Flattening Gamma: Radiometric Terrain Correction for SAR Imagery

Enabling intercomparison of synthetic aperture radar (SAR) imagery acquired from different sensors or acquisition modes requires accurate modeling of not only the geometry of each scene, but also of systematic influences on the radiometry of individual scenes. Terrain variations affect not only the position of a given point on the Earth's surface but also the brightness of the radar return as expressed in radar geometry. Without treatment, the hill-slope modulations of the radiometry threaten to overwhelm weaker thematic land cover induced backscatter differences, and comparison of backscatter from multiple satellites, modes, or tracks loses meaning. The ASAR & PALSAR sensors provide state vectors and timing with higher absolute accuracy than was previously available, allowing them to directly support accurate tie-point-free geolocation and radiometric normalization of their imagery. Given accurate knowledge of the acquisition geometry of a SAR image together with a digital height model (DHM) of the area imaged, radiometric image simulation is applied to estimate the local illuminated area for each point in the image. Ellipsoid-based or sigma naught (σ⁰) based incident angle approximations that fail to reproduce the effect of topographic variation in their sensor model are contrasted with a new method that integrates terrain variations with the concept of gamma naught (γ⁰) backscatter, converting directly from beta naught (β⁰) to a newly introduced terrain-flattened γ⁰ normalization convention. The interpretability of imagery treated in this manner is improved in comparison to processing based on conventional ellipsoid or local incident angle based σ⁰ normalization.

I. Introduction

The normalization of synthetic aperture radar (SAR) imagery for systematic terrain variations is required for meaningful multi-sensor or even single-sensor multi-track intercomparisons. Accurate backscatter estimates enable more robust use of the retrieved values in applications such as the monitoring of deforestation, land-cover classification, and delineation of wet snow covered area. Accurate estimates of backscatter in the presence of severe terrain furthermore relax constraints on same-orbit exact-repeat observations for change detection: this enables shorter temporal intervals between observations, especially given wide swath imagery, and also opens the door to multi-sensor backscatter overlays. Where the local terrain is ignored due to either lack of DHM-availability or runtime-constraints leading to a need for a simpler Earth model, then the quality of the retrievable backscatter estimate is compromised. This paper extends prevailing traditional concepts of backscatter normalization, introducing a new standard known as terrain-corrected gamma naught.