henhaobio
Sign in

Nature CommunicationsEcologyPeer-reviewed journal article

A rapidly closing window for coral persistence under global warming

Yves‐Marie BozeciD, Arne A. S. AdamiD, Beatriz Arellano NavaiD, Anna K. CresswelliD, Vanessa Haller‐BulliD, Takuya IwanagaiD, Liam LachsiD, Samuel A. MatthewsiD, Jennifer K. McWhorteriD, Kenneth R. N. AnthonyiD, Scott A. CondieiD, Paul R. HalloraniD, Juan-Carlos OrtiziD, Cynthia RiginosiD, Peter J. MumbyiD

PDFDOIPreprintCodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Summary paragraphCoral reefs around the world are threatened by recurrent marine heatwaves causing mass coral bleaching and mortality1. Mitigating future warming impacts requires strategic management2 that adopts a long-term lens. Global analyses of projected heatwaves3,4 are critical for decision-making but how disturbance refugia, coral life-histories and adaptation interact with warming is unknown. Here, we simulate coral eco-evolutionary dynamics across >3,800 individual reefs of Australias iconic Great Barrier Reef under the current suite of climate projections5. We forecast a rapid coral decline before mid-century regardless of emissions scenario. Under the most likely warming6, corals may continue to decline throughout the century. However, corals may recover after mid-century provided that warming is sufficiently slow, allowing thermal adaptation to keep pace with temperature changes - as projected if warming does not exceed 2{degrees}C. Higher emission scenarios would drive most reefs to a near collapse. Future resilient reefs were mostly found in bleaching refugia, which also possessed a greater diversity of thermal phenotypes. Although cool-adapted corals disperse to thermal hotspots, there was no evidence of gene swamping undermining thermal adaptation. While opportunities exist to build management strategies that promote adaptation in thermal refugia and warm spots, curbing the rate of warming by 2050 is crucial for coral persistence.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (25 Jan 2025) · CC BY 4.0

Eco-evolutionary simulations of 3,806 Great Barrier Reef reefs project coral cover falling to ∼14% by 2040 under every emission scenario; recovery afterwards needs warming below 2°C.

Reef forecasts have mostly asked when sea temperatures will pass present-day mortality thresholds, leaving out coral evolution, life-history diversity, larval connectivity and refugia. Here an individual-based model of the Great Barrier Reef (GBR) gains heritable heat tolerance and is driven by downscaled CMIP6 (Coupled Model Intercomparison Project Phase 6) projections, so evolutionary and demographic rescue can be weighed against the pace of warming.

Key results

  1. SuggestedMean coral cover falls steeply by 2040 under every emission scenario. After that, cover recovers only if warming stays below 2°C. Intermediate warming brings continued decline, and high emissions bring near-total loss.

    Median of likely GBR-mean cover 14% in 2040 (interquartile range 8–24%) vs 39% in 2023, a 60% decline. Shared Socioeconomic Pathway SSP2-4.5 (∼2.7°C): 11% (5–29%) by mid-century, 4% (1–16%) by 2100, >75% of reefs under 5% cover. SSP1-1.9 and SSP1-2.6 recover after mid-century, only SSP1-1.9 regaining present levels; SSP3-7.0 and SSP5-8.5 near total loss by 2080-2100. · Future coral cover trajectories

  2. SuggestedSimulated thermal adaptation proceeds at a similar pace under all scenarios, so outcomes hinge on the warming rate. Near 2°C, fast-growing heat-sensitive corals keep recovering and juvenile densities stay near historical levels; under stronger warming all taxa decline alike.

    Peak adaptation rates of 1.1-1.4°C-week (Degree Heating Week, DHW, units) per decade occur in the first half of the century, then decelerate, likely constrained by the 8°C-week tolerance limit. Under SSP2-4.5 and SSP3-7.0 all taxa decline at comparable rates. Under SSP1-1.9 and SSP1-2.6, cover starts recovering before mid-century while temperatures are still rising. · Scope for natural adaptation; Discussion

  3. SuggestedBleaching frequency and intensity explain future reef state more than cyclones, water quality or crown-of-thorns starfish (CoTS) outbreaks. Thermal refugia and larval hubs more often hold high cover, though both advantages fade with higher emissions late in the century.

    Linear models of z-scored predictors over 532,840 to 761,200 reef trajectories per SSP, at 2050 and 2100. Refugia/warm spots: below 10th/above 90th percentile of annual DHW; larval barrens/hubs: same cut-offs on external larval supply. High-cover reefs far more common in refugia under SSP1-2.6, and under SSP2-4.5/SSP3-7.0 by mid-century; refugia mainly southern (20–22°S) and far-northern (10–12°S). · Spatial variability of eco-evolutionary…; Drivers of eco-evolutionary changes; Drivers of ecosystem changes

  4. SuggestedThe model shows no gene swamping: inflow of cool-adapted larvae from refugia did not curb local thermal adaptation. Because refuge reefs have larger populations, they can still export a relatively high absolute number of heat-tolerant larvae, though as a smaller proportion.

