SPE-198586-MS

A New Continuous Waterflood Operations Optimization for a Mature Oil

Field by using Analytical Workflows that Improve Reservoir Characterization

Atul Yadav and Anton Malkov, Wintershall Dea; Essam Omara, Suez Oil Company; Ahmed El-Hawari, Wintershall
Dea; Davud Davudov, University of Oklohoma; Yusuf Danisman, Queensborough Community College; Ashwin
Venkatraman, Resermine Inc.

Copyright 2019, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Gas & Oil Technology Showcase and Conference held in Dubai, UAE, 21 - 23 October 2019.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Water flooding is an established method to increase oil production. In this research, we present a novel
approach that uses data mining techniques on the operations data on a complex mature oil field located
in the Gulf of Suez, currently being water flooded. We show how such methods help improve reservoir
characterization for this specific field is particularly challenging because of its geological complexity and
field performance.
The continuous recording of production and injection data presents a new opportunity to apply analytical
approaches to reservoir management. Such approaches provide an alternative to the traditional history-
match model update and prediction that is not only time-consuming but also carry forwards all subsurface
uncertainties.
A combination of qualitative (cross-correlation analysis) and quantitative analysis (capacitance resistance
model) is used to obtain an overall waterflood injection strategy for this Gulf of Suez field. In this
manuscript, we focus on the analysis obtained from cross-correlation analysis. The presented analysis helps
identify connectivity between wells in the reservoir during waterflood. The method presented is also adapted
to specific characteristics of this field - water drive production in this model.
We present evidence of how salinity data can be used to further justify the linkages between the different
wells obtained from the cross-correlation analysis. We also show comparison between results from this
analytical technique and the streamline approach. This comparison with salinity and streamlines helps
benchmark the model results especially in cases where such secondary data (salinity/streamlines) are not
available. The results presented in this research can be adapted to any waterflooded field to optimize
recovery at frequent intervals, where injection and production data is continuously available.

Introduction
Mature oil fields typically present challenges of increased water production and hence, water handling.
Considering the geological complexity and field performance behavior, reservoir characterization to
optimize waterflood is a big challenge. To minimize the reservoir uncertainties and increased understanding
2 SPE-198586-MS

of the field performance behavior integrated reservoir study was conducted in 2016. The acceptable history
matched model was used to estimate remaining oil potential, maintain/ increase the current production
levels, and optimize the water injection rate. Generally, history matched model needs to be updated
throughout the life of the producing fields as new sub-surface data is acquired. The integrated reservoir
modeling studies can be time-consuming that do not enable quicker decision making around operational
activities.
The continuous recording of production and injection data presents new opportunities to apply novel
analytical techniques to understand interwell connectivity in the reservoir. The current ability to store and
analyse data coupled with advances in the ability to interpret big data sets has helped create an independent
tool kit that provides analysis without the geological model. In addition, the geological information such as
pre-existing faults as well as comingled or disconnected nature of production between different layers can
be integrated into such analysis to obtain and improve analysis from the analytical models.
Refunjol and Lake (1999) have previously used Spearman coefficient to compare the correlation between
production and injection data to determine preferential flow direction. Spearman coefficient is a non-
parametric approach (Siegel (1996); Albertoni and Lake 2003) for measuring correlation between two
signals. Heffer et al (1997) used Spearman's rank correlation to determine fracture correlation. More
recently, Venkatraman et al (2019) showed how a combination of cross-correlation and capacitance
resistance model can be applied for a mature field operation to obtain interwell connectivity and through
that increase oil recovery.
We analyze the results using the Pearson's cross correlation analysis measure to obtain a qualitative
analysis of the field. We also apply Spearman's rank correlation analysis for the GOS field that helps compare
injection and production data. The objective is to present a comparison between the analytical approach and
the streamline approach to show consistency in reservoir characterization. The effective injector producer
pairs identified form an important component of the field development. In the following sections, we discuss
the systematic approach for field development decision making to deliver the optimal model.

