Ice ridge density signatures in high-resolution SAR images

Abstract

Abstract. The statistics of ice-ridging signatures were studied using high-resolution (1.25 m) and medium-resolution (20 m) SAR images over the Baltic Sea ice cover, acquired in 2016 and 2011, respectively. Ice surface profiles measured by the 2011 airborne campaign were used as validation data. The images did not delineate well the individual ridges as linear features. This was assigned to the random occurrence of ridge rubble arrangements that generate bright SAR returns. Instead, the ridging signatures were approached in terms of the local density of bright returns selected by a variably bright-pixel percentage (BPP). Density was quantified by counting bright-pixel numbers (BPNs) in pixel blocks with variable side length L. A statistical model for BPN distributions was determined by considering how the BPN values change with the BPP and was found to apply over a wide range of values for BPP and L. The statistical approach was also able to simulate a higher-BPP image when seeded by a low-BPP image. It was also found to apply to surface profile data analysed by counting ridge sail numbers in profile segments of variable length L. This provided a statistical connection between the bright-pixel density and the ridge density. The connection was studied for the 2011 data in terms of surface rubble coverage estimated both from the medium-resolution image and from the surface profiles. Apart from a scaling factor, both were found to follow the same distribution.