Download as pdf or txt
Download as pdf or txt
You are on page 1of 35

Understanding of AVO and its use in Interpretation

Many wish to learn about Amplitude versus Offset (AVO) for interpretation believing that AVO is
a direct hydrocarbon indicator (DHI). AVO is not a DHI it is one of the tools we utilize in exploratory data
analysis (EDA) and AVO anomalies can help us to focus our attention upon locations for further
consideration. AVO should not be the sole reason for drilling a prospect we need to look at various
factors including:
1) The CDP that produced the AVO anomaly;
2) How was the data acquired and processed;
3) How does the anomaly fit with geology;
4) How well the AVO fits well modeling;
5) Is the AVO present after inversion;
6) And, Pore pressure and seal integrity.
This paper focuses on reflective seismic data and attributes and touches briefly upon seismic
inversion and well modeling but does not deal with pore pressure or seal integrity. It offers at the end a
summary of how we should present prospects and the questions we should be asking ourselves during
the interpretation. It emphasis a integrated approach and is a part of a larger course that I teach.
We begin with the acquisition of seismic data and the definition of a Common Depth Point (CDP)
because AVO is found within the CDP domain. Though our stack products may reflect anomalous
amplitudes it is due to the AVO in the CDP domain. This paper emphasis an integrated Prestack Gather
Interpretation approach towards AVO.

Recording of Seismic Field Data


When recording seismic data in the field we set off a source which creates energy. This energy
goes into the earth and is either reflected or transmitted.

Figure 1:

When a non-normal incidence wave strikes the interface between two


rockboundaries we have the creation of P-waves and S-waves. The S-waves are
transverse waves so the particle motion is perpendicular to the direction of
propagation so we need three components to record the shear wave. These Swaves created at the rock boundaries may also be called converted waves.

The reflected energy waves then go back up to the receivers and are recorded over a
predetermined time period (called the record length). The receivers are hydrophones in water and
geophones on land. When recording the data we reduce it further by recording the data at a sample
rate. This allows reducing of the number of samples but retaining the wavelets of the incoming waves.
The sample rate actually limits the frequencies we record. We can not record anything beyond the
Nyquist frequency. This is due to the fact any frequencies beyond the Nyquist frequency will be aliased.
If we record the data at 2 mseconds then the sampling frequency is 1/.002 seconds which is 500 Hz. The
Nyquist frequency is then half of the sampling frequency which is 250 Hz. Anything, which is above 250
Hz, will not be properly sampled so it can be properly restored (aliased). When recording the data we
apply a field filter to limit the bandwidth to the useable frequencies such as Out-225 Hz where the upper
limit is near but not equal to the Nyquist frequency.

Figure 2:

Data sampled at 4 msec with a frequency of 125 Hz (A). The wavelet is restored
as it should be. When we increase the frequency to 167 Hz we see that the wave is
aliased with a low frequency imprint (red curve).

Figure 3:

Diagram showing a marine field file/shot record.

What is a CDP
A Common Depth point or CDP is an actual physical point on the earth represented by latitude
and longitude or X and Y coordinates where we record all the data for the subsurface. Common depth
point generally lies halfway between the receiver and the source. It is better to use a 2D seismic line to
understand a CDP. Common Depth Point is made up of several shot-receiver pairs. This is done so that if
we have a bad shot or a bad receiver it will not affect the data. The fold of the data represents the
number of shot-receiver pairs that make up a CDP. If we have a fold of 6 then we have 6 shots-receivers
that make up the CDP. This creates redundancy and allows us to stack or sum of the data into a single
trace.
A CDP is composed of traces which are spaced at offsets or distances from the shot. Each offset
or trace is from a different shot and is recorded at a different receiver. The offsets near the shot are
referred to as the Near Offsets and the offsets furthest away from the shot are the Far Offsets. The Near
Offsets tend to be contaminated by shot noise and the Far Offsets are contaminated by multiples.
The reflected signal from a rock interface is hyperbolic in shape and to flatten this hyperbola we
utilize the following equation:
2
VRMS
=

X n2
(T02 Tn2 )

Equation 1: T Equation (Taner and Koehler, 1969).


Where:

Vrms is the root mean square velocity or the velocity used to flatten the event.
Tn is two way reflection time.
T0 is two way vertical time.
Xn is the offset at Tn.

This is why seismic is composed of offset, time and velocity. Once flattened, we can sum all the
traces together assuming that the noise will not follow a hyperbolic moveout and will stack out when
we stack the data together. This is how we produce stack data.
The fold of data or number of traces within a CDP for 2D seismic line is defined by the equation:
Fold= Group Interval * Number of Channels /2*shot point interval.

Equation 2:

Seismic Fold

Figure 4:

Diagram showing the definition of a CDP or Common Depth Point. A CDP is made up of
traces from different shot points recorded at different receivers so as to minimize the
noise within the data. As can be seen this CDP is made up of three shots. The first shot
(1) is at shot point 201 and is recorded at receiver 200. The second shot (2) is shot at
shot point 202 and receiver 199; and the third shot (3) is shot at shot point 203 and
receiver 198. The CDP is at station 200.5. When we record the signal we find the primary
signal to be hyperbolic and utilize a hyperbolic equation to flatten the signal.

Understanding what a CDP is important because this is the basic building block of seismic and it
is the basis for Amplitude versus offset.

What is Normal Moveout (velocity)


The purpose of velocity analysis is to choose a velocity, which will flatten the true reflectors
within the seismic gathers so that we can achieve the optimal stacking of these reflectors. The multiples
or noise that are within the data will not be aligned. This velocity is referred to as stacking velocity
(Vstack) and it is approximately equal to the Root Mean Square Velocity (VRMS). The difference
between Vstack and VRMS is VRMS is calculated by the NMO equation and Vstack is picked from flat
gathers and semblance based coherency.
A multiple is energy which is reflected upwards from a primary and then strikes a boundary such
as the water bottom, is reflected down again and then strikes the same reflector again and bounces
back up to be recorded.

