Journal of Physics: Conference Series

PAPER • OPEN ACCESS

Bayesian Theorem Application to Model Reservoir Facies Distribution on

Deltaic Depositional Environment, Case Study of Browse Basin
To cite this article: M M Adeyosfi and A Harris 2021 J. Phys.: Conf. Ser. 2019 012083

View the article online for updates and enhancements.

This content was downloaded from IP address 213.255.248.93 on 26/10/2021 at 16:39

The 10th National Physics Seminar (SNF 2021) IOP Publishing
Journal of Physics: Conference Series 2019 (2021) 012083 doi:10.1088/1742-6596/2019/1/012083

Bayesian Theorem Application to Model Reservoir Facies

Distribution on Deltaic Depositional Environment, Case Study
of Browse Basin

M M Adeyosfi1 and A Harris2

1
Reservoir Geophysics Graduate Program, Physics Dept, Universitas Indonesia
Jl. Salemba Raya no. 4, Jakarta 10430, Indonesia
2
Physics Department, FMIPA, Universitas Indonesia
Kampus UI Depok, Depok 16424, Indonesia
Email: [email protected]

Abstract. Successful Hydrocarbon exploration and reservoir characterization always related

with good understanding of geology and geophysics aspect. One of the fundamental issues is to
get reliable parameter distribution and quantify the confidence level of the parameter model in
3D. The case study in this paper will be applied in Browse field that and the result will be a
distribution model in 3D of facies and hydrocarbon fluid. The workflow that will be introduced
in this paper is the combination between rock physics analysis, simultaneous seismic inversion
and Bayesian estimation theorem. Rock physics analysis includes well log conditioning and
correlation analysis between reservoir parameter (porosity, saturation, Vshale, etc) with seismic
parameter (acoustic impedance, vp/vs, shear impedance, etc) to obtain facies classification in
well log scale. The simultaneous seismic inversion method is used to obtain seismic parameter
cube to be correlated with rock physics result to drive the facies distribution. Bayesian estimation
theorem assembles initial knowledge about a model before observing the inversion attributes.
The probability density function later will be used to drive the facies distribution combined with
well log data and seismic data; and also estimate the confidence level distribution in 3D. The
result can be used to identify new play in the Browse field.

1. Introduction
Browse basin is a northeast-trending depocenter, located at offshore NW of Australia. This basin has
high potential of gas hydrocarbon accumulation from Paleozoic, Mesozoic and Cenozoic sediments, but
as yet unproduced. One of the reservoir rocks that consist hydrocarbon is sandstone of Plover formation
from the early Jurassic (Figure 1). The Plover Formation was deposited during periods of active faulting
during Jurassic extensions and has been, therefore, interpreted as a syn-rift succession (Struckmeyer et
al., 1998 [1]; Blevin et al., 1998 [2]). The sandstone depositional environment of Plover Formation is
mainly fluvial to tidally influenced deltaic system with some mouth bars (Tovaglieri & George., 2012
[3]). Consequently, thickness and lateral continuity of this unit are highly variable throughout the
Browse Basin.
Due to variable lateral continuity and depositional environment behaviour, it is more difficult to
distribute the facies distribution only with statistical method. In this paper the author used a combination
of rock physics analysis, simultaneous inversion, and Bayesian theorem application to characterize the
facies distribution in the studied area. The rock physics analysis consists of parameter extraction and
sensitivity analysis to define the facies in well based. Continued with simultaneous inversion to
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
The 10th National Physics Seminar (SNF 2021) IOP Publishing
Journal of Physics: Conference Series 2019 (2021) 012083 doi:10.1088/1742-6596/2019/1/012083

characterize the physics parameter distribution in field scale of the studied area. And finally, with
Bayesian theorem, multiple physics parameter from simultaneous inversion can be used to extract the
facies distribution and probability cube in the studied area.

Seismic Cross Section

Browse Basin Map

Figure 1. General Browse Basin Stratigraphic Column, seismic cross section and browse basin map
[4]
2. Method
This study used angle stack 3D seismic and well log data as primary input. Other input such as velocity
cube and borehole image log were used as cross validation and low frequency building supplementary
data. The analysis is started with data loading and QC, continued with well tie, wavelet extraction,
petrophysical analysis, borehole image processing, seismic interpretation, sensitivity analysis,
simultaneous seismic inversion, and finished with the Bayesian theorem application. There will be facies
distribution result from this analysis is a hydrocarbon bearing distribution cube.

2.1 Rock Physics

Rock physics method explains the reservoir rock in subsurface with physics properties such as porosity,
rigidity, and compressibility. These properties are affecting on how the seismic wave travels trough
formation. The objective of rock physics analysis in this paper is to find relationship between rock
properties in subsurface with observed seismic response (Figure 2).

2
The 10th National Physics Seminar (SNF 2021) IOP Publishing
Journal of Physics: Conference Series 2019 (2021) 012083 doi:10.1088/1742-6596/2019/1/012083

Figure 2. Accoustic Impedance (AI) and VpVs cross plot showing facies classification of gas bearing
sand, water bearing sand and shale.

