Joint wave-equation inversion of time-lapse seismic data

Background

There is a wide range of published work on the most important considerations for time-lapse seismic monitoring. For example, Batzle and Wang (1992) outline important rock and fluid relationships; Lumley (1995), Rickett and Lumley (2001), Calvert (2005), and Johnston (2005) discuss important processing and practical applications; and Lefeuvre et al. (2003), Whitcombe et al. (2004), and Zou et al. (2006) showed successful case studies. Ayeni and Biondi (2008) discuss additional considerations and previous work related to seismic monitoring of hydrocarbon reservoirs.

Nemeth et al. (1999), Kuhl and Sacchi (2001), Clapp (2005), and Valenciano (2008) have shown that linear least-squares wave-equation migration of seismic data improves structural and amplitude information. We demonstrate that an extension of least-squares migration to the time-lapse imaging can improve time-lapse amplitude information, especially if all available data are jointly inverted. Previous authors have discussed joint-inversion applications, including impedance inversion (Sarkar et al., 2003), ray-tomography (Ajo-Franklin et al., 2005) and wave-equation velocity analysis (Albertin et al., 2006). Lumley et al. (2003) show that improvements can be made to time-lapse processing through simultaneous processing. Dynamic imaging strategies that utilize aspects of spatio-temporal regularization have also been discussed in other scientific disciplines (Schmitt and Louis, 2002; Zhang et al., 2005; Kindermann and Leitao, 2007; Schmitt et al., 2002).

A joint wave-equation inversion formulation of the time-lapse imaging problem has the advantage that the attenuation of image differences is based on the physics of wave propagation, making it less susceptible to removal of true time-lapse changes than conventional methods. The method proposed by Ajo-Franklin et al. (2005) for tomographic inversion can be directly extended to wave-equation inversion, and it actually forms a first step in the RJMI formulation. Such direct extension to wave-equation migration is too expensive, requiring at least one set of migration and modeling per survey per iteration. In most practical inversion problems, parameter selection requires that the inversion procedure be carried out more than once. By pre-computing the Hessian operators, we are able to test different regularization schemes and parameters for the inversion at several orders of magnitude cheaper than directly solving the least-squares migration problem. In addition, we avoid the use of matching filters which can have unpredictable effects on time-lapse changes within the reservoir (Lumley et al., 2003).

Joint wave-equation inversion of time-lapse seismic data