Global predictions of coral reef dissolution in the Anthropocene

arising from K. Davis et al. Communications Earth & Environment https://doi.org/10.1038/s43247-021-00168-w (2021)

Coral reef frameworks are constructed by calcifying organisms and are highly sensitive to ocean acidification. Shifting baselines in seawater chemistry have already had measurable impacts on net ecosystem calcification (G_net) on coral reefs¹, and projections of ocean acidification portray a poor future for reefs in the Anthropocene². While experimental approaches have revealed much about this trajectory, we lack a clear understanding of: i) the drivers and predictors of net calcification at ecosystem scales, and ii) accurate predictions of when ecosystem calcification will reach net dissolution in the 21st century.

Through a meta-analysis approach, the recent study in Communications Earth & Environment by Davis et al.³ provides important insights into ecosystem-scale calcification on coral reefs. Based upon 53 publications spanning 36 coral reef sites around the world, the study provides a more nuanced understanding of the global drivers of G_net. Cover of reef calcifiers (predominantly corals) and depth are key predictors of global ecosystem calcification, with evidence of seasonality and wave action as additional factors influencing G_net³. The meta-analysis outlines important knowledge gaps and research needs and highlights the limited data available for assessing changes in ecosystem calcification at the same reefs through time.

Under future projections, ocean acidification is expected to shift coral reefs from a state of net calcification to net dissolution through reductions in pH and aragonite saturation states (Ω_a)^4,5. The exact timing of this is unclear, in part due to methodological differences, but estimates of when coral reefs will cross a tipping point to net dissolution vary substantially from 2031 to 2082⁶, 2070⁷, and 2060 to 2080⁴. Through the compilation of G_net from a subset of sites with repeated measurements (6 of the 36 available coral reefs; n = 29 of the available 116 surveys), Davis et al.³ extrapolate linear predictions of G_net decline (1975–2017) to conclude that average global net-zero calcification will occur around the year 2054, based on a decline in G_net of 4.3 ± 1.9% yr⁻¹.

Extrapolating estimates of G_net into the 21st century based upon the available historical data is complex. We identify four issues with this approach:

• 1. Declines in G_net are modelled as a linear function. The model assumes that declines in G_net are linear in nature, but relationships between Ω_ar and G_net are inherently non-linear from the level of organismal differences in carbonate mineralogy⁸ to ecosystem-level interactions with water flux, light, and temperature^9,10. Increases in ocean temperature, be it spatial (e.g., depth, latitudinal) and/or temporal (e.g., diel, seasonal, inter-annual), interact to accelerate G_net. The ability of modest levels of warming to enhance G_net may at least in part buffer the impacts of ocean acidification on calcification rates², thereby delaying the transition to net dissolution. However, marine heatwaves and persistent ocean warming will have serious consequences for G_net, as will amplifying storms, cyclones and crown-of-thorns starfish outbreaks. Yet, local increases in (or recovery of) coral cover can rapidly increase G_net over a matter of years¹¹, further highlighting i) the nonlinearity in the environmental drivers of G_net and ii) the inherent complexity in accounting for ecosystem-scale feedbacks in future predictions of G_net.

• 2. Extrapolation of limited local studies to global trends. By modelling the linear decline in repeatedly-surveyed sites over time, Davis et al.³ expect average global net-zero calcification around 2054. While this result is predicted at global scales, the underlying data represents just six Pacific reefs, of which three are from the Great Barrier Reef, Australia (Fig. 1). Given the global footprint of coral reefs (Fig. 1), predictions of global decline from this dataset are inherently limited in scale. When grouped by season (to account for the effect of temperature on G_net), the majority of sites are represented by just two data points (Table 1). Within locations, linear declines in G_net are mostly nonsignificant, except for Lizard Island, which is represented by the greatest sample size and is parallel to declines in coral cover (Fig. 1, Supplementary Code, see Code Availability Statement). The simple linear analysis is further complicated by environmental variability among seasons and wave exposure (Fig. 1). Predictions of declines in G_net based upon the limited available data mask local-scale variability, and the lack of repeated measures data within seasons hinders accurate extrapolation to global trends.

  Fig. 1: Global changes in coral reef ecosystem calcification (G_net) over time.

  

  Trends in the subset of repeated measures surveys (n = 29) used by Davis, et al.³ to predict global declines in G_net (mmol m⁻² d⁻¹). Data separated by location (a–f) and season. Studies are placed in the context of the global footprint of coral reefs where colours denote the three main biogeographic regions: Pacific, Indian, and Atlantic Oceans.

