You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/262544278

High-resolution acoustic impedance inversion to characterize turbidites at


Marlim Field, Campos Basin, Brazil

Article  in  Interpretation · August 2014


DOI: 10.1190/INT-2013-0137.1

CITATIONS READS

10 929

3 authors:

Tatiane Nascimento Paulo T. L. Menezes


Universidade Federal Fluminense Rio de Janeiro State University
1 PUBLICATION   10 CITATIONS    85 PUBLICATIONS   265 CITATIONS   

SEE PROFILE SEE PROFILE

Igor Braga
Universidade Estadual do Norte Fluminense
11 PUBLICATIONS   45 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Bacia do Rio Almada-Ilheus View project

Multi-Physics Methods applied to oil industry View project

All content following this page was uploaded by Paulo T. L. Menezes on 31 July 2014.

The user has requested enhancement of the downloaded file.


t Technical paper

High-resolution acoustic impedance inversion to characterize turbidites


at Marlim Field, Campos Basin, Brazil
Tatiane M. Nascimento1, Paulo T. L. Menezes2, and Igor L. Braga3
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Abstract
Seismic inversion is routinely used to determine rock properties, such as acoustic impedance and porosity,
from seismic data. Nonuniqueness of the solutions is a major issue. A good strategy to reduce this inherent
ambiguity of the inversion procedure is to introduce stratigraphic and structural information a priori to better
construct the low-frequency background model. This is particularly relevant when studying heterogeneous
deepwater turbidite reservoirs that form prolific, but complex, hydrocarbon plays in the Brazilian offshore basins.
We evaluated a high-resolution inversion workflow applied to 3D seismic data at Marlim Field, Campos Basin, to
recover acoustic impedance and porosity of the turbidites reservoirs. The Marlim sandstones consist of an Oligo-
cene/Miocene deepwater turbidite system forming a series of amalgamated bodies. The main advantage of our
workflow is to incorporate the interpreter’s knowledge about the local stratigraphy to construct an enhanced
background model, and then extract a higher resolution image from the seismic data. High-porosity zones were
associated to the reservoirs facies; meanwhile, the nonreservoir facies were identified as low-porosity zones.

Introduction deterministic methods as their first choice (Tetyukhina


High-resolution reservoir characterization studies al- et al., 2010). The incorporation of structural and strati-
low detailed lithology-fluid prediction in the various graphic features in the inversion process decreases the
stages of the reservoir life cycle. Commonly, this is a inherent nonuniqueness of the inversion results (Mer-
complex task, in which interpreters usually have to in- letti and Torres-Verdín, 2006). These inputs allow one
tegrate a huge amount of data with different degrees of to differentiate between similar mathematical solutions,
accuracy and spatial resolution, such as 3D seismic data on the basis on their geologic viability (Pendrel and Van
(uniformly sampled information over large areas with Riel, 1997).
low vertical resolution), well logs (high-resolution dis- The nonuniqueness in the seismic inversion pro-
tributed along irregularly distributed well paths) and cedure is mainly due to the band-limited frequency con-
geologic conceptual models. tent, missing the highest and lowest frequencies even
Inversion of seismic data is a widely used tool to es- for noise-free data, of the registered seismic waves. This
timate reservoir geometry and petrophysical characteri- can be partially mitigated by the introduction of exter-
zation (Kelamis et al., 1995). One benefit of the seismic nal independent a priori information, e.g., coming from
inversion process is to increase the frequency content the wells, to determine the lowest and highest fre-
of the seismic data, which is band limited (Latimer et al., quency content to be incorporated into the inverse
2000). Furthermore, the result of inversion also allows model. In practice, the inversion of seismic data is more
lithology differentiation and hopefully the estimation of complicated (Bosch et al., 2010) due to the presence
the fluid content, and not only the interface geometry as of noise in real data, forward-modeling simplifica-
given by the amplitude attribute (Avseth et al., 2005; tions needed to estimate the solutions in a reasonable
Contreras et al., 2005). Because acoustic impedance time, and uncertainties in well-to-seismic ties (veloc-
is a property related to lithology, the inversion process ities used in depth-to-time conversions), wavelet extrac-
enables greater inference of subsurface geology in seis- tion, and links between the reservoir and elastic
mic interpretation (Goodway et al., 1997). parameters.
Seismic inversion procedures can be divided into The standard inversion workflow commonly uses a
two main categories (Bosch et al., 2010): deterministic stratal slicing technique for building a low-frequency
and stochastic methods. Geoscientists tend to use the model in model-based inversions (Russell and Toksöz,

1
Universidade do Estado do Rio de Janeiro (UERJ), Faculdade de Geologia, Rio de Janeiro, Brazil. E-mail: tatianegeologa@gmail.com.
2
UERJ, Departemento de Geologia Aplicada, Faculdade de Geologia, Rio de Janeiro, Brazil. E-mail: ptarsomenezes@pq.cnpq.br.
3
Invision Geophysics, Rio das Ostras, Brazil. E-mail: igor@invisiongeo.com.br.
Manuscript received by the Editor 30 August 2013; revised manuscript received 3 February 2014; published online 16 May 2014. This paper
appears in Interpretation, Vol. 2, No. 3 (August 2014); p. T143–T153, 10 FIGS., 1 TABLE.
http://dx.doi.org/10.1190/INT-2013-0137.1. © 2014 Society of Exploration Geophysicists and American Association of Petroleum Geologists. All rights reserved.

