American Oil and Gas Reporter - September 2015 - 92

SpecialReport: Reservoir Characterization
FIGURE 6
Broadband Seismic and Broadband/Conventional Inversions
Overlaid with P-Impedance Well Log Data
high

Broadband Seismic

Broadband Inversion

low

sparsity assumptions. If the algorithm
uses the input model up to a specified
maximum frequency and a gap is left between it and the lowest frequency of the
seismic data, some algorithms still are
able to correctly predict the model.
Figure 5 shows a sparse inversion of
the layered sandstone impedance model
in Figure 1. These results are from using
only the low frequencies up to 2 Hz,
while the seismic data extend down to 6
Hz. The sparsity increases from red to
blue to green. Even though there is a gap
of 4 Hz, it is closed by the inversion algorithm. Of course, this method only
works for sparse models and will fail for
applications such as gradational transitions
between layers.
The reason for bringing this up is that
the power of broadband over conventional
data sometimes is inferred using sparse
synthetic models. Depending on the inversion algorithm, conclusions from sparse
models need to be scrutinized carefully.
Even the example shown in Figure 1 arguably should not be used to prove the
value of broadband data, unless the inversion algorithm is stated clearly. In
fact, the inversion on synthetic models
carried out with a sparsity-assumption
algorithm would produce similar results
for both 2-6 Hz and 1-2 Hz frequency
gap cases.

Inversion QC
There are numerous examples of the
amazing quality of broadband seismic
data compared with conventional seismic
data. The inversion results are no less astonishing, even if one ignores the cases
where the conventional inversion may
be biased by the low-frequency model.
Broadband inversions give a better match

92 THE AMERICAN OIL & GAS REPORTER

Conventional Inversion

high

low

to blind control wells. Geobodies are
more clearly defined, and lithology classification can be performed with less uncertainty.
However, away from well control, the
inversion QCs are always interpretive.
An interpreter can find himself arguing
whether a detail is real or false: Is the
one inversion result better or worse? It
was suggested that the right-hand panel
of Figure 2 was the preferred result, but
how can that be proven fully unless we
have a blind well going through that feature? What is the value of a clearer image
or reduced side lobes? In the end, only
the interpreter and the oil and gas company
can put value on the features seen in
seismic inversions. The inversions need
to go through the whole cycle of interpretation, drilling and production to prove
their value.
A commonly used QC for inversion
results is where the inversions are overlaid
by a control well. Here, too, the interpreter
must be careful to judge what he sees, as
most inversions use the well log information as a starting model in the inversion
and the low frequencies from the wells
to fill the gap with the seismic data. So,
in fact, a conventional inversion generally
will look better when compared at the
well control, since it is using more of the
well information in the inverted result.
An example is shown in Figure 6.
The same high cut-filtered (100 Hz) well
log is overlain on the broadband and
conventional inversion results. The 2-6
Hz frequency band comes from seismic
in the broadband and from the well log
in the conventional inversion. The uncertainty in the inversion in these frequencies would make the casual user incorrectly prefer the conventional inversion

away from well control. In fact, the well
control in broadband inversions gives a
better estimate of the uncertainty for
reservoir characterization. A more appropriate QC to compare broadband and
conventional inversions is the match to
blind control wells.
Although the results of broadband
seismic inversion can be fully proven
only after the whole cycle of interpretation,
drilling and production, it delivers real
value to reservoir characterization. Seismic
wavelet estimation is of critical importance
to broadband and all inversions. Variability
in time, space and offset are all important
and deserve attention.
An abundance of cases favor broadband
data over conventional data, and this is
also true for seismic inversion. Favoring
one inversion result over the other by investigating their match at well control
can be biased by the extent to which the
control well is used as the starting model
for inversion.
r
PETER R. MESDAG is technical
product manager at CGG GeoSoftware.
He joined Jason when it was founded
in 1985 and has served in a variety of
technical and managerial positions
over the years, including as geoscience
adviser and product champion. Fugro-Jason was acquired by CGG 2013.
Mesdag holds a Ph.D. in physics from
Delft University of Technology.
MENNE D. SCHAKEL is a geoscientist at CGG GeoSoftware, specializing in seismic inversion and reservoir
characterization. Before joining CGG,
he served as a geophysicist at BP, a
post-doctoral researcher in geophysics
and petrophysics at Delft University
of Technology, and a geoscientist at
Argo Geological Consultants. Schakel
holds a B.S. in earth science from
Utrecht University, an M.S. in geophysics from Universiteit Utrecht, and
a Ph.D. in petrophysics and geophysics
from Delft University.

Coming In October

With multiple large-scale facilities
soon scheduled to commence operations, reports examine the implications
of U.S. LNG exports for domestic producers.

American Oil and Gas Reporter - September 2015

Table of Contents for the Digital Edition of American Oil and Gas Reporter - September 2015