Probabilistic Machine Learning for Short-Term Atmospheric Seeing Forecasting

Abstract

Accurate short-term forecasting of atmospheric seeing is of significant interest for astronomical observations and free-space optical communication systems, where rapid variations in optical turbulence can affect system performance and operational planning. In this work, we benchmark machine learning models for predicting atmospheric seeing up to two hours in advance, with a particular focus on probabilistic forecasting and uncertainty quantification. We compare two probabilistic approaches, Gau…