Innovative SAR-optical data fusion for reflectance time series reconstruction in vegetation-covered regions

• Innovative SAR-Optical fusion for seamless cloud-free reflectance time series. • Superior accuracy in reconstructing missing data scenarios. • Lightweight architecture requires no additional training data. • Proven effectiveness in grassland and burnt area monitoring. Frequent cloud cover leads to gaps in remote sensing image time series, posing significant challenges for agriculture, grassland, and forest monitoring applications. This study proposes an innovative method to generate reflectance time series to address this issue, enhancing data reliability and application scope. We adopted a lightweight reconstruction model based on the architecture of deep learning. This model integrates the gapped Sentinel-2 reflectance time series caused by cloud cover with Sentinel-1 synthetic aperture radar (SAR) data to produce continuous high-density time series data. During the model training and evaluation, we selected time series image data from Sentinel-1 and Sentinel-2 in China and the United States, covering various typical geographical and climatic environments. The experimental results indicate that the proposed method performs excellently, significantly improving the completeness and accuracy of reflectance data, especially under conditions of prolonged cloud contamination leading to data gaps. To further validate the practical application of the model, we conducted a test in the grassland region of Mongolia to restore surface burnt areas. The results showed that this method performs excellently in restoring and detecting changes in burnt areas, significantly improving the accuracy and efficiency of detection. In summary, the method proposed in this study provides an effective solution to the problem of gaps in remote sensing image time series caused by cloud cover. This method not only enhances the reliability and practicality of remote sensing data but also demonstrates its broad potential and application prospects in various downstream applications, especially in fields such as agricultural monitoring and grassland disturbance detection.