You are on page 1of 7

Lithology color-coded seismic sections:

The calibration of AVO crossplotting


to rock properties
Downloaded 07/08/16 to 159.178.22.27. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

By RICHARD VERM and FRED HILTERMAN


Geophysical Development Corporation
Houston, Texas

I n the late 1960s, several oil companies noticed that, in en- short, interpretation of phase reversal responses is in need of
vironments of young elastic sediments, large seismic ampli- additional tools for lithologic identification (see the report by
tudes were associated with gas-saturated sands. This method Ross and Kinman in SEG’s 1994 Expanded Abstracts).
of correlating lithology to normal incidence (NI) reflectivities If the area1 rock properties are known, one method of
was appropriately named the bright spot technique. However, lithologic identification would be to investigate the clustering
it quickly became apparent that not all large amplitudes were of NI reflectivities from different lithologic boundaries. This
necessarily associated with gas reservoirs and, much to our clustering is inferred in Figure 1 by plotting the three types of
chagrin, not all gas reservoirs had large amplitudes. For the bright spot techniques against the NI axis. As expected, there
next decade, geophysicists tried numerous techniques to re- will be some overlap of the clusters with respect to the NI
solve the ambiguity associated with lithologic identification seismic attribute. Let’s now introduce a second attribute
by means of the seismic attribute NI. This met with various which can be used with NI for a better discrimination of the
degrees of success. Finally, Ostrander’s work on amplitude lithologic boundaries.
variation with offset (AVO) led to the development of another If the S -wave to P -wave velocity ratio is approximated
seismic attribute(s) which improved our discrimination of by 1/2 and terms which are insignificant below 30”
lithologies (see “Plane-wave reflection coefficients for gas are dropped, then Shuey’s AVO equation (GEOPHYSICS
sands at nonnormal angles of incidence,” GEOPHYSICS 1984). 1985) can be reduced to two terms (see Appendix). One
But proper application of this new attribute along with NI term, NI, and the other, the far-offset reflectivity (PR).
reflectivity raised several questions. Among them were “How This simplification of the linear approximation of the reflec-
are the two seismic attributes related to rock properties which tion coefficient, expressed as a function of angle of
can then be correlated to lithology?” and “How can two seis- incidence, is
mic attributes be combined into a single display parameter?”

From bright spot to AVO interpretation. The bright spot


methodology, which is based on NI reflectivity, actually in- NI and PR are defined as
volves three different reservoir scenarios as shown in Figure
1. Each involves both the water-saturated state, represented
by a rectangle, and the hydrocarbon-saturated state, repre- =
sented by a circle. Their sizes correspond to the relative mag-
nitude of the NI reflectivities. For ease of discussion, the
reservoir will be considered sand and the hydrocarbon gas. where and represent P -wave velocity, density, and
Now, for a shale/sand interface, these three scenarios are clas- Poisson’s ratio, respectively. Notice that the expression for
sified by their change in NI reflectivity from the water-satu- PR is not a reflectivity function as commonly defined by
rated state (yellow) to the gas-saturated state (red). They are: geophysicists, but it is convenient to refer to it as the Poisson
reflectivity.
Dim spot — a large positive amplitude reduces to a smaller Most geophysicists are familiar with the acoustic imped-
positive amplitude, ance function and how it varies for different lithologies, but
Phase reversal — a small positive amplitude changes to a the significance of Poisson’s ratio is not that well known. This
small negative amplitude, and finally, begs the question: “What does a Poisson’s ratio curve look
Bright spot — a negative amplitude increases to a larger like?”
negative amplitude. In a sand/shale sequence, the Poisson’s ratio curve com-
puted from borehole measurements often resembles the SP
The dim spot method of interpretation is normally associ- well-log curve (Figure 2), the long-respected lithologic curve
ated with large acoustic impedances. In these environments, used by log analysts. One main difference between the Pois-
single-trace inversion of the stack section has been success- son’s ratio curve and the SP curve is that, when a sand is gas-
fully employed to infer lithology. Undoubtedly, the interpre- saturated, the SP curve becomes suppressed (looks more like
tation of bright spot reflections has been most successful for a shale) while the Poisson’s ratio curve deviates farther from
both interpreting lithology and making estimates of sand the shale baseline. The similarity of the SP and Poisson’s
thickness. Interpretation of phase reversal reflections, on the ratio curve suggests that geophysicists have a lithologic curve
other hand, has been difficult. In fact, geologic faults are often which is readily available from an AVO inversion based on
mistakenly introduced when phase reversals do occur. In the above equation.

