Static Reservoir Modelling by Comparing PDF

UNIVERSITY OF MANCHESTER
Static Reservoir Modelling Comparing Inverse

Distance Weighting to Kriging Interpolation
Algorithm in Volumetric Estimation. Case
Study: Gullfaks Field
A dissertation submitted to The University of Manchester for the degree of Master of Science
in the Faculty of Engineering and Physical Sciences
2016
Student ID: 9679414
School of Earth and Environmental Sciences

Static Reservoir Modelling Comparing Inverse Distance Weighting to Kriging Interpolation Algorithm in
Volumetric Estimation. Case Study: Gullfaks Field
List of Contents
Table of Contents
List of Contents ....................................................................................................................................... 1
List of Figures ......................................................................................................................................... 3
List of Tables .......................................................................................................................................... 4
List of Graphs ......................................................................................................................................... 4
List of Equations ..................................................................................................................................... 4
Abstract ................................................................................................................................................... 5
Declaration .............................................................................................................................................. 6
Copyright ................................................................................................................................................ 6
Dedication ............................................................................................................................................... 7
Acknowledgement .................................................................................................................................. 7
1 INTRODUCTION .......................................................................................................................... 8
1.1 Problem Definition.................................................................................................................. 8
1.1.1 Objectives ....................................................................................................................... 9
2 LITERATURE REVIEW ............................................................................................................. 10
2.1 Inverse Distance Weighting .................................................................................................. 10
2.2 Kriging .................................................................................................................................. 11
3 GEOLOGICAL SETTING ........................................................................................................... 12
3.1 Location ................................................................................................................................ 12
3.2 Background ........................................................................................................................... 12
3.3 Regional Geology ................................................................................................................. 13
3.4 Local Geology ....................................................................................................................... 13
3.5 Petroleum System ................................................................................................................. 14
3.5.1 Source ........................................................................................................................... 14
3.5.2 Reservoir ....................................................................................................................... 15
3.5.3 Seal and Trap ................................................................................................................ 15
4 MATERIALS AND METHODS .................................................................................................. 16
4.1 Structural Modelling ............................................................................................................. 16
4.1.1 Fault modelling ............................................................................................................. 16
4.1.2 Pillar gridding and boundary definition ........................................................................ 16
4.1.3 Horizon modelling and depth conversion ..................................................................... 16
1|Page
4.1.4 Zonation and layering ................................................................................................... 16

4.2 Property Modelling ............................................................................................................... 17
4.2.1 Geometrical modelling.................................................................................................. 17
4.2.2 Facies modelling ........................................................................................................... 17
4.2.3 Statistical analysis of data ............................................................................................. 17
4.3 Petrophysical Modelling ....................................................................................................... 18
4.3.1 Variogram modelling .................................................................................................... 18
4.3.2 Modelling ...................................................................................................................... 19
4.4 Volumetric Analysis ............................................................................................................. 20
4.5 Upscaling .............................................................................................................................. 20
5 RESULTS AND INTERPRETATIONS ...................................................................................... 22
5.1 Petrophysical Analysis .......................................................................................................... 22
5.2 Facie Modelling .................................................................................................................... 22
5.3 Variogram Modelling............................................................................................................ 23
5.3.1 Anisotropy..................................................................................................................... 27
5.4 Petrophysical Modelling of Interpolating Methods .............................................................. 27
5.5 Assessment of Interpolation Methods ................................................................................... 29
5.5.1 Porosity and Permeability models ................................................................................. 29
5.6 Statistical Evaluation ............................................................................................................ 31
5.7 Volumetric Results................................................................................................................ 35
5.7.1 Cross-validation of recoverable hydrocarbons per reservoir segments ........................ 40
5.8 Upscaling .............................................................................................................................. 41
6 Discussion ..................................................................................................................................... 43
6.1 Variogram ............................................................................................................................. 43
6.2 Statistics and Cross-validation .............................................................................................. 44
6.2.1 Accuracy and Precision................................................................................................. 45
6.2.2 Consistency and Errors ................................................................................................. 45
6.3 Volumetrics ........................................................................................................................... 46
6.4 Limitations ............................................................................................................................ 47
6.5 Uncertainty Analysis ............................................................................................................. 47
7 CONCLUSION ............................................................................................................................. 49
8 RECOMMENDATIONS .............................................................................................................. 51
8.1 Further Studies ...................................................................................................................... 51
Bibliography ......................................................................................................................................... 52
2|Page
APPENDIX ........................................................................................................................................... 55
A. Materials and Methods .............................................................................................................. 55
B. Results and Interpretations ........................................................................................................ 60
C. Workflow .................................................................................................................................. 71
List of Figures
Figure 1 Geographical block location and structural framework of the study area (Gullfaks
field) modified after (NPD, 2016). .......................................................................................... 12
Figure 2 Stratigraphic log of the Gullfaks fields indicating the lithology and quality of the
two main reservoir zones of the Brent Group study area modified after (Fossen &
Hesthammer, 2000). ................................................................................................................. 14
Figure 3 Fluvial facie model displaying (A) Plan view of Top Tarbert zone multi-dimensional
flow pattern and (B) Side view showing the facie variations within the Tarbert and Ness
zones. ....................................................................................................................................... 23
Figure 4 Horizontal Variogram maps of upscaled porosity and permeability data ................. 24
Figure 5 Vertical semivariance and sample variogram of upscaled porosity data .................. 25
Figure 6 Horizontal semivariance and sample variogram of upscaled porosity data .............. 26
Figure 7 Cross-plot relationships between porosity and permeability for (A) Upscaled
properties, (B) Kriging Realisation 1 properties, (C) Kriging Realisation 2 properties and (D)
Kriging Realisation 3 properties. ............................................................................................. 28
Figure 8 Porosity distributions through the model indicating a tri-modal histogram
distribution relating to the facies.............................................................................................. 29
Figure 9 Permeability distributions through the model indicating a tri-modal histogram
distribution relating to the facies.............................................................................................. 30
Figure 10 Cross-sections of (A) Realisation 2 IDW porosity model, (B) Realisation 2 Kriging
porosity model, (C) Realisation 3 IDW porosity model, (D) Realisation 3 Kriging porosity,
(E) IDW permeability model, (F) Kriging permeability model and (G) Facie model ............. 31
Figure 11 Spider charts showing porosity comparison of normalised summary statistics of (A)
Realisation 2 and (B) Realisation 3 ......................................................................................... 33
Figure 12 Spider charts showing permeability comparison of normalised summary statistics
of (A) Realisation 2 and (B) Realisation 3............................................................................... 34
Figure 13 Cross-validation analyses between IDW and Kriging petrophysical properties in
the Gullfaks fields. (A) Realisation 2 porosity analysis, (B) Realisation 3 porosity analysis
and (C) Permeability analysis .................................................................................................. 35
Figure 14 Volumetric comparison per reservoir zones for realisation 3 showing (A) Kriging
recoverable oil, (B) IDW recoverable oil, (C) Kriging recoverable gas and (D) IDW
recoverable gas......................................................................................................................... 36
Figure 15 Cross-validation analysis between IDW and Kriging Volumetric results in the
Gullfaks fields for Realisation 2. (A) Segment by segment analysis for recoverable oil and
(B) Segment by segment analysis for recoverable gas ............................................................ 40
3|Page
Figure 16 Cross-validation analysis between IDW and Kriging Volumetric results in the
Gullfaks fields for Realisation 3. (A) Segment by segment analysis for recoverable oil and
(B) Segment by segment analysis for recoverable gas ............................................................ 41
Figure 17 Comparison of reservoir zone in (A) Kriging fine grid, (B) IDW fine grid, (C)
Kriging coarse grid, (D) IDW coarse grid for recoverable oil and (E) Kriging fine grid, (F)
IDW fine grid, (G) Kriging coarse grid, (H) IDW coarse grid for recoverable gas. ............... 42
List of Tables
Table 1 Variogram parameters used in porosity and permeability modelling ......................... 19
Table 2 Volumetric Analysis of the Two (2) Realisations based on Kriging and Inverse
Distance Weighting algorithms................................................................................................ 21
List of Graphs
Graph 1 Histogram graph of recoverable oil for Realisation 2 and 3 indication percentage
error bars .................................................................................................................................. 37
Graph 2 Histogram graph of recoverable gas for Realisation 2 and 3 indication percentage
error bars .................................................................................................................................. 37
Graph 3 Segment by segment comparison of IDW, ID2W, ID4W and Kriging of recoverable
STOIIP for Realisation 2 ......................................................................................................... 38
GIIP for Realisation 2 .............................................................................................................. 38
STOIIP for Realisation 3 ......................................................................................................... 39
GIIP for Realisation 3 .............................................................................................................. 39
List of Equations
Equation 1 Inverse Distance Weighting method ..................................................................... 10
Equation 2 Collocated cokriging method ................................................................................ 11
Equation 3 Volumetric Calculation ......................................................................................... 20
The final word count: 9,848
4|Page
Abstract
Being able to understand and characterise the behaviour of the subsurface and combining it
with a suitable statistical method gives a higher level of confidence in the model produced.
Interpolation of porosity and permeability data with minimum error and high accuracy is,
therefore, essential in reservoir modelling. The most widely used algorithm for interpolation
is kriging which with enough well data is the best linear unbiased estimator. This research
sought to compare the applicability and competitiveness of inverse distance weighting (IDW)
method using power index of 1, 2 and 4 to kriging when there is sparse data (14 wells) due to
time and budget constraints to calculate hydrocarbon volumes in a fluvial-deltaic reservoir.
After defining the reservoir extent and fluvial facies, different variograms were computed to
adequately quantify the variations in the distribution of porosity and permeability due to
reservoir heterogeneity. Interpolation results, estimated from descriptive statistics, showed
that IDW with a power index of 1 outperformed kriging. The differences in the interpolated
values were, however, insignificant but IDW with power index of 1 indicated the least error
estimation. Volume calculations also showed a marginal difference of 0.08 between IDW
power index of 1 and kriging in the reservoir zones. Cross-validation of hydrocarbon volumes
based on dividing the reservoir into fault segments resulted in an RMSE of ±2.228 MMbbls
and 0.273 BCF for recoverable oil and gas respectively. With inverse distance weight
exhibiting least errors and higher accuracy, the volumetric and statistical results confirms that
when there is less well data in a fluvial-deltaic reservoir, the best porosity and permeability
interpolation choice should be inverse distance weighting method with power index of 1.
5|Page
Declaration
This dissertation is the student’s original work unless referenced clearly to the contrary, and
that no portion of the work referred to in the dissertation has been submitted in support of an
application for another degree or qualification of this or any other university or other institute
of learning;
Copyright
The author of this dissertation (including any appendices and/or schedules to this dissertation)
owns certain copyright or related rights in it (the “Copyright”) and he has given The
University of Manchester certain rights to use such Copyright, including for administrative
purposes.
ii. Copies of this dissertation, either in full or in extracts and whether in hard or electronic
copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as
amended) and regulations issued under it or, where appropriate, in accordance with licensing
agreements which the University has entered into. This page must form part of any such
copies made.
iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual
property (the “Intellectual Property”) and any reproductions of copyright works in the
dissertation, for example graphs and tables (“Reproductions”), which may be described in
this dissertation, may not be owned by the author and may be owned by third parties. Such
Intellectual Property and Reproductions cannot and must not be made available for use
without the prior written permission of the owner(s) of the relevant Intellectual Property
and/or Reproductions.
iv. Further information on the conditions under which disclosure, publication and
commercialisation of this dissertation, the Copyright and any Intellectual Property and/or
Reproductions described in it may take place is available in the University IP Policy, in any
relevant Dissertation restriction declarations deposited in the University Library, and The
University Library’s regulations.
6|Page
Dedication
I dedicate this project work to my parents for having faith enough to invest in me. I pray and
hope that their labour will not be in vain and that they will live long to enjoy the fruits of
their labour.
Acknowledgement
All praise and thanks to the Almighty God for His blessings and mercies. I would like to
show my gratitude to the Otchere and Hammond families for their continued love and
support.
I wish to express my gratitude to Ghana National Petroleum Corporation-Oil and Gas

