Elastic wavefield tomography with physical model constraints

Abstract

Data-domain elastic wavefield tomography is an effective method for updating multiparameter elastic models that exploits much of the information provided by observed multicomponent data. However, poor illumination of the subsurface by P- and S-waves often prevents reliable updates of the model parameters. Moreover, differences in illumination, amplitude, and wavelength between P- and S-waves can distort the intrinsic physical relationships between the reconstructed model parameters. We have developed a method for elastic isotropic wavefield tomography that explicitly constrains the relationship between P- and S-wave velocities. By incorporating a model constraint term in the objective function, we confine P- and S-velocity updates to a physical area defined by prior information, for example, by laboratory measurements or well logs. We have determined that this physical constraint yields models that are more physically plausible, compared with models obtained using only the data misfit objective function.

Normalized nonzero-lag crosscorrelation elastic full-waveform inversion

“…However, estimating elastic parameters makes the inverse problem even more ill-posed. Regularization techniques can reduce the null space of the ill-posed inverse problem (Asnaashari et al, 2013;Wang et al, 2013;Duan and Sava, 2016). With some assumptions made of the subsurface (such as smoothness or sharpness), an additional penalty or a constraint can be added to the data-fitting objective or the estimated parameters, respectively.…”

Facies-constrained FWI: Toward application to reservoir characterization

“…There are inherent trade-offs between model parameters, which cannot be resolved independently using just the data misfit (Alkhalifah and Plessix, 2014;. Incorporating facies-based and other constraints sets bounds in the model space and helps obtain more accurate and geologically plausible results (Asnaashari et al, 2013;Duan and Sava, 2016;Zabihi Naeini et al, 2017).…”