Figure 5:

Water Bottom multiples are waves which bounce between the water bottom and the
top of the water. Each time the wave bounces we refer to it as a different order of water
bottom multiple. We have 1st order water bottom multiples which are the first bounce,
and then we have the 2nd order which is the 2nd bounce and so forth. These are
referred to as the water bottom multiple reverberations. Peg leg multiples are waves,
which are caught in between the layers of rocks.

This is why multiples always occur at a multiple of two of the primary. If the primary is recorded
at 2 seconds the multiple will be at 4, 6, 8 seconds. Multiples are hard to remove and it is important that
we do remove them from the seismic data. Utilizing velocities is the best way to identifying the multiple.
The multiple will have the same velocity as the primary. As mentioned above as we increase in depth
the velocity increases so multiples at 4, 6 and 8 seconds will be slower than the primary reflectors and
when the velocities are applied will appear to be too fast or curved down in the CDP.
We have different types of multiples such as the water bottom multiple which was described
above; multiples generated from near surface hard layers; and Peg leg multiples that are waves, which
are caught in between the layers of rocks. Water bottom multiples occur in the seismic at multiples of
the water bottom. PEG legs are harder to find and they have a smaller discrimination in move out than
the water bottom multiples and are harder to remove. Water bottom multiples generally have a lower
velocity and frequency than primary energy at the same two-way time. This is due to the additional
frequency attenuation on the multiples that have traversed a longer path through the earth filter. This
makes them easier to identify and subsequently remove. Peg leg multiples generally have a similarly
velocity and frequency component to the adjacent primary energy and this makes their identification
and removal a lot more difficult.

(A)

Figure 6:

(B)

On the left hand side (A) is a model of a CDP gather without velocities. We utilize an
equation called the T equation to calculate the root mean square velocity which is the
velocity to flatten the primary reflectors. On the right (B) is the same model of a CDP
gather with the proper velocities applied. When the proper velocities have been applied
to real events we have flat reflectors. A multiple has the velocity of the layer in which it
is a multiple of. When velocities are applied to the multiple it is not flattened but
appears still as a hyperbola concave down because the velocity is too fast.

We can use velocity analysis pick the optimal Vstack; and, to identify possible multiple reflectors within
the data. We generally utilize three tools to do this:

1.

The first is applying the velocities picked directly to the gathers to see how flat the gathers are.
By doing this we can also recognize possible reflectors which are multiples so that we may not
pick these events.

2.

A second tool is semblance-based coherency measures where velocities are estimated by


maximizing a coherence measure with respect to the hyperbolic parameter.

3.

A third method is overlaying the picked velocities on the seismic data to see how the velocities
relate to the geology.

What is Acoustic Impedance and Reflection Coefficient


When we send an elastic wave in the earth we have reflection off of the rock interfaces and if
we record the reflection directly above where we create the downgoing wave we will see that we have
just the primary wave or P-wave. This occurs on the near offsets or near incidence angles (0-15 degrees)
on the seismic data. If we were to invert the near offsets we would have the acoustic impedance. What
affects the acoustic impedance is changes in the density and the P-wave velocity. These changes can be
changes within the rock such as changes in porosity or they can be changes within the fluids of the rock.
Acoustic Impedance is a physical rock property. It is the product of the product of the density
() and the P-wave velocity (Vp). Any medium that can support wave propagation may be described as
having impedance. We can actually use well logs (sonic and density logs) to calculate the acoustic
impedance log. Sonic logs measure transit time which is measured in microseconds/feet (1/velocity*106) and density is measured in gm/cc so if we divide the density by the sonic log *10-6 we obtain the
Acoustic Impedance log. One of the physical properties we invert seismic for is the Acoustic Impedance.
When a seismic wave encounters a boundary between two different materials with different
impedances, some of the energy of the wave will be reflected off the boundary, while some of it will be
transmitted through the boundary. We can calculate the amplitude of the wave created at the boundary
utilizing the equation for the Reflection Coefficient which is:
Reflection Coefficient=(AI2-AI1)/ (AI2+AI1)

Equation 3: Reflection Coefficient


where AI is the Acoustic Impedance. The Reflection Coefficient describes the amplitude at zero offset
We can use Acoustic Impedance and Reflection Coefficient to create a synthetic seismogram
from well data (sonic and density). Synthetic Seismograms is seismic data created from well data such as
a sonic log and a density log. If we don't have density log we can calculate density using things like
Gardner's equation. With synthetic seismograms and top of formations we can tie the seismic to the
well data and begin to understand which reflector is which in our interpretation.

To make a synthetic seismogram we need velocity information that can come from sonic logs,
and density information that may come from a density log or it can be derived from velocity information
utilizing petrophysical equations. The sonic log should be check shot corrected to stretch the seismic log
to a surface seismic scale. Once we have the velocity and density information we can create a reflection
coefficient log that gives the amplitude of the reflection coefficients at the interfaces. We then convolve
the reflection coefficients with one of the following wavelets: Ricker, Klauder, Ormsby, Butterworth or a
user defined wavelet usually derived from the seismic. Each wavelet has its unique characteristics. We
then convolve the wavelet with the acoustic impedance log and see which synthetic seismogram better
fits the seismic data utilizing cross correlation which calculates the similarity between the seismic and
the synthetic. Stretching and squeezing may be necessary due to dispersion between seismic velocities
and sonic log velocities. If the velocity varies with frequency, that is dispersion. Seismic frequencies are
below 120 Hz and well logs frequencies are in the range of kilohertz. If we have an actual processed VSP
it is better to tie the seismic with then the synthetic seismogram. The VSP is the seismic recorded in the
borehole so it matches the seismic better.

Figure 7:

Diagram of a synthetic seismogram.

Mis-Ties of Synthetics

Issue

Solution

Reflectors seem to come in at the Synthetics usually need to be stretched or squeezed to fit data. This is usually because of an
wrong times.
inadequate measurement of the velocities in the wellbore.
The wiggles look different.

Try adjusting the frequency and phase content of the wavelet used to create the synthetic.

Extra reflections on the real


data.

Multiples

The wiggles are a lot stronger on


the real data than on synthetics,
especially at depth.
AGC is on the seismic data

Table 1::

Causes of mis-ties between synthetic seismograms and well data.

