Fast Fourier Convolutional GAN for 30 m Clear-Sky Land Surface Temperature Gap-Free Reconstruction

2026-07-22

AI Summary

This study addresses the challenge of extensive data gaps in high-resolution (30 m) satellite-derived land surface temperature (LST) caused by cloud cover. To this end, the authors propose a multimodal Fast Fourier Convolution Generative Adversarial Network (FFC-GAN), which, for the first time, incorporates fast Fourier convolution into LST reconstruction by synergistically fusing optical and synthetic aperture radar (SAR) data. Relying solely on near-globally available auxiliary inputs, the method enables all-weather, seamless LST reconstruction. By leveraging global receptive fields in the frequency domain to effectively capture long-range dependencies, the model maintains robust performance even under severe data loss scenarios with over 70% cloud coverage. The reconstructed LST achieves consistently low error across all quantiles, with an interquartile range of scene-averaged RMSE between 0.8 and 1.8 K, significantly enhancing accuracy and robustness in large missing regions.

Abstract

Satellite-derived Land Surface Temperature (LST) provides spatially comprehensive data that ground stations cannot match. However, its utility is frequently limited by severe data gaps due to the presence of clouds. As LST is essential for understanding land-atmosphere interactions, numerous methods have been proposed to address this challenge. Yet, the development of a scalable and adaptable pipeline for generating gap-free LST datasets and reconstructing cloud-contaminated pixels remains challenging. Moreover, the reconstruction of extensive missing regions in fine-spatial-resolution observations is particularly difficult. To address this challenge, we propose a Multimodal Fast Fourier Convolutional GAN for reconstructing cloud-contaminated pixels in fine-resolution (30 m) Landsat imagery to generate gap-free clear-sky LST products. The method leverages Fast Fourier Convolution to enable a global receptive field across the image, and is guided by a stack of data consisting of satellite observations and Synthetic Aperture Radar (SAR) data. Across all LST quantiles, the interquartile range of scene-averaged RMSE (computed over reconstructed pixels) is consistently between 0.8 K and 1.8 K. The proposed approach enables the recovery of extensive missing regions, including scenes with more than 70% cloud-induced gaps, while relying on auxiliary data that are readily available at a near-global scale.

Problem

• Land Surface Temperature
• cloud gap filling
• high-resolution reconstruction
• missing data recovery
• satellite remote sensing

Innovation

• Fast Fourier Convolution
• Generative Adversarial Network
• Land Surface Temperature
• Gap-filling
• Multimodal Fusion