&EPA
/430 | May 2015 | www.epa.gov
United States
Environmental Prot
Agency
Sources Contributing Inorganh
Species to Drinking Water Intake
During Low Flow Conditions on the
Allegheny River in Western
Pennsylvania
United States Environmental Protection Ag
Office of Research and Development
-------
-------
United States
Environmental Protection
Agency
Sources Contributing Inorganic Species to Drinking Water Intakes during
Low Flow Conditions on the Allegheny River in Western Pennsylvania
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC
May 2015
EPA/600/R-14/430
-------
DISCLAIMER
The U.S. Environmental Protection Agency (EPA), through its Office of Research and Development
(ORD), funded and conducted this research. The views expressed in this report are those of the authors
and do not necessarily reflect the views or policies of EPA. It has been subjected to Agency review and
approved for publication. Mention of trade names or commercial products does not constitute an
endorsement or recommendation for use.
-------
TABLE OF CONTENTS
DISCLAIMER ii
TABLE OF CONTENTS iii
LIST OF TABLES iv
LIST OF FIGURES vi
PREFACE vii
ABBREVIATIONS viii
AUTHORS x
ACKNOWLEDGEMENTS xi
E. EXECUTIVE SUMMARY 1
E.I Background 1
E.2 Objectives 1
E.3 Approach 1
E.4 Results 2
E.5 Conclusions 2
E.6 Limitations 3
1. INTRODUCTION AND PURPOSE OF STUDY 5
1.1 Need for Research 7
1.2 Research Objectives 8
1.3 Source Apportionment and River Transport Modeling 8
2. METHODS 10
2.1 Sampling Domain 10
2.2 River Sampling and Discharge Source Location 11
2.3 Source Sample Collection 18
2.4 Sample Analysis 19
2.5 PMF Receptor Model 22
2.6 Statistical Analysis 23
3. QUALITY ASSURANCE AND QUALITY CONTROL 24
3.1 Quality Control Results for River and Source Sample Analysis 24
3.2 Quality Control Results for PMF Analysis 32
3.3 Quality Assurance Assessments 32
4. RESULTS AND DISCUSSION 34
4.1 River Sample Composition 34
4.2 Source Profiles 39
4.3 PMF Receptor Modeling Results 41
4.4 PMF Bromide Sensitivity and Hybrid Analysis 56
5. Summary and Conclusions 60
5.1 River Measurements 60
5.2 Measured Source Profiles 61
5.3 PMF Source Apportionment Results 61
5.4 Application of Source Apportionment Modeling to Surface Water 63
6. REFERENCES 65
APPENDIX A A-l
APPENDIX B B-l
-------
LIST OF TABLES
Table 1. Waste water treatment facility and receiving stream information 12
Table 2. Ion chromatography method detection limits (MDL) 20
Table 3. Inductively coupled plasma-optical emission spectrometry method detection limits (MDL) 21
Table 4. High-resolution-inductively coupled plasma mass spectrometry method detection limits
(MDL) 21
Table 5. Ion chromatography recovery of performance evaluation samples 25
Table 6. Inductively coupled plasma-optical emission spectrometry recovery (R) of certified (Cert)
materials, reference materials, and sequence calibration accuracy checks 26
Table 7. High-resolution-inductively coupled plasma mass spectrometry recovery (R) of certified
reference materials 28
Table 8. Analytical precision for laboratory duplicates 29
Table 9. Average sequential sampling precision (%) 30
Table 10. Summary of river and PDWS sampling site field blanks (mg/L) 31
Table 11. Summary of CWTF sampling site field blanks (mg/L) 31
Table 12. Downstream (S03, S04, 805) and upstream (SOI) concentrations (mg/L) 35
Table 13. Average Br/Cl ratio and key species concentrations (mg/L) for measured source profiles 38
Table 14. PDWS intake low flow intake concentrations (mg/L) 39
Table 15. CWTF wastewater volumes treated in 2012 (barrels) 40
Table 16. Positive Matrix Factorization measured versus predicted regression statistics 44
Table 17. Positive Matrix Factorization high-bromide source type profiles and minimum and maximum
estimates from the displacement algorithm (mg/L) 45
Table 18. Positive Matrix Factorization low-bromide source type profiles and minimum and maximum
estimates from the displacement algorithm (mg/L) 45
Table 19. Positive Matrix Factorization (PMF) bromide concentrations by PMF analysis and sampling
site (mg/L) 53
Table 20. Positive Matrix Factorization (PMF) chloride concentrations by PMF analysis and sampling site
(mg/L) 54
Table 21. Positive Matrix Factorization (PMF) nitrate concentrations by PMF analysis and sampling site
(mg/L) 55
Table 22. Positive Matrix Factorization (PMF) sulfate concentrations by PMF analysis and sampling site
(mg/L) 56
Table 23. PMF sensitivity analysis results (mg/L) 57
IV
-------
Table Al. Summary of Allegheny River species concentrations (mg/L); n (%) number of
valid samples A-2
Table. Al (continued ) Summary of Allegheny River species concentrations (mg/L); n (%) number of
valid samples A-3
Table A2. Summary of Blacklick Creek species concentrations (mg/L); n (%) number of
valid samples A-4
Table A2. (continued) Summary of Blacklick Creek species concentrations (mg/L); n (%) number of
valid samples A-5
Table Bl. Allegheny PMF analysis parameters.1 B-l
Table B2. Blacklick PMF analysis parameters B-2
Table B3. Combined Allegheny and Blacklick PMF analysis parameters B-2
-------
LIST OF FIGURES
Figure 1. Fundamental research questions posed for each identified stage 6
Figure 2. Schematic representation of commercial wastewater treatment, discharge to surface waters, and
impact on downstream public drinking water plants (focus area of this project is circled in
red) 7
Figure 3. Shale gas plays in the contiguous U.S 10
Figure 4. Allegheny River (blue) and Blacklick Creek to Kiskiminetas River (red) sampling sites, and
location of major surface water discharges. The arrows show the direction of river flow 14
Figure 5. USGS Gauge Station Data for 2012 showing the lower daily mean discharge levels during
summer and fall 15
Figure 6. USGS mean discharge rates and gage IDs 17
Figure 7. Diagram of sampling sites, rivers, and sources 18
Figure 8. EPA PMF observed (blue) vs. predicted (red) time series plot for bromide by site (mg/L) 43
Figure 9. EPA PMF source profile plot showing the species concentration (left axis) and the percentage of
species associated with each source (right axis) 47
Figure 10. EPA PMF time series plot of normalized factor contributions (average contribution = 1 for
each factor) by site and date 48
Figure 11. EPA PMF source contribution plot for the 3 sources with elevated bromide concentrations.. 49
Figure 12. Comparison of the measured (_M) and Positive Matrix Factorization (_P) profiles 50
Figure 13. Positive Matrix Factorization (PMF) combined analysis (Allegheny and Blacklick) bromide
concentrations (mg/L) for PMF source types by river sampling site 51
Figure 14. Distribution of CWTF and FGD + AMD bromide source contributions (%) to S05_B 52
Figure 15. Subtraction of AMD bromide from the FGD + AMD source to provide AMD and FGD
estimated source contributions 58
Figure 16. Hybrid PMF source contributions of bromide by sampling site and facility discharges 59
Figure 17. Summary of Br/Cl and bromide mean concentrations for measured sources (Table 13) 62
Figure 18. Summary of median PMF bromide source contributions for PDWS intakes 63
Figure Al. Diagram showing receptor and transport model input requirements and model outputs Al
VI
-------
PREFACE
The U.S. Environmental Protection Agency (EPA) is conducting a study of the potential impacts of
hydraulic fracturing for oil and gas on drinking water resources. This study was initiated in Fiscal Year
2010 when Congress urged the EPA to examine the relationship between hydraulic fracturing and
drinking water resources in the United States. In response, EPA developed a research plan (Plan to Study
the Potential Impacts of Hydraulic Fracturing on Drinking Water Resources} that was reviewed by the
Agency's Science Advisory Board (SAB) and issued in 2011. A progress report on the study (Study of
the Potential Impacts of Hydraulic Fracturing on Drinking Water Resources: Progress Report), detailing
the EPA's research approaches and next steps, was released in late 2012 and was followed by a
consultation with individual experts convened under the auspices of the SAB.
The EPA's study includes the development of several research projects, extensive review of the literature
and technical input from state, industry, and non-governmental organizations as well as the public and
other stakeholders. A series of technical roundtables and in-depth technical workshops were held to help
address specific research questions and to inform the work of the study. The study is designed to address
research questions posed for each stage of the hydraulic fracturing water cycle:
Water Acquisition: What are the possible impacts of large volume water withdrawals
from ground and surface waters on drinking water resources?
Chemical Mixing: What are the possible impacts of surface spills of hydraulic fracturing fluid
on or near well pads on drinking water resources?
Well Injection: What are the possible impacts of the injection and fracturing process on
drinking water resources?
Flowback and Produced Water: What are the possible impacts of surface spills of flowback
and produced water on or near well pads on drinking water resources?
Wastewater Treatment and Waste Disposal: What are the possible impacts of inadequate
treatment of hydraulic fracturing wastewaters on drinking water resources?
This report, Sources Contributing Inorganic Species to Drinking Water Intakes during
Low Flow Conditions on the Allegheny River in Western Pennsylvania, is the product of one of the
research projects conducted as part of the EPA's study. It has undergone independent, external peer
review in accordance with Agency policy and all of the peer review comments received were considered
in the report's development.
The EPA's study will contribute to the understanding of the potential impacts of hydraulic fracturing
activities for oil and gas on drinking water resources and the factors that may influence those impacts.
The study will help facilitate and inform dialogue among interested stakeholders, including Congress,
other Federal agencies, states, tribal government, the international community, industry, non-
governmental organizations, academia, and the general public.
VII
-------
ADQ
AMD
CBM
cfs
CWTF
EST
EPA
FGD
GPD
HF
HR-ICPMS
1C
ICP-OES
km
MDL
MOD
mg/L
mL
MPS
MW
NPDES
PDWS
ABBREVIATIONS
audit of data quality
acid mine drainage
coal bed methane
cubic feet per second
centralized waste treatment facility
Eastern Standard Time
(U.S.) Environmental Protection Agency
flue gas desulfurization
gallons per day
hydraulic fracturing
high resolution magnetic sector field - inductively coupled plasma mass
spectrometry
ion chromatography
inductively coupled plasma - optical emission spectrometry
Kilometers
method detection limits
million gallons per day
milligrams per liter
Milliliters
multi-probe system
Megawatt
National Pollutant Discharge Elimination System
public drinking water system
VIM
-------
PDW
PMF
POTW
QA
QAPP
QC
TDS
TSA
USAGE
USGS
v/v
w/v
uS/cm
ug/L
public drinking water
Positive Matrix Factorization
publicly owned treatment works
quality assurance
Quality Assurance Project Plan
quality control
total dissolved solids
technical systems audit
U.S. Army Corps of Engineers
U.S. Geological Survey
volume per volume
weight per volume
microsiemens per centimeter
micrograms per liter
IX
-------
AUTHORS
Gary A. Norris,1 Kasey D. Kovalcik,1 Matthew S. Landis,1 Amy Bergdale,2 Carry Croghan,1
Ali S. KamaT
1 U.S. EPA Office of Research and Development, Research Triangle Park, NC 27711
2 U.S. EPA Region 3, Office of Monitoring and Assessment, Wheeling, WV 26003
+ Current Affiliation: Oak Ridge Institute for Science and Education, Oak Ridge, TN 37831-0117
-------
ACKNOWLEDGEMENTS
We thank the facilities for allowing site access for the collection of outfall samples for this research; the
Pennsylvania Department of Environmental Protection's Northwest and Southwest offices for providing
access to NPDES permit records; and EPA Region 3 for assistance in selection of study domains and
providing source information.
XI
-------
E. EXECUTIVE SUMMARY
E.I Background
This study is a component of the U.S. Environmental Protection Agency's (EPA) study of hydraulic
fracturing for oil and gas and its potential impact on drinking water resources, and addresses the research
question, "Wastewater Treatment and Waste Disposal: What are the possible impacts of inadequate
treatment of hydraulic fracturing wastewater on drinking water resources?" (U.S. EPA 2014b). The
Allegheny River and its tributaries in western Pennsylvania are affected by many different types of
contaminant sources including centralized waste treatment facilities for oil and gas wastewater, coal-fired
electric power generating stations, acid mine drainage from historic mining wastes, current mining
operations, natural oil seepage, industrial manufacturing facilities, publicly owned treatment plants that
treat municipal sewage, and industrial facility sewage treatment plants. These sources discharge a
mixture of contaminants into surface waters, some examples of which are the anions bromide, chloride,
sulfate, and nitrate. The Allegheny River is the source of raw water for thirteen (13) public drinking
water systems serving over half a million people in western Pennsylvania. Understanding sources of
contaminants in drinking water is critical due to their potential impacts on drinking water quality (States
etal.2013).
Centralized waste treatment facilities for oil and gas wastewater that discharged treated wastewater to
surface wasters in western Pennsylvania during this study primarily but not exclusively treat conventional
oil and gas wastewater, and most conventional wells in Pennsylvania are stimulated or hydraulically
fractured (PA DEP 2012). The centralized waste treatment facilities for oil and gas wastewater have
treatment processes to remove solids, but do not effectively remove bromide and chloride (Ferrar et al.
2013). Discharged bromide can lead to increased levels of brominated disinfection byproducts in
downstream drinking water treatment plants (Richardson et al. 2007, States et al. 2013, Parker et al.
2014) which may pose human health risks. Discharge of treated (centralized waste treatment facilities for
oil and gas wastewater) and untreated (via publicly owned treatment plants) wastewater from oil and gas
production in the Marcellus region was substantially reduced in May 2011 due to a request by the
Commonwealth of Pennsylvania asking companies to voluntarily stop sending their Marcellus wastewater
to these facilities (PADEP 201 la, 201 Ib; Wilson and Van Briesen 2012). The request was based on
concerns over increased bromide levels at public drinking water system intakes and associated increases
in disinfection byproducts within public drinking water system finished water in Pittsburgh (PADEP
201 la). Previous studies have focused on characterizing centralized waste treatment facilities for oil and
gas wastewater discharges since they are known sources of bromide (Ferrar et al. 2013). However, these
studies did not consider other critical sources of contaminants for the Allegheny River; or assess the
impact of the discharges on downstream public drinking water systems.
E.2 Objectives
The objectives of this study were to quantify the cumulative contribution of treated oil and gas wastewater
from centralized waste treatment facilities for oil and gas wastewater that primarily treat hydraulic
fracturing wastewater, and to distinguish that contribution from other potential sources on bromide
concentrations at two public drinking water system intakes located on the Allegheny River.
E.3 Approach
Centralized waste treatment facilities for oil and gas wastewater discharges are a known major source of
chloride, bromide, and other anions. The contribution of these contaminants from centralized waste
-------
treatment facilities for oil and gas wastewater as well as other sources to public drinking water system
intakes was evaluated by taking the following approach: (1) development of chemical source profiles, or
fingerprints, for all sources upstream of two public drinking water system intakes on the Allegheny River
by collecting outfall samples from specific sources within facilities and combined river outfalls, (2)
collection and chemical characterization of river samples from multiple sites upstream and downstream of
centralized waste treatment facilities for oil and gas wastewater, electric generating stations, industrial
facilities, and at the public drinking water system intakes, and (3) analysis of the river sample data with
the EPA Positive Matrix Factorization receptor model to quantify the contribution of sources to anion
levels at the public drinking water system intakes.
E.4 Results
Daily samples were collected from six river sampling sites, two public drinking water system intakes, and
the discharge tanks at two centralized waste treatment facilities for oil and gas wastewater for two weeks
in summer and fall 2012 during low river flow conditions. Chemical species profiles were collected for
centralized waste treatment facilities for oil and gas wastewater, coal fired power plants (cooling tower,
flue gas desulfurization scrubber, demineralizer, coal pile runoff, coal ash), industrial manufacturing
processes, publicly owned treatment plants, coal bed methane, acid mine drainage, oil seep, and coal mine
runoff. Based on the Positive Matrix Factorization multiple sampling site analysis, the predominant
sources of bromide at the public drinking water system intakes were treated wastewater discharged from
centralized waste treatment facilities for oil and gas wastewater and flue gas desulfurization, while
publicly owned treatment plants and acid mine drainage were sources of nitrate and sulfate.
E.5 Conclusions
This research applied a
technique referred to as
"source apportionment" to
quantify source contributions
for a number of common
discharge sources. Source
measurements were collected
to provide reference
information for apportioning
contaminant sources in the
Alleghany watershed,
including from centralized
wastewater treatment
facilities that treat wastes
including oil and gas
wastewater; coal-fired power
plants with and without flue
gas desulfurization; industrial
manufacturing facilities;
municipal and industrial
wastewater treatment plants; active coal mine runoff; and acid mine drainage. During the study period,
which was focused on low flow conditions, we found that centralized wastewater treatment facilities and
KEY FINDINGS
The results demonstrate that the 2 public drinking water
intakes studied are impacted by multiple sources
contributing various inorganic species, including centralized
wastewater treatment facilities, power generating stations,
and acid mine drainage.
Source measurements provide a signature or profile for
numerous bromide sources.
The predominate sources of bromide at the 2 public drinking
water intakes studied were wastewaters discharged from
including centralized wastewater treatment facilities and
coal-fired power plants with flue gas desulfurization .
CWTFs contributed nearly all the bromide at 1 intake, while
both centralized wastewater treatment facilities and coal
fired power plants with flue gas desulfurization contributed
to bromide levels at the second intake.
-------
coal fired power plants with flue gas desulfurization are contributing bromide to two Allegheny River
public drinking water system intakes. Acid mine drainage from historical mining activities also
contributes bromide, however, the magnitude of the contribution was 9% at one of the intakes. This study
collected a large amount of data and had five key findings:
Source measurements demonstrated a range of Br/Cl ratios for bromide sources, suggesting the
ratio can assist in differentiating the contributions from these sources.
Centralized waste treatment facilities for oil and gas wastewater, which are known to treat
hydraulic fracturing wastewaters, are a major source of bromide at the two public drinking water
system intakes in this study with a contribution of 89% and 37%, respectively.
Flue gas desulfurization wastewater is another source of bromide at public drinking water system
intakes. Flue gas desulfurization median percent contributions ranged from 50 to 59% at one of
the public drinking water system intakes, which varies daily due to changes in discharges. The
coal-fired electrical generating stations in this study domain burned upper Pennsylvanian and
Monongahela formation bituminous coal (e.g., Pittsburgh #8), which contains naturally high
levels of bromine.
The combination of bromide transported from centralized waste treatment facilities for oil and
gas wastewater, FGDs, and acid mine drainage explains 88-89% of the bromide at one of the
intakes, and 96% of the bromide at a second intake.
This research study demonstrates the efficacy of source apportionment techniques to quantify
contaminant impacts in complex river systems with multiple source discharges.
Understanding the sources will guide efforts to control exposures to drinking water contaminants
of concern such as brominated disinfection byproducts.
E.6 Limitations
There are important limitations and uncertainties in the information included in this report:
In Pennsylvania, hydraulic fracturing is commonly used in conventional and unconventional oil
and gas production using both vertical and horizontal wells. Although most of the wastewater
from oil and gas operations in western Pennsylvania is associated with hydraulic fracturing, it
was not possible to determine with certainty the exact mix of hydraulic fracturing and non-
hydraulic fracturing oil and gas wastewater treated by the commercial wastewater treatment
facilities during each sampling event since reporting on accepted waste streams is submitted on
an annual basis rather than for each daily delivery received by the plant. During the study period,
both centralized waste treatment facilities for oil and gas wastewater accumulated the oil and gas
wastewater from individual deliveries into large on-site storage tanks prior to batch treatment. As
a result, the wastewater from numerous individual wells was combined prior to treatment and
discharge.
Samples were collected from one large river system and one small river system, with different
source contributions, chemistry, and flow rates. Data from these two river systems were
combined in a Positive Matrix Factorization receptor modeling analysis. The ability of the
Positive Matrix Factorization model to resolve flue gas desulfurization as a source varies
depending on the sampling sites included in the analysis. The results reported above reflect the
combined analysis of both river system contributions. A sensitivity analysis provides results from
alternate site inclusion (see section 4.4, Table 23). Regardless of the site combinations, Positive
-------
Matrix Factorization was able to distinguish a centralized waste treatment facility source, either
as a single source (subset of sites) or a combined flue gas desulfurization and centralized waste
treatment facility source (all sites).
