Grad Seminar: Toward Enhanced Sea Ice Parameter Estimation: Fusing Ice Surface Temperature with the AI4Arctic Dataset using Convolutional Neural Networks
Abstract
Accurate Arctic sea ice mapping is essential for navigation and climate monitoring, but is hindered by remote sensing uncertainties and data limitations. The recent AI4Arctic dataset combines Sentinel-1 Synthetic Aperture Radar (SAR), AMSR2 brightness temperature, ERA-5 reanalysis data, and ice charts to enhance deep learning-based mapping approaches. However, AI4Arctic excludes thermal infrared data, which could improve predictions in regions where SAR and passive microwave measurements are challenging to interpret.
This study explores the use of VIIRS ice surface temperature (IST) to enhance multi-task sea ice parameter estimation. A subset of AI4Arctic is co-registered with IST, and we explore the impacts of integrating this data at both the input- and feature-levels. Using three variations on the U-Net architecture, we estimate sea ice concentration, stage of development, and floe size. Predictions are evaluated using sea ice maps and established performance metrics. Models incorporating IST outperform the AI4Arctic baseline, highlighting its potential. Practical limitations and key impacts of IST integration are also discussed.
Presenter
Lily de Loe, MASc candidate in Systems Design Engineering
Join in person, EC4 2101A