Bayesian Uncertainty Estimation in the Presence of Tomographic Model Error
A. Vlassopoulou, R. Felicio, C. Hagen, I. Jones, M. Ackers and S. Schjelderup
Event name: 81st EAGE Conference and Exhibition 2019
Session: Poster: Velocity Model Estimation A
Publication date: 03 June 2019
Info: Extended abstract, PDF ( 664.98Kb )
Price: € 20
Whether or not we build a parameter field model, or deliver a subsurface image, our industry has been sadly lacking in attempting to assign ‘error bars’ to any of the products created. Given that we can never obtain a “correct” model based on measured data, we need to assess how suitable the derived approximate model or resultant image, is. It transpires that this is an extremely difficult task to undertake in a quantitative manner. There are certain minimum acceptance criteria, which tell us that at least the derived model explains the observed data, namely, flat image gathers following migration with the obtained model, which also match all available well data (at least to within some specified acceptance threshold), but these criteria do not tell us how good the model or image is. Here we adopt a Bayesian analysis of tomographic model error so as to quantify image positioning uncertainty, but more specifically, in this work we consider the effect of the quality of the initial model on the final uncertainty estimation, demonstrating quantitatively how prior model uncertainty affects final posterior image positioning uncertainty estimates.