A human-machine collaborative approach for high-resolution monitoring of suspended sediment dynamics in data-scarce and optically complex waters
Global coastal suspended sediment concentration (SSC) plays a crucial role in regulating erosion, accretion, and geomorphological evolution. Although traditional remote sensing offers broad spatiotemporal coverage, high-resolution SSC retrieval remains challenging due to the lack of in situ calibrat...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-08-01
|
| Series: | International Journal of Applied Earth Observations and Geoinformation |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1569843225003619 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Global coastal suspended sediment concentration (SSC) plays a crucial role in regulating erosion, accretion, and geomorphological evolution. Although traditional remote sensing offers broad spatiotemporal coverage, high-resolution SSC retrieval remains challenging due to the lack of in situ calibration data. We propose a transfer-based human–machine collaborative learning approach that embeds expert knowledge to significantly reduce reliance on field data. The approach follows the Human-in-the-Loop paradigm and begins by integrating limited in situ samples with expert-labeled SSC regions from Sentinel-MODIS (2017–2024) imagery to construct an initial training set, where SMOTE is applied to address class imbalance. Expert feedback is then used to identify and correct misclassified areas, and the resulting high-confidence samples are incrementally incorporated to reconstruct local training subsets. Finally, expert-informed feature adjustment and geometric priors are integrated through hierarchical transfer learning, enabling the model to iteratively adapt with small-sample updates. Applied to China’s eastern coastal waters, the proposed approach demonstrates high predictive accuracy, explaining 95.1% of the variance in limited field observations. It identifies a 1.50% overall decline in SSC since 2017, primarily driven by reduced fluvial sediment input, while paradoxically revealing a 4.26% increase in the Yangtze Estuary due to anthropogenic shoreline modifications that enhance nearshore sediment retention. The model also captures climate-induced ocean stratification effects, which suppress vertical energy exchange, limiting sediment resuspension and lateral redistribution and thereby altering SSC distribution patterns. Overall, the approach enables accurate SSC retrieval under severe observational data scarcity, providing a scalable and transferable solution for sediment monitoring in optically complex coastal environments. |
|---|---|
| ISSN: | 1569-8432 |