Ground water contributions were not identified as a significant source of bromide in the source
apportionment analysis. However, any hydrologic contribution from ground water may be
accounted for in one of the background sources (e.g., acid mine drainage, suspended sediments).
This report is based on two sampling campaigns in summer and fall 2012 and is not intended to
quantify bromide source contributions from all centralized waste treatment facilities or other
sources on all the public drinking water system intakes along the Allegheny River during other
time periods.
-------
1. INTRODUCTION AND PURPOSE OF STUDY
Hydraulic fracturing is an important means of accessing one of the nation's most vital energy resources,
oil and natural gas. Advances in technology, along with economic and energy policy developments, have
spurred a dramatic growth in the use of hydraulic fracturing across a wide range of geographic regions
and geologic formations in the United States for both oil and gas production. As the use of hydraulic
fracturing has increased, so have interests about its potential impact on human health and the
environment, including possible effects on drinking water resources. Based on the increasing prevalence
of hydraulic fracturing in facilitating increased domestic production of natural gas, the U.S.
Environmental Protection Agency's (EPA) study of hydraulic fracturing for oil and gas and its potential
impact on drinking water resources emphasizes the impact of hydraulic fracturing operations in shale
formations containing natural gas. Portions of the research, however, may provide information on
hydraulic fracturing in other types of oil and gas reservoirs, such as coal beds and tight sands.
The purpose of the EPA's study is to assess the potential impacts of hydraulic fracturing on drinking
water resources, if any, and to identify the driving factors that may affect the severity and frequency of
such impacts. To answer these questions, EPA identified a set of research activities associated with each
stage of the hydraulic fracturing water lifecycle (Figure 1), from water acquisition through the mixing of
chemicals and actual fracturing to post-fracturing production, including the management of hydraulic
fracturing wastewaters (commonly referred to as "flowback" and "produced water") and ultimate
treatment and disposal. This report focuses on research activities developed to investigate the last stage
of water use in hydraulic fracturing operations (Wastewater Treatment and Waste Disposal) and the
associated research question: What are the possible impacts of inadequate treatment of hydraulic
fracturing wastewaters on drinking water resources (Figure 1).
In 2012, EPA identified two centralized waste treatment facilities for oil and gas wastewater (CWTFs)
that were treating and discharging hydraulic fracturing wastewaters into the Allegheny River or its
tributaries in Western Pennsylvania. EPA then conducted three (3) seasonal two-week sampling
campaigns to collect CWTF discharge, river water, and downstream public drinking water system
(PDWS) raw water intakes. The goal of this effort was to identify if hydraulic fracturing wastewaters
were contributing to anion concentrations (e.g., bromide) at PDWS intakes. EPA then undertook a source
apportionment modeling study to quantitatively determine the impact of CWTF discharges on
contaminants measured at downstream PDWS (Figure 2) during low flow river discharge periods
(summer and fall 2012). This analysis improved EPA's understanding of how contaminants in the treated
hydraulic fracturing effluent disperse when discharged to surface waters and impact downstream PDWS.
EPA also assessed how other sources of contamination (e.g., coal-fired power plants, acid mine drainage)
impact contaminant concentrations in the river. In addition, the sampling, analysis, and source
apportionment modeling developed and presented in this report provide an approach that can be applied
-------
to other river systems. Ultimately, the results of this study and the future application of the research tools
presented will provide communities, states, tribes, and industry with sound scientific knowledge on
understanding potential impacts of hydraulic fracturing on drinking water resources, and the protection of
those resources for the future.
Fracturing Operations
Water Acquisition
Chemical Mixini
Well Injection
Flowback and
Produced Water
Wastewater Treatment
and Waste Disposal
Fundamental Research Question
What are the potential impacts of large volume water withdrawals
from ground and surface waters on drinking water resources?
What are the possible impacts of surface spills on or near well pads
of hydraulic fracturing fluids on drinking water resources?
What are the possible impacts of the injection and fracturing process
on drinking water resources?
What are the possible impacts of surface spills on or near well pads of
flowback and produced water on drinking water resources?
What are the possible impacts of inadequate treatment of hydraulic
fracturing wastewaters on drinking water resources?
Figure 1. Fundamental research questions posed for each identified stage.
-------
Wastewater
reatment Plant
Drinking
ater Treatment
lant ^d
Figure 2. Schematic representation of commercial wastewater treatment, discharge to surface waters, and
impact on downstream public drinking water plants (focus area of this project is circled in red).
1.1 Need for Research
Oil and gas wastewater may contain inorganic salts, radioactive substances, heavy metals, and volatile
organic substances originating from the producing formation (Vengosh etal., 2014; Balaba and Smart,
2012). Oil and gas CWTFs, including those that treat hydraulically fractured wastewater, have processes
to remove solids but do not effectively remove monovalent ions such as bromide and chloride (Ferrar et
al. 2013), which pass through the treatment process and can be discharged into surface waters.
Previously published studies have focused on characterizing a single CWTF discharge, since these
discharges are known sources of bromide (Ferrar et al. 2013; Hladik et al. 2014; Warner et al. 2013).
However, these approaches have been limited to investigating near field downstream enhancement, did
not distinguish the contribution of CWTFs relative to other sources of the same contaminant(s), and did
not directly evaluate the impact of the discharges on downstream PDWS.
Discharge of treated (CWTF) and untreated (via publicly owned treatment plants (POTWs)) wastewater
from oil and gas production in the Marcellus region was substantially reduced in May 2011 due to a
request by the Commonwealth of Pennsylvania asking companies to voluntarily stop sending their
Marcellus wastewater to these facilities (PADEP 201 la, 201 Ib; Wilson and Van Briesen 2012). The
request was based on concerns over increased bromide levels at PDWS intakes and associated increases
in disinfection byproducts within PDWS finished water in Pittsburgh (PADEP 201 la). However,
treatment and discharge of wastewater from other oil and gas production (conventional and
unconventional) continues (Vengosh et al. 2014). Most conventional wells in Pennsylvania require
-------
hydraulic fracturing stimulation due to reservoir characteristics (PA DEP 2014c). In 2012, a total of
992,137 barrels of conventional oil and gas wastewater were treated by the two CWTFs evaluated in this
study (PA DEP 2014a). The CWTFs also treated basic sediment, drilling fluid, fracking fluid, and
servicing fluid wastewater. Conventional oil and gas wastewater has similar composition to Marcellus
shale wastewater and both have highly variable compositions (Haluszczak et al. 2013; Lutz et al. 2013;
Wilson and Van Briesen 2012; Wilson et al. 2013).
This study was a component of the U.S. EPA study of hydraulic fracturing for oil and gas and its potential
impact on drinking water resources, and addresses the research question, "Wastewater Treatment and
Waste Disposal: What are the possible impacts of inadequate treatment of hydraulic fracturing wastewater
on drinking water resources?" (U.S. EPA 2014b). Monovalent ions such as bromide and chloride, are
present in high concentrations in oil and gas wastewater (mean bromide range of 602 to 973 mg/L and
mean chloride range of 68,375 to 99,800 mg/L; Ferrar et al. 2013). They are not effectively removed in
CWTFs, and the treated wastewater is discharged to the Allegheny River and its tributaries (Ferrar et al.
2013). Depending on a complex array of factors, the resulting elevated bromide concentrations in the
Allegheny River can lead to the formation of brominated disinfection byproduct analogs during the
drinking water treatment process.
Other sources also discharge substantial quantities of bromide to the Allegheny River and contribute to
the overall river bromide burden. Samples were collected from National Pollutant Discharge Elimination
System (NPDES) facility outfalls at coal fired generating stations, POTWs, coal bed methane, and
industrial facilities. This study identified and quantified the individual contributions from the source
types identified in the study domain (e.g., CWTF, flue gas desulfurization (FGD), acid mine drainage
(AMD)) to the elevated bromide concentrations observed in two downstream PDWS raw water intakes in
order to quantify and provide context for the contribution from oil and gas wastewater relative to other
sources.
The current study is responsive to the research question since most conventional wells in Pennsylvania
are hydraulically fractured. However, this study was not able to specifically classify the amount of
hydraulically fractured and non-hydraulically fractured oil and gas wastewater treated by the CWTFs for
summer and fall 2012 two week sampling campaigns.
1.2 Research Objectives
The objectives of this study were to quantify the cumulative contribution of treated oil and gas wastewater
from multiple CWTFs that treat hydraulic fracturing wastewater, and to distinguish that contribution from
other potential sources on bromide concentrations at two PDWS intakes located on the Allegheny River.
1.3 Source Apportionment and River Transport Modeling
Source apportionment techniques such as receptor modeling can be used to quantify the sources
contributing to water quality degradation (or contamination) based on mathematical modeling that uses a
combination of measured water sample species concentrations and discharge source profiles to quantify
the contribution of specific source types to observed contaminants. A few previous surface water source
apportionment studies have been conducted in the U.S. which focused on quantifying the contribution of
contaminants from application of roadway deicing materials and POTWs, and have used a combination of
source and river measurements to evaluate relative source contributions using bromide to chloride ratios
(Kelly et al. 2010). Soonthornnonda and Christensen (2008) utilized Positive Matrix Factorization (PMF)
8
-------
to both identify and quantify water contaminant sources in Milwaukee, Wisconsin. Henry and
Christensen (2010) compared results obtained with Unmix and PMF for (i) an artificial air pollution data
set and (ii) an actual sediment polychlorinated biphenyls (PCB) data set from a heavily contaminated
freshwater estuary (Sheboygan, WI). PMF has been widely used for air pollution source apportionment,
as well as to evaluate sediment data (Assefa et al. 2013; Norris et al. 2014; Praipipat et al. 2013; Zou et
al. 2013). This is the first source apportionment study in the Allegheny River in western Pennsylvania, to
quantify contributions from multiple potential sources of halides at public drinking water system (PDWS)
raw water intakes. The complex mixture of current and historical energy extraction and production, as
well as industrial manufacturing sources, represented a significant challenge.
Traditionally, river transport modeling is used to estimate river contaminant concentrations downstream
of a discharge location based on mathematical simulations of contaminant movement, attenuation of
contaminants downstream of a discharge, and river transport time (Jobson 1996). While many excellent
water dispersion models are available, they cannot readily be applied to a specific river system like the
Allegheny River without calibration and validation. Detailed information on channel geometry and
dispersion coefficients required for accurate modeling are not commonly accessible nor can they be
calculated using readily available hydraulic information. Receptor based source apportionment modeling
requires no a priori knowledge of river system dynamics or discharge sources. Receptor models identify
the source factors and quantify their contributions based solely on contaminant measurement data without
specifying numerous model input parameters (Bielski 2012), mass discharge rates, or conducting tracer
studies (Jobson 1996) needed for traditional river transport modeling. The input data requirements and
outputs for these two modeling approaches are shown in Appendix Figure Al. Two main differences
between these modeling approaches make them complementary: (i) source apportionment results are
based on actual samples collected at specific river locations and time periods while transport modeling
provides estimated concentrations at specified locations for a range of river flow and discharge
conditions, and (ii) source apportionment quantifies the combined contribution of a source type (e.g.,
CWTFs) whereas a transport model simulates the transport and dispersion of discharges from each
individual source or facility. Together, the two modeling approaches can be used to identify missing
sources, incorrect discharge rates, or the need to adjust modeling parameters.
-------
2. METHODS
Successful application of source apportionment on surface water domains requires a combination of river
and source discharge sampling, comprehensive chemical analysis of samples, and application of receptor
modeling to identify and quantify sources impacting a sampling location. Knowledge of source discharge
chemical characteristics (fingerprints) and their discharge locations are also critical for interpreting and
evaluating the source apportionment results.
2.1 Sampling Domain
Western Pennsylvania was chosen as the region to focus our research efforts because (i) the Marcellus is
one of the largest shale plays in the U.S. (Figure 3), (ii) development and production in the Marcellus was
increasing, (iii) Pennsylvania has a limited capacity for deep well injection of oil and gas waste water
with only eight EPA approved Class IID brine disposal wells, (iv) Pennsylvania allows hydraulic
fracturing oil and gas wastewater to be discharged into POTWs where it is diluted or treated by CWTFs
with subsequent discharge to surface waters, (v) thirteen (13) PDWSs downstream of POTW and CWTF
discharges utilize the Allegheny river as their source of raw drinking water, and (vi) PDWSs along the
Lower 48 states shale plays
Cody
H lard-
Banter-
Mancos "*
Greater
Ores'
Rw
, BaMi
Bakken
Ittom Powder River (SSSioi
3siff*~
Moriterey-
Temblor
eP^
iuiL^^^j n
C>r,»b. ^^Jpc.^^
/Easin
'
"-
""-Marcellus
Denver
Basin
Monterey-
3artta Ma-a
v^ntu-a. Los
AngfrtK
Ba&ns
Eneelto
Hulky^erofcee Platform
Karadox;E3«n Pierre ~^*
Le.,s--* Woodford
Floyd- Valley iRoo*
. Heal ' Proymce
0 100 200 300 400
N
A
Shale pUys
CLrren! ptays
Prospective plays
Stacked plays
' tfiKea shae &
cnait pay
_- - Hieflonepay
Snaltowesf youngest -Maedsnaiel
Intermediate depth< age ognt doostore-
Deepest' ofo
-------
The Allegheny River drains a catchment area of approximately 30,300 km2 in the Pennsylvania and New
York region of the northwestern Appalachian Plateau (Pennsylvania Fish and Boat Commission, 2011).
Most reaches of the Northern Allegheny flow over locally-derived river sediment overlying thick layers
of glacial outwash (sand and gravel). The Allegheny headwaters flow in a northwesterly direction for
approximately 90 km and cross the New York southern border, where the headwaters begin to flow
almost due west for another 77 km before cutting south into the head of the Allegheny Reservoir and back
into Pennsylvania. The flow leaves the reservoir via the Kinzua Dam, and meanders to the southwest for
319 km. Approximately 203 km of the Allegheny River downstream of the Kinzua Dam remains free-
flowing. Further downstream, 116 km of the Allegheny River are impounded and regulated by eight
navigatin dams, and, like the upstream Allegheny Reservoir, they are controlled and maintained by the
U.S. Army Corp of Engineers (USAGE). When the Allegheny River reaches its confluence with the
Monongahela River in Pittsburgh, it is classified as a low-gradient seventh-order system (White et al.,
2005) and classified by EPA as a large river (Flotemersch et al., 2006). Between the sampling sites in
this study, the Allegheny River is confined within a narrow, severely meandering valley with precipitous
side slopes.
The Allegheny River and its tributaries in western Pennsylvania can be affected by many different types
of contaminant sources including CWTFs, coal-fired electric power generating stations, AMD from
historic mining wastes, current mining operations, natural oil seepage, industrial manufacturing facilities,
POTWs, and industrial facility sewage treatment plants. These various sources discharge a mixture of
contaminants into surface waters including anions such as bromide, chloride, sulfate, and nitrate. The
Allegheny River is the source of drinking water for 13 water systems serving over half a million people in
western Pennsylvania. Understanding the sources of contaminants to water is critical due to their
potential impacts on drinking water quality (States et al. 2013). In many parts of the U.S. most of the oil
and gas well produced wastewater is disposed through deep well injection (Gregory et al., 2011). In
Pennsylvania, the majority of oil and gas well wastewater is either trucked out of state for deep well
injection or disposed of using alternative methods (Wilson and Van Briesen 2013; Veil 2010). In
addition, 1 percent of the conventional wastewater was used for roadway deicing and dust control in 2012
(Skalaketal. 2014).
2.2 River Sampling and Discharge Source Location
The EPA Office of Research and Development (ORD) worked with EPA Region 3 personnel to identify
two CWTFs in the study domain treating hydraulic fracturing wastewater, discharging to surface waters,
with downstream PDWS intakes. The study plan included one CWTF discharging into a large river
system, and one discharging into a small river system. Other suggested selection criteria included: (i) a
minimum total discharge from a CWTF of 40,000 GPD, (ii) presence of a PDWS intake downstream
(<65-85 km), and (iii) no substantial tributary inputs between the CWTF discharge and the PDWS intake.
Details concerning the CWTF facilities that meet these criteria are detailed below in Table 1.
11
-------
Table 1. Waste water treatment facility and receiving stream information.
^^^^^^m
Approximate Discharge Volume (GPD)
Receiving Stream
Receiving Stream 2010 Mean Daily Flow
CWTF_A
50,000
Allegheny River
1640
CWTF_B
67,000
Blacklick Creek
327
(cubic feet sec-1 (cfs))
Nearest Downstream Drinking Water Intake (km)
51
90
Samples were collected along the two river systems to evaluate the transport and dispersion of bromide
and other inorganic species to the closest downstream PDWS intake. All sampling sites and sources were
located along the Allegheny River and the Blacklick Creek to the Kiskiminetas River (Blacklick Creek,
Two Lick Creek, Conemaugh River, Loyalhanna Creek, and Kiskiminetas River) as shown in Figure 4.
Sites were selected due to their proximity to CWTFs, PDWS intakes, as well as other known sources.
Allegheny and Blacklick sites are identified with an "A" and "B", respectively. Samples were collected
at five (5) sites on each river system: upstream of the CWTF (SOI), the CWTF treated wastewater (S02),
downstream of the CWTF outfall (803), an intermediate location (S04), and finally the closest
downstream PDWS intake (805). River, CWTF, and PDWS samples were collected using automated
water samplers (Teledyne Isco Model 6712, Lincoln, Nebraska) to simultaneously collect daily samples at
all of the Allegheny and Blacklick study domain sites. An 800 mL daily composite sample was collected
in acid cleaned polypropylene bottles at each site by sampling two 400 mL aliquots, one at 09:00 and one
at 12:00 Eastern Standard Time (EST). In order to determine the sampling precision of the sampler,
collocated sequential samples (e.g., bottle 1 and bottle 2) were collected at the two PDWS intakes and a
river site using the Isco sampler.
All of the Allegheny study domain sites were on the Allegheny River with S01_A upstream of the
sampled CWTF to provide the chemical composition of the river before it is impacted by the CWTF
discharge, and S03_A at a distance downstream of the CWTF where it could be considered initially well
mixed. S04_A was downstream of a coal-fired electric generating station (no FGD) and S05_A was at
the PDWS intake. An additional CWTF was also located 105 km upstream of S01_A on the Allegheny
River (CWTF_C).
The Blacklick study domain sampling sites start on the Blacklick Creek and end at the confluence of the
Kiskiminetas and Allegheny Rivers. The Blacklick upstream sampling site (S01_B) and the site
downstream of the CWTF (S03_B) are both on the Blacklick Creek prior to its confluence with the Two
Lick Creek. The next downstream site (S04_B) is after the confluence of the Conemaugh River with the
Blacklick Creek, and a flood control reservoir with a dam. After S04_B, the Conemaugh joins with
Loyalhanna Creek to form the Kiskiminetas River. The PDWS intake S05_B is on the Allegheny River
after the confluence with the Kiskiminetas River. Site S05_B is 97.8 kilometers downstream of S05_A.
The USAGE manages the water discharge rate of the Allegheny River in the study domain by controlling
water discharge volumes directly into the river from the Kinzua reservoir, and into tributaries from the
Tionesta, Union City, and Woodcock reservoirs (USAGE, 2014). This active management provides for
12
-------
flood control, power generation cooling water requirements, water quality management, and downstream
navigation requirements. The water levels in the Upper Allegheny have historically been (i) highest in
the late fall and winter during reservoir drawdown in preparation for snow melt and spring precipitation,
resulting in a relatively high assimilative capacity from source discharges, and (ii) lowest in the summer
resulting in elevated concentrations of contaminants (USGS, 1993; USGS, 2013). As a result, the
Allegheny River summer and fall low flow periods are typically characterized by water discharge rates
that are less than 5000 cfs (Figure 5).
13
-------
Upstream
105 km of SOI A
*>
+
*
LEGEND
Sampling Site * S01_A
Centralized Waste cn-> A
, A ,^tf- /\
Treatment
Public Drinking /r\ cn^ A
Water Intake
Publicly-Owned ^
Waste Treatment**
Municipal Waste ^
Treatment
Cooling Tower ,0.
Outfall & Ash Pond
Flue-Gas Desulfurization £
Acid Mine Drainage ^
Mining Operation
Steel Mill
Coal Bed Methane
Oil Seepage
036 12 18 24
Kilometers
Figure 4. Allegheny River (blue) and Blacklick Creek to Kiskiminetas River (red) sampling sites,
and location of major surface water discharges. The arrows show the direction of river flow.
