Studies of Antarctic sea ice concentrations from satellite data and their applications

Abstract

Large regional and temporal changes in the global sea ice cover have been observed recently. Because of the relevance of such changes to climate change studies it is important that key ice concentration data sets currently used for evaluating such a phenomenon are assessed for accuracy and interpreted properly. Sea ice concentrations derived from passive microwave data using the Bootstrap and NASA Team algorithms are shown to be generally consistent but are also observed to differ by 10–35% in large areas within the ice pack, especially in the Weddell Sea, Amundsen Sea, and Ross Sea regions. Comparative analyses of such passive microwave data with coregistered visible and infrared (i.e., Landsat, AVHRR, and OLS) data show a predominance of thick consolidated ice in these areas and good agreement with the Bootstrap algorithm results. The relatively low values from the NASA Team algorithm results are likely due to layering within the ice and snow and/or surface flooding, which are known to affect the polarization ratio. Large seasonality in the physical characteristics and emissivity of the ice cover is also observed, and in predominantly new ice regions, the ice concentrations from passive microwave data are usually lower than retrievals from Landsat and OLS data in which the thresholding technique is used. The passive microwave results are biased because of relatively low emissivity of new ice, but they may be more useful since the bias allows for the identification of areas of significant divergence and polynya activities. Such areas need to be identified since heat and salinity fluxes are proportionately increased in these areas compared to those from the thicker ice areas. Time series data from 1978 through 2000 also show a slight but insignificant positive trend of 0.17±0.33%/decade in ice extent which is consistent with slight continental cooling during the period. This is a big contrast to the observed negative trend of about −3%/decade in the Arctic sea ice cover. It should be noted, however, that because the overlap period for key instruments is just 1 month, the error due to changes in sensor characteristics, calibration, and threshold for the 15% ice edge may not be negligible.