Learning Subglacial Bed Topography from Sparse Radar with Physics-Guided Residuals

2025-11-18

AI Summary

This study addresses the challenge of low-accuracy subglacial bed topography reconstruction from sparse and spatially heterogeneous radar observations. We propose a physics-guided residual learning framework structured as “residual-over-prior,” integrating multi-scale physical constraints—including mass conservation, streamwise total variation, Laplacian damping, ice-thickness non-negativity, and progressive prior consistency. The architecture employs a DeepLabV3+ decoder with a ResNet-50 encoder and is trained end-to-end using confidence-weighted Huber loss. Our key contribution lies in jointly optimizing physical plausibility and spatial consistency, thereby significantly enhancing generalization in data-sparse regions. Evaluated on two Greenland subregions, the method outperforms U-Net, Attention U-Net, FPN, and conventional CNNs in reconstruction accuracy. The resulting topographies exhibit high structural fidelity and explicit physical interpretability, enabling direct application in ice-sheet modeling and cartographic workflows.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Machine Learning: Multimodal Learning
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Accurate subglacial bed topography is essential for ice sheet modeling, yet radar observations are sparse and uneven. We propose a physics-guided residual learning framework that predicts bed thickness residuals over a BedMachine prior and reconstructs bed from the observed surface. A DeepLabV3+ decoder over a standard encoder (e.g.,ResNet-50) is trained with lightweight physics and data terms: multi-scale mass conservation, flow-aligned total variation, Laplacian damping, non-negativity of thickness, a ramped prior-consistency term, and a masked Huber fit to radar picks modulated by a confidence map. To measure real-world generalization, we adopt leakage-safe blockwise hold-outs (vertical/horizontal) with safety buffers and report metrics only on held-out cores. Across two Greenland sub-regions, our approach achieves strong test-core accuracy and high structural fidelity, outperforming U-Net, Attention U-Net, FPN, and a plain CNN. The residual-over-prior design, combined with physics, yields spatially coherent, physically plausible beds suitable for operational mapping under domain shift.

Problem

• Predicts subglacial bed topography from sparse radar observations
• Improves accuracy using physics-guided residual learning framework
• Enhances generalization for ice sheet modeling under domain shift

Innovation

• Physics-guided residual learning predicts bed thickness
• DeepLabV3+ decoder with ResNet-50 encoder architecture
• Multi-scale physics constraints ensure physically plausible reconstruction