GEOPHYSICS, VOL. 68, NO. 3 (MAY-JUNE 2003); P. 849–862, 20 FIGS., 2 TABLES.

10.1190/1.1581037

AVO polarization and hodograms: AVO strength and polarization product

Patrice Nsoga Mahob∗ and John P. Castagna‡

attributes are considered valuable for evaluating anomalous

ABSTRACT
An alternative approach to identifying amplitude-
variation-with offset (AVO) anomalies is to consider the
AVO polarization in the AVO intercept–AVO gradient
(A-B) plane. This method does not require deviations or
separations from a background trend exhibited in tra-
ditional crossplots such as intercept-gradient (A-B) or
near trace–far trace (N-F). A benefit of the hodogram
or polarization method is that the wavelet is taken into
consideration. Crossplotted intercept and gradient are
polarized along a “background trend” for nonanoma-
lous events and at angles different from the “background
trend” for anomalous events. This allows recognition of
anomalous behavior otherwise buried in a background.
Attributes resulting from this methodology include
(1) the polarization angle, (2) the polarization angle
difference, (3) the AVO strength, (4) the polarization
product, and (5) the linear-correlation coefficient. These
different attributes can then be used to enhance AVO
interpretation. Synthetic modeling for a succession of
gas and brine layers encased in shale units indicates that
the method can potentially be an effective hydrocarbon
indicator. Application of the method to a real seismic
dataset shows polarization anomalies associated with
hydrocarbons.
INTRODUCTION
Amplitude-variation-with-offset (AVO) attributes have
been used in a variety of applications in exploration and
production. They can be applied effectively in discriminat-
ing hydrocarbon-filled reservoirs (Ostrander, 1984). More re-
cently, to improve prospect evaluations in new areas such
as deep offshore environments, AVO attributes are being
used as an analysis tool for quantitative prospect ranking
(e.g., Adamick et al., 1994, Cardamone et al., 1998). AVO
seismic amplitude responses on large 3D datasets (e.g., Barton
and Gullette, 1996). Evolving applications of AVO attributes
are the detection and characterization of fractured reservoirs
(e.g., Lefeuvre, 1994; Rueger and Tsvankin, 1995; Ramos,
1996) and in lithology determination (e.g., Nada and Shrallow,
1994).
Crossplotting AVO attributes helps in establishing trends
against which anomalous amplitude behavior can be seen
(e.g., Smith and Gidlow, 1987; Foster et al., 1997; Castagna
et al., 1998; Sams, 1998; Ross, 2000). Successful use of an AVO
crossplot requires a deviation of anomalous events (presumed
hydrocarbon-saturated reservoirs) from a well-defined “back-
ground” trend. However, when there is no deviation from the
background trend, the AVO crossplot cannot be used as an
AVO indicator. Rather, determining preferred orientations of
the sample points in the intercept-gradient (A-B) plane is an
alternative approach (Keho, 2000; Keho et al., 2001; Nsoga
Mahob and Castagna, 2002). This approach does not require
deviations from a background trend and takes into consid-
eration the wavelet as it is convolved with the reflectivity
series.
At any given interface, sample points resulting from a re-
flection have a preferred orientation and can be spread across
the four quadrants in the A-B plane (intercept-gradient space).
The angle defining any preferred orientation in the intercept-
gradient space is called the polarization angle. Nonanomalous
events related to shales and brine sands can exhibit a well-
defined orientation (or background angle).
In this paper, we investigate the following polarization at-
tributes: the polarization angle, the polarization angle differ-
ence, the AVO strength (distance from the origin in A-B plane),
the product of strength and polarization angle difference, and
the linear-correlation coefficient. In this study, we investigated
polarization attributes for a synthetic model of a succession of
gas and brine layers encased in shale units and then for a real
data case where we compared conventional AVO attributes
with the polarization attributes.

Manuscript received by the Editor October 26, 2001; revised manuscript received November 5, 2002.
∗
Formerly University of Oklahoma, Norman, Oklahoma; presently BP Exploration Inc., 900 East Benson Boulevard, Anchorage, Alaska 99519.
E-mail: Patrice.Mahob@bp.com.
‡University of Oklahoma, School of Geology and Geophysics, 100 East Boyd Street, Room 810, Norman, Oklahoma 73019. E-mail: castagna@ou.edu.
°c 2003 Society of Exploration Geophysicists. All rights reserved.

