--> ABSTRACT: Bayesian Inversion Technique to Identify Spatial Continuity in Rock Properties Using Level Set Methods. Real Case Study on Horn Mountain Field, Deepwater Gulf of Mexico, by Dadi, Sireesh K.; #90142 (2012)
[First Hit]

Datapages, Inc.Print this page

Bayesian Inversion Technique to Identify Spatial Continuity in Rock Properties Using Level Set Methods. Real Case Study on Horn Mountain Field, Deepwater Gulf of Mexico

Dadi, Sireesh K.*1
(1) Geology and Geophysics, Texas A&M University, College Station, TX.

The flexibility introduced by bayesian techniques to update the geologic shapes based on hierarchical data is gaining increasing attention. The results of an investigation to use Bayesian inversion techniques to identify the spatial continuity in rock properties are presented. A new workflow is presented to integrate the well logs and seismic data to determine the sub seismic facies distribution. The resultant facies distribution includes both deterministic and probabilistic estimation. The most certain deterministic facies can be freezed as hard data and the probabilistic facies distribution is used as soft constrain during depositional facies simulation obtained using techniques like Multi Point Geostatistics. Facies clusters which can be classified with a low classification error from Previous HitelasticNext Hit properties constitute the deterministic facies and a probabilistic approach (Bayesian Naïve Bayes) is followed for clusters with high classification error.

A multi scale multi attribute (MS-MA) analysis is used to deal with the scale difference between well logs, seimic and the reservoir model and identify sub seismic lithology. A novel approach is introduced to identify attributes which increase value to the classification problem. The facies from the core interval are extended to the reservoir interval using neural network techniques and the resulting facies distribution is used as training set to determine facies in rest of the wells without cores. An attempt as been made to explain the pitfalls of using inverse Previous HitimpedanceNext Hit cubes with a high correlation coefficient between well log Previous HitimpedanceNext Hit and seismic Previous HitimpedanceNext Hit. Histogram equalization, usually performed in image processing is implemented to correct the derived seismic inverse Previous HitimpedanceTop.
 

 

AAPG Search and Discovery Article #90142 © 2012 AAPG Annual Convention and Exhibition, April 22-25, 2012, Long Beach, California