The sonic log is key to a good synthetic seismogram. There may be problems with the sonic log
such as:
1)

Cycle skipping (Wash out zones);

2)

Short logging runs, or gaps in sonic log coverage;

3)

Relative pressure differences between the drilling fluid and the confining stress of the rocks
around the wellbore;

4)

Gaps in sonic log coverage;

5)

And, shale alteration.

We need to rebuild the sonic log to remove these problems because they will affect the synthetic
seismogram.

How do we do well modeling with Linearized Zoeppritz Equations


Unfortunately seismic is not that simple. When we record data we do so with multiple receivers
and that means we have different offsets and the energy is striking the interfaces not at right angles so
we have the creation of shear data. To describe this behavior we need to use the Zoeppritz equation.
The Zoeppritz equations give exact coefficients for idealized transmission, reflection, and conversion
events; their complicated structure necessitates the use of non-linear inversion techniques. To deal with
this we have formulated linearized versions of the Zoeppritz equation.
The linear approximation of Zoeppritz equations theoretically fits amplitude changes to offsets.
It does this utilizing the Reflection Coefficient at zero offset (Rp, NI, and Intercept), which is constant as
the first term in all the equations. The second term that is a particular AVO attribute, such as shear wave

reflectivity, Poisson reflectivity, or gradient. The difference between the linear approximations is the
assumptions we need to make in order to simplify the Zoeppritz equations which affects the complexity
and accuracy of the approximation used (Burianyk, 2000).
Bortfeld (1961)
In 1961 Bortfeld created one of the earliest linearized equations to the Zoepprtiz
equations. The Aki and Richards (1980) equation is derived from Bortfeld. Not many talk about Bortfeld
but some prefer Bortfeld to Shueys.

Fluid term

Rigidity term

Where =Vp, =Vs and =density

Equation 4: Bortfeld (1961) Equation


If we look at the rigidity term it is primarily shear velocity and shear velocity is not affected by
the changes in the fluid so the rigidity term for gas and for water are almost equal. Fluids within the rock
affect the fluid term.
Aki and Richards (1980)

R ( ) =

1
1

(
+
)+ (
4
2

2
2

2
2

1
) sin 2 +
(tan 2 sin 2 )

We can drop the last term so it becomes

R ( ) =

1
1

(
+
)+ (
4
2

2
2

2
2

) sin 2

Equation 5: Aki and Richards (1980) Equation


The Aki and Richards (1980) linearized approximation is the starting point for most AVO
inversion work.

Shueys Equation (1985)


Shuey (1985) attempted to make the equation into a linear equation of y=mx+b where m is the
gradient and b is the intercept thus this is where we get the terms intercept and gradient. If we were to
plot amplitude versus square of sin where is the angle of incidence of a single seismic event then the
intercept would be the zero offset stack and the gradient is the slope of the best-fit line through these
points.
Change in the gradient could reflect a change in fluid but may also reflect a change in lithology.
This approximation is good to about 30 degrees of incidence angle.

Figure 8:

The mathematical formula for the 3-term Shuey AVO (1985) Attributes: intercept (A),
gradient (B) and curvature (C). It shows the relationship between intercept and
curvature to isolate the density term. The curvature term generally is unstable because
of the noise upon the gathers. It generally is derived where we have the NMO stretch
upon the gathers.

Shuey was also the first to basically tie AVO attributes with angle stacks. We can also see that if
we utilize the Intercept and curvature we can get density. Theoretically this is possible but the curvature
term tends to be unstable.

Smith and Gidlow (1987)


Smith and Gidlow (1987) attempted to work with P-wave and Shear-wave reflectivity. As stated
above the P-wave Reflectivity is going to reflect changes in fluid and S-wave reflectivity changes in
rigidity or matrix, which comes back to Bortfelds (1961) equation of fluid and rigidity term. We can also
relate Shueys (1985) equation to Smith and Gidlow (1987). Let

1
Rp = ( + ) R S = 1 ( + )
2
2

B= (

2
2

4
2

2
2

2
2

)=
2

2
2

(2


+
)

1
1

1
(2
+
) then B = R p 2 RS
then B =
2
2

R ( ) = R P + ( R P 2 RS ) sin 2 where NI= R P and B = R P 2 RS and the fluid factor is equal to

2( R P RS ) . Rs=A-B and shear reflectivity discriminates porosity.


Truncating the third term of the Shuey equation is equivalent to constraining the P-wave velocity
reflectivity to zero

= 0

which is clearly not appropriate in a variable velocity earth. This is what leads to the bias in the estimate
of the gradient (B). Smith and Gidlow (1987) suggested a more geologically realistic constraint to
improve the stability of the inversion.
They suggested using the Gardner et al. (1974) empirical relationship linking density and P-wave
velocity to establish a relationship linking the density and P-wave velocity reflectivity thus Smith and
Gidlows equation assumes that the density follows Gardners equation.
Hilterman and Verm (1995) Equation
Hilterman and Verm (1995) equation tries to relate back to the earlier work of Ostrander (1984)
and Koefoed (1955) where they utilize Near Impedance and Poisson Reflectivity. The equation is

where NI is equal to

Equation 6: Near Impedance


and Poisson reflectivity or PR is

Equation 7: Poisson Reflectivity


The Poisson Ratio or is:

Equation 8: Poisson Ratio


Hilterman and Verm (1995) equation is one of the better equations to use to model the seismic
AVO response using Vp, Vs and from well data or petrophysical equations. Using Excel spreadsheet a
rock calculator can be built.

Figure 9:

Excel spreadsheet that utilizes Vp, Vs, and to calculate the amplitude with offset of
two different rock.
We can relate Hilterman and Verm (1995) equation to Shueys (1985):

So we see that PR=(A+B) where B is the gradient and A is NI.

