Fore reef communities were assembled in three outdoor flumes in Mo'orea, which were assigned randomly to a pCO₂ treatment targeting ambient (400 μatm), 700 μatm, and 1300 μatm pCO₂. The elevated pCO₂ treatments approximated atmospheric conditions projected for about the year 2140 under representative concentration pathways (RCP) 2.6, 4.5, and 8.5. Treatments were maintained for one year beginning in late Austral spring (November 2017), and actual pCO₂ treatments over the year differed from target values (described below). In brief, each flume consisted of a working section that was 5.0-m long, 30-cm wide and filled to ~ 30-cm depth with ~ 500 L of seawater. The fixed and unfixed communities within each fume occupied a 4.7 × 0.3 m portion of the floor of the working section of each flume. Seawater was circulated continually through a return section, and was supplied with fresh seawater at ~ 5 L min^-1. Seawater was pumped from Cook’s Bay (14-m depth) and filtered through sand (pore size ~ 450–550 µm) before entering the flumes. With this pore size, small particulates passed through the filter and were added to the flumes where they were available as food for heterotrophic organisms.

Fore reef communities:

The reef communities were assembled to correspond to the mean percent cover of the major benthic space holders recorded in 2006 at 17-m depth on the fore reef of the north shore of Mo'orea. A historic community structure (rather than present day) was used because 2006 represented the long-term community structure on this reef, and it created the capacity to compare aspects of the present experiment with a previous experiment. Based on six sites sampled around Mo'orea in 2006, the community structure in the flumes was targeted to ~ 11% cover of Pocillopora spp., ~ 8% massive Porites spp., 8% Acropora spp., and ~ 53% reef rock. This construct created a community with ~ 27% coral cover, which was slightly lower than the actual mean coral cover in 2006 (32%), because the remaining 14 genera of scleractinians and Millepora contributed 5% coral cover. The Pocillopora conformed to the classic morphology of P. verrucosa, but it is likely that other Pocillopora spp. were present in the flumes. Likewise, Acropora spp. were selected to represent A. hyacinthus and A. retusa, which were common on the fore reef when the experiment was completed, and colonies of these species were scattered haphazardly among the flumes. Given the morphological complexity of Acropora spp., it is possible that other species were placed into the flumes. Pieces of coral rubble (~ 11.5-cm diameter) were added to achieve ~ 29% cover. Coral and rubble were haphazardly scattered along the working section of each flume to approach the targets for percentage cover, and this resulted in portions of the flumes having slightly different covers of coral. This was important for the central 2.4-m portion of the flume, where community members were fixed to allow the community structure to be quantified monthly using planar photographs. In the adjacent portions of the flumes, community members were unfixed (and rested on the floor of the flumes) so that they could be removed monthly for buoyant weighing (described below). 

Corals and rubble were collected from ~ 17-m depth on the north shore fore reef, epoxied (Z-Spar A788, Pettit Marine Paint, Rockaway, NJ, USA) to plastic bases, and placed in a seawater table for at least 2 d before being added to the flumes. This time allowed the epoxy to cure and for the corals to recover from collection. Fore reef communities were assembled in the flumes on 27 October 2017, where they were maintained under ambient seawater conditions until 3 November. At this time, treatment pCO₂ levels were initiated in two flumes (one remained at ambient pCO₂), with pCO₂ gradually increased to target values over 24 h.

Physical and chemical parameters (see "Supplemental Files" for data access):

Seawater was circulated in the flumes at ~ 0.1 m s^-1 using a pump (Wave II 373 J s^-1, W. Lim Co., El Monte, CA, USA), and flow speeds were measured across the working sections using a Nortek Vectrino Acoustic Doppler Velocimeter. This flow speed was ecologically relevant for 15-m depth on the fore reef of Mo’orea (14-y mean = 0.065 m s^-1). The flumes were exposed to natural sunlight that was reduced with a blue filter (LEE #183, Lee Filters, Andover, England) to photon flux densities (PFD) in the range of photosynthetically active radiation (400–700 nm) that approximated those at 17-m depth. Light in the flumes was measured continuously (at 0.0006 Hz) using cosine-corrected sensors (Odyssey, Dataflow Systems Ltd, Christchurch, New Zealand) that recoded PAR. Odyssey sensors were calibrated with a Li-COR meter [LI-1400, Li-COR Biosciences, Lincoln, NE, USA] attached to a 2p sensor [LI 192A]). Temperatures in the flumes were regulated with chillers (heaters were not required) and were maintained close to the mean monthly seawater temperature at 17-m depth on the fore reef.

