HotSAM: A Hotspot-Spatial-Adaptive Multi-Resolution framework for efficient long-term deformation monitoring in terrestrial radar interferometry

Long-term deformation monitoring using ground-based interferometric radar (GBIR) inherently generates massive datasets. Current full-resolution processing paradigms face severe computational and storage bottlenecks, primarily because they expend equal resources on vast stable areas and localized deformation zones. To overcome this redundancy, this study proposes HotSAM (Hotspot-Spatial-Adaptive Multi-resolution), a novel framework that shifts from the conventional “one-size-fits-all” approach to a resource-adaptive strategy. The framework employs a two-stage processing scheme that first rapidly delineates active deformation zones using low-resolution analysis and an enhanced deformation hotspot detection algorithm. This allows for computationally intensive, fine-resolution processing to be selectively applied only where necessary. Beyond efficiency, the framework incorporates a joint error estimation model designed to simultaneously decouple instrument repositioning errors from atmospheric phase screens (APS), ensuring robust time-series analysis in discontinuous monitoring campaigns. In addition, a phase-matching geocoding approach is developed to establish geographic mapping of the radar-derived deformation products without requiring dedicated external ground control points. Extensive validation using both simulated and real-world GPRI-II datasets demonstrates the framework’s superior performance. Compared to traditional workflows, HotSAM achieves a 73.04 % reduction in processing time and a 96.95 % reduction in storage requirements, while maintaining high consistency in deformation results with only sub-millimeter deviations (i.e., 0.79 mm). Furthermore, quantitative sensitivity analysis of key parameters, including multi-look factors and deformation gradients, provides operational guidelines for optimal deployment. HotSAM offers a scalable, high-efficiency solution for long-term geohazard monitoring and establishes a potential blueprint for big-data processing in future terrestrial applications.