--> --> Abstract: Formation Evaluation of Barnett Shale by Kohonen Self Organizing Maps – An Example from North East Fort Worth Basin, by Atish Roy, Roderick Perez, and Kurt J. Marfurt; #90124 (2011)

Datapages, Inc.Print this page

Making the Next Giant Leap in Geosciences
April 10-13, 2011, Houston, Texas, USA

Formation Evaluation of Barnett Shale by Kohonen Self Organizing Maps – An Example from North East Fort Worth Basin

Atish Roy1; Roderick Perez1; Kurt J. Marfurt1

(1) Geology and Geophysics, University of Oklahoma, Norman, OK.

Automatic seismic facies analysis aims to classify similar seismic traces based on amplitude, phase, frequency and other seismic attributes. This research shows how this technique helps in formation evaluation in the Barnett Shale.

Recent studies show that the Barnett Shale is not a thick laterally homogeneous massive column of rock. Wireline logs reveal different parasequences which can be associated with the depositional environment of the shale. More importantly, the same shale can be classified into several rock types based in their geomechanical behavior. This work will help to visualize the variation in shale and possible relationship between these rock types and their seismic expression.

There are many different supervised and unsupervised clustering algorithms. In supervised training the clusters are pre-defined and patterns are assigned to it in subsequent trainings. Unsupervised clustering is data driven without any a priori information. The self-organizing map (SOM) is one of the most effective unsupervised pattern recognition techniques, and it can be used for the automatic identification of seismic facies.

For the present analysis we have considered seismic data within the Barnett Shale from north east Fort Worth Basin. The Upper Barnett and the Lower Barnett shale have been analyzed separately. This technique then involves in over-defining the number of initial clusters from the input dataset; training them according Kohonen SOM neighborhood training rule and later mapping these clusters against the continuous 2D Hue Saturation Value (HSV) colorbar. The colored output map of seismic facies is then interpreted in conjunction with well and post stack inversion results from the same zone.

We observe that the different colors (or clusters) in the output seismic facies map represent shale variation within the survey. The different seismic facies distribution of the Upper and the Lower Barnett shale is also visible from their seismic facies maps. We correlate output seismic facies maps to well logs and post stack inversion results to estimate geomechanical rock types within the Barnett Shale.