A Novel Approach to Generate Large-Scale InSAR-Derived Velocity Fields: Enhanced Mosaicking of Overlapping InSAR Data

Large-scale deformation fields are crucial for monitoring seismic activity, landslides, and other geological hazards. Traditionally, the acquisition of large-area, three-dimensional deformation fields has relied on GNSS data; however, the inherent sparsity of these data poses significant limitations. The emergence of Interferometric Synthetic Aperture Radar (InSAR) data offers an alternative, enabling the retrieval of large-area, high-resolution deformation velocity fields. Nonetheless, the processing of InSAR data is often complex, time-consuming, and requires substantial storage capacity. To address these challenges, various research institutions have developed online InSAR processing platforms. For instance, the LiCSAR processing platform provides interferometric images covering approximately 250 km × 250 km, facilitating scientific applications of InSAR data. However, the transition from individual interferograms to large-scale, three-dimensional deformation fields often requires additional processing steps, including ramp correction within the images, mosaicking between adjacent images, and the joint inversion of InSAR observations from different viewing angles. In this paper, we propose a novel method for splicing several individual InSAR velocity fields into continent-scale InSAR velocity maps, which takes along-track and cross-track mosaicking into consideration. This method integrates GNSS data with InSAR data and also considers the additional constraint of data overlap region. The efficacy of this methodology is substantiated through its implementation in InSAR observations of the eastern Tibetan Plateau. In some tracks, there are overlapping areas on the east and west sides, and the line-of-sight (LOS) value can be effectively corrected by using these overlapping areas with similar size for two cross-track mosaics. The root mean square error (RMSE) of these tracks was reduced by about 4% to 8% on average when verified using true values of GNSS data compared to no cross-track mosaic. In addition, a significant improvement of 30% in RMSE reduction was achieved for some tracks.