Shifts in coral reef biogeochemistry and resulting acidification linked to offshore productivity

Significance

Ocean acidification is hypothesized to have a negative impact on coral reef ecosystems, but to understand future potential impacts it is necessary to understand the natural variability and controls of coral reef biogeochemistry. Here we present a 5-y study from the Bermuda coral reef platform that demonstrates how rapid interannual acidification events on the local reef scale are driven by shifts in reef biogeochemical processes toward increasing net calcification and net respiration. These biogeochemical shifts are possibly linked to offshore productivity that ultimately may be controlled by large-scale climatological and oceanographic processes.

Abstract

Oceanic uptake of anthropogenic carbon dioxide (CO₂) has acidified open-ocean surface waters by 0.1 pH units since preindustrial times. Despite unequivocal evidence of ocean acidification (OA) via open-ocean measurements for the past several decades, it has yet to be documented in near-shore and coral reef environments. A lack of long-term measurements from these environments restricts our understanding of the natural variability and controls of seawater CO₂-carbonate chemistry and biogeochemistry, which is essential to make accurate predictions on the effects of future OA on coral reefs. Here, in a 5-y study of the Bermuda coral reef, we show evidence that variations in reef biogeochemical processes drive interannual changes in seawater pH and Ω_aragonite that are partly controlled by offshore processes. Rapid acidification events driven by shifts toward increasing net calcification and net heterotrophy were observed during the summers of 2010 and 2011, with the frequency and extent of such events corresponding to increased offshore productivity. These events also coincided with a negative winter North Atlantic Oscillation (NAO) index, which historically has been associated with extensive offshore mixing and greater primary productivity at the Bermuda Atlantic Time-series Study (BATS) site. Our results reveal that coral reefs undergo natural interannual events of rapid acidification due to shifts in reef biogeochemical processes that may be linked to offshore productivity and ultimately controlled by larger-scale climatic and oceanographic processes.

Ocean acidification (OA) resulting from rising atmospheric CO₂ (1–3) and the associated declines in surface seawater pH and saturation state with respect to CaCO₃ minerals such as aragonite (Ω_aragonite = [Ca²⁺][CO₃^2-]/K_sp*, where K_sp* is the ion solubility product) have raised concerns on the potential consequences to marine calcifiers and ecosystems (4, 5). Reductions in Ω_aragonite have been found to negatively affect organismal CaCO₃ production (6) while accelerating bioerosion and CaCO₃ dissolution (7, 8). Hence, it has been hypothesized that coral reefs could shift from a condition of net calcification to net dissolution, with some model estimates predicting a transition for worldwide reefs at atmospheric CO₂ levels of 560 ppm (5, 7).

Despite growing concern about the vulnerability of coral reefs, a lack of long-term measurements has prevented direct observation of anthropogenic OA owing to increasing atmospheric CO₂ in these environments. Additionally, short-term observations have revealed large variability and modification of reef seawater CO₂-carbonate chemistry on diurnal and seasonal timescales as a result of coral reef biogeochemical processes such as photosynthesis, respiration, calcification, and CaCO₃ dissolution (9, 10). It has been hypothesized that these natural processes, quantified as net ecosystem production (NEP = gross primary production − autotrophic and heterotrophic respiration) and net ecosystem calcification (NEC = gross calcification − gross CaCO₃ dissolution), could modulate local seawater chemistry such that the rate of acidification on coral reefs is significantly different from the open ocean (11). Consequently, the anthropogenic OA signal could be either alleviated or exacerbated by reef biogeochemical processes (10–12). Centuries-long records of reef pH derived from ¹¹B of coral cores reveal large variability in pH over decadal timescales (13), possibly indicative of the dynamic nature of reef biogeochemical processes. However, these records of pH alone (which are nonetheless inferred rather than directly measured) cannot explicitly reveal the drivers behind the observed variations in seawater pH. Long-term measurements of reef (and offshore) biogeochemistry are therefore necessary to understand the natural variation and controls (whether they be biological, physical oceanographic, or climatic) on reef CO₂-carbonate chemistry, and how this will change under future OA and climate change scenarios.

