Sea ice classification using Bayesian statistics

Using the scatterometer data from the ERS-2 satellite, a new classification algorithm for sea/ice discrimination, based on Bayesian statistics is introduced. Tests for inner consistency are described and the results discussed. The influence of spatial and temporal averaging is examined. Also a comparison with ice maps obtained by other models and measurements is made. It is compared with IFREMER ice maps for ERS and Quikscat, and with surface temperature data retrieved from ECMWF. This algorithm may be applied to the ASCAT scatterometer on board of the MetOp satellites as well.