post stack acoustic impedance
TRANSCRIPT
-
8/13/2019 Post Stack Acoustic Impedance
1/18
Post Stack Acoustic Impedance (AI) Inversion:"Basics and Usage
"By Ajay Badachhape"Seismic Analysis roup
Seismic Imaging !echnology enter
#$amples
1. Seismic Amplitude vs. AI Inversion section for deepwater
GOM amplitude anomaly: Grand Canyon
2. Te Comparison of Seismic Amplitude! "ecursive Trace
Inte#ration! and Sparse Spi$e Inversion sections for
Stratton %ield! Sout Te&as
3. "eservoir 'roperties from Inverted AI "esults: 'orosity
from Inverted AI for te (ellow Sand interval in te )rsa
field
Basics and Usage
1. Overview2. Te *avelet
+. %unctionality
,. 'ractical Concerns and -imitations
. /istorical Metodolo#y
0. "eservoir 'roperties from Inverted "esults
http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%201http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%201http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%202http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%202http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%202http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%203http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%203http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%203http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Overviewhttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Wavelethttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Functionhttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Practicalhttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Historicalhttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Reservoirhttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%201http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%201http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%202http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%202http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%202http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%203http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%203http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Example%203http://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Overviewhttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Wavelethttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Functionhttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Practicalhttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Historicalhttp://upstream.ho.conoco.com/sitc/SAG/Acoustic_Impedance.html#Reservoir -
8/13/2019 Post Stack Acoustic Impedance
2/18
%vervie&
On a basin-wide scale seismic data is the best tool currently available
to predict subsurface properties of rocks and fluids. It has the
resolution and depth of imaging to provide valuable information that
can be used to predict reservoir properties. Acoustic impedance (AI)
inversion is a term used to designate methodologies that attempt to
compute or estimate rock properties directly from measured (e.g.,
seismic) data. AI is the only rock property (or combination of rock
properties velocity multiplied by density) that can be directly
estimated from seismic data.
!"his is an estimate since modeled results do not necessarily produce a uni#ue match to the measured seismic data.
"he convolutional model for seismic reflection data assumes the
measured seismic trace is composed of the earth reflectivity
convolved with the seismic wavelet. "he earth reflectivity is generated
by interfaces which have acoustic impedance contrasts given by the
following basic e#uation$
Rc=2v2-
1v1
2v2+
1v1
where Rcis the reflection coefficient for the interface between layers
% and &, which have velocities and densities given respectively by
v%,
1and
v&,
2 .
-
8/13/2019 Post Stack Acoustic Impedance
3/18
!he 'avelet
"he wavelet is a key part of the inversion and well log calibration
process. It is the transfer function that provides the link between well
log acoustic impedance and the seismic data. "he wavelet is scaled
and spectrally shaped so that when it is convolved with the reflectivity
series derived from the acoustic impedance log, it produces synthetic
seismic traces of the appropriate fre#uency content and amplitude to
match the seismic trace(s) closest to the well. In a given data set ('
volume, & line, or set of & lines ac#uired and processed at the
same time and in the same manner), one single wavelet typically may
be used to tie all the well logs to the seismic data and produce a good
match between the synthetic trace and nearest seismic trace for each
well. ome of the assumptions for this to be true include the
following$
imilar range of depths, times (not too steep structurally so that
one well is at %*** ms. and another is at &+** ms.), and
geologic strata so that the fre#uency content of the seismic
data will be similar
imilar range of seismic amplitudes so that the wavelet from
one well with low amplitude reflections near it would produce a
wavelet with a low peak amplitude while another one in a one
of larger amplitude reflections would result in a wavelet with ahigh peak amplitude
o large artifacts present in the seismic data at one or a few of
the wells only this includes fault shadow, salt diffractions, out
of plane energy, multiples, etc.
-
8/13/2019 Post Stack Acoustic Impedance
4/18
If all assumptions are met, but similar wavelets are not produced from
each of the wells, typically, the log data for some or all the wells are
#uestionable and additional log processing will be necessary to edit
the acoustic logs (sonic and density) prior to use for inversion.
etermining the wavelet is e/tremely critical since most of the newer
inversion methods e/tract a wavelet from the well log and seismic
data, then assume the wavelet is known and use it to invert for the
acoustic impedances (opposite of creating a synthetic from the well).
