Early-arrival waveform inversion for near-surface velocity and anisotropic parameters: inversion of synthetic data

Introduction

In a separate paper (Shen, 2012), I showed that in the case of early-arrival waveform inversion, changes in vertical velocity and $\epsilon$ affect data kinematics much more than $\delta$ changes do. Since $\delta$ has a minimal effect on data kinematics, I proposed joint inversion for vertical velocity and $\epsilon$ , while holding $\delta$ fixed. For such inversion, I proposed three parametrizations for joint inversion of vertical velocity and anisotropic parameter $\epsilon$ . Naive parametrization is not suitable for joint inversion, since it results in no updates to $\epsilon$ . Velocity parametrization and logarithmic slowness parametrization both seem more suitable.

In this paper, I compare the inversion results of each parametrization using two synthetic examples, one laterally invariant vertical-gradient background model with Gaussian anomalies, and one that is part of the BP 2002 velocity model with anisotropic parameters. First I will describe the equation used and the model parametrization. Then I will compare various inversion results using the two synthetic examples.

Early-arrival waveform inversion for near-surface velocity and anisotropic parameters: inversion of synthetic data