Journal of Conference Abstracts

An Efficient Method for Calculating Finite-difference Seismograms After Model Alterations

Johan O. A. Robertsson & Chris H. Chapman (chchapman@slb.com)

Schlumberger Cambridge Research, High Cross, Madingley Road, Cambridge, CB3 0EL, U.K.

A wide variety of seismic modelling, processing and inversion algorithms require the calculation of the seismic response after incremental local alterations to a seismic model. Finite-difference simulations are required for the full model, even though changes to the model are often restricted to a small sub-volume. An example in exploration seismics is time-lapse experiments where changes in the model are restricted to the reservoir. More generally, we may wish to test various scenarios in order to design experiments, or to refine simulations as our knowledge about the model increases.

We describe a new method that allows full finite-difference simulations for the complete model to be corrected for local changes while only requiring calculations in the sub-volume and its neighbourhood. It involves first generating the FD synthetics over the entire model, during which the wavefield is recorded around one or several sub- volumes. Alterations to the seismic properties are then introduced within these sub-volumes. The FD response can then be efficiently updated from simulations that only include the vicinity of the sub- volumes. The method, referred to as the "FD-injection" method, thus enables significant reductions of computational cost both in terms of the number of calculations and memory for storage of material parameters and variable fields. It is ideally suited for problems that involve multiple simulations for models that differ only in a small sub-volume. The only approximation involved is the omission of second-order interactions between model changes in the sub-volumes and structure outside these volumes.

The method will be illustrated on a 2-D time-lapse seismic problem where computational savings on the order of a factor of 15 were made possible. Potential savings for 3-D are far greater.