You are on page 1of 10

CSIRO PUBLISHING

Exploration Geophysics, 2014, 45, 154–163


http://dx.doi.org/10.1071/EG13070

A description of seismic amplitude techniques

James Shadlow
Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Kufpec Australia Pty Ltd, 7/100 Railway Road, Subiaco, WA 6008, Australia.

Email: james.shadlow@kufpec.com.au

Abstract. The acquisition of seismic data is a non-invasive technique used for determining the sub surface geology.
Changes in lithology and fluid fill affect the seismic wavelet. Analysing seismic data for direct hydrocarbon indicators (DHIs),
such as full stack amplitude anomalies, or amplitude variation with offset (AVO), can help a seismic interpreter relate the
geophysical response to real geology and, more importantly, to distinguish the presence of hydrocarbons. Inversion is another
commonly used technique that attempts to tie the seismic data back to the geology. Much has been written about these
techniques, and attempting to gain an understanding on the theory and application of them by reading through various journals
can be quite daunting. The purpose of this paper is to briefly outline DHI analysis, including full stack amplitude anomalies,
AVO and inversion and show the relationship between all three. The equations presented have been included for
completeness, but the reader can pass over the mathematical detail.

Key words: amplitude variation with offset (AVO), direct hydrocarbon indicators (DHIs), seismic inversion.

Received 15 August 2013, accepted 20 January 2014, published online 27 February 2014

Introduction impedance of the earth. This can in turn be used to derive seismic
The first use of seismic was to identify potential hydrocarbon traps volumes of other rock properties.
by looking purely at the structures imaged. It was not until the late
1960s that geophysicists began commonly using a relationship The seismic experiment – reflectivity
between seismic amplitude brightening and structure for clastic
Seismic data is recorded by stimulating the earth with an acoustic
reservoirs (Chopra and Marfurt, 2007; Hilterman, 2001). This
source at some point A and recording the waves at some point B
marked the start of direct hydrocarbon indicator (DHI)
after they have travelled through the earth. The waves travel
identification and analysis. Many wells were drilled based on
through the earth, giving off a series of incident, reflected and
the identification of ‘bright spots’ without properly understanding
transmitted waves (Figure 1). The first thing to note here is that the
the seismic response or relating it back to the geology. This
resulted in many wells being drilled on seismic anomalies caused
by anomalous volcanic sediments or shales, as the geological
interpretation was inaccurate (Brown, 2005). At the same time as
REFLECTED
DHI analysis was becoming widely used, more detailed analysis S-WAVE
of the seismic amplitude phenomenon was performed. This INCIDENT
Rps
included analysis of the amplitude variation that is present P-WAVE REFLECTED
φ1 P-WAVE
with the change in angle or offset in the seismic gathers.
Koefoed (1955) first showed how amplitudes vary with offset Rpp
Medium 1
at a boundary. Many others including Castagna and Backus Vp1, Vs1, ρ1 θ1
(1993), Ostrander (1984) and Shuey (1985) have studied this
variation both empirically and mathematically. Their work
provided the basis for amplitude variation with offset (AVO) Interface
analysis, whereby changes in the amplitude with offset can be
used as a lithological and fluid fill discriminator. This method also θ2
needs to be applied with an understanding of the subsurface Medium 2
TRANSMITTED
geology otherwise the anomalies may be misinterpreted. Vp2, Vs2, ρ2
φ2 P-WAVE
The methods discussed to this point are used to specifically Tpp
identify hydrocarbons, but indirectly give indications of the
geology. The ultimate end point of relating the geophysics to
the geology is inversion, which provides a detailed image of the
TRANSMITTED
subsurface geology and can also be used to infer the presence S-WAVE
of hydrocarbons. Seismic inversion is carried out by reverse Tps
engineering the recorded seismic data. With the help of well
logs, a model of the subsurface is constructed which together Fig. 1. Schematic showing the incident, reflected and transmitted waves.
with an estimated wavelet is used to image the actual acoustic Modified from Castagna and Backus (1993).

Journal compilation  ASEG 2014 www.publish.csiro.au/journals/eg


A description of seismic amplitude techniques Exploration Geophysics 155

sum of the reflected and transmitted waves equals the incident Bright spot
(reservoir Al less than
wave (equation 1): surrounding shale)

R þ T ¼ Incident wave ð1Þ


This relationship states that the overall energy of the wave
cannot be increased as it goes through various layers. Also, the Phase change
energy ‘lost’ from the wave due to attenuation is not physically (reservoir Al similar than
surrounding shale)
lost, but used up by moving the particles in the earth – no energy
Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

is created or destroyed as the wave travels through the earth.