R ( ) = NI (1 sin 2 ) + PR sin 2
R ( ) = NI + ( PR NI ) sin 2
R ( ) = NI + (( NI + B ) NI ) sin 2

R ( ) = NI + B sin 2
If we look at the equation

R( ) = NI cos 2 + PR sin 2 we see that at an angle of incidence

of 0 degrees it is predominately the Acoustic Impedance while at an angle of incidence of 45 degrees it is


a mixture of the Near Impedance and the Poisson Ratio.
Change in the Poisson Reflectivity could reflect a change in fluid but may also reflect a change in
lithology. It is good to about 30 degrees of incidence angle and it assumes /=2. We will revisit
PR=(A+B) later when we look at AVO attributes.
Understanding Amplitude Versus Offset through simplified modeling
To understand the behavior of the amplitudes with offset we utilize Hilterman and Vern (1995)
and the Vp, Vs, and for the wet and gas cases for all three Rutherford and Williams (1989) classes of
AVO sands and graph the amplitude versus the angle of incidence (figure 10).
If we analyze the results we change from Class 3 AVO sand to a class 1-AVO sands the acoustic
impedance increases in value going from a trough (negative) passing through zero and going to a peak
(positive). The slope or rate of change in amplitude also increases from Class 3 AVO sand to class 1-AVO
sands. The slope seems to be related to the porosity of the rock, the higher the porosity the lower the
slope.
The comparison of the wet sand versus the gas sand shows the near impedance drops in
magnitude between the two respectively. This means the fluids in the rock affect the near impedance
more than the slope. The slope for the wet and gas trends are almost parallel for the class 1 and class 2
sands.

Figure 10:

Graph of amplitude versus angle of incidence for the three Rutherford and Williams
(1989) classes of AVO sands. These graphs show the fluids in the rock affect acoustic
impedance and the second term generally contains the lithological information.

In summary we see that we change from Class 3 AVO sand to a class 1-AVO sands the acoustic
impedance increases in value going from a trough (negative) passing through zero and going to a peak
(positive). We also see the slope or rate of change in amplitude seems to be related to the porosity of
the rock, the higher the porosity the lower the slope (matrix term) and the near impedance reflects the
fluid content (fluid term).

What is Amplitude versus Offset


Amplitude versus Offset or AVO is this change of amplitude within the CDP gather. We see the
slope or rate of change in amplitude seems to be related to the porosity of the rock and the near
impedance reflects the fluid content. Why this phenomenon is occurring is because the P-wave velocity
is sensitive to fluid changes while Shear wave velocities are not. If we have a gas in the rock we have a
decrease with the P-wave velocity but the Shear wave velocity remains roughly the same. With nonnormal angles of incidence we create shear wave in the data. This change in Vp/Vs creates the AVO or
the anomalous behavior of amplitude with angle. Vp/Vs is proportional to Poisson Ratio so when we
have changes in Vp/Vs we will see change in terms of Poisson ratio.
AVO began with the realization of bright spots within the seismic data particularly the stack
data. At this time we were utilizing Dip Moveout and Migration to properly image the data. This allowed
us to work only with stacks. With the advance of Prestack Time Migration we began to work with CDP's
and do Prestack analysis on Prestack Time Migrated gathers.

As mentioned above Acoustic Impedance or AI is one of the main drivers of AVO. When we
describe a class of AVO we define it by the acoustic impedance between the encasing shales and the
reservoir sand.
1)

Class 1 AVO sands are higher impedance sands than the encasing shales. These are
generally Mesozoic and Paleozoic sands (Roden et al., 2005). The seismic response for
the AVO gas sand is illustrated in figure 7B. Noticed on the upper reflector we have a
peak (positive amplitude);

2)

Class 2 AVO sands are equal in acoustic impedance as the surrounding shales. These
moderated compacted sands (porosity 15% to 25%) are found offshore and onshore.
Gross interval velocities of the sands are typically between 8,500 and12,500ft/s (2,650
3,650m/s) (Roden et al., 2005). The seismic response for the AVO gas sand is illustrated
in figure 7D. Noticed on the upper reflector we have a small trough almost zero in
amplitude (small negative amplitude);

3)

Class 3 (bright spots) is the classical bright spots where the gas sands have impedance
lower than the encasing shales. These reservoirs commonly have porosities >25% and
contain gas or high gas-to-oil ratio (GOR) oils. The gross interval velocity of these sands
is usually <8,500ft/s (2,650m/s) (Roden et al., 2005). The seismic response for the AVO
gas sand is illustrated in figure 7F. Noticed on the upper reflector we have a strong
trough almost zero in amplitude (strong negative amplitude).

Figure 11:

Modeling of the amplitude response at different angles of incidence for the different
Rutherford and Williams (1989) classes of reservoir sands taken from Schulte (2007).
The wet sands do not display anomalous responses but on the gas sand cases we have a
brightening or increase in the magnitude of the amplitude.

Summary of Amplitude Versus Offset


Most interpreters within the industry utilize either Hilterman and Verms (1995) NI-PR or
Shueys (1985) A-B for the AVO attributes from their respective linearized equations. We see that the
linearized AVO equations follow the trend of a fluid term, which is the P-wave reflectivity, or zero offset
stack and a matrix or lithology term which is the second term. Utilizing the graph of the Angle of
incidence versus the Reflectivity we can make some statements about the intercept (A) and gradient (B)
and the Poisson Reflectivity (PR) for the different classes of AVO.

Table 2:

Table 3:

AVO Class

NI

PR

Class 3 Gas
sand

Large negative

Small negative

Large negative

Class 2 Gas
sand

Small negative/positive

Large negative

Large negative

Class 1 Gas
Sand

Small positive

Large negative

Large negative

AVO attributes responses for gas sand.

AVO Class

NI

PR

Class 3 Wet
sand

Small negative

Small negative

Small negative

Class 2 Wet
sand

Less in magnitude than


gas sand

Large negative

Smaller in magnitude than


Gas sand

Class 1 Wet
Sand

Larger positive than gas


sand

Large negative

Smaller in magnitude than


Gas sand

AVO attributes responses for wet sand.

We can also make conclusions about how the different classes of AVO will be affected by
changes in porosity or age of the rock, or depth or when we look at the AVO our we looking at the
affects of fluids in the rock (fluid term) or the effects of matrix (matrix term).

