(Sustainable Energy Engineering) Fred Aminzadeh - Hydraulic Fracturing and Well Stimulation (Sustainable Energy Engineering) - Wiley-Scrivener (2019)

Hydraulic Fracturing
and
Well Stimulation
Scrivener Publishing
100 Cummings Center, Suite 541J
Beverly, MA 01915-6106
Publishers at Scrivener
Martin Scrivener (martin@scrivenerpublishing.com)
Phillip Carmical (pcarmical@scrivenerpublishing.com)
Book Series
Sustainable Energy Engineering
Hydraulic Fracturing and
Well Stimulation
Edited by
Fred Aminzadeh
This edition first published 2019 by John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA
and Scrivener Publishing LLC, 100 Cummings Center, Suite 541J, Beverly, MA 01915, USA
© 2019 Scrivener Publishing LLC
For more information about Scrivener publications please visit www.scrivenerpublishing.com.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or other-
wise, except as permitted by law. Advice on how to obtain permission to reuse material from this title
is available at http://www.wiley.com/go/permissions.
Wiley Global Headquarters

111 River Street, Hoboken, NJ 07030, USA
For details of our global editorial offices, customer services, and more information about Wiley prod-
ucts visit us at www.wiley.com.
Limit of Liability/Disclaimer of Warranty

While the publisher and authors have used their best efforts in preparing this work, they make no rep-
resentations or warranties with respect to the accuracy or completeness of the contents of this work and
specifically disclaim all warranties, including without limitation any implied warranties of merchant-
ability or fitness for a particular purpose. No warranty may be created or extended by sales representa-
tives, written sales materials, or promotional statements for this work. The fact that an organization,
website, or product is referred to in this work as a citation and/or potential source of further informa-
tion does not mean that the publisher and authors endorse the information or services the organiza-
tion, website, or product may provide or recommendations it may make. This work is sold with the
understanding that the publisher is not engaged in rendering professional services. The advice and
strategies contained herein may not be suitable for your situation. You should consult with a specialist
where appropriate. Neither the publisher nor authors shall be liable for any loss of profit or any other
commercial damages, including but not limited to special, incidental, consequential, or other damages.
Further, readers should be aware that websites listed in this work may have changed or disappeared
between when this work was written and when it is read.
Library of Congress Cataloging-in-Publication Data
ISBN 978-1-119-55569-8
Cover image: Oil Rig - Luchschen | Dreamstime.com

Cover design by Kris Hackerott
Set in size of 11pt and Minion Pro by Manila Typesetting Company, Makati, Philippines
Printed in the USA
10 9 8 7 6 5 4 3 2 1
Contents
Foreword xiii
Part 1: Introduction 1
1 Hydraulic Fracturing, An Overview 3
Fred Aminzadeh
1.1 What is Hydraulic Fracturing? 4
1.2 Why Hydraulic Fracturing is Important 5
1.3 Fracture Characterization 8
1.4 Geomechanics of Hydraulic Fracturing 11
1.5 Environmental Aspects of Hydraulic Fracturing 14
1.6 Induced Seismicity 18
1.7 Case Study: Fracturing Induced Seismicity in California 23
1.8 Assessment of Global Oil and Gas Resources Amenable
for Extraction via Hydraulic Fracturing 27
1.9 Economics of HF 27
1.10 Conclusions 28
Acknowledgement 30
References 30
Part 2: General Concepts 35

2 Evolution of Stress Transfer Mechanisms During Mechanical
Interaction Between Hydraulic Fractures and Natural Fractures 37
Birendra Jha
2.1 Introduction 37
2.2 Physical Model 39
2.3 Mathematical Formulation 40
2.4 Numerical Model 43
2.5 Simulation Results 44
v
vi Contents
2.6 Effect of Hydraulic Fracturing on Natural Fractures 46

2.7 Conclusion 49
References 50
3 Primer on Hydraulic Fracturing Concerning Initiatives
on Energy Sustainability 53
Michael Holloway and Oliver Rudd
3.1 Hydraulic Fracturing 54
3.1.1 Environmental Impact – Reality vs. Myth 54
3.1.2 The Tower of Babel and How it Could be the Cause
of Much of the Fracking Debate 55
3.1.3 Frac Fluids and Composition 57
3.1.4 Uses and Needs for Frac Fluids 57
3.1.5 Common Fracturing Additives 58
3.1.6 Typical Percentages of Commonly Used Additives 60
3.1.6.1 Proppants 61
3.1.6.2 Silica Sand 63
3.1.6.3 Resin Coated Proppant 65
3.1.6.4 Manufactured Ceramics Proppants 65
3.2 Additional Types 66
3.3 Other Most Common Objections to Drilling Operations 66
3.3.1 Noise 67
3.4 Changes in Landscape and Beauty of Surroundings 68
3.5 Increased Traffic 69
3.6 Chemicals and Products on Locations 70
3.6.1 Material Safety Data Sheets (MSDS) 72
3.6.1.1 Contents of an MSDS 73
3.6.1.2 Product Identification 73
3.6.1.3 Hazardous Ingredients of Mixtures 74
3.6.1.4 Physical Data 74
3.6.1.5 Fire & Explosion Hazard Data 75
3.6.1.6 Health Hazard Data 76
3.6.1.7 Reactivity Data 76
3.6.1.8 Personal Protection Information 77
3.7 Conclusion 77
Bibliography 78
Contents vii
4 A Graph Theoretic Approach for Spatial Analysis of Induced

Fracture Networks 79
Deborah Glosser and Jennifer R. Bauer
4.1 Background and Rationale 80
4.2 Graph-Based Spatial Analysis 83
4.2.1 Acquire Geologic Data and Define Regional
Bounding Lithology 84
4.2.2 Details of the Topological Algorithm 85
4.2.2.1 Data Acquisition, Conditioning and Quanta 85
4.2.2.2 Details of the k-Nearest Neighbor Algorithm 86
4.2.3 The Value of the Topological Approach Algorithm 86
4.3 Real World Applications of the Algorithm 87
4.3.1 Bradford Field: Contrasting the Graph-Based
Approaches; k Sensitivity 87
4.3.1.1 Data Sources 88
4.3.1.2 Results 88
4.3.2 Armstrong PA: Testing the Algorithms Against a
Known Leakage Scenario 88
4.3.2.1 Data Sources 90
4.3.2.2 Results 90
4.4 Discussion 91
4.4.1 Uses for Industry and Regulators 93
4.5 Conclusions 93
Acknowledgements 94
References 94
Part 3: Optimum Design Parameters 99

5 Fracture Spacing Design for Multistage Hydraulic Fracturing
Completions for Improved Productivity 101
D. Maity, J. Ciezobka and I. Salehi
5.1 Introduction 101
5.2 Method 103
5.2.1 Impact of Natural Fractures 104
5.2.2 Workflow 107
5.2.3 Model Fine-Tuning 108
5.2.4 Need for Artificial Intelligence 109
viii Contents
5.3 Data 110

5.4 Results 114
5.4.1 Applicability Considerations 120
5.5 Concluding Remarks 121
Acknowledgement 122
References 122
6 Clustering-Based Optimal Perforation Design
Using Well Logs 125
Andrei S. Popa, Steve Cassidy and Sinisha Jikich
6.2 Objective and Motivation 127
6.3 Technology 128
6.4 Clustering Analysis 129
6.4.1 C-Means (FCM) Algorithm 130
6.5 Methodology and Analysis 131
6.5.1 Available Data 131
6.6 Applying the FCM Algorithm 134
6.7 Results and Discussion 136
6.8 Conclusions 139
Acknowledgements 139
References 139
7 Horizontal Well Spacing and Hydraulic Fracturing Design
Optimization: A Case Study on Utica-Point Pleasant Shale Play 141
Alireza Shahkarami and Guochang Wang
7.2 Methodology 143
7.2.1 The Base Reservoir Simulation Model 143
7.3 Optimization Scenarios 147
7.4.1 Base Reservoir Model – A Single Well Case 148
7.4.2 Multi-Lateral Depletion – Finding the Optimum
Number of Wells 148
7.4.3 Completion Parameters 151
7.4.4 Second Economic Scenario, Reducing the Cost
of Completion 153
7.5 Conclusion 154
Acknowledgments 156
Contents ix
Part 4: Fracture Reservoir Characterization 159

Ahmed Ouenes
Introduction 159
References 161
8 Geomechanical Modeling of Fault Systems Using the Material
Point Method – Application to the Estimation of Induced
Seismicity Potential to Bolster Hydraulic Fracturing
Social License 163
Nicholas M. Umholtz and Ahmed Ouenes
8.2 The Social License to Operate (SLO) 165
8.3 Regional Faults in Oklahoma, USA and Alberta, Canada
used as Input in Geomechanical Modeling 166
8.4 Modeling Earthquake Potential using Numerical Material
Models 168
8.5 A New Workflow for Estimating Induced Seismicity
Potential and its Application to Oklahoma and Alberta 173
8.6 The Benefits of a Large Scale Predictive Model and Future
Research 178
8.7 Conflict of Interest 179
References 179
9 Correlating Pressure with Microseismic to Understand
Fluid-Reservoir Interactions During Hydraulic Fracturing 181
Debotyam Maity
9.2 Method 182
9.2.1 Pressure Data Analysis 182
9.2.2 Microseismic Data Analysis 186
9.3 Data 187
9.4 Results 188
9.4.1 Pitfalls in Analysis 196
9.5 Conclusions 196
9.6 Acknowledgements 197
References 197
x Contents
10 Multigrid Fracture Stimulated Reservoir Volume Mapping

Coupled with a Novel Mathematical Optimization Approach
to Shale Reservoir Well and Fracture Design 199
Ahmed Alzahabi, Noah Berlow, M.Y. Soliman
and Ghazi AlQahtani
10.2 Problem Definition and Modeling 203
10.2.1 Geometric Interpretation 203
10.2.1.1 Fracture Geometry 203
10.2.2 The Developed Model Flow Chart 204
10.2.3 Well and Fracture Design Vector Components 204
10.3 Development of a New Mathematical Model 204
10.3.1 Methodology 207
10.3.2 Objective Function 207
10.3.3 Assumptions and Constraints Considered
in the Mathematical Model 207
10.3.3.1 Sets 208
10.3.3.2 Variables 208
10.3.3.3 Decision Variables 208
10.3.3.4 Extended Sets 208
10.3.3.5 Constant Parameters 209
10.3.3.6 Constraints 209
10.3.4 Stimulated Reservoir Volume Representation 210
10.3.5 Optimization Procedure 211
10.4 Model Building 212
10.4.1 Simulation Model of Well Pad and SRV’s Evaluation 214
10.5 Results and Discussions 216
10.6 Conclusions and Recommendations 216
References 218
Appendix A: Abbreviations 220
Appendix B: Definition of the Fracturability Index Used
in the Well Placement Process 220
Appendix C: Geometric Interpretation of Parameters Used
in Building the Model 221
11 A Semi-Analytical Model for Predicting Productivity of Refractured
Oil Wells with Uniformly Distributed Radial Fractures 227
Xiao Cai, Boyun Guo and Gao Li
11.2 Mathematical Model 229
11.3 Model Verification 231
Contents xi
11.4 Sensitivity Analysis 231

11.5 Conclusions 233
References 234
Appendix A: Derivation of Inflow Equation for Wells
with Radial Fractures under Pseudo-Steady State Flow
Conditions 235
Part 5: Environmental Issues of Hydraulic Fracturing 243

Introduction 243
References 245
12 The Role of Human Factors Considerations and Safety Culture
in the Safety of Hydraulic Fracturing (Fracking) 247
Jamie Heinecke, Nima Jabbari and Najmedin Meshkati
12.2 Benefits of Hydraulic Fracturing 250
12.3 Common Criticisms 250
12.4 Different Steps of Hydraulic Fracturing and Proposed
Human Factors Considerations 252
12.5 Hydraulic Fracturing Process: Drilling 254
12.6 Hydraulic Fracturing Process: Fluid Injection 257
12.7 Fracking Fluid 258
12.8 Wastewater 258
12.9 Human Factors and Safety Culture Considerations 259
12.9.1 Human Factors 259
12.9.1.1 Microergonomics 260
12.9.1.2 Macroergonomics 260
12.9.2 Safety Culture 261
12.10 Examples of Recent Incidents 263
12.11 Conclusion and Recommendations 265
Acknowledgment 266
References 266
13 Flowback of Fracturing Fluids with Upgraded Visualization
of Hydraulic Fractures and Its Implications on Overall Well
Performance 271
Khush Desai and Fred Aminzadeh
13.2 Assumptions 272
13.3 Upgraded Visualization of Hydraulic Fracturing 273
xii Contents
13.3.1 Concept 273

13.3.2 Results 274
13.4 Reasons for Partial Flowback 275
13.4.1 Fracture Modelling 275
13.4.2 Depth of Penetration 276
13.4.3 Closing of Fractures 277
13.4.4 Chemical Interaction of Fracturing Fluids 277
13.5 Impact of Parameters under Control 278
13.6 Loss in Incremental Oil Production 279
13.7 Conclusions 280
13.8 Limitations 281
References 281
Appendix A 282
14 Assessing the Groundwater Contamination Potential
from a Well in a Hydraulic Fracturing Operation 285
Nima Jabbari, Fred Aminzadeh and Felipe P. J. de Barros
14.2 Risk Pathways to the Shallow Groundwater 288
14.3 Problem Statement 289
14.4 Mathematical Formulation 290
14.5 Hypothetical Case Description
and the Numerical Method 291
14.7 Conclusion 297
References 298
Index 303
Foreword
Since its inception in 1947, hydraulic fracturing has had a greater influ-
ence on hydrocarbon production than any other technology. Drilling and
fracturing of horizontal wells starting in the late 1980’s enabled further
increases in production. The oil industry took the new technology to a
new level, creating the shale revolution, making the USA the world’s largest
producer of crude oil.
The advancement of technology also brought new challenges in tech-
nological, environmental, human, and geopolitical dimensions. Each chal-
lenge will have to be understood and addressed. This book collects papers
that address the environmental and human dimensions in simple terms in
order to clarify issues and dispel misconceptions.
Exploring the technological dimension, the book includes chapters that
address many of the issues that are very unique to unconventional reser-
voirs, including chapters on Reservoir-Fluid interaction, optimization of
fracture spacing, and optimization of fracture placement.
M. Y. Soliman, PhD, PE, NAI

Chair of the Petroleum Engineering Department
William C Miller Endowed Chair
The University of Houston
xiii
Part 1
INTRODUCTION
Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (1–33)

1
Hydraulic Fracturing, An Overview*
Fred Aminzadeh *
University of Southern California
Abstract
This article provides an overview of the state of the art in hydraulic fracturing, a
controversial topic of the last decade. To Frack or not to Frack, That is the Question;
was posed at a meeting of the Western Regional Society of Petroleum Engineers.
The fact is, we have witnessed an intense debate over hydraulic fracturing’s eco-
nomic benefits and the role it has played in securing energy independence, and
its ill effects (perceived or real) during the past decade. Many, in particular those
in the fossil energy industry, consider shale oil/gas, with the associated horizontal
drilling and hydraulic fracturing, as one of the major developments in the oil and
gas industry of the past two decades. Others, especially many environmentalists,
consider fracking proponents as public enemy number one.
Different sections of this entry attempt to highlight different scientific facts about
hydraulic fracturing, the common-sense environmental concerns, and the respective
economic ramifications. After a brief overview of the principles of hydraulic fractur-
ing in section “What Is Hydraulic Fracturing?”, we discuss the importance of hydrau-
lic fracturing in section “Why Hydraulic Fracturing Is Important.” This is followed
by fracture characterization (section “Fracture Characterization”) and geomechanics
(section “Geomechanics of Hydraulic Fracturing”). They examine the natural frac-
tures in the subsurface and how one can characterize them, how hydraulic fracturing
helps to expand the natural fractures and/or create new (stimulated) fractures as well
as the underlining rock mechanics properties and the related stress regime.
Section “Environmental Aspects of Hydraulic Fracturing” addresses different
environmental concerns about hydraulic fracturing. This includes the potential
ground water contamination, amount of water used for hydraulic fracturing the
water disposal process, and methane emission concerns. Another environmental
concern is the risk of induced seismicity or man-made earthquakes. Given the
Email: faminzad@usc.edu
*Adopted from Aminzadeh, F., 2018, Hydraulic Fracturing, An Overview, Journal of Sustainable
Energy Engineering, Vol. 6, Issue 4, pp. 204–228.

3
4 Hydraulic Fracturing and Well Stimulation
publicized controversy on whether and to what extent hydraulic fracturing cre-

ates induced seismicity section “Induced Seismicity” is dedicated to deal with this
issue. The key message of this article is the best way to answer the question with
which we began, namely to frack or not to frack, lies with science; the hope is, with
sound scientific and engineering investigation, truth will prevail - “veritas omnia
vincit.”
Keywords: Hydraulic fracturing (HF), shale resources, energy resources,

micro-seismic data, economic factors, environmental impact of HF,
fracture characterization, geomechanics, induced seismicity, ground water
contamination, stress field
1.1 What is Hydraulic Fracturing?

Hydraulic fracturing (HF) is an oil and gas operation used to recover hydro-
carbon resources that are trapped in low-permeability shale and other
lithologies. Over geologic time periods, these resources were formed by
the maturation of kerogen, the organic precursor of petroleum. However,
unlike conventional oil and gas where hydrocarbons migrate into the res-
ervoir from a separate source, the hydrocarbon source and reservoir rock
share low permeability, forming an unconventional system. While there
are significant volumes of hydrocarbon trapped in unconventional reser-
voirs, the extremely low natural permeability of the formations impedes
commercial production using conventional techniques. HF is a process
that involves injection of large volumes of water (several million gallons),
sand, and small volumes of chemical additives to increase oil or natural gas
flow from low permeability formations. The large pressure associated with
injection of “fracturing fluid” creates new fractures and extends existing
fractures that enhance hydrocarbon flow, while sand mixed with injected
fluid holds the new and existing fractures open. Some of the injected fluid
flows back to the wellbore and is pumped to the surface, or is injected back
to the reservoir. Figure 1.1 is an example of a typical HF configuration.
HF is usually performed on horizontal or directional wells, oftentimes
with the well azimuth being perpendicular to the direction of maximum
horizontal in situ stress. Figure 1.1 shows a schematic display of the hor-
izontal well and the enlarged shale fractures are shown in the bottom left
part of Figure 1.1 display of the horizontal well and the enlarged shale frac-
tures are shown in the top right with the multistage (3 stages) fracturing.
Although HF has been used since the 1950s, over the last decade it has
been the subject of intense public debate. Some of the concern has been
over its potential impacts on drinking water, the potential for emission of
Hydraulic Fracturing, an Overview 5
Private well
USDW
Municipal water well:
<1000 ft.
Shale fractures
Additonal steel
casings and cement
to protect
groundwater
Protective steel casing
(Not to scale) Approximate distance

from surface: 6000 feet
Steel casing lines the well and is cemented in place to prevent any communication up the wellbore as the fracturing job is pumped
or the well is produced. Shallow formations holding fresh water that may be useful for farming or public consumption are separated
from the fractured shale by thousands of feet of rock.
Figure 1.1 An example of a typical HF configuration [1].
gas, and the associated induced seismicity or manmade earthquakes. For

more on various aspects of HF debate and its benefits and drawbacks, see
HF 101 [2], California Council on Science and Technology (CCST) report
on well stimulation [3], and HF in Colorado [4].
1.2 Why Hydraulic Fracturing is Important

HF and horizontal drilling have played a key role in making shale a signif-
icant part of the fossil energy resources globally, especially in the United
States (US). According to the US Energy Information Administration [5],
95% of new US wells drilled in 2016 were hydraulically fractured. This
accounts for two-thirds of total US natural gas production half of US crude
oil production. While shale oil and gas has contributed substantially to US
reserves and greatly impacted US oil production (increasing by about 3.2
million barrels/day), it is sobering to note that its contribution to global
reserves is rather modest. US reserves of shale oil amount to 40 billion
barrels, an amount comparable to the annual global consumption of oil [6].
Thus, the availability of this resource does end the dominance of conven-
tional oil as the primary source of global energy, and it limits the political
and economic leverage major oil producing countries have enjoyed in the
past.
It should be noted that since 2015, due to lower oil and gas prices, we
have seen a noticeable decline in shale related activities, from exploration
to drilling and production. According to the Baker Hughes rig count data,
at peak in March 2015 there were 1900 active oil rigs operating in the var-
ious US basins [7]. Following the price collapse, the rig count dropped
to about 400 in June 2016 and then began climbing again in response to
higher oil prices. As of early 2018 the number of oil rigs stood at about
1000. The relative speed with which these operations can be turned on
and off has a profound influence on the ability of oil producers to raise oil
prices by limiting their own production.
The increased production of shale oil in the USA has made it easier
for the world oil market to withstand several incidents when production
and export from certain countries was halted from war or other political
upheavals [8].
Producing shale oil can cost upwards of $60/ barrel. Prior to 2006 oil pric-
ing hovered around $40/barrel, making HF cost prohibitive. Subsequently,
largely in response to increasing global demand, the oil price rose to above
$100/barrel (with some spikes as high as $140/barrel). At these higher
prices, shale oil production by HF was profitable and it spurred drilling
activities in many US basins. The profitability can vary significantly from
one play to the next. Figure 1.2 shows an earlier peak for shale production
for different US plays that arrived in 2015, after which it dipped through
2017, with the rate of decline being an indication of the price sensitivities
in different plays. Since 2017, production increased in all plays.
Modern HF combined with horizontal drilling allows multiple wells to
be drilled from one surface location, reducing the size of the drilling area
U.S tight oil production- selected plays

million barrels of oil per day
Eagle Ford (TX) 6.5
Bakken (ND & MT) 6.0
Spraberry (TX & NM Permian)
5.5
Bonespring (TX & NM Permian)
Wolfcamp (TX & NM Permian) 5.0
Delaware (TX & NM Permian) 4.5
Yeso & Glorieta (TX & NM Permian) 4.0
Niobrara-Codell (CO, WY)
3.5
Haynesville (LA, TX)
Utica (OH, PA & WV) 3.0
Marcellus (PA, WV, OH & NY) 2.5
Woodford (OK) 2.0
Granite wash (OK & TX)
1.5
Austin Chalk (LA & TX)
Monterey (CA) 1.0
0.5
0.0
2004 2006 2008 2010 2012 2014 2016 2018
Sources: EIA derived from state administrative data collected by DrillingInfo Inc. Data are
through September 2018 and represent EIA’s official tight oil estimates, but are not
survey data. State abbreviations indicate primary state (s).
Figure 1.2 The explosion of shale production since 2010 [5].

above ground by as much as 90%. Fracking is the key to unlocking vast US

shale resources, freeing up oil and natural gas that previously was inaccessi-
ble while protecting groundwater supplies and the environment. America’s
shale energy revolution began with extensive US government agencies and
contractors (e.g., Department of Energy, DOE; Gas Technology Institute,
GTI). This was further accelerated by privately financed efforts (e.g., oil
and gas operators and service companies). In both cases, the primary
driver was new technology.
The notion of fracturing rocks to improve permeability is an old one.
Based on an idea proposed by Floyd Farris of Stanolind Oil and Gas
Corporation (later, Amoco and then BP), attempts at producing oil in tight
formations were made by using mixtures of naphthenic acid and palm oil
(napalm). The process, “Hydrafrac,” was patented in 1949 and licensed to
Haliburton.
The Devonian shale basins of the Western Appalachian, Michigan, and
Illinois Basins were known to contain vast quantities of gas, but the low
permeability of the rock limited production from this resource. The declin-
ing US reserves of natural gas prompted the DOE and the GTI to sponsor
research and development efforts for technologies for assessing and pro-
ducing gas from this resource. These efforts led to techniques such as the
slick-water fracture, horizontal drilling, and microseismic technology for
characterizing and exploiting these resources. This included several DOE
supported demonstration projects in Ohio, Texas, and New Mexico to val-
idate these technologies (API 2018) [9].
The development of commercial fracking is largely due to the efforts of
Mitchell Energy, who drew on the government-funded work, and despite
numerous setbacks persisted for decades in improving the technologies
and driving down their costs. Of course, the rise in the cost of crude oil in
2008 helped in the process. Once hydraulic fracking became economically
feasible, it was widely deployed by many companies, and experiential learn-
ing drove down the costs. Unleashing natural gas from shale resulted in a
sharp decline of its cost from about $15/Mbtu in 2008 to around $3/Mbtu
in 2014. About one third of the natural gas is used for producing hydrogen,
which in turn is used industrially to refine fuels or produce fertilizers. The
large availability of natural gas at relatively low cost has allowed US refin-
eries to increase their output and increase exports of finished products.
While those against HF may argue otherwise, the industry proponents
maintain shale gas has also had a positive environmental impact. Since
natural gas is not easily exported, its increased availability in the US has
resulted in a decline in its price and made it the fuel of choice for electric
power production. Many US power producers switched from burning coal
to burning natural gas. In 2005, electricity generation from coal and gas
in the US amounted to 2,000 TWh and 760 TWh, respectively; by 2017
the two sources were each generating about 1,200 TWh, while the total
electricity generation has held steady at 4,000 TWh. Thus, over 700 TWh
of coal generation has switched to gas with the attendant environmental
benefit of reduced carbon dioxide (CO2) emissions.
The positive economic impact of shale-related natural gas and oil
resources is undeniable. By helping to lower power and materials costs,
as well as stimulating economic activity and positive impact on the labor
market for a variety of businesses like service and supply companies, HF
has supported growth across an economy that otherwise has struggled in
recent years. According to an IMF Report, the shale gas revolution has had
a macroeconomic impact between 0.3% and 1% of US GDP [10]. Many
performance factors are used to assess the effectiveness and safety of HF.
Among them are better understanding of fracture azimuth/geometry/
length/distribution, Stimulated Reservoir Volume, and potential hazards
(induced seismicity and HF fluid leakage). We will revisit the environmen-
tal implications of HF in a different section.
To summarize, one of the key factors impacting shale production is the
economics and oil and gas prices, as discussed earlier. However, there are
other factors that have impacted the economics of shale resources, includ-
ing the evolution of the technologies, including subsurface characterization
and advances in HF, ability to monitor creation of new fractures, and asso-
ciated microseismic events. The next two sections provide some insight on
the science behind HF and how the advances in the related science and
technology can further improve the economics of production from shale
and other tight formations.
1.3 Fracture Characterization

The basic idea behind new fracture generation is illustrated by Figure 1.3.
Induced seismicity and stimulated fractures are caused in most cases by
changes in pore fluid pressure and/or change in stress in the subsurface,
especially in the presence of (a) faults with specific properties and orienta-
tions and (b) a critical state of stress in the crust. The factor that appears to
have the most direct correlation in regard to stimulated fracture is the net
fluid balance or the total balance of fluid introduced into or removed from
the subsurface. Additional factors may also influence the way fluids affect
the subsurface.
σv
σh σh
σv
Normal fault
Figure 1.3 Generation of fractures due to application of horizontal and vertical stress and
fluid injection during HF.
Microseismic mapping associated with hydraulic fracture stimulations

is used to extract the fracture geometry information, such as heights and
length. Since a large amount of shear stress is created around a hydraulic
fracture, existing natural fractures shear and create microseismic events
that allow one to estimate the length and the height of the man-made frac-
ture [11]. The assessment of fracture complexity using microseismicity has
yielded good results since it reflects the general area where the natural frac-
tures were subjected to high shear stress and interacted with the hydrau-
lic fractures. It is almost impossible to determine the interaction between
created fractures with different levels of complexity. Furthermore, the
technology to differentiate between stimulated, unchanged, and damaged
fracture systems is still evolving. Microseismic imaging is an important
technology for tracking fracture creation and reactivation of preexisting
fractures occurring during various industrial operations, particularly
hydraulic fracture treatments of unconventional reservoirs. Often, micro-
seismic locations are used to interpret the fracture geometry, although
additional insights into the fracture deformations can also be extracted
through source characterization. The resulting source deformation inves-
tigations provide additional information about the fracturing and are
sometimes interpreted as representing the entire spectra of deformations.
Furthermore, new, unconventional, passive seismic imaging techniques
such as coherency-based stacking methods and emission tomography [12,
13], provide the potential for identification of low signal-to-noise ratio
(SNR) emissions such as smaller microseismic events or low frequency
creep failure [14].
In certain cases, the observed microseismicity is found to correlate to
shear deformation, although HR is generally considered a tensile fractur-
ing process. The corresponding timeframe is also tied to the relative high
frequency content of the seismic recording equipment and much more

rapid than the relative slow and hence aseismic tensile fracture dilation
occurring throughout the injection period. Fracture movements can also
potentially occur at some distance away from the HF associated with pres-
sure and/or pressure changes resulting from the fracturing. Although the
fracture geometry can often be inferred directly from the microseismicity,
attempts to understand the fracture deformation can be a challenge. Maity
and Aminzadeh [15] report on the use of Microseismic, 3D seismic, and
well log data combination to characterize subsurface fractures by devel-
oping a neural network-based “Fracture Zone Identifier” (FZI) attributes.
Equation 1 shows the relationship between different FZI attributes and the
porosity, different types of velocities and density. Such an implicit relation-
ship is established using a specially designed neural network. Figure 1.4
shows a three-dimensional display of subsurface characterization of the
natural fracture zones. Obviously, after HF, these fracture zones would be
altered and the newly created or expanded fractures can be characterized
using similar techniques.
With the aid of repeated microseismic measurement data and possibly
4D seismic, the FZI attribute can also be used to quantify changes in the
size and nature of the fractures and their growth. Other techniques such
as the use of statistical distributions in seismicity can also shed valuable
insights into the fracturing process and its relation to the structural fea-
tures within the reservoir. Such methods examine seismicity as a function
of the in situ stress state within the reservoir.
C B'
A
D'
400 C'
B A'
800
Depth (m)
1200
1600
2000
100
50 D 80
60
FZIn = F{фu,Zu,Vpu, Vsu, Pu, VEu,} 100 40
20
150
Figure 1.4 Fracture mapping using MEQ, seismic, & petrophysical data using ANN-
based hybrid FZI attribute mapping, from Maity and Aminzadeh [15].
1.4 Geomechanics of Hydraulic Fracturing

Geomechanical evaluation and modeling in the oil and gas industry is used
to predict reservoir parameters, such as bottom hole pressure, in situ rock
stresses, modulus of elasticity, leak-off coefficient, formation porosity, and
permeability. It is no coincidence that, with increased demand for HF, geo-
mechanical simulations have gained popularity as well. For example, such
evaluations allow the determination of reservoir response to fracturing
fluid injections. Starting with a discrete fracture network (DFN), one can
model both the aseismic and microseismic modes of deformation. This
helps interpret the resulting injection pressures, rates, volumes in conjunc-
tion with seismicity data, improving the interpretation of the stimulated
fracture networks.
Preexisting fractures and the in situ stress state have a significant bear-
ing on HF operations. In particular, hydraulic fracture propagation and
hydraulic vs. natural fracture interactions can be correlated with the
observed microseismicity distribution. Recent studies have shown strong
correlation between microseismicity distribution and in situ reservoir
properties such as faults and rock brittleness [16, 17]. Recent studies
involving ground truth measurements from the HF Test Site in the Permian
Basin [18] have shown that hydraulic fracture growth and complexity can
be significant and the propped rock is relatively small in comparison with
the overall stimulated rock [19]. New mathematical techniques are being
developed to model for such significant fracture propagation complexities.
Geomechanical models are usually very sensitive to input parameters,
including stress anisotropy, completion effectiveness to fracture initiation,
and the DFN characteristics. Correlating the microseismic data with the
corresponding geomechanical deformation would increase the confidence
in the geomechanical simulation and provides an important framework to
interpret the microseismicity as the geomechanical response of the reser-
voir. A linear elastic fracture mechanics (LEFM) model can describe stress
distribution around a single hydraulic fracture. The full stress field solution
is valid everywhere in the domain and is a function of stress boundary
conditions and fracture geometry.
Predicting and controlling the initiation and cessation of a hydraulic
fracture remains a challenge due to compositional and poromechanical
heterogeneity, which causes stress concentration, and due to inelasticity
at the grain scale. The presence of organic matter also affects the rockme-
chanical properties and directionality in fracture initiation and propaga-
tion processes. Overall, understanding the fracturing behavior of a rock
at the microscale plays a critical role in predicting the performance of HF.

Fracture initiation and propagation behaviors and estimation of fracture
toughness and directionality could be accomplished through the use of
microscale geomechanical scratch tests [20]. The fracture toughness and
directionality values of tight sandstone and shale samples are compared
to understand the effect of lithology on fracture initiation and propaga-
tion processes, respectively. This is followed by initializing the model in
a principal state of stress with the maximum principal compression of
9000 psi in the x-direction. The principal stresses in the y and z direction
are 6000 psi. There is a zero-displacement boundary condition on all the
boundaries. Figure 1.5 shows the discretized grid used for the simulation,
with the injection well placed in center and perforated in layer 3, which is
the middle layer in this five-layered model.
We discussed microseismic monitoring in the “Fracture Characterization”
section, providing the data on the effectiveness of stimulation (HF) oper-
ation. During fracture stimulation operations, a change of the in situ
stresses and rock volume generates microseismicity. Mapping the spatial
and temporal pattern of these microseismic events is used for tracking
the progress of HF propagation. Microseismic monitoring can also serve
as a complementary indicator of changes in fluid pathways and mechan-
ical changes in the treatment well. The deformation of the fracture wall
opening relates to the fluid pathway through P-waves. In addition, at the
fracture tips, the shear stress regime may create S-waves which represent
the extent of the fracture wall.
The primary data used for evaluation of microseismic events are waveform
measurements obtained from the surface, downhole receivers, or both. A
“hybrid” microseismic survey design, comprised of both surface and borehole
Figure 1.5 The process of generating a 3D mesh of the domain for two-way coupled flow
geomechanics [20].
geophones, allows for optimum coverage for effective characterization of the

subsurface [21]. These data provide information related to strain and stress
changes in the reservoir as well as the nature of the physical processes such as
induced fracturing of the rock or slippage on preexisting fractures. Waveform
measurements can also indicate undesired fracture growth or fault activation.
The processed microseismic waveform recording describes the mag-
nitude and direction of the rock movements. Vector and spectral fidelity
define the accuracy of the measuring content, leading to the construction
and calibration of a model of seismic P-wave and S-wave. The accuracy of
microseismic interpretation depends on estimated event hypocenters, the
accuracy of the velocity model, and the monitoring geometry. Thus, low
signal-to-noise ratio is one of the greatest challenges for the processing of
microseismic data. Indirectly, field and surface deployment challenges can
be significant and generally act as the primary constraint in survey geom-
etries and resulting detectability.
One of the most important concepts in analyzing microseismic data is
the seismic moment, energy, and magnitude which are used to character-
ize the size of microseismic events, incremental fracture surface areas, and
lengths. This will lead to the construction of the effective stimulated vol-
ume (ESV) providing the information on the complexity of the hydraulic
fracture network. Consequently, seismic moment or cumulative moment
can provide insights into fracture behavior during stimulations. Energy-
based approach is used to calibrate fracture model with microseismic data.
The approach presents an energy-based numerical algorithm which stim-
ulates a vertical hydraulic fracture by parameter selection of elastic, stress,
and material properties defined in a layered 3D geologic medium [22].
To provide the information about the mechanism of failure at fracture
sites, the moment tensor inversion (MTI) concept can be used. MTI describes
the orientation, magnitude, and slip of individual microseismic events. The
decomposition of each moment tensor into different components provides
us with the estimation of the different failure mode: shear slip, tensile open-
ing, expansion, and their corresponding ratios. It also gives other processes
the orientation of local fracture planes and the direction of shear slip. MTI
concept allows constructing DFNs, representing the distribution, orientation
shape, connectedness, and fluid properties of hydraulically induced fractures
[23]. In addition, they can also detect complex fracturing behavior such as
fault activation or fracture propagation alongside bedding plans [24].
Given an estimate of the fracture network in the subsurface, our mod-
els will ascertain whether a particular Subsurface Fluid Injection and
Production (SFIP) strategy will cause failure at a particular site. A more
accurate statement would refer to the probability of failure occurring at a
given site, reflecting on the observation that this inference is predicated on

a particular (simplified) model and a finite set of data, introducing uncer-
tainty in the calibration and ensuing prediction. Hazard maps are them-
selves a function of the selected model and data, and their sensitivity with
respect to additional data and/or model refinement can be used to design
data acquisition efforts and model refinement through multiscale model-
ing. Accordingly, a hierarchical probabilistic model should be considered,
where its parameters reflect subscale effects, and whose sampling distribu-
tions can be accurately estimated from data. Sensitivity of the hazard maps
with respect to these parameters will then reflect the value of additional
measurements and additional model complexity.
1.5 Environmental Aspects of Hydraulic Fracturing

HF has been the subject of intense debate for its potential environmental risk
factors. This has prompted much scrutiny of HF by regulators and the gen-
eral public. Many in the environmental community believe biodiversity may
suffer as a result of land development and water supply damage. Methane
leakage may be much higher than previously reported by the EPA [25] and
this has severe implications for the climate. Fracking could exacerbate the
issue, which outweighs the potential CO2 pollution reduction by using clean
gas in place of oil/coal. Also, magnitude 2+ earthquakes have increased in
frequency in Texas and Oklahoma as well as the more common induced
seismicity below magnitude 2, although much of the seismicity is attributed
to the waste water injection not to HF directly. While it is difficult to predict,
the ill effects of HF could be mitigated by research and policy in changes.
The four main areas of concern are considered to be (a) ground water
contamination, (b) too much water used in the process and its disposal
process, (c) methane emission, and (d) induced seismicity. The first three
issues will be addressed in this section, while the induced seismicity related
concerns will be the subject of section “Induced Seismicity.”
(a) Ground-Water Contamination.

The CCST conducted an in-depth study of the issue and
released three-volume CCST Independent Report in
2016 titled: Advanced Well Stimulation Technologies
in California. While there are several key findings, the
general conclusions are: “There are no publicly reported
instances of potable water contamination from subsur-
face releases in California Well stimulation technologies,
as currently practiced in California, do not result in

a significant increase in seismic hazard.” Overall, in
California, for industry practice of today, the direct envi-
ronmental impacts of well stimulation practice appear to
be relatively limited. It should be emphasized that these
conclusions may not be easy to generalize elsewhere for
various reasons, including: Difference in geology and res-
ervoir parameters, difference in the volume of HF jobs
in California compared to many other States with more
extensive shale production, and stricter California State
regulations than those of Federal and many other States.
Some of these studies have been used by the American
Petroleum Institute (API) and other oil and gas indus-
try advocates to point out no additional regulations on
HF are required and many of the existing ones should be
scaled back. For example, API argues its ongoing work-
shop series “Commitment to Excellence in HF” is one
of the tools that the oil and natural gas industry uses to
reinforce with regulators, remind lawmakers, and educate
the public on industry’s commitment to and leadership on
safety, health, and environmental protection. Its website
continuously revises the standards related to HF. They can
be seen on the HF section of API’s website [9]. Others may
argue that many compounds used for HF may be harm-
ful, especially if enough care is not given to avoid entry
of the fracking fluid to the water column. Among com-
pounds used in HF operation are: acids, sodium chloride,
polyacrylamide, ethylene glycol, borate salts, sodium/
potassium carbonate, glutaraldehyde, guar gum, citric
acid, and isopropanol. Each of these compounds is used
for specific purposes such as helping to dissolve minerals
and initiating fissures in rock before fracturing, delaying
breakdown of the gel polymer chains, minimizing the
friction between fluid and pipe, preventing scale depos-
its in the pipe, maintaining fluid viscosity with tempera-
ture increase, ensuring effectiveness of other components,
eliminating bacteria in the water, thickening the water for
sand suspension, preventing precipitation of metal oxides,
and increasing the viscosity of the fracture fluid.After the
wells on a pad are drilled, cased, and cemented, the hori-
zontal part of the production pipe is perforated with small
holes in the casing, to expose the wellbore to the shale.

Then, fracking fluid is pumped into the well under high
pressure to create microfractures in the shale and free
the natural gas or oil. The sand in fracking fluid keeps
the fractures open after the pressure is released, and the
chemicals are chief agents to reduce friction and prevent
corrosion. Federal and State regulatory agencies normally
provide guidelines for energy development in the United
States, including HF. Among such regulations related to
shale resources are: Clean Water Act, Clean Air Act, Safe
Drinking Water Act, National Environmental Policy Act,
Resource Conservation and Recovery Act, Emergency
Planning and Community Right to Know Act. Some of
these regulations have recently been under re-assessment
and possibly revisions since mid-2017. Debate contin-
ues on whether, and to what extent, HF could be found
responsible for ground water contamination. It is under-
stood that a healthy debate should be based on science and
best practices in engineering as well extensive modeling
and simulation, combined with field monitoring. Some
examples of such research and development (R&D) efforts
are in [26] and [27]. Figure 1.6 shows an example of such
work. Recent studies at the HF Test Site in Midland Basin
highlight the minimal environmental impact of large
Conceptual Model Simulation Result Hazard Index

S˚ =0.3
w –5
Cumulative Distribution Function
1 x 10
Injection Monitoring Aquifer Boundary
0.9 2
Well Well
0.8
0.7 1.5
Vadose Zone 0.6
C/C0
K
1
0.5 1
Aquifer 0.4
0.9
Impervious Layer 0.3 0.5 0.8 99th %ile= 9.57
0.7 95th %ile= 6.03
0.2 90th %ile= 5.12
0.1 0 2 0 0.6 50th %ile= 0.82
10 10
Time |y| 0.5
F
Sand Layer S˚ =0.3 0.4

w x 10–5
1 0.3
Monitoring Well
0.9 2
0.2
0.8
0.1
0.7 1.5
0
0.6 0 1 2 3 4 5 6 7 8 9 10
C/C0
K
0.5 1 HI
0.4
Shale Reservoir 0.3 0.5
0.2
0.1 0 2 0
10 10
Time |y|
Figure 1.6 Numerical modeling to quantify risk factors for water contamination
associated with HF, Jabbari et al. [26].
36.5°N
1cm 1cm
36.3°N
(a)
36.1°N (c)
1cm
Parkfield
35.9°N
35.7°N
(b) (d) (e)
35.5°N 120.8°W 120.6°W 120.4°W 120.2°W 120°W
Figure 1.7 A comparison of induced seismicity in a field and laboratory setting with the
associated faulting.
pad scale fracturing operations on ground water and air

quality [28]. During this test, which involved fracturing
of eleven new horizontal wells and two re-fractured wells,
there was no migration of hydrocarbon or fracturing fluid
into the nearby Edward-Trinity Plateau aquifer.
(b) Amount of Water Used.
A recent report by Ceres highlighted the concern that
water used for fracking may be depleting water resources,
particularly in arid regions, and thus exacerbating the
water shortage [29]. Fracking of wells typically requires
the use of several million gallons of water. However, the
amount of water per se is not very large when compared
with other uses of water. In the ten years from 2005 and
2014, fracking operations in Texas—the state with most
frack wells—used 250 billion gallons of water, or on aver-
age about 25 billion gallons a year [30]. However, over-
all annual water usage in Texas is 22 million acre-feet, or
roughly 7,200 billion gallons of water. In other words, the
water used for fracking in Texas represents less than 0.5%
of water used each year. Viewed from this perspective,
water use for fracking does not appear to cause shortages.
Often though, the challenge with water use arises from the
fact that it must be sourced locally – trucking in millions
of gallons of water is expensive. Thus, if local groundwater

supplies are limited, fracking operations could aggravate
the water shortage. Other efforts to minimize the negative
impact of HF due to the excessive amount of water usage,
especially in the areas with severe water shortage prob-
lems, include recycling the waste water and HF fluid [31].
(c) Methane Emissions.
Similar concerns have been raised on the potential emis-
sions of CO2 from power generation and methane emission
from HF. Others argue such emissions are at their lowest
point in nearly 30 years, in large part due to the widespread
use of natural gas. Furthermore, it is argued that methane
emissions from the oil and natural gas industry make up
just 4% of total US greenhouse gas emissions. Nevertheless,
the debate can be settled by installing monitoring stations
in major HF operations to provide an accurate assessment
of the potential risk and develop operational procedures to
minimize and mitigate such risk. For additional details on
the Methane emission and other air pollution impacts of
HF, see the work by different US national laboratories and
others [32].
1.6 Induced Seismicity

The potential for earthquake-like activity induced due to production of oil
and natural gas has become a topic of recent interest, owing to environ-
mental concerns. The National seismic hazard maps in the United States
are given by Petersen et al. [33]. It has been suggested that subsurface injec-
tion and production activities might factor into the risk of induced seis-
micity. The extensive use of HF to stimulate production from shale plays
has led many to speculate that the rise in hydraulic stimulation operations
may have sparked a concurrent increase in seismicity. As the majority of
mapped faults in California are considered by the US Seismic Hazard Map
to be of higher risk, understanding the potential seismic sensitivity of the
region to hydraulic stimulation takes on greater importance.
There are, however, positive aspects. The material below is extracted
from Frequently Asked Questions (FAQ) related to Induced Seismicity
that has been released for public knowledge. It was also provided to the
California Council on Science and Technology (CCST) that sought liter-
ature and data for reviews of well stimulation technologies in California
and ISC submitted its recent study on “HF and Induced Seismicity in
California-A case study” [34].
a. What is induced seismicity?

Induced seismicity is seismic activity that occurs as a result of human
activity (anthropogenic). Seismic activity refers to the frequency, type, and
size of earthquakes. The vast majority of earthquakes are natural, caused by
stresses that cause fracturing of the rock in the earth’s crust. Examples of
processes that might cause induced seismicity include subsurface wastewa-
ter injection, geothermal energy generation, and surface-water reservoir
impoundment. Most seismic events related to induced seismicity cannot
be felt by humans because they have very low magnitudes (less than 2 on
the Richter scale) that can only detected by specialized instruments.
b. What causes induced seismicity?

The main cause of induced seismicity is increased fluid pressure in rock
pores that reduces natural friction and allows slippage of rocks along a
fault. Researchers investigating causes of induced seismicity have docu-
mented that fluid pressures have a role in seismicity. As explained by the
DOE, “pore pressures act against the weight of rock and forces holding
the rock together; if the pore pressures are low, then only the imbalance of
natural in situ earth stresses will cause an occasional earthquake. If, how-
ever, pore pressures increase, then it would take less of an imbalance to
cause an earthquake.” A recent study using data from hundreds of hydrau-
lically fractured wells in sedimentary basins in Canada has shown a strong
spatial correlation between induced earthquakes and high pore-pressure
gradients [35]. Cases have recently been observed in Oklahoma and Ohio,
where seismic activity has been associated with the disposal of large vol-
umes of wastewater into deep geologic formations. Such injection can
disrupt the balance of stress within the rock, especially when not paired
with corresponding fluid withdrawal. Other potential sources of induced
seismicity include underground mining and chemical interaction of fluids
with rock materials.
c. Can HF cause “induced seismicity”?

HF is a well stimulation practice that uses the injection of fluid to open flow
channels in tight formations to produce the hydrocarbons locked within
them. The practice has been in use for over sixty years. A report published
in 2013 by the National Academies of Science, NAS [36], concluded that
HF does not pose a high risk for inducing seismic events that could be felt
by humans. The U.S. Geological Survey (USGS) also found that HF is not
a substantial cause of induced seismicity, and that “only very rarely” is

HF the cause of any earthquakes that can be felt by humans. The “micro-
earthquakes” that HF creates are too small to be felt or cause structural
damage. HF has been associated with only minor seismic events at two U.S.
locations, neither in California. Evidence of fault activation from HF has
been observed in the past [37], but these rarely cause earthquakes of signif-
icant magnitude to impact humans or structures. The largest known earth-
quake sequence was observed in Western Canada, where a fault system
connected with the basement rock and took fluid during injection asso-
ciated with HF operations [38]. Accordingly, the DOE has also concluded
that HF is “rarely, if ever, a hazard when used to enhance permeability in
oil and gas or other types of fluid-extraction activities.” Furthermore, the
USGS finds that “although the disposal process has the potential to trig-
ger earthquakes…very few of the more than 30,000 wells designed for this
purpose appear to cause earthquakes.”
d. How can one determine whether a seismic event is natural or induced?

Ellsworth [39] addresses several issues in connection with natural versus
induced seismicity. Davis and Frohlich [40] set forth a simple set of yes or
no questions that could be used as a simple diagnostic tool in determining
whether a seismic event was likely induced. The questions were as follows:
i. Are these events the first known earthquakes of this char-

acter in the region?
ii. Is there a clear correlation between injection and seismicity?
iii. Are the epicenters near wells (within five kilometers)?
iv. Do some earthquakes occur at or near injection depths?
v. If not, are there known geologic structures that may chan-
nel flow to sites of earthquakes?
vi. Are changes in fluid pressure at well bottoms sufficient to
encourage seismicity?
vii. Are changes in fluid pressure at hypocentral locations suf-
ficient to encourage seismicity?
From these, an event may be scored with more “yes” answers indicating
an increased likelihood of an event having been induced.
e. Has HF caused earthquakes in California?

A recent study by the USC Induced Seismicity Consortium (ISC) found
no significant correlation between HF and induced seismic activity in
California. This study evaluated 30 years of seismic data in areas of oil pro-
duction where HF has been used in California and found little or no cor-
relation between oil field activities in general and seismic activity.
The ISC study compared seismic activity from 1980 to 2013 with oil
field activity, including HF, at locations throughout California. Oil field
activities and well locations were obtained from the California Division
of Oil, Gas, and Geothermal Resources (DOGGR) and FracFocus [41].
Seismic activity data was compiled from the Northern and Southern
California Earthquake Centers and included epicenter locations of seismic
activity and the magnitudes and depths of seismic events. Database maps
were prepared to display oil and gas well locations, well type (e.g. oil or
gas producer, water flood injector, steam flood injector, water disposal),
type of activity (drilling, producing, injecting, completing, stimulating, or
abandoning), and the depth of the activity. For more details, see Section 8
where a couple of case histories are highlighted.
Composite maps were prepared, comparing locations of seismic events
and known geologic faults with oil and gas well locations and oil field
activities. Statistical analyses were conducted to determine whether there
was any correlation between seismic events and oil field activities. Little or
no correlation was found. For example, in northern California, a total of
303,609 seismic events were recorded from 1980–2013. The main area of
northern California where oil and gas activities occurred is the Sacramento
area, far from known fault zones, where oil is produced from shallow depths
(less than 6 kilometers). Of the total 303,609 seismic events in northern
California, only 210 events were in the Sacramento area and only 3 seismic
events had magnitudes greater than 3; none were greater than 4.
f. Does induced seismicity from wastewater injection by the oil industry differ
from California in other states?
The practice of deep wastewater injection is commonly used in states
throughout the U.S. and is strictly regulated by state and federal laws.
California has strict regulations governing subsurface wastewater injec-
tion. Injection wells used by the oil industry in California are different
from the injection disposal wells linked to earthquakes in other states like
Ohio, Oklahoma, Texas, and Arkansas. In California oil fields, wastewa-
ter is reinjected back into the formation after the oil is removed. In other
areas of the U.S., wastewater is often injected into rocks with little poros-
ity and storage capacity. Pore pressures in the rock increases dramatically
as the limited available pore space is filled and additional injected water
causes the rock to break, causing an induced seismic event. As a result,
a University of California Santa Barbara geophysicist, Craig Nicholson,
concluded that “very little of the state’s [California] earthquake activity can
be tied in any way to reinjection…there’s not a connection like there is in
the central and eastern United States.”
g. How do we know if an earthquake was caused by induced seismicity?

Seismologists conduct extensive analyses of earthquake data and poten-
tial anthropogenic factors to determine whether an earthquake is natural,
triggered, or induced. More information is needed to improve researchers’
understanding of the “triggering” mechanisms of injection and production-
related induced seismicity as well as any irregularities that naturally
occurring earthquakes may have.
Recent studies by USC and the California Institute of Technology indi-
cate that there is not a significant correlation between earthquakes and oil
and gas field operations. A major reason for this is that natural earthquake
epicenters are typically at depths below five kilometers, while oil field oper-
ations are conducted at depths shallower than two kilometers, typically less
than one kilometer.
h. Are there any documented cases of injuries or property damage caused by

induced seismicity?
There have been no documented cases in California where injuries or
property damage have been caused by induced seismicity. However, prop-
erty damage has been documented from shaking thought to be a result of
wastewater injection into rock formations adjacent to known faults.
According to the National Research Council’s Committee on the
Induced Seismicity Potential in Energy Technologies Report [36], virtu-
ally all induced seismicity attributed to energy development has been small
in magnitude and unable to be felt. The same committee also concluded
that HF does not pose a high risk to triggering noticeable seismic activity.
Advanced HF and horizontal drilling are the technology engines driving
America’s ongoing energy renaissance – surging oil and natural gas pro-
duction that ranks first in the world. This oil and natural gas production,
enabled by HF, strengthens U.S. energy security, boosts the economy, and
lowers consumer energy costs. In addition, the increased use of cleaner-
burning natural gas is the main reason U.S. greenhouse gas emissions
from electricity generation are at their lowest level in nearly 30 years. For
decades, HF has been used safely thanks to proven engineering and effec-
tive industry risk management practices and standards, as well as federal
and state regulations.
Induced seismicity is a complex issue and the knowledge base sur-
rounding it is rapidly changing. A one-size-fits-all approach isn’t practical
because of the significant differences in local geology and surface condi-

tions, including population, building conditions, infrastructure, critical
facilities, and seismic monitoring capabilities. As such, state regulators are
best positioned to address potential issues linked to oil and gas injection
wells in their state.
States are developing diverse strategies for avoiding, mitigating, and
responding to potential risks as they locate, permit, and monitor Class
II disposal wells. Many state regulators work with experts from govern-
ment agencies, universities, private consultants, and industry experts on
these issues. Effective planning involves identifying where there’s risk of
harm from a seismic event because people and property are located nearby.
Again, state regulators are best able to make these assessments and plan
adaptive responses in the event of a quake, such as adding seismic mon-
itoring, adjusting injection rates and pressures, suspending injection well
operations, or halting injection altogether and shutting in a well.
Both HF and the underground disposal of produced waters from oil
and natural gas operations are proven to be safe and environmentally reli-
able. Industry, academia, and government entities are clearly committed
to pursuing further research to better understand the complex science and
physical mechanisms associated with induced quaking events.
1.7 Case Study: Fracturing Induced Seismicity

in California
This section includes highlights from a case study focused on establish-
ing a correlation between HF operations in California oilfields and seismic
activity. Regions with a significant level of hydraulic stimulation operations
were considered and HF activity was compared to records of seismic activ-
ity for the same period of investigation. A qualitative methodology is used
to differentiate potentially induced events from natural seismic events. The
materials presented here are adapted from Aminzadeh and Aminzadeh
et al. [42].
The potential for earthquake-like activity induced due to the production
of oil and natural gas has become a topic of recent interest, owing to envi-
ronmental concerns. It has been suggested that subsurface injection and
production activities might factor into the risk of induced seismicity. The
extensive use of HF to stimulate production from shale plays has led many
to speculate that the increase in incidences of hydraulic stimulation opera-
tions may have sparked a concurrent spike in seismicity (Ellsworth 2013).
As the majority of mapped faults in California fall in higher risk areas of
the US Seismic Hazard Map (Patterson et al., 2014) and Induced Seismicity
Map (ISM), [43] derived from the publicly available data in California
[44], understanding the potential seismic sensitivity of the region to
hydraulic stimulation takes on greater importance.
The prolific nature of seismic activity in California, owing to an extensive
network of faulting, hinders the task of differentiating potential induced
seismic events from natural events. The task of locating potential induced
events benefits from extensive data, however these data are not organized
in a fashion conducive manner.
The case history we are highlighting here is from the San Joaquin Valley,
California. The seismic catalogue used in this study was provided by the
Southern California Seismic Network (SCSN). The catalog gives the loca-
tion, date, depth, magnitude, and parameter uncertainties for over 470,000
events since 1932. The magnitude used in this study was local magnitude
(ML), with a magnitude of completeness (MC) demonstrated by Hutton and
Jones [45] to have significantly improved with installation of more seis-
mic stations. In Aminzadeh [1], there is a review of the induced seismicity
study conducted in San Joaquin Valley, California. A detailed statistical
assessment of the induced seismicity potential within this region can be
found in Goebel et al. [46]. Figure 1.8 is an example showing the distribu-
tion of different types of seismicity in the San Joaquin Valley. A first spatio-
temporal test was conducted using records of hydraulic fracture operations
conducted in the Bakersfield area from 2000 to 2013. The maximum true
vertical depth for the fractured well dataset is just over 12,900 feet (3.9 km)
and, thus, hypocenters for potential induced seismic events are expected to
fall within a reasonable vicinity of the True Vertical Depth (TVD). Based on
a simulation of pressure transfer using COMSOL and taking into account
error in depth calculation of hypocenter, we considered events less than
10 km depth for our induced seismicity study. This is a more lenient depth
constrain compared to many previous studies of induced seismicity, for
which the limit was chosen around a five-kilometer depth. The left side of
Figure 1.8 displays locations of HF operations and earthquake epicenters
from 2000 to 2013. The right map in Figure 1.8 displays HF operations and
only those earthquakes with a hypocentral depth less than ten kilometers.
Following the filter treatment, no correlation remains.
The number of HF and other SFIP operations were examined from 2000
to 2013. HF operations took place in five oilfields in the region. However,
no spatial-temporal relationship was found between seismic events and
fracturing activities. Some seismic events have been observed in some
places between fracturing and induced seismicity, but it would be unrealis-
tic to assume such correlations as a universal phenomenon.
7.5 0 7.5 15 22.5 30 km
Belridge South
Elk Hills
Figure 1.8 All seismic events in study area, San Joaquin Valley, California (left), filtering of
seismic events based on an event depth less than 10 km in San Joaquin Valley (right), Jabbari
et al. [27].
Our case history also attempts to distinguish between induced and

tectonic seismicity. The b-value analysis may provide insight into better
ways to distinguish induced and tectonic seismicity. Goebel et al. [46]
identified several target areas where the petroleum industry activity is
collocated with seismic activity and relatively low b-values. B-values are
generally low close to the San Andreas Fault, possibly suggesting higher
stress regimes compared to the more distant regions from active faults. The
spatial b-value maps are a promising first step in delineating seismicity in
low stress regimes from seismicity in higher stress regimes, which conse-
quently have higher fault activation potential.
Figure 1.9 (top) shows spatial variations in b-value in the study region
in San Joaquin Valley. The faulting (tectonics)-related b-values generally
seem to be lower (normally below 1, highlighted in blue colors). This can
be compared against the higher b-values (in red colors) mostly in the areas
Hydraulic Fracturing Wells
North
Rose Shafter
Round
Mountain
Lost Hills
Belridge
Elk Hills
Buena Vista
0 5 10 15
No Frac Wells
Figure 1.9 Spatial variations in B-values in the study region. (Left), color coded number of
HF in a few sizable oil and gas fields in San Joaquin Valley, (Right) [29].
with a number of active oil and gas fields. Figure 1.9 (bottom) shows a few
large oil fields with the color coded number of HF jobs.
The probability of activating faults due to anthropogenic influences is
likely highest in regions where active faults, fluid injection wells, and low
b-values have been encountered. This is the case, e.g., within the southern
part of the study region. In the proximity of active faults, both different
types of seismicity may occur and more detailed studies are required to
differentiate induced from natural (tectonic) seismicity. Areas of relatively
high b-values are likely connected to smaller ambient stresses and seismic
events are less likely to grow to large sizes.
See additional references on: What can microseismic tell us about
hydraulic fracturing [47], Geomechanical approach for microseismic frac-
ture mapping [48], Finite element method based modeling of hydraulic
fracturing [49], Flowback of Fracturing Fluids with upgraded visualization
of hydraulic Fracturing and its Implication of on Overall Well Performance
[50], Simulation of Hydraulic Fracturing-Induced Permeability Stimulation
using Coupled Flow and Continuum Damage Mechanics [51], and Oil,
the Next Revolution [52].
1.8 Assessment of Global Oil and Gas Resources

Amenable for Extraction via Hydraulic Fracturing
Shale resources as well as tight oil and gas are known to exist in parts of
Europe, Australia, Russia, China, and elsewhere, but this resource has
not been fully characterized by test wells. National Energy Technology
Laboratory (NETL) highlighted the technology developments that have
led to the emergence of shale gas development in a large scale [53]. The US
Energy Information Agency has published assessments of shale gas and
oil resources in various countries and basins; according to the compila-
tion published in 2015, the technically recoverable wet shale gas and tight
oil in 46 countries is 7,577 tcf and 419 billion barrels (Gbbl), respectively.
These values are larger than the proved global reserves of 6,832 tcf of nat-
ural gas and 240 Gbbl of oil [54]. Shale gas resources in the USA, Canada,
and Mexico are 623, 573, and 545 tcf, respectively. Outside North America,
significant resources are present in China (1100 tcf), Argentina (800 tcf),
Algeria (707 tcf), and Australia (430 tcf). Technically recoverable global
tight oil resources are estimated at 419 bbl. Substantial tight oil resources
are found in the USA (78 Gbbl), Russia (75 Gbbl), China (32 Gbbl), Libya
(26 Gbbl), and the UAE (23 Gbbl).
Whereas production of shale gas and oil in the US has been commer-
cially deployed for over a decade now, it has not been widely deployed
elsewhere. Factors limiting the development include public resistance to
the technology, availability of adequate water, and lack of suitable geo-
logic conditions. If the experience in the US is replicated in other coun-
tries, tight gas and oil resources could play a very large role in the global
energy scene.
1.9 Economics of HF
The total investment for a HF well must cover not only the cost of drill-
ing and completing the well, but also of obtaining land rights and per-
mits. These costs vary with location, and there is a broad range of costs.
According to the Energy Information Agency, the costs can range between
three and ten million dollars a well. Depending on the geology, the wells
produce different amounts of oil and gas, with oil being the more valuable
product, and the one that determines the economics. There are wells that
only produce gas, but, because there is already an oversupply of gas such
that no new dry wells are being developed, only those with takeoff con-
tracts are in production. It is a shame that the associated gas from oil-
producing wells in Bakken and other basins are being flared for lack of
pipeline infrastructure.
For new HF wells, the average breakeven cost of crude oil from differ-
ent basins currently ranges between $30 and $55 per bbl. The total cost
depends on the geographic location, depth of the target, horizontal length
of the well (offset), number of HF stages, and many other factors. For
example, in the Wolfcamp Delaware basin, the average breakeven cost is
$42/bbl, while in the Bakken basin it is $52. These costs are considerably
lower than they were in 2008 (~$70/bbl), when fracking first became com-
mercially viable, thanks to the spike in crude oil prices. Innovations such
as methods for better locating resources, more efficient drilling, and mul-
tipad drilling have markedly reduced costs. In face of losing market share,
in 2012, the Organization of the Petroleum Exporting Countries (OPEC)
and Russia decided to increase their production and drive down the price
of oil from $100/bbl to $50/bbl. At $50/bbl, producing conventional oil is
still profitable. The expectation was that the competition would be driven
out of business at this reduced price.
However, because of the improved technologies, producing oil from
many of tight formations was still profitable. Operations at those that were
not profitable were simply turned off temporarily and could be brought
on-line quickly when conditions became favorable. The rapidity with which
HF wells can be turned off or brought back on stream is a game changer in
the geopolitics of oil. Furthermore, the new focus on “refracking” to extend
the life of the shale oil and gas fields may become a new normal.
1.10 Conclusions
The key factors impacting shale production are environmental concerns
and economics. In California, environmental impacts of HF practices are
limited when compared to other energy resource extraction, such as urban
oil drilling. However, this is only the case because of extensive State regu-
lations on oil drilling and waste disposal in California and chance geologic
conditions and reservoir parameters that allow for safe disposal, and con-
clusions about California may not be applied elsewhere.
More generally, States engaging in HF, including California, will
aggravate water shortages where local groundwater supplies are limited.
Recycling wastewater must be a priority in these cases. In addition, more

methane leakage monitoring stations and assessments are needed in major
HF operations across the country, to inform potential climate and health
risks and develop operational procedures to minimize and mitigate direct
and indirect impacts. HF operation can be made much safer, avoiding
groundwater contamination and induced seismicity, if wastewater is not
injected into rocks with little porosity and storage capacity, which often
causes rock breakage. The probability of activating faults with HF is most
likely in regions containing active faults, fluid injection wells, and high
stress fields.
The reality is that, for regions where HF is economically profitable or
inevitable, it will persist and contribute heavily to the energy economy.
Moving forward, more environmental regulations, including HF specific
regulation and best practices training programs, will help avoid the con-
cerns surrounding fracking. The States must also better survey geologic
conditions (including rock porosity) to improve waste disposal methods.
In spite of many advances made in different aspects of HF, many new
technologies need to be developed and implemented in order to make
the operations more efficient, reduce cost, and minimize environmental
footprints.
State regulators are best able to regulate HF and plan adaptive responses
to earthquakes, and have a duty to follow in the footsteps of Californian
regulators. Meanwhile, private and public sector research into applications
of machine learning, DNA fingerprinting of reservoirs, “Shock-Waves” or
super waves and Pulsed Power Plasma Stimulation will provide alternative
approach to create similar or better results as the conventional HF. With
the aid of repeated microseismic measurement data and possibly 4D seis-
mic, the subsurface may be more accurately characterized and modeled
to achieve safety and efficiency goals. Other techniques such as the use of
statistical distributions in seismicity can also shed valuable insights into
the fracturing process and its relation to the structural features within the
reservoir.
The debate on the benefits and potential risks of HF and other oil and
gas operations will continue. Independent of one’s stance on this debate,
the need for additional research and more accurate assessment of the
associated risk factors, along with the development of both monitoring
programs and modeling approaches for alleviating the concerns is appar-
ent. Ultimately, such work will lead to the mitigation of risks associated
with the development of sound policies and strategies based upon strong
science and engineering fundamentals while giving full consideration to
environmental safety and economic factors.
Acknowledgement
The author acknowledges the financial support of the USC Induced
Seismicity Consortium (isc.usc.edu) members, valuable input from
Ripudaman Malhotra, the co-author of A Cubic Mile of Oil, and Kelly Rose
of NETL.
References
1. Department of Energy. Hydraulic Fracturing Technology, https://energy.
gov/fe/hydraulic-fracturing-technology
2. G.E. King, HF 101: What every representative, environmentalist, regulator,
reporter, investor, university researcher, neighbor and engineer should know
about estimating frac risk and improving frac performance in unconven-
tional gas and oil. In: SPE 152596, SPE HF Technology Conference, pp.
3. California Council on Science and Technology, Advanced Well Stimulation
Technologies in California: An Independent Review of Scientific and
Technical Information, July (2016). Accessible by: https://ccst.us/wp-content/
uploads/160708-blm-report.pdf
4. Colorado Oil & Gas Conservation Commission, Frequently Asked Questions
About HF (2011). Accessible by: https://cogcc.state.co.us/Announcements/
Hot_Topics/Hydraulic_Fracturing/Frequent_Questions_about_Hydraulic%20
Fracturing.pdf
5. U.S. Energy Information Administration (2018) Petroleum and other liquids
data. https://www.eia.gov/petro leum/data.php#crude
6. BP Statistical Review of Global Energy (2017). Accessible by: https://
www.bp.com/en/global/corporate/energy-economics/statistical-review-
of-world-energy.html, https://www.forbes.com/sites/arielcohen/2018/12/21/
americas-oil-and-gas-reserves-double-with-massive-new-permian-
discovery/#328c4cd92c91
7. Baker Hughes, Baker Hughes Rig Count. Accessible by: http://phx.corporate-
ir.net/ phoenix.zhtml?c=79687&p=irol-rigCountOverview
8. FT (2018). Accessible by: https://www.ft.com/content/a8314d06-0741-11e8-
9650- 9c0ad2d7c5b5
9. API (2018). Accessible by: https://www.ft.com/content/a8314d06-0741-
11e8-9650- 9c0ad2d7c5b5or http://www.api.org/oil-and-natural-gas/wellsto-
consumer/exploration-and-production/hydr aulic-fracturing
10. L. Minh-Thong Le (2018) An assessment of the potential for the development
of the shale gas industry in countries outside of North America. Heliyon
4:e00516. https://doi.org/10.1016/j.heliyon.2018.e00516
11. Warpinski NR, Mayerhofer MJ, Vincent MC, CipollaCL, Lolon EP (2009)
Stimulating unconventional reservoirs: maximizing network growth while
optimizing fracture conductivity. J Can Pet Technol 48(10):39–51. SPE-

