Monitoring Estuarine Habitats and Threats at a Regional Scale Using Aerial Photography, Object-Based Image Analysis and Deep Learning

Abstract Estuarine macrophytes, including seagrass, mangrove, and saltmarsh species, are ecological foundation species that significantly influence biodiversity, ecosystem processes, and provide various cultural and economic benefits. These habitats are, however, facing numerous threats from human disturbances which operate over a range of spatial and temporal scales. Effective management of estuarine macrophytes requires an understanding of their spatial distributions, the disturbances that threaten them, and how these may be changing over time. Here we describe how high-resolution aerial imagery combined with object-based image analysis (OBIA) and deep learning is being used to understand ecological changes and inform management of threats in estuaries in New South Wales, Australia. Although the full extent of threats to macrophytes may be widespread, the individual disturbances can occur over small areas, for example shading from jetties and pontoons, scour from boat mooring chains and offroad vehicle damage to saltmarshes, meaning that high resolution mapping is required. User and producer accuracy of the maps was > 90% for almost all habitat types; the high accuracy for subtidal seagrasses was due in part to strict image capture requirements and extensive field validation. This mapping technique is well suited for using the same imagery for mapping subtidal and intertidal estuarine macrophytes, some of which occurs in small patches or narrow fringing bands. Examples are provided to demonstrate how data generated from these maps can be used and some of the possible limitations are discussed.