Full text loading...
-
Estimation of Spatial Uncertainties in Tomographic Images
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 77th EAGE Conference and Exhibition - Workshops, Jun 2015, Volume 2015, p.1 - 4
Abstract
Velocity models derived from the inversion of seismic (or other) data often lack a clear indication of the associated spatial uncertainties, which are as important for the interpretation as the velocity models themselves.
Most common model appraisal methods only address the amplitudes of the velocity model parameters but are not able to estimate the error distribution of the location of the retrieved anomalies very well. We also expect that the distribution of the spatial uncertainties is not isotropic, but reflects the data coverage of the model space.
The approach presented here was designed as a test to study if accurate and quantitative estimates of spatial uncertainties (e.g., in meters) could be obtained from the analysis of equivalent models (similar models with the same data fit)
Our uncertainty analysis consists of a Monte Carlo-type perturbation of the velocity model to obtain a range of equivalent models. These equivalent models, taken collectively, indicate the spatial uncertainty. Due to the otherwise extremely large computational cost, the perturbations are guided by the a posteriori covariance matrix.
This presentation focuses on the method and its requirements, the lessons learned and the consequences for the application to third-party velocity models.