Trough (decrease in Al)
The amplitude or energy of the reflected and transmitted wave
is related to the density and P-wave velocity change between Peak (Increase in Al)
Dim spot
the two layers. The bigger the contrast between the initial layer Increasing reflector strength
(reservoir Al greater than
due to increase Al contrast
and the reflecting layer, the more energy is reflected rather than surrounding shale)
transmitted. The data loses overall amplitude due to attenuation Water leg

with depth; the proper representation of this phenomenon is Water leg


only possible with seismic data which is processed to preserve
amplitude.
When amplitudes were originally analysed, they were
assumed to be zero offset, meaning that a source at some point
A is co-incident with a receiver at some point B. The zero offset
reflectivity of each geological layer can therefore be described by Fig. 2. Different types of DHIs. Modified from Brown (2004).
the zero offset Zoeppritz equation (equation 2).
r2 VP2  r1 VP1
RP ¼ ð2Þ Reservoir acoustic impedance Reservoir acoustic impedance
r2 VP2 þ r1 VP1 higher than overburden lower than overburden
where RP = P-wave reflectivity; r1 = density of the first layer; Bright spot
r2 = density of the second layer; VP1 = P-wave velocity of the first Typically young clastic
reservoirs
layer; and VP2 = P-wave velocity of the second layer.
Dim spot
Observe that equation 2 is specifically written for the P-wave Typically old reservoirs/
reflectivity. At zero offset, it is physically impossible to produce carbonates
a shear wave because shear waves move the rock particles
orthogonal to the direction of wave propagation, and as such Phase change
need to interact with the rock particles at an angle less than 90. Hydrocarbon-bearing reservoir (little difference in
Water-bearing reservoir hydrocarbon and water
The seismic wave front moves through the earth as a zone of reservoir)
compression, which has the starting phase of a peak. When the
compressional front passes from one geological layer to another, a
Phase change/bright
reflected wave is generated with the phase of either a peak or a spot combination
trough. If the reflecting layer is harder than the layer that the wave Increasing positive impedance Increasing negative impedance

was initially travelling in, then a peak will be reflected back. If


Fig. 3. The change in acoustic impedance (AI) for the different direct
the reflecting layer is softer than the initial layer then a phase hydrocarbon indicators (DHIs). All DHIs are due to a decrease in AI. The
change will occur due to rarefaction and the reflected wave will movement from left to right demonstrates the change in reservoir properties
be a trough (SEG Normal Polarity). This can easily be shown by from water filled properties to hydrocarbon filled properties. Modified from
evaluating equation 2 numerically. Layer 1 is harder than layer 2 if Brown (2004).
the impedance (i.e. product of velocity and density) is greater in
layer 1 than that of layer 2.
density, low velocity) hydrocarbon fill to higher impedance
(higher density, higher velocity) water fill. They do not
DHIs necessarily have to be flat, as lateral velocity variations in the
DHI identification uses the knowledge that the introduction of oil overburden may distort the time image, but flatspots are usually
or gas into a rock results in a decrease in density and P-wave unconformable. In some cases, flatspots can also represent paleo-
velocity. The introduction of hydrocarbons into a water-filled contacts. An example of this can occur in carbonates, where a
reservoir produces different seismic effects depending on the residual gas contact in combination with acidic ground water can
acoustic impedance contrast between the reservoir and result in the diagenesis of a ‘hard ground’ (Brown, 2004).
surrounding rocks. If the water filled reservoir is surrounded
by higher impedance rocks, the amplitude of the trough associated Pitfalls of DHIs
with the reservoir brightens with the introduction of The most important factor affecting the correct interpretation of
hydrocarbons. In a water-filled reservoir that is surrounded by a DHI is the phase of the data. If the phase is 180 different to
lower impedance rocks, dimming of the peak associated with the what the interpreter thinks it is, then a hard seismic event, such as a
reservoir will occur with the introduction of hydrocarbons. In this volcanic intrusion, limestone or conglomerate could be targeted
latter case, a phase change may be observed if the introduction of for drilling rather than a bright gas bearing sand. This leads to a
hydrocarbons reduces the reservoir impedance to be less than that second factor, which is the geological understanding of the
of the surrounding rocks (Figures 2 and 3). Flatspots are a slightly reservoir and what type of DHI is expected. Not all bright
different type of DHI, in that they relate to a change of fluid fill at spots are due to hydrocarbon filled reservoirs. If the reservoir
the bottom of a reservoir. Flatspots are always acoustically hard as rock is harder than the surrounding rock, as is the case with most
flatspots are caused by a change from low impedance (low limestones, then a bright spot cannot be used to indicate the
156 Exploration Geophysics J. Shadlow

presence of hydrocarbons. Third, the relationship between the


TR
amplitude anomaly and structure or a stratigraphic pinch-out
is crucial. If there is no relationship between the amplitude Wavelet
Inflection
anomaly and the interpreted structure of the reservoir unit, points B = Wavelet breadth
B/2 = Peak – trough time
then the amplitude is either a stratigraphic effect (e.g. pinch 2T0
2T0 = First zero crossing interval
outs), lithological effect (e.g. coal beds) or due to tuning. This b
is because a fluid contact (gas or oil) is expected to occur at a

Maximum absolute amplitude of composite wavelet


50 2.0
constant depth across a field unless hydrodynamically tilted or
Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

distorted in time due to laterally varying overburden velocities.

