&EPA
   United States
   Environmental Protection
   Agency

EPA 600/R-l 1/047 I May 2011 I www.epa.gov
                   Proceedings of the Technical Workshops
                   for the Hydraulic Fracturing Study:
                   Fate and Transport
   U.S. Environmental Protection Agency

-------
S-EPA
EPA 600/R-11/047
    May 2011
    United States
    Environmental Protection
    Agency
  Proceedings of the Technical Workshops
  for the Hydraulic Fracturing Study:
  Fate and Transport
  Office of Research and Development
  US Environmental Protection Agency
  Washington, DC

-------
                               Table of Contents

List of Figures	ii
List of Tables	iii
Introduction	1
Workshop Participants	3
Agenda	5
Summary and Abstracts from Theme 1: Contaminant Identification, Transformation and
Transport	1
  Summary of Presentations for Theme 1: Contaminant Identification, Transformation and
  Transport	2
  Summary of Discussions Following Theme 1: Contaminant Identification, Transformation and
  Transport  Presentations	4
  Abstracts for Theme 1: Contaminant Identification, Transformation and Transport
  Presentations	8
    A Simple Approach  Identifying Contaminants of Most Concern to Underground Drinking
    Water Supplies from Hydraulic Fracturing Operations	9
    Contaminant Characterization of Effluent from Pennsylvania Brine Treatment Inc.,
    Josephine Facility Being Released into Blacklick Creek, Indiana County, Pennsylvania:
    Implications for Disposal of Oil and Gas Flowback Fluids from Brine Treatment Plants	10
    Fracing  & Associated Media Composition in Colorado	15
    Comparison of Hydraulic Fracturing Fluid Composition with Produced Formation Water
    following Fracturing- Implications for Fate and Transport	22
    Fate and Transport  of Select Compounds of Potential Concern (COPC) in Fracing Fluids... 28
    Fate and Transport  Evaluation of Potential Release Scenarios during Hydraulic Fracturing
    Operations	35
    A Protocol to Characterize Flowback Fluid Contamination of Drinking Water	41
Summary and Abstracts from Theme 2: Impacts of Hydraulic Fracturing on Natural Transport
Systems	44
  Summary of Presentations from Theme 2: Impacts of Hydraulic Fracturing on Natural
  Transport  Systems	45
  Summary of Discussions Following Theme 2: Impacts of Hydraulic Fracturing on Natural
  Transport  Systems Presentations	46
  Abstracts for Theme 2: Impacts of Hydraulic Fracturing on Natural Transport Systems	51
    Rock Failure and Permeability Enhancement in Tight Gas  Hydraulic Fracturing	52
    Fracture Growth in  Layered and Discontinuous Media	59
    Flow of  Gas and Water in Hydraulically Fractured Shale Gas Reservoirs	67
    Role of Induced and Natural Imbibition in Frac Fluid Transport and Fate in Gas Shales	70
Summary and Abstracts from Theme 3: Models to Predict Transport	79
  Summary of Presentations on Theme 3: Models to Predict Transport	80
  Summary of Discussions Following Theme 3: Models to Predict Transport Presentations	81
  Abstracts for Theme 3: Models to Predict Transport	83
    Modeling Philosophies & Application	84

-------
    Modeling Drinking Water Related Human Health Risks from Hydraulic Fracturing Additives
    	91
Revisiting the Major Discussion Points of the Technical Presentation Sessions	94
Poster Abstracts	97
    Review of Ground water Quality Data Surrounding Fracing Operations	98
    Control Banding as a Means of Hazard Identification & Characterization for Chemicals.. 101
Glossary of Terms	106

-------
                                  List of Figures
Note: The List of Figures does not include the figures from the Posters section beginning on page
97.

Figure 1. A generalized and a specific form of conceptual site model	36
Figure 2. Bivariate plot of Br/CI versus Cl concentrations for Marcellus frac fluids, Pennsylvanian
Devonian sedimentary basin brines, and representative road salt contamination and septic
effluent. Note the clear separation between the flowback fluid, brines and other sources of
contamination (Kight and Siegel, 2011)	42
Figure 3. Structural Permeability Diagram for New Albany Shale ( u=0.6)	53
Figure 4. Fracture orientation in horizontal wells	53
Figure 5. Interaction of multiple fractures in a horizontal well. Green  represents closed
fractures. Note the fractures turning away from each other to follow the path of least
resistance	54
Figure 6. Fracture aperture distribution (in cm) after 3 hours of pumping in Barnett shale	55
Figure 7. Distribution of pore pressures (MPa) in the formation; minimum  principal effective
stress (Barnett shale, 0.09 m3/s per fracture; 3 hrs) (Rawal & Ghassemi, 2011)	56
Figure 8. Microseismic map  shows network growth and the potential stimulated volume in
shale (GTI-NAS-Project)	57
Figure 9. Geometry for stress effects	59
Figure 10. Mineback photo of fracture propagating across interface	60
Figure 11. Fracture toughness data from MWX	61
Figure 12. Photograph and line drawing of fracture behavior crossing interfaces	62
Figure 13. Schematic of types of observed fracture behavior crossing interfaces	62
Figure 14. Fracture crossing discontinuities	63
Figure 15. Example of range of fracture architectures for a single initiation site, (modified from
Fisher etal, 2002)	72
Figure 16. Example fracture architecture models generated using Meyer & Assocs. MFRAC and
MSHALE fracture simulation software. (A-above) Simple planar fracture, (B-upper right) discrete
fracture network (DFN) with proppant primarily deposited in the principal fracture, (C-right)
DFN with proppant distributed in lateral fractures. (DFN Figures courtesy of Meyer & Assocs.) 73
Figure 17. General representative drainage capillary pressure curves for gas shales of low k
(0.00005 mD; red triangles)  and high k (0.002 mD; blue squares) measured using air-Hg and
converted to equivalent reservoir CH4-Brine pressures. Also shown are measured water
saturations for natural imbibition of core from as-received saturation (black circles). Grey
dashed curves represent generalized imbibition curves modeled from low-k sandstones	75
Figure 18. Threshold entry capillary pressures (Pte) versus specific in situ Klinkenberg
permeability for lithic sandstones (blue squares) and  representative gas shales (open triangles).
Pte were measured using air-Hg and converted to equivalent reservoir CH4- Brine pressures.
Data for shale show continuity with trend for sandstones and siltstones. Relationship can be
expressed: Pte = 12.25 kf0'424	76

-------
                                  List of Tables
Note: The List of Tables does not include the figures from the Posters section beginning on page
97.
Table 1. Predicted Downhole Fate for Hydraulic Fracturing Fluid Components	23
Table 2. Adapted from Table 4-1, Chapter 4, EPA Report 816-R-04-003. Hazard and toxicological
information sources from MSDS according to EPA. Biocide not included in original table.
Identified by multiple sources as a biocide for hydraulic fracturing	34
Table 3. Hypothetical release concentrations (my rough estimates)	37

-------
                                  Introduction

The Hydraulic Fracturing Study
In its Fiscal Year 2010 budget report, the U.S. House of Representatives Appropriation
Conference Committee identified the need for a study of the impacts of hydraulic fracturing on
drinking water resources. EPA scientists are undertaking a study to better understand its
potential impacts. The scope of the proposed research includes the full lifespan of water in HF,
from acquisition of the water, through the mixing of chemicals and actual fracturing, to the
post-fracturing stage, including the management of flowback and produced water and its
ultimate treatment and disposal.

EPA has included stakeholder concerns in the planning process of the study from its inception.
EPA engaged stakeholders in a dialogue about the study through a series of webinars and
facilitated public meetings held May-September 2010. Four technical workshops were hosted
by EPA during February and March 2011 to explore the following focus areas: Chemical &
Analytical Methods, February 24-25; Well Construction & Operations, March 10-11; Fate &
Transport, March 28-29; and Water Resource Management, March 29-30. EPA submitted a
draft study plan to the Science Advisory Board (SAB) in February 2011 and the peer review of
the draft study plan was held on March  7-8, 2011. (At the time the technical workshop
proceedings were developed, the SAB had not given its official review of the study plan to EPA.)

The goal of the technical workshops was three-fold: (1) inform EPA of the current technology
and practices being used in hydraulic fracturing, (2) identify research related to the potential
impacts of hydraulic fracturing on drinking water resources, and (3) provide an opportunity for
EPA scientists to interact with technical  experts. EPA invited technical experts from the oil and
natural gas industry, consulting firms, laboratories, state and federal agencies, and
environmental organizations to participate in the workshops. EPA will use the information
presented in this documents to inform research that effectively evaluates the relationship
between HF and drinking water.

An initial report of results from the study is expected by late 2012 with an additional report
expected in 2014.
About the Proceedings
These proceedings provide an overview of the fifteen presentations and two posters given on
fate and transport at the Technical Workshop for the U.S. EPA Hydraulic Fracturing Study held
on March 28-29, 2011. This workshop consisted of three sessions or themes: Theme 1-
Contaminant Identification, Transformation and Transport; Theme 2- Impacts of Hydraulic
Fracturing on Natural Transport Systems; and Theme 3- Models to Predict Transport. The
proceedings include abstracts of the presentations and a summary of the discussions that took
place during the workshop. The presentations from this workshop are not part of the
proceedings document, but may be accessed at http://epa.gov/hydraulicfracturing.

-------
This is the third of four technical workshops on topics relating to the EPA Hydraulic Fracturing
Study. The other three workshops are: Chemical and Analytical Methods (Feb. 24-25), Fate and
Transport (Mar. 28-29), and Water Resources Management (Mar. 29-30). Proceedings will be
available separately for the other three workshops.

The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.
Editorial Team for the Proceedings
The attendees at the Fate and Transport workshop were selected based on information
submitted to EPA during the attendee nomination process. Presenters, a workshop lead, and
theme leads were selected from the pool of attendees, once again, based on the information
submitted to EPA during the attendee nomination process. The workshop lead, JP Nicot of the
University of Texas at Austin, assisted EPA in finalizing details for the workshop and served as
the lead editor of the proceedings document. The theme leads—Angus McGrath of Stantec for
Theme 1, Ahmad  Ghassemi of Texas A&M University for Theme 2, and Andrew Havics of pH2,
LLC/QEPA for Theme 3—served as editors for their respective themes.

-------
Workshop Participants

Doug
Chrystal
Amy
Uni
Jeanne
Susan
Alan
David
Leslie
Brian
Robin
Jill
George
Dennis
James
Ahmad
Bill
Fred
Andrew
Zhong
Charles
David
Dawn
David
Steve
Kurt
Debra
Angus
JoAnn
Larry
Jean-
Philippe
Gary
Carl
Stan
Pete
Name
Beak
Beasley
Bergdale
Blake
Briskin
Burden
Byrnes
Cramer
Cronkhite
D'Amico
Danesi
Dean
Deeley
Degner
Farmerie
Ghassemi
Godsey
Hauchman
Havics
He
Hillenbrand
Jacobi
Kaback
Kargbo
Kraemer
McCoy
McElreath
McGrath
McMahon
Murdoch
Nicot

Norris
Palmer
Paxton
Penoyer
Affiliation
US Environmental Protection Agency
US Environmental Protection Agency
US Environmental Protection Agency
Majitox for Gastem USA
US Environmental Protection Agency
US Environmental Protection Agency
Chesapeake Energy Corp.
Conoco-Phillips
US Environmental Protection Agency
US Environmental Protection Agency
US Environmental Protection Agency
US Environmental Protection Agency
Shell Exploration & Production Company
Range Resources
ITT Water & Wastewater
Texas A&M University
Geo Logic Environmental Services, LLC
US Environmental Protection Agency
pH2, LLC
Range Resources
US Environmental Protection Agency
Chesapeake Energy
AMEC Geomatrix, Inc.
US Environmental Protection Agency
US Environmental Protection Agency
US Geological Survey
Chesapeake Energy Corporation
Stantec
Baker Hughes
Clemson University
The University of Texas at Austin

US Environmental Protection Agency
Idaho National Laboratory
US Geological Survey
National Park Service

-------
            Name
                                      Affiliation
Robert
Christopher
David
Harry
Manu
Donald
Khin-Cho
Joel
Denise
Conrad
Norman
Jim
Ron
Rick
Dean
Nathan
Mike
Dollis
Puls
Quina
Russell
Schurrlll
Sharma
Siegel
Thaung
Thompson
Tuck, P.E.
Volz
Warpinski
Weaver
Wilhelm
Wilkin
Willberg
Wiser
Worden
Wright
US Environmental Protection Agency
Ecology and Environment, Inc
QEP Energy Co.
CNX Gas/CONSOL Energy
Gradient
Syracuse University
US Environmental Protection Agency
Stantec
Halliburon Energy Services, Inc.
Graduate School of Public Health
Pinnacle - A Halliburton Service
US Environmental Protection Agency
US Environmental Protection Agency
US Environmental Protection Agency
Schlumberger
US Environmental Protection Agency
DOI\Bureau of Land Management
QEPA

-------
                                    Agenda
        Technical Workshops for the Hydraulic Fracturing Study
                            Fate & Transport • March 28-29, 2011

                                 US EPA Conference Center
                             One Potomac Yard (South Building)
                                   2777 S. Crystal Drive
                         Arlington, VA 22202 Room S-4370 and 4380


   March 28, 2011
8:30 am    Registration

9:00 am    Welcome
          Fred Hauchman, Director of the Office of Science Policy, EPA Office of Research and
          Development
          JP Nicot, Workshop Lead, University of Texas at Austin
          Pat Field, Facilitator, Consensus Building Institute

Theme 1: Contaminant Identification,  Transformation & Transport
9:15 am    Technical Presentation Session 1: Chemical Transformation and Identification
          A Simple Approach Identifying Contaminants of Most Concern to Underground Drinking
          Water Supplies from Hydraulic Fracturing Operations, Carl Palmer, Idaho National
          Laboratory
          Contaminant Characterization of Effluent from Pennsylvania Brine Treatment, Inc.,
          Josephine Facility Being Released into Blacklick Creek, Indiana County, Pennsylvania:
          Conceptual Models of Exposure Pathways to Recreationalists, Private Well Water Users
          and Municipal Water Systems, Conrad Volz, University of Pittsburgh
          Fracing & Associated Media Composition in Colorado, Andrew Havics, pH2, LLC/QEPA

10:30 am   Break

-------
10:40 am   Technical Presentation Session 2: Chemical Mobility and Potential Pathways
           Comparison of Hydraulic Fracture Fluid Composition with Produced Formation Water
           Quality Following Fracturing: Implications for Fate and Transport, Debra McElreath,
           Chesapeake Energy Corporation
           Fate and Transport of Select Hydraulic Fracturing Compounds of Potential Concern, Angus
           McGrath, Stantec
           Fate and Transport Evaluation of Potential Release Scenarios during Hydraulic Fracturing
           Operations, George Deeley, Shell Exploration & Production
           A Protocol to Characterize Flowback Water Contamination of Shallow Waters from Shale
           Gas Development, Donald Siegel, Syracuse University

12:10 pm   Lunch

Theme 2: Impacts of Hydraulic Fracturing on Natural Transport Systems
1:00 pm    Technical Presentation Session 3: Fracture Propagation
           Rock Failure and Stimulated Volume in Hydraulic Fracturing, Ahmad Ghassemi, Texas A&M
           University
           Fracture Growth in Layered and Discontinuous Media, Norman Warpinski, Pinnacle - A
           Halliburton Service

2:00 pm    Break

2:10 pm    Technical Presentation Session 4: Fluid and Gas Flow in Fractured Formations
           Flow of Gas and Water in Hydraulically Fractured Shale Gas Reservoirs, Zhong He, Range
           Resources
           Characterizing Mechanical and Flow Properties using Injection Falloff Tests, David Cramer,
           ConocoPhillips
           Role of Induced and Natural Imbibition in Fracturing Fluid Transport and Fate in Gas
           Shales, Alan Byrnes, Chesapeake Energy Corporation

3:25 pm    Break

3:35 pm    Revisit the Major Discussion Points of the Technical Presentation Sessions
           JP Nicot, Workshop Lead, University of Texas at Austin
           Angus McGrath, Theme 1 Lead, Stantec
           Ahmad Ghassemi, Theme 2 Lead, Texas A&M  University

4:00 pm    Adjourn for the Day

-------
   March 29, 2011
Theme 3: Models to Predict Transport
8:00 am    Technical Presentation Session 5: Demonstration of Models and Determinations of
          Model Accuracy
          Modeling Versus the Real World of Hydraulic Fracturing, Denise Tuck, Halliburton
          Modeling Philosophies and Application, Andrew Havics, pH2, LLC/QEPA
          Modeling Drinking Water Related Human Health Risks from Hydraulic Fracturing Additives,
          Manu Sharma, Gradient

9:15 am    Revisit the Major Discussion Points of the Technical Presentation Session
          JP Nicot, Workshop Lead, University of Texas at Austin
          Andrew Havics, Theme 3 Lead, pH2, LLC

9:45 am    Closing Discussions
          Jeanne Briskin, Hydraulic Fracturing Research Task Force Leader, EPA Office of Research
          and Development
          JP Nicot, Workshop Lead, University of Texas at Austin

-------
Summary and Abstracts from Theme 1: Contaminant
   Identification, Transformation and Transport

-------
    Summary of Presentations for Theme 1: Contaminant Identification,
                         Transformation and Transport

The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.

Technical Presentations
The first set of technical presentations in this theme addressed chemical transformation and
identification.

Carl Palmer, Idaho National Laboratory, described an approach for categorizing and ranking
contaminants to determine which may be  of most concern regarding impacts to ground water
from hydraulic fracturing. This approach has been employed for in situ oil shale development to
help identify contaminants that are likely to pose the greatest risk based on the contaminants'
mobility, concentration, and toxicity. Mr. Palmer presented a graphical representation of
mobility versus toxicity and identified contaminants of the greatest concern as those with both
high mobility and high toxicity. Mr. Palmer also described the general methodology and
limitations of this approach.

Conrad Volz, University of Pittsburgh, gave an overview of a field sampling and contaminant
characterization study of the effluent from the Pennsylvania Brine Treatment facility (a public
owned treatment works, or POTW) in Josephine, Pennsylvania. The POTW accepts only
wastewater from the oil and gas industry,  including flowback water from Marcellus Shale gas
extraction operations. The study estimated the 24 hour mean concentration of a range of
chemicals and compounds including barium (Ba), bromides (Br~), strontium (Sr), benzene, 2-
butoxyethanol (2-BE), chlorides (Cl~), magnesium (Mg), total dissolved solids (TDS), sulfate (S042~
), and pH. The concentrations found were compared to various drinking water, human
exposure, and aquatic health guidelines and standards with many contaminants found at
concentrations that exceeded standards and guidelines. Dr. Volz also discussed the implications
of processing flowback fluids at numerous POTWs. Dr. Volz then provided environmental public
health recommendations based on the results of the analysis.

Andrew Havics, pH2, LLC/QEPA, described results of a 2008 study on HF impact and estimated
risk in four basins in Colorado. He conducted an assessment based on samples from pit solids
and fluids, fracture fluids, drilling fluids, early and late flowback fluids/produced water, and
soils (for background analyses). Potential chemicals of concern were identified and selected for
risk assessment purposes. Mr. Havics emphasized the importance of determining background
concentrations of potential contaminants. He described the limitations of classic risk
assessments in this specific application and described his study methodology, including
statistical assumptions and study limitations. Mr. Havics also noted the importance of
establishing a process for addressing non-detects in the chemical analyses.

-------
The second set of technical presentations addressed chemical mobility and potential pathways.

Debra McElreath, Chesapeake Energy Corporation, discussed implications of predicted fate and
transport of HF chemicals, fate evaluation, and transport evaluation. Temperature and pressure
affect the fate and transport of HF fluid components. Chesapeake compared vendor-supplied
information on HF fluids, mixed HF fluids, and produced formation water following HF activities
over the life of the well. The study demonstrated that produced water volumes vary based on
location and tend to decrease overtime. In addition, chemical concentrations can stabilize at
different times in different shale plays and produced formation waters are highly variable in
composition within and between shale formations. Ms. McElreath recommended that TDS,
chloride, and divalent cations, which are present in high concentrations in formation waters,
could be used as indicator chemicals.

Angus McGrath, Stantec, described a series of common and unconventional compounds of
interest (COIs) in HF fluids. He discussed each compound's chemical properties and fate and
transport characteristics, including mobility and  persistence. In addition, he discussed the
implications for each with respect to drinking water quality. Drawing on his fate and transport
findings including the lack of persistence,  Dr.  McGrath noted that monitoring for any of these
compounds may only be warranted if impacts from more concentrated reagents in fracturing
fluids are detected in wells.

George Deeley, Shell Exploration and Production, described considerations for fate and
transport evaluation of potential release scenarios during HF operations. Developing a
conceptual site model  (CSM)  is the first step in the detection and prevention of potential
problems. Appropriate CSMs consider all controls in the HF process as reflected in existing
regulations and  best management practices.  Dr. Deeley emphasized the importance of focusing
on realistic release scenarios, chemicals, pathways, and receptors, as well as the importance of
establishing background concentrations and understanding the uncertainty of data and
analytical methods.

Donald Siegel, Syracuse University, discussed a protocol to characterize flowback fluid
contamination of drinking water and determine  its source. Much of the concern over flowback
fluids stems from elevated concentrations of trace elements, radium-226, and trace organic
substances according to Dr. Siegel. However, these constituents geochemically react and
naturally attenuate, stated Dr. Siegel, and thus elevated concentrations do not clearly identify
the extent of flowback fluid contamination in drinking water. Dr. Siegel proposed that
concentrations and ratios of dissolved halogens (hloride, bromide, and fluoride, and  iodide)
allow for identification of contamination from shale gas or oilfield brines. He described
geochemical mixing  models that allow for the identification of small amounts of flowback
water, even when other salinity sources previously contaminated the water.

-------
 Summary of Discussions Following Theme 1: Contaminant Identification,
                 Transformation and Transport Presentations
The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.

The mobility/toxicity matrix. A participant asked if original fracture fluid concentrations were
considered in the mobility/toxicity assessment. The presenter clarified that the provided
example was from an oil shale and did not involve fracture fluids,  but that the matrix approach
is applicable to other scenarios. A participant asked why the two order of magnitude red uction
in mobility was selected as a differentiation point in the mobility/toxicity matrix. The presenter
explained that this was not a standard number but was selected based on travel times in the
system.

Treating produced water at the Josephine facility. A participant asked whether the plots of
decreasing barium and strontium concentrations over a 24-hour period represented conditions
following cessation of waste disposal. The presenter clarified that surface disposal at the
Josephine facility is a continuous operation and concentrations vary over time, and added that
some POTWs discharge periodic pulses of effluent. The presenter indicated that the rate of
discharge from the Josephine facility is approximately  150,000 gallons per day, and studies are
currently investigating pore water and annual flow. He added that the facility uses settling
ponds and sulfate (to remove barium) to treat the produced water and no techniques are
applied to treat for salinity or chloride. Another participant asked  if a person  would be less
affected by barium than by the high salinity of this  water. The presenter agreed and noted that
there are drinking water wells less than 100 yards downstream of the Josephine plant. He
added that the capture curves of wells like these indicate that water can be pulled directly from
the stream, which is a source  of concern.

The presenter clarified that the Josephine facility has been in operation since the mid-1980s
and currently takes only oil and gas fluids; it previously treated conventional  brine from oil
operations and now receives Marcellus Shale fluids instead. The presenter indicated that the
Josephine facility processed 15 million gallons between July 2010  and December 2010. A
participant asked if salinity had previously been a concern at  the facility, and  the presenter
responded that the facility had not  received much attention until  it began processing Marcellus
Shale flowback water. The presenter agreed with a participant's point that chloride levels have
likely been high for years.  He added that high levels of arsenic have also been found in ground
water samples throughout Pennsylvania, though  these are likely natural background
concentrations.

A participant asked if the study considered influent to the plant or conditions upstream from
the plant. The presenter clarified that his team did  not have access to those potential sample
locations and relied on information published by the Marcellus Shale Coalition, the Society of
Petroleum Engineers, and  other sources. However, the presenter  mentioned that there are a

-------
number of ongoing projects addressing these issues including one study that is investigating
impacted versus unimpacted drainages throughout Washington and Greene counties. The
study is investigating Ba, Sr, several anions, 2-BE, thallates, and BTEX (benzene, toluene,
ethylbenzene, and xylenes).

A participant asked how actual discharge values compare to permit limits at the Josephine
facility. The presenter indicated that the permit allows for extremely high levels of total
dissolved solids (TDS) and chloride. He compared this to the Dunkard Creek fish kill in 2009,
noting that golden algae could grow at this site, especially in the summer, when flow decreases
in the stream. A participant confirmed that the Josephine plant is grandfathered under
Pennsylvania's new TDS regulations. Another participant asked if the sludge from the facility
was analyzed. The researchers did not have access to the sludge, but have made observations
regarding its volume and transport. The presenter expressed concern about leaching from
landfills, though a participant described the result of leaching studies that indicated  no risk
from leaching.

2-butoxyethanol (2-BE). Participants noted that 2-BE is a common  solvent present in household
cleaners and other products and is expected to be found in effluent from wastewater treatment
plants. They, therefore, suggested that it is important to quantify the background
concentration, which is challenging. A participant added that the ATSDR (Agency for Toxic
Substances and Disease Registry) reports that 2-BE was primarily tested on rats and  mice, which
are more prone to hematological reactions than people are. The participant noted that the
difference in reactions between rodents and humans results in a great  uncertainty factor built
into 2-BE toxicology.

Safety of fish consumption. A participant asked whether studies have been done to determine
the safety offish consumption with regard to flowback contamination. The presenter clarified
that no studies have been done; however, the same watersheds have been studied for a
number of factors. For example, the presenter noted that Pittsburgh is also a center for coal-
fired power plants, which contribute to mercury, selenium, and arsenic buildup in fish. Another
presenter said that there are plans to study fish consumption during 2011 and that this analysis
is essential because of the popularity of recreational fishing in Pennsylvania. The state issues
fish advisories, but a participant noted the Pennsylvania Department of Environmental
Protection (PA DEP) does not have the support to do much sampling although there are
currently fish  advisories for mercury and for a number of bioaccumulative contaminants,
including some pesticides.

Concentration of flowback fluid versus injected fluid. A participant asked about the
concentration of contaminants in fracturing fluids and pits in the Colorado study. The
participant noted that operators have control over injected fluids. The  presenter noted that
there is a significant difference between concentrations in fracture fluids and early and late
flowback, and he recommended that participants refer to the risk  assessment document
available on QEPA's Web site (http://www.qepa.com). The presenter also noted that data for
pit fluids are also available but were not presented in this workshop.

-------
Fingerprinting shale formations. A participant noted there are large differences in radionuclides
in shale formations regionally, geographically, vertically, and horizontally. However, the
participant added that there is likely not enough data available for fingerprinting.

Soil and degradation. A participant asked about microbial activity in soils compared to the deep
subsurface, and the presenter agreed that biological  activity is much lower in deeper
formations and  microbial degradation should be much slower. A participant asked about
degradation of the  biocide described by him as a microbially active compound  that studies
show that it degrades rapidly. The participant indicated that the biocide has toxic effects but
that there is a dosing limit to the biocide below which it no  longer functions as a  biocide and
can be degraded by microbes.

Principal component analysis. A participant asked whether Dr. Deeley had used principal
component analysis for comparing two different sources. The presenter clarified  that he has
not used this technique.

Prior water contamination. A participant asked about the impacts of prior water contamination.
The presenter stated  that previous road salt contamination should not pose a problem because
the bromide/chlorine ratio is tiny compared to formation brine, but there may be other
parameters. The presenter also suggested that iodide is harder to analyze than bromide or
chlorine. A participant added that water recycling is popular in  Pennsylvania, and another
participant asked if recycling flowback fluid would complicate analyses. In response, the
presenter clarified that the composition of the new flowback can be  incorporated into the
mixing model.

Fracture fluid versus produced water. A participant asked Ms. McElreath whether she looked at
the total mass of chemicals to determine how much fracture fluid returned when comparing
fracture fluid  and produced water. The  presenter indicated  that, in general, 20%  to 40% of the
fracture fluid  is  returned to the surface, while approximately 70% remains in the  formation.
However,  she stated that this varies by play and within the formation. Another participant
noted that toluene  and benzene came back in produced water, and asked whether those
contaminants could be byproducts or products from  biological  reactions. The presenter
responded that  these are not likely to be byproducts from biological reactions.

Conceptual models. A participant asked for a  description of a model for fracture fluid that is
injected at a depth  of 6,000 feet and migrates into an aquifer. The presenter explained that one
would need to consider how to produce a fracture reaching the surface. The presenter then
noted that the basic energy balance must be  considered, and it is important to determine
whether there are natural controls. He  suggested that other things to consider are whether
there are man-made controls to contain those pressures, and if other potential pathways (such
as abandoned wells) exist, how deep they extend, and whether they were completed or closed.
The presenter added  that it is essential to consider whether the models predict reasonable
scenarios.

-------
Availability of previous studies. A participant asked whether there have been previous efforts to
study oil fields and noted that it would be useful for EPA's Office of Research and Development
to access that work. In the late 1980's, there were EPA studies on  produced water and drilling
mud waste for RCRA considerations, and those studies are available on the EPA Web site.

-------
 Abstracts for Theme 1: Contaminant Identification, Transformation and
                          Transport Presentations
Abstracts were submitted to U.S. EPA by the presenters for use in this proceedings document.
              Not all presenters submitted abstracts of their presentations.

The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.

-------
 A Simple Approach Identifying Contaminants of Most Concern
     to Underground Drinking Water Supplies from Hydraulic
                          Fracturing Operations
                          Carl D. Palmer and Earl D. Mattson
                              Idaho National Laboratory

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.
There is concern that hydraulic fracturing operations for natural gas extraction could potentially
contaminate groundwater supplies. We illustrate an approach we have employed for in-situ oil
shale development to help identify contaminants that are likely to pose the greatest risk based
on the contaminants' mobility, concentration, and toxicity. Mobility, addressed from the
sorption characteristics of the compounds, is estimated under the expected subsurface
geochemical conditions (pH, Eh, total organic carbon) using published experimental data and
linear free energy relationships. We use the ratio of the concentration to some guideline
concentration (e.g., an MCL) as a measure of the "toxicity" of that contaminant.  A plot of
mobility versus toxicity provides a simple graphical view of the relative risk for the transport of
various contaminants from the near field to the far field. A Euclidean norm centered on a point
of unit toxicity and low mobility can be used to order potential contaminants of  concern. We
illustrate the application of this approach using an oil shale retorting example and discuss
limitations in the assumptions and available data. Applying an approach such as  this to
chemicals used in hydraulic fracturing fluids could help set research and development priorities
for additional research.

