摘要:To better understand the decline of one of earth’s most biodiverse habitats, coral reefs, many survey programs employ regular photographs of the benthos. An emerging challenge is the time required to annotate the large volume of digital imagery generated by these surveys. Here, we leverage existing machine-learning tools (CoralNet) and develop new fit-to-purpose programs to process and score benthic photoquadrats using five years of data from the Smithsonian MarineGEO Network’s biodiversity monitoring program at Carrie Bow Cay, Belize. Our analysis shows that scleractinian coral cover on forereef sites (at depths of 3–10 m) along our surveyed transects increased significantly from 6 to 13% during this period. More modest changes in macroalgae, turf algae, and sponge cover were also observed. Community-wide analysis confirmed a significant shift in benthic structure, and follow-up in situ surveys of coral demographics in 2019 revealed that the emerging coral communities are dominated by fast-recruiting and growing coral species belonging to the genera
Agaricia and
Porites. While the positive trajectory reported here is promising, Belizean reefs face persistent challenges related to overfishing and climate change. Open-source computational toolkits offer promise for increasing the efficiency of reef monitoring, and therefore our ability to assess the future of coral reefs in the face of rapid environmental change.