14
-------
ZUSGS
USGS 03025500 Allegheny River at Franklin, PA
-10000
Daily nean discharge
Period of approved data
Figure 5. USGS Gauge Station Data for 2012 showing the lower daily mean discharge levels during
summer and fall.
Mean discharge rates for low flow conditions are shown in Figure 6 along with U.S. Geological Survey
(USGS) gage station ID numbers (USGS 2014). USGS mean discharge data were used to evaluate the
differences in river flow between sites since it directly impacts the dilution of the facility discharges. The
CWTF_A discharged into the Allegheny River, which had a river flow rate of 2702 cubic feet per second
(cfs), and the river flow increased to 3734 cfs at the PDWS intake (S05_A). The CWTF_B discharged
into Blacklick Creek, which had a river flow rate of 68 cfs and increased to 886 cfs before the confluence
of the Kiskiminetas and Allegheny River. Figure 6 also shows that the contribution from tributaries is
very different between the Blacklick and Allegheny sites. River discharge (volumetric flow rate) is also
impacted by smaller tributaries without USGS gage data, direct storm water runoff, and hydrologic
ground water interactions.
The discharges into the Allegheny are diluted by a large river flow in comparison to the Blacklick
sampling sites. Blacklick sampling sites S01_B (upstream) and S03_B (downstream CWTF) are on the
Blacklick Creek which has a mean flow rate of 68 cfs. This is approximately three percent of the mean
flow in the Allegheny (2702 cfs upstream of CWTF_A), and therefore provides much lower dilution
capacity. The Conemaugh Reservoir has contributions from the Conemaugh River (443 cfs), Two Lick
Creek (74 cfs), and Blacklick Creek (68 cfs), which provides greater dilution of the discharges.
15
-------
Numerous wastewater source discharges are found along the Allegheny River and its tributaries that
impact the study domain. A simplified diagram of the rivers, creeks, and discharges is depicted in Figure
7. The volumes of wastewater treated by the CWTF_A and CWTF_B are listed in Table 1. The AMD
treatment facility on the Conemaugh River is also a CWTF, and it treated 5% of the wastewater treated by
CWTF_B (21,607 barrels). In addition, a CWTF discharged to Crooked Creek, and it treated 34% of the
volume treated by CWTF_A (193,872 barrels). The river segment with the largest number of source
contributions is between S03_B and S04_B (two coal-fired generating stations with FGD, one coal-fired
generating station without FGD, and a CWTF). Also, AMD contributes to all of the sites as noted in
Sams and Beer (2000). The flow rates and sources on Two Lick Creek, Blacklick Creek, and Conemaugh
River create a complex source mixture. The Conemaugh Dam (Figure 6, Figure 7) holds back and
integrates the flow from these rivers.
16
-------
1,972
Oil CYft.4.
340
v
No '
Data
sn A
-
8(B_A
Saiidv Cf«?ek
<
ft OHW13W
3,734
SOS^B, StNi.A - PPWS <
No
Onto
Jill
|?2
S86 :
Lo%.iili.innaCrLvk S04.B
r>!.'.:-|...-|-;v:".> ,-fD.t": '
5,382
Stti i
74
S02 B-CWTF(8> J*'
S03.B <> x?
-443
Figure 6. USGS mean discharge rates and gage IDs.
17
-------
Loyalhanna Conemaugh River
Generating Station* =
Treated cooling tower, coal pile
and ash wastewater
Two Lick Creek
POTW, Generating Station*
Blacklick Creek
POTWs, FGD,
Generating Station*
Figure 7. Diagram of sampling sites, rivers, and sources.
2.3 Source Sample Collection
Source discharge grab samples were collected from National Pollutant Discharge Elimination System
(NPDES) permitted outfalls that discharged to surface waters in the study domain to generate source
profiles (fingerprints). Internal outfall samples were collected in facilities with multiple permitted
wastewater sources before they were combined into the river outfalls to allow for chemical
characterization associated with each process. Source outfall samples were collected with acid-cleaned
polypropylene dippers and transferred into acid-cleaned polypropylene bottles. Source sample bottles
were rinsed three times with the wastewater, and two to four replicate samples were collected at each
outfall or sampling location. After collection, samples were double bagged with sealed bags, placed in
locked coolers with ice, and shipped overnight to EPA laboratories.
The location of sources upstream of the PDWS intakes where source grab samples were collected are
shown in Figure 4 and Figure 7. Treated outfall samples were collected from five coal-fired electric
power generating stations, three of which had spray lime FGD equipment and did not utilize calcium
bromide addition for supplemental mercury emission control (capacity was 1,884 megawatt, or MW, for
18
-------
two and 2,012 MW for the other). Two of the three FGDs discharged scrubber wastewater continuously,
while the third had a batch treatment process. Additionally, two of the three FGD outfalls were not
located near the generating facilities (3 - 28 km away), as shown in Figure 4. The other two generating
stations burn coal and have capacities of 585 and 95 MW. Additional outfall samples (cooling tower, ash
and coal pile runoff, and demineralizer) were collected at these facilities to elucidate the composition of
other wastewater sources. Four sewage treatment outfall samples were collected, representing a range of
discharges (1.3, 2, 0.11, and 0.0004 million gallons per day, or MGD). Samples were collected from
three industrial metals operations: specialty steel, specialty metals, and metals coating. Coal bed methane
(CBM) samples were collected at one facility with two internal outfalls with wastewater from
approximately 268 wells. Samples were also collected to represent untreated AMD, treated AMD, oil
seepage (water with visible oil residue), and active coal mine runoff.
The complexity of the source mixtures between the Allegheny and Blacklick sites were substantially
different. The sampling domain in the upper reaches of the Allegheny River had a larger volumetric
discharge rate and fewer source discharges were present. The Blacklick sampling domain contained more
source discharges, including AMD from historical mining activities, and a range of creek and river
tributary flow discharge rates.
2.4 Sample Analysis
In the field, a YSI (Yellow Springs, OH) Model 556 hand-held Multi-Probe System (MPS) was used to
electronically log instantaneous pH measurements at the river sampling sites and at the CWTFs. A three-
point calibration for pH (pH= 4.0, 7.0 and 10.0) was performed each day using calibration standards
(Orion pH Buffer Bottles, Thermo Scientific, Beverly, MA). An evening drift check was performed. If
the difference in pH was not within 0.2 of the calibration standard, data for that day were flagged as being
invalid.
Ion chromatography (1C) was used to quantify dissolved inorganic anions, including bromide, chloride,
sulfate, and nitrate. Analysis of total extractable elements was performed with both inductively coupled
plasma - optical emission spectrometry (ICP-OES) and high-resolution magnetic sector field -
inductively coupled plasma mass spectrometry (HR-ICPMS) to cover a wide concentration range of trace
and ultratrace species. Prior to the inorganic analyses, the samples were analyzed for specific
conductivity using a two point calibration with a Mettler Toledo (Columbus, OH) Model S47-K meter
equipped with an InLab Model 731 probe. These data guided volumetric and gravimetric dilution
determinations for instrumental analyses. By diluting the dissolved solids in analyzed samples, the
potential to cause sample spectral and polyatomic interferences, saturation of the instrument detector, and
other analytical issues (e.g., memory effects, internal standard suppression, peak broadening) were
minimized. The dilutions can also have the adverse consequence of diluting trace species in the parent
sample down to below method detection limit (MDL) in the analytical aliquot.
Major ions were quantified by simultaneous injection on two Dionex (Sunnyvale, CA) Model ICS-2000
instruments following a modified U.S. EPA Method 300.1 (U.S. EPA, 1997). All samples were filtered
prior to analysis with 1C Millex 0.20 (im PTFE syringe filters (Millipore). Anion separation was achieved
using a 200 (iL injection loop, AS 18 and AG18 analytical and guard columns (Dionex), and an isocratic
potassium hydroxide eluent. Cation separation was achieved with a 25 (iL injection loop, CS16 and
CG16 analytical and guard columns (Dionex), and an isocratic methanesulfonic acid method. Analytical
19
-------
procedures were the same for the river and outfall source samples, except that the outfall and CWTF
samples required additional dilutions. The dilution volume was based on the sample conductivity.
Samples were acidified and microwave digested using EPA Method SW846 3015A (U.S. EPA 2007)
prior to analysis by ICP-OES and HR-ICPMS analysis.
Samples were acidified to 2.0% nitric acid (volume by weight, or v/w) and allowed to leach for 7 days to
provide time for the elements to leach from particles (Graney et al. 2004; Landis et al. 2002). Samples
were then acidified to 0.5% hydrochloric acid (volume by volume, or v/v), oxidized with 0.05% hydrogen
peroxide (v/v), and immediately microwave digested following EPA Method 3015A (U.S. EPA, 2007a).
After microwave extraction the samples were vacuum filtered through 47 mm cellulose nitrate filters.
Fifteen (15) mL aliquots were then poured off for ICP-OES and HR-ICPMS analyses. Elemental analysis
for major elements was performed with a PerkinElmer (PE) Optima 4300 DV ICP-OES following EPA
Method 200.7 (U.S. EPA, 1994). Trace elemental analysis was performed using a ThermoFinnigan
(Bremen, Germany) Element2 double focusing HR-ICPMS using EPA Method 6020A (U.S. EPA,
2007b), utilizing all three resolution settings in a multi-element quantitative analysis (U.S. EPA 2014c).
Sample data was considered valid only if analyte concentrations did not exceed the highest concentration
calibration standard nor were less than the lowest concentration standard as described in EPA Methods
300.1 and 8000B (U.S. EPA, 1996). For several species, certified reference materials were used to extend
the reportable linear range. If a single sample was analyzed at more than one dilution producing multiple
valid results within the linear range, specific results were prioritized relative to the analytical method. For
1C, the result from the least diluted sample was chosen, because the technique is more robust, which also
provided the quantification for the majority of the species. For the ICP methods (ICP-OES and HR-
ICPMS), the most diluted sample was chosen due to the trace nature of the technique and the ability of the
plasma to be altered due the heavily loaded matrices.
As a matter of instrument performance, MDLs for 1C, ICP-OES, and HR-ICPMS are provided in Table 2,
Table 3, and Table 4, respectively. 1C limits were based upon repeated injections of the lowest calibration
standard over multiple sequences (n=3). ICP-OES detection limits were determined from the analysis of
repeated reagent blanks and the lowest calibration standard over several calibration sequences. HR-
ICPMS detection limits were based upon the analysis of the microwave digested nitric acid blanks that
were analyzed over multiple analytical sequences to minimize bias.
Table 2. Ion chromatography method detection limits (MDL).
Species MDL (mg/L)
Bromide 0.0063
Chloride 0.0258
Fluoride 0.0039
Nitrate 0.0169
Sulfate 0.0129
20
-------
Table 3. Inductively coupled plasma-optical emission spectrometry method detection limits (MDL).
Species
Ba
Ca
Fe
K
Li
Mg
Mn
Na
S
Si
Sr
MDL (mg/L) I
0.000619
0.00419
0.00320
0.0585
0.00336
0.00218
0.000937
0.0122
0.0349
0.00445
0.000792
Table 4. High-resolution-inductively coupled plasma mass spectrometry method detection limits (MDL)
Species
Ag
Ba
Be
Bi
Cd
Ce
Cs
Dy
Gd
La
Mo
Nd
Pb
Pd
Rb
Rh
Sb
Sm
Sr
Tb
Th
Tl
Resolution
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
MDL (no/L)
0.0113
0.506
0.0542
0.0126
0.107
0.015
0.0059
0.0039
0.0086
0.0290
0.161
0.0286
0.0380
0.0231
0.0776
0.0055
0.0339
0.0109
0.283
0.0148
0.0097
0.0279
Isotope
U
W
Y
Al
Co
Cr
Cu
Fe
Mn
Ni
P
S
Si
Sn
Ti
V
Zn
As
K
Se
Resolution
Low
Low
Low
Medium
Medium
Medium
Medium
Medium
Medium
Medium
Medium
Medium
Medium
Medium
Medium
Medium
Medium
High
High
High
MDL (ng/L) I
0.0064
0.0170
0.0323
1.85
0.178
0.106
0.323
1.97
0.203
0.473
1.16
62.1
19.0
0.111
0.0964
0.109
0.438
0.104
6.55
0.453
21
-------
2.5 PMF Receptor Model
The PMF version 5.0 multivariate receptor model was used to identify sources, quantify contributions,
and determine uncertainties (U.S. EPA 2014a). Multivariate receptor models solve the chemical mass
balance between measured species concentrations and source profiles by decomposing speciated sample
data into two matrices: factor contributions and profiles (Hopke 1991). One of the important receptor
modeling assumptions is that species are conserved and not transformed or deposited between the source
outfall and the receptor or sampling site.
A speciated data set can be viewed as a data matrix X (entire matrices are denoted by capital boldface
letters) of dimensions n by m, in which n samples and m chemical species were measured. Rows and
columns of X and of related matrices are indexed by / and j, respectively. The goal of multivariate
receptor modeling, for example with PMF, is to identify the number of factors p, the species profile /of
each factor (mass fraction), and the amount of mass g contributed by each factor to each individual
sample that solve the chemical mass balance between measured species concentrations and factor profiles.
This calculation is show in Equation 1 :
Where e-^ is the residual for each sample/species and c;j is the modeled solution of x^. Multivariate
receptor models calculate the factor profiles and contributions, which are concentrations, based only on
the measured data. In PMF, measured profile information can be used to constrain model results.
Additionally, measured source information can also be used as constraints to reduce rotational ambiguity
(Amato and Hopke 2012; Norris et al. 2009; Paatero 1997; Paatero and Trapper 1994), which represents
the range of profiles and contributions that can equivalently reproduce the measured data after
implementing non-negativity constraints (Henry 1987; Paatero et al. 2002). The PMF solution minimizes
the object function Q (Equation 2), based upon the estimated data uncertainties uij.
(2)
The results were evaluated within EPA PMF 5.0 using plots and statistics.
One key component of PMF is the use of user-provided sample species uncertainties to scale the residuals
in minimization, which allows for down weighting of more uncertain species or samples. PMF sample
uncertainties were developed using the sequential sampling and the precision for each species (see
Section 3.1). The uncertainty was calculated as the average percent difference multiplied by the species
concentration plus the method detection limit (MDL) for river or outfall samples. Data that were below
detection limit were replaced by MDL/A/2 and an uncertainty of 3*MDL.
The PMF displacement algorithm was used to evaluate the stability of the solution and to generate factor
uncertainties (Paatero etal. 2014). Displacement perturbs the individual fitted source profile values until
the object function reaches a defined change in Q, and the range of perturbed values is used to determine
22
-------
the upper and lower source profile species interval. The uncertainty or interval range can be high for low
concentration species in a PMF factor, and these species were not used to identify the source. PMF
provides both factors and their contributions to each species, and each factor was identified to be a source
type based on (1) comparison of the PMF source type chemical species ratios to measured profiles and (2)
the outfall locations.
2.6 Statistical Analysis
Data processing and all statistical analyses were performed using SAS v.9.4 (SAS Institute). The
Kruskal-Wallis non-parametric statistical test was used for comparing species concentrations between the
Allegheny and Blacklick sampling sites and the CWTFs. A level of significance of a=0.001 was used.
Sequential precision was calculated as the absolute percent difference for replicate and sequential sample
analysis.
23
-------
3. QUALITY ASSURANCE AND QUALITY CONTROL
Quality assurance (QA) and quality control (QC) procedures were implemented by following two Quality
Assurance Project Plans (QAPPs): Hydraulic Fracturing Wastewater Source Apportionment Study
(Revision No. 2, February 2013), and the QAPP for Surface Water Source Apportionment Model
Applications (August, 2014). The EPA PMF model was developed under an additional QAPP, EPA
Positive Matrix Factorization (EPA PMF) QAPP (approved 9/27/13).
Quality control samples were used where appropriate and available for assessing potential contamination
of field sampling materials, and spiked samples were used to assess recovery of inorganic species from
the water samples. Sequential Isco samples were collected to determine overall sampling and analytical
precision. Data quality reviews were conducted by EPA staff and Alion Science and Technology (EPA
contract EP-D-10-070). The QA and QC results described below apply to the methods described in
section 2.3 and 2.4.
3.1 Quality Control Results for River and Source Sample Analysis
Conductivity replicate analysis was performed on 10% of the samples, and the observed precision (1 -
relative percent difference, n=54) was 99 ± 1%, while the sequential sample precision (n=123) was 98 ±
5%. External 1C, ICP-OES, and HR-ICPMS calibrations were verified by regression statistics (r2),
National Institute of Standards and Technology (NIST) Standard Reference Materials (SRM), and
secondary source calibration standards (1C in Table 5), ICP-OES in Table 6, and HR-ICPMS in Table 7).
Performance verification samples were from the National Institute of Standards and Technology Standard
Reference Materials (NIST SRM; Gaithersburg, MD), Environmental Resource Associates (ERA;
Golden, CO), High Purity Standards Certified Reference Material (HP CRM; Charleston, SC), and United
States Geological Survey. The 1C performance evaluation samples (Table 5) were analyzed only once so
that no standard deviation is available. Digestion procedures were monitored with digestion blanks, lab
control samples, and spiked unknown samples and their recoveries. Instrument analytical sequences were
monitored for precision and accuracy by continuing calibration blanks, verification standards, continuing
check samples, and duplicate sample analysis.
The analytical sequence precision from laboratory duplicates for the ICP-OES and HR-ICPMS analyses is
summarized in Table 8. Only species that were quantified in 50% or more of the laboratory duplicate
samples are reported in the table.
For field precision analysis, two sequential samples were collected by the Isco sampler (container 1 and
container 2) during defined collection periods: S05_A and S05_B during spring, summer, and fall
campaigns; S03_B during spring; SO4_B during summer; and S03_A during fall. Table 9 shows the
average sequential sampling precision by analytical method.
Field blanks were collected at all sites (SOI, S02, S03, S04, 805) during the field monitoring campaigns.
The field blank sample bottles were left uncapped in the Isco samplers in the same manner as the
deployed sample collection bottles for two to three days to capture the longest time when a sample would
be exposed to potential contamination. The field blank river concentrations for anions and cations were
low, as shown in Table 10. The CWTF field blanks for anions and cations are higher, as shown in Table
11 since S02_A was collected inside a treatment facility and S02_B was collected near the top of a large
holding tank. Levels found in the field blanks were determined to have no effect on sample values, due to
24
-------
the relatively low concentrations of field blanks with respect to collected samples (river, CWTF, and
manually collected samples). As a result, reported sample concentrations were not blank corrected.
Table 5. Ion chromatography recovery of performance evaluation samples.
Species
Fluoride
Chloride
Nitrate
(asN)
Bromide
Sulfate
Sodium
Potassium
Magnesium
Calcium
ERA 505
Target Recovery
(mg/L) %
ERA 693
Target Recovery
(mg/L) %
ERA 698
ERA 5262
Target Recovery Target Recovery
(mg/L) % (mg/L) %
4.48
102
19.8
9.53
55.3
98
98
103
7.82
142
6.35
205
236
33.8
94
108
92
99
99
0.28
90
25
-------
Table 6. Inductively coupled plasma-optical emission spectrometry recovery (R) of certified (Cert) materials, reference materials, and sequence
calibration accuracy checks.
Standard Reference Material
(SRM)1640a (n = 31)
SRM 1643e (n = 31)
High Purity (HP),
CRM-TMDW
(n = 14)
USGS,
M-172 (n = 31)
ERA-500 (n = 3)
Cert., R%,
ERA custom mix (n = 3)
Species [igli. Mean ± SD1, [igli. Mean ± SD, [igli. Mean ± SD, [igli. Mean ± SD, [igli. Mean ± SD, mg/L Mean ± SD,
Al
As
B
Ba
Ca
Cd
Ce
Co
Cr
Cu
Fe
K
Li
Mg
Mn
Mo
Na
Ni
P
Pb
50.7
4.4 (a)
318.1
151.4
5542.7
3.9
19.6
39.7
83.7
36.4
562.8
1031.5
39.4
42.6
3075.7
22.2
11. 7 (a)
96
92
106
101
100
100
98
99
98
100
98
98
98
94
99
89
97
± 16
± 24
± 4
± 4
± 4
± 15
± 3
± 2
± 4
± 8
± 4
± 4
± 3
± 7
± 4
± 5
± 20
138.8
59.0
154.0
531.0
31500.0
6.4
26.4
19.9
22.2
95.7
1984.0
17.0
7841.0
38.0
118.5
20230.0
1984.0
19.2
99
95.5
104.7
99.8
96.5
103.8
95.1
101.0
115.9
101.4
96.8
105.8
97.1
99.8
100.9
98.0
94.0
94.7
± 6
± 4.4
± 5.7
± 3.4
± 3.1
± 12.6
± 3.1
± 4.0
± 17.4
± 8.9
± 3.2
± 12.0
± 2.9
± 3.2
± 3.0
± 3.0
± 0.1
± 12.3
120.0
80.0
50.0
35000.0
10.0
25.0
20.0
100.0
2500.0
20.0
9000.0
40.0
100.0
6000.0
60.0
40.0
96
99
100
100
98
97
100
100
99
98
99
101
99
100
92
97
± 5
± 3
± 2
± 3
± 8
± 2
± 3
± 3
± 2
± 7
± 3
± 2
± 2
± 2
± 1
± 3
97.1
0
123 ± 15
8360.0 106
3800.0 105 ± 4
4730.0 104 ± 5
12500.