    Mean heat tolerance distributions of heat-sensitive taxa were comparable between larval barrens and hubs. By mid-century refuge reefs averaged 1-2°C-week lower mean tolerance than warm spots but showed greater diversity of tolerance values. Of GBR reefs, 13% are bleaching refugia that can supply larvae to 58% of the system. · Drivers of eco-evolutionary changes; Discussion

  5. ShownOn average, the 2008-2023 hindcast reproduces observed GBR-wide coral cover and heatwave losses with little bias. Regional errors are larger, and the 2020 heatwave is mispredicted.

    Mean annual error (SD) vs Australian Institute of Marine Science monitoring: –0.3% (± 4.7%) GBR-wide; –3.1% (± 8.6%), +1.7% (± 5.4%), –3.3% (± 5.6%) in northern, central, southern regions. 2016: 38% simulated decline (31-45%) at 8-10°C-week, matching changes on 117 monitored reefs; 2017 and 2022 also reproduced, 2020 not (monitoring showed no significant mortality). · Modelling coral eco-evolutionary dynamics; Figure legends

Methods, data and reproducibility

ReefMod-GBR is an individual-based model of six coral groups on 3,806 reefs linked by larval dispersal. The authors added a heritable colony-level heat tolerance trait (SD 2.5°C-week, limits ±8°C-week, heritability 0.3). After a 2008-2023 hindcast, the model ran to 2100 under downscaled CMIP6 heat stress from ten models and five SSPs, with runs weighted by equilibrium climate sensitivity.

  • Australian Institute of Marine Science (AIMS) Long-Term Monitoring Program and Marine Monitoring Program coral cover surveys, used to initialise 2008 cover and check hindcasts · corals (GBR reef communities) · n = N=117 (2016), 119 (2017), 138 (2020), 192 (2022) reefs for heatwave comparisons
  • NOAA Coral Reef Watch satellite maximum annual DHW at 5-km resolution, 2008-2023 · n = 3,806 reefs
  • CMIP6 daily sea surface temperature projections from ten atmosphere-ocean general circulation models under five SSPs (2014-2100), downscaled to 10 km · n = 47 climate realisations
  • Heat stress experiment on fragments from a shallow outer reef crest in Palau, used to set within-group variance of heat tolerance · Acropora digitifera
  • In situ bleaching mortality from the 2016 GBR mass bleaching, plus a depth profile of bleaching at Wood Reef and 11-203 Reef · corals
  • Biophysical larval dispersal simulations with the eReefs-GBR4 hydrodynamic model · corals; crown-of-thorns starfish (Acanthaster spp.) · n = six spawning seasons; 25 simulated coral spawning events
  • High-resolution (10 m) remote-sensing reef maps used to set each reef's colonisable coral habitat area · n = 3,806 reefs
  • Fine-scale coral community model sites on Moore Reef, Central GBR · corals · n = 171 reef habitat sites
Methods & tools
ReefMod-GBR (REEFMOD.7.0_GBR), CMIP6, semi-dynamical shelf-sea downscaling, NOAA Coral Reef Watch, eReefs-GBR4, 4MW wave model, MIROC-6, AIMS Long-Term Monitoring Program, AIMS Marine Monitoring Program, univariate breeder's equation, linear models, bootstrap resampling
Software & environment
Model version named as REEFMOD.7.0_GBR; no programming language, dependency versions or container are stated.
Reproducibility gaps
No hardware or runtime is reported for the simulation ensemble (20 runs per climate model and scenario). The Moore Reef fine-scale model and the model simplifications are described only in Supplementary Texts, which are not included in the text.
Limitations (6)
  • Variation in heat tolerance, which selection acts on, was parameterised from one sensitive taxon (Acropora). An imposed ±8°C-week tolerance limit notably shapes outcomes, and how adaptive scope differs among taxa is unknown. · stated by the authors
  • The model omits life-history traits that warming, or stronger thermal resistance, might alter. It also omits evolution of the algal symbionts that influence coral growth and heat tolerance. · stated by the authors
  • The downscaled projections hindcast more extreme DHWs than were observed for 2014-2023. This possible warm bias may overstate the projected decline in the first 15 years, though other model assumptions lean optimistic. · stated by the authors
  • Warming projections vary widely among the CMIP6 models, giving high uncertainty within and between scenarios. This uncertainty grows over time in ways the authors find hard to quantify. · stated by the authors
  • Larval connectivity (six water years, 2010-2016) and suspended sediment (2010-2018 simulations) are recycled from hindcast conditions through 2100. Changes in dispersal or water quality under warming are therefore not represented. · noted in this summary
  • The sensitivity analysis varied one parameter at a time by ±20% for a single climate model (MIROC-6) and one cyclone run. It does not test parameter interactions or robustness across climate models. · noted in this summary

Read this if you model how populations persist under climate change and want a worked example combining heritable thermal tolerance, larval connectivity and downscaled CMIP6 projections in a reef-wide individual-based model.

Citations
38
FWCI
20.03
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISMcManus '21Bonforti '25Matz '19Matz '17Lachs '23Vogt-Vincen '25
78% semantic matchEcology
L. C. McManus, D. L. Forrest, E. W. Tekwa+7 more·bioRxiv·Feb 4, 2021