Assessing the Transferability of Neural Network-Based Sea Ice Classification between the Arctic and Antarctic

Abstract

In this study, we trained a supervised neural network for sea ice classification using X-band SAR images obtained during an Antarctic fieldwork campaign and applied it to Arctic Xband data acquired during sea ice fieldwork campaigns. The results revealed a pronounced correlation between the predicted sea ice class and measured snow depth in the Arctic. This correlation is likely attributed to the Arctic acquisitions taking place during the cold winter conditions characterized by thicker snow layers, which resemble conditions in Antarctica.