Key points are not available for this paper at this time.
• A novel LOSFLC framework for salt marsh mapping; • 10 m resolution wetland maps for the YRD and JDS from 2017 to 2024; • Spartina alterniflora removal improve native salt marsh habitat; • Wetlands face new ecological risks from declines in salt marsh and land areas. Coastal wetland patterns in China have undergone dramatic changes in recent years due to the invasion and subsequent management of Spartina alterniflora (S. alterniflora). The ecological benefits of clearing S. alterniflora still lack sufficient exploration. SAR data can provide complete time series observations covering the full phenological cycle of salt marsh vegetation, thereby offering a valuable data source for accurate wetland monitoring. However, the similarity in backscatter values among salt marsh vegetation types poses a challenge for wetland classification. To address this issue, we propose a layer-wise classification framework based on local optimal spatiotemporal features fusion (LOSFLC). We selected the Yellow River Delta wetland (YRD) and Jiuduansha wetland (JDS), which differ in geographical location and have been severely invaded by S. alterniflora, as the study areas. Using 443 Sentinel-1 images, we generated 10-m resolution wetland maps from 2017 to 2024. Classification experiments using LightGBM, XGBoost, and RF demonstrated that LOSFLC improved OA by 1. 64%–3. 44% compared to RFE. Under the optimal scheme, the OA reached 96. 07% and 96. 30% for the YRD and JDS, respectively. In addition, long-term mapping results revealed that the removal of S. alterniflora has achieved significant restoration outcomes, with native salt marsh species regaining habitat and Phragmites australis becoming the dominant type. Nonetheless, the reduction in salt marsh vegetation and land areas highlights emerging risks for wetland ecosystems. The proposed method provides a new technical reference for accurate SAR-based wetland mapping.
Li et al. (Wed,) studied this question.