114173-PA. https://doi.org/ 10.2118/114173-PA
12. Lacazette A, Vermilye J, Fereja S, Sizking C (2013) Ambient fracture imag-
ing: a new passive seismic method. In: Unconventional resources technology
conference (URTeC), Denver, 12–14 Aug 2013
13. Maity D, Salehi I, Ciezobka J (2014) Semblance weighted emission mapping
for improved hydraulic fracture treatment characterization. In: SEG annual
meeting, Denver, 26–31 Oct 2014
14. Das I, Zoback MD (2013) Long-period long-duration seismic events during
hydraulic stimulation of shale and tight-gas reservoirs – Part 2: location and
mechanisms. Geophysics 78(6):KS97–KS105
15. Maity D, Aminzadeh F (2015) Novel fracture zone identifier attribute using
geophysical and well log data for unconventional reservoirs. Interpretation
3(3):T155–T167. https://pdfs.semanticscholar.org/0ccb/4a21a8dd7d4a4b1bb
2be4f8ce0192811aa9e.pdf
16. Wessels S, Kratz M, Pena ADL (2011) Identifying faultactivation during
hydraulic stimulation in the Barnett shale: source mechanisms, b values,
and energy release analysis of microseismicity. In: SEG annual meeting, San
Antonio, 18–23 Sept
17. Maity D, Ciezobka J (2018) Correlating microseismicity with relevant geo-
physical and petrophysical data to understand fracturing process during
hydraulic stimulation: a case study from the Permian Basin. In: SEG annual
meeting, Anaheim, 15–19 Oct
18. Ciezobka J, Courtier J, Wicker J (2018) Hydraulic fracturing test site (HFTS) –
project overview and summary of results. In: Unconventional resources tech-
nology conference (URTeC), Houston, 23–25 July
19. Maity D, Ciezobka J, Eisenlord S (2018) Assessment of in-situ proppant place-
ment in SRV using through fracture core sampling at HFTS. In: Unconventional
resources technology conference (URTeC), Houston, 23–25 July
20. Ante MA, Manjunath GL, Aminzadeh F, Jha B (2018) Microscale laboratory
studies for determining fracture directionality in tight sandstone and shale
during HF. In: Unconventional resources technology conference (URTeC),
Houston, 23–25 July 2018. https://doi.org/10.15530/urtec2018-2903021
21. Maity D, Aminzadeh F (2013) A new approach towards optimized passive seis-
mic survey design with simultaneous borehole and surface measurements. In:
Joint PSAAPG/SPE/PSSEPM/PCS-SEG conference, Monterey, 19–25 Apr 2013
22. Boroumand N, Eaton D (2016) Interpreting microseismic data from HF:
understanding subsurface deformation through numerical modeling,
recorder. Can Soc Explor Geophys 41(09)
23. Eyre TS, Van Der Baan M (2015) Introduction tomoment tensor inversion of
microseismic events. In: GeoConvention, Calgary, 4–8 May 2015
24. Tan Y, Engelder T (2016) Further testing of the bedding-plane-slip model
for hydraulic-fracture opening using moment-tensor inversions. Geophysics
81(5). https://library.seg.org/doi/10.1190/geo20150370.1
25. EPA (2018) http://science.sciencemag.org/content/early/2018/06/20/science.

aar7204
26. Jabbari N, Aminzadeh F, de Barros F (2016) Hydraulic fracturing and the
environment, risk assessment for groundwater contamination. Stoch Environ
Res Risk Assess 30:10. https://doi.org/10.1007/s00477-0161280-0
27. Jabbari N, Aminzadeh F, de Barros F (2015) Assessingthe groundwater con-
tamination potential from a well in a HF operation. J Sustain Energy Eng
3(1):66–79
28. Eisenlord S, Hayes T, Perry K (2018) Environmental impact analysis on the
hydraulic fracture test site (HFTS). In: Unconventional resources technology
conference (URTeC), Houston, 23–25 July
29. Ceres (2016) HF & water stress: water demand by the numbers. https://www.ceres.
org/resources/reports/hydraulic-fracturing-waterstress-water-demand-numbers
30. Fracking in Texas, Ballotpedia. https://ballotpedia. org/Fracking_in_Texas
31. Lebas R, Lord P, Luna D et al. (2013) Development anduse of high-TDS recy-
cled produced water for crosslinked-gel-based HF. Presented at the SPE HF
technology conference, Woodlands, 4–6 Feb. SPE-163824-MS. https://doi.
org/10.2118/163824-MS
32. Hydralicfracturing.com (2018) http://www.hydraulic fracturing.com/#/?
section=air-emissions
33. Petersen MD, Moschetti MP, Powers PM, Mueller CS,Haller KM, Frankel
AD, Zeng Y, Rezaeian S, Harmsen SC, Boyd OS, Field N, Chen R, Rukstales
KS, Luco N, Wheeler RL, Williams RA, Olsen AH (2014) Documentation
for the 2014 update of the United States national seismic hazard maps. U.S.
Geological Survey open-file report 2014–1091, 243p. https://doi.org/10.333/
ofr20141091. URTeC: 2903021
34. Aminzadeh F, Goebel THW (2013) Identifying induced seismicity in active
tectonic regions: a case study of the San Joaquin Basin, California. AGU fall
meeting, Abstract: S31F-06,
35. D.W. Eaton and R. Schultz, Increased likelihood of highly overpressured shale
formations. Geophys. J. Int. 214(1), 751–757 (2018). https://doi. org/10.1093/
gji/ggy167
36. NAS (2013). Accessible by: https://www.nap.edu/catalog/13355/induced-
seismicity- potential-in-energy-technologies
37. A.D.L. Pena, S. Wessels, A.R. Gunnell, K.J. Numa, S. Williams-Stroud, L.
Eisner, M. Thornton, and M. Mueller, Fault or Frac? Source Mechanism and
B-value Detection of Fault Fracturing - A Barnett Case Study. 73rd EAGE
Conference and Exhibition (2011). DOI: 10.3997/2214-4609.20149726.
38. X. Bao and D.W. Eaton, Fault activation by HF in western Canada. Science
(2016). DOI: 10.1126/science.aag2583.
39. W.L. Ellsworth, Injection-induced earthquakes. Science 341(6142), 1225942
(2013). DOI: 10.1126/science.1225942
40. Davis and Frohlich (2013). Accessible by: https://www.sciencedirect.com/
science/article/pii/S0264817213000846
41. Frac focus (2018). Accessible by: http://fracfocus.org/

42. F. Aminzadeh, A. Tiwari, and R. Walker, Seismic Events and HF Activities.
AGU meeting in San Francisco, December 2014 (2014).
43. A. Tiwari, R. Walker, and F. Aminzadeh, HF and Induced Seismicity in
California. HF (HF) Journal, August 2014 (2014).
44. DGGR (2015). Accessible by: https://maps.conservation.ca.gov/doggr/
wellfinder/doggr-help.html
45. L.K. Hutton and L.M. Jones, Local magnitudes and apparent variations in
seismicity rates in Southern California (1993).
46. T.H.W. Goebel, E. Hauksson, F. Aminzadeh, and J.-P. Ampuero, An objec-
tive method for the assessment of possibly fluid-injection induced seismic-
ity in tectonically active region in central California. J. Geophys. Res. (2015).
doi:10.1029/2015JB.
47. S. Maxwell S (2014) What can microseismic tell us about HF? In: AAPG/
SEG/SPWLA Hedberg conference, Austin, 7–11 Dec 2014
48. Hosseini M, Aminzadeh F (2014) A geomechanical approach for microseis-
mic fracture mapping. In: AAPG/SEG/SPWLA Hedberg conference, Austin,
7–11 Dec 2014
49. Khodabakhshnejad A, Aminzadeh F (2014) An extended finite element
method based modeling of HF. In: AAPG/SEG/SPWLA Hedberg confer-
ence, Austin, 7–11 Dec 2014
50. Desi K, Aminzadeh F (2016) Flowback of fracturing fluids with upgraded
visualization of HF and its implication of on overall well performance. J
Sustain Energy Eng 4(3–4):250–261
51. Samnejad M, Aminzadeh F, Jha B (2017) Simulation of HF-induced perme-
ability stimulation using coupled flow and continuum damage mechanics,
SPE-187250-MS. In: ATCE conference in San Antonio, 9–11 Oct 2017
52. Maugeri L (2012) Oil, the next revolution. Discussion paper 2012-10, Belfer
Center for Science and International Affairs, Harvard Kennedy School, June
2012
53. NETL(2011)https://www.netl.doe.gov/file%20library/research/oil-gas/
Shale_Gas_March_2011.pdfNetl
54. U.S. Energy Information Administration (2018) https://www.eia.gov/analysis/
studies/worldshalegas/, BP 2018
Part 2
GENERAL CONCEPTS

2
Evolution of Stress Transfer Mechanisms
During Mechanical Interaction Between
Hydraulic Fractures and Natural Fractures
Birendra Jha
*
Mork Family Department of Chemical Engineering and Materials Science,

University of Southern California, Los Angeles, USA
Abstract
Design and control of hydraulic stimulation jobs in naturally fractured rock is
challenging due to interaction between injection-induced fractures and natural fra
ctures. Dynamic stress transfer between the two fracture systems and the result-
ing slip response of the natural fractures are key to understanding the temporal
evolution of hydraulically- and mechanically-stimulated reservoir volumes. Here,
we present a computational model of interaction between induced and natural
fractures based on numerical simulation of coupled flow, deformation, and fail-
ure processes. By capturing the time dependence of stress and energy transfer
mechanisms between hydraulic and natural fracture systems, our model allows
us to better understand the mechanical response of natural fractures to hydraulic
fracturing for the purpose of upscaling such responses in field-scale simulation
models.
Keywords: Geomechanics, hydraulic fracturing, upscaling
2.1 Introduction
Hydraulic stimulation of naturally fractured rock is a common operation
for fluid extraction from low permeability hydrocarbon-bearing rock and
geothermal reservoirs [1–3]. Yet, our understanding of the interaction
between hydraulically induced fractures and natural fractures in these
Email: bjha@usc.edu

37
reservoirs is far from complete and suffers from a lack of knowledge of

the physical mechanisms that control such interactions [4–7]. It is known
that natural fractures can be activated through elastic effects even before
hydraulic communication is established between the hydraulic fracture
and natural fractures [8]. However, such elastic interaction mechanisms
evolve with time leading to evolution in stresses and slip on pre-existing
fractures, which makes the estimation of permeability enhancement and
seismicity from hydraulic stimulation a challenging task. Recent compu-
tational and experimental studies in this area have provided significant
insight into understanding these interaction mechanisms and their sensi-
tivity to fracture propagation [9–11], in-situ stress anisotropy [10, 12, 13],
angle between hydraulic and natural fractures [12, 14, 15], viscosity of the
fracturing fluid [7], coupling between flow and mechanics [7], mode of
failure [16], and injection rate and pressure [17]. The focus of some of these
computationally expensive studies has been on understanding the intersec-
tion between a hydraulic fracture and a natural fracture [12, 14], modeling
the stress shadow effect in presence of multiple propagating fractures [10],
maximizing connectivity between a hydraulic fracture and a natural frac-
ture network [12], and estimating the stimulated reservoir volume in terms
of the number of natural fractures activated by hydraulic fracturing [7],
natural fracture density and brittleness [18], and permeability enhance-
ment [16]. However, a detailed study of the temporal evolution of stress
and slip on a natural fracture during a hydraulic fracturing event is miss-
ing. In some of the earlier work, hydraulic fracturing has been modeled as
an instantaneous event [7], which precludes studying the time evolution of
natural fracture response to hydraulic fracturing. Our goal is to describe
the individual mechanisms of interaction between the hydraulic and natu-
ral fractures in terms of reduced-order models that can be used in upscaled
reservoir simulation models of flow and deformation in real fields. Such
upscaled models are used for field development planning and reservoir
management purposes and can benefit from the speed gained by reduc-
ing the complexity associated with fracture models. Capturing the time
dependence of mechanical response of pre-existing fractures to hydrau-
lic fracturing is the key to a successful reduced-order model of hydraulic
stimulation of naturally fractured rock.
In this article, we study the time evolution of mechanical interaction
between hydraulic and natural fractures during opening of a hydraulic
fracture. We use a computational approach to simulate the coupled pro-
cesses of fluid flow, mechanical deformation, and rock failure processes in
a simplified setting of one hydraulic fracture interacting with four natural
fractures. We use a one way coupled sequential solution scheme in which
Evolution of Stress Transfer Mechanisms 39
the flow problem is solved independent of mechanical deformation and

failure. We analyze evolution of slip and traction on the natural fractures
in both shear and normal directions as the hydraulic fracture is created. In
agreement with the earlier studies [7], we find that the shear induced on
the natural fractures as a result of opening a hydraulic fracture depends on
the relative position and orientation of the natural fractures with respect
to the hydraulic fracture and the principal stress directions. We find spatial
and temporal heterogeneity in the failure response of natural fractures that
can influence the permeability enhancement and microseismicity resulting
from hydraulic fracturing. We explain such heterogeneity in terms of stress
and energy transfer from the hydraulic fracture to the natural fractures.
2.2 Physical Model

We consider a plane stress model under compression with one injection-
induced hydraulic fracture (HF) and four pre-existing natural fractures
(NF1, NF2, NF3 and NF4) as shown in Figure 2.1. The fractures are not con-
nected to each other hydraulically. The boundary conditions are such that the
maximum principal stress, σ1, is aligned with the x-axis and the minimum
principal stress, σ3, is aligned with the y-axis. The relative angle between the
hydraulic fracture and σ1 is 0 degree and the relative angle between the nat-
ural fracture and σ1 is 45 degree. Water is injected at a constant bottomhole
pressure at the center of HF, which then travels away from the center and
NF1 NF3
HF
Ly 1
NF2 NF4
y
x
Lx
Figure 2.1 Schematic of the physical domain used to study interaction between a hydraulic
fracture (HF) and four natural fractures (NF1, NF2, NF3, and NF4). The midpoint of the
hydraulic fracture is shown as a blue dot. Midpoints of the natural fractures are shown as
red dots. The maximum principal stress is oriented along the x-direction and the minimum
principal stress is oriented along the y-direction. The other two boundaries are fixed in the
respective normal directions to prevent translation or rotation of the domain.
along the fracture in both directions. The injection pressure is maximum at

the center of HF and decreases as we get closer to fracture tips. The injection
pressure increases non-linearly in time, and it causes opening of HF, expan-
sion of the domain, and changes in the state of stress around the natural
fractures.
2.3 Mathematical Formulation

The linear momentum balance equations governing the quasi-static mechan-
ical equilibrium of a porous medium in absence of body forces are as follows
·σ=0 (2.1)
where the total stress tensor σ (x,y,t) is related to the pore pressure p(x,y,t)
and the effective stress tensor σ (x,y,t) through the equation, σ = σ – bp1,
where 1 is the identity tensor and b is the Biot coeffcient. Note that tension
is positive in our sign convention. In 2D, the system of partial differential
equations become
xx xy
0,
x y
(2.2)
xy yy
0.
x y
The total stress can be decomposed as σ = σ0 + δσ, where the initial stress
σ0 results from the boundary tractions and the stress change δσ results from
hydraulic fracturing. The mechanical equilibrium equations become
0 b ( p) 0 b p, (2.3)
where p = p0 + δp with p0 being the initial pressure and δp being the change
in pressure. We assume a slightly compressible single-phase flow for the
flow problem. The pressure can be obtained by solving the fluid mass bal-
ance equation given as [19]
1 p
b V v f (2.4)
M t t
where V V /K dr is the volumetric strain, V is the volumetric effective

stress, Kdr is the drained bulk modulus, v is the Darcy velocity, M is the Biot
modulus and f is the volumetric injection rate per unit bulk volume of the
medium. We use an analytical solution of Eqn. (2.4) under the assump-
tion of weak coupling from the mechanics problem to the flow problem.
A weak solid-to-fluid coupling is usually observed in gas reservoirs under
high degree of consolidation where the Biot modulus is small and the
drained bulk modulus is large. We assume a linear elastic solid behavior
u ut
given by δσ = Cε, where s
u is the infinitesimal strain
2
tensor defined in terms of the displacement vector u = [ux,uy]t, and C is
the rank-4 elastic tensor. When using the engineering notation to express
stress and strain as vectors, the elastic tensor can be expressed as a matrix
in terms of Young’s modulus E and Poisson’s ratio v for a homogeneous
isotropic solid assumed here.
We model the compressible fluid behavior with the fluid density deter-
mined by a constant fluid compressibility c f 1 d f.
f ap
We use the cohesive element approach to explicitly represent fractures
in our model. We calculate traction and slip on the fracture surfaces as
they evolve during hydraulic stimulation. Detailed description of our
fracture model is provided elsewhere [19, 20]. Here, we provide a brief
overview. We use the finite element method to solve the mechanics equa-
tion (Eqn. (2.1)). The weak form in presence of one single fracture is as
follows [19]:
Ly Lx
s
: bp1 d l pf n d l pf n d
0 0 f f
Lx Ly
3 dx 1 dy ,
0 0
(2.5)
where η is the suitably chosen vector test function, l = σ n = σn + pfn is the

effective traction vector on the fracture surface with a surface normal n, pf
is the fracture pressure, and Γf+ and Γf– are the positive and negative sides of
the fracture surface. We can decompose the traction vector into its initial
value and the change from the initial value as l = l0 + δl. We can rewrite
the equation in terms of the initial stresses on the right hand side and the
change in stresses on the left hand side:
Ly Lx
s
: d ld ld
0 0 f f
Lx Ly Ly Lx Ly Lx
s
3 dx 1 dy : 0 d pd
0 0 0 0 0 0
l0 d p f 0nd p f nd
f f f
l0 d p f 0nd p f nd
f f f
(2.6)
We use a one-way coupled approach where we assume no pressure

change (drained condition) outside the hydraulic fracture, i.e. δp = 0, and
we use an analytical pressure solution [21] inside the hydraulic fracture
to compute the change in HF pressure, δpf. The displacement vector u
and the change in effective traction vector δl are the two unknowns in the
mechanical problem. Eqn. (2.6) is the displacement equation and the sec-
ond equation is provided by the definition of slip as the difference between
the displacements on the positive and negative sides of the fracture:
u d u_d dd (2.7)
f f f
where d is the slip vector. We use the Mohr-Coulomb theory to model

frictional slip on the fractures. The theory can be extended to the tensile
stress regime to model the Griffth failure (Brace, 1960). Here, we assume a
sufficiently large tensile strength such that the hydraulic fracture does not
grow in the long axis direction (x-direction), and the natural fractures are
under a compressive state of stress such that they fail only under compres-
sion. Following the Mohr-Coulomb theory, when the shear traction on the
fracture exceeds the friction stress τf = τc + μfσ N, where τc is the intrinsic
cohesion, μf is the coefficient of friction, and σ N = l·n is the effective normal
traction, the slip vector d is updated to accommodate the change in trac-
tion required to honor τ = τf . The shear traction magnitude is computed as
τ = |l – σ Nn|. Opening of HF is modeled with a positive change in HF pres-
sure, δpf > 0, that corresponds to a tensile change in the effective normal
traction. In the current model, we neglect hydraulic fracture propagation
and fluid leak-off, and we focus on early-time interaction between hydrau-
lic and natural fractures. In particular, we quantify the evolution of state
of stress and slip on the natural fractures as a function of opening of the

hydraulic fracture. We use the standard finite element method with bi-lin-
ear elements in the bulk and linear interface elements on the fractures to
discretize Eqns. (2.6–2.7).
2.4 Numerical Model

We solve the quasi-static mechanical equilibrium problem and the fracture
contact problem, Eqns. (2.6–2.7), using PyLith [20] to calculate displace-
ments and stresses in the 2D domain and slip and tractions on the 1D frac-
tures. We use the finite element grid shown in Figure 2.1 with Lx = 200 m and
Ly = 100 m and a 100 m long hydraulic fracture. We use triangular elements
in the matrix and linear interface elements along the fractures. We consider
a simulation with the following values: σ1 = σxx (Lx,y,0) = –40 MPa and σ3 =
σyy (x,Ly,0) = –30 MPa as the boundary tractions, and p0 = p(x,y,0) = 10 MPa
as the initial pressure. We assume E = 56.82 GPa and ν = 0.26. We use a static
friction model with τc = 0 and μf = 0.1, which, for the values chosen above
for σ1 and σ3, results in shear slip on the natural fractures as soon as hydraulic
fracturing begins. We assume that water is injected in the center of HF at
a constant bottomhole overpressure of δpf(Lx/2,Ly/2,t) = 44 MPa. The other
parameters are as follows: water compressibility of 1.45E-08 per Pa, water
viscosity of 1 cP, HF permeability of 1 darcy, and HF porosity of 0.5. We
calculate the evolution of HF pressure δpf(x,Ly/2,t) along the length of HF at
discrete time steps (Figure 2.2).
50
t 120 sec
40
pf, Mpa
30 1 sec
20
10
0
50 100 150
x, m
Figure 2.2 Time evolution of overpressure along HF calculated using the analytical solution
of the fluid mass balance equation for a constant pressure injection in a porous medium.
A non-uniform time stepping is chosen with finer resolution in the

beginning to resolve the early time transient and coarser resolution later as
the pressure reaches a pseudo steady state value. We conduct the simula-
tion for 0 < t < 121 sec. Initial stresses balance boundary tractions ensuring
zero initial displacements.
2.5 Simulation Results

Opening of the hydraulic fracture with time leads to an evolution of the
state of stress and resulting deformation in the domain, which drives the
time-dependent response of the natural fractures. Overpressure causes
opening of HF resulting in uy > 0 above HF and uy < 0 below HF (Figures
2.3 and 2.4). Opening is asymmetric in the y-direction because the top
boundary is a compression boundary and the bottom boundary is a fixed
ux
100
0
meter
–0.07
y, m
–0.035
–0.035
0
0 x, m 200
(a) 0.0
uy
0.035
0.07
(b)
Figure 2.3 Displacement field at t = 121 sec resulting from a balance of boundary
conditions and hydraulic fracturing induced stress changes. The horizontal displacement,
ux, changes sign as we move from HF to the top and bottom boundaries of the domain.
Its magnitude is asymmetric along x due to asymmetry in the right and left boundary
conditions. Magnitude of the vertical displacement, uy, is asymmetric along y because of
the asymmetry in the top and bottom boundary conditions. For the ease of visualization,
the mesh is distorted with displacements magnified by a factor of 50, and the color scale
is chosen to show the variation in ux and uy and not the minimum and maximum values
of these quantities, which are –0.022 m and 0.026 m for and –0.018 m and 0.092 m for uy.
The dash lines in the upper figure indicate the location of the profile plots in Figure 2.4.
100 100
80 80
60 60
y, m
40 40
xx
20 ux 20 yy
uy
xy
0 0
–0.05 0 0.05 –20 0 20
(a)) Displacement, m ((b) MPa
20
0.04
Displacement, m
MPa
0 0
–0.04
–20
0 50 100 150 200 0 50 100 150 200

(c) x, m (d) x, m
Figure 2.4 Profiles of displacements and stress changes orthogonal (upper row) and parallel
(lower row) to HF at t = 121 sec. Orthogonal profiles are at x = 70 m and parallel profiles are
at y = 25 m along dash lines shown in Figure 2.3a. (a) Displacement discontinuity across HF
is visible. ux changes sign above HF (y > 50 m) and similar to xx . uy is negative below HF
and positive above HF, as expected. (b) The vertical stress change, yy , is compressive due
to opening of HF. The horizontal stress change, xx , is also compressive in the immediate
vicinity of HF but becomes tensile as we approach the traction boundary on the top because
of upward displacement. The shear stress change, xy, due to HF opening is larger above
HF because of a larger upward displacement of the upper surface of HF. (c) Across the center
of HF, ux is asymmetric due to the asymmetry in the horizontal boundary conditions, and
uy is symmetric. Both displacements have discontinuities at x = 25 m and x = 175 m where
we cross NF2 and NF4, respectively. Stresses are piecewise continuous because we use linear
displacement elements.
displacement boundary. Displacement in the x-direction, ux, reflects the

stress field generated by the balance of boundary tractions, fixed displace-
ment boundaries, and pore pressure. The compressional boundaries on top
and right, with the principal stress ratio of σ1/σ3 = 1.5, and the fixed bound-
aries on bottom and left cause ux > 0 on the left half of the fracture surface
and ux < 0 on the right half of the fracture surface.
2.6 Effect of Hydraulic Fracturing on Natural Fractures

The initial values of tractions on NF can be computed analytically using
Cauchy’s formula. We know that the effective normal and shear traction on
the fracture surfaces are given as follows [8]
1 1
N xx yy xx yy cos 2 pf
2 2 (2.8)
1
xx yy sin2
2
This results in the following initial values: N (t 0) 20 MPa MPa and

τ(t = 0) = 0 on HF, N (t 0) 25 MPa and τ(t = 0) = 5 MPa on NF1
and NF4, which have their normal vectors such that α = 135 degree, and
N (t 0) 25 MPa and τ(t = 0) = –5 MPa on NF2 and NF3, which have
their normal vectors such that α = 45 degree. The ratio of shear to effective
normal traction of 0.2 at t = 0 is higher than the static friction of μf = 0.1
resulting in slip on natural fractures. Traction and slip on the natural frac-
tures evolve dynamically as the hydraulic fracture opens under a time-de-
pendent pressure increase from injection. To understand the evolution
of tractions on the natural fracture surfaces, we analyze the evolution of
stresses in elements adjoining the midpoints of the natural fractures as
shown in Figure 2.5.
Opening of the hydraulic fracture causes sliding of the natural frac-
tures (Figure 2.6a) even in absence of hydraulic communication. This
is reported as dry microseismic events in the literature [7]. The slid-
ing direction depends on the relative angle α between a natural frac-
ture and the maximum principal stress orientation. The shear traction
magnitudes on NF1, NF2, and NF3 decrease due to slip-induced relax-
ation (Figure 2.6c). The effective normal compressions on NF1, NF2, and
NF3 decrease because opening of HF leads to vertical expansion of the
domain and tensile changes around the natural fractures (Figure 2.6d).
NF4 behaves differently from the other three natural fractures due to its
position in the domain close to the maximum principal stress boundary
and the fixed bottom boundary. The shear traction on NF4 drops from
5 MPa to 2.48 MPa at the first time step and remains constant as it slips.
As explained in Figure 2.5 caption, the effective compression on NF4
increases causing a frictional stabilization away from the failure line in
the stress space (Figure 2.7a).
0 σxy
NF1
–10 NF2
MPa
σ΄yy NF3
NF4
–20
σ΄xx
–30
0 50 100
Time, sec
Figure 2.5 Time evolution of the effective stresses with opening of the hydraulic fracture.
The evolution is plotted for elements near the midpoints of the four natural fractures. All
natural fractures experience a decrease in compression and an increase in shear magnitude
due to opening of HF. NF4 experiences a relatively smaller decrease in compression and
largest increase in shear, which results in a compressive change in its normal traction
compared to tensile changes observed on the other three natural fractures (Figure 2.6).
0.01 0.01
NF dnormal, m
NF dshear, m
NF1
NF2
0 NF3 0
NF4
–0.01 –0.01
0 50 100 0 50 100
(a) Time , sec (b)
5 –20
–22
NF σ΄n, MPa
NF τ, MPa
–24
0
–26
–28
–5 50 100 –300 50 100
0
(c) (d)
Figure 2.6 Time evolution of slip (top row) and traction (bottom row) at the midpoints
of the four natural fractures during hydraulic fracturing. (a) Positive shear values on NF1
and NF4 indicate left-lateral shear and negative shear values on NF2 and NF3 indicate
right-lateral shear. (b) Normal slip values are zero indicating no opening or closing of the
natural fractures. (c) Shear traction magnitude decreases for all natural fractures as they
slip. This decrease continues longer for NF1, NF2, and NF3 than for NF4. (d) Normal
compression decreases on NF1, NF2, and NF3 thereby destabilizing them. It increases
on NF4, which has a stabilizing effect. Positive values indicate opening or tension and
negative values indicate closing or compression of the fractures.
3
0.01
2 Time Time
1
NF dshear, m
NF1
NFτ, MPa
0 NF2
NF3 0
–1 NF4
e
–2 Tim
–0.01
–3
–30 –25 –20 0 0.05 0.1
(a) NF σ΄n, MPa (b) HF center dnormal, m
3 –20
2 –22
NF σ΄n, MPa
1
NFτ, MPa
–24
0
–26
–1
–2 –28
–3 –30
0 20 40 0 20 40
(c) HF Ptip, MPa (d) HF Ptip, MPa
Figure 2.7 Evolution of traction and slip on natural fractures due to opening and
pressurization of the hydraulic fracture. (a) Stress paths of tractions at the midpoints of
the four natural fractures showing that NF1 and NF4 start sliding under positive (left-
lateral) shear tractions whereas NF2 and NF3 start sliding under negative (right-lateral)
shear tractions. The dash lines are the failure lines at μf = 0.1. The initial stress state
( N , )t 0 ( 25, 5) MPa falls outside the chosen scales of the plot. (b) Shear slips of
natural fractures increase monotonically with the opening at the center of HF. (c) Shear
tractions on the natural fractures decrease monotonically with the HF tip pressure.
(d) Effective normal compressions decrease on NF1, NF2, and NF3 and increase on NF4
with increasing HF tip pressure.
The stimulated reservoir volume (SRV) can be quantified in terms of the

energy that is transferred from the hydraulic fracture to natural fractures
(Figure 2.8a). We observe in our simulation that opening of HF transfers
elastic energy to NF1, NF2, NF3, and NF4, where the energy per unit frac-
ture surface area is calculated as l·d. The fastest rate of energy transfer is
observed for NF4 that experiences highest amount of slip and compression
during hydraulic fracturing. Microseismicity of natural fractures can also
be related quantitatively to hydraulic fracturing (Figure 2.8b). We calculate
2
the moment magnitude as, M w log10 M0 6.07 where M0 G | d | d
3 f
0.03
NF strain energy/area, MPa
0.5
Moment magnitude , Mω
0.4
0.02
0.3
0.2
0.01 NF1
NF2 0.1
NF3
NF4
0 0
0 2 4 0 0.05 0.1
(a) HF strain energy/area, MPa (b) HF center dnormal, m
Figure 2.8 (a) Energy transfer from the hydraulic fracture to the natural fractures is
monotonic. Growth in NF2 and NF4 strain energy are larger due to larger shear traction
and slip near the fixed bottom boundary. (b) The seismic moment magnitudes of natural
fracture slip events evolve with the opening of hydraulic fracture. Natural fractures are
grouped based on their position with respect to the traction boundary in the minimum
principal stress direction.
is the seismic moment, G is the shear modulus, and |d| is the slip magni-
tude [22]. We find that natural fractures are grouped by their proximity to
the traction boundary in the minimum principal stress direction i.e. NF2
and NF4, which are farther from the top boundary, are in a group with
a faster increase in the moment magnitude compared to NF1 and NF3,
which are closer to the top boundary.
2.7 Conclusion
We investigated the effect of hydraulic fracturing on pre-existing natural
fractures by analyzing the time-evolution of tractions and slips of natural
fractures with opening of the hydraulic fracture. We observe that the pro-
files of normal and shear slip along the natural fractures depend on the rel-
ative position of the natural fracture with respect to the hydraulic fracture,
the fixed displacement boundaries, and the minimum and maximum prin-
cipal stress orientations. We explain the heterogeneity in slip response of
the natural fractures in terms of the displacement and stress profiles paral-
lel and orthogonal to the hydraulic fracture. The energy transfer approach
in our hydraulic fracture–natural fracture interaction model can be used
to capture slip-induced permeability enhancement and microseismic
response of natural fractures during field scale simulations of hydraulic
fracturing. We are enhancing our framework to allow for two-way coupling
among the processes of fluid flow, deformation, and seismicity. We are also
implementing the capability to allow fracture propagation. The proposed
model allows for improved prediction of stimulated reservoir volume and
induced seismicity during hydraulic stimulation.
References
1. Hubbert, M.K. and Willis, D.G., Mechanics of hydraulic fracturing. Trans.
Am. I. Min. Met. Eng., 210, 6, 153–163, 1957.
2. Hossain, M.M., Rahman, M.K., Rahman, S.S., A shear dilation stimulation
models for production enhancement from naturally fractured reservoirs.
SPE J., 7, 2, 183–195, 2002, doi: 10.2118/78355-PA. SPE-78355-PA.
3. Curtis, J.B., Fractured shale-gas systems. Am. Assoc. Petrol. Geologists Bull.,
86, 1921–1938, 2002.
4. Lamont, G.C. and Jassen, F., The effects of existing fractures in rocks on the
extension of hydraulic fractures. J. Petrol. Technol., 15, 203–209, 1963.
5. Blanton, T.L., An experimental study of interaction between hydraulically
induced and pre-existing fractures. Paper SPE 10847 presented at the SPE/
DOE unconventional gas recovery symposium, Pittsburg, Pennsylvania, May,
1982.
6. Huang, J., Safari, R., Mutlu, U., Burns, K., Geldmacher, I., McClure, M.,
Jackson, S., Natural-hydraulic fracture interaction: Microseismic obser-
vations and geomechanical predictions. unconventional resources technol-
ogy conference (URTeC), Denver, Colorado, USA, 25–27 August, 2014, doi:
10.15530/urtec-2014-1921503.
7. Nagel, N.B., Sanchez-Nagel, M.A., Zhang, F., Garcia, X., Lee, B., Coupled
numerical evaluations of the geomechanical interactions between a hydrau-
lic fracture stimulation and a natural fracture system in Shale Formations.
Rock Mech. Rock Eng., 46, 581–609, 2013, doi: 10.1007/s00603-013-0391-x.
8. Jaeger, J.C., Cook, N.G.W., Zimmerman, R.W., Fundamentals of Rock
Mechanics, fourth edition, Blackwell Publishing, Malden, MA, 2007.
9. Olson, J.E., Bahorich, B., Holder, J., Examining hydraulic fracture-natural
fracture interaction in hydrostone block experiments. SPE 15261, Society of
Petroleum Engineers, 2012, doi: 10.2118/152618-MS.
10. Kresse, O., Weng, X., Gu, H., Wu, R., Numerical modeling of hydraulic frac-
tures interaction in complex naturally fractured formations. Rock Mech. Rock
Eng., 46, 555–568, 2013, doi: 10.1007/s00603-012-0359-2.
11. Warpinski, N.R. and Teufel, L.W., Influence of geologic discontinuities on
hydraulic fracture propagation. J. Petrol. Tech., 209–220, 1987.
12. Zangeneh, N., Eberhardt, E., Bustin, R.M., Investigation of the influence of
natural fractures and in situ stress on hydraulic fracture propagation using a
distinct—element approach. Can. Geotech. J., 52, 926–946, 2015.
13. Aimene, Y.E. and Ouenes, A., Geomechanical modeling of hydraulic frac-
tures interacting with natural fractures-Validation with microseismic and
tracer data from the Marcellus and Eagle Ford. Interpretation-J. Sub., 3,
71–88, 2015, doi: 10.1190/INT-2014-0274.1.
14. Potluri, N., Zhu, D., Hill, A.D., Effect of natural fractures on hydraulic
fracture propagation. SPE European Formation Damage Conference, The
Netherlands, May 2005, 2005, SPE 94568.
15. Gu, H., Weng, X., Lund, J., Mack, M., Ganguly, U., Suarez-Rivera, R.,
Hydraulic fracture crossing natural fracture at nonorthogonal angles: A crite-
rion and its validation. SPE Hydrualic Fracturing Technology Conference and
Exhibition, The Woodlands, Texas, 24–26 January 2011, 2011, SPE 139984.
16. Nassir, M., Settari, A., Wan, R., Prediction of stimulated reservoir volume
and optimization of fracturing in tight gas and Shale with a fully elasto-plas-
tic coupled geomechanical model. SPE J., 19, 5, 771–785, 2014.
17. Cheng, W., Jin, Y., Chen, M., Reactivation mechanism of natural fractures by
hydraulic fracturing in naturally fractured shale reservoirs. J. Nat. Gas Sci.
Eng., 23, 431–439, 2015.
18. Ouenes, A., Aissa, B., Boukhelf, D., Fackler, M., Estimation of stimulated
reservoir volume using the concept of Shale Capacity and its validation
with microseismic and well performance: Application to the Marcellus and
Haynesville. SPE Western North American and Rocky Mountain Joint Regional
Meeting held in Denver, Colorado, USA, 16–18 April 2014, 2014, SPE 169564.
19. Jha, B. and Juanes, R., Coupled multiphase flow and poromechanics: A com-
putational model of pore pressure effects on fault slip and earthquake trigger-
ing. Water Resour. Res., 50, 3776–3808, 2014, doi: 10.1002/2013WR015175.
20. Aagaard, B.T., Knepley, M.G., Williams, C.A., A domain decomposition
approach to implementing fault slip in finite-element models of quasi-static
and dynamic crustal deformation. J. Geophys. Res. Solid Earth, 118, 3059–
3079, 2013, doi: 10.1002/jgrb.50217.
21. Carslaw, H.S. and Jaeger, J.C., Conduction of Heat in Solids, second edition,
Clarendon Press, Oxford, 1959.
22. Hanks, T.C. and Kanamori, H., A moment magnitude scale. J. Geophys. Res.,
84, 2348–2350, 1979.
3
Primer on Hydraulic Fracturing
Concerning Initiatives on
Energy Sustainability
Michael Holloway1* and Oliver Rudd2
1
NCH Corporation, Irving, TX, USA
2
Independent Consultant, TX, USA
Abstract
Hydraulic fracturing (also known as fracking, fracing, and other variations) of
rock deep beneath the Earth’s surface to release petroleum product has become
a contentious subject across the globe. This practice is not to be confused with
drilling or extraction. Fracing is the process of using fluid power to fracture rock
to release gas and sometimes crude oil. It is not drilling, although drilling must be
done to establish a well in order to pump fluid thus fracturing rock to release prod-
uct. Certain countries have actually outlawed the practice of hydraulic fracturing
claiming that ground water and air pollution increase due to the practice. There is
also the claim that the comfort of life is adversely affected. Legitimate concerns are
always available and examples to purport a concerned view are magnified, in the
authors’ opinion. The intent of this paper is to provide a correct and balanced view
of fracturing underground rock with fluids in order to release a product to pro-
duce energy. The concept of using water to do work is nothing new. Pumping fluid
below ground in order to fracture rock to release gaseous petroleum is, however, a
relatively new practice. It is done with surprising precision as well as environmen-
tal concern yet it is interesting how the public reacts to the practice in relation to
other techniques used throughout the world. This paper will explore the materials
used as well as the concerns most common to the practice.
*Corresponding author: michael.holloway@nch.com

53
3.1 Hydraulic Fracturing

When people say that you cannot squeeze blood out of a turnip, it means
that you cannot get something from a person, especially money that they
do not have. You cannot squeeze blood from a turnip but you can release
trapped natural gas from rock – at least that is what is being accomplished
now throughout North America. One hundred years ago no one thought
it possible. Very few if any contemplated the idea. Natural gas, which is
primarily methane, has been proven to be an excellent fuel source. It can
be safely burned to create heat to power engines, boilers in factories and
homes as well as powering turbines for generating electricity. Projections
on natural gas volumes trapped under ground suggest a near inexhaustible
supply of this product yet with such abundance spawns controversy. A pop-
ular and economical technique relies on the gas from subterranean sources
requires fracturing rock bed. This process is actually carried out naturally
everyday with water or magma. Magma may flow into rock beds, super-
heating water to generate steam. The resulting pressure of the expanding
water molecule can be so great that it can lift and separate thousands of
tons of rock deep beneath the Earth’s surface. This same practice can be
carried out artificially (induced) using high powered pumps and various
liquid compounds. This technique, combined with new horizontal direc-
tional drilling machines, has enabled the harvest and distribution of natu-
ral gas. But at what cost? Does this practice contribute to greenhouse gas?
Does it create earthquakes? Does it contaminate the ground water supply?
These are important ideas to consider yet, and with proper examination
and logic, we are confident you will gain insight and reason in a practice
fueled by profit and civil concern.
No sides are taken in the following pages. This work is not intended to
be pro-industry or anti-fracturing. Namely, it aims to educate the general
public on what hydraulic fracturing really is, how it is conducted and what
possible harms may or may not come as a result.
3.1.1 Environmental Impact – Reality vs. Myth

In today’s society, it is really easy for organizations – be it the general
media, political groups, local organizations, unions or religious associa-
tions – to spread their beliefs to the public and push whatever agenda or
ideals they may have. These beliefs could be successfully put forward with
good intentions, successfully put forth with bad intentions or, in many
cases, put forward with good intentions but have a negative result. Sadly,
Primer on Hydraulic Fracturing 55
it seems human nature dictates that the first opinion heard or the opinion
heard the loudest and with the most hyperbole will be what the public
comes to believe. In time, once something is believed by enough people
and stated as “fact” long enough, the general public will no longer even
bother looking into facts and it will become part of the fabric of beliefs in
our society – for instance, a few examples of this phenomenon are: 1) you,
in fact, cannot see The Great Wall of China from the moon (not even
close); 2) the Sherlock Holmes character never once said “Elementary, my
dear Watson”; and 3) Neil Armstrong actually said, “One small step for
[a] man, one giant leap for mankind.”
As far as hydraulic fracturing is concerned, the aspect given the most
attention by press and most all concerned organizations is the impact it
may have on the environment. The question of environmental impact
through fracking is, to say the least, a very emotional topic and by far the
most polarizing issue; however, a great deal of analysis indicates that the
most significant environmental risks attributed to fracking are similar to
risks long associated with all drilling operations – including groundwa-
ter contamination due to inadequate cementing and/or well construction,
risks associated with trucking, leaks from tanks and piping and spills from
waste handling. This all-encompassing blame has given industry all the
ammunition needed to claim effects attributed to hydraulic fracturing are
overstated, not based on good science, or related to processes other than
hydraulic fracturing.
Due to the great ongoing controversy over alleged impacts from frack-
ing, many public groups have become deeply suspicious of the trustwor-
thiness and overall motives of the oil and gas industry. These suspicions are
continuously intensified by two things:
1. ongoing mistrust of data and findings due, in great part, to

semantics, and
2. by the industry initially refusing to disclose the chemical
makeup of fracking fluids and the additives used to enhance
hydraulic fracturing.
3.1.2 The Tower of Babel and How it Could be the Cause

of Much of the Fracking Debate
Most everyone has heard the story, or has a general understanding, of the
Tower of Babel from the Old Testament. In the Biblical account of this
story, humanity was attempting, as a unified group, to build a tower in
Mesopotamia to reach the heavens…only to have their efforts brought to

a halt by one of the most effective means imaginable. The efforts of this
united group of people were not thwarted by military force, or by weather,
or even by sickness and injury…their efforts in this undertaking were
thwarted by speech. The simple fact of this story is that building of the
tower came to a halt once the unified people were confounded by speech
and no longer able to communicate to work together. Now far be it for this
work to compare modern day hydraulic fracturing with the construction,
and subsequent stop in construction, of the Tower of Babel, but much of
the confusion, name-calling and general mistrust between groups on this
subject can be attributed to a difference in communication. Maybe once
this communication gap is bridged more effective talks can be established
between industry and concerned public – in place of wasting time on mis-
trust and name-calling. Hopefully, this work can help to bridge that gap.
It can be easily considered that a very large portion of negativity toward
hydraulic fracturing is actually attributable to processes other than hydrau-
lic fracturing. In the discussions between industry and the public, a great
deal of this problem can boil down to an issue of semantics…the oil and gas
industry has a narrow view of what fracking entails (including just those
processes related to the actual process of fracking while on location con-
ducting the fracking operation) while the general public is more inclined
to include many more activities commonly related to fracking (water and
sand trucking, product and equipment transport and storage, water dis-
posal), under the heading of “fracking.” This can cause misunderstandings
and skewed data in that many of the processes included by the general
public are utilized in many, if not all, drilling practices and are hard to put
solely under the heading of “fracking” when in actuality they could just as
easily be under the heading “completions” or “production.”
As has been discussed many times in the media and will undoubtedly be
discussed again and again and again ad nauseam, there are many proven
environmental impacts caused by drilling operations and processes related
to drilling. This fact can almost assuredly not be disproven by industry
personnel and can be a concern by the public in their feelings on gas well
completion and production activities. However, by the same token, there
have been well over a million wells that have gone through the process
of hydraulic fracturing, as is defined by industry, with not one reported
instance of ever impacting a fresh water aquifer. With the current issue
of semantics, public concerns can include many drilling processes while
industry can fall back on the fact that the industry definition of fracking has
never impacted fresh water in the ways commonly claimed by the media
for public consumption…and the debate can rage on with both sides being
right and both sides being wrong while never taking steps to come together
on a common goal.
3.1.3 Frac Fluids and Composition

The use of hydraulic fracturing for oil and gas exploration in the US has
become highly controversial with one of the greatest points of contention
between the public and industry being the makeup of frac fluids and their
possible impacts on public health and the environment. This has become
such a hot topic with many segments of the public for two reasons:
1. if a concern exists about the pumping of fluids into any

structure, then the most concern will naturally be centered
on what is being pumped, and
2. a great deal of suspicion arose and were intensified when
the oil and gas industry initially balked at the disclosing
the chemical makeup of fluids used to enhance hydraulic
fracturing.
This has become a major argument point for the concerned public
because, basically, “if you have nothing to hide why would you not want
to disclose it?” Take, for example, a small child walks into a room with
his hands behind his back and will not show you what is there for sev-
eral minutes…and then only does when forced. Well, even if it turns out
he was holding something as harmless as a feather behind his back it will
make you suspicious all the same wondering “what was he doing with that
feather!”
To make this contentious subject a little clearer, this section will pro-
vide descriptions of why frac fluids are needed, what general chemicals are
needed and used, relative amounts of chemicals in frac fluid composition,
proppants – the different types and uses, a discussion on slickwater, and a
discussion on present regulations and standards for industry disclosure of
frac fluid compositions.
3.1.4 Uses and Needs for Frac Fluids

There are a great deal of varied chemicals used every day in oil produc-
tion wells during all phases of drilling, completions and production. These
chemicals can include cements used to seal the annulus to protect the pipe
and surrounding formation from damage through wells exceeding the pro-
ducing and stimulation requirements placed on the pipe, temperature, and
even natural ground stresses. An example is corrosion inhibitors. These

chemicals help pipe and connection seals remain within design specifica-
tions to prevent failures. Corrosion prevention and treating chemicals may
also be necessary due to operational and field changes – even after well
completion and during production.
These chemicals can be used in much the same fashion as fracturing;
however, chemicals in well operation are applied in smaller quantities,
at lower pressure and in a regular maintenance driven schedule during a
well’s life. Just like the maintenance driven chemicals utilized during oper-
ations, chemicals serve numerous necessary functions to insure successful,
safe and efficient hydraulic fracturing operations. The following provides a
comprehensive look at common chemical additives utilized in the current
fracturing industry.
3.1.5 Common Fracturing Additives

First, there is no one formula for how much each of the following addi-
tives are used in a given fracturing fluid; however, the following section is
intended to present a brief description of some of the most commonly used
additives and a general percentage breakdown of each that has been widely
reported and is, therefore, easily verified by anyone wishing to do so.
Also, each well differs in the number, type and amount of additives (please
note: the term “additives” is used to include water, sand and chemicals to
allow for a discussion of each under one heading) in a successful fracture
treatment – “typically” between 3 and 12 additives depending on the condi-
tions of the specific well to be fractured and characteristic of the surround-
ing formations. Additives utilized in hydraulic fracturing operations are
intended to serve specifically engineered uses – such as biocides to control
microorganism/bacterial growth, corrosion inhibitor to prevent corrosion of
pipe, viscosity agents to carry proppant, gelling agents to improve proppant
placement, friction reduction to decrease pump friction and reduce treat-
ing pressure, oxygen scavengers to also aid in corrosion prevention in metal
pipes, and acids to help remove drilling mud buildup damage.
Fluids (typically water) – usually approximately 98%–99% of the

total volume – used to create the fractures in the formation
and to carry a propping agent (typically silica sand) which is
deposited in the induced hydraulic fractures to keep them
from closing up.
Hydrochloric acid (example is 15% HCl) – usually approxi-
mately 500 to 2,000 gallons per three thousand gallons of frac
fluid – used to help dissolve minerals and help remove dam-

age near the well bore by cleaning out cement around pipe
perforations, and also helps initiate fissures in the rock matrix.
Corrosion inhibitor (example is ammonium bisulfate) – usually
approximately 0.2%–0.5% of acid total volume, resulting in
approximately 5–10 gallons – used only in instances when
acid is used to prevent pipe corrosion.
Biocides (examples are sodium hypochlorite or chlorine dioxide) –
usually approximately 0.005%–0.05% of the total volume –
used to control bacterial growth in the water injected into
the well and prevent pipe corrosion.
Friction reducers (examples are polyacrylamide based com-
pounds) – usually approximately 0.025% of total volume –
used to reduce pipe friction and pressure in the piping
required to pump fluids.
Gelling agents (examples are guar gum and cellulose) – not often
used – used to thicken water-based solutions and help in
suspension and transport of prop-pants into formation.
Crosslinking agent (examples include boric acid, titanate and
zirconium) – used to enhance abilities of the gelling agent to
even further aid in transport of prop-pant material.
Breaker solution – when cross-linking additives are added,
a breaker solution is commonly added in the frac stage to
cause the enhanced gelling agent to break down into a sim-
pler fluid so it can be readily removed from the wellbore
without carrying back the sand/proppant material.
Oxygen scavenger (example includes ammonium bisulfate) –
used to prevent corrosion of pipe by oxygen.
Iron control and stabilizing agents (examples are citric acid and
acetic acid) – used to keep iron compounds in soluble form
to prevent precipitation.
Surfactant – usually approximately 0.5 to 2 gallons per thousand
gallons of frac fluid – is used to promote flow of the fluids
used in the fracturing process.
Scale Inhibitor (example is ethylene glycol) – seldom used – used
to control the precipitation of specific carbonate and/or sul-
fate minerals.
Proppants (examples are sand, resin coated sand or man-made
ceramic particles) – usually approximately 1%–1.9% of total
volume – used to hold fissures open so gas and oil can be
extracted.
Now, I am sure many of you that have seen this type of information before
are now expecting to see one of those “other uses” tables telling you that frack-
ing fluid must be safe due to the “ingredients” of fracking fluids having such
everyday uses such as: scale inhibitors having the same chemicals as wind-
shield washer fluid, friction reducers having the same chemicals as many
makeup products, surfactants being basically the same as shampoo products,
proppants being play sand and hydrochloric acid also being swimming pool
cleaner; these may be true in the strictest sense of the word, but this type of
listing can also be very misleading and insincere, in that most all chemicals can
be used for many different things, but are still not something with which you
necessarily want to come in contact. For example, ammonium nitrate is com-
monly used in agriculture as a high-nitrogen fertilizer, nitromethane is a com-
monly used industrial solvent and Ryder trucks are commonly used to move
families and their belongings to their dream homes – while these are also three
of the common “ingredients” used in the tragic April 1995 Oklahoma City
bombing of the Alfred P. Murrah Federal Building, which killed 168 people.
This is, of course, a comparison made for shock value, but it is meant as such to
stick in your memory as how these sorts of comparisons can be manipulated
and to drive home the fact that the best policy is to study upon facts when you
see a comparison like this, and make an informed decision for yourself.
3.1.6 Typical Percentages of Commonly Used Additives

Fracturing fluids are varied to meet the specific needs of each location;
however, evaluating the widely reported percentage volumes of the frac-
turing fluid components reveals the relatively small volume of additives
that are present. Overall, the concentration of additives in most fracturing
fluids is a relatively consistent 0.5% to 2%, with water and proppants mak-
ing up the remaining 98% to 99.5%. Keep in mind, however, that a typical
fracturing job uses upwards of five million gallons of fracturing fluid, so
a small percentage amount may actually result in a great deal of chemical
usage, no matter how diluted it may be.
As you can imagine, the overall composition of fracturing fluids var-
ies among companies and the drilling location. However, as a pretty good
baseline, fracturing fluids typically contain:
Approximately 90% water

Approximately 9.5% proppant materials
Approximately 0.5% chemicals – this percentage varies, but is
typically between
0.5–1.0% by weight of total fluid
Gelling
agent Scale pH adjusting
KCI 0.056% inhibitor agent
0.06% 0.043% 0.011%
Surfactant Breaker
0.085% 0.01%
Crosslinker
0.007%
Water Other
and sand Iron control
0.49% 0.004%
99.51%
Corrosion
inhibitor
0.002%
Biocide
Friction 0.001%
reducer Acid
0.088% 0.123%
Graph 1 Volumetric percentages of additives in fracturing fluids from Modern Shale Gas
Development in the United States.
As described in previous sections, the chemical additives are included

in fracking fluids to tailor the fluids to the requirements of the specific geo-
logical situation. The very popular chart above taken from Modern Shale
Gas Development in the United States demonstrates typical volumetric per-
centages of additives that were used for a typical hydraulic fracturing treat-
ment of a Fayetteville Shale horizontal well (Graph 1).
3.1.6.1 Proppants
Proppants are pretty hard to make into anything fun, exciting or entertain-
ing…as they are, for the most part, made up of sand or a manufactured
facsimile of sand. Sure, if you want to be poetic you can refer to proppants
as the only materials the operators want to remain downhole in the frac-
tures. If you really want to think poetically, feel free to consider a prop-
pant’s life as one of making its way from origins mined within the Earth
only to return to its final resting place deeper within the Earth’s fractures.
As discussed earlier, proppants are simply materials (typically silica sand,
resin coated silica sand, or manufactured ceramics) used to prop open the
open fractures to promote flow and eventual extraction of hydrocarbons.
As simple as prop-pants may seem, the estimated amount of proppant used
in industry has grown tenfold since 2000. In some regions it is not uncom-
mon to see upwards of four million pounds of proppant used per well and
represent up to 5% of well costs. The growth in proppant usage is generally
attributed to operators realizing better well completion techniques with
more proppant per stage and better well pad techniques with more laterals
and fracturing stages per pad.
Even considering the accelerated growth in the last decade, the evolu-
tion of proppant usage has been slow to develop over the industry lifetime
as a whole.
Consider that the first frac job was conducted in 1947, utilizing a
reported approximate 20,000 pounds of uncoated frac sand, and man-
ufactured ceramic proppant was not first used until 1983...or 36 years
later. Then, approximately one year later, resin coated proppant was first
introduced. As with most all technologies, as new techniques continue to
develop, proppants will surely evolve further to increase effectiveness and
efficiency in hydraulic fracturing.
No matter the type of proppant used, the most important characteristics
for a proppant are particle size distribution, crush resistance, shape and
sphericity (or roundness). Proppant materials are carefully sorted for size
and sphericity to provide an efficient conduit for production of fluid from
the reservoir to the well-bore. Grain size is critical because a proppant
must reliably fall within certain size ranges to coordinate with downhole
conditions and completion design (Figure 3.1).
Proppant shape and hardness qualities are also very important to the
efficiency and effectiveness of a fracturing operation. A coarser proppant
allows for higher flow capacity due to the larger pore spaces between
grains, but it may break down or crush more readily under high closure
stress, and rounder, smoother proppant shapes allow for better permeabil-
ity (Figure 3.2).
Another important quality that must be taken into consideration is its
hardness with respect to the formation. If the proppant is unable to embed
in the formation, something referred to as point load occurs, which leads
to higher flow capacity, but the proppant will break easier. However, if the
proppant is able to embed in the formation, it is referred to as embedment,
Sufficiently placed
& sized proppant No proppant
- Effective return - - No return -
Individual
fracture
Return flow
Insufficiently placed
& sized proppant
- Ineffective return -
Figure 3.1 Proppant size and placement. Courtesy of the authors.