Interpretation / August 2014 T143


1991). In some areas, the construction of this low-fre- highs caused by faults, and irregular deposition due
quency model is hampered due to a wide variation in to the movement within salt layers. The described sce-
the lithology distribution throughout the investigated nario corresponds precisely to what is found in the
area. Examples in offshore Brazil include deepwater giant Marlim Field (Figure 1), where Oligocene and Mio-
channel fan, switching delta lobe deposits, structural cene sand-rich turbidites form important reservoir
rocks. These turbidites were deposited in deepwater
settings associated with slope and continental rise de-
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

posits (Freire, 1989). The Marlim sandstones occur


dominantly as tabular bodies with more than 95% of
massive sand with excellent porosity, up to 30% (Souza
et al., 1989). The porosity is controlled by carbonate
cementation, Carvalho et al. (1995) identified four dif-
ferent diagenetic reservoir facies within Marlim sand-
stones. Deepwater turbidite reservoirs, like Marlim,
usually show a complex distribution of sand bodies
with thicknesses many times smaller than the vertical
resolution limit of the seismic data (Ribeiro, 2012).
Therefore, a more detailed description of the reservoir
can only be achieved through the inversion procedure.
In the present paper, we show the results of a high-
resolution deterministic inversion workflow to recover
acoustic impedance and predict porosity distribution of
the Marlim turbidite reservoirs. The main advantage of
the proposed workflow is to better estimate the low-fre-
quency model through the so-called HorizonCube
model builder (de Groot et al., 2010; Brouwer et al.,
2012) that incorporates stratigraphic knowledge of
the main sedimentary sequences of the Campos Basin
in the study area. A priori independent information
given by four available wells in the area plays a funda-
mental role in different steps of the proposed workflow.
Beside the traditional well-to-seismic ties for wavelet
Figure 1. Location map of the seismic data set at Marlim
estimation, available stratigraphic information in the
Field, Campos Basin, Brazil. P1, P2, P3, and P4 are available
wells. Coordinates are in UTM, zone 39S. Marlim reservoir wells was used to help in the seismic interpretation
outline is shown in the black line. L1, location of the inter- and build the high-resolution low-frequency model.
preted seismic line. Finally, the well logs provided constraints to the abso-
lute impedance and porosity estimations
in calibrating the inversion.

Campos Basin stratigraphic setting


The sedimentary section of the Cam-
pos Basin can be divided into three meg-
asequences associated to the following
tectonic development stages: rift, transi-
tional, and drift (Figure 2).
The Barremian lacustrine sediments
of Lagoa Feia Formation overlie the
Hauterivian Cabiúnas basalts (120–
130 Ma); these basalts represent the eco-
nomic basement of the Campos Basin.
The Lagoa Feia sediments are consid-
ered to be the most important source
rocks in the Campos Basin. The Aptian
sequence contains, from base to top:
conglomerates, carbonates, and evapor-
itic rocks deposited in a period of
tectonic quiescence. This transitional
Figure 2. Simplified stratigraphic chart of the Campos Basin, modified from stage represents the beginning of the
Guardado et al. (2000). drift phase in which the sediments are

T144 Interpretation / August 2014


associated to the first seawater inflows through the northern, and southern portions of the reservoir. To the
Walvis Ridge (Leyden et al., 1976). east, the trap is structural, where the accumulation ends
The drift stage starts with a marine megasequence against a normal listric fault associated with the under-
characterized by the Albian/Cenomanian shallow-water lying moving salt. According to Mello et al. (1994), this
calcarenites and calcilutites of the Macaé Formation. fault represents the main route of oil migration from the
The marine Upper Cretaceous to Paleogene deepwater presalt source rocks to the turbidite reservoir.
clastic section (Carapebus Formation) is composed by In late 1996, early 1997, Geco-Prakla acquired for
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