We have measured and characterized the seawater CO₂-carbonate chemistry across the Bermuda coral reef platform monthly between June 2007 and May 2012 to investigate temporal and spatial variability in seawater pCO₂, pH, and Ω_aragonite. These parameters were calculated based on surface seawater measurements of temperature, salinity, total dissolved inorganic carbon (DIC =[CO₂]+[HCO₃⁻]+[CO₃^2-]), and total alkalinity [TA = excess of bases over acids relative to a reference state (14)] at four discrete locations on a transect traversing the coral reef platform (Fig. S1). Given the large variability in reef CO₂ system parameters on seasonal timescales, a 5-y record is too short to detect the secular trend of anthropogenic OA. However, contemporaneous monthly measurements from the nearby Bermuda Atlantic Time-series Study (BATS) station, located ∼80 km southeast of Bermuda in the open ocean of the North Atlantic subtropical gyre (Fig. S1), allow us to decipher variations in the offshore source water chemistry from changes occurring on the coral reef owing to local biogeochemical processes.

Fig. S1.

Bathymetry of Bermuda coral reef platform with time series study sites. Reef time series sites are denoted by white squares (Tynes Bay, TB; Dock Yard, DY; Mid Platform, MP; North Channel, NC). At each station, water samples were collected monthly for measurement of DIC, TA, temperature and salinity from June 2007 to May 2012. The white arrow indicates the general direction of the BATS study site, located ∼80 km offshore (31°50′ N, 64°10′W). Credit: Mandy Shailer, Department of Conservation Services, Government of Bermuda.

Relative changes in DIC and TA reflect the biogeochemical partitioning of carbon between the inorganic and organic carbon cycles on the reef (15, 16) and the balance of NEP, NEC, and air–sea CO₂ gas exchange. NEC changes DIC and TA in a ratio of 1:2, with net calcification (NEC > 0) acting to draw down DIC and TA, causing a decrease in seawater pH and Ω_aragonite. NEP mainly alters DIC, with net organic carbon production (NEP > 0) reducing DIC concentrations as CO₂ is used and increasing pH and Ω_aragonite. Air–sea CO₂ gas exchange affects DIC only, but typically exerts minor influence on DIC relative to NEC and NEP in reef environments with residence times of a few days, such as Bermuda (17, 18). Consequently, comparison of salinity normalized changes in seawater DIC and TA (nDIC and nTA) between the reef and offshore can be used to evaluate the relative contribution and variability in reef NEC and NEP over time and space. Deviations between reef and BATS nTA concentrations reveal the relative extent of reef NEC, whereas deviations in nDIC corrected for the influences of NEC and air–sea CO₂ gas exchange reveal the relative extent of reef NEP. In the present study, we used these relationships to evaluate the influence of biogeochemical processes on reef seawater carbonate chemistry and the relative attribution of NEC and NEP to changes in reef pH and Ω_aragonite (see Materials and Methods for further details).

Results and Discussion

Long-Term OA Signal Offshore.

Observations since 1983 at BATS and Hydrostation S (another long-term time series near Bermuda) reveal a rise in surface seawater nDIC by ∼1.20 ± 0.09 μmol kg⁻¹⋅y⁻¹ (2), driven by the oceanic uptake of anthropogenic CO₂ (Fig. 1). Consequently, surface seawater offshore Bermuda has become less alkaline, with temperature- and salinity-normalized pH and Ω_aragonite dropping by −0.05 and −0.25 units, respectively, since 1983 (2). (Note that temperature- and salinity-normalized data are referred to throughout this discussion.) Increasing nDIC has driven most of these changes whereas nTA has remained relatively constant. Superimposed on the steady rise in nDIC over time, a seasonal oscillation driving large-amplitude swings in pH and Ω_aragonite is apparent. This reveals the influence of physical, biological, and climatic processes on surface ocean carbon dynamics, with the predominant driver being seasonal phytoplankton blooms as observed in the average annual cycle (i.e., climatology, Fig. 2).

