Wave-equation based multi-parameter linearized inversion with joint-sparsity promotion

The successful application of linearized inversion is affected by the prohibitive size of the data, computational resources required, and how accurately the model parameters reflects the real Earth properties. The issue of data size and computational resources can be addressed by combining ideas from sparsity promoting and stochastic optimization, which can allow us to invert model perturbation with a small subset of the data, yielding a few PDE solves for the inversion. In this abstract, we are aiming at addressing the issue of accuracy of model parameters by inverting density and velocity simultaneously rather than only using velocity. As a matter of face, the effects of density and velocity variations towards the wavefield are very similar, which will cause energy leakage between density and velocity images. To overcome this issue, we proposed a incoherence enhanced method that can reduce the similarity between the effect of density and velocity. Moreover, the location of structural variations in velocity and density are often overlapped in geological setting, thus in this abstract, we also exploit this property with joint-sparsity promoting to further improve the imaging result.