Professional Documents
Culture Documents
net/publication/368674339
CITATIONS READS
33 262
5 authors, including:
All content following this page was uploaded by Pooria Kianoush on 03 March 2023.
A R T I C L E I N F O A B S T R A C T
Keywords: Seismic velocity is considered the best attribute related to formation pressure changes. Integrating seismic at
Seismic attributes tributes and well-logging data through seismic inversion predicts the reservoir characteristics across the field
Post-stack data with the highest accuracy. This study especially presents seismic velocity for the whole south Azadegan Field in
Seismic inversion
SW Iran for carbonate formations. The considered dataset includes 3D seismic data, vertical seismic profiling
Vertical seismic profiling (VSP)
(VSP), logging data of 23 wells, and geological information. Here, we estimated the interval velocity using post-
Interval velocity cube
stack migration velocity, seismic inversion, and the relationship between the acoustic impedance (AI) model and
the sonic log to predict formation pressure. As a result, the correlation coefficient of 0.71 and a high inversion
accuracy (8.76% relative error) is concluded. The actual and predicted P-wave (Vp) correlation coefficient is
calculated as 0.74 and all sevens as 0.79 using an AI seismic attribute. Thus, the estimated Vp agrees with the
original well-log values. Inverted AI cubes in the deeper formations of the field are about 8000-15000 [(m/s)*(g/
cm3)], which could be referred to as calcareous formations. The correlation of the Vp cube resulting from the
Sequential Gaussian simulation (SGS) considering co-kriging with the AI, with the initial velocity cube using the
inverse distance weighted (IDW) method being 0.54 is more than the same method applied with interval
migration velocity trend in co-kriging. The anisotropy of the final Vp cube for the vertical variogram range is
96m, and for major and minor directions is 11850 m.
* Corresponding author: Corner of Kausar Blvd., Deh Haghi (Ahang) St., Southern Nabard Blvd., Pirouzi St., Tehran, Iran, Postal code: 1777613651, Contact
number: +9821-33722831-7, Fax: +9821-33717140 http://fte.azad.ac.ir/
E-mail addresses: Pooria.kianoush@gmail.com (P. Kianoush), ghodratollah46@gmail.com (G. Mohammadi), saa.hosseini@srbiau.ac.ir (S.A. Hosseini),
keshavarzn@ripi.ir (N.K.F. Khah), P_Afzal@azad.ac.ir (P. Afzal).
https://doi.org/10.1016/j.ringps.2023.100051
Received 17 December 2022; Received in revised form 10 February 2023; Accepted 11 February 2023
Available online 20 February 2023
2666-8289/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
2
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
deterministic method in all 23 wells. The correlation matching was 2 Amplitude: stratigraphy information, acoustic impedance,
99.6%, and the uncertainty described as the relative error was about 3 Frequency: information about the characteristics of the reservoir,
8.5%, which was unique due to the very high volume of data and the 4 Energy absorption: fluid information and permeability (Rai et al.,
very large number of wells. Also, other innovations of the present 2020).
research include using new capabilities of Petrel (2016) software for
conditional programming (e.g., sequential and nested conditional ex In reflection seismology, a seismic attribute is a quantity extracted or
pressions to combine logs and cubes in a single model and remove out- derived from seismic data that can be analyzed to enhance the infor
of-range values). Some conditional codes for complete Compressional mation that might be more subtle in a traditional seismic image, leading
and Shear velocity logs are shown in Appendix A. to a better geological or geophysical interpretation of the data (Kad
khodaie & Kadkhodaie, 2022).
2. Methodology Seismic attributes of Post-stacking can be extracted on a time slice or
between two horizons. Each seismic attribute could be surfaced or
2.1. Seismic reflections volumetric. Surface attributes are calculated on a surface or a horizon.
Volumetric attributes are calculated between two definite times or be
Seismic reflections are some function of acoustic impedance (veloc tween two interpreted horizons. These attributes help us to interpret
ity in density), affecting reservoir pressure. On the other hand, the type seismic data from different structural and stratigraphic aspects. In this
of formation fluid affects the sound wave velocity. Shear and compres way, they accompany us in a better understanding of reservoir charac
sional waves (Vp and Vs) respond differently to fluids and rocks and teristics (Yadav et al., 2022). Acoustic impedance obtained from seismic
different reservoir pressures. This phenomenon can have the following inversion can also be considered an attribute and was among the first
two uses: developed (Li & Liu, 2022). Common attributes include coherence, az
imuth, dip, instantaneous amplitude, response amplitude, response
1 Prediction of abnormal pressure from velocity data before drilling phase, instantaneous bandwidth, amplitude versus offset (AVO), and
2 Examining fluid movement and dynamic reservoir pressure changes spectral decomposition. A seismic attribute that can indicate the pres
using 4D seismography (Diethart-Jauk & Gegenhuber, 2018; ence or absence of hydrocarbons is known as a direct hydrocarbon in
Maurya et al., 2020). dicator (Haque et al., 2022).
