Attar A 2017 PHD Thesis.

EVALUATION OF THE IMPACT OF RESERVOIR HETEROGENEITY
AND MIXING ON LOW SALINITY WATERFLOODING
AHMED ZAKARIYA ATTAR
Doctor of Philosophy (PhD)
Earth Science and Engineering
Imperial College London
Supervised by:
Professor Ann Muggeridge

Professor Peter King
DECLARATION
I declare that the work presented in this thesis is my own. Where other sources
of information have been used they have been referred to.
Ahmed Zakariya Attar
1
COPYRIGHT NOTICE
The copyright of this thesis rests with the author and is made available under a Creative Commons
Attribution Non-Commercial No Derivatives licence. Researchers are free to copy, distribute or transmit the
thesis on the condition that they attribute it, that they do not use it for commercial purposes and that they do
not alter, transform or build upon it. For any reuse or redistribution, researchers must make clear to others
the licence terms of this work.
Ahmed Zakariya Attar: Evaluation of the Impact of Reservoir Heterogeneity and Mixing on Low Salinity
Waterflooding, PhD, © 2017
SUPERVISORS
Professor Ann Muggeridge
Professor Peter King
Time Frame:
2013-2017
ABSTRACT
This study investigated the impact of reservoir heterogeneity and physical diffusion/dispersion on low
salinity waterflooding (LSW) using detailed numerical simulation. Two different methods for modelling the
impact of LSW on relative permeability and hence on recovery were investigated: the salinity threshold
2
model (ECLIPSE and STARS) and the ion exchange model (GEM). These simulators were validated by
comparing their predictions with the analytical solution for immediate LSW.
Physical diffusion/dispersion were characterized using the dimensionless transverse dispersion number
(NTD) which was shown to be a robust measure for determining their impacts on the performance of LSW.
The impact of numerical diffusion was evaluated for a range of grid resolutions and a correlation developed
for estimating the maximum physical longitudinal dispersion that can be captured for a given grid resolution.
The impact of transverse numerical diffusion was found to be very small.
On laboratory scales, it was found that diffusion improves the oil recovery in layered high net-to-gross
models but reduces the effectiveness of continuous LSW in models containing a thick shale filled with
connate water. On the reservoir scale, the effect of dispersion on LSW was found to be minor compared with
reservoir heterogeneity. The adverse effect of an aquifer was found to be most significant during the
production of the connate water bank. If low salinity water is injected immediately as a slug, it was found
that the best slug size is around 0.4-0.6 PV assuming a total water injection of 1 PV. If it is injected after a
conventional water flood then the slug size increases to 1.3 PV assuming a total water injection of 2 PV.
Finally, an analytical pseudoization method was proposed to upscale LSW by modifying Hearn’s method for
layered reservoirs to include the connate water bank. This method was shown to give improved predictions
over those obtained by the original Hearn’s method when compared using a 2D layered model.
PUBLICATIONS
Part of this thesis were present in the following publications:
3
1. Attar, A., Muggeridge, A.H. 2015. Impact of Geological Heterogeneity on Performance of Secondary
and Tertiary Low Salinity Water Injection. Paper SPE 172775 proceedings of the SPE Middle East Oil
& Gas Show and Conference, Manama, Bahrain, 8-11 March.
2. Attar, A., Muggeridge, A.H. 2016. Evaluation of Mixing in Low Salinity Waterflooding. Paper SPE
179803 proceedings of the SPE EOR Conference at Oil and Gas West Asia, Muscat, Oman. 21-23
March.
ACKNOWLEDGEMENT
4
First, I would like to thank my academic supervisors for every support I had received during my PhD
research. I admit that my technical knowledge in reservoir engineering has significantly improved
throughout the years I spent at Imperial College London.
Second, I would like to thank Saudi Aramco management for funding my PhD research at Imperial College
London. The career development department are also thanked for their continuous support during my study.
Third, I would like to thank the support staff at Imperial College London and especially the Earth Science
and Engineering Department for creating a suitable environment for me to conduct my research.
Fourth, I would like to thank the Computer Modelling Group LTD for providing the STARS and the GEM
software. Also, I would like to thank Schlumberger for providing the ECLIPSE software.
Finally, I would like to thank my family for their support and encouragement during my study.
CONTENTS
CHAPTER 1: INTRODUCTION………………………………………………………………… 17
5
CHAPTER 2: LITERATURE REVIEW…………………………………………………………. 22
2.1 Low Salinity Waterflooding………………………………………………………...... 22
2.1.1 Description…………………………………………………………………....... 22
2.1.2 Methods for Modelling Low Salinity Waterflooding………………………….. 25
I. Salinity Threshold Model…………………………………………………...... 26
II. Ion Exchange Model………………………………………………………… 27
III. A Model that Captures Two Pore Scale Mechanisms……………….. 28
2.1.3 Pore Scale Mechanisms………………………………………………………... 28
2.1.4 Evidence Demonstrating the Success of Low Salinity Waterflooding………… 36
I. Coreflood Experiments……………………………………………………….. 36
II. Field Applications………………………………………………………….... 38
2.2 Mechanisms Limiting Reservoir Scale Recovery…………………………………....... 40
2.2.1 Mixing………………………………………………………………………….. 40
2.2.2 Heterogeneity………………………………………………………………....... 46
I. Quantifying Heterogeneity…………………………………………………… 47
II. Heterogeneity Effects on Various Displacement Processes ……………… 52
Effect on Vertical and Areal Sweep…………………………………………. 52
Crossflow and Mixing between Layers……………………………………… 53
Presence of Waterfilled Shales………………………………………………. 54
2.2.3 The Effect of Aquifers on Low Salinity Waterflooding……………………….. 55
2.3 Upscaling…………………………………………………………………………...... 56
2.4 Summary……………………………………………………………………………... 60
CHAPTER 3: MODELLING LOW SALINITY WATERFLOODING…………………………. 62
3.1 Reservoir and Fluid Properties Used for Modelling Low Salinity Waterflooding….. 62
3.2 Analytical Model for Low Salinity Waterflooding………………………………….. 63
Injection of Low Salinity Water post High Salinity Waterflooding………………… 71
3.3 Method……………………………………………………………………………..... 73
3.3.1 Grid Refinement Study and Validation………………………………………... 73
3.3.2 Validation of the Different Simulators to Model Low Salinity Waterflooding.. 74
3.3.3 Impact of Low Salinity Waterflooding Model on Predictions………………… 75
I. Salinity Threshold Model…………………………………………………..... 75
3.3.4 Modelling of Diffusion………………………………………………………… 76
A. Longitudinal Diffusion……………………………………………………… 76
B. Transverse Diffusion………………………………………………………… 77
3.4 Results………………………………………………………………………………... 79
3.4.1 Grid Refinement Study and Validation……………..…………..…………..…. 79
3.4.2 Validation of the Different Simulators to Model Low Salinity Waterflooding... 82
3.4.3 Impact of Low Salinity Waterflooding Model on Predictions………………… 83
I. Salinity Threshold Model…………………………………………………...... 83
3.4.4 Modelling of Diffusion………………………………………………………… 86
A. Longitudinal Diffusion…………………………………………………….... 86
B. Transverse Diffusion………………………………………………………… 91
3.5 Conclusions………………………………………………………………………… 92
CHAPTER 4: IMPACT OF HETEROGENEITY ON LOW SALINITY WATERFLOODING
EXCLUDING MIXING…………………………………………...…………..…………..……... 95
4.1 Continuous Low Salinity Waterflooding…………………………………………….. 95
4.1.1 Description of Models Used…………………………………………………… 95
A. 2D Layered Models with Simple Heterogeneity……………………………. 95
B. Areal Models Taken from SPE 10 Model 2 with Complex Heterogeneity…. 97
6
4.1.2 Results of Continuous Low Salinity Waterflooding…………………………… 98
4.2 Injection of Low Salinity Water as a Slug…………………………………………… 104
4.2.1 Analytical Method……………………………………………………………... 104
4.2.2 Simulation Study………………………………………………………………. 107
4.3 Conclusions and Recommendations…………………………………………………. 116
CHAPTER 5: IMPACT OF HETEROGENEITY AND PHYSICAL DIFFUSION AND
DISPERSION ON CONTINUOUS LOW SALINITY WATERFLOODING…………………... 119
5.1 Quantifying Diffusion/Dispersion…………………………………………………… 119
5.1.1 Characterizing Diffusion/Dispersion Using the Transverse Dispersion
Number (NTD)………………………………………………………………… 120
5.1.2 Quantifying the Impact of Diffusion on Oil Recovery using Df ……………… 122
5.2 Method……………………………………………………………………………...... 122
5.2.1 Immediate Injection of Low Salinity Water…………………………………… 125
A. Laboratory Scale Models……………………………………………………. 125
B. 3D Reservoir Scale Models with Complex Heterogeneity………………….. 126
C. A Reservoir with a Transition Zone Connected to an Aquifer……………… 127
5.2.2 Injection of Low Salinity Water Post High Salinity Waterflooding…………… 128
5.3 Results………………………………………………………………………………... 130
I. Two Layer Model………………………………………………………… 130
II. Five Layer Model………………………………………………………... 135
III. Shaly Sand Models……………………………………………………... 139
CHAPTER 6: IMPACT OF HETEROGENEITY AND PHYSICAL DIFFUSION AND
DISPERSION ON THE INJECTION OF LOW SALINITY WATER AS A SLUG……………. 161
6.1 Method……………………………………………………………………………...... 161
6.2 Results………………………………………………………………………………... 165
7
CHAPTER 7: UPSCALING LOW SALINITY WATERFLOODING………………………….. 190
7.1 Why Choose Hearn’s Method rather other Two-Phase Upscaling Methods?.............. 190
7.2 Description of Hearn’s Method……………………………………………………… 191
7.3 Modifications of Hearn’s Method by the Inclusion of Buckley Leverett Analysis…. 191
7.3.1 Upscaling High Salinity Waterflooding by a Modified Hearn’s Method……… 191
7.3.2 Upscaling Low Salinity Waterflooding by a Modified Hearn’s Method……… 192
7.4 Method……………………………………………………………………………….. 197
7.4.1 Upscaling Low Salinity Waterflooding……………………………………….. 197
A. Original Hearn’s Method……………………………………………………. 197
B. Modified Hearn’s Method…………………………………………………… 197
7.5 Evaluation of New Method……………….………………………………………….. 198
7.6 Results……………….……………………………………………………………….. 198
7.6.1 Upscaling High Salinity Waterflooding by a Modified Hearn’s Method……… 198
7.6.2 Upscaling Low Salinity Waterflooding………………………………………... 200
A. Original Hearn’s Method……………………………………………………. 200
B. Modified Hearn’s Method…………………………………………………… 202
I. Two Layer Model………………………………………………………… 202
II. Three Layer Model………………………………………………………. 204
III. Five Layer Model…………………………………………..…………... 206
7.7 Summary……………………………………………………………………………... 208
CHAPTER 8: CONCLUSIONS AND RECOMMENDATIONS………………………………... 210
8.1 Main Findings…………..…………..…………..…………..…………..……………. 210
8.2 Recommendations on How to Design and Model a Low Salinity Waterflood……… 212
8.3 Recommendations for Future Work…………..…………..…………..…………..…. 214
APPENDIX A: Sample CMG STARS input file…………..…………..…………..…………..… 216
REFERENCES …………………………………………………………………………………... 222
LIST OF FIGURES
Figure 1.1: The analytical Buckley Leverett solution for LSW (modified from Jerauld et al. 2008)….. 19
8
Figure 2.1: A typical set of relative permeability curves for LS and HS water………………...………. 23
Figure 2.2: The possible effect of mixing on the performance of a LS water flood……………………. 24
Figure 2.3: Water cut from two fields after LSW(Vledder et al. 2010, Korrani et al. 2014)…...……… 25
Figure 2.4: Incremental oil recovery from a tertiary LSW experiment (Jerauld et al. 2008)…...……… 25
Figure 2.5: Validation of the CMG GEM model with an experimental data (Dang et al. 2013)….…… 28
Figure 2.6: Validation of the coupled UTCHEM-IPhreeqc model with an experimental data (Korrani
et al. 2014)……………………………………………………………………………. …………..…… 28
Figure 2.7: Description of the fine scale as a proposed pore scale mechanism to recover oil by LSW
(Tang and Morrow 1999)………………………………………………………………….……………. 30
Figure 2.8: A typical structure of clays that are present in sandstone reservoirs (Hughes et al. 2012) 31
Figure 2.9: Illustration of the double layer expansion as a result of low salinity water injection
(Korrani et al. 2016)………………………………………………………. …………..…………..…… 32
Figure 2.10: The concentration of the divalent ions in the effluent water from an Alaskan field
(Jackson et al. 2016) …………………………………………………………………………………… 32
Figure 2.11: A schematic of a tube that represents the flow of desalinated water in the reservoir and
how it can recovery additional oil due to osmosis (Sandengen and Arntzen 2013) …………..……….. 33
Figure 2.12: Schematic of the experiment setup to investigate the effect of LS water on the
wettability of the mica sheets using a DSA instrument (Nasralla et al. 2013)…………………. ……... 34
Figure 2.13: The measured contact angle between mica/oil/brine at different brine concentrations
using a DSA instrument (Nasralla et al. 2013)…………………………………. …………..…………. 35
Figure 2.14: Schematic of the glass substrate model experiment to investigate the relationship
between the injected brine salinity and the mineral plate wettability (Mahani et al. 2014)……………. 35
Figure 2.15: Results of the contact angle measurements from the glass substrate model experiment to
investigate the relationship between the injected brine salinity and the mineral plate wettability
(Mahani et al. 2014)………………………………………………………….…………..…………….. 36
Figure 2.16: A typical coreflood experiment setup to investigate the potential of improving oil
recovery by LSW (Hadia et al. 2011)………………………………………………………….……….. 36
Figure 2.17: Different flow regimes based on NTD values (Lake and Hirasaki 1981)………………… 42
Figure 2.18: Best fit match of field and laboratory measured longitudinal dispersion (Mahadevan et
al. 2003)………………………………………………………………………...…………..…………... 43
Figure 2.19: Results of a simulation study proposing that numerical dispersion can approximate
physical dispersion in a 2D quarter five spot injection of LS water as slugs in a tertiary recovery stage
(Jerauld et al. 2008)…………………………………………………………..…………..…………….. 44
Figure 2.20: Comparison between UTCHEM simulation output, Muskat analytical solution for the
effect of stratification on the effluent profile and 1D C/D equation using a fitting parameter of the
apparent dispersivity (Coats et al. 2009)……………………………………………………………….. 45
Figure 2.21: Scales of geological heterogeneity (Henson et al. 2002)…………………………………. 46
Figure 2.22: An illustration showing a typical curve for plotting the (Fm) vs. (Hm) to quantify
reservoir heterogeneity using the Lorenz coefficient (Lake and Jensen 1989)………………………… 48
Figure 2.23: A typical F vs ∅ plot for different heterogeneity cases (Shook and Mitchell 2009)……… 50
Figure 2.24: The breakthrough time of the injected fluid in a miscible flood from each of the areal
layers taken from SPE 10 model 2 (Rashid et al. 2012)…………………………………..……………. 52
Figure 2.25: Different types of flow regimes in HSW of layered reservoirs (Baker 1998)……………. 53
9
Figure 2.26: Channelling in a fluvial channel sands through high k areas (Baker 1998)………... ……. 53
Figure 2.27: The incremental oil gained beyond HSW as a function of the fraction of LS water
injected into aquifers (Mahani et al. 2011)……………………………………………………..………. 55
Figure 2.28: A schematic of the periodic boundary condition (Pickup et al. 1994)…………… ……… 57
Figure 2.29: A schematic of a 2D stratified reservoir used in Hearn’s method to generate pseudo
relative permeabilities in HSW (Hearn 1971)………………………………..……….…………..……. 58
Figure 3.1: Relative permeability curves used for modelling LSW…………………………………… 63
Figure 3.2: Typical fractional flow curves for LS and HS water……………………………………… 66
Figure 3.3: A typical water saturation profile for a LS waterflood…………………………….……… 66
Figure 3.4: The water saturation of the two shock fronts and slope of the two lines that are used to
estimate their velocities in the reservoir model………………………………………………….……... 67
Figure 3.5: The analytical produced water cut by immediate LSW compared with that by immediate
HSW………………………………………………………………………………...…………..……… 70
Figure 3.6: The analytical water saturation profile at 2 PV of immediate LSW and HSW…….……… 70
Figure 3.7: Fractional flow curves for LSW post HSW………………………………………………... 72
Figure 3.8: The water saturation profile for LSW post HSW…………………………………………... 72
Figure 3.9: Schematic of the 1D simulation model used to evaluate the longitudinal diffusion in a
single phase flow…………………………………………………………………………..…………… 77
Figure 3.10: Schematic of the system used to evaluate the impact of truncation errors on the
simulation of transverse diffusion……………………………………………………………..……...... 78
Figure 3.11: Water breakthrough time (PV) for different number of grid blocks……………………... 79
Figure 3.12: The error difference in reproducing the analytical solution of LSW by the different
number of grid blocks. …………………………………………………………………………………. 80
Figure 3.13: The log of error difference in reproducing the analytical solution of LSW by the
different number of grid blocks against the log of the grid block size…………………………………. 80
Figure 3.14: Produced water cut versus and water salinity PVI for different number of grid blocks
compared with the analytical Buckley Leverett solution for LSW (ECLIPSE)………………………... 81
Figure 3.15: Produced water cut and water salinity versus PVI for different number of grid blocks
compared with the analytical Buckley Leverett solution for LSW (STARS)………………………….. 81
Figure 3.16: Water saturation at 0.4 PV predicted by ECLIPSE, CMG GEM and CMG STARS for
immediate LSW in a 1D homogeneous reservoir compared with the analytical Buckley Leverett
solution for LSW ……………………………………………………………………..…………..……. 82
Figure 3.17: Water saturation profile for various salinity thresholds…………………………..………. 83
Figure 3.18: Water cut versus PVI for various salinity thresholds…………………………….……….. 84
Figure 3.19: Oil recovery versus PVI for various salinity thresholds………………………………….. 84
Figure 3.20: Water saturation profile at 0.4 PV for immediate LSW for different scenarios using the
ion exchange model…………………………………………………………………..………….……... 85
Figure 3.21: Illustration of the impact of physical dispersion for a 1000 grid block simulation model
in which a LS water displaces a HS water………………………………………………………….…... 86
Figure 3.22: Back-calculated Pe compared with actual Pe for different number of grid blocks………. 86
Figure 3.23: The linear region of the back-calculated Pe versus actual Pe plot for different number of
grid blocks ………………………………………………………………………….………….……….. 88
Figure 3.24: The critical and plateau Pe versus the number of grid blocks……………………..……… 89
10
Figure 3.25: The linear region of the back-calculated Pe versus actual Pe plot for different number of
grid blocks using a CFL time step…………………………………….………….…………………….. 89
Figure 3.26: Critical Pe versus the number of grid blocks using a CFL time step…………..…………. 90
Figure 3.27: The salt concentration profile for different horizontal grid resolution at different Pe……. 90
Figure 3.28: Comparison between the stabilized salt concentration from the upper producer well (F1)
as predicted analytically and as predicted by the simulation model using a 200 × 20 grid for a range
of values of KT/v……………………………………………………………….…………..………….… 91
Figure 3.29: Effect of vertical grid resolution on the prediction of effluent salt concentration in the
upper well (F1) for a range of values of KT/v………………………………………………..…………. 91
Figure 3.30: The salinity distribution at 2 PV for different vertical grid resolution at two different
values of KT/v………………………………………………………………………….…………..…… 92
Figure 4.1: Relative permeability curves used for the 2D layered models to evaluate the impact of
reservoir heterogeneity on LSW……………………………………………………………..…………. 96
Figure 4.2: Permeability variations for the 2D layered models used to evaluate the impact of
heterogeneity on LSW…………………………………………………………………………..……… 97
Figure 4.3: Permeability distribution in areal layers 1 and 63 taken from SPE 10 model 2…….……… 97
Figure 4.4: Oil recovery comparison at 1 PV of immediate waterflooding in 2D layered cases……… 99
Figure 4.5: Drop in oil recovery for immediate LSW and LSW post HSW for the 2D layered cases
with simple heterogeneity variations……………………………………………………..…………..… 99
Figure 4.6: Oil saturation distribution at 1 PV for immediate LSW and immediate HSW in 2D
layered models with isotropic permeabilities…………………………………………………………... 100
Figure 4.7: Salinity distribution at 1 PV of immediate LSW in 2D layered models with isotropic
permeabilities…………………………………………………………………………………..……….. 100
Figure 4.8: Comparison between the drop in oil recovery for immediate waterflooding due to
heterogeneities with isotropic permeabilities case and the case with zero vertical permeabilities…….. 101
Figure 4.9: Oil saturation distribution at 1 PV of immediate LSW for two cases: with isotropic
permeabilities and with zero vertical permeabilities…………………………………………………… 101
Figure 4.10: Oil recovery at 1 PV of immediate LSW and immediate HSW versus the vorticity
heterogeneity index for layers 1 to 85 taken from SPE 10 model 2……………………………………. 102
Figure 4.11: Oil saturation distribution at 1 PV of immediate LSW and immediate HSW for layers 9
and 63 taken from SPE 10 model 2………………………………………………………………..…… 103
Figure 4.12: Water breakthrough (PVI) of immediate LSW and immediate HSW versus the vorticity
heterogeneity index for layers 1 to 85 taken from SPE 10 model 2…………………..…………..…… 103
Figure 4.13: Illustration of slug disintegration process in a layered reservoir (Wright et al. 1987) …… 106
Figure 4.14: Relative permeability curves used estimating of the best slug size of LS analytically 106
Figure 4.15: Estimation of the bypassing point analytically for different slug sizes of injecting LS
water using different permeability ratios between the layers……………………………….………….. 107
Figure 4.16: Estimation of the best slug size of immediate injection of LS water by the analytical
method at different permeability ratios between the layers…………………………..…………..……. 107
Figure 4.17: Oil recovery comparison at 1 and 2 PV for different slug sizes of injecting LS water into
2D layered models compared with continuous LSW………………………………..…………..…….. 108
Figure 4.18: Oil saturation distribution at 2 PVI for different slug sizes of immediate injection of LS
water into 2D layered models……………………………………………………………..……………. 109
Figure 4.19: Simulation results of oil recoveries obtained by different slug sizes of immediate 110
11
injection of LS water as a percentage of oil recovered obtained by continuous LSW at 2 PV….……..
Figure 4.20: Estimation of the best slug size of immediate injection of LS water by the analytical
method at different permeability ratios between the layers…………………………..…………..…… 110
Figure 4.21: Oil recovery at 2 PVI for slug sizes of LSW post immediate HSW in 2D layered models 112
Figure 4.22: Oil saturation distribution at 2 PV of total water injection for different slug sizes of
injecting LS water post immediate HSW in 2D layered models…………………………….…………. 113
Figure 4.23: Comparison between conventional waterflooding and injecting different slug sizes of
low salinity water post conventional waterflooding for the HL case…………………………………... 114
Figure 4.24: Oil recovery at 1 PV of total water injection versus Hv for different slug sizes of
immediate injection of LS water and LSW post immediate HSW in five layers ( SPE 10 model 2) ….. 115
Figure 4.25: Oil recovery at 1 PV of different slug sizes of immediate injection of LS water post
HSW for five heterogeneity realisations taken from SPE 10 model 2……………………..…………... 116
Figure 5.1: Schematic of the 2D simulation model used to identify flow regions……………….…….. 120
Figure 5.2: Plot of produced salt concentration versus PVI for three different NTD values ………….. 121
Figure 5.3: The water salinity distribution at 2 PVI for three different NTD values…………………… 121
Figure 5.4: The PVI for 50 % water salinity at the producer versus NTD……………………….……... 122
Figure 5.5: The LS water breakthrough versus NTD …………………………………………..……… 122
Figure 5.6: The HS and LS water relative permeability curves used in the simulations……….……… 122
Figure 5.7: Comparison between the fractional flow (fw) curves for the different sets of relative
permeability curves…………………………………………..………………………………………… 123
Figure 5.8: Comparison between the water saturation profiles (calculated analytically) at 0.06 PV for
both sets of relative permeability curves…………………………………………..…… 124
Figure 5.9: Comparison between the simulation outputs using different relative permeability curves.. 124
Figure 5.10: Permeability distribution in the 3D model taken from SPE 10 model 2 to evaluate the
impact of physical dispersion on LSW……………………………………………………..………….. 127
Figure 5.11: Model used to evaluate the effect of dispersion on LSW in a reservoir with an aquifer…. 128
Figure 5.12: Permeability variations for the 2D layered models used to evaluate the impact of
diffusion and reservoir heterogeneity on the injection of LS water ………….…………..…………..… 128
Figure 5.13: Cumulative oil recovery obtained from LS water injection into a two layer model as a
function of NTD and the thickness of the high k layer using two different salinity thresholds………… 131
Figure 5.14: Produced salt concentration versus PVI from LSW in a two layer model as a function of
NTD and the thickness of the high k layer using two different salinity thresholds….…………..……... 132
Figure 5.15: Oil recovery when the produced salt is 50 % as a function of NTD and the thickness of
the high k layer for LSW in a two layer model using two different salinity thresholds……………….. 133
Figure 5.16: Incremental oil recovery obtained by LSW in a two layer model as a function of NTD
and the thickness of the k permeability layer using two different salinity thresholds…….……………. 133
Figure 5.17: Oil saturation and water salinity distributions at 1 PV of LSW in a two layer model
using salinity thresholds of 1,000-30,000 ppm………………………………………………… 134
Figure 5.18: Oil saturation and water salinity distributions at 1 PV of LSW in a two layer model
using salinity thresholds of 1,000-7,000 ppm…………………………………………………..……… 134
Figure 5.19: Oil recovery obtained by LSW in a five layer model as a function of NTD and layer’s k 135
Figure 5.20: Produced salt concentration versus PVI from LSW in a five layer model as a function of
NTD and layer permeability using two different salinity thresholds…………………………..………. 136
Figure 5.21: Oil recovery at the time when the salinity at the producer is midway between the HS and 136
12
LS values for the two different five layer models using two different salinity thresholds……………..
Figure 5.22: The incremental oil recovery at 1 PV of LSW in a five layer model as a function of NTD
and layer permeability using two different salinity thresholds………………………………………… 137
Figure 5.23: Oil saturation and water salinity distributions at 1 PV of LSW in a five layer model as a
function of NTD and layer permeability for the salinity thresholds of 1,000-30,000 ppm…………….. 138
Figure 5.24: Oil saturation and water salinity distributions at 1 PV of LSW in a five layer model as a
function of NTD and layer permeability for the salinity thresholds of 1,000-7,000 ppm….………….. 138
Figure 5.25: Fractional decrease in the incremental oil recovery from LSW at 1 PVI as a function of
NTD from the two-layer models with a shale adjacent to the sand using two different thresholds…… 139
Figure 5.26: Fractional decrease in incremental oil recovery from LSW in a sand containing a water
saturated shale for different lengths as a function of NTD using two different salinity thresholds…… 140
Figure 5.27: Oil saturation and salinity distributions in a model with a thick shale saturated with
saline water above a thin oil bearing sand for the salinity thresholds of 1,000-30,000 ppm…………… 141
Figure 5.28: Oil saturation and salinity distributions in a model with a thick shale saturated with
saline water above a thin oil bearing sand for the salinity thresholds of 1,000-7,000 ppm…………… 141
Figure 5.29: The effect of shale porosity and system length on incremental oil recovery from LSW in
a sand containing a water saturated shale as a function of NTD using two different thresholds………. 142
Figure 5.30: Oil recovery obtained by injecting LS water into a 3D reservoir model taken from SPE
10 model 2 for the different scenarios using two different salinity thresholds…………….…………… 143
Figure 5.31: Oil saturation and water salinity distributions for immediate LSW in a 3D reservoir
model taken from SPE 10 model 2 using salinity thresholds of 1,000-30,000 ppm…..………….……. 145
Figure 5.32: Oil saturation and water salinity distributions for immediate LSW in a 3D reservoir
(SPE 10 model 2) using salinity thresholds of 1,000-7,000 ppm………………..…………..…………. 146
Figure 5.33: Oil recovery for several scenarios of LSW in an oil zone connected to an aquifer………. 147
Figure 5.34: Oil saturation and salinity distributions for LSW in a reservoir with an aquifer…………. 148
Figure 5.35: Oil recoveries at NTD = 1 compared with the no diffusion case for injecting LS water
post high salinity waterflooding into five different permeability realisations…………..……………… 151
Figure 5.36: Salinity and oil distributions at 1 PV for injecting LS water post conventional
waterflooding into five different permeability realisations for salinity thresholds 1,000-30,000 ppm… 152
waterflooding into five different permeability realisations for salinity thresholds 1,000-7,000 ppm…. 153
Figure 5.38: Oil recovery at 1 PV for injecting LS water post high salinity water into a 3D reservoir
(SPE 10 model 2) for two different salinity thresholds…………..…………..…………..……………. 155
waterflooding into a 3D reservoir (SPE 10 model 2) for salinity thresholds 1,000-30,000 ppm……… 156
waterflooding into a 3D reservoir (SPE 10 model 2) for salinity thresholds 1,000-7,000 ppm……….. 157
Figure 5.41: Oil recovery and the drop in oil recovery due to dispersion (Df) at 5 PV of oil zone for
injecting LS water post conventional waterflooding into a reservoir connected to an aquifer…………. 158
Figure 6.1: Permeability variations for the 2D layered models used to evaluate the impact of diffusion
and reservoir heterogeneity on the injection of LS water as a slug ………………….…………..…….. 162
Figure 6.2: Oil recoveries from various slug sizes of injecting LS water as function of diffusion for
five different permeability realisations for salinity thresholds of 1,000-30,000 ppm……………..…… 166
Figure 6.3: Oil recoveries from various slug sizes of injecting LS water as function of diffusion for 167
13
five different permeability realisations for salinity thresholds of 1,000-7,000 ppm…………..………..
Figure 6.4: The effect of salt diffusion on oil recoveries obtained by injecting different slug sizes of
LS into a five layer model as a function of layer permeability using two different thresholds………… 168
Figure 6.5: Water salinity and oil saturation distributions at 1 PV of injecting different slug sizes of
LS water into the LHLHL case using salinity thresholds of 1,000-30,000 ppm……….…………..…... 170
Figure 6.6: Water salinity and oil saturation distributions at 1 PV of injecting different slug sizes of
LS water into the LHLHL case using salinity thresholds of 1,000-7,000 ppm……….…………..…… 171
Figure 6.7: Oil recovery obtained by injecting different slug sizes of LS water in a 3D reservoir
model (SPE 10 model 2) for the different scenarios using two different salinity thresholds………….. 173
Figure 6.8: Oil saturation and salinity distributions at 1 PVI for different slug sizes of injecting LS
water into a 3D reservoir (SPE 10 model 2) using salinity thresholds of 1,000-30,000 ppm. ………… 174
Figure 6.9: Oil saturation and salinity distributions at 1 PVI for different slug sizes of injecting LS
water into a 3D reservoir taken from SPE 10 model 2 using salinity thresholds of 1,000-7,000 ppm… 175
Figure 6.10: Oil recoveries obtained by different slug sizes of injecting LS water into a reservoir
connected to an aquifer…………………………………………………………………………..…….. 176
Figure 6.11: Oil recoveries from different slugs sizes of LS water injected post high salinity
waterflooding for two different permeability variations…………..…………..…………..……………. 178
Figure 6.12: The oil recoveries at NTD = 1 compared to the no diffusion case for injecting different
slug sizes of LS water post high salinity waterflooding for two different permeability variations…….. 179
Figure 6.13: Salinity and oil saturation distributions at 2 PVI for injecting different slug sizes of LS
water post high salinity waterflooding into the LHLHL case…………..…………..…………..……… 180
Figure 6.14: The oil recoveries from different slug sizes of LS water injected post high salinity
waterflooding into a 3D reservoir (SPE 10 model 2) …………..…………..…………..…………..…. 182
Figure 6.15: Oil recoveries from different slug sizes of LS water injected post high salinity
waterflooding into a 3D reservoir (SPE 10 model 2) as a function of the oil recovery obtained from
continuous LSW at 2 PVI…………..…………..…………..…………..…………..…………..……… 182
Figure 6.16: Oil recoveries at NTD = 1 compared to the no dispersion case for injecting different slug
sizes of LS water post high salinity waterflooding into a 3D reservoir (SPE 10 model 2) …………… 182
Figure 6.17: Oil saturation and salinity distributions at 2 PVI for injecting different slug sizes of LS
water post high salinity waterflooding into a 3D reservoir (SPE 10 model 2) …………..……………. 184
Figure 6.18: Oil recoveries for different slug sizes of injecting LS water post high salinity
waterflooding into a reservoir connected to an aquifer…………..…………..…………..…………….. 185
Figure 6.19: Oil recoveries at NTD = 0.2 compared with the no dispersion case for injecting different
slug sizes of LS water post high salinity waterflooding into a reservoir connected to an aquifer……… 185
Figure 6.20: Oil saturation and salinity distributions at 10 PVI for injecting LS water post high
salinity waterflooding into a reservoir connected to an aquifer…………..…………..…………..…….. 186
Figure 7.1: Relative permeability for HSW……………………………….…………...…………......... 191
Figure 7.2: Regions of the water saturation profile for HSW …………..…………...…………............. 192
Figure 7.3: Relative permeability and fractional flow curves for LSW………………………………... 193
Figure 7.4: Graphical representation to estimate the slope of the two shock fronts for LSW…………. 194
Figure 7.5: Water saturation profile at 0.2 PVI calculated analytically using the BL solution of LSW 195
Figure 7.6: Regions of water saturation profile based on the salinities of each layer (LSW)………… 195
Figure 7.7: Pseudo relative permeability for HSW…………………………………………................... 198
Figure 7.8: Oil recovery, water cut and average pressure for HSW in a three layer model……………. 199
14
Figure 7.9: Pseudo relative permeability for LSW (Original Hearn’s Method)………………………... 200
Figure 7.10: Oil recovery, water cut and average pressure for LSW in a two layer model (Original
Hearn’s method)………………………………………………………………………………………... 201
Figure 7.11: Pseudo relative permeability for LSW in a two layer model……………........................... 202
Figure 7.12: Oil recovery, water cut and average pressure for LSW in a two layer model…………….. 203
Figure 7.13: Pseudo relative permeability for LSW in a three layer model……………......................... 204
Figure 7.14: Oil recovery, water cut and average pressure for LSW in a three layer model…………… 205
Figure 7.15: Pseudo relative permeability for LSW in a five layer model……………........................... 206
Figure 7.16: Oil recovery, water cut and average pressure for LSW in a five layer model …………… 207
LIST OF TABLES
Table 2.1: Different brine concentrations used in the experiment by Nasralla et al. (2013) to
investigate the effect of LSW on the wettability of the mica sheets using a (DSA) …………..………... 33
Table 2.2: Incremental oil recovery by LSW in various coreflood experiments……………..…………. 37
Table 2.3: Discussion on the cores’ wettability alteration due to LS water injection in various
laboratory experiments………………………………………………………………………..…………. 37
Table 2.4: Brine compositions for the initial brine and the injected LS brine used in the coreflood
experiments to investigate the potential of improving oil recovery by LSW ………………….……….. 38
Table 2.5: Incremental oil recovery by LSW in various field applications…………………..………… 38
Table 2.6: Clay content, water salinities and possible reasons for unsuccessful LSW in various field
applications…………………………………………………………………………………………..…... 39
Table 3.1: Rock and fluid properties used for modelling LSW…………………………………………. 62
Table 3.2: The analytical solution up to 2 PV of immediate injection of LS water into 1D
homogeneous model………………………………………………………………………….………….. 69
Table 3.3: Comparison between the shock fronts of HSW and LSW………………………..………….. 70
Table 3.4: Scenarios of number of grid blocks used for the grid refinement study…………..…………. 73
Table 3.5: Ion concentrations for connate water and the injected LS water used in CMG GEM……….. 74
Table 3.6: The interpolation parameter values for changing from HS water relative permeability curve
to LS relative permeability curve in CMG GEM……………………………………..…………..……... 74
Table 3.7: Summary of different salinity thresholds to evaluate their impacts on LSW………………... 75
Table 3.8: Different scenarios tested for modelling LSW using the ion exchange model…….……… 76
Table 3.9: Physical dimensions and properties of the 1D simulation model used to evaluate the
longitudinal diffusion in single phase flow………………………………………………..…………..… 77
Table 3.10: KT/v values used in 2D simulation model to evaluate the impact of truncation error on the
simulation of transverse diffusion……………………………………………….…………..…………... 78
Table 3.11: The critical Pe (the value of Pe above which numerical diffusion contributes significantly
to the mixing) as a function of the number of grid blocks used to discretize the model………………... 88
Table 4.1: Reservoir and fluid properties used in the 2D layered models to evaluate the impact of
heterogeneity on LSW…………………………………………………………………….…………..…. 96
Table 4.2: Description of the 2D layered models used to evaluate the impact of heterogeneity on 96
15
LSW…………………………………………………………………………………………………..…..
Table 4.3: Reservoir and fluid properties used for the estimation of the best slug size of LS water
analytically………………………………………………………………………….…………..……….. 106
Table 4.4: List of different slug sizes of LSW post HSW that were used in five different layered
models with simple heterogeneity variations……………………………………………………..……... 111
Table 4.5: List of different slug sizes of LSW post HSW used in the five heterogeneity realisations
taken from SPE 10 model 2…………………………………………………………………..………….. 115
Table 5.1: The model dimensions, fluid and rock properties used in a single phase model to
characterize diffusion using NTD …………………………………….…………..…………..…………. 120
Table 5.2: Cases examined for the impact of physical diffusion/dispersion and heterogeneity on
continuous LSW………………………………………………………………………….…………..….. 125
Table 5.3: The rock and fluid properties used in the laboratory scale models to evaluate the impact of
transverse diffusion on LSW……………………………………………………….…………..………... 126
Table 5.4: Description of the 2D layered models used to evaluate the impact of heterogeneity on LSW
post HSW using laboratory scale models with five different heterogeneity variations…………..……... 128
Table 5.5: Cases for PV of injected LS water post HSW that were used in five different layered
models with simple heterogeneity variations…………..…………..…………..…………..……………. 129
Table 5.6: Cases for PV of LSW post HSW that were used in a 3D reservoir (SPE 10 model 2) …….... 129
Table 5.7: Cases for PV of LSW post HSW that were used in a reservoir connected to an aquifer…….. 130
Table 5.8: Values used to calculate NTD at laboratory and reservoir scales…………..…………..……. 149
Table 6.1: Cases examined to evaluate the impact of physical diffusion/dispersion and heterogeneity
on injecting LS water as a slug…………………………………………….…………..…………..……. 161
Table 6.2; Slug sizes of LSW post HSW that were used in the LHLHL and HLHLH cases…………… 164
Table 6.3: Slug sizes of LSW post HSW that were used in a 3D reservoir (SPE 10 model 2) …………. 164
Table 6.4: Slug sizes of LSW post HSW that were used in a reservoir connected to an aquifer………... 165
Table 7.1: Reservoir and fluid properties used in the 2D model to test the proposed analytical method.. 191
Table 7.2: Average water saturation and average salinity for each region……………………………… 196
Table 7.3: Scenarios for testing the analytical method to upscale LSW………………………………... 198
Table 7.4: Pseudo relative permeability calculated by Hearn’s method………………………………… 200
CHAPTER 1: INTRODUCTION
16
Oil demand is increasing continuously as the global population grows (Zaid and Al-Dousari 2007). Many oil
fields across the globe are becoming mature and therefore need the deployment of enhanced oil recovery
methods. Low salinity waterflooding (LSW) is one of these methods. It is a relatively new method that has
caught the attention of various researchers in the oil industry (Tang and Morrow 1999, McGuire et al. 2005,
Lager et al. 2006, Jerauld et al. 2008, RezaeiDoust et al. 2009, Ligthelm et al. 2009, Austad et al. 2010,
Nasralla et al. 2013, Dang et al. 2015, Kuznetsov et al. 2015, Law et al. 2015). However, the success of its
deployment is affected by both reservoir heterogeneity which influences both sweep efficiency and the
degree of mixing between the injected low salinity water and in situ connate water. Increased mixing can
reduce the microscopic displacement of the process. To evaluate how both of these factors affect the
performance of low salinity waterflooding, a reservoir simulation study is needed. However, the results of
the simulator are affected by the numerical methods implemented and the number of grid blocks.
Low salinity waterflooding is defined as the injection of reduced/desalinated water into the reservoir to
enhance the oil recovery beyond the secondary stage. The improvement of oil recovery obtained by reducing
the water salinity compared with that obtained by high salinity water was first observed by Bernard (1967)
using coreflood experiments. However, since fresh water was used in these experiments for the low salinity
water, the idea of using it as enhanced oil recovery method was not further developed until the 1990s. Tang
and Morrow (1997) observed that decreasing the injected brine salinity increases the oil recovery in
laboratory experiments. Following these experiments, low salinity waterflooding was further investigated by
various researchers in the oil industry (for example: McGuire et al 2005, Lager et al. 2006, Hassenkam et al.
2012). Moreover, BP will implement low salinity waterflooding as an enhanced oil recovery method in the
Clair Ridge field (Robbana et al. 2012). This decision was made based on core flood experiments that have
confirmed the potential of improving oil recovery by low salinity waterflooding.
It is widely accepted in the oil industry that low salinity waterflooding recovers more oil by altering the
reservoir rock’s wettability making it more water wet (Jerauld et al. 2008, Nasralla et al. 2013, Dang et al.
2015, Kuznetsov et al. 2015, Law et al. 2015). However, the exact mechanism driving this wettability
alteration at the pore scale is still debated and is a subject of ongoing research (Jackson et al. 2016). To date,
several pore-scale mechanisms have been proposed: fines production (Tang and Morrow 1999), pH increase
(McGuire et al 2005), multicomponent ion exchange (Lager et al. 2006, Hassenkam et al. 2012) and double
layer expansion (Ligthelm et al. 2009, Nasralla and Nasr-EL-Din 2014), salting-in (RezaeiDoust et al. 2009),
or osmosis (Sandengen et al. 2016). The underlying pore scale mechanism for low salinity waterflooding is
still not fully understood. A critical review by Jackson et al. (2016) showed that each one of these
mechanisms sometimes showed promising results using laboratory experiments while it failed in other
experiments. This leads to the requirement of further research on the pore scale mechanism. However there
is very little research investigating how the recovery from low salinity water injection performance on the
field scale is modified by geological heterogeneities on all length scales from the core to that of the spacing
between wells.
17
In case low salinity waterflooding is selected to be implemented on field scale, then reservoir simulation is
needed to forecast the predictions of oil recoveries by low salinity waterflooding. Low salinity waterflooding
is modelled in reservoir simulators by two main methods. The first method proposed by Jerauld et al. (2008)
assumes the relative permeability curves are linearly dependent on the brine salinity within specified
thresholds, while the other method models the ion exchange that occurs between the injected low salinity
brine and the clay minerals before modifying the relative permeabilities. This results in low salinity
waterflooding producing two shock fronts (Figure 1.1) whereas there is only one shock front in conventional
high salinity waterflooding (Figure 1.1). The leading shock front is the region between the high salinity
water and the connate water and the trailing shock front is the region between the low salinity water and the
high salinity water. In between these shock fronts, a connate water bank is formed. The size of the connate
water bank is determined by the relative permeability curves and the salt diffusion and dispersion. This
diffusion and dispersion results in mixing between the saline connate water and the injected low salinity
water which in turn can reduce the effectiveness of the low salinity flood. Both of these mechanisms can
delay the improvement in recovery or indeed prevent the process being effective if low salinity water is
injected as a slug and the mixing causes the low salinity water slug to be destroyed. It is therefore essential
to carry out a detailed simulation study to investigate how reservoir heterogeneity affects the performance of
low salinity waterflooding.
1
Sf S1
Fractional flow
S2
high salinity water
low salinity water
0
0
Swc 1
Water saturation
18
1
S1
Water saturation
Sf
Connate Water S2 Swc
Bank
0
0 distance (ft) 3500
high salinity water low salinity water
Figure 1.1: The analytical Buckley Leverett solution for low salinity waterflooding. Low salinity waterflooding has two shock fronts which leads to
three regions: oil saturated zone (front), connate water bank (middle) and low salinity water zone (back). S2 is the leading shock front saturation and
S1 is the trailing shock front saturation. In comparison, there is only one shock front (Sf) in high salinity waterflooding. Swc is the connate water
saturation. They are described in details in chapter 3. The size of the connate water bank is determined by the relative permeabilities to oil and water
and the oil and water viscosities. The size of the connate water bank is modified by mixing. The effect of mixing on low salinity waterflooding is
further investigated in chapter 5 (modified from Jerauld et al. 2008).
While the impact of reservoir heterogeneity on conventional high salinity waterflooding has been studied
extensively and its effects are well known in the oil industry (e.g. Osman and Tiab 1981, Sorbie et al. 1987,
Ahmed et al. 1988, Cinar et al. 2004), its impacts on low salinity waterflooding have not been investigated.
In conventional waterflooding, vertical and areal sweep efficiencies are affected by heterogeneity.
Permeability contrast in layered reservoirs plays a key factor in influencing the vertical sweep efficiency. If
it is large and there is a very high displacement velocity, vertical sweep efficiency will be very poor leading
to an early water breakthrough (Coll et al. 2000, Baker 1998). Channelling may occur through high
permeability areas leading to a significant amount of bypassed oil in the low permeability areas which leads
to poor areal sweep. Since low salinity waterflooding has a similar mobility to conventional waterflooding, it
is anticipated that the effect of heterogeneity on its displacement will to some extent be similar to
conventional waterflooding. However, since the analytical solution of low salinity waterflooding is different
from conventional waterflooding, it is possible that heterogeneity will have different effects on low salinity
waterflooding.
While modelling reservoir heterogeneity requires very detailed simulation models, it is not always practical
to simulate the flow in such detailed models due to computation capability and time. This is particularly the
case in low salinity waterflooding as a very fine grid is needed to resolve the two shock fronts. It is therefore
essential to upscale from the fine grid to a coarse grid that contains effective flow properties. The literature
contains many methods (for example: Hearn 1971, Kyte and Berry 1975, Barker and Dupouy 1999,
Muggeridge and Hongtong 2014) for upscaling conventional waterflooding, however as yet there is no
published method for upscaling low salinity waterflooding.
Given the above information, it is essential to conduct a detailed simulation investigation study to evaluate
the impact of reservoir heterogeneity and the degree of mixing on the performance of low salinity
waterflooding. The main objectives of this thesis are to:
• Compare the various modelling methods for low salinity waterflooding with the analytical solution
and find what sensitivity parameters affect the simulation results and their robustness
19
• Conduct a grid refinement study to quantify how numerical dispersion affects the prediction of oil
recoveries by low salinity waterflooding
• Evaluate the impact of reservoir heterogeneity on the performance of low salinity waterflooding
using reservoir simulators
• Evaluate the impact of mixing between injected low salinity water and in situ connate water on low
salinity waterflooding and find how it affects the size of the connate water bank in addition to
finding how it is influenced by reservoir heterogeneity
• Find the best slug size of low salinity waterflooding as a function of reservoir heterogeneity and the
degree of mixing on both the laboratory and field scales
• Propose an analytical method for upscaling and reducing the dimensionality of 2D layered models
to 1D homogeneous models
In this study, the capillary pressure is neglected. In addition, only one rock type is used for both the
analytical solution of low salinity waterflooding and for the simulation study.
Thesis structure
Chapter 2 starts by describing low salinity waterflooding, its analytical solution, how it alters the rock’s
wettability, various methods for modelling it, several proposed pore scale mechanisms and field and
laboratory experiments that confirm its success in recovering additional oil. Then, three different factors that
affect low salinity waterflooding performance at the reservoir scale are discussed: mixing between connate
water and low salinity water, reservoir heterogeneity and the presence of an aquifer. Finally, the need to
upscale geological models to coarse models with effective properties that represent heterogeneity and
compensate for numerical dispersion is discussed.
Chapter 3 validates the chosen simulator by comparing the various methods of modelling low salinity
waterflooding to an analytical solution. In addition, a grid refinement study is conducted, first by excluding
mixing then by including it. When including mixing, a numerical dispersion study is conducted to find out
how much it affects the modelling of physical dispersion. Finally, the sensitivity of low salinity
waterflooding modelling to the choice of salinity thresholds is studied.
Chapter 4 evaluates the impact of reservoir heterogeneity on low salinity waterflooding excluding mixing.
First, low salinity water is injected continuously, then as a slug. In the slug mode, the best slug size is related
to various scales of reservoir heterogeneity in both laboratory and field scales.
Chapter 5 includes the effect of mixing in the investigation of reservoir heterogeneity on low salinity
waterflooding. Mixing is quantified by the use of the transverse dispersion number (NTD) (defined by Lake
and Hirasaki 1981). In this chapter, several heterogeneous models are used at the laboratory and reservoir
scales. On the laboratory scales, layered sands and shaly sands are used. On the reservoir scale, more
realistically geological heterogeneous reservoirs from the SPE 10 model 2 (Christie and Blunt 2001) and a
reservoir with a transition zone and an aquifer are used.
20
Chapter 6 includes the effect of mixing in the investigation of reservoir heterogeneity on the injection of
low salinity water as a slug. Mixing is quantified by the use of the transverse dispersion number (NTD)
(defined by Lake and Hirasaki 1981). In this chapter, several heterogeneous models are used at the
laboratory and reservoir scales. On the laboratory scales, layered sands are used. On the reservoir scale, more
realistically geological heterogeneous reservoirs from the SPE 10 model 2 (Christie and Blunt 2001) and a
reservoir with a transition zone and an aquifer are used. The best slug size is related to the degree of mixing
and reservoir heterogeneity.
Chapter 7 proposes an analytical method to upscale and reduce the dimensionality of 2D layered reservoir
to 1D homogeneous model by the modification of Hearn’s method to include the effect of connate water
bank.
Chapter 8 concludes the thesis and proposes areas for future research.
CHAPTER 2: LITERATURE REVIEW

The main purpose of this research is to evaluate the impact of reservoir heterogeneity and mixing on low
salinity waterflooding. In order to conduct this research it is vital to critically review the literature and find
the missing research areas that were not considered by previous work particularly on the effect of mixing
and reservoir heterogeneity on the performance of low salinity waterflooding.
Chapter 2 starts by describing low salinity waterflooding, various methods used to model it, the pore scale
mechanisms that are believed to improve oil recovery and laboratory experiments and field trials that show
low salinity waterflooding as a promising enhanced oil recovery method. Next, various mechanisms that
21
limit the success of low salinity waterflooding on the field scale are discussed: mixing, reservoir
heterogeneity and the presence of aquifers.
Finally, since it is not always economically feasible to build sufficiently fine grid models to capture the
detailed geological properties that might affect the low salinity water response, upscaling low salinity
waterflooding is discussed to include effective properties that represent the fine grid reservoir heterogeneity
and minimize numerical diffusion.
2.1 Low Salinity Waterflooding
2.1.1 Description
Low salinity waterflooding is defined as the injection of reduced/desalinated water into the reservoir to
enhance the oil recovery over that obtained by injecting seawater. Previous experiments by Tang and
Morrow (1999) suggested that the salinity of the injected water should be in the range of 1,000-7,000 ppm
compared with the typical salinities of seawater (30,000-54,000 ppm) (Nasralla et al. 2013). Other
researchers have reported that the low salinity water effect is not observed for salinities above 5,000 ppm
(Suman et al. 2013, Morrow and Buckley 2011, McGuire et al. 2005).
However, this range cannot be generalised as the thresholds are affected by the initial rock’s wettability. The
influence of the thresholds on the oil recovery by low salinity waterflooding is investigated in chapter 3.
On the Darcy scale, low salinity waterflooding increases oil recovery by altering the rock wettability from
oil/intermediate wet to more water wet. It is widely accepted that wettability alteration is the primary cause
of additional oil recovery (Nasralla et al. 2013, Dang et al. 2015, Kuznetsov et al. 2015, Law et al. 2015,
Jerauld et al. 2008).
Wettability is defined as the preference of one fluid to adhere to the rock surface (Nasralla et al. 2013). Since
the rock surface is not usually smooth, it is not possible to measure the wettability on the rock surface
directly. Instead, there are three main methods to measure the wettability: the Amott-Harvey method, the
USBM method and contact angle measurement on mineral surfaces that are present on the rock surface
(Bortolotti et al. 2010).
In the Amott-Harvey and USBM methods, the core’s bulk wettability is estimated from the calculated
Amott-Harvey wettability index (I) and the USBM wettability index (W). In both methods, the rock is
initially filled with oil and then water is injected into the core to displace the oil. The reverse displacement
occurs in the next step. In the Amott-Harvey method, I is calculated based on the volumetric ratios of the
displacing and displaced fluids in both displacement processes (Amott 1959). In the USBM methods, W is
estimated from the areas under the capillary pressure generated from both displacement processes
(Donaldson et al. 1969).
The third method for measuring wettability is by measuring the contact angle that is formed between a
mineral plate/oil/water. The plate is formed of the main mineral found in the reservoir rock of interest.
The plate is submerged in water and then an oil droplet is introduced to the system. When equilibrium is
reached, the contact angle is measured which indicates the mineral plate’s wettability (Anderson 1986).
22
Unlike conventional waterflooding which has only one shock front, low salinity waterflooding has two
shock fronts as shown previously in Figure 1.1. The oil saturated zone is ahead of the first (leading) shock
front while the low salinity water zone is behind the second (trailing) shock front. In between these two
shock fronts, a connate water bank is formed (Jearuld et al. 2008). The derivation of the analytical solution
of low salinity waterflooding is shown in chapter 3.
When low salinity water is injected into the reservoir, it will either fully displace all connate high salinity
water or mix with it to form intermediate salinity water. This intermediate salinity water reduces the residual
oil saturation but not as much as low salinity water as shown in Figure 2.1 (i.e. its relative permeability
curves fall between the relative permeability curves of low salinity water and the relative permeability
curves of high salinity water). Therefore, it is less effective than low salinity water (Jerauld et al. 2008). It is
anticipated that mixing resulting from diffusion/dispersion and reservoir heterogeneity affects low salinity
waterflooding by smearing the trailing shock front in the water saturation profile as shown in Figure 2.2.
This is a simulation result of a 1D laboratory scale model with dimensions of 30 × 3 × 3 cm that is
discretized into 200 grid blocks. The molecular diffusion is explicitly modelled and the calculated NTD
(defined in equation 2.11 in section 2.1) is 57. The impact of mixing on low salinity wateflooding is
evaluated in chapter 5. This mixing is caused by diffusion and dispersion and exacerbated by heterogeneity
and crossflow. The effects of mixing and reservoir heterogeneity are investigated in this thesis.
0.8
Relative permeability
0.6
0.4
0.2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Water saturation
krwH
krwH kroH
kroh krwL
krwL kroL
kroL krwinter
krwinter krointer
krointer
Figure 2.1: Relative permeability curves for high salinity water and low salinity water taken from a core experiment by Callegaro et al.
(2013). krwh and kroh are the water and oil relative permeability curves for high salinity waterflooding. krwL and kroL are water and oil relative
permeability curves for low salinity waterflooding (from Callegaro et al. 2013). It is anticipated that when low salinity water mixes with
connate water, intermediate salinity water forms. Its flow is controlled by the dashed relative permeability curves in this figure (krwinter and
krointer). This intermediate salinity water reduces the residual oil saturation but not as much as low salinity water, so it is less effective than
low salinity water.
23
1
0.9
0.8
0.7
Water saturation
0.6
0.5 analytical solution
0.4 with mixing
0.3
0.2
0.1
0
0 10 20 30 40
Distance (cm)
Figure 2.2: Mixing between the injected low salinity water and the connate water alters the performance of low salinity waterflooding by
smearing the trailing shock front. This is a simulation result of 1D laboratory scale model with dimensions of 30 × 3 × 3 cm that is
discretized into 200 grid blocks. The molecular diffusion is explicitly modelled and the calculated NTD (defined in equation 2.11 in section
2.1) is 57. The impact of mixing on low salinity waterflooding is evaluated in chapter 5.
Evidence of Connate Water Bank and Oil Bank during Low Salinity Waterflooding
Jerauld et al. (2008) proposed that during low salinity waterflooding, a connate water bank forms between
both shock fronts. In addition, they proposed that an oil bank occurs in tertiary low salinity waterflooding as
in most tertiary enhanced oil recovery methods. In this section, actual data from both field trials and a
laboratory experiment are provided to support that both banks form during low salinity waterflooding.
Low salinity waterflooding was implemented in two fields: as a water flood post primary oil recovery stage
in the Omar field in Syria (operated by Al Furat Petroleum Company) (Vledder et al. 2010) and as a
waterflood post high salinity waterflooding in the Endicott field in Alaska (operated by BP) (Korrani et al.
2014).
In the Omar field, it was not planned to implement low salinity waterflooding instead of high salinity
waterflooding but rather low salinity water was injected as a river was closer to the injection wells compared
to the sea (Law et al. 2014). In the Endicott field, low salinity water was injected as a field trial to test its
potential following successful core flood experiments (Korrani et al. 2014).
The actual produced water cut from both fields is shown in Figure 2.3. In the Omar field where low salinity
water was injected as the first recovery process, a connate water bank was formed and the produced water
cut stayed roughly constant for some time. In the Endicott field where low salinity water was injected after a
waterflood, the produced water cut decreases for some time which indicated the arrival of an oil bank.
24
Oil bank
Figure 2.3: Produced water cut from the Omar field (top) and the Endicott field (bottom). In the Omar field, a connate water bank was
formed and the produced water cut stayed roughly constant for some time. In the Endicott field, the produced water cut decreased for some
time which indicated the arrival of an oil bank (The top plot is from Vledder et al. (2010) and the bottom plot is modified from Korrani et al
2014).
In addition to the above field data, an oil bank was observed in experimental data from a tertiary low salinity
waterflooding experiment (Jerauld et al. 2008). The incremental oil production from this experiment was
measured and compared with results from a simulation model using the salinity threshold model to modify
the relative permeabilities and replicate the change in wettability during low salinity waterflooding. The
simulation model showed an excellent match with the experimental data (Figure 2.4). The experimental
results produced an oil bank. The analytical Buckley Leverett solution of injecting low salinity water post
high salinity waterflooding is described in chapter 3.
Experimental data
Simulation model
Figure 2.4: Incremental oil recovery from a tertiary low salinity waterflooding experiment, an oil bank formed (modified from Jerauld et al. 2008).
2.1.2 Methods for Modelling Low Salinity Waterflooding