Fig. 1.

Time series of seawater nTA, nDIC, pH, and Ω_aragonite observed at BATS and Hydrostation S from 1983 to 2014. (A) Long-term observations of nTA (blue) and nDIC (orange) reveal relatively stable nTA over time, but increasing nDIC due to uptake of anthropogenic CO₂. Change over time in pH (B) and Ω_aragonite (C) as driven by time-dependent changes in nTA (blue area) and nDIC (orange area). The thick black line represents total change in both pH and Ω_aragonite, with changes relative to initial observations.

Fig. 2.

Reef and BATS seawater carbonate chemistry observations, 2007–2012. (A–D) Time series of seawater nTA, nDIC, pH, and Ω_aragonite observed at BATS (blue) and across the reef platform (orange) from 2007 to 2012. Reef data are shown as individual markers per site with average of four sites shown as a line with ±1SD shaded; (E–H) Climatology of same seawater CO₂ parameters with ±1SD shaded. Rapid acidification events are evident in the short-term observations on the reef (A–D), particularly in the summer of 2010. The climatology reveals summertime drawdown in reef nTA and nDIC, whereas reef pH and Ω_aragonite remain relatively constant year-round. pH and Ω_aragonite have been temperature- and salinity-normalized to values of 23.1 °C and 36.6 g kg⁻¹, respectively, for both BATS and reef data.

Near-Shore Biogeochemical Processes Dictate Reef Acidification.

Over the 5-y study period between 2007 and 2012, no apparent trends in seawater CO₂-carbonate chemistry parameters resulting from rising atmospheric CO₂ are observed on the coral reef platform (Fig. 2). However, large depletions in nTA and nDIC, relative to BATS, are observed each year with the extent of the drawdown variable on the interannual timescale. Offshore variability of seawater carbonate chemistry at BATS is representative of conditions in the subtropical gyre of the North Atlantic Ocean surrounding Bermuda and provides a direct comparison with onshore variability on the Bermuda reef. Despite these depletions in reef nTA and nDIC, reef pH and Ω_aragonite remain comparatively constant while offshore a predictable seasonal pattern of acidification persists throughout the 5-y study (Fig. 2).

Occasionally, rapid acidification events are observed on the reef platform, particularly in August 2010, when the largest difference in pH between the reef and BATS (ΔpH_REEF−BATS = −0.14, Fig. 3) was observed. The climatology of the reef pH and Ω_aragonite shows that the range of variability is greater and average values are slightly lower during the summer compared with the rest of the year, coinciding with the drawdown and maximum variability in nTA and nDIC (Fig. 2 E–H).

Fig. 3.

Influence of NAO and offshore primary productivity on reef carbonate chemistry dynamics. Time series of (A) NAO [monthly in gray and winter (December–March), mean in black]; (B) wind speed (green) and mixed-layer depth (MLD) (black); (C) primary production (PP); nTA (D) and nDIC (E) at BATS (squares) and across the reef platform (circles); (F) NEC (blue), NEP (red), and air–sea CO₂ gas exchange (green); and (G) the contributions of NEC, NEP, and air–sea gas exchange to pH differences between BATS and the reef (ΔpH_{REEF– BATS}). Negative NAO winter events enhance storm activity in subtropical waters surrounding Bermuda, deepening the MLD, which brings colder, relatively nutrient-rich waters to the surface and results in increased PP. During 2010 and 2011, deep winter MLD coincided with intensified spring blooms and enhanced late summer/fall NEC, causing greater drawdown of nTA across the reef and consequently driving acidification. In contrast, a positive winter (December–March) NAO state and consequently weaker PP signal in 2008 resulted in a subdued reef NEC signal the following summer.