Learning Foundation (GNPC-OGLF) for financing my Master’s Degree which culminated in
this project and to Schlumberger for the use of their Petrel® and Techlog® software to
undertake this project.
I am especially grateful to my supervisor Dr David Hodgetts for his constructive discussions

and useful software tips and solutions. Last but not the least, I would like to give my sincere
gratitude to my colleagues and friends, who have generously provided insightful comments
on the project work.
7|Page
1 INTRODUCTION
1.1 Problem Definition

Geostatistical analysis, involving the manipulation of data of localised samples for the
prediction of hydrocarbon volumes and its flow simulation, is an essential tool for stochastic
reservoir modelling. It has been used since the eighties to describe regional variable
behaviour by using mathematical theories (Malvic, 2006). Presently, it is used in petroleum
geoscience for the spatial distribution of regional variables (e.g. porosity and permeability)
using variogram analysis and multiple interpolation techniques and simulations. Geostatistics
can be applied in all phases of geological modelling so it is very critical to the future of a
petroleum field and its stakeholders. Incorrect estimations have the following effect in a
petroleum field;
1. Miscalculation of Hydrocarbon Initially In Place (HIIP).
2. Wrong capital investment decisions.
3. Incorrect facility planning and well placement.
4. Incorrect forecast of recovery factor.
Currently, the application of Kriging and Sequential Gaussian Simulations gives improved
reservoir models in terms of porosity distribution that are very smooth without the ‘bull-eyes’
effect. Kriging techniques and stochastic simulations generally give the best spatial
estimation and the description of uncertainty in the reservoir respectively when there is
enough well data (Malvic, 2006). The negative impacts are the time of computational power
required for its implementation and problems in inferring a correct linear correlation model
due to inadequate well data.
The Inverse Distance Weighting method (IDW) on the other hand, is a simple and time
sufficient method of estimating which does not require as much data in comparison with
Kriging (Al-Hassan & Adjei, 2015). It has been proven to be effective and reliable in ‘some’
aspects. IDW gives you explicit control over the influence that distance has over
petrophysical properties; an advantage you don’t have with Kriging (Naoum & Tsanis, 2004).
This method also has its shortcomings such as; the assumption of interpolated result quality
can decrease, no account of clustering, variogram parameters and scale ignored and can over-
predict for positively skewed or high nugget effects. In view of its shortcomings, Inverse
Distance Weighting method is not reliable in all deposits hence the emphasis on ‘some’
aspects.
8|Page
Due to the time factor and data inadequacy, Inverse Distance Weighting method and Kriging
method are compared in the Gullfaks field for the Tarbert and Ness reservoir zones in
Norway to verify if Inverse Distance Weighting method can be used as a suitable alternative
algorithm when there is sparse well data and time constraint to drill new wells.
1.1.1 Objectives
The objectives of this project are to;
1. Employ the two methods of estimation on the selected area of study

and compare the results produced.
2. Predict the Hydrocarbon Initially In Place (HIIP).
3. Find an alternative time efficient method in cases where there are inadequate well
data and time constraints.
9|Page
2 LITERATURE REVIEW
There have been gluts of studies in many disciplines on Inverse Distance Weighting method
and Kriging in interpolating values (Isaaks & Srivastava, 1989; Matheron, 1989; Goovaerts,
1997; Webster & Oliver, 2001; Rocha, et al., 2012). Both methods are used in many
disciplines with some preferring one or the other. There have been claims that IDW is better
in sedimentary deposits (Al-Hassan & Adjei, 2015) whiles others also attribute Kriging to
being the Best Linear Unbiased Estimator (Skøien, et al., 2006). All interpolation methods
are based on Tobler’s first law of geography, which states “Everything is related to
everything else, but near things are more related than distant things” (Tobler, 1970) hence
choosing an interpolation method should usually be based on knowledge of the depositional
environment of the reservoir as such there should not be only 1 acceptable way of
interpolating results. As such the existence of uncertainty in a reservoir model is as a result of
our lack of subsurface understanding and the knowledge that there is that which we do not
know. This report seeks to exhausts all geological understanding and effects before
comparing a deterministic approach to a statistical method to establish the best interpolator.
2.1 Inverse Distance Weighting

The inverse distance weighting (IDW) is a deterministic method that estimates the values of
an attribute at unsampled points by applying a weighting factor that is based on a distance
reverse function to each sample with a measured parameter. The method is based on the
assumption that measured values have a greater similarity and correlation on unsampled
points closer to it and its weight decreases with distance especially when the power parameter
also increases. The accuracy of IDW is affected mainly by the choice of power parameter and
size of neighbourhood search radius which is subjective (Webster & Oliver, 2001). The
weights can be expressed as:
Equation 1 Inverse Distance Weighting method
1
⁄𝑑𝑃
𝜆𝑖 = 𝑖
∑𝑖=11⁄ 𝑝
𝑛
𝑑𝑖
Where di is the distance between the sample and estimated points, p is a power parameter,
and n represents the number of sampled points used for the estimation.
10 | P a g e
2.2 Kriging
The type of stochastic method employed in this research is the Kriging (Gslib) in
combination with collocated cokriging which uses secondary variables and explicitly
accounts for spatial cross-correlation between measured and interpolated variables
(Goovaerts, 1997). This method requires the development of a cross-covariance that is quite
difficult as the relationship between variables needs to be well understood. Co-kriging was
used in defining the cross relationship between porosity and acoustic impedance from seismic
(Doyen, 1988), hence the use of this method in the upscaling and distribution of porosity in
reservoirs. The collocated co-kriging can be mathematically expressed as (Xu, et al., 1992):
Equation 2 Collocated cokriging method
𝑍 ∗ (𝑥0 )−𝑚1 𝑛 [𝑍1 (𝑥𝑎 )−𝑚1 ] [𝑍2 (𝑥𝑎 )−𝑚2 ]

= ∑𝑎=1
1
𝜆 1𝑎 +𝜆2
𝜎1 𝜎1 𝜎𝑧
Where σ1 and σ2 are stationary standard deviations associated with means m1 and m2 for
primary and secondary variables Z1 and Z2 respectively whiles λ is the weight assigned and
Z1 (xo) is the primary variable estimate.
11 | P a g e
3 GEOLOGICAL SETTING
3.1 Location
The Gullfaks field is situated within the Norwegian borders of the North Sea. The discovery
of this field was made in 1978 in the North Eastern part of block 34/10 in the offshore licence
PL050. Its exact location from the town of Bergen is approximately 175 km north-west. The
offshore block which is mainly within a complicated 10-25 km wide NNE-SSW trending
fault blocks covers the eastern half of a total acreage of 55 km2 (Petterson, et al., 1990). See
Figure 1 below.
Gullfaks Shetland
Basin
Figure 1 Geographical block location and structural framework of the study area (Gullfaks field) modified after
(NPD, 2016).
3.2 Background
The block licence owned by Statoil (85%) as major operators, Norsk Hydro (9%) and Saga
Petroleum (6%) as minor operators had its first exploration well spudding on June 20, 1978.
The Jurassic section of the Brent sandstone of the Brent Group was intersected by the first
well and recorded an oil column height of about 162 metres. This initiated the drilling of four
12 | P a g e
new exploration wells, mainly in the western part of the field, which was hugely successful.
Three of the wells (34/10-3, -4 and -5) intersected more oil filled Brent sandstone. The other
well (34/10-6) was dry because it intersected 60 metres of thick water bearing Cook and
Statfjord regressive sandstone Formations (Petterson, et al., 1990).
After five (5) years of drilling 14 wells within the Gullfaks field, the successful exploration
was concluded. To summarise, the majority of wells (10) proved economic discoveries of
hydrocarbon with 3 wells having intersected water-bearing sandstones and 1 well abandoned
with reasons not published (Petterson, et al., 1990).
3.3 Regional Geology

The Gullfaks field, located within the East Shetland Basin which is a major North Sea Basin
that has yielded many fields, is mainly associated with west-northwest-east-northeast oriented
faults that create a domino of discrete fault blocks. The basin can be found in between the
United Kingdom and Norway surrounded by the Viking Graben, East Shetland platform and
Magnus basin as shown in Figure 1 above. The domino fault structure that characterises the
basin was created by extensional faulting system that occurred during the Eurasia and North
America rifting caused during the Mesozoic. The reactivation of the graben boundary faults
resulted in a continued extension and thermal subsidence that controlled sediments input
during the Tertiary or possibly Quaternary (Prosvirnov, 2010).
3.4 Local Geology

Based on the report by Petterson, et al., 1990 and several development wells, key information
about the complex geology of the Gullfaks field and reservoir zones is given as follows:
1. Many seismic surveys were not able to identify many of the smaller but important
faults within the basin due to its resolution.
2. The faults that were able to be interpreted on the seismic were either sealing (having
sand-to-sand contact) or not thereby impacting fluid flow across them.
3. Seismic remapping and improved resolution of the extremely faulted locations
identified the Brent sandstones to be much deeper than initially interpreted, with the
Tarbert and Ness Formations being the main reservoir zones in the western part of the
Gullfaks field.
4. New wells identified that the initial water bearing Cook Formation had oil filled sands
in two fault blocks.
13 | P a g e
5. Additional well control reports identified that the Tarbert has good lateral correlation
whiles Ness, slightly eroded to the west of the fault blocks, has moderate lateral
continuity over wide areas. A brief stratigraphy is shown in Figure 2 below.
Figure 2 Stratigraphic log of the Gullfaks fields indicating the lithology and quality of the two main reservoir zones of
the Brent Group study area modified after (Fossen & Hesthammer, 2000).
3.5 Petroleum System