Proppant roundness = Effective

return and strength
Proppant irregularity = Less

effective return and weakness
Figure 3.2 Proppant shape. Courtesy of the authors.
Point load proppant

Embedded proppant
Higher flow capacity, Lower flow capacity,

increased proppant fragility increased proppant strength
Figure 3.3 Proppant hardness. Courtesy of the authors.
which results in the load pressure spreading out over the proppant area,
increasing the breaking point but also lowering flow capacity. Embedment
is also a function of particle size (Figure 3.3).
Even though most all proppant materials are naturally occurring, includ-
ing manufactured ceramic proppants, with relatively low amounts of addi-
tional engineering necessary, the logistics in procuring and transporting
proppants can be daunting. Logistical considerations include coordination
of manufacturing material resources, transportation costs, and possibly a
substantial monetary investment in equipment necessary for processing
and material handling facilities.
3.1.6.2 Silica Sand

While the all-encompassing term for the material “sand” is generally used
for pretty much all forms of broken down granules of minerals or rocks,
to be specific falling between silt and gravel in the spectrum of sizes. There
are, however, many varieties of sand in the world, each with their own
unique composition and qualities. We all like to picture the white sandy
beaches of vacation destinations, for example, which are made up primar-
ily of limestone that has been broken down. Then there are also many black
sands either volcanic in origin or containing magnetite. Other sands have

high levels of iron in them, and so are rich and yellow in color.
The type of sand utilized for proppant materials is silica sand, which is,
by far, also the most commonly used type of proppant. Silica sand, unlike
many other “ingredients” of frac fluid, is more of a natural resource than
an engineered product. Silica sand proppant is, in a simplistic description,
made up of the most common mineral in the Earth’s continental crust…
quartz. Silica sand is simply quartz that over the years, through the work
of time and several erosion forces, has been broken down into tiny gran-
ules. Even though silica sand is a relatively common material, silica sand
used for proppant is a specifically selected and utilized product. Proppant
quality silica sand is a direct function of both the original depositional
environmental and some slight mechanical processing, if necessary. Silica
sand used for proppant is chosen for its round spherical shape and com-
monly graded particle distribution…unlike the common sand you might
find at the beach or on a playground, which often feels gritty when rubbed
between the fingers.
In addition to the oil and gas industry, there is some competition
between other industries for the bulk of silica sand, as industrial grade
silica sand has a wide range of uses. This resource is also commonly used
in the manufacture and preparations of various types of glass, in water fil-
tration, sand blasting, as fill and as an ingredient in industrial concrete,
in the metal casting industry to make cores and molds, and ironically it
is also used in the creation of highly flame-resistant industrial molds and
construction materials for the kilns used in the manufacture of the sintered
ceramic and bauxite proppants.
Even considering all the helpful and positive uses for silica sand across
several different industries, there are some possible hazards related to
its use. Because of the fine grains involved in silica sand, it can present a
health risk if not properly handled. Care must be taken to keep the silica
sand out of the lungs during use, and all materials containing more than
0.1% of silica sand must be clearly labeled. Workplace health applications
also need to be in place and enforced – failure to wear a proper respirator
or mask can result in lung irritation, and prolonged exposure can cause a
chronic condition known as silicosis.
Silicosis is a form of lung disease resulting from occupational exposure
to silica dust over a period of years – causing a slowly progressive fibrosis of
the lungs, impairment of lung function and even a heightened susceptibil-
ity to tuberculosis of the lungs. Silicosis can also progress and worsen even
after someone is no longer exposed to the silica dust, causing long term
effects and shortness of breath years later. Also, in the year 2000, the World
Health Organization determined crystalline silica is “associated with sili-

cosis, lung cancer and pulmonary tuberculosis” in classifying it as a Group
I carcinogen “based on sufficient evidence carcinogenicity in humans and
experimental animals.”
3.1.6.3 Resin Coated Proppant

As the name suggests, and to describe in the most simplistic of terms, resin
coated proppant is exactly that – silica sand coated with resin. Resin coat-
ing silica sand proppant is utilized for two main functions:
1. to spread the pressure load more uniformly to improve the

crush resistance of the silica sand particles
2. to keep pieces together that were broken due to high closure
stress from down hole pressure and temperature – this not
only prevents broken pieces from flowing into the borehole,
but also prevents these same broken pieces from returning
to the surface during flowback production operation.
Currently there are two types of resin coated proppants, Pre-cured and
Curable. Pre-cured is the “original” technology in which the resin coating on
the silica sand grains is fully cured prior to injection into the fractures. The
newer, curable technology has often been described, and I believe very well
described, as having a coating that is not completely “baked” or hardened.
Curable resin coated prop-pants are used at a little more than half cure so
that when the proppant is pumped downhole it can finish curing in the frac-
tures with down hole pressure and temperature. The advantage to curable
proppant technology is that it allows the individual proppant grains to bond
together in the fracture – resulting in the grains bonding together uniformly
in strength when temperature and pressure reach a appropriate levels.
3.1.6.4 Manufactured Ceramics Proppants

A third commonly used type of proppant is the manufactured ceramic prop-
pant. This is a proppant generally manufactured from a type of ceramic
material – typically non-metallurgic bauxite or kaolin clay. Bauxite is an alu-
minum ore from which most aluminum is extracted, while kaolin is one of
the most common minerals, occurring in abundance from chemical weath-
ering of rocks in hot, moist climatic soils like tropical rainforest areas.
Both bauxite and kaolin are utilized as proppants because of their
superior strength characteristics which are further enhanced through a
process known as sintering. The sintering process is conducted in high-

temperature kilns that are used to bake the bauxite or kaolin powder after
it has been made into specifically sized particles. This process decreases
the water content in the bauxite and kaolin to make them more uniformly
shaped for size roundness and spherical shape. The desired results of this
process are that the manufactured ceramic proppants can be engineered to
withstand high levels of downhole pressure (closure stress).
3.2 Additional Types

As more is learned through the ongoing processes and further advances
are made in technology, additional types of proppants are sure to come up.
One current trend is toward the usage of “waste” material – including glass,
metallurgical slags, and even rock cuttings produced to the surface during
oil and gas drilling. The re-use of rock cuttings from gas drilling operations
is especially attractive, since not only does it re-use a common waste prod-
uct in industry, but it is also utilizing sources indigenous to the locality –
which will cut down wastes while also cutting down on transportation and
overhead costs. However, the other possibilities are also quite attractive in
that agreements can be made with landfills, metallurgical operations and
glass companies to recycle and re-use their wastes in lieu of land filling.
3.3 Other Most Common Objections to Drilling

Operations
Now you will notice this discussion of common objections to drilling oper-
ations is a little different than other objections to drilling presented in this
book for two reasons:
1. there is no attempt to separate the operations related to

fracking from all other drilling operations – this is simply
because most nuisances related to one operation are the
same for all (for instance, additional traffic is additional traf-
fic no matter the origination), and
2. the nuisances described in this section are not written of or
discussed in a quantifiable way – in other words, this discus-
sion is not centered on an amount, but the simple fact that
it exists.
A reason for this is because a lot of data is collected for the other aspects
related to fracking operations to try and prove/disprove their existence,
when the nuisances discussed in this section are easily seen as in existence
(just spend a few minutes along any road used for drilling operations, and
this will become abundantly clear). Please keep in mind that there are
many additional nuisances absorbed by those living near drilling loca-
tions or related roadways, so this listing is far from comprehensive. The
following are merely what most see as the most common and are not pre-
sented in any order of magnitude. That decision has to be made by each
individual – one person may be more affected by noise, while another is
much more concerned with dust.
3.3.1 Noise
Noise conditions are usually one of the first things to change and be
noticed by local landowners. An increase in noise is also one of the most
continuous nuisances related to operations. Drilling and completing a
well – from the pad construction to the final completion of the well –
takes several weeks and utilizes many different types equipment. This
additional equipment can include additional trucking, construction and
drilling equipment. The noise concerns usually begin with the additional
traffic brought to an area during pad construction, then continue with
the noises associated with equipment and trucking required to construct
a pad, only to be followed by the large amount of noise related to rig
construction and operation throughout the well drilling process. Then,
once the well site is completed, there may come the additional sounds of
compressors used during ongoing production activities.
When you think of noise concerns related to the oil and gas industry,
the first thing that commonly comes to mind is the big noisy rig or maybe
the noisy traffic coming back and forth. These are, of course, very real and
valid concerns; however, the thing that is quite possibly the most notable
noise nuisance related to the oil and gas industry, due to length of time, is
the compressor. For the most part, the heavy rig work and heavy truck traf-
fic lasts approximately one to two months – while the compressor, while
not as loud, can continue for a much longer amount of time (months to
even years).
Gas compressors are normally the largest equipment remaining after the
well development process is complete and are utilized for something called
gas lift. Gas lift is used in wells that have insufficient reservoir pressure
to produce efficiently on their own. The gas lift process involves inject-
ing gas through the tubing-casing annulus to aerate the fluid to reduce
its density. Following aeration of the fluid the formation pressure is then
able to better lift the oil column up the wellbore. For pad sites where long
term compressor use is anticipated, especially in rural communities where
serenity is the norm and even the slightest ongoing noise can be heard
clearly for long distances, operators have addressed compressor noise con-
cerns with remote siting (trying to locate the compressors on the part of
the pad farthest from homes), noise tampering sound walls, and directing
compressors with fans away from homes. However, even with the measures
presently taken to mitigate ongoing sound issues, additional work must be
done and technology developed to work toward a solution.
3.4 Changes in Landscape and Beauty of

Surroundings
Several different types of pollution are commonly mentioned in relation to
the oil and gas industry – including water pollution, soil pollution, air pol-
lution and, as presented in the previous section, noise pollution. However,
one that may be overlooked to the majority of the public, but certainly not
overlooked to those affected, is visual pollution. Visual pollution is an aes-
thetic issue, “referring to the impacts of pollution that impair one’s ability
to enjoy a vista or view.” Now, with the possible exception of the immense
number of billboards lining our nation’s highways, not many things meet
the definition of visual pollution as much as a drill rig.
Drill rigs utilized in most unconventional well drilling typically can
range from approximately 50 feet to 100 feet in height. Couple the height
of the drill rig with the ongoing movement and dust related to drilling, and
it is easy to imagine how this would be bothersome to those adjacent to rig
locations. One mitigation attempt for this problem would be the usage of
lower height rigs. However, the undesirable trade-off for a lower height rig
is the necessary extended time on location for smaller rigs.
Ironically, horizontal drilling techniques commonly related to uncon-
ventional well drilling and hydraulic fracturing locations can actually be
considered a “semi-solution” to this problem. Pads used for horizontal
drilling commonly include multiple laterals on one location, in which the
drilling of multiple wells literally means moving the rig over as little as
twenty feet from one location on the pad to the next. This allows wells to
be drilled from one location without the necessity of moving the rig and
drilling in several locations – which would only disturb that many more
possible visual pollution points. This also allows for accelerated drilling
time due to lessened rig movement time, a reduction in the number of
necessary lease roads and drill pad locations, fewer necessary pipelines and
fewer tank batteries.
3.5 Increased Traffic

Another nuisance commonly cited by those living in oil production areas
is the drastic amount of added traffic it creates. This is not necessarily the
type of traffic most of us think of when we hear the word. Traffic related
to the oilfield includes all the initial traffic to bring in heavy equipment
for pad construction and eventually the rig itself, followed by traffic for
well completion and fracking activities (to get a taste for what this is like,
consider the amount of sand used in each frac job then consider how many
separate truckloads that would be), then the ongoing traffic related to haul-
ing produced water and oil from the locations until some sort of pipeline
infrastructure can be put in place.
Also, keep in mind that many of the areas affected by oil and gas oper-
ations are rural and do not, quite simply put, have the proper roadways
for the larger size or amount of traffic vehicles that come with industry
operations. Not only does the added traffic add additional wear and tear
to the local roadways, the narrower two-lane, and sometimes even more
narrow gravel roads, cause very unsafe driving conditions for the industry
and local resident vehicles alike.
The answers to the traffic problems may seem obvious – do something
to lessen the amount of traffic or do something to improve the roads –
but finding ways to turn those answers into reality is something much
more difficult than may first appear. The first, “do something to lessen the
amount of traffic,” would include:
1. the need to either use fewer (but larger) transport vehicles –

resulting in additional hazardous conditions with the larger
vehicles on the narrow rural roads, or
2. the need to install a pipeline infrastructure to transport pro-
duced water and/or oil – which comes with the obvious con-
cerns related to pipeline installation and location.
The second, “do something to improve the roads,” would depend on the
type of road to be improved. Improving and widening gravel type lease and
rural roads is a less daunting task than improving paved city/county roads
due to ease of obtaining the proper materials and the fewer restrictions put
on maintenance. However, making improvement to city and county roads
would include needing to clear a wider right of way all along the road to be
widened and the city/county would need to have the funds set aside for this
task…which is a time consuming process.
3.6 Chemicals and Products on Locations

Another common point of contention for residents in oil producing areas
is on-site storage of chemicals and products. For decades, one of the big-
gest drivers for public concern has been the identity and amounts of chem-
icals stored on pad locations during all phases of the well completion and
production process. This can include fuels used on location, the makeup
of drilling fluids (water based, oil based or synthetic), fracking chemicals
and additives stored on locations, chemicals kept on-site during ongoing
production, and even the products – produced water and/or oil – stored on
site as recovered from the well during operations.
The general public would be pretty surprised at how little chemical and
products are stored on site during construction, drilling and frack opera-
tions. For the most part, oilfield operations have become such a stream-
lined and efficient operation that operators will know how much of a
given chemical product will be necessary and, for the most part, make all
attempts to have the chemicals arrive on location as close to when needed
as possible to avoid storage. This process is beneficial to the operator in that
it cuts down on the time taken up by storing chemicals only to return for
them when needed, cuts down on waste from unused or outdated chemi-
cals, cuts down on equipment needed to maneuver chemicals if you have
them delivered directly to point of need, allows for more working space on
the pad, and also helps avoid a great deal of logistical problems related to
maneuvering equipment around storage areas.
Once production operations are in place and wells begin producing, the
fluids – produced water and oil – are often stored on site in large tanks while
awaiting transport off-site. Safeguards put in place to protect the environ-
ment and public from tank releases include consistent measurements by
pumpers, high level shut down sensors, continued equipment observations
and maintenance, and secondary containments in place around the tanks
to contain any fluids that may release. Secondary containments may be
constructed of properly packed and integrity tested earthen materials or up
to specifically designed and manufactured metal containments with plas-
tic liners. No matter the materials used in construction, secondary con-
tainments must be sufficiently large enough to contain all the fluids that
could possibly escape the tanks plus sufficient extra space for “worst case
scenario” rainfall. This amount is calculated for each region of the country
based on historic rainfall data.
Even with attempts to minimize the amount of on-site storage, some
chemical and product storage is unavoidable, and there are very valid
concerns, including potential spills, leaks, tank or container overfill, and
even the chance of traffic accidents on location or roadways leading to
releases of chemicals and/or products. Release events could range from
relatively small amounts from equipment leaks to possibly hundreds of
barrels from tank release. Two regulatory measures in place to manage and
oversee on-site chemical storage conditions are requiring Spill Prevention
Countermeasure and Control (SPCC) plans and SARA reporting.
SPCC plans are documents required by all facilities having the potential
to discharge oil to navigable waters of the U.S. and meeting one or both
of the following: greater than 1,320 gallons (31.4 bbls) aggregate above-
ground storage in equipment, drums, tanks, totes, tanks greater than
55 gallons in size; or, greater than 42,000 gallons total underground storage
capacity. Just to clarify, aggregate refers to adding up separate amounts of
all storage vessels…you can have one 1,320 gallon tank or ten 132 gallon
tanks and they would be equal under the SPCC requirements. Also, the
“having potential to discharge oil to navigable waters of the U.S.” is left up
to regulatory discretion to calculate, and has come to include pretty much
anywhere in the U.S. you could imagine. SPCC plans are, to keep it simple,
engineer-stamped documents that must be created for all facilities meet-
ing the above conditions that include a list of spill response procedures,
an emergency notification phone list, inspection procedures and schedule,
training requirements, site figures, site chemical and product storage vessel
types and sizes and containment calculations to prove sufficient contain-
ment is given to contain the largest possible spill amount.
SARA reporting, or possibly better known as the federal “right to
know,” requires quarterly and annual reporting of chemical storage
details (types of chemicals, amounts and dates of storage) for all facilities
which used more than 10,000 pounds per year of the chemical exceeding
the threshold quantity. This requirement means a facility storing more
than 10,000 pounds of a given chemical in a year must report that chem-
ical and amount. This program is intended as the “right to know” for
emergency responders and emergency services that may respond to an
emergency situation on the location so they will be able to adequately
prepare for what may be stored on site. The drawback of this program
as related to the oil and gas industry is that, with quarterly reporting,
by the time a chemical has been reported the oilfield function requiring
the chemical has normally been long complete and the chemicals are no
longer on site. This basically means that once the chemical is reported as
being on a location it is no longer there; however, as previously stated,
oilfield operations have become such a streamlined process that if you
know what has been reported for a previous location by a specified oper-
ator you can, for the most part, expect much the same chemicals and
products stored at following locations. If you are really curious about all
the chemicals used at a site, ask to receive a copy of the Material Safety
Data Sheet of the chemicals used.
3.6.1 Material Safety Data Sheets (MSDS)

Anytime a company produces for sale or uses a chemical a Material Safety
Data Sheet (MSDS) has to be written on the product and on file when
used. Occupational Safety and Health Administration (OSHA) estimates
that there are over 650,000 hazardous chemicals used daily in the United
States, and that hundreds more will be added this year alone. To address
the physical and health hazards of these chemicals, OSHA finalized the
Hazard Communication Standard (HCS) on November 25, 1983. The pur-
pose of the HCS is to “ensure that the hazards of all chemicals produced or
imported are evaluated, and that information concerning their hazards is
transmitted to employers and employees.” (29 CFR 1910.1200(a)(1)).
Employers are under obligation to use labels, MSDS, and other infor-
mation to evaluate both the physical and health hazards created by the use
of chemicals in their workplace, establish a program that addresses these
hazards and train workers to minimize their exposure. According to an
OSHA Executive Summary, “Chemical information is the foundation of
workplace chemical safety programs. Without it, sound management of
chemicals cannot occur. The HCS has made provision of hazard informa-
tion about chemical products an accepted business practice in the United
States. There is now a whole generation of employers and employees who
have never worked in a situation where information about the chemicals in
their workplace is not available.”
Manufacturers or importers of chemicals must create or obtain a
MSDS for every hazardous chemical that they produce or import (29 CFR
1910.1200(g)), and supply the appropriate one with a customer’s first pur-
chase, and any time the MSDS changes. (29 CFR 1910.1200(g)(6)(i)).
Employers are not required to evaluate information on a MSDS. (29 CFR
1910.1200(d)(1)). They do, however, have a duty to study and to use it to
“develop, implement and maintain... a written hazard communication pro-

gram” to ensure the safety of their workers (29 CFR 1910.1200(e) (1)). To
help address worker safety “at all times,” OSHA requires employers to make
MSDS “readily accessible during each work shift to employees when they
are in their work area(s).” (29 CFR 1910.1200(g)(8)). OSHA permits elec-
tronic and other forms of access to MSDS, as long as there are “no barriers to
immediate employee access in each workplace.” (29 CFR 1910.1200(g)(8)).
3.6.1.1 Contents of an MSDS

Frac site workers as well as anyone working in an industry or market that
uses chemicals will have access to an MSDS for any chemical that they
may have contact with. Interestingly enough, consumer products also have
MSDSs. In fact, your local hardware store has a complete file of MSDSs
for all the consumer chemicals they sell and many department stores do
as well. If you ever want to know the dangerous effects of a particular
insecticide or cleaner you can refer to the MSDS for detailed information.
Beware though; the information contained within an MSDS can be a bit
foreboding. Like pharmaceuticals and over-the-counter medications, the
warnings typically are meant to take into consideration any and all dan-
gers that may happen upon exposure. Without a working knowledge of
the terms and criteria put forth in the MSDS, the layperson could quickly
become horrified with the prospect of using a product only to experience
dizziness, dry mouth or shortness of breath (which seems to be the uni-
versal response to everything from aspirin to Zoloft ). The following is
an explanation which is provided to help you interpret the information
found on manufacturers’ MSDSs. While the format of these data sheets
varies from manufacturer to manufacturer, certain components appear
on each sheet.
3.6.1.2 Product Identification

The MSDS shall provide the name and address of the manufacturer and
an emergency phone number where questions about toxicity and chemical
hazards can be directed.
Product Name: Commercial or marketing name.

Synonym: Approved chemical name and/or synonyms.
Chemical Family: Group of chemicals with related physical and
chemical properties.
Formula: Chemical formula, if applicable; i.e., the conventional

scientific definition for a material.
CAS Number: Number assigned to chemicals or materials by
the Chemical Abstracts Service.
3.6.1.3 Hazardous Ingredients of Mixtures

The MSDS shall describe the percent composition of the substance, listing
chemicals present in the mixture. If it was tested as a mixture, it lists chem-
icals which contribute to its hazardous nature. Otherwise, it lists ingredi-
ents making up more than 1% and all carcinogens.
The OSHA permissible exposure limit (PEL), National Institute for
Occupational Safety and Health (NIOSH) recommended exposure limit
(REL), and/or the American Conference of Governmental Industrial
Hygienists (ACGIH) threshold limit value (TLV) will also be listed, if
appropriate.
The OSHA PEL is the regulated standard, while the others are recom-
mended limits. The PEL is usually expressed in parts per million parts of
air (ppm) or milligrams of dust or vapor per cubic meter of air (mg/m3). It
is usually a time weighted average (TWA) – concentration averaged over
an eight hour day. Sometimes, a short term exposure limit (STEL) may be
listed. The STEL is a 15 minute TWA which should not be exceeded. A
ceiling limit, (c), is a concentration which may not be exceeded at any time.
A skin notation means that skin exposure is significant in contributing to
the overall exposure.
3.6.1.4 Physical Data

The MSDS shall outline the physical properties of the material. The infor-
mation may be used to determine conditions for exposure. For example,
one can determine whether or not a chemical will form a vapor (vapor
pressure), whether this vapor will rise or fall (vapor density), and what the
vapor should smell like (appearance and odor). This could help determine
whether to use a fume hood or where to place ventilators. The following
information is usually included:
Boiling Point: temperature at which liquid changes to vapor state.

Melting Point: temperature at which a solid begins to change to
liquid.
Vapor Pressure: a measure of how volatile a substance is and
how quickly it evaporates. For comparison, the VP of water
(at 20°C) is 17.5 mm Hg, Vaseline (non-volatile) is close to

0 mm Hg, and diethyl ether (very volatile) is 440 mm Hg.
Vapor Density (air = 1): weight of a gas or vapor compared
to weight of an equal volume of air. Density greater than
1 indicates it is heavier than air, less than 1 indicates it is
lighter than air. Vapors heavier than air can flow along just
above ground, where they may pose a fire or explosion
hazard.
Specific Gravity (water = 1): ratio of volume weight of material
to equal volume weight of water.
Solubility in Water: percentage of material that will dissolve in
water, usually at ambient temperature. Since the much of
the human body is made of water, water soluble substances
more readily absorb and distribute.
Appearance/Odor: color, physical state at room temperature,
size of particles, consistency, odor, as compared to com-
mon substances. Odor threshold refers to the concentration
required in the air before vapors are detected or recognized.
% Volatile by Volume: Percentage of a liquid or solid, by volume,
that evaporates at a temperature of 70°F.
Evaporation Rate: usually expressed as a time ratio with ethyl
ether = 1, unless otherwise specified.
Viscosity: internal resistance to flow exhibited by a fluid, normally
measured in centistokes time or Saybolt Universal Secs.
Other Pertinent Physical Data: information such as freezing
point is given, as appropriate.
3.6.1.5 Fire & Explosion Hazard Data

The MSDS shall include information regarding the flammability of the
material and information for fighting fires involving the material.
Flashpoint: the lowest temperature at which a liquid gives off

enough vapor to ignite when a source of ignition is present.
Auto-ignition Temperature: the approximate temperature at
which a flammable gas-air mixture will ignite without spark
or flame. Vapors and gases will spontaneously ignite at lower
temperatures in oxygen than in air.
Flammable Limits: the lower explosive limit (LEL) and upper
explosive limit (UEL) define the range of concentration of
a gas or vapor in air at which combustion can occur. For
instance, an automobile carburetor controls this mixture –

too lean (not enough chemical) or too rich (not enough air,
as when you flood your engine) will not ignite.
Extinguishing Media: appropriate extinguishing agent(s) for the
material.
Fire-fighting Procedures: Appropriate equipment and methods are
indicated for limiting hazards encountered in fire situations.
Fire or Explosion Hazards: Hazards and/or conditions which
may cause fire or explosions are defined.
3.6.1.6 Health Hazard Data

The MSDS shall define the medical signs and symptoms that may be
encountered with normal exposure or overexposure to this material or its
components. Information on the toxicity of the substance may also be pre-
sented. Results of animal studies are most often given, i.e., LD50 (mouse) =
250 mg/kg. Usually expressed in weight of chemical per kg of body weight.
LD50 or lethal dose 50 is the dose of a substance which will cause the death
of half the experimental animals. LC50 is the concentration of the sub-
stance in air which will cause the death of half the experimental animals.
Health hazard information may also distinguish the effects of acute (short-
term) and chronic (long-term) exposure.
3.6.1.7 Reactivity Data

The MSDS shall include information regarding the stability of the material
and any special storage or use considerations.
Stability: “unstable” indicates that a chemical may decompose

spontaneously under normal temperatures, pressures, and
mechanical shocks. Rapid decomposition produces heat and
may cause fire or explosion. Conditions to avoid are listed in
this section.
Incompatibility: certain chemicals, when mixed may create haz-
ardous conditions. Incompatible chemicals should not be
stored together.
Hazardous Decomposition Products: chemical substances which
may be created when the chemical decomposes or burns.
Hazardous Polymerization: rapid polymerization may produce
enough heat to cause containers to explode. Conditions to
avoid are listed in this section.
3.6.1.8 Personal Protection Information

The MSDS shall include general information about appropriate personal
protective equipment for handling this material. Many times, this section
of the MSDS is written for large scale use of the material. Appropriate per-
sonal protection may be determined by considering the amount of the
material being used and the actual manipulations to be performed.
Eye Protection: recommendations are dependent upon the irri-

tancy, corrosiveness, and special handling procedures.
Skin Protection: describes the particular types of protective gar-
ments and appropriate glove materials to provide personnel
protection.
Respiratory Protection: appropriate respirators for conditions
exceeding the recommended occupational exposure limits.
Ventilation: air flow schemes (general, local) are listed to limit
hazardous
3.7 Conclusion
One hundred and fifty years ago when crude oil was first being extracted,
the damage done to the environment was nothing short of a nightmare.
In some areas little has changed but in many instances companies take
extraordinary precautions. Many years ago, we did not understand the
ramifications of pollution. Today, much work is underway to address what
is being understood as environmental concerns. The work that goes into
preparing a fracing well site today in assuring that the chemicals used are
innocuous while maintaining the physical integrity of the surrounding
land is taken into consideration primarily for legal reasons as well as busi-
ness concerns. If environmental laws are in place, then work shall be struc-
tured accordingly. The challenge is to enact law that makes sense according
to empirical evidence. It is also vital that those using the process and chem-
icals as well as the public understand the technology. Where industry in
general fails is when it rushes forward and only applies the letter of the law
and does not push for higher requirements. Legislation is often left to the
most verbal and most passionate. It is fair to say that when emotion runs
high, logic wanes. If we are to utilize the gifts that the Earth has bestowed
upon us then it is more than fair to assume a protective role going forward
and exploring proper protocols to ensure that the environment and com-
fort of life remain as balanced as it was found.
Bibliography
Fjaer, E. (2008). “Mechanics of hydraulic fracturing”. Petroleum related rock
mechanics. Developments in petroleum science (2 ed.). Elsevier. p. 369. ISBN
978-0-444-50260-5. http://books.google.co.uk/books?id=l6CfasFxhzYC&pg=
PA369&lpg=PA369.
Laubach, S.E.; Reed R.M., Olson J.E., Lander R.H. & Bonnell L.M. (2004).
Coevolution of crack-seal texture and fracture porosity in sedimentary
rocks: cathodoluminescence observations of regional fractures“. Journal of
Structural Geology (Elsevier) 26 (5): 967–982. doi:10.1016/j.jsg.2003.08.019.
http://www.sciencedirect.com/science/article/pii/S0191814103001858.
“Definition of frac job”. Oilfield Glossary. Schlumberger. http://www.glossary.oil-
field.slb. com/Display.cfm?Term=frac%20job.
Office of Air and Radiation U.S. Environmental Protection Agency (November
2010). “General Technical Support Document for Injection and Geologic
Sequestration of Carbon Dioxide: Subparts RR and UU Greenhouse Gas
Reporting Program”. Office of Air and Radiation U.S. Environmental
Protection Agency. http://www.epa.gov/cli-matechange/emissions/down-
loads10/Subpart-RR-UU_TSD.pdf.
Jack E. Whitten, Steven R. Courtemanche, Andrea R. Jones, Richard E. Penrod,
and David B. Fogl (Division of Industrial and Medical Nuclear Safety,
Office of Nuclear Material Safety and Safeguards (June 2000). “Consolidated
Guidance About Materials Licenses: Program-Specific Guidance About Well
Logging, Tracer, and Field Flood Study Licenses (NUREG-1556, Volume 14)”.
US Nuclear Regulatory Commission. http://www.nrc. gov/reading-rm/
doc-collections/nuregs/staff/sr1556/v14/#_1_26.
Anthony Andrews et al. (30 October 2009) (PDF). Unconventional Gas Shales:
Development, Technology, and Policy Issues. Congressional Research Service.
p. ?. http://www.fas.org/sgp/crs/misc/R40894.pdf.
Bill McKibben (8 March 2012). “Why Not Frack?”. The New York Review of Books 59
(4). http://www.nybooks.com/articles/archives/2012/mar/08/why-not-frack/.
David Biello (30 March 2010). “What the Frack? Natural Gas from Subterranean
Shale Promises U.S. Energy Independence--With Environmental
Costs”. Scientific American. http://www.scientificamerican.com/article.
cfm?id=shale-gas-and-hydraulic-fracturing.
(PDF) Chemicals Used in Hydraulic Fracturing (Report). Committee on Energy
and Commerce U.S. House of Representatives. April 18, 2011. http://
democrats.energy-commerce.house.gov/sites/default/files/documents/
Hydraulic%20Fracturing%20 Report%204.18.11.pdf
4
A Graph Theoretic Approach for Spatial
Analysis of Induced Fracture Networks
Deborah Glosser1,2* and Jennifer R. Bauer3,4
1
United States Department of Energy, National Energy Technology Laboratory,
Pittsburgh, PA, USA
2
Oak Ridge Institute for Science Education, Oak Ridge, TN
3
United States Department of Energy, National Energy Technology Laboratory
Albany, Oregon, USA
4
United States Department of Energy, National Energy Technology Laboratory, AECOM,
Albany, Oregon, USA
Abstract
Drilling induced fractures are generated when excessive stresses around a bore-
hole cause tensile failure of the wellbore wall. If stress concentrations are great
enough, compressive failures can form in the region surrounding the wellbore,
leading to wellbore breakout, and the potential compromise of wellbore integrity.
Another category of induced fracture networks are hydraulically induced frac-
tures, which are generated by the injection of pressurized fluids into the subsur-
face. Overlapping induced fracture networks between collocated wellbores may
increase pathways in the subsurface, and create the potential for unwanted fluid
leakage. The generation of induced fractures is greatly dependent upon the struc-
tural and geological characteristics. Probabilistic-based simulations are often used
to model fracture systems. Several methods for modeling local fracture networks
have been proposed in the literature. These models often involve the generation of
randomly located fractures, and may have limited capabilities for honoring engi-
neered fractures such as induced fracture networks. We present a graph theoretic
approach for identifying geospatial regions and wellbores at increased risk for sub-
surface connectivity based on wellbore proximity and local lithologic character-
istics. The algorithm is coded in Matlab, and transforms 3 dimensional geospatial
data to graph form for rapid computation of pairwise and topological relationships
between wellbores (nodes), and the spatial radius of induced fractures (edges).
*Corresponding author: deborah.glosser@netl.doe.gov

79
Induced fracture reaches are represented as cylinders with a radius r, based on

literature derived ranges for fracture lengths for different lithologies (e.g. shale,
sandstone). The topological algorithm is compared to a standard graph-based
k-nearest neighbor algorithm to demonstrate the value of incorporating lithologic
attributes in graph-based fracture models. The algorithms are applied to two sce-
narios using Pennsylvania wellbore and lithologic data: a subset of data from the
Bradford field, as well as a known leakage scenario in Armstrong County. The
topological algorithm presented in this paper can be used to complement exist-
ing fracture models to better account for the reach of induced fractures, and to
identify spatial extents at increased risk for unwanted subsurface connectivity. As
a result, the method presented in this paper can be part of a cumulative strat-
egy to reduce uncertainty inherent to combined geologic and engineered systems.
The model output provides valuable information for industry to develop environ-
mentally safe drilling and injection plans; and for regulators to identify specific
wellbores at greater risk for leakage, and to develop targeted, science-based moni-
toring policies for higher risk regions.
Keywords: Graph theory, spatial analysis, hydraulic fracturing
4.1 Background and Rationale

Commercial hydrocarbon drilling began as early as the 1800s in West
Virginia and Pennsylvania [1]. Since then, human engineering of the
subsurface has expanded worldwide to include gas storage [2]; CO2 injec-
tion [3]; unconventional resource exploration [4]; and injection of haz-
ardous waste [5]. As the character and degree of subsurface activities has
expanded, it has become increasingly important to develop methods and
techniques capable of addressing the interactions between engineered
features and local geology. Information provided by such techniques is of
critical value to both industry and regulators: leakage via wellbores and
fracture networks is a documented concern [6], and the data provided
by such methods is imperative for the development of environmentally
safe drilling and injection plans, as well as science-based monitoring and
plugging plans for regions (or particular wellbores) at greater risk of con-
nectivity and leakage.
Engineered and induced fractures, such as drilling induced fractures
(DIFs) and hydraulically induced fractures (HIFs), are important phe-
nomena that can occur when a wellbore is drilled. DIFs are generated
when stresses around a borehole are in excess of those required to cause
tensile failure of the wellbore wall [7]. If stress concentrations are great
enough, compressive failures can form in the region surrounding the
wellbore, leading to wellbore breakout, and the potential loss of wellbore
Spatial Analysis of Induced Fracture Networks 81
integrity [8]. HIFs are pressure induced fractures that are generated when
fluid is injected at high pressure into subsurface formations. If induced
fracture networks around collocated wellbores intersect, there may be an
increased likelihood of communication between wellbores, and the poten-
tial for unwanted fluid leakage between such networks. This is particularly
of concern for older, structurally unstable wellbores, which may lack ade-
quate casing or cementing necessary for zonal isolation [9]. Furthermore,
wellbore spatial densities are likely to be relatively high in regions with
historical drilling activity: Before regulations requiring minimum well-
bore spacing were implemented, it was a common practice to drill multiple
wellbores in close proximity (Figure 4.1). Furthermore, regions with exten-
sive drilling histories likely have higher wellbore densities as a result of the
multiple centuries of exploration of resources in those regions [1].
The generation of induced fractures is greatly dependent upon the
structural and lithological characteristics of local geology, which is often
difficult to accurately characterize in the absence of expensive geophysi-
cal surveys. Consequently, probabilistic-based simulations are often used
to model such fracture systems. Several methods for modeling local frac-
ture networks have been proposed in the literature [10–12]. These models
often involve the generation of randomly located fractures, with varying
degrees of user defined connectivity controls. Because of the importance
of wellbore locations; spatial densities; and the potential for overlapping
induced fracture networks to create fluid flow pathways, it is important to
Figure 4.1 Historical photograph of spatially dense wellbores in TItusville, PA (http://

www. acceity.org/2010/09/oil-in-them-thare-hills/).
account for the probable radius of influence of induced fractures around a

wellbore, as determined by the local geologic and geospatial attributes. A
model capable of incorporating these factors would complement advanced
fracture and fracture flow modeling methods. The data provided by such
an approach would allow for improved identification of spatial areas
at higher risk for communication between wellbores and geologic net-
works, and allow for the identification of specific regions and wellbores
at increased risk for unwanted leakage. Additionally, such data would
provide critical information needed for finer scale study or simulation of
these spatial extents and wellbores. The data provided by such an approach
supports a range of cumulative risk reduction and modeling strategies
due to improved uncertainty constrains regarding subsurface engineered-
geologic system characteristics. The information provided by this data is of
critical importance to both industry and regulators. Due diligence for well-
bore construction and injection plans – as well as monitoring and plugging
of wellbores – requires data on the relative likelihood of leakage potential
in particular areas and through specific wellbores. The model presented in
this paper supports these uses.
Because fracture and fracture flow models tend to cross into the realm
of big data, computational efficiency is an important consideration for
the development of an integrated method for modeling the relationship
between geologic and engineered systems. Traditional geospatial models
often require access to financially costly methods such as a geographic
information system (GIS). Such traditional geospatial models tend to be
either cell or raster based, and hence potentially computationally expen-
sive when applied to big data sets. The developing field of graph theory
allows for the representation, storage, and manipulation of geospatial data
in the form of graphs, which can provide improved computational effi-
ciency when certain graph theoretic data structures are employed [13].
Broadly speaking, graph theory is a field of mathematics and computer
science which involves the study of graphs [14]. Graphs are mathematical
structures that can represent real world data, and may be used to model
pairwise relationships between objects. A graph structure consists of ver-
tices (sometimes called “nodes”) and edges. Or, stated more rigorously, a
graph G, consists of two discrete sets, V (vertices) and E (edges). The pres-
ent work specifically uses what is called an “undirected graph,” in which
the elements of E are unordered pairs of vertices. The vertex set of a graph
G is denoted by V(G), and the edge set as E(G) [13]. Graphs are naturally
suited for visual representation of spatial relationships: Although the graph
structure itself can be either list or matrix based, graph diagrams – such
as the example shown in Figure 4.2 – show how elements in a graph lend
Figure 4.2 A drawing of an example graph structure showing vertices (blue circles) and
edges (orange lines).
themselves to visualization. The vertices in a graph can represent almost

any type of data (both abstract and discrete), and likewise, the edges can
represent multiple types of relationships between the vertices. Edges of a
graph are often associated with a weight function, w(e), that maps each
edge e in E to a number. The types of relationships represented by the edges
and their associated weights can be as simple as Euclidean distance [14], or
as complex as spatio-temporal interactions in complex networks [15, 16].
The graph-based “topological” model presented in this paper (coded in
Matlab) employs such graph structures in a novel method for characterizing
the spatial radius of influence of induced fractures around a wellbore, and the
spatial extents potentially at greater risk for unwanted fluid migration. The
algorithm is further compared and contrasted to a standard graph-based
k-nearest neighbor algorithm, to demonstrate the importance of incorpo-
rating lithologic factors into induced fracture and wellbore connectivity
models. The topological model can be used to complement existing frac-
ture models to better account for the reach of induced fractures around a
wellbore, and to identify potentially connected wellbores and spatial extents
for additional investigation as part of a cumulative strategy to reduce uncer-
tainty inherent to combined engineered and geologic systems.
4.2 Graph-Based Spatial Analysis

The general workflow for the model is shown in Figure 4.3, and referenced
and described in detail in the forthcoming sections. The model algorithm
is coded in Matlab and is developed for easy integration with other com-
mercially available software.
Acquire data
Define input
parameters
Create Create edges

vertices kNN approach Return
graph
Define topologic Generate fracture Create edges analysis
parameters propogation volume topologic approach
Figure 4.3 Workflow for preforming a graph analysis to spatially assess induced fractures.
4.2.1 Acquire Geologic Data and Define Regional Bounding

Lithology
Subsurface geology – particularly deeper lithology and structure – is
impossible to accurately characterize. Even with access to expensive geo-
physical surveys or map databases, it is often the case that only limited
information on the rock type or in situ structural characteristics of strata
at certain depth intervals is known [17]. However, regional-scale geolog-
ical databases (maps, surveys, core logs, stratigraphic columns, or well
logs) representing depth-dependent lithology and stresses are generally
freely available from state agencies [18] and other public sources. The
information provided by such resources can provide sufficient informa-
tion from which to determine the most geologically brittle and/or over-
stressed lithology in a given region. Such a lithology can be considered
to be the “bounding lithology”, or the regional geologic media that pro-
vides the physical bounding conditions for the maximum fracture radius
in that region. Arguably, these freely available data sources provide the best
information for determining the bounding lithology, given the inherent
ambiguities in interpretation of geophysical surveys (such as gravity or
magnetic anomalies). Furthermore, given the technological limitations of
resolving small scale fractures in geologic media (as well as the changes to
fracture networks caused by drilling), finer scale data sources would not
necessarily yield much more useful information regarding the present state
of the geology.
Once the bounding lithology is identified from the map surveys, a range
for the probable fracture radii can be assigned based off existing literature
(Table 4.1). One of the key features of the topological model described in
this paper is the adaptability of the analysis for user defined needs: Allowing
the user to select the range of fracture radii based on the bounding lithology
Table 4.1 Literature derived values for the average fracture radii for a bounding
lithology type [17, 31–34].
Shale Sand
Induced fracture radius (m) 67 113
and associated values reported in Table 4.1 (or other user defined values),
gives the model flexibility for a range of geologic conditions.
4.2.2 Details of the Topological Algorithm

The topological algorithm developed in this model is based on the principle
of cylindrical intersection. Wellbore point data are treated as graph nodes,
and are imported by the user. The wellbore points (x, y, z) are converted
to graph structure by the program. This is accomplished by transforming
the geographic coordinates of the wellbore data to an orthogonal, earth
fixed frame of reference and Cartesian coordinate system, so that curvature
effects can be simplified, and Euclidean distances can be calculated in com-
mon units such as meters [19]. Each wellbore node has an associated radius
of influence (r), based on the literature derived bounding lithology values
(as described in the previous section). The bounding lithology for each
wellbore is represented as a numeral in the Matlab program (1 for shale,
2 for sandstone, etc), and the data array for each wellbore node is tagged
with its bounding lithology value (alternatively, this numeral value can be
entered manually for each wellbore within Matlab, or it can be entered as a
user defined radius length value within the program). The bounding lithol-
ogy represents the most fracture prone strata that the wellbore penetrates,
and hence approximates the physical boundary of the potential fracture
reach of a wellbore intersecting that strata. Computationally, this is rep-
resented as a cylinder’s radius. The model computes the associated radius
of influence around each wellbore node in 3-dimensions. This results in a
series of finite cylinders of radius r, associated with each graph node. The
algorithm computes the intersection of the cylinders on the graph, and
when such intersections are identified, an edge is drawn between the origi-
nating nodes. This edge indicates that these originating wellbore nodes fall
within the potential fracture reach zone of each other.
4.2.2.1 Data Acquisition, Conditioning and Quanta

The non-zero graph entries for the topological algorithm are stored in the
form of a sparse matrix, making storage and manipulation of large data
quanta computationally feasible via compressed sparse row indexing.

However, certain practical data acquisition and conditioning procedures
are necessary to apply the algorithm to the geologic and geospatial data that
the algorithm is designed for. As described in prior sections, it is necessary
for the user to have a priori knowledge of the predominant geologic char-
acteristics in the subsurface. These data should be derived from geologic
maps, surveys, or well logs. In certain cases – particularly where a user is
running the algorithm over large spatial extents with multiple wellbores –
it may be necessary for the user to perform an initial GIS-based analysis
overlaying wellbore data with geologic layers to determine the bounding
lithology. The bounding lithology of each wellbore must then be con-
verted to a numeric representation (e.g. “1” for shale, “2” for sandstone),
and stored as an element in the row associated with the wellbore points,
before it is imported for analysis into the Matlab program. The topological
algorithm is also capable of allowing the user to define by hand, both the
length of the radius of influence around each wellbore, or simply select by
hand the bounding lithology from the pre-populated values within Matlab,
when running the algorithm for smaller datasets.
4.2.2.2 Details of the k-Nearest Neighbor Algorithm

Many traditional geospatial models rely on nearest-neighbor associations
to assess spatial relationships between geospatial features. For this reason,
a standard k-nearest neighbor algorithm is presented and applied to the
same data as the topological algorithm, as a means to demonstrate the value
added by the latter. Details of the knn algorithm can be found in multiple
literature sources [20–22]. The basic principle behind the knn approach
is that the data points (here, well-bores) exist in a metric feature space.
The algorithm is only distance based, and does not integrate any informa-
tion on lithology or wellbore spatial densities. The user must define the k,
or number of nearest neighbors around each wellbore, and the algorithm
makes a distance-based selection.
4.2.3 The Value of the Topological Approach Algorithm

To demonstrate the value added by the newly developed topological algo-
rithm, it is compared and contrasted to the standard knn algorithm. The knn
algorithm establishes an edge between each node and its k closest neigh-
bors (in Euclidean distance) for some user-specified integer k. The topo-
logical approach – which is the primary output of the method – considers
both geologic and topologic relationships by computing edges based on
bounding lithologies and associated radii of influence. Comparing and

contrasting the outputs from two algorithms serves to highlight the impor-
tance of accounting for subsurface geologic features in representing real
world geospatial data in graph form. A real world example is demonstrated
in the sections below.
4.3 Real World Applications of the Algorithm

4.3.1 Bradford Field: Contrasting the Graph-Based
Approaches; k Sensitivity
Commercial hydrocarbon exploration has been occurring in the Bradford
field in Pennsylvania since the 1800s [9]. The extensive drilling history in
the Bradford field results in a high spatial density of wellbores; particularly
older “legacy” well-bores which are likely to be structurally unsound and
poorly sealed [1] (Figure 4.4). The model is demonstrated using a synthetic
subset of spatially continuous well-bore data in this region.
Bradford oil & Gas field
Pennsylvania wellbores
Status of wellbores in bradford eld
Active
Plugged &/or abandoned
Other
N 0 1 .5 3 4. 5 6
Kilometers
Figure 4.4 Distribution and known status of wellbores within the Bradford Oil & Gas
field, Pennsylvania.
4.3.1.1 Data Sources

The Bradford PA wellbore data used in this analysis were obtained from
the National Energy Technology Laboratory, United States Department of
Energy (NETL/DOE), and were part of an aggregated dataset produced
using the methodology described by Dilmore et al. [9] and Glosser et al.
[1]. Initial data provenance for the dataset includes, aeromagnetic surveys
[23], digital databases [24, 25], and historical maps and minerals reports
[26, 27]. Bounding lithologies were determined either by the individual
wellbore records (where available), or by performing a geospatial overlay
of the wellbore locations with a freely available geologic map [28].
4.3.1.2 Results
The two algorithms – the knn algorithm and the topological algorithm,
were executed on these data. 50 wellbore locations from the Bradford field
were subsampled, and associated bounding lithology values for the well-
bores were chosen. The knn algorithm was run for three nearest neighbor
scenarios: k = 1 (Figure 4.5f); k = 2 (Figure 4.5e); and k = 3 (Figure 4.5d).
For the topological algorithm, the results are presented in 3 dimensions
with the induced fracture radius of influence for each wellbore (Figure
4.5a); in 2D form with the radius of influence (Figure 4.5b); and in simple
graph form showing only the edge connections (Figure 4.5c). Unlike the
knn approach – where edges are drawn to each wellbore (node’s) k nearest
neighbors, in the topological approach, the algorithm draws an edge if and
only if a wellbore (node) is within the induced fracture radius of influ-
ence of another wellbore (node). That is – if the induced fracture radius of
influence of wellbore nodes overlap, then an edge is drawn. It is apparent
from the results that the knn algorithm is, by definition, sensitive to the
number of neighbors, with the edge connections and overall graph con-
nectivity varying greatly based on this value. It is further apparent that
even when only one nearest neighbor is selected, the geometry of the edge
connections is considerably different than the geometry of the edges in the
topologic approach.
4.3.2 Armstrong PA: Testing the Algorithms Against a Known

Leakage Scenario
In March 2008, a pressurization of surface casing in a newly drilled oil and
gas well in Armstrong County, PA resulted in migration of fluids through
two other producing wells [29], mediated by connectivity in the subsurface
0
0.2
0.4
Depth
0.6
0.8 1.2
1 1 1.2
1.2 0.8 1
1 0.6 0.8
0.8 1 0.4 0.6
0.5 0.6 0.4
Long
0 0 0.2 0.4 Lat Long 0.2 0.2 Lat
(a) (b) 0 0
1.2
1 1
0.8 1.2 0.8
1
0.6 1 0.6 0.8
0.8
0.4 0.6 0.4 0.6
Long 0.4 0.4
0.2 0.2
0.2 0.2
(c) 0 0 Lat (d) 0 0
1.2 1.2
1 1
0.8 1.2 0.8
1.2
0.6 1 0.6 1
0.8 0.8
0.4 0.6 0.4 0.6
0.2 0.4 0.2 0.4
0.2 0.2
(e) 0 0 (f) 0 0
Figure 4.5 Representative graph output: Graph representation of wellbore points (n =50).
Blue dots represent wellbores (vertices), red lines represent edges, green ellipsoids
represent radii of influence (a) topological approach in 3D; (b) topological approach in 2D
form; (c) topological approach in 2D form (d) knn approach (k = 3) in 2D form (e) knn
approach (k = 2); (f) knn approach (k = 1). Graph axes are represented non-dimensionally.
fracture network. Like the Bradford field, Armstrong County has an exten-
sive regional history of oil and gas exploration. To test the performance
of the graph-based spatial analysis on a real world leakage scenario, both
the knn algorithm (k = 1) and the topological algorithm were applied to a
subset of wellbores near the pressurized well that caused the leakage event.
4.3.2.1 Data Sources

The leakage event (“Dayton Investigation”) was identified from a
Pennsylvania Department of Environmental Protection (PADEP) report
on oil and gas well stray gas cases [29], and the wellbore associated with
the event was located in the PADEP Office of Oil and Gas Management
Compliance Report [30]. A subset of 77 Armstrong County wells from the
NETL/DOE Pennsylvania wellbore dataset were spatially selected using
ArcGIS, and exported to a .csv file for import to the Matlab program.
4.3.2.2 Results
Both the knn (k = 1) algorithm and the topological algorithm were applied
to the subset of Armstrong County data. Results are shown in 2D form in
Figure 4.6. Both graph-based analyses identified a cluster of wellbores as
associated with the leakage event: However, the knn algorithm erroneously
identified subsurface connectivity between several more wellbores than
were reported as being associated with the stray gas leakage. Overall, the
Pennsylvania wellbores
Status of wellbores
Active
Dayton incident well
Plugged &/or abandoned
N Other
0 500 1,000 1,500 2,000
Meters Armstrong country, PA
Figure 4.6 Location of Dayton Incident Well.

knn algorithm (A) results in a qualitatively well connected graph, suggest-

ing that there is extensive subsurface connectivity in the region, even with
a k of 1. The wellbore associated with the leakage (black arrow) is shown
to be connected to at least 6 neighboring wells via overlapping induced
fracture networks. In contrast, in the topological algorithm (B and C), the
graph is far less connected; and the affected wellbore is connected to two
other affected producing wellbores. The topological algorithm also identi-
fied three other clusters of connected wellbores and fracture networks in
the region.
4.4 Discussion
The spatial locations of wellbores, as well as the geologic characteristics
of the reservoir, place physical controls on the formation of induced frac-
tures. Many fracture and fracture flow models do not honor wellbore loca-
tions and related engineered fractures; or, they consider only the spatial or
geologic attributes of the wellbores or reservoir in stochastically generating
random fractures. The importance of considering both spatial and geo-
logic attributes in identifying areas at greater risk for overlapping induced
fracture networks is highlighted by comparing the topological to a simple
distance-based nearest neighbor algorithm. Once identified, these regions
can be targeted for finer scale modelling, and used as part of a cumulative
modeling strategy to constrain uncertainty in the subsurface. The results
of the model are also of value as a standalone piece of data for the develop-
ment of science-based wellbore drilling, injection, and risk management
plans.
As shown in Figure 4.5, a nearest neighbor approach is by nature sensi-
tive to the selected k number of neighbors. The subgraphs produced by the
knn algorithm contrast considerably from the subgraphs produced by the
topological approach. In particular, the knn approach suggests far greater
connectivity between nodes than does the topologic approach, particularly
when higher values of “k” are chosen. Conversely, when low values of “k”
are chosen, several potential connections between nodes may be missed,
since the algorithm chooses the nearest neighbor of a node, and not its
range of influence on other nodes.
Since the purpose of this model is to identify probable regions of sub-
surface connectivity – and wellbores and wellbore clusters at greater risk
for unwanted fluid migration – the knn approach gives a substantially less
accurate representation of the spatial extents that are likely to contain over-
lapping induced fracture networks. First – the k is user defined, and the
algorithm output is greatly dependent on the user selection. Second – the

knn algorithm does not take into account the geologically-based induced
fracture radius in drawing edges: This results in a connected graph, which
is falsely suggestive of subsurface connectivity, and hence, higher risk for
fluid leakage through the subsurface and through wellbores. In neglecting
to consider geophysical attributes, knn relies solely on wellbore neighbor-
hoods, and not actual induced fracture networks in defining connectivity.
In contrast, the topological approach considers both geology and well-
bore spatial locations in assessing the probability of induced fracture net-
work overlaps. Edges are drawn if and only if a wellbore falls within the
radius of influence of a nearby wellbore. Hence, the subgraphs are repre-
sentative of both geologic and wellbore networks, and give a more realistic
(and in a sense, more conservative) representation of probable connectivity.
Furthermore, unlike knn, the topological approach is not susceptible to user
(a) (b)
(c)
Figure 4.7 (a) Results of knn algorithm on Dayton PA area wellbore (k = 1) (n = 77);
(b) Results of topological algorithm on Dayton PA wellbores (n = 77); (c) Results of
topological algorithm on Dayton PA wellbores, zoomed in on wellbores associated with
leakage. Black arrows point to overpressurized wellbore. All results are shown in 2D form
for clearer visual representation of results.
bias in the selection of neighbors: the algorithm itself automatically identifies

the induced fracture radius and number of associated connections (edges),
based on literature supplied values for lithologic fracture reach.
When tested against a known leakage scenario, the topological algo-
rithm identifies two producing wellbores associated with the leakage event
(Figure 4.7b and 4.7c). Although the knn algorithm also identifies affected
wellbores, it also falsely identifies several other nearby wellbores as being
connected (Figure 4.7a). In addition, the overall knn graph is qualitatively
connected, resulting in a likely erroneous (overly connected) representa-
tion of the subsurface induced fracture network connectivity, even with
a k value of 1. In contrast, the topological algorithm identifies 4 clusters
of connected wellbores (Figure 4.7b), including the wellbores associated
with the leakage event. The overall topological graph is qualitatively less
connected than the knn graph, and appears to provide a more realistic
representation of the connectivity of wellbores and induced fracture net-
works. The example application serves to highlight the importance of the
geologic factors in assessing regions with probable natural and engineered
flow pathways.
4.4.1 Uses for Industry and Regulators

There is an ongoing need in both industry and regulatory domains for
science-based tools from which to develop risk management programs.
This is particularly true in complex systems such as subsurface geologic
systems, especially those impacted by hydraulic fracturing. The potential
for subsurface leakage through induced fracture networks and existing
wellbores is a concern to regulators, operators, and public stakeholders.
The ability to provide better predictive tools for spatial regions or wellbores
at risk for such events meets a critical need for the development of sound
risk management strategies. The method presented in this paper can pro-
vide valuable information to stakeholders, and helps to reduce uncertainty
inherent to these complex systems.
4.5 Conclusions
Overlapping induced fracture networks between collocated wellbores may
increase communication in the subsurface, and create the potential for
unwanted fluid flow. The generation of induced fractures is greatly depen-
dent upon the structural and lithological characteristics of local geology,
which is often difficult to accurately characterize in the absence of costly
geophysical surveys. A robust, adaptable method for analyzing the spatial

regions and wellbores at higher risk for subsurface induced fracture connec-
tivity has been developed and presented. The result produced by the method
is based on geologic data, and provides a sound basis for reduction of uncer-
tainty inherent in subsurface systems. It is shown that the topological graph
theory algorithm is a potentially powerful tool for rapid characterization of
subsurface geospatial data. The topologic graph algorithm has several advan-
tages over the knn algorithm: it is not susceptible to user error in the selec-
tion of a “k”. And, in accounting for geologic factors, it provides a physically
based, and hence more realistic, assessment of probable subsurface connec-
tivity. The algorithm has been successfully demonstrated using a real world
leakage scenario. Because subsurface modeling efforts tend to occupy the
realm of “big data,” the method increases modeling efficiency in two ways:
First, the graph structures employed by the method allow for rapid computa-
tions involving big data sets; and second, the method can be used to identify
spatial extents at greater risk for induced fracture network communication,
and hence targeted fracture and fracture flow modeling. The method output
can be transformed back into a geographic coordinate system, and/or inte-
grated into existing fracture or fracture flow modeling software, as part of a
cumulative modeling strategy for risk mitigation. The information provided
by this approach can be used by regulators and industry in developing sound
risk management plans related to hydraulic fracturing operations.
Acknowledgements
This work was completed as part of National Energy Technology Laboratory
(NETL) research for the Department of Energy’s Complementary Research
Program under Section 999 of the Energy Policy Act of 2005. This
research was supported in part by an appointment to the National Energy
Technology Laboratory Research Participation Program, sponsored by the
U.S. Department of Energy and administered by the Oak Ridge Institute
for Science and Education. The authors wish to acknowledge the techni-
cal feedback from Kelly Rose, Russell Schwartz, Circe Verba, and Grant
Bromhal. Deborah also thanks her family for their support, with special
recognition to Miriam and Michael Miller.
References
1. D. Glosser, K. Rose, and J. Bauer, Spatio-temporal analysis to constrain
uncertainty in wellbore datasets: an adaptable analytical approach
in support of science-based decision making. Journal of Sustainable

Energy Engineering, in press (2016).
2. State Oil and Gas Groundwater Investigations and Their Role in
Advancing Regulatory Reforms a Two-State Review: Ohio and Texas,
Groundwater Protection Council (2011).
3. K.M Keranen, H.M. Savage, G.A. Abers, and E.S. Cochran, Potentially
induced earthquakes in Oklahoma, USA: Links between wastewater
injection and the 2011 Mw 5.7 earthquake sequence. Geology 41(6),
699–702 (2013).
4. D. Soeder, S. Sharma, N. Pekney, L. Hopkinson R. Dilmore, B. Kutchko,
et al., An approach for assessing engineering risk from shale gas wells in
the United States. In International Journal of Coal Geology, Special Issue:
Unconventional Natural Gas and the Environment (2014) (in press).
5. W. Ellsworth, Injection induced earthquakes. Science. 341(6142),
1225942 (2013).
6. J. A. Montague and G. F. Pinder, Potential of hydraulically induced
fractures to communicate with existing wellbores. Water Resources
Research 51(10), 8303–8315 (2015).
7. M. TIngay, J. Reinicker, and B. Muller, Borehole breakout and
drilling-induced fracture analysis from image logs. World Stress Map
Project (2008).
8. M. D. Zoback, D. Moos, and L. Mastin, Well Bore Breakouts and in
Situ Stress, J. Geophys. Res., USGS Staff Published Paper, 90, B7, pp.
5523–5530 (1985).
9. R.M Dilmore, J.I. Sams, D.B. Glosser, and K.M. Carter, Spatial and
temporal characteristics of historical oil and gas wells in Pennsylvania:
implications for new shale gas resources. Environmental Science and
Technology 49, 12015–12023 (2015).
10. National Energy Technology Laboratory, FracGen and NNFlow soft-
ware v. 14.9, https://edx.netl.doe.gov/dataset/fracgen-and-nffow-
version-14-9 accessed 4/11/2016.
11. P. Valko and M.J. Economides, Propagation of Hydraulically Induced
Fractures – A Continuum Damage Mechanics Approach. International
Journal of Rock Mechanics, Mineral Science, and Geochemistry 31(3)
221–229 (1994).
12. C. Zeeb, E. Gomez-Rivez, P. Bons, S. Virgo, and P. Blum, Fracture
network evaluation program (FraNEP): A software for analyzing 2D
fracture trace-line maps. Computers and Geoscience 60, 11–22 (2013).
13. J. Pach, Geometric Graph Theory, Mathematical Institute of Hungarian
Academy of Sciences, http://dcg.epf.ch/page-117408-en.html accessed
4/11/2016.
14. L. D. Cohen and R. Kimmel, Global minimum for active contour

models: A minimal path approach. International Journal of Computer
Vision 24(1), 57–78 (1997).
15. J. Pereira-Leal, A. Enright, and C. Ouzounis. Detection of functional
modules from protein interaction networks. PROTEINS: Structure,
Function and Bioinformatics 54, 49–57 (2004).
16. S. Yook, Z. Oltvai, and A. Barabasi, Functional and topological char-
acterization of protein interaction networks. Proteomics 4, 928–942
(2004).
17. M. Alizadeh, Z. Movahed, R. Junin, R. Mohsin, M. Alizadeh, and
M. Alizadeh, Finding the Drilling Induced Fractures and Borehole
Breakouts Using Image Logs. Journal of Advanced Research 10(1),
9–30 (2015).
18. West Virginia Geologic and Environmental Survey, Interactive
Mapping http://ims.wvgs.wvnet.edu/index.html (2015).
19. N.-N.G. Intellingence Agency of EUA, World Geodetic System 1984,
WGS-84. p. 3 (1984).
20. W. Dong, C. Moses, and K. Li, Efficient k-nearest neighbor graph con-
struction for generic similarity measures. WWW 577–586 (2011).
21. V. Garcia, E. Debreuve, and M. Barlaud, Fast k nearest neighbor
search using GPU. 2008 IEEE Comput. Soc. Conf. Comput. Vis. Pattern
Recognit. Work. 2, 1–6 (2008).
22. Y. P. Mack and M. Rosenblatt, Multivariate k-nearest neighbor density
estimates. J. Multivar. Anal. 9(1), 1–15 (1979).
23. R. Hammack, G. Veloski, and J. Sams, Rapid Methods for Locating
Existing Well Penetrations in Unconventional Well Development
Areas of Pennsylvania, Unconventional Resources Technology
Conference, 20–22 July, San Antonio, SPE-178558-MS (2015).
24. PA DCNR. The Pennsylvania Internet Record Imaging System/Wells
Information System (PA*IRIS/WIS): Gateway to Detailed Oil and Gas
Well Information Data http://www.pairis.state.pa.us accessed during
2013.
25. Pennsylvania Spatial Data Access Center, Oil and Gas Wells in
Pennsylvania, http://www.pasda.psu.edu/ accessed during 2013.
26. R. Arnold and W.J. Kemnitzer, Petroleum in the United States and
possessions: a presentation and interpretation of the salient data of
geology, technology, and economics of petroleum in each state and
possession treated according to the conventional major field divi-
sions; Harper & brothers (1931).
27. US Geological Survey. Minerals Yearbook Digital Collection http://
digital.library.wisc.edu/1711.dl/EcoNatRes.MineralsYearBk.
28. Pennsylvania Spatial Data Access Center, Geology of Pennsylvania,

http://www.pasda.psu.edu/ accessed during 2015.
29. Department of Environmental Protection, Stray Natural Gas
Migration Associated with Oil and Gas Wells http://www.dep.
state.pa.us/dep/subject/advcoun/oil_gas/2009/Stray%20Gas%20
Migration%20Cases.pdf Accessed during 2016.
30. DEP Office of Oil and Gas Management Compliance Report, inspec-
tion ID 1692786 http://www.depreportingservices.state.pa.us/Report
Server/Pages/ReportViewer. aspx?/Oil_Gas/OG_Compliance accessed
during 2016.
31. M. Brudy and M.D. Zoback, Drilling induced tensile well fractures:
implications for determination of in situ stress orientation and magni-
tude. International Journal of Rock Mechanics and Mining Sciences 36,
191–215 (1999).
32. J.R. de Druezy, P. Davy, and O. Bour, Hydraulic properties of two
dimensional random fracture networks following a power law length
distribution. Water Resources Research 37(8), 2065–2078 (2001).
33. N.E. Odling, Scaling and connectivity of joint systems in sandstones
from western Norway. J. Struct. Geol. 19, 1257–1271 (1997).
34. O. Bour and P. Davy, Connectivity of random fault networks follow-
ing a power law fault length distribution. Water Resour. Res. 33, 1567–
1583 (1997).
Part 3
OPTIMUM DESIGN PARAMETERS

5
Fracture Spacing Design for Multistage
Hydraulic Fracturing Completions
for Improved Productivity
D. Maity*, J. Ciezobka and I. Salehi
Gas Technology Institute, Des Plaines, Illinois, United States
Abstract
Over the years, productivity studies conducted for horizontal multistage comple-
tions have shown significant stage-wise variability. Optimizing such completions
could hold the key to unlocking true value from shale reservoirs and improving
well economics. Traditional hydraulic fracturing programs use the same fracture
design along laterals without any consideration for changes in the reservoir and
wellbore conditions. Methods using mechanical rock properties require expen-
sive petrophysical logging data, while those involving use of drill cuttings can be
highly resource intensive and time consuming. In this paper, we introduce a novel
approach, which utilizes routinely available data such as measurements made
while drilling and petrophysical data as available within a fracture spacing design
framework. We validate our approach through application on multiple wells by
comparing results from our workflow with post completion production logging.
Finally, we highlight the potential advantages and pitfalls in our approach and
present a roadmap for future implementation in different plays.
Keywords: Artificial intelligence, hydraulic fracturing, fracture spacing, fracture

design, drilling data
5.1 Introduction
While the advent of multistage hydraulic fracturing in long lateral com-
pletions has revolutionized shale oil and gas production, the process still
*Corresponding author: debotyam.maity@gastechnology.org

101
lacks the robust understanding of what happens downhole within the res-
ervoir and what leads to the significant variability in productivity from
completed stages. A significant amount of work has gone into improving
our understanding of these completions, and some of these observations
have provided enough information to try and improve fracture spacing
design. Traditional approach to completion design involves use of tran-
sient rate-time analysis to identify key design parameters (permeability
and fracture half lengths) and using them to predict well performance [1].
Stage spacing is a critical design parameter, which is impacted by consid-
erations of reservoir permeability [2], stress shadowing effects [3, 4], SRV
considerations [5], and economic considerations such as net present value
[6]. Other more elaborate techniques at optimizing stage spacing include
use of microseismic data [7], and fracture network modeling [8] to name
a few. In practical applications, what is desirable is to take a holistic view
of the completion process and utilize as much data and analysis as pos-
sible for design [9]. While holistic design techniques are in place, most
methods do not account for variability in reservoir and completion effects
along the long laterals. From production logs and distributed acoustic and
temperature sensing data, we know that many clusters show insignificant
to no production, creating zones with very low productivity. This clearly
indicates that the “one size fits all” approach creates sub-optimal fracture
design, and this has been abundantly recognized by the industry [10]. We
believe that while design issues (such as fracturing efficiency) are import-
ant; formation quality is critical, as sections with lower quality should have
modified fracture density to provide for adequate drainage.
Many novel approaches have been suggested in the past few years,
which involve a thorough investigation of the reservoir properties and
the geomechanical aspects of completion in shale formations. One recent
example highlighting an engineered fracture spacing design approach uses
characterization of reservoir and completion quality, which are used to
predict proper stage placement [11]. Microseismic monitoring and other
geophysical tools can also allow for improvements in design based on
observations [12]. Generating pseudo logs for lateral sections of wellbore
based on observations from the vertical pilots is an established technique
[13] for understanding laterals and improving associated completions.
Methods looking at fracability alone and utilizing stochastic optimization
techniques have been evaluated [14]. The need for running wireline petro-
physical logs for at least the vertical pilots, and the need for core analysis
and correlation are well understood. These are not routinely available, and
therefore, we felt there was a need for a technique that can be applied on
any well to provide a quick optimal design suggestion based on the historic
Multistage Hydraulic Fracturing Completions 103
field data available for the play in question and the mud log data from the
well under consideration.
For this study, we wanted to devise a technique that can systematically
distribute fracture stages for more effective drainage of the reservoir with-
out the use of expensive wireline or logging while drilling data. What is
unique about our completion design approach is the use of mud log data
for completion design in the absence of any wireline petrophysical or
geomechanical data from that area. This is expected to work reasonably
well, provided a predictive model for rock properties can be developed. In
our approach, we use a hybrid AI (Artificial Intelligence) based modeling
workflow to predict geomechanical properties where stage spacing design
utilizes mud log gas shows, as well as predicted geomechanical rock prop-
erties, within a predefined design framework.
5.2 Method
For most new field development programs in unconventional plays, verti-
cal pilot wells are drilled, cored, and logged in order to gain a robust under-
standing of the formation before completions can be designed. These pilots
provide valuable insights into the rock, including the mineralogy, in situ
stress state, organic content, lithology, porosity, etc. to name a few. This
wealth of data can be used to predict the behavior of the well laterals drilled
from the pilot. Using data from such pilots, we propose to design predic-
tive models for reservoir properties, which can be obtained from mud log
data alone. This allows for wider applicability of the design methodology
without compromising on upfront well completion costs. We propose a
hybrid neural network (Neuro-Fuzzy) workflow, which uses mud log data
and geomechanical predictive models to design a fracture density model,
which can then be used to place fracture clusters.
Mud logs typically provide estimates for observed gas shows, Gamma
ray for geo-steering purposes, and rate of penetration data. We understand
that Gamma ray logs can provide indications of shale layers, which have
higher natural radioactivity. Gas shows could indicate possible productive or
non-productive zones and also potential naturally fractured zones. However,
the observed gas shows are influenced by rate of penetration which in turn
can be impacted by a multitude of factors, not all of which are due to reser-
voir conditions. The gamma ray tool is also influenced by the erratic drilling
speeds and varying wellbore conditions encountered during drilling, in gen-
eral. In order to develop the suggested design framework, we have to answer
some important questions posed at the outset. Most importantly we want to
see how these parameters relate to zonal productivity potential, whether the
impact is verifiable, and which parameters are needed for reasonable design
solutions. To find these answers, we use available data from multiple wells
from the Marcellus shale play, and verify through observations the necessary
framework for the proposed design approach.
5.2.1 Impact of Natural Fractures

The initial step is to understand how some factors may play an important role
within our fracture spacing design framework. Marcellus and other shale plays
are known to have varying natural fracture distributions, and depending on
the in-situ condition as well as the properties of the injected fluid/proppant,
these could significantly enhance the productivity of the stimulated well. In
the Marcellus play, prior data suggests presence of natural fracture swarms as
a result of local stress perturbations occurring over geologic timelines [15].
These natural fracture swarms are known to contribute significantly to overall
production by providing additional surface area for gas to move from matrix
to the connected fractures and eventually to the producing well.
Identification of naturally fractured zones is a key element in accurate
understanding of well behavior but this is not easy to achieve due to the
need for use of indirect measurement techniques or proxies to identify the
zones where the reservoir is fractured a-priori. While there are many avail-
able techniques for fracture characterization in reservoirs, we use available
microseismic data from one of the wells under study (henceforth Well #1)
to characterize fractures. This is made possible due to the ways in which
hydraulic fractures interact with naturally fractured rock and the impact
such interaction has on the final fractured rock volume in terms of fracture
network complexity, fracture network dimensions, and magnitude distri-
bution of the microseisms [16].
In this context, we look at two different properties evaluated based on
the distribution of induced microseismicity associated with the hydraulic
fracturing process. The first is the b value distribution, which is obtained
from the Gutenberg-Richter law, providing the relationship between the
magnitude of the seismic event and the total number of earthquakes in any
given region and time period of at least that magnitude [17]. Higher b value
is indicative of a larger portion of small earthquakes compared to bigger
ones. Since in the presence of natural fracture swarms, many re-activations
are expected, b values tend to be higher when hydraulic fractures inter-
act with such zones [18]. In this study, we look at the overall distribution
of events and their b value estimates for every completed stage and try
to interpret post completion production logs. We expect zones showing
higher b value to be indicative of presence of natural fractures and there-

fore, they should correlate strongly with gas shows and production log data
run for some of these test wells. Similarly, higher length-to-width ratio (or
the ratio of the two principal dimensions of the microseismic event cloud)
is indicative of an elongated perturbed zone with lower degree of com-
plexity in the created fracture network. On the other hand, a lower length-
to-width ratio suggests a more complex network, which could be due to
substantive interaction of the propagating hydraulic fractures with natural
fractures [19]. Figure 5.1 shows length-to-width aspect ratio as mapped
1.5
Normalized measure
Microseismic derived “b–value”

