Regional classes of sea ice cover in the East Antarctic pack observed fromsatellite and in situ data during a winter time period

Ice concentration data alone are often of limited use in many process, and modeling studies as different ice regimes of approximately 100% concentrati on can have significantly different heat flux, albedo, and other surface pr operties. Current ice concentration algorithms perform poorly in regions of predominantly thin or highly fragmented ice, which constitute a significan t proportion of the pack in East Antarctica. The impact of the sea ice cove r on high latitude air-sea interactions and marine ecology depends not only on ice extent and concentration but also on the ice-type composition of th e pack. An unsupervised ice classification scheme, using data from four cha nnels of the SSM/I, is presented and tested as a means of gaining important additional, complementary information on surface type. Class interpretatio n is by comparison with AVHRR, ERS-1 SAR, and near-coincident digital aeria l photography and in situ data. The classification does a reasonable job at consistently differentiating the large-scale constituent regimes, includin g the outer marginal ice zone, the interior pack, and a transition zone sep arating the two. Given the short period of data analyzed, the cluster maps appear to be generally coherent and consistent through time as the pack cha nges in response to synoptic-scale atmospheric forcing, although the robust ness of the technique needs further testing over longer time periods. An ob served crossover int he meridional brightness temperature profiles is a dom inant and consistent feature which marks the transition from unconsolidated and wet ice in the marginal ice zone to more consolidated ice with a thick er and drier snow cover in the interior pack. Ambiguities occur at the boun daries of some of these regimes due to sensor resolution limitations and th e mixing of different ice types and open water. Also, some ice classes (lik e brash ice) cannot be distinguished from forming pancake ice. Furthermore, although the signature of the inner pack is usually distinct from that of the outer pack, we observed one extraordinary swell propagation event which led to ice fracturing and surface wetting, and significantly altered the s urface classification. The results of this multiparameter study underline t he importance of using multisensor systems synergistically to improve inter pretation of passive microwave data and better characterize the complex Ant arctic pack. Published by Elsevier Science Inc.