3.5.1 Source
The Draupne Formation provides the primary deposition of the main source rock in the
Gullfaks field. The migration process of the source rock had additional inputs from the Drake
and Heather Formations during transportation and final deposition in the Jurassic rift phase.
The hydrocarbons in the Brent reservoir is sourced from the high generation potential of the
Kimmeridge clay Formation which is upper Jurassic in age with the main kitchens located
within the Tampen spur with sources to the North, South and East from the "MØre kitchen",
14 | P a g e
"Oseberg kitchen" and "Troll kitchen" respectively. The accumulation of the hydrocarbons
migrated from the North of the Snorre field (Petterson, et al., 1990).
3.5.2 Reservoir
The Brent sandstone of the Brent Group constitutes the main oil bearing reservoir of the
Gullfaks field which is of middle Jurassic age. Older sandstones of Lower Jurassic and Upper
Triassic age from the other reservoir Formations. The additional reservoir zones come from
the Cook, Statfjord and Lunde Formations. The two reservoir zones relevant to this research
are the Tarbert Formation, formed in a marginal marine environment by reworked delta plain
deposits and Ness Formation, floodplain deposits with isolated fluvial sandstone. The
reservoir situated at a relatively shallow depth exhibits excellent reservoir properties with
porosity being as high as 0.3 v/v and permeability values around 4000 mD. These excellent
petrophysical properties are as a result of poor cementation resulting in loose unconsolidated
sands. The bad side of this is the probable production of sand with oil (Petterson, et al.,
1990).
3.5.3 Seal and Trap

Erosional Cretaceous mud and siltstones form the vertical seal. It can be correlated laterally
along the horizontal length of the reservoir. Fault transient analysis and identification of
different oil-water-contacts suggest that majority of the faults are sealing. These faults also
form domino structural traps with tilted and rotated fault blocks bounding the east, structural
horst complex to the west and strongly faulted accommodation zones in the middle of the
reservoir. The tilted blocks also create stratigraphic traps where the reservoir terminates
against the sealing Cretaceous mud and siltstones (Ahmadov, et al., 2012).
15 | P a g e
4 MATERIALS AND METHODS

The data used for this research consists of interpreted seismic data, petrophysical analysis and
a fault model provided by the University of Manchester. The UTM zone was inputted as
ED50-UTM31.
4.1 Structural Modelling
4.1.1 Fault modelling

This forms the first stage of critical decision making where interpreted faults were quality
checked to determine if they were geologically and geometrically reasonable to impact flow
behaviour. Due to the absence of fault transmissibility data all faults were assumed to be
sealing hence dividing the reservoir into twenty (20) segments.
4.1.2 Pillar gridding and boundary definition

To define the horizontal resolution and the skeleton framework of the model a pillar grid with
I and J increments of 100 m was used as well as setting the major bounding faults as part of
the grid boundary. The J and I trends were set to the major North-South trending faults and its
perpendicular faults respectively. Figure A 1 in Appendix A shows all the fault models in the
mid skeleton pillar grid.
4.1.3 Horizon modelling and depth conversion

Based on well and seismic interpretation, four (4) horizons were created in their correct
stratigraphic sequence. Well tops were used to define horizon tops with the horizon surfaces
as inputs. Based on literature, the basal contact types were set accordingly incorporating
relevant faults pertaining to the horizons.
A Domain conversion of the 3D grid and fault model was done using a velocity model to
bridge the gap between time and depth using the seismic reference datum to create a
realisation of the subsurface independent of time. The layer cake method for depth
conversion was used for this process with the velocity model V=V0+KZ used due to the
relative relationship between depth and time, velocity values for K becomes negative. The
resulting 3D grid was used throughout the project.
4.1.4 Zonation and layering

Zonation and layering define the vertical resolution of the model, creating a division based on
sequence stratigraphy thereby making the modelling process easier. The 3D grid cells are
defined by the layers and pillars as they make the resolution grid finer. Isochore points were
16 | P a g e
created between well tops to define the zones. Layers were then created to define the
thickness, geological deposition and orientation between horizons with an appropriate
division applied to create a geologically reasonable termination of the layers.
4.2 Property Modelling

Properties between wells were distributed using the below processes to define reservoir
heterogeneity and correlation with well data.
4.2.1 Geometrical modelling

Some properties (Bulk volume, cell angle, well index and above contact) were generated
using pre-defined variables to serve as quality checks and aid in volumetric calculations and
operations in petrophysical properties. The petrophysical and facie properties from well logs
were then up-scaled to assign a value to each grid cell. The up-scaled cells, displayed in
Appendix A Figure A 2 for some selected wells, were displayed beside the original log in the
well section to verify that flow baffles have also been up-scaled.
4.2.2 Facies modelling

Discrete facies data from the up-scaled properties were distributed throughout the model grid
using the pixel-based method Sequential Indicator Simulation (SIS) due to the inadequacy of
well data and high uncertainty. This stochastic method uses a probability distribution function
to populate facies from well data. The inherent variations in the spatial distribution of the
facie were defined using variograms and histograms to distribute the facie property. The
variogram parameters were based on modern fluvial analogues such as the Ob River in
Russia and the Nile River in Egypt. A summary of facie variogram parameters is illustrated in
Table A 1 in Appendix A.
4.2.3 Statistical analysis of data

To better understand input data and determine the method to use in populating the reservoir
unit, summary statistics of the well data were carried out to determine if the properties belong
to the same statistical group so as to combine them for statistical analysis. This helps in the
truncation of erroneous data as a result of noise and to correct for skewness that will have an
impact on the final output. The analysis in Appendix A Table A 2 and Table A 3 shows the
statistics of the porosity and permeability properties respectively for some selected well and
the distributional accuracy and similarity between its properties based on the Kolmogorov-
Smirnov test. Minimal observed skewness across the wells was within an acceptable range of
-1.5 and 1.5. However, the same cannot be said of the permeability data hence a much larger
17 | P a g e
acceptable range of -5 to 5 was applied. An output truncation and cox-box process were then
applied to the wells that fell out of the hypothetically acceptable range.
4.3 Petrophysical Modelling

For the purpose of this project, the distribution of the porosity and permeability values
throughout the model grid were based on the moving average and kriging algorithms. Data
from well logs were upscaled and used as hard data in the interpolation of petrophysical
parameters. The porosity-permeability relationship used in the co-kriging correlation
coefficient can be seen in Figure A 3 in Appendix A.
4.3.1 Variogram modelling

Variography modelling was done to help determine the lateral and vertical variations in
porosity and permeability distribution hence define the spatial anisotropy of the model and
correlation between the input properties. An inaccurate variogram model will be detrimental
to the project, impacting volumetric calculations and the dynamic simulation of the reservoir.
As such, different possible variograms were generated with only those showing significant
geologically reasonable variations used in the rest of the project. Due to limited data on facie
distribution, geological intuition, analogues and well spacing were used in generating
appropriate variograms. The vertical and horizontal variograms were modelled for each facie
based on the fluvial facie model and were used in conditioning the model per zone to help
account for heterogeneity and variation in fluid flow. For both methods, three (3) possible
realisations were created from three different variogram parameters illustrated in Table 1
below.
18 | P a g e
Table 1 Variogram parameters used in porosity and permeability modelling
Major Minor
FLUVIAL Vertical
dir. dir. Azimuth
FACIES Range
Range Range
Floodplain 5000 5000 2 19
Levee 1000 500 12 90
Porosity Channel 5000 3000 2.5 90
Crevasse
850 350 15 45
Realisation Splay
1 Floodplain 1500 1500 15 19
Levee 400 250 5 45
Permeability Channel 5000 3000 2.5 45
Crevasse
1000 850 11 45
Splay
Levee 1200 500 5 90
Porosity Channel 5000 3000 2.5 90
Crevasse
850 350 10 45
Realisation Splay
2 Floodplain 1500 1500 15 19
Levee 400 250 5 45
Crevasse
1000 850 11 45
Splay
Levee 1000 500 12 90
Porosity Channel 5000 2000 3 90
Crevasse
850 350 5 40
Realisation Splay
3 Floodplain 1500 1500 15 19
Levee 400 250 5 45
Crevasse
1000 850 11 45
Splay
4.3.2 Modelling
The porosity model was used as a correlation coefficient for the input secondary variable for
the permeability model. Since primarily porosity and permeability is a factor of facie
distribution, a different porosity model was created using the variogram parameters for the
fluvial facies model to define its anisotropy.
19 | P a g e
Interpolation process was by the Kriging (Gslib) and Inverse Distance Weighting (IDW)
methods of power index 1, 2 and 4. Figure A 3 in Appendix A illustrates the histogram
comparison of the realisations. Realisation 2 and 3 were selected for further analysis due to
their distribution being more geologically reasonable based on their spatial continuity and
variations from porosity-permeability cross plots.
The property calculator was used to define the water saturation which was expressed as a
function of height above fluid contact, porosity and permeability due to the absence of core
data. The Net to Gross property was also defined using the property calculator as 0.8 based
on Gullfaks literature and limited well logs used in this model.
4.4 Volumetric Analysis

The quantification of the total potential hydrocarbon in place was estimated based on the
porosity models. In the absence of pressure data and very limited well logs, a constant fluid
contact was created at a depth of -1880 m and -2010 m for gas-oil-contact and oil-water-
contact respectively (Schlumberger, 2010). The required parameters (total volume, Net:
Gross, effective porosity and hydrocarbon saturation) needed for the calculation of the
hydrocarbon volumes was estimated from the reservoir model. The stock tank oil (gas)
initially in place was subsequently computed using Equation 3.
Equation 3 Volumetric Calculation
𝐺𝑅𝑉×𝑁/𝐺×𝑆𝑜 ×∅
𝑆𝑇𝑂𝐼𝐼𝑃 (𝐺𝐼𝐼𝑃) = 𝐵𝑜 (𝐵𝑔)
With GRV the gross volume of rock, N/G the net: gross, So the initial oil saturation, the
porosity and Bo (Bg) formation volume factor. The Bo, Bg and recovery factor was given a
constant value of 1.656, 0.006 and 62% for oil and 74% for gas respectively (Amirbayov, et
al., 2013). Table 2 below shows the total hydrocarbon volumes of the Gullfaks field for all
algorithms and realisations.
4.5 Upscaling
Using a finer and coarser grid size of 50 by 50 and 500 by 500 respectively, the structures
and properties in the original grid were upscaled. Most-of averaging volume weighted was
used to select the most common facie for the upscaled cell. Arithmetic averaging volume
weighted was used to upscale the porosity models for realisation three (3) whiles flow based
20 | P a g e
averaging method was used for upscaling the permeability models for realisation three (3) to
define permeability tensors.
Table 2 Volumetric Analysis of the Two (2) Realisations based on Kriging and Inverse Distance Weighting algorithms
REALISATION 2 REALISATION 3
VOLUMETRICS VOLUMETRICS
ALGORITHMS
STOIIP GIIP STOIIP GIIP
(MMBBLS) (BCF) (MMBBLS) (BCF)
Kriging 446.59 89.91 478.04 103.40
Inverse
Distance 496.91 103.65 496.91 102.84
Weighting
Inverse
Distance
515.78 110.18 515.78 109.55
squared
Weighting
Inverse
Distance
522.07 114.17 522.07 113.36
quadruple
Weighting
21 | P a g e
5 RESULTS AND INTERPRETATIONS
5.1 Petrophysical Analysis

