SPE 101996
Integrating Seismic-Derived Porosity Into a Geocellular Model—A Case Study
Gary Mercado, Hidemasa Ikawa, Mohamed Elhami, and Adel Belgaid, ADMA-OPCO
Copyright 2006, Society of Petroleum Engineers
This paper was prepared for presentation at the 2006 Abu Dhabi International Petroleum
Exhibition and Conference held in Abu Dhabi, U.A.E., 5–8 November 2006.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than
300 words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
Abstract
The goal of this study was to create a 3D seismic based
porosity volume for specific carbonate reservoirs zones and
integrate this derived porosity into a geocellular model
containing 257 points of well control porosity. The 3D marine
OBC seismic data covered a 1500 km2 area of a super-giant
Middle East oil field. The Late Cretaceous reservoirs in the
field consists of limestone and dolomitic limestone and target
zones are subdivided into 3 alternating porous and dense
zones. The porous zones range from 30-40 feet thick.
The data was inverted to pseudo acoustic impedance using a
well-driven model based technique. Sixty two wells were used
to create the initial impedance model for the seismic inversion
mainly to provide low frequency trend which tied the log data
to the seismic data. A neural network process was applied that
utilized the seismic inversion data with several other seismic-
derived attributes, in order to maximize the correlation values
in reservoir zones and minimized PHIE prediction error in the
created porosity volume in time domain.. The seismic based
porosity information was depth converted and read into a
geocellular model and used as trend data.
Geostatistical methods are commonly used to integrate seismic
derived porosity, since it honors well data as well as the
degree of correlation between seismic pseudo-porosities and
log porosities. This method can create a 3D model that takes
advantage of the well log data's fine vertical resolution and the
seismic attribute's lateral resolution. The geostatistical method
used to achieve this is collocated cokriging. Collocated
cokriging can be utilized with two methodologies: kriging
(interpolation) and conditional simulation. The first approach
is a deterministic method, whereas the latter approach attempts
to honor the spatial variability of the conditioning data and
also provides multiple realizations.
The lateral distribution of the porosity, controlled by the
seismic data in between the wells enable an accurate estimate
the inter-well petrophysical properties which had a great
impact on the uncertainty in the volumetric calculations of
these reservoirs. Validation of this approach was blindly
tested by the new drilling locations in the field.
Introduction
In order to better understand how reservoir porosity affect
fluid and pressure performance in a giant Middle East
carbonate field, a model based inversion of a 2002 vintage 3D
seismic data set was undertaken. The relationship between the
inversion data and a host of additional seismic attributes was
used in a neural network process to derive porosity. This
seismically derived porosity was then incorporated into a
geocellular model using two collocated methodologies.
Validation of this methodology was blind tested at well
locations.
Acoustic Relationships
A giant Middle East field contains several continuous and
aerially stratified porous zones in a carbonate Lower
Cretaceous formation. These porous zones are dramatically
expressed in sonic and density well log data and are reflected
in the 2002 vintage OBC 3D seismic data coverage
(1500Km2) as low velocity intervals encased in a high
velocity matrix. Extensive regional well log control (over 300
wells) verifies this stratagraphic relationship and continuity
field wide (Figure-1). Simple Colored Inversion of the 3D
seismic data supports the log data relationships but is unable to
directly quantify porosity values.
Seismic Porosity Derivation
Seeking to quantify the porosity for inclusion in a field wide
geocellular model, a more advanced methodology involving a
well driven model based inversion was needed. A contractor
inversion and neural network modeling software package was
used for this. Basically the two-part procedure involved
deriving a model-based inversion data set, followed by a
multi-attribute neural-network derivation to produce a porosity
volume to enhance the resolution.
2 SPE 101996
The 3D acoustic seismic time data was first tied to well
control at 62 well locations scattered throughout the field
where petrophysically corrected sonic and density logs
existed. A model was built tying the 3D seismic and Acoustic
Impedance from well together at each well location. The
seismic data was then inverted using the model as regional low
frequency guide, producing what is termed as the model-based
inversion dataset.