Background
The mature oil field is one of the offshore fields located in Gulf of Suez. The original field name has been
changed to GOS field for the sake of confidentiality. The GOS field is situated within central province off
the eastern coast of the Gulf of Suez. The field was discovered in 1977 having a very complex structure,
trending to the northeast-southwest direction.
The complex pre-Miocene structure has been severely disturbed by many faults aligned in various
directions at different levels. Faults divided the reservoirs into 42 compartments, which are in lateral
communication but with different degree through fault juxtapositions, however vertical communication
among flow units are almost impeded. In addition to a post-Paleozoic unconformity, the presence of
post-Eocene and intra-Rudeis erosional unconformities in the stratigraphic section enhances the structural
complexity (Geipel, 2014).
The stratigraphy of GOS field consists of a sedimentary sequence of Paleozoic rocks. The oil has
been produced from sandstone reservoirs of Raha and Nubia having Cenomanian and Lower cretaceous-
paleozoic age respectively. Figure1 shows the complex reservoir structure and reservoir units (El-Hateel
et al, 2014).
SPE-198586-MS 3

Figure 1—Structure contour map and reservoir section of the mature field.

GOS Reservoir
The field comprises of three main reservoir faults blocks (A, B, and C). The initial oil production was
started in the year 1983 and further field development has been performed by drilling 42 wells from three
offshore platforms A, B, and C as shown in figure 2. The reservoir subunit has been defined based on
the palaeontological, lithological, core description, log analysis, and production performances. There are 9
reservoir sub-units, starting at the top of the reservoir namely Upper Raha, Middle Raha, Lower Raha, Unit-
III, Lower unit-III, Unit-IIB, Unit-IIA, Unit-IA, and Unit-I. The various degree of lateral communication
exists among reservoir sub-units through juxtapositions faults blocks but vertical communication is impeded
by presence of horizontal shale barriers. In addition, observed RFT data confirms the absence of vertical
communication among reservoir sub-units (Omara et al 2014).

Figure 2—Schematic showing reservoir development overview of different platforms

The reservoir depth is from 10,070 ft. to 12,500 ft. TVDSS and oil leg reaches up to 2000 ft.
Aforementioned 9 reservoir units have wide range of porosity and permeability. The range of permeability is
from 40 to 1300 mD and porosity from 5 to 17 %. According to reservoir property distribution Upper Raha,
Lower Raha, Unit-III, Lower unit-III, Unit-IIB, Unit-I have good reservoir properties for the fluid flow
compared to rest of the units. The observed initial reservoir pressure is 5632 psi at datum depth of 11,400
ft. TVDSS. The reservoir oil gravity ranges from 18.5 to 26.1° API. The main reservoir drive mechanisms
are bottom aquifer support plus peripheral water injection.
As mentioned earlier, oil production started in April 1983 and waterflooding was initiated with two
injectors in the year 1985 to give pressure support in the upper reservoir units. The field development was
4 SPE-198586-MS

achieved by drilling wells into different fault blocks. Till now, the field has produced a cumulative oil
production of ~289 MMSTB.
A previous integrated reservoir modeling study was conducted that used a holistic view of the field to
provide optimum solutions based on all decision variables and constraints to achieve maximum NPV, oil
production and minimum water cut (Yadav et al 2018). This study remained partly helpful because of the
complex reservoir structure and the underlying uncertainties further made it difficult to characterize the
fluid flow in the reservoir. Such integrated numerical models need to be updated throughout the life of
producing field. It makes it even more challenging and cumbersome when fields are complex to model
(static to dynamic) especially, when it is necessary to obtain quick decisions.
A field-wide waterflood optimization strategy needs careful analysis of the production performance of
wells in each segment/block, connectivity among injection and producer blocks, aquifer connectivity. Given
all the history of data recorded, this mature field also provides an opportunity to apply analytical techniques
to identify thief zones as well as effective injector-producer pairs. The quick analysis avoids the cumbersome
workflows associated with an integrated numerical model. This can help, especially, in not just reducing
water recirculation but also identify effective wells that can help increase oil recovery.
In this manuscript, we present how the analytical models help in the characterization of these mature
and complex reservoirs. We also present a workflow of using additional acquired data obtained during
operations– salinity of produced water to further verify results obtained from the analytical model. The
streamlines obtained from the dynamic model are further used to validate the results from this approach.