3
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
impedance is the resistance of rocks to the propagation of elastic waves. 2016; Liu et al., 2022). Seismic inversion combines seismic and well data
These attributes represent the lithology of the layer and, to some extent, to predict rock properties (lithology, fluid content, porosity) across a
indicate the reservoir fluid in the layers. As Fig. 1, different rocks have survey. These rock properties can be used to identify hydrocarbon and
different acoustic impedance values, but acoustic impedance values for reservoir. Seismic inversion removes the imprints of the wavelet in the
different rocks can overlap (Maurya et al., 2020). seismic data by deconvolution and then by converting the result into
Observing the changes in acoustic impedance (AI) inside a layer with impedance (Lindseth, 1979; Veeken et al., 2020; Veeken & Da Silva,
a specific lithological feature makes it possible to understand the 2004). Seismic inversion workflow and mechanism are shown in Fig. 2
changes in the rock facies inside the layer. Thus the acoustic impedance and Appendix Fig. B.1.
acts as a seismic attribute. Because the velocity of the wave passing
through the layers changes with increasing depth and the lithological 2.3.1. Wavelet estimation
change, denser and stronger rocks increase the velocity of the wave The modern seismic inversion methods require seismic data and a
(Maurya et al., 2020). wavelet estimated from the data. Typically, a reflection coefficient series
On the other hand, increasing the porosity of the reservoir reduces from a well within the boundaries of the seismic survey is used to esti
the wave velocity and the acoustic impedance. The effect of gas also has mate the wavelet phase and frequency. Accurate wavelet estimation is
a significant role in the P-wave velocity in the layers (Adelinet et al., critical to the success of any seismic inversion. The inferred shape of the
2018; Maurya et al., 2020). seismic wavelet may strongly influence the seismic inversion results
and, thus, subsequent assessments of the reservoir quality.
2.3. Seismic Inversion Wavelet amplitude and phase spectra are estimated statistically from
either the seismic data alone or from a combination of seismic data and
Seismic inversion transforms seismic reflection data into a quanti well control using wells with available sonic and density curves. After
tative rock-property description of a reservoir. It may be pre-stack, post- the seismic wavelet is estimated, it is used to estimate seismic reflection
stack, deterministic, random, or geostatistical, and typically includes coefficients in the seismic inversion (Conzález-Veloza et al., 2020;
other reservoir measurements such as well logs and cores (Chen et al., Fichtner & Trampert, 2011).
4
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
5
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 5. Structural map of the Abadan Plain. Major anticlines appear as elongated domes. The location of maps and seismic profiles discussed are outlined (Abdollahie
Fard, 2019).
6
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 7. Three-dimensional geological model of South Azadegan Field using seismic sections and drilling data along with the location of used wells.
Eq. (3) integrates the seismic traces (Sk) and exponentiates the result interval velocity of the depth-time pairs is calculated using an inversion
to provide impedance traces. Variable γ is a scaling factor for the method designed to minimize errors in the travel time selection.
approximation in BLI. Applying this recursive equation enables seismic Assuming that si =1/vi is the interval slowness of layer i, vi is the interval
data to be directly inverted for acoustic impedance. However, these velocity of layer i, and ti is the measured travel time of the receiver to the
algorithms have various problems (Lindseth, 1979; Maulana et al., 2016; depth of zi (i = 1, 2… N); ti, si, and zi are related by the following linear
Russell, 2017). relation (Eq. (4)).
⎡ ⎤ ⎡ ⎤⎡ ⎤
t1 Z1 0 … 0 S1
2.3.4. Pre-stacking Inversion ⎢ t2 ⎥ ⎢ Z1 Z2 … 0 ⎥⎢ S2 ⎥
Pre-stacking inversion changes seismic into P-impedance, S-imped ⎢ ⎥ ⎢
⎢ . ⎥ ⎢ . . . . ⎥⎢ . ⎥
⎥⎢ ⎥
ance, and density by integrating well and seismic data (Dufour et al., ⎢ ⎥=⎢
⎢ . ⎥ ⎢ . . . . ⎥⎢ . ⎥
⎥⎢ ⎥ (4)
⎢ ⎥ ⎢ ⎥⎢ ⎥
2002). The two techniques used for pre-stacking are simultaneous and ⎣ . ⎦ ⎣ . . . . ⎦⎣ . ⎦
elastic inversion. Note that those technics required global wavelets and a tN Z1 Z2 … ZN SN
background model (Fig. 4). Differences and Similarities of Post and
Pre-stack methods are shown in Table 1. The interval velocity is found by reversing the travel time of the
borehole velocity taken from the well by minimizing the absolute value
of x2-1 (Cordier, 1985).
2.4. Seismic Wave Velocity Resources
( )2
1 ∑N
timeasured − tipredicted
Seismic velocities are usually considered to be seismic properties. In x2 = (5)
N i=1 σi
the Earth’s upper crust, however, these are influenced strongly by
confining stresses, pore pressures, fluid saturations, temperature, and Here, the measured t-time is predicted to be the transition time
porosity, to name only a few. This contribution focuses on how these recorded at depth zi, and the predicted t-time transition is predicted
different factors interact, resulting in a seismic velocity that will finally using Eq. (5).
be observed. Some seismic velocity resources are Vertical Seismic For estimating the pore pressure with velocity data, the relationship
Profiling (VSP), Sonic Log, Stacking velocity, tomography, and acoustic between effective stress and velocity in sediments under normal pres
impedance (AI) inversion. The following two seismic velocities used in sure has been proposed by Bowers (1995) as Eqs. (6) and (7):
this study are examined (Bahmaei & Hosseini, 2020; Etminan et al.,
2012; Rointan et al., 2021). V = V0 + Aσ B (6)
[ ]1
2.4.1. Vertical Seismic Profiling (VSP) V − V0 B
σ= (7)
Vertical seismic profiling is one of the seismic methods in which A
seismic waves generated at the ground surface are recorded by geo
Where V0 is the velocity of unconsolidated fluid-saturated sediments,
phones located at different depths of the well. The VSP method is based
and A and B describe the variation in velocity with increasing effective
on seismic wave propagation in heterogeneous environments theories
stress (σ) and can be derived from offset well data (Bowers, 1995; 2002).
and can reconstruct the seismic wave propagation field and calculate
seismic data at the wellbore. This method can also be used to imagine
2.4.2. Acoustic impedance inversion
around the well or to reconstruct between several wells.
