Validation and update of 3-D velocity models by inversion of seismic and well-log data

Arnaud Berlioux

arnaud@sep.Stanford.EDU

ABSTRACT

In this paper, I propose a new procedure for estimating the validity of 3-D velocity models combining seismic data and well-log information. The method I describe gives the best least-square 3-D model for the velocity. It has even more potential for determining how accurate a velocity model is by estimating a range of possible models and giving a measure of local uncertainty about the velocity. I propose to use a 3-D prestack seismic survey as well as the well-log curves available on the site of the survey to evaluate and update a 3-D velocity model. Performing 3-D velocity analysis on the 3-D prestack seismic data, I will get an initial velocity model. From the well-log information, I plan to extract velocities at the well locations. I will then employ two different approaches to derive an accurate 3-D velocity model. The first is based on the least-square inversion of the well-log-derived velocities and uses the seismic velocities as a constraint, by applying for a conjugate gradient method. The second approach is a simulation of the velocity at each node of the 3-D model, using a sequential Gaussian simulation algorithm based on a generalized least-square inverse technique (kriging). I also intend to estimate local and global uncertainties about the velocity model I have derived.