From seismic interpretations, the lateral extent of the reservoir is approximated to be 7010 m.
Figure B 1 in Appendix B shows the interpreted North/East – South/West tilted reservoir and
seal horizons from the seismic inline 550. The reservoir is intersected by North/West -
South/East trending faults. From the interpretations made on the 14 wells, two (2) reservoir
zones were identified in all the wells namely Tarbert and Ness. Tarbert, having approximate
porosity and permeability ranges of 0.01 – 0.3 v/v and 20 - 1510 mD respectively, has an
average thickness of 92 m which is subdivided into three (3) due to laminated shales acting as
barriers with Tarbert 1 acting as the seal to the Ness reservoir zone. The Ness zone,
subdivided into two (2), averages a thickness of approximately 85 m with moderate sand
continuity. It exhibits porosity and permeability ranges of approximately 0.01 – 0.25 v/v and
30 – 1200 mD respectively. With these results, shown in Appendix B Figure B 2, the Tarbert
can be classified as a very good reservoir whiles the Ness is classified as a moderate
reservoir.
5.2 Facie Modelling

There is the possibility of running several realisations of facie distribution but due to
inadequate time one realisation depicting as accurate as possible the facie distribution was
used. The conceptual geological model below displays the channel sands realistically
showing the synchronological and geometrical arrangement of the facie and its cross
relationship with each other. Using the Karoo basin as an analogue to verify the reliability of
the model, the geo model clearly depicts the facie variability in the reservoir showing a multi-
dimensional flow pattern as shown in Figure 3 (A) below. The model, Figure 3 (B) below,
depicts the good lateral continuity of the Tarbert zone as well as the moderate connectivity of
the sand channels in the Ness zone. It also accurately identifies the laminated shale barriers
within the reservoir zones.
22 | P a g e
Figure 3 Fluvial facie model displaying (A) Plan view of Top Tarbert zone multi-dimensional flow pattern and (B)
Side view showing the facie variations within the Tarbert and Ness zones.
5.3 Variogram Modelling

The horizontal variogram map of the upscaled porosity property shows the lateral anisotropy
implied from geological knowledge of the Gullfaks with the major and minor directions
reading 175º and 85º respectively. The upscaled permeability property also gave major
direction as 161º and minor direction as 71º. This is shown in the horizontal variogram map
in Figure 4 below. These major and minor directions gave an unambiguous indication as to
the lag distances to use in the variogram model depicting spatial variations in the x and y
directions due to inadequate well data.
23 | P a g e
The vertical semivariance, shown in Figure 5 , indicates a likely high nugget effect at a 2.1 m
range and a fining upward trend. The two (2) horizontal semivariances for porosity in Figure
6 below indicate a nugget value of zero (0) and an increasing variability. The horizontal
semivariances for the upscaled permeability data in Appendix B Figure B 3 allude to three (3)
cycles, a zero (0) nugget value and an increasing variability. The vertical semivariance,
Appendix B Figure B 4, also gave high nugget value of 0.48 at a range of 2.1 m.
The observed cyclicity is attributed mainly to the sparse data but can also be caused by the
underlying geological periodicity that is the different depositional periods of the channel
sands, floodplain shales and non-channel levee and crevasse splay (hole effect). The
increasing variability is also due to the direction used, with the vertical range being shorter
than the horizontal range due to a much larger horizontal separation between data points.
These results, not divided into facies, provided additional knowledge to the analogues for the
variogram parameters conditioned to facies.
Horizontal Variogram Map of Upscaled Porosity Data Horizontal Variogram Map of Upscaled Permeability Data
Figure 4 Horizontal Variogram maps of upscaled porosity and permeability data
24 | P a g e
Figure 5 Vertical semivariance and sample variogram of upscaled porosity data
25 | P a g e
Figure 6 Horizontal semivariance and sample variogram of upscaled porosity data
26 | P a g e
5.3.1 Anisotropy
The results of the above variograms in both the major and minor directions exhibit different
ranges in both porosity and permeability upscaled data at approximately the same sill. This is
usually typical of a fluvial meandering deposit where properties such as primary porosity and
permeability have larger ranges along flow direction as against directions perpendicular to the
flow. For the porosity data, the vertical variogram reaches a lower sill than the horizontal
variogram due to areal trends that cause the vertical variogram not to reach full sill of the
porosity property as the horizontal variogram did. The vertical sill of the permeability data
reaches a higher sill value than the horizontal sill as a result of variance increase from
stratification or layering. There is also a decrease in the spatial correlation of the horizontal
variograms for both porosity and permeability with distance. In general, the porosity and
permeability data exhibits both zonal and geometric anisotropy behaviour.
5.4 Petrophysical Modelling of Interpolating Methods

All the methods applied in all the realisations show a positive relationship between the
interpolated and upscaled porosity and permeability data with a correlation coefficient of
approximately 1 and 0.7 respectively. The statistical values were derived using linear
functions Y = 1*X+0 and log(Y) = 0.00248548*X+1.50812 for porosity and permeability
respectively. This resulted in porosity and permeability covariance results of 0.0035 and
57.53 respectively. Porosity and permeability cross-plot comparison of inverse distance
weight with kriging also follows similar trends. In general, the trends indicate that the
channel sands are more porous and permeable as compared to the floodplain background
shales. There are some instances where the background shales exhibit higher porosity and
permeability. Primarily, petrophysical properties are as a result of facie deposition but these
can also be impacted on by diagenesis. Secondary processes that can cause sand channels to
be less porous and permeable are the degree and rate of cementation and compaction by
overlying sediments. Fractures and dissolution of minerals can also result in shales having an
increased porosity and permeability. The petrophysical cross-plot analysis thus portrays the
properties in association with facie distribution and reflects heterogeneity. The difference in
each of the kriging realisation is due to the different variogram parameters used hence
realisation 1 was adjudged not to show the full heterogeneity of the reservoir. Realisation 2
and 3 of kriging porosity and permeability cross-plots were used as a basis for comparison in
this research due to them being more geologically reasonable. Figure 7 below shows the
cross-plot relationship between porosity and permeability for all the kriging realisations with
27 | P a g e
the cross plot relationship for the three (3) IDW algorithms illustrated in APPENDIX B
Figure B 5.
Figure 7 Cross-plot relationships between porosity and permeability for (A) Upscaled properties, (B) Kriging
Realisation 1 properties, (C) Kriging Realisation 2 properties and (D) Kriging Realisation 3 properties.
28 | P a g e
5.5 Assessment of Interpolation Methods
5.5.1 Porosity and Permeability models

Figure 8 and Figure 9 below shows the 3D distribution of the porosity and permeability
model for Realisation 3 respectively. The histogram illustrates the distribution based on the
facies with high percentages of porosity and permeability indicating sand channel facies.
From the visual cross-section analysis, both algorithms show slight differences and
adequately conform to the facies in the fluvial facie model as illustrated in Figure 10 below.
The permeability distribution is as a result of using the porosity model as a secondary
variable. The Kriging algorithm made use of the collocated kriging algorithm in which
different zonal correlation coefficient was used but no coefficient was used for the Inverse
Distance Weighting methods.
Figure 8 Porosity distributions through the model indicating a tri-modal histogram distribution relating to the facies.
29 | P a g e
k
j
i
k
j
i
Figure 9 Permeability distributions through the model indicating a tri-modal histogram distribution relating to the
facies.
30 | P a g e
A B
C D
E F
LEGEND
G
Figure 10 Cross-sections of (A) Realisation 2 IDW porosity model, (B) Realisation 2 Kriging porosity model, (C)
Realisation 3 IDW porosity model, (D) Realisation 3 Kriging porosity, (E) IDW permeability model, (F) Kriging
permeability model and (G) Facie model
5.6 Statistical Evaluation

Summary of the descriptive statistics for IDW, ID2W, ID4W and Kriging is presented in
Figure 11 and Figure 12 as a normalised spider chart with the actual values in Appendix B
Table B 1 and Table B 2. There is obvious good correlation between all methods as shown by
the spider charts below but IDW exhibits strong comparable results to kriging. In all the
algorithms used, none over- or under-estimated the observed data values of the upscaled
31 | P a g e
porosity and permeability at their respective depth. In realisation 2 porosity interpolation,

IDW achieved the best comparable results in the measurement of standard error, standard
deviation, confidence level and sample variance. The comparably low standard error and
standard deviation, with a marginal difference of 0.064 and 0.097 respectively, indicate the
accuracy with which the values represent the sample mean of the porosity property with a
good distribution from the mean. The low variance also shows how closely related the values
are in the interpolated IDW and Kriging porosity. Realisation 3 showed similar results but
IDW values are closely related to Kriging with a marginal difference of 0.012 for the
standard error and 0.004 for standard deviation showing a strong relationship hence it being
the better of the two realisations. Results of Inverse Distance Weights in realisation 3 can be
seen to differ away from the kriging results as the power index increases. The permeability
values gave similar values for both realisations due to the variograms being the same. It also
proves that IDW gives comparable values to kriging than ID2W and ID4W with the power
index affecting its variation from the kriging interpolated values henceforth further analysis
were based on IDW and kriging only with the other power indices of Inverse distance
Weights used to cross check results.
A complete assessment of the model performance, as shown in Figure 13, involves

correlation coefficient, covariance, R2, average absolute errors and root-mean-squared error
as a cross-validation analysis for IDW to test its precision, accuracy and consistency. The
results indicate a strong correlation between realisation 3 IDW modelled porosity and Kriging
porosity values. Results of root-mean-squared error also affirm the similarity with ±0.0446
for realisation 2 and ±0.0216 for realisation 3 indicating that results from realisation 3 are
slightly more consistent. The average absolute errors of realisation 2 and 3 also recorded
figures of 0.003 and 0.002 respectively with realisation 3 IDW model being more precise.
Both realisations of porosity model show a strong positive linear relationship between IDW
and Kriging values based on the correlation coefficients with 0.83 and 0.90 for realisation 2
and 3 respectively. Results of the permeability cross-validation shown in the plot below
indicate 0.918 as the correlation coefficient, 0.22 as an average absolute error and ±55.63 mD
as a root-mean-square error and they are the same for both realisations as the same variogram
was used.
32 | P a g e
Figure 11 Spider charts showing porosity comparison of normalised summary statistics of (A) Realisation 2 and (B)
Realisation 3
33 | P a g e
Figure 12 Spider charts showing permeability comparison of normalised summary statistics of (A) Realisation 2 and
(B) Realisation 3
34 | P a g e
A B
Figure 13 Cross-validation analyses between IDW and Kriging petrophysical properties in the Gullfaks fields. (A)
Realisation 2 porosity analysis, (B) Realisation 3 porosity analysis and (C) Permeability analysis
5.7 Volumetric Results