104 ± 5
1350.0 102
530.0
753.0
859.0
827.0
102
101
99
101
689.0 91 ± 1
215.0 104 ± 2
214.0 100 ± 1
842.0 92 ± 1
795.0 102 ± 2
1850.0 101 ± 2
575.0 102 ± 1
1720.0 97 ± 2
45.4 101.5 ± 1.
26
-------
Standard Reference Material
(SRM)1640a (n = 31)
SRM 1643e (n = 31)
High Purity (HP),
CRM-TMDW
(n = 14)
USGS,
M-172 (n = 31)
ERA-500 (n = 3)
ERA custom mix (n = 3)
Species
S
Sb
Se
Si
Sn
Sr
Ti
Tl
V
Zn
Cert.,
Mg/L
4.1 (a)
18.9 (a)
5231.1
124.5
14.7
55.8
R%,
Mean
82
95
101
100
99
101
± SD1,
± 33
± 17
± 3
± 4
± 6
± 5
Cert.,
M9/L
56.9
11. 7 (a)
315.2
36.9
76.5
R%,
Mean
94.9
125.9
97.8
99.4
101.7
± SD,
± 4.0
± 27.0
± 4.6
± 2.2
± 5.5
Cert.,
M9/L
10 (a)
10.0 (a)
250.0
10.0 (a)
30.0
70.0
R%,
Mean
84
108
100
135
99
101
± SD,
± 18
± 20
± 2
± 37
± 3
± 2
Cert.,
Mg/L
4105.7
5697.4
54.0
10.3
R%,
Mean ±
109 ±
104 ±
105 ±
100 ±
SD,
3
2
5
9
Cert.,
M9/L
460.0
632.0
291.0
65.1
744.0
931.0
1330.0
R%,
Mean
98
100
101
99
101
100
96
± SD,
± 1
± 1
± 2
± 3
± 2
± 1
± 2
Cert., R%,
mg/L Mean ± SD,
56.8 102.0 ± 0.6
Standard deviation
27
-------
Table 7. High-resolution-inductively coupled plasma mass spectrometry recovery (R) of certified reference materials.
NIST SRM 1640a
NIST SRM 1643e
Target
Species (M9/kg)
Al
As
Ba
Cu
Fe
K
Mn
Mo
Ni
Pb
Rb
Sb
Sr
U
Zn
52.6
8.01
150.6
85.07
36.5
575.3
40.07
45.24
25.12
12.005
1.188
5.064
125.03
25.15
55.2
Digest
R%
102
94
101
96
100
100
99
105
96
102
97
99
103
102
94
Daily
Accuracy
R%Mean3 (
(n=31) SD%
104
101
101
104
104
104
105
106
103
97
103
99
105
95
100
Target
(Mg/kg)
138.33
58.98
531.00
22.2
95.7
1984
38.02
118.5
60.89
19.15
13.8
56.88
315.2
76.5
Digest
R%
103
87
105
88
100
98
97
106
92
102
101
95
103
80
Daily
Accuracy
R%
108
91
108
96
105
105
105
112
101
98
108
96
111
84
ERA Mix A"
Target
(M9/L) %
ERA Mix B"
530
753
827
842
795
1850
575
1720
460
632
65.1
105
103
101
99
101
105
101
106
101
108
95
65400 116
1330 102
1 Microwave Digested Lab Control Sample
2 Standard deviation
3 Daily sequence calibration accuracy checks (not microwave digested)
4 Diluted within Linear Dynamic Range prior to HR-ICPMS analysis
28
-------
Table 8. Analytical precision for laboratory duplicates.
Species
Al
As
Ba
Ca
Cu
Fe
K
Mg
Mn
Mo
Na
Ni
P
Pb
Rb
S
Sb
Si
Sr
U
Zn
HR-ICPMS1 Average
Relative Percent Difference (± SD) (n=96)
3 (±5)
16 (±30)
2 (±4)
3 (±9)
7 (±17)
3 (±7)
4 (±8)
16 (±43)
3 (±4)
9 (±35)
2 (±5)
1(±1)
5 (±13)
17 (±33)
2 (±4)
10 (±29)
5 (±10)
ICP-OES2 Average Relative Percent
Difference (±SD) (n=120)
9 (±8)
3 (±4)
4 (±9)
12 (±9)
6 (±11)
5 (±8)
4 (±8)
5 (±13)
6 (±15)
2 (±2)
5 (±12)
4 (±6)
1 High-resolution-inductively coupled plasma mass spectrometry
2 Inductively coupled plasma-optical emission spectrometry
29
-------
Table 9. Average sequential sampling precision (%
Species
Ba
Br
Ca
Cl
Cu
F
Fe
Fe
K
K
Li
Mg
Mn
Mn
Mo
Nitrate (N03-)
Na
Ni
Pb
S
Sulfate (S042-)
Si
Sr
Sr
Zn
Analysis
ICP-OES
1C
ICP-OES
1C
HR-ICPMS
1C
HR-ICPMS
ICP-OES
HR-ICPMS
ICP-OES
ICP-OES
ICP-OES
HR-ICPMS
ICP-OES
HR-ICPMS
1C
ICP-OES
HR-ICPMS
HR-ICPMS
ICP-OES
1C
ICP-OES
HR-ICPMS
ICP-OES
HR-ICPMS
Precision1
96.5
99.3
96.4
99.6
87.7
97.6
66.4
71.6
96.4
93.7
90.9
96.3
92.9
76.4
96.3
92.6
96.0
88.4
68.2
96.9
98.0
91.5
97.1
96.3
70.9
Precision = (100 - 100*(absolute value (first sequential sample - second sequential sample)/average))
30
-------
Table 10. Summary of river and PDWS sampling site field blanks (mg/L).
Fluoride
Chloride
Bromide
Sulfate
Nitrate
Lithium
Sodium
Ammonium
Potassium
Magnesium
Calcium
Table 1 1 . Summary
Fluoride
Chloride
Bromide
Sulfate
Nitrate
Lithium
Sodium
Ammonium
Potassium
Magnesium
Calcium
N
16
30
0
4
22
28
1
6
0
30
24
N
36
102
0
87
95
2
104
102
63
66
80
% detected
33
95
0
80
87
2
95
94
59
62
76
of CWTF sampling
% detected
47.1
88.2
0
11.8
64.7
82.4
2.94
17.6
0
88.2
70.6
Mean
Standard
Deviation
0.00600 0.00452
0.0814
0.0114
0.0307
0.0027
0.0292
0.0142
0.0281
0.0129
0.0235
site field
Mean
0.00476
1.02
0.0394
0.0200
0.0956
0.0976
0.00532
0.0393
0.113
0.0582
0.0123
0.0325
0
0.0309
0.0102
0.0271
0.0140
0.0189
blanks (mg/L).
Standard
Deviation
0.00183
3.21
0.0423
0.0289
0.226
0.00421
0.0461
0.428
Median
0.00475
0
0
0
0
0
0
0
0
0
.0679
.0061
.0227
.0027
.0235
.0120
.0165
.0097
.0226
Median
0.00445
0
0.
.121
0394
0.00445
0.
0.
0.
0.
0.
0322
0976
0026
0206
0188
Max
0.0268
0.334
0
0.055
0.186
0.0027
0.168
0.0414
0.156
0.0529
0.135
Max
0.0075
12.8
0
0.0762
0.0985
0.892
0.0976
0.0108
0
0.186
2.12
min
0.0024
0.0228
0
0.0007
0.004
0.0027
0.0018
0.0002
0.0119
0.0045
0.0003
min
0.0027
0.0246
0
0.0026
0.0022
0.009
0.0976
0.0026
0
0.0006
0.0119
31
-------
3.2 Quality Control Results for PMF Analysis
PMF results were evaluated using observed versus predicted regression and time series plots, and the
displacement algorithm was used to evaluate the stability of the results. Known source discharge
chemical signatures (fingerprints) were also used to evaluate the results as well as to interpret the
measured source profiles. All of the results were numerically stable based on the displacement evaluation
(no swaps at dQmax = 4), and the regression results for observed versus predicted had a coefficient of
determination greater than 0.9 for the strong or non-down-weighted species.
3.3 Quality Assurance Assessments
QA assessments of field data collection and laboratory analyses were performed at multiple levels. A
summary of reviews and outcomes is provided below.
On-site QA assessment of field sampling procedures and adherence to protocols was performed
via a Technical Systems Audit (TSA) on May 3, 2012. Field sampling activities were observed at
five individual sites, and two corrective actions were implemented. The first action, implemented
immediately, was to ensure that the scribe consistently repeated the sampling result (as called out
by the researcher taking the reading) to confirm that correct measurements were being recorded in
the field. The second corrective action added the logging of cooler temperatures during sample
collection in addition to the logging of cooler temperature during shipment of the samples to
Research Triangle Park, NC.
On-site QA assessment of laboratory sample storage, processing, and analysis processes for 1C,
ICP-MS, and ICP-OES analysis was performed via a Technical Systems Audit (TSA) on July 16-
17, 2012. Using bar codes and scanners to track samples was noted to be best practice. No
corrective actions were deemed necessary based on the findings of this audit.
Surveillance audits were conducted during pH and conductivity measurements (February 22,
2013), 1C analysis (March 5, 2013), ICP-OES analysis (May 9, 2013), and HR-ICPMS analysis
(February 19, 2014) prior to performance evaluation sample analysis. The assessments found that
protocols were being properly implemented, and no corrective actions were deemed necessary.
Performance evaluation samples were analyzed using blind samples obtained from a third party
vendor (ERA) forthe conductivity (June 14-15, 2013), 1C (June 14-15, 2013), ICP-OES (June 19,
June 27, and July 1, 2013), and HR-ICPMS analyses (February 20, 2014). All samples analyzed
were within the acceptance limits except for fluoride for the 1C analysis, which was
approximately 1.5% below the acceptance range. All fluoride concentrations were flagged as
being only estimated values.
The analytical laboratory performed ongoing review of calibration, continuing calibration
verification checks, QC recovery and background assessments, and instrumental performance
parameters. Any samples identified by QC as being out of range were reanalyzed or flagged as
needed.
Analytical laboratory results were always QA reviewed by the analyst to ensure that the results
were complete and accurate. QC results were summarized and examined to ensure overall data
quality objectives were met.
Audits of data quality (ADQs) were performed on the 1C pH/conductivity data (July 2013), 1C
data (August/September 2013), ICP-OES data (February 2014 and March/April 2014), and HR-
ICPMS data (April 2014). The ADQs helped to ensure that the measurement data were
32
-------
accurately transcribed into data analysis files, calculations were correctly performed, and data
qualifiers (flags) were appropriately assigned. All findings that may have potentially affected
data quality were either addressed immediately during the audit, or prior to use of the data in any
report. Any data that did not meet the designated quality criteria were not used, or were qualified
accordingly.
33
-------
4. RESULTS AND DISCUSSION
A total of 117 river and PDWS intake samples were collected during summer (July 5-19, 2012) and 114
samples were collected during fall (September 19-October 4, 2012). In addition, 47 samples were
collected from CWTFs, and 71 outfall or source samples were collected with two to four samples for each
source type at a facility. River and outfall samples were analyzed using conductivity, 1C, ICP-OES, and
HR-ICPMS. The concentration data were analyzed with EPA PMF, and the results were compared to
measured outfall profile concentrations and Br/Cl ratios, which are indicative of sources (Andreasen and
Fleck 1997, Davis et al. 1998). Species that were quantifiable in approximately 90% or more of the
Allegheny or Blacklick downstream samples (803, S04, 805) included: bromide (1C), chloride (1C),
fluoride (1C), nitrate (1C), sulfate (1C), barium (ICP-OES and HR-ICPMS), calcium (ICP-OES), iron
(ICP-OES and HR-ICMPS), potassium (ICP-OES and HR-ICMPS), lithium (ICP-OES), magnesium
(ICP-OES), manganese (ICP-OES and HR-ICPMS), sodium (ICP-OES), sulfur (ICP-OES), silicon (ICP-
OES), strontium (ICP-OES and HR-ICMPS), aluminum (HR-ICPMS), copper (HR-ICPMS),
molybdenum (HR-ICPMS), nickel (HR-ICPMS), lead (HR-ICPMS), rubidium (HR-ICPMS), antimony
(HR-ICPMS), and zinc (HR-ICPMS). These species were considered for the PMF analysis except for Ba,
which could form barium sulfate (BaSO4), a salt, in the high sulfate concentrations on the Blacklick Creek
and would potentially precipitate out of solution and settle into sediments before reaching the downstream
sites (Lee et al. 2002).
The CWTF contribution as well as other sources were quantified using a comprehensive source
apportionment modeling approach including: (i) developing chemical source profiles, or fingerprints, for
all sources located upstream of two PDWS intakes on the Allegheny River by collecting outfall samples
from specific sources within facilities and combined river outfalls; (ii) collecting and chemically
characterizing river samples from multiple sites upstream and downstream of CWTFs, generating
stations, industrial facilities, and at the PDWS intakes, and (iii) using the EPA implemented Positive
Matrix Factorization (PMF) source apportionment model to quantify the contribution of sources to anion
levels at the PDWS intakes. Although the two PDWS intakes in this study were both on the Allegheny
River, their relative location with respect to the numerous source discharges warranted a source
apportionment approach, since the approach is not reliant on the many unavailable model specifications
needed by other modeling approaches such as deterministic dispersion modeling. One PDWS intake is in
the upper reaches of the Allegheny River 51 km downstream of a CWTF with few tributaries and other
discharges between, while the other is located in the middle reaches of the river 1.3 km downstream from
a river confluence with a major tributary (Kiskiminetas River). The Kiskiminetas River drains a
catchment area containing a complex mixture of discharge sources and elevated contaminant
concentrations. This second PDWS intake was the sampling location with the highest raw water intake
bromide concentration, and it was located downstream of several CWTFs. The second PDWS intake
represents a challenge for river transport modeling due to the complex nature of the larger catchment
basin, numerous source discharges, intermittent discharges, and confluence mixing dynamics.
4.1 River Sample Composition
Upstream and downstream differences in halide concentrations between the Allegheny and Blacklick
study domains were evaluated to assess whether the combination of sources and flow rates impacted the
observed concentrations. Species concentrations for all upstream and downstream samples are
summarized and presented in Table 12. Bromide concentrations upstream of the CWTFs on the
34
-------
Table 12. Downstream (803, S04, 805) and upstream (SOI) concentrations (mg/L).
Allegheny Sites
Species
Br
(1C1)
Ca
(ICP-OES2)
Cl
(1C)
Fe
(HR-ICPMS3)
Mg
(ICP-OES)
Mn
(HR-ICPMS)
N03
(1C)
Na
(ICP-OES)
S04
(1C)
Statistic
mean (std)
median (IQR)
min - max
mean (std)
median (IQR)
min - max
mean (std)
median (IQR)
min - max
mean (std)
median (IQR)
min - max
mean (std)
median (IQR)
min - max
mean (std)
median (IQR)
min - max
mean (std)
median (IQR)
min - max
mean (std)
median (IQR)
min - max
mean (std)
median (IQR)
Downstream
0.0925 (0.044)
0.0827 (0.0425)
0.037 - 0.28
17.8(1.72)
17.8(2.7)
13.8-21.6
23.7 (4.56)
22.8 (4.35)
16.6-41.6
0.192(0.107)
0.146(0.125)
0.0767 - 0.541
3.63 (0.285)
3.64 (0.44)
3.02 - 4.27
0.0716 (0.0433)
0.0548 (0.0685)
0.0215-0.191
0.42 (0.309)
0.335 (0.535)
0.0178-1.2
12.6(1.93)
12.3(2)
9.27-20
9.06 (0.783)
8.97 (0.94)
Upstream
0.0718(0.0254)
0.0721 (0.0446)
0.0329-0.117
15.2(1.19)
15.3(1.4)
12.4-17.6
20.8 (2.93)
21 (4.9)
16.2-25.8
0.176(0.0903)
0.154(0.097)
0.071 - 0.384
3.09(0.191)
3.11 (0.25)
2.56 - 3.44
0.0607 (0.0326)
0.0536 (0.059)
0.0181-0.123
0.59 (0.366)
0.545 (0.558)
0.0188-1.26
11.2(1.45)
11.5(2.3)
8.18-13.5
7.47(0.341)
7.49 (0.31)
Blacklick Sites
Downstream
0.432 (0.71)
0.187(0.307)
0.0635 - 4.36
52.8 (26.2)
48.8 (35.3)
23.3-168
64.6 (97.5)
35.7 (29)
21.6-621
0.943(3.15)
0.336 (0.402)
0.0791-32
15(6.28)
13.2(11.8)
6.93-27.8
0.281 (0.303)
0.136(0.3)
0.0279-1.02
2.56(1.05)
2.39(1.72)
0.823 - 4.75
38.7 (40.2)
28.7(19.7)
15.2-255
172 (92.7)
157(169)
Upstream
0.0696 (0.0175)
0.0726 (0.0287)
0.0352 - 0.0889
56.3(19.7)
53.8 (26.4)
25.6 - 89.9
19.6(1.79)
19.6(2.2)
14.1-23.2
2.6 (4.43)
1.33(1.39)
0.479 - 23.7
15.7(5.31)
16.1(8.1)
6.79 - 24.4
0.808 (0.207)
0.913(0.34)
0.417-1.06
1.83(0.384)
1.8(0.34)
1.04-2.76
24.3(8.11)
23.3(10.7)
13.9-40.7
244 (87.3)
242 (133)
35
-------
Allegheny Sites
Species
Si
(ICP-OES)
Sr
(ICP-OES)
Statistic
min - max
mean (std)
median (IQR)
min - max
mean (std)
median (IQR)
min - max
Downstream
7.75-11.7
0.958 (0.397)
0.86(0.631)
0.198-1.87
0.0814(0.0166)
0.077(0.0167)
0.0607-0.144
Upstream
6.78-8.6
0.961 (0.335)
1.01(0.472)
0.134-1.55
0.0655 (0.00886)
0.0641 (0.0136)
0.0483 - 0.0838
Blacklick Sites
Downstream
46.2 - 370
2.26(1.29)
1.91(1.31)
0.531 - 5.47
0.677(1.23)
0.388 (0.386)
0.132-7.55
Upstream
93.9-377
4.24 (0.842)
4.52(1.56)
2.7-5.21
0.341 (0.118)
0.341 (0.177)
0.151-0.531
' Ion chromatography;2 Inductively coupled plasma-optical emission spectrometry;3 High-resolution-inductively coupled plasma mass spectrometry.
36
-------
Allegheny River and Blacklick Creek were not significantly different during the low flow sampling
campaigns (summer and fall 2012), with median concentrations of 0.072 and 0.073 mg/L, respectively.