One of the essential steps for acoustic impedance inversion is to build
Also, using the information from Sonic and density logs, subsurface
an initial model. This model is based on sonic logs (Bahmaei & Hosseini,
construction models can be prepared. In this method, the transmission
2020); also, acoustic impedance is one of the critical properties of rock
time of the seismic wave is less than the surface seismic surveys, so the
that can be obtained from seismic data during inversion. Seismic
signal-to-noise ratio increases because the receivers are close to the
inversion is a method of restoring lost frequency data by eliminating the
reflective surface. The resolution of reflective data is higher than that of
wavelet effect and completely reconstructing the earth’s frequency
seismic data. However, zero-offset data is slightly spread around the
band. Seismic inversion is to convert limited band seismic data to
well, and its data is usually suitable for drilling information (Paul et al.,
broadband acoustic impedance grids at each normal move-out (NMO) or
2021).
common midpoint (CMP) survey. Therefore, seismic inversion is a pro
cess that tries to estimate the acoustic impedance model close to the
2.4.1.1. Determine interval velocity using velocity data in wells. The
7
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Table 2 seismic data from the Azadegan anticline demonstrate the un
Division of South Azadegan Field based on the average cubic thickness of conformities and erosional surfaces due to the uplifting of basement-
geological layers. cored horsts.
Row Formation Formation Formation Average Dominant
Top (m) Base (m) Thickness Lithology 3.2. Constructing of the structural geological model
(m)
1 Aghajari 0 1272.3 1272.3 Marl and After defining the network range, horizontal (I) and vertical (J) axes,
sandstone number of nodes, and well’s layout, the 3D geological structure of the
2 Gachsaran 1272.3 1630.65 358.35 Anhydrite
field was modeled. (Fig. 6). Then, the In-line and X-line ranges of the
and
Claystone amplitude values were added, and a seismic cube with post-stack data
3 Asmari and 1630.65 2368.25 737.6 Sandstone was created by the arithmetic seismic resampling method of Petrel
Pabdeh and (2016) Software.
Limestone The Studied field formations are modeled based on the interpretation
4 Gurpi 2368.25 2590.05 221.8 Limestone
5 Tarbur 2590.05 2757.85 167.8 Limestone
of time-domain seismic horizons data and correlated with drilling
(Member) and marl geological information. Depth-domain seismic horizons have been
6 Ilam and 2757.85 2866.05 108.2 Limestone constructed as separate surfaces from the Aghajari to the Gotnia For
Laffan and mations. Due to the lack of complex fault systems in the area, the
claystone
geological model has been built with a simple network of Petrel (2016)
7 Sarvak 2866.05 3506.9 640.85 Limestone
8 Kazhdumi 3506.9 3733.95 227.05 Shale, software (Fig. 7 and Table 2).
Limestone
and 4. Results
Sandstone
9 Dariyan 3733.95 3896 162.05 Limestone
4.1. VSP interval velocity model
and marl
10 Gadvan 3896 3966.55 70.55 Marl, shale
and The interval velocity model was prepared using the relationship
limestone between checkshots and Vertical Seismic Profiling (VSP) velocity data
11 Khalij 3966.55 4071 104.45 Sandstone
and with depth changes in wells with information according to Eq. (8),
(member) and
Limestone
and the coefficients V0 and K were determined with a correlation coef
12 Upper 4071 4228.05 157.05 Limestone ficient of 0.95 (Fig. 10).
Fahliyan
13 Lower 4228.05 4589.1 361.05 Limestone V = V0 + K ∗ Z (8)
Fahliyan
In this relation, K is the constant conversion factor of changing deep
14 Garau 4589.1 4783 193.9 Limestone
and horizon layers to the average interval velocity, and V0 is the surface
claystone layer velocity.
15 Gotnia 4783 4931 148 Anhydrite The data from Ilam, Sarvak, Kazhdumi, Gadvan, Fahliyan, and Garau
and
Formations (from the upper Cretaceous to the Jurassic horizon) has been
limestone
16 Najmeh 4931 4959 28 Anhydrite used to match the data of deep seismic horizons.
and
V = V0 = Vint (9)
limestone
17 Sargelu 4959 5068 109 Limestone In order to convert the depth seismic horizon layers to average ve
and shale
18 Alan 5068 5107 39 Anhydrite
locity, due to the lack of surface seismic horizon in the surface Aghajari
and layer used Eq. (9) and for other layers from Eq. (8) with constant values
limestone of V0 = 1984.61 and K = -0.3721 calculated according to the table
19 Muss 5107 5199 92 Limestone below. Finally, the average velocity of each formation is calculated using
20 Neyriz 5199 5590 391 Limestone
Eq. (10), and the average velocity map of each layer is prepared sepa
and
anhydrite rately; its results are summarized in the table below.
(m)
Vavg = 2000 ∗ Z (m)/TWT (ms) (10)
s
actual earth model based on the initial seismic models and estimated
wavelets (Vicêncio et al., 2022). In this formula, TWT is the wave travel time in milliseconds, Z is the
depth in meters, and Vavg is the average layer velocity in meters per
3. Geological Setting second. A VSP Log and initial model sample in the Azadegan Field are
shown in Figs. 8 and 9.
