Measuring coral calcification under ocean acidification: methodological considerations for the ⁴⁵Ca-uptake and total alkalinity anomaly technique

Coral preparation and acclimation to pH treatment

Stylophora pistillata was collected by scuba diving from a depth of 4–8 m near the Interuniversity Institute for Marine Sciences (IUI) in the Red Sea, Eilat, Israel (29°30′N 34°55′E), in an area supervised by the Israel Nature and Parks Authority. Only a single colony was used for the calcification measurements in order to remove the effect of genetic variability for the tested treatments. The colony was fragmented into 36 fragments and these were suspended on nylon threads in flow-through water tables supplied with natural seawater under a constant temperature of 25 ± 0.5 °C, regulated by an array of 300 W BluClima aquarium heaters (Ferplast Spa, Vicenza, Italy) in an air-conditioned room. This setting allowed the tissue to grow over the exposed skeleton, creating microcolonies (definition sensu Almoghrabi, Allemand & Jaubert, 1993). All pieces were checked to be free of boring organisms. Light was provided by a metal halide lamp (14,000 K, 400 W/D, Osram GmBH, Germany) and photosynthetically active radiation (PAR) measured as ∼170 µmol photons m⁻² s⁻¹ (10 h L:14 h D photoperiod; quantum irradiance meter LiCor, Lincoln, Nebraska, USA). Microcolonies were either exposed to present day ambient pH_T 8.1 (pH in total scale; 8.08 ± 0.03; pCO₂ = 396 µatm) or pH_T 7.5 (7.46 ± 0.04; pCO₂ = 2, 057 µatm; Table 2), which corresponds to a predicted pCO₂ level for 2,300 with Ω_arag close to 1 (Caldeira & Wickett, 2005) and was chosen to provoke some degree of coral dissolution. The microcolonies were maintained under these conditions for over 16 months (thus, relative to most other studies acclimation time was very long, cf. Table 1; Table S1). pH was tightly regulated and monitored over the acclimation period (see below). Corals were fed each week with freshly hatched Artemia sp. nauplii and a mixture of crushed fish.

Control of seawater pCO₂

Seawater was pumped from a depth of 30 m into 1,000 L tanks, where the pH was manipulated to reach a fixed value. The incoming seawater presented a very stable chemistry during the whole experiment period: salinity of 40 ± 0.2‰, pH_T of 8.09 ± 0.02, total alkalinity of 2, 505 ± 0 µeq kg⁻¹, as reported by the Israel National Monitoring Program (NMP) of the Gulf of Eilat (NMP, 2016). A pH electrode (S-200C, Sensorex, CA, USA) located in each water table was connected to a pH controller (Aquastar, IKS ComputerSysteme GmbH, Karlsbad, Germany), monitoring the pH (software Timo, Matuta, Germany) and adjusting the bubbling of CO₂ (from a CO₂ cylinder) accordingly. Well-mixed water from each tank flowed continuously into the corresponding section (150 L) of the water table and pH was monitored daily. Seawater samples from the water tables were taken five times over the acclimation period to monitor TA (TA was stable with 2,494 ± 11 and 2,496 ± 9 µeq kg⁻¹ for pH_T 7.5 and 8.1, respectively; Table 2).

Incubation procedure for measurements of calcification

After 16 months of pH acclimation in water tables, microcolonies (2.5–3.0 cm in length, N = 6 per treatment and time point) were incubated in sealed (no gas exchange; no headspace) incubation vessels (40 mL Septa vials with additional paraffin wrap, clear borosilicate glass; 2.5 ×10 cm; Thomas Scientific, Inc, Swedesboro, NJ, USA) containing ∼40 mL filtered seawater (FSW; 0.45 µm) at the two pH levels for 2 h, 4 h, and 6 h. Temperature and light intensity during the incubation were similar to those provided during the preceding 16 months of acclimation (see above), while an orbital shaker provided constant water motion. Applicability of the shaker in contrast to conventionally used stir bars was tested in a preliminary experiment (Fig. S1). Prior to the incubation, the colony surface and attached nylon strings were carefully cleaned of epiphytes and algae. Microcolonies in the incubation vials were hanging without any wall contact by using a T-shaped string (Fig. S2). The TA and ⁴⁵Ca experiments were carried out over two consecutive days, using the same microcolonies and time of day to avoid effects of diurnal variations in physiology (Edmunds & Spencer-Davies, 1988). After TA measurements, the respective microcolonies were returned to the water table overnight and used for ⁴⁵Ca measurements the following day. This allowed minimizing physiological differences due to handling or cumulative stress between both measurements. The chosen repeated measures design clearly represents a practical compromise. Although the dual measurement of the ⁴⁵Ca-uptake into the skeleton and the TA change in the surrounding seawater would have been preferable, the subsequent alkalinity measurement of radioactive seawater with a radioactive contamination of the titrator and the associated equipment was not justifiable for reasons of lab safety (note however that such an approach was taken by Tambutté et al., 1995). While day-to-day variations in calcification cannot be ruled out with certainty, we found no significant statistical effect of this in the complete dataset (i.e., no within subjects effect in separate 2 h, 4 h, 6 h incubations; Table 3).