Currently, there are three main methods to model low salinity waterflooding: the salinity threshold model,
the ion exchange model and a model (developed by Kuznetsov et al. (2015)) that captures two pore scale
mechanisms to recover oil by low salinity waterflooding. In all three models, the change in wettability
resulting from low salinity water injection is modelled by modifying the relative permeability curves. In the
salinity threshold model, the relative permeability curve is changed from the high salinity water curve to low
salinity water curve in response to a change in the salinity of the water in a grid cell, where the salinity is
25
essentially a tracer. In the ion exchange model, the exchange of ions between the water and the rock walls is
modelled and then the relative permeability curves are modified as a function of a selected ion (typically
Ca++) adsorbed on the rock walls.
I. Salinity Threshold Model
Low salinity water injection improves oil recovery by making the reservoir more water wet. This is
represented by using more water wet relative permeability curves as salinity is reduced. The user inputs a
low salinity pair of relative permeability curves (more water wet) and a high salinity pair of relative
permeability curves (more oil wet). In this model, proposed by Jerauld et al. (2008), salinity is tracked
through the reservoir, like a simple tracer. The salinity is used directly to calculate the relative permeability.
In addition, there are two salinity thresholds that are assumed and specified in the simulator: the upper
salinity threshold and the lower salinity threshold. When the salinity is higher than the upper salinity
threshold, the high salinity water relative permeability curves are used while when the salinity is lower than
the lower salinity threshold salinity threshold the low salinity water relative permeability curves are used.
When the salinity is between these two thresholds, then linear interpolation is used to determine the relative
permeabilities for oil and water as per the following equations (following Jerauld et al. 2008):
𝐻𝐻𝐻𝐻 (𝑆𝑆 ∗ ) 𝐿𝐿𝐿𝐿 (𝑆𝑆 ∗ )
𝑘𝑘𝑟𝑟𝑟𝑟 = 𝜃𝜃 𝑘𝑘𝑟𝑟𝑟𝑟 + (1 − 𝜃𝜃)𝑘𝑘𝑟𝑟𝑟𝑟 (2.1)
𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 = 𝜃𝜃 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 + (1 − 𝜃𝜃)𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 (2.2)
𝑃𝑃𝑐𝑐𝑐𝑐𝑐𝑐 = 𝜃𝜃 𝑃𝑃𝑐𝑐𝑐𝑐𝑐𝑐 + (1 − 𝜃𝜃)𝑃𝑃𝑐𝑐𝑐𝑐𝑐𝑐 (2.3)
𝜃𝜃 = 1 𝑛𝑛 ≥ 𝑛𝑛𝑢𝑢
𝑛𝑛− 𝑛𝑛𝑙𝑙
𝜃𝜃 = 𝑛𝑛𝑙𝑙 < 𝑛𝑛 < 𝑛𝑛𝑢𝑢
𝑛𝑛𝑢𝑢 − 𝑛𝑛𝑙𝑙
𝜃𝜃 = 0 𝑛𝑛 ≤ 𝑛𝑛𝑙𝑙 (2.4)
𝑆𝑆𝑜𝑜 − 𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜
𝑆𝑆 ∗ = (2.5)
1 − 𝑆𝑆𝑤𝑤𝑤𝑤 − 𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜
where:
𝑘𝑘𝑟𝑟𝑟𝑟 and 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 are water and oil relative permeabilities
superscripts HS and LS are high salinity and low salinity water
𝑃𝑃𝑐𝑐𝑐𝑐𝑐𝑐 is capillary pressure
𝑆𝑆 ∗ is a normalised water saturation for an intermediate salinity water
𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜 is the residual oil saturation for the intermediate salinity water, it is linearly interpolated between the
high salinity water and the low salinity water
𝑆𝑆𝑤𝑤𝑤𝑤 is the connate water saturation
𝑆𝑆𝑜𝑜 is oil saturation
𝑛𝑛 is the salinity of the water, 𝑛𝑛𝑢𝑢 is the upper salinity threshold and 𝑛𝑛𝑙𝑙 is the lower salinity threshold
𝜃𝜃 is the weighting factor, 𝜃𝜃 = 1 at the upper salinity threshold while 𝜃𝜃 = 0 at the lower salinity threshold
The salinity threshold model was validated by experimental data and showed excellent agreement as shown
previously in Figure 2.5 (Jerauld et al. 2008).
Although this model captures the wettability alteration as the main cause of the additional oil recovery using
low salinity water by reducing the residual oil saturation, it does not elaborate on the pore scale mechanism
26
that is causing this wettability alteration. Examples of commercial simulators that utilize this model are
ECLIPSE and CMG STARS.
II. Ion Exchange Model
In sandstone reservoirs that contain clay minerals and cations, injecting low salinity water disturbs the
chemical equilibrium in the reservoir (Dang et al. 2015). When low salinity water is injected, clay minerals
exchange the initially adsorbed organo-metallic complexes (which were formed due to the bonding of oil
with cations) with the uncomplexed cations that are present in the injected low salinity water thus oil is
released. This model assumes that multicomponent ion exchange (described in section 2.1.3) is the main
driving pore scale mechanism for low salinity waterflooding to alter the rock’s wettability towards more
water wet.
There are three main geochemical reactions that are thought to occur during the injection of low salinity
water: intra-aqueous reactions, ion exchange reactions and mineral dissolution reactions (Dang et al. 2015)
Ion Exchange Reactions
The following two reactions are thought to occur during low salinity waterflooding (Dang et al. 2013):
1 1
𝑁𝑁𝑁𝑁+ + (𝐶𝐶𝑎𝑎 − 𝑋𝑋2 ) ↔ (𝑁𝑁𝑁𝑁 − 𝑋𝑋) + 𝐶𝐶𝐶𝐶++ (2.6)
2 2
1 1
𝑁𝑁𝑁𝑁 + + (𝑀𝑀𝑀𝑀 − 𝑋𝑋2 ) ↔ (𝑁𝑁𝑁𝑁 − 𝑋𝑋) + 𝑀𝑀𝑀𝑀++ (2.7)
2 2
where:
X is the clay mineral available in the sandstone rock. It has a subscript of 2 since it is attached to divalent
positive ions (𝐶𝐶𝐶𝐶++ or 𝑀𝑀𝑀𝑀++) in the clay, while it is 1 when attached with single positive charge (𝑁𝑁𝑁𝑁 + ).
In this model the exchange of ions between the water and the rock walls is calculated and then the relative
permeabilities are determined as a function of the equivalent fraction of a selected ion on the rock walls
(typically Ca++). In this case the user has to specify the relative permeability curves to be used for different
equivalent fractions of the ion of interest. In the simplest case the relative permeabilities are calculated using
equations 2.1-2.5, the only difference being that n is now the equivalent fraction of the ion of interest rather
than salinity, 𝑛𝑛𝑢𝑢 is the upper threshold for the equivalent fraction of that ion and 𝑛𝑛𝑙𝑙 is the lower threshold
for that ion.
Examples of commercial simulators that utilize this model are the coupled modelling of UTCOMP with
IPhreeqc as proposed by Korrani et al. (2014) and the CMG GEM as proposed by Dang et al. (2013). Both
methods used in these simulators were validated by experimental data as shown in Figures 2.5 and 2.6.
27
Figure 2.5: The CMG GEM model shows an excellent match when validated with experimental data (from Dang et al. 2013).
Figure 2.6: Coupled UTCHEM-IPhreeqc model shows an excellent match with experimental data for both types of waterflooding: high
salinity waterflooding and low salinity waterflooding (from Korrani et al. 2014).
Unlike the salinity threshold model, the ion exchange model does not relate the oil recovery directly with the
salinity of the injected brine (Cotterill 2014).
III A Model that Captures Two Pore Scale Mechanisms of Low Salinity Waterflooding
Recently, Kuznetsov et al. (2015) developed a modelling method that captures two pore scale mechanisms:
multicomponent ion exchange and double layer expansion. The model analysed the type of force (Van Der
Waals, ligand exchange or cation bridging) that binds the absorbed ions and the clay surface. Coreflood
experiments from the literature were used to build the simulation model. Generally, the model is in a good
agreement with the experimental data. However, it requires further validation and testing by field data.
Currently, there is no commercial simulator that utilizes this model. Therefore, it is not used in this thesis.
2.1.3 Pore Scale Mechanisms
While the focus of this thesis is primarily on modelling low salinity waterflooding at the Darcy scale and
investigating how its efficiency is impacted by mixing and reservoir heterogeneity, some attention is given to
critically review the proposed pore scale mechanisms that cause the wettability alteration. In this section a
28
critical review is provided for these mechanisms. The injected low salinity water alters rock wettability
towards more water wet. However, due to the complex interaction between the rock/oil and brine, the exact
driving mechanism for this wettability alteration is still debatable and is a subject for further research
(Jackson et al. 2016). To date, several pore scale mechanisms have been proposed: fines production (Tang
and Morrow 1999), pH increase (McGuire et al 2005), multicomponent ion exchange (Lager et al. 2006,
Hassenkam et al. 2012) and double layer expansion (Ligthelm et al. 2009), salting-in (RezaeiDoust et al.
2009), or osmosis (Sandengen and Arntzen 2013).
1. Fines Production
Tang and Morrow (1999) proposed that oil recovery is improved by low salinity waterflooding through fines
production. Initially at irreducible water saturation, the polar components of the oil droplets are adsorbed by
clay minerals (fines) that exist as small particles within the pores and hence the pore becomes mixed-wet.
Unlike high salinity water, low salinity water causes the clay to swell. If the clay contains kaolinite, these
particles might become detached from the pore walls (Sheng 2014, Jackson et al. 2016).
The stripping (detachment) of the clay particles from the pore wall depends on the balance between the
mechanical forces and colloidal forces (DLVO) (Tang and Morrow 1999). The mechanical forces include
the viscous forces (that promote detachment) and capillary forces which result from the adsorption of crude
oil to these clay particles. Colloidal forces between these clay particles depend on the balance between the
electrostatic repulsion and van der Waals attractive forces (Tang and Morrow 1999). Injecting low salinity
water causes the electrical double layer to expand which increases the stripping (detachment) forces on the
clay particles. As a result, the clay particles are mobilized and thus oil is produced as shown in Figure 2.7.
29
Figure 2.7: An illustration of the proposed pore scale mechanism of recovering more oil by low salinity waterflooding due to fines
migration. Initially, the polar components of the oil droplets are adsorbed by clay minerals (fines) and hence the rock wall becomes mixed-
wet. Injecting low salinity water causes the clay to swell and if the clay contains kaolinite, they might be detached from the pore walls.
Injecting low salinity water causes the electrical double layer to expand which increases the stripping (detachment) forces of the clay
particles. As a result, the clay particles are mobilized and thus oil is produced (Tang and Morrow 1999).
However, fines production cannot be considered as the only pore scale mechanism that causes wettability
alteration since coreflood experiments conducted on sandstone rocks by Zhang et al. (2007) showed
additional oil recovery by low salinity waterflooding without the production of clays in the effluent fluids.
2. pH Increase
Low salinity waterflooding may recover more oil due to the observed pH increase in the effluent water
compared with the slight increase using high salinity waterflooding. The assumption is that initially when
low salinity water is injected, the concentration of OH- is increased due to the rock’s mineral reactions with
the cations present in the low salinity water (since injecting low salinity water disturbs the chemical
equilibrium of the system). As a result, the injected water becomes more alkaline and hence its pH is
increased (McGuire et al. 2005, Nasralla et al. 2013). These mineral reactions do not occur when injecting
high salinity water as its salinity is similar to the connate water salinity and hence the chemical equilibrium
is not disturbed. After low salinity water injection, the acid components that are present in the oil are
saponified and thus surfactants are generated. These surfactants lower the interfacial tension and the oil-
30
water contact angle, so the system wettability is changed towards more water wet. Finally, the oil recovery is
increased (McGuire et al. 2005, Jackson et al. 2016).
Further investigation by Lager et al. (2006) confirmed that there is no relation between the acidity and the
increased oil recovery by low salinity waterflooding which is not consistent the hypothesis that pH increases
is the pore scale mechanism of altering the system wettability.
3. Multicomponent Ion Exchange (MIE) and Double Layer Expansion
Lager et al. (2006) proposed that oil recovery is improved by low salinity waterflooding due to
multicomponent ion exchange (MIE) which causes the double layer to expand (Ligthelm et al. 2009).
Initially, polar compounds of oil are bonded with the divalent ions thus organo-metallic complexes are
formed. As a result, the rock becomes oil wet. Clays that are present in sandstone reservoirs usually have a
negative surface charge due to the imbalance in the structural charge between the aluminium octahedral
sheet and the silicon tetrahedral sheet as shown in Figure 2.8 (Austad et al. 2010, Hughes et al. 2012).
Figure 2.8: A typical structure of clays that are present in sandstone reservoirs. The imbalance in the structural charge between the aluminium
octahedral sheet and silicon tetrahedral sheet usually causes the clay’s surface charge to be negative (from Hughes et al. 2012).
Clays attract the cations (Ca++ and Mg++) that are present in the connate water because they are negatively
charged. As a result, a layer is formed (known as the Stern layer) as shown in Figure 2.9. In addition,
negative components that are present in the oil are attached to the edge of the Stern layer to form another
layer called a diffuse layer (Figure 2.9). These two layers are known as a double layer (Austad et al. 2010,
Hughes et al. 2012, Kuznetsov et al. 2015, Jackson et al. 2016, Sheng 2014, Nasralla and Nasr-El-Din 2014,
Ligthelm et al. 2009). When low salinity water is injected, the organo-metallic complexes are replaced by
uncomplexed cations. Injecting low salinity water causes the repulsion forces between the initially adsorbed
oil droplets (negatively charged) and the clay surface to increase hence releasing the oil droplets thus the
system wettability is changed towards more water wet as shown in Figure 2.9 (Lager et al. 2006, Hughes et
al. 2012, Sheng 2014, Kuznetsov et al. 2015, Jackson et al. 2016). This mechanism is supported by the
observed initial reduction in the divalent ions in the effluent water at a coreflood experiment (Lager et al.
2006) as shown in Figure 2.10.
31
Figure 2.9: Illustration of double layer expansion as a pore scale mechanism for low salinity waterflooding (a) Initially, cations in the
connate water are attracted to the negatively charged clay surface to form the Stern layer. The cations attract the negatively charged
components in the oil to form another layer called a diffuse layer. These two layers form a double layer. The thickness of the double layer
increases as the salinity of the injected brine decreases. (b) Injecting high salinity water slightly increases the thickness of the double layer.
(c) When injecting low salinity brine, the thickness of the double layer significantly increases which results in increasing the repulsion
forces between the initially adsorbed oil droplets to the clay minerals and the clay minerals thus releasing the oil droplets and changing the
wettability towards more water wet (from Korrani et al. 2016).
Figure 2.10: An initial reduction in the concentrations of the divalent ions in the effluent water is observed in a coreflood experiment.
This confirms that there is a multicomponent ion exchange that are thought to occur at the clay surface that replaces the originally
adsorbed organo-metallic complexes with the uncomplexed cations present in the injected low salinity brine. This leads to an increase
in oil recovery (from Sorbie and Collins 2010).
This mechanism suggests that injecting low salinity water promotes the cation exchange between the organo-
metallic complexes and the uncomplex ions and hence oil recovery is increased. However, this mechanism
does not cause a change in pH as was observed in a study by Austad et al. (2010) due to low salinity water
injection. It was suggested by Austad et al. (2010) that the local increase in pH was due to the exchange at
the clay surface between protons (H+) and the initially adsorbed cations. Therefore, MIE is not the primary
pore-scale mechanism for wettability alteration due to low salinity waterflooding.
4- Salting-in
The solubility of oil’s polar compounds in water decreases when salt is added to the formation water.
RezaeiDoust et al. (2009) named this process as salting out. During the injection of low salinity water, the
reverse process takes place, which is salting-in. Injecting water with a salinity that is below a critical value
increases the solubility of oil in water compared to conventional waterflooding (seawater) (RezaeiDoust et
al. 2009).
32
When low salinity water is injected, the solubility of oil in the injected low salinity water is increased.
Therefore, the oil initially adsorbed on the clay surface dissolves in the low salinity water meaning the clay
surface becomes more water wet and the incremental oil recovery is increased due to changes in the relative
permeability curves.
However, studies conducted by Austad et al. (2010) indicated no significant difference in the desorption
development between high and low salinity water injection which clearly contradicts with this mechanism.
6- Osmosis
Sandengen and Arntzen (2013) proposed that during low salinity waterflooding, the rock’s wettability is
changed from initially an oil wet to a water wet by osmosis. The assumption is that the rock is oil wet and oil
surrounds connate water droplets. The oil acts as a semi-permeable membrane that allows the movement of
water but not ions (Sandengen and Arntzen 2013, Sandengen et al. 2016). However, injecting water with a
lower salinity than the connate water can result in osmosis through the oil as the water tries to dilute the
more saline connate water. Injecting water with a similar salinity to the connate water does not promote
osmosis through the oil as shown in Figure 2.11 (Sandengen and Arntzen 2013). The tube (100 µm)
represents the flow of water in a reservoir while the oil film (50 µm) represents a typical pore diameter.
Figure 2.11: A schematic sketch of a tube that represents the flow of desalinated water in the reservoir and how it can recover additional
oil due to osmosis as a proposed pore scale mechanism for low salinity waterflooding. Desalinated water flows through the tube at a
constant rate of 1 m/d and diffuses through the oil membrane at a constant diffusion rate. Desalinated water diffuses through the oil
membrane due to the difference in chemical potential of desalinated water and saline water. As a result of the increase in osmotic
pressure, oil will flow to the saline water and through to the main tube. This diversion of desalinated water away from the main tube can
relocate oil (from Sandengen and Arntzen 2013).
The assumption here is that low salinity water (desalinated water in Figure 2.11) flows through the tube at a
constant rate of 1 m/d and diffuses through the oil membrane at a constant diffusion rate. Low salinity water
diffuses through the oil membrane due to the difference in chemical potential of low salinity water and saline
water. As a result of the increase in osmotic pressure, oil will flow to the high salinity water (saline water in
Figure 2.11) and through to the main tube. This diversion of low salinity water away from the main tube can
relocate oil (Sandengen and Arntzen 2013).
It is proposed that in a porous media, low salinity water diffuses through the oil which acts as a semi-
membrane to the connate water. With continuous injection of low salinity water, the size of the connate
water droplets increases (due to the increase in the osmotic pressure gradient) until the oil droplets are
released thus the system wettability is changed towards more water wet (Sandengen and Arntzen 2013).
This hypothesis was confirmed using a simple visualization experiment using a glass tube by (Sandengen
and Arntzen 2013). The glass tube included oil droplets, high and low salinity brines. The experiment
showed movement of oil droplets as a result of low salinity brine diffusing to high salinity brine. When
brines with the same salinity were used, oil droplets did not move.
33
However, this mechanism cannot be the only pore scale mechanism for wettability alteration by low salinity
waterflooding since coreflood experiments conducted by Ashraf et al. (2010) found that low salinity
waterflooding improves the oil recovery in neutral wet cores.
Wettability Alteration
Although the exact mechanism that is driving low salinity waterflooding for improving the oil recovery is
still uncertain, it is widely accepted that wettability alteration is the primary cause of improved oil recovery
(Nasralla et al. 2013, Dang et al. 2015, Kuznetsov et al. 2015, Law et al. 2015, Jerauld et al. 2008). Since
sandstones surface are not smooth, mineral plates are often used to represent simplified versions of
sandstones (Bortolotti et al. 2010, Anderson 1986).
Laboratory experiments by (Drummond and Israelachvili 2004, Nasralla et al. 2013 and Mahani et al. 2014)
have confirmed the wettability alteration of mineral plates by low salinity waterflooding.
A. Drop Shape Analysis Instrument Experiment
Mica is a mineral similar to the clays found in sandstones. It is often used to investigate the effects of salinity
and oil composition on wettability to minimize the effect of sandstone roughness on quantifying wettability
(Nasralla et al. 2013). A drop shape analysis (DSA) instrument was used to measure the contact angle
between mica/oil/brine as shown in Figure 2.12. The experiment was conducted at 1,000 psi and 212°F.
Table 2.1: Different brine concentrations used in the
experiment by Nasralla et al. (2013) to investigate the
effect of low salinity water on the wettability of the mica
sheets using a drop shape analysis instrument (DSA).
Brine Salinity (1,000 ppm)
Formation water (FW) 174
Seawater (SW) 54
Aquifer water (AQ) 5
10% diluted aquifer water 0.5
(10% AQ)
Figure 2.12: Schematic of the experimental setup used to investigate the

effect of low salinity water on the wettability of the mica sheets using a drop
shape analysis instrument (DSA). Mica sheets were used as they represent the
clays found within sandstones (modified from Nasralla et al. 2013).
Four brines with a wide range of salinities were used in the experiment as summarized in Table 2.1. The
aquifer water was sourced from a shallower depth than the formation (Nasralla et al. 2013). The experiment
was conducted for each brine. First, the mica was submerged in the brine. Once equilibrium between them
was achieved, an oil droplet was added. Then, the contact angle between mica/oil/brine was measured. These
steps were repeated for the different types of brine.
Figure 2.13 shows the contact angle between mica/oil/brine for the different types of brine. The contact
angle between mica/oil/brine decreases as the brine salinity decreases. This indicates that low salinity water
alters the system wettability towards more water wet. There is a slight increase in the contact angle when
changing the brine from formation water (FW) to seawater (SW) which is different from the decrease in
contact angle for brine with lower salinities. Seawater is not considered as a low salinity water since it has
34
been reported in the literature by other workers that the low salinity water effect on improving the oil
recovery is for injected water with salinities below 5,000 ppm (Suman et al. 2013, Morrow and Buckley
2011, McGuire et al. 2005) or below 7,000 ppm as reported by Tang and Morrow (1999).
Contact angle measurement from DSA instrument experiment (Nasralla et al. 2013)
SW FW: Formation Water
FW SW: Seawater
AQ
10% AQ AQ: Aquifer water from
a shallower depth
10% AQ: 10% diluted
aquifer water
Figure 2.13: The measured contact angle between mica/oil/brine at different brine concentrations using a drop shape analysis instrument.
Mica was selected as it represents the clay found within sandstones and to minimize sandstones roughness on quantifying wettability. The
contact angle decreases as the salinity decreases which indicates that low salinity water changes the system wettability towards more water
wet. There is a slight increase in the plot when changing from formation water (FW) to seawater (SW). Seawater is not considered as a low
salinity water since it has been reported in the literature by other researchers that the low salinity water effect on improving the oil recovery
is for injected water with salinities below 5,000 ppm (Suman et al. 2013, Morrow and Buckley 2011, McGuire et al. 2005) or below 7,000
ppm as reported by Tang and Morrow (1999). This figure is modified from Nasralla et al. (2013).
B. Glass Substrate Model Experiment

In this experiment, clay minerals adhered to a glass substrate were chosen as they represent the clays found
within sandstones (Mahani et al. 2014). The objective of this experiment was to investigate the relationship
between the injected brine salinity and the rock wettability. The experimental setup is shown in Figure 2.14.
Oil droplets were attached to this model and the system was initially submerged in a high salinity brine of
25,900 ppm. Then, a low salinity brine was injected and the contact angle between the oil droplets and the
glass surface was monitored continuously using CCD camera and analysed using an imaging software.
Figure 2.14: Schematic of the glass substrate model experiment to investigate the relationship between the injected brine salinity and the
mineral plate wettability. Clay minerals adhered to a glass substrate were used as they represent the clays found within sandstones and
minimize the effect of sandstone roughness on quantifying wettability (modified from Mahani et al. 2014).
Figure 2.15 shows the contact angle measurement during the experiment for three brine concentrations:
1,600 ppm, 3,200 ppm, 6,500 ppm. Initially, two forces acted on the oil droplets in opposite direction: the
adhesion force (downward direction) between the clay and the oil droplets and the buoyancy force (upward
direction) (Ilic 2012). The contact angle between clay/oil/water initially decreased. Once equilibrium
between the two forces was reached, the contact angle stayed constant. Then, low salinity water was added
to the system (LS on Figure 2.15) and the contact angle decreased indicating a change in the system
wettability towards more water wet.
35
Contact angle measurement from the glass substrate model experiment (Mahani et al. 2014)
Brine salinity: 6,500 ppm Brine salinity: 3,200 ppm Brine salinity: 1,600 ppm
Start of low salinity
Start of low salinity
water injection Start of low salinity water injection
water injection
Figure 2.15: Results of contact angle measurement from the glass substrate model experiment by Mahani et al. (2014) to investigate the
relationship between the injected brine salinity and the mineral plate wettability. Mineral plates were used as they represent the clays found
within sandstones and minimize the effect of sandstone roughness on quantifying wettability. Initially, oil droplets were attached to this
model and the system was initially submerged in a high salinity brine and the system were left until equilibrium between the buoyancy and
adhesion forces were reached. Then, low salinity water was added to the system causing the contact angle to decrease indicating a more
water wet system (modified from Mahani et al. 2014).
The laboratory experiments by Nasralla et al. (2013) and Mahani et al. (2014) indicated that low salinity
water changes the rock wettability towards more water wet thus wettability alteration is the primary cause of
the low salinity water effect. However, coreflood experiments by Ashraf et al. (2010) found that oil recovery
increases by low salinity water as the system wettability is changed from water wet to neutral wet.
2.1.4 Evidence Demonstrating the Success of Low Salinity Waterflooding
Several coreflood experiments and field applications have confirmed that low salinity waterflooding
recovers more oil as a first displacement and as a displacement post a high salinity waterflood. In this
section, the oil recoveries obtained by both types of waterflooding are compared for various laboratory
experiments and field scale applications. In this thesis, low salinity water is either injected immediately at
the initiation of oil recovery or post immediate injection of high salinity water.
I. Coreflood Experiments
A typical coreflood experiment setup is shown in Figure 2.16. Table 2.2 lists several coreflood experiments
that were conducted to demonstrate the potential of using low salinity waterflooding to recover more oil as a
first displacement and as a displacement post a high salinity waterflood.
Figure 2.16: A typical coreflood experimental setup to investigate the potential of improving oil recovery by low salinity waterflooding (from Hadia
et al. 2011).
36
Table 2.2: Incremental oil recovery by low salinity waterflooding in various coreflood experiments.
Displacement post a high

First displacement
Experiment by salinity water flood
(Additional oil recovered in %)
(Additional oil recovered)
Yes Yes
Agbalaka et al. (2009) (2-4 % depending on the brine (5-25 % depending on the
composition) brine composition)
Cissokho et al. (2010) Yes (10%) Yes (10%)
Hadia et al. (2011) Not tested Yes (8%)
Zhang and Morrow (2006) Yes (9-15 %) Yes (7-16%)
Hadia et al.(2012) Not tested Yes (1-5%)
Austad et al. (2010) Not tested Yes (5-10%)
Callegaro et al. (2013) Not tested Yes (5-8%)
Robbana et al. (2012) Yes (9-13%) Not tested
Seccombe et al. (2008) Not tested Yes (7%)
Skrettingland et al. (2011) Not tested No
Thyne and Gamage (2011) Not tested No
The above results have shown variations in the additional oil recovery by low salinity waterflooding. These
variations are probably related to the initial wettability of the cores, the composition of the initial brine and
the injected low salinity brine and the presence of divalent ions in both the initial brine and the injected low
salinity brine. Table 2.3 summarizes the researchers’ discussions on the cores’ wettability alteration due to
low salinity water injection. From Table 2.3, it seems that the efficiency of low salinity waterflooding
depends on the initial wetting of the core (Skrettingland et al. 2011, Hadia et al. 2012). The more oil wet or
intermediate wet the core is, the more likely that the low salinity waterflooding will be effective.
Table 2.3: Discussion on the cores’ wettability alteration due to low salinity water injection in various laboratory experiments.
Experiment by Discussions on wettability alteration due to low salinity water injection

Agbalaka et al. (2009) the system wettability changes towards more water wet
Cissokho et al. (2010) did not discuss
Hadia et al. (2011) the system wettability changes towards more water wet
Zhang and Morrow (2006) did not discuss
Hadia et al.(2012) the injection of low salinity water was effective when the core was not initially water wet
Austad et al. (2010) the system wettability changes towards more water wet
Callegaro et al. (2013) the system wettability changes towards more water wet
Robbana et al. (2012) did not discuss
Seccombe et al. (2008) did not discuss
the effectiveness of the injection of low salinity water depends on the initial wetting
Skrettingland et al. (2011)
condition of the system
Thyne and Gamage (2011) did not discuss
Table 2.4 summarizes the salinities of the initial brine and the injected brine used in these experiments in
addition to the presence of divalent ions (Ca++ and Mg++) in both the initial brine and the injected brine. In
some coreflood experiments, low salinity water injection improves the oil recovery without the presence of
divalent ions (Agbalaka et al. 2009, Cissokho et al. 2010, Hadia et al. 2012, Callegaro et al. 2013). This
suggests that the low salinity water effect in improving oil recovery is related to the brine salinity and not
necessarily the presence of divalent ions.
Generally, the incremental oil recovery by low salinity waterflooding increases as the clay content in core
samples increases. The high clay content (9-10 %) in the core samples that were tested by Austad et al.
(2010) contributed in having a high incremental oil recovery (5-10%) by low salinity waterflooding while its
37
low content (< 2 %) in the core samples that were tested by Thyne and Gamage (2011) resulted in having no
incremental oil recovery by low salinity waterflooding. However, the efficiency of low salinity
waterflooding does not always depend on the clay content. Some of the core samples in the coreflooding
experiments by Skrettingland et al. (2011) showed no incremental oil recovery by low salinity waterflooding
although the clay content was high (5-35 %).
Table 2.4: Brine compositions for the initial brine and the injected low salinity brine used in the coreflood experiments to investigate the potential of
improving oil recovery by low salinity waterflooding.
Brine salinity divalent ions (Ca++ and Mg++)

Experiment by Initial brine Injected low salinity Initial Injected low Clay Content in the core
(1,000 ppm) brine (1,000 ppm) brine salinity brine
Agbalaka et al. (2009) 40 Varied: 20,10 No No Not mentioned
Yes, except for
Cissokho et al. (2010) 50 Varied from 0.1 to 10 Yes 9.2 % (no kaolinite)
one scenario
Hadia et al. (2011) 43 Varied: 4, 0.4 Yes Yes From 1.5 to 6 %
Relatively high (mostly
Zhang and Morrow (2006) 38 0.4 Yes Yes
kaolinite)
Hadia et al.(2012) 38 Varied: 0.9, 3.8 No No 6.5% (mostly kaolinite)
Austad et al. (2010) 100 0.7 Yes Yes 9-10 % (mostly illite)
Callegaro et al. (2013) 70 0.7 No No 2-7 % (mostly illite)
Robbana et al. (2012) 34 1.5 Yes Yes 3%
Seccombe et al. (2008) 143 9 Yes Yes 12-25% (mostly kaolinite)
Skrettingland et al. (2011) 34 0.5 Yes Yes 5-35 % (mostly kaolinite)
Thyne and Gamage (2011) 38 3.8 Yes Yes < 2 % (mostly illite)
II. Field Applications

To date, several pilot tests have been conducted on the field scale to evaluate the potential of low salinity
waterflooding as an improved oil recovery method. Table 2.5 lists the outcomes of some of these pilot tests.
Table 2.5: Incremental oil recovery by low salinity waterflooding in various field applications.
Additional oil
Reported by Field/Reservoir Field/Reservoir description recovery by low
salinity waterflooding
• Deposited as braided stream sands
Subzone K3A in the • Sands fairly continuous with some
Seccombe et al. (2010) Endicott formation, Alaska discontinuous shales Yes (10%)
• Average permeability is 800 md
• Lower Rutbah formation: shallow marine and
tidal channel sands
Vledder et al. (2010) Omar, Syria Yes (10-15%)
• Mulussa F formation: fluvial channel sands
and fold plain shales
• Complex package of mainly inhomogeneous
stratified fluvial sandstone
• Heterogeneous nature of reservoir is
Skrettingland et al.
Snorre, North field characterized by limited vertical to lateral No
(2011)
communication
• Reservoir permeability varies from 100-
4,000 md
26 fields within the
Thyne and Gamage Minnelusa formation in the • Quartz-rich sandstone with minor amount of
No
(2011) Powder River Basin, anhydrite and dolomite
Wyoming
• Sandstone dominated succession with minor
siltstones and mudstone
Mahani et al. (2011) Sijan, Syria No
• Tidal channel deposits with permeability up
to 10 Darcy
38
The variations in these field results suggest that the efficiency of low salinity waterflooding depends on
different factors. Table 2.6 summarizes the discussions on possible reasons for not having successful low
salinity waterflooding. It appears that the success of low salinity waterflooding depends on the clay content
and the initial wettability.
Generally, the incremental oil recovery by low salinity waterflooding increases as the clay content in
sandstones increases. The high clay content (20%) in sandstones of the Endicott formation corresponds with
a high incremental oil recovery (10%) by low salinity waterflooding (Seccombe et al. (2010)) while the low
content (< 2 %) in fields within the Minelusa formation resulted in having no incremental oil recovery by
low salinity waterflooding (Thyne and Gamage (2011)). However, the efficiency of low salinity
waterflooding does not always depend on the clay content as field applications of low salinity waterflooding
in the Snorre, North field by Skrettingland et al. (2011) showed no incremental oil recovery although the
clay content was relatively high (5-35%). The initial wetting condition for the Snorre North field was weakly
water wet. This was optimal for conventional seawater injection thus injecting low salinity water did not
improve the oil recovery. On the other hand, injecting low salinity water in the Omar field which was
initially oil wet resulted in having a high incremental oil recovery (10%).
In the Sijan field, there was no incremental recovery by low salinity waterflooding. Low salinity water was
injected in the aquifer zone. It is proposed that mixing between the saline aquifer water and low salinity
water might be the reason for an unsuccessful low salinity waterflood (Mahani et al. 2011).
Table 2.6: Clay content, water salinities and possible reasons for unsuccessful low salinity waterflooding in various field applications.
Wettability alteration Proposed reasons by the authors for

Water salinity
Reported by Clay content due to the injection of unsuccessful low salinity water
(1,000 ppm)
low salinity water injection
Connate water:
Seccombe et al. Not specified
20 % Not discussed Not applicable
(2010) Injected water:
Not specified
The rock was initially an
0.5 to 4 % (of oil wet, the injection of
Vledder et al. Connate water: 90
which 95-100 low salinity water Not applicable
(2010) Injected water: 2.2
% kaolinite) changed the wettability
towards more water wet
The rock was initially a
neutral wet to a weakly
• The rock was initially a neutral
water wet.
wet to a weakly water wet.
The initial wetting
5-35 % • The initial wetting condition is
Skrettingland et Connate water: 34 condition is already
(mostly already optimal for seawater
al. (2011) Injected water: 0.4 optimal for seawater
kaolinite) injection. Injection of low salinity
injection. Injection of low
water leads to a marginal increase
salinity water leads to a
in oil recovery
marginal increase in oil
recovery
Thyne and <2% Connate water: 9 to 136
Not discussed No mobile fines
Gamage (2011) (mostly illite) Injected water: 0.7
There was no additional recovery
from low salinity waterflooding. Low
salinity water was injected in the
Mahani et al. Not Aquifer water: > 200 aquifer zone. It is proposed that
Not discussed
(2011) mentioned Injected water: 0.5 mixing between the saline aquifer
water and low salinity water might be
the reason for an unsuccessful low
salinity waterflood
39
2.2 Mechanisms Limiting Reservoir Scale Recovery
There are various mechanisms that can affect the overall oil recovery on the reservoir scale when injecting
low salinity water. These mechanisms have to be carefully considered and evaluated when implementing
low salinity waterflooding projects on the field scale. These include mixing between the connate water and
the injected low salinity water, the reservoir heterogeneity and the presence of aquifers. In this section, these
mechanisms and their impacts on low salinity waterflooding are discussed.
2.2.1 Mixing
Mixing between the injected low salinity water and in situ high salinity water can affect the efficiency of low
salinity waterflooding (Mahani et al. 2011, Jerauld et al. 2008). Mixing is related to molecular diffusion and
physical dispersion and is enhanced when the flood front becomes distorted by heterogeneity as this
increases the surface area over which diffusion and dispersion can occur. The concept of diffusion and
dispersion in enhanced oil processes such as miscible gas flooding has caught the attention of many
researches in the oil industry, however little attention was given to its impacts on low salinity waterflooding.
When two fluids with different concentrations are in contact, molecules from the fluid with the higher
concentration randomly move (spread) to the fluid with the lower concentration. Later, a mixed diffuse zone
is formed. Finally, concentration is uniformly distributed across them (Perkins and Johnston 1963). The flux
𝐽𝐽 of one of the components is related to the concentration c of that component by Fick’s Law:
𝜕𝜕𝜕𝜕
𝐽𝐽 = − 𝐷𝐷𝑜𝑜 (2.8)
𝜕𝜕𝜕𝜕
where
J is the diffusion flux, mol/L2t
𝐷𝐷𝑜𝑜 is the molecular diffusion coefficient. L2/t
c is the concentration, mol/ L3
x is the position, L
In a porous medium, diffusivity is lower than it would be in a bulk fluid as the random movement of the
molecules is restricted within the pores.
Physical dispersion is the mixing of fluids in a porous medium that results from molecular diffusion and
variations in fluid velocity within and between pores. Their combined effect is given by:
𝐷𝐷𝑜𝑜
𝐾𝐾𝑙𝑙 = + 𝛼𝛼𝑙𝑙 𝑣𝑣 … … … … (2.9)
𝐹𝐹𝜙𝜙
𝐷𝐷𝑜𝑜
𝐾𝐾𝑡𝑡 = + 𝛼𝛼𝑡𝑡 𝑣𝑣 … … … … (2.10)
𝐹𝐹𝜙𝜙
where
Do is the molecular diffusion coefficient, L2/t
Kl is the longitudinal dispersion coefficient (in the direction of flow), L2/t
Kt is the transverse dispersion coefficient (perpendicular to the direction of flow), L2/t
40
F is the formation resistivity, (dimensionless), assumed to be 1
𝜙𝜙 is the porosity, (dimensionless)
v is the flow velocity in the porous medium, L/t
𝛼𝛼𝑙𝑙 and 𝛼𝛼𝑡𝑡 are the mechanical longitudinal and transverse dispersivities, L
At the reservoir scale, diffusion is negligible compared to dispersion (Lake and Hirasaki 1981).
When a soluble substance is added to a solvent that is moving slowly in a tube, the concentration of the
soluble substance depends on the balance between the variations in the velocity through the cross sectional
area and the molecular diffusion (Taylor 1953-1954). Taylor (1953-1954) showed analytically that
dispersion along the tube is controlled by increased longitudinal diffusion. This is known as Taylor
dispersion.
Based on Taylor’s work on dispersion in a tube, Lake and Hiraskai (1981) showed that transverse dispersion
in combination with layering in a porous medium can result in increased longitudinal dispersion. They
proposed that for a two layer porous medium the importance of advection to dispersion could be quantified
by a transverse dispersion number (NTD):
𝐿𝐿 𝐾𝐾𝑡𝑡
𝑁𝑁𝑁𝑁𝑁𝑁 = … … … … … (2.11)
𝐻𝐻 2 𝑣𝑣1
where
𝑣𝑣1 is the pore velocity of the displacing fluid in the high permeability layer, L/t
H is the medium thickness, L
L is the medium length, L
Figure 2.17 shows three different flow regimes based on the value of NTD. At a very small NTD, transverse
dispersion is negligible so the produced concentration profile depends on the permeability variations
between the layers and the longitudinal dispersion of each layer (NTD = 0 in Figure 2.17a). In this case, the
model behaves as a two layer system.
At a moderate NTD, transverse dispersion causes the fluids to move between layers thus distorting the
isoconentration lines. The resulting concentration profile is affected by both heterogeneity and transverse
dispersion (NTD moderate in Figure 2.17b).
At a large NTD, transverse dispersion causes the isoconcentration lines to be vertical across the system’s
thickness resulting in having an effluent concentration profile that shows effective longitudinal dispersion
that is larger compared to that seen in the profile shown in Figure 2.17a. This case shows Taylor dispersion.
The model behaves as a single layer system with increased longitudinal dispersion (NTD large in Figure
2.18c). In this case, the size of the area with intermediate concentration is large (since longitudinal
dispersion is increased) compared with the previous two cases. As a result, the effectiveness of the injected
fluid with lower concentration will be reduced. For example, in chemical flooding when injecting the
chemicals in slugs followed by a chase fluid, mixing would result in diluting the effect of the chemical slug
thus reducing the overall oil recovery (Wheat and Dawe 1988). A simulation study by Sharafi and
41
Jamialahmadi (2017) showed that dispersion also affects the displacement efficiency of viscoelastic polymer
flooding which results in lowering the oil recovery.
a)
b)
c)
Figure 2.17: Different flow regimes based on NTD values At a very small NTD, transverse dispersion is negligible so the produced concentration
profile depends on the permeability variations between the layers and the longitudinal dispersion of each layer. The model behaves as a two layer
system. At a moderate NTD, transverse dispersion causes the fluids to move between layers thus distorting the isoconentration lines. The resulting
concentration profile is affected by both heterogeneity and transverse dispersion. At a large NTD, transverse dispersion causes the isoconcentration
lines to be vertical across the system’s thickness resulting in having an effluent concentration profile that shows effective longitudinal dispersion
that is larger compared with the isoconcentration lines from the longitudinal dispersion of each layer. This is Taylor dispersion. The model
behaves as a single layer system with increased longitudinal dispersion (from Lake and Hirasaki 1981).
At the pore scale, small scale heterogeneity causes additional mixing in miscible flooding between the
displacing fluid and the displaced fluid due to the disturbed front between them (Perkins and Johnston 1963,
Alkindi et al. 2011, Mahadevan et al. 2003). However, Coats et al. (2009) argue that heterogeneity alone
does not causes mixing and that often the apparent dispersion seen in production profiles is purely a result of
commingling of fluids from different layers in the wellbore rather than actual mixing in the reservoir.
It is anticipated that the value of longitudinal dispersion increases with length scale. Mahadevan et al. (2003)
presented results from tracer displacements observed on a range of scales up to 100 kilometres as shown in
Figure 2.18. They attributed this to the Taylor dispersion resulting from heterogeneity on a range of length
scales. There is an uncertainty in the level of physical dispersion on the field scale. For example, the
commingling of fluids in the wellbore is called an apparent mixing and it does not necessarily mean that
physical mixing actually occurs within the pore space.
42
Figure 2.18: Best fit match of field and laboratory measured longitudinal dispersion. The plot suggests that physical longitudinal dispersivity
increases with length scale (from Mahadevan et al. 2003).
Mixing can be modelled either by explicitly modelling the physical dispersion in the model as an input
parameter or assuming that the physical dispersion can be approximated by the numerical dispersion by the
simulator (Jerauld et al. 2008).
Jerauld et al. (2008) conducted a simulation study on the injection of low salinity water as slugs in a quarter
five spot using a fine grid 2D model (162 × 162) and a coarse grid 2D model (25 × 25).
The model dimensions were 702 m × 732 m. Low salinity water was injected as slugs following the injection
of 0.4 PV of conventional water. For the fine grid model, physical dispersivity was set as 12.42 m for the
longitudinal dispersivity and 0.22 m for the transverse dispersivity. The longitudinal dispersivity is about 2%
of the model’s length.
For the coarse grid model, physical dispersion was not modelled, so the flow was dominated by numerical
dispersion. The numerical longitudinal dispersivity is estimated by:
1
𝛼𝛼𝑙𝑙 (𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛) = (2.12)
2 𝑁𝑁
where:
𝛼𝛼𝑙𝑙 (𝑛𝑛𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢) is the numerical longitudinal dispersivity, (dimensionless)
N is the number of horizontal grid blocks
By substituting N to be 25, the numerical longitudinal dispersivity becomes:
1 1
𝛼𝛼𝑙𝑙 (𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛) =
= =2%
2 𝑁𝑁 2 × 25
In this case the numerical dispersion in the coarse grid model is equal to the physical dispersion in the fine
grid model. Figure 2.19 shows the plot of the incremental oil recovery versus the total pore volume injection
for both models. The predictions by both models are very close which confirm that numerical dispersion in
the coarse grid model can approximate the physical dispersion.
43
Figure 2.19: Results of a simulation study by Jerauld et al. (2008) proposing that numerical dispersion can approximate physical dispersion in a
2D quarter five spot injection of low salinity water as slugs following the injection of 0.4 PV of conventional water. The dashed lines are the
results for the coarse grid model (25 × 25) while the solid lines are the results for the fine grid model (162 × 162). The predictions by both models
are very close which confirm that numerical dispersion in the coarse grid model can approximate the physical dispersion (from Jerauld et al.
2008).
While Mahadevan et al. (2003) suggested that physical dispersivity (𝛼𝛼) increases with length scale (as
shown previously in Figure 2.19), Coats et al. (2009) argue that physical dispersivity (𝛼𝛼) is scale
independent and it is always a constant value of 0.003 m. They suggested that dispersion is the sum of the
apparent dispersivity (𝛼𝛼𝑎𝑎 ) (due to conformance or heterogeneity) and the physical dispersivity (𝛼𝛼) (which is
always 0.003 m regardless of the length scale). Their approach was to match the actual concentration from a
field tracer with the 1D convection/diffusion equation of a miscible displacement:
⎛ ⎞ ⎛ ⎞
1 1 − 𝑄𝑄𝐷𝐷 𝑒𝑒 𝑁𝑁𝑃𝑃𝑒𝑒𝑎𝑎 1 + 𝑄𝑄𝐷𝐷
𝐶𝐶 = 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 ⎜ ⎟+ 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 ⎜ ⎟ … … . . (2.13)
2 ⎜ 2
𝑄𝑄𝐷𝐷 ⎟ ⎜ 𝑄𝑄 ⎟
2 �𝑁𝑁 2 �𝑁𝑁 𝐷𝐷
𝑃𝑃𝑒𝑒𝑎𝑎 𝑃𝑃𝑒𝑒𝑎𝑎
⎝ ⎠ ⎝ ⎠
where:
C is the concentration (fraction of 1)
𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 (𝑥𝑥) = 1 − 𝑒𝑒𝑒𝑒𝑒𝑒(𝑥𝑥)
𝑄𝑄𝐷𝐷 is the dimensionless pore volume injected
𝑁𝑁𝑃𝑃𝑒𝑒𝑎𝑎 is the apparent Peclet number (dimensionless)
In stratified reservoirs, variations in the permeability affect the effluent concentration profiles. The analytical
solution of the effluent concentration profiles is given by Muskat method (Muskat 1949) as follows:
𝐻𝐻
∫ 𝐹𝐹(𝑄𝑄𝐷𝐷𝐷𝐷 )𝑘𝑘(𝑧𝑧) 𝑑𝑑𝑑𝑑
𝐶𝐶(𝑄𝑄𝐷𝐷 ) = 0 (2.14)
𝑘𝑘� 𝐻𝐻
where
𝑘𝑘(𝑧𝑧)
𝑄𝑄𝐷𝐷𝐷𝐷 = 𝑄𝑄 (2.15)
𝑘𝑘 𝐷𝐷
𝐻𝐻
∫ 𝑘𝑘(𝑧𝑧) 𝑑𝑑𝑑𝑑
𝑘𝑘� = 0 (2.16)
𝐻𝐻
where
𝑄𝑄𝐷𝐷𝐷𝐷 is the pore volume injected into a given layer z (dimensionless)
44
C(𝑄𝑄𝐷𝐷𝐷𝐷 ) is the concentration (fraction of 1)
𝐹𝐹(𝑄𝑄𝐷𝐷𝐷𝐷 ) is any function that defines the concentration profiles in each layer
𝑘𝑘(𝑧𝑧) is permeability as a function of layers, md
H is the reservoir thickness, L
If the input physical dispersivity (𝛼𝛼) is zero, 𝐹𝐹(𝑄𝑄𝐷𝐷𝐷𝐷 ) is a step function from 0 to 1 at 𝑄𝑄𝐷𝐷𝐷𝐷 = 1. However, if
the input physical dispersivity (𝛼𝛼) is nonzero, 𝐹𝐹(𝑄𝑄𝐷𝐷𝐷𝐷 ) is represented by the 1D convection/diffusion
equation that was defined previously.
Coats et al. (2009) used a transmission test that was taken from Mahadevan et al. (2003) paper. This test was
modelled using a five layer model with dimensions of 9.14 × 6.09 m that was discretised into 30 × 5 grid
blocks. The permeabilities in the z-direction were 200, 500, 800, 1,100 and 1,400 md. The physical
dispersivity (𝛼𝛼) in the simulation was 0.46 m. The UTCHEM simulator was used.
The simulation output using 𝛼𝛼 = 0.46 𝑚𝑚, the 1D C/D equation with apparent dispersivity of 𝛼𝛼𝑎𝑎 = 0.46 𝑚𝑚
and the Muskat analytical solution for two cases: 𝛼𝛼 = 0.46 𝑚𝑚 and 𝛼𝛼 = 0 are plotted against the injected
fluid volume until a total injection of 1 PV in Figure 2.20.
There is an exact match between the UTCHEM simulation results (blue circles) and the Muskat analytical
solution (using 𝛼𝛼𝑎𝑎 to be 0.46 m) that reflected conformance (blue line). In addition, the UTCHEM
simulation results matches the 1D C/D equation using 𝛼𝛼𝑎𝑎 = 0.46 𝑚𝑚 (red line).
This figure clearly shows that conformance (or stratification) (when using 𝛼𝛼𝑎𝑎 to be 0.46 m) affects the
effluent concentration profile to look or appear like there is a physical mixing (known as an apparent
mixing) (from Coats et al. 2009).
Figure 2.20: Comparison between the UTCHEM simulation output (of 𝛼𝛼 = 0.46 𝑚𝑚) (blue circles), Muskat analytical solution for the effect of
stratification (or conformance) on the effluent concentration profiles (using input dispersivities of 0 and 0.46 m (blue and grey lines respectively )
and the 1D C/D equation using a fitting parameter of the apparent dispersivity 𝛼𝛼𝑎𝑎 to 0.46 𝑚𝑚 (red line). There is an exact match between the
UTCHEM simulation results (blue circles) and the Muskat analytical solution (using 𝛼𝛼𝑎𝑎 = 0.46 m) that reflected conformance (blue line). In addition,
the UTCHEM simulation results matches the 1D C/D equation using 𝛼𝛼𝑎𝑎 = 0.46 𝑚𝑚 (red line). This figure clearly shows that conformance (or
stratification) (when using 𝛼𝛼𝑎𝑎 = 0.46 m) affects the effluent concentration profile to look or appear like there is a physical mixing (known as an
apparent mixing) (from Coats et al. 2009).
45
2.2.2 Heterogeneity
In this section, heterogeneity effects on limiting the overall reservoir scale recovery are discussed.
Heterogeneity is a major factor to consider when planning any water injection project. Heterogeneity effects
on conventional waterflooding have been extensively studied and they are well understood (Osman and Tiab
1981, Sorbie et al. 1987, Ahmed et al. 1988, Cinar et al. 2004). However, little attention has been given to
their effects on low salinity waterflooding. They tend to reduce the macroscopic sweep efficiency especially
if the flow is viscosity dominated (large injection rates with large permeability contrast) (Coll et al. 2000,
Baker 1998).
The modelling of the impact of heterogeneity on waterflooding usually requires very fine grid blocks
because geological heterogeneity occurs at all levels from the pore (micro-scale) level to the depositional
system (mega-scale) level as illustrated in Figure 2.21 (Jones et al. 1995, Henson et al. 2002). At large
scales, the contrast in permeability between layers and the vertical communication between them can cause
viscous cross flow which can affect the overall sweep efficiency (Kjonsvik et al. 1994). At small scales,
laminations cause trapping of oil (Jones et al. 1995).
Figure 2.21: Scales of geological heterogeneity (Henson et al. 2002).
Before discussing heterogeneity effects on low salinity waterflooding and reviewing its effects on other
displacement processes, a critical review of the various methods of quantifying heterogeneity is provided.
The best calculation method is chosen for quantifying heterogeneity when investigating its effects on low
salinity waterflooding later on this thesis.
46
It is anticipated that low salinity waterflooding is influenced by reservoir heterogeneity as it involves a
similar gravity contrast between fluids to conventional waterflooding. A complication to low salinity
waterflooding compared with conventional waterflooding is that mixing occurs between the injected low
salinity water and the connate water (similar to chemical flooding). Mixing might reduce the effect of low
salinity waterflooding by forming water with intermediate salinities. These intermediate salinities do not
alter the rock’s wettability similar to the low salinity water (Jerauld et al. 2008). Therefore, low salinity
waterflooding becomes less effective. One of the purposes of this thesis is to evaluate the impact of
heterogeneity on the performance of low salinity waterflooding and compare it to conventional
waterflooding.
2.2.2 I Quantifying Heterogeneity
There are several ways to quantify heterogeneity in a given reservoir. They are divided mainly into static
indices and dynamic indices.
Static Indices
Static indices estimate reservoir heterogeneity without modelling any flow in the reservoir. The Dykstra and
Parsons coefficient VDP (Dykstra and Parsons 1950) was one of the earliest coefficients to quantify reservoir
heterogeneity. The permeability data set is sorted in an increasing order of permeability and VDP is calculated
according to the following equation:
𝑘𝑘50 − 𝑘𝑘16
𝑉𝑉𝐷𝐷𝐷𝐷 = (2.17)
𝑘𝑘50
where:
VDP is the Dykstra and Parsons coefficient (dimensionless)
𝑘𝑘50 is the mean permeability and 𝑘𝑘16 is the 16th percentile permeability (both in md)
For homogeneous reservoirs, 𝑉𝑉𝐷𝐷𝐷𝐷 is zero while 𝑉𝑉𝐷𝐷𝐷𝐷 is 1 in extremely heterogeneous reservoirs. It was
observed by Jensen and Lake (1988) that slight variations in the estimation of 𝑉𝑉𝐷𝐷𝐷𝐷 when it is more than 0.7
cause significant changes in the predicted oil recovery. Therefore, an improved method by Jensen and Currie
(1990) was proposed to estimate 𝑉𝑉𝐷𝐷𝐷𝐷 as per the following equation:
𝑉𝑉𝐷𝐷𝐷𝐷 = 1 − 𝑒𝑒 −𝜎𝜎𝑘𝑘 (2.18)
𝑛𝑛 𝑛𝑛 2
0.25 1 1
𝜎𝜎𝑘𝑘 = �1 + �� 𝑙𝑙𝑙𝑙 𝑘𝑘𝑖𝑖 − � � 𝑙𝑙𝑙𝑙 𝑘𝑘𝑖𝑖 � � (2.19)
𝑛𝑛 − 1 𝑛𝑛 − 1 𝑛𝑛
𝑖𝑖=1 𝑖𝑖=1
where:
𝑉𝑉𝐷𝐷𝐷𝐷 is the improved Dykstra Parsons coefficient (dimensionless)
𝜎𝜎𝑘𝑘 is approximately the standard deviation of 𝑙𝑙𝑙𝑙 𝑘𝑘 (dimensionless)
n is the number of permeability samples
Schmalz and Rhame (1950) developed another coefficient to quantify heterogeneity in reservoirs. They
called it the Lorenz coefficient (𝐿𝐿𝑐𝑐 ). It can be calculated from the plot of flow capacity (Fm) versus the
cumulative thickness (Hm), where Fm and Hm are defined by:
47
∑𝑖𝑖=𝑚𝑚
𝑖𝑖=1 𝑘𝑘(𝑖𝑖) . ℎ(𝑖𝑖)
𝐹𝐹𝑚𝑚 = (2.20)
∑𝑖𝑖=𝑛𝑛
𝑖𝑖=1 𝑘𝑘(𝑖𝑖) . ℎ(𝑖𝑖)
∑𝑖𝑖=𝑚𝑚
𝑖𝑖=1 ℎ(𝑖𝑖)
𝐻𝐻𝑚𝑚 = (2.21)
∑𝑖𝑖=𝑛𝑛
𝑖𝑖=1 ℎ(𝑖𝑖)
The plotted points are arranged in a decreasing value of permeability (k). Finally, the Lorenz coefficient 𝐿𝐿𝑐𝑐 is
the shaded area shown in Figure 2.22. For a homogeneous reservoir, Lc = 0 while it approaches 1 for highly
heterogeneous reservoirs.
Figure 2.22: An illustration showing a typical curve for plotting the flow capacity (Fm) versus the cumulative thickness (Hm) to quantify reservoir
heterogeneity using the Lorenz coefficient (which is the shaded area shown in this plot) (from Lake and Jensen 1989).
In an immiscible displacement such as conventional waterflooding, the injected water often flows
preferentially through the high permeability streaks within the reservoir (channelling). Static indices do not
consider such effects when quantifying heterogeneity. However, dynamic indices which involve running a
simulation, capture such effects (Lake and Jensen 1989).
Dynamic Indices
Koval (1963) developed an empirical equation to relate reservoir heterogeneity and viscous fingering in
miscible displacement processes as follows:
𝑘𝑘 = 𝐻𝐻 𝐸𝐸 (2.22)
4
𝜇𝜇𝑜𝑜 0.25
𝐸𝐸 = �0.78 + 0.22 � � � (2.23)
𝜇𝜇𝑠𝑠
where
H is the heterogeneity factor (dimensionless)
𝜇𝜇𝑜𝑜 is oil viscosity and 𝜇𝜇𝑠𝑠 is the solvent viscosity (cp)
H is equal to 1 in homogeneous systems, however it is greater than 1 for heterogeneous systems.
Essentially H is a factor that increases the mobility ratio. k is related to the dimensionless breakthrough time
𝑡𝑡𝑑𝑑 by:
48
1
𝑘𝑘 = (2.24)
𝑡𝑡𝑑𝑑
Shook and Mitchell (2009) developed another coefficient to quantify the heterogeneity in reservoirs. They
called it The Dynamic Lorenz Coefficient Lc. Streamline simulation is used in developing this coefficient.
The first step is to solve for pressure in a grid block. Then, streamlines are constructed perpendicular to the
pressure contours. These streamlines describe the fluid flow in the grid block. The fluid flow equations are
then solved based on these streamlines (Idrobo et al. 2000, Thiele et al. 2010). The streamline simulator is
allowed to run for several time steps until steady state is reached. Once steady state is reached, the
volumetric flow rates (𝑞𝑞𝑖𝑖 ) through each streamline and the time it takes the fluid to travel through these
streamlines (known as time of flight (𝜏𝜏𝑖𝑖 )) are used to calculate the volume of each streamline as per the
following equation:
𝑉𝑉𝑝𝑝𝑝𝑝 = 𝑞𝑞𝑖𝑖 𝜏𝜏𝑖𝑖 (2.25)
where:
𝑉𝑉𝑝𝑝𝑝𝑝 is the volume of streamline i (L3)
𝑞𝑞𝑖𝑖 is the volumetric flow rate (L3/t)
𝜏𝜏𝑖𝑖 is the time of flight (t)
After that, the streamlines are sorted in decreasing values of 𝜏𝜏𝑖𝑖 .
Then, F and 𝛷𝛷 are calculated as follows:
∑𝑖𝑖𝑗𝑗=1 𝑞𝑞𝑗𝑗 ∑𝑖𝑖𝑗𝑗=1 𝑉𝑉𝑝𝑝𝑝𝑝
𝐹𝐹𝑖𝑖 = 𝑎𝑎𝑎𝑎𝑎𝑎 Φ𝑖𝑖 = (2.26)
∑𝑁𝑁
𝑗𝑗=1 𝑞𝑞𝑗𝑗 ∑𝑁𝑁
𝑗𝑗=1 𝑉𝑉𝑝𝑝𝑝𝑝
where:
i refers to any layer
N is the number of layers
𝐹𝐹𝑖𝑖 is the flow capacity of layer i (dimensionless)
Φ𝑖𝑖 is the storage capacity of layer i (dimensionless)
F is plotted against 𝛷𝛷 and thus the Dynamic Lorenz Coefficient Lc is calculated by:
1
𝐿𝐿𝑐𝑐 = 2 �� 𝐹𝐹𝐹𝐹Φ − 0.5� (2.27)
0
Figure 2.23 shows a typical F vs. 𝛷𝛷 for 3 different heterogeneous reservoirs. The plot is a straight line for
the homogeneous case and Lc is zero. For a highly heterogeneous system, Lc approaches 1 and most of the
flow comes from a very small portion of the reservoir’s pore volume.
49
Figure 2.23: A typical F vs. 𝛷𝛷 plot for three different heterogeneity cases. For a homogeneous case, the plot is a straight line and the dynamic
Lorenz coefficient (Lc) is zero. For a highly heterogeneous case, Lc approaches 1 and most of the flow comes from a very small portion of the
reservoir’s pore volume (modified from Shook and Mitchell 2009).
A more recent method of quantifying heterogeneity was developed by Rashid et al. (2012). This method uses
a pressure solver to calculate the vorticity (Ω) of the displacement velocity (𝜔𝜔) for different geological
heterogeneities. The vorticity is a measure of the local rotation of a fluid within its velocity field (Rashid et
al. 2012). Darcy’s law for describing flow in a porous media is as follows: (neglecting gravity and capillary
effects):
𝑘𝑘
𝑣𝑣 = − ∇𝑃𝑃 (2.28)
𝜇𝜇
where:
v is the flux per unit cross sectional area, L/t
k is the permeability tensor, (md)
𝜇𝜇 is the dynamic viscosity of the fluid, (cp)
𝑃𝑃 is the scalar pressure field (psi)
In a 2D (x-y plane) heterogeneous porous media with anisotropic permeabilities (i.e.𝑘𝑘𝑥𝑥 ≠ 𝑘𝑘𝑦𝑦 ), the vorticity
is given by the curl of the Darcy velocity:
1 𝜕𝜕𝑘𝑘𝑦𝑦 𝜕𝜕𝜕𝜕 𝜕𝜕𝑘𝑘𝑥𝑥 𝜕𝜕𝜕𝜕 𝜕𝜕 2 𝑃𝑃
(∇ × 𝑣𝑣)𝑧𝑧 = − �� − �+ �𝑘𝑘 − 𝑘𝑘𝑥𝑥 �� =
𝜇𝜇 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝑦𝑦
1 𝜕𝜕𝜕𝜕𝜕𝜕𝑘𝑘𝑦𝑦 𝜕𝜕𝑙𝑙𝑛𝑛𝑘𝑘𝑥𝑥 𝜕𝜕 2 𝑃𝑃
= − �� − 𝑣𝑣𝑥𝑥 �+ �𝑘𝑘 − 𝑘𝑘𝑥𝑥 �� (2.29)
𝜇𝜇 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝑦𝑦
In a 2D Cartesian coordinate system, the changes in the velocity field can be related to the disturbance in the
interface between the displacing fluid and the displaced fluid as follows. The assumption here is that the total
velocity does not change with time.
50
Darcy’s velocity vector at any location 𝑟𝑟 (𝑥𝑥0 , 𝑦𝑦0 ) is given by:
𝑣𝑣 (𝑥𝑥, 𝑦𝑦) = 𝑣𝑣0 (𝑥𝑥0 , 𝑦𝑦0 ) + 𝐽𝐽. 𝛿𝛿𝛿𝛿 (2.230)
where:
v is the velocity vector, L/t
𝛿𝛿𝛿𝛿 is a small distance away from the original location of 𝑟𝑟 (𝑥𝑥0 , 𝑦𝑦0 ), L
𝜕𝜕𝑢𝑢0 𝜕𝜕𝑢𝑢0
⎛ 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 ⎞
𝐽𝐽 = ⎜ (2.31)
𝜕𝜕𝑣𝑣0 𝜕𝜕𝑣𝑣0 ⎟
⎝ 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 ⎠
where:
𝑢𝑢0 is the velocity in the x-direction (L/t) and 𝑣𝑣0 is the velocity in y-direction (L/t)
In single phase flow, velocity will change due to reservoir heterogeneity. By decomposing 𝐽𝐽 into symmetric
and antisymmetric components, equation 2.30 becomes:
𝜕𝜕𝑢𝑢0 𝜕𝜕𝑢𝑢0 𝜕𝜕𝑣𝑣0 𝜕𝜕𝑢𝑢0 𝜕𝜕𝑣𝑣0
2 + 0 −
1⎛ 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 ⎞ 1⎛ 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 ⎞
𝑣𝑣 (𝑥𝑥, 𝑦𝑦) = 𝑣𝑣0 (𝑥𝑥0 , 𝑦𝑦0 ) + ⎜ 𝛿𝛿𝛿𝛿 + ⎜ ⎟ 𝛿𝛿𝛿𝛿 (2.32)
2 𝜕𝜕𝑢𝑢 0 𝜕𝜕𝑣𝑣0 𝜕𝜕𝑣𝑣0 ⎟ 2 𝜕𝜕𝑣𝑣0 𝜕𝜕𝑢𝑢0
+ 2 − 0
⎝ 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 ⎠ ⎝ 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 ⎠
The second (middle) term in the equation is called the rate of strain tensor (𝛾𝛾̇ ) and the last term (right) in the
equation is called the vorticity (Ω). Vorticity (Ω) is defined by:
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
Ω= − (2.33)
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
Darcy’s law for a two phase flow with the inclusion of gravity term is given by:
𝒗𝒗𝑻𝑻 = −𝑘𝑘𝜆𝜆 𝑇𝑇 [∇𝑃𝑃] (2.34)
where:
𝒗𝒗𝑻𝑻 is the total velocity (𝑢𝑢 𝑇𝑇 , 𝑣𝑣𝑇𝑇 ), (L/t)
𝜆𝜆 𝑇𝑇 is the total mobility (dimensionless)
By taking the curl (rotation) of the total velocity and combining equations 2.33 and 2.34 with the assumption
that the fluid mobility depends on the saturation of the displaced fluid only, the vorticity can be defined by:
𝜕𝜕𝜕𝜕𝜕𝜕 𝑘𝑘 𝜕𝜕𝜕𝜕𝜕𝜕 𝑘𝑘
Ω=� 𝑣𝑣𝑇𝑇 − 𝑢𝑢 � (2.35)
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝑇𝑇
Using a numerical simulator, the vorticity of each grid block can be calculated using a single phase pressure
solver.
In a heterogeneous reservoir, the heterogeneity can be quantified by the following equation:
1 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 (Ω)
𝐻𝐻𝑣𝑣 = 𝐶𝐶𝑣𝑣 = (2.36)
𝐶𝐶𝑣𝑣 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 (Ω)
where:
𝐻𝐻𝑣𝑣 is the vorticity heterogeneity index (dimensionless)
Ω is the vorticity, (1/t)
51
𝐶𝐶𝑣𝑣 is the coefficient of variation (dimensionless)
As the reservoir’s heterogeneity increases, Hv approaches zero, while it approaches infinity when it is a
homogeneous system.
Rashid et al. (2012) used 85 areal layers taken from SPE 10 model 2 (Christie and Blunt 2001) to evaluate
heterogeneity characterization in a miscible flood. The breakthrough time of the injected fluid in each layer
is plotted against three different measures of heterogeneity in Figure 2.24.
Figure 2.24 shows that unlike the Dynamic Lorenz coefficient and Dykstra Parsons coefficient, the vorticity
heterogeneity index correlates very well with the breakthrough time especially for very heterogeneous
reservoirs (labelled as Upper Ness in Figure 2.24).
In this thesis, the vorticity heterogeneity index is used to quantify heterogeneity for complex and geological
heterogeneous reservoirs.
Figure 2.24: The breakthrough time of the injected fluid in a miscible flood from each of the areal layers taken from SPE 10 model 2 (Christie and
Blunt 2001). Unlike the Dynamic Lorenz coefficient and Dykstra Parsons coefficient, the vorticity heterogeneity index correlates very well with
the breakthrough time especially for very heterogeneous reservoirs (labelled as Upper Ness). Figure is from Rashid et al. (2012).
2.2.2 II Heterogeneity Effects on Various Displacement Processes:

Effect on Vertical and Areal Sweep
Heterogeneity affects the vertical and areal sweep efficiency. In conventional waterflooding, the
permeability contrast in layered reservoirs has a significant impact on the vertical sweep efficiency. The flow
regime is controlled by three different effects: viscous, gravity and capillary. The flow is considered as
capillary dominated when the injected water prefers to flow through the lower permeability layers leaving oil
unswept in the higher ones. Fluids are segregated in the case of gravity dominated flow in the thick and high
permeability layers (Figure 2.25). Viscous dominated flow occurs when the injected water prefers to flow in
the higher permeability layers leaving oil unswept in the low permeability layers (Coll et al. 2000). The
dominant flow regime can be determined using the gravity-viscous number and the capillary viscous number
(Baker 1998). At high injection rates (i.e. when viscous effects dominate), the injected water sweeps the high
permeability layers leading to a poor vertical sweep efficiency (Baker 1998, El-Khatib 1985). Similarly,
heterogeneity affects the areal sweep efficiency as channelling may occur in fluvial channel sands through
high permeability areas leading to huge amount of bypassed oil in the low permeability areas. As a result, a
poor areal sweep is observed (Baker 1998) as shown in Figure 2.26.
Since low salinity waterflooding has a very similar mobility to conventional waterflooding, it is anticipated
that the effect of heterogeneity on its sweep is similar to conventional waterflooding. However, mixing of
the injected low salinity water in a high permeability layer by the connate water from a low permeability
52
layer might affect both the vertical sweep efficiency and the microscopic displacement efficiency. This is
further investigated later in this thesis.
Figure 2.25: Different types of flow regimes in conventional waterflooding Figure 2.26: Channelling in fluvial channel sands through high
of layered reservoirs. In gravity dominated flow, the permeability contrast permeability areas leads to a huge amount of bypassed oil in the low
is minor and the water is much denser than oil leading the water to slump permeability areas thus a poor areal sweep is observed (from Baker
to the bottom layers while in a viscous dominated flow, the permeability 1998).
contrast effect on the flow is increased when the mobility ratio is high
which causes the water to preferentially sweep the high permeability layers
(from Baker 1998).
Cross Flow and Mixing Between Layers

