The implications of ocean acidification are acute for calcifying organisms, notably tropical reef corals, for which accretion generally is depressed and dissolution enhanced at reduced seawater pH. We describe year-long experiments in which back reef and fore reef (17-m depth) communities from Moorea, French Polynesia, were incubated outdoors under pCO₂ regimes reflecting endpoints of representative concentration pathways (RCPs) expected by the end the century. Incubations were completed in 3–4 flumes (5.0 × 0.3 m, 500 L) in which seawater was refreshed and circulated at 0.1 m s^-1, and the response of the communities was evaluated monthly by measurements of net community calcification (NCC) and net community photosynthesis (NCP). For both communities, NCC (but not NCP) was affected by treatments and time, with NCC declining with increasing pCO₂, and for the fore reef, becoming negative (i.e., dissolution was occurring) at the highest pCO₂ (1067–1433 µatm, RCP8.5). There was scant evidence of community adjustment to reduce the negative effects of ocean acidification, and inhibition of NCC intensified in the back reef as the abundance of massive Porites spp. declined. These results highlight the risks of dissolution under ocean acidification for coral reefs and suggest these effects will be most acute in fore reef habitats. Without signs of amelioration of the negative effects of ocean acidification during year-long experiments, it is reasonable to expect that the future of coral reefs in acidic seas can be predicted from their current known susceptibility to ocean acidification.

Common infrastructure

Experiments were conducted in flumes, consisting of a 5.0 × 0.3 m working section filled to ~ 0.3-m depth with 500L of seawater. Four flumes were used for both experiments, but equipment malfunction resulted in one flume being dropped from the analysis of the fore reef communities. See Supplementary Information in published article for more details.

Communities were exposed to treatments targeting ambient (~ 400 µatm), 700 µatm, 1000 µatm, and 1300 µatm pCO₂ that approximate atmospheric conditions expected by ~ 2140 under different RCPs. Experiments began in late Austral spring in 2015 (back reef) and 2017 (fore reef), and extended 12 months with monthly measurements of community metabolism (this study) and community structure.

The flumes were filled with seawater pumped from 14-m depth in Cook’s Bay and filtered through sand (~ 450–550 µm pore size), and was continuously added to the flumes at 300 L h^-1. Seawater was circulated in each flume at 0.1 m s^-1 using a pump (W. Lim Wave II 373 J s^-1), and flow speeds were measured in the working section using a Nortek (Boston, MA) Vectrino Acoustic Doppler Velocimeter. A flow speed of 0.1 m s^-1 is ecologically relevant to the back reef, and the fore reef at 15-m depth.

The seawater was temperature-controlled with chillers that matched the monthly temperatures to the ambient seawater in Cook’s Bay. Temperatures were increased from ~ 27 °C in November to ~ 29 °C around April and May, then back to ~ 27 °C the following November. Temperature was recorded using loggers (Onset Hobo Pro 2, ± 0.2 °C) and a certified thermometer (± 0.05 °C, model 15-077, Fisher Scientific). Sunlight was shaded with neutral density mesh to approximate the light at 2-m depth in the back reef, or 17-m depth on the fore reef, and for the fore reef, blue filters (LEE #183, Lee Filters, Andover, England) were used to simulate light at 17-m depth. Photosynthetically active radiation (PAR) was recorded using cosine-corrected loggers (Odyssey, Dataflow Systems, Ltd., Christchurch, New Zealand), calibrated against a certified Li-Cor cosine sensor (± 5% resolution, LI 192A, Li-Cor, Lincoln, NE, USA) in units of photon flux density (PFD, µmol photons m^-2 s^-1) Flumes were covered with clear UV-transparent acrylic lids to prevent rain from entering and to reduce wind speed at the air-seawater interface.

The 5-m working section was divided into a central 2.4-m portion and two flanking portions, each 1.3 m in length. The central portion included a 0.3-m deep sediment box that was filled with sediment for the back reef community, but was covered and sealed for the fore reef community where sand accumulations are rare. The sand in the sediment box for the back reef was arranged with a smooth surface, and it lacked the ripples in surface texture that can enhance advective exchange with seawater. Members of the benthic community from each habitat were scattered along the working section to create a community composition benchmarked against the empirical benthic community. Community members in the central portion of the flumes were attached to a metal grid and were not removed from the flume so that they could be quantified photographically. In the flanking portions, community members were removed monthly to measure buoyant weight that was used in the analysis of the ecological responses to treatments. The communities included a few invertebrates (e.g., coral ectosymbionts such as crabs and brittle stars) with herbivores (e.g., sea urchins) to aid in the control of algae. Routine cleaning (~ weekly) removed excess algal biomass on the walls and other exposed surfaces of the flumes.