shale, marls, and sandstone turbidites, deposited in a Petrobras the 3D seismic data interpreted in this study
period of general tectonic quiescence and of continued (Sansonowski et al., 2007). The survey area spanned
subsidence. These turbidite systems comprise the most greater than 720 km2 (Johann et al., 2009), and the ac-
important petroleum reservoirs in the Campos Basin quisition parameters are shown in Table 1. This data set
(Bruhn et al., 2003). Progradational siliciclastic sequen- became public in 2003 when the Brazilian Petroleum
ces characterize the remaining Neogene section. In the Agency (ANP) provided a poststack time-migrated,
deep water of the Campos Basin, the deposition of this zero-phase American polarity (Brown and Abriel,
megasequence was strongly conditioned by salt tecton- 2014), 3D seismic amplitude cube. The ANP also pro-
ics (Mohriak et al., 1996). vided check shots, stratigraphy, and composite well
logs (density, sonic, porosity, and gamma ray) of four
wells in the study area (Figure 1).
Marlim Field data set
The giant Marlim Field is located in the northeastern
portion of the offshore Campos Basin, Brazil. This oil Inversion workflow
field, discovered in 1985 at water depths ranging from To interpret the Marlim reservoir, we applied a
600 to 1200 m (Figure 1), had an initial estimate of over four-step inversion workflow, namely, (1) seismic data
6.4 billion STB with 1.97 billion barrels recoverable filtering, (2) well-to-seismic tie, (3) HorizonCube and
(Candido and Cora, 1978). Oil production started in acoustic impedance inversion, and (4) porosity pre-
1991, and Marlim shortly became the largest producing diction.
field in Brazil with daily production of 610,000 barrels
(Johann et al., 2009). Seismic data filtering
The reservoirs in Marlim Field (Figure 1) consist of Seismic interpretation is highly dependent on the
an Oligocene/Miocene deepwater turbidite system. quality and resolution of the seismic data (Davogustto
They form a series of amalgamated sandstone bodies et al., 2013). Frequently, even a small amount of noise
known as Marlim sandstone (Peres, 1993). The turbi- contaminating the amplitude data can mask subtle seis-
dite sedimentation occurred under the influence of a mic events. Data filtering is then necessary to reduce
regional northwest–southeast transfer faults system. noise and enhance the signal-to-noise ratio of the seis-
This regional system, reactivated in the Tertiary, pro- mic data. To that end, we applied a dip steering median
vided the pathway for the turbidite sedimentation, filter (Brouwer and Huck, 2011; Qayyum and de Groot,
reworking, and redistribution of several deepwater 2012) to attenuate the high-frequency random noise
reservoirs in the Campos Basin (Cobbold et al., 2001). from the seismic amplitude data. The main advantage
The Marlim sandstone shows a remarkable signature
in the seismic amplitude data (Figure 3). Fainstein et al.
(2001) associate this high-amplitude seismic response Table 1. Seismic acquisition parameters.
to the impedance contrast between the high-porosity
reservoirs and the overlying high-speed shale. The
Acquisition
Marlim trap is stratigraphic (pinch out) in the western,
Area (km2 ) 720
Number of cables 6
Spreads (offset, m) 0–148–3535.5
Channels/cable 288
Shot point interval (m) 25
Receiver interval (m) 12.5
Cable interval (m) 50
Sample rate (ms) 1
Bin size (m) 12.5 × 25
Cable depth (m) 9
Azimuth (°) 123
Nominal fold 72
Figure 3. Seismic signature of the Marlim reservoir (turbi- Traces∕km2 230,400
dite), modified from Candido and Cora (1978).

Interpretation / August 2014 T145


of this approach is that the median operator removes well-to-seismic tie at all suitable wells. After that gener-
outliers in the data and retain steps in data values as en- ation, crosscorrelation and validation can be used to se-
countered in nearby faults, so the filtering process pre- lect one wavelet to invert the seismic data. Usually, the
serves the traces’ amplitudes and edges and enhances as correlation rate of the seismic-to-well tie depends on
continuous events (Qayyum and de Groot, 2012). There- the quality of the well and seismic data and also the
fore, it facilitates structural interpretation because it en- time/depth relationship.
ables more accurate autotracking and allows the We edited the original sonic and density well-log data
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

interpreter to extract more detail from the seismic data. of all four available wells to remove spurious values.
In Figure 4, we show the comparison between the The sonic and density logs were multiplied to yield
original amplitude data (Figure 4a) and the dip steering the acoustic impedance logs (Figure 5). The well data
median filtered data (Figure 4b). It is possible to ob- were then converted to time using a check-shot-cali-
serve in the filtered data an overall improvement, asso- brated time-depth curve. After that, we established
ciated with the high-frequency noise content reduction. the well-to-seismic tie in time by using a statistical
As a result, we obtain a sharper subsurface image. An wavelet extraction from the data (Nascimento et al.,
important aspect of the filtered data is the better defi- 2013). The central idea is to estimate a wavelet that in-
nition of small amplitude reflections inbetween the ma- creases the correlation between synthetics and the seis-
jor horizons. The enhancement on the lateral continuity mic data. Therefore, an appropriate choice of interval,
of these small reflections is a necessary condition for a length, frequency, and taper filter is absolutely neces-
precise structural horizon mapping of the HorizonCube. sary for a better estimation of the wavelet (Salleh
The corresponding data-driven multihorizon tracking and Ronghe, 1999).
algorithm is based on following the dip and azimuth In the present work, we extracted the wavelet using
of the seismic data. According to Brouwer et al. a 500 ms window centered at 2650 ms. In Figure 5, we
(2012), the use of smoothed dip fields is preferable be- show the extracted wavelet with its spectrum and also
cause they are more continuous than amplitude fields
the good match between the synthetic and seismic data
and less prone to noise.
at the P1 well. The synthetics were calculated by con-
volving the extracted wavelet with the acoustic imped-
Well-to-seismic tie ance computed from the available well logs. The
The well-to-seismic tie is the best way to relate the
correlation was done at the same time window of the
seismic response to rock properties by using synthetic
wavelet estimation. The correlation coefficient is 83%.
seismograms generated from well logs (White, 1997). It
These results are representative for all four wells in
is widely used for interpretation because it enables the
the study area. The 17% mismatch between the seismic
correct identification of mapped seismic reflections and
and synthetic data can be explained by two different
also optimizes the correlation between the acoustic
geologic causes. Above the top of the Marlim reservoir
impedance model and seismic data. A successful inver-
(TM), we have a dipping siliciclastic sequence (see
sion scheme starts with the wavelet determination for a
Figure 3), a 2D situation, whereas our convolutional
model is essentially 1D. Below the base
of the Marlim reservoir (BM), we have
highly heterogeneous marls/carbonate
sections that lead to a lower correlation
between the signals.

