Direct Waveform Inversion for Irregularly Layered Media

Abstract Full waveform inversion is based on iterative perturbation approaches and often encounters convergence problems when the global initial model is far from the true model. As an alternative solution, our direct waveform inversion (DWI) can accurately invert for the acoustic properties of layered media without requiring a global initial model and furthermore DWI convergence is unconditional. However, previous studies of DWI focused only on inverting 1D horizontally layered media using incident plane waves. Its performance in higher dimensions is yet to be evaluated. Now, we have developed a new DWI scheme designed to invert for 2D irregularly multilayered acoustic media using point-source data. Leveraging the advantages of boundary integral equations in handling irregular layer boundaries, we derived a new wavefield representation and implemented it in our 2D DWI. We incorporated a localized imaging and inversion process for determining the geometries of the irregular interfaces. Similar to our previous 1D DWI, our new 2D DWI explicitly enforces the time–space causality property of the wavefields, retaining the key advantage of not requiring a global initial model and avoiding local minimum and cycle-skipping pitfalls. We tested our 2D DWI on two irregular layered models using explosive point sources. The results confirm that our 2D DWI can effectively and accurately invert for P-wave velocities as well as irregular subsurface geometries.