Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Time Series Classification of Supraglacial Lakes Evolution over Greenland Ice Sheet

Emam Hossain, Md Osman Gani, Devon Dunmire, Aneesh Subramanian, Hammad Younas

TL;DR

This work tackles the problem of understanding supraglacial lake evolution on the Greenland Ice Sheet by framing lake types as a time-series classification task. It introduces a computationally efficient RPS-GMM pipeline that uses a single representative sample per class to distinguish refreeze, drain, and buried lakes, leveraging HV_anom from Sentinel-1 and p_water from Sentinel-2. Across 777 labeled lakes, the model achieves 85.46% accuracy with HV_anom alone and 89.70% with both features, outperforming several ML/DL baselines under limited training data. The approach offers a robust, data-efficient tool for monitoring meltwater dynamics and informing ice-sheet mass-balance assessments, with potential extensions to incorporate additional physical features.

Abstract

The Greenland Ice Sheet (GrIS) has emerged as a significant contributor to global sea level rise, primarily due to increased meltwater runoff. Supraglacial lakes, which form on the ice sheet surface during the summer months, can impact ice sheet dynamics and mass loss; thus, better understanding these lakes' seasonal evolution and dynamics is an important task. This study presents a computationally efficient time series classification approach that uses Gaussian Mixture Models (GMMs) of the Reconstructed Phase Spaces (RPSs) to identify supraglacial lakes based on their seasonal evolution: 1) those that refreeze at the end of the melt season, 2) those that drain during the melt season, and 3) those that become buried, remaining liquid insulated a few meters beneath the surface. Our approach uses time series data from the Sentinel-1 and Sentinel-2 satellites, which utilize microwave and visible radiation, respectively. Evaluated on a GrIS-wide dataset, the RPS-GMM model, trained on a single representative sample per class, achieves 85.46% accuracy with Sentinel-1 data alone and 89.70% with combined Sentinel-1 and Sentinel-2 data. This performance significantly surpasses existing machine learning and deep learning models which require a large training data. The results demonstrate the robustness of the RPS-GMM model in capturing the complex temporal dynamics of supraglacial lakes with minimal training data.

Time Series Classification of Supraglacial Lakes Evolution over Greenland Ice Sheet

TL;DR

This work tackles the problem of understanding supraglacial lake evolution on the Greenland Ice Sheet by framing lake types as a time-series classification task. It introduces a computationally efficient RPS-GMM pipeline that uses a single representative sample per class to distinguish refreeze, drain, and buried lakes, leveraging HV_anom from Sentinel-1 and p_water from Sentinel-2. Across 777 labeled lakes, the model achieves 85.46% accuracy with HV_anom alone and 89.70% with both features, outperforming several ML/DL baselines under limited training data. The approach offers a robust, data-efficient tool for monitoring meltwater dynamics and informing ice-sheet mass-balance assessments, with potential extensions to incorporate additional physical features.

Abstract

The Greenland Ice Sheet (GrIS) has emerged as a significant contributor to global sea level rise, primarily due to increased meltwater runoff. Supraglacial lakes, which form on the ice sheet surface during the summer months, can impact ice sheet dynamics and mass loss; thus, better understanding these lakes' seasonal evolution and dynamics is an important task. This study presents a computationally efficient time series classification approach that uses Gaussian Mixture Models (GMMs) of the Reconstructed Phase Spaces (RPSs) to identify supraglacial lakes based on their seasonal evolution: 1) those that refreeze at the end of the melt season, 2) those that drain during the melt season, and 3) those that become buried, remaining liquid insulated a few meters beneath the surface. Our approach uses time series data from the Sentinel-1 and Sentinel-2 satellites, which utilize microwave and visible radiation, respectively. Evaluated on a GrIS-wide dataset, the RPS-GMM model, trained on a single representative sample per class, achieves 85.46% accuracy with Sentinel-1 data alone and 89.70% with combined Sentinel-1 and Sentinel-2 data. This performance significantly surpasses existing machine learning and deep learning models which require a large training data. The results demonstrate the robustness of the RPS-GMM model in capturing the complex temporal dynamics of supraglacial lakes with minimal training data.
Paper Structure (17 sections, 10 equations, 4 figures, 1 table, 1 algorithm)

This paper contains 17 sections, 10 equations, 4 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Supraglacial Lakes
  4. Machine Learning and Deep Learning
  5. Background
  6. Reconstructed Phase Space
  7. Gaussian Mixture Models
  8. Maximum Likelihood Classifier
  9. Evolution of Supraglacial Lakes
  10. Experiments
  11. Data Collection and Preprocessing
  12. Methodology
  13. Results and Discussion
  14. RPS-GMM Performance Evaluation
  15. Comparison against Existing ML/DL Models
  16. ...and 2 more sections

Figures (4)

  • Figure 1: Cumulative melting days over the GrIS in 1981, 2001, and 2021. Source: National Snow and Ice Data Center (NSIDC) nsidc2021greenland
  • Figure 2: Comparison of backscatter signal received from within a lake (blue) and its vicinity (purple).
  • Figure 3: Evolution of different types of supraglacial lakes over time. The grey line shows the time series of backscatter differences (left y-axis) and the red dots represent the percentage of water coverage from each S2 image observation (right y-axis). Black-and-white and color images are from Sentinel-1 and Sentinel-2 satellites, respectively.
  • Figure 4: Performance of the RPS-GMM models