Acoustic impedance inversion and


the HorizonCube
Seismic inversion is the process of
restoring the subsurface impedance
model or any other elastic property from
seismic data. This can be achieved by
several approaches, the most common
of which is to formulate the solution
as the minimization of an objective
function representing the difference be-
tween the modeled data and the ob-
served seismic. For many decades, the
stationary convolution model has been
used to shape a seismic trace, but it fails
Figure 4. Upper panel: left, original amplitude data; right, correspondent am-
plitude spectrum. Lower panel: left, dip steered median filtered data; right, cor- to take into account propagation effects
respondent amplitude spectrum. Note the overall enhancement of the filtered such as absorption and dispersion that
data associated to the noise reduction in the high-frequency range (above 75 Hz). occur on a real seismic data set.

T146 Interpretation / August 2014


The inversion scheme herein adopted (Braga, 2011) mation of the subsurface impedance. The advantage of
takes advantage of the ability that the continuous wave- the proposed method is that the wavelet can be esti-
let transform (CWT) has to represent non stationary mated from the redundant crosscorrelation of closely
phenomenon (Appendix A). We define the CWT of a sig- dilated wavelets, where the inversion process can take
nal as the crosscorrelation of a dilated and translated care of small variations in the pulse form. The method-
version of a reference wavelet with the seismic signal ology proposed by Braga (2011) is similar to the Mallat
under analysis. This shows how well the signal corre- and Zhong (1992) and Wapenaar et al. (2005) ap-
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

lates with each wavelet at each time instant (Braga, proaches in the use of CWT to extract subsurface infor-
2011). The decomposed seismic trace is a 2D complex mation from seismic data.
matrix in which the maximum energy at each scale rep- The HorizonCube is a 3D stack of horizons (de Groot
resents the most likely region to be the true reflectivity et al., 2010) typically spaced in the order of the seismic
series. By isolating the maximum energy and computing sampling interval. At least two framework horizons are
the inverse CWT, one can reconstruct a good approxi- needed for bounding the calculations in between them.

Figure 5. Well-to-seismic tie at well P1 with synthetic constructed using the extracted wavelet. Upper panel: well-log suite of well
P1. TM corresponds to the top, and BM corresponds to the base of the Marlim reservoir. Marlim turbidite shows high porosity, low
density, and acoustic impedance. Lower panel: synthetic and observed seismic amplitude, extracted wavelet and correspondent
amplitude spectrum.

Interpretation / August 2014 T147


In the present work, we mapped four main framework mic inversion, this trend is added to seismic inversion
horizons using the stratigraphic control given by the results to obtain an absolute acoustic impedance esti-
four wells, from bottom to top (Figure 6a): Albian, Blue mate (Figure 6d). In Figure 7, we display the 0.83 cor-
Marker, Oligocene, and Miocene. These horizons corre- relation coefficient between the acoustic impedance
spond to the main regional stratigraphic units in the computed from the well logs and the recovered abso-
study area. The Albian horizon corresponds to the lute acoustic impedance at P1 well.
top of Macaé Formation, composed of shallow water
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

carbonate section (Candido and Cora, 1978). The lower Porosity prediction
Oligocene Blue Marker is a basinwide maximum flood- Porosity prediction from seismic data constrained by
ing surface. The Oligocene horizon represents the tur- well control can significantly improve reservoir charac-
bidite system of the Carapebus Formation, and the terization (Angeleri and Carpi, 1982; Dolberg et al.,
Miocene horizon is on the top of that sequence. A 2000). Seismic porosity maps are useful in providing in-
dip-steered autotracker algorithm was used (de Groot formation about the porosity spatial variation away
et al., 2010). Beside speed, awareness of fault detection from the well control. The estimated porosities are rou-
is a significant advance of the algorithm as it automati- tinely used as steering constraints to calculate, trends,
cally stops at mapped fault planes. Multihorizon large-scale volumetric estimations, and also reveal
tracking is then applied to map several horizons within sweet spots within the reservoir.
the confined region between the framework horizons The relationship between porosity and acoustic
(Figure 6b). impedance in the study area can be estimated by cross-
To establish a background model, we filtered the plotting these attributes at all four available wells (Fig-
acoustic impedance wells logs in the 0–10 Hz low- ure 8a). To that end, we estimated a linear relationship
frequency range and then interpolated the filtered for the whole data set that can be described by a single
acoustic impedance along the mapped horizons with ac- linear regression having a correlation coefficient of
curate stratigraphic control (Figure 6b). Typical exam- 0.75. Data points in Figure 8a are color coded by
ples of interpolation algorithms are inverse distance gamma-ray logs, a good lithology indicator. As shown
and kriging (Huck et al., 2010). In the present work, in Figure 5, the Marlim reservoir is easily identified
we have chosen the inverse distance-based algorithm. as a low-density, high-porosity zone. In Figure 8a, we
As a result, we obtained the detailed low-frequency separated identified gamma-ray data points with high
model shown in Figure 6c. This model contains only in- porosities (0.30 threshold) as associated to low imped-
formation about the low-frequency trend that is not ance values. In Figure 8b, we plot the selected points
available in the seismic amplitude data. After the seis- against their position at the P1 well. One can note that