The overall distributions of downstream concentrations were affected by the intermittent CWTF
discharges due to each facility's operation schedule, with CWTF_A discharging on 17 days and CWTF_B
discharging on 15 days of the total 25 sampling days. Median bromide concentrations at the downstream
sites (803, S04, S05) were significantly higher on the Blacklick (0.187 mg/L) compared to the Allegheny
(0.083 mg/L). The relatively high downstream site bromide impacts in the Blacklick study domain were a
function of mean CWTF discharge volumes during the study period from CWTF_A (75,116 gallons/day)
and CWTF_B (58,551 gallons/day), discharging into the Allegheny River (2,702 cfs) and Blacklick Creek
(68 cfs), respectively, with their large differences in water discharge rate (Figure 6) and resulting dilution
capacities. Median chloride concentrations were significantly higher at the Blacklick downstream sites
(35.7 mg/L) compared to upstream (19.6 mg/L). Chloride concentrations at the Allegheny downstream
sites were also significantly higher but the magnitude of absolute difference (~4 mg/L) was smaller than
the Blacklick observations. The aggregate mean chloride concentration of 64.3 mg/L for the Blacklick
downstream sites reflects the relatively high chloride concentrations measured at site S03_B during
CWTF_B discharge days. The Blacklick sites also had much higher sulfate concentrations (median
downstream=157 mg/L) compared to the Allegheny sites (median downstream=8.97 mg/L), indicating
higher AMD contributions on the Blacklick. Iron and Mn were highest at the Blacklick upstream site due
to AMD, and Si was also elevated at this site due to the influence of runoff and rapid changes in river
discharge rate in the creek. Although Ca and Na were elevated in the CWTF and FGD source samples
and CWTFs discharge significant levels of Sr (Table 13), these species did not have large differences
between Blacklick upstream (median Ca=53.8 mg/L, median Na=23.3 mg/L, median Sr=0.341 mg/L) and
downstream sites (median Ca=48.8 mg/L, median Na=28.7 mg/L, median Sr=0.388 mg/L). High peak
concentrations relative to the mean were observed for most of the species indicating variability in natural
or industrial sources influencing the sites. The maximum bromide concentration for each sampling
domain was measured at site S03 closest to each respective CWTF discharge, with 4.36 mg/L and 0.28
mg/L observed on the Blacklick Creek and Allegheny River, respectively.
Twelve target elements were selected for the species summary and PMF analysis (bromide, chloride,
nitrate, sulfate, Ca, K, Mg, Na, Si, Sr, Fe, and Mn), because they were valid for almost all of the samples,
were present in the high bromide sources, and also captured sediment species. The Appendix Tables Al
and A2 have a summary of these species concentrations for each site as well as the number of valid
samples. Sequential samples were included in the summary, which accounts for the increased number of
samples at S03_A, S04_B, S05_A, and S05_B.
A summary of the PDWS intake (raw water) bromide, chloride, nitrate, and sulfate concentrations are
shown in Table 14 and the median bromide concentration increased 0.0468 mg/L from S05_A to S05_B
due to a combination of sources downstream of S05_A and the confluence of the Allegheny and
Kiskiminetas Rivers. The largest difference in concentrations between the intakes was found for sulfate
which increased from a median of 9.58 mg/L at S05_A to 82.6 mg/L at S05_B.
As expected the low river discharge period bromide concentrations presented in Table 14 are significantly
higher than the concentrations measured at the same sites during the spring (high river discharge; April 30
- May 14, 2012) with median concentration of 0.038 and 0.035 mg/L for S05_A and S05_B, respectively.
37
-------
Table 13. Average Br/Cl ratio and key species concentrations (mg/L) for measured source profiles.
Source
CWTF2_A
CWTF_B
CWTF_C
Cooling Towers
Treated FGD3
Demineralizer
Treated Coal Pile
Runoff
Treated Coal Ash
Wastewater
Treated Industrial
Wastewater
POTW4
Treated AMD6 and
Oil and Gas
Wastewater
CBM7
AMD
Oil Seep
Coal Mine Runoff
N1
23
23
1
5
3
1
2
2
3
4
1
2
1
1
1
Br/Cl
Ratio
0.0111
0.0097
0.0100
0.0030
0.0146
0.0001
0.0031
0.0247
0.0010
0.0003
0.0063
0.0077
0.0058
0.0119
0.0055
684
808
658
0.651
187
<0.006
0.169
2.48
0.259
0.04185
5.34
35.1
0.124
<0.06
1.76
9420
12100
10266
248
2780
27.9
136
384
396
34.3
684
284
92.9
65.7
75.3
n
61700
83200
66110
216
12800
82.6
54.0
101
254
138
844
4560
21.4
5.04
322
<16.9
<16.9
<16.9
28.5
183
4.23
10.9
11.4
235
98.5
<4.06
0.875
<0.016
9
<0.169
15.9
23800
31700
21900
162
844.0
2290
225
126
245
96.2
629
2290
10.8
7.66
536
218
577
487
809
2080
2100
902
1190
482
24.5
2420
358
469
7.36
787
n
395
1010
247
1.12
23.7
0.130
0.533
2.29
0.374
0.0999
6.68
8.80
0.555
0.312
0.606
1 for CWTFs, N is the number of samples collected during summer and fall campaigns; for all other source types,
N is the number samples collected at different facilities.
2 centralized waste treatment facility
3 flue gas desulfurization
4 publicly owned treatment works
5 one POTW had an average bromide concentration of 0.0418 mg/L, and the three other POTW shad bromide
concentrations below the bromide method detection limit of 0.0063 mg/L.
6 acid mine drainage
7 coal bed methane
38
-------
Table 14. PDWS intake low flow intake concentrations (mg/L).
Site
S05_A
S05_B
Statistic
mean
median
IQR1
maximum
mean
median
IQR
maximum
Bromide
0.0867
0.0852
0.0436
0.166
0.143
0.132
0.0455
0.205
Chloride
23.6
23.5
4.45
32.1
30.4
29.7
6.20
37.6
Nitrate
0.327
0.284
0.477
1.20
2.05
2.07
1.12
3.88
Sulfate
9.62
9.58
0.960
10.9
82.2
82.6
40.9
129
1 Interquartile range (15^-25^ percentile)
4.2 Source Profiles
Inorganic species compositions of sources must be known in order to identify the PMF factor profiles.
Using the same extraction and analysis techniques for both river samples and source samples allowed for
direct comparison of PMF-generated profiles and measured profiles. However, only CWTF and AMD
source profile information had been previously published (Ferrar et a/.2013; Warner et al. 2013).
Multiple source samples from facilities discharging to the surface waters in this study were critical for (1)
an evaluation of source variability and (2) calculation of an average profile that was representative of the
source contributions at the PDWS intakes.
In light of the voluntary diversion of the Marcellus waste away from CWTFs discharging to
surface waters ( PADEP 201 Ib; Wilson and Van Briesen, 2012), we characterized the 2012
wastewater treated by the CWTF_A and CWTF_B as to whether or not it originated from a
hydraulically fractured well. The characterizations were based on our review of the conventional
and unconventional wastewater annual reports of treated waste provided to the Pennsylvania
Department of Environmental Protection (PA DEP 2014a). The Pennsylvania Department of
Environmental Protection well waste classification is based on the formation that produces the oil
or gas. Conventional waste is derived from a formation above or below the Elk Sandstone.
Unconventional waste results from a geologic shale formation. Both conventional and
unconventional wells typically require stimulation by hydraulic fracturing although much greater
volumes of fluid are required for unconventional wells (PA DEP 2012). The total volume of
unconventional wastewater treated during the course of this study was low and CWTF_A did not
have any reported unconventional wastewater while CWTF_B had a total of 1,585 barrels.
Conventional wastewater was further classified as stimulated or hydraulically fractured, not
fractured, or unknown based on a comprehensive review of over 7,250 well file records in the
Pennsylvania Department of Environmental Protection Internet Record Imaging System (PA DEP
2014c). The total volumes of wastewater (drilling fluid waste, fracking fluid waste, produced
fluid) treated by the CWTFs during this study were 572,464 barrels and 419,673 from CWTF_A
and CWTF_B, respectively (Table 15). Less than 12% of the wastewater was classified as
unknown due to missing information in the well completion report. Of the total treated oil and
gas wastewater treated, 76% (CWTF_A) and 92% (CWTF_B) were determined to be from
hydraulically fractured wells. In 2012, the CWTFs also treated basic sediment (CWTF_A total
barrels = 110, CWTF_B total barrels = 5,808), drilling fluid (CWTF_A total barrels = 46,454.40,
39
-------
CWTF_B total barrels = 12,324), and servicing fluid (CWTF_A total barrels = 300, CWTF_B
total barrels = 707). During the study period, both CWTFs accumulated the oil and gas
wastewater from individual deliveries into large on-site storage tanks prior to batch treatment. As
a result, the wastewater from numerous individual wells was combined prior to treatment and
discharge. Accordingly, although these data suggest that there is a high probability that
hydraulically fractured waste water was being treated during the sampling period, we do not have
specific information as to the wastewater being discharged on days of sampling.
Table 15. CWTF wastewater volumes treated in 2012 (barrels).
CWTF_A
Hydraulically
Fractured
No
Unknown
Yes
All
Total
69,451
68,246
434,766
572,464
Percent
12%
12%
76%
100 %
CWTF_B
Total
16,097
19,144
384,431
419,673
Percent
4%
5%
92%
100 %
Table 13 provides a summary of the source sample concentrations for bromide, chloride, nitrate, sulfate,
Na, Ca, Sr, and Br/Cl ratio. The Br/Cl ratio can be used to identify brine or salt sources (Davis et al.
1998), and it was used in this study to compare the PMF and measured source profiles. Outfalls typically
capture wastewater from multiple processes as described in the NPDES permits, and the wastewater is
generally treated prior to discharge to meet discharge limits. Other water contaminant source types
include AMD from historic and ongoing mining activities, and natural oil seepage. Road salt has been
reported as a significant source of chloride (Kelly et al. 2010), but this source was not expected to have a
significant contribution, since this report focuses on summer and fall seasons when roadway deicing
materials are not typically applied. Furthermore, bromide concentrations in the roadway deicing
materials used in this region were extremely low (0.0033% and 0.0063% by weight) for two road salt
piles sampled near S04_B. Finally, samples were collected from two ground water-supplied PDWS
facilities to characterize the ground water that could contribute to the Allegheny River. One of the water
facilities was supplied by three wells, and the other was supplied by four. The mean concentrations were
0.241, 62.7, and 10.4 mg/L for bromide, chloride, and sulfate, respectively. Ground water can
hydraulically exchange with surface water, and this potential source of bromide was not identified in the
source apportionment analysis. This result was expected since background bromide concentrations in the
Allegheny have been reported to be low (States et al., 2013), and the upper and middle reaches of the
river flow over locally-derived river sediment overlying thick layers of glacial outwash (sand and gravel).
Separation of sources requires that sources have a difference in relative species compositions and that
their contribution varies at the multiple sampling sites. Two key sources of bromide and chloride were
CWTFs and FGDs, with FGDs having higher sulfate and nitrate concentrations, and CWTFs having
higher Sr concentrations. Both bromide and chloride are conserved in surface waters due to their
solubility (Davis et al. 1998). Sources with elevated sulfate concentrations included treated and untreated
AMD, CBM, treated coal ash and coal pile runoff, active coal mine runoff, and cooling towers. CBM
wastewater and untreated AMD also had high Fe concentrations. The oil seepage, however, had elevated
40
-------
Fe and low sulfate. Treated AMD was low in Fe (0.0464 mg/L) compared to untreated AMD (36.0
mg/L). The industrial metals treated wastewater source had more unique species with elevated F (5.09
mg/L) and Mo (1.45 mg/L), and low Fe (0.06 mg/L) due to the type of industries. POTWs had a nitrate
concentration of 98.5 mg/L and only one of the POTWs had measurable bromide concentrations (the
facility located upstream of S01_A). The two samples collected from that facility had similar bromide
concentrations (0.406 and 0.431 mg/L), which were low compared to a nearby CWTF (657.6 mg/L).
These sources combine in the surface water after they are discharged (Figure 7), and the ratios of the
measured profiles species along with the known outfall locations were critical for evaluating the PMF
results.
The two highest sources of bromide were the CWTF (average of 746 mg/L) and FGD (187 mg/L). The
Br/Cl ratios for these two sources were different, with average ratios of 0.0104 and 0.0146 for CWTF (A
and B only) and FGD, respectively. Upper Pennsylvanian and Monongahela formation bituminous coal
(e.g., Pittsburgh #8) that is currently being burned by coal-fired power plants in this region of the U.S. has
naturally high levels of bromine (Seere and Lee, 2009). The FGD had higher concentrations of nitrate
and sulfate compared to the CWTF. However, other sources contribute to the Blacklick sampling sites
including nitrate from POTWs and sulfate from AMD (Figure 4, Figure 7). The CWTF Br/Cl ratio was
similar to the ratio of 0.0080 reported by Warner et al. (2013) and the ratio of 0.0091 reported by Hladik
et al. (2014). Discharge schedules for CWTF_A were between 06:00 and 13:00 EST Monday through
Friday, while CWTF_B discharges were more sporadic and ranged from nine to 39 hours based on river
conductivity measurements. The FGD discharges were continuous for two of the three generating
stations; the other had a batch treatment process.
Comparison of coal-fired generating discharges to other studies was limited due to the lack of published
data. Wilson et al. (2013) reported that there was no statistically significant difference in total dissolved
solids (TDS) between coal-fired power plant wastewater and coal mine discharge. The similar
concentrations in the sources indicate either that they were collected at non-FGD power plants, or that the
FGD discharge was not captured in the sample. Our study results show that treated coal pile waste had
much lower concentrations of bromide and chloride (bromide = 0.169 mg/L, chloride = 136 mg/L)
compared to FGD wastewater (bromide =187 mg/L, chloride = 2,780 mg/L). In addition, a coal mine
discharge sample was collected as part of this study, and the conductivity was 8,273 |o,S/cm, compared to
29,725 nS/cm for one of the FGDs.
4.3 PMF Receptor Modeling Results
River sampling data from multiple sites were used in the PMF analysis to provide a range of source
contributions and to take advantage of the study design, which focused on selecting sites that were
upstream and downstream of CWTFs and other sources. Species concentration data for all river sampling
sites were initially combined together in the PMF analysis; however, the magnitude of the CWTF,
POTW, and AMD contributions for the Blacklick sites dominated the source contributions at the
Allegheny sites. This was due to the low discharge flow rate of the Blacklick Creek (68 cfs) compared to
the Allegheny River (2,702 cfs), and the substantial AMD impairment of the Blacklick Creek (Sams and
Beer 2000).
The Allegheny and Blacklick sampling site data were then evaluated separately by PMF to determine the
source contributions for each set of sample sites, excluding S05_B, which was on the Allegheny River.
41
-------
Lastly, in order to understand the source contributions to the PDWS intakes, PMF was used to evaluate
the combined Allegheny and Blacklick downstream sites and PDWS samples (S05_A and S05_B).
Twelve species were included in all PMF analyses: bromide, chloride, nitrate, sulfate, Ca, K, Mg, Na, Si,
Sr, Fe, and Mn. Tables with the PMF settings and Q values for the three analyses are provided in the
Appendix Tables Bl to B3.
PMF analysis of species concentrations for the Allegheny sites (S01_A, S03_A, S04_A, and S05_A)
found four factors, identified as CWTF, sediments, POTW, and AMD based on the source profiles in
Table 13 and Br/Cl ratios. Measured source profiles were also used to determine if any of the PMF
factors contained multiple sources. Bromide in the PMF POTW factor exceeded the measured source
profile, indicating that PMF was not able to definitively separate the source based only on the
measurement data. When this factor was constrained using the Br/Cl ratio in Table 13, the bromide level
decreased in the POTW source and Q increased 0.06%, indicating a small change from the optimum PMF
solution. This constrained result was used, and the resultant median bromide concentration at the PDWS
intake (S05_A) was found to be 0.0553 mg/L from the CWTF, with no contributions from the POTW or
sediments. The PMF results also showed downstream transport of bromide from a CWTF that
contributed 0.0743 mg/L to S01_A. Additional bromide was contributed by CWTF_A resulting in a
combined contribution of 0.0803 mg/L at S03_A. Downstream transport was also observed with nitrate
from POTWs. S01_A had anitrate concentration of 0.538 mg/L from multiple upstream POTWs.
Additional nitrate was contributed by a POTW upstream of site S03_A, resulting in 0.454 mg/L at that
site. AMD sulfate contributions increased from S01_A (2.19 mg/L) to S05_A (5.20 mg/L,) which
follows the increase in bituminous coal mining activity near the downstream sites (PA DCNR 2014).
Three factors were found in the PMF analysis of river sample data for the Blacklick sites (S01_B, S03_B,
and S04_B) which were identified as CWTF, POTW, and AMD. The PMF POTW profile had elevated
bromide, which was partially removed with the same constraint as used for the Allegheny analysis;
however, the bromide remained high in the profile (0.0626 mg/L), and the AMD profile contained
bromide (0.0349 mg/L) as well. PMF results demonstrated that the CWTF_B discharge had a large
contribution to the bromide source concentrations at the immediate downstream site (S03_B). The
median was determined to be 0.231 mg/L with an IQR of 0.817 mg/L and a maximum of 4.55 mg/L.
These high and variable concentrations were due to the low flow in the Blacklick Creek which did not
significantly dilute the CWTF discharge. An FGD source was not identified in the Blacklick site analysis
even though the FGD source had high bromide levels and two generating stations with FGDs were
discharging into the Blacklick domain. The PMF results were investigated by comparing discharge rates
using the Pennsylvania Department of Environmental Protection Discharge Monitoring Records and the
CWTF and FGD profile concentrations in Table 13 (PA DEP 2014b). Average daily discharge of
bromide from two FGD discharges (396 kg/day) was higher than CWTF_B (161 kg/day), and the
bromide contributed from these discharges and from the AMD treatment and CWTF facility was
quantified at the sampling site after the Conemaugh Dam (S04_B). Based on the mass discharge analysis
and the CWTF profile contributing the majority of the bromide, the CWTF source was capturing the
combined CWTF and FGD sources.
Based on the results from each river, the PMF analysis of the combined Allegheny and Blacklick
downstream sites (S03_A, S04_A, S05_A, S04_B, and S05_B, but excluding site S03_B), was used to
evaluate the FGD and CWTF contributions on the two PDWS intakes. This combination of sites also
42
-------
provides a focus on the sources contributing to the PDWS intakes, since it only uses the Blacklick sites
downstream of the Conemaugh Dam (S04_B and S05_B; Figure 7). The combined analysis was the
focus of this report, and the steps used to interpret the PMF are shown below.
1. Select the number of factors based on the number of expected sources and evaluate increasing or
decreasing the number of factors.
2. Evaluate the observed vs. predicted concentrations to evaluate if the residual (observed -
predicted) is similar across sites and the regression statistics.
3. Evaluate the stability of the solution using the displacement algorithm and reduce or increase the
number of factors if the solution is not stable. Assess the minimum to maximum uncertainty
range provided by the displacement algorithm.
4. Compare the PMF factors to measured profiles and identify the sources using species ratios.
5. Evaluate the source contributions at each sampling site to determine if the source contributions
increase downstream of outfall or are not present at a site, such as the Allegheny sites (S01_A to
S05_A) that were not downstream of a FGD.
6. Evaluate different site combinations in a multiple site dataset and evaluate if a PMF profile
represents multiple sources.
The observed vs. predicted bromide concentration plot for each site shows excellent agreement, as shown
in Figure 8. Data from each of the sampling sites are displayed by site and sorted by date. The observed
versus predicted regression statistics for all of the species in Table 16 show good agreement, except for Fe
and Mn which were categorized as weak to reduce their influence on the results.
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Bromide
concentration
(mg/L)
PMF results-
each site's
contributions
sorted by date
S04_B S05_B S03_A S04_A S05_A
Figure 8. EPA PMF observed (blue) vs. predicted (red) time series plot for bromide by site (mg/L).
43
-------
Table 16. Positive Matrix Factorization measured versus predicted regression statistics.
Species
BMC1
CIJC
NOsJC
SCHJC
Ca_ICP-OES2
KJCP-OES
MgJCP-OES
Na_ICP-OES
SiJCP-OES
SrJCP-OES
Fe_HR-ICPMS3
Mn_HR-ICPMS
Intercept
0.002
-0.036
-0.023
-0.085
1.338
0.064
0.112
0.052
0.035
0.006
0.132
0.042
Slope
0.990
1.001
0.925
1.002
0.943
0.957
0.980
0.994
0.956
0.944
0.193
0.282
Coefficient of
Determination
1.00
1.00
0.90
1.00
0.99
0.98
1.00
0.99
1.00
0.98
0.44
0.55
1 Ion chromatography
2 Inductively coupled plasma-optical emission spectrometry
3 High-resolution-inductively coupled plasma mass spectrometry
The stability of the solution was first confirmed by evaluating the displacement output. A stable solution
has zero swaps for all of the factors at the lowest displacement level. PMF factors were identified by
comparing them to the measured profiles and the Br/Cl ratios. PMF profiles and Br/Cl ratios are shown
for high bromide sources in Table 17 and low bromide sources in Table 18. Sources with low bromide
concentrations (Table 17) had minimum displacement error estimates of zero, indicating they are not
significant bromide sources. The AMD profile in Table 18 is from the Blacklick site PMF analysis since
it was able to separate an AMD source. PMF source profiles represent the source fingerprint in the
diluted river sample, so the concentrations are lower than the measured profiles. In addition, a sediment
source type with Fe, Si, and Mn was found in PMF analyses.
