Effect of regularization in the migration of time-lapse data

With 4D seismic studies the aim is to look at signal changes in the reservoir which can improve our understanding of the dynamic reservoir system and enable us to optimize production. For a successful study, noise must be minimized as the magnitude of the signal related to the 4D effect is generally weak and the same order of magnitude as the noise. There are many sources of non-repeatability in marine 4D seismic studies from differences in acquisition parameters, differences in environmental factors, and also from data processing effects. When we compare time-lapse datasets, this nonrepeatable noise will remain and impede the interpretation of the data. Pre-stack Kirchhoff Time Migration (PreSTM) is a processing technique utilized in seismic exploration and reservoir monitoring. Such a migration scheme is attractive as it allows the implementation of spatially and temporally varying velocity fields to migrate the data. It is only suitable for datasets with smooth velocity contrasts as any sharp velocity changes (e.g., shallow surface heterogeneities) would leave residual energy visible as a migration smile. Even with smooth velocity variations, another form of migration noise comes from the sub-optimal interference of the migration operator. As the PreSTM implementation uses a symmetrical operator (not the case for Pre-stack Depth Migration), the best possible constructive and destructive interference requires square bins with the midpoint position of the traces at the bin centre. In addition, the traces in a single offset volume should be mono-azimuth (or contain an equally distributed azimuth content for each bin) and mono-offset. When this is not the case residual migration noise will remain as an imprint of the acquisition pattern. In a 4D study where each dataset has slightly different acquisition parameters, each vintage will also have its own pattern of migration noise. This noise is uncorrelated between vintages and appears as energy on the difference section. The use of weighting schemes inside the PreSTM algorithm can help reduce noise by normalizing the migration output. However, regularizing the data on input optimizes the constructive and destructive interference, this handles amplitudes on seismic events as well as attenuating migration smiles. In this paper we consider the level of migration noise from sub-optimal recording positions in a constant velocity medium for a Kirchhoff PreSTM algorithm. In addition, we outline some processing techniques to reduce the level of migration noise.