Data and Code for "A Deep Learning-Based Downscaling Model for Arctic Sea Ice Concentration Projections"

This archive provides the data and code supporting the study “A Deep Learning-Based Downscaling Model for Arctic sea ice concentration projections.” It includes essential input data, training and evaluation scripts, statistical results, source data for figures, independent-period predictions for 2024–2025, and downscaled CMCC-CM2-VHR4 sea ice concentration projections for 2030–2050. Variable-ablation results are provided for 2015 and 2022. The complete ERA5 and observational source datasets, along with training caches, are not included. For details on model configurations, evaluation protocols, file contents, and reproduction requirements, see DATA_AND_CODE_GUIDE.md.