Although L2 matched-filters can remove bandwidth and phase differences between time-lapse surveys, the filter amplitudes may be biased by the presence of random noise in the surveys. If a further step of amplitude equalization is not taken after matched-filtering, then artifacts may appear in difference images that are not related to genuine production-related changes.
Matched-filtering aside, the aim of amplitude balancing is to equalize the amplitude of the non-reservoir related seismic signal between surveys, so that the difference image consists of only production-related changes and a small amount of background noise. Matching energy between surveys will only achieve this goal if the signal-to-noise levels are approximately equal in both cases, or the noise levels are very low.
We have demonstrated these amplitude problems on a realistic synthetic example, and shown that by calculating the signal-to-noise levels before cross-equalization, the correct scaling factor can be applied to suppress the amplitude bias.