On the improvement of sea ice classification by fusing C- and L-band SIR-C polarimetrìe quantities

The benefits of combined C-band with L band polarimetric quantities for supervised sea ice classification in the Eastern Weddell Sea, Antarctica, are investigated. In experiments, we compared the performance of a maximum likelihood classifier when used with the combined preferred polarimetric parameter sets against using the individual preferred polarimetric parameter sets. The relationship between classification accuracy and preferred number of polarimetric parameters to use for classification was examined, as well as whether locally linear embedding (LLE)[1] can be used to reduce the dimensionality of the parameter set. Combining deal-frequency polarimetric quantities often improves classifier accuracy against using individual single-frequencies. With the increase of dimensionality of the preferred polarimetric parameter set, the classification using high-dimensionality results in improvements over the smaller subsets or does not make a statistically significant difference. Using all available polarimetric quantities is recommended over the test site, data fusion with locally linear embedding (LLE) does not offer any benefit for sea ice classification.