1
0.5
0
0 1000 2000 3000 4000 5000 6000
(a)
1.5
Normalized measure
Microseismic derived W/L ratio

1
0.5
0
0 1000 2000 3000 4000 5000 6000
(b)
1.5
1 R² : 0.81132
Norm. L/W ratio
0.5
−0.5
−0.5 0 0.5 1 1.5
(c) Norm. b-value
Figure 5.1 Stage-wise distribution of microseismic derived (a) ‘b value’ and (b) length-to-
width ratio for a study well. Subplot (c) shows cross plot of the two showing a strong
positive correlation between the two parameters.
with borehole microseismic data and how it correlates with evaluated b

values for the same stages.
Figure 5.2 shows another example where b value has been compared with
production logs and other relevant data to highlight these observations and
how they correlate with mud log gas shows. We observe a poor correlation
between b values and observed fracture density from image logs, which is
expected as image logs are subject to interpretation errors (and so is b-value
analysis). However, image logs only provide a snapshot of fractures at the
wellbore unlike b-value, which defines the spatio-temporal seismicity dis-
tribution. We observe reasonable correlation between sections showing very
high flow contribution and sections indicating highly complex fractured
zones from b value distribution. Finally, we observe a reasonable correla-
tion between production log and highly fractured sections of the reservoir, as
well as a reasonably strong correlation between production log and high gas
composition from mud log gas show data. The observed correlation between
mud log gas shows and production log data over certain depth intervals has
been observed for multiple wells and provides one element governing our
stage/cluster spacing optimization workflow. Even though the correlation is
not perfect, in conjunction with gamma log readings and rate of penetration
data, a strong correlation between the observed production and modeled
geomechanical properties governing production in shale reservoirs should
be possible as it may take care of some of the outlier observations.
1.5
measure
1
Norm.
0.5
0
(a) 0 1000 2000 3000 4000 5000 6000
1.5
measure
1
Norm.
0.5
0
(b) 0 1000 2000 3000 4000 5000 6000
1.5
measure
1
Norm.
0.5
(c) 0 1000 2000 3000 4000 5000 6000
20
% G as
10
0
(d) 0 1000 2000 3000 4000 5000 6000
Lateral length (ft.)
Figure 5.2 (a) Microseismic derived ‘b value’, (b) image log derived fracture density,
(c) mud log gas shows (red: 1 and blue: 0), and (d) production log for a study well. The
properties (except production log data) are normalized between 0 and 1. The black insert
in subplot (d) indicates the section of the wellbore where production log could not be run.
Based on the observations, our design workflow involves utilizing rele-

vant routinely logged data from mud logs (gas shows, rate of penetration,
and gamma ray) and models for rock properties such as Young’s Modulus
and Poisson’s Ratio using data from Well #1. These in turn are used to pre-
dict rock brittleness, which we correlate with another brittleness measure
from lithological distribution to validate the brittleness function before
actual use. This is because lithology has an impact on rock properties
and therefore, rock brittleness can be considered a function of Young’s
Modulus and Poisson’s Ratio [20]. Broadly speaking, Increasing Young’s
Modulus or decreasing Poisson’s Ratio is indicative of more brittle forma-
tions. This modeled brittleness is then used in conjunction with gas shows
to identify the optimal hydraulic fracture/cluster density along the lateral.
The basic framework governing our design is to provide for more cluster
density in regions susceptible to lower productivity behavior in order to
improve overall production from the completed lateral. Based on this frac-
ture density model, clusters are populated along the length of the lateral by
honoring the background modeled density values.
5.2.2 Workflow
Apart from the pre-completion drilling data, the proposed workflow
requires some rock properties derived from specialty logging such as dipole
sonic or spectral azimuthal gamma for representative lithologic layers. This
data is typically available for a given field, especially new development felds
where a single or multiple pilot wells are drilled before a full field develop-
ment drilling program is implemented. This is necessary to model for the
same properties based on routine mud log data by deriving the necessary
models. The workflow involves the following steps:
1. Training wells are nominated based on availability of rele-

vant specialty logging data and the exact position of the well
in relation to various shale sub-layers. The original high-
resolution mud log data is inverted by passing through a low
pass filter, and the filtered data is compared with the origi-
nal so as to make sure that the univariate statistics show a
reasonably good match. Multiple inputs are generated using
multiple filters.
2. The filtered mud log data are used as input to design a feed
forward back-propagation neural network model. The model
is trained to predict desired output rock properties (Young’s
Modulus and Poisson’s ratio). The model design includes the
usual training, validation and testing phase. Care is taken to

prevent over-fitting of the data as well as having a reasonable
network in terms of size. We use a network with single hidden
layer and 10:1 ratio for number of hidden layer nodes when
compared with number of sample points. Models specific to
defined lithological layers are tagged and stored for application.
3. For application of the design, data from the candidate well is
identified, and it is segmented based on predefined litholog-
ical layers. Corresponding models are applied after careful
filtering of data using the filters obtained in step 1.
4. Based on the generated geomechanical models, a rock brit-
tleness parameter is computed. The resulting parameter is
combined within the predefined hydraulic fracture density
design framework shared later. It works on the basis of par-
tially weighting mud log gas shows and modeled properties
to get the final density values. The density model is normal-
ized and fractures/clusters are placed based on the behavior
of this modeled parameter.
5.2.3 Model Fine-Tuning

We need to highlight that there are multiple input properties being used in
the modeling process where each comes with varying degrees of associated
uncertainties. As an example, modeled rock properties have a high degree
of uncertainty due to modeling errors, particularly at large offset from
design wells. Similarly, mud log gas shows can sometimes show erroneous
readings due to gas flow into the wellbore downstream of the drilling bit.
Similarly, the framework relating modeled rock properties and gas shows
with naturally fractured zones in the reservoir is loosely defined due to a
lack of adequate corroborative data. Moreover, multiple input sets are gen-
erated from singular properties using variable filter parameters. All these
add up to create a highly non-unique solution space and therefore, finding
the right framework for combining these parameters to define a hydraulic
fracture density model can be a challenge.
In order to tide over these uncertainties, we use a fuzzy classification tech-
nique to identify the definition boundaries with adequate fuzziness so as to
classify sections of the lateral in terms of cluster spacing design by taking
into account the underlying uncertainty as well. At the same time, if pro-
duction logs are available and the broad framework is well defined (such
as highly brittle rock and high gas shows should lead to a lower modeled
hydraulic fracture spacing density, etc.), we can try to generate the best
possible model (and correspondingly, the best possible fuzzy classifier) to

match the designed fracture density with the observed production behavior
post completion. This is accomplished by using an evolutionary algorithm
to minimize a predefined error function which tries to match the inverse
of modeled fracture density with the observed cluster wise production. The
fuzzy rules set which defines the modeling framework we used is as follows:
• Rule #1: Low modeled brittleness and low gas shows imply
very high density.
• Rule #2: Medium modeled brittleness and low gas shows
imply very high density.
• Rule #3: High modeled brittleness and low gas shows imply
high density.
• Rule #4: Low modeled brittleness and medium gas shows
imply high density.
• Rule #5: Medium modeled brittleness and medium gas
shows imply medium density.
• Rule #6: High modeled brittleness and medium gas shows
imply medium density.
• Rule #7: Low modeled brittleness and high gas shows imply
low density.
• Rule #8: Medium modeled brittleness and high gas shows
imply low density.
• Rule #9: High modeled brittleness and high gas shows imply
very low density.
Here the ‘density’ values indicate final fracture density (or perforation
cluster density) recommendation to be made by the designed fracture den-
sity model. We do note that these rules suggest relatively lower fracture
cluster count for the so called sweet spots. Since the decision on how much
to frac and where is a highly complex one with well economics playing a
major role, the workflow is adaptable enough so that the rules can be flipped
with the high density recommendations changed to low density recommen-
dations and vice versa. This approach is useful in cases where specific well
intervals have a predictable behavior and sensitivity to stimulation.
5.2.4 Need for Artificial Intelligence

There are three computational elements using AI techniques used in this
work-flow. We need to consider the need for using said methods in this
study. We understand that though broad relationships between Gamma
Ray measurements and rock properties are expected due to influence of

clay content on said properties, the relation may not always hold due to
other influences. The same holds true for hydrocarbon indicator used for
modeling (mud log gas shows). Due to this non-linearity and in-exactness
in the relationship between geomechanical properties and gamma ray,
Artificial Neural Nets are ideally suited since they can map highly non-
linear relations if properly modeled and calibrated and are very robust in
handling noisy data [21].
Furthermore, a broad correlation between modeled and observed prop-
erties and the desired application (hydraulic fracture spacing design) can
be easily defined [e.g. higher gas shows and higher modeled brittleness
leads to lower cluster density, etc.]. However with the high uncertainty in
the available inputs, a simplistic framework for combining said properties
may not capture the relationships accurately. A classic solution to such a
classification problem is to use a Fuzzy Inference System. They are easy
to understand as they are governed by fuzzy rules, which are semantic in
nature, even though the underlying evaluation is mathematical. They have
the ability to optimally search for the best classifier set definitions to match
the observed data. They are simple yet highly adaptable and can work
with the imprecise and incomplete data that we have [22].
Finally, as stated already, the designed fracture density and observed pro-
duction behavior mismatch is minimized by using an evolutionary search
routine. The big advantage with using such an approach is that it is highly
scalable and adaptable and can be used to solve for multi-dimensional,
non-differential, non-continuous and non-parametric problems. They are
intuitive and very easy to build and therefore provide an optimal search algo-
rithm for the problem at hand [23].
5.3 Data
We apply this approach to three wells from two separate well pads (sep-
arated by 10’s of miles from one another). The data from Well #1 for Pad
#1 is used as the design data as the well had open-hole logging carried out
for the vertical pilot as well as the horizontal section of the well. Two other
wells were used as application wells for validating the models as each had
a production log available for independent validation. These include Well
#2 associated with the same well pad as Well #1 and Well #3 associated
with Pad #2. Figure 5.3 shows the various lithological layers of relevance
across Well #1. We can clearly observe that the well lateral intersects two
layers (Zone C or the target zone and zone B which is the overburden lower
4900
Marcellus
Onondaga
5000 Top cherry valley
Top zone A
Top zone B
5100 Top zone C
Top zone D
Well #1
5200
TVD (ft)
5300
5400
5500
5600
4000 5000 6000 7000 8000 9000 10000 11000 12000
MD (ft.)
Figure 5.3 Well #1 cross-section showing well track in reference to the varying lithologic
formations in the Marcellus shale play. The lateral is restricted to Zones C & D.
Marcellus layer) of interest, and we use the available data to model the geo-
mechanical properties for these two layers.
Since we have extensive wireline logging done for this long lateral (Well
#1), we can use the data from the logs to estimate geomechanical and other
properties, which in turn form the basis for our model design framework
using artificial neural nets. Apart from the standard mud log gas shows avail-
able from drilling records, other wireline tools run for this well (both for the
horizontal lateral and the vertical sections) include standard measurements
such as density, porosity, and resistivity, as well as lithological tools to iden-
tify mineralogy and organic content. A thorough petrophysical analysis was
carried out for the entire logged wellbore, and the geomechanical proper-
ties were ascertained using the lithology data. Gamma from both the actual
wireline logging run and the mud log data was correlated to validate applica-
bility for other “application” wells which lack similar wireline logs.
Using the geomechanical properties and the mud log data available for
well #1, the entire dataset was pruned such that two separate datasets were
generated. The inputs were expanded using multiple filtering bandwidths
to extract features at different frequency spectrums, which might hold
physical meaning and therefore are valuable in the modeling process. The
inputs were in turn used to develop two separate models for two separate
reservoir (shale) units, namely Zone C and Zone B as discussed earlier and
observed from Figure 5.3. However, based on the well trajectory along the
lateral, the two sections have varied data density in terms of available sam-
pling points. From the mud logs, while Zone C has approximately 600 data
points, zone B only has approximately 70 data points, which makes results
from the model defined for this zone susceptible to more errors. Figure 5.4
shows sample rock property (Poisson’s Ratio) modeled for these two layers,
and we can observe relatively higher errors for the Lower Marcellus Layer
compared to the target layer due to said mismatch.
For network training, care is necessary to prevent either over-fitting
or non-representative dataset generation. Care is taken in the neural net-
work design with the ratio of number of network nodes to the number of
data samples kept at less than 0.10. Also, segments from each Zone are
combined making sure that they are representative of varying behavioral
aspects of the property being modeled (such as sudden rise or drop in
value). The final network models are chosen based on the minimized error
observed within the network validation process (Figure 5.5).
With the individual geomechanical models as well as the composite
brittleness model ready for use, the next step is to identify the best fuzzy set
definitions corresponding to the rules defined earlier. Figure 5.6 shows the
final identified rules set which corresponds with the best match between
the predicted fracture density models and the observed production log
behavior for Well #1.
Once the model is ready for application, all of the model design param-
eters are stored for later use with application test scenarios. These include
Original
Modeled
(a) 5760 5780 5800 5820 5840 5860 5880 5900 5920
(b) 5760 5780 5800 5820 5840 5860 5880 5900 5920
Original
Modeled
(c) 5800 6000 6200 6400 6600 6800 7000 7200
(d) 5800 6000 6200 6400 6600 6800 7000 7200

MD (ft.)
Figure 5.4 (a) Modeled Poisson’s Ratio compared with actual log derived values and
(b) corresponding error mismatch for Zone B and (c) Modeled Poisson’s Ratio compared
with actual log derived values and (d) corresponding error mismatch for Zone C.
1.0 Data
Fit
0.9 Y= T
0.8
Modeled property
0.7
0.6
0.5
0.4
R = 0.9074
0.3
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Target property
Figure 5.5 Validation results from a sample neural network training run showing
minimized error in modeled property (Poisson’s Ratio) compared to target property
(actual log derived data). Note: The target data was normalized before modeling run.
Low Medium High
(in)
(a) 0 50 100
in
Low Medium High
(in)
(b) 0 50 100
in
Very Low Medium High Very
low high
(in)
(c) 0 50 100
in
Figure 5.6 Optimal fuzzy set definitions based on training well data for (a) modeled
brittleness, (b) mud log gas shows, and (c) output fracture density guidance.
the number of layers in the network, number of nodes in the hidden, input
and output layers, trained weights associated with each node in the net-
work, activation function s associated with each node, biases within the
network, etc. These saved models are in turn applied to any dataset from
“application” wells to generate hydraulic fracture/cluster density maps. The
final fracture density maps need to be res-called in order to make sure
that sufficient “maximum” and “minimum” fracture spacing is maintained
before final fracture or perforation cluster placement. These maxima and
minima limits can be independently evaluated using other stage design
approaches discussed in the introduction to this study.
5.4 Results
Before the models can be applied to other wells, they need to be catego-
rized based on lithological layers associated with each portion of the well-
bore to be analyzed. This is done by using the same approach as used for
the training well (Well #1) as discussed earlier. Once the wellbore has been
characterized, these segmented subsections are evaluated for rock proper-
ties by using the corresponding rock property predictive artificial neural
net derived models. We will share the design results obtained using the
models for the two application wells (Well #2 and Well #3) and compare
the observations with production log data. This will allow validation of the
observed results using independent production log results, which is critical
as initial production is the key for rapid return on investment in shale gas
wells. We again note that while Well # 2 belongs to the same pad as the
training well (Well #1), Well #3 is located 10’s of miles from the first pad
and incorporates a different completion design.
Before we look into the application wells (Well #2 & Well #3), we apply
the derived models to data from the training well itself (Well #1) using the
segmented modeling approach. Based on the wellbore location in reference to
lithological units (Figure 5.3), separate models are applied to the dataset along
the wellbore. Figure 5.7 shows the results for this particular test case and as
expected, we get a good match between the suggested fracture density derived
from the proposed workflow and the production log data for Well #1.
We note that the original completion for Well #1 involved 18 stages
with 4 perforation clusters per stage and an inter-stage separation of ~280
feet. The modified spacing design as suggested by the workflow is shown
in Figure 5.7(b). As observed from these results, significant mismatch is
observed close to 6900 ft. (measured depth), which could be due to the
wellbore lying either very close to or at the interface of zone B & zone C
Production log
inverse of designed density
%age match: 82.1302 Poor match
Good match
6000 6500 7000 7500 8000 8500 9000 9500 10000 10500
(a) MD(ft.)
5400
TVD(ft.)
5450
5550
6000 6500 7000 7500 8000 8500 9000 9500 10000 10500
(b) MD(ft.)
Designed clusters
# stages: 18 # Clusters/Stage: 4 Confidence - low
Smoothening Window: 40 ° of randomness: 0.01 Confidence - high
Equalspaced separation: 70 feet Rescaling: 0.75
Confidence: 50 Cluster density - rescaled
Cluster density - original
(c) 6000 7000 7500 8000 8500 9000 9500 10000 10500
MD(ft.)
Figure 5.7 Figure showing (a) the comparison between modeled fracture density and
available production log, (b) TVD behavior along lateral with traditional equally spaced
perforation locations, and (c) the actual fracture spacing recommendation based on the
model results for Well #1 incorporating both zone B & zone C models using segmented
modeling approach. Red arrow indicates section with significant mismatch between
production and designed fracture density.
(Figure 5.3). This would make it difficult to interpret which model is the
right model and applicable for corresponding sections of the wellbore.
For Well #2, we ran three tests with the first test incorporating the model
associated with zone C for the entire lateral (Case 2A), the second test incor-
porating the model associated with zone B for the entire lateral (Case 2B),
and the third test incorporating segment wise modeling using both models
based on location of the lateral in relation to the lithological units (Case 2C).
Figure 5.8 shows the modeling and design results for Case 2A, Figure 5.9
shows the results for Case 2B, and Figure 5.10 shows the results for Case 2C.
We note that the original completion profile for Well #2 involved 14
stages with 4 perforation clusters per stage with an inter-stage separa-
tion of ~300 feet. We can clearly see sections along the wellbore where the
design recommendation suggests sparser clusters and other sections which
suggest denser cluster spacing. This correlates well with the production log
results with sections suggesting denser clusters showing lower productivity
and vice versa. This is desirable considering the defined modeling framework
discussed earlier. However, for Case 2A & Case 2B, the predicted fracture
spacing design does not match well with the observed stage-wise productivity
behavior at some locations (identified by red arrow). This can be attributed to
the model applicability issue in certain sections of the wellbore, depending on
Production log
%age match: 87.2792 inverse of designed density
Poor match
Good match
6000 6500 7000 7500 8000 8500 9000 9500 10000

(a) MD(ft.)
5400
TVD(ft.)
5450
5500
5550
6000 6500 7000 7500 8000 8500 9000 9500 10000
(b) MD(ft.)
Designed clusters
Equalspaced separation: 78 feet Confidence: 50 Cluster density - rescaled
Rescaling: 0.75 Cluster density - original
(c) 6000 6500 7000 7500 8000 8500 9000 9500 10000
MD(ft.)
perforation locations, and (c) the actual fracture spacing recommendation based on
the model results for Well #2 incorporating model from zone C. Green arrow indicates
section with significant mismatch between production and designed fracture density,
which corresponds with wellbore section falling outside zone C.
Production log
Good match
(a) 6000 6500 7000 7500 8000 8500 9000 9500 10000
MD(ft.)
5400
TVD(ft.)
5450
5500
5550
(b) 6000 6500 7000 7500 8000 8500 9000 9500 10000
MD(ft.)
Designed clusters
(c) 6000 6500 7000 7500 8000 8500 9000 9500 10000
MD(ft.)
the model results for Well #2 incorporating model from zone B. Green arrows indicate
sections with significant mismatch between production and designed fracture density,
some of which corresponds with wellbore section falling outside zone B.
Production log
Good match
(a) 6000 6500 7000 7500 8000 8500 9000 9500 10000
MD(ft.)
5400
5450
TVD(ft.)
5500
5550
6000 6500 7000 7500 8000 8500 9000 9500 10000
(b) MD(ft.)
Designed clusters
(c) 6000 6500 7000 7500 8000 8500 9000 9500 10000
MD(ft.)
modeling approach. We do not see any significant mismatch between the predicted
fracture density behavior and the productivity of completed perforation clusters.
whether the well track is within the zone defining the applied geomechanical
model or not.
For Case 2C, we observe a much better match along the entire completed
lateral, and this is due to segmented modeling approach where the correct
model (based on the location of the wellbore in reference to the lithological
units) is used (Figure 5.10). This validates the applicability of the proposed
approach for wells within proximity of the well used in training our models.
For Well #3, the completion design was significantly different with 27
stages and 4 perforation clusters per stage with an inter-stage separation of
200 feet. Once again we generate results incorporating the model associ-
ated with zone C for the entire lateral (Case 3A), the second test incorpo-
rating model associated with zone B for the entire lateral (Case 3B) and the
third test incorporating segment wise modeling using both models based
on location of lateral (Case 3C). Since Well #3 is at an offset of 10’s of miles
from Well #1, we expect the results to be not as robust as was the case with
Well #2. Figure 5.11 shows the modeling and design results for Case 3A,
Figure 5.12 shows the results for Case 3B, and Figure 5.13 shows the results
for Case 3C.
For Case 3A & Case 3B, the predicted fracture spacing design does not
match well with the observed stage-wise productivity behavior at sections
Production log
Good match
8500 9000 9500 10000 10500 11000 11500 12000 12500 13000 13500
(a)
MD(ft.)
8200
TVD(ft.)
8300
8400
8500
8500 9000 9500 10000 10500 11000 11500 12000 12500 13000 13500
(b) MD(ft.)
# stages: 27 # Clusters/Stage: 4 Designed clusters

Confidence - low
(c)
8500 9000 9500 10000 10500 11000 11500 12000 12500 13000 13500
MD(ft.)
the model results for Well #3 incorporating model from zone C. Green arrows indicate
some of which corresponds with wellbore section falling outside zone C.
Production log
%age match: 84.6568 inverse of designed density
Poor match
Good match
8500 9000 9500 10000 10500 11000 11500 12000 12500 13000 13500
(a)
MD(ft.)
8200
TVD(ft.)
8300
8400
8500 8500 9000 9500 10000 10500 11000 11500 12000 12500 13000 13500
(b) MD(ft.)
# Clusters/Stage: 4 Designed clusters

# stages: 27 Confidence - low
Confidence: 50
Cluster density - rescaled
(c) 8500 9000 9500 10000 10500 11000 11500 12000 12500 13000 13500
MD(ft.)
model results for Well #3 incorporating the model from zone B. Green arrows indicate
some of which corresponds with wellbore section falling outside zone B.
Production log
Good match
(a) 8500 9000 9500 10000 10500 11000 11500 12000 12500 13000 13500
MD(ft.)
8200
TVD(ft.)
8300
8400
8500
8500 9000 9500 10000 10500 11000 11500 12000 12500 13000 13500
(b) MD(ft.)
# stages: 27 # Clusters/Stage: 4 Designed clusters
° of randomness: 0.01 Confidence- low
Smoothening Window: 50 Confidence- high
Equalspaced separation: 51 feet Confidence: 50
Rescaling: 0.75 Cluster density - rescaled
(c) 8500 9000 9500 10000 10500 11000 11500 12000 12500 13000 13500
modeling approach. Green arrows indicate sections with significant mismatch between
production and designed fracture density.
highlighted using red arrows. This can be attributed to the model applica-
bility issue in certain sections of the wellbore, depending on whether the
track is within the zone defining the applied geomechanical model or not,
as observed with the earlier test case. Other issues include the robustness
of the model at separation of 10’s of miles as well as issues with inadequate
data for zone B model as highlighted earlier.
For Case 3C (Figure 5.13) using segmented modeling approach, we
observe a much better match along the completed lateral. However once
again, there are small sections of the lateral where the fracture placement
recommendation based on density model does not match well with the
production log observations.
Next we highlight the issue of model robustness due to data inadequacy.
As observed from Figure 5.12, a certain section of the modeled fracture
density along the wellbore shows significantly poor results suggested by the
consistent high values from ~9300 feet to ~11700 feet (measured depth).
This is because the model from zone B is poorly defined due to lack of
adequate data as discussed earlier. While the laterals for both Well #1 and
Well #2 fall mostly within the target zone C, significant sections of the well-
bore corresponding to the identified depth interval for Well #3 fall within
overburden zone B (as observed in Figure 5.13 from ~10300 feet measured
depth to ~11000 feet measured depth). Therefore, these erroneous artifacts
are observed for both Case 3B and 3C, which makes use of geomechanical
models from zone B.
5.4.1 Applicability Considerations

Based on our results, we can say with some degree of confidence that this
approach can be useful in designing completions (stage or cluster spacing)
of wells within the same well pad provided major sections of the wellbore
do not fall very close to or at the interface between geologically distinct
layers with significant variability in geomechanical properties. Moreover,
presence of local faulting or completion of nearby wells post drilling oper-
ations of the candidate well can also have a significant impact on results.
Since the proposed method works with multiple models, based on a seg-
mented modeling approach, it is critical that each model is robust as well
as well-defined, and its reliability should be ascertained before application.
For wells at significant offset from those wells used for model design,
applicability can suffer depending on lithologic variability across spa-
tial distance, as well as other formation properties. However, our results
show that properly designed models and segmented modeling approach
can still provide reasonably good fracture density maps and spacing
recommendations.
(a)
(b)
(c)
Figure 5.14 (a) Well schematic showing localized sweet spot due to intersecting natural
fracture swarm (green box) and (b) hydraulic fracture density framework suggested in
this study with (c) potential alternate framework along the wellbore lateral.
Beyond the questions regarding effectiveness of the models, the meth-

odology used, and applicability close to and away from those wells used
for model training and design, a more fundamental question is the efficacy
of the design framework proposed in this study. Figure 5.14 highlights the
broad framework in question as well as one possible alternative framework
to highlight this issue.
Since the question of how to proceed with the completion design frame-
work is a complex one with well economics playing an integral part in any
decision making process, a more thorough investigation and decision mak-
ing based on particulars of the wells being completed using this approach
is essential. As an example, the decision on which framework to choose
could be decided by price factors (gas vs. oil/condensate rich play) as well
as reservoir related considerations (Clay richness, natural fractures, etc. to
name just two).
5.5 Concluding Remarks

We have introduced a fracture spacing design approach which makes use
of routinely collected mud log data apart from some reference wireline
specialty logs to model for complex geomechanical properties of the rock
surrounding the well-bore. These models along with observed gas shows
are used to propose variable perforation cluster spacing along the well-
bore laterals. We have demonstrated this approach to be useful at small
(100’s of feet) as well as large (10’s of miles) geographical offsets from the
wells used to train said models. The proposed methodology identifies local
sweet spots that require less stimulation and areas where more stimula-
tion is needed. By redistributing the hydraulic fracture density along the
wellbore, we aim to balance stimulation costs and long-term production
performance.
While many methods have been proposed over the years which utilize
data from such specialty wireline logs to predict well behavior or recom-
mend completion design, the key discriminator with our proposed frac-
ture spacing design methodology is the ability to apply the technique at
geographically far off wells without having to update the geomechanical
models. The key is to use a well-defined modeling framework and produc-
tion log or other completion quality attributes (such as from fiber-optic
data) to constrain the designed models so that they can mimic well behav-
ior upon completion with reasonable accuracy.
Based on the results we have observed from multiple wells, includ-
ing those shared in this study, we hypothesize that this approach should
work in most situations provided proper care is taken before applying this
approach. However, the proposed approach needs to be validated as it may
not hold under many situations depending on economic considerations.
For future work, we propose to carry out extensive modeling studies and
generate guidelines for applicability under varying scenarios as suggested
in this work.
In the future, we plan on using a fuzzy or probabilistic classifier to decide
which model to be used depending on the closeness of the well track to a
particular lithologic boundary. This is due to significant uncertainty rang-
ing from 10’s to 100’s of feet when it comes to layer boundaries and exact
well location, which can make the decision making on models to be used
for design very non-representative. We hope to test this approach on mul-
tiple wells in other shale plays (Permian Basin). We also expect to conduct
similar design work for multiple wells which are geographically spread out
and validate these observations.
Acknowledgement
This work was supported by Research Partnership to Secure Energy for
America (RPSEA) project number 11122-20. We also acknowledge WPX
Energy for providing access to active hydraulically fractured wells of
opportunity in Marcellus shale play and Schlumberger for acquiring the
wireline and production logs used in this study.
References
1. J.W. Crafton and D. Anderson, Use of extremely high time-resolution pro-
duction data to characterize hydraulic fracture properties, Presented at
the SPE Annual Technical Conference and Exhibition, San Antonio, Texas,
24–27 September, SPE-103591-MS, http://dx.doi.org/10.2118/103591-MS
(2006).
2. P. Ye, L. Chu, I. Harmawan, et al., Beyond linear flow analysis in an
unconventional reservoir. Presented at the SPE Unconventional Resources
Conference, The Woodlands, Texas, 10–12 April, SPE-164543-MS, http://dx.
doi.org/10.2118/164543-MS (2013).
3. K. Wu and J.E. Olson, Investigation of the impact of fracture spacing and
fluid properties for interfering simultaneously or sequentially generated
hydraulic fractures. SPE Production & Operations 28(4), 427–436 (2013).
4. M. Jonathan and J.L. Miskimins, Optimization of hydraulic fracture spacing in
unconventional shales. Presented at the SPE Hydraulic Fracturing Technology
Conference, The Woodlands, Texas, 6–8 February, SPE-152595-MS, http://

dx.doi.org/10.2118/152595-MS (2012).
5. C. Cipolla and J. Wallace, Stimulated reservoir volume: A misapplied con-
cept? Hydraulic Fracturing Journal 1(2), 54–75 (2014).
6. X. Ma, T. Plaksina, and E. Gildin, Optimization of placement of hydraulic
fracture stages in horizontal wells drilled in shale gas reservoirs, Presented
at the Unconventional Resources Technology Conference, Denver, Colorado,
12–14 August, SPE-168790-MS, http://dx.doi.org/10.1190/URTEC2013-151
(2013).
7. A. Baig, T.I. Urbancic, K.C. Mace, et al. Assessing the spacing of stages in plug
and Perf completions through seismic moment tensor inversion, Presented
at the SPE Hydraulic Fracturing Technology Conference, The Woodlands,
Texas, 6–8 February, SPE-152547-MS, http://dx.doi.org/10.2118/152547-MS
(2012).
8. L.W. Bazan, S.D. Larkin, M.G. Lattibeaudiere, et al., Improving production
in the Eagle Ford Shale with fracture modeling, increased fracture conduc-
tivity, and optimized stage and cluster spacing along the horizontal wellbore,
Presented at the Tight Gas Completions Conference, San Antonio, Texas, 2–3
November, SPE-138425-MS, http://dx.doi.org/10.2118/138425-MS (2010).
9. R.L. Kennedy, R. Gupta, S.V. Kotov, et al., Optimized shale resource devel-
opment: Proper placement of wells and hydraulic fracture stages, Presented
at the Abu Dhabi International Petroleum Conference and Exhibition,
Abu Dhabi, UAE, 11–14 November, SPE-162534-MS, http://dx.doi.
org/10.2118/162534-MS (2012).
10. J. Ciezobka, Marcellus shale gas project, RPSEA, August 2013, http://www.
rpsea.org/files/3629/ (accessed 11 May 2015) (2013).
11. B. Ajayi, I.I. Aso, I.J. Terry, et al. Simulation design for unconventional
resources. Oilfield Review 25(2), 34–46 (2013).
12. N.R. Warpinski, M.J. Mayerhofer, E.J. Davis, et al., Integrating fracture diag-
nostics for improved microseismic interpretation and stimulation mod-
eling, Presented at the Unconventional Resources Technology Conference,
Denver, Colorado, 25–27 August, SPE-2014-1917906-MS, http://dx.doi.
org/10.15530/urtec-2014-1917906 (2014).
13. R. Borstmayer, N. Stegent, A. Wagner, et al., Approach optimizes completion
design. The American Oil & Gas Reporter 54 (8), 79–88 (2011).
14. A. Jahandideh and B. Jafarpour, Optimization of hydraulic fracturing design
under spatially variable shale fracability, Presented at the SPE Western North
American and Rocky Mountain Joint Meeting, Denver, Colorado, 17–18 April,
SPE-169521-MS, http://dx.doi.org/10.2118/169521-MS (2014).
15. T. Engelder, G.G. Lash, and R.S. Uzcategui, Joint sets that enhance produc-
tion from Middle and Upper Devonian gas shales of the Appalachian Basin.
AAPG Bulletin 93(7), 857–889 (2009).
16. B. Bahorich, J.E. Olson, and J. Holder, Examining the effect of cemented
natural fractures on hydraulic fracture propagation in hydrostone block
experiments. Presented at the SPE Annual Technical Conference and

Exhibition, San Antonio, Texas, 8–10 October, SPE-160197-MS, http://dx.
doi.org/10.2118/160197-MS (2012).
17. B. Gutenberg and C.F. Richter, Seismicity of the Earth and Associated
Phenomena, second edition, Princeton University Press, Princeton, NJ (1954).
18. N. Boroumand, Hydraulic fracture b-value from microseismic events in dif-
ferent regions. Presented at GeoConvention, Calgary, Canada, 12–16 May
(2014).
19. I.A. Salehi and J. Ciezobka, Controlled hydraulic fracturing of naturally frac-
tured shales: A case study in the marcellus shale examining how to identify
and exploit natural fractures. Presented at the SPE Unconventional Resources
Conference, The Woodlands, Texas, 10–12 April, SPE-164524-MS, http://dx.
doi.org/10.2118/164524-MS (2013).
20. R. Rickman, M.J. Mullen, J.E. Petre, et al., A practical use of shale petro-
physics for stimulation design optimization: All shale plays are not clones
of the Barnett shale, Presented at the SPE Annual Technical Conference and
Exhibition, Denver, Colorado, 21–24 September, SPE-115258-MS, http://dx.
doi.org/10.2118/115258-MS (2008).
21. S. Haykin, Neural Networks: A Comprehensive Foundation, second edition,
Prentice Hall, Saddle River, NJ (1998).
22. T.J. Ross, Fuzzy Logic with Engineering Applications, third edition, John
Wiley & Sons, West Sussex, UK (2009).
23. D.E. Goldberg, Genetic Algorithms in Search, Optimization and Machine
Learning, first edition, Addison-Wesley Longman, Boston, MA (1989).
6
Clustering-Based Optimal Perforation
Design Using Well Logs
Andrei S. Popa1*, Steve Cassidy1 and Sinisha Jikich2
1
Chevron North America Exploration and Production, Bakersfield, California, USA
2
University of Pittsburgh, Pittsburgh, Pennsylvania, USA
Abstract
In an effort to better understand the well performance in one of the Chevron’s
assets in San Joaquin Valley, a study was conducted to evaluate the perforation
strategies and capture best practices. Well completion through perforation is typ-
ically performed using bare essential technology such as wireline logs and per-
foration guns. For basic reservoir formations, simple rules-of-thumb are used
for perforation spacing and interval lengths. These are rarely validated by other
methods, such as production logging and micro-seismic monitoring. For more
challenging lithology, a more appropriate approach would be to place perforation
clusters in target formations with similar properties. The research presents an effi-
cient use of fuzzy clustering technology for identification of the optimum perfo-
ration strategy in a challenging waterflood diatomite reservoir. The methodology
was applied on all newly drilled wells in the reservoir (within the last two years),
and we found that this new approach improved our understanding over previ-
ous practices, not only by designing optimum perforations, but also an increased
production was observed. Cluster analysis is the task of grouping a set of objects
in such a way that objects in the same group (called a cluster) are more similar to
each other (in some sense or another) than to those in other clusters. There are two
commonly used types of clustering methods: hard and fuzzy clustering. In hard
clustering, data is divided into distinct clusters, where each data element belongs
to exactly one cluster. In fuzzy clustering, data elements can belong to more than
one cluster, and associated with each element is a set of membership levels. The
fuzzy clustering algorithm, also known as Fuzzy C-Mean (FCM) algorithm, was
applied to log data of wells from different areas of a reservoir. Based on the cluster-
ing results, the workflow then identified whether the perforation was performed
*Corresponding author: andreipopa@chevron.com

125
on “good” regions (sand) or on “bad” regions (shale bedding). This information

allowed the evaluation of the perforation jobs executed and also allowed capturing
best practices and design changes for future well completions. The case study pres-
ents a simple, yet efficient workflow to extract additional information from logs
and improve completion strategies and perforation design. The methodology is
flexible and can be applied to any well where complex lithology creates a challenge
in defining the optimum perforation intervals.
Keywords: Fuzzy C-Mean, well perforation design, log analysis, hydraulic

fracturing
6.1 Introduction
The Lost Hills oil field is one of several significant oil fields located in the
San Joaquin basin of central California [1]. The field is located about 45
miles northwest of Bakersfield in Kern County, Figure 6.1. The producing
structure is a northwest-southeast trending double-plunging anticline [2].
The anticline is about 12 miles long and about 1 mile wide with high dip-
ping flanks reaching up to approximatively ~20 degrees.
The main reservoir is relatively shallow, and the producing interval var-
ies from 600 feet to a maximum of 1200 feet along the crest of the anti-
cline. The most productive unit is the Belridge diatomite of the Miocene
Figure 6.1 Lost Hills field location.

Clustering-Based Optimal Perforation Design 127
Monterey Formation. The diatomite is a biogenic siliceous deposit formed

by the shells of diatoms with varying amounts of detrital material, prin-
cipally clay and sand [3]. These deposits form a series of interbedded
sequences within the entire productive interval, whereby thinner zones are
sandier diatomite with better capacity for flow, while others are more com-
pact with very poor permeability.
Since its discovery in 1910, the reservoir was developed and produced in
primary recovery. Hydraulic fracturing was initiated in the mid-1980’s and
the technology proved to be a great avenue for increasing recovery given
the natural low permeability of the reservoir. Waterflooding was introduced
in the early 1990’s, initially localized and as a mitigation process to over-
come the compaction of the highly compressive diatomite rocks. The water
injection reduces compaction due to reduction of net effective stress on the
reservoir rock. Compaction appears to be driving the crossflow from lower
permeability to higher permeability layers and can be significant when adja-
cent layers have strongly contrasting permeabilities. The waterflood process
proved to be very successful and contributed to additional recovery. As a
result, the water injection was extended to the entire field. The initial devel-
opment consisted of 2 ½ acre staggered patterns; however, with time, these
patterns were infilled to 1 ¼ acre. After the 2000’s a relatively small part of
the field was further infilled to 5/8 acre patterns.
6.2 Objective and Motivation

The rapid and aggressive field development started in the late 1980’s when
advances in hydraulic fracturing technology led to significant production
increase and improved economics. Early on, technology employed multi-
stage (plug and perf) hydraulic fracturing, followed by multi-stage coiled
tubing fracturing technique, and after the mid 2000’s, use of Cobra jet/max
fracs. Over the years the number of stages varied, back and forth, from
a minimum of 5–6 to as many as 19 stages per well. The underlying rea-
son was the full coverage of the hundred plus feet formation. However, the
most interesting observation was the significant variance of the production
performance of the wells after stimulation. The range of expected outcome
was observed to vary anywhere from as low as 10–20 bopd to as high as
400–500 bopd at peak. Thus, multiple efforts were entertained over the
years to try to establish a correlation between fracturing treatment attri-
butes (including well design and completion) and well productivity. One
of the most comprehensive hydraulic stimulation studies was performed
in 2008 when more than 440 wells, completed between 1998 and 2007,
were analyzed to drive an optimum fracturing-treatment design to maxi-

mize production. The effort consisted of an overwhelming data collection
exercise including cultural data, wellbore schematics, perforations, treat-
ment by stage, additives by stage, pumping schedule, geological info such
as formation markers (tops and bottoms), well production, and pattern
production. While the study revealed some observations, it did not provide
a silver bullet for frac completion optimization.
The development of the field continues today with an active drilling pro-
gram consisting of up to 50 wells per year. With every single producing
well undertaking a hydraulic fracturing treatment prior to starting pro-
duction, the opportunity to optimize the well completion and fracturing
treatment is still of high interest.
The main objective of the study is to investigate the impact of perforation
placement and design on fracture execution and ultimately on the perfor-
mance response of newly drilled wells in different development areas across
the subject field. Additional motivation for the authors was to investigate
if using only basic log data and clustering technologies reveals any hidden
patterns, learnings, best practices and opportunities for future well comple-
tion optimization. It is a helpful benchmark exercise to analyze and contrast
different perforations strategies driven by different lithology across the field.
6.3 Technology
The well perforation strategy for this field was strongly linked to the number
of completion frac stages planned for the well, and vice-versa. Therefore,
a large number of frac stages, which would provide better frac coverage
across the productive formation, would require an increased number of
perforation sets. However, it was shown that one likely non-uniformity
in well production is that fluid that is injected with intention to gener-
ate fractures from all perforation clusters is diverted to only a portion of
the clusters leaving many clusters unstimulated [4]. Industry experience
shows that there is an opportunity for improvement of recovery efficiency
by stimulating effectively all perforation clusters along the well. Several
approaches to improve stimulation uniformity have been proposed. In
the “engineered completion” approach [4], well log data was used so that
perforation clusters can be placed in regions of similar stress or brittle-
ness index. Cipolla details new algorithms and an integrated workflow that
could improve fracture treatment staging in both vertical and horizontal
wells [5]. In a newer approach, Pierce and Bunger, promoted uniform stim-
ulation of shale gas reservoirs by strategically locating entry points so that
stress interaction between the hydraulic fractures promotes rather than

suppresses their stimulation growth [6]. This approach was inspired by
computational studies showing that the “stress shadowing”, or more gen-
erally stress interaction can lead to either suppression or enhancement of
multiple fracturing growth.
Thus, there is a need for geologists and completion engineers to work to
establish the best perforation placement along the wellbore, such as what
fracs would initiate and propagate, to offer the best fracturing cover.
In the case of the studied field the main observation indicated that well
performance across the entire field shows significant variability, yet this is not
clearly explained by any past stimulation and/or reservoir studies. However, it
appears that production outcome follows a characteristic distribution which
can be divided into three categories of wells: poor, average, and good. This
classification inspired the application of clustering technologies in an attempt
to research the underlying nonlinear correlations between formation charac-
teristics, perforation placement, and production response.
6.4 Clustering Analysis

Cluster analysis was originated in anthropology by Driver and Kroeber in
1932 and introduced to psychology by Zubin in 1938 and Robert in 1939.
The technology was famously used by Cattell in 1943 for trait theory clas-
sification in personality psychology.
Cluster analysis or clustering is a generic name for a set of statistical or
machine learning algorithms that aim to detect distinct groups in a sample
of objects, these groups usually being called clusters. In contrast to dis-
criminant analysis, the cluster analysis does not require a group structure
known a priori. This specific characteristic makes cluster analysis attractive
as an exploratory tool [7].
Clustering can be achieved by various algorithms that differ signifi-
cantly in their notation of what constitutes a cluster and how to efficiently
define them. Popular notions of clusters include: groups with small dis-
tances among the cluster members, dense areas of the data space, and
intervals or particular statistical distributions. Clustering can, therefore,
be formulated as a multi-objective optimization problem. The appropriate
clustering algorithm and settings (for example, distance function, density
threshold, or the number of expected clusters) depend on the individual
data set and intended use of the results.
Typical cluster models include [8]: connectivity models such as
Hierarchical Clustering [9], centroid models such as K-mean Algorithm
[10], distribution models such as Expectation-Maximization (EM) algo-

rithms [11] where multivariate normal distributions are assumed, den-
sity models such as Density-based spatial clustering of applications with
noise (DBSCAN) [12], subspace models such as two-mode-clustering, and
graph-based models [13].
A “clustering” is essentially a set of such clusters, usually containing all
objects in the data set. Additionally, it may also specify the relationship of
the clusters to each other. Clustering can be roughly distinguished as hard
clustering, where each object belongs to a cluster or not, and, soft cluster-
ing where each object belongs to each cluster to a certain degree.
In the next section, the Fuzzy C-Mean (FCM) algorithm [Bezdek, 1981;
Nock, 2006] is reviewed.
6.4.1 C-Means (FCM) Algorithm

In contrast to k-mean clustering, where each data point belongs to exactly
one cluster, fuzzy clustering associates a membership level to each of the
clusters for every data point. These membership levels indicate the strength
of the association between that data element and a particular cluster. Fuzzy
clustering is a process of assigning these membership levels, and then using
them to assign data elements to one or more clusters.
One of the most widely used fuzzy clustering algorithms is the Fuzzy
C-Mean (FCM) algorithm. Similar to the k-mean algorithm described in
the last section, the FCM algorithm attempts to partition a finite collection
of n elements (x1, x2, ..., xn) into a collection of k fuzzy clusters with respect
to some given criteria. Given a finite set of data, the FCM algorithm not
only returns a list of k cluster means = {μ1, μ2, ..., μk}, but also a partition
matrix W = wi,j [0,1], i = 1, ..., n, j = 1, ... k, where each element in matrix
W represents the degree to which data point xi belongs to cluster j.
The minimization objective function in the FCM algorithm is as follows:
n k 2
arg min s wi , j x1 j
i 1 j 1
(6.1)
where:
1
wi , j 2
m 1 (6.2)
k x1 j
l 1 x1 j
Note that (6.1) is different from k-means algorithm’s objective function

by the addition of the membership values wi,j and the fuzzifier m , with
m ≥ 1. The fuzzifier m determines the level of cluster fuzziness. A large m
results in smaller membership values wi,j, and hence, fuzzier clusters. In the
extreme case that m = 1, the memberships wi,j converge to 0 or 1, which is
the same as a crisp partitioning. In the current study, a value for m = 2 was
assigned, as commonly used in the literature.
6.5 Methodology and Analysis

As stated in the objective section, the goal of this study was to analyze the
impact of perforations placement and design on the well production out-
come. Therefore, the study focuses on researching the nonlinear “mapping”
between the available logs, existing perforations along the 600+ feet of pro-
ductive formation, and production outcome of the wells. The following sec-
tion will describe the data considered in this study and the methodology.
6.5.1 Available Data

The three main sources of data consisted of reservoir characterization
(wireline logs), well completion (perforations depth and footage), and well
production performance (peak rate, first 3 months, etc.). A total number of
132 wells were drilled in the last two years and had available data for anal-
ysis. A map of the field is presented in Figure 6.3. The field was divided in
three areas, as shown in Figure 6.3. Ten wells were randomly selected from
each area for analysis. The motivation behind the field division and well
selection was to capture any potential trends associated with field stratigra-
phy and depositional environment. Additionally, the wells were located in
areas with different waterflooding maturity. The map shows the active
existing wells with cyan. In contrast, the newly drilled wells within the last
two years are shown in black.
In the reservoir characterization domain, six types of log data were col-
lected for all the wells. These logs are listed below, together with a brief
explanation.
1. Deep Resistivity Log (DRES). This represents the deepest

reading resistivity log. It is a measure of the ability of the for-
mation to carry an electric current. The deepest resistivity is
often believed to read past any near wellbore contamination
from drilling mud.
Figure 6.2 Reservoir structure.
2. Gamma Ray (GR). This is a measure of the natural radio-

activity of the formation. The tools physically count particles
emitted by the formation. These counts are then calibrated
to an API scale.
3. Neutron Porosity (NPHI). This is a measure of the capacity
of the formation to attenuate neutron radiation. It is most
sensitive to the presence of hydrogen, either in the form of
water or hydrocarbons. In “clean” formations it is easily cor-
related to the porosity of the formation.
4. Permeability Log (PERM). This is a transform measurement
that uses other logs’ information and attempts to estimate
the formation permeability. For accuracy and validation,
it needs to be calibrated to laboratory measurements of
permeability taken from whole core plugs.
5. Porosity Log (PHIT). This represents the total porosity. This
track is calculated from the other logs and is an estimate of
the true pore space capable of holding fluids.
Area 1
Area 2
Area 3
Figure 6.3 Field map showing the division of the three zones.
6. Shale Volume (VSH). The volume of shale is the fraction of

the rock volume that is shale. It is calculated from the other
logs. It is a critical input into the calculation of total porosity
and water (or oil) saturation.
In addition to the reservoir characterization data, completion informa-

tion such as the depth and total perforated footages was collected for each
of the wells. The perforations data was merged with the log data and stored
in the same repository table. A snapshot of the sample data containing the
data for the study is shown in Figure 6.4. For every half foot, there is a
binary indicator (column COMPL_OPEN) to represent whether perfora-
tions are present. The total perforation footage for each well is calculated
using this track data.
In terms of production performance, several production indicators were
considered. The following indicators were collected for all the wells: cumu-
lative lifetime production, peak daily oil production, first month aver-
age production, first 3-month average daily production rate, peak (max)
3-month daily production, etc. In this study, we present the results for the
peak 3-month daily production as the output production indicator.
Figure 6.4 A snapshot of the sample data showing collected log tracks and perforation
completion.
6.6 Applying the FCM Algorithm

For the purpose of this paper, we present the results of the ten wells sam-
pled from each of the three areas of the field (see Figure 6.1). The FCM
algorithm was applied to cluster the log data for each well. Initially, all six
log types were used in the exercise; however, it was found that the cal-
culated tracks (PERM, PHIT, VSH) did not bring additional value to the
assessment. The final cluster analysis was performed using only the three
measured log tracks. While originally the data was grouped into three clus-
ters, to achieve higher definition in the partition zones we used five fuzzy
clusters in the study. Thus the linguistic representation of the clusters,
which describe the quality of the reservoir, can be defined as: bad, poor,
medium, fair, and good.
With the clustering performed for each well, the process identified
whether the perforations were shot in the “good” (sand) regions or on the
“bad” (shale) regions.
The following step process was employed to cluster the regions within
the reservoir formation for each sampled well:
1. For every half foot, the DRES, GR, and NPHI log data was
input into the FCM algorithm. Also, the measurements that
were less than 1000 feet were disregarded as not being part
of the reservoir.
2. Run the FCM algorithm on xi = [DRESi, GRi, NPHIi], where
i represents each half-foot measurement, on all the available
log data for the well. The number of clusters, k, was set to be
5 in all the experiments.
3. The mean of each of the five clusters was returned from
the FCM algorithm. The clusters were ranked by the DRES
log, where the one with the least DRES log was labeled as
the “bad” (shale) region and the one with the largest DRES
was labeled as the “good” (sand) region. An example of the
cluster means returned by the FCM algorithm is shown in
Figure 6.5.
4. The results of the clustering allow for associating each well
with a cluster by depth. The clustering results for two sam-
pled wells are depicted in Figure 6.5.
5. The final step was to calculate the perforation footage in the
cluster label “good.”
The procedure presented is explained through the two plots presented

in Figure 6.5 above. The continuous lines represent the DRES, GR, and
NPHI log data every half foot, while the blue “dots” at the right of the
graphs represent the footage perforation existing in the wells. Lastly, the
–100 –50 0 50 100 150 200 –100 –50 0 50 100 150 200
1000 1000
1200 1200
1400 1400
Depth, ft
Depth, ft
1600 1600
1800 1800
2000 2000
2200 2200
2400 2400
DRES C1 DRES C1
GR C2 GR C2
NPHI C3 NPHI C3
Perforated C4 Perforated C4
C5 C5
Figure 6.5 Clustering results by depth for two sample wells. In this figure, C1 represents
the shale region, and C5 represents the sand region. The actual perforation regions were
also labeled by the dark purple regions.
column to the far right represents the “cluster” track where every half foot
is assigned to one of the five clusters.
This study was carried out using only fuzzy clustering. Other method-
ologies, such as hard clustering of ants colony optimization, can be applied
to cluster the data.
The procedure presented is explained through the two plots presented
in Figure 6.5 above. The continuous lines represent the DRES, GR, and
NPHI log data every half foot, while the blue “dots” at the right of the
graphs represent the footage perforation existing in the wells. Lastly, the
column to the far right represents the “cluster” track where every half foot
is assigned to one of the five clusters.
It should be noted that current field practices were rather inconsis-
tent and selection of the perforations was accomplished through a visual
inspection of each log separately. Certain logs may show particular signa-
tures in the “good regions”, however, this was not very intuitive. The clus-
tering allowed not only for a better definition of the regions but also for
corroboration with the perforation intervals and production response.
6.7 Results and Discussion

Two plots were generated to investigate the correlation between perfora-
tion placement and production performance. Figure 6.6 shows the plot of
3-month peak production over the total perforation footages for all the
sampled wells, while Figure 6.7 depicts the plot of 3-month peak produc-
tion over the total perforation footages on the “good” regions (identified by
the FCM algorithm) for all the sampled wells.
In Figure 6.6, with two exceptions, most of the wells have about 40 feet
total perforations, however their 3-month peak production ranged from
about 200 BPD up to more than 1,600 BPD. No visible correlation can be
observed from the plot. In contrast, in Figure 6.7, a clearly visible correla-
tion can be observed between the 3-month peak production and the foot-
age perforated in “good” regions. The correlation stays valid for all three
areas of the field.
Similar analysis was carried out and plots generated for the other produc-
tion indicators: peak daily oil production, first month average production,
peak (max) 3-month daily production. Generally the plots showed the same
trends with the 3-month peak; however the interpretations were impaired
of both data availability (missing) and variability. The field does not benefit
from any simulation models due to its complex geologic earth model and
comingled production. Thus, a history match could not be performed.
1800
Peak 3-month production (bopd)

1600
1400
1200
1000
Zone 1
800
Zone 2
600
Zone 3
400
200
0
0 50 100 150
Total footage perforated per well (ft)
Figure 6.6 3-month peak production vs. total perforation footages for all sampled wells.
1800
Peak 3-month production (bopd)
1600
1400
1200
1000
Zone 1
800
Zone 2
600
Zone 3
400
200
0
0 50 50 100 150
Footage perforated in "good" zones (C5) (ft)
Figure 6.7 3-month peak production vs. total perforation footages on good regions
identified by the FCM algorithm.
To better understand the data driven findings, a complex 3D fracture

simulation model (Fracpro) was run for the same well in two scenarios.
The stress profile used was derived from the available sonic logs run in a
test well in the field. Consequently, both reservoir and treatment design
were kept the same and the only difference between the runs was the place-
ment of the perforation cluster. The two model runs are depicted in Figure
6.8 and Figure 6.9 respectively. To the left of each display there is the stress
profile, a shaded area characterizing the “good” cluster and the perforation
placement. To the left one can see the fracture length and conductivity at
the end of pumping treatment.
Strat well
31.33 min
0.000
0.550
1.100
Proppant coverage lb/ft2

1.650
Cluster 5
2.200
2.750
3.300
3.850
4.400
4.950
5.500
6.285 m/sec
263 1244 2226 100 200 300

TXSP psi Fracture Penetration (ft)
Figure 6.8 Fracture modeling. Proppant coverage at the end of pumping, perforations at
2680 ft.
Strat well
368.63 min
0.000
0.450
0.900
Proppant coverage lb/ft2

1.350
Cluster 5
1.800
2.250
2.700
3.150
3.600
4.050
4.500
6.285 m/sec
263 1244 2226 100 200 300

TXSP psi Fracture penetration (ft)
Figure 6.9 Fracture modeling. Proppant coverage at the end of pumping, perforations at
2750 ft.
The fracture simulator provided an effective representation of the frac-

ture conductivity in the case of the two scenarios. In the first scenario,
Figure 6.8, the fracture modeling with the perforations at 2680 ft. show
high fracture conductivity (cyan and red color on the right chart) over
the “good” region at the end of treatment. In contrast, in the second sce-
nario, when perforations are placed at 2750 ft. below the “good” region,
the simulation shows poor or low fracture conductivity over the “good”
formation at the end of treatment, Figure 6.9. That supports previous
published findings, whereby the fluid injected with intention to generate
fracture across all formation of interest actually initiates and creates high
conductivity paths at the perforation cluster, potentially leaving the rest
of the productive formation likely unstimulated. The simulations support

the fuzzy clustering analysis, which revealed that perforations placed in
the “good” regions of the formations return the highest production output.
6.8 Conclusions
The paper presents a simple, yet efficient methodology to define the “good”
footage regions for potential perforation in complex and heterogeneous
reservoirs using FCM clustering. The study summarizes a step-by-step pro-
cedure that can assist the geologist and completion engineers in designing
the perforation plan. The methodology demonstrates that the perforation
design can be improved if one uses all the log data available for the well
(such as DRES, GR, and NPHI) as well as the additional information defin-
ing the behavior of the logs. Using Artificial Intelligence (AI) techniques
such as fuzzy clustering analysis methods, the well log data is effective in
identifying rock intervals with similar properties.
The clustering techniques demonstrate an unmatched value by exploit-
ing hidden information from multiple attributes and drawing knowledge
to assist in practical engineering applications. The methodology is flexible
and can be applied to any well where complex lithology creates a challenge
in defining the optimum perforation intervals.
Acknowledgements
The authors would like to thank Chevron North America Exploration and
Production for allowing the publication of this work.
References
1. S.A. Graham and L.A. Williams, Tectonic, depositional, and diagenetic
history of Monterey Formation (Miocene), central Sand Joaquin basic,
California. AAPG Bulletin 69, 385–411 (1985).
2. J. Brink and T. Patzek, Lost hills field trial - incorporating new technology
for reservoir management. SPE 77646-MS, SPE Annual Technical Conference
and Exhibition, 29 September-2 October, San Antonio, Texas (2002).
3. M.S. Bruno and C. Bovberg, Reservoir compaction and subsurface subsid-
ence above the Lost Hills Field, California, Proc. 33rd US Symp. On Rock
Mechanics, AA Balkema, Rotterdam, pp. 263–272 (1992).
4. R. Slocombe, A. Acock, C. Chadwick, E. Wigger, A. Viswanathan, K. Fisher,

and R. Reischman, Eagle Ford Completion Optimization Strategies Using
Horizontal Logging Data” SPE-168693. Unconventional Resources Technology
Conference, 12–14 August, Denver, CO (2013).
5. C.L. Cipolla, X. Weng N. Ond, T. Nadaraja, U. Ganguly, and R. Malpani,
New algorithms and integrated workflow for tight gas and shale completions
SPE-146872, SPE Annual Technical Conference and Exhibition, 30 October-2
November, Denver, CO (2011).
6. A. Peirce and A. Bunger, Interference fracturing: Nonuniform distribu-
tions of perforation clusters that promote simultaneous growth of multiple
hydraulic fractures. SPE Journal 20 (April 2015).
7. F.W. Wilmink and H.T. Uytterschaut, Cluster analysis, history, theory and
applications. Multivariate Statistical Methods in Physical Anthropology,
pp. 135–175 (1984).
8. L. Kaufma and P. Rousseeuw, Finding Groups in Data: An Introduction to
Cluster Analysis, John Wiley, New York, NY (1990).
9. T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning,
Springer, New York, NY (2009).
10. J.B. MacQueen, Some methods for classification and analysis of multivar-
iate observations, Proceedings of 5th Berkeley Symposium on Mathematical
Statistics and Probability (1967).
11. A.P. Dempster, N.M. Laird, and D.B. Rubin, Maximum likelihood from
incomplete data via the EM algorithm. J. Roy. Stat. Soc., Series B, 39(1), 1–38
(1977).
12. M. Ester, H.P. Kriegel, J. Sander, and X. Xu, A density-based algorithm for
discovering clusters in large spatial databases with noise, Proceedings of the
Second International Conference on Knowledge Discovery and Data Mining
(1996).
13. P. Drineas, A. Frieze, R. Kannan, S. Vempala, and V. Vinay, Clustering in
large graphs and matrices. Machine.
7
Horizontal Well Spacing and Hydraulic
Fracturing Design Optimization: A Case
Study on Utica-Point Pleasant Shale Play
Alireza Shahkarami* and Guochang Wang
Saint Francis University, Petroleum and Natural Gas Engineering, Loretto, Pennsylvania
Abstract
Recent drilling and hydraulic fracturing activities in the Utica-Point Pleasant shale
play have recorded true vertical depth of over 13,500 feet. Drilling wells at this
depth is very costly and challenging. With the current commodity pricing, drilling
in such conditions becomes unaffordable.
One immediate solution to the current low energy prices is optimizing well
spacing to enhance hydrocarbon recovery and, thus, the commercial feasibility
of the project. Horizontal well spacing constitutes a fundamental parameter for
the success of a shale-drilling venture. Determining the optimum horizontal well
spacing in shale reservoirs represents a challenging task because of the complexity
of controlling factors. These factors can be categorized into three groups: geologi-
cal, engineering, and economic.
Geological modeling and reservoir simulation are the standard tools utilized in
the industry to integrate these controlling factors. In this study, we employed these
tools to perform sensitivity analysis of reservoir characteristics and future produc-
tion optimization for a deep drilling case study in the Utica-Point Pleasant forma-
tions. We sought to find the optimum horizontal well spacing scenario as well as
hydraulic fracturing design, in order to attain the highest net present value for 50
years of gas production. Our reservoir model represented a portion of Utica-Point
Pleasant formations at the depth of 13,000 feet and the dry gas window. A com-
mercial reservoir simulator was coupled with an optimization algorithm to reach
the best solution with a minimum simulation cost. Although the outcome of our
study is subjective to the chosen asset, the workflow provides a good example of
horizontal well spacing and hydraulic fracturing design optimization.
*Corresponding author: ashahkarami@francis.edu

141
Keywords: Inter-lateral well spacing, production optimization, utica shale,

point pleasant formation
7.1 Introduction
Inter-lateral spacing (well spacing) is one of the biggest decisions in a field
development in unconventional shale plays. Not having optimized well
spacing can either cost the operators a lot of money or leaves a lot of money
on the table. The well spacing must be close enough to help create as much
of the stimulated reservoir volume (SRV) as possible; however, it must be
wide enough to minimize fracture interference [1] (well to well interfer-
ence and “frac hits”) and over-capitalization in a field development [2].
Various tools such as rate transient analysis (RTA) and numerical simula-
tions can be used to optimize well spacing [3, 4]. At the very beginning of
a field development, little data is available; therefore, it is very important to
run sensitivity analysis and uncertainty investigation to assess the effect of
uncertainties involved in reservoir characteristics of the field performance.
Once pressure and production history data are available, the model can be
calibrated to insure all of the assumptions are practically feasible.
Well spacing is impacted by a combination of three types of factors
defining the geological characteristics, the engineering process, and the
economic constraints. The geological and reservoir characteristics assure
the quality of the reservoir and the hydrocarbon presence. The engineering
design provides the deliverability of the well. Finally, the economic factors
warrant that the engineering project leads to profit [5].
The most important parameters that must be heavily studied are matrix
permeability, fracture half-length (impacted by completions design), reser-
voir properties, capital expenditure, operating costs, and hydrocarbon pric-
ing. The hydraulic fracturing enhances the reservoir permeability and leads to
more recovery. The fracture permeability can be upscaled for fluid flow simu-
lation using Oda’s method [6, 7]. In a high commodity pricing environment,
it would make more sense to place the wells closer apart; however, low com-
modity pricing will dictate farther well spacing. A rock that has high matrix
permeability would be better suited for a larger well spacing, but a rock that
has lower matrix permeability is better optimized in a tighter well spacing.
When designing well spacing, it is crucial to take completions design into
account. For example, if 1,000 feet of well spacing is designed for a field devel-
opment, it is very important to design the completions job in a fashion that
would yield 500 feet fracture half-length by pumping enough sand and water,
along with optimizing stage spacing, cluster spacing, and perforations design.
Hydraulic Fracturing Design Optimization 143
Therefore, a big portion of achieving the desired well spacing is creating com-
pletion jobs that would guarantee the attainment of such goals.
Deep dry Utica located in both Pennsylvania and Ohio has sparked the
interest of a lot of the operating companies in the past two years due to its
attracting geology, massive production volumes, and the advantageous time
value of money created in the production of as much volume as possible in
the first 3 to 5 years of the life of the well. The biggest challenges in developing
deep dry Utica have been high capital expenditure, lack of production his-
tory and reservoir rock characteristics, and optimal management of pressure
drawdown to avoid prop-pant crushing/embedment and geo-mechanical
effects. All of these challenges will be resolved and understood with time as
this has been the case for all of the other shale plays developed to date.
In order to create long-term value for the shareholders, the well spacing
must be selected based on a spacing paradigm that yields the highest NPV.
Production volumes could be important to the strategic development of a
company, but the single economic parameter that creates long-term value
for the shareholder of a company is NPV [5]. In this study, the well spacing
was optimized using NPV. Optimal economical well spacing is achieved
by having a solid understanding of fracture half-length, conductivity, and
matrix permeability, which are typically very hard to obtain in unconven-
tional shale plays and reservoir modeling. Therefore, various sensitivities
must be run to understand the impact of these variables.
In this study, we utilized a numerical reservoir simulation in order to
model a dry gas reservoir 13,000 feet deep into a formation that resembles
the Utica-Point Pleasant play. Initially by drilling one well, we studied the
recovery after 50 years of production over a constant drainage area. Next,
we looked for increasing the recovery factor by adding multiple wells.
Finally, we wanted to find the optimum number of wells that leads to the
highest NPV over 50 years of production.
7.2 Methodology
7.2.1 The Base Reservoir Simulation Model
Our reservoir model covers a 1,003 acres area. For the sake of comparison,
we kept the drainage area constant for the different well spacing scenarios.
The drainage area is divided into 35,000 grid blocks, and the grid block size
is 50 feet by 50 feet. The top of the reservoir is 13,000 feet deep, and the pay
zone thickness is 200 feet, allotting into two layers of 100 feet thickness.
These layers resemble the structure of Utica-Point Pleasant play. The reser-
voir contains dry gas, methane.
We used a commercial reservoir simulation package—CMG-BUILDER

and IMEX [8, 9] to build and run the reservoir model. Table 7.1 lists the
reservoir reference data for this study. The reference data was selected after
consulting with multiple operators in the region. Figure 7.1 depicts a 3D
view of the reservoir. In addition, Figure 7.2 shows the relative permeability
Table 7.1 Reservoir characteristics of the base reservoir model.

