Improved phase gradient stacking for landslide detection

Abstract

The advanced interferometric synthetic aperture radar (InSAR) provides an effective tool to detect landslides over a large area. However, it is greatly affected by atmospheric delays and phase unwrapping errors in a complex environment and requires massive calculations and analysis. These factors hinder InSAR from reliably and rapidly identifying landslides. In this study, we propose an improved phase gradient stacking (IPGS) method, which effectively suppresses atmospheric delay disturbance, topographic residuals, and noise while enhancing local deformation signals. The temporally stacked phase gradients with a preset step along four directions are merged to form a phase gradient map. It avoids complicated unwrapping and massive time series analysis. The simulation experiment demonstrates the improvement to traditional methods by combining four directions and a specific step. The IPGS method achieves a comparative landslide detection as the classical SBAS method in terms of Sentinel-1 datasets covering Danba County. Even for some small-scale landslides that are difficult for SBAS to detect, the phase gradients are distinct. A field investigation validates the reliability of IPGS-detected landslides. It provides an effective tool for large-scale, rapid, and reliable detection of geological disasters.