Waterflooding History
In the GOS field, the water injection started after two years of initial field production. The well A3B and
A2A started injecting water in the year 1985. GOS field has moderate to good aquifer support plus peripheral
water injection (Figure 3).
SPE-198586-MS 5

Figure 3—Schematic showing the peripheral nature of water injection

Initially, well production started commingled from several layers having different reservoir properties.
Due to variable pressure and reservoir heterogeneities across different layers, in addition to crossflow in
well water showed up in some layers. The operator simply isolated the watered-out layers due to which
field production continued to decline. Having experienced such a difficulty in the past, new strategy for
upcoming side-tracks has a bottom-up strategy in each well and considers production from single layer.
According to Baily (2000), it is crucial to understand the distinction among different types of water. The
water sweeping the oil in the reservoir is obviously more important as it determines the well productivity
and recoverable reserves in the field. The second type namely good water is produced together with oil
which gradually increases the WOR with time. It cannot be shut-off without losing reserves, because it
flows commingled with oil. Proper management of good water can lead to good oil production until the
point WOR reaches to an economical limit. The third type namely bad water produces no oil, WOR is above
the economic limit. There are multiple reasons for this bad water - operational and reservoir elements such
as casing, tubing or packer leaks, moving oil-water contact, presence of fractures or faults.
In the GOS field, we have observed all the elements which have caused excessive water production.
There are 9 reservoir units that have lateral communication, but vertical communication is prevented by
interbedded shale layer. The wells producing from bottom reservoir unit have produced excessive water
either originating from aquifer or injection well, the main reasons attributed to reservoir heterogeneities,
higher vertical permeability. Most of the wells having multilayer production suffered an unwanted water
production problem, the watered-out layer having high permeability zone brought either injected or aquifer
water towards the producer. In the absence of crossflow among reservoir units, the watered-out zones were
isolated using mechanical and rigid shutoff fluids.
6 SPE-198586-MS

As an example, the top most reservoir layer called Upper Raha has suffered early water breakthrough
problems due to the presence of a high-perm channel between injector and producer. This is a good example
of bad water, where well R-5 started production with ~ 4000 bopd but due to rapid water cut development
up ~ 90 % within a couple of months the producer had to shut-in leaving behind a significant amount of
remaining reserves.
In summary, the peripheral/edge water from aquifer and injectors through a complex reservoir faulted
reservoir structure leads to poor areal and vertical sweep efficiency. It is necessary to divert the injected water
away from the already swept area. This warrants not just continuous monitoring but also quick decision
making that is provided by the below presented analytical method.

Qualitative analysis using Pearson and Spearman cross correlation analysis

Cross-correlation is a signal processing technique to compare two-time series signals – in this case signals
from injector wells and producer wells. Data from each injector and surrounding producers are used for
cross-correlation analysis. This helps to identify the most effective injector-producer pairs for further use in
quantitative analysis that gives redistribution of water for higher oil recovery. The Pearson cross-correlation
measure (r) for two-time series x and y are given as follows,

(1)

Here, μx and μy are the means of time series x and y respectively. Also, σx and σy are the standard-deviation
of time-series x and y respectively while N is the total data points in the time series x and y and τ stands
for the time lag.
Additionally, following Refunjol and Lake (1999), Spearman rank correlation (rs) can also be used to
quantify the relationship between injector and producer well pairs as follows,

(2)

Here dn is the difference between rankings of the nth data points. We adapt this measure to production
injection flow rates measured with respect to time. We also refer these values as production and injection
time series data. The injection data is series x while series y is the production data. The following steps are
taken to find the cross-correlation values –
1. A time window (based on the total data available) is established for comparison that determines the
number of data points N in every series x and y.
2. A radius of investigation helps determine producer wells of interest around every single injector. We
take a high enough radius of investigation to ensure that every injector will be surrounded by all the
producer pairs for this GOS field.
3. The Pearson cross-correlation (Eq 1) and Spearman rank correlation (Eq 2) are calculated for every
injector-producer pair and repeated for varying values of time lag (τ). The time lag is representative
of the time it takes for any injector well signal to reach the producer well.
4. This time lag is varied from zero to half of the total historical data available.
5. The maximum Pearson cross-correlation and Spearman rank correlation values are recorded for every
injector and producer pair. The time lag when this maximum value occurs is also recorded.
The injector producer pairs with higher cross-correlation values (maximum value above 0.6) at lower
phase lags can be considered as potential thief zones. This implies any change in the injection data is
immediately observed (lower time lag) in the producer reflected by the high cross-correlation value. On the
other hand, higher cross-correlation values which occur at higher time lag correspond to injector producer
SPE-198586-MS 7

