Cyclic 1D matching of time-lapse seismic data sets: A case study of the Norne Field

Discussion

The sequential cyclic 1D search method provides a robust alternative to a full 3D search for cross-correlation peaks. By iteratively estimating displacements in a cyclic manner using variable correlation gates and lags, it is possible to obtain high-resolution estimates of displacement components (Figures 1 to 3). These displacement components contain different information about changes within and around the reservoir and may be interpreted qualitatively (Figures 4 to 6 ). Analysis of the estimated displacements show that in general, velocity within the reservoir decreses with time (Figure 7). A velocity decrease of up to 5 percent is observed between 2001 and 2006 (Figure 7(c)).

There are inherent errors caused by considering only vertical displacements in time-lapse seismic data sets (Figure 8(a)). In the current example, this approximation leads to time-shift overestimation below the reservoir, and time-shift underestimation above the reservoir. Errors in time-shift estimates, which are more pronounced around dipping reflectors, will map into errors in velocity change and vertical-strain estimates (Figures 8(b) and 8(c)).

Filter parameters derived from the proposed method (Figure 9) enable computation of the optimal filters at each trace location. Before computing time-lapse images, the monitor data sets are shifted into alignment with the baseline by interpolating them with the estimated displacements. After data alignment (Figures 10(b) and 11(b)), optimal match-filtering removes residual artifacts, thereby improving the time-lapse data-quality (Figures 10(c) and 11(c)). Finally, especially at steeply dipping flanks, considering only vertical displacements can lead to erroneous time-lapse amplitude changes (Figure 12).

Cyclic 1D matching of time-lapse seismic data sets: A case study of the Norne Field