Over the recent decades, advances in remote sensing have transformed our understanding of the global distribution of coral reefs. High-resolution optical and spectral satellite data have enabled major mapping initiatives, delivering unprecedented detail of shallow reef extent at global scales. However, these approaches remain largely constrained to shallow waters (typically <15 m), leaving the spatial extent of deeper reef habitats unresolved.
This limitation is significant, as mesophotic coral ecosystems (MCEs; ~30–150 m) are increasingly recognised for their ecological importance, while representing a substantial proportion of global reef area. Estimating the spatial extent of these ecosystems requires approaches that extend beyond optical-only based methods.
Here, we present a scalable geospatial modelling framework that integrates bathymetric derivatives and environmental covariates alongside optical predictors to estimate the likelihood of coral reef occurrence across full depth gradients, with initial application to the Great Barrier Reef and Coral Sea. In parallel, we introduce a complementary, open-access, multi-modal reference dataset that synthesises observations from multiple sources to support model training and validation.
Together, these approaches should help to improve our knowledge of the spatial extent of deep reef ecosystems.