Precise coregistration, stationary and non-stationary azimuth offsets and challenges for TOPS time-series analysis

Time-series analysis of Synthetic Aperture Radar (SAR) images acquired with Terrain Observation with Progressive Scan (TOPS) mode requires precise coregistration of a stack of TOPS images to a reference coordinate system with accuracies better than 0.001 pixel in azimuth direction. This accuracy is an order of magnitude larger than the achievable accuracy using geometry based coregistration techniques with satellite precise orbits and Digital Elevation Models. To avoid phase discontinuities at the regions of burst overlap in TOPS interferograms, azimuth offsets obtained with geometry-based techniques need to be refined for possible misregistration. In the case of stationary azimuth misregistration (i.e., constant azimuth misregistration or slowly varying misregistration in the azimuth direction), Enhanced Spectral Diversity (ESD) technique can be used to adjust the azimuth offsets precisely and achieve coregistration accuracies better than 0.001 of an azimuth cell. We present an approach to estimate a time-series of azimuth misregistration using a Network-based Enhanced Spectral Diversity (NESD) method that reduces the impact of temporal decorrelation on coregistration. We evaluate the NESD performance using different stacks of TOPS images acquired by Sentinel-1 over different regions. Standard deviation of the estimated misregistration time-series for different stacks varies from 1.1e-3 to 2e-3 of the azimuth resolution, equivalent to 1.6 to 2.8 cm orbital uncertainty in azimuth direction. These values fall within the 1-sigma orbital uncertainty of the Sentinel-1 orbits and imply that orbital uncertainty is most likely the main source of the constant azimuth misregistration between different TOPS acquisitions. We further discuss the sources of non-stationary azimuth misregistration (i.e., azimuth misregistration which varies significantly in range or azimuth direction) and possible solutions to overcome these difficulties. We demonstrate how deviation of the SAR processing geometry (e.g., during SAR image formation) from the real TOPS acquisition geometry leads to significant azimuth misregistration and phase discontinuity in TOPS interferograms. We also discuss non-stationary azimuth offsets in TOPS data induced by inhomogeneity in ionospheric propagation delay.