849

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
850 Nsoga Mahob and Castagna

ALGORITHM depend on the data set in hand. However, a rule of thumb is

presented below.
Concepts
For a time window about a single reflection from a given Equations governing the attributes
interface, the AVO intercept and gradient have a preferred
orientation in the A-B plane (Figure 1). The angle defining the Some equations and principles related to AVO hodograms
preferred orientation in the intercept-gradient space is called are determined by analogy with three-component vertical seis-
the polarization angle. Nonanomalous reflections related to mic profile (VSP) data analyses (DiSiena et al., 1981; Esmersoy,
shales and brine sands may exhibit a small range of orientations 1984).
with a dominant background angle. Hence, reflections with an- Polarization angle.—The polarization characteristics of a
gles different from the background angle can be considered as seismic event change in time. Therefore, the angle of polar-
anomalous. Therefore, the angle of polarization can potentially ization is characterized by the preferred orientation within a
be used in identifying AVO anomalies of any Rutherford and time window for a given time sample point (Figure 3). The po-
William (1989) class (I, II, III), or Castagna and Swan (1997) larization angle can be determined by eigenvector analysis as
class (IV). suggested by Keho (2000) for AVO hodograms and Esmersoy
One of the main benefits of this approach is the enhancement
of seismic anomalies that either exhibit small anomalies or are
embedded in the background trend (Figure 2) using traditional
crossplot-derived AVO indicators. For example, an event cor-
responding to a gas sand whose points are plotted close to the
background trend on the A-B crossplot will not show a large
separation (distance from the trend). However, such an event
may show up as a large anomaly based on polarization angle
and related attributes.

Window-size analysis.—The size of the time window is very

important in computing the polarization attributes. There is
an optimum size that gives a good temporal resolution of seis-
mic events. For a given preferred orientation or polarization
direction, the magnitude of the attribute will have the max-
imum value. The polarization attributes resulting from very
small windows are noisy, whereas attributes from very large
windows do not represent temporally the seismic reflections FIG. 2. AVO intercept and gradient crossplot showing an
(Nsoga Mahob, 2001). The optimum window size will greatly anomalous trend that is embedded in a background trend.

FIG. 1. Schematic diagram of a reflection at a boundary. The reflectivity series is convolved with a wavelet, and
the resulting AVO intercept (A) and gradient (B) traces are crossplotted. Note that the points are spread across
all the quarants.

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 AVO Strength and Polarization Product 851

(1984) for polarization analysis of three-component VSP. The of polarization at a time sample is
formulation and derivation of the polarization vector compo- µ ¶
nents are described in Appendix A. −1 Py
φ = tan , (1)
The polarization angle, φ, is determined for a sliding time Px
window. The size of the time window should be from one-half
to a wave period (Keho, 2000). For any given window, the angle where Px and Py are the components of the eigenvector (see
Appendix A).
The values of the polarization angle range from −90◦ to
+90◦ .

Polarization angle difference.—The polarization angle dif-

ference, 1φ, is the difference between the polarization angle
and the “background” angle or trend angle:
1φ = φ − φtrend , (2)

FIG. 3. Angle of polarization definition. The angle φ is mea-

sured counterclockwise from the point and the horizontal axis.
At any point M on the hodogram, a unit polarization vector
P (PX , PY ) can be computed. The polarization direction is the
preferred orientation in the time window [−N , N ].

FIG. 5. Linear-correlation coefficient attribute (r ) and sample

FIG. 4. Definition of the AVO strength attribute, L, as is related points scattering. (1) Small values of r are related to high scat-
to the minimum and maximum values of the AVO intercept (A) tering of data points within the analysis window. (2) High values
and gradient (B). of r correspond to small scattering of data points.