The PMF profiles are displayed in Figure 9. Blue bars show the PMF profile species concentrations
(mg/L), and red boxes show the percent of the species associated with the source. Normalized
contribution (the average of the normalized contributions for a source = 1) time series by site are shown
in Figure 10, and the contributions were evaluated to determine if the increase in the PMF contribution
was consistent with known outfall locations. Measured source profiles, PMF profiles, and contribution
time series were used to identify the four source types:
1. CWTF factor, characterized by high bromide and chloride and a Br/Cl ratio of 0.0066, compared
to the measured profile average of 0.0104. In addition, it had low sulfate and nitrate
concentrations. Peak bromide concentrations were observed at S04_B, which is downstream of
two CWTFs.
2. FGD + AMD factor, characterized by: high bromide, chloride, sulfate, and nitrate; and a Br/Cl
ratio 0.016, which was similar to the FGD measured profile average of 0.0146 but higher than the
AMD ratio of 0.0058. This factor may also be impacted by coal ash wastewater, which has a
Br/Cl ratio of 0.0247. Generating stations, plus AMD from historic mining and coal mining.
44
-------
Table 17. Positive Matrix Factorization high-bromide source type profiles and minimum and maximum
estimates from the displacement algorithm (mg/L).
Species
Br
Cl
N03
S04
Ca
K
Mg
Na
Si
Sr
Fe
Mn
Profile
0.0894
13.5
0
6.23
7.00
0.450
1.41
6.14
0
0.0485
0.0166
0.00226
1 Centralized waste treatment
2 Flue gas desulfurization and
Table 18.
estimates
Positive
from the
Combined CWTF1
Br/CI = 0.007
Min
0.0745
11.0
0
2.25
5.57
0.344
1.16
4.99
0
0.0392
0.00928
0.000656
facility
acid mine drainage
Combined FGD + AMD 2
Br/CI = 0.016 |
Max
0.135
21.1
0.0330
9.93
11.8
0.797
2.42
9.97
0.306
0.0766
0.0613
0.0209
Matrix Factorization low -bromide source
displacement algorithm (mg/L).
Profile
0.068
4.24
0.662
50.8
7.01
0.293
3.23
4.21
0.234
0.0822
0.00491
0.00991
type profiles and
Combined POTW1 Combined Sediment
Species
Br
Cl
N03
S04
Ca
K
Mg
Na
Si
Sr
Fe
Mn
Br/CI =
Profile Min
0
25.6
46.4
12.8
34.7
40.0
27.9
28.6
18.3
15.3
27.7
9.95
0
17.2
28.0
0
25.7
29.2
17.0
20.4
0
6.91
1.37
0
0
Max Profile
27.6 0.00716
52.1 5.82
84.8 0
36.3 2.07
53.2 5.12
63.0 0.379
43.4 1.14
46.8 3.46
38.7 0.785
30.7 0.0219
39.7 0.112
26.7 0.0530
Br/CI = 0
Min
0
1.59
0
0
2.26
0.148
0.751
1.41
0.626
0.0139
0.0839
0.0423
Max Profile
0.0456 0.0349
13.9 8.45
0.139 0.246
11.2 110
10.4 24.2
0.767 1.59
2.28 6.14
7.44 10.4
1.16 2.18
0.0525 0.219
0.174 0.257
0.0701 0.481
Min
0.0303
0
0.189
40.2
1.85
0
1.94
1.34
0.00489
0.0605
0
0
Max
0.0790
6.16
0.887
64.4
9.20
0.419
4.06
5.319
0.429
0.0999
0.0246
0.0161
minimum and maximum
BlacklickAMD2
Br/CI
= 0.0041
Min
0.00
3.453
0.127
101
21.3
1.43
5.40
7.88
2.05
0.139
0.243
0.454
Max
0.0444
9.69
0.607
127
27.7
1.84
7.66
11.7
2.27
0.227
0.262
0.530
1 Publicly owned treatment works
2 Acid mine drainage
45
-------
runoff contributed to all of the sites on the Blacklick to Kiskiminetas, and the high sulfate on this
source may be due to a combination of the FGD and AMD sources.
3. Sediment factor, characterized by high concentrations of Si, Fe, and Mn. The variability in this
source is associated with changes in the Allegheny River flow rate after rainfall.
4. POTW factor, characterized by high concentrations of nitrate, low bromide, and a Br/Cl ratio of
zero, compared to the measured profile ratio of 0.0003. S04_B and the sites on the Allegheny had
POTW contributions.
The PMF normalized contributions are multiplied by the PMF profiles to generate the sample source
contributions. The POTW source in Figure 9 has a low bromide contribution with a concentration less
than 0.001 mg/L in contrast to the bromide contributions for the CWTF, FGD + AMD, and sediment
sources which are shown in Figure 11. The concentration range for each of the time series plots is
different with the CWTF and FGD + AMD plots having similar ranges of approximately 0.4 mg/L and
the sediment range is less than 0.02 mg/L. The highest combined bromide concentrations were observed
at S04_B after the Conemaugh River Dam which is downstream of CWTF and FGD discharges. The
CWTF and FGD + AMD bromide source contributions are similar for S05_B, while S05_A only has
contributions from the CWTF source. The sediment source contributions are more uniform across all of
the sampling sites.
PMF used the river sampling data to solve the chemical mass balance between measured species
concentrations and source profiles by decomposing the speciated sample data into two matrices: factor or
source contributions and profiles (Hopke 1991). Since the PMF profiles are based on the river samples,
and the source concentrations have been diluted by the time they reach the sampling location, ratios are an
ideal way to compare the measured profiles and the PMF source type profiles. Comparing the
concentration patterns also helps confirm the sources, as shown in Figure 12. For example, the measured
CWTF and PMF profiles have higher concentrations of chloride and sodium.
46
-------
Base Factor Profiles
Legend: % of Species
CZ1 Cone of Species
10
10°
10-'
10'2
10''
10"
10*
10°
s 1
ts O
10
10"4
10'
10°
10':
10"'
10"*
10'
10°
10''
10-
10'3
CWTF
FGD + AMD
Sediment
POTW
100
80
60
40
20
0
100
80
60
40
20
0
100
0
100
80
60 -n
40
20
? »
Figure 9. EPA PMF source profile plot showing the species concentration (left axis) and the
percentage of species associated with each source (right axis). Blue bars show the PMF profile
species concentrations (mg/L), and red boxes show the percent of the species associated with the
source. The two PMF sources with significant bromide were CWTF and FGD+AMD sources with
each source contributina over 40% of the bromide.
47
-------
Base Factor Contributions
CWTF
FGD + AMD
PMF results
each site's
contributions
sorted by date
Sediment
POTW
S04 B
805 B
S03 A
S04 A
SOS A
Figure 10. EPA PMF time series plot of normalized factor contributions (average contribution = 1
for each factor) by site and date. The contribution axis scale was set to show the variation in the
source contribution and each plot has a different scale.
48
-------
Base Factor Concentrations
0.4 -
0.00
CWTF
FGD + AMD
Sediment
PMF results-
each site's
contributions
sorted by date
0.4
03
0.2
01
00
0.3
0.2
0.1
0 0
0.01
0.00
S04 B
SOS B
SOS A
S04 A
SOS A
Figure 11. EPA PMF source contribution plot for the 3 sources with elevated bromide concentrations.
49
-------
100000
^
^ * -^ ,s
e' /' /' /'
Br BCI BN03 BS04 Na «Ca
Figure 12. Comparison of the measured (_M) and Positive Matrix Factorization (_P) profiles.
Bromide source contributions for each sampling site are shown in Figure 13 and Table 19. The boxplots
show the daily variation in each source contribution. S04_B has the largest range in concentrations
because the CWTF_B had variable daily discharge levels. All of the bromide at the S05_A PDWS intake
(median = 0.079 mg/L) originated from upstream CWTFs. A combination of discharges from CWTFs
and FGD + AMD contributed to the concentrations at the S05_B PDWS intake, resulting in median
bromide contributions of 0.054 mg/L from the CWTFs (37%) and 0.086 mg/L from FGDs (59%).
The relative CWTF and FGD+AMD source contributions at S05_A and S05_B were further evaluated by
plotting the distribution of the daily sample contribution to the total bromide (Figure 14). The S05_A
distributions show that the CWTF contributed over 80% of the bromide in the majority of the samples and
the FGD+AMD contribution was 10% or less (only one sample had a 5% contribution and the rest had
zero). For site S05_B the distribution shows that majority of daily bromide contributions ranges between
30 - 60% for CWTF, and 40 - 70% for FGD + AMD.
50
-------
j" 0.25 -
| 0.20-
« 0.15 -
| 0.10 -
2 0.05-
0.00 -
0.30 -
j" 0 25 -
|* 0.20 -
0> 015-
o
'£ 0.10-
2 005-
0 00 -
0.30 -
j" 0 25 -
f 0.20 -
a, 0.15-
E 010-
2 0.05-
0.00 -
0.4 -j
=3, 0.3-
T 0.2 -
;g
0 0.1 -
a
0.0 -
0.4 -[
j"
"Si 03 -
^T 0.2 -
O
0 01 -
m
0.0 -
S03_A
T
+ t
CWTF FGD + AMD POTW Sediments
S04_A
T
+ :
CWTF FGD + AMD POTW Sediments
S05_A
CWTF FGD + AMD POTW Sediments
CflA R
_L T
V
CWTF FGD + AMD POTW Sediments
S05_B
^_
CWTF FGD + AMD POTW Sediments
Source
Figure 13. Positive Matrix Factorization (PMF) combined analysis (Allegheny and Blacklick) bromide
concentrations (mg/L) for PMF source types by river sampling site. The figure shows the elevated
contributions associated with centralized waste treatment facilities (CWTFs) and flue gas desulfurization
scrubbers (FGDs). AMD = acid mine drainage. POTW = publicly owned treatment works.
51
-------
0 J
30
25
20 -
15 -
10 -
5 -
0
805 AFGD+AMD
20 40 60 80 100 0 20 40 60 80 100
Percent Contribution Percent Contribution
10
6 -
2 -
805 BFGD+AMD
20 40 60 80
Percent Contribution
100
20 40 60 80
Percent Contribution
100
Figure 14. Distribution of CWTF and FGD + AMD bromide source contributions (%) to S05_B.
Median concentrations of bromide, chloride, nitrate, and sulfate contributed by each source type and
sampling site are provided for the three PMF analyses in Table 19, Table 20, Table 21, and Table 22,
respectively. The bromide concentration for the sediment source was 0.0 mg/L for the Allegheny PMF
analysis, and the combined sites had median bromide concentrations that ranged from 0.0054 to 0.0097
mg/L (Table 19). The sediment source most likely represents a background concentration; no sediment
samples were collected as part of this study.
52
-------
Table 19. Positive Matrix Factorization (PMF) bromide concentrations by PMF analysis and sampling
site (mg/L).
Allegheny
Site
S01_A
S03_A
S04_A
S05_A
Blacklick
Site
S01_B
S03_B
S04_B
Combined
Site
S03_A
S04_A
S05_A
S04_B
S05_B
CWTF1
N
30
26
29
26
N
28
29
26
N
26
29
26
26
26
Median
0.074
0.080
0.060
0.055
CWTF
Median
0.0004
0.23
0.280
CWTF
Median
0.088
0.082
0.079
0.13
0.054
IQR"
0.044
0.037
0.037
0.041
IQR
0.045
0.82
0.15
AMD2
Median
0.010
0.012
0.025
0.025
AMD
Median
0.060
0.043
0.0048
IQR
0.0090
0.0064
0.012
0.0075
IQR
0.049
0.0530
0.0050
FGD5 + AMD
IQR
0.031
0.027
0.040
0.18
0.0320
Median
0.0
0.0
0.0
0.28
0.086
IQR
0.0
0.0
0.0
0.065
0.053
POTW3
Median
0.0
0.0
0.0
0.0
POTW
Median
Sediments
IQR
0.0
0.0
0.0
0.0
IQR
Median
0.0
0.0
0.0
0.0
IQR
0.0
0.0
0.0
0.0
0.043 0.0064
0.033
0.123
POTW
Median
0.0
0.0
0.0
0.0
0.0
0.018
0.038
Sediments
IQR
0.0
0.0
0.0
0.0
0.0
Median
0.0071
0.0053
0.0097
0.0059
0.0054
IQR
0.0046
0.0076
0.0081
0.0042
0.0033
1 Centralized waste treatment facility
2 Acid mine drainage
3 Publicly owned treatment works
4 Interquartile range
' Flue gas desulfurization
53
-------
Table 20. Positive Matrix Factorization (PMF) chloride concentrations by PMF analysis and sampling site
(mg/L).
Allegheny
Site
S01_A
S03_A
S04_A
S05_A
Blacklick
Site
S01_B
S03_B
S04_B
Combined
Site
S03_A
S04_A
S05_A
S04_B
S05_B
N
30
26
29
26
N
28
29
26
N
26
29
26
26
26
CWTF1
Median
9.03
9.75
7.27
6.72
CWTF
Median
0.0505
31.2
37.8
CWTF
Median
13.2
12.4
11.9
19.2
8.14
IQR"
5.30
4.45
4.46
4.99
IQR
6.11
111
20.5
AMD2
Median
4.17
4.94
9.93
9.91
AMD
Median
14.6
10.4
1.17
IQR
3.63
2.58
4.77
3.04
IQR
11.9
12.8
1.22
FGD5 + AMD
IQR
4.6
4.1
6.0
27.7
4.81
Median
0.0
0.0
0.0
17.6
5.31
IQR
0.0
0.0
0.0
4.1
3.3
POTW3
Median
5.01
4.23
1.81
1.74
POTW
Median
6.41
5.01
18.5
POTW
Median
3.53
5.31
4.56
15.8
12.0
Sediments
IQR
4.23
4.20
2.52
3.34
IQR
0.957
2.63
5.74
Median
2.84
4.10
2.37
4.92
IQR
4.81
4.93
5.38
4.44
Sediments
IQR
3.63
5.06
5.57
5.6
4.02
Median
5.77
4.34
7.87
4.80
4.39
IQR
3.70
6.15
6.55
3.45
2.67
1 Centralized waste treatment facility
2 Acid mine drainage
3 Publicly owned treatment works
4 Interquartile range
5 Flue gas desulfurization
The highest concentrations of chloride were observed on the Blacklick, with contributions from the
CWTF, FGD + AMD, and POTW sources (Table 20). Measured chloride concentrations were higher in
both CWTFs and the FGD, and a large increase was found at both S03_B and S04_B. The increase in
chloride from the CWTFs on the Allegheny was much smaller due to higher river discharge and resulting
increase in dilution capacity.
High concentrations of nitrate were found on the AMD and CWTF sources, indicating that the elevated
concentrations measured at S03_B impacted the ability of PMF to separate the nitrate source (Table 21).
The combined analysis had nitrate contributions from both the FGD + AMD source and the POTW. The
measured FGD profile had significant nitrate, and the generating stations have on-site POTWs which also
discharge nitrate. The combination of these sources could be the reason for the nitrate contribution.
PMF sulfate concentrations in Table 22 show that including S03_B in the Blacklick site analysis impacted
PMF's ability to apportion sulfate only to the AMD source, since the POTW source had a significant
sulfate contribution at S01_B, S03_B and S04_B. The measured profiles had high sulfate concentrations
for both the FGD and AMD profiles, and low sulfate for the POTW profile. Furthermore, the discharges
54
-------
from these sources combine at S04_B, and have a large impact on sulfate at the PDWS intake after the
confluence of the Kiskiminetas and the Allegheny (S05_B).
Table 21. Positive Matrix Factorization (PMF) nitrate concentrations by PMF analysis and sampling site
(mg/L).
Allegheny
Site
S01_A
S03_A
S04_A
S05_A
Blacklick
Site
S01_B
S03_B
S04_B
Combined
Site
S03_A
S04_A
S05_A
S04_B
S05_B
CWTF
N
30
26
29
26
N
28
29
26
N
26
29
26
26
26
Median
0.0008
0.0009
0.0007
0.0006
CWTF
Median
0.0002
0.11
0.13
CWTF
Median
0.0
0.0
0.0
0.0
0.0
IQR4
0.0005
0.0004
0.0004
0.0004
IQR
0.022
0.39
0.073
AMD*
Median
0.0
0.0
0.0
0.0
AMD
Median
0.43
0.30
0.034
IQR
0.0
0.0
0.0
0.0
IQR
0.35
0.37
0.036
FGD5 + AMD
IQR
0.0
0.0
0.0
0.0
0.0
Median
0.0
0.0
0.0
2.7
0.83
IQR
0.0
0.0
0.0
0.63
0.51
POTW
Median
0.54
0.45
0.19
0.19
POTW
Median
1.3
1.0
3.7
POTW
Median
0.25
0.38
0.32
1.1
0.85
Sediments
IQR
0.45
0.45
0.27
0.36
IQR
0.19
0.53
1.2
Median
0.0
0.0
0.0
0.0
IQR
0.0
0.0
0.0
0.0
Sediments
IQR
0.26
0.36
0.40
0.40
0.29
Median
0.0
0.0
0.0
0.0
0.0
IQR
0.0
0.0
0.0
0.0
0.0
1 Centralized waste treatment facility
2 Acid mine drainage
3 Publicly owned treatment works
4 Interquartile range
' Flue gas desulfurization
55
-------
Table 22. Positive Matrix Factorization (PMF) sulfate concentrations by PMF analysis and sampling site
(mg/L).
Allegheny
Site
S01_A
S03_A
S04_A
S05_A
Blacklick
Site
S01_B
S03_B
S04_B
Combined
Site
S03_A
S04_A
S05_A
S04_B
S05_B
CWTF1
N
30
26
29
26
N
28
29
26
N
26
29
26
26
26
Median
0.0072
0.0078
0.0058
0.0054
Median
0.024
15
18
Median
6.1
5.7
5.5
8.9
3.8
IQR"
0.0042
0.0036
0.0036
0.0040
CWTF
IQR
2.9
53
9.8
CWTF
IQR
2.2
1.9
2.8
13
2.2
AMD2
Median
2.2
2.6
5.2
5.2
AMD
Median
190
136
15
IQR
1.9
1.4
2.5
1.6
IQR
155
167
16
FGD5 + AMD
Median
0.0
0.0
0.0
210
64
IQR
0.0
0.0
0.0
48
39
POTW3
Median
3.3
2.7
1.2
1.1
POTW
Median
71
55
204
POTW
Median
3.8
5.7
4.9
17
13
Sediments
IQR
2.7
2.7
1.6
2.2
IQR
11
29
63
Median
1.8
2.7
1.5
3.2
IQR
3.1
3.2
3.5
2.9
Sediments
IQR
3.9
5.4
6.0
6.0
4.3
Median
2.1
1.6
2.8
1.7
1.6
IQR
1.3
2.2
2.3
1.2
0.95
1 Centralized waste treatment facility
2 Acid mine drainage
3 Publicly owned treatment works
4 Interquartile range
5 Flue gas desulfurization
4.4 PMF Bromide Sensitivity and Hybrid Analysis
PMF results are based solely on the measurement data included in the model. Therefore adding or
removing data from a sampling site(s) may materially impact the apportionment results. The sensitivity
of the numerical receptor modeling results was investigated by including or removing the sampling sites
on the Blacklick Creek (S01_B and S03_B). These sites had a high contribution from AMD, and the
CWTF source contributions at S03_B were very high due to the low dilution capacity from the creek.
The PMF results for S03_B also demonstrated that species were not being conserved due to the reaction
of CWTF source with the AMD in the creek. A PMF analysis was conducted with all sites, including
S01_B and S03_B, during low river discharge and pH conditions (pH range = 4.42 - 6.83). The analysis
extracted both a brine source (bromide, chloride, sodium) and a source composed of Fe and Mn. This
additional PMF source was hypothesized to be a precipitate composed of Fe, Mn, and Ba formed when
the higher pH CWTF discharge (pH range = 9.47 - 9.78) reacted in the AMD impacted Blacklick Creek
(Lee et al. 2002). This precipitate would tend to settle in the river/reservoir sediment rather than being
56
-------
transported downstream of the Conemaugh Dam and would change the S01_B creek composition
downstream of the CWTF_B discharge. A sensitivity analysis evaluated the impacts of the following
combinations of sites included in the PMF analysis:
1. All sites (S01_A, S03_A, S04_A, S05_A, S01_B, S03_B, S04_B, S05_B).
2. All sites except site S03_B which had a high CWTF impact due to the relatively low flow
(S01_A, S03_A, S04_A, S05_A, S01_B, S04_B, S05_B).