While little research has been conducted on the effect of cross flow and mixing between layers on low
salinity waterflooding, extensive research has been conducted on conventional waterflooding (Osman and
Tiab 1981, Zapata and Lake 1981, Goddin et al. 1966, Ahmed et al. 1988) and polymer flooding (Wright et
al. 1987, Zapata and Lake 1981, Sorbie et al. 1987).
Cross flow affects the displacement of oil by the injected high salinity water depending on the balance
between capillary, viscous and gravity forces (Cinar et al. 2004, Haq and Reis 1993). The capillary-viscous
number is used to determine if the flow is dominated by capillary forces as per the following equation
(Shook et al. 1992, Coll et al. 2000, Rashid 2012):
𝜆𝜆0 𝑟𝑟𝑟𝑟 𝜎𝜎�𝐾𝐾𝑥𝑥 𝜙𝜙
𝑁𝑁𝑝𝑝𝑝𝑝𝑝𝑝 = (2.37)
𝐿𝐿 𝑈𝑈𝑡𝑡
where:
𝑁𝑁𝑝𝑝𝑝𝑝𝑝𝑝 is the capillary-viscous number (dimensionless)
𝜆𝜆0 𝑟𝑟𝑟𝑟 is the mobility of water at residual oil saturation (dimensionless)
𝜎𝜎 is the interfacial tension between oil and water (N/m)
𝐾𝐾𝑥𝑥 is absolute permeability in the horizontal direction (m2)
𝜙𝜙 is the porosity (dimensionless)
𝐿𝐿 is the reservoir length (m)
𝑈𝑈𝑡𝑡 is the total fluid velocity (m/s)
Capillary effects are negligible when 𝑁𝑁𝑝𝑝𝑝𝑝𝑝𝑝 < 0.1 while it is considered as significant when 𝑁𝑁𝑝𝑝𝑝𝑝𝑝𝑝 > 5 (Shook
et al. 1992, Rashid 2012). In a capillary dominated waterflood displacements, capillary cross flow causes the
53
injected water to flow from the high permeability layer to the adjacent low permeability layer thus reducing
the areal sweep efficiency of the high permeability layer (Haq and Reis 1993).
Another dimensionless number (𝑁𝑁𝑔𝑔𝑔𝑔 ) is used to determine if the water displacement is a viscous dominated
flow or a gravity dominated flow as per the following equation (Shook et al. 1992, Coll et al. 2000, Rashid
2012):
𝐾𝐾𝑥𝑥 𝜆𝜆0 𝑟𝑟𝑟𝑟 ∆𝜌𝜌 𝑔𝑔 cos(𝛼𝛼) 𝐻𝐻
𝑁𝑁𝑔𝑔𝑣𝑣 = (2.38)
𝑈𝑈𝑡𝑡 𝐿𝐿
where:
𝑁𝑁𝑔𝑔𝑔𝑔 is the gravity-viscous number (dimensionless)
𝐾𝐾𝑥𝑥 is absolute permeability in the horizontal direction (m2)
𝜆𝜆0 𝑟𝑟𝑟𝑟 is the mobility of water at residual oil saturation (dimensionless)
∆𝜌𝜌 is the density difference between water and oil (kg/m3)
𝑔𝑔 is the acceleration due to gravity (m/s2)
𝛼𝛼 is the angle of dip for the reservoir (degrees)
𝐻𝐻 is the thickness of the reservoir (m)
𝑈𝑈𝑡𝑡 is the total fluid velocity (m/s)
𝐿𝐿 is the reservoir length (m)
If 𝑁𝑁𝑔𝑔𝑔𝑔 < 0.1, the water displacement is viscous dominated (Shook et al. 1992, Rashid 2012) while when it is
greater than 10 it is gravity dominated (Baker 1998). When the injection rate is high, cross flow has minor
effects on the oil recovery as the injected water preferentially sweeps the high permeability layer leading to
earlier water breakthrough and leaving a bypassed oil in the low permeability layers (El-Khatib 1985).
In chemical flooding, the magnitude of cross flow largely depends on the slug size of the injected chemicals
and the mobility ratios (Wright et al. 1987, Zapata and Lake 1981). When the slug has a higher mobility
compared to the chase fluid mobility, most of the high permeability layers are swept and the effect of
crossflow is negligible. However, a lower mobility slug compared to the chase fluid mobility promotes
crossflow from the higher permeability layers to the lower permeability layers (Wright et al. 1987).
Presence of Waterfilled Shales
In some cases, shales fully saturated with connate water are present within sands in reservoirs. These have to
be carefully considered when implementing low salinity waterflooding projects in sandy reservoirs. If the
shales are fully saturated with more salty connate water, they may increase the salinity of the injected low
salinity water by mixing and thus might affect the overall oil recovery. While little research has been
performed investigating the effects of shale on low salinity waterflooding, extensive research has been done
on other EOR processes such as steam flooding (Henson et al. 2002, Trigos et al. 2010). It is possible
that connate water in shales may also affect surfactant and alkaline-surfactant-polymer flooding.
In steam flooding, high temperature steam is injected into reservoirs to recover highly viscous oil. The
effectiveness of this process is severely affected by the presence of shales (Henson et al. 2002, Trigos et al.
54
2010). Shales often act as baffles to flow so they distort the steam flood front as well as resulting in heat
loss. As a result, the overall oil recovery will be reduced. Shales have higher thermal conductivity compared
with sand, so most of the injected steam will tend to heat the shales rather than sands. Therefore, the energy
of heating is reduced due to shales thus oil recovery from the sands is reduced (Trigos et al. 2010).
For low salinity waterflooding, a recent simulation study by Brodie and Jerauld (2014) indicated that the
effect of salt diffusion on the performance of low salinity waterflooding is most significant on thin sands
adjacent to thick waterfilled shales. However, this study did not correlate the oil recoveries with the
transverse dispersion number (NTD) thus did not consider the flow regime (whether it is diffusion
dominated, heterogeneity dominated or in between).
Based on these previous findings, it is vital to investigate in this thesis the effects of the presence of shales
on the effectiveness of low salinity waterflooding.
2.2.3 The Effect of Aquifers on Low Salinity Waterflooding
The presence of aquifers plays a huge factor in improving the oil recovery by conventional waterflooding
especially if they result in large water influx. However, in some cases when there are many vertical fractures,
the aquifer water quickly moves to the producer leading to a rapid increase in the produced water cut thus
reducing the overall oil recovery.
In low salinity waterflooding, the presence of aquifers can affect the performance of low salinity
waterflooding when the more saline aquifer water mixes with the injected low salinity water and thus
reduces its efficiency. To date, the effect of the presence and strength of aquifers on the efficiency of low
salinity waterflooding has been studied by various researchers (Law et al. 2014, Mahani et al. 2011,
Bedrikovetsky et al. 2015, Akhmetgareev and Khisamov 2015).
A simulation study on the field scale conducted by Law et al. (2014) found an incremental oil recovery of
4% with the inclusion of a strong aquifer in their model. However, the increase cannot be generalized as
mixing of low salinity water by the aquifer water was not considered.
Another simulation study by Mahani et al. (2011) on a field scale evaluated the effect of the fraction of the
injected low salinity water into aquifers on oil recovery. Figure 2.27 shows the incremental oil gained
beyond conventional waterflooding as a function of the low salinity water injected into the aquifer. The
incremental oil recovery is significantly decreased when the fraction of the injected low salinity water into
the aquifer is greater than 0.7.
Figure 2.27: The incremental oil gained beyond conventional waterflooding as a function of the fraction of the low salinity water injected
into the aquifer. The effect of this fraction significantly decreases the incremental oil recovery when the value is greater than 0.7. Nve stands
for the vertical equilibrium number used in their simulation model (from Mahani et al. 2011).
55
The effect of the presence of aquifers on low salinity waterflooding was analysed by Bedrikovetsky et al.
(2015) and Akhmetgareev and Khisamov (2015). It was found that the presence of aquifers had two different
effects on the efficiency of low salinity waterflooding. The presence of an aquifer showed an adverse effect
by reducing the efficiency of low salinity waterflooding at injectors completed close to the aquifer as they
experienced mixing by the aquifer water. However, the presence of an aquifer showed no effects on the
performance of low salinity waterflooding at injectors completed away from the aquifer. As there are
different effects, the degree of mixing of the injected low salinity water by the aquifer water in addition to
the well completion are further investigated later in this thesis.
2.3 Upscaling
It is often not practical to run full field fine detailed geological model (~millions of grid blocks) due to
limitations in CPU time and capacity (Salazar and Villa 2007). In addition, sensitivity analyses often require
many simulation runs. Therefore, upscaling the fine grid model to a much coarser grid model (~thousands of
grid blocks) is required. The coarse grid model must capture the effective reservoir properties of the fine grid
model.
To date, there are no methods that have been developed to upscale low salinity waterflooding. There are two
main challenges that the upscaling method has to overcome: using a relatively coarse grid model that
represents the effective reservoir properties and minimizing numerical dispersion that usually arises in the
coarser grid model. Once these two main challenges are well resolved, the coarse grid model can be utilized
in conducting the required sensitivity analysis for low salinity waterflooding projects.
In the literature, there are two main upscaling methods that represent the effect of heterogeneities in the
coarse grid model: single phase upscaling and two phase upscaling. In single phase upscaling, only the
absolute permeabilities of the fine grid model are upscaled in the coarse grid model while in two phase
upscaling, relative permeabilities are upscaled.
There are three main methods in single phase upscaling: heuristic, deterministic and stochastic methods
(Renard and De Marsily 1996). In the stochastic methods, the absolute permeabilities of the fine grid model
are assumed to be randomly distributed. Then, the permeabilities in the coarse grid model are determined
from solving the following stochastic partial differential equations (Renard and De Marsily 1996):
𝑣𝑣 = −𝐾𝐾 ∇𝑃𝑃 … … … … … … (2.39)
∇. 𝑣𝑣 = 0 … … … … … … … . . (2.40)
E (𝑣𝑣) = −𝐾𝐾𝑒𝑒𝑒𝑒 𝐸𝐸(∇𝑃𝑃) … … . . (2.41)
where:
𝑣𝑣 is the Darcy velocity (L/t)
𝑃𝑃 is the pressure (psi)
E (𝑣𝑣) is the mathematical expectation of the Darcy velocity and 𝐸𝐸 (𝛻𝛻𝛻𝛻) is the mathematical expectation of
the pressure gradient
𝐾𝐾 is the permeability tensor and 𝐾𝐾𝑒𝑒𝑒𝑒 is the effective permeability tensor (md)
56
In the heuristic methods, the absolute permeabilities in the coarse grid model are calculated based on
averaging the absolute permeabilities in the fine grid model. They are computationally faster than the other
methods. In the deterministic methods, flow based simulations in the fine grid model are used to determine
the effective permeabilities that will be used in the coarse grid model (Renard and De Marsily 1996, Zhou
2010). The resulting effective permeabilities depend on the boundary conditions that are assumed in the flow
based simulations using the fine grid model. These boundary conditions are generally classified into:
periodic boundary conditions, and constant pressure across the horizontal direction with no flow across the
vertical direction (Zhou 1996, Christie 2001). For periodic boundary conditions (PBCs), it is assumed that
each coarse grid cell can be represented by a repetitive (periodic) unit cell (Durlofsky 1991) as shown in
Figure 2.28. Unlike other boundary conditions which generate negative elements in some of the diagonal
tensor for a highly heterogeneous reservoir, PBCs always generate positive definite and symmetric matrix
(Pickup et al. 1992, Pickup et al. 1994, Renard and De Marsily 1996, Christie 1996). Therefore, PBCs is
more robust (Pickup et al. 1992, Pickup et al. 1994, Christie 1996, Renard and De Marsily 1996).
Figure 2.28: When upscaling using periodic boundary conditions, each coarse grid cell can be represented by a repetitive (periodic) unit cell. P is the
pressure and nx refers to the x direction while ny refers to the y-direction (from Pickup et al. 1994).
Since single phase upscaling involves running flow based simulations using a single fluid, it cannot capture
the effects of heterogeneity on two-phase flow such as channelling during low salinity waterflooding. Two-
phase upscaling generally can capture the impact of various heterogeneities during waterflooding such as
channelling (Christie 1996, Muggeridge and Hongtong 2014). It is classified into two main methods: steady
state upscaling and dynamic pseudoization methods.
Steady state upscaling methods are generally faster in computation than dynamic pseudo relative
permeability methods. In addition, running the full fine scale model is not required in steady state upscaling
methods (Jonoud and Jackson 2008, Ekrann and Aasen 2000, Odsaeter et al. 2015). However, the limitation
of using these methods is that they require more than one rock type (i.e. more than one set of relative
permeability curves) when generating the pseudo relative permeabilities that will be used in the coarse grid
model. In this thesis, only one rock type will be used for upscaling low salinity waterflooding so steady state
upscaling methods will not serve the purpose of generating pseudo relative permeabilities that are utilized
when upscaling low salinity waterflooding.
57
In dynamic pseudoization methods, pseudo relative permeabilities are generated from running the fine grid
simulation models. The purpose is to compensate for numerical dispersion (Barker and Dupouy 1999). There
are two practical challenges that arise when using these methods: the choice of the number of grid blocks,
well positioning and rates (Barker and Dupouy 1999). For every grid block in the coarse model, there is a
different set of pseudo relative permeabilities, in addition if there is a sensitivity study in which it is
necessary to change the number of grid blocks, the set of pseudo relative permeabilities will be changed. In
addition, if the well locations or rates have to be changed as sensitivity parameters, then a new set of pseudo
relative permeabilities must be generated.
In low salinity waterflooding, pseudo relative permabilities are difficult to generate using dynamic
pseudoization methods since the displacement is complicated by two shock fronts in addition to the presence
of the connate water bank. There are two set of relative permeability curves (one for high salinity water and
the other one for low salinity water) and hence 2 set of pseudos (Muggeridge and Hongtong 2014). These
factors make the generation of the pseudo relative permeabilities very challenging to generate for low
salinity waterflooding.
Instead of the above upscaling methods, upscaling low salinity waterflooding in this thesis is based on
modifications of Hearn’s method (Hearn 1971). This method generates pseudo relative permeabilities for
conventional waterflooding from a 2D stratified reservoir such as illustrated in Figure 2.29.
Figure 2.29: A schematic of a 2D stratified reservoir used in the Hearn method to generate pseudo relative permeabilities in conventional
waterflooding (from Hearn 1971).
The resulting pseudo relative permeabilities are used in the 1D homogeneous model. The following
assumptions are made when deriving the pseudo relative permeabilities (Hearn 1971):
• The reservoir is stratified with different rock properties for each layer
• Water and oil are incompressible
• The displacement front is sharp with only oil flowing at the irreducible water saturation (Swc) ahead
of the shock front and only water flowing behind the shock front at the residual oil saturation (Sor).
• Gravity and capillary forces are negligible
• The reservoir has only one rock type (i.e. one set of relative permeability curves)
• Crossflow is negligible
58
The layers are sorted in a decreasing permeability value with the flooding order from top to bottom. The
average water saturation, the pseudo relative permeabilities are calculated at the outlet end of the model as
follows:
At n = 0 (before water breakthrough)
∑𝑁𝑁
𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖 𝑆𝑆𝑤𝑤𝑤𝑤𝑤𝑤
𝑆𝑆𝑤𝑤0 = (2.42)
∑𝑁𝑁
𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖
𝑘𝑘�
𝑤𝑤0 = 0 (2.43)
𝑘𝑘�
𝑜𝑜0 = 𝑘𝑘𝑜𝑜𝑜𝑜 (2.44)
st
At n = 1,2,…N-1 (breakthrough of 1 top layer,…..,breakthrough of all layers except the bottom one)
∑𝑁𝑁 𝑁𝑁
𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖 (1 − 𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜 ) + ∑𝑖𝑖=𝑛𝑛+1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖 𝑆𝑆𝑤𝑤𝑤𝑤𝑤𝑤
𝑆𝑆𝑤𝑤𝑤𝑤 = (2.45)
∑𝑁𝑁𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖
𝑘𝑘𝑤𝑤𝑤𝑤 ∑𝑛𝑛𝑖𝑖=1 𝐾𝐾𝑖𝑖 ℎ𝑖𝑖

𝑘𝑘�
𝑤𝑤𝑤𝑤 = (2.46)
∑𝑁𝑁𝑖𝑖=1 𝐾𝐾𝑖𝑖 ℎ𝑖𝑖
𝑘𝑘𝑜𝑜𝑜𝑜 ∑𝑁𝑁
𝑖𝑖=𝑛𝑛+1 𝐾𝐾𝑖𝑖 ℎ𝑖𝑖
𝑘𝑘�
𝑜𝑜𝑜𝑜 = 𝑁𝑁 (2.47)
∑𝑖𝑖=1 𝐾𝐾𝑖𝑖 ℎ𝑖𝑖
At n = N (breakthrough of last bottom layer)
∑𝑁𝑁
𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖 (1 − 𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜 )
𝑆𝑆𝑤𝑤𝑤𝑤 = (2.48)
∑𝑁𝑁
𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖
𝑘𝑘�
𝑤𝑤𝑤𝑤 = 𝑘𝑘𝑤𝑤𝑤𝑤 (2.49)
𝑘𝑘�
𝑜𝑜𝑜𝑜 = 0 (2.50)
where
N is the total number of layers
𝑆𝑆𝑤𝑤0 is the average water saturation at outflow face before the water breakthrough of layer N
ℎ𝑖𝑖 is the thickness of layer i (m)
𝜙𝜙𝑖𝑖 is the porosity of layer i
𝐾𝐾𝑖𝑖 is the absolute permeability of layer i (md)
𝑆𝑆𝑤𝑤𝑤𝑤𝑤𝑤 is the connate water saturation of layer i
𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜 is the residual oil saturation of layer i
𝑘𝑘𝑜𝑜𝑜𝑜 is the oil relative permeability at the connate water saturation
𝑘𝑘𝑤𝑤𝑤𝑤 is the water relative permeability at the residual oil saturation
𝑘𝑘�
𝑤𝑤0 is the pseudo water relative permeability for the average water saturation before the water breakthrough
of layer N
𝑘𝑘�
𝑜𝑜0 is the pseudo oil relative permeability for the average water saturation before the water breakthrough of
layer N
𝑆𝑆𝑤𝑤𝑤𝑤 is the average water saturation at outflow face before the water breakthrough of all layers except the
bottom one (N)
59
𝑆𝑆𝑤𝑤𝑤𝑤 is the average water saturation at outflow face after the water breakthrough of layer N
𝑘𝑘�
𝑤𝑤𝑤𝑤 is the pseudo water relative permeability for the average water saturation after the water breakthrough of
any layer between 1< n< N

𝑘𝑘�
𝑜𝑜𝑜𝑜 is the pseudo oil relative permeability for the average water saturation after the water breakthrough of
any layer between 1< n< N

𝑘𝑘�
𝑤𝑤𝑤𝑤 is the pseudo water relative permeability for the average water saturation after the water breakthrough of
layer N
𝑘𝑘�
𝑜𝑜𝑜𝑜 is the pseudo oil relative permeability for the average water saturation after the water breakthrough of
layer N
The limitation of this method is that it does not consider the connate water bank and the presence of two
shock fronts, so further modifications are required to improve the generations of the pseudo relative
permeability when upscaling low salinity waterflooding. This will be shown later in chapter 7.
Another challenge that arises when upscaling the fine grid model to the coarse model is to minimize
numerical dispersion. Numerical dispersion affects the two shock fronts of the low salinity waterflooding
especially the trailing shock fronts. The connate water bank that can result in mixing between the injected
low salinity water and connate water is most affected by numerical dispersion as discussed earlier. Jerauld et
al. (2008) attempted to capture physical dispersion by the numerical dispersion. It was suggested that 10 grid
blocks between wells are sufficient to capture the physical dispersion. However, this is not always practical
in the coarse grid model as the number of grid blocks might be significantly lower (Muggeridge and
Hongtong 2014). The representation of physical dispersion by numerical dispersion will be investigated and
evaluated in chapter 4.
Muggeridge and Hongtong (2014) have developed an analytical upscaling method to minimize numerical
dispersion in low salinity waterflooding by generating pseudo relative permeability from conventional
waterflooding. The advantage of this method over other pseudoization methods (Kyte and Berry 1975,
Barker and Dupouy 1999) is that it does not require the output of a fine scale two- phase flow model for
generating the pseudo relative permeabilities.
One of the objectives of this thesis is to develop an analytical upscaling method that consider the effect of
the connate water bank by modifying Hearn’s method (will be shown later in chapter 7).
2.4 Summary
The literature has shown that there is no evaluation of the impact of reservoir heterogeneity and mixing on
low salinity waterflooding
Low salinity waterflooding was confirmed to improve oil recovery by several laboratory experiments and
field applications. It is widely accepted that it recovers oil by modifying the rock wettability towards more
water wet (Nasralla et al. 2013, Dang et al. 2015, Kuznetsov et al. 2015, Law et al. 2015, Jerauld et al.
2008). Due to the complex interaction between the brine/oil/minerals, the pore scale mechanism is debatable.
60
Low salinity waterflooding is different from conventional waterflooding since it has two shock fronts
compared to just one in conventional waterflooding. In between these shock fronts, a connate water bank
forms. The presence of this connate water bank has been observed by laboratory experiments and field
applications.
It is anticipated that mixing occurs between the injected low salinity water and the connate water especially
in heterogeneous systems as a result of overlapping the connate water banks. The literature has shown that
there is a lack of previous studies on this particular effect on low salinity waterflooding, so one of the aims
of this research is to investigate mixing effects on the performance of low salinity waterflooding in the
presence of various heterogeneity scales using detailed simulation study.
The effect of reservoir heterogeneity on conventional waterflooding is well known (such as the presence of
high permeability streaks within the reservoir and the presence of shales). However, low salinity
waterflooding is different from conventional waterflooding as it involves mixing between the injected low
salinity water and connate water. It is anticipated that reservoir heterogeneity affects the performance of low
salinity waterflooding more than ordinary waterflooding. While the literature has shown that most of the
current research is focusing on the pore scale mechanism of low salinity waterflooding, the impact of
reservoir heterogeneity on low salinity waterflooding was not investigated. Consequently, this research will
evaluate these effects in details using a detailed simulation study.
Previous studies have shown also that the presence of aquifers affects the performance of low salinity
waterflooding (Law et al. 2014, Mahani et al. 2011, Bedrikovetsky et al. 2015, Akhmetgareev and Khisamov
2015). However, these studies did not consider that when the aquifer water encroaches into the oil reservoirs
it will mix with the injected low salinity water thus reducing its efficiency. In this thesis, a simulation study
will be conducted to investigate the effect of mixing on the performance of low salinity waterflooding.
While modelling reservoir heterogeneity requires a very fine grid block model, it is not always practical to
run it due to limitations in computation capability and time frame. The literature has shown limited work on
upscaling low salinity waterflooding. Therefore, one of the objectives of this research is to propose an
analytical upscaling method for low salinity waterflooding.
61
CHAPTER 3: MODELLING LOW SALINITY WATERFLOODING
The aim of this chapter is to investigate how low salinity waterflooding is impacted by: the choice of the
number of grid blocks that minimizes numerical dispersion and hence ensures that the physical longitudinal
and transverse diffusion are properly modelled and the method used to model the transition from using high
salinity relative permeabilities to low salinity relative permeabilities and the choice of the thresholds. The
number of grid blocks must be chosen to ensure that the simulation is dominated by physical diffusion rather
than numerical diffusion. As noted in chapter 2, there are two methods used in black oil simulators to relate
the relative permeability curves to the salinity: the simple salinity threshold model proposed by Jerauld et al.
(2008) that is implemented in (ECLIPSE and CMG STARS) and the ion exchange model proposed by (Dang
et al. 2013) that is implemented in CMG GEM. It is important to understand the differences between these
two models and how these impact the predictions as the rest of the thesis uses one of these models. The
impact of the low salinity waterflooding model on predictions is investigated for both the salinity threshold
model and the ion exchange model. For the salinity threshold model, the choice of the salinity thresholds and
its impact on low salinity waterflooding is investigated while for the ion exchange model the choice of the
interpolation parameter, the values to change from the high salinity water relative permeability curve to the
low salinity water relative permeability curve and the cation exchange capacity and their impacts on low
salinity waterflooding are investigated.
3.1 Reservoir and Fluid Properties Used for Modelling Low Salinity Waterflooding
The rock and fluid properties used for modelling low salinity waterflooding in this chapter are shown in
Table 3.1. The relative permeability curves for high salinity water and low salinity water are shown in Figure
3.1. They are taken from a template input file from ECLIPSE simulator that shows how to model a low
salinity water flood. The high salinity water relative permeability curves shows the characteristics of a
𝑚𝑚𝑚𝑚𝑚𝑚
mixed/oil wet rock since its end point water relatie permeability value (𝑘𝑘𝑟𝑟𝑟𝑟 ) is less than that for the low
salinity water relative permabilty curve. This figure shows different connate water saturation which is not
physically possible (i.e. low salinity waterflooding does not change the connate water saturation). However,
the value of the connate water saturation to low salinity waterflooding in the field is not meaningful as the
only water saturations that are greater than the trailing shock fronts are seen in the reservoir and modelled in
the simulator. This is illustrated in section 3.2.
Table 3.1: Rock and fluid properties used for modelling low salinity waterflooding in this chapter.
Property Value
Model dimensions (x, y, z) (m) 1,020 × 100 × 50
Porosity 0.2
Permeability (x, y, z) (md) 100
Initial reservoir pressure (bar) 265
Rock compressibility (bara-1) 0.3 × 10-5
Water compressibility (bara-1) 4.6 × 10-5
Oil density (kg/m3) 850
Water density (kg/m3) 1,000
Oil viscosity (cp) 0.47
Water viscosity (cp) 0.5
Oil formation volume factor (rm3/sm3) 0.999
Water formation volume factor (rm3/sm3) 1.03
Connate water salinity (1,000 ppm) 30
62
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Sw
krwH
krwh kroH
kroh krwL
krwL kroL
kroL
Figure 3.1: Relative permeability curves used for modelling low salinity waterflooding in this chapter.
3.2 Analytical Model for Low Salinity Waterflooding

The following assumptions are made for the analytical Buckley Leverett solution of low salinity
waterflooding:
• One dimensional flow in a homogeneous porous media with isotropic permeabilities
• Fluids are incompressible
• Gravity and capillary effects are ignored
• No salt diffusion
• The salt concentration does not alter the volume of the water phase
• There is an instantaneous transition from the high salinity water (HS) to the low salinity water (LS)
(i.e. the flow is controlled by either the high salinity water relative permeability curve or the low
salinity water relative permeability curve depending on the salinity of the water)
The analytical Buckley Leverett solution of low salinity waterflooding is similar to that for polymer flooding
(Pope (1980), Green and Willhite 1998)), however the adsorption and inaccessible pore volume terms are
not applicable here (Jerauld et al. 2008).
The continuity equation for low salinity water is different from that for high salinity water in the dependence
of the water salinity on both space (x) and time (t). Therefore, there are two continuity equations: one for the
water phase and the other for the salt component as follows:
Water continuity equation:
𝜕𝜕(𝑆𝑆𝑤𝑤 𝐶𝐶𝑤𝑤 ) 𝑞𝑞𝑖𝑖 𝜕𝜕(𝑓𝑓𝑤𝑤 𝐶𝐶𝑤𝑤 )
+ = 0 … … . (3.1)
𝜕𝜕𝜕𝜕 𝐴𝐴𝐴𝐴 𝜕𝜕𝜕𝜕
Salt component continuity equation:
𝜕𝜕(𝑆𝑆𝑤𝑤 𝐶𝐶𝑖𝑖 ) 𝑞𝑞𝑖𝑖 𝜕𝜕(𝑓𝑓𝑤𝑤 𝐶𝐶𝑖𝑖 )
+ = 0 … … … … . . ( 3.2)
𝜕𝜕𝜕𝜕 𝐴𝐴𝐴𝐴 𝜕𝜕𝜕𝜕
𝐶𝐶𝑤𝑤 is the water component
𝐶𝐶𝑖𝑖 is the amount of salt that is dissolved in water, it can be either the high salinity (HS) or the low salinity
(LS), there are no intermediate salinities (ppm)
63
𝑆𝑆𝑤𝑤 is the water saturation (dimensionless)
𝑞𝑞𝑖𝑖 is the injection rate (L3/t)
A is the cross sectional area (L2)
𝜙𝜙 is the porosity (dimensionless)
𝐿𝐿 is the model length (𝐿𝐿)
The continuity equations can be rewritten in a dimensionless form. The dimensionless distance 𝑥𝑥𝐷𝐷 and
dimensionless time 𝑡𝑡𝐷𝐷 is defined by:
𝑞𝑞𝑖𝑖
𝑡𝑡𝐷𝐷 =
𝑡𝑡 … … … … (3.3)
𝐴𝐴𝐴𝐴𝐴𝐴
𝑥𝑥
𝑥𝑥𝐷𝐷 = … … … … … . … (3.4)
𝐿𝐿
By using the dimensionless time and distance, equations 3.1 and 3.2 becomes:
𝜕𝜕(𝑆𝑆𝑤𝑤 𝐶𝐶𝑤𝑤 ) 𝜕𝜕(𝑓𝑓𝑤𝑤 𝐶𝐶𝑤𝑤 )
+ = 0 … … . (3.5)
𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷
𝜕𝜕(𝑆𝑆𝑤𝑤 𝐶𝐶𝑖𝑖 ) 𝜕𝜕(𝑓𝑓𝑤𝑤 𝐶𝐶𝑖𝑖 )
+ = 0 … … … … . . ( 3.6)
𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷
Neglecting gravity and capillary effects, the fractional flow of the water phase (𝑓𝑓𝑤𝑤 ) is calculated by:
1
𝑓𝑓𝑤𝑤 = … … … . . (3.7)
𝑘𝑘 𝜇𝜇
1 + 𝑟𝑟𝑟𝑟 𝑤𝑤
𝜇𝜇𝑜𝑜 𝑘𝑘𝑟𝑟𝑟𝑟
where: 𝑘𝑘𝑟𝑟𝑟𝑟 and 𝑘𝑘𝑟𝑟𝑟𝑟 are the relative permeabilities to oil and water, respectively
𝜇𝜇𝑜𝑜 and 𝜇𝜇𝑤𝑤 are the oil and water viscosities, respectively
The mass balance of the water phase equation is:
𝜕𝜕𝑆𝑆𝑤𝑤 𝜕𝜕𝑓𝑓𝑤𝑤
+ = 0 … … … … … . . ( 3.8)
𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷
Expanding the salt component continuity equation (equation 3.6) gives:
𝜕𝜕𝐶𝐶𝑖𝑖 𝜕𝜕𝑆𝑆𝑤𝑤 𝜕𝜕𝐶𝐶𝑖𝑖 𝜕𝜕𝑓𝑓𝑤𝑤
𝑆𝑆𝑤𝑤 + 𝐶𝐶𝑖𝑖 + 𝑓𝑓𝑤𝑤 + 𝐶𝐶𝑖𝑖 = 0 … … … … . . ( 3.9)
𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷
Grouping the common terms in equation 3.9 gives:
𝜕𝜕𝑓𝑓𝑤𝑤 𝜕𝜕𝑆𝑆𝑤𝑤 𝜕𝜕𝐶𝐶𝑖𝑖 𝜕𝜕𝐶𝐶𝑖𝑖
𝐶𝐶𝑖𝑖 � + � + 𝑆𝑆𝑤𝑤 + 𝑓𝑓𝑤𝑤 = 0 … … … … . . ( 3.10)
𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷
The term in between the brackets is zero from equation 3.8, so equation 3.10 reduces to:
𝜕𝜕𝐶𝐶𝑖𝑖 𝜕𝜕𝐶𝐶𝑖𝑖
𝑆𝑆𝑤𝑤 + 𝑓𝑓𝑤𝑤 = 0 … … … … . . ( 3.11)
𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷
It is assumed that there is no salt diffusion, so there are no intermediate salinities, therefore the salinity
changes instantaneously from the high salinity to the low salinity. The salt component (𝐶𝐶𝑖𝑖 ) is a function of
space (x) and time (t):
𝑑𝑑𝐶𝐶𝑖𝑖 = 𝑑𝑑𝑑𝑑𝐷𝐷 + 𝑑𝑑𝑑𝑑 … … … . . (3.12)
𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷 𝐷𝐷
64
Since the there is no overall change to the salt concentration 𝑑𝑑𝐶𝐶𝑖𝑖 = 0, so equation 3.12 becomes:
𝑑𝑑𝑑𝑑𝐷𝐷 = − 𝑑𝑑𝑑𝑑 … … … . . (3.13)
𝜕𝜕𝜕𝜕𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷 𝐷𝐷
Substituting equation 3.13 in equation 3.11 gives:
𝜕𝜕𝐶𝐶𝑖𝑖 𝑑𝑑𝑑𝑑𝐷𝐷 𝜕𝜕𝐶𝐶𝑖𝑖
𝑆𝑆𝑤𝑤 �− � + 𝑓𝑓𝑤𝑤 = 0 … … … … . . ( 3.14)
𝜕𝜕𝜕𝜕 𝑑𝑑𝑑𝑑𝐷𝐷 𝜕𝜕𝜕𝜕𝐷𝐷
𝑑𝑑𝑑𝑑𝐷𝐷
Solving for :
𝑑𝑑𝑑𝑑𝐷𝐷
𝑑𝑑𝑑𝑑𝐷𝐷 𝑓𝑓𝑤𝑤
� = … … ( 3.15)
𝑑𝑑𝑑𝑑𝐷𝐷 𝐶𝐶 𝑆𝑆𝑤𝑤
𝑖𝑖
The analytical Buckley Leverett solution for low salinity waterflooding is different from conventional
waterflooding. Unlike conventional waterflooding that has one shock front, low salinity waterflooding has
two shock fronts. The oil saturated zone is ahead of the first (leading) shock front while the low salinity
water zone is behind the second (trailing) shock front (Jerauld et al. 2008). In between these two regions, a
connate water bank is formed. A typical fractional flow curve and water saturation profile for a low salinity
waterflood are shown in Figures 3.2 and 3.3.
The change in salt concentration from high salinity to low salinity causes a saturation shock (trailing shock
front). The velocity of this shock is given by:
𝑑𝑑𝑑𝑑𝐷𝐷 𝑓𝑓𝑤𝑤,𝐿𝐿𝐿𝐿 − 𝑓𝑓𝑤𝑤,𝐻𝐻𝐻𝐻
� =� � … … ( 3.16)
𝑑𝑑𝑑𝑑𝐷𝐷 𝑆𝑆 𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 − 𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻
𝑤𝑤,𝐿𝐿𝐿𝐿
The water saturation at this trailing shock front (𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 ) is determined by drawing a tangent line to the
fractional flow curve for the low salinity water (red curve in Figure 3.2). When this tangent line intersects
with the fractional flow curve for the high salinity water (blue curve in Figure 3.2), the water saturation of
the leading shock front (𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 ) is determined. The velocity of the leading shock front is given by:
𝑑𝑑𝑑𝑑𝐷𝐷 𝑓𝑓𝑤𝑤,𝐻𝐻𝐻𝐻 − 0
� =� � … … ( 3.17)
𝑑𝑑𝑑𝑑𝐷𝐷 𝑆𝑆 𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 − 𝑆𝑆𝑤𝑤𝑤𝑤
𝑤𝑤,𝐻𝐻𝐻𝐻
where: 𝑆𝑆𝑤𝑤𝑤𝑤 is the connate water saturation.

For any saturation (𝑆𝑆𝑤𝑤 > 𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 ), the velocity of this saturation value is estimated from the slope of its
tangent line:
𝑑𝑑𝑑𝑑𝐷𝐷 𝑑𝑑𝑓𝑓𝑤𝑤
� = … … … … … … … (3.18)
𝑑𝑑𝑑𝑑𝐷𝐷 𝑆𝑆 𝑑𝑑𝑆𝑆𝑤𝑤
𝑤𝑤
65
1
Sw,LS , fw,LS
Sw,HS , fw,HS Sw,LS
Sw,HS
Sw
fw
Swc
0
0 Swc Sw 1
HS LS X
Figure 3.2: Typical fractional flow curves for low salinity (LS) and Figure 3.3: A typical water saturation profile for a low salinity
high salinity (HS) water curves. A tangent line is drawn from the waterflood. The first (leading) front refers to the region between the
(0,0) point to the low salinity fractional flow curve (LS) and the high salinity water and the connate water in the reservoir. The second
trailing shock front water saturation is determined ( Sw,LS ). When (trailing) shock front refers to the region between low salinity water
this line crosses the fractional flow curve for the high salinity and the high salinity water. In between these two regions, a connate
water, the leading shock front water saturation is determined water bank is formed.
(Sw,HS ). The velocity of trailing shock front can be estimated from
the slope of this tangent line. The velocity of the leading shock
front is determined from the slope of the line that connects the
connate water saturation (Swc) and the leading shock front
saturation ( Sw,HS ).
Example
The following example illustrates how to analytically calculate the produced water and oil rates at several
pore volume injections and how to generate a water saturation profile for a 1D low salinity waterflood using
the reservoir model described in section 3.1. Low salinity water is injected immediately. First, the two shock
front saturations and the slope of the two lines are calculated as shown in Figure 3.4.
66
1
0.9 Sw,HS = 0.52
0.8 Sw,LS = 0.72
𝒅𝒅𝒇𝒇𝑯𝑯𝑯𝑯,𝒘𝒘
0.7 = 2.06 𝒅𝒅𝒇𝒇𝑳𝑳𝑳𝑳,𝒘𝒘
= 1.32
𝒅𝒅𝑺𝑺𝒘𝒘
Fractional Flow
0.6 𝒅𝒅𝑺𝑺𝒘𝒘
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Sw
HS LS
Figure 3.4: The water saturation of the shock fronts and the slope of the two lines that are used to estimate their velocities in
the reservoir model.
The times of the leading shock front and the trailing shock front breakthrough are given by:
𝑃𝑃𝑃𝑃 ⎛ 1 ⎞
𝑡𝑡𝐵𝐵𝐵𝐵,𝐻𝐻𝐻𝐻 = .⎜ ⎟ … … … … . . (3.19)
𝑞𝑞𝑖𝑖 𝑑𝑑𝑓𝑓𝑤𝑤
�
𝑑𝑑𝑆𝑆𝑤𝑤 𝑆𝑆
⎝ 𝑤𝑤,𝐻𝐻𝐻𝐻 ⎠
𝑃𝑃𝑃𝑃 ⎛ 1 ⎞
𝑡𝑡𝐵𝐵𝐵𝐵,𝐿𝐿𝐿𝐿 = .⎜ ⎟ … … … … . . (3.20)
𝑞𝑞𝑖𝑖 𝑑𝑑𝑓𝑓𝑤𝑤
�
⎝ 𝑤𝑤,𝐿𝐿𝐿𝐿 ⎠
where:
𝑡𝑡𝐵𝐵𝐵𝐵,𝐻𝐻𝐻𝐻 , 𝑡𝑡𝐵𝐵𝐵𝐵,𝐿𝐿𝐿𝐿 are the times of the leading shock front breakthrough and the trailing shock front
breakthrough, respectively.
PV: is the model’s pore volume in L3
𝑞𝑞𝑖𝑖 : is the injection rate in L3/t
𝑑𝑑𝑓𝑓𝑤𝑤
� : is the slope of the line that connects the connate water saturation and the leading shock front
saturation as shown in Figure 3.4

�
: is the slope of the tangent line at the trailing shock front saturation as shown in Figure 3.4
67
The time of the leading shock front breakthrough is calculated as follows:
𝑃𝑃𝑃𝑃 ⎛ 1 ⎞ 6,286,124.3 1
𝑡𝑡𝐵𝐵𝐵𝐵,𝐻𝐻𝐻𝐻 =
.⎜ ⎟ = × = 1,589.9 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
𝑞𝑞𝑖𝑖 𝑑𝑑𝑓𝑓𝑤𝑤 1,916.9 2.1
�
⎝ 𝑤𝑤,𝐻𝐻𝐻𝐻 ⎠
The time of the trailing shock front breakthrough is calculated as follows:
𝑃𝑃𝑃𝑃 ⎛ 1 ⎞ 6,286,124.3 1
𝑡𝑡𝐵𝐵𝐵𝐵,𝐿𝐿𝐿𝐿 =
.⎜ ⎟ = × = 2,511.8 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
𝑞𝑞𝑖𝑖 𝑑𝑑𝑓𝑓𝑤𝑤 1,916.9 1.3
�
⎝ 𝑤𝑤,𝐿𝐿𝐿𝐿 ⎠
The water cut versus pore volume of low salinity water injected can be generated as per the following steps.
The producing water cut at the breakthrough of the leading shock front:
1- The water saturation of the leading shock front is estimated from Figure 3.4.
2- The fractional water value is read from the high salinity water curve in Figure 3.4.
3- � is calculated as shown in Figure 3.4.
4- The water oil ratio (WOR) is calculated as per the following equation:
𝐵𝐵𝑜𝑜 𝑓𝑓𝑤𝑤
𝑊𝑊𝑊𝑊𝑊𝑊 = … … . (3.21)
𝐵𝐵𝑤𝑤 (1 − 𝑓𝑓𝑤𝑤 )
where:
𝐵𝐵𝑜𝑜 is the oil formation volume factor in reservoir bbl/STB
𝐵𝐵𝑤𝑤 is the water formation volume factor in reservoir bbl/STB
𝑓𝑓𝑤𝑤 is the value from step 2
5- The produced oil flow rate (𝑞𝑞𝑜𝑜 ) is calculated as per the following equation:
𝑞𝑞𝑖𝑖
𝑞𝑞0 = … … … . . (3.22)
𝐵𝐵𝑜𝑜 + (𝐵𝐵𝑤𝑤 × 𝑊𝑊𝑊𝑊𝑊𝑊)
6- The produced water flow rate (𝑞𝑞𝑤𝑤 ) is calculated as per the following equation:
𝑞𝑞𝑤𝑤 = 𝑞𝑞𝑜𝑜 × 𝑊𝑊𝑊𝑊𝑊𝑊 … … … … . (3.23)
7- The total produced flow rate (𝑞𝑞𝑡𝑡 ) is calculated as per the following equation:
𝑞𝑞𝑡𝑡 = 𝑞𝑞𝑜𝑜 + 𝑞𝑞𝑤𝑤 … … … … (3.24)
8- The produced water cut (WC) is calculated as per the following equation:
𝑞𝑞𝑤𝑤
𝑊𝑊𝑊𝑊 = … … … … (3.25)
𝑞𝑞𝑡𝑡
9- The pore volume of low salinity water injected (𝑄𝑄𝑖𝑖 ) up to the breakthrough of the leading shock
front is calculated by:
1
𝑄𝑄𝑖𝑖 = … … (3.26)
�
68
10- The cumulative low salinity water injected (𝑊𝑊𝑖𝑖𝑖𝑖𝑖𝑖 ) in (L3) at 𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 is calculated as per the following
equation:
𝑊𝑊𝑖𝑖𝑖𝑖𝑖𝑖 = 𝑃𝑃𝑃𝑃 × 𝑄𝑄𝑖𝑖 … … … … . (3.27)
11- The time (t) required to inject 𝑊𝑊𝑖𝑖𝑖𝑖𝑖𝑖 is given by :
𝑊𝑊𝑖𝑖𝑖𝑖𝑖𝑖
𝑡𝑡 = … … … … . . . (3.28)
𝑞𝑞𝑖𝑖
Steps 1 to 11 are repeated for 𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 with the only difference in step 3 being the value of �
instead.
Then, for all water saturations (𝑆𝑆𝑤𝑤 > 𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 ), steps 1 to 11 are repeated but the only difference in step 3 being
that � is estimated from the tangent line at this 𝑆𝑆𝑤𝑤 .
𝑤𝑤
The values of produced water cut up to 2 PV of immediate injection of low salinity water are shown in Table
3.2 and are plotted in Figure 3.5.
Table 3.2: The analytical solution up to 2 PV of immediate injection of low salinity water into the 1D homogeneous model.
𝑑𝑑𝑓𝑓𝑤𝑤 𝑞𝑞𝑜𝑜 𝑞𝑞𝑤𝑤 𝑞𝑞𝑡𝑡 𝑄𝑄𝑖𝑖 𝑡𝑡

𝑆𝑆𝑤𝑤 𝑓𝑓𝑤𝑤 � WOR 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑊𝑊𝑊𝑊 𝑊𝑊𝑖𝑖𝑖𝑖𝑖𝑖 (bbls)
𝑑𝑑𝑆𝑆𝑤𝑤 𝑆𝑆 � � � � � � (PV) (days)
𝑤𝑤
𝑑𝑑𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑𝑑𝑑
0.520 0.660 2.063 1.883 6.524 × 102 1.228 × 103 1.880 × 103 0.653 0.485 3.048 × 106 1.589 × 103
0.720 0.904 1.305 9.097 1.849 × 102 1.682 × 103 1.866 × 103 0.901 0.766 4.817 × 106 2.511 × 103
0.748 0.959 1.285 22.880 78.031 1.785 × 103 1.863 × 103 0.958 0.778 4.892 × 106 2.552 × 103
3 3 6
0.750 0.962 1.209 24.438 73.246 1.790 × 10 1.863 × 10 0.961 0.827 5.198 × 10 2.712 × 103
0.770 0.980 0.650 47.226 38.614 1.824 × 103 1.862 × 103 0.979 1.540 9.678 × 106 5.049 × 103
0.775 0.983 0.554 55.682 32.851 1.829 × 103 1.862 × 103 0.982 1.804 1.134 × 107 5.916 × 103
3 3 7
0.778 0.984 0.504 61.465 29.808 1.832 × 10 1.861 × 10 0.984 1.985 1.247 × 10 6.510 × 103
0.779 0.985 0.488 63.524 28.856 1.833 × 103 1.861 × 103 0.985 2.049 1.288 × 107 6.721 × 103
The positions of the leading shock front ( 𝑋𝑋𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 ) and the trailing shock front ( 𝑋𝑋𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 ) and any water
saturation (𝑆𝑆𝑤𝑤 > 𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 ) at any time before their breakthrough are given by the following equations:
𝑞𝑞𝑖𝑖 𝑡𝑡
𝑑𝑑𝑆𝑆𝑤𝑤 �𝑆𝑆
𝑋𝑋𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 = … … … … … . (3.29)
𝐴𝐴 𝜙𝜙
𝑞𝑞𝑖𝑖 𝑡𝑡 �
𝑋𝑋𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 = … … … … . . (3.30)
𝐴𝐴 𝜙𝜙
𝑞𝑞𝑖𝑖 𝑡𝑡
𝑤𝑤
𝑋𝑋𝑆𝑆𝑤𝑤 = … … … … . (3.31)
𝐴𝐴 𝜙𝜙
where:
A is the model’s cross sectional area given by multiplication of width by thickness (L2)
𝜙𝜙 is the porosity
�
is estimated from the tangent line at this 𝑆𝑆𝑤𝑤 .
𝑤𝑤
69
The water saturation profile at 0.2 PV of immediate injection of low salinity water is shown in Figure 3.6
and compared with the water saturation profile of immediate high salinity waterflooding.
1 1
0.9 0.9
0.8 0.8
Water saturation
0.7
Produced water cut
0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0 0 200 400 600 800 1000 1200
0 0.5 1 1.5 2 Distance (m)
PVI
low salinity waterflooding high salinity waterflooding low salinity waterflooding high salinity waterflooding
Figure 3.5: The analytical produced water cut by the immediate Figure 3.6: The analytical water saturation profile at 2 PV of immediate
injection of low salinity water compared with that by immediate high injection of low salinity water compared with high salinity
salinity waterflooding. Water breaks through earlier using high salinity waterflooding. Water breaks through earlier in the high salinity
waterflooding compared with low salinity waterflooding. In addition, waterflooding compared with low salinity waterflooding.
the producing water cut at the breakthrough is higher using high salinity
waterflooding compared with low salinity waterflooding. For low
salinity waterflooding, the producing water cut stays constant during
the production of the connate water bank between the injection of 0.46
PVI and 0.73 PVI.
Comparison between the Mobility Ratios of High Salinity and Low Salinity Waterflooding
The mobility ratios at the shock front(s) are compared for high salinity waterflooding (it has only one shock
front 𝑆𝑆𝑤𝑤,ℎ ) and low salinity waterflooding (two shock fronts). The water saturation for each shock front for
both type of waterflooding are shown in Table 3.3.
Table 3.3: Comparison between the shock fronts of high salinity and low salinity waterflooding.
High salinity waterflooding Low salinity waterflooding
Shock front Water saturation of the shock front Shock front Water saturation of the shock front
Leading shock front (𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 ) 0.52

𝑆𝑆𝑤𝑤,ℎ 0.58
Trailing shock front (𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 ) 0.72
High Salinity Waterflooding

Shock front 𝑆𝑆𝑤𝑤,ℎ mobility ratio (M):
𝑘𝑘𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,ℎ � 𝑘𝑘𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,ℎ � 0.029 0.239
+ +
𝑀𝑀𝑠𝑠 =
𝜇𝜇0 𝜇𝜇𝑤𝑤
= 0.47 0.5 = 0.062 + 0.478 = 0.255
𝑘𝑘𝑟𝑟𝑟𝑟 (𝑆𝑆𝑤𝑤𝑤𝑤 ) 1 2.128
𝜇𝜇0 0.47
Low Salinity Waterflooding
Leading shock front 𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 mobility ratio:
𝑘𝑘𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 � 𝑘𝑘𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 � 0.073 0.169
+ +
𝑀𝑀𝑠𝑠 =
= 0.47 0.5 = 0.155 + 0.338 = 0.232
(𝑆𝑆
𝑘𝑘𝑟𝑟𝑟𝑟 𝑤𝑤𝑤𝑤 ) 1 2.128
𝜇𝜇0 0.47
70
Trailing front 𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 mobility ratio:
𝑘𝑘𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 � 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 � 0.013 0.2
+ + 0.027 + 0.400
𝜇𝜇𝑜𝑜 𝜇𝜇𝑤𝑤
𝑀𝑀𝑠𝑠 = = 0.47 0.5 = = 0.877
𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 � 𝑘𝑘𝑟𝑟0 �𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 � 0.169 0.07 0.338 + 0.149
+ +
𝜇𝜇𝑤𝑤 𝜇𝜇0 0.5 0.47
The leading shock front for low salinity waterflooding has almost the same mobility ratio as the high salinity
water front. However, the trailing shock front of low salinity waterflooding has a much higher mobility ratio
than the others, which indicates it has a less favourable displacement.
Injection of Low Salinity Water post High Salinity Waterflooding
In this analysis, it is assumed that the reservoir has been waterflooded by high salinity water down to
residual oil saturation before the injection of low salinity water. The example described in the previous
section is used here to illustrate the analytical Buckley Leverett solution of injecting low salinity water post
high salinity waterflooding. Two shock fronts form as shown in Figure 3.8. Ahead of the leading shock front
is the region of residual oil saturation (Sor) and behind the trailing shock front is the region with low salinity
water. In between these two shock fronts, an oil bank is formed. The fractional flow curves for injecting low
salinity water post high salinity waterflooding is shown in Figure 3.7. A tangent line is drawn from the (0,0)
point to the low salinity fractional flow curve (LS) and the trailing shock front water saturation is determined
(Sw,LS). When this line crosses the fractional flow curve for the high salinity water, the water saturation of the
oil bank is determined (Sw,HS). The velocity of the trailing shock front can be estimated from the slope of this
tangent line. The velocity of the leading shock front is determined from the slope of the line that connects the
residual oil saturation (Sor) and the water saturation of the oil bank ( Sw,HS )
The slopes of the leading shock front and trailing shock front are calculated as follows:
𝑑𝑑𝑓𝑓𝑤𝑤 𝑓𝑓𝑤𝑤,𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑓𝑓𝑤𝑤,𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 1 − 0.68

� = = = 1.36
𝑑𝑑𝑆𝑆𝑤𝑤 𝑆𝑆 𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 0.75 − 0.52
𝑑𝑑𝑓𝑓𝑤𝑤 𝑓𝑓𝑤𝑤,𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 0.95

� = = = 1.32
𝑑𝑑𝑆𝑆𝑤𝑤 𝑆𝑆 𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 0.72
The times of the leading shock front and the trailing shock front breakthrough in pore volume (PV) are given
by:
1 1
𝑡𝑡𝐵𝐵𝐵𝐵,𝐻𝐻𝐻𝐻 = = = 0.73 𝑃𝑃𝑃𝑃
𝑑𝑑𝑓𝑓𝑤𝑤 1.36
1 1
𝑡𝑡𝐵𝐵𝐵𝐵,𝐿𝐿𝐿𝐿 = = = 0.76 𝑃𝑃𝑃𝑃
𝑑𝑑𝑓𝑓𝑤𝑤 1.32
71
Sor = 0.75
1
0.9
Sw, LS = 0.72
0.8
Sw, HS = 0.52
0.7
0.6
fw
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Sw
high salinity waterflooding low salinity waterflooding
Figure 3.7: Fractional flow curves for injecting low salinity water post high salinity waterflooding. A tangent line is drawn from the (0,0) point to the
low salinity fractional flow curve (LS) and the trailing shock front water saturation is determined ( Sw,LS ). When this line crosses the fractional flow
curve for the high salinity water, the water saturation of the oil bank is determined (Sw,HS ). The velocity of the trailing shock front can be estimated
from the slope of this tangent line. The velocity of the leading shock front is determined from the slope of the line that connects the residual oil
saturation (Sor) and the water saturation of the oil bank ( Sw,HS ).
0.9
0.8 Sw, LS Sor
0.7
Water saturation
0.6
0.5
Sw, HS
0.4
0.3 Oil Bank
0.2
0.1
0
0 200 400 600 800 1000
Distance (m)
Figure 3.8: The water saturation profile for injecting low salinity water post high salinity waterflooding at 0.6 PVI. Two shock fronts form: ahead of
the leading shock front is the region of residual oil saturation (Sor) and behind the trailing shock front is the region with low salinity water. In between
these two shock fronts, an oil bank is formed.
72
Mobility ratios of the two shock fronts
Leading shock front 𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 mobility ratio:
𝑘𝑘𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 � 𝑘𝑘𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 � 0.073 0.169
+ +
𝑀𝑀𝑠𝑠 =
= 0.47 0.5 = 0.155 + 0.338 = 0.493
𝑘𝑘𝑟𝑟𝑟𝑟 (𝑆𝑆𝑜𝑜𝑜𝑜 ) 0.5 1
𝜇𝜇𝑤𝑤 0.5
Trailing front 𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 mobility ratio:

𝑘𝑘𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 � 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 � 0.013 0.2
+ + 0.027 + 0.400
𝜇𝜇𝑜𝑜 𝜇𝜇𝑤𝑤
𝑀𝑀𝑠𝑠 = = 0.47 0.5 = = 0.877
𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 �𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 � 𝑘𝑘𝑟𝑟0 �𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 � 0.169 0.07 0.338 + 0.149
+ +
𝜇𝜇𝑤𝑤 𝜇𝜇0 0.5 0.47
The trailing shock front of low salinity waterflooding has a much higher mobility ratio than the leading
shock front which indicates it has a less favourable displacement.
3.3 Method
3.3.1 Grid Refinement Study and Validation
The performance of low salinity waterflooding is affected by the number of grid blocks used in the
simulation. Therefore, the aim of this section is to investigate this effect excluding the modelling of physical
dispersion and determine the smallest number of grid blocks that minimizes the numerical dispersion and is
close to the analytical solution. The ECLIPSE simulator that utilizes the salinity threshold model was used
for this study.
A simple 1D homogenous model was used for this study with dimensions of 1,020 × 100 × 50 m. The
reservoir consists of one injector and one oil producer 1 km apart. The injected water salinity was 1,000 ppm
and the injection rate was set at a constant rate of 304 sm3/d. The production rate was set at a constant rate of
304 sm3/d. The connate water salinity was 30,000 ppm. The salinity thresholds were set to be midway
between the connate water salinity and the injected salinity (i.e. 14,000-15,000 ppm) to minimize the linear
interpolation effects of relative permeability of intermediate salinities and reproduce the analytical solution
of low salinity waterflooding that assumes the instantaneous transition from high salinity water relative
permeability curves to low salinity water relative curves. In this study, six different numbers of grid blocks
were used as listed in Table 3.4.
Table 3.4: Scenarios of number of grid blocks used for the grid refinement study.
Scenario # Number of grid blocks
1 10
2 26
3 60
4 98
5 194
6 1000
7 98 with high salinity waterflooding
73
3.3.2 Validation of the Different Simulators to Model Low Salinity Waterflooding
As described in chapter 2, there are two main methods to model low salinity waterflooding. The first one
models low salinity waterflooding using the salinity thresholds to modify the rock’s permeability curves
with the salinity of the injected brine whilst treating salinity as a simple tracer (Jerauld et al. 2008) while the
other one models the ion exchange between the clay surface and the injected brine’s divalent ions using the
resulting divalent ion amounts to calculate the relative permeability (Dang et al. 2013). Examples of
commercial simulators that utilize the first model are ECLIPSE and CMG STARS while CMG GEM utilizes
the second model.
Both models assume that low salinity water injection improves recovery by making the reservoir more water
wet. Both represent this by using more water wet relative permeability curves as salinity is reduced. The user
inputs a low salinity pair of relative permeability curves (more water wet) and a high salinity pair of relative
permeability curves (more oil wet). In the simple threshold model, proposed by Jerauld et al. (2008), salinity
is tracked through the reservoir, like a simple tracer. In the more complex model, the salinity of the water
changes according to a set of ion exchange reactions. In the model proposed by Jerauld et al. (2008), the
salinity is used directly to calculate the relative permeability whereas in the ion exchange model, the divalent
ion content is used to determine the relative permeability
The 1D model described in the grid refinement study (section 3.3.1) was used here. It was discretized into
1000 horizontal grid blocks.
In ECLIPSE and CMG STARS (both use the simple salinity threshold model, equations 2.1-2.5 on page 26)
the low and high salinity thresholds were set to 14,000 ppm and 15,000 ppm to minimize the linear
interpolation effects of relative permeability of intermediate salinities and reproduce the analytical solution
of low salinity waterflooding that assumes the instantaneous transition from high salinity water relative
permeability curves to low salinity water relative permeability curves respectively.
In CMG GEM (which uses the ion exchange model) the ion concentrations for both high and low salinity
water are listed in Table 3.5. The modelling of low salinity waterflooding depends on the cation exchange
capacity (CEC) and the interpolation parameter for changing the relative permeability curves. The closest
scenario of modelling low salinity waterflooding using CMG GEM to the analytical Buckley Leverett
solution is the scenario that uses the equivalent fraction of Na-X as the interpolant with values shown in
Table 3.6 (this will be shown later in section 3.4.3 B).
Table 3.5: Ion concentrations for the connate water and the injected low salinity water used in CMG GEM.
Property Na+ Ca++ Mg++ Cl- Salinity (1,000 ppm)
Connate water (mol/l) 0.0132 1.4× 10-3 1.74× 10-4 0.0167 30
Low salinity water (mol/l) 4.2 × 10-4 4.5 × 10-5 5.5 × 10-6 5.31× 10-4 1
Table 3.6: The interpolation parameter values for changing from high salinity water relative permeability curve to low salinity relative permeability
curve in CMG GEM.
Property Value
Na-X equivalent fraction for the high salinity relative permeability curve (mol/l) 0.0132
Na-X equivalent fraction for the low salinity relative permeability curve (mol/l) 0.009
Cation exchange capacity (CEC) (equivalent/m3) 5
74
3.3.3 Impact of Low Salinity Waterflooding Model on Predictions
Modelling low salinity waterflooding is affected by different modelling parameters in the salinity threshold
model and the ion exchange model.
In the salinity threshold model, the choice of the salinity thresholds affect the performance of low salinity
waterflooding. In the ion exchange model, the performance of low salinity waterflooding can be affected by
the choice of the interpolation parameter, the values of the interpolation parameter to change between the
relative permeability curves and the cation exchange capacity.
In this section, the impacts of these different parameters on modelling low salinity waterflooding are
investigated and compared with the analytical Buckley Leverett solution of low salinity waterflooding using
a 1D homogeneous model (previously described in section 3.1).
A-Salinity Threshold Model
The relative permeability values are linearly dependent on the salinity values within the salinity thresholds.
For this section, the effect of the salinity thresholds on low salinity waterflooding is investigated. Six
different salinity thresholds were tested in this evaluation with 1000 grid blocks as summarized in Table 3.7.
In each scenario (except scenario 4), the injected water salinity was 1,000 ppm and the connate water salinity
was 30,000 ppm.
Table 3.7: Summary of different salinity thresholds used to evaluate their impacts on low salinity waterflooding.
Scenario # Lower salinity threshold (1,000 ppm) Upper salinity threshold (1,000 ppm)
1 1 7
2 1 1.5
3 14 15
4 14 15 (with injected salinity water of 13.5)
5 15 29.5
6 29 29.5
B-Ion Exchange Model

Modelling low salinity waterflooding depends on the choice of the interpolation parameter, the values to
change from the high salinity water relative permeability curve to the low salinity water relative permeability
curve and the cation exchange capacity.
The following two chemical reactions are thought to occur during low salinity waterflooding (Dang et al.
2013):
1 1
𝑁𝑁𝑁𝑁+ +
(𝐶𝐶𝐶𝐶 − 𝑋𝑋2 ) ↔ (𝑁𝑁𝑁𝑁 − 𝑋𝑋) + 𝐶𝐶𝐶𝐶++
2 2
1 1
𝑁𝑁𝑁𝑁 + + (𝑀𝑀𝑀𝑀 − 𝑋𝑋2 ) ↔ (𝑁𝑁𝑁𝑁 − 𝑋𝑋) + 𝑀𝑀𝑀𝑀++
2 2
where: X is the clay mineral available in the sandstone rock. It has subscript of 2 since it is attached with
divalent positive ion (𝐶𝐶𝐶𝐶++ or 𝑀𝑀𝑀𝑀++) in the clay, while it is 1 when attached with single positive charge
(𝑁𝑁𝑁𝑁+ ).
The interpolation parameter for the relative permeability curve can be either the mole concentration of the
ions (for example Ca++) or the equivalent fraction of ions that are adsorbed by the mineral surface (for
example Na-X). Table 3.8 shows the seven scenarios tested in modelling low salinity waterflooding using the
75
ion exchange model. The values in Table 3.8 for the high salinity water relative permeability curves and the
low salinity water relative curves correspond to the mole concentrations of either 𝑁𝑁𝑁𝑁+ and Ca++ for the high
salinity water and low salinity water as listed in Table 3.5. However, in scenarios 3, 6 and 7 in Table 3.8, the
values for low salinity water relative permeability curves were set to be greater than the mole fractions of
low salinity water. The purpose was to test how they affect the modelling of low salinity waterflooding when
compared to the analytical Buckley Leverett solution of low salinity waterflooding.
For any mole concentrations that fall in between the values of high salinity water and low salinity water in
Table 3.8, their relative permeability curves are linearly interpolated between the high and low salinity water
relative permeability values.
Table 3.8: Different scenarios tested for modelling low salinity waterflooding using the ion exchange model (see equations 2.1-2.5 on page 26 and the
discussion on page 27 on chapter 2)
Values to change between the relative permeability curves Cation exchange

Interpolation
Scenario High salinity water relative Low salinity water relative capacity
parameter
permeability curve permeability curve (equivalent/m3)
1 Ca++ 0.0014 0.0000445 50
2 concentration 0.0014 0.0000445 0.5
3 (mol/l) 0.0014 0.0009 0.5
4 0.0132 0.000419 50
5 Equivalent 0.0132 0.00419 50
fraction of Na-X
6 (mol/l) 0.0132 0.009 50
7 0.0132 0.009 5
3.3.4 Modelling of Diffusion