pCO₂ regimes were assigned randomly to the four flumes in each experiment, and the flumes were oriented with their long axis aligned north-south. Seawater carbonate chemistry was not controlled in the ambient flumes, but in the other three flumes was regulated with CO₂ gas to approach target values. CO₂ was added to the treatment flumes to alter seawater pH relative to a set point (controlled with AquaControllers, Neptune Systems, Morgan Hills, CA) that operated a solenoid to regulate the supply of CO₂, except in the ambient flume; ambient air was bubbled into all flumes. The system was programmed to apply a nocturnal downward pH adjustment of ~ 0.1 unit to simulate conditions on the reefs of Moorea.

Seawater pH on the total hydrogen scale (pH_T) was measured daily using a hand-held meter (with a DG 115-SC electrode, Mettler Toledo, Columbus, OH, USA) that was calibrated with TRIS buffer (SOP 6a). These records were used together with temperature to adjust the pH set points of the AquaController to approach target values. Seawater carbonate chemistry was calculated weekly using pH_T and the total alkalinity (A_T) recorded once during the day and night. A_T was measured using open-cell, acidimetric titrations (SOP 3b) with an automatic titrator (T50 Mettler Toledo) operated with a DG 115-SC probe and filled with certified acid (A. Dickson, Scripps Institute of Oceanography [SIO]). The accuracy and precision of the analyses were determined from reference materials (CRMs, batches 158 and 172, A. Dickson, SIO). Relative to CRMs, determinations of A_T maintained an accuracy of 1.7 ± 0.3 µmol kg^-1 to 2.7 ± 0.4 µmol kg^-1, and a precision of 1.8 ± 0.1 µmol kg^-1 (n = 475). Calculations to determine seawater carbonate chemistry were made using temperature, salinity, and pH_T, in the R package Seacarb.

Calculations of benthic metabolism

Community metabolism was assessed using net community calcification (NCC) and net community productivity (NCP) that were calculated using the alkalinity anomaly method (Equation 1 in published paper), and changes in dissolved oxygen (DO see Supplementary Information) (for NCP, Equation 2 in published paper).

NCC and NCP were measured under ambient conditions in all flumes at the start of the experiments, the pCO₂ treatments were initiated, and NCC and NCP again were measured. For the back reef community, NCC and NCP were measured biweekly or monthly for the first four months (to March 2016) and then monthly until the end of the experiment. Single measurements every month revealed variance attributed to weather and, therefore for the fore reef community, NCC and NCP were recorded on three days every month, with the three days usually consecutive. Each day of measurements consisted of six determinations, two in the morning (06:00–09:00 hrs and 09:00–12:00 hrs), two in the afternoon (12:00–15:00 hrs and 15:00–18:00 hrs), and two at night (18:00–24:00 hrs and 24:00–06:00 hrs). These times were accurately recorded, but each period varied to accommodate logistical constraints. Each determination was completed with the flumes operating in closed-circuit mode, during which the inflow of fresh seawater was halted. The flumes were flushed for ~ 30 minutes between each incubation by re-initiating the inflow of seawater at ~ 5 L min^-1, thereby exchanging ~ 25–30% of the seawater between incubations.

For the back reef community, for which a single day of measurements was obtained monthly, missing values were interpolated using a third order polynomial (for NCC) or a fourth order polynomial (NCP) for empirical values against time of day (Carpenter et al. 2018). NCC and NCP were calculated as hourly rates (mmol CaCO₃ m^-2 h^-1 and mmol O₂ m^-2 h^-1, respectively) with positive values reflecting net CaCO₃ accretion and a net release of photosynthetically-fixed O₂, and negative values reflecting net dissolution of CaCO₃ and a net uptake of O₂, all respectively. NCC was integrated over the day and night to gain insight into the relative effects of treatment on carbonate accretion (that occurs mostly during the day) versus dissolution (that occurs mostly at night), and NCP was integrated over the day to evaluate net photosynthesis. Local day length was used to calculate day and night rates, with day length varying from 13.17 hrs in December to 11.08 hrs in June.