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
852 Nsoga Mahob and Castagna

where φ is the polarization angle and φtrend is the background where Amin is the minimum signed value within the time win-
polarization angle. dow of the analysis of A and Bmin is the corresponding B at
The background polarization angle or trend angle is com- Amin , and Amax is the maximum signed value within the time
puted from a larger time window that can be several hundred window of the analysis of A and Bmax is the corresponding B
milliseconds long. We should note that as the V P /VS ratio de- at Amax .
creases with depth (or two-way time) and as wave propagation
effects accumulate or the signal-to-noise ratio varies, the back-
ground angle could change (Castagna et al., 1998).
The polarization angle difference attribute should visually
magnify any polarization angle anomaly, thus enhancing visual
detection of the seismic amplitude anomaly.

AVO strength.—The AVO strength is the measure of the

distance of the hodogram points from the origin within the
time window of the analysis. It is one way to measure the AVO
hodogram magnitude, as mentioned in Keho et al. (2001). The
sample points from the intercept (A) and the gradient (B)
traces, on the plot can be considered as a cloud of points of
a certain length (Figure 4). The strength, L, is defined as

L = L min + L max , (3)

with
q
L min = A2min + Bmin
2
(4)

and FIG. 7. AVO hodogram of the event for the top of the B Sand,
q D Sand, G1 Sand, and I Sand. The B Sand and I Sand are gas
L max = A2max + Bmax
2 , (5) sands; the D Sand and G1 Sand are brine sands. The aperture
of the data is less or equal to 32◦ .

FIG. 6. Extracted AVO intercept (A) and gradient (B) traces along with the synthetic gather. The first event
around 1450 ms is the top of B Sand reflection.

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 AVO Strength and Polarization Product 853

Polarization product.—The product of AVO strength and
polarization angle difference, also called the polarization prod-
uct, is a measure of the magnitude of the AVO effect along
the trace. Large seismic amplitude anomalies will exhibit large
values, whereas small values will be related to nonanoma-
lous events. This attribute, L1φ, can be used to identify AVO
anomalies of magnitude above noise level.
Linear-correlation coefficient.—The linear-correlation coef-
ficient, r , of the polarization analysis is the measure of how well-
defined the polarization spread is (Figure 5). This attribute can
exhibit various effects resulting from seismic processing such
as residual normal moveout (NMO), NMO stretching, and/or
migration artifacts (Dong, 1996, 1998; and Ross, 2000). The
linear-correlation coefficient, r , is defined as (Rawlings et al.,

FIG. 8. Display of intercept trace, gradient trace, polarization angle, AVO strength, and square of linear-
correlation coefficient for the model.

FIG. 9. Display of the synthetic gather, the product of AVO strength and polarization angle difference, and the
linear-correlation coefficient for the model.

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
854 Nsoga Mahob and Castagna

1998) A synthetic common-depth-point (CDP) gather was gener-

ated using the full elastic wave algorithm of the AVO mod-
(Cov(At+i , Bt+i )) 2
r2 = , (6) eling module of the AVO Hampson-Russell software. A zero-
Var(At+i ) × Var(Bt+i ) phase Ricker wavelet with a dominant frequency of 40 Hz and a
where Cov and Var are the covariance and the variance,
respectively.
In our application, we use r 2 which ranges from 0, when there
is very high scattering of hodogram points, to 1, when there is
no scattering of hodogram points about the polarization trend
within the analysis window.
MODELING EXAMPLE

Model parameters
Consider a flat-layered model made of a succession of gas-
and water-saturated sand units encased in shale or silt units
(Table 1) based on well log data from the Northwest Shelf of
Australia. The B Sand, C Sand, I Sand, and M Sand are gas
sands. The D Sand, G1 Sand, and L Sand are water-saturated
sands. The G Sand is a tight gas sand. The model elastic
parameters are presented in Table 1.

FIG. 10. Stacked seismic line with known gas- and brine-sand FIG. 11. Display of a NMO-corrected CDP gather close to the
intervals. The gas-sand zones are indicated in dark gray; well used in the study. The interval of interest is between 2900
brine-sand intervals are shown in light gray. and 3300 ms.

Table 1. Elastic parameters of the flat-layered model.

