Analysis of consistency in first-year sea ice classification potential of C-band SAR polarimetric parameters

Abstract

The consistency in first-year sea ice classification potential of C-band SAR polarimetric parameters was analyzed by comparing the results of two studies conducted for the same ice types under different geophysical settings. The SAR images used in the comparison were acquired at an incidence angle difference of 4°. Probability density functions, grey level parameter images, and classification statistics derived using k-means classifier were used in the comparative analysis. The investigation showed that not all polarimetric parameters exhibit consistency in their classification performance under different geophysical settings. Out of the 20 polarimetric parameters analyzed, 12 demonstrated high levels of classification consistency between the two studies. Among these 12 parameters, only four possessed high classification accuracy and could be applicable for sea ice classification under variable environmental conditions. The parameters that showed the highest classification accuracies in both the studies were found to be inconsistent in their ice type separation capabilities. The signatures of these parameters differed for one or more ice types when compared between the two studies. The utility of these parameters in individual sea ice classification studies is recommended but their relevance in generalized sea ice classification scheme is unclear.