3.1. Geological model based on seismic interpretation Based on this, the average velocity of the surface Aghajari Formation
with a constant rate of 1984.6 m/s has been calculated, the highest
The Azadegan oilfield is located in the transition zone between the average velocity in the range of 2760- 2900 m/s in the northeast side is
Arabian plate and the Zagros basin (Fig. 5). The Zagros orogeny has related to the Gotnia Formation, and the lowest is related to the Gach
changed the shape and fracture of the subsurface layers due to the saran Formation with 2180-2250 m/s in the southwest direction of the
presence of shale and marl in the Cenozoic formations and the reduction studied area (Table 3). An example of the layer velocity maps in the
of tectonic stress. Although they still have a sealing role in restraining lower Fahliyan Formation is presented in Fig. 11.
the migration and vertical loss of oil (Du et al., 2016). The seismic profile According to Fig. 14, the average velocity of the lower Fahliyan
across Azadegan high shows a steep fault system in the Jurassic and Formation is between 2330 and 2760 m/s, and the highest values of that
underlying sedimentary rocks (Abdollahie Fard, 2019; Morgan, 1999). are visible in the northeast with red contour and the lowest in the
The Azadegan dome is a complex horst. Seismic data of the Azadegan southwest with purple contour.
structure show steep faulting in the core of the anticline. These faults die
up in the upper Jurassic Gotnia Formation. Exploratory drilling and
8
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 8. a) VSP travel times (Two Way Time), b) Average and Interval Velocity of Well A-001.
Fig. 9. Location of studied wells along with initial model of exploratory wells with VSP data in South Azadegan Field.
4.2. Construction of seismic acoustic impedance (AI) inversion cube Petrel (2016) software (Fig. 12.a). In the next step, by using the data of
checkshots control points and vertical seismic profiling (VSP) in seven
As explained in the methodology, seismic data inversion is a process wells with information on velocity and depth relationships in terms of
that tries to estimate the acoustic impedance model close to the real V=V0+KZ (Eq. (8)) with a correlation coefficient of 0.95, and using the
earth model based on the primary seismic models and the estimated software of Petrel (2016), seismic data is converted from time-domain to
wavelets. depth-domain data (Fig. 12.b).
4.2.1. Converting post-stack seismic data from time domain to depth 4.2.2. Construction of depth-domain seismic data cube
domain At this stage, by determining the In-line and X-line ranges of the
In order to perform the inversion with deterministic method, first, amplitude values and using the Geometrical Modeling section and the
the post-stack seismic data as time-domain were outputted from the “Arithmetic Seismic Resampling” method of Petrel (2016) software, a
Hampson Russell 08 software (HRS-8, 2010 model) and entered into the post-stack seismic data cube has been created.
9
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 10. Correlation coefficient of average velocity data of check-shot and VSP points and depth to determine coefficients of velocity model.
4.2.5.3. The root mean square (RMS) error. Based on this, the highest
4.2.3. Entering seismic horizons into Petrel (2016) software RMS error (difference between actual values and synthetic data) of
Interpreted sections, geological data at the top and base of the for acoustic impedance is 3669.5 [(m/s)*(g/cm3)] in A-025 well, and the
mations, and the wells’ layout have been entered, and correlations have lowest is 1326.43 in the well A-018. The total RMS error is 2305.8, with
been studied in seismic sections (Fig. 13). a correlating rate of 99.61%.
Since the data of seismic horizons are in depth-domain data, in order Table 4 shows the correlating matching, relative, and the RMS error
to inversion all input data to HRS-8 software, they must be in the format values of the inverted acoustic impedance (AI) for synthetic inversion
of depth-domain data. The required data include depth seismic data data and original data at the location of exploratory wells.
cube, depth seismic horizons, completed velocity and density logs from
the ground’s surface to the end of the well, the updated wavelet, and the 4.2.6. Acoustic impedance inversion analysis
post-stack data analysis are based on the model base method. After At this stage, to analyze the inversion of acoustic impedance (AI), the
converting to point data, deep seismic horizons have been exported from seismogram of the degree of correlation and error between the artificial
Petrel software and entered into HRS-8 software. seismic log and the original seismic data, along with the modified time-
dependent wavelet in the studied wells, is drawn separately. Based on all
4.2.4. Entering converted time-domain to depth-domain seismic data and the relative error values, the compatibility and RMS error of the inver
performing inversion sion with the Post-stack data in the studied wells are within the
After determining the coordinate range in UTM-WSG84 and zone acceptable range. Therefore the accuracy of the inversion has been
39S, the converted depth-domain seismic data cube has been entered confirmed. For example, the seismogram of the error values of Wells A-
into the HRS-8 software as separate depth-domain seismic horizons. 010 and A-025 are presented in Fig. 16.a and 16b.
In the next step, after entering the corrected logs into the HRS-8 Thus, the seismic inverted AI cube with the highest degree of cor
software, the time and frequency graph of the time-dependent wavelet relation and the lowest relative error between the Vp and density with
has been updated (Fig. 14). synthetic seismic mapping is constructed. Then the above data is used to
An example of the acoustic impedance (AI) cross-section obtained model the interval velocity cube by the Sequential Gaussian simulation
from synthetic seismic mapping from the surface of the Aghajari to the (SGS) method and co-kriging with the cube of the seismic acoustic
Gotnia Formations is presented in Fig. 15. impedance.