Base model reservoir characteristics
Rock Compressibility 5.8e-6 l/psi @ 3000 psi
Reservoir Temperature 175 °F
Gas Density at Standard Condition 0.58 Air = 1
Reference Pressure and Depth
Pressure 12,350 psi
Depth 13,000 ft
Gradient Pressure 0.95 psi/ft
Water-Gas Contact (DWGC) 14,000 ft
Grid top (ft) 2015-01-01
13,100
13,090
13,080
13,070
13,060
13,050
13,040
13,030
13,020
13,010
13,000
Figure 7.1 Reservoir top 3D model built in CMG-BUILDER. The drainage area covers
1,003 acres and the reservoir includes 200 feet pay zone divided into two layers.
1.00
Krw vs Sw
0.80
Kr - relative permeability
0.60
0.40
0.20
0.00
0.20 0.36 0.52 0.68 0.84 1.00
Sw
0.80
Krg vs Sl
0.64
Kr - relative permeability
0.48
0.32
0.16
0.00
0.20 0.35 0.50 0.65 0.80 0.95
Sl
Figure 7.2 Relative permeability curves for this study. These curves are modified from the
correlations available in the CMG-BUILDER.
curves utilized in this study. These curves are modified from the correla-
tions available in the CMG-BUILDER. The base reservoir model has eight
percent porosity and one micro-Darcy permeability.
The wells are drilled and perforated in the bottom layer of the Point
Pleasant formation. The length of the horizontal lateral is 7,000 feet. In
order to simulate the hydraulic fractures, we refined the grid blocks in a

planar direction perpendicular to the wellbore. Figure 7.3 demonstrates an
example of planar hydraulic fracturing design.
The production is performed under the constant bottom-hole pressure
of 2,000 psi for a period of 50 years, starting in 2015. The volumetric esti-
mation of initial gas in place is 316.69 BCF. Table 7.2 summarizes the eco-
nomic inputs required for the economic analysis study.
Well-S
Well-3
Well-4
Well-E
Well-S
Figure 7.3 Horizontal wells and planar hydraulic fracturing design.
Table 7.2 Economic analysis inputs.

Economic analysis inputs
Drilling Cost $ 5,000,000 per 7,000 lateral
Hydraulic fracturing $ 300,000 per stage
Royalty 20% of Production
Operating Cost $ 1.00 per MCF
Natural Gas Price $ 2.50 per MCF
Yearly Discount Rate 10%
7.3 Optimization Scenarios

In order to address the main objective of this study, well spacing optimi-
zation, we created five drilling scenarios in a constant drainage area (1,003
acres). Table 7.3 summarizes the number of wells and the corresponding
well spacing selections. In addition to the well spacing, we wanted to find
the optimum number of completion stages, and hydraulic fracture charac-
teristics such as permeability, width, and half length. Table 7.4 shows the
selected ranges for these parameters.
Figure 7.4 demonstrates the workflow that we followed in this study.
This work-flow comprises five general steps. The first two steps include
building the base model and defining the ranges for the key parameters.
Next, we create our objective function considering the field net present
value (NPV) as the global goal to maximize. The process continues with
coupling the reservoir model with an optimization algorithm. The optimi-
zation algorithm selects the values of the adjustable parameters from the
Table 7.3 Five scenarios of inter-lateral spacing and the corresponding well
spacing over a constant drainage area (1,003 acres) in this study.
Scenario Well spacing (ft) #Wells
1 500 8
2 750 6
3 1,000 5
4 1,250 4
5 1,500 3
Table 7.4 Range of completion parameters.

Parameters
Hydraulic Fracturing Spacing 150–450 ft
Number of Stages 46–15
Fracture Permeability 1–10 Darcy
Fracture Width 0.0001–0.005 ft
Fracture Half Length 150–400 ft
Define the Build the Define the Coupling

objective the reservoir NPV and
key base
functon for model to an well spacing
parameters geological
well spacing optimization optimization
and their model
ranges in shale algorithm
reservoirs
Figure 7.4 The workflow for the well spacing and completion parameters optimization study.
range that the user has provided earlier. Then the optimization algorithm
calls the simulator, and, in turn, the simulator runs the models and calculates
the NPV values based on the economic analysis parameters. The output of
this workflow is a collection of reservoir models, which allows us to select
the optimum parameters based on the highest field NPV value. In this study,
the whole process of optimization was carried out using CMG-CMOST [8].

7.4.1 Base Reservoir Model – A Single Well Case
Figure 7.5 shows a scenario of the reservoir model with the base reservoir
characteristics (Table 7.1) and just one well drilled. The figure shows the pro-
file of gas rate (blue curve) and cumulative gas production (orange curve)
for 1, 5, 10, 20, and 50 years after production starts. On the right side of each
graph, the corresponding pressure distribution is also pictured. As it is clear
from Figure 7.5, after 50 years of production, a significant portion of the
reservoir is still untouched and does not contribute to gas production. This
is a good example of how reservoir management methods may leave hydro-
carbon behind. In addition, the total recovery factor in the scenario depicted
is around 25%. Therefore, the need for drilling more wells is expected.
7.4.2 Multi-Lateral Depletion – Finding the Optimum Number

of Wells
Varying the well spacing, we designed five drilling scenarios on a fixed
drainage area (Table 7.3). The optimization results suggest that the
Gas production over 50 years

80000 1000
1 yr 5 yrs 10 yrs 20 yrs 50 yrs
Cumulative gas, MMSCF

70000
Gas rate, MMSCF/day

60000
100
50000
Cumulative gas SC (MMSCF)
40000
Gas rate SC (MMSCF/day)
30000
10
20000
10000
0 1
2015 2025 2035 2045 2055 2065
Time, date
10 years
20 years
50 years
5 years
1 year
Figure 7.5 The cumulative gas production, gas flow rate, and pressure distribution for
different time steps (1, 5, 10, 20, and 50 years).
highest production belongs to 1,000 feet spacing between the laterals

(Figure 7.6), which corresponds to the 5 wells. The second most pro-
lific drilling scenario could be a 4 wells scenario with 1,250 feet spac-
ing. In general, we judge the profitability of a project based on the NPV.
Therefore, Figure 7.7 compares the optimum NPV for the drilling sce-
narios. While the 5 well scenario yields the highest NPV for this field, the
4 well scenario produces our second best NPV. The 4 well scenario could
be an acceptable decision, particularly if there is limited capital budget
for drilling and completion.
In oil and gas development plans, short-term production could be
vital in returning the initial investment. Figure 7.8 compares the recov-
ery factor (RF) after 5, 30, and 50 years for all of the drilling scenarios. As
it is shown in this graph, the 5 well scenario at 1,000 feet spacing has the
highest recovery factor during all three time periods. In addition, con-
sidering just the 5 year recovery factor, a 6 well project built at 750 feet
spacing could lead to a higher recovery factor than a 4 well scenario built
at 1,250 feet spacing. However, we should also consider the limitations
on capital budget. An important observation resulting from these sce-
narios is that almost all of them, except the 3 well (1,500 feet) scenario,
yield more than a 50% recovery factor (almost two thirds of the ultimate
production) after just 5 years of production. This is mainly because pro-
duction of wells is limited to the stimulated reservoir volume.
Field cumulative gas production after 50 years

250
Cumulative gas production, BCF
245
245.09
240
235 235.27 240.02
230
225
223.27
220
215
213.73
210
500 750 1000 1250 1500
Well spacing,ft
8 6 5 4 3
Number of wells
Figure 7.6 Field cumulative gas production after 50 years for five drilling scenarios.
NPV optimization for 50 years of production

140.0
130.0 127.4
119.1
120.0
108.7
110.0
NPV, $M
100.0 96.0
90.0
80.0 74.3
70.0
500 750 1000 1250 1500
Well spacing, ft
8 6 5 4 3
Number of wells
Figure 7.7 The results of net present value optimization study for five drilling scenarios.
Figure 7.9 compares the capital and operating costs of these projects
with the NPV values. As it is expected, a higher number of wells will lead to
a higher capital cost. However, completion projects typically require a sig-
nificant share of capital cost. In addition, a higher production will increase
the operational cost.
Field recovery factor over different time periods

RF after 50 years RF after 30 years RF after 5 years
100.0%
90.0% 81.6% 80.0%
78.3% 79.9%
74.3% 75.5% 77.0%
80.0%
70.3% 71.2%
Recovery factor (RF), %
66.3% 65.5%
70.0%
59.0% 56.4%
60.0% 54.3%
50.0% 42.7%
40.0%
30.0%
20.0%
10.0%
0.0%
500 750 1000 1250 1500
Well spacing, ft
8 6 5 4 3
Number of wells
Figure 7.8 Field recovery factor (RF) over different time periods for all the drilling scenarios.
Capital cost, $M Operating cost, $M Optimum NPV, $M

210 (194.4)
(174.7) (172.7)
180 (162.3)
150 127.4 (136.2)
119.1
120 108.7
$M
96.0
(88)
90 74.3
(66) (67)
60 (53.6)
(40.2)
30
0
500 750 1000 1250 1500
Well spacing (ft)
8 6 5 4 3
Number of wells
Figure 7.9 Comparison of capital and operating costs with NPVs for different drilling scenarios.
7.4.3 Completion Parameters

Table 7.5 summarizes the outputs of the optimization process for the com-
pletion parameters. Based on these results, a higher fracture conductivity
constitutes the best solution. However, the fracture half-length is regulated
by the well spacing, as expected. For instance, for the well spacing values
152
Table 7.5 Optimized completion parameters.
Fracture spacing Fracture conductivity Fracture half

Scenario Well spacing (ft) # Wells (ft) # Stages (md-ft) length (ft)
1 500 8 350 20 50 250
2 750 6 350 20 47.7 350
3 1000 5 250 28 48.7 400
4 1250 4 250 28 50 400
5 1500 3 250 28 50 400
Hydraulic Fracturing and Well Stimulation
of 500 and 750 feet, fracture half-lengths are 250 and 350 feet, respectively.
Obviously, higher values of fracture half-length will cause interference
between the fractures (“frac hit”) and loss of injected fluids. The number of
fracture stages depends on the space between the stages (fracture spacing).
In addition, the cost of completion has a direct relationship with the num-
ber of stages. Therefore, an economic optimization analysis is required. We
could suggest a higher number of stages for a lower number of wells per a
constant drainage area and vice versa.
7.4.4 Second Economic Scenario, Reducing the Cost

of Completion
The optimization process is driven by the commodity costs. We created
another scenario by reducing the cost of completion from $300,000 to
$200,000 per stage. Figure 7.10 compares the NPV values for the two sce-
narios of completion costs for all drilling cases. As it is clear from this image,
over the 50 years of production, we observe almost a constant shift (11–12%
of NPV) between these scenarios.
An interesting refection results from comparing the recovery factor
values after 5 years of production for the two cost scenarios depicted in
Figure 7.11. For well spacing cases of 750 and 1,000 feet, the reduction
of completion costs leads to a recovery factor increase of 4.6% and 2.4%,
respectively. These changes could be pivotal in decision-making, since in
NPV optimization for 50 years of production

HF cost scenario 1 HF cost scenario 2
150.0 141.3
140.0 131.3
130.0 122.5
NPV, $M
120.0 127.4 108.0

110.0 108.7 199.1
100.0 90.3
90.0 96.0
80.0
74.3
70.0
500 750 1000 1250 1500
Well spacing, ft
8 6 5 4 3
Number of wells
Figure 7.10 Comparison of NPV between two costs of completion scenarios.

Field RF comparison after 5 years between two cost scenarios

RF after 5 years-HF cost scenario 1 RF after 5 years-HF cost scenario 2
100.0%
90.0% 4.6% RF
2.4% RF
Recovery factor (RF), %
80.0% increase
increase
70.0% 63.6% 66.3% 66.2%
59.0% 56.4% 59.0%
60.0% 54.3% 54.3%
50.0% 42.7% 44.7%
40.0%
30.0%
20.0%
10.0%
0.0%
500 750 1000 1250 1500
Well spacing, ft
8 6 5 4 3
Number of wells
Figure 7.11 Field recovery factor comparison after 5 years between two completion costs
scenarios.
5 years of gas production shale reservoirs could bear more than 70% of
ultimate production. The main reason behind the increase of recovery
factor values goes back to having a higher number of completion stages
due to the reduction of cost (Table 7.6). However, for the well spacing
of 500 feet, the even increase of stages from 28 to 35 will not impact the
production level after five years. This might be because of the tight well
spacing that already covers all of the reservoir volume.
7.5 Conclusion
In addition to the reservoir quality and engineering factors, the commod-
ity costs and capital expenditure guide oil and gas development plans. In
this project, we utilized geological modeling and reservoir simulation to
study the profitability of a shale dry gas asset in the Utica-Point Pleasant
formations. Due to the very tight formations, horizontal drilling and multi-
stage hydraulic fracturing were planned to increase the reservoir deliver-
ability. Our reservoir model simulates a portion of Utica-Point Pleasant
formations at 13,000 feet deep and 200 feet thick. The drainage area covers
1,003 acres. Our simulation results showed that due to the tight shale for-
mation, drilling just one well would leave a significant amount of natural
gas behind. We wanted to find the optimum number of lateral wells over
Table 7.6 Optimized completion parameters for the completion cost scenario of $200K per stage.
Fracture Fracture conductivity Fracture half

Scenario Well spacing (ft) # Wells spacing (ft) # Stages (md-ft) length (ft)
1 500 8 350 20 47.8 250
2 750 6 250 28 36.8 350
3 1000 5 250 28 48.2 400
4 1250 4 200 35 49.5 400
5 1500 3 200 35 49 400
Hydraulic Fracturing Design Optimization
155
the drainage area. By integrating the reservoir characteristics, economic

data, and operational parameters the best drilling and completion scenar-
ios were projected.
Based on two different completion cost scenarios, $300,000 and $200,000
per stages, the optimum NPV for 50 years of production results from a 5
well drilling project. This number of wells includes 1,000 feet well spacing
and 28 stages for each well. According to our optimization analysis, this
project leads to a 77.4% recovery factor over 50 years.
Drilling horizontal wells is a costly adventure and requires the availabil-
ity of a significant capital budget. For this case study, a drilling scenario
with just 4 well and 1,250 feet well spacing might be an attractive project.
Although a 4 well scenario results in 6.5% less NPV compared to a 5 wells
project, it will decrease the capital cost by 20%.
In shale formations, the main part of production happens in the first
couple of years. In our case study, almost 70% of ultimate production
occurs in the first five years of production. This could be considerable since
it can return the initial investment quickly. Our results demonstrate that a
5 well scenario gives the highest production after 5 years. However, chang-
ing the cost of completion can render the 4 and 6 well scenarios.
Finally, although the outcome of our study is subjective to the chosen
asset, the practice provides a good example of horizontal well spacing and
hydraulic fracturing design optimization.
Acknowledgments
The authors wish to extend their appreciation to Computer Modeling Group
(CMG) for providing the software applications for reservoir simulation.
References
1. A. Awada, M. Santo, D. Lougheed, D. Xu, and C. Virues, Is that interference?
A work flow for identifying and analyzing communication through hydrau-
lic fractures in a multiwell pad. SPE J. (2016). DOI: 10.2118/178509-PA
2. R. Barree, S. Cox, J. Miskimins, J. Gilbert, and M. Conway, Economic opti-
mization of horizontal well completions in unconventional reservoirs. PE
Hydraulic Fracturing Technology Conference, The Woodlands, Texas, USA,
Society of Petroleum Engineers (2014). DOI: 10.2118/168612-MS
3. A.S. Ziarani, C. Chen, A. Cui, D.J. Quirk, and D. Roney, Fracture and wellbore
spacing optimization in multistage fractured horizontal wellbores: Learnings
from our experience on Canadian unconventional resources. International
Petroleum Technology Conference. Kuala Lumpur, Malaysia: International

Petroleum Technology Conference (2014).
4. T.C. Frick, Importance of economics in production and reservoir engineer-
ing. J. Petrol. Technol. 10(9), 11–12 (1958). DOI: 10.2118/1026-G
5. F. Lalehrokh and J. Bouma, Well spacing optimization in eagle ford. SPE/
CSUR Unconventional Resources Conference – Canada. Calgary, Canada,
Society of Petroleum Engineers (2014). DOI: 10.2118/171640-MS
6. M. Oda, Permeability tensor for discontinuous rock masses. Geotechnique.
35, 483 (1985).
7. P. K. Ghahfarokhi, The structured gridding implications for upscaling model
Discrete Fracture Networks (DFN) using corrected oda’s method. J. Petrol.
Sci. Eng., 153, 70–80 (2017).
8. Computer Modeling Group, Computer Modelling Group Manual (2015).
9. C. L. Cipolla, E. P. Lolon, J. C. Erdle, and B. Rubin, Reservoir modeling
in shale-gas reservoirs. Soc. Petrol. Eng. J. 13(4) (2010, August 1). DOI:
10.2118/125530-PA
Part 4
FRACTURE RESERVOIR
CHARACTERIZATION
Ahmed Ouenes
Introduction
The role of discontinuities and the major impact they can have on the out-
come of hydraulic fracturing has been recognized since the early days of
field experiments carried by the Department of Energy and Gas Research
Institute. Warpinski and Teufel [1] described many of these issues and
attempted to model them with the best tools they had at their disposal at
that time. Unfortunately, that initial effort was not continued, and the shale
revolution was built on that initial knowledge and did not expand it sig-
nificantly. With the oil crisis of 2014–2017, followed by financial pressures
imposed by investors in 2018–2019 who were not informed about the risks
of frac hits and well interreferences, the entire industry has to catch up on
years of “natural fracture denial” where the hydraulic fracturing design of
thousands of wells used software and technologies that does not include
any representation of the natural fractures. New evidence from field exper-
iments such as the one conducted at the Hydraulic Fracturing Test Site I
industry consortium (HFTS-1) is again highlighting the importance of the
natural fractures.
The recent findings of HFTS-1 are well summarized in the September
2018 Journal of Petroleum Technology article titled “Real Fractured Rock is
So Complex it’s Time for New Fracturing Models” [2]. 600ft of core taken in
a hydraulically fractured Wolfcamp reservoir in the Permian Basin, USA
showed a more complex reality than what is accounted for in most hydrau-
lic fracturing design and analysis software. This includes the interaction
between hydraulic and natural fractures, which is largely ignored or poorly
accounted for in most software currently used to model hydraulic fractures
and their resulting geometry. Rassenfos [2] emphasized in his summary of
multiple recent publications describing HFTS 1 findings, that the “fracture
height is overrated. While microseismic testing indicated that fractures grew
up about a 1000ft, the height of the propped fractures- the fractures most

160 Part 
likely to produce oil and gas was about 30ft.” Rassenfos summary article
discusses the important role played by the natural fractures.
The impact of natural fractures and discontinuities in general on hydraulic
fracturing can be seen at different scales and manifest itself in different forms.
In this section, four articles attempt to illustrate this impact while showing
the variety of data that can be used to enhance our understanding and val-
idate our new models. The four articles also illustrate the various numerical
approaches which may be used to incorporate the effects of discontinuities.
In Umholtz and Ouenes, the effects of faults and discontinuities is exam-
ined at a regional scale where local changes in stress created by a complex
network of faults could inhibit or promote earthquakes resulting from
stress perturbations created by deep water injection or hydraulic fracturing.
Most of the past work in this area was limited to the conventional study
by seismologists of fault stability which frequently was restricted to one
or a limited number of these faults thus not capturing the true stress field
created by a complex fault network. The authors used the Material Point
Method (MPM) to incorporate a large number of faults over a regional scale
to numerically compute the various stress properties that could be used to
compute an Induced Seismicity Potential (ISP) that was validated using
limited public data from Oklahoma. The use of numerical methods such
as MPM applied to continuum mechanics augmented with discontinuities
could provide the necessary means to capture the effects of the faults and
fractures if adequate data is provided to model these critical discontinuities.
Maity’s section shows the various data that can be used to capture the
effects of natural fractures during hydraulic fracturing. In this article, some
of challenges related to data collection and availability are highlighted.
Maity studied two Marcellus wells that had not only microseismic data
but also production and image logs which provided a quantification of
the natural fractures at different stages. To quantify the performance of
the hydraulic fracturing at each stage, Maity used a modified Nolte-Smith
diagnostic tool to correlate different diagnostic parameters to production
logs, microseismic and fracture density estimated from image logs. The
microseismic data was quantified using the b-value. When plotting the
various diagnostic plots against the production logs, b-value from micro-
seismicity, and natural fracture from images logs, Maity found some weak
correlations when using the production performance. However, stronger
correlations were found between the b-value from microseismic with the
extensional growth parameters compared to natural fracture interaction
related parameters. The b-value showed no correlation with dilatational
fracture swelling related parameters. Lastly, Maity observed no appar-
ent correlation between the derived fracture density and the modeled
Part  161
diagnostic parameters. This was expected since the image logs provide
indication on the near wellbore region and cannot estimate the presence
or absence of actual fractures in the formation. The author emphasized the
multiple challenges associated with each type of data and the pitfalls that
could result from not addressing data quality issues.
Given all the challenges capturing the effects of natural fractures with
the data that could possibly be available, Alzahabi et al. used a more engi-
neering oriented approach to capture the effects of natural fractures and
the resulting interaction with hydraulic fractures is homogenized into a
Stimulated Reservoir Volume (SRV) with a certain average property. Given
a certain SRV distribution estimated with simple geometrical properties
of a fracture geometry, the goal is to find the best well positions using an
optimization problem based on Integer Programming.
Continuing into more computational approaches, Cai et al. use a semi-
analytical approach to predict the well performance in the case of refractur-
ing. The authors use the concept of fracture-matrix cross-flow for a single
fractured wellbore to derive their equations. The semi-analytical approach
assumes some reservoir properties and focuses on the number of hydraulic
fractures and their impact on the well productivity. Testing the approach
against actual performances shows that the approach overestimates the
well performance by 10%. The authors attribute this overestimation to
the uncertainties in the reservoir parameters which will always remain an
issue when using analytical and semi-analytical approaches. The sensitivity
analysis conducted by the authors shows the increase in productivity as the
number of hydraulic fractures increases but the benefits of such an increase
levels out when the fracture conductivity reaches 2000 md-in.
With these four articles, the reader is exposed to the variety of problems
encountered in representing of the natural and hydraulic fractures and
the consequences they have on the well performance. The important take
away, is that multiple approaches exist and are in continuous development
to create a toolbox that engineers can use to solve their many hydraulic
fracturing challenges.
References
1. Warpinski, N.R. and Teufel, L.W., Influence of geologic discontinuities on
hydraulic fracture propagation. J. Petroleum Technol., 39, 02, pp.209-220,
1987.
2. Rassenfos, S., Real Fractured Rock is so Complex it’s Time for New Fracturing
Models, Journal of Petroleum Technology. 19 September 2018.
8
Geomechanical Modeling of Fault Systems
Using the Material Point Method –
Application to the Estimation of
Induced Seismicity Potential to Bolster
Hydraulic Fracturing Social License
Nicholas M. Umholtz and Ahmed Ouenes*
FracGeo, The Woodlands TX, USA
Abstract
In an effort to promote awareness and understanding of the phenomena of induced
seismicity, geomechanical modeling is applied to large publicly available datasets
to demons trate the potential for bolstering social license. The Material Point
Method (MPM) is used to simulate the interaction of fault systems with regional
stresses. By combining mechanical results of the simulation to create induced seis-
micity potential (ISP) proxies, maps are generated to express areas of high and
low inducement potential of seismic events. The results are compared to recent
earthquake epicenter and injection well data. High coincidence of earthquake epi-
centers with regions of predicted high induced seismicity potential suggests the
workflow presented could be deployed to quantify the risk of induced seismicity
associated with the location of high-volume injection wells. The addition of a tool
to assess the impact of location, not only injection volumes, is another critical
step towards responsible regulation of injection wells, and mitigation of induced
seismic events.
Keywords: Induced seismicity, geomechanical modeling, material point method,

Oklahoma earthquakes, Alberta earthquakes, social license, water disposal
*Corresponding author: aouenes@fracgeo.com

163
8.1 Introduction
Recent media coverage has highlighted a major shortcoming in the indus-
try of unconventional resource production, inducement of seismic events.
Seismicity in certain areas of the midcontinent North America has increased
drastically in the past few years. This has come to be understood as a result
of pore pressure, and stress field perturbation due to water injection. Water
disposal and hydraulic fracturing practices both introduce quantities of
water to the subsurface which may alter the mechanical quasi-equilibrium
established through geologic time. Such rapid perturbation of established
stress fields can in some cases cause earthquakes. Because many of these
operations are based in historically seismically quiescent regions, even
magnitude 3–5 earthquakes may cause significant damage. The public’s
perception and media’s coverage of this phenomena has produced a signif-
icant discussion topic, which impedes the rise of Shale 2.0, which will seek
to achieve a more efficient development of shale resources. It is clear that
operators do not understand the mechanical impact of such injections, as
unpredicted seismic events have been induced at an increasing rate and of
increasing magnitude. Regulators in Oklahoma have taken steps to reduce
the risk of operators inducing seismic events, however the lack of means to
quantify induced seismicity potential has hindered deployment of effective
policy. The new tougher regulations imposed in Oklahoma are limited to
reducing injection volumes and shutting down a limited number of water
disposal wells around areas affected by recent earthquakes. In other words,
the regulators are mostly reacting to the increasing earthquakes rather than
proactively regulating the disposal process. With the advent of induced
seismicity, institutionalized trust is compromised once again, and this time
during a serious shift in the economies of unconventional production,
underpinning not just the importance of the social license to operate, but
the effect such public support can have when attempting to deploy the vast
infrastructure necessary to realize unconventional petroleum resources at
half of the value which they were previously produced.
In order to address these concerns, a workflow is proposed, to com-
bine the Material Point Method (MPM) and continuous fracture modeling
(CFM) technologies with large geologically and geophysically constrained
datasets. This workflow is applied to investigate the relationship between
complex regional fault systems, and their collective impact on stress fields
in a region. The mechanical outputs of this modeling workflow can be
combined into proxies, and used to investigate areas of the model with
stress fields more likely to be perturbed through high-volume injections
Geomechanical Modeling of Fault Systems 165
or hydraulic stimulation. Plotting these proxies as a continuous distribu-

tion on a map provides a practical tool that leads to very fast results that
are simple to interpret, and which leverages large amounts of data and
advanced modeling. This simple tool can be utilized by geoscientists, reg-
ulators, insurance companies, and the public, whose support for the petro-
leum industry is critical for the social license to operate.
8.2 The Social License to Operate (SLO)

In Thompson & Boutilier [1], the concept of the social license to operate,
and its direct relationship with the economic viability of an industry, is dis-
cussed. Traditionally, industries with large aerial, and temporal impacts on
a community, such as mining, have been considered in such a light. Now,
with the advent of unconventional resource production, the manufactur-
ing approach of deploying hydraulic fracturing for unconventional petro-
leum resource production from shales (fracking) has drastically increased
the footprint of the petroleum industry. This increased footprint elevates
the potential for increased adverse effects as a result of industry presence
and practices. Considered in this study is induced seismicity.
In the pyramid of social trust, there are many boundaries which must
be surmounted to approach an institutionalized state of trust on the part of
the public (Figure 8.1). The public perception of the industry must evolve
through legitimacy, to credibility, and finally trust, to provide the commu-
nity support necessary to develop resources most efficiently.
Trust
Psychological Institutionalized
boundary
identification trust
Credibility
Approval boundary
Socio-political Interactional
Acceptance legitimacy trust
Withheld/ Legitimacy
withdrawn boundary Economic
legitimacy
Figure 8.1 Diagrams showing the importance of credibility and trust in promoting
social license. Critical to the highest level of public support, institutionalized trust, is
sociopolitical legitimacy, and a transition from credible operation to trusted operations.
Modified from Thomson and Boutilier [1].
In order to increase the public stakeholders in the industry, these

boundaries: legitimacy, credibility, and trust, must be satisfied. In the cur-
rent price environment, the economic legitimacy will be realized again in
Shale 2.0 as new technologies drive production costs lower into a sustain-
able range. Interactional trust and the perception of socio-political legiti-
macy will only be established if new technologies can also serve to educate
the public, and enhance regulators’ ability to ensure safe operations. If all
of these conditions are satisfied, institutionalized trust can be established,
maximizing public stakeholders’ engagement in unconventional resource
production.
The inability of regulators to stay ahead of such a significant indus-
try impact as induced seismicity has cautioned landowners and the
general public’s support of the unconventional petroleum industry. As
volatile energy commodity prices, driven by geopolitical factors, con-
tinue to challenge the development of North American shale resources,
it becomes exceedingly important to increase public stakeholders in the
industry.
Historically, there have been many setbacks to the public’s perception
of the petroleum extraction industry. With the advent of hydraulic frac-
turing, a major concern which developed in the public’s eye was consider-
ation of aquifer contamination. Sophisticated studies have been leveraged
to both absolve most hydraulic fracturing from association with this con-
tamination, and highlight shortcomings in completions, which have led
to safer practices and more thorough inspections on the industry’s part
to mitigate the hazard of aquifer contamination [2]. In a similar fashion,
sophisticated multidisciplinary modeling tools are deployed in this study
to continue a rigorous and scientific exploration of the potential drivers
of induced seismicity.
8.3 Regional Faults in Oklahoma, USA and Alberta,

Canada used as Input in Geomechanical Modeling
To address the issue of induced seismicity, collaboration between leading
experts in many fields and disciplines is the only viable course of action.
No single expert, whether they be hydrologically, seismically, or mechan-
ically inclined, will provide a satisfactory way to address the issue. It is
through the synthesis of these many disparate topics, that comprehensive
models may be developed to help understand the ways in which humans
may perturb evolving stress fields within the earth’s crust, and induce seis-
mic events.
Oklahoma scientists and regulators have taken a stand to address this

growing concern amongst their constituents. To this end, large, publicly-
available datasets are being considered in numerous respects. Here, fault
information (Figure 8.2) will be used as input in a geomechanical simula-
tion that uses an advanced numerical model [3], the results of which will
be compared to injection well and seismic epicenter data (Figure 8.3).
Regulators in Alberta, Canada are also concerned, and have also hosted
publicly available regional fault datasets, examined in Figure 8.4.
Oklahoma fault map
Figure 8.2 Fault map of Oklahoma [12]. These fault data are used as input to the
geomechanical models simulated in this study.
Magnitude 3+ earthquakes 2006-present Cumulative injection volume 2006–2013
Figure 8.3 Available earthquake and Underground Injection Control (UIC) well data
from 2006-Present [13, 14]. Note the lack of earthquakes in multiple areas where large
volumes were injected.
(c)
(a)
(b)
(d)
Figure 8.4 (a) Extent of the Montney/Duvernay shale play and considered area of
interest, (b) Regional faults mapped by the Alberta Geologic Survey, (c) Faults in the
study area between the cities of Fox Creek and Saint John where earthquakes of 4.2 and
4.5 magnitude, respectively attributed by Alberta Geologic Survey to hydraulic fracturing,
were recorded in 2015, (d) major faults extracted for the geomechanical simulation and
subjected to the regional stress.
Using these publicly available datasets, and new geomechanical mod-

eling techniques, a workflow is proposed to develop fast industry tools
to assist oil and gas companies, insurance companies, and regulators in
addressing these public concerns.
A geomechanical workflow which incorporates geological & geophys-
ical data, with geomechanics [4, 5], is deployed on a regional scale to
model the stress heterogeneities which exist as a results of the interaction
between far-field regional stresses, and complex subsurface fault systems.
Development and refinement of such a model on an oil or gas field level,
where 3D seismic is available to accurately map all the existing faults, will
provide a truly predictive tool to assess the risk of inducing a seismic event
in any area where it is possible to leverage actual geologic and geophysical
data.
8.4 Modeling Earthquake Potential using Numerical

Material Models
In attempting to develop a predictive model for seismicity, and especially
induced seismicity resulting from water disposal or hydraulic fracturing,
the primary drivers of the phenomena must be characterized. An earth-

quake is the surface wave we experience standing on the earth as defor-
mation equilibrates after stresses, built up through tectonic loading over
time, release due to critical loading of a fault past a frictional restraint. It
is therefore important to quantify the interaction between complex fault
systems and regional stresses.
What is a rock but a material, a material which to the ease of examina-
tion is populated with regular lattices, defined by varying minerals. The
interfaces of the minerals, and in some rocks, the altered mineral inter-
faces, provide defects in the material where the potential to be separated is
much higher than the atomic bonds within the mineral lattices. Through
time, deforming the material will nucleate along the interfaces, and gener-
ate voids, or line deformations, which when given some amount of throw,
in the very large scale are considered as faults. When considering the entire
material, these defects will serve as a preferential nucleation zone for fur-
ther deformation, especially when these fractures are in certain orienta-
tions with respect to the principle stress axes. A more complete discussion
of the history of the study of failure in rocks can be found in Scholz [6].
It is therefore critical, for any model purporting to act in a predictive
capacity, with regards to induced seismicity and seismic potential, to real-
istically capture the interaction of fractures (faults) with confining stresses,
and furthermore, the interaction of multiple fractures with each other and
such confining stresses. Successful prediction of the stress fields which arise
from such interaction, will provide the ability to assess fault stability, and
consequently seismicity potential within an evolving stress environment.
Previous approaches to constraining the effects of earthquakes on sub-
sequent earthquake potential have focused on the Coulomb failure criteria,
when a Coulomb stress, Cf, exceeds a value
Cf = τβ + μ(σβ + ρ)
where τβ is the shear stress on the fracture plane, σβ is the normal stress on
the fracture plane, p is the pore pressure, and μ is the coefficient of friction
along the fracture plane [7].
In these models, the Coulomb stress change, which is the difference
between the Coulomb stress distribution before and after a major fault slip,
highlights regions of stress perturbation, in many cases associated with
subsequent seismicity or aftershocks. To better understand these complex
stress fields and their relation to earthquakes on a field or regional scale,
a new geomechanical workflow will be used and compared to previous
results.
A major research topic that has preoccupied mechanical engineers

for decades is modeling materials with discontinuities (cracks, fractures,
faults). This is a critical research area since it is related to material failure
which can cause major losses of life in areas such as aeronautics, and civil
engineering. During the last decades major advances have been made in
the study of crack propagation, crack intersections and material failure
prevention. However, in most industries the number of fractures studied
or modeled remain small compared to the problems found in the petro-
leum industry where rock failure is sometimes the goal. In hydraulic frac-
turing, the petroleum industry attempts to create a stimulated permeability
by creating rock failure through hydraulic fracturing and leveraging the
presence of natural fractures and the existing localized rock weakness
where earth discontinuities have been formed throughout geologic time.
When working at the scale of a well, this problem of studying the interac-
tion between hydraulic fractures and a complex network of natural frac-
tures poses a major engineering challenge. Modeling this large number of
fractures and their propagation and interaction poses major computational
and modeling challenges to classical numerical simulation methods such
as finite elements. When modeling the geomechanical behavior of a large
area that covers an entire oilfield or a producing trend the same problems
arise since the number of faults present in the area could become over-
whelming to most classical numerical simulation methods. To address this
issue of handling a large number of discontinuities (natural fractures or
faults) in geomechanical simulations, the Material Point Method (MPM)
was introduced by Aimene and Nairn [8]. The introduction of the MPM
method was accompanied with and continues to drive the ability to create
new workflows that use quantitative data and information from geology,
geophysics and geomechanics [4, 5].
The Material Point Method (MPM) is a meshless method developed as
a potential tool for numerical modeling of dynamic solid mechanics prob-
lems [9]. It represents an alternate approach, with alternate characteris-
tics, for solving problems traditionally studied by Dynamic Finite Element
Methods. In MPM, a material body is discretized into a collection of
points, called particles. It uses a background mesh as a computation space
which allows the model to capture rapid and large deformation. Solid body
boundary conditions are applied to the grid and/or on the particles. At
each time step, the particle information is extrapolated to the background
grid, to solve these equations. Once the equations are solved, the grid-
based solution is used to update all particle properties such as position,
velocity, acceleration, stress and strain, state variables, etc. This combi-
nation of Lagrangian (particles) and Eulerian (grid) methods has proven
useful for solving solid mechanics problems. Nairn [10] extended MPM
to handle explicit cracks, resulting in the CRAMP method. By imposing
discrete discontinuities in the model, CRAMP is able to simulate fractured
materials using elastic fracture mechanics.
The major inputs for MPM simulations applied to a study area are four-
fold: 1) the distribution of rock geomechanical properties, such as Young’s
Modulus, in the study area, 2) the distribution and properties of the discon-
tinuities such as natural fractures or faults, 3) the distribution of the pore
pressure, and 4) the boundary conditions representing the far field stresses
acting on the study area. The MPM formulation is used to solve the clas-
sical dynamic continuum mechanics equations that include the momen-
tum equation. The results of the MPM simulation is the usual stress and
strain at the end of the dynamic simulation when the study area reached a
quasi-equilibrium. More details on MPM and its use in the geomechani-
cal modeling of multiple fractures and faults can be found in Aimene and
Ouenes [3].
The MPM geomechanical computation was used to investigate a variety
of geomechanical and engineering issues related to unconventional petro-
leum production at the well scale [3, 4]. Through several field validations,
at the well and pad scales, this workflow has been demonstrated to success-
fully capture local stress variations, and predict complicated microseismic
distribution in a number of stimulated wells. In this study, the workflow
is now deployed on a regional scale inquiry to capture the interaction
between large, complex, regional fault systems, and regional stresses. At
the regional scale, the distribution of elastic properties and pore pressure
used in the MPM simulation are considered constant. The discontinuities
are represented by the interpreted regional faults. To test the validity of
these assumptions at the regional scale, the results from MPM simulations
are compared to previous studies conducted by seismologists studying the
distribution of earthquakes near a fault.
A simple case study using a boundary element method, from King
and Cocco [7], is reproduced using MPM. The model presented by King
and Cocco [7] plots a Coulomb stress change (Figure 8.5a), as a result of
movement of the fault in the model. An approximation of Coulomb stress,
captured by outputs of the MPM model, is compared to the distribution
of stress in the original model, and is strikingly similar (Figure 8.5a,b).
A proxy for Coulombs Stress (PCS) (Figure 8.5b) is defined as:
PCS = 0.5 * (σH − σh) * (sin(2β) − 0.4 * cos(2β)) (8.1)

− 0.5 * μ * (σH + σh) + μ * p
(a) (b) (c)
Figure 8.5 (a) Coulombs stress change as computed by King & Cocco [7], (b) An
approximation of Coulomb stress computed in this work using MPM, (c) An Induced
Seismicity Potential, ISP1, computed in this work using MPM.
where σH and σh are maximum and minimum principal stresses respec-

tively, μ is the coefficient of friction, and β represents the angle between the
considered fault and the compressive stress orientation.
The goal of computing PCS and other possible Induced Seismicity
Potential or (ISP) proxies is to demonstrate the possibilities for new
quantifications of energy in the medium which may elucidate the regions
most likely to express seismicity related to a fault in a specific stress field.
Specifically, in this work a newly derived ISP1 is used,
ISP1 = |τ13| * |(σ11 − σ33) − (σH − σh)| (8.2)
where τ13 is shear stress, σ11 is the maximum principle stress at the par-
ticle in the simulation, σ33 is the minimum principle stress at a particle,
σH is the input maximum horizontal stress, and σh is the minimum hor-
izontal stress input to the model. This quantity captures a particle’s sub-
jection to shear stress and stress field perturbation due to the presence of
a fracture. Plotting the proxy highlights areas near the fault that display
an asymmetric distribution of aftershocks proximal to and oriented along
the fault (approximately N-S), with quiescent regions surrounding the ele-
vated occurrence of aftershocks, and with a secondary concentration of
aftershocks distributed E-W further from the simulated fault (Figure 8.5c).
Advances in processing power, and using MPM to simulate fracture
mechanics, is desirable when investigating earthquakes as it allows for
more realistic and comprehensive inputs to be considered by the model.
Advancing from models with only a single or handful of faults, a model is

considered in this study which captures hundreds of regional faults, simu-
lated in discrete sections, yielding over 2000 independent fracture planes
to be simulated. An induced seismicity proxy, which is not dependent on
calculations involving individual fracture plane orientation, greatly reduces
the computational inefficiencies which have plagued previous inquiries.
8.5 A New Workflow for Estimating Induced

Seismicity Potential and its Application
to Oklahoma and Alberta
For decades, the industry of petroleum exploitation has developed organi-
zational structures that “siloed” professionals, which could have stymied
multidisciplinary innovation. The ever-evolving economies of petroleum
extraction continue to appreciate the value of data gathering, and espe-
cially the analysis and synthesis of those data into valuable information.
To circumvent the shortcomings of siloing, and enhance the translation
of value between geologists and geophysicists (G&G) and engineers, a 3G
workflow that leverages the quantitative and simultaneous use of geology,
geophysics and geomechanics was introduced to the industry [3, 5]. The
use of a 3G workflow provides the fundamental framework within which
to develop numerous subsidiary workflows to combine, enhance, and apply
large G&G datasets with powerful mechanical modeling tools to address
multiple issues related to unconventional resource production.
Here, the workflow introduced by Umholtz & Ouenes [11] is applied
to understand how large, regional-scale fault networks serve to perturb
far-field stresses defined by tectonic loading. The resulting stress field attri-
butes of such a model are then combined to generate an Induced Seismicity
Proxy, which serves to identify regions of the model more or less likely
to be perturbed by nearby or farther away large injection volumes. The
resulting ISP predictions depend entirely on the preliminary input fault
network, which is a work in progress [12].
Two areas in Oklahoma are of elevated interest to regulators (Figure
8.6), the area around the large population center of Oklahoma City, and an
area slightly northwest, which has hosted an increasing number of seismic
events since 2009. These will serve as the focus of the current study given
the initial fault interpretation released by Oklahoma regulators [12].
Additionally, an area in Alberta, Canada (Figure 8.4), which does
not yet demonstrate a significant number of seismic events, is mod-
eled to demonstrate the workflow in a region where a few large induced
Simulation 2
Study area
Simulation 1
EFM
EFM
Figure 8.6 Equivalent Fracture Models (EFM) derived from an area of interest in
Oklahoma for input into geomechanical models.
seismicity events exceeding magnitude 4 have been attributed to hydrau-

lic fracturing.
With a powerful tool, able to capture the interactions between stress
fields and complicated fracture systems, the next step in developing an
induced seismicity proxy is selecting the appropriate outputs of the model
to represent areas of preferential fault growth in the evolving stress field.
As many authors modeling fault propagation have noted, Coulomb stress
change appears to highlight areas of historical aftershock generation related
to ruptures along large-scale faults. Building on these efforts, an alterna-
tive induced seismicity proxy is developed which combines both the shear
strain and differential stress changes in the model as a result of tectonic
loading on the regional fault system. This geomechanical proxy captures
both the preferentially stored energy, represented by differential stress
(σHmax − σhmin), and the shear strain on the particles (dv/dx and du/dy), cap-
turing regions likely to accommodate slip along critically oriented planes.
dv du
ISP ( H max h max )
dx dy
The results of the simulation using the preliminary fault interpretation

of Holland [12] are summarized with the ISP proxy plotted in Figure 8.7.
Regions of elevated ISP are warmer red colors, while areas of low ISP are
cooler blue colors. Many of the regions of elevated ISP are found proximal
ISP
High
Low
Figure 8.7 Result of the geomechanical proxy for ISP. Compared to historical earthquakes
and underground injection control well data.
to fault systems, especially those critically oriented. Some regions of high

ISP however, form conduits connecting less active areas of the fault net-
work. Comparison to earthquakes data and injection well data can help
substantiate these interpretations.
Figure 8.7 compares the ISP proxy to earthquake data. We see that many
of the earthquakes observed in Oklahoma fall into areas of high ISP, and
many are located near critically oriented faults. Areas of low predicted ISP
are often lacking in induced seismicity.
Further comparison of the dataset, with consideration of injection well
data, shows interesting features in low ISP zones. Namely, that some of
the largest volumes injected, do not coincide with seismicity, and these
instances are all in areas predicted as low ISP (Figure 8.8).
For these reasons, the predicted ISP map (Figure 8.7) is a promising tool
to help understand and regulate induced seismicity in Oklahoma by add-
ing a missing critical component: location of the injection wells as com-
pared to the geomechanical interaction between the regional stress and
the regional fault network. Most regulatory efforts made to reduce induced
seismicity potential are targeted at reducing injection volumes. These pre-
liminary results indicate that the location of these injections may serve as a
primary control over the expression of seismic events as the result of stress
field perturbation through high volume injections.
We also note that the preliminary fault interpretation of Holland [12]
and its use in the resulting ISP shows some inconsistencies with recent
major earthquakes (Figure 8.9). It is obvious from these recent earthquakes
as well as previous earthquakes in that area that the initial fault interpreted
by Holland [12] continues towards the North East and does not stop as
shown in the preliminary fault interpretation. This observation shows the
importance of accurately mapping all of the faults in high seismicity areas.
This detailed mapping requires a statewide coordinated effort and can
only succeed if all the seismic surveys and wells available in these areas are
used. The proposed MPM geomechanical simulation and its resulting ISP
ISP
High
Low ISP regions Prague fault: high ISP
Large injection volumes Many earthquakes
Limited earthquakes Limited injec
Low
Figure 8.8 Zoom of Figure 8.7 that includes to the east the Prague fault, Oklahoma City
and very large injection volumes in areas with no earthquakes. Notice that most of the
earthquakes occur in high ISP areas and large volumes are injected north and west of the
Prague fault in low ISP areas without causing any earthquake in their vicinity. Size of blue
circles represent injected volumes. Size of red stars represent magnitude of earthquakes.
0 60
km
Figure 8.9 Zoom of Figure 8.8.3 to show seismicity NE of interpreted fault in Major
County.
could be easily and quickly updated if a more complete fault interpretation

becomes available in the future.
The same process could be applied to Alberta where the faults shown in
Figure 8.4d and their discrete representation shown in Figure 8.10a, could
be used as input in the geomechanical workflow which will calculate the
ISP shown in Figure 8.10b. The areas with red colors in Figure 8.10b could
potentially have a higher probability of causing earthquakes as a result of
water disposal or even large volume hydraulic fracturing.
The implication of these results is that a predictive tool, which can be
deployed over large areas, can be constrained in such a fashion that it is an
understandable representation of the spatial variation of the potential to
induce seismic events for regulators, E&P companies, insurance companies,
and the public. Such a tool can be envisioned by considering the following
maps for large areas in Oklahoma (Figure 8.11) and Alberta (Figure 8.12).
High
Low
(a) (b)
Figure 8.10 (a) Equivalent Fracture Model derived from the major fault map of Figure
8.2d (b) Resulting geomechanical output showing the complex distribution of the areas of
high potential induced seismicity.
ISP
High
Low
Figure 8.11 Derived ISP maps plotted along with injection volumes and earthquake data
over a large area of the state of Oklahoma.
High
Low
Figure 8.12 Derived ISP maps plotted in an area of Alberta where multiple large
magnitude earthquakes attributed to hydraulic fracturing were recorded in Fox Creek
and Fort St. John in 2015.
8.6 The Benefits of a Large Scale Predictive Model

and Future Research
This study demonstrates the potential of MPM and its use in a 3G work-
flow to develop tools which may be valuable to regulators, insurance
companies, as well as oil and gas companies in addressing induced seis-
micity risks. A workflow which incorporates geological and geophysical
datasets, with powerful mechanical modeling techniques, was applied to
develop a tool for understanding induced seismicity potential in a region.
The benefits of the approach outlined in this study are the use of a phys-
ically accurate modeling technique, which leverages large G&G datasets
and produces results which may be combined into an induced seismicity
potential proxy. This proxy can be plotted to create a tool which is inter-
pretable by the general public, geoscientists, and regulators alike. Various
ISP proxies are explored, which can be simulated on complex regional fault
systems constrained by geologic and geophysical datasets. These ISP maps
could be quickly recomputed as the interpreted fault system is updated

by using both available or future seismic surveys and wellbores, as well as
other sources of information.
This model will be improved by incorporating more realistic fault
geometries and imposing variable material properties and pore pressure
which reflect strati-graphic and depth ranges. Additionally, imposing
injection wells as input to the model will help understand the magnitude
of induced events. Finally, allowing faults in the model to propagate and
intersect with other discontinuities, a feature available in the MPM geo-
mechanical simulator, may help understand the time dependency of these
events. In addition to enhancing the model, different approaches to val-
idating the model may be explored, especially examination of principle
stress axes rotation throughout the study area and comparison to robust,
publicly available data.
8.7 Conflict of Interest

The authors wish to confirm that there are no known conflicts of interest
associated with this publication and there has been no financial support for
this work that could have influenced its outcome. The entire study reported
in this publication is the outcome of FracGeo internal R&D efforts. The
results of the study can be obtained free of charge by the public, any private
or public company, state or federal government agency, or non-profit orga-
nization by contacting directly the authors.
Acknowledgements
The authors thank Arman Khodabakhshnejad for his help with the MPM
simulation of the one fault case shown in Figure 8.6.
References
1. I. Thompson and R.G. Boutilier, Modeling and measuring the social license
to operate: Fruits of a dialogue between theory and practice, http://social-
icense.com/publications/Modelling%20and%20Measuring%20the%20
SLO.pdf (2011).
2. T.H. Darrah, A. Vengosh, R.B. Jackson, N.R. Warner, and R.J. Poreda,
Noble Gases Identify the Mechanisms of Fugitive Gas Contamination
in Drinking-water Wells Overlying the Marcellus and Barnett Shales.

Proceedings of the National Academy of Sciences Proc Natl Acad Sci USA
111(39), 14076–4081 (2014).
3. Y. Aimene and A. Ouenes, Geomechanical modeling of hydraulic fractures
interacting with natural fractures — Validation with microseismic and tracer
data from the Marcellus and Eagle Ford. Interpretation 3(3), SU71–SU88
(2015). DOI: 10.1190/ INT-2014-0274.1
4. A. Ouenes, N. Umholtz, and Y. Aimene, Using geomechanical modeling to
quantify the impact of natural fractures on well performance and microseis-
micity: Application to the Wolfcamp, Permian Basin, Reagan County, Texas.
Interpretation 4(2), SE1–SE15, (April 2016). DOI: 10.1190/INT-2015-0134.1
5. A. Ouenes, Y. Kiche, L. Ouhib, R. Smaoui, M. Paryani, S. Poludasu, A.
Bachir, and D. Balogh, Efficient Development of Unconventional Reservoirs
Using 3G Workflows – Breaking the Silos to Achieve True Integration with
Multidisciplinary Software. First Break 34 (May 2016).
6. C. Scholz, The Mechanics of Earthquakes and Faulting, Cambridge University
Press, New York (2002). DOI: 10.1017/CBO9780511818516
7. G.C.P. King and M. Cocco, Fault interaction by elastic stress changes: new
clues from earthquake sequences. Advances in Geophysics 44, 1–38 (2001).
8. Y.E. Aimene and J.A. Nairn, Modeling Multiple Hydraulic Fractures Interacting
with Natural Fractures Using the Material Point Method. Society of Petroleum
Engineers, Canada (2014). DOI: 10.2118/167801-MSD
9. Sulsky, Z. Chen, and H. L. Schreyer, A particle method for history-dependent
materials. Computer Methods in Applied Mechanics and Engineering 118,
179–196 (1994).
10. J.A. Nairn, Material Point Method calculations with explicit cracks. Computer
Modeling in Engineering and Science 4, 649–663 (2003).
11. N. Umholtz and A. Ouenes, Optimal Fracing Near Faults - Quantifying the
Interaction Between Natural and Hydraulic Fractures Using Geomechanical
Modeling, Society of Petroleum Engineers, Canada (2015). DOI:
10.2118/176932-MS
12. A. Holland, Peliminary Fault Map of Oklahoma, http://www.ou.edu/con-
tent/ogs/data/fault.html (2015).
13. USGS Earthquake Archives, ANSS ComCat, http://earthquake.usgs.gov/
earthquakes/search
14. Oklahoma Corporation Commission (OCC). Oil and Gas Data Files,
Accessed January 5, 2015. http://www.occeweb.com/og/ogdatafles2.htm
9
Correlating Pressure with Microseismic to
Understand Fluid-Reservoir Interactions
During Hydraulic Fracturing
Debotyam Maity
*
Gas Technology Institute, Des Plaines, IL
Abstract
Hydraulic fracturing of shales and other tight formations has gained tremendous
prominence over the past decade or so, and is set to grow in importance with
increasing global energy demands. In recent years, new methods, which develop
upon older techniques, have been proposed for the interpretation of multi-stage
hydraulic fracturing data. A major issue has been validating these methods for
wider applicability within the industry, considering some limitations that these
techniques pose as a result of certain inherent assumptions.
In this paper we take up a recently proposed diagnostic technique and analyze
completion data from multiple shale gas wells. We correlate the results with avail-
able microseismic monitoring data and validate the interpretations made using
this approach. Using actual stage-wise production contributions information from
production logs, we also highlight the limitations of the current approach when it
comes to predicting overall stage-wise completion quality.
Keywords: Hydraulic fracturing, pressure, microseismic, completion, diagnostics
9.1 Introduction
Analyzing pressure data as a means to understand treatment behavior has
been a part and parcel of the oil and gas industry for many decades now.
Pressure data can help identify fracturing behavior, proppant transport
issues, screen out situations, limited entry calculations, etc., to name just a
Email: Debotyam.Maity@gastechnology.org

181
few of its uses. However, with the increasing use of multi-stage hydraulic
fracturing treatments, many of these traditional techniques no longer hold
validity, and newer techniques are considered desirable for improved treat-
ment diagnostics. In this paper we consider a recently proposed diagnostic
technique, which builds upon a very popular real time fracturing analysis
technique, and demonstrate the utility as well as pitfalls associated with
using such techniques.
9.2 Method
9.2.1 Pressure Data Analysis
Simple fracture models were proposed by Perkins and Kern [1] who
suggested that the fracturing pressure at the wellbore should be a power
function of treatment time (Eq. 9.1) with a large value of the exponent
indicating better fluid containment within the developing fracture.
p(t) te (9.1)
This and other work by Nordgren [2] formed the basis for the real time
completion analysis technique as proposed by Nolte and Smith [3]. They
concluded that fracture propagation may follow one of four predefined
modes based on the slope of net pressure plotted against time as shown
in Figure 9.1. The basic power law equation guiding the propagation of
hydraulic fractures is given as follows:
log (p−pclosure) e × log(t) (9.2)
Recent proposals include modifications to the original Nolte-Smith

approach to take into account the intermittent nature of fracture propa-
gation and the well-known phenomenon of natural and hydraulic fracture
interaction during treatment in shale plays [4, 5]. Developing upon the
power function defined by Eq. 9.1, and considering the reference time (ti)
for fracture growth initiation, we have:
p−pi = C (t−ti) (9.3)
p e
t ti eC t ti (9.4)
t
Correlating Pressure with Microseismic 183
p
t ti e p pi (9.5)
t
Based on Nolte’s generalization [6] for the bounds applicable on Eq.

9.1 for non-Newtonian fluids, we expect the exponent to range close to
0.25. If the fracture is dilating, the exponent will have a value close to 1. If
the growing hydraulically created fracture interacts with natural fracture
swarms, we expect rapid loss of fracturing fluid into the natural fracture
system with the pressure behavior to be similar to fast leak off scenario
(Figure 9.1, Mode IV) provided the pressure is stable or increasing. At the
same time, we understand that as soon as pressure drops due to such fluid
loss, the fissures tend to close as it drops below the stress holding them
open. However upon closure, the pressure starts to rise again and there
is a corresponding opening of closed fissures. This opening-closing cycle
should produce rapid fluctuations in the exponent as an indicator of natu-
ral fracture interaction during treatment as will be shown later.
This modified Nolte-Smith approach with varying reference time has
been detailed by Pirayesh and others [4]. We use the same approach to esti-
mate time variant exponent ‘e’ during treatment and use the results in our
analysis. The calculations involve evaluating the exponent and constant at
each point of time (during treatment) and then integrating the results from
the reference to the current time.
dp / dt
et (9.6)
pn pi / tn ti
pn pi
ct et (9.7)
tn ti
tn
1
E et dt (9.8)
tn ti
ti
tn
1
C ct dt (9.9)
tn ti
ti
III
Net pressure
II
I
IV
Time
Figure 9.1 Nolte-Smith Interpretation Guide (log Pnet vs. log T). Mode I indicates
constrained height with unrestricted extensional growth; Mode II indicates Stable growth;
Mode III indicates restricted extension or screenout conditions; Mode IV indicates
unstable fracture height growth.
Next we can estimate BHP by using the E and C values and identify
the error from the actual BHP value available either through downhole
pressure instrumentation or surface pressure data corrected for downhole
conditions.
BHPest = pn + C (tn−ti)E (9.10)
ε = |BHJPest – BHPcalc.| (9.11)
If the calculated error mismatch, ε, is higher than a predefined threshold,

the initial reference point (ti) is shifted to the current evaluation time. This
provides us with a varying value for exponent e for the entire treatment,
and this can then be used for our diagnostic interpretations. As an exam-
ple, Figure 9.2 shows pressure match obtained for a reference stage and
the error in evaluation. We note that for this analysis, we used a threshold
limit of 5 psi in order to obtain a close match between the estimated BHP
and the calculated BHP. We note that it is preferable that we use downhole
pressure data instead of calculating BHP from surface measurements. The
potential pitfalls of using BHP calculated from surface data compared to
downhole measurements will be looked into later on.
After validation of adequate accuracy in the net pressure prediction, the
exponent variability with completion time can be interpreted as required.
Figure 9.3 shows the variability of exponent ‘e’ for the same treatment data
subset shown earlier, and we can clearly identify those sections of the treat-
ment where we either had Mode I, Mode III or Mode IV fracturing based
x 10 4
1.202 Calc. BHP
Modeled BHP 15
1.2
Δ Pressure (psi)
Pressure (psi)
1.198 10
1.196
5
1.194
0
64 66 68 62 64 66 68
(a) Time (minutes) (b) Time (minutes)
Figure 9.2 (a) Comparison of calculated BHP and estimated BHP (Eq. 9.10) and (b)
observed ΔP mismatch for a short window extracted from completion data for a sample
hydraulic fracturing stage.
0.5
Exponent, e
-0.5
-1
62 63 64 65 66 67 68 69
Time (minutes)
Figure 9.3 Modeled modified Nolte-Smith exponent ‘e’. The region highlighted from e =
0.75 to 1 indicates restricted flow and fracture dilation, the region highlighted from e = 0.1
to 0.3 indicates unrestricted fracture extension and from e = -1 to -0.1 indicates natural
fracture interactions or rapid height growth.
on the distribution of the exponent values. We observe that there is signifi-

cant interaction with natural fractures and we also observe that towards the
end of the period in question, extensional growth of the fracture network
occurs with fluid injection. The exact range used for the zones indicating
the three modes are: -1.00 to -0.10 for mode IV; 0.10 to 0.30 for Mode I and
0.75 to 1.00 for Mode III.
While real time diagnostic application has already been studied for multi
stage shale gas completions by Soliman [5], we want to look at the pre-
dicted fracturing behavior and identify possible ways to characterize the
completion effectiveness using the modeled parameters. This is achieved
by defining certain diagnostic parameters computed for each stage. For the
three modes (I, III and IV) identified, we calculate two separate diagnostic
parameters as defined below:
tend
DP1 S tts | e |Δt (9.12)

t 1
tend
DP2 S tts Δt (9.13)

t 1
Here δ represents the Kronecker delta, tend represents the end of evalu-
ation period, and tS is the set with the time stamps that correspond with
the value of exponent ‘e’ falling within the three modes in question. Finally,
Δt is the time step in seconds. The reason for using the above definitions
for the diagnostic parameters DP1-S and DP2-S was that the extent that the
exponent ‘e’ stays within each modal region should provide an indication
of how much extensional growth, dilatational growth as well as interaction
with natural fractures occurs during completion.
9.2.2 Microseismic Data Analysis

Microseismic events occur due to stress changes and pressure variations
as a result of fluid being injected into the formation during hydraulic frac-
turing [7]. These events occur due to small movements over very small
rock area in either shear, tensile or complex mode failure [8]. In the 1940’s,
Gutenberg and Richter identified a relationship between the frequency of
earthquakes and their magnitudes, and this relation has been observed to
be universal and hold for local and micro level seismicity as well [9]. The
relation is defined as follows:
log10 N = 1−bM (9.14)
Where M is the magnitude, N is the number of events within a particu-

lar magnitude/ time or spatial bin. Constants ‘a’ and ‘b’ are derived empir-
ically from observed seismicity distribution. Figure 9.4 shows a sample
b-value calculation based on 30 year data available from the Global CMT
(Centroid Moment Tensor) catalog for the state of California (United
States) highlighting the linear nature of the distribution outside the roll-off
N = No. earthquakes M
102 b=1
100
6 7 8 9
Magnitude (M)
Figure 9.4 Gutenberg-Richter magnitude frequency relationship for earthquake data

from California (CMT catalog).
regions and the expected slope (b-value) of 1 for tectonic earthquakes

[10]. While evaluating b-value, care should be taken to ft the data using
maximum likelihood method [11]. Datasets with very few earthquakes
should not be considered for analysis. Also, error in magnitude can cause
significant uncertainty in the slope estimation as well as identifying the MC
(Magnitude of completeness). In our analysis, we assume that the errors in
magnitude within our catalogs are negligible.
For local seismicity scenarios, variations in b-value can be attributed
to reservoir heterogeneity, thermal gradients across the perturbed zone,
applied stress through injection as well as other factors [12]. For induced
fracturing, this relationship was originally observed by Scholz [13] and has
been further validated in recent years [14, 15]. Finally, recent studies have
also indicated some correlation between observed b-value distribution and
the local stress regime guiding rock failure [16, 17]. In general, when it
comes to local or microseismicity, b values less than 1 are rare and repre-
sent compressive failure modes; b values close to one indicate strike-slip
regime similar to observations during failure of faults (from earthquake
seismology) and b values higher than 1 are indicative of extensional failure
modes.
9.3 Data
We apply this technique on two separate wells from the Appalachian Basin
(Marcellus shale gas play). Both the wells were completed as essentially dry
gas wells with reasonable water cuts. The choice was based on the fact that
both these wells involved microseismic monitoring during completion,
which was a necessary element of our analysis. Well # 1, located in north-

ern Pennsylvania, involved the pumping of ~ 300,000 lbs. of proppant for
each stage with a total of 18 stages. Well # 2 located in western Pennsylvania
again involved the pumping of ~ 300,000 lbs. of proppant for each individ-
ual stage with a total of 13 stages. Both completions were monitored by
downhole sliding microseismic monitoring arrays placed in nearby hori-
zontal observation wells. For b-value analysis, we used microseismic event
catalogs generated by using the standard P (compressional) and S (shear)
wave travel time inversion approach. Production logs were run post treat-
ment for both the wells, and these are also used in our study. For Well # 1,
we also have data from OBMI log, and we also use the interpreted stage-
wise fracture density from this log for our analysis.
9.4 Results
As stated earlier, we apply both these techniques, namely modified Nolte-
Smith approach, to fracture diagnostics as well as stage wise b-value
mapping on the data from the two wells under study. For both wells, we
generate the diagnostic parameters (DP1-S and DP2-S) for each of the frac-
turing modes under study defined by subscript S (Mode I, III and IV).
Based on the identified parameter values, we cross-correlate the same with
available production log data (fractional gas flow) from post completion
production logging runs. Figure 9.5 shows how the modeled diagnostic
parameter results for Well # 1 correlate with observed stage-wise produc-
tivity from production logs. We notice weak positive correlation for both
mode I and mode IV results but no correlation with mode III results.
Figure 9.6 shows the same parameters evaluated for Well # 2 under
study and their correlation with stage-wise productivity.
Once again, we observe weak positive correlation for both the mode
I and mode IV parameters and no correlation whatsoever with mode III
results. Since observations from both sets of analysis seem to validate each
other, we can argue that this observation is non-unique and should hold
for more completions. We do note that the correlations are weakly posi-
tive and may not signify much. However, in essence, these observations
indicate that the productivity from any hydraulically fractured stage shows
slight correlation with the extent of extensional fracture growth taking
place during the treatment as well as any interaction with natural frac-
tures observed during treatment. Moreover, we can clearly state that the
degree of dilatational fracture growth has no impact on the productivity
of the completed zone. These observations are also intuitive as we would
140 600
120 500
100 400
80
DP 2-l
300
DP 1-l
60
200
40
100
20 y = 353.25x + 73.603 y = 1502.6x + 286.31
R2 = 0.3066 R2 = 0.3206
0 0
0 0.05 0.1 0.15 0 0.05 0.1 0.15
(a) Fractional gas (b) Fractional gas
1600 1800
1400 1600
1200 1400
1200
1000
DP 1-lII
DP 2-lII
1000
800
800
600
600
400 400
200 y = 397.23x + 905.92 200 y = 478.99x + 989.48
R2= 0.0032 R2 = 0.0041
0 0
0 0.05 0.1 0.15 0 0.05 0.1 0.15
(c) Fractional gas (d) Fractional gas
0 1400
0 0.05 0.1 0.15
y = −1178.8x − 318.41 1200
−100
R2 = 0.2997
1000
−200
DP 2-lV
800
DP1-lV
−300 600
−400 400
200 y = 2384.8x + 772.05