Figure 6. HorizonCube high-resolution acoustic impedance inversion workflow. (a) Four main mapped framework horizons,
(b) numerous autotracked horizons, (c) realistic low-frequency model, and (d) high-resolution absolute acoustic impedance.

T148 Interpretation / August 2014


these estimates correlate to Marlim sandstones. The inversion workflow, it is possible to clearly distinguish
same correlation can be established at all other wells. between reservoir and nonreservoir facies.
Our results are in accordance with previous geologic
knowledge of the studied Marlim sandstone Sansonow- Conclusions
ski et al. (2007). These authors define the reservoir fa- Seismic inversion is a key tool for reservoir charac-
cies as a highly friable clean sandstone with low silt, terization because it can provide petrophysical informa-
clay, and calcite content. Core analyses at various wells tion that better describes the subsurface. One important
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

register an average porosity of 30% and a mean per- aspect in the inversion procedure is the correct choice
meability of 2000 mD for the Marlim sandstone (Sanso- of the background model because it can minimize the
nowski et al., 2007). inherent ambiguity of the process. This is particularly
We used the calculated relationship ϕ ¼ 0.53 − relevant in complex geologic areas where a standard
ð4.36 × 10−5 ÞAI, where ϕ is the porosity and AI is the model building workflow may introduce model errors.
acoustic impedance, to convert the acoustic impedance In the present work, we applied a high-resolution
cube to a porosity cube. In Figure 9, we show a high- deterministic inversion workflow to estimate acoustic
resolution image of the predicted porosity distribution impedance and porosity of turbidites reservoirs at
along the Oligocene horizon. The high-porosity bodies Marlim Field, Campos Basin, Brazil. The proposed
(ϕ > 0.30) correspond to Marlim sandstones, whereas workflow incorporates the interpreter’s stratigraphic
the surrounding shale and shaly sand facies systemati- knowledge to better estimate the low-frequency model.
cally show lower porosity trends (ϕ < 0.30). All drilled A priori independent information of four wells in the
wells in the study area penetrated the Marlim horizon at study area provided important constraints to several
high-porosity zones (Figure 9). With a high-resolution steps of the proposed workflow. The main reservoirs
picture of the subsurface as provided by the proposed facies (clean sandstones) of the Marlim Field were

Figure 8. (a) Relation between porosity and acoustic imped-


ance for wells P1, P2, P3, and P4. The best-fit linear has a cor-
relation factor (r) of 0.75. Data points are color coded by
gamma-ray log values (color scale shown in the figure).
Figure 7. Upper panel: crossplot between inverted acoustic The black rectangle separates a region with low impedance
impedance (IP-abs-mm) versus upscaled well-log acoustic and high porosity values. (b) Picked points in Figure 8a plot-
impedance (IP-0-40-MD) at P1 well. Lower panel: comparison ted at their actual position along well P1. The background im-
between the recovered acoustic impedance along an arbitrary age is the recovered acoustic impedance magnified at the
section and the upscaled well-log acoustic impedance at P1 Marlim turbidites interval. The selected points coincide with
well. the reservoir facies.

Interpretation / August 2014 T149


For this, consider a seismic trace sðtÞ
as a function of time and its Fourier
transform ^ sðf Þ ¼ F s ðf Þ as a complex
function of frequency f . Whereas in
the convolutional model, the seismic
trace is represented by the convolution
of the reflectivity series and a wavelet
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

function, the goal is to develop a decon-


volution filter in the wavelet domain.
The Morlet wavelet is well suited to
the spectral decomposition of the seis-
mic signal and can be analytically ex-
pressed in time and frequency domain
as the pair of Fourier transform given by

1 i2πf 0 t −t2 ∕2
ψðtÞ ¼ e e ↔ψ^ ðf Þ
π 1∕4
pffiffiffi 2
¼ π 1∕4 2e−1∕2ð2πf −2πf 0 Þ ; (A-1)

where f 0 is a fixed parameter that rep-


resents the center frequency of the
wavelet. The wavelet transform of the
seismic trace s~ ðτ; αÞ ¼ W s ðτ; αÞ is a com-
plex function of time τ and a variable
scale α defined by α ¼ f ∕f 0 . The CWT
can be expressed as the inverse Fourier
Figure 9. Map of the predicted porosity distribution along Marlim horizon: P1, transform of the product of ^ s and ψ c
P2, P3, and P4 are the available wells. The Marlim turbidites (clean sandstones)
are clearly highlighted as high-porosity bodies. Shale and shaly sand (nonreser-
(Addison, 2002):
voir) facies show lower (below 0.28) porosities. Coordinates are in UTM, zone Z
pffiffiffi ∞
39S. s~ ðτ; αÞ ¼ α ^ c αei2πf τ df ; (A-2)
sðf Þψ
−∞

where ^ c are, respectively, the Fourier transform


s and ψ
identified as high-porosity zones (porosities above
of the input seismic trace sðtÞ and the complex conju-
30%), whereas the nonreservoirs facies systematically
gate of the Morlet wavelet.
show low-porosity values.
To establish the relationship between the represen-
tation of the CWT and nonstationary filters, we define a
Acknowledgments nonstationary filter Gðf ; tÞ as a function of time and fre-
We would like to thank Interpretation’s editor, Y. quency. An example of that, might be a nonstationary
Sun, associate editors R. Newrick, D. Herron, J. wavelet operating in reflectivity series represented by
Sumner, Y. DelMoro, R. Wegner, and four anonymous s0 ðtÞ. Consider s^0 ðtÞ as the Fourier transform of the re-
reviewers for the valuable suggestions that helped to flectivity series. By applying the nonstationary filter G
improve the text. We thank dGB Earth Sciences for and the inverse Fourier transform, we obtain
making freely available the open-source seismic inter- Z ∞
pretation software OpendTect. We acknowledge LA-
sðtÞ ¼ s^0 ðtÞGðf ; tÞei2πf t df : (A-3)
GEX/UERJ and Invision Geophysics for making −∞
available its computational park. P. Menezes thanks the
support provided by research grant from CNPq. Data In the wavelet domain, we can apply an equivalent op-
set made available by ANP under the support of the eration to retrieve the original reflectivity series s0 ðtÞ by
CNPq Project coordinated by A. Evsukoff. We also designing an inverse filter G−1 ≡ Θðτ; αÞ. Then, the de-
thank the additional support by CNPq project sired output ðs0 ðtÞÞ without the effect of a transient
no. 470742/2011-9. wavelet is obtained through the inverse wavelet trans-
form Addison (2002):
Appendix A Z Z
1 dτdα
sðtÞ ¼ s~0 ðtÞΘðτ; αÞψ τ;α ðtÞ 2 ; (A-4)
High-resolution inversion c α τ α
Braga (2011) uses a formulation in the wavelet do-
main, in which a series of filters to trace the trace op- where c is a normalization constant to balance the
erations can be designed and efficiently implemented. wavelet energy. In practice, because the CWT is

T150 Interpretation / August 2014


Their use in exploration and produc-
tion: 31st Annual, 496–533.
Brouwer, F., A. Huck, N. Hemstra, and I.
Braga, 2012, Extracting full-resolution
models from seismic data to minimize
systematic errors in inversion: Method
and examples: The Leading Edge, 31,
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

546–554, doi: 10.1190/tle31050546.1.


