A Rule-Based Automatic Approach for Mapping Intertidal Seagrass Meadows Using Optical and Synthetic Aperture Radar Images

Abstract

Intertidal seagrass meadows are experiencing considerable fading caused by climate changes and anthropogenic stressors, which pose threats to the achievement of Sustainable Development Goals (SDGs). Thus, monitoring the dynamics of intertidal seagrass has become urgent. Satellite monitoring of intertidal seagrass meadows remains challenging due to similar spectral characteristics among neighboring coastal vegetation and the impact of tidal fluctuations. Thus, we presented a novel algorithm called rule-based Automatic Mapping through integrating Optical and SAR images for intertidal Seagrass meadows (AMOSS). First, we revealed that seagrass has a lower backscatter coefficient compared to that of mangroves or salt marshes in VH (vertical transmit and horizontal receive) polarization. Utilizing the distinct difference, we excluded mangroves or salt marshes and extracted the low-tide level zones with seagrass from Sentinel-1 images using the Otsu algorithm. Intertidal seagrass meadows were then mapped automatically using a multi-binary classification method with Sentinel-2 images. Furthermore, we developed a change detection method to track the dynamic changes in intertidal seagrass meadows from 2019 to 2023. The algorithm was applied to 15 intertidal seagrass meadows, ranging from tropical to subpolar zones, to demonstrate its robustness and generality. The overall accuracy of the AMOSS algorithm exceeds 84% across all 15 study sites, indicating its effectiveness even in complex seascape settings that include various intertidal vegetation types. This rule-based automatic algorithm enables large-scale identification, comprehensive surveys, and continuous monitoring of intertidal seagrass meadows, potentially uncovering previously unknown seagrass meadows.