Concept-driven extraction of the Antarctic marginal sea ice zone from remote sensing image time series

Sea ice plays a significant role in global climate change. Marginal ice zone (MIZ) is defined as the transition zone between open ocean and pack ice where intensive air–ice–ocean–wave​ interactions between the ocean and the atmosphere occur. This definition of MIZ is rather vague, which affects its mapping. Previous data-driven methods extracted MIZ from single source data and conveniently used a single classification threshold. In this study, we propose to apply a concept-driven top down extraction method. We use a fuzzy inference system (FIS) to integrate multiple ice properties based upon the MIZ definition, and use membership functions to quantify the uncertainty of the threshold for several ice characteristics. This concept-driven method is applied to mapping the Antarctic MIZ between 2010 and 2021. Results show that a FIS successfully combines the ice concentration and ice thickness, while it allowed us to map the MIZ as objects with vague boundaries, thus well-presenting its indeterminate nature in space.