Spatio-temporal Nowcasting of Atmospheric Optical Turbulence for LEO Satellite-to-Ground Laser Communication Networking

Satellite-to-ground laser communication (SGLC) is an emerging candidate for future high-speed feeder links in mega Low Earth Orbit (LEO) constellation networks, which is constrained by the highly dynamic and non-stationary nature of atmospheric optical turbulence during rapid transit windows. To address the critical necessity for network scheduling, this study captures transient atmospheric states by deriving high-resolution Fried parameter ($r_0$) fields, rather than relying on static atmospheric characterization. We propose the Diffusion Schrödinger Bridge Graph Transformer (DSBGT) for minute-scale spatiotemporal nowcasting. The DSBGT architecture employs a Graph Wavelet Network-based Transformer (GWNformer) backbone to resolve complex spatiotemporal dependencies via dynamic graphs and parallel attention mechanisms. To tackle the non-stationarity of optical turbulence, a Diffusion Schrödinger Bridge is incorporated to establish probabilistic optimal transport mappings from current features to future $r_0$ distributions. Experimental results reveal that the DSBGT model achieves a mean absolute error (MAE) of less than 0.24 cm across all four seasons, which represents an accuracy improvement over existing models. These findings highlight the model's value for enabling proactive operational link scheduling and dynamic resource optimization in future LEO-SGLC networks.