Two-way time apparent thickness (ms)


40

Tuning

e
tud
pli
Amplitude effects are not simply related to the lithology or 30

Am
hydrocarbon fill. The seismic response will also ‘tune’, where
the wavelet increases in seismic amplitude through constructive 1.0
interference as the geological beds get closer together to a
20 s
maximum at half-wavelet thickness in two-way time (TWT) nes
ick
(1/4 wavelet thickness in depth). At this point, the peak and b/2 Th
TR TR = λ /4.6 = 1/(3*Fdom)
trough corresponding to the top and base of a geological event are
10 b/2 = λ /4 = 1/(2.6*Fdom)
separated by half a wavelet, so have maximum constructive
interference, resulting in maximum amplitude. a = λ /8.5

If the beds become closer than half-wavelet thickness in TWT,


0 0.0
then the seismic wavelet thickness stays constant but the
0 a 10 20 30 40 50
amplitude decreases until the event can no longer be identified. b/2
Figures 4 and 5 show how amplitude relates to thickness and what Two-way time true thickness (ms)
the seismic response looks like as an event goes through tuning.
Equations 3 (Kallweit and Wood, 1982) and 4 (Brown, 2004) give Fig. 4. Tuning curve for a Ricker wavelet. Modified from Kallweit and
the tuning thickness in time for the Ricker wavelet and sinusoidal Wood (1982).
wavelets respectively (b = trough-to-trough spacing of a Ricker
wavelet, v = velocity, l = wavelength, FDom = dominant frequency). reservoir at maximum amplitude. For example, if a Ricker
   
b 1 v wavelet has a frequency of 35 Hz in sediments with an interval
Ricker wavelet : ¼  ð3Þ velocity of 4000 m/s then the maximum amplitude occurs when
2 2:6FDom 2
    the top and base of a geological event are separated by 11 ms or
1 1 v 22 m. This can then be tied to well data to use the amplitude to
Sinusoidal wavelet : l ¼  ð4Þ
4 2FDom 2 predict the reservoir thickness of thin beds.
To obtain the tuning thickness in depth, it is simply a matter of
multiplying the TWT tuning thickness by velocity and dividing AVO and the reflectivity equations
by two. As such, tuning can be used as a measure of minimum As previously discussed, the presence of hydrocarbons alters
resolution but more importantly indicates the thickness of the the rock properties within a field. This can cause amplitude

Wedge model (Ricker wavelet)


Thickness (m)
–60

–40

–20
Depth from top sand (m)

0 5 10 15 20 25 30 35 40 45
0 Top sand

20

40 Constant thickness
Base sand
Intra-sand resolution
60
ambiguous

80

100

120

Fig. 5. Wedge model showing tuning. Notice wavelet thickness is constant when the layer thickness is less than 20 m.
Modified from Kallweit and Wood (1982).
A description of seismic amplitude techniques Exploration Geophysics 157
Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Fig. 6. Schematic showing seismic data acquisition, its relationship to angle gathers and the AVO response. Modified
from Kunjan (2010).

anomalies, where the amplitude is either enhanced or decreased, Table 1. Relationship between DHIs and AVO anomalies.
depending on the reservoir lithology. One parameter which is
DHI classification AVO class
changed with the introduction of hydrocarbons is the ratio of the
compressional wave and the shear wave, and this change Dim spot (dimming peak with offset) Class 1, 2
translates to a change in reflectivity with offset on the seismic Phase change (peak to trough with offset) Class 2, 3
gathers (Hilterman, 2001). Seismic gathers are the raw data from Bright spot (increasing trough with offset) Class 3, 4
which seismic stacks are created. Full stack seismic data is
Poisson’s ratio:
generated by stacking all the offset traces together (Figure 6).
As such, DHIs can be formed from AVO effects. g2
s¼ ð6Þ
The Zoeppritz equations and its approximations (Appendix A) 2g  2
 2
describe how amplitude varies with offset on the gathers. One
where g ¼ VP
VS
such approximation, the Bortfeld equation (equation 5) (Bortfeld,
1961), shows that the zero angle term is the same as the P-wave AVO classes, relation to geology and intercept
reflectivity, or in other words the zero offset reflectivity (equation
and gradient attributes
2). When amplitude is plotted against offset angle, the zero offset
reflectivity is the intercept (Figure 6). As the angle increases, the There are four main classes of AVO anomalies as classified by
relationship with the shear wave velocity becomes more several authors (Castagna et al., 1985; Rutherford and Williams,
important due to the sin2y term. This defines the slope of the 1989). The most commonly recognised class is Class 3. Full stack
amplitude in Figure 6, and is commonly referred to as the gradient amplitude anomalies are related to lateral changes in the acoustic
term. As defined by the Aki Richards equation (Appendix A) (Aki impedance whereas AVO anomalies are related to changes in Vp
and Richards, 1980), the increase in dependence on the shear and Vs. However, the lateral changes in acoustic impedance that
wave term results in a change in the Poisson’s ratio (equation 6). form full stack amplitude anomalies can be due to changes in Vp
and Vs as a result of hydrocarbon fill. As such, it is possible to infer
Rði Þ ¼ R0 þ Rsh sin2 ði Þ þ RP tan2 ði Þ sin2 ði Þ ð5Þ a relationship between AVO and full stack amplitude anomalies
where (Table 1). However, care must be taken as full stack anomalies
may be caused by other changes, such as porosity preservation
due to hydrocarbon fill. The changes in reflectivity with respect
DVP to offset and associated AVO classes are summarised in Figures 7
RP ¼
2VP and 8.
Class 1 AVO anomalies are the hardest to identify as the
Dr
Rr ¼ reflectivity naturally decreases with offset due to attenuation of
2r the seismic signal. Class 3 is the simplest to identify as the
R0 ¼ RP þ Rr ¼ Intercept amplitude increases with offset, which acts against the natural
  attenuation. Class 4 AVO anomalies are similar to Class 3,
1 DVP Dr DVS whereby the reservoir has a lower acoustic impedance (AI)
Rsh ¼ k  2k ¼ Gradient
2 VP r VS than the seal. However, the change in gradient slope is due
  to different Vs properties (Castagna et al., 1998). The Class 4
2VS 2
k¼ response is generally limited to coals or shallow unconsolidated
VP sands.
158 Exploration Geophysics J. Shadlow

