"
ABSTRACTSubsurface models of lithology are often poorly constrained due to the lack of
dense well control. Although limited in vertical resolution, high-quality threedimensional
(3-D) seismic data usually provide valuable information regarding
the lateral variations of lithology. In this thesis, I will show how Bayesian
approach can be used to generate seismically constrained models of
lithology. Unlike cokriging-based simulation methods, this method does not
rely on a generalized linear regression model, which is inadequate when
combining discrete variables, such as lithology indicator; and continuous
variables, such as seismic attributes. This method uses a Bayesian updating
rule to construct a posterior probability distribution function of lithoclasses by
using a priori information from well data and the seismic likelihood to constrain
the 3-D geological scenarios produced by geostatistical technique, which is
then sampled sequentially at all points in space to generate a set of
realizations. The realizations define alternative, equiprobable lithologic
models. The methodology was applied to delineate productive reservoir zone
in Boonsville, Texas. To achieve better result in the Bayesian Sequential
Indicator Simulation, I used acoustic impedance obtained from a seismic
inversion process as the attribute to constrain the simulation. It is expected
that by using this attribute, the separation of the litho-class conditional
distribution will be well defined and at the same time minimizing the overlaps
between the two distributions. The lithology classification obtained from BSIS
is then integrated with the result of the seismic inversion to clearly delineate
the productive zone in the field."
