Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Data-Driven Uncertainty-Aware Forecasting of Sea Ice Conditions in the Gulf of Ob Based on Satellite Radar Imagery

Stefan Maria Ailuro, Anna Nedorubova, Timofey Grigoryev, Evgeny Burnaev, Vladimir Vanovskiy

TL;DR

This work presents a novel data-driven approach for sea ice condition forecasting in the Gulf of Ob, leveraging sequences of radar images from Sentinel-1, weather observations, and GLORYS forecasts to enhance forecast accuracy and model robustness, crucial for reliable operations in volatile Arctic environments.

Abstract

The increase in Arctic marine activity due to rapid warming and significant sea ice loss necessitates highly reliable, short-term sea ice forecasts to ensure maritime safety and operational efficiency. In this work, we present a novel data-driven approach for sea ice condition forecasting in the Gulf of Ob, leveraging sequences of radar images from Sentinel-1, weather observations, and GLORYS forecasts. Our approach integrates advanced video prediction models, originally developed for vision tasks, with domain-specific data preprocessing and augmentation techniques tailored to the unique challenges of Arctic sea ice dynamics. Central to our methodology is the use of uncertainty quantification to assess the reliability of predictions, ensuring robust decision-making in safety-critical applications. Furthermore, we propose a confidence-based model mixture mechanism that enhances forecast accuracy and model robustness, crucial for reliable operations in volatile Arctic environments. Our results demonstrate substantial improvements over baseline approaches, underscoring the importance of uncertainty quantification and specialized data handling for effective and safe operations and reliable forecasting.

Data-Driven Uncertainty-Aware Forecasting of Sea Ice Conditions in the Gulf of Ob Based on Satellite Radar Imagery

TL;DR

This work presents a novel data-driven approach for sea ice condition forecasting in the Gulf of Ob, leveraging sequences of radar images from Sentinel-1, weather observations, and GLORYS forecasts to enhance forecast accuracy and model robustness, crucial for reliable operations in volatile Arctic environments.

Abstract

The increase in Arctic marine activity due to rapid warming and significant sea ice loss necessitates highly reliable, short-term sea ice forecasts to ensure maritime safety and operational efficiency. In this work, we present a novel data-driven approach for sea ice condition forecasting in the Gulf of Ob, leveraging sequences of radar images from Sentinel-1, weather observations, and GLORYS forecasts. Our approach integrates advanced video prediction models, originally developed for vision tasks, with domain-specific data preprocessing and augmentation techniques tailored to the unique challenges of Arctic sea ice dynamics. Central to our methodology is the use of uncertainty quantification to assess the reliability of predictions, ensuring robust decision-making in safety-critical applications. Furthermore, we propose a confidence-based model mixture mechanism that enhances forecast accuracy and model robustness, crucial for reliable operations in volatile Arctic environments. Our results demonstrate substantial improvements over baseline approaches, underscoring the importance of uncertainty quantification and specialized data handling for effective and safe operations and reliable forecasting.

Paper Structure

This paper contains 27 sections, 4 equations, 9 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Data
  3. Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. Funding
  8. Author contributions statement
  9. Competing interests
  10. Availability
  11. Appendix

Figures (9)

  • Figure 1: The map (a) and coordinates (b) of meteorological stations used, along with the target region outlined in red. Available sea surface area is 120,559 km$^2$. The area of interest is 51,262 km$^2$. (c) Examples of colorized SAR images
  • Figure 2: Mean cell-wise sea ice concentration correlation between several data sources: Sentinel-1 SAR sentinel1sar, GLORYS operative and reanalysis GLORYS, and AMSR modis+amsr
  • Figure 4: (a) Frequency of missing values in SAR imagery smoothed with a month-wide rolling window. (b-c) Distribution of distances between consecutive missing values across all subsets, calculated frame-wise between frames with any valid data, and pixel-wise at fixed locations.
  • Figure 5: Examples of images before filtration (the first line) and after (the second line). Percentage of noise is a ratio of detected imagery artifacts intensity, percentage of sea-ice is the ratio of residual intensity of variance from noise- and ice-free frame $c_0$, that is presented in the upper left corner of the figure. Images are colorized according to sentinelhub.
  • Figure 6: The RMSE percentage improvement over persistence baseline for each month (a), and for each lead time in days (b), over the test subset. The colormap is shared.
  • ...and 4 more figures