Layer Name Thickness (ft) V P (m/s) VS (m/s) ρ(g/cm3 ) Poisson’s ratio

1 Barrow Group 1000 3640 2000 2.45 0.33
2 B (gas) 300 3530 2390 2.27 0.10
3 shale/silt 200 3610 2040 2.42 0.30
4 C (gas) 200 3625 2235 2.35 0.10
5 shale/silt 200 3450 1900 2.30 0.31
6 D (brine) 200 3915 2540 2.40 0.25
7 shale/silt 200 3615 2025 2.49 0.30
8 G (tight gas) 250 3985 2435 2.45 0.20
9 shale/silt 175 3755 2020 2.47 0.30
10 G1 (brine) 250 3830 2425 2.37 0.20
11 shale/silt 200 3740 2125 2.45 0.30
12 I (gas) 300 3550 2415 2.33 0.10
13 silt 200 3960 2080 2.45 0.30
14 L (brine) 200 4140 2555 2.43 0.20
15 shale/silt 200 3995 2140 2.40 0.28
16 M (gas) 300 3830 2540 2.33 0.10
17 shale/silt 350 4320 2460 2.50 0.30

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 AVO Strength and Polarization Product 855

length of 200 ms was used. The range of offsets modeled varies methodology. Known hydrocarbon intervals are compared to
from 0 to 16405 ft (5000 m). the derived attributes.

Polarization attribute generation Seismic data

A gradient analysis was performed to extract the intercept
(A) and the gradient (B) traces using a maximum incidence
angle of 32◦ . The resulting intercept and gradient traces are
depicted in Figure 6.
Representative hodograms of events corresponding to the
top of B Sand, D Sand, G1 Sand, and I sand are shown in
Figure 7. A time window size of 20 ms, corresponding to about
80% of the period of the seismic data was used to compute the
polarization attributes displayed in Figure 8. A constant back-
ground angle of −20◦ was used to compute the polarization
angle difference.
From the synthetic results (Figure 9), note that the porous gas
sands correspond to large product of strength and polarization
angle difference (L1φ), whereas brine sands do not. The tight
gas sand is represented by a very small value of the polarization
product.
REAL CASE EXAMPLE
Polarization attributes are computed using real seismic data
from the Northwest Shelf of Australia to investigate the
One prestack time-migrated (PSTM) 2D line, extracted from
a 3D survey, is used for this study (Figure 10). The dominant
frequency of the seismic data is roughly 30 Hz, and the ap-
proximate tuning thickness is about 30 m (or 18 ms two-way
time) in the reservoir section of interest. Some of the known
gas and brine intervals are highlighted on the seismic sections
(Figure 10). A CDP gather close to the well of interest is de-
picted in Figure 11.
Conventional AVO attribute generation
To perform the gradient extraction, the smoothed corrected
sonic curve at a nearby well location was used for the velocity
function. The following constraints are set during the analy-
sis: range of incidence angles = 8–32◦ , range of offsets = 280–
3160 m. The resulting AVO product (A × B) and scaled
Poisson’s ratio change (0.5A + 0.5 B) (Verm and Hilterman,
1995) sections for the line are depicted in Figures 12 and 13,
respectively. The known hydrocarbon and brine zones are high-
lighted and colorcoded. Overall, note that porous gas intervals
correspond to larger AVO products and scaled Poisson’s ratios

FIG. 12. Display of the AVO product (A × B) attribute along the seismic line. The red boxes and arrows indicate the gas-sand
intervals and the light blue boxes and arrows show the brine-sand zones.

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
856 Nsoga Mahob and Castagna