As described in chapter 2, mixing can affect low salinity waterflooding by increasing the salinity of the low
salinity water and decreasing the salinity in the connate water bank. This may alter the performance of low
salinity waterflooding. In general, mixing results from molecular diffusion and velocity dependent
dispersion. In this chapter, velocity dependent dispersion is ignored. The CMG STARS simulator was
selected for this study as molecular diffusion can be explicitly modelled. However, modelling physical
diffusion can be affected by numerical diffusion that arises from the discretisation of the flow equations
(Lantz 1971). Numerical diffusion affects the results in both directions: parallel to flow (longitudinally) and
perpendicular to flow (transversely). Therefore, it was essential to a) validate the formulation used to model
diffusion in the numerical simulator and b) ensure that the simulations were dominated by physical rather
than numerical diffusion. This was achieved by performing grid refinement studies of longitudinal and
transverse diffusivity between low salinity water and high salinity water using the chosen simulator (CMG
STARS). In this section, longitudinal diffusion and transverse diffusion were evaluated.
A-Longitudinal Diffusion
This study used a series of 1D simulations of low salinity water displacing high salinity water. There was no
oil in this model as the system was initially completely filled with high salinity water. The physical
dimensions and properties of the model are given in Table 3.9. The injection and production rates were set to
be equal to 60 cm3/d.
76
In the analysis of mixing, the dimensionless Peclet number (Pe) was used to quantify the relative importance
of transport by advection to that by diffusion or dispersion:
𝑣𝑣𝑣𝑣
𝑃𝑃𝑃𝑃 = … … … … … … … . (3.32)
𝐾𝐾
where L is a characteristic length (in this chapter, the distance between the injection and production wells)
and K is the diffusivity or dispersivity (L2/t). The higher the Pe then the greater is the impact of transport by
advection. In the simulation model, the salt diffusivity was varied to achieve Pe between 1 and 20,000.
Different horizontal grid resolutions were used to study the impact of numerical diffusion on the predictions
of longitudinal physical diffusion: 10, 50, 100, 200, 400 and 1000 horizontal grid blocks. A schematic of the
model with 50 grid blocks is shown in Figure 3.9.
Table 3.9: Physical dimensions and properties of the 1D simulation
model used to evaluate the longitudinal diffusion in single phase flow.
Property Value
Model dimensions (cm) 30 × 3.2 × 3.2
Permeability (md) 100
Porosity 0.2
H2O: 0.9
High salinity water mole fractions
Salt: 0.1
H2O: 0.99
Low salinity water mole fractions
Salt: 0.01
Figure 3.9: Schematic of 1D simulation model used to evaluate the
longitudinal diffusion in a single phase flow. This study used different
horizontal grid resolutions to study the impact of numerical diffusion on
the predictions of longitudinal physical diffusion. This figure shows the
model when it is discretized into 50 grid blocks. The model dimensions
are 30 × 3.2 × 3.2 cm. Low salinity water is injected along the left hand
side of the model and then produced with the combined high salinity
water along the right hand side of the model.
For each simulation, the Pe based on input dispersivity was compared with the Pe evaluated from the
produced salt concentration versus time using the method of Brigham (1974):
2
2.38
𝑃𝑃𝑃𝑃𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = � � … … … … … . . (3.33)
𝑈𝑈80 − 𝑈𝑈20
where
𝑉𝑉𝑖𝑖
−1
𝑈𝑈 = 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 … … … … . . (3.34)
� 𝑉𝑉𝑖𝑖
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉
and Vi is the pore volume injected and U80 and U20 were calculated at 80% and 20% concentrations of
produced low salinity water, respectively.
B-Transverse Diffusion
The impact of grid resolution on the ability of the simulator to model transverse diffusion was investigated
using a 2D model with the flow configuration shown in Figure 3.10. This mimics the setup used in
laboratory experiments to measure transverse diffusion and dispersion (see for example Blackwell, 1962,
Alkindi et al. 2011). The physical size of the simulated domain, its permeability and porosity were the same
as those used in modelling longitudinal diffusion (Table 3.9). The salinities of the initial water and the
77
injected water are the same as those used in modelling longitudinal diffusion (Table 3.9). The ratio of
transverse diffusivity to velocity (KT / v) was varied as listed in Table 3.10.
Table 3.10: KT / v values used in 2D simulation model to evaluate the impact of truncation error on the simulation of transverse diffusion.
Scenario # KT / v (cm)
1 1.71×10-4
2 1.71×10-3
3 3.8×10-3
4 7.6×10-3
5 1.24×10-2
6 2.09×10-2
7 5.90×10-2
8 0.209
Figure 3.10: Schematic of the system used to evaluate the impact of truncation error on the simulation of transverse diffusion. Low and high
salinity water are injected simultaneously, at the same rate and in parallel into a two dimensional system. Salt diffuses into the low salinity water
across the interface between the two fluids. The width of the mixing zone at the outlet of the system can then be used to calculate the transverse
diffusion.
Low salinity water was injected along the top half of the inlet face at a constant rate of 30 cm3/d and high
salinity water was injected at the same rate into the bottom half of the inlet face. Two production wells were
completed in the opposite face, one in the top half and one in the bottom half model. Salt then diffused
transversely across the interface between the two fluids. This resulted in the production from the top well
containing a higher salinity than that in the low salinity water whilst the produced fluid from the bottom
stream contained less salt than the injected high salinity water stream. When the concentrations of the
produced streams had stabilized (steady state flow), each concentration was used to back calculate KT using
the analytical solution given in the following equation (Alkindi et al. 2011):
∞
1 2 𝑛𝑛2 𝜋𝜋 2 𝐿𝐿
−𝐾𝐾𝑇𝑇 2
𝐹𝐹1 = 𝛼𝛼 + 2 � 2 sin(𝑛𝑛𝑛𝑛𝑛𝑛) sin(𝑛𝑛𝑛𝑛𝑛𝑛) 𝑒𝑒 𝑤𝑤 𝑣𝑣 … … … (3.35)
𝜋𝜋 𝛽𝛽 𝑛𝑛
𝑛𝑛=1
where:
F1 is the stabilized concentration from producer 1
α is the fraction of the width where injector 1 is completed, in this model it is 0.5
β is the fraction of the width where producer 1 is completed, in this model it is 0.5
L and w are the system length and width, respectively (L). This was compared with the input KT.
78
3.4 Results
3.4.1 Grid Refinement Study and Validation
The water saturation profile at 0.2 PVI for the different scenarios are plotted in Figure 3.11 and compared
with the analytical Buckley Leverett solution for low salinity waterflooding. This plot shows that very fine
grid (1000 grid blocks) reproduces very close water saturation profile to the analytical solution. Coarsening
the simulation model results in incorrect predictions of earlier water breakthrough of the leading shock front
and later arrival of trailing shock front. The effect of numerical dispersion on the 10 grid block scenario is so
large that it results in one shock front.
1
0.9
0.8
0.7
Water saturation
analytical solution
0.6 10 grid blocks
0.5 26 grid blocks
0.4 50 grid blocks
0.3 98 grid blocks

194 grid blocks
0.2
1000 grid blocks
0.1
0
0 200 400 600 800 1000 1200
Distance (m)
Figure 3.11: Water saturation profile for the different number of grid blocks compared with the analytical solution. The effect of numerical dispersion
on the 10 grid block scenario is very large that it results in there being only one shock front compared with the two shock fronts predicted by the
analytical solution for low salinity waterflooding. The numerical dispersion effects on coarse grid blocks (10 grid blocks) results in incorrect
predictions of earlier water breakthrough of the leading shock front and later arrival of trailing shock front.
To estimate the minimum number of grid blocks that is required to reproduce the analytical solution to
within 1% error, the average difference of the magnitude in the water saturation profile between the
analytical solution and the simulation model is calculated for each scenario and plotted as a log-log plot in
Figure 3.12. This figure suggests that a very fine grid (1000 grid blocks) is required to reproduce the
analytical solution to within 1% error. Figure 3.13 shows that the slope of the line is less than 1 which
suggests that the order of convergence is a first order.
79
100 1.6
1.4 y = 0.5631x + 0.1666

R² = 0.9905
1.2
10
1
log (error)
Error (%)
0.8
1 0.6
0.4
0.2
0.1
0
1 10 100 1000
0 0.5 1 1.5 2 2.5
Number of grid blocks log (Δx)
Figure 3.12: The error difference in reproducing the analytical solution of Figure 3.13: The log of error difference in reproducing the analytical
low salinity waterflooding by the different number of grid blocks. A very solution of low salinity waterflooding by the different number of grid
fine grid (1000 grid blocks) is required to reproduce the analytical solution blocks against the log of the grid block size. The order of convergence
to within 1% error. is a first order since the slope of the line is less than 1.
The water cut versus PVI and the produced water salinity are plotted for the different grid refinements in
Figure 3.14. The water cut plots are also compared with the analytical Buckley-Leverett solution for low
salinity waterflooding. It can be seen that the water saturation in the connate water bank fluctuates
unphysically for the coarsest grid simulations. This is probably due to the changes in salinity causing rapid
changes in relative permeability and affecting production as a) this is not seen in the plots of water saturation
versus distance before water breakthrough and b) these fluctuations occur roughly when the produced
salinity drops to around the threshold values. The reason behind these fluctuations in water cut is an
unphysical instability. Similar fluctuations are observed using STARS simulator (Figure 3.15).
80
1 35
0.9
30
Produced water salinity (1,000 ppm)

0.8
0.7 25
0.6
20
Water cut
0.5
15
0.4
0.3 10
0.2
5
0.1
0 0
0 0.3 0.6 PVI 0.9 1.2 1.5 0 0.3 0.6 PVI 0.9 1.2 1.5
analytical solution 10 grid blocks analytical solution 10 grid blocks
26 grid blocks 50 grid blocks 26 grid blocks 50 grid blocks
98 grid blocks 98 grid blocks
Figure 3.14: a) water cut vs PVI b) produced water salinity vs PVI

The water cut versus PVI and the produced water salinity for different grid refinements compared with the analytical Buckley-Leverett
solution for low salinity waterflooding. It can be seen that the water saturation in the connate water bank fluctuates unphysically for the
coarsest grid simulations. This is probably due to the changes in salinity causing rapid changes in relative permeability and affecting
production as a) this is not seen in the plots of water saturation versus distance before water breakthrough and b) these fluctuations occur
roughly when the produced salinity drops to around the threshold values. The reason behind these fluctuations in water cut is an unphysical
instability.
1 35
0.9
30
Produced water salinity (1,000 ppm)
0.8
0.7 25
0.6
20
Water cut
0.5
0.4 15
0.3 10
0.2
5
0.1
0 0
0 0.3 0.6 0.9 1.2 1.5 0 0.3 0.6 0.9 1.2 1.5
PVI PVI
analytical solution 10 grid blocks analytical solution 10 grid blocks
26 grid blocks 50 grid blocks 26 grid blocks 50 grid blocks
98 grid blocks 98 grid blocks
Figure 3.15: a) water cut vs PVI b) produced water salinity vs PVI

The water cut versus PVI and the produced water salinity for different grid refinements as predicted by STARS compared with the analytical
Buckley-Leverett solution for low salinity waterflooding. The water cut predictions are similar to those by ECLIPSE.
81
3.4.2 Validation of the Different Simulators to Model Low Salinity Waterflooding
Figure 3.16 compares the water saturation profile at 0.2 PVI of immediate low salinity waterflooding as
predicted by ECLIPSE, CMG GEM and CMG STARS simulators compared with the analytical Buckley
Leverett solution. CMG STARS and ECLIPSE simulators show excellent match with the analtycai solution.
On the other hand, CMG GEM shows excellent match with the analytical solution for the leading shock front
only. However, there are two main differences between CMG GEM and the analytical solution: 1) the water
saturation profile behind the trailing shock front does not increase smoothly towards the injection well in the
GEM simulations results. 2) The connate water bank saturation is slightly higher than the BL solution (~ 3%
higher) and also the position of the trailing shock front is 9 meters behind that for the BL solution (the
trailing shock front speed as predicted by GEM is slower than that for the BL solution). The wiggles in the
water saturation profiles of CMG GEM are also present in results presented by CMG so are presumably a
'feature' of the formulation in GEM whilst the mismatch between the connate water bank saturation and the
speed of the trailing shock fronts is related to the difference in the modelled oil viscosities in the BL and
GEM. The fractional flow curves depend on the ratio of the water viscosity to the oil viscosity. In the ion
exchange modelling in GEM, the oil viscosity is computed through equations that involve several
parameters. The initial oil viscosity based on the mixture of hydrocarbons in the model was found to be
around 0.21 cp. After running the model to 0.2 PVI, the oil viscosity was increased to around 0.26 cp (close
to the injector) decreasing with distance to 0.24 cp (close to the producer).
Since both methods generate similar results, the salinity threshold model (ECLIPSE and CMG STARS) was
selected to be the main modelling method for the remainder of this thesis as its simulation running time is
significantly faster than the ion exchange model.
1
0.9
0.8
0.7
Water saturation

0.5 CMG STARS
0.4 ECLIPSE
0.3 CMG GEM
0.2
0.1
0
0 200 400 600 800 1000 1200
Distance (m)
Figure 3.16: Water saturation profiles at 0.2 PVI as predicted by ECLIPSE, CMG GEM and CMG STARS for immediate injection of low salinity water
into a 1D homogeneous model (discretized into 1000 grid blocks) compared with the analytical Buckley Leverett solution for low salinity
waterflooding. ECLIPSE and CMG STARS show excellent match with the analytical solution. On the other hand, CMG GEM shows excellent match
with the analytical solution for the leading shock front only. However, there are two main differences between CMG GEM and the analytical
solution: 1) the water saturation profile behind the trailing shock front does not increase smoothly towards the injection well in the GEM
simulations results. 2) The connate water bank saturation is slightly higher than the BL solution (~ 3% higher) and also the position of the trailing
shock front is 9 meters behind that for the BL solution (the trailing shock front speed as predicted by GEM is slower than that for the BL
solution).
82
3.4.3 Impact of Low Salinity Waterflooding Model on Predictions
A- Salinity Threshold Model
The water saturation profile at 0.4 PV is shown in Figure 3.17 for different salinity thresholds. The (1,000-
1,500 ppm) case has the slowest trailing shock front compared with the analytical solution while it has the
fastest leading shock front. The range between its thresholds is narrow and close to the injected water
salinity of 1,000 ppm. On other the hand, the (29,000-29,500 ppm) case has the fastest trailing shock front
and the slowest leading shock front. The (14,000-15,000 ppm) case is the closest one to the analytical
solution. Due to numerical dispersion, the shock fronts for all cases are not vertical.
1
0.9
0.8
0.7
Water saturation
0.6
0.5
0.4
0.3
0.2
0.1
0
0 200 400 600 800 1000 1200
Distance (m)
1,000-7,000 ppm 1,000-1,500 ppm
14,000-15,000 ppm (14,000-15,000 ppm inj at 13,500 ppm
15,000-29,500 ppm 29,000-29,500 ppm
analytical solution
Figure 3.17: Water saturation profiles for various salinity thresholds, increasing the thresholds to be midway between the salinities of the injected
water and connate water results in the closest numerical predictions compared with the analytical solution.
Producing water cuts versus pore volume water injected are plotted in Figure 3.18. The leading shock front
breaks through for the (1,000-1,500 ppm) case earlier than the analytical solution. For the (29,000-29,500
ppm) case, the leading shock front breaks through last compared to other cases. This case also has the lowest
producing water cut of all the cases. The (1,000-1,500 ppm) case has the highest producing water cut at
breakthrough. The trailing shock front breaks through first in the (29,000-29,500 ppm) case while it breaks
through last in the (1,000-1,500 ppm) case. The closest case to the analytical solution is the (14,000-15,000
ppm) case.
Mixing between low salinity water and connate water (connate water bank) occurs in all cases but with
different degrees of mixing and connate water bank size. In the (29,000-29,500 ppm) case, the arrival of the
second shock front is way before all the other cases and also the arrival of the first shock front is slightly
later than all cases. Therefore, it is the case least affected by the connate water bank.
The least affected case to the numerical dispersion is (14,000-15,000 ppm) case (i.e. the case where the
salinity thresholds are midway between the salinities of the connate water and injected water).
From Figure 3.19, the oil recoveries for the six cases vary up to 5 % during the production of the connate
water bank (0.4 – 0.9 PV), while they are the same for the rest of the simulation. It should be noted that the
83
(29,000-29,500) ppm case has the highest oil recovery during this period as it is the least affected by the
mixing zone.
0.9
0.8
0.7
1,000-7,000 ppm
0.6
1,000-1,500 ppm
Water cut
0.5
14,000-15,000 ppm
0.4 14,000-15,000 ppm inj at

13,500 ppm
0.3 15,000-29,500 ppm
0.2 29,000-29,500 ppm
0
0 0.5 1 1.5 2 2.5
PVI
Figure 3.18: Water cut versus PVI for various salinity thresholds, the water breakthrough from analytical solution is earlier than the ones predicted
by simulation.
0.9
0.8
0.7 1,000-7,000 ppm
0.6 1,000-1,500 ppm

Oil recovery
0.5
14,000-15,000 ppm
0.4
14,000-15,000 ppm inj
13,500 ppm
0.3
15,000-29,500 ppm
0.2
29,000-29,500 ppm
0.1
0
0 0.5 1 1.5 2 2.5
PVI
Figure 3.19: Oil recovery versus PVI for various salinity thresholds, the only variations in oil recoveries between all cases is during the production
of the connate water bank (0.4-0.9 PV).
84
B- Ion Exchange Model
The water saturation profile at 0.2 PV of immediate injection of low salinity water for the different scenarios
compared with the analytical Buckley Leverett solution are shown in Figure 3.20.
When setting the Ca++ concentration as the interpolation parameter for the relative permeability curves, the
simulator predicts a later arrival of the trailing shock front compared with the analytical solution (i.e. a larger
connate water bank) with a higher water saturation in the connate water bank. When setting the equivalent
fraction of Na-X as the interpolation parameter, the simulation predictions are closer to the analytical
solution especially for the trailing shock front. This is similar to the findings by Dang et al (2015). The
simulation predictions are even closer to the analytical solution when the cation exchange capacity (CEC) is
reduced from 50 (equivalent/m3) to 5 (equivalent/m3). Therefore, as the amount of ions that are adsorbed by
the clay is reduced, the ion exchange model predictions are closer to the analytical solution.
The values to change from the high salinity water relative permeability curve to the low salinity water
relative permeability curve have a minor effect on modelling low salinity waterflooding compared with the
choice of the interpolation parameter and the value of cation exchange capacity.
1
0.9
0.8
0.7
Water saturation
0.6
0.5
0.4
0.3
0.2
0.1
0
0 200 400 600 800 1000 1200
Distance (m)
analytical solution Ca++, CEC 50,highsal 0.0014 lowsal 0.0000445
Ca++, CEC 0.5,highsal 0.0014, lowsal 0.0000445 Ca++, CEC 0.5,highsal 0.0014, lowsal 0.0009
Na-x, CEC 50,highsal 0.0132, lowsal 0.000419 Na-x, CEC 50,highsal 0.0132, lowsal 0.00419
Na-X, CEC 50,highsal 0.0132, lowsal 0.009 Na-x, CEC 5,highsal 0.0132, lowsal 0.009
Figure 3.20: Water saturation profile at 0.2 PV of immediate injection of low salinity water for different scenarios using the ion exchange model
compared with the analytical Buckley Leverett solution. The choice of the interpolation parameter affects the modelling of low salinity
waterflooding. When setting the Ca++ concentration as the interpolation parameter, the arrival of the trailing shock front is delayed compared to
the analytical Buckley Leverett solution (i.e. larger size of connate water bank) in addition to higher water saturation of the connate water bank.
When setting the equivalent fraction of Na-X as the interpolation parameter, the predictions are closer to the analytical solution especially for the
trailing shock front. When using low values of cation exchange capacity (CEC), the simulation predictions are closer to the analytical solution
The values to change between the relative permeability curves has a minor effect on modelling low salinity waterflooding compared to the
choice of the interpolation parameter and the cation exchange capacity.
85
3.4.4 Modelling of Diffusion
A-Longitudinal Diffusion
Figure 3.21 shows the salt concentration in the produced water for several input Pe using 1000 grid blocks.
The front between low and high salinity becomes progressively sharper as the Pe increases and the influence
of advection becomes more important than diffusion.
100
Produced salt concentration
90
80
70
60
50 Actual Peclet=1.75
40 Actual Peclet=26
30 Actual Pe=1725
20
10
0
0 0.5 1 1.5 2 2.5
PVI
Figure 3.21: Illustration of the impact of physical diffusion for a 1000 grid block simulation model in which low salinity water displaces high
salinity water. The higher the Pe, the sharper the front. At low Pe diffusion is more important.
Figure 3.22 compares the back-calculated Pe from the produced salt concentration profile with those
calculated from the input data as a function of the number of grid blocks used in the simulation. It can be
seen that at higher input Pe, the back-calculated Pe are constant irrespective of the value of the input Pe.
This is because the physical diffusion is very low and so numerical diffusion dominates over physical
diffusion for these values.
180
160
160
140 145
120 120
Back-calculated Pe
100 100
80
65
60
40
20 20
0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
Actual Pe
1000 GB 400 GB 200 GB
100 GB 50 GB 10 GB
Figure 3.22: Back-calculated Pe compared with actual Pe for different numbers of grid blocks. At high input Pe the back-calculated Pe is constant
because numerical diffusion dominates over physical diffusion.
86
Figure 3.23 shows the relationship between the input and back-calculated Pe for Pe < 70 for different grid
resolutions. It can be seen that for low Pe the back-calculated Pe is linearly related to the input Pe and that
the range of Pe over which this is true increases with the number of grid blocks used in the simulation. It is
interesting to note that there is a constant offset of ~ 2.5 between back-calculated and input Pe and that the
linear relationship does not have a gradient of 1. If there was a one-to-one correspondence between the input
and back-calculated Pe then a zero offset and a gradient of 1 would be expected. The gradient increases
towards 1 as the grid size increases but even at 1000 grid blocks it is still only 0.9. This is because the time-
step was not controlled in these simulations but the simulator (running in fully implicit mode) was allowed
to select the time-step itself. Numerical diffusion is a function of both time-step and grid discretization (see
Lantz, 1971). The 1000 grid block simulation was repeated using a constant time-step equal to that given by
the Courant-Friedrichs-Lewy (CFL) criterion and in this case the back-calculated Pe was linearly related to
the input Pe with a gradient of 1.003 although there was still a constant offset of about 2.5. It is not clear
why for this case the straight line does not start at (0,0) point. This could be due to higher order
discretization errors.
The fact that the back-calculated Pe is systematically lower than the input Pe when the time-step is not
controlled suggests that in all standard simulations there is a contribution of numerical diffusion and possibly
numerical dispersion (resulting from higher order truncation errors) to the observed mixing, even when very
fine grids are used. It appears that 200 grid blocks seems to be the smallest number of grid blocks required to
capture physical diffusion since refining the model further only results in a minor decrease in the numerical
diffusion.
87
y = 1.0196x + 1.9048
60 R² = 0.9998
y = 0.9061x + 2.6525
50 R² = 0.9977
y = 0.8974x + 2.6238
Back-calculated Pe
40 R² = 0.998
y = 0.8685x + 2.5688
R² = 0.9969
30
y = 0.8144x + 2.7622
R² = 0.9926
20
y = 0.7699x + 2.6098
R² = 0.9909
10 y = 0.505x + 2.3235
R² = 0.9437
0
0 10 20 30 40 50 60 70
Actual Pe
1000 GB 400 GB
200 GB 100 GB
50 GB 10 GB
controlled timestep (1000 GB) Linear (1000 GB)
Linear (400 GB) Linear (200 GB)
Linear (100 GB) Linear (50 GB)
Linear (10 GB ) Linear (controlled timestep (1000 GB))
Figure 3.23: The linear region of the back-calculated Pe versus actual Pe plot for different numbers of grid blocks. At low actual Pe (high
physical diffusion) there is a linear relationship between the input and back-calculated Pe. The linear relationship ends at different values of actual
Pe. This value is defined as the critical Pe. Above this value, the relationship of the calculated Pe and actual Pe changes to non-linear relationship.
This critical Pe increases as the model refines. The slope of the lines are less than 1 due to the fact that the time-step was not controlled in all the
cases except in the controlled time-step (1000 grid block case) which is labelled by yellow squares. However, there is still an offset of around 2
due to the contribution of numerical dispersion resulting from higher order truncation errors.
To ensure that the simulation model is dominated by physical diffusion rather than numerical diffusion, the
modelled diffusion should fall within the linear region in Figure 3.23. The end of the linear region for any
number of grid blocks is defined as the critical Pe. Above this critical Pe, the relationship between the
generated Pe by the simulation and the input Pe is no longer linear thus numerical diffusion dominates. This
critical Pe varies as a function of grid blocks in the model and is tabulated for the different grid sizes in
Table 3.11. Figure 3.24 plots both this critical Pe and the plateau Pe against the number of grid blocks. It can
be seen that there is a non-linear relationship between these Pe and the number of grid blocks whereas
Lantz’s analysis of numerical diffusion suggests there should be a linear relationship. It is assumed that this
non-linear relationship is because the time-step was not constant in the simulations.
Table 3.11: The critical Pe (the value of Pe above which numerical diffusion contributes significantly to the mixing) as a function of the number of
grid blocks used to discretize the model.
Number of grid blocks Critical Pe
10 12
50 26
100 35
200 42
400 50
1000 60
88
180 y = 32.49ln(x) - 56.344
160 R² = 0.9875
140
120 Critical Pe
100 Plateau
Pe y = 10.557ln(x) - 13.486
80 Log. (Critical Pe)
R² = 0.9958
60
Log. (Plateau)
40
20
0
0 200 400 600 800 1000 1200
Number of grid blocks
Figure 3.24: The critical and plateau Pe vary logarithmically with the number of grid blocks.
The simulations were repeated with a constant time-step equal to that given by the Courant-Friedrichs-Lewy
(CFL) criterion for all number of grid blocks. Figure 3.25 shows the linear region of the back-calculated Pe
versus the input Pe using a constant time-step equal to that given by the CFL criterion. The critical Pe for
different numbers of grid blocks using a constant time-step equal to that given by the CFL criterion are
plotted in Figure 3.26. The relationship between the critical Pe (using the CFL time-step) and the number of
grid blocks is linear. There is no critical Pe for the 10 grid block case as the relationship between the
calculated Pe and input Pe is a horizontal line.
60
50 y = 1.0192x + 1.92
R² = 0.9998
Back-calculated Pe
40
30
y = 0.8867x + 2.203
R² = 0.9981
20 y = 0.7628x + 2.4349
R² = 0.9773
y = 0.7057x + 1.9083
10 R² = 0.9935
y = 0.688x + 1.2781
R² = 0.9261
0
0 10 20 30 40 50 60
Actual Pe
1000 GB 400 GB 200 GB 100 GB
50 GB 10 GB Linear (1000 GB) Linear (400 GB)
Linear (200 GB) Linear (100 GB) Linear (50 GB) Linear (10 GB)
Figure 3.25: The linear region of the back-calculated Pe versus the input Pe plot for different numbers of grid blocks using a constant time-step
equal to that given by the Courant-Friedrichs-Lewy (CFL) criterion. The critical Pe (where the linear region ends) decreases as the number of grid
blocks decreases. For the 10 grid block case, there is no critical Pe as the relationship between the back-calculated Pe and the input Pe is a
horizontal line.
89
60
y = 0.0437x + 7.6221
50 R² = 0.9977
40
Critical Pe
30
20
10
0
0 200 400 600 800 1000 1200
Number of grid blocks
Figure 3.26: Critical Pe for different numbers of grid blocks using a constant time-step equal to that given by the Courant-Friedrichs-Lewy (CFL)
criterion. The relationship is linear. There is no critical Pe for the 10 grid block case as the relationship between the back-calculated Pe and the input
Pe is a horizontal line.
Figure 3.27 shows the salt concentration profiles at 1 PVI for the 10, 100 and 1000 grid blocks at different
Pe. When the Pe is greater than 59, numerical diffusion dominates the flow in the 10 grid blocks case. This
is why the salinity distributions at 1 PV for the Pe values of 59 and 1700 are similar. This is not observed in
the 100 and 1000 grid blocks. Therefore, as the horizontal grid is refined, the effect of numerical diffusion
decreases.
The producer is 30 cm away from the injector

The high salinity water mole fraction is 0.1 while the low salinity water mole fraction is 0.01
The direction of injection is from left to right
10 grid blocks 100 grid blocks 1000 grid blocks

0.07 0.07 0.07
Salt concentration (mole fraction)
0.06
0.05 0.05
0.05
0.04
0.03 0.03
0.03
0.02
0.01 0.01
0.01
0 -0.01 0 10 20 30 -0.01 0 10 20 30
0 10 20 30
Distance (cm) Distance (cm) Distance (cm)
Pe = 1700 Pe = 59 Pe = 1.7 Pe = 1700 Pe = 59 Pe = 1700 Pe = 59
Pe = 1.7 Pe = 1.7
Figure 3.27: The salt concentration profiles at 1 PVI for the 10, 100 and 1000 grid blocks at three different Pe. This figure shows that when Pe is
greater than 59, numerical diffusion dominate the flow for the 10 grid blocks, such that the salinity profiles for Pe = 59 and 1700 are almost
identical. As the grid is refined, the effect of numerical diffusion decreases.
90
%7UDQVYHUVH'LIIXVLRQ
)LJXUH FRPSDUHV WKH VDOW FRQFHQWUDWLRQ LQ WKH SURGXFWLRQ IURP WKH XSSHU ZHOO SUHGLFWHG E\ WKH
VLPXODWLRQXVLQJDîJULGZLWKWKDWREWDLQHGDQDO\WLFDOO\IURPHTXDWLRQIRUDUDQJHRIWUDQVYHUVH
GLIIXVLYLWLHV,WFDQEHVHHQWKDWWKHVLPXODWLRQUHVXOWVDUHLGHQWLFDOWRWKHDQDO\WLFDOVROXWLRQVXJJHVWLQJWKDWD
YHUWLFDO GLVFUHWL]DWLRQ RI JULG EORFNV SURYLGHV D FRQYHUJHG VROXWLRQ WR WKH PRGHOOLQJ RI WUDQVYHUVH
GLIIXVLRQ
0.95
0.9
0.85 Simulation
0.8
analytic test
0.75
)
0.7
0.65
0.6
0.55
0.5
0 0.05 0.1 0.15 0.2 0.25
.7Y

)LJXUH&RPSDULVRQIRUWKHVWDELOL]HGVDOWFRQFHQWUDWLRQIURPWKHXSSHUSURGXFHUZHOO)DQDO\WLFDOO\DQGE\WKHVLPXODWLRQPRGHOXVLQJD
îJULGIRUDUDQJHRIYDOXHVRI.7 Y&OHDUO\JULGEORFNVWUDQVYHUVHWRWKHSULQFLSDOIORZGLUHFWLRQLVVXIILFLHQWWRUHVROYHDOOOHYHOVRI
WUDQVYHUVHGLIIXVLRQVLPXODWHG
)LJXUHFRPSDUHV)DVDIXQFWLRQRI.7YSUHGLFWHGE\WKHVLPXODWRUZKHQWKHQXPEHURIJULGEORFNVLQ
WKH YHUWLFDO GLUHFWLRQ LV YDULHG IURP WR ,W FDQ EH VHHQ WKDW WKHUH LV D JRRG SUHGLFWLRQ RI WUDQVYHUVH
GLIIXVLRQIRUDUDQJHRIYDOXHVRI.7YDVORQJDVDWOHDVWJULGEORFNVDUHXVHGWRGLVFUHWL]HHDFKOD\HU
1
0.95 nz=2
0.9 nz=4
0.85
nz=6
0.8
)
0.75 nz=10
0.7 nz=20
0.65
0.6
0.55
0.5
0 0.05 0.1 0.15 0.2 0.25
.7Y

)LJXUH (IIHFW RI YHUWLFDO JULG UHVROXWLRQ RQ WKH SUHGLFWLRQ RI VDOW FRQFHQWUDWLRQ LQ WKH HIIOXHQW RI WKH XSSHU ZHOO ,W FDQ EH VHHQ WKDW D
PLQLPXPRIJULGEORFNVSHUOD\HUDUHQHHGHGWRPRGHOWKHSK\VLFDOWUDQVYHUVHGLIIXVLRQ
91
Figure 3.30 shows the salinity distribution at 2 PVI (steady state flow) for the 20, 4 and 2 grid block cases
for two different KT / v ratios. As the ratio increases, the mixing zone increases. The mixing zone is captured
better as the vertical grid is refined. The 2 grid block scenario is not sufficient enough to capture the mixing
zone.
-3 -2
KT / v = 3.8 × 10 cm Direction of flow KT /v = 1.2 × 10 cm
NZ = 20 grid blocks NZ = 20 grid blocks Salt mole
fraction
NZ = 4 grid blocks NZ = 4 grid blocks
NZ = 2 grid blocks NZ = 2 grid blocks
Figure 3.30: The water salinity distribution at 2 PV for the 20, 4 and 2 grid block cases for two different values of KT / v. This figure shows that as
the ratio increases, the mixing zone increases and is captured better in the 20 grid block case. It also shows that the 2 grid block scenario does not
capture the mixing zone very well when compared to the other resolution cases.
3.5 Conclusions
The aim of this chapter was to investigate several factors that affect the modelling of low salinity
waterflooding: numerical dispersion caused by the grid resolution, the method used to model the transition
from using high salinity relative permeabilities to low salinity relative permeabilities and its impact on
predictions and the smallest number of grid blocks needed to capture the longitudinal and transverse
diffusivities.
First, the numerical dispersion effect on modelling low salinity waterflooding was investigated and
compared against the analytical solution. It was found that 1000 grid blocks reproduces the analytical
solution within 1 %. The error difference between the analytical solution and the simulated model using any
number of grid block was found to vary linearly with a first order of convergence.
Generally, it was found that numerical dispersion in low salinity waterflooding results in an earlier
breakthrough of the leading connate water bank. The water cut after the breakthrough of the connate water
bank is higher when there is higher numerical diffusion. The increase in water cut as a result of the trailing
shock front is much slower in coarse grid models compared with very fine grid models. This means that care
needs to be taken when predicting water production for designing surface facilities. Typical coarse grid
simulation models will predict a lower water production rate later in field life than is actually the case. Very
92
fine grid models are needed to capture this water cut development properly because the trailing shock front
is less self-sharpening than the leading shock front.
Second, physical diffusion was included in the grid refinement study. Velocity dependent dispersion was
ignored. The study used a 1D model to investigate the impact of numerical diffusion on the prediction of
physical longitudinal diffusion and a 2D model to investigate the impact of numerical diffusion on the
prediction of transverse physical diffusion.
It was found that the smallest number of horizontal grid blocks that sufficiently captures most of the physical
longitudinal diffusion and minimizes the numerical diffusion is 200 while it was found that more than 2 grid
blocks were needed vertically to capture the physical transverse diffusion.
For longitudinal diffusion, there is a critical Pe that can be used to determine when physical diffusion
dominates over numerical diffusion this can be calculated from the input diffusivity, frontal advance rate and
system length but its value depends non-linearly on the number of grid blocks. There is also a plateau Pe
which describes the level of numerical diffusion in a simulation and also depends non-linearly on the number
of grid blocks. Even when physical diffusion dominates over numerical diffusion there is a contribution to
the diffusion from the numerical truncation error that varies with the number of grid blocks. This is due to
higher order truncation errors in combination with the variable time-step size chosen by the simulator. All
these considerations suggest that simply using a chosen number of coarse grid blocks to represent the effects
of physical diffusion and dispersion on mixing in low salinity waterflooding is only a very crude
approximation.
Third, low salinity waterflooding models were validated by the analytical Buckley Leverett solution. The
salinity threshold model (ECLIPSE and CMG STARS) and the ion exchange model (CMG GEM) produce
generally the same water saturation profile as the analytical Buckley Leverett solution so it was decided to
choose the salinity threshold model as the low salinity waterflooding model for the remainder of this thesis
as its simulation running time is significantly faster than the ion exchange model. For the salinity threshold
model, the simulation predictions are closer to the analytical solution when the salinity thresholds are
midway between the injected low salinity water and the connate water so that the change from low salinity
water to high salinity water is almost instantaneous and the region of the intermediate salinities are narrow.
The simulation results showed unphysical oscillations in water cut during production of the connate water
bank, when a coarse grid model was used. These were probably due to using a coarse grid and having a very
small range in salinity between the upper and lower salinity thresholds. As a result, the volume of mobile oil
and water could change quite suddenly in the production well block. This highlights the importance of
choosing a suitable range for the salinity thresholds as well as demonstrating that problems may arise when
simulating low salinity waterflooding using very coarse grids such as are often used in field scale simulation.
Finally, the impact of low salinity waterflooding model on predictions was investigated for both models: the
salinity threshold model and ion exchange model. For the salinity thresholds model, the impact of the
salinity thresholds on the performance of low salinity waterflooding was investigated. It was found that the
best salinity thresholds that predicts the closest results to the analytical solution were when they are midway
93
between the injected water salinity and the connate water salinity. The maximum effect of the salinity
thresholds on the oil recovery (5 % difference) are during the production of the connate water bank. The
effect on the saturation fronts are as follows: earlier water breakthrough (leading shock front) when the
salinity thresholds are narrow and closer to the injected water and delayed arrival of trailing shock front with
higher water cut when the salinity thresholds are narrow. At field scale, the choice of the salinity thresholds
in modelling low salinity waterflooding may have significant effects on the prediction of oil recovery and the
producing water cut. In case the actual upper salinity threshold is much higher than the modelled one in the
simulator, then the predicted oil recoveries by the simulator will be lower (~5 %) with higher water cut
during the production of the connate water bank. This might affect the decision to implement low salinity
waterflooding in a field scale. As there is an uncertainty in estimating the salinity thresholds for any field, it
is recommended to conduct coreflood experiments to estimate these salinity thresholds that will be used in
the simulation model for better predictions.
For the ion exchange model, the choice of the interpolation parameter, the values to change from the high
salinity water relative permeability curve to the low salinity water relative permeability curve and the cation
exchange capacity and their impacts on low salinity waterflooding were investigated. It was found that
setting the equivalent fraction of Na-X as the interpolation parameter predicted a similar water saturation
profile to the analytical Buckley Leverett solution compared with the Ca++ concentration. In addition, when
the cation exchange capacity (CEC) was about 5 equivalent/m3 (i.e. the amount of ions that are absorbed by
the clay) the predictions are closer to the analytical Buckley Leverett solution. Moreover, it was found that
the values to change from high salinity water relative permeability curve to the low salinity water relative
permeability curve had a minor effect on modelling low salinity waterflooding compared with the choice of
the interpolation parameter and the value of cation exchange capacity.
94
CHAPTER 4: IMPACT OF HETEROGENEITY ON LOW SALINITY
WATERFLOODING EXCLUDING MIXING
Reservoir heterogeneity alters the vertical and areal sweep efficiency of low salinity waterflooding. As
previously discussed in chapter 2, it is considered as one of the main factors that affect the success of
deploying low salinity waterflooding on the field scale. Consequently, in this chapter, a detailed simulation
study is conducted to find its effects on the performance of low salinity waterflooding. The study is
conducted without explicitly modelling mixing (which will be investigated in chapter 5).
The study is conducted first using continuous low salinity waterflooding then using a slug followed by
conventional water in field scale models. For both modes, two main models were used: 2D layered models
with simple heterogeneity variations and areal models with complex heterogeneity taken from the SPE 10
model 2 (Christie and Blunt 2001). An analytical method is presented for estimating the best slug size of low
salinity water.
4.1 Continuous Low Salinity Waterflooding
4.1.1 Description of Models Used
A- 2D Layered Models with Simple Heterogeneity
The fluid and reservoir properties used throughout this chapter are shown in Table 4.1 and the relative
permeability curves are shown in Figure 4.1. The difference in connate water saturation between the two
relative permeability curves was discussed in section 3.1.
The model dimensions and their isotropic permeabilities are shown in Table 4.2. Figure 4.2 shows the
simulation models with their permeability variations. In all scenarios, the salinity of the injected water was
1,000 ppm and the salinity thresholds were 1,000-7,000 ppm. These thresholds were chosen as they
represent the more realistic thresholds as confirmed by previous experiments (Tang and Morrow 1999).
One injection well and one producer well were used in the models. Both wells were completed vertically in
all layers. The injection well was located in the first column (the left side of the model) and set to have a
constant injection rate of 125 sm3/d while the producer well was located in the last column (the right side of
the model) and set to have a constant production rate of 125 sm3/d. For all models, low salinity water was
injected immediately and compared with immediate conventional waterflooding in addition to injecting low
salinity water post the breakthrough of immediate conventional waterflooding. In this chapter, water density
was kept constant at 1,000 kg/m3 for both brines (low and high) as it has negligible effects on the
performance of low salinity waterflooding. This was confirmed by comparing simulation runs including and
ignoring the brine density difference. Oil recoveries obtained by both models were identical. The ECLIPSE
simulator that utilizes the salinity threshold model was selected for this evaluation.
95
Table 4.1: Reservoir and fluid properties used in the 2D layered models to evaluate the impact of heterogeneity on low salinity waterflooding.
Property Value
Porosity 0.2
Initial reservoir pressure 265 bar
Rock compressibility 0.3 × 10-5 bara-1
Water compressibility 4.6 × 10-5 bara-1
Oil density 850 kg/m3
Water density 1,000 kg/m3
Oil viscosity 0.47 cp
Water viscosity 0.5 cp
Oil formation volume factor 0.999 rm3/sm3
Water formation volume factor 1.03 rm3/sm3
Connate water salinity 30,000 ppm
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Sw
krwH
krwh kroH
kroh krwL
krwL kroL
kroL
Figure 4.1: Relative permeability curves used for the 2D layered models to evaluate the impact of reservoir heterogeneity on low salinity
waterflooding.
Table 4.2: Description of the 2D layered models used to evaluate the impact of heterogeneity on low salinity waterflooding.
Case
Property
Homogeneous LH HL LHLHL HLHLH
Model dimensions
1,041 × 100 × 25 m
(x, y, z)
Number of grid
100 × 1 × 50
blocks (x, y, z)
Grid block size
10.41 × 100 × 0.5 m
(dx, dy, dz)
The top 20z (10 The top 20z (100 Pattern of the top 10z Pattern of the top 10z
Permeability (z) 100 md md) the bottom md) the bottom (10 md), (100 md),
30z (100 md) 30z (10 md next 10z (100 md), next 10z (10 md),
96
Homogeneous LH
HL LHLHL HLHLH
Permeability (md)
10 100
Figure 4.2: Permeability variations for the 2D layered models used to evaluate the impact of heterogeneity on low salinity waterflooding.
B- Areal Models Taken from SPE 10 Model 2 with Complex Heterogeneity

SPE 10 model 2 was used as it produces more complex heterogeneity realisations compared to the 2D
layered models that had simple heterogeneity realisations. It was created by Christie and Blunt (2001) from
the data of Floris et al. (2001). The model dimensions were 368 × 675 × 52 m discretized with 60 × 220 × 85
grid blocks with a grid block size of 6 × 3 × 0.6 m.
In this investigation, the model was sliced into 85 layers. Each areal layer had a 60 × 220 × 1 grid. The top 35
layers represented the Tarbert formation with less heterogeneous layers compared with the bottom 50 layers
from the Upper-Ness formation that were more heterogeneous. Figure 4.3 shows the permeability
distribution of layers 1 and 63.
Layer 1 (Tarbert) Hv = 0.70 Layer 63 (Upper-Ness) Hv = 0.45
Permeability (md)
1000
100
10
0.1
Figure 4.3: Permeability distribution in areal layers 1 and 63 taken from SPE 10 model 2.
97
Before starting water injection, it was essential to quantify the heterogeneities for these 85 layers. The
vorticity heterogeneity index Hv (previously defined in chapter 2) was used for quantifying their
heterogeneities. Rashid et al. (2012) found that for layers 1-35 (Tarbert formation) Hv varies from 0.54 to
0.84. However, for layers 36-85 (Upper-Ness formation) it varies from 0.31 to 0.67. Figure 4.3 illustrates the
permeability distribution for layers 1 and 63 with Hv of 0.7 and 0.45 respectively.
In this section, first the relationship between the vorticity heterogeneity index and the oil recoveries obtained
from both low and high salinity waterflooding is investigated, then its relationship with water breakthrough
from both types of waterflooding is analysed.
In this evaluation, one injection well and one production well were used with constant rates for each layer.
Both wells were horizontal. Depending on the heterogeneity variation of each layer, the rates were set
between 1.5 to 10 m3/d. The injection well was located in the first row of the model while the oil producer
was located in the last row of the model and separated by 675 m. For all layers, low salinity water was
injected immediately and compared with immediate conventional waterflooding.
4.1.2 Results of Continuous Low Salinity Waterflooding
In this section, the outcomes of the evaluation of the impact of heterogeneity on injecting immediate low
salinity water continuously are presented. The results are shown for both immediate injection of low salinity
water and injection of low salinity water post immediate injection of conventional water. First, the results are
shown for the 2D layered models with simple heterogeneity, then for the areal models taken from SPE 10
model 2 with complex heterogeneity.
A- 2D Layered Models with Simple Heterogeneity Variations
The oil recovery obtained at 1 PV of water injection from all cases is shown in Figure 4.4. There is a clear
effect of reservoir heterogeneity on oil recovery in both types of waterflooding. Equation 4.1 was used to
quantify the drop in oil recovery for each layered case. The calculated drops in oil recovery for each case are
plotted in Figure 4.5.
𝑜𝑜𝑜𝑜𝑜𝑜 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑓𝑓𝑓𝑓𝑓𝑓 𝑡𝑡ℎ𝑒𝑒 ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 − 𝑜𝑜𝑜𝑜𝑜𝑜 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑎𝑎𝑎𝑎 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
% 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 𝑖𝑖𝑖𝑖 𝑜𝑜𝑜𝑜𝑜𝑜 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = × 100 (4.1)
𝑜𝑜𝑜𝑜𝑜𝑜 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑓𝑓𝑓𝑓𝑓𝑓 𝑡𝑡ℎ𝑒𝑒 ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
98
0.9 18
0.8 16
0.7 14
Drop in oil recovery (%)

0.6 12
Oil recovery
0.5 10
0.4 8
0.3 6
0.2 4
0.1 2
0 0
Homogeneous LH HLHLH HL LHLHL
LH HL LHLHL HLHLH
Immediate conv Immediate lowsal immediate conv immediate lowsal
lowsal post immediate conv lowsal post immediate conv
Figure 4.4: Oil recovery comparison at 1 PV of immediate water Figure 4.5: Drop in oil recovery for immediate injection of low salinity
injection into 2D layered cases. The oil recovery from immediate water and injecting low salinity water post immediate injection of
injection of low salinity water is always greater than the oil recovery conventional water for the 2D layered cases with simple heterogeneity
from immediate injection of conventional water, but the differences in oil variations. For the HL, LHLHL and HLHLH, the drop in oil recoveries
recoveries are not the same for all cases. by immediate injection of low salinity water compared to the injection of
For the HLHLH and HL cases and assuming total water injection of 1 conventional water are larger than the drop in oil recoveries in the LH
PV, the injection of low salinity water post the breakthrough of case. This is due to mixing caused by viscous cross flow between the
conventional water recovers almost as much oil as injecting low salinity injected low salinity water and connate water, while in the LH case,
water immediately since gravity and viscous forces act together to gravity and viscous forces act together to drive the injected low salinity
improve the vertical sweep efficiency. water to sweep the bottom layer with high permeability thus leaving the
top layer with low permeability unswept.
With the exception of the LH case, low salinity waterflooding is more affected by reservoir heterogeneity
than conventional waterflooding. In the HL and LHLHL cases, the percentage drops in oil recovery with low
salinity waterflooding are 4 % more than the drop in conventional waterflooding. In the HLHLH case, the
percentage drop in low salinity waterflooding is even more (about double the drop in conventional
waterflooding).
Figure 4.6 shows the oil saturation distribution at 1 PV of immediate low salinity water injection and 1 PV
of immediate injection of conventional water. In the homogeneous case, most of the layers have a residual
oil saturation of around 11%, however in the heterogeneous cases, several factors cause the reduction in the
efficiency of low salinity waterflooding. Recovery factor is the product of the microscopic displacement and
the macroscopic sweep. Low salinity water injection improves the microscopic displacement by modifying
the relative permeability values krw and kor and reducing the residual oil saturation Sor. Reservoir
heterogeneity reduces the macroscopic sweep (reducing the volume of the reservoir actually contacted by the
injected low salinity water).
Viscous cross flow causes mixing between the injected low salinity water and the connate water for
immediate low salinity waterflooding which results in having water with intermediate salinities thus
reducing the performance of low salinity waterflooding in the more heterogeneous cases as shown in Figure
4.7. Gravity causes low salinity water in the HL case to slump into the low permeability layer at the bottom
leading to a better vertical sweep efficiency compared to the LH case.
99
Immediate injection of low salinity water Immediate injection of conventional water
Direction of flow Homogeneous LH
Homogeneous LH
HL LHLHL HLHLH HL LHLHL HLHLH
Oil saturation
0.11 0.29 0.46 0.63 0.80
Figure 4.6: Oil saturation distribution at 1 PV of water injection for immediate low salinity water and immediate injection of conventional water into
2D layered models with isotropic permeabilities. Reservoir heterogeneity affects the macroscopic sweep in low salinity waterflooding more than
conventional waterflooding. Viscous cross flow causes mixing between the injected low salinity water and connate water.
Homogeneous LH
Direction of flow
HL LHLHL HLHLH
Salinity (1,000 ppm)
1 4 7 11 14 17 20 24 27 30
Figure 4.7: Salinity distribution at 1 PV of immediate injection of low salinity water into 2D layered models with isotropic permeabilities. Viscous
cross flow causes mixing between low salinity water and connate water which results in having water with intermediate salinities (green) thus
reducing the performance of low salinity waterflooding in more heterogeneous cases. There is a difference between the vertical sweep efficiency in
the LH case and HL case due to the effect of gravity which causes low salinity water to slump into the bottom layer with low permeability in the HL
case compared to the poor vertical sweep efficiency of the top layer with low permeability in the LH case.
To evaluate the impact of viscous cross flow for immediate low salinity waterflooding, the simulations were
repeated with assuming zero vertical permeabilities. Figure 4.8 shows the comparison of the drop in oil
recoveries between the 2D layered cases with isotropic permeabilities and the 2D layered cases with zero
vertical permeabilities. The drop in oil recoveries for the cases with zero vertical permeabilities are more
than the cases with isotropic permeabilities. Viscous crossflow enhances oil recovery in heterogeneous
reservoirs by promoting the flow of water into the low permeability layers thus improving the overall
vertical sweep efficiency. However, this crossflow adversely affects the microscopic displacement efficiency
in the low salinity water floods. This is because the crossflow results in mixing between the low salinity
water and the high salinity connate water bank in the adjacent layer, reducing the benefit of the low salinity
100
flood. Figure 4.9 shows the comparison of oil saturation distribution at 1 PV of immediate injection of low
salinity water in the cases with isotropic permeabilities and the cases with zero vertical permeabilities.
30 30
25 25

20 20
15 15
10 10
5 5
0 0
LH HL LHLHL HLHLH LH HL LHLHL HLHLH
immediate conv immediate lowsal Immediate conv Immediate lowsal
Figure 4.8: Comparison between the drop in oil recovery for immediate waterflooding due to heterogeneities with isotropic permeabilities case and
the case with zero vertical permeabilities. For the case of isotropic permeabilities (left plot), the drop in oil recoveries obtained by immediate low
salinity waterflooding due to heterogeneities is about (5-15%) which is lower than the case with zero vertical permeabilities (25-30%) (right
plot).Viscous cross flow actually enhances the oil recovery in the heterogeneous reservoirs for the case with isotropic permeabilities compared to the
case with zero vertical permeabilities because it promotes the flow of low salinity water into the low permeability layers thus increasing the overall
vertical sweep efficiency in all layered cases.
Isotropic permeabilities Zero vertical permeabilities

Homogeneous LH Homogeneous LH Direction of flow
Direction of flow
HL LHLHL HLHLH HL LHLHL HLHLH
Oil saturation
0.11 0.29 0.46 0.63 0.80

Figure 4.9: Oil saturation distribution at 1 PV of immediate low salinity waterflooding for two cases: with isotropic permeabilities and with zero
vertical permeabilities. For the case with isotropic permeabilities, viscous cross flow promotes the flow of low salinity water into the low
permeability layers thus increasing the overall vertical sweep efficiency in all layered cases. There is a difference between the vertical sweep
efficiency in the LH case and HL case due to the effect of gravity which causes low satiny water to slump into the low permeability layer in the HL
case compared to the poor vertical sweep efficiency of the low permeability layer at the top in the LH case. However, for the case with zero vertical
permeabilities (no viscous cross flow), the low permeability layers were less swept leading to poor vertical sweep efficiency in all layered cases.
101
The oil recoveries obtained by immediate low salinity waterflooding and immediate conventional
waterflooding after the injection of 1 PV are plotted against vorticity heterogeneity index for all layers taken
from the SPE 10 model 2 (Figure 4.10).
0.8
Oil recovery at 1 PV of water injection
y = 0.5631x + 0.229
R² = 0.7757
0.7
0.6
Immediate lowsal
0.5 Immediate conv

y = 0.4909x + 0.1788
R² = 0.7688 Linear (Immediate lowsal)
0.4
Linear (Immediate conv)
0.3
0.2
0.2 0.4 0.6 0.8 1
Hv
Figure 4.10: Oil recovery at 1 PV of immediate low salinity water injection and immediate conventional water injection versus the vorticity
heterogeneity index for layers 1 to 85 taken from SPE 10 model 2. Since the slopes of both lines are almost similar, the performance of immediate
low salinity waterflooding is affected by reservoir heterogeneity to the same extent as immediate conventional waterflooding. For both types of
waterflooding, oil recovery varies linearly with the vorticity heterogeneity index, the less heterogeneous the reservoir is (high Hv value), the higher
the oil recovery is.
Oil recoveries vary linearly with the vorticity heterogeneity index for both types of waterflooding. The
regression values for both linear relationships are good at around 0.77. The separation between the two
straight lines is slightly smaller when the reservoir heterogeneity increases, and it is larger when the
reservoir heterogeneity decreases. Apparently, reservoir heterogeneity can have a severe impact on oil
recovery by both types of waterflooding. At the macroscopic scale, the sweep efficiency is poor in the very
heterogeneous reservoirs (low Hv values) due to the effect of channelling of the invaded water through the
higher permeability areas.
The oil saturation distribution at 1 PV of immediate low salinity waterflooding and immediate conventional
waterflooding for layers 9 and 63 taken from SPE 10 model 2 are shown in Figure 4.11. Low salinity
waterflooding has swept most of layer 9 (less heterogeneous) compared with layer 63 (more heterogeneous)
which was poorly swept due to the water channelling through the high permeability areas. Similarly,
conventional waterflooding swept most of layer 9 compared to layer 63 but the overall oil saturation is
higher (light blue) since the residual oil saturation was not reduced as in the case of low salinity
waterflooding (dark blue).
102
Layer 9 (Tarbert) Hv = 0.82
Injector (lowsal) Injector (conventional)
Oil saturation
Permeability (md) 1
0.9
1000 0.8
0.7
100 0.6
0.5
10 0.4
0.3
1
0.2
0.1 0.1
0
Oil producer Oil producer

Layer 63 (Upper-Ness) Hv = 0.45
Injector (lowsal) Injector (conventional)
Permeability (md) Oil saturation

1
1000 0.9
100 0.8
0.7
10 0.6
0.5
1 0.4
0.3
0.1
0.2
0.1
0
Oil producer Oil producer
Figure 4.11: Oil saturation distribution at 1 PV of immediate injection of low salinity water and immediate injection of conventional water for layers
9 and 63 taken from SPE 10 model 2. In both types of waterflooding, sweep efficiency is poor in layer 63 due to the effect of channelling of the
invaded water through the higher permeability areas.
Figure 4.12 shows the plot of water breakthrough from both types of waterflooding versus the vorticity
heterogeneity index for all areal layers taken from the SPE 10 model 2.
0.5
0.45
y = 0.4205x + 0.0378
R² = 0.7547
0.4
Water breakthrough (PVI)
0.35
0.3 y = 0.3879x + 0.037 Immediate lowsal

R² = 0.7543
0.25 Immediate conv
0.2 Linear (Immediate lowsal)

Linear (Immediate conv)
0.15
0.1
0.05
0
0.2 0.4 0.6 0.8 1
Hv
Figure 4.12: Water breakthrough (PVI) of immediate injection of low salinity water and immediate injection of conventional water versus the
vorticity heterogeneity index for layers 1 to 85 taken from SPE 10 model 2, water breaks through in immediate conventional waterflooding earlier
than in immediate low salinity waterflooding. For both types of waterflooding, the more heterogeneous the reservoir is, the earlier the water
breakthrough is due to channelling of the invaded water through high permeability areas.
103
Similar to the oil recovery plot, water breakthrough varies linearly with the vorticity heterogeneity index.
The regression values for both straight lines are 0.75. Generally, water breaks through in conventional
waterflooding earlier than in low salinity waterflooding which is consistent with high salinity water injection
being more oil wet than low salinity water injection.
It can be concluded from oil recovery and water breakthrough plots that the sweep efficiency of low salinity
waterflooding is reduced as the reservoir heterogeneity increases but not significantly more than
conventional waterflooding.
4.2 Injection of Low Salinity Water as a Slug
Low salinity water is normally injected as a slug of finite volume during both immediate injection and
injection post immediate conventional waterflooding. This is because the water is obtained from desalination
plants or another often distant water source. Once the reservoir starts producing water (normally saline) this
water is too dirty to pass through a desalination plant or be disposed over offshore without further expensive
processing so it is typically re-injected. It is crucial to know how big this slug has to be to obtain a good
(economic) improved oil recovery as this will drive the design of the surface facilities and indeed whether
low salinity water injection is economically feasible. Previous work conducted by Jerauld et al. (2008)
indicated that the best slug size of low salinity waterflooding is around 0.4 PV since at this slug size most of
the oil is contacted by the injected low salinity water. However, the impact of reservoir heterogeneity on slug
size was not included in their study. The same two models that were used in continuous low salinity
waterflooding (section 4.1) are used here.
It is often economically infeasible to inject low salinity water continuously due to the need to reinject
produced water (for example offshore) or due to limitations in desalination capacity. In this section, the
impact of reservoir heterogeneity is evaluated when low salinity water is injected as a slug in both immediate
low salinity waterflooding and injection of low salinity water post immediate conventional waterflooding.
The purpose of this section is to find the best slug size of low salinity water that will recover oil almost as
much as if it is injected continuously to test the analytical solution. First, an analytical method is presented to
find this best size and then a simulation study follows.
4.2.1 Analytical Method
The best slug size of low salinity water is required to be estimated prior to implementing in the field to avoid
significant cost increase when increasing the slug size above this best size or having a less efficient low
salinity waterflood that results in negligible or no incremental oil recovery if a smaller slug size is chosen,
such that the slug loses its integrity before arriving at the production well.
Therefore, it is essential to estimate this best slug size of low salinity water. In this section, it is estimated
using an analytical approach defined by Wright et al. (1987). It is assumed that that there is minimal mixing
between the injected low salinity water and the in situ high salinity water in addition to negligible gravity
and capillary forces.
104
During the injection of a low salinity water slug in layered reservoirs (illustrated in Figure 4.13), the injected
low salinity water moves faster in the high permeability layer and then breaks through much more quickly
than in the low permeability layer. If the chase high salinity water flows in the same high permeability layer
and passes the horizontal position of the slower moving low salinity water slug in the low permeability layer,
bypassing occurs and the low salinity water slug will disintegrate as shown in Figure 4.13. The assumption is
that once this bypassing occurs, the low salinity effect will be reduced by transverse diffusional mixing and
crossflow between the layers. The objective of this analysis is to estimate the location of this horizontal
position (bypassing point) at various slug sizes and the best slug size accordingly.
To estimate the position of the bypassing point (Xbp) at any slug size of low salinity water, the following
equation is used:
𝑉𝑉
𝑋𝑋𝑏𝑏𝑏𝑏 = (4.2)
1
�(1 − 𝑓𝑓𝑎𝑎 ) �1 − 𝑅𝑅 � + 𝑓𝑓𝑎𝑎 (𝑅𝑅21 − 1)�
32
where:
Xbp is the bypassing point in the fraction of the model’s length (dimensionless)
V is low salinity water slug size (in pore volume)
𝑓𝑓𝑎𝑎 is the fraction of the model’s pore volume in the high permeability layer (dimensionless)
subscript 1 refers to the oil phase, subscript 2 is for low salinity water and subscript 3 is for the chase high
salinity water
R is defined as per the following equations:
(𝑅𝑅𝑐𝑐 + 1)
𝐸𝐸 = � (4.3)
𝑅𝑅𝑐𝑐 𝑀𝑀 + 1
𝑅𝑅𝑐𝑐 (𝐸𝐸 + 1)
𝑅𝑅32 = (4.4)
1
�𝐸𝐸 + 𝑀𝑀�
M is the mobility ratio between any two fluids

Rc is the conductance ratio (permeability contrast)
105
Slug Oil
Slug Oil
a)
Slug Oil
Slug Oil
Chase fluid Slug Oil

b)

c) Chase fluid Slug Oil
Figure 4.13: Illustration of slug disintegration process in a layered reservoir. In these cases, low oil recovery is achieved as a result of mixing between
the chase fluid and the slug ahead of it (modified from Wright et al. 1987).
The HL model (previously described in section 4.1.1A) is used to estimate the best slug size of low salinity
water analytically. The model’s rock and fluid properties are shown in Table 4.3 with the same relative
permeability curves that are used in the simulation models throughout this chapter as shown in Figure 4.14.
The difference in connate water saturation between the two relative permeability curves was discussed in
section 3.1.
Table 4.3: Reservoir and fluid properties used for the estimation of the 1
best slug size of low salinity water analytically.
0.9
0.8
Property Value
Porosity 0.2 0.7
0.6
Rc (permeability contrast) Three scenarios: 2, 5 and 10
0.5
Oil viscosity (cp) 0.47
0.4
Water viscosity (cp) 0.5 0.3
Oil density (kg/m3) 850 0.2
Water density (kg/m3) 1,000 0.1
Layers a and b thicknesses (m) 10 and 15 0
System length (m) 1,041 0 0.2 0.4 0.6 0.8 1
System width (m) 100 Sw
krwH
krwh kroH
kroh krwL
krwL kroL
kroL
Figure 4.14: Relative permeability curves used for the estimation of the
best slug size of low salinity water analytically.
Three scenarios were conducted with permeability contrasts of 2, 5 and 10. The positions of the bypassing
point were calculated at different slug sizes of low salinity water and are plotted in Figure 4.15. The best
slug size of low salinity water (Vbp) is defined when Xbp=1 (at the producer location) and calculated as per
the following equation:
1 1
𝑉𝑉𝑏𝑏𝑏𝑏 = (1 − 𝑓𝑓𝑎𝑎 ) �1 − � + 𝑓𝑓𝑎𝑎 𝑅𝑅𝑐𝑐 𝑀𝑀21 �1 − � (4.5)
𝑅𝑅32 𝑅𝑅21
For different conductance ratios, the best slug size of low salinity water was calculated and plotted versus the
conductance ratio in Figure 4.16. For reservoirs with large permeability contrast between layers, the best
slug size of low salinity water needs to be very large to avoid bypassing of low salinity water in the low
permeability layer and avoid slug disintegration.
106
1 1.8
0.9
Bypassing point (fraction of model's length)
1.6
0.8
0.7 1.4
0.6 1.2
0.5
Vbp (PVI)
1
0.4
0.3 0.8
0.2 0.6
0.1
0 0.4
0 0.5 1 1.5 2 0.2
Slug size of immediate injection of low salinity
water (PV) 0
Rc=2 fa=0.4 (Xbp) Rc=5 fa=0.4 (Xbp) 0 5 10 15
Rc=10 fa=0.4 (Xbp) Rc (conductance ratio)
Figure 4.15: The slug size of low salinity water needed to avoid bypassing Figure 4.16: The best slug size of immediate injection of low salinity
while the injected low salinity water is progressing through the layers water increases with the permeability contrast (conductance ratio),
increases from 0.35 PVI using Rc = 2 to 1.6 PVI using Rc = 10. which means in reservoirs with large permeability contrast between
layers, a very large slug size of immediate low salinity waterflooding
is needed to avoid slug disintegration.
4.2.2 Simulation Study

In this section, the slug size of immediate low salinity waterflooding and the slug size of the injection of low
salinity water post immediate conventional waterflooding are evaluated as a function of reservoir
heterogeneity. Similar to continuous low salinity waterflooding, mixing is excluded (it will be investigated
later in chapter 5).
A-2D Layered Models with Simple Heterogeneity
In this section, the slug sizes of low salinity water are examined as an immediate injection and as an
injection post immediate conventional waterflooding.
Slug Sizes of Immediate Injection of Low Salinity Water into 2D Layered Models
In this section, various slug sizes (0.4, 0.6, 0.8, 1, 1.5 and 1.75 PV) for immediate low salinity waterflooding
are evaluated and compared with several heterogeneity variations. The oil recoveries obtained after 1 and 2
PV of water injection are shown in Figure 4.17. In this case low salinity water was injected as soon as
recovery was initiated.
107
0.18 0.18
0.15 0.15
Incremental oil recovery
Incremental oil recovery

0.12 0.12
0.09 0.09
0.06 0.06
0.03 0.03
0 0
Homogeneous LH HLHLH HL LHLHL Homogeneous LH HL LHLHL HLHLH
0.4 PV lowsal 0.6 PV lowsal 0.8 PV lowsal

