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, Gaussian Processes (GPs) and a normalizing-flow- based model (FloTS), against deterministic recurrent neural network (RNN) and long short-term memory (LSTM) baselines. All models are trained and evaluated on seeing measurements from the Maunakea summit. We assess the models using both deterministic metrics, including the root mean square error and the Pearson correlation coefficient, and probabilistic metrics that evaluate the quality and calibration of predictive uncertainty estimates. The results show that deep learning approaches outperform standard recurrent networks for short-term forecasting, with LSTM and FloTS achieving the best overall predictive accuracy. While Gaussian Processes provide a useful probabilistic baseline, their predictive distributions are limited by the Gaussian assumption. FloTS combines competitive forecasting accuracy with flexible uncertainty quantification and produces the most informative probabilistic forecasts.
Versions
Related
- SHAPE: Solar Heliospherical Analysis and Prediction of Eruptions A machine learning approach to solar flare forecasting
- Aurora Hunter: A Two-Stage Framework for Probabilistic Visibility Forecasting
- PROSWIN: Probabilistic Solar Wind Speed Forecasting Using Deep Distributional Regression From Solar Images
- Electric Potential Patterns Forecasting in the Southern Hemisphere with Deep Learning Techniques
- Seeing in the Dark: Gravitational Waves from Resonant Production of Dark Photon Dark Matter
BibTeX
@misc{medlej2026probabilistic,
title = {{Probabilistic Machine Learning for Short-Term Atmospheric Seeing Forecasting}},
author = {Mary-Joe Medlej and Rahul Srinivasan and Simon Prunet and Aziz Ziad and Christophe Giordano},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.29654},
eprint = {2610.29654},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.29654},
}
RIS
TY - UNPB
TI - Probabilistic Machine Learning for Short-Term Atmospheric Seeing Forecasting
AU - Mary-Joe Medlej
AU - Rahul Srinivasan
AU - Simon Prunet
AU - Aziz Ziad
AU - Christophe Giordano
PY - 2026
DA - 2026/10/21
PB - arXiv
DO - 10.48550/arxiv.2610.29654
UR - https://arxiv.org/abs/2610.29654
AB - 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, Gaussian Processes (GPs) and a normalizing-flow- based model (FloTS), against deterministic recurrent neural network (RNN) and long short-term memory (LSTM) baselines. All models are trained and evaluated on seeing measurements from the Maunakea summit. We assess the models using both deterministic metrics, including the root mean square error and the Pearson correlation coefficient, and probabilistic metrics that evaluate the quality and calibration of predictive uncertainty estimates. The results show that deep learning approaches outperform standard recurrent networks for short-term forecasting, with LSTM and FloTS achieving the best overall predictive accuracy. While Gaussian Processes provide a useful probabilistic baseline, their predictive distributions are limited by the Gaussian assumption. FloTS combines competitive forecasting accuracy with flexible uncertainty quantification and produces the most informative probabilistic forecasts.
ER -