Class 1 Class 2P Class 2

Class 3 Class 4

Amplitude vs angle
0.3
Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Class 1

0.2

0.1
Fig. 8. Intercept–gradient plot showing different AVO responses (modified
from Hampson-Russell (2009)). Class 3 is the most commonly observed
anomaly.
Amplitude (unitless)

Poisson’s ratio change (SPRC) (equation 7), shear reflectivity


Class 2

0.0 (equation 8) and fluid factor (equation 9). These attributes are
0 10 20 30
calculated on the gathers at each sample point on each trace, and
can then be stacked to generate an entire volume (Figure 9).
The SPRC equation results in a seismic attribute that can
distinguish fluid fill, where A and B come from the first two terms
–0.1 of the Aki Richards approximation of the Zoeppritz equations.
Equation 7 is derived assuming that there is a background trend
which is water wet, where Vp/Vs has a value of 2 and the Poisson’s
ratio is 1/3. The shear reflectivity is essentially the shear
impedance (SI), but calculated from the intercept and gradient
Class 3/4

information. The fluid factor, from Fatti’s approximation of the


–0.2
Zoeppritz equations as expressed in Smith and Gidlow (1987),
shows the deviation from the Castagna mudline (Figure 10),
which is due to oil or gas presence in the reservoir. The fluid factor
can also be calculated from the deviation from the mudline on an
intercept–gradient plot. Both the mathematical and graphical
–0.3 methods for the fluid factor are equivalent (Veeken and
Angle (degrees) Rauch-Davies, 2006) (Figure 11). SPRC and fluid factor are
dependent on density, unlike Vp/Vs ratios and Poisson’s ratio
Fig. 7. Castagna’s AVO classifications and how the AI changes with offset. (presented later), which are both independent of density.
Modified from Castagna and Swan (1997).
9
A þ B ¼ Ds ð7Þ
4
The gradient of the curves in Figure 7 can be plotted against the 1
RS ¼ ðA  BÞ ð8Þ
intercept, as shown in Figure 8. Seismic volumes for the intercept 2
and gradient (sometimes referred to as A and B respectively) can VS
be generated by converting each gather into a seismic trace DF ¼ RP  1:16 RS ð9Þ
VP
containing either the intercept or the gradient as the amplitude
information and then stacking the data. From this, hydrocarbon To assist in the analysis of AVO effects, various cross plots
effects can be isolated using the plot in Figure 8. extracted from log data can be created. Examples of these plots
There are many different AVO attributes which can be include:
calculated. All of these attributes attempt to identify the
- P impedance against Vp/Vs
hydrocarbon bearing sediments by either isolating the
- P impedance against porosity
associated decrease in Vp, divide the small Vs increases by
- lm against rm
the Vp decrease or both. The increase in Vs in hydrocarbon
bearing sediments is actually caused by a decrease in rock The first plot can be used to discriminate fluid fill from
density. This is shown later in equation 11 where Vs is lithology. For example, fluid fill can be separated from
inversely proportional to density. porosity effects, as a change in porosity will not significantly
Basic attributes can be calculated based on the intercept and change the Vp/Vs ratio, whilst hydrocarbon effects will mostly
gradient, such as the product (intercept  gradient) or intercept alter the Vp/Vs ratio (Figure 12). The second plot can also be used
minus gradient, which are useful for identifying Class 3 and Class to discriminate fluid fill. Presence of hydrocarbons will result in a
1 AVO responses respectively on seismic. Other more detailed lowering of the P impedance, resulting in a shift in the points on
attributes can also be calculated from Zoeppritz approximations the graph (Figure 13). This effect is more noticeable with higher
(Appendix A). These include, but are not limited to, scaled porosities.
A description of seismic amplitude techniques Exploration Geophysics 159

800 ms TWT DIP LINE – AVO processed


–SPRC anomalies
Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

1000 ms TWT

1200 ms TWT
Anomaly 7 – Yellow = Decrease
in scaled Poisson’s ratio - interpreted
gas sands
5 km

Fig. 9. Example of SPRC attribute seismic volume generated from equation 8 being applied to gathers and then stacked.