0e are now going to assume the wavelet is constant and for each
seismic trace, solve for an impedance model that produces a
reflectivity series that when convolved with the known wavelet,
produces a synthetic that matches the seismic trace). An e/ample of
a wavelet e/traction and synthetic1seismic comparison is seen in
2igure %.
Once the wavelet amplitude, fre#uency content, and phase have
been e/tracted from the seismic data over the inversion time gate as
accurately as possible, the inversion algorithm can automatically
account for tuning and sidelobe events. 3/tra events that are due to
constructive interference from wavelet sidelobes as well as tuning
responses are purely wavelet phenomena. Once the wavelet has
been accurately e/tracted, inversion can essentially eliminate these
seismic artifacts. 2or this reason, maps of inverted AI anomalies
usually represent the true location of anomalies and are of different
sie and shape (typically) than the seismic amplitude anomalies.
-
8/13/2019 Post Stack Acoustic Impedance
5/18
4eduction of most of the tuning and sidelobe responses places the
anomalies in their proper position.
igure : 'avelet e$traction and seismic to synthetic tie
A single wavelet is ade#uate in almost all cases. Occasionally, a very
long time gate is to be inverted (e.g., typically more than & seconds of
data), in which case two or more inversion time windows may have to
be run separately and smoothly merged later. 5aterally varying
wavelets are a tricky problem since the wavelets are estimated at the
well positions (typically, but not always), and simple linear
e/trapolation between the well1wavelet positions is not ade#uate to
properly model where the wavelet changes actually take place. "hese
-
8/13/2019 Post Stack Acoustic Impedance
6/18
wavelet changes may be due to e/tremely anisotropic areas, gas
chimneys, salt, faults, comple/ geologic areas such as overthrusts,
etc. (typically areas where the lateral seismic fre#uency content
changes drastically). "hese situations can be modeled fairlywell using
inversion, but the approaches necessarily vary by situation and need
to be handled on an individual basis.
-
8/13/2019 Post Stack Acoustic Impedance
7/18
unctionality
AI inversion is useful for a variety of reasons, including the following$
3nhances resolution compared to seismic data
6rovides a geologically consistent, layer-based cube or section
of acoustic impedance data from seismic data
6rovides the ability to detect small lateral changes in acoustic
impedance within layers
6rovides the ability to delineate possible reservoir ones more
accurately than seismic data (due to removal of wavelet effects
such as tuning and false events caused by sidelobe
interference)
3nhances fault detection and delineation
3nhances delineation of fluid contacts
epicts data as layer information rather than interface data
6rovides the ability to use ' visualiation techni#ues on
inverted data to view reservoir geometries
7akes interpretation1tracking of events in low amplitude (and
other difficult) ones easier
6rovides the ability to convert AI to other parameters such as
porosity, net sand, w, etc., (providing that crossplot analysis of
well log data show appropriate trends to 8ustify a conversion).
As an illustration, 2igure %.% shows a seismic section with two
potential targets indicated as trough-peak pairs with amplitude
anomalies. "he deeper target has better trough development, but the
-
8/13/2019 Post Stack Acoustic Impedance
8/18
following peak is not as strong as in the upper target. 2igure %.&
shows the inversion results for the same section. "he upper target is
a thicker, porous sand, and the lower target is actually two thin sands
that are not as porous and have a fairly thick shale interval in
between. "he base of the second sand in the lower target is below
the interpreted base of the sand from the seismic.
"he inversion illustrates that the most anomalous part of the section
is the very hard (acoustically) shale in between the two targets. "his
(presumably dewatered) shale has an e/tent that closely corresponds
to the area in which the upper sand is seen to be lower in acoustic
impedance and most porous. It should also be noted that the
inversion has now produced information about the layers, not 8ust the
interfaces. "he tracked horions on the seismic are the centers of the
trough-peak pairs that comprise the interpreted targets. "he inversion
results show each horion tracked on a trough to be the top of the
sand units, while the horions tracked on the peaks are the bases of
the sands (the second sand in the lower target was not resolved in
the seismic). "he intervening hard shale is not identifiable as such on
the seismic it is manifested as strong peak development at the
bottom of the upper target, and as strong trough development at the
top of the lower target due to constructive interference which
increased the strength of the amplitude anomalies.
