Extraction of winter total sea-ice concentration in the Greenland and Barents Seas from SSM/I data

Publisher Website
Google Scholar
Abstract
Significantly different sea-ice concentrations estimated by the well known Bootstrap and NASA/Team SSM/I algorithms are found to occur when the brightness temperature of horizontally polarized radiation is depressed, possibly as a result of ice layers in the snow cover. Furthermore, discontinuous ice concentrations, which do not reflect real concentration variations, sometimes occur when the Bootstrap algorithm switches between polarization and frequency schemes. The Bristol algorithm is designed to overcome these problems, and is described in this paper. In an initial evaluation against 10 cloud-free Advanced Very High Resolution Radiometer (AVHRR) winter scenes of the Greenland and Barents Seas, the Bristol algorithm, with a correlation coefficient (c corr) of 0·88 and r.m.s. error (e r.m.s.) of 5·2 per cent, outperforms the NASA/Team (c corr = 0·82,e r.m.s. =6·2 per cent), Bootstrap (c corr = 0·74,e r.m.s. = 8·3 per cent) and AES/York (c corr = 0·82,e r.m.s. =6·4 per cent) algorithms, although further work is required to confirm its possible advantages.