Coral reefs are among the most biodiverse and socially important ecosystems on Earth. The Coral Triangle (CT) alone is home to 75% of all known coral species and provides protein supplies to one billion people, yet even optimistic climate scenarios predict catastrophic consequences for coral reef ecosystems by 2100. Antropogenic stressors, including temperature rising and overfishing, threaten coral biodiversity, thus understanding how reef connectivity, biodiversity and resilience are shaped by climate variability would improve chances to establish sustainable management practices.
Designing effective marine protected areas over large spatio-temporal scales is crucial to preserve coral ecosystems, but still challenging due to the complex interactions between ecoregions connectivity and climate variability, usually impossible to be addressed with the currently available computational resources. Here we combine machine learning algorithms with physical intuition applied to sea surface temperature anomalies over a 24-year period to analyze ecoregions and connectivity response to climate variability, and to assess bleaching recovery potential of central Indo-Pacific coral reefs. Our outcomes additionally quantify the impacts of the El Niño Southern Oscillation on reefs biodiversity and resilience, finding that resilience is higher for north-equatorial reefs, and that the outstanding biodiversity of the Coral Triangle is dynamic in time and space, and benefits from ENSO in its negative and neutral phases. La Niña conditions amplify the large-scale exchange of genetic material between the CT and the Indian Ocean, while such exchange is enhanced between the CT and the central Pacific in neutral years. By means of machine-learning approaches and physical-grouded data analysis we contextualize the CT biodiversity maintenance, species richness dynamics and it time evolution, opening new routes to effectively monitoring its future.