The observed drawdown in reef nTA is mainly caused by reef NEC, with modifications in nDIC driven both by reef NEC and NEP (the influences of air–sea gas exchange of CO₂ are minimal, Fig. 3). Consequently, seasonal changes in NEC and NEP are ultimately driving the observed pH and Ω_aragonite variability on the reef, with increasing NEC causing acidification and increasing NEP having the opposite effect. Summertime increases in NEC produce a reoccurring reduction in seawater pH, with the strength of NEC corresponding to the extent of pH reduction (Fig. 3). In contrast, NEP fluctuates with no discernible seasonal cycle. When increases in NEC correspond with decreases in NEP, as they did during the summer of 2010, we observe the most extensive reduction in reef seawater pH and Ω_aragonite (Fig. 3).

Reef Biogeochemistry Responds to Offshore Forcing.

Our results reveal that seasonal and interannual changes in NEC and NEP are responsible for the observed changes in temperature- and salinity-normalized pH and Ω_aragonite, but what is the driver of these changes? Pronounced acidification events, such as during the summer of 2010, occurred during periods of anomalously high NEC and low NEP, indicating shifts in reef biogeochemical processes toward increased calcification and heterotrophy. Generally, calcification is thought of as an autotrophy-enhanced process with photosynthetic drawdown of seawater CO₂ concentrations elevating Ω_aragonite and providing conditions favorable for organismal CaCO₃ deposition (19). In contrast, respiration releases CO₂, reduces Ω_aragonite, and therefore is thought to decrease calcification. Thus, our finding of enhanced calcification during periods of increased heterotrophy runs counter to some modeling and field-based studies (10, 19). However, numerous experiments have shown that if corals, the dominant reef calcifiers, are well fed (such as by zooplankton) or have nutritionally replete diets, they will have greater rates of both tissue and skeletal growth (20–24). Some results have even demonstrated that corals may be able to counteract reduced rates of calcification resulting from OA by increasing feeding rates (23). Furthermore, increased heterotrophy and elevated tissue growth in corals have been found to cause overall respiration rates to increase (21). Therefore, we speculate that external pulses of nutrition to the reef could have enabled both the anomalously high summertime calcification and shifts to increasing heterotrophy responsible for the observed acidification events.

Measurements at BATS indicate that phytoplankton blooms were enhanced during our study period, particularly in the winter and spring of 2010 and to a lesser degree 2011 (Fig. 3). Whereas the Sargasso Sea surrounding Bermuda is typically oligotrophic, enhanced spring phytoplankton blooms have previously been documented and are hypothesized to be linked to the North Atlantic Oscillation (NAO) (25, 26), a climate mode describing differences in pressure between the atmospheric subtropical high-pressure system near the Azores and the subpolar low-pressure system near Iceland (27). A negative winter NAO state infers a southern shift in the Gulf Stream and winter storm track, causing deeper mixed layers in the Sargasso Sea due to enhanced winds (25, 27). With deeper mixed layers come lower sea surface temperatures and greater entrainment of nutrients, resulting in productivity blooms and increased mesozooplankton abundance (25, 26, 28–30).

Since measurements began at BATS, the winter NAO state has been primarily positive, with sporadic neutral to slightly negative events causing documented increases in offshore productivity (25, 26). However, the winter 2010 event (and to a lesser degree 2011) was anomalously negative (Fig. 3), and coincident with an abnormally large spring bloom. At the onset of the 2010 event, a strongly negative shift in reef NEP (i.e., net respiration) coincided with a deep mixed-layer depth (MLD) and consequent upwelling of nutrients [as indicated by anomalously low sea-surface temperature (SST) and high nDIC waters at BATS, Figs. S2–S4] (25). Reef nTA values that were elevated above those at BATS indicate that the shift in NEP coincided with platform-wide net CaCO₃ dissolution (Fig. 3). Whereas increased offshore primary productivity occurred in the winter to spring season of 2010 and 2011, the following late summer and fall, relative measurements of reef NEC, as inferred by alkalinity anomalies, reached their highest values over the 5-y measurement period, signifying intensified community calcification (max. NEC 2007: 39.8 μmol kg⁻¹; 2008: 30.9 μmol kg⁻¹; 2009: 42.2 μmol kg⁻¹; 2010: 50.5 μmol kg⁻¹; 2011: 50.0 μmol kg⁻¹, Fig. 3), and coincided with negative NEP values. Noticeably, the following winters in 2011 and 2012 also had NEC rates higher compared with previous years (Fig. 3). Cross-correlation analysis revealed the potential time dependence between offshore and reef processes, with maximum correlation between offshore primary productivity and MLD observed when productivity lagged MLD by 0 to 1 mo (r > 0.8, Fig. 4). In contrast, reef biogeochemical processes lagged offshore productivity by several months. Reef NEC showed the strongest correlation with offshore productivity at a lag of 4 mo (r = 0.77), whereas negative NEP (i.e., heterotrophy) was most strongly correlated at a lag of 6 mo (r = −0.45, Fig. 4). These correlations were statistically significant at the 99.5% confidence level.

