Evaluation of Spectral Indices for Differentiating Melt Pond from Sea Ice.

Spectral indices using two reflectance bands were evaluated to implement fast differencing the melt pond from sea ice. The spectral indices comprise simple band-ratio calculations and these ratio-based methods can compensate light variations within individual images. The spectral indices developed for emphasizing ice, snow and water were evaluated with fine snow, melting snow having a shallow puddle of water on top and sea water reflectance spectra selected from ex-isting spectral libraries. Pairs of two bands of green and SWIR (shortwave infrared), red and SWIR, green and NIR (near infrared), and green and red wavelength regions were used as input bands for the spectral indices. The differences of the spectral indices between fine snow and water and be-tween melting snow and sea water for every pairs were calculated to compare performance of the spectral indices. The indices using green and SWIR band pairs and red and SWIR band pairs showed larger differences between fine snow and sea water and between melting snow and sea water than the indices using green and NIR band pairs and green and red band pairs. The highest separability inferred from averaged index differences between fine snow and water and between melting snow and water was obtained from the pairs of red and SWIR bands. The results can be applied to planning of a spectral imager specialized for polar region or to choose reasonable bands from existing remote sensing sensors.