Table B 3 in Appendix B shows the total recoverable hydrocarbons in place for the Gullfaks
field. IDW proves the best comparable algorithm with kriging with a marginal difference of
0.113 and 0.153 for oil and gas respectively in realisation 2. IDW also recorded marginal
difference of 0.039 for oil and 0.005 for gas in realisation 3. To further test the level of
comparison of IDW to kriging, the total recoverable hydrocarbon was calculated for each
reservoir zone. This will also form the basis as to the zones with the highest amount of
hydrocarbons to produce from. Results of this are illustrated in Graph 1 and Graph 2 below
which also confirms realisation 3 as the better of the two as a result of a detailed statistical
and geologically modelled variogram. Realisation 3 gave a percentage difference between
35 | P a g e
Kriging and IDW of approximately 8% in both recoverable oil and gas and this is illustrated
in Figure 14 below. The Tarbert zone, being classified as a very good reservoir, has the
highest amount of oil and gas. The Gullfaks, proving profitable hydrocarbons, will then be
divided into segments based on the fault model hence further analysis done to quantify the
volumes in each segment and test the comparable results of IDW, ID2W, ID4W and Kriging
are shown in the Graph 3–Graph 6 below. The results also confirm that IDW is the feasible
alternative algorithm to use with realisation 3 still proving better results.
Figure 14 Volumetric comparison per reservoir zones for realisation 3 showing (A) Kriging recoverable oil, (B) IDW
recoverable oil, (C) Kriging recoverable gas and (D) IDW recoverable gas.
36 | P a g e
HISTOGRAM OF RECOVERABLE OIL IN GULLFAKS

RESERVOIR ZONES FOR REALISATION 2 AND 3
SHOWING ERROR BAR WITH PERCENTAGE
160
140
RECOVERABLE OIL (MMBBLS)
120
100 Realisation 2 Kriging STOIIP

(MMBBLS)
80 Realisation 2 IDW STOIIP
(MMBBLS)
60
Realisation 3 Kriging STOIIP
(MMBBLS)
40
Realisation 3 IDW STOIIP
20 (MMBBLS)
0
Top Tarbert Tarbert2 - Tarbert1 - Top Ness - Ness1 - Top
- Tarbert 2 Tarbert1 Top Ness Ness1 Etive
RESERVOIR ZONES
Graph 1 Histogram graph of recoverable oil for Realisation 2 and 3 indication percentage error bars
HISTOGRAM OF RECOVERABLE GAS IN GULLFAKS

RESERVOIR ZONES FOR REALISATION 2 AND 3
SHOWING ERROR BARS WITH PERCENTAGE
35
30
RECOVERABLE GAS (BCF)
25
20 Realisation 2 Kriging GIIP (BCF)
15 Realisation 2 IDW GIIP (BCF)

Realisation 3 Kriging GIIP (BCF)
10
Realisation 3 IDW GIIP (BCF)
5
0
Top Tarbert Tarbert2 - Tarbert1 - Top Ness - Ness1 - Top
- Tarbert 2 Tarbert1 Top Ness Ness1 Etive
RESERVOIR ZONES
Graph 2 Histogram graph of recoverable gas for Realisation 2 and 3 indication percentage error bars
37 | P a g e
RECOVERABLE STOIIP COMPARISON BY

SEGMENTS FOR REALISATION 2
160
150
140
RECOVERABLE STOIIP (MMBBLS)
130
120
110
100
90 KRIGING [2]
80
70 IDW [2]
60 ID2W [2]
50
40 ID4W[2]
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
SEGMENTS
Graph 3 Segment by segment comparison of IDW, ID2W, ID4W and Kriging of recoverable STOIIP for Realisation
2
RECOVERABLE GIIP COMPARISON BY

45
40
35
RECOVERABLE GIIP (BCF)
30
25 KRIGING [2]
20 IDW [2]
15 ID2W [2]
ID4W[2]
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
SEGMENTS
Graph 4 Segment by segment comparison of IDW, ID2W, ID4W and Kriging of recoverable GIIP for Realisation 2
38 | P a g e
RECOVERABLE STOIIP COMPARISON BY

160
150
140
RECOVERABLE STOIIP (MMBBLS)
130
120
110
100
90 KRIGING [3]
80
70 IDW [3]
60 ID2W [3]
50
40 ID4W [3]
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
SEGMENTS
Graph 5 Segment by segment comparison of IDW, ID2W, ID4W and Kriging of recoverable STOIIP for Realisation
3
RECOVERABLE GIIP COMPARISON BY

45
40
35
RECOVERABLE GIIP (BCF)
30
25 KRIGING [3]
20 IDW [3]
15 ID2W [3]
ID4W [3]
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
SEGMENTS
Graph 6 Segment by segment comparison of IDW, ID2W, ID4W and Kriging of recoverable GIIP for Realisation 3
39 | P a g e
5.7.1 Cross-validation of recoverable hydrocarbons per reservoir segments

Cross-validation of the results, shown in Figure 15 below, gives the relationship between
IDW and Kriging of realisation 2 an average absolute error of 0.047 and 0.015 for
recoverable oil and recoverable gas respectively. The root-mean-square error also indicates
an estimation of ±2.4155 MMbbls of recoverable oil and ±1.169 BCF of recoverable gas.
Realisation 3, illustrated in Figure 16 below, shows a stronger relationship between IDW and
Kriging recording average absolute errors of 0.029 for recoverable oil and 0.00735 for
recoverable gas. It further indicates consistency with a root-mean-square error of ±2.228
MMbbls and ±0.273 BCF for recoverable oil and gas respectively.
Figure 15 Cross-validation analysis between IDW and Kriging Volumetric results in the Gullfaks fields for
Realisation 2. (A) Segment by segment analysis for recoverable oil and (B) Segment by segment analysis for
recoverable gas
40 | P a g e
Figure 16 Cross-validation analysis between IDW and Kriging Volumetric results in the Gullfaks fields for
Realisation 3. (A) Segment by segment analysis for recoverable oil and (B) Segment by segment analysis for
recoverable gas
5.8 Upscaling
Upscaling results in a loss of information and changes in the geological model. There were
slight variations in the statistical comparison of IDW and Kriging in both the fine and coarse
upscaled grids. The differences were negligible as shown in Appendix B Table B 4.
Appendix A Table B 5 shows the comparison of volumetrics of Realisation 3 fine and coarse
upscaled grid between IDW and kriging algorithms. The difference between the total field
recoverable hydrocarbons of the original grid was the same for fine grid upscale. The coarse-
scale grid reduced the original grid total recoverable hydrocarbon difference from 3.9% to
1.3% for oil and 0.5% to 0.4% for gas. For the reservoir zones, Figure 17, the fine-scale grid
41 | P a g e
did not impact the original grid recoverable oil for both Kriging and IDW but it rather
decreased the average difference between recoverable gas of the two algorithms from 8% to
6%. The use of a coarser grid for the reservoir zones instead reduced the average percentage
difference of the original grid recoverable oil between kriging and IDW from 8% to 7% but
did not impact significantly the volumes of recoverable gas. Upscaling results further show
the strong correlation between IDW and Kriging regardless of the grid size used hence
enhancing its competitiveness as an alternative to Kriging.
Figure 17 Comparison of reservoir zone in (A) Kriging fine grid, (B) IDW fine grid, (C) Kriging coarse grid, (D) IDW
coarse grid for recoverable oil and (E) Kriging fine grid, (F) IDW fine grid, (G) Kriging coarse grid, (H) IDW coarse
grid for recoverable gas.
42 | P a g e
6 Discussion
The discussion is mainly biased to the porosity interpolated values because it was used in the
reservoir volumetric analysis. Permeability results were however moderately discussed so as
to give an indication of what to expect during flow simulation.
6.1 Variogram
A detailed variogram capturing the underlying spatial variation of the reservoir is not only an
important prerequisite for static modelling but dynamic flow simulation. Variograms define
the tortuosity and connectivity between data points clearly impacting the flow behaviour
within the reservoir. This is illustrated in many texts (Armstrong, 1984; Journel & Huijbregts,
1991; Olea, 1995; Goovaerts, 1997) although none has been able to give a robust and
systematic workflow to produce a geologically legitimate variogram model. The different
variogram parameters used to determine the 3-D variations within the model was based on
the summary statistics of the upscale properties. In characterising the petrophysical properties
data set, geological knowledge on flow barriers and diagenetic processes was considered.
Prior to variogram modelling, the data analysis tool was used to transform the data points to a
standardised normal distribution that were slightly skewed to make the variogram
interpretation easier.
The variogram, conditioned to facies from the conceptual fluvial facie model, depicted a lot
of variogram properties that indicated geologic process at different scales. The vertical
variograms for both porosity and permeability exhibited areal trends indicating zonal
anisotropy which is attributed to the zonations within the reservoir. The major and minor
horizontal variogram parameters were inferred from the variogram used to model the fluvial
facie as primarily porosity and permeability are controlled by facie distribution. The
moderations to the parameters were to capture diagenetic processes that vary within the
reservoir. The experimental variogram identified interfacies variations hence indicating zonal
anisotropy in the reservoir due to these stratifications. The differences in the range of spatial
correlation also suggest geometric anisotropy behaviour within the model. The different
anisotropic behaviour corresponds to a specific geological interpretation which is typical of a
fluvial-deltaic reservoir. To define all these behaviours, large directional ranges were used
due to sparse data and this was combined with analogue data (Kupfersberger & Deutsch,
1999) and published variogram parameters (Gringarten & Deutsch, 2001). The parameters
used in modelling the variogram were somehow different due to the number of data points.
43 | P a g e
To compute a better variogram to get a better understanding of the spatial variations, outcrop
data or simulation results of a similar field should be used as qualitative inputs. A better
documentation of simulation and outcrop data of a similar producing field will go a long way
in accurately characterising a variogram consistently thereby reducing uncertainty. This
somehow makes an inference that to get the best kriging results more data (wells) is
paramount. The anisotropic ratio used in the Inverse Distance Weighting modelling was also
based on the ranges used in the kriging variograms so as to make a good comparison of the
interpolation methods that characterizes the same porosity and permeability variations.
This project clearly defined anisotropic trends and directions which were used as an informed
input in defining variograms for each reservoir facie with the initial semi-variance model
suggesting that the valid model type to best fit the porosity and permeability data is
exponential. The non-shale facie ranges were kept relatively smaller due to a much larger
internal variation with the opposite applying to the floodplain facie. In the variograms models
for the reservoir petrophysical properties, the majority had a nugget to sill ratio of less than
0.1 indicating a strong spatial structure with a few showing a weak spatial structure with a
nugget to sill ratio of about 0.6. The weak spatial structure is generally associated to the
vertical variograms that mean the majority of the data variations are as a result of the short
lag distance.
6.2 Statistics and Cross-validation