AUGUST 1995 THE LEADING EDGE 847


When the bright spot methodology was discussed earlier,
a plot of NI reflectivities for a shale/water-saturated sand and
shale/gas-saturated sand (Figure 1) illustrated that NI reflec-
tivity always decreases when gas is introduced into the pore
spaces. Figure 3 has a similar role; PR decreases when gas is
introduced. Thus, the reflectivities from a shale/gas-saturated
sand interface will always plot on the NI-PR graph below and
to the left of the corresponding shale/water-saturated sand in-
terface. If the formations are reversed so that the interface has
Downloaded 07/08/16 to 159.178.22.27. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

sand/shale, then the reflectivities associated with gas plot


above and to the right of those associated with water.
The NI-PR graph in Figure 3 is color-coded to indicate the
AVO responses for various NI-PR reflectivity values. For in-
stance, NI-PR reflectivity values that plot in the blue region,
Figure 1. Bright spot classification of three distinct such as the illustrated gas-sand point, will have AVO
changes in reflection amplitude when a water-saturated responses that increase in magnitude with increasing offset.
reservoir is replaced by a hydrocarbon-saturated reser- This is indicated by the blue AVO plot on the left side of
voir. The term “bright spot” was adapted to indicate that Figure 3.
the interpretation was based on the amplitude of the
normal incidence reflectivity.

Figure 3. NI-PR crossplot of two reflection events. The


upper point (blue box) is for a shale/water-saturated sand
interface; the lower point (yellow box) is for the fluid
replacement, shale/gas-saturated interface.

The concept of separating the NI-PR graph into three


regions of AVO responses was introduced by Hilterman in
1987 in SEG’s Continuing Education course “Seismic Lithol-
ogy.” The color-coded areas in the NI-PR graph are separated
by three lines that are defined by the AVO equation presented
earlier. The three lines are:

AVO Condition Equation Line


Constant amplitude
with offset RC(0°) = RC(30°) NI = PR
Figure 2. Comparison of measured SP curve to a Far trace changes phase RC(30°) = 0 N I = –PR/3
Poisson’s ratio curve derived from measured P-wave Magnitude increases with
and dipole S-wave curves. offset if PR 0 NI=0 PR axis

An example of a water-saturated sand and a gas-saturated The color-coded plots displayed on the left side of Figure
sand beneath the same shale formation will illustrate the in- 3 are the AVO responses that fall between these three lines. To
creased lithologic discrimination available when using the illustrate the similarities to the bright spot classification, the
two reflectivities, NI and PR. The respective P-wave veloc- Rutherford and Williams classification (GEOPHYSICS 1989)
ity, density and Poisson’s ratio values for the media are: shale of AVO anomalies is also shown in Figure 3. Typically, class
(2940 m/s, 2.33 gm/cm3, .36); wet sand (3033 m/s, 2.26 1 AVO anomalies are associated with dim spots and class 3
gm/cm3, .30) and gas sand (2960 m/s, 2.12 gm/cm3, .16). AVO anomalies describe bright spots. Class 2 AVO anom-
With these rock properties, the shale/wet sand reflectivities alies overlap the range of dim spots and phase reversals and
are NI = –.04 and PR = –.13 and the shale/gas sand reflectiv- can be associated with either in an NI reflectivity analysis of
ities are NI = –.04 and PR = –.37. These reflectivity values seismic events.
can be plotted on a graph (Figure 3). Using the actual reflectivity values for the two points plot-