Total alkalinity anomaly technique

Net calcification rates of corals were determined using the TA anomaly technique (Chisholm & Gattuso, 1991). Water samples were collected for TA at the beginning and end of the incubation from each treatment to determine carbonate chemistry and stored until analysis in sealed, bubble-free brown glass vials (borosilicate glass) in the dark at 4 °C. Calcification rates were calculated from the difference between TA measured at the beginning and the end of each incubation period (ΔTA; corrected for blank values from filtered seawater only incubations with N = 3), according to the equation by Schneider & Erez (2006). Rates were normalized to skeleton dry weight: ${\displaystyle \text{Calcification}\left(\mathrm{\mu }\mathrm{mol CaCO}3{\mathrm{h}}^{-1}{\mathrm{g}}^{-1}\text{dry skeleton}\right)=\frac{\frac{\Delta \text{TA}}{2}\times \left({\mathrm{V}}_{\mathrm{vessel}}-{\mathrm{V}}_{\mathrm{coral}}\right)\times \text{1,000}\times 1.028}{\mathrm{W}\times \mathrm{T}}}$ ΔTA (in meq kg⁻¹); V _vessel is the volume of the experimental vessel (in mL); V _coral is the displacement volume of the coral (in mL); 1.028 is the density of seawater in the northern Gulf of Eilat (in g mL⁻¹); W is skeleton dry weight (in g) and T is the incubation duration (in hours).

⁴⁵Ca technique

Measurements of GC were based on the improved ⁴⁵Ca protocol (Tambutté et al., 1995). Microcolonies were placed in filtered seawater (FSW; 0.45 µm) with a total activity of 360 kBq (as ⁴⁵CaCl₂, 1958.18 MBq mL⁻¹, Perkin-Elmer Life and Analytical Sciences). Dead microcolonies, killed with 2% formaldehyde, were included in the experiment as a control for isotopic exchange (Al-Horani et al., 2005) and three incubation vessels were left without microcolonies to serve as a seawater-only control. To determine the specific activity, 100 µL aliquots were taken at the beginning and end of each incubation period. Following the labelling period, specimens were immersed in 600 mL FSW for 1 min, and then rinsed (5 ×1 min) with 10 mL of ice-cold glycine-high calcium medium (50 mM CaCl₂, 950 mM Glycine, pH 8.2). Labelled specimens were then incubated for 30 min in vessels containing 20 mL of FSW. Water motion was provided by a shaker. Following efflux incubation, microcolony tissue was removed using 2 M NaOH for 20 min at 90 °C. Following tissue hydrolysis, the skeleton was first rinsed with 1 mL NaOH (Houlbrèque, Tambutté & Ferrier-Pagès, 2003), then thoroughly rinsed with FSW, followed by double distilled water. The solution from the first rinse was added to the tissue hydrolysate, while the remaining rinsing solutions were discarded. Finally, skeletons were dried at 70 °C, weighed, and subsequently dissolved in 12 M HCl. Samples (500 µL) of skeleton digest or tissue hydrolysate were added to 10 mL Ultima Gold AB (PerkinElmer) scintillation liquid and measured on a scintillation counter (Tri-carb 1600TR, Packard). Sample count rates (CPM, total counts per minute) were obtained several times over two days and the average difference between readings was 0.8 ± 0.54%. Total calcification was then calculated from the activity recorded in the skeleton digest and seawater control samples and normalized as µmol CaCO₃ per skeleton dry weight using the formula: ${\displaystyle \text{Calcification}\left(\mathrm{\mu }\text{mol CaCO}3{\mathrm{h}}^{-1}{\mathrm{g}}^{-1}\text{dry skeleton}\right)=\frac{\left({\mathrm{Activity}}_{\mathrm{sample}}\times \frac{1.17}{{\mathrm{Activity}}_{\mathrm{seawater}}}\right)}{\mathrm{W}\times \mathrm{T}}}$ where Activity_sample is the total counts per minute (CPM) in the dissolved skeleton sample; Activity_seawater is the total CPM in 100 µL seawater sample (control); 1.17 is the amount of Ca²⁺ in 100 µL ambient seawater (in µmol); W is skeleton dry weight (in g) and T is the incubation duration (in hours). The amount of ⁴⁵Ca uptake by dead specimens (covered with tissue) was subtracted from the amount measured in intact (live) specimens.