10
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 11. Average velocity map of the lower Fahliyan Formation based on seismic horizons and VSP data.
Fig. 12. a) Primary seismic data transferred from HRS-8 to Petrel (2016) as time-domain, b) Seismic data conversion from time-domain to depth-domain.
11
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 13. An example of a depth-domain section of seismic data with entering formations’ top, depth-domain seismic sections, and the location of exploratory wells in
the studied field.
Fig. 14. Wavelet time response, amplitude and phase response and history of a) Initial wavelet (Wave 0 on the left sides), and b) Constructed wavelet (Wave 1 on the
right sides).
4.2.7. Entering AI inversion as depth-domain data into Petrel (2016) the final values of inverted acoustic impedance at low depths are mostly
software in the range of 8000-15000 [(m/s)*(g/cm3)], which can be in the range
After entering the acoustic impedance data obtained from the syn of calcareous formations. The Aghajari surface Formation, with a value
thetic seismogram inversion with the highest correlation and the lowest of less than 8000 [(m/s)*(g/cm3)], is located in the Marley and shale
acceptable error, using the Geometrical Modeling command and the formations, the results of which are highly consistent with the geological
Seismic Resampling method of Petrel (2016) software, the initial cube of samples during drilling. The highest acoustic impedance values are in
the acoustic impedance is made according to the depth data. In the next the field’s lower part in the range of the lower Fahliyan Formations to
step, the out-of-range model data, which were values less than 4000 Gotnia.
[(m/s)*(g/cm3)], have been removed. The initial acoustic impedance
model resulting from a seismic inversion with depth-domain data and
the modified model with removing out-of-range values is presented in 4.3. Seismic migration interval velocity model
Fig. 17.
In order to complete the acoustic impedance model resulting from After entering the post-stack seismic data and constructing the
seismic inversion, The empty parts of the inverted acoustic impedance relevant petro-physical model, data points of interval seismic migration
(AI) cube have been completed using the acoustic impedance cube ob velocity were calculated with Petrel (2016) software. Finally, a seismic
tained from the compressional velocity and density logs with a corre migration velocity cube was constructed. Most variations of migration
lation coefficient of 0.7 (Figs. 18 and 19). Based on the results obtained, velocity data are in the 5500-600m/s (Fig. 20.a and 20b). The AI and
interval migration cubes were used to select the final velocity model.
12
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 15. Examples of seismic acoustic impedance inversion (AI) section using HRS-8 software in a) Aghajari to Pabdeh Formations, b) Kazhdumi to Got
nia Formations.
Table 4
The correlating machine, relative, and the RMS errors values of the inverted acoustic impedance (AI) for synthetic inversion data and original data at the location of
exploratory wells.
Type of Error Wells Minimum Maximum Average
4.4. Estimating compressional velocity (Vp) and density with seismic (Fig. 21 and Table 5).
attributes The correlation coefficient of the cross-sectional diagram of the
actual and predicted P-wave (target diagram) using an AI seismic
Emerge module of HRS-8 software was used to estimate the well attribute has been calculated as 0.74 and all seven attributes as 0.79.
parameters, including compressional velocity (Vp) and density, and Hence, the increase of the correlation coefficient of +0.05 shows the
finally, their multiple products as acoustic impedance (AI). After inte improvement of the relationship. It means that the estimated values of P-
grating well logs and seismic data, the Emerge module, using seismic wave velocity are in good agreement with the original Well log values.
attributes, determines a suitable set of attributes to determine the
seismic velocity in the well area. 4.4.2. Estimating density from seismic data
What is analyzed here is not the seismic data but a set of multiple Suitable attributes for estimating density (gr/cm3) from seismic data
seismic attributes extracted from these data. Using a set of attributes is are extracted by the Emerge module of HRS-8 software. Like what was
more beneficial than using raw seismic data. After that, these data can mentioned about P-wave estimation from seismic data, density is also
train the system to estimate the P-wave velocity property. estimated from seismic data and based on the analysis of suitable at
The present studies have been carried out using post-stack seismic tributes. According to Table 6 and Fig. 22, by evaluating the training and
data, sonic, and density log data of four important wells, A-001, A-002, validation curve, it was found that the first three seismic attributes of the
A-004, and A-006, in the studied fields. selected attributes are suitable for density estimation. Therefore, the
density volume was estimated using these three seismic attributes and
4.4.1. Estimating compressional velocity (Vp) density on the seismic data cube.
The Emerge module of HRS-8 software extracts suitable attributes for Analyzing and checking the validation curve in density estimation
estimating compressional velocity from seismic data. Then, it was shows that the three attributes in Fig. 23 are among the selected attributes
necessary to select the optimal number of seismic attributes. For this suitable for density estimation. The estimation was done based on single
purpose, the validation diagram was used, and some suitable attributes attribute analysis and then multi-attribute analysis. Fig. 23.a shows the
were selected. Seven attributes were suitable due to the reduction of the cross plot of density against the attribute of inverse acoustic impedance
estimation error by increasing each attribute to the previous attributes (1/AI); As seen in the figure, the correlation coefficient is 0.57.
13
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 16. Analysis the seismogram of relative error, correlation, and RMS error values of the inversion with post-stack data in a) Well A-010, b) Well A-025.
