SPE

International

Society of Petroleum Engineers

.

SPE 98827

Modelling Channelized Systems: A Robust Approach

Ikechukwu Chike and 2Mike Onyekonwu
1

1. Laser Engineering and Resources Consultants Ltd.

2. PTDF Gas Research Group, University of Port-Harcourt.

Copyright 2005, Society of Petroleum Engineers Inc.

The result is a model that incorporates morphologic
This paper was prepared for presentation at the 29th Annual SPE International information and petrophysical trends from analogue
Technical Conference and Exhibition in Abuja, Nigeria, August 1-3, 2005.
while accurately describing the rock and fluid property
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 using well data.
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 Introduction
Engineers, its officers, or members. Papers presented at SPE meetings are subject
to publication review by Editorial Committees of the Society of Petroleum Engineers.
The architecture of the Niger Delta channel sands
Electronic reproduction, distribution, or storage of any part of this paper for results from an interplay of one or more of the various
commercial purposes without the written consent of the Society of Petroleum
Engineers is prohibited. Permission to reproduce in print is restricted to an abstract depositional energies (fluvial, wave and tides). These
of not more than 300 words; illustrations may not be copied. The abstract must
contain conspicuous acknowledgement of where and by whom the paper was
interactions and the abundant supply of fluvial
presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836, sediments make the geometry of the channels
U.S.A., fax 01-972-952-9435.
complex1. To reproduce the architecture and
subsequently model the petrophysical setting is thus
Abstract very challenging.
The importance of capturing the architecture of Over the past few years, a lot of work has been done in
depositional facies within a system cannot be the modelling of channel systems, Schatzinger et al.,
overemphasized. For a channel system as in most cases 1999 emphasized the link between lateral cyclicity of
in the Niger Delta, the problem of accurately petrophysical properties and the scale of primary
describing the geometry of these channels exists. bedding features, which can be transferred from
Ideally, object based models are best suited to outcrops directly into the subsurface because scaling
reproduce facies architecture, but this is generally problem are avoided. Hauge et al., 2003, showed how
hindered by the rigidity of this algorithm. Moreover, sandbody geometry can be modelled using object
the application of outcrop analogue in the object model models. The studies by Knox et al., 1999 suggested that
to reproduce the subsurface could be stalled by the fluvial reservoirs deposited early in an intermediate
differences in diagenesis between the outcrop and the frequency cycle contains several narrow laterally
reservoir. Thus, the need to modify the algorithm in isolated channel belts, whereas adjacent reservoir
modelling the system arises. deposited late in the cycle(as in our case) contains a
single broad channel belt that is internally
We adopted a robust object and pixel based model to heterogeneous. The dimension of complex channel of
capture the reservoir heterogeneity while honouring deep water deposit (Beaubouef, 2004) and several
hard data. The approach includes integrating several analogue geometry documented by Robert, 2004 are
available options for a particular channel setting in the used as guide in choosing the geometry. These
software used (for the object based model) to come up distributions provide constraints for conditioning the
with a flexible descriptor. Depositional trend were also area, shape, placement, and preferred orientations of
built to capture the upward fining sequence of tidal sedimentary units in reservoir models2.
channel sands. This was accomplished by building a
complex transformation sequence to ensure accurate
petrophysical description within the facies using a pixel The major and most prevalent method is the application
model. of outcrop analogue in an object based multipoint
geostatistics technique. This has proved quite helpful in
2 I. Chike and M. Onyekonwu SPE 98827

