A Pattern‐Based Strategy for InSAR Phase Unwrapping and Application to Two Landslides in Colorado

Abstract

AbstractInterferometric synthetic aperture radar (InSAR) has been successfully used to map ground displacements associate with landslides. One challenge with InSAR is that the basic measurement of interferometric phase takes values between 0 and 2π instead of values representing total displacement relative to a stable reference frame. Phase unwrapping is necessary to reconstruct measurements of total displacement for use in quantitative analysis. Unwrapping approaches often assume that the absolute phase difference between two neighboring pixels should be smaller than π. In the presence of noise or high‐strain rates associated with fast‐moving landslides, aliasing of the phase (undersampling of the wrapped signal) can result in unwrapping errors and misinterpretation of total displacement. Here we use a pattern‐based strategy for phase unwrapping of InSAR observations of fast‐moving landslides, where we determine the unwrapped deformation field that is most similar to a scaled reference displacement map. We describe a range of metrics that we use to evaluate the most appropriate scaling for each interferogram. We also demonstrate the range of conditions where they perform well using synthetic data. For evaluation of the results, we generated UAVSAR wrapped interferograms over the Slumgullion landslide in Colorado where phase aliasing for interferograms with temporal baselines larger than 7 days is common. We compare the interferograms unwrapped with our approach against results from range offsets (pixel tracking). The comparison demonstrates that our approach can be used for time spans much larger than those where traditional phase unwrapping performs well.