0.4 PV lowsal 0.6 PV lowsal 0.8 PV lowsal lowsal 1.0 PV lowsal 1.5 PV lowsal 1.75 PV lowsal
lowsal
Figure 4.17: Oil recovery comparison at 1 and 2 PV of immediate low salinity water injection into 2D layered models for various slug sizes compared
with continuous injection of low salinity water. At 1 PV of total water injection, a slug size of at least 0.6 PV is beneficial in heterogeneous reservoirs
increasing to 0.8 PV in highly heterogeneous reservoirs. At 2 PV, the slug of low salinity water sizes increase to 1.5 PV and even to 1.75 PV in
highly heterogeneous reservoirs.
At 1 PV of water injection, the best slug size of immediate low salinity water injection is 0.6 PV for the
homogeneous and LH cases while it increases to 0.8 PV for the HLHLH, HL and LHLHL cases. At 2 PV
water injection, the best slug size for the homogeneous and LH cases is 1.5 PV while it increases to 1.75 PV
in the rest of the cases. For the homogeneous and LH cases, the slug size is found to be 60% at 1 PV and
increased to 75% at 2 PV. In the HL, LHLHL and HLHLH cases, the best slug sizes is 80% at 1 PV
increased slightly to 87% at 2 PV. Generally, as the duration of the water injection increases, the best slug
sizes increases. The oil saturation distribution at 2 PV for slug sizes of 1.5 and 1.75 PV is compared with
continuous injection of immediate low salinity water are shown in Figure 4.18. 1.5 PV slug of low salinity
water injection is effective for the homogeneous and LH cases since most of the layers are swept similar to
continuous low salinity waterflooding. However, 1.5 PV slug of low salinity water injection is not as
effective as continuous low salinity water injection for the rest of the cases. At this slug size, mixing of the
injected low salinity water with the saline connate water that is caused by viscous cross flow in addition to
the gravity severely reduces the effect of low salinity waterflooding. Therefore, the overall macroscopic
sweep efficiency is reduced resulting in much lower oil recovery at 2 PV compared with continuous low
salinity waterflooding. This is why the best slug of low salinity water is 1.75 PV for these cases. More low
salinity water injection is needed to minimize mixing. However in this chapter, mixing is not exclusively
modelled. Its effects are further investigated in chapter 5.
108
Direction of flow
Lowsal (homogeneous) 1.5 PV slug (homogeneous) 1.75 PV slug (homogeneous)
Lowsal (LH) 1.5 PV slug (LH) 1.75 PV slug (LH)
Lowsal (HL) 1.5 PV slug (HL) 1.75 PV slug (HL)
Lowsal (LHLHL) 1.5 PV slug (LHLHL) 1.75 PV slug (LHLHL)
Lowsal (HLHLH) 1.5 PV slug (HLHLH) 1.75 PV slug (HLHLH)
Oil saturation
0.11 0.29 0.46 0.63 0.80
Figure 4.18: Oil saturation distribution at 2 PVI for different slug sizes of immediate low salinity water injection into 2D layered models. 1.5 PV slug
of immediate low salinity water injection is effective for the homogeneous and LH cases since most of the layers are swept similar to continuous low
salinity waterflooding. In the other cases, mixing of the injected low salinity water with the connate water and the connate water bank severely
reduces the effect of low salinity waterflooding. This mixing is due to viscous cross flow and numerical diffusion. Viscous cross flow between the
layers and gravity reduces the overall macroscopic sweep efficiency of waterflooding resulting in much lower oil recovery at 2 PVI compared with
continuous low salinity waterflooding. This is why the best slug size of low salinity water injection is 1.75 PVI for these cases. More low salinity
water injection is needed to minimize mixing.
109
Comparison between the Simulation Models and the Model for Predicting the Slug Size of Immediate
Low Salinity Waterflooding
In this section, the predictions of the slug sizes by the simulation models are compared to the analytical
model. Since the analytical model neglects the effect of viscous cross flow, two cases of simulation models
are compared with the analytical model: one with maximum viscous cross flow between layers (i.e. isotropic
permeabilities) and the other with no viscous cross flow between layers (i.e. vertical permeabilities are zero).
Three different permeability ratios were used (also known as the conductance ratio, (Rc)): 2, 5 and 10. For
each of these permeability ratios, the slug sizes of immediate low salinity waterflooding were varied: 0.3,
1.0, 1.5, 1.8 PVI. The oil recoveries obtained by each of these scenarios are plotted (Figure 4.19) as a
percentage of oil recovery obtained by continuous injection of low salinity water at 2 PV. The best slug size
of immediate low salinity waterflooding by the analytical model is shown in Figure 4.20.
For the case with zero vertical permeabilities, the best slug size predicted by the simulation models is around
1.0 PVI using Rc = 2 and increases to 1.8 PVI using Rc = 10 (solid lines in Figure 4.19). This generally
matches the analytical model (Figure 4.20). However, for the case with maximum viscous cross flow
(isotropic permeabilities), the percentage of oil recovered by the small slug sizes using Rc = 2 and 5 (dashed
lines in Figure 4.19) are lower than the slug sizes in the cases with zero vertical permeabilities. This is due to
mixing between the injected low salinity water and connate water that is caused by viscous cross flow thus
affecting the performance of low salinity waterflooding.
100 1.8
% of oil recovered compared to 2 PV of continuous
98 1.6
96
1.4
94
low salinity waterflooding
1.2
92
Vbp (PVI)
90 1
88 0.8
86 0.6
84
0.4
0 0.5 1 1.5 2
Slug size of immediate low salinity water injection
0.2
(PVI)
Zero vertical perm Rc = 10 Zero vertical perm Rc = 5 0
Zero vertical perm Rc = 2 Isoperm Rc = 10 0 5 10 15
Isoperm Rc = 5 Isoperm Rc = 2 Rc (conductance ratio)
Figure 4.19: Simulation results of oil recoveries obtained by different slug Figure 4.20: The best slug size of immediate low salinity
sizes of immediate low salinity water injection as a percentage of oil waterflooding (by the analytical method) increases as the
recovered obtained by continuous low salinity waterflooding at 2 PV. For the conductance ratio increases which generally agrees with the
case with zero vertical permeabilities (no viscous cross flow), the best slug simulation predictions.
size increases from around 1 PV using Rc = 2 to 1.8 PV using Rc = 10 which is
generally matching the analytical model. However, when setting the vertical
permeabilities to be equal to the horizontal permeabilities in the other cases
(dashed lines), the best slug sizes of Rc = 2 and 5 are much larger than 1 PV
which suggests that viscous cross flow affects the simulation predictions.
110
Slug Sizes when Injecting Low Salinity Water Post Immediate Injection of Conventional Water into
2D Layered Models
The best slug size to use when injecting low salinity water post immediate conventional waterflooding is
evaluated for the 2D layered cases with simple heterogeneity variations. The best slug sizes are summarized
in Table 4.4.
Table 4.4: List of different slug sizes of injected low salinity water post conventional waterflooding that were used in five different layered models
with simple heterogeneity variations.
Case
Scenario Pore volume of water injected
Immediate conventional water injected up to
0.43 0.32 0.27 0.28 0.34
breakthrough (PVI))
Injecting low salinity water at the Slug size of injected low salinity water post
1.57 1.68 1.73 1.72 1.66
breakthrough of immediate immediate conventional waterflooding (PVI)
conventional waterflooding
Total water injected (PVI) 2 2 2 2 2
Immediate conventional waterflooding (PVI) 0.8 0.8 0.8 0.8 0.8

Injecting low salinity water post
Slug size of injected low salinity water post
0.8 PV of injecting immediate 1.2 1.2 1.2 1.2 1.2
immediate conventional waterflooding (PVI)
Figure 4.21 shows the oil recovery at 2 PV of total water injected for different slug sizes of low salinity
water post conventional waterflooding. Generally, the earlier injection of low salinity water post
conventional waterflooding, the higher the oil recovery is obtained. However, the balance between viscous
cross flow and gravity affects the best slug size of low salinity water depending on the heterogeneity
variations.
For the HL case, the oil recovered by injecting low salinity water post immediate conventional waterflooding
is relatively insensitive to slug size which results in very similar oil recoveries at 2 PV of total water
injected. Gravity drives the injected low salinity water into the lower layer with low permeability while the
viscous forces drive the injected low salinity water into the top layer with high permeability. These two
forces compete so that the final oil recoveries are almost similar for different slug sizes of injected low
salinity water. Figure 4.22 shows the oil saturation distribution at 2 PV of total water injection for three
different slug sizes of low salinity water injected post immediate conventional waterflooding for the 2D
layered models. For the different slug sizes, there is a region of higher oil saturation within the low
permeability layer (a green circle). The size of this circle decreases as the slug size of low salinity water
decreases. This is due to the fact that at larger sizes, most of the top high permeability layer is swept and
then mixing via cross flow occurs between the low permeability layer and high permeability layer. At
smaller slug sizes (injecting low salinity water post 1.6 PV of injecting conventional water), mixing via cross
111
flow is less compared to the larger slug sizes. This is not seen when compared with conventional
waterflooding in Figure 4.23.
For the LH case, gravity and viscous forces act together to impact the vertical sweep efficiency by driving
low salinity water to channel through the bottom layer with high permeability leaving a huge amount of
unswept oil in the top layer with low permeability. Therefore, a poor vertical sweep efficiency is observed as
shown in Figure 4.22. Unlike the HL case, the LH case is sensitive to the slug size of low salinity water.
Delaying the start of low salinity water injection post immediate conventional waterflooding results in much
reduced overall oil recoveries compared to immediate injection of low salinity water.
For the LHLHL and HLHLH cases, the macroscopic sweep efficiency is similar to the HL case since the
gravity and viscous forces compete to minimize the viscous channelling of water. However, the drop in the
overall oil recoveries by small slugs of low salinity water is slightly lower in the LHLHL and HLHLH cases
compared to the drop in the HL case. This is due to the fact that they have larger surface area of low-high
permeability contact compared to the HL case which increases the viscous cross flow.
For the homogeneous case, viscous forces are negligible compared to gravity since the permeabilities are
constant. This leads to unswept oil in the top layer (due to gravity) and a reduction in overall oil recovery
when injecting low salinity water post immediate conventional waterflooding compared to immediate
injection of low salinity water.
0.9
0.8
Oil recovery at 2 PV of total water injection
0.7
0.6 Immediate conv

Injecting lowsal post BT of conv
0.5
Injecting lowsal post 0.8 PV of conv
0.4 Injecting lowsal post 1.2 PV of conv
0.3 Injecting lowsal post 1.6 PV of conv

Immediate lowsal
0.2
0.1
0
Figure 4.21: Oil recovery at 2 PVI of different slug sizes of low salinity water post immediate conventional waterflooding into 2D layered models.
Generally, the earlier injection of low salinity water post conventional waterflooding, the higher the oil recovery is obtained.
112
Homogeneous
Lowsal post 0.8 PV of conv Lowsal post 1.2 PV of conv Lowsal post 1.6 PV of conv
LH
HL
LHLHL
HLHLH
Oil saturation
0.11 0.29 0.46 0.63 0.80

Figure 4.22: Oil saturation distribution at 2 PV of total water injection for different slug sizes of injecting low salinity water post immediate injection
of conventional water into 2D layered models. In the HL case, gravity drives the injected low salinity water to flow into the lower layer with low
permeability while viscous forces drive the injected low salinity water to flow into the top layer with high permeability. These two forces compete so
that the final oil recoveries are similar for different slug sizes of injected low salinity water. In the LH case, gravity and viscous forces act together to
impact the vertical sweep efficiency by driving low salinity water to channel through the bottom layer with high permeability leaving a significant
amount of unswept oil in the top layer with low permeability resulting in poor vertical sweep efficiency In the LHLHL and HLHLH cases, the
macroscopic sweep efficiency is similar to that in the HL case since gravity and viscous forces are competing. This minimizes the viscous
channelling of water. In the homogeneous case, viscous forces are small compared to gravity since the permeabilities are uniform. This leads to
leaving unswept oil in the top layer (due to gravity) and a significant reduction in overall oil recovery when delaying the start of low salinity water
injection post immediate injection of conventional water compared to immediate injection of low salinity water.
113
HL
Conventional waterflooding
Oil saturation
0.11 0.29 0.46 0.63 0.80

Figure 4.23: Comparison between conventional waterflooding and injecting different slug sizes of low salinity water post conventional waterflooding
for the HL case. The oil saturation distributions are shown at 2 PVI. A region with high oil saturation (a green circle) appears in the low salinity water
figures indicating a mixing via viscous cross flow between the low permeability layer (bottom) and the high permeability layer (top). The size of this
circle decreases as the slug size of low salinity water decreases. This is due to the fact that at larger sizes, most of the top high permeability layer is
swept and then mixing via cross flow occurs between the low permeability layer and high permeability layer. At smaller slug sizes (injecting low
salinity water post 1.6 PV of injecting conventional water), mixing via cross flow is less compared to the larger slug sizes. This is not seen when
compared with conventional waterflooding (bottom figure).

In the previous section, the slug size of injected low salinity water was evaluated for a simple type of
reservoir heterogeneity. In this section, an evaluation is performed for more complex heterogeneities from
geologically realistic models taken from the SPE 10 model 2 for both immediate low salinity waterflooding
and the injection of low salinity water post immediate conventional waterflooding.
Slugs Sizes of Immediate Injection of Low Salinity Water into Areal Models Taken from SPE 10
Model 2
The low salinity water slug sizes used in these simulation runs are: 0.4, 0.6 & 0.8 PV. They are compared
with continuous injection of immediate low salinity water and immediate conventional waterflooding.
Five heterogeneity realizations representing a wide range in heterogeneity (Hv = 0.4 (very heterogeneous
reservoir) to Hv = 0.81 (less heterogeneous reservoir)) were selected from the SPE 10 model 2 for this
analysis. Oil recoveries at 1 PV of total water injection for these realizations are plotted in Figure 4.24.
114
0.8
Immediate conv
0.7
Oil recovery at 1 PV of total water
0.4 PV of Immediate lowsal

0.6
0.5 0.6 PV of Immediate lowsal

injection
0.4 0.8 PV of Immediate lowsal
0.3
Immediate lowsal
0.2
Injecting lowsal post BT of immediate
0.1 conv
0
0.81 0.73 0.63 0.49 0.4
Hv
Figure 4.24: Oil recovery at 1 PV of total water injection versus Hv for different slug sizes of immediate low salinity waterflooding and injecting low
salinity water post immediate conventional waterflooding into five layers taken from SPE 10 model 2. The best slug size of immediate low salinity
waterflooding is between 0.6-0.8 PV and is insensitive to the degree of heterogeneity.
The best slug size of immediate low salinity waterflooding for all five cases is between 0.6- 0.8 PV and is
insensitive to the degree of heterogeneity. It was previously highlighted by Jerauld et al. (2008) that the best
slug size of injecting low salinity water is 0.4 PV. However, this evaluation indicates that reservoir
heterogeneity requires the best slug size to be larger than this.
Slug Sizes of Injecting Low Salinity Water Post Immediate Injection of Conventional Water into Areal
Models Taken from SPE 10 Model 2
The slug sizes of injecting low salinity water post immediate conventional waterflooding are evaluated for
the same five heterogeneity realisations for immediate low salinity waterflooding. The slug sizes of injecting
low salinity water post conventional waterflooding are summarized in Table 4.5. Low salinity water
injection was begun immediately after the breakthrough of the preceding conventional waterflood
Table 4.5: List of different slug sizes of injected low salinity water post conventional waterflooding that were used in the five heterogeneity
realisations taken from SPE 10 model 2.
Hv
0.81 0.73 0.63 0.49 0.40
0.34 0.29 0.28 0.22 0.17
Injecting low salinity water at the breakthrough (PVI))
breakthrough of immediate Slug size of injected low salinity water post
0.66 0.71 0.72 0.78 0.83
conventional waterflooding immediate conventional waterflooding (PVI)
Injecting low salinity water post 0.4
PV of immediate conventional 0.6 0.6 0.6 0.6 0.6
waterflooding
waterflooding
waterflooding
115
Oil recovery at 1 PV of total water injected is plotted for all scenarios and compared with immediate low
salinity waterflooding and immediate conventional waterflooding in Figure 4.25. Higher oil is recovered
using the larger slugs of injecting low salinity water post immediate conventional waterflooding.
0.8
0.7
0.6
Oil recovery at 1 PV of total water injection
Immediate conv
0.5
Injecting lowsal post BT of conv water
0.4
Injecting lowsal post 0.4 PV of conv water
0.3
0.2
0.1
Immediate lowsal
0
0.81 0.73 0.63 0.49 0.4
Hv
Figure 4.25: Oil recovery at 1 PV of different slug sizes of injecting low salinity water post conventional waterflooding for five heterogeneity
realisations taken from SPE 10 model 2. Higher oil is recovered using the larger slugs of injecting low salinity water post immediate conventional
waterflooding.
4.3 Conclusions and Recommendations

In this chapter, the effect of reservoir heterogeneity on the performance of low salinity waterflooding was
investigated for both continuous injection and as a slug injection. Two main models were used for the
evaluation: 2D layered models with simple heterogeneity variations and areal models with more complex
heterogeneity taken from SPE 10 model 2. For both models, the injection of low salinity water was evaluated
for both immediate injection and injection post immediate conventional waterflooding.
For continuous injection of low salinity water into 2D layered models with simple heterogeneity variations,
the study revealed that heterogeneity affects low salinity waterflooding and conventional waterflooding to a
similar extent. However, due to the anticipated mixing via viscous cross flow of the injected low salinity
water with the in situ high salinity water, the resulting incremental oil recovery by low salinity waterflooding
is reduced slightly in very heterogeneous reservoirs compared with the same model realizations when using
conventional waterflooding. Further simulation runs using the 2D layered models and assuming zero vertical
permeabilities (i.e. no viscous cross flow between layers) indicated that low salinity waterflooding is more
affected by reservoir heterogeneity as the low permeability layers are unswept. This results in poor vertical
sweep efficiencies. In layered reservoirs with a large permeability contrast, the efficiency of low salinity
waterflooding is reduced due to two main reasons: channelling of the injected low salinity water front
through the high permeability layers and reduced trailing front velocities through these high permeability
layers due to mixing via viscous cross flow with the other lower permeability layers. The first factor will
lead to poor macroscopic sweep efficiency and the second factor will lead to lower incremental oil recovery
116
by low salinity waterflooding. For such types of layered reservoirs with a high degree of heterogeneity, the
resulting incremental oil recovery might not be sufficient to justify the high initial capital costs and related
operational cost of the required continuous desalination of the injected water.
For the case of continuous injection of low salinity water into areal layers with realistic permeabilities taken
from SPE 10 model 2 (Christie and Blunt 2001), the study revealed that the incremental oil recovery by
immediate low salinity waterflooding over conventional waterflooding is slightly less than when injecting in
a less heterogeneous reservoirs. However, low salinity waterflooding still improves recovery over that
obtained from conventional waterflooding.
For the case of injecting low salinity water immediately as slugs into the 2D layered models, the best slug
size was estimated analytically in two layers case following the approach by Wright et al. (1987). It
increases as the permeability contrast between the layers increases assuming negligible gravity and capillary
forces. It increases from 0.3 PVI for a permeability contrast of 2 to 1.6 PVI for a permeability contrast of 10.
These agree with the findings for the case with zero vertical permeability (i.e. no viscous cross flow).
However, further simulations using isotropic permeabilities in the models indicated that the best slug sizes
need to be larger especially when using low permeability contrasts case (Rc = 2). The analytical solution is a
quick method to estimate the slug size of injecting low salinity water immediately in a field scale. However,
it cannot be used without a prior knowledge of the vertical permeability in the reservoir. At very low kv/kh
ratios, viscous cross flow is negligible thus the analytical method can be used for the estimation of the slug
size of injecting low salinity water. However at very high kv/kh ratios, the analytical method cannot be used
as viscous cross flow affects the performance of low salinity waterflooding thus impacts the best slug size.
Therefore, in such cases the analytical method will lead to lower predictions of the actual best slug size
needed.
For the 2D layered models, when increasing the total water injection to 2 PV, the best slug size of immediate
low salinity waterflooding appears to increase significantly from 1.5 PV for less heterogeneous case to 1.75
PV for more heterogeneous case. This suggests that when implementing low salinity waterflooding
immediately in a field case, a detailed simulation study must be carried out prior to implementing in the field
to determine the best slug size of injecting low salinity water. The related operational cost of desalinating the
injected water is considered to be a major element when planning for the project. For very heterogeneous
cases, the slug size of immediate low salinity waterflooding is expected to be large and therefore might not
be feasible due to the expected low incremental oil recovery. This is related to the fact that mixing via a
viscous cross flow between layers is expected and might severely reduce the performance of low salinity
waterflooding.
For the case of injecting low salinity water immediately as slugs into the areal layers taken from SPE 10
model 2, the best slug size of immediate low salinity waterflooding was found to be 0.6- 0.8 PV and is
insensitive to the degree of heterogeneity. It was previously highlighted by Jerauld et al. (2008) that the best
slug size of 0.4 PV. However, this evaluation indicates that reservoir heterogeneity requires the best slug size
to be larger than this.
117
For the case of injecting low salinity water as slugs post immediate injection of conventional water into both
2D layered models and areal layers taken from SPE 10 model 2, it is recommended to inject low salinity
water post the breakthrough of conventional waterflooding. Delaying the start of low salinity water injection
would lead to lower incremental oil recovery due to mixing via a viscous cross flow with the conventional
water. It is anticipated that in 3D models, gravity would cause additional mixing when compared to 2D
models.
118
CHAPTER 5: IMPACT OF HETEROGENEITY AND PHYSICAL
DIFFUSION AND DISPERSION ON CONTINUOUS LOW SALINITY
WATERFLOODING
In previous chapters, the impact of reservoir heterogeneity was found to have a significant effect on the
performance of low salinity waterflooding. The simulation findings indicate that mixing occurs between the
injected low salinity water and connate water due to a combination of viscous and gravity driven crossflow
and numerical diffusion. This mixing can improve or reduce the effectiveness of low salinity waterflooding.
However the simulations did not model physical diffusion and dispersion. As noted in chapter 2, there is
some discussion in the literature as to whether molecular diffusion and dispersion can influence oil recovery
in the field scale (Mahadevan et al. 2003, Coats et al. 2009) and particularly on its effect on low salinity
waterflooding (Jerauld et al. 2008). In this chapter, the impact of physical diffusion and dispersion on low
salinity waterflooding is investigated. This is achieved by using the CMG's STARS simulator as it models
physical diffusion and dispersion as well as low salinity waterflooding using a salinity threshold model
similar to that implemented in Schlumberger's Eclipse. First, physical diffusion in a system with two layers
is characterized by the transverse dispersion number (NTD) using a single phase simulation on the laboratory
scale. Then, the impact of heterogeneity and physical diffusion/dispersion on continuous low salinity
waterflooding is investigated using laboratory scale models and reservoir scale models. This includes
injection of low salinity water immediately and injection of low salinity water post high salinity
waterflooding.
For the laboratory scale models, three different sets were modelled: a two layer model to investigate the
impact of layer thickness on low salinity waterflooding, a five layer model to investigate the effect of
permeability variations on low salinity waterflooding and a two layer model where one layer was a
sandstone and the other layer was a very low permeability shale saturated with a high salinity water to
investigate the effect of the presence of shales on low salinity waterflooding.
For the reservoir scale models, two different sets were modelled: a ten layer model with complex
heterogeneity taken from SPE 10 model 2 (Christie and Blunt 2001) and a reservoir with a transition zone
that is connected to an aquifer to investigate the effect of the presence of an aquifer on low salinity
waterflooding.
Since the salinity thresholds pose an impact on the predictions and outcomes of the simulation models, they
were considered to be a sensitivity parameter in most of the simulation models conducted on this chapter.
5.1 Quantifying Diffusion/Dispersion
In this chapter, the impact of physical diffusion/dispersion on oil recovery in heterogeneous reservoirs is
quantified using Df (will be defined in section 5.1.2). Physical diffusion/dispersion is characterized by the
transverse dispersion number (NTD) using a single phase simulation study as defined in section 5.1.1.
119
5.1.1 Characterizing Dispersion/Diffusion Using the Transverse Dispersion Number (NTD)
In this section, physical longitudinal and transverse dispersion/diffusion (defined in equations 2.9 and 2.10)
is quantified using the transverse dispersion number NTD (defined in equation 2.11). The velocity dependent
dispersion (𝛼𝛼𝑙𝑙 and 𝛼𝛼𝑡𝑡 ) term in equations 2.9 and 2.10 is ignored, so the longitudinal diffusion is equal to the
transverse diffusion. NTD is used to characterize the flow dominant regions. The heterogeneity variations in
the reservoir affect the sweep efficiency while physical diffusion affects the microscopic displacement. In
order to determine when each term dominates the flow, NTD is used. A single phase simulation was used
with two horizontal layers as shown in Figure 5.1. The model was selected to be horizontal to neglect the
effect of gravity. Two wells were completed: one injection well was completed in the columns along the left
hand face of the model and one production well was completed in columns along the right hand face of the
model. The injection and production rates were constant and equal such that the average pore velocity within
the model was 1.2 cm/hr (similar to the frontal advance rate seen in the field.
INJ PROD
Permeability (md)
100
Figure 5.1: Schematic of the 2D simulation model used to identify flow regions. The model dimensions were 30 × 3.2 × 3.2 cm. It was discretized
into 200 × 22 × 1. The Y/X aspect ratio was changed to 3 instead of 1 to make the figure easier to view. Low salinity water was injected along the
left hand face of the model and then produced with the combined high salinity water along the right hand face of the model.
The simulations were performed on laboratory scale models but the use of a dimensionless number (NTD)
enables the reservoir engineer to apply the results to the reservoir scale, following Rapoport (1955), amongst
others. Capillary pressure was assumed to be negligible in all cases. One layer had a permeability of 100 md
and the other layer had a permeability of 1 md. Several simulations were performed, varying the NTD values
from 0.001 to 111. This was achieved by varying the transverse diffusivity. These simulations were
performed in a horizontal model so gravity did not affect the flow. The model dimensions in addition to the
fluid and rock properties are listed Table 5.1.
Table 5.1: The model dimensions in addition to the fluid and rock properties used in a single phase model to characterize diffusion using NTD from
0.001 to 111.
Property Value
Physical dimensions (cm) 30 × 3.2 × 3.2
Number of grid blocks 200 × 22 × 1
100
Permeabilities (md)
1
Porosity 0.2
Water: 0.9
High salinity water mole fractions
Salt: 0.1
Water: 0.99
Low salinity water mole fractions
Salt: 0.01
120
Figure 5.2 shows the produced salt concentration versus pore volumes injected for three different values of
NTD. Figure 5.3 shows the salinity distribution in the model at 2 PVI for the same three values of NTD.
When NTD is low (0.001) then the impact of transverse diffusion is low, low salinity water travels
preferentially in the high permeability layer and breakthrough of low salinity water is controlled by the
permeability contrast between the two layers. When NTD is high then transverse diffusion is high, so the
salinity of the water in the low permeability layer is increased and conversely salt diffuses rapidly (due to the
high transverse diffusion) from the low permeability layer into the high permeability layer. Breakthrough of
low salinity water is much earlier and is determined by the effective longitudinal diffusion that results from
the combination of flow along the layers and the transverse diffusion between them.
100 Direction of flow

Salt Concentration
Produced salt concentration (%)
90
(mole fraction)
80 NTD = 0.001
70
60
50 NTD=0.001
40 NTD = 0.1
NTD=0.1
30
NTD=111
20
10
NTD = 111
0
0 0.5 1 1.5 2
PVI
Figure 5.2: Plot of produced salt concentration versus time for three different Figure 5.3: the salinity distribution in the model at 2 PVI. Flow is
NTD values. dominated by advection when NTD is low and dominated by
transverse diffusion when NTD is high.
Figure 5.4 shows the PVI at which the produced salinity is halfway between the injected salinity and the
salinity of the connate water as a function of NTD. In a homogeneous system the produced salinity would be
midway between the injected low salinity water and high salinity water at 1 PVI so this shows the impact of
layering on transverse diffusion. Figure 5.5 shows the time at which low salinity water breaks through, again
as a function of NTD. In both figures, it can be seen that there is a change of behaviour for 0.1 < NTD <1 .
Flow is dominated by advection/heterogeneity for NTD < 0.1 and dominated by transverse diffusion for NTD
> 1. This is consistent with what was seen by Lake and Hirasaki (1981) in their simulations.
121
PVI for 50 % salinity at producer 1 0.3
0.9 0.25
LSW breakthrough (PV)

0.8 0.2
0.7 0.15
0.6 0.1
0.5 0.05
0.4 0
0.001 0.01 0.1 1 10 100 1000 0.001 0.01 0.1 1 10 100 1000
NTD NTD
Figure 5.4: The PVI for 50 % salinity at producer. The PVI value Figure 5.5: The change from flow dominated by advection and
increases as NTD increases until it reaches a peak value when (NTD = heterogeneity (layering) to flow dominated by diffusion occurs when
0.2) then decreases. This NTD marks the turning point from reservoir 0.1 < NTD < 1.
heterogeneity dominated flow to diffusion dominated flow.
5.1.2 Quantifying the Impact of Diffusion on Oil Recovery Using Df

Following Brodie and Jerauld (2014), the impact of diffusion on the incremental oil recovery achieved is
characterized by low salinity water injection using:
𝑅𝑅𝑅𝑅𝑑𝑑𝐿𝐿𝐿𝐿 − 𝑅𝑅𝑅𝑅 𝐻𝐻𝐻𝐻
𝐷𝐷𝑓𝑓 = … … . (5.1)
𝑅𝑅𝑅𝑅𝐿𝐿𝐿𝐿 − 𝑅𝑅𝑅𝑅𝐻𝐻𝐻𝐻
where 𝑅𝑅𝑅𝑅𝑑𝑑𝐿𝐿𝐿𝐿 is the oil recovery from low salinity waterflooding including molecular diffusion, 𝑅𝑅𝑅𝑅 𝐿𝐿𝐿𝐿 is the oil
recovery from low salinity waterflooding excluding molecular diffusion and 𝑅𝑅𝑅𝑅 𝐻𝐻𝐻𝐻 is the oil recovery from
conventional waterflooding. Df varies from 0 to greater than 1. At a value of 1, molecular diffusion has no
effect on low salinity waterflooding while at 0 there is no benefit from low salinity waterflooding whilst
greater than 1 suggests diffusion improves recovery. In this chapter, the oil recovery is measured at 1 PVI.
5.2 Method
Relative Permeability Curves
The high and low salinity water relative permeability curves (taken from a core flood experiment by
Callegaro et al. (2013)) used in this chapter are shown in Figure 5.6.
1
0.8
0.6 krwH
kroH
0.4
krwL
0.2 kroL
0
0 0.2 0.4 0.6 0.8
Sw
Figure 5.6: The high and low salinity water relative permeability curves used in the simulations.
122
These relative permeability curves were chosen as they are more realistic since they are taken from a core
flood experiment instead of the synthetic relative permeability used in the previous chapters. To compare
both sets of relative permeability curves, the fractional flow curves are shown in Figure 5.7. The fractional
flow curves for the relative permeability curves used in this chapter has a higher velocity of the leading
𝑑𝑑𝑑𝑑𝐻𝐻𝐻𝐻,𝑤𝑤
shock front which results in an earlier water breakthrough. In addition due to the larger difference in
𝑑𝑑𝑆𝑆𝑤𝑤
𝑑𝑑𝑑𝑑𝐻𝐻𝐻𝐻,𝑤𝑤 𝑑𝑑𝑑𝑑𝐿𝐿𝐿𝐿,𝑤𝑤
the velocities of its two shock fronts ( and ) a larger size of connate water bank forms compared
𝑑𝑑𝑑𝑑𝑤𝑤 𝑑𝑑𝑑𝑑𝑤𝑤
with the fractional flow curves used in the previous chapters. This is confirmed by comparing the water
saturation profile at 0.06 PV for both sets of relative permeability curves (Figure 5.8).
1 1
0.9 Sw,HS = 0.52 0.9 Sw,HS = 0.37
Sw,LS = 0.52
Sw,LS = 0.72
0.8 𝒅𝒅𝒇𝒇𝑯𝑯𝑯𝑯,𝒘𝒘 0.8 𝒅𝒅𝒅𝒅𝑯𝑯𝑯𝑯,𝒘𝒘
= 2.06 = 𝟓𝟓. 𝟏𝟏𝟏𝟏 𝒅𝒅𝒅𝒅𝑳𝑳𝑳𝑳,𝒘𝒘
𝒅𝒅𝑺𝑺𝒘𝒘 𝒅𝒅𝒇𝒇𝑳𝑳𝑳𝑳,𝒘𝒘 𝒅𝒅𝒅𝒅𝒘𝒘 = 𝟏𝟏. 𝟖𝟖𝟖𝟖
0.7 = 1.32 Fractional Flow 0.7 𝒅𝒅𝒅𝒅𝒘𝒘
Fractional Flow
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Sw Sw
HS LS HS LS
a) b)
Figure 5.7: Comparison between the fractional flow (fw) curves for the different sets of relative permeability curves: a) fw curves for the relative
permeability curves used in the previous chapters and b) fw curves for the relative permeability curves used in this chapter. The latter has a higher
𝑑𝑑𝑑𝑑
velocity of the leading shock front 𝐻𝐻𝐻𝐻,𝑤𝑤 which results in an earlier water breakthrough. In addition, due to the larger difference in the velocities of
𝑑𝑑𝑑𝑑𝑤𝑤
𝑑𝑑𝑑𝑑𝐻𝐻𝐻𝐻,𝑤𝑤 𝑑𝑑𝑑𝑑𝐿𝐿𝐿𝐿,𝑤𝑤
the shock fronts ( and ) of the latter fw curves, a larger size of connate water bank forms compared with the former fw curves.
𝑑𝑑𝑑𝑑𝑤𝑤 𝑑𝑑𝑑𝑑𝑤𝑤
123
1
Water Saturation
0.8
0.6
0.4
0.2
0
0 200 400 600 800 1000 1200
Distance (m)
LS HS
a) Water saturation profile at 0.06 PVI for the relative permeability curves used in the previous chapters and the model described in section 3.1
0.8
Water Saturation
0.6
0.4
0.2
0
0 200 400 600 800 1000 1200
Distance (m)
LS HS
b) Water saturation profile at 0.06 PVI for the relative permeability curves used in this chapter and the model described in section 3.1
Figure 5.8: Comparison between the water saturation profiles (calculated analytically) at 0.06 PV for both sets of relative permeability curves and the
model described in section 3.1. The velocity of the leading shock front using the relative permeability curves in this chapter (b) is much higher than
that using the relative permeability curves in the previous chapters (a) which results in a quicker water breakthrough. In addition, the size of the
connate water bank for the latter case is larger compared to the former case.
To test how low salinity wateflooding is impacted by using a different set of relative permeability curves, the
model described in section 3.1 is used. The simulation outputs of the oil recoveries and produced water cut
for both sets of relative permeability curves (Figure 5.9). Again, an earlier water breakthrough and larger
size of connate water bank is seen when using the relative permeability curves in this chapter.
0.9 1
0.8 0.9
0.7 0.8
Oil recovery (fraction)
Water cut (fraction)
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2 0.2
0.1 0.1
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
PVI PVI
LSW (previous chapters rel perm) LSW (previous chapters rel perm)
HSW (previous chapters rel perm) HSW (previous chapters rel perm)
LSW (chapter 5 rel perm) LSW (chapter 5 rel perm)
HSW (chapter 5 rel perm) HSW (chapter 5 rel perm)
a) Oil recovery plot b) Water cut profile
Figure 5.9: Comparison between the simulated outputs of the model described in section 3.1 using different sets of relative permeability curves. LSW
stands for low salinity waterflooding while HSW stands for high salinity waterflooding. Using the relative permeability curves in this chapter results
in earlier water breakthrough and larger size of the connate water bank.
124
5.2.1 Immediate Injection of Low Salinity Waterflooding
The impact of heterogeneity and physical diffusion/dispersion on low salinity waterflooding is evaluated
using various models from the laboratory scale to the reservoir scale. Table 5.2 summarizes the cases
examined for continuous low salinity waterflooding.
Table 5.2: Cases examined for the impact of physical diffusion/dispersion and heterogeneity on continuous low salinity waterflooding.
Properties
Scale Model Number of grid blocks Physical dimensions Permeabilities Range of values
varied
Case 1 Case 2
100 md Layer thickness
Two layer 200 × 1 × 10 30 × 3.2 × 3.2 cm 0.317 0.951
1 md (cm)
2.853 2.219
Realisation 1:
1000, 100, 10, 1, 0.1
Two different Permeability of
Five layer 200 × 1 × 50 30 × 3.2 × 3.2 cm Realisation 2:
realisations layers (md)
1000, 500, 250, 125,
Laboratory
62.5
Shale thickness
2, 8
(cm)
Sand: 1.2 d System length
Sand + shale 200 × 1 × 50 30 × 1 × 10 cm 30, 150, 300, 1500
Shale: 0.1 md (cm)
Shale porosity
5, 20
(%)
complex
Ten layer heterogeneity
model taken that
from SPE 10 represents a
60 × 60 × 10 1,200 × 600 × 20 ft Not applicable Not applicable
model 2 geologically
(Christie and realistic field
Blunt 2001) (0.001 to
20,000 md)
Oil zone:
Reservoir
50 × 1 × 7
Transition zone: Oil zone:
Reservoir
1,000 × 50 × 35 ft
with a 50 × 1 ×10
Aquifer: Transition zone:
transition
1,000 × 50 × 20 ft 100 md Not applicable Not applicable
zone Top 35 ft.: 50 × 1 × 7
Aquifer:
connected to Next 40 ft.: 50 × 1 × 4
an aquifer 1,000 × 50 × 355 ft
Next 40 ft.: 50 × 1 × 2
Next 80 ft.: 50 × 1 × 2
Bottom 160 ft.: 50 × 1 × 2
A-Laboratory Scale Models

In this part, the impact of transverse diffusion on oil recovery from low salinity water injection into a range
of 2D layered models is examined on the laboratory scale. The use of a dimensionless number (NTD)
following Rapoport (1955), suggests that the results can be extended to any scale (e.g. reservoir scale).
Capillary pressure is assumed to be negligible in all cases. The fluid properties are summarized in Table 5.3.
Similar to chapter 4, brine density was not modelled as it has a negligible effect on low salinity
waterflooding.
125
Table 5.3: The rock and fluid properties used in the laboratory scale models to evaluate the impact of transverse diffusion on low salinity
waterflooding.
Property Value
Oil and water viscosity (cp) 0.4
Oil and water density (kg/m3) 829, 1000
Porosity 0.2
Connate water salinity (1,000 ppm) 30
Injected water salinity (1,000 ppm) 1
High salinity water:
0.35
Residual oil saturation
Low salinity water:
0.28
Two scenarios:
1- Lower threshold: 1
Upper threshold: 30
Salinity thresholds (1,000 ppm)
2- Lower threshold: 1
Upper threshold: 7
For all models, a 2D line drive was used. The injection well was completed in the columns along the right
hand face of the model and the production well was completed in columns along the left hand face of the
model. The injection and production rates were constant and equal such that the average pore velocity within
the model was 1.2 cm/hr (similar to the frontal advance rate seen in the field). These simulations were
performed in heterogeneous models initially saturated with oil and a connate water saturation of 0.24. One
set considered a two layer model to investigate the impact of layer thickness on low salinity waterflooding,
another set of simulations used a five layer model with two different orderings of the layers to investigate the
effect of permeability variations on low salinity waterflooding and a final set considered a two layer model
where one layer was a sandstone and the other was a very low permeability shale saturated with high salinity
water. This last model was similar to that investigated by Brodie and Jerauld (2014) to ascertain the impact
of saturated shale layers on the effectiveness of low salinity waterflooding. In this section, the impact of
shale thickness, system length and shale porosity are investigated.
In all these cases the impact of varying transverse diffusivity while keeping the injection rate constant on oil
recovery was investigated. The grid sizes for the different models and the range of properties varied are
listed in Table 5.2. These models were vertical so gravity could potentially affect the displacement (although
the density difference between fluids was small). Two different salinity thresholds were used in modelling
low salinity waterflooding: 1,000-30,000 ppm and 1,000-7,000 ppm. The salinity thresholds of 1,000-7,000
ppm were chosen as they represent the more realistic thresholds as confirmed by previous experiments (Tang
and Morrow 1999). However, the salinity thersholds of 1,000-30,000 ppm are not realistic. They were
chosen to see the maximum effect of the linear interpolation of relative permeabilities (high and low salinity
water) on modelling low salinity waterflooding since all of the intermediate salinities fall within the linear
interpolation region.
B- 3D Reservoir Scale Models with Complex Heterogeneity (from SPE 10 Model 2)
In this section, the impact of physical dispersion and heterogeneity on low salinity waterflooding is
evaluated using 3D models built from a much more complex heterogeneity model taken from SPE 10 model
2 (Christie and Blunt 2001). This model represents a geologically realistic field.
126
SPE 10 model 2 was previously described in chapter 4. However, neither the impact of gravity nor physical
dispersion was evaluated previously. In this section, the model contains the first top ten layers from SPE 10
model 2 (Tarbert formation). The permeability varies from 0.001 to 20,000 md as seen in Figure 5.10.
PERM (md)
Figure 5.10: Permeability distribution in the 3D model taken from SPE 10 model 2 to evaluate the impact of physical dispersion on low salinity
waterflooding.
The model dimensions were 1,200 × 600 × 20 ft with 60 × 60 × 10 grid blocks so each grid block had
dimensions of 20 × 10 × 2 ft. One injection well and one production well were completed in all layers with a
distance of 1,340 ft between them. With regards to modelling physical dispersion, three scenarios were run:
one without the inclusion of dispersion, one with a transverse dispersivity that is 30 times less than the
longitudinal dispersivity as suggested by Mahadevan et al. (2003) (NTD = 0.18) and the last one with a
transverse dispersivity that is 30 times less than a longitudinal dispersivity that is 5 % of the reservoir length
(NTD = 0.47). Unlike in the 2D layered models where the velocity dependent dispersion was excluded, in
the 3D models, the velocity dependent dispersion is significant on the reservoir scale and the molecular
diffusion is negligible. Similar to the laboratory scale models, two different salinity thresholds were used in
modelling low salinity waterflooding: 1,000-30,000 ppm and 1,000-7,000 ppm. The salinity thresholds of
1,000-7,000 ppm were chosen as they represent the more realistic thresholds as confirmed by previous
experiments (Tang and Morrow 1999). However, the salinity thersholds of 1,000-30,000 ppm are not
realistic. They were chosen to see the maximum effect of the linear interpolation of relative permeabilities
(high and low salinity water) on modelling low salinity waterflooding since all of the intermediate salinities
fall within the linear interpolation region.
C- A Reservoir with a Transition Zone Connected to an Aquifer
Many reservoirs in the oil industry are connected to a large aquifer. Previous work by Law et al. (2014) and
Mahani et al. (2011) have found that low salinity waterflooding is affected by the presence of aquifers.
However, their studies did not include the dilution of the low salinity water effect by the aquifer water due to
physical dispersion. In this section, a simple 2D model (field scale) with an oil zone and a transition zone
connected to a large aquifer was used to evaluate the effect of physical dispersion on the efficiency of low
salinity waterflooding. The homogeneous model used for this evaluation is summarized in Table 5.2 and is
shown in Figure 5.11. The relative permeability curves for high and low salinity water, the salinities of the
aquifer water and the injected water are similar to the laboratory scale models. The salinity thresholds were
127
set to be 1,000-7,000 ppm. These thresholds were chosen as they represent the more realistic thresholds as
confirmed by previous experiments (Tang and Morrow 1999).
With regards to modelling physical dispersion, three scenarios were run: one without the inclusion of
dispersion, one with a transverse dispersivity that is 30 times less than the longitudinal dispersivity as
suggested by Mahadevan et al. (2003) (NTD = 0.07) and the last one with a transverse dispersivity that is 30
times less than a longitudinal dispersivity that is 5 % of the reservoir length (NTD = 0.2).
Oil saturation
Figure 5.11: Model used to evaluate the effect of physical dispersion on injecting low salinity water in an oil zone connected to an aquifer.
5.2.2 Injection of Low Salinity Water Post High Salinity Waterflooding

A-Laboratory Scale Model
Five different 2D layered models with simple heterogeneity variations were used to investigate the effect of
diffusion and reservoir heterogeneity on the injection of low salinity water post high salinity waterflooding.
The model dimensions and their permeability variations are listed in Table 5.4. Figure 5.12 shows the
simulation models with their permeability variations.
Table 5.4: Description of the 2D layered models used to evaluate the impact of heterogeneity on the injection of low salinity water post high salinity
water using laboratory scale models with five different heterogeneity variations.
Case
Property
Model dimensions
29.8 × 3.17 × 3.17 cm
(x, y, z)
Number of grid
200 × 1 × 20
blocks (x, y, z)
Grid block size
10.41 × 100 × 0.5 cm
(dx, dy, dz)
The top 18z (10 The top 2z (100 Pattern of the top 5z Pattern of the top 2z
Permeability (z) 100 md md) The bottom md) The bottom (10 md), (100 md),
2z (100 md) 18z (10 md) next 2z (100 md) next 5z (10 md)
Homogeneous LH
HL LHLHL HLHLH
Permeability (md)
1 100
Figure 5.12: Permeability variations for the 2D layered models used to evaluate the impact of diffusion and reservoir heterogeneity on the injection
of low salinity water post high salinity waterflooding.
128
One injection well and one production well were used in the models. Both wells were completed vertically in
all layers. The injection well was located in the last column along the right hand face of the model and set to
have a constant injection rate of 0.04 cm3/min while the production well was located in the first column
along the left hand face of the model and set to have a constant production rate of 0.04 cm3/min. Two
different salinity thresholds were used in modelling low salinity waterflooding: 1,000-30,000 ppm and
1,000-7,000 ppm. The salinity thresholds of 1,000-7,000 ppm were chosen as they represent the more
realistic thresholds as confirmed by previous experiments (Tang and Morrow 1999). However, the salinity
thersholds of 1,000-30,000 ppm are not realistic. They were chosen to see the maximum effect of the linear
interpolation of relative permeabilities (high and low salinity water) on modelling low salinity waterflooding
since all of the intermediate salinities fall within the linear interpolation region. Table 5.5 summarizes the
cases examined for the injection of low salinity water post high salinity waterflooding.
Table 5.5: Cases for PV of injected low salinity water post high salinity waterflooding that were used in five different layered models with simple
heterogeneity variations.
Case
0.17 0.03 0.03 0.05 0.05
Injecting low salinity water at the breakthrough (PVI))
breakthrough of immediate Volume of injected low salinity water post
0.83 0.97 0.97 0.95 0.95
conventional waterflooding immediate conventional waterflooding (PVI)
Volume of injected low salinity water post
B- 3D Reservoir Scale Models with Complex Heterogeneity (from SPE 10 Model 2)

Table 5.6 summarizes the cases examined for the injection of low salinity water post high salinity
waterflooding.
Table 5.6: Cases for PV of injected low salinity water post conventional waterflooding that were used in a 3D reservoir scale model with complex
heterogeneity taken from SPE 10 model 2.
Case
No dispersion NTD = 0.47
Immediate conventional water injected up to breakthrough
0.1 0.1
Injecting low salinity water at the (PVI))
breakthrough of immediate conventional Volume of injected low salinity water post immediate
0.9 0.9
waterflooding conventional waterflooding (PVI)
Total water injected (PVI) 1 1
Immediate conventional waterflooding (PVI) 0.4 0.4
Injecting low salinity water post 0.4 PV of
Volume of injected low salinity water post immediate
injecting immediate conventional 0.6 0.6
conventional waterflooding (PVI)
waterflooding
Immediate conventional waterflooding (PVI) 0.6 0.6
Injecting low salinity water post 0.6 PV of
Volume of injected low salinity water post immediate
injecting immediate conventional 0.4 0.4
conventional waterflooding (PVI)
waterflooding
129
Table 5.7 summarizes the cases examined for the injection of low salinity water post high salinity
waterflooding.
Table 5.7: Cases for PV of injected low salinity water post conventional waterflooding that were used in a reservoir with a transition zone connected
to an aquifer.
Case
Scenario PV of oil zone
No dispersion NTD = 0.07 NTD = 0.2
Immediate conventional water injected up
0.2 0.2 0.2
to breakthrough (PV of oil zone)
Injecting low salinity water at the
breakthrough of immediate
immediate conventional waterflooding 4.8 4.8 4.8
(PV of oil zone)
Total water injected (PV of oil zone) 5 5 5
Immediate conventional waterflooding
0.5 0.5 0.5
(PV of oil zone)
Injecting low salinity water post 0.5 PV
of injecting immediate conventional
immediate conventional waterflooding 4.5 4.5 4.5
waterflooding
(PV of oil zone)
1 1 1
(PV of oil zone)
Injecting low salinity water post 1 PV of
injecting immediate conventional
immediate conventional waterflooding 4 4 4
waterflooding
(PV of oil zone)
2 2 2
(PV of oil zone)
immediate conventional waterflooding 3 3 3
waterflooding
(PV of oil zone)
3 3 3
(PV of oil zone)
Volume of injected low salinity water
post immediate conventional 2 2 2
waterflooding
waterflooding (PV of oil zone)
5.3 Results
5.3.1 Immediate Injection of Low Salinity Waterflooding
A- Laboratory Scale Models
In this section, the outcomes of the evaluation of the impact of physical diffusion and heterogeneity on
continuous low salinity water injection using the laboratory scale models are presented.
I. Two Layers
Figure 5.13 shows the cumulative oil recovery versus PVI for three different values of NTD for the two
different salinity thresholds and Figure 5.14 shows the produced salt concentration versus PVI.
When the salinity thresholds are 1,000-30,000 ppm, most of the salinities fall between the salinity thresholds
so the flow of water with intermediate salinities is controlled by the intermediate salinity water relative
permeability curves that are calculated using the linear interpolation between the high and low salinity water
relative permeability curves. As a result, the overall oil recovery for all cases is higher compared with the
case using salinity thresholds of 1,000-7,000 ppm. Another key difference is that the separation between the
oil recovery curves for the thin high k layer case is larger for the 1,000-30,000 ppm salinity thresholds
compared with the other salinity thresholds. This means that for the thin high k layer case, low salinity
130
ZDWHUIORRGLQJ LV PRUH DIIHFWHG E\ SK\VLFDO GLIIXVLRQ ZKHQ WKH UDQJH RI WKH VDOLQLW\ WKUHVKROGV LV ODUJHU
+RZHYHUIRUWKHSSPVDOLQLW\WKUHVKROGVZKHQWKHKLJKNOD\HULVWKLFNHUWKHRLOUHFRYHU\LV
PRUHDIIHFWHGE\GLIIXVLRQHVSHFLDOO\GXULQJWKHSURGXFWLRQRIWKHFRQQDWHZDWHUEDQN39'XULQJ
WKLV SHULRG SK\VLFDO GLIIXVLRQ GRPLQDWHV WKH IORZ DQG VLQFH WKH UDQJH EHWZHHQ WKH VDOLQLW\ WKUHVKROGV DUH
QDUURZFRPSDUHGZLWKWKHRWKHUVDOLQLW\WKUHVKROGVPRVWRIWKHVDOLQLWLHVLQWKHPRGHODUHDERYHWKHXSSHU
VDOLQLW\ WKUHVKROG VR WKH IORZ RI ZDWHU ZLWK LQWHUPHGLDWH VDOLQLWLHV LV FRQWUROOHG E\ WKH KLJK VDOLQLW\ ZDWHU
UHODWLYHSHUPHDELOLW\FXUYHV$VDUHVXOWRLOPRYHVPRUHVORZO\FRPSDUHGZLWKWKHRWKHUVDOLQLW\WKUHVKROGV
ZKHUH IORZ LV FRQWUROOHG E\ HLWKHU LQWHUPHGLDWH VDOLQLW\ ZDWHU UHODWLYH SHUPHDELOLW\ FXUYHV RU ORZ VDOLQLW\
ZDWHUUHODWLYHSHUPHDELOLW\FXUYHV7KLVKDVUHVXOWHGLQDORZHURYHUDOORLOUHFRYHU\IRUWKHVDOLQLW\WKUHVKROGV
RISSP
/DERUDWRU\H[SHULPHQWVE\7DQJDQG0RUURZKDYHVKRZQWKDWLQFUHPHQWDORLOUHFRYHU\LVREVHUYHG
ZKHQ WKH LQMHFWHG ZDWHU VDOLQLW\ LV EHORZ SSP FRPSDUHG ZLWK WKH FRQQDWH ZDWHU RI SSP
7KHUHIRUH LW LV PRVW OLNHO\ WKDW WKH VDOLQLW\ WKUHVKROGV IRU DQ\ PRGHO DUH FORVHU WR SSP WKDQ
SSP,WFDQEHFRQFOXGHGWKDWDWWKHHQGRIGHSOR\LQJDORZVDOLQLW\ZDWHUSURMHFWLQDQ\WZR
OD\HUPRGHOWKDWWKHSHUIRUPDQFHRIORZVDOLQLW\ZDWHUIORRGLQJLVPRVWDIIHFWHGE\SK\VLFDOGLIIXVLRQDVWKH
WKLFNQHVVRIWKHKLJKSHUPHDELOLW\OD\HUGHFUHDVHV+RZHYHUWKHHIIHFWLVPRUHSURQRXQFHGIRUWKLFNHUKLJK
SHUPHDELOLW\OD\HUVEXWRQO\GXULQJWKHSURGXFWLRQRIWKHFRQQDWHZDWHUEDQN
35 35
Connate water bank
30 30
Oil Recovery (%)
Oil Recovery (%)
25 25
20 20
15 15
10 10
5 5
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
PVI PVI
NTD=0.002 (thin high k) NTD=0.002 (thin high k)
NTD=0.2 (thin high k) Thin high k layer: NTD=0.2 (thin high k) Thin high k layer:
NTD=21 (thin high k) 0.317 cm NTD=21 (thin high k) 0.317 cm
NTD=0.006 (thick high k) Thick high k layer: NTD=0.006 (thick high k) Thick high k layer:
NTD=0.2 (thick high k) 0.951 cm NTD=0.2 (thick high k) 0.951 cm
NTD=21 (thick high k) NTD=21 (thick high k)

)LJXUHDVDOLQLW\WKUHVKROGVSSPEVDOLQLW\WKUHVKROGVSSP
&XPXODWLYHRLOUHFRYHU\REWDLQHGIURPORZVDOLQLW\ZDWHULQMHFWLRQLQWRDWZROD\HUPRGHOVKRZLQJWKHLPSDFWRI17'DQGWKHWKLFNQHVVRIWKHKLJK
SHUPHDELOLW\OD\HUD)RUWKHSSPVDOLQLW\WKUHVKROGVWKHLPSDFWRI17'RQRLOUHFRYHU\LVPRUHSURQRXQFHGIRUWKLQKLJKNOD\HUEIRU
WKHSSPVDOLQLW\WKUHVKROGVWKHLPSDFWRI17'RQRLOUHFRYHU\LVVLJQLILFDQWIRUWKHWKLFNKLJKNOD\HUFDVHGXULQJWKHSURGXFWLRQRIWKH
FRQQDWHZDWHUEDQN39,VLQFHGLIIXVLRQGRPLQDWHVWKHIORZGXULQJWKLVWLPHDQGPRVWRIWKHVDOLQLWLHVLQWKHPRGHODUHDERYHWKHXSSHU
VDOLQLW\ WKUHVKROG VR WKH IORZ RI ZDWHU ZLWK LQWHUPHGLDWH VDOLQLWLHV LV FRQWUROOHG E\ WKH KLJK VDOLQLW\ ZDWHU UHODWLYH SHUPHDELOLW\ FXUYHV 7KLV KDV
UHVXOWHGLQDQRLOUHFRYHU\WKDWLV VLJQLILFDQWO\ORZHUWKDQWKHRWKHUVDOLQLW\WKUHVKROGV
131
100 100

80 80
60 60
40 40
20 20
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 PVI 0.6 0.8 1
PVI
NTD=0.002 (0.3 cm high k thickness) NTD=0.002 (0.3 cm high k thickness)
NTD=21 (0.3 cm high k thickness) NTD=21 (0.3 cm high k thickness)
NTD=21 (0.9 cm high k thickness) NTD=21 (0.9 cm high k thickness)
Figure 5.14: a) salinity thresholds: 1,000-30,000 ppm b) salinity thresholds 1,000-7,000 ppm
Produced salt concentration versus time. Generally there is no notable difference in the produced salt concentration when changing the salinity
thresholds because the effluent concentration profiles depend on the diffusion constant only.
Figure 5.15 shows the oil recovery at the time when the water salinity in the producer has fallen to halfway
between the connate water and the injected low salinity water for the two different salinity thresholds. By
this measure there is a clear change in oil recovery with NTD in both models. Generally, the curves show
that recovery from the model with the thinner high permeability layer is more sensitive to transverse
diffusion. Although the curves are similar for the thin high k layer case for both salinity thresholds, there is a
clear difference when the high k layer is thicker. For the 1,000-30,000 ppm salinity thresholds case, the
curve is constant and then increases to a maximum value at NTD ~ 1 and then decreases, while for the 1,000-
7,000 ppm salinity thresholds, it decrease at NTD ~ 0.4 and then increases and decreases again. This
confirms that the effect of diffusion is more significant during the production of the connate water bank
which is about the time that produced salt concentration is about halfway between the connate water and the
injected low salinity water (50 %).
132
25 25
Oil recovery at 50 % LSW concentration (%)
Oil recovery at 50 % LSW concentration (%)

20 20
15 15
10 10
5 5
0 0
0.0001 0.001 0.01 0.1 1 10 100 1000 0.0001 0.001 0.01 0.1 1 10 100 1000
NTD NTD
thin high k thick high k thin high k thick high k
Oil recovery when the produced salt has fallen to 50 % as a function of NTD. The peak oil recovery is at NTD ~ 1. For lower values of NTD, oil
recovery is dominated by layering whilst for higher values it is dominated by diffusion.
Figure 5.16 shows the incremental oil recovered by low salinity water injection as a function of NTD for the
two salinity thresholds. Diffusion has very little impact on the incremental oil recovery from low salinity
water injection when the high k layer is thicker. More surprisingly it shows that diffusion actually improves
incremental oil recovery when the higher k layer is thinner. This is because the salt in the high salinity
connate water bank in the lower k layer is decreased as a result of diffusion of the salt into the low salinity
water in the higher k layer. A key observation is that for the thin high k layer case, the Df curves are different
for the two salinity thresholds especially at intermediate NTD values (0.1-1). Df value increases in this
region for the salinity thresholds of 1,000-30,000 ppm while it decreases and then sharply increases when
NTD is 1. To better understand the reason behind this differences, oil saturation and water salinity
distributions at 1 PVI are shown in Figures 5.17 and 5.18.
3.5 3.5
3 3
2.5 2.5
2 2
Df
Df
1.5 1.5
1 1
0.5 0.5
0 0
0.0001 0.001 0.01 0.1 1 10 100 1000 0.0001 0.001 0.01 0.1 1 10 100 1000
NTD NTD
thin high k thick high k thin high k thick high k
Incremental oil recovery from low salinity water injection (Df ) as defined by equation 5.1 as a function of NTD (measured at 1 PVI). Maximum
incremental oil recovery is achieved at NTD ~ 1. At very low values of NTD, diffusion has a negligible effect on low salinity waterflooding so Df is
1.
133
Direction of flow
Oil saturation at 1 PVI Water salinity at 1 PVI Oil saturation at 1 PVI Water salinity at 1 PVI
NTD = 0.006 (thin high k) Salinity (1,000 ppm)

NTD = 0.002 (thick high k)
30
27
24
21
NTD = 0.2 (thin high k) NTD = 0.2 (thick high k)
18
15
13
10
NTD = 21 (thin high k) NTD = 21 (thick high k)
7
Oil saturation
Figure 5.17: Oil saturation and water salinity distributions at 1 PV of low salinity water injected into a two layer model using salinity thresholds of
1,000-30,000 ppm. The model with a thin high k layer is shown on the left and the model with the thicker high k layer is shown on the right.
Transverse diffusion, as quantified by NTD, has the most effect on oil recovery in the model with the thinner high k layer. At 1 PVI, transverse
diffusion appears to improve oil recovery by allowing more low salinity water into the low permeability layer and improving sweep. The effect is
largest in the case with the ‘thief’ zone.
Direction of flow
Oil saturation at 1 PVI Water salinity at 1 PVI Oil saturation at 1 PVI Water salinity at 1 PVI

NTD = 0.006 (thin high k) 30 NTD = 0.002 (thick high k)
27
24
21
NTD = 0.2 (thin high k) NTD = 0.2 (thick high k)
18
15
13
NTD = 21 (thin high k) 10 NTD = 21 (thick high k)

7
Oil saturation
Figure 5.18: Oil saturation and water salinity distributions at 1 PV of low salinity water injected in a two layer model using salinity thresholds of
1,000-7,000 ppm. A key difference compared with the other salinity thresholds is that for the thin high k layer at NTD = 0.2, a very high oil saturation
region is seen on the lower left corner (yellow and orange) since most of the intermediate water salinities that formed due to diffusion are above the
upper salinity thresholds so the flow of water with intermediate salinities is controlled by the high salinity water relative permeability curves which
resulted in a lower oil recovery compared with the no diffusion case ( i.e. a drop in the values of Df ).
134
II. Five Layers
Oil recovery is plotted versus PVI in Figure 5.19 for three NTD values and their corresponding produced salt
concentration versus PVI is shown in Figure 5.20. Again it can be seen that the permeability of the layers has
a much greater effect on oil recovery than transverse diffusion as quantified by NTD for the salinity
thresholds of 1,000-30,000 ppm. However, diffusion has a greater impact on oil recovery for both
heterogeneity realisations for the salinity thresholds of 1,000-7,000 ppm especially during the production of
the connate water bank. The reason behind this is similar to the two layer model, during the production of
the connate water bank, diffusion dominates the flow so most of the associated produced water salinities are
above the upper salinity threshold so the flow of water with intermediate salinities is controlled by the high
salinity water relative permeability curves. Another key observation on the 1,000-7,000 salinity thresholds is
that the effect of high diffusivity (NTD = 108) in the double k case is much greater compared with the log k
case because heterogeneity effects are much smaller compared with the log k case while diffusion effects are
equal for the two different salinity thresholds.
45 45
40 40
35 35
Oil recovery (%)
Oil Recovery (%)
30 30
25 25
20 20
15 15
10 10
5 5
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
PVI NTD=0.006 (log k increase)
PVI
NTD=0.006 (log k increase)
NTD=0.2 (log k increase) NTD=0.2 (log k increase)
log k increase (md): log k increase (md):
NTD=64 (log k increase) NTD=64 (log k increase)
0.1, 1, 10, 100 and 1000 0.1, 1, 10, 100 and 1000
NTD=0.01 (double k increase) NTD=0.01 (double k increase)
double k increase (md): double k increase (md):
NTD=0.1 (double k increase) NTD=0.1 (double k increase) 62.5, 125, 250, 500 and 1000
62.5, 125, 250, 500 and 1000
NTD=108 (double k increase) NTD=108 (double k increase)
Oil recovery obtained by low salinity water injection into a five layer model as a function of NTD and layer permeability. Heterogeneity appears to
have a much greater impact on recovery than transverse diffusion, regardless of the value of NTD for the 1,000-30,000 ppm. However, diffusion has a
more pronounced effect on oil recovery for both heterogeneity realisations for the salinity thresholds of 1,000-7,000 ppm with the greatest effect
during the production of the connate water bank because diffusion dominates the flow so most of the associated produced water salinities are above
the upper salinity threshold so the flow of water with intermediate salinities is controlled by the high salinity water relative permeability curves.
Another key observation on the 1,000-7,000 salinity thresholds is that the effect of high diffusivity (NTD = 108) in the double k case is much greater
compared with the log k case because the effect of reservoir heterogeneity is much smaller compared with the log k case while diffusion is equal for
the two different salinity thresholds.
135
100 100
Produced low salt concentration (%)