Fig. S2.

Times series of seawater carbonate chemistry parameters measured at Hydrostation S and BATS 1983–2012. (A) Surface seawater temperature (°C); (B) surface seawater salinity (g kg⁻¹); (C) TA (μmol kg⁻¹; in situ in blue; salinity-corrected, i.e., nTA, in orange); (D) DIC (μmol kg⁻¹; in situ in blue; salinity-corrected, i.e., nDIC, in orange); (E) seawater pCO₂ (μatm; in situ in light blue; salinity- and temperature-corrected in red); (F) seawater pH (in situ in light blue; salinity- and temperature-corrected in red); (G) Ω_arag (in situ in light blue; salinity- and temperature-corrected in red).

Fig. S4.

Time series of seawater carbonate chemistry parameters measured at the four cross-platform reef sites (TB, DY, MP, and NC) 2007–2012. (A) Surface seawater temperature (°C); (B) surface seawater salinity (g kg⁻¹); (C) TA (μmol kg⁻¹; in situ in blue; salinity-corrected, i.e., nTA, in orange); (D) DIC (μmol kg⁻¹; in situ in blue; salinity-corrected, i.e., nDIC, in orange); (E) seawater pCO₂ (μatm; in situ in light blue; salinity- and temperature-corrected in red); (F) seawater pH (in situ in light blue; salinity- and temperature-corrected in red); (G) Ω_arag (in situ in light blue; salinity- and temperature-corrected in red).

Fig. 4.

Time-dependent correlations between offshore and reef processes. MLD at BATS shows maximum positive correlation with offshore primary productivity (PP) when it precedes PP by 1 mo (r = 0.83). In contrast, reef NEC and negative NEP (i.e., heterotrophy) show maximum correlation with PP when they lag PP by 4 (r = 0.77) and 6 mo (r = −0.45), respectively. The correlations were significant at the 99.5% confidence level, indicated by the black dashed lines.

Fig. S3.

Time series of seawater carbonate chemistry parameters measured at Hydrostation S and BATS 2007–2012. (A) Surface seawater temperature (°C); (B) surface seawater salinity (g kg⁻¹); (C) TA (μmol kg⁻¹; in situ in blue; salinity-corrected, i.e., nTA, in orange); (D) DIC (μmol kg⁻¹; in situ in blue; salinity-corrected, i.e., nDIC, in orange); (E) seawater pCO₂ (μatm; in situ in light blue; salinity- and temperature-corrected in red); (F) seawater pH (in situ in light blue; salinity- and temperature-corrected in red); (G) Ω_arag (in situ in light blue; salinity- and temperature-corrected in red).

