Understanding the Seismic Expression of Complex Turbidite Reservoirs Through Synthetic Seismic Forward Modeling: 1D-Convolutional Versus 3D-Modeling Approaches
Tomas van Hoek, Boudewijn Salomons, 2006. "Understanding the Seismic Expression of Complex Turbidite Reservoirs Through Synthetic Seismic Forward Modeling: 1D-Convolutional Versus 3D-Modeling Approaches", Reservoir Characterization: Integrating Technology and Business Practices, Roger M. Slatt, Norman c. Rosen, Michael Bowman, John Castagna, Timothy Good, Robert Loucks, Rebecca Latimer, Mark Scheihing, Hu Smith
Download citation file:
Synthetic seismic forward modeling has been used for many years to gain a better understanding of the seismic expression of subsurface geology and to ensure consistency between quantitative models and available data. With improvement in static model-building capabilities, increased computing power, and the ongoing need to optimally use seismic information to condition exploration and production models, synthetic seismic modeling approaches have evolved towards 3D modeling of realistic and complex input models.
The 1D-convolutional method of generating 3D synthetic seismic models is computationally very fast and convenient to apply. However, influences of spatially varying lateral resolution, acquisition, processing, and overburden effects on the resulting seismic image are fully or partially neglected. Given the simplifying assumptions of the 1D-convolutional modeling method, it is important to understand the degree to which results are representative of the actual seismic expression of the subsurface geology. It is desirable to know under which circumstances the 1D-convolutional approach can be assumed to be a sufficiently close approximation and under which conditions the more sophisticated 3D techniques are required.
As a contribution to addressing this question, two suites of 3D synthetic seismic models were constructed from high resolution, realistic, and representative static facies models of complex turbidite reservoir architecture; one using the 1D-convolutional method and the other employing a 3D-modeling technique. The latter approach honors lateral resolution, processing, acquisition, and overburden effects. Comparison of results of the two methods suggests potential pitfalls in applying inferences from the 1D method in reservoir characterization (e.g., lithofacies distribution, net-to-gross, and connectivity).