848 THE LEADING EDGE AUGUST 1995


ted in Figure 3, an interpretation based solely on NI ampli- discrimination has been discussed by others (such as Gidlow
tude would indicate an amplitude separation of .04 - (-.04) et al. at SEGIEAEG Summer Research Workshop, 1992,
= .08 when going from the water-saturated state to the gas- How Useful is Amplitude-Versus-Offset (AVO) Analysis?)
saturated state. However, the absolute separation of the two The crossplot of NI versus PR differentiates the water-wet
plotted points is .25. The NI-PR separation is three times the sand from the gas-charged sand. But this represents only one
vertical, or NI, separation. The increased separation is the ad- CDP location and the significance of two-way time is lost on
vantage of the two-reflectivity lithologic analysis (AVO in- the crossplot. To be effective, the crossplot needs to be con-
terpretation) over the NI reflectivity analysis (bright spot verted into a single trace format for the CDP location from
interpretation). which the NI and PR reflectivities were extracted.
Downloaded 07/08/16 to 159.178.22.27. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

With the introduction of Poisson reflectivity, the discrim- Transforming a single trace into color is relatively
ination of different lithologic boundaries has increased but, as straightforward. Discrete ranges of trace amplitudes are as-
stated before, there are now essentially two reflectivities for signed different colors. By adding another axis to the color
each CDP. How can these two be combined to produce a sin- table, two traces can be mapped into color essentially the
gle seismic section? How can the different lithologic reflec- same way that a single trace is mapped. The arrows in Figure
tion clusters be conveniently defined on a single seismic 4 illustrate the procedure. For a specific time, a value from the
section? Are AVO product sections (such as NI*PR) the only NI trace is located on the NI axis in the color matrix table.
approach? The corresponding PR value at the same time is located on
the PR axis. The color sent to the display is determined by
C rossplotting NI and PR reflectivities. Typical NI and PR where the two lines intersect in the color matrix table.
reflectivity traces for a phase reversal scenario are shown in By crossplotting NI and PR traces on top of the color ma-
the left side of Figure 4. The upper reflection in each of the trix table, colors can easily be assigned to bins for identifica-
two reflectivity traces represents a shale/wet sand boundary tion of different lithologies. In particular, by assigning yellow
to the pattern of the upper event (water-wet reflection) and
red to the pattern of the lower event (gas-charged reflection),
a color trace is produced in the right side of Figure 4 which
discriminates the water-wet sand from the gas-charged sand.
By comparing NI and PR crossplots near a well to the known
lithologic boundaries from the well logs, a color matrix table
can be designed to calibrate the local rock properties to the
seismic data. This can be best illustrated with a model.

Figure 4. Color-coding a color matrix table to transform


the two attribute traces in the left side into a single
attribute trace in the right side.

while the lower reflection represents a shale/gas sand bound-


ary. The NI trace exhibits a positive reflection for the water
sand boundary while the NI trace at the gas sand boundary
has a phase reversal response. The magnitude of these two NI
reflections is the same. However, the PR trace exhibits a two-
fold increase in magnitude from the water-wet state to the
gas-charged state. Figure 5. Sand truncation models illustrating stack
The center portion of Figure 4 is obtained by plotting NI responses for both a water- and gas-saturated sand.
and PR for each time sample. As an example, the center point
of the wavelet associated with the upper reflector has a posi-
tive NI value and a negative PR value and is thus plotted in
the second quadrant. The series of points that goes from the S tratigraphic model example. Figure 5 shows stack sec-
second quadrant to the fourth quadrant represents the upper tions for a class 2 AVO model. The upper section depicts the
shale/wet sand reflection while the shale/gas sand reflection wet sand response; the lower section is the gas response. Am-
is represented by the first-to-third quadrant series of points. plitudes for the shale/wet sand boundary are smaller than
In order to illustrate problems that might exist in real data, the amplitudes associated with shale/shale boundaries in the
PR wavelet for the upper reflector was intentionally shifted 2 same section. In the figure’s lower part, the corresponding
ms with respect to the corresponding NI wavelet. The ellipti- gas response is even weaker...in fact, the reflection event goes
cal shape of the water-wet sand crossplot resulted. This prin- away. This prospect would be difficult to promote with the
ciple of crossplotting two AVO attributes for lithologic lithologic rejoinder: “Drill when the reflection disappears!”

