How does one correct for the effects of topography on radar backscatter?
Topography within a radar image of the Earth affects its geometry and also modulates its brightness. How should one go about removing the effects of topography on radar backscatter to establish a “level playing field” when comparing multiple images that were acquired from different directions? In a recent publication, David Small introduces a new method by improving the sensor model to better incorporate terrain modulations and then compensate for them.
Estimating the local viewable area corresponding to every SAR pixel by integrating across the whole height model all the while testing for local radar shadow
The equations used to calculate SAR backscatter were originally developed using an ellipsoidal Earth model that is appropriate in the flatlands of Kansas or over water surfaces. The equations have been in need of an overhaul to incorporate land terrain effects. Radar effects such as layover, foreshortening, and shadow need to be incorporated in the sensor model that underlies the equations. This paper shows that conventional local incident angle approaches fail to adequately capture the effects in their sensor model. When the gamma nought backscatter convention is integrated with use of a digital height model for the first time, the local viewable area is estimated for each individual SAR pixel, and terrain-normalised backscatter estimates are output. Conventional local incident angle based terrain correction approaches that adopt the sigma nought backscatter convention are deficient in ignoring the effects of radar foreshortening and shadow within their sensor model. The flatness achieved with conventional methods was compared directly with the new method over a region with relatively homogenous land cover. The backscatter was shown to be significantly flatter when the new gamma nought flattening was applied.
Terrain-normalisation of the radar backscatter allows the label “terrain-corrected” to be applied not only to the geometry, but also the radiometry of the SAR image. Once the effects of terrain have been flattened from the radar backscatter, then thematic differences caused by local land cover variations that were previously outshone by stronger terrain-effects become clearly visible. For example, during the springtime snow melt season in the Alps, the distribution of wet snow extent appears dark, but that signal is mixed with terrain undulations in SAR imagery unless radiometric terrain correction (RTC) is applied.
Good terrain-normalisation for the first time enables inter-comparisons between diverse sets of SAR images, mitigating the requirement for an exact repeat track. For example, using the Envisat ASAR wide swath mode: one must wait at least 30 days for an exact repeat image, but can revisit an area on the Earth viewed from a different angle in 3 days or even less. Improved inter-comparability of imagery therefore improves the potential temporal resolution available for land cover monitoring. The technique is widely applicable, and relevant to all current and future SAR systems. Data from the upcoming Sentinel-1 and Radarsat Constellation Missions alone could potentially offer multiple observations over Europe and Canada every day.
Contact
David Small - david.small@geo.uzh.ch
Reference
Small, D. (2011). Flattening Gamma: Radiometric Terrain Correction for SAR Imagery. IEEE Transactions on Geoscience and Remote Sensing , Vol. 49, Issue 8, pp. 3081-3093.