−500 R2 = 0.3076
0
0 0.05 0.1 0.15
−600
(e) Fractional gas (f) Fractional gas
Figure 9.5 Production data (fractional gas flow) compared with extensional diagnostic
parameters (a) (DP1-I), (b) DP2-I, dilatational diagnostic parameters (c) DP1-III, (d) DP2-III,
natural fracture interaction diagnostic parameters (e) DP1-IV and (f) DP2-IV for Well # 1.
180 700
160 600
140
500
120
400
DP1-l
DP2-l
100
80 300
60
200
40
y = 2.9381x + 81.959 100 y = 11.798x + 324.91
20
R2 = 0.4891 R2 = 0.4554
0 0
0 10 20 30 0 10 20 30
(a) Fractional gas (b) Fractional gas
1800 2000
1600 1800
1400 1600
1200 1400
1200
DP2-lII
DP2-lII
1000
1000
800
800
600
600
400 400
200 y = 5.4387x + 1058.1 y = 3.4788x + 1089.9
200
R2 = 0.0122 R2 = 0.0041
0 0
0 10 20 30 0 10 20 30
(c) Fractional gas Fractional gas
(d)
0 1000
0 10 20 30 900
−50
y = −7.1587x −225.18 800
−100 R2 = 0.3822
700
−150
600
DP2-lV
−200 500
DP1-lV
−250 400
300
−300
200
−350 y = 13.746x + 572.19
100
R2 = 0.4271
−400 0
0 10 20 30
−450
Fractional gas Fractional gas
(e) (f)
Figure 9.6 Production data (fractional gas flow) compared with extensional diagnostic
not expect fractures ballooning due to fluid fill-up to have any impact on
productivity as any far field fracturing due to stress perturbation may not
be contributing due to them being spatially isolated. Extensional growth
and growth into natural fractures, on the other hand, should lead to more
productivity due to higher fractured area through the connected hydrauli-
cally created as well as natural fractures.
Next we look at the b-value analysis results and how they correlate with
the modeled parameters shared above. Careful selection was made from
the microseismic event catalogs to make sure that the analysis was valid,
including removing outliers. Those stages with very low event count were
not considered in this analysis (stages 1 through 5 for well # 1). Finally,
fitting to identify slope was done using maximum likelihood technique as
mentioned earlier. We understand that higher b values (and consequently
higher fractal dimensions) are indicative of more complex fractured net-
work or values higher than 1, indicating extensional fracture growth [18].
In our analysis, for all stages studied, we found b-values close to or higher
than 1. Any b-value close to or higher than 2 could be a result of microseis-
mic data quality or could indicate fluid-rich completions. Also, as observed
from Figures 9.7 and 9.8, b-value shows a much stronger correlation with
extensional growth parameters compared to natural fracture interaction
related parameters. Finally, b-value shows no correlation with dilatational
fracture swelling related parameters.
This is again expected as extensional growth through fracture propa-
gation should create extensive three dimensional microseismicity and so
should interaction with natural fractures. However, we note that with fluid
flling up dilating fractures, seismicity will be limited in size (small shear
tip or far field failures). Next, we look at correlation with OBMI log in
Figure 9.9.
As observed from Figure 9.9, we can clearly see no apparent cor-
relation between the derived fracture density from OBMI logs and the
modeled diagnostic parameters. This is expected since the OBMI logs
provide a snapshot of wellbore or near wellbore fracturing and cannot
estimate the presence or absence of actual fractures in the formation.
Eyeballing the modeling results using the pressure data can provide
indicators to make judgement calls by identifying stages with significant
interactions with natural fractures during treatment. Figure 9.10 shows
sample stages from Well # 1 showing the modeled exponent. We validate
the observations by comparing the corresponding b-values as shown in
Figure 9.11.
As observed from results for the two stages shared in Figure 9.10, the
stage corresponding to Figure 9.10a (stage ‘A’) shows a relatively lower
3.5 3.5
3 3
2.5 2.5
b-value
b-value
2 2
1.5 1.5
1 1
0.5 y = 0.0116x + 0.9782 0.5 y = 0.0028x + 1.0139

R2 = 0.5601 R2 = 0.5757
0 0
0 50 100 150 0 200 400 600
DP1-l DP2-l
(a) (b)
3.5 3.5
3 3
2.5 2.5
b-value
b-value
2 2
1.5 1.5
1 1
0.5 y = −0.0002x + 2.2792 0.5 y = −0.0002x + 2.2418

R2 = 0.0248 R2 = 0.0162
0 0
0 500 1000 1500 2000 0 500 1000 1500 2000
DP1-lII DP2-lII
(c) (d)
3.5 3.5
3 3
2.5 2.5
b-value
b-value
2 2
1.5 1.5
1 1
y = −0.0026x + 1.1803 0.5 0.5 y = −0.0014x + 1.2298

R2 = 0.2976 R2 = 0.2775
0 0
−600 −400 −200 0 0 500 1000
DP1-lV (f) DP2-lV
(e)
Figure 9.7 Microseismic derived b-value compared with extensional diagnostic

3 3
2.5 2.5
2 2
b-value
b-value
1.5 1.5
1 1
0.5 y = 0.0067x + 1.348 0.5 y = 0.0016x + 1.3941

R2 = 0.5404 R2 = 0.5164
0 0
0 50 100 150 200 0 200 400 600 800
DP1-l DP2-l
(a) (b)
3 3
2.5 2.5
2 2
b-value
b-value
1.5 1.5
1 1
0.5 y = 0.0002x + 1.8431 0.5 y = 0.0002x + 1.8588

R2 = 0.0568 R2 = 0.0574
0 0
0 500 1000 1500 2000 0 500 1000 1500 2000
DP1-lII DP2-lII
(c) (d)
3 1000
900
2.5 800
700
2
600
b-value
b-value
1.5 500
400
1 300
200
y = −0.0022x + 1.421 0.5 y = 364.95x −68.817
100
R2 = 0.4581 R2 = 0.4384
0 0
−600 −400 −200 0 0 1 2 3
DP1-lV DP2-lV
(e) (f)
Figure 9.8 Microseismic derived b-value compared with extensional diagnostic

140 y = −0.2374x + 62.421 140 y = −0.0631x + 63.648

R2 = 0.0172 R2 = 0.021
120 120
100 100
80 80
OBMI
OBMI
60 60
40 40
20 20
0 0
0 50 100 150 0 200 400 600
(a) DP1-l (b) DP2-l
140 y = 0.0271x + 14.913 140 y = 0.0249x + 14.756
R2 = 0.0268 R2 = 0.0262
120 120
100 100
80 80
OBMI
OBMI
60 60
40 40
20 20
0 0
0 500 1000 1500 2000 0 500 1000 1500 2000
DP1-lII (d) DP2-lII
(c)
y = 0.0865x + 70.236 140 140 y = −0.0443x + 66.422
R2 = 0.025 R2 = 0.0199
120 120
100 100
80 80
OBMI
OBMI
60 60
40 40
20 20
0 0
−600 −400 −200 0 0 500 1000
(e) DP1-lV (f) DP2-lV
Figure 9.9 OBMI log derived fracture density compared with extensional diagnostic
0
e
-1
20 25 30 35 40 45 50 55 60 65
(a) Time (minutes)
1
0
e
-1
25 30 35 40 45 50 55 60 65
(b) Time (minutes)
Figure 9.10 Modeled exponent ‘e’ for two stage treatments during proppant injection
phase. Subplot (a) shows lower productivity for stage ‘A’ compared to subplot (b) showing
productivity for stage ‘B’ from production log data.
103
102
Cumulative number
Cumulative number
102
101
101
b = 1.1 +/- 0.09 b = 1.49 +/- 0.1

a = 0.8218 a = 0.52611
100 100
-3 -2.5 -2 -1.5 -1 -0.5 0 -2.5 -2 -1.5 -1 -0.5
(a) Magnitude (b) Magnitude
Figure 9.11 Results from b-value analysis for (a) stage ‘A’ and (b) stage ‘B’ using
Maximum Likelihood approach.
degree of interactions with natural fractures. Moreover significant sec-

tions of the treatment show dilatational behavior compared to stage cor-
responding to Figure 9.10b (stage ‘B’). These observations are validated
by actual observations from production log data with stage ‘A’ showing
1.135% gas flow contribution compared to 14.139% for stage ‘B’ from
the post completion production log. Furthermore, the corresponding
b-value evaluation shows significant differences with stage ‘B’ showing
a higher value (1.5), indicating higher fracture complexity compared to
stage ‘A’ (1.0).
9.4.1 Pitfalls in Analysis

We need to carefully consider the limitations and assumptions made in
our analysis and highlight adequate caution required when conducting
similar analyses. Before looking into the models themselves, the most
important constraint with any diagnostic methodology is the actual data
quality collected from the field and used in the analysis. With pressure data,
downhole pressure measurement is extremely rare in long multi-stage hor-
izontal hydraulic fracturing programs due to cost issues. Downhole data is
necessary, as an assumption of net pressure being equal to the calculated
BHP can be highly flawed due to uncertainties in models used for these
calculations, including corrections for air entrainment, frictional losses
and impact of proppant loads. With microseismic data, errors in inver-
sion, limitations of array design, deployment issues, etc. can significantly
alter the overall quality of the microseismic catalogs and thereby under-
mine any interpretations. When it comes to the actual physical fracture
model used, we need to highlight that it is by very nature highly simplistic.
Moreover, the variability in exponent, though very useful, can be inter-
preted in multiple ways. As an example, negative exponent values could
mean natural fractures as suggested in this work. But at the same time, they
could also mean uncontrolled rapid height growth into lower closure stress
zones (particularly with decreasing pressure) or interaction with local
faults. Similarly, with b-value analysis, sometimes the data artifacts require
manual mapping of the slopes resulting in significant non-uniqueness in
the selected values. Therefore we believe that for proper utilization of this
technique, more robust ways of analyzing data, as well as analysis of other
information in addition to those shared in this work, may be necessary. All
of these considerations will influence our future work with new wells.
9.5 Conclusions
Novel completion diagnostic techniques such as those applied in this study
provide valuable tools, which can be very useful in helping people under-
stand the behavior of long lateral multi-stage hydraulically fractured wells.
Judicious selection of data, analysis methodology and a careful consider-
ation of potential pitfalls are also necessary in order to add value to any
such diagnostic workfow. In this study, we have demonstrated two ways
of identifying potential fracture growth mechanisms available today and
have tried to correlate the two to highlight a reasonable match between
the results as per our observations. However, a more careful analysis of
methods as well as better models are deemed necessary before any such
technique can find widespread use within the fracking industry. This is
because of a high degree of uncertainty and the non-uniqueness possible
in the interpretations.
9.6 Acknowledgements
This work was supported by RPSEA project number 11122-20. I also
acknowledge WPX Energy for providing access to active hydraulically frac-
tured wells of opportunity in the Marcellus shale play, and Schlumberger
for acquiring the production log data used in this study. I would also like to
acknowledge valuable contributions and constructive suggestions by Iraj
Salehi and Jordan Ciezobka from the Gas Technology Institute.
References
1. T.K. Perkins and L.R. Kern, Widths of hydraulic fractures. J. Petr. Tech. 13,
937–949 (1961).
2. R.P. Nordgren, Propagation of a vertical hydraulic fracture. Soc. Petrol. Eng.
J. 253, 306– 314 (1972).
3. K.G. Nolte and M. Smith, Interpretation of fracturing pressures. J. Petr. Tech.
33, 1767– 1775 (1981).
4. E. Pirayesh, M.Y. Soliman, and M. Rafiee. Make decision on the fly: A new
method to interpret pressure-time data during fracturing – application to
frac pack. SPE Annual Technical Conference and Exhibition (2013).
5. M.Y. Soliman, M. Wigwe, A. Alzahabi, and E. Pirayesh, Analysis of fractur-
ing Pressure data in heterogeneous shale formations. Hydraulic Fracturing J.
1(2), 8–13 (2014).
6. K.G. Nolte, Determination of fracture parameters from fracturing pressure
decline, SPE Annual Technical Conference and Exhibition (1979).
7. N.R. Warpinski, S.L. Wolhart and C.A. Wright, Analysis and prediction of
microseismicity induced by hydraulic fracturing. SPE J. 9(1), 24–33 (2004).
8. M.F. Kanninen and C.H. Popelar: Advanced Fracture Mechanics, pp. 563,
Oxford University Press, New York (1985).
9. B. Gutenberg and C.F. Richter: Seismicity of the Earth and Associated
Phenomena, pp. 16–25, Princeton University Press, Princeton, New Jersey
(1949).
10. A.M. Dziewonski, T.A. Chou, and J.H. Woodhouse, Determination of earth-
quake source parameters from waveform data for studies of global and
regional seismicity. J. Geophys. Res. 86, 2825–2852 (1981).
11. K. Aki, Maximum likelihood estimate of b in the formula log (N) = a - bM

and its confidence limits. Bull. Earthq. Res. Inst. Tokyo Univ. 43, 237–239
(1965).
12. J. Farrell, S. Husen, and R.B. Smith, Earthquake swarm and b-value charac-
terization of the Yellowstone volcano tectonic system. J. Volcanol. Geoth. Res.
188, 260–276 (2009).
13. C.H. Scholz, The frequency-magnitude relation of microfracturing in rock
and its relation to earthquakes. Bull. Seismol. Soc. Am. 58, 399–415 (1968).
14. S. Wessels, M. Kratz, and A. De La Pena, Identifying fault identification
during hydraulic simulation in the Barnett shale: Source mechanism, B val-
ues, and energy release analyses of microseismicity. 81st SEG Annual Meeting
Expanded Abstracts 30, 1463–1467 (2011).
15. S. Sil, B. Bankhead, C. Zhou, and A. Sena, Analysis of B value from Barnett
shale microseismic data, 74th EAGE Conference & Exhibition (2012).
16. M. Grob and M. van der Baan, Inferring in situ stress changes by statisti-
cal analysis of microseismic event characteristics. The Leading Edge 30(11),
1296–1302 (2011).
17. M. Grob and M. Van der Baan, Statistical analysis of microseismic event
characteristics helps monitor in-situ stress changes, 21st Canadian Rock
Mechanics Symposium (2012).
18. T. Urbancic, A. Baig, and S. Bowman, Utilizing B-values and fractal
dimensions for characterizing fracture complexity, CSPG/ CSEG/ CWLS
GeoConvention (2010).
10
Multigrid Fracture Stimulated Reservoir
Volume Mapping Coupled with a Novel
Mathematical Optimization Approach to
Shale Reservoir Well and Fracture Design
Ahmed Alzahabi1*, Noah Berlow2, M.Y. Soliman1 and Ghazi AlQahtani3
1
Petroleum Engineering Department, UH Energy Research Park (ERP),
University of Houston, Houston
2
Department of Electrical and Computer Engineering, Texas Tech University,
Lubbock, USA
3
Special Simulation Studies Unit Saudi Aramco
Abstract
This paper introduces a grid-based fracture-stimulated reservoir volume (SRV)
concept. SRV is defined as a volume occupied by fluid in a fracture, whether cre-
ated or caused by intersection with natural fractures.
Fracturing of the optimum zones is believed to contribute to higher hydrocar-
bon production from shale and tight formations. This requires choosing placement
of fractures along the designed path of horizontal wells to maximize expected SRV.
Additionally, a new linear programming-based approach to mathematically
optimize the placement of SRV in shale reservoirs is presented. The approach may
be useful in pad drilling and fracturing as well as development of applications for
use with shale formations.
This work aims to globally optimize the placement of surface well pads, the
location and number of wells attached to the pads, and the location of the frac-
tures throughout the wells. This optimum placement will also take into account
numerous practical constraints, including the length of wells, the number of wells
associated with a pad, numerous overlap constraints inherent in unconventional
gas and oil well development, the spacing between wells and fractures, etc.
*Corresponding author: amalzahabi@uh.edu

199
One approach to this challenge to optimize is based on maximizing the fractur-

ability index (FI) values assigned to the cells of the model values explored by the
final network, and will be constrained by the previously mentioned considerations
as well as a global maximum number of wells and a maximum development bud-
get. In addition, the mathematical framework allows for easy extensibility to other
constraints, and can be customized based on these constraints.
Keywords: Stimulated reservoir volume, shale reservoir, optimization, well and

fracture placement design, fracture network, well pad
10.1 Introduction
Simulated reservoir volume (SRV) has a long history of use in defining
the effect of fracturing in shales. Substantial evidence from sonic logs and
production data from shale wells shows that certain segments of the wells
make up 70% of the total production of wells. This paper presents a con-
cept for identifying SRV in shale rock. Creating hydraulic fractures leads to
fracture network growth. Fracture growth interaction with existing natural
fractures causes complexity. Complexity is a resultant network of induced
and existing fractures. SRV is used to account for this resultant complexity.
These complex networks have a substantial impact on well performance in
shale and tight rocks. The shape of SRV can be predicted from stimulated
and shear propped fractures, while the volume can be correlated with frac-
ture network length. Britt et al. [1] & Cipollaet et al. [2] discussed geome-
chanics of a shale prospective and fracture complexity.
The size of the SRV is correlated to treatment volume based on micro-
seismic measurement. Figure 10.1 shows the relationship between treat-
ment volume and fracture network length for five vertical Barnett shale
wells, modified after Fisher et al. [3, 4].
Mayerhofer et al. (2008) introduced the SRV concept as a 3D size of
created fracture network, and defined SRV as a correlation parameter for
well performance. Mayerhofer et al. [5] linked SRV with well performance
of shale reservoirs. A direct relationship is demonstrable between fracture
network length and SRV, as shown in Figure 10.1 for Barnett shale wells
(modified after Fisher et al. [3]).
Anderson et al. [6] defined SRV linked to the horizontal well by stimu-
lated reservoir width, areal extent and fracture half-length. Zhou et al. [7]
introduced a method for identifying anisotropic regions in unconventional
hydrocarbon reservoirs. Anisotropy can be indicative of sweet spot zones
for fracturing and for drilling a productive well. Seismic amplitude data
from receivers along two orthogonal lines radiating from a seismic source
Fracture Stimulated Reservoir Volume Mapping 201
30000
Fracture network length = –4E–05 Vol.2 + 2.108 Vol. + + 1920.2
R2 = 0.9839
25000
Fracture network length, ft.
20000
15000
10000
5000
0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
Fluid volume bbl
Figure 10.1 Fracture network length as a function of fluid volume injected for five
vertical wells of Barnett (modified after Fisher et al. [3]).
is used. Sil et al. [8] introduced a method to calculate fracture parameters

from common well log data. Fracture parameters can indicate sweet spot
zones in unconventional rock. Microseismic mapping is currently used to
map SRV in shale rocks. It is also used as a tool to diagnose the effective-
ness of the created hydraulic fractures, especially in multistage fracturing
in horizontal wells. Zhang et al. [9] introduced the SRV equation as follows:
SRV SRA H f Ai Hp , (10.1)

i 1
where Hf is the fracture height; SRA is the stimulated reservoir area; Hp and
is the reservoir thickness. One limitation is implicit: it seems to be only
valid when hp = hf.
Cheng et al. [10] presented an SRV formula as a function of average
fracture length, average height of fractures, maximum number of fractures,
maximum horizontal stress and minimum horizontal stress as follows:
H
nc
2 h
SRV 4 x hi i 2
(10.2)
1 H
1
h
where
xi = Average length of the fractures in the ith cluster
hi = Average height of the fractures in the ith cluster, m
nc = Number of clusters
σH = Maximum horizontal in situ stress, N/m2
σh = Minimum horizontal in situ stress, N/m2
Additionally, a linear programming-based approach to mathematically

optimize SRV is presented. This newly developed optimization approach
improves the placement of fractures in quantifiably better zones in shale
reservoirs, to guarantee optimality of the reservoir development plan given
the available data and modeling constraints. This approach will be useful in
pad drilling and development of applications applied to shale formations.
This work will lead to the global optimization of the placement of surface
pads, location and design of wells attached to the pads, and location of the
fractures (SRV) throughout the wells. This design will also take into account
other practical design constraints, including length of wells, number of wells
associated with a pad, numerous overlap constraints inherent in unconven-
tional gas and oil well development, etc. The development will be optimized
based on maximization of the FI values explored by the final network, and
will be constrained by the previously mentioned considerations, as well as a
global maximum number of wells and a maximum development budget. In
addition, the mathematical framework allows for easy extensibility to other
constraints, and can be customized based on the problem constraints.
Fracture parameters such as half-length, azimuth, width and height can be
used extensively in the fracture modeling process. Fracture geometry can be
modeled instead through SRV, as an estimated fracture volume can give a bet-
ter description of fracture parameters. SRV is represented by grid like geome-
try that has a value of FI greater than the predefined threshold of 0.5. The SRV
consists of a group of cells as shown in Figure 10.2, and SRV varies from one
stage to the next. The next section details the geometric interpretation of the
SRV representation, that is, a group of grids characterized by high values of FI.
Before entering into discussion of the mathematical definition, it is neces-
sary to set the objective of developing a method to predict SRV location and
number using an input map of FI’s. Such a method, when coupled with math-
ematically developed code, could help in exploitation of the shale resource
with the minimum number of wells and minimum number of fracture stages.
The paper aims at:
• Planning and automating an optimum well path and opti-

mum fracture design in shale and tight formation;
Figure 10.2 Multigrid based SRV.
• Establishing a process of choice of maximum SRV for future

initiation of fractures;
• Finding out the optimum number of fracture stages;
• Optimizing the number of SRV’s.
Previous work [11] identified an index to help prioritize fracture position

and scheduling. Mathematical optimization using Integer Programming
(IP) proved its superior performance in vertical well placement (for details
on its performance, see AlQahtani et al. [12]).
Computational concepts such as dynamic programming and graph the-
ory may be useful in exploration of algorithms applied to a wide range of
oil and gas optimization topics, the most important part of which is the
computational methods used to solve them, so that an optimum placement
can be obtained. Since the problem is a mathematically based method, the
problem definition is first outlined in the next section.
10.2 Problem Definition and Modeling

10.2.1 Geometric Interpretation
10.2.1.1 Fracture Geometry
Hydraulic fracture geometry dimensions may be calculated using analyti-
cal approaches based on net pressure, fluid and rock properties. Another
common approach is microseismic monitoring, which fits a rectangular

box to the microseismic event locations along the horizontal well path.
SRV can be estimated based on the volume of the rectangular or principal
component box, or by summing a series of volumetric boxes (e.g.,
Mayerhofer et al. (2008)). For a realistic approach, SRV is used as a repre-
sentation of fractures connected to the wells. A stimulated approximation
of grids is used to represent the hydraulic fracture.
10.2.2 The Developed Model Flow Chart

In this paper an approach for placing surface well pads and fractures in
shale rock is shown in Figure 10.3. The solution is presented in Figure 10.3
and Figure 10.4 as follows.
Figure 10.5a and b illustrate the two allowable designs for placing frac-
tures and then SRV’s.
10.2.3 Well and Fracture Design Vector Components

1. Number of transverse fracture stages per well (5–50).
2. Number of wells of single pad (5–40).
3. Number of perforation per stage (1–6).
4. Length of horizontal well (6,000–10,000 ft).
5. Half Length of fracture (200–600 ft).
6. Spacing (wells, fractures), (500–1,600 ft).
7. Pay zone thickness, (200–1,000 ft).
8. Reservoir boundary dimensions (Ye, Xe), (rectangular shape).
9. Variable Stimulated Reservoir Volume (VSRV) with differ-
ent variable conductivity.
10. The formation is heterogeneous.
11. The transverse fractures fully penetrate the majority of for-
mation, except 10 ft. from the boundary. Fractures are con-
tained within the formation.
12. Multiple transverse fractures are not identical (varied frac-
ture conductivity, and varied fracture propped characteris-
tics like length and network width).
10.3 Development of a New Mathematical Model

In this section, a description of the mathematical model used to solve
the above-mentioned problem is given, consisting of an introduction of
Direct global solution
Load reservoir parameters

and data
Generate FI values for

reservoir
Collect user designed

optimization parameters
User selects desired pad

locations
Identify viable well pad locations
Preprocess: generate frac SRV for

all wells in well pad reach
Preprocess: generate well pad,

well, frac LP model
Store model
All models built?
Join all well pad models via overlapping

constraints
Solve global model via MLP solver
Save solved global model and export for

simulator
End
Figure 10.3 First recommended flowchart to be used as a utility in the optimization

process; it shows the interface between pad design and SRV’s.
Local pad –> global merging
Load reservoir parameters and

data
Generate FI values for

reservoir
Collect user-designed
optimization parameters
User selects desired pad

locations
Identify viable well pad

locations
Preprocess: generate frac SRV

for all wells in well pad reach
Preprocess: generate well pad,

well, frac LP model
Solve well pad model via MLP solver
Store model solution
All models solved?
Convert models from LP to link-based

models
Merge and trim models
Save solved global model and export

for simulator
End
Figure 10.4 Second recommended approach of optimization process.

(a) (b)
Figure 10.5 Four different SRV’s are located, whether staggered, or staggered and
overlapping designs. (a) Staggered design. (b) Staggered and overlapping design.
methodology, Objective Function, the essential sets, variables, and con-

stant parameters followed by presentation of the optimization procedure.
10.3.1 Methodology
The following describes the numerical formulation of the newly developed
technology.
10.3.2 Objective Function

The problem can be formulated as follows:
max Amn Xmn FkYk (10.3)
An alternative objective function is to maximize the net income obtained

from unconventional reservoirs. The net income is calculated as the differ-
ence between total income from total hydrocarbons produced and the total
capital and operating expenses, including optimum wells and fractures.
10.3.3 Assumptions and Constraints Considered

in the Mathematical Model
Sets, variables and decision variables are assumed as follows. For detailed
explanation, see geometric interpretation of parameters in Appendix C.
where,
10.3.3.1 Sets
1. Amn, total FI values unlocked by the stage from node m to
node n
2. Fk, total FI values unlocked by fracturing at node k
3. Lmn, length of stage from node m to node n
4. EI(n), edges inbound to node n
5. EO(n), edges outbound from node n
6. S(n), set of stages that connect to or pass through node n,
represented as edges in the network
7. Ψ(n), set of starting nodes mutually exclusive with w
8. E, set of valid edges for the network
9. Λ(m, n), set of fractures accessible from the stage between
node m and node n
10. Φ(n), set of nodes whose fracture SRV would intersect the
fracture SRV of node n
11. Ω(n), set of stages intersected or interfered with by fractur-
ing node n
10.3.3.2 Variables
1. Pi,j,k, total FI values unlocked by fracturing or placing well at
node i, j, and k
10.3.3.3 Decision Variables

m
2. w binary variable equal to 1 if well m goes through node
i , j ,k
i,j,k (m = 1…maxwell)
3. fim, j ,,nk binary variable equal to 1 if fracture n, extending from
well m, goes through node i,j,k (n = 1… maxfrac)
4. Wn, well origin for single well originating at node n,
continuous
5. Xmn, flow of connections from node m to node n, continuous
6. Υn, fracturing of node n, continuous
7. Snm, usage of stage connecting node m to node n, binary
10.3.3.4 Extended Sets

8. ψw (m,i,j,k), set of nodes that cannot have well if wim, j ,k 1 (10.4)
mn
9. ψf (m,n,i,j,k), set of nodes that cannot have well fi , j ,k 1 (10.5)
10. Ωw(m,i,j,k), set of nodes that cannot have well if due to

constraints on well orientation angle (10.6)
11. Ωf (i,j,k), set of nodes neighboring node (i,j,k) along the plane
of minimum stress
10.3.3.5 Constant Parameters

1. MinSpace: Minimum spacing between fractures
2. MinStageSpace: Minimum spacing between stages
3. MaxStageSpace: Maximum spacing between stages
4. MaxLength: Maximum allowed length for a single well
5. MaxStages: Maximum number of active stages in a well
6. MaxFPS: Maximum number of fractures per stage
7. MaxWells: Maximum number of wells per pad
8. MaxPads: Maximum number of pads per reservoir
9. MaxPadCost: Maximum cost for global development
10. MaxCost: Maximum cost for global development
11. Thickness: Reservoir thickness
12. WC: Cost for initial construction of a well originating from
a pad
13. PCmn: Cost for construction of a well segment from node m to n
10.3.3.6 Constraints
The problem is subject to the following constraints:
1. Fracture half-length length (Xf) < 0.9 (D1 + D2) (10.7)

2. The fracture propagates only along the predefined direction
of maximum stress at each location (10.8)
3. Fracture half-length (Xf) < Xe & Xf <500 ft. (10.9)
4. There is no fracturing in angled paths
5. Snm {0, 1} [active stage variable, may not be used.] (10.10)
6. 0 ≤ Wα,n ≤ 1 (10.11)
7. 0 ≤ Xmn ≤ 1 (10.12)
8. 0 ≤ Υk ≤ 1 (10.13)
9. Xnp Xmn 0 [flow conservation
p|(n , p ) EO (n ) m|(m , n ) EI (n )
with no sink] (10.14)
10. Xm , n 1, (m , n ) S(n) [non-overlap constraint] (10.15)
11. Xw , n 1, n (n) [non-overlap constraint] (10.16)

12. E ( Xij Lij ) MaxLength [constrains length of well] (10.17)

(i, j)
13. Yk MaxStages [constrains number of stages in well] (10.18)

k
14. W ,kMaxWells [constrains number of wells per pad]

k
(10.19)
15. Xmn MaxFPS Yk 0 [access to fractures for
k (m , n )
active stages; limited to straight, level stages] (10.20)
16. Yk 1 [non-overlap of fracture/SRV] (10.21)
k (n )
17. Yk 1 [minimum spacing between two frac-

k |dist (n , k ) MinSpace
tures, focused on node n] (10.22)
18. | (n)| Yn Xij | (n)| [non-overlap of fracture
( i, j) (n)
SRV for node k with stages that the SRV would intersect or
be too close to (within 50 ft.)] (10.23)
19. WC W ,k (X ij PCij ) FC Yk MaxCost (10.24)

k (i, j) E k
Equation (10.3), corresponds to the objective function which computes

the total FI values. Equations (10.4), (10.5) and (10.6) ensure that these
nodes cannot have a well due to fracture-reorientation constraint angle
from the minimum horizontal stress direction, as confirmed in equation
(10.7). Equations (10.7) and (10.9) establish the upper limit of the fractures’
half-length. Equation (10.8) ensures that the fractures propagate only in
the direction of the minimum horizontal stress. Equations (10.11), (10.12)
and (10.13) establish the continuous range for well origin, connection flow
and fracturing nodes. Equation (10.14) guarantees flow conservation with
no sink. Equations (10.15) through (10.18) define three constraints includ-
ing non-overlap, length of well and number of fracture stages.
10.3.4 Stimulated Reservoir Volume Representation

Ideal drainage area of each SRV is shown below in Figure 10.6 as SRV1,
SRV2 and SRV3. It is represented in our work through stimulated grids
around each created fracture.
Figure 10.7 shows one horizontal well connected with four different
SRV’s in a form of multistage transverse fractures. Figure 10.8 shows the
same number of SRV’s connected to four different wells.
Well origin region

(Pad reach) Pad location at surface
SRV 2 SRV 1
SRV 3
Figure 10.6 Drainage volume for different SRV for different fracture stages branched out
from a single horizontal well.
Figure 10.7 One Horizontal well passes through four different SRV’s. Model domain
shows the molded portion of our reservoir along one horizontal well of 10,000 ft. The four
different separate regions represent four different volume SRV’s.
10.3.5 Optimization Procedure

The procedure includes four steps as outlined below:
Step 1: Obtain a detailed 3D map of geomechanical properties

of an unconventional shale reservoir using industry stan-
dard sonic techniques.
Step 2: Use Fracturability Index algorithm [11] and previously
defined SRV to create 3D distribution of the reservoir qual-
ity shale model.
Figure 10.8 Combination of 4 wells and SRV for the model before optimization, where
the blue cells represent the FI below the cut-off.
Step 3: Run the Field Level Optimization procedure developed

in this work to get the optimum number of wells and frac-
tures, as well as the placement of the wells and fractures.
Step 4: Perform numerical (computational) simulations for the
optimum results.
The sequential steps in this process, is shown in the flow charts provided
in Figure 10.3 and Figure 10.4.
The automated process is a direct tool utilizing two software programs:
the global optimizer and the reservoir simulator. The first step begins
with a spreadsheet input file that is built to facilitate the entry data of the
Fracturability Indices data, or any other input maps. The currently used set
of input data maps is generated using the correlation published in Alzahabi
et al. [13].
The number of optimum wells, number of optimum fractures within the
wells, and the spacing between wells and fractures will be suggested using
the approach. It is believed that fracturing the optimum zones will contrib-
ute to higher hydrocarbon production from shale and tight formations.
10.4 Model Building

Many wells from the Permian basin are analyzed to help build a represen-
tative geochemistry map and mineralogical Index. Petrophysical log data
from two wells in Wolfcamp were used with geostatistical techniques to
construct a detailed geological model that is used here for testing. Many
relationships of parameters (e.g. porosity, permeability, quartz, clay con-
tent, E) were tested in this work to understand the Wolfcamp shale reser-
voir. The reservoir model has many 2D layers. The 2D nodes have FI values
assigned to it. The commercial reservoir Simulator Eclipse and Petrel were
used to populate the properties, then the correlation of FI was programmed
to generate the quality maps of Fracturability Index values, denoted by
(FI), as shown in Figure 10.9. The terminology quality map was introduced
by Da Cruz et al. (2004) and is commonly used in conventional reservoirs
in identifying producing regions. The heterogeneous properties of the res-
ervoir model are represented in the FI values. These input maps serve as
input for the optimization developed model.
These quality map generations were applied for each layer in the reser-
voir model.
The importance of the new model proposed here lies in its simplicity
and relative accuracy for the theory used in this work. More importantly,
it is based on easily obtained maps that are becoming more available in
today’s applications of shale characterization. Unlike many of the avail-
able SRV prediction tools, the new model does not require information
obtained from real microseismic data. Therefore, the SRV can be estimated
before drilling many wells in the reservoir.
Figure 10.10 compares a generated SRV before and after applying the
filter of FI = 0.5 on a chosen fracture stage of the shale model. Figure 10.10
shows the difference in grid-based modeling of one SRV before and after
removing the cells which have values of FI < 0.5.
Map of FI values for 2D reservoir example

50
Scale
45
0.7
40
0.6
35
0.5 30
0.4 25
20
0.3
15
0.2
10
0.1 5
0 10 20 30 40 50
Figure 10.9 Fracturability index distribution for the middle layer in the 80 × 80 × 5 model.
(a) (b)
Figure 10.10 A comparison between SRV, before and after applying the filter of FI. Note that
some grids with assigned values of FI < 0.5 were removed in Figure 10.10b. (a) One chosen
SRV before applying the filter of FI > 0.5. (b) One SRV after applying the filter of FI > 0.5.
10.4.1 Simulation Model of Well Pad and SRV’s Evaluation

To build a fast and accurate mathematically optimized approach to locate
wells and SRV’s in the shale reservoir model, a model with random natu-
ral fractures distribution to represent the complexity of shale rock is built.
Figure 10.11 shows a comparison between optimized placements of fracks
in terms of SRV versus uniform optimized cases; the parameters of the
reservoir model are listed in Table 10.1.
Number of fracks vs. cum. oil prod.

140
Optimized
Cum oil prod., BBls thousands
120
Uniform distribution
100
80
60
40
20
0
0 2 4 6 8 10 12 14
Number of fracks
Figure 10.11 Optimized placement of the same size SRV versus uniform distribution of
SRV along one horizontal well.
Table 10.1 Parameters of the reservoir model used in validating the developed
model.
Well type Horizontal wells
Reservoir dimension, ft. 10,000 × 10,000
Fracture half length, ft. 500, 600, 650, 700, 750
Number of hydraulic fractures/well 5–50
Number of wells/pad 2–40
Length of horizontal well (6,000–10,000 ft.)
Half length of fracture, SRV half (200–600 ft.)
length
Spacing (wells, fractures) (500–1,600 ft.)
Stage width 200 ft.
1. Input Parameters
i. 3D FI excel sheet heat map
ii. Min Fracture stage (SRV) spacing
iii. Reservoir dimensions X, Y, Z
iv. Number of wells: 2 wells (budget constraints), assum-
ing we have two pads at the surface, each pad has 2
wells
v. Horizontal well length 6,000 ft
vi. Overlap constraints
vii. Spacing between wells and neighboring fractures
2. Output Parameters
viii. Number of fractures stages (SRV) per well and per the
pad
ix. The location of surface well pads per the model.
x. The locations Fracture stages (SRV) along the hori-
zontal wells
xi. Optimum scheduling of fracture stages (numbered
rank based on sum of FI values per the stage).
xii. Well spacing
xiii. Fracture dimensions (length & width assuming frac-
ture propagates as a network through the whole frac-
ture stage)
xiv. Predict SRV’s location and number using an input

map of FI’s.
xv. An optimum well path and optimum fracture design
placement in shale and tight formations
Assumption
• Symmetric planar bidirectional propagation of the fracture

from the wellbore occurs perpendicular to the direction of
minimum principal stress.
• Wells are in the direction of the minimum horizontal stress
(predefined by the user).
10.5 Results and Discussions

This section is to evaluate the performance of FI via reservoir simulation.
The Permian basin evaluated consists of hundreds of fractures per well.
Table 10.2 shows the data ranges used to develop the model.
As the optimization technique has evolved, statistical algorithms and
many software packages have been developed to improve the under-
standing of fracture stages and the SRV concept, and complex simu-
lations are being implemented to take into account a greater number
of variation in input parameters; however, the importance of consid-
ering the power of FI correlation and its predefined cut-offs remains
paramount.
The model presented in this work is based on coupling the sweet spot proxy
and optimization tool. It requires input maps of calculated Fracturability
Indices. Contrary to detailed microseismic-based techniques, it requires
downhole sensing tools.
10.6 Conclusions and Recommendations

In this paper an analysis of placing optimum wells and optimum frac-
tures in the sweet spot regions of shale reservoir is presented. These sweet
spots are known as grid assigned high values of FI. The time and num-
ber of required stages to reach the objective function was investigated.
The fracture spacing and well spacing was assumed to be non-even. SRV
size is not equal among stages due to the heterogeneous nature of the rock
Table 10.2 Properties of the reservoir model used in validating the developed
model.
Mineralogical properties
Parameters Minimum value Maximum value
Quartz, wt. % 6.00 75.0
Calcite wt. % 0.00 84.0
Clay wt. % 3.00 49.0
Pyrite wt. % 0.00 8.00
Petrophysical properties
Photoelectric Index (Pe), 2.61 5.71
barns/electron
Density ρz, g/cc 2.41 2.71
Geomechanical properties
E, psi 0.38 E6 9.75 E6
ν, ratio 0.02 0.38
Reservoir properties
Initial reservoir pressure, psia 5,330
Thickness, ft. 900
Model dimensions 80 × 80 × 5
Porosity, % Avg. : 9
Permeability, Nano-darcy Avg. : 208
represented in varied FI values. The advantage of this helps in reducing the

cost of placing wells and fracture stages.
The concluding remarks from this paper are listed below.
In this work, a mathematical optimization approach for the placement
of horizontal wells and hydraulic fractures within shale reservoirs was
developed. The approach is able to provide a design that gives the opti-
mal predicted stimulation of sweet spot locations that are identified by the
use of the Fracturability Index. The technique suggests the optimal num-
ber of wells and fractures needed in order to drain the shale reservoir by
achieving maximum contact area, while respecting the physical and eco-
nomic constraints.
i. A model that includes the coupling of geomechanical

and mathematical optimization was determined for the
well data by use of a sophisticated Integer Programming
approach. It is believed that the proposed model arrived
at in this analysis is the best of its kind in the industry.
A comparison of our proposed model versus published
models (although published models are based on other
non-optimal algorithms), shows better results in terms of
accuracy in placing fractures. As a final recommendation,
more refined models could be proposed in future work
involving the collection of more data.
ii. Geometric placement of SRiV’s and hydraulic fracture
stages in shale and tight formations should be replaced by
coupled approaches of sweet spot indices and optimiza-
tion methodologies.
References
1. Britt, L.K. and Schoeffer, J., The geomechanics of a shale play: What makes a
shale prospective? 2008. SPE-125525.
2. Cipolla, C.L., Warpinski, N.R., Mayerhofer, M.J., Lolon, E.P., Vincent, M.C.,
The relationship between fracture complexity, reservoir properties, and frac-
ture treatment design. Presented at the 2008 SPE Annual Technical Conference
held in Denver, Colorado, USA, 21–24 September, 2008, SPE-115769, doi:
10.2118/115769-MS.
3. Fisher, M.K., Wright, C.A., Davidson, B.M., Goodwin, A.K., Fielder, E.O.,
Buckler, W.S., Steinsberger, N.P., Integration fracture mapping technologies
to optimize stimulations in the Barnett Shale. Presented at the SPE Annual
Technical Conference and Exhibition held in San Antonio, Texas, 2002,
SPE-77441.
4. Fisher, M.K., Heinze, J.P., Harris, C.D., Davidson, B.M., Wright, C.A., Dunn,
K.P., Optimizing horizontal completions techniques in the Barnett Shale
using microseismic fracture mapping. Paper 90051 presented at the SPE
Annual Technical Conference and Exhibition, Houston, 26–29 September,
2004, doi: 10.2118/90051-MS.
5. Mayerhofer, M.J., Lolon, E.P., Warpinski, N.R. et al., What is stimulated
reservoir volume? Presented at SPE Shale Gas Production Conference, Fort
Worth, Texas, Society of Petroleum Engineers, 2010, SPE-119890, doi:

10.2118/119890-MS.
6. Anderson, D.M., Nobakht, M., Moghadam, S., Mattar, L., Analysis of produc-
tion data from fractured shale gas wells. Presented at the SPE Unconventional
Gas Conference held in Pittsburgh, Pennsylvania, 2010, SPE-131787, doi:
10.2118/131787-MS.
7. C. Zhou and S. Sil, Fracture identifcation from azimuthal migrated seismic
data. Patent US20130201795, 2013.
8. S. Sil, R. Keys, R. Baishali, D. Foster, Methods for seismic fracture parameter
estimation and gas flled fracture identifcation from vertical well log data.
Patent EP2506039A3, 2013.
9. Zhang, S.C., Lei, X., Zhou, Y.S., Xu, G.Q., Numerical simulation of hydraulic
fracture propagation in tight reservoirs by volumetric fracturing. Pet. Sci., 12,
674–682, 2015, doi: 10.1007/s12182-015-0055-4.
10. Cheng, Y., Guo, B., Wei, N., Prediction of fracture population and stimu-
lated reservoir volume in shale gas/oil reservoirs. Presented at the SPE Asia
Unconventional Resources Conference and Exhibition held in Australia, 2015,
SPE-176833, doi: 10.2118/176833-MS.
11. Alzahabi, A., Soliman, M., Berlow, N., Spinner, T., 2016.
12. AlQahtani, G., Alzahabi, A., Kozyreff, E., Farias, I., Soliman, M., A compari-
son between evolutionary metaheuristics and mathematical optimization to
solve the wells placement problem. Adv. Chem. Eng. Sci., 3, 4A, 30–36, 2013,
doi: 10.4236/aces.2013.34A1005.
13. Alzahabi, A., Soliman, M.Y., AlQahtani, G.D., Bateman, R.M., Asquith, G.,
Fracturability index maps for fracture placement in shale plays. Hydraul.
Fract. J., 2, 1, 8–18, 2015. ISSN 2373-8197.
Appendix A
Abbreviations
E Young’s modulus
v Poisons’ ratio
FCD Dimensionless conductivity
K Permeability
Δσh Difference between minimum and maximum hori-
zontal stress
FI Fracturability Index
E Young’s Modulus, psi
E Plane Strain Modulus, psi
E Normalized Plane Strain Modulus
v Poisson’s ratio
ρ Density, lb./ft3
x(i,j) X_Y location of a fracture in the reservoir represents
the location (I, j) in the shale formation grid
X Coordinate axis along well path, ft.
Y Coordinate axis along fracture path, ft.
BHP Bottomhole pressure, psi
Qg Gas fow rate, Mscf/d
cf Formation compressibility, psi-1
Lf Fracture half length, ft.
Pi Initial reservoir pressure, psi
L Well Lateral length, ft.
ΔX,ΔY Model grid dimensions in x and y direction, ft.
Pnet Net pressure, psi
Xe, Ye Rectangular Reservoir shape dimensions in x and y
direction, ft.
W Horizontal well
F Fracture stage
Dmin Min well spacing, ft.
Appendix B
Definition of the Fracturability Index Used in the Well
Placement Process
In this model every cell is assigned a number based on the calculated FI
using geomechanical properties assigned to each cell. The FI consists of a
range between 0 and 1. The objective function is achieved through a sum

of the total FI values unlocked by the stage from node m to node n and the
total FI values unlocked by fracturing at node k.
Appendix C
Geometric Interpretation of Parameters Used in

Building the Model
Figure A10.12 demonstrates 5 wells attached to one pad. Figure A10.13 shows
one SRV ideal shape and ideal geometry. Figures A10.14 through A10.20 list
all geometric interpretation of network connections in wells and fractures.
Figure A10.14 shows a single well designated by originating at node
moving in the direction of minimum horizontal stress, where shows
continuous well connections from node m to n. Figure A10.15 shows a
Pad
D1
D2
Figure A10.12 Combination of 5 wells originating from one well pad.

Frac network height, h Frac network width, w

σ max σ min
Frac stage half length, Xf
Figure A10.13 Stimulated reservoir volume (SRV) for one fracture stage.
Well origin region

(Pad reach) Pad location at surface
Variable distance
e.x. space of 3,
Angular
next node 4 cells
movement of 1
away
(max allowed)
Well origin
Wn Xmn
(n)
Straight, level well path.

Only connection that allows
for fracking
Figure A10.14 Basic network connections in a well.
continuous flow of nodes representing fracture origin. Figure A10.16

defines the bounds of each individual valid SRV. Figure A10.17 differen-
tiates valid versus invalid fracture locations considering minimum frac-
ture distance. Figures A10.17 and A10.18 describe the possible geometrical
representation of the well path considering other existing SRV’s. Figure
A10.19 introduces the main dimensions of one possible SRV including
fracture stage height and width. Figure A10.20 explains how SRV is repre-
sented in our model, whereas discretized SRV is an approximation of SRV
ellipse 3D, as represented in Figure A10.13.
Yk = 0
Yk = 1
Node m Node n
Po tential fracture
locations (^(m, n))
Figure A10.15 Hydraulic Fracture stage nodes representation.
Min frac
distance
Invalid frac
(too close to
other frac)
Figure A10.16 Valid versus invalid nodes for fracture stage.

Frac
height
Frac
width
Invalid frac*
Invalid frac* (intersects SRV)
(intersects SRV)
*Handled by
Φ(n)
Figure A10.17 Three-dimensional representation of fracture intersection constraints.
Valid well path! Invalid well path*

(within 50 ft of frac)
Frac
height
Invalid well path*
(intersects SRV)
Frac
width
Invalid frac
Invalid well path* (intersects SRV)
(intersects SRV)
*Handled by
Ω(n)
Figure A10.18 Valid paths for neighboring wells.
Fracture height
Frac center,
well node
Fracture network width/

fracture stage length
Figure A10.19 Fracture stage parameters.

σmin
σmax Discretized SRV (sampled as approximated ellipse
centered at frac with height and width determined
by frac type)
Frac center
Frac half length
Approximate 50 ft gap (will need

to be adjusted based on scale)
Figure A10.20 Discretized SRV for one fracture stage.

11
A Semi-Analytical Model for Predicting
Productivity of Refractured Oil Wells with
Uniformly Distributed Radial Fractures
Xiao Cai1*, Boyun Guo1 and Gao Li2
1
University of Louisiana at Lafayette
2
Southwest Petroleum University
Abstract
Refracturing is a cost effective method enhancing the sustainability of both
depleted and good quality reservoirs. It is generally believed that refracturing
technology does not always work in all reservoir conditions. It is highly desirable
to have a quantitative method for predicting well productivity to select refractur-
ing well candidates. Assuming radial fractures around wellbore are uniformly dis-
tributed, a semi-analytical model was derived in this study to predict productivity
of refractured oil wells under pseudo steady state flow conditions. Result of a field
case study indicates that the mathematical model over-estimates well productivity
by about 10%. Sensitivity analysis with the mathematical model shows that the
productivity of refractured wells increases non-linearly with the number of radial
fractures around the wellbore. Sensitivity analysis with the mathematical model
also shows that the productivity of refractured wells increases non-linearly with
fracture conductivity, but the benefit of increasing fracture conductivity levels out
beyond 2000 md-inch. Therefore, there is no need to create very high-conductivity
fractures in refracturing operations. This paper provides petroleum engineers a
handy tool to assess well candidates for refracturing.
Keywords: Refracturing, oil wells, productivity, analytical model
*Corresponding author: xxc6743@louisiana.edu

227
11.1 Introduction
Seeking sustainable methods of fossil fuels and addressing energy sustain-
ability are major contents of sustainable energy engineering. Refracturing
is considered one of the important technologies used to strengthen the
sustainability of the oil and gas reservoirs. It is a hydraulic fracturing oper-
ation conducted on a well which has already been hydraulically fractured.
Refracturing can sustainably increase the production rate and enhance
ultimate oil/gas recovery from not only the depleted reservoir, but also the
good quality reservoir. In the down turn of the oil and gas industry, com-
panies are realizing that refracturing old wells is a more budget-friendly
way to achieve cash flow than drilling and completing new wells [1, 2],
so it is also a sustainable method used to increase the cost effciency of oil
companies.
While some people are certain that it works quite well, others are still
unsure of the success rate of refracturing due to variation of reservoir con-
ditions [3]. It is highly desirable to have a quantitative method for assessing
reservoir conditions to select wells for refracturing.
For previously fractured wells, it is possible to create a secondary frac-
ture that is perpendicular to the first. The secondary orthogonal fracture
can be created within a certain time window that, in turn, depends on res-
ervoir properties [4]. The optimum time window for refracturing is in the
order of months to years. Roussel and Sharma [5] presented a systematic
methodology for selecting candidate wells for refracturing using produc-
tion data. It considers the stress reorientation occurring around a fractured
well, causing the refracture to propagate orthogonally to the initial fracture
in underdepleted sections of the reservoir.
Ren et al. [6] presented a novel method to determine the refracturing
strategy and priorities of the candidate wells by comprehensive analysis,
which considers reservoir properties, remaining oil distributions, and
induced stress fields. According to Miller et al. [7], reservoir depletion
from parent wells can create localized “pressure sinks.” When a nearby
infill well is fractured, the hydraulic fracture grows towards areas of lowest
stress (the depleted area) which can damage the parent well. Subsequently,
the production potential of the new well is decreased because it is now
draining from the same reservoir area as the parent well. With the advent
of modern diversion-based refracturing techniques, it is possible to create
a “pressure barrier” around the parent well that helps redirect the new frac-
turing treatment into other, virgin areas of the reservoir. By refracturing
the parent well first, this may not only preserve the original production of
Productivity of Refractured Oil Wells 229
the parent well, but it may also increase the production of the parent well,
as well as the new infill wells. Although refracturing adds to the capital
expenditure of the project, project economics may be more favorable with
an investment in parent well refracturing.
A gap exists between reservoir data and well refracturing design. Particularly,
a simple and accurate mathematical model is needed for quantitatively pre-
dicting the productivity of refractured oil wells. This study fills the gap.
No literature has been found to disclose any method for predicting pro-
ductivity of refractured wells where multiple radial fractures are created.
Methods for predicting productivity of a well with a single planar fracture
of infinite and finite conductivity are well documented in the literature
[8–11]. The concept of matrix-fracture cross flow presented by Guo and
Schechter [10] for single-fractured wellbore is employed in this study to
develop a semi-analytical model for quantitatively predicting productivity
of refractured wells.
11.2 Mathematical Model

It is expected that multiple radial fractures are created in refracturing the
same pay zone. Initiation and extension of these fractures can be simulated
using commercial software [5]. Assuming uniformly distributed radial
fractures (Figure 11.1), an analytical model was derived in this study to
investigate the long-term productivity of refractured oil wells. Derivation
Radial Fractures
k= kXky
hf
Lf
Figure 11.1 Sketch of radial fractures around a refractured vertical well.

of the analytical model is given in Appendix A. The resultant well produc-

tivity model is summarized in this section.
The well production rate in Darcy’s units is expressed by Eq. (11.1). The
derivation of Eq. (11.1) is shown in the Appendix A. Eq. (11.1) is created based
on the Eq. (A11.18) in Appendix A for the purpose of practical application.
Lf
4nkh f Pd
Q dx (11.1)
B0 tan( ) x
rw
where Q is liquid production rate, n is number of fractures, k kx k y is

the geometrical mean of horizontal permeabilities, hf is fracture height, μ is
liquid viscosity, Bo is the formation volume factor of oil, a = π/n is the half
angle between fractures, rw is the wellbore radius, Lf is fracture length, and
pd is expressed as
pd C1 x J1 (2i c x ) C2 x N1 (2i c x ) (11.2)
where J1 is a Bessel Function of the first kind in the order of 1 and N1 is a

Bessel Function of the second kind in the order of 1 and
4k
c (11.3)
wk f (tan )
pdT12
C1 (11.4)
T02T11 T12T01
pdT11
C2 (11.5)
T02T11 T12T01
T01 rw J1 (2i c rw ) (11.6)
T02 rw N1 (2i c rw ) (11.7)

T11 L f J1 (2i c L f ) (11.8)
T12 L f N1 (2i c L f ) (11.9)
pd pe pw (11.10)
where pe and pw are reservoir pressure and wellbore pressure, respectively.

For any given data set, the integral in Eq. (11.1) can be evaluated
numerically. The numerical integration procedure has been coded
in an MS Excel spreadsheet which is available from the authors upon
request. This spreadsheet was utilized in the field case study and sensi-
tivity analyses that are presented in the following sections. Because the
mathematical model involves analytical solutions for pressure drawdown
determination and numerical integration for flow rate calculation, it is
called a semi-analytical model.
11.3 Model Verification

No clean data set is found for refractured wells to verify the reliability of
the mathematical model. A real well completed with high energy gas frac-
turing was employed to check the model accuracy. It is the well PIC-25
in the Pirital Field located in Maturin, Venezuela. The well was tested in
Cretaceous and Naricual-5 formations of super-low permeability [12].
Propellant-assisted perforating was performed in 6 intervals of 9 feet long
each in the Naricual-5 formation. Test-derived and estimated data are
shown in Table 11.1. The well was perforated in a 2,500 psi underbalance
condition by spotting diesel in the tubing prior to perforating. A stabilized
oil rate of about 100 BOPD was achieved in a 20 ft perforated interval.
The model-predicted interval production rate is 113 stb/day, which is an
over-prediction by 13%.
11.4 Sensitivity Analysis

The reliability of the mathematical model depends on the accuracy of
model parameter values. These parameters with high uncertain values
Table 11.1 Estimated reservoir and fracture properties in the Naricual-5 formation.
Horizontal permeability kx 0.08 md
Horizontal permeability ky 0.08 md
Fluid viscosity 8 cp
Oil formation volume factor 1.1 rb/stb
Wellbore radius 0.328 ft
Number of fractures 6
Fracture length 20 ft
Fracture height 4.5 ft
Fracture width 0.1 inch
Fracture permeability 1000 md
Reservoir pressure 12472 psia
Wellbore pressure 5339 psia
include the number of fractures, fracture permeability, and fracture width.

A sensitivity analysis was carried out using the baseline data in Table 11.1.
Figure 11.2 shows the effect of the number of fractures on well productiv-
ity, while all other parameter values remain unchanged. It indicates that
well productivity increases non-linearly with the number of fractures.
This is because the drainage distances between fractures get shorter as
the number of fractures increases. The rate of increase gets higher as the
number of fractures grows. Thus, the optimal design of refracturing should
consider creating a complex fracture network in order to maximize well
productivity.
Equation (11.3) implies that fracture width and fracture permeabil-
ity appears in the model as a product called fracture conductivity (wkf).
Therefore, there is no need to analyze the effect of each of the parameters.
Figure 11.3 illustrates the effect of fracture conductivity on well produc-
tivity while all other parameter values are kept constant. It demonstrates
that well productivity increases non-linearly with fracture conductivity.
The benefit of increasing fracture conductivity levels out beyond 2000
md-inch. Therefore, there is no need to create high-conductivity fractures
in refracturing operations.
250
Well production rate (stb/day)
200
150
100
50
0
0 2 4 6 8 10
Number of fractures
Figure 11.2 Calculated effect of number of fractures on well productivity.
350
Well production Rate (stb/day)
300
250
200
150
100
50
0
0 2000 4000 6000 8000 10000
Fracture conductivity (md-inch)
Figure 11.3 Calculated effect of fracture conductivity on well productivity.
11.5 Conclusions
Assuming uniformly distributed radial fractures around wellbore, a
semi-analytical model was derived in this study to predict productivity of
refractured oil wells under a pseudo steady state flow condition. Field case
study and sensitivity analyses were performed with the semi-analytical
model. The following conclusions are drawn:
1. Results of a field case study for a well completed with high

energy gas fracturing indicate that the mathematical model
over-estimates well productivity by about 10%. This incon-

sistency is attributed to the uncertainties in estimating
model parameter values.
2. Sensitivity analysis with the mathematical model shows that
the productivity of refractured wells increases non-linearly
with the number of radial fractures around the wellbore.
This is because the drainage distances between fractures get
shorter as the number of fractures increases. Because the
rate of increase gets higher as the number of fractures grows,
optimal design of refracturing should consider forming a
complex fracture network.
3. Sensitivity analysis with the mathematical model also shows
that the productivity of refractured wells increases non-
linearly with fracture conductivity, but the benefit of increas-
ing fracture conductivity levels out beyond 2000 md-inch.
Therefore, there is no need to create high-conductivity frac-
tures in refracturing operations.
Acknowledgements
The authors are grateful to Chevron USA and the UNOCAL Corporation
for providing the LA Board of Regents Chevron I and Chevron II
Professorships in Petroleum Engineering and UNOCAL Professorship in
Engineering throughout this study.
References
1. R. Khusainov, B. Ganiev, A. Karimova, and O. Karpova, Refracturing is the
best way to develop hard-to-recover reserves in Romashkino oilfield con-
ditions. Presented at SPE Russian Oil and Gas Exploration and Production
Technical Conference and Exhibition, Moscow, Russia, 14–16 October
(2014).
2. E. Urban, D. Orozco, A. Fragoso, K. Selvan, and R. Aguilera, Refracturing
Vs. infill drilling - A cost effective approach to enhancing recovery in shale
reservoirs. Presented at Unconventional Resources Technology Conference,
San Antonio, Texas, USA, 1–3 August (2016).
3. M. C. Vincent, Refracs: Why do they work, and why do they fail in 100 pub-
lished field studies? Presented at SPE Annual Technical Conference and
Exhibition, Florence, Italy, 19–22 September (2010).
4. N. P. Roussel and M. M. Sharma, Quantifying transient effects in altered-

stress refracturing of vertical wells. SPE J. 15, 770–782 (2010).
5. N. P. Roussel and M. M. Sharma, Selecting candidate wells for refracturing
using production data. SPE Production & Operations 28, 36–45 (2013).
6. J. Ren, J. Guo, Y. Deng, L. Zhang, and H. Lin, Refracturing strategy for seven-
springs oilfield in china based on geological and engineering characteristics.
Presented at SPE/IATMI Asia Pacific Oil & Gas Conference and Exhibition,
Nusa Dua, Bali, Indonesia, 20–22 October (2015).
7. G. Miller, G. Lindsay, J. Baihly, and T. Xu, Parent well refracturing: economic
safety nets in an uneconomic market. Presented at SPE Low Permeability
Symposium, Denver, Colorado, USA, 5–6 May (2016).
8. R.G. Argawal, R.D. Carter, and C.B. Pollock, Evaluation and prediction of
performance of low-permeability gas wells stimulated by massive hydraulic
fracturing. J. Petrol. Technol. 31, 362–372 (1979).
9. H. Cinco-Ley and F. Samaniego, Transient pressure analysis for fractured
wells. J. Petrol. Technol. 33, 1749-1766, (1981).
10. B. Guo and D.S. Schechter, A simple and rigorous IPR equation for vertical
and horizontal wells intersecting long fractures. J. Can. Petrol. Technol. 38,
46–54 (1999).
11. J. Li, B. Guo, D. Gao, and C. Ai, The effect of fracture face matrix damage on
productivity of fractures with infinite and finite conductivities in shale gas
reservoirs. SPE Drilling & Completion 27, 347–353 (2012).
12. J. Ramirez, J. Barrera, and R. Romero, Propellant-assisted perforating in
high-pressure and temperature wells at Campo Bosque in Northern Monagas
State. Presented at SPE Annual Technical Conference and Exhibition, New
Orleans, Louisiana, USA, 30 September–3 October (2001).
Appendix A: Derivation of Infow Equation for Wells

with Radial Fractures under Pseudo-Steady State
Flow Conditions
Radial fractures around a vertical wellbore are illustrated in Figure 11.1.
The following assumptions are made in deriving a well inflow model:
1. The rock in the pay zone is homogeneous.

2. Reservoir pressure is above the bubble point so that single
phase liquid flow dominates.
3. Linear fluid flow governed by Darcy’s Law prevails in the rock.
4. All fractures are identical in geometry.
5. Fluid flow from the rock directly to the wellbore is negligible.
6. Pseudo-steady state flow prevails in the fractured region.
p
ry
da
un
Bo
ow
-fl
No
r a A Lf
o x
x
Δx
Fractures
Figure A11.1 A plan view of fractures around a wellbore.
Figure A11.1 shows a plan view of radial fractures around a wellbore. It

can be shown that the half-angle between fractures is a = π/n, where n is
the number of fractures.
Consider the fluid flow from point P to point A at a radial distance,
x, from the wellbore centerline. The fluid flow rate, q(x), over a short
interval of fracture dx can be formulated. Under a pseudo-steady state
flow condition, the fluid production is driven by the expansion of fluid
and rock in the reservoir. The total compressibility of the reservoir is
defined as:
1 V
ct (A11.1)
V p
Differentiation of Eq. (A11.1), with respect to time, gives an expression

of the flow rate
p V
ctV q( x ) (A11.2)
t t
where the fluid volume is expressed as
x tan( ) ( x Δx )tan( )
V Δh f h f tan( )xΔx (A11.3)
2
The pressure decline rate ∂p/∂t in Eq. (A11.2) is then expressed as
p q( x ) q( x )
(A11.4)
t ctV ct h f tan( )xΔx
It is well known that the following equation governs linear flow in

porous media:
2
p ct p
2 (A11.5)
y k t
where the effective permeability in the y-direction, perpendicular to the

fracture orientation, can be estimated based on directional permeabilities,
i.e. k kx k y
Substituting Eq. (A11.4) into Eq. (A11.5) yields:
2
p q (x )
2 (A11.6)
y kh f tan( )xΔx
Integrating Eq. (A11.6) one time yields:
p q( x )
y c1 (A11.7)
y kh f tan( )xΔx
where c1 is an integration constant and can be determined using the no-flow

boundary condition
p
0. (A11.8)
y y x tan( )
Applying Eq. (A11.8) to Eq. (A11.7) gives
q( x )
c1 (A11.9)
kh f Δx
Substituting Eq. (A11.9) into Eq. (A11.7) gives
p q( x ) y
1 (A11.10)
y kh f Δx x tan( )
Separating variables, Eq. (A11.10) is changed to
q( x ) y
dp 1 dy (A11.11)
kh f Δx x tan( )
which is integrated to get
q( x ) y2
p y c2 . (A11.12)
kh f Δx 2 x tan( )
where the integration constant, c2, can be determined using the boundary
condition at the fracture face
p y 0 p f ( x ). (A11.13)
where pf(x) is the pressure in the fracture at point x. Applying Eq. (A11.13)
to Eq. (A11.12) gives
c2 = pf(x) (A11.14)
Substituting Eq. (A11.14) into Eq. (A11.12) results in
q( x ) y2
p p f (x ) y (A11.15)
kh f Δx 2 x tan( )
Along the no-flow boundary, y = x tan(a), the pressure is pe. Eq. (A11.15)
demands
2kh f Δx
q( x ) [ pe p f ( x )] (A11.16)
x tan( )
The fluid velocity in the matrix in the y-direction at the fracture face at
point x can be expressed as
q( x ) 2k
v( x ) [ pe p f ( x )] (A11.17)
h f Δx x tan( )
The cumulative flow rate of fluid collected by a fracture interval between

well-bore and point A can be determined based on v(x) as
Lf x
2kh f
Q( x ) 2 v( x )h f dx 2 [ pe p f ( x )]dx (A11.18)
x tan( )
x Lf
If the average width of the fracture is w, Darcy Velocity, vf(x), in the

fracture can be formulated by dividing Eq. (A11.18) by the cross-sectional
area of the fracture:
Q( x )
v f (x ) (A11.19)
wh f
Applying Darcy’s Law to the flow along the fracture gives
k f dp f ( x )
v f (x ) (A11.20)
dx
Coupling Eq. (A11.20) and Eq. (A11.19) yields
Q( x ) k f dp f ( x )
wh f dx (A11.21)
Substituting Eq. (A11.18) into Eq. (A11.21) and rearranging the latter
gives
x
dp f ( x ) 2k
2 [ pe p f ( x )]dx (A11.22)
dx wk f x tan( )
Lf
Differentiation of Eq. (A11.22), with respect to x, yields
d 2 p f (x ) 4k
2
[ pe p f ( x )] (A11.23)
dx wkx x tan( )
Defning pd as the pressure drawdown at point x in the fracture
pd = pe – pf (x). (A11.24)
and
4k
c (A11.25)
wk f tan( )
Substituting Eqs. (A11.24) and (A11.25) into Eq. (A11.23) yields
d 2 pd c
pd (A11.26)
dx 2 x
The first boundary condition is given at the wellbore and expressed as
pd x rw pd pe pw (A11.27)
The second boundary condition is expressed as
pd|x Lf 0. (A11.28)
The general solution to Eq. (A11.26) is
pd C1 x J1 (2i c x ) C2 x N1 (2i c x (A11.29)
where J1 is a Bessel Function of the first kind in the order of 1 and N1 is a

Bessel Function of the second kind in the order of 1. The constants, C1 and
C2, are determined based on the boundary conditions to be
pdT12
C1 (A11.30)
T01T11 T12T01
and
pdT11
C2 (A11.31)
T02T11 T12T01
where
T01 rw J1 (2i c rw ) (A11.32)
T02 rw N1 (2i c rw ) (A11.33)
T11 L f J1 (2i c L f (A11.34)
T12 L f N1 (2i c L f (A11.35)

Part 5
ENVIRONMENTAL ISSUES OF
HYDRAULIC FRACTURING
Introduction
Population growth and the vital need for energy are among important chal-
lenges to be tackled by nations across the globe. Fossil fuels have long been
relied upon to meet the energy demand despite their limited quantity and
environmental problems such as greenhouse gas emission which is broadly
recognized as a worldwide issue encountered when heavy hydrocarbon fuels
are burnt. Moving towards a more sustainable future necessitates harnessing
renewable sources of energy and combining them with non-renewable sources
to create a flexible energy portfolio. However, renewables are not yet ready to
go online because of their high intermittency and some availability limitations.
More extensive research and investigation are warranted until reaching a point
where renewables will be fully able to quench the world’s thirst for energy.
In a large country such as the United States, the need and desire to lessen
dependency on imported conventional energy sources has always been a
point of discussion. According to the United States Energy Information
Administration (USEIA) the country is trending downward in the use of
imported oil, after the peak year of 2005 with showing an export number
greater than import in 2015 for the first time in a long time [1]. Shifting
from conventional sources to the more sustainable renewables requires
substantial time and effort, primarily to ensure that the energy production
from renewables is economically viable [2]. Natural gas, as a candidate for
short and even mid-term energy solutions, has garnered attention of both
energy companies and energy demanding countries. As a light hydrocar-
bon fuel, natural gas is easily accessible and is dependable which means
unlike some renewable energy sources, natural gas’s performance is not a
function of environmental conditions such as sunlight or wind, nor is this
energy source intermittent. When compared to heavier hydrocarbon fuels,
natural gas emits less harmful by-products when burnt. Hydraulic frac-
turing can help easing the process of translation from heavy hydrocarbon
fuels to a more sustainable energy resource by producing natural gas. This
technology, however, is not without deficiencies when it comes to protect-
ing the environment.

