China's Sub-meter Time-series Mangrove Map (STMM)

STMM (Sub-meter Time-series Mangrove Map) is the first national, sub-meter resolution mangrove time-series dataset for China, spanning 1990–2020. Derived using the Sub-meter Mangrove Transfer Learning Mapping (SMTLM) framework, STMM resamples Landsat imagery to sub-meter scale and employs a SegFormer-based deep learning model with layer-wise fine-tuning to overcome the scarcity of historical high-resolution samples. The dataset achieves overall accuracies of 93.4%–97.2% with F1-scores of 0.93–0.97, and accurately delineates mangrove boundaries, small and fragmented patches, and internal features such as creeks and open gaps. Compared with existing 30 m and 10 m products, STMM retrieves much finer spatial detail and 2–4 times more mangrove patches, providing a robust benchmark for assessing long-term mangrove dynamics, conservation outcomes, and ecological restoration in China.