Statistical learning as a new approach for optical turbulence forecasting

For the new generation of Extremely Large Telescopes the knowledge of atmospheric turbulence conditions is become an information of primary importance to design and optimize all focal instrumentation. In the same way, the forecast of these atmospheric conditions is also of interest to allow both flexible scheduling and long term site testing. Until now we have used weather forecast tools coupled with turbulence models to predict turbulence conditions. In addition, we are developing a predictive statistical learning tool, using a large atmospheric database. Since 2015, the Calern Observatory hosts the Calern Atmospheric Turbulence Station (CATS) which measures during daytime and nighttime, ground meteorological conditions, vertical profiles of the C2n and all relevant integrated parameters characterizing the optical turbulence. This large CATS database is used as input for our predictive statistical learning tool. This latter should take into account more closely the local specificities, seasonal variations and day/night transitions. The results from these turbulence predictive models and statistical learning tools are presented and discussed.