I N T E R P R E T E R ’ S C O R N E R — C O O R D I N AT E D BY EZEQUIEL GONZALEZ

Application of elastic impedance: A case study

from Mahanadi Basin, India
Sakshi G. Saha 1 , Ashish K. Singh 1 , and Shakuntala Mangal 1

Abstract is characterized by sand/shale/clay alternations. The reservoirs

Over the last decade, seismic inversion has acquired the sta- are channelized fan lobes deposited on a proximal part of the
tus of being a routine interpretation step during exploration as basin floor with bypassed zones on slope. The two gas reservoirs
well as development of fields. For detailed reservoir character- in the study well are designated as Res-1 at a depth of 2880 m
ization, prestack inversion is desirable; however, poststack in- true vertical depth subsea (TVDSS) of early Pliocene age, and
version is a reasonably effective tool. The inversion starts with Res-2 at a depth of 3427 m TVDSS of late Miocene age. Res-1
calibration of the well with seismic data. The synthetic seismo- is 27 m thick and Res-2 is 54 m thick. Res-1 consists mainly of
gram is tied with stacked seismic data on the basic premise that sandstone with minor siltstone and clay stone units (Figure 1a).
the seismic data (full stack) mimics the zero-offset data and The sandstone is well sorted, firm, friable, moderately hard, and
therefore should match with the zero-offset synthetic seismo- cemented at places with good visible porosity. The other reser-
gram. Generally, a good match is obtained. However, during voir, Res-2, also consists of shale-sand alterations. The sand-
the course of many inversion exercises, it is observed that the stone is of two types: The first is similar to Res-1, but the second
synthetic-to-seismic match is not good in parts of the drilled type of sandstone is moderately hard to hard, cemented with
interval. There could be many reasons for this partial mismatch. abundant carbonaceous specks and calcareous cement/matrix
One of the important reasons is that the well data represent (Figure 1b). The effective porosities of Res-1 and Res-2 are 30%
one-dimensional data, whereas the seismic data are stacked over and 15%, respectively. The deeper reservoir, Res-2, is more com-
an angle range. In other words, the well data are equivalent to pact and has suffered loss of porosity as a result of cementation.
zero-offset data, whereas seismic is not. The poor calibration at The background claystone units are similar in both cases.
the well location leads to lack of confidence in inversion results. The reservoirs Res-1 and Res-2 are separated by 547 m of
The possible solution, such as simultaneous or full-waveform sediments and appear as high-amplitude anomalies at 2800 m
inversion, involves a lot of computation time and hardware re- and 3150 m, respectively, on the full-stack seismic section (Fig-
quirements. An alternative approach is introduced which uses ure 2a). The P-sonic, S-sonic, and density logs recorded in the
elastic impedance instead of normal P-impedance for synthetic well are shown in Figure 2b. Res-1 shows a clear and substantial
match and inversion of stacked data. The synthetic seismic gen-
erated from elastic impedance shows an improved match with
seismic data. The approach is quick, reliable, easy, and not heavy
on computational resources. The result is inverted volumes with
better hydrocarbon delineation and characterization ability.

Introduction
Seismic data inversion, poststack as well as prestack, has
become a basic tool in the exploration and production industry
and is used with different objectives throughout the life cycle
of a prospect, from discovery to development. The exercise, im-
mediately after drilling an exploratory well, starts with well-
to-seismic tie, which is not always a simple task. Frequently,
it is observed that the match between seismic and synthetic is
less than satisfactory, even after considering wavelet and other
log data conditioning. The seismic forward-modeling tool does
not help in understanding the problem. In such a situation,
the interpretability and reliability of inverted data for reservoir
characterization and reservoir property distribution become a
cause for concern. This necessitates finding an alternative way to
carry out reliable seismic inversion or to go for more advanced
prestack inversion.
In this case study, an exploratory well located at 350-m wa-
ter depth off the east coast of India in Mahanadi Basin has been Figure 1. Lithologs showing the nature of gas reservoirs (a) Res-1 and
analyzed. It encountered two gas reservoirs. The well lithology (b) Res-2 encountered at the well.