AUGUST 1995 THE LEADING EDGE 849


Downloaded 07/08/16 to 159.178.22.27. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Figure 6. Synthetic CDP responses for the sand trunca- Figure 9. NI*PR product section for gas-saturation model
tion models displayed in Figure 5. The rock properties with known lithologic model as color background.
were associated with the red traces of the two previously
shown models. The red middle trace is the location for which CDP gather
responses for both the wet sand and gas sand were generated.
The AVO responses, shown by the CDP gathers in Figure 6,
are only slightly better for lithologic identification. In the
water-wet case, there is a positive NI response which decays
quickly with offset. When the sand contains gas, the NI
amplitude decreases by a factor of two. More importantly, at
8500 ft offset (where offset equals depth in this model), the
reflection amplitude has completely changed polarity. If the
CDP gathers were muted at this offset or if the acquisition
was offset limited, would the gas-sand response be recog-
nized? In short, both the stack section and the CDP gathers
show subtle responses for differentiating the wet sand from
the gas sand. So let’s examine the reflection clusters on the
NI-PR crossplot for the CDP gather located at the red trace in
the lower section of Figure 5.
Figure 7, the crossplot of the extracted NI and PR reflec-
tivities, has several features of note:

1) Shale/shale reflections (green) cluster along a –45° line


Figure 7. Crossplot of NI-PR reflectivity traces corre- running from the upper left to lower right,
sponding to the gas-saturated model shown in Figure 5. 2) Shale/water-wet sands and water-wet sands/shale
The reflectivity traces were bandlimited with a zero- reflections cluster along the shale line, but just on the outside
phase wavelet.
(yellow),
3) Shale/gas sands and gas sands/shale reflections lie near
the PR axis (red); the large red dots are individual points
associated with the gas-sand response.

There is a large overlap of lithologic clusters with respect


to the NI axis. Therefore, it is not surprising that the ampli-
tude of the NI section (Figure 8) does not discriminate dif-
ferent lithologies. (The color background of Figure 8
represents actual model lithology.) In fact, as shown in Fig-
ure 9, the amplitude of the AVO product section (NI*PR) is
no better at discriminating the lithologies than the NI section.
Obviously, more help is needed. So let’s examine the color
crossplot display of NI-PR where the NI traces are overplot-
ted (Figure 10). The color background was generated by the
method associated with Figure 4. For ease in identifying
sands, a sand percentage curve for the model is superimposed
with the gas sand shaded in red.
Figure 8. NI reflectivity section for the gas-saturation Notice the strong correlation of the color background in
model with known lithologic model as color background. Figure 10 to the color backgrounds in Figures 8-9. There is

850 THE LEADING EDGE AUGUST 1995


one main difference - the color background in Figure 10 rently, a popular method for displaying the two reflectivities
comes from the seismic data itself! By using the color matrix is to multiply them together to produce the product term
table (the lower left side of Figure 10), the NI-PR crossplot NI*PR. The contours of NI*PR are overlain on the reflectiv-
yields an excellent estimate of the actual lithology. Also, this ity clusters. On the left side, notice how the contour nearest
color matrix table is similar to the well log crossplot of NI and the NI and PR axes cuts through all three lithologies. Obvi-
PR lithologies shown in Figure 7. This demonstrates the cal- ously, a display of NI*PR cannot distinguish between lithol-
ibration of rock properties to seismic data for the generation ogy for class 2 AVO anomalies because they will have similar
of a seismic lithologic section. values of NI*PR. The inability of the AVO product sections
One salient feature of Figure 10 is that the trace values are to discriminate class 1 and class 2 sands has been previously
Downloaded 07/08/16 to 159.178.22.27. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

