4D elastic AVO inversion of seismic data from the Sleipner CCS project

Abstract

In seismic inversion, the goal is to reconstruct unknown subsurface model variables, such as elastic velocities, from seismic measurements. This thesis focuses on the application of a novel adjoint-state based elastic Amplitude Versus Offset (AVO) inversion technique to characterize reservoir behavior during CO2 injection into a shallow aquifer in the North Sea. The deterministic inversion methodology employed minimizes the objective function using the L-BFGS optimization method, leveraging the adjoint-state numerical technique for efficient gradient computation. The study involves creating a synthetic 4D elastic model representing pre- and post-injection conditions applying the inversion technique to real seismic data from the 1994 and 2010 vintages of the Sleipner Carbon Capture and Storage (CCS) project. The study includes a methodology for wavelet estimation and seismic-to-well tie processes, including how to deal with challenges related to noise and wavelet non-stationarity.

The results demonstrate the method’s effectiveness in analyzing the reservoir’s response to CO2 injection, with notable changes observed between 1994 and 2010. The inversion accurately characterizes variations in acoustic impedance and showed a good agreement with well log data, although uncertainties in seismic velocities and density were noted.

This study marks the first application of this inversion method on real data, indicating its potential to real-world reservoir characterization. However, results could be further refined by including improved low-frequency models and using more less approximate AVO theory, such as the Zoeppritz equations or the elastic wave equation. Future research should also explore viscoelastic AVO inversion and integrate rock physics models to enhance the understanding of reservoir behavior under CO2 injection. Extending the methodology to other CO2 storage projects could validate the technique’s generalizability and identify opportunities for optimization.