多源SAR地表形变多尺度时空融合方法及城市应用OACHSSCD

Fine-scale monitoring of land subsidence in the complex environments of coastal megacities is crucial for urban safety and disaster mitigation.Since single Synthetic Aperture Radar(SAR)sensors struggle to balance wide-area continuous coverage with local fine-scale resolution,we propose a spatiotemporal fusion method for multi-source SAR data that captures both multi-scale spatial features and non-linear temporal evolution.Spatially,we construct a reconstruction model based on frequency separation and robust surface adjustment.By employing a reliability weighting strategy,high-resolution SAR local details are adaptively injected into the wide-area deformation background,effectively eliminating systematic biases and step jumps at data boundaries.Temporally,to overcome the time-lag inherent in traditional static weighting and unidirectional filtering,a state-space-based Forward-Backward Kalman Smoother(FBKS)Smoother is introduced to enforce bidirectional assimilation constraints using full time-series contextual information.The proposed method was applied to Sentinel-1 and TerraSAR-X data of Shanghai(2023-2024)and validated by concurrent leveling measurements.Results demonstrate that embedding local TerraSAR-X micro-deformations into the wide-area Sentinel-1 subsidence background successfully unifies point,line,and surface deformation features.Quantitative temporal validation further indicates that the FBKS dynamic smoothing effectively suppresses outliers and high-frequency noise,preserving long-term trend consistency and enhancing temporal robustness.The reconstructed subsidence field provides reliable technical support for understanding urban deformation mechanisms and ensuring critical infrastructure safety.