By adding the number of attributes (using the multi-attribute tech impedance inversion cubes as seismic migration velocity cubes sepa
nique), the correlation coefficient between density and attributes (three rately. Both have been re-modeled after calculating the correlation co
attributes of "1/AI", "Quadrature Trace", and "Integrated Absolute efficients of their cubes with the initial Vp cube modeled with the
Amplitude") reaches 0.60 (Fig. 23.b). The density volume is estimated by inverse distance weighted (IDW) method (Appendix Fig. B.3 and
applying this relationship between seismic data (seismic attributes) and Fig. 24).
density on the seismic data cube. As a result, the correlation coefficient of compressional velocity cube
(Vp) resulting from SGS (combined with the co-kriging method with the AI
5. Discussion inverse seismic cube) and the initial velocity cube using the inverse dis
tance weighted (IDW) method is 0.54 (Appendix Fig. B.4.a), as well as the
5.1. Determining the final velocity model by combining SGS and co- correlation coefficient of the Vp cube resulting from SGS (combined with
kriging methods co-kriging with the interval seismic velocity cube) and the initial velocity
cube using the IDW method, is 0.51 (Appendix Fig. B.4.b). Therefore, to
For determining the final velocity model, the completed data of model the effective pressure using the Bowers method, it is recommended
compressional velocity (Vp) logs are re-modeled using a sequential to use the compressional velocity cube obtained from SGS combined with
Gaussian simulation (SGS) combined co-kriging with acoustic co-kriging with the seismic acoustic impedance (AI) cube method.
14
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 17. Removal of out-of-range values of the primary Geometrical Modeling of Seismic AI (values less than 4000 [(m/s)*(g/cm3)].
Fig. 18. Determining acoustic impedance (AI) correlation coefficient from seismic inversion and log data.
Fig. 19. Final AI cube resulting from seismic inversion and log data integration with r of 0.7.
15
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 20. a) Calculating point’s data of migration velocity, b) Interval migration velocity cube based on seismic post-stack data.
16
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 21. Training and validation errors of P-wave velocity data of studied field wells.
Table 5
A number of suitable attributes for compressional velocity (Vp) estimation.
Row Target Final Attribute Training Error Validation Error
Table 6
A number of suitable attributes for density (gr/cm3) estimation.
Row Target Final Attribute Training Error Validation Error
6. Conclusions
17
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. 23. Cross plot of actual and predicted density using a) a single seismic attribute of reverse acoustic impedance (1/AI) with a correlation coefficient of 0.57, b)
multiple attributes (1/AI, Quadrature Trace, and Integrated Absolute Amplitude) with a correlation coefficient of 0.60.
Fig. 24. Secondary compressional velocity cube by SGS method and co-kriged with a) acoustic impedance (AI) cube resulting from seismic inversion, b) interval
seismic migration velocity cube.
Table 7
Experimental Variogram computation for final velocity cubes.
Direction Azimuth Dip Number of lags Lag distance Search radius Band width Tolerance angle Lag tolerance Thickness
Table 8
Results of Gaussian Variogram of compressive velocity cube obtained by combining SGS and Co-kriging methods with acoustic impedance (AI) in studied field.
Direction Nugget Sill Range Number of Pairs Anisotropy range (m)
stack data, the uncertainty described as the average relative error was impedance in deeper formations of the field (Gotnia Formation) are
8.76%. often in the range of 8000-15000 [(m/s)*(g/cm3)], which could be
The correlation coefficient of actual and predicted P-wave (target referred as calcareous lithology. The computed acoustic impedance
diagram) using only AI as attribute has been calculated as 0.74 while obtained from predicted Vp and density cubes shows a good correlation
considering seven seismic attributes the correlation increased to 0.79. with the inverted acoustic impedance of the well-data as r=0.71 and the
The same approach applied to estimate density logs out of well locations, relation used for infilling well properties which well logs do not exist. As
presents correlation improvement from 0.57 to 0.60. a result, the correlation coefficient of the Vp cube resulting from SGS
The highest average velocity in the range of 2760-2900 m/s in the (combined with a co-kriged with the AI inverse cube) and the initial
northeast of the study area is related to the Gotnia Formation, and the velocity cube using the IDW method is more than the same method with
lowest is related to the Gachsaran Formation with 2150-2280 m/s in the seismic interval migration velocity. It is suggested to use SGS (combined
southwest direction of the study area. Cube values of inverted acoustic with a co-kriged with the AI inverse cube) in similar oilfields to validate
18
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Declaration of Competing Interest The present work is from a Ph.D. dissertation on Mining Engineering-
Mineral Exploration from Islamic Azad University, South Tehran
The authors declare that they have no known competing financial Branch, hosted by the Research Institute of Petroleum Industry (RIPI).
interests or personal relationships that could have appeared to influence The authors consider it necessary to express their sincere gratitude to the
the work reported in this paper. esteemed experts of the RIPI and Exploration Directorates of the Na
tional Iranian Oil Company (NIOC).
Data availability
Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.ringps.2023.100051.
Appendices
Appendix A: Conditional programming for complete Compressional and Shear velocity logs data
Conditional programming with petrel (2016) software to calculate the empty part values (U)1 of compressional velocity for surface Aghajari to
Gotnia Formations in all the 23 exploratory wells studied in the South Azadegan Field.
Vp temp = 29.83 ∗ GR + 2911 (A.1)
Vp Full Final = If (Vp Full = U, Vp Full2, If (V2 = U, Vp temp, if (Vp = U, Interval velocity VSP, Vp))) (A.5)
Conditional programming by Petrel (2016) software to calculate the shear velocity by combining Litho_base (measured in Japan’s TRC laboratory)
and DSI logs.
