Sea ice type classification from multichannel passive microwave datasets

Passive microwave sensors have been described as the ¿workhorses¿ of sea ice observation. Frequency and polarization dependent variations in microwave emissivity are a function of the ice thickness, age, salinity, temperature, structure, snow cover and overall surface conditions. Whilst there are numerous algorithms for deriving sea ice area concentration which utilise a limited subset of the available channels, the use of unsupervised classification schemes has been limited. This paper examines a cluster analysis approach to automatic classification of the data and demonstrates that by clustering using emissivities from all channels it is possible to obtain classifications of the different ice regimes both in the seasonal and perennial ice cover.