-------
  Contaminant Characterization of Effluent from Pennsylvania
  Brine Treatment Inc., Josephine Facility Being Released into
 Blacklick Creek, Indiana County, Pennsylvania: Implications for
 Disposal of Oil and Gas Flowback Fluids from Brine Treatment
                                   Plants
 Conrad D. Volz, DrPH, MPH; Kyle Ferrar, MPH; Drew Michanowicz, MPH, CPH; Charles Christen,
  DrPH, MEd; Shannon Kearney, MPH, CPH; Matt Kelso, BS; and Samantha Malone, MPH, CPH
                             University of Pittsburgh

 The statements made during the workshop do not represent the views or opinions of EPA. The
          claims made by participants have not been verified or endorsed by EPA.
This report contains results from sampling and analysis of wastewater effluent entering
Blacklick Creek, Indiana County Pennsylvania from the Pennsylvania Brine Treatment (PBT)
Josephine Facility conducted by the Center for Healthy Environments and Communities (CHEC).
The PBT-Josephine Facility accepts only wastewater from the oil and gas industry, including
flowback water from Marcellus Shale gas extraction operations. This report describes the
concentrations of selected analyzed contaminants in the effluent water and compares the
contaminant effluent concentrations to standards, guidelines and criteria set by federal and
state regulatory and investigative agencies for the protection of human and aquatic health.

Sampling Methodology and Concentrations of Contaminants in Effluent Water
from Pennsylvania Brine Treatment Facility, Josephine Plant
CHEC conducted sampling of wastewater as it was discharged into Blacklick Creek, Indiana
County, Pennsylvania from the  PBT-Josephine  Facility on December 10, 2010. Samples were
taken at 3-hour intervals over the course of one 24-hour period. These samples were analyzed
for listed inorganic and organic species by R. J. Lee Inc, a PA State Certified Laboratory
(Certificate # 006).

The concentrations of analyzed contaminants in this effluent of primary environmental public
health importance, which may also stress aquatic life, include: barium (Ba) [mean, 27.3 ppm;
maximum, 37.0 ppm];  bromides (Br) [mean, 1068.8 ppm; maximum, 1100.0 ppm; strontium (Sr)
[mean, 2983.1 ppm, maximum  3120.0 ppm]; benzene [mean 0.012 ppm; maximum  0.013 ppm]
and 2 butoxyethanol (2-BE) [mean 59ppm; maximum 66 ppm]. Contaminant concentrations of
ecological and secondary drinking water importance include: chlorides (Cl) [mean 117,625 ppm,
maximum 125,000 ppm]; magnesium (Mg) [mean 1247.5 ppm; maximum 1300.0 ppm]; total
dissolved solids (TDS) [mean  186,625 ppm; maximum 190,000 ppm]; sulfate (S04) [mean 560
ppm; maximum 585 ppm], and  pH [mean 9.58 units; maximum 10 units].
                                      10

-------
Comparisons of Effluent Contaminant Concentrations to Standards, Guidelines
and Criteria set by Federal and State Regulatory and Investigative Agencies for
the Protection of Human and Aquatic Health
Levels of contaminants in effluent from the PBT- Josephine Facility were interpreted according
to comparisons with applicable federal and state standards and recommended guidelines for
both human and aquatic health.  Barium had a mean concentration in effluent of 27.3 ppm
(maximum of 37 ppm); this is approximately 14 times the United States Environmental
Protection Agency (EPA) maximum concentration limit (MCL) of Ba in drinking water of 2 ppm.
The EPA consumption concentrations 'water and organism' and 'organism alone' for barium are
both 1 ppm. The levels  of barium in the effluent are over 27 times these consumption
concentrations. The U.S. EPA criteria maximum concentration (CMC) and the EPA criteria
continuous concentration (CCC), both for protection of aquatic health,  are 21 ppm and 4.1 ppm,
respectively; the mean  level of barium in effluent exceeds these criteria by 1.3 and 6.7 times,
respectively. The mean level of barium in effluent water was 4, 4.73, and 9 times the derived
drinking water minimum risk level (MRL) for intermediate  and chronic exposures in adult men,
adult women, and children, respectively.

The EPA recommended limit for strontium (Sr) in finished municipal drinking water is 4  ppm.
The mean concentration of Sr  in  PBT-Josephine effluent water is 2981.1 ppm (over 745  times
the recommended level). The MRL for Sr set by the Agency for Toxic Substances and Disease
Registry (ATSDR) for intermediate length exposure  is 2 mg/kg of body weight/day. The sampled
mean level of strontium in PBT-Josephine effluent water was over 43, 51.7, and 97.9 times the
derived drinking water  MRL for intermediate exposures in adult men, adult women, and
children, respectively.

Bromide in water is of concern because of its ability to form brominated analogs of drinking
water disinfection by-products (DBP). Specifically, bromide can be involved in reactions
between chlorine and naturally occurring organic matter in drinking-water, forming brominated
and mixed chloro-bromo byproducts, such as trihalomethanes or halogenated acetic acids.
There is general agreement that  bromide levels in fresh-water sources  be kept below about 100
ppb. The PBT- Josephine facility discharged effluent into Blacklick Creek with a measured mean
concentration of bromide of 1068.8 ppm, which is 1,068,800 ppb. This  is 10,688 times the 100
ppb level at which authorities  become concerned.

The mean level of benzene, a known carcinogen, in outfall effluent from PBT-Josephine was
0.012 ppm or 12 ppb. The drinking water MCL for benzene is 5 ppb, thus effluent levels were
above twice the drinking water MCL. The EPA consumption, water and  organism risk level for
benzene is 2.2 ppb in water; the  mean level of benzene in  PBT-Josephine effluent water is
almost 6 times this criterion. The organism-only risk level for benzene is 50 ppb in water, and
the mean level of benzene in effluent water is 24%  of this guideline. The measured Benzene
value was 0.6 times, 1.2 times, and 1.5 times the derived drinking water MRL for chronic
exposures in adult men, women  and children, respectively.
                                         11

-------
2-butoxyethanol (2-BE) is a glycol ether and is used as an anti-foaming and anti-corrosion agent,
as well as an emulsifier in slick-water formulations for Marcellus Shale gas extraction. The mean
concentration of 2-BE in the effluent exceeded derived drinking water MRL's for 2-BE for both
acute and intermediate exposure for adult men and women and children.

Contaminants with secondary MCLs (SMCL) and aquatic receptor effects that were measured in
the PBT-Josephine Facility effluent include magnesium, manganese, chlorides, sulfates, and
total dissolved solids (IDS). Magnesium was found in the effluent with a mean concentration of
1,247.5 mg/L, which is 24,950 times the EPA Mg SMCL of .05 mg/L. The mean concentration of
Manganese in the effluent was .08 mg/L, and the SMCL for Manganese concentration in
drinking water is .05 mg/L. Therefore, the SMCL for manganese concentration is 62.5% lower
than the concentration in the effluent. The mean concentration of chlorides in the sample
analysis was 117,625 mg/L, which is 470.5 times the SMCL for chlorides in drinking water of 250
mg/L. To protect aquatic communities, the criteria maximum concentration (CMC) for chlorides
in surface water is 860 mg/L, and the criteria continuous concentration (CCC) for chlorides in
surface water is 230 mg/L. The mean concentration of chlorides measured in samples was 138
times the CMC and 511 times the CCC. The mean concentration of sulfates in the sample
analysis was 560 mg/L, or 2.2 times the SMCL for sulfates in drinking water (250 mg/L). The
SMCL for total dissolved solids (TDS) in drinking water is 500 mg/L, and the mean concentration
of TDS measured  in samples was 186,625 mg/L, or 373 times the SMCL.

Levels of strontium and 2-BE exceeded the NPDES reporting requirement set by the
Pennsylvania DEP of 100 ppb and 500  ppb, respectively, for discharge of a toxic substance
regularly or irregularly, respectively.

Masses of Contaminants Entering Blacklick Creek
CHEC has information from the Pennsylvania Department of Environmental Protection  (PA DEP)
that the PBT - Josephine Facility treated 15,728,241 gallons of oil and gas wastewater in the 6
month period from July 1, 2010 to December 31, 2010. Using this figure as the amount of
effluent wastewater exiting the Josephine outfall and using the mean level of each contaminant
found in the effluent over the sampling period of the study, the masses of contaminants with
important human and ecological consequences discharged from the PBT, Josephine Facility into
Blacklick Creek in the last 6 months of 2010 are projected to be: barium -1627 kg (3588
pounds); strontium -177,712 kg (391,856 pounds; 196 tons); bromides -63,708 kg (140,476
pounds; 70.2 tons); chloride - 7,011,631 kg (15, 460,646 pounds; 7,730 tons); sulfate - 33,382
kg (73,607 pounds; 36.8 tons); 2-BE - 3517 kg (7,755 pounds; 3.88 tons); and TDS - 11,124,733
kg (24,530,036 pounds; 12,265 tons).

Potentially Exposed Populations
Recreationalists are at high risk of being exposed to outfall contaminants through ingestion,
inhalation and through dermal exposure. The outfall  of the Josephine Facility is easily accessible
to users of nearby rails-to-trails pathways, and there are indications that anglers frequent the
                                         12

-------
area.1 Additionally, children wade and swim in the creek during warmer weather, and regional
watershed websites indicate that paddlers use the creek for canoeing and kayaking. 2-BE
released into Blacklick Creek may be ingested by swimmers in the creek, as this pollutant can
become airborne and present an inhalation hazard to anglers, swimmers and boaters. It is also
taken in to the body via dermal absorption. Anglers catching and eating fish from upstream or
downstream of the effluent outfall are at risk for exposure to multiple contaminants that were
sampled in this study.

CHEC has developed maps showing numerous private water wells in the immediate vicinity of
Blacklick Creek downstream from the effluent discharge. Private well water users are at risk of
exposure to contaminants in effluent being released into Blacklick Creek because these private
wells may capture water from the creek when the well  pump rate is sufficiently high. High
pump rates can occur especially during peak  usage by residents.

The first identified municipal drinking water intake downstream of this discharge is at Free port,
Pennsylvania on the Allegheny River. Populations served by the Freeport authority and  water
authorities downstream of Freeport are at potential risk for exposure to contaminants
identified in effluent, as well as other contaminants in Marcellus Shale flowback water that
were not sampled for in this study.

Implications of Effluent Discharge from the PBT - Josephine Facility Discharge
for Exposures to Other Contaminants Known to be Present in Marcellus Shale
Flowback Fluids and a Regional Appreciation of These Results
Of particular environmental public health significance is that Marcellus Shale flowback water
contains other contaminants, in addition to those analyzed for in this study, which have health
consequences  if ingested, inhaled,  and/or absorbed through the skin. While we make no
statements regarding the presence of other contaminants in this effluent water being
discharged into Blacklick Creek, it is imperative that additional testing be conducted
immediately by federal  and state health and  enforcement agencies to determine if other
contaminants of public  health significance are entering this watershed.

Oil and gas wastewater and Marcellus shale flowback fluids are being disposed of in "brine
treatment" facilities and at Publicly Owned Treatment Works (POTW's) throughout the
Commonwealth of Pennsylvania and in Ohio, West Virginia, and New York. Therefore, the
ramifications of disposal of large quantities of oil and gas wastewater through ineffectual brine
treatment plants and POTW's needs further evaluation throughout the region to determine its
impact on stream and river systems and public drinking water supplies, as well as to
recreationalists and  private well water users.
1 Blacklick Creek has been classified as a 'trout stocking" stream.
 US EPA. 2004. "Evaluation of impacts to underground sources of drinking water by hydraulic fracturing of coalbed methane
reservoirs (Final)." Office of Water, June.
 New York State Department of Environmental Conservation (NYSDEC). 2009. "Draft Supplemental Generic Environmental
Impact Statement on the Oil, Gas and Solution Mining Regulators Program—Well Permit Issuance for Horizontal Drilling and

-------
Recommendations
   •   Operations at this plant should be halted until all contaminants of human and aquatic
       health concern in accepted oil and gas fluids are known and it can be determined that
       the treatment processes used at the plant effectively remove these contaminants from
       the fluids being treated.
   •   All approaches to the effluent discharge area and a reasonable distance downstream (at
       least 100 meters) from streamside and landside should be posted with warning signs.
       These signs should discourage any use of and/or contact with stream water.
   •   An advisory to all anglers should be issued stating that fish taken  from this stream, both
       upstream and  downstream may be contaminated in order to discourage fish take and
       consumption.
   •   Studies to determine the levels of all potential Marcellus Shale flowback fluid
       contaminants  in  downstream water, sediments and pore water should be undertaken
       immediately.
   •   Residential and other private well water users downstream of the effluent outfall of the
       PBT-Josephine Facility should be advised that there may be contaminants in their well
       water and discouraged from using it for drinking, cooking or bathing until such water is
       tested for continuous safe use.
   •   Municipal water authorities downstream of this outfall should be notified of the
       contaminants found in effluent from the PBT- Josephine Facility, of other possible
       contaminants  in  Marcellus Shale flowback fluids and oil and gas wastewater, and that
       there are other treatment facilities and POTW's in the Blacklick, Conemaugh, and
       Kiskikiminetas drainages that accept and discharge oil and gas waste fluids into surface
       water.
   •   All municipal water authorities at reasonable distances downstream of "brine
       treatment" and POTW's  accepting Marcellus Shale flowback fluids and other oil and gas
       wastewater in the region extending eastward across Ohio, Pennsylvania and West
       Virginia and New York should be notified of these results.
   •   The  PA DEP and other state and federal regulatory authorities should immediately
       review all surface water discharge permits granted to brine treatment facilities and
       POTW's that accept Marcellus Shale flowback fluids and oil and gas wastewater, to
       ensure that 2-BE concentrations being discharged are below all applicable standards,
       guidelines and criteria. This review should be informed by results of this report but
       should be extended to all known contaminants in flowback and other oil and gas
       wastewater.
                                          14

-------
       Tracing & Associated Media Composition in Colorado
                   Andrew A Havics1, CHMM, CIH, PE and Dollis Wright2
                                          , LLC
                                       2QEPA

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.
The fracing process begins with wellpad siting, proceeds through completion and ends with
production (and the eventual decommissioning or abandonment). The steps in the process
include: site selection, well pad development, drilling, fracing, and production. The chemical
composition of media during the fracing as well as naturally occurring constituents present in
the natural resources all add to the constraints and characteristics of fate, transport, exposure
and projected risk. The classical risk assessment process can be divided roughly into: 1) Hazard
Identification, 2) Dose-Response Determination, 3) Exposure Assessment, and 4) Risk
Characterization, followed by Risk Management (including policy development) and preceded
as well as intermingled with research (NRC, 1983, 1994). Within the Exposure Assessment
aspect is the fate and transport of chemicals and subsequent exposure. In terms of this risk
framework, chemical composition plays a strong initial role in  Hazard Identification but is also
relevant in terms of dose-response, exposure pathway determination, fate and transport
property selection, and risk assessment. The process and consideration of chemical selection
are presented with  regard to the investigation of fracing impact in four energy basins in
Colorado completed in the spring of 2008. The focus will be on three media, flowback material,
frac fluids, and produced waters, although other media and subsequent pathways were
considered and are  discussed in part here.

Risk Assessment in Regards to the Identification of Constituents for Analytical
Evaluation
As just mentioned, the  identification and eventual selection of chemicals for consideration in
risk assessment is part of the classic risk assessment (RA) process (NRC 1983, 1994). Although
the identification and selection of chemicals seems limited in scope  and limited in interaction to
hazard identification, it can be driven by regulatory requirements or public concerns. It can also
impact the cost, eventual selection of exposure pathways,  and bring to light underlying issues in
the RA process and  the interweaving of policy with the science.

The steps in the process of drilling and fracing produce a variety of media (frac fluids, produced
water, waste pit solids, etc.), all of which should be considered in a holistic approach to both
understanding and managing risk in the Oil & Gas (O&G) Industry.

Selection of Chemicals for Analysis
In 2008, in a project funding by the Colorado Oil and  Gas Association (COGA), QEPA, pH2

                                         15

-------
through QEPA, and URS were contracted to devise a sampling and analytical plan as part of a
risk assessment for proposed changes in the O&G regulations in Colorado (COGCC, 2008). This
first necessitated the identification of chemicals for analytical evaluation (CAE) and eventual
selection of the chemicals (URS, 2008). The Material Safety Data Sheets (MSDS) were gathered
by requesting input from 7 of the COGA member companies and reviewing the submissions.
More than 100 products were looked at and broken into reported relative sub-quantities (%) by
Chemical Abstract Service (CAS) number. MSDS are required to report any standard listed
chemicals that make up more than 1% of the chemical composition, or >0.01% if the chemical is
carcinogenic (OSHA, 2008). From this, the beginnings of a CAE list was produced, which
included glycols and pH as a surrogate for acids and bases. A list of standard chemicals of
interest in the O&G industry (BTEX, PAHs, Boron, Chloride, etc.) were also added (CDPHE, 2007;
COGCC, 2008). Because of a lack of clarity with regard to total extractable petroleum
hydrocarbons (TEPH), these were analyzed for both Diesel Range Organics (DRO) and Motor Oil
Range Organics (MRO)  to allow relative  hydrocarbon grouping if desired. Primary metals (and
metalloid) included were the eight Resource Conservation and Recovery Act (RCRA) metals
consisting of arsenic, barium, cadmium, chromium, lead, mercury, silver, and selenium. An
additional fifteen target analyte list (TAL) metals were also added from the EPA Method 6020A
Method list and included Aluminum, Antimony, Beryllium, Calcium, Cobalt, Copper, Iron,
Magnesium, Manganese, Nickel, Potassium, Sodium, Thallium, Vanadium & Zinc. Based on a
review of potential agents associated with raw material derived from subsurface deposits, gross
alpha and gross beta were selected. Analysis for specific isotopes at this phase was considered
unwarranted by the risk assessor. Consideration for chemicals recommended by local
environmental groups was also undertaken; this was also accounted for post-sampling in terms
of reviewing tentatively identified compounds (TICs) for relevance and comparing to groups of
chemicals used, and the MSDS product list. Basin usage for the products was also recorded in
the event that significant findings relevant to geologic formation(s) was(were) discovered. A
small subset of samples was analyzed by EPA's Toxic Characteristic Leaching Procedure (TCLP)
for eight RCRA metals,  plus pH, reactive sulfides, and reactive cyanides to evaluate waste
disposal considerations.

Statistical Considerations
In any sampling plan, several consideration with regard to chemicals should be made, including,
but not limited to: a) number of samples for the intended use, b) minimum limit of detection
(LOD) and limit of quantification (LOQ) for a constituent, c) background levels, d) level of quality
of sampling and analysis, e) statistical application, and f) appropriate selection of a method.
One should never take  a sample before  knowing what one will communicate once the results
are in. The desire would be to collect multiple samples from each media and ones that are
representative of a typical media by energy basin and type of drilling or fracing operation. For
example, one should collect both early and late flowback samples and one should  consider the
regulatory impact of required oil-water separation in certain basins, etc. In terms of return on
statistical data, five to six samples (base on a normal distribution)  place the mean  as reasonably
estimated. However, to reasonably estimate the standard deviation more than twenty samples
would be necessary.  In the case of RA, much of the decision making is driven in orders of
magnitude (QEPA, 2008). Therefore, five to six samples per media setting provides a good basis

                                          16

-------
to work from, even in consideration of right-skewed distribution (e.g., lognormal). The
subsequent quasi-policy and quasi-science decision of selecting an appropriate estimator
(mean, max, upper confidence limit one sided at 95% [UCL1,95]) can then be utilized by a risk
manager. In terms of LOD and LOQ for analytical method selection, regulatory levels in
Colorado (COGCC, 2008; CDPHE, 2007) were reviewed along with risk assessor pre-estimates of
effective dilution-attenuation factors to determine relevant methodologies. Thus, by
recognizing end risk calculation relevancy, PAH detection levels were set at standard levels,
which are higher than that achievable by more sophisticated (and costly) methods.

Media and Sample Collection
Sample quality is important, particularly if the analysis has broader policy implications.  In the
2008 project in Colorado, URS personnel collected the samples independent from the risk
assessors (QEPA & pH2) and independent from the labs used for analysis. Statistical analyses
were performed by both URS and pH2/QEPA with pH2 directing the parameters. Samples were
collected at a variety of sites-55 in all—to represent  four energy basins in Colorado; these
basins are the Denver-Julesburg (DJ), Piceance, Raton, and San Juan. The media types selected
were in consideration of the RA and included: pit solids, pit fluids, drill fluids, frac fluids,
flowback fluids, produced water and background soils. It should be noted that some of  the
materials are co-mingled with other fluids and moved between pits as multiple-pads or sites are
developed. Sample analysis for both solids and liquids were separated, and sets of dissolved
and total constituent analyses were performed in most cases. Decisions for sample media
categorization (fluid/solid) were ultimately determined by the risk assessor after input from
URS and the lab.

Samples were collected at points representative  of current drilling and fracing operations, both
near and  away from residences and within differing  hydrogeologic and geologic conditions.
Multi-point composite samples were collected to achieve better representation (except for
VOCs due to potential constituent loss). There was a high frequency of co-located field
duplicates (45%) and MS/MSDs (30%) QC samples collected. In addition, rigorous paper, photo,
and video documentation were also performed to incorporate with the Level IV Quality Control
(QC) data packages and analysis by an NELAC certified laboratory. Also, samples were analyzed
for >170 constituents using EPA-approved methods  (e.g., VOCs by 8260, SVOCs by 8270, Glycols
by 8015). Thorough data validation was performed resulting in >99% data usability.  Other
Precision, Accuracy, Representativeness, Completeness, and Comparability (PARCC) parameters
were satisfactory (URS, 2008).

Analytical Results for Flowback Materials, Frac Fluids, and Produced Waters
Analytical sampling results from the COGA study completed in 2008 represented  more  than
52,000 data points for pit solids, liquids, fracing fluid, flowback, and drilling fluids. Analytical
data included BTEX, PAHs, metals (primary eight  RCRA plus secondary), gross alpha, gross beta,
boron, and glycols.
                                          17

-------
Chemicals not Detected
For the solid media, 43 VOCs were reported as Not Detected (ND) in every solid sample, as
were 57 semivolatile compounds (SVOCs). In addition, reactive cyanide and reactive sulfide
were not detected for reactivity, corrosivity, and ignitability (RCI) analyses performed on solid
samples.

The list of non-detect (ND) constituents for liquid media was not as extensive as the list for
solids. A total of 39 VOCs and 48 SVOCs were reported as ND in every sample that was analyzed
as a liquid. Although the liquids list is shorter, not every constituent that was reported as ND for
liquid samples was also reported as ND for solids. In total, the two lists share 81 common
constituents, including 35 VOCs and 46 SVOCs. Reactive sulfide and cyanide are not included in
this figure because RCI analyses were not performed for liquid samples.

Flowback
A total of twenty four base samples (plus duplicates) of flowback fluids were collected and
analyzed throughout the four basins. One of the flowback fluid samples collected in the DJ
Basin was analyzed as a solid due to the high amount of suspended sediment present in the
sample. The following constituents were detected in 100 percent of the flowback fluid samples:
barium, benzene, boron, chloride, ethylbenzene, naphthalene, nickel, toluene, total xylenes,
trimethylbenzene (TMB), and TEPH. A few constituents of significance were often  below
detectable levels, for example: 37% of gross beta, 69% of gross alpha, and 84% of anthracene
were ND, whereas ones like BTEX, 1,2,4-TMB, and 1,3,5-TMB, boron, and chloride  were always
detected. Although the max concentrations for flowback fluids were 270 and 4,030 pCi/L for
gross alpha and beta, respectively, the comparable pit fluids were only 17 and 174 pCi/L,
respectively

Frac Fluids
Two frac fluid samples (plus a duplicate) were collected and analyzed in the Piceance  Basin. One
of the frac fluid samples collected was analyzed as a solid due to the high amount  of suspended
sediment present in the sample. The following constituents were detected in 100 percent of the
frac fluid samples: barium, benzene, boron, chloride, ethylbenzene, gross beta, naphthalene,
nickel, sulfate, toluene, total xylenes, TMB, and TEPH.

Produced Water
Produced water samples were collected in the  Raton and San Juan Basins. Altogether, 10
produced water samples (plus duplicates) were collected between the two basins. The
following constituents were detected in 100 percent of produced water samples: barium,
boron, chloride, and nickel. These constituents occur naturally in formation waters, and at least
a portion of the detected concentration for each constituent is likely due to natural
background.

In regard to other PCOCs, benzene was detected in 5 produced water samples; ethyl benzene
was detected in  3 produced water samples; naphthalene, toluene, total xylenes, TMB, and

                                          18

-------
TEPH were detected in 4 produced water samples. In general, the PCOCs were detected at a
higher frequency in produced water from the San Juan Basin than from the Raton Basin.

Chemicals Matching MSDS Constituents
Only 8 constituents out of more than 100 found to be present in MSDS reviewed. The
constituents found include: propanol, 2-butoxyethanol, ethylene glycol, n-heptane,
isopropanol, naphthalene, 1,2,4-TMB, and ethanol. The constituents detected in most of the
media are 1,2,4-TMB and naphthalene. Solvents and fracing agents were the most common use
of these 8 constituents in the O&G industry in CO.

As a caveat, the detection of a chemical listed in an MSDS for a product in a particular media
does not necessarily mean that it came from that product.  It only means it could have come
from it, but for some it just as easily could have come from a natural occurring deposit. Also, a
chemical's presence does not mean that it is a significant risk either.

Narrowing the Selection  for Risk Assessment Purposes
From the CAE, a select group of those chemicals meeting one of the following characteristics
were selected for assessment of risk: a)  present in either significant amounts (near the
proposed COGCC Table 910-1 values for instance), b) or those with a significant frequency of
presence (e.g, TMBs, BTEX, most metals), or c) those with a concern because of significant
usage (e.g., glycols, barium, chloride), or those thought to be of concern but having little prior
test data (e.g., PAHs, gross alpha and gross beta).

These were then considered as constituents in the following media scenarios:

      Solids placed in Pits
      Liquids in Subsurface Pits
      Fracing Fluid  placed in pits
      Fracing Fluid  placed in containers
      Produced Water placed in containers
      Produced Water placed in pits
      Drilling fluids in drilling
      Drilling fluids in pits

Details of the RAare provided in QEPA, 2008.

Limitations
As with any assessment there are a number of gaps or limitations imposed or resulting from the
manner in which this RA was commissioned. The first is that groundwater as a resource was the
prime focus, thus air was not  considered in as great as detail; nor is it relevant for the scope this
workshop. Secondly, samples were from operations in place in Spring of 2008, not prior,  nor
post. Other seasons  may result in different concentrations, e.g., VOCs. Some  practices from the
past (diesel fuel in the drilling) or more common today (treatment or recycling of produced

                                         19

-------
waters) were not accounted for. Also, only produced water data from 3 energy basins was
collected and there was limited drilling fluid & frac fluid samples compared to that desired by
the risk assessor.

In terms of other RA fate & transport aspects, there was no verification of actual depth to
groundwater on-site, no verification of actual hydrogeologic properties (hydraulic conductivity,
head difference), although neighboring data was gathered and evaluated. All estimated
exposure doses for base risk were therefore modeled using conservative parameters for the
potential chemicals of concern (PCOCs), i.e, maximum or UCLi;95°/0.

There are other limitations created by classical risk assessment guidance (ASTM, 2002; EPA
1989a, 1989b, 1991a, 1991b), which includes failure to consider background concentrations
(e.g., arsenic), basic solubility under site-specific situations, and toxicological constraints. One of
the toxicology constraints is that barium sulfate, normally used as a drilling fluid, is expected as
the primary source of the barium, yet it is neither very soluble in many instances, nor very toxic
compared the BaCI2 upon which the Reference Dose (RfD) for the element is based (EPA, 2005).
In RA it is also important to gauge the general level of influence of one parameter versus
another in a  RA. For chemicals, the RfDs, Reference Concentrations (RfCs), Benchmark Dose
Levels (BMDL), Minimal Risk Levels (MRLs), and Slope Factors (SFs) drive the primary order of
magnitude of risk, and generally use safety-uncertainty margins in the range of 10-100 already.
For instance  BaCI2 uses a safety-uncertainty factor of 300 (EPA, 2005). If is useful to consider
this in ranking the value of chemicals and their health hazards in scenarios like the one
presented here. In the cases where there is no relevant acceptable value for toxicity, other
means, such  as a control banding approach, could be applied (Nelson, et al., 2011).

Acknowledgments
The study upon which a large portion of this presentation is based was funded by the Colorado
Oil and Gas Association (COGA). I would also like to acknowledge Stacey Malerba, Project
Manager and Mark Leverson, PG, Principal Hydrogeologist, with  URS  for their work and
assistance on the sampling, analytical, databasing, and hydrogeology aspects. In addition, Mike
Paules, EH&S Manager, Williams Production RMT Company; Gerald Jacob, PhD, Environmental
- Regulatory Manager, Pioneer Natural Resources USA Inc.; Scot Donate, Bill Barrett
Corporation, are to be commended for sharing their field expertise in pits and processes. And
finally, Ms. Kim Zielinski, CIS Specialist, from URS, who kindly provided some additional
background data forgeostatistic application in the project.

