Classification of First-Year Arctic Sea Ice using Spaceborne Hyperspectral Imagery

The Arctic region is experiencing major transformations due to climate change. This includes sea ice loss that affects a positive feedback system between the Earth’s albedo and melting ice. In situ measurements of sea ice is difficult due to location and finances, which is why using hyperspectral remote sensing is ideal for studying the ice. This study analyzed hyperspectral data from the Hyperion satellite in the program ENVI to classify types of sea ice based on class of melt. The classification process used atmospheric correction (FLAASH) to remove water retrieval error then data reduction was achieved using Minimum Noise Refraction. The n-D Visualizer and the Pixel Purity Index tool allowed selection of unique endmembers to create a classification map in Spectral Angle Mapper. Field spectra measurements taken by another group in another location of sea ice, are compared with the classification created to improve accuracy. Afterwards a subsistence hunting area is speculated based on data. This method is proven to correctly differentiate between sea ice types by accuracy assements.