Characterizing Derawan seagrass cover change with time-series Sentinel-2 images

Abstract

In recent decades, seagrass ecosystems are declining globally. Providing spatial information of seagrasses has become growing needs in which data from satellite remote sensing platforms are proven to be useful. While multi-temporal archives of commercial and open access imagery are available, monitoring activity is rarely done frequently, limiting the information of seagrass dynamics particularly in tropical region. Here, we utilized time-series Sentinel-2 images to map seagrass percentage cover at intra-annual basis in two years’ time frame based on empirical method, while also compared maximum likelihood and random forest classifier to map dominant benthic cover type and determined seagrass spatial extent. We found that random forest classifier outperformed maximum likelihood classifier. Furthermore, a pattern of seagrass percentage cover change was present. Our results demonstrate how freely available Sentinel-2 images and limited field data could be useful in monitoring seagrasses at dense time. We anticipate this study to be a starting point for more complicated seagrass change analysis at denser time and link potential change drivers spatially.