the capturing of the reservoir heterogeneity. However,

it is very difficult, to find a true analogue for any
reservoir (unless the very formation crops out in the
“neighbourhood”). More so, the outcrop records are
enigmatic3, reducing our confidence in their geometry
as analogue for an object model, since object models
are often applied in modelling facies architecture where
the geometry are well known. This fact has led to the
use of pixel based approach by some authors in
modelling the architecture of channel reservoirs3. This
does not however reproduce the geometries as the
object base models. Their argument is in the flexibility
of the pixel based algorithm. Hence increasing the
flexibility of the object based algorithm is important, Figure 1: Conceptual 3D view of fluvial-wave-tide
for it to be used. Thus, a robust algorithm of an object influence of the Niger delta (after Oomkens,1974)
based modelling technique is paramount in reproducing
the facies architecture and subsequently the rock Facies Modelling
properties, using a Sequential Gaussian simulation with The lithofacies mentioned above form the main basis
trends4 (pixel based). A detailed transformation for defining the architecture of the facies. The facies
sequence is used to ensure that both the facies and well log was constructed using the calculated volume of
data are honoured. For instance we used a depositional shale (Vsh) by applying cut-offs. The Vsh log was
trend to ensure the upward fining of some reservoir chosen in place of the Gamma ray log used by some
parameter in tidal channels. authors because of the presence of hot sands. This
resulted to the following facies: tidal channel sands
The objective of this paper is to present a robust (Green), shaly sand barriers (Red) and shales (Purple).
algorithm (“flexible object” – pixel based) that Figure 2 shows the facies.
incorporates morphologic information and
petrophysical trends from analogue (documented in
literatures) and well data analysis, bearing in mind the
major constraints.

Reservoir Geology and Lithofacies

The reservoir is bounded to the north and south by an
East-West trending boundary faults. These primary
faults constitute the main trapping mechanism of the
reservoir. The minor faults strike in an East-West
direction and assume an en-echelon arrangement with
the major boundary faults. All the faults are
downthrown southwards as with most reservoirs in the
area. The occurrence of few minor faults suggests
minor tectonics and recent sedimentation. The sand
marks the end of a second depositional cycle of the Y Figure 2: Showing various lithofacies
sand, rapidly deposited in a progradational deltaic The tidal channel sands are modelled as objects that are
sequence. The multiple, fining-upwards sequences placed in a background of the shale5. This is the general
occur in an overall coarsening-upward sequence mark point process. However, the internal algorithm is
observed from well logs, typical of tidal sands of the modified to suit our purpose. Instead of modelling
Niger Delta, suggesting a truncation of the coastal single cut and fill channels we used multi-channel
barrier sands initially deposited by the fluvial wave model(channels internally composed of genetically
interactions. The offshore mud is therefore overlain by related channels1), to represent the complex geometry1,6
the tidal channel sand. The sand is however intercalated appropriately. In this model the initial focus was on the
by minor shales. A conceptual model of the reservoir is belts properties before attention is shifted to the
shown in Figure 1. individual channels inside the channel belts. Each
channel belt is basically described by its reference point
in the reservoir, azimuth, dimension, and its amplitude
and sinuosity. The reference point and azimuth
3 Modelling Channelized Systems: A Robust Approach SPE
98827

orientation defines the channel belt axis. Based on the

drawn channel belt axis, a channel belt line is drawn
with a specified amplitude and sinuosity to give the
channel belt’s horizontal fluctuations around the
channel belt axis. Based on the drawn channel belt line,
the axes of each of the channels inside the channel belt
are drawn as displacements of the channel belt line. The
stacking of the channels in the belt was achieved by
simulating the channel axes from two normal
distributions (Figure 3), the standard deviations of the
two distributions were specified as belt dimensions,
while the “expectation values” were such that the
average axis is identical to the channel belt line. Figure
4 shows the hierarchical modelling of multichannel
belts. This algorithm gives an anastomosed channel
pattern. This solves the problem of using single channel
dimension.

Figure 4: Illustration for modelling multiple channel

belts

Trend that are normal to the average strike direction

were defined using the center of the “simulation box”.

Two methods were used to control the total number of

objects; constant value and 1D trend function, Z
(simbox). They were used within a tolerance of 0.005.
This defines how close the volume fraction used must
be at the end of the simulation. The 1D trend function
Figure 3: Normal distribution of channel axes were calculated using VPC(which represent the
fraction of facies
computed in the wells at different levels parallel to a
reference). Lateral variations in the proportions of each
facies were also taken into account since the wells are
at different locations. Constant values (global volume
fraction) were later modified using results of the VPC,
this gave a more realistic facies distribution. The
fraction of Shaly sands lithofacies were specified as a
fraction of the volume of channels (and not as a
fraction of the channel system volume).

The channel dimension: Vertical thickness and channel

axis width were specified using two normal
distributions. A correlation of factor of 0.6 was used to
relate the relationship between the vertical thickness
and channel axis. The orientation of tidal channels
supports low ratio of channel length to valley length1,
we used 1.2 to represent this ratio.