pairs that are related and more effective for waterflood operations. Alternatively, results for low cross-
correlation results (maximum value below 0.3 at any time lag) are analyzed and discussed.
The results obtained from the above analysis helps in the qualitative classification of injectors in the
field. Because this analysis corresponds to only production and injection data, it also reflects the changes
occurring in the geology during injection in case of a mature field being waterflooded for a long time. Also,
additional data wherever available must be used to further verify results from this analysis. In this research,
we use additional data like salinity and streamlines to benchmark model results to further gain confidence
in predictions, especially in cases where additional data is not available. The results of the cross-correlation
analysis for GOS field is presented in the next section.

Results and Discussion

A large radius of investigation is chosen so that all possible pairs of injectors and producers are considered
in this analysis. The cross-correlation values obtained for all possible well pairs for different lag values are
stored. These values ranged from −1 to 1. In this study, negative values are considered as zero correlation.
The cross-correlation values along with the lag value associated help identify potential thief zones, effective
injector producer pairs, as well as the well pairs that are not connected.

Pearson and Spearman cross-correlation analysis on GOS

Given the vast history of waterflooding in GOS, we choose production and injection data corresponding to
more later stages of production which is from 01/01/2014 to 11/01/2018 (59 months). For the chosen time
period active injector and producer wells are shown in Figure 4. In this study, the results from lag values
that correspond to 50% of the entire data (up to 30 months) have been considered.

Figure 4—Schematic showing the surface coordinates of injection and production wells
8 SPE-198586-MS

Figure 5 illustrates the results obtained from both Pearson and Spearman correlations for A1 and A11
injector producer pair at different time lag values. Results indicate that both models have similar and hence,
comparable results as shown in Figure 6. This outcome also confirms that the nature of the injector producer
data is close to a linear relationship.

Figure 5—Cross correlation results for A1 - A11 injector-producer pair between 2014–
2018 at different time lag (a) Pearson cross correlation b) Spearman rank correlation

Figure 6—Comparison of Pearson and Spearman correlation results for

A1 - A11 injector-producer pair between 2014–2018 at different time lag

The Pearson and Spearman model results with high correlation values (above 0.6) are summarized in
Tables 1 and 2, respectively. It can be seen from Tables 1 and 2 that injectors A1 and A3B mostly contribute
to block A producers whereas injector A9A distributes water to all blocks. Block B wells receive injected
fluid mostly because of A9A injector. Table 3 summarizes consistent results from both correlations helping
realize that A9A is the injector well impacting all production wells in all the blocks. Figure 7 shows a
schematic of connection between different wells as identified in Table 3. It should be noted that A1 injector
also shows correlation with producer well C4 that is further discussed in the next section.
SPE-198586-MS 9

Figure 7—Schematic showing correlation between injector and producer wells as summarized
in Table 3. The x and y axis are surface coordinates of the wells as in Figure 4 and the
thickness of the line connecting injector producer pair reflects the strength of the correlation.

Table 1—Pearson cross-correlation results at high correlation values (>0.6)

Table 2—Spearman cross-correlation results at high correlation values (>0.6)

Table 3—Consistent results (c) from Pearson and Spearman at high correlation values (>0.6)

Non-correlated pairs (below 0.3) besides high correlated pairs, can also be identified. This has two
practical applications; a) Determine injector and producer pairs that are not correlated and hence referred to
as non-correlated pairs, b) Determine non-correlated producers around an injector to reduce the number of
unknowns for more sophisticated models such as capacitance resistance (CRM) (Venkatraman et al.2019).
10 SPE-198586-MS

Results for non-correlated pairs obtained from Pearson and Spearman models are summarized in Tables 4
and 5 respectively. Table 6 summarizes correlated pairs that are consistent from both Pearson and Spearman
cross-correlation models. It can be observed that most of the production wells are potentially correlated with
injector A9A which is consistent with previous results. Moreover, injectors A1 and A3B do not contribute
to most of the wells from block B.

Table 4—Pearson cross-correlation results at low correlation values (<0.3)

Table 5—Spearman cross-correlation results at low correlation values (<0.3)

Table 6—Consistent no connection results (x) from Pearson and Spearman (correlation values <0.3)

A complete summary of results is presented in Table 7 which shows the highly correlated pairs (marked
as c), as well as non-correlated pairs (marked as x) consistent across both the Pearson and the Spearman's
cross-correlation analysis.