Descriptive statistical properties like correlation coefficient, standard deviation, regression,
range, covariance and input data values are important as these can influence the selection of a
spatial interpolation method (Collins & Bolstad, 1996). The results from this paper clearly
infers that realisation 3 being the best choice statistically is based on the measurement
accuracy, data and spatial distribution as this impacts interpolation accuracy conforming to
the study by (MacEachren & Davidson, 1987). Some papers (Al-Hassan & Adjei, 2015)
report that ID2W rather gives accurate comparisons to kriging but with porosity and
permeability modelling, smoothing results in the loss of extreme values hence an increase in
the polynomial of Inverse Distance Weighting resulted in increasing disparity from kriging.
Even though all four methods recorded similar minimum and maximum values for porosity
and permeability properties, the distribution based on the deviation, variance, low skewness
and its median proves that the lower the power index used for Inverse Distance Weighting,
the more similar it is to kriging statistically. The standard error and deviation increased as the
44 | P a g e
power index was increased as illustrated in Appendix B Table B 1 and Table B 2. IDW
recorded the relatively lesser standard error and deviation values than kriging. This does not
subsequently prove IDW to be better but rather shows its competitive nature when few input
data points (14 wells) are used. The bulls-eye and smoothing effect caused by Inverse
Distance Weighting and Kriging respectively does not represent the raw data geologically
and this is as a result of the polynomial regression of the two algorithms (Myers, 1990).
6.2.1 Accuracy and Precision

The cross-validation analysis between kriging and IDW models for realisations 2 and 3 as
shown in Figure 7 above shows a cluster of data points with realisation 2 being relatively
diffused. The cluster of the kriging and IDW interpolated porosity values gives correlation
coefficients of 0.8 and 0.9 for realisation 2 and 3 respectively. The latter indicates a strong
relationship between the two algorithms than the former although they both show a strong
positive relationship. The regression analysis also showed that realisation 3 shows a stronger
IDW to kriging relationship with values of 0.8 as compared to 0.7 for realisation 2. The
accuracy of IDW to kriging was assessed by using the average absolute error. Although there
are other methods (coefficient of residual mass and mean prediction errors) to evaluate this,
the average absolute error was enough to give an accurate measurement in combination with
standard error. The average absolute errors for realisation 2 and 3 IDW/Kriging cross plot
were 0.003 and 0.002 respectively. This criterion shows that interpolation errors of IDW to
kriging were relatively less showing a competitive performance of less than 1% for both
realisations. The relatively lower average absolute error of realisation 3 indicates its superior
accuracy over realisation 2.
6.2.2 Consistency and Errors

The interpolation results of IDW showed a high variation in the upscaled data producing
results that clearly captured the minimum and maximum porosity and permeability raw data
points. It, however, produced unrealistic values at well locations as it is affected by the
uneven distribution and spatial correlation of well data points predicting less accurate values.
This is evident by the root-mean-squared error (RMSE) porosity results obtained from
realisation 2 and 3 showing approximately ±4% and ±2% respectively providing a measure of
error size (Li & Heap, 2014). The results, indicating the precision and consistency of IDW to
kriging, were acceptable as it falls within the acceptance range of 10%. The RMSE and
absolute error of permeability results of approximately 2% for both also suggest that the
model can be accepted for the two algorithms. By quantifying the systematic and
45 | P a g e
unsystematic errors into RMSE (Willmott, 1981), the recorded small errors of the porosity
values indicate a good performance of the model. The proportion of realisation 3 RMSE was
less (2.2%) than that of realisation 2 (4.5%) suggesting the superiority of realisation 3. The
almost similar standard error values for realisation 3 porosity gave the inference that the
deviation will be similar resulting in equal precision in the interpolated values. The equal
sample variance results of 0.002 between IDW and Kriging for realisation 3 porosity
indicates reasonably precise interpolated values. This research paper clarifies that every
geoscience parameter should be taken as uncertain as even the reliable and best interpolation
algorithms do not accurately capture subsurface variability hence the error parameters
recorded. This difference in RMSE and absolute error between IDW and Kriging were
insignificant hence IDW is not prone to biased interpolation estimates. The equal standard
deviation of IDW and Kriging porosity values also reveals the consistency of the two
methods. The low but equal standard errors and deviations of the two methods for both
realisations indicate a slight difference between them and they gave comparable, consistent,
precise and acceptable interpolated values with less than 5% difference in both cases. Both
methods are acceptable for interpolation as the 95% confidence intervals are less than 10% of
the predicted mean. The recorded 95% confidence intervals for both porosity and
permeability realisations prove that IDW shows less uncertainty than kriging.
6.3 Volumetrics
Both visual assessment and statistical comparisons showed that IDW was most likely to give
the best and consistent volumetric results to kriging as compared to ID2W and ID4W which
contradicts a lot of papers (Al-Hassan & Adjei, 2015) but agrees with many as well
(Robinson & Metternicht, 2005; Sajid, et al., 2013). This research does not explicitly put this
contradiction to rest but the findings based on cross-validation results shows compelling
consistency and precision of IDW over ID2W and ID4W. From the zonal hydrocarbon
volumes of realisation 3, the differences between kriging, IDW, ID2W and ID4W is as a result
of smoothing caused by increasing the polynomial. Further comparison as a result of dividing
the reservoir into segments produced slight disparity between the performance of kriging and
the three power polynomials of Inverse Distance Weighting. Whilst the volumetric results
were consistent with the statistics, it was somehow at odds with results of (Al-Hassan &
Adjei, 2015), who reported that ID2W was more accurate to kriging than the other power
polynomials. A possible explanation for this discrepancy is that his work is based on using
ordinary kriging to model gold mineralisation that has been transported and deposited in a
46 | P a g e
sedimentary environment hence forming around the circumference of quartz conglomerates

whilst this paper is about using kriging (collocated cokriging) to define pore spaces and its
spatial connectivity which is primarily controlled by deposition but has been affected by
secondary processes. Upscaling results do not change any of the results significantly albeit
further strengthening the positive correlation between IDW and Kriging.
Undoubtedly kriging is the best linear unbiased estimator of petrophysical properties in

reservoir modelling. The method, though robust, gives good results and better uncertainty
characterisation when enough well data is attained hence the requirement for more wells to be
drilled. Inverse distance, on the other hand, proved that with 14 wells it can give improved
geological models, petrophysical distribution and volume calculations. This is based on the
inference drawn from the results that it outperformed kriging by recording the least standard
error, variance, uncertainty and similar standard deviation results. This can be possible if
more time is spent on the variogram model as this has an impact on the interpolation method
to be used. Defining different variograms to better fit the petrophysical data had an impact on
the volumetric report for the field and the inferences of the results of this project. IDW has
proved to be the most appropriate alternative to kriging based on this research and previously
published papers (Robinson & Metternicht, 2005; Sajid, et al., 2013; Husanovic & Malvic,
2014; Al-Hassan & Adjei, 2015).
6.4 Limitations
The whole point of this project is the use of fewer well data hence the true heterogeneity of
the reservoir was not captured. This was, however, defined by the variogram used for facie
modelling. The project had very sparse well log data with the only logs available being GR,
facies, porosity and permeability logs. This clearly made it difficult to determine several
factors like fluid contacts. Although depths from literature were used as fluid contacts depth,
the sealing capacity of the faults would have caused the fluid contacts to differ from each
reservoir segments. The pressure data, if provided, could have made it possible to determine
the different contacts and also know the segments that are in communication. This would
have changed the hydrocarbon volumes but not affected the results achieved.
6.5 Uncertainty Analysis

Analogue variogram parameters, simplifications, seed number and assumptions used in
building the fluvial facie, porosity and permeability models creates uncertainty in the model.
The inherent uncertainty in this research is captured in the error analysis of the interpolation
47 | P a g e
algorithms. The use of one (1) realisation for the fluvial facie model causes variability of
possible outcomes. In quantifying the spatial correlation, inadequate data led to the
introduction of substantial uncertainty. Analogue information from published works was used
to address these uncertainties and measure adequately the spatial variability within the
reservoir. The use of three different variogram parameters represented statistically as close as
possible the heterogeneity nature of the reservoir with the exception of realisation 1 which
was not used in further analysis. The use of at least one (1) horizontal well will provide
invaluable data input for variogram modelling of horizontal ranges.
To run a simulation on the model, an uncertainty analysis should be performed to achieve a

conservative model. To study the uncertainty based on the fluvial facie model and its effect
on hydrocarbon volumes, a Monte–Carlo stochastic sampling algorithm should be carried out
to sample random uncertain variables.
48 | P a g e
7 CONCLUSION
The choice of a suitable interpolation method with high accuracy, precision, consistency and
minimum errors is of chief importance in reservoir modelling. This research was divided into
three (3) important parts; quantification of spatial variation within the reservoir, assessing the
accuracy of the interpolated porosity and permeability properties using Kriging and IDW of
various power index algorithms and hydrocarbon volumes calculation in the reservoir.
The general reservoir variogram accounted for two (2) types of anisotropic behaviour
between the horizontal and vertical semivariances:
1. Zonal anisotropy caused by areal trends and stratifications.

2. Geometric anisotropy caused by varying ranges in 3 dimensions.
Inferences drawn from the statistical results of the interpolated properties of this research are
that:
1. Any of the four algorithms can be used as a suitable algorithm for modelling porosity
and permeability properties due to marginally insignificant differences with respect to
accuracy, consistency, and precision.
2. From the statistical parameters, realisation 3 recorded the least interpolation errors
and the most consistency and accuracy with IDW outperformed ID2W and ID4W by
recording the most comparable results to kriging.
3. Cross-validation of realisation 3 IDW and kriging relationship proved to be the best
realisation by recording the least average errors and highest correlation with both
having absolute errors of less than 1%.
4. This study, however, concludes that statistically in a fluvial-deltaic reservoir, Inverse
Distance Weighting algorithm with a power index of 1 is a suitable and better
alternative to kriging when there is sparse well data.
This conclusion is based on a systematic approach to accurately compare the interpolation

ability of IDW to kriging.
Similar results were achieved in the assessment of hydrocarbon calculations with the
following inferences drawn from the results:
1. Comparisons by reservoir zones reported realisation 3 Inverse Distance Weighting

method as the better of the two realisations.
49 | P a g e
2. Realisation 2 volumetric comparison between IDW and kriging were about 14% for
recoverable oil and 17% for recoverable gas which is above the accepted threshold of
10% for this research. However, realisation 3 results fell below the acceptable cut-off
for differences recording about 8% for both recoverable oil and gas.
3. Reservoir segments cross-validation and correlation analysis results indicate IDW to
have no significant difference to kriging with absolute errors of 3% for recoverable oil
and 0.7% for recoverable gas.
4. IDW results exhibited a correlational difference to kriging of ±2.2 MMbbls for
recoverable oil and ±0.3 BCF of recoverable gas. With IDW showing statistically
fewer interpolation errors and deviation, it exhibits a higher accuracy than kriging
volumetric results.
5. Grid upscaling, which normally affects models by causing a loss of geological
features and extreme porosity values, did not impact the results. It rather
complemented the robustness of IDW in both fine and coarse grid upscale.
50 | P a g e
8 RECOMMENDATIONS
1. More time needs to be spent on variogram models with different parameters not only
for facies but zones as well to adequately capture zonal anisotropy.
2. With enough additional well data, object based modelling should be used for facie
modelling for a more realistic representation of the channel pattern and connectivity.
3. When there are less than 15 wells, Inverse Distance Weighting method should be
employed in porosity and permeability modelling in a fluvial-deltaic shallow marine
reservoir. When more wells are drilled and information is attained, variogram model
can be improved and kriging can be used to attain very accurate result. This will give
better value for money.
4. Inverse Distance Weighting method could be used to estimate hydrocarbon volumes
when there is sparse data, time and budget constraints.
8.1 Further Studies