References
ASTM: £1739-95(2002), Standard Guide for Risk-Based Corrective Action Applied at Petroleum
       Release Sites. ASTM, Conshohocken, PA. 2002.
Colorado Department of Public Health and Environment, Hazardous Materials and Waste
       Management Division, Table 1 Colorado Soil Evaluation Values (CSEV) - December 2007
Colorado Oil  and  Gas Conservation Commission, Draft Rules for Oil and Gas Development in
       Colorado, (HB 1298 & HB 1341), March 31, 2008.
                                          20

-------
EPA: Risk Assessment Guidance for Superfund, Volume I, Human Health Evaluation Manual,
      Part A, Interim Final. EPA 540/1-89/002. USEPA: Washington, DC. 1989a.
EPA: Risk Assessment Guidance for Superfund, Volume II, Environmental Evaluation Manual,
      Interim Final. EPA 540-1-89-001. USEPA: Washington, DC. March, 1989b.
EPA: Risk Assessment Guidance for Superfund, Volume I, Human Health
Evaluation Manual, Part B: Development of Risk-Based Preliminary Remediation Goals, Interim.
      EPA 540/R-92/003. USEPA: Washington, DC. December, 1991a.
EPA: Risk Assessment Guidance for Superfund, Volume I, Human Health Evaluation Manual,
      Part C: Risk Evaluation of Remediation Alternatives. USEPA: Washington, DC. December,
      1991b.
EPA: Toxicological Review of Barium and Compounds (CASRN 7440-39-3), EPA-635-R-05-001,
      June, 2005.
National Research Council (NRC): Risk Assessment in the  Federal Government: Managing the
      Process. NAP, Washington, DC. 1983.
National Research Council (NRC): Science and Judgment in Risk Assessment. NAP, Washington,
      DC. 1994.
Nelson, Deborah Imel, Stephen Chiusano, Anne Bracker,  Lance Erickson, Charles Geraci, Martin
      Harper, Carolyn Harvey, Andrew Havics, Mark Hoover, Thomas Lentz, Richard Niemeier,
      Susan Ripple, Erica Stewart, Ernest Sullivan, and David Zalk: Guidance for Conducting
      Control Banding Analyses. AIHA, Fairfax, VA. 2007.
OSHA: 29 CFR 1910.1200 (Hazard Communication Standard). July 1, 2008.
QEPA: Pathway Analysis and Risk assessment (PARA) For  Solids and Fluids Used In Oil and Gas
      Exploration and Production in Colorado, pp. 1-930, June 2008.
URS: Field Activities Report for Characterization of Exploration and Production Pit Solids and
      Fluids in Colorado Energy Basins, June 4, 2008.
                                         21

-------
   Comparison of Hydraulic Fracturing Fluid Composition with
 Produced Formation Water following Fracturing - Implications
                           for Fate and Transport
                                  Debra McElreath
                           Chesapeake Energy Corporation

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.
Chesapeake Energy Corporation (Chesapeake) as a part of an evaluation of produced formation
water had examined the composition of hydraulic fracturing chemicals used in natural gas
production wells. Samples of the hydraulic fracturing fluid prior to the addition of proppant had
been analyzed as well as time series samples of produced formation water for a natural gas well
located in a shale formation west of the Mississippi River (Location 1) and another well in the
eastern U.S. (Location 2). Some of the data evaluated is subject to attorney-client privilege
(hereinafter "privileged data"). The major conclusions from Chesapeake's evaluation of
produced formation water data are:
   •  Produced formation water is highly variable within and between shale plays.
   •  The elevated temperature and pressure affect the fate and transport of hydraulic
      fracturing fluid components and can create safety issues for sampling.
   •  Analytical techniques used for chemical and radiochemical characterization of the
      produced formation water must be robust to the matrix interferences presented by
      total dissolved solids.
   •  The most reliable sentinel compounds appear to be total dissolved solids, chloride and
      divalent cations. The concentrations of these components are related to each other and
      are also related to the formation water volume.
   •  The concentration of total dissolved solids is predictive of the concentration of other
      species.
   •  The concentration of indicator parameters for hydraulic fracturing mixtures, such as
      chloride, sulfate, and boron, are overshadowed by the naturally-occurring
      concentrations of these parameters in formation water.
   •  The presence of NORM is delayed and associated with higher percentages of formation
      water in the produced water with increasing time.

Based on the predicted  downhole behavior of the hydraulic fracturing fluid, Chesapeake
designed a sampling program for hydraulic fracturing fluid and produced formation water in
order to understand the fate of hydraulic fracture chemicals in the produced formation water.
The sampling design incorporated a review of chemicals used in two Chesapeake wells during
hydraulic fracturing. Produced formation water samples were taken in a time series, hours to
days, following hydraulic fracturing. The analyte list Chesapeake utilized was the complete list
found at 40 CFR Part 264, Appendix IX, and commonly associated  with groundwater monitoring

                                        22

-------
supplemented with three glycols. All of the analyses were conducted using EPA analytical
methods and were performed by a single NELAC-certified laboratory.
Table 1. Predicted Downhole Fate for Hydraulic Fracturing Fluid Components
      Component
          Acid
Reacts with minerals to create salts, water and CO:
   Cor os o  Inhibit           with pipe surfaces, broken down by micro-organisms or returned in
                      produced formation water
      Iron Control      Reacts with minerals to create salts, water and CO?

      Anti- Bacteria I     Broken down by micro-organisms or small amount returned in formation
         Agent        produced water

     Scale Inhibitor     Attaches to the formation, majority returns with produced formation water

    P   »   o          Remains in formation, broken down by micro-organisms or small amount
                      returned  m formation produced water
       Surfactant
Returned with produced formation water or produced natural gas
      Gelling Agent     Broken down by breaker and returns with produced formation water
        Breaker
Reacts with 'gel' and "crossknker" to form ammonia and sulfate salts which
are returned in produced formation water
       Crosslmker
Combines with the "breaker" m the formation to create salts that are returned
in  produced formation water
  Sourc* CheMpMfcr 2010  MvctOui Sh«l« Hydraulic Fncturmg Fact Sheet Ji4y
Since it was predicted that many of the hydraulic fracture fluid chemicals would undergo
transformation, the focus of this paper is indicator parameters, such as total Kjeldahl nitrogen,
ammonia, sulfate, and sodium, which are used as surrogates for the presence of breakdown
products of the hydraulic fracturing chemicals. In many cases, analytical methods are not
available to analyze a compound but instead the analysis is conducted on a compound's
                                           23

-------
predicted components since these results can be combined to provide an estimate of the
compound's concentration. Focusing on indicator or sentinel parameters is also cost effective
and does not require use of exotic or yet-to-be developed methodology.

There are significant issues regarding sampling of produced formation water. Natural gas is
contained under high pressure in specialized equipment that is not designed for producing high
quality environmental samples. The sample matrix itself presents challenges such as foaming
and changes in surface tension. Analytical techniques are also impacted by the presence of
elevated concentrations of total dissolved solids and chloride. The inorganic and wet chemistry
methods were most affected by the presence of high total dissolved solids. EPA Method 8015
has been found to be insufficiently robust to overcome the  matrix issues which are attendant to
the analysis of produced formation water. The range of detection limits seen in the available
data sets ranged from 10,000 to 50,000 u.g/L. It does not appear that the glycol methods can
provide meaningful results for samples with these matrix issues. EPA Method 8270 has some
utility for larger glycols, such as glycol ethers; however for the smaller, more soluble, ethylene
and diethylene glycols, the extraction methods are not useful. Radiochemistry methods are
particularly affected by the elevated concentrations of barium and total dissolved solids. Since
hydraulic fracture fluid is a product rather than an environmental sample, certain standard
requirements for environmental samples, such as holding time, are not applicable.

Hydraulic fracturing service vendor-supplied data was available to compile the water volume,
proppant volume, and specific hydraulic fracturing products and related volumes used for
fracturing of the individual well. Component information for each of the hydraulic fracturing
products was drawn from the Material Safety Data Sheets. These data were utilized to calculate
the concentrations of the  individual chemicals used in the hydraulic fracture fluid for each of
the two subject wells. For some, surrogate chemical species, e.g. sodium or chloride, were
calculated for comparative purposes.

Produced formation water volumes have been estimated for the Location 1 well. Typically,
produced water volumes decrease markedly with time once a natural gas well is in production.
In the first ten days of production, about 600,000 gallons or 60,000 gallons per day are
produced; between days 11 and 365, the volume drops to approximately 8,400 gallons per day.
In subsequent years of production, the well would be expected to produce about 4,200 gallons
per million standard feet of gas (MMCF) of gas produced average for the life of the well.
Chesapeake used approximately 4 million gallons of water for drilling and fracture stimulation.

For the Location 1 well, a hydraulic fracture fluid containing the following compounds was
utilized: biocide (gluteraldehyde and an alcohol); beaker (ammonium persulfate); iron control
(sodium compound); friction reducer (polymer and a hydrocarbon); crosslinker (polyol and
borax); corrosion and scale inhibitors (alcohols, organic acids, and sodium salt of a polymer);
and acid (hydrochloric acid). Therefore, boron, sodium, sulfate, and total Kjeldahl nitrogen were
expected to be useful as surrogate analytes.
                                          24

-------
Analytical results were available for the hydraulic fracture fluid (pre-injection) and produced
formation water samples in time series for 6 hours, 1 day, 2 days, 9 days, and 30 days following
fracturing. The results for several surrogate/indicator species, total Kjeldahl nitrogen, ammonia,
chloride, total dissolved solids, sodium, boron and glycols, were summarized. Data were also
available for two radium isotopes (Ra-226 and Ra-228). Since the hydraulic fracture fluid did
contain a hydrocarbon carrier, benzene and toluene were also evaluated (most other volatile
organic compounds and semi-volatile organics were below detection limits).

In general, the concentrations for the analytes of interest increased, especially the total
dissolved solids, chloride and sodium, with time following hydraulic stimulation. These
increases reflect increasing percentages of formation water entering the produced water
volume. It should be noted that the sodium, chloride and boron concentrations in hydraulic
fracture fluids were rapidly overshadowed by the naturally-occurring concentrations of these
compounds on formation water. The increasing concentrations of the nitrogenous compounds
probably reflect degradation of the nitrogen compounds in the hydraulic fracture fluid and
microbial activity.

The results for the divalent cations, barium and  strontium concentrations were examined for
relationship with total dissolved solids concentrations. The correlation coefficient for barium
and total dissolved solids was 0.998 and that for strontium and total dissolved solids was 0.935.
The correlation between chloride and total dissolved solids was 0.943. The relationship
between total dissolved solids and radium-226 and radium-228 appear to be similar to that of
the other divalent cations. It is evident that the  presence of total dissolved solids  can  be used as
a sentinel parameter.

The calculated concentration of glycols was 55,000 u.g/L; however, the analytical result for the
hydraulic fracture fluid was 35,800 u.g/L. As the  total dissolved solids increased in the 12 hour
and day 2 samples, the detection limit for EPA Method 8015 increased to <10,000 u.g/L These
results point to the limitations of Method 8015 for glycol analyses in produced formation water.

Produced formation water volumes have been estimated for the Location 2 well located in the
eastern U.S. Typical produced  water volumes decrease markedly with time a natural gas well is
in production. In the first ten days of production, about 600,000 gallons or 60,000 gallons per
day are produced; between days 11 and 365, the volume dropped to approximately 8,400
gallons per day. In subsequent years of production, the well would be expected to produce less
than 200 gallons per MMCF average for the life  of the well. Chesapeake used approximately 3.4
million gallons for fracture stimulation.

For the Location 2 well, the hydraulic fracture fluid contained the following: biocide (sodium
salt, sodium  hydroxide, and a bromide salt); breaker (sodium and potassium salts); iron control
(citric acid); friction reducer (water soluble nitrogenous-based polymer); gel (guar gum, a
hydrocarbon, and polymer); corrosion and scale inhibitors (alcohol, glycol and an  amide);
surfactant (alcohol, glycol and a hydrocarbon); and acid (hydrochloric acid). Sodium, chloride,
                                          25

-------
total Kjeldahl nitrogen, and sulfate were expected to serve as surrogates for the components of
the fracture fluid.

Analytical results were available for the hydraulic fracture fluid (pre-injection) as well as for the
subsequent produced  formation waters at 6 hours, 12 hours, 5 days, 10 days and 30 days
following fracture stimulation. The results for surrogate and indicator parameters were
evaluated. These included sulfate, total Kjeldahl nitrogen, ammonia, total dissolved solids,
sodium, glycol and 2-butoxyethanol. Examination of the glycol and 2-butyoxyethanol results
reveals the difficulty in accurately quantitating glycol using the available EPA approved method.
Benzene and toluene results were found to increase within 12 hours to concentrations well
above that  in the original hydraulic fracture fluid. The occurrence of these compounds is
attributed to natural occurrence within the natural gas production zone.

The results for the divalent cations, barium and strontium concentrations were examined for
relationship with total dissolved solids concentrations. The correlation coefficient for barium
and total dissolved solids was 0.966 and that for strontium and total dissolved solids was 0.988.
The correlation between chloride and total dissolved solids was 0.933. There appears to be a
similar relationship between total dissolved solids and radium-226 and radium-228 as well. It is
evident that the presence of total dissolved solids can be used as a sentinel parameter.

The appearance of gross alpha, gross beta, two radium isotopes and uranium-238 in produced
formation water following hydraulic fracturing represents an example of transport of naturally-
occurring materials from the shale formation into these waters and the time at which
concentrations appear to stabilize varies considerably from shale play to shale play. For
locations in the western  U.S., the measured activity for these radiochemistry parameters varies
considerably. The range of gross alpha and gross beta is from 620 to 4,000 pCi/L (mean value
1,750 pCi/L) and 250 to 1,200 pCi/L (mean value 760 pCi/L),  respectively. Radium 226 and 228
appear together with the radium  228 being the lesser in terms of activity. No uranium-238 was
detected. The activities encountered in the well samples from the eastern U.S. covered a wider
range of activity levels and exhibited higher maximum values. When results for a single location
are examined, there does appear to be a relationship with increasing total dissolved solids. This
appears to be true for results for  both Location 1 and Location 2.

The major conclusions from the review of produced formation water data are:
    •  Produced formation water is highly variable within and between shale plays.
    •  The elevated temperature and pressure affect the fate and transport of hydraulic
      fracturing fluid components and can create safety issues for sampling.
    •  Analytical techniques used for chemical and radiochemical characterization of the
      produced formation water must be robust to the matrix interferences presented by
      total dissolved  solids.
    •  The most reliable sentinel compounds appear to be total dissolved solids, chloride and
      divalent cations. The concentrations of these components are related to each other and
      are also related to the formation water volume.
                                          26

-------
The concentration of total dissolved solids is predictive of the concentration of other
species.
The concentration of indicator parameters for hydraulic fracturing mixtures, such as
chloride, sulfate, and boron, are overshadowed by the naturally-occurring
concentrations of these parameters in formation water.
The presence of NORM is delayed and associated with higher percentages of formation
water in the produced water with increasing time
                                    27

-------
  Fate and Transport of Select Compounds of Potential Concern
                          (COPC) in Tracing Fluids
                                 Angus E. McGrath
                               Stantec Consulting Inc.

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.

Use of proprietary mixtures of reagents in fracing fluids injected  in deep (>1,000 feet) zones, in
an effort to liberate natural gas, has led to considerable controversy regarding the potential
contamination of shallower (<500 feet) drinking water aquifers. This paper focuses on the
different classes of compounds identified in fracing fluids, and discusses:

   1. their properties in soil/sediment and groundwater,
   2. their potential fate  in the environment, and
   3. the potential problems analyzing for them.

The Canadian Society for Unconventional Gas (Understanding Hydraulic Fracturing, 2011)
identifies the following as the typical composition of fracing fluid (citing All Consulting, based on
data from a fracing operation in the Fayetteville Shale, 2008):

   •  Water and Sand = 99.51%
   •  Other 0.49% =
          o  Acid = 0.123%
          o  Friction Reducer = 0.088%
          o  Surfactant = 0.085%
          o  KCI = 0.06%
          o  Gelling Agent = 0.056%
          o  Scale Inhibitor = 0.043%
          o  pHAdusting Agent = 0.011%
          o  Breaker = 0.01%
          o  Crosslinker = 0.007%
          o  Iron Control = 0.004%
          o  Corrosion  Inhibitor = 0.002%
          o  Biocide = 0.001%

Table 2 (Adapted  from the  EPA, 2004) provides examples of typical compounds used in each
class of chemicals listed above. This paper will focus on the less common compounds included
in the list. The transport and fate of the compounds listed below is well understood and
treatment options have been identified where  warranted, so they will not be evaluated:
   •  Muriatic acid or hydrochloric acid;
   •  Guargum;

                                        28

-------
   •   Diesel (BTEX);
   •   Ethanol and isopropanol;
   •   Sodium chloride;
   •   Formic acid;
   •   Fumaric and adipic acids; and
   •   Boric acid.

This paper will focus on the compounds below which are not commonly encountered in the
environment and/or whose fate and transport is not clearly understood:
   •   Ethylene glycol;
   •   Mono or di-ethanol amine;
   •   2,2 Dibromo-3-nitrilopropionamide, a
   •   biocide;
   •   2-Butoxy ethanol; and
   •   Diammonium peroxosulfate (ammonium persulfate).


Compounds of Potential Concern (COPCs)
Ethylene glycol (EG) is a chemical commonly used  in a variety of applications such as airplane
deicing and radiator fluid. It is used in analytical laboratories and found as an impurity in
alcohols and other compounds. It is a polar liquid that is miscible in water and due to its low
Henry's constant and octanol-water partitioning coefficient (Kow), is unlikely to volatilize or
adsorb onto carbon or soil organic matter. One of the main reasons EG is not a common analyte
for environmental investigations is because it is not considered very toxic (not regulated by the
EPA although states, such as Pennsylvania, have set a regulatory goal of 14 milligrams per liter,
mg/L) and it is a common laboratory contaminant leading to false positive detections in
environmental samples at concentrations as high as 3 to 5 mg/L. With respect to its
environmental fate, although it is quite  mobile in soil and groundwater, EG also biodegrades
readily under aerobic and anaerobic conditions which are common in areas with methane
contamination through the following reaction pathway (Dwyer, 1983, Huang, 2005):

       ethylene glycol -> acetaldehyde  -> ethanol -> acetate -> methane + C02
Nitrate can serve as the electron acceptor in anaerobic biodegradation of EG (Schramm and
Schink, 1991), and nitrate is a common contaminant in rural groundwater as a result of
agricultural practices and/or septic field contamination of drinking water.

EG is unlikely to be significantly retarded by adsorption and therefore maybe a potential
concern for groundwater transport. Because it degrades readily and is not considered to be
persistent in the environment, it is not likely to be a significant problem for groundwater
migration.
                                          29

-------
Mono- or di-ethanol amine is a polar, colorless liquid that, like most amines, is weakly basic. It is
commonly used as a foaming agent and used in shampoo and cosmetics. The amine group
allows it to be a surface active compound that can adsorb onto mineral surfaces although it has
a low KOW (estimate of its adsorption onto soil organic matter) and Henry's constant limiting its
adsorption onto carbon and its potential to be stripped from water. As a result, diethanolamine
will be difficult to treat in groundwater using conventional water treatment technologies such
as granular activated carbon and air-stripping.

It undergoes moderate to rapid  biodegradation and is not considered persistent. Knapp et al.
(1996) isolated an anaerobe that uses diethanolamine as a carbon source and found that it
grew better under reducing conditions in the presence of nitrate than under aerobic conditions.
Therefore, due to its lack of persistence in the environment, diethanolamine should not pose  a
significant risk of migration.

2,2 Dibromo-3-nitrilopropionamide (DBNPA) is a water soluble (15 g/L) solid that is categorized
as a "quick-kill" biocide marketed by Dow for protecting reverse osmosis membranes, paper
mills, and offshore oil flooding operations. The  definition of quick-kill comes from the relatively
effective biocidal properties of the compound.  Dow (2006) sells a formulation of DBNPA as a
biocide, that includes EG to increase DBNPA solubility. Their marketing material promotes the
short persistence of DBNPA in the environment.

Dow (2006) and Exner et al. (1973) explain that DBNPA reacts with nucleophiles (or reducing
agents, I, HS, HSOs , SzCv, and SOs2) in aqueous systems to form cyanoacetamide and bromide
as shown in the reaction below:
  N-C-CBr2-CONH2+2NaHS03-i-2H/)    ->     N»C-CH2-CQNH2+ 2HjS04 + 2NaBr
      (DBNPA)                                  (Cyanoacetamide)

Cyanoacetamide is subsequently hydrolyzed to cyanoacetic acid, its amide, and malonic acid.
Exner et al. (1973) evaluated the degradation of DBNPA in soil/groundwater and surface water
assessing the half-life for hydrolysis, nucleophilic substitution (as described above) and
exposure to sunlight. They found that hydrolysis is rapid at pH 7.4 with a half-life of
approximately 21 hours. The half-life of hydrolysis increases to 155 hours at a pH below 6.
When they evaluated degradation rates in soil, they found that biodegradation under aerobic
conditions occurred with a half-life of 6 to 15 hours and formed no measurable degradation
byproducts besides ammonia and bromide. In order to assess the role of adsorption versus
biodegradation, they washed the soil with deionized water, filtered it and added DBNPA to the
extracted water. No  byproducts were observed besides ammonia and bromide. The reaction
was observed at a pH of 5.8, therefore the reaction was believed to be biologically mediated
since hydrolysis was  determined to be negligible below a pH of 6.

An evaluation of DBNA toxicity after reaction with activated sludge (simulating wastewater
treatment) indicated that the DBNPA was deactivated by the reaction (Gartiser and Ulrich,
2003). The study did  not evaluate for the  presence degradation  products, and attributed the

                                         30

-------
change to adsorption onto the activated sludge. The results of the Exner et al. (1973) appear to
indicate that the deactivation was likely due to biodegradation, enhanced by the microbial
enriched activated sewage sludge.

The available data appears to confirm that DBNPA has a short half-life in soil and sediments and
should biodegrade rapidly in the environment. Under reducing conditions and above a pH of 7.4
the toxicity of the compound should be rapidly deactivated. Therefore, due to its short half-life,
DBNPA is unlikely to pose a significant risk of migration.

2-Butoxy ethanol is a colorless liquid that is miscible in water and most organic solvents. It is
used as a solvent in many surface coatings and fast drying paints and  lacquers. It is also found  in
many cleaning agents and is therefore a common contaminant in domestic households,
businesses and their wastes. Beihoffer and Ferguson (1994) reported  that concentrations of 2-
butoxyethanol were observed in aqueous samples from a municipal and an industrial landfill in
the USA at concentrations ranging from <0.4 to 84 mg/l.

Given the common structure of its functional groups, it is relatively easily biodegraded. Howard
et al. (1991) determined an aerobic surface water half-life of 1 to 4 weeks and a soil half-life of
2 to 8 weeks in soil.

The WHO reports that 2-butoxy ethanol has a low Kow and Henry's law constant making it both
poorly adsorbed by organic matter in soil and not likely to volatilize into the gas phase once
dissolved in water. Therefore it will be difficult to treat in solution using either adsorption or
air-stripping.

Groundwater monitoring for 2-butoxy ethanol as a trace constituent in fracing fluids will be
complicated by the fact that it is commonly found in house hold cleaning products and paints
which maybe spilled at residential sites impacting shallow groundwater. Many domestic wells
exhibit contamination from both septic and other domestic sources which could include 2-
butoxy ethanol.

Although it is unlikely to be significantly retarded by adsorption, because 2-butoxy ethanol is
not considered to be persistent in the environment, it is unlikely to pose a significant risk for
contaminant migration. Household cleaning products and paints  may pose a greater risk for
exposure.

Diammonium  peroxosulfate or ammonium persulfate is a salt comprised of two common  ions,
ammonium and persulfate. Persulfate is commonly used in the environmental remediation field
to oxidize petroleum hydrocarbons, chlorinated solvents and other organic contaminants. It
degrades into sulfate and can form strong sulfate and hydroxyl radicals when activated by
ferrous iron or heat. Because it is a reactive oxidant, it is not persistent in soil or groundwater
and will react to form sulfate within a few weeks depending on how reducing groundwater
conditions are. As an oxidant, persulfate is one of the more slow  reacting oxidants under
                                          31

-------
standard temperature and pressure and in the absence of a catalyst such as ferrous iron or
heat.

Once reacted to form sulfate and ammonium ions, the main concern with this compound is that
it increases the total dissolved solids (IDS) in groundwater and may create algal blooms in
surface waters because it may increase the nitrogen concentration in surface waters. Both ions
are easily removed from impacted groundwater through ion exchange or reverse osmosis
treatment. It is not anticipated to pose a significant risk for migration because it will  react with
organic carbon and  reduced materials in the aquifer.


Priorities for Groundwater Monitoring and Treatment
Based on the findings of this evaluation, none of the compounds identified appear to pose a
significant risk of migration due to lack of persistence. Their chemical structure, reactivity
and/or biodegradability lower their persistence in the environment and limit their potential to
impact drinking water wells. That is not to say that screening and monitoring for these
compounds is not warranted if there is a reason to believe, based  on hydrogeologic data and
other evaluations, that the fracing fluids have the potential to reach a drinking water well.
                                         32

-------
References
Beihoffer J, C. Ferguson. 1994. Determination of selected carboxylic acids and alcohols in
      groundwater by GC-MS. Journal of Chromatographic Science, 32, p!02-106.
Dow. 2003. Antimicrobial 7287 Water Treatment Microbiocide. Product Information Sheet.
Dwyer, D.E. and J.M. Tiedje. 1983. Degradation of Polyethylene Glycols by Methanogenic
      Consortia. Applied and Environmental Microbiology, July, p 185-190.
EPA Report. 2004. Evaluation  of Impacts to Underground Sources of Drinking Water by
      Hydraulic
Fracturing of Coalbed Methane Reservoirs. Chap. 4: Hydraulic fracturing fluids. EPA 816-R-04-
      003.
Exner, J.H., G.A. Burk, and D. Kyriacou. 1973. Rates and products of decomposition of 2,2-
      dibromo-3-nitrilopropionamide. J. Agr. Food Chem., 21(5), p 838-842.
Gartiser, S. and E. Urich. 2003. Elimination of cooling water biocides in batch tests at different
      inoculumconcentrations. SETAC Conference. Poster.
Howard ,PH,  R.S. Boethling, W.F. Jarvis, W.M. Meylan, E.M. Michalenko. 1991. Handbook of
environmental degradation rates. Chelsea,  Ml, Lewis Publishers Inc.
Huang, Yi-Li.  2005. Anaerobic Biodegradation of Polyethylene Glycols Using Sludge Microbes.
      Process Biochemistry 40, p 207-211.
Knapp, J. S., N. D. Jenkey, C. C. Townsley. 1996. The anaerobic biodegradation of
      diethanolamine by a nitrate reducing bacterium. Biodegradation, 7 (3), p 183-189
Schramm, E.  and B. Schink. 1991. Ether-Cleaving Enzyme and Diol Dehydratase Involved in
      Anaerobic Polyethylene Glycol Degradation by a New Acetobacterium sp.
      Biodegradation, 2, p 71-79.
                                         33

-------

Product


W«.fgH>lg.g™
ur^rplpd,™,


Lifi»s» T*,**
fo.r.ir^^er.t



AsCHS ireat'r^l-

iVd^ior***
8t«,a«-
Table 1; £»*acactertstks of Undiluted ChcoiicaH loiinrt in Byttrauik Manunn^ Huids ^ Based on NtSDSs)
Chemical Composition iftformartiGMn Hazards. IiiforwiatioR1 Toxicolg0c4l Ititernsatiofi

• Can at latai it *igs-i"t»d
2 1 »V I rr
3. r,ai» ^ ^ [
;! v7r«*t'" f :an™ue wpon - .>,_- «ve. Sk r, ana fesp-ratao- tract .rr«t,c^
* F^Tnttfois ] /*nr«* .
» IV*, ^ J. JU* 'iSf t '^T „ (*fr iS'itr,
2i CtrfyJenff G^c^ • Ccm&Lntsb^ d*1*" !' i*->ri- # > *r *i *» i ,- "M* i yi
» C**n j i * « • ** S«* ->r fft- t Yie tE*
IS '^^ufTi E*ffiSfciC*at^ tl*t"af'^fiiSt# M^y &** mlMT'jr HTittStlflig- May 1"^ ifl ^ n ' *t w
• tc s»v« -arsd skin
H '^OPTK,* ; • »'»^''j[ ;f *Mllu«0 • CtrotK **t«b/w.(-^r^.v: »n.v «u« ^ve- .M frU-wy *««fe
*| >*Tot Ah ^ anii MI I * . .

"i) Etf**s *iS'i
4) Pclyjslycc.*- sttiet
11 H«l,«f,)o-.c «kl , ' **,<*£**<,*.*»** • Chro«
-------
  Fate and Transport Evaluation of Potential Release Scenarios
                during Hydraulic Fracturing Operations
                                   George Deeley
                               Shell Upstream Americas

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.
Introduction
Fate and transport includes the processes that control the behavior of chemicals if they are
released from some source by escaping existing natural or manmade barriers. Therefore, the
mechanism and chance of release must be determined before prioritizing any potential release
scenarios.  Determination of potential  receptors must then be determined  for the realistic
potential release scenarios. With realistic sources and receptors identified, fate and transport
analysis may proceed. Fate and transport  analysis will then require input data on appropriate
chemicals, hydrogeological data, and geochemical data within a proper modeling framework.
Fate and transport results are used to  determine appropriate collection of data to fill gaps,
validate models, or eliminate pathway from further consideration.

Scope
While focusing on fate and transport of chemicals (exposure pathways) and processes related
to potential impacts of hydraulic fracturing on drinking water, there will also be  some
description of the selection of reasonable  release  scenarios and receptors. The  fate and
transport processes  will, in general, be similar to those found with other potential chemical
releases to soil and  groundwater such as surface spills, impoundment leaks, and leaks from
underground storage tanks. The major differences will be in some potential release scenarios
such as cement jobs, pipe strings, and fractures, which are normally prevented  with engineered
and natural controls. This process demands the evaluation of these controls before proceeding
to further evaluation steps. Also,  the  chemical nature of potentially released material in
upstream  operations has been characterized as high volume and low toxicity.1 Therefore,
upstream  materials such  as  produced water are  of low  potential  risk  when managed
appropriately.

Conceptual Model
A conceptual site model (CSM) is a "...written or pictorial representation of an environmental
system and the biological, physical, and chemical processes that determine the transport of
contaminants from sources through environmental media to environmental receptors with the
system."2 Barriers (controls or defenses) may also be part  of the system (Figure 1). Whatever
the form used, a CSM is critical for determining prioritizing reasonable release scenarios for fate
and transport evaluation.
                                         35

-------
                                         BARRIERS
                             Controls

Pirx Siring

I'eiviem Jc*

-
?
Lft
f
=
I
¥
-
•
M.'.,r.|. l.r.r
Pn^v.ri ',-...
                                \
Defences
    ©CAE-EGYPT http://www.cae-eg.com/bowtie.html

HiiidliXiVL

>

11
T
/
?
"
'••
1

I
              SOURCE
                         CONTROLS
                                         PATHWAY       REMEDIATION   RECEPTOR
                                       < fait ,ind iHutspon)
             Figure 1. A generalized and a specific form of conceptual site model.

Realistic Potential Release Scenarios
The example CSM in Figure IFigure  1 shows a potential pathway from a cement job. There
would  be no need to  evaluate further  if the controls for the cement job are adequate to
prevent  release.  Controls  may include  cement standards,  and  pressure  tests.  However,
proceeding with development of a potential release scenario may be necessary if controls were
considered weak and  simple modifications  are not sufficient to provide confidence in  the
controls.