3. All Blacklick sites downstream from the Conemaugh Dam and Allegheny sites (S01_A, S03_A,
S04_A, S05_A, S04_B, S05_B).
The results of the PMF sensitivity analysis are summarized in Table 23 and the results from the combined
analysis are shown at the bottom of the table.
Table 23. PMF sensitivity analysis results (mg/L).
All Sites
PMF Profile Br/CI ratio
S05_A Bromide
S05_B Bromide
AMD Estimate
(No S03_B)
PMF Profile Br/CI ratio
S05_A Bromide
S05_B Bromide
No sampling sites on Blacklick
Creek (No S03_B, S01_B)
PMF Profile Br/CI ratio
S05_A Bromide
S05_B Bromide
Combined Analysis
Source
AMD + Sediments
0.0023
0.0094
Sediments
01
0
0
CWTF +FGD
0.0075
0.076
0.13
POTW
0.0031
0.00064
0.0091
CWTF +FGD
0.0066
0.080
0.13
POTW
0
0
0
Sediment
0
0
0
Sediment
0.0021
0.019
0.010
POTW
0
0
0
CWTF
0.0062
0.071
0.051
AMD
0
0.0012
0.013
FGD + AMD
0.018
0
0.085
(No S03_B, S01_B, S01_A)
PMF Profile Br/CI ratio
S05_A Bromide
S05_B Bromide
CWTF
0.0066
0.079
0.054
FGD +AMD
0.016
0
0.086
Sediment
0.0012
0.0097
0.0054
POTW
0
0
0
The measured profile Br/CI ratios and known discharge locations were used to evaluate each of the
solutions presented in Table 23. CWTF ratios were similar to the measured profiles (0.0097 to 0.011);
FGD + AMD ratios were similar to the FGD ratio (0.0146) and not AMD (0.0058). Including or not
including Blacklick Creek sampling data provided different results.
Although the mass discharge rates described in Section 4.3 indicated that the FGD discharges contributed
a significant bromide contribution, only the third PMF modeling scenario (no sampling sites on Blacklick
Creek included) and the combined analysis identified a separate FGD source. A hybrid analysis using
results from these two PMF analyses was conducted to discriminate the FGD and AMD contributions,
57
-------
since these two sources were initially combined by PMF. The PMF modeling results for the second
scenario presented in Table 23 (excluded site S03_B) were able to separate bromide into a combined
CWTF and FGD source and an AMD source (high sulfate). The AMD contribution from this analysis
was subtracted off the combined analysis site source contributions to provide estimated FGD and AMD
bromide concentrations as shown in Figure 15.
Figure 16 shows the large bromide source contribution difference downstream of the Kiskiminetas River
CD
0.3
0.25
0.15
0.1
0.05
S
3
e
o
CO
£
O
Site
FGD - AMD
AMD
FGD
Figure 15. Subtraction of AMD bromide from the FGD
estimated source contributions.
AMD source to provide AMD and FGD
and Allegheny River confluence. The impact of the source discharges clearly increased the Allegheny's
mean bromide concentration, from sites S05_A to S05_B, by 0.056 mg/L. After adjusting for the
bromide from AMD at S05_B (0.013 mg/L), the FGD contribution decreased from 59% to 50%. These
results represent the median percent contribution, which varies on a daily basis due to changes in
discharge volumes (e.g., CWTFs only discharging on weekdays).
58
-------
Crooked Creek |
D'C°a/%,
Generating Station*
Two Lick Creek
POTW, Generating Station*
POTWs, FGD,
Generating Station*
Figure 16. Hybrid PMF source contributions of bromide by sampling site and facility discharges.
59
-------
KEY FINDINGS
5. SUMMARY AND CONCLUSIONS
This research applied a technique referred to as "source apportionment" to quantify source contributions
for a number of common discharge sources. Source measurements were collected to provide reference
information for apportioning contaminant sources in the Alleghany watershed, including from centralized
wastewater treatment facilities (CWTFs) that treat wastes including oil and gas wastewater; coal-fired
power plants with and without flue-gas desulfurization (FGD); industrial manufacturing facilities;
municipal and industrial
wastewater treatment plants;
active coal mine runoff; and
acid mine drainage (AMD).
The study investigated the
sources of inorganic species
such as bromides and other
anions contributing to
contaminants at PDWS intakes
on the Allegheny River during
low river discharge conditions
in the summer and fall of
2012. The study included: (i)
river sampling from multiple
sites and source profile
collection; (ii) river and source
sample analysis with
conductivity, 1C, ICP-OES,
and HR-ICPMS; and (iii)
receptor modeling of
measurement data. Results using the study design and methods described in this report showed that
during the period of this study, the CWTFs that treated hydraulically fractured oil and gas wastewater,
along with other wastes, were a significant source of bromide at both PDWS intakes. Multiple sources
contributed chloride, nitrate, and sulfate, including CWTFs, FGDs, AMD, and POTWs to the
downstream PDWS intakes.
5.1 River Measurements
River contaminant concentrations at the Allegheny and Blacklick sites differed significantly. While the
upstream bromide concentrations were similar for the Allegheny and Blacklick sites, downstream
concentrations were much higher on the Blacklick. The mean bromide concentration at the PDWS intake
located 98 river kilometers upstream from the Allegheny/Kiskiminetas confluence was 0.0867 mg/L
(S05_A); while it was 0.143 mg/L (S05_B) just downstream of the confluence. The source
apportionment analysis was used to investigate the sources associated with the observed increase in
bromide between the two intakes, in particular to understand the possible relative contribution from
CWTFs that treat waste from oil and gas operations - the majority of which is hydraulic fracturing
wastewater.
The results demonstrate that the 2 public drinking water
intakes (PDWS) studied are impacted by multiple sources
contributing various inorganic species, including centralized
wastewater treatment facilities (CWTFs), power generating
stations, and acid mine drainage (AMD).
Source measurements provide a signature or profile for
numerous bromide sources.
The predominate sources of bromide at the 2 public drinking
water intakes studied were wastewaters discharged from
CWTFs and coal-fired power plants with flue-gas
desulfurization (FGD). CWTFs contributed nearly all the
bromide at 1 intake, while both CWTFs and FGDs
contributed to bromide levels at the second intake.
Publically owned wastewater treatment plants and acid mine
drainage were the predominate sources of nitrate and sulfate.
60
-------
5.2 Measured Source Profiles
The measured profiles showed differences in the source Br/Cl ratios (Figure 17). CWTF and FGD sources
contributed the highest bromide concentrations. The generating stations had multiple bromide sources,
including coal ash and coal pile runoff. The FGD Br/Cl ratio (0.0146) was higher than the ratio for both
CWTF (0.0104) and AMD (0.0058) and this suggested that Br/Cl ratios can be used as an indicator to
separate the multiple sources contributing bromide to the Allegheny River. This allowed for the
identification of a combined FGD and AMD source impacting S04_B, which was downstream from the
FGD discharges. Coal ash wastewater from generating stations also had a high Br/Cl ratio (0.0247)
compared to CWTF (0.0146); its contribution could not be separated from the FGD source.
5.3 PMF Source Apportionment Results
PMF was used to evaluate the river sample data from multiple sites, and a sensitivity analysis evaluated
the impact of including or excluding sites from the analysis. The results were sensitive to the inclusion of
the Blacklick Creek site samples (S01_B, S03_B), which were collected from a creek impacted by both a
CWTF and AMD from historical mining. The measured profile Br/Cl ratios were used to identify the
sources along with the location of the outfalls. In addition, CWTF and FGD mass discharge rates were
calculated using the measured profile concentrations. FGDs significantly contributed to anion
concentrations in the S04_B samples. As shown in Figure 4 and Figure 7, S04_B was located downstream
of the Conemaugh Dam and was impacted by many sources. High sulfate levels on the Blacklick to
Kiskiminetas indicated that AMD was a major source. Therefore, despite having a low measured bromide
contribution, AMD needed to be accounted for in the PMF analysis.
61
-------
1000
0.04
0.00
Figure 17. Summary of Br/Cl and bromide mean concentrations for measured sources (Table 13).
A combination of two PMF analyses was used to calculate the bromide sources at the PDWS intakes: (1)
downstream combined sites only, and (2) a hybrid using two PMF analyses to separate the AMD/FGD
source factor for bromide that could not be resolved in the combined site analysis. Both the combined
and hybrid PMF results generated similar overall contributions including all source categories. For
example, CWTF contributions at PWDS intake S05_B was 37% for the combined analysis and 37% for
the hybrid analysis. The hybrid results are useful because they provided a basis to estimate the FGD
bromide contributions (50% at S05_B) in the absence of AMD (9% at S05_B).
Figure 18 presents the bromide source contributions for both PMF analyses of the PDWS intakes and
highlights the differences between the two analyses. These results represent the median percent
contribution which varies daily due to changes in discharges. The two results provide a range in
contributions for the FGD source.
62
-------
Combined
Hybrid
Sediment
FGD Mo/0
+AMD
0%
POTW
0%
S()5_A, PDWS Intake
FGD
0%
Sediment AMD
POTW 11% ,o/0
0% I
Sediment PQTW
.0%
S05_B, PDWS Intake
Sediment
4%
POTW
0%
Figure 18. Summary of median PMF bromide source contributions for PDWS intakes.
This source apportionment study also evaluated the sources contributing other inorganic contaminants to
the PDWS intakes and the other sampling sites The measured source profiles (Table 13) included
chloride, nitrate, and sulfate; and the source contributions for each of these contaminants from CWTF,
FGD+AMD, POTW, and Sediment sources are presented in Tables 20, 21, and 22. For the combined
PMF analysis, CWTFs contributed the highest amount of chloride to downstream sites (8.14 to 19.2
mg/L), followed by FGD + AMD (5.31 to 17.6 mg/L), and POTWs (3.53 to 15.8 mg/L). Measured
source profiles generally supported these receptor model results with CWTFs (CTWT_A, CWTF_B)
having an average concentration of 72,450 mg/L, and FGD having concentrations approximately 5 times
less (12,800 mg/L). Sources of nitate included both POTWs (0.25 to 1.1 mg/L) and FGD (0.83 to 2.7
mg/L), which is consistent with the measured profiles (FGD =183 mg/L, POTW =98.5 mg/L). Sulfate
sources were CWTF (3.8 to 8.9 mg/L), FGD +AMD (64 to 210 mg/L), and POTW (3.8 to 17 mg/L).
Both FGD and AMD had high measured source concentrations of sulfate with 2080 mg/L and 469 mg/L,
respectively.
5.4 Application of Source Apportionment Modeling to Surface Water
This study is the first of its kind to demonstrate the application of source apportionment techniques to
quantify a complex array of source contributions to measured contaminant concentrations at PDWS raw
63
-------
water intakes. The source apportionment results are based solely on measured sample data and require no
a priori information on river discharge sources or river dispersion characteristics. Stable PMF results
were found when applying a multiple site source apportionment design commonly used for air pollution
source apportionment studies to this surface water application. We provide source profiles for a number
of common discharge facilities that may be relevant to apportion sources in other locations. We have
shown that multiple sources contribute to concentrations of anions at two Allegheny PDWS intakes,
including discharges from CWTF, FGD, and AMD sources. Source apportionment can improve our
understanding of the magnitude of the impact for the various sources. Understanding the sources will
guide efforts to control exposures to drinking water contaminants that are of concern to human health
such as brominated disinfection byproducts. These results can inform strategies for source mitigation or
treatment optimization using the source contribution summaries. Ultimately, the results of this study and
the future application of the research tools presented will provide communities, states, tribes, and industry
with sound scientific knowledge on understanding potential impacts of hydraulic fracturing on drinking
water resources, and the overall protection of those water resources for the future.
64
-------
6. REFERENCES
Andreasen, D. C., and W.B. Fleck. 1997. Use of bromide: chloride ratios to differentiate potential sources
of chloride in a shallow, unconfmed aquifer affected by brackish-water intrusion. Hydrogeol. J.
5(2): 17-26.
Amato, F., and P. K. Hopke. 2012. Source apportionment of the ambient PM2.5 across St. Louis using
constrained positive matrix factorization. Atmospheric Environment 46: 329-337.
Assefa, A. T., A. Sobek, K. L. Sundqvist, I. Cato, P. Jonsson, M. Tysklind, and K. Wiberg. 2013.
Temporal trends of PCDD/Fs in Baltic Sea sediment cores covering the 20th century.
Environmental Science and Technology 48(2): 947-953.
Balaba, R. S., and R. B. Smart. 2012. Total arsenic and selenium analysis in Marcellus shale, high-salinity
water, and hydrofracture flowback wastewater. Chemosphere 89: 1437-1442.
Bielski, A. 2012.Modelling of mass transport in watercourses at unsteady states. Environment Protection
Engineering. 38(3): 159-173.
Chapman, E. C., R. C. Capo, B. W. Stewart, C. S. Kirby, R. W. Hammack, K. T. Schroeder, and H. M.
Edenborn. 2012. Geochemical and strontium isotope characterization of produced waters from
Marcellus Shale natural gas extraction. Environmental Science and Technolology 46(6): 3545-
3553.
Davis, S. N., D. O. Whittemore, and J. Fabryka-Martin. 1998. Uses of bromide/chloride ratios in studies
of potable water. Groundwater 36(2): 338-350.
Ferrar, K. J., D. R. Michanowicz, C. L. Christen, N. Mulcahy, S. L. Malone, and R. K. Sharma. 2013.
Assessment of effluent contaminants from three facilities discharging Marcellus Shale wastewater
to surface waters in Pennsylvania. Environmental Science and Technology 47(7): 3472-3481.
Flotemersch, J. E.; Stribling, J. B.; Paul. M. J. 2006. Concepts and Approaches for the Bioassessment of
Non-wadeable Streams and Rivers. EPA 600-R-06-127.
Graney, J. R., M. S. Landis, and G. A. Norris. 2004. Concentrations and solubility of metals from indoor
and personal exposure PM2.5 samples. Atmospheric Environment 38(2): 237-247.
Gregory, K. B.; Vidic, R. D.; Dzombak, D. A. 2011. Water Management challenges associated with the
production of shale gas by hydraulic fracturing. Elements 7:181-186.
Haluszczak, L. O., A.W. Rose, and L. R. Kump. 2013. Geochemical evaluation of flowback brine from
Marcellus gas wells in Pennsylvania, USA. Applied Geochemistry 28: 55-61.
Henry, R. C. 1987. Current factor analysis receptor models are ill-posed. Atmospheric Environment
21(8): 1815-1820.
Henry, R. C., and E. R. Christensen. 2010. Selecting an appropriate multivariate source apportionment
model result. Environmental Science and Technology 44(7): 2474-2481.
Hladik, M. L., M. J. Focazio, and M. Engle. 2014. Discharges of produced waters from oil and gas
extraction via wastewater treatment plants are sources of disinfection by-products to receiving
streams. Science of the Total Environment 466: 1085-1093.
Hopke, P. K. 1991. An introduction to receptor modeling. Chemometrics and Intelligent Laboratory
Systems 10(1): 21-43.
Jobson, Harvey E. 1996. Prediction of traveltime and longitudinal dispersion in rivers and streams. United
States Geological Survey.
65
-------
Kelly, W. R., S. V. Panno, K. C. Hackley, H. Hwang, A. T. Martinsek, and M. Markus. 2010. Using
chloride and other ions to trace sewage and road salt in the Illinois Waterway. Applied
Geochemistry 25: 12.
Landis, M. S., A. F. Vette, and G. J. Keeler. 2002. Atmospheric mercury in the Lake Michigan basin:
influence of the Chicago/Gary urban area. Environmental Science and Technology 36(21): 4508-
4517.
Lee, G., J. M. Bigham, and G. Faure. 2002. Removal of trace metals by coprecipitation with Fe, Al and
Mn from natural waters contaminated with acid mine drainage in the Ducktown Mining District,
Tennessee. Applied Geochemistry 17(5): 569-581.
Lutz, B. D., A. N. Lewis, and M. W. Doyle. 2013. Generation, transport, and disposal of wastewater
associated with Marcellus Shale gas development. Water Resources Research 49(2): 647-656.
Norris, G. A., R. Duvall, S. Brown, and S. Bai. 2014. EPA Positive Matrix Factorization (PMF) 5.0
fundamentals and user guide. EPA/600/R-14/108.
Norris, G. A., R. Vedantham, K. Wade, P. Zahn, S. Brown, P. Paatero, S. Eberly, and C. Foley. 2009.
Guidance document for PMF applications with the Multilinear Engine. EPA/600/R-09/032.
Paatero, P. 1997. Least squares formulation of robust non-negative factor analysis. Chemometrics and
Intelligent Laboratory Systems 37(1): 23-35.
Paatero, P., and U. Tapper. 1994. Positive matrix factorization: A non-negative factor model with optimal
utilization of error estimates of data values. Environmetrics 5(2): 111-126.
Paatero, P., P.K. Hopke, X. H. Song, and Z. Ramadan. 2002. Understanding and controlling rotations in
factor analytic models. Chemometrics and Intelligent Laboratory Systems 60(1): 253-264.
Paatero, P., S. Eberly, S. Brown, and G. Norris. 2014. Methods for estimating uncertainty in factor
analytic solutions. Atmospheric Measurement Techniques 7(3): 781-797.
PA DCNR (Pennsylvania Department of Conservation and Natural Resources). 2014. Distribution of
Pennsylvania coals. Available at: http://www.dcnr.state.pa.us/topogeo/econresource/coal/
PA DEP (Pennsylvania Department of Environmental Protection). 201 la. Letter to EPA Region 3
Regional Administrator. Available at: http://www.epa.gov/region03/marcellus_shale/pdf/letter-
padep-to-epa7-26-l 1 .pdf
PA DEP (Pennsylvania Department of Environmental Protection). 201 Ib. DEP calls on natural gas
drillers to stop giving treatment facilities wastewater. Available at:
http://www.portal.state .pa.us/portal/server.pt/community/newsroom/14287?id=17071&typeid=l
PA DEP (Pennsylvania Department of Environmental Protection). 2012. ACT 13 frequently asked
questions. Available at:
http://files.dep. state .pa.us/OilGas/OilGasLandingPageFiles/Act 13/Act 13 FAQ .pdf
PA DEP (Pennsylvania Department of Environmental Protection). 2014a. Oil & gas reporting. Available
at: https://www.paoilandgasreporting.state.pa.us/publicreports/Modules/Waste/WasteHome.aspx
PA DEP (Pennsylvania Department of Environmental Protection). 2014b. NPDES eDMR Data System.
Available at:
http://www.depreportingservices.state .pa.us/ReportServer/Pages/ReportViewer.aspx?/NPDES/E
DMRPA
PA DEP (Pennsylvania Department of Environmental Protection). 2014c. The Pennsylvania Internet
Record Imaging System/ Wells Information System (PA*IRIS/WIS): Gateway to Detailed Oil
66
-------
and Gas Well Information Data. Available at:
http://dcnr.state.pa.us/topogeo/econresource/oilandgas/pa iris home/index.htm
Parker, K.M., T. Zeng, J. Harkness, A. Vengosh, and W.A. Mitch. 2014. Enhanced Formation of
Disinfection Byproducts in Shale GasWastewater-Impacted Drinking Water Supplies. Environ.
Sci.Tedmol. 48:11161-11169.
Pennsylvania Fish and Boat Commission. 2011. Three Rivers Management Plan - A Strategy for
Managing Fisheries Resources of the Allegheny, Monongahela, and Ohio Rivers.
http://www.fish.state.pa.us/water/rivers/threerivers/ThreeRiversMgmtPlan.pdf
Praipipat, P., L. A. Rodenburg, and G. J. Cavallo. 2013. Source apportionment of poly chlorinated
biphenyls in the sediments of the Delaware River. Environmental Science and Technology 47(9):
4277-4283.
Richardson, S.D., M.J. Plewa, E.D. Wagner, R. Schoeny, and D.M. DeMarini. 2007. Occurrence,
genotoxicity, and carcinogenicity of regulated and emerging disinfection by-products in drinking
water: A review and roadmap for research. Mutation Reseach. 636:178-242.Sams, J. I., and K. M.
