Recent years have seen remarkable technological advances in seismic data acquisition and processing. It is now commonplace for marine vessels to acquire a terrabyte of 3D seismic data per day and for some of the larger data processing centers to achieve daily throughput in excess of 10 terrabytes of seismic data. These exponential increases in available data represent huge data management and data interpretation challenges to our industry.

There are clear similarities between the seismic exploration industry and the Internet in terms of the volume of information that is available for analysis; it therefore makes perfect sense to deploy data mining tools and methodologies developed for other industries to address the needs of the oil and gas exploration business.

"Data mining is a process that uses a variety of data analysis tools to discover patterns and relationships in data that may be used to make valid predictions" (Two Crows Corp.). The data employed for gas exploration in this study are seismic attributes, and the data enrichment process employs a neural network classification scheme.

Seismic attributes are a specific class of mathematical constructs of the propagated seismic wavefield. However, many attributes are simple numerical derivatives that provide little additional discrimination, from one to the other. The primary goal of the data mining exercise is to establish appropriate seismic attributes that, in combination, afford the maximum discrimination of hydrocarbon indicators.

The phased workflow represents a progressive information enrichment process. The initial phase, conducted in the absence of well data, generates attributes appropriate for the prospects being mined. The attributes are next organized using statistical tools to select those affording the appropriate discrimination. The next phase organizes the multi-attribute response into a manageable set of discriminating classes, thereby enriching the information content concealed in the attributes. Final phases' seek to extract knowledge from the classification by calibration to known prospectivity determined from well data, yielding such properties as lithology, porosity, and fluids.

Each phase delivers a data product in its own right, so the seismic data miner can select the appropriate number of phases for the task at hand.

This seismic data mining workflow will be discussed as it applies to multiple seismic attribute volumes calculated from a 3D dataset acquired on behalf of Forest Oil Corporation over the Ibhubesi Field in the Orange River Basin, RSA.

End_of_Record - Last_Page 15---------------