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

TL;DR

This paper introduces EFDiff, a diffusion-based super-resolution framework guided by Earth foundation models, significantly improving land surface temperature resolution from highly degraded thermal data.

Contribution

The paper presents a novel Earth foundation model-guided diffusion framework for super-resolution, leveraging geospatial embeddings to enhance reconstruction quality in remote sensing.

Findings

EFDiff outperforms baseline super-resolution methods on a large global benchmark.

Cross-attention conditioning by Earth foundation models is more effective than channel concatenation.

The framework is broadly applicable to remote sensing problems with pretrained geospatial representations.

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-$\epsilon$ 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.