Validation of sea ice drift vector extraction from AMSR-E and SSM/I data by using MODIS data

Sea ice drift vectors extracted from satellite passive microwave sensors data are widely used today. However, the validation of which has not been done much so far. In this study, the sea ice drift vectors automatically extracted from pairs of the satellite passive microwave sensor AMSR-E images acquired in one or two days' interval were validated by comparing with sea ice drift vectors manually extracted from pairs of cloudless MODIS images acquired simultaneously with AMSR-E images. The cross correlation algorithm was used for automatically extracting sea ice drift vectors from passive microwave sensor images. Total of six pairs of cloudless MODIS images were selected for the validation. The six pairs of images consist of three from the Sea of Okhotsk, two from the Bering Sea, and one from the Antarctic Ocean. In order to evaluate the sensor differences, the sea ice drift vectors were also extracted automatically from a pair of SSM/I images. The sea ice drift vector extraction accuracy of AMSR-E images was 96% in the best case and was 55% in the worst case under the permissible error range of 6 km. This result suggests the usefulness of AMSR-E for sea ice drift monitoring. As for the polarization difference, the accuracies of H polarization were higher than V polarization in some cases, but were reverse in other cases. By comparing MODIS images with simultaneously collected passive microwave sensor images, it has become clear that the changes of sea ice distribution, including deformation and condition change of sea ice, sometimes reduce and sometimes increase the advantage of H polarization against V polarization. The accuracy of extracted vectors from AMSR-E images was higher than that of SSM/I reflecting the IFOV difference of the two sensors. However, since there were four hours observation time difference between SSM/I and the other two sensors, the accuracy of the SSM/I should be noted as a reference.