Sensitivity study should be carried out to check the reliability of the variogram parameters
used to indicate its impact on reservoir flow behaviour. Quantifying the impact will clearly
underline the importance of variogram modelling. This will also better test the robustness of
Inverse Distance Weighting method.
51 | P a g e
Bibliography
Ahmadov, M. et al., 2012. Technical Report - Part A and B Improved Oil Recovery from Gullfaks -
IOR Challenge 2. [Online]
Available at: http://www.ipt.nynu.no/~kleppe/pub/Gullfaks-Reports-2013/G1-technical.pdf
[Accessed 3 August 2016].
Al-Hassan, S. & Adjei, D., 2015. Competitiveness of Inverse Distance Weighting Method for the
Evaluation of Gold Resources in Fluvial Sedimentary Deposits: A Case Study. Journal of
Geosciences and Geomatics, 3(5), pp. 122-127.
Amirbayov, T. et al., 2013. Subsea compression Gullfaks Sør satellite field EiT - Gullfaks village
2013 group 1. [Online]
Available at: http://www.ipt.ntnu.no/~kleppe/pub/Gullfaks-Reports-2013/G1-technical.pdf
[Accessed 18 July 2016].
Armstrong, M., 1984. Improving the estimation and modelling of the variogram. In: Geostatistics for
Natural Resources Characterization: Reidel. Dordrecht, Holland: s.n., pp. 1-20.
Collins, F. C. J. & Bolstad, P. V., 1996. A comparison of spatial interpolation techniques in

temperature estimation. In: Proceedings of the Third International Conference/Workshop on
Integrating GIS and Environmental Modeling. Santa Barbara: s.n.
Doyen, P. M., 1988. Porosity from Seismic Data: A Geostatistical Approach. Geophysics 53.
Fossen, H. & Hesthammer, J., 2000. Possible absence of small faults in the Gullfaks Field, northern
North Sea: implications for downscaling of faults in some porous Sandstone. Journal of Structural
Geology, Volume 22, pp. 851-863.
Goovaerts, P., 1997. Geostatistics for natural resources evaluation, New York: Oxford University
Press.
Gringarten, E. & Deutsch, C. V., 2001. Variogram Interpretation and Modeling. Mathematical
Geology, 33(4), pp. 507-534.
Husanovic, E. & Malvic, T., 2014. Review of deterministic geostatistical mapping methods in
Croatian hydrocarbon reservoirs and advantages of such approach. NAFTA, 65(1), pp. 57-63.
Isaaks, E. H. & Srivastava, R. M., 1989. An introduction to applied geostatistics, New York: Oxford
University Press.
Journel, A. G. & Huijbregts, C. J., 1991. Mining Geostatistics. London: Academic Press.
Kupfersberger, H. & Deutsch, C. V., 1999. Methodology for Integrating Analog Geologic Data in 3-D
Variogram Modelling. AAPG Bulletin, 83(8), pp. 1262-1278.
Li, J. & Heap, A. D., 2014. Spatial interpolation methods applied in the environmental sciences: A
review. Environmental Modelling & Software, Volume 53, pp. 173-189.
MacEachren, A. M. & Davidson, J. V., 1987. Sampling and Isometric Mapping of Continuous
Geographic Surfaces. The American Cartographer, 14(4), pp. 299-320.
52 | P a g e
Malvic, T., 2006. Basics of geostatistics theory. [Online]

Available at: http://studylib.net/doc/7895958/basics-of-geostatistics-theory
[Accessed 17 August 2016].
Matheron, G., 1989. Estimating and Choosing: An Essay on Probability in Practice. 1st ed. Berlin:
Springer-Verlag.
Myers, R. H., 1990. Classical and Modern Regression with Applications. 2nd ed. Boston: PWS-Kent
Publishing Company.
Naoum, S. & Tsanis, I. K., 2004. Ranking Spatial Interpolation Techniques Using a GIS-Based DSS.
Global Nest: the Int. J., 6(1), pp. 1-20.
NPD, 2016. [Online]

Available at: http://www.norskpetroleum.no/en/production/field/gullfaks/
[Accessed 5 May 2016].
Olea, R. A., 1995. Fundamentals of semivariogram estimation, modeling and usage. In: Stochastic
modeling and geostatistics: Principles, methods and case studies. s.l.:AAPG Computer
Appliancations in Geology, pp. 27-36.
Petterson, O. et al., 1990. The Gullfaks field: geology and reservoir development, Netherlands:
Springer.
Prosvirnov, M., 2010. Evidence for tectonic control deposition of the Brent Group, northern North
Sea, Stavanger: University of Stavenger, Norway.
Robinson, T. P. & Metternicht, G., 2005. Testing the performance of spatial interpolation techniques
for mapping soil properties. Computer and Electronics in Agriculture, Volume 50, pp. 97-108.
Rocha, M. M., Yamamoto, J. K., Watanable, J. & Fonseca, P. P., 2012. Studying the influence of a
secondary variable in Collocated Cokriging estimates. Anais da Academia Brasileira de Ciéncias,
84(2), pp. 335-346.
Sajid, A. H., Rudra, R. P. & Parkin, G., 2013. Systematic evaluation of kriging and inverse distance
weighting methods for spatial analysis of soil bulk density. Canadian Biosystems Engineering, 55(1),
pp. 1.1-1.3.
Schlumberger, 2010. Petrel Introduction G&G Course. Houston: s.n.
Skøien, J. O., Merz, R. & Bloschl , G., 2006. Top-Kriging - geostatistics on stream networks.
Hydrology and Earth System Sciences, Volume 10, pp. 277-287.
Tobler, W. R., 1970. A computer movie simulating urban growth in the Detroit region. Economic
Geography, Volume 46, pp. 234-240.
Webster, R. & Oliver, M. A., 2001. Geostatistics for Environmental Scientists. West Sussex: John
Wiley & Sons Ltd.
Willmott, C. J., 1981. On the validation of models. Physical Geography, 2(2), pp. 184-194.
53 | P a g e
Xu, W., Tran, T. T., Srivastava, R. M. & Journel, A. G., 1992. Integrating Seismic Data in Reservoir
Modelling: The Collocated Cokriging Alternative. In: Annual Technical Conference of the SPE, 67,
Washington, Proceedings, pp. 833-842.
54 | P a g e
APPENDIX
A. Materials and Methods
Figure A 1 Mid-Skeleton Pillar Grid of the Gullfaks field showing active fault models within the reservoir zone.
55 | P a g e
Figure A 2 Well section showing original logs and upscaled logs for wells C1 and C2.
56 | P a g e
Figure A 3 Porosity-Permeability Cross plot showing facie distribution and relationship for Reservoir zones Tarbert
and Ness.
Figure A 4 Histogram comparison of IDW, ID2W, ID4W and Kriging for (A) Porosity Realisation 1, (B) Porosity
Realisation 2, (C) Porosity Realisation 3 and Permeability models
57 | P a g e
Table A 1 Variogram parameters used in facie modelling
Major Minor
Fluvial Facies dir. dir. Vertical Range Azimuth
Range Range
Levee 1000 500 10 25
Channel 3500 1500 10 25
Crevasse Splay 850 500 10 25
Table A 2 Summary statistics of porosity data for wells A10, A15, B2, B4, C1 and C2
STATISTICAL WELLS POROSITY DATA

PROPERTIES A10 A15 B2 B4 C1 C2
Possible Values 1835 552 2322 2524 933 1092
Min Value 0.103 0.023 0.091 0.091 0.025 0
Max Value 0.327 0.224 0.276 0.265 0.247 0.209
Median 0.228 0.184 0.211 0.204 0.110 0.092
Arithmetic Mean 0.212 0.155 0.183 0.180 0.115 0.088
Average Deviation 0.045 0.045 0.044 0.040 0.025 0.025
Standard Deviation 0.05 0.054 0.048 0.045 0.037 0.036
Variance 0.003 0.003 0.002 0.002 0.001 0.001
Skewness -0.383 -1.128 -0.506 -0.838 1.218 0.473
Kurtosis -1.228 -0.226 -1.467 -1.102 1.765 1.654
Normality Test: 2.61E- 3.21E- 2.39E-

0 0 6.02E-17
Kolmogorov-Smirnov 30 32 26
Quantile (Q1-10%) 0.139 0.054 0.114 0.108 0.074 0.044

Quantile (Q2-25%) 0.160 0.129 0.123 0.119 0.097 0.070
Quantile (Q3-50%) 0.228 0.184 0.211 0.204 0.110 0.092
Quantile (Q4-75%) 0.252 0.192 0.220 0.212 0.122 0.102
Quantile (Q5-90%) 0.267 0.197 0.225 0.218 0.178 0.119
58 | P a g e
Table A 3 Summary statistics of permeability data for wells A10, A15, B2, B4, C1 and C2
STATISTICAL WELLS PERMEABILITY DATA

PROPERTIES A10 A15 B2 B4 C1 C2
Possible Values 1835 552 2322 2524 933 1092
Min Value 4.275 2.939 3.479 8.586 1.306 1.035
Max Value 1672 918 1471 2801 952 1289
Median 102.651 96.845 107.203 119.067 77.811 74.786
Geometric Mean 102.465 66.707 106.209 123.917 71.793 62.712
Average Deviation 80.186 89.782 71.737 75.973 94.660 116.282
Standard Deviation 127.469 122.143 109.088 135.174 146.719 175.809
Variance 16248 14919 11900.3 18272 21527 30909
Skewness 3.958 1.924 3.426 6.879 2.758 3.023
Kurtosis 29 6 23 94 9 12
Normality Test:
0 1.60E-13 0 0 2.07E-32 0
Kolmogorov-Smirnov
Quantile (Q1-10%) 44.572 11.138 48.258 61.348 16.082 8.812

