Structural interpretation - 3-D Seismic Interpretation - page 81

Once the fault polygons have been added to the map, the horizon surface can easily
be gridded and contoured; if the horizon has been picked on a dense grid from the
3-D survey then a gridding step is not required, though it may still be useful as a way
to apply smoothing and fill in small gaps in the interpretation. Unlike the case of a
sparse 2-D dataset, where control of the gridding is a critical part of the interpretation,
it is usually sufficient to use quite simple algorithms in view of the density of the
picked horizon data. There are often problems, however, caused by minor inaccuracies
in picking. Ideally a picked horizon should extend up to the picked fault plane and
terminate exactly at the fault cut, on both the upthrown and downthrown sides, and
the fault polygon on the map should match exactly the horizon cutout on the section.
In practice, it would be extremely time-consuming to ensure total precision in picking
horizons and faults on every line, and it is common to find some horizon values that
belong to the upthrown side actually plotting on the downthrown side of the mapped
polygon, and vice versa. This then causes anomalous grid and contour values near the
fault, which is sometimes only aesthetically displeasing but which sometimes could
lead to error in estimating the throw on a fault that might be critical to trap integrity. In
such cases, the anomalous values have to be removed by detailed manual editing of the
picks or of the grid values; alternatively, software is available to check the consistency
of horizon and fault picks and to repair minor inconsistencies automatically.

3.2.3 Autotrackers

The snag with the procedure outlined in the previous section is that it is very time-
consuming to interpret horizons manually across a large 3-D dataset, even with the aid
of the automatic book-keeping provided by the interpretation software. A target horizon
with complicated structuration may have to be tackled in this way, but in many cases
the interpreter wants also to map simpler horizons; these may for instance be above the
target zone and needed for depth conversion. In these cases, a semi-automated picking
method is able to give good results in a fraction of the time needed for a full manual
interpretation (Dorn, 1998 ) .
The general idea is illustrated in fig. 3.16 . Suppose we are trying to track a horizon that
is a strong positive loop, and have identified it on one trace of a line, e.g. at a well. Then
we might look for the pick on the next trace along the line as follows. Knowing that
the horizon is at time T ms on the initial trace, we can look at a window from T − δ
to T + δ ms on the next trace, choosing δ so that the correct pick should be within this
window; this depends on the local dip of the reflector, and a reasonable value can easily
be found by inspection of a section display. We assume for the moment that no faults
are present. It is then possible to get the software to make a pick automatically within
the window; the simplest approach would be to take the time of the largest positive
amplitude. The process can then be extended to a third trace, using a window centred
on the pick found on the second trace. In this way, the horizon could be picked along an