Identify the potential release scenarios by examining the  hydraulic fracturing processes that
could potentially release material  in a manner that could impact drinking water resources.
Gather  information  critical to  this  potential scenario. These data  might  include  process
information and controls, identification of volume and type of additives used in the  process,
and total volumes of hydraulic fracturing fluids that might be released if controls could fail.

Returning to the potential  for  a failed cement job, there are numerous standards  through
regulation or guidance in the industry to prevent such an occurrence.3'4 In this case then, it may
be determined  that existing standards and  regulations mitigate cement jobs as a  reasonable
release scenario. No further analysis is required in this case.

For the  sake of discussion,  assume that there is found to be potential for a release to surface
soil somewhere in hydraulic fracturing process from a control failure. Also assume that it was
not due to noncompliance  to any applicable regulations. If about 50,000 gallons of hydraulic
fracturing fluids5 were  released, this would  represent a maximum volume from one hydraulic
fracturing treatment substage.
                                          36

-------
As much information as possible on the release  liquid  is now required. With a hydraulic
fracturing fluid this would  include its make-up. It  is primarily water,  sand, and about 0.5%
(5,000 ppm) of up to about 8 additives.6 These additives  may include  a gellant (guar), buffer
(sodium   hydroxide),   breaker   (NaCI),  friction   reducer   (mineral  oil),    antimicrobial
(gluteraldehyde),  acid (HCI), and  surfactant (citrus terpenes  and isopropanol). If this were
flowback solution, there might be salt (NaCI), hydrocarbons, and ions  from formation  water.
Composition of the hydraulic fracturing fluid and formation water may  vary greatly depending
upon  reservoir requirements and properties.

At this point, the  source release  scenario,  release volume, and  release composition will be
known to some extent. In this case: A surface release of 50,000 gallons of hydraulic fracturing
fluid containing 50% original hydraulic fracturing fluid and 50%  formation water to surface soil.
There would of course be changes to the original hydraulic fracturing fluid such as a loss of
proppant to the producing formation.

Assume a roughly estimated composition of spilled material  (fluid and  formation water) as
shown in Table 3.
Table 3. Hypothetical release concentrations (my rough estimates)
Water (from fluid and formation water)
Guar (from fluid)
Sodium Hydroxide (from fluid)
Sodium Chloride (from fluid and formation
water)
Mineral Oil (from fluid)
Gluteraldyhyde (from fluid)
Hydrochloric Acid (from fluid)
Hydrocarbon (from formation water)
Citrus Terpenes (from fluid)
Isopropanol (from fluid)
NORM (from formation water)
97.8%
0.02%
0.005%
2.0%
0.04%
0.0005%
0.06%
0.05%
0.005%
0.005%

200 ppm
50 ppm
20,000 ppm
400 ppm
5 ppm
600 ppm
500 ppm
50 ppm
50 ppm
Managed if present
Fate and Transport
Data are now required on hydrogeology, geochemistry, and chemical properties. This also must
include background chemical information (naturally occurring or anthropogenic) for potentially
impacted drinking water sources that might interfere with fate and  transport evaluation.
Typical  background chemical constituents  and  parameters are  salts,  pH,  metals, naturally
occurring hydrocarbons, and anthropogenic hydrocarbons.
                                          37

-------
For  each component,  available  chemical  and  physical  property data  are  required  for
subsequent  evaluation.  These data may include dissolution, precipitation, degradation rates,
volatility, sorption, and dispersivity data.

At this point modeling may be performed based on site parameters. In the absence of the detail
required for sophisticated modeling, screening models may be used to conservatively estimate
fate and transport. For reactive inorganic compounds,  a  mineral  equilibrium model such as
minteqa2 may  provide  information on the behavior of  chemicals  such as barium, strontium,
etc.7 Most of these metals will fall out of solution due to solid precipitation or adsorption at the
levels found in the source material. Soluble salts can be  conservatively treated as remaining in
solution. All  chemicals in solution will be subject to potential transport through the unsaturated
zone and movement with groundwater. Concentrations  in Groundwater will tend to  decrease
with distance downgradient from the source due to mechanical  mixing (dispersion)  and
biodegradation. This decrease over distance can be defined as a dilution attenuation factor for
the saturated zone. This dilution factor is site specific and dependent on horizontal distance to
the  point   of  extraction,  biodegradation  rate,  retardation,  groundwater flow  velocity,
dispersivity,  and source size. This  relation between groundwater concentration in the source
area and downgradient from the source can be described by the following steady-state solution
for groundwater8:
                     exp
                               -, 1 +
                                     4 A a rR
                                  x erf
erf
where:
Cx
x
R
v
ax
ay
az
dissolved salt at a distance x from the source (mg/L)
horizontal distance to point of extraction (cm)
first-order degradation constant - salt does not degrade (day"1)
constituent retardation factor (dimensionless)
specific discharge (cm/day)
longitudinal dispersivity (cm)
transverse dispersivity (cm)
vertical dispersivity (cm)
source width  (cm)
source depth (cm)
Validation
Once any modeling is completed, it cannot be considered representative of a system until it is
validated, especially where there is uncertainty on any model factors. Uncertainly has to be
addressed through data gap analysis and appropriate data gathering to fill those gaps. Also,
field sampling and analysis will be required to test the model.

Chemicals must be selected for analysis based on their predicted ability to reach sampling
points. Chemicals that are biodegraded readily (benzene, citrus terpenes, isopropanol),
precipitated out of solution (barium), adsorbed to soils (polymers), or reacted (HCI) will not
move a significant distance. Therefore, they may not be measureable downgradient. Upstream
                                           38

-------
fate and transport modeling and monitoring exercises have been most successful when
performed on salts and divalent ions. Potential impacts may also be evaluated using ion ratios,
isotopic methods, and trilinear diagrams.

Remember that any spill from a hydraulic fracturing operation will be of limited mass and
volume so one cannot assume an infinite source, i.e., the source is finite and will disappear with
time. Immobile chemicals will stay near the release area to be left in place or readily removed.

Risk Assessment
Knowing the distance to a receptor, an exposure rate can be determined. This can be evaluated
with accepted risk exposure models for various drinking water receptors.

Conclusions
Processes are available for evaluating the fate and transport of selected reasonable release
scenarios in hydraulic fracturing operations. These methodologies consider all controls in the
hydraulic fracturing process as reflected in existing regulations and best management practices
before selecting scenarios. Background chemicals, both anthropogenic and natural, must be
considered in any analysis. Tools must be used that are  appropriate for uncertainty in available
data and analytical techniques.

References
1USEPA, October 2002, Exemption  of Oil  and  Gas  Exploration and  Production  Wastes from
Federal Hazardous  Waste Regulations,  United  States  Environmental  Protection Agency,
Washington,  D.C., accessed at:    http://www.epa.gov/osw/nonhaz/industrial/speciaI/oil/oil-
gas.pdf
2ASTM  International,  2008, Standard Guide  for  Developing Conceptual  Site Models for
Contaminated Sites,  E 1689-95 (Reapproved 2008), ASTM International, 100 Barr Harbor Drive,
PO Box C700, West Conshohocken, PA 19428-2959.
3DOE, May 2009, State Oil  and  Natural Gas Regulations Designed  to Protect Water  Resources,
prepared by Ground Water Protection Council for  U.S. Department  of Energy, accessed at:
http://www.gwpc.org/home/GWPC  Home.dwt.
4API,  October  2009,  Hydraulic Fracturing Operations  -  Well  Construction  and Integrity
Guidelines, Guidance Document HF1, First Edition, American Petroleum Institute, Washington,
D.C., accessed at: http://www.api.org/.
5DOE, April 2009, Modern Shale Gas Development in the United States -A Primer, prepared by
Ground  Water  Protection  Council  for U.S.  Department  of  Energy,   accessed  at:
http://fossil.energv.gov/news/techlines/2009/09024-Shale Gas Primer Released.html.
6Energy in Depth, 2009, A  Fluid  Situation: Typical Solution  Used in  Hydraulic  Fracturing,
accessed at: http://www.energyindepth.org/frac-fluid.pdf.
7USEPA, March 1991, Minteqa2/Prodefa2 - A Geochemical Assessment Model for
Environmental Systems: Version 3.0  User's  Manual, United States Environmental Protection
Agency, Athens, GA,  accessed at:
http://www.epa.gov/ceampubl/mmedia/minteq/USERMANU.PDF
                                          39

-------
8Domenico, P. A., 1987, An analytical model for the multidimensional transport of a decaying
contaminant species, J. of Hydrology, 91, 49-58.
                                         40

-------
   A Protocol to Characterize Flowback Fluid Contamination of
                                Drinking Water
                        Donald I. Siegel1 and Melody D. Kight, Esq.2
                                  Syracuse University
        2State University of New York, College of Environmental Science and Forestry

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.

The Appalachian Basin of New York, Pennsylvania and West Virginia contains one of the largest
gas plays in the world in the organic-rich Paleozoic-age Marcellus and Utica Shales. The
hydrocarbon-extraction industry has begun drilling and hydrofracturing (fracking) to produce
methane sorbed in the shale lamina. One concern relates to the potential for flowback fluids
and methane from the producing formation to contaminated drinking waters because of
improper grouting of vertical well casings and/or surface disposal.

Much of the concern over flowback fluids stems from elevated concentrations of trace
elements, radium-226, and trace organic substances. But, these constituents geochemically
react and naturally attenuate. For example, reactive iron and manganese in the anoxic flowback
fluid precipitate upon contact with oxygen. Also, barium  in solution precipitates upon contact
with dissolved  sulfate. Small concentrations of dissolved  organic carbon in flowback fluid,
including possible aromatic hydrocarbons and glycols, biodegrade once discharged to fresh
waters - analogous to natural attenuation associated with UST spills with orders of magnitude
greater concentrations of dissolved organic substances. Finally, rock formations can provide
many trace substances common to that found in flowback fluid. For example, the Marcellus
Shale in Appalachia contains barium and strontium mineralization that seals fractures (Siegel et
al., 1987; Chamberlain et al., 1986). In short, elevated concentrations of trace metals and many
dissolved organic compounds do not provide the means to unequivocally identify the extent of
flowback fluid contamination of drinking waters.

Formation waters mixing with shale bed methane flowback fluids have high salinities associated
with formation brines. But, in the Northeastern United States, road salt and septic discharge of
salts associated with water softeners also contribute salinity to surface waters (Mullaney et al.,
2009). Seawater intrusion can occur near coastal areas. Also, the surface disposal of acid mine
drainage from  coal mines can elevate the salinity of drinking water by the addition of calcium
and sulfate, among other solutes (Tetra Tech, 2009).

Fortunately, the dissolved halogen elements associated with salinity: chloride, bromide, and
fluoride, and iodide, can provide the means to unequivocally identify where flowback fluids or
oilfield brines have contaminated drinking waters. Mineral-water reactions minimally affect
these solutes, the ratios of which can distinguish among salinity sources (e.g. Panno et al., 2006;
                                         41

-------
          1.0000
          0.1000
          0.0100
       1
       SL  oooio
          0.0001
          0.0000
* Frac Flow Bat* (PA DEP 2010)
• Devonian Formation Brine (Dresel
1987)
Groundwater near Known salted
road (Panno 2006)
•Groundwater near septic effluent
(Panno 2006)
















































"
•
• .
v:
* %







	 	








•^

•

















































„













*








• ^
V*
















                          10
                                     100        1000
                                        [Cl] (mg/L)
                                                            •oooo
                                                                       1CCCCC
    Figure 2. Bivariate plot of Br/CI versus Cl concentrations for Marcellusfrac fluids,
    Pennsylvanian Devonian sedimentary basin brines, and representative road salt
    contamination and septic effluent. Note the clear separation between the flowback fluid,
    brines and other sources of contamination (Kight and Siegel, 2011)

Townsend, and Whittemore, 2005). Therefore, any contaminants associated with flowback
water that do not naturally attenuate must move with flowback fluid halogen salinity.
We have heuristically investigated using halogen and major solute ratios and chemical mixing
models to develop a protocol to characterize potential flowback water contamination of
drinking waters in the Appalachian Basin. Our preliminary results (Kight and Siegel, 2011) show
that plotting Br/CI ratio versus Cl concentrations distinguishes the presence of flowback fluid
contamination from other salinity  sources (Figure 2).

The Br/CI ratio of Marcellus flowback fluids also show that Appalachian Basin brine probably
provides the source for flowback salinity, not dissolving rock salt as previously proposed (Blauch
et al., 2009). The Br/CI ratio of flowback fluid would have to be at least an order of magnitude
smaller to reflect halite dissolution because bromide is excluded from halite when it
precipitates. Either hydrofracking-induced vertical fractures beyond the contacts of the
formation with more permeable and brine-filled formations, or the Marcellus contains
disseminated brine, yet unrecognized.

Our geochemical mixing models enable us to identify extraordinarily small amounts of flowback
water in mixtures, even when other salinity sources previously contaminated drinking waters.
                                           42

-------
Our protocol using halogen concentrations coupled to mixing models needs to be applied on a
site-by-site basis. That is, the background chemical composition of potential drinking-water
receptors must be obtained as well as representative analyses of flowback waters produced
near them. Every shale gas methane basin has its own geochemical halogen fingerprint.
However, once we have identified flowback from halogen ratios, the presence of solutes and
substances that could potentially compromise drinking waters can be addressed without the
fear of false positives.

We are continuing our work by addressing additional combinations of halogens and major
solutes and characterizing the most plausible source for methane in drinking waters by using a
combination of mixing models, coupled to methane isotopes and trace gases.

References Cited
Blauch, M.E., Myers, R.R., Moore. T.R., Lipinski, B.A., 2009, Marcellus Shale Post-frac Flowback
      Waters-Where is All the Salt Coming From and What are the Implications?, In
      Proceedings: Eastern Regional  Meeting of Society of Petroleum Engineers, Paper SPE
      125740, Charlestown, West Virgina, 23-25 September 2009.
Chamberlain, S.C., Dossert, W.P.  and Siegel, D.I., 1986, A new paragenesis and new localities for
      the  barium carbonate, witherite, Canadian  Mineralogist, vol. 24, p. 79-90.
Dresel P.E., 1985. The geochemistry of oilfield brines from western  Pennsylvania. Master's
      thesis, Pennsylvania State Univ.
Kight, M. and Siegel, D.I., 2011, A protocol for characterize flowback water contamination to
      shallow waters from shale gas development. Abstracts and Program, NE Geological
      Society of America Meeting, Pittsburgh, PA, p. 76.
Panno, S.V., K.C. Hackley, H.H. Hwang, S.E. Greenberg, I.G. Krapac, S. Landsberger, and D.J.
      O'Kelly, 2006, Characterization and Identification of Na-CI Sources in Ground Water,
      Ground Water, Vol. 44, p. 176-187.
Rose, A.W.  and Dresel, P.E., 1989, Deep brines in Pennsylvania: in Water Resources in
      Pennsylvania, Availability, Quality, and Management, S.K. Majumdar, E.W. Miller, and
      R.R. Parizek (eds.), The Pennsylvania Academy of Science, pp. 421-431.
Siegel, D.I., Chamberlain, S.C. and Dossert, W.P., 1987, The isotopic and chemical evolution of
      mineralization in septarian concretions: Evidence for episodic paleohydrogeologic
      methanogenesis, Geological Society of America, vol. 99, p. 385-394.
Davis, S.N., Whittemore, D.O.; and  Fabryka-Martin, J, 1998, Uses of chloride/bromide ratios in
      studies of potable water,  Ground Water, Vol. 36, p. 338-350.
Mullaney, J.R., Lorenz, D.L. and Arntson, A. D., 2009, Chloride in Groundwater and Surface
      Water in Areas Underlain by the Glacial Aquifer System, Northern United States, U.S.
      Geological Survey, National Water-Quality Assessment Program, Scientific Investigations
      Report  2009-5086, 54p.
Tetra Tech  NUS, Inc., 2009, Evaluation of High TDS Concentrations in the Monongahela River,
      126p.
Townsend,  M.A. and Whittemore, D.O., 2005, Identification of nitrate and chloride souces
      affecting municipal wells waters of the city of McPherson, Kansas, Kansas Geological
      Survey  Open File Report 2005-34, 24p.

                                          43

-------
Summary and Abstracts from Theme 2: Impacts of Hydraulic
        Fracturing on Natural Transport Systems
                         44

-------
      Summary of Presentations from Theme 2: Impacts of Hydraulic
                   Fracturing on Natural Transport Systems

The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.

Technical Presentations
The first set of technical presentations in this theme addressed fracture propagation.

Ahmad Ghassemi, Texas A&M University, discussed rock failure and stimulated volume with
regard to HF. He described the stress and pressure changes resulting from HF treatments, the
variables that affect rock failure, and how this information is used to predict stimulated rock
volume (SRV), the volume of rock with fracture-induced enhanced permeability. Dr. Ghassemi
explained that stimulation often results in the formation of a complex fracture structure, which
is important  to consider. He explained the assumption that Mode I fracture propagation
dominates; however, microseismic monitoring shows that shear fracture also does occur and at
times is the dominant mode. Therefore, Dr. Ghassemi recommended that mixed-mode
fractures and slip and shear dilation should  be incorporated into models. Dr. Ghassemi added
that the large volume of failed rock tends to redistribute the stresses in the  rock mass and can
modify the nearby rock permeability. The rock mass heterogeneity and the time-dependent
behavior of intact rock and fractures are important in the permeability evolution of the
stimulated volume, according to Dr. Ghassemi.

Norman Warpinski, Pinnacle -A Halliburton Service, discussed vertical fracture growth and the
effect of heterogeneities in layered and discontinuous media. He indicated that fracture
patterns are  not simple to model; fractures are complex with many complex individual strands.
Mr. Warpinski stated that vertical fracture growth is influenced by many factors that vary
between and within reservoirs, and that in situ stress distribution is the dominant influence,
according to  Mr. Warpinski, but sedimentary interfaces, natural fractures, and other
heterogeneities can also affect fracture behavior, and layering and interfaces resulting in
inefficient growth.

The second set of technical presentations addressed  fluid and gas flow in fractured formations.

Zhong He, Range Resources, discussed the flow of gas and water in hydraulically fractured shale
gas reservoirs, focusing mainly on flow taking place after the fracture treatment. Pressure
gradients are the driving factor for fluid flow: a pressure sink in the depleted zone drives gas
through the fractures and into the wellbore. While a  portion of the injected fluid is returned to
the surface through the  wellbore, the presenter stated  the majority of the fluid is trapped by
the shale formation and becomes immobile. Mr. He emphasized that physical principles
preclude the migration of gas and water away from the stimulated zone.
                                         45

-------
David Cramer, ConocoPhillips, discussed the use of injection fall-off tests for characterizing
mechanical and flow properties. He described the procedure for fall-off testing and provided
details on two specific types of analysis: (1) fracture closure analysis for determining in situ
stress and identifying non-ideal fracture closure mechanisms, and (2) after-closure analysis for
determining reservoir flow behavior, reservoir flow capacity or transmissibility, and initial
reservoir pressure. He stated that the information gathered from these analyses can be used to
predict the results of the HF treatment (fracture geometry, proppant placement, fracture
conductivity, etc.) and to inform modifications of the treatment design.

Alan Byrnes, Chesapeake Energy,  described the role of induced and natural imbibition in
fracturing fluid fate and transport. He described calculations for estimating stimulated reservoir
volume and presented data from experiments and models demonstrating that a majority of the
injected fluid leaks off into the fracture face, resulting in an "invaded zone" of elevated HF fluid
concentrations 2-6 inches thick. As pressure in the reservoir drops, Mr. Byrnes indicated that
strong capillary pressures hold these imbibed fluids in place.
     Summary of Discussions Following Theme 2: Impacts of Hydraulic
           Fracturing on Natural Transport Systems Presentations
The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.

Horizontal and vertical fractures. The presenters clarified that heterogeneity in the vertical
direction influences stresses, and other rock mechanical properties like ductility or brittleness
will also affect hydraulic conductivity. They stated that above a depth of approximately 1,500
feet, the dominant component of fractures will shift from vertical to  horizontal. The presenters
indicated that differences in overburden pressure are unlikely to affect fracture orientation.
They noted that controlling fracture orientation would require controlling the in-situ stress
since, at depth, almost every fracture is vertical. A participant asked whether the operator
would know if horizontal activity were occurring,  and the presenters responded that the
operator would know if overburden stress were exceeded and would be aware of changes in
pressure.

Shallow fracture depths. A participant asked about the shallowest depths for HF. While most
operators' shallowest fracture jobs have been 1,000-1,200 feet, a participant indicated that HF
is performed for some ground water cleanup work and other applications at depths of only a
few meters.

Distance of perforation clusters. A participant asked about the spacing of perforation clusters
on horizontal wells. A presenter responded that the optimum spacing is determined
empirically.
                                          46

-------
Fracture fluids and different formation types. A participant asked if different types of fracture
fluids are used for deep shale gas and shallow coalbed methane (CBM) jobs. The presenters
explained that water-based fluids and crosslinked gels are both used in shales. The presenters
stated that, in general, viscous fluids do not create the desired level of complexity, though in
certain situations, denser fluids are used to encourage downward fracture growth. They stated
that borate systems are more popular for CBM, and higher-viscosity fluids tend to perform
better in coal formations. Participants noted that tight sands are treated similarly to shales,
though tight sands do have more complexities.

Propagation rates. A participant asked about the rate of the propagation across lamina. The
presenter responded that while  laboratory experiments have been performed in layered rock,
there is no way to directly measure these propagation rates in the field. Participants suggested
that microseismic techniques or pressure profiles might be used to calculate them. Other
participants noted that, due to the small scale and complexity of the layered systems, this
would be very difficult to model.

Models simulating fracture growth. A participant noted that most of the complexities
introduced  into fracture growth  models result in  predictions of less fracture growth than the
models would otherwise predict. A participant asked if any uncertainties would lead to a higher
fracture growth than predicted.  The presenters clarified that models are used much less than
empirical knowledge when designing fracture jobs. In addition, they noted that less complexity
does not necessarily equate to less fracture propagation, but simply indicates a more complex
geometry and higher levels of uncertainty. They also stated that information on permeability
can be related to flowback  performance to gain understanding of the properties of the fracture
network.

Conceptual models for fluid migration. A participant asked for more information on conceptual
models that would involve  fractures propagating from deep reservoirs to shallow aquifers. The
presenters emphasized the importance of understanding the  in situ stress profile and lithologic
variations. In the field, the  presenters noted that diagnostic microseismic techniques provide
data on fracture height. In  general, presenters indicated that stress conditions controls fracture
height, and fluid pressure within the formation is also important to understand. The presenters
indicated that extreme vertical fracture growth is very unlikely.

Electromagnetic (EM) methods. A participant mentioned magnetotellurics and asked whether
electromagnetic methods are used to monitor fracture propagation. A presenter indicated that
EM methods are frequently used for geothermal  applications and are also sometimes used in
the oil and gas industry.

Production termination. A participant asked when a shale gas system would become a pressure
sink. A presenter explained that  this is based on permeability; if most of the pressure is in
microfractures, a depressurized  zone will  be created when that pressure is dissipated. The
presenter noted that the production  decline curve provides information on dissipating
pressures within fractures,  and matrix depletion is a function of the permeability of the matrix.

                                          47

-------
Imbibition. A participant asked about collecting post-injection core samples to confirm that
water is held immobile in the rock. Another participant noted that because shale formation
production is fairly new, this has not been done. However, data from lab-tested core samples
do indicate that water is trapped in the matrix. Another participant asked about the six-inch
invasion zone. The presenter clarified that the six-inch area is the induced imbibition zone, and
that after this point, natural imbibition factors go into effect. A participant asked if imbibition
causes changes in the rock, damaging the shale. The presenters answered that operators try to
produce water from the well as quickly as possible following a HF job so that the water in the
formation is not working against gas production since water in the near-fracture face
environment slows down the production rate. In addition, the presenters indicated that the
imbibition process is not static; it does not stop when pumping stops. Other participants added
that the more successful wells are generally the ones where the least amount of injected water
is returned, meaning that the formation is porous and permeable enough to hold the water.

Pressure fall-off tests. A participant asked about the effect of viscosity changes on pressure test
curves. The presenter stated that the viscosity of the reservoir fluid is an important parameter
to know or estimate (it is the u. in kh/u., the expression for rock transmissibility). However,
changes in the viscosity of the injected fluid (due  to pressure changes) do not impact the
pressure test curve. A participant noted  that, in his experience, relative permeability curves are
more asymmetric than the one included in the presentation. The presenter clarified that the
curves in the presentation are generic curves, not specific to gas shales. The presenter
explained further that relative permeability curves for gas shales are not inconsistent with
those for tight sands; the curves are generally asymmetric, and the shape of the curve varies
depending on the formation. The presenter indicated that water relative permeability curves
are not very well defined for gas shales because the permeability of the rock to water is very
low (in the range of a few nD or less).

Pressure buildup in plugged wells. A participant noted that, in Upstate New York, some people
are concerned that pressure will increase in a plugged well and cause a blowout. The presenters
were not aware of any instances of this.  One participant noted that pressure can build up over
long periods of time; however, he suggested that with good engineering practices, the most
extreme pressure experienced by the well occurs during the initial completion. In addition,  he
noted that multiple plugs set at different depths would protect against this kind of event.
Another participant added that layers above the cement plug would be at hydrostatic pressure,
in contrast to the underpressured,  depleted reservoir. The participant indicated that this would
prevent a dangerous pressure buildup.

Potassium chloride (KCI) substitutes. A participant asked about potassium chloride (KCI)
substitutes. The presenters clarified that KCI is expensive in large quantities and that other
chemicals, such as tetramethylammonium chloride, can be used instead. However, they
indicated that KCI and KCI substitutes are often not necessary at all, depending on the
properties of the formation. Participants noted that they do not use KCI substitutes in the
Marcellus and the Eagle Ford.

                                          48

-------
Microseismic data and vertical fracture heights. The presenters clarified that, in the slides
showing microseismic data, aquifer depth corresponds to the depth of the deepest water wells
per county, based on U.S. Geological Survey (USGS) data. A participant asked if vertical fracture
heights could reach 1,000-1,500 feet. The presenters suggested that fracture height growth will
vary as a function of the formation and the properties of the vertical stress profile. However,
participants indicated that, fundamentally, pressure and energy constraints would prevent the
creation of induced fractures that large.

Fluid migration through abandoned wells. A participant asked if native brine could travel up an
unplugged abandoned well and reach sources of drinking water, in a case where the gas
reservoir was overpressurized. A participant noted that this would require gas lift or another
drive mechanism. Another participant described a situation where an operator fractured into
offset wells and saw water at the surface. However, the presenters noted that moving proppant
from the  injection well to the offset well would require a continuous string of proppant from
one well to the other, which is difficult to achieve. Participants emphasized the need for
monitoring and coordination when there are nearby wells. One participant noted that, while
well communication does occur, operators are not aware of any resulting environmental
problems.

Buoyantly rising fluids. Participants discussed the possibility of fluids rising in a slug, similar to
magma rising buoyantly through layers of rock. Some  participants argued that this type of
movement would require a complete absence of leakoff, which seems extremely unlikely, even
in crystalline rock. In addition, they said that the horizontal stresses that act to close the
fracture would have to be counteracted. One participant described a situation where this may
have happened in the Gulf of Mexico. Another participant referenced studies showing a
correlation between the thickness of the Marcellus Shale and the amount of radon in residents'
basements. However, another participant noted that this is diffusive transport. One participant
suggested modeling the buoyant rise, and another participant noted that an abandoned well
conduit would be a situation with zero leakoff.

Alternative fracturing fluid systems. A participant asked about using gas for HF instead of water.
Another participant stated that one operator has been using gas in HF for years,  mostly in
subpressurized reservoirs. The participant gave an example of the use of nitrogen or carbon
dioxide in foam fracturing. A participant added that nitrogen is used in shale reservoirs with
very low underpressures,  and another participant asked about HF with liquid carbon dioxide
("dry frac"). The presenter responded that dry frac systems are primarily used in Canada,
because that is where the few blenders of this fluid system are located. He noted that liquid
carbon dioxide is used in super-depleted, very low permeability formations where any water in
the wellbore would create a problem, and these systems are very expensive and rarely
economically viable. In addition, he indicated that proppant transport is not very good. The
presenter noted that it is more  likely that an operator would use foam instead, which  has most
of the benefits of the dry fracture with better proppant placement and a lower price.
                                          49

-------
Participants noted that thinner fracture fluids would, in general, lead to more complexity in
fracture growth.
                                           50

-------
    Abstracts for Theme 2: Impacts of Hydraulic Fracturing on Natural
                              Transport Systems
Abstracts were submitted to U.S. EPA by the presenters for use in this proceedings document.
              Not all presenters submitted abstracts of their presentations.

The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.
                                      51

-------
     Rock Failure and Permeability Enhancement in Tight Gas
                             Hydraulic Fracturing
                                   Ahmad Ghassemi
                                Texas A&M University

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.


Introduction
Generally, a tight gas  reservoir is defined  by its low  permeability, however,  it  has been
suggested (Holditch, 2007) to define a tight gas reservoir as one that "cannot be produced at
economic flow rates or recover economic volumes of natural gas unless a special technique is
used to stimulate production." Large  hydraulic fracture  treatments, often from a horizontal
wellbore or multilaterals must be  used to increase the  recovery efficiency  in the reservoir.
Fracture  conductivities of 10  mD/ft  or higher  appear  to be necessary for economic  gas
production.

Shale gas reservoirs have heterogeneous geological and geomechanical characteristics that
pose challenges to accurate prediction of their response  to hydraulic fracturing. Experience in
shale gas formations shows that stimulation often results in formation of a complex fracture
structure, rather than the planar  fracture aligned with the maximum principal stress. The
fracture complexity arises from intact rock and rock mass textural characteristic and the in-situ
stress and their interaction with applied loads. Open and  mineralized joints and interfaces, and
contact between rock units play an  important role in fracture network complexity which affects
the rock mass permeability and  its  evolution with time. Currently,  the  mechanisms that
generate these fracture systems  are  not  completely  understood,  and can generally be
attributed to lack of in-situ stress contrast, rock brittleness, shear reactivation of mineralized
fractures, and textural heterogeneity.