Figure 12:

Graph of the different classes of AVO and the associated properties such as the effect of
fluids, porosity, depth and age. The fluids in the rock mainly cause a Class 3 AVO; the
porosity of the rock is large almost like unconsolidated clean sand; the age of the rock is
relatively young; and the depth is shallow. The changes in the class 1 AVO on the other
hand are due to changes within the rock matrix such as changes in porosity rather than
changes than changes in fluid content; the porosity is small so it is a tighter reservoir;
the rock is relatively older and it is deeper in the section.

Affect of Phase of the Gathers on the Intercept and Gradient


To calculate intercept (A) and gradient (B) we want the seismic data to be zero phase data. This
is an important criterion for AVO work. A zero phase wavelet is a symmetrical wavelet with the apex of
the peak at the rock interface (figure 13).

Figure 13:

Zero phase wavelet.

How we determine this initially is by tying either a synthetic seismogram or a VSP to the seismic
data and rotate the data until we have a good cross correlation. Cross correlation is a measure of
similarity of two waveforms as a function of a time lag applied to one of them. When we have a good
cross correlation between the seismic and the synthetic seismogram then we should have zero phase
data.

The phase rotation should be done to the gathers and AVO attributes recalculated. This will then
allow for better AVO attributes to be generated.

AVO in Stacked Data


Though we have described what the gathers or prestack data looks like for the different classes
of AVO most interpreters work with Near, Far and Full offset stacks in their interpretation.

Figure 14:

Modeling of the different stack responses for the 3 classes of AVO for the different fluid
types taken from Schulte (2007).

Class 1 AVO sands are known as dim out because we begin with a strong positive peak and go to a
trough on the far offsets. When we stack this together we have cancellation and the amplitude on
the full stack is almost 0 for a gas sand. The near impedance is a peak for the Class 1 sand and the
fars are a small trough.
Class 2 AVO sands have a small peak or trough for the near offsets increasing in magnitude with the
far offsets. This causes the stacking response on the full offset stack and the near offset stack to be
weak, and the far offset stack has a strong trough.
Class 3 AVO sands are known as the bright spot. The full offset stack, near offset stack, and far offset
stack are strong in amplitude with the far offset stack being stronger in amplitude. The change
between the nears and the fars is not as dramatic as the other classes of AVO. Most of our

AVO/amplitude technology is designed around the class 3 AVO. The bright spot/class 3-AVO sand
has become synonymous with what most consider as AVO. We can utilize the stack volumes in order
to build attributes that will illuminate AVO anomalies. The most common AVO attribute from stack
data is the far offset stack. We can see from the different stack responses that when we have gas for
Class 3 and Class 2 AVOs we have a strong negative (trough) response but when the rock is wet we
have a small negative response (trough) response.
Since we are measuring the difference in amplitudes between nears and fars many think Far
minus Near may be a good attribute but the problem with a class 3 is there is little change between the
nears and fars so we create another attribute referred to as the Enhanced Restrictive Gradient (ERG)
which is ((Far-Nears)*Far) (Barton and Gullette, 1996) which causes the ERG to brighten up when we
have class 2 or 3s. By multiplying the (Fars Nears) by the Fars we are enhancing or emphasizing the
Fars. The best way to utilize this attribute is to use the envelope of the energy for the Nears and the Fars
then we have the absolute values.

Figure 15:

Modeling a class 3 AVO. The orange is the part that will be summed up to produce the
near offset stack and the green will be the far offset taken from Schulte (2007). Notice
there is little difference between the Nears and the Fars and if there was to be
subtraction of the Nears and Fars it will be close to zero.

Figure 16:

Modeling a class 2 AVO. The orange is the part that will be summed up to produce the
near offset stack and the green will be the far offset taken from Schulte (2007).

Class AVO

Near

Far

Erg

Class 3 Gas
Sand

Large Negative

Larger Negative

Large Positive

Class 2 Gas
Sand

Small Positive or
Negative

Large Negative

Large Positive

Class 1 Gas
Sand

Positive

Small Negative

Small Positive

Table 4:

Stacking responses for the Near, Far and ERG.

AVO Attributes, Crossplots and Prestack Gathers


And how to apply these concepts into interpretation
We have looked at the theory of AVO and utilized simplified modeling in order to understand
what the Intercept, Gradient and Poisson Reflectivity (PR) will look like for the different Rutherford and
Williams (1989) classes of gas sands. We will now look at what can we utilize for interpretation of
reflective data and establish guidelines to be utilized in the interpretation of AVO. AVO is not a direct
hydrocarbon indicator. The only known DHI is a flat spot. The AVO indicates we have anomalous
amplitudes and we need to look further into what is causing these anomalous amplitudes with prestack
gathers and well modeling.

Figure 17:

Example of a phase reversal and a flat spot in seismic data. As shown the phase
reversal and flat spot are due to the drop of the P-wave velocity due to
hysrocarbons.

AVO Attributes
We have discussed what the intercept (A), gradient (B) and Poisson Reflectivity (PR) look like for
the different Rutherford and Williams (1989) classes of gas sands. We will know look at how we can
utilize them in interpretation and rules to govern their interpretation.
Factors which may affect the linear relationship between intercept and gradient are
(Whitcombe et al, 2004):
1) Residual velocity;
2) Third Term Move out;
3) NMO stretch (Keho et al, 2001);
4) Multiples;
5) Gain;
6) And overburden effects.
If we do not compensate for these factors then there may be issues in the AVO analysis and the
attributes maybe compromised with leakage of the fluid term into the matrix/lithology term (Hermann
and Cambois, 2001; Cambois, 1998; Cambois, 2002). These factors can cause a large difference in
magnitude between the fluid term and the matrix term which may affect our AVO attributes and
crossplots. All but one of these factors can be taken care of in what we refer to as gather conditining for
AVO which takes the Kirchhoff Prestack Time Migrated gathers and attempts to flatten them with
residual velocity and third term move out and remove any residual noise. We may also run radon again
to remve any residual multiples.
To compensate for the difference in the Intercept and gradient in magnitude we f we normalize
A and B then Vp/Vs=2. This will also take care of the overburden affect.
Hydrocarbon Indicator (HCI) Swan (1993)
The hydrocarbon indicator is (A*B). It works best when we have a negative intercept and a
negative gradient. When we have a Class 1 AVO gas sand we have a positive intercept and a strong
negative gradient which will produce a strong negative hydrocarbon indicator.