Calculation of seawater carbonate system

Value of TA of accurately weighed seawater samples (12.3 g) were measured to the second end point (Almgren, Dyrssen & Fonselius, 1983) using an automatic potentiometric titration (Mettler-Toledo GmbH, DL67 titrator; resolution of burette volume: 1/5,000). The TA was determined in triplicate and computed using the Gran equation (DOE, 1994) with pH values lower than 3.9 for creating the Gran plot. Accuracy of TA analysis was checked against certified seawater reference material prepared by Andrew Dickson (Scripps Institution of Oceanography). The differences between triplicate samples were less than 6 µeq kg⁻¹. The pH electrodes (Mettler-Toledo DG-111–SC; Stockholm, Sweden) were calibrated daily before using the titrator (the manufacturer’s technical specifications are ±0.03 pH for both precision and accuracy). The acid concentration was 0.049 N HCl (JT Baker, Phillipsburg, NJ). In the series of experiments that compared gross and net calcification, a new titrator was utilized: a Metrohm 862 compact Titrosampler that uses at least 35 g seawater samples (autosampler combined with titrator; potentiometric measuring accuracy of ±0.003 with resolution of 0.001; resolution of burette volume: 1/10,000). Hence, experimental samples containing only 40 mL, were diluted by a factor of three (to have enough water for triplicate measurements of TA) and the acid concentration was set to 0.025 M. TA was calculated using the first derivative of the curve for the evaluation of the exact end point. Prior to measurements, water samples were filtered (0.22 µm EMD Millipore Millex sterile syringe filters). Water samples for analysis were stored in darkness at 4 °C in brown glass bottles, filled up to the top with a gas tight screw, and processed within two weeks of collection.

pH measurements were carried out using a CyberScan pH meter (pH/Ion 510 Eutech Instruments with automatic temperature compensation) and CyberScan gel-filled pH combination electrode. Prior to experiments, the pH electrode was calibrated against National Bureau of Standards (NBS) scale buffers of 4.01, 7.00 and 10.00 (Mettler Toledo) at 25 °C and was soaked in seawater for at least 1 h before measurement. The manufacturer’s technical specifications of the pH meter were 0.01 pH for resolution and ±0.01 (standard error) for accuracy.

Components of the carbonate system (pCO₂, CO${}_3^{2-}$, HCO${}_3^{-}$, DIC concentrations and Ω_arag) were calculated from TA along with pH values, temperature and salinity using the CO2SYS program, version 1.03 (Lewis, Wallace & Allison, 1998; Pierrot, Lewis & Wallace, 2006). The pH_NBS were shifted onto the total pH scale (pH_T) by subtracting −0.11 (Zeebe & Wolf-Gladrow, 2001), which includes a minor correction for [SO${}_4^{2-}$] and the stability constant of HSO${}_4^{-}$ at a salinity of 40.7‰. The thermodynamic carbonate dissociation constants for activity scales (K1 = 5.845 ± 0.008 and K2 = 8.945 ± 0.013) were attained from Mehrbach et al. (1973) and the refit from Dickson & Millero (1987). The input parameters for pressure (10 dbar), total P (0.03 µmol kg⁻¹) and total Si (0.68 µmol kg⁻¹) were obtained from the NMP of the Gulf of Eilat (NMP, 2016).

