Exploring the Potential of Latent Embeddings for Sea Ice Characterization using ICESat-2 Data

2025-04-25

AI Summary

ICESat-2 sea ice elevation data suffer from scarce manual annotations and challenging optical co-registration. Method: We propose an unsupervised autoencoding embedding framework that integrates LSTM and CNN into a spatiotemporal autoencoder to learn low-dimensional latent representations directly from unlabeled photon-counting elevation sequences, followed by UMAP for interpretable nonlinear dimensionality reduction and visualization. Contribution/Results: This is the first work to apply unsupervised autoencoding embedding to ICESat-2 sea ice topographic representation. Without human labels, it significantly improves intra-class compactness and inter-class separability, preserves clustering structure effectively, and outperforms raw features in discriminative capability—thereby substantially reducing reliance on sparse optical co-registration annotations.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Segmentation
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web

Abstract

The Ice, Cloud, and Elevation Satellite-2 (ICESat-2) provides high-resolution measurements of sea ice height. Recent studies have developed machine learning methods on ICESat-2 data, primarily focusing on surface type classification. However, the heavy reliance on manually collected labels requires significant time and effort for supervised learning, as it involves cross-referencing track measurements with overlapping background optical imagery. Additionally, the coincidence of ICESat-2 tracks with background images is relatively rare due to the different overpass patterns and atmospheric conditions. To address these limitations, this study explores the potential of unsupervised autoencoder on unlabeled data to derive latent embeddings. We develop autoencoder models based on Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN) to reconstruct topographic sequences from ICESat-2 and derive embeddings. We then apply Uniform Manifold Approximation and Projection (UMAP) to reduce dimensions and visualize the embeddings. Our results show that embeddings from autoencoders preserve the overall structure but generate relatively more compact clusters compared to the original ICESat-2 data, indicating the potential of embeddings to lessen the number of required labels samples.

Problem

• Reducing reliance on manual labels for sea ice classification
• Exploring unsupervised autoencoder for ICESat-2 data embeddings
• Improving efficiency in sea ice characterization using latent features

Innovation

• Unsupervised autoencoder for latent embeddings
• LSTM and CNN models for sequence reconstruction
• UMAP for dimensionality reduction and visualization