Dominant Limestone lithology : Vs = − 0.01068Vp2 + 1.5106Vp − 2.2008 (A.6)
Dominant Limestone
(A.8)
Marl and Shale lithology : Vs = 0.5243Vp + 0.0451
Vs Litho base = If(DEPT < 1290, (0.738 ∗ Vp Full Final/1000 − 0.5653) ∗ 1000, If(DEPT >= 1290 and
DEPT < 2330, (0.5243 ∗ Vp Full Final/1000 + 0.0451) ∗ 1000, If(DEPT >= 2330And DEPT < 3300, ( − 0.1068∗
(A.9)
Pow(Vp Full Final/1000, 2) + 1.5106 ∗ Vp Full Final/1000 − 2.2008) ∗ 1000, If(DEPT >= 3300 and
DEPT < 3890, (0.5243 ∗ Vp Full Final/1000 + 0.0451) ∗ 1000, If(DEPT >= 3890andDEPT < 4640, ( − 0.1068∗
Pow(Vp Full Final/1000, 2) + 1.5106 ∗ Vp Full Final/1000 − 2.2008) ∗ 1000, U)))))
Vs Full Final = If (Vs Litho base < > Vs DSI, Vs DSI, Vs Litho base) (A.10)
Appendix B
1
U: Undefined
19
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. B.2. Model-based inversion starts with an initial model that is updated iteratively until the residual trace is minimized to a threshold value (Veeken & Da
Silva, 2004).
Fig. B.3. Method of constructing secondary compressional velocity cube (Vp) using SGS method combined with co-kriging with a) acoustic impedance cube (AI)
resulting from seismic inversion and b) seismic interval migration velocity cube.
20
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
Fig. B.4. a) Correlation coefficient of Vp1 cube (resulting from combined SGS and co-kriged with AI cube) and initial velocity cube with IDW method, b) Correlation
coefficient of Vp2 cube (resulting from SGS and co-kriged with migration velocity cube method) and initial velocity cube with IDW method in the south Azade
gan Field.
21
P. Kianoush et al. Results in Geophysical Sciences 13 (2023) 100051
for Natural Gas Subsurface Systems 51–92. https://doi.org/10.1016/B978-0-323- Rointan, A., Soleimani Monfared, M., Aghajani, H., 2021. Improvement of seismic
85465-8.00007-8. velocity model by selective removal of irrelevant velocity variations. Acta Geod.
Li, C., Liu, X., 2022. Three-term AVO inversion using group total variation regularization. Geophys. 56 (1), 145–176. https://doi.org/10.1007/s40328-020-00329-x.
J. Appl. Geophys. 207, 104854 https://doi.org/10.1016/j.jappgeo.2022.104854. Ronquillo Jarillo, G., Markova, I., Markov, M, 2018. Acoustic reflection log in
Lindseth, R.O., 1979. Synthetic sonic logs – A process for stratigraphic interpretation. transversely isotropic formations. J. Appl. Geophys. 148, 1–7. https://doi.org/
GEOPHYSICS 44, 3–26. https://library.seg.org/doi/abs/10.1190/1.1440922?journa 10.1016/j.jappgeo.2017.11.005.
lCode=gpysa7. Russell, B.H., 2017. Introduction to Seismic Inversion Methods: Society of Exploration
Liu, G., Zhang, L., Wang, Q., Xu, J., 2022. Data-driven seismic prestack velocity inversion Geophysicists [Research]. SEG Library. https://doi.org/10.1190/1.9781560802303.
via combining residual network with convolutional autoencoder. J. Appl. Geophys. Shahbazi, A., Ghosh, D., Soleimani, M., Gerami, A., 2016. Seismic imaging of complex
207, 104846 https://doi.org/10.1016/j.jappgeo.2022.104846. structures with the CO-CDS stack method. Stud. Geophys. Geod. 60 (4), 662–678.
Liu, Y., Dong, L., 2012. Influence of wave front healing on seismic tomography. Sci. https://doi.org/10.1007/s11200-015-0452-6.
China Earth Sci. 55 (11), 1891–1900. https://doi.org/10.1007/s11430-012-4441-0. Shahbazi, A., Soleimani Monfared, M., Thiruchelvam, V., Ka Fei, T., Babasafari, A.A.,
Mallick, S., 1995. Model-based inversion of amplitude-variations-with-offset data using a 2020. Integration of knowledge-based seismic inversion and sedimentological
genetic algorithm. GEOPHYSICS 60, 939–954. investigations for heterogeneous reservoir. J. Asian Earth Sci. 202, 104541 https://
Maulana, Z.L., Saputro, M.D., Latief, F., 2016. In: New Modified Band Limited Impedance doi.org/10.1016/j.jseaes.2020.104541.
(BLIMP) Inversion Method Using Envelope Attribute IOP Conf. Ser.: Earth Environ. Shakiba, S., Asghari, O., Khah, N.K.F., 2018. A combined approach based on MAF
Maurya, S.P., Singh, N.P., Singh, K.H., 2020. Post-stack Seismic Inversion. In: Maurya, S. analysis and AHP method to fault detection mapping: A case study from a gas field,
P., Singh, N.P., Singh, K.H. (Eds.), Seismic Inversion Methods: A Practical Approach. southwest of Iran. J. Appl. Geophys. 148, 8–15. https://doi.org/10.1016/j.
Springer International Publishing, pp. 39–80. https://doi.org/10.1007/978-3-030- jappgeo.2017.11.003.
45662-7_3. Soleimani, M., 2016. Seismic imaging by 3D partial CDS method in complex media.