Produced low salt concentration (%)
80 80
60 60
40 40
20 20
0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
PVI PVI
NTD=0.006 (Log k increase) NTD=0.006 (Log k increase)
NTD=0.2 (log k increase) NTD=0.2 (log k increase)
NTD=64 (log k increase) NTD=64 (log k increase)
NTD=0.01 (double k increase) NTD=0.01 (double k increase)
ND=0.1 (double k increase) ND=0.1 (double k increase)
NTD=108 (double k increase) NTD=108 (double k increase)
Produced salt concentration versus PVI. There is no noticeable difference between the two different salinity thresholds because the effluent
concentration profiles depend on the diffusion constant only.
Figures 5.21 shows that NTD does indeed quantify transverse diffusion with a slightly higher oil recovery at
NTD = 1 for the salinity thresholds of 1,000-30,000 ppm.
40 40
Oil recovery at 50 % LSW concentration
Oil recovery at 50 % LSW concentration
35 35
30 30
25 25
20 20
15 15
10 10
5 5
0 0
0.0001 0.001 0.01 0.1 1 10 100 1000 0.0001 0.001 0.01 0.1 1 10 100 1000
NTD NTD
Log k Increase double k increase Log k Increase double k increase
Oil recovery at the time when the salinity at the producer is midway between the high and low salinity values for the two different five layer
models. NTD again appears to quantify transverse diffusion in the reservoir.
Figure 5.22 shows that, as for the two layer model, diffusion can actually improve the incremental oil
recovery from low salinity water injection as it reduces the salinity of the connate water bank in the lower
permeability layer. However, at intermediate NTD values (0.01-0.1) diffusion decreases the oil recovery for
both permeability realisations using salinity thresholds of 1,000-7,000 ppm. This is due to the fact that
diffusion is small compared with reservoir heterogeneity at these intermediate NTD values. Low levels of
diffusion increase the salinity of the injected low salinity water in the layers with intermediate permeabilities
(middle layers). This results in oil that is not contacted by the injected low salinity water. This is because
136
most of these intermediate salinities are above the upper salinity threshold so the flow of water with
intermediate salinities is controlled by the high salinity water relative permeability curves. As a result, the
overall oil recovery is lower than the no diffusion case. The effect of diffusion on oil recovery can be
visualized by inspecting the oil saturation and water salinity distributions at 1 PVI at different values of NTD
for the two different salinity thresholds (Figures 5.23 and 5.24).
1.4 1.4
1.2 1.2
1 1
0.8 0.8
0.6 0.6
Df
Df
0.4 0.4
0.2 0.2
0 0
0.0001 0.001 0.01 0.1 1 10 100 1000 0.0001 0.001 0.01 0.1 1 10 100 1000
NTD NTD
Log k Increase double k increase Log k Increase double k increase
The incremental oil recovery from low salinity waterflooding (Df ) as defined by equation 5.1. at 1 PV of low salinity water injected into a five layer
model as a function of NTD. Generally, diffusion improves the oil recovery for NTD > 1. However, for salinity thresholds of 1,000-7,000 ppm,
diffusion decreases oil recovery at 0.01 < NTD < 0.1 since the effect of diffusion is small compared with the effect of reservoir heterogeneity at
these intermediate NTD values. Low levels of diffusion increase the salinity of the injected low salinity water which results in oil that is not
contacted by the injected low salinity water. This is because most of these intermediate salinities are above the upper salinity threshold so the flow
of water with intermediate salinities is controlled by the high salinity water relative permeability curves. As a result, the overall oil recovery is
lower than the no diffusion case.
137
Direction of flow
Oil saturation at 1 PV Water salinity at 1 PV Oil saturation at 1 PV Water salinity at 1 PV
Salinity
NTD = 0.006 (log k increase) (1,000 ppm) NTD = 0.01 (double k increase)
30
27
24
NTD = 0.2 (log k increase) 21 NTD = 0.1 (double k increase)
18
15
13
NTD = 64 (log k increase) 10 NTD = 108 (double k increase)
1
Oil saturation
Figure 5.23: Oil saturation and water salinity distributions at 1 PV of low salinity water injected into a five layer model as a function of NTD for the
salinity thresholds of 1,000-30,000 ppm. It can be seen that transverse diffusion has only a small effect on oil recovery. The major impact is due to
permeability variations.
Oil saturation at 1 PV Water salinity at 1 PV Direction of flow Oil saturation at 1 PV Water salinity at 1 PV
NTD = 0.006 (log k increase) Salinity NTD = 0.01 (double k increase)
(1,000 ppm)
30
27
24
NTD = 0.2 (log k increase) NTD = 0.1 (double k increase)
21
18
15
13
NTD = 64 (log k increase) NTD = 108 (double k increase)
10
1
Oil saturation
Figure 5.24: Oil saturation and water salinity distributions at 1 PV of low salinity water injected into a five layer model as a function of NTD for the
salinity thresholds of 1,000-7,000 ppm. A key difference compared with the other salinity thresholds is that higher oil saturations are observed (light
green) in the double k cases since most of the intermediate water salinities that formed due to diffusion are above the upper salinity thresholds so the
flow of water with intermediate salinities is controlled by the high salinity water relative permeability curves which resulted in a lower oil recovery
compared with the no diffusion case (i.e. a drop in Df values compared with the other salinity thresholds).
138
III. Shaly Sand Lab Scale Models
Effect of Shale Thickness
Figure 5.25 shows the incremental oil recovered from low salinity water injection as a function of NTD for
two different shale thicknesses using two different salinity thresholds. It can be seen that, as might be
expected, the thicker shale has a greater, adverse impact on the oil recovery when transverse diffusion is
higher. It is noted that in this case the effect is seen for NTD > 0.001. The thicker shale has a larger impact
because it contains a larger volume of saline water that reduces the beneficial effects of low salinity water
when there is transverse diffusion. High diffusivity (NTD > 0.01) results in a reduction in oil recovery from
low salinity waterflooding for the salinity thresholds of 1,000-7,000 ppm and the thick shale case (Df < 0).
Similar to the five and two layer sand models, for high diffusivity, most of the intermediate salinities in the
model are above the upper salinity threshold so the flow of water with intermediate salinities is controlled by
the high salinity water relative permeability curves. Therefore, the overall oil recovery is lower compared
with the no diffusion case.
1.2 1.2
1 1
0.8 0.8
0.6 0.6
0.4 0.4
Df
Df
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
0.000001 0.00001 0.0001 0.001 0.01 0.1 1 0.000001 0.00001 0.0001 0.001 0.01 0.1 1
NTD NTD
8 cm shale 2 cm shale 8 cm shale 2 cm shale
Fractional decrease in the incremental oil recovery from low salinity water injection (Df ) as defined by equation 5.1 at 1 PVI as a function of NTD
from the two layer models with a shale saturated with high salinity water adjacent to the sand. The thicker shale has a larger impact because it
contains a larger volume of saline water that reduces the beneficial effects of low salinity water when there is transverse diffusion. High diffusivity
(NTD > 0.01) results in a reduction in oil recovery from low salinity waterflooding for the salinity thresholds of 1,000-7,000 ppm and the thick
shale case (Df < 0). Similar to the five and two layer sand models, for high diffusivity, most of the intermediate salinities in the model are above
the upper salinity threshold so the flow of water with intermediate salinities is controlled by the high salinity water relative permeability curves.
Therefore, the overall oil recovery is lower compared with the no diffusion case.
Effect of System Length

Df is plotted against NTD for four system lengths in Figure 5.26. It can be seen the incremental oil recovered
from low salinity waterflooding decreases as transverse diffusion increases. It would appear that transverse
diffusion becomes significant for NTD > 0.01 for the salinity thresholds of 1,000-30,000 ppm while it is at a
lower value of NTD > 0.001 for the salinity thresholds of 1,000-7,000 ppm. The adverse effect of transverse
diffusion is slightly larger in shorter systems in the 1,000-30,000 ppm threshold case but it is still significant
in the 1,000-7,000 ppm case.
Figures 5.27 and 5.28 show the oil saturation and water salinity distribution in a 30 cm model and a 300 cm
model for three different values of NTD. The producing sand is the lower layer in all cases. Transverse
139
diffusion has a greater effect at high values of NTD and in the model with a lower aspect ratio. This diffusion
results in a higher salinity water in the sand and a reduction in recovery from low salinity water injection.
1.2 1.2
1 1
0.8 0.8
0.6 0.6
0.4 0.4
Df
0.2 0.2
Df
0 0
-0.2 -0.2
-0.4 -0.4
0.000001 0.00001 0.0001 0.001 0.01 0.1 1 10 0.000001 0.00001 0.0001 0.001 0.01 0.1 1 10
NTD NTD
L=30 cm (L/H=3) L=150 cm (L/H=15) L=30 cm (L/H=3) L=150 cm (L/H=15)
L=300 cm (L/H=30) L=1500 cm (L/H=150) L=300 cm (L/H=30) L=1500 cm (L/H=150)
The fractional decrease in incremental oil recovery from low salinity waterflooding (Df ) as defined by equation 5.1 in a sand containing a water
saturated shale for different system lengths, as a function of NTD. Recovery is reduced by transverse diffusion and the effect is slightly worse in
shorter systems. The adverse effect of transverse diffusion is slightly larger in shorter systems in the 1,000-30,000 ppm thresholds while it is
significant in the 1,000-7,000 ppm.
140
Direction of flow

Salinity
NTD = 0.0007 (L = 30 cm) (1,000 ppm) NTD = 0.0007 (L = 300 cm)
30
27
24
NTD = 0.02 (L = 30 cm) 21 NTD = 0.02 (L = 300 cm)

18
15
13
NTD = 0.36 (L = 30 cm) 10

NTD = 0.36 (L = 300 cm)
7
1
Oil saturation
Figure 5.27: Oil saturation and water salinity distributions in a model with a thick shale saturated with saline water above a thin oil bearing sand for
the salinity thresholds of 1,000-30,000 ppm. At very low NTD, molecular diffusion is small so there is minimal diffusion between the high salinity
connate water from the shale and the low salinity water that has been injected into the sand. At NTD = 0.02, transverse diffusion has resulted in a
higher salinity in the sand, sufficient to reduce the oil recovery from low salinity water injection. At NTD = 0.36, high diffusivity increases the rate at
which the saline water within the shale diffuses into the low salinity water within the sands which results in reducing the oil recovery in the shorter
model. The impact is slightly less in the longer model
Direction of flow
NTD = 0.0007 (L = 30 cm) Salinity (1,000 ppm) NTD = 0.0007 (L = 300 cm)
30
27
24
21
NTD = 0.02 (L = 30 cm) NTD = 0.02 (L = 300 cm)
18
15
13
10
NTD = 0.36 (L = 30 cm) NTD = 0.36 (L = 300 cm)
7
1
Oil saturation
Figure 5.28: Oil saturation and water salinity distributions in a model with a thick shale saturated with saline water above a thin oil bearing sand for
the salinity thresholds of 1,000-7,000 ppm. A significant drop in oil recovery can be seen in both cases due to the fact that most of the reservoir ends
up with intermediate water salinities that are above the upper salinity threshold so the flow of water with intermediate salinities is controlled by the
high salinity water relative permeability curves which results in a lower oil recovery.
141
Effect of Shale Porosity
Figure 5.29 shows the impact of shale porosity (and hence volume of saline water in the shale) and system
length on incremental oil recovered from low salinity water injection as a function of NTD. At low values of
NTD, when diffusion has little effect, then incremental oil recovery is a maximum. A significant reduction in
incremental oil recovery occurs when NTD > 0.01 as seen in the previous simulations with a shale for the
salinity thresholds of 1,000-30,000 ppm while it occurs at a much lower value of NTD > 0.001 for the
salinity thresholds of 1,000-7,000 ppm. This reduction is greater when the shale porosity is higher and this
has a greater effect on recovery than system length. At the very highest levels of diffusion there is no benefit
from injecting low salinity water (i.e. the oil recovery obtained by low salinity waterflooding is similar to
that by conventional waterflooding).
1.2 1.2
1 1
0.8 0.8
0.6 0.6
Df
0.4 0.4
Df
0.2 0.2
0 0
-0.2 -0.2
-0.4 -0.4
0.000001 0.00001 0.0001 0.001 0.01 0.1 1 10 0.000001 0.00001 0.0001 0.001 0.01 0.1 1 10
NTD NTD
L=30 cm (5% shale porosity) L=30 cm (5% shale porosity)
The effect of shale porosity and system length on incremental oil recovery from low salinity waterflooding (Df) as defined by equation 5.1 in a sand
containing a water saturated shale. Transverse diffusion begins to adversely affect incremental oil recovery when NTD > 0.01 for the salinity
thresholds of 1,000-30,000 ppm while it occurs at NTD > 0.001 for the salinity thresholds of 1,000-7,000 ppm. A higher shale porosity reduces the
incremental oil recovery more as there is more saline water available to degrade the injected low salinity water.
Comparing the Effect of Shale Porosity Results to Previous Study by Brodie and Jerauld
While the study presented here was conducted on a laboratory scale model, it is believed that the results can
be extended and used as a reference in any scale even at reservoir scale as they are presented in terms of a
dimensionless number (NTD). A previous simulation study by Brodie and Jerauld (2014) conducted on a
reservoir scale model is compared with the above results. In their model, reservoir A is a sandstone which
has inter-bedded shales. In one of their sensitivity studies, only physical longitudinal dispersivity was varied.
They neglected modelling transverse diffusion so their simulations were dominated by transverse numerical
diffusion instead. They found that diffusion has the greatest effect on low salinity waterflooding when low
salinity water is injected in a sand adjacent to a thick shale. It is anticipated that their results and findings
cannot be generalized as neither did they consider the effect of transverse diffusion nor did they correlate
their results using NTD.
142
B-3D Reservoir Scale Models with Complex Heterogeneity (from SPE 10 Model 2)
Figure 5.30 compares the oil recovery obtained from three scenarios of injecting low salinity water. The
results confirm that the effect of dispersion is more pronounced on oil recovery during the production of the
connate water bank especially for the salinity thresholds of 1,000-7,000 ppm. During this period, most of the
intermediate water salinities are above the upper salinity threshold so the flow of water with intermediate
salinities is controlled by the high salinity water relative permeability. Therefore, the overall oil recovery is
lower compared to the no dispersion case. However when comparing the overall oil recoveries by the
different scenarios, the difference is small so the flow is always dominated by heterogeneity variations and
the effect of dispersion tends to be minor.
40 40
35 35
30 30
Oil recovery (%)
Oil recovery (%)

25 25
20 20
15 15
10 10
5 5
0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
PVI PVI
no dispersion NTD = 0.47 NTD = 0.18 no dispersion NTD = 0.47 NTD =0.18
Oil recovery for the three scenarios in the 3D reservoir model. The effect of dispersion on the oil recovery is much smaller compared to the shaly
sand 2D layer model at laboratory scale. In field scale, heterogeneity variations dominates the flow for the displacement of oil by low salinity
waterflooding where dispersion has minor effect. During the production of connate water bank, the effect of dispersion on oil recovery is noticeable
(~ 2 % difference) for the salinity thresholds of 1,000-7,000 ppm compared to the other salinity thresholds. During this period, most of the
intermediate salinities are above the upper salinity threshold so the flow of water with intermediate salinities is controlled by the high salinity water
relative permeability. Therefore, the overall oil recovery is lower compared with the no dispersion case.
Figures 5.31 and 5.32 show the oil saturation and water salinity distributions across layers 1 and 10 for two
scenarios at two times (0.3 PVI and 1 PVI) for both salinity thresholds. These two scenarios are: the no
dispersion case and the case using a transverse dispersion is 30 times less than the longitudinal dispersion as
suggested by Mahadevan et al. (2003) (NTD = 0.18).
Using the salinity thresholds of 1,000-30,000 ppm, at early times (0.3 PVI), more areas were contacted by
intermediate salinities due to dispersion compared to the no dispersion case. These intermediate salinities fall
between the salinity thresholds, so the flow of water with intermediate salinities is controlled by intermediate
water salinity relative permeability curves. The overall oil recovery is slightly higher than the no dispersion
case. However for the salinity thresholds of 1,000-7,000 ppm, at the same time (0.3 PVI), most of
intermediate salinities are higher than the upper salinity threshold so the flow of water with intermediate
salinities is controlled by the high salinity water relative permeability curves. Therefore, the oil recovery by
the dispersion case at 0.3 PVI is less than the no dispersion case.
143
At 1 PVI, for the salinity thresholds of 1,000-30,000 ppm, more area was contacted by intermediate water
salinities due to dispersion leading to higher oil recovery. However, the incremental oil recovery by
dispersion is not that high (only 1%) due to the fact that the permeability variations in this realistic model are
very large (i.e. some areas are 0.001 md compared to 20,000 md). For the salinity thresholds of 1,000-7,000
ppm, the oil recovery by the no dispersion case is slightly higher due to the fact that for the dispersion case,
the intermediate water salinities are higher than the upper salinity threshold so the flow of water with
intermediate salinities is controlled by the high salinity water relative permeability curves.
144
Permeability distribution
Layer 1 Layer 10
PERM (md)
Oil saturation distribution Oil saturation distribution

Direction of flow
At 0.3 PV At 0.3 PV
Layer 1 Layer 10
Layer 1 Layer 10
Oil saturation
At 1 PV At 1 PV
Layer 1 Layer 10 Layer 1 Layer 10
Water salinity distribution Water salinity distribution

At 0.3 PV
Salinity At 0.3 PV
Layer 1 Layer 10 (1,000 ppm) Layer 1 Layer 10
30
27
24
21
18
At 1 PV At 1 PV
Layer 1 Layer 10 15 Layer 1 Layer 10
13
10
Figure 5.31: Oil saturation and water salinity distributions for immediate injection of low salinity water into a 3D reservoir taken from SPE 10 model
2 using salinity thresholds of 1,000-30,000 ppm. Two cases are shown: no dispersion and NTD = 0.18 cases at two times. At early times (0.3 PVI),
more areas were contacted by intermediate water salinities due to dispersion compared to the no dispersion case. These intermediate salinities fall
between the salinity thresholds so the flow of water with intermediate salinities is controlled by intermediate water salinities relative permeability
curves. The overall oil recovery are slightly higher than the no dispersion case At 1 PVI, more areas were contacted by intermediate water salinities
due to dispersion leading to higher oil recovery. However, the incremental oil recovery by dispersion is not that high (only 1%) due to the fact that the
permeability variations in this realistic model are very large (i.e. some areas are 0.001 md compared to 20,000 md).
145
Layer 1 Layer 10
PERM (md)
Oil saturation distribution Oil saturation distribution

No dispersion Direction of flow NTD = 0.18
At 0.3 PV At 0.3 PV
Oil Saturation
At 1 PV At 1 PV
Layer 1 Layer 10
Layer 1 Layer 10

At 0.3 PV Salinity At 0.3 PV
Layer 1 Layer 10 (1,000 ppm) Layer 1 Layer 10
30
27
24
21
At 1 PV 18 At 1 PV
13
10
Figure 5.32: Oil saturation and water salinity distributions for immediate injection of low salinity water into a 3D reservoir taken from SPE 10 model
2 using salinity thresholds of 1,000-7,000 ppm. Two cases are shown: no dispersion and NTD = 0.18 at two times. At early times (0.3 PVI), more area
was contacted by intermediate water salinities due to dispersion compared to the no dispersion case. Most of these intermediate salinities are higher
than the upper salinity threshold so the flow of water with intermediate salinities is controlled by high salinity water relative permeability curves. The
overall oil recovery are lower than the no dispersion case at 0.3 PVI. At 1 PVI, more areas were contacted by intermediate water salinities due to
dispersion. However, the oil recovery by the no dispersion case is slightly higher due to the fact that for, the dispersion case, most of intermediate
water salinities are higher than the upper salinity threshold so the flow of water with intermediate salinities is controlled by the high salinity water
relative permeability curves.
146
C-A Reservoir with a Transition Zone Connected to an Aquifer
Figure 5.33 shows the oil recovery for the three scenarios in addition to conventional high salinity
waterflooding. The effect of dispersion on oil recovery is most significant during the production of the
connate water bank. During this period, most of the intermediate salinities are above the upper salinity
threshold so the flow of water with intermediate salinities is controlled by the high salinity water relative
permeability curves. Following this time, dispersion has less effect, so the curves start to converge.
40
35
30
25
Oil recovery (%)
20
15 Salinity thresholds
1,000-7,000 ppm
10
5
0
0 1 2 3 4 5 6
PVI (oil zone)
no dispersion NTD = 0.2 NTD=0.07 conventional waterflooding
Figure 5.33: Oil recovery for several scenarios of injecting low salinity water into an oil zone connected to an aquifer. The maximum effect of
dispersion on oil recovery is during the production of the connate water bank since most of the intermediate salinities are above the upper salinity
threshold so the flow of water with intermediate salinities is controlled by the high salinity water relative permeability curves. Following this time,
dispersion has less effect, so the curves start to converge.
Figures 5.34 shows the oil saturation and water salinity distributions at two times for two scenarios. At 2 PV
of the oil zone injected, intermediate salinities were formed due to dispersion (NTD = 0.07) especially in the
region that is closer to the aquifer (green in the salinity figure) while in the region away from the aquifer, oil
was contacted by low salinity water. Most of the intermediate salinities (green in the salinity figure for NTD
= 0.07) are above the upper salinity threshold so the flow of water is controlled by the high salinity water
relative permeability curves. Therefore, the microscopic displacement in the regions close to the aquifer has
been reduced to that of a high salinity waterflood. However, in the regions away from the aquifer (red in the
salinity figure for NTD = 0.07), the microscopic displacement has not been reduced. As a result, the oil
recovery is lower compared to the no dispersion case while it is still higher than that obtained by
conventional waterflooding (as shown in Figure 5.33). This suggests that more low salinity water injection is
needed to compensate for this decreases due to dispersion. A previous study by Law et al. (2014) found that
the presence of an aquifer increased the oil recovery by low salinity water as it supported the pressure.
However, dispersion was not modelled in their study although the injectors were close to the oil water
contact. The effect of the presence of aquifers on low salinity waterflooding was also analysed by
Bedrikovetsky et al. (2014). It was found that since low salinity water was injected into an aquifer zone, low
incremental oil recoveries by low salinity waterflooding were observed.
147
Oil saturation distribution Direction of flow Oil saturation distribution
Oil saturation
At 2 PV of oil zone At 2 PV of oil zone

Salinity
(1,000 ppm)
30
27
24
21
18
15
13
10
7 Upper threshold
1 Lower threshold
Figure 5.34: Oil saturation and water salinity distributions for immediate injection of low salinity water into a reservoir connected to an aquifer. Two
cases are shown: no dispersion and NTD = 0.07 at two times. At 2 PV of oil zone injected, intermediate salinities were formed due to dispersion
(NTD = 0.07) especially in the region that is closer to the aquifer (green in the salinity figure) while in the region away from the aquifer, oil was
contacted by low salinity water. Most of the intermediate salinities (green in the salinity distribution figure for NTD = 0.07) are above the upper
salinity threshold so the flow of water is controlled by the high salinity water relative permeability curves. Therefore, the microscopic displacement in
the regions close to the aquifer has been reduced to that of a high salinity waterflood. However, in the regions away from the aquifer (red in salinity
figure for NTD = 0.07), the microscopic displacement has not been reduced. As a result, the oil recovery is lower compared to the no dispersion case
while it is still higher than that obtained by conventional waterflooding. This suggests that more low salinity water injection is needed to compensate
for this decreases due to dispersion.
148
Range of NTD Values in the Field
Evidence from laboratory experiments suggest that NTD in laboratory scale systems (Hassinger and Von
Rosenberg 1968, Blackwell 1962, Perkins and Johnston 1963) is within the range of 5.6 × 10-3 < NTD < 0.04.
Extending the system to a reservoir scale increases the longitudinal dispersion to about 2-5% of the system
length (Mahadevan et al. 2003, Jerauld et al. 2008). Consequently, the transverse dispersion increases and so
does the NTD to values around 0.2 (see discussion earlier in this chapter). At this value (NTD ~ 0.2), the
flow is at the turning point from reservoir heterogeneity dominated flow to dispersion dominated flow.
However, the system is very unlikely to experience NTD values greater than 5 which suggests that reservoir
heterogeneity affects the performance of low salinity waterflooding more than dispersion.
NTD depends on the system length, thickness, the pore velocity in the high permeability layer and the
transverse dispersivity (equation 2.11). Table 5.8 shows the values used to calculate NTD for both laboratory
scale and reservoir scale.
Table 5.8: Values used to calculate NTD at laboratory scale and reservoir scale in this chapter.
Laboratory scale model Laboratory units converted to field units Reservoir scale model
3
𝐿𝐿 (cm) 29.80 0.97 ft 𝐿𝐿 (ft) 1.34 ×10
𝑣𝑣1 (cm/min) 0.01 0.47 ft/d 𝑣𝑣1 (ft/d) 15.89

-4 -4
𝐾𝐾𝑡𝑡 (cm2/min) 6.20 ×10 9.61 ×10 ft2/d 𝐾𝐾𝑡𝑡 (ft2/d) 0.87
𝐻𝐻 (cm) 3.17 0.10 ft 𝐻𝐻 (ft) 20
𝑁𝑁𝑁𝑁𝑁𝑁 0.18 𝑁𝑁𝑁𝑁𝑁𝑁 0.18
Table 5.8 shows that while 𝐾𝐾𝑡𝑡 increases significantly at reservoir scale, the system length and thickness also
increases. As a result, NTD is almost the same. This shows that the use of a dimensionless number (NTD) is
a robust measure of dispersion at any scale.
Oil recoveries for the NTD = 1 scenario are compared with the no diffusion scenario for the five different
permeability realizations in Figure 5.35 using Df for both salinity thresholds.
Using the salinity thresholds of 1,000-30,000 ppm, diffusion increases the oil recoveries in the HL and LH
cases when low salinity water is injected post the breakthrough of high salinity water (i.e. at 0.05 PV). The
uniform salinity front promotes the injected low salinity water to contact the low permeability layers. The oil
in these low permeability layers are recovered as they were contacted by the low salinity water compared to
the no diffusion case. This increase in oil recovery is less in the LHLHL case (at 0.05 PV Df is 1.8 compared
to 3.3 in the LH and HL cases). In the HLHLH and homogeneous cases, diffusion has minor effects on the
oil recovery when injecting low salinity water post the breakthrough of high salinity water (0.05-0.17 PV).
The effect of diffusion on oil recovery is different depending on the volume of the low permeability layers in
the model. Diffusion increases the oil recovery in the HL, LH, and LHLHL cases as their low permeability
volumes are larger compared to the HLHLH and homogenous cases (i.e. where the volume of low
149
permeability is zero since it is has a uniform permeability of 100 md). Generally, the increase in oil recovery
due to diffusion decreases as the start of injecting low salinity water is delayed from the breakthrough of
high salinity water to 0.6 PV. This is due to the fact that low salinity water mixes with larger volumes of
injected high salinity water thus the microscopic displacement is reduced for most cases.
Using the salinity thresholds of 1,000-7,000 ppm, when low salinity water is injected post the breakthrough
of high salinity water, diffusion has similar effect on increasing the oil recoveries for most cases to the
salinity thresholds of 1,000-30,000 ppm. However, diffusion has a significant effect on decreasing the oil
recoveries for all cases as the start of injecting low salinity water is delayed to 0.6 PV of high salinity water
injected. This is due to the fact that most of the intermediate salinities are above the upper salinity threshold
so the flow of water with intermediate salinities is controlled by the high salinity water relative permeability
curves. As a result, the microscopic displacement is reduced to that of high salinity waterflood. This has
resulted in a significant decrease in oil recoveries for all cases (Df ~ 0.3).
Water salinity and oil saturation distributions for the LHLHL case at 1 PV for both salinity thresholds are
shown in Figures 5.36 and 5.37. Using the salinity thresholds of 1,000-30,000 ppm (Figure 5.36), when low
salinity water is injected post the breakthrough of high salinity water, the uniform salinity front at NTD =1
has contacted all layers to form intermediate salinities. The flow of these intermediate salinities is controlled
by the intermediate salinity water relative permeability curves that are linearly interpolated between the high
and low salinity water relative permeability curves. At any water saturation, the relative permeability to oil at
intermediate salinities is higher than the relative permeability to oil at high salinity. High diffusivity has
improved the sweep efficiency but reduced the microscopic displacement compared to the no diffusion case,
however the overall oil recovery due to diffusion is higher (Df ~ 1.8). When the injection of low salinity
water is delayed to 0.6 PV of high salinity water injected, the injected low salinity water mixes with larger
volume of high salinity water thus the microscopic displacement is reduced.
Using the salinity thresholds of 1,000-7,000 ppm (Figure 5.37), when low salinity water is injected post the
breakthrough of high salinity water, diffusion has improved the sweep efficiency since the uniform salinity
front has contacted all layers. However, when the start of low salinity water injection is delayed to 0.6 PV of
high salinity water injected, diffusion has a significant impact on reducing the oil recoveries. This due to the
fact that low salinity water mixes with larger volume of high salinity water. In addition, Most of the
intermediate salinities are above the upper salinity thresholds so the flow of water is controlled by the high
salinity water relative permeability curves. Therefore, the microscopic displacement is reduced to that of
high salinity waterflood. This has resulted in an overall lower oil recovery compared to the no diffusion case
(Df ~ 0.3).
150
3.5 3.5
3 3
2.5 2.5
2 2
Df
Df
1.5 1.5
1 1
0.5 0.5
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Start of Injecting low salinity water (PV) Start of Injecting low salinity water (PV)
HL LH HLHLH LHLHL Homogeneous HL LH HLHLH LHLHL Homogeneous
Figure 5.35: a) salinity thresholds 1,000-30,000 ppm b) salinity thresholds 1,000-7,000 ppm
The oil recoveries (Df ) as defined by equation 5.1 at NTD = 1 are compared with the case of no diffusion for two different salinity thresholds. The
effect of diffusion on the oil recoveries by low salinity waterflooding depends on the choice of the salinity thresholds. This is related to the fact that
most of the intermediate salinities for the 1,000-30,000 ppm fall between the salinity thresholds so the flow of water with intermediate salinities is
controlled by intermediate water salinity that are linearly interpolated between the high and low salinity water relative permeability curves. In these
cases, the effect of diffusion is less compared with the other salinity thresholds (1,000-7,000 ppm) where most of the intermediate water salinities are
above the upper salinity threshold so the flow of water with intermediate salinities is controlled by the high salinity water relative permeability
curves. For the salinity threshold of 1,000-7,000 ppm, the oil recoveries due to diffusion are much lower compared to the other salinity thresholds
when the start of low salinity water injection is delayed to 0.6 PV of high salinity water injected. Generally, for both salinity thresholds, diffusion
increases the oil recoveries when low salinity water is injected post the breakthrough of high salinity water (0.05-0.17 PV). Diffusion promotes the
injected low salinity water to contact low permeability layers which had resulted in a better sweep efficiency. However, the increase in oil recovery is
not the same for all cases. In the HL, LH and LHLHL cases, the overall volume of the low permeability layers is larger compared to their volume in
the homogenous (where it is zero) and HLHLH cases.
151
Direction of flow
Salinity at 1 PV Salinity at 1 PV
No diffusion Salinity
NTD = 1
(1,000 ppm)
Lowsal post BT of conv Lowsal post BT of conv
30
27
24
21
18
Lowsal post 0.6 PV of conv Lowsal post 0.6 PV of conv
15
13
10
Oil saturation at 1 PV 4 Oil saturation at 1 PV

No diffusion 1
NTD = 1
Lowsal post BT of conv Oil saturation Lowsal post BT of conv
Figure 5.36: Water salinity and oil saturation distributions at 1 PV of injecting low salinity water post conventional waterflooding into the LHLHL
case using the salinity thresholds of 1,000-30,000 ppm. When low salinity water is injected post the breakthrough of high salinity water, the uniform
salinity front at NTD = 1 has contacted all layers to form intermediate salinities. The flow of these intermediate salinities is controlled by the
intermediate salinity water relative permeability curves that are linearly interpolated between the high and low salinity water relative permeability
curves. At any water saturation, the relative permeability to oil at intermediate salinities is higher than the relative permeability to oil at high salinity.
High diffusivity has improved the sweep efficiency but reduced the microscopic displacement compared to the no diffusion case, however the overall
oil recovery due to diffusion is higher (Df ~ 1.8). When the injection of low salinity water is delayed to 0.6 PV of high salinity water injected, the
injected low salinity water mixes with larger volume of high salinity water thus the microscopic displacement is reduced.
152
Direction of flow
No diffusion NTD = 1
30
27
24
21
18
15
13
10
Oil saturation at 1 PV 4 Oil saturation at 1 PV

No diffusion 1
NTD = 1
Lowsal post BT of conv Oil saturation Lowsal post BT of conv
Figure 5.37: Water salinity and oil saturation distributions at 1 PV of injecting low salinity water post conventional waterflooding into the LHLHL
case using the salinity thresholds of 1,000-7,000 ppm. When low salinity water is injected post the breakthrough of high salinity water, diffusion has
improved the sweep efficiency since the uniform salinity front has contacted all layers. However, when the start of low salinity water injection is
delayed to 0.6 PV of high salinity water injection, diffusion has a significant impact on reducing the oil recoveries. This due to the fact that low
salinity water mixes with larger volume of high salinity water injected. In addition, most of the intermediate salinities are above the upper salinity
thresholds so the flow of water is controlled by the high salinity water relative permeability curves. Therefore, the microscopic displacement is
reduced to that of high salinity waterflooding. This has resulted in an overall lower oil recovery compared to the no diffusion case (Df ~ 0.3).
153
B- 3D Reservoir Scale Model with Complex Heterogeneity (From SPE 10 Model 2):
Figure 5.38 shows the oil recovery for two different salinity thresholds. The effect of dispersion on oil
recovery depends on the choice of the salinity thresholds and the volume of the high salinity water that was
injected before low salinity water. Using the salinity thresholds of 1,000-30,000 ppm, when low salinity
water is injected after the breakthrough of high salinity water, dispersion increases the oil recovery since it
promotes intermediate salinities to contact low permeability areas thus the sweep efficiency is improved
although the microscopic displacement is reduced. Most of these intermediate salinities are between the
salinity thresholds so the flow of water with intermediate salinities is controlled by the intermediate salinity
water relative permeability curves that are linearly interpolated between the high salinity and low salinity
water relative permeability curves. As a result, the overall oil recovery for the salinity thresholds of 1,000-
30,000 ppm is higher with the inclusion of dispersion than that with no dispersion. When the start of low
salinity water injection is delayed to about 0.6 PV of high salinity water injected, the incremental oil
recovery due to dispersion is lower (0.5%) compared to 2% for the case when low salinity water is injected
post the breakthrough of high salinity water (0.1 PV). This is due to the fact that low salinity water mixes
with larger volumes of the injected high salinity water thus the microscopic displacement is reduced.
Using the salinity thresholds of 1,000-7,000 ppm, dispersion has decreased the oil recovery. This is due to
the fact that most of the intermediate salinities are above the upper salinity threshold so the flow of water is
controlled by the high salinity water relative permeability curves. As a result, the microscopic displacement
is reduced to that of a high salinity waterflood.
154
35 35
34 34
33 33
32 32
31 31
Oil recovery
Oil recovery
30 30
29 29
28 28
27 27
26 26
25 25
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
Start of injecting low salinity water (PV) Start of injecting low salinity water (PV)
no dispersion NTD =0.47 no dispersion NTD =0.47
Figure 5.38: a) salinity thresholds 1,000-30,000 ppm b) salinity thresholds 1,000-7,000 ppm
Oil recovery at 1 PV of injecting low salinity water post high salinity waterflooding into a 3D reservoir with complex heterogeneity taken from SPE
10 model 2. For the salinity thresholds of 1,000-30,000 ppm, when low salinity water is injected post the breakthrough of high salinity water,
dispersion has improved the sweep efficiency but reduced the microscopic displacement which resulted in a higher oil recovery compared to the no
dispersion case. The incremental oil recovery due to dispersion decreases as the start of low salinity water injection is delayed. This is due to the fact
that low salinity water mixes with larger volume of high salinity water thus the microscopic displacement is reduced and hence the incremental oil
recovery is lower. This is not the case for the salinity thresholds of 1,000-7,000 ppm since most of the intermediate salinities are above the upper
salinity thresholds so the flow of water with intermediate salinities is controlled by the high salinity water relative permeability curves. As a result,
the microscopic displacement is reduced to that of a high salinity waterflood and accordingly the oil recovery is reduced compared to the no
dispersion case.
Figure 5.39 shows the oil saturation and water salinity distributions using the salinity thresholds of 1,000-
30,000 ppm. When low salinity water is injected post the breakthrough of high salinity water, most oil
volumes trapped in low permeability areas are displaced due to the dispersion causing additional oil recovery
in the case when NTD = 0.47 compared to the case with no dispersion. When the start of low salinity water
injection is delayed to 0.6 PV of high salinity water injected, the injected low salinity water mixes with
larger volume of high salinity water thus the microscopic displacement is reduced. Therefore, the
incremental oil recovery due to dispersion is lower compared to the case when low salinity water is injected
post the breakthrough of high salinity water.
However, for the salinity thresholds of 1,000-7,000 ppm (Figure 5.40), this is only partially true. Dispersion
causes the low salinity water to contact the high salinity water in the low permeability layer thus displaces
oil towards the producer but with much lower effective permeability since most of the salinities in the
reservoir are above the upper salinity threshold so the flow of water with intermediate salinities is controlled
by the high salinity water relative permeability curves.
155
Layer 1 Layer 10
PERM (md)
Oil saturation at 1 PVI Direction of flow Oil saturation at 1 PVI

No diffusion NTD = 0.47
Layer 1 Oil saturation
Layer 10 Layer 1 Layer 10

Salinity
No diffusion (1,000 ppm) NTD = 0.47
30
27
24
21
18
Lowsal post 0.6 PV of conv 15 Lowsal post 0.6 PV of conv

Layer 1 Layer 10 13
Layer 1 Layer 10
10
Figure 5.39: Water salinity and oil saturation distributions at 1 PV of injecting low salinity water post conventional waterflooding into a 3D reservoir
taken from SPE 10 model 2 using the salinity thresholds of 1,000-30,000 ppm. Oil volumes trapped in the low permeability areas are displaced due to
dispersion causing additional oil recovery in the case of NTD = 0.47 compared to the no dispersion case. When the start of low salinity water
injection is delayed to 0.6 PV of high salinity water injected, the injected low salinity water mixes with larger volume of high salinity water thus the
microscopic displacement is reduced. Therefore, the incremental oil recovery due to dispersion is lower compared to the case when low salinity water
is injected post the breakthrough of high salinity water.
156
Layer 1 Layer 10
PERM (md)
No diffusion NTD = 0.47
Layer 1 Layer 10 Oil saturation Layer 1 Layer 10

Salinity at 1 PV Salinity Salinity at 1 PV

No diffusion (1,000 ppm) NTD = 0.47
30
27
24
21
18
Lowsal post 0.6 PV of conv 15 Lowsal post 0.6 PV of conv
Layer 1 Layer 10 13
Layer 1 Layer 10
10
4
1
Figure 5.40: Water salinity and oil saturation distributions at 1 PV of injecting low salinity water post conventional waterflooding into a 3D reservoir
taken from SPE 10 model 2 using the salinity thresholds of 1,000-7,000 ppm. Oil volumes trapped in the low permeability areas are being displaced
due to dispersion causing additional oil recovery in the case of NTD = 0.47 compared to the no dispersion case. However, due to the fact that most of
the intermediate salinities are above the upper salinity threshold so the flow of water with intermediate salinities is controlled by the high salinity
water relative permeability curves. As a result, the microscopic displacement has been reduced to that of a high salinity waterflood and accordingly
the overall oil recovery is lower compared with the no dispersion case.
157
Figure 5.41a and 5.41b shows the oil recovery for three different cases and the drop in oil recovery due to
dispersion Df , respectively. The effect of dispersion on oil recovery is significant as the start of low salinity
water injection is delayed (i.e. when low salinity water is injected post 0.6 PV of high salinity water injected,
the drop in oil recoveries due to dispersion (Df ) is around 0.4 compared to (Df ~ 0.9) when low salinity water
is injected post the breakthrough of high salinity water (0.2 PV)). This is due to the fact that the injected low
salinity water mixes with larger volume of high salinity water in addition to the aquifer water which results
in much lower oil recovery.
40 1
Salinity Thresholds Salinity Thresholds
1,000-7,000 ppm 1,000-7,000 ppm
38 0.9
Oil recovery at 5 PV of oil zone
0.8
36
Df at 5 PV of oil zone
0.7
34
0.6
32
0.5
30
0.4
28 0.3
0 1 2 3 4 0 1 2 3 4
Start of injecting low salintiy water (PV of Start of injecting low salintiy water (PV of
oil zone) oil zone)
no dispersion NTD = 0.07 NTD = 0.2 NTD = 0.07 NTD = 0.2
Figure 5.41: a) Oil recovery at 5 PV of oil zone b) (Df) as defined by equation 5.1 at 5 PV of oil zone
Oil recovery and the drop in oil recovery due to dispersion (Df ) at 5 PV of oil zone for injecting low salinity water post conventional waterflooding
into a reservoir connected to an aquifer. The effect of dispersion on oil recovery is significant as the start of low salinity water injection is delayed
(i.e. when low salinity water is injected post 0.6 PV of high salinity water injected, the drop in oil recovery due to dispersion (Df ) is 0.4 compared to
(Df ~ 0.9) when low salinity water is injected post the breakthrough of high salinity water (0.2 PV)). This is due to the fact that the injected low
salinity water mixes with larger volume of high salinity water in addition to the aquifer water which results in a much lower oil recovery.

This chapter has investigated the effect of physical diffusion/dispersion and reservoir heterogeneity on the
performance of low salinity waterflooding when it is injected continuously. This included injection of low
salinity water immediately and injection of low salinity water post high salinity waterflooding.
First, the chapter started by characterizing mixing using the transverse dispersion number (NTD). The flow is
dominated by reservoir heterogeneity when NTD < 0.2, while it is dominated by diffusion when NTD > 1.
Then, the impacts of physical diffusion/dispersion and reservoir heterogeneity on low salinity waterflooding
were investigated using the CMG STARS simulator. In most cases, two salinity thresholds were considered
as a sensitivity parameter when modelling low salinity waterflooding: the first one assumed that the
thresholds are equal to the injected low salinity water and the connate water while the other one assumed that
the thresholds are close to the injected low salinity water.
158
Immediate Injection of Low Salinity Water
The investigations included: laboratory scale with simple heterogeneity variations, reservoir scale with more
complex heterogeneity and finally a reservoir with the inclusion of an aquifer.
On the laboratory scale, it was found that the effects of diffusion on low salinity waterflooding depends on
the salinity thresholds. If the salinity thresholds are closer to the salinity of the injected water, the effect is
more pronounced: during the production of the connate water bank for a two layer model with a thick high k
layer, during the production of the connate water bank for a five layer model with a double decrease in
layer’s permeability and during the production of the connate water bank for a short system with a thick
shale. The findings in the shaly sand model is similar to the findings by Brodie and Jerauld (2014). However,
in their models, physical transverse diffusion was set to zero which resulted in an unrealistic NTD = 0. Their
simulations were dominated by transverse numerical diffusion instead. Their oil recoveries were not
correlated to NTD. It is believed that the findings in this chapter are more robust.
On the reservoir scale, using a 3D model with more complex heterogeneity, it was found the dispersion gives
the maximum effect on oil recovery during the production of the connate water bank. However, dispersion
had minor effects on the oil recovery by low salinity waterflooding compared to reservoir heterogeneity at
this scale. This is due to the fact that the permeability variations in this realistic model at reservoir scale are
very large (i.e. some areas are 0.001 md compared to 20,000 md) and the displacement of oil by low salinity
waterflooding is mainly through the high permeability areas while very small portions of the low
permeability areas are contacted by high dispersivities.
When including an aquifer in the model, a similar effect of dispersion was found during the production of
the connate water bank. During this time, most of the intermediate salinities are above the upper salinity
permeability curves. As a result, the oil recovery at this time is lower compared with the no dispersion case.
After the production of the connate water bank, dispersion has less effect on low salinity waterflooding so
the overall oil recovery for both cases are about the same.
The findings in this chapter can be extended to deploying low salinity waterflooding in field projects. It is
believed that typical NTD values in the field ranges from 10-4 to about 0.2. Even in the extreme cases (when
the transverse dispersion is 30 times less than a longitudinal dispersion that is 5 % of the reservoir length),
NTD would approach 0.4. In most cases, it is anticipated that the flow would be dominated by reservoir
heterogeneity and dispersion would have a minor effect. However, if the reservoir has a large volume of
shale (with high porosity) that is adjacent to a sandstone, then the effect of dispersion would be more
pronounced based on the findings of this chapter and the resulting oil recovery would be sharply decreased.
The presence of an aquifer was found to have an effect in decreasing the oil recovery by low salinity water
especially during the production of the connate water bank and with a transverse dispersion that is 30 times
less than the longitudinal dispersion as suggested by Mahadvean et al. (2003). Previous study by Law et al.
(2014) found that the presence of an aquifers increased the oil recovery by low salinity water as it supported
the pressure. However, dispersion was not modelled in their study although the injectors were close to the oil
159
water contact. In addition, the effect of the presence of aquifers on low salinity waterflooding was analysed
by Bedrikovetsky et al. (2014). It was found that since low salinity water was injected into an aquifer zone,
low incremental oil recoveries by low salinity waterflooding were observed.
Overall, the effect of dispersion is to reduce oil recovery from continuous low salinity waterflooding at
intermediate times, although the ultimate recovery will be the same as without dispersion.
Injection of Low Salinity Water Post High Salinity Waterflooding
On the laboratory scale, it was found that diffusion has different effects on low salinity waterflooding
depending on the choice of the salinity thresholds, the start of injecting low salinity water post high salinity
waterflooding and the overall volume of the low permeability layers. Diffusion improves the oil recovery
when low salinity water is injected post the breakthrough of high salinity water as it promotes the low
salinity water to contact the low permeability layers. The incremental oil recovery due to diffusion is much
higher in the cases with larger overall volume of low permeability layers. Diffusion was found to
significantly decrease the oil recovery when the start of low salinity water injection is delayed and especially
if the salinity thresholds are closer to the injected water salinity.
On the reservoir scale, dispersion increases the oil recovery as it promotes the low salinity water to contact
the low permeability layers thus improving the sweep efficiency. The incremental oil recovery due to
dispersion is lower as the start of low salinity water injection post high salinity waterflooding is delayed.
This is due to the fact that low salinity water mixes with large amount of high salinity water thus the
microscopic displacement is reduced. However, this is not the case when the upper salinity threshold is
closer to the injected water salinity. This is because most of the intermediate salinities are above the upper
salinity threshold so the flow of water is controlled by high salinity water relative permeability curves. As a
result, the microscopic displacement is reduced to that of a high salinity waterflood and hence the oil
recovery is lower compared to the no dispersion case.
For the case of a reservoir connected to an aquifer, it was found that dispersion has a significant effect on the
oil recovery if the injection of low salinity water post high salinity water is delayed from the breakthrough of
high salinity water. This is due to the fact that the injected low salinity water mixes with larger volume of
high salinity water in addition to the aquifer water which results in much lower oil recovery.
160
CHAPTER 6: IMPACT OF HETEROGENEITY AND PHYSICAL
DIFFUSION AND DISPERSION ON THE INJECTION OF LOW SALINITY
WATER AS A SLUG
In chapter 5, the impact of heterogeneity and physical diffusion/dispersion was investigated on continuous
low salinity waterflooding. In this chapter, the investigation is performed when low salinity water is injected
as a slug. It is often economically infeasible to inject low salinity water continuously due to lack of
availability of low salinity water, the need to dispose of saline produced water by reinjection or due to
limitations in desalination capacity. The purpose of this chapter is to find the best slug size of low salinity
water that will recover oil almost as much as when it is injected continuously. The investigation is conducted
using simulation models at laboratory scale and reservoir scale. For the reservoir scale models, two different
sets were modelled: a ten layer model with complex taken from SPE 10 model 2 (Christie and Blunt 2001)
and a reservoir with a transition zone connected to an aquifer to investigate the effect of an aquifer on low
salinity waterflooding. The impact of physical diffusion/dispersion and reservoir heterogeneity is evaluated
first on low salinity water when it is injected immediately and then when it is injected post high salinity
water. Since the salinity thresholds pose an impact on the predictions and outcomes of the simulation
models, they were considered to be a sensitivity parameter in most of the simulation models conducted on
this chapter.
6.1 Method
6.1.1 Immediate Injection of Low Salinity Water
Table 6.1 summarizes the cases examined for immediate injection of low salinity water as a slug.
Table 6.1: Cases examined to evaluate the impact of physical diffusion/dispersion and heterogeneity on injecting low salinity water as a slug.
Slug size of low salinity
Scale Model Number of grid blocks Physical dimensions Permeabilities
water (PV)
Homogeneous 100 md
The top 18z (10 md) The
LH
bottom 2z (100 md)
The top 2z (100 md) The
HL
bottom 18z (10 md) 0.1, 0.3, 0.4, 0.5, 0.6 and
29.8 × 3.17 × 3.17
Laboratory 200 × 1 × 20 Pattern of the top 5z continuous low salinity
cm
LHLHL (10 md), water injection (1 PV)
next 2z (100 md)
Pattern of the top 2z
HLHLH (100 md),
next 5z (10 md)
complex heterogeneity
Ten layer model that represents a 0.1, 0.3, 0.6 and
taken from SPE 60 × 60 × 10 1,200 × 600 × 20 ft geologically realistic continuous low salinity
10 model 2 field (0.001 to 20,000 water injection (1 PV)
md)
Oil zone:
50 × 1 × 7
Transition zone: Oil zone:
Reservoir
A Reservoir 50 × 1 ×10 1,000 × 50 × 35 ft
0.4, 0.7, 1, 2, 3 and
with a transition Transition zone:
Aquifer: continuous low salinity
zone and 1,000 × 50 × 20 ft 100 md
Top 35 ft: 50 × 1 × 7 water injection (5 PV of oil
connected to an Aquifer:
Next 40 ft: 50 × 1 × 4 zone)
aquifer 1,000 × 50 × 355 ft
Next 40 ft: 50 × 1 × 2
Next 80 ft: 50 × 1 × 2
Bottom 160 ft: 50 × 1 × 2
161
Five different 2D layered models with simple heterogeneity variations were used to investigate the effect of
diffusion and reservoir heterogeneity on the injection of low salinity water as a slug. The model dimensions
and their permeability variations are listed in Table 6.1. Figure 6.1 shows the simulation models with their
permeability variations.
Homogeneous LH
HL LHLHL HLHLH
Permeability (md)
1 100
Figure 6.1: Permeability variations for the 2D layered models used to evaluate the impact of diffusion and reservoir heterogeneity on the injection
of low salinity water as a slug.
One injection well and one production well were used in the models. Both wells were completed vertically in
all layers. The injection well was located in the last column along the right hand face of the model and set to
have a constant injection rate of 0.04 cm3/min while the production well was located in the first column
along the left hand face of the model and set to have a constant production rate of 0.04 cm3/min. For all
models, low salinity water was injected immediately and then followed by high salinity water. The slug sizes
of low salinity water were varied to represent a wide range: 0.1, 0.3, 0.4, 0.5, 0.6 PV in addition to
continuous low salinity water injection (1 PV). For each permeability realisation, two scenarios were tested:
one scenario without diffusion and the other scenario with the modelling of molecular diffusion that results
in NTD = 1. Two different salinity thresholds were used in modelling low salinity waterflooding: 1,000-
30,000 ppm and 1,000-7,000 ppm. The salinity thresholds of 1,000-7,000 ppm were chosen as they represent
the more realistic thresholds as confirmed by previous experiments (Tang and Morrow 1999). However, the
salinity thersholds of 1,000-30,000 ppm are not realistic. They were chosen to see the maximum effect of the
linear interpolation of relative permeabilities (high and low salinity water) on modelling low salinity
waterflooding since all of the intermediate salinities fall within the linear interpolation region.
B-3D Reservoir Scale Model with Complex Heterogeneity (from SPE 10 Model 2)
The 3D model described previously in chapter 5 was used here but instead low salinity water was injected as
a slug followed by high salinity water. The slug sizes of low salinity water were varied to represent a wide
range: 0.1, 0.3, 0.6 PV in addition to continuous low salinity water injection (1 PV).
With regards to modelling physical dispersion, three scenarios were run: one without the inclusion of
162
times less than a the longitudinal dispersivity that is 5 % of the reservoir length (NTD = 0.47).
Two different salinity thresholds were used in modelling low salinity waterflooding: 1,000-30,000 ppm and
1,000-7,000 ppm. The salinity thresholds of 1,000-7,000 ppm were chosen as they represent the more
realistic thresholds as confirmed by previous experiments (Tang and Morrow 1999). However, the salinity
thersholds of 1,000-30,000 ppm are not realistic. They were chosen to see the maximum effect of the linear
interpolation of relative permeabilities (high and low salinity water) on modelling low salinity waterflooding
since all of the intermediate salinities fall within the linear interpolation region.
The 2D model described previously in chapter 5 was used here but instead low salinity water was injected as
a slug followed by high salinity water. The slug sizes of low salinity water were varied to represent a wide
range: 0.4, 0.7, 1, 2, 3 PV of oil zone in addition to continuous low salinity water injection (5 PV of oil
zone). With regards to modelling physical dispersion, three scenarios were run: one without the inclusion of
times less than a longitudinal dispersivity that is 5 % of the reservoir length (NTD = 0.2). The salinity
thresholds were set to be 1,000-7,000 ppm. These thresholds were chosen as they represent the more realistic
thresholds as confirmed by previous experiments (Tang and Morrow 1999).
For all cases, the salinity thresholds were set be 1,000-7,000 ppm. These thresholds were chosen as they
represent the more realistic thresholds as confirmed by previous experiments (Tang and Morrow 1999). As
observed in chapter 5, the effect of diffusion/dispersion is more significant on low salinity waterflooding
(LSW) and accordingly on oil recovery using this salinity thresholds.
It was observed in chapter 5 that the effect of diffusion on layered models with simple variations depends on
the volume of low permeability in the model in addition to the surface area of the low-high permeability
contact. In this section, the LHLHL and HLHLH scenarios were selected to be the only two cases for
investigating the slug sizes of injecting low salinity water post high salinity waterflooding. Table 6.2
summarizes the different slug sizes used in this investigation. The start of low salinity water injection was
varied to be: after the breakthrough of conventional water and after the producing water cut has reached 90%
from conventional waterflooding. For all cases, the total water injected was 2 PV.
163
Table 6.2: Slug sizes of injected low salinity water post high salinity waterflooding that were used in the LHLHL and HLHLH.
Start of low salinity water injection
Slug size of low Post the breakthrough of Post 90% water cut from
Pore volume of water injected
salinity water (PVI) conventional water conventional waterflooding
LHLHL HLHLH LHLHL HLHLH
Injected conventional water before LSW (PVI)) 0.05 0.05 0.65 0.65
Slug sizes of injected low salinity water post
0.10 0.10 0.10 0.10 0.10
Volume of chase high salinity water 1.85 1.85 1.25 1.25
0.30 0.30 0.30 0.30 0.30
0.40 0.40 0.40 0.40 0.40
0.50 0.50 0.50 0.50 0.50
0.60 0.60 0.60 0.60 0.60
1.00 1.00 1.00 1.00 1.00
B-3D Reservoir Scale Model with Complex Heterogeneity (from SPE 10 Model 2)
With regards to modelling physical dispersion, two scenarios were run: one without the inclusion of
dispersion and one with a transverse dispersivity that is 30 times less than a longitudinal dispersivity that is 5
% of the reservoir length (NTD = 0.47). Table 6.3 summarizes the different slug sizes used in this
investigation. The start of low salinity water injection was varied to be: after the breakthrough of
conventional water and after the producing water cut has reached 90% from conventional waterflooding. For
all cases, the total water injected was 2 PV.
Table 6.3: Slug sizes of injected low salinity water post high salinity waterflooding that were used in a 3D reservoir taken from SPE 10 model 2.
No dispersion NTD = 0.47 No dispersion NTD = 0.47
0.1 0.1 0.1 0.1 0.1
0.3 0.3 0.3 0.3 0.3
0.6 0.6 0.6 0.6 0.6
1.0 1.0 1.0 1.0 1.0
1.3 1.3 1.3 1.3 1.3
164
With regards to modelling physical dispersion, two scenarios were run: one without the inclusion of
dispersion and one with a transverse dispersivity that is 30 times less than a longitudinal dispersivity that is 5
% of the reservoir length (NTD = 0.2). Table 6.4 summarizes the different slug sizes used in this
investigation. The start of low salinity water injection was varied to be: after the breakthrough of
conventional water and after the producing water cut has reached 90% from conventional waterflooding. For
all cases, the total water injected was 10 PV.
Table 6.4: Slug sizes of injected low salinity water post high salinity waterflooding that were used in a reservoir connected to an aquifer.
No dispersion NTD = 0.2 No dispersion NTD = 0.2
0.40 0.40 0.40 0.40 0.40
0.70 0.70 0.70 0.70 0.70
1.00 1.00 1.00 1.00 1.00
2.00 2.00 2.00 2.00 2.00
3.00 3.00 3.00 3.00 3.00
4.00 4.00 4.00 4.00 4.00
5.00 5.00 5.00 5.00 5.00
6.2 Results
6.2.1 Immediate Injection of Low Salinity Water
The results are compared at 1 PVI. The oil recoveries are shown in Figures 6.2 and 6.3 as a fraction of the oil
recovery (FRC) obtained from continuous low salinity waterflooding for both salinity thresholds.
It was found in chapter 4 that the best slug size of low salinity water for layered case is around 0.6-0.8 PV
without modelling diffusion compared to 0.4 PV as suggested previously by Jerauld et al. (2008). When
modelling diffusion (Figures 6.2 and 6.3), it can be seen that the best slug size for both salinity thresholds is
around 0.4-0.6 PV. At 0.4 PV, FRC increases from 0.9-0.92 for HLHLH and LHLHL cases to 0.9 to 0.95 for
other cases. This means that for heterogeneous cases, a slug has to be increased to recover additional oil.
With regards to the choice of the salinity thresholds in modelling low salinity waterflooding, when the
salinity thresholds are 1,000-7,000 ppm, diffusion (NTD = 1) increases the salinity of the injected low
165
salinity water and thus reduces the efficiency of low salinity waterflooding by forming intermediate salinities
especially at very small slug sizes (0.1-0.4 PV). Most of these intermediate salinities are above the upper
salinity threshold so the flow of water with intermediate salinities is controlled by the high salinity water
relative permeability curves. As a result, the overall oil recoveries are lower compared with the no diffusion
case.
HL LH
1 1
0.95 0.95
0.9 0.9
0.85 0.85
FRC
FRC
0.8 0.8
0.75 0.75
0.7 0.7
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
LSW slug size (PV) LSW slug size (PV)
no diffusion NTD=1 no diffusion NTD=1
HLHLH LHLHL
1 1
0.95 0.95
0.9 0.9
FRC
0.85 0.85
FRC
0.8 0.8
0.75 0.75
0.7 0.7
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Homogeneous
1
0.95
0.9
0.85
FRC
0.8
0.75
0.7
0 0.2 0.4 0.6 0.8 1
LSW slug size (PV)
no diffusion NTD=1
Figure 6.2: Oil recoveries from various slug sizes of low salinity water as a fraction of oil recovery from continuous low salinity waterflooding at 1
PV (FRC) are plotted as function of diffusion for five different permeability realisations for salinity thresholds of 1,000-30,000 ppm. The main
observations are: 1- generally, the best slug size of low salinity water for both cases is around 0.4-0.6 PV, 2- with regards to diffusion, at very small
slug sizes of low salinity water (0.1 PV), the gap between the two curves are larger for the LH and HL cases.
166
HL LH
1 1
0.95 0.95
0.9 0.9
0.85
FRC
0.85
FRC
0.8 0.8
0.75 0.75
0.7 0.7
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
HLHLH LHLHL
1 1
0.95 0.95
0.9 0.9
0.85 0.85
FRC
FRC
0.8 0.8
0.75 0.75
0.7 0.7
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Homogeneous
1
0.95
0.9
0.85
FRC
0.8
0.75
0.7
0 0.2 0.4 0.6 0.8 1
LSW slug size (PV)
no diffusion NTD=1
Figure 6.3: Oil recoveries from various slug sizes of low salinity water as a fraction of oil recovery from continuous low salinity waterflooding at 1
PV (FRC) are plotted as a function of diffusion for five different permeability realisations for the salinity thresholds of 1,000-7,000 ppm. Diffusion
(NTD = 1) increases the salinity of the injected low salinity water and thus reduces the efficiency of low salinity waterflooding by forming
intermediate salinities especially at very small slug sizes (0.1-0.4 PV). Most of these intermediate salinities are above the upper salinity threshold so
the flow of water is controlled by the high salinity water relative permeability curves. As a result, the overall oil recoveries are lower compared to
the no diffusion case.
167
Oil recoveries for the NTD = 1 scenario are compared with the no diffusion scenario for the five different
permeability realizations in Figure 6.4 using Df for both salinity thresholds.
4 4
3.5 3.5
3 3
2.5 2.5
2 2
Df
Df
1.5 1.5
1 1
0.5 0.5
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
HL HL LSW slug size (PV)
LSW slug size (PV) LH
LH
HLHLH HLHLH
LHLHL LHLHL
Homogeneous Homogeneous
Figure 6.4: a) salinity thresholds:1,000-30,000 ppm b) salinity thresholds: 1,000-7,000 ppm

