Application of an Envelope Inversion and FWI Workflow to Frontier Exploration in Argentina

Spectrum’s Mark Roberts explains the application of  modified envelope inversion as “an initial step to provide a workflow suitable for executing projects in a robust manner and short turn around and illustrate the effectiveness on a dataset from offshore Argentina.” This information was presented at SEG 2018 in Anaheim, California.

Summary
FWI has increasingly become a standard part of the velocity model building process due to the ability of the two-way wave-equation to generate much higher resolution velocity models compared with ray tracing. Challenges with non-linearity in FWI require accurate starting models and low frequency data, making the technique generally unsuitable for frontier exploration. In this abstract I look at modifying the envelope inversion method as an initial step to provide a workflow suitable for executing projects in a robust manner and short turn around and illustrate the effectiveness on a dataset from offshore Argentina.

Figure 1: Envelope inversion

Introduction
Frontier exploration often involves processing lines over vast areas with varying geology and little a priori information. Often seismic data in frontier areas is targeting specific licensing rounds and can require processing be completed in a timespan of weeks. These unique situations often present additional challenges to seismic processing projects in more mature areas. The quality of the velocity model remains a primary driver of image quality for most depth-imaging projects and FWI has proven an important tool for generating high resolution models. However, FWI suffers from cycle skipping leading to requirements for low-frequencies, long offset data and/or an accurate starting model (Sirgue, 2006).

Envelope Inversion (EI) (Bozdag et al, 2011) holds much promise and has been extensively studied on synthetic data. In this abstract I address a number of issues to improve the performance on real data and illustrate the resulting method on a real dataset from offshore Argentina. Wu and Chen (2017), observe the challenges with using the Born approximation for computing the gradient when there is a large shift between the observed and synthetic datasets. To mitigate the deficiency in the Born gradient computation, in this abstract I implement a Rytov-style gradient computation loosely following the work of Luo and Schuster (1991). Another challenge with applying EI to real data is the sensitivity to dynamic data: for computational efficiency an acoustic constant density finite-difference approach is used and in order to mitigate the impact of poro-elastic effects on the dynamic information a trace normalization strategy has been implemented.

Click to read the EXTENDED ABSTRACT here:
Application of an Envelope Inversion and FWI workflow to frontier exploration in Argentina