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

Apparent displacement vectors by cyclic 1D search

Components of the apparent displacement vectors between time-lapse seismic data sets can be efficiently computed using a cyclic 1D search approach (Hale, 2009). This method involves a sequential cyclic 1D search for cross-correlation peaks in different directions. This removes the need for a computationally expensive 3D search for cross-correlation peaks or the inaccurate assumption that displacements are only in the vertical direction. To improve the resolution of displacement estimates in all directions, the input data are first band-passed to remove very low frequencies. This preprocessing step serves a similar purpose as the prediction-error filtering step of Hale (2009).

This implementation utilizes correlation gates and lags of varying sizes and an acceptance criterion based on the relative values of the cross-correlation peaks. To first capture the low-frequency, high-amplitude displacements, this iterative procedure starts with large correlation gates and lags. The correlation gates and lags are then decreased as a function of iteration. Displacement estimates are filtered based on the displacement size and cross-correlation peak. For each iteration, the results (shifted image(s) and accumulated displacements) from one direction (e.g. vertical) are the inputs to the next direction (e.g. in-line). For each direction, shifts are accumulated after interpolating the image and displacement components from the previous step. Algoirthm 3 summarizes this procedure.

The estimated displacements are used to align the baseline and monitor data sets prior to time-lapse image computation. They are also useful for estimating velocity and geomechanical changes.

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