Quantile (Q2-25%) 69.054 20.607 71.866 82.166 36.483 27.722
Quantile (Q3-50%) 102.651 96.845 107.203 119.067 77.811 74.786
Quantile (Q4-75%) 164.135 173.466 164.466 175.020 147.183 170.094
Quantile (Q5-90%) 268.901 267.356 253.568 264.154 282.326 332.241
59 | P a g e
B. Results and Interpretations
7010 m
BCU
Tarbert
Ness
Figure B 1 Horizons and faults interpretation showing the lateral extensivity of the Gullfaks field and intersected
wells at seismic Inline 550
60 | P a g e
Figure B 2 Well section showing the reservoir zonal division, fluvial facies, porosity and permeability logs of wells C 6
and C7
61 | P a g e
Figure B 3 Horizontal semivariance and sample variogram of upscaled permeability data
62 | P a g e
Figure B 4 Vertical semivariance and sample variogram of upscaled permeability data
63 | P a g e
Figure B 5 Porosity/Permeability cross plots comparison for (A) Realisation 1, (B) Realisation 2, (C) Realisation 3 and
(D) Upscaled data
64 | P a g e
Figure B 6 Top Tarbert surface map of IDW and Kriging Realisation 3
65 | P a g e
Table B 1 Summary descriptive statistics of Kriging, IDW, ID2W and ID4W interpolated porosity and permeability
properties for realisation 2
REALISATION 2 POROSITY
STATISTICS
IDW ID2W ID4W KRIGING
Mean 0.127 0.125 0.124 0.124
Standard Error 8.958E-05 10.097E-05 11.078E-05 8.419E-05
Median 0.116 0.115 0.115 0.116
Mode 0.119 0.119 0.119 0.207
Standard Deviation 0.041 0.046 0.051 0.037
Sample Variance 0.002 0.002 0.003 0.001
Kurtosis 0.409 0.220 0.069 -0.070
Skewness 0.868 0.820 0.779 0.577
Range 0.292 0.292 0.292 0.292
Minimum 0.003 0.003 0.003 0.003
Maximum 0.295 0.295 0.295 0.295
Sum 26763 26295 26053 24520
Count 210388 210388 210388 197741
Confidence Level
1.756E-04 1.979E-04 2.171E-04 1.650E-04
(95.0%)
REALISATION 2 PERMEABILITY
Mean 73.060 75.705 78.498 69.854
Standard Error 0.234 0.252 0.273 0.237
Median 34.487 34.438 35.204 30.650
Mode 2.059 4.609 2.059 88.091
Sample Variance 10816.967 12619.850 14794.220 11046.863
Kurtosis 18.751 22.743 27.707 21.102
Skewness 3.391 3.779 4.192 3.754
Range 1508.177 1508.177 1508.177 1508.177
Minimum 1.362 1.362 1.362 1.362
Maximum 1509.539 1509.539 1509.539 1509.539
Sum 14463443 14986985 15540011 13762557
Count 197966 197966 197966 197020
Confidence Level
0.458 0.495 0.536 0.464
(95.0%)
IDW gives comparable values to Kriging than ID2W and
INFERENCE
ID4W
66 | P a g e
Table B 2 Summary descriptive statistics of Kriging, IDW, ID2W and ID4W interpolated porosity and permeability
properties for realisation 3
REALISATION 3 POROSITY
STATISTICS
Mean 0.127 0.125 0.124 0.123
Standard Error 9.187E-05 10.211E-05 11.167E-05 9.296E-05
Median 0.115 0.115 0.115 0.116
Mode 0.119 0.119 0.119 0.207
Sample Variance 0.002 0.002 0.003 0.002
Kurtosis 0.377 0.214 0.053 -0.388
Skewness 0.901 0.827 0.780 0.519
Range 0.292 0.293 0.293 0.293
Minimum 0.003 0.003 0.003 0.003
Maximum 0.295 0.295 0.295 0.295
Sum 26619 26337 26106 25034
Count 210245 210245 210245 203513
Confidence Level
1.801E-04 2.001E-04 2.189E-04 1.822E-04
(95.0%)
REALISATION 3 PERMEABILITY
Mean 73.060 75.705 78.498 69.854
Standard Error 0.234 0.252 0.273 0.237
Median 34.487 34.438 35.204 30.650
Mode 2.059 4.609 2.059 88.091
Sample Variance 10816.967 12619.850 14794.220 11046.863
Kurtosis 18.751 22.743 27.707 21.102
Skewness 3.391 3.779 4.192 3.754
Range 1508.177 1508.177 1508.177 1508.177
Minimum 1.362 1.362 1.362 1.362
Maximum 1509.539 1509.539 1509.539 1509.539
Sum 14463443 14986985 15540011 13762557
Count 197966 197966 197966 197020
Confidence Level
0.458 0.495 0.536 0.464
(95.0%)
IDW gives comparable values to Kriging than ID2W and
INFERENCE
ID4W
67 | P a g e
Table B 3 Volumetric results of the Gullfaks field for Realisation 2 and 3 showing the comparative results of IDW to
Kriging
REALISATION 2
Algorithm KG IDW Marginal diff
STOIIP GIIP STOIIP GIIP STOIIP GIIP
Volumes
(MMBBLS) (BCF) (MMBBLS) (BCF) (MMBBLS) (BCF)
Field Recoverable 446.590 89.912 496.910 103.600 0.113 0.153
Zone
Top Tarbert -
50.320 12.749 69.190 16.100 0.375 0.263
Tarbert 2
Tarbert2 -
100.640 16.139 113.220 21.370 0.125 0.324
Tarbert1
Tarbert1 - Top
94.350 17.234 94.350 17.690 0 0.027
Ness
Top Ness - Ness1 113.220 17.481 125.800 19.320 0.111 0.105
Ness1 - Top Etive 88.060 26.310 94.350 29.170 0.071 0.109
Average 0.137 0.165
REALISATION 3
Volumes
Zone
Top Tarbert -
56.610 14.656 69.190 15.790 0.222 0.077
Tarbert 2
Tarbert2 -
106.930 18.788 113.220 21.190 0.059 0.128
Tarbert1
Tarbert1 - Top
100.640 18.964 94.350 17.900 0.063 0.056
Ness
Top Ness - Ness1 119.510 18.894 125.800 19.280 0.053 0.021
Ness1 - Top Etive 94.350 32.101 94.350 28.680 0 0.107
Average 0.079 0.078
68 | P a g e
Table B 4 Upscaled grid porosity results
POROSITY Coarse Grid Fine Grid

STATISTICS IDW KRIGING IDW KRIGING
Mean 0.131 0.125 0.128 0.115
Standard Error 37.996E-05 10.211E-05 0.000 0.000
Median 0.124 0.115 0.125 0.111
Mode 0.099 0.119 0.207 0
Sample Variance 0.001 0.002 0.001 0.002
Kurtosis 0.204 0.214 -0.185 0.691
Skewness 0.652 0.827 0.232 0.170
Range 0.256 0.293 0.260 0.280
Minimum 0 0.003 0.000 0
Maximum 0.256 0.295 0.260 0.280
Confidence Level
7.448E-04 2.001E-04 8.28E-04 8.402E-05
(95.0%)
69 | P a g e
Table B 5 Volumetric results of the Gullfaks field for Fine and Coarse grid upscaling showing the comparative results
of IDW to Kriging
FINE GRID UPSCALE

Algorithm KG IDW Marginal diff
Volumes
Zone
Top Tarbert - 56.610 15.574 69.190 14.585 0.222 0.063
Tarbert 2
Tarbert2 - 106.930 20.518 113.220 18.717 0.059 0.088
Tarbert1
Tarbert1 - Top 100.640 17.834 94.350 18.858 0.063 0.057
Ness
Top Ness - Ness1 119.510 19.529 125.800 18.999 0.053 0.027
Ness1 - Top Etive 94.350 29.382 94.350 31.748 0.000 0.081
Average 0.079 0.063
COARSE GRID UPSCALE
Volumes
Zone
Top Tarbert - 62.900 15.574 69.190 17.622 0.100 0.132
Tarbert 2
Tarbert2 - 106.930 20.518 113.220 22.460 0.059 0.095
Tarbert1
Tarbert1 - Top 94.350 17.834 88.060 16.633 0.067 0.067
Ness
Top Ness - Ness1 119.510 19.529 125.800 19.741 0.053 0.011
Ness1 - Top Etive 94.350 29.382 88.060 25.921 0.067 0.118
Average 0.069 0.084
70 | P a g e
C. Workflow
Structural Modelling
Reservoir Grid & Boundary Make Horizons, Zones & Domain Conversion
Definition Layering
Quality Check Interpreted Data
Property Modelling
Statistical Analysis
Volume Calculation
Grid Upscaling
71 | P a g e

Static Reservoir Modelling by Comparing PDF

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Static Reservoir Modelling by Comparing PDF

Diunggah oleh

Hak Cipta:

Format Tersedia

UNIVERSITY OF MANCHESTER

Static Reservoir Modelling Comparing Inverse

Student ID: 9679414

School of Earth and Environmental Sciences

4.1.4 Zonation and layering ................................................................................................... 16

The final word count: 9,848

I wish to express my gratitude to Ghana National Petroleum Corporation-Oil and Gas

I am especially grateful to my supervisor Dr David Hodgetts for his constructive discussions

1.1 Problem Definition

1. Employ the two methods of estimation on the selected area of study

2.1 Inverse Distance Weighting

Equation 1 Inverse Distance Weighting method

Equation 2 Collocated cokriging method

𝑍 ∗ (𝑥0 )−𝑚1 𝑛 [𝑍1 (𝑥𝑎 )−𝑚1 ] [𝑍2 (𝑥𝑎 )−𝑚2 ]

3.3 Regional Geology

3.4 Local Geology

3.5 Petroleum System

3.5.3 Seal and Trap

4 MATERIALS AND METHODS

4.1 Structural Modelling

4.1.1 Fault modelling

4.1.2 Pillar gridding and boundary definition

4.1.3 Horizon modelling and depth conversion

4.1.4 Zonation and layering

4.2 Property Modelling

4.2.1 Geometrical modelling

4.2.2 Facies modelling

4.2.3 Statistical analysis of data

4.3 Petrophysical Modelling

4.3.1 Variogram modelling

Table 1 Variogram parameters used in porosity and permeability modelling

4.4 Volumetric Analysis

Equation 3 Volumetric Calculation

5 RESULTS AND INTERPRETATIONS

5.1 Petrophysical Analysis

5.2 Facie Modelling

5.3 Variogram Modelling

Figure 4 Horizontal Variogram maps of upscaled porosity and permeability data

Figure 5 Vertical semivariance and sample variogram of upscaled porosity data

Figure 6 Horizontal semivariance and sample variogram of upscaled porosity data

5.4 Petrophysical Modelling of Interpolating Methods

5.5 Assessment of Interpolation Methods

5.5.1 Porosity and Permeability models

5.6 Statistical Evaluation

porosity and permeability at their respective depth. In realisation 2 porosity interpolation,

A complete assessment of the model performance, as shown in Figure 13, involves

5.7 Volumetric Results

HISTOGRAM OF RECOVERABLE OIL IN GULLFAKS

100 Realisation 2 Kriging STOIIP

HISTOGRAM OF RECOVERABLE GAS IN GULLFAKS

20 Realisation 2 Kriging GIIP (BCF)

15 Realisation 2 IDW GIIP (BCF)

RECOVERABLE STOIIP COMPARISON BY

RECOVERABLE GIIP COMPARISON BY

RECOVERABLE STOIIP COMPARISON BY

RECOVERABLE GIIP COMPARISON BY

5.7.1 Cross-validation of recoverable hydrocarbons per reservoir segments

6.2 Statistics and Cross-validation

6.2.1 Accuracy and Precision

6.2.2 Consistency and Errors

sedimentary environment hence forming around the circumference of quartz conglomerates

Undoubtedly kriging is the best linear unbiased estimator of petrophysical properties in

6.5 Uncertainty Analysis