Authors: Christopher Yen*, Brown University, Andrey N Petrov, University of Northern Iowa
Topics: Remote Sensing, Cryosphere, Coastal and Marine
Keywords: Hyperspectral, imaging, Alaska, sea ice, classification, melt index, polar bear
Session Type: Paper
Start / End Time: 5:20 PM / 7:00 PM
Room: Oak Alley, Sheraton, 4th Floor
Presentation File: No File Uploaded
Hyperspectral imaging for remote sensing was developed in the past few decades, and provide rich information in spatial and spectral areas. Hyperspectral imagers on spaceborne platforms capture near-continuous spectra and provide valuable insight on Earth materials such as Arctic sea ice. Arctic sea ice is a dynamic interface between the atmosphere and the ocean, and an important component of the global climate system. Knowledge of sea ice conditions are important for marine industries as well as local peoples. Furthermore, the study of sea ice is critical to understanding local ecology such as polar bears in a rapidly changing climate. Visible, infrared, and passive/active microwave sensors generally study sea ice parameters such as extent and thickness. This study presents an analysis of spring first-year sea ice off the coast of Point Barrow, Alaska using EO-1 Hyperion data. Atmospheric correction was done using the FLAASH module, then noise and data dimensionality reduction was applied with Minimum Noise Fractionation transformation. Pixel Purity Index and n-D Visualizer were used for user-defined endmember selection, and classification maps was made using Spectral Angle Mapper and Spectral Information Divergence. Accuracy assessment was performed using higher resolution EO-1 Advanced Land Imager data taken concurrently with the Hyperion data, as well as comparison with field spectra of sea ice taken by another research group in a different part of the Arctic. A normalized difference ice melt index (NDIMI) was devised to highlight melting sea ice, and a polar bear selection probability map was created.