Bayesian inversion for hurricane parameters based on underwater ambient noise levels

A method is presented for determining the significant parameters, maximum wind speed and radius of maximum wind speed, of the surface winds associated with a hurricane. The method is based on Bayesian inversion, using Markov chain Monte Carlo sampling. Underwater acoustic measurements are used to estimate parameters in the axisymmetric Holland model for hurricane surface winds. This method is validated first using synthetic data which shows that unbiased estimates are obtained. Applying the method to field measurements taken during the passage of Typhoon Fanapi in 2010 gives an accurate estimate of the reported maximum wind speed. The model assumes a linear dependence of low-frequency underwater noise on the surface wind speed, within the range 15 m s-1-50 m s-1. The slope derived based on the previous results is 0.48 s m-1, adopted in the current study as a universal constant. The significant parameter estimations for hurricane classifications are validated by comparing estimated sound pressure levels to measurements.