Scale property Monte Carlo driven inversion of surface wave data

Analysis of surface wave data can be made using probabilistic approaches, e.g. Monte Carlo methods that employ a random or pseudorandom generator. A method like this is required to efficiently avoid local minima, evaluate nonuniqueness in the solution and estimating the values and uncertainties of the model parameters. The pure Monte Carlo method applied to surface wave inversion becomes efficient with the introduction of a "smart sampling" rule which exploits the scale property (scaling of the modal solution with the wavelength) of the solution. Introducing this property in the Monte Carlo inversion focuses the scan of model space on high probability density zones. Each model is scaled before evaluating the misfit in order to bring the theoretical dispersion curve obtained by forward algorithm closer to the experimental. An applicative example is presented to support our hypothesis. The main advantage of the proposed approach, based on scale property of dispersion curves, is the possibility of using a pure Monte Carlo method with a limited number of simulations. This leads to a solution which accounts for data uncertainties evidencing equivalent final models. The possible bias of the result from a wrong choice of the initial model is significantly reduced