Visibility in synthetic aperture radar satellite data: metric formulation, observation scheduling, and orbit design

Abstract

Earth observation satellites serve as vital information gatherers for effectively addressing some of humanity’s most pressing challenges including management of limited resources and minimization of losses from disasters. Synthetic aperture radar (SAR) is a type of active remote sensing instrument that operates in the microwave portion of the electromagnetic spectrum and is a preferred Earth observation system thanks to the reliable imagery it can collect in all illumination and weather conditions. SAR data is acquired using a side-looking viewing geometry in which the radar is pointed perpendicular to the satellite platform’s direction of motion. This viewing geometry, in conjunction with the illuminated terrain’s topography, results in geometric distortions termed layover and shadow. These distortions degrade the utility of the collected imagery since they effectively obscure portions of the image and preclude extraction of actionable insights. While geometric distortions will be everpresent in SAR imagery, their location and coverage can be manipulated by controlling the relative orientation between the satellite and the illuminated topography. Such manipulation has historically been infeasible for legacy SAR satellites that collect globally consistent data sets under rigid operating requirements. However, the recent advent of commercial SAR satellite constellations has re-framed the practicality of carefully tuned observation geometries that maximize region of interest visibility. Commercial SAR constellations operate on a task-wise basis that grants data end-users flexibility in specifying desired observation parameters including acquisition times and observation geometries. However, a mismatch between on-orbit capabilities and delivered data quality exists due to a lack of formalized tools for planning observations with maximum region of interest visibility. Specifically, no systematic method for identifying visibility-favorable observation geometries exists. This dissertation addresses this gap in a stepwise approach. First, an extension of opensource radar processing software is developed that enables prediction of layover and shadow in a 2D distortion mask for any satellite-target relative geometry. Visibility metrics are then defined to represent the favorability of a particular observation geometry with respect to a distortion mask. The computation of visibility metric scores at geometries spanning the entire sample space enables creation of visibility maps that completely characterize the visibility characteristics of a given region of interest. To broaden the suitability of visibility maps for observation planning, a set of generalizable visibility maps are created to enable estimation of region of interest visibility characteristics in mission scenarios that are computationallyconstrained and information-limited. Visibility maps are then directly integrated into satellite operations by developing the first SAR observation scheduling algorithm that explicitly accounts for visibility. Finally, visibility is considered in the orbit design process to establish general guidance on optimal repeat ground track orbit parameters for pre-defined region of interest visibility characteristics. Region of interest visibility improvements of up to 90% are obtained for individual tasks when using the observation planning tools developed in this dissertation. Constellation-wide visibility improvements of 18% are achieved with modest reductions in traditional performance measures when integrating visibility into observation scheduling. Two-fold improvements in the visibility characteristics of observation opportunities are attained for orbits designed to maximize overpass geometry quality. The contributions of this dissertation are timely, given the concurrent proliferation of flexible, high-resolution SAR observation capabilities, and lay the groundwork for enabling the acquisition of SAR data that is maximally useful for limited resource management, disaster response, and other applications.