AVO Class

Table 5:

Intercept

Gradient

AB

(A)

(B)

positive

Strong
negative

Strong negative

2P

Small positive

Strong
negative

Strong negative

Small negative

Strong
negative

Strong positive

Large negative Small negative

Strong positive

Hydrocarbon indicator responses for the different classes of AVO.

Fluid Factor

Figure 18:

Diagram depicting fluid factor.

The Fluid Factor is a combination of A and B which will produce a zero result for the background
trend. If we assume

= 2 and =a.25 so
= .25
then F = 1.252 A + .58B .and this equation fits

the form Ak1+Bk2 where kn are constants. If we normalize A and B then the fluid factor becomes A+B.
Both A+B and RP-2RS have the following characteristics that we require in order to make an AVO
attribute a good attribute is:
1)

In nonpay zones approximates zero (Castagna and Smith, 1994);

2)

Almost always negative for shale over gas sands intervals and significantly more negative than
for shale over brine sands (Castagna and Smith, 1994).

Now if we look at Hilterrman and Verm (1995) equation we see that Poisson Reflectivity equals A+B. So
A+B is a scaled Poissons Ratio which reflects / which reflects water saturation. Personal
experience has shown A+B is a robust attribute.
Fluid Factor, Far Offset Stack and Elastic Impedance

2
The Fluid Factor = A + B sin

where f is the fluid factor angle (Smith and Gidlow,

2003). If the angle of incidence of the far offset stack approximates the fluid factor angle then the far
offset stack approximates a fluid factor section. This explains why some see the Far Offset stack
resembling the Fluid Factor. In some companies the Far Offset Stack is utilized as an AVO attribute. In
elastic impedance we invert the far offset stack and if this far offset stack approximates a fluid factor
section we are inverting the fluid factor section.

AVO Crossplots
Cross plotting enables the simultaneous and meaningful evaluation of two attributes be they
reflective attributes or inversion attributes (Chopra, Alexeev, and Xu, 2003). The goal of utilizing these
two attributes is to isolate lithology and fluid types into clusters allowing for the identification of
background that is composed of reflections (Chopra, Alexeev, and Xu, 2003) from water bearing
sandstone and shaly sequences and anomalous points that may represent oil and gas bearing rocks.
The power behind AVO analysis is within the interactive cross plot. We can cross plot various
attributes to see what is anomalous and project it back onto any seismic section and indexing the
gathers that produced the seismic section.
We need to look at where the AVO anomaly cross plots compared to the well data. If the AVO
has been determined to be a Class 2 then it should occur in the zone where we have a class 2. If the
seismic and the well modeling do not correspond then there are issues in the seismic data, which we will
discuss about in a later section, that are causing leakage of the intercept into the gradient. These issues
will also affect any inversions we attempt to do on the data.

Figure 19:

Diagram comparing the raw PSTM gathers which have issues in the seismic data and the
conditioned gathers which have been processed to correct some of the issues within the
Seismic Data.

By relating the intercept-gradient cross plot, seismic section and gathers together we can
understand the anomalies better. The tendency to ignore the Prestack Data may cause us to drill on
false anomalies. The majority of the interpreters attempt to reduce the data and disk space for their
interpretation and utilize just stack sections in their AVO analysis foregoing analyzing the Prestack
Gathers.
Theory of the Cross Plot

Figure 20:

Diagram of the mechanisms of the AVO cross plot. The V p/Vs or background lines
depends upon the fluid type (Castagna, Han and Batzle, 1995) so the AVO responses for
gas sands must fall upon the same linear trend. Their position on the linear trends
depends upon the acoustic impedance contrast with the surrounding rocks (Foster and
Keys, 1999).

The Vp/Vs rato depends upon the fluid type (Batzle, Han and Castagna, 1995) so the AVO
responses for gas sands must fall upon the same linear trend. Their position on the linear trends
depends upon the acoustic impedance contrast with the surrounding rocks (Foster and Keys, 1999). The
acoustic impedance is dependent upon factors such as fluid and porosity. As mentioned before changes
in pore fluids for the top of the sand can be see with a decrease in the Intercept or movement to the left
in the intercept-Gradient cross plot. Increases in porosity also causes the Intercept to decrease or shift
to the left. The question is does the anomalous response reflect a change in porosity or a change in fluid.
The best way to answer this is to analyze the relationship of the anomaly to mapped structure ( Simm,
White, and Uden, 2000).

Figure 21:

Cross plot of Near Impedance versus Poisson Reflectivity for the different Rutherford
and Williams (1989) classes of AVO sands. The top of gas sand reflections tends to plot
below the background trend and the bottom of gas sand reflections plot above the
background trend (Castagna et al., 1998). The background trend is related to the V p/Vs
ratio. Class 1 (A) has a positive intercept and negative gradient so it plots in the 4th
quadrant of the Cartesian axis (light brown trapezoid). Class 2 (B) has a positive or
negative intercept and a negative slope so it plots around the 0 on the x-axis in
quadrants 3 and 4 (light blue trapezoid) and the Class 3 (C) has a negative intercept and
gradient so it is in quadrant 3 (yellow trapezoid). The arrows in each graph represent the
fluid factor. As can be seen Class 3 and 2 fluid factors are large but the Class 1 AVO has a
small fluid factor.

