Introduction

In modern marine surveys with good signal quality, it is often easy to identify move-out curves for velocity analysis or migration velocity analysis (MVA), and automatic algorithms can pick semblance panels with little manual intervention. The main challenge becomes determining the depth migration velocity model that best fits the picked move-outs. But in other datasets, especially land datasets, it can be challenging even to hand-pick initial RMS velocity fields. A variety of practical solutions are used to overcome this challenge. One of the most common is the use of super-gathers, which are sets of adjacent common-midpoint (CMP) gathers treated as single larger CMP gathers. By including more traces in the computation of semblance panels, the signal-to-noise ratio (SNR) improves. However, the size of super-gathers must be limited in areas with steep dips because travel-time surfaces start to cut across reflections in the midpoint direction (Figure ). A typical size for super-gathers is 3x3 or 3x5, with a larger size often used in the cross-line direction, which is likely to be flatter. When dips are so steep that the data is close to being spatially aliased, even 3x3 super-gathers can be problematic. To overcome this problem, it can be helpful to compute semblance panels on individual CMPs and then to stack adjacent semblance panels. This approach is unlikely to degrade the quality of the final panel, but stacking the all-positive semblance panels does not yield the same SNR improvements that are possible by including more data in a single semblance computation. For horizon-based velocity analysis, super-gathers are sometimes locally flattened using interpreted horizons. Because of the need for human interpretation, this approach is practical only for select horizons.

This report shows that full-volume dip fields can be used to make such corrections on the full volume of velocity panels. Dip fields can typically be computed automatically on stack sections. Because these sections, migrated or unmigrated, have higher SNR than prestack data, automatic dip scans are much more reliable than automatic velocity scans. The dip field is then used as input for a dip corrected velocity analysis. An example from a 3D land dataset shows that the quality of semblance panels improves, especially in the near surface, and that larger super-gathers can be used.

The dip corrections can be viewed either as extensions to standard move-out equations or as simplifications of common reflection surface (CRS) processing, which uses a series of searches for move-out curvature, dip, and reflection curvature to generate parametric descriptions of pre-stack data (e.g., Jäger and Hubral (2001)). CRS processing scans for semblance along travel-time surfaces in super-gathers similar to the surfaces described here, but searches in super-gathers are simultaneous searches for multiple parameters and are intended to be automatic. While convenient when they work, automatic searches may become instable in noisy data. The dip corrections in this report make use of a similar representation of move-out in super-gathers, but the application is changed to allow for manual picking and easier integration into standard processing workflows.

 

supergather_moveout Figure 1 Semblance surface in a 2D super-gather. Uncorrected velocity analysis treats the top of hyperbolas as flat, but precomputed dips can be used to match the shape of reflections more accurately. This is also the surface for a 2D CRS stack. Note that stacking along this surface is equivalent to first summing along the hyperbolas (a conventional NMO stack), and then summing along the apexes (a post-stack slant-stack).