Vs – Vp relationship Intercept–gradient plot


6000
Vp = 1.16 Vs + 1360 2.5
Wet trend (shale/water-
5000 bearing sands)
2.0
Top hydrocarbon

4000 Base hydrocarbon


1.5
Vp (m/s)

3000 Siltstone 1.0


Wolfcamp Shale
Oil Shale
Devonian Shale
0.5
2000
Gradient

Pierre Shale
Grayson Shale
Gulf Coast Sediments
1000 Shale Sonic Log –2.5 –1.5 –0.5 0.5 1.5 2.5
Tosayas Clay Point
Quartz
Water
–0.5
Wet trend
0
0 500 1000 1500 2000 2500 3000 3500 4000 –1.0
Vs (m/s)
Divergence from –1.5
Fig. 10. Castagna’s mudrock line. Modified from Castagna et al. (1985). wet trend = fluid
factor –2.0
A more recent method is the plotting of lm against rm,
commonly referred to as lmr (lambda mu rho, LMR) cross- –2.5
Intercept
plotting. In this case, l and m, the Lamé constants, are calculated
from Vp and Vs, combined with density (r) and then plotted Fig. 11. Fluid factor from intercept–gradient plot. The fluid factor attribute is
against each other. l is incompressibility, and m is related to defined by the distance from the wet rock trend to the data points. Points shown
rigidity. lr and mr can be calculated as shown in equations 10–14. are based on synthetic dataset.
Note that mr is the square of the shear impedance,
SI. Hydrocarbons can be identified by points having low lr
ZP2  2ZS2 ¼ lr ð13Þ
values (Figure 14).
From these cross-plot methods and others, hydrocarbons can ZP2 ¼ ðrVP Þ2 ¼ ðl þ 2mÞr ð14Þ
be identified and highlighted through inversion and then cross-
plotting the actual seismic data to highlight anomalies.
Inversion
sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
l þ 2m As discussed at the beginning of this paper, seismic data is the
VP ¼ ð10Þ response measured from an acoustic wave reflecting from a
r lithological boundary. Inversion is the process by which this
rffiffiffi seismic reflectivity can be back calculated into the acoustic
m
VS ¼ ð11Þ impedance values that caused the seismic reflection. This is
r
done by deconvolving the seismic back to the actual rock
ZS2 ¼ mr ð12Þ properties by estimating a wavelet and calculating an operator,
160 Exploration Geophysics J. Shadlow

3.0

PayP or Reef
2.8

WetP or Reef 2.6


Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

2.4

Vp/Vs
PayLow PorR...

2.2

WetLow PorR... 2.0

1.8
TightCarb

1.6

Shale

5e+06 6e+06 7e+06 8e+06 9e+06 1e+07 1.1e+07 1.2e+07 1.3e+07 1.4e+07 1.5e+07

P-Impedance [kg/m^3*m/s]

Fig. 12. Log derived points showing the gas effect (yellow points) can be seen by the red points moving to the yellow
points. The porosity effect can be seen by the carbonate changing along the P impedance axis.

0.5 80
Gas zone
70 Gas
Oil zone sands Cemented/tight Carbonates
0.4
Water zone 60 sands
Mu Rho

Poly. (gas zone)


50
Porosity (v/v)

0.3 log. (oil zone)


Shaly gas sand
Poly. (water zone) 40 Clastics/carbonate
threshold
30
0.2
Shale
20
Porous gas sand cut-off
0.1 0
0 20 40 60 80 100 120 140 160
Lambda Rho
0.0
10 000 15 000 20 000 25 000 30 000 35 000 40 000 45 000 50 000 55 000 60 000
Fig. 14. Well log derived LMR cross-plot. Note that the gas zone is to the left
Acoustic impedance (unitless)
and dependent on the lr term. Modified from Goodway et al. (1997).
Fig. 13. Log derived P impedance plotted against porosity. Note that the
hydrocarbon fill effects shift the points to the left on the P impedance plot.
Modified from Carter et al. (2005). Inversion methods
Recursive/band limited inversion
usually from well logs. It should be noted that seismic data is Band limited inversion is the process whereby the acoustic
band limited (e.g. data frequency between ~8 and 100 Hz). As the impedance is given for the first seismic layer, and the subsequent
inversion process attempts to provide a broadband model of the layers are calculated recursively from the first layer by utilising
earth’s impedance, low frequency and high frequency data may equation 2. A background drift correction model can be applied
be included in the inversion process. The low frequency data can using a sonic curve. The output inversion has the same frequency
be a model, stacking velocity data or a time/velocity curve from spectrum as the input seismic, hence being band limited.
checkshot data. The high frequency data are the input well logs.