2.2 Simultaneous Inversion

Simultaneous seismic inversion method is an inversion technique with Aki-Richards formula to
calculate the reflectivity in different offset in pre-stack seismic data or angle stack. Simultaneous
inversion is used to characterize reservoir rock and fluid parameter to inversion result, such as, P-
impedance, S-impedance, density, and VP/Vs. Shear wave (S wave) is being used to characterize fluid
and compressional wave (P wave) that sensitive to change of fluid in the pore and S wave is related to
interaction with rock matrix.
In seismology, seismic wave reflectivity and transmission in bedding boundary is formulated by
many scientists, one of them is equation formula from Knott-Zoepprits. This equation is simplified by
Aki-Richards, known by Aki-Richards equation (1980) [5]:
2
1 ∆𝑉𝑝 ∆𝜌 𝑉 ∆𝑉𝑠 ∆𝜌 1 ∆𝑉𝑝
𝑅 (𝜃 ) ≈ ( + ) − 2 ( 𝑠 ) (2 + ) x 𝑠𝑖𝑛2 𝜃 + tan2 𝜃 (1)
2 𝑉𝑝 𝜌 𝑉𝑝 𝑉𝑠 𝜌 2 𝑉𝑝

The assumption that being used is change of property (ΔVp/Vp, ΔVs/Vs and Δρ/ρ) relatively small,
hence second order can be ignored and ϴ with a value less than 90o (Ma, 2002 [6]). Equation (1) can be
simplified with S wave impedance and P wave impedance parameter as below:
2 2
∆𝐼 𝑉 ∆𝐼𝑠 𝑉 ∆𝜌
𝑅(𝜃) ≈ (1 + tan2 𝜃 ) 2𝐼𝑝 − 8 (𝑉𝑠 ) sin2 𝜃 − [tan2 𝜃 − 4 (𝑉𝑠 ) sin2 𝜃] 2𝜌 (2)
𝑝 𝑝 2𝐼𝑠 𝑝

Fatti et al [7] in 1994 simplified the (2) equation by removing ρ parameter where the Vs/Vp ratio is
around 0.5 and small angle as below:

(3)
During the simultaneous inversion, one of the algorithms that's being used is simulated annealing. A
simulated annealing algorithm is a vector solution that optimized with an acoustic impedance value,
followed by n shear impedance value, where n is amount of layers. The algorithm will search for lowest
cost function with Monte-Carlo method (Figure 3).

3
The 10th National Physics Seminar (SNF 2021) IOP Publishing
Journal of Physics: Conference Series 2019 (2021) 012083 doi:10.1088/1742-6596/2019/1/012083

Figure 3. Global search covering possible solutions with Monte Carlo in simulated annealing
algorithm [8]
The given initial model is reflection coefficient of zero-offset P and S impedance at interface I can be
calculated by the equation below:

(4)
Hence the ratio of average Is/Ip between layer i-1 and layer I can be calculated by the equation below:

(5)
By replacing the equation (5) and (4) into the equation (3), reflection coefficient R (ϴ) can be calculated
in every offset and every layer boundary.

2.3 Bayesian Theorem

A Bayesian theorem in short understanding can be viewed as conditional probabilistic. The theorem
itself stated if we have two events x and y, we can design the probability of occurrence of each event by
P(x) and P(y). If we consider of the probability of these two events, we can design this as joint probability
of (P(x,y)) that is given by:

𝑃(𝑥, 𝑦) = 𝑃(𝑥 |𝑦)𝑃(𝑦) = 𝑃(𝑦|𝑥)𝑃(𝑥) (6)

In This case, P(x/y) is the probability of x to occur given that y already occurred, the same situation
applied to P(y/x) and P(y/x) can be defined as conditional probability. From equation (6), if we want to
know how the probability of y is, given x already happened, we can re-arrange the equation into:

(7)
To complete equation (7), according to the total law of probability:
𝑃(𝑥) = 𝑃(𝑥 |𝑦)𝑃(𝑦) + 𝑃(𝑥 |𝑦 𝑐 )𝑃(𝑦 𝑐 ) (8)
In equation (8) above, the P (yc) is defined as the probability of not having y to be happening.
Therefore the complete Bayesian equation to define the conditional probability of y to be happened
given x that already happened is:

(9)
The product of Bayesian calculation is a posterior PDF from each alternative. Below is a probability
density function (PDF) for hydrocarbon bearing facies. Hydrocarbon bearing facies has low AI and low
Vp/Vs ratio (Figure 4).

