Target-Oriented Time-Lapse Full Waveform Inversion Using Marchenko Redatumed Data

Abstract

Time-lapse full waveform inversion (TLFWI) is a powerful tool for monitoring spatial and temporal variations in the physical properties of Earth, offering valuable insights into the dynamic evolution of reservoirs. However, the influence of complex overburden (such as scattering responses and multiple reflections) and limited illumination of the target area can reduce the spatial resolution of interest targets. To alleviate this problem, we propose a target-oriented inversion method. It uses Marchenko redatuming to mimic the acquisition of sources and receivers close to the reservoir, after which the redatumed datasets are utilized for the joint time-lapse inversion. One benefit of this method is that it only requires a kinematically accurate overburden velocity model. A secondary advantage is that TLFWI becomes computationally more efficient because of the reduced dimension of model parameters. This method has two key points: First, the normalized zero-lag cross-correlation objective function is employed in TLFWI to mitigate issues related to amplitude distortions in the redatumed data, particularly at far offsets; second, we introduce a joint inversion strategy and the total variation regularization to stabilize the inverted solution. Three numerical experiments involving a Chevron 2014 model, a Marmousi model and a complex salt dome model demonstrate that the proposed method can achieve higher-resolution inversion results compared to traditional methods while reducing the computational cost of standard TLFWI. Specifically, the method demonstrates excellent performance in the normal overburden model, whereas in the complex salt dome model, despite the challenges posed by geological complexity that limit its application, it still achieves meaningful monitoring results. Overall, it provides a valuable and effective tool for time-lapse monitoring.