Abstract We present a new method for ambient noise tomography, which provides robust velocity images accompanied with meaningful estimates of local resolution and uncertainty. We use the Subtractive Optimally Localized Averages (SOLA) Backus–Gilbert inference method to determine surface wave velocities as local (unbiased) averages of the “true” model parameters. Regularization on the model itself is not needed, but a compromise is sought between the local resolution and uncertainty. Then, we construct a 2D grid which efficiently accommodates the lateral variations on resolution at period. Based on this grid, we perform a probabilistic inversion to estimate Vs ${V}{s}$ at depth, efficiently propagating the frequency‐dependent uncertainties. Finally, we demonstrate the effectiveness of this method by deriving a robust 3D Vs ${V}{s}$ model of Western Europe, using 13 years of continuous noise recordings.

Read original article

Read original article