Stimulated Volume and Permeability Enhancement
The idea of stimulation by hydraulic fracturing is to create a  large volume of fractured rock with
enhanced permeability. Many tight gas reservoirs are characterized by high deviatoric stresses
and hard, naturally fractured rock. Stimulation treatments  in such reservoirs may result from
slip on pre-existing critically stressed fracture systems and or  creation of new fractures. It is
generally believed that fracturing is caused by both shear and tensile failure. Shear slippage is
induced by altered stresses near the tip of the fractures as well as by increased pore pressure in
response to leakoff through the fracture "walls". In view of this, it has been suggested that
increased viscosity promotes tensile failure and can lower complexity (Cipolla et al.  2008).
Accordingly, water fracs are used where shear failure is anticipated to dominate (Chipperfield,
ST., Wong, J.R., Warner, D.S. et al. 2007). According to Cramer (2008), water is used as a base
fluid in most unconventional reservoir treatments.

                                         52

-------
To determine intact rock failure and joint slippage, a failure criterion is employed. There are
many failure criteria for the sliding of jointed rock masses but often the Mohr-Coulomb failure
criterion is used. By combining a stress analysis with a criterion, one can assess the effects of
increasing pore pressure on rock by generating a structural permeability diagram. This map that
shows the AP required to reactivate joints of different orientations (e.g., Nygren and Ghassemi,
2005; Nelson et al. 2007) during fracture stimulation treatments at high treating pressures. One
such map  is shown in Figure 3 for the New Albany Shale.
                                                          AP
                                                         (psi/ft)
                                                           .0.28
           Figure 3. Structural Permeability Diagram for New Albany Shale ( u,=0.6).
                      Figure 4. Fracture orientation in horizontal wells.

Most field implementation of stimulation involves creation of multiple hydraulic fractures and
stimulation of the neighboring rock volume by compression and pore pressure increase.
                                           53

-------
Normally hydraulic fracturing is performed in horizontally-drilled wells. The geometry and
propagation direction of a hydraulic fracture will mostly depend on the drilling direction of
horizontal well and the in situ conditions as shown in Figure 4.

It is generally accepted that hydraulic fractures propagate perpendicular to the least principal
stress. In shallower environments where the least principal stress is vertical, a fracture will grow
horizontally. At some depth where the increase in overburden causes the least principal stress
to be horizontal, the predominant fracture growth geometry will be vertical. Variations in
stresses between different lithology in vertical sequences of rocks can cause fracture growth in
a contained manner and generate length, or allow it grow vertically upwards or downwards. In
addition to the in-situ stress, fracture growth will depend on many factors such as natural
fractures, bed laminations, and other characteristics of a reservoir including the formation pore
pressure in the reservoir. The pore pressure will affect the effective in situ stresses, and can
further affect the post-fracturing deformation of rock and its natural  fractures which will, in
turn, influence the path of the hydraulic fracture (Koshelev and Ghassemi, 2001).

Fracture interaction
Multiple stage hydraulic fracturing is popular in the stimulations of tight gas reservoirs.
\
.
\
i


1 }
1
*• . f
T ' '
lip ' x
•x •**
/
y
»
• 	
4
;
Figure 5. Interaction of multiple fractures in a horizontal well. Green represents closed fractures.
Note the fractures turning away from each other to follow the path of least resistance.
                                           54

-------
   Figure 6. Fracture aperture distribution (in cm) after 3 hours of pumping in Barnett shale.
Estimating Stimulated Volume
It  is believed that microseismic events (Figure 8) are mainly created as a result of shear
slippages around the hydraulic fractures (Albright and Pearson (1982); Warpinski et al. (2001);
Rutledge et al. (2003)). Shear slippage is  induced  by altered stresses  near the tip of the
fractures as well as shear slippages related to leakoff induced pore pressure changes.

Accepting that failure of the formation around a hydraulic fracture is caused by pore-pressure
and  stress  perturbations,  the stimulated reservoir volume (volume of  "failed" rock in the
reservoir) can be assessed using the areal extent of the micro-seismic cloud (Plamer et al. 2005;
Jun and Ghassemi, 2005).  However, it should be this procedure for evaluation of stimulated
volume and fracture surface area is based on the assumption that energy release is exclusively
related to fluid penetration, which may not always hold true. The micro-seismic record may
also be used to detect hydraulic connection with the outside zone.

Prediction of enhanced permeability
The methodology of predicting the permeability in the failed region around a fracture is based
on a trial and error procedure: (i) use the pressure profile at shut-in, and guesses a value  for
permeability, K; (ii) for a selected net fracture pressure, predict the  failed  rock volume (FRV)
using the stress analysis; (iii)  vary K until the FRV matches the  particular trend-line of the
stimulated reservoir volume from induced seismicity at the given net  fracturing pressures. This
method is based on the equivalent permeability for the failed  rock and does not consider the
time dependent behavior of rock and the fractures that are created. Furthermore, it is assumed
that the MEQ's are related to local pore pressure perturbations.

The interaction  of the multiple hydraulic fracture stimulation on the larger scale flow regime is
not clearly understood. The large stimulated volume that is generated tends to redistribute the
                                          55

-------
stresses within the crust and can cause changes in nearby rock permeability. In this context, the
presence of faults (active and inactive) need be considered.
                           Ml l I  I : !).» 17 4 ft f.5 11.4 HI 4 I
                          '
Figure 7. Distribution of pore pressures (MPa) in the formation; minimum principal effective
stress (Barnett shale, 0.09 ms/s per fracture; 3 hrs) (Rawal & Ghassemi, 2011).
                                           56

-------
                                                                             Fracture Mapping Results
                                                                                         DPI-2485-21
                                                                                        3. Discussion
               -2.000      -1.500     -1.000       -500
                                                                    500       1 000       1 500
     3.500
     3000
     2.500
     2000
    H
     1.500
     1 000
     500
                 DPI 2426-21
                                         *   *
Orientation Shot 1
              Orientation Shot 2
      \
                                           DPI 2486-211
                                           iDPI 2485-21
                                                                      Stage 1
                                                                      Stage 2
                                                                      Stage 3
                                                                      Stage 4
                                                                      Stage 5
                                                                      Stage 6
                                                                      Stage 7
                                                                      Stage 8
Figure 8. Microseismic map shows network growth and the potential stimulated volume in shale
(GTI-NAS-Project).
                                                  57

-------
Chipperfield, ST., Wong, J.R., Warner, D.S. et al. (2007). Shear dilation diagnostics: A new
       approach for evaluating tight gas stimulation treatments. SPE Hydraulic Fracturing
       Technology Conference, College Station, Texas USA. SPE 106.
Cipolla, C.L., Warpinski, N.R., Mayerhofer, M.J. et al. (2008). The relationship between fracture
       complexity, reservoir properties, and fracture treatment design. SPE Annual Technical
       Conference and Exhibition, Denver, Colorado, USA. SPE 115.
Cramer, D.D. (2008). Stimulating unconventional reservoirs: lessons learned, successful
       practices, areas for improvement. SPE  Unconventional Reservoirs Conference, Keystone,
       CO. SPE# 114172.
Ge, J., Ghassemi, A. (2011). Permeability enhancement in shale gas reservoirs after stimulation
       by hydraulic fracturing. Proc. 45th U.S. Rock Mechanics Symposium, June 26 -29, San
       Francisco.
Ge, J., Ghassemi, A. (2008). Analysis of failure  potential around a hydraulic fracture in jointed
       rock. Proc. 42nd U.S. Rock Mechanics Symposium, June 29 -July 2, San Francisco.
Koshelev, V. & Ghassemi, A. (2003). Numerical modeling of stress distribution and crack
       trajectory near a fault or a natural fracture. Soil-Rock America Symp., Boston.
Nelson, E.J., et al. (2007). Using geological information to fracture optimize stimulation
       practices in the Cooper Basin, Australia. Petroleum Geoscience, (13), 3-16.
Nygren, A. and Ghassemi, A. (2005). Influence of cold water injection on critically stressed
       fractures in Coso Geothermal Field, CA. 40th U.S. Rock Mech. Symp., Anchorage, Alaska,
       June 25-29.
Palmer, I. D., Moschovidis, Z. A., and Cameron, J. R. (2005). Coal failure and consequences for
       coalbed  methane wells. SPE 96872, ATCE, Dallas, 9-12 Oct 2005.
Peasron,  C. (1981). The relationship between micro-seismicity and high pore pressure during
       hydraulic stimulation experiments in low permeability rocks. J. Geophy. Res., 86, B9,
       7855.
Rawal, C., Ghassemi, A. (2011).  Poroelastic rock  failure analysis around multiple hydraulic
       fractures using a BEM/FEM model. Proc. 45th U.S. Rock Mechanics Symposium, June 29
       -  29, San Francisco.
Rutledge, J., and Philips, W.  (2003). Hydraulic stimulation of natural fractures as revealed by
       induced  microearthquakes, Carthage Cotton Valley gas field, east Texas, Geophysics,
       68(2), 441-452.
Warpinski, N.R., Wolhart, S.L., and Wright, C.A. (2001). Analysis and prediction of
       microseismicity induced by hydraulic fracturing.  SPE 71649.
                                          58

-------
       Fracture Growth in Layered and Discontinuous Media
                                    Norm Warpinski
                             Pinnacle - A Halliburton Service
 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.

Fracture behavior in the vicinity of layered and discontinuous rock masses has been the subject
of numerous papers. The major factors that have been investigated are stress variations,
modulus variations, fracture toughness variations, interface properties, high permeability
zones, combined layering and interfacial behavior, and fluid pressure gradient changes. Of
these, stress changes are clearly the largest influence on fracture growth across layers and
stress bias is clearly the largest factor in the development of complexity in discontinuous media.
Nevertheless, many of the other factors play a significant role in cases where the stress
contrasts are not large and in the general development of complex fractures.

In Situ Stress
The in situ stress contrasts clearly have the most significant effect on fracture height growth.
The importance of stress was recognized early on (e.g.,  Perkins and Kern 1961) and has been
extensively studied in modeling (e.g., Simonson et al. 1978, Voegele et al. 1983, Palmer and
Luiskutty 1985), mineback tests (Warpinski et al. 1982), and numerous laboratory experiments.
Fracture height growth can be easily restricted if the layers above and below have higher stress
than the reservoir rock, and this is a common occurrence in sedimentary basins.

An equilibrium (static) analysis of the Linear Elastic Fracture Mechanics behavior of a fracture
surrounded by rocks with higher stress was first given by Simonson et al. (1978) for a symmetric
case (stresses above and below are equal). Given the geometry in Figure 9, an equation can be
written as
 2r      i .  _lf /Z^   KIc
:-[CT2-CTl]sin  — --=£=
            71
              H
where P is the net pressure in the fracture, 
-------
conservative since there are other dynamic factors that affect the amount of height growth that
will occur. Similar equations can be developed for non-symmetric stress contrasts, but more
complete dynamic analyses are usually performed  in fracture models.

Layer Material Property Differences
While Simonson et al. (1978) show that a material property interface in an ideal situation could
blunt fracture growth, years of fracturing
experience (Nolte and Smith 1979), fracture
diagnostic monitoring (Warpinski et al. 1998,
Wright et al. 1999), mineback testing
(Warpinski et al. 1982),  and other research
(Smith et al. 1982; Teufel and Clark 1984;
Palmer and Sparks 1990) have shown that this
is not the case. Figure 10 shows an example of
a dyed water fracture that has propagated
through an interface from a  low modulus
material into a high modulus material
(Warpinski et al. 1982).  A more complete
discussion of the role of the  interface has been
given by Cleary (1978), where the complexities
of the interface, the micromechanics of the
fracturing process, the potential for blunting
and twisting (no longer  only mode I fracture
growth), and various other factors make the
problem difficult to analyze with standard
analysis tools. What is clear from these studies
is that crossing interfaces requires additional
energy and can hinder vertical growth.
  High-modulus
     material
        Propagation
         Direction
                          Low-modulus
                            material

Figure 10. Mineback photo of fracture
propagating across interface.
Modulus contrasts clearly have an effect on the width of the fracture and can be expected to
enhance or restrict fluid flow appropriately. Cleary (1980) provided a time-constant analysis of
the effect of modulus, while Van Eekelen (1980) developed a relationship based on relative
height changes in the layers, given by
                                     .              l
                                     og — +- 3 + ^-	1
                                                              '
As discussed by Van Eekelen (1980) and Smith et al. (2001), these effects are generally small
and cannot be expected to provide significant containment of fractures. Gu and Siebrits (2008)
also show that low modulus layers surrounding a higher modulus pay zone can be restrictive
due to a lowered stress intensity factor, but this also depends on the relative fracture
toughness of the different materials.
                                          60

-------
Fracture Toughness
Fracture toughness can have a very significant impact on fracture growth, and a large value of
KIC can either induce a high pressure, restrict the height, or both. For a homogeneous
formation, the stress intensity factor at the top of the fracture can be computed if the net
stress distribution is known by
                                   1   MIZ  ( . I
                                   m*-™^-
                                                 H/2-y
where p(y) is the net stress distribution vertically. If the stress intensity factor exceeds the
fracture toughness of the material, the fracture will propagate. Obviously, the situation
becomes more complex (and not analytic) for layered materials with different elastic
properties, but the equation above gives a rough estimate of the fracture stability.
Laboratory experiments have generally shown that
fracture toughness varies over only a limited range
(e.g., Hsiao and El Rabaa 1987), which suggests that
fracture toughness effects will be rather limited.
Figure 11 shows a compendium of fracture
toughness measurements made at the DOE MWX
experiment that shows the relatively small range for
both reservoir and non-reservoir rocks. However,
the scale dependence of fracture toughness (or
potentially other types of tip effects) is not well
understood for large scale fractures, so there may
be potential for fracture containment due to this
mechanism (Shlyapobersky et al 1998).
                                                     4000
                                                     5000
                                                     6000
7000
                                                     8000
                                                     9000
                                                                O  O
                                                            A

                                                               A

                                                            A   A
                 0  A

                   0
                                                                  OA
                                                          O Sandstones
                                                          A Non-reservoir lithologies
                                                        0    500   1000  1500   2000   2500
                                                            Fracture Toughness (psi-Vin)
                                                 Figure 11. Fracture toughness data from
                                                 MWX.
Interfaces
It is well known that weak interfaces can blunt
fracture growth, and such a mechanism is often
cited for the use of KGD (Khristianovich, Geertsma and De Klerk) models (Nierode 1985).
Examples of blunting have been noted in mineback experiments (Warpinski et al 1982,
Warpinski and Teufel 1987, Jeffrey et al. 1992, Zhang et al. 2007) and laboratory experiments
(Anderson 1981, Teufel and Clark 1984). While it is generally expected that weak interfaces will
be most important at shallow depths where friction due to the overburden stress is a minimum,
other factors such as overpressuring or embedded particulates (equivalent to a fault gouge) can
clearly minimize frictional effects even at great depths. Weak interfaces have the potential of
totally stopping vertical fracture growth, initiating interface fractures, or causing offsets in the
fracture. In addition to restricted growth effects, weak interfaces above and below the
reservoir can decouple the fracture walls (Barree and Winterfeld 1998, Gu et al. 2008), resulting
in poor coupling of the fracture pressure in the reservoir to the fracture outside of the weak
                                          61

-------
interfaces. This reduced coupling would create narrower fractures in the layers across the
interface and much wider fractures within the reservoir rock.

Many mechanism, such as those described above and others, can be bundled together to
describe fracturing across a succession of interfaces. The possibility that such layered media
could contain hydraulic fractures has been derived from fracture diagnostic information
(Warpinski et al. 1998, Wright et al. 1999,  Griffin et al. 1999). It is easy to conceive of multiple
mechanisms serving to blunt, kink, offset,  bifurcate, and restrict growth in various layers, much
as a composite material hinders fracture growth across it. Various methods are now being used
to model such behavior (Wright et al. 1999, Miskimmins and Barree 2003, Weijers et al. 2005).

Several of the mechanisms can be seen in  Figure 12, which is a mineback photo of a fracture
propagating upward across several interfaces. The left-hand side is the unaltered photograph,
while the right-hand side has the fracture accentuated with a line drawn over it. There is
kinking, offsetting, and bending occurring  as the fracture makes its way through the layers. In
other cases, additional fractures are initiated or some fractures are terminated.

           Figure 12. Photograph and line drawing of fracture behavior crossing
           interfaces.
Figure 13. Schematic of types of
observed fracture behavior
crossing interfaces.
                               Figure 13 shows a schematic of several types of behavior that
                               have been observed in minebacks or laboratory tests. The
                               result of these behaviors could be any combination of
                               complexity, restriction, or termination of the fracture as it
                               propagates across the layered medium. Restrictions should
                               be common if kinking or offsets occur, as the width in the
                                          62

-------
kink or offset will necessarily be less than in the vertical part of the fracture due to both
geometric and stress considerations.

Discontinuities
Any heterogeneities and discontinuities can modify the
propagation behavior of fractures in a rock mass. Figure
14 shows an example of a fracture that is crossing
unhealed natural fractures (Warpinski et al. 1981), which
is also equivalent to the case of a weak interface with
some permeability along the interface. This example
shows offsets of the fractures at a location  that is very
close to the wellbore. Cement was used as  the fracturing
fluid for this test in order to preserve the width of the
fracture. Such offsets would clearly restrict fracture
growth because of the narrower width of the fracture in
the offset and the possibility of sand bridging.

There have been many studies of the factors that
influence fracture growth across discontinuities (e.g.,
Teufel 1979). These studies have demonstrated the
effects of stress, angle of approach, and various material
properties in blunting or offsetting fractures. These types
of offsets are likely responsible for much of the
complexity observed in hydraulic fractures  in  cores
(Warpinski et al. 1993, Branagan et al 1996) and mineback tests. They prevent fractures from
propagating as a single planar feature and instead force it into multiple, variably connected,
intersecting components. This complexity makes it difficult for fractures to grow large distances
as planar features.
Figure 14. Fracture crossing
discontinuities.
High permeability interval
High permeability zones can also terminate vertical fracture growth by dehydrating the slurry
through high leakoff. Coals are excellent examples of zones where fracture growth might be
terminated by this mechanism.
Summary
Hydraulic fracture growth is influenced by a multiplicity of factors that are common in any
reservoir. Of most importance is the in situ stress distribution, but interfaces, natural fractures,
and other heterogeneities may also significantly affect behavior.
                                           63

-------
References
Anderson, G.D. 1981. Effects of Friction on Hydraulic Fracture Growth near Unbonded
       Interfaces in Rocks. SPEJ 21:21-29.
Baree, R.D. and Winterfeld, P.M. 1998. Effects of Shear Planes and Interfacial Slippage on
       Fracture Growth and Treating Pressures. Paper SPE 48926 presented at the SPE Annual
       Technical Conference and Exhibition. New Orleans, Louisiana, 27-30 September.
Branagan, P, Peterson, R, Warpinski, N, and Wright, T. 1996. Results of Multi-Site Project
       Experimentation in the B-Sand Interval: Fracture Diagnostics and Hydraulic Fracture
       Intersection. Gas Research Institute Report GRI-96/0225, Chicago, Illinois.
Cleary, M.P.  1978. Primary Factors Governing Hydraulic Fractures in Hetrogeneous Stratified
       Porous Formations. Paper 78-Pet-47 presented at the 1978 ASME ETC Conference.
       Houston, Texas, Nov 5-9.
Cleary, M.P.  1980. Analysis of Mechanisms and Procedures for Producing Favourable Shapes of
       Hydraulic Fractures. Paper SPE 9260 presented at the 55th SPE Annual Fall Technical
       Conference and Exhibition. Dallas, Texas. 21-24 September.
Griffin, L.G., Wright, C.A., Davis, E.J., Weijers, L, and Moschovidis, Z.A. 1999. Tiltmeter Mapping
       to Monitor Drill Cuttings Disposal. Proceedings of the 37th Annual Rock Mechanics
       Symposium. Vail, Colorado. 2:1033-1040, 6-9 June.
Gu, H. and Siebrits, E. 2008. Effect of Formation Modulus Contrast on Hydraulic Fracture Height
       Containment. SPE Production and Operations 23:2-170-176.
Gu, H., Siebrits, E. and  Sabourov, A. 2008. Hydraulic-Fracture Modeling with Bedding Plane
       Interfacial Slip. Paper 2008 presented at the SPE Eastern Regional/AAPG Eastern Section
       Joint  Meeting. Pittsburgh, Pennsylvania. 11-15 October.
Hsiao, C. and El Rabaa, A.W. 1987. Fracture Toughness Testing of Rock Cores. Presented at the
       28th U.S. Symposium on Rock Mechanics. 141-148. Tucson, Arizona, 29 June - 1 July.
Jeffrey, R.G., Byrnes, R.P.,  Lynch, P.A. and Ling, D.J. 1992. An Analysis of Hydraulic Fracture and
       Mineback Data for a Treatment in the German Creek Coal Seam.  Paper SPE 24362
       presented at the SPE Rocky Mountain Regional Meeting. Casper,  Wyoming, 18-21 May.
Miskimmins, J.L. and Barree, R.D. 2003. Modeling of Hydraulic Fracture Height Containment in
       Laminated  Sand and Shale  Sequences. Paper SPE 80935 presented at the SPE Production
       Operations Symposium, Oklahoma City, Oklahoma, 22-25 March.
Nierode, D.E. 1985. Comparison of Hydraulic Fracture Design Methods to Observed Field
       Results. JPT 37:1831-1839.
Nolte, K.G. and Smith,  M.B. 1981.  Interpretation of Fracturing Pressure. JPT 33:1767-1775.
Palmer, I.D. and Luiskutty, C.T. 1985. A Model of the Hydraulic Fracturing Process for Elongated
       Vertical Fractures and Comparisons of Results with Other Models. Paper SPE 13864
       presented at the SPE/DOE  Low Permeability Gas Reservoirs Symposium. Denver,
       Colorado, 19-22 May.
Palmer, I.D. and Sparks, D.P. 1991. Measurement of Induced  Fractures by Downhole TV Camera
       in Black Warrior Basin Coalbeds. JPT43:270-275; 326-328.
Perkins, T.K.  and Kern, L.R. 1961. Widths of Hydraulic Fractures. JPT 13:937-949.
Shlyapobersky, J., Issa  M.A., Issa, M.A., Islam, M.S., Dudley, J.W., Shulkin, Y. and Chudnovsky, A.
       1998. Scale Effects on  Fracture Growth Resistance in Poroelastic  Media. Paper SPE
                                          64

-------
       48929 presented at the SPE Annual Technical Conference and Exhibition. New Orleans,
       Louisiana, 27-30 September.
Simonson, E.R., Abou-Sayed, A.S., and Clifton, J.J. 1978. Containment of Massive Hydraulic
       Fractures. SPEJ 18:27-32.
Smith, M.B., Rosenberg, R.J. and Bowen, J.F. 1982. Fracture Width - Design vs. Measurement.
       Paper SPE 10965 presented at the SPE Annual Technical Conference and Exhibition. New
       Orleans, Louisiana, 26-29 September.
Smith, M.B., Bale, A.B., Britt, L.K., Klein, H.H., Siebrits, E., and Dang, X. 2001. Layered Modulus
       Effects on Fracture Propagation, Proppant Placement, and Fracture Modeling. Paper SPE
       71654 presented a the SPE Annual Technical Conference and Exhibition. New Orleans,
       Louisiana, 30 September - 3 October.
Teufel, L.W. 1979. An Experimental Study of Hydraulic Fracture Propagation in Layered Rock.
       Phd Dissertation. Texas A&M University. College Station, Texas.
Teufel, L.W. and Clark, J.A. 1984. Hydraulic Fracture Propagation in Layered Rock: Experimental
       Studies of Fracture Containment. SPEJ 24:19-32.
Van Eekelen, H.A. 1982. Hydraulic Fracture Geometry: Fracture Containment in Layered
       Formations. SPEJ 22: 341-349.
Voegele, M.D., Abou-Sayed, A.S. and Jones, A.H. 1983. Optimization of Stimulation Design
       through the Use of In-Situ Stress Determination. JPT 35:1071-1081.
Warpinski, N.R., Northrop, D.A., Schmidt, R.A., Vollendorf, W.C., and Finley, S.J. 1981. The
       Formation Interface Fracturing Experiment: An In Situ  Investigation of Hydraulic
       Fracture Behavior Near a Material Property Interface. Sandia National Laboratories
       Report SAND81-0938. June 1981.
Warpinski, N.R., Fnley, S.J., Vollendorf, W.C., O'Brien, M., and Eshom, E. 1982. The Interface
       Test Series: An In Situ Study of Factors Affecting the Containment of Hydraulic Fractures.
       Sandia National Laboratories Report SAND81-2408. February 1982.
Warpinski, N.R., Schmidt, R.A., and Northrop, D.A. 1982. In Situ Stresses: The Predominant
       Influence on Hydraulic Fracture Containment. JPT.  34:653 - 664.
Warpinski, N.R. and Teufel, L.W. 1987. Influence of Geologic Discontinuities on Hydraulic
       Fracture Propagation. JPT 39-1: 209.
Warpinski, N.R., Lorenz, J.C., Branagan, P.T., Myal, F.R., and Gall, B.L. 1993. Examination of a
       Cored Hydraulic Fracture in a Deep Gas Well. SPE Production & Facilities 8-3:150.
Warpinski, N.R., Branagan,  P.T., Peterson, R.E., and Wolhart, S.L. 1998 An Interpretation of M-
       Site Hydraulic Fracture Diagnostic Results. Paper SPE 39950 presented at the SPE Rocky
       Mountain Regional/Low Permeability Reservoirs Symposium, Denver Colorado, 5-8
       April.
Weijers, L., Wright, C., Mayerhofer,  M. and Cipolla, C. 2005. Developing Calibrated Fracture
       Growth Models for Various Formations and Regions across the United States. Paper SPE
       96080 presented at the SPE Annual Technical Conference and Exhibition. Dallas, Texas,
       9-12 October.
Wright, C.A., Weijers, L., Davis, E.J. and Mayerhofer,  M. 1999. Understanding Hydraulic Fracture
       Growth: Tricky but Not Hopeless. Paper SPE 56724 presented at the SPE Annual
       Technical Conference and Exhibition. Houston, Texas, 3-6 October.
                                          65

-------
Zhang, X., Jeffrey, R.G., and Thiercelin, M. 2007. Effects of Frictional Geological Discontinuities
       on Hydraulic Fracture Propagation. Paper SPE 106111 presented at the SPE Hydraulic
       Fracturing Technology Conference. College Station, Texas. 29-31 January.
                                           66

-------
    Flow of Gas and Water in Hydraulically Fractured Shale Gas
                                    Reservoirs
                                       Zhong He
                            Range Resources Appalachia, LLC

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.
Underground fluid flow is primarily controlled by two physical factors: hydraulic conduits and
pressure gradients.  Both are required, or fluids will  not move. In their natural state, shale
formations are very impermeable, which means that there  are virtually no  natural hydraulic
conduits in the rock. Because of this, shale has often acted as a cap rock and effectively limited
and/or prevented fluids from escaping or migrating into other geologic formations over millions
of years (i.e., geologic time).

The flow capacity of the rock can be quantified by permeability. The permeability of the shale
matrix typically ranges from tens of nanodarcy to hundreds of nanodarcy (1 nanodarcy equals
10~6 microdarcy or 10~9 darcy). The shale matrix has such ultra-low permeability because of its
very small  pore size, which typically is on the order of tens of nanometer.  Although  natural
fractures may exist in shale formations, most of them are filled with minerals in their in-situ
conditions.

Because shale is so impermeable in its natural state, technologies such as horizontal drilling and
hydraulic  fracturing  are necessary to  enable  economical  production of  gas from  these
formations. The  process of hydraulic fracturing involves  creating  man-made  fractures or
fracture networks (i.e., hydraulic conduits) by pumping water and proppants (typically sand) at
high rates down the wellbore. By placing proppant into the conduits, they can be held open
over time so  gas can be effectively produced from the formation.  By drilling horizontal wells,
multiple  hydraulic fractures can be created in a single wellbore, which significantly increases
the ability of gas and water to flow out  of the shale.

The horizontal and vertical extent of hydraulically induced fractures are typically limited, being
confined by  such factors  as in-situ stress differences, formation leak-off  and the relative
properties of the target shale formation and surrounding geological strata. When in-situ stress
contrasts  are high,  propagation  of the hydraulic fracture  is prevented because  the  stress
contrasts serve as a barrier to fracture growth. Even in the absence of stress barriers, formation
leak-off will always arrest the fracture  height growth, meaning the injected fracturing fluid will
be absorbed into the strata with enough porosity and permeability, therefore stopping fracture
extension. Typically, hydraulic fractures grow on the order of only a few hundred feet vertically
and hundreds of feet horizontally. To  evaluate fracture geometry, there is a service industry
that collects data during fracture treatments. In addition, numerous hydraulic fracture models

                                          67

-------
have been developed to model fracture geometry. Considering that gas shales are often buried
several thousand feet (sometimes more than ten thousand feet) below the surface of the earth,
even large hydraulic fractures would still be confined many thousands of feet below the earth's
surface.

Not all induced fractures will result in conductive pathways between the formation and the
wellbore. After a fracturing treatment, the in-situ stress of the formation will close some of the
induced fractures, typically those without proppant. Those unpropped fractures may lose their
width,  and become disconnected with the propped fractures. Therefore, the effective  post-
treatment propped fracture lengths and  heights are always less than the  induced  lengths and
heights achieved during treatment.

During production, the horizontal wellbore serves as a pressure sink (i.e., the pressure is much
lower in the wellbore than in the surrounding shale formation),  causing the gas to flow from
the shale formation (high pressure  environment) into the fractures, and through the fractures
to the wellbore (low pressure environment). Since the shale  matrix has ultra-low permeability,
flow of gas in the unstimulated shale zone is minimal. Virtually no conduits allow water to flow
through  the  unstimulated  zone.  Consequently,  the gas/water  movement is  within the
stimulated  zone and towards  the  wellbore since the  pressure  gradient  is  in that  direction.
Migration of gas and water away from the stimulated zone is precluded.

Wells may be  shut-in  periodically.  During the shut-in,  pressure will  build  up within the
stimulated  shale zone.  However, the pressure within  the stimulated zone will be always less
than the pressure outside the  stimulated zone. A minimum amount of gas will keep flowing
from the unstimulated shale zone into the stimulated shale  zone. Again, migration of gas and
water away from the stimulated zone  is precluded.

Fracturing  treatment in  shale gas reservoirs typically uses water  as the fracture fluid to
propagate the fractures and  transport sands. A portion of this frac water will be produced up
the wellbore,  which is  isolated from the surrounding rock by  the steel casing  strings and
cement, back to surface during production. It is often referred to as flowback water. The flow
of water mainly exists within the fractures. Both water and  gas  flows together as multiphase
flow within the fractures toward the wellbore. The flow capacities of water and gas depend on
the relative permeabilities of each  phase, which are functions of the water saturation in the
fractures and  matrix. Initial water rates are high, but they decline quickly as water saturation is
reduced. The  water production typically tends to stabilize at low rates after a short period of
production. Over many years, about 20-40 % of the injected water will be produced  back.