1
Reliance Industries Ltd. http://dx.doi.org/10.1190/tle33111268.1

1268 THE LEADING EDGE November 2014

Downloaded from http://pubs.geoscienceworld.org/tle/article-pdf/33/11/1268/3095349/gsedge_33_11_1268.pdf
by CNOOC user
 lowering of P-impedance associated with drop in both P-wave Once it has been established that the observed mismatch is
velocity and density compared to surrounding shale. On the a result of AVA behavior, there are two options for poststack
other hand, Res-2 shows practically no contrast in P-impedance. inversion:
However, a lowering of density is observed but is offset by the Option 1: Use near-angle stack or near-trace seismic data
increase in P-wave velocity while computing P- impedance. only, which may have a good match with synthetic data and
carry out inversion. However, because it is known that the near-
The problem trace data are generally noisy, it might lead to noisy inversion
The synthetic seismogram generated from the well logs output. Therefore, Option 1 is not preferable.
shows a good match at Res-1 but not at Res-2. The seismic data In this case study, the inversion output from Option 1 (Fig-
show high-amplitude events corresponding to both Res-1 and ure 5a) shows no impedance anomaly corresponding to Res-2
Res-2, but a synthetic seismogram shows a high-amplitude (which is absent in near-stack seismic data in the first place) and
event corresponding only to Res-1, and no recognizable ampli- thus misses exploiting the seismic attribute for delineating the
tude event is seen corresponding to Res-2 (Figure 3). deeper reservoir.
The considerable amplitude mismatch corresponding to Res- Option 2: Use full-angle stack and invert with poor well-
2 results in poor correlation coefficient in the vicinity of Res-2 to-seismic tie. Lowering of P-impedance is seen on the inverted
(Figure 3). Forward modeling carried out using the well data volume at both levels, i.e. Res-1 and Res-2 (Figure 5). However,
also confirms the expected poor seismic response of the lower corresponding anomalies on well data are not seen for Res-2, and
reservoir Res-2, as shown in Figure 3b. therefore, it cannot be interpreted in the absence of good cali-
The well-to-seismic mismatch might lead to a false conclu- bration. There is gross estimation error of P-impedance property
sion that the seismic data are not reliable and the strong seis- for Res-2, as seen from the well log and pseudolog comparison
mic response is the result of some unexplained noise or remnant in Figure 5b.
multiple in the data, therefore making it unsuitable for inversion. The next-best alternative is to carry out prestack inversion,
The challenge is to analyze the cause of mismatch between which will take care of the AVA effect in wavelet estimation, but
synthetic and seismic and to understand whether the seismic the process is time consuming and hardware intensive.
data can be used effectively for reliable
reservoir characterization in such cases.

Data analysis
To understand the seismic response
of reservoirs more closely, the AVA be-
havior of the gathers has been analyzed
using the Hampson-Russell amplitude
variation with offset (AVO) or ampli-
tude variation with angle (AVA) pack-
age. The AVA analysis shows a general
increase of amplitude with incidence
angle (Figure 4) for both reservoirs.
The amplitude response correspond-
ing to Res-1 is a large negative at near Figure 2. (a) Seismic and (b) log response of reservoirs Res-1 and Res-2 on full-stack seismic
angle, becoming more negative at far, section.
exhibiting the Class III nature of AVA.
Conversely, the amplitude correspond-
ing to Res-2 is a small negative (near
zero) at near angles, rapidly becoming
more negative at far, exhibiting the
Class II nature of AVA. Res-2 shows
a smaller intercept but a larger gradient
than Res-1 (Figure 4).
The AVA analysis makes it evi-
dent that the observed strong-ampli-
tude event corresponding to Res-2 in
stacked data is a result of the contribu-
tion of higher amplitude from the far
angles. The zero-offset synthetic data
are not expected to incorporate any Figure 3. (a) Well-to-seismic calibration indicating a good match at Res-1 and significant
AVA effect, and hence the mismatch mismatch at Res-2. (b) Forward modeling using blocked logs also shows weak seismic response
between the two. corresponding to Res-2.

1270 THE LEADING EDGE November 2014

Downloaded from http://pubs.geoscienceworld.org/tle/article-pdf/33/11/1268/3095349/gsedge_33_11_1268.pdf
by CNOOC user
 Figure 4. AVA response as seen from seismic gather data for (a) Res-1 and (b) Res-2. Res-2 has Class II AVA signature, whereas Res-1 exhibits
a Class III anomaly.

An alternative approach
A simple and robust approach, yet
as effective as the prestack inversion,
has been adopted here to overcome the
problem. The workflow is as follows:

• Load full-stack seismic data (know

the angle range).
• Load the well-log data after con-
ditioning (P-sonic, S-sonic, and
density).
• Generate elastic-impedance log for
same angle range as for seismic data.
• Extract wavelet using seismic data
and elastic-impedance log. Figure 5. Poststack inversion results from (a) Option 1 and (b) Option 2. The pseudo-impedance
• Perform well-to-seismic tie and re- log shows a reasonably good match with well impedance. However, no impedance anomaly is
fine the wavelet. observed corresponding to Res-2 in the first case, whereas mismatch is observed at Res-2 in the
second case.
• Perform inversion using full-stack
seismic data. offers a strong distinguishing method for the hydrocarbon
sand, especially for the Class II AVA anomalies similar to
For the analysis of this case study, full-stack seismic data with Res-2.
angle range 0–45° have been studied using Jason’s JGW software. Reflectivity versus incidence angle was modeled to generate
Elastic impedance log for the angle range 0–45° has been generated a synthetic gather (Figure 6b). The generated synthetic gather
using P-sonic, S-sonic, and density logs recorded at the well. The for different incidence angles demonstrates a Class II anomaly
elastic impedance is analogous to P-impedance at a nonzero angle at Res-2 with near-zero amplitude trough in the near angles
of incidence (Connolly, 1999). The equation to compute elastic im- becoming more prominent with increasing angle. However, at
pedance at an angle of incidence ș (Whitcombe, 2002) is: Res-1, there is minor change in amplitude.
The well-to-seismic tie of synthetic generated using the
. elastic-impedance log data (0–45°) with full-stack data (Figure
7) reveals that the match at the deeper, laminated sand-shale
gas reservoir, Res-2, has improved considerably. The match be-
Here, a = 1+ sin 2θ , b = -8Ksin 2θ , c = 1-4 Ksin 2θ and K = VS 2/ VP 2 tween synthetic and seismic for the shallow gas reservoir Res-1
Vp , Vs , and ȡ are the parameters as observed at the well and remains good.
are available from a standard log suite. VP0 , VS0 , and ȡ0 are the av- As a final step of the workflow, full-stack data are inverted
erage value of these parameters over a background shale region. using a new wavelet derived after calibration of full-stack data
Typically, for θ = 0; EI= [Vp* ρ], i.e., acoustic impedance. using the elastic-impedance log. The inverted impedance vol-
Various elastic-impedance logs calculated at different ume shows low-impedance anomalies corresponding to both gas
incidence angles are shown in Figure 6a. Elastic impedance zones (Figure 8). A much improved correlation between well