Data normalization

For accurate determination of TA changes in the incubation volume, coral volume was determined by measuring its displacement weight in seawater following Archimedes’ principle and assuming an approximate density of 1.028 g cm⁻³ for Red Sea seawater. Normalized calcification rates were obtained by determining skeleton dry weight, using a vibra balance (Shinko Denshi Co., Ltd., Japan; accuracy ∼1 mg).

Statistical analysis

Due to the paired nature of the measurements, the overall dataset was analysed with a repeated measures ANOVA. This way, the within subjects effects allow to test for consistent differences in calcification rates between both estimation methods (TA vs. ⁴⁵Ca) as a function of incubation time (2 h, 4 h, 6 h), seawater pH_T (8.1 vs. 7.5) and the combination of both, whereas the between subject effects test whether overall calcification rates differ between pH and incubation time. All calcification rates were cube root transformed to achieve normality as tested by the Shapiro–Wilk test. Differences between estimation methods due to seawater pH were also tested for the 2 h time point, using the overall fitted model (repeated measures ANOVA as above) as well as a paired t-test for each pH.

The relationship between absolute estimates of calcification (in nmol) between TA and ⁴⁵Ca was evaluated using geometric regression fitting and by directly comparing ⁴⁵Ca/TA ratios over a range of calcification values. Geometric regression was preferred over linear regression since no causality can be established between both variables and both have an inherent measurement error. Note that geometric regression equations still depend on the assignment of x and y. To test whether the relationship between both methods is affected by seawater pH, ⁴⁵Ca/TA-ratios from both treatments after 2 h were tested with a t-test. All statistical analyses were performed in JMP 11.2.1 (SAS Institute, Cary, NC, USA).

For comparison with literature data from Smith & Kinsey (1978) and Tambutté et al. (1995), the software WebPlotDigitizer (Rohatgi, 2015) was employed to extract the data points from the original publication graphs, since the raw data is no longer available (S Smith & E Tambutté, pers. comm., 2015). The extracted data is provided as supplementary information for future reference (Table S2).

Effect of incubation time and estimation method on derived calcification rate

During the 16 months of acclimation at both pH treatments all coral microcolonies survived and remained visibly healthy (extended polyps; no bleaching; substantial increase in size). Considering data from all incubation periods, calcification rate estimates from both techniques were not significantly different, nor were there consistent differences between both techniques related to pH or incubation time (Fig. 2 and Table 3). Overall, calcification rates were significantly higher at reduced pH with no significant effect of incubation time (Fig. 2 and Table 3). Since only a single biological replicate was tested, this ecological finding is not of significance and will not be further discussed here.

Considering only the calcification rates obtained from 2 h of incubation as physiologically useful data (see later discussion) shows that absolute estimates of calcification differed between both methods depending on pH (Fig. 2 and Table 3). Calcification estimates from the TA anomaly technique were on average 5% and 21% higher than the ⁴⁵Ca estimate, for ambient and reduced pH, respectively. However, this difference was only statistically significant for the reduced pH treatment (paired t-test, t = 4.486, p = 0.0065^∗, N = 6). The implications of this finding, including previously published experimental data, will be discussed in the following paragraph.

Can dissolution simply be derived from the difference between net and gross calcification?

Net calcification rates can be altered as result of changes in dissolution and/or gross calcification (Fig. 1). Distinguishing true changes in gross calcification (as result of biological impairment) from changes in gross dissolution rates is therefore of importance to assess the real impact of pH stress. Paradoxically, the few studies that measure GC and NC that would allow an estimate for dissolution found higher rates for NC in comparison to GC (Rodolfo-Metalpa et al., 2015; Tambutté et al., 1995), with the exception of Smith & Kinsey (1978). Some techniques might not be suitable at all for such calculations, as indicated by a 7- to 11-fold overestimation of calculated GC, when using rates of NC and dissolution from the buoyant weight technique and comparing it with ⁴⁵Ca GC rates (Rodolfo-Metalpa et al., 2015). In this case, the raw data even suggested that buoyant weight-based dissolution rates in the cold-water coral Desmophyllum dianthus at pH 7.7 were larger than actual ⁴⁵Ca GC rates, despite measurable NC (as calculated from provided raw data; R Rodolfo-Metalpa, pers. comm., 2015). Correlating blank-corrected calcification values from the TA and ⁴⁵Ca technique from the 2 h dataset at both pH treatments confirms the strong significant positive correlation between both techniques (Fig. 3A), consistent with previous findings (Smith & Kinsey, 1978; Tambutté et al., 1995). Nevertheless, for fully covered coral colonies (this study; Tambutté et al., 1995), it still suggests that TA tends to overestimate calcification values relative to the ⁴⁵Ca technique because ⁴⁵Ca/TA estimate ratios tend to be below 1 (Fig. 3B).

