Global coral reef ecosystems exhibit declining calcification and increasing primary productivity

Coral reefs are under increasing threat from global climate change, including ocean acidification and rising thermal stress that trigger bleaching and reduce resilience. Understanding ecosystem functions such as net ecosystem calcification (Gnet: balance of calcification and dissolution) and net organic productivity (Pnet: balance of photosynthesis and respiration) is central to predicting reef persistence. Historically, reefs were near net-zero or slightly positive in Pnet, but recent observations show changes linked to stress events and environmental change. Increasing Pnet relative to Gnet can indicate shifts from coral- to algal-dominated states. Higher Gnet typically occurs on reefs with higher coral cover and fewer stressors, whereas declining or net-dissolving calcification reflects stress or low live coral cover. Two main approaches estimate ecosystem calcification: community census methods and hydrochemical methods; the latter integrate processes at ecosystem scale and are the basis for this meta-analysis. Prior case studies have identified potential drivers (e.g., depth, temperature, coral cover), but global consensus is lacking and local or mesocosm relationships may not scale. This study compiles in situ, hydrochemical investigations to identify global drivers of Gnet and assess temporal trends, hypothesizing interactive controls by biogeochemical parameters, climate change, and reef state, and anticipating declines in Gnet with increasing Pnet over time due to reductions in coral cover and reef condition.

Previous work has used both community census and hydrochemical approaches to estimate coral reef calcification. Census methods can apportion contributions by taxa but often rely on literature-based growth rates that may not match in situ conditions, leading to large uncertainties. Hydrochemical methods, based on changes in seawater carbonate chemistry, integrate ecosystem-scale calcification and productivity in space and time and avoid species-level biases. Manipulative mesocosm experiments have demonstrated roles for light, aragonite saturation state (Ωar), and community composition on metabolic rates, yet may not capture the complexity of natural reefs. Case studies suggest controls by calcifier cover, hydrodynamics (wave action, depth), temperature, light, nutrients, and Pnet, but relationships can be site-specific and nonlinear. Skeletal records and regional assessments indicate widespread declines in coral calcification over recent decades, with warming often implicated more strongly than acidification in low-latitude regions. Phase shifts from coral to algal dominance have been linked to reduced resilience and ecosystem services, with increasing Pnet sometimes signaling post-disturbance algal establishment. Knowledge gaps include underrepresentation of equatorial reefs, inconsistent reporting of auxiliary variables (light, uncertainty, nutrients), and limited nighttime measurements despite nighttime dissolution being common and sensitive to acidification.

Systematic literature review and meta-analysis focused on peer-reviewed estimates of coral reef net ecosystem calcification (Gnet) derived via in situ hydrochemical methods that included both day and night observations. Searches (Google Scholar) combined terms such as coral reef, metabolism, carbonate budgets, calcification, ecosystem, and productivity, including reference chaining. Inclusion required diel-integrated Gnet or sufficient data to calculate it; studies with major external carbon sources (e.g., riverine or groundwater inputs) or invalidated assumptions were excluded. Data collection was finalized in April 2020. The final dataset comprised 116 diel-integrated calcification rates from 53 publications across 36 reef sites in 11 countries. Qualitative variables compiled: study year, location, sampling duration (days), wave exposure (exposed, moderate, protected), seasonal heat type (summer–autumn vs winter–spring), methodology (slack water, flowing water/Eulerian/Lagrangian, chambers with multiple benthos, offshore TA anomaly, benthic gradient flux), nighttime production status (net calcifying vs dissolving), and reef state (degraded vs healthy/unspecified). Quantitative variables compiled: diel-averaged aragonite saturation state (Ωar), temperature, calcifier cover (corals + coralline algae), depth, NO3− and PO43−, Pnet and Gnet, and reported uncertainties. Pnet estimates from both oxygen- and DIC-based methods were included to maximize temporal coverage. Hydrochemical approaches quantify ecosystem-scale calcification and do not resolve taxa-specific contributions. Statistical analyses used linear mixed effects regression (LMER) in R (lme4), with restricted maximum likelihood, incorporating fixed effects for candidate predictors (Pnet, latitude, duration, season heat type, methodology, reef state, Ωar, temperature, log-transformed calcifier cover, depth, and categorical wave action) and a random intercept for location to account for site-level variability. Due to frequent missing values, a backward-selection process guided by AIC/BIC iteratively removed variables to increase sample size and improve model parsimony. Collinearity checks, residual diagnostics (homoscedasticity, linearity), Cook’s distance for influence, and Type II Wald chi-square tests (car package) assessed significance. Tukey contrasts with Bonferroni-Holm correction (multcomp) supported within-factor comparisons. A refined analysis subset examined seasonal temperature effects on Gnet by comparing the same sites across seasons (excluding degraded reefs). Long-term trends used repeated surveys at the same sites within the same seasonal bins across years. Underreporting of variance precluded inverse-variance weighting; nutrients and some auxiliary variables were often unavailable and omitted from models.