Zero offset inversion Coloured inversion


The simplest type of inversion is zero offset inversion, also Coloured inversion is a modern update on band limited
referred to as full stack inversion, which is the process where inversion, where the inversion is equal to the seismic
full stack seismic data are inverted into a rock property such as multiplied by some operator. In practice, this usually equates
acoustic impedance (AI). This assumes that the recorded seismic to a 90 phase change. Both coloured inversion and band limited
is all zero offset data and that there are no AVO effects. There are inversion are not tied to the well, and can be referred to as relative
many different methods that the data can be inverted by, of which inversions, as the inversion amplitude numbers are only relative
the details of some are presented below. between layers, and not tied absolutely to a well.
A description of seismic amplitude techniques Exploration Geophysics 161

Sparse spike inversion


lnðZS Þ ¼ k lnðZP Þ þ kC þ DLS
Sparse spike inversion assumes that the seismic comes from ð15Þ
dominant AI spikes embedded in a background of smaller spikes lnðrÞ ¼ m lnðZP Þ þ mC þ DLd
(Hampson-Russell, 2009). The well log is processed for the
dominant spikes and the spikes are convolved with a wavelet
where Kc, mc = constants, interpreted on Zp against Zs or r plot;
to calculate the inversion. This method can be particularly useful
and DLs, DLd = deviation from background trend, interpreted on
in carbonates, as large impedance spikes can be present.
cross-plots.
Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Model based inversion LMR inversion


In model based inversion, a low frequency impedance model is LMR inversion uses the relationship between the Lame
built first based on impedances calculated from well logs located constants and Vp and Vs to invert the data (equations 10–14).
in the seismic data that is to be inverted. The low frequency model First the data is inverted to Zp and Zs using the Fatti equation, and
is projected away from the wells along seismically derived then transformed into lr and mr. Simultaneous and LMR
horizons, where the impedance values are interpolated inversion both provide AI and SI volumes as final results.
between the wells, so that there is a low frequency impedance
available at each seismic trace. Model based inversion then
convolves a wavelet with the low frequency model and checks Elastic impedance
the result against the actual seismic trace. The modelled Connolly (1999) first pioneered a method for properly
impedance is then adjusted at each trace until the errors accounting for the changes in reflectivity with offset by
between the model and the seismic are minimised. analysing the Aki Richards equation (Appendix A), which can
be written in the form of equation 16. He observed that the
elastic impedance (EI), or impedance that properly accounts
Stochastic inversion for changes in angle, should be able to be written in the form
Stochastic inversion is a geostatistical method used to account of equation 17. From that point he then solved equation 16 to get
for the non-uniqueness of any inversion. A large number of the elastic impedance equation given in equation 18. Equation 18
inversions are run, all of which honour the seismic and well implies that the AI can be calculated at each different offset
data, and are combined with a variogram to determine the most angle. Conolly’s equation is still valid for the zero offset case, as
likely outcome. shown by equation 19.
1 DAI 1
RP ðÞ   lnAI ð16Þ
AVO inversion – inversion with offset 2 AI 2
1 DEI 1
AVO inversion uses the same algorithms as zero offset inversion, RP ðÞ   lnEI ð17Þ
but differs with the input seismic data. Zero offset inversion takes 2 EI 2
a final full stack volume whereas AVO inversion uses either the EIðÞ ¼ VP
ð1þsin2 Þ ð8Ksin2 Þ ð14Ksin2 Þ
VS r ð18Þ
angle gathers (prestack inversion) or angle stacks for inverting.  2
This method takes into account any AVO affects that may be in the where K ¼ VS
VP
data by calculating the compressional and shear impedances. As a
EIð0 Þ ¼ AI ¼ rVP ð19Þ
measurement of Vs is required as part of the input, either dipole
sonic log data or an approximation of Vs from Vp is also needed. Connolly’s equation did not properly scale the amplitude
There are several different methods for AVO inversion, some changes between the near and the far offsets. His equation was
of which are listed below. These methods are used in conjunction particularly unstable at angles approaching 90 offset due to
with the previously listed methods (i.e. sparse spike simultaneous the siny term. This was resolved in Whitcombe and co-
inversion, SSSI). workers’ equations (Whitcombe et al., 2002) for extended
elastic impedance (EEI), given in equation 20. EI and EEI can
be used to good effect for analysing fluid at far offsets. Results are
Independent inversion usually a near stack inversion and a far stack inversion, which
Independent inversion makes use of the compressional (P) and are proportionally compared for differences which may be AVO
shear (S) reflectivities that can be generated from gathers using related.
Fatti’s approximation of the Zoeppritz equations (Appendix A), "  2   2   #
VP ð1þsin Þ VS ð8Ksin Þ r ð14Ksin Þ
2

and is the simplest method of AVO inversion. These reflectivities EEIðÞ¼VP0 r0


are then independently inverted using the same techniques VP0 VS0 r0
discussed in zero offset inversion to generate P impedance and ð20Þ
S impedance volumes.
By making the substitution tanc = sin y, equation 20 can be
2

reformulated as
Simultaneous inversion
RðcÞ ¼ R0 þ Rsh tan c ð21Þ
This method simultaneously inverts the partial stacks or
gathers, using an individual wavelet for each angle range. This reformulation allows the calculation of the extended
Density and Zs are also solved simultaneously by using the elastic impedance for offsets that have no physical
Fatti equations. Zp is plotted against Zs and density on a significance, such as negative values or values greater than
log–log plot and then the relationship for Zs and density is 90. The advantage of this is that some of these values
solved simultaneously, assuming a water wet lithology approximate other physical properties, such as bulk modulus
(equation 15). (c = 10), l (c = 19) and m (c = –58) (Whitcombe et al., 2002).
162 Exploration Geophysics J. Shadlow

