Mangrove species mapping in coastal China using synthesized Sentinel-2 high-separability images

The absence of national-scale mangrove species maps has hindered the precise estimation of their blue carbon storage ecological value evaluation, and effective management of protected areas. Mangroves typically grow in harsh intertidal environments, with non-mono species distributed together, and exhibit varied species compositions and appearances along the latitudes. Previous studies demonstrated the effectiveness of mapping localscale mangrove species using fine data sources from commercial satellites and unmanned aerial vehicles. However, because of cost and resource limitations, these approaches cannot be easily applied to areas with large latitudinal spans, where mangroves are commonly distributed. To fill the gap in national-scale mangrove species mapping, this study proposes a comprehensive solution by identifying high-separability images from synthesized Sentinel-2 images to constrain costs, designing a new classification scheme, applying probabilistic classification to exclude mixed communities, and partitioning the areas to control latitudinal variations. This approach was used to map nine mangrove species in China for 2020. The resultant map for partitioned areas had overall accuracies and kappa coefficients ranging from 83.8 to 86.4% and 0.806-0.849, respectively. The produced map exhibited a performance comparable to that of previous local-area maps derived from fine data sources and sophisticated methods. To the best of our knowledge, this study is the first to reveal the national-scale mangrove species distribution using remote sensing, report the mangrove species composition in seven mangrove national nature reserves, and analyze hotspots of nine mangrove species using adaptive kernel density estimation. Furthermore, this study provides a novel report on the distribution and concentration of two widely planted exotic mangrove species, Sonneratia apetala and Laguncularia racemosa. The produced map can be used as a training sample for future mangrove species mapping, enabling the generation of detailed maps and exploration of identification mechanisms of mangrove species. This study will benefit blue carbon research by providing a foundation for representative sampling that can be used for biomass and carbon estimation. Furthermore, it provides a valuable basis for evaluating ecological value of mangroves and their precise management.