-
8/13/2019 Post Stack Acoustic Impedance
9/18
Practical oncerns and *imitations
"he AI inversion results may be thought of as 9pseudo-logs9, which
can only be calibrated at well locations. :owever, since AI inversion
results are modeled rather than measured data, they can be prone to
some problems. 2or e/ample, the bandlimited nature of the input
(seismic) data produces results that are non-uni#ue i.e., many AI
9pseudo-logs9 produce reflectivity traces that can be convolved with
the wavelet to produce e#ually valid synthetic1seismic correlations.
"his is especially true given the bandlimited nature of seismic data
(and therefore the inverted results) and emphasies the need for a
valid low- fre#uency model and appropriate, geologically reasonable
constraints (if the inversion method incorporates the use of limits on
the results). "he AI constraints placed upon the inversion results and
the accuracy of the low-fre#uency AI model away from the wells is
crucial to producing the best possible results.
AI inversion results are therefore not a magic bullet. 4esults have to
be analyed to ensure that the data falls within ranges that are
geologically feasible and make sense when compared to the
available well data, velocity trends, or other data. "he inversion
method also has to take into account the results of other methods of
analysis, including A;A1amplitude analysis techni#ues. 0hen
inversion results do not agree with another method such as A;A
analysis, there should be a geologic reason for the discrepancy (e.g.,
presence of a hard shale above or below the target one which
complicates the reflection image but was not accounted for in the
other method).
-
8/13/2019 Post Stack Acoustic Impedance
10/18
+istorical ,ethodology
AI inversion in a basic sense may be thought of as taking ero-phase
seismic data and applying a
-
8/13/2019 Post Stack Acoustic Impedance
11/18
Another method that appro/imates inversion is simple bandpass
filtering and phase rotation (which can be cascaded to apply to the
entire time window), which may or may not also be merged with a
low-fre#uency AI model. 2iltering is the fastest and cheapest method,
however, neither filtering nor trace integration methods e/tend the
bandwidth or resolution beyond the seismic band. "he introduction of
low fre#uencies e/tends the bandwidth on the low-fre#uency side so
that geologic trends are introduced, but nothing more is gained.
-
8/13/2019 Post Stack Acoustic Impedance
12/18
-eservoir Properties .rom Inverted -esults
ince the results of inversion are models of physical properties of the
earth, analysis of well data in the area can be used to produce
estimates of other properties that can be derived from acoustic
impedance. 2or e/ample, 2igure '.% shows a crossplot with a clear
trend between acoustic impedance and density (which is easily
translated to density porosity) and 2igure '.& shows a map of the
results of converting acoustic impedance to porosity for one one in
an area. 2or comparison purposes, 2igure '.' shows the minimum
seismic amplitude (ma/imum trough development) for the same
interval. 2igure '.& shows the effects of removal of tuning and
sidelobe events the anomalies are correctly positioned, they have
changed in sie and shape, and they provide better fault resolution of
what is now apparent as fault-bounded porous sand bodies.
=onversion of inverted data to other properties is only as good as the
inverted results plus the trend used for conversion. =areful analysis
of the inverted results and the information used to create the formula
or trend (crossplots or trend curves from wells in the area, etc.) used
to convert the impedances to another property must be done. It is
easy to produce dubious results from data that do not form clear
trends. =alibration to well data and comparison with results from
other methods or geologic information is critical.
0hen used in concert with all available data, AI inversion can be a
valuable tool. 4ecognition of its strengths and weaknesses is
necessary to properly utilie it and compare1supplement it with other
analysis methods.
