Coherent Azimuth Ambiguity Removal from Along-Track Interferometric Synthetic Aperture Radar Data for the Harmony Mission

Abstract

Coherent azimuth ambiguities in interferometric synthetic aperture radar (InSAR) can be a major systematic error source if accompanied by a low suppression capability. They induce biases in the interferometric phase, which limits the achievable performance and hence, degrades the viability of remote sensing data for scientific usage. This work derives, evaluates and compares algorithms to remove the effect of coherent azimuth ambiguities on 2- and 3-channel along-track interferometry (ATI). The interferometric performance is evaluated under the assumption of identically independently distributed data, which is simulated using a simplified one-dimensional (azimuth) InSAR channel model with coherent ambiguities. This linear time-invariant InSAR channel is further described analytically to ease the development of removal algorithms. The InSAR channel configuration bases on the future Earth Explorer 10 Harmony mission, which exhibits a low azimuth ambiguity suppression capability. The mission goal is to observe ocean and sea surface currents to refine global Earth system models. To simulate ocean and sea surface scenes, a frozen multiplicative noise model is used, where parameters of interest are described from empirical data in different sea states. One investigated algorithm on interferogram level (2 channels) minimizes the mean square error (MSE) of the interferometric estimate by means of a linear function of interferometric looks on the scene. The work derives this linear estimator from the previously found analytical channel model. It achieves removal of the effect of coherent azimuth ambiguities at a slight degradation of overall interferometric performance. It is further compared, which improvement of the Cramér-Rao bound (CRB) is made by an additional third channel. Using this third channel, the suppression capability of two well-known algorithms from harmonic analysis are evaluated. Both algorithms -- minimum variance distortionless response (MVDR), also known as Capons beamformer, and multiple signal classification (MUSIC) -- achieve suppression of the effect of coherent ambiguities, but, residual biases pose a limit on the minimum achievable performance.