  Table 1 Breakdown of repeated measures surveys (n = 29) used by Davis, et al.³ to project global declines in G_net when accounting for the season and seasonal bins (H hot, C cold, A all, GBR Great Barrier Reef).

• 3. Interacting environmental drivers of G_net. While calcifier cover, temperature, season, depth and wave exposure can substantially alter measures of G_net (Fig. 2 in³), these drivers are considered as additive effects in Davis et al.³ and are not included in modelling future predictions of G_net (likely due to the limited number of studies with repeated measures). The linear model predicts global net dissolution in 2054, yet the confidence intervals surrounding this prediction are broad, and only 16% of the variance in the fixed effects can be explained by the temporal trajectory ‘year’ (Supplementary Code, see Code Availability Statement). Of the six sites used in the analysis, in only one location was G_net consistently quantified in the same season (Table 1). In Palau, measurements were quantified from two different habitats (reef flat and lagoon), each with a single summer and autumn survey (Table 1). Exposed reef flats and sheltered lagoons have distinct biogeochemistries and metabolisms¹², which confounds predictions of G_net. Data compiled within the same linear model without environmental variables treated as such (e.g., random effects) confound precise estimates of spatial and temporal change.

• 4. Historical and contemporary patterns of G_net. In assessing changes in biogeochemical processes on coral reefs, historical datasets represent a critical baseline with which to assess recent change. The inclusion of an early study from Kaneohe Bay, Hawaii, complicates the analysis. Kinsey¹³ noted his data were ‘sparse and somewhat erratic’, that values of G_net were not weighted across seasons, and that values were over-represented by sampling in September when near-maximum metabolic rates occur (see footnote Table 38). Thus, the value of G_net of 241 mmol m⁻² d⁻¹ in 1976 (Fig. 1a) from Kinsey¹³, which was used to benchmark a 40–43% decline in G_net between 1976 and 2009, is likely to be an overestimate. While historical datasets can be paired with contemporary studies at the scale of individual reefs, exact locations within reefs also vary. For example, historical measurements at Davies Reef in 1981 were conducted on the reef flat ~60 m from the windward margin¹⁴, whereas contemporary measurements were conducted on the protected (lagoonal) side of the reef flat¹⁵, some 300 m from the windward margin. Similarly, while the four studies (n = 7 surveys) from Lizard Island, Great Barrier Reef (1975–2019), were conducted on the same reef flat, they were situated > 1 km apart^16,17. Spatial and temporal differences in environmental parameters (tides, depth, waves) and benthic structure (coral cover, seasonal algal blooms) confound paired measurements of G_net between historical and contemporary studies, and further complicate accurate predictions.

While meta-analysis of global datasets can yield important insights into the global drivers of coral reef ecosystem calcification, we reason that they are inherently limited in their capacity to accurately predict future states. The key conclusion of Davis et al.³ is that we can expect to observe net-zero calcification in coral reefs around 2054, yet the 95% confidence interval from their model places tipping point this anywhere between 2030 and somewhere beyond 2150 (Supplementary Code, see Code Availability Statement). Rationalising that net dissolution will occur “around 2054” overlooks this high degree of variability and uncertainty. Given the points raised above, we agree with the conclusions of an earlier study by Davis et al.¹¹ in that ‘…the predictive capabilities of ecosystem metabolism studies may be constrained by using the available short-term datasets to represent long-term calcification trends’.

In light of the future predicted shifts in ocean chemistry², global predictions of coral reef dissolution are increasingly important. Given the environmental complexities and heterogeneity of coral reefs, the magnitude of declines in G_net will inherently be variable, non-linear and hard to predict. Regardless, the meta-analysis by Davis et al.³ provides important insight into the future of G_net on coral reefs, including that i) reef slopes are likely to be more impacted than reef flats due to increased G_net at depth and extreme diel fluctuations in shallow water, and ii) reefs that support high levels of calcifiers and sustained coral cover may buffer declines in G_net due to changes in aragonite saturation state.

From a media coverage and policy perspective, conclusions of net-zero calcification in coral reefs around 2054 are striking and high-impact¹⁸. We argue that highlighting the uncertainty surrounding such estimates is essential. In future projections, local buffers in ocean chemistry and climate refugia may enhance adaptive capacity to the negative impacts of ocean acidification in future decades^5,19,20, while recovery of coral communities can rapidly increase G_net¹¹. Identifying bright spots where coral reef calcification may be sustained long into the 21^st century can build optimism and action at the interface between science and policy²¹, and sustain motivation for adopting emissions pathways that will ensure a future for coral reefs in the Anthropocene.