Systematic review of the uncertainty of coral reef futures under climate change, datasets

Climate change impact syntheses, such as those by the Intergovernmental Panel on Climate Change (IPCC), consistently assert that limiting global warming to 1.5°C is unlikely to safeguard most of the world’s coral reefs. This prognosis primarily stems from 'excess heat’ threshold models, which assume that widespread coral bleaching predictably occurs when temperatures accumulate beyond a specific threshold. Our systematic review of research projecting coral reef futures to climate change (n=79) revealed that 'excess heat' models constituted only one third (32%) of all studies but attracted a high proportion (68%) of citations in the field. We observed that most methods employed deterministic cause-and-effect rules rather than probabilistic relationships, impeding the field's ability to estimate uncertainties of coral reef futures. In attempting to assess the consistency of projected impacts, we aimed to identify common coral reef metrics under the same emissions scenarios. However, disparate choices in metrics and emissions scenarios hindered a cohesive synthesis and limited the exploratory analysis to a small fraction of available studies. We found substantial discrepancies in expected impacts to coral reefs, suggesting that some 'excess heat' models may project more extreme impacts than other methods. Drawing on lessons from the field of climate change science, we propose that an IPCC ensemble-like approach to generating probabilistic projections for coral reef futures is feasible. Successful implementation will require improved coordination among modeling efforts to select common output metrics and emission scenarios, addressing existing geographical biases, among other gaps in current modeling efforts.

Methods
We conducted a comprehensive literature search using the Thomson Reuters Web of Science database to identify studies that projecting the impacts of climate change on shallow tropical and sub-tropical coral reefs. This search, adhering to PRISMA guidelines, yielded 2705 peer-reviewed articles, which we refined to 79 relevant articles published between 1999 and 2023 based on a specific selection criteria (Dataset 1). These studies were categorized into five major methodology types and further classified based on their approaches to simulating heat stress. Key characteristics such as the model output variables, spatial scale, and geographic area of each study were extracted, along with their methodological approaches, assumptions, and the techniques used.

Our study aimed to assess and compare the projected impacts and uncertainties of various model types using a meta-analysis approach. The database of 79 studies was considered for inclusion in the exploratory meta-analysis based on specific criteria (view published article and supplementary methods for detailed list and Supplementary Figure 1). Briefly, to enable a meaningful analysis, we identified the three most frequently used model outputs in our database. Among those, only studies that provided: 1) sufficient data for projection estimates and uncertainty measures to be reliably extracted or calculated, 2) reported end-of-century projections, and 3) used a baseline period between 2000 and 2015, were selected for the exploratory meta-analysis. In cases where projection and uncertainty estimates were presented in figures, values were extracted using PlotDigitizer, where possible.