Reflection Tomography by Depth Warping: a Case Study Across the Java Trench

Accurate subsurface velocity models are crucial for geological interpretations based on seismic depth images. Seismic reflection tomography is an effective iterative method to update and refine a preliminary velocity model for depth imaging. Based on residual move-out analysis of reflectors in common image point gathers, an update of the velocity is estimated by a ray-based tomography. To stabilize the tomography, several preconditioning strategies exist. Most critical is the estimation of the depth error to account for the residual move-out of the reflector in the common image point gathers. Because the depth errors for many closely spaced image gathers must be picked, manual picking is extremely time-consuming, human biased, and not reproducible. Data-driven picking algorithms based on coherence or semblance analysis are widely used for hyperbolic or linear events. However, for complex-shaped depth events, purely data-driven picking is difficult. To overcome this, the warping method named non-rigid matching is used to estimate a depth error displacement field. Warping is used, for example, to merge photographic images or to match two seismic images from time-lapse data. By matching a common image point gather against its duplicate that has been shifted by one offset position, a locally smooth-shaped displacement field is calculated for each data sample by gather matching. Depending on the complexity of the subsurface, sample tracking through the displacement field along predefined horizons or on a simple regular grid yields discrete depth error values for the tomography. The application to a multi-channel seismic line across the Sunda subduction zone offshore Lombok island, Indonesia, illustrates the approach and documents the advantages of the method to estimate a detailed velocity structure in a complex tectonic regime. By incorporating the warping scheme into the reflection tomography, we demonstrate an increase in the velocity resolution and precision by improving the data-driven accuracy of depth error picks with arbitrary shapes. This approach offers the possibility to use the full capacities of tomography and further leads to more accurate interpretations of complex geological structures.