1272 THE LEADING EDGE November 2014

Downloaded from http://pubs.geoscienceworld.org/tle/article-pdf/33/11/1268/3095349/gsedge_33_11_1268.pdf
by CNOOC user
 Figure 6. (a) Elastic impedance as a function of incidence angle. Res-2 separates out from the background at higher angles. (b) Synthetic gather
for different incidence angles demonstrates a Class II anomaly at Res-2.

elastic-impedance log and pseudo-impedance log is

observed at both reservoirs.

Results
The good match between well data and in-
verted data leads to the confidence in reliability
of this output for reservoir delineation as well as
characterization and establishes that the full-stack
data are more representative of subsurface elastic
impedance (combination of P-impedance, S-im-
pedance, and density) rather than of P-impedance
alone.
The impedance maps from both P-impedance
and elastic-impedance volumes have been extract-
ed for reservoir zones (Figure 9). The comparison
shows a much improved delineation of gas sands
in elastic-impedance domain for both Res-1 and Figure 7. Comparison of full-stack seismic data with synthetic data generated using
the elastic-impedance log (0-45°) shows marked improvement in well-to-seismic ties
Res-2 (Figure 9). However, the deeper reservoir at deeper reservoir, Res-2.
(Res-2), which is barely distinguishable in acoustic
domain, has been mapped as a well-defined elastic-
impedance anomaly. Any interpretation based on
the P-impedance volumes might have missed the
gas sand completely. The drilled well has proved
gas presence in this anomaly. Another well location
based on these results, northeast of the drilled well,
has been confirmed for future drilling.

Conclusions
The mismatch between full-stack seismic
data and synthetic data has been analyzed in
detail. The AVA behavior has been found to be
responsible for strong mismatch between seismic
and synthetic data during the inversion process,
especially in Class II type reservoirs. The anal-
ysis of AVA behavior has led to the conclusion
that the presence of strong AVA in seismic data
Figure 8. Poststack inversion result using new approach brings out the imped-
affects the full-stack data, which are usually con- ance anomaly at both the reservoirs and shows good match with the elastic-
sidered as equivalent of zero-offset data. impedance log.

1274 THE LEADING EDGE November 2014

Downloaded from http://pubs.geoscienceworld.org/tle/article-pdf/33/11/1268/3095349/gsedge_33_11_1268.pdf
by CNOOC user
 The inversion attempted with mis-
matching seismic and well data is bound to
result in misleading inversion outputs that
are unsuitable for reservoir characteriza-
tion studies.
The approach discussed in this case
study uses elastic impedance for inver-
sion of full-stack data, which results in
reliable impedance volume suitable for
reservoir characterization. The approach
exploits information in offset data, is sim-
ple to use, and is easy on computational
resources.

Acknowledgments
The authors thank the manage-
ment of Reliance Industries Limited
for permitting us to publish this work,
carried out using its data and resources.
The authors thank Satyesh K. Bhan-
dari (formerly with RIL, currently with
Figure 9. Comparison of P-impedance map with elastic-impedance map for (a) Res-1 and (b)
GSPC, India) for guidance and Bhaga- Res-2. The areal extension of Res-2 anomaly is well defined in the elastic-impedance map.
ban Das (Reservoir Characterization
Group) for his interest, encouragement, References
and support. Connolly, P., 1999, Elastic impedance: The Leading Edge, 18, no.
4, 438–452, http://dx.doi.org/10.1190/1.1438307.
Corresponding author: sakshi.saha@ril.com Whitcombe, D. N., 2002, Elastic impedance normalization: Geo-
physics, 67, no. 1, 60–62, http://dx.doi.org/10.1190/1.1451331.

1276 THE LEADING EDGE November 2014

Downloaded from http://pubs.geoscienceworld.org/tle/article-pdf/33/11/1268/3095349/gsedge_33_11_1268.pdf
by CNOOC user