reflectivities and not impedances; therefore, the NI-PR cross- pointed out by Castagna (GEOPHYSICS 1994).
plot colors essentially correspond to the boundaries of litho- However, there is a slight separation of the clusters along
logic changes. the PR axis so, in principle, the lithologies should be separa-
An AVO option that is still desired is the ability for a trace ble. An easy way to accomplish our goal is to apply a trans-
to represent lithology by a range of numeric values rather form to the NI and PR reflectivities which essentially rotates
than a color. With numeric coding, 3-D volumes could be au- them 45°, as shown on the right side of Figure 11. The rela-
tomatically searched for lithologic horizons. To accomplish tive positions of the lithologic clusters have not changed.
this, the cluster behavior of a class 2 AVO reflector in a However, when the rotated NI*PR contours are examined,
sand/shale sequence will be examined. each lithology gets a unique value. In fact, with the rotation,
Typical NI and PR values for different lithologic bound- gas sands should appear as large-amplitude peaks on an
aries encountered in a class 2 AVO model have been plotted NI*PR section. Essentially, the rotated NI and PR reflectivi-
on the left side of Figure 11. The shale/gas sand and the ties now behave as a class 3 gas sand. For this special case,
shale/water sand clusters occupy almost the same range in NI when the rotation is 45°, the rotated product is equivalent to
reflectivity, and the NI amplitudes are relatively small. Cur- the unrotated expression (PR2 - NI2)/2.
The original NI and PR traces used in Figure 9 were ro-
tated and the new product traces are equivalent to a bright
spot anomaly (Figure 12). The actual lithology of the model
is color coded in the background to show the correspondence

Figure 10. NI reflectivity section for gas-saturation model


with estimated lithology as color background. The litho-
logic estimation was generated by the NI-PR color-coded
crossplot.
Figure 12. Rotated NI*PR product section for gas-
saturation model with known lithologic model as color
background. Trace amplitudes now represent different
lithologic boundaries.

of high amplitudes to gas sands, mi ddle amplitudes to wet


sands, and low amplitudes to shales.

G ulf of Mexico field data. With a feel for the behavior of


class 2 AVO anomalies with model data, let’s look at some
field results. These data are Miocene epoch rocks from the
offshore Gulf of Mexico. The only significance about data
from the Miocene is that they exhibit class 2 AVO anomalies
in this area of the Gulf.
Figure 11. Axis rotation of NI and PR attributes to change Figure 13 depicts a portion of a migrated relative-ampli-
a class 2 AVO anomaly (left side) to a class 3 AVO tude section. This line traverses a discovery well which en-
anomaly (right side). countered 110 ft of gas sand. The gas zone is a high-velocity,

AUGUST 1995 THE LEADING EDGE 851


Downloaded 07/08/16 to 159.178.22.27. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Figure 13. Migrated section from portion of the Gulf of Figure 15. Estimated lithologic section for Gulf line based
Mexico where class 2 AVO anomalies are present. Well on NI-PR color-coded crossplotting. Color matrix table is
penetrated 110 ft of gas sand at 3.25 s. designed for class 2 AVO anomalies.

Figure 14. NI*PR product section for Gulf line shows no Figure 16. Rotated NI*PR product section for Gulf line.
significant indication of the known ll0-ft gas sand. The ll0-ft known gas sand is displayed as a red event to-
ward the bottom of the well.
low-density sand with respect to the encasing shale forma- expected on the downthrown side of the faults, exactly where
tion. Thus, NI reflectivity is close to zero. If the migrated sec- several yellow packages are shown in Figure 15.
tion is examined carefully at the well location, it will be noted As a final demonstration of convening this class 2 AVO
that no significant reflection amplitude is associated with the anomaly into a class 3 anomaly, the NI*PR production after
gas sand which occurs at 3.25 s. CDP gathers associated with a 45° rotation is displayed in Figure 16. In this case, the col-
the well location do not show any appreciable AVO anomaly ors were chosen to highlight water-wet sands (yellow), gas-
either. tilled sands (red), and high-acoustic impedance shale/shale
Initially, the NI and PR reflectivities were extracted from boundaries (green). In Figure 16, the lack of yellow events in
these data and a product term was produced as shown in Fig- the shallower portion of the well is consistent with borehole
ure 14. The high and low amplitudes for the product term are information which showed almost exclusively shale. With re-
represented by red and yellow colors. No red or yellow events spect to the indication of a gas sand just beneath the bottom
cross the borehole in Figure 14, proving that NI*PR products hole location, this color crossplot analysis was performed
are not always reliable gas sand indicators (particularly for after the completion of the well.
class 2 AVO reflectors).
Now, if an NI-PR color crossplot is generated using a color Conclusions and observations. For many years, log ana-
matrix table for class 2 AVO anomalies, then the gas zone lysts have crossplotted two borehole attributes for lithologic
pops out as a bright red event as shown in Figure 15. In the discrimination. Only recently have geophysicists utilized this
color matrix table, the gray diagonal identifies the shale principle by extending the one reflectivity analysis (NI) to a
zones, yellow represents water-wet sands and, of course, red two reflectivity analysis (NI and PR). By colorcoding the NI-
is for the gas zones. Notice how the lithology on this display PR matrix, a more effective discriminator of lithologies de-
is consistent with geologic principles. This area of the Gulf is velops because the separation of reflection clusters increases.
dominated by growth faults and the sand deposits should be Color was introduced to illustrate this separation along with

