IMMI's performance with different seismic acquisition parameters and random noise
Sergio J. Romahn, Kristopher A. H. Innanen
IMMI stands for iterative modelling, migration and inversion. It proposes to incorporate standard processing techniques into the process of full waveform inversion (FWI). Following IMMI’s philosophy, we use a phase shift plus interpolation (PSPI) migration with a deconvolution imaging condition to obtain the gradient, and well velocity to scale the gradient into a velocity perturbation. The above contrasts with the use of a two-way wave migration method (such as reverse time migration RTM), and the use of an approximation of the inverse Hessian matrix or a line search to ﬁnd the scale, as is done in standard FWI. We show the suitability of estimating the subsurface velocity model by applying IMMI’s approach using a synthetic example. The results conﬁrms that the gradient obtained with PSPI provides an adequate direction to minimize the objective function, and that well calibration produces an efﬁcient scale to convert the gradient into a velocity perturbation. We evaluated the performance of the inversion when the maximum offset and the source interval are changed with and without the presence of random noise. Generally speaking, larger offsets and higher shot density generate better results, specially in the presence of noise. Higher folds, produced by large offsets and small source interval, improve the inversion result because the gradient is obtained by stacking the migrated data residuals.