ReefFM-SCS: physics-guided foundation-model-assisted weakly supervised framework for fine and stable coral reef extent mapping in the South China Sea

Coral reef extent mapping remains difficult in shallow, optically complex environments affected by water depth, optical attenuation, benthic mixing, and tidal disturbance. This challenge is acute in the South China Sea, where reefs are highly heterogeneous, manual annotation is costly, and existing global boundary products often lack local precision. We developed ReefFM-SCS, a physics-guided, foundation-model-assisted weakly supervised framework for fine and stable coral reef extent mapping. The framework integrates Allen Coral Atlas priors, bathymetry, the diffuse attenuation coefficient (Kd), spectral indices, and a dual-branch optical/physical-prior segmentation network. Rather than serving as the final segmenter, SAM2 is used only for candidate generation; reef-scale adaptive physical constraints guide pseudo-label construction, while quality-weighted multi-temporal fusion integrates annual predictions from 2016 to 2025. The resulting product represents the multi-year stable geomorphic reef-flat extent across major islands and reefs in the South China Sea with high spatial continuity. Compared with the Allen Coral Atlas baseline, precision, recall, F1 score, and IoU improved from 0.84, 0.86, 0.84, and 0.74 to 0.93, 0.95, 0.93, and 0.88, while the boundary metric increased from 0.22 to 0.53. The resulting product provides a consistent spatial baseline for reef habitat mapping, environmental assessment, and related regional applications.

Authors

Institutions

Publication Details

Journal
International Journal of Digital Earth
Published
2026-11-17
DOI
https://doi.org/10.1080/17538947.2026.2750721
Primary Topic
Coral and Marine Ecosystems Studies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

ReefFM-SCS: physics-guided foundation-model-assisted weakly supervised framework for fine and stable coral reef extent mapping in the South China Sea

Bo Ai, Jiaqi Yao, Xue Liu, Chongjun Feng et al.
International Journal of Digital Earth
Coral and Marine Ecosystems Studies
article

ReefFM-SCS: physics-guided foundation-model-assisted weakly supervised framework for fine and stable coral reef extent mapping in the South China Sea

Bo Ai, Jiaqi Yao, Xue Liu, Chongjun Feng, Xinming Tang, Zhen Wen
article en

Abstract

Coral reef extent mapping remains difficult in shallow, optically complex environments affected by water depth, optical attenuation, benthic mixing, and tidal disturbance. This challenge is acute in the South China Sea, where reefs are highly heterogeneous, manual annotation is costly, and existing global boundary products often lack local precision. We developed ReefFM-SCS, a physics-guided, foundation-model-assisted weakly supervised framework for fine and stable coral reef extent mapping. The framework integrates Allen Coral Atlas priors, bathymetry, the diffuse attenuation coefficient (Kd), spectral indices, and a dual-branch optical/physical-prior segmentation network. Rather than serving as the final segmenter, SAM2 is used only for candidate generation; reef-scale adaptive physical constraints guide pseudo-label construction, while quality-weighted multi-temporal fusion integrates annual predictions from 2016 to 2025. The resulting product represents the multi-year stable geomorphic reef-flat extent across major islands and reefs in the South China Sea with high spatial continuity. Compared with the Allen Coral Atlas baseline, precision, recall, F1 score, and IoU improved from 0.84, 0.86, 0.84, and 0.74 to 0.93, 0.95, 0.93, and 0.88, while the boundary metric increased from 0.22 to 0.53. The resulting product provides a consistent spatial baseline for reef habitat mapping, environmental assessment, and related regional applications.

International Journal of Digital EarthVol. 19(2)
Tianjin Normal University (CN), Southeast University (CN), Shandong University of Science and Technology (CN)
Coral and Marine Ecosystems Studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon