Exploration Insights December 2019/ January 2020 | Page 16

16 | Halliburton Landmark
core, the play is thin (15–30 m) and production is
limited by, and highly sensitive to, play thickness
(Table 3). In the northeastern core, the play is
considerably thicker (100–200 m), but production
is limited by reservoir porosity, reflected in the
high importance weighting of porosity (Table 3).
CONCLUSIONS
» Random Forest classifier produces the
most accurate ML models predicting
initial production success of individual
wells to 97% accuracy, providing insight
into the plays from which the data are
trained only. The potential for widespread
application of this model increases as
data are collected from a larger range of
unconventional plays.
» This ML method has application in
the petroleum industry as a means to
streamline production and eliminate
expenditure on unnecessary data
collection, providing a method of
understanding which measurements are
most important as defined by features
most correlated to production.
» Comparison of the Bakken and Marcellus
plays demonstrates that production rate
within two prolific resource plays can be
sensitive to different subsurface variables
and that this is fundamentally related
to the geological differences between
the plays. Characterization of geological
heterogeneities is, therefore, integral to
achieve success, both in U.S. plays and
analogous plays worldwide.
» We have found that pore pressure
is highly influential in the production
success of the majority of plays.
However, importance between geological
parameters and production does vary
significantly between plays. This is
fundamentally related to three geological
heterogeneities between resource plays,
as demonstrated by the comparison of
the Bakken and Marcellus plays.
Exploration Insights | 17
REFERENCES
Production data used within this study is from ShaleWellCube
(version 13-06-2019) and is provided courtesy of Rystad
Energy.
McGlade, C., Speirs, J. & Sorrell, S. (2013). Methods of
estimating shale gas resources - Comparison, evaluation and
implications. Energy. [Online]. 59. p.pp. 116–125. Available
from: http://dx.doi.org/10.1016/j.energy.2013.05.031.EIA
(2018). Permian Basin Wolfcamp Shale Play. (October).
Flender, S. (2019). Data is not the new oil. [Online].
2019. Towards Data Science. Available from: https://
towardsdatascience.com/data-is-not-the-new-oil-
bdb31f61bc2d.
Géron, A. (2017). Hands-On Machine Learning with Scikit-
Learn & TensorFlow. 1st Ed. Sebastopol, CA: O’Reilly Media,
Inc.
Dimitriadis, S.I. & Liparas, D. (2018). How random is the
random forest? Random forest algorithm on the service of
structural imaging biomarkers for Alzheimer’s disease: From
Alzheimer’s disease neuroimaging initiative (ADNI) database.
Neural Regeneration Research. 13 (6). p.pp. 962–970.
Patel, S. (2017). Chapter 2: SVM (Support Vector Machine)
— Theory. [Online]. 2017. Medium. Available from: https://
medium.com/machine-learning-101/chapter-2-svm-support-
vector-machine-theory-f0812effc72.
Du, W., Du, W., Zhan, Z. & Zhan, Z. (2002). Building decision
tree classifier on private data. Proceedings of the IEEE
international conference on Privacy, security and data mining-
Volume 14. [Online]. p.pp. 1–8. Available from: http://portal.
acm.org/citation.cfm?id=850784.
Jarvie, D.M. (2012). Shale Resource Systems for Oil Resource
Systems: Part 2 - Shale-oil resource systems. AAPG Memoir.
97. p.pp. 89–119.
Zagorski, W.A., Bowman, D.C., Emery, M. & Wrightstone,
G.R. (2011). An overview of Some Key Factors Controlling Well
Productivity in Core Areas of the Appalachian Basin Marcellus
Shale Play *. 110147. p.p. 90122.
DISCLAIMER
This article is a synthesis based upon published data
and information, and derived knowledge created within
Halliburton. Unless explicitly stated otherwise, no
proprietary client data has been used in its preparation.
If client data has been used, permission will have been
obtained and is acknowledged. Reproduction of any
copyrighted image is with the permission of the copyright
holder and is acknowledged. The opinions found in
the articles may not necessarily reflect the views and/
or opinions of Halliburton Energy Services, Inc. and its
affiliates including but not limited to Landmark Graphics
Corporation .
Earth Model Award: Rewarding
Excellence in Master’s Level
Research
The Earth Model Award is the culmination of an
international geoscience competition that rewards
excellence in master’s level research. It was
established as the Neftex ® Earth Model Award
in 2012, in affiliation with the Geological Society
of London, to foster the link between industry
and academia. The award ceremony for the 2018
Earth Model Award was held at Halliburton’s
LIFE2019 conference in Houston, Texas, U.S.A.
This year, the winner of the Earth Model Award
was Aasmund Olav Løvestad from the University
of Bergen for his master’s project on fluvial
reservoirs in Utah. The second prize went to
James Lovell-Kennedy from the University of
Manchester, and third place was awarded to
Landon Lockhart from the University of Texas.
Here’s what our winners have to say about the
2018 Earth Model Award…
It was a great honor to receive the 2018 Earth
Model Award for excellence in master’s level
research for the work I did on fluvial reservoirs
in Utah. Emerging researchers are fortunate to
have the encouragement and support offered by
awards such as this. The LIFE2019 conference
was wonderful and a great opportunity to meet
members of the international industry, and to
discover what the leading universities in the
field are focusing on. My deepest gratitude
goes to The Geological Society of London and
Halliburton Landmark for this award. Finally, my
sincere thanks and appreciation in supporting my
research go to The University of Bergen and to
Associate Professor Christian Haug Eide for his
great mentorship and invaluable insight, which
were truly inspiring.

~ Aasmund Olav Løvestad, University of
Bergen
Upon completion of my master’s degree at the
University of Manchester, I responded to the
advert from Halliburton Landmark and submitted
my thesis to the 2018 Earth Model Award
competition. My thesis looked at ‘Assessing
the provenance and contribution of local versus
regional drainage systems for the Upper Triassic
fluvial deposits, High Atlas, Morocco.’ It involved
collaboration between the North Africa Research
Group (NARG) and Sound Energy to try to
improve understanding of reservoir quality fluvial
sands in the Triassic of Eastern Morocco.
To my surprise, I won second place in the 2018
Earth Model Award, and as a result was invited
to attend LIFE2019 in Houston, Texas, where I
was given the chance to present my master’s
research to the conference attendees. By taking
part in the Earth Model Award program and
attending LIFE2019, I had a fantastic opportunity
to see the work Halliburton is doing, allowing
me to appreciate the value of industry-academic
collaboration in solving the challenges facing
the oil and gas industry. I would like to thank
the organizers of the award and all Halliburton
staff, who were incredibly helpful, friendly, and
supportive throughout the process.

~ James Lovell-Kennedy, University of
Manchester
It was a great honor to win the 2018 Earth
Model Award and to present my research at the
Halliburton LIFE2019 conference. This award
affirms that I am on the right path, pursuing what I
love and am passionate about. I was also pleased
that UT Austin and my advisors were recognized
through this award. My thesis was a group effort
that required a tremendous amount of time and
energy on behalf of everyone involved.
I had a memorable time attending LIFE2019
and the award ceremony. The conference was
highly informative and well-organized. I enjoyed
the opportunity to meet and socialize with
professionals in the oil and gas industry, and to
celebrate the great achievements of the other Earth
Model Award winners. I want to thank Halliburton
for sponsoring the event and rewarding the next
generation of geoscientists for their hard work.

~ Landon Lockhart, University of Texas