- Dataset scope: 53 publications, 116 diel-integrated Gnet estimates, 36 reef sites in 11 countries. Australian reefs contributed 35% (mostly Great Barrier Reef). 51% Northern vs 49% Southern Hemisphere; mid-latitudes (15–28°) dominated (72% of studies). Over 40% of studies occurred in the last decade. - Metabolic ranges: Gnet ranged from −90 to 667 mmol m−2 d−1 (mean 124.1 ± 106.6). Pnet ranged from −870 to 1240 mmol m−2 d−1 (mean 65.1 ± 254.3). Six of 116 studies reported diel net dissolution; 32% reported negative Pnet. - Nighttime processes: 67% of reefs were net dissolving at night. Reefs with nighttime dissolution had significantly lower diel Gnet (χ2 = 31.066, p < 0.001, n = 84), indicating daytime calcification often fails to offset nighttime dissolution. - Global drivers (LMER): Depth and calcifier cover significantly predicted Gnet; wave action and seasonality showed suggestive effects. • Depth: Significant negative effect (χ2 = 4.788, p = 0.029, n = 84). Each 1 m increase in depth decreased Gnet by 14.8 ± 6.8 mmol m−2 d−1 (holding other factors constant). • Calcifier cover: Significant positive effect (χ2 = 15.723, p < 0.001, n = 84). A 10% increase in calcifier cover increased Gnet by 4.1 ± 1.0 mmol m−2 d−1. Relationship is nonlinear, with strong sensitivity when cover <20% (Gnet ≈ 42.5·log(calcifier cover) + 120.2). • Wave action: Not significant at α = 0.05 (χ2 = 5.597, p = 0.061, n = 84) but retained, indicating potential influence. • Seasonality: Became significant after removing five outliers (χ2 = 6.737, p = 0.035, n = 79); summer–autumn > winter–spring Gnet on average. • At sites with multi-season data (n = 26), Gnet increased with higher temperature across seasons (χ2 = 22.232, p < 0.001), with an empirical relationship ΔGnet% = 22.49·ΔT − 41.605. • No significant global effect detected for methodology, study duration, latitude, reef state, Pnet, or Ωar in the final models. - Temporal trends (repeat sites): Across repeatedly surveyed reefs (n = 29 Gnet and n = 26 Pnet surveys), since the 1970s: • Pnet increased by 3.0 ± 0.8 mmol m−2 d−1 yr−1 (p < 0.001). • Gnet declined by 1.8 ± 0.7 mmol m−2 d−1 yr−1 (p < 0.001), equivalent to a mean decline of 4.3 ± 1.9% yr−1 across sites (three of four sites with ≥3 surveys declined at −5.5 ± 3.9% yr−1; one site increased +0.4% yr−1). • Extrapolation indicates global net-zero calcification around 2054 at the current linear rate of decline. - Reef condition: Degraded/recovering reefs had lower Gnet (64.2 ± 10.5) than healthy/unspecified reefs (137.5 ± 11.7 mmol m−2 d−1), though reef state was not retained in the LMER due to potential confounding by location.