The channel form were modelled to allow enough

flexibility in the model for well conditioning and to
generate realistic channel appearance. Their attributes:
4 I. Chike and M. Onyekonwu SPE 98827

Width/thickness rugosity, vertical amplitude and

sinuosity were modified to meet these criteria.
Width/thickness rugosity attribute describes small scale
width-thickness variation along an object. Two 1D
Gaussian fields were used, one for width variations and
the other for thickness variations. Each field has the
following attributes: relative amplitude and absolute
range. The relative amplitude multiplies the channel
width, to ensure that the narrow channels have smaller
width variations than wider channels. This is an
improvement on past algorithms. The absolute range
gives a correlation range for the width/thickness
variation.

The channel variations were controlled by using

vertical amplitude and sinuosity to restrict the vertical
fluctuations around the channel axis.

Clustering of channels and channel belts were much

because of the high Sand/gross4 ratios observed from Figure 5: Facies model
logs (shown in Figure 2.0)
Petrophysical Modelling
Within the channel sands we observed the shaly sand The algorithm(model) for petrophysical modelling was
lithofacies3. This we modelled as small facies object, based on the application of statistical and geological
ellipsoidal in form and oriented relative to the channels’ transformations and spatial correlation. The statistical
principal axis. They constitute the barrier to flow within and geological transformations approximates the
the channels(K less than 10mD) as the sand is generally petrophysical parameters(φ, NTG, Sw) to a Gaussian
clean(K greater than 300mD). We simulated the width, distribution, after known geological trends and
thickness and length of the barrier independently from variability have been taken into account. While the
three Gaussian distributions with specified mean and spatial correlation captures the similarity between two
variability values. The barriers were distributed until observations, which provides information on the
the desired barrier fraction of channel volume, confidence with which the value of a cell not penetrated
specified is reached5. by the well can be estimated5. The essence of this
complex transformation sequence and spatial continuity
The entire channel system modelling is terminated after
estimation is to ensure that both facies and
reaching the set number of iterations. To set the number
petrophysical well data are correctly honoured, and the
of iteration required, we considered the complexity of
resultant characterization is valid. The absolute
the model, number of object to be generated and
permeability in the 3-D model was estimated using the
number of wells. Ideally the simulation should
Wyllie-Rose model by assuming an irreducible water
terminate successfully after having reached the
saturation of 12%. The irreducible water saturation
minimum number of iterations. However, to give room
assumption is required to obtain reasonable results in
for tolerance, a maximum number of iteration (slightly
the transition zone and the aquifer8. Figure 6
higher than the minimum) was set to enable
summarizes the process. The petrophysical models
convergence of simulation. Where this is not met, the
(Figures 7-9) are consistent with known trend in the
end temperature of an annealing function was reduced
Niger delta. Within the facies, the K values decreased
to prioritise the convergence criterion. This is a very
significantly10-13.
important process.

Having defined the entire channel setting a single

realization was generated without conditioning to the
wells. This is useful to evaluate the attribute settings.
Ten Equiprobable images of the facies model were then
generated5,7-9. This is an iterative algorithm, which
makes a proposed new state in each iteration, and then
accepts or rejects it based on the likelihood under the
model.
5 Modelling Channelized Systems: A Robust Approach SPE
98827

Figure 8: Permeability model condition to facies

Figure 6: Summary of Petrophysical modelling

process

Figure 9: Water saturation model condition to facies

Multiple equiprobable images of the petrophysical

Figure 7: Porosity model condition to facies. models were generated to assess the range of
uncertainty associated with reservoir model5-7. This is
important, because the volumetric predictions of
reservoir performance are normally based on the best
technical model. The best technical model was
determined by ranking the various realizations. A
tedious but very accurate method used was a pre
simulation run of all the most likely realizations.
Streamline simulation concept was used to minimize
the number of models used for the pre simulation run.
6 I. Chike and M. Onyekonwu SPE 98827

Reservoir volume (STOIIP) was generated for each of Characterization-Recent Advances, AAPG
the realizations simulated. However, the volumes from Memoir 71, p. 321-332.
the “most likely” model (which compared favourably
with an independent material balance calculation) was 4. Hauge, R., Syversveen, A. R. and MacDonald,
used for the initialization. This is another technique of A. C.: “Modelling Facies Bodies and
ranking the realizations. Petrophysical Trends in Turbidite Reservoirs,”
paper SPE 84053, presented at the SPE
Annual Technical Conference and Exhibition
held in Denver, Colorado, 5-8 September
Nomenclature 2003.

