AI and the World's Hard Problems · Physics · arXiv · 2026-10-23
OTProf: estimating high-resolution profiles of optical turbulence ($C_n^2$) from reanalysis using deep learning
OTProf is a deep-learning method that estimates high-resolution vertical profiles of optical turbulence (Cn²), which are important for ground-based astronomy and free-space optical communication, from coarse ERA5 reanalysis data. Trained and tested in the Netherlands against WRF-based Cn² data, it reproduced the vertical structure more accurately than the Hufnagel-Valley baseline and estimated the Fried parameter and scintillation index more accurately. Performance dropped in the Canary Islands, where the terrain differs markedly, and bias correction using a few local simulations helped only partly, so domain adaptation needs more work.