Since frac water is  in contact  with the shale matrix  through fracture surfaces, water-phase
imbibition also plays an important  role in water flow. The imbibition effect is caused by the
capillary pressure between the  gas and water phases. The lower the reservoir permeability, the
higher the capillary pressure  will be. In low-permeability reservoirs such as shales, the capillary
pressure can  be thousands of psi. Once water is imbibed  into  the micropores of the  shale
matrix, it will quickly become immobile and therefore be retained  in the matrix permanently.

                                          68

-------
As discussed before, some induced but unpropped fractures may lose their width and become
disconnected from the propped fractures during the initial flowback period and long-term. In
this case, the water filling these fractures will become trapped and  remain immobile  during
production operations.

Because integrated reservoir models consider reservoir geology, the physics of fluid flow in
porous media, the nature of the fracturing treatment, production conditions, etc., they  can be
used to effectively quantify the  flow of fluids in subsurface formations. These models show that
the injected  and produced fluids are  contained within the shale formation or in the formations
immediately adjacent to them. These formations are  thousands of feet below the surface.
                                          69

-------
 Role of Induced and Natural Imbibition in Frac Fluid Transport
                            and Fate in Gas Shales
                                    Alan P. Byrnes
                            Chesapeake Energy Corporation

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.
Abstract
Hydraulic fracture modeling and fracture surface area calculations determined from production
data analysis and reservoir numerical flow simulation support estimates of created hydraulic
fracture (frac) surface areas of 24-60 MM sq ft for representative hydraulic fracture treatment
designs. Approximately 30+10% of the frac fluids are recovered and the remaining 70+10%
leaks off into the fracture face resulting in elevated frac fluid saturations within 2-6 inches of
the fracture face. Although natural imbibitions capillary forces can generally be ignored in
conventional reservoirs, in gas shales these forces can range from 200-2,000 psi. Drainage and
imbibition capillary pressure analysis indicate that the elevated near frac-face saturations are
not in equilibrium and capillary forces act to naturally imbibe the fluid back into the reservoir
away from the frac face. This process occurs over weeks to months and can be slowed or halted
by gas pressure decrease associated with well production. The frac fluids imbibed into the
formation are effectively locked in place with the native brine by capillary forces both during
the life of the well production and for geologic periods of time after reservoir depletion.

Introduction
Gas shale reservoirs characteristically exhibit low porosity (§ = 3-10%), low in situ specific
permeability (kj = 50-2,000 nD), low water saturation (Sw = 10-50%) and thickness of H = 50-
400 ft over large regions. To achieve economic gas production rates for these matrix properties,
flow to the wellbore is enhanced using multi-stage hydraulic fracture stimulation. Present
optimum well designs vary among operators and with reservoir properties but can be broadly
characterized as comprising horizontal wells with 4,500+1,000 ft of lateral length and with up to
60+20 fracture clusters along the wellbore. In a "large" hydraulic fracture stimulation (frac) as
much as 5+2 million pounds of proppant may be  used, transported by 120,000+20,000 bbls of
frac fluid, where 30+10% of the frac fluid is typically recovered during production.
Understanding the transport and fate of these frac fluids is important for environmental and
optimum well stimulation design reasons.

Frac fluid flows into and out of the rock formation through the frac face can be characterized as
comprising three flow periods: 1) an induced  imbibition period during and immediately
following the frac treatment and dominated by pressure-induced leakoff, 2) a natural imbibition
period when the well is shut-in following stimulation and capillary forces influence frac fluid
redistribution, and 3) flow out of the formation resulting from pressure drawdown in the
                                         70

-------
fracture and when capillary and viscous forces are potentially competing. Numerous studies
have explored hydraulic fracture modeling, which implicitly involves transport or flow of frac
fluids both in induced and natural fractures and into the formation through the created fracture
face. Published studies have extensively explored issues involving leakoff of fluids during
fracture creation and after pumping during fluid pressure decay. Foundational work by Nolte
(e.g., 1979, 1986, 1993) explored the relationships describing fluid leakoff associated with initial
spurt loss and subsequent filter-cake limited pressure-dependent leakoff. Additional work has
explored and summarized previous work on such issues as fluid loss in natural fractures
(Warpinski, 1990), effective fracture length (Barree et a I, 2003; Cipolla et a I, 2008), leakoff and
permeability (Meyerhofer and Economides, 1997), and  fracture modeling (Barree, 1983; Meyer
et al, 1990; Cipolla et al, 2011).

Fluid flow is influenced by two forces; 1) viscous force associated with induced pressure
differences resulting from pumping or well production,  and 2) capillary force associated with
interfacial tension among fluids and the rock pore surfaces. Capillary forces can generally be
ignored in reservoirs with k > 0.01 mD because reservoir methane-brine capillary forces are
generally only 1-100 psi and these forces are small compared to viscous forces associated with
flowing pressure drops. In contrast, very low permeability reservoir (0.000001 mD < k < 0.001
mD) threshold entry methane-brine capillary pressures  (Pte = Pc,Sw=i) range from approximately
Pte = 200-2,000 psi and  increase with decreasing water saturation. At these levels the influence
of capillary pressure on fluid movement and distribution cannot be ignored and can play a
significant role in the transport and fate of frac fluids. The influence of high water saturations
near the frac face were investigated by Holditch (1979)  and more recently by Cheng (2010).
Cheng's analysis illustrated the significant role that natural imbibition can play in fluid
distribution but did  not fully explore properties at high capillary pressures.

This brief abstract will utilize the above work and laboratory data to broadly analyze the
transport and fate of frac fluids in representative hydraulic fractures and illustrate that the low
fluid recoveries from frac treatments are consistent with the fracture  and rock properties and
that once frac fluids are imbibed into the reservoir capillary forces act to imbibe them away
from the frac face and hold them in place with capillary pressure forces of hundreds to
thousands of pounds per square inch (psi). The initial high water saturations result in near-frac
face blockage and reduced gas flow rates but natural imbibition results in a decrease in Sw over
time and "cleanup"  of the frac face and consequent increase in gas flow rates. The cleanup
period is influenced  by  such variables as the volume of water introduced, the permeability and
effective water permeability of the reservoir, whether the well is flowing or shut-in, the near-
frac face pressure, and  the specific capillary pressure properties of the reservoir rocks and can
occur over periods of weeks to months.

Frac Fluid Composition
Many variables are involved in fracture fluid chemistry design. Prior to pumping any fluid
systems, fluid-rock core measurements are used to determine the minimum fluid additives
necessary in each play to prevent formation damage  from drilling or fracture fluids. The
majority of the shale plays in North America are treated with a large percentage of "slickwater."

                                           71

-------
Slickwater is predominantly fresh water with four to eleven chemical additives at a combined
concentration of 1,000-6,000 parts per million by volume (ppmv) or 0.1-0.6 percent by volume
of the liquid pumped. Light gels are often used at the end of a stage to transport higher sand
concentrations. Chesapeake Energy's Green Frac™ program was initiated in 2009 to eliminate
additives not critical to successful completion and to replace necessary additives with more
environmentally benign chemicals.

Hydraulic Fracture Architecture and Induced Frac Fluid Imbibition
Numerous models exist for hydraulic fracture architectures and can broadly be classified as
ranging from simple planar fractures to complex fracture networks (Figure 15). The nature of
the fracture architecture that develops at any given location is a function of numerous variables
including but not limited to: magnitude of direction of horizontal stress field; vertical stress
profile; regional and local principal stress anisotropy; presence and orientation of one or more
natural fracture sets; rock elastic properties; fracture toughness; vertical and lateral
heterogeneity of rock properties; frac fluid properties; frac pump rates, pressures, and times
between injection period; proximity to frac barriers, nearby wells, and adjacent well histories;
and reservoir rock properties including porosity, permeability,  relative permeabilities, initial
water saturations, capillary pressure properties, pore throat size distribution, etc.

Based on the influence and interaction of these variables, fracture architecture will vary among
different shale plays, within a shale play, along a given horizontal well, and potentially even
within a given frac stage.  Microseismic data can be interpreted to support the predominance of
a given frac architecture within given areas. For the range of frac architectures that can occur,
frac modeling indicates that the proppant is deposited in a region representing only 20-50% of
                                              the total fracture system (Figure 16).
   Figure 15. Example of range of fracture
   architectures for a single initiation site.
   (modified from Fisher et o\, 2002).
     Simple Fracture
                                              Total effective surface area can be estimated
                                                using frac modeling, pressure transient
                                                analysis (PTA), production data analysis
                                                (PDA), and numerical flow simulation (NFS).
                                                These methods provide non-unique
                                                solutions that model observed pressure
                                                and flow behavior through time.
                                                Production decline analysis provides a
                                                solution for AVk (Area * k°'5). Knowing in
                                                situ reservoir effective gas  matrix
                                                permeability from core analysis, it is
                                                possible to define the total effective
                                                fracture surface area from  the early
                                                production and pressure data that is
                                                characterized  by transient unsteady-state
                                                flow (e.g., Miller et a I, 2010). These
methods indicate that "large" hydraulic fracture stimulation treatments utilizing approximately
120,000 bbls of fluid and 5 million pounds of proppant create effective fractures with surface
     'Complex Fracture
    With Fissure Opening
Complex Fracture
   Network
                                           72

-------
     See*   IHUWb
    > - ....
            W'flltsi
                            : •_::
Figure 16. Example fracture architecture models
generated using Meyer & Assocs. MFRAC and
MSHALE fracture simulation software. (A-
above) Simple planar fracture, (B-upper right)
discrete fracture network (DFN) with proppant
primarily deposited in the principal fracture, (C-
right) DFN with proppant distributed in lateral
fractures.  (DFN Figures courtesy of Meyer &
Assocs.)
 areas of approximately 12 MM sq ft (12 * 106 sq ft). Assuming the effective fracture surface
 area is 20-50% of the total, the total fracture surface area created is approximately 24-60 MM
 sq ft. This range in values is consistent with the total surface area estimated for either a simple
 planar fracture, compared to the high proppant concentration area (Figure 16), or for a central
 principal fracture or fracture set surrounded by unpropped or stranded lateral complex
 fractures (Figure 16). Generally, limited by material balance constraints and the PDA-defined
 surface area, a simple planar fracture potentially exhibits greater height or more effective half-
 length than the DFN architectures.

 Although the three fracture models shown in Figure 16 differ in architecture,  PDA and
 numerical flow simulation show that for the same effective fracture surface area, the initial
 transient unsteady-state flow is identical and that differences in production do not occur until
 inter-frac interference begins. This issue is highly relevant to optimum gas well production.
 However, for frac fluid transport, the surface area created by the fracing process, and the
 surface area into which frac fluids are injected, is approximately or can be exactly the same for
 all three architectures. Differences in fracture architecture do  not necessarily significantly
 change the total surface area into which frac fluids flow.
                                            73

-------
Fracture modeling shown in Figure 16 provides the basis for defining the distribution of frac
fluid in the near-frac face region for all fractures shown. Volumes of frac fluid induced to imbibe
into the formation are defined by various forms of the leakoff equation that describes initial
spurt loss (i.e., the initial loss of fluid before a filter cake is formed) followed by filter-cake
controlled leakoff. In it simplest ID form (Carter, 1957), the total fluid leakoff rate can be
characterized by:

       q(0 = 4 0!A(t) C/(«0)0'5  dA                                       [1]

where q(t) is the fluid loss rate at time t, A(t) is the fracture area of one face, C is the total
leakoff coefficient (including the initial spurt loss coefficient), t0 is the time of fracture area
creation.

Although injected volumes are greater near the fracture initiation location due to longer times
of injection, and  are greater in higher permeability rocks, the volume of frac fluid injected at
any given point on the frac face can be very approximately estimated assuming the total
volume injected  is uniformly distributed over the frac face surface area. For 120,000 bbls
(670,000 ft3) of frac fluid pumped, assuming this is uniformly  injected into 24-60 MM sq ft, then
the depth of penetration (Dffp) of a 100% saturated interval, in a rock with (() = 0.06 is Dffp ~ 2-6
inches.

Typically, following a large fracture stimulation treatment a well will be produced to recover as
much frac fluid as possible and then the well will be shut in for different time periods
depending on operator practices, surface facilities construction or hookup, or pipeline
scheduling. This initial production removes most of the frac fluid in the effective propped
fracture and the  shut-in period initiates the time of natural imbibition. Whether the well is
shut-in or begins production, following induced imbibition the frac fluids in the near-frac region
are influenced by natural imbibition resulting from capillary pressure forces.

Natural Imbibition
                                           74

-------
 Elevated frac water saturations in the near frac-face region are not in capillary pressure
 equilibrium. Air-mercury capillary pressure curves (converted to equivalent reservoir-condition
 methane brine pressures), generally representing the bounding range of those observed for gas
 shales with specific in situ Klinkenberg permeability ranging between 0.0002 mD > ki > 0.00005
 mD, illustrate that threshold entry methane-brine capillary pressures (Pte = Pc,Sw=i) range from
 approximately 2,000 psi > Pte > 250 psi and increase with decreasing water saturation (Figure
 17). These threshold entry pressures are consistent with threshold entry pressure-permeability
 relationships exhibited by lithic low-permeability sandstones and siltstones (Figure 18). Because
 the drainage curves in Figure 17 were measured using air-mercury, they represent drainage
 conditions where all pore surfaces are wetted by the wetting phase. The capillary pressures
 required to achieve or maintain the brine saturations in the present-day reservoirs (0.2 < Sw <
 0.5) do not exist in the reservoirs today. It can be hypothesized that the low reservoir water
 saturations were created by displacement of connate brine from the rock pore space during oil
Figure 17. General representative
drainage capillary pressure curves
for gas shales of low k (0.00005 mD;
red triangles) and high k (0.002 mD;
blue squares) measured using air-Hg
and converted to  equivalent
reservoir CH4-Brine pressures. Also
shown are measured water
saturations for natural imbibition of
core from as-received saturation
(black circles). Grey dashed curves
represent generalized imbibition
curves modeled from low-k
sandstones.
 10000


u
    10
               •High K
                Water Imbibition
                General Imbibition
      0.0  0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9  1.0
              Wetting Phase Saturation
 or gas generation when the formation rock was more porous, permeable and exhibited a lower
 capillary pressure. Additionally, the development of intra-kerogen porosity during catagenesis,
 which did not require displacement of brine, increased the total porosity and consequently
 decreased brine saturation because the brine saturation is referenced to the total pore volume.
 Water adsorption measurements indicate that kerogen surfaces exhibit mixed wettability and
 portions are both hydrophobic and hydrophilic. Under these conditions, the capillary pressure
 curves shown represent an end-member condition and methane-brine capillary pressure curves
 at reservoir conditions can exhibit lower slopes but still require similar threshold entry
 pressures because these are determined by the interparticle pore system of the water-wet
 mineral grains. An alternate model for capillary pressure-saturation conditions is that the
 reservoirs are presently undersaturated and are not in capillary equilibrium.  For this
 nonequilibium condition, the reservoir would presently be working to naturally imbibe water
 from surrounding formations but may be limited by availability of water, extremely low water
 relative permeability, and potential partial influence of a mixed wettability condition.
                                           75

-------
                     1000
                 c
                 LU
                 2
                 o
                 -c "i?
                 11
                 H 2
                 O 3
                 •ES
100
                 m
                 c >,
                 ll
                 I'5-
                 S ra
                 .!= O
                 o
                 o
                 «
                 o
                 a:
                      0.000001 0.00001 0.0001 0.001  0.01   0.1    1    10   100
                                In situ Klinkenberg Permeability (mD)
                                                               1000
 Figure 18. Threshold entry capillary pressures (Pte) versus specific in situ Klinkenberg
 permeability for lithic sandstones (blue squares) and representative gas shales (open
 triangles). Pte were measured using air-Hg and converted to equivalent reservoir CH4- Brine
 pressures. Data for shale show continuity with trend for sandstones and siltstones.
 Relationship can be expressed: Pte = 12.25
                    -0.424
The wettability and imbibitions properties of gas shales can be tested by performing imbibition
capillary pressure measurements on core in as-received condition. Typically wettability is
measured using the Amott or USBM methods but these methods are experimentally difficult to
perform on gas shales due to their low permeability and high capillary pressures. A simple
limiting condition test is to perform a natural imbibition test which represents the condition of
brine imbibition at low gas-brine capillary pressure and the resulting equilibrium brine
saturations. For this test, if kerogen surfaces are hydrophobic then brine will only be imbibed
into the water-wet portions of the mineral-lined pore pace and a trapped residual gas
saturation, representing the gas in the kerogen pores, will result. If the kerogen pore surface is
sufficiently hydrophilic then water is imbibed into the complete pore space and residual
trapped gas saturations are low. The imbibition data shown in Figure 17 for the condition of 1
psi capillary pressure indicate that gas shales are capable of naturally imbibing water leaving
very low residual gas saturations.

Research on the exact imbibition capillary pressure curve shape is on-going but curve shapes
characteristic of low-permeability sandstones and  siltstones (Byrnes and Cluff, 2009) are likely
to be representative (Figure 17). These curves indicate that imbibition capillary pressure forces
between the elevated water saturations near the frac face and the lower water saturations in
the reservoir produce a capillary pressure drive mechanism of hundreds to thousands of psi.
This force  acts to naturally imbibe the frac water away from the frac face and into the
formation where it is held in place by those  same forces.
                                           76

-------
If the well is produced before the high water saturations are reduced near the frac face then
the elevated near frac-face water saturations can be stabilized by the lower capillary pressures
that can result from gas pressure depletion. When a well is fully pressure depleted it can be
projected that the reservoir would naturally imbibe water from surrounding intervals. The time
period over which this occurs would be a function of many variables including the formation
effective water permeability, surrounding formation effective water permeabilities and
capillary pressures in the formation and in the surrounding formations.

Discussion
Hydraulic fracture modeling and fracture surface area calculations determined from pressure
decay analysis and reservoir numerical flow simulation support estimates of created  hydraulic
fracture surface areas of 24-60 MM sq ft. Approximately 30+10% of the frac fluids are
recovered and the remaining 70+10% leaks off into the fracture face resulting in elevated frac
fluid saturations within 2-6 inches of the frac face. Drainage and imbibitions capillary pressure
analysis indicate that these saturations are not in equilibrium and capillary forces act to imbibe
the fluid back into the reservoir away from the frac face over a period of weeks to months and
can be slowed or halted by gas pressure decrease associated with well production. The frac
fluids imbibed into the formation are effectively locked in place by capillary pressure forces of
hundreds to thousands of pounds per square inch (psi) both during the life of the well
production and for geologic periods of time after reservoir depletion. Further research is
needed and is ongoing as to the exact shape of the  imbibition capillary pressure curves and
imbibition water relative permeability curves to fully quantify the imbibitions process and
timing.

Selected References
Barree, R.D.,  Cox, S.A., Gilbert,  J.V., and Dobson,  M.:  "Closing  the  Gap: Fracture Half Length
      from Design,  Buildup, and Production  Analysis," SPE 84491,  Proc.  2003 SPE Annual
      Technical Conference and Exhibition, Denver, Colorado, 5-8 October.
Barree, R.D, 1983, "A  Practical Numerical Simulator for Three Dimensional Fracture Propagation
      in Heterogeneous Media," SPE 12273, presented at the 1983 SPE Symposium on
      Reservoir Simulation, San Francisco, CA, Nov. 15-18, 12 pgs.
Byrnes, A.P., and R.M. Cluff, 2009, "Analysis of Critical Permeablity, Capillary Pressure and
      Electrical Properties for Mesaverde Tight Gas Sandstones from Western U.S. Basins",
      Final Report, U.S. Department of Energy contract #DE-FC26-05NT42660, 355 pgs.
Cipolla, C., X. Weng, M. Mack, U. Ganguly, H. Gu, 0. Kresse, and C. Cohen, 2011, "Integrating
      Microseismic Mapping and Complex Fracture Modeling to Characterize Fracture
      Complexity," SPE 140185-MS, SPE Hydraulic Fracturing Technology Conference, 24-26
      January 2011,  The Woodlands, Texas, USA, 22 pgs.
Fisher, M.K., Davidson, B.M., Goodwin, A.K., Fielder, E.G., Buckler, W.S., and N.P. Steinberger,
      2002,  "Integrating  Fracture  Mapping  Technologies to  Optimize Stimulations  in  the
      Barnett  Shale,"  SPE  77411-PA_P,  presented at  the  2002 SPE  Annual  Technical
      Conference and Exhibition, San Antonio, Texas, USA, Sept 29-Oct  2, SPE Production and
      Facilities, May, p. 85-93,
                                          77

-------
Mayerhofer, M.J., and M.J. Economides, 1997, "fracture-injection-test Interpretation: Leak-Off
       vs. Permeability, SPE 28562, SPE Production and Facilities, Nov., p. 231-236.
Meyer,  B.R., Cooper,  G.D., and S.G. Nelson, 1990,  "Real-Time  3-D  Hydraulic  Fracturing
       Simulation: Theory and  Field  Case Studies," SPE 20658-MS, Proc. SPE Annual Technical
       Conference and Exhibition, 23-26 September 1990, New Orleans, Louisiana, p. 417-432.
Miller, M.A., Jenkins, C., and R.  Rai, 2010, "Applying Innovative production Modeling techniques
       to  Quantifying fracture  Characteristics, Reservoir properties, and  Well Performance in
       Shale Gas  Reservoirs," SPE 139097, presented at 2010 SPE Eastern Regional Meeting,
       Morgantown, West Virginia, USA, Oct 12-14, 12 pgs.
Nolte, K.G., 1986, "A General Analysis of Fracturing Pressure Decline With Application to Three
       Models," SPE Formation Evaluation, Dec., p. 571-583.
Nolte, K.G., Mack, M.G., and Lie, W.L.: "A Systematic Method  of Applying Fracturing Pressure
       Decline  : Part  I,"  SPE 25845,  Proc., SPE Rocky Mountain Regional/Low Permeability
       Reservoirs Symposium, Denver (1993), p.  31-50.
Warpinski, N., 1988, "Dual Leak-off Behavior in Hydraulic Fracturing of Tight, Lenticular Sands,"
       SPE 18259, SPE Production Engineering, p. 243-252.
                                          78

-------
Summary and Abstracts from Theme 3: Models to Predict
                    Transport
                        79

-------
    Summary of Presentations on Theme 3: Models to Predict Transport
The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.

Technical Presentations
This set of technical presentations addressed demonstration of models and determinations of
model accuracy.

Andrew Havics,  pH2 LLC, discussed modeling philosophies and applications. He described key
aspects of modeling for risk assessment, fate and transport, and exposure, including
assumptions, model selection, selection of input parameters, sensitivity, and validation. Mr.
Havics also emphasized the importance of understanding the limitations of models, referring to
the quote, "All models are wrong and some are useful." He pointed out that models are most
often tied to a region or area for which specific site data is required. He also noted that once a
model is chosen it would create a framework from which parameter selection would then drive
the model output.

Denise Tuck, Halliburton, provided information on modeling HF  fluid fate and transport. She
gave an overview of potential migration pathways, discussed key input parameters, and
reviewed sources of data for these parameters, including HF fluid additive information from
industry, information collected as  part of spill response measures, and information collected
during well installation and stimulation.  Ms. Tuck recommended that EPA identify  key marker
compounds for fate and transport evaluation. She also recommended that EPA utilize all
available data, particularly historic data on spills/releases, and assess human health risks
associated with drinking water contamination. She emphasized that data collection at the time
of installation is  important. She also indicated possible confounding factors to consider, such as
naturally migrating gases and abandoned wells.

Manu Sharma, Gradient, described an approach for modeling drinking water-related human
health risks from HF fluid additives. He described toxicity and exposure implications for two
surface release scenarios (acute and diffuse) and various  migration pathways, using available
information on HF fluid composition, flowback characteristics, and chemical toxicity. Mr.
Sharma concluded that, based on the resulting dilution attenuation factors (DAFs), which are
conservatively high, human health risks associated with HF additives and flowback are very
small.
                                         80

-------
 Summary of Discussions Following Theme 3: Models to Predict Transport
                                   Presentations
The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.

Clarifying items from the technical presentations. A participant asked about the log-normal
hydraulic conductivity plot in Mr. Navies' presentation (and included in the abstract) Modeling
Philosophies and Application. This plot includes data from 28 sites in the Denver-Julesburg (DJ)
and Piceance Basins. The results reflect fluid flow in the shallow subsurface. Mr. Havics added
that the model was deterministic and that the results show concentrations at the leading edge
of the plume. Another participant asked why Mr. Sharma's analysis did not include an
impoundment scenario. The presenter indicated that while an impoundment scenario was not
considered relevant to this particular investigation, impoundment results are available and
have been filed with the New York Department of Environmental  Conservation (NYDEC). A
participant asked which metals were included in Mr. Sharma's analysis (described in the
abstract Modeling Drinking Water Related Human Health Risks from Hydraulic Fracturing
Additives). The list of metals is included on  Slide 17 of Mr. Sharma's presentation; radium was
not included because, in this initial data set from the Marcellus Shale, the radium detection
frequency was very low. Participants noted that this low detection frequency may be related to
limitations of the analytical methods used.

Stray gas migration. A participant asked about the industry's position on stray gas migration.
The presenters indicated that, in their experience, most (though not all) gas migration incidents
are related to natural stray gas migration and that they have not seen evidence that gas
migration incidents are  related to HF at depth. Another participant suggested that pumping
ground water for refracturing water needs  could lead to depressurization of the shallow
aquifer, possibly resulting in methane outgassing or desorbing—in other words, an indirect
source of gas migration. Other participants indicated that this could be possible, especially for
pads with large numbers of wells.

Probabilistic modeling. A participant asked  about the possibility of using simple deterministic
models in a Monte Carlo approach to characterize uncertainty. The presenters responded that
probabilistic approaches can be very useful. However, according to the presenters, challenges
include identifying available data and selecting relevant underlying distributions.

Risk and DAFs. A participant asked about the basis for the vadose  zone DAFs in Mr. Sharma's
presentation. The presenter explained that the DAFs are the  result of simple analytical solutions
that consider various processes separately. The presenter indicated that a more complex
numerical model would  allow the user to vary the organic fraction and other parameters. The
presenters noted that the DAFs are quite conservative and are comparable to previous EPA
results. Another participant asked if surface water impacts were quantified. The lowest annual
average mean daily discharge was calculated based on data from  USGS gauging stations and

                                         81

-------
records. This DAF was on the order of 60,000 on an annual average basis. A participant asked
for a comparison of DAFs relative to underground storage tank (UST) petroleum releases. These
two scenarios are quite different; the  UST release is a non-aqueous phase liquid release, which
is modeled as an infinite source. The DAFs are much smaller because the  plume is constantly
fed by concentrations that are at or near the solubility limit for the compound. The HF scenario,
on the other hand, is a discrete spill in the dissolved phase. One presenter recommended a
paper that provides information on maximum plume sizes for UST releases.

Sources of information. A participant noted that spill information and other types of state data
are generally available online in well-by-well format and not compiled into databases. Other
participants also indicated  that data are often presented in well-by-well format, though
Colorado does maintain a database and the state allows outside groups to perform some data
mining. Participants noted  that some states are more willing to provide data than others; in
addition, states all collect different types of information. Another participant explained that
data mining efforts can be  extensive; even matching well names and  numbers can be a
challenge.

Mixing in bedrock. A participant asked about the DAF for mixing in bedrock (DAFBR), specifically
whether the migration would be more similar to slug movement or diffusion and mixing. The
presenter explained that this scenario is similar to slug movement, though it only involves
gradual migration through pore spaces (not fractures), and that this type  of migration would
occur on the scale of thousands of years. According to the presenter, the  DAF is based on mass
loss over a distance in addition to dilution factors, and above a certain zone, migration is more
convection based than diffusivity based. Also, the presenter noted that all of the fluids moving
through this system are essentially exposed to activated carbon over large surface areas.

Octanol-water portioning coefficient (Kow) analyses. A participant asked about the limitations of
log KOW analyses. A  presenter noted that different lab tests can be run with different soil
materials, but that  Koc analyses might be better, especially when clays are present. Another
participant asked whether ionic and non-ionic surfactants should be considered separately
when determining Kow-  While it would be possible to compile the data and separate the types
of surfactants, a presenter noted that it is generally easier and more cost-effective to consider
them together and select conservative values. A participant indicated that there is a lot of
existing information on this topic, though it is generally available only through older or foreign
journals.
                                          82

-------
             Abstracts for Theme 3: Models to Predict Transport
Abstracts were submitted to U.S. EPA by the presenters for use in this proceedings document.
              Not all presenters submitted abstracts of their presentations.

The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.
                                       83

-------
                  Modeling Philosophies & Application
                   Andrew A Havics, CHMM, CIH, PE1 and Dollis Wright2
                                          , LLC
                                       2QEPA

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.
Models are mere surrogate test fields for answering a question(s) or solving a problem
expediently. They can be physical analogs or computer-based (Bear, 1972). Regardless, they all
begin with a framework and a set of assumptions and limitations that go along with that
framework. As a result, all models are wrong, but some are useful. Assumptions and limitations
begin before selecting models but also arise from the selection or specialized application of a
model. In terms of fracing and Risk Assessment (RA), there are several places where models and
parameters must be chosen to complete hazard identification including chemical selection, fate
& transport, exposure assessment, and risk determination aspects.  Furthermore, to evaluate
the model(s) chosen, and at a minimum qualitative assess or rank their values of data
(response/output), an analysis of sensitivity and validity of the models should be completed.
The level of complexity in the model, its use, and evaluation will be based on a number of
factors derived for the purposes, objectives and goals of the model's use, many of which may
be directed by policy.