φ = Porosity 5. Mohan, K. and Perez, G. 2002, Applied

geostatistics for reservoir characterization, p. 2
NTG = Net to gross – 15, 140 – 146
K = Permeability 6. Beaubouef, T. T.:”Deep-water Leveed-
Channel Complex of the Cerro Toro
mD = Millidarcy Formation, Upper Cretaceous, Southern Chile”
paper AAPG Bulletin, v. 88, NO. 11
SW = Water saturation
(November 2004), pp. 1471-1500
STOIIP = Stock tank oil initially in place
7. Jerry, L. J., Lake, L. W., Corbett, P.W. M.,
VPC = Vertical proportion curve and Goggin, D. J. 1997, Statistics for
Petroleum Engineers and Geoscientists, p 143
-165, 115 – 118.

Conclusions 8. Chike, I., Robert, B. and Onyekonwu, M.

A robust algorithm that models the tidal channel sands O.:”Preservation of Reservoir Heterogeneity
of the Niger delta has been introduced. The model after Upscaling: a Niger delta Case Study”
showed a realistic level of reservoir heterogeneity that paper AAPG Bulletin, in press
greatly improves reservoir simulation exercise.
9. Hamman, J.G., Buettner, R.E., Caldwell,
The characterization of the channels show that though D.H..: “A Case Study of a Fine Scale
seismic attribute data are important in conditioning Integrated Geological, Geophysical,
facies modelling, in their absence, the robust algorithm Petrophysical, and Reservoir Characterization
introduced works well too. and Simulation with Uncertainty Estimation”.
paper SPE 84274, presented at the SPE
Uncertainties can not be ruled out in any reservoir Annual Technical Conference and Exhibition
model, due to lack of sufficient data. However, a held in Denver, Colorado, 5-8 October 2003.
geological software that allows reliable ranking of the
multiple realizations and presimulation run reduces the 10. Schatzinger, R.A., Tomutsa, L.:”Multiscale
uncertainties in the model. Heterogeneity Characterization of Tidal
Channel, Tidal delta, and Foreshore Facies,
References Almond Formation Outcrops, Rock Springs
Uplift Wyoming, 1999, in Schatzinger, R. and
1. Reading, H.G.: “Sedimentary Environment Jordan, J., eds., Reservoir Characterization-
and Facies”. Oxford, 1991, p. 190-209 Recent Advances, AAPG Memoir 71, p. 45-
56.
2. Robert, S. t.:”Geomorphology: An Approach
to Determining Subsurface Reservoir 11. Castle, J. W. and Byrnes, A. P.: “Petrophysics
Dimensions”. AAPG Bulletin, V 88, NO. 8 of Lower Silurian sandstones and integration
(AUGUST 2004), p. 1123-1147. with the tectonic-stratigraphic framework,
Appalachian basin, United States” paper
3. Seifert, D.:”Evaluation of Field Development AAPG Bulletin, v.89, NO. 1 (January 2005),
Plans Using 3D Reservoir Modelling,1999, in pp. 41-60.
Schatzinger, R. and Jordan, J., eds., Reservoir
7 Modelling Channelized Systems: A Robust Approach SPE
98827

12. Knox, P. R., Barton, M. D.:” Predicting 13. Chike, I., Egbele, E., Ukaonu, C. and
Interwell Heterogeneity in Fluvial-Deltaic Onyekonwu, M. O..: “Congenital reservoir
Reservoirs: Effects of Progressive rock characterization of weak hydraulic
Architecture Variation Through a Depositional properties” paper SPE 88968, presented at the
Cycle from Outcrop and Subsurface Annual Technical Conference and Exhibition
Observations, 1999, in Schatzinger, R. and held in Abuja, Nigeria 2-4 August 2004.
Jordan, J., eds., Reservoir Characterization-
Recent Advances, AAPG Memoir 71, p. 57-72