than the brine-sand intervals do. However, brine sands also ex- tributes at the well location are depicted in Figures 16–20,
hibit large values. A crossplot of AVO intercepts and gradients respectively.
and the corresponding seismic section is shown in Figure 14.
There is a clear separation from the defined background trend RESULTS
of the seismic reflections related to the porous gas sand at Figures 12–13 and 15–19 indicate that the porous gas sands
2910 ms, but the other known gas sand at roughly 3200 ms can be better identified on the polarization product section
is not exhibited as anomalous on both the crossplot and the than on the conventional AVO attributes, where gas and brine
seismic section. sands can exhibit the same signature. The lateral extent of the
Polarization attribute computation
Table 2. Average background angle values for the seismic
The extracted intercept (A) and gradient (B) traces for the line. Two time windows are considered: 2500–2700 ms and
2600–2800 ms. The angles obtained from the other seismic lines
seismic lines were used to compute the polarization attributes: are similar to the values presented in the table.
(1) polarization angle (φ), (2) polarization angle difference
(1φ), (3) AVO strength (L), (4) polarization product (prod- Window 1 Window 2
uct of strength and polarization angle difference (L1φ), and Trace (2500–2700 ms) (2600–2800 ms)
(5) square of linear-correlation coefficient (r 2 ). A 24-ms slid-
431 −12.60 −22.20
ing window was chosen for the computation. Since the domi- 432 −13.40 −12.60
nant frequency of the seismic data is approximately 30 Hz, the 433 −12.60 −13.40
time window for the computation is roughly 0.727T (where 434 −22.20 −15.80
T is the seismic wave period of the data), a value within the 435 −22.20 −15.80
suggested range (Keho, 2000). A constant background polar- 436 −19.00 −12.60
437 −12.60 −16.60
ization angle of −20◦ was used for the entire trace to calcu- 438 −17.40 −15.80
late the polarization angle difference. The background value 439 −15.80 −22.20
is determined after examining the polarization angles along 440 −16.60 −12.60
a series of traces (12) of the seismic line, particularly out- 441 −18.20 −17.40
side the zone of interest (Figure 15, Table 2). The five at- Average −16.60 −16.09

FIG. 13. Display of the scaled Poisson’s ratio change (0.5A × 0.5 B) attribute along the seismic line. The red boxes and arrows
indicate the gas-sand intervals and the light blue boxes and arrows show the brine-sand zones.

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 AVO Strength and Polarization Product 857

sand bodies seem well defined on the polarization attributes,
whereas the delineation is not clear on the AVO product sec-
tion or on the scaled Poisson’s ratio attribute.
Known hydrocarbon and brine intervals for the case-study
seismic line exhibit different signatures on the polarization at-
tributes. Overall, gas-sand zones are indicated by a large po-
larization product (positive), whereas brine-sand zones exhibit
smaller (or negative) values. This is validated by the high val-
ues of the square of linear-correlation coefficient (≥0.60) in
gas intervals, but intervals of large polarization product corre-
sponding to brine sands have small values of r 2 (≤0.2), mean-
ing that there is a high scattering of time sample points about
the polarization trend within these analysis windows. A large
polarization product with large r 2 identifies every productive
gas zone. The single large polarization product associated with
brine had a low r 2 .
CONCLUSIONS
It has been shown from the synthetic results that porous gas
sands correspond to large polarization product (L1φ), whereas
brine sands do not. This is in agreement with the real data result.

FIG. 14. (a) Crossplot of AVO intercepts and gradients. The purple indicates the top of gas-sand reflections,
yellow illustrates the base of gas-sand reflection, and blue represents the background. (b) Overlay of intercept
traces and color zones generated from the crossplot in (a).

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
858 Nsoga Mahob and Castagna

FIG. 15. Background polarization angle determination for seismic line 2. Two time windows are considered for the analysis:
2500–2700 ms and 2600–2800 ms. The background angle calculated at a trace from each window is the arithmetical average of
the polarization angles within the window. The trace average values for each are presented in Table 2. After examination of the
values, a rounded value of −20◦ is chosen for the attribute computation.

