Improvement of interferometric SAR coherence estimates by slope-adaptive range common-band filtering

The accuracy of SAR interferometric coherence estimates depends on the precision of several processing steps. In particular decorrelation can occur if range common-band filtering does not perform optimally. Typically a planar surface is adopted which introduces additional decorrelation in case of sloped terrain. To take into account topographic variations a slope-adaptive range common-band filtering method has been developed. A DEM is used to simulate an unwrapped interferogram and the fringe rate is used as driver for the filter size in the range common-band filtering step. Tests with several spaceborne interferometric SAR datasets confirmed the robustness of the method. The improvement of the coherence increased for increasing perpendicular baseline. As a consequence, the fringe visibility also greatly improved. To quantify the improvement of the coherence estimates with the slope-adaptive range common-band filtering, we considered the variation of classification in forest mapping, i.e. an application in which accurate coherence estimates are needed. With improved coherence the classified forest stem volume agreed better with forest maps derived from other remote sensing datasets.

I. Introduction

Correct estimation of the interferometric SAR coherence plays a crucial role in several applications of the coherence itself as well as of the interferometric phase. For forest biomass retrieval for example, even a slight error in the coherence estimate can lead to an estimate of biomass having a relevant deviation from the actual value. The uncertainty of the interferometric phase estimate is strictly linked to the corresponding coherence level, which in turn means that decorrelation introduced by the processing chain could lead to substantial errors when estimating terrain elevation, displacements etc. Furthermore, the robustness of phase unwrapping is enhanced if the coherence is increased.