852 THE LEADING EDGE AUGUST 1995


axis rotation to discriminate the lithologies with a numeric If the approximations = and = are made and
value. the last term is assumed insignificant for angles less than 30°,
It is not difficult to see that additional NI-PR transforms or then equation A2 reduces to
search functions are available for assigning the lithologic
clusters a numeric value. These transformations are neces- (A3)
sary to convert the two attributes of a 3-D AVO analysis into
a single 3-D volume that can be automatically searched for where PR = Poisson reflectivity =
lithology with existing technology.
One purpose of this article was to illustrate that class 2
Downloaded 07/08/16 to 159.178.22.27. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

AVO inversion. At each time sample, the AVO coefficients


AVO anomalies can be mapped as class 3 AVO anomalies
NI and PR are extracted from a CDP gather by fitting equa-
with an AVO product term. However, the application of color
tion A3 to the CDP amplitudes. This requires a knowledge of
crossplotting works just as effectively for class 3 anomalies.
RMS velocity in order to convert the CDP gathers from a
When class 1 anomalies are examined with the color cross-
function of offset to an approximate function of incident
plotting method, it appears that NI and PR inversions to
acoustic impedance and Poisson’s ratio are necessary. Also,
angle.
other rocks besides sand and pore fluids besides gas are ap-
propriately analyzed by AVO crossplotting. Discussion. The following points should be considered con-
cerning AVO inversion:

Acknowledgements: Geophysical Development Corporation 1) Equation A3 is applied for AVO inversion, not forward
appreciates the seismic data made available by Oryx Energy and the modeling.
technical assistance provided by Christopher Ross. Of course, a 2) Equation A3 can be compared to the conventional AVO
project such as this required the assistance of many colleagues. The
equation
authors appreciate the petrophysical model work conducted by
Mark Wilson, the discussions with Jim DiSiena, and the software
development of Luh Liang.

where a is NI and b is associated with the second line on the


Appendix right side of the equation Al.
3) A slowly varying estimate of a can be extracted from
Shuey’s approximation of Zoeppritz’s equation. If equa- the NMO velocity field prior to AVO inversion. With an em-
tions 8, 12, and 13 in Shuey’s 1985 article are incorporated pirical transform such as the mudline equation (see Castagna
into his equation 14, the following results: et al., GEOPHYSICS 1985), can be estimated so that the
term in equation A3 can be replaced by the more exact
expression before AVO inversion.
4) A similar AVO inversion technique developed by
Gidlow et al. is based on the equation
(A1)
where = reflectivity at the incident angle
NI = P-wave normal incidence reflectivity, where NI, is the S-wave normal incidence reflectivity,
= average of the P-wave velocities across an + An advantage of Gidlow’s formulation is
interface, that an estimate of does not have to be known before AVO
= average of the S-wave velocities, inversion. After AVO inversion and with an estimate of
= average of the Poisson’s ratios, the difference of the normal incidence P-wave and S-wave
= average of the incident and transmitted angles, reflectivities yields
and ho are changes in and across the
interface.

In order to separate the NI term from the term, this


equation is rewritten as
The right side results from the differentiation of Shuey’s
equation 8. The above expression is very similar to PR in its
ability to differentiate different lithologic boundaries.
5) When = the AVO coefficients have the follow-
ing relationship

(A2)

AUGUST 1995 THE LEADING EDGE 853

You might also like