Abstract
Detecting transient, low-amplitude displacements such as slow slip events (SSEs) using Interferometric Synthetic Aperture Radar (InSAR) and its time-series analysis is challenging because of noise effects. Although they have been connected to large earthquakes, validation of their generality requires more samples. Installed observation stations are not required for satellite SAR; thus, it has the potential to detect unknown SSEs in scarce observation areas. However, owing to the high computational cost and dependence on the user’s visual interpretation of InSAR-related techniques, an objective signal detection method for InSAR time series is urgently needed for efficient deformation detection. Therefore, this study developed an anomaly detection model from the sliding window k-nearest neighbour method, which was tuned for the InSAR time-series to detect unknown SSEs and statistically validated its capabilities. Model validation was performed using a synthetic SAR time-series dataset comprising simulated surface displacement and atmospheric delay turbulent components. The synthetic dataset included 100 sets of simulated SAR time series for three cases with maximum line-of-sight displacements of 3.8, 1.3, 1.0 cm, and a control case without displacement. The displacement detection capabilities of the cases were statistically evaluated. For 3.8 and 1.3 cm displacements, mean anomaly time series showed significantly higher values compared with the 95% error range (±1.96σ) of the control. Our implemented model detected transient displacements such as SSEs with at least a 1.3 cm displacement under the situation of current satellite SAR observation archives.

This publication has 16 references indexed in Scilit: