Summary
We selected seismograms from a 3D seismic survey acquired at the Waggoner Ranch, Texas to characterize the distribution of porosities in each lithofacie. Porosity images were obtained using cokriging and the facies using the Self-Organizing Map (SOM) technique. The reservoir geology is a sand and shale sequence with small amounts of limestone as markers. The SOM uses reflection attributes that are able to capture the three geological units. The seismic lines are first converted to impedance and then with the help of the porosity and impedance logs are transformed to porosity images using geostatistics. We compute vertical and horizontal variograms as well as cross variograms of the soft (impedance) and hard data (porosity logs). The SOM image correlates with the lithological units and the seismic reflection lines as well. The SOM image identifies the sedimentary structure of the shale and sands, and the porosity image provides the distribution of porosity in the specific sand and shale units of the reservoir.
Introduction
In this paper, we use well log data and seismic lines from the Waggoner oil reservoir located in northeast Texas. Production of this field is primarily from shallower Permian horizons, where thin sandstone and limestone formations represent alternating, fairly rapid transgressive and regressive marine sequences. The reservoir is a sand-shale sequence that is characterized at the borehole and seismic scales. The well log data and the lithofacies are described in detail by Parra et al., 2006, and selected seismic refection lines calibrated with well logs are given in Parra et al., 2015.
In this paper, we process the seismic lines to capture the lithofacies from reflection data. We use the self-organizing map introduced by Kohonen (1982), (2001), (2013) that is one of the most successful neural network algorithms applied to unsupervised classification (Roy et al. 2010, 2012; Singh et al. 2004; Taner et al., 2001). The goal is to discover some underlying structure of the data by a dimensionality reduction method. SOM clusters data such that of the statistical relationships between multidimensional data are converted into a much lower dimensional latent space that preserves the geometrical relationships among the data points.
The objective is to find the three geological units of the sand-shale sequence with the SOM method from the seismic data, then correlate these units with the lithological column of an existing well intersecting the seismic cross-line 1176. The final analysis will integrate the SOM image with the porosity image, which is derived via cokriging using seismic data and well logs. This allows the interpreter to identify the high and low porosity zones within the sand and shale lithofacies. This interpretational lithofacies finder combined with porosity maps, helps us characterize and evaluate the reservoir. The selection of SOM over other traditional clustering techniques is based on preliminary experiments, which showed that the results obtained with SOM show more realistic classifications that relate better with the seismic reflection attributes and layered structure of the reservoir.