Brown, A. R., and W. L. Abriel, 2014, The
polarity of zero-phase wavelets: Inter-
pretation, 2, no. 1, 19F, doi: 10.1190/
INT2013-1202-PS.1.
Bruhn, C., J. T. Gomes, C. Lucchese, Jr.,
and P. Johann, 2003, Campos Basin:
Figure A-1. Illustration of the retrieving of a reflectivity series via wavelet trans- Reservoir characterization and man-
form. (a) Original reflectivity series, (b) seismic trace produced by the convo-
agement-historical overview and future
lution of the original series of Figure 6a with a 40-Hz wavelet, and
(c) retrieved reflectivity series after applying the inverse wavelet transform. challenges: Presented at OTC.
Candido, A., and C. A. Cora, 1992,
The Marlim and Albacora giant fields,
Campos Basin, offshore Brazil: AAPG Memoir 54,
nonorthogonal, the spectral representation is redun- 123–135.
dant in time and frequency. This redundancy allows Carvalho, M. V. F., L. F. De Ros, and N. S. Gomes, 1995,
the calculation of the inverse transform by sum- Carbonate cementation patterns and diagenetic reser-
ming the real part at all frequencies (Torrence and voir facies in the Campos Basin Cretaceous turbidites,
Compo, 1998). offshore eastern Brazil: Marine and Petroleum Geology,
Figure A-1 illustrates the process of wavelet trans- 12, 741–758, doi: 10.1016/0264-8172(95)93599-Y.
form via deconvolution, showing a very accurate recov- Cobbold, P. R., K. E. Meisling, and V. S. Mount, 2001,
ery of the original reflectivity series, which can be
Reactivation of an obliquely rifted margin, Campos
confirmed by comparing the original (Figure A-1a) with
and Santos basins, southeastern Brazil: AAPG Bulle-
the recovered one (Figure A-1c). Once we recovered
tin, 85, 1925–1944, doi: 10.1306/8626D0B3-173B-11D7-
reflectivity, the impedances can be calculated by simple
8645000102C1865D.
integration of the trace (Braga, 2011).
Contreras, A., C. Torres-Verdin, W. Chesters, K. Kvien, and
M. Globe, 2005, Joint stochastic inversion of petrophys-
ical logs and 3D pre-stack seismic data to assess the
References spatial continuity of fluid units away from wells: Appli-
Addison, P. S., 2002, The illustrated wavelet trans- cation to a Gulf-of-Mexico deepwater hydrocarbon res-
form handbook: Introductory theory and applications ervoir: Presented at SPWLA 46th Annual Logging
in science, engineering, medicine and finance: CRC Symposium.
Press. Davogustto, O., M. C. de Matos, C. Cabarcas, T. Dao, and
Angeleri, G., and R. Carpi, 1982, Porosity prediction from K. J. Marfurt, 2013, Resolving subtle stratigraphic
seismic data: Geophysical Prospecting, 30, 580–607, features using spectral ridges and phase residues: Inter-
doi: 10.1111/j.1365-2478.1982.tb01328.x. pretation, 1, no. 1, SA93–SA108, doi: 10.1190/INT-2013-
Avseth, P., T. Mukerji, and G. Mavko, 2005, Quantitative 0015.1.
seismic interpretation: Applying rock physics tools de Groot, P., A. Huck, G. de Bruin, N. Hemstra, and J. Bed-
to reduce interpretation risk: Cambridge University ford, 2010, The horizon cube: A step change in seismic
Press. interpretation!: The Leading Edge, 29, 1048–1055, doi:
Bosch, M., T. Mukerji, and E. F. Gonzalez, 2010, Seismic 10.1190/1.3485765.
inversion for reservoir properties combining statistical Dolberg, D. M., J. Helgesen, T. H. Hanssen, I. Magnus, G.
rock physics and geostatistics: A review: Geophysics, Saigal, and B. K. Pedersen, 2000, Porosity prediction
75, no. 5, 75A165–75A176, doi: 10.1190/1.3478209. from seismic inversion, Lavrans Field, Halten Terrace,
Braga, I. L., 2011, Tecnicas multiespectrais aplicada a Norway: The Leading Edge, 19, 392–399, doi: 10.1190/1
fluxos de inversao e caracterizacao de reservatorios .1438618.
de hidrocarbonetos: Ph.D. thesis, Universidade Estad- Fainstein, R., G. Jamieson, A. Hannan, N. Biles, A. Krueger,
ual do Norte Fluminense. and D. Shelander, 2001, Offshore brazil santos basin ex-
Brouwer, F., and A. Huck, 2011, An integrated workflow to ploration potential from recently acquired seismic data:
optimize discontinuity attributes for the imaging of Presented at 7th International Congress of the Brazilian
faults, in Attributes: New views on seismic imaging — Geophysical Society.

