Lightning the Night with Generative Artificial Intelligence

2025-06-25

AI Summary

To address the fundamental limitation of conventional remote sensing—disrupted continuous observation at night due to insufficient visible-light illumination—this paper proposes RefDiff, the first method leveraging generative diffusion models for nighttime visible-band surface reflectance retrieval. RefDiff fuses multi-spectral thermal infrared brightness temperature data from the Advanced Geosynchronous Radiation Imager (AGRI) aboard FY-4B satellite and incorporates an ensemble averaging strategy to enhance prediction stability. It further enables pixel-level uncertainty quantification. Validation across 0.47 μm, 0.65 μm, and 0.825 μm bands yields structural similarity (SSIM) scores of 0.90. Comparative evaluation against VIIRS nighttime products demonstrates accuracy approaching daytime observational fidelity, particularly outperforming conventional approaches in complex and optically thick cloud regimes. RefDiff thus establishes a novel paradigm for all-weather meteorological monitoring and numerical weather prediction.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The visible light reflectance data from geostationary satellites is crucial for meteorological observations and plays an important role in weather monitoring and forecasting. However, due to the lack of visible light at night, it is impossible to conduct continuous all-day weather observations using visible light reflectance data. This study pioneers the use of generative diffusion models to address this limitation. Based on the multi-band thermal infrared brightness temperature data from the Advanced Geostationary Radiation Imager (AGRI) onboard the Fengyun-4B (FY4B) geostationary satellite, we developed a high-precision visible light reflectance retrieval model, called Reflectance Diffusion (RefDiff), which enables 0.47~μmathrm{m}, 0.65~μmathrm{m}, and 0.825~μmathrm{m} bands visible light reflectance retrieval at night. Compared to the classical models, RefDiff not only significantly improves accuracy through ensemble averaging but also provides uncertainty estimation. Specifically, the SSIM index of RefDiff can reach 0.90, with particularly significant improvements in areas with complex cloud structures and thick clouds. The model's nighttime retrieval capability was validated using VIIRS nighttime product, demonstrating comparable performance to its daytime counterpart. In summary, this research has made substantial progress in the ability to retrieve visible light reflectance at night, with the potential to expand the application of nighttime visible light data.

Problem

• Retrieve nighttime visible light reflectance using thermal infrared data
• Overcome lack of visible light for continuous weather monitoring
• Improve accuracy and uncertainty in nighttime reflectance estimation

Innovation

• Generative diffusion models for night reflectance
• Multi-band thermal infrared data conversion
• Ensemble averaging improves accuracy significantly