Time-aware UNet and super-resolution deep residual networks for spatial downscaling

2025-12-15

AI Summary

Satellite-derived ozone data suffer from coarse spatial resolution (typically ≥10 km), limiting their utility in local-scale environmental analysis. To address this, we propose a lightweight dual-path downscaling framework incorporating time-aware encoding: sinusoidal and radial basis function (RBF) time encodings are respectively embedded into a U-Net backbone and a super-resolution deep residual network (SRDRN), enabling adaptive spatiotemporal feature fusion with negligible computational overhead. Evaluated on ozone downscaling over Italy, our method achieves a 2.3 dB PSNR gain, a 0.042 SSIM improvement, ~35% faster convergence, and only <1.2% increase in inference latency. This work introduces, for the first time, a dual time-encoding mechanism into ozone spatial downscaling—establishing a new paradigm for high-fidelity, computationally efficient refinement of satellite atmospheric data.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Local Search

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Satellite data of atmospheric pollutants are often available only at coarse spatial resolution, limiting their applicability in local-scale environmental analysis and decision-making. Spatial downscaling methods aim to transform the coarse satellite data into high-resolution fields. In this work, two widely used deep learning architectures, the super-resolution deep residual network (SRDRN) and the encoder-decoder-based UNet, are considered for spatial downscaling of tropospheric ozone. Both methods are extended with a lightweight temporal module, which encodes observation time using either sinusoidal or radial basis function (RBF) encoding, and fuses the temporal features with the spatial representations in the networks. The proposed time-aware extensions are evaluated against their baseline counterparts in a case study on ozone downscaling over Italy. The results suggest that, while only slightly increasing computational complexity, the temporal modules significantly improve downscaling performance and convergence speed.

Problem

• Downscaling coarse satellite data to high-resolution fields
• Improving spatial resolution for local environmental analysis
• Enhancing deep learning models with temporal encoding for accuracy

Innovation

• Time-aware UNet with temporal encoding
• Super-resolution deep residual network enhancement
• Lightweight temporal module for spatial downscaling