Beer. 2000. Effects of coal-mine drainage on stream water quality in the Allegheny and
Monongahela River basin: Sulfate transport and trends. U.S. Department of the Interior, U.S.
Geological Survey, National Water-Quality Assessment Program.
Seere, S. D.; Lee, C.W. Evaluation of the Impact of Chlorine on Mercury Oxidation in a Pilot-Scale Coal
Combustor - the Effect of Coal Blending, United States Environmental Protection Agency,
Office of Research and Development, EPA/600/R-09/021, 2009, Washington, D.C.
Skalak, K.J., M.A. Engle, E.L. Rowan, G.D. Jolly, K.M. Conko, A.J. Benthem, and T.F. Kraemer. 2014.
Surface disposal of produced waters in western and southwestern Pennsylvania: Potential for
accumulation of alkali-earth elements in sediments. International Journal of Coal Geology 126:
162-170.
Soonthornnonda, P., and E. R. Christensen. 2008. Source apportionment of pollutants and flows of
combined sewer wastewater. Water Research 42(8): 1989-1998.
States, S., G. Cyprych, M. Stoner, F. Wydra, J. Kuchta, J. Monnell, and L. Casson. 2013. Marcellus Shale
drilling and brominated THMs in Pittsburgh, Pa., drinking water (PDF). Journal-American Water
Works Association 105(8): E432-E448.
U.S. EPA (U.S. Environmental Protection Agency). 1994. EPA Method 200.7 Revision 4.4,
Determination of metals and trace elements in water and wastes by Inductively Coupled Plasma -
Atomic Emission Spectrometry, Office of Research and Development, Cincinnati, OH. Available
at: http://www.epa.gov/sam/pdfs/EPA-200.7.pdf.
U.S. EPA (U.S. Environmental Protection Agency). 1996. EPA Method 8000b, Determinative
Chromatographic Separations. Available at:
http://www.epa.gov/solidwaste/hazard/testmethods/sw846/pdfs/8000b.pdf.
U.S. EPA (U.S. Environmental Protection Agency). 1997. EPA Method 300.1 Determination of Inorganic
Anions in Drinking Water by Ion Chromatography Revision 1.0. Available at:
http://water.epa.gov/scitech/methods/cwa/bioindicators/upload/2007_07_10_methods_method_30
01.pdf.
U.S. EPA (U.S. Environmental Protection Agency). 2007a. EPA Method 3015A, microwave assisted acid
digestion of aqueous samples and extracts. Available at:
http://www.epa.gov/osw/hazard/testmethods/sw846/pdfs/3015a.pdf.
67
-------
U.S. EPA (U.S. Environmental Protection Agency). 2007b. EPA Method 6020A, Inductively Coupled
Plasma-Mass Spectrometry. Available at:
http://www.epa.gov/osw/hazard/testmethods/sw846/pdfs/6020a.pdf.
U.S. EPA (U.S. Environmental Protection Agency). 2014a. EPA Positive Matrix Factorization software
and user guide. Available at: http://www.epa.gov/heasd/research/pmf.html.
U.S. EPA (U.S. Environmental Protection Agency). 2014b. EPA's study of hydraulic fracturing for oil
and gas and its potential impact on drinking water resources. Available at:
http://www2.epa.gov/hfstudy.
U.S. EPA (U.S. Environmental Protection Agency). 2014c. Operation and maintenance of the Element 2
high-resolution inductively coupled plasma mass spectrometry instrument. Environmental
Characterization and Apportionment Branch Standard Operating Procedure 98.2.
U. S. Army Corps of Engineers. http://www.lrp.usace.army.mil/Missions/WaterManagement.aspx.
USGS (U.S. Geological Survey), 1993. National Water Summary 1990-1991: Hydrologic Events and
Stream Water Quality, Available at: http://pubs.usgs.gov/wsp/2400/report.pdf.
USGS (U.S. Geological Survey), 2013, Water-resources data for the United States, Water Year 2012:
U.S. Geological Survey Water-Data Report WDR-US-2012, site 03025500, Available at:
http://wdr.water.usgs.gov/wv2012/pdfs/03025500.2012.pdf.
USGS (U.S. Geological Survey). 2014. USGS surface-water historical instantaneous data forthe Nation:
Build Time Series. Available at: http://waterdata.usgs.gov/nwis/uv/?referred module=sw.
Veil J. Water Management Technologies Used by Marcellus Shale Gas Producers; prepared for U.S.
Department of Energy Office of Fossil Fuel Energy, National Energy Technology Laboratory,
2010. http://www.veilenvironmental.com/publications/pw/Water_Mgmt_in_Marcellus-final-
iull0.pdf.
Vengosh, A., R. B. Jackson, N. Warner, T. H. Darrah, and A. Kondash. 2014. A critical review of the
risks to water resources from unconventional shale gas development and hydraulic fracturing in
the United States. Environmental Science and Technology 48(15): 8334-8348.
Warner, N. R., C. A. Christie, R. B. Jackson, and A. Vengosh. 2013. Impacts of shale gas wastewater
disposal on water quality in western Pennsylvania. Environmental Science and Technology
47(20): 11849-11857.
White, D.; Johnston, K.; Miller, M. 2005. Ohio River Basin. In A.C. Benke and C.E. Gushing (Eds.).
Rivers of North America. Elsevier Academic Press. Burlington, Massachusetts. 374-424.
Wilson, J. M., and J. M. Van Briesen. 2012. Oil and gas produced water management and surface
drinking water sources in Pennsylvania. Environmental Practice 14(4): 288-300.
Wilson, J. M., Y. Wang, J. M. Van Briesen. 2013. Sources of high total dissolved solids to drinking water
supply in southwestern Pennsylvania. Journal of Environmental Engineering 140(5).
Zou, Y., E. R. Christensen, and A. Li. 2013. Characteristic pattern analysis of polybromodiphenyl ethers
in Great Lakes sediments: a combination of eigenspace projection and positive matrix
factorization analysis. Environmetrics 24(1): 41-50.
68
-------
APPENDIX A
Comparison of Receptor and River Transport Conceptual Models and
Summary Statistics for River, CWTF, and PDWS Intake Samples
Receptor Modeling
Transport Modeling
Results for sampling sites
and study period and no
parameters need to be
specified, based on
measurements at the
intake.
River or
PDWS intake
samples
Sample
Analysis -
Chemical
Speciation
i
r
Speciated
concentration and
uncertainty files
Identify sources
and constraints
Results for specified sampling
locations and can be used to evaluate
impacts under various flow and
discharge conditions, many parameters
need to be specified.
Source type
profiles and
contributions to
each sample
(known and
unknown)
Total
contaminant
contribution to
each sampling
location
Measured
chemical
composition
of sources
Disharge
contaminant
concentration
Discharge
schedule
Discharge
Rates
Figure Al. Diagram showing receptor and transport model input requirements and model outputs.
A-1
-------
Table Al. Summary of Allegheny River species concentrations (mg/L); n (%) number of valid samples.
Species Statistic
Br - 1C n (%)
mean (std)
median (IQR)
min - max
Ca - OES n (%)
mean (std)
median (IQR)
min - max
Cl - 1C n (%)
mean (std)
median (IQR)
min - max
Fe - MS n (%)
mean (std)
median (IQR)
min - max
Mg - OES n (%)
mean (std)
median (IQR)
min - max
Mn - MS n (%)
mean (std)
median (IQR)
min - max
S01_A
31 (100%)
0.0718(0.0254)
0.0721 (0.0446)
0.0329-0.117
30 (97%)
15.2(1.19)
15.3(1.4)
12.4-17.6
31 (100%)
20.8 (2.93)
21 (4.9)
16.2-25.8
31 (100%)
0.176(0.0903)
0.154(0.097)
0.071 - 0.384
30 (97%)
3.09(0.191)
3.11(0.25)
2.56 - 3.44
31 (100%)
0.0607 (0.0326)
0.0536 (0.059)
0.0181-0.123
S02_A
23 (100%)
684 (76.4)
697(101)
536 - 788
22 (96%)
9420(1050)
9470(1560)
7680-11500
23 (100%)
61700 (6210)
61900 (8100)
48800 - 73400
8 (67%)
0.183(0.12)
0.149(0.159)
0.0485 - 0.383
23 (100%)
969(154)
1010(155)
472-1140
10 (83%)
0.33 (0.357)
0.144(0.445)
0.0296-1.12
S03_A
45 (98%)
0.104(0.0596)
0.0836 (0.0399)
0.037 - 0.28
45 (98%)
16.8(1.6)
16.6(2.1)
13.8-21.3
46 (100%)
24.2 (6.04)
22.5 (4.2)
16.6-41.6
42 (91%)
0.202(0.134)
0.145(0.145)
0.0767 - 0.541
45 (98%)
3.41 (0.224)
3.41 (0.33)
3.02 - 4.06
42 (91%)
0.0645 (0.0437)
0.0417 (0.0726)
0.0215-0.183
S04_A
30 (100%)
0.0851 (0.0302)
0.0779 (0.0378)
0.0379-0.154
29 (100%)
17.9(1.78)
18.2(3.5)
14.5-20.1
30 (100%)
23.3 (3.46)
22.5 (4.6)
17.3-31.3
29 (100%)
0.154(0.0762)
0.125(0.045)
0.0767 - 0.389
29 (100%)
3.64 (0.272)
3.65 (0.49)
3.09 - 3.98
29 (100%)
0.0655 (0.0423)
0.041 (0.055)
0.0265-0.168
S05_A
52 (100%)
0.0867
/A AQ4 "7\
(0.031/)
0.0852
/A A/1QC\
(0.043b)
0.038-0.166
51 (100%)
18.6(1.34)
18.5(1.8)
16.3-21.6
52 (100%)
23.6 (3.53)
23.5 (4.45)
17.6-32.1
50 (98%)
0.206
ir\ r\no H \
(0.0931)
0.178(0.106)
0.0956 - 0.476
51 (100%)
3.81 (0.205)
3.79 (0.27)
3.4-4.27
51 (100%)
0.081
0.0686
/A A7QO\
(0.0/oz)
0.0299-0.191
A-2
-------
Table. Al (continued ) Summary of Allegheny River species concentrations (mg/L); n (%) number of
valid samples.
Species
N03-IC
Na - OES
S04-IC
Si -OES
Sr - OES
Statistic
n (%)
mean (std)
median (IQR)
min - max
n (%)
mean (std)
median (IQR)
min - max
n (%)
mean (std)
median (IQR)
min - max
n (%)
mean (std)
median (IQR)
min - max
n (%)
mean (std)
median (IQR)
min - max
S01_B
30 (100%)
0.59 (0.366)
0.545 (0.558)
0.0188-1.26
30 (97%)
11.2(1.45)
11.5(2.3)
8.18-13.5
31 (100%)
7.47 (0.341)
7.49(0.31)
6.78 - 8.6
30 (97%)
0.961 (0.335)
1.01 (0.472)
0.134-1.55
30 (97%)
0.0655 (0.00886)
0.0641 (0.0136)
0.0483 - 0.0838
S02_B
22 (100%)
23800 (2560)
24000(4100)
19100-30000
23 (100%)
218(61.7)
198(62)
145-385
23 (100%)
0.738(0.132)
0.744 (0.236)
0.481-0.96
22 (100%)
395 (202)
343(173)
218-1190
S03_B
45 (100%)
0.594 (0.285)
0.652 (0.4)
0.0182-1.11
45 (98%)
12.9(2.51)
12.5(2.3)
9.27 - 20
46 (100%)
8.44 (0.426)
8.35 (0.7)
7.75 - 9.69
45 (98%)
0.94 (0.332)
0.86 (0.408)
0.446-1.64
45 (98%)
0.085 (0.0227)
0.0782(0.016)
0.0607-0.144
S04_B
27 (100%)
0.283 (0.237)
0.243 (0.328)
0.0284 - 0.807
29 (100%)
12.2(1.6)
11.9(2.1)
9.68-16.2
30 (100%)
9.02 (0.72)
8.93 (0.67)
7.98-11.7
29 (100%)
0.888 (0.467)
0.697 (0.814)
0.219-1.66
29 (100%)
0.079(0.0118)
0.0746(0.019)
0.0627-0.105
S05_B
44(100%)
0.327 (0.292)
0.284 (0.477)
0.0178-1.2
51 (100%)
12.6(1.45)
12.5(1.8)
10-16.2
52 (100%)
9.62 (0.635)
9.58 (0.96)
8.5-10.9
51 (100%)
1.01(0.407)
1.06(0.548)
0.198-1.87
51 (100%)
0.0797(0.0115)
0.077(0.0171)
0.0613-0.105
A-3
-------
Table A2. Summary of Blacklick Creek species concentrations (mg/L); n (%) number of valid samples.
Species
Br -1C
Ca - OES
Cl -1C
Fe -MS
Mg -OES
Mn -MS
Statistic
n (%)
mean (std)
median (IQR)
min - max
n (%)
mean (std)
median (IQR)
min - max
n (%)
mean (std)
median (IQR)
min - max
n (%)
mean (std)
median (IQR)
min - max
n (%)
mean (std)
median (IQR)
min - max
n (%)
mean (std)
median (IQR)
min - max
S01_B
27 (100%)
0.0696 (0.0175)
0.0726 (0.0287)
0.0352 - 0.0889
28 (100%)
56.3(19.7)
53.8 (26.4)
25.6 - 89.9
28 (100%)
19.6(1.79)
19.6(2.2)
14.1-23.2
28 (100%)
2.6 (4.43)
1.33(1.39)
0.479 - 23.7
28 (100%)
15.7(5.31)
16.1(8.1)
6.79 - 24.4
28 (100%)
0.808 (0.207)
0.913(0.34)
0.417-1.06
S02_B
29 (100%)
808 (65.9)
812(80)
713-1020
29 (100%)
12100 (975)
12100 (900)
8820-13600
29 (100%)
83200 (15900)
85800 (6400)
7010-109000
14 (93%)
0.213(0.0692)
0.212 (0.069)
0.0663-0.371
29 (100%)
749 (94.6)
749 (88)
579 - 979
15 (100%)
0.154(0.068)
0.158(0.091)
0.0399 - 0.269
S03_B
29 (100%)
0.995(1.32)
0.327 (0.99)
0.0635 - 4.36
29 (100%)
74.1 (34.8)
64(31.4)
26.2-168
29 (100%)
140(185)
48.7(107)
21.6-621
29 (100%)
2.99(6.19)
1.05(1.36)
0.298 - 32
29 (100%)
16.6 (5.4)
15.7 (8.7)
6.93-27.4
28 (97%)
0.797(0.195)
0.853 (0.297)
0.418-1.02
S04_B
42 (100%)
0.428(0.146)
0.419(0.215)
0.178-0.797
42 (100%)
65.9 (9.76)
66.4(12.7)
41.6-87.1
42 (100%)
58.3(13.9)
57.6 (22)
35.2 - 93.7
40 (95%)
0.36 (0.243)
0.307 (0.218)
0.0791-0.942
42 (100%)
21.2(3.18)
21.4(3.6)
13.2-27.8
40 (95%)
0.181 (0.103)
0.163(0.103)
0.0279 - 0.438
S05_B
56 (100%)
0.143(0.0277)
0.132 (0.0455)
0.102-0.205
54 (96%)
31.1(5.39)
29.3 (7.6)
23.3-42.3
56 (100%)
30.4 (3.69)
29.7 (6.2)
24.4-37.6
56 (100%)
0.301 (0.152)
0.249(0.213)
0.137-0.8
54 (96%)
9.22(1.76)
8.64 (2.68)
6.94-12.9
56 (100%)
0.0945(0.0314)
0.0909(0.0513)
0.049-0.156
A-4
-------
Table A2. (continued) Summary of Blacklick Creek species concentrations (mg/L); n (%) number of valid
samples.
Species
N03 - 1C
Na - OES
S04-IC
Si -OES
Sr - OES
Statistic
n (%)
mean (std)
median
(IQR)
min - max
n (%)
mean (std)
median
(IQR)
min - max
n (%)
mean (std)
median
(IQR)
min - max
n (%)
mean (std)
median
(IQR)
min - max
n (%)
mean (std)
median
(IQR)
min - max
S01_B
30 (100%)
0.59 (0.366)
0.545 (0.558)
0.0188-1.26
30 (97%)
11.2(1.45)
11.5(2.3)
8.18-13.5
31 (100%)
7.47(0.341)
7.49(0.31)
6.78 - 8.6
30 (97%)
0.961 (0.335)
1.01(0.472)
0.134-1.55
30 (97%)
0.0655
(0.00886)
0.0641 (0.0136)
0.0483 - 0.0838
S02_B
22 (100%)
23800 (2560)
24000(4100)
19100-30000
23 (100%)
218(61.7)
198(62)
145-385
23 (100%)
0.738(0.132)
0.744 (0.236)
0.481-0.96
22 (100%)
395 (202)
343(173)
218-1190
S03_B
45 (100%)
0.594 (0.285)
0.652 (0.4)
0.0182-1.11
45 (98%)
12.9(2.51)
12.5(2.3)
9.27 - 20
46 (100%)
8.44 (0.426)
8.35 (0.7)
7.75-9.69
45 (98%)
0.94 (0.332)
0.86 (0.408)
0.446-1.64
45 (98%)
0.085 (0.0227)
0.0782 (0.016)
0.0607-0.144
S04_B
27 (100%)
0.283 (0.237)
0.243 (0.328)
0.0284 - 0.807
29 (100%)
12.2(1.6)
11.9(2.1)
9.68-16.2
30 (100%)
9.02 (0.72)
8.93 (0.67)
7.98-11.7
29 (100%)
0.888 (0.467)
0.697(0.814)
0.219-1.66
29 (100%)
0.079(0.0118)
0.0746 (0.019)
0.0627-0.105
S05_B
44 (100%)
0.327 (0.292)
0.284 (0.477)
0.0178-1.2
51 (100%)
12.6(1.45)
12.5(1.8)
10-16.2
52 (100%)
9.62 (0.635)
9.58 (0.96)
8.5-10.9
51 (100%)
1.01 (0.407)
1.06(0.548)
0.198-1.87
51 (100%)
0.0797(0.0115)
0.077(0.0171)
0.0613-0.105
A-5
-------
APPENDIX B
PMF Model Analysis Parameters and Settings
Table Bl. Allegheny PMF analysis parameters.1
Species BMC, CIJC, N03JC, S04JC, Ca_OES, K_OES, Mg_OES, Na_OES, SLOES, Sr_OES,
Fe_MS, Mn_MS
Number of Samples 124
Number of Sampling Sites 5
Number of Factors 4
Species Categories Fe_MS (Weak)
Number of Base Runs 50
Random Seed 12
Robust mode Yes
Non-converged runs No
Q (Robust) 1330
Q(True) 1350.3
Q (True)/Qexpected 1.85
DISP (dQmax = 4) 000
Constraints % dQ 0.10
Constraint Equation [POTW|Br_IC] - 0.0003 * [POTW|CI_IC] = 0
1 Qexpected = m * n - p * (m+n) where Qexpected is the degrees of freedom with m chemicals,
n samples, and p factors.
B-l
-------
Table B2. Blacklick PMF analysis parameters.
Species
Number of Samples
Number of Sampling Sites
Number of Factors
Species Categories
Number of Base Runs
Random Seed
Robust Mode
Non-converged runs
Q (Robust)
Q (True)
Q (True)/Qexpected
DISP (dQmax = 4)
Constraints %dQ
Constraint Equation
BMC, CIJC, N03JC, S04JC, Ca_OES, K_OES, Mg_OES, Na_OES, SLOES, SrjDES,
Fe_MS, Mn_MS
90
3
3
Fe_MS (Weak)
50
2
Yes
No
4355.65
6323.13
10.06868
000
0.12
[POTW|Br_IC] - 0.0003 * [POTW|CI_IC] = 0
Table B3. Combined Allegheny and Blacklick PMF analysis parameters.
Species
Number of Samples
Number of Sampling
Sites
Number of Factors
Species Categories
Number of Base Runs
Random Seed
Robust Mode
Non-converged runs
Q (Robust)
Q (True)
Q (True)/Qexpected
DISP (dQmax = 4)
Constraints %dQ
Constraint Equation
BMC, CIJC, N03JC, S04JC, Ca_OES, K_OES, Mg_OES, Na_OES, SLOES, SrjDES,
Fe_MS, Mn_MS
90
5
Fe_MS (Weak), Mn_MS (Weak)
50
67
Yes
None
2787.78
3283.05
4.3774
0000
NA
NA
B-2
-------
xvEPA
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
Office of Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
------- |