Morgan, P., 1999. Azadegan Field Geophysical Interpretation [Technical]. ConocoPhillips Journal of Petroleum Science and Engineering 143. https://doi.org/10.1016/j.
UK LTD. jseaes.2020.104541.
Paul, S., Ali, M., Chatterjee, R., 2021. Prediction of velocity, gas content from neural Sompotan, A.F., Pasasa, L.A., Sule, R., 2011. Comparing models GRM, refraction
network modeling and estimation of coal bed permeability from image log in coal tomography and neural network to analyze shallow landslide. J. Eng. Technol. Sci.
bed methane reservoirs: Case study of South Karanpura Coalfield, India. Results 43, 161–172.
Geophys. Sci. 7, 100021 https://doi.org/10.1016/j.ringps.2021.100021. Swarbrick, R.E., Seldon, B., Mallon, A.J., 2005. Modelling the Central North Sea pressure
Pennebaker, J.E.S., 1968. Seismic data indicate depth magnitude of abnormal pressures. history. Petroleum Geology Conferences Ltd.
Spring meeting of southern district 184–191. Tape, C., Liu, Q., Maggi, A., Tromp, J., 2010. Seismic tomography of the southern
Qadri, S.M.T., Islam, M.A., Shalaby, M.R., El-Aal, A.K.A., 2021. Reservoir quality California crust based on spectral-element and adjoint methods. Geophys. J. Int. 180
evaluation of the Farewell sandstone by integrating sedimentological and well log (1), 433–462. https://doi.org/10.1111/j.1365-246X.2009.04429.x.
analysis in the Kupe South Field, Taranaki Basin-New Zealand. J. Pet. Explor. Prod. Terzaghi, K., Peck, R.B., Mesri, G., 1996. Soil mechanics in engineering practice, 3rd Ed
11 (1), 11–31. https://doi.org/10.1007/s13202-020-01035-8. ed. John Wiley & Sons.
Radwan, A.A., Nabawy, B.S., Abdelmaksoud, A., Lashin, A., 2021. Integrated Tong, P., Zhao, D., Yang, D., Yang, X., Chen, J., Liu, Q., 2014. Wave-equation-based
sedimentological and petrophysical characterization for clastic reservoirs: A case travel-time seismic tomography - Part 1: Method. Solid Earth 5 (2), 1151–1168.
study from New Zealand. Gas Sci. Eng. 88, 103797 https://doi.org/10.1016/j. https://doi.org/10.5194/se-5-1151-2014.
jngse.2021.103797. Veeken, P., Kashubin, A., Curia, D., Davydenko, Y., Priezzhev, I., 2020. From data
Radwan, A.A., Nabawy, B.S., Shihata, M., Leila, M., 2022. Seismic interpretation, conditioning, depth imaging and reservoir characterization to machine learning.
reservoir characterization, gas origin and entrapment of the Miocene-Pliocene First Break 38 (6), 71–77.
Mangaa C sandstone, Karewa Gas Field, North Taranaki Basin, New Zealand. Mar. Veeken, P.C.H., Da Silva, M., 2004. Seismic inversion methods and some of their
Pet. Geol. 135, 105420 https://doi.org/10.1016/j.marpetgeo.2021.105420. constraints. First Break 22, 47–70.
Radwan, A.E., 2021. Modeling pore pressure and fracture pressure using integrated well Vicêncio, H., Teves-Costa, P., Caetano, P.S., 2022. Seismic site effects in Setúbal county
logging, drilling based interpretations and reservoir data in the giant El Morgan oil (Portugal) using remi technique. Results Geophys. Sci. 9, 100037 https://doi.org/
field, Gulf of Suez, Egypt. J. Afr. Earth. Sci. 178, 104165 https://doi.org/10.1016/j. 10.1016/j.ringps.2021.100037.
jafrearsci.2021.104165. Virieux, J., 1986. P-SV wave propagation in heterogeneous media: Velocity-Stress finite-
Rahimi, M., Riahi, M.A., 2020. Static reservoir modeling using geostatistics method: a difference method. GEOPHYSICS 51 (4), 889–901.
case study of the Sarvak Formation in an offshore oilfield. Carbonates Evaporites 35 Yadav, A., Mondal, S., Chatterjee, R., 2022. Geophysical analysis to delineate a Class-I
(2), 62. https://doi.org/10.1007/s13146-020-00598-1. AVO prospect in the offshore east coast of India: A case study. J. Appl. Geophys. 206,
Rai, N., Singha, D.K., Shukla, P.K., Sain, K., 2020. Delineation of discontinuity using 104794 https://doi.org/10.1016/j.jappgeo.2022.104794.
multi-channel seismic attributes: An implication for identifying fractures in gas Yu, G. (2010). Method improve pore pressure prediction. The American oil and gas
hydrate sediments in offshore Mahanadi basin. Results Geophys. Sci. 1-4, 100007 Reporter. www.aogr.com.
https://doi.org/10.1016/j.ringps.2020.100007. Yu, Z., Liu, Y., 2022. A robust migration velocity analysis method based on adaptive
Riahi, M.A., Fakhari, M.G., 2022. Pore pressure prediction using seismic acoustic differential semblance optimization. J. Appl. Geophys. 207, 104851 https://doi.org/
impedance in an overpressure carbonate reservoir. J. Pet. Explor. Prod. Technol. 10.1016/j.jappgeo.2022.104851.
https://doi.org/10.1007/s13202-022-01524-y.
22