Observations, Modeling, and Applications of ERS-ENVISAT Coherence Over-Land Surfaces.

European Remote Sensing satellite (ERS) ENVISAT coherence is a new repeat-pass interferometric synthetic aperture radar product characterized by a very short repeat-pass interval between acquisitions (28 mm). In this paper, we investigate the properties of two 28-mins coherence images acquired over the regions of Paris, France, and Prague, Czech Republic, as well as a 35-day coherence image from Prague. Bare soils showed higher coherence compared to forests and urban areas. The range of coherence was larger for 28-mm pairs, perpendicular baseline closer to 2 km, and stable weather conditions between acquisitions. Coherence modeling shows that because of the long baseline: 1) significant surface decorrelation occurs when slopes are not accounted for in the common-band filtering and for variation of topography in the resolution cell and 2) volume decorrelation is significant. Temporal decorrelation seems to be relevant for vegetation even at the 28-mm interval time scale. Observations and modeled coherence were found to be in good agreement. Land cover classification using coherence and backscatter has been tested for the Paris scene for mapping the four major classes. The classification accuracy was 86% when an edge-eroded version of the reference land cover map was used. It decreases to 68% when the full land cover map was used as reference because of local mismatches between the coherence image and the land cover map, and the reduced common band in range. The high coherence difference between forests and bare fields suggests the possibility to use the ERS-ENVISAT coherence for forest/nonforest mapping and estimation of biophysical properties of short vegetation. [ABSTRACT FROM AUTHOR]