FRInGE; Full-Resolution InSAR timeseries using Generalized Eigenvectors

InSAR time-series analysis at full SAR resolution is crucial to improve our understanding of many geophysical processes such as near-fault deformation, aseismic creep, landslides, volcanic deformation and in addition monitoring infrastructures in urban regions. However, stack processing and time-series analysis at full SAR resolution and using all possible interferometric pairs in the era of operational SAR missions (e.g., Sentinel-1 and NISAR) is challenging.

We present FRInGE, our newly developed InSAR time-series analysis toolbox which implements the state-of-the-art methods to efficiently explore the full capability of modern SAR observations to produce ground displacement time-series at full resolution of SAR images at pixels dominated by Permanent Scatterers (PS) or Distributed Scatterers (DS). For DS pixels, FRInGE uses all possible interferometric pairs. Starting from a coregistered stack of SAR images, FRInGE evaluates the temporal distribution of the backscatter of SAR images at each pixel to identify statistically self-similar neighbors in a pre-defined window. The neighborhood map also helps to identify PS pixels. FRInGE forms a full covariance matrix of interferometric pairs for each DS pixel over its neighborhood. The wrapped phase time-series of the DS pixels is estimated as the phase of the generalized eigen vector corresponding to the largest eigen values of the covariance matrix. The phase series of the PS pixels are also extracted and together with the estimated phase series of DS pixels can be unwrapped through a minimum span tree network using two-dimensional available unwrappers. The estimated unwrapped phase time-series of DS and PS pixels inclusive of all components (e.g., ground displacement, geometrical residuals, propagation delay) can be post-processed through available tools to estimate displacement time-series. FRInGE enables efficient generation of a full coherence matrix (all possible pairs) which can be used for Damage Proxy Mapping and change detection. We will demonstrate the superior performance of the technique compared to regularly down-sampled displacement time-series for different case studies including the post-seismic deformation of the Mw 7.3 2017 Sarpol-e Zahab earthquake and aseismic fault movements in Northern California.