Interpretation / August 2014 T151


Freire, W., 1989, Campos basin deepwater giant fields: Pre- Pendrel, J. V., and P. V. Van Riel, 1997, Estimating porosity
sented at OTC. from 3D seismic inversion and 3D geostatistics: 67th An-
Goodway, B., T. Chen, and J. Downton, 1997, Improved nual International Meeting, SEG, Expanded Abstracts,
AVO fluid detection and lithology discrimination using 834–837.
Lamé petrophysical parameters; λρ, μρ, λμ fluid stack, Peres, W. E., 1993, Shelf-fed turbidite system model and its
from P and S inversions: 67th Annual International application to the Oligocene deposits of the Campos
Meeting, SEG, Expanded Abstracts, 183–186. Basin, Brazil: AAPG Bulletin, 77, 81–101.
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Guardado, L. R., A. R. Spadini, J. S. L. Brandão, and M. R. Qayyum, F., and P. de Groot, 2012, Seismic dips help
Mello, 2000, Petroleum system of the Campos Basin, unlock reservoirs: American Oil & Gas Reporter, 60,
Brazil, in M. R. Mello, and B. J. Katz, eds., Petroleum 75–79.
systems of South Atlantic margins: AAPG Memoir 73, Ribeiro, N., 2012, Joint inversion of multi-component seis-
317–324. mic data for reservoir characterization of an offshore
Huck, A., G. Quiquerez, and P. de Groot, 2010, Improving Campos Basin field, Brazil: First Break, 30, 87–91.
seismic inversion through detailed low frequency model Russell, B., and M. N. Toksöz, 1991, Comparison of post-
building: 72nd Annual International Conference and Ex- stack seismic inversion methods: 61st Annual Inter-
hibition, EAGE, Extended Abstracts, F026. national Meeting, SEG, Expanded Abstracts, 876–878.
Johann, P., R. Sansonowski, R. Oliveira, and D. Bampi, Salleh, M. S., and S. S. Ronghe, 1999, Reservoir characteri-
2009, 4D seismic in a heavy-oil, turbidite reservoir off- zation on thin sands in South West Ampa 21 area
shore Brazil: The Leading Edge, 28, 718–729, doi: 10 (BLK11) using seismic inversion: 69th Annual Inter-
.1190/1.3148415. national Meeting, SEG, Expanded Abstracts, 1568–
Kelamis, P. G., N. A. Akbar, M. H. Al-Fares, R. B. Lathon, 1571.
and J. J. Kim, 1995, A deterministic approach to reser- Sansonowski, R. C., R. M. de Oliveira, N. M. d. S. R. Júnior,
voir description via 3-D poststack inversion and petro- D. Bampi, and L. F. C. Junior, 2007, 4D seismic interpre-
physical analysis: 65th Annual International Meeting, tation in the Marlim Field, Campos Basin, offshore Bra-
SEG, Expanded Abstracts, 322–324. zil: 77th Annual International Meeting, SEG, Expanded
Latimer, R. B., R. Davidson, and P. Van Riel, 2000, An in- Abstracts, 2872–2877.
terpreter’s guide to understanding and working with Souza, J., J. Scarton, A. Candido, C. E. Cruz, and C. A. G.
seismic-derived acoustic impedance data: The Leading Cora, 1989, The Marlim and Albacora Fields: Geophysi-
Edge, 19, 242–256, doi: 10.1190/1.1438580. cal, geological, and reservoir aspects: Presented at
Leyden, R., H. Asmus, S. Zembruscki, and G. Bryan, 1976, OTC.
South Atlantic diapiric structures: AAPG Bulletin, 60, Tetyukhina, D., L. J. van Vliet, S. M. Luthi, and K. Wape-
196–212. naar, 2010, High-resolution reservoir characterization
Mallat, S., and S. Zhong, 1992, Characterization of signals by an acoustic impedance inversion of a Tertiary deltaic
from multiscale edges: IEEE Transactions on Pattern clinoform system in the North Sea: Geophysics, 75,
Analysis and Machine Intelligence, 14, 710–732, doi: no. 6, O57–O67, doi: 10.1190/1.3506039.
10.1109/34.142909. Torrence, C., and G. P. Compo, 1998, A practical guide to
Mello, M. R., E. A. M. Koutsoukos, W. U. Mohriak, and G. wavelet analysis: Bulletin of the American Meteorologi-
Bacoccoli, 1994, Selected petroleum systems in brazil, cal Society, 79, 61–78, doi: 10.1175/1520-0477(1998)
in L. B. Magoon, and W. G. Dow, eds., The petroleum 079<0061:APGTWA>2.0.CO;2.
system — From source to trap: AAPG Memoir 60, 499– Wapenaar, K., R. Ghose, G. Toxopeus, and J. Fokkema,
499. 2005, The wavelet transform as a tool for geophysical
Merletti, G., and C. Torres-Verdín, 2006, Accurate detec- data integration: Integrated Computer-Aided Engineer-
tion and spatial delineation of thin-sand sedimentary se- ing, 12, 5–23.
quences via joint stochastic inversion of well logs and White, R., 1997, The accuracy of well ties: Practical proce-
3D prestack seismic amplitude data: Presented at SPE dures and examples: 67th Annual International Meeting,
Annual Technical Conference and Exhibition. SEG, Expanded Abstracts, 816–819.
Mohriak, W. U., J. M. Macedo, and R. T. Castellani, 1996,
Salt tectonics and structural styles in the deep-water
province of the Cabo Frio region, Rio de Janeiro, Brazil, Tatiane M. Nascimento received a
in M. P. A. Jackson, D. G. Roberts, and S. Snelson, eds., B.S. (2011) and an M.S. (2013) in geol-
Salt tectonics: A global perspective: AAPG Memoir 65, ogy from Rio de Janeiro State Univer-
273–304. sity, Brazil. In 2012–2013, she worked
Nascimento, T. M., P. T. L. Menezes, and I. G. Braga, 2013, as a consultant at Invision Geo-
physics. Her research interests in-
Reservoir characterization by acoustic inversion Mar-
clude seismic inversion and seismic
lim Field: Presented at 13th International Congress of interpretation. She is a member of
the Brazilian Geophysical Society. the Brazilian Geophysical Society.

T152 Interpretation / August 2014


Paulo T. L. Menezes received a B.S. (1986) in geology Igor Braga received a Ph.D. in (2011)
from Rio de Janeiro State University, Brazil, an M.S. (1990) in exploration and petroleum engi-
in geophysics from the Federal University of Pará, Brazil, neering from Universidade Estadual
and a Ph.D. (1996) in geophysics from the National do Norte Fluminense, Brazil. His re-
Observatory of Rio de Janeiro. In 1996–1997, he worked search interests include seismic
as a consultant at Geomag Aerolevantamentos Ltda. Since processing, seismic inversion, and
1997, he has been a professor at Rio de Janeiro State Uni- reservoir characterization. He is the
versity. He also works as a geophysicist at Petrobras. His cofounder of Invision Geophysics,
Downloaded 06/11/14 to 179.218.11.75. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

research interests include interpretation of potential-field, where he is the CEO. He is a member of SEG, EAGE,
seismic, and electromagnetic data. He is a member of SEG, and the Brazilian Geophysical Society.
EAGE, and SBGf.

Interpretation / August 2014 T153

View publication stats

You might also like