First year sea ice characterization from Quad-pol H-A-α classification

摘要

At spring when solar irradiance and air temperature turn snow and ice into an array of surface scatterers, point scatterers become the main signature of ridged environments. This information is better retained in un-filtered T3 matrixes. On another hand a Lee filtered T3 matrix gives information about the diversity of a distributed target which is equally indicative of the higher variability within ridges. In order to keep both in a single set of data, we computed an hybrid matrix composed of the alpha parameter from a single look T3 matrix and the entropy of the same matrix from a Lee 3×3 matrix. As revealed running a Wishart classification algorithm, this approach increases greatly the contrast between ridged and flat areas. The statistics presented in the results section were computed from areas characterized as a) linear ridges, b) rubble fields, c) type 1 un-deformed ice and, d) type 2 un-deformed ice. Un-deformed type 1 ice dates back to the previous fall freeze-up period while un-deformed type 2 ice is a thinner ice formed later through the winter season when the ice pack open up under certain wind conditions and new ice can develop into a flat section. While this technique couldn't be used during most of the winter season, this may present a great potential to extract narrow linear structures when snow wetness increase surface scattering and therefore the occurrence of single and double bounce scattering mechanisms.