地理加权弱监督贝叶斯高分辨率变换器用于全球 Arctic 海冰浓度映射与不确定性估计
摘要
尽管全球 Arctic 海冰高分辨率映射及其可靠对应的不确定性对于运营海冰浓度（SIC）制图至关重要，但由于关键挑战（如冰签名特征的微妙性、SIC标签的不准确、模型不确定性以及数据异构性），该任务备受挑战。本研究提出了一种新颖的贝叶斯高分辨率变换器方法，用于利用 Sentinel-1、RADARSAT 星座任务（RCM）和高级微波扫描仪2号（AMSR2）数据实现 200 米分辨率的全球 Arctic SIC 映射与不确定性量化。首先，为提高小且微妙的海冰特征（如裂缝/裂口、Pond 和冰 floe）提取效果，我们设计了一种具有全局和局部模块的新型高分辨率变换器模型，以更好地辨别海冰图案的细微差异。其次，针对低分辨率和不准确的 SIC 标签，我们设计了地理加权弱监督损失函数，在区域层面对模型进行监督，而非像素级别，并优先考虑纯淡水和冰堆签名，同时减轻边缘冰区（MIZ）模糊性的影响。第三，为提高不确定性量化， we designed a Bayesian extension of the proposed Transformer model, treating its parameters as random variables to more effectively capture uncertainties. Fourth, to address data heterogeneity, we fuse three different data types (Sentinel-1, RCM, and AMSR2) at decision-level to improve both SIC mapping and uncertainty quantification. The proposed approach is evaluated under pan-Arctic minimum-extent conditions in 2021 and 2025. Results demonstrate that the proposed model achieves 0.70 overall feature detection accuracy using Sentinel-1 data, while also preserving pan-Arctic SIC patterns (Sentinel-1 R² = 0.90 relative to the ARTIST Sea Ice product).
引用
@article{arxiv.2603.03503,
title = {Geographically-Weighted Weakly Supervised Bayesian High-Resolution Transformer for 200m Resolution Pan-Arctic Sea Ice Concentration Mapping and Uncertainty Estimation using Sentinel-1, RCM, and AMSR2 Data},
author = {Mabel Heffring and Lincoln Linlin Xu},
journal= {arXiv preprint arXiv:2603.03503},
year = {2026}
}
备注
23 pages, 20 figures