Table 7—Summary of consistent results from Pearson and Spearman correlations for high correlated (C)
and non-correlated (x) pairs. A9A contributes to all blocks but A1 and A3B only contributes to block A.

Given the difference in salinity between in-situ water (120,000 – 160,000 mg/l) and injected water
(25,000 – 30,000 mg/l), change in salinity of water in the producer well can be used to verify if injected
water is breaking out. Further, the streamlines plotted from injectors help indicate most effective injector-
SPE-198586-MS 11

producer pairs. If the injector-producer pairs are effective as identified in this analysis, a streamline density
map would be denser around these pairs. These verification steps are explained in the next section

Verification using salinity data

A drop-in salinity measured at the producer well, given the large differences between in-situ and injected
water salinity, implies the breakout of injected water. We use this difference in salinity between in-situ
and injected water to verify the presence of correlated and non-correlated pairs identified by model results.
This serves as a benchmark for results obtained from the model and hence gain more confidence in model
predictions, especially when such salinity data is not available.
The salinity measured at producer wells is shown in Figure 8. Table 6 identifies all well pairs that are
not correlated and consistent with both Pearson and Spearman's rank correlation. If there is any interaction
with the injected water and producer wells of Block A, one can expect intermediate salinity in these wells.
Given that only low salinity values are observed in block A, this implies that the model results are consistent
with salinity observations.
12 SPE-198586-MS

Figure 8—Salinity measurement at producer wells in (a) Block A (b) Block B and (c) Block C
SPE-198586-MS 13

By combining results from Tables 1–3 we can confirm that the salinity in wells of block A has low
salinity concentration (producer wells A11, A12, A4, A5A, A8E, and B2) which corresponds to that of
injected water while that of producer wells from block B (producer wells B6, B7, B10, B11, and B13)
has an intermediate salinity suggesting that breakthrough of injected water is occurring through one of the
perforations (because one observes comingled production at the surface).

Verification using streamlines

An integrated model developed previously was used to generate streamlines and compare connectivity
identified by cross-correlation analysis. The streamlines around injectors A1, A3B and A9A are given in
Figure 9. It can be seen that all three injectors have strong connections to wells in block A whereas A9A is
the one that mostly contributes to block B wells that is consistent with cross-correlation analysis.

Figure 9—Streamline results for A3B (blue), A1 (dark brown) and A9A (orange)

Discussion
Both Pearson and Spearman correlations are simple, fast and efficient measure to analyze production data
and determine both effective and non-correlated injector-producer pairs. On the other hand, both models
have certain limitations and challenges.
Pearson correlation is parametric test which is more suitable when data is close to normally distributed,
whereas Spearman rank correlation is a nonparametric model and can be used even if the data is not normally
distributed. Even though these two measures on our dataset have the same results (Figure 6), they may not
be true for other datasets. In summary, if Pearson correlation is used for analysis, it should be complemented
by Spearman rank correlation before making predictions and analyzing the results.
Like any statistical model, the presented model results themselves are strong function of the quality of the
data. Outliers and some noise data points may result in incorrect outcome which could be misleading. As an
example, based on the coordinates and salinity data, producer well A12 is expected to have connection with
A1 and/or A3B injectors which is not really observed from the results of both correlations. Pre-processing
14 SPE-198586-MS

of outliers associated with the dataset is critical before running the model and interpretation of the results.
The specific treatment of outliers varies based on the information available for the well.
The correlation values obtained in this analysis are between −1 and 1, with positive values representing
a perfect correlation between the injector-producer pair. However, negative correlations are not physical as
they imply decrease in production with an increase in injection. This may result owing to the interaction
between the injectors themselves. In this study, negative values are considered as zero correlation.
In many scenarios, several injectors will have similar injection schedules which will result in identical
input for the model. These injectors will likely show similar outcomes in terms of connectivity which can
be misleading. In the case study presented, when Pearson correlation results are analyzed from Table 1,
results are similar from both injectors A1 and A9A. This awareness means that the high correlation obtained
between injector A1 and producer C4 needs to be further analyzed and treated more carefully.

