# Processing of VSP data

Series Investigations in Geophysics Öz Yilmaz http://dx.doi.org/10.1190/1.9781560801580 ISBN 978-1-56080-094-1 SEG Online Store

We now discuss the basic steps in VSP processing. After some trace editing, VSP processing starts with the separation of the downgoing waves from the upcoming waves (reflections). One separation technique is based on f − k filtering. Examine the zero-offset VSP in Figure 11.4-2a and note that the upcoming and downgoing waves have opposite dips. Because of this, each wave type should map in a different half plane in the f − k domain. Hence, downgoing waves can be suppressed with an f − k dip filter, thereby leaving only the reflection and associated multiples that constitute the upcoming waves (Figure 11.4-2b).

A VSP data set may not always have uniform receiver sampling in depth — a prerequisite for f − k filtering. The problems of edge effects and amplitude smearing often are observed on data after f − k filtering (frequency-wavenumber filtering).

An alternate approach to extracting upcoming waves is to use median filtering [1]. To start, apply first-arrival time shifts to traces in a VSP data set VSP(z, t) to flatten the downgoing waves. (First arrivals now are at t = 0.) Then apply a median filter to each horizontal array of samples, VSP(z, t = constant). Median filtering is best explained by an example. Consider the following array of numbers: (-1,2,1,2.5,1.5). Reorder its elements from small to large values: (-1,1,1.5,2,2.5). The median of the series then is the middle sample, 1.5. Median filtering rejects noise bursts and any event that is not flat. Apply the median filter to the VSP data set with flattened downgoing waves to yield the downgoing waves. This result can be subtracted from the input to obtain the upcoming waves. The last step involves unflattening the data.

The next VSP processing step involves datuming all receivers to the well head (D in Figure 11.4-1a). From Figure 11.4-1, this static correction is the same as correcting each trace by an amount that is equal to the one-way traveltime down to the corresponding receiver location. For example, trace C is corrected by an amount equal to the traveltime associated with raypath DC.

The static corrections are followed by deconvolution and filtering (Figure 11.4-2c). In principle, the deconvolution operators can be designed from downgoing or upcoming waves. These deconvolution operators then are applied to traces of the upcoming wave profile. The common practice is to use the downgoing waves to design the deconvolution operators. This is because downgoing waves on a VSP record are much stronger than upcoming waves. Hence, designing deconvolution operators from downgoing waves has an advantage in that the operators are based on stronger signal, with more emphatically represented multiples [1].

The last step involves stacking the traces in Figure 11.4-2c. Stacking normally includes a narrow corridor along the region in which upcoming and downgoing waves coincide. The resulting trace, repeated a few times, is shown in Figure 11.4-2d. To a large degree, corridor stacking prevents the multiples that do not merge with the downgoing wave path from being stacked in. The trace in Figure 11.4-2d can be considered an alternative to a zero-offset synthetic seismogram derived from the sonic log; thus, it can be compared to the CMP stack (not shown here) at the well location.

Figure 11.4-3a shows raw data from an offset VSP data set. By using the median filtering scheme described earlier, downgoing waves are extracted from the raw data (Figure 11.4-3b). The subtraction of downgoing waves from original data (followed by another median filtering for suppression of noise bursts) yields upcoming waves (Figure 11.4-3c). The upcoming waves then are deconvolved (Figure 11.4-3d) using operators designed from the downgoing waves of Figure 11.4-3b. This step normally is followed by static corrections. For nonzero-offset data, we also must correct for moveout resulting from the offset separation between the well head and the shot location. From Figure 11.4-1, this dynamic correction involves mapping the traveltime associated with raypath ABCD to 2DE. An NMO-corrected upcoming-wave profile can be compared with the surface seismic at the well location, provided the subsurface consists of horizontal layers with no lateral velocity variations.

## References

1. Hardage (1983), Hardage, B. A., 1983, Vertical seismic profiling: Geophysical Press.
2. Alam and Millahn, 1986, Alam, A. and Millahn, K., 1986, Interactive model-based VSP-CDP transform: Presented at the Symp. on Practical Aspects of Modeling in Exploration and Development, Kristiansand, Norway.