AVO Well Modeling


Rock physics is different than petrophysics. With rock physics we want to describe rock
properties in terms of physical properties that will affect how seismic waves will pass through them
(Dewar, 2001). The goal is to develop a relationship between the rock properties and the observed
seismic response and then develop how we can detect these properties seismically so we can utilize it
within interpreting the seismic data for prospects. We do this by understanding the elastic properties of
the pore fluid and the rock frames and develop models through various equations such as BiotGassmann, or the creation of pseudo-gathers to understand the rock-fluid interactions (Dewar, 2001)
Petrophysicists are interested in developing physical properties such as water saturation,
porosity, permeability, volume of shale, etc which are related to production parameters. Petrophysicists
are not concerned about seismic but about wellbore measurements to contribute to the reservoir
description and some of their findings may be useful to the rock physicists (Dewar, 2001).
Rock physicists can first take the well data and cross plot it so that we can see the sensitivity to
petrophysical properties such as acoustic impedance versus porosity, or Vp/Vs versus water saturation.
If there is no sensitivity to petrophysical properties at this scale then the seismic will provide little or no
value to show petrophysical properties.

Figure 22:

Seismic interpretation involves the integration of well data, petrophysics, AVO


modeling, and geology.

One of the questions we have in AVO analysis is: what is causing the AVO anomaly we see in the
seismic data. Is this being caused by a change in fluid or a change in the rock such as the increase in
porosity and how do these changes manifest themselves in the seismic response. If the changes in the
rock are manifest with changes in the fluid content can we utilize a 4D seismic survey to begin to under stand what is incorrect in the reservoir model similar to how engineers utilize production history match ing to look for improvements in the reservoir model (Pennington, Acevedo, Haataja, and Minava, 2001).

One of the pitfalls of AVO is the anomaly is caused by high porosity clean wet sands. Sometimes
these high porosity wet sands can produce amplitudes similar to hydrocarbon sands. Knowledge of well
control in the area, and Biot-Gassmann fluid substitution modeling can help us with this problem
(Roden, Forrest, and Holeywell, 2005).

Figure 23:

Work flow for seismic modeling from Bulloch (1999).

AVO Inversion
Seismic inversion is a non-unique process which involves transforming seismic reflection data
into a quantitative rock property such as Acoustic Impedance or Elastic Impedance. It is non-unique
because elastic parameters are what we can measure with the seismic which are; , , and and these
properties are influenced by many rock and fluid variables. Hence, any predictions made about the rock
properties from elastic parameters will be ambiguous.

The benefiits of seismic inversion are:


1) Seismic inversion brings the seismic AVO attributes over to physical properties that can be
compared directly with the well log data (Hansen et al, 2000). We can interpret the inversion on
a gross regional basis or on a detailed reservoir scale;
2) Broader band due to lower frequencies in data and reduction of the wavelet effects, tuning and
side lobes creating higher frequencies and reducing the effects of wavelet tuning in the data;

3)
4)
5)
6)
7)

Wavelet represents rock (geological) layer and not rock interfaces;


Modeling and inclusion of layer stratigraphy;
Calibration to well data;
Attenuation of random noise;
And, improved interpretability of seismic horizons

Figure 24:

Diagram showing the difference between a reflective wavelet and an inverted wavelet.
With reflection seismology and zero phase data the rock interface is at the apex of the
peak. When we invert the data the zero crossing becomes the top and bottom of the
reflector.

Most oil and gas companies now use seismic inversion to increase the resolution and reliability
of the data and to improve estimation of rock properties including porosity and net pay.

Figure 25:

Diagram showing the workflow of the data for inversion.

Limitations to Inversion
1) Knowledge of the seismic wavelet;
2) Inversion impedances must be consistent with impedance model constructed from well logs;
3) AVO attributes intercept, gradient and curvature are statistically correlated and can be affected

by the noise in the data so instead of inverting the AVO attributes we can invert the partial
stacks.

AVO Interpretation/Conclusion
One of the first steps is to do well modeling to understand what may be apparent within the
seismic. We need to be able to utilize Biot-Gassmann fluid substitution to understand changes in the
gather due to lithology and fluid changes. We then need to utilize seismic crossplotting of attributes
such as intercept and gradient, and multiple seismic AVO attributes to identify areas where we have
anomalous amplitudes. We then need to look at the Prestack gathers across this area to determine that
these anomalous amplitudes were not caused by processing. When we look at these anomalous
amplitude zones we are looking for AVO anomalies that appear to be (Roden et al., 2005):
a. Consistent within the mapped target area on the stacked data area;
b. Match the geology especially down dip (down dip conformance);
c. Few unexplained anomalies;

d. Are consistent with direct hydrocarbon indicators such as flat spots in the seismic;
e. And, match decreases in the interval velocity.
The main points we are trying to prove in exploration with the interpretation of seismic data are:
1) What is the depositional setting, source rocks, migration pathways, reservoir rock and seal.
2) What is the timing and is it conducive to the formation of a reservoir.
3) What are the extents of the reservoir (AVO).
4) Examination of synthetic seismograms (well) ties with seismic and top of formations.
Justification of the phase and statement of the phase.
5) Is there closure, does the AVO match the closure, is there possibility of leaking.
6) Is there other information such as flat spots or phase reversal, which may prove there are
hydrocarbons present. Flat spots are created by fluid contacts and appear to be going against
the structural trend.
7) How does the well models tie into the seismic.
8) How does the AVO tie into production history (wet, gas and oil) from previous wells.
9) What is the size of the hydrocarbon column, and the thickness of the reservoir sand.
10) What is the net to gross.
11) Is this economical estimate based upon what we know and the margins of error.
12) What are the risks for drilling and is it feasibility to drill this.
The attributes we utilize should be substantiated by the well modeling; the phase should be
described through explanations why it was chosen; the attributes should match the geology; they should
conform to the structure; and all of this should tell a neat story where we can then analyze risk and
decide which wells to drill first. AVO anomalies that are random or fit pore pressure need to be defined
and explained. The presentation should bring together all the well work, modeling, pore pressure work
and inversions.
Many interpreters utilize the attributes to prove a reservoir but the attributes should be used to
illuminate and discriminate the size of the reservoir and not prove the reservoir. The proof in the
reservoir comes in linking all the elements together and then insure that we tie back to the geology. The
attributes need to be chosen to highlight the above questions. The interpreters need to understand the
acquisition and processing and how it affected their data. By doing weekly reporting during the
acquisition and processing and showing improvements on the data the interpreter can defend the data.

