Bayesian inversion for hurricane parameters based on underwater sound levels

Hurricanes classification has been a critical task for meteorologists working to mitigate loss of life and property. In situ observations using aircrafts and remote observations based on radar and satellite are commonly used. We report on investigations of an acoustic method for hurricane classification using measured underwater sound levels. The relation between the sound level and the surface wind speed is approximately linear, within certain ranges. The slope of the linear relation is believed to be universal. The Holland model for hurricane surface winds is a commonly used axisymmetric approximation. The method investigated inverts for the parameters of the Holland model using sparse measurements of the sound levels under a hurricane. The method is validated first using the synthetic data and then applied to field measurements. It is shown that the maximum wind speed and radius of maximum wind speed in axisymmetric hurricane models are estimated accurately using Bayesian inversion.