InSAR Time-Series Analysis With a Non-Gaussian Detector for Persistent Scatterers

Abstract

Persistent scatterer interferometric synthetic aperture radar (PS-InSAR) is a remote sensing technique that maps patterns of crustal deformation with large spatial coverage and fine resolution. By identifying a network of reliable points called persistent scatterers (PS) that are used to improve estimates of decorrelation and atmospheric noise, PS-InSAR extracts measurements in areas that are difficult to analyze with traditional InSAR. PS detection has traditionally relied on Gaussian-based models despite the fact that radar backscatter is highly non-Gaussian (NG) in nature. Here, we demonstrate a novel PS detector, abbreviated as the NG detector, that is designed to account for the NG behavior observed in radar backscatter. We showcase its performance using two case studies: 1) Kilauea Crater in Hawaii and 2) the city of Corcoran in California's Central Valley. The NG detector finds significantly more PS than its Gaussian equivalent in both areas. Compared to the existing StaMPS PS algorithm, the NG detector finds slightly fewer PS in Kilauea and more PS in the Central Valley. We retrieve the deformation time-series over both areas using the NG detector and compare it with those derived from the Gaussian-based maximum-likelihood estimator detector, StaMPS, and also the small baseline subset (SBAS) method and GPS. We find that all results are consistent in Kilauea while the NG detector is the only method to accurately track the deformation measured by GPS in the Central Valley. Overall, our results suggest that PS-InSAR performance over challenging terrain can be improved by incorporating NG statistics into PS detection schemes.

Keywords

• Interferometric synthetic aperture radar (InSAR)
• non-Gaussian (NG) detection
• persistent scatterers (PS)
• time-series