From Bureau of Economic Geology, The University of Texas at Austin (
For more information, please contact the author.

American Association of Petroleum Geologists Annual Meeting, Dallas, April 18-21, 2004

Integrating Detailed Stratigraphic Architecture with 3D Seismic for High-resolution Reservoir Modeling

Hongliu Zeng, Charles Kerans, and Stephen Ruppel


Generating high-resolution reservoir models from band-limited seismic data is a major challenge. One solution is to utilize wireline logs to build a priori impedance models and conduct model-based seismic inversion for a broad-banded reservoir model. However, it is commonly difficult to map detailed stratigraphic architecture from low-resolution seismic data. Without such information, inversion results are prone to be either low resolution or erroneous when depicting reservoir architecture.

Progressive inversion helps solve these problems by allowing interpreters to build multiple initial models and to perform multiple inversions. The first initial model is a low-accuracy model based on a few prominent and reliable geologic boundaries and related seismic horizons. However, even this first inversion commonly reveals more geologic details than do original seismic data. Additional horizons are then added from interpretation of the first inversion to create a new and more accurate initial model that better fits the geologic interpretation of core, well logs, and outcrop analogs. A new inversion will provide yet more geologic details and higher resolution. This process may be repeated until inversion resolution is satisfactory for reservoir-model-building applications.

We present preliminary results on applying this methodology to a carbonate outcrop model and 3-D seismic data sets in Cogdell and Fullerton fields, all in west Texas. These examples show that after several rounds of inversion, 20- to 40-ft-thick stratigraphic and reservoir units can be resolved. This is a significant improvement over the 100-ft resolution commonly obtained from seismic amplitude data.