Enhancing 4-D Landslide Monitoring and Block Interaction Analysis With a Novel Kalman-Filter-Based InSAR Approach

Abstract

In recent years, Synthetic Aperture Radar Interferometry (InSAR) has become widely utilized for slow-moving landslide monitoring due to its high resolution, accuracy, and extensive coverage. By integrating data from various orbits/platforms and monitoring sources, one-dimensional (1-D) line-of-sight (LOS) InSAR measurements can be explored to infer three-dimensional (3-D) movements. However, inconsistencies in observation times among different orbits and monitoring sources pose challenges in accurately capturing dynamic 3-D movements over time (referred to as 4-D). In this study, we propose a novel method, termed KFI-4D that incorporates spatiotemporal constraints into the traditional Kalman filter. This enhancement transforms the underdetermined problem of 4-D movement acquisition into a dynamic parameter estimation problem, enabling precise monitoring of landslide movements. The KFI-4D method was evaluated using both synthetic data sets and real data from the Hooskanaden landslide, demonstrating an improvement exceeding 50% in root mean square errors (RMSEs) compared to conventional methods. Additionally, the high-resolution characteristics of InSAR-derived 4-D movements allow for the analysis of strain invariants, providing insights into block interactions and landslide dynamics. Our findings reveal that strain invariants effectively indicate the distribution and activity of landslide blocks and slip surfaces as well as their response to triggers. Notably, abnormal signals identified in strain invariants prior to the catastrophic event at Hooskanaden suggest potential for early warning of landslides. The future integration of data from advanced satellites, such as NISAR, ALOS4 PALSAR3, and Sentinel-1C, is expected to further enhance the KFI-4D method's capabilities, improving temporal resolution and early warning potential for landslide monitoring.