Partitioning uncertainty in the projection of coral collapse using Large Ensembles of multiple Earth System Models

Anthropogenic climate change exposes coral reefs to increasingly frequent, intense and persistent events of elevated sea surface temperatures. These events adversely affect coral health and can lead to coral reef collapse. The timing of such collapse is uncertain, due to (1) biological uncertainty related to the adaptation and acclimation capacity of corals themselves and (2) environmental uncertainty related to the uncertainty in the Earth’s warming rate (climate sensitivity), natural variability and future anthropogenic emissions. Earth System Models (ESMs) provide a powerful tool for predicting how environmental conditions might evolve over the coming century and thereby contribute to the collapse of coral reefs. In this project we use Large Ensemble experiments from 4 ESMs to assess the model, scenario and initial-condition or natural variability dependence of predicting coral reef collapse. We find that the collapse of all tropical coral reefs occurs this century for all 4 ESMs if a business-as-usual (RCP8.5) emissions pathway is followed and if corals undergo no adaptation or acclimation. For nearly all coral reef locations, however, natural variability is a larger source of uncertainty than model differences — highlighting an irreducible uncertainty in predicting the timing of coral collapse. We find moderate mitigation (RCP4.5) could save up to 50% of the coral reef locations. We also explore the role of changing internal variability in the acceleration of coral collapse.