Π-ML: a dimensional analysis-based machine learning parameterization of optical turbulence in the atmospheric surface layer

Abstract

Turbulent fluctuations of the atmospheric refraction index, so-called optical turbulence, can significantly distort propagating laser beams. Therefore, modeling the strength of these fluctuations (Cn2) is highly relevant for the successful development and deployment of future free-space optical communication links. In this Letter, we propose a physics-informed machine learning (ML) methodology, Π-ML, based on dimensional analysis and gradient boosting to estimate Cn2. Through a systematic feature importance analysis, we identify the normalized variance of potential temperature as the dominating feature for predicting Cn2. For statistical robustness, we train an ensemble of models which yields high performance on the out-of-sample data of R² = 0.958 ± 0.001.

This dataset contains  

• The 14 trained Pi-ML ensembles
• The frozen source code used to generate the ensembles. Check Github for more up-to-date code: https://github.com/mpierzyna/piml.
• The publication describing the Pi-ML methodology