A new paper on arXiv introduces LC-TSRNet, a land-cover-guided two-stage network for reconstructing annual MODIS surface reflectance time series. The authors note that MODIS surface reflectance products frequently contain data gaps caused by cloud contamination and sensor malfunctions, which restricts their use in long-term and large-scale studies.
The paper points out that most existing reconstruction methods rely on cloud-free data, a condition that is often unavailable. LC-TSRNet instead incorporates land-cover guidance in a two-stage design, offering a different approach to filling these gaps.
Because the abstract is the only source available, details on the network's architecture and experimental results are not covered here. The significance lies in addressing a persistent data-quality problem that affects many environmental monitoring applications.