Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bayesian Transformer for Pan-Arctic Sea Ice Concentration Mapping and Uncertainty Estimation using Sentinel-1, RCM, and AMSR2 Data

Mabel Heffring, Lincoln Linlin Xu

TL;DR

The paper introduces a high-resolution Bayesian Transformer tailored for Pan-Arctic sea ice concentration (SIC) mapping with explicit uncertainty quantification. It combines global (GloFormer) and local (LoFormer) attention modules to better capture subtle ice features and uses decision-level fusion of Sentinel-1, RCM, and AMSR2 data to address heterogeneity. By treating model parameters as random variables and applying Bayes by Backpropagation, the approach yields calibrated uncertainty estimates and improves SIC accuracy relative to non-Bayesian and other UQ methods. The work demonstrates robust uncertainty maps, particularly in the marginal ice zone, and provides a practical framework for reliable Pan-Arctic SIC mapping with potential for extension to other multi-sensor, multi-modal EO tasks.

Abstract

Although high-resolution mapping of Pan-Arctic sea ice with reliable corresponding uncertainty is essential for operational sea ice concentration (SIC) charting, it is a difficult task due to some key challenges, e.g., the subtle nature of ice signature features, model uncertainty, and data heterogeneity. This letter presents a novel Bayesian Transformer approach for Pan-Arctic SIC mapping and uncertainty quantification using Sentinel-1, RADARSAT Constellation Mission (RCM), and Advanced Microwave Scanning Radiometer 2 (AMSR2) data. First, to improve feature extraction, we design a novel high-resolution Transformer model with both global and local modules that can better discern the subtle differences in sea ice patterns. Second, to improve uncertainty quantification, we design a Bayesian extension of the proposed Transformer model, treating its parameters as random variables to more effectively capture uncertainties. Third, 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 tested on Pan-Arctic datasets from September 2021, and the results demonstrate that the proposed model can achieve both high-resolution SIC maps and robust uncertainty maps compared to other uncertainty quantification approaches.

Bayesian Transformer for Pan-Arctic Sea Ice Concentration Mapping and Uncertainty Estimation using Sentinel-1, RCM, and AMSR2 Data

TL;DR

The paper introduces a high-resolution Bayesian Transformer tailored for Pan-Arctic sea ice concentration (SIC) mapping with explicit uncertainty quantification. It combines global (GloFormer) and local (LoFormer) attention modules to better capture subtle ice features and uses decision-level fusion of Sentinel-1, RCM, and AMSR2 data to address heterogeneity. By treating model parameters as random variables and applying Bayes by Backpropagation, the approach yields calibrated uncertainty estimates and improves SIC accuracy relative to non-Bayesian and other UQ methods. The work demonstrates robust uncertainty maps, particularly in the marginal ice zone, and provides a practical framework for reliable Pan-Arctic SIC mapping with potential for extension to other multi-sensor, multi-modal EO tasks.

Abstract

Although high-resolution mapping of Pan-Arctic sea ice with reliable corresponding uncertainty is essential for operational sea ice concentration (SIC) charting, it is a difficult task due to some key challenges, e.g., the subtle nature of ice signature features, model uncertainty, and data heterogeneity. This letter presents a novel Bayesian Transformer approach for Pan-Arctic SIC mapping and uncertainty quantification using Sentinel-1, RADARSAT Constellation Mission (RCM), and Advanced Microwave Scanning Radiometer 2 (AMSR2) data. First, to improve feature extraction, we design a novel high-resolution Transformer model with both global and local modules that can better discern the subtle differences in sea ice patterns. Second, to improve uncertainty quantification, we design a Bayesian extension of the proposed Transformer model, treating its parameters as random variables to more effectively capture uncertainties. Third, 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 tested on Pan-Arctic datasets from September 2021, and the results demonstrate that the proposed model can achieve both high-resolution SIC maps and robust uncertainty maps compared to other uncertainty quantification approaches.

Paper Structure

This paper contains 8 sections, 8 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Dataset Overview
  4. High-Resolution Transformer Architecture
  5. Bayesian Transformer Model
  6. Results and Analysis
  7. Conclusion
  8. Acknowledgement

Figures (4)

  • Figure 1: Proposed high-resolution Transformer architecture, $f^{\omega}(.)$, with an among token Transformer block for modeling global context (GloFormer) and a within token Transformer block for modeling local detail (LoFormer).
  • Figure 2: Pan-Arctic SIC (top) and corresponding uncertainty (bottom) from the High-Resolution Bayesian Transformer for September 4th, 2021. SIC and uncertainty are derived from (a) Sentinel-1 (200m resolution), (b) RCM (200m resolution), and (c) AMSR2 (5km resolution), and (d) is the final mosaic of all data sources with Sentinel-1 layered on top, followed by RCM and AMSR2.
  • Figure 3: Local visual comparison of SIC derived from Sentinel-1 on September 4th, 2021, where (a) NASA Team SIC, (b) Sentinel-1 HV Imagery, (c) Sentinel-1 HH Imagery, (d) Deterministic Transformer SIC, (e) Mean Monte Carlo Dropout SIC, (f) Mean Epoch Ensemble SIC, and (g) Mean Bayesian Transformer SIC (our approach).
  • Figure 4: Local visual comparison of SIC derived from RCM on September 4th, 2021, where (a) NASA Team SIC, (b) RCM HV Imagery, (c) RCM HH Imagery, (d) Deterministic Transformer SIC, (e) Mean Monte Carlo Dropout SIC, (f) Mean Epoch Ensemble SIC, and (g) Mean Bayesian Transformer SIC (our approach).