The oil recoveries ( Df ) as defined by equation 5.1 at NTD = 1 are compared with the case of no diffusion for two different salinity thresholds. The
effect of diffusion on the oil recoveries by low salinity water depends on the choice of the salinity thresholds. This is related to the fact that most of
the intermediate salinities for the 1,000-30,000 ppm fall between the salinity thresholds so the flow of water with intermediate salinities is controlled
by intermediate water salinities that are linearly interpolated between the high and low salinity water relative permeability curves. In these cases, the
effect of diffusion is less compared with the other salinity thresholds (1,000-7,000 ppm) where most of the intermediate water salinities are above
the upper salinity threshold so the flow of water with intermediate salinities is controlled by the high salinity water relative permeability curves. For
the salinity threshold of 1,000-7,000 ppm, the oil recoveries are much lower for small slug sizes of low salinity water and then significantly increase
as the slug size increases especially for the LH, HL and LHLHL cases. In the HL, LH and LHLHL cases, the overall volume of the low permeability
layers is larger compared to their volume in the homogeneous (where it is 0) and HLHLH cases so diffusion promotes the injected low salinity water
to sweep these layers compared with the no diffusion case.
Using the salinity thresholds of 1,000-30,000 ppm, diffusion causes the oil recoveries to increase in the HL
and LH cases when increasing the slug size of low salinity water from 0.1 PV (Df ~ 2) to 0.6 PV (Df ~ 3.3).
In the NTD = 1 scenario, the uniform salinity front promotes the injected low salinity water to contact the
low permeability layers. The oil in these low permeability layers is recovered as they were contacted by the
low salinity water compared to the no diffusion case where they were contacted by only high salinity water.
As a result, the overall oil recovery due to diffusion is increased as the slug size of low salinity water is
increased. This increase in oil recovery is less in the LHLHL (at 0.6 PV, Df is 1.8 compared to 3.3 in the LH
and HL cases). Diffusion has the least effect on the homogenous case as there is no permeability variations.
Using the salinity thresholds of 1,000-7,000 ppm, the effect of diffusion on the oil recovery for all five cases
is more pronounced. For a very small slug size (0.1 PV), diffusion decreases the oil recovery compared to
the no diffusion case since most of the intermediate salinities are above the upper salinity threshold so the
flow of water with intermediate salinities is controlled by high salinity water relative permeability curves.
For the homogeneous and LH cases, at a very small slug size (0.1 PV), diffusion almost completely destroys
the benefit of low salinity waterflooding (Df ~ 0). However, the effect of diffusion at this small slug size is
not the same for the HLHLH and LHLHL cases since they contain different volumes of low permeability
and there is a different surface area of low-high permeability contact. In addition, the effect of viscous cross
flow is increased. For all five cases, diffusion increases the oil recovery when the slug size of low salinity
168
water is increased. For the HL, LH and LHLHL cases, diffusion significantly increases the oil recoveries for
a slug size of 0.6 PV (Df ~ 3) compared to the homogeneous and HLHLH cases (Df ~ 1). This is due to the
fact that in the HL, LH and LHLHL cases, the volume of the low permeability layers is larger compared to
their volume in the homogeneous case (i.e. where the volume of low permeability is zero since it is has a
uniform permeability of 100 md) and in the HLHLH cases. In the LH, HL and LHLHL cases, diffusion
results in the injected low salinity water being able to sweep the low permeability layers which did not occur
in the no diffusion case (i.e. Df values are increased significantly). The injected low salinity water was not
destroyed compared with the case when the slug size was small.
Water salinity and oil saturation distributions for the LHLHL case at 1 PV for both salinity thresholds are
shown in Figures 6.5 and 6.6. There is a uniform salinity front when NTD = 1. Using the salinity thresholds
of 1,000-30,000 ppm (Figure 6.5), for the 0.1 PV slug size, the uniform salinity front has contacted all layers
(but not horizontally all the way to the producer due to its small size) to form intermediate salinities. The
flow of these intermediate salinities is controlled by the intermediate salinity water relative permeability
curves that are linearly interpolated between the high and low salinity water relative permeability curves. At
any water saturation, the relative permeability to oil at intermediate salinities is higher than the relative
permeability to oil at high salinity. High diffusivity has improved the sweep efficiency but reduced the
microscopic displacement compared to the no diffusion case, however the overall oil recovery due to
diffusion is higher (Df ~ 1.5).
Using the salinity thresholds of 1,000-7,000 ppm, for the 0.1 PV slug size, the uniform salinity front has
contacted all layers (but not horizontally all the way to the producer due to its small size) to form
intermediate salinities. However, since most of these intermediate salinities that formed due to diffusion are
above the upper salinity threshold, flow is controlled by the high salinity water relative permeability curve.
Again, high diffusivity has improved the sweep efficiency and reduced the microscopic displacement
compared to the no diffusion case but the microscopic displacement has been reduced to that of high salinity
flooding so the overall oil recovery due to diffusion is lower (Df ~ 0.5). Diffusion has reduced the effect of
low salinity waterflooding at this very small slug size. When the slug size is increased to 0.6 PV for both
salinity thresholds, diffusion has improved the sweep efficiency compared to the no diffusion case thus their
overall oil recoveries are increased compared with the no diffusion case. The salinity fronts are uniform due
to diffusion while the displacement fronts are not uniform in oil saturation since diffusion only occurs in a
single phase displacement (water) while the water-oil displacement is a two phase immiscible displacement
that is controlled by their relative permeabilities and their mobility ratio.
169
Direction of flow
Salinity
0.1 PV slug (1,000 ppm) 0.1 PV slug
30
27
24
21
18
0.6 PV slug 0.6 PV slug
15
13
10
1
Oil saturation at 1 PV Oil saturation at 1 PV
Oil saturation
Figure 6.5: Water salinity and oil saturation distributions at 1 PV of injecting different slug sizes of low salinity water into the LHLHL case using
salinity thresholds of 1,000-30,000 ppm. For the 0.1 PV slug size, the uniform salinity front has contacted all layers (but not horizontally all the way
to the producer due to its small size) to form intermediate salinities. The flow of these intermediate salinities is controlled by the intermediate salinity
water relative permeability curves that are linearly interpolated between the high and low salinity water relative permeability curves. At any water
saturation, the relative permeability to oil at intermediate salinities is higher than the relative permeability to oil at high salinity. High diffusivity has
improved the sweep efficiency and reduced the microscopic displacement compared to the no diffusion case but the overall oil recovery due to
diffusion is higher (Df ~ 1.5). When the slug size is increased to 0.6 PV, diffusion had improved the sweep efficiency and the microscopic
displacement to that of a low salinity flood thus the overall oil recovery is increased compared to the no diffusion case. It can be seen that the salinity
front is uniform due to diffusion while the displacement front is not uniform in oil saturation since diffusion only occurs in a single phase
displacement (water) while the water-oil displacement is a two phase immiscible displacement that is controlled by their relative permeabilities and
their mobility ratio.
170
Salinity at 1 PV Direction of flow Salinity at 1 PV
Salinity
0.1 PV slug (1,000 ppm) 0.1 PV slug
30
27
24
21
0.6 PV slug 18 0.6 PV slug
15
13
10
4
Oil saturation at 1 PV Oil saturation at 1 PV
1
Oil saturation
0.6 PV slug
0.6 PV slug
Figure 6.6: Water salinity and oil saturation distributions at 1 PV of injecting different slug sizes of low salinity water into the LHLHL case using
salinity thresholds of 1,000-7,000 ppm. For the 0.1 PV slug size, the uniform salinity front has contacted all layers (but not horizontally all the way to
the producer due to its small size) to form intermediate salinities. However, since most of these intermediate salinities are above the upper salinity
threshold so the flow of water with intermediate salinities is controlled by the high salinity water relative permeability curves. High diffusivity has
improved the sweep efficiency but reduced the microscopic displacement to that of high salinity water so the overall oil recovery due to diffusion is
lower (Df ~ 0.5) Diffusion has reduced the effect of low salinity waterflooding at this very small slug size. When the slug size is increased to 0.6 PV,
diffusion has improved the sweep efficiency thus the overall oil recovery is increased compared to the no diffusion case. The salinity fronts are
uniform due to diffusion while the displacement fronts are not uniform in oil saturation since diffusion only occurs in a single phase displacement
(water) while the water-oil displacement is a two phase immiscible displacement that is controlled by their relative permeabilities and their mobility
ratio.
171
B- 3D Reservoir Scale Model with Complex Heterogeneity (From SPE 10 Model 2):
Figure 6.7 shows the oil recovery for two different salinity thresholds. There are three key differences
between the two plots.
First, using a large slug size of low salinity water (0.5 PV and larger), dispersion increases the oil recovery
using the salinity thresholds of 1,000-30,000 ppm. Dispersion promotes intermediate salinities water to
contact low permeability layers thus the sweep efficiency is improved although the microscopic
displacement is reduced. Most of these intermediate salinities are between the salinity thresholds so the flow
of water with intermediate salinities is controlled by the intermediate salinity water relative permeability
curves that are linearly interpolated between the high salinity and low salinity water relative permeability
curves. As a result, the overall oil recovery for the salinity thresholds of 1,000-30,000 ppm is higher with the
inclusion of dispersion than that with no dispersion. However, when using the 1,000-7,000 ppm, most of the
intermediate salinities that were formed due to dispersion are higher than the upper salinity threshold so the
flow of water with intermediate salinities is controlled by the high salinity water relative permeability
curves. Therefore, the overall oil recovery for the 1,000-7,000 ppm is lower with the inclusion of dispersion
than that with no dispersion.
Second, the best slug size of low salinity water using the salinity thresholds of 1,000-30,000 ppm is around
0.6 PV compared with 1 PV using the salinity thresholds of 1,000-7,000 ppm.
Third, for the salinity thresholds of 1,000-30,000 ppm, the oil recovery when including transverse dispersion
is slightly higher than that with no dispersion. However, for the salinity thresholds of 1,000-7,000 ppm, for
the case with a transverse dispersion that is 30 times less than the longitudinal dispersion as suggested by
Mahadevan et al. (2003) (NTD = 0.18), the oil recovery in the dispersion case is always less than the case
with no dispersion for any slug size of low salinity water. At this level of dispersion, dispersion increases the
salinity in the low salinity water slug which results in intermediate salinities that are above the upper salinity
permeability curves. Therefore, the overall oil recovery is lower compared with the no dispersion case.
However, transverse dispersion is not sufficiently high to dominate the flow thus not all the low permeability
areas are contacted by the low salinity water. When using a transverse dispersivity that is 30 times less than
a longitudinal dispersivity that is 5 % of the reservoir length (NTD = 0.47) with the salinity thresholds of
1,000-7,000 ppm, dispersion actually improves the oil recovery for large slug sizes due to the fact that at this
large dispersion, dispersion dominates the flow over the reservoir heterogeneity. The salinity front is uniform
so it promotes the injected low salinity water to contact the low permeability areas in the reservoir thus the
overall sweep efficiency is improved and accordingly the oil recovery is improved.
172
35 35
33 33
31 31
Oil recovery (%)
Oil recovery (%)

29 29
27 27
25 25
23 23
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
no dispersion NTD = 0.18 NTD=0.47 no dispersion NTD = 0.18 NTD = 0.47
Figure 6.7: a) Salinity thresholds:1,000-30,000 ppm b) salinity thresholds: 1,000-7,000 ppm

Oil recovery for three scenarios. For the salinity thresholds of 1,000-30,000 ppm, transverse dispersion that is 30 times less than the longitudinal
dispersion as suggested by Mahadevan et al. (2003) (NTD = 0.18) has improved the sweep efficiency and reduced the microscopic displacement
which has resulted in higher oil recoveries compared with no dispersion especially for large slug sizes. This is not observed in the 1,000-7,000 ppm
since most of the intermediate salinities are above the upper salinity thresholds so the flow of water with intermediate salinities is controlled by the
high salinity water relative permeability curve which has resulted in lower oil recovery compared with the no dispersion case. For both salinity
thresholds, when the transverse dispersivity is 30 times less than a longitudinal dispersivity that is 5 % of the reservoir length (NTD = 0.47) with the
salinity thresholds of 1,000-7,000 ppm, dispersion actually improves the oil recovery. The salinity front is uniform so it promotes the injected low
salinity water to contact low permeability areas in the reservoir thus the overall sweep efficiency is improved and accordingly the oil recovery is
improved.
Figure 6.8 shows the oil saturation and water salinity distributions using the salinity thresholds of 1,000-
30,000 ppm for two slug sizes (0.1 and 0.6 PV) at 1 PVI for two scenarios: no dispersion and with a
transverse dispersion that is 30 times less than the longitudinal dispersion as suggested by Mahadevan et al.
(2003) (NTD = 0.18). Generally, most oil volumes trapped in low permeability areas are displaced due to the
large dispersion causing additional oil recovery in the case when NTD = 0.18 compared to the case with no
dispersion. However, for the salinity thresholds of 1,000-7,000 ppm (Figure 6.9), this is only partially true.
Dispersion causes the low salinity water to contact the high salinity water in the low permeability layer thus
displaces oil towards the producer but with much lower effective permeability. This is due to the fact that
most of the salinities in the reservoir are above the upper salinity threshold so the flow of water with
intermediate salinities is controlled by the high salinity water relative permeability curves thus the
microscopic displacement is reduced to that of a high salinity waterflood.
173
Layer 1 Layer 10
PERM (md)


Oil saturation

Layer 10
Salinity at 1 PVI Salinity at 1 PVI

No dispersion 30 NTD = 0.18
0.1 PV slug 27
0.1 PV slug
21
18
15
0.6 PV slug 13
0.6 PV slug
10
Figure 6.8: Oil saturation and water salinity distributions at 1 PVI for different slug sizes of injecting low salinity water into a 3D reservoir taken
from SPE 10 model 2 (Christie and Blunt 2001) using salinity thresholds of 1,000-30,000 ppm. Two cases are shown: no dispersion and with
transverse dispersion that is 30 times less than the longitudinal dispersion as suggested by Mahadevan et al. (2003) (NTD = 0.18). Oil volumes
trapped in the low permeability areas are displaced due to the large dispersion causing additional oil recovery in the case of NTD = 0.18 compared to
the no dispersion case.
174
Layer 1 Layer 10
PERM (md)
Direction of flow
Oil saturation at 1 PVI Oil saturation at 1 PVI

Oil saturation
Layer 1

Salinity
(1,000 ppm)
Salinity at 1 PVI 30 Salinity at 1 PVI
No dispersion 27
NTD = 0.18
0.1 PV slug 24 0.1 PV slug
21
18
15
13 Layer 10
10
Figure 6.9: Oil saturation and water salinity distributions at 1 PVI for different slug sizes of injecting low salinity water into a 3D reservoir taken
from SPE 10 model 2 (Christie and Blunt 2001) using salinity thresholds of 1,000-7,000 ppm. Two cases are shown: no dispersion and with
transverse dispersion that is 30 times less than the longitudinal dispersion as suggested by Mahadevan et al. (2003) (NTD = 0.18). Oil volumes
trapped in the low permeability areas are being displaced due to the large dispersion causing additional oil recovery in the case of NTD = 0.18
compared to the no dispersion case. However, due to the fact that most of the intermediate salinities are above the upper salinity threshold so the flow
of water is controlled by the high salinity water relative permeability curves. Therefore, the overall oil recovery is lower compared with the no
dispersion case.
175
The results are compared at 5 PVI. The oil recoveries are shown in Figure 6.10 as a fraction of the oil
recovery (FRC) obtained from the continuous injection of low salinity water. The best slug size of injecting
low salinity water is estimated when (FRC) is 0.95. From this figure, the best slug size of injecting low
salinity water into a reservoir connected to an aquifer depends on dispersion. For the no dispersion case, the
best slug size is around 2 PV of oil zone increasing to 3 PV of oil zone for the case of a transverse dispersion
that is 30 times less than a longitudinal dispersion that is 5 % of reservoir length (NTD = 0.2). This is due to
the fact that smaller (but still large) slug sizes of low salinity water (1-2 PV of oil zone) dispersion destroys
the benefit of low salinity water.
100
95
Salinity Thresholds:
90 1,000-7,000 ppm
FRC
85
80
75
70
0 1 2 3 4 5 6
LSW slug size (PV of oil zone)
no dispersion NTD = 0.2 NTD = 0.07

Figure 6.10: Oil recoveries from various slug sizes of injecting low salinity water into a reservoir connected to an aquifer as a fraction of oil
recovery from continuous low salinity waterflooding (FRC) at 5 PV of the oil zone. The oil recoveries are plotted for three different cases: no
dispersion, transverse dispersion that is 30 times less than a longitudinal dispersion that is 5 % of reservoir length (NTD = 0.2) and with transverse
dispersion that is 30 times less than the longitudinal dispersion as suggested by Mahadevan et al. (2003) (NTD = 0.07). The best slug size is
estimated when the fractional oil recovery by the slug of low salinity water (FRC) is 95% of the oil recovery by continuous LSW at 5 PV oil zone.
This plot suggests that the best slug size of injecting low salinity water into a reservoir connected to an aquifer depends on the transverse
dispersion. It is around 2 PV of oil zone for the no dispersion case compared to 3 PV for the case of NTD = 0.2. This slug size was chosen since
injecting larger slug sizes recover additional 5% of oil. This is due to the fact that smaller (but still) slug sizes of low salinity water (1-2 PV of oil
zone) dispersion destroys the benefit of low salinity water.
176
For all models, the best slug size of low salinity water injected post conventional waterflooding is estimated
when the oil recovery (FRC) is around 95% of the oil recovery obtained by continuous low salinity
waterflooding. This is represented by the following equation:
𝑜𝑜𝑜𝑜𝑜𝑜 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑏𝑏𝑏𝑏 𝑎𝑎 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑜𝑜𝑜𝑜 𝑙𝑙𝑙𝑙𝑙𝑙 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 ℎ𝑖𝑖𝑖𝑖ℎ 𝑠𝑠𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
𝐹𝐹𝑅𝑅𝑅𝑅 = … … … . . (6.1)
𝑜𝑜𝑜𝑜𝑜𝑜 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑏𝑏𝑏𝑏 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑙𝑙𝑙𝑙𝑙𝑙 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
For each case, the dominator of equation 6.1 depends on the diffusion/dispersion case. For the laboratory
scale models, if the case is “no diffusion” then the dominator of this equation will be oil recovery by
continuous low salinity waterflooding using “no diffusion”. Similarly is the case for NTD =1.
For the reservoir scale models, if the case is “no dispersion” then the dominator of this equation will be the
oil recovery by continuous low salinity waterflooding using “no dispersion”. Similarly is the case for NTD =
0.47 (as in the case for SPE 10 model 2) or NTD = 0.2 (as in the case for a reservoir connected to an aquifer).
The results are compared at 2 PVI. The oil recoveries for the different slug sizes of low salinity water are
shown as a fraction of the oil recovery (FRC) obtained from continuous low salinity waterflooding at 2 PVI in
Figure 6.11 for both the LHLHL and HLHLH cases. FRC is more sensitive to the slug size of low salinity
water for the case with high diffusivity (NTD = 1) compared to the no diffusion case. High diffusivity
promotes the injected low salinity water to contact the low permeability layers. As a result, the sweep
efficiency is improved but the microscopic displacement is reduced. At a very small slug size (0.1 PV), FRC
is around 0.75 for the case with high diffusivity (NTD = 1) compared to 0.9 for the case with no diffusion. In
addition, FRC depends on the start of low salinity water injection for the case with high diffusivity for slug
sizes of (0.3-0.6 PV) while FRC is almost identical for the no diffusion case when the start of low salinity
water injection is varied. The best slug size of low salinity water for the case with no diffusion is around 0.5
PV. For high diffusivity (NTD = 1), the best slug sizes of low salinity water increases to 0.6 PV when low
salinity water is injected at the breakthrough of high salinity water (BT of conv in Figure 6.11). If the
injection of low salinity water is delayed until the producing water cut from high salinity water has reached
90% water cut (90 WC in Figure 6.11), the best slug size of low salinity water increases to 1 PV. These are
similar to the best slug size when low salinity water is injected immediately (section 6.2.1 A).
177
1 1
0.95 0.95
0.9 0.9
0.85 0.85
FRC
0.8 0.8
FRC
0.75 0.75
0.7 0.7
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
LHLHL no diffusion post BT of conv HLHLH no diffusion post BT of conv
LHLHL NTD=1 post BT of conv HLHLH NTD=1 post BT of conv
LHLHL no diffusion post 90 WC of conv HLHLH no diffusion post 90 WC of conv
LHLHL NTD=1 post 90 WC of conv HLHLH NTD=1 post 90 WC of conv
Figure 6.11: a) LHLHL model b) HLHLH model

Oil recoveries from different slug sizes of low salinity water injected post high salinity waterflooding as a fraction of oil recovery from continuous
low salinity waterflooding (FRC) at 2 PVI for two different permeability variations: LHLHL (left figure) and the HLHLH (right figure). The start of
low salinity water injection was varied: post the breakthrough of high salinity water (BT of conv) and when the producing water cut from high
salinity water has reached 90% (90 WC of conv). The oil recoveries are more sensitive to the slug size for high diffusivity. Diffusion promotes the
injected low salinity water to contact the low permeability layers thus the sweep efficiency is improved but the microscopic displacement is reduced.
The best slug size of low salinity water for the case with no diffusion is around 0.5 PV increasing to 0.6 PV for the case of (BT of conv) with high
diffusivity. It increases to 1 PV for the case of (90 WC of conv) with high diffusivity.
The oil recovery for the NTD = 1 scenario is compared with the no diffusion scenario for the LHLHL and
HLHLH cases in Figure 6.12 using Df. The effect of diffusion on the oil recovery depends on the slug size of
low salinity water, the start of low salinity water injection and the volume of low permeability in the model.
For both heterogeneity realisations, at a very small slug size (0.1 PV), diffusion decreases the oil recovery
since it reduces the microscopic displacement. Most of the intermediate salinities are above the upper
salinity threshold so the flow of water with intermediate salinities is controlled by the high salinity water
relative permeability curves. If low salinity water is injected post the breakthrough of high salinity water, the
effect of diffusion on decreasing the oil recovery (Df ~ 0.4) is less compared to the case when it is injected
after the producing water cut from high salinity water has reached 90% (Df ~ 0.1). In the latter case, more
high salinity water was injected before low salinity water which promotes more mixing between the injected
low salinity water and high salinity water. This has reduced the microscopic displacement and thus the
overall oil recovery is lower compared to the other case. At large slug sizes (0.6-1 PV), diffusion increases
the oil recovery as it promotes low salinity water to contact the low permeability layers thus the sweep
efficiency is improved. The increase in oil recovery due to diffusion is higher for the LHLHL model
compared to the HLHLH model because it contains larger volume of low permeability.
178
4.5 4.5
4 4
3.5 3.5
3 3
2.5 2.5
2 2
Df
Df
1.5 1.5
1 1
0.5 0.5
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
LHLHL NTD=1 post BT of conv HLHLH NTD=1 post BT of conv
LHLHL NTD=1 post 90 WC of conv HLHLH NTD=1 post 90 WC of conv
Figure 6.12: a) LHLHL model b) HLHLH model

The oil recoveries for the case of NTD = 1 ( Df ) as defined by equation 5.1 are compared to the no diffusion for injecting different slug sizes of low
salinity water post high salinity waterflooding for two different permeability variations: LHLHL model (left figure) and HLHLH model (right figure).
The start of low salinity water injection was varied: post the breakthrough of high salinity water (BT of conv) and when the producing water cut from
high salinity water has reached 90% (90 WC of conv). At a very small slug size (0.1 PV), diffusion decreases the oil recovery since it reduces the
microscopic displacement to that of a high salinity flood. For the case of (BT of conv), the effect of diffusion on decreasing the oil recovery (Df ~ 0.4)
is less compared to the case of (90 WC of conv) where it is (Df ~ 0.1). In the latter case, more high salinity water was injected before low salinity
water which promotes more mixing between the injected low salinity water and high salinity water. This has reduced the microscopic displacement
and thus the overall oil recovery is lower compared to the other case. At a large slug size (0.6-1) PV, diffusion increases the oil recovery as it
promotes low salinity water to contact the low permeability layers thus improving the sweep efficiency. The increase in oil recovery due to diffusion
is higher for the LHLHL model compared to the HLHLH model because it contains larger volume of low permeability.
Water salinity and oil saturation distributions for the LHLHL case at 2 PV are shown in Figure 6.13. For the
0.1 PV slug size, diffusion has increased the salinity of the injected low salinity water to intermediate
salinities. Most of these intermediate salinities are above the upper salinity threshold so the flow of water is
controlled by high salinity waterflooding. This has reduced the microscopic displacement to that of a high
salinity flood. As a result, high permeability layers were not contacted by low salinity water and thus the oil
recovery is lower compared to the no diffusion case. However, this is not the case for the 1 PV slug size.
Diffusion promotes the injected low salinity water to contact the low permeability layers. This has improved
the sweep efficiency but reduced the microscopic displacement. As a result, the overall oil recovery is higher
compared to the no diffusion case.
179
Salinity at 2 PVI Direction of flow Salinity at 2 PVI
0.1 PV slug of lowsal Salinity

0.1 PV slug of lowsal
(1,000 ppm)
Post BT of conv Post 90 WC of conv Post BT of conv Post 90 WC of conv
30
27
24
21
1 PV slug of lowsal 18 1 PV slug of lowsal
Post BT of conv Post 90 WC of conv

15 Post BT of conv Post 90 WC of conv
13
10
Oil saturation at 2 PVI 4

Oil saturation at 2 PVI
No diffusion 1
NTD = 1
0.1 PV slug of lowsal Oil saturation 0.1 PV slug of lowsal
1 PV slug of lowsal 1 PV slug of lowsal
Figure 6.13: Water salinity and oil saturation distributions at 2 PVI for injecting different slug sizes of low salinity water post conventional
waterflooding into the LHLHL case. For the 0.1 PV slug size, diffusion (NTD =1) has increased the salinity of the injected low salinity water to
intermediate salinities. Most of these intermediate salinities are above the upper salinity threshold so the flow of water is controlled by high salinity
waterflooding. This has reduced the microscopic displacement to that of a high salinity flood. As a result, high permeability layers were not contacted
by low salinity water and thus the oil recovery is lower compared to the no diffusion case. However, this is not the case for the 1 PV slug size.
Diffusion (NTD =1) promotes the injected low salinity water to contact the low permeability layers. This has improved the sweep efficiency but
reduced the microscopic displacement. As a result, the overall oil recovery is higher compared to the no diffusion case.
180
B- 3D Reservoir Scale Model with Complex Heterogeneity (from SPE 10 Model 2)
The results are compared at 2 PVI. The oil recoveries for different slug sizes of injecting low salinity water
post high salinity waterflooding are shown in Figure 6.14 and as a fraction of the oil recovery (FRC) obtained
from continuous low salinity waterflooding at 2 PVI in Figure 6.15. Generally, the effect of dispersion on
low salinity waterflooding is minor compared to reservoir heterogeneity. This is due to the fact that for this
realistic model the permeability variations are very large (some areas are 0.001 md compared to 20,000 md).
Oil recoveries for the NTD = 0.47 scenario are compared with the no dispersion scenario in Figure 6.16
using Df. For the no dispersion case, FRC is almost identical when the start of low salinity water injection is
varied. However, this is not the case for the case with high dispersivity (NTD = 0.47). High dispersivity
promotes more mixing of low salinity water with the injected high salinity water. This has reduced the
microscopic displacement since most of the intermediate salinities are above the upper salinity threshold so
the flow of water is controlled by high salinity waterflooding. As a result, the oil recovery is lower compared
to the no dispersion case for small slug sizes (0.1-0.3 PV). For the case of NTD = 0.47 when injecting low
salinity water post the breakthrough of high salinity water, Df decreases when the slug size is increased from
0.1 to 0.3 PV. The percentage of increase in oil recovery for the no dispersion case is higher than that for
NTD = 0.47 (Figure 6.15) so Df for 0.3 PV slug would be lower than for the smaller slug size (0.1 PV).
When the slug size is large (1-1.3 PV), dispersion has a minor effect on low salinity waterflooding, so the
overall oil recovery is similar to the no dispersion case (Df ~ 1). Generally, injecting low salinity water when
the produced water cut from conventional waterflooding has reached 90% water cut recovers more oil
compared with injecting low salinity water post the breakthrough of high salinity water. For the latter case,
the volume of the injected chase high salinity water is larger which results in increasing the salinity of the
injected salinity water and thus reducing the microscopic displacement.
The best slug size is around 1.3 PV for both cases compared with the best slug size of injecting low salinity
water immediately (0.6-1 PV). This would be expected as low salinity water mixes with larger amount of
high salinity water thus more low salinity water is needed to compensate for the decrease in oil recovery due
to dispersion.
181
37 1
35 0.95
33 0.9
Oil recovery
31 0.85
FRC
29 0.8
27 0.75
25 0.7
0 0.5 1 1.5 2 0 0.5 1 1.5
no dispersion post BT of conv no dispersion post BT of conv
NTD = 0.47 post BT of conv NTD = 0.47 post BT of conv
no dispersion post 90 WC of conv no dispersion post 90WC of conv
NTD = 0.47 post 90 WC of conv NTD = 0.47 post 90 WC of conv
Figure 6.14: Oil recoveries from different slug sizes of low salinity water Figure 6.15: Oil recoveries from different slug sizes of low salinity
injected post high salinity waterflooding into a 3D reservoir with water injected post high salinity waterflooding as a fraction of the oil
complex heterogeneities taken from SPE 10 model 2. The start of low recovery (FRC) obtained from continuous low salinity waterflooding at 2
salinity water injection was varied: post the breakthrough of high salinity PVI. FRC is almost identical when the start of low salinity water
water (BT of conv) and when the producing water cut from high salinity injection is varied. However, this is not the case for the case with high
water has reached 90% (90 WC of conv). Generally, the effect of dispersivity (NTD = 0.47). High dispersivity promotes more mixing of
dispersion on low salinity waterflooding is minor compared to reservoir low salinity water with the injected high salinity water which reduced
heterogeneity. This is due to the fact that at this realistic model the the microscopic displacement. As a result, the oil recovery is lower
permeability variations are very large (some areas are 0.001 md compared to the no dispersion case for small slug sizes (0.1-0.3 PV).
compared to 20,000 md). The best slug size is around 1.3 PV for both cases compared with the
best slug size of injecting low salinity water immediately (0.6-1 PV).
This would be expected as low salinity water mixes with larger amount
of high salinity water thus more low salinity water is needed to
compensate for the decrease in oil recovery due to dispersion.
1.2
0.8
0.6
Df
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
LSW slug size (PV)
NTD = 0.47 post BT of conv NTD = 0.47 post 90 WC of conv
Figure 6.16: The oil recoveries for the case of NTD = 0.47 ( Df ) as defined by equation 5.1 are compared to the no dispersion case for injecting
different slug sizes of low salinity water post high salinity waterflooding into a 3D reservoir with complex heterogeneities taken from SPE 10
model 2. For NTD = 0.47 when injecting low salinity water post the breakthrough of high salinity water, Df decreases when the slug size is
increased from 0.1 to 0.3 PV. The percentage of increase in oil recovery for the no dispersion case is higher than that for NTD = 0.47 (Figure
6.15) so Df for 0.3 PV slug would be lower than for the smaller slug size (0.1 PV). When the slug size is large (1-1.3 PV), dispersion has a minor
effect on low salinity waterflooding, so the overall oil recovery is similar to the no dispersion case (Df ~ 1). Generally, injecting low salinity water
when the produced water cut from conventional waterflooding has reached 90% WC recovers more oil compared with injecting low salinity water
post the breakthrough of high salinity water. For the latter case, the volume of the injected chase high salinity water is larger which results in
increasing the salinity of the injected salinity water and thus reducing the microscopic displacement.
182
Water salinity and oil saturation distributions are shown in Figure 6.17. At 0.1 PV slug, dispersion has
increased the salinity of the injected low salinity water to intermediate salinities. Most of these intermediate
salinities are above the upper salinity threshold so the flow of water is controlled by the high salinity
waterflooding. Therefore, the microscopic displacement is reduced to that of a high salinity flooding and as a
result the oil recovery is lower compared to the no dispersion case. At 1.3 PV slug, dispersion promotes the
injected low salinity water to contact the low permeability layers thus the sweep efficiency is improved but
the microscopic displacement is reduced. However, due to the large volume of injected high salinity water
before low salinity water, low salinity water mixes with this high salinity water which has resulted in similar
oil recoveries to the no dispersion case (Df ~ 1).
183
Layer 1
PERM (md)
Direction of flow
Oil saturation at 2 PVI Oil saturation Oil saturation at 2 PVI

0.1 PV slug of lowsal 0.1 PV slug of lowsal



30
No dispersion 27 NTD = 0.47

24
Post BT of conv Post 90 WC of conv 21

Post BT of conv Post 90 WC of conv
18
15
13
10
7
Figure 6.17: Oil saturation and water salinity distributions at 2 PVI for injecting low salinity water post high salinity waterflooding into a 3D
reservoir with complex heterogeneities taken from SPE 10 model 2. At 0.1 PV slug, dispersion has increased the salinity of the injected low salinity
water to intermediate salinities. Most of these intermediate salinities are above the upper salinity threshold so the flow of water is controlled by the
high salinity waterflooding. Therefore, the microscopic displacement is reduced to that of a high salinity flooding and as a result the oil recovery is
lower compared to the no dispersion case. At 1.3 PV slug, dispersion promotes the injected low salinity water to contact the low permeability layers
thus the sweep efficiency is improved but the microscopic displacement is reduced. However, due to the large volume of injected high salinity water
before low salinity water, low salinity water mixes with this high salinity water which has resulted in similar oil recoveries to the no dispersion case
(Df ~ 1).
184
The results are compared at 10 PVI. The oil recoveries are shown in Figure 6.18 as a fraction of the oil
recovery (FRC) obtained from continuous low salinity waterflooding at 10 PVI. Oil recoveries for the case
with high dispersivity (NTD = 2) are compared with the no dispersion scenario in Figure 6.19 using Df. The
best slug size of injecting low salinity water post conventional waterflooding into a reservoir connected to an
aquifer depends on dispersion. The best slug size is around 4 PV of oil zone for the case with no dispersion
increasing to 5 PV of oil zone for the case with high dispersivity (NTD = 0.2). At small slug sizes, dispersion
destroys the benefit of low salinity water by reducing the microscopic displacement since most of the
intermediate salinities are above the upper salinity threshold so the flow of water is controlled by high
salinity waterflooding as shown in the oil saturation and water salinity distributions in Figure 6.20. At large
slug sizes, most of the reservoir has been swept and dispersion has a minor effect on low salinity
waterflooding since the model is homogeneous and the volume of low permeability is zero
It was shown in section 6.2.1 C that when low salinity water is injected immediately, the best slug size is
around 2-3 PV which is smaller than when it is injected post high salinity waterflooding. This would be
expected as low salinity water mixes with larger amount of high salinity water thus more low salinity water
is needed to compensate for the decrease in oil recovery due to dispersion.
1 1
0.9
0.95
0.8
0.9 0.7
0.6
0.85 0.5
Df
FRC
0.4 Salinity Thresholds

0.8 Salinity Thresholds
1,000-7,000 ppm 0.3 1,000-7,000 ppm
0.75 0.2
0.1
0.7 0
0 1 2 3 4 5 0 1 2 3 4 5
LSW slug size (PV of oil zone) LSW slug size (PV of oil zone)
no dispersion post BT of conv
NTD = 0.2 post BT of conv
no dispersion post 90 WC of conv NTD = 0.2 post BT of conv NTD = 0.2 post 90 WC of conv
NTD = 0.2 post 90 WC of conv
Figure 6.18: Oil recoveries for different slug sizes of injecting low Figure 6.19: Oil recoveries at NTD = 0.2 ( Df ) as defined by equation
salinity water post high salinity waterflooding into a reservoir connected 5.1 compared with the no dispersion case. At small slug sizes, dispersion
to an aquifer as a fraction of oil recovery from continuous low salinity destroys the benefit of low salinity water by reducing the microscopic
waterflooding at 10 PVI (FRC). The results are shown as a fraction of the displacement since most of the intermediate salinities are above the
oil recovery obtained from continuous low salinity waterflooding at 10 upper salinity threshold so the flow of water is controlled by high
PVI. The best slug size depends on dispersion. The best slug size is salinity waterflooding.
around 4 PV of oil zone for the case with no dispersion increasing to 5
PV of oil zone for the case with high dispersivity (NTD = 0.2).
Dispersion destroys the benefit of low salinity water at small slug sizes.
185
2 PV slug of lowsal Oil saturation 2 PV slug of lowsal
Salinity at 10 PV Salinity (1,000 ppm)

Salinity at 10 PV
30
27
24
21
18
15
5 PV slug of lowsal 13 5 PV slug of lowsal
10
7
Upper threshold
1
Lower threshold
Figure 6.20: Oil saturation and water salinity distributions at 10 PVI for injecting low salinity water post high salinity waterflooding into a reservoir
connected to an aquifer. At 2 PV slug, dispersion destroys the benefit of low salinity water by reducing the microscopic displacement since most of
the intermediate salinities are above the upper salinity threshold so the flow of water is controlled by high salinity waterflooding At 5 PV, most of the
reservoir has been swept and dispersion has a minor effect on low salinity waterflooding since the model is homogeneous and the volume of low
permeability is zero.
186
This chapter has investigated the effect of physical diffusion/dispersion and reservoir heterogeneity on the
performance of low salinity waterflooding when it is injected as a slug. The injection of low salinity water as
a slug is the most realistic case since it is often not economically feasible to inject low salinity water
continuously due to lack of availability of low salinity water, the need to dispose of saline produced water by
reinjection or due to limitations in desalination capacity. The impact of physical diffusion/dispersion and
reservoir heterogeneity was evaluated first on low salinity water when it is injected immediately and then
when it is injected post high salinity water. CMG STARS simulator was used for this investigation.
Two salinity thresholds were considered as a sensitivity parameter for immediate low salinity water injection
when modelling low salinity waterflooding: the first one assumed that the thresholds are equal to the injected
low salinity water and the connate water while the other more realistic one assumed that the thresholds are
close to the injected low salinity water. The investigations looked into injecting low salinity water
immediately and post high salinity waterflooding. The investigation of slug size for low salinity post
conventional water injection just used the more realistic thresholds.
Immediate Injection of Low Salinity Water
The investigations looked into both laboratory scale and reservoir scale models. Assuming total water
injection of 1 PV, in the laboratory scale models, the best slug size of low salinity waterflooding was found
to be around 0.4-0.6 PV. The effect of diffusion was found to be most significant for very small slug sizes
(0.1 PV) when the upper salinity threshold is close to the injected water salinity due to the fact that most of
the intermediate salinities in the model are above the upper salinity threshold so the flow of water with
intermediate salinities is controlled by the high salinity water relative permeability curves. In this case, the
overall oil recovery is similar to that by conventional waterflooding.
On the reservoir scale with more complex heterogeneities, assuming total water injection of 1 PV, it was
found that the best slug size of the injected low salinity water depends on the upper salinity threshold. When
the upper salinity threshold was close to the injected water salinity, the best slug size was found to be around
1 PV, while it was found to be around 0.6 PV when the upper salinity threshold was close to the connate
water salinity.
On the reservoir scale, the level of dispersion was found to have different effects on oil recovery by low
salinity waterflooding when the upper salinity thresholds is close to the injected water salinity. When the
transverse dispersion is 30 times less than the longitudinal dispersion as suggested by Mahadevan et al.
(2003), it was found that it lowers the oil recovery compared with the no dispersion case. Dispersion
increases the salinity of low salinity water as it travels through the reservoir which results in intermediate
salinities that are above the upper salinity threshold so the flow of water is controlled by the high salinity
water relative permeability curves. Therefore, the overall oil recovery is lower than that in the no dispersion
case. However, transverse dispersion is not sufficiently high that it dominates the flow thus not all the low
permeability areas are contacted by the low salinity water. If a transverse dispersion is assumed to be 30
187
times less than a longitudinal dispersion that is 5 % of the reservoir length with an upper salinity threshold
close to the injected water salinity, it was found that dispersion improves the oil recovery when the low
salinity water slug is large because dispersion dominates the flow. This means there is good conformance all
over the reservoir.
Based on these results, the main effect of dispersion, in combination with heterogeneity, is to degrade the
low salinity slug. This means than an even larger slug size is needed (compared with the sizes discussed in
the previous chapter) in order to maximise recovery. The results suggest the slug size needs to be between
0.6 and 1 PV in size, depending on the salinity thresholds. This is effectively continuous low salinity
injection.
For the case of injecting low salinity water as slugs into a reservoir connected to an aquifer, it was found that
dispersion increases the best slug size of injecting low salinity water from 2 PV to around 3 PV of oil zone.
This is due to the fact that smaller (but still large) dispersion destroys the benefit of low salinity water.
Injection of Low Salinity Water Post High Salinity Waterflooding
On the laboratory scale, assuming a total water injection of 2 PV, the best slug size of low salinity water was
found to be 0.5-0.6 PV when low salinity water is injected at the breakthrough of conventional water. This is
similar to the best slug size when injecting low salinity water immediately. However, it increases to 1 PV
when the start of low salinity water injection is delayed until the producing water cut has reached 90% water
cut. This is due to the fact that more high salinity water is injected and low salinity water mixes with it so the
microscopic displacement is reduced. As a result, more low salinity water injection is needed to compensate
for the decrease in oil recovery due to diffusion. At very small slug sizes, diffusion destroys the benefit of
low salinity waterflooding by increasing the salinity of the injected water thus the microscopic displacement
is reduced. At very large slug sizes, diffusion improves the oil recovery by reducing the water salinity in the
low permeability layer thus the sweep efficiency is improved although the microscopic displacement is
reduced. The incremental oil recovery due to diffusion is larger when the model contains larger volumes of
low permeability.
On the reservoir scale with more complex heterogeneities, assuming a total water injection of 2 PV, the best
slug size of low salinity water was found to be 1.3 PV compared with (0.6-1 PV) when low salinity water is
injected immediately. This would be expected as low salinity water mixes with the larger pre-existing
volume of high salinity water thus more low salinity water is needed to compensate for the decrease in oil
recovery due to dispersion. At very small slug sizes, dispersion decreases the oil recovery as it increases the
salinity of the injected low salinity water so the microscopic displacement is reduced. At large slug sizes,
dispersion promotes the injected low salinity water to contact the low permeability layers thus the sweep
efficiency is improved but the microscopic displacement is reduced. Since a large amount of high salinity
water is injected before low salinity water, mixing between the low and high salinity water results in similar
oil recovery to the no dispersion case. Generally, the effect of dispersion on low salinity waterflooding is
minor compared with reservoir heterogeneity at the reservoir scale. This is due to the fact that the
permeability variations are very large (some areas are 0.001 md compared to 20,000 md) so the displacement
188
is mainly through the high permeability areas and only small portions of the low permeability areas are
contacted by high dispersivities.
For the case of injecting low salinity water as slugs post high salinity waterflooding into a reservoir
connected to an aquifer, it was found that dispersion increases the best slug size of injecting low salinity
water from 4 PV to around 5 PV of oil zone assuming a total water injection of 10 PV of oil zone. This is
due to the fact that smaller (but still large) dispersion destroys the benefit of low salinity water. This best
slug size is larger than that when injecting low salinity water immediately (2-3 PV). This would be expected
as low salinity water mixes with larger amount of high salinity water thus more low salinity water is needed
to compensate for the decrease in oil recovery due to dispersion.
These results suggest that low salinity water needs to be injected continuously in previously water flooded
fields, in order to gain a worthwhile EOR benefit.
189
CHAPTER 7: UPSCALING LOW SALINITY WATERFLOODING
It is often difficult or impossible to model the impact of geological heterogeneity on the performance of low
salinity waterflooding due to computational constraints. A very fine grid may be needed to capture the
spatial variations in permeability and porosity which in turn can result in excessive run times and huge
volumes of output data. Therefore, it is essential to upscale the fine grid model to a coarse grid that captures
the effective flow properties.
This chapter proposes an analytical method to upscale low salinity waterflooding in layered models by
modifying Hearn’s method. First, the reasons for choosing this method rather than other two-phase upscaling
methods are discussed and then the pseudo relative permeabilities are generated using the conventional
Hearn’s method without any modifications. Finally a new analytical method is proposed by modifying
Hearn’s method to consider and include the effect of the connate water bank and the predictions are tested
using a 2D layered model with various number of layers: two, three and five layers. The chapter ends with
discussing the limitations of using this analytical method.
7.1 Why Choose Hearn’s Method rather than other Two-Phase Upscaling Methods?
Unlike single phase upscaling, two-phase upscaling generates pseudo relative permeabilities that capture the
impacts of detailed heterogeneity effects such as channelling on two-phase flow. They are divided into two
categories: unsteady state methods and steady state methods. Unsteady state methods (such as the Pore
Volume Weighted method) minimize the numerical dispersion that arises when upscaling. However, they
require running the fine scale model to use their outputs for generating the pseudo relative permeabilities
(Kyte and Berry 1975, Barker and Dupouy 1999).
Steady state methods (such as capillary limit and viscous limit) do not require the outputs from running a full
fine scale simulation for generating the pseudo relative permeabilities (Jonoud and Jackson 2008, Ekrann
and Aasen 2000, Odsaeter et al. 2015). However, they require more than one rock type to be present in the
fine scale model.
In this thesis, only one rock type will be used in the investigations so steady state upscaling will not serve the
purpose of generating pseudo relative permeabilities that will capture the impact of geological heterogeneity
on low salinity waterflooding.
First, the pseudo relative permeability curves are generated using Hearn’s method without any modifications
then the new analytical method is proposed.
190
7.2 Description of Hearn’s Method
Hearn (1971) showed that pseudo relative permeabilities for conventional waterflooding can be generated
from a 2D layered reservoir and used in a 1D homogenous model. To generate these pseudo using Hearn’s
method, a constant pressure boundary condition is required so that the pressure difference between the
injector and producer is the same for each layer. While this method was initially developed for high salinity
waterflooding, it will be modified to upscale low salinity waterflooding. The equations for generating pseudo
relative permeabilities using Hearn’s method were described in chapter 2.
7.3 Modifications of Hearn’s Method by the Inclusion of Buckley Leverett Analysis
In this section, an analytical method is proposed to modify Hearn’s method by the inclusion of Buckley
Leverett solution of immiscible displacement. The method is described first for high salinity waterflooding
then for low salinity waterflooding.
7.3.1 Upscaling High Salinity Waterflooding by a Modified Hearn’s Method
The original Hearn’s method (Hearn 1971) generates the pseudo relative permeability curves by ordering the
layers as per their absolute permeabilities from top to bottom. Then, the pseudo relative permeability values
are determined based on the water breakthrough of each layer. The calculation assumes that the waterflood is
piston like, i.e. only residual oil remains in all areas that have been waterflooded whereas the water
saturation is equal to initial water saturation elsewhere. In this section, it is assumed that the displacement in
each layer can be modelled using Buckley Leverett theory, meaning there is a shock front followed by a
rarefaction. The advance of the shock front and saturations in the rarefaction depend upon the gradient of the
fractional flow curve for those saturations. A new method is proposed to improve upscaling high salinity
waterflooding based on the Buckley Leverett solution of immiscible displacement. A 2D model with three
layers is used to illustrate the proposed method. The model dimensions in addition to the rock and fluid
properties are shown in Table 7.1. The permeability ratio between the layers is 3. The relative permeability
curves used for this model are shown in Figure 7.1.
Table 7.1: Reservoir and fluid properties used in the 2D simulation
1
model to illustrate the proposed analytical method.
0.9
Property Value
Model dimensions (x, y, z) 1,000 × 100 × 10 m 0.8
Number of grid blocks (x, y, z) 100 × 1 × 20 0.7

Grid size (dx, dy, dz) 10 × 100 × 0.5 m
0.6
Porosity 0.2
Permeability (layers 1 ,2, 3) 100 , 33, 11 md 0.5 krwH
Initial reservoir pressure 265 bar 0.4
kroH
Rock compressibility 0.3 × 10-5 bara-1
0.3
Water compressibility 4.6 × 10-5 bara-1
Oil density 850 kg/m3 0.2
Water density 1,000 kg/m3 0.1
Oil viscosity 2 cp
0
Water viscosity 1 cp
0 0.2 0.4 0.6 0.8
Oil formation volume factor 0.999 rm3/sm3
Sw
Water formation volume factor 1.03 rm3/sm3
Figure 7.1: Relative permeability curves for high salinity waterflooding.
191
High salinity water is injected along the left hand face of the model at a constant rate of 60 m3/d and
produced along the right hand face of the model.
First, the water saturation profile is generated and is divided into four regions as shown in Figure 7.2. Then,
for each region the pseudo relative permeability values are calculated using the following equations:
∑𝑁𝑁
𝑖𝑖=1 𝐾𝐾𝑖𝑖 ℎ𝑖𝑖 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟
𝑘𝑘�
𝑟𝑟𝑟𝑟 = (7.1)
∑𝑁𝑁
𝑖𝑖=1 𝐾𝐾𝑖𝑖 ℎ𝑖𝑖
∑𝑁𝑁
𝑘𝑘�
𝑟𝑟𝑟𝑟 = (7.2)
∑𝑁𝑁
where
𝑘𝑘� �
𝑟𝑟𝑟𝑟 and 𝑘𝑘𝑟𝑟𝑟𝑟 are the pseudo water and oil relative permeability, respectively.
The 2D layered model was homogenised into a 1D model with an absolute permeability that is calculated by
taking the arithmetic average of the individual layer permeabilities. The pseudo relative permeabilities for
high salinity water were used in the 1D homogenous model.
Hearn’s method assumes no cross flow between layers (kv = 0). To evaluate the effect of cross flow between
layers on the simulation predictions using the pseudo relative permeabilities, the vertical permeability was
varied for the 2D layered model: (kv = 0), (kv =0.01) kh and (kv = kh). In all cases, a constant pressure
boundary condition was used so the pressure difference between the injector and producer is the same for
each layer.
0.8
4
0.7 3
0.6 2
0.5
layer 1
0.4
Sw
layer 2
0.3 layer 3
1
0.2
0.1
0
0 500 1000 1500 2000 2500 3000 3500
Distance (ft)
Figure 7.2: Regions of the water saturation profile for high salinity waterflooding at 0.16 PV.
7.3.2 Upscaling Low Salinity Waterflooding by a Modified Hearn’s Method

In this section, an analytical method is proposed to improve upscaling low salinity waterflooding based on
modifying Hearn’s method to include the connate water bank. The method is described for a two layer
reservoir and it can be extended to a multi-layer reservoir. The model dimensions, fluid and rock properties
are the same as those for high salinity waterflooding (Table 7.1). However, for low salinity wasterflooding a
two layer model (permeabilities: 100 and 33 md). The connate water salinity is 30,000 ppm and the injected
192
low salinity water is 1,000 ppm. The salinity thresholds are 1,000-7,000 ppm. These thresholds are chosen as
they represent the more realistic thresholds as confirmed by previous experiments (Tang and Morrow 1999).
Low salinity water is injected along the left hand face of the model at a constant rate of 60 m3/d and
produced along the right hand face of the model. The relative permeability curves and their corresponding
fractional flow curves are shown in Figures 7.3a and 7.3b, respectively. The difference in connate water
saturation between the two relative permeability curves was discussed in section 3.1
1 1
0.9 0.9
0.8 Sw,LS = 0.66
0.8
Fractional flow
0.7 0.7 Sw,HS = 0.47
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Sw
Sw
krwH
krwh kroH
kroh krwL
krwL kroL
kroL
HS LS
Figure 7.3: a) Relative permeability curves b) Fractional flow curves

Relative permeability curves for high salinity water are labelled by H while for low salinity water they are labelled by L. The tangent line
is drawn to the low salinity water fractional flow curve to estimate the two shock front saturations. Sw,HS is the leading shock front
saturation and Sw,LS is the trailing shock front saturation.
First, the water saturation profile at 0.2 PVI is calculated based on the analytical Buckley Leverett solution
for each layer.
𝑡𝑡 𝑣𝑣𝑖𝑖
𝑑𝑑𝑆𝑆𝑤𝑤 𝑆𝑆𝑤𝑤
𝑥𝑥𝑆𝑆𝑤𝑤 = (7.3)
𝜙𝜙
where
𝑥𝑥𝑆𝑆𝑤𝑤 is the position of any water saturation value, L
t is the time at which the saturation profile is calculated, t
𝑣𝑣𝑖𝑖 is the frontal velocity for any layer, L/t
𝜙𝜙 is the porosity, (dimensionless)
is the slope of the tangent line at any water saturation (estimated from the fractional flow curves in
Figure 7.4).
193
1
0.9
𝒅𝒅𝒇𝒇𝑳𝑳𝑳𝑳,𝒘𝒘
0.8 𝒅𝒅𝒇𝒇𝑯𝑯𝑯𝑯,𝒘𝒘
= 2.39 𝒅𝒅𝑺𝑺𝒘𝒘
= 1.37
Fractional flow 0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Sw
HS LS
𝑑𝑑𝑓𝑓𝐿𝐿𝐿𝐿,𝑤𝑤
Figure 7.4: The slope of the tangent line to the LS curve ( ) is used to determine the position of the trailing shock front
𝑑𝑑𝑓𝑓𝐻𝐻𝐻𝐻,𝑤𝑤
for each layer. The position of the leading shock front for each layer is determined from the slope of the line ( ) that
connects the connate water saturation and the leading shock front. For any saturation of Sw greater than the trailing shock
front water saturation, the slope of the tangent line is calculated to determine the position of this Sw for each layer.
For each layer, there are three different regions in the saturation profile.
Region 1:
This region is ahead of the leading shock front (for any Sw < 0.47). It starts from 𝑥𝑥𝐻𝐻𝐻𝐻,𝑤𝑤 and ends at the
producer location and it has a constant Sw value of 0.2.
𝑑𝑑𝑓𝑓𝐻𝐻𝐻𝐻,𝑤𝑤
𝑑𝑑𝑆𝑆𝑤𝑤 𝑆𝑆𝑤𝑤 = 0.47
𝑥𝑥𝐻𝐻𝐻𝐻,𝑤𝑤 = (7.4)
∅
Region 2:
This region is the connate water bank (0.47 < Sw < 0.66). It starts from 𝑥𝑥𝐿𝐿𝐿𝐿,𝑤𝑤 and ends at 𝑥𝑥𝐻𝐻𝐻𝐻,𝑤𝑤 and it has a
constant Sw value of 0.47.
𝑑𝑑𝑓𝑓𝐿𝐿𝐿𝐿,𝑤𝑤
𝑑𝑑𝑆𝑆𝑤𝑤 𝑆𝑆𝑤𝑤 = 0.66
𝑥𝑥𝐿𝐿𝐿𝐿,𝑤𝑤 = (7.5)
∅
Region 3:
This region is behind the trailing shock front (for any Sw > 0.66). It starts at the injector and ends at 𝑥𝑥𝐿𝐿𝐿𝐿,𝑤𝑤 .
𝑑𝑑𝑓𝑓𝐿𝐿𝐿𝐿,𝑤𝑤 𝑑𝑑𝑓𝑓𝑤𝑤
For every Sw, 𝑥𝑥𝐿𝐿𝐿𝐿,𝑤𝑤 is calculated from the slope of by using equation 7.3 but using the for that
𝑑𝑑𝑆𝑆𝑤𝑤 𝑑𝑑𝑆𝑆𝑤𝑤
Sw.
The resulting water saturation profile for both layers at 0.2 PVI are shown in Figure 7.5.
194
1
0.9
0.8
0.7
0.6
0.5 layer 1
Sw
0.4 layer 2
0.3
0.2
0.1
0
0 500 1000 1500 2000 2500 3000 3500
Distance (ft)
Figure 7.5: Water saturation profile at 0.2 PVI calculated analytically using the Buckley Leverett solution. The leading shock front
progresses in the reservoir at different rates due to variations in the layers’ permeabilities.
Second, the water saturation profile is divided into five regions based on the salinities of each layer as
highlighted in Figure 7.6.
1
0.9 5
0.8
4
3
0.7
0.6 2
0.5
Sw
layer 1
0.4
0.3 layer 2
1
0.2
0.1
0
0 500 1000 1500 2000 2500 3000 3500
Distance (ft)
Figure 7.6: Regions of the water saturation profile at 0.2 PV of low salinity water injection based on the salinities of each layer.
The average water saturation and salinity for each region are calculated using the following equations:
∑𝑁𝑁
𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖 𝑆𝑆𝑤𝑤𝑤𝑤
𝑆𝑆�
𝑤𝑤 = (7.6)
∑𝑁𝑁𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖
∑𝑁𝑁
𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖 𝑆𝑆𝑤𝑤𝑤𝑤 𝐶𝐶𝑖𝑖
𝐶𝐶̃ = (7.7)
∑𝑁𝑁
𝑖𝑖=1 ℎ𝑖𝑖 𝜙𝜙𝑖𝑖 𝑆𝑆𝑤𝑤𝑤𝑤
where
𝐶𝐶̃ is the average water salinity in ppm for layers i to N, in this case 1 to 2
i refers to any layer, in this case 1 and 2
N is the total number of layers, in this case it is 2
195
The layer salinities, the average water saturation and the average salinity for each region are summarized in
Table 7.2.
Table 7.2: Average water saturation and average salinity for each region.
Region Layer Sw 𝑆𝑆�

Salinity 𝐶𝐶̃ (1,000 ppm)
𝑤𝑤 (Equation 7.4) (1,000 ppm) (Equation 7.5)
1 0.20 7
1 0.20 7.00
2 0.20 7
1 0.47 7
2 0.33 7.00
2 0.20 7
1 0.66 1
3 0.43 2.39
2 0.20 7
1 0.70 1
4 0.58 3.41
2 0.47 7
1 0.75 1
5 0.70 1
2 0.66 1
The next step is to generate the pseudo relative permeability for each region. The general equations for
calculating the water and oil pseudo relative permeability are as follows:
∑𝑁𝑁
𝑘𝑘�
𝑟𝑟𝑟𝑟 = (7.8)
∑𝑁𝑁
∑𝑁𝑁
𝑘𝑘�
𝑟𝑟𝑟𝑟 = (7.9)
∑𝑁𝑁
where
𝑘𝑘� �
𝑟𝑟𝑟𝑟 and 𝑘𝑘𝑟𝑟𝑟𝑟 are the pseudo water and oil relative permeability, respectively.
Note: since there are two different salinities in the reservoir, the pseudo calculated in this step is for the
water phase irrespective of salinity, the method for calculating the corresponding pseudo for the low and
high salinity water curves is described below.
Regions 1 and 2:
This is a high salinity water region, so there is no pseudo relative permeability for the low salinity water
curve, therefore the pseudo relative permeabilities calculated above are for the high salinity water curves.
Region 5:
This is a low salinity water region, so there is no pseudo relative permeability for the high salinity water
curve, therefore the pseudo relative permeabilities calculated above are for the low salinity water curves.
Region 3 and 4:
These regions have intermediate salinity water. The relation between the pseudo relative permeability for the
intermediate salinity at the average water saturation (in Table 7.2) and their values at high and low salinity
water is as follows:
𝑘𝑘� � ̃ ��
𝑟𝑟𝑟𝑟 = 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 �1 − 𝐶𝐶 � + 𝐶𝐶 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 (7.10)
𝑘𝑘� � ̃ ��
𝑟𝑟𝑟𝑟 = 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 �1 − 𝐶𝐶 � + 𝐶𝐶 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 (7.11)
196
The calculated pseudo water and oil relative permeabilities (𝑘𝑘� �
𝑟𝑟𝑟𝑟 and 𝑘𝑘𝑟𝑟𝑟𝑟 ) in the above equations are for the
intermediate salinity water. It is assumed that the pseudo high salinity water relative permeabilities (𝑘𝑘�
𝑟𝑟𝑟𝑟𝑟𝑟
and 𝑘𝑘�
𝑟𝑟𝑟𝑟𝑟𝑟 ) are the same as the rock high salinity water relative permeability curves. In the case that the
average water saturation of these two regions are between two tabulated water saturations in the high salinity
water relative permeability, then its pseudo high salinity water relative permeability values are linearly
interpolated using the two tabulated water saturations and their corresponding relatitve permability values.
Then, the pseudo low salinity water relative permeability values (𝑘𝑘� �
𝑟𝑟𝑟𝑟𝑟𝑟 and 𝑘𝑘𝑟𝑟𝑟𝑟𝑟𝑟 ) for this average saturation
are calculated using equations 7.8 and 7.9 as the average salinity is already known.
7.4 Method
7.4.1 Upscaling Low Salinity Waterflooding
A- Original Hearn’s Method
To generate the pseudo relative permeabilities using Hearn’s method, it is assumed that two separate floods
occur. The first one is for high salinity water displacing oil and the other one is for low salinity water
displacing oil in addition to the presence of connate water that has the same salinity as the injected low
salinity water. Since the connate water salinities are the same as the injected water salinities for both floods,
there is no connate water bank. The two layer model (described in Table 7.3) is used for testing the original
Hearn’s method for upscaling low salinity waterflooding.
The pseudo relative permeabilities for the two floods are generated using Hearn’s methods.
The 2D layered model was homogenised into a 1D model with an absolute permeability that is calculated by
the arithmetic average of the individual layer permeabilities. The pseudo relative permeabilities for both high
salinity and low salinity water were used in the 1D homogenous model.
Hearn’s method assumes no cross flow between layers (kv = 0). Therefore, the vertical permeability was set
to be zero. A constant pressure boundary condition was used so the pressure difference between the injector
and producer is the same for each layer.
B- Upscaling Low Salinity Waterflooding Using A Modified Hearn’s Method
The method described in section 7.3 is used for upscaling low salinity waterflooding. Three different layered
models (described in table 7.3) were used to test this method.
For all cases, the 2D layered model was homogenised into a 1D model with an absolute permeability that is
calculated by the arithmetic average of the individual layer permeabilities. The pseudo relative
permeabilities for both high salinity and low salinity water were used in the 1D homogenous model.
Hearn’s method assumes no cross flow between layers (kv = 0). Therefore, the vertical permeability was set
to be zero for all cases. A constant pressure boundary condition was used so the pressure difference between
the injector and producer is the same for each layer.
197
7.5 Evaluation of New Method
In this section, the proposed analytical method (described in section 7.3) is tested using reservoir scale
models with different numbers of layers as shown in Table 7.3. For all cases, the vertical permeability was
set to zero since Hearn’s method assumes that there is no cross flow between layers. The model dimensions,
rock and fluid properties are provided in Table 7.1.
Table 7.3: Scenarios for testing the analytical method to upscale low salinity waterflooding.
Case Permeabilities (md) Layer thickness (m)
Layer 1: 100 Layer 1: 5
Two layer model
Layer 2: 33.3 Layer 2: 5
Layer 1 100 Layer 1 3.3
Three layer model Layer 2: 33.3 Layer 2: 3.3
Layer 3: 11.1 Layer 3: 3.3
Layer 1: 100 Layer 1: 2
Five layer model Layer 3: 11.1 Layer 3: 2
Before testing this method for upscaling low salinity waterflooding, the method is tested for high salinity
waterflooding using the three layer model and the results obtained using the original Hearn’s method are
compared with a 1D model without upscaling. This test used the original high salinity relative permeability
curves. Then, the proposed upscaling method is tested for low salinity waterflooding using the two layer
model, the three layer model and the five layer model.
7.6 Results
7.6.1 Upscaling High Salinity Waterflooding by a Modified Hearn’s Method
The resulting pseudo relative permeability curves are plotted in Figure 7.7 and compared with the pseudo
relative permeability curves generated by the original Hearn’s method in addition to the rock relative
permeability curves.
0.9
0.8
0.7
krwH
KrwH
0.6 kroH
KroH
pseudo krwH (original
pseudo krwh (original Hearn)
Hearn)
0.5
pseudo kroh (original Hearn)
pseudo kroH (original Hearn)
0.4 pseudo krwH (modified
pseudokrwh (modifiedHearn)
Hearn)
0.3 pseudo kroH (modified Hearn)
pseudo kroh (modified Hearn)
0.2
0.1
0
0 0.2 0.4 0.6 0.8
Sw
Figure 7.7: The pseudo relative permeability curves generated for high salinity waterflooding by modifying Hearn’s method compared with the
rock relative permeability curves and the pseudo relative permeability created using the original Hearn’s method for a three layer model.
198
The oil recoveries, water cut and average reservoir pressure obtained by the 1D homogenized model using
the modified Hearn’s method are shown in Figure 7.8 and compared with the 2D layered model in addition
to the 1D homogenised model using the original Hearn’s method and the 1D homogenised model using the
rock relative permeability curves. The oil recovery obtained by the 1D model with pseudo generated by the
modified Hearn’s method (dashed purple line) are much closer to the 2D model with kv = 0 (light blue line)
than the 1D model with pseudo generated by the original Hearn’s method (dashed green line). This modified
Hearn’s method is based on the assumption that there is no cross flow between layers, so it is not applicable
for any case where kv > 0.
a) b)
0.8 1
0.7 0.9
0.8
0.6
0.7
Water cut (%)

0.5
Oil recovery
0.6
0.4 0.5
0.3 0.4
0.3
0.2
0.2
0.1
0.1
0 0
0 0.5 PV 1 1.5 2 0 0.5 1 1.5 2
PV
2D kv=0 (constant P) 2D kv=0 (constant P)
2D kv=0.01 kh (constant P) 2D kv=0.01 kh (constant P)
2D kv=kh (constant P) 2D kv=kh (constant P)
1D with original rel perm 1D with original rel perm
1D Upscaling (Original Hearn) 1D Upscaling (Original Hearn)
1D Upscaling (Modified Hearn) 1D Upscaling (Modified Hearn)
c)
270
Average reservoir pressure (bar)
250
230
210
190
170
150
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
PV
2D kv=0 (constant P) 2D kv=0.01 kh (constant P)
2D kv=kh (constant P) 1D with original rel perm
1D Upscaling (Original Hearn) 1D Upscaling (Modified Hearn)
Figure 7.8: a) oil recovery vs PVI b) water cut vs PVI c) average reservoir pressure vs PVI
Oil recovery, water cut and average reservoir pressure for high salinity waterflooding in a three layer model. The cases shown are: the 2D
heterogeneous model with cases of various vertical permeabilities, the 1D homogeneous model with the pseudo relative permeabilities generated
using the modified Hearn’s method, the 1D homogeneous case with pseudo relative permeabilities generated using the original Hearn’s method and
the 1D homogeneous case using original relative permeability curves. The oil recovery obtained by the 1D model with pseudo generated by the
modified Hearn’s method (dashed purple line) are much closer to the 2D model with kv = 0 (light blue line) than the 1D model with pseudo generated
by the original Hearn’s method (dashed green line). This modified Hearn’s method is based on the assumption that there is no cross flow between
layers, so it is not applicable for any case where kv > 0.
199
7.6.2 Upscaling Low Salinity Waterflooding
A- Original Hearn’s Method
The calculated pseudo relative permeabilities for high salinity and low salinity waterflooding are listed in
Table 7.4 and compared with the original relative permeability curves in Figure 7.9.
Table 7.4: Pseudo relative permeability calculated by Hearn’s method.
High salinity water Low salinity water
Sw (average) kwh (pseudo) kroh (pseudo) Sw (average) krwL (pseudo) kroL (pseudo)
0.200 0.000 1.000 0.150 0.000 1.000
0.475 0.375 0.249 0.550 0.263 0.250
0.750 0.500 0.000 0.900 0.350 0.0000
0.9
0.8
0.7 krwH
KrwH
0.6 kroH
KroH
krwL
KrwL
0.5
kroL
KroL
0.4
pseudo
pseudo krwH
krwH
0.3 pseudo
pseudo kroHkroH
0.2 pseudo
pseudo krwLkrwL
pseudo
pseudo kroLkroL
0.1
0
0 0.2 0.4 0.6 0.8 1
Sw
Figure 7.9: Pseudo relative permeability curves for high salinity and low salinity water using Hearn’s method and assuming that two separate
floods occur: The first one is for high salinity water displacing oil and the other one is for low salinity water displacing oil in addition to the
presence of connate water that has the same salinity as the injected low salinity water. Since the connate water salinities are the same as the
injected water salinities for both floods, there is no connate water bank.
the original Hearn’s method are shown in Figure 7.10 and compared with the 2D layered model in addition
to the 1D homogenised model using the rock relative permeability curves. It is clear from the figure that
upscaling the pseudo relative permeability using Hearn’s method without any modifications does not
improve the results obtained from the 1D model over these obtained just using the rock curves.
200
a) b)
1 1
0.9 0.9
0.8 0.8
0.7 0.7
Water cut (%)