Assumptions, leading to Limitations
Beginning with the assumptions, there are de facto policy aspects. The first is whether the
parameters for the models will be either (a) discrete  - single point values, or (b) stochastic -
continuous function of the probability of a value. By choosing a discrete model, one must then
decide what are the appropriate parameters, constraints, or additional assumptions. For
instance, should one use a Maximum Likelihood Estimator (MLE), Mean, Median, Maximum, or
some stochastic upper limit as the single point value  input. Compiled or repeated use  of
maximums or extreme values in equations, such as hydrogeologic data or human exposure
estimates, increase the likelihood of overestimating risk and in some cases the cost of
subsequent actions; but this  must be weighed against benefits of communicating the level of
conservatism and the simplicity of using it as a screening tool. At some point, the weight of
evidence surrounding the parameter such as distance to point of exposure (POE) must be
assessed and a professional judgment made. In certain cases, regulatory restrictions will
determine initial estimates. In the case of fracing, the source material composition by  basin will
influence the selection of a parameter, e.g., the chemicals of interest and their anticipated
concentrations. Again, regulatory stipulations, such as pre-treatment can and will affect the
selection of an appropriate input value.  Furthermore the geology & hydrogeology of a region or
local area can also influence chemical selection and fate and transport parameters, whether the
selection of PAHs from coalbed deposits or hydraulic conductivity of a formation. Based on

-------
these variations, it is clear from the science that a single model of predicting fate, transport, or
risk (or even one method of regulatory control) should not be applied to all locations.
Limitations
There are limitations in model selection, either constrained from the geology/hydrogeology, or
constrained from the chemical side (metal versus organic), from the media of concern selected
or from the pathways of anticipation. The limitation of sufficient, good quality or robust data
will certainly restrict the use of stochastic estimates, but will also require a good analysis of
sensitivity. The models themselves can (and usually are) created to permit a bias in the way of
either over- or under-prediction of transport or exposure depending on how values are
selected. Thus, a listing of all input values as well as the model structure should  be  available in
any prediction  using a model. There are limitations that are derived from precision of the
model, and those that come from precision of the parameters. Calibrating the model to known
field scenarios  or lab-based experiments can provide an estimate of that precision. An analysis
of sensitivity should follow to gage the relative importance of model variability versus model
sensitivity.

Model Selection
By default, model selection restricts options. The  more complex the model, generally the more
costly, the more difficult to assess precision, the more difficult to understand the results and
communicate them. One should consider the value of model refinement and model complexity
relative to any  gain in understanding the ultimate endpoint - risk. Regardless, in selecting a
model, boundary conditions must be chosen and applied. In some  models, these will drive fate
and transport more than others. Because boundary conditions are likely to differ between
geologic formation, application of one model from one formation to another should proceed
cautiously.

Toxicology in the RA Model
In terms of RA, orders of magnitude are the norm. For toxicologists 3 times 3 is 10, and this
level of rounding or semi- to multi-order-of-magnitude math pervades. First, one must
recognize that  parameters for doses that represent safe levels from toxic endpoints are
selected with certain safety and uncertainty factors built in. For non-cancer agents these are
typically 10- to 100-fold lower than No Observed Adverse Effect Levels (NOAELs) which are
usually 10-50 times lower than Lowest Observed Adverse Effect Levels (LOAELs) [Dourson,
1996; Haber, 2002]. These are based on many studies indicating ratios of low or no response to
a response between species and over differing times (days-months-years) and generally
represent a conservative estimate times another conservative estimate. If less data is available
or equivocal, a Benchmark Dose Level  (BMDL) approach can be applied first; it brings a more
stochastic approach to another deterministic aspect [EPA, 2000]. Examples are Barium (300-
fold factor) [EPA, 2005] and Benzene (<10 factor)  [ATSDR, 2007]. For cancer agents, risk is
usually presumed to follow a straight from a  projected 95% confidence limit to zero, thus is
intentionally conservative in its application. The ultimate acceptable risk level is a policy
decision but usually ranges in the 1 in  100,000 to  1 in 1,000,000 for residential settings and 1 in

-------
10,000 to 1 in 100,000 for commercial or industrial settings [Kocher, 1991]. For comparison
sake, one can consider two scenarios - a) struck by lightning and b) killed in a vehicle accident,
both over a lifetime. The first, the lightning strike, is on the order of 6 in 1,000,000, the second
(vehicular death) is about 1-2 in 100 people. Basic risk aside, there are assumptions or defaults
that must be made with regard to multiple pathways from fate and transport, multiple routes
of exposure, multiple toxic endpoints, and multiple mixtures. The most sensitive toxic endpoint,
say liver damage, would generally be used. Generally, all doses are added for multiple routes of
exposure or multiple pathways. As for mixtures of chemicals, this is a policy aspect and can
range from no summing of risk, to summing only like toxic endpoints, to summing all aspects.
Even after consideration of risk is made, there are potential limitations such as background
amounts of an agents (e.g., arsenic in soils), and lack of tox data for additives or proprietary
mixtures. The lack of data might be able to be handled using a control banding technique
(Nelson, et al., 2011). Ultimately, the chemical selection and fate and transport aspects cannot
be divorced from the models or the remainder of the RA, and science and policy must find
consensus.

Fate & Transport
The selection or input of physical-chemical properties such as solubility, retardation, and Log
Octanol Water partition coefficient (Log Kow), can drive an equation in terms of fate and
transport. There are, however,  certain generalities as well as  pitfalls in their selection and use.
Kow within a group of chemicals can be estimated within reason, but the value can easily affect
transport estimates but will  also affect dermal exposure. Barium Sulfate (BaS04) is generally
considered to be relatively insoluble, but it can become more soluble, thus available for
transport and human uptake, in the presence of high chlorides (Templeton, 1960). High
chlorides were found present in our study and the dissolved Barium also tended to rise with
increasing chlorides. Despite these effects, the effect of the hydrogeology in the Colorado
pathway scenarios was not affected by this in terms of showing a significant risk.

Hydrogeologic parameters can vary widely from region to region and formation to formation.
Thus, either a conservative selection or field data should be used. Because hydraulic
conductivity (Kha) can significantly influence transport models, a review of 28 sites in the DJ and
Piceance were selected for detailed assessment of hydrogeological parameters including
hydraulic conductivity.  Each location was reviewed to establish local a geologic and
hydrogeologic setting, and data from the vicinity on these aspects was then gathered. Kha
ranged from 0.01-0.0000001 cm/sec, with most values in the 0.001-0.0001  range, and the
hydraulic conductivity appeared log normally distributed, which was expected. Using this data
(and other representative data  from the region) for Benzene with a retardation value of 1.5, the
velocity of the benzene might be expected to be 27.5 ft/year at a Kha of 1E-3 versus 0.028 ft/yr
at a Kha of 1E-6. One can see the necessity for gathering and entering relevant data into a
model. For the modeling in the Colorado study, a Kha of 3.63E-3 was ultimately used, driven by
regulatory concerns [CDPHE, 2007].

Similarly for leaching from pits, the depth to groundwater (GW), can strongly affect transport.
For the Colorado study, water well logs from <= 1.5 miles from each well pad in the DJ and

-------
Piceance were pulled from state records and depth to GW evaluated. A total of 42 water wells
provided sufficient data for analysis. The distribution appeared lognormal. Values ranged from
3 to 315 feet with the 5% quantile at 7.8 feet. Ultimately, 3.3, 9.8, and 20 feet (1, 3, 6.1 m)
estimates were used  in the modeling. The wells logs also were used to estimate point of
exposure (POE) assuming the distance from the oil well directly to the water well. The  results
ranged from 72 to 2,420 meters (236-7,939 feet) with a 5% quantile at  135 m (442 feet). For the
study, the minimum distance of 72 meters was selected by the risk assessor as a POE for the
subsequent modeling; however, a 5% quantile value would likely be a reasonable maximum
likelihood estimator (MLE). One should note that the 72 m is not likely as regulation restrictions
have mandated setbacks to wells of 91-183 m (300-600 feet) with more stringent distances
proposed [COGCC, 2008], again, a limitation imposed from outside the  science. As a surrogate
reference, one could  consider plume data from 604 actual sites from other states for known
significant pure product releases (only diluted fluids and solids are expected  present in the O&G
sites) reveal 75% are  under 200 ft and  most are shrinking [Newell, 1998].

 Exposure
Several assumptions go into the exposure assessment process and a  number of limitations also
arise from this. The first is the selection of what is the exposed resource/population at risk, and
it may be a natural resource such as water quality, a defined  ecosystem to include fish or ducks,
or human health. The toxicity of copper exposure to humans is less significant than to fish; the
use of a typical surface water in Colorado isn't quite the recreation use that it might be for
typical default scenario the EPA RAGS  [EPA, 1989a, 1989b, 1991a, 1991b] or the ASTM RBCA
[2002] and thus parameters of exposure might be overestimated. On the other hand, dust
generation may be more than that usually considered by EPA or ASTM, as might be irrigation
for grass. Furthermore, the determination of the need for a specific fate and transport model to
assess these pathways will be controlled by their selection in the first place. Because the
selection of completed pathways can be significant in any RA, there is a strong need for
accurate transport models. The discrete selection of what is and what isn't complete must be
considered in light of typical patterns of use, along with distances and relevant geophysical
parameters.  It is often driven by the media selected as "contaminated" or of concern.  In the
study in Colorado a number of media were selected and each relevant pathway was assessed
with some form of fate and transport model. Media were:

       Solids placed in Pits
       Liquids in  Subsurface Pits
       Fracing Fluid placed in pits
       Fracing Fluid placed in containers
       Produced Water placed  in containers
       Produced Water placed  in pits
       Drilling fluids in drilling
       Drilling fluids in pits

Thus, leaching from pits and migration to groundwater followed by transport to a residential
well was modeled. On the other hand, because houses are not built on waste pits and  distance

-------
to houses are so great, direct vapor intrusion was not considered a complete pathway. Because
each completed pathway leads to exposure, the dose from each was assessed as additive. Thus
the removal of a pathway also removes a potential dose. However, because multiple pathways
and thus exposures and doses are modeled on the same population (e.g., person), an
overestimate of real exposure is likely to arise. This is where stochastic modeling is very useful.
The RA was completed using a variety of assumptions, details of the RA are provided in the
QEPA report [QEPA, 2008].

Sensitivity
A sensitivity assessment should be performed on any model, even if crudely done, unless the
RA itself evaluates multiple scenarios such that sensitivity to basic parameters is obvious. That
said, the purpose of a sensitivity analysis is: (a) to quantify the uncertainty in the calibrated
model caused by uncertainty in the estimates of parameters, stresses, and boundary
conditions, and (b) to identify the model inputs that have the most influence on model
calibration and predictions. ASTM has provided guidelines for assessing GW models in this
fashion (ASTM, 2002). A direct sensitivity analysis involves checking the response after varying a
particular parameter through a range of values, or alternatively, taking the derivative of the
response function and plotting that. Because not all parameters are based on the same scale, a
normalized sensitivity measure can be used to more equally evaluate different parameters
(Norton, 2008). Even so, this author has found normalizing to the data range (R) or 5%-95%
quantile range provides a better representative of response relative to the expected range of
values for a parameter.

Validation
It is useful at some point to validate the prediction of a model. This can be accomplished by pre-
planned injections, correlating actual (unintentional) releases, running parallel models, or
evaluating health outcomes or using biomarkers in the case of RA. The first option has been
accomplished for petroleum releases by intentionally injection product into  outdoor locations
and indoors in laboratory setting, but not at great depths. Correlations to actual field data have
been limited or proprietary in nature. Parallel models are run in select situations but don't offer
real world calibration, in the case of the Domenico model used in the Colorado study, it has
been evaluated for some aspects and found to have errors where the  longitudinal dispersivity
parameters are high or where the Peclet number (Pe)  is low (Guyonnet, 2004; West, 2007) or
the study, the Pe was in the range of 80-180. As a follow up in the 2008 study, a review of a
2006 groundwater data study (URS, 2006) from wells in one basin for the O&G drilling locations
was conducted, and additional data reviewed (as available) for the  site locations in the 2008 RA.
For the 2006 study, no Benzene (the most likely contaminant given the local conditions) was
detected. For the 2008 review some BTEX compounds were identified, but none of the
chemicals was over their respective drinking water limits; also, confounding sources could not
be ruled out where detections were observed.

Acknowledgments
The study upon which a portion of this presentation is based was funded by the Colorado Oil
and Gas Association (COGA). I would also like to acknowledge Mark Leverson, PG, Principal

-------
Hydrogeologist with URS for his work and assistance on the hydrogeology aspects.

References
ASTM: El 739-95(2002), Standard Guide for Risk-Based Corrective Action Applied at Petroleum
      Release Sites. ASTM, Conshohocken, PA. 2002.
ASTM: D5611-94, Standard Guide for Conducting a Sensitivity Analysis for a Ground Water Flow.
      ASTM, Conshohocken, PA. 2002
ATSDR: Toxicological Profile for Benzene, Draft. ATSDR, August, 2007.
Colorado Department of Public Health and Environment, Hazardous Materials and Waste
      Management Division, Table 1 Colorado Soil Evaluation Values (CSEV) - December 2007
Colorado Oil and Gas Conservation Commission, Draft Rules for Oil and Gas Development in
      Colorado, (HB 1298 & HB 1341), March 31, 2008.
Bear, Jacob: Dynamics of Fluids in Porous Media. Dover, NY. 1972.
Dourson, M., S Felter, and D. Robinson: Evolution of Science-Based Uncertainty Factors in
      Noncancer Risk Assessment. Reg Tax Pharmacol. 24:108-120.1996.
EPA: Risk Assessment Guidance for Superfund, Volume I, Human Health Evaluation Manual, Part
      A, Interim Final. EPA 540/1-89/002. USEPA: Washington, DC. 1989a.
EPA: Risk Assessment Guidance for Superfund, Volume II, Environmental Evaluation Manual,
      Interim Final. EPA 540-1-89-001. USEPA: Washington, DC. March, 1989b.
EPA: Risk Assessment Guidance for Superfund, Volume I, Human Health
Evaluation Manual, Part B: Development of Risk-Based Preliminary Remediation Goals, Interim.
      EPA 540/R-92/003. USEPA:  Washington, DC. December, 1991a.
EPA: Risk Assessment Guidance for Superfund, Volume I, Human Health Evaluation Manual, Part
      C: Risk Evaluation of Remediation Alternatives. USEPA: Washington, DC. December,
      1991b.
EPA: Final Guidelines for Exposure Assessment. US EPA, Washington, DC. Vol 57, Federal
      Register No.  104. May 29, 1992.
EPA: Review of risk levels for 32 states. 1996.
EPA: Benchmark Dose Technical Guidance Document, DRAFT, External Review Draft, EPA630R-
      00001. 2000.
EPA: Toxicological Review of Barium and Compounds  (CASRN 7440-39-3), EPA-635-R-05-001,
      June, 2005.
Guyonnet, D. and C. Neville:  Dimensionless Analysis of Two Analytical Solutions for 3D Solute
      Transport in  Ground Water. J Contaminant Hydrology 75:141-153. 2004.
Haber, L, and A. Maier: Scientific Criteria Used for the Development of Occupational Exposure
      Limits for Metals and Other Mining-Related Chemicals.  Reg Toxicol Pharmacol. 36:262-
      279. 2002.
Kocher, David, and F. Owen Hoffman: Regulating Environmental Carcinogens: Where do we
      draw the line? ES&T. 25(2):1986-1989.1991.
Nelson, Deborah Imel, Stephen Chiusano, Anne Bracker, Lance Erickson, Charles Geraci,  Martin
      Harper, Carolyn Harvey, Andrew Havics, Mark Hoover, Thomas Lentz, Richard Niemeier,
      Susan Ripple, Erica Stewart, Ernest Sullivan, and David Zalk: Guidance for Conducting
      Control Banding Analyses. AIHA, Fairfax, VA. 2007
Newell, CJ, and JA Connor: Characteristics of Dissolved Hydrocarbon Plumes, Results from Four

-------
       Studies. API Newsletter December, 1998.
Norton, JP: Algebraic sensitivity analysis of environmental models. Environ Model & Soft
       23:963-972. 2008.
QEPA: Pathway Analysis and Risk assessment (PARA) For Solids and Fluids Used In Oil and Gas
       Exploration and Production in Colorado, pp. 1-930, June 2008.
Templeton, C.: Solubility of Barium Sulfate in Sodium Chloride Solutions from 25° to 95° C. J
       Chem Eng Data 4:514-516. 1960.
URS: Phase 1, Hydrogeologic Characterization of the Mamm Creek Area In Garfield County,
       March 23, 2006 - for Board of County Commissioners Garfield County, Colorado.
West, M.R., B.H. Kueper, and M.J. Ungs: On the  Use and Error of Approximation in the
       Domenico (1987) Solution. Ground Water 45(2)126-135. 2007.

-------
   Modeling Drinking Water Related Human Health Risks from
                      Hydraulic Fracturing Additives
   Manu Sharma, M.S., P.E.; David E. Merrill, M.S.; Ari S. Lewis, M.S.; Sam A. Flewelling, Ph.D.
                                       Gradient

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.
Potential impact of hydraulic fracturing (HF) activities on drinking water aquifers is being
studied by the United States Environmental Protection Agency (US EPA) at the request of
Congress. Although HF has been widely used for natural gas development for many years, this
topic is receiving greater attention in the media and the scientific community as large new
natural gas reserves are being proposed for development. A number of regulatory agencies,
                         r\
including the US EPA (2004),  have previously assessed potential impacts to drinking water
aquifers,  and concluded that HF activities are not expected to affect drinking water aquifers.
Among them, is a comprehensive  evaluation undertaken by the New York State Department of
Environmental Conservation (NYSDEC), which published a Draft Supplemental Generic
Environmental Impact Statement  (SGEIS) addressing permitting requirements for the
development of natural gas production wells in the Marcellus Shale formation (NYSDEC, 2009).3
Our presentation relies on the HF-related information presented in the NYSDEC SGEIS, and
evaluates potential human health risks associated with HF-related releases that could affect
groundwater or surface water.
Using the HF fluid composition of  the "model HF fluid" used by NYSDEC, as well as HF flowback
fluid composition from the Marcellus Shale reported in the SGEIS for samples from
Pennsylvania/West Virginia, and published information on chemical toxicity, we examined
potential HF fluid release scenarios and their associated potential impacts on human health.
We focused on possible contamination of drinking water resources - in particular  either
groundwater aquifers or surface water bodies.

Exposure Analysis
We examined potential  contamination of drinking water resources and quantified risk to
human health for the following  scenarios:
 US EPA. 2004. "Evaluation of impacts to underground sources of drinking water by hydraulic fracturing of coalbed methane
reservoirs (Final)." Office of Water, June.
 New York State Department of Environmental Conservation (NYSDEC). 2009. "Draft Supplemental Generic Environmental
Impact Statement on the Oil, Gas and Solution Mining Regulatory Program—Well Permit Issuance for Horizontal Drilling and
High-Volume Hydraulic Fracturing to Develop the Marcellus Shale and Other Low-Permeability Gas Reservoirs." Division of
Mineral Resources, September.

-------
Groundwater
   •   Migration of HF fluid additives that remain in the Marcellus Shale after fracturing, up
       through the overlying shale and multiple bedrock layers to an overlying drinking water
       aquifer.

   •   Spills/releases of HF fluid to the surface during HF operations or from flowback
       management (i.e., pumping, handling), and subsequent migration to a nearby drinking
       water well.

Surface Water

   •   Spills/releases of HF fluid to the surface during HF operations or from flowback
       management at a surface impoundment, and migration to a nearby stream/river.

For each scenario, we adopted conservative (health-protective) assumptions that tend to
overstate, rather than  understate, the potential for human exposure via drinking water. For
example:

   •   We examined shallow drinking water aquifers with water tables (WTs) from ~15 to 30
       feet (ft) below the surface (5 to 10 meters).

   •   We examined scenarios for shallow drinking water wells, ranging from 65 to 165 ft deep
       (20 to 50 meters).

   •   We assigned a hydraulic conductivity to the drinking water aquifer that is lower (less
       mixing and dilution) than typical values for productive aquifers of the Southern Tier of
       New York State (NYS).

   •   We assumed no "setback" for the surface releases in our analysis of impacts to shallow
       drinking water wells or surface waters, whereas setback requirements are typically used
       for well siting purposes.

Overall, these assumptions are conservative (health-protective) and expected to yield an upper-
bound estimate of human health risks.

Toxicity Evaluation
We adopted established risk analysis methods to evaluate chemical toxicity and potential
human exposures. Agency-established toxicity criteria (e.g., drinking water  standards, or risk-
based benchmarks) were available for most of the model HF fluid and flowback constituents.
For HF additives lacking these "agency-established" health drinking water benchmarks or
toxicity factors, we developed risk-based concentrations (RBCs) for drinking water based on
published toxicity data in order to evaluate the health risks of the HF additives.

-------
Risk Analysis Results
None of the conservatively-modeled HF/flowback constituent concentrations in shallow
groundwater and surface water exceeded a risk-based concentration for drinking water.
Furthermore, our analysis confirms that migration of HF fluid additives from the Marcellus
Shale up through overlying bedrock to a surface aquifer is an implausible contamination
pathway. Even if such a pathway were plausible, the rate of migration would be such that the
dilution/attenuation of groundwater would be significant, thereby reducing the model HF fluid
additive concentrations in drinking water (from the overlying aquifer), to concentrations well
below health-based standards/benchmarks and not pose a threat to human health.

Conclusions
To summarize, our analysis indicates that even using conservative (health-protective) exposure
assumptions and a combination of agency-developed/Gradient-derived toxicity factors, the
potential human health risks associated with model HF fluid additives and measured flowback
constituents via drinking and household use of water are expected to be insignificant, as
defined by agency-based guidelines:

   •   The migration of HF additives from the Marcellus Shale formation to overlying potable
       aquifers is  implausible, given the thickness of the overlying confining rock layers and the
       effective hydraulic isolation that these overlying layers have provided for millions of
       years (resulting in trapping of the natural gas). Even using extreme case assumptions,
       the migration of HF additives from the Marcellus Shale to potable aquifers would not be
       sufficient to exceed health-based drinking water concentrations.

   •   Human health risks associated with surface releases of the model HF fluid additives are
       also expected to be insignificant due to attenuation mechanisms which are expected to
       reduce concentrations in potable aquifers and surface waters to levels well below
       health-based drinking water concentrations. In addition, typically used setback
       requirements and mitigation measures are expected to further protect these water
       resources and human health.

-------
      Revisiting the Major Discussion Points of the Technical
                            Presentation Sessions
The statements made during the workshop do not represent the views or opinions of EPA. The
claims made by participants have not been verified or endorsed by EPA. Any mention of trade
names or commercial products does not constitute endorsement or recommendation for use.

JP Nicot of the University of Texas at Austin, the workshop lead, described  the key points
covered by the technical presentations and discussion in the context of a risk assessment. The
first step of the assessment is to identify the potential sources of contamination: fracturing fluid
itself, flowback water, or produced water. The next step is to identify the situations under
which contamination could occur: a surface spill, a defect in the well bore, poor well
construction, or leakage through natural pathways. The third step is determining the
composition  of the fluids of concern, which could correspond to component(s) of the injected
fluid, compounds mobilized from the formation, or degradation products from reactions in the
subsurface. The  next steps are identifying pathways,  such as the fracture system itself and
abandoned wells, and determining the  fate of chemicals. Dr. Nicot identified monitoring as a
topic that was not covered in this session. He also emphasized the importance of developing a
conceptual model before moving forward with any type of mathematical modeling or
laboratory experiment. He added that modeling and  analysis tends to focus on the shallow
subsurface, bypassing the larger unknowns of fluid flow in deep aquifers and injection zones.

Angus McGrath  of Stantec, the Contaminant Identification, Transformation, and Transport
theme lead, summarized highlights of the Theme 1 presentations. He mentioned the four-
quadrant mobility vs. toxicity risk matrix and suggested it be expanded to cover more aspects of
fate and transport such as  persistence and degradation. Dr. McGrath also described the data
included in the presentations and the importance of determining the source of drinking water
or surface water contamination. He described the methods and forensic tools discussed  by the
presenters, as well as the idea of a conceptual model. He cautioned that conceptual models
should be flexible enough to allow for future developments. He emphasized the importance of
collecting background data and having  a tiered approach to monitoring. Dr.  McGrath also
described key fate and transport issues for future work, including: developing conceptual
models of pathways of fracture fluid release; evaluating chemicals for relative risk of release
and potential harm; evaluating geochemical processes and assessing relative risk; quantifying
chemicals in  potential sources; and designing monitoring approaches. He also provided
suggestions for the format of future workshops, recommending that summaries of previous
workshops be made available and that  direction be provided on the material discussed and the
priorities for each theme.

A participant added to Dr.  McGrath's list of key issues, noting that it is critical to distinguish
between shallow HF and HF at depth. CBM techniques use similar fracture fluids, and there is
more potential for contamination due to proximity to drinking water aquifers.

-------
Ahmad Ghassemi of Texas A&M University, the Impacts of Hydraulic Fracturing on Natural
Transport Systems theme lead, summarized the key points from Theme 2. First, he stated that
HF creates a large stimulated volume, which enhances the rock's permeability and enables
production. The resulting fracture system is often complex, involving the interaction of injection
pressures, in-situ stresses, and rock properties. Dr. Ghassemi noted that, despite this
complexity, diagnostic tools allow us to characterize the fracture geometry and  identify out-of-
zone fractures. Large height growth is possible, but existing stress gradients tend to limit it and
maintain the stimulated volume within the zone of interest. Dr. Ghassemi described how the
pressure regime of the stimulated reservoir volume favors the flow of gas into the well and is
not conducive to gas migration out of the stimulated zone. In addition, the  petrophysical
characteristics of shales favor retention of the injected water. He concluded that there is still a
need for analysis of the processes of fracture creation to improve the confidence of fracture
design.

Andrew Havics of pH2, LLC, the Models to Predict Transport theme lead, listed the key points
raised during the Theme 3 presentations. First, he  noted that models  have  limitations, and
validation is important both for screening and sensitivity purposes. He stated that model
selection should be based on parameter selection, as well as applicability to the region or site
and whether discrete or stochastic results are desirable. Data collection at the time of well
installation is critical, and data from actual spills and releases are very valuable to  modeling.
Confounding factors, such as those related to stray gas migration, should also be considered.
Mr. Havics noted that case studies tend to focus on extreme events and should  be considered
within the framework of the underlying distribution. He  also repeated the conclusions of Mr.
Sharma's presentation: for these conditions, the calculated results show very low  health risks
associated with the modeled scenarios. Mr. Havics noted that the DAFs presented by Mr.
Sharma are consistent with previous EPA modeling. He also emphasized that poor well
construction and abandonment present significant potential for problems.

Bob Puls of US EPA, the technical lead for the Hydraulic Fracturing Study, thanked  the leads,
presenters, and participants and assured all  of the attendees that  the information from the
workshop will be incorporated  into EPA's study. Dr.  Puls summarized  key  items from the
workshop. First,  he  noted the importance of identifying  which chemicals are most toxic, are
used most frequently and  at the  highest concentrations, and are the  most mobile; the
presenters emphasized that these characteristics  are most likely to have an impact on human
health.  Dr. Puls also noted that  some  presentations  indicated  that  potential chemicals of
concern occur both  in HF additives and in flowback/produced water according  to presenters.
He  also observed some  presenters believe  contaminants like methane, naturally occurring
radioactive material (NORM), and high TDS waters may pose more of a risk than the fracture
fluids themselves. Dr. Puls noted it is essential to monitor and understand how these fluids are
managed at the surface and ultimately disposed. He  also observed  that several  presenters
thought flowback characterization is useful, but it is important to determine the  sources of the
monitored fluid components. Dr. Puls  noted the  importance of investigating  existing data to
provide a realistic context for the case studies and the EPA study as a whole.  Dr. Puls also noted
that several presenters suggested that the likelihood of fluid transport up  from the injection

-------
zone into shallow aquifers seems low and that other pathways, such as abandoned wells, may
be linked to more significant risks. Dr.  Puls thanked all of the participants and organizers again
and concluded the workshop.

-------
                               Poster Abstracts

  Posters were submitted to US EPA for display during the technical workshops. Authors also
             submitted poster abstracts for use in this proceedings document.
Figures and Tables referred to by the authors in the poster abstracts are found in the respective
 posters. Poster figures and tables are not listed in the List of Figures (page ii) and List of Tables
                                      (page iii).

 The statements made during the workshop do not represent the views or opinions of EPA. The
 claims made by participants have not been verified or endorsed by EPA. Any mention of trade
  names or commercial products does not constitute endorsement or recommendation for us

-------
    Review of Groundwater Quality Data Surrounding Tracing
                                  Operations
                 Dollis M. Wright1 and Andrew A Havics2, CHMM, CIH, PE
                                       *QEPA
                                     2pH2, LLC

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.

 Figures and Tables referred to by the authors in the poster abstracts are found in the respective
 posters. Poster figures and tables are not listed in the List of Figures (page ii) and List of Tables
                                     (page Hi).
Introduction
With the increase in activity of the Oil and Gas Industry on the Western Slopes, comes a
heightened awareness of real and perceived health concerns attributed to this activity. A
common misperception is that the presence of a chemical equates to the chemical causing a
health impact. A review of existing data on domestic wells located within targeted Colorado
communities where fracing operations occur was performed by Quality Environmental
Professional Associates (QEPA). This activity was conducted as part of a larger study to address
data gaps regarding chemicals used in oil and gas activities and their risk to human health.1

Objective
A fate and transport evaluation of the constituents was performed for water as well other
primary media (soil, air). The water pathways were reviewed and analyzed using qualitative and
quantitative methods to determine if a completed pathway is possible or likely under
customary operating conditions. As part of the evaluation of these pathways, available
sampling from private/domestic wells and documented complaints in the targeted area were
reviewed.

Data Reviewed
In an effort to evaluate groundwater as a potential pathway of exposure, QEPA reviewed data
from the following sources:
   1.  Citizen Complaint data from the Garfield County Oil & Gas Office
   2.  Baseline and post drilling water well data within V* mile  radius of the drilling pad as
      provided by Operators in the targeted operations area from 2007-2008. (See figure 1)
   3.  Data from 495 permitted water wells (1963 - 2005) in a  2006 study.2

Results
There were a total of 271 complaints registered in the Garfield County Oil and Gas office by
citizens. The majority of the complaints focused on odor and water well issues (90%). There

                                         98

-------
were 64 complaints concerning water quality and impact to water wells. It is important to note
that none of the concerns were confirmed to be Oil & Gas Operations related, (again, see figure
1)

Part of the 2006 URS study objective was to address the vulnerability of surface water and
groundwater resources in the area to impacts from gas well development. A review of this data
showed  that Benzene (a primary contaminant of concern in the QEPA 2008 (Pathway Analysis
and Risk Assessment Reporti) was not detected in any of the sample results. Some additional
findings are that given natural attenuation, the fate of chemicals spilled or released in oil and
gas operations do not readily move very far from their point of origin. This is supported by four
separate studies covering 604 sites (FL, TX, CA, and a general US database) where
plume distances for known significant pure product releases (only diluted fluids and solids are
present  in the Oil and Gas reviewed) revealed that 75% are under 200 ft and most are shrinking
plumes.3

It is important to note that these wells are located in 110 square mile area where 978 gas wells
were drilled. The number of wells completed in each aquifer is shown below:
    o  Alluvial (Al)aquifer: 48 wells
    o  Wasatch (Wt) aquifer: 388 wells
    o  Both aquifers (Al and Wt): 5 wells
    o  Unidentified or incomplete logs: 9 wells

Operators in the targeted study area provided maps and sampling results of private well water
within a 1/z mile area of the Pad where samples were taken for the QEPA 2008 PARA (Pathway
Analysis and Risk Assessment). None of the private well  samples exceeded drinking water
standards. (See figures 2  & 3)

Conclusions
Based on a review of the  sampling data and record of complaints:
    o  Groundwater is not a past or present completed pathway of exposure to chemicals
      associated with oil and gas operations for citizens living within the targeted study area.
    o  In an area where there is considerable oil and gas drilling activity, Benzene was not
      detected in any private well water samples.
    o  Citizen concerns surrounding drinking water quality and oil and gas activities were not
      confirmed.