Given the experimental evidence linking increased heterotrophy to higher calcification rates in corals (20–24) and the statistical significance of our cross-correlation analysis, we hypothesize that lateral advection of offshore blooms as well as nutrient upwelling, both of which were exacerbated during the winters of 2010 and 2011, possibly due to the NAO state, provided external pulses of nutrition to the reef. These pulses of nutrition enabled short-term shifts in reef NEC and NEP toward increasing calcification and heterotrophy, respectively, and it was these biogeochemical shifts that ultimately caused the observed changes in seawater pH and Ω_aragonite (Fig. 5). Additional evidence in support of this hypothesis is provided by coral cores from Bermuda that exhibit thicker layers of CaCO₃ deposition, and therefore greater rates of calcification, corresponding to years of negative SST anomalies (indicative of deeper MLD) (31, 32), negative winter NAO events (33), and presumably higher offshore productivity (25, 26).

Fig. 5.

Conceptual model of interannual variations in winter NAO state and resulting shifts in reef biogeochemistry (modified from ref. 23). (A) During periods of positive or neutral winter NAO states, spring blooms are weakened. With less advection of external biomass onto the coral reef, the reef community tends toward autotrophy and weaker calcification, resulting in enrichment of seawater Ω_aragonite when moving offshore to inshore across the reef platform (contour plot C). (B) When the winter NAO state is negative, spring blooms are enhanced with greater primary productivity and zooplankton biomass. Increased advection of offshore biomass shifts reef community metabolism toward greater heterotrophy and calcification, resulting in reduction of seawater Ω_aragonite and increased acidification. (C) Ω_aragonite as a function of nTA and nDIC with data from a positive (2008, light gray circles) and negative (2010, dark gray squares) NAO winter index year. Type II linear regression lines are shown for the positive (n = 33; m = 0.869 ± 0.124 SD; b = 582.111 ± 254.920 SD; r = 0.773) and negative years (n = 48; m = 1.029 ± 0.112 SD; b = 240.322 ± 230.700 SD; r = 0.798). Note the steeper slope (corresponding to greater acidification) during the negative NAO index year driven by a higher frequency of acidification events (shift to +δNEC and –δNEP, red arrows). Average BATS nDIC and nTA ±1 SD is shown by the black circle.

However, without direct measurements of phytoplankton and zooplankton quantities on the reef, nor biogeochemical measurements extending directly off the platform, our hypothesis linking reef biogeochemistry to offshore productivity relies on the assumption that BATS is representative of waters advecting onto the reef and that reef water residence time remained relatively unchanged (17). Additionally, discrepancies within our time series indicate that year-to-year subtleties in climatic and oceanographic processes may have cascading effects, with slight changes in offshore upwelling and phytoplankton blooms having subsequently large effects on reef biogeochemical processes and carbonate chemistry. For instance, unlike the 2010 event, the 2011 bloom in offshore primary production and summertime increase in NEC corresponded with a much smaller shift to heterotrophy; although summer acidification once again occurred, it was not as pronounced an event as 2010 (Fig. 3). It remains unclear why such differences in reef biogeochemistry were observed during years of relatively similar offshore productivity blooms, although it is possible they relate to variations in the timing, magnitude, and taxonomic composition (34) of the bloom (26), which are related to the strength of stratification, mixing, and the nutrient reservoir within the North Atlantic subtropical mode water (35). For instance, the 2010 spring bloom was noted for a shift from cyanobacteria in the preceding years to pico/nanoeukaryotic algae (34), having potentially cascading effects on higher trophic levels such as zooplankton (30), and ultimately affecting reef biogeochemical processes.

Similar to our observations from the Bermuda coral reef platform, coral coring studies from other locations have revealed that reef pH varies greatly on annual to decadal timescales, although the oceanographic and climatic mechanisms driving such variations may be site-specific (13). Using ¹¹B as a proxy for seawater pH, coral cores from Flinder’s Reef in the Great Barrier Reef revealed large interdecadal variations in reef pH of up to 0.3 units over the past few centuries, which ref. 13 postulated were driven by changes in reef residence time and ultimately the Interdecadal Pacific Oscillation. However, those authors did not address whether the observed changes in pH could result from changes in reef biogeochemical processes. In the present study, contemporaneous measurements of residence time on the Bermuda coral reef platform (17) lead to the conclusion that changes in residence time or other physical oceanographic parameters on the reef platform were likely small or nonexistent, and could not explain the observed variations in seawater carbonate chemistry. In contrast with the study by ref. 13, direct measurement of seawater carbonate chemistry rather than inference through paleoproxies allows for quantification of reef biogeochemical processes, and their contribution to acidification. Whereas coral cores provide us with a longer perspective on OA, as well as natural subdecadal cycles of acidification, a record of pH alone does not provide a mechanistic explanation of why or how such fluctuations occur.

