Introduction

Seismic reservoir monitoring is a new technology that uses multiple time-lapse 3D seismic surveys to directly image fluid movements, pressure/temperature fronts or other effects of production in the subsurface Lumley (1995). Unfortunately, different generations of 3D seismic can exhibit seismic differences unrelated to reservoir production, caused by non-repeatability of seismic acquisition and processing artifacts. The aim of cross-equalization is to remove processing and acquisition differences between time-lapse seismic surveys, so comparison between them can be interpreted in terms of genuine fluid-related changes.

Cross-equalization of post-stack seismic datasets typically includes the following generic elements:

1.

Survey realignment to a common grid.

2.

Amplitude balancing to scale the data to the same amplitude/energy level.

3.

Bandwidth equalization to compensate for different source wavelets.

4.

Phase matching to correct residual wavelet phase differences.

5.

Spatial and temporal re-registration to correct the effects of geometry errors, differential statics, or different velocity functions used for NMO and migration.

In a previous SEP report, Rickett (1997) discussed the use of L₂ matched-filtering to simultaneously correct bandwidth, phase and static shifts between surveys. Unfortunately, the amplitudes of the L₂ filters are biased by the presence of uncorrelated noise differences between the surveys, and the amplitude of coherent events may not be equalized correctly.

Similarly, any scheme that is based on equalizing the energy in the two surveys, implicitly assumes that their random noise levels are equal, which may or may not be the case.

We present an approach that scales the amplitudes in two data sets based on their relative signal-to-noise levels, and test the approach on a synthetic example that has many features observed on 4D field datasets.