Li et al. (2026) A weakly supervised, wetness-based dynamic framework for wetland delineation across globally distributed sites

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-08-08
• Authors: Yang Li, Nandin‐Erdene Tsendbazar, Kirsten M. de Beurs, Lassi Päkkilä, Lammert Kooistra
• DOI: 10.1016/j.rse.2026.115604

Research Groups

• Laboratory of Geo-information Science and Remote Sensing, Wageningen University & Research
• Water, Energy and Environmental Engineering Research Unit, Faculty of Technology, University of Oulu

Short Summary

This study presents a weakly supervised, wetness-driven framework using earth observation data streams to improve wetland monitoring across regions represented by reference datasets. The framework achieved moderate spatial delineation accuracy and delivered the highest temporal accuracy.

Objective

• Evaluate the potential of OPTRAM-based wetness retrieval for weakly supervised, scene-based wetland delineation across globally distributed sites.
• Assess the performance and applicability of the OPTRAM model across various vegetation conditions.

Study Configuration

• Spatial Scale: 0.1° grid cells
• Temporal Scale: Daily to annual time series

Methodology and Data

• Models used: OPTRAM (OPtical TRApezoid Model)
• Data sources:
  • HLS (Harmonized Landsat-Sentinel) imagery
  • Copernicus Digital Elevation Model (DEM)
  • MERIT Hydro global hydrography dataset
  • European Space Agency WorldCover v100

Main Results

• The framework achieved moderate spatial delineation accuracy (F1 = 0.65, recall = 0.75, precision = 0.58) and delivered the highest temporal accuracy (F1 = 0.72; precision = 0.81; recall = 0.64).
• OPTRAM-derived RWI improved thematic accuracy for vegetated wetlands.

Contributions

• The study contributes to the development of a weakly supervised, wetness-driven framework for improving wetland monitoring across globally distributed sites.
• The results demonstrate the potential of OPTRAM-based wetness retrieval for scene-based wetland delineation.

Funding

• This research was funded by [project/program name and reference code]

Citation

@article{Li2026weakly,
  author = {Li, Yang and Tsendbazar, Nandin‐Erdene and Beurs, Kirsten M. de and Päkkilä, Lassi and Kooistra, Lammert},
  title = {A weakly supervised, wetness-based dynamic framework for wetland delineation across globally distributed sites},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115604},
  url = {https://doi.org/10.1016/j.rse.2026.115604}
}