0.6 0.6
Oil recovery
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
PV PV
1D upscaling (original Hearn) 1D upscaling (original Hearn)
c)
270
260
Average pressure (bar0
250
240
230
220
210
200
190
180
170
0 0.5 1 1.5 2
PV
2D kv=0 (constant P) 1D with original rel perm
1D upscaling (original Hearn)
Oil recovery, water cut and average reservoir pressure for low salinity waterflooding in a two layer model. The cases shown are: the 2D
heterogeneous case, the 1D homogeneous case with the pseudo relative permeabilities generated using the original Hearn’s method and
the 1D homogeneous case using original relative permeability curves. It is clear from the figure that upscaling the pseudo relative
permeabilities using the original Hearn’s method in the 1D homogeneous case does not improve the oil recovery predictions obtained
from the 1D model over these obtained just using the rock curves.
201
B- Upscaling Low Salinity Waterflooding Using A Modified Hearn’s Method
I. Two Layer Model
The resulting pseudo relative permeability for the high and low salinity water curves are plotted in Figure
7.11 and compared with the rock relative permeability.
0.9
0.8
KrwH
krwH
0.7
Relative Permeability
KroH
k
roH
0.6 KrwL
k
rwL
KroL
0.5
kroL krwH
pseudo
0.4 pseudo k
pseudo kroHrwH
pseudo k
0.3 pseudo krwLroH
pseudo k
pseudo kroLrwL
0.2
pseudo kroL
0.1
0
0 0.2 0.4
Sw 0.6 0.8 1
Figure 7.11: Pseudo relative permeability curves generated for both low and high salinity water in a two layer model (by modifying Hearn’s
method to include the connate water bank) compared with the rock relative permeability curves.
to the 1D homogenised model using the rock relative permeability curves. The oil recovery obtained by 1D
homogeneous with pseudo relative permeabilities generated by the modified Hearn’s method (dashed green
line) are slightly improved over that obtained by the 1D homogeneous model with original relative
permeability curves. However, these predictions are much better than those obtained by the original Hearn’s
method (shown in Figure 7.9). The largest difference between the oil recovery obtained by the modified
Hearn’s method and that obtained by the 2D heterogeneous model (blue line) starts at the time when the
leading shock front of the lower permeability layer breaks through until the trailing shock front breaks
through (0.7-1.5 PV). This suggests that further modifications are needed to capture the connate water bank
effect especially for the low permeability layer.
202
a) b)
0.8 1
0.9
0.7
0.8
0.6 0.7
Oil recovery
Water cut (%)

0.5 0.6
0.4 0.5
0.4
0.3
0.3
0.2
0.2
0.1 0.1
0 0
0 0.5 1 1.5 2 0 0.5 PV 1 1.5 2
PV
1D Upscaling (modified Hearn) 1D Upscaling (modified Hearn)
c)
270
Average pressure (bar)
250
230
210
190
170
0 0.5 1 1.5 2
PV
1D Upscaling (modified Hearn)
Oil recovery, water cut and average reservoir pressure for low salinity waterflooding in a two layer model. The cases shown are: the 2D
heterogeneous case, the 1D homogeneous case with the pseudo relative permeabilities generated using the modified Hearn’s method and
the 1D homogeneous case using the original relative permeability curves. The oil recovery obtained by 1D homogeneous with pseudo
relative permeabilities generated by the modified Hearn’s method (dashed green line) are slightly improved over that obtained by the 1D
homogeneous model with original relative permeability curves. However, these predictions are much better than those obtained by the
original Hearn’s method (shown in Figure 7.9).
203
II. Three Layer Model
0.9
0.8
0.7
KrwH
krwH
0.6 KroH
kroH
KrwL
0.5 krwL
KroL
kroL
0.4 pseudo krwH
pseudo k rwH
pseudo kroH
0.3 pseudo k roH
pseudo krwL
pseudo k rwL
0.2 pseudo kroL
pseudo kroL
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Sw
Figure 7.13: Pseudo relative permeability curves generated for both low and high salinity water in a three layer model (by modifying Hearn’s method
to include the connate water bank) compared with the rock relative permeability curves.
line) are much improved over that obtained by the 1D homogeneous model with original relative
permeability curves. The largest difference between the oil recovery obtained by the modified Hearn’s
method and that obtained by the 2D heterogeneous model (blue line) starts at the time when the leading
shock front of the intermediate permeability layer breaks through until the trailing shock front breaks
through (0.4-1 PV). This difference is much larger compared with the two layer model. This suggests that
further modifications are needed to capture the connate water bank effect especially for the intermediate
permeability layer.
204
a) b)
0.8 1
0.9
0.7
0.8
0.6
0.7
Water cut (%)

0.5
Oil recovery
0.6
0.4 0.5
0.3 0.4
0.3
0.2
0.2
0.1 0.1
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
PV PV
c)
270
250
230
210
190
170
0 0.5 1 1.5 2
PV
Oil recovery, water cut and average reservoir pressure for low salinity waterflooding in a three layer model. The cases shown are: the 2D
the 1D homogeneous case using the original relative permeability curves. The oil recovery obtained by 1D homogeneous model with
pseudo relative permeabilities generated by the modified Hearn’s method (dashed green line) are much improved over that obtained by
the 1D homogeneous model with original relative permeability curves. The difference between the oil recovery obtained by the (1D
Upscaling (modified Hearn) and that obtained by the 2D heterogeneous case is much larger compared with the two layer model. This
suggests that further modifications are needed to capture the connate water bank effect especially for the intermediate permeability layer.
205
III. Five Layer Model
0.9
0.8
krwH
KrwH
0.7
kroH
KroH
0.6
krwL
KrwL
0.5
kKroL
roL
0.4 pseudo krwH

pseudo KrwH
0.3 pseudo kroH

pseudo kroH
pseudo krwL
pseudo krwL
0.2
kroL
pseudokroL
pseudo
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Sw
Figure 7.15: Pseudo relative permeability curves generated for both low and high salinity water in a five layer model (by modifying Hearn’s method
to include the connate water bank) compared with the rock relative permeability curves.
line) are much improved over that obtained by the 1D homogeneous model with original relative
permeability curves. However, they are still not in a good match with the 2D heterogeneous case. The water
breakthrough of all layers are delayed compared with the 2D heterogeneous case. This suggests that while
this method predict better results than the 1D homogeneous model with original relative permeability curves,
it requires further modifications especially for multi layered reservoirs.
206
a) b)
0.8 1
0.7
0.6 0.8
0.5
Oil recovery
water cut (%)

0.6
0.4
0.3 0.4
0.2
0.2
0.1
0
0
0 0.5 1 1.5 2
0 0.5
PV PV 1 1.5 2
c)
270
250
230
210
190
170
0 0.5 PV 1 1.5 2

Oil recovery, water cut and average reservoir pressure for low salinity waterflooding in a five layer model. The cases shown are: the 2D
the 1D homogeneous case using the original relative permeability curves. The oil recovery obtained by 1D homogeneous with pseudo
relative permeabilities generated by the modified Hearn’s method (dashed green line) are much improved over that obtained by the 1D
homogeneous model with original relative permeability curves. However, they are still not in a good match with the 2D heterogeneous
case. The water breakthrough of all layers are delayed compared with the 2D heterogeneous case. This suggests that while this method
predict better results than the 1D homogeneous model with original relative permeability curves, it requires further modifications
especially for multi layered reservoirs.
Discussion
It was shown in this chapter that when reducing a 2D layered model to 1D homogeneous model using
pseudo relative permeabilities by modifying Hearn’s method, the simulation predictions are much improved
compared with those obtained by the original Hearn’s method for both high salinity and low salinity
waterflooding. However, when the number of layers in the reservoir is large (3 to 5 layers), the difference in
the oil recovery predictions by the 2D layered model and the 1D model with pseudo relative permeabilities is
large especially during the water breakthrough of the intermediate permeability layers. This due to the fact
that low salinity waterflooding is not a piston like displacement as assumed by Hearn’s method. However, to
date, there are no methods that have been developed to upscale low salinity waterflooding. Therefore, this
method is considered to be one of the first developed methods to upscale low salinity waterflooding.
Alternatively, upscaling low salinity waterflooding can be improved by a different method. This method is
based on the analytical Buckley Leverett solution of low salinity waterflooding. First, the fractional flow is
207
calculated at the breakthrough of the two shock fronts from each layer. Then, the mean water saturation in
the 2D layered model and the total mobility of each layer is calculated at these times. Accordingly, the
average total mobility is calculated and the pseudo krw and kro are calculated by multiplying the average total
mobility by the fractional flow for these times. This proposed method was originally proposed by Stone
(1991) but requires modifications to account for the low salinity water effect. Due to the limitations in the
time frame of this PhD research, this method was not tested and is recommended for further investigations.
Moroever, for reservoirs with complex heterogeneities upscaling from a fine scale model to a coarser model
can be achieved using streamline based approaches (Ates et al. 2003). First, the velocity field and pressure
distrbutions are calculated in the fine scale model. Then, streamlines are constrcted across the fine scale
model. Using the concept of time of flight, the water saturations are calculated for each streamline. Unlinke
the conventional simulations that solve the water saturations in a 2D or 3D model, in streamline based
simulations the water saturations are solved in 1D model based on the time of flight coordinates. At the end,
the find scale model is ranked into a smaller vertical model. The proposed method in this chapter can be
applied into a smaller vertical model.
Limitations of the Analytical Method
While the analytical method has generated pseudo relative permeability curves that improve the prediction of
oil recovery and water cut when reducing a 2D layered model to a 1D homogeneous model, there are several
limitations to this method:
• It only reduces the dimensionality of the model from a 2D heterogeneous case to a 1D homogeneous
case. There is no further reduction in grid resolution in the horizontal direction. A further study on
pseudoization for the compensation of numerical dispersion is essential.
• The method assumes a Buckley-Leverett like displacement (mobility ratio close to 1). This
assumption is valid for most fields as they contain medium oil viscosities (close to water viscosity).
However, for viscous oil reservoirs this assumption is not valid and this method will not be
applicable when there is viscous be fingering or viscous cross flow.
• It can only be used in layered models. More complex heterogeneous reservoirs require further
upscaling methods (such as pressure solvers or renormalization methods) to upscale the effective
permeabilities and determine the pseudo relative permeabilities for each layer.
7.7 Summary
In this chapter, an analytical method was proposed to upscale low salinity waterflooding in layered models.
The method involves modifying the existing Hearn’s method to include the effect of the connate water bank.
The inclusion of the connate water bank requires the identification of different regions based on the salinity
of each layer. From these regions, the average water saturation and salinities are estimated.
This method was chosen from other two phase upscaling methods as it does not require more than one rock
type when generating the pseudo relative permeabilities. The pseudo relative permeabilities are estimated by
averaging the values of krw and kro for the individual layers similar to Hearn’s method.
208
This analytical method showed improved results compared with those obtained by the original Hearn’s
method. However, it still requires further improvements especially at the breakthrough of the two shock
fronts from the intermediate permeability layers (in case of 3 to 5 layer reservoirs). This due to the fact that
low salinity waterflooding is not a piston like displacement as assumed by Hearn’s method. However, to
date, there are no methods that have been developed to upscale low salinity waterflooding. Therefore, this
method is considered to be one of the first developed methods to upscale low salinity waterflooding. The
proposed upscaling method can be coupled with the use of streamline based approaches especially for
reservoir with complex heterogeities. Upscaling low salinity waterflooding could be further improved, by
modifying Stone’s (1991) method, by generating the pseudo relative permeabilities from the fractional flow,
average total mobility and average water saturation at the breakthrough of the two shock fronts from each
layer. However due to limitations in the time frame of this PhD research, this method was not tested and it
requires further investigation.
There are three main limitations in using the analytical method proposed in this chapter. First: it only reduces
the dimensionality of the layered model from 2D to 1D. It does not coarsen the model horizontally so it does
not minimize numerical dispersion that arises when upscaling in the horizontal direction. Second, this
method does not apply in non-Buckley-Leverett like displacements or viscous cross flow (e.g. heavy oil
reservoirs). Finally, it cannot be used alone in reservoirs with more complex heterogeneity. Before applying
this method, further upscaling methods (such as pressure solvers or renormalization methods) are required to
upscale the effective permeabilities for each layer.
209
CHAPTER 8: CONCLUSIONS AND RECOMMENDATIONS
8.1 Main Findings
In this study, the capillary pressure was neglected. In addition, only one rock type was used for both the
analytical solution of low salinity waterflooding and for the simulation study.
The main objective of this thesis was to investigate how the performance of low salinity waterflooding
(LSW) is affected by reservoir heterogeneity and physical diffusion/dispersion using detailed simulation
studies at the laboratory and reservoir scales. The investigation used various heterogeneity models ranging
from simple layer to areal and 3D models with more complex heterogeneities. Before conducting this
simulation study, several factors that affect the modelling of LSW were investigated: the choice of the
modelling method, the effect of numerical dispersion, the smallest number of grid blocks that capture the
physical longitudinal and transverse diffusion and the choice of the salinity thresholds.
LSW was modelled using two different methods: a simple salinity threshold model (ECLIPSE and CMG
STARS) and a threshold model using results from a more complex ion exchange model (GEM). Both
methods model LSW by modifying the rock’s relative permeability curves. The first uses the salinity of the
brine whilst the second method uses the divalent ion content derived from a model of the exchange between
the clay in the reservoir and the brine’s divalent ions. The flow behaviour predicted by the three different
simulators was validated by comparison against the analytical solution of LSW. They reproduced similar
results to the analytical solution.
It was found that 1000 grid blocks reproduces the analytical solution within 1 %. The error difference
between the analytical solution and the simulated model using any number of grid block was found to vary
linearly with a first order of convergence. In addition, the smallest number of horizontal grid blocks that
captures most of the physical diffusion (for Peclet number (Pe) < 70) and minimizes the numerical diffusion
was found to be 200 while it was found that more than 2 grid blocks were needed vertically to capture the
physical transverse diffusion. This minimum number of horizontal grid blocks depends on Pe. To capture the
physical longitudinal diffusion, Pe has to be below a critical Pe that can be used to determine when physical
diffusion dominates over numerical diffusion. The critical Pe depends non-linearly on the number of grid
blocks. There is also a plateau Pe which describes the level of numerical diffusion in a simulation and also
depends non-linearly on the number of grid blocks. Even when physical diffusion dominates over numerical
diffusion there is a contribution to the diffusion from higher order truncation errors in combination with the
variable time-step size chosen by the simulator. All these considerations suggest that simply using a chosen
number of coarse grid blocks to represent the effects of physical diffusion and dispersion on mixing in LSW
is only a very crude approximation.
It was found that the simulator predicts the results closest to the analytical solution when the salinity
thresholds are close together and midway between the injected water salinity and the salinity of the connate
water. The maximum effect of the salinity thresholds on the oil recovery (5 % difference) are during the
production of the connate water bank. The effect on the saturation fronts are as follows: earlier water
210
breakthrough (leading shock front) and delayed arrival of the trailing shock front with higher water cut when
the salinity thresholds are narrow and closer to the injected water salinity.
For the ion exchange model, the choice of the interpolation parameter which is used to change from the high
salinity (HS) water relative permeability curve to the low salinity (LS) water relative permeability curve and
the cation exchange capacity (CEC) and their impacts on LSW were investigated. It was found that the
simulation predictions are closer to the analytical Buckley Leverett solution when setting the interpolation
parameter to be the equivalent fraction of Na-X compared with the case when setting it to be the Ca++
concentration. In addition, when the CEC (i.e. the amount of ions that are absorbed by the clay) was about 5
equivalent/m3, the predictions were closer to the analytical Buckley Leverett solution. Moreover, it was
found that the values to change from HS water relative permeability curve to the LS water relative
permeability curve had a minor effect on modelling LSW compared with the choice of the interpolation
parameter and the value of CEC.
The impact of reservoir heterogeneity on LSW was first evaluated without modelling physical
diffusion/dispersion using 2D layered models and areal models with complex heterogeneity taken from SPE
10 model 2 (Christie and Blunt 2001). LS water was first injected continuously then as a slug. In the layered
models, it was found that heterogeneity has a larger effect on LSW compared to high salinity waterflooding
(HSW) due to the mixing via viscous crossflow while it has a similar effect on LSW to HSW in the areal
models. Assuming a total water injection of 1 PV, the best slug size when injecting LS water into layered
models was about 0.4-0.8 PV increasing to 0.6-0.8 PV when it is injected into areal models. When LS water
is injected post immediate injection of conventional water into both layered models and areal layers with
complex heterogeneity, it is recommended to inject LS water post the breakthrough of HSW. Delaying the
start of LS water injection leads to lower incremental oil recovery due to mixing via viscous cross flow with
the conventional water.
Next, the investigation included physical diffusion/dispersion which were characterized by the use of a
dimensionless number (NTD) (Lake and Hirasaki 1981). Before conducting this investigation, a single-phase
flow simulation study was conducted to characterize the flow regime based on NTD. It was found that the
flow is dominated by reservoir heterogeneity when NTD < 0.2 while when NTD >1 the flow is dominated by
diffusion. The use of NTD was found to be very robust for characterizing the effect of diffusion on flow
performance.
For two-phase flow, the impact of physical diffusion/dispersion and reservoir heterogeneity on LSW was
investigated using several simulation models: laboratory scale models that included layered models and
shaly sand, a 3D reservoir scale model taken from SPE 10 model 2 and a reservoir scale model with a
transition zone connected to aquifer. The investigations looked into the injection of LS water immediately at
the initiation of recovery and then post HSW.
For the case of immediate injection of LS water, LS water was first injected continuously then as a slug. At
the laboratory scale, the effect of salt diffusion on LSW was found to be more significant: when the
211
thickness of the high permeability layer decreases, when the sand is adjacent to a thick shale and during the
production of the connate water bank in the layered models especially when setting the upper salinity
threshold to be close to the salinity of the injected water. It was found that the best slug size of injecting LS
water into laboratory scale layered models is around 0.4-0.6 PV for immediate LS water injection increasing
to 0.6-1 PV when it is injected into reservoir scale models with more complex heterogeneities. At the
reservoir scale, it was found that the effect of dispersion is minor compared to reservoir heterogeneity. The
effect of an aquifer on LSW was found to be significant during the production of the connate water bank.
The best slug size of injecting LS water into an aquifer was found to be 2-3 PV of oil zone assuming a total
water injection of 5 PV.
For the case of injecting LS water post HSW, assuming a total water injection of 2 PV, the best slug size of
injecting LS water into laboratory scale models was found to be 0.5-1 PV compared with 0.4-0.6 PV for
immediate LS water injection. At the reservoir scale, it was found that the best slug size of injecting LS
water post HSW is around 1.3 PV compared with 0.6-1 PV when LS water is injected immediately. This
would be expected as LS water mixes with the larger amount of HS water present in the reservoir thus more
LS water is needed to compensate for the decrease in oil recovery due to dispersion. The best slug size of
injecting LS water into an aquifer was found to be 4-5 PV of oil zone assuming a total water injection of 10
PV.
Finally, an analytical method was proposed to upscale LSW from a 2D layered model to a 1D homogeneous
model. This method is based on generating pseudo relative permeabilities by modifying Hearn’s method to
include the connate water bank. This analytical method gives improved predictions over those obtained by
the original Hearn’s method when compared using a 2D layered model. However, it requires further
improvement especially at the breakthrough of the two shock fronts from the intermediate permeability
layers (in case of 3 to 5 layer reservoirs). This due to the fact that LSW is not a piston like displacement as
assumed by Hearn’s method. However, to date, there are no methods that have been developed to upscale
LSW. Therefore, this method is considered to be one of the first developed methods to upscale LSW.
8.2 Recommendations on How to Design and Model a Low Salinity Waterflood
This section provides recommendations to the reservoir engineer on how to design and model a LS
waterflood project in a field scale based on the findings of this thesis. Before planning a LS waterflood, it is
essential to acquire a prior knowledge of: the reservoir heterogeneity, the salinity thresholds, the presence of
an aquifer and the presence of waterfilled shales. These factors contribute to the success of LSW.
The incremental oil recovery by LSW depends on the reservoir heterogeneity. Reservoirs with large
permeability variations would only gain an incremental oil recovery by LSW from high permeability areas.
Even if the dispersivity is high, the incremental oil recovery from low permeability layers would be
marginal. Generally, dispersion has a minor effect on the oil recovery at the reservoir scale compared to the
effect of reservoir heterogeneity. Less heterogeneous reservoirs with smaller permeability variations would
benefit more from LSW compared to more heterogeneous reservoirs.
212
The prior knowledge of the salinity thresholds is essential when modelling LSW as they could result in
incorrect predictions of oil recovery (up to 5 % difference). These thresholds should be estimated from
laboratory experiments prior to conducting a simulation study.
The presence of an aquifer is likely to reduce recovery if the injectors are planned to be completed close to
the aquifer. This is due to the fact that the injected LS water will mix with aquifer water. As a result, LS
water would be less effective. Therefore, it is recommended to complete the injectors away from the aquifers
if possible.
If the reservoir contains waterfilled shales within sands then LSW could be affected especially if they are
thick. In this case, it is unlikely to improve recovery.
It is recommended to implement LSW immediately at the initiation of the recovery from a new field.
However, if there is lack of availability LS water, then it is recommended to inject 0.6-1 PV slug of LS water
assuming a total water injection of 1 PV.
If LSW is planned to be implemented as a displacement post HSW, it is recommended to start injecting it as
soon as possible. Delaying its start would lead to a lower incremental oil recovery due to the mixing between
LS water and HS water. If LS water is planned to be injected as slugs post HSW, then a slug size of 1.3 PV
is recommended assuming a total water injection of 2 PV. This large slug size of LS water might not be
operationally practical. This is due to the fact that either the produced water, which requires desalination to
be reinjected, contains oil droplets that might block the desalination equipment thus reducing its efficiency
or the produced water would be disposed.
Ideally, at least 200 grid blocks horizontally and 2 grid blocks vertically (per each layer) should be used to
capture most of the physical diffusion (Pe < 70) and minimize the numerical diffusion. Using coarser grid
blocks would lead to incorrect predictions of oil recovery by LSW as numerical diffusion dominates. It is
recognised that it is usually impractical to use this level of grid refinement in field studies. In the absence of
a suitable pseudoization method to compensate for numerical diffusion the engineer should remember that
their coarse model will probably predict early breakthrough of the connate water bank and reduced water cut
around the time of the LS water arrival, compared with reality.
With regards to the method of modelling LSW, the salinity thresholds method is recommended since: it was
validated by experimental data (Jerauld et al. 2008), its simulation running time is significantly faster than
the ion exchange model and it showed excellent comparison with the analytical Buckley Leverett solution of
LSW in this thesis.
The dimensionless number (NTD) should be used to characterize the impact of transverse physical
diffusion/dispersion on modelling LSW at different heterogeneities. It enables the reservoir engineer to use
the results of this thesis on any scale. The best slug sizes of injecting LS water into different models was
found both analytically and by using the simulation models. The analytical solution is a quick method to
estimate the best slug size of injecting LS water at very low kv/kh where the effect of viscous crossflow is
213
negligible. However, it cannot be used at high kv/kh as the effect of viscous crossflow is significant and the
predicted best slug sizes can then be estimated by the use of a simulation model.
8.3 Recommendations for Future Work
In this thesis, only one rock type was used for all simulation models in addition to the analytical solution of
low salinity waterflooding. However, for reservoirs with multiple rock types, the outcomes of this study
might be different. This is especially the case if there are two rock types with different connate water
saturations. An example is a case where there is a low permeability layer (with a high connate water
saturation) that is adjacent to a high permeability layer that has a much lower connate water saturation. In
this case, a larger connate water bank forms in the low permeability layer compared with the high
permeability layer. This will change the relative timing of water breakthrough in the two layers, the rate of
transverse mixing and the water cut changes over time. This case requires further analysis and is
recommended as a topic for future research.
In addition, the capillary pressure effect was ignored. However, in fractured carbonate reservoirs or
reservoirs with very large permeability contrasts, capillary pressure is important where oil is trapped in low
permeability matrix that is surrounded by water (Blunt 2017). This trapped oil is recovered by spontaneous
imbibition of water into the low permeability matrix. The imbibition rate depends on the rock’s wettability.
It is larger in strongly water wet rocks compared to mixed wet and oil wet rocks. It is widely accepted in the
oil industry that low salinity waterflooding recovers more oil by altering the reservoir rock’s wettability
making it more water wet (Jerauld et al. 2008, Nasralla et al. 2013, Dang et al. 2015, Kuznetsov et al. 2015,
Law et al. 2015). Therefore, the imbibition rate will be increased due to low salinity water injection as a
result of wettability alteration. The possible enhancement of spontaneous imbibition as a result of low
salinity water injection in fractured carbonate reservoirs (such as Middle Eastern fields with large oil
reserves) or reservoirs with very large permeability contrast is recommended as a topic for future research.
In this thesis, transverse dispersion was varied to investigate its effect on LSW. However, a further study is
recommended to investigate if transverse dispersion increases with length scale. It is suggested to use
heterogeneous models with different length scales. Then, the transverse dispersion is back-calculated from
the effluent profiles.
The analytical method presented in this thesis for upscaling LSW cannot be used in reservoirs with more
complex heterogeneity. Before applying this method, further upscaling methods (such as pressure solvers or
renormalization methods) are required to upscale the effective permeabilities for each layer. Upscaling LSW
could be further improved by generating the pseudo relative permeabilities from the fractional flow, average
total mobility and average water saturation at the breakthrough of the two shock fronts from each layer.
However due to limitations in the time frame of this PhD research, this method was not tested and it requires
further investigation. Moreover, the proposed upscaling method in this thesis can be coupled with the use of
streamline based approaches especially for reservoir with complex heterogeities.
214
It is recommended to conduct laboratory experiments with various heterogeneities (such as shaly sand) that
include injecting LS water to compare with the findings of this thesis.
Most of the investigation in this thesis was carried out using the salinity threshold model which linearly
interpolates between the high and low salinity water relative permeability curves for intermediate salinities.
Further laboratory experiments are recommend to compare this proposed method to actual data.
In addition, laboratory experiments are needed to measure LSW relative permeability curves. The current
practice to measure LS water relative permeability curves is as follows: initially a core is filled with oil and
connate HS water, then LS water is introduced until residual oil saturation is reached. The measured effluent
oil and water are history matched with a simulation model. This simulation model assumes a pore scale
mechanism to model LSW which is still a subject of ongoing research (Pouryousefy et al. 2016,
Akhmetgareev et al. 2015, Callegaro et al. 2013).
215
APPENDIX A
This is a sample CMG STARS input file to model a 2D laboratory scale model where low salinity water is
injected immediately. The molecular diffusion is modelled using the keywords: “DIFFI_WAT 'NACL'
CON” in the x-direction, “DIFFJ_WAT 'NACL' CON” in the y-direction” and DIFFK_WAT 'NACL' CON”
in the z-direction.
RESULTS SIMULATOR STARS 201210

INUNIT LAB
WSRF WELL 1
WSRF GRID TIME
WSRF SECTOR TIME
OUTSRF GRID SO VELOCSC W
OUTSRF WELL LAYER NONE
OUTSRF SPECIAL MOLEFRAC 'PRODN' 'NACL'
WPRN GRID 0
OUTPRN GRID NONE
OUTPRN RES NONE
**$ Distance units: ft
RESULTS XOFFSET 0.0000
RESULTS YOFFSET 0.0000
RESULTS ROTATION 0.0000 **$ (DEGREES)
RESULTS AXES-DIRECTIONS 1.0 -1.0 1.0
** ============== GRID AND RESERVOIR DEFINITION =================
*GRID *CART 200 1 10 ** One-dimensional grid
*DI *CON 0.149 **8cm
*DJ *CON 3.17
*DK *CON 0.317
**$ Property: NULL Blocks Max: 1 Min: 1
**$ 0 = null block, 1 = active block
NULL CON 1
*POR *CON 0.2
PERMI KVAR
7*1 3*100
PERMJ EQUALSI
PERMK EQUALSI
**$ Property: Pinchout Array Max: 1 Min: 1
216
**$ 0 = pinched block, 1 = active block
PINCHOUTARRAY CON 1
*END-GRID
** ============== FLUID DEFINITIONS ======================
** Two aqueous and two noncondensible gas components
**$ Model and number of components
MODEL 3 3 3 2
COMPNAME 'water' 'NACL' 'C18H38'
** ----- -------- -------- --------
CMM
18 58.4 254.502
PCRIT
22060 26000 22000
TCRIT
373 3600 373
KVTABLIM 14.7 20000 10 100
** Reference conditions
PRSR 100
PSURF 101
TSURF 16.85
MASSDEN
0.001 0.001121 0.0008
VISCTABLE
** temp
80 0.5 0.5 0.5
*ROCKFLUID
RPT 1 WATWET
** ============== ROCK-FLUID PROPERTIES ======================
INTCOMP 'NACL' WATER
*INTLIN
**SET 1 HIGH SALINITY
KRINTRP 1
DTRAPW 0.007
** Water-oil relative permeabilities
**Sw Krw Kro
217
**$ Sw krw krow
**$ Sw krw krow
SWT
0.24 0 1
0.27 0.013320597 0.517935384
0.3 0.033488337 0.252889649
0.33 0.057424201 0.115057573
0.36 0.084190576 0.0480306
0.39 0.113280145 0.018011667
0.42 0.144365977 0.005888046
0.45 0.177216472 0.001604437
0.48 0.21165736 0.000339434
0.51 0.247551881 0
0.54 0.284789316 0
0.57 0.323277861 0
0.6 0.362939929 0
0.62 0.39 0 ** table 1 high salinity
**$ Sl krg krog
**$ Sl krg krog
SLT
0.24 1 0
1 0 1
**SET 2 LOW SALINITY
KRINTRP 2
DTRAPW 0.001
** Water-oil relative permeabilities
**Sw Krw Kro
**$ Sw krw krow
**$ Sw krw krow
SWT
0.25 0 1
0.28 0.007208404 0.673176445
0.31 0.018122122 0.440672544
0.34 0.031074949 0.279327974
0.37 0.045559499 0.170537982
0.4 0.061301239 0.099612244
0.43 0.078123252 0.055182432
218
0.46 0.095900206 0.028658466
0.49 0.114537799 0.013733457
0.52 0.13396202 0.005937352
0.55 0.154112955 0.00223927
0.58 0.174940926 0.000698524
0.61 0.196403945 0.00016435
0.64 0.218465975 0
0.67 0.241095684 0
0.7 0.264265533 0
0.72 0.28 0 ** table 2 low salinity
**$ Sl krg krog
**$ Sl krg krog
SLT
0.25 1 0
1 0 1
DIFFI_WAT 'NACL' CON 7.014e-2
DIFFJ_WAT 'NACL' CON 7.014e-2
DIFFK_WAT 'NACL' CON 7.014e-2
*TORTU *NOPORSAT
*INITIAL
VERTICAL OFF
INITREGION 1
REFPRES 102
REFDEPTH 0
DWOC 0.01
PRES CON 102
*TEMP *CON 80
SW CON 0.24
SO CON 0.76
MFRAC_WAT 'water' CON 0.97
MFRAC_WAT 'NACL' CON 0.03
*NUMERICAL
** ============== NUMERICAL CONTROL ======================
*TFORM ZT
*ISOTHERMAL
*DTMAX 2.0
*SORDER *RCMRB
219
*RANGECHECK *OFF
NORM PRESS 10 ZO 0.05 ZNCG 0.4 ZAQ 0.4
CONVERGE PRESS 0.01 ZO 0.001 ZNCG 0.001 ZAQ 1e-005
*RANGECHECK *ON
*RUN
** ============== RECURRENT DATA ======================
*TIME 0
*DTWELL .05
**
** *WELL 1 'INJTR' ** Well list
**$
WELL 'INJTR'
** *WELL 2 'PRODN'
**$
WELL 'PRODN'
PRODUCER 'PRODN'
OPERATE MAX STL 0.04159 CONT REPEAT
** LINEAR PRESSURE DROP
** I J K
**$ rad geofac wfrac skin
**$ UBA ff Status Connection
GEOMETRY K 0.01 1.0 1.0 0.0
PERF GEO 'PRODN'
1 1 1 1.0 OPEN FLOW-TO 'SURFACE'
1 1 2 1.0 OPEN FLOW-TO 1
**
INJECTOR UNWEIGHT 'INJTR'
220
INCOMP WATER 0.999 0.001 0.0
OPERATE MAX STW 0.04159 CONT REPEAT
** I J K WI
** WI?
**$ UBA wi Status Connection
GEOMETRY K 0.01 1.0 1.0 0.0
PERF GEO 'INJTR'
200 1 1 1.0 OPEN FLOW-FROM 'SURFACE'
200 1 2 1.0 OPEN FLOW-FROM 1
TIME 1
TIME 3
TIME 6
TIME 9
TIME 12
TIME 15
TIME 18
STOP
221
REFERENCES
1. Agbalaka, C.C., Dandekar, A.Y., Patil, S.L., Khataniar, S., Hemsath, J.R. 2009. Coreflooding Studies to
Evaluate the Impact of Salinity and Wettability on Oil Recovery Efficiency. Transport in Porous Media
76 (1): 77-94.
2. Ahmed, G., Castanier, L.M., Brigham, W.E. 1988. An Experimental Study of Waterflooding From a
Two-Dimensional Layered Sand Model. SPE Reservoir Engineering 3 (1): 45-54.
3. Akhmetgareev, V., Khisamov, R. 2015. 40 Years of Low-Salinity Waterflooding in Pervomaiskoye

Field, Russia: Incremental Oil. Paper SPE 174182 proceedings of the SPE European Formation Damage
Conference and Exhibition, Budapest, Hungary, 3-5 June.
4. Alkindi, A., Al-Wahaibi, Y., Bijeljic, B., Muggeridge, A. 2011. Investigation of longitudinal and
transverse dispersion in stable displacements with a high viscosity and density contrast between the
fluids. Journal of Contaminant Hydrology 120-121:170-183.
5. Amott, E. 1959. Observations Relating to the Wettability of Porous Rock. Petroleum Transactions,
AIME 216: 156-162.
6. Anderson, W.G. 1986. Wettability Literature Survey- Part 1: Rock/Oil/Brine Interactions and the
Effects of Core Handling on Wettability. Journal of Petroleum Technology 38 (10): 1125-1144.
7. Ashraf, A., Hadia, N.J., Torsaeter, O., Tweheyo, M.T. 2010. Laboratory Investigation of Low Salinity
Waterflooding as Secondary Recovery Process: Effect of Wettability. Paper SPE 129012 proceedings of
the SPE Oil and Gas India Conference and Exhibition, Mumbai, India, 20-22 January.
8. Ates, H., Bahar, A., El-Abd, S., Charfeddine, M., Kelkar, M., Datta-Gupta, A. 2003. Ranking and
Upscaling of Geostatistical Reservoir Models Using Streamline Simulation: A Field Case Study. Paper
SPE 81497 proceedings of the SPE 13th Middle East Oil Show & Conference, Bahrain, 9-12 June.
9. Austad, T., RazaeiDoust, A., Puntervold, T. 2010. Chemical Mechanism of Low Salinity Water
Flooding in Sandstone Reservoirs. Paper SPE 129767 proceedings of the 2010 SPE Improved Oil
Recovery, Tulsa, Oklahoma, USA, 24-28 April.
10. Baker, R. 1998. Reservoir Management for Waterfloods-Part II. The Journal of Canadian Petroleum
Technology 37 (1): 12-17.
11. Barker, J.W., Dupouy, P. 1999. An analysis of dynamic pseudo-relative permeability methods for oil-
water flows. Petroleum Geoscience 5: 385-394.
12. Bedrikovetsky, P., Zeinijahromi, A., Badalyan, A., Ahmetgareev, V., Khisamov, R. 2015. Fines-
Migration-Assisted Low-Salinity Waterflooding: Field Case Analysis. Paper SPE 176721 proceedings
of the SPE Russian Petroleum Technology Conference, Moscow, Russia, 26-28 October.
13. Bernard, G. G. 1967. Effect of Floodwater Salinity on Recovery of Oil from Cores Containing Clays.
Paper SPE 1725 proceedings of the 38th Annual California Regional Meeting of the Society of
Petroleum Engineers of AIME, Los Angeles, California, 26-27 October.
14. Blackwell, R.J. 1962. Laboratory Studies of Microscopic Dispersion Phenomena. Society of Petroleum
Engineers Journal 2: 1–8.
222
15. Blunt, M.J. 2017. MULTIPHASE FLOW IN PERMEABLE MEDIA A PORE-SCALE PERSPECTIVE
pages 406-407, 427. Cambridge University Press, Cambridge, United Kingdom.
16. Bortolotti, V., Macini, P., Srisuriyachai, F. 2010. Wettability Index of Carbonatic Reservoirs and EOR:
Laboratory Study to Optimize Alkali and Surfactant Flooding. Paper SPE 131043 proceedings of the
CPS/SPE International Oil & Gas Conference and Exhibition, Beijing, China, 8-10 June.
17. Brigham, W.E. 1974. Mixing Equations in Short Laboratory Cores. Society of Petroleum Engineers
Journal 14 (1): 91-99.
18. Brodie, J., Jerauld, G. 2014. Impact of Salt Diffusion on Low-Salinity Enhanced Oil Recovery. Paper
SPE 169097 proceedings of the SPE Improved Oil Recovery Symposium, Tulsa, USA, 12-16 April.
19. Callegaro, C., Bartosek, M., Masserano, F., Nobili, M., Parracello, V.P., Pizzinelli, C.S., Caschili, A.
2013. Opportunity of Enhanced Oil Recovery Low Salinity Water Injection: From Experimental Work
to Simulation Study up to Field Proposal. Paper 164827 proceedings of the EAGE Annual Conference
& Exhibition incorporating SPE Europec, London, United Kingdom, 10-13 June.
20. Christie, M.A. 1996. Upscaling for Reservoir Simulation. Journal of Petroleum Technology 48 (11):
1004-1010.
21. Christie, M.A. 2001. Flow in porous media – scale up of multiphase flow. Current Opinion in Colloid
& Interface Science 6 (3): 236-241.
22. Christie, M.A., Blunt, M.J. 2001. Tenth SPE Comparative Solution Project: A Comparison of Upscaling
Techniques. SPE Reservoir Evaluation & Engineering 4 (4): 308-317.
23. Cinar, Y., Jessen, K., Berenblyum, R., Juanes, R., Orr, F.M. 2004. An Experimental and Numerical
Investigation of Crossflow Effects in Two-Phase Displacements. Paper SPE 90568 proceedings of the
SPE Annual Technical Conference and Exhibition, Houston, Texas, 26-29 September.
24. Cissokho, M., Boussour, S., Cordier, P., Bertin, H., Hamon, G. 2010. Low Salinity Oil Recovery on
Clayey Sandstone: Experimental Study. Petrophysics 51 (5): 305-313.
25. Coats, K. H., Whitson, C. H., Thomas, L.K. 2009. Modeling Conformance as Dispersion. SPE
Reservoir Engineering and Evaluation 12 (1): 33-47.
26. Coll, C., Muggeridge, A.H., Jing, X.D. 2000. Regional Upscaling: A New Method to Upscale
Waterflooding in Heterogeneous Reservoirs for A Range of Capillary and Gravity Effects. Paper SPE
59337 proceedings of the 2000 SPE/DOE Improved Oil Recovery Symposium, Tulsa, Oklahoma, USA,
3-5 April.
27. Cotterill, S. Low Salinity Effects on Oil Recovery, MSc Thesis, Imperial College London, 2014.
28. Dang, C., Nghiem, L., Nguyen, N., Chen, Z., Nguyen, Q. 2015. Modeling and Optimization of Low
Salinity Waterflood. Paper SPE 173194 proceedings of the SPE Reservoir Simulation Symposium,
Houston, Texas, USA, 23-25 February.
29. Dang, C.T.Q., Nghiem, L.X., Chen, Z., Nguyen, Q.P. 2013. Modeling Low Salinity Waterflooding: Ion
Exchange, Geochemsitry and Wettability Alteration. Paper SPE 166447 proceedings of the SPE Annual
Technical Conference and Exhibition, New Orleans, 30 September-2 October.
223
30. Donaldson, E.C., Thomas, R.D., Lorenz, P.B. 1969. Wettability Determination and Its Effect on
Recovery Efficiency. Society of Petroleum Engineers Journal 9 (1): 13-20.
31. Drummond, C., Israelachvili, J. 2004. Fundamental studies of crude oil- surface water interactions and
its relationship to reservoir wettability. Journal of Petroleum Science and Engineering 45: 61-81.
32. Durlofsky, L.J. 1991. Numerical Calculation of Equivalent Grid Block Permeability Tensors for
Heterogeneous Porous Media. Water Resources Research 27 (5): 699-708.
33. Dykstra, H. and Parsons, R.L. 1950. The Prediction of Oil Recovery by Waterflooding, Secondary
Recovery of Oil in the United States pages 160-174. API, Dallas 2nd edition.
34. Ekrann, S., Aasen, J.O. 2000. Steady-State Upscaling. Transport in Porous Media 41 (3): 245-262.
35. El-Khatib, N. 1985. The Effect of Crossflow on Waterflooding of Stratified Reservoirs. Society of
Petroleum Engineers Journal 25 (2): 291-302.
36. Floris, F.J.T., Bush, M.D., Cuypers, M., Roggero, F., Syversveen, A-R. 2001. Methods for quantifying
the uncertainty of production forecasts: A comparative study. Petroleum Geoscience 7: S87-S96.
37. Goddin, C.S., Craig, F.F., Wilkes, J.O., Tek, M.R. 1966. A Numerical Study of Waterflood
Performance In a Stratified System With Crossflow. Journal of Petroleum Technology 18 (6): 763-771.
38. Green, D.W. and Willhite, G.P. 1998. Enhanced Oil Recovery pages 43-45. SPE Textbook Series Vol.6,
Richardson, Texas.
39. Hadia, N.J., Hansen, T., Tweheyo, M.T., Torsaeter, O. 2012. Influence of Crude Oil Components on
Recovery by High and Low Salinity Waterflooding. Energy & Fuels 26: 4328-4335.
40. Hadia, N., Lehne, H.H., Kumar, K.G., Selboe, K., Stensen. J.A., Torsaeter, O. 2011. Laboratory
Investigation of Low Salinity Waterflooding on Reservoir Rock Samples from the Froy Field. Paper
SPE 141114 proceedings of the SPE Middle East Oil and Gas Show and Conference, Manama, Bahrain,
25-28 September.
41. Haq, S., Reis, J.C. 1993. Predicting Capillary Crossflow in Layered Reservoirs. Paper SPE 26651
proceedings of the 68th Annual Technical Conference and Exhibition of the Society of Petroleum
Engineers, Houston, Texas, 3-6 October.
42. Hassenkam, T., Matthiesen, J., Pedersen, C.S., Dalby, K.N., Stipp, S.L.S., Collins, I.R. 2012.
Observation of the low salinity effect by atomic force adhesion mapping on reservoir sandstones. Paper
SPE 154037 proceedings of the Eighteenth SPE Improved Oil Recovery Symposium, Tulsa, Oklahoma,
USA, 14-18 April.
43. Hassinger, R.C., Von Rosenberg, D.U. 1968. A Mathematical and Experimental Examination of
Transverse Dispersion Coefficients. Society of Petroleum Engineers Journal 8 (2): 195-204.
44. Hearn, C.L. 1971. Simulation of Stratified Waterflooding by Pseudo Relative Permeability Curves.
Journal of Petroleum Technology 23 (7): 805-813.
45. Henson, R., Todd, A., Corbett, P. 2002. Geologically Based Screening Criteria for Improved Oil
Recovery Projects. Paper SPE 75148 proceedings of the SPE/DOE Improved Oil Recovery Symposium,
Oklahoma,13-17 April.
224
46. Hughes, D. Law, S., Pitt, G. 2012. Low Salinity EOR “State of Play” Review”. Aberdeen: Senergy
(GB) Limited.
47. Idrobo, E.A., Choudhary, M.K., Datta-Gupta, A. 2000. Swept Volume Calculations and Ranking of
Geostatistical Reservoir Models Using Streamline Simulation. Paper SPE 62557 proceedings of the
2000 SPE/AAPG Western Regional Meeting, Long Beach, California, 19-23 June.
48. Illic, D. 2012. Experimental study of the Low Salinity effect using glass plate cell. Rijswijk: Shell
Rijswijk B.V.
49. Jackson, M.D., Vinogradov, J., Hamon, G., Chamerois, M. 2016. Evidence, mechanisms and improved
understanding of controlled salinity waterflooding part 1: Sandstones. Fuel 185: 772-793.
50. Jensen, J.L., Currie, I.D. 1990. A New Method for Estimating the Dykstra-Parsons Coefficient To
Characterize Reservoir Heterogeneity. SPE Reservoir Engineering, 5 (3): 369-374.
51. Jensen, J.L., Lake, L.W. 1988. The Influence of Sample Size and Permeability Distribution on
Heterogeneity Measures. SPE Reservoir Engineering 3 (2): 629-637.
52. Jerauld, G.R., Lin, C.Y., Webb, K.J., Seccombe, J.C. 2008. Modeling Low-Salinity Waterflooding. SPE
Reservoir Evaluation & Engineering 11 (6): 1000-1012.
53. Jonoud, S., Jackson, M.D. 2008. Validity of Steady-State Upscaling Techniques. SPE Reservoir
Evaluation and Engineering 11 (2): 405-416.
54. Jones, A., Doyle, J., Jacobsen, T., and Kjonsvik, D. 1995.Which sub-seismic heterogeneities influence
waterflood performance? A case study of a low net-to-gross fluvial reservoir. Geological Society,
London, Special Publications 84: 5-18.
55. Kjonsvik, D., Doyle, J., Jacobsen, T., Jones, A. 1994. The Effects of Sedimentary Heterogeneities on
Production From A Shallow Marine Reservoir- What Really Matters?. Paper SPE 28445 proceedings of
the European Petroleum Conference, London, 25-27 October.
56. Korrani, A.K.N., Jerauld, G.R., Sepehrnoori, K. 2014. Coupled Geochemical-Based Modeling of Low
Salinity Waterflooding. Paper SPE 169115 proceedings of the SPE Improved Oil Recovery
Symposium, Tulsa, Oklahoma, USA, 12-16 April.
57. Korrani, A.K.N, Jerauld, G.R., Sepehrnoori, K. 2016. Mechanistic Modeling of Low-Salinity
Waterflooding Through Coupling Geochemical Package With a Compositional Reservoir Simulator.
SPE Reservoir Evaluation & Engineering 19 (1): 142-162.
58. Koval, E.J. 1963. A Method for Predicting the Performance of Unstable Miscible Displacement in
Heterogeneous Media. Society of Petroleum Engineers Journal 3 (2): 145-154.
59. Kuznetsov, D., Cotterill, S., Giddins, M.A., Blunt, M.J. 2015. Low-Salinity Waterflood Simulation:
Mechanistic and Phenomenological Models. Paper SPE 174615 proceedings of the SPE Enhanced Oil
Recovery Conference, Kuala Lumpur, Malaysia, 11-13 August.
60. Kyte, J.R., Berry, D.W. 1975. New Pseudo Functions To Control Numerical Dispersion. Society of
225
61. Lager, A., Webb, K. J., Black, C. J. J., Singleton, M., Sorbie, K. S. 2006. Low Salinity Oil Recovery:
An Experimental Investigation. Paper SCA2006-36 proceedings of the International Symposium of the
Society of Core Analysts, Trondheim, Norway, 12-16 September.
62. Lager, A., Webb, K.J., Collins, I.R., Richmond, D.M. 2008. LoSal Enhanced Oil Recovery: Evidence of
Enhanced Oil Recovery at the Reservoir Scale. Paper SPE 113976 proceedings of the 2008 SPE/DOE
Improved Oil Recovery Symposium, Tulsa, Oklahoma, USA, 19-23 April.
63. Lake, L.W., Hirasaki, G.J. 1981. Taylor’s Dispersion in Stratified Porous Media. Society of Petroleum
Engineers Journal 21 (4): 459-468.
64. Lake, L.W., Jensen, J.L. 1989. A Review of Heterogeneity Measures Used in Reservoir
Characterization. Paper SPE 20156. General for the Society of Petroleum Engineers.
65. Lantz, R.B. 1971. Quantitative Evaluation of Numerical Diffusion (Truncation Error). Society of
66. Law, S., Sutcliffe, P., Fellows, S. 2014. Secondary Application of Low Salinity Waterflooding to
Forties Sandstone Reservoirs. Paper SPE 170725 proceedings of the SPE Annual Technical Conference
and Exhibition, Amsterdam, 27-29 October.
67. Law, S., McDonald, A., Fellows, S., Reed, J., Sutcliffe, P.G. 2015. Influence of Clay Content and Type
on Oil Recovery Under Low Salinity Waterflooding in North Sea Reservoirs. Paper SPE 175506
proceedings of the SPE Offshore Europe Conference & Exhibition, Aberdeen, Scotland, 8-11
September.
68. Ligthelm, D.J., Gronsveld, J., Hofman, J.P., Brussee, N.J., Marcelis, F., Van Der Linde, H.A. 2009.
Novel Waterflooding Strategy by Manipulation of Injection Brine Composition. Paper SPE 119835
proceedings of the 2009 SPE EUROPEC/EAGE Annual Conference and Exhibition, Amsterdam, 8-11
June.
69. Mahadevan, J., Lake, L.W., Johns, R.T. 2003. Estimation of True Dispersivity in Field-Scale Permeable
Media. SPE Journal 8 (3): 272-279.
70. Mahani, H., Berg, S., Llic, D., Bartels, W.-B., and Joekar-Niasar, V. 2014. Kinetics of Low-Salinity-
Flooding Effect. SPE Journal preprint.
71. Mahani, H., Sorop, T.G., Ligthelm, D., Brooks, A.D., Vledder, P., Mozahem, F., Ali, Y. 2011. Analysis
of Field Responses to Low-Salinity Waterflooding in Secondary and Tertiary Mode in Syria. Paper SPE
142960 proceedings of the SPE EUROPEC/EAGE Annual Conference and Exhibition, Vienna, Austria,
23-26 May.
72. McGuire, P.L., Chatham, J.R., Paskvan, F.K., Sommer, D.M., Carini, F.H. 2005. Low Salinity Oil
Recovery: An Exciting New EOR Opportunity for Alaska’s North Slope. Paper SPE 93903 proceedings
of the 2005 SPE Western Regional Meeting, Irvine, CA, USA, 30 March-1 April.
73. Morrow, N., Buckley, J. 2011. Improved Oil Recovery by Low-Salinity Waterflooding. Journal of
Petroleum Technology 63 (5): 106-112.
74. Muggeridge, A.H., Hongtong, P. 2014. An Upscaling Methodology for EOR. Paper We A27
proceedings of ECMOR XIV-14th European Conference on the Mathematics of Oil Recovery, Catania,
Sicily, Italy, 8-11 September.
226
75. Muskat, M. 1949. Effect of Permeability Stratification in Cycling Operations. Transactions of AIME
179 (1): 313-328.
76. Nasralla, R.A., Bataweel, M.A., Nasr-El-Din, H.A. 2013. Investigation of Wettability Alteration and
Oil-Recovery Improvement by Low-Salinity Water in Sandstone Rock. Journal of Canadian Petroleum
Technology 52 (2): 144-154.
77. Nasralla, R.A., Nasr-El-Din, H.A. 2014. Double-Layer Expansion: Is It a Primary Mechanism of
Improved Oil Recovery by Low-Salinity Waterflooding?. SPE Reservoir Engineering 17(1): 49-59.
78. Odsaeter, L.H., Berg, C.F., Rustad, A.B. 2015. Rate Dependency in Steady-State Upscaling. Transport
in Porous Media 110 (3): 565-589.
79. Osman, M.E., Tiab, D. 1981. Waterflooding Performance and Pressure Analysis of Heterogeneous
Reservoirs. Paper SPE 9656 proceedings of the Middle East Oil Technical Conference of the Society of
Petroleum Engineers, Manama, Bahrain, 9-12 March.
80. Perkins, T.K., Johnston, O.C. 1963. A Review of Diffusion and Dispersion in Porous Media. Society of
81. Pickup, G.E., Jensen, J.L., Ringrose, P.S., Sorbie, K.S. 1992. A Method for Calculating Permeability
Tensors using Perturbed Boundary Conditions. Paper proceedings of the ECMOR III-3rd European
Conference on the Mathematics of Oil Recovery, Delft, 17 June.
82. Pickup, G.E., Ringrose, P.S., Jensen, J.L., Sorbie, K.S. 1994. Permeability Tensors for Sedimentary
Structures. Mathematical Geology 26 (2): 227-250.
83. Pope, G.A. 1980. The Application of Fractional Flow Theory to Enhanced Oil Recovery. Society of
84. Pouryousefy, E., Xie, Q., Saeedi, A. 2016. Effect of multi-component ions exchange on low salinity
EOR: Coupled geochemical simulation study. Petroleum 2 (3): 215-224.
85. RezaeiDoust, A., Puntervold, T., Strand, S., Austad, T. 2009. Smart Water as Wettability Modifier in
Carbonate and Sandstone: A Discussion of Similarities/Differences in the Chemical Mechanism.
Energy & Fuels 23: 4479-4485.
86. Rapoport, L. A. 1955. Scaling Laws for Use in Design and Operation of Water-Oil Flow Models. AIME
Petroleum Transactions 204: 143-150.
87. Rashid, B., Muggeridge, A.H., Bal, A., Williams, G. 2012. Quantifying the Impact of Permeability
Heterogeneity on Secondary-Recovery Performance. SPE Journal 17 (2): 455-468.
88. Rashid, B. Simplifying Reservoir Models by Flow Regime, PhD Thesis, Imperial College London,
2012.
89. Renard, Ph., De Marsily, G. 1997. Calculating equivalent permeability: a review. Advances in Water
Resources 20 (5-6): 253-278.
90. Robbana, E., Buikema, T., Mair, C., Williams, D., Mercer, D., Webb, K., Hewson, A., Reddick, C.
2012. Low Salinity Enhanced Oil Recovery- Laboratory to Day One Field Implementation- LoSal EOR
227
into the Clair Ridge Project. Paper SPE 161750 proceedings of the Abu Dhabi International Petroleum
Exhibition & Conference, Abu Dhabi, 11-14 November.
91. Salazar, M.O., Villa, J.R. 2007. Permeability Upscaling Techniques for Reservoir Simulation. Paper
SPE 106679 proceedings of the 2007 SPE Latin American and Caribbean Petroleum Engineering
Conference, Buenos Aires, Argentina, 15-18 April.
92. Sandengen, K., Kristoffersen, A., Melhuus, K., Jøsang, L.O. 2016. Osmosis as Mechanism for Low-
Salinity Enhanced Oil Recovery. SPE Journal 21 (4): 1227-1235.
93. Sandengen, K., Arntzen, O.J. 2013. Osmosis During Low Salinity Water Flooding. Paper A19
proceedings of the IOR 2013-17th European Symposium on Improved Oil Recovery, St. Petersburg,
Russia, 16-18 April.
94. Schmalz, J.P., Rahme, H.D. 1950. The Variation of Waterflood Performance with Variation in
Permeability Profile. Prod.Monthly, 15 (9): 9-12.
95. Seccombe, J., Lager, A., Jerauld, G., Jhaveri, B., Buikema, T., Bassler, S., Denis, J., Webb, K., Cockin,
A., Fueg, E., Paskvan, F. 2010. Demonstration of Low-Salinity EOR at Interwell Scale, Edicott Field,
Alaska. Paper SPE 129692 proceedings of the 2010 SPE Improved Oil Recovery Symposium, Tulsa,
Oklahoma, USA, 24-28 April.
96. Seccombe, J.C., Lager, A., Webb, K., Jerauld, G., Fueg, E. 2008. Improving Waterflood Recovery:
LoSalTM EOR Field Evaluation. Paper SPE 113480 proceedings of the 2008 SPE/DOE Improved Oil
Recovery Symposium, Tulsa, Oklahoma, USA, 19-23 April.
97. Sharafi, M.S., Jamialahmadi, M. 2017. Investigation and quantifying effect of dispersion on oil
recovery in the process of viscoelastic polymer flooding using numerical method. Petroleum Science
and Technology 34 (15): 1333-1339.
98. Sheng, J.J. 2014. Critical review of low-salinity waterflooding. Journal of Petroleum Science and
Engineering 120: 216-224.
99. Shook, M., Li, D., Lake, L.W. 1992. Scaling Immiscible Flow through Permeable Media by
Inspectional analysis. In Situ 16 (4): 311-349.
100. Shook, G.M., Mitchell, K.M. 2009. A Robust Measure of Heterogeneity for Ranking Earth Models: The
F-PHI Curve and Dynamic Lorenz Coefficient. Paper SPE 124625 proceedings of the 2009 SPE Annual
Technical Conference and Exhibition, New Orleans, Louisiana, USA, 4-7 October.
101. Skrettingland, K., Holt, T., Tweheyo, M.T., Skjevark, I. 2011. Snorre Low-Salinity-Water Injection-
Coreflooding Experiments and Single-Well Field Pilot. SPE Reservoir Evaluation & Engineering 14
(2): 182-192.
102. Sorbie, K.S., Wat, R.M.S., Rowe, T.C. 1987. Oil Displacement in Heterogeneous Cores: Analysis of
Recovery Mechanisms. Paper SPE 16706 proceedings of the 62nd Annual Technical Conference and
Exhibition of the Society of Petroleum Engineers, Dallas, Texas, 27-30 September.
103. Sorbie, K.S., Collins, I.R. 2010. A Proposed Pore-Scale Mechanism for How Low Salinity
Waterflooding Works. Paper SPE 129833 proceedings of the 2010 SPE Improved Oil Recovery
Symposium, Tulsa, 24-28 April.
228
104. Stone, H.L. 1991. Rigorous Black Oil Pseudo Functions. Paper SPE 21207 proceedings of the 11th SPE
Symposium on Reservoir Simulation, Anaheim, California, 17-20 February 1991.
105. Suman, Y.K., Shirif, E., Ibrahim, H., Ala-Ktiwi, A. 2014. Evaluation of Low Saline “Smart Water”
Enhanced Oil Recovery in Light Oil Reservoirs. World Journal of Engineering and Technology 2: 13-
22.
106. Tang, G-Q., Morrow, N.R. 1999. Influence of brine composition and fines migration on crude
oil/brine/rock interactions and oil recovery. Journal of Petroleum Science and Engineering 24: 99-111.
107. Tang, G.Q., Morrow, N.R. 1997. Salinity, Temperature, Oil Composition and Oil Recovery by
Waterflooding. SPE Reservoir Engineering 12 (4): 269-276.
108. Taylor, G. 1953. Dispersion of soluble matter in solvent flowing slowly through a tube. Proceedings of
The Royal Society 219 (1137): 186-203.
109. Taylor, G. 1954. Conditions under which dispersion of a solute in a stream of solvent can be used to
measure molecular diffusion. Proceedings of The Royal Society 225 (1163): 473-477.
110. Thiele, M.R., Batycky, R.P., Fenwick, D.H. 2010. Streamline Simulation for Modern Reservoir-
Engineering Workflows. Journal of Petroleum Technology 62 (1): 64-70.
111. Thyne, G., Gamage, P. 2011. Evaluation of the Effect of Low Salinity Waterflooding for 26 Fields in
Wyoming. Paper SPE 147410 proceedings of the SPE Annual Technical Conference and Exhibition,
Colorado, 30 October- 2 November.
112. Trigos, E.M., Gonzales, A.R., Pinilla, J.M., Munoz, S.M., Mercado, D.P. 2010. Feasibility Study of
Applying Steamflooding in a Reservoir With High Shale/Sand: Teca Field. Paper SPE 132775
proceedings of the Trinidad and Tobago Energy Resources, Port of Spain, Trinidad, 27-30 June.
113. Vledder, P., Fonseca, J. C., Wells, T., Gonzalez, I., Ligthelm, D. 2010. Low Salinity Water Flooding:
Proof of Wettability Alteration On A Field Wide Scale. Paper SPE 129564 proceedings of the 2010
SPE Improved Oil Recovery Symposium, Tulsa, Oklahoma, USA, 24-28 April.
114. Wheat, M.R., Dawe, R.A. 1988. Transverse Dispersion in Slug-Mode Chemical EOR Processes in
Stratified Porous Media. SPE Reservoir Engineering 3 (2): 466-478.
115. Wright, R.J., Wheat, M.R., Dawe, R.A. 1987. Slug Size and Mobility Requirements for Chemically
Enhanced Oil Recovery Within Heterogeneous Reservoirs. SPE Reservoir Engineering 2 (1): 92-102.
116. Zaid, A.M., Al-Dousari, M. 2007. Forces Driving Oil and Gas Demand Cycles. Paper IPTC 11277
proceedings of the International Petroleum Technology Conference, Dubai, 4-6 December.
117. Zapata, V.J., Lake, L.W. 1981. A Theoretical Analysis of Viscous Crossflow. Paper SPE 10111
proceedings of the 56th Annual Fall Technical Conference and Exhibition of the Society of Petroleum
Engineers of AIME, San Antonio, Texas, 5-7 October.
118. Zhang, Y., Morrow, N.R. 2006. Comparison of Secondary and Tertiary Recovery With Change in
Injection Brine Composition for Crude Oil/Sandstone Combinations. Paper SPE 99757 proceedings of
the 2006 SPE/DOE Symposium on Improved Recovery, Tulsa, Oklahoma, USA, 22-26 April.
229
119. Zhang, Y., Xie, X., Morrow, N.R. 2007 Waterflood Performance by Injection of Brine With Different
Salinity for Reservoir Cores. Paper SPE 109849 proceedings of the 2007 SPE Annual Technical
Conference and Exhibition, Anaheim, California, USA, 11-14 November.
120. Zhou, H. Algebraic Multiscale Finite-Volume Methods For Reservoir Simulation, PhD Thesis, Stanford
University, 2010.
230

Attar A 2017 PHD Thesis.

Uploaded by

Document Informationclick to expand document informationlow salinity waterflooding modelling

Document Informationclick to expand document information

Copyright:

Available Formats

Attar A 2017 PHD Thesis.

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Attar A 2017 PHD Thesis.

Uploaded by

Copyright:

Available Formats

EVALUATION OF THE IMPACT OF RESERVOIR HETEROGENEITY

AND MIXING ON LOW SALINITY WATERFLOODING

AHMED ZAKARIYA ATTAR

Doctor of Philosophy (PhD)

Earth Science and Engineering

Imperial College London

Professor Ann Muggeridge

Ahmed Zakariya Attar

CHAPTER 2: LITERATURE REVIEW

2.1.2 Methods for Modelling Low Salinity Waterflooding

Figure 2.12: Schematic of the experimental setup used to investigate the

B. Glass Substrate Model Experiment

Displacement post a high

Experiment by Discussions on wettability alteration due to low salinity water injection

Brine salinity divalent ions (Ca++ and Mg++)

II. Field Applications

Wettability alteration Proposed reasons by the authors for

Figure 2.21: Scales of geological heterogeneity (Henson et al. 2002).

2.2.2 II Heterogeneity Effects on Various Displacement Processes:

Cross Flow and Mixing Between Layers

𝑘𝑘𝑤𝑤𝑤𝑤 ∑𝑛𝑛𝑖𝑖=1 𝐾𝐾𝑖𝑖 ℎ𝑖𝑖

any layer between 1< n< N

any layer between 1< n< N

3.2 Analytical Model for Low Salinity Waterflooding

where: 𝑆𝑆𝑤𝑤𝑤𝑤 is the connate water saturation.

saturation as shown in Figure 3.4

The time of the trailing shock front breakthrough is calculated as follows:

𝑑𝑑𝑓𝑓𝑤𝑤 𝑞𝑞𝑜𝑜 𝑞𝑞𝑤𝑤 𝑞𝑞𝑡𝑡 𝑄𝑄𝑖𝑖 𝑡𝑡

Leading shock front (𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 ) 0.52

High Salinity Waterflooding

𝑑𝑑𝑓𝑓𝑤𝑤 𝑓𝑓𝑤𝑤,𝑆𝑆𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑓𝑓𝑤𝑤,𝑆𝑆𝑤𝑤,𝐻𝐻𝐻𝐻 1 − 0.68

𝑑𝑑𝑓𝑓𝑤𝑤 𝑓𝑓𝑤𝑤,𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 0.95

0.8 Sw, LS Sor

0.3 Oil Bank

Trailing front 𝑆𝑆𝑤𝑤,𝐿𝐿𝐿𝐿 mobility ratio:

B-Ion Exchange Model

Values to change between the relative permeability curves Cation exchange

3.3.4 Modelling of Diffusion

0.4 50 grid blocks

0.3 98 grid blocks

1.4 y = 0.5631x + 0.1666

Produced water salinity (1,000 ppm)

Figure 3.14: a) water cut vs PVI b) produced water salinity vs PVI

Figure 3.15: a) water cut vs PVI b) produced water salinity vs PVI

0.6 analytical solution

0.3 CMG GEM

0.4 14,000-15,000 ppm inj at

0.2 29,000-29,500 ppm

0.1 analytical solution

0.7 1,000-7,000 ppm

0.6 1,000-1,500 ppm

The producer is 30 cm away from the injector

10 grid blocks 100 grid blocks 1000 grid blocks

Salt concentration (mole fraction)

Salt concentration (mole fraction)

NZ = 4 grid blocks NZ = 4 grid blocks

NZ = 2 grid blocks NZ = 2 grid blocks