Back reef experiment, 2015-2017

The back reef experiment began in November 2015, and the communities were assembled to mimic the community found in the back reef of Moorea in 2013. Community members were collected from ~ 2-m depth on the north shore. Initially the communities had ~ 25% coral cover with 11% massive Porites spp., 7% P. rus, 4% Montipora spp., 3% Pocillopora spp., and ~7% crustose coralline algae (CCA) with 4% Porolithon onkodes, and 3% Lithophyllum kotschyanum. Corals and CCA were attached to plastic bases (using Z-Spar A788). Coral rubble (~ 1-cm diameter) was added to ~ 5% cover, and the remainder of the floor of the flume was sand. Communities were assembled on top of the sand and were augmented with a few holothurians (~ 8-cm long, Holothuria spp.) and macroalgae (Turbinaria ornata and Halimeda minima) that were added to ecologically relevant cover for the back reef in 2013 (~ 4–5% cover). The communities were established in the flumes by 12 November 2015, and were maintained under ambient pCO₂ for 5 days until pCO₂ treatments were established in three flumes; pCO₂ was increased over 24 h to establish target values.

Fore reef experiment, 2017-2018

The fore reef experiment began in November 2017, and communities were assembled in the flumes to mimic the community found at 17-m depth on the fore reef of Moorea in 2006. An historic community was used as a target because it better reflected the long-term representative fore reef community in this location, and because it allowed a more direct comparison with previous short-term incubations completed with fore reef communities. Community members were collected from 17-m depth on the north shore fore reef. Initially, the communities had ~ 27% coral cover with ~ 11% Pocillopora spp., ~ 8% massive Porites spp., ~ 8% Acropora spp. and 53% reef rock. The Pocillopora conformed to P. verrucosa, but multiple species probably were present. Likewise, the Acropora spp. targeted A. hyacinthus and A. retusa, which are common on the fore reef of Moorea, but possibly included other species. Corals were attached to plastic bases (with Z-Spar A788). The reef rock consisted of rubble (~ 12-cm diameter) and the flora and fauna with which it was associated. The communities were established in the flumes by 27 October 2017 and were maintained under ambient pCO₂ for 7 days, when the pCO₂ treatments were initiated over 24 h to reach target values. Because of equipment malfunctions, the fore reef experiment was completed with 3 flumes that targeted ambient conditions (400 µatm), 700 µatm, and 1300 µatm pCO₂.

Statistical analysis

NCC and NCP were analyzed separately for each experiment, and in both cases, linear, mixed-effects models were fitted using restricted maximum likelihood (REML) methods, in which treatments (pCO₂ regimes) and time (i.e., measurement day) were fixed effects (covariates), and light, temperature, and coral cover were random covariates. Four analyses were conducted for each experiment, with each employing a different dependent variable (24h-NCC, day-NCC, night-NCC, and NCP). For the fore reef experiment, the 3 measurements month^-1 were averaged to provide a single value for each month to facilitate a contrast with the back reef experiment. To evaluate changes in metabolism over time, the average of the monthly measurements was registered against the second of the three-day measurement sequence. Light was integrated over a day (mol photons m^-2 d^-1) on the day that community metabolism was measured (back reef), or was averaged over the 3 measurement periods (fore reef). Temperature was the mean daily temperature over the month preceding the metabolism measurements, and coral cover was calculated from planar photographs.

To understand the integrative balance between inorganic (NCC) and organic (NCP) carbon sequestration, the relationships between NCC and NCP (by hour) were explored using scatter plots, with associations tested using Pearson correlations. Linear relationships were displayed using Model I regressions. To test for variation in NCC–NCP relationships over time, the slopes and elevations of the NCC–NCP relationships were compared among seasons using ANCOVA. The seasonal contrast was created by segregating results from November–February (near Austral summer), March–May (near austral autumn), June–August (near Austral winter), and September–November (near Austral spring).

To test for linear relationships between 24h-NCC and pCO₂, Model I linear regressions were fit to the results by month (n = 3–4 pCO₂ treatments) using the single measure month^-1 for the back reef experiment, and the 3 measures month^-1 for the fore reef experiment. To test for variation over time (year) in the sensitivity of the 24h, NCC–pCO₂ relationship, Model I linear regression was used, in which the slopes of the 24h-NCC–pCO₂ relationships by time were the dependent variable, and the independent variable was time (days over the incubation year).

Statistical analyses were completed using Systat 13.0 software (Inpixon, Palo Alto, USA), and the statistical assumptions of ANOVA were tested using graphical analysis of residuals.