Previous work by Smith and Roth (in Smith & Kinsey, 1978) incubated tips of Acropora formosa for 30 min in test tubes containing ⁴⁵Ca-labelled seawater, whereas Tambutté et al. (1995) improved this ⁴⁵Ca protocol and labelled microcolonies of S. pistillata in seawater for up to 3 h. Both studies confirmed a strong correlation (r = 0.95–0.99) between the TA and ⁴⁵Ca estimates, but found opposing results with regard to absolute calcification estimates between these two methods. While the slope of the geometric regression was similar between both studies and similar to the obtained values here (Fig. 3A), the regression and its extrapolated intercepts relative to the 1:1 line suggested opposite findings: (1) the presence of ⁴⁵Ca in the skeleton even without measurable changes in TA (Smith & Kinsey, 1978; Figs. 3A and 3B) and (2) changes in TA without detectable ⁴⁵Ca incorporation in the skeleton (Tambutté et al., 1995; Figs. 3A and 3B). Physical adsorption of ⁴⁵Ca to the exposed skeleton surface at the fracture site, rather than biologically mediated incorporation, was suggested as the source for the overestimation by the ⁴⁵Ca method by Smith & Kinsey (1978). The use of completely tissue-covered microcolonies was intended to minimize this effect in the study of Tambutté et al. (1995) and the resulting negative intercept was interpreted as loss of radioactivity during washing steps or general time lag due to the ⁴⁵Ca loading of extracellular and tissue compartments. Since both studies chose an overall line-fitting method to compare TA and ⁴⁵Ca estimates instead of comparing the values directly, it was not apparent that the absolute level of over- or underestimation of either method was not constant, but largely dependent on the absolute amount of deposited CaCO₃ (Fig. 3B). This is expected given that the relative overestimation of ⁴⁵Ca adsorption at exposed skeletal parts, or underestimation due to ⁴⁵Ca loss, should represent a small, constant value that is related to surface and porosity of the skeleton and is not proportional to the amount of deposited CaCO₃. Thus, the influence of such artefacts on the ⁴⁵Ca/TA relationship becomes less important for larger calcification values (as can be seen in Fig. 3B). We confirmed that the ⁴⁵Ca isotopic equilibration of dead, covered skeletons (used as blanks) was rather constant after 2 h and 6 h, with 0.14 ± 0.03 and 0.19 ± 0.02 µmol CaCO₃ g⁻¹ dry skeleton (mean ± SD; N = 3), respectively. Figure 3B illustrates that for a sufficiently high CaCO₃ deposition (>1,500 nmol from TA estimate), the ⁴⁵Ca/TA ratio approaches a rather constant value in each study with 1.19 ± 0.02 (mean ± SE, N = 3, Smith & Kinsey, 1978), 0.77 ± 0.01 (N = 10; Tambutté et al., 1995), 0.94 ± 0.06 (N = 5; this study pH_T 8.1), and 0.85 ± 0.03 (N = 5; this study pH_T 7.5). Using only these values and considering only the studies with fully-covered microcolonies, TA estimates are consistently larger than ⁴⁵Ca estimates. Furthermore, this ratio is not significantly different between the pH treatments in our study (One-Way ANOVA, F_1,8 = 1.930, p = 0.2022).