Conclusions
The application of analytical techniques to a mature and complex waterflood GOS field is presented. Cross-
correlation analysis that includes Pearson and Spearman's rank correlation were presented to determine
connectivity between injectors and producers in the reservoir. The following conclusions can be drawn -
1. The results obtained from Pearson and Spearman's correlation are similar for the dataset used. Well
pairs with high connectivity and no connectivity were determined based on consistency between the
two correlation values. The most significant result is that A9A is the only injector that is connected
to majority of the wells in all the blocks. The other two injectors (A1 and A3B) are limited in their
capability and only connected to wells from block A.
2. The additional data available for the GOS field- salinity and streamlines generated by a previous
integrated model helped verify connectivity between the wells. This helps verify model results and
in this case, it was used to benchmark model results especially in cases where such additional data
is not available.
3. The proposed technique can be extended as a qualitative approach to obtain thief zones (high
correlation values at low phase lags) to avoid water recirculation, a problem with mature waterfloods.
In the presence of additional data like salinity data, results from cross-correlation analysis can help
benchmark and further gain confidence in model predictions. This approach can be used for screening
and identifying injector wells to shut for effective water management in mature fields where many
injector wells are present and there is absence of salinity data.
4. The application of analytical models on historical data in mature fields presents an opportunity to
form an additional toolset for quicker decision making without geological data.
5. This proposed technique is used to identify non-correlated pairs injector producer pairs that are
further used to reduce number of unknowns solved for Capacitance Resistance modeling that form a
quantitative analysis for an effective waterflood strategy to increase oil recovery in the mature field.

Acknowledgment
The authors of this paper thank the management of the Wintershall Dea for giving permission to publish
this paper. We further wish to thank SUCO management for support in this project. Resermine (MOST
PROMISING STARTUP, OTC 2018) provides a cloud platform for reservoir modeling that includes
several proprietary modules in subsurface data analytics. The results presented in this manuscript are from
Resermine's subsurface analytics platform configured for Wintershall DEA.

Nomenclature
RFT = Repeat formation tester
SPE-198586-MS 15

TVDSS = True vertical depth subsea

MMSTB = Million standard barrel
ft. = Feet
API = American petroleum institute
psi = Pounds per square inch
mD = Millidarcy
SCAL = Special core analysis
PVT = Pressure, Volume, Temperature
ppm = Parts per million
mg/ l = Milligram per litre

References
1 Albertoni, A., & Lake, L. W. (2003). Inferring interwell connectivity only from well-rate
fluctuations in waterfloods. SPE Reservoir Evaluation & Engineering, 6(01), 6–16.
2 B. Bailey, M. Cratree, J. Tyrie et al: "Water Control" Oilfield Review 2000 12 (1): 30–51.
3 Geipel Peter, 2014: "3D static model report" Internal report DEA company, Hamburg, Germany.
4 Heffer, K. J., Fox, R. J., McGill, C. A., & Koutsabeloulis, N. C. (1997). Novel techniques show
links between reservoir flow directionality, earth stress, fault structure and geomechanical
changes in mature waterfloods. SPE Journal, 2(02), 91–98.
5 Omara E., El Hawary, Nosseir M. et al: "Identifying Opportunities in a complex Mature Oil
Reservoir" presented at the 2014 International Petroleum Technology Conference, Doha, Qatar
Jan 19 - 22
6 Refunjol, B.T., and Lake, L.W.,, 1999. Memoir 71, Chapter 15: Reservoir Characterization Based
on Tracer Response and Rank Analysis of Production and Injection Rates.
7 Siegel, S., & Castellan, N. J. (1956). Nonparametric statistics for the behavioral sciences (Vol. 7).
New York: McGraw-hill.
8 Venkatraman, A., Malkov, A., Yadav, A., Davudov, D., Awemo, K., Hag, T., & Chen, X. (2019,
April). A New Qualitative and Quantitative Analytics Approach on Waterflood Operations Data
for Improving Oil Recovery. In IOR 2019–20th European Symposium on Improved Oil Recovery.
9 Yadav, A., Omara, E., El-Hawari, A., Malkov, A., Mba, B. B., & Michelle, E. (2018, October).
Deterministic or Probabilistic Approach for Mature Oilfield Development: Isn't it all about
Decision Making? (Russian). In SPE Russian Petroleum Technology Conference. Society of
Petroleum Engineers.