Bayesian Selection of Atmospheric Profiles from an Ensemble Data Assimilation System using Infrasonic Observations of May 2016 Mount Etna Eruptions

Abstract

AbstractObserved infrasound arrivals are considered as a way to provide additional knowledge of the middle atmosphere for numerical weather prediction (NWP) models. To do so, a Bayesian approach, based on the discrepancies between predicted and observed infrasound arrival characteristics, is adopted for selecting the most likely atmospheric states, in terms of temperature and wind velocity profiles, among an ensemble of NWP model analyses. The predicted characteristics, in terms of trace velocity and the back azimuth angle, are computed using a three‐dimensional ray‐tracing model and atmospheric profiles extracted once a day from ensemble of analyses produced by the Météo‐France global NWP model. The performance of the method is demonstrated using a set of thousands of volcanic eruptions that correspond to an upsurge in volcanic activity of Mount Etna (Sicily) during May 2016. It is shown that the Bayesian approach allows the identification of the most likely members and hence provides additional knowledge on the atmosphere truly probed by the infrasound. When the atmospheric temperature and wind profiles are reasonably meeting the effective sound speed assumption requirements, the members retrieved are similar in terms of effective sound speed but not in terms of wind and temperature. For improving both temperature and wind velocity profile selection, additional infrasound arrival characteristics are then required (e.g., travel time, amplitude, etc.). The confidence on the retrieved profiles is also sensitive to the time of validity of the meteorological analyses and on the definition of the surface conditions involved in the estimation of the predicted trace velocities.