Lihe et al. (2026) Reconstructing cloud-free Sentinel-2 time series under complex degradations with state-driven spatio-temporal modeling

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-09-12
• Authors: Ziyang Lihe, Qiangqiang Yuan, Jiang He, Liupeng Lin, Xianyu Jin, Huanfeng Shen, Liangpei Zhang
• DOI: 10.1016/j.rse.2026.115661

Research Groups

School of Geodesy and Geomatics, Wuhan University
Collaborative Innovation Center of Geospatial Technology, Wuhan, China
Chair of Data Science in Earth Observation, Technical University of Munich, Munich, Germany
School of Resource and Environmental Sciences, Wuhan University, Wuhan, China
State Key Laboratory of Information Engineering, Survey Mapping and Remote Sensing, Wuhan University, Wuhan, China

Short Summary

This paper proposes a novel approach for reconstructing cloud-free Sentinel-2 time series under complex degradations using state-driven spatio-temporal modeling. The proposed method achieves robust reconstruction in global scenarios.

Objective

• Reconstruct high-quality satellite image time-series (SITS) under complex degradations such as clouds, blur, and sensor noise.

Study Configuration

• Spatial Scale: Global scale with extensive evaluations on globally sampled sites.
• Temporal Scale: Long-term Earth surface dynamics monitoring with a focus on temporal continuity of high temporal frequency optical observations.

Methodology and Data

• Models used: State-driven Spatio-Temporal Reconstruction approach through Directed Information Routing (STRIDE) which includes state-aware directed information routing (SDIR), manifold-mapped temporal evolution (MMTE), asymmetric spatio-temporal transformer (ASTT), and auxiliary degradation estimation module.
• Data sources: Sentinel-2 satellite data, synthetic aperture radar (SAR) data.

Main Results

• The proposed STRIDE approach achieves robust SITS reconstruction in global scenarios with improved fidelity of remote sensing indices.
• Extensive evaluations demonstrate the superiority of our approach over existing methods.

Contributions

• Original contribution is the development of a state-driven spatio-temporal modeling approach for reconstructing long-term SITS under complex degradations.
• The proposed method addresses the limitations of existing models by incorporating continuous temporal correlation and degradation-aware routing mechanisms.

Funding

• This research was funded by the Collaborative Innovation Center of Geospatial Technology, Wuhan, China.

Citation

@article{Lihe2026Reconstructing,
  author = {Lihe, Ziyang and Yuan, Qiangqiang and He, Jiang and Lin, Liupeng and Jin, Xianyu and Shen, Huanfeng and Zhang, Liangpei},
  title = {Reconstructing cloud-free Sentinel-2 time series under complex degradations with state-driven spatio-temporal modeling},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115661},
  url = {https://doi.org/10.1016/j.rse.2026.115661}
}