Zoning seagrass protection in Lap An Lagoon, Vietnam using a novel integrated framework for sustainable coastal management

Seagrass is a key factor of the nature-based solution to climate change impacts, however, this resource has been lost and degraded worldwide in both of area and habitats. Protection of extant seagrass often requires a zoning approach to fit the local conditions and to identify the spatial prioritization. In this study, we integrated the state-of-the-art machine learning model (CatBoost) and Sentinel-2 multi-spectral imagery for mapping the extent of seagrass meadows, combined with the multi-criteria evaluation (MCE), fuzzy logic (Sigmoidal membership function), and analytic hierarchy process (AHP) technique in a GIS database to score the ecological protection zoning of seagrass ecosystem in Lap An lagoon, Vietnam. Seagrass map retrieved from Sentinel-2 multi-spectral imagery was used as the constrain factor, whilst salinity, water depth, substratum, and distance to aquaculture sites were conditioning factors. Our results presented accurate mapping of seagrass meadows in the study site (scores of overall accuracy, precision, and F1 are 0.93, 0.90 and 0.92, respectively) and indicated 22.05 ha (scores 0.66"”0.99) in high, 18.63 ha (scores 0.33"”0.66) in medium, and 10.80 ha (scores 0"”0.33) in low priority of protection in the southern and the eastern southwest parts of the lagoon, and the areas closed to the aquaculture sites, respectively. Our novel integrated approach to map the priority zones is useful for sustainable protection and management of seagrass meadows and provides a framework to strengthen the application of remote sensing and GIS-based techniques for further conservation of seagrass globally.