4
The 10th National Physics Seminar (SNF 2021) IOP Publishing
Journal of Physics: Conference Series 2019 (2021) 012083 doi:10.1088/1742-6596/2019/1/012083

Figure 4. Probability Density Function (PDF) for HC bearing, water bearing and shale facies
3. Result and Discussion
3.1 Seismic Inversion Result
The simultaneous seismic inversion was done in Plover formation only as target formation. The results
are acoustic impedance cube (AI), Vp/Vs cube and density cube. The relative misfit of the inversion
result is relatively low which is 27% and the residual cube has low amplitude means the simultaneous
seismic inversion result is valid to be used for further analysis.
The acoustic impedance in the Plover formation is relatively low compared to Triassic formation and
there are some low acoustic impedance reflecting high permeability – gas bearing facies or shale facies.
The Montara formation above Plover formation has a mostly lowest acoustic impedance related to shale
facies developed in this formation. The cross section with well data in figure 5 below shows good
coherency between acoustic impedance from simultaneous seismic inversion with acoustic impedance
from well data.

Figure 5. SW-NE cross section of the Acoustic Impedance Cube with Kronos-01, Poseidon-02 and
Boreas-01, shows good coherency between inversion result and well data

Another simultaneous inversion result is Vp/Vs cube. NE-SW cross section of Vp/Vs cube with some
wells at figure 6 below shows good coherency between the inversion result with Vp/Vs log data from
well. The Vp/Vs in Triassic and Montara formation mostly have high value. The Plover formation Vp/Vs
is relatively low compared to Triassic formation and Montara formation, with some very low Vp/Vs
value at lower Plover formation related to gas bearing sand facies.

5
The 10th National Physics Seminar (SNF 2021) IOP Publishing
Journal of Physics: Conference Series 2019 (2021) 012083 doi:10.1088/1742-6596/2019/1/012083

Figure 6. SW-NE cross section of a VP / Vs Cube with Kronos-01, Poseidon-02 and Boreas-01,
shows good coherency between inversion result and well data
3.2 Hydrocarbon Bearing Facies Distribution
The probability density function of the gas bearing facies, water bearing facies and shale from acoustic
impedance and Vp/Vs cross plot is being used to predict the hydrocarbon bearing facies distribution.
Similar to sand facies distribution that act as a reservoir, the hydrocarbon bearing facies are mostly
distributed at the lower part of the Plover formation. Based on the NE-SW cross section at figure 7
below, most of the highest hydrocarbon bearing facies probability is at the lower part of the formation.
Cross validation with well log data such as water saturation shows good coherency, hence the facies
prediction with Bayesian theorem is valid.

Figure 7. SW-NE cross section of hydrocarbon bearing facies probability Cube with Kronos-01,
Poseidon-02 and Boreas-01, shows good coherency between Bayesian theorem and well data (water
saturation)
The probability cube can be extracted into horizon probe (geobody) where the facies distribution in field
scale can be seen clearly. The hydrocarbon bearing facies distribution shows gas accumulation mostly
on the NE part of the field, similar to mud log data and petrophysical analysis result. The geometry of
the hydrocarbon bearing facies distribution shows that the sand with gas bearing only at NE part, the
sand facies at SW part is mostly filled with water (Figure 8).

6
The 10th National Physics Seminar (SNF 2021) IOP Publishing
Journal of Physics: Conference Series 2019 (2021) 012083 doi:10.1088/1742-6596/2019/1/012083

Probability Map of Gas Bearing Facies at Plover Fm Gas Bearing Facies Distribution with accumulation
sketch

Figure 8. Probability Map of Hydrocarbon Bearing Facies Distribution shows gas mostly accumulated
at NE part of the field
4. Conclusions
The application of combination between rock physics, simultaneous seismic inversion and Bayesian
theorem application is proven to be a powerful tool to predict facies distribution in tectonically active
and poor facies lateral continuity condition compared to geostatistical method. The studied area shows
the sand distribution is NE-SW trend similar with browse basin axis with geometry as a channel. The
gas accumulation is mainly on the NE part of the field.
Acknowledgments
Authors wishing to acknowledge my colleagues, Abdul Kholiq, Dyna Mariana, Ian Kainama & Levina
Aurelia Wirjowedojo, whose given author guidance and input during the analysis process. The author
also would like to express gratitude to Australia government (National Offshore Petroleum Information
Management System) who provide author this wonderful published data.

References
[1] Struckmeyer 1998 The Sedimentary Basins of Western Australia 2: Proceedings of the Petroleum
Exploration Society of Australia Symposium, pp. 345-367
[2] Blevin 1998 The Sedimentary Basins of Western Australia 2: Proceedings of the Petroleum
Exploration Society of Australia Symposium, pp. 369-395
[3] F. Tovaglieri and A. George 2012 AAPG Annual Convention and Exhibition.
[4] ConocoPhillips (Browse Basin) Pty Ltd 2009.
[5] K. Aki and P. G. Richards 1980 Quantitative Seismology: Theory and Methods, (San Fransisco:
W.H. Freeman Company), p. 932pp.
[6] X. Q. Ma 2002 Geophysics, 67(6) pp. 1692-2041
[7] Fatti 1994 Geophysics, 59 pp. 1362-1376
[8] Schlumberger 2019 Petrel Quantitative Analysis Training Book, (Schlumberger)