The neural network procedure takes the model based
inversion, plus 43 other derived seismic attributes and cross
correlates them against the PHIE (porosity) curves from the 62
points of well control. The model based inversion, plus the 5
attributes that maximized the correlation and minimized the
error were chosen as input to the non-linear neural network
processing. Rigorous attribute and processing parameter
determination was performed in order to minimize run time
and volume size yet retain data quality and minimize potential
error. Even with this care, the processing to create the full
field neural network porosity volume took 39 days of
continuous computer runtime.
The neural network porosity results display a very good
correlation to well log PHIE. All well correlation values
exceed 85% and blind test values were at 75% (Figure 2).
Error ranged from 3-5%. Each porosity zone was then
interpreted and mapped (Figure-3). Geophysicist and geologist
then worked together to incorporate the porosity into the
geocellular model.
Geostatistical Porosity Distribution
One of the great strengths of geostatistics is that it can
quantitatively combine diverse types of data. Geostatistical
methods are commonly used to integrate seismic derived
porosity, since it honors well data as well as the degree of
correlation between seismic pseudo-porosities and log
porosities. This application was used to create a 3D fine-scale
model that takes advantage of the well log data's fine vertical
resolution and the seismic attribute's lateral resolution. The
most common geostatistical method used to achieve this is
collocated Co-kriging. Collocated Co-kriging was utilized
with two approaches: kriging (deterministic) and stochastic
(conditional simulation). The first approach is a deterministic
method offering one unique solution, while the latter approach
attempts to honor the spatial variability of the conditioning
data and also provides multiple solutions or realizations.
Model Methodology
To build the petrophysical portion of the geocellular model,
log porosity was distributed using two methods,
deterministically (using krigging) and conditionally (using the
stochastic method). Each method was controlled and guided
by well data only. The result of these distributions was then
individually re-gridded using the seismically derived porosity
through the collocated co-krigging methodology. The Seismic
porosity volume of 8-10 ft. vertical resolution was transferred
to the 3D geocellular model and was re-sampled to the level of
the model-layering scheme. Then the geostatistical method,
collocated co-kriging, was used to distribute the porosity in the
model taking advantage of both the well log data vertical
resolution and the seismic attribute lateral resolution for
porosity distribution (Figure-4).
The two independent geostatistical approaches (deterministic
& stochastic) were applied successfully and compared. In both
cases, the lateral distribution of the porosity controlled by the
seismic data in between the wells provided a more accurate
estimate the inter-well petrophysical properties. This has had a
great impact on reducing the uncertainty in the volumetric
calculations of these reservoirs. The ultimate validation of the
two methods was blindly tested by the drilling of a new well
location in the field. The deterministic approach was found to
give the more accurate result and superior tie to the newly
drilled well.
CONCLUSIONS
The derivation of porosity from seismic data utilizing a model
based inversion and several seismic attributes in a neural
network process allowed improved resolution, accuracy and
porosity discrimination. This method was verified by
extensive blind testing. Incorporating the seismically derived
porosity information provided a more robust porosity
distribution in between wells in the giant Middle East
carbonate oil field. Of the two geostatistical distribution
methods, deterministic and co-located co-krigging, a more
accurate solution was found using the deterministic approach
for both blind testing and new well bore penetration.
Acknowledgments
The authors would like to thank the senior management of
ADNOC and ADMA-OPCO for their continous support and
for permission to publish this paper.
References
1. Cacas, M. C., Letouzey, J. & Daniel, J. M.: “Object-
Oriented Modelling of Jointed Oil Reservoirs, Paper 4-03
presented at the 1998 60th Eage Conf, Leipzig,.
2. Garcia, M.H., Journel, A.G., and Aziz, K., “Automatic
Grid Generation for Modeling Reservoir Heterogeneities”,
SPE Reservoir Engineering (May1992), 278-284.
3. Durlofsky, L.J., Jones, R.C., and Milliken, W.J., “A New
Method for the Scale Up of Displacement Processes in
Heterogeneous Reservoirs”, 4th European Conference on
the Mathematics of Oil Recovery, R ros, Norway, 7-10
June, 1994.
4. Saad, N., Kalkomey, C.T., and Quenes, A., “Optimal
Gridding of Stochastic Models for Scale-up”, 4th
European Conference on the Mathematics of Oil
Recovery, R ros, Norway, 7-10 June, 1994.
SPE 101996 3