Note also that this formulation is equivalent to the first two terms Carter, D., Mandhiri, D., Park, R. K., Asjhari, I., Syaiful Basyuni, S., Birdus,
in equation 5. S., Bradfield, J. P., Iriawan, A., Nasfiah, M., and Nugroho, M. A. A., 2005,
Interpretation methods in exploration of Oligocene-Miocene carbonate
Pitfalls of inversion reservoirs, offshore northwest Madura, Indonesia: Thirtieth Annual IPA
Convention & Exhibition, August 2005, Jakarta, 179–215.
Any inversion is only as good as the input model. Therefore any Castagna, J. P., and Backus, M. M., 1993, Offset-dependent reflectivity –
problems with the seismic or input well logs will be included in the theory and practice of AVO analysis: SEG
inversion. There are also infinitely many solutions to inversion – Castagna, J. P., and Swan, H. W., 1997, Principles of AVO crossplotting:
the results are not unique. A fairly coarse initial model should be The Leading Edge, 16, 337–344. doi:10.1190/1.1437626
Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

used for model based approaches so that the final inversion is not Castagna, J. P., Batzle, M. L., and Eastwood, R. L., 1985, Relationship
biased towards a preconceived model. Also, post stack inversion between compressional and shear-wave velocities in clastic silicate rocks:
cannot handle any AVO effects that may be in the dataset. The Geophysics, 50, 571–581. doi:10.1190/1.1441933
inversion is also strongly influenced by the quality of the input Castagna, J. P., Swan, H. W., and Foster, D. J., 1998, Framework for AVO
gradient and intercept interpretation: Geophysics, 63, 948–956.
seismic stacks or gathers. If the seismic quality is poor, the
doi:10.1190/1.1444406
inversion will also be poor. AVO inversion required gathers to Chopra, S., and Marfurt, K., 2007, Seismic attributes for prospect
be as flat as possible, or angle stacks to be spectrally balanced and identification and reservoir characterisation: SEG Monogaph 11.
time aligned. Only true amplitude data should be used for Connolly, P., 1999, Elastic impedance: The Leading Edge, 18, 438–452.
inversion, as any amplitude scaling such as automatic gain doi:10.1190/1.1438307
control will alter the ‘lithological’ response that the inversion Fatti, J. L., Vail, P. J., Smith, G. C., Strauss, P. J., and Levitt, P. R., 1994,
is trying to model. Detection of gas in sandstone reservoirs using AVO analysis: A 3-D
seismic case history using the geostack technique: Geophysics, 59,
1362–1376. doi:10.1190/1.1443695
Conclusions Goodway, B., Chen, T., and Downton, J., 1997, Improved AVO fluid
The seismic response is caused by geological and fluid changes. detection and lithology discrimination using Lamé petrophysical
Amplitude and AVO analysis can be used to show any fluid parameters: ‘lr’, ‘mr’ and ‘l/m fluid stack’, from P- and S-inversions:
effects in the data. Inversion can be used to determine the SEG Technical Program Expanded Abstracts, Dallas, 183–186.
Hampson-Russell, 2009, Hampson Russell strata and AVO workshop - course
lithological parameters that affect the seismic as well as any
notes: Hampson Russell Software.
fluid affects that may be present. The changes in amplitude Hilterman, F. J., 2001, Seismic amplitude interpretation: SEG Monograph 4.
recognised on seismic are due to changes in density and Kallweit, R. S., and Wood, L. C., 1982, The limits of resolution of zero-phase
acoustic velocity. AVO effects are caused by a change in the wavelets: Geophysics, 47, 1035–1046. doi:10.1190/1.1441367
VP, VS and r. All interpretation of DHIs, AVO and inverted data Koefoed, O., 1955, On the effect of Poisson’s ratios of rock strata on the
must fit with the geological framework for the area – the results reflection coefficients of plane waves: Geophysical Prospecting, 3,
must be able to relate to the real rocks. This is only a very brief 381–387. doi:10.1111/j.1365-2478.1955.tb01383.x
outline of DHIs, AVO and inversion, and is only intended as an Kunjan, B. K., 2010, Bahamas – a large unconventional biogenic gas play in
introduction. There are many other papers, books and courses the Taranaki Basin – New Zealand: Petromin.
that give a far more detailed description of technique, which are Ostrander, W. J., 1984, Plane-wave reflection coefficients for gas sands at non-
normal angles of incidence: Geophysics, 49, 1637–1648. doi:10.1190/
readily available.
1.1441571
Rutherford, S. R., and Williams, R. H., 1989, Amplitude-versus-offset
Acknowledgements variations in gas sands: Geophysics, 54, 680–688. doi:10.1190/1.1442696
Shuey, R. T., 1985, A simplification of the Zoeppritz equations: Geophysics,
I would like to thank Bala Kunjan, Naomi Osman, Irfan Yuliandri, for their 50, 609–614. doi:10.1190/1.1441936
proof reading and suggestions. Smith, G. C., and Gidlow, P. M., 1987, Weighted stacking for rock property
estimation and detection of gas: Geophysical Prospecting, 35, 993–1014.
doi:10.1111/j.1365-2478.1987.tb00856.x
References Veeken, P. C. H., and Rauch-Davies, M., 2006, AVO attribute analysis and
Aki, K., and Richards, P. G., 1980, Quantitative seismology: theory and seismic reservoir characterization: First Break, 24, 41–52.
methods: W. H. Freeman and Co. Whitcombe, D. N., Connolly, P. A., Reagan, R. L., and Redshaw, T. C., 2002,
Bortfeld, R., 1961, Approximations to the reflection and transmission Extended elastic impedance for fluid and lithology prediction:
coefficients of plane longitudinal and transverse waves: Geophysical Geophysics, 67, 63–67.
Prospecting, 9, 485–502. doi:10.1111/j.1365-2478.1961.tb01670.x Wiggins, R., Kenny, G. S., and McClure, C. D., 1983, A method for
Brown, A., 2004, Interpretation of three-dimensional seismic data: AAPG determining and displaying the shear-velocity reflectivities of a
Memoir 42. geologic formation: European Patent Application 0113944.
Brown, A., 2005, Pitfalls in 3-D seismic interpretation: AAPG 3-D Seismic Zoeppritz, K., 1919, Erdbebenwellen viiib on the reflection and propagation of
Symposium, March 2005, Denver. seismic waves: Gottinger Nachrichten, 1, 66–84.
A description of seismic amplitude techniques Exploration Geophysics 163