References
QEPA, Pathway Analysis and  Risk Assessment for Pit Solids, Fluids and Air in Colorado Energy
      Basins, June 2008.
URS Phase 1 Hydrogeologic Characterization of the Mamm Creek Field Area in Garfield County,
      March 23, 2006.
Newell,  Characteristics of Dissolved Hydrocarbon Plumes, Results from Four Studies, API
      Newsletter, December, 1998.
                                         99

-------
QH<
           Review of Groundwater Quality Data Surrounding  Fracing Operations
                                              Do/ItsM.  Wright, Andrew A.  Havrcs '
  Introduction
Data Reviewed
  A review of existing data on domestic wefts
  located within targeted Colorado communities
  where tracing operations occur was performed


  Objective	

  i*HMtn available data from targeted area on the
  impact of oil and gas operations on private wed
  water quality (Sato*/s a diagram snowing an
  example of a well pad (blue crcfe) and the
  location of sampled domestic welts (pmk circies)
  withma ',mle radius of the pad)

 •
•Baseline and post dolling water we! data
 were retrieved from Operators in the
 targeted operations area from 2007-2006
•CompUNnt togs of the targeted area County
 Health Department were reviewed (1998 -
 2008 soe graph to the nght for number and
 type of complaints by year)
•Data from 495 permitted water wets (1983-
 2005) ' These weds are located in 110 square
rrete area where 978 gas wets were drilled
                                               The above graph is charactetued data for one we*
                                               over a I 1/2 year penod
•
•
•



. .







.




1
^y «>
                                             Conclusions
                                             •Groundwater quality was second only to odor
                                              complaints m communities where there were
                                              (racing operations
                                             •Benzene was not detected m (he 495
                                              characterized domestic water we«s
                                             •None of the water samples taken from
                                              private wefts located within'.mite radius of
                                              the fraong operations exceeded dnnlong
                                                 rstandards
                                                                                             References
                                             1 URS Phase 1 Hydrogeotogic Charactenzaton
                                               of the Mamm Creek Field Area in GarfteW
                                               County  March 23 2006

                                             Acknowledgments	
                                             This data is part of (he COGA Pathways Analyst:
                                             and Risk Assessment for Solids and Fluids Used
                                             m OH and Gas Exploration and Production in
                                             Colorado QEPA 2008
                                                               100

-------
       Control Banding as a Means of Hazard Identification &
                      Characterization for Chemicals
                           Andrew A Havics, CHMM, CIH, PE
                                      pH2, LLC

 The statements made during the workshop do not represent the views or opinions of EPA. The
           claims made by participants have not been verified or endorsed by EPA.

 Figures and Tables referred to by the authors in the poster abstracts are found in the respective
 posters. Poster figures and tables are not listed in the List of Figures (page ii) and List of Tables
                                      (page Hi).
Control Banding (CB) or Performance-Based Occupational Exposure Limits (PB-OELs) have been
used intensively in the pharmaceutical and biological industries and are becoming more broadly
applied [Naumann, 1996, Nelson, 2007] Control banding can be thought of as a grouping or
banding strategy paralleling classical risk assessment components (see Figure 1 above). These
components include: Hazard Identification, Dose-Response Relationship, Exposure Assessment,
Risk Assessment, and Risk Management. The Hazard Identification and Dose-Response aspects
go hand in hand and form a beginning basis for determining inherent risk of a material by
reviewing and assessing toxicological parameters and then calculating an acceptable reference
value for dose or dose rate. For classical risk assessment, this value is a reference dose (RfD) for
non-carcinogens or a slope factor (SF) for carcinogens. For Control Banding, these two
components are merged into one step with compounds or materials classified by group, using
similar parameters but in a semi-quantitative manner.

One of these methods is the Control of Substances Hazardous to Health (COSHH) process
[Brooke, 1998]. For this method, Risk Phrases called R-phrases are used to establish a
chemical's band; these can be found on Material Safety Data Sheets (MSDS).  Example R-
Phrases and a subsequent grouping scheme by R-phrases is presented in Figure 2 [Brooke,
1998]. A modification of this scheme is provided in Table 1 below [Havics, 2008]. This modified
scheme was created by and partially validated using over 100 occupational limits and a
companion toxicological database [Havics, 2008]. If has been further modified here to include
two aspects: a) the addition of another 10 fold safety factor to adjust from occupational limits
to those for the general population  (this adjustment is consistent with findings of others
[Haber, 2007]), and b) conversion to oral dose rates (RfDs) from inhalation concentrations. This
second aspect results in an intermediate equivalent RfD conversion parameter, referred to here
as RfD*. The molecular weight (MW) of the agent must be multiplied by the intermediate RfD*
to provide a final RfD to account for the conversion from parts per million (ppm) to milligrams
per cubic meter (mg/m3) in the original scheme to the present application.  Table 2 provides the
ranges of RfDs and RFD*s for categorization using the modified process.

                                        101

-------
Although the scheme is based on ranges or bands, an estimated point value for calculation
purposes is also provided for each grouping. Because the groups are numerically arranged in
orders of magnitude (factors of 10), the point estimate represents a average logic value of the
range for that category. Examples of standard constituents and products from Colorado fracing
operations reviewed in 2008 [URS, 2008; QEPA, 2008] are shown in Table 3. The examples of
fracing products are then categorized as shown in Table 4 using this scheme. Not every R-
phrase will fit in the same category. As a result, the weighting of the two most conservative
categories are assessed before selecting the most appropriate one. In the case of cancer agents,
the most conservative is used.  For other aspects, the corresponding health  effects for the R-
phrases are reviewed along with the general amount of data available on the chemical and the
number of risk phrases in a category. If it is well-categorized, then a lesser category may be
selected, otherwise the more conservative category is chosen. Where a product has more than
one category for a component, the most conservative category is selected for that product.
Following this, intermediate RfD*s are then estimated followed by the RfD equivalents. These
are provided for the same examples as before in Table 5. Also, actual RfDs or Slope Factors
converted into equivalent RfDs are provided for comparison. The results suggest that the
process tends to be more conservative for known chemicals. Also, by default, poorly
characterized chemicals or products will have a tendency to be shifted to more conservative
values using this process.
References
1.      Brooke, A UK Scheme to Help Small Firms Control Health Risks from Chemicals,
       Toxicological Considerations, Ann Occ Hyg, 42, 6, 377-390, 1998.
2.      Haber, L, and A. Maier: Scientific Criteria Used for the Development of Occupational
       Exposure Limits for Metals and Other Mining-Related Chemicals.  Reg Toxicol
       Pharmacol. 36:262-279. 2002.
3.      Havics, A., Susan Ripple,  MS, CIH (DOW Chemical), Bob Sussman, PhD, DABT
       (Safebridge), Tracy Kimmel, PhD, DABT  (Safebridge), Bruce Naumann, PhD, DABT
       (Merck): A Comparison of the Hazard and Dose-Response Characterization of AIHA
       WEEL Chemicals Using 2 Different Control Banding Approaches"; RT254 Control
       Banding: Fundamentals,  Underlying Issues, Applications and Strategies for
       Implementation, at the American Industrial Hygiene Conference & Exposition (AIHCE),
       2008, May 31-June 5, 2008, Minneapolis, MN.
4.      Naumann, Bruce, Edward Sargent, Barry S. Starkman, William J. Fraser, Gail T. Becker,
       and G. David Kirk: Performance-Based Exposure Control Limits for Pharmaceutical Active
       Ingredients. AIHA J 57(l):33-42. 1996.
5.      Nelson, Deborah Imel, Stephen Chiusano, Anne Bracker, Lance Erickson, Charles Geraci,
       Martin Harper, Carolyn Harvey, Andrew Havics, Mark Hoover, Thomas Lentz, Richard
       Niemeier, Susan Ripple, Erica Stewart, Ernest Sullivan, and David Zalk: Guidance for
       Conducting Control Banding Analyses. AIHA, Fairfax, VA. 2007.
6.      National Research Council (NRC): Risk Assessment in the Federal Government: Managing
       the Process. NAP, Washington, DC. 189  pp. 1983.

                                        102

-------
National Research Council (NRC): Science and Judgment in Risk Assessment. NAP,
Washington, DC. 651 pp. 1994.
QEPA: Pathway Analysis and Risk assessment (PARA) For Solids and Fluids Used In Oil
and Gas Exploration and Production in Colorado, pp. 1-930, June 2008.
URS: Phase 1, Hydrogeologic Characterization of the Ma mm Creek Area In Garfield
County, March  23, 2006- for Board of County Commissioners Garfield County, Colorado,
2006.
                                  103

-------
              p
  H2.    LLC
Control   Banding  as  a  Means  of  Hazard  Identification
                         &  Characterization  for Chemicals
                          Andrew  Anthony "Tony"  Havics,  CHMM.  CIH,  PE
 II	"I llAMlllllKll Hi "I I'ril	AMI i- llA'.cil ui iii|iAliiulAl I M	If I in ill •. I I'll I )l I M 11 M >• lirrn H'.i-il Illl ri iMyrly
 III HIM Mli.lllilAi fiilli.il ,11 nl I ill tin,: I,.il iinlirli if. ,1,"I .111- |	mi' 	r Ill 11.ii II', .ii'i'llnl | M.VIIII.II 111  I'I'll,  MI-IMIM,
 2007|  Control binding c«n bi thought of « • grouping or binding ttritigv p« »"Hing tlank«l rink MUHmint
 IMItlMlillf III'. ('."(• hut il I- I  il M"."   I llf.r  i MI MM in ir.nl'. Illi lullr-  l.v  ,IM| 1'lrnl III, .ill MM |)	 llh'.pi'll'.r.
  '•I it HIM' Mil- I -i1	1 r A'.-1".	ill I'1!' I A•-•'••.'Ml	 ,VM I I11' I M.i 11.11.'>*n mil  I Mr 11 AMI 11 III rill It 11 Al It'll .11 |i I I)," r
 >'> • i	• .I-M, , I • (.><> 11,ml I In 11.11 nl .11 ii I MM Ml  I hi T Illi ill ir tl.l> I' I "I h'lt'i i nil ill II Mllii'Miiil I I*. I. Ml  ,» iii.ilni l.il liy
 IfVlnwII'K ,11111 ,l"rv Illf I ll.lt III mMl ,1! 11,II  linrlrl- .111, I Mini I .ill lll.llmi; .111 ,11 ' r(ll ,ll,lr I r-l r I r M I r- VAl'Irl'n llM-.r Ml
 tlir.r l Air   I Ml i  l.v.th ,il i l-.l ,!•• r< Miiriil  Mil-  V,I|M i-1'. A1 rln ri|i r '!"' f I (Ml) I I Ml null I AM lil'iurM-. m  .1 -Inpr IAI I ill
 I'.I I III I I.III lint!'	  I II, I  t nil, ill I',,11 n I n it  Hit- r I ,,.	11 MI ir nl   ii ,• in, i r,'M lulu M	lf|, i...	 ,,t,niii,l' Ml
 MUVlUI- < l.r • III I'll IIV IMI  M- nil; 'liMll.il |'.n .Hurl ri '.  lull III * -rill , |,I,IMMI .it In- MiAMMri  II in- M| I lir-.r nirtlniil'.
 • II,r I Ml III Ml Ml  '.Mil M	 I I,I,,mini I' IM I I ".ill  11 III J' .1 I I 11 |l M II '• •'.  111	1 .•  l')')H| I I M I III' I Mr I IIMI I I'M.I I'lllA'.r-
 l»lll»rt l< "I'MIM". ,1,'	-ft I I	.Mllll'h .11 llniilli ,il • h.llnl 11 ii" i' i  in  In' I tn in ll MM M.Vll.'ll '.AlHy llAl.l'.h"!-!'.
 I MM V.I I ..1lll|i|i" II I'hiAM", Alii I .l''lli-ri|,|i-nl KM" l Mil it Mlirnir liy l< M|,,AM-- I- i""''•n"-i! Ill I luulr /  [III mil r
 I'J'IHI  A Mlll'lllll All'	I '111- '.' l|r-MI'- r. ('I 'H'lilt-il IM Lllll'' I 11 rl" w  | I l,ll' 11 •  .'MnH| I I ir. 11II' I III Ini I M | ir, Mr I.I.M •
 i M>.it f,l liv .ii ill i >.n 11 .ill; \i ill'Lilril M'.iin' , H,,'  I MM in , MM,iMiMi.il III nil • .mil .1 i MMI|I.II mill Mi-1 , tl,i,'i ,il l,,v i il.il.l
 lllAvli '  .'III III I  II 11,1'. I	 I Mil IIMI  I	III [I'll hi'M' Ml Illi lull" I');M ,1'l'ri I'   II III'' .11 It III lull  I ll .ill ill I in III h'lll
 '.Airly IAI till In AI||M'.I Iliitll in I MpAlnniAl Ililill •• I M lhn>,r I in I Mr yniriAl innmlAlltMl Illn-. ,vl|ir.l mriil I', t "ii'.l-.ti-i ii
 WIMil/Millniti nl iitlli-n. Illabri /Illl/l)  .Mi'l ll) i iiiivnuliiiilii innlilinr iAlnt(|(|IM limn IllllAlmliill 11 Hi i rill ml liini.
 .11 I
           '1,1
 • i In-ill	ilir im "'MI ,i|i|'ln.illi>M  l.ilil'' / I'M
 modltlid (iinint AIMmiiili Ilir 11 Imiiw It U»i
 I nil pii'.i" I'. ,»|MI in MVI'Iril hn r,v ll I'IIMII'IMI.-  IU-, .IT.I- I Mr |-	|i- ,v " 1111111-11, .ill, ,11 I nifi! IM intlri • nl
 i unit nil ml i' 11.1	1  hii HIM (M tin! <".iirTiftt*i>pn>t»ntt ii jtvirigi Ingy, valuvof HiBiflngB tor that cat •gory
 I K,til It'll" Ml ' I ,1111 1,11,1 'MM'III I ir III' ,1111 I I	I, I I I ' I I Mill , III In, |l III I I .11 |l II' M|ll'l, II lull M'l  , h.-nir  l|"t ''VI y I' pln.l' r will III IM I In' • ,111	ll I'l'in y  A', ,1 I I" nil I Mr wi|-1 ll IM I "M I n' M,', M ,,,'>' I
 conwrvfttlvi uttyorlvk uni AUffcimvnt bvrom %•!•! (flgiln1 HUM.! .ICIH "in i.*l §oni  I" ''"'' •»•  of r»nn»r Agtntv
 till' 	I  , i.iri'i v,ll K"- I'  *,' '-'I I "I MI Mr I ,1'. |iril •.  Ihr i MI ,r'.(" I ||i III if hr;i|l || i-||r< h. I'll ll|r |' pin ,-i-.r-. A'l-
 invlpwml altinu; wllh ihi* Knnriiii aiiiiHiiii nt il«t» »v»IUtilf MM 11 in iliniili A! anil ihe munlir-i nf i  ^K phrA&n In A
 iair-ijiiiy  Ir It It wnil i Aif«"i l/»il, thin • liuir t«t»|ory rray bi ulictid, nthirwlM thimoti iiKiMivMlvn
 I.Oi'I'iliy  I- I I	  Wlll'li' I  |" I I'll II' ll,l' IM	 Ill,II	 I .ll"|',lll','lMI 1	|	I  Mil- 	I 	,,|llv»
 i.lli-l'iil',,  I' .rlrilr-illMi 11,.,I ,111.11,,, I  I ,illin,.;lM.- Mil'  Illi n iMr-ili.,1 r I'll )*'  nr lln-n I" I nii..lnl Mlltn.vril liy I III- I'll)
 li'"l Mil" ' link,ili'iii CM)' ,II"|MMHI|H| i,, 	i',,  	  lli>c»tultttiig|«lthAl lh> prniittlcntltto bl
 III'	II' I "I l',t! IV Mil I.MMWII 'I	II'  M' M ||V  Ml'1,1,111 I'lnMIV I M .11 ,11 t t-M ; I'l I I I	ll' Ml  |	IMI I' 'A' ,11 IMVr .1
 tindincy to bi ihlftid to rnor* conMrvntlvi v»lui> utlng thlt proun


 Figure 1.  NAS/NRC Risk Assessment/Management  Paradigm
                • -rw^iwiM
           NX-~-I—
m—^JL—m  »-jir •,.'»',•  I*IM
 I            AaMMIrtiil
                              nr ,., •      .,, ,i
                                                                        |w? l> ptiretritrom M£D5 ortlmlUrto twyfonre yrvOuctt nr ch«mlul%lntr>ftnttiin1 band I in
                                                                                      point III npotut* and nil cnrt> ti iiant Bandt jnr jwit » »gnri4> Hung wltK «»POI,JIT d«u ml
                                                                                           ,  ,...,. ^
                                                                                                     ....
                                                                                                                                  .
                                                                                                                                  f                .it
                                                                                                                                           • ' •«'
                                                                                                                                  • M    .   .     i >»x:' a MC«
                                                                                                                                          •
                                                                                                                                  !•>>>. i. •  i	'
                                                                                                                                  •             ....
                                                                                          •
                                                                                          I
                                                                                          • ••
                                                                                           i» (-.-..
                                                                                           II I«M rf •
                                                                                       Source:
                                                                                                   Table 1.  Modified COSHH Categories
A

ty*>TIUnU
•.i ,. '..-l
K5
148
HoR
Ml U>MI R
•i in in •
>«OtoSJO«
Vap.ii
fi
rlarmKjfan
Slty^V r"Jp'K.i|r9

M2U
rll'1
(12?

•l) 1 10 t
»Sto SO


TI»I«. r.orro.ive,
•ta

(U3.24,^
414 1!
M?
a-li 41
•am to u 1
>OSlo 8

fi
V»r>' T, • -
ntsri»»*»lon

UK. 27. X
modua
R4&51>?4.'J5
Rflu.fll.a?. «
•t^ni*
,,l,hlf .I'M
•USppm
V() llS
ac
• i- i .
nu.Rsa

•r-OtWf
<-*oa

                                                                                                    Rnplritory Saml«i«ri
                                                                                    Hal Sp« Jl«d but. rt to «a by A D Cjt«|ari

                                                                                         104
                                                                                                                                                               - Ngvlo. A., rt Jl, 3UUS.

-------
2.
                        Table 2.   RfD and  RfD*  Categories
       Est Equivalent
     General Population
         Value for
          Solids
            RfD
       Useful General
         Population
      RfD Point Value
       Est Equivalent
     General Population
          Value for
      Liquids/Vapors
           RfD*
       Where RfD =
       (RfD*) X (MW)
       Useful General
         Population
     RfD* Point Value
                             mg/kg
                              0.046
                           >0.03-0.3+
                             mg/kg
                              0.094
>0.014-0.140
    mg/kg
                                           0.0046
  >0.003-03
    mg/kg
                                           0.0094
>0.0014-
  0.014
  mg/kg
                                                          0.00046
>O.Q003-
  0.003
  mg/kg
                                                          0.0009
>0.00014-
  0.0014
  mo/kg
                                                                        0.000046
>0.00003-
  0.0003
  mg/kg
                                                                         0.00009
<=0.00014
  mo/kg
                                                                                      0.0000046
<=0.00003
  mg'kg
                                                                                      0.000009
           Table 3.  Example Fracing Materials from Colorado
Product Name
CL-37 Crosslinker
LGC-S
Rowback fluids
Produ ct Use
Drilling,
Fracturing
Drilling,
Completion,
Fracturing
Fracturing
Major Constituent
Glycerine
Propanol
Triethanolarnine zirconate
Guaf Gum
Benzene
Toluene
Elhylbenzene
Xylenes
CAS No.
56-81-5
71-23-8
101033^14-7
»uuu-^u-u
71-43-2
1 08-88-3
1 00-41-4
1330-20-7
                                                                  ase fiespanseChw*cteri» jn ,-nn y PJ: - t'dy:,xd :,-*-*-n t ftwry -  I't^TTw-n ^Ili^rn J . fraser, G»l T. flecker, «nd G. Cwd Kirk: PerfcrTTidnce 6«sed C^iostre Ccrrtral limits far fti .xm JT«J t -:^ Ac< ve
H^wierfts. ^AHWJ i7U)3342. !99S.
:-. Ndsaa Deborah rnd . Stephen Chius»rta Anns tirxter. Lance Cricteon, Charles Ger«i, Martin Harper, Carolyn Harvey, Andrew H-aic*. Marie Hoover, Thomas term (tidnard
Nienieier, Susan ftipple, Eric* Stawart Ernest SJIivjn,^nd fcawd Zalfc -^m3a/>cvjwCig3 fl/iaVifi ^AHA ftiA^ac, VJI. /'i-'i
& N*tiand  te«fx:hCai^dMNft>:i:«a*fl5*M/nMtMth?^                                 Ir
?! Watrana/ Ae-secvch CcxncW jtvACi' Sctenxani /lament wi /WiV flisps vnevit. NAP, Washinjtoa DC. 6Sl op . 1994
£. Q.£PA:Aat^war'flr*^^a^^^ai^isme^t//)fl^flJ/^^^a^>:AuA±; Uied"An O^ attf Stu fjvhmtfofi onSftrad
H u IL  (*h«e i H/drq^eola^ii: Oil«rch23,200^ -far ft
                                                                                                                                      Table 4.  Example Fracing Material Categorization
Agent
Critera



Benzene


Toluene


Ethylbenzene


Xylenes


Guar Gum
CL-37
Overall

Glycerine
Propanol
Triethanolarnine
Zirconate
A
R36
R48
NoR
All other ft

Rff
R38
R65
R11
R65
R67
R11
R36
R38
R10
R36
R38
•


No R Phrases
•
R11
R67
R36
R38
E
R20
R21
R22








R20
(")

R20
R22
•





•
(•)

C
R23, 24, 25
R34, 35
R37
R41.43
R4B/20/21/22
R36


R48/20
M

K,7
V

R37
KJ


•/•



R41
("t
R37
•
D
R26, 27, 28
R40 Care. Cat 3
R48/23/24/25
R60, 61, 62, 63

R48/23/24/25
M

R63
•S















t
R42
R45.49
R46, R68


R45
•


















                                                                                                                                                  Table  5.  Example RfD Equivalents
Agenl
Benzene
Toluene
Ethylbenzene
Xylenes
LC-S
Guar G urn
CL-37 Overall
Glycerine
Propanol
Triethanolarnine
Zirconate
Hazard Ran King
E
0
C
C
A
C
A
B
C
Eat WO'
u. OuOuutiy
000009
0 0006
0.0009


0.094
0.0094
0.00094
Est ttO Equivalent
u Ouu/
0008
01
0.1
0.046
056
8.7
o.se
o.es
Actual RfD
o.ijuu/
Modified from Slope
factor
02
0.1
2
NA
t\
-------
                                   Glossary of Terms
The sources of the definitions found in this glossary are noted at the end of each definition.
Sources include the following:

Abbreviated Source       Full Source Name
SPE                      Society of Petroleum Engineers Exploration & Production Glossary
                         (http://www.spe.org/glossary/wiki/doku.php/)
Schlumberger            Schlumberger Oilfield Glossary
                         (http://www.glossary.oilfield.slb.com/default.cfm)


ABBREVIATIONS
2-BE 2-butoxyethanol
ASTM American Society for Testing and Materials
ATSDR Agency for Toxic Substances and Disease Registry
BMDL benchmark dose level
BTEX benzene, toluene, ethylbenzene, and xylene
CAE chemicals for analytical evaluation
CAS Chemical Abstract Service
CB control banding
CBM coalbed methane
CCC criteria continuous concentration
CHEC Center for Health Environments and Communities
CMC criteria maximum concentration
COGA Colorado Oil and Gas Association
COGCC Colorado Oil and Gas Commerce Commission
COI compound of interest
COSHH Control of Substances Hazardous to Health
CSM conceptual site model
DBNPA 2,2-dibromo-3-nitrilopropionamide
DRO diesel range organics
EG ethylene glycol
EM electromagnetic
FRV failed rock volume
GWgroundwater
HF hydraulic fracturing
Kha hydraulic conductivity
Kow octanol-water partitioning coefficient
LOAEL lowest observed adverse effect level
LOD limit of detection
LOQ limit of quantification
MCL maximum contaminant level
MLE maximum likelihood estimator
MMCF million standard feet of gas
MRL minimum risk level
MRO motor oil range organics
MSDS Material Safety Data Sheets
MW molecular weight
ND not detected
NELAC National Environmental Laboratory Accreditation Conference
                                              106

-------
NFS numerical flow simulation
NOAEL no observed adverse effect level
NORM naturally occurring radioactive material
NPDES National Pollutant Discharge Elimination System
NYDEC New York Department of Environmental Conservation
NYS New York State
OSHA Occupational Safety and  Health Administration
PA DEP Pennsylvania Department of Environmental Protection
PAH polycyclic aromatic hydrocarbon
PARCC Precision, Accuracy, Representativeness, Completeness, and Comparability
PB-OEL performance-based occupational exposure limit
PBT Pennsylvania Brine Treatment
PCOC priority contaminants of concern
PDA production data analysis
POE point of exposure
POTW publicly owned treatment works
PTA pressure transient analysis
QEPA Quality Environmental Professional Associates
RA risk assessment
RAGS Risk Assessment Guide for Superfund
RBC risk-based concentrations
RBCA risk-based corrective action
RCI reactivity, corrosivity, and ignitability
RCRA Resource Conservation and Recovery Act
RfC reference concentration
RfD reference dose
SF slope factor
SGEIS supplemental generic environmental impact statement
SMCL secondary maximum contaminant level
SVOC semivolatile organic compounds
TAL target analyte list
TCLP toxic characteristic leaching procedure
IDS total dissolved solids
TEPH total extractable petroleum hydrocarbons
TIC tentatively identified compounds
TMB trimethyl benzene
USBM United States Bureau of Mines
USGS US Geological Survey
LIST underground storage tank
VOC volatile organic compounds
WHO World Health Organization
WT water table
GLOSSARY

Amott imbibition test a comparison of the imbibition potential of water and oil into a rock. It is possible for the
   same rock to imbibe both water and oil, with water imbibing at low in situ water saturation, displacing excess
   oil from the surface of the rock grains, and oil imbibing at low in-situ oil saturation, displacing excess water.
   (Schlumberger)
connate brine the natural brine occupying the pore spaces. Usually this water is at equilibrium with the minerals in
   the formation. (SPE)
                                                 107

-------
fall-off test The measurement and analysis of pressure data taken after an injection well is shut in. These data are
   often the easiest transient well-test data to obtain. Wellhead pressure rises during injection, and if the well
   remains full of liquid after shut-in of an injector, the pressure can be measured at the surface, and bottomhole
   pressures can be calculated by adding the pressure from the hydrostatic column to the wellhead pressure. Since
   most water-injection wells are fractured during injection, and injection wells often go on vacuum, the fluid level
   can fall below the surface. Dealing with this complication requires reverting to bottomhole pressure gauges or
   sonic devices. (Schlumberger)
filter-cake 1. the layer of solids stranded on the face of permeable formations by liquids driven into the rock by
   pressure differential towards the formation. When sized correctly the filter cake may completely stop losses.
   (SPE)
   2. The residue deposited on a permeable medium when a slurry, such as a drilling fluid, is forced against the
   medium under a pressure. Filtrate is the liquid that passes through the medium, leaving the cake on the
   medium. Drilling muds are tested to determine filtration rate and filter-cake properties. Cake properties such as
   cake thickness, toughness, slickness and permeability are important because the cake that forms on permeable
   zones in the wellbore can cause stuck pipe and other drilling problems. Reduced oil and gas production  can
   result from reservoir damage when a poor filter cake allows deep filtrate invasion. A certain degree of cake
   buildup is desirable to isolate formations from drilling fluids. In openhole completions in high-angle or
   horizontal holes, the formation of an external filter cake is preferable to a cake that forms partly inside the
   formation. The latter has a higher potential for formation damage. (Schlumberger)
flowback The process of allowing fluids to flow from the well following a treatment, either in preparation for a
   subsequent phase of treatment or in preparation for cleanup and returning the well to production.
   (Schlumberger)
imbibition  absorption and adsorption of fluids into the pores of the rock. (SPE)
kerogen An initial stage of oil that never developed  completely into crude. Typical of oil shales. (SPE)
leakoff The magnitude of pressure exerted on a formation that causes fluid to be forced into the formation. The
   fluid may be flowing into the pore spaces of the rock or into cracks opened and propagated into the formation
   by the fluid pressure. (Schlumberger)
modulus of elasticity Modulus refers to stress at a predetermined level of elongation, usually at 100% elongation.
   The higher the modulus of a compound, the more apt it is to recover from loading or localized force and the
   better is its resistance to extrusion. (SPE)
pressure transient analysis The analysis of pressure changes over time, especially those associated with small
   variations in the volume of fluid. In most well tests, a limited amount of fluid is allowed to flow from the
   formation being tested and the pressure at the formation monitored over time. Then, the well is closed and the
   pressure monitored while the fluid within the formation equilibrates. The analysis of these pressure changes
   can provide information on the size and shape of the formation as well as its ability to produce fluids.
   (Schlumberger)
shut-in stop a well from flowing and close the valves. (SPE)
slickwater a water base fluid with only a very small amount of a polymer added to give friction reduction benefit.
   (SPE)
spurt loss the initial loss of fluids from a mud or frac fluid,  before the walk cake can be formed. (SPE)
                                                  108

-------
    Printed with vegetable-based ink on paper that
    contains a minimum of 50% post-consumer fiber
    content and processed chlorine free.

United States
Environmental Protection
Agency
PRESORTED STANDARD
 POSTAGES FEES PAID
          EPA
   PERMIT NO. G-35
Office of Research and Development (8101R)
Washington, DC 20460

Official Business
Penalty for Private Use
$300

-------