Estimation of primaries by sparse inversion with scattering-based multiple predictions for data with large gaps

We propose to solve the Estimation of Primaries by Sparse Inversion problem without any explicit data reconstruction, from a seismic record with missing near-offsets and large holes in the acquisition grid which does not produce aliasing but otherwise causes large errors in the multiple prediction. Exclusion of the unknown data as an inversion variable from the process is desirable, since it sidesteps possible issues arising from fitting observations that are also unknowns in the same inversion problem. Instead, we simulate their multiple contributions by augmenting the forward prediction model for the total wavefield with a scattering series that mimics the action of the free surface reflector within the confines of the unobserved trace locations. Each term in this scattering series simply involves convolution of the predicted wavefield once more with the current estimated surface-free Green's function at these unobserved locations. We investigate the necessary modifications to the primary estimation algorithm to account for the resulting nonlinearity in the modeling operator, and also demonstrate that just a few scattering terms are enough to satisfactorily mitigate the effects of near-offset data gaps during the inversion process. Numerical experiments on synthetic data show that the final derived method can significantly outperform explicit data reconstruction for large near-offset gaps. This is achieved with a similar computational cost and better memory efficiency compared to explicit data reconstruction. We also show on real data that our scheme outperforms pre-interpolation of the near-offset gap.