at mean collection depth of 2 m), rapid un ltered natural seawater turnover rate (1 hr), and additional water circulation provided by seawater pumps (10-15 cm s -1 ) to replicate light and ow conditions on the natural reefs as closely as possible (Fig. 2), which may help to explain the observed insensitivity of coral calci cation to low pH relative to many previous laboratory studies. In addition, corals may show threshold responses to acidi cation such that they are able to maintain calci cation rates under a 0.2 pH unit reduction yet experience reduced calci cation rates at higher levels of acidi cation 17,18,35 . Unlike acidi cation, elevated temperature reduced coral community calci cation and coral abundance by nearly half due to bleaching, mortality, and reduced skeletal extension among the survivors (Fig. 3, Table S3). In contrast, net calci cation by rubbleassociated communities declined under ocean acidi cation conditions yet was insensitive to warming (Fig. 3, Table S3). These results suggest that the measured reductions in calci cation for mesocosm communities 22,23,36 and natural communities 6,30,37 due to acidi cation are driven largely by processes occurring within the reef framework and soft sediments rather than by the corals themselves. The calci cation budget of the mesocosms exceeded that explained by the corals and rubble, and this additional carbonate production was likely from the growth of coralline algae, coral recruits, and other organisms which formed thick, calci ed crusts on the mesocosm walls (Fig. 1). Future reefs will undoubtedly experience a major decline in growth due to the loss of corals from heat stress, and reduced calci cation by the reef framework under acidi cation. Nonetheless, our ndings indicate that under mitigated CO 2 emissions consistent with current commitments under the Paris Climate Agreement, reefs may be capable of maintaining positive carbonate balances and net calci cation despite these climate change stressors.
Corals are ecosystem engineers, yet coral reef biodiversity is derived largely from the array of algae, invertebrates, and microbes which live within, among, and upon the reefs. Coral reefs occupy less than 0.2% of the sea oor but are home to an estimated 32-38% of all marine species 38 . Yet almost nothing is known about how this biodiversity will respond to ocean acidi cation, warming, or the combination of both factors. To determine how algal, microbial, and non-coral invertebrate composition varied within each treatment, at the end of the experiment we 1) retrieved 3-tiered settlement tile arrays (modi ed Autonomous Reef Monitoring Structures 39 ) which had recruited diverse benthic assemblages while soaking in the mesocosms for the duration of the experiment; 2) sampled the coral-associated algal endosymbionts and coral-associated microbes, as well as the water column-associated microbes; and 3) sampled the mesocosms for benthic, eshy algae. The settlement tile arrays mimicked the threedimensional structure of the reef framework and provided a standardized tool with which to examine this often-overlooked cryptobenthic community. The settlement tiles provided four data sets: 1) benthic cover by functional group from tile image analysis; 2) sponge species richness from individual morphological identi cation with DNA barcoding con rmation; 3) coralline algal richness from individual morphological identi cation; and 4) overall metazoan richness from DNA metabarcoding of the homogenized biomass scraped from each tile array [39][40][41] . Given the relatively short reproductive cycles of many algae and invertebrate species (weeks to months), they experienced multiple generations over the course of the experiment, providing a time-integrated measure of the treatment effects on community composition and abundances. Throughout the course of the experiment the benthic community transitioned from early colonizing species to a mature and diverse community that underwent seasonal variation in abundance similar to adjacent reef communities (JV, KDB, MAT, CPJ, and RJT, pers. obs.) Benthic cover analyses revealed that only a subset of the functional groups responded signi cantly to the treatments. Calcifying vermetid gastropods declined under acidi cation but increased under warming (Fig. 4, Table S3). Calcifying coralline algal cover increased with warming, non-calcifying turf algal cover decreased with warming, but neither group responded to acidi cation (Fig. 4, Table S3). Encrusting green algal cover increased in the combined future ocean scenario relative to the other treatments (Fig. 4, Table  S2). Other calcifying taxa (including bivalves and serpulid worms) were unaffected by either low pH or elevated temperature (Fig. 4, Table S2,& S3). The lack of an acidi cation effect on the benthic cover of most calcifying taxa is unexpected considering that individual species are often particularly sensitive to reduced pH in other studies 9 . Overall benthic community structure, however, was unaffected by pH, indicating that it is not changes in the dominance of functional groups, but rather the relative sensitivity of species within functional groups that alter this structure. Benthic functional group community structure differed only by temperature, and this effect was driven largely by separation of the control and ocean warming treatments in a community ordination (Fig. 5, Table S4).
Species richness of coralline algae and most other heavily calci ed taxa were not signi cantly different among treatments 41 , whereas the richness of some non-calci ers declined under acidi cation (Table S3).
Total metazoan richness on the settlement tiles was signi cantly reduced by acidi cation, signi cantly enhanced by warming, and these two factors offset each other in combination, resulting in species richness estimates that were not signi cantly different between the control and the combined future ocean scenarios 41 . The water column-associated microbial species composition and richness in the mesocosms did not differ signi cantly among treatments (Table S4). In contrast, the coral-associated microbial communities showed higher richness under warming and shifted structure depending on both temperature and pH (Table S3 & S4). Richness of coral-associated algal symbionts (Symbiodinaceae, resolved to the genera Cladocopium and Durusdinium) increased under warming (ocean warming and combined future ocean treatments) because a higher proportion of the Montipora corals that survived those treatments hosted both symbiont genera rather than Cladocopium alone (Table S1). The proportional contribution of each type to the overall symbiont community, however, was unchanged by any of the treatments (Table S1). Likewise, eshy algal species richness did not respond to either warming or acidi cation (Table S3). Hence, taxonomic and functional groups showed variable responses to the experimental treatments (Fig. 4). However, when all taxonomic datasets (sponges, coralline algae, and metabarcoding of metazoans from settlement tiles, coral-associated microbes, water columnassociated microbes, coral-associated algal endosymbionts, eshy algae, and corals) were pooled to examine the effects of low pH and elevated temperature on proportional changes in overall species richness, the number of species was not signi cantly affected by any of the treatments (Fig. 5, Table S1).