As anthropogenic OA continues unabated, and the scientific community tries to elucidate the impacts on coral reefs, it is becoming increasingly apparent that this problem must be viewed in the context of natural climatic and oceanographic drivers as well as the ability of reefs to partly modify the surrounding seawater chemistry (11). Similar to the offshore environment where anthropogenic CO₂ trends are only discernible after decades of observations (2, 3), confounding local biological processes easily mask the OA signal on coral reefs. Adding complexity to the system is the proposed connection to offshore processes and climatic phenomena such as the NAO, which exerts demonstrable control on biogeochemical processes (26, 30) and seawater CO₂ dynamics (36), but has yet to be extensively examined as a driver of coral reef, or more generally, coastal biogeochemistry (37). Continued and expanded observations will allow the scientific community to address these links in near-shore environments, as well as discern the effect of anthropogenic CO₂.

Materials and Methods

Sampling DIC and TA.

Surface seawater samples from the Bermuda coral reef platform were collected once a month at 0.5–1-m depth using a 5-L Niskin bottle. TA and DIC samples were collected according to standard protocols (14) using 250-mL Kimax brand glass sample bottles. Samples were immediately poisoned with 100 μL saturated solution of HgCl₂. Temperature was measured in the Niskin bottle while samples for salinity were collected in glass bottles and later analyzed using an autosalinometer (Guildline Instruments). DIC was analyzed coulometrically using a UIC CM5011 CO₂ coulometer combined with a versatile instrument for the determination of total inorganic carbon and titration alkalinity (VINDTA) 3C (Marianda Inc) or single-operator multiparameter metabolic analyzer (SOMMA) system, alternatively based on infrared absorption using an automated infrared inorganic carbon analyzer (AIRICA) (Marianda, Inc) and a Li-Cor 7000 as the detector. TA was analyzed based on potentiometric acid titrations (∼0.1 N HCl) using a VINDTA3S (Marianda Inc). Performance and precision of the instruments were regularly verified using certified reference material (CRM) prepared by A. Dickson at Scripps Institution of Oceanography (SIO). The accuracy and precision of replicate CRMs on any given day of analyses were typically in the range of ±2–4 μmol kg⁻¹ for both TA and DIC.

Since 1983, the surface seawater at the Hydrostation S (32°10′N, 66°30′W) and BATS (31°50′ N, 64°10′W) sites has been sampled approximately monthly for T, S, DIC, and TA, as well as other biogeochemical parameters. The sampling frequency of the time series has not been exactly uniform, ranging from 9 to 12 sampling cruises per year during the first few decades of collection to 14–15 sampling cruises per year more recently. Between 1983 and 1989, C. D. Keeling at SIO managed sample collection and analysis. Since 1989 samples were collected using 500-mL Pyrex bottles (replaced by 250-mL Kimax bottles in the early 2000s), which were quickly poisoned with HgCl₂ and sealed until analysis at Bermuda Institute of Ocean Sciences (BIOS). Analysis occurred typically within a few months of collection (2). Similar to the reef platform samples, DIC samples were analyzed using coulometric methods with an SOMMA system. TA was analyzed using a manual alkalinity titrator until the 2000s, when it was then replaced by a VINDTA 2S system. Precision of the instruments was verified on a daily basis using CRMs, with an accuracy and precision of typically <0.2% for replicate analyses (2).

