Research on InSAR Coherence Proxy and Optimization Method for Interferometric Network Construction in the Era of InSAR Big Data

With the explosive growth of interferometric synthetic aperture radar (InSAR) data, we have entered the era of InSAR big data. A key challenge in big data InSAR processing is efficient construction of an optimal interferometric network from the vast number of available interferometric pairs. The conventional small spatiotemporal baseline set method aims to enhance the overall coherence of the interferometric network by constraining temporal and spatial baselines; however, it still presents several shortcomings in practice. Coherence-based interferometric network optimization methods have emerged to improve deformation monitoring accuracy and spatial coverage density. Nevertheless, calculating coherence for all possible interferometric pairs is computationally inefficient and hinders the implementation of near-real-time InSAR processing. To address this issue, we analyzed long-term coherence variations with respect to temporal baselines, spatial baselines, and seasonal changes, and developed a coherence proxy model to estimate a priori coherence for interferometric pairs. Leveraging this model, we proposed an interferometric network optimization method that integrates graph theory algorithms, specifically the minimum spanning tree and phase triangular completion, to ensure robustness and temporal consistency. The coherence proxy model was validated using both synthetic data and real data from five representative regions across China with diverse land cover types, and the proposed interferometric network construction method was likewise evaluated on both synthetic and real datasets. The results demonstrate that the coherence proxy model performs well on both synthetic and real data, requiring only approximately 300–400 interferograms to reliably model the proxy and estimate the prior coherence of all potential interferometric pairs. The proposed network construction method yields a highly coherent and structurally robust interferometric network, enabling high-accuracy and high-coverage deformation monitoring. For Sentinel-1 data spanning three years, it reduces the coherence computation workload by up to 90%. In the case of incremental network construction with newly acquired data, the method can completely eliminate the heavy computational burden of coherence estimation, thereby supporting the advancement of near-real-time InSAR processing.