244 Part 
Hydraulic fracturing is an old technique which has been around for sev-
eral decades. Extensive use of fracturing in recent years is attributed to
advances in horizontal drilling and the capability of reaching deep target
formations with lower costs. Concerns have been raised on wide-spread
application of hydraulic fracturing in oil and gas and other operations
despite of the benefits of this technology [3, 4]. Unique safety and environ-
mental considerations must be taken to mitigate concerns and ensure safe
operations [5]. In particular, little is known about long-term environmen-
tal consequences given the relatively short history of extensive horizontal
drilling.
Potential environmental impacts can be classified into three major cat-
egories: air, induced seismicity, and water and wastewater. Air pollution
in general is linked to natural gas development activities, including frac-
turing. For example, a human health risk assessment study by McKenzie
et al. [6] on a gas field in Garfield County, Colorado suggested that people
living at a distance greater than half a mile from the site were in less dan-
ger than the ones living at a lesser distance. According to a more recent
study by Moore et al. [7], particulate matters smaller than 2.5 μm in size
(PM2.5) and Nox are found in diesel emissions during preproduction (i.e.
drilling and hydraulic fracturing activities) of unconventional natural gas
production.
Subsurface injection operations normally generate low magnitude
(smaller than 2 in Richter scale) seismic events which are termed as
micro-earthquakes or microseismic [8]. In a few cases of hydrofracking
jobs, seismic events have been felt by nearby residents and the operator has
been forced to stop the injection. In 2011, two seismic events (2.3 and 1.5
magnitude in Richter scale) were recorded close to Blackpool, Lancashire
in the U.K. -- the operator had to halt the fracturing operation in nearby
Bowland Shale formation [9]. Generally, the risk of generating serious
earthquakes as a result of hydraulic fracturing is low when compared to the
deep-well injection process, which shows higher probabilities of observed
larger seismic events [8, 10, 11]. Historically, the oldest injection-induced
seismic events are those of the Rocky Mountain Arsenal waste injection
site in Denver, Colorado [12]. The most recent events were generated in
Youngstown, Ohio [13] and central Oklahoma [14], both in 2011. The
former is also known to be the largest injection-induced event with a 5.6
magnitude [8].
Potential surface water [15, 16] and groundwater contaminations are
caused by either fracturing fluid chemicals or volatile compounds from
deep formations [17–20]. The returned fluid handling and treatment is
another source of possible environmental complications [16, 21–24].
Part  245
Section 5 includes three chapters: Chapter 5-1 addresses major con-

tributing factors to the safety of hydraulic fracturing. This chapter further
covers analyzing different steps of a typical fracturing job and discussing
root-causes of example incidents and accidents and the public perception.
Post fracturing fluid recovery is the topic of Chapter 5-2 where the fate of
fracturing fluid within the reservoir has been investigated and responsi-
ble factors for low recovery of large portions of injected fluid have been
identified. Chapter 5-3 wraps up Section 5 by introducing and numerically
solving a potential groundwater contamination problem in the domain
of hydraulic fracturing where a breach in the well results in solute move-
ment in the adjacent formation and contamination of nearby underground
water resources.
References
1. U.S. Energy Information Administration, 2018.
2. Spellman, F., Environmental Impacts of Hydraulic Fracturing, CRC Press,
2012.
3. Kargbo, D.M., Wilhelm, R.G., Campbell, D.J., Natural gas plays in the
Marcellus Shale: challenges and potential opportunities. Environ. Sci.
Technol., 44, 15, 5679–84, 2010.
4. USEPA, Assessment of the Potential Impacts of Hydraulic Fracturing for Oil
and Gas on Drinking Water Resources, 2015.
5. Jabbari, N., Ashayeri, C., Meshkati, N., Leading Safety, Health, and
Environmental Indicators in Hydraulic Fracturing. SPE Western Regional
Meeting, Garden Grove, CA, 27-30 April 2015, 2015.
6. McKenzie, L.M. et al., Human health risk assessment of air emissions from
development of unconventional natural gas resources. Sci. Total Environ.,
424, 79–87, 2012.
7. Moore, C.W. et al., Air Impacts of Increased Natural Gas Acquisition,
Processing, and Use: A Critical Review. Environ. Sci. Technol., 2014.
8. Ellsworth, W.L., Science (New York, N.Y.), 341, 6142, 1225942, 2013.
9. Pater, C. De and Baisch, S., Geomechanical study of Bowland Shale seismicity,
2011.
10. Goebel, T. et al., A probabilistic assessment of waste water injection induced
seismicity in central California. American Geophysical Union Fall Meeting,
San Francisco, 2014.
11. Aminzadeh, F., Tiwari, A., Walker, R., Correlation between Induced
Seismic Events and Hydraulic Fracturing activities in California. American
Geophysical Union Fall Meeting, San Francisco, 2014.
12. Healy, J. and Rubey, W., The Denver earthquakes. Science (New York, N.Y.),
161, 3848, 1301–1310, 1968.
246 Part 
13. Ohio Department of Natural Resources, Preliminary Report on the North Star
1 Class II Injection Well and the Seismic Events in the Youngstown, Ohio, Area,
2012.
14. Holland, A., 2011.
15. Hammer, R. and VanBriesen, J., In fracking’s wake: new rules are needed to
protect our health and environment from contaminated wastewater, Natural
Resources Defense Council, 2012.
16. VanBriesen, J., Wilson, J., Wang, Y., 2014.
17. Osborn, S.G. et al., Methane contamination of drinking water accompanying
gas-well drilling and hydraulic fracturing. Proc. Natl. Acad. Sci. U.S.A., 108,
20, pp.8172–6, 2011.
18. Vengosh, A. et al., The Effects of Shale Gas Exploration and Hydraulic
Fracturing on the Quality of Water Resources in the United States. Procedia
Earth Planet. Sci., 7, pp.863–866, 2013.
19. Llewellyn, G.T. et al., Evaluating a groundwater supply contamination inci-
dent attributed to Marcellus Shale gas development. Proc. Natl. Acad. Sci.,
112, 20, pp.6325–6330, 2015.
20. Jabbari, N., Aminzadeh, F., de Barros, F.P.J., Hydraulic fracturing and the
environment: risk assessment for groundwater contamination from well cas-
ing failure. Stochastic Environ. Res. Risk Assess., 31, 1527–1542, 2017, doi:
10.1007/s00477-016-1280-0.
21. Gregory, K.B., Vidic, R.D., Dzombak, D.A., Water Management Challenges
Associated with the Production of Shale Gas by Hydraulic Fracturing.
Elements, 7, 3, pp.181–186, 2011.
22. USEPA, Study of the Potential Impacts of Hydraulic Fracturing on Drinking
Water Resources, 2012.
23. Gordalla, B.C., Ewers, U., Frimmel, F.H., Hydraulic fracturing: a toxicolog-
ical threat for groundwater and drinking-water? Environ. Earth Sci., 70, 8,
3875–3893, 2013.
24. Glazer, Y.R. et al., Potential for Using Energy from Flared Gas for On-Site
Hydraulic Fracturing Wastewater Treatment in Texas. Environ. Sci. Technol.
Lett., 1, pp.300–304, 2014.
25. Jabbari, N., University of Southern California, Los Angeles, California,
2016, Retrieved from http://digitallibrary.usc.edu/cdm/ref/collection/
p15799coll40/id/231045.
12
The Role of Human Factors Considerations
and Safety Culture in the Safety of
Hydraulic Fracturing (Fracking)
Jamie Heinecke1, Nima Jabbari2* and Najmedin Meshkati3
1
Senior Student, Daniel J. Epstein, Department of Industrial & Systems
Engineering, Viterbi School of Engineering, University of Southern California
2
Ph.D. Candidate, University of Southern California, Sonny Astani Department of Civil
and Environmental Engineering, Viterbi School of Engineering, University of Southern
California, Los Angeles, CA
3
Professor, Sonny Astani Department of Civil/Environmental Engineering,
Daniel J. Epstein, Department of Industrial & Systems Engineering,
Viterbi School of Engineering, University of Southern California
Abstract
Hydraulic fracturing is a well stimulation method frequently used in oil and gas
operations in order to facilitate the flow movement and increase production in
tight and low permeability formations. Along with horizontal drilling, hydraulic
fracturing has helped the United States in meeting its enormous need for energy.
Despite of the advantages such as boosting economy and energy-independency,
there are potential environmental and safety issues associated which rise many
questions on the future of this technology. More specifically, recent accidents and
incidents have heightened public and industry concerns. This paper addresses
major contributing factors to the safety of fracturing by analyzing different steps
of the operation and providing associated human factors and safety culture con-
siderations. Primary human factor root-causes of three recent fracking accidents
have been analyzed, and recommendations as how to proactively address these
considerations in the entire hydraulic fracturing life-cycle process are proposed.
Keywords: Hydraulic fracturing, fracking, safety, human factors, accident

causation, safety culture
*Corresponding author: jabbari@usc.edu

247
12.1 Introduction
The United States currently has access to some of the largest shale oil and gas
reserves in the entire world. “Since 2007, discoveries of unconventional gas
including shale gas have more than doubled the estimate of North American
reserves to 3,000 trillion cubic feet, enough to meet 100 years of demand”
[1]. Shale rock is a unique type of rock formation that lies over a mile below
the surface and contains a significant amount of gas trapped inside tiny
pockets within the rock [2]. Until recently, companies found it difficult to
extract these precious resources using conventional drilling methods.
However, the improvement of hydraulic fracturing (also known as fracking)
in recent years, has allowed companies to take advantage of these reserves.
Fracking allows gas to flow to the surface by pumping water under extreme
pressure down into the well in order to fracture the rock formation. The pro-
cess behind hydraulic fracturing is not entirely new, but became especially
advantageous when combined with another technique called horizontal
drilling. This combination is so effective that according to the U.S. Energy
Information Administration (EIA), production of natural gas will meet con-
sumption needs by 2019 and 12% of all production will be exported by 2040
[3], as shown in Figure 12.1. In fact, it is primarily due to fracking in the U.S.
that already, according to a recent article in the Wall Street Journal (August
12, 2014), “since March 2008, oil production has increased 58% and
natural-gas output has risen 21%, making the U.S. the world’s largest producer
of both fuels, according to federal and international agency statistics”
History 2011 Projections

40
Net exports, 2040 (12%)
Total production
30
Net imports, 2011 (8%)
Total consumption
20
10
Net imports
0
-10
1990 2000 2010 2020 2030 2040
Figure 12.1 Total U.S. natural gas production, consumption, and net imports in the
reference case, 1990–2040 (trillion cubic feet) [3].
The Role of Human Factors Considerations 249
(emphasis added); and “jobs directly related to oil and gas production have
nearly doubled in the past 10 years to 697,000, government data shows” [4].
Fracking will become an increasingly important topic of discussion as it
starts to play a larger role in America’s energy future. Human factors con-
sideration should be the foundation of this rapidly growing technology to
ensure that it is implemented correctly in order to minimize the drawbacks
of the fracking process.
The potential benefits of fracking could be enormous for the United
States, but only if executed properly. Due to the increasing pressure to find
alternative energy sources, hydraulic fracturing is being implemented in a
greater capacity than ever before. “In a decade, shale gas has risen from 2
percent of US natural gas production to 37 percent. The US has overtaken
Russia as the world’s largest natural gas producer” [5]. However, such rapid
and enormous growth poses significant safety risks. As companies rush
to capitalize on this opportunity, they tend to over-look many safety and
environmental concerns such as increased seismicity as a result of waste
injection [6] and groundwater contamination [6]. These challenges have
been crucial enough in some cases to make the cities take serious actions to
regulate, monitor, and even halt the fracking operations just like the cases
of Pavillion, Wyoming in 2009 [7] and New York in 2010 [8].
Furthermore, according to the National Institute of Occupational Safety
and Health (NIOSH) [9], “the oil and gas extraction industry has an annual
occupational fatality rate of 27.5 per 100,000 workers (2003–2009) - more
than seven times higher than the rate for all U.S. workers. The oil and gas
extraction industry employed approximately 435,000 workers in 2010. The
annual occupational fatality rate in this industry is highly variable; this
variation is correlated with the level of drilling activity in the industry.
Fatality rates are higher when there is an increased number of active drill-
ing and workover rigs.” An accident in drilling-related operations, like in
other contexts, can be characterized as “an error with sad consequences”
[10]. Nevertheless, a correct understanding of the root-causes of the afore-
mentioned so-called “error” in terms of instances of human-machine,
human-task, and/or human-organization mismatches can be greatly con-
tributed to its prevention [11–14].
To minimize accidents which could result in fatalities and nonfatal
injuries and the possibility of other negative effects associated with frack-
ing, one must consider the human factors issues involved in each stage of
the process. By examining each stage of hydraulic fracturing, the specific
human factors issues involved in the overall process can be addressed and
ultimately both the safety and the efficiency of the entire system can be
improved.
As discussed by [15], the fracturing operation is known to have sev-

eral challenges to the environment and also to the society ranging from
groundwater contamination and air pollution to increasing traffic and
incidents associated. However, the purpose of this paper is to shed light on
the root causes of accidents and to analyze potential risk areas that could
lead to future accidents. This paper does not extensively examine the envi-
ronmental impacts of fracking, but instead it focuses on the safety of work-
ers and well integrity from a human factors perspective. It will also analyze
the human, technical, and organizational factors that contribute to well
control related issues and worker safety oversight.
12.2 Benefits of Hydraulic Fracturing

According to experts, shale gas has the potential to “dramatically alter
the energy supply picture for North America and potentially the world as
other countries are just beginning to determine the extent of their own
unconventional resources” [16]. The United States has the opportunity
to outperform the rest of the world if it can improve fracking techniques
by fixing often overlooked human factors related issues. As the world’s
leader in this technology, the United States would generate an immense
stimulus. Current operations have already shown positive economic
effects on local industries; “the most immediate has been in employment –
more than 1.7m jobs have been created” [5]. Expanding operations will
greatly help a struggling American middle class while easing our energy
dependence and reducing carbon emissions (natural gas is a light hydro-
carbon with fewer negative consequences than the heavier hydrocarbons
that are more commonly used for fuel). Despite such potential benefits,
fracking has drawn significant criticism, largely as a result of human factor
issue oversight.
12.3 Common Criticisms

Fracking is not without imperfections and can be improved upon to reduce
the risks associated with the process. Media discussions about fracking
tend to focus on the possibility of long-term risks:
Despite the arguable benefits of fracking, some public health officials,

environmentalists and scientists are not convinced that it is worth the
potential risks. Many groups are concerned over the lack of long-term
research into the effects the chemicals used in hydraulic fracking have
on the people and environment around it, as well as some grey areas
in industry regulation [16].
The majority of critiques raise concerns over the contamination of fresh

water supplies and argue that, “despite the industry’s claims that hydraulic frac-
turing is a safe and proven technology […] there have been many allegations
that hydraulic fracturing had led to the contamination of drinking water in
many communities” [2]. Residents in communities located in Wyoming, New
York, and Colorado have insisted that their water tastes and smells abnormal
since fracking began in the local area. In response, fracking companies claim
that it is impossible that contaminants were able to reach the aquifer level
because the fracturing process is thousands of feet away from the water table.
Through numerical simulations it has been shown that in some extreme cases
and rare events, failure in the system may result in contaminating the forma-
tion nearby the well and the groundwater aquifer eventually [17].
Critics also claim, in addition to water contamination, that fracking has
caused more earthquakes in the area. However, while fracking does create
seismic activity, “the energy released by rock breaking during shear frac-
turing has a magnitude about one hundred thousand times smaller than
the magnitude of the slightest ‘felt’ earthquake (magnitude ~3.0) (USGS
calculator)” [18]. The cause of these earthquakes is not the act of drilling
and fracturing, but rather has been traced to the disposal wells used to
store the contaminated flow-back fluid (e.g. [5]). “These wells are located
thousands of feet underground, encased in layers of concrete and usu-
ally store the waste from several different wells” [19]. The wells can cause
faults to slip resulting in an earthquake. The Rocky Mountain Arsenal in
Colorado and north-central Arkansas are two well-documented cases in
areas which link fluid injection disposal wells with earthquakes [20]. Both
areas have seen an increase in the size and frequency of earthquakes near
the underground wastewater disposal wells. Only a handful of the thou-
sands of disposal wells have been linked with seismic activity, but because
of the magnitude of this possible risk, new protocols should be established
to deal with the disposal of waste fluid. For example, many companies are
beginning to send their waste to a water treatment facility. Although this is
likely a more expensive option, it seems to provide a promising alternative
to underground disposal wells. Microseismic monitoring and data gath-
ering during the fracturing operation [21], applications of sophisticated
reflection seismology methods [22] and more specifically using real-time
microseismic monitoring accompanied by keep tracking of pore-pressure
changes [15] can play crucial roles in mitigating the issue and in increas-
ing the safety of the hydrofracking operation. Also, in a study done in
the San Joaquin Basin of California, it has been shown that the seismic
events related to the oil and gas activities can be distinguished from the
natural events using semiology measures such as fractal dimension and
b-value [23].
The final major area of concern for critics is the composition of frack-
ing fluids used to break open the shale formation. Although many critics
argue these fluids include harmful chemicals, “evaluating the relative vol-
umes of the components of a fracturing fluid reveals the relatively small
volume of additives that are present. Overall the concentration of addi-
tives in most slickwater fracturing fluids is a relatively consistent 0.5% to
2% with water making up 98% to 99.2%” [24]. However, it is important
to remember when analyzing these percentages that millions of gallons
of water are used in each drilling operation and the long term effects of
what these chemicals will do in the ground is unknown. Furthermore,
the exact composition of the fracking fluid used by each company is
not fully revealed because it is considered proprietary, which causes
additional unease. These risks and benefits of fracking are affected by
operations in each stage and can be improved through human factors
management.
In the end, if these allegations are true, experts say that it is not due to
the scientific fracking technique but instead is a result of human error. This
illustrates the key obstacle fracking must overcome: the science behind
fracking is valid, but in accounting for human factors, it becomes a much
less reliable process that may be prone to significant and costly problems.
Also, it is worthwhile to note that the concerns and worriedness about the
fracturing are important issues to be addressed through communications
with the public media. It, thus, necessitates proactively conducting sessions
and talks on the scientific facts regarding the hydraulic fracturing and try
to stay away from the myths [25–27].
12.4 Different Steps of Hydraulic Fracturing and

Proposed Human Factors Considerations
Hydraulic fracturing operation includes different steps starting from
drilling vertically and horizontally all the way to injection and returning
the fluid from the well. On the other hand, human factor considerations
are vital to any system, but they are especially important when a tech-
nique is still being perfected. The improvements made by studying the
human factors of the operation could allow fracking projects to realize
their enormous potential. Figure 12.2 briefly discusses major steps and
Examples of Human Factors

Fracking Steps Description Considerations1
Vertical Well is spudded vertically up to a specified Using proper lifting and material handling
Drilling depth. techniques.
Human-machine interface
Communications
Monitoring/Supervisory Control
Job hazards (identification & conrol);
PPEs; (”Electricity hit”) [29]
Vertical Vertical casing is inserted in place to Standard operating procedures
Casing function as a barrier to the surrounding
environment.
Drilling operation and equipment

monitoring
By reaching the kick-off point, drilling bit Using proper lifting and handling
Horizontal starts deviating from the vertical direction, techniques.
Drilling creates a bend, and advances in horizontal
Human-machine interface
direction
Communications
Monitoring/Supervisory Control
Job hazards (identification & control);
PPEs; (”Electricity hit”) [29]
Horizontal
Horizontal casing is placed after the Following cementing standards and
Casing horizontal drilling is finished. Horizontal protocols
section in entirely located in the target Information monitoring (”Supervisory
formation. Control”)
Monitoring/information processing/signal
Perforation Horizontal casing is perforated using detection.
electrical charges.
Situational awareness
Ensure the blow-out preventer system

Injection Fracking slurry is injected through an functions properly.
injection pipe inside the casing and Information processing/monitoring
formation is cracked.
Decision making
Ensure waste is handled properly

according to regulations.
Returning Part of the injected fluid is returned back to Situational awareness and continuous
Waste the surface and is temporarily stored on- microseismic monitoring of formation
site. The waste is dumped later on. condition and integrity to avoid
triggering seismic events [5]
Continuous motioning and
characterization of waste health hazards
Wastewater is injected underground through
Wastewater a number of deep injection wells. Standard Operating Procedures/PPEs
Injection Total system comprehension and shared
mental model of the underground
1Will be further discussed in the section of hydraulic fracturing process as well as the section of human factors
Figure 12.2 Hydraulic fracturing steps (inspired from [28]) and related major human factors.
important human factors issues pertained to each step. Safety culture

and its associated elements (or traits) such as organizational leadership,
communication, training, and fatigue management, are cross-cutting
phenomena that potentially affect every single step of the fracturing
operation.
12.5 Hydraulic Fracturing Process: Drilling

The beginning of the fracking operation involves a drilling stage. In this
stage, a drill bit mounted on the end of a drill pipe creates the well. The
hole extends to a level just below the aquifer zone and a casing (termed as
surface casing) is inserted into the hole. Cement is pumped between the
casing and the hole in order to seal off the wellbore from the freshwater.
This process is continued further down the hole until there are multiple
layers of cement to prevent any leakage into the water supply. When the
drill reaches the appropriate length, it is then drilled horizontally into the
shale formation (as shown by Figure 12.3).
Casing is placed and cement is again pumped the full length of the well-
bore. During this drilling process, the drill bit requires maintenance which
is called “tripping pipe.” This is a multiple step job that could be potentially
extremely dangerous for workers.
The first stage of tripping pipes involves setting slips that attach around
the stem of the drill. Generally, there are three workers involved in this
process. The potential hazards of this stage include getting fingers or other
body parts pinched between the slips or slip handles in addition to a risk
of muscle strain due to poor lifting techniques. It is imperative for worker
safety that they are trained to use proper hand placement when handling
the slips and use proper lifting techniques to prevent possible injury [31].
The next stage involves setting the kelly (large hexagonal steel struc-
ture) over the hole. During this process there are three identifiable possible
Ground Water Aquifers

Conductor Casing
Surface Casing
Intermediate Casing
Production Casing
Horizontal Well Vertical Well
500 ft. Radius

Vertical Fractures in Vertical Wells
Producing Formation
Vertical Fractures in Horizontal Well
Figure 12.3 View of vertical and horizontal drilling [30].

hazards. The first is the possible release of excess drilling mud which can get
on the workers skin or create a slippery floor surface. In order to counter
this, it is important to follow the proper protocol of shutting down the mud
pumps and using a kelly that has a mud saver valve. The second hazard is
the possibility of being struck by slip handles when the drill string is spun.
There are alternative technologies such as a pipe spinner that companies
should consider investing in to increase worker safety. The last hazard is
the possibility of getting struck by the kelly when it is being placed over
the hole. Again, proper training can prevent any incidents of this hazard
occurring [31].
The third stage requires workers to latch elevators onto the pipe to pre-
pare for lifting or lowering the pipe. Potential risks include getting hands
pinched, caught, or being struck by elevators. Supervisors need to make
sure that workers are correctly following latching procedures and elevators
should be inspected and maintained after each use.
Tripping the pipe also requires the worker to climb the oil derrick and
work from an area called the monkeyboard (component 4 of Figure 12.4).
There are numerous hazards to address in this stage. Workers can fall when
they are climbing the ladder or while working from the monkeyboard. It
is very important for the worker to use a climb assist device and to wear
a body harness to protect against falls. They should be wearing the proper
protective clothing including a hard hat, gloves, and safety-toed footwear.
Companies need to ensure that slip-resistant coatings are put on all work-
ing surfaces. Additional hazards include getting caught between the pipe
and various objects and being struck by dropped objects. Workers need to
exercise additional caution when work is being done over head and to use
proper hand placement [31].
The next stage involves using a pair of tongs to break out and disconnect
the pipe. Tongs are tools that consist of two long arms that aid in seizing or
holding an object. If these tongs slip or backlash during operations, seri-
ous injury could occur to the worker. It is important to implement proper
break out procedure and workers should be outside of the 4 foot tong rota-
tion area (as seen in Figure 12.5). Communication is essential between the
driller and the floor hands that are operating the tongs [31].
The final stage is to rack the pipes. Crew members need to be wary of
getting their hands pinched between pipes or getting feet crushed under a
stand of pipe. It is important for workers to use proper hand positioning
and to properly position their feet away from pipe stands.
The risks in each process described above may be mitigated through
human factors consideration. In the drilling stage, proper training and exe-
cuting proper lifting and handling techniques will eliminate many risks.
Drilling Rig Components*
1. Crown Block and Water

Table
2. Catline Boom and Hoist
Line
3. Drilling Line
4. Monkeyboard
5. Traveling Block
6. Top Drive
7. Mast
8. Drill Pipe
9. Doghouse
10. Blowout Preventer
11. Water Tank
12. Electric Cable Tray
13. Engine Generator Sets
14. Fuel Tanks
15. Electric Control House
16. Mud Pump
17. Bulk Mud Components
Storage
18. Mud Pits
19. Reserve Pits
20. Mud Gas Separator
21. Shale Shaker
22. Choke Manifold
23. Pipe Ramp
24. Pipe Racks
25. Accumulator
Equipment used in drilling
Figure 12.4 Drilling rig components [26].
Hazardous
Area
Mouse
Hole
Hole
Center RADIUS
4 FEET
Caution Area
Figure 12.5 Drilling rig floor hazardous area layout with tong swing radius [31].
Due to the fracking boom, more and more inexperienced workers are being
hired and put to work without sufficient training. This puts both the workers
and the environment at risk. Companies must invest in thorough training to
keep their workers safe. Such training is also in the company’s best interest
because the reduction of personal hazards improves job performance.
12.6 Hydraulic Fracturing Process: Fluid Injection

The fluid injection stage can begin once the drilling is complete. First, the
casing along the horizontal portion of the well must be perforated. To do
this, a perforating gun is sent down to the desired location and a small
explosive charge is set off using an electrical current. The charge punctures
the well casing and part of the shale rock creating fissures to allow the frac-
ture fluid to enter the rock formation (Figure 12.6).
Fracking fluid consists of roughly 90% water, 9.5% sand, and 0.5%
other chemicals. These “other chemicals” have been a topic of controversy
as well. Although the percentage is small, there are millions of gallons of
fracking fluid being used so the overall impact is significant. This fluid is
pumped under high pressure down into the wellbore creating fissures in
the shale formation. These fissures allow the trapped gas to flow into the
wellbore and up to the surface for capture [32]. Workers insert a plug into
the wellbore and the process is repeated multiple times throughout the
horizontal wellbore, which can be over a mile long. When this is complete,
the plugs are removed and the wastewater is pulled back up to the surface
and stored in open pits until it can be safely disposed of. The sand from
the fracking fluid prevents the fissures in the shale formation from closing,
Perforating Cement
Gun
Detonation
Cord
Jet
Charge Casing Formation
Figure 12.6 Components of the well perforation process [30].

creating pathways for the gas to continually flow into the wellbore and up
to the surface to be harvested. Each stage of this fracking process presents
a few key human factor issues that must be addressed.
12.7 Fracking Fluid

The first topic that should be addressed from the human factors perspective
is the fracking fluid. Fracking fluid contains powerful chemicals including
hydrochloric acid, formaldehyde, and ethylene glycol. However, the real
threat to worker health is the sand. “NIOSH’s recent field studies show
that workers may be exposed to dust with high levels of respirable crys-
talline silica during hydraulic fracturing. Respirable crystalline silica is the
portion of crystalline silica that is small enough to enter the gas-exchange
regions of the lungs if inhaled; this includes particles with aerodynamic
diameters less than approximately 10 micrometers” [31]. NIOSH findings
indicate that of 116 samples collected, 47% showed silica exposures higher
than the OSHA permissible exposure limit and 79% were above the rec-
ommended exposure limit. This exposure to silica is dangerous for work-
ers because “breathing silica can cause silicosis. Silicosis is a lung disease
where lung tissue around trapped silica particles reacts, causing inflam-
mation and scarring and reduces the lungs’ ability to take in oxygen” [33].
Because these workers are forced to breathe silica repeatedly they “are at
a greater risk of developing silicosis. Silica can also cause lung cancer and
has been linked to other diseases, such as tuberculosis, chronic obstructive
pulmonary disease, and kidney and autoimmune disease” [34]. Some pos-
sible solutions to limit the exposure of silica are to limit the distance sand
falls through the air to reduce dust kick up, limit the time spent in high
risk areas, and install a dust collection system on the machines that release
dust. Providing respirators will also help reduce the amount of exposure
the worker’s lungs receive. Proactive measures to protect employees should
be a primary concern for gas companies. Not only is it extremely costly for
companies to cover missed work, medical bills and lawsuits, but they will
also foster greater public support by showing a concern for their workers’
well-being, thereby improving the fracking industry’s reputation.
12.8 Wastewater
The second human factors topic associated with the fracking process is
the flow back wastewater. Flow back fluids contain not only the harsh
chemicals found in fracking fluids but also other contaminants and natu-
rally occurring radioactive materials from exposure to the shale formation.
The fluid is usually stored in temporary pits or steel tanks until it can be
properly treated or disposed of. Correctly containing these fluids within
a lined pit is especially important for reducing the risk of contaminating
shallow ground water, a common concern among critics:
The failure of a tank, pit liner, or the line carrying fluid (“flowline”)
can result in a release of contaminated materials directly into surface
water and shallow ground water. Environmental clean-up of these
accidentally released materials can be a costly and time consuming
process. Therefore, prevention of releases is vitally important [24].
This stage possesses the highest likelihood of contaminating fresh ground

water due to human error. Luckily, new systems such a closed loop fluid han-
dling systems are being implemented throughout the industry to avoid the
use of open air pits. They work by “keeping fluids within a series of pipes
and tanks throughout the entire fluid storage process. Since fluid is never
in contact with the ground, the likelihood of groundwater contamination is
minimized” [24]. As this stage is the most at risk, there needs to be strict
standards set in place to reduce the risk of any spills due to mishandling or
human errors.
12.9 Human Factors and Safety Culture Considerations

Human factors and safety culture have been implicated as critical contribut-
ing factors to major accidents in both up- and down streams in the oil and gas
industry. BP Deepwater Horizon (2010) [35–37] and BP Texas City Refinery
fire and explosion (2005) [38] are among the famous examples of incidents.
Operations of every major step in the aforementioned “Fracking Steps”
on Figure 12.2, which includes Vertical Drilling, Vertical Casing, Horizontal
Drilling, Horizontal Casing, Perforation, Injection, and Returning Waste
are sensitive to and can seriously be compromised by the human factors
and safety culture considerations.
12.9.1 Human Factors

Human factors, or ergonomics, is a scientific field concerned with improv-
ing the productivity, health, safety, and comfort of people, as well as the
effective interaction between people, the technology they are using, and
the environment in which both must operate. While both human factors
and ergonomics are used interchangeably, the term ergonomics is used

when focusing on how work affects people [39]. This concerns issues such
as fatigue due to prolonged monitoring tasks, injuries due to unsafe work-
station, and errors due to a confusing console layout.
A straight-forward justification for the need for ergonomic consider-
ations is that the technological systems are being controlled by humans;
therefore, they should be designed with the human operator’s physical and
psychological needs, capabilities and limitations in mind. Ergonomics can
be divided into two related and complementary areas of concentration:
1- microergonomics and 2- macroergonomics. Micro- and macroergonomic
approaches build upon each other and concentrate on the introduction,
integration and utilization of technology, and its interface with the end-user
population. Their overall objective is to improve the safety and efficiency
of the intended technological system [40].
12.9.1.1 Microergonomics
Microergonomics, also called human engineering, addresses the relationship
between human, equipment and physical environment. It is focused on the
human-machine system level and is, for example, concerned with the design
of individual workstations, work methods, tools, control panels and displays.
Microergonomics includes studies of the human body sizes, known as anthro-
pometrics, physical and psychological abilities and limitations, information
processing, and human decision-making and error. It is noteworthy that, in
the context of the control room environment:
“The Human Error Probability (HEP) will be reduced by factors of

2 to 10 if the workstation (display and controls) are improved by the
incorporation of standard human engineering concepts” [41].
Microergonomics aims to reduce incompatibilities between operator abili-

ties and system requirements. The following sample represents additional
areas of microergonomics consideration: Materials Handling, Handtool Design
and Use, Machinery Design, Workstation Design, and Workplace Environment.
12.9.1.2 Macroergonomics
Ergonomics at the macro level, macroergonomics, is focused on the overall
people-technology system level and is concerned with the impact of techno-
logical systems on organizational, managerial, and personnel (sub-) systems.
Macroergonomics includes areas such as training, management, the plan-
ning process, information systems, internal review/inspection programs,
performance measurement systems, reward structure, initial employee qual-

ifications assessments, and personnel selection criteria [42]. Additional areas
on which macroergonomics focuses include: Job Analysis, Training,
Communications, Policies and Procedures, and Organizational Design.
12.9.2 Safety Culture

The human factors considerations, although are of great importance to frack-
ing, consist only of necessary conditions for achieving a safe and reliable oper-
ation; to make it sufficient, one should take into account safety culture. Safety
culture can be a common mode of failure for such incidents as it plays a vital
role in shaping a complex technological system resiliency, robustness and
defense-in-depth:
“Because of their diversity and redundancies, the defense-in-depth

will be widely distributed throughout the system. As such, they are
only collectively vulnerable to something that is equally widespread.
The most likely candidate is safety culture. It can affect all elements in
a system for good or ill.” [43]
The safe and efficient operation of fracking is a function of the interac-

tions among its human (i.e., personnel and organizational) and engineered
subsystems. More specifically, interactions of its Human, Organizational,
and Technological (i.e., engineered) (HOT) subsystems, within their over-
all operational milieu – could determine the safety culture [44].
The connection of these three (HOT) subsystems, in the context of the
total system, is represented in Figure 12.7. This simplified and symbolic
demonstration depicts only one critical system’s reality – the role of each
subsystem as a link in a chain – in the integrity of the whole system. It
does not, however, show all the needed subsystems’ interactions and
interrelationships.
The chain metaphor is also helpful in understanding the effects of output
or production load, produced by the system, on its individual subsystems.
Any increase in the output level or the capacity utilization rate imposes
strain on all subsystems.
Obviously, the chain (system) could break down if any link breaks down.
This may occur if either all the links (subsystems) are not equally strong
and designed for handling the additional load, or if they are not adequately
prepared and reinforced to carry the extra load in a sustainable fashion.
Major accidents at complex, large-scale technological systems have been
caused by break downs of the weakest links in this chain, which are most
often the human or organizational subsystems.
Interactive
Effect
Human
Organization
Technology
Volume of Output
Figure 12.7 Major subsystems of a technological system.
Finally, Safety culture can also be characterized as the “overlay” or direct

result of sound HOT interactions that should have been incorporated
during work system’s design and nurtured and maintained during opera-
tion stages. Furthermore, an organization’s safety culture, as a system com-
posed of behaviors, practices, policies, and structural components, cannot
flourish or succeed without interactions and harmony with its environ-
ment – the societal or national culture in which the organization which
runs the technological system must operate. In other words, safety culture
should be considered in the context of national culture, which could seri-
ously affect its effectiveness [45, 46].
Creating and nurturing a positive safety culture basically means to
instill thinking and attitudes in organizations and individual employees
that ensure safety issues are treated as high priorities. An organization fos-
tering a safety culture would encourage employees to cultivate a question-
ing attitude and a rigorous and prudent approach to all aspects of their
job, and would set up necessary open communications between line work-
ers and mid- and upper management. These safety culture characteristics
are equally applicable both to the operating companies as well as to their
cognizant/designated governmental regulatory safety agency.
It is noteworthy that the Institute of Nuclear Power Operations (INPO)
has recently developed a seminal guideline and code of practice for safety
culture in nuclear power industry, Traits of a Healthy Nuclear Safety Culture
(2013), that could (and should) be adopted and utilized for this purpose by
the fracking industry [47].
12.10 Examples of Recent Incidents

Fracking accidents and incidents can have multiple, complex and interact-
ing causes. However as depicted in the following Figure 12.8, such events
can result into three major categories of losses and adverse consequences:
People-related, Environment-related, and Product/Process-related. The
collective and overall impact of these three types of losses on the system’s
productivity can be attributed to their negative impact on resource utiliza-
tion; which could be significant, as they simultaneously reduce, lower or
erode the “Output”. In order to compensate for the inefficient utilization
of productive resources which should have gone into the final product, the
“Input” should be increasing. The ultimate result of this vicious cycle, in
the short-term, is low overall fracking “Productivity Ratio” and negative
“Safety, Health, Environmental, Economical, and Social Impacts”. And in
the long-term, depending on the severity of the three types of losses, major
regulatory and/or political ramifications can follow.
Fracking mishaps and accidents
Prevention Strategies Prevention Strategies

Human-Systems Integration (HSI) Engineering: Equipment, Design, Material,
(Human factors considerations, Operation, Maintenance, Hard and soft safety barriers
Situational awareness, Physical and mental Regulations
Workload analysis, etc. Process/
People
Safety culture Product
Procedures
Standard Operating Procedure (SOP)
Emergency Operating Procedure
(EOP)
Loss Environment Loss
Incidents Low/
Injuries Interruption
Accidents / Water of
production
Loss
Spills, Emissions
Contamination
Unintended
changes in
geological
formations and
stability
Output
Productivity Ratio =
Input
Final Outcomes
Safety, Health, Environmental, Economical, and Social Impacts
Figure 12.8 Risks/loses and impacts of fracking mishaps and accidents (adapted from [48]).
As discussed before and is also depicted in Figure 12.8, human factors

and safety culture considerations comprise a major loss prevention strategy
with an overarching effect on people, environment and product/process.
Example 1: Exxon Mobil subsidiary spills wastewater into the

Susquehanna River
Recent reports of fracking issues reveal that many of the problems are not
created by the technical procedure but are a result of human error. In 2010,
the Dallas News reported on a fracking spill outside Pittsburgh where a stor-
age tank was leaking water and chemicals into a nearby river. According to
the article, the XTO Energy Fracking Company that was responsible for the
operation was forced to pay a $100,000 fine and to implement operations
changes that cost nearly $20 million [49]. However, the leak was avoidable
and not an inherent problem of fracking procedures. Irresponsible human
actions caused the leak: “a valve on the storage tank had been left open
and the drainage plug removed,” additionally, “XTO had ordered its water
recycling contractor to another drilling site in West Virginia” [49]. Such
a simple error was extremely costly for this company, yet it was entirely
avoidable. Forcing the water-recycling contractor to move to another
project before the current one was secure represents a human factor error
associated with upper management attempting to maximize profits while
disregarding important safety and oversight concerns.
Example 2: Report cites inadequate management of risks

Poor decision-making by management is not the only possible cause of
human error. As an incident in northern Alberta revealed, worker disre-
gard can have disastrous implications. According to the report, “workers
failed to recognize and properly assess a number of issues that led to the
perforation and fracturing above the base of groundwater protection” [50].
These errors resulted in the crew directly fracking into underground water
table where 42 cubic meters of propane gel entered an aquifer. However,
the company’s response to this incident indicates that such errors were
a result of complete disregard for procedures. The company’s Chief
Operating Officer, Rob Morgan, added, “all of the personnel who were
involved in this particular circumstance are no longer with the company”
[50]. Unfortunately, this drastic incident has fueled criticisms of fracking,
yet it was a result of human error. Accidentally fracking above the aquifer
level is something that should never happen. As shown in Figure 12.9, the
operator missed the pay zone by over 800 meters. Operators need to be
correctly reading the equipment that monitors the depth levels. If lack of
experience of the operator was the issue then the company should evaluate
~50 meters NE
GL
Deep Monitoring
Well
Accidental
perforations at
136 mMD
177 mm Surface
Casing set to
500 mMD 606 mMD
Base of Ground
Water Protection
600 m 114 mm Production
KOP 697 mMD Previous to incident perforated and
Casing Set to
hydraulically fractured intervals
2503 mMD
Bridge Plug
1557 mMD
1000 mMD
Planned Perforation
Interval
1486–1486.6 mMD
Figure 12.9 Well cross-section schematic [50].
its training program. The company should also evaluate the monitoring
equipment to ensure that it is ergonomically correct and easy to use. If
equipment is not adjusted to ergonomic standards then the chance for
operator error is much higher.
Example 3: Mud spill in Pennsylvania state forest

Another incident occurred in January 2010 when “an estimated 8,000 gal-
lons to 12,000 gallons of mud used by Anadarko E&P Co. Inc. for drilling
operations over-flowed at the well site due to human error […] [Luckily, the
mud] didn’t spread far enough to contaminate any surface waters, ground
water, or wetlands in the area” [51]. Companies should be investing in addi-
tional monitoring equipment to have assurance in each stage of the opera-
tion. By breaking down the operation they can evaluate where the problem
areas lie and what they need to do in order to improve their process. Lack of
monitoring equipment in addition to inexperienced operators can be devas-
tating to not only to the company but to the environment as well.
12.11 Conclusion and Recommendations

By understanding and strengthening each of these interrelationships, the
safety of the hydraulic fracturing process can be improved upon. According to
a report released in 2012 by the JRC (Joint Research Centre), human-related
causes were present in 14% of offshore accidents at the very minimum. It is

likely that human-related causes were present in even more accidents, how-
ever, no systematic analysis to identify human/organizational failures were
performed in the other events [52]. It would be reasonable to assume that
onshore fracking accidents have similar numbers if not more.
Fracking has enormous potential for the United States in our quest for
cleaner energy. The economic benefits from this process are something that
even the industry’s stoutest critics have a hard time denying. “Last year, a
group from Yale estimated that shale gas production contributes over $100
billion to U.S. consumers annually. Jobs have been created, many landown-
ers have benefited financially, and lower gas prices have provided relief for
consumers in the form of lower heat and electricity bills. In comparison,
the authors estimated that the cost of ground-water contamination could
be $250 million per year, which is 1/400th the benefit” [53]. However, we
should not settle for the status quo. By strengthening current regulations
and improving human factors and safety culture, we can enhance the safety
of the workers, reduce accidents, and minimize adverse environmental
impact of fracking.
Acknowledgment
Authors would like to express their sincere gratitude to Dr. Iraj Ershaghi,
Dr. Donald Paul of University of Southern California (USC), and Dr. Wayne
D. Pennington of Michigan Technological University for their insights and
analyses concerning technical issues affecting the hydraulic fracturing pro-
cess. This work, however, should not necessarily be construed as their rep-
resentative position(s).
References
1. S.L. Sakmar, Shale gas developments in North America: An overview of the
regulatory and environmental challenges facing the industry, SPE North
American unconventional gas conference and exhibition, The Woodlands,
Texas. pp. 1, 3, 7 (2011).
2. K. Barillaro, Process of Fracking, http://shalestuff.com/education/fracking/
fracking (2012).
3. J.J. Conti, U.S. Energy Information Administration, Office of Integrated and
International Energy Analysis, Annual energy outlook 2013, http://www.eia.
gov/forecasts/aeo/ (2013).
4. Democrats warming to the energy industry: Lawmakers weigh benefits of

fracking boom against opposition from environmentalists; some Republicans
are skeptical, Wall Street Journal. p. A4, August 12 (2014).
5. D. Yergin, US energy is changing the world again, Financial Times, http://
www.ft.com/intl/cms/s/b2202a8a-2e57-11e2-8f7a-00144feabdc0,Autho-
rised=false.html?_i_location=http://www.ft.com/cms/s/0/b2202a8a-2e57-
11e2-8f7a-00144feabdc0.html?s iteedition=intl&siteedition=intl&_i_referer=.
p. 2. (2012).
6. K.M. Keranen, M. Weingarten, G.A. Abers, B.A. Bekins, and S. Ge, Sharp
increase in central Oklahoma seismicity since 2008 induced by massive
wastewater injection. Science 1255802 (2014).
7. D. DiGiulio, R.T. Wilkin, C. Miller, G. Oberlev. Investigation of ground water
contamination near Pavillion, Wyoming, EPA. (2011).
8. D. Rahm, Regulating hydraulic fracturing in shale gas plays: The case of
Texas. Energy Policy 39(5), 2974–2981 (2011).
9. National Institute for Occupational Safety and Health (NIOSH). Oil and Gas
Extraction, Inputs: Occupational Safety and Health Risks, http://www.cdc.
gov/niosh/programs/ oilgas/risks.html
10. J. Reason, Human Error, p. 162, Cambridge University Press, New York, NY.
(1992).
11. J. Rasmussen, The definition of human error and a taxonomy for techni-
cal system design, in New Technology and Human Error, J. Rasmussen, K.
Duncan, and J. Leplat (Eds.), John Wiley & Sons, New York, NY. (1987).
12. J. Rasmussen, Human error and the problem of causality in analysis of acci-
dents. Invited paper for Royal Society meeting on Human Factors in High
Risk Situations, 28-29 June, 1989, London, England. (1989).
13. N. Meshkati. Human Factors in Large-Scale Technological Systems’
Accidents: Three Mile Island, Bhopal, Chernobyl. Ind. Crisis Q 5, 133–154
(1991).
14. N. Meshkati. Where do accidents come from? The critical role of human and
organizational factors in the safety and reliability of technological systems,
Rivista Tecnica dell’ANPAC (Italian Pilots’ Union Technical Publication), 2,
14–16. (2000).
15. M.D. Zoback and A.J. Arent, Shale gas: Development opportunities and chal-
lenges, the Bridge, Nat. Acad. Eng. 44(1), (2014).
16. S. Tuthill, Understanding fracking: Arguments for and against natural gas
extraction, http://www.accuweather.com/en/weather-news/fracking-environ
mental-health/17444221. p. 1, (2013).
17. N. Jabbari, F. Aminzadeh, F. de Barros, B. Jafarpour, Hydraulic fracturing and
potential groundwater contamination risk. Presented at the 248th American
Chemical Society National Meeting & Exposition, Paper Number: 315, San
Francisco, CA, 10–14 (2014).
18. K.E. George, Hydraulic fracturing 101, J. Petrol. Technol. April 34–42, (2012).
19. T. Henry, How oil and gas disposal wells can cause earthquakes, https://
stateimpact. npr.org/texas/tag/earthquake/.p. 1, (2012).
20. S. Horton, Disposal of hydrofracking waste fluid by injection into subsurface
aquifers triggers earthquake swarm in central Arkansas with potential for
damaging earthquake. Seismol. Res. Lett. 83(2), 250–260 (2012).
21. N.R. Warpinski, Understanding Hydraulic Fracture Growth, Effectiveness,
and Safety through Microseismic Monitoring, ISRM International
Conference for Effective and Sustainable Hydraulic Fracturing. International
Society for Rock Mechanics, (2013).
22. W.D. Pennington, Reservoir geophysics. Geophysics 66(1), 25–30 (2001).
23. F. Aminzadeh, T.H.W. Goebel, Identifying induced seismicity in active tec-
tonic regions: a case study of the San Joaquin Basin, California. AGU Fall
Meeting, Abstract: S31F-06, San Francisco, California. (2013).
24. Frac Focus, Chemical Disclosure Registry, http://fracfocus.org/ (2013).
25. F. Aminzadeh, To Frack or not to Frack. That is the question. Los Angeles
Science Center Series on “Science Matters”, June 8 (2013).
26. N. Jabbari, F. Aminzadeh, F. de Barros, Hydraulic Fracturing, Points and
Counter Points. Urban Water Institute 20th Annual Water Policy Conference
San Diego, CA, August 16 (2013).
27. F. Aminzadeh and D. Paul, Induced Seismicity Intersection of Science,
Public and Regulation, IOGCC Annual Meeting, November 5. Long Beach
California. (2013).
28. I. Ershaghi and Q. Qianru, Aspects of Oilfield Related Accidents. SPE Annual
Technical Conference and Exhibition. Society of Petroleum Engineers.
(2013).
29. M. Economides and T. Martin, Modern Fracturing: Enhancing Natural Gas
Production, Energy Tribune Publishing Inc., Houston. (2007).
30. American Petroleum Institute, Hydraulic fracturing operations- well con-
struction and integrity guidelines, API Guidance Document HF1, doi:
Product No. GHF101, (2009).
31. OSHA, Oil and gas well drilling and service etool, http://www.osha.gov/
SLTC/etools/oilandgas/index.html (2014).
32. Clean Water Action, Fracking: The process, http://www.cleanwateraction.
org/page/fracking-process (2013).
33. J.A. Merchant, U.S. Department of Health and Human Services, National
Institute for Occupational Safety and Health, Occupational respiratory dis-
eases, Publication No. 86-102, p. 266, (1986).
34. F.L. Rice, U.S. Department of Health and Human Services, National Institute
for Occupational Safety and Health, Health effects of occupational exposure
to respirable crystalline silica, Publication No. 2002-129.p. 5, (2002).
35. National Academy of Engineering and National Research Council. Interim
Report on Causes of the Deepwater Horizon Oil Rig Blowout and Ways to
Prevent Such Events. The National Academies Press, Washington, DC. (2010).
36. I. Ershaghi and D. Luna, Validation process for human factor in com-
plex upstream oil and gas operations. SPE 147654, SPE Annual Technical
Conference and Exhibition, Denver, CO, (2011).
37. M. Tabibzadeh and N. Meshkati, Learning from the BP Deepwater Horizon
accident: Risk analysis of human and organizational factors in negative pres-
sure test. Environ. Syst. Decis. 34(2), 194–207 (2014).
38. United States Chemical Safety and Hazard Investigation Board. Investigation
Report, Refinery Explosion and Fire, BP Texas City, Texas. Report No. 2005-
04-I-TX. (2007).
39. Eastman Kodak Company. Ergonomic Design for People at Work (Vol. 1),
Lifetime Learning Publications, Belmont, California. (1983).
40. N. Meshkati, Critical Human and Organizational Factors Considerations
in Design and Operation of Petrochemical Plants. Proceedings of the First
International Conference on Health, Safety & Environment in Oil and Gas
Exploration and Production, Society of Petroleum Engineers (SPE), 11-14
November, The Hague, The Netherlands, Volume I, 627–634 (Paper # SPE
23275). (1991).
41. A.D. Swain and H.E. Guttmann, Handbook of Human Reliability Analysis
with Emphasis on Nuclear Power Plant Applications. Final Report (NUREG/
CR-1278). Washington, D.C. U.S. Nuclear Regulatory Commission. pp. 11–15,
(1983).
42. H.W. Hendrick, Macroergonomics: A concept whose time has come. Human
Factors Soc. Bull. 30(2), 1–3, (1987).
43. J.T. Reason, A Life in Error: From Little Slips to Big Disasters. Ashgate,
Burlington, VT, (2013).
44. N. Meshkati, Human factors in process plants and facility design, in Cost-
Effective Risk Assessment for Process Design, R. Deshotels and R. Zimmerman
(Eds.), pp.113, 115, McGraw-Hill, Inc., New York, NY. (1995).
45. N. Meshkati, Cultural context of nuclear safety culture: A conceptual
model and field study, in Nuclear Safety: A Human Factors Perspective,
J. Misumi, B. Wilpert, R. Miller (Eds.), pp. 61–75, Taylor and Francis,
London. (1999).
46. M.J. Gelfand, M. Frese, E. Salmon, Cultural influence on errors: Preventions,
Detections, and Management, in Errors in Organizations, D.A. Hofmann and
M. Frese (Eds.), pp. 273–315, Taylor & Francis, Routledge, New York, NY.
(2011).
47. Institute of Nuclear Power Operations. Traits of a Healthy Nuclear Safety
Culture. Institute of Nuclear Power Operations. INPO 12-012. http://
pbadupws.nrc.gov/docs/ ML1303/ML13031A707.pdf. (2013).
48. N. Meshkati, Preventing Accidents and Explosions in the Petroleum
Industry: The Critical Role of Human Factors. Invited plenary presentation
at the First International Congress on Medicine in the Petroleum Industry,
organized by Pemex, the State-owned oil company of Mexico. (1992).
49. J. Osborne, Exxon mobil subsidiary faces criminal charges over 2010 fracking
spill, Dallasnews, http://www.dallasnews.com/business/energy/20130911-exx-
on-mobil-subsidiary-faces-criminal-charges-over-2010-fracking-spill.ece. p. 1,
(2013).
50. C. Linnitt, Alberta finds mismanagement of errors causes fracking water
contamination. http://www.desmogblog.com/2012/12/22/alberta-finds-
mismanagement-errors-causes-fracking-water-contamination-alberta.
p.1.(2012).
51. R. Swift, Spill in state forest moves gas drilling moratorium debate, Times-
Tribune, http://thetimes-tribune.com/news/spill-in-state-forest-moves-gas-
drilling-morato-rium-debate-1.705590, p. 1, (2010).
52. M. Christou and M. Konstantinidou, Safety of offshore oil and gas operations:
Lessons from past accident analysis, JRC Scientific and Policy Reports, http://
publications.jrc.ec.europa.eu/repository/bitstream/111111111/27463/1/off-
shore-accident-analysis-draft-final-report-dec-2012-rev6-online.pdf (2012).
53. R. Rapier, Both sides mislead when it comes to fracking, http://blogs.wsj.
com/experts/2013/11/14/both-sides-mislead-when-it-comes-to-fracking
p. 1, (2013).d
13
Flowback of Fracturing Fluids with
Upgraded Visualization of Hydraulic
Fractures and Its Implications on
Overall Well Performance
Khush Desai* and Fred Aminzadeh†
The Mork Family Department of Chemical Engineering & Materials Science,

Viterbi School of Engineering, University of Southern California,
Los Angeles, California, USA
Abstract
The increasing popularity of hydraulic fracturing follows some achievements in the
USA, where it has become a proven technology in the stimulation of tight reser-
voirs. Nevertheless, the physics behind the process are not completely understood,
particularly in the domain of post fracturing fluid recovery. In many instances, the
recovery of large portions of injected fracturing fluid has not been successful. In
this research, the goal was to identify and evaluate the responsible factors.
The scope of our study includes determining the fate of the fracturing fluid
within the reservoir, and calculating the loss in incremental production as a conse-
quence of that outcome. The information can be used to more effectively predict the
performance of stimulated wells with hydraulic fractures over time. Also, an estima-
tion of incremental oil recovery post treatment can be calculated more accurately.
This new knowledge is beneficial to many of the participants - service provider
companies will enjoy a clear understanding of the treatment; operating companies
should perform more reliable economic analyses; the individual states could real-
ize increased accuracy in the quantification of reserves; and regulatory agencies
can better determine the probability of these fluids in groundwater contamination.
Keywords: Flowback, fracturing fluids, hydraulic fracturing, streamtubes,

well performance
*Corresponding author: khushdes@usc.edu

†
Corresponding author: faminzad@usc.edu

271
13.1 Introduction
A streamtube model has been developed as a tool to predict well perfor-
mance post hydraulic fracturing treatment. It is also useful in ascertaining
various possibilities of representative streamtube models that would be
characteristic of the reservoir.
As stated earlier, major portions of injected fracturing fluid cannot be
recovered. Various physical phenomena impact that reality. Several sen-
sitivity analyses have been conducted in order to correctly prioritize the
contributing issues.
The loss of potential production has been calculated using different
input parameters resulting from damage caused by not recovering injected
fracturing fluids. Exponential decline has been applied for the calculation
of well production rates over time. Results were obtained for administering
the simulation for a finite time period.
An ongoing integrated work considering the causes of partial flowback
is the in situ stress analysis in a reservoir. The purpose is to incorporate the
stress distribution along the horizontal section. More importantly, the frac-
ture closing might be a function of proppant travelling and stress regimes
surrounding the wellbore. Each fracture stage will be classifed based upon
stress regime at a distance from the wellbore using seismic data, and with
image and sonic logs as secondary data.
13.2 Assumptions
Our assumptions from this research are as follows:
• Vertical well with PKN Fracture Geometry

• Formation of a single fracture
• Homogeneous formation and Isotropic formation
• No heat transfer effects
• Water wet rock with initial water saturation equal to irre-
ducible water saturation
• Comparable properties of pad and fracturing fluid
• Exponential decline for producing wells
• Effective injection rates
• Negligible effects of gravity
• Uniform proppant distribution
Flowback of Fracturing Fluids 273
13.3 Upgraded Visualization of Hydraulic Fracturing

13.3.1 Concept
Production from any given well is the result of formation fluids being
transported as a consequence of pressure differential. The path that the
fluid takes within the reservoir can be visualized as a network of very small
diameter tubes. Formation fluids flow within the tubes, which are known
as streamtubes or flowtubes. A sizeable number of these streamtubes exist
within the reservoir. It is important to note that only those streamtubes
that are hydraulically connected to the wellbore contribute to well produc-
tion. The flowrate of well post hydraulic fracturing treatment is dependent
upon the distribution of the different sizes of streamtubes and the velocity
of fluid flowing within them.
Figure 13.1 shows a visualization of streamtubes. It is a Computational
Fluid Dynamics (CFD) Simulation of the flow of water within a 7 mm core
of Belgian sandstone. It was carried out by inCT, Materialize and TotalSim.
The warmer colours in the picture represent higher fluid velocity.
The upgraded visualization of hydraulic fracturing involves looking at
the treatment as the rearrangement of grains. During a hydraulic fractur-
ing procedure, the grains are rearranged so that more of the larger diame-
ter streamtubes are created. In so doing, the area of contact between fluid
and rock per unit volume of formation fluid is reduced. The result is a
narrowing of boundary effects; thus, an increase in the average velocity
of fluids within the tubes is observed. That increase in fluid velocity leads
to an increase in the throughput, which has been realized as an enhanced
production rate post hydraulic fracturing operation.
Low velocity
streamtubes
7 mm
High velocity
streamtubes
7 mm
Figure 13.1 CFD simulation of water within 7 mm Belgian Sandstone [7].