The study addresses the global drivers and temporal evolution of coral reef ecosystem calcification using an integrated hydrochemical meta-analysis. The strong dependence of Gnet on calcifier cover confirms that losses in coral and other calcifying benthos directly reduce ecosystem calcification potential. The significant depth effect highlights hydrodynamic and light-driven constraints on calcification, potentially via light attenuation, altered boundary layer dynamics, and thermal stratification. While wave action and seasonality showed weaker but suggestive roles, temperature exerted a clear positive influence on Gnet within sites across seasons, consistent with increased calcification up to thermal optima. However, at broader scales the temperature signal is masked by heterogeneity, and warming beyond thresholds leads to bleaching and sharp calcification declines, underscoring nonlinearity and context dependence. Temporally, the consistent increase in Pnet alongside declining Gnet supports phase-shift dynamics, where algal dominance expands following disturbances, reducing reef resilience, biodiversity, and ecosystem services. The observed Gnet decline (4.3% yr−1 on average) outpaces reported declines in calcifier cover alone (≈1.8% yr−1), implying additional mechanisms, including sub-lethal stress that transiently suppresses calcification, shifting community composition, and enhanced nighttime dissolution. The projection of net-zero calcification around 2054 aligns with independent lines of evidence (sediment dissolution experiments and skeletal records) indicating accelerating declines in reef carbonate budgets. Contrary to expectations from mesocosms and some local studies, Ωar did not emerge as a primary global driver of Gnet in this synthesis, suggesting that local hydrodynamics, community structure, and thermal stress modulate or overshadow direct acidification effects at the ecosystem scale. The prevalence of nighttime dissolution and its association with lower diel Gnet emphasize the importance of including nocturnal processes in assessments. Collectively, these results refine global understanding of drivers of coral reef carbonate production and indicate that continued warming and disturbance regimes will likely intensify declines.

This meta-analysis identifies depth and benthic calcifier cover as the dominant global predictors of coral reef ecosystem calcification, with wave action and seasonality also influential. Over the past five decades, reefs have experienced a significant decline in Gnet (≈4.3 ± 1.9% yr−1; −1.8 ± 0.7 mmol m−2 d−1 yr−1) and a concurrent increase in Pnet (≈+3.0 ± 0.8 mmol m−2 d−1 yr−1), consistent with a shift toward algal dominance and reduced carbonate production. Extrapolating observed trends suggests reefs may reach net-zero calcification around 2054. The findings indicate that Ωar is not a reliable primary predictor of long-term global Gnet, whereas community structure and hydrodynamics are critical. Warmer seasonal temperatures can enhance Gnet locally, but increasing frequency and severity of heat stress and bleaching will likely negate any benefits and accelerate declines. Future research should prioritize: (1) comprehensive, standardized reporting of auxiliary variables (including uncertainties, light/PAR, nighttime rates, carbonate chemistry, and nutrients); (2) increased coverage of underrepresented equatorial and highly biodiverse regions (e.g., Coral Triangle); (3) long-term repeated ecosystem-scale observations across seasons; and (4) improved quantification of hydrodynamic metrics (e.g., residence time, wave energy) to refine predictive models.

- Data availability and reporting: Many studies lacked key auxiliary variables; fewer than 10% reported all core numerical parameters (Gnet, Pnet, depth, calcifier cover, temperature, Ωar). Underreporting of variance prevented inverse-variance weighting. Nutrients and light/PAR were rarely available in standardized units. - Sample representativeness: Equatorial reefs are underrepresented (<10% of studies despite comprising ~26% of reefs), potentially biasing global inferences. Some regions had multiple repeat studies while others lacked coverage. - Methodological constraints: Wave action was categorized (exposed/moderate/protected) rather than quantified as continuous energy metrics; residence time was omitted due to limited, inconsistent, or error-prone reporting. Hydrochemical methods integrate ecosystem-scale processes but cannot apportion taxa-specific contributions. - Model limitations: Backward selection increased sample size but may omit interacting variables; missing data reduced the initial model to 46 of 116 studies. Five seasonal outliers influenced seasonality results. Potential confounding between reef state and location precluded including reef state in the final LMER. - Temporal projections: The 2054 net-zero estimate assumes linear continuation of trends; actual trajectories may be nonlinear due to increasing frequency and intensity of mass bleaching and extreme events.