Context and objectives
Accurately assessing the effects of climate change on Antarctica is crucial for understanding global sea-level rise and the health of the polar climate system. However, current Regional Climate Models (RCMs) operate at coarse spatial resolutions and fail to capture the fine-scale processes that drive surface mass balance (SMB) and meltwater dynamics. At the same time, in-situ observations are sparse, and existing satellite products lack the physical constraints required for high-fidelity climate downscaling.
The ClimaVision project aims to overcome these limitations by developing a novel, physically informed downscaling framework. This framework integrates multi-source satellite observations with deep learning–based super-resolution techniques to enhance the spatial resolution and physical consistency of climate variables over Antarctica. The project’s three primary objectives are:
• To develop deep learning models that incorporate physical constraints (e.g., elevation, conservation laws) for climate downscaling.
• To ensure consistency between high- and low-resolution representations of climate variables.
• To apply the resulting downscaling framework to improve estimates of SMB and surface melt across Antarctic regions.
Project outcome
Expected scientific results
ClimaVision is expected to yield the following key scientific contributions:
• A novel framework for physically informed deep learning downscaling in climate science.
• High-resolution, physically consistent datasets of SMB and surface melt across Antarctica.
• Improved understanding of the spatial variability and drivers of SMB and melt processes.
• Open-source publications and codebases enabling replication and future research by the community.
• Quantitative assessments of model performance, transferability, and uncertainty across different RCM inputs.
Expected products and services
• Open-source Python packages and model architectures for physically constrained climate downscaling.
• High-resolution downscaled datasets for SMB and surface melt (multiple RCM sources, multiple years).
• Evaluation reports comparing model outputs across regions and timescales.
• Public-facing documentation, tutorials, and training materials via GitHub and the project website.
• Peer-reviewed scientific publications detailing methodological advances and climate insights.
| Project leader(s): | KULeuven - Department of Earth and Environmental Sciences (EES) | |
| Location: |
Continent:
|
|
| Website: | https://earthmapps.io/projects/ClimaVision.html | |