Figure 13.2 Narrowing down of capillary tubes, HiWAY by Schlumberger [8].
This concept has been applied by Schlumberger in the technology

named ‘HiWAY’ for hydraulic fracturing treatments. In this technique,
they attempt to distribute proppants within the fracture with specialized
blending equipment that injects the proppant in pulses so that stable chan-
nels of larger diameters are created, which result in infinite fracture con-
ductivity. Figure 13.2 shows a schematic of the concept.
13.3.2 Results
All the points in Plot 13.1 are the possibilities of the initial production
rate of the well post hydraulic fracturing treatment. The magnitudes of the
increase in the well production rate comply with the fact that the produc-
tion rate of the well will be high in case of higher cumulative volume of
connected tubes (higher reservoir permeability) and vice versa for a fixed
number of set (The term ‘set’ or ‘sets’ here and onwards in this paper means
the number of possibilities of range of radius of the streamtube in the dis-
tribution). Also, for a fixed volume of connected tubes the production rate
450 No increments or decrements

10% incremental unconnected tubes
400
350
Production (Bbls/day)

300 10% decrement in connected tubes
20% decrement in connected tubes
250 30% decrement in connected tubes
40% decrement in connected tubes
200
150
100
6 8 10 12 14 16
Sets
Plot 13.1 Comparison of change in throughput.

of the well would decrease with increasing number of sets. These observa-
tions validate the model. Below are observations from Plot 13.1:
• Production rises by increasing the communication of natu-

ral fractures that result from the hydraulic fractures.
• Production rates increase as the number of sets decrease,
which is logical because a lesser number of sets would corre-
spond to better hydraulic fracturing treatment.
• Production rates increase as cumulative cross sectional area
is enlarged, which is appropriate as a higher cross sectional
area means more stream-tubes being hydraulically con-
nected, and hence higher production.
• The magnitude of increase or decrease in the flowrate is
directly and linearly related to magnitude of the increase or
decrease in the cumulative cross sectional area of hydrauli-
cally connected streamtubes.
• The production rate changes non-linearly with a higher
number of sets.
Using the data of distribution of the streamtubes up to the drain-

age radius of a well, a model can be built to characterize the network of
streamtubes. When using the output of that model as input for this model,
the performance of that well following treatment can be predicted with
reasonable certainty.
In addition, this model can be used to find the representative distribu-
tion of streamtubes of the entire reservoir after the hydraulic fracturing
treatment. It is challenging, as there can be varying scenarios of distribu-
tions that may show the same flowrate of the well. The possibilities can
be converged by using information from other sources including, but not
limited to, seismic, core analysis and logs.
13.4 Reasons for Partial Flowback

13.4.1 Fracture Modelling
The Geertsma-De Klerk (KGD) and Perkins-Kern-Nordgen (PKN) mod-
els are widely accepted two dimensional models for fracture geometry. An
assumption was made wherein the length of the fracture is considerably
more than the fracture height. In such a scenario, the PKN [1] model is
more applicable.
The following equations have been used to calculate fracture length,

width, and the average width with time during the treatment:
1/5
Q 3t 4 E
L 51.46
(1 v 2 ) h 4
1/ 4
(1 v 2 )Q L
Wmax 0.389
E
Wavg 0.628 Wmax
Where,
L is fracture half-length in ft.
Q is injection rate in BPM
t is time in minutes
h is formation thickness in ft.
E is young’s modulus in PSI
v is poisson’s ratio which is unitless
μ is viscosity in cp
13.4.2 Depth of Penetration

It is essential to know the distribution of the fracturing fluid within the
reservoir to properly characterize which part of that fluid flows back.
Those sections of the hydraulic fracture that are close to the wellbore
will experience exposure to higher pressure for longer durations of time
during the treatment.
For a fracture to propagate, signifcant treatment pressures must be
applied from the surface to provide sufficient pressure differential at the tip
of the fracture. Farther from the wellbore, pressure is diminished due to
turbulence and friction pressure. The magnitudes of pressure experienced
in each section of wellbore will progressively increase as the fracture pene-
trates deeper in the reservoir [2].
Taking the effect of both of these aspects into consideration, a plot has
been developed to obtain a visualization of distribution of fracturing fluid
within the reservoir. Since the viscosity of fracturing fluid is much higher
than the reservoir fluids, the relative permeability of the fracturing fluid is
high. As a result, fingering takes place which leads to deeper penetration of
the fracturing fluids into the reservoir. Plot 13.2 was obtained for the depth
of penetration of fluids over the fracture half length.
10
9
Depth of penetration (ft)
8
7
6
5
4
3
2
1
0
0 100 200 300 400 500 600 700 800
Fracture half length (ft)
Plot 13.2 Fracture half-length vs. depth of penetration.
13.4.3 Closing of Fractures

At the conclusion of a hydraulic fracturing treatment, the surface pumps
are shut down, which causes the pressure in the fracture to decline
instantaneously. Due to the existence of stresses within the rock, the
fractures close off. Ideally, the segment of fracture that closes is directly
dependent upon proppant distribution within the fracture [3]. For sim-
plicity, in this model the assumption is to provide uniform distribution
of proppants.
The length of fracture that closes is the section where proppants have
not been placed [4]. A criteria was used to find the region of fracture where
proppants were unable to reach. The criteria is that proppants can pene-
trate in the fracture only until the width of fracture is twice the diameter
of proppant.
In Plot 13.2, a red line has been drawn at a distance of 133.56 feet from
the end of the fracture half length. It means the amount of fracturing fluid
lost on the right side of the red line is potentially unrecoverable. Its vol-
ume has been calculated to be 19.31 bbls, which is 4.29% of the volume of
injected fracturing fluid.
13.4.4 Chemical Interaction of Fracturing Fluids

Once the well is producing, subsequent to hydraulic fracturing treatment,
a comingled flow might occur from reservoir and fracturing fluids. After
a certain amount of time, only the flow of formation fluids is contribut-
ing and dominating. From observation, it is unlikely to be due to all the
fracturing fluids being produced. As production continues, the pressure
transience gets deeper into the reservoir. The pressure differential should be
enough for the fracturing fluid to flow within the reservoir. However, frac-
turing fluids react with reservoir rock and fluids over time, which decrease
their mobility, and thus require higher differential pressure to overcome the
capillary pressure and flow within the reservoir. In reality, applying such
high differential pressures in the reservoir may not be practical, leading to
unproduced injected fracturing fluids within the reservoir. This phenome-
non is a major factor in high volumes of fracturing fluid remaining in the
reservoir.
Calculated Values:
• Additional irrecoverable volume of injected fracturing fluid =

455.49 bbls
• Recoverable volume of injected fracturing fluids = 786.67 bbls
• Total recoverable oil = 3121.86 bbls (Virgin zone) +
3540.02 bbls (Affected zone) = 6661.88 bbls
13.5 Impact of Parameters under Control

The data analysis shows that approximately 45% to 80% of injected frac-
turing fluids are recoverable under certain conditions. From 20% to 55%
of that fluid stays in the reservoir, of which 4.29% is due to the closing of
fractures, and 15.71% to 50.71% is caused by the interaction of fractur-
ing fluid with reservoir fluids and rock. It is very evident that we get a
higher percentage of loss of fracturing fluid due to the interaction of those
injected fluids with reservoir rock and fluids. The fracturing fluid can be
more easily recovered by both decreasing viscosity of fracturing fluid post
treatment and increasing the time for comingled flow.
From sensitivity analysis, 344.94 bbls of incremental oil per 1 cp of
decrease in fracturing fluid viscosity post treatment is achieved as per Plot
13.3. However, 5.26 bbls of incremental oil per 1 hr of increase in time of
comingled flow is achieved as per Plot 13.4. This means the performance of
stimulated well is highly sensitive to the viscosity of fracturing fluids post
treatment. Breaker is one of the additives of fracturing fluid for decreasing
its viscosity post treatment. Hence, the performance of the stimulated well
is dependent on the performance of breaker.
5000
Additional irrecoverable volume (bbls)
4500
4000
3500
3000
2500
2000
1500
1000
500
0
0 5 10 15 20 25 30
(cp)
Plot 13.3 μ vs. recoverable oil volume.
10000
Additional irrecoverable volume (bbls)
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
0 200 400 600 800 1000 1200 1400
t (hrs)
Plot 13.4 Time vs. additional irrecoverable fracturing fluid volume.
13.6 Loss in Incremental Oil Production

The fracturing fluid that stays back in the reservoir acts as a restriction to
the reservoir fluids flowing into the fracture. As a result, there is skin asso-
ciated with fluids not flowing back. That skin can cause a decrease in pro-
duction rate over time and therefore lessen the cumulative oil production.
900
800
700
600
qo (bbls/day)
500
400
300
200
100
0
0 100 200 300 400 500 600 700 800 900 1000
Time, post comingled flow (days)
Without skin due to fracturing fluid With skin due to fracturing fluid
Plot 13.5 Decline curve.
In the case of Plot 13.5, an incremental production of 36,676 bbls of oil

could be achieved over the duration of 900 days if the formation damage
due to unproduced fracturing fluid had been avoided [5].
13.7 Conclusions
We conclude that by combining the streamtube model approach with
the knowledge of geomechanics of the field, the well performance after
hydraulic treatment can be predicted with reasonable accuracy.
Partial flowback of fracturing fluids is primarily a result of:
• High pressure injection of viscous fracturing fluid leading to

larger depths of penetration where it becomes very difficult
to get enough pressure differential to overcome the capillary
pressure of fracturing fluids. Furthermore, the interaction
of fracturing fluid with reservoir rock and fluids substan-
tially reduces the mobility of fracturing fluid. The situation
is aggravated by high viscosity of fracturing fluids post treat-
ment. This is a major contributor in the loss of fracturing fluid
in the reservoir and its impact on overall well performance.
• The closing of fractures following treatment causes the
injected fluid to be placed in sections of the reservoir which
are relatively hard to access. This is a minor contributor in
the loss of fracturing fluid in the reservoir and its impact on
overall well performance.
In this model, 45% to 80% of injected fluid can potentially be recovered.

The skin caused by unproduced fracturing fluids can prevent us from
producing tens of thousands of barrels of oil over the life of the well.
Wells with lower skin values are much more sensitive to the skin due to
unproduced injected fracturing fluids.
13.8 Limitations
All the estimates of the magnitude of loss of fracturing fluid are low when
considering only one fracture. However, in reality, the estimates would
increase substantially due to the complex geometry of fracture where a
single fracture does not exist.
In actuality, the distribution of the proppants within the fracture is not
uniform. For that reason, the fracture may close off in the early or middle
region, making the rest of the length of the fracture highly ineffective. In
that case the loss of fracturing fluid due to closing of fractures may be much
more than what has been previously calculated. Usually, the concentration
of proppants is greater in the lower parts of a fracture due to the effect of
gravity.
One would encounter losses attributable to heterogeneity, natural frac-
tures and faults in real life which have not been taken into account in this
model. That could be one of the contributing factors for further losses of
injected fracturing fluid impacting the observation of 2–26% flowback [6].
References
1. T.K. Perkins and L.R. Kern, SPE-89-PA widths of hydraulic fractures. J Pet
Technol 13(9), 937–949 (1961).
2. T. Martin and P. Valko, Hydraulic Fracture Design for Pressure Enhancement,
Modern Fracturing Enhancing Natural Gas Production, Houston, Texas.
3. R.A. Woodroof, M. Asadi, and M.N. Warren, SPE-82221-MS Monitoring
fracturing fluid flowback and optimizing fracturing fluid cleanup using
chemical Frac Tracer, SPE European Formation Damage Conference, 13–14
May, The Hague, Netherlands (2003).
4. http://www.corelab.com/ps/hydraulic-fracture-design
5. http://www.fekete.com/SAN/WebHelp/FeketeHarmony/Harmony_
WebHelp/Content/HTML_Files/Reference_Material/Analysis_Method_
Theory/Blasingame_Theory.htm
6. Q. Zhou, R. Dilmore, A.N. Kleit, and J. Yilin, SPE-173364-MS Evaluating
Fracturing Fluid Flowback in Marcellus Using Data Mining Technologies.
7. http://biomedical.materialise.com/cases/cfd-simulation-water-flow-
sandstone-using-micro-ct
8. http://www.slb.com/services/completions/stimulation/sandstone/hiway_
channel_fracturing.aspx?t=2
Appendix - A
Calculations for Upgraded Visualization:
R2 p
Vavg
8 x
Where,
Vavg is average velocity in m/s
R is radius in m
μ is viscosity in kg/m/s
P is change of pressure in kg/m2
x is change in distance in m
Calculations for Damage due to Fracturing Fluids:
k r
S frac S 1 ln s
ks rw
Sfrac is skin due to unproduced fracturing fluid and its unitless

S is total skin due to everything but unproduced fracturing fluid
and its unitless
k is virgin reservoirs permeability in mD
ks is damaged reservoir permeability in mD
rs is radius of damaged region in ft
rw is wellbore radius in ft
Calculations for Flowrate of Well:
k0
( Pwf Pi ) h
0
q0
1688 Ct rw2
162.6 BO log (0.868 S )
t
Where,
q is flowrate in bbl/day
Pwf is wellbore flowing pressure in psi
Pi is the initial reservoir pressure in psi
k0 is relative oil permeability and its unitless
μ0 is oil viscosity in cp
h is thickness of formation in ft
Bo is the formation volume factor in rb/stb
ϕ is porosity and it’s unitless
Ct is the total compressibility of the reservoir in psi–1
rw is wellbore radius in ft
t is time in hrs
S is skin and its unitless
Calculations for Change in Flowrate of Well with Time:
qt = qi * exp(–ta)
Where,
qi is initial production rate in stb/day
qt is production rate at time t in stb/day
t is time in days
a is decline factor and its unitless
14
Assessing the Groundwater
Contamination Potential from a Well
in a Hydraulic Fracturing Operation
Nima Jabbari1*, Fred Aminzadeh2 and Felipe P. J. de Barros1
1
University of Southern California, Sonny Astani Department of Civil and
Environmental Engineering, Los Angeles, CA, USA
2
University of Southern California, Mork Family Depart of Chemical Engineering
and Material Science, Los Angeles, CA, USA
Abstract
The introduction of hydraulic fracturing as a new technology for shale gas produc-
tion from low-permeability geologic formations has revitalized natural gas as an
abundant, economic, and cleaner source of energy. However, hydraulic fracturing
has raised several environmental concerns, from air pollution and drinking water
contamination to on-site accidents and risks associated with potential seismic
activities induced by subsurface fluid injection. Some of the potential incidents
and accidents could be mitigated by adding more precautions to the different
stages of an operation. Integrity of the injection well is one of the most important
factors to be considered when dealing with water resources contamination. This
work focuses on groundwater contamination potential in a leakage accident from
the casing during hydraulic fracturing. A numerical flow and transport modeling
approach is adopted to explore and quantify the associated risk of groundwater
contamination under various hypothetical scenarios. The goal is to investigate a
few cases with different geological and operational parameters and, also, to eval-
uate the importance of the well integrity, especially along the interval in which
well casing is inside an aquifer. The results of the work show that if the well integ-
rity is compromised, the groundwater will be contaminated by the chemicals in a
relatively short duration. It is also shown that the result of the stochastic work is
dependent on how well one knows the failure scenario. The appropriate choice of
*Corresponding author: jabbari@usc.edu

285
the probability distribution function for the leakage point is of high importance in
simulations such as these.
Keywords: Hydraulic fracturing, well integrity, groundwater, environmental
14.1 Introduction
Hydraulic Fracturing is an old reservoir stimulation method that has been
acknowledged in petroleum industry since fifty years ago. The technology
aims at increasing the oil and gas production through a sequence of pro-
cesses including drilling vertical or/and horizontal wells, injecting highly
pressurized fluid, cracking the matrix formation, and creating fissures,
which, in turn, increases permeability values of the reservoir [1]. As stated
by Daneshy [2], hydraulic fracturing boosts the production through chang-
ing the flow pattern regime toward the well from radial to linear. At the
same time, horizontal wellbores help in obtaining a larger reservoir-well
contact which, in turn, results in higher production numbers. A combina-
tion of advanced horizontal drilling technologies and hydraulic fracturing
has become very popular lately in producing oil and natural gas from shale
and other tight formations [3, 4]. Despite the popularity and the common
use, this technique still has imperfections that raise questions on the safety
and environmental aspects of the process [5, 6]. In addition to the safety,
environmental problems are also important since the long-term effects of
the hydraulic fracturing activities are not very well understood yet [7]. This
fact creates lots of questions for the environmentalist and the public. The
environmental impacts can be viewed in three categories: air, water, and
induced seismicity.
Air pollution, in general, is linked to several different activities car-
ried out during natural gas development including drilling and fracturing
[8–10].
Subsurface injection operations normally generate low magnitude
(smaller than 2 in Richter scale) seismic events which are termed as
micro-earthquakes or microseismic [11]. In a few cases of hydrofracking
jobs, seismic events have been felt by the nearby residents and make the
operator stop the injection. As an example, in 2011 two seismic events
(2.3 and 1.5 magnitude in Richter scale) were recorded near Blackpool,
Lancashire in the United Kingdom which made the operator halt its frack-
ing operation in the nearby Bowland Shale formation [12]. Generally,
the risk of generation of serious earthquakes as a result of hydraulic frac-
turing is low when compared to the deep-well injection process, which
Groundwater Contamination Potential 287
shows higher probabilities of observing larger seismic events [11, 13, 14].
Historically, the oldest injection-induced seismic events are those of the
Rocky Mountain Arsenal waste injection site in Denver, Colorado [15].
The most recent events were generated in Youngstown, Ohio [16] and cen-
tral Oklahoma [11], both in 2011. The former is also known to be the larg-
est injection-induced event with a 5.6 magnitude [11].
Chemicals used in a typical fracturing job vary and range from benign
(e.g. guar gum), to toxic (e.g. Tetramethylammonium chloride), to very
toxic materials (e.g. Kathon which is a biocide) [17]. Injecting these
chemicals under the ground, if not practiced under safe and sound condi-
tions, can an increase in the risk of having ground or surface water bodies
contaminated. The importance of a solid well system with full integrity is
pronounced here. Handling and management of returned fluid is another
aspect of the fracturing operation which can potentially be a threat to the
health of drinking water resources [18–21].
The well integrity issue is a quite well-known phenomenon in injec-
tion operations such as carbon sequestration and deep waste injection
[22–27]. If the well integrity is not maintained, groundwater can be a tar-
get for the contaminants originating from the initial injectant or found in
the returned fluid [28, 29]. From the human health point of view, ground-
water pollution is critical, more specifically for regions with water short-
age and high demand for groundwater tables [3, 9]. Underground failure
and upward fluid migration has been listed as one of the important risk
pathways to the shallow groundwater [30]. Damages in well integrity
and leakage of fracking fluid into adjacent formation, poor design of the
fracturing job and fracture growth in overburdened formations, reactiva-
tion of faults as permeable pathways for the fluid, and fluid flow through
abandoned wells are along risk pathways discussed by the United States
Environmental Protection Agency (USEPA) [3, 18]. However, among the
discussed risk pathways, the chance of upward fluid migration from deep
formations to the shallow aquifers is reduced as a result of a number of
factors. Absence of a hydraulic connectivity between the aquifer and shale
formation because of geometrical features and physical barriers, makes it
unlikely for reservoir fluid to migrate upward, mostly, as stated by several
authors. The separation between a shallow aquifer and the shale forma-
tion is too large to make the upward fluid migration into the aquifer hap-
pen in a short time-scale [4, 31, 32]. Furthermore, shale layers are usually
capped with tight formations acting as natural barriers [33]. Artificially
created fractures may penetrate into the overburden formation, however,
the distance to which they can propagate upward is limited and the frac-
ture is very unlikely to pass through the overburden, reach the aquifer, and
connect it to the shale layer [31, 34, 35]. In addition, when the production
starts fluid flows toward the horizontal section of the well, upward migra-
tion would be less probable to take place [32, 36]. Meanwhile, hydraulic
fracturing operations near conductive faults might change the story and,
as declared by Gassiat, Gleeson, Lefebvre, and McKenzie [37], under some
circumstances, reservoir fluids could reach the aquifer through the per-
meable fault, although, due to the slow rate of movement, the mechanism
takes place in a long time (thousands of years).
14.2 Risk Pathways to the Shallow Groundwater

Risk pathways to shallow aquifers can be divided into two major catego-
ries: (1) over ground accidental spills from the pits used for temporary
storage of the returned fluid or from accidents while transporting fluids
or stemming from drilling activities (2) underground failure and upward
fluid migration [30].
Upward fluid migration to the aquifer can take place through differ-
ent pathways. USEPA suggests that some of the possible scenarios for the
upward flow migration are as follows:
1. Damages in well integrity as a consequence of corrosion or

improper cementing job. Leakage of fracking fluid into the
formation can be a possible outcome of this scenario.
2. Poor design of the fracking job leading to fracture growth in
overburden formations and providing hydraulic connections
between the shale gas reservoir and aquifers above. In some
cases, in presence of hydrocarbon zones above the shale layer,
the situation becomes more critical as the fluid is capable of
carrying hydrocarbon while traveling to shallower depths.
3. Fault reactivation resulted from the fracking process and
provided permeable pathways for the fluid.
4. Fracturing the overburden formation and creating pathways
into wells in the vicinity of the fracturing well (abandoned
wells in most cases). In this situation, fluid might find its way
through the aquifer using the adjacent wells and through
issues in well integrity [3, 18].
Well integrity issues described in scenario 1 are further divided into two
categories [28]: Behind the casing movement of the fluid toward the well
head in which fluid finds the space between cementing and casing as a
possible pathway (annular flow) and; leakage from the well in a radial pat-
tern so that the fluid can reach the nearby formation and the chemicals can
undergo different transport mechanisms (leak flow).
King [1] has also named surface casing rupture as one of the poten-
tial leakage scenarios in a fracturing operation. Real-time pressure mon-
itoring helps increase the chance of leakage discovery and, in a way, the
pressure drop followed from a major leakage in pipe enables the opera-
tors to discover the leakage in five minutes [17]. However, when there is
a smaller leakage event happening, the failure remains undetected during
the hydraulic fracturing operation [17]. It is therefore, important to closely
investigate the aftermath of a leakage event, particularly when the rupture
is within the aquifer or at a point close to the aquifer.
On the same subject, a probability bound analysis study has been done
on risk of water pollution in Marcellus shale [30]. The study concluded
that, among other risk pathways such as spills from on-site activities and
wastewater disposal, well casing failure is an important contributor to the
contamination risk where it shows lower than 0.01 m3 and 9 m3 contami-
nation volumes for best-case 50th percentile and worst-case 50th percen-
tile respectively [30]. Furthermore, a study by Browning and Smith [24] on
deep well injection in Louisiana, Nebraska, Michigan, and Pennsylvania
suggests malfunctioning casings and tubings as two reasons for gaining
failure in mechanical integrity tests. According to the same study, out of
10,000 wells that had undergone the integrity test, 50% of them fail to pass
with casing failure as the main contributor to the failure rate (responsible
for 45% to 85% of failure). Among the failed tests due to casing problems,
22% showed failure in the only protective layer so that the injection waste
was able to leak into the near formation. Although the data from this study
might not be directly applicable to fracking wells because of the difference
between waste injection and fracking fluid injection, it still gives ideas on
chance of possible failures of casing in a fracking well [24].
14.3 Problem Statement

This study is focused on the casing failure during the injection phase of
hydraulic fracturing followed by a leakage flow into the aquifer. The fate of
the contamination plume is investigated through application of a numeri-
cal solver for flow and transport equations. Various scenarios in this study
are devised by altering hydrogeological parameters such as a homogeneous
anisotropic permeability field and porosity, as well as operational param-
eters (i.e. leakage rate and depth of the leakage point). It is noteworthy to
state that the failure scenario of concern in this study falls into the type
of incidents classified as low probability. The small probability is mainly
attributed to the pressure monitoring during the operations and the fact
that the leakage incident normally causes an observable pressure drop [17].
However, the leakage phenomenon is still worth studying, as the risk asso-
ciated could be significant. One should note that risk, here, is defined as
the multiplication of probability of an event by the consequence, and the
aftermath of such a leakage scenario near or into the aquifer can be quite
remarkable. With the low probability and high consequence, the hypothet-
ical case studied here is considered to be a “rare” event. To investigate such
a scenario a 2D numerical simulation is used and the effects of different
parameters on the results are studied. The parameters of interest in this
study are selected to be the concentration magnitude of the chemical and
the arrival time at the well location.
14.4 Mathematical Formulation

The chemical transport within the aquifer is modeled using a 2D flow field.
The permeability field is set to be anisotropic homogenous and the poros-
ity is kept constant. The transport equation is as follows [38]:
(c )
(c u D c) q (14.1)
t
in which c is the concentration of the chemical in the fluid phase, φ is

the medium’s effective porosity, ρ is the water density, u is the specific dis-
charge, and q is sink or source. The concentration varies in space x = (x,z)
and time t. Darcy’s law for a fully saturated system is given by:
1
u k( p g z) (14.2)
with k denoting an absolute permeability tensor and p, u representing pres-

sure, and viscosity of water, respectively. The two dimensional diffusion
and mechanical dispersion tensor is defined as follows:
D= dmI |u| dl E(u ) dt E (u ) (14.3)

where dm is the molecular diffusion coefficient and dl and dt are the longi-
tudinal (parallel to the flow) and transverse (perpendicular to the flow in
vertical direction) mechanical dispersion coefficients. The Euclidian norm
of the specific discharge is given by:
|u| u12 u22 (14.4)
The orthogonal projections in equation 3 are defined as:
1 u12 u1u2
E(u) u)
and E (u ) I E(u (14.5)
|u|2 u2u1 u22
As for the boundary conditions used for the system of equations, we

consider the Neumann type in which mass flux is prescribed on the bound-
ary (Γ):
ρu · υ = g2 on Γ (14.6)
where u is the unit normal outward to Γ and ρu·υ gives the projection of
the flux vector on a unit normal vector of the boundary. In equation 14.6,
g2 stands for the prescribed flux on the boundary. The initial condition is
given as:
p(x,0) = p0(x), x Ω (14.7)
with p0(x) being the hydrostatic pressure and Ω denoting the entire domain
of interest.
14.5 Hypothetical Case Description

and the Numerical Method
Figure 14.1 shows the leakage as a point source on the well casing. The layers
are stacked on top of each other with the 20 meter under-saturated layer on
top. The aquifer is also 20 meters thick and is separated from a much thicker
sand formation through a 2 meter thick impervious layer. The 800 meter
thick sand formation is playing a role as the overburden of the shallow shale
Injection Monitoring
well well
Vadose zone
Aquifer
Impervious layer
Sand layer
Shale reservoir
Figure 14.1 Schematic representation of failure in vertical pipe and contamination plume
in the aquifer.
layer extending from a depth of 840 to 900 meters. In order to investigate the
effect of groundwater flow movement on the contamination, a monitoring
well is placed 100 meters away from the injection well. The numbers here are
adapted from average properties of shallow shale plays reported by the U.S.
Energy Information Administration [39].
In the numerical analysis, the aquifer and the impervious layer are
modeled using a two dimensional ECLIPSE grid with dimensions of 210
m × 20 m. ECLIPSE is a simulation software for subsurface multi-phase
and multi-component flow and transport equations (see equations 14.1
through 14.7) [40]. The stochastic simulations are then conducted in order
to find the probability of exceedance from the threshold in the monitor-
ing well. The chemicals used in the fracturing fluid are modeled using a
passive tracer and the results are reflected in measurements of one quan-
tity of interest which is the concentration level at the monitoring well. The
Monte-Carlo simulation is adapted to quantify the probability of crossing
the specified concentration threshold.
The fracturing operation is assumed to be carried out in 6 stages, each
2.5 hours long with 10 hours of relaxation time, with the injection rate of
11,500 m3/d (50 bbl/ min). It is also presumed that the chemical concen-
tration is equal to two weight percent of the fracturing slurry and the sim-
ulation time is 500 years. The parameters of concern for this study include
porosity and anisotropic permeability of the aquifer, as well as the location
of the leakage point with respect to the aquifer top and the leakage rate.
Each parameter is randomly drawn from different probability distribution
functions. Table 14.1 shows the parameters which are constant over the
course of simulation.
Stochastic simulations are conducted in order to evaluate the sensitivity
of the model to different parameters and also to determine the probability
of passing the contaminant threshold in the aquifer. The threshold is set to
be 1 mg/l. Figure 14.2 shows the workflow used to incorporate randomness
in the model. In here, the Monte-Carlo simulation is used to calculate the
probabilities. The number of Monte-Carlo simulations and the probability
Table 14.1 Parameters of the model.

Parameter Value
Aquifer Velocity 1.1 cm/d
Pressure Gradient in Aquifer 0.02 m/m
Tracer molecular diffusion coefficient 10-5 cm2/s

Tracer longitudinal dispersivity 0.5 m
Tracer transverse dispersivity 0.05 m
Tracer initial concentration 22 kg/m3
Value randomly drawn from specified

probability density function
Simulation run
Saving the results
Figure 14.2 Random simulations workflow.

Table 14.2 Parameters of the random simulations.

Parameter Distribution Mean S.D.
Leakage rate (m3/d) Uniform 1.05 0.55
Leakage depth (m) Uniform 9.5 4.91
Poisson’s 18 18
Sand permeability (mD) Log-normal 103.2 23.4
Sand porosity Truncated Normal 0.175 0.04
Monte Carlo simulations number 1000
density functions (PDF) used for selecting the parameters are shown in
Table 14.2. Log-normal and truncated normal distributions are chosen in
a way to pick aquifer permeability and porosity values from ranges 150 to
270 mD and 0.10 to 0.30, respectively. Selected PDFs for permeability and
porosity are in line with the petroleum engineering convention. The ranges
chosen for permeability and porosity represent a permeable aquifer with
an average permeability of 220 mD and an average porosity of 0.2. Also, the
vertical permeability is selected as one-tenth of the horizontal permeability
by convention. For the leakage location, two different PDFs are chosen. The
first one is a uniform one extending from one to eighteen meters inside the
aquifer with the assumption of absence of any information on the statistics
of leakage on the wall of a vertical well. Whereas, the second is set to be a
Poisson’s distribution in which the probabilities increase as the depth takes
larger values in a manner where the maximum probability is gained for the
depth equal to the aquifer bottom. The main reason for working with two
different PDFs assigned for the leakage location is to assess the changes in
results when one takes into account the fact that exertion of more stresses
on the well casing in higher depths increases the probability of failure.

The CDF of the normalized concentration at the monitoring well are shown
in Figure 14.3 and Figure 14.4 for the cases of uniform and Poisson’s PDFs,
respectively. For the first time-step shown (i.e. 0.27 years), the concentra-
tion plume is not yet reaching the monitoring well. At a time of 2.74 years,
we can observe concentration values in the monitoring well for both cases
of leakage PDF. The difference is that when working with the Poisson’s PDF,
0.27 yr 2.74 yr
1 1
CDF
CDF
0.5 0.5
00 1 2 3 4 5 0
1 2 3
C/C 0 x 10–5 C/C0 x 10–4
27.4 yr 100 yr
1 1
Concentration
Threshold
CDF
CDF
0.5 0.5
0 0
0 2 4 6 8 0 1 2 3 4 5
C/C0 x 10–3 C/C0 x 10–5
Figure 14.3 Cumulative distribution functions (CDFs) of concentration value in the

monitoring well – uniform distribution for the leakage location.
0.27 yr 2.74 yr
1 1
CDF
CDF
0.5 0.5
0 0
0 1 2 3 4 5 6 8
–5
C/C0 x 10 C/C0 x 10–4
27.4 yr 100 yr
1 1
Concentration
Threshold
CDF
CDF
0.5 0.5
00 2 4 0 5
0
6 8 0 1 2 3 4
C/C x 10–3 C/C0 x 10–5
Figure 14.4 Cumulative distribution functions (CDFs) of concentration value in the

monitoring well – poisson’s distribution for the leakage location.
the concentration values show higher numbers as opposed to the case with
uniform PDF. The probability of exceedance is also larger when using the
Poisson’s PDF. The same holds for the next time-step shown (27.4 years),
with a difference that the concentrations are roughly ten folds larger. In
year 100 and beyond, the concentration values in the monitoring well
show small numbers, mainly because of the mass transport from the open
Table 14.3 Probabilities calculated for different scenarios.

P(C≥Cth) Leakage Point PDF
Time (yr) Uniform Poisson’s
0.27 0 0
2.74 0.12 0.15
27.4 0.95 0.98
100 0 0
1
0.9
0.8
0.7
0.6
F(time)
0.5
0.4
0.3
0.2
0.1
0
2 4 6 8 10 12 14 16
(a) Time (y)
1
0.9
0.8
0.7
0.6
F(time)
0.5
0.4
0.3
0.2
0.1
0
0 5 10 15
(b) Time (y)
Figure 14.5 Arrival time CDF – Uniform (a) and Poisson’s (b) PDFs for the leakage depth.
Table 14.4 Statistics of arrival time (years).

Leakage point PDF Mean Variance Min. Max. Range
Uniform 6.73 5.82 2.22 15.74 13.52
Poisson’s 5.57 2.98 1.7 14.74 13.04
boundary of the model. The calculated probabilities of passing the thresh-

old value of the chemical scenarios are also listed in Table 14.3.
In general, the probabilities take their maximum values at year 27.4 and
show reduction from this time forward. Also, the case with Poisson’s PDF
for the leakage depth shows slightly more critical results.
The problem can also be viewed from the time perspective. The time
of interest here (arrival time) is defined as the first moment after which
the concentration level in the monitoring well goes beyond the specified
threshold (1 mg/l). Figure 14.5 illustrates the CDF results for two cases of
the leakage scenario (i.e. Uniform and Poisson’s PDF for the leakage depth)
and Table 14.4 displays the statistics for different stochastic scenarios.
On average, it takes between 6.73 and 5.57 years for the contaminant to
go above its threshold in the monitoring well in this geological setting. It
is also noticeable how replacing the uniform PDF by the Poisson’s one can
affect the results.
14.7 Conclusion
Hydraulic fracturing operations are not yet without deficiencies and can
potentially be harmful to the environment if not practiced in a sound and
safe manner. This work focuses on possible groundwater contamination
issues stemmed from hypothetical failure and leakage in the well casing
when the injection takes place. Permeability and porosity of the aquifer,
leakage depth, and leakage rate are the parameters considered in this study.
Higher importance is given to the leakage depth by assigning two different
PDFs for this parameter and comparing the results. The results are reported
as the concentration value of a tracer as well as the arrival-time of a specific
threshold in the monitoring well. It can be inferred from the results that,
under the special scenario and specific geological settings introduced in
this study, the concentration in the monitoring well can exceed the thresh-
old limit. It is also shown that using Poisson’s PDF results in more critical
results as the concentrations reported increase and the average arrival-time
decreases. Keeping the importance of leakage location in mind, one can
now impede failure scenarios by placing stronger obstacles in front of the

fluid when the injection is happening. In general, some of the hydraulic
fracturing drawbacks can be mitigated by setting stricter regulations and
improving the design and well integrity. It is noteworthy to mention that
a comprehensive study in a three-dimensional (3D) domain is required to
give more solid ideas on the scenarios. 3D models help in achieving more
realistic results, however, 2D simulations can still be helpful in understand-
ing the general concepts. Also, one step forward in such studies could be
increasing the uncertainty dimension by taking into account even more sto-
chastic parameters, including the ones related to the physics of the well such
as number of breaches on the well location. Those analyses require more
advanced probabilistic methods to overcome the burden of the computation.
References
1. G.E. King, Hydraulic Fracturing 101: What Every Representative,
Environmentalist, Regulator, Reporter, Investor, University Researcher, Neighbor
and Engineer Should Know about Estimating Frac Risk and Improving Frac
Performance in Unconventional Gas and Oil Wells, SPE Hydraulic Fracturing
Technology Conference, The Woodlands, TX, pp. 1–80 (2012).
2. A. Daneshy, Hydraulic fracturing to improve production. SPE Tech. 101,
14–17 (1990).
3. USEPA (US Environmental Protection Agency), Study of the Potential
Impacts of Hydraulic Fracturing on Drinking Water Resources, (2012).
4. J.D. Arthur, B.K. Bohm, B.J. Coughlin, M.A. Layne, and A L L Consulting,
Evaluating Implications of Hydraulic Fracturing in Shale Gas Reservoirs,
SPE Americas E&P Environmental and Safety Conference, pp. 23–25 (March
2009).
5. J. Heinecke, N. Jabbari, and N. Meshkati, The role of human factors consider-
ations and safety culture in the safety of hydraulic fracturing (fracking). JSEE
2 (2), 130–151 (September 1, 2014). doi:10.7569/JSEE.2014.629509.
6. N. Jabbari, C. Ashayeri, and N. Meshkati, Leading Safety, Health, and
Environmental Indicators in Hydraulic Fracturing, SPE Western Regional
Meeting, Garden Grove, CA (2015).
7. T. Engelder, R.W. Howarth, and A. Ingraffea, Should fracking stop? Nature
477 (7364), 271–275 (2011).
8. L.M. McKenzie, R.Z. Witter, L.S. Newman, and J.L. Adgate, Human health
risk assessment of air emissions from development of unconventional natu-
ral gas resources. Sci. Total Environ. 424, 79–87 (2012).
9. D.M. Kargbo, R.G. Wilhelm, and D.J. Campbell, Natural gas plays in the
Marcellus shale: challenges and potential opportunities. Environ. Sci. Technol.
44 (15), 5679–5684 (August 1, 2010). doi:10.1021/es903811p.
10. C.W. Moore, B. Zielinska, G. Pétron, and R.B. Jackson, Air impacts of
increased natural gas acquisition, processing, and use: a critical review.
Environ. Sci. Technol. 48 (15), 8349–8359 (March 18, 2014). doi:10.1021/
es4053472.
11. W.L. Ellsworth, Injection-induced earthquakes. Science 341 (6142), 1225942
(2013).
12. C.J. De Pater, and S. Baisch, Geomechanical Study of Bowland Shale
Seismicity, Synthesis Report, 57 (2011).
13. R. Walker, F. Aminzadeh, and A. Tiwari, Correlation between Induced
Seismic Events and Hydraulic Fracturing Activities in California, AGU Fall
Meeting Abstracts, Volume 1, p. 05 (December 2014).
14. T. Göbel, A comparison of seismicity rates and fluid-injection operations
in Oklahoma and California: implications for crustal stresses. The Leading
Edge 34 (6), 640–648 (2015).
15. J. Healy, and W. Rubey, The denver earthquakes. Science 161 (3848), 1301–
1310 (1968).
16. ODNR (Ohio Department of Natural resources), Preliminary Report on the
Northstar 1 Class II Injection Well and the Seismic Events in the Youngstown,
Ohio, Area, (2012).
17. B.C. Gordalla, U. Ewers, and F.H. Frimmel, Hydraulic fracturing: a toxico-
logical threat for groundwater and drinking-water? Environ. Earth Sci. 70
(8), 3875–3893 (August 13, 2013). doi:10.1007/s12665-013-2672-9.
18. USEPA (US Environmental Protection Agency), Assessment of the Potential
Impacts of Hydraulic Fracturing for Oil and Gas on Drinking Water
Resources, (2015).
19. R. Hammer, and J. VanBriesen, In fracking’s wake: new rules are needed to
protect our health and environment from contaminated wastewater. Nat.
Resour. Defense Coun. 11 (May 2012).
20. Y.R. Glazer, J.B. Kjellsson, K.T. Sanders, and M.E. Webber, Potential for using
energy from flared gas for on-site hydraulic fracturing wastewater treatment
in Texas. Environ. Sci. Technol. Lett. 1 (7), 300–304 (2014). doi:10.1021/
ez500129a.
21. J.M. VanBriesen, J.M. Wilson, and Y. Wang, Management of Produced Water
in Pennsylvania: 2010-2012, Shale Energy Engineering, 107–113 (2014).
22. J.M. Nordbotten, M.A. Celia, S. Bachu, and H.K. Dahle, Semianalytical solu-
tion for CO2 leakage through an abandoned well. Environ. Sci. Technol. 39
(2), 602–611 (2005). doi:10.1021/es035338i.
23. C.F. Tsang, J. Birkholzer, and J. Rutqvist, A comparative review of hydro-
logic issues involved in geologic storage of CO2 and injection disposal of liq-
uid waste. Environ. Geol. 54 (8), 1723–1737 (August 4, 2007). doi:10.1007/
s00254-007-0949-6.
24. L.A. Browning, and J.D. Smith, Analysis of the Rate of and Reasons for
Injection Well Mechanical Integrity Test Failure, SPE/EPA Exploration and
Production Environmental Conference, San Antonio, TX (1993).
25. R.J. Davies, S. Almond, R.S. Ward, R.B. Jackson, C. Adams, F. Worrall,
L.G. Herringshaw, J.G. Gluyas, and M.A. Whitehead, Oil and gas wells
and their integrity: implications for shale and unconventional resource
exploitation. Mar. Petrol. Geol. 56, 239–254 (March 2014). doi:10.1016/j.
marpetgeo.2014.03.001.
26. G.E. King, and D.E. King, Environmental risk arising from well-construction
failure-differences between barrier and well failure and estimates of failure
frequency across common well types locations. SPE Prod. Oper. 28 (4), 323–
344 (October 2013).
27. T. Watson, and S. Bachu, Evaluation of the potential for gas and CO2 leakage
along wellbores. SPE Drill. Completion. 24 (1), 115–126 (March 2009).
28. M.D. Holloway, and O. Rudd, Fracking: The Operations and Environmental
Consequences of Hydraulic Fracturing, John Wiley & Sons, Beverly, MA
(2013).
29. N. Jabbari, F. Aminzadeh, F. de Barros, and B. Jafarpour, Hydraulic Fracturing
and Potential Groundwater Contamination Risk. Presented at the 248th
American Chemical Society National Meeting & Exposition, Paper Number:
315, San Francisco, CA, 10–14 (2014).
30. D.J. Rozell, and S.J. Reaven, Water pollution risk associated with natural
gas extraction from the Marcellus shale. Risk Anal. 32 (8), 1382–93 (August
2012). doi:10.1111/j.1539-6924.2011.01757.x.
31. M. Zoback, S. Kitasei, and B. Copithorne, Addressing the Environmental
Risks from Shale Gas Development, Worldwatch Institute, Washington, DC
(2010).
32. R.W. Howarth, A. Ingraffea, and T. Engelder, Natural gas: should fracking
stop? Nature 477 (7364), 271–275 (September 15, 2011). doi:10.1038/477271a.
33. J.D. Arthur, B.K. Bohm, and M.A. Layne, Hydraulic Fracturing Considerations
for Natural Gas Wells of the Marcellus Shale, The Ground Water Protection
Council Annual Forum, Cincinnati, OH (2008).
34. M. Fisher, and N. Warpinski, Hydraulic-fracture-height growth: real data.
SPE Prod. Oper. 27 (1), 8–19 (November 2012).
35. S.A. Flewelling, M.P. Tymchak, and N. Warpinski, Hydraulic fracture height
limits and fault interactions in tight oil and gas formations. Geophys. Res.
Lett. 40 (14), 3602–3606 (July 28, 2013). doi:10.1002/grl.50707.
36. A. Kissinger, R. Helmig, A. Ebigbo, H. Class, T. Lange, M. Sauter, M. Heitfeld,
J. Klünker, and W. Jahnke, Hydraulic fracturing in unconventional gas reser-
voirs: risks in the geological system, part 2. Environ. Earth Sci. 70 (8), 3855–
3873 (June 22, 2013). doi:10.1007/ s12665-013-2578-6.
37. C. Gassiat, T. Gleeson, R. Lefebvre, and J. McKenzie, Hydraulic fracturing
in faulted sedimentary basins: numerical simulation of potential contami-
nation of shallow aquifers over long time scales. Water Resour. Res. 49 (2)
(December 12, 2013). doi:10.1002/2013WR014287.
38. Z. Chen, G. Huan, and Y. Ma, Computational Methods for Multiphase Flows
in Porous Media (Computational Science and Engineering), Society for
Industrial and Applied Mathematic, Philadelphia, PA (2006).
39. USEIA (U.S. Energy Information Administration), Review of Emerging
Resources: U.S. Shale Gas and Shale Oil Plays, US Energy Information
Administration, (2011).
40. Schlumberger, ECLIPSE: Reservoir Simulation Software, Reference Manual,
(2010).
Index
Additives Biot modulus, 41

common fracturing, 58–60 Boiling point, 74
manufactured ceramics proppants, Bottomhole pressure (BHP), 11, 39,
65–66 146, 184–185
percentages of commonly used, Bowland Shale formation, 286
60–66 Bradford field, 87–88
proppants, 61–63 Breaker solution, 59
resin coated proppant, 65 Bulk modulus, 41
silica sand, 63–65 B-value analysis, 25–26, 106, 160, 188,
type and amount, 58 191–193
Air pollution, 18, 68, 244, 286
Alberta Geologic Survey, 168 California Council on Science and
Alfred P. Murrah Federal Building, 60 Technology (CCST) report, 5,
American Petroleum Institute (API), 14–15, 18
15 California Well stimulation
Ammonium nitrate, 60 technologies, 14–15
Anisotropy, 200 Carbon dioxide (CO2) emissions, 8
Anti-fracturing, 54 Care, in neural network design, 112
Appearance, 75 Cartesian coordinate system, 85
Armstrong PA, leakage scenario, Cauchy’s formula, 46
88–91 CDFs. See Cumulative distribution
Artificial intelligence, 109–110 functions (CDFs)
Artificial neural nets, 110, 111, 114 Chemicals
Auto-ignition temperature, 75 interaction of fracturing fluids,
277–278
Back-propagation neural network manufacturers/importers, 72
model, 107–108 storage, 70–72
Base reservoir simulation model, Clean Air Act, 16
143–146 Clean Water Act, 16
Bauxite, 65–66 Clustering analysis
Bessel function, 230–231 available data, 131–134
Big data, 94 Fuzzy C-Mean algorithm, 130–131
Biocides, 59 Lost Hills oil field, 126–127
Biot coefficient, 40 methodology and analysis, 131

303
304 Index
objective and motivation, 127–128 Data acquisition, 85–86

results and discussion, 136–139 Database maps, 21
technology, 128–129 Data conditioning, 85–86
CMG-BUILDER, 144–145 Dayton Incident Well, 90
Coal, electricity generation from, 8 Decision variables, stimulated
Cobra jet/max fracs, 127 reservoir volume, 208
Coherency-based stacking methods, 9 Deep resistivity log (DRES), 131
Cohesive element approach, 41 Deep-well injection process, 286
“Commitment to Excellence in HF,” 15 Density-based spatial clustering
Completion of applications with noise
cost of, 153–154 (DBSCAN), 130
engineered completion approach, Department of Energy (DOE), 7, 20
128 Depth-dependent lithology, 84
parameters, 151–155 Devonian shale basins, 7
real time completion analysis Discrete fracture network (DFN), 11
technique, 182 Division of Oil, Gas, and Geothermal
Composite maps, 21 Resources (DOGGR), 21
Computational efficiency, 82 Drilling
Computational Fluid Dynamics (CFD) hydrocarbon, 80
Simulation, 273 operations, 66–68
Constant parameters, stimulated rig components, 256
reservoir volume, 209 rig floor hazardous area layout, 256
Constraints, stimulated reservoir safety culture, 254–257
volume, 209–210 Drilling induced fractures (DIFs),
Continuous fracture modeling (CFM) 80–81
technologies, 164 Drill rigs, 68
Corrosion inhibitors, 58, 59 Dynamic Finite Element Methods, 170
Cost of completion, reducing, 153–154
Coulomb stress change, 169, 171, 174 Earthquakes
CRAMP, 171 Alberta, 166–168, 173–178
Crosslinking agent, 59 in California, 20–21
Crude oil caused by induced seismicity, 22
average breakeven cost of, 28 Oklahoma, 166–168, 173–178
production, 5 potential using numerical material
rise in cost of, 7 models, 168–173
Cumulative distribution functions ECLIPSE software, 292
(CDFs) Economics
leakage scenario, 297 constraints, 218
of normalized concentration, factors, 29, 142
294–295 of hydraulic fracturing, 27–28
Curable resin coated proppant, 65 inter-lateral well spacing, 146
shale gas, 8
Darcy’s units, 230 Utica-Point Pleasant shale play, 146
Darcy velocity, 41 Edward-Trinity Plateau aquifer, 17
Index 305
Effective stimulated volume (ESV), 13 causes of, 272

Elastic effects, 38 chemical interaction of fracturing
Embedment, 62–63 fluids, 277–278
Emergency Planning and Community closing of fractures, 277
Right to Know Act, 16 depth of penetration, 276–277
Emission tomography, 9 impact of parameters under control,
Energy-based approach, 13 278–279
Energy Information Administration incremental oil production, loss in,
(EIA), 5, 248 279–280
Energy Information Agency, 27 limitations, 281
Energy Policy Act of 2005, 94 reasons for partial, 275–278
Energy resources Flowrate of well, calculations for, 283
extraction, 28 Fluid injection, 257–258
fossil, 5 Fluid mass balance equation, 40, 43
sustainable, 243 Fluids, 58
Engineered completion approach, 128 Fossil fuels, 243
Environmental impacts, 14–18, 54–55, Frac fluids, 4, 16, 258
243–245 calculations for damage, 282
Equivalent Fracture Models (EFM), chemical interaction of, 277–278
174, 177 and composition, 57
Ergonomics, 259–261 uses and needs for, 57–58
Euclidean distance, 83 FracFocus, 21
Evaporation rate, 75 Frac stages, 128
Expectation-Maximization (EM) Fracture contact problem, 43
algorithms, 130 Fracture geometry, 203–204
Extended sets, stimulated reservoir Fracture network, 80–83, 200–201
volume, 208–209 graph-based spatial analysis, 83–87
Extinguishing media, 76 real world applications, 87–91
Exxon Mobil subsidiary spills Fracture spacing, 101–103
wastewater into the Susquehanna applicability considerations,
River, 264 120–121
Eye protection, 77 artificial intelligence, 109–110
data, 110–114
Fault stability, 160, 169 method, 103–104
Fault zones, 21 model fine-tuning, 108–109
Federal and State regulatory agencies, natural fractures, 104–107
16 results, 114–120
Finite element method, 26 workflow, 107–108
Fire & explosion hazard data, 75–76 Fracture Zone Identifier (FZI)
Fire-fighting procedures, 76 attributes, 10
Flammable limits, 75–76 Fracturing rocks, 7
Flashpoint, 75 Frequently Asked Questions (FAQ),
Flowback 18
calculated values, 278 Friction reducers, 59
306 Index
Friction stress, 42 random simulations workflow, 293,

Fuzzy classification technique, 294
108–109 results and discussion, 294–297
Fuzzy C-Mean (FCM) algorithm, shallow groundwater, 288–289
130–131, 134–136 statistics of arrival time, 297
Fuzzy Inference System, 110 subsurface injection operations, 286
upward fluid migration, 287
Gamma ray (GR), 103, 132 Gutenberg-Richter law, 104
Gas compressors, 67
Gas lift, 67–68 Hazard Communication Standard
Gas resources, extraction, 27 (HCS), 72
Geertsma-De Klerk (KGD) models, Hazard maps, 14
275 Hazardous decomposition products,
Gelling agents, 59 76
Geographic information system (GIS), Hazardous polymerization, 76
82 Health hazard data, 76
Geologic data, 84–85 HF. See Hydraulic fracturing (HF)
Geomechanics HiWAY, 274
evaluation, 11–14 Holistic design techniques, 102
of hydraulic fracturing, 11–14 Horizontal drilling, 7
microseismic mapping, 26 Human Error Probability (HEP), 260
proxy, 174–175 Human factors
regional faults, 166–168 considerations, 252–253
Geometric interpretation and safety culture, 259–261
of parameters, 221–225 Human, Organizational, and
stimulated reservoir volume, Technological (HOT) subsystems,
203–204 261–262
Global Centroid Moment Tensor Hybrid microseismic survey design, 12
(CMT), 186–187 Hybrid neural network workflow, 103
Graphs, 82–83 Hydrafrac, 7
Graph-based approaches, 87–88 Hydraulically induced fractures
Graph-based spatial analysis, 83–87 (HIFs), 80–81
Graph-based “topological” model, 83 Hydraulic fracturing (HF)
Graph theory, 82 accidents, 263–265
Greenhouse gas emission, 243 additional types, 66
Griffth failure, 42 additives, common fracturing,
Groundwater contamination, 14–17 58–60
hypothetical case description, additives, percentages, 60–66
291–294 benefits of, 250
mathematical formulation, 290–291 characterization, 8–10
numerical method, 291–294 chemical interaction of, 277–278
parameters of model, 293 chemicals and products on
probabilities calculation, 296 locations, 70–72
problem statement, 289–290 closing of, 277
Index 307
definition, 4 Hydraulic Fracturing Test Site I

diagnostics, 182, 188 industry consortium (HFTS-1),
disposal process, 20 159–160
drilling, 66–68, 254–257 Hydrocarbon drilling, 80
earthquakes in California, 20–21 Hydrochloric acid, 58–59
economics of, 27–28 Hydrogen, 7
energy transfer from, 49
environmental impacts, 14–18, 28, Illinois Basins, 7
54–55, 243–245 IMEX, 144
example, 4, 5 Incompatibility, 76
fault activation from, 20 Induced seismicity, 14–18
field and laboratory setting, 16 causes, 19
fluid injection, 257–258 definition, 19
frac fluids, 57–58 documented cases of injuries,
fracking fluid, 258 22–23
geomechanics of, 11–14 earthquake caused by, 22
horizontal and vertical stress, 8, 9 in field and laboratory setting, 17
on horizontal/directional wells, 4 hydraulic fracturing causing,
human factors considerations, 19–20
252–253 in California, case study, 23–27
importance, 5–8 natural vs., 20
initiation and cessation, 11–12 new workflow for estimating,
intersection constraints, 224 173–178
mathematical formulation, 40–43 property damage caused by, 22–23
modelling, 275–276 from wastewater injection, 21–22
on natural fractures, 46–49 Induced Seismicity Consortium (ISC),
numerical model, 43–44 20–21
oil and gas resources extraction via, Induced Seismicity Map (ISM), 24
27 Induced Seismicity Potential (ISP),
physical model, 39–40 160, 172
pressurization, 48 in Alberta, 178
risk factors, 16, 29 geomechanical proxy for, 174–175
simulation results, 44–45 injection volumes, 177
stage nodes representation, 223 Induced Seismicity Potential in Energy
stage parameters, 224 Technologies Report, 22
time evolution of effective stresses, Industry, and regulators, 93
46–47 Inflow equation, refractured oil wells,
Tower of Babel, 55–57 235–241
upgraded visualization of, 273–275, Injuries, documented cases of, 22–23
282 In situ reservoir properties, 11
valid paths for neighboring wells, In situ stress, 11, 12
224 In-situ stress anisotropy, 38
wastewater, 258–259 Instantaneous event, 38
Hydraulic fracturing Test Site, 11, 16 Integer Programming (IP), 161, 203
308 Index
Inter-lateral well spacing, 142 Macroergonomics, 260–261

base reservoir simulation model, Magma, 54
143–146 Manufactured ceramics proppants,
capital and operating costs, 151 65–66
completion parameters, 151–155 Marcellus play, 104
cost of completion, 153–154 Mass balance equation, 40
economic analysis inputs, 146 Material Point Method (MPM), 160,
field cumulative gas production, 150 164, 170–173, 175, 178–179
multi-lateral depletion, 148–151 Material Safety Data Sheets (MSDS),
net present value optimization 72–73
study, 150 contents of, 73
optimization scenarios, 147–148 fire & explosion hazard data, 75–76
single well case, 148 hazardous ingredients of mixtures,
Iron control, 59 74
health hazard data, 76
Kaolin, 65–66 personal protection information, 77
k-mean clustering, 130 physical data, 74–75
k-nearest neighbor algorithm, 83, 86 product identification, 73–74
Armstrong PA, 90 reactivity data, 76
Bradford field, 88, 89 Matlab program, 83, 85, 86
connected graph, 92 Mechanical equilibrium equations, 40
on Dayton PA area, 92 Melting point, 74
wellbores, 91–93 Methane, 54
Kronecker delta, 186 emissions, 18
k sensitivity, 87–88 leakage, 14
Michigan Basins, 7
Landscape, changes in, 68–69 Microearthquakes, 20
Large scale predictive model, 178–179 Microergonomics, 260
Leakage scenario, 88–91 Microseismicity
Length-to-width ratio, 105 data analysis, 186–187
Lethal dose 50, 76 distribution, 11
Linear elastic fracture mechanics of natural fractures, 48
(LEFM) model, 11 Microseismic mapping
Linear momentum balance equations, 40 extracting fracture geometry
Linear programming-based approach, information, 9–10
202 geomechanical approach for, 26
Lithology, 84–85 stimulated reservoir volume, 201
Lost Hills oil field Microseismic monitoring, 204
clustering analysis, 129–130 data, 187–188
location, 126 design based on observations, 102
objective and motivation, 127–128 evaluation of, 12
technology, 128–129 pitfalls in analysis, 196
Lower explosive limit (LEL), 75–76 results, 188–195
Lower Marcellus Layer, 112 Microseismic waveform, 13
Index 309
Midland Basin, 16 multistage hydraulic fracturing,

Model fine-tuning, 108–109 104–107
Modern Shale Gas Development in the pre-existing, 39
United States, 61 Natural gas, 7–8, 54, 243, 248
Mohr-Coulomb theory, 42 Network training, 112
Moment tensor inversion (MTI), 13 Neural network design, 112–114
Monkeyboard, 255 Neuro-Fuzzy workflow, 103
Monte-Carlo simulation, 292–293 Neutron porosity (NPHI), 132
MSDS. See Material Safety Data Sheets Nitromethane, 60
(MSDS) Noise, 67–68
Mud logs, 103, 107, 111–112 Nolte-Smith approach, 182–185, 188
Mud spill in Pennsylvania state forest, Non-uniform time stepping, 44
265 Numerical material models, 168–173
Multi-lateral depletion, 148–151
Multi-stage coiled tubing fracturing Occupational Safety and Health
technique, 127 Administration (OSHA), 72
Multistage hydraulic fracturing, Executive Summary, 72
101–102 permissible exposure limit, 74
applicability considerations, Oda’s method, 142
120–121 Odor, 75
artificial intelligence, 109–110 Oil
data, 110–114 extraction, 27
method, 103–104 production, loss in incremental,
model fine-tuning, 108–109 279–280
natural fractures, impact of, One-size-fits-all approach, 22–23, 102
104–107 One-way coupled approach, 42
pilot wells, 103 Optimization
results, 114–120 procedure, 211–212
workflow, 107–108 process, 205, 206
scenarios, 147–148
National Academies of Science (NAS), Organization of the Petroleum
19 Exporting Countries (OPEC), 28
National Energy Technology Overpressure, 44–45
Laboratory (NETL), 27, 88 Oxygen scavenger, 59
National Energy Technology
Laboratory Research Pads, 68
Participation Program, 94 Pennsylvania Department of
National Environmental Policy Act, 16 Environmental Protection
National Institute of Occupational (PADEP) report, 90
Safety and Health (NIOSH), 249 Percentage volatile by volume, 75
National seismic hazard maps, 18 Perkins-Kern-Nordgen (PKN) models,
Natural fractures, 38 275
hydraulic fracturing effect on, 46–49 Permeability log (PERM), 132
microseismicity of, 48 Permian Basin, 11
310 Index
Permissible exposure limit (PEL), 74 Rate transient analysis (RTA), 142

Personal protection information, 77 Reactivity data, 76
Petroleum extraction industry, 166 Real time completion analysis
Plane stress model, 39 technique, 182
Point load, 62 Real-time pressure monitoring, 289
Poisson’s ratio, 41, 107, 112 Recovery factor (RF), 149, 151
Porosity log (PHIT), 132 Recycling wastewater, 29
Prague fault, 176 Reduced-order models, 38
Pre-cured resin coated proppant, 65 Refractured oil wells, 28
Pre-existing fractures, 11, 38, 39 fracture conductivity on, 233
Pressure inflow equation, 235–241
barrier, 228 mathematical model, 229–231
bottom hole, 11, 39, 146 matrix-fracture cross flow, 229
data, 181 model verification, 231
data analysis, 182–186 number of fractures on, 233
overpressure, 44–45 pressure barrier, 228
real-time pressure monitoring, 289 pressure sinks, 228
sinks, 228 radial fractures, 229
vapor pressure, 74–75 reservoir and fracture properties,
Probabilistic-based simulations, 81 232
Probability density functions (PDF), secondary orthogonal fracture, 228
293–297 sensitivity analysis, 231–234
Production performance, 133 Regional bounding lithology, 84–85
Product storage, 70–72 Regional faults, 166–168
Property damage, caused by induced Regional-scale geological databases, 84
seismicity, 22–23 Regulators, industry and, 93
Proppants, 59, 61–63 Research and development (R&D), 16
coverage at end of pumping, 138 Reservoir volume, 8. See also
hardness, 63 Stimulated reservoir volume
manufactured ceramics, 65–66 (SRV)
resin coated, 65 Resin coated proppant, 65
shape, 63 Resource Conservation and Recovery
silica sand, 63–65 Act, 16
size and placement, 62 Respiratory protection, 77
Proxy for Coulombs Stress (PCS), “Right to know” program, 71
171–172 Rocky Mountain Arsenal waste
Pseudo-steady state flow conditions, injection site, 244, 251, 287
235–241 Ryder trucks, 60
PyLith, 43
Safe Drinking Water Act, 16
Qualitative methodology, 23 Safety culture
Quanta, 85–86 accidents, 263–265
Quasi-static mechanical equilibrium, common criticisms, 250–252
40, 43 considerations, 261–263
Index 311
drilling, 254–257 Shear traction magnitude, 42

Exxon Mobil subsidiary spills Short term exposure limit (STEL), 74
wastewater into the Susquehanna Signal-to-noise ratio (SNR) emissions,
River, 264 9, 13
fluid injection, 257–258 Silica sand, 63–65
fracking fluid, 258 Silicosis, 64–65
human factors considerations, Siloed professionals, 173
252–253 Sintering process, 66
Mud spill in Pennsylvania state Skin protection, 77
forest, 265 Slick-water fracture, 7
poor decision-making by Social license to operate (SLO),
management, 264–265 165–166
wastewater, 258–259 Soil pollution, 68
San Joaquin Valley, seismicity in, Solubility in water, 75
23–27 Southern California Seismic Network
SARA reporting, 71 (SCSN), 24
Scale inhibitor, 59 Sparse matrix, 85–86
Semi-analytical approach, 161, 231 Spatial analysis, 83–87
Sets, stimulated reservoir volume, 208 Specific gravity, 75
Shale 2.0, 164 Spill Prevention Countermeasure and
Shale gas, 7 Control (SPCC) plans, 71
economic impact of, 8 SRV. See Stimulated reservoir volume
production of, 27 (SRV)
Shale oil, 5–7, 28, 248 Stability, 76
Shale reservoir, 200–203 Stabilizing agents, 59
flow chart, 204 Stage spacing, 102
geometric interpretation, 203–204 Stanolind Oil and Gas Corporation, 7
mathematical model, 204–212 State regulators, 29
model building, 212–216 Stimulated reservoir volume (SRV), 48,
recommendations, 216–218 142, 161
well and fracture design vector assumptions and constraints,
components, 204 207–210
Shale resources, 7 complexity, 200
characterization, 27 developed model flow chart, 204
development, 164 discretized for one fracture stage,
economics of, 8 225
exploitation, 202 drainage volume for, 211
North American, 166 equation, 201
regulations, 16 geometric interpretation, 203–204
Shale rock, 200, 201, 248 input parameters, 215
Shale volume (VSH), 133 linear programming-based
Shallow groundwater, risk pathways to, approach, 202
288–289 mathematical model, 204–207
Shear stress, 9, 12, 45, 172 methodology, 207
312 Index
microseismic mapping, 201 real world applications, 87–91

model building, 212–214 value of, 86–87
multigrid based, 203 Tower of Babel, 55–57
network connections, 222 Traditional geospatial models, 82, 86
objective function, 207 Traffic problems, 69–70
for one fracture stage, 222 True Vertical Depth (TVD), 24
optimization procedure, 205, 206,
211–212 Underground Injection Control (UIC)
output parameters, 215–216 well data, 167
overlapping design, 207 Undirected graph, 82
physical and economic constraints, United States Environmental
218 Protection Agency (USEPA), 287
properties of reservoir model, 217 Upper explosive limit (UEL), 75–76
recommendations, 216–218 Upscaled models, 38
representation, 210–211 U.S. Energy Information
results, 216 Administration (EIA), 5, 248
size of, 200 U.S. Geological Survey (USGS), 19–20
staggered design, 207 US Seismic Hazard Map, 18, 24
well and fracture design vector Utica-Point Pleasant shale play
components, 204 base reservoir simulation model,
well pad and evaluation, 214–216 143–146
Streamtube model capital and operating costs, 151
assumptions, 272 completion parameters, 151–155
limitations, 281 cost of completion, 153–154
upgraded visualization of hydraulic economic analysis inputs, 146
fracturing, 273–275, 282 field cumulative gas production, 150
Stresses, 84 location, 143
field, 11, 45, 160, 164, 169, 172, multi-lateral depletion, 148–151
174 net present value optimization
friction, 42 study, 150
plane stress model, 39 optimization scenarios, 147–148
shadowing, 129 single well case, 148
shear, 9
in situ stress, 11, 12 Vapor density, 75
volumetric effective stress, 41 Vapor pressure, 74–75
Stress tensor, 40 Variables, stimulated reservoir volume,
Subsurface Fluid Injection and 208
Production (SFIP) strategy, 13 Vector test function, 41
Subsurface geology, 84–85 Velocity model, 13
Surfactant, 59, 60 Ventilation, 77
Viscosity, 75
Time weighted average (TWA), 74 Visual pollution, 68
Topological algorithm, 85–86 Volumetric effective stress, 41
big data, 94 Volumetric strain, 41
Index 313
Wastewater, 258–259 spatial locations of, 91

disposal, 164, 168, 177, 289 Well integrity, 287–289, 298
injection, induced seismicity from, Well pad, 214–216
21–22 Well perforation design, 128, 257
recycling, 29 Well placement process, 220–221
Waterflooding, 127, 131 Well spacing, 142
Water pollution, 68, 287, 289 Wolfcamp Delaware basin, 28
Water usage, amount of, 17–18
Weak solid-to-fluid coupling, 41 Young’s modulus, 41, 107, 171
Wellbores, 81–82
k-nearest neighbor algorithm, 91–93 Zero-displacement boundary
point data, 85 condition, 12

(Sustainable Energy Engineering) Fred Aminzadeh - Hydraulic Fracturing and Well Stimulation (Sustainable Energy Engineering) - Wiley-Scrivener (2019)

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

(Sustainable Energy Engineering) Fred Aminzadeh - Hydraulic Fracturing and Well Stimulation (Sustainable Energy Engineering) - Wiley-Scrivener (2019)

Uploaded by

Copyright:

Available Formats

Hydraulic Fracturing

Wiley Global Headquarters

Limit of Liability/Disclaimer of Warranty

Library of Congress Cataloging-in-Publication Data

Cover image: Oil Rig - Luchschen | Dreamstime.com

Printed in the USA

Part 2: General Concepts 35

2.6 Effect of Hydraulic Fracturing on Natural Fractures 46

4 A Graph Theoretic Approach for Spatial Analysis of Induced

Part 3: Optimum Design Parameters 99

5.3 Data 110

Part 4: Fracture Reservoir Characterization 159

10 Multigrid Fracture Stimulated Reservoir Volume Mapping

11.4 Sensitivity Analysis 231

Part 5: Environmental Issues of Hydraulic Fracturing 243

13.3.1 Concept 273

M. Y. Soliman, PhD, PE, NAI

Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (1–33)

University of Southern California

Fred Aminzadeh (ed.) Hydraulic Fracturing and Well Stimulation, (3–33)

publicized controversy on whether and to what extent hydraulic fracturing cre-

Keywords: Hydraulic fracturing (HF), shale resources, energy resources,

1.1 What is Hydraulic Fracturing?

(Not to scale) Approximate distance

Figure 1.1 An example of a typical HF configuration [1].

gas, and the associated induced seismicity or manmade earthquakes. For

1.2 Why Hydraulic Fracturing is Important

U.S tight oil production- selected plays

Figure 1.2 The explosion of shale production since 2010 [5].

above ground by as much as 90%. Fracking is the key to unlocking vast US

1.3 Fracture Characterization

Microseismic mapping associated with hydraulic fracture stimulations

frequency content of the seismic recording equipment and much more

1.4 Geomechanics of Hydraulic Fracturing

at the microscale plays a critical role in predicting the performance of HF.

geophones, allows for optimum coverage for effective characterization of the

given site, reflecting on the observation that this inference is predicated on

1.5 Environmental Aspects of Hydraulic Fracturing

(a) Ground-Water Contamination.

as currently practiced in California, do not result in

holes in the casing, to expose the wellbore to the shale.

Conceptual Model Simulation Result Hazard Index

Sand Layer S˚ =0.3 0.4

35.5°N 120.8°W 120.6°W 120.4°W 120.2°W 120°W

pad scale fracturing operations on ground water and air

of gallons of water is expensive. Thus, if local groundwater

1.6 Induced Seismicity

a. What is induced seismicity?

b. What causes induced seismicity?

c. Can HF cause “induced seismicity”?

a substantial cause of induced seismicity, and that “only very rarely” is

d. How can one determine whether a seismic event is natural or induced?

i. Are these events the first known earthquakes of this char-

e. Has HF caused earthquakes in California?

g. How do we know if an earthquake was caused by induced seismicity?

h. Are there any documented cases of injuries or property damage caused by

because of the significant differences in local geology and surface condi-

1.7 Case Study: Fracturing Induced Seismicity

7.5 0 7.5 15 22.5 30 km

Our case history also attempts to distinguish between induced and

Hydraulic Fracturing Wells

1.8 Assessment of Global Oil and Gas Resources

Recycling wastewater must be a priority in these cases. In addition, more