A combined remote sensing/modelling approach for the retrieval of sea ice thickness in East Antarctica

Sea ice thickness is an essential parameter for the estimation of energy, mass and momentum exchange and sea ice mass balance in the Southern Hemisphere. In situ methods of estimating the ice thickness distribution cannot provide the necessary areal coverage, while the coarse resolution of numerical models is unsuitable for parameterising small-scale processes. Remote sensing technology has the potential to address these issues of areal coverage and spatial resolution. A procedure for estimating the distribution of sea ice thickness using remote sensing data from the Advanced Very High Resolution Radiometer sensor is presented. This procedure is based on the method described by Yu and Rothrock [1996] for Arctic sea ice and relies upon the assumptions a) that the surface temperature of thin sea ice is closely related to its thickness and b) that there is a linear vertical temperature gradient across the recently-formed sea ice. Considerable modifications have been made to the Yu and Rothrock method in order to apply it to Antarctic conditions. Broadband albedo, ice/snow surface temperature and near-surface air temperature were estimated for cloud-free AVHRR pixels identified by an \expert system\" algorithm. These variables were used as inputs to a thermodynamic model to estimate ice thickness for two areas of the East Antarctic sector during the austral winter of 1995 and spring of 1996. The sensitivity of the model was examined with reference to the estimation of near-surface air temperature effective thermal conductivity of snow bulk transfer coefficients for heat and evaporation and the calibration of the satellite sensor's gain. Resultant ice thickness maps and frequency distribution histograms were compared with concurrent ship-based sea ice observations and digital aerial photography. Comparisons were also made with historical (\"climatological\") data including drill hole measurements ship-based observations and Special Sensor Microwave/Imager-derived sea ice extent data. In spite of the relative lack of concurrent data to validate the results the analysis of the 1996 data showed good agreement with ship-based observations from previous years and with estimates of areal coverage obtained from passive microwave data. The ice thickness distribution in the Prydz Bay area for the months of October and November is also consistent with ice drift patterns regional iceberg distribution and ocean circulation described in the literature for that area. Important by-products of the method proposed are maps of ice/snow surface physical temperature and albedo. Comparisons with near-contemporary digital aerial photography demonstrated the procedure's ability to resolve features such as icebergs and large floes and the transition between areas of open water and/or thin ice and areas of thick snow covered ice. The model's response to the presence of more than one surface type within a single pixel depends on their radiometric characteristics and relative concentration. The small number of images from August 1995 precluded a more detailed analysis but it was clear that the larger solar zenith angles that occur in winter significantly affected the result. In spite of these problems it was possible to interpret the thickness distribution in terms of the meteorological and geophysical conditions"