Sea Ice
Sea ice research focuses on understanding its dynamics, distribution, and impact on climate, with a primary objective of improving prediction capabilities for navigation and environmental monitoring. Current research heavily utilizes deep learning, employing convolutional neural networks (CNNs), U-Nets, and transformers to analyze satellite imagery (SAR and optical) for sea ice classification and forecasting, often incorporating multi-temporal and multi-spectral data fusion. These advanced models aim to improve accuracy and efficiency compared to traditional physical models, particularly for short-term operational forecasting in key regions. Improved sea ice prediction is crucial for maritime safety, resource management, and a more comprehensive understanding of Arctic climate change.
Papers
Comparison of Cross-Entropy, Dice, and Focal Loss for Sea Ice Type Segmentation
Rafael Pires de Lima, Behzad Vahedi, Morteza Karimzadeh
Enhancing sea ice segmentation in Sentinel-1 images with atrous convolutions
Rafael Pires de Lima, Behzad Vahedi, Nick Hughes, Andrew P. Barrett, Walter Meier, Morteza Karimzadeh