-
8/13/2019 Post Stack Acoustic Impedance
13/18
&le 1Seismic Amplitude vs. AI Inversion section for deepwater
GOM amplitude anomaly: Grand Canyon
igure /: Seismic section over amplitude anomaly/!&o stacked amplitude anomalies are delineated 0y the tracked hori1ons as trough2peakpairs/
igure /3: Inversion results over amplitude anomaly/!his is a 0road 0and inversion result in &hich the #arth,odel .rom 423 +1/ is derived .romthe &ell log and the 3254 +1/ data is dra&n .rom inversion results using the 6asoneoscience 'ork0ench7s onstrained Sparse Spike inversion algorithm/!he upper seismic anomaly (0lue to green hori1ons) is sho&n to 0e due to the presence o.a thick8 porous sand/ !he lo&er seismic anomaly (yello& to cyan hori1ons) is t&o thinnersands &ith a signi.icant shale 0reak 0et&een/ !he second sand in this deeper target is
-
8/13/2019 Post Stack Acoustic Impedance
14/18
0elo& the original ma$imum peak that &as tracked on the seismic as 0eing the 0ase o. theanomaly/ !his inversion result sho&s that the most anomalous event in the section is thehigh impedance shale 0et&een the t&o target anomalies/ !he high impedance shale ispresuma0ly a de&atered shale that has appro$imately the same lateral e$tent as theporous sand development/ !he high impedance shale caused constructive inter.erence o.0oth anomalies and 0rightened them on the seismic data/
-
8/13/2019 Post Stack Acoustic Impedance
15/18
Example 2The Comparison of Seismic Amplitude, Recursive Trace Integration, and Sparse Spike Inversion
sections for Stratton ield, South Texas
Areas o. interest include the sections 0elo& the green / hori1on (just a0ove /5seconds) &here resistivity spikes (cyan &iggle) indicate hydrocar0on presence/ !hemagenta &iggle is the impedance log that sho&s &hich log events created the seismicevents/
Figure 2.2: Recursive Trace Integration results for cross-line 154.
The trace integration process has transforme interface ata into la!er ata. This process oes not increase
resolution an oes not incorporate lo" fre#uenc! $vertical tren% information.
-
8/13/2019 Post Stack Acoustic Impedance
16/18
2igure &.'$ parse pike Inversion results for cross-line %+?."he inversion results closely match the well results. and morphology and lateral porositychanges are evident. A greater amount of detail is observable for a given layer compared to2igures &.%-&.&. "he background acoustic impedance has a low-fre#uency trend that is seen as agradual increase in acoustic impedance with depth.
-
8/13/2019 Post Stack Acoustic Impedance
17/18
Example !Reservoir "roperties from Inverted AI Results#
$A%I&'()CE'TER)"orosit* rom Inverted AI for the +ellow Sand interval in the rsa field
Figure 3.1: &rossplot of 'coustic Impeance, (ensit!, an )amma Ra! $color%.
This crossplot from a eviate "ell in the eep"ater )*+ rsa fiel sho"s a clear san tren that isifferent from the shale tren. The sans $"hite to !ello" to orange, re, an green colors% are lo"er inacoustic impeance an ensit! than the shales $green to lue% in the upper right of the plot. 'n impeance
of aout 23, g/cc0ft/sec can e use as a cutoff et"een fairl! clean to clean sans versus san! shales
to shales. Impeances elo" 23, g/cc0ft/sec can e calirate on the asis of a linear tren et"een 'I
an ensit! $"hich is easil! converte to ensit! porosit!%. This tren can e use to compute porosit!
from 'I.
-
8/13/2019 Post Stack Acoustic Impedance
18/18
Figure 3.2: +ap of orosit! compute from Inverte 'I an &rossplot ata.
This porosit! map for the ello" san reservoir in the rsa fiel sho"s clear patterns that inicate the
sans are fault oune an have t"o provenances the sans in the lo"er half of the plot proal! came
from the east-northeast, "hile the sans in the upper half proal! came from the north-north"est. The
porosit! trens closel! match the "ell ata. The 'I anomalies have een correctl! positione ue to the
removal of most of the "avelet phenomena $e.g., tuning an sieloe events%. &onversion of the 'I ata toporosit! for a given one provies aitional information aout the #ualit! of the sans.
igure 9/9: ,inimum Seismic Amplitudes over target 1one/!he seismic amplitude anomalies are not in the same position nor the same si1e andshape as the inversion results indicate/ !he .aults are not as clearly de.ined and the sandsto the south do not appear to 0e .ault 0ounded/
pecial thanks to r. 4obert =orbin for editing and to 4e/ 7c@inleyfor reviewing this article.