When interpreting the relationship between the ⁴⁵Ca and TA anomaly technique one must consider two important aspects: (1) the correlational strength, which demonstrates the linear relationship between both methods over a range of values, and (2) the numerical ratio itself, which represents the relative difference in the calcification estimate between both methods and is assumed to provide an estimate for the occurrence of dissolution. For the first aspect, we conclude that both methods correlate well in both pH treatments. For the second point however, we have to conclude that the apparent tendency for GC/NC ratios to be <1 in the relevant studies that used fully covered microcolonies contradicts the generally assumed relationship of NC ≤ GC. This is furthermore problematic since TA estimates in our experiment are factually a measure of GC, because we demonstrated in a separate experiment, that dissolution (if occurring) was not detectable. We tested bare skeletons of similar size in the same vials over a slightly longer period of time (3 h) and even lower pH (pH 8.1, 7.6, and 7.3 with N = 9 each) and found that 25 of the obtained 27 ΔTA values were less than 18 µeq kg⁻¹ and therefore below the recommended detection limit of at least three times the precision of the used titrator (Table S4). Thus, for the case of no detectable dissolution, both methods would represent measures of GC here, but still provide different estimates for gross calcification, especially at reduced pH. Even if dissolution occurred under reduced pH conditions, the ⁴⁵Ca/TA ratio shifted in the wrong direction, erroneously indicating that NC increases relative to GC. This violates the assumed relationship of NC = GC–GD and the similar results of this study and the study of Tambutté et al. (1995) raises the general question of whether a comparison of these two sensitive methods is really appropriate to provide accurate values for GD. Limited by sensitivity of common methods, assessing gross dissolution and its relation to net calcification simultaneously is clearly a challenge.

Changes in seawater chemistry

Substantial changes in the carbonate chemistry (TA, DIC, pH, pCO₂, HCO${}_3^{-}$, CO${}_3^{2-}$ and Ω_arag) were recorded inside the incubation vessels during the time course of the experiment at both pH treatments (Table 4, Table S3). A general decline in DIC was correlated with an increase in seawater pH and a shift within the carbon equilibrium. For an incubation period beyond two hours, a substantial pH shift (+0.4–0.9 units) with CO_2(aq) depletion of more than 70%, as well as considerable changes (≥10%) in seawater DIC were observed irrespective of pH (Table 4).

At constant temperature and salinity, alterations in carbonate system parameters such as pH and TA are driven mainly by biological activities of the coral holobiont such as photosynthesis, respiration, and calcification (Schulz & Riebesell, 2013). Photosynthesis and calcification decrease the seawater DIC, while respiration and dissolution increase the DIC. However, only calcification and dissolution alter the TA (Chisholm & Gattuso, 1991; Zeebe & Wolf-Gladrow, 2001). The photosynthetic activity of the dinoflagellate symbiont is the main driver for the observed changes in total DIC and DIC speciation (with subsequent pH shift) as CO₂ is removed and fixed in biomass at a much higher rate than HCO${}_3^{-}$/CO${}_3^{2-}$ incorporation into the skeleton. The stronger changes in carbonate chemistry at reduced pH are likely to be the result of two interacting factors: (1) Potentially altered photosynthetic/respiratory rates at reduced pH (Jury et al., 2013; Page et al., 2016), and (2) a generally lower buffering capacity of CO₂-enriched seawater, specifically at pH_T 7.5, where seawater reaches its minimum buffering capacity (Delille et al., 2005; Egleston, Sabine & Morel, 2010; Riebesell et al., 2007; Suzuki, 1998). Near pH_T 7.5, where DIC∼TA, any changes in DIC and TA will cause a similar magnitude change in [CO₂], [H⁺], and Ω and a sharp change in pH (Egleston, Sabine & Morel, 2010).

Best practice guidelines recommended changes in TA should be at least 3- to 10-fold (Langdon, Gattuso & Andersson, 2010) the analytical precision of the instrument (e.g., ΔTA ∼6 − 20 µmol kg⁻¹ seawater), but these changes along with changes in DIC should not exceed 3% (Schulz et al., 2009) or 10% (Langdon, Gattuso & Andersson, 2010) of the absolute values. Given the observed stronger changes in seawater chemistry at lower pH over time in our experiment, one can see that the trade-off between using a small incubation volume to accurately measure changes in TA versus using a large volume to minimize changes in carbonate chemistry becomes a crucial issue for measurements under OA conditions, especially in closed systems. For calcification measurements with symbiotic corals at reduced pH in closed vessels, incubation time must be minimized as long as the seawater carbonate parameters are not kept constant. In addition, increasing oxygen saturation in closed vessels due to photosynthesis is also an important factor to consider when measuring zooxanthellate corals in the light. In a similar experiment, very high pO₂ levels (200–250%) were recorded after 4 h and 6 h of incubation, at both pH treatments and overall O₂-productivity significantly declined with prolonged incubation (Fig. S3A, S3B) likely due to a negative feedback of high oxygen levels on photosynthesis and general physiology. Hyperoxic conditions promote the generation and accumulation of reactive oxygen species (Gerschman et al., 1954) with potentially negative effects that interfere with the normal physiological performance and the subsequent estimate of calcification rate.