All the data from well logs to AVO attributes to inversions need to be utilized to build the case why we
should drill the prospect.
Prestack Gather
Conditioning

Geophysics Stacks and


Gathers

AVO
Attributes

Interval
Velocities

Practical

Cross Plots
Geology

Interpretation System

Horizons and
faults

Synthetics, Fluid Substitutions and Cross Plots


Engineering
Production
history

Well Logs

Theoretical

Decisions on
drilling
Integration of all disciplines

Figure 26:

Seismic interpretation involves the bringing together of a lot of information into


an interpretation system to best integrate the information into the
interpretation. Most interpreters focus on the stacks and seldom utilize all the
information they can to enhance their interpretation.
Bibliography

Aki, K., and Richards, P.G., 1980, Quantitative Seismology: Theory and Methods: W.H. Freeman and Co.
Barton, J., and Gullette, K., 1996, Reconnaissance AVO techniques in the Niger Delta (ERG attribute),
AAPG, Caracas, Venezuela.
Batzle, M., and Wang, Z., 1992, Seismic properties of pore fluids, Geophys., 57, 1396-1408.
Bortfeld, R. (1961) Approximations to the reflection and transmission coefficients of plane longitudinal
and transverse waves. Geophys. Prosp., 9, 485-502.
Bulloch, T.E., 1999, The Investigation of Fluid Properties and Seismic Attributes for Reservoir
Characterization, Master Thesis Michigan Technology University
Burianyk, M., 2000, Amplitude-vs-Offset and Seismic Rock Property Analysis: A Primer, CSEG Recorder
November 2000.

Cambois, G., 1998, AVO attributes and noise: pitfalls of crossplotting, SEG Abstract 1998.
Cambois, G., 2002, A new approach to the fluid factor leads to elastic inversion without shear log, Soc.
Expl. Geophys., Expanded Abtracts.
Castagna, J. P., and Smith, S.W., 1994, Comparisons of AVO indicators: A modeling study, Geophysics
Vol. 59, No. 12 P. 1849-1855
Castagna, J.P., Han, D., and Batzle, M.J., 1995, Issues in rock physics and implication for DHI
interpretation, The Leading Edge
Castagna, J.P., Swan, H.W., and Foster, D.J., 1998, Framework for AVO gradient and intercept
interpretation, Geophysics, Vol. 63, P. 948-956
Chopra, S., Alexeev V., and Xu, Y., 2003, Successful AVO and Cross-plotting CSEG Recorder November
2003.
Dewar, J., 2001, Rock Physics For the rest of us an Informal Discussion, CSEG Recorder May, 2001, p 4250
Foster, D.J., and Keys, R.G., 1999, Interpreting AVO responses: 69th Ann. Internat. Mtg., Soc. Expl.
Geophys., Expanded Abstracts, 748-751.
Gardner, G.H.F., Gardner, L.W., and Gregory, A.R., 1974, Formation velocity and density - The diagnostic
basis for stratigraphic traps, Geophys., 39, 770-780.
Gassmann, F., 1951, Uber die Elastizitt Porser Medien, Vier. der Natur. Gesellschaft in Zrich, 96, 1-23.
Herrmann, P., and Cambois, G, 2001, Statistical properties of seismically, derived AVO attributes, Soc.
Expl. Geophys., Expanded Abtracts.
Hilterman, F., and Verm, R., 1995, Lithology color-coded seismic sections: The calibration of AVO crossplotting to rock properties, The Leading Edge, 14, 847-853.
Keho, T., Lemanski, S., and Raja, B., 2001, The AVO hodogram: Using polarization to identify anomalies,
The Leading Edge, November, p.1214-1224
Koefoed, O., 1955, On the effect of Poisson.s ratios of rock strata on the reflection coefficients of plane
waves: Geophys. Prosp., 3, 381-387.
Pennington, W.D., Acevedo, H., Haataja, J.I., and Minava, A., 2001, Seismic time-lapse surprise at Teal
South: that little neighbor reservoir is leaking!. The Leading Edge.
Ostrander, W.J., 1984, Plane-wave Reflection Coefficients for Gas Sands at Non-normal Angles of
Incidence: Geophysics, 49, 1637-1648.

Roden, R., Forrest, M., and Holeywell, R., 2005, The Impact of Seismic Amplitudes on Prospect Risk
Analysis, The Leading Edge, July 2005; v. 24; no. 7; p. 706-711
Roden, R., Jones, G., Castagna, J., 2005, The impact of pre-stack data phase on the AVO interpretation
workflow, SEG/Houston 2005 Annual Meeting
Rutherford, S.R., and Williams, R.H., 1989, Amplitude versus offset variations in gas sands, Geophysics,
Vol. 54, P 680-688
Schulte, B.W., 2007, Bridging Seismic (AVO) Prestack Interpretation and Well Modeling, Geophysical
Society of Houston lunch
Shuey, R. T., 1985, A simplification of the Zoeppritz equations: Geophysics, 50, 609-614.
Simm, R., White, R., and Uden, R., 2000, The anatomy of AVO crossplots, The Leading Edge, February
2000.
Smith, G. C., and Gidlow, P. M., 1987, Weighted stacking for rock property estimation and detection of
gas: Geophys. Prosp., 35, 993-1014.
Smith, G.C. and Gidlow, P., 2003, The fluid factor angle and the cross plot angle, 73nd Ann. Internat.
Mtg: Soc. of Expl. Geophys.,
Swan, H. W., 1993, Properties of direct AVO hydrocarbon indicators, in Castagna, J. P., and Backus, M.
M., Eds., Offset-dependent reflectivity - Theory and practice of AVO analysis: Soc. Expl. Geophys., 78-92.
Taner, M.T., and Koehler, F., 1969, Velocity spectra-digital computer derivation and applications of
velocity functions, Geophysics, Vol. 34, 859-888.
Whitcombe, D.N., Dyce, M., McKenzie C.J.S., Hoeber, H., 2004, Stabilizing the AVO gradient, Soc. Expl.
Geophys., Expanded Abstracts.

You might also like