A Workflow to Quantify Velocity Model Uncertainty

Abstract

Summary A typical depth migration velocity model building project delivers final velocity model attributes and their associated seismic products. The only quantitative measures of the reliability of these data are provided through comparison with any available auxiliary data or from analysis of volumetric residual move-out. The non-linearity inherent within the tomography used to derive the earth model gives rise to multiple realizations of a solution model, which similarly honour the constraining data and yield the same convergence criterion. Therefore in isolation such data provides little useful evidence of the reliability of any individual model. In this study we propose to use multiple realizations of the solution space in order to derive estimates of the uncertainty of an individual model. The method performs multiple random perturbations of the target model followed by tomographic inversion. A solution model population is then selected based on analysis of the volumetric residual move-out; they will comprise models exhibiting a similar level of residual move-out as the target model. This subset of realizations is then used to directly derive a model parameter variance attribute. It also provides for subsequent error envelope analysis studies at key target intervals.