Appendix A
Zoeppritz equations
The Zoeppritz equation (equation A-1) (Zoeppritz, 1919) evaluates the interactions of a seismic wave at all angles, not only at a zero
offset. This can be simplified into the various different approximations in Table A-1. It should be noted that the Bortfeld (1961), Aki and
Richards (1980) and Shuey (1985) equations are the only true approximations of the Zoeppritz equations; Wiggin’s (Wiggins et al.,
1983) and Fatti’s (Fatti et al., 1994) equations are only mathematical equivalents of the Aki Richards equation.
2 31
 sin1  cos1
Downloaded 01/27/16 to 130.15.241.167. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

sin2 cos2
2 3 2 3
RP ð1 Þ 6 7 sin1
6 cos1  sin cos  sin 7
6 7 6 1 2 2 7 6 7
6 RS ð1 Þ 7 6 7 6 cos1 7
6 7 6 V r V 2
V r V V 7 6 7 ðA-1Þ
6 7¼6 P1 2 S2 P1 2 S2 P1
cos22 7 6 7
6 TP ð1 Þ 7 6 sin21 V cos21 r V 2 V cos21 r1 VS12 7 6 sin2 7
4 5 6 S1 1 S1 P2 7 4 15
6 7
TS ð1 Þ 4 VS1 r2 VP2 r2 VS2 5 cos21
 cos21 sin21 cos22  sin22
VP1 r1 VP1 r1 VP1

Table A-1. Details of the different Zoeppritz equation approximations (Aki and Richards, 1980; Bortfeld, 1961; Fatti et al., 1994; Shuey, 1985;
Wiggins et al., 1983).

Approximation name Equation Simplifications


Aki Richards RP ðÞ ¼ a DV
VP
P
þ b DV
VS
S
þ c Dr
r a ¼ 2cos1 2 
h i2
b ¼ 4 VVSP sin2 
   
2
c ¼ 0:5  2 VVSP sin2 
h i
Dr
Shuey RP ðÞ R0 ¼ 12 DV VP þ r
P

h i
Ds
 R0 þ A0 R0 þ ð1sÞ 2

sin2  A0 ¼ B  2ð1 þ BÞ 12s
1s
DVP

þ 12 DV
VP ðtan   sin Þ
P 2 2
B ¼ DVPVP
VP þ Dr
r

Bortfeld Rði Þ ¼ R0 þ Rsh sin2 ði Þ þ RP tan2 ði Þsin2 ði Þ DVP
RP ¼
2VP
Dr
Rr ¼
2r
R0 ¼ RP þ Rr
 
1 DVP Dr DVS
Rsh ¼ k  2k
2 VP r VS
 
2VS 2
k ¼
VP
 
Wiggins RP ðÞ ¼ R0 þ Rsh sin2  þ RP tan2 sin2  ¼ Bortfeld equation 1 DVP Dr
R0 ¼ þ
2 VP r
 2  2
1 DVP VS DVS VS Dr
Rsh ¼ 4 2
2 VP VP VP VP r
DVP
RP ¼
2VP

Fatti RP ðÞ ¼ C1 R0 þ C2 RS þ C3 RD C1 ¼ 1 þ tan2 


VS2 2
C2 ¼ 8 sin 
VP2
VS2 2 1
C3 ¼ 2 sin   tan2 
VP2 2
 
1 DVP Dr
R0 ¼ þ
2 VP r
 
1 DVS Dr
RS ¼ þ
2 VS r
Dr
RD ¼
r

www.publish.csiro.au/journals/eg

You might also like