Asynchronous Remote Sensing Time-Series Fusion for Cloud Removal and Anytime Reconstruction

2026-05-26

AI Summary

This work addresses the challenge of reconstructing long-term, cloud-free time series from asynchronously and irregularly acquired Sentinel-1 SAR and Sentinel-2 optical imagery. To this end, the authors propose AGFlow, a novel spatiotemporal generative flow matching model that integrates timestamp-conditioned alignment, joint spatiotemporal denoising, and on-demand query mechanisms. AGFlow enables seamless fusion of unpaired, asynchronous SAR and optical data without requiring pre-aligned inputs or fixed observation times, allowing for flexible generation of clear surface reflectance images at arbitrary timestamps. Evaluated on the RESTORE-DiT benchmark, AGFlow outperforms existing methods by reducing MAE and RMSE by 16–19% for fully missing frames, demonstrating significantly improved reconstruction quality over long temporal gaps and superior cloud removal performance.

Abstract

Frequent cloud cover severely limits the usability of Sentinel-2 (S2) optical time series for Earth surface monitoring. Sentinel-1 (S1) SAR provides all-weather complementary observations, but practical S1/S2 fusion remains difficult because acquisitions are irregular and asynchronous. Many existing approaches assume temporally aligned inputs (or require external nearest-date matching) and typically restore only observed timestamps, limiting reconstruction under long gaps and preventing on-demand synthesis. We propose AGFlow (Time Aligned Generative Flow Matching), a spatiotemporal flow-matching model for S1/S2 cloud removal and time-series reconstruction with three capabilities: (1) timestamp-conditioned internal alignment that fuses asynchronous S1 and cloudy S2 observations without preprocessing-based pairing; (2) spatiotemporal, context-aware denoising that models spatial structure jointly with temporal dynamics (rather than independent per-pixel time series); and (3) anytime querying, enabling generation of cloud-free S2 frames at both observed and user-specified timestamps within the monitoring window. We evaluate on the RESTORE-DiT benchmark protocol with quantitative metrics, qualitative comparisons, and component ablations. AGFlow notably improves fully missing-frame reconstruction (MAE and RMSE reduce by 16-19% over RESTORE-DiT) and provides reliable reconstructions under persistent gaps, while also yielding competitive cloud removal performance and flexible temporal querying for downstream tasks such as dense vegetation monitoring.

Problem

• cloud removal
• asynchronous fusion
• time-series reconstruction
• Sentinel-1/Sentinel-2
• anytime querying

Innovation

• asynchronous fusion
• flow matching
• anytime reconstruction
• spatiotemporal denoising
• cloud removal