Seawater carbonate chemistry was uncontrolled in one flume (ambient, ~ 400 μatm pCO₂), and controlled in two others to simulate conditions arising from seawater pCO₂ targeted at 700 μatm and 1300 μatm. Seawater pH was not altered in the ambient flume, but was controlled in the treatment flumes by bubbling CO₂ into the seawater to alter pH relative to a set-point (regulated using an Aquacontroller, Neptune Systems, Morgan Hill, CA, USA) that operated a solenoid supplying pure CO₂ gas to a diffuser stone submerged in each flume. A diurnal upward pH adjustment of ~ 0.1 unit was applied to the two treatment flumes to simulate natural diurnal variation in seawater pCO₂ on the reef of Mo’orea. The ambient flume also maintained a diurnal variation in pCO₂ with a night time pH ~ 0.1 unit lower than in the daytime. Ambient air was bubbled continuously into all flumes. Periodic measurements of pCO₂ in the flumes confirmed that nocturnal pCO₂ met, or exceeded day-time target values (described in results).

Throughout the experiment, logging sensors (described above) recorded PAR, and temperature (Hobo Pro v2 [± 0.2 °C], Onset Computer Corp., Bourne, MA, USA). pH was measured daily on the total hydrogen ion scale (pH_T) using a handheld meter (see below). The values from the temperature and pH measurements were used to adjust the thermostat and pH-set points to achieve target pCO₂ values. Seawater carbonate chemistry (pH and A_T) and salinity were measured during the day (14:00 hrs) and night (20:00 hrs) and were obtained weekly. A bench-top conductivity meter (Thermo Scientific, Orionstar A212, Waltham, MA, USA) was used to measure salinity. The remaining parameters of the seawater carbonate system were calculated from temperature, salinity, pH_T, and A_T, using the R package Seacarb.

pH_T was measured using a DG 115-SC electrode (Mettler Toledo, Columbus, OH, USA) that was calibrated with a TRIS buffer. A_T was measured using open-cell, acidimetric titration using a certified titrant with an automatic titrator (T50, Mettler Toledo) fitted with a DG 115-SC electrode (Mettler Toledo). The accuracy and precision of measurements were determined by processing certified reference materials (CRMs batch numbers 158 and 172; from A. Dickson Laboratory, Scripps Institution of Oceanography, CA, USA), against which measured values of A_T maintained an accuracy of 1.7 ± 0.3 μmol kg^-1 (n = 15) and precision of 1.8 ± 0.1 μmol kg^-1 (n = 475).

Response variables:

See more details about coral weight and Gnet calculation in the related dataset "Edmunds et al. 2020 ICES: pCO2 flume - Coral weight" https://www.bco-dmo.org/dataset/924603.  

Community structure:
* This section describes the methodology of the "percent_area" column in this dataset.
* The images of each flume (1-3) are in the "Supplemental Files" section of this dataset.

The effects of the treatments on the community structure were described using photographs recorded monthly in planar view. The image-based technique strengthened the ability to address the effects of OA on the community ecology of coral reefs, which frequently is recorded using planar photographs (including in Mo'orea). Photographs were recorded in ambient light using a GoPro Hero 4 camera (12 MP, 3-mm focal length) that was fitted to a stand and positioned on the upper edge of the flumes to record the benthic community through the air-water interface. At each sampling, the camera was sequentially moved along the flume to record the community in the middle 2.4 m of the working section using ~ 15 contiguous photographs.

Photographs were analyzed using ImageJ software after they were stitched together to make a single image for each sampling. This image covered the ~ 2.4-m length of the central portion of the flume where the corals were secured to a plastic-coated metal grid with a mesh size of 5 × 5 cm. The stitching of photographs sometimes was imperfect due to parallax errors, and in such cases, separate pictures were evaluated to assess organism size. The planar area of living tissue on corals was quantified by outlining organisms in ImageJ, after scaling the image using the metal grid as a size reference. Organism size (cm²) was expressed as a percentage of the area (7200 cm²) occupied by the fixed portion of the community. The summed area of community members was used to determine cover of the benthic community, and the areas of each organism were used to quantify growth (and shrinkage). Where organisms died, their area was set to zero.

Organism identifiers (Taxon, LifeSciences Identifier (LSID), name as appears in Species column):

Acropora hyacinthus, urn:lsid:marinespecies.org:taxname:207044, A. hyacinthus
Acropora retusa, urn:lsid:marinespecies.org:taxname:430653, A. retusa
Porites, urn:lsid:marinespecies.org:taxname:206485, Massive Porites
Pocillopora damicornis, urn:lsid:marinespecies.org:taxname:206953,
Pocillopora verrucosa, urn:lsid:marinespecies.org:taxname:206954,