Data fusion for mapping coral reef geomorphic zones: Possibilities and limitations

Mapping the geomorphology of coral reefs provides key information to scientists and managers about the distribution, extent and

structure of reef landforms. Geomorphological zones within a reef system are underpinned by geological and environmental

gradients in physical and biological processes, usually resulting in well-defined and clearly recognisable landforms. Mapping of such

zones has been traditionally undertaken by visual interpretation of remotely sensed imagery, with mapping performance constrained

by the operator‟s contextual interpretation and/or imagery characteristics. However, mapping criteria are subjective and often not

transferable to other sites. This study explores a semi-automatic, GEO-Object-Based Image Analysis (GEOBIA) approach to

mapping intra-reef geomorphological zones based on fusing high-resolution satellite imagery and seamless elevation data. The

method is applied to Quickbird and Worldview2 imagery of two coral reefs in Australia: Bet Reef, an intertidal lagoonal reef

platform in the central Torres Strait; and Lizard Island, a fringing reef in northern Great Barrier Reef (GBR). Combining optical and

bathymetric information considerably improved classification results from ~ 80% to ~ 90% overall accuracy. Rule sets developed

based on this data fusion approach have the potential to be transferable between different reef types across geographic settings.