Wang et al. (2026) HiT_DS: A Modular and Physics-Informed Hierarchical Transformer Framework for Spatial Downscaling of Sea Surface Temperature and Height

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-01-15
• Authors: Min Wang, Weixuan Liu, Rong Chu, Shouxian Zhu, Guanghong Liao
• DOI: 10.3390/rs18020292

Research Groups

Not specified in the abstract.

Short Summary

This paper introduces HiT_DS, a modular hierarchical Transformer framework, for high-resolution downscaling of Sea Surface Temperature (SST) and Sea Surface Height (SSH) fields, demonstrating improved reconstruction accuracy and enhanced physical fidelity across distinct oceanic regions.

Objective

• To develop and validate HiT_DS, a modular hierarchical Transformer framework, for high-resolution downscaling of Sea Surface Temperature (SST) and Sea Surface Height (SSH) fields to overcome limitations of low spatial resolution in fine-scale oceanographic analyses.

Study Configuration

• Spatial Scale: Two dynamically distinct oceanic regions.
• Temporal Scale: Not specified in the abstract.

Methodology and Data

• Models used: HiT_DS (a modular hierarchical Transformer framework) which integrates:
  • Enhanced Dual Feature Extraction (E-DFE) using depth-wise separable convolutions.
  • Gradient-Aware Attention (GA).
  • Physics-Informed Loss Functions.
• Data sources: Satellite observations (Sea Surface Temperature and Sea Surface Height data).

Main Results

• HiT_DS achieves improved reconstruction accuracy for downscaled SST and SSH fields.
• HiT_DS demonstrates enhanced physical fidelity in the reconstructed fields.
• The framework allows for selective module combinations tailored to regional dynamical conditions.

Contributions

• Introduction of HiT_DS, a novel modular hierarchical Transformer framework for high-resolution oceanographic data downscaling.
• Integration of specific modules (E-DFE, GA, Physics-Informed Loss Functions) to address multiscale feature representation, emphasize high-gradient structures, and ensure physical consistency.
• Provision of an effective and extensible approach for downscaling Sea Surface Temperature and Sea Surface Height data.

Funding

Not specified in the abstract.

Citation

@article{Wang2026HiTDS,
  author = {Wang, Min and Liu, Weixuan and Chu, Rong and Wang, Xidong and Zhu, Shouxian and Liao, Guanghong},
  title = {HiT_DS: A Modular and Physics-Informed Hierarchical Transformer Framework for Spatial Downscaling of Sea Surface Temperature and Height},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18020292},
  url = {https://doi.org/10.3390/rs18020292}
}