These days many unconventional plays are being challenged by low commodity prices. During market downturns, many operators look for ways to "squeeze" more information out of their seismic and well control data to reduce the risk of a dry hole or a poorly performing well. Operators in the Eagle Ford and other shale plays must account for changing stratigraphy and facies to properly locate horizontal wells for optimum perforation intervals; but to accurately understand their geologic distribution requires resolution on seismic data.

In our methodology we employ Principal Component Analysis (PCA) to identify and quantify the key attributes of any given class that are the most independent and prominent in the data set. Usually a dozen or fewer attributes rise to the top and can be included in the second step which is a Self-Organizing Map (SOM) (Roden et al., 2015). This is a learning machine data classification process wherein winning neurons reveal natural data clusters, form discernable systems tracts, and enable specific calibrations or visual correlations in this application as the seismic data quality was unusually high. Known pitfalls of coherent noise such as the presence of multiples or acquisition footprints were not detected here. A third weapon in the arsenal is an interactive 2D ColorMap that can be queried and inverted to "extract" the natural clusters or geobodies from the "forest" of 3D data.

Eagle Ford stratigraphy is often associated with thin beds and facies well below conventional seismic resolution that change both vertically and laterally. Over this seismic survey area which is nearly 15 miles in the dip direction and over 14.5 miles in the strike direction (216 sq. mi.), a high-graded group of instantaneous attributes detected 16 different winning neurons that represent the various facies present in the Eagle Ford Shale over a 14ms window (70–84 ft.). The facies architecture of the entire Eagle Ford Group (Fairbanks et al., 2016), which includes the underlying Basal Clay Shale, Eagle Ford Shale, and the overlying Eagle Ford Marl, are defined by 26 different winning neurons over 28ms (210–252 ft.). Individual facies units as thin as one sample interval of 2ms (10–12 ft.) can be resolved and are calibrated in nine vertical calibration wells by time depth corrections that were carefully computed and then applied to formation tops and edited log curves. The careful analysis of X-ray diffraction and saturation information from five cores whose variations which were incorporated in the corroboration of the SOM results were found to be both systematic and compelling.

This content is only available via PDF.
You can access this article if you purchase or spend a download.