A B S T R A C TWith the increasing use of permanently installed seismic installations, many of the issues in time-lapse seismic caused by the lack of repeatability can be reduced. However, a number of parameters still influence the degree of reliability of 4D seismic data. In this paper, the specific impact of seawater velocity variations on time-lapse repeatability is investigated in a synthetic study. A zero-lag time-lapse seabed experiment with no change in the subsurface but with velocity changes in the water column is simulated. The velocity model in the water column is constant for the baseline survey while the model for the repeat survey is heterogeneous, designed from sea salinity and temperature measurements in the West of Shetlands. The difference section shows up to 80% of residual amplitude, which highlights the poor repeatability. A new dynamic correction which removes the effect of seawater velocity variations specifically for permanent installations is developed. When applied to the synthetic data, it reduces the difference residual amplitude to about 3%. This technique shows substantial improvement in repeatability beyond conventional time-lapse cross-equalization.
I N T R O D U C T I O NOne of the main issues encountered during 4D seismic experiments is the degree of repeatability between successive surveys. It often determines the degree of confidence in the interpretation of the time-lapse signature. Metrics have been developed to quantify that degree of confidence. The most used are the NRMS (which measures the normalized root mean square difference residual amplitude) and the predictability (which measures the correlation between two traces). Their equations are given in Appendix A. Over the past years, factors affecting repeatability have been studied extensively and solutions or guidelines have been laid out in order to minimize their effects. From the acquisition point of view, dedicated 4D surveys have proved to be more repeatable than the use of legacy data sets. Swanston et al. (2003) compared the repeatabilities obtained for legacy and 4D dedicated sur- * Now at: veys. In their examples, dedicated 4D experiments tended to achieve an NRMS value of around 30% to 40% while it is more likely to exceed 60% for 4D experiments using data not purposely acquired. Such experiments include the Alba Field (towed streamers and seabed receivers) where the NRMS is 89% (Hanson et al. 2003). Similarly, dedicated 4D processing helps to reduce the discrepancies between data sets. Smith et al. (2001) showed that careful 4D reprocessing reduced the difference residual amplitude from 74.6% to 41.3%. New 4D specific processing techniques have also been developed to remove non-production-related difference signals. Cross-equalization (Ross, Cunningham and Weber 1996), warping (Rickett and Lumley 2001), singular-value decomposition (Reid et al. 2005) and geostatistics (Lecerf and Coleou 2002) are some of the methods that are now used to improve the quality of the 4D signature. However, some parameters, often ...