The problem of fundamental changes in seawater chemistry and their negative feedback for coral physiology can essentially be captured in the relationship between coral biomass, chamber volume, and incubation time. The presented experimental data clearly confirm that these factors define the applicability and limits of closed chamber incubations as highlighted previously, but become especially important in the context of measurements under OA conditions (Chisholm & Gattuso, 1991; Langdon & Atkinson, 2005; Langdon, Gattuso & Andersson, 2010; Schulz et al., 2009). Note that coral size rather than coral surface area is used in the following, since surface area as a size proxy was not always available from the below discussed studies. For comparison, in the present experiment, the ratio between incubation volume (mL) and coral size (cm) was 13 − 16 (the corresponding V_vessel∕V_coral-ratios for this branching species were 20 to 44) and incubations beyond 2 h already caused large shifts in carbonate parameters and pO₂. This should be a general point of consideration, since previous studies that used the same coral species, similar conditions of temperature and light intensity (∼25 °C; ∼150–170 µmol photons m⁻² s⁻¹), and used similar (10–17; Furla et al., 2000) or even smaller (6–8; Tambutté et al., 1995) incubation volume/coral size-ratios, incubated corals for up to 3 h (incubation volumes: 7–10 mL), thus most likely experiencing similar shifts in the carbonate species and/or O₂ concentration. The other extreme, employing very large incubation volumes relative to coral size, bears the risk of approaching the detection limit of the TA titration method, even with longer incubation times. For example, ΔTA over 2 h approached less than three times the analytical precision for most of the treatments for a volume/size ratio of 225 in the study of Takahashi & Kurihara (2013). It should be favourably noted that the studies that experienced this issue and used alkalinity changes as low as 3 µmol kg⁻¹ to calculate calcification rates belong to the few publications that explicitly report the measured changes in the carbonic acid system in the incubation vessels (Hossain & Ohde, 2006; Ohde & Hossain, 2004; Takahashi & Kurihara, 2013). Our experimental data reaffirm the necessity of reporting changes in the carbonic-acid system parameters not only for pre-experimental pH acclimation, but also for actual incubation experiments that yield calcification data, following previously suggested best practice guidelines (Andersson & Mackenzie, 2012; Langdon, Gattuso & Andersson, 2010). The need for a valid description of the whole carbonate system is further emphasised by the study of Hoppe et al. (2012) on the uncertainties in the calculated carbonate chemistry when using only two measured parameters.

In order to avoid substantial changes in the seawater chemistry in incubation vessels, it might be useful to estimate the maximum permissible coral size based on known estimates of the species-specific calcification rate, the known incubation volume and by applying the 3–10% rule for the permitted absolute TA change. For our case, given a Red Sea seawater TA of ∼2,500 µeq/kg, TA must not change by more than ∼75 µmol L⁻¹ (3%-rule) over the total incubation time, corresponding to ∼3,000 nmol in a 40 mL chamber (note that the real incubation volume is actually V_chamber–V_coral). For an expected calcification rate of ∼250 nmol CaCO₃ cm⁻² h⁻¹ for a Northern Red Sea Stylophora pistillata (e.g., Krueger et al., 2017), the maximal permitted coral surface area of the fragment used for a 1 h incubation must be smaller than 12 cm² (3,000 nmol/250 nmol) or 6 cm² for a 2 h incubation. Considering the varying performance in calcification, photosynthesis, and respiration between coral species and environmental conditions (e.g., pH, light, temperature), preliminary experiments might also benefit in determining minimal vessel volumes, depending on coral size and respective incubation time in closed systems.