A Comparison of Layer-Based and Sample-Based Inversions for Time-Lapse AVO Problems
Event name: 80th EAGE Conference and Exhibition 2018
Session: AVO Technologies
Publication date: 11 June 2018
Info: Extended abstract, PDF ( 329.92Kb )
Price: € 20
Time-lapse seismic data is a challenge to process and invert for meaningful interpretation results. A key idea is to maintain a consistent earth model through the various seismic vintages, and ensure the requisite forward models are generated from this consistent evolving earth model. Layer and contact based models are attractive in that reservoir properties can be attached to specific and controlled layers, and the saturation and geomechanical effects are easier to specify and control. Their main weakness is that significant interpreter input is required for definition of model layers, and this also results in a significant vulnerability to misspecification. By contrast, sample based models require rather less interpreter input, but are more difficult to render consistent with saturation and geomechanical laws. However, with the increasing use of discrete lithofacies-based inversion on the time or depth lattice, some of these consistency requirements can be usefully expressed using discrete fluid variables. This work sketches the theory for time-lapse inversion for both a marked-point Bayesian MCMC layer-based inversion, and also a discrete lattice/facies based inversion, with illustrative examples based on an active CO₂ storage monitoring project.