The square of the linear-correlation coefficient (r 2 ) provides
an indication of the reliability of the result. For a good en-
hancement of AVO interpretation, the polarization attributes
should be used in conjunction with the correlation coefficient.
The polarization product and the linear-correlation coefficient
seem to be the most useful attributes for the synthetic and real
data investigated.
The study results from the polarization methodology suggest
that:
1) Polarization attributes should be considered as an alter-
native approach to identifying AVO anomalies.
2) Polarization attributes can enhance AVO interpretation.
3) Polarization attributes can potentially be used as a re-
connaissance tool to identify possible hydrocarbon (gas)
intervals.
AVO polarization attributes are potentially useful hydro-
carbon indicators. For a synthetic model and real seismic
data example, large polarization products combined with high
linear-correlation coefficients were found to correlate with the
presence of hydrocarbons. However, this technique will not
work properly if the signal-to-noise ratio of the data is very
poor. In addition, a seismic event, corresponding to a very thin
gas sand, hidden within the sidelobe of a large background
event, would not be detectable. This method might fail when
analyzing very low frequency data.
AVO strength and polarization product attributes enhance
and highlight amplitude anomalies related to gas sands and
brines better than conventional AVO attributes.
ACKNOWLEDGMENTS
The authors express sincere thanks to Chevron Overseas
Petroleum Inc. for making the data available to us. We thank
BP for financial support. We also thank Dr. Bill Lamb for use-
ful discussions and advice in programming the polarization at-
tributes. Thanks to Hampson-Russell Software Services for use
of its AVO software. The authors show their sincere appreci-
ation to Herbert Swan and the other reviewers for their con-
structive comments.
REFERENCES
Adamick, J., Hall, D., Skoyles, D., DeWildt, J., and Erickson, J., 1994,
AVO as an exploration tool; Gulf of Mexico case studies and ex-
amples: 64th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded
Abstracts, 1107–1111.
Barton, J., and Gullette, K., 1996, Reconnaissance amplitude versus
offset techniques in the Niger Delta (abs.): AAPG Bulletin, 80,
1272.

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 AVO Strength and Polarization Product
FIG. 16. Display of the polarization angle (φ) attribute along the seismic line. The black arrows indicate known gas-sand zones; the
white arrows show brine-sand intervals.
Cardamone, M., Marini, I., and Bertelli, L., 1998, New 3D visualization
and analysis tools improve prospect evaluation in a deep offshore
environment (abs.): AAPG Bulletin, 82, 1898.
Castagna, J. P., and Swan, H. W., 1997, Principles of AVO crossplotting:
The Leading Edge, 16, 337–342.
Castagna, J. P., Swan, H. W., and Foster, D. J., 1998, Framework for
AVO gradient and intercept interpretation: Geophysics, 63, 948–
956.
DiSiena, J. P., Gaiser, J. E., and Corrigan, D., 1981, Horizontal com-
ponents and shear wave analysis of three-component VSP data, in
Toksoz, M. N., and Stewart, R. R., Eds, Handbook of geophysical
exploration, vol. 14B: Advanced concepts: Geophysical Press, 177–
188.
Dong, W., 1996, Fluid line distortion due to migration stretch: 66th Ann.
Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1345–
1348.
——— 1998, AVO detectability against tuning and stretching artifacts:
Geophysics, 64, 494–503.
Esmersoy, C., 1984, Polarization analysis, rotation and velocity estima-
tion in three-component VSP, in Toksoz, M. N., and Stewart, R. R.,
Eds, Handbook of geophysical exploration, vol. 14B: Advanced con-
cepts: Geophysical Press, 236–255.
Foster, D. J., Keys, R. G., and Reilly, J. M., 1997, Another perspective
on AVO crossplotting: The Leading Edge, 16, 1233–1237.
Keho, T. H., 2000, The AVO hodogram: Using polarization to identify
anomalies: 70th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded
Abstracts, 118–121.
Keho, T. H., Lemanski, S., Ripple, R., and Tambunan, B. R., 2001, The
AVO hodogram: The Leading Edge, 20, 1214–1224.
Lefeuvre, F., 1994, Fractured related anisotropy detection and analysis:
“and if P-waves were enough?”: 64th Ann. Internat. Mtg., Soc. Expl.
Geophys., Expanded Abstracts, 942–945.
Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
859
Nada, H., and Shrallow, J., 1994, Evaluating geophysical lithology
determination method in the central offshore Nile Delta, Egypt:
64th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts,
1112–1113.
Nsoga Mahob, Z. P., 2001, AVO polarization attributes and hodograms:
Ph.D. diss., Univ. of Oklahoma.
Nsoga Mahob, Z. P., and Castagna, J. P., 2002, AVO hodograms and
polarization attributes: The Leading Edge, 21, 18–27.
Ostrander, W. J., 1984, Plane-wave reflection coefficients for gas sands
at nonnormal angles of incidence: Geophysics, 49, 1637–1648.
Ramos, A. C. B., 1996, Three-dimensional AVO analysis and
anisotropic modeling applied to fracture characterization in coalbed
methane reservoirs, Cedar Hill field, San Juan Basin, New Mexico:
Ph.D. diss., Colorado School of Mines.
Rawlings, J. O., Pantula, S. G., and Dickey, D. A., 1998, Applied regres-
sion analysis: A research tool, 2nd ed.: Springer-Verlag New York,
Inc.
Ross, C. P., 2000, Effective AVO crossplot modeling: A tutorial: Geo-
physics, 65, 700–711.
Rueger, A., and Tsvankin, I., 1995, Azimuthal variation of AVO re-
sponse for fractured reservoirs: 65th Ann. Internat. Mtg., Soc. Expl.
Geophys., Expanded Abstracts, 1103–1106.
Rutherford, S. R., and Williams, R. H., 1989, Amplitude-versus-offset
variations in gas sands: Geophysics, 54, 680–688.
Sams, M., 1998, Yet another perspective on AVO crossplotting: The
Leading Edge, 17, 911–917.
Smith, G. C., and Gidlow, P. M., 1987, Weighted stacking for rock
property estimation and detection of gas: Geophys. Prosp., 35, 993–
1014.
Verm, R., and Hilterman, F., 1995, Lithology color-coded seismic sec-
tions, The calibration of AVO crossplotting to rock properties: The
Leading Edge, 14, 847–853.
860 Nsoga Mahob and Castagna