For both reef and BATS measurements, seawater CO₂ chemical parameters were calculated using CO2SYS (cdiac.ornl.gov/ftp/co2sys/) with measured TA and DIC at in situ temperature and salinity conditions, and stoichiometric constants defined by ref. 38. The mean of the four reef sites at each time point was used to represent “reef” measurements. Temperature and salinity normalized seawater CO₂ chemical parameters for BATS and reef sites were calculated using nDIC, nTA, and the average salinity and temperature observed across the reef platform (36.63 g kg⁻¹ and 23.1 °C, respectively). To account for the nonuniform sampling intervals, which occurred primarily in the BATS dataset, samples occurring in the same month were averaged (2). Short-term data at BATS and on the reef flat were clipped to exactly 5 y to prevent any seasonal weighting. The climatology was calculated by averaging the seasonal cycle of a given parameter (Fig. S4), with the anomalies determined by subtracting the climatology from the time series of each parameter (2).

Computing Reef Biogeochemical Processes and Contribution to Acidification.

The contributions of nDIC and nTA to the temperature and salinity normalized pH and Ω_aragonite at BATS and on the reef platform were calculated by using the temporal changes in nDIC and nTA. The starting pH and Ω_aragonite were determined based on the initial nDIC and nTA, with the final values for nDIC and nTA at each time point sequentially stepped through and the pH and Ω_arag calculated for each step. To calculate reef NEC, NEP, and air–sea CO₂ gas exchange, offshore BATS and reef data were interpolated onto evenly sampled datasets, with BATS assumed to be representative of waters flowing onto the reef platform. Relative NEC (μmol kg⁻¹) was calculated based on depletions in alkalinity from BATS to the reef using the following equation assuming constant residence time and average depth:

$$\text{NEC}=-\frac{1}{2}\left(nT{A}_{\text{reef}}-nT{A}_{\text{BATS}}\right).$$

Air–sea gas exchange of CO₂, which impacts only the DIC concentration, was calculated using in situ reef pCO₂ values (calculated from in situ DIC, TA, T, and S data) and the equations described by refs. 39, 40. Wind and barometric pressure for air–sea flux calculations were measured at the Bermuda international airport, and assumed to be representative of conditions on the reef platform. Ten-min sustained wind speed at 10 m and barometric pressure data taken at 3-h intervals were obtained from the World Meteorological Organization and Bermuda Weather Service, and subsequently binned into months, averaged, and interpolated to match our measurement intervals for the BATS and reef biogeochemistry data. Accounting for the effects of air–sea gas exchange and NEC on DIC mass balances, relative NEP (μmol kg⁻¹) was subsequently calculated according to the following equation:

$$\text{NEP}=-{\text{NEC}-\text{CO}}_2\hspace{0.17em}\text{gas}\hspace{0.17em}\text{exchange}-\left({\text{nDIC}}_{\text{reef}}{-\text{nDIC}}_{\text{BATS}}\right).$$

The attributions of NEC, NEP, and air–sea CO₂ gas exchange of reef to BATS differences in pH (ΔpH_{REEF – BATS}, Fig. 5) were calculated as previously described for nDIC and nTA contributions, with the impacts of each biogeochemical process on nDIC and nTA calculated and the resulting nDIC and nTA used to then calculate pH.

Climatological and Offshore Primary Productivity Data.

Monthly mean NAO index data were obtained from the NOAA Climate Prediction Office (www.cpc.ncep.noaa.gov/products/precip/CWlink/pna/nao.shtml) with the winter NAO state taken as the average of December to March. Depth-integrated primary productivity data were calculated using a trapezoidal integration of primary productivity data from the upper 140 m at BATS. Primary production is calculated as the mean difference in ¹⁴C uptake between light and dark bottle incubations of seawater samples. The MLD was computed as the depth where temperature was less than 0.5 °C cooler than surface temperatures (29).

Cross-correlations between offshore primary productivity, MLD, and reef NEC and NEP processes were calculated using the MATLAB function xcorr, with correlation coefficients normalized such that autocorrelations at zero lag were exactly 1. Confidence intervals of 99.5% were calculated using the equation ± 2.58/√N, where n = 60.