When Earth Foundation Models Meet Diffusion: An Application to Land Surface Temperature Super-Resolution

2026-04-18

AI Summary

This study addresses the challenge of reconstructing land surface temperature (LST) at extreme spatial degradations, such as 32× downsampling, by proposing EFDiff—a novel framework that integrates the Earth foundation model Prithvi-EO-2.0 with a diffusion model. For the first time, high-resolution multispectral reflectance is leveraged via cross-attention mechanisms to provide geospatial embeddings as priors that guide thermal image super-resolution. Evaluated on a global benchmark comprising 242,416 co-registered Landsat samples, EFDiff significantly outperforms existing methods, demonstrating that cross-attention guidance surpasses conventional channel concatenation strategies by simultaneously preserving pixel-level fidelity and enhancing perceptual realism. The framework includes two variants: EFDiff-ε and EFDiff-x₀.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Land surface temperature (LST) super-resolution is important for environmental monitoring. However, it remains challenging as coarse thermal observations severely underdetermine fine-scale structure. In this paper, we propose Earth Foundation Model-guided Diffusion (EFDiff), a novel framework for super-resolution under extreme spatial degradation. EFDiff uses the Prithvi-EO-2.0 Earth foundation model to encode high-resolution multispectral reflectance into geospatial embeddings, which are injected into the denoising network via cross-attention to guide fine-scale reconstruction from highly degraded observations. We study two variants, EFDiff-$ε$ and EFDiff-$x_0$, which offer complementary trade-offs between perceptual realism and pixel-level fidelity. We evaluate EFDiff under an extreme $32\times$ scale gap using a globally diverse benchmark comprising 242,416 co-registered Landsat thermal-reflectance patches. Results show that EFDiff consistently outperforms baseline methods and that cross-attention conditioning by EFM is more effective than HLS channel concatenation. Although we present EFDiff in the context of LST super-resolution, the framework is broadly applicable to remote sensing problems in which pretrained geospatial representations can guide generative reconstruction.

Problem

• Land Surface Temperature
• Super-Resolution
• Remote Sensing
• Spatial Degradation
• Thermal Imagery

Innovation

• Earth Foundation Model
• Diffusion Model
• Super-Resolution
• Cross-Attention
• Land Surface Temperature