Revealing real-world hidden 3D atmospheric turbulence from imaging

Learning the 3D structure of atmospheric turbulence poses challenges due to randomness, complex dynamics, and spatiotemporal coupling. We propose a turbulence effects-induced probing and evolutionary framework (TEPEF) that leverages optical degradation in multi-view images to learn real-world 3D turbulence. Trained on a large-scale dataset of 131,240 sequences (2,327,540 frames) and enhanced by a semi-supervised strategy, TEPEF estimates the refractive index structure constant (C_n²), a key measure of turbulence strength. TEPEF demonstrates high accuracy in quantifying atmospheric turbulence, achieving excellent agreement between predictions and ground truth for both probing and evolution tasks and substantially outperforming existing methods. When applied to predicting atmospheric coherence length and “seeing,” TEPEF achieves results with errors below 3%, offering a precise and cost-effective approach to astronomical site selection. These results highlight the promise of deep learning from turbulence-induced imaging for advancing atmospheric turbulence research.