One of the
best ways to tie seismic data back to a "ground truth" is through
comparison with a sonic log. This comparison is the link between
seismic travel time and depth from the well log.
Although
this link may seem obvious, there are influences in both data types
that may make the tie difficult to make.
This first
part of a two-part discussion serves as a guide for the geologist
or geophysicist to use when addressing issues with a problematic
well to seismic tie in compacted rock environments.
Overview
Let's briefly
examine the differences in the way seismic data and sonic logs measure
"the same thing."
First,
seismic velocities are deduced statistically to provide the best
stack of the reflectors in the data. Stacking is done to collapse
a volume of measurements into a single vertical reflectivity profile.
This may involve the sampling of hundreds of thousands of cubic
feet of rock for any one stacked trace.
Due to
the large amount of data required to produce any single seismic
trace, the statistics are quite robust.
A sonic
log, as opposed to seismic, measures velocity more directly. The
actual borehole measurement made is interval transit time (reciprocal
velocity). All sonic log measurement methods sample a volume very
near the well bore, over a short vertical interval and amount to
sampling perhaps a few thousand cubic feet of rock for an entire
well.
Assuming
that the well bore is in good condition and that there are no other
known problems with the logging environment, the sonic tool is capable
of recording a very accurate interval transit time profile with
depth.
Clearly,
each of these measurement techniques has sampled very different
volumes of rock in order to determine velocity and reflectivity
at the same physical location. Therefore, we should not necessarily
expect there to be a 1:1 correspondence between all reflectors seen
in these two data sets.
Because
the well to seismic ties are done mostly after final seismic processing,
we will assume that the seismic data are of high quality with no
significant AVO effects, and that the velocity profile associated
with each trace cannot easily be improved. However, there are several
items that need to be addressed with the sonic log velocity profile
before we can expect a good tie.
What we
do not want to do is to stretch and squeeze the synthetic seismogram
to force a match with the seismic, as this will litter the sonic
log with unreasonable velocity artifacts.
Instead,
we need to deterministically edit and calibrate the sonic log, by
comparing the sonic log to other wireline data as well as the seismic
data.
Sonic
Log Problems
It is important
to note that using a sonic log to tie to the seismic data is a very
sensitive numerical operation.
Because
we wish to know the cumulative time from the surface down to any
reflector, we need to sum the sonic log in time. By summing, we
greatly exaggerate any systematic problems with the sonic log.
With the
exception of noise spikes, all of the problems with sonic log data
discussed below make the transit time too slow. When combined and
summed, these errors can render a sonic log useless.
Raw sonic
log problems include:
- Cycle skips and noise.
- Short logging runs,
or gaps in sonic log coverage.
- Relative pressure
differences between the drilling fluid and the confining stress
of the rocks around the wellbore.
- Shale alteration
(principally clay hydration from the drilling fluid).
Specifics
First,
in order to be useful, a sonic log must represent actual rock velocities.
Spike noise and cycle skips do not represent true rock measurements,
and therefore must be removed from the sonic log. Spike noise can
be easily removed by "de-spiking."
Cycle skips
occur when the sonic tool records an arrival that is not correct
(typically one "cycle" in the wave train late). The most common
cause of cycle skipping is badly washed out zones. When they represent
a significant problem, an intelligent data replacement scheme is
required.
Figure
1 illustrates a sonic log where the shales are badly washed
out, causing frequent cycle skips and some spike noise.
Gaps in
sonic log coverage need to be handled smoothly. Often, there are
some log data in this gap, just not sonic data. If we can model
a pseudo sonic from another curve and then replace the missing sonic
data, we will have a well-behaved synthetic seismogram.
If we must
model large vertical intervals without real sonic data, we also
need to be able to accurately estimate the low frequency component
(burial trend) of the earth's velocity profile.
Figure
2 compares actual raw sonic data with a pseudo sonic modeled
from the deep resistivity data in an interval where the borehole
conditions are not conducive to recording a good sonic log. This
pseudo sonic can now be used to replace bad sonic log data or to
fill in gaps in sonic coverage where resistivity data exist.
Shale alteration
is a problem where the in-situ shales are desiccated. During the
process of drilling, these dry shales are brought into contact with
the drilling fluid, which can cause swelling and fracturing of the
shales, as well as chemical alteration of the constituent clays.
Figure
3 shows the relationship between interval transit time and deep
conductivity. This parabolic trend is used to estimate the magnitude
of shale alteration.
Relative
pressure differences between the drilling fluid and the confining
stress of the rocks around the wellbore will have an effect on the
sonic log. Since different logging runs typically use different
mud systems, separate sonic log runs will likely need unique velocity
calibrations to match the seismic data.
Figure
4 illustrates the difference in the corrections required for
different logging runs.
Conclusions
If some or
all of the problems described above are present in a well, the likelihood
of achieving a quality tie to the seismic data is slim. However,
if these issues are addressed in a deterministic way and calibrated
to the seismic data, high quality informative ties can be made.
High quality
ties can be used for many purposes, including phase determination,
relative wavelet extractions, seismic inversions, effective stress
calculations, etc.
Next
month we'll address solutions to these common sonic log problems
with real life data comparing synthetic seismograms from a raw sonic
log that exhibits all of the problems listed above to the same log
after all of the problems have been corrected.