Remote Sensing Identification and Mapping of Spartina Alterniflora in Chinese Mainland Coastal at 2 m Spatial Resolution

Spartina alterniflora (S. alterniflora)has become the most invasive alien plant along the coast of mainland China, severely threatening coastal ecosystem security. Accurate identification and mapping ofS. alternifloraare fundamental for its management, resurgence monitoring, and prevention of secondary invasion. To address the need for nationwide precise monitoring ofS. alternifloraand the challenges associated with multi-temporal, cross-scene high-resolution satellite remote sensing, this study utilized 2m spatial resolution GF-1 satellite remote sensing data covering the Chinese mainland coastal area in 2021. We proposed a Multi-Level Feature Cross-Fusion Dual Attention Mechanism Network (RMD-Net) tailored forS. alterniflorasegmentation. Additionally, we developed a pseudo area re-training semi-supervised learning method based on the Pseudo-Focal loss function, suitable forS. alterniflorasegmentation under large-area, small-sample conditions. This method achieves precise identification and mapping ofS. alternifloraalong the Chinese mainland coastal area in 2021. The proposed method demonstrates excellent segmentation performance and mapping accuracy, providing a technical foundation and data support for the precise monitoring and control ofS. alterniflorain China.