FIG. 17. Display of the polarization angle difference (1φ) attribute along the seismic line. The black arrows indicate known gas-sand
zones; the white arrows show brine-sand intervals.

FIG. 18. Display of the AVO strength (L) attribute along the seismic line. The black arrows indicate known gas-sand zones; the
white arrows show brine-sand intervals.

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 AVO Strength and Polarization Product 861

FIG. 19. Display of the polarization product (L × 1φ) attribute along the seismic line. The black arrows indicate known gas-sand
zones; the white arrows show brine-sand intervals.

FIG. 20. Display of the square of linear-correlation coefficient (r 2 ) attribute along the seismic line. The black arrows indicate known
gas-sand zones; the white arrows show brine-sand intervals.

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
862 Nsoga Mahob and Castagna

APPENDIX A
POLARIZATION-VECTOR COMPONENTS DERIVATION

The formulation of the polarization vector is derived from The eigenvectors corresponding to the characteristic roots (λ)
the correlation matrix Rm that is used to compute the eigen- calculated from equation (A-3). The two vectors are orthogo-
vectors (Esmersoy, 1984): nal. Nsoga Mahob (2001) showed that the components of the
vector corresponding to the larger eigenvalue are
1 XN
Rm = r (i)r T (i), (A-1) Ã !
2N + 1 i=−N Px
=
where N is half of the length of the time window (in sample
Py
points) and r (i) represents the observed data in the time win-  X 
dow of interest. The subscript m is the center sample point of √ At+i Bt+i
 2 
the time window [-N , N ], which is rectangular.  vÃ
i

The matrix Rm from equation (1) can be expanded in the  1 u !2 Ã !2
[1 + D] 2 u X X X 
A-B plane as follows:  t4 
 At+i Bt+i + A2t+i − 2
Bt+i ,
   
X
N X
N  i i i 
 √ 
 A2t+i At+i Bt+i   2 1 
1  i=−N  [1 + D] 2
Rm =  i=−N ,
2N + 1   2
XN XN
2 
At+i Bt+i Bt+i (A-5)
i=−N i=−N
where
(A-2) Ã !
X X
where At+i is the AVO intercept value at time sample t, and A2t+i − 2
Bt+i
Bt+i is the AVO gradient value at time sample t. Rm is a 2 × 2 i i
D=v à !2 à !2
symmetric matrix, and its eigenanalysis can be done efficiently. u
u X X X
The eigenvalues are obtained by solving the equation t4 A + A2t+i − 2
t+i Bt+i Bt+i
|Rm − λI| = 0, (A-3) i i i

(A-6)
where λ represents the eigenvalues or characteristic roots and
I is the unity matrix: with i = −N , . . . , N . Px and Py are normalized so that a unit
à ! polarization vector is considered. That is,
1 0
I= . (A-4)
0 1 kPk = 1. (A-7)

Downloaded 16 Oct 2012 to 75.148.212.146. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/