Our experiments support predictions that future reefs will have less coral and reduced calci cation rates. Yet, when experimental mesocosms were exposed to future ocean levels of simultaneous acidi cation and warming, consistent with current commitments under the Paris Climate Agreement (but in the absence of local stressors), we did not observe coral extirpation, reef community collapse, or net community carbonate dissolution. Instead, responses to ocean acidi cation, warming, and the combined future ocean stressors varied greatly among taxa, and many showed no signi cant effect of the treatments. Despite repeated severe annual heat stress (24 DHW annually), overall species richness was unchanged in future ocean conditions relative to present-day. The species richness or benthic cover of calci ers (such as coralline algae) that have been predicted to do poorly based on previous work instead showed unexpected persistence, while some non-calci ers (such as turf algae) declined under low pH or elevated temperature. Our results highlight the challenges of scaling up single-species studies to predict the outcome in diverse reef communities. Higher levels of acidi cation and warming, either due to unmitigated CO 2 emissions or during periodic marine heatwaves, will undoubtedly result in more severe effects on reef ecosystems and reduced capacity to provide valuable ecosystem services.
The results of our long-term experimental study, which includes the most diverse spectrum of organismal and community responses examined to date, demonstrate that future ocean acidi cation and warming pose grave threats to coral reef ecosystems, and that unmitigated CO 2 emissions will undoubtedly result in severe damage to these iconic ecosystems. However, these results also suggest that future reefs are likely to maintain a degree of function and avoid global collapse under realistic levels of ocean acidi cation and warming so long as local stressors can be mitigated 42 . Our study predicts that the community structure of reefs under future ocean warming will shu e signi cantly, undoubtedly altering ecosystem function, but the ultimate outcome of such changes depends on species interactions and the speci c ecological functions affected. Previous studies have underscored the importance of also mitigating local stressors for maintaining reef resilience 17,43,44 , especially because such local stressors increase the magnitude of coral loss from marine heatwaves 42 . The reef communities in these mesocosms were free from local anthropogenic stressors such as destructive shing practices, sedimentation, and coastal pollution, which may have facilitated their persistence in the face of future ocean conditions. Overall, our results support projections that scleractinian corals will have lower abundance on future reefs, but also suggest that they will not be extirpated, that reef calci cation will decline, yet will remain positive, and that a substantial fraction of reef biodiversity and ecosystem function could be preserved, even in the face of climate change, if serious action is taken to limit local human stressors on these iconic ecosystems.

Declarations
Acknowledgments: Thanks Coralline algae richness: CAS with assistance from CPJ and CMS. Settlement tile metabarcoding: MAT, with assistance from JV, MKW, KDB, ATG, CPJ and RJT. Coral-associated algal symbionts: MRdS with assistance from CPJ and RJT. Water column-associated microbes: WS with assistance from CEN. Fleshy algae: NHY with assistance from CPJ and CMS. The rst draft of the paper was written by CPJ. All authors contributed to discussion, data interpretation, manuscript revisions, and all approve the nal manuscript.
Competing interests: Authors declare that they have no competing interests.
Data and materials availability: These data will be submitted to an online repository to enable open access with a persistent link and DOI. For the purpose of review, the datasets generated and analyzed herein are available from the corresponding author upon request.  Mesocosms were covered in 30% shade cloth to replicate irradiance at mean collection depth (2 m).
Maximum instantaneous irradiance was about 1730 µmol m-2 s-1 in the mesocosms and 2470 µmol m-2 s-1 in the air. Water circulation in the mesocosms was provided by seawater pumps (10-15 cm s-1) and seawater turnover rate was 1 hr. See Table 1 and Fig. S1 for additional chemistry data, and see the Supplementary Information for additional details.  Table S3 for test results.   Benthic community structure and overall species richness in the mesocosms. (a) Non-metric multidimensional scaling ordination illustrating the effects of ocean acidi cation and warming on benthic community structure from settlement tiles colonized in the mesocosms for the control (light blue), ocean acidi cation (dark blue), ocean warming (light red), and combined future ocean treatments (dark red) (n = 6 tile arrays per treatment). Ellipses overlaid on scatterplots show the standard deviation around the centroid for each group. Community structure differed according to temperature but not pH, which was driven largely by separation of the control and ocean warming treatments. (b) Treatment effects on overall species richness in the mesocosms (data shown as proportional variation in species richness relative to the maximum observed richness among treatments) derived from samples of sponges, coralline algae, and metabarcoding of metazoans from settlement tiles, coral-associated microbes, water column-associated microbes, coral-associated algal endosymbionts, eshy algae, and corals (n = 8 datasets representing thousands of species). Box-plots show the median as center line, box limits are upper and lower quartiles, whiskers are 1.5x interquartile range, and there were no outliers. Treatment effects were not signi cant. See Table S1, & S4 for test results.